Thursday, December 12, 2013

Enthalpy/entropy compensation

The weak forces that govern protein-protein association and aggregation are of course intimately connected with hydration, partly for example via hydrophobic interactions but also through the electrostatic consequences of water removal at the interfaces. These effects are examined in a continuum solvent model by Sergio Hassan at the NIH in Bethesda and colleagues (A, Cardone et al., J. Phys. Chem. B 117, 12360; 2013 – paper here). They use the model to examine the barnase-barstar complex, and say that it could be extended to multi-protein systems.

A rather exquisite water-mediated mechanism of enzymatic activity is reported by Qiang Cui at Wisconsin-Madison and colleagues (P. Goyal et al., PNAS 110, 18886 – paper here). They describe a computational study of cytochrome c oxidase, which uses oxygen reduction to pump protons across a membrane. During this process a glutamate residue is believed to act as a temporary proton donor. Cui and colleagues show that the proton affinity of this Glu is controlled by the degree of hydration in an internal hydrophobic cavity, which is itself governed by protonation of a substituent on the heme group 10Å away, triggering movement of a loop gating the cavity’s entrance.

Enthalpy/entropy compensation in protein-ligand binding is the topic of a nice study by George Whitesides’ group in a special issue of the European Physical Journal (doi:10.1140/epjst/e2013-01818-y – paper here). The group examines this issue further in a paper in JACS (B. Breiten et al., JACS 135, 15579; 2013 – paper here), where they look at the human carbonic anhydrase system that they have also studied earlier (see P. W. Snyder et al., PNAS 108, 17889; 2011). This new paper expands on that earlier work, in which a series of ligands modified with various thiazole-based sulfonamides was used to probe the effects of solvent rearrangements within the binding cavity. They find that the changes in the free energy of binding, and the contributions from enthalpy and entropy, are largely determined by the rearrangements or displacements of water, and voice a telling conclusion: “This water- centric view of ligand binding – and H/S-compensation – cannot be rationalized by the lock-and-key principle and suggests that the molecules of water surrounding the ligand and filling the active site of a protein are as important as the structure of the ligand and the surface of the active site.”

Myoglobin binds various ligands, such as O2, CO and NO, in an internal cavity. It also binds water in internal sites, but crystallographic studies have given conflicting views on where these cavities are and how occupied they are. Shuji Kaieda and Bertil Halle at Lund have used deuteron and 17O magnetic relaxation dispersion spectroscopy to probe these hydration sites and the dynamics of the water molecules occupying them (J. Phys. Chem. B 117, 14676; 2013 – paper here). They find that the waters come and go on a microsecond timescale in all four sites, despite their significant separation, and conclude that these dynamics share a global mechanism involving transient penetration of the protein by hydrogen-bonded chains that may intermittently ‘flush’ the cavity network.

More on small-molecule solute perturbations to protein structure: Warren Beck and colleagues at Michigan State University report fluorescence spectroscopic measurements on zinc-substituted cytochrome c in the presence of guanidinium ions, in an effort to discover why Gdm+ perturbs the protein dynamics by making it apparently more flexible (J. Tripathy et al., J. Phys. Chem. B 117, 14589; 2013 – paper here). They conclude that part of this response comes from a change in structure of the protein’s hydration shell because of direct binding of Gdm+ to the protein surface.

How do magnesium ion channels attain their high specificity in the presence of high concentrations of calcium? Todor Dudev at the Academia Sinica and Carmay Lim at National Tsing Hua University in Taiwan say that the differences in the metals’ hydration shells are the key (JACS 135, 17200; 2013 – paper here). Their calculations suggest that the hexacoordinated magnesium ions polarize the bound water more strongly, resulting in stronger ion-water-protein interactions.

How water diffuses at the surface of a lipid membrane has been studied before, but it can be hard to ensure that just the water molecules immediately at the interface are probed. Robert Bryant at the University of Virginia Charlottesville and colleagues claim to be able to do that using magnetic relaxation dispersion spectroscopy for phospholipid vesicles in deuterated water (K. G. Victor et al., J. Phys. Chem. B 117, 12475; 2013 – paper here). They report that the water molecules explore the interface via essentially two-dimensional diffusion over the vesicle surface, with a translational correlation time that seems relatively long: about 70 ps. This modified dimensionality of water motions, they say, stems from an excluded-volume effect.

The molecular basis of anaesthesia is still unclear, although it seems to involve low-affinity binding of the anaesthetic to proteins. Hai-Jing Wang of UNC and colleagues say that hydration water plays a crucial role in this process (H.-J. Wang et al., J. Phys. Chem. B 117, 12007; 2013 – paper here). They use NMR to look at the binding of volatile halogenated alkanes (known anaesthetics) to BSA, which has binding pockets for such molecules. They find that binding occurs only once the hydration level of the protein surpasses a critical threshold. The molecular details of what this hydration water does remain to be elucidated, but Wang et al. suppose that the threshold hydration level is needed to “establish a favourable free-energy landscape for the binding of anaesthetics to the pre-existing binding sites.”

Peter Rossky and Lauren Kapcha at Texas at Austin have introduced a new hydrophobicity scale for protein constituents that works at the level of individual polar and non-polar atoms rather than classifying each residue as a whole (J. Mol. Biol. 10.1016/j.jmb.2013.09.039 – paper here). They say that it gives an appropriate measure of hydrophobicity in cases where a residue-based approach fails – without being any more computationally intense – and that it shows that this atomistic level of detail is therefore sometimes indispensable.

Another computational challenge is addressed by Sergio Pantano of the Institut Pasteur de Montevideo in Uruguay and coworkers (H. C. Gonzalez et al., J. Phys. Chem. B 117, 14438; 2013 – paper here). They describe a way to combine the common SPC, TIP3P and SPC/E atomistic water models with a coarse-grained model called WatFour or WT4, demonstrating its effectiveness via (among other things) simulations of the β1 domain of streptococcal protein G.

There’s a curious and interesting paper by Akira Yamakata of the Toyota Technical Institute in Nagoya and colleagues on the observation of hydration shells of ions being destroyed by electrochemical control of the ions’ binding to an electrode surface (A. Yamakata et al., JACS 135, 15033; 2013 – paper here). The researchers use a platinum electrode rendered hydrophobic by coverage with a CO monolayer, and use time-resolved IR spectroscopy to monitor the hydration shells of tetrapropylammonium (Pr4N) and sodium ions as they are pulled onto the surface by the electric field. They can monitor the destruction of the hydration shells as this happens, and say that while that of Pr4N is rather easily destroyed at high field so that the ion interacts directly with the CO layer, the shell of sodium is more rigid and remains intact.

Lawrence Pratt and colleagues at Tulane University use simulations to demonstrate that the hydrophobic interactions between hard-sphere solutes in water are more attractive and endothermic than is predicted by the molecular-scale Pratt-Chandler theory (M. I. Chaudhari et al., PNAS 10.1073/pnas.1312458110 – paper here). This shows what an improved theory will have to shoot at.

Peter Hamm and coworkers at the University of Zurich have used 2D Raman-THz spectroscopy to probe collective intermolecular modes of pure liquid water, and conclude that there are different types of hydrogen-bonded networks – albeit only on a very short (100-fs) timescale, which is not compatible with controversial proposals for persistent heterogeneity of these networks like that of Anders Nilssen et al. (J. Savolainen et al., PNAS 10.1073/pnas.1317459110 – paper here). This seems to add to the growing view that this apparent controversy is best seen as a question of the different timescales being examined, and that at room temperature any ‘two-state’ picture is very rapidly averaged away.

Finally, a curiosity about the macroscopic mechanical roles of water in plants. Peter Fratzl of the Max Planck Institute of Colloids and Interfaces in Potsdam and colleagues look at how changes in hydration can produce stresses and movement in plants via swelling and shrinkage (L. Bertinetti et al., Phys. Rev. Lett. 111, 238001; 2013 – paper here). The free energy needed for such actuation can be considered to be provided, at least in part, by the formation of new hydrogen bonds. The authors propose a model that allows them to estimate the free energy made available by water absorption into woody plant tissue, and find that (except in the case of nearly dry tissue) it amounts to about 1.2 kT per water molecule, suggesting that the water bound to the macromolecules in the tissue acquires about one additional hydrogen bond for every eight molecules. Frankly, I don’t fully understand the arguments here: the authors say that “This would suggest that the main driving force for water absorption in non living plant tissues is a phase transition of water to a liquid state, characterized by a stronger H-bond network, occurring when H2O is confined within the macromolecular components of the wood cell wall material.” I’m going to have to get back to you on that.

Thursday, November 14, 2013

Using water for drug design

Alfonso García-Sosa at the University of Tartu in Estonia has published a paper getting to grips with precisely the question that I have long wanted to see addressed: how to employ water molecules in design strategies for drug binding (J. Chem. Inform. Model. 53, 1388; 2013 – paper here). He has analysed over 2,000 crystal structures of hydrated and non-hydrated ligand-receptor complexes (including many drugs), and finds that bridging water molecules are an effective strategy for tight binding. He concludes that such a binding mechanism could be beneficially targeted, and that “if a tightly bound, bridging water molecule is observed in the binding site, attempts to replace it [in a designed drug that binds competitively] should only be made if the subsequent ligand modification would improve also its ligand efficiency, enthalpy, specificity, and pharmacokinetic properties.” This is just the kind of large-scale study that is needed to extract the advantages or otherwise, and the effective modes, of using water molecules in ligand design.

It seems increasingly likely that the mechanism of hydrophobic association, once considered in terms of a static picture involving expulsion of ‘bound’ water, is in fact associated with dynamical effects, in particular the fluctuations in the hydrogen-bonded network around the hydrophobes. That view is emphatically supported in a paper by Aljaz Godec and Franci Merzel of the National Institute of Chemistry in Ljubljana, working with Jeremy Smith at Oak Ridge (Phys. Rev. Lett. 111, 127801; 2013 – paper here). Their MC simulations of two hydrophobic particles (of about the size of a methane molecule) coming together shows that crossing over the desolvation barrier to the associated state involves a large collective fluctuation in hydration water in which the intervening hydrogen-bonded clusters are mostly displaced from the inner to the outer hydration shell. As the authors conclude, “a complete description of hydrophobic association can be obtained only by explicitly considering collective fluctuations involving many-body correlations between water molecules.”

