Altered structural fluctuations in duplex RNA versus DNA: a conformational switch involving base pair opening

Abstract

DNA and RNA are known to have different structural properties. In the present study, molecular dynamics (MD) simulations on a series of RNA and DNA duplexes indicate differential structural flexibility for the two classes of oligonucleotides. In duplex RNA, multiple base pairs experienced local opening events into the major groove on the nanosecond time scale, while such events were not observed in the DNA simulations. Three factors are indicated to be responsible for the base opening events in RNA: solvent–base interactions, 2′OH(n)–O4′(n+1) intra-strand hydrogen bonding, and enhanced rigid body motion of RNA at the nucleoside level. Water molecules in the major groove of RNA contribute to initiation of base pair opening. Stabilization of the base pair open state is due to a ‘conformational switch’ comprised of 2′OH(n)–O4′(n+1) hydrogen bonding and a rigid body motion of the nucleoside moiety in RNA. This rigid body motion is associated with decreased flexibility of the glycosyl linkage and sugar moieties in A-form structures. The observed opening rates in RNA are consistent with the imino proton exchange experiments for AU base pairs, although not for GC base pairs, while structural and flexibility changes associated with the proposed conformational switch are consistent with survey data of RNA and DNA crystal structures. The possible relevance of base pair opening events in RNA to its many biological functions is discussed.

INTRODUCTION

RNA is one of the key biological macromolecules, being involved in a variety of processes that require specific molecular recognition, including transcription and translation (1). Our understanding of the varied roles of RNA has increased rapidly in the past two decades, particularly since the discovery of catalytic RNA (2,3), which has been followed by the elucidation of the extensive structural and catalytic contributions of RNA to the ribosome (4,5). Central to its diverse biological functions, RNA molecules adopt a variety of complex three-dimensional (3D) structures, showing it to be significantly more polymorphic than DNA (6–8). For example, the large ribosomal subunit (9) includes various non-canonical conformations of RNA, including those important for interactions with proteins, while small ribozymes assume four distinct tertiary structural types, with known variations amongst each type (5). Other examples are the structures of transfer RNA (10). In all of these RNA structures, double-helical, duplex regions are combined with various stem–loops, bulges and mismatches. In contrast, DNA assumes primarily double-helical, duplex structures.

The double-helical, duplex forms of RNA and DNA are their most studied 3D structures. Helical, duplex RNA is conformationally more confined than DNA, with duplex RNA primarily being of the A form while DNA duplexes are able to assume the B, A or Z forms, among others (11–15), with the assumed form being influenced by both sequence and the environment (16–20). The more variable nature of duplex DNA is supported by its ability to incorporate non-standard nucleotides without disrupting the overall B-form conformation (21). In addition, DNA can assume the A form in hybrid duplexes with RNA, while RNA usually cannot fit into the B-form context, with such structures being unstable (22–24). Thus, there is an inconsistency in the structural variability of RNA versus DNA. While RNA assumes a wider variety of 3D structures as compared to DNA, at the secondary (i.e. helical) level it is less polymorphic than DNA. Experimental studies have clearly shown this apparent contradiction. However, experimental approaches alone may be limited in understanding this phenomenon.

Understanding of the structure and dynamics of oligonucleotides at the atomic level has been greatly facilitated by theoretical approaches. Multiple examples of the application of both quantum mechanical (QM) (25–30) and molecular dynamic (MD) based studies of nucleic acids are available in the literature (31–48). With respect to RNA, the orientation of the 2′OH group, which is believed to be the dominant factor responsible for the structural and dynamical differences between RNA and DNA, is not readily deduced from experimental data (49,50). Therefore, interpretation of experimental data is equivocal regarding the role of the 2′OH on RNA structure. MD simulations, on the other hand, can provide an atomic detail picture to help elucidate subtle structural effects associated with the 2′OH (51,52). Clearly, MD simulations will make a significant contribution in elucidating oligonucleotide structural and dynamic properties.

In this manuscript, we report results from MD simulations indicating differential structural fluctuations of duplex RNA relative to DNA, leading to increased base pair opening events on the nanosecond time scale in the former. These opening events are shown to be related to altered hydration and a ‘conformational switch’ unique to RNA that is comprised of a rigid body behavior of the nucleoside subunits and 2′OH(n)–O4′(n+1) hydrogen bonds. In addition, published experimental results and data from surveys of the nucleic acid database (NDB) (53) are presented to support the conclusions based on the MD simulations.

