9
RNA Solvation: A Molecular Dynamics Simulation Perspective Pascal Auffinger Eric Westhof Institut de Biologie Mole ´ culaire et Cellulaire du CNRS, Mode ´ lisations et Simulations des Acides Nucle ´ iques, UPR 9002, 15 rue Rene ´ Descartes 67084 Strasbourg Cedex, France Received 21 February 2001; accepted 15 March 2001 Published online 18 December 2001 Abstract: With the availability of accurate methods to treat the electrostatic long-range interac- tions, molecular dynamics simulations have resulted in refined dynamical models of the structure of the hydration shell around RNA motifs. The models reviewed here range from basic Watson–Crick to more specific noncanonical base pairs, from “simple” double helices to RNA molecules display- ing more complex tertiary folds, and from DNA/RNA hybrid double helices to RNA hybrids formed with a chemically modified strand. © 2001 John Wiley & Sons, Inc. Biopoly (Nucleic Acid Sci) 56: 266 –274, 2001 Keywords: RNA solvation; molecular dynamics; electrostatic long-range interactions; hydration shell INTRODUCTION Due to the continuous developments in experimental techniques, a large array of data describing the mac- roscopic and microscopic hydration features of bio- molecular systems give shape to our current solute/ solvent interaction model. 1–6 Clearly, all the newly generated data need to be integrated in a coherent way. Molecular dynamics (MD) simulation tech- niques contribute to this process. Here, we will con- centrate on the main contributions of MD simulations with explicit representations of the solvent to the study of the hydration of RNA molecules. An outline of several common methods used to characterize the structural and dynamical features of the hydration shell, followed by an account of results from MD simulations on RNA fragments describing water as well as ion binding characteristics, will be given. Additional information on nucleic acids, 5,7–12 and more generally, on biopolymer 13,14 hydration fea- tures, can be found in several recent articles and reviews. METHODS Structure: Characterization of Hydration Sites Crystallography has unambiguously established that some water molecules occupy spatially well-defined hydration Correspondence to: Eric Westhof; email: [email protected] strasbg.fr; Pascal Auffinger; email: P.Auffi[email protected] Biopolymers (Nucleic Acid Sciences), Vol. 56, 266 –274 (2001) © 2001 John Wiley & Sons, Inc. 266

RNA solvation: A molecular dynamics simulation perspective

Embed Size (px)

Citation preview

Page 1: RNA solvation: A molecular dynamics simulation perspective

RNA Solvation: A MolecularDynamics SimulationPerspective

Pascal AuffingerEric Westhof

Institut de BiologieMoleculaire

et Cellulaire du CNRS,Modelisations et Simulations

des Acides Nucleiques,UPR 9002,

15 rue Rene Descartes67084 Strasbourg Cedex,

France

Received 21 February 2001;accepted 15 March 2001

Published online 18 December 2001

Abstract: With the availability of accurate methods to treat the electrostatic long-range interac-tions, molecular dynamics simulations have resulted in refined dynamical models of the structure ofthe hydration shell around RNA motifs. The models reviewed here range from basic Watson–Crickto more specific noncanonical base pairs, from “simple” double helices to RNA molecules display-ing more complex tertiary folds, and from DNA/RNA hybrid double helices to RNA hybrids formedwith a chemically modified strand. © 2001 John Wiley & Sons, Inc. Biopoly (Nucleic Acid Sci)56: 266–274, 2001

Keywords: RNA solvation; molecular dynamics; electrostatic long-range interactions; hydrationshell

INTRODUCTION

Due to the continuous developments in experimentaltechniques, a large array of data describing the mac-roscopic and microscopic hydration features of bio-molecular systems give shape to our current solute/solvent interaction model.1–6 Clearly, all the newlygenerated data need to be integrated in a coherentway. Molecular dynamics (MD) simulation tech-niques contribute to this process. Here, we will con-centrate on the main contributions of MD simulationswith explicit representations of the solvent to thestudy of the hydration of RNA molecules. An outlineof several common methods used to characterize thestructural and dynamical features of the hydration

shell, followed by an account of results from MDsimulations on RNA fragments describing water aswell as ion binding characteristics, will be given.Additional information on nucleic acids,5,7–12 andmore generally, on biopolymer13,14 hydration fea-tures, can be found in several recent articles andreviews.

METHODS

Structure: Characterization of HydrationSites

Crystallography has unambiguously established that somewater molecules occupy spatially well-defined hydration

Correspondence to: Eric Westhof; email: [email protected]; Pascal Auffinger; email: [email protected] (Nucleic Acid Sciences), Vol. 56, 266–274 (2001)© 2001 John Wiley & Sons, Inc.

266

Page 2: RNA solvation: A molecular dynamics simulation perspective

sites in the vicinity of nucleic acid atoms.1,15–18 Unfortu-nately, even for high resolution structures, not all of thesehydration sites can be observed. In order to understandcompletely the physicochemical properties of these systemsand given the symbiotic solvent/nucleic acid relationship, itis essential to be able to determine with a good precision thelocation of these hydration sites. For that purpose, severaltechniques have been developed.

