Global stabilization of rRNA structure by ribosomal proteins S4, S17 and S20

Summary

Ribosomal proteins stabilize the folded structure of the rRNA and enable the recruitment of further proteins to the complex. Quantitative hydroxyl radical footprinting was used to measure the extent to which three different primary assembly proteins, S4, S17 and S20, stabilize the 3D structure of the E. coli 16S 5′ domain. The stability of the complexes was perturbed by varying the concentration of MgCl₂. Each protein influences the stability of the rRNA tertiary interactions beyond its immediate binding site. S4 and S17 stabilize the entire 5′ domain, while S20 has a more local effect. Multi-stage folding of individual helices within the 5′ domain shows that each protein stabilizes a different ensemble of structural intermediates, that include non-native interactions at low Mg²⁺. We propose that the combined interactions of S4, S17 and S20 with different helical junctions bias the free energy landscape toward a few RNA conformations that are competent to add the secondary assembly protein S16 in the next step of assembly.

Ribosome biogenesis is one of the most important synthetic tasks undertaken by the cell. Among bacteria, short generation times require that several hundreds of ribosomes are produced per minute in order to produce the tens of thousands of ribosomes needed by each new cell ^1;2. A critical question is how the RNA and protein components of the ribosome interact to achieve the native subunits and avoid nonfunctional complexes.

Studies on the reconstitution of the 30S ribosomal subunit have shown that binding of ribosomal proteins is coupled to folding of the rRNA ^3;4. The addition of the primary assembly proteins, which bind directly to the 16S rRNA, stabilize the helices with which they interact, and enable further assembly of each 30S domain by pre-ordering the binding sites for secondary and tertiary assembly proteins ^5;6. The extent to which individual ribosomal proteins stabilize the structure of the rRNA beyond their immediate binding site is thus directly connected to the hierarchy and cooperativity of assembly ^4;7.

The 5′ domain of the 16S rRNA is the first to be transcribed in vivo and forms a separate tertiary structure that makes up the body and spur of the small subunit (Fig. 1) ^8;9. Three primary assembly proteins S4, S17 and S20 each recognize and bind a different helix junction within the 5′ domain (Fig. 1a). Protein S17 binds helices 7 and 11 near the central junction ¹⁰. Its flexible loops extend from a β-barrel ¹¹ to contact different regions of the rRNA. Amino acids 13-17 in S17 stabilize a K-turn in helix 11, while aa 29-34 thread between helix 11 and helix 21 to reach the 560-region at the junction with the central domain ¹². Mutations in aa 67-68, which stitch together helices 7 and 11, result in defective 30S assembly ¹³. Protein S20 forms a three-helix bundle that interacts with the lower junction between helices 7-10 and the loop of helix 13 ^12;14. S20 also contacts helix 44 in the 30S subunit, helping to correctly position the 3′ minor domain¹⁴.

Figure 1. Protein interactions in the E. coli 5′ domain.

(a) Secondary structure of the 16S rRNA with 5′ domains nucleotides 21-562 in black. Schematic of protein interactions is redrawn from Ref. 12. (b) Structure of the 5′ domain in the E. coli 30S ribosome (2avy; ⁶⁰), with S4 in pink, S17, green, S20, yellow. Ribbon made with nuccyl (L. Jovine) and Pymol (Delano Scientific).

Protein S4 binds the junction between 5 helices (3, 4, 16-18) in the “upper” part of the 5′ domain ^15;16;17, and makes additional interactions to helix 17 (A496), helix 1 (A9) and the tip of helix 21 ¹² (Fig. 1a). S4 is one of two 30S proteins required for nucleation of subunit assembly ¹⁸. Consistent with this role, a comparison of base chemical modification and backbone protections suggested that S4 strongly influences the structure of the 16S rRNA outside its immediate binding site ¹⁹. Temperature-dependent changes in the conformation and stability of the 16S rRNA provided evidence for reorganization of the S4-16S complex ^20;21. Reorganization of S4 complexes may be linked to refolding of the N-terminal domain of S4, which is disordered in the free protein ²². By contrast, there is no evidence for temperature-dependent reorganization of the S17 and S20 complexes ²³.

We wished to understand in more detail how the three primary binding proteins in the E. coli 16S 5′ domain contribute to assembly of the 30S ribosome. To make the system more amenable to quantitative analysis, we used a transcript that contains only the E. coli 16S 5′ domain. The 5′ domain RNA forms a stable complex with proteins S4, S16, S17 and S20 ²⁴. Proteins S12 and S5 also interact with the 5′ domain residues within the 16S rRNA at a later stage of 30S assembly ^19;25, but do not stably bind the 5′ domain complexes ²⁴.

