Computational Approaches to Investigate How Biological Macromolecules Can Be Protected in Extreme Conditions
Authors are with Johnson Research Foundation, Dept. of Biochemistry and Biophysics, School of Medicine, Univ. of Pennsylvania, Philadelphia, PA 19104. Direct inquiries to author Vanderkooi (E-mail: email@example.com).
Authors are with Johnson Research Foundation, Dept. of Biochemistry and Biophysics, School of Medicine, Univ. of Pennsylvania, Philadelphia, PA 19104. Direct inquiries to author Vanderkooi (E-mail: firstname.lastname@example.org).
ABSTRACT: Water is required to hydrate molecules, but under cold conditions water freezes and under dehydrating conditions water evaporates—thus presenting a dilemma for organisms that live in extreme environments. Organisms have developed various strategies for protection against extreme temperatures and dehydration. In this review, we describe how the interaction of water and 2 natural cryoprotectants, namely glycerol and sugars, can be studied at the molecular level. Techniques using infrared spectroscopy and computation are described. In the case of glycerol, H-bonding of water to the OH groups of glycerol limits the amount of water available to form ice and prevents crystallization at low temperatures. For aldohexopyranose sugars, the different isomeric forms have different water H-bonding networks, which are consistent with isomeric-dependent activities. By studying the strategies used in nature, derivatives for use in food preservation can be more readily developed.
The metabolic pathways of strawberries, wheat, insects, dogs, humans, and most of the life on earth are basically the same—having a very similar set of enzymes catalyzing analogous reactions. Therefore, it would follow that all organisms require a similar environment for survival. Yet, this is not the case. Whereas most organisms survive within a narrow range of living conditions, some kinds of life can survive extremes by adapting to situations that would otherwise be lethal. When an organism goes into a senescent state that enables it to survive extreme temperature, desiccation, or osmotic stress, it faces the same problem that we encounter in storing food. How are the macromolecules required for life maintained over long periods of time under adverse conditions such that the recovery of cellular architecture and molecular structure is possible?
Nature has developed special mechanisms to protect the important molecules of biology. By looking at these mechanisms, we can begin to understand how to engineer systems that will preserve food products and compounds such as vaccines and medicines. A key player in the story is water. Water is required to hydrate molecules; for instance, the folding of proteins into active molecules cannot proceed without water. But for the organism facing severe environmental stresses, water is not an ideal solvent. At cold temperature, water freezes and forms disruptive ice crystals that damage cellular structures. In conditions of low humidity or high osmolality, water leaves cells, thus altering the hydration of macromolecules and their ability to participate in reactions necessary for life.
There exists more than 1 form of protection against these vagaries of water. Therefore, an interesting question arises: What common features do the various protectants have that enable them to protect biological structures from damage associated with water stress? This seemingly straightforward question is not easily addressed by experimental techniques—especially when the interactions of interest involve water molecules. As a result, computational methods have gained popularity for investigating how the interactions between protective solutes and water may be responsible for the protection of organisms against environmental extremes.
In this review, we explain the general principles behind protection from water stress and describe some of the common classes of protectants found in nature. Then, we demonstrate how computational techniques can be employed to determine at the molecular level the principles behind water stress protection. Finally, we give examples of how the protective solutes glycerol and small sugars such as glucose interact with water as determined by computational methods.
General Actions of Compensatory Solutes
Protective mechanisms are found throughout prokaryotic and eukaryotic organisms ranging from bacteria and yeast to plants (Chen and others 2006) and animals (Storey 1997). Many organisms use low molecular weight organic solutes such as polyhydric alcohols, carbohydrates, and simple amino acids for these functions. Common to these molecules, collectively referred to as compensatory solutes, are characteristics that make them ideal for protection against water stress: high solubility, nontoxicity, compatibility with other biological molecules, and efficient production from metabolic pathways. Compensatory solutes are accumulated in high concentrations (up to 2 M) and provide protection from water stress by colligatively raising the osmolality of body fluids or intracellular solutions (Storey 1997). As the environmental temperature is lowered below 0 °C, water molecules crystallize to form ice until the osmolality of the unfrozen water fraction yields a solution melting point equal to the current temperature. Therefore, in organisms lacking a protective mechanism, a larger fraction of extracellular water is converted to ice at a given subzero temperature, which leads to the damage of cells and tissue caused by expansion of the solution. Additionally, intracellular water is lost to the extracellular ice mass, thus producing a high osmotic stress on internal components of the cell.
