Escherichia coli HU protein is a major component of the bacterial nucleoid. HU stabilizes higher order nucleoprotein complexes and belongs to a family of DNA architectural proteins. Here, we report that HU is required for efficient expression of the sigma S subunit of RNA polymerase. This rpoS-encoded alternative σS factor induces a number of genes implicated in cell survival in stationary phase and in multiple stress resistance. By analysis of rpoS–lacZ fusions and by pulse-chase experiments, we show that the efficiency of rpoS translation is reduced in cells lacking HU, whereas neither rpoS transcription nor protein stability is affected by HU. Gel mobility shift assays show that HU is able to bind specifically an RNA fragment containing the translational initiation region of rpoS mRNA 1000-fold more strongly than double-stranded DNA. Together with the in vivo data, this finding strongly suggests that, by binding to rpoS mRNA, HU directly stimulates rpoS translation. We demonstrate here that HU, an abundant DNA-binding, histone-like protein, is able specifically to recognize an RNA molecule and therefore play a role in post-transcriptional regulation.
Escherichia coli HU is a major protein component of the bacterial nucleoid (Rouviere-Yaniv and Gros, 1975; Rouviere-Yaniv, 1978). It was first characterized as a histone-like protein because of its capacity to introduce negative supercoiling into relaxed circular DNA molecules in the presence of topoisomerase I and to compact DNA (Rouviere-Yaniv et al., 1979). The role of HU in modulating the topology of the bacterial chromosome was further substantiated by the observation of a reverse correlation between HU levels and topoisomerase I activity: a decrease in HU concentration is coupled with an increase in relaxing activity, whereas increasing HU concentration decreases topoisomerase I activity (Bensaid et al., 1996). HU binds DNA with low affinity in a sequence-independent manner, but displays a high affinity to some DNA structures such as DNA junctions, nicks, gaps and bulges under stringent conditions (Pontiggia et al., 1993; Bonnefoy et al., 1994; Castaing et al., 1995; Kamashev et al., 1999; Pinson et al., 1999).
We show here that the histone-like protein HU, required by E. coli for survival under starvation, is another factor that participates, like Hfq, dsrA, H-NS, ClpXP and others, in the regulation of RpoS expression. We demonstrate that HU specifically binds an RNA containing the rpoS translation initiation region and that the translation of RpoS is greatly affected in cells lacking HU. These results led us to reconsider the role of the histone-like protein HU as an RNA binder in addition to being a DNA binder.
HU deficiency impairs the expression of RpoS-dependent HPII catalase
HU-deficient cells, the hup mutants, suffer multiple growth defects and show increased sensitivity to various stress conditions such as heat shock, cold shock, osmotic and oxidative stress (Wada et al., 1988; Huisman et al., 1989; M. Wery, L. Claret and J. Rouviere-Yaniv, unpublished). The response to oxidative stress includes the induction of the enzymes HPI and HPII, which catalyse the dismutation of hydrogen peroxide. The same enzymatic activity also accumulates during the stationary phase. One of the two E. coli catalases, HPII, is dependent on RpoS for its expression (Touati et al., 1991). To investigate whether RpoS expression was affected by HU, we therefore examined whether the hup mutants were defective in catalase(s) induction. Figure 1 shows that the activity of the σS-dependent HPII catalase, encoded by the katE gene, is high during the transition to stationary phase in a wild-type background, whereas its activity drops to almost undetectable levels in hupAB mutants. In contrast, the activity of RpoS-independent HPI catalase is not altered in the absence of HU (Fig. 1).
In the absence of HU, cellular σS levels are reduced
Decreased activity of RpoS-dependent HPII catalase in the absence of HU, coupled with other phenotypic changes observed in the hupAB mutants, such as increased osmosensitivity (M. Wery and J. Rouviere-Yaniv, unpublished data), suggested that the accumulation of σS might be perturbed in cells lacking HU. To test this hypothesis, the cellular content of σS was measured in the wild-type strain and in a hupAB mutant by immunoblotting with a σS-specific monoclonal antibody. Figure 2 shows that, as is well established (Lange and Hengge-Aronis, 1994), the level of σS protein gradually increases as E. coli cells approach the stationary phase to reach at least a 10-fold induction level (left lanes). On the contrary, in cells lacking HU, there was a strong decrease in the level of σS accumulation in late-exponential and stationary phase cells in both rich (LB) and minimum (M9) medium (right lanes). To test whether these low levels of σS observed were really caused by the absence of HU and were not just the consequence of secondary mutations accumulating in the hupAB mutant, we introduced into this mutant a plasmid carrying either one or other of the HU-encoding genes. The presence of a functional hup gene restored the accumulation of RpoS to nearly wild-type levels (data not shown). This result confirmed that HU is necessary for normal rpoS accumulation during transition into the stationary phase.
