Monday, April 13, 2015

Hydrophobic or just solvophobic?

As I mentioned in the previous post, the notion of a dewetting transtion – in effect, capillary condensation driven by enhanced density fluctuations – that drives hydrophobic attraction has yet to be fully integrated with the question of whether this is a generic solvophobic effect or something specific to water’s hydrogen-bonded network. David Chandler’s picture of a dewetting transition occurring between extended hydrophobic surfaces for a lateral size scale of around 1 nm or more has tended to focus on the impossibility of maintaining the integrity of the H-bonded network in this geometry. But it may be that the density depletion and enhanced fluctuations on which this picture is predicated are more general features of solvophobicity. Rick Remsing and John Weeks at the University of Maryland speak to this question in a preprint that aims to dissect this hydrophobic interaction into components related to hydrogen bonding and to longer-ranged dispersion and electrostatic forces between the solvent molecules (http://www.arxiv.org/abs/1502.05220). Their conclusions are so nicely summarized in the paper that I can’t do better by paraphrasing them:
“We employ short ranged variants of the SPC/E water model to show that small scale solvation and association in water is governed by the energetics of the hydrogen bond network alone. However when the solute is large and the hydrogen bond network is broken at the hydrophobic interface, water behaves in a manner qualitatively similar to a simple fluid, with unbalanced LJ attractions dominating the solvation behavior.”

For example, without LJ attractions in the solvent, there is no dewetting-induced hydrophobic attraction of two fullerene molecules. (This implies that the crossover between “small” and “large” solutes lies somewhere between the sizes of methane and C60.) In other words, dewetting here is nothing other than regular (albeit barrier-less) capillary evaporation of a solvent, and not a “water effect” at all. Which, if it’s right, means that we might want to think about speaking of a “hydrophobic interaction” at small scales but a “solvophobic interaction” at large scales. But I’d like also to know how this fits with Ronen Zangi’s study indicating that there’s actually a repulsion between fullerenes in water, mentioned in an earlier post. In other words, how potential-dependent is all this?

They’ve been busy. In another contribution, Remsing and Weeks add another variant to the many efforts to develop hydrophobicity scales for biomolecules. This one is based on electrostatics, which has the advantage of being able to predict water-mediated hydrophilic interactions as well as hydrophobic ones (JPCB jp509903n; 2015 – paper here). They begin with a nice description of efforts so far, making the fundamental distinction between “surface-based” methods which aim to use the biomolecular surface properties alone, and “water-based” methods in which the effects of surface topography and neighbouring chemical functionality on the hydrogen-bond network of the local hydration sphere are taken into account. Their new method calls into the latter category, but is computationally inexpensive as it aims to characterize the long-wavelength collective electrostatic response of the water to the surface in question. Not only does this distinguish between hydrophilic and hydrophobic surfaces, but it accounts for different types of hydrophilic surfaces, e.g. those the polarize the water molecules in different orientations. This allows them to identify situations where the approach of two hydrophilic surfaces might induce a water-mediated interaction because of the commensurate polarization of the water network.

That water penentrates into carbon nanotubes, despite its hydrophobic nature, has been confirmed both in simulations and in experiments. In a preprint (http://arxiv.org/abs/1501.0608), Hemant Kumar and colleagues at the Indian Institute of Science in Bangalore ask whether this is driven by entropy or energy. Previous studies have given conflicting answers, but on the basis of MD simulations using the two-phase thermodynamic method Kumar et al. conclude that both energy (the carbon-oxygen LJ interaction) and entropy (for low occupancy, at least) support the filling process. This seems consistent with the findings of J. P. Huang at Fudan University in Shanghai and colleagues that water flows several times faster through non-straight (zigzag) carbon nanotubes than through straight ones, owing to the greater LJ interactions in kinked channels (T. Qiu et al., JPCB 119, 1496; 2015 – paper here).



