Small RNA binding to the lateral surface of Hfq hexamers and structural rearrangements upon mRNA target recognition

Abstract

The bacterial Sm-like protein Hfq is a central player in the control of bacterial gene expression. Hfq forms complexes with small regulatory RNAs (sRNAs) that use complementary “seed” sequences to target specific mRNAs. Hfq forms hexameric rings, which preferably bind uridine-rich RNA 3′ ends on their proximal surface and adenine-rich sequences on their distal surface. However, many reported properties of Hfq/sRNA complexes could not be explained by these RNA binding modes. Here, we use the RybB sRNA to identify the lateral surface of Hfq as a third, independent RNA binding surface. A systematic mutational analysis and competition experiments demonstrate that the lateral sites have a preference for and are sufficient to bind the sRNA “body,” including the seed sequence. Furthermore, we detect significant structural rearrangements of the Hfq/sRNA complex upon mRNA target recognition that lead to a release of the seed sequence, or of the entire sRNA molecule in case of an unfavorable 3′ end. Consequently, we propose a molecular model for the Hfq/sRNA complex, where the sRNA 3′ end is anchored in the proximal site of Hfq, whereas the sRNA body, including the seed sequence, is bound by up to six of the lateral sites. In contrast to previously proposed arrangements, the presented model explains how Hfq can protect large parts of the sRNA body while still allowing a rapid recycling of sRNAs. Furthermore, our model suggests molecular mechanisms for the function of Hfq as an RNA chaperone and for the molecular events that are initiated upon mRNA target recognition.

The Hfq protein acts as a central hub in the control of bacterial gene expression by small RNAs (sRNAs) (1–3). Hfq belongs to the large family of Sm and Sm-like [(L)Sm] RNA binding proteins and assembles into homohexameric rings (4). Similar to its heteroheptameric homologs in eukaryotes, the Hfq ring has a so-called proximal RNA binding site, where it accommodates up to six nucleotides (one per monomer), with a preference for uridines and a discrimination against guanines (5–7). In addition to the proximal surface, the distal surface of the Hfq ring also binds RNA, with a preference for adenine-rich sequences, and covering up to 18 nucleotides—two (8) or three (9) per monomer.

The regulatory sRNAs that bind Hfq are Rho-independent transcription units and consequently share a terminator stem-loop structure with a stretch of single-stranded uridines at their 3′ end (10); their remaining bodies are structurally very diverse, with additional hairpins and single-stranded uridine-rich regions (11, 12). Depending on the metabolic and environmental conditions, sRNAs are differentially expressed (13, 14) and regulate gene expression via base pairing, frequently to entire sets of partially complementary mRNAs (15, 16). The respective mRNA targeting (seed) sequence is usually located at the 5′ end of the sRNA (11, 12).

Hfq stabilizes sRNAs in the absence of their targets and facilitates base-pairing to the mRNAs (17–19); it also helps trigger subsequent steps, such as the repression of translation and/or the acceleration of decay, but also mRNA activation (2, 3, 16, 20). Furthermore, there is a rapid exchange and competition of sRNAs for Hfq under physiological conditions, such that Hfq can be regarded as a node for the integration of parallel sRNA signaling pathways (21–24).

In an attempt to explain the selectivity of Hfq for sRNAs over other RNAs in the cell, we recently found that the proximal site of Hfq preferably binds the 3′ hydroxyl group in the context of the uridine-rich sRNA terminator ends (5) and not internal uridine-rich sites, as previously inferred from RNA complexes of the eukaryotic Sm heteroheptamers (6, 25–27). This finding challenged the prevailing view of how sRNAs interact with Hfq. However, whereas terminator recognition provides a unifying explanation for the recognition of structurally diverse sRNA (5, 28), many other physiological aspects cannot be explained by an exclusive interaction with the terminator end.

Therefore, we took a highly systematic approach, exploiting the advantages of quantitative size-exclusion chromatography and combining structure-based protein mutations with tailor-made RNA substrates to separately probe the individual RNA binding sites on Hfq. This approach led to the discovery of the lateral surface of the Hfq ring as an additional independent RNA binding surface that is largely responsible for binding the sRNA body, and to the detection of significant conformational rearrangements in the Hfq/sRNA complex upon mRNA target recognition. As a result, we obtain a radically different and much more detailed molecular model for the interaction of sRNAs with Hfq that explains many of the observed properties in a better way, and that stimulates new hypotheses on the molecular events following mRNA target recognition.

