Commentary

Are denatured proteins ever random coils?

Robert L. Baldwin* and Bruno H. Zimm†

*Department of Biochemistry, Beckman Center, Stanford University School of Medicine, Stanford, CA, 94305-5307; and †Department of Chemistry, University of California, San Diego, La Jolla, CA, 92093-0314

o polymer chemists, the term random and he used optical rotatory dispersion, the denatured protein is predisposed to Tcoil means a chain molecule whose the forerunner of circular dichroism used adopt specific backbone conformations at backbone coils randomly in three-dimen- today, to detect secondary structure. Tan- the start of refolding, when stronger hy- sional space, with a conformation de- ford found that denatured proteins in 6 M drogen bonds come into play (compare scribed by the Gaussian probability func- guanidinium chloride can be described as ref. 9). tion (1). To protein chemists, the term random coils in the sense of being devoid Their view of the denatured protein random coil carries an added special of all secondary and tertiary structure. He connects with the hierarchic model (re- meaning: the backbone conformation of pointed out, however, that thermally de- views, refs. 10–12) in which folding begins every amino acid residue, described by its natured proteins in water have optical in the backbone by forming specific local phi, psi pair of backbone angles, is inde- rotatory dispersion spectra indicative of structures, and these then interact during pendent of the conformations of neigh- some secondary structure, possibly resid- subsequent stages to make tertiary inter- boring residues. This is Flory’s isolated- ual native structure. actions that establish the native topology. pair hypothesis (2). If it is valid for Modern NMR studies of denatured The hierarchic model implies that the denatured proteins in specified solvent proteins are able to use 13C- and 15N- denatured protein is poised, ready to conditions, then the random coil protein isotopic labeling, combined with three- adopt specific backbone structures once in these conditions has no structure char- dimensional spectral resolution ap- the denaturant is diluted out. The results acteristic of the native protein, which proaches, to assign resonances and of Pappu et al. show that favored backbone could be used to guide the first steps of the measure coupling constants (6) in individ- conformations already preexist in the de- refolding process. Evidently, study of the ual amino acid residues. Using this ap- natured protein. The hierarchic model is folding problem would be greatly simpli- proach, Hennig et al. (7) find no evidence being used with some success currently to fied if conditions could be found in which for nonrandom structure in denatured hen predict folding rates and transition state denatured proteins assume the random lysozyme (8 M urea, pH 2.0) except for structures (reviews, refs. 13 and 14). A coil conformation. In proposing the hydrophobic clusters, and they suggest main reason for some skepticism about COMMENTARY isolated-pair hypothesis, Flory was moti- that their results can be described to a the hierarchic model has been that local vated by his own experimental analysis of good approximation by a statistical ran- backbone structure cannot be detected in the chain dimensions of some peptides: dom coil model. The distribution of resi- peptides from the small protein CI2 while see especially his later work on poly-L- due conformations found in this way mutational evidence (‘‘phi values’’) points alanines (3). agrees satisfactorily with a random coil to nucleation of folding at the level of Pappu et al. in this issue of PNAS (4) model for denatured proteins (COIL) de- tertiary interactions (15). However, a re- now find that the isolated-pair hypothesis rived from the Protein Data Bank, by cent mutational study (16) of the structure is valid for alanine peptides only within a making a library of residue torsion angles responsible for the formation of the 14–38 restricted region of the Ramachandran from all residues except those included in disulfide bond at the start of refolding, map, near the ␤-strand or extended re- regular secondary structures (8). when pancreatic trypsin inhibitor refolds gion. For example, steric exclusion by con- The concept of denatured proteins as oxidatively, reveals that a backbone tacts between residues separated by 3–6 random coils in specified solvent condi- ␤-hairpin, present at a level too low to be peptide bonds is pronounced near the tions allows great freedom in hypothesiz- detected by ordinary methods, is respon- right-hand ␣-region of the map. The result ing different models for the initial events sible. Formation of the 14–38 disulfide is to stiffen the peptide chain and reduce in folding. However, the recognition that bond stabilizes the hairpin structure and the number of possible conformations. denatured proteins are not in fact random the overall process acts as a funnel to Pappu et al. obtain their results by exhaus- coils focuses interest on possible struc- direct folding. tive enumeration of all possible confor- tures in the denatured protein. Pappu et al. Pappu et al. (4) point out an additional mations, using a simple potential function (4) find that steric clashes among residues reason for testing the validity of the in which the only variable parameter is separated by 3–6 units eliminate large isolated-pair hypothesis. When a dena- the strength of an interaction treated as a numbers of backbone conformations in tured protein folds to its unique native peptide hydrogen bond, which might certain regions of the Ramachandran conformation, there is a large decrease in include contributions from related map. Their result is particularly striking backbone conformational entropy, which interactions. because it is obtained with alanine pep- is difficult to measure or estimate accu- Tanford (5) studied whether denatured tides, because alanine has only a methyl rately. Its value is needed to quantitate the proteins are in fact random coils in 6 M group for a side chain. Steric clash should thermodynamics of folding. Most esti- guanidinium chloride, a denaturant that be enhanced with larger side chains, es- mates of the entropy change on folding are ␤ unfolds nearly all water-soluble proteins. pecially -branched side chains. The cal- based on the assumption that the dena- He used hydrodynamic properties such as culations of Pappu et al. almost certainly tured protein is a random coil. First intrinsic viscosity to characterize the over- will stimulate experimental efforts to find all dimensions of the polypeptide chain, preferred conformations in short pep- after reducing any disulfide bonds present, tides. A main result from their work is that See companion article on page 12565.