When a solute particle is excited (electronically, vibrationally, rotationally), how does the (water) solvent relax to the perturbation? Rossend Rey at the Polytechnic University of Catalonia and James Hynes at Colorado have examined this question using linear response theory, and conclude that most of the absorbed energy is transferred to hindered rotations (librations) of the water molecules – and that mostly in the first hydration shell (JPCB jp5113922; 2015 – paper here).

The M2 proton channel of influenza virus A is the target of several flu drugs. These appear to bind to the channel and disrupt the proton transport, which seems to involve water clusters in the channel. By calculating the energetics of pore blockers at different sites in M2, Michael Klein at Temple University and coworkers offer insights into the mechanisms of drug action that might guide the identification of new types of inhibitor (E. Gianti et al., JPCB 119, 1173; 2015 – paper here). The general principle is to dehydrate the pore by replacing the water clusters with the ligand scaffold, and the results here show that this is most effectively done when the ligand scaffold mimics the water-cluster contour, while also preserving the interactions that the cluster made with the protein.

Another water-containing channel is explored by Ai Shinobu and Noam Agmon at the Hebrew University of Jerusalem, and I like their opening line: “Internal water molecules in proteins are conceivably part of the protein structure”. Quite so. They look at the lone water molecule in the “barrel” of the proton-conducting green fluorescent protein, using MD simulations to examine how water exchange occurs following photoexcitation, which opens up the channel transiently (JPCB 119, 3464; 2015 – paper here). The water molecule is shifted by the formation of a water wire through a temporary “hole in the barrel” connecting the chromophore with the bulk: a weak spot between strands of the β-barrel. This wire provides a route for protons to leak out of the channel, and the authors think that this might in fact supply the dominant mechanism for proton escape from the protein. The water motion, meanwhile, involves interactions with hydrogen-bonding residues that result first in sub- and then super-diffusive motion.

How conformationally stable are proteins when dehydrated for storage? One way to find out is to look at water adsorption isotherms as a function of humidity for different conformations. This is what Pablo Debenedetti and coworkers at Princeton have done for the Trp-cage mini-protein using simulations of different protein matrices: crystal, powder, and thermally denatured powder (S. B. Kim et al., JPCB 119, 1847; 2015 – paper here). All three matrices display so-called type II adsorption isotherms, in which there is hysteresis between adsorption and desorption across most of the humidity range. The isotherms are all of similar shape, showing little sensitivity to the degree of ordering in the proteins. Moreover, all show similar changes in swelling behaviour and hydrogen-bonding content as a function of humidity, except for the degree of intra-protein H-bonding, which unsurprisingly depended on the degree of folding in the monomers.

A great deal of attention is now being focused on proteins that have no great degree of conformational regularity in the first place: intrinsically disordered proteins, such as the tau protein which regulates microtubule formation in the nervous system. Dysfuntions in tau can lead to protein aggregation and fibril formation of the sort associated with neurodegenerative diseases such as Alzheimer’s. Martin Weik at Grenoble and coworkers have used neutron scattering and MD simulations to compare the dynamical coupling of protein and solvent through the protein dynamical transition (at 240 K) for the tau IDP and a representative globular protein, the maltose binding protein (G. Schirò et al., Nature Commun. 6, 6490; 2015 – paper here). There is a general notion that the dynamical transition corresponds with the onset of hydration-water translational motion on the protein surface. But although IDPs also show a dynamical transition, they have considerably more solvent-accessible surface than globular proteins, so it is by no means clear that one can expect the same kind of water-protein coupling to apply. But it seems that it does. Martin and colleagues find that in both cases the onset of water translational diffusion seems to coincide with that of large-amplitude protein conformational fluctuations, of the sort needed for functional behaviour. Thus this connection seems to be independent of the protein’s folding state.

