Modeling hydration mechanisms of enzymes in nonpolar and polar organic solvents Nuno M. Micae ˆ lo and Cla ´ udio M. Soares Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Universidade Nova de Lisboa, Oeiras, Portugal The ability of enzymes to work in media other than water is now widely accepted, and is the basis of exten- sive basic research on enzyme catalysis and many bio- technological applications [1]. The fact that most enzymes have evolved in an aqueous environment in living cells does not mean that they cannot be trans- ferred and be functional in a completely different kind of medium [2–4]. Our recent molecular modeling stud- ies have depicted the molecular mechanism of the effects of different hydration percentages on the struc- tural [5] and enantioselective properties of enzymes [6] when placed in organic solvents such as hexane. Many experimental studies in the field of nonaqueous enzy- mology have focused their attention on demonstrating that the amount of water in the organic medium plays an important role in controlling the catalytic properties of the enzymes [5–9]. These studies have shown that when enzymes are used in organic solvents, water reta- ins its fundamental role in controlling the physical properties of the enzyme, and this role probably can- not be taken by other solvent. In such systems, the effect of water is complicated to investigate, because this solvent is distributed in several phases; it can be in the vapor phase, adsorbed to the support material, dis- solved in the organic liquid phase, or bound to the enzyme [10]. Of the total water added to the organic medium, the effect of the organic solvents on the enzyme seems to be primarily due to the water that is bound to the enzyme [7,11]. This bound water is usually measured experimentally in terms of the thermodynamic activity of water, assuming that, for enzymatic reactions Keywords enzyme hydration; organic solvents; protein modeling; water clusters Correspondence C. Soares, Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Universidade Nova de Lisboa, Av. da Repu ´ blica, Apartado 127, 2781-901 Oeiras, Portugal Fax: +351 21 4433644 Tel: +351 21 4469610 E-mail: claudio@itqb.unl.pt Website: http://www.itqb.unl.pt/pm (Received 21 December 2006, revised 1 March 2007, accepted 8 March 2007) doi:10.1111/j.1742-4658.2007.05781.x A comprehensive study of the hydration mechanism of an enzyme in non- aqueous media was done using molecular dynamics simulations in five organic solvents with different polarities, namely, hexane, 3-pentanone, diisopropyl ether, ethanol, and acetonitrile. In these solvents, the serine protease cutinase from Fusarium solani pisi was increasingly hydrated with 12 different hydration levels ranging from 5% to 100% (w ⁄ w) (weight of water ⁄ weight of protein). The ability of organic solvents to ‘strip off’ water from the enzyme surface was clearly dependent on the nature of the organic solvent. The rmsd of the enzyme from the crystal structure was shown to be lower at specific hydration levels, depending on the organic solvent used. It was also shown that organic solvents determine the struc- ture and dynamics of water at the enzyme surface. Nonpolar solvents enhance the formation of large clusters of water that are tightly bound to the enzyme, whereas water in polar organic solvents is fragmented in small clusters loosely bound to the enzyme surface. Ions seem to play an import- ant role in the stabilization of exposed charged residues, mainly at low hydration levels. A common feature is found for the preferential localiza- tion of water molecules at particular regions of the enzyme surface in all organic solvents: water seems to be localized at equivalent regions of the enzyme surface independently of the organic solvent employed. Abbreviations FF, force field; MD ⁄ MM, molecular dynamics ⁄ molecular mechanics; SPC, single point change. 2424 FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS carried out in different media for a certain enzyme at fixed water activity, the enzymes have equivalent amounts of water bound [12]. This approach to expres- sing water content in organic solvents has been a standard in nonaqueous enzymology, simplifying the interpretation and prediction of changes in enzyme performance. It is commonly reported that the same enzyme placed in different aqueous ⁄ organic mixtures with the same water activity has similar catalytic con- stants. This fact supports the idea that it is the enzyme-bound water that modulates the catalytic properties of the enzyme. However, this relationship does not hold for polar solvents, especially at high water activity [10]. It was hypothesized that solvents such as alcohols are able to partially replace the role of bound water, acting as ‘water replacers’ in promo- ting enzyme activity [10]. The failure of water activity to predict the critical amount of water needed for enzyme activity suggests that the organic solvent also has a role in modulating the structure and dynamics of the enzyme, probably by taking part in the solvation mechanism of the enzyme. It is clear that a concise molecular picture of the solvation mechanism of enzymes in nonaqueous solvents is needed. Reviewing what is known about protein hydration takes us back to early protein hydration studies on dry proteins. There is a striking similarity between the water-adsorption properties of proteins in air and in organic solvents [13]. Early studies of Yang & Rupley [14] and Rupley et al. [15], based on calori- metric measurements of the heat capacity of the lyso- zyme–water system, detailed the mechanism of the hydration process of dry proteins. The authors [14] pointed out that the hydration steps for lysozyme resemble the Hill model for the localized adsorption of adsorbate onto a heterogeneous surface [16]. They interpreted the Hill model as follows: at low cover- age, the adsorbate is dispersed on the surface; the increase in adsorbate leads to the formation of a con- densed phase of clusters; the clusters grow until the surface is nearly covered and only the weakest sites remain open; condensation of adsorbate over these regions completes the adsorption process. With this model, Yang proposed that water clusters can be viewed as mobile arrangements centered on polar regions of the protein surface that increase in size and number as water is added. Protein–water adsorp- tion isotherms in organic solvents and in the gas phase [13] have shown that, at low water activity, water adsorption by proteins suspended in nonpolar organic solvents or by proteins equilibrated with water from a gas phase are similar. This has led to the conclusion that the presence of an organic solvent has little effect on the interaction between the protein and water in this water range. Parker et al. [8] have detailed the mechanism of enzyme hydration (using subtilisin Carlsberg) in non- polar solvents using sensitive NMR experiments with deuterated water. Their work clearly shows that in nonpolar solvents (hexane, toluene, and benzene) water is preferentially localized in the most polar regions of the enzyme. The majority of the enzyme surface is in direct contact with the organic solvent, and the forma- tion of a monolayer of water over the protein surface is thermodynamically unfavorable. However, no polar organic solvent was used in this study. A similar study [17], using water sorption isotherms of lysozyme in nonpolar and polar organic solvents, previously sug- gested the same mechanism of protein hydration. Io- nizable sites are hydrated first, followed by polar and nonpolar sites. However, when the sorption isotherms of toluene and n-propanol in the same water activity range are compared, they differ due to the competition of the organic solvent for the enzyme hydration sites. The hydration mechanism of enzymes in nonaque- ous solvents seems to be dictated by many factors, most of which have been addressed in the previous cited reports. Not only do the properties of the organic solvent determine the relative partition of water at the enzyme surface at specific sites, but additionally, this solvent has a significant role in the solvation mechan- ism of the enzyme. In this context, it would be import- ant for a molecular interpretation of the effects of the different quantities of enzyme-bound water in non- polar and polar organic solvents if the number of water molecules boun d to the enzyme could be precisely measured, characterized, and localized [10,18]. Our work is a comprehensive study of the hydration mech- anism of the enzyme cutinase in nonpolar (hexane, di- isopropyl ether, 3-pentanone) and polar (ethanol, acetonitrile) organic solvents with increasing hydration levels. We have tried to determine how enzyme hydra- tion occurs in these media, and what the role is of the different organic solvents in the solvation mechanism of the enzyme. Results and Discussion Enzyme structure Protein molecular dynamics ⁄ molecular mechanics (MD ⁄ MM) simulations in organic solvents have been reported for several model systems [5,6,19–27]. These simulation studies and the ones presented in this work typically involve one suspended isolated single protein molecule surrounded by water, ions, and organic N. M. Micae ˆ lo and C. M. Soares Modeling hydration mechanisms of enzymes FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS 2425 solvent. Real nonaqueous experimental studies are usu- ally carried on an immobilization support; however, the system would be more complex if other factors, such as protein aggregation, played an important role in the protein function. Our approach to the study of protein structure and dynamics in such media has been focused on understanding how proteins are affected by the different hydration conditions when placed in organic solvents [5,6]. These studies have shown that the structure, dynamics and enantioselectivity of cu- tinase in hexane can be optimized within a specific water hydration range [5,6,27]. A more complete understanding of the structural properties of this enzyme in organic solvents is shown in Fig. 1. Cu- tinase was simulated in five different organic solvents of increasing dielectric constant with different hydra- tion conditions. The water range studied for each solvent is a key aspect in understanding how the structural properties of the enzyme are modulated by the hydration level. Testing different organic sol- vents allows us to determine the role played by the organic solvent in the stabilization ⁄ destabilization of Fig. 1. Average rmsd of Ca atoms of cutinase from the X-ray structure in (A) hexane, (B) diisopropyl ether, (C) 3-pentanone, (D) ethanol, and (E) acetonitrile, with different water percentages. Calculations of rmsd deviations were done in the 3–7 ns period for each simulation and for all replicas. Error bars are estimated from the SE of five to seven replicate simulations. Modeling hydration mechanisms of enzymes N. M. Micae ˆ lo and C. M. Soares 2426 FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS the enzyme structure. The rmsd of the Ca atoms of the protein fitted against the X-ray structure shows that the enzyme structure is slightly different in each organic solvent. There are low rmsd values for the enzyme in ethanol, and high values in acetonitrile. For hexane, diisopropyl ether, and 3-pentanone, the enzyme Ca rmsd values at the different water percent- ages are within 0.16 and 0.27 nm. How do these data compare with experimental data? The molecular struc- tures of several enzymes soaked in organic solvents have been determined by crystallographic studies [28–34], but only limited changes