Tanford's Top Model

Of course, as soon as I criticize PNAS, something interesting comes out in it, although what interests me often has nothing to do with what interests people generally. In this case I am talking about protein folding, so Elan, you can go browse some other site. More specifically, I am talking about the unfolding of proteins, and how it is accomplished by urea, the lovely little molecule over to the right. Chemical unfolding (also called denaturation) of proteins using urea or guanidine hydrochloride has been a popular method of investigating protein stability and the kinetic determinants of folding pathways. My own work with Drew and Marshall made use of guanidine denaturation to determine the stability of eglin c, but Dave Shortle's examination of residual structure in denatured eglin c used urea. Blindingly useful as this little molecule is, experiments with it have been plagued by one niggling little question: exactly how does it work?

More than 40 years ago, Charles Tanford proposed a model of urea-denaturation energetics based on the individual contributions of the protein backbone and side chain, and the assumption that all these contributions were additive. From this excellent work, it was thought that the side-chains contributed significantly to the transfer free energies (ΔG(tr)) for each amino acid, and that interactions between urea and the side-chains were therefore substantially responsible for urea's denaturant capability. The observation that the m-value of a given protein — which essentially measures how good urea is at unfolding it — correlates with the surface area exposed on unfolding was taken as support of this proposition.

Looking at urea, it is difficult to see why this might be so. Urea is incredibly soluble in water (up to 10 M), on account of its ability to form numerous hydrogen bonds. That ionic or polar side chains such as D, E, K, R, Q, N, S, and T might prefer interactions with urea to interactions with other side chains is easy to understand, but these residues tend to be on the surface of the protein anyway. Even if urea occupies the hydrogen-bonding sites of these side chains, it's difficult to understand why this would lead to anything other than the destruction of loop-loop interactions. When we come to hydrophobic side chains the idea that urea solubilizes them becomes even harder to fathom. What sort of interactions would this very polar molecule enjoy with nonpolar side chains?

This has led some to claim that interactions with the backbone are equally (or more) important in urea denaturation, a proposition supported by some simulations. In this regard, however, simulations are always questionable because the progress in replicating the properties of water alone in simulation has been slow, let alone water acting in concert with a solute like urea.

Auton, Holthauzen, and Bolen claim to have now perfected Tanford's work by correcting the ΔG(tr) that underlie the energetic predictions for the activity coefficients of glycine, which are elemental to the predicted ΔG(tr) for all amino acids. They find that, given these new ΔG(tr), the Tanford model can be used to predict the m-values of a number of proteins. More significantly, they find that the primary contributor to the denaturation effect of urea is the peptide backbone, not the side-chains. In fact, they propose that the contribution of the backbone has been underestimated by as much as 40 (kcal/mol)/M.

They convincingly demonstrate their claim in Figure 3 of their paper (left). I would have preferred a different scale for this graph (a glance at Figure 2 will explain why they used the one they did), but even so the linear correlation between their predictions and the observations is fairly compelling. The dashed red line I've added here is roughly along the correlation you get if you use Tanford's original values (if you're curious this can be found in the paper's Figure 2). The upper and lower bounds pointed out here refer to the degree of residual structure in the denatured state. The implication from this figure is not only that the backbone is the primary determinant of the m-value, but also that most proteins have neither a fully random coil (lower bound) or a compact denatured state (upper bound). According to Auton et al. the central line actually approximates a self-avoiding random coil. Deviations from this line may reflect the existence of more or less residual structure in the denatured state of the various proteins.

Another implication of this paper is that the perceived relationship between m-value and exposed surface area in the denatured state is actually standing in for a simpler relationship. The analysis here suggests that although the exposed surface area is dominated by the side chains, the energetics are dominated by the backbone. However, the surface area should be roughly correlated with the number of peptide moieties available for interactions with urea. Thus, the length of a protein chain may be the most significant determinant of the m-value. If the argument of the paper is correct, the m-value should be directly correlated with the number of hydrogen-bonding-competent groups exposed in the denatured state.

This paper is not as clearly written as it could be, and will be a little difficult to digest for those who haven't thought seriously about chemistry in a while. Nonetheless, with a dedicated effort it should be possible for even a non-expert to follow the argument. I'm not sure about the delay, but I think PNAS makes articles open-access after 6 months, at which point you might want to check it out.

Auton, M., Holthauzen, L.M.F., and Bolen, D.W. "Anatomy of energetic changes accompanying urea-induced protein denaturation." Proc. Nat. Acad. Sci. 104 (2007) p. 15317-15322.