Does flexibility increase protein stability?

Since my favorite physicist of the blogosphere put up a great post on the basics of NMR in preparation for a geeky post on the subject, I figure I'll add in a geeky post of my own. As Chad mentions in his post, the dependence of the resonance frequency on local chemical structure allows us to get a great deal of information about covalent bonds and their conformation from NMR spectra. In addition, certain NMR experiments can provide us with detailed information about the motions of atoms with respect to one another. Although the various approaches differ in the level of detail they offer, the sum of NMR dynamics experiments fairly effectively cover the range from picosecond motions to month-long fluctuations. My particular interest in these approaches is in their use to dissect intramolecular signaling in proteins.

The major part of my graduate studies at UNC under Andrew Lee was devoted to understanding the pathways by which changes in dynamics are propagated in eglin c (explore this protein at the PDB). Eglin c is an excellent subject for this kind of study because it has very good spectroscopic properties, and because it has no discernable allosteric properties at all. This means we can do a lot of different experiments on it, and also that our observations about long-range dynamic interactions are likely to be generalizable to any protein that has a hydrophobic core, not just allosteric ones.

My work with eglin c involved making mutations to the protein and determining what dynamic changes resulted. In the course of this I discovered several mutants that caused significant changes in the motions of amino acid side chains on the timescale of picoseconds to nanoseconds (1,2). This is the range covered by the Lipari-Szabo model-free formalism (3). One mutation in particular, V54A (read: valine 54 to alanine), caused almost universal rigidification of the core of eglin c on this timescale. However, the protein was significantly less stable.

This is curious because a lack of stability is known to correlate with an increased frequency of localized unfolding of the backbone. In elements of secondary structure, hydrogens attached to amide nitrogens tend to be caught in hydrogen bonds. When placed in a solution of deuterated water, these hydrogens can only be replaced by deuterium when the hydrogen bond breaks (this is called hydrogen-deuterium exchange or HX). This process can be monitored by NMR (and less specifically by mass spectrometry), and is known to speed up when the structural elements are less stable. So the V54A mutant is more flexible on this slower timescale, and less flexible with respect to some faster processes.

The significant caveat here is that we only looked at some groups in the protein. For reasons that are a bit too complex to discuss in this post, it was only feasible to observe the dynamics of methyl groups. This is a problem because the core of eglin c has a substantial number of aromatic side chains in it. Any picture of dynamic behavior would be incomplete without examining these side chains and quantifying the changes in HX behavior, which I did not do.

In an upcoming article in Biochemistry, Josh Boyer and Andrew Lee address these deficiencies (4). In order to look at the aromatic side chains, they employ a clever biosynthetic labeling scheme devised by Akke's group (5). While this only allowed them to look directly at δ-carbons of these residues, the general rigidity of aromatic rings meant that they could generalize the observations to essentially the whole side chain. They found using this approach that the dynamic response of aromatic side chains to the V54A mutation was more heterogeneous than that of the methyl side chains. As you can see in the figure I have shamelessly stolen from their paper, the central portion of the core appears to have rigidified across all residue types, while the edges of the core and the outward-facing residues of the protein all appear to have become somewhat more flexible. The location of V54 is shown here in black.

This heterogeneity of the dynamic response was also observed when the HX results were taken into account. The distribution of responses, however, is very unexpected. The top portion of the figure at left shows rigidified side chains as a transparent surface, and destabilized backbone amides (as measured by HX) as solid surfaces. Note that the two regions of destabilized amides are linked by contiguous side chains. The lower figure shows side chains that have become more flexible as transparent pink surfaces. The solid backbone surface indicates a region of the protein that has become more stable. You will notice that this stabilized region lies a significant distance from the mutation site (though as the figure above shows it is connected to it by a network of contiguous side chains), and that it corresponds with a bundle of aromatic sidechains that have become more flexible. A significant portion of the protein is clearly destabilized, as was expected from the previous results on V54A. However, the co-localization of side-chain rigidity and backbone instability (and vice-versa) doesn't jive with our expectations.

Our physical intuitions suggest that rigid things are stable and flexible things are not, and in many cases these intuitions have been borne out—in studies of backbone dynamics of proteins from thermophilic organisms, for example. However, the present results can be rationalized if one supposes that conformational entropy makes a significant contribution to stability. Indeed, because the structure of V54A is known to be essentially unchanged from wild-type (2), the enthalpic contributions (hydrogen bonds, charge-charge interactions, etc.) will not be altered. Entropic contributions will therefore dominated the observed changes in stability.

Rigid regions of a protein may have no place to disperse thermal energy other than localized unfolding, while flexible regions may have the option to dump some of that energy into side-chain fluctuations. When mutations increase the available space for those fluctuations, therefore, stability may be expected to increase, and vice versa. Obviously, the benefits of flexibility to ordered backbone structure will diminish as the fluctuations approach a magnitude that permits solvent to permeate the site.

Boyer and Lee's results, though probably expected by those who have followed previous research in eglin c, do not mesh with our macroscopic understanding of the relationship between rigidity and stability. However, they are not so outlandish that they can't be rationalized in terms of what we already know about proteins. An investigation of this anticorrelation in other eglin c mutants (promised in the paper) would be welcome. In addition, it will be interesting to see if this phenomenon is observed in other proteins (especially thermophilic proteins). As new labeling techniques and experimental approaches expand the range of molecules that can be effectively investigated using NMR, we may learn significantly more about the role of conformational entropy in the stabilization of protein folds.

1. Clarkson, M.W., Lee, A. (2004). Long-Range Dynamic Effects of Mutations Propagate Through Side Chains in the Serine Protease Inhibitor Eglin C. Biochemistry, 43(39), 12448-12458. DOI: 10.1021/bi0494424

2. Clarkson, M., Gilmore, S., Edgell, M., Lee, A. (2006). Dynamic Coupling and Allosteric Behavior in a Non-Allosteric Protein. Biochemistry, 45(25), 7693-7699. DOI: 10.1021/bi060652l

3. Lipari, G., Szabo, A. (1982). Model-free approach to the interpretation of nuclear magnetic resonance relaxation in macromolecules. 1. Theory and range of validity. Journal of the American Chemical Society, 104(17), 4546-4559. DOI: 10.1021/ja00381a009

4. Boyer, J.A., Lee, A.L. (2008). Monitoring Aromatic Picosecond to Nanosecond Dynamics in Proteins via 13C Relaxation: Expanding Perturbation Mapping of the Rigidifying Core Mutation, V54A, in Eglin c. Biochemistry DOI: 10.1021/bi702330t

5. Teilum, K., Brath, U., Lundstrom, P., Akke, M. (2006). Biosynthetic ¹³C Labeling of Aromatic Side Chains in Proteins for NMR Relaxation Measurements. Journal of the American Chemical Society, 128(8), 2506-2507. DOI: 10.1021/ja055660o