in protein structure were detected as compared with conventional aqueous crystals. Some authors argue that this approach could hardly give any other answer, because if major con- formational changes were to exist, it would be unlikely that the crystal packing could be maintained [18]. Other methods are available that can provide struc- tural information on these proteins in solution. CD studies of a-chymotrypsin in organic solvents have shown a clear correlation between water content and secondary structure of the enzyme [9,35]. Fluorescence measurements of enzymes in organic solvents have also been used to investigate the structural changes of enzymes. Kijima and coworkers [36,37] have shown that a-chymotrypsin enantioselectivity and fluorescence properties are correlated with the solvent composition. These findings suggest a more dynamic picture of enzyme structure rearrangement when enzymes are placed in different organic solvents and have different hydration levels. The enzyme Ca rmsd measurement from our simulations at different water percentages and organic solvents show that different solvation con- ditions yield different enzyme structural properties. This result is in agreement with the common experi- mental observation that the solvent composition in nonaqueous systems is able to affect enzyme structural properties. Further analysis of the rmsd (Fig. 1) suggests the existence of minima in the rmsd data versus water per- centage in the less polar organic solvents (hexane, diisopropyl ether, and 3-pentanone). It is possible to see that the structure of cutinase deviates less from the X-ray structure when it is placed in hexane with 7.5% water. In the case of cutinase in diisopropyl ether, we obtained the lowest rmsd at 30% water. With a slightly polar medium such as 3-pentanone, a local rmsd minimum was observed at 40%; however, we also observed low rmsd values at very low water per- centages for this organic solvent. It can be seen that the optimal water content that minimizes the difference from the X-ray structure is displaced to higher water levels as we increase the polarity of the organic solvent, as seen experimentaly [38]. The dependence of enzyme structural properties on different water con- tents in organic solvents is a well-documented phenom- enon observed for several enzymes. In some cases, a bell-shaped behavior of structural and biocatalytic properties is observed. The rmsd data of our model enzyme placed in the less polar organic solvents with different hydration levels resemble this type of bell- shape behavior; this is clear in the case of hexane and diisopropyl ether. However, this phenomenon is not clearly seen in our simulations in the case of polar, water-miscible organic solvents such as ethanol and acetonitrile. This may be because only a small amount of water is retained at the enzyme surface in the case of polar organic solvents relative to nonpolar organic solvents, as detailed below. Water at the enzyme surface Spatial probability density A key aspect of this work is the analysis of the local- ization of water at the enzyme surface. We have suc- cessfully identified regions of high density of water in close contact with the protein for the different organic solvents tested. In Fig. 2, we show the spatial distribu- tion probability of water at 25% water content for the simulations in organic solvents and for the fully hydra- ted simulation. This result was obtained by calculating the water probability density from all configurations of the last 3 ns for each organic solvent and for all repli- cates. Probability densities were chosen in order to drawn contours at percentiles approximately between 93% and 98% for hexane, diisopropyl ether, and 3-pentanone, and at 99% for ethanol, acetonitrile, and water, giving the clearest picture for the preferential hydration regions near the enzyme surface. From our simulations, we see that water does not partition appreciably to the organic solvent phase in the case of hexane, diisopropyl ether, and 3-pentanone, whereas in ethanol and acetonitrile, a considerable amount of water is found in the organic phase (results not shown). Note that in the fully hydrated system, we are also able to identify regions of higher density of water molecules. The water arrangement over the enzyme surface shows that there is no evidence of a complete water monolayer covering the enzyme (in fact, a monolayer would be possible at water percentages higher than 50%, although that never occurs), in accordance with previous suggestions that this would be thermodynam- ically unfavorable [8]. The water distribution at the enzyme surface is clearly localized in certain regions, leaving other parts of the enzyme surface in direct N. M. Micae ˆ lo and C. M. Soares Modeling hydration mechanisms of enzymes FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS 2427 contact with the organic solvent. This clearly suggests that the role of the organic solvents should not be undervalued, given that a significant enzyme surface area is solvated by the organic solvent in all organic solvents tested. This means that water per se might not account entirely for the solvation process of the enzyme. An important discussion point that arises from our studies is that the preferential binding sites of water at the enzyme surface seem to be independent of the organic solvent. This is supported by the observation that the water spatial probability distributions seem to be equivalent for the same enzyme, regardless of the organic solvent used (Fig. 2). This appears to be true if we look at Fig. 2 and try to compare the spatial prob- ability densities of water in hexane, diisopropyl ether, and 3-pentanone. They show that water is distributed over similar regions of the enzyme in these three sol- vents. In the case