Resolving the Motional Modes That Code for RNA Adaptation

See allHide authors and affiliations

Science  03 Feb 2006:
Vol. 311, Issue 5761, pp. 653-656
DOI: 10.1126/science.1119488


Using a domain elongation strategy, we decoupled internal motions in RNA from overall rotational diffusion. This allowed us to site-specifically resolve a manifold of motional modes in two regulatory RNAs from HIV-1 with the use of nuclear magnetic resonance spin relaxation methods. Base and sugar librations vary on a picosecond time scale and occur within helical domains that move collectively at diffusion-limited nanosecond time scales. Pivot points are short, functionally important, and highly mobile internal loops. These spontaneous changes in RNA conformation correlate quantitatively with those that follow adaptive recognition of diverse targets. Thus, ligands may stabilize existing RNA conformations rather than inducing new ones.

Ribonucleic acids (RNAs) must adaptively change their conformation to meet the diverse requirements of their biological functions (13). The same RNA element can adopt conformations when assembling into a ribonucleoprotein complex that differ from those it adopts when carrying out its function (4, 5). Similarly, ribozymes take on markedly different conformations during their catalytic cycles in order to meet the unique demands of substrate binding, catalysis, and product release (6, 7).

The dynamic properties of RNA structures that are the basis of this plasticity remain poorly understood. This is largely due to difficulties in experimentally resolving complex superpositions of motional modes, each having unique amplitudes and time scales. These can include local librations at picosecond time scales, collective domain motions at nanosecond to milli-second time scales, and overall Brownian rotational diffusion at nanosecond to microsecond time scales (Fig. 1A). The individual contributions of these motional modes to spectroscopic observables cannot be readily resolved, especially when modes are physically coupled. For example, domain motions can change the RNA hydrodynamic shape and therefore the overall rotational diffusion. When these two motional modes have similar time scales, a dynamical coupling (8, 9) develops that renders their spectroscopic contributions inseparable (Fig. 1A).

Fig. 1.

Decoupling motional modes in RNA by domain elongation. (A) Collective motions lead to coupled changes in rotational diffusion described by the principal axis of the diffusion tensor, which is assumed parallel to the long axis of the RNA. (B) Decoupling collective motional modes by domain elongation. (C) NMR-invisible elongation of TAR RNA. The wild-type TAR loop is replaced by the more stable UUCG loop. Lines indicate Watson-Crick base pairs that are hydrogen-bonded, as detected by the JNN correlation spectroscopy (COSY) experiment (28). Isotopically unlabeled residues are shown in gray. Two terminal G-C base pairs are added to domain I in E-TAR to maximize yields by in vitro transcription. (D) Two-dimensional (2D) 1H-13C heteronuclear single-quantum coherence (HSQC) spectra of the aromatic region of E-AU-TAR and E-GC-TAR (in red) overlaid on corresponding spectra of nonelongated TAR (in black) in free form (FREE) and bound to ARG (+ARG). Asterisks denote resonances that belong to two terminal guanine (G–21 and G–22) and cytosine (C+21 and C+22) residues in E-AU-TAR. Examples of bulge and neighboring residues that undergo ARG-induced chemical shift perturbations are highlighted in circles.

We describe a domain elongation strategy that allows us to resolve picosecond local motions and nanosecond domain motions by nuclear magnetic resonance (NMR) spectroscopy. We applied this strategy to two regulatory RNAs from HIV-1. The strategy involves extending the size of an RNA terminal domain through a stretch of Watson-Crick base pairs (Fig. 1B). This results in an elongated RNA (E-RNA) that has a hydrodynamic shape (and therefore a rotational diffusion profile) that is less sensitive to domain motions. Furthermore, by slowing down overall tumbling, the elongation broadens the sensitivity of NMR relaxation, which is limited to internal motions occurring at time scales faster than overall tumbling (10). To avoid increasing NMR spectral overlap, we prepared two constructs in which stretches of either unlabeled A-U (E-AU-RNA) or unlabeled G-C (E-GC-RNA) base pairs were used for elongation in a background of uniformly 13C/15N-labeled G-C or A-U nucleotides, respectively (Fig. 1C). The two constructs allowed acquisition of NMR data over the entire RNA target while keeping elongation residues “NMR-invisible.”

We prepared (11) elongated constructs of the transactivation response element (TAR) from HIV-1 (12), a classic example of an RNA that can adaptively change its conformation and bind diverse targets (Fig. 1C) (1320). The E-AU-TAR and E-GC-TAR chemical shifts were in excellent agreement with their nonelongated TAR counterpart, both in free form and when bound to argininamide (ARG) (a mimic of the TAR cognate protein target Tat) (Fig. 1D) (fig. S1) (13). This finding, together with degenerate 1H chemical shifts observed for elongation residues (fig. S1), provided strong evidence that elongation residues adopt the expected helical structure without affecting TAR.

