To determine whether Hfq affects the abundance and the decay-rate of the rpsO mRNA in vivo, we determined the stability of the corresponding transcripts in a set of isogenic wt (hfq⁺), hfq^- and Hfq overproducing strains containing pTX367 5, a pGEM3 derivative expressing the hfq gene from its own promoter. The decay-rate of the rpsO mRNA was measured after inhibition of transcription by rifampicin. The more abundant mRNA species detected on Northern blots are the 420 nucleotide monocistronic transcript (P1-t1) and the 502 nucleotide RNase III-processed rpsO-pnp mRNA (P1-RIII). Fig. 1 shows that amounts of both rpsO transcripts are more abundant in the hfq⁺ strain than in the hfq^- mutant. There is in fact 10 times more P1-t1 mRNA in the wild type strain than in the hfq mutant, and 2.2 times more in hfq⁺/pTX367 overproducing cells than in the wild type bacteria transformed with the empty vector (compare time 0 in rifampicin experiments). Surprisingly, this drop in intracellular concentration is associated with stabilization expected to cause an accumulation of the rpsO transcript. Indeed the half-life of the rpsO mRNA is significantly longer in the hfq mutant (1.79 +/- 0.13 min) than in the wild type strain (0.98 +/- 0.12 min), thus confirming that Hfq can activate RNA decay 25. The fact that no difference in stability was observed when Hfq was overproduced (half-life 0.98 +/- 0.16 min) relative to the control strain (half-life 1.17 +/- 0.14 min), suggests that the Hfq level in the wt strain is sufficient to exert a maximum destabilizing effect on the rpsO mRNA decay (Fig. 1 and Table 1). The data above clearly indicate that the drop in rpsO mRNA levels associated with Hfq deficiency does not result from its destabilization. On the contrary, we observed an increase in the rpsO mRNA stability, which should, in theory, result in its accumulation. It means that the drop in the rpsO mRNA level is, in fact, partially compensated by its stabilization.

Previous investigations suggested that Hfq could affect rpsO mRNA stability by modulating both the poly(A)-dependent and the RNase E-mediated degradation pathways. Besides its stimulating effect on the synthesis of poly(A) tails, Hfq was shown to protect poly(A) tails from exoribonucleolytic degradation in vitro and to affect the length and the frequency of oligo(A) tails in vivo 26272829. Moreover, Hfq was found to protect the rpsO mRNA from RNase E cleavage in vitro 29. To evaluate the role of Hfq in the poly(A)- and the RNase E-dependent degradation pathways in vivo, we compared the stability of the rpsO transcript in hfq⁺ and hfq-deficient strains when either of these pathways was inactivated. For this purpose, we used the pcnB null-mutation to fully inactivate poly(A) polymerase and the rne3071 allele allowing inactivation of thermosensitive RNase E at the non permissive temperature.

Fig. 2A shows that Hfq has no significant impact on the decay-rate of the P1-t1 rpsO transcript in pcnB cells. Half lives are 1.32 +/- 0.09 min in hfq⁺ and 1.39 +/- 0.10 min in hfq^- in the first phase of the decay (time 0 to 4 min) and, in contrast to what happens in pcnB⁺ cells, that rpsO mRNA seems to be slightly more stable in the hfq^- strain 4 min after rifampicin addition. Moreover, the drop in the rpsO mRNA concentration resulting from Hfq inactivation is less marked in the absence of PAP I (2.3 times instead of 10 times in the PAP I-containing cells).

These data suggest that Hfq does not significantly affect the RNase E dependent degradation pathway, which accounts for RNA decay in the pcnB null-mutant 40. Moreover, if extended to other strains, this conclusion implies that the stabilization that occurs in hfq^- cells may not result from the impairment of the RNase E-mediated pathway. It is therefore likely, that changes of rpsO mRNA stability and concentration described above require active PAP I; a prediction which prompted us to verify how Hfq affects the decay-rate of the rpsO transcript when thermosensitive RNase E was inactivated.

We have ascertained that polyadenylation contributes to degradation of the rpsO mRNA in the absence of RNase E. As expected, the rpsO P1-t1 transcript was significantly stabilized when thermosensitive RNase E was inactivated 35. Moreover, Hfq had a strong effect on the rpsO mRNA abundance. While the stability of the P1-t1 transcript rose from 6.99 +/- 0.14 min. to 11.18 +/- 0.14 min. the steady-state amount of the transcript was reduced 7.7 times in the absence of Hfq (Fig. 2B). These data reinforce the hypothesis above that Hfq deficiency has a stabilizing effect on the rpsO mRNA, which in theory should cause an accumulation of RNA. However, this stabilization is completely masked by a drop in mRNA concentration whose origin is independent of RNA stability. Moreover, they also confirm that the extent of the impact of Hfq on RNA stabilization and concentration depends on the presence of PAP I in the cell. It is worth noting again, that this effect is also observed in wild type strains where the degradation of the rpsO mRNA is mostly carried out by RNase E. It must be recalled here, that the RNase E and PAP I dependent pathways are somehow related and can substitute for one another 41.

