Mg²⁺ regulates transcription of mgtA in Salmonella Typhimurium via translation of proline codons during synthesis of the MgtL peptide

Significance

Mg²⁺ is an important signal for the regulation of virulence and thermotolerance in Salmonella enterica. Transcription of the mgtA gene, which encodes a transporter for Mg²⁺, is highly induced by Mg²⁺ limitation. The 5′ leader of the mgtA mRNA encodes mgtL, a 17-codon, proline-rich open reading frame, whose translation controls the transcription of mgtA: efficient translation of mgtL results in termination of transcription upstream of the mgtA protein-coding region, whereas slow or incomplete translation of mgtL antagonizes this termination. We show that the proline codons in mgtL present an impediment to translation at low but not high Mg²⁺ concentrations, and thereby, we provide a model for how Mg²⁺ is coupled to the regulation of mgtA transcription.

Abstract

In Salmonella enterica serovar Typhimurium, Mg²⁺ limitation induces transcription of the mgtA Mg²⁺ transport gene, but the mechanism involved is unclear. The 5′ leader of the mgtA mRNA contains a 17-codon, proline-rich ORF, mgtL, whose translation regulates the transcription of mgtA [Park S-Y et al. (2010) Cell 142:737–748]. Rapid translation of mgtL promotes formation of a secondary structure in the mgtA mRNA that permits termination of transcription by the Rho protein upstream of mgtA, whereas slow or incomplete translation of mgtL generates a different structure that blocks termination. We identified the following mutations that conferred high-level transcription of mgtA at high [Mg²⁺]: (i) a base-pair change that introduced an additional proline codon into mgtL, generating three consecutive proline codons; (ii) lesions in rpmA and rpmE, which encode ribosomal proteins L27 and L31, respectively; (iii) deletion of efp, which encodes elongation factor EF-P that assists the translation of proline codons; and (iv) a heat-sensitive mutation in trmD, whose product catalyzes the m¹G37 methylation of tRNA^Pro. Furthermore, substitution of three of the four proline codons in mgtL rendered mgtA uninducible. We hypothesize that the proline codons present an impediment to the translation of mgtL, which can be alleviated by high [Mg²⁺] exerted on component(s) of the translation machinery, such as EF-P, TrmD, or a ribosomal factor. Inadequate [Mg²⁺] precludes this alleviation, making mgtL translation inefficient and thereby permitting mgtA transcription. These findings are a significant step toward defining the target of Mg²⁺ in the regulation of mgtA transcription.

Magnesium is important for many cellular processes, including enzymatic activity, nucleoside triphosphate-dependent phosphorylation reactions, and integrity of macromolecules and membranes (1). Furthermore, Mg²⁺ homeostasis is connected to thermotolerance in the food pathogen Salmonella enterica, because the survival of this organism at lethal high temperatures can be dramatically increased by the overproduction of Mg²⁺ transport proteins (2). Salmonella has three uptake systems for Mg²⁺: MgtA, MgtB, and CorA. The transcription of mgtA and mgtB, which is flanked by mgtC and mgtR in an operon, is inducible over a hundred-fold by Mg²⁺ limitation, whereas the transcription of corA is not regulated by Mg²⁺ (3).

Transcription of mgtA and the mgtCBR operon is dependent on the PhoQP two-component system, in which the inner membrane protein PhoQ carries out phosphorylation and dephosphorylation of the DNA-binding transcriptional regulator PhoP in response to periplasmic stimuli (4). The PhoQP system regulates directly or indirectly the transcription of 5% of the genes of Salmonella and Escherichia coli, including the phoPQ operon itself and genes involved in virulence, membrane composition, antimicrobial peptide resistance, and acid stress resistance (5). The kinase activity of PhoQ is stimulated by diverse signals, including low concentrations of Mg²⁺ (6), acidic pH (7), and a number of antimicrobial peptides (8). Because the cytoplasm of macrophages and phagosomes is acidic and limiting for Mg²⁺, it has been proposed that Salmonella uses the PhoQP system to induce virulence genes needed for growth inside host cells (6, 9, 10).

Superimposed on PhoQP-dependent regulation, there is a second layer of control of mgtA transcription. The mgtA mRNA has a 264 nucleotide-long 5′ leader region (LR) that contains self-complementary sequences predicted to form mutually exclusive secondary structures (stem loops “A” and “B” vs. “C”; Fig. 1) (11). It was proposed that the 5′ LR mRNA functions as a riboswitch that can adopt alternative secondary structures depending on intracellular concentrations of Mg²⁺ and thereby regulate whether transcription is terminated upstream of mgtA or allowed to continue (11). Unaccounted in this model was the presence of a short ORF, called mgtL, in the mgtA 5′ LR that encodes a proline (Pro)-rich leader peptide highly conserved in Enterobacteriaceae (Fig. 1) (12, 13). The role of mgtL is reminiscent of the regulatory functions of short ORFs in the trp, his, and other amino acid biosynthetic operons, in which the efficiency of translation of the leader peptide regulates termination or read-through of transcription into the rest of the operon (14). However, opposing conclusions were reached in previous work for the role of Pro in the regulation of mgtA expression: Park et al. (12) suggested that low levels of Pro-charged tRNA^Pro increase expression of mgtA, whereas Zhao et al. (13) concluded that Pro has no role in mgtA regulation. Subsequent discovery of the involvement of the Rho protein in the regulation led to a model in which high Mg²⁺ concentrations favor the formation of stem loops A and B, which expose a Rho-utilization (rut) site for this transcription termination factor, whereas low Mg²⁺ concentrations promote formation of stem loop C, which sequesters the rut site and precludes termination (11, 15).