The observation by Amish Patel, Shekhar Garde and their colleagues that hydrophobic association is faster near a hydrophobic surface, which I have mentioned before in this blog, is fleshed out in detail by these researchers in a new paper (S. Vembanur et al., J. Phys. Chem. B 117, 10261; 2013 – paper here). The key issue is that the interfacial water, like that at the air-liquid interface, is more easily displaced, lowering the desolvation barrier to association. Naturally, this means that hydrophobic surfaces might be used to catalyse aggregation processes. I wonder if it might also help to explain Barry Sharpless’s observations some years back of an increase in organic reaction rates at the air-water interface?

Water diffusion close to lipid membranes is not like that close to solid surfaces, Roland Netz at the Free University of Berlin and colleagues say (Y. von Hansen et al., Phys. Rev. Lett. 111, 118103; 2013 – paper here). Their MD simulations reveal that, whereas at solid surfaces water diffuses faster laterally than perpendicularly, at lipid membranes the opposite is true. This seems to be because the lipids present a rough surfaces, with some head groups protruding from the leaflets, so that lateral diffusion takes place amidst a rough free-energy landscape and is correspondingly hindered. As a result, lateral motion in the fluid phase at the membrane surface involves a distinct perpendicular component to the trajectories: water molecules wander in local energy traps before moving up and away an then “descending” elsewhere.



The dynamics of interfacial water at lipid membranes is also probed by Kevin Kubarych and colleagues at the University of Michigan using IR spectroscopy (D. G. Osborne et al., J. Phys. Chem. B jp4049428 – paper here). They infer this from the spectral diffusion of a labelled cholesterol derivative in the membrane, which reflects the dynamics of its hydration water, and find that these dynamics are three times slower than those of the same molecule in bulk water. The paper largely establishes this as a nice tool that might be used to look at variations in water dynamics at different locations in a membrane.

One explanation for the apparent long-ranged hydrophobic interaction measured between adsorbed monolayers in the surface force apparatus invokes electrostatic interactions between the restructured monolayers themselves (see, for example, Jacob Klein’s work: Phys. Rev. Lett. 96, 038301; 2006 and 109, 168305; 2012). Max Berkowitz and colleagues at UNC investigate this idea using a lattice model to look at how a surfactant coated charged (mica) surface might become reconfigured into heterogeneous structures (C. Eun et al., J. Phys. Chem. B jp405979n – paper here). They find that indeed an initially uniform surfactant monolayer can become transformed into a non-uniform surface covered with patches of surfactant bilayer separated by patches of bare (water-covered) mica. This simple model system might help to understand not just Klein’s studies but also the somewhat related issue of formation of lipid rafts in mixed-lipid membranes.

Another ongoing discussion is the mechanism by which osmolytes stabilize proteins. Piotr Bruzdziak and colleagues at Gdansk University of Technology in Poland examine that issue by using IR spectroscopy and DSC to study the effects of various osmolyes (TMAO, Gly, NMG, DMG) on the stability of lysozyme (P. Bruzdziak et al., J. Phys. Chem. B 117, 11502; 2013 – paper here). In contrast to many recent simulation studies of this problem, the authors here conclude that the osmolyte-protein interaction is indirect, the osmolytes “enhancing” water structure with a consequent “tightening” of protein folding.

OK, why not touch on another controversial topic too: the molecular basis for the dormancy of bacterial spores, and in particular whether this involves an ‘altered state’ of water. That’s addressed in a preprint (arxiv.org/abs/1309.5033) by Bertil Halle at Lund and colleagues, using deuteron magnetic relaxation dispersion to look at the state of water in spores of B. subtilis. Two models for the structure of the core aqueous phase have been proposed: a gel, in which mobile water permeates a macromolecular network, and a glass, in which everything inside the cells (including water) forms a solid amorphous phase. The NMR results clearly support the former picture – the water remains mobile, albeit with rotational motion about 15 times slower than the bulk.

Calcium release-activated calcium (CRAC) channels play an important role in cell signalling, and their dysfunction is associated with cardiac arrythmia and immunodeficiency problems. Thanks to a recent crystal structure of the pore region of a CRAC channel, Michael Klein and colleagues at Temple University have been able to investigate the mechanism of ion permeation through simulations (H. Dong et al., PNAS 110, 17332; 2013 – paper here). They say that a central hydrophobic region of the channel is the crucial region for switching on and off, and that it is the hydration of this region that controls ion transport: a small change in the number and orientation of water molecules here significantly alters the local electrostatic field. In other words, this is another channel that is water-regulated.

Moving onto another channel: the high water permeability of aquaporins may be a result of an optimization of shape at the pore entrance to minimize viscous dissipation, according to a study by Laurent Joly of the University of Lyon 1 and colleagues (S. Gravelle et al., PNAS 110, 16367; 2013 – paper here). Their finite-element calculations suggest that the hour-glass profile of the pore is particularly efficient at reducing dissipation, and that the optimal opening angle of 5-20 degrees is in the range of those observed in the proteins.

Aquaporins have of course a well resolved crystal structure, but many membrane proteins are intrinsically disordered, especially in the regions that extend beyond the lipid membrane and into the extracellular matrix. Songi Han of UCSB and colleagues suggest that hydration dynamics – in their case as measured by Overhauser dynamic nuclear polarization NMR – can provide a proxy for at least locating these extended protein segments (C.-Y. Cheng et al., PNAS 110, 16838; 2013 – paper here). The technique relies on the apparent existence of a gradient of diffusion dynamics up to 3 nm away from the membrane surface (really?). The authors can seemingly also use the technique to look at protein structure parallel to the membrane surface, as in the case of an amyloid-forming protein called alpha-synuclein associated with Parkinson’s disease. They report a sinusoidal variation in water retardation close to the lipid surface which they interpret as the result of the alpha-S coiled into an alpha helix embedded laterally in the membrane with its axis about 1-3 Å below the phosphate head groups, while the C-terminus end floats freely and disordered above the membrane surface.



Water dynamics generally found to be retarded in the vicinity of proteins. But NMR measurements on crystals of the protein Crh (protein catabolite repression Hpr) have indicated that water in the crystals may remain mobile down to at least -30 C – in other words, it seems more mobile than bulk water at these temperatures (Böckmann et al., J. Biomol. NMR 45, 319; 2009). How can this be? To answer that question, Anja Böckmann of the University of Lyon, who did the NMR wirk, has collaborated with Wilfred van Gunsteren at ETH Zurich and colleagues to conduct MD simulations of the crystalline form of Crh (D. Wang et al., J. Phys. Chem. B 117, 11433; 2013 – paper here). They find that indeed while retarded dynamics of hydration water are seen at 291 K, at 200 K (close to the glass transition) the rotational and translational mobility of this water is greater than in bulk. But while introducing artificial perturbations such as rigidifying the protein or switching off protein-solvent electrostatics does induce some enhanced mobility of either translation or rotation, the cause of the overall water mobility at this temperature remains obscure.

While on the subject of glassy low-temperature behaviour, Thomas Loerting of the University of Innsbruck and colleagues have reported a second glass transition in amorphous ice at 116 K, which they say is related to the coexistence of two highly viscous metastable liquid-like phases: a low-density phase with a glass transition of 136 K and a high-density phase which undergoes this second glass transition at 116 K (K. Amann-Winkel et al., PNAS 110, 17720; 2013 – paper here). Thus, they say, the putative HDL and LDL “can both be observed experimentally” – although the results don’t directly address the question of whether the two can be interconverted by a first-order liquid-liquid transition ending at a critical point.

A stark and rather remarkable reminder of how much still remains to be understood about the hydrogen bond is provided by Michele Ceriotti at Oxford and colleagues (PNAS 110, 15591; 2013 – paper here). Their ab initio simulations that incorporate nuclear quantum effects show that the protons in water’s hydrogen bonds undergo fluctuations that take them towards the oxygen acceptor atoms for significant fractions of time, sometimes leading to spontaneous proteolysis. What’s more, these events are strongly correlated among neighbouring hydrogen bonds, so that perturbations to the hydrogen-bonded network seem likely to modulate this distinctly non-classical effect in significant ways.

The discussion continues of how ions alter the dynamics of water in their hydration shells. MD simulations by Ana Vila Verde and Reinhard Lipowsky at the MPI for Colloids and Interfaces in Potsdam look at the effect of ion pairs (here magnesium sulphate and caesium chloride) to see to what extent the slowdown of water dynamics in cooperative between the anion and cation (J. Phys. Chem. B 117, 10556; 2013 – paper here). They find that the cooperative slowdown can be intense (especially for ions of high charge density), but only in the first hydration shell. I note their comment “our results do not support the notion that the Hofmeister series is due primarily to long-range effects of ions on water properties. Instead, they point to the possibility that models of ion solvation may focus primarily on the first hydration layer without excessive loss of accuracy.”

Directional transport of water through a carbon nanotube connecting two reservoirs, driven by electric fields, has been reported in simulations by several teams over the past few years. Hangjun Wu of Zhejiang Normal University in Jinhua and colleagues now add another instance (X. Zhou et al., J. Phys. Chem. B 117, 11681; 2013 – paper here). They show that a vibrating charge that deforms the nanotube away from its halfway point (thus introducing the asymmetry that underpins the ratchet-like effect) will create a directional flux of water. But the direction of that flux is dependent on the displacement of the nanotube wall, and can even change sign when the displacement gets very large.

Daryl Eggers at San José State University continues to develop his picture of solvation water as a reactant in solution equilibria. With Brian Castellano he considers what happens when one includes in the thermodynamic treatment of binding equilibria a representation of how water is perturbed next to solutes (so that it is not ‘equivalent’ on both sides of the equilibrium) (B. M. Castellano & D. K. Eggers, J. Phys. Chem. B 117, 8180; 2013) – paper here). The spirit of the exercise is nicely encapsulated thus: “When water is viewed as a coreactant, it becomes apparent that the traditional equation for the standard-state Gibbs free energy of binding represents an unbalanced equation.” In fact, this means that the equilibria and binding constants become concentration-dependent. Perhaps this might account for discrepancies between different ways of experimentally measuring the enthalpy changes of some reactions.

