Genetic Analysis of Riboswitch-mediated Transcriptional Regulation Responding to Mn2+ in Salmonella*

Background: Divalent cation binding to riboswitch RNAs regulates expression of their transporter genes in bacteria. Results: Mn2+ interacts with Salmonella riboswitches characterized from Mn2+ transporter mntH and Mg2+ transporter mgtA to modulate transcription of the downstream coding region. Conclusion: Specific riboswitches control gene expression in response to Mn2+ in bacteria. Significance: This is the discovery of a Mn2+ riboswitch. Riboswitches are a class of cis-acting regulatory RNAs normally characterized from the 5′-UTR of bacterial transcripts that bind a specific ligand to regulate expression of associated genes by forming alternative conformations. Here, we present a riboswitch that contributes to transcriptional regulation through sensing Mn2+ in Salmonella typhimurium. We characterized a 5′-UTR (UTR1) from the mntH locus encoding a Mn2+ transporter, which forms a Rho-independent terminator to implement transcription termination with a high Mn2+ selectivity both in vivo and in vitro. Nucleotide substitutions that cause disruption of the terminator interfere with the regulatory function of UTR1. RNA probing analyses outlined a specific UTR1 conformation that favors the terminator structure in Mn2+-replete condition. Switch sequence GCUAUG can alternatively base pair duplicated hexanucleotide CAUAGC to form either a pseudoknot or terminator stem. Mn2+, but not Mg2+, and Ca2+, can enhance cleavage at specific nucleotides in UTR1. We conclude that UTR1 is a riboswitch that senses cytoplasmic Mn2+ and therefore participates in Mn2+-responsive mntH regulation in Salmonella. This riboswitch domain is also conserved in several Gram-negative enteric bacteria, indicating that this Mn2+-responsive mechanism could have broader implications in bacterial gene expression. Additionally, a high level of cytoplasmic Mn2+ can down-regulate transcription of the Salmonella Mg2+ transporter mgtA locus in a Mg2+ riboswitch-dependent manner. On the other hand, these two types of cation riboswitches do not share similarity at the primary or secondary structural levels. Taken together, characterization of Mn2+-responsive riboswitches should expand the scope of RNA regulatory elements in response to inorganic ions.

Manganese is a redox-active metal, and the manganese ion Mn 2ϩ plays a pivotal role in both eukaryotic and prokaryotic organisms as a required or preferred cofactor in many met-alloenzymes including DNA and RNA polymerases, kinases, and varied redox enzymes (1,2). The bacterial pathogen Borrelia burgdorferi even utilizes manganese to bypass host defense by eliminating the need for iron (3). Various metal divalent ion transporters are able to mediate the uptake of Mn 2ϩ , thus maintaining the Mn 2ϩ cytoplasmic concentration. In eukaryotic cells, two natural resistance-associated macrophage proteins, Nramp1 and Nramp2, are characterized as divalent cation transporters (reviewed in Refs. 4 and 5). Nramp1, particularly, is a proton-divalent cation antiporter mediating the uptake of Mn 2ϩ , Co 2ϩ , Fe 2ϩ , Zn 2ϩ , and others (6). This transporter, expressed exclusively in macrophages, is regarded as a host resistance factor against different intracellular pathogens probably by depleting divalent cations from these bacteria (for reviews see Refs. 4,7,and 8). Consistently, lack of Nramp1 causes an inability of the murine macrophage to destroy intracellular Salmonella enterica serovar typhimurium, Leishmania donovani, and Mycobacterium bovis (9).
As reported from several studies, the intracellular Mn 2ϩ level of bacteria is detected at an order of 0.01 mM; however, expression of Mn 2ϩ transporters can raise this level by more than 10-fold (0.2-0.3 mM) and readily to the millimolar range under specific conditions (1, 10 -12). Many bacteria develop an Nramp1-dependent mechanism to transport Mn 2ϩ , as the mntH gene, which encodes an Nramp1 homolog, has been characterized in both Gram-positive and Gram-negative bacteria (13). In Escherichia coli, MntH mediates uptake of Mn 2ϩ and several transition metal divalent ions including Cd 2ϩ , Co 2ϩ , Fe 2ϩ , and Zn 2ϩ (14). Additionally, the sitABCD loci in Salmonella encode a member of the ABC-type ATPase superfamily that mediates Mn 2ϩ transport (12). Both MntH and SitABCD are highly selective for Mn 2ϩ over other divalent cations. It seems that SitABCD is mostly active under alkaline pH conditions (12).
Mn 2ϩ uptake is important for virulence in pathogenic bacteria. A Salmonella strain harboring mutations at both mntH and sitABCD loci exhibited an avirulent phenotype in a mouse infection model (15). On the other hand, Mn 2ϩ overload causes cytotoxicity regardless of its biological importance (16). Bacteria establish Mn 2ϩ homeostasis mainly by modulating expression of the Mn 2ϩ transporters. A transcriptional repressor, MntR, plays a major role in regulating mntH expression in Bacillus subtilis. Mn 2ϩ can bind to MntR to facilitate its binding to the mntH promoter via a palindromic sequence, 5Ј-TTT-GCCTTAAGGAAAC-3Ј, resulting in repression of mntH transcription (17). Transcriptional regulators with low identity (ϳ30%) with B. subtilis MntR have also been characterized in many Gram-negative bacteria (18,19). In E. coli and Salmonella, binding to Mn 2ϩ facilitates these MntR proteins to interact with a different palindromic sequence termed MntR-box, 5Ј-AAACATAGCAAAGGCTATGTTT-3Ј, thus implementing repression of mntH transcription (18,19). The mntH transcription is also partially repressed by Fe 2ϩ via a global iron regulator, Fur, which specifically binds Fe 2ϩ and targets a Furbinding site in the mntH promoter (11,20). Importantly, inactivation of Fur disrupts Fe 2ϩ -dependent repression of mntH transcription but retains Mn 2ϩ -dependent repression (19). Besides the negative regulation, mntH transcription is activated through the H 2 O 2 -sensing regulator OxyR, which binds to the OxyR-binding site in the promoter (11,19). It is known that Fe 2ϩ , but not Mn 2ϩ , has a high reactivity with peroxide, which generates the reactive hydroxyl radical through a Fe 2ϩ -mediated Fenton reaction. Because Mn 2ϩ is regarded as an antioxidant to counter the effect of Fe 2ϩ , facilitation of Mn 2ϩ import in response to oxidative stress may allow Mn 2ϩ to replace Fe 2ϩ in some metalloenzymes to prevent protein damage caused by reactive oxygen species (21).