Results

Lateral RNA Binding Sites on Hfq Form a Major sRNA Binding Surface.

Previous binding experiments have shown that Salmonella typhimurium RybB sRNA still binds Hfq hexamers when 3′ end recognition is blocked by a 2′–3′ cyclic phosphate (RybB-cP), and even when the proximal RNA binding site of Hfq is occupied by another RNA molecule (5). Like for the natural sRNA 3′ end (RybB-OH), the binding stoichiometry between RybB-cP and Hfq remains equimolar [one RNA molecule (25 kDa) per protein hexamer (67 kDa)], as independently determined by quantitative size-exclusion chromatography and static laser light scattering (Fig. S1). This finding is true even with a threefold excess of Hfq, indicating that the sRNA body does not bind more than a single Hfq ring (Fig. S1C). These experiments suggest that in addition to the proximal site, another RNA binding surface on Hfq contributes to sRNA binding. We therefore used RybB-cP as a sensitized probe that would no longer depend on the proximal site in Hfq binding experiments and should hence reveal the additional sRNA binding surfaces in the context of suitably mutated Hfq hexamers.

RNA binding mutants of Hfq were generated in the context of the Salmonella typhimurium protein, St Hfq, and were based on existing Hfq/RNA crystal structures to specifically impair RNA binding to the proximal [Hfq (F42A)] (5) or the distal RNA binding site [Hfq (Y25A Q52A)] (9). As an additional potential RNA interaction surface, we also mutated a conserved, positively charged patch that appears six times on the lateral surface of the Hfq hexamer [Hfq (R16S R17A R19A K47E)] (Fig. S2). Using analytical size-exclusion chromatography, we first verified that all Hfq mutants still formed hexameric rings and confirmed that the mutations were functional and specific (Fig. S3). Indeed, the Hfq (F42A) mutation selectively prevented the interaction of a hexauridine substrate [oligo-(U)₆] with the proximal RNA binding site, whereas the Hfq (Y25A Q52A) mutation impaired the interaction of oligo-(A)₂₀ RNA with the distal surface (Fig. S3 B and G). As expected, mutation of the lateral surface had no effect on oligo-(U)₆ or oligo-(A)₂₀ RNA binding (Fig. S3 D and H).

Binding of RybB-cP, however, was not affected by mutating the proximal or the distal surface of Hfq; instead, it depended entirely on an intact lateral surface (Fig. 1). Although individual residues had been investigated before (29, 30), the lateral surface of Hfq had not been implicated as a major sRNA binding surface. Therefore, we tested the general importance of the six lateral sites for sRNA binding using another sensitized sRNA, S. typhimurium RprA-cP (31), and found again that the mutation of the lateral surface efficiently abolished sRNA binding (Fig. S4). Hence we conclude that the lateral RNA binding sites on Hfq are highly relevant for the interaction, likely with many sRNAs.

Fig. 1.

RybB sRNA interacts with the lateral RNA binding surface of St Hfq. Analytical size-exclusion chromatography of St Hfq RNA binding mutants (solid lines, colored according to the mutated surface) in the presence of RybB-cP sRNA (black, solid lines). The cartoons indicate the positions of the mutations on the molecular surface of the Hfq hexamer (Hfq₆; Protein Data Bank ID code: 2YLB) (5). Starting concentrations of Hfq₆ and of RybB-cP are indicated. Elution profiles show apparent concentrations for Hfq₆ and RybB-cP, calculated from the relative absorption properties of the components. Elution profiles for the RNA substrate alone (black, dashed lines) are superimposed. (A) Hfq wild-type, proximal side view. The RNA secondary structure is drawn as derived from covariance analyses (55, 56) and experimental probing (11). Uridine-rich sequences from the sRNA body are in blue, the terminator structure is in green, and the 2′–3′ cyclic phosphate is in orange. (B) Hfq (F42A), proximal side view. (C) Hfq (Y25A Q52A), distal side view. (D) Hfq (R16S R17A R19A K47E), proximal side view. Only the mutation of the lateral surface (D) abolishes the interaction with RybB-cP. See Fig. S3 for additional controls.

Hfq Binding Involves Uridine-Rich Sequences from the sRNA Body and the 3′ Hairpin Structure.

