|Home | About | Journals | Submit | Contact Us | Français|
CspA, a small protein that is highly induced by cold shock, is encoded by a monocistronic mRNA of 428 nucleotides (nt) whose half-life and abundance are greatly increased following cold shock. We show here that in vitro cspA mRNA can bind multiple copies of Hfq, a hexameric Sm-like protein which promotes a variety of RNA-RNA interactions. Binding of the first Hfq hexamer occurs with an apparent Kd (dissociation constant) of <40 nM; up to seven additional hexamers can bind sequentially at higher concentrations. Known ligands of Hfq, including the small regulatory RNA, RyhB, compete with cspA mRNA. Several experiments suggest that the first binding site to be occupied by Hfq is located at or near the 3′ end of cspA mRNA. The consequences of limited Hfq binding in vitro include nearly total inhibition of RNase E cleavage at a site ~35 nt from the 3′ end of the mRNA, stimulation of polyadenylation by poly(A) polymerase 1, and subsequent exonucleolytic degradation by polynucleotide phosphorylase. We propose that Hfq may play a facilitating role in the metabolism of cspA mRNA.
Differential mRNA stability is an important means of achieving posttranscriptional regulation of gene expression in Escherichia coli and many other organisms (1, 5). CspA, a small cold shock protein (11, 16, 17, 28), is an excellent case in point. During normal exponential growth, CspA is almost undetectable; moreover, the half-life of its mRNA is extremely short, having been estimated at 10 to 47 s (12, 15). Following the onset of cold shock, the cspA mRNA undergoes a structural rearrangement (9) and is stabilized significantly as its half-life increases to >10 min (2, 6, 7, 10, 11, 13). Thus, the abundance of CspA increases substantially. In later stages of the cold shock response, levels of CspA fall, accompanied by a reduction in the stability of its mRNA as cells adapt to growth at a lower temperature (10). The precise mechanisms by which cspA mRNA is initially stabilized following cold shock and later destabilized at the end of the adaptive period remain to be elucidated. Nonetheless, it is clear that differential stability is inherent to the cspA mRNA itself. First, its 5′ untranslated region (UTR) is a major determinant of cold sensitivity (6, 7). Second, cspA mRNA is susceptible to RNase E in either the normal degradosome or in the modified degradosome isolated from cold-shocked cells (16). This observation and others reviewed by Beran et al. (1) imply that modification of the enzymes of mRNA decay cannot explain differential mRNA stability following cold shock. Rather, changes in translational efficiency and mRNA stability at the onset of cold shock offer the most likely explanation for the induction of expression of CspA (13). Both RNase E and polynucleotide phosphorylase have been reported to act on cspA mRNA in vivo and/or in vitro (6, 16, 37), but the mechanisms by which these enzymes may determine the differential stability of cspA mRNA are unclear.
We were fortuitously prompted to investigate whether Hfq, a widely distributed member of the Lsm family of RNA-binding proteins (reviewed in references 3, 33, and 36), might intervene in regulating CspA expression. Hfq functions as a general regulator of gene expression, particularly through its role in mediating the action of many small regulatory RNAs (sRNAs) (33). It has also been implicated in regulating translation (28, 34, 35), promoting polyadenylation (15, 20, 25), protecting RNAs from endo- or exonucleases (8, 26), and in participating in cellular responses to various stresses (36). The Hfq hexamer contains two RNA binding faces (22, 24, 31). The proximal face binds small RNAs and tRNAs while the distal face can interact with poly(A) (21, 22, 24, 31). In addition, Hfq has been reported to bind to or alter the properties of several enzymes involved in RNA turnover, including RNase E, polynucleotide phosphorylase, and poly(A) polymerase (PAP) (15, 25, 27). In the most striking example, evidence has been presented that Hfq binds to the C-terminal “scaffold” domain of RNase E in lieu of the resident protein ligands, enabling the action of the small RNAs, SgrS and RyhB (27). The many apparent functions of Hfq and the pleiotropy of hfq mutations (32) have elevated the interest in this protein.
