The Escherichia coli Sm‐like host factor I (Hfq) protein is thought to function in post‐transcriptional regulation by modulating the function of small regulatory RNAs. Hfq also interferes with ribosome binding on E. coli ompA messenger RNA, indicating that Hfq also interacts with mRNAs. In this study, we have used stimulation of group I intron splicing in vivo and a modified in vitro toeprinting assay to determine whether Hfq acts as an RNA chaperone. Hfq was able to rescue an RNA ‘folding trap’ in a splicing defective T4 bacteriophage td gene in vivo. Enzymatic analysis showed that Hfq affects the accessibility of the ompA start codon, as well as other bases within the ribosome‐binding site, explaining its negative effect on ribosome binding. We also show that the Hfq‐induced structural changes in ompA mRNA are maintained after proteolytic digestion of the protein, which classifies Hfq as an RNA chaperone.
Escherichia coli host factor I (Hfq) was first described as a factor required for bacteriophage Qβ replication (Fernandez et al., 1968). The 11.2‐kDa Hfq protein forms a hexameric ring‐shaped structure, and belongs to the large family of Sm and Sm‐like proteins that have RNA binding activity (Møller et al., 2002; Schumacher et al., 2002; Zhang et al., 2002).
E. coli Hfq binds a number of small RNAs, including those transcribed from the genes rprA (Wassarman et al., 2001), dsrA (Sledjeski et al., 2001) and oxyS (Zhang et al., 1998). All three of these RNAs modulate translation of the rpoS gene, which encodes the σS subunit of RNA polymerase, and Hfq is thought to be required for their function. In addition, we have shown that Hfq represses translation of the E. coli ompA gene by interfering with ribosome binding, which results in rapid degradation of ompA messenger RNA (Vytvytska et al., 2000). Thus, Hfq appears to have both direct and indirect functions in the control of RNA translation, by its targeting of mRNAs and regulatory RNAs, respectively.
The nucleocapsid protein of human immunodeficiency virus and the hnRNP A1 protein enhance the activity of the hammerhead ribozyme (Herschlag et al., 1994), and the E. coli proteins StpA (Zhang et al., 1995; Clodi et al., 1999) and ribosomal protein S12 (Coetzee et al., 1994) are known to facilitate splicing of the bacteriophage T4 td intron. Because these proteins are thought to assist in RNA folding by preventing misfolding or by resolving misfolded RNA structures, the term ‘RNA chaperone’ has been used to describe them (Herschlag, 1995). Another hallmark of RNA chaperones is their stimulation of intermolecular annealing events between complementary nucleic acids (Tsuchihashi & Brown, 1994). Hfq has been shown to stimulate the interactions between spot42 RNA and galK mRNA (Møller et al., 2002), and between OxyS RNA and fhlA mRNA (Zhang et al., 2002). In addition, an electron microscopy study indicated that binding of Hfq to the 3′‐terminal region of phage Qβ plus‐strand RNA induces a conformational change that brings the 3′‐end of the RNA close to the S‐site/M‐site domain. There, the RNA is presented to the bound replicase, enabling it to initiate replication (Barrera et al., 1993). In this study, we have used an in vivo RNA‐chaperone assay, using an RNA ‘folding trap’ in the T4 phage td intron (Clodi et al., 1999), and an assay using ompA mRNA as a target in vitro, to test the RNA‐chaperone activity of Hfq.
