Genome-wide transcriptome analyses of several bacterial species have recently uncovered a hitherto unappreciated amount of antisense transcription. However, the physiological role, regulation and significance of such antisense transcripts are presently unclear. We now report the identification of a cis-encoded 1.2 kb long antisense RNA – termed AmgR – that is complementary to the mgtC portion of the mgtCBR polycistronic message from Salmonella enterica. The mgtCBR mRNA specifies the MgtC protein, which is necessary for survival within macrophages, virulence in mice and growth in low Mg2+; the Mg2+ transporter MgtB with no apparent role in virulence; and the membrane peptide MgtR mediating MgtC degradation. Expression of AmgR dimished both MgtC and MgtB protein levels in a process requiring RNase E but independent of RNase III, the RNA chaperone Hfq, and the regulatory peptide MgtR. Inactivation of the chromosomal amgR promoter increased MgtC and MgtB protein levels and enhanced Salmonella virulence. Surprisingly, AmgR transcription is governed by the regulatory protein PhoP, which also directs transcription of the sense mgtCBR mRNA. AmgR may function as a timing device that alters MgtC and MgtB levels after the onset of PhoP-inducing conditions.
The transcriptome profile of bacteria can be exceedingly complex. In addition to the mRNAs corresponding to annotated open reading frames and the non-coding RNAs of known function such as rRNAs, tRNAs and regulatory small RNAs, vast amounts of cis-encoded antisense transcripts have been recently uncovered in a variety of bacterial species including Helicobacter pylori (Sharma et al., 2010), Escherichia coli (Mendoza-Vargas et al., 2009), Vibrio cholerae (Liu et al., 2009) and Listeria monocytogenes (Toledo-Arana et al., 2009). Unfortunately, the conditions that govern the synthesis of these antisense RNAs, as well as their physiological role and mechanism of action remain largely unknown. We now report the identification of a novel antisense RNA that, surprisingly, is under direct control of the transcription factor promoting expression of the sense RNA that it regulates.
In this paper, we describe the identification and characterization of a 1.2 kb long cis-encoded antisense RNA that is transcribed from a promoter located in the unusually long mgtC–mgtB intergenic region. We demonstrate that, unexpectedly, the regulatory protein PhoP directs transcription of both the sense mgtCBR transcript and the newly identified antisense RNA. The antisense RNA is critical for the production of physiological amounts of the MgtC and MgtB proteins because altering expression of the antisense RNA affected Salmonella's ability to grow in low Mg2+ and to cause a lethal infection in mice.
Identification of a PhoP-activated antisense RNA complementary to the mgtC mRNA
The mgtC and mgtB coding regions are separated by 219 nucleotides (Fig. 1B), raising the possibility of the downstream mgtB gene being transcribed from a separate promoter than the upstream mgtC gene. However, mgtB expression appears to be strictly dependent on the promoter located upstream of the mgtC gene because a MudJ transposon insertion in the mgtC gene or a four-nucleotide substitution in the PhoP binding site in the mgtC promoter abolished expression of both the MgtC and MgtB proteins (Fig. S2).
To gain further insight into the regulation of the mgtCBR operon, we carried out Northern blot analysis with RNA harvested from wild-type Salmonella grown in minimal media with different Mg2+ concentrations. We observed multiple bands in samples corresponding to bacteria grown in very low (i.e. 10 µM) Mg2+, but not following growth in high Mg2+, when we used a double-stranded DNA probe corresponding to the leader region preceding the mgtC open reading frame (Fig. S3). We then re-probed the blots using strand-specific riboprobes corresponding to the same region and found multiple bands when examining the sense strand (data not shown) but a single 1.2 kb long transcript corresponding to the antisense strand for mgtC. The latter transcript was produced at maximal levels following growth in very low (i.e. 10 µM) Mg2+ and could not be detected in samples corresponding to bacteria grown in high Mg2+ (Fig. 2A). In agreement with these results, S1 nuclease protection assays identified a transcript originating in the mgtC–mgtB intergenic region that was produced when Salmonella was grown in 10 or 25 µM Mg2+ but not in 1 mM Mg2+ (Fig. 2B).
To map the 5′ end of the identified antisense RNA, we used 5′ RACE (rapid amplification of cDNA ends) and obtained three different transcription start sites in five independent clones. One site is present 145 nucleotides downstream of the mgtC stop codon, corresponding to the mgtC–mgtB intergenic region and in the same general area found in the S1 mapping experiment (Fig. 2B). The other two 5′ RACE products are located within the mgtC coding region 148 and 154 nucleotides upstream of its stop codon; their origin and significance remain unknown as products of the corresponding sizes were not detected by Northern blot or S1 mapping. Thus, during growth in very low Mg2+Salmonella produces an antisense transcript that initiates in the mgtC–mgtB intergenic region (Fig. 1B) and extends beyond the 5′ end of mgtC coding region (Fig. 1A). We termed this transcript AmgR for antisense for mgtC regulator.