The present investigation was prompted by the unexpected finding that cspA mRNA can bind multiple Hfq hexamers (see below). There is no prior evidence that Hfq plays a role in the cold shock response or in the metabolism of cspA mRNA in particular. Nonetheless, given its frequent involvement in posttranscription regulatory mechanisms and the hints that Hfq interacts with key components of the mRNA decay apparatus, it would be surprising if Hfq did not play at least a facilitating role in RNA metabolism at some point during adaptation of E. coli to growth at low temperature.
Strain MG1693 (thi rph) and its derivatives, SK5665 (rne1), SK5691 (pnp7), SK7988 (ΔpcnB), SK10023 (hfq1::Ω Kanr), SK10272 (hfq1::Ω Kanr ΔpcnB) and SK10364 (hfq1::Ω Kanr pnp7), were obtained from Sidney Kushner, Department of Genetics, University of Georgia. Strain MA06 [BL21(DE3)/pLysS/pET21b-Hfq] was obtained from Eric Massé (Université de Sherbrooke). This strain was used for the overexpression of Hfq. Plasmids encoding Hfq mutants Y25A [Hfq(Y25A)] and Hfq(Y55A) (24) were supplied by Andrew Feig, Wayne State University (Detroit, MI), and subsequently modified to bring the CspA coding sequence in frame with the C-terminal hexahistidine tag encoded by the vector. The arabinose-regulated expression vector pBAD28 (14) was obtained from Jon Beckwith (Harvard Medical School). The plasmid pCspA-1 has been described previously (16); pCspA-Δ5′ was derived from it by amplification of chromosomal DNA using primers 5′-CGCAGAATTCAATACGACTCACTATAGGGAACGTAACCAGCCTGTAATC and 5′-GGACGGATCCTGAAAACATTTAAAAAAATCCCCGCC. The plasmid includes the 3′-terminal 76 residues of cspA (the last six codons and the entire 3′ UTR) under the control of the T7 promoter in a pUC18 backbone. A construct encoding RyhB was similarly constructed using the primers 5′-GGCGCGAATTCGCGATCAGGAAGACC and 5′-GTCGCCTGCAGTAAAAAAGCCAGCACCC.
Degradosomes were purified and used as described previously (4). Hfq was purified (18) from strain MA06 grown in LB medium supplemented with chloramphenicol and carbenicillin at 37° with vigorous aeration to an A600 of 0.4 and then induced with isopropyl-β-d-thiogalactopyranoside (IPTG; 1.25 mM). Growth continued for 3 h. Cells were harvested by centrifugation and washed once with 10 mM Tris-HCl, 100 mM NaCl, and 3 mM EDTA, pH 8; the pellet was then frozen at −20°. Cells were thawed and resuspended in 50 mM Tris, pH 8, 200 mM NaCl, 5 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, 0.8 μg/ml leupeptin, 2 μg/ml aprotinin, 0.8 μg/ml pepstatin A, and 160 units of DNase I. Cells were subjected to three 10-s pulses of sonication at 30% maximum intensity on a Branson 250 sonicator. This lysate was clarified by centrifugation in a JA20 rotor at 11,200 rpm for 45 min at 4°C. The resultant supernatant was passed over a Talon 7 column twice. The column was then washed with 2 volumes of 10 mM Tris-HCl, pH 7.8, 200 mM NaCl, and 5 mM imidazole and eluted in a step gradient of the same buffer containing 25, 50, 100, and 250 mM imidazole. Fractions were analyzed by SDS-PAGE, and the second elution fraction, containing the majority of the Hfq, was applied to a Source Q column in 10 mM Tris-HCl, pH 7.8, 5% glycerol, 0.1 mM EDTA, and 0.5 mM dithiothreitol (DTT) and eluted with the same buffer containing a gradient of 100 to 800 mM NaCl. Following analysis by SDS-PAGE, appropriate fractions were pooled and concentrated. This material was applied to a Sephacryl 300 column in 10 mM Tris-HCl, pH 7.8, 5% glycerol, 0.1 mM EDTA, 0.5 mM DTT, and 300 mM NaCl. Fractions were analyzed by SDS-PAGE, and the four peak fractions containing Hfq were pooled and quantified.