MATERIALS AND METHODS

MD simulations were performed using the program CHARMM (54) with the CHARMM27 nucleic acid force field (55,56), the modified TIP3P water model (57) and published sodium and chloride parameters (58). Starting structures were either the canonical A or B form (59) of the respective sequences and were prepared with QUANTA (Accelrys Inc.). For UAAGGAGGUGAU (IX in Table 1) (PDB identifier: 1sdr) (60) the starting structure was the crystal structure. Starting structures were overlaid with a pre-equilibrated solvent box consisting of water and sodium ions or NaCl, as required. The solvent box was extended 8 Å beyond the nucleic acid non-hydrogen atoms. Solvent molecules that were within 1.8 Å of nucleic acid non-hydrogen atoms were deleted. The number of sodium ions in the solvent box was adjusted to yield NaCl concentrations of ∼0.3 and 1 M by deleting sodium ions that were furthest from nucleic acid or adding sodium and chloride ions at random positions. The number of chloride ions added was selected to neutralize the systems for the high salt concentration (1 M). Periodic boundary conditions were simulated using the CRYSTAL module (61) in CHARMM. Each system was minimized for 500 Adopted Basis Newton–Raphson (ABNR) steps with mass-weighted harmonic constraints of 2.0 kcal/mol/Å on the non-hydrogen atoms of the oligonucleotide. The minimized system was then subjected to a 20 ps MD simulation in the constant volume, isothermal (NVT) ensemble in the presence of the harmonic constraints. The integration time step was 2 fs and SHAKE (62) was used to constrain all covalent bonds involving hydrogens. In all calculations, electrostatics were treated via the particle mesh Ewald method (63,64), Lennard–Jones interactions were truncated at 12 Å with a switch smoothing function from 10 to 12 Å and the non-bond atom lists were updated heuristically. Production simulations were performed for 5 ns in the isothermal isobaric (NPT) ensemble at 300K with the Leap-Frog integrator. The NPT ensemble was achieved using Hoover chains (65) for temperature control with a thermal piston mass of 1000 kcal/mol/ps² at 300K with pressure control to 1 ATM via the Langevin Piston method (66) with a piston mass of 600 a.m.u. and the piston collision frequency set to 0. In selected cases the simulations were extended to 7 ns to ensure that the final 4 ns had converged (see below). The final 4 ns from each trajectory were used for analysis. For the identification of hydrogen bonds, distance cutoffs of 2.5 and 3.5 Å were used for hydrogen–heteroatom and heteroatom–heteroatom pairs, respectively. A base pair is defined as being in the open state when the N1(purine)–N3(pyrimidine) hydrogen bond distance exceeds 3.5 Å.

Table 1. Duplex sequences, starting conformations and salt concentrations for performed simulations.

IDa	Duplex sequenceb	Starting conformation	NaCl conc. (M)
I	GAGUACUC	A	0.3, 1
II	GCGAGUACUCGC	A	0.3, 1
III	d(GAGUACUC)	A	0.3, 1
IV	d(GCGAGUACUCGC)	A	0.3, 1
V	d(GAGTACTC)	A	0.3, 1
VI	d(GAGTACTC)	B	0.3, 1
VII	CGCGAUCGCG	A	0.3, 1
VIII	CCUUUCGAAAGG	A	0.3, 1
IX	UAAGGAGGUGAUc	A	0.3, 1

^aThe roman numerals identifiers are used for referring to the corresponding nucleic acid duplexes throughout the manuscript.

^bThe central six bases used for analysis are printed in bold.

Experimental survey data of oligonucleotide crystal structures was based on the NDB (53). Structures compiled included duplex structures for RNA, B-form DNA and A-form DNA. In addition to the RNA duplex structures, tRNAs, ribozymes and RNA in the ribosome were analyzed. These structures were selected based on the NDB classifications and only structures with a resolution of 2.0 Å or better were used for the survey of duplex A-DNA, B-DNA and A-RNA. Based on this criteria, 74, 68 and 35 duplex B-DNA, A-DNA and A-RNA duplexes, respectively, were identified. Most of the tRNA, ribosome and ribozyme structures have lower resolutions (>2.0 Å) and therefore all nine tRNA, 65 ribosomal RNAs and nine ribozymes were included in the survey.

RESULTS AND DISCUSSION

A variety of duplexes have been investigated in the proposed study, with the majority of sequences studied in two salt concentrations. Sequence I was initially selected as it represents the control sequence in an experimental study on GU mismatches in RNA (67); our long-term goal is to elucidate the physical origins of the unique stability properties of these mismatches. During these efforts we came upon the interesting results presented in this manuscript. Results of the GU mismatch study will be published elsewhere (Pan and MacKerell, in preparation). Sequence II was chosen to study the impact of end effects on the observed structural and dynamic properties. ‘DNA’ sequences III and IV have the same base pair sequence and composition as I and II and were selected to test how the RNA 2′OH group impacts the structural and dynamic properties while maintaining the uracil bases. DNA sequence V was selected as it includes both the omission of the 2′OH groups and the U→T base switch. Sequence VI is the same as V except the simulation was initiated from the B conformation, allowing for convergence based on the starting structure to be tested. Sequences VII and VIII were included as they have been the subject of imino proton exchange experiments (68). Finally, RNA duplex IX was included as its sequence differs significantly from the other oligonucleotides and a crystal structure is available (60). This collection of oligonucleotides allows for the influence of sequence and chemical differences between RNA and DNA on the observed differences in the fluctuations between RNA and DNA to be investigated, as well as allowing for direct comparisons with experimental data.

Structural and dynamical properties

Average root mean square differences (RMSD) for the simulations are presented in Table 2. In all cases both the canonical A and B conformations are used as reference states, despite the fact that only two of the simulations were started from the B form (Table 1). The DNAs starting from the A form, including those containing U bases, all experienced an A→B transition, although the longer sequences took more time to converge (see Supplementary Material, Fig. S1). These simulations (sequence IV) were extended to 7 ns to ensure that the last 4 ns of the trajectories were properly converged for analysis. It can be seen that in both low (0.3 M) and high (1 M) salt concentrations, the RMSDs of RNA simulations are distinctly smaller when the reference state was the A form versus the B form. In addition, for sequence IX the average RMS differences with respect to the crystal structure were 1.5 and 1.3 Å at low and high salt concentrations, respectively, further indicating that the presented calculations yield the experimentally appropriate conformations.

Table 2. Average RMS differences (Å).