One widely used method is based on three-dimensional(3D) grids.19 The space around the average structure isdivided into small volume elements (usually close to 0.1Å3). Then, structures extracted from the simulation at reg-ular time intervals are fitted onto the average model andwater molecules are binned into the above defined volumeelements. Drawing the calculated densities at different con-tour levels allow to visualize the hydration sites. Thismethod gives generally a good picture of the hydration shellof molecules that are not too big and not subject to markedlylarge motions. For flexible molecules, an obvious drawbackof this approach is that, due to the fitting procedure, someregions like the extremities of long DNA duplexes are lesswell defined than the center regions. Consequently, theirhydration patterns are similarly less well defined.19

A more local approach has been developed in order todeal with this issue.20–22 This method consists in overlayingspecific structural elements, such as Watson–Crick basepairs surrounded by water molecules belonging to their firsthydration shell, in order to form “hydrated building blocks”similar to those derived from statistical analysis of x-raydiffraction data.23–25 These hydrated building blocks areFourier transformed into pseudo-electron density maps thatcan be drawn at different contour levels in order to delineatethe best defined hydration sites (Figure 1). At the base-pairstep level, this method should be able to help in the visu-alization of extended hydration patterns such as the waterspines observed in some DNA duplexes.

In parallel, radial distribution function approaches arevery common and give a one dimensional view of thedistribution of the solute/solvent distances. In order to get amore accurate three-dimensional view of the distribution ofwater molecules around solute atoms, the spatial distribu-tion function, or SDF, which integrate radial and angularcoordinates, was developed.26 Alternatively, a computation-ally efficient method to describe the organization of wateraround biomolecules has been proposed.27,28 It is based ona statistical mechanical expression of the water-density dis-tribution in terms of particle correlation functions.

Dynamics: Determination of ResidenceTimes

The determination of the residence times of water moleculesis not straightforward and is generally based on the estima-tion of the lifetime of specific solute-solvent hydrogenbonds.20–22,29,30 One possible definition considers that theresidence lifetime for any water molecule in the vicinity ofa hydrophilic atom equals the amount of time over which ahydrogen bond between both partners exists. Arbitrary time

limits are sometimes introduced specifying that, in order toconsider a water molecule as resident, the hydrogen bond inwhich it is engaged should not be broken for more than afew picoseconds (�1 or 2 ps). However, the use of timelimits is probably not absolutely required since it has beenobserved that a water molecule that escapes from a givenhydration site almost never comes back to the same siteduring a nanosecond MD simulation. Thus, the total timeover which a water molecule forms a hydrogen bond with agiven solute atom is a good approximation of its residencetime. From these data, profiles giving the number of watermolecules bound to a specific atom with respect to thelifetime of a given hydrogen bond can be drawn (Figure 2).Although these profiles may slightly change if the simula-tion time is extended,22 they are atom specific. Thus, even,if the values derived from these profiles have to be consid-ered on a semiquantitative basis, especially for the watermolecules with long residence times, the comparison be-tween profiles calculated with comparable protocols fordifferent atoms is very useful. Likewise, by recording thetime over which water molecules establish hydrogen bondswith two or three neighboring atoms, one can estimate upperlimits for water bridge lifetimes, such as the frequentOR(n) . . . Ow . . . OR(n � 1) hydrogen-bond pattern ob-served in RNA.20,21,29,30

Other methods rely on the time spent by a water mole-cule in a 5 Å sphere around a given solute atom31 or usecoordination correlation function approaches.14,32,33 Worthmentioning is that none of the above methods can givestatistically exact values for the residence times of the moststrongly bound water molecules, since the length of a MDsimulation should be many times longer than the longestcalculated residence times.

Force-Field Issues

Despite real difficulties attached to the inclusion of polar-ization and charge transfer effects into current two-bodyforce fields,34 recent improvements of empirical forcefields, reviewed by Cheatham and Kollman,12 and of thetreatment of long-range electrostatic interactions,35,36 haveled to more accurate MD simulations. Yet, it has beenproposed that simulations of DNA in explicit water aresurprisingly insensitive to small perturbations in the force-field or local environment on current time scales.37 ForDNA, the use of TIP3P38 or SPC/E39 water models wasfound to result in comparable hydration patterns. However,the water densities associated with the TIP3P model, whichhas the highest diffusion rate, appear more “blurred” thanthose calculated with the SPC/E model.37 Besides, a MDsimulation of a protein in a crystal environment led to theconclusion that results obtained with the SPC/E model arein much better agreement with neutron-scattering experi-ment data than those collected with the TIP/3P model.40

Hence, an estimation of the quality of a water model withrespect to dynamical data, which are very scarce, has still tobe achieved.