Accumulation of compensatory solutes within the body fluids or intracellular spaces can mitigate the damages caused by these stresses (Mazur 1984). When accumulated in body fluids, compensatory solutes limit the amount of extracellular water that can crystallize to form ice at low temperatures. With less ice produced, volume expansion of the solution is limited and the mechanical stress on cells and tissues is lessened. Additionally, the initial rate of formation of ice upon nucleation is diminished due to changes in the microstructure of the solution that slow crystal propagation and reduce the size of ice crystals. Smaller ice crystals are less likely to damage cellular structure.
Alternatively, accrual of compensatory solutes within intracellular spaces protects biological macromolecules from the loss of cellular water during freezing, desiccation, or osmotic stress. Although compensatory solutes allow some water loss to extracellular spaces, much of the intracellular water is retained through bonding interactions with these solutes. This interaction maintains proper hydration of biological macromolecules while at the same time limiting the ability of the “bound” water to leave the cell or to crystallize.
The importance of compensatory solutes in biological macromolecule stabilization is described through preferential hydration (Shimizu and Smith 2004). In preferential hydration, proteins are stabilized as a result of entropically favored surface minimization stemming from changes in water activity upon addition of cosolvent molecules (Brandts and Hunt 1967; Gerlsma 1968; Gekko and Timasheff 1981; Betting and others 2001). As small compensatory solutes, or cosolvents, are added to the solution, the distribution of water and cosolvent molecules at the protein surface changes relative to the bulk solution. For compensatory solutes, the concentration of cosolvent molecules is lower in the direct vicinity of the protein than in bulk solution—water molecules preferentially hydrate the protein. As a result, exposure of protein hydrophobic regions becomes less thermodynamically favorable in the water–cosolvent mixture than in water alone. While the specific molecular forces behind preferential hydration are yet unknown, the strong affinity of compensatory solutes for water may be involved in preferential hydration by indirectly influencing water structure in the hydration shell of the protein (Galinski and others 1997). Although most organisms do not encounter temperature or dehydration extremes severe enough to cause macromolecular destabilization, the compensatory solutes used for these types of stress protection have been shown capable of providing stabilization (Crowe and others 1987).
Meanwhile, stabilization of cell membranes is thought to occur through more direct interaction of compensatory solutes such as glucose and trehalose with the polar head groups of membrane phospholipids. Two nonmutually exclusive models exist for how these solutes are able to stabilize membranes. Under the water replacement model, sugar molecules replace water molecules bound to the phospholipid head groups and stabilize these groups through H-bonding (Crowe and Crowe 1992), while the vitrification model suggests that sugar molecules form a high viscosity, glassy state surrounding the phospholipids which prevents molecular mobility (Koster and others 1994). Regardless of which mechanism is active in a given case, compensatory solutes limit or prevent damage to the cell membrane caused by fusion or lipid phase transitions that can lead to the loss of cellular contents particularly in freeze/thaw situations—referred to as “nutrient drip” in food science.
Whether water stress-protection stems from reduction of freezable water or stabilization of biological macromolecules, how compensatory solutes interact with water is central to understanding how organisms survive environmental extremes. Synthetic protectants to aid in food quality and preservation may be more readily designed with these details of the water–solute interaction known.
Examples of Compensatory Solutes Present in Nature
As mentioned previously, polyhydric alcohols are 1 class of water stress protectants found in nature. The polyhydric alcohol glycerol, containing 3 hydroxyl groups attached to an alkyl backbone, is one of the least structurally complex, yet most prevalent, protectants. As an intermediate of the glycolytic pathway produced from glycerol-3-phosphate, glycerol is a simple metabolite found in most organisms. However, in many organisms, glycerol is also utilized as a cryoprotectant. From a chemical point of view, the highly hydroxylated nature and flexibility of glycerol are fundamental to its ability to interact with water. In a mixture of water and glycerol, the infrared (IR) absorption of the OH stretch from each compound cannot be distinguished (Zelent and others 2004). Since the frequency of the OH stretch is a function of the force constant of the bond, the OH groups of water and glycerol have very similar chemistry. In other words, the OH of glycerol acts very much like the OH of water.
Glycerol is a major solute in the hemolymph of insects (Wyatt and Meyer 1959) and is known to protect both their larval and pupal developmental stages (Kukal and others 1988; Izumi and others 2006). Other larger polyhydric alcohols such as sorbitol and mannitol are also effective at providing water stress protection in small insects (Salvucci and others 1998). For example, larvae of the freeze-tolerant gall fly, Eurosta solidaginis, exhibit seasonal accumulation of glycerol and sorbitol as a protection against lower winter temperatures. At the peak of polyhydric alcohol synthesis, levels of glycerol and sorbitol reach 250 and 150 mM, respectively (Joanisse and Storey 1994). The concentration of glycerol and the specific interactions this molecule makes with water have been shown to greatly affect water structure and the ability to form ice (Dashnau and others 2006).
