Thursday, October 13, 2011

Lum-Chandler-Weeks under the microscope

Shekhar Garde and his colleagues have shown in several recent publications how hydration of solutes and its consequences, such as conformational changes in polymers, can be significantly altered near the air-water interface. The more general effect of interfaces on hydrophobic phenomena is now probed by Garde along with David Chandler, Amish Patel and others (A. Patel et al., PNAS doi 10.1073/pnas.1110703108 – paper here). They look at how interactions between hydrophobic solutes of various sizes from sub-nm to several nm are altered close to the interface of water with self-assembled monolayers of various surface chemistries, from hydrophobic to hydrophilic. They find that the driving force for the assembly of hydrophobic particles is smaller near a hydrophobic surface than it is in bulk, and decreases with increasing temperature, in contrast to the bulk (and to hydrophilic surfaces). This implies that hydrophobic surfaces should act as catalysts for the unfolding of proteins. It might also account for how chaperonins work: their initially hydrophobic surfaces help misfolded proteins to unfold, and then ATP-driven conversion of the walls to hydrophilic release the unfolded protein from the wall so that it might fold again in bulk.

Misfolding in the context of Lum-Chandler-Weeks theory is also the subject of a paper by Ruhong Zhou and colleagues (Z. Yang et al., J. Phys. Chem. B 115, 11137 (2011) – paper here). They have looked at whether dewetting transtions play a role in the assembly of two amyloidogenic beta-sheets. For the peptides considered there are two conformations that lead to stacking of beta-sheet pairs: one containing water between the sheets in a 2D slab-like geometry, the other with the water in a 1D tube-like geometry. In both cases the sheets are brought together by a drying transition, but surprisingly this is stronger for the slab-like than the tube-like case. This is attributed here to the different surface roughness of the two packing modes: the ‘staggered’ packing in the slab-like case, which is rougher, disrupts the H-bonding network of the intervening water to a greater extent.

Some further new insights into the Lum-Chandler-Weeks dewetting mechanism for hydrophobic interactions are offered by two recent papers. Li and Walker appear to see the cross-over region between the mechanisms of hydrophobic hydration at small (<1 nm) and large (>1 nm) scales that this theory predicts, by measuring the free energy of hydration of individual monomers of various size in hydrophobic polymers, using an AFM to pull the chains out from a collapsed conformation (Li, I. T. S. & Walker, G. C. PNAS 108, 16527-16532; 2011 – paper here). At a monomer size of around 1 nm in the temperature region of 50 C or so, they find a switch in the hydration entropy from negative to positive (this crossover size reduces to just 3.5 Å at 150 C). Shekhar Garde and Amish Patel have published a commentary on it (10.1073/pnas.1113256108).

And Garde and Patel have joined forces with David Chandler and others in a preprint ( – paper here) that attempts to unravel the issue of why some protein subunits (such as melittin) seem to aggregate via dewetting while others (such as BphC) do not. They find that at model (SAM) hydrophobic surfaces, simulations show that the statistics of large-amplitude fluctuations in the density of interfacial water are altered relative to the Gaussian stats of both bulk water and water at a hydrophilic surface. In other words, despite similar average interfacial water densities, the fluctuations reveal the proximity of the hydrophobic interface to a dewetting transition. This tuning of the interfacial water is, they argue, common to biological systems, where it induces a strong sensitivity to small changes in conformation, allowing the system to take advantage of the phase transition in engineering biomolecular function (in a manner analogous to the finely balanced wetting or drying of ion channels for ‘vapour-lock’ gating). Although melittin and BphC lie on opposite sides of this transition, small modifications to both can tip the balance one way or the other. This offers what seems to me to be a persuasive argument that dewetting is relevant to hydrophobic aggregation even if it does not exactly provide the mechanism for it in all (or even in most) cases: the transition is, if you like, ‘there’ even if it doesn’t manifest itself.

Alenka Luzar and her colleagues have considered this issue from the perspective of whether or not cavitation can take place at the protein-protein interface (J. Wang et al., PCCP 10.1039/c1cp22082a – paper here). They present a lattice model which offers a fast method for predicting if cavitation can happen, and find that part of the surface of melittin is sufficiently hydrophobic to permit this on a timescale that is consistent with that seen in the earlier simulations.

Alenka and her colleagues have also examined how this putative crossover length-scale for hydration behaviour is influenced by charge on the solute (J. Stat. Phys. 10.1007/s10955-011-0337-1 – paper here). They conclude that, for moderate charge, the electrostatic contribution to the solvation free energy is in fact essentially independent of solute curvature, because of a compensation between counterion shielding and the dielectric screening of water – the solvation free energy remains more or less a function only of solute surface area.

To what extent protein-ligand binding requires an atomistic description of changes in hydration is a crucial question, not least for attempts to design synthetic ligands in drug development. Ulf Ryde at Lund University and colleagues have looked at this issue by comparing the ability of continuum methods to predict binding free energies for four different protein-ligand pairs with quite different degrees of solvent exposure at the binding site (S. Genheden et al., JACS ja202972m – paper here). They find that the continuum methods often perform badly, particularly for cases with a greater degree of solvent exposure. We need to know precisely where the waters are and where they go.

And that is somewhat elucidated by Nan-jie Deng at Rutgers University and colleagues for the case of two synthetic inhibitors of HIV-1 protease, Nelfinavir and Amprenavir (N.-j. Deng et al., J. Phys. Chem. B jp204047b – paper here). The binding of these drugs is apparently entropically driven, but the question is where that entropic contribution comes from. A classical view would be inclined to attribute it to the release of water from the binding cleft, but it seems that this isn’t so: these MD simulations suggest that any entropy gain there is more than offset by the restriction of ligand rotation and vibration on binding. Instead, the favourable entropic contribution seems to come from desolvation of the ligand.

A different view of water’s influence on bimolecular recognition is provided by Stacey Wetmore and colleagues at the University of Lethbridge in Alberta (F. M. V. Leavens et al., J. Phys. Chem. B jp205424z – paper here). They have looked at how water molecules can affect the pi-pi interactions between DNA and DNA-binding proteins. Solvating water molecules seem to have essentially no influence on the strength of pi-pi stacking for histidine and adenine, but do weaken the interaction if the histidine is protonated. The latter interaction, however, remains in all cases stronger than the former.

Protons seem to be delivered to membrane proteins such as the proton pump cytochrome c oxidase via some kind of surface-enhanced, two-dimensional transport at the membrane surface. This has been previously postulated as a series of jumps between ionisable groups (phosphate and carbonyl) at the membrane surface. But that notion is challenged by Peter Pohl at the Johannes Kepler University in Linz and colleagues, whose fluorescence measurements of proton transfer at membranes show that proton transport can be equally fast in the absence of ionisable groups (A. Springer et al., PNAS 108, 14461-14466; 2011 – paper here). They conclude that it is probably the network of interfacial water molecules that is responsible instead for the rapid proton motion: as they say, “water structuring at the interface seems to be mandatory for providing the pathway”.

It has of course been long thought that protons may be delivered to the interior of an enzyme via chains of water molecules. That process is studied by heme peroxidase by Emma Lloyd Raven and coworkers at the University of Leicester (I. Efimov et al., JACS ja2007017 – paper here). Kinetic isotope effects reveal that the proton pathway utilizes a Grotthus-like shuttling of protons along a pathway towards the ferryl oxygen that involves three bound waters and two arginine residues.

A model system for studying such proton transfer in confined geometry is reported by Bradley Habenicht and Stephen Paddison at the University of Tennessee in Knoxville (J. Phys. Chem. B jp205787f – paper here). They use MD simulations to look at how protons are transported within carbon nanotubes whose inner walls are functionalized with perfluoro sulphonic acid groups. If these groups are spaced far apart (~8 Å), they tend to be individually hydrated by clusters of water molecules with little interaction between them, and correspondingly reduced acidic proton dissociation. But there is also a pronounced effect of confinement at small nanotube diameters: in the smaller tubes there are stronger interactions between the walls and the water molecules, which can lead to break-up of the hydrogen-bonded network of waters linking the sulphonic acid groups. That network may be restored by polarized charges of fluorine atoms attached to the nanotube walls.

ATP hydrolysis in the active cleft of actin plays an important role in the state of its filamentous form F-actin, affecting its rigidity and its binding of regulatory proteins. There is water in this active site, but it hasn’t previously been clear what, if anything, it does. Marissa Saunders and Greg Voth at Chicago have clarified this through MS simulations based on the crystal structure (J. Mol. Biol. 413, 279-291; 2011 – paper here). They say that the ordered waters help the protein to flatten and brings about a conformational change that promotes ATP hydrolysis. These changes also stabilize the charge on the phosphate and accelerate the deprotonation of the catalytic water involved in hydrolysis. In short, the bound water helps to organize the active-site geometry.