The 5Ј-untranslated region (5Ј-UTR) of particular bacterial genes can exert a regulatory effect on either transcription elongation to the downstream region or translation initiation of the open reading frame (ORF). Many of these 5Ј-UTRs are riboswitches that bind a specific signal molecule, thus forming an alternative conformation via switching between mutually exclusive base pairs to modulate transcription or translation of the downstream region (for recent reviews see Refs. [22][23][24]. The signal molecules, which are mainly metabolites present in the cytoplasm, interact with the ligand-binding domain (or aptamer) of the riboswitches. Most commonly, the riboswitch domain in a nascent transcript can cause a transcription termination by forming an intrinsic transcription terminator (22,23,25). It has been shown that inorganic ions such as Mg 2ϩ and F Ϫ can serve as ligands to interact with specific riboswitches (26 -28). It is generally believed that inorganic cations play an important role in neutralizing negatively charged phosphate groups that come into close proximity in the transition states of RNA folding (for a recent review see Ref. 29). Essentially, Mg 2ϩ contributes to the folding of all large RNAs (29). Meanwhile, particular RNA molecules display high specificity to interact with Mg 2ϩ . The 5Ј-UTR of two Mg 2ϩ transporter genes, mgtA from Salmonella and mgtE from Bacillus, responds to Mg 2ϩ through a riboswitch domain regardless of having no homology at the primary or secondary structure, which subsequently facilitates a transcription termination of the downstream coding region (26,27). It has been shown that six Mg 2ϩ ions residing in the mgtE riboswitch stabilize the conformation with a Rho-independent terminator, thus facilitating mgtE transcription termination (27). Also, a regulatory mechanism that controls transcription termination is involved in the mgtA Mg 2ϩ riboswitch function via a Rho-dependent terminator (30) as an open read-ing frame encoding a 17-residue leader peptide that is translated within the 5Ј-leader region (LR) 2 (31,32).
In this article, we have provided evidence of Salmonella response to Mn 2ϩ through divalent cation riboswitches. By conducting in vivo gene expression assays, RNA structural probing, mutational analysis, and in vitro transcription experiments, we have established that a 5Ј-UTR of the mntH mRNA functions as a regulatory element by sensing Mn 2ϩ to determine whether transcription reads through into the mntH coding region or stops within the 5Ј-UTR. We also have demonstrated that the Mg 2ϩ riboswitch can exert its regulatory effect by responding to Mn 2ϩ in a manner similar to Mg 2ϩ .

MATERIALS AND METHODS
Bacterial Strains, Growth Conditions, and Oligonucleotides-All S. enterica serovar typhimurium strains were derived from the wild-type strain ATCC14028s and are listed in Table 1. Bacteria were grown at 37°C in Luria-Bertani (LB) broth or in N minimal medium (33), pH 7.4, supplemented with 0.1% casamino acids and 38 mM glycerol. MnCl 2 and MgCl 2 were added to the required concentrations. When necessary, antibiotics were added at final concentrations of 50 g/ml for ampicillin, 20 g/ml for chloramphenicol, and 50 g/ml for kanamycin. E. coli DH5␣ and BL21-Gold (DE3) were used as hosts for the preparation of plasmid DNA and protein production, respectively. Oligonucleotides used in this study are described in Table 2.
Construction of Strains with Chromosomal Deletions, lac Fusions, FLAG Fusions, and Site-directed Mutations-Strains harboring deletions and FLAG fusion were generated as described previously (34). If needed, the antibiotic resistance cassette was removed using plasmid pCP20. Deletion of the mntH, mntR, and fur genes was carried out using primer pairs 144 and 145, 1253 and 1254, and 1262 and 1263, respectively, to amplify the kanamycin resistance cassette (Km R ) from plasmid pKD4 and integrate the resulting PCR product into the chromosome. A lac gene was integrated in the deleted chromosomal mntH location using plasmid pKG137, which was inserted into the FLP recombination target sequence generated after the Km R cassette was removed (35). mntH-lac strain which contained substitution of the mntH 5Ј-UTR, stem-R[mut], was constructed using primer pair 145 and 1575 to amplify the PCR product from the chromosomal DNA of the ⌬mntH::Km strain; the product was electroporated into wild-type cells harboring pKD46, and then Km R colonies were selected, and substitution was confirmed. Construction of the strain harboring a chromosomal copy of the mntH-FLAG (C terminus) fusion was carried out using primer pair 7 and 8 to amplify the kanamycin resistance cassette (Km R ) from plasmid pKD4 and integrate the resulting PCR product into the chromosome of a Salmonella strain that carried a chromosomal corA-FLAG fusion (our laboratory collection).
Plasmid Construction-All plasmids used in this study are listed in Table 1. pYS1300 was constructed using PCR fragments containing mntH full-length UTR1 generated with primer pair 1307 and 1308 and wild-type 14028s chromosomal DNA as template. These fragments were digested with PstI and XhoI and then ligated between the PstI and XhoI sites of pYS1000 (36). Derivatives of pYS1300 with nucleotide substitutions were constructed using the GeneTailor site-directed mutagenesis system (Invitrogen) with Platinum Taq Polymerase High Fidelity (Invitrogen) and pYS1300 DNA and the following primer pairs: for pYS1301, 1341 and 1342; and for pYS1331, 1543 and 1544. pYS1466 was constructed using PCR fragments containing the mntR coding region generated with primer pair 1630 and 1631 and 14028s chromosomal DNA as template; these were digested with NdeI and XhoI and then ligated between the NdeI and XhoI sites of pET28a (EMD Biosciences). pYS1008 (a FLAG epitope in the C terminus of MntH) was constructed as follows. A PCR fragment containing the mntH coding region with its Shine-Dalgarno sequence that was generated with a primair pair 12 and 13 and 14028s chromosomal DNA as template was digested with BamHI and HindIII and cloned into pUHE21-2lacI q that had been digested with the same enzymes. Derivatives of pYS1008 with nucleotide substitutions were constructed through site-directed mutagenesis (described above) using the following primer pairs: 20 and 21 for pYS1013 (D34E) and 23 and 24 for pYS1014 (E102D). pYS1020 was constructed using a PCR fragment containing the corA coding region with its Shine-Dalgarno sequence that was generated with primer pair 105 and 106 and 14028s chromosomal DNA as template was digested with BamHI and HindIII and cloned into pUHE21-2lacI q , that had been digested with the same enzymes. The pYS1010 derivative pYS1011, with nucleotide substitutions at stem-loop B, was constructed through site-directed mutagenesis using primer pair 17 and 18.
In Vitro Transcription Assays Using E. coli RNA Polymerase Holoenzyme-Linear DNA templates containing the P lac1-6 promoter region, the mntH full-length UTR1, and the first 62 nucleotides (nt) of the lacZ ORF were generated by PCR from pYS1300 and its derivatives using primers 241 and 1302. The template harboring the full-length mgtA 5Ј-UTR was amplified using plasmid pYS1010 DNA and primers 201 and 248. A control template containing the P lac1-6 promoter and the first 151 nt of the lacZ gene was amplified from pYS1000 (36) using primers 241 and 565. In vitro transcription was carried out as described (37). Briefly, 1 unit of E. coli RNA polymerase 70 holoenzyme (Epicentre) was incubated with 0.5 g of template DNA in 35 l of transcription buffer, which contained 100 mM Tris-HCl, pH 8.0, 100 mM NaCl, 0.2 mM EDTA, 0.2 mM DTT, 50 mg/ml BSA, and 0.35 mM Mg 2ϩ , at 37°C for 30 min to form open complexes. RNA synthesis was initiated by adding 15 l of NTP mixture, which contained 0.32 mM ATP, CTP, and GTP, 0.1 mM UTP, and 2 Ci of [␣-32 P]UTP (PerkinElmer Life Sciences). After a 10-min incubation at 37°C, transcripts were precipitated with 5 l of 3 M sodium acetate, pH 5.5, and 150 l of ethanol, separated in a 6% denaturing polyacrylamide gel electrophoresis, and detected by autoradiography. When necessary, Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ were added to the required concentrations. The length of transcripts was determined by using a DNA sequencing ladder generated from a PCR product amplified with primers 241 and 32 P-labeled 1302 from pYS1300 (for mntH UTR1) or primers 248 and 32 P-labeled 201 (for mgtA 5Ј-LR), and degraded using the Maxam-Gilbert sequencing reaction.