Cultures expressing mutant forms of Hfq were grown essentially as described above but with kanamycin in place of chloramphenicol and carbenicillin. Cells were harvested by centrifugation, washed once with 50 mM Tris-HCl, 0.5 M NH4Cl, 20 mM imidazole, and 5% glycerol, pH 7.5, and frozen at −20°C. Cells were thawed and resuspended in 25 ml of the same buffer, subjected to sonication as above, and treated with 100 units of DNase I for 1 h on ice. Debris was removed by centrifugation as above, and the supernatant was passed over a Talon 7 column twice. The column was washed with 5 volumes of 50 mM Tris-HCl, 0.5 M NH4Cl, 20 mM imidazole, and 5% glycerol, pH 7.5, followed by 5 volumes of 50 mM Tris-HCl, pH 7.5, 1.5 M NH4Cl, and 5% glycerol and eluted in two steps: 5 volumes of 50 mM Tris-HCl, pH 7.5, 0.5 M NH4Cl, 250 mM imidazole, and 5% glycerol, followed by 5 volumes of 50 mM Tris-HCl, pH 7.5, 1 M NH4Cl, 8 M urea, and 5% glycerol. Eluted fractions were analyzed by SDS-PAGE. Hfq was predominantly found in the first elution fraction and was dialyzed against 50 mM Tris-HCl, pH 7.5, 0.25 M NH4Cl, 1 mM EDTA, and 10% glycerol and quantified.
Internally labeled RNAs were prepared by transcription of linear templates (16). Binding of Hfq to various RNAs was assayed by electrophoretic mobility shift. Radiolabeled RNA (0.2 pmol) was incubated for 30 min at 37° with 0 to 100 nM Hfq in 50 μl of 25 mM Tris-HCl, pH 8.0, 1 mM EDTA, 80 mM NaCl, 1% glycerol, and 0.01% dodecyl maltoside before the addition of 20 μl of loading dye (0.25% bromphenol blue, 0.25% xylene cyanol FF, 50 mM EDTA, pH 8.0, and 50% glycerol). In competition experiments, the method of Lee and Feig (21) was employed. Radiolabeled RNA (0.2 pmol) was incubated for 30 min at ambient temperature with 60 nM Hfq in 25 μl of 50 mM Tris-HCl, pH 8.0, 250 mM NH4Cl, and 10 mM MgCl2. Various amounts of cold mRNA were added in a 12.5-μl volume of the same buffer and incubated a further 30 min before the addition of 20 μl of loading dye. In either case, samples were separated on a 6% polyacrylamide (36:1) gel in Tris-borate-EDTA buffer at 4°. Gels were fixed in 5% acetic acid-5% ethanol, rinsed with distilled water, dried, and exposed to a PhosphorImager screen.
Oligonucleotide blocking experiments were performed by mixing 0.4 pmol of radiolabeled RNA with 4 nmol of oligonucleotide in 10 μl of H2O, heated to 99° for 60 s, held at 42° for 25 min, and then chilled on ice. Each RNA-oligonucleotide complex was then mixed with various amounts of Hfq in a final volume of 50 μl of electrophoretic mobility shift buffer (see above) and held at ambient temperature for 30 min. Agarose loading dye was added, and samples were separated as described above.
RNase assays were performed as described previously (4, 16). Samples containing Hfq were made 0.1% in SDS and then extracted with phenol-chloroform-isoamyl alcohol (25:24:1) and concentrated by ethanol precipitation prior to analysis.