The 5′ domain RNA forms all of its predicted tertiary backbone contacts in 20 mM MgCl₂ in the absence of proteins ²⁶. However, tertiary interactions in helices 15 and 17 require more than 5 mM magnesium to form completely, while other helices are protected less strongly in the naked RNA than in native 30S ribosomes, suggesting that the RNA tertiary structure is dynamic in the absence of proteins. By contrast, the RNP containing ribosomal proteins S4, S16, S17 and S20 assembles cooperatively in less than 5 mM Mg^{2+ 27}. Thus, the 5′ domain proteins stabilize the folded rRNA under physiological conditions.

Previous footprinting experiments on the E. coli 16S 5′ domain showed that assembly passes through at least two equilibrium intermediates when S4, S17 and S20 are bound. In the ‘native-like’ intermediate, the core of the 5′ domain RNA including helix 15 has the same structure as in the mature 30S subunit, but helix 3 is out of contact with helices 18 and 12. In the ‘non-native’ intermediate, helix 3 packs against helix 18, but the core around helix 6a is misfolded ²⁷. The secondary assembly protein S16 preferentially stabilizes the native-like intermediate in 1 mM MgCl₂, smoothing the pathway of assembly. This leads to a conformational switch in 3-5 mM MgCl₂ that repositions helix 3 between helices 18 and 12 and stabilizes the helix 18 pseudoknot (or 530-loop). Thus, the ribosomal proteins not only stabilize rRNA tertiary interactions but can also change the path of assembly, avoiding unproductive RNA conformations.

To understand how individual proteins in the 5′ domain stabilize the rRNA tertiary structure and shape the early steps of assembly, we probed the stability of the RNA tertiary interactions in single protein complexes using hydroxyl radical footprinting. The extent of hydroxyl radical cleavage correlates with the solvent accessibility of the RNA backbone ^28;29, which in turn reflects the sum of folding equilibria that expose or protect an individual ribose. In general, we expect the RNA tertiary interactions to form in less Mg²⁺ if the bound protein stabilizes the RNA structure, because stable RNAs typically require less Mg²⁺ to fold than unstable RNAs ^30;31. A correlation between RNA stability and protein binding has been observed for other proteins that bind structured RNAs, such as group I ribozymes ^32;33;34. Therefore, under equilibrium conditions, the concentration of Mg²⁺ required to protect each segment of the RNA reveals how the protein changes the folding free energy of the rRNA. It also provides information about the structures of the most stable (or highly populated) assembly intermediates.

Here, we report that proteins S4 and S17 alone stabilize RNA interactions throughout the the 5′ domain of the 16S rRNA. Thus, individual primary assembly proteins increase the global stability of the 16S 5′ domain. However, tertiary interactions distant from the bound protein form less cooperatively than those that are in direct contact with the protein. Comparison with our previous results on complexes containing all three proteins and S16 suggests that non-native intermediates are preferentially depopulated as more proteins join the complex. Such synergistic interactions between RNA binding proteins may be an important mechanism for increasing the fidelity of RNP assembly.

Results

Magnesium Dependence of S17 binding

We first sought to confirm that the 30S 5′ domain proteins interact more strongly with the folded rRNA, by investigating the Mg²⁺-dependence of binding under the conditions of our experiments. Geobacillus stearothermophilus S4 was previously shown to bind the 16S rRNA more tightly in 4-8 mM MgCl₂ than in KCl alone, consistent with its binding to a folded form of the rRNA ²¹. We also found that S4 can bind a minimal rRNA substrate containing the five helix junction comprising its binding site ³⁵. The S4 binding site is more preorganized in the minimal RNA than in the 5′ domain RNA (16S nt 21-562), and this correlates with slightly stronger affinity for the minimal RNA ³⁵.

To determine whether Mg²⁺ also increased the affinity of S17 for the 5′ domain RNA, nitrocellulose filter binding assays were performed in 0 to 30 mM Mg²⁺. When the Mg²⁺ concentration was ≤ 2 mM, the maximum extent of binding was 20% (Figure 2a). Above 2 mM MgCl₂, filter retention increased up to 55% and the apparent K_d dropped from 31 nM in 3 mM MgCl₂ to 8.3 nM in 10 mM MgCl₂ (Figure 2b). Non-ribosomal RNAs such as polyA or tRNA were unable to compete with the radiolabeled complex, indicating that the interaction with the 5′ domain is sequence-specific. We were unable to measure the dependence of S20 binding on Mg²⁺ because of poor retention on nitrocellulose filters and a small gel mobility shift.

Figure 2. Mg²⁺-dependent binding of protein S17.

Protein S17 was incubated with 1 nM 5′ domain at 37°C in 0-30 mM MgCl₂ (see Methods). (a) Fraction RNA retained on nitrocellulose filters versus S17 concentration. Circles, 30 mM MgCl₂; triangles, no MgCl₂. Symbols and error bars represent the mean and standard deviation of three trials. (b) Equilbrium dissociation constants in 3-30 mM Mg²⁺ concentration. Less than 20% RNA was bound in 0 and 2 mM MgCl₂.