Simple monosaccharides are widely used by nature to stabilize cells. These sugars resemble polyhydric alcohols in that they contain both OH groups and an alkyl backbone. However, monosaccharides also contain either a ketone or aldehyde group. The aldehyde and keto groups of sugars allow them to cyclize—therefore producing a more rigid structure than that found in polyhydric alcohols. Consequently, as will be discussed later, the structure of these molecules allows the formation of intramolecular H-bond networks that can greatly impact interaction with water and surrounding water structure (Dashnau and others 2005a).
Glucose, one of the most common 6-carbon monosaccharides found in biology, is used by the wood frog, Rana sylvatica, for water stress protection (Storey 1997). Following receipt of peripheral stimuli, glucose production in the liver increases dramatically and the molecule is distributed throughout the body of the organism. Maximal levels of glucose reach concentrations of 200 to 300 mM in core internal organs and 25 to 50 mM in more peripheral organs (Storey 1997). The differential deposition of glucose in the organs allows the wood frog to freeze and thaw in a beneficial pattern (Rubinsky and others 1994). Freezing occurs gradually from the outside in—allowing core internal organs the time to eliminate water to body cavities and thus reduce water in extracellular spaces where it can damage cells (Costanzo and others 1993). Due to the high concentration of glucose in core organs, thawing occurs from the inside out, thus allowing vital functions to resume ahead of peripheral organ activation.
More complex sugars such as the disaccharide trehalose are also used for protection. Trehalose is a disaccharide composed of 2 glucose molecules joined by an α–(1,1) linkage. The reducing end of glucose is tied up in the glycosyl linkage, and for this reason trehalose is a nonreducing sugar. Although trehalose is composed of 2 glucose units, its protective properties have been found to differ drastically from that of the monosaccharide. The importance of trehalose is evident by the fact that it is used in many species, including bacteria, yeast, fungi, insects, other invertebrates, and plants (Elbein 1974). The larva of insects Trabutina mannipara and Najacoccus sperentinus, found in the Sinai Desert and other places in the Middle East, secrete it (Leibowitz 1944) and it is thought by some to be the manna described in Exodus of the Bible. Trehalose is part of the heat shock response in yeast (Nwaka and Holzer 1998) and it participates in the stabilization of heat shock transcription factor (Bulman and Nelson 2005).
This brief summary shows that many phylogenetically diverse species have a mechanism for cryoprotection and/or prevention of irreversible changes during desiccation. Yet, relatively few types of molecules are used as cryoprotectants. The molecules have the properties that they are uncharged or zwitterionic and they are derived from common metabolic intermediates. They are also stable chemically. For all these molecules, their protectant action must in some way be due to their interaction with water.
Computational Methods Used to Study Solute-Induced Water Structure
Computational methods have become increasingly popular for studying the interaction of solutes with water at the molecular level. A primary reason behind the increased use of computational procedures is that these methods have become more accessible to researchers; they can now be carried out using inexpensive, but powerful, desktop computers equipped with commercially available software. In this section, we provide an overview of general computational methods that can be followed to gain insight into how protection occurs at the molecular level. The flowchart presented in Figure 1 outlines the major steps in this process. In the next section, we show how the output from these methods can be used to determine how water interacts with protective solutes—glycerol and carbohydrates.
When using computational chemistry methods, there are 2 major processes that are performed. First, one must obtain or develop a structure for the molecule of interest. In the section on structure development, we outline common ways in which a structure can be constructed. Once a structure is developed, one must obtain a trajectory that describes positions of atoms in the system over time. The section on trajectory production describes the steps required to produce a trajectory.
A number of resources are available from which to obtain an initial structure. Generally, molecules can be obtained by either building the structure manually or by obtaining coordinates through experimental methods that include nuclear magnetic resonance (NMR) and X-ray crystallography. The method for generating a structure that can be used in computational studies is dependent mainly upon the size of the molecule of interest. For smaller macromolecules such as sugars and single amino acids, programs such as PyMOL (DeLano 2002), ACD/ChemSketch (2005), and ChemDraw can be used to manually build structures. Then, quantum mechanics programs can be used to optimize these structures as described in the section on building and optimizing a structure with quantum mechanics.
For larger molecules such as proteins, coordinates from NMR or X-ray crystallography structures are often available in protein data banks. However, if a structure is obtained through a source such as the Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank, the file containing atomic coordinates should be checked to ensure all hydrogen and heavy atoms are present. If these atoms are absent, they will need to be added before the structure is optimized. Also, if a complete structure is not available, one can be generated through homology modeling in which a structure for the protein of interest is created by modeling after a sequentially or structurally similar protein. Since the size of these structures makes optimization using quantum mechanics too computationally demanding, molecular mechanics programs can be used to minimize structures as described in the section on building and minimizing a structure with molecular mechanics.