To determine which sequence and secondary structure elements of the RybB sRNA interact with the lateral sites of Hfq, we designed a systematic series of RybB mutants and analyzed their effect on sRNA binding. Again, we used sensitized RNA constructs terminating in a 2′–3′ cyclic phosphate to eliminate the interaction of the 3′ hydroxyl group with the proximal site of Hfq. We focused on the uridine-rich stretches within RybB, because probing experiments previously had identified internal uridine-rich sequences as primary interaction sites for Hfq (11, 32–34). Indeed we find that mutation of the first uridine-rich stretch within RybB-cP (RybB_mut1-cP, U⁹UUUCUUU to A⁹AAACAAA) as well as the mutation of the second uridine-rich stretch (RybB_mut2-cP, U²⁷UUU to G²⁷CGC) abolish Hfq binding in size-exclusion chromatography (Fig. 2 A and B). However, the two stretches are not sufficient on their own, because a corresponding minimal fragment (RybB_Δ31–78-cP; Fig. 2C), which lacks the Rho-independent transcription terminator, fails to stably associate with Hfq. Consequently, we also modified the terminator stem-loop and found it to be crucial as well. A shortening and likely destabilization of the stem by the replacement of nucleotides 40–68 by a GAAA tetra-loop (RybB_Δ40–68-cP; Fig. 2D) resulted in complex dissociation during gel filtration, whereas a substitution of the hairpin with an alternative sequence was tolerated (RybB_hp-cP; Fig. 2E). Furthermore, the truncation of the 3′ terminal oligo-(U) sequence of RybB-cP to only two remaining 3′ uridines (RybB_Δ76–78-cP; Fig. 2F) strongly reduced Hfq binding, which indicates an alternative binding mode for the 3′ terminal uridines in the context of RybB-cP, where these nucleotides are not buried in the proximal site of Hfq (i.e., they bind on the lateral surface).

Fig. 2.

The specific binding of the sRNA body to the lateral sites of Hfq depends on the sequence and structure of RybB. Analytical size-exclusion chromatography of RybB-cP sRNA variants (black, solid lines) in the presence of St Hfq (red, solid lines). Starting concentrations of Hfq₆ and of RybB-cP are indicated. Elution profiles show apparent concentrations for Hfq₆ and RybB-cP, calculated from the relative absorption properties of the components. Elution profiles for the RNA substrate alone (black, dashed lines) are superimposed. The mutated elements are highlighted in red in the context of the RybB secondary structure. RNA peaks labeled by an asterisk are present in the absence of protein and not indicative of Hfq binding. See Fig. S5 for additional details. (A) RybB_mut1-cP, (B) RybB_mut2-cP, (C) RybB_Δ31–78-cP, (D) RybB_Δ40–68-cP, (E) RybB_hp-cP, and (F) RybB_Δ76–78-cP.

For the unmodified RybB RNA, these results support an sRNA binding model where the natural 3′ hydroxyl end is anchored in the proximal site of Hfq and where several of the lateral sites provide multiple contacts all along the sRNA body. These contacts are not simply an unspecific interaction with the phosphoribose backbone, but they depend both on the sequence and on the structure of the body (Fig. 2 and Fig. S5). Given the competition of sRNAs for Hfq (21–24), this mode of binding can likely be extended to many other sRNAs, explaining how highly variable body architectures can yet be bound specifically.

Three RNA Binding Surfaces on Hfq Act Independently.

Hfq has been suggested to act as an RNA interaction platform and to promote sRNA/mRNA base-pairing by simultaneous binding of both substrates on different surfaces (17, 19, 35). To clarify whether a single Hfq hexamer can indeed bind two different RNA substrates simultaneously (36), we used RybB-cP as an sRNA and oligo-(A)₂₀ RNA as an unstructured, adenine-rich substrate for the distal surface (Fig. 3 and Fig. S6A). RybB-cP and oligo-(A)₂₀ each stably associate with Hfq in gel filtration (Fig. 3A and B); however, they do not interact in the absence of Hfq (Fig. 3D).

Fig. 3.

