Sequestration of a two-component response regulator by a riboswitch-regulated noncoding RNA

See allHide authors and affiliations

Science  22 Aug 2014:
Vol. 345, Issue 6199, pp. 940-943
DOI: 10.1126/science.1255083

A dual-action RNA switch for expression

Riboswitches are short segments of RNA that bind small molecules and switch between two different conformations, thereby regulating gene expression (see the Perspective by Chen and Gottesman). DebRoy et al. and Mellin et al. find a new class of riboswitches—in two different species of bacteria—that are both part of and regulate the production of a noncoding RNA. Each riboswitch ensures that a particular metabolic pathway is only activated in the presence of an essential small-molecule cofactor. In the absence of the cofactor, the full-length non-coding RNA is made and binds a regulator protein, preventing the regulator protein from inappropriately activating the metabolic pathway.

Science, this issue p. 937 and p. 940; see also p. 876


Riboswitches are ligand-binding elements contained within the 5ʹ untranslated regions of bacterial transcripts, which generally regulate expression of downstream open reading frames. Here, we show that in Listeria monocytogenes, a riboswitch that binds vitamin B12 controls expression of a noncoding regulatory RNA, Rli55. Rli55, in turn, controls expression of the eut genes, whose products enable ethanolamine utilization and require B12 as a cofactor. Defects in ethanolamine utilization, or in its regulation by Rli55, significantly attenuate Listeria virulence in mice. Rli55 functions by sequestering the two-component response regulator EutV by means of a EutV-binding site contained within the RNA. Thus, Rli55 is a riboswitch-regulated member of the small group of regulatory RNAs that function by sequestering a protein and reveals a distinctive mechanism of signal integration in bacterial gene regulation.

Ethanolamine is an abundant molecule in the vertebrate intestine (1, 2), and genes of the ethanolamine utilization pathway (eut) are widely conserved in pathogenic bacteria (35). This includes the Gram-positive intracellular human pathogen Listeria monocytogenes, in which eut expression has been shown to be up-regulated in the intestine during infection of mice (6), which suggests that ethanolamine is important for Listeria pathogenesis. In Enterococcus faecalis, eut expression is activated in response to ethanolamine by a two-component response regulator, EutVW (7, 8). In Salmonella enterica, ethanolamine utilization requires vitamin B12 as a cofactor (9), and we noted the presence of a B12-binding riboswitch located upstream of the first gene in the eut locus of L. monocytogenes (Fig. 1A) (10), which suggested that eut expression might also be regulated in response to B12 availability.

Fig. 1 Control of ethanolamine utilization (eut) genes by a B12 riboswitch–regulated ncRNA.

(A) The eut locus. Green arrow denotes B12 riboswitch. Vertical arrows denote positions of ANTAR elements (13) (B) RNA-seq coverage of eut operon. (C) Regions deleted in the Δrli55 and Δribo strains are indicated by solid black lines. Northern blot of the rli55 transcript in the (D) wild-type strain with probe indicated by thick dashed line in schematic or (E) wild-type (wt), Δribo, or Δrli55 strains with probe indicated by thin dashed line in schematic. Ethidium bromide staining of ribosomal RNA is shown as a loading control. Quantification of bands from three experiments is shown below each blot. Expression of the (F) eutV, eutC, and eutT genes was evaluated by qRT-PCR in the indicated strains. LB, Luria broth. Values represent means ± SEM, n = 3; *P < 0.05; ns, not significant.

To investigate a role for B12, we examined expression of the eut locus in response to B12 and ethanolamine by RNA sequencing (RNA-seq) (Fig. 1B) and quantitative reverse transcription polymerase chain reaction (qRT-PCR) (fig. S1). We observed expression of the eutVW genes under all conditions, albeit at low levels, which suggested that the cell maintains a pool of EutVW to sense and respond to ethanolamine. In contrast, higher-level expression of eutVW and expression of other eut genes require both B12 and ethanolamine (Fig. 1B). These data indicated B12 is required to activate eut expression and suggested that the B12 riboswitch does not prevent transcription of the eut locus in the presence of B12, as might be expected for a classical riboswitch (11).

To clarify whether the riboswitch has a role in the B12-dependent regulation of eut expression, we examined transcription of the riboswitch locus (Fig. 1C) in response to B12 and ethanolamine. We were unable to detect any long transcript, which might extend into the downstream eut locus. However, we did detect a ~450-nucleotide (nt) transcript, Rli55 (10, 12), in the absence of B12 (Fig. 1D), and a smaller (~200-nt) transcript that accumulated in the presence of B12 (Fig. 1D). This result suggested that the riboswitch mediates transcription termination of the ~450-nt Rli55 transcript in response to B12, and any remaining long transcript is rapidly degraded (fig. S2A). A strain with a deletion in the B12 riboswitch (Δribo) constitutively expressed the long Rli55 transcript under all conditions (Fig. 1E and fig. S2B), which confirmed B12-dependent regulation by the riboswitch. Conversely, neither the long or short Rli55 transcripts were detected in a strain (Δrli55) in which the entire rli55 locus was deleted (Fig. 1E and fig. S2B). Thus, the B12 riboswitch determines whether Rli55 is expressed as a long or short transcript in response to B12.

