Our result confirmed the older hypothesis of Krimm that unfolded peptides and proteins favor PII structure rather than random coil. Still this hypothesis was not accepted by many, and had been attacked in the literature. The result surprised us, since we were unaware of Krimm’s work when we began the experiments. The conclusion that there is local structure in unfolded proteins has several important implications, despite the fact that it is still not generally accepted. It is now suspected that a significant fraction of proteins are only weakly folded in vivo, and unfold on extraction from the crowded milieu in cells. This makes it timely to define the structure present in unfolded proteins. Our work redefines the starting state for folding as one with significant order, changing the interpretation of folding "funnels" for example. The funnel metaphor postulates a vast conformational ensemble of energetically equivalent unfolded states in the fully disordered coils at the top of the funnel. Our results reduce this entropy a great deal, and in doing so provide one more response to Levinthal’s paradox. Finally I should mention that the conformation of short di-, and tri-peptides are also dominantly PII as found by other spectroscopic studies which we became aware of later on. These short peptides have been used to calibrate force fields for calculations of biological polymers in water. The finding that such very short chains favor PII conformation in water means that several force fields and/or water models in wide use need re-examination.

Proteins are the workhorses in any cell, giving cells shape, mobility, energy, and nutrients. Most proteins become active by virtue of "folding" into a specific intricate 3D structure, the so-called native state. Many proteins have the capacity to regain their native state after heating destroys that structure. This process is among the most complex and poorly understood today, starting with the state of proteins when they are unfolded. This paper presents evidence that there is a local structure in unfolded proteins that guides the early stages of folding, and thus speeds up the folding reaction. Surprisingly this structure is related to that in the protein collagen found in our skin and bones, but not the structure of native globular proteins. In addition to its presence in unfolded proteins, evidence is accumulating that the structure we describe is an intermediate in formation of amyloid fibrils, the aggregates that accumulate in Alzheimer’s disease and other disorders. I should add that the paper took over a year to get published because an initial reviewer rejected it as completely wrong and suggested we had not read the literature in this field!

The idea for this project emerged from a talk I had with George Rose at a meeting in Poland attended also by Buzz Baldwin. Rose had just described his calculations on short alanine peptides that suggested steric interactions alone should prevent them from acquiring random coil structure, and boldly predicted we would find structure in such peptides. Since short chains of alanine rapidly become insoluble in water as the chain length increases, I mentioned that an undergraduate in my lab, Anders Olson—currently a grad student at Cal Tech—had found a way to solubilize chains of alanine by adding pairs of basic side chains at the ends of the sequence. To minimize interactions of the basic side chains with residues in the middle of the chain, we used bases with side chains shorter than Lys for the caps. Thus we came up with the design for XAO, an 11 mer peptide model in which seven alanines are flanked by pairs of short chain basic side chains at the N and C termini. Anders synthesized the peptide, and measured its CD spectrum. The NMR behavior of the peptide gave us the first surprise: each NH proton in the peptide had a distinct chemical shift. The spectrum was assigned by Kevin (Zhengshuang) Shi—currently a postdoc at Penn—who measured the coupling constants of the backbone amides as a function of temperature. Unexpectedly, the coupling constants change (increase) substantially with temperature. This is inconsistent with a coil, for which we expect the values to be invariant or to change insignificantly with temperature. We could then determine the phi dihedral angles of the backbone. The values were low, but still between those for α and β, hence could still be a mixture of the two structures as the coil model would predict. We could eliminate the possibility of enough alpha helix because the characteristic helical NOEs were absent. On the other hand, the coupling constants were too low to permit substantial beta structure. This left us with PII as the only dominant alternative.