We reasoned that AmgR was likely to be transcriptionally controlled by the PhoP protein given that PhoP's cognate sensor PhoQ is activated during growth in low Mg2+ (Garcia Vescovi et al., 1996). In agreement with this prediction, S1 mapping demonstrated that the AmgR product detected in wild-type Salmonella grown in 10 or 25 µM Mg2+ was not present in the isogenic phoP mutant (Fig. 2B). Still, because PhoP is required for transcription of the sense mgtCBR transcript (Soncini et al., 1996), it was possible that the phoP mutant lacked AmgR because of AmgR destabilization resulting from the absence of the mgtCBR mRNA with which it is largely complementary (Fig. 1A). Indeed, AmgR was not detected in a Salmonella strain with a four-nucleotide substitution in the PhoP box of the mgtC promoter that abolishes mgtC expression (Fig. 2B). Therefore, we re-evaluated the possibility of PhoP being necessary for AmgR transcription initiation by carrying out S1 experiments with RNA harvested from four isogenic strains with a mutation inactivating the chromosomal copy of the mgtC promoter, a wild-type or mutant phoP gene, and carrying either the plasmid vector pUH21-2lacIq or plasmid pmgtC, where the mgtC gene is transcribed from a derivative of the lac promoter that is PhoP independent, and which does not include the sequences corresponding to the amgR promoter. AmgR was produced by the phoP+ strain carrying plasmid pmgtC when grown in the presence of the gratuitous inducer of the lac promoter IPTG but not in its absence (Fig. 2C). Conversely, AmgR was not detected in the phoP mutant strain carrying pmgtC regardless of whether IPTG was added to the culture (Fig. 2C). AmgR was also absent from the strains harbouring the plasmid vector pUH21-2lacIq (Fig. 2C). These data define AmgR as a novel PhoP-dependent transcript that is apparently unstable in the absence of the complementary mgtCBR mRNA.
AmgR transcription from a heterologous promoter silences MgtC and MgtB expression
We initially compared the levels of the MgtC and MgtB proteins in isogenic Salmonella strains harbouring either plasmid pBAD-amgR, where AmgR is expressed from a heterologous promoter, or the plasmid vector pBAD18. Western blot analysis revealed that both MgtC and MgtB were produced in the vector-harbouring cells grown in 10 µM Mg2+ but not when grown in 25 µM, 50 µM or 1 mM Mg2+ (Fig. S4A), reflecting that transcription of the mgtCBR operon is regulated by the levels of Mg2+ in the media (Snavely et al., 1991; Soncini et al., 1996). By contrast, neither MgtC nor MgtB were detected in bacteria carrying pBAD-amgR at all investigated Mg2+ concentrations (Fig. S4A).
To examine the specificity of AmgR regulatory action, we introduced pBAD-amgR (or the vector control pBAD18) into isogenic strains harbouring chromosomal lac transcriptional fusions in either the mgtB or mgtA coding regions. (The mgtA gene encodes a protein that is 50% identical to MgtB and is transcribed from a PhoP-activated promoter located elsewhere in the genome.) The mgtB-lac strain produced less β-galactosidase activity when carrying pBAD-amgR than when harbouring the plasmid vector (Fig. S4B). This effect appears to be specific for mgtB-lac because the mgtA-lac strain exhibited slightly higher β-galactosidase activity when harbouring pBAD-amgR than with the plasmid vector at intermediate Mg2+ concentrations (Fig. S4C). The higher mgtA-lac expression displayed by the strain with pBAD-amgR may reflect derepression of the Mg2+-responding riboswitch in the mgtA leader (Cromie et al., 2006) possibly caused by a decrease in cytosolic Mg2+ resulting from AmgR-promoted lowering of MgtC and/or MgtB protein levels (Fig. S4A).
The AmgR-expressing plasmid pBAD-amgR impaired growth of wild-type Salmonella in 10 µM Mg2+ (Fig. S4D). This likely reflects downregulation of MgtC levels (Fig. S4A) and that MgtC is required for growth in low Mg2+ (Blanc-Potard and Groisman, 1997). Importantly, the pBAD-amgR-carrying bacteria grew as well as those with pBAD18 at 1 mM Mg2+ (Fig. S4E), indicative that AmgR's inhibitory effect is limited to low Mg2+ conditions. These results demonstrated that AmgR can function in trans to diminish the amounts of the MgtC and MgtB proteins.
Inactivation of the amgR promoter enhances MgtC and MgtB expression and growth in low Mg2+
The phenotypes described above could be due to AmgR being expressed at non-physiological levels from a heterologous promoter in a multi-copy number plasmid. Thus, to test the role of AmgR when produced at normal levels, we constructed a strain with a three-nucleotide substitution in the chromosomal copy of the predicted −10 region of the amgR promoter (Fig. 1B), which, as anticipated, failed to produce AmgR (Fig. S5). We then compared the amounts of MgtC and MgtB proteins present in crude extracts from wild-type and amgR strains grown at three different Mg2+ concentrations for 4 and 5 h. The amgR promoter mutation had two striking effects: first, expression of the MgtC protein following 5 h growth in 10 µM Mg2+ when none is detected in the wild-type strain at this time (Fig. 3A); and second, expression of both the MgtC and MgtB proteins in bacteria grown in 50 µM Mg2+ for 5 h, a condition under which these proteins are not observed in the wild-type strain (Fig. 3A and B).
The amgR promoter mutant grew better than the wild-type strain in 10 µM Mg2+ (Fig. 3C), suggesting that MgtC is a limiting factor for Salmonella's ability to grow in low Mg2+. By contrast, wild-type and amgR promoter mutant strains grew similarly in 1 mM Mg2+ (Fig. 3D), a condition where neither MgtC nor MgtB are produced at detectable levels (Fig. 3A and B). Cumulatively, these data indicate that AmgR normally downregulates the levels of the MgtC protein, and to lesser extent of the MgtB protein.