We next tested different 16S fragments in competitive binding assays to determine what regions contribute to S17 binding. Dissociation constants determined by competition of ³²P-labeled 5′ domain RNA with either unlabeled 16S rRNA or 5′ domain RNA were similar (K_d ≈ 6 nM; data not shown). Thus, regions of the 16S rRNA outside the 5′ domain do not contribute significantly to the stability of the S17 complex. An rRNA fragment that includes the S17 binding site in helices 7 and 11 (nt 111-313) was bound less strongly by S17, however. In direct titrations with this rRNA fragment, half the RNA formed stable S17 complexes with a K_d = 12 nM, while the other half formed unstable complexes (K_d > 400 nM; data not shown). Because the minimal RNA contains nearly all the residues that directly contact S17, these results suggested that other helices in the 5′ domain contribute to S17 binding, perhaps by correctly orienting helices 7 and 11. Thus, the S17 (and S4) complexes are influenced by a wide network of rRNA interactions that are Mg²⁺-dependent. Conversely, we expected protein binding to have a similarly distributed effect on folding of the 5′ domain.

Proteins stabilize the rRNA tertiary structure

To determine whether single ribosomal proteins stabilize the structure of the 16S 5′ domain outside their immediate binding sites, complexes with S4, S17 or S20 were probed by hydroxyl radical footprinting in 330 mM KCl and 0 to 20 mM MgCl₂ (Fig. 3a and Methods). The difference between these experiments and previous footprinting experiments ^16;20 is that the Mg²⁺ concentration was varied to perturb the folding free energy of the RNA, allowing the stabilities of various complexes to be compared.

Figure 3. Fe(II)-EDTA footprinting of single protein complexes.

Complexes with the 16S 5′ domain were formed at 37°C in 0-35 mM MgCl₂ before cleavage with hydroxyl radical (see Methods). (a) Representative data with S17. Lanes N, no treatment; H, RNA in 80 mM K-Hepes; K, lane H plus 330 mM KCl; A, G, sequence ladders; 30S, native 30S ribosomes. Extension with primer annealing at nt 433 is shown. Protections explained by predicted RNA-RNA contacts in the 30S subunit are shown. (b-d) Fractional saturation of backbone protection (Y) versus Mg²⁺ concentration, relative to RNA in Hepes (Y = 0) and native 30S ribosomes (Y = 1). The data were fit to two or four-state models (Methods). Black, RNA only; pink, +S4, green, +S17; yellow, +S20. (b) nt 288-290 (helix 11); (c) nt 481-483 (helix 17); (d) nt 501-502 (helix 18).

The footprinting results on individual protein complexes were compared to previous footprinting experiments on the naked 5′ domain RNA under the same conditions ²⁷. The extent of protection was normalized to control reactions on the RNA in 80 mM Hepes (unfolded) and native 30S subunits (completely folded). Many residues throughout the 5′ domain were protected in less Mg²⁺ when bound by a ribosomal protein than in the naked RNA (Fig. 3b-d). Thus, each primary binding protein stabilizes the folded structure of the 5′ domain to a large extent.

The domain-wide effects of each protein on the stability of the tertiary interactions can be appreciated from histograms for Mg²⁺-induced protection of the RNA backbone (Fig. 4). In the absence of protein, most folding midpoints are above 13 mM MgCl₂ (Fig. 4a). (The 5′ domain RNA is less well folded in 330 mM KCl than in 120 mM NH₄Cl, which we used in a previous study ²⁶.) In the presence of S4, however, only 10% of nucleotides require this much Mg²⁺ to be protected (Fig. 4b). Overall, S4 had the most dramatic effect on the stability of the folded RNA, with S17 and 20 being somewhat less stabilizing (Fig. 4c,d). S4, S17 and S20 together stabilized the RNA more than each protein alone (Fig. 4e), while the addition of S16 to the complex allowed nearly all the RNA interactions to form in low Mg²⁺ (Fig. 4f).

Figure 4. Protein-dependent perturbation of rRNA tertiary interactions.

The midpoints for Mg²⁺-dependent saturation of individual contacts along the RNA backbone were sorted into 0.3 mM bins. For residues with two transitions, the highest midpoint was used. Individual parameters are listed in Table S1. Black, RNA only; pink, S4; green, S17; yellow, S20; purple, S4+S17+S20; blue, S4+S16+S17+S20. Data for three and four protein complexes are described elsewhere ²⁷.