Duplex	0.3 M NaCl				1 M NaCl
	Heavy	Base	Sugar	Phos	Heavy	Base	Sugar	Phos
Versus A form
I	2.7 ± 0.5	2.6 ± 0.5	3.2 ± 0.7	2.6 ± 0.5	2.5 ± 0.7	2.4 ± 0.7	3.0 ± 0.9	2.5 ± 0.7
II	2.0 ± 0.6	1.8 ± 0.5	2.6 ± 0.7	2.0 ± 0.6	1.6 ± 0.6	1.5 ± 0.6	1.9 ± 0.7	1.6 ± 0.5
III	2.4 ± 0.3	1.8 ± 0.2	3.8 ± 0.5	2.4 ± 0.4	2.5 ± 0.3	1.9 ± 0.3	3.8 ± 0.5	2.4 ± 0.4
IV	2.4 ± 0.4	1.7 ± 0.3	3.6 ± 0.5	2.5 ± 0.4	2.6 ± 0.3	1.9 ± 0.3	3.8 ± 0.5	2.7 ± 0.4
V	2.4 ± 0.3	1.7 ± 0.3	3.9 ± 0.5	2.4 ± 0.4	2.3 ± 0.3	1.6 ± 0.2	3.7 ± 0.5	2.3 ± 0.4
VI	2.6 ± 0.3	1.7 ± 0.3	4.3 ± 0.4	2.5 ± 0.3	2.4 ± 0.3	1.6 ± 0.2	4.1 ± 0.3	2.4 ± 0.3
VII	2.3 ± 0.6	2.2 ± 0.4	3.2 ± 1.1	2.0 ± 0.6	2.4 ± 0.4	2.8 ± 0.6	2.2 ± 0.3	2.0 ± 0.3
VIII	1.4 ± 0.4	1.2 ± 0.4	2.1 ± 0.6	1.3 ± 0.4	1.7 ± 0.5	1.5 ± 0.4	2.4 ± 0.8	1.6 ± 0.5
IX	1.5 ± 0.4	1.7 ± 0.5	1.7 ± 0.4	1.3 ± 0.3	1.3 ± 0.4	1.1 ± 0.4	1.8 ± 0.6	1.2 ± 0.4
Versus B form
I	4.2 ± 0.5	3.4 ± 0.5	5.3 ± 0.7	4.6 ± 0.5	4.2 ± 0.7	3.3 ± 0.7	5.3 ± 0.9	4.6 ± 0.6
II	3.6 ± 0.4	2.5 ± 0.4	4.8 ± 0.6	4.0 ± 0.4	3.4 ± 0.3	2.3 ± 0.3	4.6 ± 0.3	3.8 ± 0.3
III	1.6 ± 0.3	1.1 ± 0.3	1.9 ± 0.4	1.9 ± 0.4	1.7 ± 0.4	1.2 ± 0.3	1.9 ± 0.5	2.0 ± 0.5
IV	1.7 ± 0.3	1.1 ± 0.3	2.8 ± 0.4	1.7 ± 0.4	1.6 ± 0.3	1.0 ± 0.2	2.6 ± 0.3	1.6 ± 0.4
V	1.8 ± 0.3	1.6 ± 0.2	2.8 ± 0.3	1.7 ± 0.4	1.9 ± 0.3	1.6 ± 0.2	2.8 ± 0.4	1.7 ± 0.4
VI	1.4 ± 0.3	1.1 ± 0.2	1.8 ± 0.4	1.7 ± 0.3	1.5 ± 0.3	1.1 ± 0.2	1.9 ± 0.4	1.8 ± 0.3
VII	3.7 ± 0.4	3.1 ± 0.3	4.7 ± 0.7	3.9 ± 0.5	3.7 ± 0.3	3.6 ± 0.5	4.4 ± 0.4	3.6 ± 0.3
VIII	6.0 ± 0.6	5.1 ± 0.7	7.4 ± 0.7	6.3 ± 0.5	5.7 ± 0.4	4.9 ± 0.4	7.0 ± 0.6	6.1 ± 0.5
IX	3.3 ± 0.3	2.4 ± 0.2	4.2 ± 0.4	3.7 ± 0.3	3.3 ± 0.3	2.3 ± 0.3	4.3 ± 0.5	3.8 ± 0.3

The canonical A- and B-form structures were used as the reference states. Averages over the final 4 ns and only non-hydrogen atoms were included. Errors represent the RMS fluctuations.

Average phosphodiester backbone dihedrals (Table 3) and sugar pseudorotation angle distributions (data not shown) confirm the conformational difference between RNA and DNA. The calculated dihedral angles are in good agreement with experimental survey data for DNA and RNA (69–71), respectively, and the sugar pucker distributions are also typical for their respective canonical conformations (25,72). In addition, analysis of helicoid parameters via the program CURVES (73) yielded results (not shown) consistent with the conclusions based on the RMS difference and dihedral angle analysis. These results indicate that the present simulations yield the expected A-like ensembles for the RNAs and B-like ensembles for the DNAs (31,74). Moreover, these results confirm the dominant role of the 2′OH group on the conformational preference for RNA (11), since all the DNA sequences lacking the 2′OH but still containing U bases underwent the transition from the A to B forms.

Table 3. Average backbone dihedral angles and their fluctuations for the central six residues (degrees)^a.

