Chiral water in DNA's hydration shell

In a clever study of DNA hydration using SFG spectroscopy, Poul Petersen and his coworkers have found that the chiral spine of hydration in the minor groove, inferred from oxygen locations for hydrated crystalline DNA by Dickerson and collaborators in the 1980s, exists also in aqueous solution under ambient conditions, and entails orientational ordering of the hydrogen bonds in the single-file water chain that fits into this narrow groove (M. L. McDermott et al., ACS Centr. Sci. 10.1021/acscentsci.7b00100; 2017 – paper here). I wrote a news story for Chemistry World on this work (here).

I applaud the ambition of Modesto Orozco of the Barcelona Institute of Science and Technology and colleagues in writing a paper called “The multiple roles of waters in protein solvation” (A. Hospital et al., JPCB 121, 3636; 2017 – paper here). There’s a title guaranteed to say to me “Read this now!” And the ambition continues in the extent of the systems they investigate with MD: a range of proteins, at a range of temperatures, some denatured, some with crowding agents, some with high concentrations of urea. They say that the results illustrate “the dramatic plasticity of water, and its chameleonic ability to stabilize proteins under a variety of conditions”, which seems a fair way to summarize the matter. I’m not sure I see any surprises here, and the denaturant effects of urea are discussed with something of a “water structure” flavour, but it’s a kind of snapshot of the sorts of things hydration water gets up to.

A more specific study of protein hydration dynamics is described by Dongping Zhong and colleagues at Ohio State University, who use tryptophan as the reporter group to characterize the dynamics at 17 sites on the surface of the β-barrel protein rat liver fatty acid binding protein (J. Yang et al., JACS 139, 4399; 2017 – paper here). They observe three quite distinct dynamical timescales. The water in the outer hydration layer is bulk-like, relaxing quickly (hundreds of fs). For the inner layer, reorientational motion happens on a few-ps timescale, while larger-scale network restructuring takes many tens of ps. The last of these seem to drive protein fluctuations on comparable timescales.

The dynamics of the protein hydration layer are examined by Biman Bagchi and colleagues of the Indian Institute of Science in Bangalore by calculating those around residues (Trp, Tyr, His) previously used as natural probes in spectroscopic studies (S. Mondal et al., arxiv preprint 1701.04861). They find a range of different timescales, including accelerated as well as retarded rotations. Since NMR measurements give average values, these findings might explain the apparently discrepancy between such studies and those (such as Zewail’s) that focus on specific residues. The protein side-chain dynamics seem particularly to influence the slow solvation component.

The role of hydration in the protein dynamical transition around 230 K has been widely debated. Prithwish Nandi and Niall English at University College Dublin find in MD simulations of lysozyme that the protein and hydration water dynamics seem to be correlated up to about 285 K, at which point the protein-water hydrogen-bond network becomes too disrupted to sustain the coupling (JPCB 120, 12031; 2016 – paper here).

However, the whole notion of coupling between the protein and hydration dynamics in the vicinity of the ~200-220 K dynamical transition is challenged by Antonio Benedetto of University College Dublin on the basis of elastic neutron-scattering from lysozyme (arxiv preprint 1705.03128). Specifically, the water begins to relax at 179 K, while the protein doesn’t do so until 195 K. It seems puzzling, and no explanation is advanced here for the discrepancy with a considerable body of earlier results.

I missed previously this nice paper from H. F. M. C. Martiniano and Nuno Galamba in Lisbon on the structure and dynamics of water around a hydrophobic amino acid (PCCP 18, 27639; 2016 – paper here). It reports MD simulations of the hydration of valine, and distinguishes between two populations of water molecules in the hydration shell: those that have have four and less than four neighbours. The latter, they say, have faster librational dynamics than bulk water and faster orientational dynamics than four-coordinated “tetrahedral” water. Meanwhile, four-coordinate water in the hydration shell are “more tetrahedral” than bulk water at all temperatures. It would seem, then, that this work argues the case for “tetrahedrality” as a useful concept for characterizing water structure, while advising caution about how it is used and interpreted for the bulk.