PNAS ͉ November 7, 2000 ͉ vol. 97 ͉ no. 23 ͉ 12391–12392 Downloaded by guest on October 1, 2021 thoughts on this entropy problem took the mer, then the number of states in the unfolded polypeptide with 100 peptide synthetic polymer chain as a guide. Con- random coil is astronomical (3100 is about units is much, much less than 3200 (there sider, for example, a chain of N carbon 5 ϫ 1047), and it is inconceivable that the are two backbone bonds per peptide unit atoms. There are three rotational confor- chain could ever find its way to fold into a about which free rotation might take mations at each carbon atom: a ‘‘trans’’ unique ‘‘native’’ structure without the aid place). Further, there are important state and two ‘‘gauche’’ states. In the of steering interactions. steering effects arising from hydrogen simplest carbon polymer, polymethylene, Clearly some caution should be used bonds and related interactions between all three have approximately the same by those inclined to make rough calcu- residues. Understanding these and re- energy within about 1 kcal͞mol. Then a lations of the backbone entropy change lated mysteries will require more studies random coil of length N has approxi- on folding by using the exponential func- like that of Pappu et al. and doubtless will mately 3N independent states. If N is about tion. The study by Pappu et al. shows that require an intimidating amount of com- 100, representative of a very small poly- the number of states accessible to an puter power.

1. Flory, P. J. (1953) Principles of Polymer Chemistry 7. Hennig, M., Bermel, W., Spencer, A., Dobson, 11. Baldwin, R. L. & Rose, G. D. (1999) Trends (Cornell Univ. Press, Ithaca, NY). C. M., Smith, L. J. & Schwalbe, H. (1999) J. Mol. Biochem. Sci. 24, 77–83. 2. Flory, P. J. (1969) Statistical Mechanics of Chain Biol. 288, 705–723. 12. Honig, B. (1999) J. Mol. Biol. 293, 283–293. Molecules (Wiley, New York). 8. Smith, L. J., Bolin, K. A., Schwalbe, H., 13. Baker, D. (2000) Nature (London) 405, 39–42. 3. Conrad, J. C. & Flory, P. J. (1976) Macromolecules MacArthur, M. W., Thornton, J. M. & Dobson, 14. Eaton, W. A., Mun˜oz, V., Hagen, S. J., Jas, G. S., 9, 41–47. Henry, E. R. & Hofrichter, J. (2000) Annu. Rev. C. M. (1996) J. Mol. Biol. 255, 494–506. 4. Pappu, R. V., Srinivasan, R. & Rose, G. D. (2000) Biophys. Biomol. Struct. 29, 327–359. Proc. Natl. Acad. Sci. USA 97, 12565–12570. 9. Srinivasan, R. & Rose, G. D. (1999) Proc. Natl. 15. Itzhaki, L. S., Otzen, D. E. & Fersht, A. (1995) J. 5. Tanford, C. (1968) Adv. Protein Chem. 23, 121–282. Acad. Sci. USA 96, 14258–14263. Mol. Biol. 254, 260–288. 6. Hu, J.-S. & Bax, A. (1996) J. Am. Chem. Soc. 118, 10. Baldwin, R. L. & Rose, G. D. (1999) Trends 16. Zdanowski, K. & Dadlez, M. (1999) J. Mol. Biol. 8170–8171. Biochem. Sci. 24, 26–33. 287, 433–445.

12392 ͉ www.pnas.org Baldwin and Zimm Downloaded by guest on October 1, 2021