Friday, May 27, 2016

Dewetting and its origins

Several thermophilic proteins are known to have internal water-filled cavities, and some are known to denature when these cavities are emptied. Could this internal water be the secret to their thermal stability? Fabio Sterpone in Paris and his colleagues have investigated that question by using MD simulations to calculate the hydration free energy of buried water for several homologous mesophilic and thermophilic proteins (D. Chakraborty et al., JPCB 119, 12760; 2015 – paper here). Their findings support that main contention: the buried water contributes favourably and significantly to the stability of the thermophilic proteins. This could therefore offer a strategy for designing proteins robust against high temperatures.

Protein folding is studied with proton NMR by Francesco Mallamace, Gene Stanley and their collaborators (F. Mallamace et al., PNAS 113, 3159; 2016 – paper here). In particular, they aim to elucidate the role of water in the folding process, taking lysozyme as the paradigmatic example. They examine the evolution of hydrophilic (amide NH) and hydrophobic (methyl and methine CH) groups as folding proceeds, both for a reversible unfolded intermediate state and the irreversible denatured state. Hydrogen bonding between amide groups and internal water seems to play a crucial role: water acts as a kind of “glue” between buried amide and carbonyl groups, while the increased mobility of hydrophobic groups as they form clusters in the folded state compensates for the loss of configurational entropy that this entails.

Studying the binding and unbinding kinetics of cavity-ligand interactions could offer valuable insights into the efficacy of drugs and drug candidates. But it’s tough to do using MD simulations, because the timescales are often too long. Bruce Berne at Columbia and his colleagues have developed a “metadynamics” scheme that allows MD to probe timescales of not just many seconds but an hour or so (P. Tiwary et al., PNAS 112, 12015; 2015 – paper here). They examine a prototypical ligand–cavity system: a fullerene in a simple hydrophobic pocket, with explicit water. They find that binding involves an abrupt dewetting transition when the fullerene is constrained to move only along the central axis of symmetry of the system, but that dewetting is continuous, and binding is 20-fold shorter-lived (around 200s), when this steric constraint is removed.


Hydration of a fullerene in a simple binding cavity.

Abrupt dewetting has been observed between two planar hydrophobic surfaces at small separations, where it is simply capillary evaporation by another name. But Matej Kanduc and Roland Netz at the Free University of Berlin find that it can also occur between a hydrophobic and mildly hydrophilic surface, if they are both polar but electrically neutral (PNAS 112, 12338; 2015 – paper here). This leads to dry adhesion of the dissimilar surfaces. Kanduc and Netz construct the full phase diagram for any pair of surfaces, showing that dry adhesion can happen even for appreciable hydrophilicity of one surface (say, contact angle of 45 degrees) if the other is markedly hydrophobic (contact angle around 120 degrees).


The phase diagram of attractive/dewetting regimes for two planar surface of different contact angle.

Masataka Nagaoka at Nagoya University and colleagues look at another dewetting process, which takes place in the binding pocket of thrombin when it binds a ligand (I. Kurisaki et al., JPCB 119, 15807; 2015 – paper here). The process is quite complicated. These MD simulations indicate that, following the establishment of a hydrogen-bonding interaction between the substrate and an Asp group in the pocket, the water is gradually removed from the pocket, going not into the bulk phase but into a water channel within the protein – which thus turns out to have a functional role.


The water channel of thrombin.

The importance of density fluctuations at the interface in nanoscale (as opposed to atomic-scale) hydrophobic interactions driven by drying transitions has been asserted for some time now by David Chandler and his colleagues. Phillip Geissler and Suriyanarayanan Vaikuntanathan recently developed this idea by showing that the Lum-Chandler-Weeks drying-induced hydrophobic attraction needs to pay close attention to capillary waves (Phys. Rev. Lett. 112, 020603; 2014). Now they and their coworkers have put this theory to the test in a coarse-grained lattice model (S. Vaikuntanathan et al., PNAS 113, E2224; 2016 – paper here). Despite the relative simplicity of the model, it seems to capture very efficiently the nanoscale density fluctuations and their relevance for hydrophobic hydration in the case of a hydrophobic solute particle at the air-water interface and a pair of nanoscale hydrophobic plates. In both cases the “minimalistic” model agrees very well with detailed atomistic simulations, suggesting that it succeeds in modeling the basic physics of these situations. What seems striking here, I think, is how this long-problematic issue of hydrophobic hydration seems to be turning back towards the “old-style” basic physics of wetting and inhomogeneous fluids (compare, for example, R. Evans & M. C. Stewart, J. Phys. Condens. Matt. 27, 194111; 2015; R. Evans & N. B. Wilding, PRL 115, 016103; 2015).