During experiments aimed at reconstituting the cleavage of small RNAs complexed to Hfq by RNase E, we tested whether other classes of RNA would also bind to Hfq. The interaction of cspA mRNA, encoding the major cold shock protein of E. coli, with Hfq was particularly striking (Fig. (Fig.1a).1a). This mRNA formed a series of complexes with Hfq with relatively high affinity. At 40 nM Hfq (Fig. (Fig.1a,1a, lane 3) over half the cspA mRNA entered a complex. At 60 to 80 nM Hfq, four additional complexes of decreasing mobility were clearly visible (Fig. (Fig.1a,1a, lanes 4 and 5). We presume that the first complex (lanes 2 and 3) consists of one hexamer of Hfq complexed to cspA mRNA while each successively slower complex contains an additional Hfq hexamer. Up to eight discrete complexes can be identified in lanes 4 to 10.
In addition to cspA mRNA, several other RNAs tested, including the rpsT mRNA and the 9S RNA precursor to 5S rRNA, formed complexes, as assayed by electrophoretic mobility shift (Fig. (Fig.2).2). The apparent affinity of Hfq for these RNAs was considerably lower than that for cspA mRNA. In the case of rpsT mRNA, less than half the input RNA was converted to complexes at 135 nM Hfq (Fig. (Fig.2a).2a). The same was true for 9S RNA (Fig. (Fig.2b).2b). As a complementary experiment, we also tested whether other known or putative RNA-binding proteins would interact with cspA mRNA. Among others (see below), we tested CspA itself as it is widely believed to be an RNA chaperone (13, 29). Purified CspA did not bind to cspA mRNA under the conditions used here at any concentration tested even in the presence of Hfq (data not shown).
Several experiments were performed to determine whether the cspA mRNA-Hfq complexes represent a specific interaction. First, we tested whether mutants known to reduce the affinity of Hfq for target RNAs, namely, Hfq(Y25A) and Hfq(Y55A) (21), would also weaken the interaction with cspA mRNA. The former mutation affects the distal face of Hfq that interacts with poly(A), whereas Hfq(Y55A) affects the proximal face that binds to small RNAs (21, 22, 24). Because both mutants were prepared with a C-terminal His6 tag, we used similarly tagged wild-type (WT) Hfq in this comparison (Fig. (Fig.1b).1b). The data in Fig. Fig.1c1c show that Hfq(Y25A) was still able to bind cspA mRNA but with somewhat reduced affinity. Complexes were clearly detectable at 80 nM Hfq(Y25A), compared to 60 nM for Hfq-His6 (compare Fig. Fig.1c,1c, lane 5, to b, lane 4). In contrast, Hfq(Y55A) was unable to bind cspA mRNA with appreciable affinity at any concentration up to 500 nM (Fig. (Fig.1d).1d). For comparison, we tested WT Hfq and both mutants for their ability to form complexes with RyhB, a small RNA known to require Hfq for its activity (23). RyhB bound to Hfq(Y25A), albeit with about half the apparent affinity of binding to WT Hfq (data not shown). Like cspA mRNA, RyhB sRNA did not bind to Hfq(Y55A) at any concentration tested up to 500 nM (data not shown). We conclude that cspA mRNA interacts with the proximal face of Hfq only, similar to small RNAs and tRNA (21, 24).
As a second test of specificity, we determined whether the binding of Hfq to cspA was sensitive to competition (Fig. (Fig.3).3). cspA mRNA itself competed effectively for Hfq binding as evidenced by the reduction in low-mobility complexes and the appearance of free cspA mRNA in the presence of as little as 8 nM competitor (Fig. (Fig.3a).3a). Likewise, RyhB sRNA also competed for Hfq (Fig. (Fig.3b).3b). In this case, however, higher concentrations of RyhB RNA were required to release free cspA mRNA than in the experiment shown in Fig. Fig.3a.3a. Finally, an RNA representing the 3′ 76 residues of cspA mRNA also competed for Hfq binding (Fig. (Fig.3c)3c) but with less effect than the full-length RNA.