Building and optimizing a structure with quantum mechanics Gaussian (Frisch and others 2004), a commercially available quantum mechanics program, implements quantum mechanics theory to predict properties from structures and energies to spectra. A guide to understanding this program is Exploring Chemistry with Electronic Structure Methods (Foresman and Frisch 1996). This program can be used to prepare small structures (of less than 50 atoms) for later use in molecular mechanics programs. Generally, once a rough starting structure for the molecule of interest is obtained or constructed, the following steps are performed in Gaussian: optimization, single-point energy stability calculation, and frequency calculation.
The goal of optimization is to find the equilibrium structure of a molecule that corresponds to an energy minimum on the molecule's potential energy surface. To perform a geometry optimization, an appropriate level of theory and basis set are selected for the calculation. A comprehensive treatment of quantum mechanics theory as related to computational chemistry and pertaining to the selection of theory and basis sets can be found in Essentials of Computational Chemistry: Theories and Models (Cramer 2004). Geometry optimization is then performed by applying a force to the molecule and then evaluating the energy and gradient, or first derivative of the energy, at that point. When the molecule is at a minimum on the potential energy surface, the forces are zero. If the molecule meets optimization criteria following optimization, that is, if forces and displacements fall within a selected threshold, the structure is considered at a local minimum. However, if the geometry optimization has not converged, the procedure is iterated until convergence is reached.
Following geometry optimization, a single-point energy stability calculation is performed to determine whether a lower energy wave function is available for the optimized structure. If the wave function is not stable, then optimization must be performed again as the structure does not correspond to a ground state. However, if it is stable, then frequency calculations can be performed.
Frequency calculations predict the direction (eigenvectors) and magnitudes (eigenvalues) of nuclear displacement by analyzing the second derivative of the energy with respect to the positions of nuclei. These calculations are run on optimized structures using the same level of theory and basis set as used in the geometry optimization. The results of the calculation include information about frequencies, intensities, and normal modes. This information is used to determine if the molecule is in the minimum energy ground state. The number of imaginary (negative) frequencies present in the output gives an indication of whether the structure is at minimum (no imaginary frequencies), represents a transition structure, or is at a saddle point (imaginary frequencies present). For example, if the molecule of interest is expected to be in the ground state, but the frequency calculation reveals imaginary frequencies, then the structure corresponds to a saddle point rather than a minimum. In this case, the molecule should be reoptimized after distorting the molecule structure along the normal mode associated with the imaginary frequency.
Generally, once a stable, optimized, ground state structure is obtained, it can be imported into a molecular mechanics program for use in trajectory production.
Building and minimizing a structure with molecular mechanics As explained earlier, quantum mechanics programs such as Gaussian are useful in preparing small molecules structures for molecular dynamics. However, this program is not appropriate for preparation of larger molecules such as proteins. In these cases, molecular mechanics is used for energy minimization. CHARMM (Chemistry at Harvard Macromolecular Mechanics) (Brooks and others 1983) is one such program used for molecular mechanics. As is the case in quantum mechanics geometry optimizations, the purpose of energy minimization in molecular mechanics is to find the molecular conformation that corresponds to the lowest potential energy state of the system. There are a number of minimization algorithms available in CHARMM for which to accomplish this task—one should consult the accompanying program documentation for a full description of each method.
It should be noted that, in most cases, the minimization process in molecular mechanics is not as essential as optimization in quantum mechanics. Since the molecules will be free to move throughout later steps and will thus sample a number of conformations, it is not imperative that an energy minimum be achieved. Rather, minimization is used primarily to achieve a reasonable starting geometry with which to use in the equilibration and dynamics steps. However, minimization is particularly useful in cases where missing atoms were manually built into an initial structure or in cases where structures were created through homology modeling. In these cases, minimization will ensure that the structure is in an acceptable configuration. Minimized structures can then be used to generate a trajectory.
CHARMM (Brooks and others 1983) can be used to generate a trajectory. Following input of the structure into the program, the main steps in running CHARMM are outlined briefly below and include topology/parameter construction, solvation, minimization, initialization, heating, equilibration, and dynamics. A more comprehensive source of information on principles and syntax can be found in the online documentation files for the program (CHARMM 2003).
