D RNA Kink-Turn ... - ACS Publications

J. Phys. Chem. B 2010, 114, 10581–10593

10581

Structural Dynamics of the Box C/D RNA Kink-Turn and Its Complex with Proteins: The Role of the A-Minor 0 Interaction, Long-Residency Water Bridges, and Structural Ion-Binding Sites Revealed by Molecular Simulations Nad’a Sˇpacˇkova´, Kamila Re´blova´, and Jir˘´ı Sˇponer* Institute of Biophysics, Academy of Sciences of the Czech Republic, Kra´loVopolska´ 135, 61265 Brno, Czech Republic ReceiVed: March 22, 2010; ReVised Manuscript ReceiVed: June 24, 2010

Kink-turns (K-turns) are recurrent elbow-like RNA motifs that participate in protein-assisted RNA folding and contribute to RNA dynamics. We carried out a set of molecular dynamics (MD) simulations using parm99 and parmbsc0 force fields to investigate structural dynamics of the box C/D RNA and its complexes with two proteins: native archaeal L7ae protein and human 15.5 kDa protein, originally bound to very similar structure of U4 snRNA. The box C/D RNA forms K-turn with A-minor 0 tertiary interaction between its canonical (C) and noncanonical (NC) stems. The local K-turn architecture is thus different from the previously studied ribosomal K-turns 38 and 42 having A-minor I interaction. The simulations reveal visible structural dynamics of this tertiary interaction involving altogether six substates which substantially contribute to the elbow-like flexibility of the K-turn. The interaction can even temporarily shift to the A-minor I type pattern; however, this is associated with distortion of the G/A base pair in the NC-stem of the K-turn. The simulations show reduction of the K-turn flexibility upon protein binding. The protein interacts with the apex of the K-turn and with the NC-stem. The protein-RNA interface includes long-residency hydration sites. We have also found long-residency hydration sites and major ion-binding sites associated with the K-turn itself. The overall topology of the K-turn remains stable in all simulations. However, in simulations of free K-turn, we observed instability of the key C16(O2′)-A7(N1) H-bond, which is a signature interaction of K-turns and which was visibly more stable in simulations of K-turns possessing A-minor I interaction. It may reflect either some imbalance of the force field or it may be a correct indication of early stages of unfolding since this K-turn requires protein binding for its stabilization. Interestingly, the 16(O2′)-7(N1) H- bond is usually not fully lost since it is replaced by a water bridge with a tightly bound water, which is adenine-specific similarly as the original interaction. The 16(O2′)-7(N1) H-bond is stabilized by protein binding. The stabilizing effect is more visible with the human 15.5 kDa protein, which is attributed to valine to arginine substitution in the binding site. The behavior of the A-minor interaction is force-field-dependent because the parmbsc0 force field attenuates the A-minor fluctuations compared to parm99 simulations. Behavior of other regions of the box C/D RNA is not sensitive to the force field choice. Simulation with net-neutralizing Na+ and 0.2 M excess salt conditions appear in all aspects equivalent. The simulations show loss of a hairpin tetraloop, which is not part of the K-turn. This was attributed to force field limitations. Introduction The box C/D RNA is a common type of small noncoding RNAs found in eukaryotic and archaeal organisms in small nucleolar RNAs (snoRNA) and small RNAs (sRNA), respectively. The box C/D RNA associates with specific proteins, leading to formation of ribonucleoprotein (RNP) complexes.1,2 The primary function of the box C/D RNA in the RNP complex is to serve as a guide for site-specific 2′-O-methylation of ribose sugars within rRNAs.3 The box C/D RNA forms a secondary structure motif called a Kink-turn (K-turn).4–8 K-turns are recurrent asymmetric internal loops with a sharp bend in the phosphodiester backbone.4 They consist of three structural parts, namely, the noncanonical (NC) and canonical (C) stems that are flanking the unpaired internal (bulge) segment (Figure 1). K-turns possess five characteristic structural signatures:4,9 canonical base pair at the end of the C-stem, two tandem trans-Hoogsteen/sugar* Corresponding author. E-mail: [email protected]. Phone: +420 541 517 133. Fax: +420 541 211 293.

Figure 1. Schematic representation and 3D structure of the box C/D RNA K-turn. Standard nomenclature for non-WC base pairs is used;10 circles represent Watson-Crick edges, squares Hoogsteen edges, triangles sugar edges, open symbols trans pairing, and closed symbols cis pairing.

edge (trans-H/SE) A/G base pairs at the end of the NC-stem, and two tertiary interactions between the stems. The first one is trans-SE/SE base pair between the 5′-end nucleotide (its 2′OH) of the bulge and adenine (its N1) of the terminal A/G

10.1021/jp102572k  2010 American Chemical Society Published on Web 07/26/2010

10582

Sˇpacˇkova´ et al.

J. Phys. Chem. B, Vol. 114, No. 32, 2010

base pair of the NC-stem (standard annotation of base pairs is used in this paper10). This interaction is essential for folding of Kink-turns.11,12 The second tertiary interaction is the A-minor interaction13 between the preceding A/G base pair in the NCstem and the terminal base pair of the C-stem. The A-minor interaction significantly contributes to the topology of K-turn and is essential for its internal structural dynamics.14 A-minor interactions are the most frequent tertiary interactions in large functional RNAs and ribonucleoprotein particles such as ribosome.13 They are formed by adenines (either unpaired or involved in non-Watson-Crick base pairs) inserted through their sugar edges into a minor groove of a canonical A-RNA double helix, preferentially at CdG base pairs. The A-minor interactions can adopt several substates.13,15 The available X-ray structures show that K-turns possess either A-minor type I or type 0 interactions. The reason why some K-turns have A-minor I and others A-minor 0 is not known. The K-turn RNA motif was first observed in the crystal structure of the complex of a human U4 snRNA and 15.5 kDa protein.6 The K-turn was also found in crystallographic studies of a molecular complex between an archaeal L7ae protein with the box C/D sRNA.16 Finally, multiple K-turns were found in the ribosomal subunits.4 The K-turn motif was extensively studied using molecular dynamics (MD) simulations with explicit representation of solvent and counterions.14,17–23 MD simulations reveal structural dynamics and flexibility of studied molecules not apparent from the crystallographic studies. For example, the K-turn from U4 snRNA, K-turns 38, 42, and 58 from 23S rRNA of Halobacterium marismortui as well as larger RNA fragments with K-turns (GTPase-associated center segment and base of A-site finger of 23S rRNA) were characterized by simulations. The simulations revealed that K-turns possess unique elbow-like anisotropic and anharmonic flexibility, so that it has been suggested that some K-turns may facilitate largescale functional movements in the ribosome.14,18,23 Biochemical studies provided in-depth insights into the physical chemistry of folding of K-turns. K-turns typically require proteins or high concentration of divalent cations to adopt the functional tightly kinked geometry.12,24–28 Thus, many K-turns play an important role in protein-assisted RNA folding. Recently, the isolated K-turn (U4 snRNA) was studied by NMR and small-angle neutron scattering method.29 This study reveals unkinked geometry of the K-turn, as expected from the biochemical investigations. As noted above, preceding simulations of isolated K-turn motifs show considerable elbow-like flexibility of K-turns which behave as flexible hinges. Local dynamics at the apex of the V-shaped molecule propagates to the attached rather stiff Cand NC-helical arms, leading to opening/closing fluctuations of both arms. Fluctuations of the folded K-turn between the most tightly closed geometry and partially open structures are connected with dynamics of the A-minor tertiary interaction, mainly (in case of the A-minor I interaction) with dynamical insertion of long-residency water molecules into its cis-SE/SE base pair.14,19,22 Note that the notion closed and open geometry may have different meaning in different studies, depending on whether they investigate dynamics within the conformational space of the folded K-turn or overall unfolding (unkinking) of the K-turn. Typically, MD simulations provide detailed sampling of fluctuations of the K-turns associated with their starting (tightly kinked) functional geometries. This dynamics within the tightly kinked state, albeit substantial, is not easily resolvable by biochemical studies. In contrast, the experiments capture equilibrium between tightly kinked functional state of the K-turn

