A Step Closer to Understanding How Proteins Fold

Release Date:
Alisa Zapp Machalek, NIGMS
(301) 496-7301

Scientists are now one step closer to solving a problem that they've struggled with for 40 years--determining how proteins fold into the three-dimensional shapes crucial for their function.

Proteins control nearly every system in our bodies. Antibodies, enzymes, and many hormones are proteins. A protein's three-dimensional structure determines which molecules it binds, which reactions it catalyzes, and, to some extent, where it is located in the cell.

It is much easier to determine the order, or sequence, of amino acids in a protein than it is to determine the protein's three-dimensional structure. Although the sequence contains all the information necessary for correct folding, scientists still cannot predict a protein's final shape from its sequence.

"If we could decipher the structures of proteins from their sequences, we could better understand all sorts of biological phenomena, from carcinogenesis to AIDS," said Dr. James Cassatt, director of the Division of Cell Biology and Biophysics at the National Institute of General Medical Sciences (NIGMS). "Then we might be able to do more about these disorders."

Now, NIGMS grantee Dr. George Rose and postdoctoral fellow Rajgopal Srinivasan at The Johns Hopkins University in Baltimore, Maryland, have developed a method that promises to come closer than ever before to cracking the protein folding code. Rose and Srinivasan describe the method, a computer program called LINUS, in a paper in Proteins: Structure, Function, and Genetics due out in mid-June.

"There are a lot of programs to predict secondary structure--small-scale, localized twists in the protein--from sequence information," said Cassatt. "But somehow Rose gets the overall structure. That makes his program unique."

The LINUS computer program, which is named after the late Nobel laureate Dr. Linus Pauling, is based on very simple assumptions. No two atoms are allowed to be in the same space at the same time; amino acids are encouraged to be in the conformations most commonly seen in proteins; and hydrophobic, or "water-fearing," amino acids are encouraged to cluster in the protected center of the protein.

In its calculations, LINUS doesn't even use the true structure for amino acids in the protein. Instead, it makes a gross approximation. Every amino acid has two parts: a backbone that is common to all amino acids, and a side chain unique to that type of amino acid. LINUS replaces each side chain--which can range from 1 to 18 atoms--with a single sphere that varies in size depending on the type of amino acid. To predict a protein's structure, LINUS divides the protein sequence into overlapping, bite-sized chunks. Starting with chunks only six amino acids long, LINUS randomly twists the amino acids into any of four possible localized structures: helix, sheet, turn, or coil.

The process repeats 5,000 times, with each trial conformation ranked according to how energetically favorable it is--a measure of how likely it is to occur in nature. If a conformation is ranked as favorable in more than 70 percent of the trial structures, and it enables interaction between some amino acids, it is frozen in position for subsequent cycles.

LINUS then starts another cycle by biting off larger overlapping chunks--12 amino acids this time--and twisting them into 5,000 new conformations. As before, LINUS freezes the most favorable conformations from this cycle, creating a new starting point for the next cycle. The process continues, with LINUS taking ever-larger bites of the protein each round. The final structure is made up of the conformations that are most favored in the last cycle.

The resulting structures provide a surprisingly good estimate of the protein's three-dimensional shape. X-ray crystallography can provide more structural detail, but it requires months, even years, to complete a crystal structure.

"Right now LINUS gives the same benefit as a poor crystal structure," Rose said. "What is that useful for? A lot of things." For example, he said, it could be used in conjunction with the Human Genome Project to classify proteins that might be involved in genetic disorders. Or it could be used to speed the progress of targeted drug design.

"The most surprising things about this work is how these exceedingly simple ideas, crudely expressed, can do so well in giving you the gross conformation of the protein," Rose said. Of the seven structures that Rose examined in his paper, LINUS predicted the gross overall shape for all but one. The predicted length and location of the more localized structures--the helices and sheets--were almost identical to known structures determined by x-ray crystallography.

With further refinements, Rose expects the accuracy of the program to increase dramatically. For example, in the research reported in his paper, Rose told LINUS to take bites no larger than 50 amino acids. Now he is modifying the program to produce accurate structure predictions using much larger chunks. He also plans to enable LINUS to predict special protein features such as disulfide bonds, prosthetic groups, and multiple subunits. Eventually, he hopes that LINUS will be useful to researchers all over the world.

"Our goal is to boil this program down into something that is sufficiently simple and sufficiently exportable that we can give it to many other labs," Rose said. "That way, instead of trying to do all these experiments ourselves, other people who are experts in certain areas can try their own experiments."

The process that LINUS uses--identifying small, localized structures, then incorporating them into the overall protein structure--is called hierarchical condensation. Although this technique has been successful in predicting protein structure, Rose does not claim that it predicts the pathway by which proteins fold in living systems.

"We do think that hierarchical condensation is what happens in one way or another when proteins fold," Rose said. "But I don't think LINUS is giving a detailed representation of any folding pathway."

Although the field of protein folding is very popular now, that wasn't always the case, Rose said. "In 1979 when I got my first NIGMS grant to work on this, protein folding was considered a problem too hard to be solved in this century," Rose said. "Now we're very optimistic."

Please acknowledge funding for this work from the National Institute of General Medical Sciences (NIGMS), a component of the National Institutes of Health that supports basic, non-disease-targeted research.

Reference: Srinivasan R, and Rose GD. LINUS--A Hierarchic Procedure to Predict the Fold of a Protein. Proteins: Structure, Function, and Genetics 1995; 22(2): 81-99.

To contact scientists available for comment, please call the NIGMS Office of Communications and Public Liaison at (301) 496-7301.