Global stabilization by protein S4

To determine which interactions are stabilized by each primary binding protein, the midpoints for protection of the RNA backbone in Mg²⁺ were projected onto the 2D and 3D structures of the 16S 5′ domain within the 30S ribosome (Figure 5). In the absence of protein, a stable core of protein-independent RNA interactions were observed adjacent to the upper five helix junction, the central junction (helix 11a) and along the interface between helices 6 and 7 (Figure 5a; ²⁷). The functionally important pseudoknot in helix 18 was partly structured in the naked rRNA, but the pattern of backbone protection was inconsistent with the structure of the pseudoknot in the 30S ribosomes ^8;9.

Figure 5. Stabilization of rRNA tertiary structure by ribosomal proteins.

Protected regions are colored according to the midpoint (C_m) of the folding transition: red, 0-2.3 mM; orange, 2.3-4.9 mM; green, 4.9-13.4 mM; blue, >13 mM. Residues having more than one transition are clustered according to the one at highest Mg²⁺. 2D schematics and 3D ribbons as in Figure 1. See Supplemental Figure S1 and Table S1 for further data. (a) 5′ domain RNA only; (b) RNA plus S4; (c) RNA plus S17; (d) RNA plus S20. Data in (a) are from ²⁷.

As expected, S4 stabilized tertiary interactions around the five helices that comprise its binding site ¹⁹ (Figure 5b). For example, tertiary interactions between helices 16 and 18 (nt 427-429) were 80% protected in the presence of S4, but unprotected with S17 or S20 (Figure S1). However, the Mg²⁺ titrations showed that tertiary interactions in the lower half of the 5′ domain are also stabilized indirectly by binding of S4, accounting for its importance in 30S assembly ¹⁸. Binding of S4 allowed the formation of RNA tertiary contacts in helices 10, 11, 12, 15 and 17, which are not stable in the naked RNA (e.g., Figure 3c,d and Figure S1).

In addition, S4 binding enhanced hydroxyl radical cleavage of nt 78-80 in helix 6 relative to the naked RNA (data not shown). Helix 6 forms a spur which protrudes from the base of 30S subunit ^8;9. Cleavage of helix 6 is enhanced even more when other proteins are added to the complex, suggesting that the spur becomes increasingly exposed to solvent as the helices in the 5′ domain become more structured. This observation supports the idea, discussed below, that the ensemble of RNA conformations becomes smaller as more proteins join the complex.

Global stabilization by protein S17

Protein S17 stabilized RNA tertiary interactions around its binding site in helices 7 and 11 (Fig. 5c), including the central junctions between helices 6, 6a, 7, 11 and 12, and a kink turn in helix 11 that allows the tip of helix 11 to dock against helix 7 (Figure 3b and Figure S1). S17 indirectly stabilized the lower junction between helices 7, 8, 9 and 10 that overlaps the binding site for S20. Thus, binding of these two proteins may be weakly cooperative. S17 also perturbed RNA interactions in helices 15, 17, and 18, which are distant from its binding site (Fig. 5c). This was unexpected, because, unlike S4, S17 has not been implicated in the nucleation of 30S assembly.

Regional stabilization by protein S20

In comparison with S4 and S17, specific effects of protein S20 were more localized to nucleotides surrounding its binding site between helices 6, 7, 8, 9, 11, and 13 (Figure 5d). Although many nucleotides were partially protected in low Mg²⁺, 40% of protected nucleotides have a Mg²⁺ midpoint for complete saturation of backbone contacts that is higher than 13 mM, compared with 10% for S4 (Fig. 4d). S20 failed to stabilize tertiary interactions in the upper half of the domain where S4 binds (helices 15-18; Fig. 5d and S1). Thus, S20 has a more modest influence on the overall structure of the 5′ domain than S4 and S17. The footprinting results are consistent with previous base probing studies and cleavage by tethered Fe(II)-EDTA complexes, which show that S20 interacts with helices at the bottom of the 5′ domain ³⁶.

Single proteins stabilize partially folded intermediates

For each single protein complex, the pattern of hydroxyl radical protection in 20-30 mM Mg²⁺ was largely consistent with the conformation of the native 5′ domain RNP expected from crystal structures of 30S subunits ^8;9, although the pseudoknot in helix 18 and the tip of helix 5 (nt 364) did not fold completely. Many residues, however, were semi-protected from hydroxyl radical cleavage in intermediate concentrations of Mg²⁺ (Figure 3).

Comparison of the footprinting results on single protein complexes and the naked RNA showed that the primary binding proteins perturb the structure of the 5′ domain RNA in very low Mg²⁺. Thus, the proteins must interact with the 5′ domain even though they do not form stable complexes under these conditions, presumably because the RNA is not stably folded. Each protein greatly increased protection of helices 6, 7, and 8 in 0.01 mM Mg²⁺, relative to the RNA alone (Figure S1). However, these interactions did not necessarily progress to the fully native domain. For example, when S4 was present, residues in helices 6 and 7 were 40-60% protected in 0.01 mM MgCl₂ relative to 30S controls. However, these backbone protections did not saturate, even when the MgCl₂ concentration was raised to 30 mM (Table S1). These results suggested that S4 stabilizes RNA conformations that never achieve the native structure. By contrast, when S17 was bound, the same core helices became fully protected in 20 mM MgCl₂ (Table S1 and Figure S1). Thus, S17 may stabilize the native conformation of the core helices more selectively than S4.