Fig. 1.

Translation of Pro codons in mgtL is the Mg²⁺-sensing stimulus for the transcriptional control of the Salmonella mgtA gene. Transcription of mgtA is regulated at two steps: activation of the promoter by PhoP, which is phosphorylated by PhoQ in response to low [Mg²⁺] and other periplasmic signals (inset 1) (4), and translation of mgtL (encoded by nucleotides 71–124) (12), which governs folding of the 5′ LR mRNA and Rho-dependent termination. At very low intracellular [Mg²⁺], ribosomes stall during translation of mgtL (inset 2a), enabling the formation of stem loop C in the RNA, which sequesters the Rho-binding site rut and allows transcription to proceed into the mgtA coding region, turning mgtA transcription ON (inset 3a). At high intracellular [Mg²⁺], translation of mgtL is rapid and complete (inset 2b), facilitating stem loop B formation, which exposes the rut site and leads to Rho binding and transcription termination, turning mgtA OFF (inset 3b). EF-P, which assists the incorporation of Pro residues into nascent proteins (16) and TrmD, which catalyzes Mg²⁺-dependent m¹G37 methylation of all three tRNA^Pro species (26), are required for rapid translation of mgtL (inset 4). Nucleotide changes in red denote mutations that repress mgtA expression, and nucleotide changes in green denote mutations that turn on mgtA expression. Pro codons and residues are highlighted in yellow. Predicted folding of the mgtA 5′ LR mRNA is based on the MFold web server (12, 40). AdoHcy, S-adenosylhomocysteine; AdoMet, S-adenosylmethionine; G37, guanine at position 37 in tRNA^Pro; m¹G37, methylation at N¹ position of G37 in tRNA^Pro; RNA-P, RNA polymerase; RpmA and RpmE, ribosomal proteins L27 and L31, respectively. Parts of this figure were reproduced from ref. 12, with permission from Elsevier.

However, several questions about the proposed transcriptional control models of mgtA remain unresolved. It is unclear whether Mg²⁺ sensing is carried out entirely by the riboswitch or whether translation of mgtL plays a role in this process and how the proposed Mg²⁺ sensing via the folding of the riboswitch is compatible with the effects of translation of mgtL on the secondary structure of the 5′ LR. Finally, it is not obvious why abundance of Pro should be physiologically connected to the availability of Mg²⁺. We addressed the mechanism of transcriptional regulation of mgtA by analysis of cis- or trans-acting mutations that altered the expression of this gene in S. enterica serovar Typhimurium. We propose that translation of the Pro codons of mgtL is sensitive to the intracellular concentrations of Mg²⁺ and provides the regulatory stimulus for the transcriptional control of mgtA.

Results

A Mutation Resulting in Poly-Pro in mgtL Induces mgtA Expression.

In the wild-type strain carrying the mgtA9226::MudJ insertion, expression of lacZY is low on MacConkey agar, giving rise to the Lac⁻ (white) phenotype. By selecting Lac⁺ (red) mutants, we isolated three classes of derivatives carrying spontaneous mutations that conferred increased expression on the mgtA–lacZ fusion. One of these carried an A86C point mutation in mgtL, which resulted in a Thr to Pro substitution at codon 6 of this ORF, generating a run of three contiguous Pro codons (Fig. 1). A stretch of three Pro residues could constitute a ribosome-stalling motif sequence (16–19). Although mgtA–lacZ expression in the wild-type strain was up-regulated sevenfold upon decreasing [Mg²⁺] from 1.6 mM (high) to 0.016 mM (low), the A86C mutant exhibited a 13-fold increased expression of the mgtA–lacZ fusion at high Mg²⁺, compared with the wild type (Fig. 2). The constitutive “ON” phenotype of the A86C mutant generated by the poly-Pro mutation suggests that ribosome stalling in mgtL could be involved in the Mg²⁺-sensing regulation of mgtA transcription. A second spontaneous mutant carried a C98T transition that resulted in an Arg to nonsense substitution at codon 10 in mgtL (Fig. 1) and exhibited a 27-fold increased expression of the mgtA–lacZ fusion at high Mg²⁺ (Fig. 2). The same allele was previously obtained in an mgtA⁺ background in the selection of mutations that confer enhanced thermotolerance (2). The elevation of transcription of mgtA by the C98T mutation and other premature nonsense mutations in mgtL (12) supports the notion that impaired translation of mgtL enables transcription to continue into the mgtA structural gene.

Fig. 2.

Mutations in mgtL and the genes encoding ribosomal proteins L27 and L31, translation factor EF-P, and methyltransferase TrmD induce mgtA expression. Strains used were TL4295 (mgtL⁺ mgtA–lacZ), TL4576 [mgtL (A86C) mgtA–lacZ], TL4575 [mgtL (C98T) mgtA–lacZ], TL5151 [rpmA (G23A) mgtL⁺ mgtA–lacZ], TL5692 (∆rpmE mgtL⁺ mgtA–lacZ), TL5332 (∆efp mgtL⁺ mgtA–lacZ), and TL5856 (S88L-trmD mgtL⁺ mgtA–lacZ). Cultures were grown at 37° (lanes 1–6) and at 30° (lanes 7 and 8), and β-galactosidase was assayed as described in Materials and Methods.

Mutations in the Ribosomal Proteins L27 and L31 Genes Induce mgtA Expression.