AmgR decreases MgtC and MgtB protein levels in an mgtR-independent fashion
We considered the possibility that AmgR action might be exerted via MgtR because this peptide promotes the FtsH-mediated degradation of the MgtC protein (Alix and Blanc-Potard, 2008) and because AmgR preferentially affects MgtC protein levels (Fig. 3A and B). Thus, we compared the MgtC and MgtB protein levels in isogenic wild-type and mgtR strains harbouring either plasmid pBAD-amgR or the vector pBAD18. MgtC and MgtB levels were lower in cells carrying pBAD-amgR than in those with the vector control, and this was true for both wild-type and mgtR Salmonella (Fig. S6A). As expected, the levels of the MgtC protein were higher in the mgtR mutant than in the wild-type strain in cells harbouring the plasmid vector (Fig. S6A), and the levels of the MgtB protein were not affected by the presence of mgtR (Fig. S6A). These results indicated that the AmgR-promoted decrease in MgtC and MgtB protein levels is not mediated by mgtR.
AmgR promotes decay of the mgtC RNA in an RNase E-dependent fashion
Regulatory RNAs use a variety of mechanisms to modulate gene expression including degradation of their mRNA targets (Krinke and Wulff, 1990; Waldbeser et al., 1995) and interference with translation of open reading frames in the mRNAs (Brantl, 2002; Gerdes and Wagner, 2007; Weaver, 2007). We determined that there were lower amounts of mgtC mRNA in wild-type Salmonella than in the amgR promoter mutant following 5 h growth in 10 µM Mg2+ (Fig. S7A). No differences were observed following growth in 10 µM Mg2+ for 2 h or in 10 mM Mg2+ for 2 or 5 h (Fig. S7A). The mRNA levels corresponding to mgtA, which was used as control, were similar in wild-type and amgR Salmonella under all investigated conditions (Fig. S7B).
To examine whether AmgR affected the stability of the polycistronic mgtCBR message, we harvested RNA from bacteria experiencing inducing PhoP conditions and then treated with rifampicin to inhibit gene transcription. In the wild-type strain, the mgtC portion of the message was less stable than the mgtB portion (Fig. 4A and B). This is due, at least in part, to AmgR because 89% of the mgtC transcript remained in the amgR promoter mutant 1 min after addition of rifampicin whereas only 59% was still present in the wild-type strain (Fig. 4A). AmgR appears to target primarily the mgtC mRNA because the wild-type strain and the amgR promoter mutant displayed similar decay rates for the mgtB mRNA (Fig. 4B) and for the mgtA mRNA (Fig. 4C), which was used as control. While the amgR promoter mutant harbours higher levels of MgtB protein than the wild-type strain when grown in 50 µM Mg2+ for 5 h (Fig. 3A), the decay rates for the mgtB mRNA have only been determined following growth in 10 µM Mg2+ for 3.5 h (Fig. 4B) and could be different under other conditions.
We wondered whether the AmgR-promoted degradation of the mgtC mRNA was mediated by RNase E, which is a major player in mRNA turnover (Deutscher, 2006). Because RNase E is essential, we compared the mgtC mRNA levels in a set of four isogenic strains with a wild-type or mutant amgR promoter, and harbouring the wild-type or a ts allele of rne (Figueroa-Bossi et al., 2009). Organisms were grown at the permissive temperature of 30°C overnight and then transferred into inducing medium (N-minimal medium containing 10 µM Mg2+) at 37°C to partially inactivate the rne gene product [We used 37°C instead of the preferred temperature of 42°C because the rne mutant failed to grow in the defined minimal medium at 42°C, as previously reported (Spinelli et al., 2008)]. We then harvested RNA at different times and examined the transcripts by Northern blot hybridization using a probe specific for the mgtC gene.
The decrease in mgtC RNA levels taking place in the wild-type strain between 3.5 and 5 h post induction at 37°C requires both AmgR and RNase E because neither the rne and amgR single mutants nor the rne amgR double mutant (Fig. 4D) exhibited the dramatic decrease in mgtC mRNA levels displayed by the isogenic rne+ amgR+ strain (Fig. 4D). More mgtC mRNA remained in the amgR mutants than in the isogenic amgR+ strains grown at the permissive temperature of 30°C (Fig. 4E; note that the AmgR-promoted decrease in mgtC mRNA required a longer incubation at 30°C), reinforcing the notion that the RNase E-dependent degradation of the mgtC mRNA requires AmgR.
The lower MgtC protein levels resulting from AmgR-promoted destabilization of the mgtC mRNA is RNase III-independent because it was still observed in an rnc mutant (Fig. S6B). Likewise, AmgR retained the ability to reduce MgtC protein levels in an hfq mutant (Fig. S6C), lacking the RNA chaperone that is essential for the activity of many trans-acting sRNAs (Wagner and Flardh, 2002; Chen et al., 2004; Katayama et al., 2005; Storz et al., 2005; Waters and Storz, 2009). Cumulatively, these data indicate that the AmgR-dependent decrease in mgtC mRNA levels requires RNase E but neither RNase III nor Hfq.