RNA Solvation 267

Page 3: RNA solvation: A molecular dynamics simulation perspective

RESULTS

Watson–Crick Base Pairs

Watson–Crick pairs constitute the most basic unitencountered in RNA structures. As such, it is ofparamount importance to gain a precise view of theirhydration. Besides, data gathered for Watson–Crickpairs are the basis for a sound comparison in theevaluation of modifications of the hydration shellassociated with insertions of non-Watson–Crick pairsand modified nucleotides into RNA structures.

A systematic approach for the study of the hydrationshells of Watson–Crick pairs has been proposed.20,21

This method consists in performing MD simulations ofregular RNA duplexes, i.e., r(CpG)12 and r(UpA)12. Inorder to get statistically accurate data, the 12 central basepairs of each duplex were overlaid and the water mole-cules surrounding them were counted. On the average,21.9 and 21.0 solvent molecules (water and ions) werefound to enter into contact with the r(GAC) and ther(A–U) pairs, respectively. Pseudoelectron density mapswere then calculated from the cloud of water moleculessurrounding each type of base pairs. High-level contoursof the density maps allowed us to delineate 22 well-defined hydration sites for the r(GAC) and r(A–U) pairs(Figure 1). The positions of the hydration sites aroundthe bases are in excellent agreement with those derivedfrom a statistical analysis of high-resolution crystallo-graphic data.24 Note that both in the crystal and insolution, these sites are better defined for the r(GAC)than for the r(A–U) pairs. As expected, the hydrationpatterns for the r(GAC) and r(A–U) pairs are different.While three water molecules are in contact with the deepgroove atoms, three (GAC) and two (A–U) water mol-ecules are located in their shallow groove. The hydrationpatterns around the backbone are essentially sequenceindependent. Well-defined hydration cones around theanionic oxygen atoms and the hydroxyl groups couldalso be observed. Water molecules with long residencetimes (�700–800 ps) are attached to the OR atoms ofthe phosphate groups (Figure 2). Other water moleculeswith residence times in the 500 ps range are mainlylocated in the vicinity of the (G/A)N7, (G)O6, (A)N6,and (U)O4 deep groove atoms. The residence times ofwater molecules bound to the shallow groove O2�-Hgroups are especially short (�100 ps; Figure 2). Theydisplay a water-like behavior characterized by a contin-ued formation and breakage of solute/solvent hydrogenbonds. Thus, although hydroxyl groups interact prefer-entially with water and tend to avoid the formation ofintramolecular O2�-H(n) . . . O4�(n � 1) hydrogenbonds, they do not favor long-lived shallow groovehydration patterns.41 A comparison with hydration pat-

terns derived for the d(GAC) and d(A–T) pairs ex-tracted from B-DNA duplexes reveals significant differ-ences that will not be discussed here.20,21

Non-Watson–Crick Base Pairs

The non-Watson–Crick pairs play major structuraland functional roles in RNA architectures42,43 and areassociated with specific hydration patterns.16,24

FIGURE 1 Different views of the hydration shell sur-rounding a r(GAC) base pair from a 2.4 ns MD simulationsof a r(CpG)12 duplex.20,22 Top: Low and high level pseudo-electron densities of water molecules calculated from thecloud of waters surrounding the base pair by using Fouriertransformation. Middle: Water molecules fitted manually inthe density peaks shown above by using the O program.79

The position of the water molecules was subsequently re-fined by using the SHELX program.80 The position of a K�

ion in the vicinity of the (G)O6 atom is shown in yellow.Note that the position of the K� nearly overlaps with theposition of a hydration site. Bottom: “Pseudo” thermalellipsoids plotted at a 50% level of probability by using theRASTEP program, which is part of the RASTER3D mo-lecular graphics package.81

268 Auffinger and Westhof

Page 4: RNA solvation: A molecular dynamics simulation perspective

G�U Pairs. In order to evaluate the structural effectsassociated with a conserved G3�U70 base-pair presentin the tRNAAla acceptor arm, the hydration of severalvariants of this motif has been investigated on thebasis of seven 2.5 ns of MD simulations.30 The au-thors showed that the G�U pair, as found in crystals,44

induces local deviations from A-form geometry, i.e.,the wild-type helix is underwound at the base-pairstep above the G�U pair and overwound at the base-pair step below the G�U pair in each case by about 7–9degrees compared to an A-form helix. They discussedthe presence of tightly bound water molecules on theshallow groove side of the G�U pair that display longresidence times (�500 ps), and correlate with watermolecules observed by crystallography and by multi-ple MD simulations for a G�U pair embedded in thetRNAAsp anticodon hairpin.29 These earlier simula-tions have also revealed that water molecules occu-pying the shallow groove hydration site in contactwith the (U)O2�, (U)O2, and (G)H21 atoms displaylong residence times (�450 ps). No long-lived hydro-gen bonds involving the (U)O2� atom were observed,but long-lived hydrogen bonds with the (U)O2 and the(G)N2—H21 atoms were detected. Remarkably, for aU3�G70 pair replacing the G3�U70 pair, a significantlysmaller (in size and magnitude) water density peak isobserved compared to that found for the wild-typebase pair.30