To what extent can the vibrations of liquid water be decomposed into essentially single-molecule bending and stretching OH modes, plus intermolecular contributions? Less than was thought, according to Andrei Tokmakoff, currently at the University of Chicago, and colleagues (K. Ramasesha et al., Nat. Chem. 5, 935; 2013 – paper here). Their studies using ultrafast 2D IR spectroscopy show that there is strong mixing between all of these modes, so that all the vibrations have some collective component. This will require some significant rethinking of relaxation and energy dissipation in water.

Thursday, October 10, 2013

In praise of the Nobels

All the three recipients of this year’s Chemistry Nobel Prize – Martin Karplus, Arieh Warshel and Michael Levitt – have of course played some part in the general exploration of water’s roles in molecular biology. But Levitt in particular has been inspirational. It was what he wrote with Mark Gerstein in Scientific American in 1998 that, more than anything else, persuaded me there was a story to be unfolded about the importance of water in biology:
“When scientists publish models of biological molecules in journals, they usually draw their models in bright colors and place them against a plain, black background. We now know that the background in which these molecules exist – water – is just as important as they are.”

Much of the popular discussion of the Nobel award has centred around the technical computational aspects of these guys’ work, almost as though they deserve praise as programmers. But it is comments like this one that reveal the deep chemical insight behind all that computer stuff. This why I view the decision with deep satisfaction.

Monday, October 7, 2013

Water and protein folding

Does water drive protein folding? That’s the title of a paper by Yutaka Maruyama and Yuichi Harano at Osaka University (Chem. Phys. Lett. 581, 85; 2013 – paper here), and as you’d expect from a title like that, they conclude that the answer is at least partly affirmative. They combine MD simulations and 3D-RISM solvation theory to calculate thermodynamic quantities involved in protein folding, and conclude that while hydration energy favours the denatured state (but is largely offset by the intramolecular interactions of the protein on folding), hydration entropy favours the folded state. In this sense, water contributes significantly to the stability of the collapsed conformation.

A perhaps more dramatic claim along these lines is made by Ariel Fernandez, now at the Argentinian Institute of Mathematics in Buenos Aires (J. Chem. Phys. 139, 085101; 2013 – paper here). He presents a model of electrostatic interactions between the protein surface and water dipoles, which suggests a principle that Ariel calls minimal episteric distortion: the protein-water interface adopts a configuration that minimizes the energetic cost of disrupting dipole interactions of the hydrogen-bonded matrix. In other words, the native fold is the one that corresponds to this minimally perturbing topology: the protein structure is the one that is least at odds with the structure of the surrounding water. As Ariel has put it, “The paper shows that the full interfacial free energy can be computed/interpreted in two different but equivalent ways: a) as elastic energy arising from the perturbation of the hydrogen-bond matrix of water, or b) as electrostatic energy stored in the anomalous polarization of water (interactions between dipoles forced not to behave in Debye manner and protein charges).” And crucially, if this interfacial free energy is omitted from the folding computations, an incorrect fold results because the structure is then “at odds with the solvent”.

John Weeks at the University of Maryland was one of the architects of the Lum-Chandler-Weeks hypothesis (J. Phys. Chem. B 103, 4570; 1999) that the hydration of small and large hydrophobes is qualitatively different, with a crossover scale of around 1 nm, and that the latter is dominated by interfacial free energies and the breaking of hydrogen bonds. (This was itself an extension of a proposal by Frank Stillinger in 1973.) Now Weeks, together with Richard Remsing at Maryland, has attempted to refine this idea by considering what roles van der Walls dispersion forces and electrostatic interactions might have in this business. Using MD simulations with the SPC/E water model, they confirm the original idea that solvation and association of small hydrophobes is dominated by hydrogen-bonding, but find that for large hydrophobes (where hydrogen bonds are broken at the interface) the attractions are dominated by Lennard-Jones dispersion forces, with water now behaving more like an ordinary liquid (J. Phys. Chem. B jp4053067 – paper here). The crossover length scale is set by the hydrogen-bond network itself, occurring when the solute is too large for this network to remain intact. This paper seems to offer an important extension of LCW/Stillinger, while showing that the basic ideas expressed in those earlier works are sound.

Further support for this picture comes from Guillaume Jeanmairet and colleagues at the École Normale Supérieure in Paris. They have recently presented a now molecular density-functional theory of water (J. Phys. Chem. Lett. 4, 619; 2013). In a preprint, they have now applied it to the hydration of hydrophobes of various sizes (G. Jeanmairet et al., arxiv.org/abs/1307.7237). They find that the original theory doesn’t work so well, in terms of predicting solvation free energies, when applied to large hydrophobes, but that supplementing it by introducing a hard-sphere component that reproduces the van der Waals picture of liquid-vapour coexistence in bulk rescues the model. This seems to be, they say, very much in line with the Lum-Chandler-Weeks picture of a crossover between “volume-driven” hydrophobic hydration at small scales and a surface-driven picture, dominated by interfacial free energies, at larger scales.

Rahul Sarma and Sandip Paul of the Indian Institute of Technology in Guwahati Assam have been rather extensively investigating the origins of osmolyte effects on protein stability using MD simulations. In their latest paper (J. Phys. Chem. B 117, 9056; 2013 – paper here) they look at the stabilizing effect of trimethylamine-N-oxide (TMAO) against the pressure-denaturation of proteins. They find that TMAO enhances the hydration number of a coiled peptide more than its extended state under pressure, whereas the opposite is seen in its absence. Partly this is an indirect effect: solvation of TMAO reduces the ‘crowding’ of water molecules at high pressure that drives them towards the relative freedom they find in the hydration shell of the extended peptide. There is also a direct effect due to the relative inefficiency with which TMAO interacts with the unfolded peptide.

Alkaline phosphatase enzymes are able to catalyse the hydrolysis of a range of different phosphate and sulphate substrates. What accounts for this promiscuity? Through first-principles simulations, Guanhua Hou and Qiang Cui at the University of Wisconsin-Madison show that this class of enzymes can support different types of transition state in the same active site (JACS 135, 10457; 2013 - paper here). The differential placement of water molecules in the active sites is implicated in some of this, although the details are complicated and apparently as yet case-specific: there is no general picture emerging yet of how these and other enzymes engineer their promiscuity.

Another water molecule playing a functional role in enzyme catalysis is identified by Walter Thiel and colleagues at the MPI für Kohlenforschung in Mülheim (B. Karasulu et al., JACS 135, 13400; 2013 – paper here). They look at the three-molecule water bridge that exists in the active site of a lysine-specific demethylase (which is involved in demethylation of histones), showing how it is stabilized by interactions with the surrounding residues and how it assists proton transfer.



Electrowetting – switching surfaces between hydrophobicity and hydrophilicity with electric fields – has useful applications in microfluidics and ink-jet printing, and might enable voltage-gating of flow through nanochannels. This provides the motivation for a study by Alenka Luzar at Virginia Commonwealth University and her colleagues of the effect of electric fields on water dynamics at the interfaces of a slit-like nanopore (M. von Domaros et al., J. Phys. Chem. C 117, 4561; 2013 – paper here). They recognize that the application of a field across the pore breaks the symmetry so that the water dipoles tend to be aligned in opposite directions with respect to the surfaces at each interface – this becomes, they say, the electrical equivalent of a chemical Janus interface. Their simulations indicate that the water density in the first hydration layer is lower for the “incoming” than for the “outgoing” field, and that the response times for orientational dynamics can be up to two orders of magnitude different (in the picosecond regime) – an effect that ought to be visible to, for example, dielectric spectroscopy.

Pavel Jungwirth, Damien Laage and their colleagues have looked at how hydrated ions affect water dynamics (G. Stirnemann et al., JACS 135, 11824; 2013 – paper here). Their simulations indicate that in dilute solution, effects on water reorientational dynamics are short-ranged and ion-specific: both retardation and acceleration may be seen, depending on the interaction strengths. But in concentrated solution, where hydration shells overlap, the effect is always a retardation and is not ion-specific. They suggest that this might explain the apparent discrepancy between the studies of water dynamics by NMR (at low salt concentration) and by infrared and THz spectroscopy (e.g. Huib Bakker), which must use high salt concentrations.

Niharendu Choudhury and colleagues at the Bhabha Atomic Research Centre in Mumbai offer a new view of the local solvation environment in pure water (J. Phys. Chem. B 117, 8831; 2013 – paper here). They say that, by considering the hydration shell of each molecule to be composed only of other waters hydrogen-bonded to it, they can explain the density anomaly without recourse to polyamorphism of the liquid state. I confess that I’ll need to digest this paper more to see where it really diverges from previous treatments; in essence it seems to posit three order parameters, namely a tetrahedral order parameter in the first hydration shell, an orientational order parameter in the second shell, and the number of hydrogen bonds. I think the point is that the first two can be subsumed in the third, and that distortions to the tetrahedral coordination arise at higher temperatures from the incursion of non-hydrogen-bonded molecules into the first shell. The resulting analysis restores a picture of water structure in terms of fluctuations in a homogeneous fluid, rather than requiring the kind of heterogeneities posited by Anders Nilsson.

Ali Hassanali at ETH and coworkers have uncovered what seems to be an extraordinary complexity and subtlety in motion of protons through water (A. Hassanali et al., PNAS 110, 13723; 2013 – paper here). (See also commentary by Edelsys Codorniu-Hernández and Peter Kusalik of the University of Calgary here.) Their first-principles simulations show that there is a great deal more to it than Grotthuss hopping. For a start, the process is very heterogeneous in time: both proton and hydroxide diffusion happen in bursts of activity followed by periods of quiescence. These motions involve medium-range correlations between the movements of many protons, which the authors characterize in terms of the connected-ring structure of the hydrogen-bonded network. The result is that protons can jump over a wide range of length scales, and not merely hope from one molecule to the next. The authors point out that “the fundamental aspects raised in this re- port are likely to open up new directions in the role of the water network in phenomena associated with the hydration of ions and macromolecules such as proteins and DNA.”