Differential PhoP binding to the amgR and mgtCBR promoters
The PhoP protein typically binds to an hexanucleotide direct repeat separated by five nucleotides – designated PhoP box – present in the promoters of its regulated targets (Zwir et al., 2005). Analysis of the amgR promoter region revealed the presence of a sequence resembling one half of a PhoP box (Fig. 1B). This raised the possibility that PhoP may regulate amgR transcription directly; and if that were to be the case, PhoP was likely to bind to the amgR promoter with less affinity than to the mgtC promoter, which has a complete PhoP box with excellent match to the PhoP box consensus sequence (Zwir et al., 2005).
We first explored PhoP binding to the amgR and mgtC promoters in vivo by chromatin immunoprecipitation using a strain expressing the PhoP-HA protein from the normal phoP promoter and chromosomal location (Shin and Groisman, 2005). PhoP occupied the mgtC promoter 2 h after Salmonella was switched from high to low Mg2+ media (Fig. 5A). By contrast, PhoP binding to the amgR promoter took place only after 4 h (Fig. 5B). This appears to reflect that there was more phosphorylated PhoP protein at 4 h than at 2 h post induction (Fig. S8).
Electrophoretic mobility shift assays demonstrated that larger amounts of purified PhoP-His6 protein were necessary to shift a DNA fragment harbouring the amgR promoter than one carrying the mgtC promoter (Fig. 5C). The observed shifts were specific to the mgtC and the amgR promoter regions because they could be competed out by the corresponding unlabelled DNA fragments but not by a non-specific competitor (Fig. 5C). Furthermore, a DNA fragment consisting of one half of the amgR promoter region could not be shifted by the same amount of PhoP-His6 protein (data not shown). These data demonstrated that the PhoP protein controls AmgR expression directly, and that it binds to the amgR promoter with less affinity than to the mgtC promoter.
AmgR regulates Salmonella virulence
Because MgtC is required for virulence in mice (Blanc-Potard and Groisman, 1997) and because AmgR reduces MgtC protein levels (Fig. 3A), we reasoned that changes in AmgR expression might impinge on Salmonella's pathogenicity. To test this notion, we carried out two sets of experiments that probed the effect of obliterating or overexpressing AmgR. First, we inoculated mice intraperitoneally with ∼103 of either wild-type Salmonella or the isogenic amgR promoter mutant and followed mouse survival for 3 weeks. Only 40% of the mice inoculated with the amgR promoter mutant survived compared with 80% of those inoculated with wild-type Salmonella (Fig. 6A). Moreover, the mice inoculated with the amgR promoter mutant died 5–7 days before those inoculated with the wild-type strain (Fig. 6A). And second, we inoculated mice intraperitoneally with ∼2 × 104 of wild-type Salmonella harbouring either the AmgR-expressing plasmid pBAD-amgR or the vector pBAD18. All the mice inoculated with the strain harbouring pBAD-amgR survived whereas 100% of the mice inoculated with the strain harbouring the control plasmid vector died within 2 weeks (Fig. 6B). (We used a larger bacterial dose than in the previous experiment to better evaluate the protective effect of AmgR overexpression.) These results showed that tight control of MgtC and/or MgtB protein levels by AmgR is critical for Salmonella virulence.
Different roles for AmgR and MgtR in controlling MgtC levels
Both the MgtR peptide (Alix and Blanc-Potard, 2008) and the AmgR RNA (Fig. 3) function as negative regulators of MgtC. However, they differ in two aspects: first, mgtR is co-transcribed with mgtC and mgtB whereas amgR is on a different transcription unit, suggesting that they act at different times after the onset of PhoP-inducing conditions. And second, MgtR does not affect MgtB levels (Alix and Blanc-Potard, 2008) whereas AmgR does (Fig. 3), which together with the first point raised the possibility of these two regulators making different contributions to Salmonella virulence.
We evaluated the MgtC and MgtB protein levels in four isogenic strains differing in whether they have functional or null alleles of mgtR and amgR. The amgR promoter mutant had higher MgtC protein levels than the wild-type strain following 5 h growth under PhoP-inducing conditions (Fig. 6C); yet, this effect was no longer apparent after 23 h (Fig. 6C). By contrast, the mgtR mutant accumulated larger MgtC amounts both at 5 and 23 h in wild-type and amgR backgrounds (Fig. 6C). As expected, accumulation of the MgtC protein in the amgR mgtR double mutant resembled that taking place in the mgtR mutant (Fig. 6C).
To compare the virulence contribution of MgtR and AmgR, we determined the survival of C3H/HeN mice when inoculated intraperitoneally with ∼3 × 103 of the four isogenic strains described above. All the animals that were infected with the Salmonella amgR promoter mutant perished within the first 16 days whereas 40% of those injected with the mgtR mutant survived, displaying a similar survival as those that were inoculated with wild-type Salmonella or the mgtR amgR double mutant (Fig. 6D).
Most examined organisms produce non-coding regulatory RNAs (Wagner and Flardh, 2002; Chen et al., 2004; Katayama et al., 2005; Storz et al., 2005; Waters and Storz, 2009). In bacteria, these RNAs can be divided into two groups based on whether they are encoded in cis or trans with respect to the target mRNA they regulate (Wagner and Flardh, 2002). Until very recently, the typical cis-acting RNA was small, encoded in a plasmid, transposon or bacteriophage genome, exhibited perfect complementarity with its target and did not require the RNA chaperone Hfq to exert its regulatory action (Brantl, 2007). This is in contrast to trans-acting RNAs, which tend to be longer, chromosomally encoded, display only partial complementarity with their targets, and are dependent on Hfq (Storz et al., 2005). However, genome-wide transcriptomic analyses now indicate that bacteria make large numbers of cis-acting antisense RNAs (Liu et al., 2009; Mendoza-Vargas et al., 2009; Toledo-Arana et al., 2009; Sharma et al., 2010), the properties of which remain largely unknown. We have now uncovered a cis-encoded antisense RNA – termed AmgR – with three salient properties: (i) it is unusually long, (ii) it functions to attenuate virulence mediated by the sense-encoded MgtC protein and (iii) its expression is under direct control of the PhoP protein, which is essential for transcription of the sense RNA that it regulates.