G�dU, I�U, and 2AA � IsoC Pairs. G�dU, I�U, and2AA � IsoC base pairs were inserted in the tRNAAla

acceptor stem described above in order to estimatedeformations introduced in the hydration shell of thewild-type structure.30 Removal of the (U)O2�—H

group (G�dU) or of the (G)NH2 group (I�U) does notalter significantly the shallow groove water densitypeak observed for the G�U pair in line with x-ray datashowing that a water molecule is systematically con-nected to the O2 atoms of G�T and I�T pairs.5 For the2AA � IsoC pair as well, a region of high water densityis seen at the same location. Thus, the occurrence ofthis hydration site is probably more related to thenotch created by the wobble geometry of the basepairs than to the presence and type of hydrophilicgroups lining this cavity.15

The residence times of water molecules connectedto all these sites are different. The longest calculatedresidence times follow the order G�U (537 ps)� 2AA�IsoC (324 ps) � U�G (224 ps) � G�dU (�180ps) � I�U (�140 ps) � GAC (�50 ps) in line withother data collected from MD simulations on G�U29

and GAC20 pairs. Thus, water densities alone are nota sufficient criterion to estimate the strength of abinding site. Residence times have to be taken intoaccount as well.

C�U Pairs. MD simulations have been conducted ona r(GGACUUCGGUCC)2 duplex that contains twoG�U pairs framing two water-mediated C�U pairs.45 Ithas been observed that the U�C pairs switch during theMD run between two types of conformations involv-ing an interaction between the (C)NH2 group andeither the (U)O2 or the (U)O4 atoms. For the confor-mations involving a (C)NH2 . . . O2(U) interaction,water molecules displaying long residence times(�300 ps; in one occurrence a 1 ns residence time isobserved) are located in the shallow groove. For theconformation involving a (C)NH2 . . . O4(U) interac-

FIGURE 2 Profiles emphasizing the difference in residence times of water molecules stronglybound to the deep groove OR atoms (�800 ps) and weakly bound to the shallow groove O2� atoms(�100 ps). Hydrogen-bonding times, calculated from 2.4 ns simulations of the r(CpG)12 and ther(UpA)12 duplexes,20–22 are plotted versus the number of water molecule bound to the OR and O2�atoms. The data have been normalized with respect to the number of molecules present in theduplex, and therefore differ from the non-normalized data presented in preceding work.20,21

RNA Solvation 269

Page 5: RNA solvation: A molecular dynamics simulation perspective

tion, stable water molecules (�300 ps) are located inthe deep groove of the duplex. These conformationsfor the C � U pairs are different from the bifurcatedhydrogen-bonded C � U pair observed in MD simula-tions of the tRNAAsp anticodon loop that involve awater-mediated base-to-backbone interaction.46

G�A Pairs. A transition between a conformation of aG�A pair involving two water molecules and a water-mediated conformation has been observed in a MDsimulation of a GCAA tetraloop.47 This water-medi-ated conformation involves most likely a long-livedwater molecule.

Modified Nucleotides

Naturally modified nucleotides are more than anec-dotic features of nucleic acids. They are present inRNAs from all three kingdoms (archaea, bacteria,eukaryote) where they occupy strategic positions.48

Besides the dU, I, 2AA, and isoC nucleotides men-tioned before, some study dealt with pseudouridines(�) and 2�O-methylations. The MD simulations de-scribed below emphasize important structural modifi-cations of the hydration shell related to the presenceof modified nucleotides embedded into RNA struc-tures.

Pseudouridines (�). Among other roles, pseudouri-dines, which display a C1�—C sugar–base linkagewith an N1—H group replacing the C5—H grouppresent in uridines, are involved in the stabilization ofthe structure of the hydration shell of specific struc-tural motifs.49,50 It has been observed in six multipleMD simulations of the yeast tRNAAsp anticodon hair-pin that the water molecules that display the longestresidence times form very stable hydrogen-bond con-tacts with the � base and the nucleic acid back-bone.29,49 These water molecules establish three hy-drogen bonds, i.e., an acceptor hydrogen bond withthe (�)N1—H group and two donor hydrogen bondswith anionic oxygen atoms of adjacent phosphategroups. Thus, a strong (�)N1—H . . . Ow bond re-places a (U)C5—H . . . Ow interaction and leads tothe stabilization of a water molecule with very longresidence times.51 This water molecule helps to lockthe base in a specific conformation with respect to itsbackbone, and thus is part of a “nucleotide/water”complex (Figure 3).