Another nail in the coffin of ‘structure-making and –breaking’: Fabio Bruni and colleagues in Rome have looked at how local solvent structure around ions affects their influence on viscosity – specifically, on how changes in ionic concentration affect the viscosity of the solution (T. Corridoni et al., J. Phys. Chem. B jp202755u – paper here). Classically, the viscosity is found to be (almost) linearly related to the concentration. It has been asserted that the magnitude of the coefficient of proportionality, denoted B, depends on how the ions perturb the water structure. Fabio et al. use neutron scattering and simulations to look for some structural parameter that can be correlated with B. They find that the nature of the (univalent) ions is all but irrelevant to the size of the percolating water clusters in solution. The change in viscosity seems to be unrelated to any structural changes in the bulk liquid, but instead pertain to changes in the local hydration shells of the ions. As a result, they say, “the particular effect of solutes ranked in the Hofmeister series must be looked at in terms of specific ion interactions with hydrophilic or hydrophobic surfaces”, and not in terms of any generalized propensity for structure-making or –breaking.

In an intriguing preprint, Jampa Maruthi Pradeep Kanth and Ramesh Anishetty at the Institute of Mathematical Sciences in Chennai propose that the hydrophobic interaction should be understood as an effect analogous to the Casimir effect (the attraction of two surfaces separated by a vacuum due to the suppression of long-wavelength electromagnetic fluctuations of the vacuum in the gap) (preprint here). Their molecular mean-field analytical method suggests that confinement alters the allowed fluctuations of the hydrogen-bond network, specifically the long-ranged correlations between water molecular orientations. Now what I’d like to know is whether an explicit connection can be made to the alleged role of fluctuations in the Lum-Chandler-Weeks model. But that seems to argue in the opposite direction, namely that fluctuations are actually enhanced in the gap owing to the destabilizing influence of the hydrophobic surfaces on the intervening water layer. It’s not clear to me whether in Kanth and Anishetty the water in the gap is, aside from the suppression of fluctuations, any different from bulk water, except perhaps for the monolayer adjacent to the surfaces…? And why would this not work for hydrophilic confinement too?

Alenka Luzar and her colleagues Christopher Daub and Dusan Bratko have just published a review of how electric fields at interfaces can modify their wettability (Top. Curr. Chem. 10.1007/128_2011_188; 2011 – paper here). Effects of this nature may play a role in the behaviour of voltage-sensitive ion channels.

Still more on denaturants. Thomas Record and colleagues at Wisconsin ask why urea is a denaturant while glycine betaine is a protein stabilizer (E. J. Guinn et al., PNAS 108, 16932-16937; 2011 – paper here). They use osmomentry and solubility measurements to look at the interactions of these molecules with 45 model proteins, and conclude that the explanation for the different behaviours lies with the details of how and where the molecules interact with the peptide surfaces. For example, urea accumulates at amide O groups, and to a lesser extent at aliphatic carbon atoms, whereas glycine betaine is excluded from them. This adds further weight to the notion that such osmolytes exert their effects via direct interactions with proteins rather than any generalized influence on ‘water structure’.

Phosphate groups turn out to be a sensitive probe of electric fields, including those that can be induced by hydration. As Steven Boxer and colleagues at Stanford show (N. M. Levinson et al., JACS 133, 13236; 2011 – paper here), electric fields perturb the vibrational spectra of organophosphates in a way that can reveal changes in hydration within partially hydrated environments, such as the active sites of enzymes.

Is there a liquid-liquid transition in confined water? That question is investigated via MD simulations by Limei Xu and Valeria Molinero at Utah (J. Phys. Chem. B jp205045k – paper here). The possibility has been raised by simulations of water in slit-like pores 2.4 nm wide (Brovchenko & Oleinikova, J. Chem. Phys. 126, 214701; 2007). Valeria and Limei use the mW water model to look at water’s behaviour in 1.5-nm hydrophilic pores at a range of temperatures and pressures up to 4000 atm. They find that at high pressures there is a signature of a somewhat abrupt but nonetheless continuous phase transition in the supercooled regime which could be interpreted as a ‘shadow’ of a L-L transition in the bulk phase which cannot itself be accessed.

A new and unusual view of the protein dynamical transition at c.200-220 K is presented in a preprint by Andrei Krokhotin and Antti Niemi at Uppsala University (paper here). They say that it this transition can be regarded as an analogue of the transition of a high-temperature superconductor to a non-superconducting pseudo-gap state. In other words, proteins can be assigned an order parameter formally equivalent to the quasiparticle wave function of superconductors. It’s not clear to me how/if this description modifies what is known already about the transition (on which, and on the role that hydration plays, there seems still to be no real consensus), but it is an original idea.

Wednesday, August 17, 2011

How hydration forces assemble protein aggregates

Dielectric spectroscopy has been widely used to study the dynamics of proteins and their hydration shells, but is somewhat hampered by ambiguities about the origin of the atomistic motions that contribute to the signals. Sheila Khodadadi at NIST and colleagues have now shown just how fraught a business this is. The dielectric response of hydrated proteins typically contains two components – a fast (20-50 ps) relaxation, thought to be due to hydration water (which is slowed down relative to the bulk), and a slow (0.5-10 ns) relaxation ascribed to ‘tightly bound water’. Khodadadi and colleagues have previously questioned the former assignment (J. Phys. Chem. B 112, 14273; 2008), and now they question the latter too (J. Phys. Chem. B 115, 6222; 2011 – paper here). They say that comparisons of MD simulations to dielectric spectroscopic data suggest that the nanosecond relaxation is in fact due to motions of the protein atoms themselves, while the hydration water relaxes much more quickly.

Cytochrome c oxidase (CcO) is one of the classic examples of an enzyme that uses embedded water wires for proton transport. But how does it avoid back transport of protons along the same route? One proposal is that the water wire becomes reoriented to prevent this. Another is that a side-chain element of the hydrogen-bonded network, Glu-286, rotates to break the chain. Shuo Yang and Qiang Cui of the University of Wisconsin now show that neither process seems to be a viable gating mechanism (Biophys. J. 101, 61; 2011 – paper here). They suspect that the vectorial proton transport is probably achieved instead in a more subtle way, by stabilization of the proton-transfer transition state by the charge distribution around the active site, to explore which will require sophisticated quantum-chemical modelling.

Bacteriorhodopsin is of course another archetype for the functional water wire. It has been claimed by Klaus Gerwert at Bochum and others that the proton pumped through the protein by light absorption is stored in a water cluster inside the channel. This idea is challenged in another paper by Cui and coworkers, soon to appear in JACS (P. Goyal et al., JACS ja201568s – paper here). They say that for the proton to be stored on water requires an implausibly high increase in the pKa of a hydronium, and argue that the stored proton is instead shared between Glu194 and Glu204. One can be sure this is not going to be the last word on the matter…

I missed an interesting paper last year by Ariel Fernández and colleagues on how proteins can ‘seal’ backbone hydrogen bonds (BHBs) at their surface against competition from hydration water by a close control of the local curvature: if the BHBs are in a sufficiently highly curved location, water cannot penetrate without compromising its own H-bonded network for purely geometric reasons (E. Schultz et al., PLoS ONE 5, e12844; 2010 – paper here). Misfolded proteins fail adequately to protect their BHBs this way. This work is rather closely related to the recent paper by Fernández and Michael Lynch at Indiana on how BHBs that are ‘poorly wrapped’ by hydrophobic groups can become sites of protein-protein aggregation, which lead to complexification of the ‘interactome’ due to the degradation of wrapping created by genetic random drift (Nature 474, 502; 2011 - paper here). I have discussed that work in the August issue of Chemistry World (here – requires a subscription).

Somewhat relevant to this work is a study of protein ‘hot spots’ where preferential binding of organic small-molecule probes bind, by Frank Guarnieri of Virginia Commonwealth University in Richmond and colleagues (J. L. Kulp III et al., JACS 133, 10740; 2011 – paper here). They use simulated annealing to look at the interactions of eight probe molecules, and water, with hen egg lysozyme, and find that all the probes bind in the known binding site, which is evidently the ‘hot spot’. This implies that ligand binding isn’t simply a case of molecular recognition – the ligand is already guided towards that site by the fact that other potential binding spots are ‘guarded’ by tightly bound water. The researchers say that these hot spots can therefore also act as sites for potential protein-protein binding.