In the above selection, we also obtained a derivative carrying the G23A mutation in the rpmA gene that resulted in a 23-fold increased expression of the mgtA–lacZ fusion at high Mg²⁺ (Fig. 2). Because the latter mutation affected a ribosomal protein, we tested the effect of other ribosomal mutations on the expression of mgtA. Deletion of the rpmE gene, which encodes the nonessential 50S ribosomal protein L31 (20), resulted in an 11-fold increased expression of the mgtA–lacZ fusion at high Mg²⁺ (Fig. 2). Both the rpmA (G23A) and ∆rpmE mutations are likely to impair the formation of peptide bonds, suggesting that the efficiency of translation of mgtL is impaired in these mutant strains. These results support the importance of translation of mgtL for the control of mgtA expression. However, deletion of rpmF, which encodes 50S ribosomal protein L32, did not alter the expression of mgtA (Fig. S1).

Fig. S1.

Deletion of genes encoding ribosomal protein L32, translation EFs EF-Tu and LepA, and methyltransferases TrmA, TrmB, and TrmE does not affect mgtA–lacZ expression. Strains TL5836 (∆rpmF mgtA–lacZ), TL 5917 (∆tufB mgtA–lacZ), TL5833 (∆lepA mgtA–lacZ), TL5856 (S88L-trmD mgtA–lacZ), TL5815 (∆trmA mgtA–lacZ), TL5817 (∆trmB mgtA–lacZ), and TL5819 (∆trmE mgtA–lacZ) were grown at 37°, and β-galactosidase was assayed as described in SI Materials and Methods.

Loss of the Translation Elongation Factor P Gene Induces mgtA Expression.

Elongation factor P (EF-P) is a universal protein that is needed for relieving ribosomes stalled during translation of stretches of multiple Pro codons (21). We constructed a ∆efp mutation and found it to confer a 10-fold increased expression of the mgtA–lacZ fusion at high Mg²⁺ (Fig. 2). During the preparation of this article, it has also been reported that loss of EF-P resulted in a 13-fold increase in the accumulation of the mgtA mRNA in cells grown with 0.5 mM Mg²⁺, as measured with quantitative RT-PCR (22). The role of EF-P is unique in this regard, because mutations in the tufA, tufB, and lepA genes, which encode other EFs (23, 24), did not have significant effects on the expression of mgtA (Fig. S1).

Mutation in the trmD Gene That Reduces m¹G37 Methylation of tRNA^Pro Induces mgtA Expression.

TrmD catalyzes m¹G37 methylation of all three species of tRNA^Pro, which is critical for translation of all Pro codons (25). Without this methylation, translation of Pro codons is prone to +1-frameshifting (25). TrmD is unique among AdoMet-dependent nucleic acid methyltransferases in that it requires Mg²⁺ for its activity (26). TrmD is essential for viability, but the heat-sensitive trmD27 (S88L) allele (27) enabled us to test whether reduction in the levels of m¹G37-tRNA^Pro would decrease the efficiency of translation of the Pro codons of mgtL. The S88L-trmD mutation, which impairs the activity of TrmD even at 30° (27), resulted in a 17-fold increase in mgtA–lacZ expression at this temperature (Fig. 2), indicating that the decrease in m¹G37-tRNA modification, and thus translation of Pro codons, increased expression of the reporter gene. Expression of mgtA–lacZ in both high and low Mg²⁺ was lower at the less permissive temperature of 37° than at 30° (Fig. S1), most likely due to a global detrimental effect on protein synthesis. In contrast, mutations in the trmA, trmB, and trmE genes, which encode other tRNA methyltransferases (28–31), did not have significant effects on mgtA–lacZ expression (Fig. S1).

Replacement of Three Pro Codons in mgtL Impairs Mg²⁺ Sensing for mgtA Expression.

To further probe the role of the Pro codons in mgtA regulation, we constructed the mgtL (C77T, C84T, C89T, or 3Pro–) mutant, in which the first three Pro codons of this short ORF were replaced with Ser, Leu, and Ser codons, respectively (Fig. 1). These mutations are not predicted to significantly alter the 5′ LR mRNA secondary structure or its ability to adopt the stem loop B or stem loop C conformation (Fig. 1). In the mgtL (3Pro–) mutant, expression of the mgtA–lacZ fusion was very low and insensitive to Mg²⁺ concentration (Fig. 3, lane 1). We ruled out the possibility that this mutation “locked” the 5′ LR mRNA secondary structure into a constitutively “OFF” conformation, because introduction of a nonsense mutation (G80T) into the strain carrying the mgtL (3Pro–) mutation resulted in high-level derepression of mgtA–lacZ (Fig. 3, lane 2). Our observation that the replacement of three Pro codons nearly abolished expression of mgtA suggests that the 5′ LR mRNA favors the formation of stem loop B and enables Rho to terminate transcription. Park et al. (12) and Zhao et al. (13) constructed mutants in which various combinations of Pro codons in mgtL were substituted by other amino acid codons, and they also found that substitutions of Pro codons in mgtL diminished the sensitivity of the expression of mgtA to the Mg²⁺ concentration.

Fig. 3.

