Protein-guided RNA dynamics during early ribosome assembly

Abstract

The assembly of 30S ribosomes requires the precise addition of 20 proteins to the 16S ribosomal RNA. How early binding proteins change the ribosomal RNA structure so that later proteins may join the complex is poorly understood. Here we use single-molecule fluorescence resonance energy transfer (FRET) to observe real-time encounters between Escherichia coli ribosomal protein S4 and the 16S 5' domain RNA at an early stage of 30S assembly. Dynamic initial S4-RNA complexes pass through a stable non-native intermediate before converting to the native complex, showing that non-native structures can offer a low free-energy path to protein-RNA recognition. Three-colour FRET and molecular dynamics simulations reveal how S4 changes the frequency and direction of RNA helix motions, guiding a conformational switch that enforces the hierarchy of protein addition. These protein-guided dynamics offer an alternative explanation for induced fit in RNA-protein complexes.

Figure 1. Fluctuations during early ribosome assembly

a, Protein S4 (tan surface) bound to the 16S rRNA in the E. coli 30S ribosome (2I2P). 5WJ colored in the ribbon and 16S schematic; rest of 5′ domain, black. S12 (dark red) binds the 50S interface side of 5WJ in the mature 30S ribosome. b, Labeling positions for FRET between S4-Cy3 (green sphere) and 5′domain h3-Cy5 (red sphere). C507 in the h18 pseudoknot is in magenta. c, smFRET traces show two state fluctuations in the distance between S4 and h3. Green: Cy3 fluorescence; red: Cy5 fluorescence; blue: FRET efficiency. The trace at 4 mM Mg²⁺ exhibits frequent S4 binding and dissociation. d, FRET histograms of wild-type complex in 20 and 4 mM Mg²⁺ and mutant RNAs in 20 mM Mg²⁺. Δh18loop lacks the teal region in (a). e, Stacking interactions by A397 and A499 stabilize the h3-h18 junction in the ribosome.

Figure 2. Dynamics of S4 binding to the 5′ domain RNA

a, S4-Cy3 was added to immobilized 5′ domain RNA labeled with Cy5 at h3. b, d, Binding traces in 20 and 4 mM Mg²⁺, respectively. Arrows indicate when S4-Cy3 binds. c, e, f, Probability density maps of synchronized FRET dynamics. Histograms show FRET distribution at the moment of binding. Orange rectangles denote initial mid-FRET population. g, Binding trace at 4 mM Mg²⁺ acquired at 10 ms resolution. h, Synchronized FRET density map at 10 ms resolution. Orange rectangle denotes the broad initial FRET distribution converging to the low FRET state.

Figure 3. Kinetic pathway for S4 binding

a, Model constructed from single molecule data. b-f, [Mg²⁺]-dependence of rate constants for initial binding (b), decay of the encounter complex (EC) (c), flipped intermediate (FI) (d), and native complex (NC) (e,f) via fast (k_-2F) and slow (k_-2S) components. g, The fraction of the slow component, A_S/(A_S + A_f), when fitting the NC decay as A_Se^−k_−2St + A_Fe^−k_−2Ft. The decay rate of FI state was comparable between initial binding measurements (red) and equilibrium dynamics (black). Error bars represent the standard error of the mean from triplicate measurements (see Methods for data statistics).

Figure 4. Modulation of the rRNA dynamics by S4 binding

a, Dynamics of the free rRNA in 20 mM Mg²⁺, based on the fluctuation in the h16-Cy3 to h3-Cy5 distance. b, Three-color FRET shows that the h16-h3 dynamics is suppressed by bound S4, while S4-h3 dynamics is observed. c, Schematic model based on the observed distances and dynamics. Colors of spheres represent the labels for three-color measurements (S4-Cy3; h16-Cy5; h3-Cy7). Isotropic motions of h16 and h3 in the free RNA are denoted by black arrows. Restricted motion of h3 when S4 binds is denoted by blue arrows. d, Distances between S4, h16, and h3 from 35 stable complexes (grey dots) were projected on each face to show the correlation between each pair of distances. The majority of molecules behave as in (c) (blue rectangles); a minor misfolded population shows higher FRET for h16-h3 and S4-h16 and stable high FRET for S4-h3 (violet rectangles). e, RNA dynamics from 150 ns all-atom MD simulation with and without S4. Thin lines trace movement (from violet to yellow) of h16 (red cylinder) and h3 (blue cylinder) during the simulation.