Results and Discussion
Assaying Hfq chaperone function in vivo
Semrad & Schroeder (1998)showed that the presence of a stop codon in the upstream exon of the phage T4 td gene inhibits RNA splicing. The splicing defect in the absence of translation was shown to be due to an exon–intron interaction that traps the pre‐mRNA in a splicing‐incompetent conformation. Making use of this ‘folding trap’ in the td intron, Clodi et al. (1999)have devised an in vivo RNA‐chaperone assay that relies on suppression of the splicing defect. To test whether Hfq, like StpA (Clodi et al., 1999), can rescue this folding trap, the E. coli hfq− strain AM111F′ was transformed with plasmids carrying the wild‐type td gene (pTZ18U–tdΔP6‐2), or a mutant form of the gene that has a stop codon 82 nucleotides upstream of the 5′ splice site (pTZ18U–tdKS82). The bacteria were also transformed with either plasmid pAZ205 (carrying the stpA gene), plasmid pDLE102 (carrying the hfq gene) or plasmid pACYC184 (an empty‐plasmid control). As shown previously (Clodi et al., 1999), overexpression of stpA (Fig. 1D, lane 2) stimulated splicing of the wild‐type td gene, resulting in an approximately two‐fold increase in the amount of spliced RNA (Fig. 1A, compare lanes 2 and 3; Fig. 1B). In contrast, expression of the hfq gene from the low copy‐number vector pDLE102 had no measurable effect on the splicing of the wild‐type td mRNA in the hfq− background (Fig. 1A, compare lanes 1 and 3; Fig. 1B). As determined by quantitative western blotting, this may have been due to the low levels of Hfq expressed in strain AM111F′ (pDLE102; pTZ18U–tdΔ P6‐2) (Fig. 1C, lane 1) as compared with the amount of StpA expressed in strain AM111F′ (pAZ205; pTZ18U–tdΔ P6‐2) (Fig. 1D, lane 2). However, like StpA (Fig. 1A, lane 5; Fig. 1B), the expression of Hfq was sufficient to rescue the splicing defect of the exonic stop‐codon mutant (Fig. 1A, lane 4; Fig. 1B). On a qualitative basis, and according to the definition of Clodi et al. (1999), the in vivo splicing assay suggests that Hfq functions as an RNA chaperone. Although this in vivo experiment does not permit distinction between direct and indirect effects of Hfq, it should be noted that the expression of Hfq from plasmid pDLE102 did not affect the StpA levels in E. coli strain AM111F′ (Fig. 1D, compare lanes 1 and 3, and lanes 4 and 6, respectively).
Hfq binding affects the structure of the ompA RBS
Our previous studies have shown that Hfq blocks 30S‐ribosome binding to E. coli ompA mRNA, and we observed that, unlike canonical translational repressors, Hfq does not protect a contiguous segment of the ompA ribosome binding site (RBS) from endoribonucleases (RNases) (Vytvytska et al., 2000). 30S‐ribosome binding protects the 5′untranslated region (UTR) of ompA from RNase E at cleavage site C, which is situated between the stem–loop structures hp1 and hp2 (Fig. 2C). Considering that the ribosome binds to the ompA RBS without destabilization of hp2, this result was in agreement with previous footprinting studies, which identified the region from −35 (±2) to +20 as the 30S boundaries (Hüttenhofer & Noller, 1994).
Changes observed in the cleavage pattern of RNA by endoribonucleases can be used to detect RNA chaperone activity. Taking into account the linear binding of the 30S ribosomal subunit, we examined the region from positions −70 to +20 of ompA 180 mRNA (comprising the first 180 nucleotides of the ompA mRNA and the 5′ untranslated region; see Fig. 2C) for Hfq‐induced changes in RNase T1 and RNase CV1 cleavage patterns. This enzymatic footprinting analysis showed that the presence of Hfq results in changes in sensitivity (both increases and decreases) to these nucleases at different positions in the −70 to +20 region (Fig. 2A). We found that the presence of Hfq results in an increase in prematurely terminated primer‐extension products within this mRNA segment (Fig. 2A, lanes 5 and 6). Therefore, we considered only protection from RNase cleavage and consistently occurring pronounced increases in RNase cleavage as relevant. As shown in Fig. 2A and C, these types of change in the pattern of cleavage by RNase CV1 were found at positions −61, −27 and −14 (protection), and at position −11 (pronounced increase compared to other signals). The alterations in cleavage at the first two positions (−61 and −27) affect nucleotide positions surrounding hp2, whereas the Hfq‐induced increased sensitivity to RNase CV1 at position −11 affected the single‐stranded conformation of the Shine–Dalgarno (SD) sequence. Moreover, the presence of Hfq rendered the G residue of the AUG start codon of ompA mRNA resistant to T1 cleavage, indicating that the start codon is not available for pairing with the anticodon of initiator transfer RNA in the presence of Hfq (Fig. 2A, lane 10). To verify these results, RNase T1 digestions were carried out in the absence or presence of Hfq, using ompA 180 mRNA labelled at the 5′ end with 32 P. Hfq protected several Gs (Fig. 2B, lanes 2 and 3; Fig. 2C) within the −70 to +20 region from T1 cleavage. In particular, the G of the AUG start codon, as well as the Gs of the ompA SD sequence, were protected by a 50‐fold molar excess of Hfq (Fig. 2B, lane 2; Fig 2C).
