Skip to Content

A New Factor in Protein Folding

I linked to this particular XKCD strip when it came out, but it came right back to mind when I saw this paper, on a new kind of intramolecular interaction in proteins (and other systems). Protein folding, macromolecular folding in general, is indeed a famously horrendous problem to attack from first principles. There’s been progress on semi-empirical approaches based on the protein structures that we already know, but there’s still a lot to understand.

Protein folding 

It all comes down to the many sorts of interactions (some attracting, some repelling) that the protein can make with itself, compared to what can be had with the solvent. Enthalpic and entropic terms are dueling it out in each case for the final free energy figure, and things like the disposition of single water molecules can make a big difference – and a difference that might totally flip from negative to positive in a different example. To counterbalance those big influences, there are a large number of very small ones, each of which adds little to the binding picture on its own, but which as a collection can send things over into particular paths.

That’s what this latest paper is talking about. Apparently, calculations have suggested that C-H bonds can have a weak interaction with the pi bonds in carbonyls. That’s definitely not one of the the bonding modes that anyone would think of first, but the paper demonstrates actual detection of (very, very small) coupling constants in 2D NMR (HMQC)  spectra that are due to through-space interactions of (say) the methyl groups in a valine with the carbonyl of an appropriately situated backbone CO group. There doesn’t seem to be much correlation with the type of CO, what amino acid it’s part of , or whether it’s involved in a hydrogen bond with some other group, etc. This experiment needed some specific isotopic labeling to get the signal/noise up to where the couplings could be detected, and the authors believe that many others will be found in other labeling experiments. They suggest that it’s not going to be just methyl groups, but OH groups and others that will prove to participate another order of weak hydrogen-bond-like interactions.


So this could be a new way to interrogate protein structures in solution, and it seems to represent a new consideration to add into protein folding calculations. There will doubtless be some systems where these interactions are buried down in the noise compared to the other factors, and others where they really do add up to something, but we don’t know much about this yet. It’s worth thinking, though, that this is a phenomenon that until now we haven’t even known about. Every time something like this is discovered, I ask myself what else we don’t know. (Hint: there’s a lot). It’s great. All that stuff is waiting out there for us.

17 comments on “A New Factor in Protein Folding”

  1. Curious Wavefunction says:

    I don’t have access to the paper, but have they compared these contacts with the “null model”? In other words, one can see lots of close contacts in protein structures that are hard to label as “bonds” or attractive interactions. As Jack Dunitz put it, “atoms have to go somewhere”, so how do we know that these are not just close contacts that are simply side products of the folding process rather than bonafide attractive interactions in their own right?

    1. I strongly agree with Ash here–there is a dangerous tendency to conflate close contacts in a crystal structure with “interactions.” There is undoubtedly a slightly attractive interaction between carbonyls and nearby alkyl side chains (driven mostly by dispersion effects, with a small amount of induction effects tossed in); however, there are attractive interactions between ALL neutral molecule fragments, so it’s unclear whether this is anything special and whether it will prove useful in understanding protein structure. It may well be that the observed contacts are a consequence of the fold, not a cause.

      1. Peter Kenny says:

        Ash/Steven, as you correctly point out, the key question is whether the NMR couplings quantify strength of interaction or proximity. As Ash has heard me say (on more than one occasion) the contribution of an intermolecular contact to affinity (or folding free energy) is not an experimental observable. It’s also worth noting that contact between polar and non-polar surfaces is not inherently repulsive even if it is destabilizing in the context of the protein-ligand complex. Even if it were possible to demonstrate (by measuring the strength of interaction as a force constant) that a protein-ligand contact in the complex was attractive in the complex, it would not necessarily follow that the contact stabilized the complex in aqueous solution. I have linked a recent article on hydrogen bonding in which some of this is discussed.

  2. Samuel Schneck says:

    It seems like this work is still thinking in terms of bonds.

    This “C-H carbonyl-pi” interaction is electrons and protons coming together in a way that we didn’t expect. Pretending that it’s some new species of chemical bond is classical chemical thinking. That approach was fine when all we were trying to figure out was which atoms are stuck to which other atoms, but this is a physics/computation problem, and not one that chemical thinking will be particularly helpful.

    What we have here is things sticking to other things in monstrously complicated ways. Making leaky abstractions so that we can pretend to have intuition about what’s going on isn’t going to be helpful.

    1. James Weiss says:

      Typos and grammar errors are really helping your argument here buddy.

    2. Derek Lowe says:

      Hmm – if every time the word “bond” came up, it were to be replaced by “interaction”, would that be better? I suppose what I’m saying is that if some of these “monstrously complicated ways” turn out to bin into roughly defined piles (what we call hydrogen bonding, the various kinds of pi-cloud effects, this new proposal, etc.), is there a problem with using those as separate concepts?

    3. Humulonimbus says:

      Samuel, you are right that this interaction is something we didn’t expect, but I don’t think that means this interaction has no chemical explanation/implication. These interactions are detected via J-couplings: an indication that the protons are in nondegenerate CHEMICAL environments.

      Maybe the term “bond” should be used more carefully, but that’s the imperfect language we have invented to describe these phenomena. H-bonds, for instance, aren’t the same things as covalent bonds, but they can indicate which bonds are likely to break and form in acid-base chemistry and are commonly used to lower activation energy barriers.