Making water modelling easier: Vijay Pande at Stanford and colleagues have simplified the AMOEBA polarisable model of water to a version they call inexpensive or iAMOEBA, which introduces a direct polarization approximation that removes the need to calculate polarization iteratively (L.-P. Wang et al., J. Phys. Chem. B 117, 9956; 2013 – paper here). Any shortfall in the polarization energy introduced by the approximation can be recovered by parametrization, so that the model accurately reproduces the water/ice phase diagram, dielectric properties and so forth.

Tuesday, August 20, 2013

Spectral tuning and pumping in rhodopsins

Inert gases in water will under some conditions not exhibit the expected hydrophobic attraction but will form a solvent-separated pair. Yuri Djikaev and Eli Ruckenstein at SUNY at Buffalo rationalize this observation in the course of developing a new approach to describing hydrophobic hydration (J. Phys. Chem. B 117, 7015; 2013 – paper here). They derive an analytical expression for the number of hydrogen bonds each water molecule has as a function of its distance from the hydrophobic particle, and find that the hydration energy can be positive even for apolar particles if they are small enough, thanks to van der Waals interactions.

On the other hand, two guanidinium ions can form a like-charge ion pair in water, as reported by Richard Saykally and colleagues at Berkeley from XAS measurements (O. Shih et al., J. Chem. Phys. 139, 035104; 2013 – paper here). Their calculations indicate that the ions form a stacked arrangement through pi* interactions.

It is very nice to see that Bruce Berne and colleagues are extending the work on drying transitions in protein assembly to their potential role in molecular recognition and ligand binding (J. Mondal et al., PNAS 110, 13277; 2013 – paper here). Their simulations of an idealized hydrophobic particle being bound in a cavity indicate that the latter can fluctuate between wet and dry states, which have distinct kinetic binding profiles.

More on water’s liquid-liquid transition: Yaping Li and colleagues at the University of Arkansas see such a first-order transition between HDL and LDL, with an associated critical point, in simulations using the so-called Water potential from Adaptive Force Matching for Ice and Liquid (WAIL), which they argue is a particularly reliable potential because it is not biased by fitting to experimental data (Y. Li et al., PNAS 110, 12209; 2013 – paper here).

And in a preprint apparently heading for the Eur. Phys. J., Hajime Tanaka presents a new model of the liquid state based on an order parameter that describes bond orientational ordering, which he says explains many of the anomalies of tetrahedrally bonded liquids (including a liquid-liquid transition) (arxiv.org/abs/1307.0621). This huge and ambitious paper also provides accounts of glass formation and fragility, quasicrystal formation and other metastable crystallized states.

Somewhat related in exploring tetrahedral order in water is a preprint destined for Chem. Phys. Lett. by Marcelo Carignano of Northwestern University and colleagues, who have compared this aspect of water structure along with hydrogen-bond lifetimes and diffusion coefficients in simulations with four different water potentials (SPC/E, TIP4P-Ew, TIP5P-Ew, Six-site) (arxiv/1307.3611). They find that there are significant differences below 270 K, depending on whether or not the models include explicit lone pairs.

Shaoyi Jiang and colleagues at the University of Washington in Seattle present a case, from simulations and NMR measurements, that the hydration differences for salt bridges in proteins contribute significantly to their relative stabilities (A. D. White et al., J. Phys. Chem. B 117, 7254; 2013 – paper here).

More on spectral tuning of photoactive sites by hydration environments. Sivakumar Sekharan and colleagues at Yale say that the chloride-pumping transmembrane retinal protein halorhodopsin in a halophile works by subtle rearrangements of waters and ions in the vicinity of the chromophore as chloride translocation progresses, inducing changes in chromophore bond lengths that affect its absorption wavelength (R. Pal et al., JACS 135, 9624; 2013 – paper here). Meanwhile, Hyun Woo Kim and Young Min Rhee at the Pohang University of Science and Technology in Korea say that the pH-dependence of emission from firefly oxyluciferin – a redshift in acidic conditions – seems to be fine-tuned by a water molecule in the active site that can mediate the dynamics of neighbouring groups to the chromophore (J. Phys. Chem. B 117, 7260; 2013 – paper here).

Valentin Gordeliy of the Université Grenoble Alpes and colleagues describe the mechanism of a different rhodopsin-like pump, an unusual proteorhodopsin from a permafrost bacterium that pumps protons in a different manner to other known rhodopsins (I. Gushchin et al., PNAS 110, 12631; 2013 – paper here). Unusually, this structure has a proton release site (a lysine) already connected to the bulk solvent by a water chain in the ground state. There’s a nice graphic that illustrates the similarities and differences in the water-containing cavities of the channel for this structure and other rhodopsins, bacteriorhodopsin and xanthorhodopsin, the closest homologue of this one (ESR):



I guess it is ultimately a cautionary tale that one should extract from the simulation study of reverse micelles by John Straub and colleagues at Boston University: they say that the structure and dynamics of the encapsulated water and the overall micelle assembly depend strongly on the force fields used (A. V. Martinez et al., J. Phys. Chem. B 117, 7345; 2013 – paper here). In any event, the micelles tend not to be spherical but undergo pronounced shape fluctuations, a point made also by earlier simulations but which tends to be ignored in interpreting experimental data.

I must confess that I’d not previously come across inhomogeneous fluid solvation theory (IFST) as a general method for calculating the effect of a solute on the solvent structure. But now David Huggins and Mike Payne at Cambridge have assessed how well it does for predicting the hydration free energies of small solutes (J. Phys. Chem. B 117, 8232; 2013 – paper here). They look at six solutes such as benzene, isobutene and methanol, all in water, and figure that IFST does pretty well in comparison with a more sophisticated theory but that accurate prediction of entropies might suffer from poor sampling of the available configurations unless the simulation times are rather long.

And here’s a curious one: George Reiter at Houston and colleagues say that water confined on the scale of about 2 nm has a different ground-state configuration of valence electrons than in the bulk (G. F. Reiter et al., Phys. Rev. Lett. 111, 036803; 2013 – paper here). They explore the issue using X-ray Compton scattering to probe water in the pores of Nafion, but they say, reasonably enough, that “Biological cell function must make use of the properties of this [nanoconfined] state and cannot be expected to be described correctly by empirical models based on the weakly interacting molecules model.”

Monday, July 15, 2013

A role for water in allostery

Rearrangement of the water network hydrating a protein can provide a mechanism for allostery, according to a study by Peter Hamm and colleagues (B. Buchli et al., PNAS 10.1073/pnas.1306323110; paper here). They insert an azobenzene photoswitch in the binding groove of a PDZ domain protein, a common system for studying allostery, such that photo-induced isomerisation induces a conformational change similar to that which occurs on ligand binding. This enables them to use fast IR spectroscopy to look at the opening of the binding groove in a precisely controlled fashion, having first characterized the equilibrium structures using NMR. Their MD simulations show that a change in water density in the vicinity of the photoswitch right after switching propagates slowly through the water network over about 100 ns until it reaches the back of the protein. They suggest that this change in hydration structure could then induce, either dynamically or structurally, a remote allosteric change in protein conformation. In such a case, this would be a particularly dramatic example of how the hydration network is really a part of the functional apparatus of the protein.

How aquaporins transport water across membranes – and specifically how they do so without also transporting protons – has been a topic of much debate. The consensus has come to focus on the so-called NPA motif in the centre of the channel, a bottleneck through which water kolecules pass in single file, which seems to prohibit proton transport via electrostatic repulsion. Urszula Kosinska Eriksson at the University of Gothenburg and colleagues have recently reported a new high-resolution crystal structure of yeast aquaporin 1 which sheds new light on the issue (Science 340, 1346; 2013 – paper here). As Jeff Abramson and Armand Vartanian of ULCA explain in an accompanying perspective (Science 340, 1294 – paper here), the structure shows that proton transport isn’t (as some have suggested) blocked by hydrogen-bonding of a single water molecule in the NPA region to two asparagines. Rather, there are two independent waters here, but the interactions with the asparagines constrain their dynamics in such a way as to effectively break the ‘water wire’ threading through the channel. The authors say that water transport then seems to happen in pairwise fashion, similar to the ion transport in potassium channels – a possible example of convergent evolution to solve related problems.

A carefully structured water cluster also seems to play an important role in oxygen evolution during the photocycle of photosystem II, as Brandon Polander and Bridgette Barry of Georgia Tech deduce using laser flashes to induce the process and following it by FTIR spectroscopy (PNAS 10.1073/pnas1306532110 – paper here). The hydrogen-bonded cluster of five water molecules, bound to the catalytic Mn4CaO5 cluster, seems to become protonated during the S1→S2 part of the cycle, and it stores the proton until a later stage of the reaction. Ammonia can poison the reaction by disrupting the water cluster.

Water molecules that gain access to the interiors of globular proteins can act as probes of the intrinsic conformational dynamics that enable proteins to function. These waters in the ‘dry’ protein interior have been studied by magnetic relaxation dispersion spectroscopy, but atomistic models are needed to interpret those results. The problem is that such deep water penetration tends to be a rare event, demanding very long (millisecond) run times for simulations. A technique has recently been developed that enables this (D. E. Shaw et al., Science 330, 341; 2010), and now Filip Persson and Bertil Halle at Lund have used the method to compare MD with the MRD experiments for bovine pancreatic trypsin inhibitor (JACS 135, 8735; 2013 – paper here). They find that some of these internal hydration sites have water residence times of several microseconds, and that the water molecules gain access along single-file hydrogen-bonded chains.

Dewetting transtions are known to be important for at least some instances of hydrophobic assembly, and Bruce Berne and colleagues at Columbia now extend their earlier studies of dewetting-induced protein collapse to look at the potential role of such transitions in the docking of hydrophobic ligands in their binding pockets (J. Mondal et al., preprint http://www.arxiv.org/abs/1305.7505). In this way the solvent dynamics, which are retarded in the concave cavity, are explicitly included in the kinetics of the binding process – the process can be parametrized through a state variable that describes whether the pocket is ‘wet’ or ‘dry’, while the ligand diffuses across a potential-energy surface that can switch between these two states.