The 1200 nt long AmgR is slightly longer that the 1000 nt long antisense RNA identified in Clostridium acetobutylicum (Andre et al., 2008) and a 750 nt antisense RNA reported in Bacillus subtilis (Eiamphungporn and Helmann, 2009). The length of the AmgR suggests that it has the potential for complementarity with the mgtC coding region as well as with the leader region of the mgtCBR transcript, extending even to the PhoP-dependent mgtC promoter (Fig. 1A). This is in contrast to the well-characterized antitoxin RNAs, which typically overlap only with the 5′ or 3′ ends of toxin-encoding mRNAs (Fozo et al., 2008).
The amgR promoter is located in the intergenic region that separates the mgtC and mgtB genes, raising the possibility of other cis-acting RNAs being encoded within operons as recently suggested for H. pylori (Sharma et al., 2010), V. cholerae (Liu et al., 2009) and L. monocytogenes (Toledo-Arana et al., 2009). Thus, the repertoire of bacterial regulatory RNAs is likely to be larger than anticipated because most genome-wide efforts to identify regulatory RNAs have typically focused on regions located between (as opposed to within) transcription units (Vogel and Sharma, 2005; Altuvia, 2007). Yet, long intergenic regions located within operons may not necessarily code for regulatory RNAs but rather be the target of a sRNA(s) as shown for the 161 bp region that separates the glmU and glmS genes in the E. coli glmUS operon (Urban et al., 2007) and for the 111 bp region separating the iscR and iscS genes in E. coli iscRSUA operon (Desnoyers et al., 2009).
Antisense RNAs affecting expression within an operon have been described for the Spot42 RNA in the galETKM operon of E. coli (Moller et al., 2002), the RNAβ regulatory RNA of the iron transport/biosynthesis operon of Vibrio anguillarum (Stork et al., 2007) and the GlmZ sRNA targeting the glmUS operon involved in amino-sugar metabolism of E. coli (Urban and Vogel, 2008). Spot42 binds to the galK ribosome binding sequence, which unlike the mRNA destabilization promoted by AmgR, it leads to translational repression of galK without affecting expression of the two upstream genes galE and galT (Moller et al., 2002). Similarly, RNAβ is transcribed within the iron uptake-biosynthesis operon (fatDCBA angRT) of the virulence plasmid pJM1, causing differential expression between the fatDCBA genes encoding the ferric siderophore transport proteins and the angRT genes encoding the siderophore biosynthesis proteins (Stork et al., 2007). In contrast to the Spot42 and RNAβ RNAs, the GlmY and GlmZ sRNAs act hierarchically on the activation of GlmS synthesis at a post-transcriptional level without affecting GlmU protein levels (Urban and Vogel, 2008).
AmgR appears to act as a timing device that allows Salmonella to modify the levels of the MgtC and MgtB proteins dynamically when Salmonella experiences PhoP-inducing conditions (Figs 3 and 6C). The AmgR effects are more dramatically observed on MgtC and this could obey the multiple biochemical functions and/or physiological roles ascribed to MgtC, which enables growth in low Mg2+, mediates Salmonella survival inside macrophages and is necessary for Salmonella's ability to cause a lethal infection in mice (Blanc-Potard and Groisman, 1997). Thus, it is possible that different amounts of MgtC may be needed for the pathogenicity and Mg2+ homeostasis activities, which have been separated genetically (Rang et al., 2007), and that AmgR functions to limit MgtC levels under particular circumstances, which would be determined by the degree of activation of PhoP protein.
The role played by AmgR is different from that carried out by the MgtR peptide, which sets the steady-state MgtC : MgtB ratio (Alix and Blanc-Potard, 2008), because: first, even though inactivation of either mgtR or the amgR promoter results in higher MgtC protein levels, the MgtR effect is observed at both 5 and 23 h after switching bacteria to low Mg2+ media (Alix and Blanc-Potard, 2008) (Fig. 6C) but AmgR plays its major role at 5 h (Figs 3A and 6C). And second, the amgR promoter mutant is hypervirulent (Fig. 6A and D) whereas the mgtR mutant retains the pathogenic behaviour of the wild-type strain (Fig. 6D) (Alix and Blanc-Potard, 2008).
Pathogenic bacteria must control the expression of its virulence determinants tightly as overexpression of particular proteins or constitutive alleles of particular genes can compromise a pathogen's ability to survive within its host rendering the bacterium as attenuated as mutants lacking the gene (Miller and Mekalanos, 1990; Humphreys et al., 2004; Mouslim et al., 2004). It is also the case that certain virulence proteins must be expressed for a defined extent of time. For example, the Salmonella effector SptP reverses the cytoskeletal rearrangements promoted by a different Salmonella effector – termed SopE – taking place during eukaryotic cell invasion (Kubori and Galan, 2003), possibly to preserve the integrity of Salmonella's replication habitat (Galan and Zhou, 2000). Our experiments now indicate that the amount of MgtC protein is a limiting factor in Salmonella's pathogenic capacity as the amgR promoter mutation, which enhances MgtC and MgtB levels (Fig. 3A), rendered Salmonella more virulent than the wild-type strain (Fig. 6A and D) [By contrast, AmgR overexpression decreased MgtC protein levels (Fig. S4A) and attenuated Salmonella virulence (Fig. 6B).] Thus, Salmonella uses the AmgR antisense RNA to limit proliferation in host tissues, perhaps because mgtC contributes to long-term infection in mice (Lawley et al., 2006).