Multi-stage folding of the 5′ domain RNA

A comparison of titration curves in different regions of the 5′ domain provided further evidence that each protein stabilizes an ensemble of intermediates which can include both native and non-native RNA structures. Not only did different regions of the rRNA fold in different Mg²⁺ concentrations, but many nucleotides folded in two or three stages (Figure 3b-d). For example, protection of nt 481-483 in the presence of S4 goes from 30% to 50% in 0.1 mM MgCl₂ and then up to 100% between 2 and 20 mM MgCl₂ (Figure 3c). Residues in helices 15 and 17, which extend down from the upper junction to pack against the core of the 5′ domain, exhibited multiphasic folding in all three of the primary binding proteins (Figure S1). The lower junction between helices 7, 8, 9 and 10 also folded in two distinct stages, particularly when S4 was bound. S17 induced two-stage folding of helices 5, 6 6a, 13 and 14, which all surround the central junction.

Interestingly, S4 and S17 had opposite effects on the kink in helix 17 (nt 481-485) that packs against a C-loop in helix 15 in the mature 30S subunit ^8;9. When S4 was present, nt 481-483 was protected in several phases while nt 484-485 remained unprotected (Figure S2). When S17 was present, nt 484-485 were 30% protected in low Mg²⁺ and 100% in 5 mM MgCl₂, while nt 481-483 were only lightly protected in 20 mM Mg²⁺. These results illustrate how each protein selectively stabilizes specific conformations of the rRNA.

Transient unfolding of the 5′ domain RNA

In addition to the multi-stage increase in protection described above, we observed transient exposure or unfolding of certain residues during the Mg²⁺ titration. The most dramatic example is the partial protection of helix 18 in no Mg, exposure in 2-10 mM MgCl₂, and reprotection in ≥ 20 mM MgCl₂, when S17 is bound to the RNA (Figure 3d, Figure S1, S3). These results suggest that S17 stabilizes an intermediate that is remodeled during folding of the 5′ domain. As discussed below, these results are consistent with a previously proposed conformational switch at helix 3 during assembly of the 16S 5′ domain ²⁷.

We also observed slight unfolding and refolding of the lower helix junction in the presence of S4, S17 and S20. Transient exposure of C217 in 0.5 mM MgCl₂ was clearly supported by the footprinting data with S20, and to a lesser extent, with S4 and S17 (Table S1, Figure S1). C217 lies in helix 10 and contacts the tip of helix 17. Other residues that may become transiently exposed are nt 196-200 on the opposite side of helix 10 which contacts helix 6, nts 66-72 (helix 6), nt 151-153 (helix 8), nt 327 (helix 13) and nt 379-380 (helix 15), although the data in low Mg²⁺ for these residues are noisy relative to the small degree of exposure. All of these residues cluster around helix 6 at the “bottom” of the 5′ domain. Thus, S20 likely stabilizes more than one conformation of the 5′ domain RNA. The idea that the core and the lower junction can adopt a conformation that is different from the one in the 30S ribosome is also supported by the protection of nt 94 in helix 6, which is accessible to solvent in the 30S ribosome.

Discussion

Stabilization of RNA structure by RNA-binding proteins

By using hydroxyl radical footprinting to measure the stability of individual tertiary contacts, we find that ribosomal proteins S4 or S17 stabilize the rRNA structure far beyond their previously defined binding sites ¹⁹, while the effects of S20 are more local. Indirect stabilization of entire RNA domains by a single protein has been observed in several large ribozymes. For example, the yeast CBP2 protein is required for splicing of the mitochondrial bI5 group I intron in physiological conditions ³⁷. The naked bI5 intron requires 40-50 mM MgCl₂ to fold and be active ^38;39. Similarly, other group I splicing factors ^34;40 and the protein component of bacterial RNase P ^41;42 enhance the catalytic activity of their cognate ribozymes by stabilizing the active conformation of the RNA or by interacting with substrates.

Like the ribozyme examples above, the 5′ domain of the 16S rRNA is rich in RNA interactions, and capable of folding independently in high salt and Mg^{2+ 26}. Thus, the tertiary structure of the 16S 5′ domain is itself sufficiently cooperative that a single protein can stabilize the entire domain. The intrinsic stability of the 16S 5′ domain also helps explain why three of the 5′ domain proteins bind the 16S rRNA without requiring other 30S proteins. As discussed below, however, S4, S17 and S20 each play a distinct role in stabilizing specific tertiary structures within the 5′ domain. Consequently, the three proteins together have the capacity to increase the specificity of the assembly pathway at physiological Mg²⁺ concentrations by constraining the structure of the rRNA.