Activation of mgtA–lacZ expression by mutations in the ribosomal proteins L27 and L31, translation factor EF–P, and methyltransferase TrmD genes requires mgtL Pro codons. Each strain carried the mgtL [C77T, C84T, C89T (i.e., 3Pro–)] mutations, mgtA–lacZ fusion, and other mutations, as indicated. Strains TL5666, TL5667 [mgtL (G80T)], TL5968 [rpmA (G23A)], TL5969 (∆rpmE), TL5967 (∆efp), and TL5848 (S88L-trmD) were grown, and β-galactosidase was assayed as described in Materials and Methods. For lanes 1 and 3–6, the β-galactosidase–specific activities were lower than the resolutions of the figure.

To further test the suggestion that the removal of three of the four Pro codons facilitates the translation of mgtL and thereby leads to termination of transcription of mgtA, we introduced the rpmA (G23A), ∆rpmE, ∆efp, and S88L-trmD mutations into the mgtL (3Pro–) mutant background. In contrast to the mgtL⁺ background, where each of these mutations elicited large increases in the expression of mgtA–lacZ (Fig. 2), these mutations did not elevate the expression of the mgtA–lacZ fusion in the mgtL (3Pro–) background (Fig. 3, lanes 3–6). These results demonstrate that the high level of expression of mgtA caused by the rpmA, rpmE, efp, and trmD mutations is due specifically to impaired translation of mgtL⁺ and not to global effects that these mutations might have on translation.

Availability of Mg²⁺ Positively Regulates mgtL Translation Without Involvement of mRNA Folding.

We hypothesize that abundant Mg²⁺ facilitates the translation of mgtL specifically at the steps of prolyl-bond formation, allowing the positioning of ribosomes to dictate folding of the 5′ LR mRNA differently at high and low Mg²⁺ (Fig. 1). To test this hypothesis, we investigated the effect of Mg²⁺ on translation of an mgtL–lacZ translational fusion. Because the 5′ LR mRNA of the mgtL–lacZ construct does not contain nucleotide sequences covering stem loops B and C, it cannot adopt alternative RNA conformations, allowing us to separate the direct effect of Mg²⁺ on mgtL translation from changes in mRNA folding. In contrast to the mgtA–lacZ transcriptional fusion, which was up-regulated by low Mg²⁺, the expression of the mgtL–lacZ translational fusion decreased in 0.016 mM Mg²⁺ compared with that seen in 1.6 mM Mg²⁺ (Fig. 4). In confirmation of our hypothesis that the Pro codons in mgtL present an obstacle to rapid translation, the rpmA (G23A), ∆rpmE, ∆efp, and S88L-trmD single mutations decreased the expression of the mgtL–lacZ fusion (Fig. S2), unlike their effects on the expression of the mgtA–lacZ fusion (Fig. 2). We examined the effect of more severe Mg²⁺ limitation on the expression of the mgtA–lacZ and mgtL–lacZ fusions by growing cells in nominally “Mg²⁺-free” M63 that contained only trace concentrations of Mg²⁺ initially. Expression of the mgtA–lacZ fusion was highest in the cells that reached stationary phase due to depletion of Mg²⁺. In contrast to the pattern of expression of the mgtA–lacZ fusion, which increased upon Mg²⁺ starvation, the expression of the mgtL–lacZ fusion decreased as a consequence of Mg²⁺ limitation, reaching a minimum at stationary phase in the Mg²⁺-free medium (Fig. 4).

Fig. 4.

Translation of mgtL responds to the extracytoplasmic Mg²⁺ concentration in a manner opposite to the transcription of mgtA. Strains TL4295 (mgtL⁺ mgtA–lacZ) and TL5508 (mgtL–lacZ) were grown at 37° to midexponential phase with 1.6 and 0.016 mM Mg²⁺ and under severe Mg²⁺ limitation in Mg²⁺-free M63-glucose for 3 h (Trace; 3 h) and 6 h to Mg²⁺ depletion (Depleted; 6 h), and β-galactosidase was assayed as described in Materials and Methods.

Fig. S2.

Translation of mgtL is adversely affected by mutations in the ribosomal proteins L27 and L31, translation factor EF-P, and methyltransferase TrmD. Each strain carried an mgtL–lacZ translational fusion, ∆mgtA, and other indicated mutations. Strains TL5508, TL5863 [rpmA (G23A)], TL5862 (∆rpmE), TL5829 (∆efp), and TL5847 (S88L-trmD) were grown at 37°, and β-galactosidase was assayed as described in SI Materials and Methods.

The his operon is regulated by attenuation via translation of the hisL ORF, which contains seven His codons (32). As a negative control, we determined the effects of Mg²⁺ availability and mutations in the translation machinery on the expression of this operon. Unlike the results obtained with mgtA–lacZ, expression of the hisD–lacZ fusion was not responsive to Mg²⁺ and was not elevated by the rpmA (G23A), ∆rpmE, ∆efp, and S88L-trmD single mutations to a similar extent as the expression of mgtA–lacZ (Fig. S3). These results further support the conclusion that induction of mgtA is the consequence of inefficient translation of Pro codons in mgtL.

Fig. S3.

Mutations in the genes encoding ribosomal proteins L27 and L31, translation factor EF-P, and methyltransferase TrmD have negligible effects on hisD–lacZ expression. Each strain carried a hisD–lacZ transcriptional fusion and other indicated mutations. Strains TL4960, TL5348 [rpmA (G23A)], TL5949 (∆rpmE), TL5344 (∆efp), and TL5856 (S88L-trmD) were grown at 37°, and β-galactosidase was assayed as described in SI Materials and Methods.