of more polar solvents such as ethanol and acetonitrile, we see that the water in these systems is found in regions also present in the nonpolar solvents. As the enzyme surface does not change dramatically when the enzyme is placed in different organic solvents, water molecules seem to populate equivalent sites that correspond to the areas of exposed charged ⁄ polar side chains hydrated to a higher or lower degree according to the polarity of the organic solvent. Nonpolar organic solvents In order to obtain a more precise picture of water at the surface of the protein, we looked at the number of water molecules within a specific layer of 0.25 nm from the surface of the protein (Fig. 3), and the ratio of Fig. 2. Spatial distribution probability density of water in (A) hexane, (B) diisopropyl ether, (C) 3-pentanone, (D) ethanol and (E) aceto- nitrile with 25% water and (F) in the fully hydrated system. The molecular surface corresponds to the average structure of cu- tinase from the 3–7 ns period, for each sol- vation system and for all replicas. For each organic solvent, two sides of the enzyme are shown in order to give a complete view of the surface. Each view of the enzyme has the same orientation in all organic sol- vents. The contours enclose regions with a probability density above 9 · 10 )6 A ˚ )3 for hexane, diisopropyl ether, and 3-pentanone, 4 · 10 )6 A ˚ )3 for ethanol, 3 · 10 )6 A ˚ )3 for acetonitrile, and 1 · 10 )6 A ˚ )3 for the fully hydrated system. Fig. 3. Average number of water molecules less than 0.25 nm away from the enzyme surface, for each organic solvent and water percentage. Error bars are estimated from the SE of five to seven replicate simulations. The total number of water molecules corres- ponding to each hydration level, for comparison with the number bound, is given in supplementary Table S3. Modeling hydration mechanisms of enzymes N. M. Micae ˆ lo and C. M. Soares 2428 FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS water to organic molecules in the region beyond this layer (supplementary Table S1), for all organic solvents and hydration levels. This criterion is intended to cap- ture the first layer of water molecules in direct contact with the enzyme. The first evidence from these curves is the fact that they resemble the shape of the water adsorption isotherms of enzymes in nonaqueous sol- vents [13,17,39,40], although in these experimental reports, water adsorption is plotted as a function of water activity rather than water content. These curves show, first, a rapid increase in bound water, followed by a second step in which there is a slow increase, and then a third step of high water activity, where again there is a sharp increase. The water range that we tes- ted seems to comprise the two initial steps. Note that the curves can easily discriminate nonpolar and polar solvents. In nonpolar solvents, water is highly retained at the enzyme surface, whereas in polar solvents, water is only weakly retained. In the particular case of hexane, most of the water is located in this first hydration shell around the enzyme, covering a large proportion of the surface area of the enzyme but not achieving full coverage, as seen in the previous section. This organic solvent is the one that allows the retention of the highest amount of water at the enzyme surface. The remaining water that is not in direct contact with the enzyme is found in secondary hydration layers. Changing to a slightly polar solvent such as diisopropyl ether, we see the same trend for the water amount curve, but in this case the amount of water in direct contact with the enzyme is slightly lower, and it decreases even more as the polarity of the organic solvent increases, as seen for 3-pentanone. This clearly suggests that, to obtain the same amount of water bound to the enzyme, we need to add more water to the system as we move to more polar solvents. The general trend for the amount of water at the enzyme surface observed for these three solvents shows that, at low water percentages, most of the water present in the system is found at the surface of the enzyme, and as water is added, it expands its coverage over the enzyme surface up to a certain limit. For instance, it is possible to see that in diisopropyl ether, the hydration of the enzyme surface reaches a saturation point at about 40% water, corresponding to 90 water molecules at a distance less than 0.25 nm from the enzyme surface. For the case of 3-pentanone, we see that at water percentages above 60%, corres- ponding to 80 water molecules, there is almost no more water retained at the enzyme surface. In general, it can be seen that as we change from apolar solvents (hexane) to a slightly polar organic solvent (3-penta- none), it becomes energetically favorable to have organic solvent molecules instead of water molecules solvating the enzyme. Polar organic solvents In water-miscible organic solvents such as ethanol and acetonitrile, the competition between the organic sol- vent and water for the enzyme surface is higher. These two organic solvents can mimic the nonbonding prop- erties of water, and can effectively compete for the polar regions of the enzyme. This is clearly seen in Fig. 3, which shows that the amount of water bound at the enzyme surface is very low as compared to the situation with nonpolar solvents. In polar organic sol- vents, the amount of water bound to the enzyme increases slowly as water is introduced to the system, showing that, to obtain the same amount of water bound to the enzyme in nonpolar solvents such as hex- ane, it is necessary to add very high amounts of water. The amount of organic solvent bound to the enzyme is very high, and the solvent clearly acts as a water re- placer in many polar and nonpolar regions of the enzyme surface. This phenomenon