Another coarse-grained description of hydration is offered by Bill Jorgensen at Yale and colleagues (X. C. Yan et al., JPCB jpcb.6b00399; 2016 – paper here). They use techniques I’m frankly not familiar with to couple a coarse-grained approach with all-atom solute models, and find that the resulting multiscale simulations compare acceptably with the all-atom case while achieving a 7-30-fold reduction in computational cost.

And on the same track, Kenichiro Koga at Okayama and colleagues report a mean-field approximation for inhomogeneous liquids to study hydrophobic hydration, specifically the solvation of methane at the air-water interface (K. Abe et al., JPCB 120, 2012; 2016 – paper here). The approach here is to consider two steps in the solvation process: creation of the cavity for a hydrophobic solute, and insertion of the solute itself.

Ions modify hydrophobic interactions in an ion-specific manner, familiar as Hofmeister effects. But to what extent do these influences depend on the arrangements and mobility of the ions? Izabela Szlufarska at the University of Wisconsin and colleagues examine that question using MD simulations of hydrophobic interactions between a nonpolar surface and nonpolar or amphiphilic nanorods mimicking β-peptides (K. Huang et al., JPCB 119, 13152; 2015 – paper here). They compare the cases of free ions in solution, and ionic groups tethered to the rods both randomly and in organized spatial arrangements (such as a row down one side). Free ions generally strengthen hydrophobic interactions according to the conventional ion-specific series, in a manner that is correlated with (and assumed to be due to) fluctuations in water density close to the surface. But these interactions can be switched on or off by different arrangements of the immobilized ions, and the ion-specific effects now don’t follow the usual Hofmeister ranking. The authors analyse the findings in a dynamical picture, concluding that “Our analysis of the structure and dynamics of water near the hydrophobic nanorod shows that dynamics is a better indicator of the specific ion effect than the static water structure.” Is this perhaps an emerging consensus for such situations?


The nanorods studied by Huang et al.: cyan groups are nonpolar, yellow are immobilized ions.

Hydration water around intrinsically disordered proteins is more mobile than that around globular proteins, according to simulations by Pooja Rani and Parbati Biswas at the University of Delhi (JPCB 119, 13262; 2015 – paper here). They find that both the translational and rotational diffusion are relatively greater for IDPs. In an earlier study, they found that the residence times of hydration waters for IDPs are relatively long (JPCB 119, 10858; 2015) – but that work doesn’t contradict this, they argue, because the enhanced mobility doesn’t actually help the water molecules leave the hydration layer.

Water wires are essential for transporting protons through the pores of the chloride/proton exchanger ClC-ec1, a member of the CLC superfamily of proteins that exchange protons for halides and other anions across cell membranes. That process is elucidated in detail in simulations by Emad Tajkhorshid at UIUC and colleagues (T. Jiang et al., JACS 138, 3066; 2016 – paper here). They examine how the water wires formed in the presence of chloride bound in the central anion binding site differ from those in the presence of fluoride, nitrate and thiocyanate, giving the channel its anion specificity. For fluoride and nitrate there are only “pseudo water wires” separated by the intervening anion, which can’t sustain proton transport; for thiocyanate there are none at all.


Water wires and their disruption/absence for various anions in the channel of the chloride/proton exchanger ClC-ec1.

Is the air-water interface acidic or basic? In other words, do hydrated excess protons or hydroxide ions have an affinity for the water surface? Simulations and experiments have sometimes seemed to give conflicting results, making this a point of controversy. Greg Voth, now at the University of Chicago, and his coworkers reported several years ago that simulations showed protons having an affinity for the interface. They now confirm this finding using state-of-the-art reactive molecular dynamics (Y.-L. S. Tse et al., JACS 137, 12610; 2015 – paper here). The (weak) preference of the hydrated proton of the water surface, they say, is enthalpically favoured and reduces the loss of hydrogen-bonding there. The hydroxide ion, meanwhile, is repelled from the interface, also for enthalpic reasons.

Small peptides don’t sample all of the conformational space theoretically available to them. Brigita Urbanc and colleagues at Drexel University in Philadelphia wonder if hydration structure has something to do with this (D. Meral et al., JPCB 119, 13237; 2015 – paper here). Using MD simulations, they show that the so-called polyproline II-like (pPII) and β-strand conformations that are prominent for GXG peptides (with X a “guest” residue) have different hydration structures, the former being enthalpically stabilized and the latter entropically. In general the nature of backbone hydration in the pPII conformations is clathrate-like, and differs for different X depending on the residue’s ability to template this structure.

It’s looking ever more as though water has an important, even pivotal role in the assembly of the β-amyloid aggregates implicated in neurodegenerative diseases. Nadine Schwierz at UC Berkeley and coworkers report that it drives the growth of these fibrils (N. Schwierz et al., JACS 138, 527; 2016 – paper here). Their simulations suggest that solvent entropy is the main driving force, and that assembly involves collective water motions. As two β strands lock together, motion of the intervening water is significantly retarded, and there is an entropic gain when this water is removed to create a dry binding interface.