Discussion

Multitiered regulation of the Salmonella Mg²⁺ transport gene mgtA has been documented over the last decade. The PhoQP two-component system was suggested to activate transcription of mgtA at low extracellular Mg²⁺ concentrations (6, 9), and the mgtA 5′ LR mRNA has been proposed to function as an Mg²⁺-sensing riboswitch that regulates read-through or Rho-dependent transcription termination at a site upstream of the mgtA structural gene in response to intracellular Mg²⁺ (11, 15). Subsequently, it was noted that translation of the Pro codon-rich mgtL ORF encoded in the mgtA 5′ LR mRNA is involved in the regulation (12, 13). However, the questions of whether Mg²⁺-sensing is mediated by direct recognition of this cation by a riboswitch and what role the translation of mgtL plays in the regulation of expression of mgtA have not been settled adequately. We propose that the intracellular concentration of Mg²⁺ positively regulates the efficiency of translation of mgtL, which in turn affects the folding of the mgtA 5′ LR mRNA during translation. This Mg²⁺-dependent regulation serves as a determinant to control transcription termination by Rho at adequate Mg²⁺ and progression of the RNA polymerase into mgtA at low Mg²⁺.

Unique Mutations That Clarify the Role of mgtL Translation in mgtA Regulation.

The A86C mutation that introduced an extra Pro codon into mgtL, generating three adjacent Pro codons, conferred high-level constitutive expression of mgtA, whereas replacement of three Pro codons with Ser, Leu, and Ser rendered mgtA uninducible. Thus, Mg²⁺ sensing was greatly compromised by both increasing and decreasing the number of Pro codons in mgtL. These findings suggest that the codon composition of the mgtL⁺ sequence slows down translation just enough to make stalling exquisitely dependent on the Mg²⁺ concentration. Incorporation of Pro or other N-alkyl amino acids is an impediment to rapid translation (17). Nascent peptides containing stretches of Pro codons can induce ribosome stalling (18), and translation of the CCC and CCU Pro codons is particularly prone to +1-frameshifting (25). Our results for the poly-Pro mutant, which exhibited constitutively high mgtA–lacZ expression (Fig. 2), support the notion that the impaired translation in mgtL, due to stalling or +1-frameshifting, allows the formation of stem loop C, which sequesters the rut site within the mRNA secondary structure and precludes transcription termination by Rho. In addition, we isolated the C98T mutation in mgtL, which had been isolated previously by others (2, 11) and was shown to result in premature translation termination (12). Despite the ability to form stem loop B conformation for Rho-dependent transcription termination, the C98T mutation in mgtL confers the constitutively ON phenotype (Fig. 2), suggesting a different mode of regulation than the previously proposed Mg²⁺-sensing riboswitch, as it is not predicted to significantly alter the 5′ LR mRNA secondary structure.

We also characterized mutations unlinked to mgtA in genes encoding trans-acting proteins that potentially could be involved in regulating mgtA expression and provide insight into the mechanism of transcriptional regulation. One was the G23A missense mutation in rpmA, which encodes the 50S ribosomal protein L27. This mutation, which generated a Gly to Glu substitution at position 8 near the N terminus of the protein product, resulted in constitutively high mgtA–lacZ expression (Fig. 2). This essential protein is located near the peptidyltransferase center (PTC) of the ribosome, and its N-terminal end is important for stabilizing A-site tRNAs during peptidyltransfer (33). Deletion of the rpmE gene, which encodes the 50S ribosomal protein L31, likewise conferred constitutively high mgtA–lacZ expression (Fig. 2). The latter protein, which can be cross-linked to P-site tRNAs (34), is also located near the PTC (20). We showed that loss of EF-P, which assists the incorporation of Pro residues into nascent peptides, also led to increased expression of mgtA at high Mg²⁺ (Fig. 2). These mutations provide evidence that the efficiency of translation of the Pro codons in mgtL is an important determinant of mgtA expression. Furthermore, because the 5′ LR of mgtA is intact in the rpmA, rpmE, efp, and trmD mutants, the fact that these mutations nevertheless conferred high-level, constitutive mgtA transcription provides evidence against the existence of a riboswitch that regulates mgtA transcription in response to [Mg²⁺].

Pro Codons Within mgtL Are Critical for Mg²⁺ Sensing.

We constructed an mgtL (3Pro–) mutant in which the first, second, and third Pro codons of mgtL were replaced by Ser, Leu, and Ser codons, respectively. In agreement with our proposed hypothesis, the mgtL (3Pro–) mutant exhibited the constitutive OFF phenotype even in low Mg²⁺ (Fig. 3). The introduction of the rpmA (G23A), ∆rpmE, and ∆efp single mutations into the mgtL (3Pro–) mutant background did not increase the expression of the mgtA–lacZ fusion (Fig. 3), unlike what was observed in the mgtL⁺ background (Fig. 2). The mgtL (3Pro–) mutant does not require ribosomal rescue by EF-P, and ribosomes are presumably able to complete translation of mgtL rapidly, which would facilitate the formation of stem loop B and thereby expose the rut site to Rho and terminate transcription.

Translation of mgtL Is Regulated by Mg²⁺.

The data in Fig. 4 show that the concentration of Mg²⁺ is positively correlated with translation of mgtL and negatively correlated with transcription of mgtA. Because the mgtA 5′ LR mRNA of the mgtL–lacZ construct is unable to form alternative stem loops B and C, our results support the inference that differential Mg²⁺-sensing control of mgtA expression is mediated by the translation of mgtL and not by the previously proposed Mg²⁺-sensing riboswitch (11).