2�O-Methylations. Methylations of the 2�-hydroxylgroups are very common in RNA molecules. Based onx-ray and NMR structures, MD simulations of ther(CGCGCG)2 and 2�-O-Me(CGCGCG)2 duplexes were

carried on.26 Interestingly, compared to the nonmodifiedstructure, a more ordered hydration shell around the O2�atoms of the methylated duplex was observed, and spe-cific shallow groove hydration sites occupied by longlived (�1ns) water molecules were detected.82

RNA Duplexes

MD simulations on regular RNA duplexes formed byWatson–Crick pairs lead to consistent views.19–21,52

RNA duplexes are found to be more rigid than B-DNAduplexes of similar sequence. Interestingly, while se-quence dependent long-lived hydration patterns (�1 ns)are found in the minor groove of B-DNA linking the O4�with the N3 or O2 atoms, no such hydration patterns arefound for RNA.21 Instead, long-lived hydration patterns(�800 ps) are located in the deep groove and involvenonsequence-dependent water bridges between adjacentphosphate groups. Indeed, there is clearly a relationbetween the shape and dynamics of the hydration pat-terns and the flexibility of the nucleic acid structures.Nevertheless, this relationship is intricate and thereforedifficult to quantify.

Hybrid Duplexes and ChemicallyModified Backbones

DNA/RNA Hybrids. Hydration patterns aroundDNA/RNA hybrids were observed to be less defined

FIGURE 3 Snapshot extracted from a MD simulation ofthe tRNAAsp anticodon hairpin showing a strongly boundwater molecule (�500 ps).29,49 This water molecule, whichis involved simultaneously in three hydrogen bonds withbase and backbone atoms, is part of a “nucleotide/water”complex.

270 Auffinger and Westhof

Page 6: RNA solvation: A molecular dynamics simulation perspective

than around an equivalent B-DNA or an RNA du-plex.19 In this study, it is shown that the more flexiblehybrid structure has less defined hydration patterns.

Phosphoramidate Analogs. Radial distribution func-tions calculated for a 10 base-pair long DNA/RNAhybrid and two 3�pnDNA–RNA and 5�pnDNA–RNAphosphoramidate analogs were used to propose anexplanation for the observed differential stability ofthese duplexes.53 Experimentally, it is known that3�pnDNA–RNA are more stable than correspondingphosphodiesters, while 5�pnDNA–RNA do not formduplexes at all.

PNA/RNA. Hydration maps were calculated forPNA/DNA and PNA/RNA duplexes by using a gridmethod and were found to be completely different forboth molecules.54 Interestingly, the PNA/DNA struc-ture exhibits a spine of hydration in the minor groovesimilar to that found for DNA, while the PNA/RNAhelix shows, like RNA duplexes,19–21 a better hydra-tion of the deep groove.

HNA/RNA. Experimentally, it is known that molec-ular association between HNA and RNA is morestable than between HNA and DNA and betweennatural nucleic acids (dsDNA, dsRNA, DNA/RNA).A thorough description of the hydration of hexitolnucleic acid duplexes (HNA/DNA and HNA/RNA)led to the observation that a better hydration of theHNA/RNA vs the HNA/DNA duplex correlates witha reduced flexibility.55

2�-Substituted Analogs. Nanosecond simulations ofa MOE(CCAACGTTGG)–r(CCAACGUUGG) du-plex were MOE stands for 2�-sugar substituted O-(2-methoxyethyl) have been reported.56 A short descrip-tion of the hydration of the duplex is given thatemphasizes the occurrence of a hydration site betweenthe O3� oxygen atom of the backbone and the oxygenatom embedded in the MOE chain.

Complex RNA Structures

RNA hairpins, which are of higher complexity thandouble-stranded structures, are more sensitive to thetreatment of electrostatic interactions than DNA du-plexes. Therefore, the first successful long (200–1000ps) MD simulations of RNA could only be achievedwhen accurate treatment of the long-range electro-static interactions became available.47,57–59 The firstpublished simulation described a water insertion eventat the level of the G � A base pair, which closes aGNRA tetraloop.47 Multiple molecular dynamics sim-

ulations conducted on the yeast tRNAAsp anticodonhairpin revealed the occurrence of well-defined hy-dration patterns associated with water molecules dis-playing long residence times (�500 ps).29,41,49,51 Theoccurrence of long-lived hydration patterns in theloop was proposed to be associated with the presenceof modified nucleotides (�). Thus, among other roles,modified nucleotides stiffen the hairpin structurethrough the stabilization of its hydration shell.