Duplex	α	β	γ	δ	ε	ζ	χ
0.3 M NaCl
I	282 ± 20	173 ± 16	63 ± 11	77 ± 5	205 ± 19	290 ± 24	200 ± 9
II	285 ± 21	172 ± 18	66 ± 12	78 ± 5	210 ± 18	285 ± 29	200 ± 8
III	298 ± 24	165 ± 20	52 ± 14	129 ± 18	194 ± 20	250 ± 31	249 ± 18
IV	299 ± 18	170 ± 18	51 ± 14	130 ± 19	192 ± 19	253 ± 28	248 ± 19
V	299 ± 18	163 ± 23	52 ± 14	129 ± 19	197 ± 24	250 ± 45	250 ± 18
VI	297 ± 27	166 ± 24	53 ± 16	126 ± 22	198 ± 27	245 ± 42	250 ± 21
VII	283 ± 22	173 ± 17	63 ± 11	78 ± 6	204 ± 23	285 ± 22	200 ± 10
VIII	283 ± 20	173 ± 17	65 ± 12	77 ± 5	207 ± 17	287 ± 20	200 ± 9
IX	283 ± 17	174 ± 18	64 ± 11	77 ± 5	208 ± 19	290 ± 19	199 ± 8
1 M NaCl
I	285 ± 21	173 ± 18	64 ± 13	77 ± 5	208 ± 18	288 ± 19	200 ± 8
II	285 ± 19	170 ± 18	64 ± 10	77 ± 5	210 ± 17	285 ± 22	200 ± 8
III	296 ± 25	165 ± 20	52 ± 14	126 ± 21	197 ± 22	249 ± 35	246 ± 22
IV	300 ± 20	169 ± 20	52 ± 14	133 ± 15	193 ± 21	249 ± 31	251 ± 16
V	298 ± 22	166 ± 22	53 ± 14	127 ± 21	193 ± 23	243 ± 48	247 ± 19
VI	299 ± 21	168 ± 21	52 ± 15	128 ± 20	191 ± 17	256 ± 24	247 ± 19
VII	280 ± 25	173 ± 19	62 ± 13	77 ± 6	206 ± 22	285 ± 22	202 ± 13
VIII	283 ± 19	172 ± 17	64 ± 12	77 ± 5	208 ± 18	286 ± 27	200 ± 9
IX	284 ± 20	173 ± 19	63 ± 12	77 ± 6	210 ± 19	291 ± 25	199 ± 9
Expt (A form)b	291	175	57	80	205	287	199
Expt (B form)b	298	168	51	134	187	262	252

^aEach dihedral was averaged over the last 4 ns for all the central six nucleosides. Errors represent RMS fluctuations.

^bSee Foloppe and MacKerell (69).

While the expected conformations were obtained from the simulations, inspection of the data in Table 2 reveals some interesting differences between RNA and DNA. The RMSDs of the bases and phosphates from the RNA simulations (I, II, VII, VIII and IX) with respect to the A form were similar, with the differences between those moieties being <0.3 Å, while with the DNA simulations (III, IV, V and VI), the differences between base and phosphate RMSDs with respect to the B form were at least 0.4 Å, with the base variations less than the phosphates. This behavior is opposite to that observed in the RNAs. This observation indicates that the structural properties of the backbone, relative to the bases, differ in RNA versus DNA.

Consistent with the structural differences in RNA and DNA are the RMS fluctuations of the bases and the backbone moieties, as shown in Table 4. In RNA the RMS fluctuations for base, sugar and phosphate moieties are similar. In contrast, the phosphates have the largest fluctuations and the bases fluctuate the least in the DNA simulations. Thus, the present results indicate that there are inherent differences in the structural variations and flexibility in RNA versus DNA, with the motions of the individual moieties in DNA being relatively independent, while in RNA those motions are more coordinated, with the nucleosides and nucleotides moving more as a rigid body.

Table 4. RMS structural fluctuations (Å) for the central six residues from the last 4 ns of the trajectories.

Duplex	All	Base	Sugar	Phosphate
0.3 M NaCl
I	1.5	1.6	1.3	1.6
II	1.3	1.4	1.2	1.4
III	0.9	0.7	0.9	1.2
IV	0.9	0.7	0.9	1.2
V	0.9	0.7	0.9	1.2
VI	0.9	0.7	0.9	1.2
VII	1.2	1.2	1.4	1.1
VIII	1.1	1.0	1.2	1.0
IX	1.0	1.0	1.1	0.9
1 M NaCl
I	1.8	1.8	1.6	1.8
II	1.0	1.0	0.9	1.1
III	0.9	0.8	1.0	1.3
IV	0.9	0.7	0.9	1.1
V	0.9	0.7	0.9	1.1
VI	0.8	0.7	0.9	1.1
VII	0.9	0.9	0.9	0.8
VIII	1.4	1.4	1.6	1.3
IX	1.0	0.9	1.2	1.0

To validate the differences in the relative mobilities of the bases and backbone moieties in RNA and DNA a survey was undertaken of duplex DNA and RNA in the NDB. Reported in Table 5 are the average B-factors for the different moieties in B-form DNA, A-form DNA and RNA duplexes along with the average differences of the B-factors for the individual crystal structures. In B-form DNA, the difference in the average B-factors upon going from the base to the phosphate is close to 11, while in A-form DNA it decreases to ∼8 and is slightly larger than 6 in RNA. Comparison of the sugar and base B-factors shows a similar pattern, where the differences are 6.2 in B DNA, 4.6 in A DNA and 3.6 in RNA. These differences are beyond the standard errors as well as being above 99% confidence levels in the two-tail t-test, with the exception of the average sugar–phosphate difference between A DNA and RNA (not shown). Thus, the survey data are consistent with the simulation results, indicating the moieties in B DNA to have the most independent flexibilities, with the differences decreased in A-form DNA and the lowest in RNA, supporting the present observation that RNA at the nucleoside and nucleotide levels moves more like a rigid body than DNA.

Table 5. Average B factors and B-factor differences of DNA and RNA duplexes from NDB database.