Construct topology and parameter files In CHARMM, topology and parameter files are used to apply empirically derived forces in the calculation of molecular trajectories. For most structures, all parameters needed for the calculation are provided within the associated program files. However, for structures that contain unusual residues or functional groups, parameters may not be present. In these cases, experiment or quantum mechanics calculation must be used to generate the missing values—force constants for bonds, angles, dihedrals, and torsions. This calculation can be performed in Gaussian. For example, to obtain the force constant associated with a bond stretch, the potential energy associated with a number bond distances can be calculated. The resulting potential energy surface can then be fit to a potential function to determine the force constant associated with the stretch. Once all parameters are defined, trajectory development can proceed.
Solvation and minimization To this point, the system has consisted only of the molecule of interest. However, as most molecules are solvated by water, explicit water molecules are added to the system. This step is done by superimposing a pre-equilibrated box of water molecules on the molecule and then removing overlapping water molecules. Then another round of minimization is performed on the entire system to adjust the water molecules with respect to the molecule.
Heating and equilibration Before a molecular mechanics simulation can be performed to obtain a trajectory, the system must be heated to the temperature at which the system is to be studied. This step is necessary since the minimized structure does not account for the structure of the molecule at nonzero temperatures. Therefore, the temperature of the system is slowly increased by incrementally increasing the velocities assigned to each atom until the system reaches the temperature at which the simulation will be run. Once the required temperature is reached, equilibration is then performed by allowing the simulation to evolve over time in order to distribute kinetic and potential energy equally throughout the system.
Dynamics Once the system has reached equilibrium, the system is then further studied to obtain a set of trajectories. The dynamics run is very similar to the equilibration run. Basically, the equations of motion are further integrated using the same time step. The main differences are the dynamics run represents the equilibrated system and the duration often lasts for much longer in order to sample a large enough portion of the trajectory for analysis. However, many of the details of dynamics are specific to the questions aimed to be addressed.
Analyzing the Interaction between Water Stress Protectants and Water
After a successful run of molecular dynamics, the resulting trajectory of the simulation can be analyzed in any number of ways. The method one chooses to investigate the trajectory is dependent upon the experimental data with which it is to be compared. Many methods of trajectory analysis produce data that are analogous to that obtained through experiment. For instance, neutron scattering experiments yield information on the distribution of distances between 2 atom types. These distributions, or radial distribution functions (RDFs), outline the probability of finding a certain atom type around a central molecule. Since a trajectory contains the positions of all atoms in the simulation, it is simple to calculate a theoretical RDF for use in comparison. Other experimental data, such as those obtained through spectroscopic methods, can also be simulated with computation methods. As the focus of our lab is spectroscopic in nature, we describe below how computational analysis has been used to specifically complement these types of experiments in the study of 2 cryoprotectant systems: glycerol/water and carbohydrate/water.
Experimentally, infrared spectroscopy has been used to study the nature of H-bond networks in solution. This method is used since the OH stretch (3000 to 3600 cm−1) and OH bend (approximately 1650 cm−1) frequencies are indicative of the overall strength of H-bonding within a sample. For the OH stretch, higher frequencies are generally associated with weaker, more geometrically distorted H-bonds, while lower frequencies correspond to stronger, more linear H-bonds. The change in frequency associated with H-bond strength is explained in terms of intramolecular force constants. Introduction and orientation of an H-bond acceptor weakens the intramolecular force constant between the H-bond donor and the atom with which it is covalently bound. As a result, the covalent bond is lengthened and the potential energy associated with the corresponding stretch decreases. The decrease in potential energy is detected as a decrease in infrared stretch frequency.
Infrared spectroscopy has been used to examine the nature of the H-bond network in glycerol/water mixtures across a range of temperatures to determine how the presence of glycerol affects ice crystallization (Dashnau and others 2006). Figure 2 illustrates how glycerol concentration affects the frequency and width of the OH stretch as a function of temperature.
At room temperature, the OH stretch exhibits a concentration dependent change in frequency; the position of the OH stretch shifts from 3403 cm–1 in water to 3337 cm−1 in glycerol, with values for glycerol/water mixtures falling in between this range. This result suggests that the H-bonds of the OH network of glycerol and its mixtures with water are arranged more linearly than those of water at room temperature. As the temperature is decreased, a phase change is clearly exhibited in the water sample between 260 K and 265 K (Figure 2, top graph). This phase change corresponds to the formation of ice crystals containing a large number of nearly linear H-bonds. The shift of the OH stretch to low frequency (below 3300 cm−1) and a narrowing of the band indicate that the OH network contains mostly linear H-bonds and vibrationally relaxes at a relatively slower rate than seen in water at room temperature. With the addition of glycerol (both in aqueous mixtures and in pure form), the sharp, low frequency peak of ice does not form—there is no detectable crystallization of water. However, the frequency of the OH stretch shifts to a position that is nearly the same as that of ice in the pure water sample. This result suggests that with a decrease in temperature, more low energy, linear bonds are forming at the expense of high-energy, distorted bonds. Yet the broad width of the peak suggests that there is not a substantial change in the relaxation rate relative to room temperature. This result is in contrast to that of pure ice and indicates that the ability of water to relax is maintained even at low temperature when in the presence of a cryoprotectant. Preservation of these dynamics may explain why proteins are stabilized by glycerol and sugar glasses (Wright and others 2003; Dashnau and others 2005b).