Daniel Sindhikara and Fumio Hirata of the Ritsumeikan University in Japan present a fast algorithm, based on the three-dimensional reference interaction site model (3D-RISM), for calculating the solvent distribution around solutes (J. Phys. Chem. B 117, 6718; 2013 – paper here). This is a wholly theoretical approach derived from the Ornstein-Zernicke equation. They say that it gives water positions and orientations that agree well with available experimental data, and demonstrate its use on HIV-1 protease.

Paul Ben Ishai at the Hebrew University of Jerusalem and colleagues have used quasielastic neutron scattering to look at salt effects on water dynamics (J. Phys. Chem. B 117, 7724; 2013 – paper here). They find that water diffusion is slower than in pure water, on average, in NaCl solution, but faster in KCl. They interpret the result in terms of structure-making and –breaking, saying that the disruption of the hydrogen-bonding network by potassium ions accounts for its apparent ‘lubricating’ effect.

The hydration state of arginine side chains can be deduced from its UV resonance-enhanced Raman spectrum, according to Sanford Asher at Pittsburgh and colleagues (Z. Hong et al., J. Phys. Chem. B 117, 7145; 2013 – paper here). Their density-functional calculations show that a particular vibration of this residue is sensitive to hydration. They use this signal to characterize differing degrees of hydration of Arg in two polyAla model peptides.

Friday, June 28, 2013

Complexity of osmolytes

More on the mechanisms of denaturation and osmolyte stabilization. Trimethylamine-N-oxide (TMAO) is known to confer protection against protein denaturation by urea. Rahul Sarma and Sandip Paul at the Indian Institute of Technology have carried out MD simulations to try to figure out why (J. Phys. Chem. B 117, 5691; 2013 – paper here). They say that TMAO interacts with a model peptide, N-methylacetamide (NMA), so as to cause some dehydration by replacing solvation water. However, this interaction with the peptide is relatively inefficient, especially because TMAO cannot donate its hydrogen to the backbone carbonyls. As a result, TMAO is not efficient at stabilizing the unfolded peptide chain. If the protein folds, then TMAO is more available for forming very strong hydrogen bonds with water, and indeed with urea. It’s a subtle argument in which the direct interactions between all four components – TMAO, urea, peptide and water – are implicated.

Much the same issues are explored in a new preprint by Bruce Berne and coworkers at Columbia (J. Mondal et al., http://www.arxiv.org/abs/1306.4642). They look at the interactions of urea and TMAO separately with a model 32-mer made of Lennard-Jones beads whose hydrophobicity can be varied. Although both small solutes interact strongly with the polymer, TMAO destabilizes extended conformations while urea stabilizes them. This seems to be because TMAO molecules are more stable (via van der Waals interactions) next to the collapsed rather than the extended polymer. The results provide further support for an interpretation of osmolyte effects that invokes direct interactions rather than indirect effects on “water structure”.

The subtlety of mixed-osmolyte solutions is revealed in simulations of a hydrophobic polymer in urea and guanidinium chloride by Payel Das at IBM Yorktown Heights and colleagues (P. Das et al., Langmuir 29, 4877; 2013 – paper here). They find that, while both of these molecules act as denaturants on their own, in combination they actually promote collapse of the polymer. It is hard to tell a simple story about why this happens, although apparently Paul Flory predicted back in 1955 that two good solvents could combine to induce polymer chain collapse. Although Gdm has the stronger interaction with the polymer, its enhanced concentration in the vicinity of the polymer attracts urea due to the favourable urea-Gdm interaction, with the result that urea is in fact preferentially adsorbed onto the polymer. This sets up a long-ranged interaction between the monomers mediated by their clouds of urea molecules, ultimately driving collapse.

The effect of progressive dehydration of a protein (lysozyme) on its structure and dynamics is studied using Raman spectroscopy by Gediminas Niaura of Vilnius University of colleagues (J. Phys. Chem. B 117, 4981; 2013 – paper here). They find that there is a structural change in the dry protein crystal that begins at about 7-10 wt% water, and that the native state is reached at about 35 wt% water (which amounts to appreciably more than monolayer coverage). In the dry state the protein is dynamically glassy, with a diminished proportion of alpha helices and an enhanced content of beta-sheet contacts.

And on the question of ‘dry’ proteins, Eric Gloaguen at the CNRS Laboratoire Francis Perrin in Gif-sur-Yvette and colleagues report that small peptides with aromatic residues will fold into hydrophobic domains even in the solvent-free gas phase, showing that this is a favourable conformation even in the earliest stages of protein folding (E. Gloaguen et al., J. Phys. Chem. B 117, 4945; 2013 – paper here).

Somedatta Pal and Sanjoy Bandyopadhyay at the Indian Institute of Technology in Kharagpur have used MD simulations to look at the complementary issue of how protein (here barstar) dynamics affect the dynamics of hydration water (J. Phys. Chem. B 117, 5848; 2013 – paper here). They compare the normal situation of the hydrated protein with that in which the protein is kept frozen, looking specifically at the effect of the change on the low-frequency vibrations of the water molecules. Freezing the protein results in stronger confinement of water bound to the protein surface, with a corresponding blue shift of the vibrational frequency for transverse water oscillations – but much less effect on longitudinal oscillations.

An intriguing paper by Nicholas Spencer and colleagues at ETH investigates hydration forces for glycoproteins using the surface force apparatus (R. M. Espinosa-Marzel et al., Biophys. J. 104, 2686; 2013 – paper here). They attach these onto hydrophobic and hydrophobic surfaces, and find that there is a rather long-ranged repulsive force (several tens of nanometres) between the surfaces. It is good to see this neglected class of biological molecules get some attention in terms of their fundamental hydration characteristics. However, I don’t buy the interpretation that the repulsion is due to some “long-ranged structuring of water”. The whole discussion around the results seems to be something of a throwback, starting with the whole notion of “vicinal water” and involving two-state water theories, kosmotropes and all the rest of the paraphernalia that tended to surround discussions of ‘water structure’ 20 years ago. “Ordering” of water over 30 nm or so just isn’t any longer consistent with what is known of other systems, and it’s been suggested to me that steric repulsion of the surfaces due to a few of the macromolecules protruding from the monolayer is a much more likely explanation of the effects observed.

But could it be that long-ranged water structuring is going to try to stage a little comeback? I ask this because it is invoked in another recent paper too, by Jan Christer Eriksson and Ulf Henriksson of the Royal Institute of Technology in Sweden (Langmuir 29, 4789; 2013 – paper here). They develop their earlier argument (Langmuir 23, 1126; 2007) that the long-ranged hydrophobic attraction might be accounted for by the formation of roughly cylindrical bridging water clusters that are “slightly more organized than the rest of the film”. Their analysis suggests that such a proposal can explain some recent measurements on water thin films with the SFA (Wang et al., J. Coll. Int. Sci. 364, 257; 2011). But I think I will stick with the bridging-nanobubbles idea.

Friday, May 10, 2013

No ice please

I’m not a fan of the Frank-Evans iceberg picture of hydrophobic hydration. But I will happily admit that, on the question of whether water hydrating small nonpolar solutes is more/less ordered and mobile than that in the bulk, the evidence is mixed. Nuno Galamba of the University of Lisbon now offers some more support for a weak version of the iceberg picture (J. Phys. Chem. B 117, 2153; 2013 – paper here). His MD simulations show that a subset of water molecules in the first hydration shell of methane have “significantly enhanced tetrahedrality and a slightly larger number of hydrogen bonds”, as well as slower reorientational dynamics, relative to the bulk. He adds that these characteristics should not be visible in the rdfs deduced from neutron scattering, explaining why they have not been seen experimentally. Whether this view extends to large hydrophobes is another matter, but these results at least argue for some small degree of water ‘ordering’, even if this is very far from ice-like.

The role of hydration water in the dynamics of hydrophobic sidechains of peptides is explored by Daniela Russo at the ILL in Grenoble and colleagues, using inelastic neutron scattering and simulations (J. Phys. Chem. B 117, 2829; 2013 – paper here). They say that the activation energy for methyl group motions increases with increasing level of hydration but eventually reaches a plateau when an extended hydrogen-bonding network is established around the group, which happens when it is essentially surrounded by a single layer of water. These sidechain dynamics seem to have a critical impact on the flexibility of the peptide as a whole, and so the degree of hydration appears to determine the onset of conformational freedom for the entire polypeptide.

Neeraj Sinha and colleagues at the Centre of Biomedical Magnetic Resonance in Lucknow have taken on the challenge of exploring water-protein interactions in a rather complex system, the helical coat protein of the filamentous virus Pf1 (R. N. Purusottam et al., J. Phys. Chem. B 117, 2837; 2013 – paper here). Using 2D proton-N15 NMR, they deduce that the filamentous assembly has a highly hydrated core which not only acts as a ‘glue’ but might also mediate the interaction of the arg44 residue with DNA.

Membrane transporters are membrane protein channels that pump directional transport of small molecules across the membrane, coupled to chemical energy sources in the cell. Their operation has generally been considered to involve a carefully orchestrated sequence of conformational changes to ensure one-way and selective transport. But it has been found that sometimes water and ions get through too, and the question arises of whether this is passive, osmotically driven ‘leakage’ or stoichiometric co-transport of these species. Emad Tajkhorshid and colleages at Illinois at Urbana investigate this question using MD simulations, for several classes of membrane transporters (J. Li et al., PNAS 110, 7696; 2013 – paper here). They find that these generally support states in which there are channels that permit passive water flux – but that this does not interfere with the coordinated vectorial transport of the primary substrate. Thus the transporters are imperfect, but not problematically so: as the authors put it, “Given the soft mechanical properties of transporter proteins, it comes as no surprise to observe harmless imperfections in the overall gating motions, which manifest themselves in the formation of water-conducting states. It would, of course, be a concern if these channels were large enough to leak the substrate, and/or long-lived to allow very large amounts of smaller species to permeate across the membrane. Neither of these aspects appears to be the observed in our results, because the leaky states are only large enough for small species such as water. Furthermore, it appears that these states only transiently rise during the transport cycle, an attribute that might make them difficult to capture experimentally.”