These changes in the accessibility of nucleotides in the ompA RBS in the presence of Hfq explain why Hfq inhibits the formation of the 30S translation‐initiation complex (Vytvytska et al., 2000; see Fig. 3B, lane 4). As suggested by structural data from the Staphylococcus aureus Hfq protein (Schumacher et al., 2002), binding of Hfq to A/U rich sequences could trigger RNA unwinding at the target site and at surrounding RNA structures, which would permit new RNA–RNA interactions. We have recently shown that the A/U‐rich sequence between hp2 and the SD sequence of ompA mRNA (Fig. 2C) is involved in Hfq binding (Vytvytska et al., 2000). The changes in the nuclease accessibility of bases upstream and downstream of hp2 (Fig. 2A and B), and downstream from the A/U sequence, are consistent with the proposed model of Hfq action (Schumacher et al., 2002); that is, Hfq binding can bring about the formation of new RNA–RNA interactions, including those that have a negative effect on the formation of stable translation‐initiation complexes on ompA mRNA.
Block of 30S binding after proteolysis of Hfq
In contrast to RNA‐binding proteins, RNA chaperones are defined by their transient action and by their dispensability for maintenance of the structural changes induced in RNA by their binding. To test whether Hfq fulfils these criteria, the toeprinting experiments illustrated schematically in Fig. 3Awere carried out using ompA mRNA. Briefly, because a translation initiation complex bound to an mRNA can block the progress of reverse transcriptase from a downstream primer, a stop signal (toeprint signal) is generated. When compared to the intensity of the readthrough signal, obtained from all mRNAs in the population to which no ribosomes are bound, the toeprint signal provides a measure for the efficiency of ternary‐complex formation at a given start codon (Hartz et al., 1988). As shown in Fig. 3Aa, in the first experiment, Hfq was added at a 90‐fold molar excess to ompA 180 mRNA, to which either primer AvaII (ompA 180–AvaII mRNA) or primer Y17 (ompA 180–Y17 mRNA) were annealed, and incubated for 10 min at 37 °C. In these conditions, formation of the translation initiation complex was found to be inhibited (Fig. 3Aa; Fig. 3B, lane 4).
In the second experiment (Fig. 3Ab), Hfq was first incubated with ompA 180–AvaII mRNA for 10 min at 37 °C and then proteolytically digested with protease K. This was verified by removal of an aliquotfrom the reaction mixture, followed by western blotting with anti‐Hfq antibodies (Fig. 3C, lane 2). Next, ompA 180–Y17 mRNA (used as an internal control) and phenylmethylsulphonylfluoride (PMSF) were added. 30S ribosomes and tRNAfMet were included in the reaction mixture, and after a 10‐min incubation at 37 °C, the primers were extended using reverse transcriptase. As shown in Fig 3B(lane 7), the proteolytic removal of Hfq did not restore translation‐initiation‐complex formation on ompA 180–AvaII mRNA, whereas the experimental conditions did not significantly affect 30S initiation on the ompA 180–Y17 mRNA internal control (Fig. 3Ab; Fig. 3B, lane 7). In addition, the primer extension signal obtained for ompA 180–AvaII mRNA showed that its integrity remained unaffected during the course of the experiment. The lack of translation‐initiation‐complex formation on ompA 180–AvaII mRNA under the conditions outlined in Fig. 3Abcan be explained by the inaccessibility of the G of the AUG start codon to RNase T1 after proteolytic removal of Hfq (Fig. 2A, lane 11).
To verify these results, the experiment shown in Fig. 3Abwas repeated, except that after proteolysis of Hfq, omp A180–AvaII mRNA was incubated for 3 min at 80 °C (see Fig. 3Ac). This was done to resolve the RNA conformation inhibitory to 30S binding. As expected, this restored formation of the 30S initiation complex (Fig. 3B, lane 8). These toeprinting studies showed that the Hfq‐induced structural changes in ompA RNA, which impede 30S binding (Fig. 3B, lane 4), are maintained after removal of the protein. Thus, from these experiments and from the results shown in Figs 1 and 2, Hfq fulfils all the criteria for it to be classed as a genuine RNA chaperone.
In vivo splicing assays.