Primary binding proteins stabilize different RNA intermediates

The changes in RNA backbone accessibility which are revealed by the footprinting data show that each protein stabilizes the tertiary structure of the 16S rRNA in different ways. Protein S4, which binds the upper five helix junction, strongly stabilizes the tertiary interactions in the upper half of the domain, and by extension, interactions between helices 15 and 17 with helices 6, 6a and 7 in the lower core of the domain (Figure 6a,b and Figure S2). Conversely, S4 stabilizes interactions with helices 5, 12 and 13 only partially.

Figure 6. Primary binding proteins stabilize different rRNA intermediates.

In low Mg²⁺, the 5′ domain forms an ensemble of partly folded states (I_C) with a subset of stable tertiary interactions. Individual proteins stabilize partly folded RNAs that may contain non-native interactions. Helices are shown as cylinders (red, most stable; gold, medium stable; green, less stable; blue, disordered). (a) Schematic showing location of helices in the 5′ domain secondary structure; (b) S4; (c) S17; (d) S20.

In contrast, protein S17 strongly stabilizes the tertiary interactions around its binding site in helices 7 and 11 and around the central junction, so that these interactions form at very low (0.01 mM) Mg²⁺ in S17 complexes (Figure 6c). Some interactions in the upper helix junction (such as with helix 16) are not stabilized at all by S17 (Figure S2e), while those in the lower junction are moderately stabilized. The footprinting data also indicate that S4 and S17 stabilize different conformations of helix 17, where it bends to make a key tertiary contact with helix 15.

S20 has the strongest effect on helices 8 and 14 that surround its binding site, so that tertiary interactions between these helices appear at much lower Mg²⁺ than in S4 and S17 complexes (Figure 6d). S20 also stabilizes tertiary interactions in helices 6, 7, 11 and 12, but not tertiary interactions in the upper half of the domain.

Ribosomal proteins stabilize more than one RNA conformation

The Mg²⁺-dependence of the footprinting results shows that each ribosomal protein forms one or more intermediate complexes with the rRNA. Not only do different residues fold at different Mg²⁺ concentrations, but many residues fold in two or three stages or become transiently exposed to cleavage. Multi-stage folding can be explained by a competition between intermediates with different structures, whose populations change as the Mg²⁺ concentration is raised ²⁷. The observed “footprint” represents the weighted average solvent accessibility of the populated intermediates ⁴³. A plateau in the level of backbone protection occurs when two intermediates have similar stabilities at a particular Mg²⁺ concentration, but different susceptibilities to cleavage by hydroxyl radical.

Protein binding lowers (or raises) the free energies of some RNA conformations but not others, driving the RNP population into newly stabilized states. Thus, in this model, the proteins direct the path of assembly by altering the free energy landscape for RNA folding ^27;44;45. This type of statistical mechanical model has been used to interpret the effects of ligands on the rates of hydrogen exchange in proteins ⁴⁶. As we have previously noted ²⁷, multi-stage protection and transient exposure of specific residues all point toward the presence of non-native structures that are remodeled during assembly of the 16S 5′ domain RNP.

The presence of equilibrium assembly intermediates agrees with data on the kinetics of assembly. Time-resolved footprinting of the 16S 5′ domain showed that the less stable tertiary contacts in helices 15 and 17 form slowly in the absence of proteins, indicating that about half the RNA population is kinetically trapped in metastable intermediates ²⁶. The kinetics of 30S reconstitution also showed that assembly occurs in stages with individual complexes taking different routes to the final structure ^44;47. While primary assembly proteins such as S4 contact their binding sites in 100 ms or less, other tertiary interactions in the 5′ domain form in 1-100 s⁴⁷.

Specificity of assembly

An important observation is that the proteins perturb the 5′ domain structure even at very low Mg²⁺ concentrations, in which the RNA is poorly folded. For example, S17 changes the footprint of the 5′ domain in less than 0.1 mM MgCl₂, yet a stable complex needs 4 mM MgCl₂ (Figure 2). S4 enables indirect protection of core helices in less Mg²⁺ than protection of helices 16 and 18 within its own binding site. Thus, the ribosomal proteins are not merely capturing the natively folded rRNA, but interacting with early folding intermediates that are structurally disordered or dynamic.