and unkinked geometries. The tightly kinked-unkinked structure transition is not reachable by the standard simulations due to the limited time scale of the simulations. Unkinking of the K-turn was recently achieved using replica-exchange umbrellasampling free-energy computations.22 Note also that, besides the leading hinge-like flexibility, the K-turn architectures show also substantial twisting flexibility components, allowing them a wide range of structural fluctuations and adaptations.14,22,23 In this study, we present MD simulations of the box C/D RNA K-turn structure (close to 1 µs in total simulation time) which substantially extend previous computational studies on K-turns. We investigate a complex of the archaeal box C/D RNA and the L7ae protein observed in the crystal,16 a modeled complex between the archaeal box C/D RNA and the human 15.5 kDa protein (based on the ability of 15.5 kDa protein to interact with the box C/D RNA8,12,30), and the isolated box C/D RNA. The box C/D K-turn experimental structure possesses a different type of the A-minor interaction between the C- and NC-stems compared with previously extensively studied ribosomal K-turns 38 and 42.14,18 The present simulations reveal that the type 0 A-minor leads to quite different local dynamics of the K-turn compared to the A-minor I interaction. Another purpose of this study was to analyze the influence of the protein binding on the structure and flexibility of the K-turn. Further, we compare performance of two variants of the Cornell et al.31 “AMBER” empirical force field: parm9932,33 and its parmbsc034 version with modified R/γ backbone torsion angle profiles. In addition, we compare simulations carried out with net-neutralizing set of Na+ cations (the most common condition in nucleic acid simulation literature) and ∼0.2 M excess salt KCl simulations. Our simulations confirm that the box C/D K-turn behaves as an elbow-like RNA motif similarly to the K-turns studied earlier. The elbow-like flexibility of the box C/D K-turn (in its folded geometry) is intimately associated with substates of the A-minor interaction between C- and NC-stem which, however, differ from ribosomal K-turns 38 and 42.14,18 The A-minor substates lead to characteristic averaged angles between the K-turn arms. The unbound box C/D K-turn shows extensive local instability of the critical trans-SE/SE interaction between the last adenine of the NC-stem and the first unpaired nucleotide. Nevertheless, this signature interaction is often replaced by a water bridge with exceptionally tightly bound water molecule, which still appears to be specific for the signature N1(A) position. In the absence of the protein, there is major structural involvement of cations at the bulge region of the K-turn which may reflect the known sensitivity of K-turn folding to ions.24,26 The simulations showed reduced flexibility and clear stabilization of the K-turn in the RNA-protein complexes. Important stabilizing role is played by water molecules forming long-residency hydration sites in the RNA-protein interface. The dynamics of the key A-minor interaction is influenced by the force field choice, while behavior of the other parts of the K-turn molecule is not substantially influenced by the force field selection. Methods The simulations were carried out using the AMBER835 and AMBER936 programs with both parm9932,33 and parmbsc034 versions of Cornell et al.31 force field for RNA while using the ff99SB37 for the protein. Systems were neutralized with standard sodium ions (radius 1.868 Å, potential well depth 0.00277 kcal/ mol) using the Xleap module of AMBER. Several simulations were carried out using potassium ions (parameters by Dang,38 radius 1.8687 Å, potential well depth 0.100 kcal/mol) and higher

MD Simulations of the Box C/D RNA

J. Phys. Chem. B, Vol. 114, No. 32, 2010 10583

TABLE 1: List of Simulations simulation name CD1 CD2 CD3 CD4 CD5 CD6 CD7 CD8 CD9 CD_pL7_1 CD_pL7_2 CD_pL7_3 CD_pL7_4 CD_pL7_5 CD_pL7_6 CD_p15_1 CD_p15_2 CD_p15_3 CD_p15_4

simulated system C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D C/D

c

box box box boxd boxd boxd boxd boxd boxd box + protein L7ae, first complex boxd + protein L7ae, first complex box + protein L7ae, second complex box + protein L7ae, second complex boxd + protein L7ae, first complex boxd + protein L7ae, first complex box + protein 15.5 kDa box + protein 15.5 kDa boxd + protein 15.5 kDa boxd + protein 15.5 kDa

ionsa

MD length (ns) 40 40 40 50 50 50 50 50 50 40 50 40 40 50 50 40 40 50 50

+

Na Na+ Na+ Na+ Na+ Na+ Na+ 0.2 M 0.2 M Na+ Na+ Na+ Na+ 0.2 M 0.2 M Na+ Na+ 0.2 M 0.2 M

KCl excess salt KCl excess salt

KCl excess salt KCl excess salt KCl excess salt KCl excess salt

force field versionb parm99 parmbsc0 parm99 parmbsc0 parm99 parmbsc0 parm99 parmbsc0 parm99 parm99 parmbsc0 parm99 parm99 parmbsc0 parm99 parm99 parm99 parmbsc0 parm99

a Net-neutralizing, if not stated otherwise. b Proteins were described by parm99SB in all simulations. c C/D box stands for the box C/D RNA; see Figure 1. d Original U3/A23 base pair replaced by canonical U-A pair.

salt conditions corresponding to 0.2 M KCl (standard AMBER parameters for chloride correspond to parameters by Dang,38 radius 2.47 Å, potential well depth 0.100 kcal/mol). An octahedral box of TIP3P39 water molecules was added around the solute to a depth of 12 Å. Simulations were carried out using the particle mesh Ewald technique40 with 9 Å nonbonded cutoff and 2 fs time step. To improve sampling, we used two versions of equilibration protocols (both considered as adequate). First equilibration protocol started by 2000 steps of minimization followed by 100 ps of MD, with the system heated from 100 to 300 K. The atomic positions of the solute molecule were fixed by 50 kcal/mol Å2 restraint. Then, four minimization and MD simulation series were carried out with restraints of 25, 5, 3, and 1 kcal/mol Å2 applied to all solute atoms. Then 100 ps of unrestrained MD followed. The other protocol started by 1000 steps of minimization followed by 100 ps of MD, with the system heated from 100 to 300 K. The atomic positions of the solute molecule were fixed by 25 kcal/mol Å2 restraint. Then, five minimization and MD simulation series were carried out with restraints of 5, 4, 3, 2, and 1 kcal/mol Å2 applied to all solute atoms. Then 50 ps of restrained (0.5 kcal/mol Å2) and unrestrained MD followed. The production MD runs were carried out with constant pressure boundary conditions and constant temperature using Berendsen weak-coupling algorithm.41 SHAKE42 constraints were applied to all hydrogens while translational and rotational center-of-mass motion was removed every 2 ps.43 The standard simulation length was around 40-50 ns. Trajectories were analyzed using Ptraj module of AMBER. Visualization was carried out using VMD software.44 Studied Systems. Table 1 summarizes simulations carried out in this study. The initial structures were taken from two crystallographic studies. The first structure is the complex of a box C/D RNA and a L7ae protein16 (PDB: 1RLG, resolution 2.7 Å), which represents starting structure for both isolated RNA and RNA-protein simulations. As the crystal contains two marginally different structures, we used randomly both geometries as starts. Other starting structures are based on the crystal structure of a U4 snRNA in the complex with a 15.5 kDa protein6 (PDB: 1E7K, resolution 2.9 Å). Here, the U4 snRNA

was replaced by the box C/D K-turn from the 1RLG structure which was aided by similarity of both K-turn motifs. All bromouracils were replaced by uracils. Results Description of the RNA Structure. The K-turn (Figure 1) consists of three parts: the C-stem, the NC-stem, and the unpaired bases which we call the bulge. The C-stem contains two WC C8dG15 and C9dG14 base pairs and is followed by GAAA tetraloop. The NC-stem starts by two trans-H/SE A7/ G19 and A20/G6 base pairs followed by cis-WC U5/U21 base pair. As shown recently, the third base pair of the NC-stem can affect fine features of the K-turn folding and evolution.26 The NC-stem is followed by RNA stem with three canonical GdC and one cis-WC/H U3/A23 base pairs. The bulge of the K-turn is formed by three unpaired C16, G17, and U18 residues in the longer strand. C16 is stacked below the C-stem and forms transSE/SE base pair with A7 of the NC-stem. G17 is stacked below the NC-stem, while U18 is bulged out. The connection of Cand NC-stems is mediated mainly by C16/A7 trans-SE/SE interaction and the A-minor interaction between A20 and C8dG15. Residues G6, A7, G17, U18, and G19 adopt C2′endo sugar puckering. In summary, the studied system can be divided into five segments, in the 5′ to 3′ direction, which we call A-RNA, NC-stem, bulge, C-stem, and tetraloop throughout this study (Figure 1). The box C/D K-turn X-ray geometry is visibly different from the previously analyzed Kt-38 and Kt-42, while it resembles Kt-U4.14,17–23 The Kt-42 and Kt-38 have the fully paired A-minor I interaction between the C- and NC-stems, with the primary signature interaction formed by trans-SE/SE G/A base pair complemented by cis-SE/SE A/C base pair (Figure 2a). In the box C/D K-turn, the receptor GdC base pair is inverted so that the signature interaction is trans-SE/SE C/A base pair (Figure 2b). More importantly, the adenosine slides toward the cytosine (i.e., toward the shorter strand side of the C-stem), bringing the adenosine sugar-phosphate segment toward the center of the minor groove and abolishing the cis-SE/SE base pair. Similar A-minor 0 interaction is seen for example in Kt-46 of the H. marismortui large ribosomal subunit and also in the Kt-U4 (Figure 2c,d). It is not immediately apparent if this structural

10584



Figure 2. Four examples of the A-minor interaction: (a) type I in Kt-42, (b) type 0 in the box C/D, (c) type 0 in Kt-46 (from H. marismortui), and (d) type 0 in Kt-U4.