A single Hfq₆ ring simultaneously binds RybB-cP via its lateral sites and oligo-(A)₂₀ RNA on its distal surface. Competition experiment analyzed by analytical size-exclusion chromatography, challenging a saturated RybB-cP/Hfq₆ complex with an excess of oligo-(A)₂₀ RNA. Elution profiles show UV absorption of the mixtures at 260 nm. Starting concentrations of Hfq₆ and of the RNA substrates are indicated. Elution profiles for individual components are superimposed (dashed lines). Arrows indicate the composition of the peaks. (A) Saturation of Hfq₆ with RybB-cP. (B) Saturation of Hfq₆ with oligo-(A)₂₀ RNA. (C) Competition experiment. Consecutive application of RybB-cP and oligo-(A)₂₀ RNA does not lead to a displacement of RybB-cP, but to a simultaneous saturation of the Hfq₆ ring with both RNA species. Double arrows indicate the reduction of free RNA substrate in A–C. (D) Control. RybB-cP and oligo-(A)₂₀ RNA do not interact in the absence of Hfq₆. Compare Figs. S6 and S7 for further analyses.

Under conditions where Hfq was saturated with RybB-cP, we added an excess of oligo-(A)₂₀ RNA and did size-exclusion chromatography. In comparison with individual binding experiments using the same starting concentrations, we find that the amount of free oligo-(A)₂₀ RNA decreases to the same extent, and that the amount of free RybB does not increase when both substrates are present (Fig. 3C). These two observations indicate that the binding of the two substrates is not mutually exclusive, and that an Hfq-mediated ternary complex is formed instead, containing one molecule of each kind.

Furthermore, we previously showed that RybB-cP binds Hfq also if the proximal site is occupied by another uridine-rich RNA oligomer, R16 RNA (5). Consistently, it is even possible to reconstitute a quaternary complex, where RybB-cP sRNA is bound to the lateral surface, oligo-(A)₂₀ to the distal binding site, and the R16 RNA probe to the proximal site (Fig. S6B). Together, these experiments show that the three types of RNA binding sites on Hfq are independent of each other, and that simultaneous binding to the different Hfq surfaces can bring RNA molecules into proximity without the requirement for direct RNA/RNA interactions.

Hfq Rings Cooperate in Oligo-(A)_n RNA Binding.

Calculation of the binding stoichiometry of the Hfq/oligo-(A)₂₀ complex showed that one Hfq hexamer binds one RNA molecule (Fig. 3B), which suggests that oligo-(A)₂₀ interacts with the distal surface in a circular way, as observed in the crystal structure of the Escherichia coli Hfq/oligo-(A)₁₅ complex, with one Hfq monomer occupying three nucleotides (9). However, conflicting stoichiometries have been reported for oligo-(A)₁₈ RNA (29, 35, 36). To elucidate how many nucleotides of a putative mRNA substrate are needed for a stable interaction with the distal site of Hfq, we analyzed complex formation of Hfq with oligo-(A)_n RNA of different lengths (n = 10, 20, 27, and 30 nucleotides) by size-exclusion chromatography. Additionally, we determined the molecular weight of the eluting complexes by multiangle laser light scattering to distinguish complexes with similar hydrodynamic radii but different composition (Fig. S7). We find that oligo-(A)₁₀ RNA does not stably interact with Hfq (Fig. S7A), whereas oligo-(A)₂₇ and oligo-(A)₃₀ RNA can even accommodate two Hfq hexamers per RNA molecule (Fig. S7 C and D). Hence, a stable association of oligo-(A)_n RNA with the Hfq distal site involves at least four of the six Hfq monomers. Furthermore, the second hexamer is preferably recruited to already existing complexes rather than distributing onto the unbound RNA molecules. This observation suggests a cooperative assembly, mediated by the preorganization of an exposed ring of adenines on the first hexamer.

Base-Pairing of RybB to an mRNA Target Causes Structural Rearrangements on Hfq.

Despite its long 5′-terminal sequence of 36 nucleotides, RybB-cP does not bind more than one Hfq hexamer, even if Hfq is in excess (Fig. S1C). Conversely, a single Hfq hexamer does not bind more than one molecule of RybB-cP (Fig. S1B). These observations are consistent with the notion that essentially all of the RybB-cP body is required for the stable interaction with a single Hfq ring and that one RNA molecule binds and/or blocks most of the six lateral binding sites. However, because the first uridine-rich stretch within RybB-cP RNA (Fig. 2) is part of the 5′-terminal seed region that base-pairs to target mRNAs (11, 12), we also investigated what happens to the Hfq/sRNA complex upon mRNA target binding (Fig. 4 and Fig. S8). As a minimal mRNA target we used a 5′ Cy3-labeled 24-mer oligonucleotide (Cy3-ompN; Fig. 4A) that had been identified as a RybB target in ompN mRNA (37) and that can be traced separately in size-exclusion chromatography via its absorption at 550 nm.