	Average B factors
	Base	Sugar	Phosphate
B-DNA	15.7 ± 0.9	20.4 ± 1.1	24.9 ± 1.3
A-DNA	16.1 ± 0.8	19.8 ± 1.0	23.5 ± 1.2
RNA	19.9 ± 1.6	23.1 ± 1.5	25.5 ± 1.6
	Average B-factor differences
	Phosphate–Base	Sugar–Base	Phosphate–Sugar
B-DNA	10.8 ± 0.7	6.2 ± 0.4	4.6 ± 0.3
A-DNA	8.4 ± 0.5	4.6 ± 0.3	3.8 ± 0.2
RNA	6.3 ± 0.6	3.6 ± 0.4	2.6 ± 0.3

Averages are over the non-hydrogen atom B factors for the respective moieties in each structure included in the survey. Average B-factor differences represent the average over all crystal structures for the respective duplex types of the differences over the respective moieties in each crystal. Errors represent the standard errors (88). Units in Ų.

Further support for the more rigid body like motion in RNA versus DNA at the nucleoside and nucleotide levels comes from comparison of the RMS fluctuations of the phosphodiester backbone dihedrals (Table 3) versus the cartesian coordinate RMS fluctuations (Table 4). In RNA, the RMS fluctuations of the dihedrals are generally smaller than in DNA, with the difference being most significant with δ and χ. In contrast, the RMS fluctuations of the cartesian coordinates of DNA are typically lower than those of the RNA. Such a difference is consistent with a model where RNA at the nucleoside and nucleotide level is moving more as a rigid body than in DNA, leading to lower dihedral fluctuations, while larger fluctuations of the nucleoside and nucleotide moieties are occurring in RNA leading to the larger cartesian RMS fluctuations.

Base pair opening

Further investigations of the differential mobility in RNA versus DNA involved monitoring the Watson–Crick (WC) base pairing. Analysis of the N1–N3 hydrogen bond distances revealed that the WC hydrogen bonds in RNA were frequently broken. Supplementary Material Table S1 summarizes the number and duration of the opening events for the central six base pairs based on the N1–N3 distance. The corresponding time series details are also shown in the Supplementary Material (Table S2 and Fig. S2). For the central six base pairs in the RNA octamers, the N1–N3 H-bonds were broken at least once, some of which were reformed over the 4 ns production trajectory. For duplex II, of which the central eight residues are the same as duplex I, four of the six central base pairs show the same property. The duration of these events covered a wide range, from a few picoseconds to a couple of nanoseconds. Less base opening events were observed at the high salt versus low salt concentration, consistent with salt stabilization of RNA (75). However, opening events did occur in the high salt conditions. Interestingly, all the base pair opening events occurred into the major groove. In contrast, DNA did not have a significant tendency toward base pair opening (Supplementary Material, Fig. S2). In all of the DNA simulations there were only four short opening events, the duration of which were all <10 ps (Table S1).

To see whether the base pair opening may be a common phenomenon in RNA, additional simulations of RNA with different sequences were undertaken (sequences VII, VIII and IX, Table 1). As may be seen in Table S1 a variety of opening events occurred in these systems, suggesting that the observed opening events in RNA are not limited to selected sequences. Of the additional sequences studied, VII and VIII have been subjected to NMR-based imino proton exchange experiments (68). In the NMR study it was observed that the lifetime of several of the AU base pairs are quite short, with most being ≤0.1 ms. In several cases the lifetimes were too short to be measured at 15°C. This contrasts lifetimes for d(AT) base pairs in DNA, which are typically 10 ms or longer (76,77). In addition, it was shown that the dissociation constants of AU pairs were similar to that of d(AT) pairs, indicating that the lifetimes of both the open and closed states are reduced in RNA versus DNA, resulting in open lifetimes of ∼1 ns being reported. This timescale is consistent with opening observed in the present calculations. Thus, the present results are in agreement with experimental data indicating enhanced opening of AU base pairs in RNA versus AT base pairs in DNA, with the timescales of those motions being on the nanosecond level.

However, several additional points require discussion. Snoussi and Leroy (68) did not see a significant increase in the opening of GC base pairs in RNA, while the present simulation data does indicate increased opening of GCs. Such a discrepancy indicates a possible bias by the force field used in the present study towards favoring opening, although other confounding factors may contribute. These include temperature differences between the experimental and MD studies, differences in ionic strengths and ions used, limited sampling in the MD simulations as well as inherent limitations in the force field, including lack of electronic polarization. Moreover, it has been shown experimentally that base pair lifetimes in the millisecond time range occur in polyA–polyU (78) and in AU pairs stacked between GC pairs (68), indicating significant sequence-specific effects. Consistent with the latter are computational studies on base flipping in RNA, where the conversion from DNA to RNA for flipping of a U stacked between two Gs (versus a T in the DNA) did not significantly change the flipping energy surface (79). Despite these limitations, the significant increase in base pair opening in AU base pairs is consistent with the experimental data, indicating the calculated observations to be representative of the experimental regimen.

2′OH(n)–O4′(n+1) hydrogen bonding

Simply switching the thymines to uracils in the context of DNA does not lead to enhanced opening, as evidenced by the results for III and IV. This is consistent with experimental studies (68) and, again, points to the dominant role of the 2′OH in the altered behavior of DNA versus RNA. To examine the role of the 2′OH group in RNA base pair opening, hydrogen bonding between adjacent sugars via the 2′OH(n) and the O4′(n+1) was analyzed for each base opening event (Supplementary Material, Table S2). Notably, at least one of the two possible 2′OH–O4′ hydrogen bonds from the two strands was formed in 133 out of the 172 opening events. This was based on the average 2′OH(n)–O4′(n+1) distance being less than 2.5 Å while the average N1–N3 distance was >3.5 Å. In 65 cases, the 2′OH(n)–O4′(n+1) distances for both hydrogen bonds were <2.5 Å. Compared with the average distance (3.3–3.5 Å) between 2′OH(n) and O4′(n+1) for the all the RNA trajectories (see Supplementary Material), it is evident that 2′OH(n) and O4′(n+1) groups had significant interactions during the base pair open states. These results indicate that 2′OH(n)–O4′(n+1) hydrogen bonding is playing an important role in the base pair opened state.