Simulations of more complex RNA systems arestill rare60 and even rarer are descriptions of theirhydration shell. Some aspects of the hydration of theflavin–mononucleotide/RNA aptamer complex weredescribed.61 Several water mediated protein/RNAcontacts observed in a 0.6 ns MD simulation of thecomplex between the human U1A protein and thehairpin II of the U1 small nuclear RNA have beenreported.62 Water mediated contacts between amino-glycosides and the hammerhead ribozyme were alsodescribed,63 and similar contacts were observed in thecrystal structure between paromomycin and a RNAoligomer.83

Ion Binding Features

The literature describing ion binding events to DNAstructures both from an experimental and theoreticalpoint of view is growing.4,11,21,33,64–66 For RNA,experimental evidence emphasize the structural in-volvement of monovalent ions that, therefore, shouldno longer exclusively be considered as neutral addi-tives.67–72 Binding sites for NH4

� cations have beencharacterized in a RNA loop, based on multiple mo-lecular dynamics simulations of the yeast tRNAAsp

anticodon hairpin.29 These ions utilize simultaneouslyabout three of their N—H bonds to interact withelectronegative atoms of the RNA. Sodium ion bind-ing events to the major groove of a RNA duplex havebeen described.19 Potassium ions are found to bindspecifically to sites located in the deep groove ofr(GpC) and r(ApU) steps but do not penetrate the deepgroove of r(CpG) and r(UpA) steps mainly due tosteric and electrostatic repulsion constrains emphasiz-ing expected sequence specific effects (Figure 4).73

Given the rapid exchange rates of water moleculesbound to monovalent ions, such ions represent inter-esting probes for locating possible binding sites ofelectropositive chemical groups.

The question of the divalent ions is more difficultto address since the dehydration steps of these ions(�10 �s for Mg2�) cannot be simulated on the cur-rently accessible time scales. However, data havebeen obtained for hexahydrated Mg2� ions bound toDNA,74 and Brownian dynamics methods have been

RNA Solvation 271

Page 7: RNA solvation: A molecular dynamics simulation perspective

used to locate Mg2� binding sites around static RNAstructures.75

Perspectives

The field of MD simulations of biological macromol-ecules is now becoming mature since reliable simu-lations can be generated on nanosecond time scales.This assertion is reasonable at least for systemsaround equilibrium or located in conformational “at-tractors.” As an outcome, the structural hydrationcharacteristics are rather well reproduced. The valid-ity of the derived lifetimes for the various hydrationsites is on a slightly less secure basis and has to beconsidered on a semiquantitative basis. The case ofcharged ions, especially divalent (Mg2�) ions, wouldstill be worse. Will MD simulations ever bridge thegap between macroscopic thermodynamical ap-proaches like the nonlinear Poisson–Boltzmann mod-el75,76 and the highly precise and localized crystalstructures of complex tRNAs?77,78 On one side, it isshown that electrostatics is sufficient for associatingMg2� ions with RNA molecules without involvingprecise hydrogen-bonding or coordination schemes.76

On the other hand, x-ray structures reveal a rich arrayof contacts between ions, waters, and RNA atomspositioned precisely in space by the three-dimensional(3D) fold. Indeed, the crystal clear perspective ofbinding sites may induce an overemphasis on the

importance of specifically bound water and ions instabilizing native tertiary folds of structured RNAs insolution. This view of static structures, as derivedfrom x-ray experiments, is mainly “enthalpic” (coor-dination, hydrogen bonds, van der Waals con-tacts, . . . ). Yet entropic contributions should not beneglected. For example, in the folding process itself,the entropic contribution of solvent release (water andions), although only indirectly observed, is primor-dial. MD simulations, by distinguishing betweenstrongly bound and more labile solvent molecules arecentral to the evaluation process of the enthalpic/entropic balance and compensation.

REFERENCES

1. Saenger, W. Principles of Nucleic Acid Structure;Springer-Verlag: New-York, 1984.

2. Jeffrey, G. A.; Saenger, W. Hydrogen Bonding in Bi-ological Structures; Springer-Verlag: Berlin, 1991.

3. Wuthrich, K. Acta Cryst 1995, D51, 249–270.4. Halle, B.; Denisov, V. P. Biopolymers 1998, 48, 210–

233.5. Auffinger, P.; Westhof, E. In Roles of Hydration on the

Structure and Dynamics of Nucleic Acids; Ross, Y. H.,Leslie, R. B., Lillford, P. J. Eds.; Technonic Publishing:Basel, 1999; pp 165–198.

6. Chalikian, T. V.; Volker, J.; Srinivasan, A. R.; Olson,W. K.; Breslauer, K. J. Biopolymers 1999, 50, 459–471.

FIGURE 4 Snapshots extracted from 2.4 ns MD simulations of the r(CpG)12 and the r(UpA)12

duplexes showing K� binding sites associated with the r(GpC) and the r(ApU) steps.20,21 The K�

ions are represented in yellow. Given specific steric and electrostatic constrains, no ion binding siteis observed in the deep groove of the r(CpG) and r(UpA) steps.