Primer Extension-Bacteria were grown to mid-exponential phase (A 600 nm of 0.4 -0.6) in 50 ml of N minimal medium, pH 7.4, containing 10 mM MgCl 2 and 0 or 0.05 mM MnCl 2 , and total RNA was isolated from the harvested bacterial cells using the SV total RNA isolation system (Promega) according to the manufacturer's specifications as described previously (38). The primer extension assay was performed in a 25-l reaction with 10 g of total RNA, 32 P-labeled primer 1301 (complementary to the 31-50 nt of the mntH ORF), 100 units of M-MuLV reverse transcriptase (Promega), and 1ϫ reaction buffer and incubated at 42°C for 2 h. The cDNA were synthesized and resuspended in 10 l of H 2 O after precipitation with ethanol. Samples (3 l) were analyzed by 6% denaturing Isolation of MntR-His 6 -E. coli BL21-Gold (DE3) harboring plasmid pET28a-mntR (pYS1466) was grown at 37°C with shaking to A 600 nm of 0.5 in 500 ml of LB medium; then IPTG (final concentration, 1 mM) was added, and the bacteria were incubated for 2 h. Cells were harvested, washed with PBS once, resuspended in 10 ml of PBS, and opened by sonication. The whole cell lysate was used for MntR-His 6 purification by mixing with His-Select nickel affinity gel (Sigma) following the instructions of the manufacturer. A pure MntR-His 6 sample was tested using silver staining (Pierce) following the instructions of the manufacturer.
DNase I Protection Assays-DNase I protection assays were carried out using DNA fragments amplified by PCR using 14028s chromosomal DNA as template. Prior to the PCR, primer 1583 was labeled with T4 polynucleotide kinase and [␥-32 P]ATP. The mntH promoter region was synthesized with primers 1582 and 32 P-labeled 1583. Approximately 25 pmol of labeled DNA and 0, 100, or 200 pmol of the MntR-His 6 protein were used in a 100-l reaction. 0.05 mM Mn 2ϩ was added to the required reactions. DNase I digestion was carried out as described previously (36) using 0.05 units of DNase I (Invitrogen) per reaction. Samples (3 l) were analyzed by 6% denaturing polyacrylamide electrophoresis by comparison with a DNA sequence ladder generated with the appropriate primer by a Maxam-Gilbert AϩG reaction. The positions of radioactive DNA fragments in the gels were detected by autoradiography.
Structural Probing of the mntH UTR1 RNA Structure-An RNA containing full-length mntH UTR1 was synthesized with a T7 RiboMax large scale RNA production system (Promega) according to the manufacturer's instructions using PCR-generated products as templates. The PCR products harboring the T7 promoter were generated using pYS1300 and its derivatives as template and primers 1312 and 1355. A DNA sequencing ladder was generated from the PCR product that was amplified using primer 1312 and 32 P-labeled 1355 and degraded using the Maxam-Gilbert reaction. Chemical modification of the RNA with dimethyl sulfate (DMS) was carried out as follows. 1 l of RNA solution (corresponding to 7.5 g) was mixed with 146 l of H 2 O, incubated at 95°C for 5 min, cooled on ice for 30 s, and placed at room temperature for 5 min.  When the specificity of UTR1 to cationic ions was clarified, Mg 2ϩ in the solution was substituted by Mn 2ϩ , Mg 2ϩ , or Ca 2ϩ at the required concentrations. 2 l of DMS (Acros) was added, and the reaction mixture was incubated at 37°C for 5 min. Then, 20 l of 3 M sodium acetate, pH 5.2, and 600 l of cold ethanol were added, and the tube was kept at Ϫ80°C for 15 min. The RNA was precipitated by centrifugation at 14,000 rpm for 15 min, and the pellet was washed with 75% ethanol and airdried. The products were reverse-transcribed using M-MLV reverse transcriptase (Promega) and 32 P-labeled primer 1355 according to the manufacturer's instructions, separated using 6% denaturing polyacrylamide gel electrophoresis, and detected by autoradiography. DMS modification of the denatured RNA was carried out in a similar fashion. Specified concentrations of Mn 2ϩ and EDTA were added as described above. The mixture was incubated at room temperature for 40 h. The product was separated and analyzed as described above for the RNA structural probing step.
␤-Galactosidase Assay-The ␤-galactosidase assay was carried out in triplicate, and the activity was determined as described previously (39). The data correspond to three independent assays conducted in duplicate, and all values are mean Ϯ S.D.

RESULTS AND DISCUSSION
A MntR-independent Mechanism Contributing to Mn 2ϩ -responsive mntH Transcription in Salmonella-0.05 mM Mn 2ϩ was able to turn off mntH transcription in Salmonella because ␤-galactosidase activity was reduced to an undetectable level in a mntH-lac strain grown in N minimal medium supplemented with Mn 2ϩ (Fig. 1). In a mutant harboring deletion of the regulator mntR gene, mntH transcription could still be repressed 2.8-fold by Mn 2ϩ although it was not turned off as in the wild type (Fig. 1). This observation implies that Salmonella may regulate Mn 2ϩ -responsive mntH transcription through a MntRindependent mechanism. When a chromosomal region located upstream of the mntH ORF and overlapping a previously characterized MntR-binding site (referred to as MntR-box; see Ref. 18) was substituted with an alternative sequence (i.e. stem-R[mut]), this MntR-independent regulation was eliminated because mntH transcription could no longer be repressed by Mn 2ϩ and therefore remained similarly activated regardless of Mn 2ϩ in a mntR stem-R[mut] mutant (Fig. 1). On the other hand, mntH transcription could still be repressed significantly by Mn 2ϩ in a stem-R[mut] mutant, implying that MntR could exert its effect by interacting with a region other than the MntR-box, in the mntH promoter ( Fig. 1). Hence, we concluded that this novel regulatory mechanism was independent from MntR but still dependent on a regulatory element located upstream of mntH ORF. Because it functions regardless of MntR, this chromosomal region most likely plays a regulatory role as more than just a MntR binding region. Importantly, this hypothesis provides a possible explanation for several previous observations drawn from a plasmid-borne mntH expression that were consistent with our observations (1,19).