clearly correlates with previous observations that polar organic solvents strip water from the enzyme surface to a higher extent than nonpolar ones [41]. This result is also in agree- ment with the observation of McMinn et al. [17] show- ing that in polar organic solvents, the amount of water bound to the enzyme at high water activities is signifi- cantly lower relative to the case when nonpolar organic solvents are used. In this case, the organic sol- vent has a significant role in solvating the enzyme, and also takes part in the modulation of the structural and dynamic properties of the enzyme. It is also of note that ethanol and acetonitrile are significantly different in their ability to strip off water from the enzyme. In ethanol, at almost all hydration levels, we find twice as much water bound to the enzyme as in aceto- nitrile; however, experiments show that more water is bound when the enzyme is equilibrated with a given water concentration in acetonitrile as compared with ethanol [17]. Water structure In our simulated systems, we see that water molecules can be found in the bulk organic phase or at the enzyme surface. In the water-miscible organic solvents, we see a considerable amount of water in the organic solvent phase (results not shown). However, in all organic solvents, water seems to be organized in clus- ters at the enzyme surface (Fig. 4A). These water clusters might have a substantial functional role in N. M. Micae ˆ lo and C. M. Soares Modeling hydration mechanisms of enzymes FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS 2429 modulating the structural and dynamic properties of the enzyme, as suggested previously. In Fig. 4A, we show the number of water clusters at the enzyme surface. A cluster is defined by two or more water molecules at a minimum distance of 0.3 nm (distance between oxygen atoms). Note that this criterion might not account for single water mole- cules interacting with polar residues or water molecules solvating the ions in solutions (as oxygen atoms in this case might be more than 0.3 nm distant). In Fig. 4B, we show the average number of water molecules per cluster. Again, the behavior in nonpolar and polar sol- vents is easily distinguishable by this property. In the presence of nonpolar organic solvents, the number of water clusters is almost identical in all hydration con- ditions up to 25% (Fig. 4A). In the nonpolar organic solvents, the number of clusters grows rapidly as water is added, indicating that water is being organized in clusters of two or more water molecules at the enzyme surface, hydrating specific spots of the enzyme. The number of clusters increases up to 25% water content, and beyond this water level, the number of clusters remains constant. This indicates that, as water is added to the system, new clusters of water are formed at available specific regions at the enzyme surface, which become fully occupied at 25% water content. The size of the clusters also grows as water is added, meaning that the water that is gradually introduced is also dis- tributed on pre-existing water clusters. Water clusters in nonpolar solvents are of similar size at low water percentages, but as water is added, the clusters in hex- ane become significantly larger than those in diisopro- pyl ether and 3-pentanone (Fig. 4B). In polar organic solvents, the water organization at the enzyme surface is different from that in the nonpo- lar organic solvents. The cluster size increases more slowly as water is added, suggesting that most of the water introduced into the system is preferentially localized in the bulk organic solution. The size of the clusters in ethanol and acetonitrile is fairly identical up to 50% water content. Above this level, water clusters in ethanol are slightly larger that those in acetonitrile. The molecular picture that arises from this analysis of polar organic solvents is that water is largely frag- mented into single water molecules and small clusters of water molecules around the protein. In the case of nonpolar solvents, water is tightly bound to the enzyme and organized in clusters that grow in number and size in proportion to the water added. Ions in nonaqueous systems Ions play an important role in nonaqueous media, as they will allow the neutralization of exposed charged residues of the enzyme that cannot form intramolecu- lar ion pairs [42]. This is evident from the X-ray struc- ture of trypsin in cyclohexane, which shows the existence of sulfate ions forming salt bridges or hydro- gen bonds with residues or water molecules [34]. Work with different ion concentrations in nonaqueous system has also shown that ions have a marked impact on enzyme activity, relative to enzymes with no added salt [43–45]. In all organic solvent simulations, we have 10 Na + and 10 Cl – docked to cutinase, as described in a previ- ous study [5]. These ions neutralize individual charged groups at the enzyme surface. We have seen that at equilibrium, water rearranges itself around the protein according to the type of the organic solvent used and the amount of water present in the system. The same seems to be true for the ions. In Fig. 5, we analyze Fig. 4. (A) Average number of water clusters for each organic sol- vent and water percentage. (B) Average size of the water clusters shown in (A). Error bars are estimated from the SE of five to seven replicate simulations. See text for details and comments regarding this analysis. Modeling hydration mechanisms of enzymes N. M. Micae ˆ lo and C. M. Soares 2430 FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS how many charged residues are neutralized by sodium (Fig. 5A) and chloride (Fig. 5B) counterions. The first evidence is that ions are not irreversibly bound to the enzyme, as some charged residues initially stabilized by ions are