Coming back to the previous paper, Ariel also believes that poor wrapping of BHBs is implicated in the aggregation of proteins in amyloid diseases – and that these are thus potentially an alarming result of evolution’s failure to deal with random drift in anything more than an ad hoc manner. The question of how hydration is related to amyloid aggregation is also considered in a preprint by Dave Thirumalai and colleagues (arxiv 1107.4820 – paper here). They argue that “water controls the self-assembly of higher-order structures”, just as the expulsion of water is a key stage in the interaction between peptides forming beta-sheet protofilaments. Specifically, water accelerates the formation of fibrils from mainly hydrophobic peptides, but slows down the aggregation of hydrophobic sequences by stabilizing them. In the higher-order hydrophobic structures, trapped water can be considered to be analogous to water confined within carbon nanotubes – in which it seems that there is a sensitive dependence of the water structure on the width of the confining region. Thus the work nicely connects studies of confined water to an important biological problem.

A different picture – I’ve not quite figured out if it competes with or complements the above picture – of self-assembly of protein fibrils is presented by Krishnakumar Ravikumar and Wonmuk Hwang at Texas A&M (JACS 133, 11766; 2011 – paper here). They compare the role of hydration forces in the self-assembly of hydrated, dry nonpolar and dry polar surfaces, represented by collagen and two amyloidic peptides, the former a ‘steric zipper’ with interdigitating hydrophobic side-chain groups. In all cases the surfaces have hydration shells, and an oscillating hydration force as the surfaces come together due to coalescence and depletion of the hydration shells. What differs is the magnitude of these. Thus the interactions have a common origin, being water-mediated in all cases. I can’t improve on what the authors say here about the role of the hydration shells: “Thus, designating a protein surface as either “hydrophobic” or “hydrophilic” may be too simplistic of a dichotomy, as surfaces in reality lie between these idealized limits. There should be no fundamental difference in the way hydration forces arise among different types of protein surfaces, even with varying affinity for water.”

The idea that water can support two liquid states in the high-pressure supercooled regime has not been directly verified experimentally, but has received enough indirect support from both experiments and simulations to have gained wide acceptance. So a preprint from David Limmer and David Chandler at Berkeley making strong claims for the non-existence of the two liquid states is bound to cause waves (arxiv 1107.0337 – paper here). On the basis of calculations using Valeria Molinero’s recently reported mW model of water, they assert that the signatures of a phase transition seen by others relate to a solid-liquid transition, or artifacts due to imperfect equilibration (which is extremely slow in this region for, e.g. ST2 water), and that there is no sign of two liquid basins anywhere in this part of the free-energy landscape. This work, and the topic in general, will be debated in a water mini-session at the Mini Stat Mech meeting at Berkeley in January (see here), which promises to be lively.

Majed Chergui at EPFL in Lausanne and colleagues have described a nice study of the time-resolved ultrafast changes in hydration of an iodide ion as it is transformed to neutral iodine by electron abstraction, probed using XAS (V.-T. Pham et al., JACS ja203882y – paper here). This transformation changes the solute from hydrophilic to hydrophobic, and so entails a considerable reorganization of the hydration shell. The hydrogens are reoriented from pointing towards the solute to pointing away from it, forming a hydrogen-bonded ‘hydrophobic cavity’. This reorganization process takes about 3-4 ps.

Achim Müller at Bielefeld and colleagues have developed a new way to study hydrophobic species encapsulated in genuinely hydrophobic cavities. They make hollow capsules from molybdenum-based polyoxometalates, with hydrophobic interior surfaces due to ligands coordinated there (C. Schäffer et al., Chem. Eur. J. 10.1002/chem.201101454 – paper here). Much less water is found inside these shells than in those with hydrophilic interior surfaces. And hydrophobic entities such as n-hexanol are spontaneously taken up into the porous shells.

Gaurav Chopra and Michael Levitt at Stanford have used state-of-the-art quantum chemical methods to map out the hydration shell of C60 (PNAS 10.1073/pnas.1110626108 – paper here). These reveal dramatic ordering of the surrounding water, evident in the time-averaged azimuthal distribution function, which is not seen when empirical force fields are used for the calculation. Moreover, ordering of the water molecules is evident as far out as about 1 nm from the C60’s centre (as I recall, the molecule is itself about 4 Å in radius). Significantly, the azimuthal water ordering around the buckyball is not due to some kind of ice-like clathration of a non-polar species but to strong dispersion attractions between the solute and solvent molecules themselves. I guess one thing this study illustrates is the sensitivity of the important details of the hydration structure of hydrophobes to the interaction potentials used.

I think I somehow failed previously to comment on an important and provocative paper by Roy Daniel and colleagues published towards the end of last year (M. Lopez et al., Biophys. J. 99, L62; 2010 – paper here). They report significant catalytic activity of pig liver esterase in near-anhydrous conditions – not merely in a non-aqueous medium, but in a ‘dry’ powder with just 3±2 water molecules per enzyme molecule. Let me quote from the abstract: “This indicates that neither hydration water nor fast anharmonic dynamics are required for catalysis by this enzyme, implying that one of the biological requirements of water may lie with its role as a diffusion medium rather than any of its more specific properties.” They admit, naturally, that it remains an open question whether this result can be generalized to all enzymes – this one might just happen to be particularly rigid. All the same, this is most definitely a finding to chew on.

Jeremy England’s paper on allostery and hydrophobic burial, mentioned in an earlier post, has now been published in Structure (paper here). To save you a click, here’s what I said before:
The paper shows that it is possible to estimate low-energy conformational changes in a protein, such as those involved in allosteric effects, on the basis simply of residue-by-residue hydrophobic effects. Specifically, he develops a method for determining the most energetically favourable way of burying hydrophobic residues, given a particular amino-acid sequence. This amounts to identifying the particular ‘burial modes’ of any given sequence. Thus, although the stabilities of conformations are doubtless multifactorial, hydrophobicity seems to be the major governing factor.

Thursday, July 7, 2011

Stabilizing and destabilizing proteins

How do osmolytes stabilize proteins against denaturation? Using mechanically induced unfolding by means of atomic-force microscopy of protein I27 (in fact 8 repeat units in a polyprotein), Julio Fernández and colleagues at Chicago find that osmolytes can act as ‘solvent bridges’ during the unfolding transition, pinning together hydrogen-bonding sites on the polypeptide backbone (L. Dougan et al., PNAS 108, 9759; 2011 – paper here). They found in earlier studies that water molecules generally act in this capacity. They now investigate this process in the presence of glycerol, which is known to enhance protein stability, and related solutes. They find that while glycerol, ethylene glycol and propylene glycol all enhance the mechanical stability, larger hydrogen-bondong osmolytes such as sucrose and sorbitol do so to a far lesser extent – MD simulations suggest that the latter are unable to penetrate the folded structure and act as stabilizing bridges, and any stabilizing influence they exert must be indirect.

The effects of methanol on protein conformational stability are even more complex, to judge from the NMR and MD results of Christian Hilty of Texas A&M and coworkers (S. Hwang et al., J. Phys. Chem. B 115, 6653; 2011 – paper here). They report that the alcohol stabilizes the secondary structure by strengthening hydrogen-bonding interactions in the backbone (a consequence of partially eliminating the water molecules that compete with these interactions), while also weakening hydrophobic interactions and thus swelling and loosening the folded structure overall. So whether methanol tightens or loosens the structure overall is a complex balance that depends on the sequence and initial structure of the protein.

Urea and guanidinium both help to denature the model protein the Trp-cage, according to the experiments and simulations of Pavel Jungwirth in Prague and colleagues (J. Heyda et al., J. Phys. Chem. B 10.1021/jp200790h – paper here). Although these two small molecules are chemically quite different, it appears that their denaturing mechanisms here are essentially the same, involving first the destabilizing displacement (positional exchange) of two proline residues within the hydrophobic core, followed by a gradual unravelling of secondary structural elements.

The conformational stability of proteins is also influenced by ions: Hofmeister ‘salting out’ agents such as the strongly solvated, high-charge carbonate and sulphate ions also stabilize the folded state, whereas weakly solvated low-charge ions such as bromide and iodide promote denaturation. This tendency overlaps with the classical denaturation properties of complex ions such as guanidinium (Gdm+). Christopher Dempsey at Bristol and colleagues have studied these effects for the cases of the sulphates and chlorides of guanidinium and tetrapropylammonium (TPA+) (C. Dempsey et al., JACS 133, 7300; 2011 – paper here). They find that the question of conformational stabilization is subtle. TPA+, like Gdm+, perturbs the stability of some proteins (specifically the tryptophan zipper trizip, a beta-hairpin peptide), but stabilizes the alpha-helices of alahel peptides, apparently because TPA+ cannot compete effectively for hydrogen bonds in H-bond-stabilized conformations. Moreover, the cation effects may be modified by the anion: sulphate counteracts the denaturing effects of Gdm+ on trizip, but has no effect on the influence of TPA+ in that case, because Gdm+ but not TPA+ forms ion pairs with sulphate. The emerging picture is thus one in which the Hofmeister-like effects of ions must be understood in the light of a detailed consideration of (i) ion hydration; (ii) anion-cation interactions; and (iii) direct ion-protein interactions.