What effect do osmolytes such as urea (a denaturant of globular proteins) have on IDPs? That question is explored by Zachary Levine and colleagues at UCSB using a combination of simulations and experiments (PNAS 112, 2758; 2015 – paper here). They find that both urea and trimethylamine N-oxide (TMAO) affect the structure of tau by shifting the distribution of existing conformations rather than by adding any new ones. The osmolytes do so by altering the balance between hydrogen-bonding and salt-bridge interactions in the individual IDPs. In doing so, urea suppresses aggregation, while TMAO promotes the formation of compact oligomers. The mechanism of the latter is subtle, stemming from changes in hydration of the IDP in the presence of TMAO in such a way as to promote aggregation entropically by releasing TMAO and water from the protein surfaces. These predictions of the simulations are borne out by experiments.

Urea can denature RNAs too. Alexander MacKerell at Maryland and colleagues have investigated why (K. Kasavajhala et al., JPCB 119, 3755; 2015 – paper here). The general idea has been that urea forms H-bonds and stacking interactions with the nucleotide bases. That’s a view that is supported by these ab initio calculations, which show that stacking via dispersion forces as well as H-bonding create cage-like complexes around the bases. For example, guanine may become surrounded by 5 urea molecules and 12 waters, pretty decisively trapping it in an unfolded conformation. Direct interactions, you see.

MacKerell also has an interesting paper with E. Prabhu Raman on the energetics of protein-ligand binding (JACS 137, 2608; 2015 – paper here). They have looked in particular at the roles of water in the binding site, and especially the classical view that the “hydrophobic effect” in ligand binding is due to the reduction of nonpolar solvent-exposed area at the binding interface and the concomitant release of more “highly structured” water from this location. To calculate changes in solvation energy as a ligand binds, they use something called Grid Inhomogeneous Solvation Theory (GIST), described by Nguyen et al. in J. Chem. Phys. 137, 044101 (2012). They calculate the various thermodynamic contributions to the binding energy for propane and methanol in several different binding pockets of the proteins Factor Xa and P38 MAP kinase. While they find that the entropy of reorganization of water in the binding pockets favours ligand binding, much as the traditional picture of an entropically driven hydrophobic effect would suggest, the picture is actually rather complex, with subtle interplay between direct protein-ligand interaction energies (even in nonpolar sites) and loss of water interaction energies that can sometimes compensate and lead to a rather small binding enthalpy. Sometimes the enthalpic and entropic changes can oppose (compensate) each other, sometimes they reinforce one another. The general message seems to be that, much as some earlier studies have shown, it is difficult to generalize about the respective contributions to the binding thermodynamics.

More on antifreeze proteins: Aatto Laaksonen and colleagues at Stockholm University show that representatives of the two major classes of AFPs (“hyperactive” and “moderately active”) have different effects on the nature of the ice-water interface when they are bound there (G. Todde et al., JPCB 119, 3407; 2015 – paper here). For the former class, they study the snow flea AFP; for the latter, the winter flounder AFP. Both AFPs increase the thickness of the interfacial region (defined as the region where water diffusion varies from 10 to 90% of the bulk liquid value), with the hyperactive AFP having the greatest effect (widening by 25-40%). This protein has ~25% more hydrophobic surface (the ice-binding side) than the wfAFP, but also 60-70% more hydrophilic surface; the authors think that it’s the first of these differences that counts the most.


The snow flea (a) and winter flounder (b) antifreeze proteins bound at the ice-water interface.

Ariel Fernandez has reported further evidence that the protein regions he calls dehydrons – parts of the backbone where hydrogen bonds in the backbone are “imperfectly wrapped” and thus solvent-exposed – may play not just a structural but also a chemical role (FEBS Letters 589, 967; 2015 – paper here). He has previously argued that water molecules in dehydron regions can act as proton acceptors because of the way that confinement prevents them from orienting their dipoles perfectly with the prevailing electric fields. Now, using quantum calculations, he expands on how this behaviour makes dehydrons activators of nucleophilic groups, and looks at some biochemical consequences, in particular for cancer-related mutations of certain kinases. The nucleophilicity induced by the dehydron, he says, turns the kinase constitutively (and hazardously) active for phosphorylation.