In agreement with this observation, CBP2 protein has been shown to interact with a near-native collapsed state of bI5 RNA and drive it into the native state ^32;48, and single molecule FRET experiments showed that non-specific binding increases the structural dynamics of the bI5 RNA ⁴⁹. Single molecule FRET experiments on other large ribozymes such as RNase P and a group II intron have uncovered multiple folding pathways for each RNA ^50;51;52;53, consistent with fluctuations between different I-states. Time-resolved footprinting of the 16S rRNA suggests a similarly diverse set of conformational states are available to the 5′domain RNA during assembly of the 30S ribosome ⁴⁷.

Synergy between primary assembly proteins

Although each 5′ domain protein stabilizes rRNA tertiary interactions far from its binding site, they each bind a different helix junction, and favor different folding intermediates. Consequently, we expect fewer RNA conformations to be allowed when all three primary assembly proteins are bound. This is exactly what we find when S4, S17 and S20 are added to the 5′ domain RNA before hydroxyl radical footprinting ²⁷. When three proteins are present, the pattern of backbone cleavage and protection becomes more coherent in low Mg²⁺, leading to pre-organization of tertiary interactions within the S16 binding site ²⁷.

Thus, the combined interactions of the primary binding proteins can make assembly more specific by driving the RNA population into a smaller number of structures that are competent for subsequent steps of 30S assembly. A similar synergistic effect may exist between S8 and S15 in the central domain of the 16S rRNA, as tethered hydroxyl radical footprinting experiments showed that the structural environment of S15 changes when S8 is bound, even though both proteins bind independently to separate helical junctions ⁵⁴.

With three proteins bound, we still observe a competition between native-like and non-native configurations of the core helices, that is only resolved in favor of the native-like intermediate when S16 is added to the complex ²⁷. In this context, ‘native-like’ only means that the conformation is similar to that in the mature subunit; we do not yet know whether such intermediates form naturally in the cell. Formation of the native-like intermediate correlates with displacement of helix 3 from its normal position between helix 18 and the tip of helix 12. Raising the Mg²⁺ a bit further flips helix 3 back into place ²⁷, stabilizing the helix 18 pseudoknot which forms part of the 30S decoding site.

This conformational switch at helix 3 requires coordination between residues in the core (helices 6, 8, 10, 11, 13), the central junction (helices 5, 6a, 12), and the upper junction (helices 3, 17 and 18). Although these structural changes are observed in single protein complexes, they are weakly coupled. Thus, S4, S17 and S20 in combination may stiffen the switch by discriminating against unproductive rRNA conformations.

Conclusions

In summary, these studies reveal that 5′ domain assembly involves a diverse landscape of conformational states that are bound and stabilized by single proteins. Understanding how the dynamics of competition between alternative conformations produce more stable and presumably productive configurations leading toward the final steps of 5′ domain assembly is of particular interest. These last steps include binding of secondary protein S16 to the 5′ domain and native tertiary interactions between helices 3, 12 and 18 ²⁷. Because these helices interact with S12, and S12 binding depends on the presence of S16, these conformational steps may form a checkpoint for incorporation of S12 into 30S ribosomes and further interactions between the 5′, central and 3′ minor domains.

In the cell, ribosome assembly is coupled to transcription, allowing the 5′ domain to fold before the 3′ end of the 16S rRNA is synthesized. Thus, RNA and protein interactions within the 5′ domain may correpond reasonably well with assembly intermediates formed in vivo. A caveat is that metastable interactions with the 16S leader, which interacts with the 5′ domain, may change the balance between alternative conformations of the 5′ domain or introduce new conformations ^55;56. Nonetheless, our results show that ribosomal proteins stabilize the structure of the rRNA far beyond their immediate binding sites and in specific ways, contributing to the cooperativity of subunit assembly.

Methods

Purification of recombinant ribosomal proteins

E. coli ribosomal proteins S4, S17 and S20 were over-expressed, purified by ion exchange chromatography and assayed for binding activity as previously described ^{27 57}.

Purification of 5′ domain

The 542 nt 5′ domain RNA was transcribed in vitro with T7 RNA polymerase from pRNA1 ²⁶ using standard methods ⁵⁸ and purified by denaturing 4% (w/v) polyacrylamide gel electrophoresis. The RNA concentration was determined by UV absorption at 260 nm and ε₂₆₀ = 5.4 × 10⁶ M^-1cm^-1.

Nitrocellulose filter binding

S17 binding reactions were carried out in Fisher low-retention tubes. Uniformly ³²P-labeled 16S rRNA 5′ domain was diluted to approximately 1 nM (10,000-25,000 cpm per 50 μL reaction) in reconstitution buffer (80 mm K-Hepes 7.6, 330 mM KCl, 20 mM MgCl₂, 0.01% Nikkol detergent, and 6 mM β-mercaptoethanol) except where noted. Serial dilutions of the protein in reconstitution buffer containing 0, 2, 3, 4, 5, 10, 20 or 30 mM MgCl₂ were kept on ice until needed. The RNA and protein were coincubated at 37°C for one hour before filtration on 25 mm nitrocellulose filters (Schleicher and Schuell BA85) as previously described ²¹. The extent of binding reached saturation in 5 min (data not shown). Dried filters were counted, and the data were fit to a Langmuir binding isotherm. For competitive binding assays, ³²P-labeled 5′ domain and S17 were incubated 20-30 minutes at 37°C as described above, then challenged with increasing concentrations of competitor RNA for 30 minutes at 37°C. The data were fit as previously described ⁵⁹.