To verify the relevance of the 2′OH(n)–O4′(n+1) hydrogen bonds to the experimental regimen, probability distributions of the O2′(n)–O4′(n+1) distances were obtained from the RNA MD simulations and from the NDB survey data. Presented in Figure 1 are O2′(n)–O4′(n+1) distance probability distributions for all the RNA simulations along with those from the NDB survey for the different structural types of RNA. In the distributions two peaks are evident. The shorter peak at ∼3 Å corresponds to a O2′(n)–O4′(n+1) hydrogen bond while the longer peak represents a non-hydrogen bond state. In the MD simulations the presence of the hydrogen bond peak is evident. This peak is also present in the survey data for tRNA, ribozymes and ribosomal RNA, while in the duplex RNA structures only a shoulder is observed in this region. Such a discrepancy is interesting. First, the observed opening events are consistent with imino proton exchange data for AU base pairs, however, in RNA A-form duplex structures the base pairs are well maintained and only a minor amount of O2′(n)–O4′(n+1) hydrogen bonding is evident from the survey data. This suggests that either the crystal environment is stabilizing the AU base pairs, thereby inhibiting the opening events, or that the opening events are short-lived such that they are not observed in the crystal structures. The latter is consistent with the fast AU opening and closing rates in the experimental study, suggesting that base pair opening events may be occurring in the crystals, however, they are not observed due to the timescale of the experiments. Consistent with our simulation results are the sugar B factors for RNA versus DNA (Table 5), where larger B-factors occur in RNA versus DNA, while the B-factor difference between structural moieties is more dramatic in DNA than in RNA. However, in the non-canonical RNA structures, where a significant number of bases are not involved in WC interactions, a high percentage of the O2′(n)–O4′(n+1) pairs are hydrogen bonded. This suggests a scenario where opening events can occur in all forms of RNA, but are typically not stable enough to be observed via crystallography in A-form duplex RNA. Only when the WC interactions are not present are O2′(n)–O4′(n+1) intrastrand pairs in hydrogen bonding orientations experimentally observed. While not conclusive, NMR data did not exclude the presence of such hydrogen bonds (49), supporting the feasibility of the present hypothesis.

Figure 1.

O2′(n)–O4′(n+1) probability distribution from all the RNA simulations (solid line) and from a survey of the NDB. Survey results for A-RNAs (dotted line), tRNAs (circles), ribosomes (squares) and the ribozyme (diamonds) are shown.

To more closely investigate the relationship of 2′OH(n)–O4′(n+1) hydrogen bonding to base pair opening, we selected two representative events, one from the sequence II simulation at 1 M NaCl and the second from the sequence VIII at 0.3 M NaCl, for further analysis. The event from duplex II lasted ∼1.4 ns (from 1.38 to 2.78 ns) and the second from duplex VIII lasted 0.14 ns (from 1.16 to 1.30 ns). The N1–N3 and corresponding 2′OH–O4′ distance time series for these events are shown in Figure 2. The 2′OH–O4′ distances fluctuate within the range of 2.5–4.5 Å when the base pair WC hydrogen bonds are present for both events. However, during base pair opening, not only do the 2′OH–O4′(n+1) distances decrease significantly, but the fluctuations of those distances become smaller, indicating how 2′OH(n)–O4′(n+1) hydrogen bond formation and the breakage of the N1–N3 hydrogen bond are correlated.

Figure 2.

Selected N1–N3 and 2′OH(n)–O4′(n+1) distances (A and B) and snapshots of base pairs with local waters or ions in the major groove (C, D and E) for the (A and C) U6-A7 base pair from the duplex II simulation at 1 M NaCl and (B and D) A9-U4 base pair from the duplex VIII simulation at 0.3 M NaCl. Also included (E) is the DNA C6-G3 base pair with local waters from the duplex V simulation at 0.3 M NaCl. In (A) and (B), black is the N1–N3 distance, and red and green are the 2′OH–O4′ distances for strands one and two, respectively. For (C), (D) and (E), hydrogen bonds are shown with dotted lines and the two spheres at 1714 ps in (C) are Na⁺ (yellow) and Cl^– (magenta) ions.

To understand if the 2′OH(n)–O4′(n+1) hydrogen bond was initiating base pair opening, the N1-N3 and 2′OH(n)–O4′(n+1) distances from strand one were followed chronologically for the two events (Supplementary Material, Table S3). During initial base pair opening, breakage of the N1–N3 hydrogen bond (distance changed from <3.5 to >3.5 Å) and formation of the 2′OH(n)–O4′(n +1) hydrogen bond (distance changed from >2.5 to <2.5 Å) occurred simultaneously at 1380 ps for one event and at 1160 ps for the other (Table S3). In the reverse process (i.e. base pair closing), the N1–N3 distance decreased from 4.2 to 3.4 Å and then to 3.1 Å from 2670 to 2674 ps for event one. However, the hydrogen bond between 2′OH and O4′(n+1) was maintained during that period, at 1.9, 1.8 and 2.0 Å, respectively. Similar results were observed for the second event. Since 2′OH–O4′(n+1) hydrogen bond formation prior to base pair opening and loss of the 2′OH–O4′(n+1) hydrogen bond prior to the base pair re-closing were not observed, it is concluded that the formation and breakage of the 2′OH(n)–O4′(n+1) hydrogen bonds are not pre-requisites for base pair opening and closing, respectively. Thus, the role of the 2′OH(n)–O4′(n+1) hydrogen bond appears to be stabilization of the open state of the base pairs rather than being responsible for initiation of such events.