HU affects rpoS expression at a post-transcriptional level
Regulation of rpoS expression is rather complex and proceeds by the control of rpoS transcription, rpoS translation or protein stability (Lange and Hengge-Aronis, 1994). Because the control of rpoS expression is exerted at multiple levels, HU might be involved in one or several of these steps. By monitoring the activity of the transcriptional rpoS742–LacZ and translational rpoS742–LacZ fusions in both the wild-type and hupAB mutant backgrounds, we first eliminated the possibility that HU affects the transcription of rpoS (Fig. 3). The activity of the transcriptional rpoS742–LacZ fusion, although very low as usually observed, is practically unaffected by the absence of HU. In contrast, we observed a substantial decrease in the activity of the translational rpoS742–LacZ fusion in late-exponential and stationary phase hupAB cells, which corroborates the data from immunoblotting experiments (Fig. 3). It can therefore be concluded that HU is involved in post-transcriptional control of rpoS expression. This finding is rather intriguing, as the DNA-binding protein HU was, until now, only known to participate in the regulation of gene expression at the transcriptional level (Aki and Adhya, 1997; Lewis et al., 1999; P. Morales, J. Rouviere-Yaniv and M. Drayfus, submitted).
HU mutants are defective in rpoS translation
RpoS stability is tightly regulated during cell growth as well as under osmotic upshift (Lange and Hengge-Aronis, 1994; Muffler et al., 1996c). The activity of the translational fusion rpoS–742, which contains the amino acid element known to be the primary target for the degradation of RpoS, reflects in consequence both the translation efficiency of rpoS mRNA and σS stability (Muffler et al., 1996c). To discriminate between an effect of HU on these two possible mechanisms, we investigated whether the marked decrease in RpoS content observed in the hupAB mutant could result from increased degradation of the protein. To measure the σS half-life in the presence or absence of HU, we carried out pulse-chase experiments with [35S]-methionine in both the wild-type and hupAB mutant backgrounds. After protein labelling, σS was immunoprecipitated using anti-RpoS monoclonal antibodies (a generous gift from R. Burgess). The results presented in Fig. 4 demonstrate that the residual RpoS molecules found in the hupAB mutant are subject to roughly the same turnover as in wild-type cells. In agreement with previous studies (Lange and Hengge-Aronis, 1994; Schweder et al., 1996), we found that σS is very unstable in exponential cells with a half-life of 3.5 min, independent of the presence of HU. In both wild-type cells and hupAB mutants, RpoS is stabilized gradually during the transition to stationary phase with half-lives of 13 and 16 min, respectively, reaching greatest stability at the onset of starvation with a half-life of more than 30 min in both backgrounds. Therefore, the effect of HU deficiency on σS levels cannot be ascribed to an enhanced σS degradation: clearly, HU does not affect σS turnover.
Finally, to investigate whether HU affects rpoS translation, we measured the rate of RpoS synthesis directly in both strains. Labelling of σS with [35S]-methionine was performed in both strains for 1 min, a time substantially shorter than the half-life of the protein (see Fig. 4), to measure the rate of RpoS synthesis rather than its stability. After immunoprecipitation and autoradiography, we observed that hupAB cells exhibit a very low rate of σS accumulation relative to the otherwise isogenic hupAB+ strain (Fig. 5). As we have shown that the transcription of rpoS is not affected in the hupAB mutant, this low accumulation must reflect a very low rate of translation. Bands corresponding to σS were quantified in relation to [35S]-methionine incorporation into total cellular protein to eliminate any possible effect of hup mutation on the total rate of translation. The rate of σS synthesis in HU-deficient cells was estimated to be about 25% of that in wild-type cells. We conclude that HU is greatly involved in the regulation of rpoS translation.