More on Hofmeister from Corinne Gibb and Bruce Gibb of the University of New Orleans, who offer a new perspective: namely, that so-called chaotropic anions in fact display bind preferentially to concave hydrophobic surfaces, thus effectively weakening hydrophobic attraction and promoting ‘salting in’ (JACS 133, 7344; 2011 – paper here). They reach this conclusion by looking at the thermodynamics of ion effects in the binding of adamantane carboxylic acid within the deep hydrophobic cavity of a cavitand – an interaction perturbed by anions binding to the cavity. Sure, the extension to proteins is purely by analogy, but it’s an intriguing take on the old problem. Meanwhile, Oleg Krasilnikov and colleague at the Federal University of Pernambuco in Brazil have considered how ions affect molecular interactions in nanopores, and find that these too seem to display a Hofmeister-like sequence of activities (C. G. Rodrigues et al., Biophys. J. 100, 2929; 2011 – paper here). They look at how simple ions alter the rate constant for the interaction of poly(ethylene glycol) with the protein pore alpha-hemolysin, and see a change consistent with the Hofmeister series for halide anions. They suggest that this results from a competition for hydration water between the ions and other solutes within the pore.

Protein aggregation is commonly suppressed by the addition of surfactants and sugars or polyols. But arginine hydrochloride has also been found to possess this capability, and unlike some conventional aggregation-suppressors it does so by reducing protein-protein interactions without seeming to affect the stability of the folded conformation. This behaviour hasn’t been fully explained, although Bernhardt Trout and colleagues at MIT have proposed that at least part of the mechanism might be non-specific entropic effects due to the exclusion of arginine from the gap between two proteins as they come together (B. M. Baynes et al., Biochemistry 44, 4919; 2005). They now refine this picture by looking at the influence of the anionic counterion, showing that these display the usual Hofmeister progression for aggregation suppression (C. P. Schneider et al., J. Phys. Chem. B 10.1021/jp111920y – paper here). The protonated arginine contains a guanidinium group, again potentially making the connection to denaturant activity – but the relevant behaviour here seems in fact to be the ability of the arginine ions to self-associate in stacks. Thus, while arginine can, like Gdm+, bind directly to the protein surface, the self-association weakens this interaction so that it does not cause denaturation yet still weakens protein-protein interactions. Schneider et al. elucidate this process further by looking, both experimentally and computationally, at the anion effects. Some anions, such as sulphate, phosphate and citrate, can enhance the cationic clustering by forming multiple hydrogen bonds, creating larger clusters and thus stronger exclusion effects on aggregation – and perhaps also slowing protein diffusion via an enhancement of solvent viscosity.

Understanding how bacteria void toxic substances from the cell interior could have a profound impact on our ability to combat antibiotic resistance. E. coli have a multi-drug efflux pump called AcrAB-TolC, in which the AcrB protein in the inner membrane binds drugs non-specifically and pumps them to the TolC exit duct. Attilio Vargiu of the University of Cagliari and colleagues have taken a close look at how this pump works, with a particular focus on the role of water molecules in carrying the extruded substrate along (R. Schulz et al., J. Phys. Chem. B 10.1021/jp200996x – paper here). The protein has several small holes that allow water molecules to enter and flow in a directional manner. This water acts as a lubricant and transport medium for the drug, but also flattens out the electrostatic profile in the channel that might otherwise cause the drugs to get stuck (perhaps by hydrogen bonding), and thus it contributes to the polyspecificity of the mechanism.

The role of water-mediated interactions in protein-substrate binding and associated drug design has been given a fair bit of attention, but W. David Wilson of Georgia State University and colleagues show that such things may be relevant to small-molecule DNA-binding agents too. They look at the binding of the synthetic molecule DB921 into the AT-rich minor groove of DNA, an interaction that might be useful for the disruption of parasite mitochondria (Y. Liu et al., JACS 10.1021/ja202006u – paper here). The binding is mediated by a water molecule, and to better understand how this works the researchers look at the effect on the structure, kinetics and thermodynamics of binding of introducing a host of modifications to DB291. This information, in particular the characteristics responsible for the water-mediated interaction, could be valuable for designing new agents that bind strongly in the minor groove in a sequence-specific manner.

Combining vibrational sum-frequency measurements with simulations containing three-body terms, James Skinner and colleagues at the University of Wisconsin say that the liquid-vapour interface of water shows no evidence of ‘enhanced’ molecular structuring such as ice-like ordering (P. A. Pieniazak et al., JACS 10.1021/ja2026695 – paper here).

The charge and pH of this interface have recently become contentious issues. Sylvie Roke at the MPI for Metals Research in Stuttgart and colleagues now offer a profile of the oil-water interface, using zeta-potential and sum-frequency scattering measurements alongside MD simulations (R. Vácha et al., JACS 10.1021/ja202081x – paper here). They say that the water orientations at the surface are like those at a negatively charged surface, even though there is no hydroxide absorption to make it so. There is nonetheless a surface charge, which comes instead from a disturbance in the balance of hydrogen-bond donors and acceptors at the interface.

The notion that the hydration water of proteins is dynamically coupled to the protein itself receives more support from a study by Nguyen Quang Vinh and colleagues at UCSB, who have used terahertz spectroscopy to look at the large-scale collective vibrations of lysozyme (N. Q. Vinh et al., JACS 133, 8942; 2011 – paper here). They find that the protein is surrounded by 150-180 water molecules (a sub-monolayer) that, in the authors’ words, ‘in terms of their picosecond dynamics behave as if they are an integral part of the protein’. THz spectroscopy seems to be emerging as a nigh-incomparable technique for probing these long-ranged collective motions.

The dynamics of hydration water are also studied by Cesare Cametti at the University of Rome ‘La Sapienza’ and colleagues, using dielectic spectroscopy (C. Cametti et al., J. Phys. Chem. B 10.1021/jp2019389 – paper here). They find that the dielectric relaxation of the lysozyme hydration sphere is bimodal at high concentrations, corresponding to tightly and loosely bound waters, although monomodal in dilute solution.

The nature of the protein dynamical transition at 200-220 K shows no sign of being resolved. Salvatore Magazù and colleagues at the University of Messina now throw a cat among the pigeons by suggesting that there is no such transition at all: it is an artefact caused by the coincidence of the system’s relaxation time with the instrumental resolution (S. Magazù et al., J. Phys. Chem. B 115, 7736; 2011 – paper here). Nonetheless, they say – can life really be this complicated? – there is a crossover of some sort at 220 K, at least for the case of lysozyme considered here, for this marks a change from Arrhenius to super-Arrhenius behaviour in the coupled hydration-water/protein motions. I think what this implies – it is a little unclear – is that there is no intrinsic, qualitative change in the protein dynamics at the ‘dynamical transition’, but rather, a change due to the coupling of these to the hydration water dynamics. I could be wrong.

Water rotational dynamics are slowed down around many small solutes, both hydrophilic and hydrophobic. So it stands to reason that the same should happen for larger solutes, such as proteins. And it does – but Ana Vila Verde and R. Kramer Campen at the FOM Institute in Amsterdam suggest that the latter is not necessarily just a straightforward extension of the former (J. Phys. Chem. B 10.1021/jp112178c – paper here). Their simulations of water dynamics around the disaccharides kojibiose and trehalose show that the sheer size of these solutes, relative to smaller ones, induces additional mechanisms of water retardation as a result of topological constraints on water motions. Thus, one can’t for example imagine that the hydration of a free amino acid is the same as that when the amino acid represents a peptide residue.

Bear that in mind, perhaps, in considering the hydration structure of glycine as deduced from ab initio calculations by Bo Liu at Henan University and colleagues. They build up the hydration shell molecule by molecule from a gas-phase picture (Y. Yao et al., J. Phys. Chem. B 115, 6213; 2011 – paper here).

Salt bridges play a role in stabilizing the glycosyl hydrolase (an enzyme with potentially important industrial applications) of the hyperthermophile Rhodothermus marinus (L. Bleicher et al., J. Phys. Chem. B 115, 7940; 2011 – paper here). But perhaps surprisingly, the enzyme also contains salt bridges that seem to be destabilizing: located in the hydrophobic core, where they might facilitate the permeation of water.