The m¹G37 Methylation of tRNA^Pro by TrmD Is Involved in the Mg²⁺-Dependent Regulation of mgtA Expression.

TrmD methylates G37 of tRNA^Leu that recognizes the CUU and CUC codons, tRNA^Arg that recognizes the CGG codon, and all three tRNA^Pro species that recognize the four Pro codons (CCC, CCU, CCG, and CCA) (35). The 17 codons of mgtL contain one CCC, one CCG, and two CCU Pro codons; one CUC and one CUU Leu codon; and one CGG Arg codon (Fig. 1). The heat-sensitive S88L-trmD mutation resulted in a dramatic increase in mgtA–lacZ expression (Fig. 2) and in a reduction in the synthesis of the mgtL–lacZ translational fusion product (Fig. S2). This observation and the result that the S88L-trmD mutation had no effect on the expression of the mgtA–lacZ fusion in the strain that also carried the mgtL (3Pro–) mutation (Fig. 3) indicate that m¹G37 methylation of tRNA^Pro is critical for efficient translation of Pro codons in mgtL.

PhoP-Dependent Activation of mgtA Is Not Completely Inhibited at High Mg²⁺ Concentrations.

One facet of the current model of Salmonella pathogenesis is that the PhoQP system, which is a global regulator of a number of virulence genes, is not active at high Mg²⁺ concentrations that are likely found in environments where bacteria are free-living but becomes activated when bacteria infect animal hosts where the concentration of Mg²⁺ is presumably low (6). The significance of regulation of pathogenesis-related genes by physiological concentrations of Mg²⁺ is controversial (3). In several strains we isolated—for example, the mgtL (A86C), mgtL (C98T), rpmA (G23A), ∆rpmE, ∆efp, and S88L-trmD mutants (Fig. 2); the derivative carrying mgtL–lacZ fusion (Fig. 3); and in mutants isolated by O’Connor et al. (2)—the mgtA–lacZ transcriptional and the mgtL–lacZ translational fusions were expressed at high levels in media containing high Mg²⁺. In fact, some of the spontaneous, constitutively ON mutants were isolated as strains that formed Lac⁺ colonies on MacConkey agar containing repressing Mg²⁺ concentrations. The expression of the mgtA gene in such mutants was still dependent on the transcriptional activator PhoP (2). These results show that despite the decreased transcription of the phoPQ operon and reduced phosphorylation of the PhoP protein at high Mg²⁺ (4), transcription could be initiated at high levels from the mgtLA promoter in these constitutively ON mutants.

Because we used derivatives of the attenuated S. enterica serovar Typhimurium strain LT2 for our work, whereas much of the characterization of the PhoQP regulon has been carried out in the pathogenic strain ATCC 14028s (6, 9), the possibility existed that lack of repression of the mgtLA promoter at high Mg²⁺ might be a feature of LT2 derivatives. However, we ruled out this possibility by observing that mgtA–lacZ fusion was expressed at similar high levels at 10 mM Mg²⁺ in LT2 and ATCC 14028s strains harboring the mgtL (C98T) mutation or deletion of the 5′ LR (∆9–244) (Fig. S4). These results indicate that there is no substantial difference in Mg²⁺ sensing by the PhoQP systems of the two strains. It has been shown (36) that when synthesized at high levels, unphosphorylated PhoP can activate transcription from some of its target promoters in the absence of the PhoQ sensor kinase and thus independently of sensing of periplasmic Mg²⁺. In our strains, PhoP is produced at wild-type levels, but nevertheless it is able to turn on transcription from the mgtLA promoter in the constitutively ON mutants even at high Mg²⁺ concentrations, suggesting that high-level phosphorylation of this protein may not be necessary for the transcriptional activation of this promoter.

Fig. S4.

Expression of mgtA is not completely shut off by high Mg²⁺ in constitutive mutants in both the attenuated LT2 strain background and in the pathogenic ATCC 14028s background. S. enterica serovar Typhimurium LT2 and ATCC 14028s strains TL4295 and TLa4255 (both mgtA–lacZ), TL4575 and TLa4261 [both mgtL (C98T) mgtA–lacZ], and TL4697 and TLa5936 [both 5′ LR (∆9–244) mgtA–lacZ] were grown at 37° to midexponential phase in glucose-M63 containing 10 mM or 0.01 mM Mg²⁺, and β-galactosidase was assayed as described in SI Materials and Methods.

Concluding Remarks

The main feature of our regulatory model for the expression of mgtA is that the Pro codons in mgtL present a “speed bump” to translation that can be overcome by high Mg²⁺ concentrations (Fig. 1). The activity of enzymes is generally dependent on Mg²⁺ concentration (1), and it has been known for a long time that Mg²⁺ starvation causes disintegration of polysomes and 70S ribosomes (37). However, we postulate that the translation of mgtL is more sensitive to Mg²⁺ concentrations than other global processes. The efficiency of translation of mgtL in turn directs the formation of the secondary structures of the 5′ LR mRNA. High Mg²⁺ concentrations facilitate rapid, complete translation of mgtL and favor a secondary structure that exposes the rut site to the Rho protein, terminating transcription upstream of mgtA. Conversely, low Mg²⁺ concentrations result in ribosome stalling or +1-frameshifting during translation of mgtL, promoting the formation of a different structure in which the rut site is hidden, and transcription can proceed into the mgtA structural gene. The 50S ribosomal proteins RpmA and RpmE, the translation factor EF-P, and the methyltransferase TrmD have roles in the expression of mgtA via translation of mgtL, highlighting an important relationship among Mg²⁺ homeostasis, efficiency of translation, and tRNA^Pro methylation. The discovery of which component or function of translation (e.g., RpmA, RpmE, EF-P, TrmD, or some other ribosomal protein) or prolyl peptide bond formation might be the specific Mg²⁺ sensor presents an exciting challenge for the understanding the regulation of mgtA transcription.