The IR spectra provide a good starting point for analyzing glycerol water mixtures. However, for a better molecular understanding of the H-bond network, computational methods are needed. The group of Sharp and others have pioneered the use of computation for the analysis of H-bond angle distributions (Madan and Sharp 1996, 1999; Sharp and Madan 1997; Gallagher and Sharp 2003a, 2003b). Using molecular dynamics trajectories such as those that are obtained by following methods similar to those discussed in previous sections of this review, the angles formed by water molecules surrounding a solute can be studied. The strength of the H-bond network detected by IR spectroscopy by way of OH stretch frequency shifts is related to the orientation of H-bond acceptor relative to H-bond donor. Therefore, by determining the distribution, P(θ), of H-bond angles (θ) in a simulation, where θ is defined as the angle formed by H—O…O, information analogous to that produced by IR can be obtained.
The H-bond angle distributions can be calculated by running the program PRAM (Gallagher and Sharp 2003a, 2003b) following generation of a molecular dynamics trajectory. A brief description of this program as it was used in the study of glycerol/water mixtures (Dashnau and others 2006) follows. In this program, all solute heavy atoms are classified according to their partial charge; a partial charge magnitude less than 0.35 is categorized as nonpolar and a magnitude equal to or greater than 0.35 is categorized as polar. Water molecules are then classified as being within the 1st solvation shell of the solute if they are within a specific distance cutoff of a solute atom (that is, 5.6 or 3.4 Å of a C or O atom, respectively). Water molecules not part of the 1st solvation shell are given the designation of bulk water (bulk). The P(θ) for bonds formed between pairs of water molecules in the 1st solvation shell or between pairs of bulk water molecules is then calculated. The P(θ) for 1st solvation shell water pairs is further broken down by the types of solute atoms inducing water structure. As demonstrated in Figure 3, in a 1st solvation shell water pair, if one of the water oxygen atoms (A) is closest to a nonpolar solute atom and the other water oxygen (B) is closest to a polar solute atom, the associated θ is classified as part of the nonpolar–polar (np) distribution.
Other possible distributions include nonpolar–nonpolar (nn) and polar–polar (pp) interactions. In addition to accumulating P(θ) for water–water H-bonds, H-bonds involving glycerol OH groups are also included. These distributions include the θ formed by the following interactions: glycerol–water (gw), intramolecular glycerol–glycerol (ggi), and intermolecular glycerol–glycerol (gg) interactions.
Figure 4 shows a typical H-bond angle distribution produced by simulation. The distribution of H-bonds tends to be bimodal, with a lower angle peak centered on 12 degrees and a higher angle peak centered on 52 degrees. These peak positions are seen for most H-bond interactions involving water at room temperature. Geometrically constrained H-bonds such as those that occur within larger molecules such as glycerol or sugars may not exhibit peaks at these locations.
To simplify comparison between dynamics runs, the ratio of the area under the 1st peak to the area under the 2nd peak is used to assess the overall nature of an H-bond population. Therefore, the A1/A2 ratio increases when the number of lower angle H-bond increases relative to the number of higher angle H-bonds. The benefit of this type of analysis is that subsets of H-bond interactions can be selected and analyzed separately rather than surveying the H-bond network in totality, as is done in IR spectroscopy.
By combining the results from computation with those from experimental methods, a number of conclusions can be drawn about the impact of glycerol concentration on water structure and crystallization. Figure 5 shows the average number of H-bonds made by water for specific interaction types as a function of glycerol concentration. First, analysis of the number of H-bonds per water molecule shows that disruption of the percolated water network corresponds to the concentration at which glycerol is present in sufficient amount to prevent ice crystallization. The percolation threshold is defined as the average number of H-bonds per water below which there is an insufficient number of water molecules to create a spanning network of water surrounding a solute molecule. The 2-dimensional percolation threshold is cited as 2.0 to 2.3 H-bonds per water (Oleinikova and others 2005). Therefore, there must be on average 2 H-bonds/water at the surface of a solute for the water network to be contiguous throughout the sample. The examination of Figure 5 shows that for the solvation shell surrounding glycerol molecules, this threshold is crossed at a molar fraction of glycerol of about 0.27—corresponding to one of the lowest concentration glycerol/water mixtures that prevent ice crystallization. At concentrations above this point, bulk water is no longer present in large quantities and the network formed by solvation shell water molecules is significantly disturbed. Without the presence of a contiguous water network, crystallization cannot occur at low temperatures.