The role of solvent molecules in base pair opening

Solvent–solute interactions are an indispensable part of the structure and function of macromolecules. Accordingly, it may be anticipated that solvent may contribute to the differential base pair opening behavior in RNA versus DNA. To investigate this possibility, the number of water molecules within 3.5 Å of the N1 and N3 atoms of each base pair was calculated for the closed states of the simulations (Supplementary Material, Table S4). For the central six base pairs, more water molecules were adjacent to the N1 and N3 atoms in RNA as compared to DNA, supporting the possibility that major groove hydration in RNA may facilitate the opening events.

Details of solvent–base interactions prior to selected base pair opening events were examined to better determine if the increased presence of water molecules near RNA base pairs facilitate base pair opening. The same two events in duplexes II and VIII discussed in the previous section were analyzed. Since the base pair opening pathway was through the major groove for all events, we only focus on the major groove side of the base pairs. For the first opening event (Fig. 2C), at 1378 ps, a water hydrogen bonds with the O4 atom (1.75 Å) of U6 from strand one and with H62 (1.83 Å) of A7 of strand two. At this point, the N1–N3 hydrogen bond is still formed. At 1380 ps, as the N1–N3 hydrogen bond breaks, the water molecule forms a hydrogen bond (1.94 Å) with N1 of A7, followed by an additional hydrogen bond with N3 of U6 at 1382 ps. After this, the base pair remained open until 2780 ps, during which at least one water or sodium chloride ions were between the two bases, forming a hydrogen bond network (see below). For the second event (Fig. 2D), a water molecule competes with H62 of A9 for hydrogen bonding with O4 of U4, disrupting the H62–O4 hydrogen bonding right before the N1–N3 hydrogen bond breakage (1158 ps). This water molecule further hydrogen bonds with both the N1 of A9 and the H3N3 of U4, marking the beginning of the base pair opening event at 1160 ps. There was at least one water molecule hydrogen bonding with the bases during the base opened state (data not shown). The WC hydrogen bond was restored for this base pair from 1300 to 1314 ps when the water molecule was released from between the two bases.

Similar interactions between water and base pairs were found for the DNA simulations. For example, in the d(GAGTACTC)₂ A-form simulation at 0.3 M NaCl (V), which involves a 10 ps base pair opening event (Table S1), the longest among all of the DNA simulations, a water molecule formed a hydrogen bond with O6 of G3 in strand one at 1398 ps (Fig. 2C). This water molecule then formed an additional hydrogen bond with the N4 of C6 in strand 2 at 1400 ps, when the base pair opened (N1–N3 distance of 3.6 Å). The base pair remained open until 1410 ps, during which the water–base hydrogen bonds were maintained. The water–base interactions were lost by 1420 ps, at which time the base pair reverted to a fully closed state. Thus, water molecules appear to be able to compete with the WC interactions via the major groove in both DNA and RNA. However, the increased presence of water molecules in the vicinity of the N1–N3 atoms of RNA may facilitate water-mediated base opening initiation events as compared to DNA. Once a base pair opening initiation does occur, additional factors in RNA lead to the actual base pair opening events. These factors include the increased rigid body motion of the individual nucleosides and the 2′OH–O4′(n+1) hydrogen bonds.

Further analysis of the interactions between water molecules and the opened base pairs reveal the dynamic nature of hydrogen bonding between water molecules and the bases during the base pair open state. For example, in the event shown in Figure 2C, a total of 12 water molecules were within 3.5 Å of N1 or N3 during the 1380–2670 ps period (Supplementary Material, Fig. S3), with different waters interacting with the open bases during that time frame. This dynamic behavior further suggests that the rigid body behavior and the 2′OH(n)–O4′(n+1) interactions in RNA stabilize the base pair open state allowing for exchange of water molecules or ions without base pair closing.

Of note is the relationship between the number of water molecules interacting with the bases and the extent of base pair opening. When one water molecule bridges the N1 and N3 atoms, the N1 and N3 separation is ∼5 Å (Fig. S2), which is typical in many of the opening events. Further separation of N1 and N3 to ∼6.5 Å or more was associated with two water molecules or ions participating in the hydrogen bond network (1664 ps, Fig. 2C). Moreover, for the event depicted in Figure 2A, no water–base hydrogen bonds were found during the 1676–1858 ps period. Instead, sodium and chloride acted as the bridging elements (1714 ps, Fig. 2C). From these results it is clear that solvent molecules play important roles, not only in triggering base pair opening, but also in maintaining the open state and influencing the extent of base pair opening.

Mechanism for RNA conformational switch

Based on the data presented above, it is proposed that structural fluctuations of both DNA and RNA combined with interactions with the surrounding solvent allow water molecules to compete for WC hydrogen bonding, initiating base pair opening. In the case of RNA, this can lead to base pair opening on the nanosecond time scale; such opening does not occur on that time scale in DNA. This additional opening in RNA is suggested to be facilitated by the increased rigid body nature of RNA at the nucleoside level as compared to DNA and hydrogen bonding between 2′OH(n) and O4′(n+1) atoms in RNA. This combination of rigid-body behavior and the 2′OH(n)–O4′(n+1) hydrogen bond therefore acts as a conformational switch allowing for increased structural variability in RNA as compared to DNA on a nanosecond timescale.