Guanidinium is a complicated osmolyte. It can act as both a protein denaturant and stabilizer, depending on the counteranion. Jan Heyda at the Institut für Weiche Materie und Funktionale Materialien in Berlin and colleagues have setout to understand why, using MD simulations and FTIR (J. Heyda et al., JACS 139, 863; 2017 – paper here). Their test peptide, an elastin-like polypeptide, was stabilized in the collapsed state by Gnd sulphate by an excluded volume effect (Gnd being depleted at the peptide/water interface). GndSCN was stabilizing at low concentrations thanks to Gnd+’s ability to crosslink the polymer chains, but at higher concentration it became a denaturant. GndCl, meanwhile, was a denaturant at all concentrations, since in this case partitioning of the chloride to the polymer surface enables recruitment of Gnd+ to the surface too, where it stabilizes the unfolded state. A very graphic example of how the details of direct interactions between polymer, anion, cation (and potentially water) all matter in figuring out what is going on.

Essentially the same team – which includes Paul Cremer, Joachim Dzubiella and Pavel Jungwirth – have put together a review of such ion-specific effects that, it seems to me, will be the go-to resource for this field for some time to come (H. I. Okur et al., JPCB 121, 1997; 2017 – paper here). I need say no more; if you want to understand how the thinking on Hofmeister has developed over the past several years, this is where to come.

Does water play the role of reactant in O-O bond formation in photosystem II? That idea has been suggested, water acting as a nucleophile that attacks a terminal oxo group. But Per Siegbahn of Stockholm University uses DFT calculations to determine the free-energy barriers for the six most plausible modes of attack and finds that these barriers are all too high (PNAS 114, 4966; 2017 – paper here) – a notion put forward previously but here refined using improved structural data and computational methods.

I didn’t even know that lipid bilayers, like proteins, show a dynamical transition around 200 K or so. But it seems they do. V. N. Syryamina and S. A. Dzuba of the Russian Academy of Sciences in Novosibirsk have studied thus for two types of phosphocholine bilayers in water using a technique (also new to me) called electron spin echo envelope modulation spectroscopy to follow hydrogen (deuterium) motions (JPCB 121, 1026; 2017 – paper here). They find that the dynamical transition in the bilayer interior at 188 K is accompanied by the onset of water motion in the first hydration layer, and that another transition around 100 K is accompanied by restricted reorientational motions of water. What I can’t tell from these results is whether there is any sign of slaving of water to lipid dynamics or vice versa.

I’m not going to pretend to understand the Bayseian model used by Nathan Baker of PNNL in Washington and colleagues to estimte small-molecule solvation free energies (L. J. Gosink et al., JPCB 121, 3458; 2016 – paper here). But it’s basically a method for aggregating many other calculational procedures, and seems to work better than any such techniques in isolation.

Mihail Barbiou of the European Institute of Membranes in Montpellier and colleagues have used artificial water channels in liposomes, made from stacked imidazoles, to investigate water transport along water wires, analogous to those that thread through aquaporins (E. Licsandru et al., JACS 138, 5403; 2016 – paper here). The channels can conduct around a million water molecules per second, a rate two orders of magnitude greater than AQPs, and also conduct protons (but not other ions) efficiently. The chirality of the channels seems to be important for producing strong dipolar orientation in the water wire. Let me also draw attention to Mihail’s nice review of artificial water channels, which includes this example, in Chem. Commun. 52, 5657 (2016) (paper here).


The water channel in stacked imidazoles.

More on water confined in pores: in MD simulations, Xiao Cheng Zeng at the University of Nebraska and colleagues see low- and high-density liquid states of water within single-walled carbon nanotubes of 1.25 nm diameter at ambient temperature (K. Nomura et al., PNAS 114, 4066; 2017 – paper here). The two phases are, however, separated by a hexagonal “tubular ice” phase (which has already been observed experimentally).