TABLE 2: Geometries of the Bulge Observed in our MD Simulations (See Figure 3) system type

substate name

description

RNA-protein CA ) X-ray C16 in the contact geometry with A7 only complex C16 in the contact CG with G17 only G17 stacked below isolated RNA GA A7 (A7/G19) GC G17 partly stacked below C16

H-bond contacts 16(O2′)-7(N1), 16(O2)-7(C2) 16(O2′)-17(N2) 16(O2′)-17(N2), 16(O2)-17(N2) 16(O2′)-17(O4′), 17(N2)-17(O2P)

variability of K-turn A-minor interactions is due to base sequence or whether it is due to protein binding and other interactions. Structural Dynamics of K-Turn. During our MD simulations, we observed several important geometrical changes in the K-turn structure. The bulge sampled different geometries in the isolated RNA and the RNA-protein complexes. The A-minor interaction exhibited structural fluctuations mainly in isolated systems. All of these changes are usually connected with formation of long-residency water bridges, and some of them are interrelated with variations of the interhelical angle of the K-turn. The original geometry of U3/A23 base pair (outside the K-turn motif) is unstable in all simulations; therefore, in many simulations, it has been manually replaced be canonical U-A base pair (see Table 1), which was stable. Bulge Region: Unexpected Instability of the C16(O2′)A7(N1) H-Bond. In the X-ray structure, C16 participates in the key tertiary 16(O2′)-7(N1) H-bond. This is the single most important interaction for structuring the tightly kinked geometries of K-turns.4,9,11,12 Surprisingly, this interaction is basically lost in simulations of free K-turn, while its behavior is connected with specific geometries of the bulge. Table 2 and Figure 3 summarize the typical geometries that are seen in simulations. The behavior is different in simulations of protein-RNA complexes and of free K-turn (Figure 4). In the RNA-protein systems, we see two substates marked as CA and CG (see Table 2 and Figure 3a). CA corresponds to the X-ray structure with 16(O2′)-7(N1) and 16(O2)-7(C2) H-bonds. Its population is ∼80% in simulations of 15.5 kDa complexes (Figure 4) with 16(O2′)-7(N1) distance of ∼2.8 Å. Alternatively, C16 makes contact with G17 (CG substate) via 16(O2′)-17(N2) H-bond. In this case, the distance between 16(O2′) and 7(N1) is around 3.5 Å or longer. This substate dominates in L7ae complex simulations where it is typically realized for 60-70% of simulation time (Figure 4). The 16(O2′) and 7(N1) interaction is, in this case, water-mediated (see below). G17 stays below A7 in both substates. In simulations of isolated K-turn, the direct 16(O2′)-7(N1) H-bond (X-ray substate CA) is quickly lost and is basically never restored (Figure 4). The contact is, nevertheless, still mediated through long-residency water molecules for most of the time (basically the orange state in Figure 4; see also below). After the initial rearrangement, we observed direct contact only

Figure 3. Stereoviews of the bulge substates observed in MD simulations: (a) structure of the RNA-protein complex with C16 in bold; orange and gray colors represent CA and CG substate, respectively; (b) structure of the isolated box C/D RNA with G17 in bold - red and gray colors represent GA and GC substate, respectively. A7 is in cyan, C16 in orange (or gray), G17 in red (or gray), G19 in green, and characteristic H-bonds in blue. See Table 2 for detail description of all substates.

between C16 and G17. G17 was found in two equally populated positions (Table 2, Figure 3b). GA substate is characterized by simultaneous contacts of 17(N2) with O2 and O2′ atoms of C16, while G17 is stacked below A7 (or below the center of the A7/ G19 base pair). GC substate is described by 17(O4′)-16(O2′) and 17(N2)-17(O2P) contacts and formation of partial stacking interaction between G17 and C16. Sugar pucker of G17 is switched from C2′-endo to C3′-endo and/or C4′-exo. Typical time development of the isolated system is the following: The initial CA geometry converts to GA substate. Then (usually at 10-30 ns) the GA substate is replaced by GC substate, which

Figure 4. Development of the 16(O2′)-7(N1) signature interaction during MD simulations.



TABLE 3: Long Residency Hydration Sites Identified in the K-Turn Area in Substates, as Specified (cf. Tables 2 and 4): Hydration Sites Positioned in the RNA/Protein Interface and Protein Atoms Participating in Hydration Site Formation Are in Italics name

localization

A-minor substate, isolated RNA

A-minor substate, RNA-protein complex

maximal water binding times

H1a H1b H1c

20(N3,O2′), 8(O2,O2′), 15(N2) 15(N2), 8(O2,O2′), 20(O2′) 20(N3), 8(O2′)

A20 to G15 shift (A*) partial vertical A20 shift (B) partial vertical A20 shift (B)

A20 to G15 shift (A*) partial vertical A20 shift (B) partial vertical A20 shift (B)

H2a

7(N1), 15(N2)

vertical A20 shift (B*)

14 ns (RNA), 10 ns (complex) 4 ns 4 ns (alternative K+ binding events up to 7 ns observed) 1 ns

a

bulge substate, isolated RNA

bulge substate, RNA-protein complex during 17(N2)-16(O2′), CG substate (competition with H7a) not substate specific

27 ns (RNA),10 ns (complex)

not substate specific not substate specific not substate specific during 7(N1)-16(O2′), CA substate (competition with H2b) when H7a is not observed

38 50 18 11

H2b

7(N1), 16(O2′,O4′), 20(O4′)

not substate specific

H3

7(N3,O2′), 16(O2)

G17/C16 stacking, GC substate

H4 H5 H6 H7a

18(O1P), PRO(O) 18(O2P), 19(N7), ASN(N) 6(O2′), GLU(OE1/OE2) 17(N2,N3,O2P), ARG(NE)a

H7b

17(O2P), PRO(O)

14 ns (RNA),7 ns (complex) ns ns ns ns

12 ns

In the 15.5 kDa complexes only.

persists until the end of the simulation. Two simulations (CD2 and CD4) remained in the GA substate for the whole simulation, and one system (CD5), after early switch to GC, returned after 30 ns back to GA. Exception from this common behavior was observed in CD8 simulation in the interval of 18-29 ns. At the beginning of this event, G17 moved below G19 while C16 passed under A7. New position of G17 allows more motion which resulted in syn to anti conformational flip of G17. After C16 movement to in-plane position with A7, we observed return G17 to its former position below A7. Nevertheless, G17 did not switch to syn and remained in anti conformation until the end of simulation. Note that syn base orientations are not uncommon in X-ray structures of K-turn bulges, including, for example, G17 in the box C/D, A30 in Kt-U4, A262 in Kt-15, and A1603 in Kt-58. Kt-46 contains U in equivalent position, which is anti. Basically, syn orientation is common for purines in position equivalent to the box C/D G17. In the RNA-protein systems, the U18 base is flipped out into the protein binding pocket and is recognized by the protein. In isolated systems, U18 base fluctuates around its initial position while not forming any contacts with the rest of the molecule. In the X-ray structure, the phosphodiester backbone in the bulge is further stabilized by two H-bond interactions: 18(O2′)-17(O1P) and 17(O2′)-18(O1P/O2P). In RNA-protein simulations, both H-bonds are stable with an interatomic distance of ∼2.8 Å. 17(O2′) preferentially interacts with 18(O1P). These H-bonds are less robust in simulations of free K-turn. While the 18(O2′)-17(O1P) interaction is observed in ∼40% of simulation time, the other H-bond is rarely sampled. Existence of these interactions is not influenced by the bulge substate transitions. In RNA-protein simulations, we also observed transient H-bond contact between A7(O2′) and G17(O6) with population of ∼25%. Note that it has been suggested that this interaction can efficiently form only when two adenosine nucleotides are involved.11 The residue C16 is stacked below G15 in all simulations. C2′-endo sugar puckering of G6, A7, G17, U18, and G19 is conserved in MD simulations except for G17 in GC substate (see above). Hydration of the KinksStructurally Important Water Bridges. We have analyzed long-residency water molecules with binding time on a scale of ∼1 ns and above (Table 3). Different substates adopted by the bulge region are associated with distinct hydration patterns. For the free K-turn, the most

Figure 5. Hydration in the kink region of the isolated K-turn molecule in dependence on G17 position: (a) in GA substate (see Table 2 for details) only hydration site H2b is observed; (b) in GC substate another hydration site H3 exists together with the H2b site. Positions of A7 and G17 are highlighted, and water molecules are shown as red balls.

important hydration site marked as H2b appears near 7(N1) atom with participation of other atoms such as 16(O2′,O4′) and 20(O4′). This site was observed in all simulations of the isolated system (Figure 5). Binding times of individual water molecules are usually up to 10-15 ns with the longest time of 27 ns. This complex hydration pocket is not influenced by changes of the G17 position. Importantly, due to the H2b hydration, the key 16(O2′)-7(N1) interaction has not been fully lost as it has been replaced by prominent and apparently very specific water bridge (H2b basically corresponds to orange color time intervals in Figure 4). Another hydration site H3 is in the cavity formed by 7(N3,O2′), and 16(O2) atoms. This hydration site appears only when G17 is partly stacked with C16 (Figure 5b). The longest binding time for this site was 14 ns. It should be noted that these are extremely long hydration events. Such hydration sites have not been reported (until now) in simulations of uncomplexed RNA molecules, with the exception of the closed catalytic pocket of the hairpin ribozyme.45,46

10586



Figure 6. Substate-specific hydration patterns observed in MD simulations of RNA-protein complexes. Left, numbering; right, stereoview (a) during CG substate (see Table 2 for details) and (b) during CA substate. Position of C16 is highlighted. Water molecules typical for each substate are as green balls, other water molecules are in red.