How the rigid body behavior and the 2′OH(n)–O4′(n+1) hydrogen bond combine to create a conformational switch is worth additional discussion. First, for the interactions associated with the 2′OH moiety to be transmitted to the base there are only two flexible regions that must be considered; the sugar pucker and the χ dihedral (i.e. glycosyl bond). Crystal surveys of the distributions of these terms show them to be narrow for A-form conformations (25,69). With the pseudorotation angle, in A-form DNA structures the distribution is centered in the vicinity of 18° and is quite narrow, while in B-form DNA structures it is centered in the vicinity of 160° and is quite broad. Importantly, the narrow distribution of pseudorotation angles also occurs in RNA crystal structures (55). Similarly, with χ, the distributions are wider for B-DNA crystal structures as compared to A-form structures, including RNA. Thus, the decreased flexibility of the sugar pucker and χ in A-form RNA structures will lead to a more rigid-body-like behavior as compared to DNA, which is dominated by the B form. This rigid body behavior is suggested to contribute to the correlation between base opening and the 2′OH–O4′(n+1) hydrogen bonds.

Importantly, the more restrained nature of χ and sugar pucker in RNA is supported by QM calculations on nucleic acid related models compounds. QM studies on energies associated with rotation about χ show the global minimum in the energy surface to coincide with the region sampled in A-form RNA, while B-form DNA is sampling a higher energy region (69,80). Since in the A form, χ is near its global minimum, decreased fluctuations would be expected as compared to the B form, where χ is sampling a higher energy region. With the pseudorotation angle, QM energy surfaces have been performed on a series of model compounds (25,55,69). These results also suggest that conformational energies associated with the sugar pucker lead to the narrow distribution seen in the north conformation in A-form crystal structures as compared to the south conformation typical to the B form. These QM results further support a model where intrinsic conformational energies of the glycosyl linkage and the sugar pucker in the A form lead to more rigid-body motions of the nucleoside in RNA as compared to DNA. The consistent picture from the present simulations, imino protein exchange experiments, crystal database surveys and the QM data support the validity of the proposed conformational switch in RNA. An image of the closed and open states from the duplex II simulation is shown in Figure 3.

Figure 3.

Snapshots of base pair closed (1.2 ns) and open (2.0 ns) states from duplex II simulation at 1 M NaCl, showing the conformational switch. Only three base pairs (G5U6A7 from strand one and U6A7C8 from strand two) are shown with the central pair involving the conformational switch colored in red. O2′H2′–O4′(n+1) hydrogen bonds present in the open state are shown as green lines. The major groove faces toward the reader in the upper panel and towards the top in the lower panel.

Conclusions

In the present study it is shown that DNA and RNA duplexes display different flexibility properties associated with the 2′OH group. The presence of the 2′OH group in RNA, allowing for the formation of 2′OH(n)–O4′(n+1) hydrogen bonds, combined with decreased intrinsic flexibility of the sugar and the χ dihedral in A-form structures leads to a proposed conformational switch in RNA that is not present in DNA. The presence of this conformational switch allows for local base pair opening events in RNA on the nanosecond timescale. Importantly, the increased base opening events in RNA versus DNA are consistent with imino proton exchange experiments for AU base pairs (68); however, increases in GC opening seen in the present calculations are not consistent with experimental data, indicating a possible bias towards the open state in the applied force field. In addition, 2′O(n)–O4′(n+1) hydrogen bonds occur in tRNA, ribozymes and ribosomal RNA crystal structures, while decreased flexibility of the sugar pucker and χ is consistent with NDB survey data for duplex DNA and RNA as well as conformational energies of model compounds based on QM calculations. The combination of simulation and experimental data strongly indicate the presence of a conformational switch in RNA that is responsible for increased base pair opening in RNA on the nanosecond time scale.

The proposed conformational switch in RNA may be important for its structural diversity and biological function. For instance, RNA folding is known to often involve kinetic traps associated with misfolded states (81,82). These misfolded states include mispairings which have been suggested to undergo pairing rearrangements prior to attaining the correctly folded state. Such conformational malleability is evidenced by the P5abc region from the Tetrahymena thermophilia self-splicing intron, which assumes different tertiary structures in different environments (81,83). Conformational heterogeneity also appears to be important for catalytic RNA. For example, the hairpin ribozyme undergoes alterations in hydrogen bonds during catalysis (84) and the hammerhead ribozyme is thought to undergo significant conformational changes in its catalytic cycle (5,85,86). Interestingly, in the hammerhead ribozyme, a 2′OH(n)–O4′(n+1) hydrogen bond has been indicated to stabilize an intermediate, possibly important for the catalytic mechanism (87). While speculative, it may be suggested that the conformational switch involving rigid-body motions at the nucleoside level and the 2′OH(n)–O4′(n+1) hydrogen bond may act to lower the barrier to these conformational transitions. For example, if the opening events observed in the present calculations occur in the vicinity of additional hydrogen bonding moieties (e.g. other bases or phosphate moieties) adjacent to the major groove, possible interactions between the opened base and those moieties may lead to further opening and larger conformational changes. Alternatively, although opening beyond 8 Å is only observed once in the present nanosecond MD simulations, such events may lead to more open states on the millisecond timescale associated with RNA folding; states that may be important for exiting misfolded conformations leading to the correctly folded structure. Further experimental studies are required to better understand the proposed conformational switch in RNA.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at NAR Online.