Extended Data Figure 1. Modification of the 5′ domain RNA preserves its structure

a, The secondary structures of the wild-type and extended 5′ domain RNAs were probed by selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE). The 5′ domain RNA (2 pmol) was annealed to unlabeled oligonucleotides and folded in HKM20 buffer (80 mM K-HEPES pH 7.6, 300 mM KCl, 6 mM 2-mercaptoethanol, 20 mM MgCl₂) before treatment with 3 mM N-methylisatoic acid (NMIA) at 42 ºC for 26 min. Modifications were detected by primer extension and quantified as previously described. Results of SHAPE chemical probing of the free RNA structure for the wild-type 5′ domain (left) and the 5′ domain with h16 and h3 extensions after annealing with oligonucleotides (right). Saturation of grey indicates reactivity with NMIA (see Methods). Dashed circles indicate nucleotides that were not detected in our primer extension assay. The results show that the extensions added for the fluorescent labeling of the rRNA do not significantly perturb the rRNA folding. b, Native PAGE folding assay of 5′dom-h3h16. Fluorescently-labeled oligonucleotide (h3P5-Cy5; 25 nM) was annealed to an equimolar concentration of extended 5′ domain RNA in 10 μL CE buffer (20 mM Na-cacodylate, 0.5 mM Na₂EDTA) for 5 min at 70 ºC and 5 min at 25 ºC. The RNA-oligonucleotide complex was then folded at 37 ºC for 30 min in varying [MgCl₂] (0–20 mM) before electrophoresis on a native 8% polyacrylamide gel containing 10 mM MgCl₂. The folding midpoint was 0.5 ± 0.1 mM MgCl₂, similar to that of the wild-type 5′ domain RNA (1.4 ± 0.2 mM) reported previously.

Extended Data Figure 2. Annealing of labeled oligonucleotides to the 5′ domain RNA

³²P-labeled oligonucleotides were annealed to the 3′ extension of h3 of 5′dom-h3 (h3P5, DNA) (a) or to the extended loop of h16 of 5′dom-h3h16 (h16P2-2, RNA) (b). Annealing reactions were performed in HK buffer and 6 mM 2-mercaptoethanol at 25 ºC. Binding data were fit to the quadratic form of a two-state binding isotherm. Apparent dissociation constants were ≤0.4 nM and 2.6 ± 0.2 nM for h3 and h16 oligonucleotides, respectively. Equilibrium constants are the average and S.D. of two or more independent trials. The lengths of the labeled oligonucleotides were varied to optimize affinity with the extended 5′ domain RNA, while avoiding perturbations to S4 binding (see Extended Data Fig. 1).

Extended Data Figure 3. S4 labeling and its binding to the rRNA

E. coli ribosomal protein S4 was over-expressed, purified, and labeled with Cy3 or Cy5 fluorescent dyes as described in Methods. a, SDS-PAGE of unlabeled protein stained with Coomassie (left) or labeled with Cy5 (right). b, c, Ensemble titration of the modified 5′ domain RNAs in HKM20 shows that S4-Cy5 binds with similar affinity as the wild type S4-rRNA complex. Extended 5′ domain RNAs annealed with h3P5-Cy3 and/or h16 oligonucleotides were titrated with S4-Cy5 in a 500 μL cuvette, and the fluorescence emission was recorded from 550 to 700 nm with 540 nm excitation (Fluorolog-3, Horiba). Excitation and emission slits were fixed at 2 nm and 5 nm, respectively. The sample was incubated at 37 ºC for 1 min after each addition. Two or more independent measurements were averaged and titration curves were fitted to a quadratic binding expression. Equilibrium dissociation constants were 5′dom-h3, 0.11 ± 0.02 nM (statistical error of the fit parameter) and 5′dom-h3h16, 0.2 ± 0.1 nM, at 37 ºC, and were comparable to that of the 5′ domain RNA with wild-type E. coli S4 (0.9 nM).