In vivo splicing assays were carried out using the E. coli hfq mutant strain AM111F′, with genotype Δ( arg–lac )U196 araD139 rpsL150 ptsF25 flbB5301 rbsR deoC relA1 hfq1::Ω (F′proAB, lacIq ZΔM15 Tn10 ) (Muffler et al., 1996). Bacteria were transformed with either pAZ205 (a low copy‐number pACYC184 derivative carrying the E. coli stpA gene under control of lacpo ; Zhang et al., 1995), pDLE102 (a low copy‐number pACYC184 derivative, carrying the E. coli hfq gene under the control of lacpo ; Sonnleitner et al., 2002), pACYC184, pTZ18U–tdΔ P6‐2 (a high copy‐number pUC18 derivative carrying the wild‐type td gene under the control of lacpo ; Galloway‐Salvo et al., 1990) or pTZ18U‐td KS82 (a high copy‐number pUC18 derivative carrying a mutation in exon 1 of td, under the control of lacpo ; Clodi et al., 1999) or combinations thereof, as described in the legend to Fig. 1. Total cellular RNA isolation and the poisoned primer reaction were both carried out as described by Clodi et al. (1999).
Toeprinting assays (Hartz et al., 1988) were carried out using 30S ribosomal subunits and tRNAfMet with ompA 180 mRNA. The plasmid pUH100 (Vytvytska et al., 2000) was used as a template for the production of ompA 180 mRNA, to which either the AvaII primer (complementary to the +110 to +87 region downstream of the ompA mRNA start codon) or the Y17 primer (complementary to nucleotides +60 to +76 downstream of the start codon) were annealed. Moloney murine leukaemia virus reverse transcriptase was pre‐mixed with deoxynucleotide triphosphates before addition to the toeprint assay reaction mixture. When included in the reaction, Hfq was added in a 90:1 molar ratio to the mRNA. Hfq protein was prepared as described by Vytvytska et al. (1998). For removal of Hfq after incubation with ompA 180–AvaII mRNA, 0.01 units of protease K were added for 10 min at 37 °C. PMSF was then added to a final concentration of 2 mM, and the toeprinting reactions were carried out in SB buffer (with 10 mM Mg‐acetate) as described in Hartz et al., 1988, with one‐fifth of the reaction mixture.
The Ava II primer, labelled at the 5′ end with 32 P (Fig. 3B), was annealed to the ompA 180 runoff transcript of Hind III‐digested pUH100, as described in Vytvytska et al. (2000). The annealed mixture then was subjected to RNase T1 or RNase CV1 digestion in the presence (90:1 molar ratio of Hfq to mRNA) or absence of Hfq (see Fig. 2A) in SB buffer. The cleavage sites were then mapped by primer extension using avian myeloblastosis virus reverse transcriptase. The concentration of the ompA 180 mRNA used was 12 nM, and 1 unit of RNase T1 or 0.01 units of RNase CV1 were used.
In addition, Rnase T1 mapping was carried out with 12 nM ompA 180 mRNA labelled at the 5′ end with 32 P. Hfq was added to the mRNA in a 50‐fold or 90‐fold molar excess (see Fig. 2B), and the mixture was incubated for 5 min at 37 °C in SB buffer before addition of 1 unit of RNase T1. After a 3‐min incubation at room temperature, the reaction was terminated by ethanol precipitation.
To quantify Hfq and StpA protein levels in the strains used for the in vivo RNA‐chaperone assay (Fig. 1A, lanes, 1–6), equimolar amounts of total cellular protein, prepared concomitantly with total RNA, were separated on 12% SDS–polyacrylamide gels. Quantitative immunoblotting was carried out using anti‐Hfq and anti‐StpA antibodies. Visualization of Hfq‐ and StpA‐specific immunocomplexes was carried out using a goat anti‐rabbit antibody, labelled with alkaline phosphatase, in a colour reaction using a nitroblue tetrazolium p‐toluidine salt (Sonnleitner et al., 2002).
We are grateful to C. Waldsich and R. Schroeder for their help and advice, and to R. Hengge‐Aronis for materials. This work was supported within the framework of the Special Research Program on ‘Modulators of RNA fate and function’ at the Vienna Biocenter by grant F1715 from the Austrian Science Fund.
- Copyright © 2003 European Molecular Biology Organisation