HU protein specifically recognizes rpoS mRNA in vitro
It has been known for a long time that HU, a DNA-binding protein, shows similar affinity for RNA (Rouviere-Yaniv and Gros, 1975; Y. Dudnick and J. Rouviere-Yaniv, unpublished data). Finding that HU is involved in rpoS regulation at the translational level, we examined whether HU could directly affect rpoS translation by binding to its mRNA. A fragment of rpoS mRNA encompassing the region from −125 nucleotides (nt) to 25 nt relative to the A of the first AUG codon was synthesized by in vitro transcription. The resulting RNA, RNA150, was used in gel mobility shift assays with the HU protein. We selected this RNA fragment because it is thought to be involved in the regulation of the initiation of rpoS translation, as it contains the Shine–Dalgarno sequence and 25 nt downstream from the start codon. RNA150 also contains a sequence complementary to the regulatory DsrA RNA (Majdalani et al., 1998). The RNA150 (rpoS mRNA ‘−125/+25’) retains the same secondary structure as in the whole rpoS mRNA when examined by a program that predicts such structures (data not shown).
The interaction of HU with double-stranded DNA is strongly salt dependent. At high salt (200 mM NaCl), HU has a very low affinity for double-stranded DNA, and HU–DNA complexes migrate as smears in polyacrylamide gels (Kamashev and Rouviere-Yaniv, 2000). In contrast, HU binds 1000-fold more strongly to some DNA structures, e.g. nicked DNA, and these specific complexes migrate as a sharp band. Moreover, HU is able to recognize these specific DNA structures, e.g. nicked DNA, in the presence of a large excess of non-specific double-stranded DNA. We performed similar gel mobility shift assays to test the specificity of the binding of HU to RNA at high salt concentration (stringent conditions). This permits clear distinction between strong/salt-independent and weak/salt-dependent interactions. Figure 6A shows that HU binds RNA150 under stringent conditions (200 mM NaCl) and forms two major complexes with this RNA. To test the specificity of this binding, we performed the gel shift retardation assay in the presence of a 5′ end-labelled specific substrate, a nicked DNA. Under identical experimental conditions, both HU–RNA150 and HU–nicked DNA complexes are formed at the same protein concentration (Fig. 6B). This experiment allows direct comparison between the dissociation constants of HU complexed to either RNA or nicked DNA. These were calculated to be 12 nM and 10 nM respectively. Similar assays were performed in the presence of 5′ end-labelled 40 bp non-specific double-stranded DNA (Fig. 6C). No complexes between HU and double-stranded DNA were detected, whereas HU binds strongly to RNA150 under salt conditions (200 mM NaCl) close to the physiological ionic strength of the cell. Hence, HU binds rpoS–RNA as it binds nicked DNA, around 1000 times stronger than double-stranded DNA (Castaing et al., 1995, Kamashev et al., 1999; Pinson et al., 1999). These results provide evidence that HU interacts with RNA150 with high specificity.
The data presented here demonstrate that HU stimulates the expression of the rpoS-encoded σS subunit of RNA polymerase. Multiple controls of rpoS expression are exerted at the levels of transcription, translation and protein stability (Lange and Hengge-Aronis, 1994). Using a combination of pulse-chase experiments and analysis of rpoS–lacZ fusions, we show that HU deficiency interferes with rpoS translation. Although the activity of the rpoS–lacZ transcriptional fusion was unchanged in the absence of HU, the rpoS–lacZ translational fusion displayed greatly reduced activity in the hupAB mutant background. As the activity of this rpoS translational fusion reflects both the efficiency of translation and the degradation of the protein, we examined RpoS stability. Pulse-chase experiments followed by immunoprecipitation of σS revealed that, in hupAB mutants, σS is subject to the same regulation of turnover as in wild-type cells. Moreover, labelling of total cellular proteins followed by immunoprecipitation of σS demonstrated that specific incorporation of labelled methionine into σS, that is the efficiency of σS synthesis compared with total protein synthesis, was reduced in cells lacking HU. Together, these data demonstrate that HU participates in the regulation of rpoS translation.