preferentially compensated by water mole- cules, particularly at high water percentages. Sodium ions bound to the enzyme seem to be highly conserved at low water percentages in all organic solvents. As we add water, sodium ions seem to be displaced from the enzyme surface to the same extent in all organic sol- vents. These ions are found free in solution, hydrated by water molecules, which, as we have seen, are organ- ized in clusters. Another common observation is the formation of Na + Cl – ion pairs and, more rarely, Na + Cl – tetrads in solution or at the enzyme surface, where one of the negative or positive pair is a charged residue. Chloride ions are also displaced from the enzyme surface by water in the same way as sodium ions. However it seems that more polar solvents such as ethanol and acetonitrile, even at low water percent- ages, are able to replace chloride ions at the enzyme surface. This phenomenon is more evident in acetonit- rile, as these chloride ions are rapidly removed from the enzyme, even at very low levels of hydration. At high levels of hydration, only one ion or even none is found bound to the enzyme. Another phenomenon seen in Fig. 5A at low water percentages in nonpolar solvents is that some of the chloride ions stabilize more than one positive charged residue, as the 10 Cl – are in the proximity of more than 10 positive charged resi- dues. What these results suggest is that ions can also be ‘stripped off’ from the enzyme surface by the water molecules. The charged residues initially stabilized by ions at low water percentages become preferentially compensated by water molecules as the water content increases. This suggests that at low hydration levels, ions are important in the stabilization of charged resi- dues, but as the system becomes ‘more aqueous’, the exposed charged residues are preferentially stabilized by water molecules. The loss of the charge-counteract- ing effect provided by the ions near exposed charged residues could also be responsible for the structural changes observed at higher water percentages in all organic solvents. Water dynamics We have seen that organic solvents with increasing polarity can structure, in different ways, the water at the enzyme surface. It is also important to question how the dynamics of the water are modulated by the presence of the different organic solvents. In a recent NMR study [46], the authors suggested a hydration model of subtilisin in tetrahydrofuran that comprised tightly bound, loosely bound and free water. To ana- lyze the water dynamics at the enzyme surface, we recorded all the hydration events of all water mole- cules in the system. A hydration event is the total time for which one water molecule is inside a layer of 0.25 nm around the enzyme surface. The analysis was done during the last 3 ns of all trajectories for water contents of 25% and 60%. All hydration events of all water molecules for a specific water content and organic solvent are collected and grouped in a fre- quency histogram. For a clear analysis, the data were fitted using the Levenberg–Marquardt method to a two-exponential equation of the form: f ðxÞ¼a þ b  e ÀcÂx þ d  e ÀfÂx where x stands for the time in ps that a water molecule is inside a layer of 0.25 nm around the enzyme surface, and f(x) is the frequency of occurrence of that period Fig. 5. Average number of (A) negatively and (B) positively charged residues neutralized by sodium and chloride ions, respectively, for each organic solvent and water percentage. Error bars are estima- ted from the SE of five to seven replicate simulations. N. M. Micae ˆ lo and C. M. Soares Modeling hydration mechanisms of enzymes FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS 2431 of time. Final parameters and standard errors are shown in supplementary Table S2. The residence times of the water molecules at the enzyme surface at 25% and 60% water content in the different organic solvents and in the fully hydrated sys- tem are shown on Fig. 6. A general overview of this figure indicates that many of the water molecules have very low residence times at the enzyme surface. How- ever, a significant proportion of the water is retained at the enzyme surface on a nanosecond time scale in all organic solvents. At 25% water content (Fig. 6A), it is possible to distinguish the effects of nonpolar and polar organic solvents on the dynamics of the water at the enzyme surface. Water molecules in nonpolar sol- vents are retained for longer periods of time at the sur- face of the enzyme than in polar organic solvents. Particularly at 25% water content, the water residence times in ethanol and acetonitrile are equivalent to those in the pure aqueous system. At 60% water con- tent (Fig. 6B), all four organic solvents (no data for hexane at this water content) modulate the hydration events in a progressive way; that is, water is retained for less time at the enzyme surface as the polarity of the organic solvent increases. It seems that, besides the differential structuring of water by the organic solvent, these water molecules organized in clusters at the enzyme surface do not behave as in a bulk aqueous solution. Their dynamic properties, with respect to the residence time at the enzyme surface, are modulated by the polarity of the organic solvent. Concluding remarks We have performed a systematic simulation study of the hydration mechanism of one enzyme in three dis- tinct classes of solvent: nonpolar organic solvents, polar organic solvents, and water. We consider the effect of five different organic solvents, with different water percentages, on the structural properties of one model enzyme. It is shown that the structural proper- ties of the enzyme in the less polar solvents (hexane, diisopropyl ether, and 3-pentanone) give a bell-shape curve, indicating