Additionally, Figure 6 builds on the results of IR spectroscopy by identifying the main type of H-bond responsible for changes in the OH frequency with glycerol concentration. As glycerol concentration increases, only the glycerol–water H-bond shows a larger distribution of angles in more linear arrangement. This result is demonstrated by the increase in A1/A2 ratio that corresponds to OH stretch shifts to lower frequency in IR. Therefore, in addition to the effect of bulk water depletion, change in geometry of the H-bond between glycerol and water molecules supports a network that is similar to that of ice, but does not crystallize at low temperature.
From the computation studies of glycerol/water mixtures, a number of conclusions can be reached about the impact of glycerol on water structure and how this relates to its protective properties. As glycerol is a colligative solute, concentration plays a major role related to the availability of bulk water for crystallization. There exists a threshold above which the amount of water present in solution is insufficient to fully hydrate glycerol molecules and the network of water is disrupted. Ice is unable to form in these solutions. In addition to this property, the nature of H-bonds between glycerol and water change as concentration is increased. As the fraction of glycerol increases, the interaction between glycerol and water strengthens; H-bonds between these molecules become stronger. This result supports the idea that strong interactions between water and cosolvents may indirectly affect water surrounding proteins in preferential hydration.
While glycerol represents one of the simplest polyhydric alcohol structures known to present protective properties, there are more complicated polyhydric alcohols found in nature that also provide these beneficial characteristics: carbohydrates. Carbohydrates, like glycerol, consist of an alkyl backbone and hydroxyl groups. However, carbohydrates are less flexible than glycerol since these molecules can form ring structures. Therefore, while the chemical composition within a class of monosaccharides is identical, the orientation of OH groups differs across stereoisomers. For example, the aldohexopyranose sugars glucose and galactose differ only in the position of the OH group on the 4th carbon. In glucose, OH-4 is oriented equatorially, while in galactose this same group is in an axial position relative to the ring.
The ring structure of carbohydrates coupled with the large number of OH groups lends to the formation of intramolecular H-bond networks and introduces the concept of H-bond cooperativity. Under H-bond cooperativity, the H-bond formed between a proton donor and proton acceptor is strengthened as additional H-bonding groups are added. For example, the H-bond between a proton donor A-H and proton acceptor B will become stronger when another group, A*-H*, is H-bonded to A-H. With the large number of OH groups in a carbohydrate, extensive intramolecular H-bond networks can form.
López de la Paz and others (2002) have used NMR and IR to study how the orientation of OH groups in carbohydrate derivatives influences H-bond cooperativity in the molecule. While this study was aimed at understanding carbohydrate recognition processes, the results are also relevant to understanding carbohydrate solvation since these intramolecular H-bond networks interact with the H-bond network of aqueous solution. A number of principles governing the formation of cooperative intramolecular H-bond networks were developed based on this research. First, the relative position of OH groups influences H-bond stability. Namely, 1,3-diols (syndiaxial) are more stable than cis-1,2-diols (cis-vicinal) or trans-1,2-diols (trans-vicinal). Second, other groups on the molecule, such as the anomeric OH group at the 1st carbon and 5-O, can affect cooperativity by influencing the directionality of networks. For example, the position of the 2-OH next to the anomeric center leads to higher acidity of the 2-OH group and polarizes H-bonding so that 2-OH acts as an H-bond donor to the 4-OH acceptor. Finally, H-bond networks that terminate within the molecule are preferentially formed in solution over networks ending in a free OH group when OH groups are in a position to do so.
Computational approaches have been used to verify and extend results of the experimental analysis described above to include the effect of H-bond cooperativity on solvation (Dashnau and others 2005a). The complete set of 16 aldohexopyranose sugars was chosen for analysis as they represent all possible OH combinations (axial/equatorial) for the 6-carbon D-sugars. Consequently, a full factorial analysis of the impact of OH position on 1st solvation shell H-bond structure could be performed. The sugars were analyzed using PRAM, as described earlier, and by performing electrostatic calculations. Briefly, electrostatic calculations were performed by running Gaussian-optimized structures through the DelPhi application, Qnifft. This program is described elsewhere in more detail (Sitkoff and others 1994; Sharp 1995).