The RNA-protein complexes contain a larger number of long-residency hydration sites (Figure 6). Except hydration sites observed in the isolated systems (i.e., H2b and H3) are also hydration sites that stabilize the intermolecular complex. Two of them, H4 (near 18(O1P)) and H5 (near 18(O2P) and 19(N7)), represent the hydration sites characterized by the longest binding times which frequently exceed 20 ns with maximum represented by just one water molecule during the whole 50 ns simulation. Another hydration site is formed near 6(O2′) atom (H6). Influence of modest geometrical changes in the bulge is visible in formation of hydration site H7a near 17(N2, N3,O2P). This hydration site appears when the direct C16(O2′)-A7(N1) H-bond is realized (Figure 6b). On the other hand, formation of the H2b site requires disruption of this H-bond. H7b site (formed on the opposite side of 17(O2P) atom) is usually formed in absence of the H7a site. Trapped water molecules at the protein-RNA interface are not surprising.47 Interactions with Ions in the Kink. In the following analysis, we report mainly ion-binding sites localized at and around the bulge. All ion-binding events reported below mean inner shell (direct) cation binding to RNA atoms. The analysis did not show any differences between Na+ and KCl simulations. Thus the results are robust within the framework of the presently available force fields. In protein-RNA complexes, we found one noticeable ionbinding site positioned near 16(N3,O2) with partial participation of 15(O6) (Figure 7). This site is occupied only occasionally even in simulations with excess salt conditions. The ion-binding time was usually around 2-3 ns, with a maximum of 6 ns. Much more significant ion binding occurred in simulations of the isolated systems. We found one ion-binding site characterized by exceptionally long binding times, positioned near 17(O6). It occurs in the GC substate, while it is absent in the GA substate (see Table 2 for substate description) and in the initial stages (i.e, the X-ray structure substate). Once formed, this site is usually occupied by only one tightly bound ion during the presence of the GC substate, with binding times from 14 to 31 ns. The ion samples two equally occupied positions around 17(O6) atom (Figure 7). The first one is near O2 and N3 atoms of C16, while the other one is near 17(N7) and 7(O2′).

Figure 7. Ion binding to K-turn. Black ball represents the position of ion in both isolated and complexed K-turns, while the gray ball represents the ion position observed only in the isolated system. In the case of the isolated K-turn, ion binding is observed exclusively during GC substate (see Table 2), where G17 is stacked below C16 (both residues highlighted) and both positions are equally populated. Water in H3 site is in red, and contacts with the ion are in green. Position of G17 in the RNA-protein complex is in black, while the other residues adopt the same position as in the isolated system.

We found several significant inner shell (direct) monovalent ion-binding sites elsewhere in the simulated structure. Many of them are positioned near N7 atoms of purines. There are two ion-binding sites at and around the GAAA tetraloopsnear 10(N7) and near 14(N7), respectively, which is in agreement with our previous results on GAGA tetraloops.48 Both sites are frequently occupied by ions with the maximal binding time around 13 ns. Other ion-binding sites are observed in the NCstem, predominantly near 23(N7), 22(N7), 5(O4), and 6(N7) atoms. The G22(N7) ion-binding site corresponds to ion binding reported by Curuksu et al.22 for Kt-38. Maximal binding times do not exceed 5 ns. In the simulation CD9, an ion also participates in the A-minor interaction (see below). U3/A23 Base Pair. The X-ray U3/A23 base pair (which is already outside the K-turn) is poorly paired. A23 adopts syn conformation and exposes its Hoogsteen edge to U3 with a single H-bond between 3(O4) and 23(N6). In addition, the X-ray geometry would result in a clash between hydrogens of 3(N3) and 23(N6), unless the amino group is pyramidal.49,50 This base pair is very unstable in simulations, while its instability often propagates to the adjacent C4dG22 base pair in the direction away the K-turn. In several simulations, we changed its geometry to the canonical cis-WC geometry. This geometry is



Figure 9. Appearance of different A-minor geometries in MD simulations: partial A20 to G15 shift A (light blue), A20 to G15 shift A* (dark blue), partial vertical A20 shift B (orange), vertical A20 shift B* (red), and G6/A20 opening I (green). Yellow color corresponds to sampling of the initial geometry.

Figure 8. Different geometries of the A-minor interaction formed between A20/G6 and C8dG15 base pairs, which were observed in our MD simulations. (a) X-ray structure, (b) partial A20 to G15 shift substate A, (c) A20 to G15 shift substate A*, (d) partial vertical A20 shift substate B, (e) vertical A20 shift substate B* (interaction with A7 is shown, thin line), and (f) G6/A20 opening with A-minor I interaction, substate I. H-bond contacts are shown in green.

stable in all simulations. The syn arrangement of A23 could possibly be a result of low resolution of the X-ray structure, which can lead to syn-anti bias in the refined structures, as common for example with ribosomal X-ray structures.51 A-Minor 0 Interaction is Very Dynamical. The box C/D A-minor type 0 interaction is formed between A20 and C8dG15 base pair and consists of trans-SE/SE C8/A20 base pair (Figure 8a). The low-resolution X-ray structure suggests three contacts, namely, 8(O2′)-20(N3), 8(O2′)-20(O2′), and 8(O2)-20(O2′). Ribose O2′ atoms can participate in these H-bonds as either proton donors (interactions with N3, O2, and O2′) or proton acceptors (interactions with N2 and O2′). All of these H-bonds cannot exist simultaneously; for example, if both O2′ atoms donate to N3 and O2, respectively, the O2′-O2′ H-bond is not realized. In simulations of uncomplexed RNA, we observed sampling of the initial A-minor geometry (with H-bond fluctuations and a strong preference for the O2′-O2′ interaction) as well as five

additional substates. We thus altogether classify six substates (Figure 8, Figure 9, Table 4). The alternative substates are more frequently (some of them exclusively) observed in parm99 force field simulations. In summary, the following substates were seen in our simulations. (1) X-ray A-minor 0 geometry with fluctuations of the 2′OH groups (Figure 8a). We observed several variants (differing by H-bond number and type) of this basic A-minor geometry which alternate on a 10-100 ps time scale due to fast reorientations of the hydroxyl groups of residues C8 and A20. Average distances between the corresponding heteroatoms oscillate around 3.0-3.2 Å when the H-bond is established. Except for these contacts an additional contact between 20(O2′) and 15(N2) appeared in simulations. (2) Substate A, partial A20 to G15 shift: (Figure 8b) A20 is slightly shifted toward G15 leading to disruption of the 8(O2′)-20(O2′) contact. The 20(O2′)-15(N2) and 20(O2′)-8(O2) distances are moderately shortened due to this shift. (3) Substate A*, A20 to G15 shift: (Figure 8c) A20 is shifted toward G15 and the A-minor interaction is realized only by 15(N2)-20(O2′) H-bond. Other H-bonds are disrupted forming a cavity between A20 and C8 which is occupied by a water molecule. (4) Substate B, partial vertical A20 shift: (Figure 8d) A20 is slightly shifted upward above the C8dG15 base pair plane, and the A-minor interaction is realized only by 8(O2′)-20(O2′) H-bond. Increased distances between G15 base and A20 ribose, and between A20 base and C8 ribose, allow extensive solvation of the bases. (5) Substate B*, vertical A20 shift: (Figure 8e) A20 is shifted upward above the C8dG15 base pair plane. The A-minor interaction is disrupted and all H-bonds disappeared. However, C8dG15 base pair, which is temporarily positioned near A7, forms another A-minor interaction (type II) with the 7(N3)-8(O2′) H-bond. This geometry represents a rather large perturbation of the initial state; however, its formation is still reversible. (6) Substate I, G6/A20 opening and A-minor I formation: (Figure 8f) formation of full A-minor I interaction between A20 and C8dG15 base pair, with additional cis-

10588



TABLE 4: Substates of the A-Minor Interaction Observed in Our MD Simulations direction of A20 movement

substate name

description

H-bond contacts

X-ray

experimental geometrya

A A* B B* I

partial A20 to G15 shift A20 to G15 shift partial vertical A20 shift vertical A20 shift G6/A20 opening

8(O2′)-20(N3), 8(O2′)-20(O2′), 8(O2)-20(O2′) 15(N2)-20(O2′), 20(O2′)-8(O2) 15(N2)-20(O2′) 8(O2′)-20(O2′) 7(N3)-8(O2′) 15(N2)-20(N3), 15(O2′)-20(O2′), 8(O2)-20(C2), 20(N1)-8(O2′)

a

horizontal horizontal vertical vertical horizontal

specific water insertion (where observed)

A-minor type

none

type 0

none A-minor A-minor A-minor G/A pair in the NC-stem

type type type type type

0 0 0 II with A7 I

Substantial dynamics of hydroxyl hydrogens observed; see the text.

Figure 10. Hydration of A-minor interaction in dependence on adopted geometry. (a) A20 to G15 shift (A*) with hydration site H1a, (b) partial vertical A20 shift (B) with hydration sites H1b and H1c, and (c) vertical A20 shift (B*) with hydration site H2a. H1c site (gray ball) can be also occupied by ions. Other hydration sites are present as red balls. Residues G6, A20, C8, and G15 in panels a and b and residues G19, A7, C8, and G15 in panel c, respectively, are highlighted.

