Physicists Take on Challenge Of Showing How Proteins Fold

How proteins fold is a central mystery of the life process that for decades has eluded explanation. But biologists are getting help on the problem nowadays from physicists, who bring quantitative theorems and new technologies to the task of showing how one-dimensional amino acid sequences determine the three-dimensional shapes of proteins. Such knowledge could guide structure-based design of drugs to treat a range of diseases now thought to be caused by misshapen proteins.

“We’re just beginning,” says Stanford University’s Steven Chu, cowinner of the 1997 Nobel Prize in physics, who has recently turned his efforts toward the protein folding question. “Everyone is just beginning. The interesting thing is, straight off, you see things you don’t understand.”

Steven Chu


Chu gave the keynote address at the third international symposium on biological physics, held recently in Santa Fe, N.M. Speakers from the United States and a dozen foreign countries focused on a number of physical approaches to biological systems, many of them involving the dynamics of protein folding.

To function properly, a protein must fold into a particular conformation, called the native state. This conformation appears to be somehow encoded in the amino acid sequence. Researchers are trying to uncover general rules governing the thermodynamics and kinetics of folding that can explain not only the process itself, but the extent to which it reflects the workings of natural selection.

Given that the product, or phenotype, of a gene is a protein, the pressure of natural selection is on the protein rather than on the DNA. To know how proteins create their functional structures is therefore necessary to understand the meaning of much genomic data being generated nowadays. In that sense, Carnegie Mellon University assistant professor of biochemistry Zheng-yu Peng and others have referred to protein folding as the “second half of the genetic code.”

COIL TO COLLAPSE: This computer model image by Vijay S. Pande, an assistant professor in the chemistry department at Stanford University, shows a heteropolymer coil and a globule collapse.


Work by Peng and colleagues strongly suggests that defective folding and/or aggregation of protein is a common mechanism for inactivation of the p16 kinease inhibitor, which normally is responsible for suppressing tumors arising from a variety of cancers (B. Zhang et al., Journal of Biological Chemistry, 271[46]:28734-7, 1996). Such research eventually could provide clues to designing small-molecule drugs that mimic the structure and function of normal p16, for patients carrying mutant forms of it.

Other studies have demonstrated that deletion of phenylalanine 508, an amino acid in the cystic fibrosis transmembrane conductance regulator, CFTR, causes proteins to misform (B.H. Qu et al., Journal of Biological Chemistry, 271[13]:7261-4, 1996).

Mutant protein likewise has been implicated in neurodegenerative diseases. Creutzfeldt-Jakob disease, kuru, and other spongiform encephalopathies are thought to be caused by misfolding of the protein-based pathogens called prions. Abnormal aggregation of the amino acid glutamine in neurons has been linked to forms of inherited ataxia and to Huntington’s disease, while aggregation or tangling of several neuronal protein types is involved in Alzheimer’s disease.

Computerized Models

“The reason we know there is information in the sequences of a biomolecule is that they’re very complicated,” comments Peter G. Wolynes, a University of Illinois professor of chemistry, biophysics, and physics. He codeveloped the energy landscape theory of protein folding, which he characterizes as “extremely controversial among biochemical people.” The theory¬≠an important aid in designing computer simulations and algorithms to predict the structure of proteins–suggests that proteins are not typical random systems but are special in possessing energy landscapes that can be described as rugged funnels (J.N. Onuchic et al., Annual Review of Physical Chemistry, 48:545-600, 1997).

“Proteins are minimally frustrated systems, and what we mean by that is, their native states make sense,” Wolynes says. He adds that the special structures within proteins have many interactions or contacts that create energy.

A question being explored now, he notes, is how does the funnel shape apply to real proteins?

The complexity of proteins makes such a question difficult to answer, as illustrated in 1968 by Levinthal’s Paradox (C. Levinthal, Journal of Chemical Physics, 65:44-5, 1968). It says, let each amino acid residue in a small, 100-residue protein have six possible conformations. To this main chain can be factored in the side chains, which incorporate the 20 common amino acids. The resultant folding possibilities are so numerous that even today, the fastest computers randomly searching through them would require thousands of years to deduce the protein’s atom-by-atom conformation. This is why it is surmised that the information is encoded in the amino acid sequence.