Fig. 4.

mRNA base-pairing displaces the sRNA body of RybB from Hfq. Reconstitution of mRNA targeting using a 5′ Cy3-labeled complementary RNA oligonucleotide (Cy3-ompN) analyzed by analytical size-exclusion chromatography. Elution profiles show UV absorption of the mixtures at 230 nm (red), 280 nm (black), and 550 nm (green). Elution profiles for individual components are superimposed (dashed lines) for Hfq₆ (230 nm, red, 5 μM), RybB-OH (280 nm, black, 7 μM), and Cy3-ompN (550 nm, green, 15 μM). Cartoons symbolize the composition and structure of the dominant complexes. (A) Secondary structure of RybB RNA, base-paired to Cy3-ompN. Cy3-ompN corresponds to a fragment of ompN mRNA that includes the translational start codon (AUG, underlined), and it can be traced selectively at 550 nm. (B and C) Binary complexes between RybB-OH and Cy3-ompN (B) or RybB-OH and Hfq₆ (C). (D and E) Higher-order complexes between Hfq₆ and Cy3-ompN (D) that can be prevented by mutating the distal surface of Hfq₆ (E). (F and G) Release of Hfq₆ from RybB-cP (carrying a modified 2′–3′ cyclic phosphate) upon Cy3-ompN target recognition. (H and I) Ternary complexes between Cy3-ompN, RybB-OH, and Hfq₆ that require a hydroxyl group at the sRNA 3′ end to prevent the release of Hfq. The asterisk indicates Hfq₆, released from RybB upon target binding (F and G), and/or higher-order complexes with excessive Cy3-ompN (D and F). See Fig. S8 for an additional control.

Cy3-ompN readily forms a base-paired duplex with RybB RNA in the absence of Hfq (Fig. 4B). In the presence of Hfq, the adenine-rich Cy3-ompN can bind to the distal site of Hfq and forms higher-order complexes that elute close to the exclusion volume (∼9 mL) of the gel filtration column (Fig. 4D). The formation of such complexes can be prevented by the Hfq (Y25A Q52A) distal site mutation (Fig. 4E). In a first experiment, we used RybB-cP to reconstitute the Hfq/sRNA complex. Upon addition of the Cy3-ompN mRNA target mimic, Hfq is surprisingly released from the RybB-cP/Cy3-ompN duplex rather than forming a ternary complex with the two RNAs (Fig. 4 F and G). Depending on the Hfq variant, the liberated protein either forms higher-order complexes again with excessive Cy3-ompN RNA [Hfq (wild-type)] (Fig. 4F) or elutes close to the RybB-cP/Cy3-ompN duplex [Hfq (Y25A Q52A)] (Fig. 4G). In a second experiment, we used RybB-OH to reconstitute the Hfq/sRNA complex. Here, upon addition of the Cy3-ompN mRNA target mimic, Hfq remains anchored to the RybB-OH/Cy3-ompN duplex via its proximal RNA binding site, and no higher-order complexes are detected with the excessive Cy3-ompN RNA (Fig. 4 H and I).

Together, these results show that the hybridization of RybB to an mRNA target strongly affects the structure of the Hfq/sRNA complex. Large parts of the sRNA body are displaced from the lateral surface, and the resulting RNA duplex does not favorably interact with Hfq [see also Soper et al. (38)]. As a consequence of these important structural rearrangements, the interaction of the sRNA with Hfq is weakened significantly, liberating molecular surfaces for the potential recruitment of downstream effectors, such as RNase E (39–41).

Discussion

In the present work we describe the discovery of RNA binding sites on the lateral surface of the hexameric ring formed by the bacterial (L)Sm protein Hfq. This surface is composed of six RNA binding patches, one per monomer, and is likely conserved in most of the bacterial species where Hfq is known, with the notable exception of certain Gram-positive species where the basic character is less evident (Fig. S2). The lateral surface has an RNA binding specificity that is distinct from the previously known proximal and distal RNA binding surfaces of Hfq, and it acts independently of them. The lateral surface is sufficient to bind RybB, RprA, and probably many other sRNAs that compete with the latter for Hfq binding in the cell.