SE/SE A20/G15 base pair paired via 15(N2)-20(N3) and 15(O2′)-20(O2′) H-bonds. However, this is accompanied with a loss of H-bond between 6(N3) and 20(N6), that is, destabilization of the G/A base pair within the NCstem. This means that the box C/D K-turn can form the full A-minor I interaction, but its formation is achieved only upon perturbation of the intrastem G/A base pair. In the RNA-protein complexes, the A-minor interaction is more stable and the systems predominantly sample the initial A-minor geometry. Two preferentially populated substates are then characterized by combination of two H-bonds: O2′-N3, O2-O2′, and O2′-N3, O2′-O2′. However, we also observed additional A-minor substates A and B in approximately 20% of simulation time (Figure 9). Hydration of the A-Minor Interaction. The alternative A-minor substates (with exception of substate A) include specific insertions of water molecules which can be summarized as follows. Substate A*: water molecule is positioned in the cavity between A20 and C8, which is formed by 20(N3,O2′), 8(O2,O2′), and 15(N2) atoms (hydration site H1a) (Figure 10a). The bound water molecule shows no exchange events during this geometrical substate leading to binding times of 14 and 10 ns for CD3 and CD_pL7_1 simulations, respectively. Substate B: one water molecule is shared by 15(N2), 8(O2,O2′) and 20(O2′) atoms (H1b site) (Figure 10b). Another hydration site (H1c) is formed between 20(N3) and 8(O2′) atoms

Figure 11. Development of the interhelical angle in MD simulations.

as the direct contact between them is disrupted. The longest water binding time was 4 ns. The H1c site can be occupied also by ions. In CD9 simulation. one K+ ion was positioned here for 7 ns. Substate B*: hydration site H2a with longest water binding times ∼1 ns mediates the contact between 7(N1) and 15(N2) atoms (Figure 10c). Substate I: common (binding times less than 0.5 ns) water bridge between 6(N3) and 20(N6) atoms in the distorted G/A base pair in the NC-stem. Elbow-like Dynamics. K-turns are known for their anisotropic elbow-like thermal fluctuations.14,17–19,22,23 We have monitored the box C/D K-turn elbow-like fluctuations using interhelical angle between the C-stem and the NC-stem, defined by three centers of mass of selected residues in the C-stem (residues 8, 9, 14, and 15), the bulge (residues 7, 16, and 19), and the NC-stem (residues 5, 6, 20, and 21). Due to geometrical changes observed at both edges of the K-turn stems (in GAAA tetraloop, see below, and the U/A base pair, see above), only residues relatively close to the Kink were selected for this calculation. If the K-turn is bound in the protein complex, we observe fluctuations of the interhelical angle in the range of 92-118° with the average value of 106° (Figure 11). The observed range of ∼25° is similar to the range of fluctuations observed in simulations of isolated Kt-38. This flexibility range is significant and would provide large flexibility if the fluctuations propagate further away from the Kink area.23 Isolated K-turn exhibits larger fluctuations of the interhelical angle of 94-129°. It frequently adopts values higher than 120°


Figure 12. Normalized histograms of interhelical angles for MD simulations of the isolated box C/D RNA.

Figure 13. Representative structures of isolated K-turn characterized by the interhelical angle of ∼100° (green), ∼110° (orange), and ∼120° (red). The A-minor interaction (line representation) adopts I (G6/A20 opening), X-ray, and B (partial vertical A20 shift) substates, respectively. The structures are superimposed over the C-stem. Terminal base pairs of the NC-stem are shown for better plasticity.

not sampled in the RNA-protein complexes (Figure 11). Figure 12 demonstrates differences between parmbsc0 and parm99 simulations of the isolated box C/D RNA. Peaks (representing maximal frequencies of interhelical angles) are found in three regions: approximately around 103, 113, and 123°. In the case of parmbsc0 simulations, there is only one peak at ∼113°, which reflects reduced flexibility of the A-minor interaction. In parm99 simulations, peaks are found also around 123 and 103°. The low value (∼100°) represents closing of the K-turn hinge and is mostly associated with A, A*, and I substates. By contrast, B and B* substates generally lead to larger angle values around 120° accompanied by opening of the K-turn hinge. Therefore, the A-minor interaction regulates the optimal angle between the stems. Influence of the interhelical angle changes on the global K-turn shape is demonstrated in Figure 13, which shows three representative structures characterized by interhelical angles of (approximately) 100, 110, and 120°. Interaction of K-Turn with Protein. The RNA-protein interaction is realized mainly by the bulge and adjacent part of the NC-stem represented by residues G6 and G19 (Figure 14). Other contacts are formed between the central part of the NCstem (phosphates of C4 and U5) and side chains of the R2helix. The C-stem does not interact with the protein. The most important contacts involve U18 (O1P, O2P, N3, and O4 atoms), which interacts mainly with backbone N and O atoms. Other residues which participate in protein-RNA H-bond network are G17 (N7, O2′), G19 (N1, O6), and G6 (O2′). All of these contacts are permanent with no significant fluctuations. On the other hand, the contacts formed by C4 and U5 phosphates fluctuate (with different behavior in each simulation). The contact of protein with the NC-stem influences the flexibility of the whole K-turn as it fixes the position of the NC-stem and the bulge and restricts the dynamics.


Figure 14. Stereoview of the box C/D RNA and L7ae protein complex. Signature amino acids in the R5-β4 loop are highlighted: valine (L7ae protein) is in orange, while arginine (single amino acid taken from the 15.5 kDa protein and overlaid over the L7ae protein) is colored by atom types. H-bonds between arginine side chain and G17 phosphate group are in red. RNA residues which participate in RNA-protein contact are in blue (residues 4-6, 17-19). Amino acid side chains which are in direct contact with the NC-stem of the K-turn molecule are shown.

Our simulations considered two proteinssL7ae and 15.5 kDa. Sequence and folding of both proteins are very similar. Binding surface of the protein is formed approximately by52,5311-12 amino acids, and half of them are identical in both proteins. The contacts are realized either by backbone N and O atoms (mainly by amino acids of different type in both proteins) or by side chain atoms which predominantly belong to the conserved amino acids. We identified one significant difference between L7ae and 15.5 kDa binding. In the 15.5 kDa complex, the phosphate group in the bulge (namely 17(O2P)) forms stable contacts with imino (NE) and amino (NH2) nitrogens of arginine (R5-β4 loop) (Figure 14). Further, arginine (NH2, NH1 atoms) also partly interacts with 16(O2P), and its position in our simulations is very similar to position observed in the original crystal structure of U4 snRNA-15.5 kDa complex.6 In the case of the pL7ae complex, the same site is occupied by valine and its shorter side chain is not sufficient for a contact with RNA. This valine belongs to signature amino acids that enable binding of L7ae protein to the K-loop motif,53–56 while 15.5 kDa protein does not recognize it.2,12,30,52 The K-loop motif is structurally similar to the K-turn, but the C-stem is replaced with a short terminal loop; that is, the K-turn is incomplete (it effectively is one-half of the K-turn).52,53 Note that in some mainly older papers the K-loop was called K-turn, which may cause some confusion. K-loop is much less frequent than K-turn. We suggest that the 17(O2P)-ARG(NE,NH2) H-bonds in 15.5 kDa complexesmayimprovestabilityoftheK-turnsignature7(N1)-16(O2′) interaction, as indicated by the simulations (see above), while valine allows more variable RNA recognition. Force Field Description of the GAAA Tetraloop Is Not Perfect. The C-stem of the box C/D RNA is capped by a GAAA tetraloop (Figure 15a), which belongs to the common group of GNRA tetraloops. Functional RNA tetraloops are known to be rather stiff with highly conserved 3D geometry, at least when the available structural experiments are considered.57,58 In the GAAA tetraloop, the first and fourth bases form trans-SE/H G10/A13 base pair, which is stabilized by a bifurcated H-bond interconnecting G(N2) with A(N7) and A(O2P). Three consecutive adenines A11-A13 form a triple purine stack. The tetraloop is also stabilized by an H-bond between 12(N7) and 10(O2′). The simulation behavior of the GAAA tetraloop is very similar to behavior of the GAGA tetraloop of the sarcin ricin domain (SRD) reported in earlier 25 ns simulations.48 Generally,

10590



Figure 15. Geometries of the GAAA tetraloop: (a) initial geometry from the crystal structure, (b) geometry with vertically oriented adenine in the tetraloop apex, and (c) example of disrupted tetraloop geometry.

