Calculating Carbohydrates

Nobody would argue with the statement that proteins are a challenge for computer modelers. Descriptions of their swirls of helices and ribbons, to say nothing of their further folding, can tax even the biggest supercomputer. But the complexity of proteins pales in comparison to that of their biochemical allies, the carbohydrates.

Though proteins may contain many thousands of atoms bunched together in elaborate structures, they tend to hold their globular shapes. Carbohydrates, by contrast, are bendy, twisty, frequently branched strands that are often even bigger than proteins. And that has made describing their behavior with computer simulations a difficult prospect.

But there's a wealth of powerful information about carbohydrates that can be gleaned from modeling, with profound implications for biology and medicine, from infection to immunity. For example, an intimate knowledge of the polysaccharide surface of a bacterium could be key in pinpointing targets for antibiotics or vaccines. And what scientists learn from studying carbohydrates could shed light on other flexible biopolymers such as RNA and provide insights into protein behavior as well.

Fifteen years ago, scientists' ability to model carbohydrates lagged far behind their ability to model proteins and nucleic acids. But carbohydrate chemists have been catching up, thanks in part to giant strides made in computer power and speed. That much was clear from a symposium on the computational chemistry of carbohydrates that was held last month at the American Chemical Society national meeting in Anaheim, Calif., and that was sponsored by the Division of Carbohydrate Chemistry.

"I would say, based on the symposium in Anaheim, that the gap has been essentially closed," said John W. Brady, chemistry professor in the food science department at Cornell University. "The modeling being done now on carbohydrates is as sophisticated as that being done on proteins."

Nearly two dozen researchers from around the globe gathered to discuss new advances in the field, from strategies for simulating solvation to new computer programs tailored to predict carbohydrates' unique characteristics. In particular, the coupling of nuclear magnetic resonance experiments with modeling is proving to be fruitful in studying carbohydrate structure and conformation.

Speakers at the meeting highlighted the increasing number of computer programs specially tailored for carbohydrates. "The carbohydrate chemists are certainly making good use of the increased computer power and improved software," remarked Alfred D. French, a carbohydrate chemist at the Department of Agriculture's Southern Regional Research Center in New Orleans.

Numerous hurdles face the carbohydrate modeler. For example, the binding sites of proteins and their matching ligands tend to be seen as locks and keys. Not so with carbohydrates. "Since they're flopping, the phenomenon of recognition is no longer easy to think about," said Robert J. Woods, associate professor of biochemistry and molecular biology at the University of Georgia, Athens, and organizer of the ACS symposium.

THAT FLOPPINESS also makes carbohydrates difficult to crystallize, so their structures are frequently elucidated with NMR, rather than the X-ray crystallography traditionally used for proteins. NMR structures represent averages, however, and it's difficult to say which conformational families are biologically most important. Models can help narrow down what shapes the carbohydrate is likely to adopt in a given environment.

The glycosidic oxygens that link together monosaccharides are the locus of much of carbohydrates' flexibility. These motions are described by changes in the internal torsion angles (the rotation angles between the glycosidic oxygen and its two neighbors). The carbohydrates don't move around freely, of course, but alternate between several preferred conformations. In fact, in a big carbohydrate, torsion angles may change while the molecule maintains its overall shape. "Carbohydrates show a lot of correlated internal motion," Woods said.

Carbohydrate motion lends itself particularly to molecular dynamics simulations--predictions of how the molecule moves over time. These low-frequency motions can require simulations of 10 to 100 nanoseconds--an expensive eternity in computer simulation time--compared with protein simulations only a few nanoseconds long.

Molecular dynamics simulations often involve modeling a molecule in a "box of water," where it can roll around, buffeted by its environment. But when a lengthy, floppy carbohydrate tumbles during the simulation, it's likely to bump into the wall of the box. To avoid those collisions requires a huge box containing mostly water, said Woods.

The importance of water in carbohydrate behavior now seems intuitive, Woods explained, but in the early days, researchers tended not to include water in their simulations. "It set the field apart from biomolecular modeling for a while because the simulation methods weren't developed by theoreticians but by experimentalists who wanted a simple model for interpreting their data," Woods said.

In biology, however, many carbohydrates are covalently bound to proteins. And as researchers began to examine this complication, "it became obvious that they had to include water," Woods said. Now, a lot of work is devoted to developing ways to treat water in simulations. For example, is it best to treat the water molecules explicitly, which costs more in computer time and power, or to treat water as a continuous background?

Brady's group has pioneered the study of how carbohydrates interact with water. With their strongly hydrogen-bonding OH groups and non-hydrogen-bonding aliphatic CH groups, carbohydrates have a complex relationship with solvents. "In many cases, there's a direct opportunity for the water to react," Brady said. For example, water can affect ring flipping in monosaccharides, or a water molecule bridging parts of the carbohydrate could affect energetics. Brady is working to bolster his theoretical studies with neutron diffraction experiments with isotopically labeled sugars.

A number of speakers at the meeting demonstrated that modeling is already having a direct impact on the study of biologically important processes involving carbohydrates.

The bacterium that causes tuberculosis, Mycobacterium tuberculosis, has one of the more impenetrable coatings known. It resists the cell-wall-piercing actions of antibiotics and can even survive inside immune cells. It's known that this impermeability stems from a close-packed layer of mycolic acids on the outside. These lipids are attached to the cell membrane through a polysaccharide. What's unusual is that this polysaccharide is made up of monosaccharides in the five-membered, or furanose, ring form. This configuration is higher in energy than the six-membered-ring pyranoses found in mammalian polysaccharides.