AmgR promotes changes in MgtC protein levels (Fig. 3A) by destabilizing the mgtC portion of the polycistronic mgtCBR mRNA (Fig. 4A) in an RNase E-dependent (Fig. 4D) but MgtR-, RNase III- and Hfq-independent fashion (Fig. S6). The Hfq-independent action of AmgR is typical of cis-acting antisense RNAs (Brantl, 2007).
Both AmgR and mgtCBR are expressed when Salmonella experiences low Mg2+ (Fig. 2) (Snavely et al., 1991; Soncini et al., 1996) by the direct transcriptional control of the PhoP protein (Figs 2 and 5). This is in contrast to most sRNAs described to date, the expression of which is typically regulated by a different regulator than the mRNA they target. Yet, it does not appear to be unique as the recently described antisense RNAs for acid resistance genes in H. pylori are also induced by acid stress (Sharma et al., 2010). The observed kinetic changes in MgtC and MgtB protein levels (Fig. 3A) probably reflect the different affinity of the PhoP protein for the amgR and mgtCBR promoters (Fig. 5C), which result in distinct promoter occupancies in vivo (Fig. 5A and B). Our findings raise the possibility of PhoP-P binding to the amgR promoter hindering transcriptional activation of the mgtC promoter, perhaps by forming a loop with PhoP-P at the mgtC promoter.
Analysis of the mgtC–mgtB intergenic region indicates that the amgR promoter and putative PhoP box sequences are conserved among salmonellae serovars differing in host specificity and disease they provoke (data not shown). Interestingly, the mgtC and mgtB genes are separated by unusually long sequences also in Yersinia spp. (300–408 bp) and in Erwinia tasmaniensis (295 bp) (Fig. S1). In the case of Yersinia spp., we could identify a conserved promoter region and a putative PhoP box sequence, raising the possibility of an AmgR-like RNA operating also in this enteric pathogen where the mgtC gene is also required for virulence (Grabenstein et al., 2006). By contrast, a TBLASTn search of the Enterobacteriaceae family genomes with the MgtR amino acid sequence did not result in a positive hit. Yet, downstream of the E. tasmaniensis mgtB gene is an open reading frame potentially coding for a highly hydrophobic 30-amino-acid long peptide, like the Salmonella MgtR peptide with which it shares six amino acids.
Bacterial strains, plasmids, oligonucleotides and growth conditions
Bacterial strains and plasmids used in this study are listed in Table S1. Oligonucleotides are listed in Table S2. All S. enterica serovar Typhimurium strains are derived from the wild-type strain 14028s (Fields et al., 1986) and were constructed by phage P22-mediated transductions as described (Davis et al., 1980). Bacteria were grown at 37°C in Luria–Bertani broth (LB) or N-minimal media (Snavely et al., 1991) supplemented with 0.1% casamino acids, 38 mM glycerol and the indicated concentrations of MgCl2. For AmgR expression from pBAD18-derived plasmids, cultures were treated with L-arabinose (0.1 mM final concentration when indicated). E. coli DH5α was used as the host for preparation of plasmid DNA. Ampicillin and kanamycin were used at 50 µg ml−1, chloramphenicol was used at 20 µg ml−1, tetracycline at 10 µg ml−1 and fusaric acid at 12 µg ml−1.
Construction of plasmid pBAD-amgR
A 1372 bp of EcoRI-HindIII fragment corresponding to the complete 1224 bp AmgR sequence and 148 bp downstream of AmgR was amplified using genomic DNA from strain 14028s and primers 7611 and 7613, and then ligated into plasmid pBAD18 that had been digested with EcoRI and HindIII to generate pBAD-amgR. DNA sequencing verified the presence of wild-type amgR sequence in the pBAD-amgR plasmid.