Mn 2ϩ Down-regulates mntH Transcription from Two Different Initiation Sites-To gain insights into this regulation, we characterized the promoter region of mntH gene by mapping its transcription initiation site. We carried out primer extension and detected two products from the wild-type strain grown in a low Mn 2ϩ condition ( Fig. 2A, lane 1). Therefore, mntH transcription is initiated from two locations, which correspond to adenosines located 76 and 18 nt upstream of the mntH ORF, respectively (Fig. 2, A and C, referred to as ϩ1 and ϩ1Ј hereinafter). Mn 2ϩ (0.05 mM) added to the wild-type culture reduced the level of both transcripts initiated from ϩ1 and ϩ1Ј by 5.1and 7.4-fold, respectively ( Fig. 2A, lane 2), suggesting that Mn 2ϩ down-regulates the transcription initiated from both starts. We examined the MntR-binding site in the mntH promoter by carrying out DNase I footprinting and found that MntR protein could protect two regions (Fig. 2B) (referred to as R1 and R2, respectively). R1 is located in the Ϫ21 and Ϫ2 nt upstream of ϩ1, which is protected by MntR mainly in the presence of Mn 2ϩ (Fig. 2B, lanes 4 -6), whereas R2 is located in the Ϫ18Ј and Ϫ1Ј nt upstream of ϩ1Ј, which is protected under the conditions we tested regardless of Mn 2ϩ (Fig. 2B, lanes 1-3  and 4 -6). Thus, MntR most likely exerts its inhibitory effect by binding to individual Ϫ10 regions upstream of ϩ1 and ϩ1Ј, respectively (illustrated in Fig. 2C). Consistent with previous results, R2 partially overlaps the MntR-box demonstrated previously (18). A pentanucleotide motif, CAAAG, is shared by the R1 and R2 sequences and is highly conserved in the mntH promoter of E. coli and Shigella (Fig. 2C), and thus it most likely

Salmonella Riboswitches Responding to Mn 2؉
APRIL 18, 2014 • VOLUME 289 • NUMBER 16 represents a consensus sequence for MntR binding in these Gram-negative bacteria.
Regulatory Role of the 5Ј-Untranslated Region from Transcription Start ϩ1 in Response to Mn 2ϩ -We examined whether the mntH 5Ј-UTR could play a role in MntR-independent regulation. Transcription initiated from ϩ1 and ϩ1Ј will produce a long 76-nt 5Ј-UTR (termed UTR1 hereinafter) and a short 18-nt 5Ј-UTR, respectively. We investigated the regulatory role of the UTR1 by constructing a plasmid (pYS1300) that carried full-length UTR1-lacZ transcriptional fusion under the control of P lac1-6 (i.e. a promoter independent of IPTG) (40). P lac1-6 is insensitive to Mn 2ϩ because the level of ␤-galactosidase from wild type harboring parental pYS1000, which carries a P lac1-6 -controlled lacZ gene (26), is similar regardless of Mn 2ϩ (Fig. 3A). ␤-Galactosidase activity from wild type harboring pYS1300 is 2.6-fold lower in the medium supplemented with Mn 2ϩ than without (Fig. 3A). It is still reduced by about 2.6-fold in an mntR fur mutant harboring pYS1300 by supple- menting Mn 2ϩ (Fig. 3A) regardless of both the MntR (R1)-and Fur-binding sites present in the UTR1 region (Fig. 2C). We reason that the chromosomal sequence upstream of UTR1, which is absent in pYS1300, may still be required for MntR and Fur to exert their effects. Taken together, these results suggest that UTR1 is sufficient to respond to Mn 2ϩ and down-regulate the mntH transcription independently from the known regulators.
It is generally believed that Mg 2ϩ and Ca 2ϩ have a similar size and charge density that can overlap Mn 2ϩ function in stabilizing the structural charge of enzymes. However, ␤-galactosidase activity from wild type harboring pYS1300 was reduced by supplementing 0.05 mM Mn 2ϩ , but not Mg 2ϩ or Ca 2ϩ (Fig. 3C), indicating that these ions could not replace Mn 2ϩ to repress transcription through UTR1. To determine whether the UTR1 would respond to a higher concentration of Mg 2ϩ , we tested a high (H) (10 mM) and low (L) (0.01 mM) Mg 2ϩ , respectively, which were used to determine the regulatory activity of the Salmonella Mg 2ϩ riboswitch, i.e. the mgtA 5Ј-LR (26). Indeed, ␤-galactosidase activity from wild type harboring pYS1010 (a plasmid that carried a P lac1-6 -controlled lacZ fusion with mgtA 5Ј-LR; also see Ref. 26) grown in N medium with 10 mM Mg 2ϩ was 29-fold lower than that with 0.01 mM Mg 2ϩ (see Fig. 3D, right panel, Mg 2ϩ L/H column). However, the ratio of the ␤-galactosidase activity from wild type harboring pYS1300 grown in high and low Mg 2ϩ was ϳ1-fold (Fig. 3D, left panel, Mg 2ϩ L/H column), indicating that Mg 2ϩ , even at a high level, had no effect on UTR1. Furthermore, when Mg 2ϩ uptake was enhanced by overexpressing a Salmonella Mg 2ϩ transporter, corA, mgtA 5Ј-LR-controlled transcription was repressed because ␤-galactosidase activity from wild type harboring pYS1010 and pYS1020 (P lac -controlled corA gene) grown in low Mg 2ϩ with IPTG (0.5 mM) was ϳ25-fold lower than that without IPTG (Fig. 3D, right panel, IPTG Ϫ/ϩ column). However, ␤-galactosidase activity from wild type harboring pYS1300 and pYS1020 remained similar in low Mg 2ϩ regardless of IPTG (Fig.  3D, left panel, IPTG Ϫ/ϩ column). Taking these results together, we concluded that the UTR1 is highly specific to Mn 2ϩ .
An Intrinsic Rho-independent Terminator within the UTR1-We observed that UTR1 nucleotides 46 -65 formed a stemloop (termed T hereinafter; Fig. 3B, middle panel) when two complementary segments, 46 CAUAGC 51 (i.e. stem-L) and 56 GCUAUG 61 (stem-R), were base-paired to each other and followed by 62 UUUU 65 (poly(U)). Hence, it is likely a Rho-independent terminator that is responsible for the regulatory function of UTR1. According to this structure, the stem-R[mut] substitution actually changed complementary bases at the right arm of this stem-loop (Fig. 3B, right panel), thus explaining why it caused disruption of the MntR-independent regulation (Fig.  1). We constructed two pYS1300 derivatives, pYS1301 and pYS1331, in which UTR1 contained sequences of stem-R[mut] and stem-L[mut] (Fig. 3B, left panel), respectively, and reconfirmed that disruption of this terminator impaired the regulatory activity of UTR1, because ␤-galactosidase activity remained at levels similar to the wild-type strain or mntR fur double mutant carrying these plasmids grown in the medium regardless of Mn 2ϩ ; this gave rise to a ratio change of ϳ1-fold with and without supplementing Mn 2ϩ (Fig. 3A, last four columns) and was also similar to the wild-type strain carrying pYS1300 grown without Mn 2ϩ (data not shown). The UTR1 sequence is highly conserved in E. coli and Shigella, and therefore, this Mn 2ϩ -responsive regulation is also most likely employed in these enteric bacteria (Fig. 2C). On the other hand, the short (18 nt) 5Ј-UTR transcribed from the ϩ1Ј start is unlikely to mediate a premature termination because it contains only three nucleotides from the region of the intrinsic terminator (Fig. 2C). ). The -fold change was determined by ␤-galactosidase activity from no Mn 2ϩ supplemented medium divided by activity from 0.05 mM Mn 2ϩ . B, the predicted Rho-independent terminator structure in mntH UTR1 (T, center panel). The highlighted sequences were substituted in stem-L[mut] (left panel) and stem-R[mut] (right panel), respectively. C, a ␤-galactosidase assay was carried out with wild type harboring pYS1300. Bacteria were grown for 6 h in N medium (0.01 mM Mg 2ϩ ). The -fold change was determined by ␤-galactosidase activity from no Mn 2ϩ supplemented medium divided by activity from 0.05 mM Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ , respectively. D, a ␤-galactosidase assay was carried out with the wild-type strain harboring pYS1020 and pYS1300 (UTR1-lacZ, left panel) or pYS1010 (mgtA 5ЈLR-lacZ, right panel). The -fold change was determined by ␤-galactosidase activity from 0.01 mM Mg 2ϩ divided by activity from 10 mM Mg 2ϩ (Mg 2ϩ L/H) and by ␤-galactosidase activity from 0.01 mM Mg 2ϩ without IPTG divided by activity from 0.01 mM Mg 2ϩ with 0.5 mM IPTG (IPTG Ϫ/ϩ). Bacteria in A, C, and D were grown for 6 h in N medium (0.01 mM Mg 2ϩ ).