The identification of the lateral RNA binding surface significantly changes our view of how regulatory sRNAs interact with Hfq. In analogy to the binding of eukaryotic spliceosomal small nuclear RNAs to the eukaryotic Sm heteroheptameric ring (25, 27), sRNA binding to Hfq was thought to rely on the interaction of uridine-rich, internal RNA sequences with the proximal RNA binding site of Hfq (6, 11, 17, 19, 42). Considering the present and previously published data (5, 28), however, this view seems to be an exception rather than the rule. Therefore, we propose a different model for sRNA binding by Hfq, where the sRNA is anchored in the proximal site via its 3′-terminal uridine-rich terminator end, whereas the sRNA body [consisting of the internal uridine-rich stretches and base-paired elements; see also Ishikawa et al. (43)] wraps around the ring and is protected by the interaction with several of the lateral sites (Fig. 5). Hence, a single Hfq ring, possibly assisted by the C-terminal tails of Hfq (44, 45), is frequently sufficient to protect an entire sRNA. According to this model, the distal RNA binding site of Hfq would interact with sRNAs only in a few exceptional cases and would rather serve for an interaction with mRNA, as previously suggested (9, 35). In this way, the distal RNA binding site could assist mRNA targeting of the Hfq/sRNA complex in addition to the 5′ terminal seed sequence of the sRNA, increasing targeting specificity. An alternative function of the Hfq distal site is described in the context of oligoadenylation-mediated degradation of the mRNA 3′ end (46–48). Here, the observed cooperativity of Hfq-binding to oligo-(A) RNA substrates may play a role.

Fig. 5.

Model for the interaction of an sRNA with hexameric Hfq. (A) Binding of an sRNA anchors its 3′ end in the proximal RNA binding site of Hfq, locating the terminator hairpin above the proximal surface. Additional sequences from the sRNA body interact with several of the lateral RNA binding sites of Hfq. For simplicity, the sRNA body does not show additional elements of RNA structure. Furthermore, the RNA could also span across the hexamer, because quite generally, RNA binding to the lateral sites does not necessarily have to involve neighboring Hfq monomers. (B) Rapid competition by other sRNA molecules can be explained by a consecutive replacement of lateral contacts. (C) Formation of the mRNA targeting complex rearranges the interaction of the sRNA with Hfq, which can lead to a release of Hfq in the case of a modified sRNA 3′ end and if Hfq fails to bind additional mRNA sequences, e.g., via its distal RNA binding surface.

The presented model for RNA binding by Hfq untangles and improves the interpretation of existing data and has several implications for how we envisage sRNAs to function in the cell.

First, the presence of the lateral binding sites provides a rationale for the increased specificity of Hfq for sRNAs compared with other terminator-containing RNAs, such as mRNAs that lack the elements to engage the lateral sites (13). Although it is not yet clear precisely which combination of sequence/structure elements adds up for an efficient binding of the sRNA body to the lateral sites, our mutational analysis indicates a rather complex balance. Considering the enormous structural variability among sRNAs, there may indeed be many ways to bind Hfq with high affinity, and even for a given sRNA there may be several conformational options to engage the lateral sites. Importantly, however, and in contrast to previous binding models (3), the presently proposed sRNA binding mode can explain how a single Hfq ring can simultaneously protect both the sRNA 3′ end and its body from nuclease attack (Fig. 5A).

Second, the presented mode of sRNA binding also explains how the rapid competition of sRNAs for Hfq can be achieved despite their high affinity (21–24). In contrast to the proximal and distal RNA binding sites, the position of the lateral sites on the Hfq ring and their respective distance are particularly well suited for a stepwise engagement of the sites, which allows for a consecutive replacement of a given sRNA without breaking all of the interactions at once. In this way, the process of sRNA dissociation is broken down into several smaller steps with smaller individual transition energies resulting in faster kinetics (Fig. 5B). The rapid recycling of sRNAs on Hfq allows for a mechanistically simple integration of sRNA-mediated gene regulation already at the level of Hfq binding, where the abundance and the affinity of a given sRNA determine the occupancy of Hfq. The complexity is increased by the presence of an intact 3′ terminator structure that provides considerable additional binding energy when its 3′ hydroxyl group docks into the proximal site. Hence, modifications of the 3′ end, such as oligoadenylation (49) or nucleolytic processing (50), are particularly well suited for additional layers of regulation.