Construction of a chromosomal mutant with nucleotide substitutions in the amgR and mgtC promoter
To create the amgR and mgtC promoter mutants, we introduced a TetR cassette in the mgtC–mgtB intergenic region and the mgtC promoter region, respectively, as follows: we generated a 1990 bp PCR product harbouring the tetRA genes using as template chromosomal DNA from strain MS7953s and primers 7802/7803 (for amgR) and 7370/7371 (for pmgtC−), which apart from the region of complementarity with the tetRA sequences they were flanked by 40 bp sequences exhibiting identity to the mgtC–mgtB intergenic region and the mgtC promoter region. The product was purified using a QIAquick PCR purification kit (QIAGEN) and used to electroporate Salmonella 14028s containing plasmid pKD46 (Datsenko and Wanner, 2000). The resulting mgtCB::tetRA and pmgtC::tetRA strains containing plasmid pKD46 (EG18750 and EG17156) were kept at 30°C. Then, we replaced the tetRA cassette in the mgtC–mgtB intergenic region and mgtC promoter region to create the amgR and mgtC promoter mutations. This was accomplished by preparing a DNA fragment harbouring nucleotide substitutions in the amgR promoter −10 sequence and PhoP box in the mgtC promoter by a two-step PCR process. For the first PCR reaction, we used primer pairs 7554/7860 and 7859/1259 for the amgR promoter, and 3552/1689 and 1690/3553 for the mgtC promoter. For the second PCR reaction, we mixed the two PCR products from the first reaction as templates and amplified the DNA fragment using primers 7554/1259 (amgR promoter) and 3552/3553 (mgtC promoter). The PCR product was purified, electroporated into EG18750 and EG17516, respectively, and plated on media containing fusaric acid to select against the tetRA genes (Maloy and Nunn, 1981), and incubated at 37°C. The transformants were further purified on fusaric acid-containing plates and screened for their ability to grow in the presence of tetracycline and ampicillin. Those tetracycline- and ampicillin-sensitive colonies were further investigated for the presence of nucleotide substitution in the amgR promoter and mgtC promoter region by sequencing a PCR product generated with primers 7554/1259 and 3552/3553 respectively. In the mgtC promoter mutant nucleotides CTTAAACAGA were substituted for nucleotides CTGGATCCGA. The nucleotide changes in the amgR promoter mutant are indicated in Fig. 1B.
Construction of strains with chromosomal deletions of the hfq, mgtR and rnc genes
Salmonella strains deleted for the hfq, mgtR or rnc genes were constructed by the one-step gene inactivation method (Datsenko and Wanner, 2000). The CmR cassette from plasmid pKD3 was amplified using primers 4718 and 4719, 10009 and 10010, and 2991 and 2993 to make the hfq, mgtR and rnc mutants respectively. The resulting PCR products were integrated into the chromosome of strain 14028s to create strains EG15349 (hfq::CmR) and EG15350 (rnc::CmR). P22 phage lysates grown in strains EG15349 and EG15350 were used to transduce wild-type 14028s and amgR promoter mutant (EG18729) Salmonella selecting for chloramphenicol resistance to generate strains EL58 (hfq::CmR), EL59 (amgR, hfq::CmR), EL60 (rnc::CmR) and EL61 (amgR, rnc::CmR) respectively. A P22 lysate grown on a chloramphenicol resistant transformant generated upon transformation with a PCR product created with primers 10009 and 10010 was used to transduce wild-type strain 14028s and amgR promoter mutant selecting for chloramphenicol resistance to generate strain EG19879 (mgtR::CmR) and EG19880 (amgR, mgtR::CmR). The structure of the mgtR chromosomal mutation was verified by DNA sequencing using primers 10009 and 10010.
Western blot analysis
Cells were grown for the indicated times in 25 ml of N-minimal media with different MgCl2 concentration as described above. Cell extracts were prepared by sonication, electrophoresed on 12% SDS-polyacrylamide gel and transferred to PVDF membranes. The MgtB and MgtC proteins were detected using anti-MgtC and anti-MgtB polyclonal antibodies (generously provided by Michael Maguire) respectively. The data are representative of six independent experiments, which gave similar results.
S1 nuclease assay
The S1 nuclease protection assay was performed as described (Garcia Vescovi et al., 1996), with RNA harvested from mid-exponential phase cultures (OD600 of 0.3–0.4) grown in 25 ml of N-minimal medium containing different MgCl2 concentration. Total RNA was isolated with RNeasy RNA Isolation Kit (Qiagen) according to the manufacturer's specifications. Primer 7554 was labelled at the 5′ end by phosphorylation with [γ-32P]-ATP using T4 polynucleotide kinase (Promega) as described (Kato et al., 2003), and a PCR product was generated using Salmonella chromosomal DNA (strain 14028s) as template and primers 7554 and 1259. The resulting labelled PCR products were used as probes with RNA prepared from wild-type (14028s) and the amgR promoter mutant (EG18729) Salmonella (Fig. 1A).
Total RNA was isolated using RNeasy Kit (Qiagen) according to the manufacturer's instructions. The purified RNA was quantified on a Nanodrop machine (NanoDrop Technologies). cDNA was synthesized using Superscript II RNaseH- reverse transcriptase (Invitrogen) and Random hexamers (Invitrogen). The mRNA levels of the mgtC, mgtB and mgtA genes were measured by quantification of cDNA using SYBR Green PCR Master Mix (Applied Biosystems, Foster city) and appropriate primers (mgtC:6962/6963, mgtB:7763/7764 and mgtA:7225/7226) and monitored using an ABI7000 machine (Applied Biosystems, Foster city). Data were normalized to the levels of 16S ribosomal RNA amplified with primers 6970 and 6971.
Overnight cultures grown in N-minimal media containing 10 mM MgCl2 were diluted 1/100 in 50 ml of fresh N-minimal medium containing 10 µM MgCl2 and grown for 3.5 h. Rifampicin was added to a final concentration of 500 µg ml−1. Incubation was continued at 37°C, 220 r.p.m., and aliquots were withdrawn prior to or 1, 2, 4, 8, 16 and 32 min after rifampicin addition, mixed with 0.2 volume of stop solution (5% water-saturated phenol, 95% ethanol), and snap-frozen in liquid nitrogen. After thawing on ice, bacteria were pelleted by centrifugation (2 min, 16 100 g, 4°C), and RNA was isolated as described above. We measured mRNA stability of each gene by quantitative real time PCR using SYBR Green PCR master mix (Applied Biosystems) as described above. For the mgtC mRNA, we used a strand-specific Taqman probe (9664) to eliminate the amgR mRNA signal during quantitative RT-PCR. Data were normalized to the levels of 16S ribosomal RNA amplified with primers 6970 and 6971.