that there is an optimum hydration level that allows the existence of a native-like structure in solution. This optimal hydration level for each organic solvent is obtained at increasing water percent- ages as we move to more polar solvents. Our study also provides a detailed molecular picture of the hydration mechanism in the organic media, as indica- ted previously by experimental findings obtained by hydration studies in organic solvents by NMR and also from water adsorption experiments. Our results show that water in nonaqueous media is organized at the enzyme surface in clusters of water molecules hydrating preferentially charged ⁄ polar residues. These clusters populate identical enzyme surface regions when the enzyme is placed in different organic sol- vents. As water is added, these clusters grow in num- ber and size. The nature of the organic solvent is able to determine the size and number of clusters. Nonpolar solvents allow the existence of large clusters of water molecules at the enzyme surface, whereas polar sol- vents fragment these clusters into smaller aggregates. Polar solvents have the ability to replace water at some enzyme surface regions and contribute effectively to the structure and dynamics of the enzyme. This means that water activity per se may not be sufficient to char- acterize the solvation of enzymes, and thus, water activity values in different organic solvent might not correlate directly with catalytic properties measured experimentally. Ions seem to be preferentially bound to the enzyme at low hydration levels. Owing to the presence of the organic solvent, water is retained for A B Fig. 6. Water residence time frequency for each organic solvent with (A) 25% water and (B) 60% water. See text for details and comments regarding this analysis. Modeling hydration mechanisms of enzymes N. M. Micae ˆ lo and C. M. Soares 2432 FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS longer at the enzyme surface, and this is more evident in solvents with very low polarity. On the other hand, in high-polarity solvents, water at the enzyme surface behaves similarly as in the fully hydrated system. This study has provided a detailed molecular picture of the hydration mechanism of an enzyme, and shown it to be clearly dependent on the nature of the organic sol- vent and water content. Experimental procedures Organic solvents force field The organic solvents employed in this study were: hexane, diisopropyl ether, 3-pentanone, ethanol, and acetonitrile. These organic solvents are commonly used in nonaqueous enzymology studies, and are parameterized for MD ⁄ MM simulations. Hexane was modeled as a flexible united atom model using the gromos96 43a1 alkane parameters [47]; diisopropyl ether was taken from Stubbs et al. [48] and adapted to the gromos96 43a1 force-field (FF) [49,50]; 3-pentanone is found in the gromos96 43a1 FF [51]; ethanol is also present in the gromos96 43a1 FF, and a new parameterization of acetonitrile [52] was recently done for this FF. These organic solvents have increasing dielectric proper- ties and different partition coefficients [53] (Table 1). The rationale for the choice of these organic solvents was to have two groups with distinct properties, those that are immiscible with water and that have low polar characteris- tics (hexane, diisopropyl ether, and 3-pentanone), and those that have polar properties and are water miscible (ethanol and acetonitrile). System setup The general simulation methodology applied in the MD ⁄ MM simulations of cutinase in nonaqueous solvents with increasing amounts of water was similar to the one that we applied in a previous study [6]. The 1.0 A ˚ resolu- tion cutinase structure of Longhi was used [54], and the protonated state of charged residues was estimated using a methodology described previously [55]. The selection of counterion positions and the different amounts of water hydrating the enzyme was done as explained in detail else- where [5]. Five to seven replicates of 12 hydration levels were chosen, ranging from 5% to 100%; see supplementary Table S3 for a complete description of the molecular com- position of each system. The replicates are composed of equivalent molecular systems (enzyme, ions, water, and organic solvent) with water molecules at slightly different positions at the enzyme surface. Different hydration ranges were chosen for each organic solvent according to the knowledge that the critical amount of water that optimizes the structural and dynamic properties of the enzyme depends on the polarity of the organic solvent used [7]. Each replicate of cutinase hydrated with a specific amount of water was placed in a dodecahedral box with a minimum distance between the protein and box wall of 0.8 nm, and solvated with an equilibrated configuration of organic sol- vent molecules at 300 K. Three replicate simulations of cutinase with ions in a fully hydrated system with single point change (SPC) water were also done. MD ⁄ MM simulations MD ⁄ MM simulations were performed with the gromacs package [56,57] using the gromos96 43a1 FF [49,50]. Bond lengths of the solute and organic solvent molecules were constrained with lincs [58], and those of water with settle [59]. Nonbonded interactions were calculated using a twin- range method [50] with short-range and long-range cut-offs of 8 A ˚ and 14 A ˚ , respectively. The SPC water model [60] was used in aqueous and in nonaqueous simulations. A reaction field correction for electrostatic interactions [61,62] was applied, taking a dielectric of 54 for the fully hydrated system with SPC water [63]. For the nonaqueous systems, the dielectric constant was chosen according to the experi- mental value reported in Table 1. The simulations were started in the canonical ensemble with initial velocities from a Maxwell–Boltzmmann