Nevertheless, relatively simplified computer models of proteins can simulate current theories about folding. Vijay S. Pande, an assistant professor in the chemistry department at Stanford University, makes cubic lattice models of designed heteropolymers. Lattice models have repeatedly shown that folding speed is determined by temperature. By analyzing the statistical properties of hundreds of folding events under simulated temperature constraints, a classical folding “pathway” can be determined for a given polymer (V.S. Pande et al., Current Opinions in Structural Biology, 8:68-79, 1997).

On the computer screen, the model can be seen beginning to fold, then disappearing, and reappearing in succession, before the native state finally asserts itself. Pande and colleagues are trying to understand folding in terms of the chemical kinetics of the transition between these folded and unfolded states. Delineation of the transitional state could provide a breakdown into a series of phases in the folding event that could reveal general dynamic properties, he says.

Physical Studies

Physicists aren’t only working in the theoretical realms; some have joined biologists in the study of real material. For example, Chu’s recent interest has been in observing the wide range of conformations assumed by individual polymer strands of DNA when they are stretched (D.E. Smith, S. Chu, Science, 281:1335-40, Aug. 28, 1998). His team currently is employing electrophoresis, fluorescent energy transfer, and confocal microscopy to examine the folding and attachment during ribosomal self-assembly of an essential subunit called S15. They also are using optical tweezers to stretch molecules of titin, a protein involved in protecting muscles from overstretching, in an attempt to witness the formation of its subunits.

At Cornell University, physics research associate Lois Pollack and colleagues are developing a new technique involving small-angle X-ray scattering to study submillisecond events in protein folding. Protein is placed in a mixing device with a buffer that compresses it. The protein folds as it flows out of the mixer. Multiple small-angle X-rays taken at different positions provide a time resolution while detecting scattered radiation from the proteins to discern whether they are compact or expanded. With this method, the researchers can monitor folding stages that occur in time intervals of less than 500 microseconds. “So what [the technique] does is change space into time, if that’s the way you want to think about it,” Pollack explains.

ne instrument now being used to study biomolecules manages to provide both the quantitative information prized by physicists and some of the intuitive information familiar to molecular biologists. The atomic force microscope (AFM), invented just 11 years ago, can visualize static and dynamic biomolecular associations. It also has the advantages of perceiving height and measuring the phase lag, or period of slowed change, as substances interact with different surfaces–for example, sticky, hard, or soft (S. Kasas et al., International Journal of Imaging Systems and Technology, 8:151-61, 1997). “This is another way we can get information that is to some extent intuitive,” notes Helen Hansma, an adjunct associate professor of physics at the University of California, Santa Barbara. “And as the maths and physics take over, it becomes somewhat less intuitive.”

Folding Phases Elude Solution

Relatively little is known for certain about the phases of protein folding. It is known that the initially unstructured polymers of polyamino acid form “secondary structures,” which are mostly alpha-helices and beta-sheets or strands. Also, side chains projecting off the main chain are formed. And, it is known that most of the energy driving the protein toward its native state is hydrophobic. Like oil, the hydrophobic side chains want to escape their aqueous environment, which they achieve by tucking themselves inside the protein. One theory, the framework model, has it that proteins first make their secondary structures, which then begin condensing toward the native state. A newer theory, the hydrophobic model, suggests that the main chain collapses due to hydrophobia before the secondary structures begin forming.

S. Walter Englander, a professor of biochemistry and biophysics at the University of Pennsylvania School of Medicine, suspects that the truth might lie somewhere between these two theories. He bases this notion on stopped-flow experiments that he and colleagues have been conducting.

The stopped-flow technique, in use for years, subjects protein to a very high concentration of a denaturant, such as urea, guanadine, or chloride, which induces the protein to unfold completely. The solution is then diluted, which almost instantaneously produces an environment conducive to refolding. Using light, a photomultiplier, and a signal such as flourescent staining, time-dependent changes in the cell can be measured. The protein folds within milliseconds, a speed that defies mathematical logic. This speed is vital, because unfolded polymers risk either being carved up by proteases, or sticking to each other and forming the aggregations associated with encephalopathies. “It’s a fairly hostile environment in the cell,” Englander notes, “so it’s important to know the real time scale of folding.”