Solvation of small molecules in water has often been described using point charges affixed to the solutes to represent polarization effects. David Cerutti at Rutgers and colleagues present a quantum chemical approach for fitting such partial charges to solutes like amino acids (D. S. Cerutti et al., J. Phys. Chem. B 117, 2328; 2013 – paper here). They say that their approach, which represents an evolution from the simplest point-charge models of previous decades, predicts substantially more polarization of amino acids than earlier efforts using AMBER force fields.

More on urea-induced protein denaturation, this time from Michela Candotti of the Institute of Research in Biomedicine in Barcelona and colleagues, who use MD, SAXS and NMR data to develop an atomistic picture of the unfolded states and the energetics of unfolding and refolding (M. Candotti et al., PNAS 110, 5933; 2013 – paper here). They conclude that urea’s denaturing influence is a combination of kinetic (disrupting stabilizing intramolecular contacts) and thermodynamic (stabilizing the extended conformation) effects. As I understand it, the results support models based on direct interactions of urea and protein, while also revealing that the urea-unfolded state is rather different to the denatured conformation that exists in pure water.

Hydration is evidently important to DNA conformation, influencing the A-B transition, interactions with DNA-binding proteins, and perhaps even affecting shape in a sequence-dependent manner. There is some evidence that the hydration water of DNA has collective dynamics of a glassy nature, and this notion is offered further support in inelastic neutron scattering experiments by Alessandro Paciaroni at the Università degli Studi of Perugia and colleagues (A. Paciaroni et al., J. Phys. Chem. B 117, 2026; 2013 – paper here). They find that at 100 K the large-wavevector scattering from water, related to coherent excitations, seems to imply a character related to amorphous ice. In other words, the interactions with DNA significantly alter the structure and dynamics of the interfacial water.

Melittin, one of the multi-subunit peptides that seems to aggregate by dewetting, is a component of been venom that acts as an antimicrobial. The melittin tetramer forms a pore that inserts into lipid membranes, and Max Berkowitz and colleagues at UNC now suggest that its effect is to create transient water-permeable channels, making the membranes leaky (K. P. Santo et al., J. Phys. Chem. B 117, 5031; 2013 – paper here). The technique they use for MD simulations can handle timescales of up to microseconds, and they show that the melittin peptides gradually aggregate into a kind of wedge that punctures the membranes and allows water to pass through, before falling apart again.

[FeFe] hydrogenase catalyses hydrogen-ion reduction to gaseous hydrogen, and could therefore be a useful biocatalyst. Simulations by Martin McCullagh and Greg Voth at Chicago now show that it seems to work by coupling electron transfer to proton transfer along a previously unknown water channel accessing the active site (J. Phys. Chem. B 117, 4062; 2013 – paper here).

There are seemingly strange rumours that fully miscible liquids are inhomogeneous on length scales of hundreds of nanometres, a conclusion suggested by some light-scattering and small-angle neutron-scattering studies. For example, Marián Sedlák of the Slovak Academy of Sciences reported such a claim in a series of papers in 2006 (e.g. J. Phys. Chem. B 110, 4329). Could these be solute clusters, or perhaps nanobubbles stabilized by adsorbed solute? Sedlák and Dmytro Rak now investigate that latter possibility (J. Phys. Chem. B 117, 2495; 2013 – paper here). They look at a range of solutes: magnesium sulphate, citric acid, urea, and t-butyl alcohol, and find no significant differences in the scattering from normal and degassed solutions, apparently ruling out the nanobubble interpretation.

Thursday, April 25, 2013

Hydration and thermostability

What makes thermophilic proteins stable? And in particular, does hydration play a role? Those questions are examined by Fabio Sterpone of the Université Paris Diderot and colleagues (O. Rahaman et al., Phys. Chem. Chem. Phys. 15, 3570; 2013 – paper here). They study two homologous proteins, one from E. coli and the other from the thermophile Sulfolobus solfataricus, using MD simulations. The average water dynamics at the protein surface is the same in both cases, despite their different amino acid compositions, being slowed by a factor of 3-5 relative to the bulk. The authors conclude that this slowdown is primarily a geometric effect due to excluded volume, which explains why the protein sequence has little influence. This doesn’t exactly answer the initial question about thermostability – but it does suggest that there is nothing special about thermophilic proteins in terms of their hydration.

More on dewetting as a mechanism for protein folding and stability. Ruhong Zhou, Robert Matthews and their colleagues at Columbia have simulated a TIM barrel protein and found drying inside clusters of hydrophobic residues, signified by strong fluctuations in water density (P. Das et al., JACS 135, 1882; 2013 – paper here). In particular, a hydrophobic cluster (ILV) at the N-terminus shows drying that is weakened or suppressed by substituting some of the hydrophobic amino acids for less hydrophobic ones. A cluster at the C-terminus, meanwhile, seems to experience only partial drying that is unaffected by such substitutions. Experiments on the structure and stability of wild-type and mutant versions seem to back up these conclusions. The authors note that ILV clusters are common in several other protein motifs too.

Huib Bakker and colleagues at the FOM Institute in Amsterdam have used Förster resonant energy transfer between water OH stretches to find out where the water hydrating lipid membranes actually is (L. Piatkowski et al., J. Phys. Chem. B 117, 1367; 2013 – paper here). Somewhat surprisingly, they find that, even at rather low hydration levels, the water molecules are not evenly distributed among the lipid head groups, but form nanoscale clusters with an average intermolecular distance of 3.4 Å.

Our picture of the water surface continues to evolve. Is the water here structurally and dynamically different from that in the bulk? That idea is supported by experiments by James Skinner and colleagues at Wisconsin in simulations using a new three-body water potential (Y. Ni et al., PNAS 110, 1992; 2013 – paper here). They calculate that the hydrogen-bond switching dynamics are retarded at the water surface by a factor of about 3, although the rotational dynamics are actually a little faster. They say that vibrational 2D sum-frequency generation spectroscopy should be a good experimental method for investigating these dynamics, and calculate what the spectra should look like.

To make optimal use of plant cellulose as a feedstock for biofuels such as ethanol, it’s necessary to get it into aqueous solution. But cellulose is hard to solvate, which is why it is important to understand its hydration structure. Sylvia McLean at Oxford and colleagues have used neutron diffraction to look at the aqueous solvation of the disaccharide cellobiose, and find that there is (as has been suggested) a hydrogen bond across the glycosidic bond linking the two sugars (W. B. O’Dell et al., PLoS ONE 7, e45311; 2012 – paper here). There is competition from water molecules for the oxygen acceptor in this bond, however, with average occupancy of 50% for both water and the intramolecular OH donor.

More evidence that a hydrogen-bonded cluster of water molecules at the catalytic site plays a crucial role on photosynthesis: Bridgette Barry and colleagues at Georgia Tech use EPR to look at the rate of reaction of an intermediate neutral radical in the proton-coupled electron transfer that leads to oxygen evolution at the reaction centre of photosystem II, which contains a catalytic Mn4CaO5 cluster (J. M. Keough et al., J. Phys. Chem. B 117, 1296; 2013 – paper here). They find that this network is rearranged during the transition between the S0 and S2 states of the catalytic cycle, and that ammonia slows oxygen evolution because it disrupts the network by displacing water.

Suzi Jarvis and colleagues at University College Dublin have been using scanning probe microscopy for some time to probe hydration forces, and in their latest contribution they use a technique called frequency modulation AFM to look at hydration forces at the interface of mica and an electrolyte (J. I. Kilpatrick et al., JACS 135, 2628; 2013 – paper here). By ‘hydration force” here they mean the monotonically decaying force with a characteristic length of a few Å, without regard to its precise origin. They find that, relative to pure water, ions introduce or accentuate oscillations in the force as a function of distance due to the formation of distinct hydration layers. They point out that there are implications for obtaining atomic-resolution AFM images in aqueous saline solution.

Uzi Landman, Gary Schuster and colleagues at Georgia Tech offer a fascinating insight into the role of hydration water around DNA in the oxidation and consequent mutation of A/T-rich regions, which my be significant in the early stages of carcinogenesis due to stalling of replication (R. N. Barnett et al., JACS 135, 3904; 2013 – paper here). Their experiments and simulations indicate that oxidation of adenine leads to proton-coupled electron transfer to thymine, mediated by a water wire. This process can explain why nearly all the mutation due to such an oxidation event happens at thymine.

It’s sobering to realise that even now the reasons for the anomalous behaviour of water’s thermodynamic response functions, such as the divergence in the heat capacity and compressibility at low temperature, are still unknown. Francesco Mallamarce, Carmelo Corsaro and Gene Stanley discuss this issue with reference to experimental data on the power spectrum of sound velocity (F. Mallamarce et al., PNAS 110, 4899; 2013 – paper here). They conclude that the anomalies are associated with a structural transformation due to the appearance of an extended hydrogen-bonding network, which gives rise to viscoelastic behaviour in the liquid.

Also on bulk water, what happens to water’s structure above its critical point has been studied using X-ray Raman spectroscopy by Christoph Sahle of the Technical University of Dortmund and colleagues (C. Sahle et al., PNAS 110, 6301; 2013 – paper here). They find that, as one might expect, distortions of the hydrogen bonds are significant above the critical point, and the average coordination of each molecule decreases to just 0.6 at 600 oC and 134 MPa.