C. Preston Moon and Karen Fleming at Johns Hopkins have a neat idea for developing a hydrophobicity scale for amino acids based on their energy of transfer from water to within a phospholipid bilayer, thus relating the measure directly to the driving forces for the assembly and stabilization of membrane proteins (PNAS 108, 10174; 2011 – paper here). Their thermodynamics measurements show some differences from the predictions of simulations, especially for the translocation of arginines – which makes the results relevant to voltage-sensitive ion channel gating mechanisms, since these involve the movement of arginines into the hydrophobic interior of the membrane.

How homogeneous are concentrated aqueous solutions? This question has been studied for methanol, which seems to aggregate to some extent in water; now Lorna Dougan at the University of Leeds and colleagues study the case of glycerol using neutron scattering, motivated by the relevance to cryoprotection (J. J. Towey et al., J. Phys. Chem. B 115, 7799; 2011 – paper here). They find that glycerol-glycerol hydrogen-bonding in the pure liquid is scarcely affected by the addition of water, while the water-water bonding is highly disrupted. In effect, the waters become isolated from one another, binding preferentially to glycerols. Thus, it seems likely that glycerol could act as a cryoprotectant by keeping water molecules apart and preventing for the formation of an ice network.

Water can penetrate the hydrophobic interior of carbon nanotubes, a fact that is being investigated for potential desalination technologies among other things. William Goddard at Caltech and colleagues take a look at what drives the filling process for different tube diameters (T. A. Pascal et al., PNAS pnas.1108073108 – paper here). For CNTs between 0.8 and 2.7 nm, the interior water phase is always more stable than the bulk, but for several different reasons. For nanotubes thinner than 1 nm, the water phase is gas-like, with an entropic driving force. For nanotubes of 1.1-1.2 nm the encapsulated phase is ice-like and enthalpy-stabilized. For nanotubes wider than 1.4 nm the interior phase is liquid-like, but stabilized by increased translational energy. The overall message is sobering in showing how very fine adjustments to the hydrogen-bonded network and the balance of interactions with the surrounding environment can create quite different phase behaviour and thermodynamic driving forces in confined situations even for very small differences in dimensions.

A new method for rapidly calculating solvation energies in water is presented by Jianzhong Wu and coworkers at the University of California at Riverside (S. Zhao et al., J. Phys. Chem. B 10.1021/jp201949k – paper here), which combines DFT with MD simulations. The researchers have so far tested it for simple ions, for which it works well. Meanwhile, Merchant and Dilip Asthagiri at JHU have a preprint (arxiv 1106.0448 – paper here) in which they examine the range of ion-specific effects in water and conclude that, at least for sodium, potassium, chloride and fluoride, these extend no more than about 4 Å, so not much more than the size of a single water molecule.

This hasn’t yet exhausted my list of papers: still to come (soon) are developments on the putative liquid-liquid transition of water and on proton transport in bacteriorhodopsin…

Tuesday, May 24, 2011

Hydration of PSII

What a beautiful crystal structure is reported by Yasufumi Umena of Osaka City University and colleagues in Nature (473, 55; 2011 – paper here). They have revealed the entire photosystem II complex at 1.9 Å resolution, showing the hydration structure of the core Mn4CaO5 cluster along with the locations of 1,300 water molecules in a hydration shell that seems to offer several hydrogen-bonded channels for protons, water molecules for photolysis, or oxygen molecules. The latter are probably formed by oxidation of some of the four waters bound to the Mn cluster. I don’t recall ever seeing before such a complex, orchestrated and carefully rationalized example of a multifunctional hydration structure.

Collagen is generally considered a structural protein, but particular collagen motifs are also recognized as substrates by collagen binding proteins such as adhesins in some human pathogens. On the basis of MD simulations and comparison with crystal structures, Luigi Vitagliano at the Consiglio Nazionale delle Ricerce in Naples and colleagues say that the regions of collagen-binding adhesin CNA involved in binding to hydrophobic parts of the collagen triple helix are inherently prone to dewetting, making them primed to bind their target by reducing the desolvation penalty (Vitagliano et al., Biophys. J. 100, 2253-2261; 2011 – paper here). Moreover, besides these hydrophobic contacts the interaction between CNA and collagen is mediated by an intricate network of 13 water molecules.

Local water densities around biological systems can be calculated fairly accurately with a computationally cheap interaction-site model (based on the Ornstein-Zernicke integral equation), rather than with a full MD simulation, according to a new study by Vijay Pande at Stanford and colleagues, who compare the two for the hydration of GroEL (M. C. Stumpe et al., J. Phys. Chem. B 115, 319; 2011 – paper here).

In drug design, the surfaces of target proteins are often mapped out to look for ‘hot spots’ where it is particularly advantageous to place functional groups in the ligand complementary to those on the protein. This process often identifies many such local minima in a rugged potential-energy surface. But that’s in vacuo for a rigid surface. One might expect that including the protein’s conformational flexibility and interactions with water will smooth out this rugged landscape. But Katrina Lexa and Heather Carlson at the University of Michigan say that it doesn’t, unless the protein is allowed its full flexibility (JACS 133, 200; 2011 – paper here). In other words, there are no short cuts: if the molecule is semi-rigid, spurious hot spots remain.

Structured water molecules bound within cytochrome c have been suspected for some time of participating in the enzyme’s electron-transfer and oxygen reduction reactions, but it hasn’t been clear exactly how. Amandine Maréchal and Peter Rich at University College London have used FTIR spectroscopy to investigate the issue (PNAS pnas.1019419108 – paper here). They find that rearrangements of up to 8 water molecules are associated with the photolysis reaction, probably forming transient hydrogen-bonded pathways for proton conduction and gating.

More roles for water in protein-ligand binding are revealed by Michelle Sahai and Philip Biggin at Oxford (J. Phys. Chem. B 10.1021/jp200776t – paper here). They consider how the GluA2 ionotropic glutamate receptor binds both glutamate and the related compound AMPA, and find two quite distinct modes of binding. The difference seems to be in the location of a single water molecule in the binding cleft. They use density functional theory to figure out why it is more favourable for the water to sit in different positions in the two different cases. That they don’t seem yet to have fully settled that question (they cannot consider entropic contributions) seems to underline how tricky it might be to use bound waters in rational drug design.

Joe Dzubiella in Berlin has an interesting preprint on the thermodynamics of hydrophobic association which claims that the curvature of the interface is important. He points out how there seems to be a crossover at small size scales (about 1 nm) between enthalpy-driven hydrophobic association at larger scales and entropy-driven at smaller scales. But for concave binding cavities this does not seem to apply: enthalpy continues to dominate even at very small scales. Joe explains this on the basis that the surface-area-based models used to describe large-scale interactions on the basis of solvent-accessible area and surface tension remain applicable at small scales so long as “the antagonistic effects on concave vs. convex bending on water interface thermodynamics are properly taken into account”. The paper will appear in a forthcoming special issue of J. Stat. Phys. dedicated to water.

There’s a nice potted summary of current understanding of antifreeze protein ice-binding mechanisms by Kim Sharp of the University of Pennsylvania in PNAS (pnas.1104618108 – paper here). It is a commentary on a new study by Garnham et al. (PNAS pnas.1100429108), which I haven’t yet got hold of. They report the crystal structure of the AFP of an Antarctic bacterium, Marinomonas primoryensis, which has a new binding motif, a parallel beta-helix. This structure has the ice-binding surface fully solvent-exposed in the crystalline state, and so is likely to be “free from crystal-packing artifacts.” Much of this surface is hydrophobic (although anchored at the edges by H-bonds), and the claim is that this induces a clathrate-like structure in the first hydration layer that is close to the structure of ice – in other words, the AFP “brings its own ‘ice’ with it”.

With an eye on the Lum/Chandler/Weeks model of hydrophobic attraction, Pablo Debenedetti and colleagues have calculated the evaporation length scale – the separation of solvophobic plates at which capillary evaporation occurs – for water and a range of organic liquids (C. A. Cerdeiriña et al., J. Phys. Chem. Lett. 2, 1000; 2011 – paper here). There’s nothing conceptually new here, but the numbers are somewhat surprising: for water they find a length scale of about 1.5 microns at atmospheric pressure, which is much larger than I’d have expected. However, this applies for purely repulsive surfaces (contact angle of 180 degrees), which is of course rarely found, and never for protein surfaces. The length is also large for the organics, such as benzene, heptane and cyclohexane, but about a factor of 3 less so – water is (somewhat) anomalous here because of its large surface tension.