The findings of this study confirmed the principles established by Lopez de la Paz and others (2002) and further explained the impact of intramolecular H-bond networks on surrounding water. Figure 7 shows the electrostatic potential maps of 4 representative sugars. These sugars were chosen as they show progressive conversion of OH groups from equatorial to axial orientation at positions 4, 2, and 3 of the sugars. As is illustrated by this figure, the presence of a syndiaxial interaction between OH-2 and OH-4 (Figure 7, talose and idose) induces stronger intramolecular H-bonds and a dramatically less multipolar potential surface than that produced when this interaction is not present. In essence, the less multipolar surface presents as relatively more hydrophobic due to the decreased ability to form H-bonds with the surrounding water molecules. As a result, the water network surrounding the sugar OH groups with less multipolar electrostatic potentials is relatively more linear than the water network surrounding the OH groups of more balanced multipolar sugars.
This phenomenon is explained as follows. In biological molecules containing a large number of OH groups, as additional intramolecular H-bonds are added to the OH network, the ratio of available H-bond donors to acceptors decreases. As the balance of H-bond donor and acceptor characteristics decreases, these regions of the molecule are less able to H-bond with surrounding water molecules—resulting in patches of relatively more ordered, linear H-bonds within the 1st solvation shell. Longer, directionally oriented networks of OH groups produce even stronger effects than shorter networks. Additionally, where networks terminate in a free OH group, surrounding water molecules are better able to H-bond and thus the distribution of H-bond angles formed between water molecules in the 1st solvation shell is more distorted. The distorted H-bonding of the 1st solvation shell indicates a disruption of strong water–water H-bonds. Distortion of these bonds will make water molecules less capable of interacting with each other at low temperature to form ice crystals.
In monosaccharides, 2 main networks can form when syndiaxial groups are present—those involving axial OH-2/4 groups or those involving axial OH-1/3 groups. However, networks involving OH-2/4 produce more linear water structure than those with OH-1/3. This result is likely because OH-2/4 can further H-bond with OH-6 and O-5 to form a long, terminated network lacking an H-bond donor. The networks formed by OH-1/3 are not within bonding distance of the OH-6 group and are thus shorter and more likely to pair with vicinal groups for a weaker H-bond or to contain a free OH group. These networks can serve as both H-bond donor and acceptor to surrounding water molecules and thus are more able to distort water structure in surrounding layers.
Although the OH-1 and OH-3 groups are inefficient in producing a long H-bond network, these groups do indirectly influence H-bonding. When these groups are equatorial, a hydrophobic patch can form on 1 side of the sugar. Figure 8 illustrates for the common biological sugars (glucose, maltose, and galactose) the electrostatic potential surrounding 1 face of the structure.
The electrostatic potential is less complex and less multipolar than that found on the opposite face of these sugars (not shown). However, the water structure surrounding the CH groups that comprise this patch is relatively more polar-like than in sugars where the patch is not present; that is, water structure is more distorted.
The interesting result from these studies is that the sugars most common to nature are those that lack extensive H-bond networks and contain a hydrophobic patch on 1 face of their structure: glucose, galactose, and mannose. The different complexity of the electrostatic potential and water structure surrounding the 2 faces of these sugars suggests possible reasons why glucose may be used as a protectant. Lack of a strong intramolecular H-bond network allows glucose to be easily solvated, but at the same time the presence of a hydrophobic patch would provide a region capable of interaction with other biomolecules (Dashnau and others 2005a).
Survival of many species depends upon a narrow range of temperature and requirements of water. Yet many organisms can exist at extreme temperatures and extremely low levels of hydration. These organisms incorporate a variety of substances that either change the properties of water or replace some of the interactions between water and biomolecules; it is by this means that they can survive. By examining the principles whereby some species can endure dehydration and extreme temperature, one can perhaps modify important agricultural species and their products to withstand a wider range of conditions. Based upon examples in nature, it is clear that there is not just 1 method in which to provide such protection.
What is clear is that one needs to think of not only the substance but also its solvation and the ways it changes water. In this review, we described a computational approach to gain insight in how water and biomolecules interact. Glycerol is used by some species to prevent freezing. The question that was asked was at what concentration must glycerol be present to prevent ice crystallization. The model that was obtained suggests that H-bonding of water to the OH groups of glycerol is required to prevent ice crystallization. The model was substantiated by infrared measurements of water and glycerol as a function of temperature. The 2nd example was the study of aldohexopyranose sugars in water. The sugar molecules had the same chemical formula and the same functional groups. Yet only glucose and galactose are found to any great extent in nature. Arrangement of water molecules is quite different around the different isomeric forms of sugars. The view obtained is that the reactivity of a particular sugar is due to water–sugar interactions, which is a function of the placement of the functional groups on the molecule.
This project was supported by the Natl. Research Initiative of the USDA Cooperative State Research, Education and Extension Service, Grant 2005-35503-16151. We thank the following people for making helpful comments on this paper: Drs. K. A. Sharp, N. V. Nucci, N. Scott.