Friday, April 12, 2013

How antifreeze proteins work

A relatively short catch-up, with much more to come…

Martina Havenith at Bochum and her colleagues have extended their previous work on the mechanism of antifreeze proteins. Previously they looked at an antifreeze glycoprotein and found, using terahertz spectroscopy, that ice-binding by the protein seems to involve a long-range retardation of water H-bond dynamics, extending up to 2 nm from the molecular surface (S. Ebbinghaus et al., JACS 132, 12210; 2010). Now they find a similar effect operating for the antifreeze protein (AFP) of the fire-coloured beetle D. canadensis, a hyperactive insect (K. Meister et al., PNAS 110, 1617; 2013 – paper here). This contrasts, however, with the mechanism of another class of AFP, called wfAFP-1, which seems to operate only by short-ranged water binding to surface OH groups (S. Ebbinghaus et al., Biophys. J. 103, L20; 2012). As the authors say, “Nature is probably more inventive than initially thought and makes use of short- and long-range water perturbation to varying degrees in different classes of AFPs”.

Meanwhile, Ido Braslavsky of Ohio University and colleagues have used microfluidic methods to study the effects of AFPs on ice nucleation, and find that the growth of ice crystals is inhibited by the irreversible surface binding of the proteins (Y. Celik et al., PNAS 110, 1309; 2013 – paper here). These results help to rule out suggestions that direct binding of the AFPs to ice is not necessary to their mode of action.

More cases of water assisting receptor-substrate recognition and catalytic activity. First, Stephen Neidle at UCL and colleagues find that a cluster of 11 water molecules in an AT region of the minor groove of DNA seems to support the binding of three different small-molecule ligands (D.-G. Wei et al., JACS 135, 1369; 2013 – paper here). This cluster stabilizes the ligand by hydrogen bonding, and is also linked to (but distinct from) the well-known spine of hydration in B-DNA. Slight differences in binding mode with this cluster seem to account for the differences in binding affinity of the ligands: in other words, it is the water network that is calling the shots.

Second, Xiaoqing Wang and Hajime Hirao at Nanyang Technological University in Singapore say that the catalysis of myo-inositol monophosphatase (IMPase, a potential target for lithium treatment of bipolar disorder) is dependent on two bound water molecules (J. Phys. Chem. B 117, 833; 2012 – paper here). One provides the hydroxide ion that attacks the bound substrate. The other, coordinated to a magnesium ion, facilitates proton transfer leading to the product.

There seems to be an increasing perception that understanding the collective vibrations of proteins – their softness or rigidity – could offer insights into their enzymatic activity. Sow-Hsin Chen at MIT and coworkers support that view with an X-ray scattering study of the collective modes of hydrated lysozyme (Z. Wang et al., J. Phys. Chem. B 117, 1186; 2013 – paper here). They find that at low hydration levels both the collective ‘soft’ phonon modes and the enzymatic activity are much weaker or absent, suggesting a causal relationship: an indicator of the now familiar plasticizing effect of hydration.

C. Cametti at “La Sapienza” University of Rome and colleagues consider another aspect of protein hydration: how high concentrations of the protein (again lysozyme) can lead to clustering and a consequent decrease in average hydration number (C. Cametti et al., J. Phys. Chem. B 117, 104; 2013 – paper here). Their measurements of the dielectric spectra from 500 MHz to 50 GHz, which probe orientational relaxation, are consistent with this hypothesis of clustering into small aggregates at concentrations above about 100 mg/mL, which was first proposed by Stradner et al. (Nature 432, 492; 2004).

There is now some debate about whether the proposed two metastable liquid phases and associated critical point claimed on the basis of some simulations is real or not. That has been disputed by David Limmer and David Chandler at Berkeley, who extend their previous negative results using several different water potentials in a new preprint. I have written a commentary on the issue here.

Friday, February 22, 2013

On icebergs

Still we ponder the nature of hydrophobic hydration: is it more ‘ice-like’ in some sense? Apparently so, say Dor Ben-Amotz and colleagues at Purdue (J. G. Davis et al., Nature 491, 582; 2012 – paper here). They use Raman spectroscopy to monitor vibrational spectra of water hydrating linear alcohols ranging from methanol to heptanol, and see more tetrahedral ordering of water molecules, and fewer weak hydrogen bonds, at low temperatures. But for hydrophobic chains longer than about 1 nm this hydration structure gives way to one in which the water is more disordered and has weaker H-bonds at higher temperatures. This scale-dependent crossover is reminiscent of that proposed by Lum, Chandler & Weeks above about 1 nm (J. Phys. Chem. B 103, 4570; 1999).

The nature of the hydration shells of methanol, ethanol and propanol have been probed using THz spectroscopy by Vladimir Matvejev of the Free University of Brussels and colleagues (J. Phys. Chem. B 116, 14071; 2012 – paper here). They estimate that the shells comprise about 14, 23, 23 and 31 molecules for methanol, ethanol, 1- and 2-propanol, respectively, and the water molecules are retarded by a factor of around 1.4.

In a somewhat similar vein, L. Luca of the University of Perugia and colleagues use GHz-THz light scattering to probe the hydration of mono- and disaccharides (J. Phys. Chem. B 116, 14760; 2012 – paper here). They find that slowing of water collective reorientation (by relatively large factors of 5-6) occurs only over relatively short distances (3-4 Å or essentially the first hydration layer), regardless of the size of the sugar molecules. This retardation involves considerably more water molecules than those few instantaneously hydrogen-bonded to the sugar.

A more generic approach to small-molecule solvation is described by Alla Oleinikova and Ivan Brovchenko of the Dortmund University of Technology (J. Phys. Chem. B jp306781y – paper here). They use MC simulations to study water structure around spherical solute particles 3-10 Å in size that vary from strongly hydrophobic to strongly hydrophilic. In all cases there is a density depletion relative to the bulk due simply to the missing-neighbour effect. For strongly hydrophobic particles there is a drying transition at the surface. Similar effects are seen for other fluids, but the directional hydrogen bonding of water enhances them.

Water in useful places: there is likely to be water molecules forging a hydrogen-bonded assembly in the mammalian photoreceptor melanopsin in the retina, which triggers the biological clock, according to Sivakumar Sekharan and colleagues at Cornell (JACS 134, 19536; 2012 – paper here). Their first-principles calculations of the active site suggest that two water molecules bridge the Schiff base and residues on the peptide, accounting for the blue shift of the optical absorption relative to the closely related rhodopsin.

Cytochrome c might control water access to its heme centre to tune the reduction potential via hydration changes, according to quantum MD simulations of Isabella Daidone of the University of L’Aquila and colleagues (C. A. Bortolotti et al., JACS 134, 13670; 2012 – paper here). They find that a yeast cytochrome has two channels that it can open to admit water, altering the reduction potential and thus refining processes of electron transfer.

How water evaporates from the air-water interface has been studied in detail by Patrick Varilly and David Chandler at Berkeley (J. Phys. Chem. B jp310070y – paper here). They find that the escape trajectory of a water molecule can be described in terms of two parameters: the distance from the instantaneous interface and the velocity along the surface normal. The results seem to imply that evaporation has zero activation energy, as some but not all experiments have suggested.

Fresh variety in the structures of ice confined to nanopores is reported by Jaeil Bai and Xiao Cheng Zeng of the University of Nebraska (PNAS 109, 21240; 2012 – paper here). Their simulations indicate that a bilayer of ice-like water molecules in slit-like pores about 7-9Å apart can be transformed under pressure to an amorphous phase and then to a very-high-density amorphous phase at 250K and 3 GPa. For rapid compression to 6 GPa an entirely new VHD ordered phase is found in which the water molecules are linked into square nanotubes.

Andreas Barth and colleagues at Stockholm University propose that studying changes in the water absorption bands in infrared spectroscopy can offer a way of quickly and remotely monitoring for binding of a ligand to a target protein in drug development (S. Kumar et al., J. Phys. Chem. B 116, 13968; 2012 – paper here). Specifically, they look at the bands diagnostic of ‘bound’ water in the binding cavity, and see this decline as ligand binding expels the water. They have so far looked at rather high protein concentrations, but think that the approach should work at lower concentrations with brighter light sources.

Hydrophobic drugs can be rendered more water-soluble by adding certain solutes known as hydrotropes. Although this has been long known, the mechanism is not clear. Seishi Shimizu of the University of York and colleagues have studied the question using thermodynamic theory (the so-called fluctuation theory of solution) and measurements of thermodynamic quantities (J. J. Booth et al., J. Phys. Chem. B 116, 14915; 2012 – paper here).

Junrong Zheng at Rice University and coworkers report evidence for strong segregation of ions and water in strong electrolyte solutions of thiocyanate ions (J. Phys. Chem. B jp310153n – paper here). The clustering is enhanced by the addition of strongly hydrated ions such as fluoride, whereas iodide ions tend to associate with the SCN clusters. In any event, the aqueous systems are strongly inhomogeneous.

Finally for now – though lots more still to catch up on – I have a brief article on the Chemistry World site (here) on a recent paper by Thomas Kühne and Rustam Khaliullin at Mainz (Nat. Commun. 4, 1450; 2013 – paper here) that seems to shed light on the arguments over XAS studies of liquid water by Anders Nilsson and colleagues (Wernet et al., Science 304, 995; 2004). Views on whether there was a controversy here in the first place will doubtless vary, but it seems that this new work provides a useful perspective on what the XAS work was showing. Anders gave me some rather extensive remarks on his view of the matter, which I thought might be usefully reported here. Forgive me for presenting them here unmediated – they will hopefully at least clarify Anders’ current position on this. Everything that follows is from him:

I find this paper to represent a very important step forward. It is very close to our original suggestion in the Wernet et al. Science paper that many molecules will be in instantaneous configurations with one strong and one weak hydrogen bond, i.e. asymmetric configurations. XAS measures the electronic structure on a timescale of a few attoseconds which means that the molecules have no time to move so we are only detecting snapshots of space-averaged frozen configurations (also stated in the original Wernet et al. paper).

However, there is a difference. In real water we expect the asymmetry between the strong and weak bonds to be much larger than based on the structures from the current simulation. It is mentioned in the methods section but not directly in the main text. Here is the underlying experimental evidence that requires no spectroscopy interpretation.