distribution at 300 K, and run for 50 ps with position restraints applied to all heavy atoms of the protein and water molecules (force constant of 10 6 kJÆmol )1 Ænm )2 ) and a temperature coupling constant of 0.01 ps, allowing the equilibration of the organic solvent. A further 50 ps of restrained simulation with the same force constant on the protein heavy atoms and temperature coup- ling constant was done for the equilibration of water mole- cules. A final step of 50 ps was done with restraints only applied to the Ca carbons of the enzyme and a temperature coupling constant of 0.1 ps. The unrestrained simulations were done in the isothermal–isobaric ensemble with an integration time step of 2 femtoseconds. The protein, ions, organic solvent and water were coupled to four separated heat baths [64] with temperature coupling constants of 0.1 ps and a reference temperature of 300 K. The pressure control [64] was implemented with a reference pressure of Table 1. Dielectric constant and partition coefficient (log P) [53] of the organic solvents employed in this work. Dielectric constant (temperature K) Partition coefficient (log P) Hexane 1.9 (293.2) 4.00 Diisopropyl ether 3.8 (303.2) 1.52 3-Pentanone 17.0 (293.2) 0.82 Ethanol 25.3 (293.2) ) 0.30 Acetonitrile 36.6 (293.2) ) 0.54 N. M. Micae ˆ lo and C. M. Soares Modeling hydration mechanisms of enzymes FEBS Journal 274 (2007) 2424–2436 ª 2007 The Authors Journal compilation ª 2007 FEBS 2433 [...]... 2051–2057 Gorman LAS & Dordick JS (1992) Organic solvents strip water off enzymes Biotechnol Bioeng 39, 392–397 Halling PJ (2000) Biocatalysis in low-water media: understanding effects of reaction conditions Curr Opin Chem Biol 4, 74–80 Modeling hydration mechanisms of enzymes 43 Morgan JA & Clark DS (2004) Salt-activation of nonhydrolase enzymes for use in organic solvents Biotechnol Bioeng 85, 456–459... Soares CM (2005) Water dependent properties of cutinase in nonaqueous solvents: a computational study of enantioselectivity Biophys J 89, 999–1008 7 Zaks A & Klibanov AM (1988) The effect of water on enzyme action in organic media J Biol Inorg Chem 263, 8017–8021 8 Parker MC, Moore BD & Blacker AJ (1995) Measuring enzyme hydration in nonpolar organic- solvents using NMR Biotechnol Bioeng 46, 452–458 9 Sasaki... solvents J Biol Inorg Chem 263, 3194–3201 12 Bell G, Halling PJ, Moore BD, Partridge J & Rees DG (1995) Biocatalyst behaviour in low-water systems Trends Biotechnol 13, 468–473 2434 13 Halling PJ (1990) High-affinity binding of water by proteins is similar in air and in organic solvents Biochim Biophys Acta 1040, 225–228 14 Yang PH & Rupley JA (1979) Protein–water interactions Heat capacity of the lysozyme–water... of a-chymotrypsin in organic solvents as studied by circular dichroism Biotechnol Tech 11, 387–390 10 Bell G, Janssen AEM & Halling PJ (1997) Water activity fails to predict critical hydration level for enzyme activity in polar organic solvents: interconversion of water concentrations and activities Enzyme Microb Technol 20, 471–477 11 Zaks A & Klibanov AM (1988) Enzymatic catalysis in nonaqueous solvents. .. MT, Dordick JS, Reimer JA & Clark DS (1999) Optimizing the salt-induced activation of enzymes in organic solvents: effects of lyophilization time and water content Biotechnol Bioeng 63, 233–241 45 Ru MT, Wu KC, Lindsay JP, Dordick JS, Reimer JA & Clark DS (2001) Towards more active biocatalysts in organic media: increasing the activity of salt-activated enzymes Biotechnol Bioeng 75, 187–196 46 Lee CS,... dependence on water activity in different organic solvents Biochim Biophys Acta 1118, 218–222 Sirotkin VA (2005) Effect of dioxane on the structure and hydration dehydration of alpha-chymotrypsin as measured by FTIR spectroscopy Biochim Biophys Acta 1750, 17–29 Sirotkin VA, Solomonov BN, Faizullin DA & Fedotov VD (2002) Sorption of water vapor and acetonitrile by human serum albumin Russ J Phys Chem 76,... GK (1994) X-ray crystal-structure of gamma-chymotrypsin in hexane Biochemistry 33, 7326–7336 Zhu GY, Huang QC, Wang ZM, Qian MX, Jia YS & Tang YQ (1998) X-ray studies on two forms of bovine beta-trypsin crystals in neat cyclohexane Biochem Biophys Acta 1429, 142–150 ´ Simon LM, Garab MKG & Laczko I (2001) Structure and activity of a-chymotrypsin and trypsin in aqueous organic media Biochem Biophys Res...ˆ N M Micaelo and C M Soares Modeling hydration mechanisms of enzymes 1 atm and relaxation times of 0.5 ps and 1.3 ps, for water or organic ⁄ water solvent simulations, respectively Acknowledgements The authors acknowledge helpful discussions with Dr ´ Antonio M Baptista and Professor Susana Barreiros, and financial support from Fundacao para a Ciencia e ¸ ˜... dynamics of chymotrypsin in hexane J Am Chem Soc 118, 6490–6498 25 Zheng Y-J & Ornstein RL (1996) A molecular dynamics and quantum mechanics analysis of the effect of DMSO on enzyme structure and dynamics: subtilisin J Am Chem Soc 118, 4175–4180 26 Yang L, Dordick JS & Garde S (2004) Hydration of enzyme in nonaqueous media is consistent with solvent dependence of its activity Biophys J 87, 812–821 27 Vidinha... LJ & Klibanov AM (1997) The crystal structure of subtilisin Carlsberg in anhydrous dioxane and its comparison with those in water and acetonitrile Proc Natl Acad Sci USA 94, 4250–4255 Schmitke JL, Stern LJ & Klibanov AM (1998) Comparison of x-ray crystal structures of an acyl-enzyme intermediate of subtilisin Carlsberg formed in anhydrous acetonitrile and in water Proc Natl Acad Sci USA 95, 12918–12923 . Modeling hydration mechanisms of enzymes in nonpolar and polar organic solvents Nuno M. Micae ˆ lo and Cla ´ udio M. Soares Instituto de Tecnologia. according to the polarity of the organic solvent. Nonpolar organic solvents In order to obtain a more precise picture of water at the surface of the protein,