UTR1 Independently Facilitates Transcription Termination-
We carried out an in vitro transcription assay in which a linear DNA template amplified from pYS1300 was used to produce a 165-nt runoff RNA containing full-length 76-nt UTR1, an upstream 9-nt linker sequence, and a downstream 80-nt lacZ coding region (Fig. 4A). In this in vitro system, the only protein component supplemented was E. coli RNA polymerase 70 holoenzyme, thus ruling out the influence from any cellular regulatory factors. The only product detected in the Mn 2ϩ -free condition was the runoff transcript (Fig. 4B, lane 1, designated  R), indicating that no terminator structure was formed to stop transcription before it read through the template. Mn 2ϩ (0.2 mM) significantly enhanced overall transcription and facilitated production of the 72-77-nt terminated transcripts (Fig. 4B,  lane 2, designated T). These terminated transcripts contained 63-68-nt UTR1 fragments, indicating that Mn 2ϩ induced the terminator structure to cause transcription termination occur-ring from the second nucleotide of the polyuridine sequence (Fig. 4A, red arrowheads). Mn 2ϩ , but no other cations tested, also enhanced production of additional truncated transcripts appearing mainly as 110-and 111-nt bands (Fig. 4B, lane 2, designated U), probably by pausing transcription at nucleotides 7 or 8 of the lacZ sequence (indicated in Fig. 4A by pink arrowheads). Thus, these truncated products should most likely be regarded as runoff transcripts as well because they stopped downstream of the UTR1. As a result, the level of runoff transcripts was reduced to 31% of the total transcripts through the Rho-independent termination in the presence of 0.2 mM Mn 2ϩ (the percentage was calculated by using the formula [R ϩ U]/ [R ϩ T ϩ U] ϫ 100 (Fig. 4B)). On the other hand, the level of runoff transcript in the reaction supplemented with Mg 2ϩ or Ca 2ϩ to the same concentration (0.2 mM) remained at 56 and 49% of the total transcripts, respectively (Fig. 4B, lanes 3 and 4), indicating that Mg 2ϩ and Ca 2ϩ caused less 72-77-nt termi- nated products than Mn 2ϩ . This observation is consistent with the in vivo expression assay (Fig. 3C), thus suggesting that UTR1 has a high specificity in response to Mn 2ϩ . This transcription termination was dependent on Mn 2ϩ concentration, because the level of runoff RNA dropped to 48% of the total transcripts when 0.1 mM Mn 2ϩ was supplemented (Fig. 4C,  lane 2) and was further reduced to 27% when the Mn 2ϩ concentration was raised to 0.2 mM (see Fig. 4C, lane 3, similar to the result in Fig. 4B, lane 2). However, it was most interrupted when the stem-R[mut] UTR1 was transcribed because the runoff transcript in Mn 2ϩ -containing reactions represented almost all (99%) of the total products (Fig. 4C, lanes 8 and 9). Furthermore, the UTR1 transcription using terminator-insensitive T7 RNA polymerase could only produce the runoff transcript regardless of Mn 2ϩ (data not shown), confirming that the terminator is essential for transcription termination at the UTR1 63-68-nt region.
Interestingly, the 110-and 111-nt truncated products were still dependent on the UTR1 because they appeared in the Mn 2ϩ -containing reactions with both the wild type and the substituted templates (Fig. 4, B, lane 2, and C, lanes 2 and 3, 5 and 6, and 8 and 9) and because there were no truncated bands representing termination at nucleotides 7 and 8 of lacZ observed in Mn 2ϩ -containing reactions using a template from pYS1000 that carried P lac1-6 -lacZ (the runoff transcript is 179 nt; see Fig. 4C, lanes 10 -12). We postulated that the UTR1 might interact with this downstream region in the presence of Mn 2ϩ to pause transcription in a terminator-independent manner. Surprisingly, we did not observe the predicted ϩ1Јtranscript, i.e. a 98-nt band, in these in vitro reactions (Fig. 4, B and C), although the ϩ1Ј-site is located within the cloned UTR1 (indicated by uppercase A, Fig. 4A). This was probably because additional regulators, which are absent in our in vitro system, are required to initiate transcription from this start.
The UTR1 Favors a Terminator Conformation in the Presence of Mn 2ϩ -Mn 2ϩ most likely induces a conformational modification of UTR1, by which this 5Ј-UTR formed the terminator structure to facilitate premature transcription termination. We predicted two UTR1 conformations mainly through canonical base-pairing (Fig. 5A). Structure 1, designated S1, possesses stem-loops A, B, and C. Based on our results, this conformation, which lacks the terminator stem-loop (T), most likely represent a structure favored in low Mn 2ϩ , thus allowing transcription to pass through UTR1. On the contrary, structure 2 (S2), which contains stem-loops D, E, and T, is likely a conformation favored in high Mn 2ϩ .
We carried out structural mapping using full-length UTR1 RNAs folded in a buffer supplemented without (ϪMn) (Fig.  5B, Ϫ) or with 0.35 mM Mn 2ϩ (ϩMn) (Fig. 5B, ϩ). The folded RNA was treated with DMS, which modifies RNA by methylating unpaired adenosines, cytidines, and guanosines. We tested wildtype RNA and observed that nucleotides A 52 , A 53 , A 54 , and G 55 were modified 2.8-, 3.5-, 5.3-, and 7.2-fold more, respectively, in Mn 2ϩ -treated RNA than in Mn 2ϩ -free RNA (Fig. 5B, lanes 4 and  3, upper panel, and quantified data in lower panel). These results are consistent with the predicted conformation in which they are located in the terminator loop T formed in S2 and thus are likely unpaired and modified more by DMS in ϩMn (i.e. high Mn 2ϩ ).
On the other hand, A 52 , A 53 , A 54 , and G 55 are paired in stem C formed in S1, and indeed, they are modified less by DMS in ϪMn (i.e. low Mn 2ϩ ). Contrastingly, A 49 and G 50 were predicted to be located in terminator stem T and paired in S2 but unpaired in a single-strand region of S1; thus they are likely modified less by DMS in ϩMn than in ϪMn. Consistently, our results showed that A 49 and G 50 were modified 4.7-and 7.2-fold less, respectively, in Mn 2ϩ -treated RNA than in Mn 2ϩ -free RNA (Fig. 5B, lanes 4 and  3).