Since mRNAs are generally protected by translating ribosome, we reasoned that the stabilization of the rpsO transcript in the Hfq deficient strains may result from an increase in translation efficiency. One could imagine that the autoregulation loop which adjusts rpsO translation efficiency to the need for S15, may convert the drop in mRNA resulting from Hfq deficiency into a stimulation of translation aiming to compensate for the poor levels of S15 synthesis. Alternatively, Hfq-mRNA interactions could directly affect translation or the autogeneous repression by S15.

These hypotheses led us to look at the behaviour of the constitutively translated rpsO transcript, which does not bear the operator recognizable by S15 and hence whose translation is independent of the S15 yield. This was achieved by using a plasmid-borne rpsO gene in which the first 122 nucleotides of the coding region were deleted, seven of these nucleotides being involved in the formation of the translational operator. This plasmid, referred to as pΔS15AUG, encodes a truncated S15 polypeptide unable to participate in the 30S assembly and to repress translation of the wild-type rpsO message 38. The stabilities of mRNAs originating from the chromosome (rpsO) and from the plasmid (rpsOΔ) were determined in hfq⁺ and hfq1 backgrounds. As was shown previously in wild-type cells the constitutively translated rpsOΔ transcript was more stable than the native rpsO mRNA due to higher translation levels 38. The greater stability of the rpsOΔ transcript was also observed in the hfq mutant (Table 1). Moreover, just as in the case of the wild-type rpsO mRNA transcript originating from the chromosome, Hfq inactivation caused both a drop in the rpsOΔ mRNA level and an increase in its stability (Table 1). These data indicate that the consequences of Hfqdeficiency on mRNA level and stability do not rely on the autoregulation of S15 synthesis. Indeed, the observed stabilization of the unregulated rpsOΔ mRNA in the hfq mutant rules out the possibility that a drop in a functional rpsO mRNA concentration provokes increased translation efficiency through the autoregulation loop, resulting in stabilization of mRNA.

Whether Hfq affects the autoregulation of the genuine rpsO mRNA bearing an intact rpsO operator recognizable by S15 was directly examined by using a specially generated strain in which the rpsO promoter and the whole translation initiation region (TIR) governed synthesis of the chimeric ß-galactosidase from the chromosomal lacZ gene. The hfq⁺ and hfq^- isogenic variants of this strain were obtained by P1 transduction, and then each of them was transformed by a plasmid, pS15, expressing the rpsO gene from its own promoter and thereby serving as a source of additional S15 in a cell, or by a parent vector pACYC184 as a control. The β-galactosidase assay showed that neither the expression of the translationally active gene fusion (cells transformed by pACYC184) nor that of the translationally repressed fusion (cells transformed by pS15) were modified by the hfq mutation (Fig. 3A). Indeed, the impact of S15 protein expression in trans on the rpsO-lacZ fusion was similar in both hfq⁺ and hfq deficient strains (in both cases the repression factor was about 9-10) (Fig. 3A). Thus, Hfq does not seem to be involved in the autogeneous control of S15 synthesis.

As one can notice (Fig. 3B), the reduction of the rpsO mRNA level in the hfq^- strain is more pronounced in strains with empty vector (lanes 1 and 4) than for the pS15 bearing strains (lanes 2 and 5) indicating that the gene dose plays a role in this effect. As expected, the drop in the rpsO mRNA abundance was complemented by the plasmid pTX381 (pHfq), a pACYC 184-derivative expressing the wild-type hfq gene from its own promoter.

It is important to note that the changes in abundance and stability of the rpsO transcripts do not depend on the nature of the inactivating mutation. We found that inactivation of the hfq gene by insertion of a Ω-cassette (hfq1) or by in-frame deletion (hfqΔ) had similar impacts. In contrast, the V43R substitution, which only partially impairs Hfq function 16 did not significantly affect the amount of the rpsO mRNA in the cell, indicating that complete Hfq deficiency is required for the reduction (Fig. 4).