Hfq is believed to bind to an AU-rich pentamer in single-stranded RNAs (3, 30) as well as to 3′ ends adjacent to Rho-independent terminators (25). Inspection of the sequence of the cspA mRNA identified a number of AU-rich tracts that could potentially bind Hfq (Fig. (Fig.4a).4a). We attempted to map such sites by toeprinting of cspA mRNA-Hfq complexes using oligonucleotide a (Fig. (Fig.4a)4a) as a primer for reverse transcription. We could not, however, reproducibly detect incomplete cDNAs with this method even under conditions where we could show independently that Hfq was bound to the RNA. Either the reverse transcriptase displaced bound Hfq, or Hfq was bound 3′ to the site to which oligonucleotide a annealed. We thus asked instead whether a bound oligonucleotide could compete with Hfq for binding. A set of partially overlapping oligonucleotides (Fig. 4a, a to e), designed to occlude potential Hfq sites near the termination codon, was annealed to RNA. The sensitivity of these complexes to RNase H showed that a presumed partial duplex formed as predicted (data not shown). In several cases (e.g., Fig. Fig.4b,4b, lane 3) the partial duplex exhibited slightly altered mobility. Each partial duplex was incubated with sufficient Hfq to form at least two slower-moving complexes (compare Fig. Fig.1a,1a, lane 3) and analyzed for complex formation (Fig. (Fig.4b,4b, lanes 7 to 12). Oligonucleotide e did not alter the spectrum of cspA mRNA-Hfq complexes (Fig. (Fig.4b,4b, lane 12). In contrast, oligonucleotides a to d largely prevented the formation of the slowest complex (Fig. (Fig.4b,4b, lanes 8 to 11). We conclude that at least one Hfq site is located near the stop codon in cspA mRNA. Since none of the oligonucleotides blocked formation of the least retarded complex, we presume that a hexamer of Hfq is also bound to the 3′ side of the terminator stem-loop (25).
Hfq can protect RNAs against both exo- and endoribonucleases (8, 26). We tested whether Hfq would protect the cspA mRNA from the action of either RNase E or polynucleotide phosphorylase. The preferred site of RNase E cleavage in the cspA mRNA is ~35 nucleotides (nt) from the 3′end (15) and is just 3′ to a presumed site of Hfq binding. Full-length cspA mRNA was almost fully resistant to RNase E cleavage in the presence of 30 nM Hfq, sufficient to form complexes of one to three hexamers per RNA (data not shown). To obtain better resolution, we repeated this experiment with a 5′-truncated cspA mRNA retaining the 3′ 76 residues, designated cspA-Δ5′ RNA, shown in Fig. Fig.5a.5a. This RNA was cleaved by degradosomes to yield products of ~41 and ~35 nt (Fig. (Fig.5b),5b), with half the substrate being converted to products within 5 min. In the presence of 50 nM Hfq, a 2.5-fold excess, however, this substrate became fully resistant to cleavage even after 20 min (Fig. (Fig.5c).5c). Moreover, Hfq formed a protective complex during digestion, as shown in Fig. Fig.5d.5d. In this case, no additional substrate was cleaved after Hfq was added at 10 min following initiation of cleavage by RNase E. As a control we also tested whether ribosomal protein S20, a small, highly basic protein known to interact directly with 16S rRNA (39), would mimic the protective effects of Hfq. RNase E digestion of cspA mRNA was unaffected by the presence of S20 (data not shown).