Unlike the wild-type RNA, DMS modified A 52 , A 53 , A 54 , and G 55 from stem-R[mut] UTR1 poorly but similarly in both ϪMn and ϩMn (Fig. 5B, lanes 6 and 7). This suggests that Mn 2ϩ could not induce stem switching in this substituted UTR1, and we predicted that these nucleotides would remain paired within stem-loop CЈ regardless of the Mn 2ϩ concentration (Fig. 5C). We reasoned that this conformation was thermodynamically favored in this substituted UTR1 because the free energy reduction in stem-loop CЈ (⌬G, Ϫ7.7 Kcal/mol, estimated using the Mfold tool) was greater than that in another conformation (ϳ⌬G, Ϫ3.8 Kcal/mol), which could only form stem-loops D and E as in wild-type S2 (⌬G, Ϫ9.6 Kcal/mol) but not T (⌬G ϭ Ϫ5.8 Kcal/mol).
UTR1 Forms Terminator Structure through a Mn 2ϩ Riboswitch-We found two CAUAGC sequences in UTR1, representing nucleotides 34 -39 and 46 -51 and alternatively basepairing stem-R sequence 56 GCUAUG 61 (Fig. 5A, sequences in dashed-line boxes in red and blue, respectively). We proposed that switching their base-pairing would contribute to the regulatory function of UTR1. In S1, the 34 CAUAGC 39 sequence is located in loop C, meanwhile forming a pseudoknot through pairs with stem-R sequence and thus playing a role as antiterminator by preventing the formation of a terminator structure (Fig. 5A). Instead, the 46 CAUAGC 51 sequence, i.e. stem-L, base-pairs with stem-R in S2 to form the terminator (Fig. 5A). Consistently, RNA probing results showed that the stem-R sequence remained paired and could not be modified by DMS in either Mn 2ϩ -free or Mn 2ϩ -treated RNAs (Fig. 5B, lanes 3  and 4). The conformational change is also shown by nucleotide A 29 , which is predicted to be paired in stem A in S1, is switched to become unpaired in loop E in S2, and is modified 3.1-fold more in Mn 2ϩ -treated RNA than Mn 2ϩ -free RNA. On the contrary, C 21 and A 22 , which appear to be unpaired in loop A in S1, but paired in stem E in S2, are modified 7.2-and 7.4-fold less, respectively, in Mn 2ϩ -treated RNA than Mn 2ϩ -free RNA ( Fig.  5B, lanes 4 and 3). On the other hand, C 21 and A 22 in stem-R[mut] UTR1 were modified significantly by DMS in both low and high Mn 2ϩ , and the level of modification was similar to that in Mn 2ϩ -free wild-type RNA; meanwhile, A 29 remained unmodified under the Mn 2ϩ concentrations tested (Fig. 5B,  lanes 6 and 7). This demonstrates that loss of the riboswitch due to inability to form stem-loop T is a reason to keep stem-R[mut] UTR1 in the thermodynamically favorite conformation regardless of Mn 2ϩ (Fig. 5C). An in vitro transcription assay using the stem-L[mut] template showed that the runoff RNA could be reduced to 92 and 88% in 0.1 and 0.2 mM Mn 2ϩ , respectively (Fig. 4C, lanes 5 and 6). We postulate that lack of the stem-L sequence 46 CAUAGC 51 may allow the alternative CAUAGC sequence, i.e. 34 CAUAGC 39 , to base pair the stem-R sequence 56 GCUAUG 61 (the same base-pairing occurred in the pseudoknot structure of S1 (Fig. 5A)), resulting in the formation of an alternative stem-loop followed by the poly(U) sequence 62 UUUU 65 . Thus, this structure could be an alternative terminator and facilitate termination of stem-L[mut] transcription, despite its functioning less efficiently than stem-loop T (88% of the runoff RNA in the mutant versus 27% in the wild type from reactions containing 0.2 mM Mn 2ϩ ; Fig. 4C, lanes 6 and 3). On the other hand, this alternative terminator could not be formed in vivo, probably for some unknown reason, because lacZ expression remained similar to the strains carrying pYS1331 grown in the medium regardless of Mn 2ϩ (Fig. 3A). Unlike the stem-L[mut] UTR1, elimination of 56 GCUAUG 61 resulted in the inability of the stem-R[mut] UTR1 to form either termina-tor structure and thus could not mediate transcription termination in vivo and in vitro (Figs. 3A and 4C).
It is worth pointing out that Mg 2ϩ was supplemented as a component in the RNA structural mapping reaction of Fig. 5B for the sake of the UTR1 transcription in vitro, which requires this cation. Thus, we carried out additional structural mapping using only one given divalent cation to fold wild-type UTR1 RNA. 0.01 mM Mn 2ϩ (referred to as low Mn 2ϩ hereinafter), also at a level analogous to the cytoplasmic concentration under the physiological condition (1), caused the UTR1 conformation to resemble that of S1, because the same set of nucleotides was modified similarly by DMS (Fig. 5D, lane 3, arrowheads) as those in the Mg 2ϩ -containing condition without supplementing Mn 2ϩ (Fig. 5B, lane 3). Concomitantly, the addition of 0.35 FIGURE 5. RNA structural probing of UTR1. A, schematic representation of the predicted secondary structure of the 76-nt mntH UTR1 from S. enterica serovar typhimurium. Left, S1 conformation, which is postulated to form in low Mn 2ϩ . Right, S2 conformation, which is predicted to form in high Mn 2ϩ . Sequences in blue and red represent nucleotides that showed stronger modification by DMS in Mn 2ϩ -free and Mn 2ϩ -supplemented reactions through detection, respectively. Asterisks in red and pink represent nucleotides cleaved more after wild-type and stem-R[mut] UTR1 were incubated with Mn 2ϩ , respectively. Dashed line boxes in red and light blue represent repeated CAUAGC sequences that base pair alternatively with 56 GCUAUG 61 . Pink lines indicate a pseudoknit formed between 34 CAUAGC 39 and 56 GCUAUG 61 . Numbering represents the location from the ϩ1 transcription site. The sequences highlighted in blue and yellow were the substitutions in stem-L[mut] and stem-R[mut], respectively. B, DMS treatment of the full-length UTR1 RNA. Wild-type and stem-R[mut] UTR1 were incubated with 0 (Ϫ) or 0.35 mM (ϩ) Mn 2ϩ . M corresponds to a ladder prepared from a Maxam-Gilbert AϩG reaction. U is the untreated sample. The blue and red arrowheads indicate nucleotides in wild-type UTR1 that are modified by more Mn 2ϩ -free and Mn 2ϩ -supplemented reactions, respectively. The numbers represent the nucleotides located in the UTR1. The corresponding modification ratio (ϩMn 2ϩ /ϪMn 2ϩ ) is calculated and shown in the right panel. C, the predicted secondary structure of substituted stem-R[mut] UTR1, which was similar to wild-type S1 except for the pseudoknit. D, comparative DMS analysis of the UTR1 RNA treated with Mn 2ϩ , Mg 2ϩ , and Ca 2ϩ , respectively. Each reaction was supplemented with only one of the given cations and carried out similar to those shown in B. Mn 2ϩ was added to 0.01 mM (L) and 0.35 mM (H), respectively; Mg 2ϩ was added to 0.35 mM (L) and 3.5 mM (H), respectively; and Ca 2ϩ was added to 0.01 mM (L) and 0.35 mM (H), respectively. mM Mn 2ϩ (a concentration presumably falling in the range accumulated in the cytoplasm when Mn 2ϩ uptake was enhanced; see Ref. 1) caused C 21 , A 22 , A 49 , and G 50 to be modified less by DMS and A 29 , A 52 , A 53 , A 54 , and G 55 to be modified more (Fig. 5, B and D, lanes 4), indicating that this high Mn 2ϩ level likely enhances UTR1 to form the same S2 structure in Mg 2ϩ -depleted and Mg 2ϩ -containing conditions.