So why did these bacteria evolve to produce polysaccharides composed of monosaccharides in the least stable ring form? One hypothesis holds that the bacterium uses furanoses because they're more flexible than pyranoses, which allows the lipids to form their tightly packed arrangement. It's analogous to a linker made of rubber, rather than brick, said Todd L. Lowary, chemistry professor at the University of Alberta, Edmonton. But until recently, that hypothesis hadn't been put to the test, experimentally or computationally.

Now, Lowary--with colleagues Christopher M. Hadad, chemistry professor at Ohio State University, and assistant professor Justin B. Houseknecht at Centre College, Danville, Ky.--is using computational methods to examine the effects of mycolic acids on the polysaccharide conformation. They selected a disaccharide fragment and its attached mycolic acids and performed simulations, randomly generating about 185,000 different conformations.

They then calculated the energies of each conformer, which gave them insight into what furanose ring conformations were most likely. The result, they found, is that the lipids do dramatically influence the carbohydrate's conformation. "They force the sugars to adopt a conformation they normally wouldn't want to adopt," Lowary said.

Though the study is preliminary, he noted, "This idea of a flexible scaffold now has some support, whereas before it was just an idea."

Group B Streptococcus is another nasty bug, responsible in particular for life-threatening illnesses in newborns. Scientists are working to develop vaccines against group B strep. Such a vaccine may contain a piece of the bacterium's carbohydrate capsule that's recognized by the immune system, stimulating the immune system to form antibodies against the bacteria. A thorough understanding of group B strep's carbohydrate coating itself is key to understanding how the antibody interacts with the bacterial surface.

CURRENT THINKING has held that, in general, the carbohydrate conformation recognized by an antibody--the epitope--can't contain more than six carbohydrate residues. Anything bigger, and the antibody can't interact with it.

But in the case of group B strep, the epitope appears to contain about 20 saccharide residues--which didn't fit in with the reigning hypothesis. To resolve that discrepancy, it was thought that the polysaccharide at some critical length changes its conformation, and that a smaller piece of that new shape is what the antibody recognizes.

Now, Woods's group has performed dynamics calculations on the carbohydrate-antibody system that contradict that notion. "We found that the antibody does indeed interact with a very large piece [of the carbohydrate]," Woods said. The segment forms a large corkscrew shape, the inner surface of which is recognized by the antibody. In fact, the polysaccharide needs to be as large as it is to form a helix. "So there's no great change in conformation that occurs," Woods said. "It's unproven experimentally, but strongly suggested by all simulations."

With knowledge of what's required generally for an antibody response comes the possibility of thinking about developing vaccines that protect against more than one strain of the bacteria. "If you understand the shape dependency, you can see if there's a common core feature you could exploit," Woods said.

A more biologically benign topic centered around anthocyanins, carbohydrates that are responsible for much of the red and orange coloring found in nature--for example, in grapes, strawberries, and flowers. Scientists are interested in these compounds because not only do they have antioxidant properties but they could also serve as natural food colorings, replacing some of the artificial dyes used currently.

The red color in anthocyanins is due to an aromatic flavylium cation that exists in an acidic environment. But at higher pH, the cation is attacked by water, generating a colorless hemiacetal. That makes anthocyanins useless as food colorings in their current form.

The color of anthocyanins with an attached aromatic acyl group tends to be stabler than that of its unacylated counterparts at higher pH. NMR studies have indicated that this is because the flavylium chromophore and the acyl group stack with one another, possibly making water less able to attack the flavylium cation.

AT THE MEETING, Karen T. Welch, a postdoctoral researcher in the chemistry department at the University of Tennessee, described her efforts to determine the characteristic collection of functional groups that impart color stability--with an eye toward designing useful food colorings.

Welch simulated the dynamics of a monoacylated anthocyanin produced by the wild carrot, taking 'snapshots' of different structures. She calculated the energies of about 1,000 different conformations, clustered into 18 families, eventually narrowing them down to two likely candidates, one of which agreed with all the NMR data. This model, Welch said, could be used to optimize the color stability of other anthocyanins.

Organic chemistry professor Göran Widmalm of Stockholm University, in Sweden, and coworkers explored the dynamics of lacto-N-neotetraose, a carbohydrate that is produced in humans and also is found in the coat of Neisseria meningitides, the bacterium responsible for meningitis. Their simulations picked out two significant conformations for the molecule, which NMR experiments also identified.

Widmalm's group also used the combination of NMR and modeling to study the binding of a complex oligosaccharide to wheat germ agglutinin (WGA). Some of the structural variations in the carbohydrate are characteristics of surfaces of metastatic cancer cells, and so might be markers of malignancy. WGA's ability to bind with these carbohydrates is of potential use in developing cancer diagnostics.

The Stockholm researchers used molecular modeling to study the interaction between WGA and the carbohydrate, which, Widmalm said, gave them six possible binding modes. They performed NMR experiments on the bound complex and compared those data with the model, picking the binding mode that best fit. "The modeling gave us different alternatives, which we could discard or confirm, showing the strength of using NMR and simulation together," he said.

Despite the success of the NMR-modeling combination, there remains a paucity of experimental data on 3-D carbohydrate structures in solution, said chemistry professor Anthony S. Serianni of the University of Notre Dame. His group is developing NMR methods to establish standard relationships between experimental data and carbohydrate structure.

In addition to developing NMR experiments, they're using the computer programs Chymesa and Glyfit, written by Serianni's posdoc Christophe Thibaudeau. The software helps them decide which conformational model is consistent with the observed data.

Perhaps the biggest computational effort has been in the development of force fields for carbohydrates. Force fields are used to calculate molecular energies and configurations. Chemists can then augment these force fields to deal with the particular characteristics of carbohydrates.

"It will be interesting to see how these force fields will converge in the future to describe the conformation and flexibility of carbohydrates," Widmalm said.