We performed similar experiments to assess the effects of Hfq on susceptibility of cspA mRNA to polynucleotide phosphorylase (PNPase), using either purified PNPase or degradosomes, in the presence or absence of PAP1 and ATP. In the first experiment, cspA mRNA was incubated with PAP1 and ATP for 10 min with or without Hfq, after which PNPase was added (Fig. 6a and b). In the absence of Hfq (Fig. (Fig.6a),6a), the oligoadenylated substrate (lane 1) was shortened within 8 min to two relatively stable species, one of ~430 nt, the size of cspA mRNA, and a second about 20 nt longer. In the presence of Hfq (Fig. (Fig.6b),6b), the poly(A) tail on cspA mRNA was noticeably longer although a fraction of the initial substrate was not modified (Fig. (Fig.6b,6b, lane 1; compare to a, lane 1). Digestion with PNPase resulted in formation of the same two species noted in the absence of Hfq; however, the majority of the polyadenylated cspA mRNA persisted for the duration of the incubation (compare Fig. Fig.6b,6b, lane 6, to a, lane 6). In contrast, when degradosomes were used as the source of PNPase (as well as RNase E and RhlB) in the experiment shown in panel c, the presence of PAP1 facilitated the degradation of cspA mRNA. Lanes 1 to 5 of Fig. Fig.6c6c show that cspA mRNA underwent a slow cleavage to generate an n ~390 nt product, as previously described (16). In the presence of PAP1 alone, the cspA mRNA was quantitatively elongated by >200 nt (Fig. (Fig.6c,6c, lane 7). It subsequently disappeared, apparently due to the action of both RNase E (as evidenced by the appearance of a 390-nt band in lane 8) and PNPase. The continued activity of PAP1 was demonstrated by the persistence of elongated products in lanes 8 to 10, albeit of decreasing intensity. In the presence of Hfq (Fig. (Fig.6c,6c, lanes 11 to 15), similar intermediates were detected although a fraction of the substrate was not polyadenylated (compare Fig. Fig.6c,6c, lane 12, with b). Polyadenylated cspA mRNA was noticeably longer (~800 nt) in the presence of Hfq. In addition, heterogeneous polyadenylated species persisted somewhat longer in the presence of Hfq (compare lanes 15 and 10 in Fig. Fig.6c6c).
Since cold shock is known to extend the half-life of cspA mRNA and since Hfq protects cspA mRNA in vitro, we asked whether Hfq was required for induction of CspA. We used Northern blotting to compare the steady-state levels of cspA mRNA just prior to cold shock, at 5 min post-cold shock, and at 60 min post-cold shock in strains carrying inactivating mutations in several genes of interest (Fig. (Fig.7).7). In strain MG 1693 (wild type), cspA mRNA was induced an average of 9-fold within 5 min of a shift to 15°C and by 73-fold after 60 min (Fig. (Fig.7).7). The extent to which cspA mRNA was induced in strains carrying mutations in pnp, pcnB, rne, or hfq was similar to that of the wild type (Fig. (Fig.7).7). The extent of induction was somewhat lower in strains carrying two mutations (hfq-1 pnp-7 or hfq-1 ΔpcnB). These strains, however, grew ~40% more slowly than the others, complicating interpretation. At later times after cold shock, the steady-state amount of cspA mRNA was up to 2.5-fold higher in strain SK10023 (hfq-1) than in its wild-type parent, MG1693, continuing for at least 10 h (data not shown). We also tested the viability of SK10023 at 37° and 15°C. At the former temperature this strain grew slightly more slowly than its parent, MG1693, consistent with earlier observations of the strong pleiotropy of the hfq-1 allele (32). Interestingly, the viability of SK10023 was severely compromised at 15°C as colonies did not form even after 5 days (data not shown). Attempts to complement the defect in SK10023 with CspA or Hfq expressed from pBAD28 (13) failed to rescue the cold-sensitive phenotype. This finding suggests that the poor viability of SK10023 at low temperature does not directly involve either hfq or cspA alone. Rather, the low-temperature defect is more likely due to a combination of deficiencies including polar effects of the insertion in the hfq-1 allele on distal genes, as documented earlier (32).
Because Hfq strongly potentiates the action of small regulatory RNAs, it has acquired the reputation of being a global regulator (3, 23, 33). This, in turn, has implied that Hfq exerts specificity in its interactions. Our data show that Hfq can bind relatively tightly to the cspA mRNA in particular and also to the rpsT mRNA and the 9S precursor to 5S rRNA, albeit with lesser affinities. Taken in conjunction with the work of Zhang et al. (38) and Lee and Feig (21), which shows that Hfq can bind a number of RNAs, including tRNAs and their precursors, our findings demonstrate clearly that Hfq exhibits only moderate ability to discriminate among RNAs. Competition experiments show that the affinity of Hfq for cspA mRNA is comparable to its affinity for a small regulatory RNA, RyhB. Since previous work has shown that Hfq is required for the proper function of RyhB in vivo (23), the observed Kd (dissociation constant) for the cspA mRNA-Hfq interaction is physiologically relevant. Nonetheless, the relatively high affinity of Hfq for cspA mRNA is surprising compared to another mRNA of similar size, rpsT mRNA, which also terminates in a Rho-independent terminator. This suggests that care must be taken in interpreting interactions between Hfq and other proteins or complexes as an RNA could provide an effective bridge (19, 25, 27).