Northern blot analysis
Overnight cultures were diluted 1/100 in 25 ml of fresh N-minimal medium and grown for 3 h or indicated times as described above. Total RNA was isolated as described above. RNA samples (5 to 20 µg) were denatured for 5 min at 65°C in loading buffer containing 40% (final) formamide, separated on 1.3% agarose gels and transferred to nylon membranes by electroblotting. The membranes were hybridized with a double-strand probe randomly labelled with [α-32P]-dCTP or [α-32P]-UTP-labelled strand-specific probes as follows: a [32P]-UTP-radiolabelled RNA probe used to detect the mgtC transcript was synthesized by using an in vitro transcription reaction (Promega, Riboprobe combination system) according to the manufacturer's instructions. A 370 bp template mgtC leader DNA (i.e. a PCR product generated with primers 6962 and 6140 and 14028s genomic DNA as template) for the transcription reaction was cloned into the pGEMT-easy vector, which has T7 and SP6 promoters enabling the synthesis of strand-specific riboprobes. Double-strand DNA probe was PCR amplified with same primers from genomic DNA and labelled with RediprimeII Random Prime Labelling System (GE healthcare).
Assays were carried out essentially as described (Bensing et al., 1996), with minor modifications. 5′ Triphosphates were converted to monophosphates by treatment of 15 µg of total RNA with 25 units of tobacco acid pyrophosphatase (Epicentre Biotechnologies, Madison) at 37°C for 60 min in a total reaction volume of 50 µl containing 50 mM sodium acetate (pH 6.0), 10 mM EDTA, 1% β-mercaptoethanol and 0.1% Triton X-100. Control RNA was incubated under the same conditions in the absence of the enzyme. Reactions were stopped by phenol chloroform extraction, followed by ethanol sodium acetate precipitation. Precipitated RNAs were dissolved in water, mixed with 500 pmol of 5′ RNA adapter (7387, Dharmacon Research), heat-denatured at 95°C for 5 min, then quick-chilled on ice. The adapter was ligated at 17°C for 12 h with 50 units of T4 RNA ligase (New England Biolabs) in a buffer containing 50 mM Tris-HCl (pH 7.9), 10 mM MgCl2, 4 mM DTT, 150 µM ATP and 10% DMSO. Phenol chloroform-extracted, ethanol-precipitated RNA (5 µg) was then reverse-transcribed with random primers (2 pmol) and Superscript II RNaseH- reverse transcriptase (Invitrogen) according to the manufacturer's instructions. The products of reverse transcription were amplified by the use of 1 µl aliquot of the RT reaction, 25 pmol of each gene-specific and adapter-specific primers (7531 and 7534), 250 µM of each dNTP, 1 unit of Platinum Taq DNA polymerase (Invitrogen), and cloned into the pGEMT-easy vector (Promega).
Two millilitres of EG13918 cells grown overnight in N-minimal medium containing 10 mM MgCl2 were washed with Mg2+-free medium and transferred into 200 ml of N-minimal medium containing 10 µM MgCl2. Cells were aliquoted immediately after transfer (time 0) or 1, 2 and 4 h later and cross-linked with 1% formaldehyde. Chromatin immunoprecipitation assays were performed and the data analysed as described (Shin et al., 2006).
Electrophoretic mobility shift assay
DNA fragments corresponding to the mgtC, amgR and ssrB promoter regions were generated by PCR using primers 6005/3553, 8119/9448 and 2072/2073 respectively, and 14028s genomic DNA as template. The DNA fragments were gel-purified with QIAquick column (Qiagen) and 100 ng of DNA labelled with T4 polynucleotide kinase and [γ-32P]-ATP. Unincorporated [γ-32P]-ATP was removed using G-50 microcolumns (GE healthcare). A total of 1 × 104 c.p.m. of labelled probe (∼6 fmol), 200 ng of poly(dI-dC) (GE healthcare) and purified His-tagged PhoP were mixed with binding buffer [20 mM HEPES (pH 8.0), 10 mM KCl, 2 mM MgCl2, 0.1 mM EDTA, 0.1 mM dithiothreitol, 50 µg ml−1 bovine serum albumin and 10% glycerol] in a total volume of 20 µl and incubated for 20 min at room temperature. Samples were then electrophoresed on 4–20% TBE gels (Invitrogen) at 4°C, and the gels were dried and autoradiographed.
Mouse virulence assays
Six- to eight-week-old female C3H/HeN mice were inoculated intraperitoneally with ∼103 or ∼104 colony forming units. Mouse survival was followed for 21 days. Virulence assays were conducted two times using five mice per strain per dose, with similar outcomes.
We thank Michael Cromie for help with the mouse virulence assays, Tammy Latifi for help with experiments, Akinori Kato for the construction of the hfq and rnc mutants, Michael Maguire for the antibodies against the MgtC and MgtB proteins, Lionello Bossi and Nara Figureoa-Bossi for strains, and Anne Blanc-Potard and three anonymous referees for comments on the manuscript. This work was supported, in part, by NIH grant AI49561 to E.A.G. who is an investigator of the Howard Hughes Medical Institute.