Third, the presence of lateral RNA binding sites on Hfq is very helpful to explain its activity as a nucleic acid chaperone (51), i.e., its ability to facilitate the rearrangement of nucleic acid structures such as the promotion of base pairs between the sRNA 5′ seed sequence and the mRNA target or the resolution of “nonproductive” sRNA/mRNA complexes. Bringing the two binding partners close in space clearly is one important aspect in this context, and Hfq with its ability to accommodate several different RNA molecules on a single ring is particularly well suited to do so (Fig. 3 and Fig. S6). Another more general and possibly also more important requirement of Hfq to act as an RNA chaperone is an ability to melt, fix, and preorient the sRNA for efficient hybridization to the target. The available crystal structures of RNA in the Hfq proximal site display a rather unfavorable conformation of the respective sequences for hybridization, with the Watson–Crick edges of the bases facing the protein (5–7). In contrast, the lateral binding sites may provide a much more favorable binding mode for target hybridization of the bound sRNA sequences. Indeed, our data show that significant portions of the RybB 5′ seed sequence contribute to Hfq binding via the lateral sites (Fig. 2), and that these sequences are readily available and released once a base-pairing partner becomes available (Fig. 4).

Fourth, and last, we show that the Hfq/sRNA complex can exist in different conformational states and that the sRNA can switch between different binding modes (Fig. 5 A and C); this is most obvious for the 5′ and 3′ ends of the sRNA. For the 3′ end, it makes a significant difference whether it is anchored in the Hfq proximal site or not (5, 28), as this can affect the degree of 3′ end processing (47, 50) or cause the release of Hfq once the sRNA has bound to its mRNA target (Fig. 4). Because the individual 3′ ends of a given sRNA species in the cell are frequently quite heterogeneous (13), this indicates distinct molecular populations with distinct regulatory potential. Regarding the 5′ end, we find that the seed sequence can alternate between an Hfq-bound state and an mRNA target-bound state where previously occluded portions of the sRNA body and of the protein surface are exposed (Figs. 2 and 4). It can easily be imagined that such conformational differences and changes are exploited in the cell to trigger downstream processes such as the recruitment of RNase E (39, 41) or the accessibility for nucleolytic cleavage and ultimate degradation (32, 52). Clearly, further experiments are needed to elucidate the mechanistic details regulating these downstream processes.

Materials and Methods

Preparation of RNA and Protein Material.

Synthetic RNA oligonucleotides [oligo-(U)₆, oligo-(A)_n, R16 (R16: 5′ GCCACUGCUUUUCUUU 3′) and Cy3-ompN RNA] were purchased as desalted material. The remaining RNA molecules were transcribed in vitro from a modified pSP64 plasmid containing a 3′-terminal HDV ribozyme that autocleaves cotranscriptionally, leaving a 2′–3′ cyclic phosphate on the target RNA. Target RNA was purified over denaturing polyacrylamide gels as described previously (53). Where necessary, the cyclic phosphate was removed using T4 polynucleotide kinase (54).

St Hfq constructs (St Hfq, GAM₁-E₁₀₂; UniProt ID code P0A1R0) were expressed as NusA fusions from a pETM60 vector in E. coli BL21(DE3) GOLD cells. NusA-Hfq was purified from the cleared lysate by Ni²⁺ affinity chromatography. After proteolytic removal of the affinity tag, Hfq was further purified by heparin affinity and size-exclusion chromatography (5). For Hfq (R16S R17A R19A K47E), the heparin affinity step was replaced by anion-exchange chromatography (Mono Q, GE Healthcare).

Analytical Size-Exclusion Chromatography.

RNA was first annealed (65 °C for 10 min, slow cooled to room temperature) in chromatography buffer [100 mM NaCl, 10 mM Tris⋅HCl (pH 8.0), 10 mM MgCl₂], then incubated with limiting amounts of St Hfq (10 min) and finally loaded onto the chromatography column [150-μL sample on a Superdex 200 10/300 GL column mounted on an ÄKTA Purifier-10 (GE Healthcare)]. Starting concentrations are shown in the figures. Elution was monitored by UV absorption at 230, 260, and 280 nm (with Cy3-ompN RNA: 230, 280, 550 nm). Apparent concentrations were calculated from the relative absorption properties of the components as described previously (5, 53). For competition experiments, elution profiles show UV absorption instead of concentrations. All experiments were done at thermodynamic equilibrium, such that the order of addition did not matter.