Hydration seems to respond to the flexibility of alicyclic systems, according to Annalisa Boscaino and Kevin Naidoo of the University of Cape Town in South Africa (J. Phys. Chem. B 10.1021/jp110248j – paper here). They find from MD simulations that molecules with a cyclopyranose framework, such as glucose, have a significantly higher hydration number than those based on a cyclohexane framework with hydroxyls, such as cyclohexanol. This seems to result from the greater rigidity of the former, enabling the formation of longer-lived hydrogen bonds to the surrounding water.

It’s generally thought that rearrangements of the hydrogen-bond network in bulk water must have a collective character. Andrei Tokmakoff at MIT and colleagues now provide evidence of that (R. A. Nicodemus et al., J. Phys. Chem. B 115, 5604; 2011 – paper here). They have used ultrafast IR spectroscopy of HOD in pure water to measure the energy barriers to spectral diffusion and reorientational relaxation, and find that the slow-decay component is consistent with collective reorganization.

Marcus Weinwurm and Christoph Dellago have calculated the vibrational spectra of single-file water molecules in narrow pores, enabling them to distinguish it from the stacked-ring structure in wider pores (J. Phys. Chem. B 115, 5268; 2011 – paper here).

Thursday, April 28, 2011

Water as glue

There is a very nice crop of papers to draw on for this post, many of which speak to very central questions in this field. First up is a paper in Nature Communications from Volkhard Helms and colleagues at Saarbrücken, which looks at the detailed role of interfacial water in the association of hydrophilic protein surfaces (M. Ahmad et al., Nat. Commun. 2, 261; 2011 – paper here). Whereas hydrophobic association has been the focus of a lot of recent attention, especially with a view to the possibility of a dewetting-induced process, hydrophilic surfaces have received less attention. That’s an important lacuna, since as the authors point out, around 70% of interfacial residues are hydrophilic. The common approach is to assume direct electrostatic interaction mediated by a continuum solvent. But the water network has a more complex role. As shown in these MD simulations of the barnase-barstar complex, water molecules mediate and stabilize the interactions between native contacts. Moreover, for electrostatic interactions to be important, the interfacial water’s dielectric constant needs to be reduced to reduce screening. This happens as a consequence of changes in the structure of the interfacial layers (the dielectric permittivity is less than 50% of the bulk value for interfacial separations of less than 1.2 nm), and it preferentially promotes electrostatic interactions normal to the surfaces. In other words, you could say that (once again, though arguably to put the cart before the horse) water does exactly what is required of it.

Jeremy England, now at Princeton, has a paper in press with Structure which shows that it is possible to estimate low-energy conformational changes in a protein, such as those involved in allosteric effects, on the basis simply of residue-by-residue hydrophobic effects. Specifically, he develops a method for determining the most energetically favourable way of burying hydrophobic residues, given a particular amino-acid sequence. This amounts to identifying the particular ‘burial modes’ of any given sequence. Thus, although the stabilities of conformations are doubtless multifactorial, hydrophobicity seems to be the major governing factor.

Calculating the surface free energies of heterogeneous surfaces exposed to water (such as protein surfaces) is tough. The approach of the Cassie equation is additive, but as Alenka Luzar and colleagues point out, that doesn’t always work (J. Wang et al., PNAS 108, 6374-6379; 2011 – paper here). They show that deviations from linear additivity result when parts of a surface are unevenly exposed to solvent. In particular, it seems that polar patches exert an inordinately strong influence, being able to ‘pin’ a droplet so that it might remain closely attached to adjacent hydrophobic patches. They examine these effects with reference to water droplets first on a functionalized graphene surface and then on the surface of melittin.

Daryl Eggers at San José State University presents an interesting approach to the energetics of reactions in aqueous solution, geared especially to biochemical equilibria, that treats the water as a reactant and product, thereby subsuming the local changes in water structure that inevitably accompany the reaction (Biochemistry 50, 2004-2012; 2011 – paper here). Here the free energy of bulk water is treated as a variable, allowing for the effects of all solutes including those that may not participate directly in the reaction (such as dissolved salts). As I understand it, this offers a means of accommodating the effects of such solutes (for example, in salting in/out) that does not make any assumptions about global changes in ‘water structure’, but represents only the global average of localized changes. Also jettisoned in that process is any insistence on putative structure-makers and structure-breakers; rather, solvation effects need be discussed only in what seems like relatively uncontentious terms of subpopulations of water with differing free energies. I haven’t yet quite figured out how one gets at these free energies in experimental terms, but I like the principle, not least because it explicitly acknowledges the role of water as a participant.

Yingkai Zhang at New York University and coworkers have used ab initio MD simulations to study the mechanism of action of histone deacetylase (HDAC) enzymes, which remove acetyl groups from histone residues and have been identified as a target for anti-cancer drugs (R. Wu et al., JACS 133, 6110-6113; 2011 – paper here). They suggest that some HDACs function via a mechanism involving a modulation of water access (a hydrogen-bonded chain of waters) to the binding pocket, in which a zinc ion in the metalloenzyme binds to its substrate. The presence or absence of water alters the dielectric constant in the binding pocket and thereby affects the strength of zinc binding.

Transitions of DNA between A, B and Z forms are thought to be associated with and perhaps driven by transitions in the nature of hydration. Karim Fahmy and colleagues at the Institute of Radiochemistry in Dresden now suggest that the same applies to the more subtle sub-transitions between the BI and BII states of the B-form (H. Khesbak et al., JACS 133, 5834-5842; 2011 – paper here). Specifically, there are two sub-populations of water molecules – one bound to phosphates, the other not – that contribute to stabilizing the two conformations via entropic effects. These water rearrangements can also be involved in interactions with DNA-binding ligands, such as the antimicrobial peptide indolicidin, by a water-mediated induced fit.

The structures of amyloid fibrillar assemblies are still far from well understood. The hydration state of the peptides is a particularly important issue in determining their stability and perhaps their mode of formation. Beat Meier at ETH and colleagues report some tricks that enable this question to be probed by NMR (Van Melckebeke et al., J. Mol. Biol. 405, 765; 2011 – paper here). For a particular prion domain called HET-s(218-289) they show that, although these protofibrils have a hydrophobic core and a semi-hydrophobic pocket, they do not engage in ‘dry’ interfibril contacts but are each surrounded by water.

The behaviour of water close to and between lipid bilayers has been much studied, but there doesn’t seem to have been much consideration of how that behaviour might feature in biological membrane processes such as fusion. This issue is investigated by Vijay Pande at Stanford and colleagues using simulations (P. M. Kasson, E. Lindahl & V. S. Pande, JACS 133, 3812-3815; 2011 – paper here). They look at the water trapped between the faces of two approaching membranes – that is, in hydrophilic confinement – and find that the dynamics are altered significantly. Specifically, the trapped water has reduced rotational entropy, and it helps the two membranes to adhere. However, the slower dynamics also retards the process of fusion itself – the formation of the ‘stalk’ that bridges the lipid membranes.

Mafumi Hishida and Koichiro Tanaka at Kyoto have also looked at the hydration of phospholipid bilayers, here experimentally using terahertz spectroscopy and SAXS (Phys. Rev. Lett. 106, 158102; 2011 – paper here). They conclude that water structure is perturbed up to 4-5 layers from the surface, over a distance of at least 1 nm, and that the average density in this hydration layer is slightly greater than that in the bulk.

Meanwhile, Joshua Layfield and Diego Troya at Virginia Tech have considered a water droplet confined between hydrophobic surfaces of self-assembled monolayers (J. Phys. Chem. B 115, 4662-4670; 2011 – paper here). Here the water motions (lateral translational diffusion) are accelerated by confinement, and this effect seems to operate over distances of more than 1 nm from the surfaces. Structural effects (preferential orientation of the water molecules) aren’t evident beyond 1 nm, however – but while the authors consider this to be a relatively short-ranged effect, I’d have been surprised to see anything structural with a longer reach.

Daniela Russo at the ILL and colleagues have used inelastic neutron scattering to probe the low-frequency densities of states of water hydrating small ‘model peptides’ (N-acetyl-leucine-methylamide, NALMA, and N-acetyl-glycine-methylamide, NAGMA) at low temperatures (Russo et al., JACS 133, 4882-4888; 2011 – paper here). At 200K they find that the hydration water for the hydrophilic NAGMA is similar to high-density amorphous ice, while that of the hydrophobic NALMA is more like low-density amorphous ice. Something similar has been reported at 100 K by Paciaroni et al. (Phys. Rev. Lett. 101, 148104; 2008), but was in that case attributed to curvature of the biomolecular surface – which is evidently not the case for these small molecules.