In order to determine whether the effect of Hfq deficiency on the rpsO mRNA could be also observed for other transcripts, we probed total RNA isolated from cells containing wt and mutant hfq alleles for rpsT, rpsB and lpp mRNAs and for the Leu1 and MetY2 tRNAs. While it has been reported that Hfq binds the rpsO mRNA 1629 and tRNAs 1819in vitro, the other three transcripts were not examined. However, the fact that mRNAs from many ribosomal protein operons co-immunoprecipitate with Hfq 20 argues in favor of such a possibility for rpsB and rpsT. The Northern blots revealed that hfq inactivation did not modify lpp mRNA or tRNA abundance (Fig. 4). In contrast, the two rpsT mRNA species were less abundant in the hfq mutant, the stronger effect being observed for the P2-t1 rpsT transcripts initiated at promoter P2. Just as in the case of rpsO, the decreased level of rpsT transcripts was accompanied by a significant increase in their stability (Table 1).

Transcripts of the rpsB-tsf operon also behave like the rpsO transcripts in response to Hfq deficiency. The rpsB and rpsB-tsf mRNAs are transcribed from a single rpsB promoter 43. The monocistronic rpsB transcripts terminate at the attenuator structure in front of tsf, while about one third of the total transcripts read through the attenuator, generating the bicistronic rpsB-tsf mRNA. The levels of both transcripts were significantly decreased by Hfq deficiency (Fig. 4, 5). Moreover, like in the case of the rpsO-lacZ transcript, the abundance of the chimeric rpsB-lacZ mRNA (the transcription product from a rpsB-lacZ translational fusion whose expression was driven by the rpsB promoter and the rpsB TIR) was also lower in the absence of Hfq in a cell (Fig. 5). Because the rpsB-lacZ construction used in these experiments bears a small deletion in the rpsB TIR, which abolishes autogeneous regulation by S2 in trans 43, this effect is not related to changes in translation control.

Interestingly, the rpsB-lacZ fusions produced similar level of β-galactosidase activity in hfq⁺ and hfqΔ isogenic strains (12150 +/- 400 and 11800 +/- 250 units of β-galactosidase in the hfq⁺ and hfqΔ strains respectively). The same was observed for the rpsO-lacZ fusion (Fig. 3A), where the ß-galactosidase activities in hfq⁺ and hfq^- strains were similar. This means that despite the reduced levels of both mRNAs (rpsB-lacZ and rpsO-lacZ) in hfq^- mutant, the relative levels of the corresponding proteins remain about the same, indicating that the reduction in mRNA concentration may be compensated by an increase of translation efficiency or by other compensatory mechanisms. The reduced levels of many transcripts for important cellular proteins (ribosomal proteins, translation factors and others) were observed in a genome-wide transcriptome analysis of a hfq mutant by Guisbert et al. 12. The drop in levels of the transcripts encoding ribosomal proteins may slow down ribosome synthesis, thereby being responsible (at least partly) for the known growth defects of the hfq mutants. In turn, the slower rate of ribosome synthesis should slow down the synthesis of other cellular proteins, so that finally the overall proteome of essential proteins is kept in balance. Thus, as revealed by Western-blotting (Fig. 6), the relative amounts of the ß-galactosidase produced from the rpsO-lacZ fusion gene and polynucleotide phosphorylase (encoding by the pnp transcripts whose level, as will be shown below, is independent of the presence of Hfq) remain about the same in hfq⁺ and hfq^- strains despite a large reduction in the rpsO-lacZ transcript level caused by Hfq deficiency.

Since the above data suggest that changes in mRNA steady-state levels cannot be caused by modifications of mRNA stability or translation efficiency, we suspected that the lack of Hfq might exert a direct negative impact on transcription yield. Previous data have shown that Hfq may act as a DNA binding protein 30 that is able to change superhelicity of plasmid DNA 5. Thus the changes in transcription yield observed in the absence of Hfq could be the consequence of modified DNA topology. If this is the case, the loss of Hfq should change the levels of the pnp mRNA, whose transcription from its own P2 promoter was reported to depend on DNA supercoiling 44. The relative abundance of the pnp transcripts was deduced from the amounts of cDNA generated by primer extension. Since the pnp mRNA is strongly destabilized by a RNase III cleavage in the 5' UTR 34, these experiments were also performed in a RNase III deficient strain (rnc mutant) where RNA concentration is expected to reflect the transcription efficiency of pnp more directly. Fig. 7 shows that Hfq deficiency has no detectable effect on the intracellular concentrations of either the primary transcript initiated at P2 or the transcript processed by RNase III. This contrasts with the bands referred to as M2 and t at the top of the autoradiograph (Fig. 7) which correspond to arrest of reverse transcription at an RNase E processing site and a stable hairpin in the rpsO-pnp bicistronic transcript initiated at the rpsO promoter (P1). Both bands diminished in intensity in hfq strain, confirming the decrease in abundance of transcripts from the rpsO promoter observed by Northern-blot technique. Thus, the lack of Hfq affects transcription yield only from a subset of promoters (e.g. rpsO, rpsB, rpsT), and this negative effect is unlikely to result from changes in DNA topology. Finally, PNPase was used as a control to verify that Hfq does not globally affect protein synthesis (see above).

