Skip to Content

Nature Doesn’t Abhor a Vacuum As Much As You’d Think

I wrote some years ago about the case of a protein that seemed to have a completely empty binding pocket – empty, as in not even any water molecules hanging around in there. There are a number of these known, and there’s a lot of arguing about them among both experimental and computational chemists. You’d think that clearing a void like that completely of solvent would be energetically costly – I mean, surely water molecules can generally find something to interact with, right? (The flip side of that argument is that they they generally can, thus the relative rarity of these empty pockets).

Here’s a new paper investigating this phenomenon, using the well-known enzyme thermolysin. It’s known to have a binding pocket one residue’s worth away from the active site that very much prefers greasy aliphatic side chains (model protein substrates show that it particularly likes valine, leucine, and isoleucine in that spot). So there’s certainly a hydrophobic cavity (as confirmed by X-ray crystal structures), but just how hydrophobic is it? According to this new work, very. The team (from Marburg) used a weak ligand that doesn’t avail itself of the pocket, but rather bridges over it, largely covering it up. Careful X-ray work quantified just what sort of electron density remained inside the cavity, and the answer is “none”. They saw no evidence for any water molecules in there. Noble gas atoms made their way in if you exposed the system to such things, but water veers off.

If proteins really can have such regions, there’s a huge amount of favorable binding waiting to be picked up by any similarly hydrophobic ligand that can access them. And so it proves in thermolysin: as the SAR of its substrates (and inhibitors) would suggest, you can pick up 40,000-fold binding affinity by dropping an isobutyl group in there. Careful isothermal calorimetry showed that this was completely due to changes in enthalpy; entropic changes were basically zero. There is no desolvation penalty to be paid by the residues in the protein pocket, because there’s nothing to desolvate. That accounting was done when the protein folded. In fact, it’s  only with the larger side chains that you start to see a desolvation penalty at all, and that comes from the water molecules around the side chains themselves. Other than that, it’s pure profit, energetically speaking.

So there really are situations where it’s more energetically favorable to have a totally empty cavity in a protein’s three-dimensional structure – a tiny zone of utter vacuum – than it is to let even one water molecule in there, even though the protein itself is swimming in the stuff. The take-home lessons of this paper are not only that these cases are real, but that they represent a huge opportunity for picking up binding affinity when they are identified. Actually, there’s at least one more lesson – that medicinal chemists’ explanations of solvation and compound binding are, at least in some cases, a bit too pat. We need to realize that, as odd as it may seem to us, that there are bubbles of vacuum in some proteins, and be ready to take advantage of that fact. And we should also be wary of any computational approaches that try to model water molecules into such spots because we think that they should be there. In some cases, they just aren’t.

20 comments on “Nature Doesn’t Abhor a Vacuum As Much As You’d Think”

  1. Wavefunction says:

    “there are bubbles of vacuum in some proteins, and be ready to take advantage of that fact.”

    Wonder if MD simulations can discover these pockets.

  2. Mark Thorson says:

    Huh! A vacuum can be a structural component! Maybe it acts as a bearing for sliding or rotational movement.

  3. Barry says:

    Hydrophobic pockets in proteins have been known since at least ’84, and Petsko’s Xenon soaking–often used to solve phase–points out many of them.

  4. AndyM says:

    The paper seems to neglect complementary and prior deuterium exchange experiments of thermolysin. Specifically, studies like “Conformational Dynamics of Free and Catalytically Active Thermolysin Are Indistinguishable by Hydrogen/Deuterium Exchange Mass Spectrometry”, Liu and Konermann, Biochemistry, 2008, 47 (24), pp 6342–6351. It’s interesting that in that study that deuterium exchange of residues 139-155, which forms a alpha helix that frames the S1′ pocket, was extremely slow.

    “A comparison with previous X-ray data suggests that these peripheral regions (to the catalytic site) undergo quite pronounced structural changes during the catalytic cycle. In contrast, active site residues exhibit only a moderate degree of backbone flexibility, and the central zinc appears to be in a fairly rigid environment. The presence of both rigid and moderately flexible elements in the active site may reflect a carefully tuned balance that is required for function. Interestingly, the HDX behavior of catalytically active thermolysin is indistinguishable from that of the free enzyme.”