The interaction between Hfq and cspA mRNA seems to reflect, in large part, the availability of single-stranded AU-rich pentamers in this mRNA (Fig. (Fig.4a).4a). However, these potential binding sites are not equivalent. First, Hfq binding occurs sequentially rather than simultaneously, as evidenced by the formation of discrete intermediates that can be detected readily in a gel mobility shift assay. Second, the presence of competitor RNAs results in the progressive, sequential conversion of higher-order Hfq-cspA mRNA complexes to lower-order complexes. This is consistent with each site displaying a different intrinsic affinity for Hfq. Third, as little as a 2- to 3-fold molar excess of Hfq is sufficient to inhibit fully the cleavage of cspA mRNA by purified degradosomes. This must reflect full occupancy of a site that precludes access by RNase E to its cleavage site ~35 nt from the 3′ end of the mRNA. This situation is highly reminiscent of the rpsO mRNA where Hfq blocks cleavage at the major RNase E site, M2, which is also distal to the stop codon (8).
Our data are fully consistent with a model in which Hfq facilitates other regulators of the metabolism of cspA mRNA rather than playing an essential role itself. During exponential growth at 37°C, Hfq has been reported to be highly abundant and to be sufficiently concentrated (~15 μM) to saturate substrates with affinities similar to those of cspA mRNA (33). Other reports, however, suggest that Hfq may be more limiting since its abundance is reported to correlate negatively with growth rate (34). In either case, cspA mRNA would be in competition for the pool of Hfq. If one or two hexamers of Hfq were bound to cspA mRNA, our data show that they would be preferentially located at or near its 3′ end (shown schematically in Fig. Fig.8).8). In such a location they would be able to promote polyadenylation (15, 20) and the destabilization of cspA mRNA by 3′ exonucleases, consistent with the observed short half-life of cspA mRNA and the in vitro data in Fig. Fig.6.6. Moreover, the dependence of cspA mRNA turnover on RNase E in vivo is modest at 30°C (16), indicating that blockage of the preferred RNase E site by Hfq would not necessarily provide protection. At the onset of cold shock, cspA mRNA is stabilized dramatically, with a concomitant increase in its translation (2, 10, 11, 13, 29). Recent data show that it undergoes a significant structural rearrangement (9). Although it is unknown whether Hfq is redistributed after a shift to low temperature, the increased abundance of cspA mRNA in the early stages of cold adaptation, coupled with its intrinsic high affinity, would permit it to compete effectively for any free Hfq. Bound Hfq would offer protection against RNase E, enhancing the stabilizing effect of increased translation (Fig. (Fig.8).8). Finally, at the end of the period of cold adaptation and the resumption of (slower) growth, cspA mRNA declines in abundance in a process that requires polynucleotide phosphorylase (37). Hfq bound to rho-independent terminators promotes polyadenylation and the action of PNPase (8, 25), consistent with our data showing that PNPase in degradosomes is capable of digesting polyadenylated cspA mRNA. The finding of preferential binding of Hfq to the 3′ end of cspA mRNA also helps to explain why cspA mRNA is sensitive to polynucleotide phosphorylase at the end of cold adaptation (37).
This work was supported by grant MOP-5396 from the Canadian Institutes of Health Research.
We thank Jon Beckwith, Andrew Feig, Sidney Kushner, and Eric Massé for their generosity in supplying strains.
Published ahead of print on 16 March 2010.