Hydroxyl radical footprinting

The 5′ domain RNA was folded 30–40 min at 37°C in reconstitution buffer containing 0-20 mM MgCl₂ before treatment with Fe(II)-EDTA as previously described ²⁶. The 5′ domain RNA was previously shown to fold in less than 5 minutes ²⁶. Protein-RNA complexes were prepared by incubating 12 pmol RNA in reconstitution buffer containing the desired MgCl₂ concentration, 15–20 minutes at 37°C, before addition of 5 molar equivalents of S17 or S20 or 4 molar equivalents of S4 in a total volume of 56 μl. The RNA was coincubated with the proteins for an additional 45 minutes at 37 °C. The optimum protein:RNA ratios and incubation time were determined as described elsewhere (P.R. and S.W., manuscript submitted).

Following assembly of 5′ domain RNP complexes, 14 μl containing 3 pmol 5′ domain RNA was removed from each reaction and treated with Fenton reaction reagents 1 min on ice as previously described ²⁹. Control reactions were carried out in parallel with each titration on native E. coli 30S ribosomal subunits and on 5′domain RNA in 80 mM K-Hepes, 80 mM K-Hepes plus 330 mM KCl, or reconstitution buffer. Samples were analyzed by primer extension and AMV reverse transcriptase as described previously ²⁶.

Data analysis

Protected regions of the RNA backbone were compared with the solvent-accessible surface area for each C4′ atom in the 5′ domain of the 16S rRNA using coordinates from the structure of E. coli 30S ribosome ⁶⁰ and the program Calc-surf ⁶¹. The extent of cleavage at each position in the 5′ domain RNA was quantified with a Molecular Dynamics Phosphorimager and normalized to a reference band whose intensity did not change over the Mg²⁺-titration. The fractional saturation (Ȳ) was determined by normalizing to the maximum extent of cleavage (Ȳ = 0) and the cleavage of 30S complexes (Ȳ = 1). The extent of cleavage was frequently but not always greatest in 80 mM K-Hepes.

Assuming that the extent of hydroxyl radical cleavage reflects the equilibrium between an “open” and “closed” state at each nucleotide, the fractional saturation (Ȳ) of each backbone protection vs. Mg²⁺ concentration (C) was fit to an isotherm for a cooperative two-state equilibrium, Ȳ = Ȳ₀ + (C/C_m)ⁿ/[1+(C/C_m)ⁿ], in which C_m is the midpoint, n is the Hill coefficient and Y₀ is the extent of protection in 330 KCl and no MgCl₂. Multiphasic transitions were fit to a four-state model in which the statistical weight of each term was taken from the equilibrium constant for the open and closed state, K_i = (C/C_m,i)^{n,i 31;62},

$$\overline{Y}=\frac{{\left(C/C_{m,2}\right)}^{n,2}+{\left(C/{\text{C}}_{m,N}\right)}^{n,N}}{1+{\left(C/C_{m,1}\right)}^{n,1}+{\left(C/C_{m,2}\right)}^{n,2}+{\left(C/C_{m,N}\right)}^{n,N}}\cdot$$ (1)

The initial state (I_core) and one intermediate are assumed to be completely exposed, while the second intermediate and the final state (N) are assumed to be fully protected. In some cases, the data could be described equally well by two sequential transitions,

$$\overline{Y}=\frac{A_1{\left(C/C_{m,1}\right)}^{n,1}}{1+{\left(C/C_{m,1}\right)}^{n,1}}+\frac{A_2{\left(C/C_{m,2}\right)}^{n,2}}{1+{\left(C/C_{m,2}\right)}^{n,2}}\cdot$$ (2)

in which A₁ and A₂ are the magnitude of change in protection with each step.

Histograms

Individual transitions were classified into four groups by sorting midpoints into 0.3 mM bins. Weakly protected positions for which a folding transition with respect to [MgCl₂] could not be determined were arbitrarily assigned a C_m = 20 mM and n = 1. For biphasic transitions, the midpoint of the second transition was assumed to represent formation of the native complex and was used in the cluster analysis. 3-D ribbons were made in Pymol (Delano Scientific, http://pymol.org/educational.html) using the crystallographic coordinates of the E. coli ribosome (2avy) ⁶⁰.