How does water freeze at liquid-vapour interfaces? Specifically, does the interface itself nucleate or suppress freezing? That’s a question relevant to a host of real-world phenomena such as ice nucleation in clouds and other atmospheric processes, but it’s been hard to study experimentally, but Amir Haji-Akbari and Pablo Debenedetti in Princeton study it computationally in a free-standing 4-nm-thick water nanofilm (PNAS 114, 3316; 2017 – paper here). Although the rate of ice nucleation in this confined geometry is seven orders of magnitude greater than that in the bulk, nucleation doesn’t start in the surface layers but rather in the (non-bulk-like) interior of the film, where the conditions favour the formation of “double-diamond” water cages that serve as the seeds for the nucleation and growth of cubic ice.

And here’s a truly surprising thing, discovered by Pablo and Amir in another paper working with Elia Altabet: making hydrophobic plates confining water to a space just over 1 nm wide more flexible by just an order of magnitude decrease in the modulus increases the evaporation rate by nine orders of magnitude, and decreases the condensation rate from the vapour by no less than 24 orders of magnitude, changing the timescale of the process from nanoseconds to tens of millions of years (Y. E. Altabet et al., PNAS 114, E2548; 2017 – paper here). This, at any rate, is what is implied by simulations for plates 3 nm square. Evaporation proceeds via the formation of bubbles at the surfaces that then grow and coalesce to form a gap-spanning cavity. For stiff plates this coalescence is rare, and so is the subsequent growth of the cavity above the critical size for nucleation of the vapour phase. For softer, more flexible plates these configurations occur much more frequently. Such a sensitivity of a drying transition to subtle changes in the mechanical properties may well have implications for processes involving hydration changes at or close to membrane proteins, and could presumably have ramifications for materials design of surfaces on which protein adhesion needs to be controlled.

Optimization of lead compounds for drug discovery is a complicated business, and when this is done by empirical combinatorial screening, the results can sometimes be counterintuitive, with nonpolar groups in the ligand juxtaposed to polar groups in the target for example. Ariel Fernandez at the Argentine Institute of Mathematics and Ridgway Scott of the University of Chicago review a method for understanding some of those apparent conundrums that involves a consideration of the relevant hydration structures, and in particular the role of what Ariel calls dehydrons (water-exposed backbone hydrogen bonds, which lead to frustration in the hydrogen-bonding arrangements of adjacent water molecules) (Trends Biotechnol. 35, 490; 2016 – paper here). Their approach uses the WaterMap software to identify “hot” water molecules that might profitably be displaced by a ligand to increase the binding energy and drug specificity.

The hydrogen-bond network of pure water is of course riddled with defects which underpin fluctuations of the network. Because of topological constraints these tend to occur in correlated pairs. Ali Hassanali at the ASICTP in Trieste and colleagues have studied these correlations using ab initio modelling (P. Gasparotto et al., J. Chem. Theor. Comput. 12, 1953; 2016 – paper here). They say that the defect pairs have some similarities to those in solid states of water, and are rather insensitive to the details of the water potentials used.

One of water’s well known “anomalies” is the decrease in viscosity with increasing applied pressure, which seems to be a consequence of a collapse of the hydrogen bonding network. This effect is larger at low temperatures, but whether that trend continues into the supercooled region hasn’t been studied previously. Now Frédéric Caupin and colleagues at the University of Lyon have investigated this effect down to 244 K and for pressures of up to 300 MPa, and find that indeed the viscosity reduction can be dramatic – by as much as 42% (L. P. Singh et al., PNAS 114, 4312; 2017 – paper here). They argue that the results can be understood by invoking a two-state model under these conditions: a mixture of a high-density “fragile” liquid and a low-density “strong” liquid.

Finally, I have taken what I hope is a somewhat fresh look at the many roles of water in molecular biology in an article for PNAS, for a special issue on water (2017 – paper here), which I hope extends the general message of my 2008 Chem Rev article (paper here) using some more recent examples.