the initial tetraloop geometry is rather flexible in all MD simulations. Together with the initial geometry, we observed two additional substates which are characterized by different positions of adenine A11 at the tetraloop apex. In the first geometry, stacking between A11 and A12 is disrupted and A11 is vertically oriented with respect to A12 (Figure 15b). The other geometry is characterized by rotation of A11 around the glycosidic bond from anti to syn orientation and by restoration of A11-A12 stacking. Almost in each simulation we can observe all three substates which alternate several times during the simulation. When considering all simulations, we find that the three substates are equally populated while their populations are not sensitive to the force field choice and salt conditions. The two alternative substates extensively populated in the simulations deviate from the native topology of GNRA tetraloops,57,58 probably indicating imperfectness of the force field description. In addition, in three simulations (CD3, CD_pL7_3, CD_p15_3), we observe complete degradation of the tetraloop (Figure 15c). It appears at 5 ns for isolated RNA and at ∼25 ns for protein-RNA complexes. The G10/A13 base pair is disrupted (except for one simulation) and all adenines (as G10 remains stacked on the adjacent GdC pair) sample new positions. This is consistent also with our extensive unpublished simulations of the SRD, indicating progressive loss of the GAGA tetraloops, albeit all other segments of the SRD are entirely stable. We also analyzed R/γ backbone dynamics of G14, which was also reported for the GAGA tetraloop.48 The X-ray structure of the box C/D RNA contains two independent structures with two combinations of G14 R/γ: 300°/70° (canonical substate) and 130°/180° (γ-trans substate). In case of parmbsc0 simulations of isolated systems, the G14 R/γ angles adopt canonical values, and short flips to other values are very rare (if any). On the other hand, γ-trans is frequently observed in parm99 simulations (∼60% of the simulation time), and the R/γ angles adopt values of 130-230°/180° (i.e., R adopts several characteristic values). This difference in force field behavior is consistent with the γ-trans penalty of the parmbsc0 force field, which was introduced to stabilize B-DNA simulations.34 In case of RNA-protein complexes, the force field difference is not so sharp, and in several parm99 simulations, we did not see any G14 γ-trans flip. This result may be affected by the limited sampling. It is to be noted, however, that the G14 R/γ dynamics is likely entirely unrelated with the overall tetraloop destabilization described above, and all populated G14 R/γ substates are in fact compatible with correct overall tetraloop structure. Discussion and Conclusions We have carried out a series (∼1 µs in total) of MD simulations of the K-turn found in the box C/D RNA and its complexes with L7ae and 15.5 kDa proteins. This study aimed

to investigate structural dynamics of K-turn with the A-minor type 0 tertiary interaction and to understand influence of the protein binding on the structural dynamics of the K-turn. We compared parm99 and parmbsc0 variants of the Cornell et al. force field while testing net-neutralizing Na+ and excess salt KCl conditions. Elbow-like Dynamics and A-Minor 0 Fluctuations. Simulations revealed substantial elbow-like dynamics of the simulated molecule associated with substates of the A-minor interaction between the C- and NC-stems. Besides the X-ray A-minor 0 geometry, we observed five alternative substates, which are seen in ∼60% of the simulation time for parm99 simulations of the isolated box C/D RNA. The substates are characterized by shifts of A20 with respect to the C8dG15 base pair (Figure 8): by horizontal A20 to G15 shift (substates A and A*), by vertical A20 shift (B and B*), and by opening of G6/A20 base pair in the NC-stem (I). Shift of A20 away from C8dG15 leads to the A-minor opening and formation of hydration cavities in the minor groove side of the C8dG15 base pair (Figure 10). These hydration sites are almost permanently occupied by longresidency water molecules, which typically are not exchanged during the substate lifetime. Parmbsc0 force field reduces the fluctuations (see below). Fluctuations in the A-minor interaction propagate to the attached arms and regulate the interhelical angle between Cand NC-stems. B and B* substates are associated with the opening of the K-turn (increase of the average interstem angle to ∼120°), while substates with horizontal shifts (A, A*, and I) lead to closing of the K-turn accompanied with decrease of the interhelical angle to ∼100°. Thus the K-turn fluctuates in a wide ∼35° range (taking all snapshots into account) of the interstem angles and is capable of acting as a molecular elbow similar to the K-turns 38 and 42.14,18 The local mechanism to achieve the flexibility is, however, different from the waterinserted A-minor I dynamics of K-turns 38 and 42. A-Minor 0 versus A-Minor I. The box C/D K-turn is capable (substate I) to temporarily reach the A-minor I interaction. The A-minor I interaction is established only at the expense of perturbation of the G/A base pair in the NC-stem that donates the adenine sugar edge for the A-minor interaction. This simulation result is perfectly consistent with experimental structures since the G/A base pair is also typically perturbed in X-ray structures of ribosomal K-turns with A-minor I interactions, such as Kt-7, Kt-38, and Kt-42. The structural and simulation data thus indicate struggle between different interactions of K-turns to be simultaneously optimized. All K-turns share common signature K-turn interactions, irrespective of whether they have A-minor I or A-minor 0 interaction.4,9 However, additional “nonsignature” nucleotides and interactions can provide fine-tuning of physicochemical properties of K-turns.26 Obviously, it is likely that the molecular

MD Simulations of the Box C/D RNA context determines whether a given K-turn is observed with A-minor I or A-minor 0 interaction. On the other side, our K-turn simulations accumulated so far indicate that K-turns do not seem to readily undergo transitions between A-minor I and A-minor 0 conformations. This indicates possible sequencedependent propensity to adopt either A-minor I or A-minor 0 interaction. Work is in progress to analyze if such propensity indeed exists and which nonsignature sequence and structural factors might contribute to it. Dynamics of the C16(O2′)-A7(N1) H-Bond. This H-bond between the first (5′-end) bulge base and the NC-stem is considered as the single most characteristic interaction of the 3D signature of K-turns. This interaction was rather stable in preceding simulations of K-turns 38 and 42 (population of the direct C16(O2′)-A7(N1) contact ∼50-80%, alternating with water-mediated substate). However, it is swiftly replaced by water-mediated contact in present simulations of isolated box C/D K-turn and is basically never restored as direct H-bond (Figure 4). We then identify two substates (both characterized by water-mediated C16(O2′)-A7(N1) interaction) where G17 is either stacked below A7 (GA substate) or below C16 (GC substate). Usually, the system first adopts the GA substate and then (after 10-30 ns) it moves to the GC substate, which persists until the end of the simulation. This behavior is not correlated with fluctuations of the A-minor interaction. The instability of the C16(O2′)-A7(N1) signature interaction is of concern and points to possible force field imbalance. Despite the low resolution, this interaction appears to be a direct H-bond in all K-turn X-ray structures. We suggest, however, that it is premature to make any such conclusion since the K-turn is known to ultimately unfold in isolation,11,12,24,25,27,28 and thus the simulations may just reflect early signs of its genuine instability. Nevertheless, it is fair to admit that force field imbalance is to be suspected, too. Despite all of the local fluctuations, none of the simulated K-turns was disrupted. The results indicate that the K-turns 38 and 42 are intrinsically more stable (at least in simulations) than the box C/D K-turn. It is important to point out that, even if the force field is not absolutely perfect, it should still be capable of correctly identifying relative difference between properties of distinct K-turns. We suggest the most likely simulations reflect subtle but real differences between the box C/D “A-minor 0” K-turn on one side and “A-minor I” K-turns 42 and 38. K-TurnsProtein Complexes. The protein interacts with the bulge and a part of the NC-stem. The structure and dynamics of the Kink are substantially influenced by the protein binding. In the RNA-protein complexes, the Kink remained almost in the initial geometry with G17 stacked below A7. In our MD simulations, we identified two different positions of C16 (Table 2, Figure 3a), which forms contacts either with A7 (CA substate) or with G17 (CG substate). The first substate is consistent with the X-ray structure with direct 16(O2′)-7(N1) signature H-bond. In the second substate, this interaction is mediated through a tightly bound water molecule. The native CA substate has a population of ∼80 and ∼40% in 15.5 kDa and L7ae complex simulations, respectively, while it is observed only in ∼7% of the time in the isolated box C/D RNA simulations (essentially at the beginning of the trajectories, Figure 4). The stabilizing role of the protein binding in this particular RNA-protein complex is clearly seen. Besides the different behavior of the signature interaction, we did not find any substantial differences between simulations with the L7ae and 15.5 kDa proteins. Both L7ae and 15.5 kDa proteins have very similar tertiary structures, and their contacts with RNA are similar. However,

J. Phys. Chem. B, Vol. 114, No. 32, 2010 10591 the 15.5 kDa complex contains one tertiary interaction (between G17 phosphate and arginine side chain of the R5-β4 loop), which is missing in the L7ae complex due to arginine to valine substitution. The simulations indicate that this contact may help to stabilize the signature C16(O2′)-A7(N1) interaction. Thus, we suggest that the 15.5 kDa protein shows better optimization of the protein-K-turn interaction than L7ae. The L7ae valine is important for interaction with RNA K-loop motif,52 which is not recognized by 15.5 kDa protein.2,30 The dynamics of the RNA molecule in the complex is mainly influenced by interaction of the NC-stem with side chains of the R2-helix. It fixes the NC-stem and does not allow its motion in the full range. Mutual RNA-protein contacts are then reinforced by strong hydration sites. The RNA-protein binding fixes the position of G17 (which is the second most mobile base of the isolated K-turn after U18) and allows only moderate C16 fluctuations. The average value of the interhelical angle remains around ∼106° as observed in the crystal structure. However, even the RNA-protein structure is able to adopt shifted A-minor geometry as observed in CD_pL7_1 simulation where we identified the A20 to G15 shift (A* substate) accompanied with formation of the characteristic hydration site and decrease of the interhelical angle. Substate A with partial A20 to G15 shift is frequently populated in RNA-protein simulations (Figure 9). Thus, albeit the protein binding visibly limits the elbowlike flexibility of the K-turn, it does not fully eliminate it. The interstem angle samples a range of instantaneous values of ∼25°. Long-Residency Hydration and Ion Binding. Various substates of K-turn and their complexes with proteins involve specific long-residency hydration sites and, to a lesser extent, ion binding. Details are provided in the Results section. Several of the water molecules are found in intricate binding pockets where a single water molecule can be present for many nanoseconds. This confirms that complex RNA tertiary topologies may involve major long-residency hydration and cationbinding sites which represent an inherent part of their local conformational variability. It is, however, fair to say that in this particular case the most salient K-turn hydration and ion-binding sites form when the K-turn deviates from the starting structure. As we noted above, we cannot rule out that details of the dynamics are affected by force field approximations. Definitely the most interesting hydration site in our simulations is the one that replaces the signature C16(O2′)-A7(N1) H-bond by a water bridge in some substates, so the interaction is not fully lost and retains the same sequence specificity as the direct interaction. We suggest that the water-bridged substate can play a role in K-turn folding/unfolding path and may be sometimes populated in real structures. However, the waterbridged state is so far not supported by direct evidence from available structural data. GAAA Tetraloop Is Not Properly Described by the Force Field. The GAAA tetraloop, albeit included in our simulations, is not part of the K-turn structure. As described in the Results section, we suggest that the force field is not accurate enough to describe its structure and we even evidenced several cases of its entire disruption. Single-stranded hairpin loops of nucleic acids are difficult systems for balanced force field description, and the presently available force fields appear to fail to correctly predict their native structures, as already indicated in the literature.59–61 Work is in progress to provide more extensive analysis of tetraloop behavior in different contexts. Description of apical single-stranded hairpin loops is a painful problem in contemporary simulations of nucleic acids, and simulations of systems with such tetraloops should be taken with great care.