Generally, when unfolded proteins in a concentrated denaturant are diluted, they immediately contract. This submillisecond “burst phase” was presumed to be an intermediate form of the native state, but Englander and colleagues recently showed otherwise. Comparing two ribonuclease A chains that were identical except that the disulfide bonds in one chain were broken and the protein therefore could not fold, the researchers observed the same initial burst of optical activity in both the disulfide-intact and disulfide-broken chains. This indicates that the burst phase is not an intermediate shape, but is merely a reaction to the dilution of the denaturant (P.X. Qi et al., Nature Structural Biology, 5:882-4, October 1998). Other work with intact cytochrome C protein resulted in similar findings (C.K. Chan et al., Proceedings of the National Academy of Sciences, 94:1779-84, 1997).

With much left to learn about protein folding, Englander believes that both theoretical and practical approaches have value. “You really have to attack a problem like this from all points of view,” he says. “The danger is, you can’t deal with representations of real protein molecules in a computer. It’s way too complicated …, and some models might be very misleading, because they’re not high-veracity. There’s a potential danger of throwing out the baby with the bath water.”

FOLDING PHASES ELUDE SOLUTION

Relatively little is known for certain about the phases of protein folding. It is known that the initially unstructured polymers of polyamino acid form “secondary structures,” which are mostly alpha-helices and beta-sheets or strands. Also, side chains projecting off the main chain are formed. And, it is known that most of the energy driving the protein toward its native state is hydrophobic. Like oil, the hydrophobic side chains want to escape their aqueous environment, which they achieve by tucking themselves inside the protein. One theory, the framework model, has it that proteins first make their secondary structures, which then begin condensing toward the native state. A newer theory, the hydrophobic model, suggests that the main chain collapses due to hydrophobia before the secondary structures begin forming.S. Walter Englander, a professor of biochemistry and biophysics at the University of Pennsylvania School of Medicine, suspects that the truth might lie somewhere between these two theories. He bases this notion on stopped-flow experiments that he and colleagues have been conducting.

The stopped-flow technique, in use for years, subjects protein to a very high concentration of a denaturant, such as urea, guanadine, or chloride, which induces the protein to unfold completely. The solution is then diluted, which almost instantaneously produces an environment conducive to refolding. Using light, a photomultiplier, and a signal such as flourescent staining, time-dependent changes in the cell can be measured. The protein folds within milliseconds, a speed that defies mathematical logic. This speed is vital because unfolded polymers risk either being carved up by proteases, or sticking to each other and forming the aggregations associated with encephalopathies. “It’s a fairly hostile environment in the cell,” Englander notes, “so it’s important to know the real time scale of folding.”

Generally, when unfolded proteins in a concentrated denaturant are diluted, they immediately contract. This submillisecond “burst phase” was presumed to be an intermediate form of the native state, but Englander and colleagues recently showed otherwise. Comparing two ribonuclease A chains that were identical except that the disulfide bonds in one chain were broken and the protein therefore could not fold, the researchers observed the same initial burst of optical activity in both the disulfide-intact and disulfide-broken chains. This indicates that the burst phase is not an intermediate shape, but is merely a reaction to the dilution of the denaturant (P.X. Qi et al., Nature Structural Biology, 5:882-4, October 1998). Other work with intact cytochrome C protein resulted in similar findings (C.K. Chan et al., Proceedings of the National Academy of Sciences, 94:1779-84, 1997).

With much left to learn about protein folding, Englander believes that both theoretical and practical approaches have value. “You really have to attack a problem like this from all points of view,” he says. “The danger is, you can’t deal with representations of real protein molecules in a computer. It’s way too complicated …, and some models might be very misleading, because they’re not high-veracity. There’s a potential danger of throwing out the baby with the bath water.”

–S.B.