  5. Peter Kenny says:

    I can’t see the article but I wonder if solubility of water in cyclohexane (0.003 M) or hexadecane (0.002 M) could have been used to set some sort of reference. I’m not sure how well the contact between water and non-confined hydrophobic surfaces is understood, especially when the hydrophobic and polar regions of molecular surface are interspersed. I’ve linked a review by Gerhard Hummer and colleagues as the URL for this comment.

    1. Barry says:

      adding a lipophilic co-solvent (e.g. THF) is likely to denature your protein. But you can find these voids by doing the crystallization in the presence of Xenon

  6. tlp says:

    Cool! One can deliberately engineer bunch of hydrophobic pockets into some protein to make a tiny vacuum cleaner! Or is it what albumin does anyway?

    1. Barry says:

      the Shokat group has been engineering just such voids into protein kinases for years

    2. NJBiologist says:

      Thank you, tlp. That gave me a much-needed laugh.

  7. Anon says:

    I wonder if this would be supported by NMR, or whether the crystal lattice is somehow stabilizing the structure to leave the cavity unfilled?

    Question: What’s the biggest hydrophobic pocket that water could get into, but doesn’t?

  8. provocateur says:

    This notion of describing an empty space in proteins as vacuum seems odd.Vacuum , in my mind, is associated with pressure and not emptiness.How do people feel abouth this space described as vacuum?

    1. Christophe Verlinde says:

      You must be an engineer or an applied physics guy to associate the term vacuum with the method to create it namely by lowering the pressure. However, vacuum comes from the latin word for void or empty space. As such the use of the term vacuum by Klebe et al. goes back to the etymology of the word.

      1. Li Zhi says:

        Not to nit-pick but a vacuum is MEASURED by pressure. So, for those of us chemists who have bought into the whole “Science as an evidence based discipline” paradigm, we have no problem in thinking that the lack of pressure in a volume in which vapor could exist is a darn good way to think about vacuum.

        1. Cymantrene says:

          There is photon gas in the void, which has a pressure as well.

  9. provocateur says:

    Well, Klebe et al. does not use the word vacuum.

  10. Probing apparently de-wetted regions says:

    Such pockets undoubtedly exist… and indeed are paid for in advance by protein folding, but I would not describe them as a vacuum, even when calculated as such on the basis of electron density and hydration free energy simulations. The most common situation is encountered when somewhat disordered hydrophobic side-chains adopt multiple conformations at low density, continuously rearranging to fill the site, but each at an occupancy too low for a crystallographer to assign at ordinary resolutions. Then, there’s the choice of all sorts of small hydrophobic pieces present at low concentrations in the mix, that can help fill the space, at low and disordered density, that are more favorable than vapor-phase water. Experience suggests that it is valuable to target such regions, for potency (if not always drug-like properties) but that the boost you get is smaller than what you’d expect if it were a frank vacuum (or water vapor) present in the middle of a cluster of ordered residues; you also need to out-bind the gamut of small, reasonably hydrophobic and disordered species that may be present and bound at low affinity (e.g. N2 in vivo; many others in slow-grown crystals), and pay the entropy cost to substantially nail down the surrounding environment.

  11. Lingle Wang says:

    Computational model predicting the existence of such empty region in protein binding pockets and their importance to ligand binding affinity appear long time ago:

  12. Anon says:

    I remember sortilin having a big internal, pain-in-the-neck cavity. Never figured out what to put in that one.

  13. Henry Rzepa says:

    I would dearly love to play with the coordinates of thermolysin as reported in the paper. Whilst they have been deposited as PDB code 5M9W, this remains “unavailable” ( ) with a status of “AUTH: processed, waiting for author review and approval”. Perhaps this in turn will be embargoed for a period once approved.

    I do find it odd that such lack of fundamental data post-publication is still allowed.

    1. Gerhard Klebe says:

      Will be released on 2017-08-16, this is the next available release date from the pdb. it does not go faster, sorry.

      The author

Leave a Reply to Christophe Verlinde Cancel reply

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

Time limit is exhausted. Please reload CAPTCHA.