The absence of HU leads to pleiotropic phenotypes and renders the cells extremely sensitive to various stress conditions (Wada et al., 1988; Huisman et al., 1989; Boubrik and Rouviere-Yaniv, 1995; Painbeni et al., 1997; M. Wery and J. Rouviere-Yaniv, in preparation). Some of these phenotypes could be explained, at least in part, by a deficiency in σS expression in cells lacking HU. We found that the expression of stationary phase-inducible HPII catalase, encoded by the RpoS-dependent katE gene, is greatly reduced in the absence of HU. This may account for the loss of adaptation to oxidative stress in hupAB cells by the indirect influence of HU on HPII synthesis through the regulation of rpoS expression.
To date, HU has been shown to participate in the formation of higher order DNA–protein complexes that modulate transcriptional regulation and recombination (Lavoie et al., 1996; Aki and Adhya, 1997; Lewis et al., 1999). We demonstrate here that HU is also able specifically to recognize an RNA molecule. Gel mobility shift assays revealed that HU binds strongly to a segment of rpoS mRNA bearing the translation initiation site and forms two strong complexes with this RNA molecule. Stringent conditions of binding (200 mM NaCl) make it possible to distinguish between specific and non-specific HU binding for DNA. By analogy with HU binding to DNA, we can postulate that, under identical stringent conditions, HU recognizes specific structures on RNA. Our results here demonstrate that, in these stringent conditions, HU exhibits a 1000-fold higher affinity in its binding to rpoS mRNA than to double-stranded DNA. Similarly, under stringent conditions, it displays no affinity to double-stranded RNA (A. Balandina and J, Rouviere-Yaniv, unpublished data). Effectively, the dissociation constant of the HU–rpoS RNA complex is 12 nM under stringent conditions, very similar to the 10 nM found for nicked DNA, which is one of the specific structures recognized by HU. It was shown recently that HU specifically recognizes most of the DNA repair and recombination intermediates and, among them, a number of DNA junctions (Kamashev and Rouviere-Yaniv, 2000). RNA secondary structure motifs are highly varied and constitute, together with double-stranded RNA, the major part of RNA molecules. According to our computer modelling of rpoS mRNA secondary structure, HU may recognize a three-way RNA junction located at or near the translational initiation region. However, the precise element defining HU-specific binding to RNA remains to be elucidated. It is, however, easy to imagine that HU will bind specifically a number of other RNAs sharing with rpoS mRNA this RNA-binding motif. Interestingly, the HU protein of Bacillus subtilis (Hbsu) has recently been shown to bind specifically to an Alu domain of small cytoplasmic RNA (scRNA), a component of an SRP-like particle (Nakamura et al., 1999).
The strong probability that HU binds rpoS mRNA in vivo based on the finding of its specific interaction in vitro raises the question as to how HU is distributed in the bacterial cell over its numerous substrates. The quantity of HU protein has been determined by us to be in the order of 30 000 dimers cell−1 at the beginning of starvation (Rouviere-Yaniv and Kjeldgaard, 1979; J. Rouviere-Yaniv, unpublished data). This corresponds to between 20 µM and 60 µM HU in the cell. Even though the concentration of free HU is lower on account of its binding to DNA, HU should certainly bind rpoS mRNA in vivo, as its binding constants for rpoS mRNA and nicked DNA are almost identical (12 nM versus 10 nM). This is in agreement with the in vivo data, which shows that, in the absence of HU, rpoS translation proceeds much less efficiently, suggesting that HU stimulates rpoS translation directly by binding to rpoS mRNA.
Two other abundant proteins have been shown to influence rpoS translation: an RNA-binding protein, Hfq, essential for rpoS translation (Brown and Elliot, 1997; Muffler et al., 1996b); and a histone-like, DNA-binding protein, H-NS (Barth et al., 1995; Yamashino et al., 1995), which participates in σS degradation and represses rpoS translation. A fuller picture of RpoS regulation is beginning to emerge, with abundant proteins such as HU, Hfq and H-NS all being implicated in rpoS regulation in response to environmental signals.