Materials and Methods

Strains and plasmids used in this work are listed in Table S1. Oligonucleotides used are in Table S2. Expression of mgtA was monitored by β-galactosidase assays (38) in strains carrying mgtA9226::MudJ lacZ transcriptional or mgtL–lacZ translational fusions grown in glucose-minimal medium 63 (M63) (2) with indicated concentrations of Mg²⁺, as described in Supporting Information.

Strains TL4575 [mgtL (C98T) mgtA9226::MudJ] and TL4576 [mgtL (A86C) mgtA9226::MudJ], which carry a nonsense and a Thr to Pro codon substitution in mgtL, respectively, are spontaneous Lac⁺ derivatives of TL4295 (mgtA9226::MudJ) isolated on MacConkey agar. We obtained a third spontaneous Lac⁺ derivative of TL4295, strain TL5151, in which the mutation giving rise to the Lac⁺ phenotype was unlinked to mgtA9226::MudJ. This mutation was identified as a G to A transition at nucleotide 23 of the rpmA gene, which encodes 50S ribosomal protein L27 (see Table S1).

Strain MA11988 (hisL::mgtL-lacZY, where mgtL–lacZ is translational fusion inserted into the hisL locus) was constructed by a “scarless” recombineering procedure (39), as elaborated in Table S1. The mgtL–lacZ fusion was transferred to the chromosomal mgtL locus by P22 transduction to generate strain TL5508 (see Table S1). Similar recombineering was used to construct strains carrying Pro codon changes in mgtL (C77T, C84T, C89T) and (see Table S1) in a strain carrying mgtA744::MudK translational fusion (C77T, G80T, C84T, C89T). The mgtA744::MudK translational fusion was exchanged with the mgtA9226::MudJ transcriptional fusion by P22 transduction to generate TL5666 [mgtL (C77T, C84T, C89T) mgtA9226::MudJ] and TL5667 [mgtL (C77T, G80T, C84T, C89T) mgtA9226::MudJ] as described in Table S1.

SI Results

Deletion of the Ribosomal Protein L32 Gene Does Not Induce mgtA Expression.

To further investigate the influence of 50S ribosomal proteins in mgtA regulation, we obtained a viable knockout of the rpmF gene from the comprehensive S. Typhimurium mutant collection constructed by Porwollik et al. (41). The RpmF or L32 protein is located near the PTC in the 50S ribosomal subunit. The ∆rpmF::Kan mutation did not affect mgtA–lacZ expression (Fig. S1), indicating that RpmA and RpmE are specific ribosomal proteins involved in translation of mgtL and transcription of mgtA.

Mutations in the EFs EF-Tu and LepA Genes Do Not Induce mgtA Expression.

EF-Tu, encoded by two nearly identical genes, tufA and tufB, brings aminoacylated tRNAs to the A site and performs GTP hydrolysis upon accurate codon–anticodon reading (23). The accommodated aminoacylated tRNA in the A site serves as the acceptor for peptidyl transfer from the P-site tRNA. After peptide bond formation, the essential translocation factor EF-G catalyzes GTP-dependent translocation to move the A-site tRNA to the P site and P-site tRNA to the E site (42). EF-4 (or LepA), ubiquitous to all cells, competes with EF-G and induces back-translocation, allowing EF-G a chance to correct errors (24). We investigated the contribution of each of these translation factors in mgtA regulation by obtaining tufB::MudJ (23) and ∆lepA::Kan knockouts (41). Neither the tufB nor the lepA mutation had a major effect on mgtA–lacZ expression (Fig. S1). A triple-mutant strain carrying the mgtA–lacZ transcriptional fusion along with the tufA (Q125R) and tufB::MudJ mutations was Lac⁻ on MacConkey agar like its tufA⁺ tufB⁺ counterpart, indicating that EF-Tu does not have a specific role in the regulation of translation of mgtL.

Mutations in the Methyltransferases TrmA, TrmB, and TrmE Do Not Induce mgtA Expression.

Modification of tRNAs is an important process for accurate identification and incorporation of amino acids into nascent peptides during translation. Multiple enzymes modify tRNAs by methylation at different positions (43). We investigated the contribution of three other methyltransferases, TrmA, TrmB, and TrmE, which modify all tRNAs at the U54, G46, and U34 positions, respectively (28–31). We obtained strains carrying the knockouts ∆trmA::Cm, ∆trmB::Cm, and ∆trmE::Cm from the S. Typhimurium mutant collection (41). Although the absence of the TrmA, TrmB, and TrmE methyltransferases may affect protein synthesis on a global scale, the ∆trmA::Cm, ∆trmB::Cm, and ∆trmE::Cm mutations had only minor effects on the expression of mgtA–lacZ expression (Fig. S1), supporting the role of TrmD as a trans-acting factor that is specific for translation of mgtL.

Translation of mgtL Is Impaired by Mutations in rpmA, rpmE, efp, and trmD.