Hydration water in drug design

Electrostatic interactions with lipid heads groups retard water molecules near the surface of a membrane. But how are those dynamics affected by a membrane protein? Lars Schäfer at the Ruhr University of Bochum and colleagues attempt to answer that question using (ODNP-enhanced) NMR and simulations to deduce water motions (O. Fisette et al., JACS 138, 11526; 2016 – paper here). They conclude that the water-protein interactions have a weaker retarding effect, and dominate only at distances of more than 10 Å above the membrane surface. Moreover, the protein (here annexin B12) and membrane effects are additive. This creates a gradient in water entropy with distance from the surface, with potential consequences for recognition and binding events involving membrane proteins.


The hydration environment of membrane protein annexin B12 in a lipid membrane.

A technique called oriented-sample solid-state NMR can supply information about the water-accessibility of individual residues of membrane proteins in situ, say Gianluigi Veglia and colleagues at the University of Minnesota (A. Dicke et al., JPCB 120, 10959; 2016 – paper here). They’ve used the method to gather this information for the archetypal small transmembrane protein sarcolipin in synthetic bilayers, and find that, as one might expect, there is a relatively smooth gradient of water accessibility with increasing depth within the membrane.

Does the denaturing effect of osmolytes such as urea and guanidinium chloride depend on concentration? Experiments using FRET and SAXS have produced conflicting results, and Robert Best at the NIH and colleagues try to resolve the matter using simulations (W. Zheng et al., JACS 138, 11702; 2016 – paper here). Their test case is the intrinsically disordered protein ACTR, for which they calculate the chain swelling and radius of gyration as a function of denaturant concentration. The protein does indeed swell to a degree proportional to concentration, but the researchers show that nevertheless the structural changes are consistent with SAXS results that appear to show no change in radius of gyration. The denaturant effects operate by a direct mechanism of weak association with the protein.

What is there is a mixture of denaturants such as urea and stabilizing osmolytes such as TMAO? It seems that TMAO can counteract urea’s effects, but how exactly is the hydrophobic interaction affected in that environment? Indrajit Tah and Jagannath Mondal at the Tata Institute in Hyderabad look into that question via simulations of a model hydrophobic polymer and polystyrene (JPCB 120, 10969; 2016 – paper here). In contrast to what is observed with proteins, for the hydrophobic polymer TMAO actually reinforces the destabilizing effect of urea. This seems to be due to the different direct interactions with the polymer chain: for proteins, TMAO is excluded from the surface while urea remains bound, whereas for the hydrophobic polymer both osmolytes may individually bind to the surface. In this latter case, exclusion of TMAO by urea in the mixed solution depletes the opportunities of TMAO to stabilize the collapsed state.

Guangzhao Zhang and colleagues at the South China University of Technology in Guangzhou look at the same denaturant-inhibiting effect of betaine, this time with lysozyme as the model protein (J. Chen et al., JPCB 120, 12327; 2016 – paper here). Using proton NMR, they conclude that in this case betaine interacts directly with urea to form dimers, removing the urea from the protein surface where otherwise it interacts directly to stabilize the denatured state.

Ion-specific “Hofmeister” effects on protein stability and aggregation are still not fully understood at the molecular level. Simon Ebbinghaus at the Ruhr University of Bochum and colleagues seek insights through thermodynamic (DSC) measurements on bovine ribonuclease A (M. Senske et al., PCCP 18, 29698; 2016 – paper here). By measuring ion effects over the whole temperature- and concentration-dependent landscape of protein stability, they find a very complicated picture, due to a complex interplay of contributions. At low concentrations, electrostatic (non-ion-specific) effects dominate, but at higher concentrations there is ion specificity. It’s hard (for me, anyway) to summarize the findings, but I believe it if fair to say that the authors are seeking a unified molecular picture that helps to explain not only ion effects but also those of non-electrolyte cosolutes on protein stability, in terms of a balance between entropic and enthalpic contributions to the excess free energy.