The hydrophobic interaction is still challenging to understand intuitively. That’s emphasized by a report by Lawrence Pratt at Tulane University and colleagues showing that, for argon-argon interactions in water, the “hydrophobic bond” is actually weakened by including solute attractive interactions (M. I. Chaudhari et al., JPCB 120, 1864; 2016 – paper here). They treat the situation using local mean field theory, and I think (but am not sure) that ultimately the reason for this result is that in effect the solute attractive interactions benefit solvation more than they do aggregation.

Might water networks in a protein binding site be amenable to a thermodynamic treatment that considers them a kind of loose, composite ligand? That seems to be in essence what Gregory Ross at the University of Southampton and colleagues are attempting to enable (G. A. Ross et al., JACS 137, 14930; 2015 – paper here). They have modified the grand canonical MC simulation technique to allow efficient calculation of the binding energy of an entire hydrogen-bonded water network, as well as to estimate the individual water-molecule affinities and the degree of cooperativity between them. They propose that the method might allow the incorporation of water networks into rational drug design.

How to handle long-ranged forces has, as John Weeks at Maryland and colleagues put it in a paper in PNAS (R. C. Remsing et al., PNAS 113, 2819; 2016 – paper here), “plagued theory and simulation alike”. They offer a solution: a way to incorporate long-ranged interactions that uses only short-ranged potentials. They demonstrate it for the case of hydration of ionic and hydrophobic species. It relies on local molecular field theory, in which the interactions are accommodated by finding a simpler, single-particle “mimic system” that can furnish an effective field that takes care of averaged “far-field” interactions in the full system under study.

Different types of antifreeze proteins have different mechanisms of action, according to Ilja Voets at the Eindhoven University of Technology and colleagues (L. L. C. Olijve et al., PNAS 113, 3740; 2016 – paper here). They have measured both the thermal hysteresis and the ice recrystallization inhibition activity for a range of AFPs, and find no correlation between the two. Both of these properties have been considered to be features of antifreeze activity, which the researchers attribute to binding to different ice planes. They conclude that antifreeze activity is many-factored, and that applications and rational design of AFPs “requires a tailored optimization to the specific purpose.”

More debate on the putative liquid-liquid transition in metastable water: Gyan Johari and José Teixeira argue that the thermodynamic arguments linking this transition to that between low- and high-density amorphous ice don’t stand up (JPCB 119, 14210; 2015 – paper here). They conclude that “The available data show that HDA is not a glass, and the presumed HDL and LDL are not normal liquids. If that is accepted, the HDL−LDL fluctuations view, the two-liquid model, and the virtual liquid−liquid phase transition would all be consequences of a false premise.” I look forward to the next round…

Two quite different forms of metastable (quasi-)liquid water are reported by Gen Sazaki and colleagues at Hokkaido University. They have previously reported such phases on the surface of ice (Sazaki et al., PNAS 109, 1052; 2012); one of the liquid phases wets the ice, the other doesn’t. Using advanced optical microscopy, they now show that these surface phases are kinetically, not thermodynamically, stable, and that they form not by the surface melting of ice but by the deposition of supersaturated water vapour on the ice surface (H. Asakawa et al., PNAS pnas.1521607113; 2016 – paper here).

Julio Fernández continues his studies of “frustrated” water at the protein interface (dehydrons, which are topologically deprived of hydrogen-bonding opportunities) with a study of the acid-base chemistry of such water, using quantum-chemical calculations (FEBS Lett. 590, 215; 2016 – paper here). He reports that water surrounding deydrons is proton-accepting, and enables a mechanism for directed Grotthuss transport of protons

Tryptophan fluorescence has been widely used to study protein dynamics and hydration. Feng Gai and colleagues at the University of Pennsylvania report that a tryptophan analogue, 5-cyanotryptophan can be used in the same way but with greater sensitivity (B. N. Markiewicz et al., JPCB 120, 936; 2016 – paper here). They show that, among other things, the new probe molecule allows them to distinguish two differently hydrated environments in a folded protein (Trp-cage).

Whether or not the result reported by Ruth Livingstone and colleagues at the MPI Mainz is relevant to water in biology isn’t clear, but it’s an interesting finding all the same. They use vibrational spectroscopy to look at water interacting with a monolayer of the detergent sodium dodecyl sulfate at the water surface, and see two distinct types of water present (R. Livingstone et al., JACS 137, 14912; 2015 – paper here). Close to the anionic head groups the waters have localized O-H stretches, but further below the monolayer there is a delocalized O-H mode. These two modes are coupled and can exchange energy. The question is obviously whether the same applies for lipid assemblies in vivo.