The name and origin of the strains used here are listed in Table 2. A DNA fragment comprising the rpsO regulatory regions (rpsO promoter and translation initiation region, positions from -188 to +58 with respect to A+1 in the initiator ATG) was amplified from the genomic E. coli DNA and inserted into pEMBLΔ 46/BamHI, HindIII in frame with the lacZ coding sequence 56. Primers used for PCR amplification were: PrpsO-for 5'-

CGTGGATCCTCGTCGCCTGGTGGTTG corresponding to the sequence from -188 to -162 upstream of rpsO and bearing the BamHI site near the 5'-terminus (italicized) and TIRrpsO-rev 5'-CAGAAGCTTGCGTCACGACCAAACTC complementary to the rpsO coding sequence from +40 to +58 and bearing the HindIII site (italicized). The resulting plasmid pErpsO188 was then used to create the ENSO derivative IBrpsO188::lacZ by homologous recombination 56. This strain was used to generate two tetracycline-resistant derivatives hfq⁺ and hfqΔ by P1 transduction as described 16.

The plasmid pS15 was created by cloning the whole rpsO gene flanked with its own promoter and terminator into BamHI and HindIII sites of pACYC184. The DNA fragment to be cloned was generated by PCR on E. coli genomic DNA using rpsO-for corresponding to the positions (-149) to (-126) relative to the rpsO translation start and bearing the BamHI site; 5'-CAGGGATCCGTCTTGCGATAACAG and rpsO-rev, complementary to the positions from +303 to +325 of the rpsO mRNA and bearing the HindIII site 5'-CCGTAAGCTTGAAAAAAGGGGCC.

Construction of strains LABrpsB208::lacZ carrying chromosomal rpsB'-'lacZ fusion with the rpsB portion comprising the rpsB promoter and TIR (positions from -208 to +41 relatively the initiator ATG codon) was described 43. Here, we used a derivative of this strain, LABrpsB208::lacZ ΔGGGU bearing a small deletion in the rpsB 5'-untranslated region (from -72 to -69), which abolishes autoregulation 43. The hfqΔ allele 16 was transferred into this derivative by P1 transduction to generate hfq⁺ anf Δhfq variants.

In the case of IBrpsO188::lacZ and LABrpsB208::lacZ, cell cultures were grown at 37°C in LB medium supplemented if necessary with tetracycline (12 μg/ml) or chloramphenicol (35 μg/ml), in the absence of IPTG (lac-promoter is closed, rpsO or rpsB promoters are active). Cells were harvested in exponential phase (OD₆₀₀≈0.4-0.5). The β-galactosidase activity was measured in clarified cell lysates 54 and expressed in nmol ONPG (o-nitrophenyl-b-D-galactopiranoside) hydrolysed per minute per milligram of total soluble proteins. Protein concentration in lysates was measured by Bradford assay (Bio-Rad).

Cellular proteins (20 μg) were separated on a 12% SDS-PAGE gel and analyzed with the use of ¹²⁵I-iodinated A-protein as described in 16. Polyclonal rabbit antibodies raised against β-galactosidase (ICN Biomedicals, Inc.) and PNPase (kindly provided by A. J. Carpousis) were used.

RNAs were prepared form bacteria grown to an A₆₅₀ = 0.35-0.4, according to 37. Templates for the synthesis of rpsO and rpsT RNA probes were obtained by PCR amplification as described 57. The lacZ template was obtained using TAATACGACTCACTATAGGGATACTGACGAAAC and GCCGTCGTTTTACAACGTC, the lpp probe using TAATACGACTCACTATAGGGTATTTAGTAGCCATGTTG and GTTCTACTCTGCTGGCAG, and the rpsB probe using TAATACGACTCACTATAGGGTTCACGAAGAACTGGTCG and ATGGCAACTGTTTCCATGCG oligonucleotides as primers. The upstream primer includes the T7 promoter indicated in italics. RNA probes were synthesized by T7 RNA polymerase yielding uniformly labeled RNAs with [α-³²P] UTP 26. 5S rRNA, tRNALeu1 and tRNAMetY were probed with the 5'-labelled oligonucleotides 5'-ACTACCATCGGCGCTACGGC, 5'-CCCCCACGTCCGTAAGGACA and 5'-CGGGTTATGAGCCCGACGA, correspondingly.

Total RNA separation on 1% agarose formaldehyde gel or acrylamide/urea gel and Northern blotting were previously described 3435. RNA levels were quantified using a PhosphoImager, and values were corrected for variations in RNA loading by hybridizing the same blot for the 5S rRNA.