A belated addition from the Bochum group: Yao Xu and Martina Havenith have provided a nice summary of recent work using THz spectroscopy to look at ps-scale collective hydrogen-bond dynamics in the hydration shells of proteins, and the role that they play in molecular-recognition processes (JCP 143, 10.1063/1.4934504; 2015 – paper here).

A further use of THz spectroscopy to investigate hydration is reported by Y Ogawa and colleagues ay Kyoto University (K. Shiraga et al., Appl. Phys. Lett. 106, 253701; 2016 – paper here). Their aim is simply to get some bulk estimate of how much of the water in a cell (they use HeLa cells) has retarded dynamics – a question explored here in the context of the now more or less obsolete notion of “biological water” in cells. Their answer: about a quarter of the total water content has reorientational dynamics slower than the bulk, presumably because of its involvement in biomolecular hydration. This is more than the 10-15% reported previously in prokaryotes and human red blood cells. But of course the absolute numbers must depend on where one places the thresholds in a dynamical continuum.

The voltage-dependent proton transport channel Hv1 is implicated in diseases ranging from cancer to some forms of brain damage. This makes it a potential drug target, and some inhibitors seem to expel bound water from the pore when they bind. Because this water forms intermittent hydrogen-bonded clusters and water wires, it seems likely that there’s a hydration-related entropic contribution to the binding free energy. With this in mind, Mike Klein at Temple University and colleagues have used modeling and simulations to look at water fluctuations in the pore, so as to identify potential binding sites (E. Gianti et al., PNAS 113, E8359 2016 – paper here). Their analysis reveals two such sites: one the binding site known already, another at the outlet of the proton pathway, both of them associated with maximal fluctuation, apparently on the brink of a drying transition, and in locations where replacement by a hydrophobic ligand is optimal. The researchers say that the second, new site should therefore also be considered as a locus of drug design.

The same group has also considered the mechanism of proton transport (S C. van Keulen et al., JPCB 10.1021/acs.jpcb.6b08339; 2016 – paper here). It’s been suggested previously that this occurs via Grotthuss hopping along a water wire. But the results of these quantum/molecular mechanics simulations suggest that instead the proton hops between three acidic residues via mediating water molecules.


How a proton (circled in purple) makes its way down the channel of Hv1: via a series of water-mediated hops between acidic residues.

More proton transport: on the basis of FTIR experiments, Udita Brahmachari and Bridgette Barry at Georgia Tech say that a crucial stage of the oxygen-forming S-cycle of photosynthesis involves the insertion of a proton into a hydrogen-bonded water network (JPCB 120, 11464; 2016 – paper here). Thus bound water here acts as a catalytic proton acceptor and donor.


The catalytic water network in the S3-S0 stage of the photosynthetic cycle.

Jonathan Nickells at Oak Ridge and coworkers have characterized the general hydration environment of green fluorescent protein using neutron scattering spectroscopy to probe the dynamics (S. Perticaroli et al., JACS 10.1021/jacs.6b08845; 2016 – paper here). These dynamics are slowed over just two hydration shells: by a factor of 4-10 in the first shell and 2-5 in the second.


The hydration of GFP.

In comparison, Keiichiro Shiraga and colleagues at Kyoto see dynamical perturbations out to three or four hydration layers (a distance of around 8.5 Å) around albumin, based on THz spectroscopy (K. Shiraga et al., Biophys. J. 111, 2629; 2016 – paper here). They say that the hydrogen-bond network in the hydration layers seems to be less defective than that in the bulk, even though there seems to be greater distortion of the network away from tetrahedral.