Aquaporins seem to play a crucial role in water regulation in arid periods during the life cycle of the major malaria vector mosquito Anopheles gambiae, according to Kun Liu of the Johns Hopkins Malaria Research Institute in Baltimore and colleagues (K. Liu et al., PNAS 108, 6062-6066; 2011 – paper here). The authors don’t say whether this makes these AQPs a potential target for controlling the spread of malaria, though I suppose that is a possible implication.

There’s still more to be understood about what hydrogen bonds are, as evidenced by a recent IUPAC working group set up to redefine them (see here). Angelos Michaelides at UCL (who I have to thank for my water-crested football shirt) and colleagues have now refined the quantum picture of the H-bond, showing how quantum nuclear effects due to the anharmonicity of the bond and the small proton mass can alter the bond strength, weakening weak H-bonds (like those in water) and strengthening strong ones (X.-Z. Li et al., PNAS 108, 6369-6373; 2011 – paper here).

There are one or two other papers I’ve still to get to, including some that folks have kindly sent to me. Apologies for that – more soon, I hope.

Thursday, February 24, 2011

Coarse-grained models

More on the mechanisms of urea-induced protein denaturation, this time from Ruhong Zhou at IBM and his colleagues in Beijing (M. Gao et al., J. Phys. Chem. B 114, 15687; 2010 – paper here). In simulations, they look at the populations of water and urea molecules around each residue in hen egg-white lysozyme, and find that some of the hydrophobic core residues stay virtually dry as unfolding proceeds in 8M urea. Moreover, the urea molecules are bound preferentially to uncharged rather than to charged residues. So the picture is that the protein swells to a molten globule state while keeping largely dry inside, because it is the urea rather than water that penetrates. this may not be a wholly general order of events, however – depending on the detailed shape and structure of the protein, water might sometimes penetrate first, as found for example by Bennion and Daggett (PNAS 100, 5142; 2003).

John Klassen and colleagues at the University of Alberta present an interesting comparison of protein-ligand dissociation constants in the hydrated and dehydrated states (L. Liu et al., JACS 132, 17658; 2010 – paper here). The latter are obtained from gas-phase measurements at 25-66 C. The kinetic stability of the associated state is significantly reduced in the hydrated phase: water stabilizes the dissociative transition state and might thereby be considered a kind of lubricant that facilitates the departure of the ligand.

David Beauchamp and Mazdak Khajehpour predicate their study of water-water interactions on enzyme activity with the supposition that the hydrogen-bond distribution in pure water is bimodal, with bonds that are high- and low-angle (J. Phys. Chem. B 10/1021/jp107556s – paper here). They say this is supported by some simulations, and that it is consistent with the (contentious) two-state model of the liquid. They also say that salts have the ability to perturb this distribution. OK… so given these assumptions, they see what added salts do to the activity of ribonuclease t1, and say that their spectroscopic and kinetic results are consistent with the notion that salts that promote the high-angle bonds stabilize the more compact and less active forms of the enzyme. Interesting ideas, but it seems a big leap from the data to the microscopic interpretation.

I recently described work by Martin Gruebele, Martina Havenith and colleagues on the mechanism of antifreeze glycoproteins, in which they argued that the biomolecules can effect long-range changes in water dynamics to inhibit freezing (S. Ebbinghaus et al., JACS 132, 12210; 2010). The authors now report similar findings for a 37-residue alpha-helical antifreeze protein from winter flounder (S. Ebbinghaus et al., Biophys. J. Biophys. Lett. in press). The find using CD and FRET on native and mutant versions that the antifreeze activity seems to be connected to a kinking of the helix, and that this is coupled to a suppression of bulk-like dynamics in the solvation water over a range of at least 3 nm, as indicated by THz spectroscopy.

Years ago, Royer et al. showed that a group of water molecules at the interface between the subunits of the dimeric haemoglobin of Scapharca clams seem implicated in the molecule’s allosteric cooperativity (Royer et al., PNAS 93, 14526; 1996). David Leitner and colleagues at the University of Nevada at Reno now look in detail at the dynamics of this cluster of waters using MD simulations based on the crystal structures (R. Gnanasekaran et al., J. Phys. Chem. B 114, 16989; 2010 – paper here). They find that those in the oxy form (11 molecules) exhibit slower relaxation than those (17) in the deoxy form, and that the water cluster, although rather static on ps timescales, can enhance energy transport across the interface of the subunits via vibrations.

The close and reciprocal interactions of water and protein dynamics, especially at low temperatures, seems to be echoed in the case of hydrated lipid bilayers, according to Peter Berntsen and colleagues at Chalmers University in Göteborg (J. Phys. Chem. B 10.1021/jp110899j – paper here). They have used dielectric rexlaxation measurements below 250 K to show that at low temperatures the water dynamics becomes increasingly dominated by the movements of the lipids, and is super-Arrhenius-like at low hydration levels.

Fast proton transport along peptide backbones can be assisted by water bridges, according to ab initio calculations by Po-Tuan Chen of the National Taiwan University of Science and Technology and colleagues (J. Phys. Chem. B 10.1021/jp107219r – paper here). They say that a two-molecule water bridge can make the transport between two adjacent carbonyl oxygens almost barrierless.

In a preprint (not sure where it is destined, but it looks like JCP format), David Chandler and his coworkers present a coarse-grained lattice model to implement the Lum-Chandler-Weeks dewetting theory of hydrophobic interactions (P. Varilly et al., arxiv: 1010.5750 – paper here). The model captures the essential features of the model at far less computational cost than full MD simulations, in particular modelling the solvent fluctuations that are essential for the dewetting mechanism.

With much the same objective of computational cheapness, Ken Dill at UCSF and colleagues present a new solvation model that they call semi-explicit self-assembly (C. J. Fennell et al., PNAS 108, 3234; 2011 – paper here). They basically construct a solute’s solvation shell as some combination of pre-computed solvation shells for simple spheres in explicit TIP3P water. They have so far only tested it here on simple small molecules such as sugars.

And Valeria Molinero and colleagues at the University of Utah have a coarse-grained model of DNA solvation with explicit water and ions (R. C. DeMille et al., J. Phys. Chem. B 115, 132; 2011 – paper here). It reproduces base-pair specificity and is computationally faster by two orders of magnitude than atomistic simulations.

There is a curious paper in Nature Structural and Molecular Biology by Nathaniel Nucci and colleagues at the University of Pennsylvania (NSMB 18, 245; 2011 – paper here) on an NMR technique to identify residence times of specific clusters of water molecules around proteins. I say curious, because these useful results are presented, particularly in a News & Views article in Nature (V. J. Hilser, Nature 469, 166; 2011 – paper here), as “challenging current dogma about protein hydration”. It seems this challenge comes from the fact that the results show that not all ‘bound’ water exchanges slowly with the surrounding solvent. But a wide range of exchange times is surely already well established, especially from simulations – the old crystallographic picture in which ‘hydration water’ is all securely bound and long-lived seemed long dead. Still, it is interesting that Nucci et al., whose method relies on confining the proteins (here ubiquitin) within reverse micelles to slow the hydration dynamics, found that water molecules with similar residence times seem to cluster on the protein surface, so that the molecules in each cluster form independent networks which exhibit intra-cluster cooperativity.

In another preprint, Giancarlo Franzese and colleagues at the University of Barcelona offer something of a mini-review of hydration structure and dynamics at protein surfaces, along with some Monte Carlo simulations that investigate cooperativity and dynamical transitions of a water monolayer hydrating a protein surface at low temperatures (arxiv preprint 1010.4984; paper here).

On the still-evolving picture of pure liquid water: Alessandro Cunsolo at Brookhaven and his colleagues study it using QENS to look at single-particle diffusion rates at 200 MPa as a function of temperature, and find that their results point to the proposed existence of a second critical point at about 220 K, and of the Widom line which ends at this point (J. Phys. Chem. B 114, 16713; 2010 – paper here).

Meanwhile, Richard Henchman and Sheeba Jem Irudayam of Manchester University in the UK propose a ‘topological’ definition of hydrogen bonding in water that offers a new description of water structure and dynamics (investigated in simulations using TIP4P/2005 water) based on the character of the H-bond network (J. Phys. Chem. B 10.1021/jp105381s – paper here). They say that in this description almost all the water molecules are H-bonded and that there are an appreciable number of ‘defects’ in which molecules are acceptors for one (trigonal) and three (trigonal bipyramidal) hydrogens rather than two.