In contrast to the increase in transcription of the mgtA–lacZ fusion by the rpmA, rpmE, efp, and trmD mutations at high Mg²⁺ (Fig. 2), these mutations reduced the expression of the mgtL–lacZ translational fusion. The fact that the expression of the mgtL–lacZ fusion is down-regulated whereas the expression of the mgtA–lacZ fusion is elevated suggests that the impaired translation of mgtL in these mutants is the consequence of a specific need for the RpmA, RpmE, EF-P, and TrmD proteins for efficient translation of this ORF rather than a global effect of these mutations on translation.

Altered TrmD Activity Does Not Induce Expression of Another Leader Peptide-Regulated System.

The transcription of the his (histidine biosynthetic) operon is regulated by attenuation via translation of the hisL leader ORF that harbors seven contiguous histidine codons (32). The expression of the hisD–lacZ fusion was not responsive to the external Mg²⁺ concentration and was not induced to a similar extent as the mgtA–lacZ fusion by the rpmA (G23A), ∆rpmE, ∆efp, and S88L-trmD single mutations (Fig. S3). Although each of these mutations as well as the Mg²⁺ concentration could have pleiotropic consequences on gene expression, these results support the conclusion that these mutations and the Mg²⁺ concentration influence the expression of mgtA by their effects on the translation of mgtL.

Expression of mgtA Is Not Completely Shut Off by High Mg²⁺ in the Attenuated LT2 and the Pathogenic ATCC 14028s Strains.

The mgtA9226::MudJ transcriptional fusion, together with the closely linked mgtL (C98T) and 5′ LR (∆9–244) mutations, was transduced from LT2 (i.e., TL1) derivatives into strain ATCC 14028s (Table S1). C98T is a nonsense mutation in mgtL and ∆9–244 is a deletion of nucleotides 9–244 of the 5′ LR of mgtA, including all of mgtL and sequences forming stem loops A, B, and C. In both constructs as well as in the wild-type background, the mgtA–lacZ fusion is transcribed from the native, PhoP-dependent promoter.

In both the LT2 and ATCC 14028s backgrounds, these mgtL mutations conferred a high level of expression of the mgtA–lacZ fusion even at 10 mM Mg²⁺, indicating that despite the reduced level of expression and reduced phosphorylation of PhoP, transcription can be initiated efficiently from the mgtA promoter at high Mg²⁺. The pH of minimal medium M63 containing 0.01 mM or 10 mM Mg²⁺ used in these experiments was 7.2, and therefore, it was not mildly acidic to result in activation of the PhoQP system.

SI Materials and Methods

Media and Growth Conditions.

Cells were grown with aeration at 37°, unless otherwise indicated. Media used were LB (44) and minimal medium M63 (2) that was made up without Mg²⁺ (Mg²⁺-free M63) and augmented with the indicated concentrations of Mg²⁺ (added as MgSO₄). The pH of M63 without and with up to 10 mM MgSO₄ was 7.2. The trace Mg²⁺ concentration in the nominally Mg²⁺-free glucose-M63 was determined by inductively coupled plasma mass spectrometry (ICP-MS) in a representative sample to be 9 μM. Carbon source used in M63 for all experiments, except for some steps of strain construction, was 10 mM glucose. When used as carbon sources, glycerol and rhamnose were at 2 g/L. His⁻ mutants were grown with 0.2 mM histidine in glucose-M63. When used, antibiotics were at the following concentrations: sodium ampicillin (Ap), 100 mg/L; chloramphenicol (Cm), 12.5 mg/L; kanamycin sulfate (Kan), 25 mg/L; and tetracycline (Tc), 20 mg/L. Resistance and sensitivity to chemicals are indicated by superscripts ^R and ^S, respectively. Solid media contained 20 g Difco agar per liter. MacConkey medium contained, per liter, 25 g MacConkey agar (Difco), 7.75 g agar (Difco), 5 g lactose, and 15 mg neutral red, in 1 L H₂O. The Mg²⁺ concentration of the MacConkey medium in one representative sample was 0.8 mM, as determined by ICP–MS.

Determination of the Effect of the Extracytoplasmic Mg²⁺ Concentration on the Expression of mgtA–lacZ and mgtL–lacZ.

Cells were grown from single colonies in LB to stationary phase, diluted 1:100 into glucose-minimal medium M63 with high (1.6 mM) or low (0.016 mM) MgSO₄. After growth to stationary phase, cells were diluted 1:25 into fresh glucose-M63 supplemented with the same respective high or low concentrations of Mg²⁺, grown to midexponential phase (∼5 × 10⁸ cells/mL), and samples were harvested for β-galactosidase assays. To impose severe Mg²⁺ limitation for the experiment in Fig. 4, cells were grown in LB to stationary phase, diluted 1:100 into glucose-M63 containing 0.16 mM MgSO₄, and grown overnight. Cells were washed three times in Mg²⁺-free M63, diluted 1:25 into fresh Mg²⁺-free glucose-M63, and grown. Samples were harvested at 3 h and 6 h of growth for β-galactosidase assays.

β-Galactosidase Assays.

Beta-galactosidase activity was determined, as described in ref. 38. Data for each experiment are the averages ± SD of results obtained for three independent cultures, each of which was assayed in two or three replicates.

Strain Construction.

The derivation of strains used for this work is presented in Table S1. All of the experiments shown in Figs. 1–4 and Figs. S1–S3 were carried out with derivatives of S. enterica serovar Typhimurium strain TL1 (LT2). The transducing phage used for strain construction was P22 HT105/1 int-201 (P22, hereafter). Transductions and the verification that transductants were P22-free and P22-sensitive were carried out as described in ref. 44.