10592



Choice of the Force Field. We have utilized two variants of the Cornell et al. force field, which both are assumed to be appropriate for RNA simulations. The parm99 and parmbsc0 force fields appear to provide equivalent (within the limits of the sampling achieved in our work) description of most parts of the present system, in agreement with other recent tests.62,63 The parmbsc0 obviously penalizes the γ-trans backbone substates. This, however, usually does not have a substantial effect on the RNA behavior, and we cannot make any recommendation which of the two force fields is closer to reality. We specifically note that the force field choice does not affect the 16(O2′)-7(N1) signature H-bond dynamics and the GAAA tetraloop behavior, the two cases where we suspect force field imperfectness. The simulations, however, reveal force field sensitivity of the fluctuations of the A-minor interaction. The K-turn structure as described by the parmbsc0 force field is stiffer and shows reduced dynamics of the A-minor interaction (alternative substates were observed for only 15% of simulation time, compared with 60% with parm99). Since the A-minor dynamics affects the overall K-turn flexibility, the choice of the force field has an effect on the elbow-like dynamics of the K-turn. With parmbsc0, values of the interhelical angle below ∼105° and above ∼120° are rarely observed. It is important to underline that the A-minor dynamics samples a range of substates which are iso-energy (they in fact may be considered as forming one flat, nonharmonic free-energy region). Therefore, very subtle modulation of the free-energy surface due to force field modification may have substantial visible effect on the actual sampling (population) of the A-minor geometries. We do not have any viable reference data to suggest which force field is closer to reality. In general, the results confirm that parmbsc0 and parm99 force fields have similar performance for RNA. We did not identify any noticeable differences between simulations with net-neutralizing Na+ environment and with ∼0.2 M excess salt KCl. Concluding Remarks The box C/D K-turn with the A-minor 0 interaction has similar overall elbow-like flexibility as K-turns with A-minor I interactions (such as K-turns 38 and 42). However, the flexibility is achieved via different local fluctuations in the K-turn. The A-minor tertiary interactions are capable to adopt series of subtly different substates that allow fine-tuning of the molecular structures and dynamics. In other words, the A-minor interactions have flat (nonharmonic) free-energy surface. Such interactions which can adopt several substates are important for flexibility and structural rearrangements in large RNAs.64,65 Specific hydration plays a key role in achieving unique flexibility of the A-minor interactions.14,19,22,66 We suggest that K-turns could be separated in several subgroups with subtle differences in their molecular interactions; that is, there appears to be visible difference in the simulation behavior between K-turns 38 and 42 with A-minor I on one side, and the box C/D RNA K-turn with A-minor 0 on the other side. While it is likely that interactions with surrounding elements affect the balance between A-minor I and A-minor 0 K-turn arrangements, fine-tuning of the sequences of the individual K-turns that does not perturb the signature (consensus) K-turn interactions may also be significant. Work is in progress to analyze relation between K-turn sequences and their behavior in molecular simulations. Protein binding substantially attenuates the flexibility of the box C/D K-turn and leads to visible stabilization of signature interactions of this specific K-turn. Nevertheless, the K-turn still

retains a non-negligible fraction of its elbow-like flexibility and might contribute to large-scale dynamical movements. The valine to arginine substitution between L7ae and 15.5 kDa protein is stabilizing the complex, at least as reflected by the dynamics. Our study confirms that MD simulation technique, although based on significant approximations, is a useful tool to provide insights into available experimental data about structural dynamics of RNA. More examples can be found in recent literature.14,17–23,45–48,60,62,63,66–91 Although the simulated structures are not always perfect, the computer model can properly reflect many features of the studied systems and even provide ideas for future experiments. Acknowledgment. This work was supported by the Ministry of Education of the Czech Republic (Grant Nos. AVOZ50040507, AVOZ50040702, MSM0021622413, and LC06030), by the Grant Agency of the Academy of Sciences of the Czech Republic (Grant Nos. KJB400040901 and IAA400040802), and by the Grant Agency of the Czech Republic (Grant No. 203/ 09/1476). References and Notes (1) Dennis, P. P.; Omer, A. Curr. Opin. Microbiol. 2005, 8, 685–694. (2) Reichow, S. L.; Hamma, T.; Ferre-D’Amare, A. R.; Varani, G. Nucleic Acids Res. 2007, 35, 1452–1464. (3) KissLaszlo, Z.; Henry, Y.; Bachellerie, J. P.; CaizerguesFerrer, M.; Kiss, T. Cell 1996, 85, 1077–1088. (4) Klein, D. J.; Schmeing, T. M.; Moore, P. B.; Steitz, T. A. EMBO J. 2001, 20, 4214–4221. (5) Kuhn, J. F.; Tran, E. J.; Maxwell, E. S. Nucleic Acids Res. 2002, 30, 931–941. (6) Vidovic, I.; Nottrott, S.; Hartmuth, K.; Luhrmann, R.; Ficner, R. Mol. Cell 2000, 6, 1331–1342. (7) Watkins, N. J.; Dickmanns, A.; Luhrmann, R. Mol. Cell. Biol. 2002, 22, 8342–8352. (8) Watkins, N. J.; Segault, V.; Charpentier, B.; Nottrott, S.; Fabrizio, P.; Bachi, A.; Wilm, M.; Rosbash, M.; Branlant, C.; Luhrmann, R. Cell 2000, 103, 457–466. (9) Lescoute, A.; Leontis, N. B.; Massire, C.; Westhof, E. Nucleic Acids Res. 2005, 33, 2395–2409. (10) Leontis, N. B.; Stombaugh, J.; Westhof, E. Nucleic Acids Res. 2002, 30, 3497–3531. (11) Liu, J.; Lilley, D. M. J. RNA 2007, 13, 200–210. (12) Szewczak, L. B. W.; Gabrielsen, J. S.; DeGregorio, S. J.; Strobel, S. A.; Steitz, J. A. RNA 2005, 11, 1407–1419. (13) Nissen, P.; Ippolito, J. A.; Ban, N.; Moore, P. B.; Steitz, T. A. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 4899–4903. (14) Razga, F.; Koca, J.; Sponer, J.; Leontis, N. B. Biophys. J. 2005, 88, 3466–3485. (15) Lescoute, A.; Westhof, E. Biochimie 2006, 88, 993–999. (16) Moore, T.; Zhang, Y. M.; Fenley, M. O.; Li, H. Structure 2004, 12, 807–818. (17) Razga, F.; Koca, J.; Mokdad, A.; Sponer, J. Nucleic Acids Res. 2007, 35, 4007–4017. (18) Razga, F.; Spackova, N.; Reblova, K.; Koca, J.; Leontis, N. B.; Sponer, J. J. Biomol. Struct. Dyn. 2004, 22, 183–193. (19) Razga, F.; Zacharias, M.; Reblova, K.; Koca, J.; Sponer, J. Structure 2006, 14, 825–835. (20) Cojocaru, V.; Klement, R.; Jovin, T. M. Nucleic Acids Res. 2005, 33, 3435–3446. (21) Cojocaru, V.; Nottrott, S.; Klement, R.; Jovin, T. M. RNA 2005, 11, 197–209. (22) Curuksu, J.; Sponer, J.; Zacharias, M. Biophys. J. 2009, 97, 2004– 2013. (23) Reblova, K.; Razga, F.; Li, W.; Gao, H.; Frank, J.; Sponer, J. Nucleic Acids Res. 2010, 38, 1325–1344. (24) Goody, T. A.; Melcher, S. E.; Norman, D. G.; Lilley, D. M. J. RNA 2004, 10, 254–264. (25) Turner, B.; Lilley, D. M. J. J. Mol. Biol. 2008, 381, 431–442. (26) Schroeder, K. T.; Lilley, D. M. J. Nucleic Acids Res. 2009, 37, 7281–7289. (27) Matsumura, S.; Ikawa, Y.; Inoue, T. Nucleic Acids Res. 2003, 31, 5544–5551.

