The Protein Folding Funnel

How Water Cages Push Polypeptide Chains Past Levinthal's Combinatorial Trap

At a Mössbauer spectroscopy workshop in Monticello, Illinois, in 1969, Columbia University molecular biologist Cyrus Levinthal performed a back-of-the-envelope calculation that haunted biophysics for two decades. Consider a modest polypeptide chain of just 100 amino acids emerging from a ribosome. Because each peptide backbone link can adopt roughly three stable rotational bond angles, the chain possesses at least $3^{100} \approx 5 \times 10^{47}$ distinct three-dimensional conformations. Even if thermal agitation allowed the chain to sample a new conformation every picosecond ($10^{-12}$ seconds), testing every state to find the single biologically active native fold would require $5 \times 10^{35}$ seconds. That duration exceeds the 13.8-billion-year age of the universe by more than twenty-five orders of magnitude.

Yet inside living cells, proteins fold reliably into their exact native geometry in milliseconds to seconds. Christian Anfinsen proved in 1961 at the National Institutes of Health that denatured bovine pancreatic ribonuclease refolds spontaneously even inside a purified glass test tube. Furthermore, folding a flexible random coil into a single compact crystal-like structure dramatically reduces the conformational entropy of the polypeptide chain ($\Delta S_{\text{chain}} < 0$). How does a protein beat both Levinthal's combinatorial clock and the second law of thermodynamics at the same time?

The thermodynamic answer was uncovered in 1959 by Princeton chemist Walter Kauzmann: a protein never folds in a vacuum. It folds inside liquid water. Roughly half of a typical protein's amino acid side chains—such as leucine, isoleucine, valine, and phenylalanine—are oily, non-polar hydrocarbons that cannot form hydrogen bonds with water. When an unfolded polypeptide sits exposed to water, surrounding $\text{H}_2\text{O}$ molecules are forced to arrange themselves into rigid, ice-like clathrate cages around every exposed oily side chain. Those cages inflict a severe loss of solvent translational and rotational entropy. When the protein collapses to bury its oily residues inside a dry central core, hundreds of caged water molecules are liberated back into the bulk liquid. The massive gain in water entropy ($\Delta S_{\text{solvent}} \gg 0$) easily pays for the loss of backbone conformational entropy ($\Delta S_{\text{chain}} < 0$), driving the total free energy $\Delta G$ negative.

The kinetic answer to Levinthal's clock arrived between 1987 and 1995. Theoretical physicists Peter Wolynes at the University of Illinois and José Onuchic at the University of California, San Diego, applied spin-glass statistical mechanics to formulate the protein folding funnel. Levinthal had implicitly assumed a flat golf-course energy surface where the chain wanders blindly on a level plain until stumbling into a single tiny hole. Wolynes and Onuchic proved that natural selection sculpted evolved protein sequences according to the principle of minimal frustration.

Instead of a flat plain, the protein's free-energy surface is shaped like a steep, multidimensional funnel. The horizontal width of the funnel at any depth represents the chain's conformational entropy $S_{\text{chain}}$, while the vertical axis measures internal plus solvent free energy. Whenever hydrophobic collapse and local alpha-helices form native contacts, the energy drops faster than the conformational width narrows. Millions of denatured molecules roll downhill simultaneously along parallel trajectories toward the narrow native minimum at the bottom of the funnel. In 2003, biophysicist William Eaton and his team at the National Institutes of Health used laser-induced temperature-jump spectroscopy to track a thirty-five-residue headpiece subdomain of the actin-binding protein villin. The polypeptide snapped from a disordered random coil into its native three-helix bundle in just 700 nanoseconds—roughly $10^{38}$ times faster than an unguided Levinthal search across its conformational space.