The spin relaxation properties of E-TAR exposed complex motional modes that were not detected in nonelongated TAR. Significant variations in resonance intensities—which, ignoring chemical exchange, report the net dynamics of a given site relative to the applied magnetic field—were observed in E-TAR. As expected, the lowest intensity indicative of overall tumbling of a well-structured helix was observed for the elongated domain I (Fig. 2A). Relative to this reference, many sites have higher intensities indicative of internal motions that are faster than overall tumbling (11). The high intensities of the bulge and neighboring residues provide evidence for a highly flexible domain-domain interface (Fig. 2A). The consistently higher intensities for base pairs in domain II relative to domain I suggest that domain II moves collectively across the flexible interface (Fig. 2A, inset). The even higher intensities for the UUCG loop indicate that it undergoes both collective and local motions (Fig. 2A). Most of these motional modes are not resolved in the intensities of nonelongated TAR (Fig. 2B). This can be attributed to couplings between domain motions and rotational diffusion as well as reduced sensitivity to internal motions due to faster overall tumbling.

Fig. 2.

RNA dynamics by motionally decoupled NMR. (A to D) Normalized resonance intensities (peak heights) measured from nonconstant-time 2D 1H-13C HSQC spectra. Shown are values for sugar C1′H1′ (diamonds) and base C2H2 (squares), C5H5 (circles), C6H6 (triangles), and C8H8 (inverted triangles) in (A) E-AU-TAR + E-GC-TAR, (B) TAR, (C) E-AU-TAR+ARG + E-GC-TAR+ARG, and (D) TAR+ARG. The intensity for each type of C-H spin is normalized to a minimum value of 0.1 independently for G-C and A-U residues. Insets show intensities for Watson-Crick residues only. The UUCG loop intensities are denoted by open symbols. (E to H) Ratios (R2/R1) of imino 15N transverse (R2) to longitudinal (R1) relaxation rates measured for guanine (circles) and uridine (diamonds) residues in (E) E-AU-TAR + E-GC-TAR, (F) TAR, (G) E-AU-TAR+ARG + E-GC-TAR+ARG, and (H) TAR+ARG. Hydrodynamically predicted R2/R1 values are denoted by open symbols.

We obtained further insight into the complex motional manifold by measuring resonance intensities in the E-TAR+ARG complex. Previous studies have shown that ARG stabilizes a coaxially aligned TAR conformation by interacting with residues at the interdomain interface (13, 14, 21). Consistent with an arrest of domain motions, ARG binding leads to a reduction in the relative intensity of most sites in domain II, including residues (e.g., UUCG loop) far removed from the ARG binding pocket (Fig. 2C). This is accompanied by a reduction in the mobilities of residues at the domain-domain interface (U23, A22, and U40) that are known to interact with one another or ARG upon complex formation (Fig. 2C) (13, 14, 21). The ARG arrest of domain motions exposes the local mobility of the UUCG loop as intensities (Fig. 2C) that correlate well (R = 0.99) with its local motional amplitudes derived independently from a previous NMR relaxation study (22). Although ARG binding induces similar intensity changes in TAR, these are significantly less pronounced and the arrest of domain motions goes completely undetected (Fig. 2D).

Independent support for domain motions in E-TAR was obtained by measuring imino 15N relaxation data (23) for guanine and uridine residues (11) (fig. S2 and table S1). The uniformly smaller ratios of transverse (R2) to longitudinal (R1) relaxation rates (R2/R1) observed for every site in domain II as compared to domain I in E-TAR confirmed the existence of domain motions that reorient every site in domain II relative to domain I (Fig. 2E) (fig. S3). In contrast, the similar R2/R1 values observed for the two hydrodynamically equivalent domains in TAR confirmed that in the absence of elongation, domain motions cannot be separated from rotational diffusion (Fig. 2F). Consistent with an arrest of domain motions, ARG binding leads to an increase in the domain II R2/R1 values in E-TAR such that they approach the values measured in domain I (Fig. 2G). In contrast, this dynamical arrest goes undetected in nonelongated TAR (Fig. 2H). The agreement between the measured E-TAR R2/R1 values and hydrodynamic calculations (24) using an extended A-form helix for domain I provides further support that elongation residues adopt the expected helical structure (Fig. 2, E and G).