MD Simulations of the Box C/D RNA (28) Wozniak, A. K.; Nottrott, S.; Kuhn-Holsken, E.; Schroder, G. F.; Grubmuller, H.; Luhrmann, R.; Seidel, C. A. M.; Oesterhelt, F. RNA 2005, 11, 1545–1554. (29) Falb, M.; Amata, I.; Gabel, F.; Simon, B.; Carlomagno, T. Nucleic Acids Res. 2010, doi: 10.1093/nar/gkq380. (30) Szewczak, L. B. W.; DeGregorio, S. J.; Strobel, S. A.; Steitz, J. A. Chem. Biol. 2002, 9, 1095–1107. (31) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179–5197. (32) Wang, J. M.; Cieplak, P.; Kollman, P. A. J. Comput. Chem. 2000, 21, 1049–1074. (33) Cheatham, T. E.; Cieplak, P.; Kollman, P. A. J. Biomol. Struct. Dyn. 1999, 16, 845–862. (34) Perez, A.; Marchan, I.; Svozil, D.; Sponer, J.; Cheatham, T. E.; Laughton, C. A.; Orozco, M. Biophys. J. 2007, 92, 3817–3829. (35) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Merz, K. M.; Wang, B.; Pearlman, D. A.; Crowley, M.; Brozell, S.; Tsui, V.; Gohlke, H.; Mongan, J.; Hornak, V.; Cui, G.; Beroza, P.; Schafmeister, C.; Caldwell, J. W.; Ross, W. S.; Kollman, P. A. AMBER 8; University of California: San Francisco, 2004. (36) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Merz, K. M.; Pearlman, D. A.; Crowley, M.; Walker, R. C.; Zhang, W.; Wang, B.; Hayik, S.; Roitberg, A.; Seabra, G.; Wong, K. F.; Paesani, F.; Wu, X.; Brozell, S.; Tsui, V.; Gohlke, H.; Yang, L.; Tan, C.; Mongan, J.; Hornak, V.; Cui, G.; Beroza, P.; Mathews, D. H.; Schafmeister, C.; Ross, W. S.; Kollman, P. A. AMBER 9; University of California: San Francisco, 2006. (37) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Proteins 2006, 65, 712–725. (38) Dang, L. X. J. Am. Chem. Soc. 1995, 117, 6954–6960. (39) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926–935. (40) Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089– 10092. (41) Berendsen, H. J. C.; Postma, J. P. M.; Vangunsteren, W. F.; Dinola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684–3690. (42) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977, 23, 327–341. (43) Harvey, S. C.; Tan, R. K. Z.; Cheatham, T. E. J. Comput. Chem. 1998, 19, 726–740. (44) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 33–38. (45) Rhodes, M. M.; Reblova, K.; Sponer, J.; Walter, N. G. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 13380–13385. (46) Ditzler, M. A.; Sponer, J.; Walter, N. G. RNA 2009, 15, 560–575. (47) Reblova, K.; Spackova, N.; Koca, J.; Leontis, N. B.; Sponer, J. Biophys. J. 2004, 87, 3397–3412. (48) Spackova, N.; Sponer, J. Nucleic Acids Res. 2006, 34, 697–708. (49) Luisi, B.; Orozco, M.; Sponer, J.; Luque, F. J.; Shakked, Z. J. Mol. Biol. 1998, 279, 1123–1136. (50) Sponer, J.; Mokdad, A.; Sponer, J. E.; Spackova, N.; Leszczynski, J.; Leontis, N. B. J. Mol. Biol. 2003, 330, 967–978. (51) Stombaugh, J.; Zirbel, C. L.; Westhof, E.; Leontis, N. B. Nucleic Acids Res. 2009, 37, 2294–2312. (52) Gagnon, K. T.; Zhang, X. X.; Qu, G. S.; Biswas, S.; Suryadi, J.; Brown, B. A.; Maxwell, E. S. RNA 2010, 16, 79–90. (53) Nolivos, S.; Carpousis, A. J.; Clouet-d’Orval, B. Nucleic Acids Res. 2005, 33, 6507–6514. (54) Duan, J. Q.; Li, L.; Lu, J.; Wang, W.; Ye, K. Q. Mol. Cell 2009, 34, 427–439. (55) Hamma, T.; Ferre-D’Amare, A. R. Structure 2004, 12, 893–903. (56) Li, L.; Ye, K. Q. Nature 2006, 443, 302–307. (57) Hsiao, C.; Mohan, S.; Hershkovitz, E.; Tannenbaum, A.; Williams, L. D. Nucleic Acids Res. 2006, 34, 1481–1491.

J. Phys. Chem. B, Vol. 114, No. 32, 2010 10593 (58) Sarver, M.; Zirbel, C. L.; Stombaugh, J.; Mokdad, A.; Leontis, N. B. J. Math. Biol. 2008, 56, 215–252. (59) Fadrna, E.; Spackova, N.; Sarzynska, J.; Koca, J.; Orozco, M.; Cheatham, T. E.; Kulinski, T.; Sponer, J. J. Chem. Theory Comput. 2009, 5, 2514–2530. (60) Ditzler, M. A.; Otyepka, M.; Sponer, J.; Walter, N. G. Acc. Chem. Res. 2010, 43, 40–47. (61) Fadrna, E.; Spackova, N.; Stefl, R.; Koca, J.; Cheatham, T. E.; Sponer, J. Biophys. J. 2004, 87, 227–242. (62) Besseova, I.; Otyepka, M.; Reblova, K.; Sponer, J. Phys. Chem. Chem. Phys. 2009, 11, 10701–10711. (63) Reblova, K.; Strelcova, Z.; Kulhanek, P.; Besseova, I.; Mathews, D. H.; Van Nostrand, K.; Yildirim, I.; Turner, D. H.; Sponer, J. J. Chem. Theory Comput. 2010, 6, 910–929. (64) Mokdad, A.; Krasovska, M. V.; Sponer, J.; Leontis, N. B. Nucleic Acids Res. 2006, 34, 1326–1341. (65) Noller, H. F. Science 2005, 309, 1508–1514. (66) Krasovska, M. V.; Sefcikova, J.; Reblova, K.; Schneider, B.; Walter, N. G.; Sponer, J. Biophys. J. 2006, 91, 626–638. (67) Auffinger, P.; Hashem, Y. Curr. Opin. Struct. Biol. 2007, 17, 325– 333. (68) Freedman, H.; Huynh, L. P.; Le, L.; Cheatham, T. E.; Tuszynski, J. A.; Truong, T. N. J. Phys. Chem. B 2010, 114, 2227–2237. (69) Hall, K. B. Curr. Opin. Chem. Biol. 2008, 12, 612–618. (70) Hashem, Y.; Auffinger, P. Methods 2009, 47, 187–197. (71) Huang, W.; Kim, J.; Jha, S.; Aboul-Ela, F. Nucleic Acids Res. 2009, 37, 6528–6539. (72) Cheatham, T. E. Curr. Opin. Struct. Biol. 2004, 14, 360–367. (73) Li, W.; Sengupta, J.; Rath, B. K.; Frank, J. RNA 2006, 12, 1240– 1253. (74) MacKerell, A. D.; Nilsson, L. Curr. Opin. Struct. Biol. 2008, 18, 194–199. (75) Orozco, M.; Noy, A.; Perez, A. Curr. Opin. Struct. Biol. 2008, 18, 185–193. (76) Romanowska, J.; Setny, P.; Trylska, J. J. Phys. Chem. B 2008, 112, 15227–15243. (77) Sanbonmatsu, K. Y.; Tung, C. S. J. Struct. Biol. 2007, 157, 470– 480. (78) Sharma, M.; Bulusu, G.; Mitra, A. RNA 2009, 15, 1673–1692. (79) Villa, A.; Wohnert, J.; Stock, G. Nucleic Acids Res. 2009, 37, 4774– 4786. (80) Reblova, K.; Fadrna, E.; Sarzynska, J.; Kulinski, T.; Kulhanek, P.; Ennifar, E.; Koca, J.; Sponer, J. Biophys. J. 2007, 93, 3932–3949. (81) Reblova, K.; Spackova, N.; Stefl, R.; Csaszar, K.; Koca, J.; Leontis, N. B.; Sponer, J. Biophys. J. 2003, 84, 3564–3582. (82) McDowell, S. E.; Spackova, N.; Sponer, J.; Walter, N. G. Biopolymers 2007, 85, 169–184. (83) Barthel, A.; Zacharias, M. Biophys. J. 2006, 90, 2450–2462. (84) Deng, N. J.; Cieplak, P. J. Chem. Theory Comput. 2007, 3, 1435– 1450. (85) Kopitz, H.; Zivkovic, A.; Engels, J. W.; Gohlke, H. ChemBioChem 2008, 9, 2619–2622. (86) Kormos, B. L.; Baranger, A. M.; Beveridge, D. L. J. Struct. Biol. 2007, 157, 500–513. (87) McCrate, N. E.; Varner, M. E.; Kim, K. I.; Nagan, M. C. Nucleic Acids Res. 2006, 34, 5361–5368. (88) Mu, Y. G.; Stock, G. Biophys. J. 2006, 90, 391–399. (89) Pitici, F.; Beveridge, D. L.; Baranger, A. M. Biopolymers 2002, 65, 424–435. (90) Sarzynska, J.; Kulinski, T. J. Biomol. Struct. Dyn. 2005, 22, 425– 439. (91) Yingling, Y. G.; Shapiro, B. A. J. Mol. Biol. 2005, 348, 27–42.

JP102572K

D RNA Kink-Turn ... - ACS Publications

Recommend Documents