Extended Data Figure 4. Exchange kinetics of docked and flipped complexes

a, Sample FRET traces are shown for mutant rRNAs in 20 mM Mg²⁺. b, The cumulative histograms of the dwell times in the high and low FRET conformations were calculated for wild-type 5′ domain RNA, C507G mutant, and Δh18loop mutant, and fit with both mono-exponential and bi-exponential decay functions. One of the triplicate sets of data is demonstrated for each (refer to Methods for number of traces). Significantly lower χ² values suggest the data are best fit with two exponential terms, except for transitions from the low FRET state to the high FRET state of the wild-type complex, which was well fit by a single exponential decay function. c, d, All of the dwell time histograms were fit with bi-exponential decay and the fitting parameters were compared between the wild-type and the mutants. Both components of the transition from the high to low FRET state were faster in the mutants than in the wild-type complex. The lifetime of the wild-type low FRET state had a single component; for the mutants we observed an additional slow component in the lifetime.

Extended Data Figure 5. Sample traces of S4 binding and binding trials

a, In 20 mM Mg²⁺, the binding occurred mostly at the low FRET state. Infrequently, we observed the dissociation and secondary binding of S4. b, At 4 mM Mg²⁺, we often observed unsuccessful and transient binding of S4. Arrows indicate the transient fluorescence signals from unstable binding. These traces also exhibit the mid FRET spike at the beginning of successful binding event. c, The portion of the molecules that form a stable complex on the first try to the total molecules that form stable complex within 5 min was plotted with varying [Mg²⁺]. The error bars represent the 95% confidence interval assuming a binary distribution. The number of molecules used was 76, 81, 318, and 170 for 2, 4, 10, and 20 mM Mg²⁺, respectively.

Extended Data Figure 6. Progression of FRET population at different [Mg²⁺]

From the synchronized maps of FRET distribution as shown in Fig. 2, the relative populations of low (0–0.35), mid (0.35–0.55), and high (0.55–0.9) FRET states were plotted at different [Mg²⁺]. In 20 and 10 mM Mg²⁺, the bound complexes started with large low FRET population that converted to high FRET population within 5–10 s. At 4 and 2 mM Mg²⁺, there were considerable mid and high FRET populations in the beginning, reflecting the broad initial FRET distribution (Fig. 2e, f). This quickly converted to the low FRET population, which was then followed by slow conversion to the high FRET population within several seconds. The number of molecules used was 112, 239, 116, and 275 for 20, 10, 4, and 2 mM Mg²⁺, respectively.

Extended Data Figure 7. Sample traces at 10 ms time resolution

Single molecule traces at higher time resolution demonstrate the heterogeneous and fluctuating behavior of the encounter complex (S4-Cy3 and 5′dom-h3-Cy5). The change of FRET in different molecules cannot be described as a single behavior. In general, the initial FRET distribution over the complexes is broad and the FRET signal converges to the relatively stable low FRET prior to the transition to the high FRET.

Extended Data Figure 8. Dynamics of free 5′dom-h3h16

a, Schematic of extensions in h16 and h3 with labeled oligonucleotides. b, Sample FRET traces showing the fluctuation between two distinct states. The frequency and distribution of these fluctuations varied between molecules. c, Mg²⁺ dependence of the molecular heterogeneity. Histograms show the relative high FRET population for each molecule (157, 162, and 74 traces for 20 mM, 5 mM, and 1 mM Mg²⁺, respectively). At higher [Mg²⁺], more molecules stay in the high FRET state for longer periods of time. High FRET between h16 and h3 does not necessarily correspond to the native structure of the 5′ domain RNA in complex with protein S4 that is represented by high FRET between S4 and h3.

Extended Data Figure 9. Examples of switching between different h16-h3 dynamics

Arrows indicate when the fluctuation dynamics switches between stable high FRET, stable low FRET, and alternating high and low FRET behaviors. The RNA was labeled as in Extended Data Figure 8.

Extended Data Figure 10. S4 binding trajectories from hybrid MD-Gō simulations

a, Simulated FRET between S4 and h3 from representative binding trajectories displays various binding pathways. b, Density map constructed from 61 successful binding trajectories. The trajectories were synchronized at the moment when the first native contact between S4 and the 5WJ is established (dotted lines in (a)). c, Sample trajectories of successful binding, showing how folding of h16 and h3 is induced by S4 binding.