We also compared modifications of UTR1 treated by Mn 2ϩ as well as Mg 2ϩ and Ca 2ϩ . We chose 0.35 and 3.5 mM Mg 2ϩ to investigate the UTR1 folding; these were used previously as low and high Mg 2ϩ conditions in structural mapping of the mgtA 5Ј-LR, respectively (26). In comparison to the stem-loop T region in S1 and S2 (Fig. 5A), A 52 , A 53 , A 54 , and G 55 , which were paired in stem C formed in S1 and unpaired in loop T formed in S2, were indeed modified less in low Mg 2ϩ than in high Mg 2ϩ (Fig. 5D, lanes 5 and 6). On the other hand, A 49 and G 50 , which resided in 46 CAUAGC 51 (stem-L), were modified more in high Mg 2ϩ than in low Mg 2ϩ , suggesting that they stayed in a singlestrand region in high Mg 2ϩ . This contrasted with the Mn 2ϩtreated reaction (Fig. 5D, lane 4), in which this switching sequence formed stem T (i.e. paired) in S2 (Fig. 5A). Thus, we postulated that Mg 2ϩ is unlikely to facilitate the formation of an intact terminator structure. In addition, we mapped the conformation of UTR1 treated with 0.01 mM (i.e. low) and 0.35 mM (high) Ca 2ϩ , respectively. We observed that more nucleotides could be modified by DMS after the RNA was treated in both Ca 2ϩ conditions (Fig. 5D, lanes 8 and 9). Particularly, both the stem-L and stem-R regions could still be modified in high Ca 2ϩ (Fig. 5D, lane 9), probably due to the relaxed structure of the UTR1. Thus, it is also unlikely for this cation to induce a terminator structure in the UTR1.
Mn 2ϩ -facilitated Cleavage at the Specific Nucleotides of the UTR1-Specific divalent metal cations have been shown to induce site-specific cleavage when they bind tRNAs (41)(42)(43). Particularly, Mn 2ϩ can induce hydrolysis of yeast tRNA Phe and Elongator tRNA Met , mainly in their D-loops, and also tRNA Glu in the anticodon loop (43). We incubated UTR1 RNA in a buffer supplemented with varying amounts of Mn 2ϩ to examine whether Mn 2ϩ could facilitate specific cleavage of UTR1 RNA when it interacted with the riboswitch domain. Specific nucleotides in which the Mn 2ϩ -facilitated cleavage took place were characterized by monitoring truncated UTR1 fragments generated from wild-type full-length UTR1 RNA. We observed RNA fragments containing 47-, 54-, and 55-nt UTR1, respectively, in which the level was proportional to the Mn 2ϩ concentration supplemented (Fig. 6A, lanes 2-4), indicating that Mn 2ϩ induced a strong cleavage of the 3Ј,5Ј-phosphodiester bond at nucleotides A 47 , A 54 , and G 55 in a concentration-dependent manner; these are located in stem-loop T (Fig. 5A). The effect of this ligand was validated by supplementation of EDTA, which significantly reduced cleavage at these sites (Fig. 6A, lane 5). However, all three of these sites were seldom cleaved when stem-R[mut] UTR1 was incubated with Mn 2ϩ at the same levels as for wild-type UTR1. It is worth pointing out that the nucleotides at these sites keep the same nucleotides in the wild type and the substituted UTR1 (Fig. 5, A and C). This suggests that cleavage would take place only if nucleotides 47, 54, and 55 were placed in stem-loop T. We also observed minor cleavages occurring at U 17 and G 42 (Fig. 6A, lane 4). Interestingly, all five of these nucleotides are located in base-paired stem regions in S1 (the conformation without Mn 2ϩ ligand), but two of them, A 54 and G 55 , are switched to loop T region in S2. We reasoned that the loop nucleotides might be flexible for Mn 2ϩ binding. In stem-R[mut] UTR1, the strong cleavage sites were U 17 and C 27 , which could be weakly or rarely cleaved in wild-type UTR1 (Fig.  6A, lanes 9 and 4). Although it remains to be investigated whether specific cleavage of the UTR1 would reflect a direct Mn 2ϩ interaction with these specific nucleotides, we postulated that Mn 2ϩ could bind to central atoms adjacent to specific cleavage sites located in stem-loop T in wild-type UTR1, but located in to stem-loop A in R[mut] UTR1. The products derived from cleavage at A 47 , A 54 , and G 55 could not be detected when wild-type UTR1 was incubated with the same amount of Mg 2ϩ or Ca 2ϩ (Fig. 6B, lane 2 versus lanes 3 and 4), which again demonstrates that the UTR1 is a riboswitch that is highly specific to Mn 2ϩ .
Salmonella mgtA Mg 2ϩ Riboswitch Can Sense Mn 2ϩ -In vitro, Mn 2ϩ can induce similar folding of the Mg 2ϩ riboswitch characterized from B. subtilis mgtE (27). However, a previous finding showed that Mn 2ϩ is unable to repress a mgtA transcription in Salmonella, which is controlled by its leader region (26). We reasoned that 0.025 mM Mn 2ϩ supplemented in the experiment could not raise the cytoplasmic concentration of Mn 2ϩ over a threshold level to act on this riboswitch. Thus, we introduced a plasmid (pYS1008) carrying the mntH-FLAG fusion gene under the control of promoter P lac into the wildtype strain carrying pYS1010. Overexpression of mntH to facilitate Mn 2ϩ uptake significantly inhibited the expression of lacZ controlled by mgtA 5Ј-LR, as ␤-galactosidase activity in the wild-type strain carrying pYS1010 and pYS1008 was 25-fold lower than that carrying pYS1010 and pUHE21 (vector) when bacteria were grown in N medium with 0.01 mM Mg 2ϩ (a low FIGURE 7. The Salmonella mgtA Mg 2؉ riboswitch also responds to cytoplasmic Mn 2؉ in a manner similar to Mg 2؉ . A, ␤-galactosidase activity was determined in the wild-type strain harboring pYS1010 and pYS1008 (pmntH). Bacteria were grown for 6 h in N medium (0.01 mM Mn 2ϩ ) supplemented with 0.5 mM IPTG and 0 or 0.01 mM Mn 2ϩ . B, ␤-galactosidase activity was determined in mntH-FLAG corA-FLAG strain (YS10211) harboring pYS1010 and pYS1008 (pmntH) or one of the pYS1008 derivatives (pYS1013 (D34E) or pYS1014 (E102D)) grown under the same conditions as in A. C, immunoblot analysis of MntH protein. The level of MntH-FLAG protein from the cultures in B was determined by Western blot. M2 anti-FLAG antibodies (Sigma) were used. Cellular lysate of the wild type (14028s) was used as a negative control of MntH-FLAG (shown as C). Production of CorA-FLAG is dependent on expression of MntH. The mntH-FLAG corA-FLAG strain harboring pUHE21 (vector) was used as a negative control of heterogeneous overproduction of MntH-FLAG (shown as Ϫ). D, ␤-galactosidase activity was determined in wild-type strain 14028s harboring pYS1008 and pYS1010 or pYS1011 grown in N medium (0.01 mM Mg 2ϩ ) supplemented with 0.5 mM IPTG and 0 or 0.01 mM Mn 2ϩ . E, in vitro transcription of a template harboring the P lac1-6 promoter and the full-length wild-type mgtA 5Ј-LR sequence conducted in buffer containing 0.35 mM Mg 2ϩ without (Ϫ) or with (ϩ) the addition (3.5 mm) of one of the tested divalent cations.