      Physics/computation problem is not mutually exclusive to a chemistry problem. And when you say “protons and electrons coming together,” sorry, that is a chemistry problem too. You’re forcing a dichotomy that just isn’t there.

      1. Samuel Schneck says:

        I didn’t make my point very well, and you’re right that these “interactions” as I termed them, are the basis of what chemists think of as “bonds.”

        Unfortunately, the division of interaction into bond and non-bond, having the luxury of considering some interactions more important than others, doesn’t work on a problem like protein folding, where we don’t understand what matters and what doesn’t.

        At some point, you might define a “hydrophobicity bond,” where two non-polar regions of the relevant species favor contact with each other over the solvent. How different is this than a reaction where the proximity of two atoms changes to the favored energetic configuration?

        Now, fair enough, I’m flattening the bond hierarchy, a hierarchy that has been, and will continue to be useful to the day to day practice of chemistry.

        I just don’t see the purpose of imposing that hierarchy in a system that tries to reason mathematically about a complex system where anything, not just what we see as “bonds” can be important.

        As to the NMR data, that’s useful information about what’s going on in the system, but I don’t see the interest or relevance of this result unless it influences our understanding of the first principles involved, which I believe is where any useful progress on this problem is going to come from.

        Then again, I’m out of my depth in a field I left a long time ago, so I should probably stop talking before I make an even bigger fool out of myself.

        Good day to you all, and may your lives be full of joy and wonder.

        1. Humulonimbus says:

          “Unfortunately, the division of interaction into bond and non-bond, having the luxury of considering some interactions more important than others, doesn’t work on a problem like protein folding, where we don’t understand what matters and what doesn’t.”

          I totally agree with you on this. This gets back to Ash’s first comment. Just because the atoms are in close proximity does necessitate that there is some sort of interaction pulling them together, they could be pushed together by the myriad of other factors we still don’t understand about how proteins fold. That’s why in my comment I said “chemical explanation/implication” – meaning it’s not clear whether this interaction is a cause or an effect. Even if it’s the latter, though, there is a glimmer of hope for a new way to activate C-H bonds, which is still interesting.

          Did not intend to shut you down. For what it’s worth, I don’t think you’re out of your depth.

  3. tlp says:

    We see only what our instruments allow us to see. Astronomers knew that since 1920-es

  4. Anon says:

    Why the surprise? Wherever any atoms and/or groups are in close proximity to each other, there is bound to be some electrostatic force and thus interaction. That’s basic stuff you learn in junior school science class, let alone high school physics.

  5. Barry says:

    I’ve never relied on non-bonded through-space J-coupling data, but it’s been in the literature for years. ‘J’ need not imply bond

  6. Nick K says:

    A stupid question perhaps, but I’ll pose it anyway: how do we know this interaction isn’t merely the Van der Waals force?

    1. tangent says:

      Sign that question yup for that question too. Would you not expect this configuration to show some dipole-polarization (Debye) force?

      (Is the point here that this specific pairing shows more attraction than existing models would have suggested?)

  7. Chris Phoenix says:

    Here’s a somewhat related question/speculation:

    Molecular machines seem to be efficient, in part, because of a very flat energy trajectory: thermal noise is enough to push them back and forth through their trajectory – nothing goes “clonk” from one state to another and dissipates energy.

    This energy trajectory is flattened, in part, by tuning springs. But a lot of the springs are entropic rather than bond-strain. And the “spring constant” of entropic springs depends on temperature.

    Thus, it will be difficult to tune a molecular machine to work over a wide range of temperatures.

    Thus, the molecular machine proteins in a warm-blooded animal can be somewhat simpler than the comparable proteins in a cold-blooded animal, and can have more degrees of freedom to evolve while retaining functionality.

    Thus, warm-blooded animals have a couple of advantages over cold-blooded animals, which may compensate for the rather high energetic cost of maintaining body temperature.

    The testable prediction is pretty obvious. I haven’t seen this discussed anywhere, but it’s not my field and I don’t really know how to do a good search.

    A corollary of this theory is that, when a warm-blooded molecular machine stops working outside its “designed” temperature range, it may not actually be denatured; it may simply be that its springs are not tuned correctly to allow thermal noise to drive it through its cycle. Kind of like how an internal combustion engine may not turn over in very cold weather: nothing has broken or even changed shape, it’s just that the starter motor is not strong enough to overcome the “energy barrier” of the cold oil.


  8. Henry Rzepa says:

    It is sometimes of interest to see what “small molecules” are up to. A search of the small molecule crystal database (CSD) reveals a number of interesting CH⋅⋅⋅π Interactions between methyl and carbonyl groups. See DOI: 10.14469/hpc/2595

    1. Paula Lario says:

      As someone who has spent many years looking at sub-atomic resolution structures, I will say that the interactions between aliphatic carbon atoms and hydroxyl and pie bonds are weakly attractive, especially methylene carbons beside electron withdrawing groups. This conclusion is based on observations of the local environment and atom placement, where alternate conformations are are possible but routinely not observed in buried but not too crowded hydrophobic environments. Yes, even with the consideration of unresolved buried water molecules. These weak interactions (I considered as Van der Waals) in my mind would affect the folding pathway just due to their numbers by creating a kinetic barriers.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.