Meanwhile, Monique Tourell and Konstantin Momot at the Queensland University of Technology have zeroed in on the single-water-molecule bridges that link parts of the peptide chains in collagen (JPCB 120, 12432; 2016 – paper here). Some experimental studies have apparently implied that the waters are “ice-like” in their dynamics, although I’m not clear quite what this is meant to imply. In any event, these MD simulations suggest otherwise: the waters exhibit strongly anisotropic rotation in which a single molecule might flip back and forth many times while remaining resident at the bridging site for more than 100 picoseconds.

When a particle is solvated at the air-water surface, fluctuations of the interface such as capillary waves may contribute to the solvation free energy – the solute might dampen the fluctuations, for example. Kaustubh Rane and Nico van der Vegt set out to quantify this using Monte Carlo simulations (JPCB 120, 9697; 2016 – paper here). They find that the contribution of fluctuations is not negligible in general, and that the dampening effect is a generic one that doesn’t depend on the chemical nature of solute or solvent. However, the strength of the interactions between ions and water will determine the magnitude of the effect of fluctuations, so that one can expect ion-specific propensities towards proximity to the water surface.

Ariel Fernández, now at the Argentine Mathematics Intitute, has been for some time exploring the role of hydrogen-bond frustration in protein-protein and protein-ligand interactions. The idea here is that such frustration contributes to interfacial tension via its effects on “non-Debye” polarization. Now he has studied how “minimal frustration” might guide these molecular assembly processes and be exploited in drug design, potentially enabling a high degree of binding selectivity (FEBS Lett. 590, 3481; 2016 – paper here).

There’s more on the use of hydration information for drug design from Gerhard Klebe of the University of Marburg and colleagues (S. G. Krimmer et al., J. Med. Chem. 59, 10530; 2016 – paper here). They say that optimizing the water layers covering ligands bound to their target – here hydrophobic inhibitors of thermolysin – can boost the enthalpic contribution to binding free energy. MD simulations enabled the prediction of high binding affinity for a series of ligands, one of which then proved to have 50 times better binding affinity than the known (and patented) parent ligand. This is a really nice piece of work, showing that it’s not just trapped or displaced water molecules that are important in drug design but also the final bound-ligand hydration profile.


The hydration structure of the best drug candidate for binding in the hydrophobic pocket of thermolysin.

Nested fullerenes, or “carbon onions”, cluster in water. Adam Makarucha have used MD simulations to look at the size- and shape-dependence of the effect (A. J. Kakarucha et al., JPCB 120, 11018; 2016 – paper here). As one might expect for a hydrophobic surface, there is layering of water at the surface of these particles, and the disruption of the hydrogen-bond network increases with increasing particle size because of the increased shape anisotropy: the tendency for the larger fullerene shells to become faceted with vertices (where the pentagonal rings sit).

To what extent are the bulk properties of pure water altered by confinement? Debates about confinement-induced changes in “water structure” have sometimes tended to overlook generic effects such as layering of a liquid close to a wall. Roland Netz and colleagues at the Free University of Berlin bring some clarity to the problem by using MS simulations to look at changes in the dielectric properties of water due to confinement-induced correlations in the polarization of neighbouring water molecules (A. Schlaich et al., Phys. Rev. Lett. 117, 048001; 2016 – paper here). They find that these effects – specifically an anticorrelation for neighbouing molecules – result in a significantly decreased dielectric response perpendicular (but not so much parallel) to a pair of walls (here consisting of closely packed decanol monolayers) separated by up to a nanometre or so. This behaviour has obvious consequences for, say, water’s ability to screen electrostatic interactions between closely spaced surfaces (of proteins or lipid membranes, say). [There’s also an APS Physics comment piece on this here]

The existence of a quasiliquid layer on the surface of ice below 273 K is now fairly well attested. Mischa Bonn and Ellen Backus of the MPI for Polymer Research at Mainz and their colleagues have studied this layer using SFG and simulations (M. A. Sánchez et al., PNAS 114, 227; 2016 – paper here). They find evidence of a stepwise transition from a single to a double bilayer of water molecules around 257 K. They say that there is evidence for the single bilayer being quasiliquid all the way down to 235 K.