HU and H-NS are both abundant DNA-binding proteins with no sequence specificity that could influence gene expression by modulating transcription. It seems that these two nucleoid-associated proteins play opposite roles in the regulation of gene expression. For example, RscA, the positive regulator of the capsular polysaccharide synthesis, is downregulated by H-NS. DsrA RNA counteracts this silencing by activating rcsA transcription (Sledjeski et al., 1996), and HU functions in the same direction by stimulating RcsA synthesis (Painbeni et al., 1993). Similarly, HU and H-NS act antagonistically in the regulation of OmpF, the outer membrane porin. Whereas HU increases the steady-state level of micF antisense RNA, the negative regulator of OmpF (Painbeni et al., 1997), H-NS acts as a transcriptional repressor of OmpF (Deighan et al., 2000).
Here, we show that HU, as already established for H-NS, exerts an influence over the control of rpoS expression by participating in the regulation of translation, although playing an opposite role. H-NS has been shown to bind RNA non-specifically and to have the properties of an RNA chaperone, but there are no data about specific H-NS/RNA binding (Falconi et al., 1988; Cusick and Belfort, 1998). The published studies leave open the question as to whether H-NS interferes directly or indirectly with the initiation of rpoS translation by influencing the transcription of an unknown factor or by another mechanism.
On the other hand, there is strong evidence that Hfq interacts with rpoS mRNA in vivo (Zhang et al., 1998) but, at the moment, there are no reports of Hfq binding to rpoS in vitro. We show that HU binds an rpoS mRNA fragment containing the translation initiation region. Two possible mechanisms as to how HU stimulates rpoS translation can be proposed: either HU modifies the RNA secondary structure to facilitate ribosome binding, or it modulates binding of other proteins such as Hfq or H-NS, or both. Yet, how proteins control rpoS translation and what is the link between the histone-like, abundant proteins HU, H-NS and Hfq are issues that will have to be clarified in order to understand the molecular mechanism(s) that control RpoS expression during the growth cycle of E. coli.
Bacterial strains and growth conditions
The bacteria used in this study were E. coli K-12, C600 (F−thr leu tonA rpsL supE lac thr leu tonA rpsL supE lac) and its hupAB derivative, hupA::Cmr, hupB::Knr (Huisman et al., 1989). RpoS fusions λRZ5:rpoS742–lacZ and λRZ5:rpoS742–lacZ (hybrid) (Lange and Hengge-Aronis, 1994) were introduced into the C600 and C600hupAB background by lambda infection, and single lysogens were verified by polymerase chain reaction (PCR) (Powell et al., 1994). Cultures were grown at 37°C under aeration in LB rich medium or M9 minimal medium, supplemented with 0.2% glucose, 0.005% vitamin B1 and amino acids (400 µg ml−1). Growth was monitored by measuring the optical density at 600 nm (OD600). The growing cultures were inoculated from freshly isolated colonies streaked on LB plates from the stock cultures stored at −70°C.
Catalase activity determination
Cells were grown to stationary phase in M9 minimal medium, and equal amounts of total cellular protein were loaded onto a 10% non-denaturing polyacrylamide gel. Negative staining for HPII catalase activity was performed as described by Touati et al. (1991).
β-Galactosidase activity was assayed using ONPG as a substrate (Miller, 1972).
SDS–PAGE and immunoblot analysis
To determine the cellular concentration of σS, cultures were grown in LB medium or in M9 minimal medium, and the cells were centrifuged briefly and resuspended in Laemmli buffer (0.01OD600 ml−1). Samples containing 10 µg of total cellular protein were separated on SDS−12% polyacrylamide gels and transferred to nitrocellulose membrane filters by electroblotting (Schleicher and Schuell). The blots were blocked for 1 h with 5% low-fat milk, washed and probed with a 1:1000 dilution of the anti-RpoS monoclonal antibody 1RS1 (Nguyen et al., 1993) and then with a 1:1000 dilution of anti-mouse IgG alkaline phosphatase conjugate (Promega). Detection was by chromogenic substrate (BCIP/NBT; Sigma).