272 Auffinger and Westhof

Page 8: RNA solvation: A molecular dynamics simulation perspective

7. Auffinger, P.; Westhof, E. Curr Op Struct Biol 1998, 8,227–236.

8. Feig, M.; Pettitt, B. M. Structure 1998, 6, 1351–1354.9. Feig, M.; Pettitt, B. M. Biopolymers 1998, 48, 199–

209.10. Feig, M.; Pettitt, B. M. J Mol Biol 1999, 286, 1075–

1095.11. Beveridge, D. L.; McConnell, K. J. Curr Opin Struct

Biol 2000, 10, 182–196.12. Cheatham, T. E.; Kollman, P. A. Annu Rev Phys Chem

2000, 51, 435–471.13. Pettitt, B. M.; Makarov, V. A.; Andrews, B. K. Curr

Opin Struct Biol 1998, 2.14. Makarov, V. A.; Andrews, B. K.; Smith, P. E.; Pettitt,

B. M. Biophys J 2000, 79, 2966–2974.15. Westhof, E. Annu Rev Biophys, Biophys Chem 1988,

17, 125–144.16. Westhof, E.; Beveridge, D. L. In Hydration of Nucleic

Acids; Franks, F., Ed.; Cambridge University Press:Cambridge, 1990; Vol 5, pp 24–123.

17. Egli, M.; Portmann, S.; Usman, N. Biochemistry 1996,35, 8489–8494.

18. Egli, M.; Tereshko, V.; Teplova, M.; Minasov, G.;Joachimiak, A.; Sanishvili, R.; Miller, R.; Maier, M. A.;An, H.; Cook, P. D.; Manoharan, M. Biopolymers2000, 48, 234–252.

19. Cheatham, T. E.; Kollman, P. A. J Am Chem Soc 1997,119, 4805–4825.

20. Auffinger, P.; Westhof, E. J Mol Biol 2000, 300, 1115–1133.

21. Auffinger, P.; Westhof, E. J Mol Biol 2001, 305, 1057–1072.

22. Auffinger, P.; Masquida, B.; Westhof, E. LNCS 2001,in press.

23. Schneider, B.; Berman, H. M. Biophys J 1995, 69,2661–2669.

24. Auffinger, P.; Westhof, E. J Biomol Struct Dynam1998, 16, 693–707.

25. Schneider, B.; Patel, K.; Berman, H. M. Biophys J1998, 75, 2422–2434.

26. Kulinska, K.; Kulinski, T.; Lyubartsev, A.; Laaksonen,A.; Adamiak, R. W. Comp Chem 2000, 24, 451–457.

27. Hummer, G.; Garcia, A. E.; Soumpasis, D. M. BiophysJ 1995, 68, 1639–1652.

28. Hummer, G.; Garcia, A. E.; Soumpasis, D. M. FaradDiscuss 1996, 103, 175–189.

29. Auffinger, P.; Westhof, E. J Mol Biol 1997, 269, 326–341.

30. Nagan, M. C.; Kerimo, S. S.; Musierforsyth, K.; Cra-mer, C. J. J Am Chem Soc 1999, 121, 7310–7317.

31. Tsui, V.; Radhakrishnan, I.; Wright, P. E.; Case, D. A.J Mol Biol 2000, 302, 1101–1117.

32. Bruge, F.; Parisi, E.; Formili, S. L. Chem Phys Lett1996, 250, 443–449.

33. Feig, M.; Pettitt, B. M. Biophys J 1999, 77, 1769–1781.34. van der Vaart, A.; Merz, K. M. J Am Chem Soc 1999,

121, 9182–9190.

35. Darden, T.; Perera, L.; Li, P.; Pedersen, L. Structure1999, 7, R55–R60.

36. Sagui, C.; Darden, T. A. Annu Rev Biophys Struct1999, 28, 155–179.

37. Cheatham, T. E.; Kollman, P. A. In Molecular Dynam-ics Simulation of Nucleic Acids in Solution: How Sen-sitive Are the Results to Small Perturbations in theForce Field and Environment?; Sarma, R. H., Sarma,M. H., Eds.; Adenine Press: New York, 1998; pp 99–116.

38. Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.J Chem Phys 1983, 79, 926–936.

39. Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. JPhys Chem 1987, 97, 6269–6271.

40. Tarek, M.; Tobias, D. J. Biophys J 2000, 79, 3244–3257.

41. Auffinger, P.; Westhof, E. J Mol Biol 1997, 274, 54–63.

42. Leontis, N. B.; Westhof, E. Quart Rev Biophys 1998,31, 399–455.

43. Hermann, T.; Westhof, E. Chem Biol 1999, 6, R335–343.

44. Masquida, B.; Westhof, E. RNA 2000, 6, 9–15.45. Schneider, C.; Brandl, M.; Suhnel, J. J Mol Biol 2001,

305, 659–667.46. Auffinger, P.; Westhof, E. J Mol Biol 1999, 292, 467–

483.47. Zichi, D. A. J Am Chem Soc 1995, 117, 2957–2969.48. Grosjean, H.; Benne, R. Modification and Editing of

RNA; American Society for Microbiology: Washing-ton, DC 2005, 1998.