Monday, February 14, 2011

Less is more

In the spirit of maintaining continuity (do some folks really look at this site every week, as I’m told?), I will aim to post more regularly rather than exhaustively. That’s to say, this is not a complete list for what I’ve seen out there, but more will follow soon.

Sheh-Yi Sheu at National Yang-Ming University in Taiwan and Dah-Yen Yang at the Institute for Molecular Science in Okazaki, Japan, present a method for deducing the free energy of the hydration shell for a biomolecule from simulations (J. Phys. Chem. B 10.1021/jp105164t – paper here). They say that water in a protein (here myoglobin) hydration shell follows fractional Brownian motion.

Flavoproteins are electron-transfer proteins involved in a range of biological processes, including cell apoptosis and DNA repair. The nature of solvation at the active sites is of central importance for their function. Dongping Zhong and colleagues at Ohio State have used ultrafast spectroscopy to characterize the dynamics of the water network at the functional site in three redox stats of a representative flavoprotein, flavodoxin (C.-W. Chang et al., JACS 192, 12741 (2010) – paper here). Rather neatly, they are able to use the intrinsic cofactor of this protein as the fluorescent probe for the experiments. They can monitor changes in the relaxation and rigidity of the local water network between the different states, for example a retardation of the relaxation from around 2.6 to 40 ps between the oxidized to semiquinone state. They propose an intimate and biologically relevant coupling between the flexibility of the solvation network and the protein.

Gating of ion channels due to cooperative drying has been suggested as the underlying mechanism for such functions. Fangqiang Zhu and Gerhard Hummer at NIH explore this notion for the ligand-gated ion channel GLIC of the bacterium G. violaceus, for which the crystal structure of the open state is solved (PNAS 107, 19814; 2010 – paper here). They find that the pore is typically water-filled in the open state, but that a very small decrease in the channel radius can induce cooperative drying. It seems that the emptying of the pore is a response to, rather than the driving force for, changes in the pore width.

Alla Oleinikova, Ivan Brovchenko and their colleague G. Singh at Dortmund have calculated the heat capacity of hydration water of hydrophobic and hydrophilic peptides from simulations (A. Oleinikova et al., Europhys. Lett. 90, 36001; 2010 – paper here). They say that around 330 K there is a sharp change in structure from a percolating to a fragmented H-bonded network, and that this coincides with a point at which the heat capacity changes from being dominated by water interactions within the hydration shell to a situation where the hydration-shell-to bulk interactions are more important. At this stage, the ‘detachment’ of the hydration network from the peptide in effect makes the biomolecule more hydrophobic.

Anrew White and Shaoyi Jiang at the University of Washington have studied the hydration of glycine and two (zwitterionic) analogues di- and trimethylglycine using MD (J. Phys. Chem. B 115, 660; 2011 – paper here). They say that all three molecules affect the water structure out to the second hydration shell, but trimethylglycine has the greatest (retarding) effect on water dynamics and, perhaps surprisingly, does not aggregate but remains well solvated even at high concentrations. This might help to account for trimethylglycine’s antifouling properties and its suppression of protein aggregation.

There is sill a tussle going on about whether bulk water at ambient temperatures is best considered in a homogeneous or two-phase framework. Lars Pettersson and Anders Nilsson have recently teamed up with theorists including Jens Norskov at Stanford to perform ab initio MD calculations which point to a mixture of low- and high-density regions at ambient conditions (A Møgelhøj et al., arxiv preprint 1101.5666; paper here). But Niall English and John Tse dispute this, saying that such inhomogeneities are of very short range and transient, being just the ordinary density fluctuations one would expect to see in an equilibrium system (Phys. Rev. Lett. 106, 037801; 2011 – paper here).

Raymond Dagastine and colleagues at the University of Melbourne report a very striking claim: that air bubbles interact with (probably virtually all) solid surfaces via a repulsive van der Waals interaction (R. F. Tabor et al., Phys. Rev. Lett. 106, 064501; 2011 – paper here). As a bubble approaches a surface under hydrodynamic flow, they say, it will therefore start to deform and flatten at the point where the repulsive force equals the Laplace pressure. Repulsive vdW forces are well known in theory, but it seemed previously that rather special combinations of materials were needed to realise them.

Thursday, February 3, 2011

Yes, I'm still here

Apologies for the long silence. This blog is still active, but Christmas (among other things) got in the way and then I face the Sysiphean task of catching up. That task is not by any means completed here, but I want to flag up that Water in Biology hasn’t expired. Much more as soon as I can manage it.

Bruce Berne, Richard Friesner and Lingle Wang at Columbia have recently introduced the notion that ligand binding in protein receptor pockets is largely driven by the displacement of water molecules that sit in an unfavourable position in the pocket, which are replaced with groups on the ligand that are complementary to the protein surface (Young et al., PNAS 104, 808; 2007; Abel et al., JACS 130, 2817; 2008). In a new paper (Wang et al., PNAS 108, 1326; 2011 – paper here) they consider what their model, called WaterMap, has to say about regions of the binding pocket that are initially dry, being highly unfavourable environments for water. Including an interaction term in the model that represents the formation of a (hydrophobic) protein-ligand interface in dry regions, they can compute binding affinities in good agreement with experiment.

Dave Thirumalai at Maryland and his coworkers offer a striking view of how amyloid fibrils self-assemble from interdigitated beta-sheets (G. Reddy et al., PNAS 10.1073/pnas.108616107 – paper not yet online). They present a MD study of the association of beta-sheets in two amyloidogenic proteins of very different sequence, one polar and the other hydrophobic. They say that in the former case the association of the sheets is mediated by one-dimensional water wires at the interface between them, which are gradually expelled. But for the hydrophobic peptides the sheets come together in something like an abrupt drying transition, as postulated previously for some protein-folding and aggregation processes. This happens much faster (nearly 1,000-fold) than the previous case, since the trapped water wires for the polar peptide create a barrier to rapid assembly. Thus, although the final structures are very similar, the mechanisms and dynamics are quite different. It would seem that this paper ties in with a new one by Ken Dill and colleagues on the mechanisms of amyloid assembly into fibrils, of which I’ve only seen the abstract (which suggests that there’s not a big emphasis on the role of the solvent here beyond the involvement of hydrophobicity) (J. D. Schmit et al., Biophys. J. 100, 450-458; 2011 – paper here).

Cytochrome c oxidase (CcO) acts as a proton pump in which the transmembrane proton motion is thought to be facilitated by a proton wire involving strategically placed water molecules. The roles of these waters are investigated by Shelagh Ferguson-Miller and colleagues at Michigan State based on high-resolution crystals structures of two mutant forms of bacterial CcO (Liu et al., PNAS 108, 1284; 2011 – paper here). In both mutants, where proton transfer is inhibited to different degrees, the overall structural changes are very small but one or more of the bound waters is eliminated. The story is not, however, quite as simple as a mere break in the water wires, but involves subtle conformational changes between oxidized and reduced forms of the metal centres: an indication that, while bound water undoubtedly plays an active role in the catalytic function, in this case that role resists reduction to a simplistic picture.

Human telomeres contain G-rich sequences that form quadruplex structures in Hoogsteen hydrogen-bonded patterns. It’s not clear what influences the stability of this unusual motif, but John Trent and colleagues at the University of Louisville in Kentucky say that hydration plays a major part (M. C. Miller et al., JACS 132, 17105-17107; 2010 – paper here). They say that previous studies of the crystal structure of these sequences have been misleading because the dehydrating agents used to cause precipitation (PEGs) may give crystal structures that are not closely related to those in solution. Instead they use acetonitrile (which is water-soluble but does not engage in hydrogen-bonding) as the cosolvent for CD and NMR solution studies, and find that stabilization of the quadruplex seems to be caused more by dehydration than by steric crowding effects.

How hydration affects energy relaxation of cytochrome c after photoexcitation has been studied using ultrafast spectroscopy by Shuji Ye of the University of Science and Technology of China in Hefei and and Andrea Markelz of SUNY at Buffalo (J. Phys. Chem. B 114, 15151-15157; 2010 – paper here). One of the main conclusions is that hydration doesn’t in fact have a great deal of influence on the initial energy dynamics: there is an initial fast (around 300 fs) conversion from the electronically excited state to a vibrationally excited ground state, which is essentially hydration-independent. But the vibrational cooling then does involve interaction with the solvent, more or less in line with the existing notion that hydration water acts as a kind of plasticizer in this molecule.

Finally for now, and not really at all relevant to the real themes of this blog but too much of a curiosity for me to ignore, there are two papers on the arxiv investigating the notorious Mpemba effect, whereby hot water is said to sometimes freeze faster than cold: see here and here.