I attach a recent study (JCP in press, Skinner et al.) where we together with Benmore’s group have undertaken a major effort to determine a more accurate O-O pair distribution function (PDF) of ambient water. This is based on 4 new x-ray diffraction data sets (only 3 are consistent at high Q), all with a much higher Q cut-off than any previous measurements where we have also taken serious care to remove any OH contributions. Figure 9a shows the O-O PDF with very small error bars with a first peak height of 2.57. If you comparethis with the O-O PDF presented in the supplementary material it is clear that the simulations have an overstructured main peak. For the PBE functional, used in the main text, the peak height is 3.3. The TPSS –D3-FF simulation shown in the supplementary information has a peak height of 3.1, still overstructured but less so than the PBE. The consequence in the asymmetry is clearly visible in terms of the weak bond energy distribution from fig. 3a and S3a where the latter has more contributions towards lower energies. If we have to further dramatically understructure the liquid down to a peak height of 2.57 in the O-O PDF we expect the asymmetry to become much larger. In my opinion their asymmetry is only a lower limit.

Most likely the underlying reason for the asymmetry is in the many-body cooperativity effects that only an electronic structure simulation can capture as demonstrated in the current paper, and which is not represented by classical force fields. It has been known that the cooperativity effect is strongest when you have one strong donor and one strong acceptor bond. Thereby for 2 hydrogen bonded structures the energy per hydrogen bond is higher than for 4 bonded or tetrahedral coordinated structures. We discussed this in a few sentences in the Wernet et al. paper with references that I have underlined with yellow in the attachment. I also attach a recent review by Lars, Congcong and myself on water (A. Nilsson et al., J. Mol. Liq. 176, 2; 2012) where we discuss the importance to further develop simulations and in particular the importance of many-body effects in an electronic structure description where it is also essential that the latter includes van der Waals interactions.

Another interesting aspect in their study is the dynamics indicating that the strong and weak bonds switch place on ultrafast time scales. This is also in line with our previous discussions. I underline in the Leetma et al. paper (J. Chem. Phys. 129, 084502; 2008) how we discussed ultrafast measurements where the strongly H-bonded OH group in one water molecule could switch places with the weak one. The difference was only that we assumed that the switching occurred via librations whereas in their simulations it is mostly via translational motions of other molecules.

We had a last sentence in our Wernet et al. paper of a more speculative nature that with mostly only one strong donor and one strong acceptor bond per molecule, one-dimensional structures should appear which could be chains or rings. I noticed that they have a sentence regarding molecular chains at each instant connecting the strong bonds. Very nice.

Another point not related to the current paper that could also be of interest for you is the recent paper by Overduin and Patey (J. Phys. Chem. B 116, 12014; 2012) that discusses our PNAS Huang et al. paper based on simulations coming to similar conclusions regarding inhomogeneities in the liquid but only using a different language in terms of concentration fluctuations of two different classes of structures.

I think Kühne and Khaliullin put the picture in terms of a symmetry breaking in order to be not too far away from the most accepted tetrahedral picture of water. But if you really look at what they claim, which is also aligned with us, is that most molecules will be in an asymmetric position at all instances when taking a snapshot. It is simply that an OH group that is either weak or strong switches on a rather fast timescale. This would mean that the symmetric position is never really seen. It is only when you average over a long time that it looks like it is mostly in the tetrahedral position. Like me make analog. Take a pendulum that swings back and forth. The speed at the end points are close to zero and at the middle position it is highest. It means that the pendulum spends most time at the end points and extremely little time at the middle position. If you take the average position it will be the middle position but it is hardly ever visited. You could in such a picture claim that the equilibrium position is in the middle if you average over a single period but the pendulum will spend very little time there. The question is how the surrounding will be affected by such a motion. If I understand correctly the paper it is the surrounding that indeed infer the asymmetry. They claim that it is the translational motion of the other molecules that provides the mechanism of the switching. This should mean in my opinion that it is the end points of the asymmetric motion that makes the interaction with the surrounding and not the time averaged position. If it would have been dominated by a more internal motion such as librations it could have had less effect on the surrounding.

In water my own belief is that it is somewhat more complicated. Let me come to my current picture of water at the end of these comments.

I think it is currently not possible to observe the asymmetry based on the current status of pump-probe IR spectroscopy. Please don't quote me on this but I believe that we are still missing some major understanding about IR spectroscopy. The vibrational life time as a local oscillator in H20 is too short. Nearly all experiments on ultrafast dynamics are based on isotope substitution with HDO impurities in either H2O and D2O and measurements of the time resolved development of the OH or OD frequency. It is then assumed that HDO has equal probabilities for all molecular positions in the liquid. This assumption has never been proven. Based on our understanding of water in terms of fluctuations of two local structures, tetrahedral (low density water) and asymmetric (high density water) the latter has more contributions to the free energy through stronger bond energy whereas the latter has more entropy. I attached a slide from one of my presentations so you can see what I mean (hit a return in slide show mode to see the motion). There you also see the switching of the two bonds. Since the asymmetric configuration will provide more entropy and the HDO is already asymmetric it is likely that HDO will reside more in asymmetric configurations. Furthermore, we can anticipate through quantum effects that also the OD and OH hydrogen bonds are different. I would assume that OD will more likely be situated with the strong bond and OH more in the weak bond. This was observed in a recent PRL by Misha Bonn's group (attached) where they showed based on sum frequency generation spectroscopy and simulations that there is a strong preference for OH to point to the vacuum and OD to the liquid side for HDO at the water-air interface. At this point I think caution needs to be exercised regarding pump probe IR since these measurements might only be probing some fraction of all possible molecular motions. We (Lars) are currently further investigating the HDO dynamics based on quantum simulations. There are also recent indications that the pump pulse disturbs the dynamics and these measurements might not always represent the equilibrium motions.

We are currently using the new x-ray laser at Stanford, the Linac Coherent Light Source (LCLS) to various problems related to water. The machine provides completely new opportunities. We are planning a future step to open up dynamics. In particular x-ray correlation spectroscopy with x-ray lasers could provide new answers. This is probably 2 years away. Since the x-ray laser is fully coherent there will be a speckle pattern due to diffraction also from a disordered material. The plan is to make probe-probe measurement to thereby study equilibrium dynamics. You split the x-ray beam into two pulses with a controllable delay and the change in the speckle pattern will give information about how molecules have moved between the two pulses. These will be most challenging experiments but hopefully can be done and would open for completely new avenue's to probe dynamics in liquids. Naturally one challenge will be to not allow the first pulse to disturb the system but water is quite forgiving for hard x-rays since it is a low Z liquid.

We have demonstrated in a number of papers that this asymmetry can't been seen in diffraction since the data is not completely perfect. The difficulty is to have a technique that is only sensitive to the hydrogen bonding asymmetry around individual molecules. You can have a total hydrogen bonding average through a linear combination of water molecules with nearly no hydrogen bonds and fully tetrahedral water that on average would be like an asymmetric species. It is the rehybridization of the molecular orbital structure leading to local OH orbitals instead of delocalized H2O orbitals that makes XAS sensitive. I attach a paper where we discuss this (Nilsson section 3.5). It has to do with O2p and O2s hybridization and XAS only provides intensity for the O2p part in the orbitals. It is quite involved. Maybe it is time for a simpler review so many can understand these effects.

Here is our current picture of water. At high temperatures (close to boiling and above) water behaves as a simple liquid where most interactions are isotropic (dominated by van der Waals interactions). This is a structure where many molecules are in very disordered shells and without having a well defined first and second shell (this structure you get from ab initio MD simulations with van der Waals functionals even at ambient temperature, see discussion in my previous mail in paper "fluctuations in ambient water"). This type of non directional orientations gives high flexibility for various motions and high entropy. As the liquid cools down the water molecules starts to stick to each other through directional hydrogen bonds. This will appear in two classes of configurations, tetrahedral and asymmetric. In the tetrahedral structures each molecules are in 4 hydrogen bonds, this provides the lowest enthalpic energy. Since cooperatively effects makes the bonds stronger if water is bonded to other waters that are also in tetrahedral structures, these start to clump together in small local regions. Since the molecules are stuck with four bonds the motion is very restricted (see the attached slide) and thereby low entropy.

The other alternative is to form asymmetric structures. Here the hydrogen bond energy per molecule is higher than in the tetrahedral structures but since it is fewer bonds the enthalpic energy is lower. With less directional hydrogen bonds you have more flexibility for motions and thereby higher entropy (see slides). In this structure there are also non isotropic molecular interactions causing interstitial molecules. These are therefore called by us and others, preferable in the supercooled community, as high density water. Another way of viewing such species is that we start not with hexagonal ice but with high density amorphous or very high density amorphous ices where we have a large collapse of the second shell. In these ices we have also a local tetrahedral bonding but with other angles towards the second shell (often call interstitials). Here we induce asymmetric distortions around these tetrahedral bonded molecules in the first shell but simultaneous keeping the interstitials ( see fluctuations in ambient water). We have also seen this in water at an interface recently published in the Nature journal Scientific Reports (see attached Kaya et al.). There are fluctuations between the tetrahedral structures and asymmetric or high density structures. As we cool the liquid down the molecules in the asymmetric structures converts more and more into the tetrahedral structures which grows in size since the enthalpic energy contributes more to the free energy with decreasing temperature. The timescales in the fluctuations between these two classes should also slow down (not yet determined). There is also a continuous change in the asymmetric structures with temperature. The switching time is expected to slow down and thereby also the amplitude in the motion with decreasing temperature. We will be approaching more and more a local tetrahedral arrangement. However, not as in hexagonal ice but more towards high density ices with still many interstitials. That the density is decreasing below 4C is simply due to that we are at the same time converting many molecules into tetrahedral structures which have a more similar structure as hexagonal ice but disordered more towards low density amorphous ice (low density water) with the second shell at the tetrahedral angle causing an open network with low density. In my opinion we have to consider both these two classes of fluctuations (between tetrahedral and asymmetric and within asymmetric) separate but most likely there should also be some coupling.

It is these fluctuations between tetrahedral and asymmetric structures plus the varying fluctuations in the asymmetric structures that depends on temperature, pressure and interactions with solutes, interfaces, biomolecules etc. We can imagine that for instance an interface disfavors tetrahedral structures (such as in Kaya et al.) creating a dominance of asymmetric high density structures but through the interaction with the interface the switching time and amplitude will be affected between the strong and weak bonds.