49. Auffinger, P.; Westhof, E. In Effects of Pseudouridyla-tion on tRNA Hydration and Dynamics: A TheoreticalApproach; Grosjean, H., Benne, R., Eds.; AmericanSociety for Microbiology: Washington, DC, 1998; pp103–112.

50. Davis, D. R. In Biophysical and Conformational Prop-erties of Modified Nucleosides in RNA (Nuclear Mag-netic Resonance Studies); Grosjean, H., Benne, R.,Eds.; American Society for Microbiology: Washington,DC, 1998; pp 85–102.

51. Auffinger, P.; Louise-May, S.; Westhof, E. Farad Dis-cuss 1996, 103, 151–174.

52. Cheatham, T. E.; Cieplak, P.; Kollman, P. A. J BiomolStruct Dynam 1999, 16, 845–862.

53. Barsky, D.; Colvin, M. E.; Zon, G.; Gryaznov, S. M.Nucleic Acids Res 1997, 25, 830–835.

54. Soliva, R.; Sherer, E.; Luque, F. J.; Laughton, C. A.;Orozco, M. J Am Chem Soc 2000, 122, 5997–6008.

55. De Winter, H.; Lescrinier, E.; Van Aerschot, A.; Herd-ewijn, P. J Am Chem Soc 1998, 120, 5381–5394.

56. Lind, K. E.; Mohan, V.; Manoharan, M.; Ferguson,D. M. Nucleic Acids Res 1998, 26, 3694–3799.

57. Cheatham, T. E.; Miller, J. L.; Fox, T.; Darden, T. A.;Kollman, P. A. J Am Chem Soc 1995, 117, 4193–4194.

58. Auffinger, P.; Westhof, E. Biophys J 1996, 71, 940–954.59. Louise-May, S.; Auffinger, P.; Westhof, E. Curr Opin

Struct Biol 1996, 6, 289–298.

RNA Solvation 273

Page 9: RNA solvation: A molecular dynamics simulation perspective

60. Zacharias, M. Curr Op Struct Biol 2000, 10, 311–317.61. Schneider, C.; Suhnel, J. Biopolymers 1999, 50, 287–

302.62. Tang, Y.; Nilsson, L. Biophys J 1999, 77, 1284–1305.63. Hermann, T.; Westhof, E. J Mol Biol 1998, 276, 903–

912.64. Lyubartsev, A. P.; Laaksonen, A. J Biomol Struct Dy-

nam 1998, 16, 579–592.65. McFail-Isom, L.; Sines, C. C.; Williams, L. D. Curr

Opin Struct Biol 1999, 9, 298–304.66. Williams, L. D.; Maher, L. J. Annu Rev Biomol Struct

2000, 29, 497–521.67. Gluick, T. C.; Gerstner, R. B.; Draper, D. E. J Mol Biol

1997, 270, 451–463.68. Basu, S.; Rambo, R. P.; Strauss-Soukup, J.; Cate, J. H.;

Ferre-D’Amare, A. R.; Strobel, S. A.; Doudna, J. A.Nat Struct Biol 1998, 5, 986–992.

69. Klosterman, P. S.; Shah, S. A.; Steitz, T. A. Biochem-istry 1999, 38, 14784–14792.

70. Batey, R. T.; Rambo, R. P.; Lucast, L.; Rha, B.;Doudna, J. A. Science 2000, 287, 1232–1239.

71. Shiman, R.; Draper, D. E. J Mol Biol 2000, 302, 79–91.

72. Nissen, P.; Hansen, J.; Ban, N.; Moore, P. B.; Steitz,T. A. Science 2000, 289, 920–930.

73. Young, M. A.; Jayaram, B.; Beveridge, D. L. Am ChemSoc 1997, 119, 59–69.

74. Young, M. A.; Beveridge, D. L. J Mol Biol 1998, 281,675–687.

75. Hermann, T.; Westhof, E. Structure 1998, 6, 1303–1315.

76. Misra, V. K.; Draper, D. E. J Mol Biol 2000, 299,813–825.

77. Jovine, L.; Djordjevic, S.; Rhodes, D. J Mol Biol 2000,301, 401–414.

78. Shi, H.; Moore, P. B. RNA 2000, 6, 1091–1105.79. Jones, T. A.; Zou, J. Y.; Cowan, S. W.; Kejldgaard, M.

Acta Cryst 1991, A47, 110–119.80. Sheldrick, G. M.; Schneider, T. R. Methods Enzymol

1997, 276, 307–326.81. Merritt, E. A.; Bacon, D. J. Methods Enzymol 1997,

277, 505–524.82. Auffinger, P.; Westhof, E. Agnew Chem Int Ed 2001, in

press.83. Vicens, Q.; Westhof, E. Structure 2001, 9, 647–658.

274 Auffinger and Westhof