Pulse labelling of cells and immunoprecipitation
Cells were grown in M9 minimal medium containing 0.2% glucose, 0.005% vitamin B1, methionine (2.4 µg ml−1) and other amino acids (400 µg ml−1). At the indicated optical densities, the culture was labelled with l-[35S]-methionine (Amersham; > 1000 Ci mmol−1) at 30 µCi of culture ml−1 for 3 min at 37°C. The sample was chased with 10 mM non-radioactive methionine, and the samples were removed at intervals, precipitated with 10% TCA and immunoprecipitated as described previously (Ito et al., 1981; Itoh et al., 1993; Rockabrand et al., 1998). For determination of the rate of σS synthesis, the samples were labelled for 1 min and immunoprecipitated immediately. After immunoprecipitation, the proteins were separated onto the SDS−10% polyacrylamide gel and detected by autoradiography. The intensity of the bands was quantified with a Molecular Dynamics Phosphorimager (400S).
The RpoS DNA fragment from −125 to +25 nt relative to the A of the first AUG codon was amplified by PCR reaction using pDTS-1 plasmids as a template with primers T1E (5′-TA CTT GAA TTC CAT TTT GAA ATT CGT TAC AAG GGG AAA) and T1H (5′-CT GTG TAC AAG CTT GAA CTT TCA GCG TAT TCT GAC TCA TAA). The DNA fragment was digested with EcoRI and HindIII and cloned under the control of the T7 promoter into pGem3Z (Promega). The [α-32P]-RNA was prepared from plasmid linearized by HindIII by in vitro transcription. Before use, the RNA was renatured by incubation at 65°C for 5 min and cooling on ice.
DNAs were constructed according to the method of Kamashev et al. (1999). The oligonucleotides used for the construction of nicked DNA are (from 5′ to 3′):
(a) AGTCTAGACT GCAGTTGAGT CCTTGCTAGG ACG GATCCCT
(c) pAGGGATCCGTCCTAGCAAGG, 5′ phosphorylated.
Double-stranded DNA was constructed from oligonucleotides a and d:
Gel mobility shift assay
HU protein was purified from E. coli strain C600 as described by Pellegrini et al. (2000). Varying amounts of HU protein were incubated with either internally 32P-labelled RNA150 (2 fmol) and [5′-32P]-nicked DNA (2 fmol) or with dsrA–RNA150 hybrid (7 fmol) for 15 min at room temperature in 15 µl of the binding buffer BB (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 0.1 mM EDTA, 0.05 mg ml−1 bovine serum albumin, 10% glycerol). Samples were loaded into 7% polyacrylamide gels (30:1) prerun for 2 h, electrophoresed at 4°C in 90 mM Tris-borate, pH 8.6, and subjected to autoradiography.
Determination of dissociation constant
An apparent dissociation constant was calculated as Kd = P0 × F/B, where P0 is the total concentration of the protein, and F and B are the concentrations of free and bound RNA, respectively, quantified in arbitrary units using a Molecular Dynamics Phosphorimager (400S). This equation assumes that the concentration of free protein is equal to the concentration of total protein in the assay, that is the RNA concentration is negligible compared with the total concentration of the protein.
MRNA secondary structure predictions
mRNA secondary structure predictions were performed according to the method of Zuker (1989) with the mulfold and loopviewer programs developed by Don Gilbert, Indiana University, Bloomington, IN, USA.
We are especially grateful to Richard Burgess and Nancy Thompson for the generous gift of the anti-RpoS monoclonal antibodies 1RS1, and to Danielle Touati for an introduction to the catalase activity determination technique and for the gift of the pDT5-1 plasmid. We thank O. Pellegrini, L. Zig, J. Oberto, M. Wery, D. Kamashev and J. Plumbridge for their help and numerous discussions. A.B. was the recipient of PhD fellowships from successively INTAS (short fellowship for young NIS scientists) and EMBO (short-term fellowship). L.C. is grateful to la Societe des Amis des Sciences for a short-term fellowship. This work was supported by the CNRS (UPR 9073) and grants from l’Association de la Recherche contre le Cancer (ARC 99-00), EDF (contrat RB97-21 & RB98-34) and le Programme en Microbiologie (MENESR).
Present addresses: †Engelhardt Institute of Molecular Biology, Vavilov Str. 32, 117984 Moscow, B-334 Russia.
‡Department of Pathology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QP, UK.