Model-free analysis (10, 25, 26) of the E-TAR 15N relaxation data (11) allowed us to quantitatively resolve overall rotational diffusion, local N-H fluctuations, and domain motions (Fig. 3A) (table S2). The observed time constant for E-TAR molecular rotational diffusion (τm = 18.9 ns) is in excellent agreement with hydrodynamic predictions (τcalc = 19.5 ns). Local N-H fluctuations have effective time constants that are on the order of picoseconds (τe = 24 to 33 ps). These fast motions have amplitudes spanning Math to 0.89 (Math varies between 0 and 1 for maximum and minimum motions). The shorter domain II displays slightly larger amplitudes on average, with the largest amplitudes observed for U38, which is consistent with its high resonance intensities (Fig. 2A, inset). This motion is likely important in the formation of a U38-A27·U23 base-triple that accompanies adaptive recognition (13, 21). The N-H sites in domain II also experience slower and larger amplitude (Math to 0.76) domain motions whose time constants (τs = 1.5 to 1.9 ns) approach the hydrodynamically predicted time constant for rotational diffusion of domain II alone (τm = 2.2 ns). Thus, the two domains reorient independently of one another at their own diffusion-limited time scales across a highly unstructured interface. These domain motions occur at time scales approaching that of TAR overall rotational diffusion (τm ∼6 ns), making separation of the two modes difficult in the absence of elongation. ARG binding arrests the domain motions in E-TAR and leads to a uniform reduction in the librational amplitudes (Math to 0.95) (Fig. 3B). The time constant for E-TAR+ARG rotational diffusion (τm = 18.4 ns) is also in excellent agreement with hydrodynamic predictions (τcalc = 18.2 ns).

Fig. 3.

Resolving nanosecond and picosecond motional modes in RNA. Vertical lines in the RNA secondary structures indicate hydrogen-bonded Watson-Crick base pairs as detected by the JNN COSY experiment (28). Shown are time constants (τ) and amplitudes (S2) for rotational diffusion (blue), collective (green), and local (brown) motions in (A) E-AU-TAR + E-GC-TAR, (B) E-AU-TAR+ARG + E-GC-TAR+ARG, and (C) E-AU-SL1m + E-GC-SL1m. Black horizontal dashed lines correspond to the hydrodynamically computed time constant for overall rotational diffusion of E-RNA and domain II alone. For E-SL1m, a range of domain II time constants is shown to indicate inclusion and exclusion of base pairs (C9-G26, U10-A25, and U11-A24) that do not have detectable hydrogenbonds.

The subnanosecond motional modes observed in E-TAR involve small kinetic barriers and therefore constitute fundamental dynamics of the RNA main chain. We confirmed the generality of these motional modes by applying elongated NMR spectroscopy (11) to characterize the dynamics of a different RNA, SL1m (27), also derived from HIV-1 (Fig. 3C). Unlike TAR, the two helical domains in SL1m are linked by a purine-rich internal loop that is juxtaposed by different Watson-Crick base pairs (Fig. 3C). Despite this different linker, the resonance intensities (figs. S4 and S5) and 15N relaxation data (figs. S6 and S7 and table S1) measured in E-SL1m revealed domain motions that once again evade detection in nonelongated SL1m. Model-free analysis (10, 25, 26) of the E-SL1m 15N relaxation data yielded local librations and collective domain motions similar to those observed in E-TAR (Fig. 3C) (table S2).

Such fundamental motional modes could provide a molecular basis for RNA structural adaptation. TAR is one of the best documented examples of an RNA molecule that can adopt different conformations and thereby bind to diverse targets. The current repertoire of such “adapted” HIV-1 TAR conformations includes eight high-resolution NMR and x-ray structures of TAR in the absence of ligands (15) and bound to Tat-derived peptides (14), divalent ions (16), and five chemically distinct small molecules (1720). These TAR structures differ significantly (structures superimpose with an all-atom root mean square deviation of 4.7 Å) both in the global orientation of helical domains (interhelical angle spanning ∼5° to ∼47°) and in the local structure of the binding pocket, which comprises the bulge and neighboring residues (fig. S8).

To assess the relationship between adaptive structural changes and internal motions, we quantified the magnitude of TAR structural adaptation by computing the mean angular variation 〈Δθ 〉 in the orientation of C-H bonds across the eight structures, using domain I as a reference for superimposing structures (11). We then compared these adaptive structural changes with the free E-TAR intensities (Fig. 2A), which provided a corresponding domain I–referenced measure of internal motions. Remarkably, the magnitude of adaptive structural change observed at a given site (Fig. 4A) was quantitatively correlated (R = 0.80) to the degree of spontaneous internal motions at the site in the free RNA (Fig. 4B). Thus, a hierarchical network of local and collective internal motional modes occurring at nanosecond and faster time scales underlies RNA's ability to adaptively change conformation.

Fig. 4.

Site-specific comparison of RNA dynamics and structural adaptation. (A) Mean angular difference 〈Δθ 〉 in the orientation of sugar and base C-H bond vectors across eight different HIV-1 TAR structures (in free form and bound to seven distinct targets) after superposition of residues in domain I. The symbols denote different C-H bonds as described in Fig. 2. The UUCG loop is excluded because it is absent from the TAR structures examined. (B) Correlation plot between 〈Δθ 〉 and the corresponding free E-TAR intensities (shown in Fig. 2A). The line corresponds to a linear best fit.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S8

Tables S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article