Mg 2ϩ condition allowing transcription to pass through the 5Ј-LR (26)), 0.01 mM Mn 2ϩ , and IPTG (Fig. 7A, right panel,  column 2 versus 1). On the other hand, lacZ expression remained similarly activated in the strains harboring pYS1010, and either pYS1008 or pUHE21, when they were grown in the medium with low Mg 2ϩ and IPTG but without 0.01 mM Mn 2ϩ (Fig. 7A, left panel, column 2 versus 1), indicating that only overexpression of mntH without adding Mn 2ϩ was unable to repress this lacZ transcription. It has been shown that Asp 34 and Glu 102 residues are conserved in Salmonella and E. coli and are essential for MntH-dependent Mn 2ϩ transport (45). We constructed two pYS1008 derivatives, pYS1013 and pYS1014, which directed the synthesis of two substituted MntH proteins in which Asp 34 and Glu 102 were changed to Glu 34 and Asp 102 , i.e. D34E and E102D, respectively. Expression of lacZ was similar in the wild-type strain harboring pYS1010 and pYS1008 or one of its derivatives under the low Mg 2ϩ condition without added Mn 2ϩ but with IPTG (Fig. 7B, columns 1-3, left panel). When 0.01 mM Mn 2ϩ was added to the strain harboring pYS1010 and wild-type pYS1008, ␤-galactosidase activity was reduced ϳ24-fold (Fig. 7B, column 1, right panel versus column 1, left panel). However, Mn 2ϩ supplementation could not change lacZ expression in the strain harboring pYS1010 and pYS1013 or pYS1014, as ␤-galactosidase activity remained similar in these strains regardless of Mn 2ϩ (Fig. 7B, columns 2 and  3, right panel versus columns 2 and 3, left panel). Although expression of D34E or E102D MntH was unable to repress mgtA 5Ј-LR transcription, immunoblot results showed that the level of mutated MntH protein produced was similar to that of the wild-type protein under the same inducing condition, all of which was much higher than the level of MntH expressed from the chromosomal locus (Fig. 7C). These results indicate that mgtA Mg 2ϩ riboswitch can sense Mn 2ϩ when Mn 2ϩ uptake is enhanced by MntH transporter. It has been shown that mgtA 5Ј-LR forms a stem-loop B at high cytoplasmic Mg 2ϩ concentrations to mediate transcription termination (26). We sought to determine whether mgtA 5Ј-LR also requires this stem-loop by constructing a pYS1010 derivative, pYS1011, which carried a substituted right arm of stem-loop B. Under low Mg 2ϩ and IPTG-containing conditions, Mn 2ϩ supplementation reduced ␤-galactosidase activity in the wild-type strain harboring pYS1010 and pYS1008 by ϳ23-fold (Fig. 7B, column 1, right  panel versus column 1, left panel), whereas ␤-galactosidase activity remained similar in the strain harboring pYS1011 and pYS1008 grown under the same conditions regardless of Mn 2ϩ (Fig. 7D, column 2, right panel versus left panel). Also, we conducted an in vitro transcription assay to determine whether mgtA 5Ј-LR could sense Mn 2ϩ and Ca 2ϩ in addition to Mg 2ϩ . When transcription of the full-length mgtA 5Ј-LR was initiated by P lac1-6 , consistent with a previous observation (26), the reactions with a low Mg 2ϩ condition (0.35 mM) mainly produced a 264-nt runoff RNA (Fig. 7E, lanes 1, 3, and 5, respectively, marked Ϫ), and a high Mg 2ϩ condition (3.5 mM) strongly facilitated transcription termination to produce a 220-nt truncated RNA (Fig. 7E, lane 2, ϩ). Adding either 0.35 mM Mn 2ϩ or Ca 2ϩ to the low Mg 2ϩ reaction induced the 220-nt truncated product to a level similar to that of Mg 2ϩ (Fig. 7E, lanes 4 and 6 versus  lane 2). Taking these findings together, we concluded that Mn 2ϩ and Ca 2ϩ , just like Mg 2ϩ , are able to act on mgtA 5Ј-LR in a manner similar to Mg 2ϩ and that the mgtA Mg 2ϩ sensor should be considered a general divalent cation sensor.
Concluding Remarks-The discovery of the Mn 2ϩ riboswitch from the Salmonella mntH gene should provide new insights into specific transcriptional regulation employing riboswitches to respond to specific divalent metal cations. It also reveals a feedback genetic control, which modulates the transcription of the bacterial Mn 2ϩ transporter gene through control of the elongation step by sensing Mn 2ϩ . As transcription of the mntH gene is regulated in response to not only Mn 2ϩ but also other cations, the biological significance of the Mn 2ϩ riboswitch may lie in the selective sensing of Mn 2ϩ but not of other divalent cations that can be transported through MntH. Mn 2ϩ also interacts with a Mg 2ϩ riboswitch in Salmonella, mgtA 5Ј-LR, and modulates transcription of the coding region of the Mg 2ϩ transporter. This suggests a coordinating regulation that contributes to the maintenance of divalent cation homeostasis by which Salmonella can modulate the level of cytoplasmic Mg 2ϩ and probably some other divalent cations by responding to Mn 2ϩ .
In Fig. 5A we have summarized a working model that describes the regulatory action of the UTR1. In low Mn 2ϩ , 34 CAUAGC 39 and the following 40 AUG 42 sequence play the role of anti-terminator motifs in S1, in which 40 AUG 42 basepairs 46 CAU 48 to form stem-loop B, and 34 CAUAGC 39 basepairs 56 GCUAUG 61 ; together they prevent these two sequences from forming terminator stem-loop T. In high Mn 2ϩ , on the other hand, the anti-terminator 20 UCAU 23 and the following 24 UAU 26 base-pair anti-terminator motifs to form stem-loop E in S2, which allows 46 CAU 48 as well as 49 AGC 51 to pair with 56 GCUAUG 61 and form stem-loop T.
Two independent mechanisms most likely participate in the regulation of the mntH expression in response to the cytoplasmic Mn 2ϩ through transcription initiated from the ϩ1 transcription start site. Low levels of cellular Mn 2ϩ allow transcription to take place when the R1 sequence is not occupied by MntR. Synthesis of MntH transporter will facilitate uptake of Mn 2ϩ , which causes an accumulation of cytoplasmic Mn 2ϩ . The Mn 2ϩ ion then binds to UTR1 as a nascent transcript, thus acting on its riboswitch motif to terminate mntH transcription before it arrives at the mntH coding region. We postulate that UTR1 serves as a fine-tuning device. When cytoplasmic Mn 2ϩ reaches further, to a higher level, it will activate MntR. Then, this regulator will bind to R1 and repress the transcription initiation from ϩ1. At that time point, it remains to be investigated how MntR can differentiate the transcription initiated alternatively from ϩ1 and ϩ1Ј. Although the Fur-box is located proximally downstream of ϩ1, it is unlikely to regulate the Mn 2ϩ -dependent transcription of mntH because Fur responds to Fe 2ϩ but not Mn 2ϩ (18,19).