We reasoned that Rli55 might act as a regulatory RNA controlling expression of the eut locus. To test this hypothesis, we examined expression of eut genes by qRT-PCR in the wild-type, Δribo, and Δrli55 strains. In the wild-type strain, eut expression was elevated only when both ethanolamine and B12 were present in the media (Fig. 1F and fig. S3, A and B), whereas in the Δribo strain, which constitutively expresses the long Rli55, eut genes were never expressed under any conditions. In contrast, in the Δrli55 strain, expression of the eut genes was high in the presence of ethanolamine alone in addition to ethanolamine and B12 together. Together, these data support a model in which Rli55 prevents the expression of the eut locus in the absence of B12, ensuring that the eut genes are expressed only in the presence of both ethanolamine (substrate) and B12 (cofactor) (fig. S3C).

We tested whether defects in ethanolamine utilization, or its regulation, impacted L. monocytogenes virulence by examining the wild-type, Δribo, and Δrli55 strains in a mouse intravenous infection model. We also tested a mutant lacking the eutB gene encoding an ethanolamine lyase subunit, which is unable to catabolize ethanolamine. The Δribo and ΔeutB strains both had significantly reduced bacterial loads at 24 hours after infection compared with the wild-type strain, and these differences increased at 48 hours and 72 hours postinfection (~10- to 50-fold) (Fig. 2, A to F). In contrast, the Δrli55 strain, in which eut expression is not inhibited, was present in amounts comparable to those in the wild-type strain in the spleen and liver. Thus, defects in ethanolamine utilization or activation of eut expression significantly attenuate L. monocytogenes virulence.

Fig. 2 Ethanolamine utilization and pathogenesis.

BALB/c mice were injected intravenously with ~4500 colony-forming units (CFU) of the indicated strain. Mice were killed at (A) and (D) 24 hours, (B) and (E) 48 hours, or (C) and (F) 72 hours, and spleens and livers were removed to assess bacterial load per organ. Results represent two independent experiments with three or four mice per group in each experiment. *P < 0.05, ns, not significant.

In E. faecalis, ethanolamine is sensed by the sensor-kinase EutW, which subsequently phosphorylates the response regulator EutV (7). Phosphorylated EutV in turn binds ANTAR elements (AmiR and NasR transcriptional antiterminator regulator) in the 5′ untranslated regions of actively transcribed eut mRNAs, which prevents the formation of a transcription terminator and consequently activates eut expression (13). In L. monocytogenes, ANTAR sites were identified upstream of the eutA and eutV genes, and a third site was identified in the rli55 locus upstream of the eutG gene and downstream of the B12 riboswitch (13). We also identified a second ANTAR site in the rli55 locus (Fig. 3A), which suggested that one or both of these ANTAR elements could be transcribed as part of the 3′ end of Rli55 RNA. Indeed, our RNA-seq data showed that, in the absence of B12, high levels of Rli55 are transcribed as a long transcript encompassing the first ANTAR element (Fig. 3A). In contrast, rli55 transcription terminates abruptly after the riboswitch in the presence of B12, which indicates that the riboswitch determines whether Rli55 is transcribed with or without an ANTAR element. This notion was supported by the detection of putative Rli55 orthologs in E. faecalis and Streptococcus sanguinis (fig. S4). In the latter, the riboswitch, in conjunction with a single ANTAR element, has undergone an inversion relative to the adjacent eutG gene, which suggests that the riboswitch and the first ANTAR element are functionally linked (fig. S4B) and that this is also the case in L. monocytogenes.

Fig. 3 Rli55 sequesters the two-component response regulator EutV.

(A) RNA-seq coverage in reads per million (RPM) of the rli55 locus from bacteria grown as indicated. (B and C) qRT-PCR of eutV and eutA genes. Vitamin B12 (B12), ethanolamine (Ea). Values represent means ± SEM, n = 3. All differences were significantly different (*P < 0.05) from the Δrli55::Empty strain in the Ea condition, except where indicated (ns). All differences were not significantly different from the Δrli55::Empty strain in the B12 + Ea condition, except where indicated. (D) Secondary structure of the ANTAR elements encoded in the boxed region of Fig. 3A. Mutations are shown in blue. (E and F) RNA-seq coverage in RPM of the rli55 and eutV loci with RNA isolated by coimmunoprecipitation of cell lysates from either EutVFLAG or EutVNOFLAG cultures grown in the presence of ethanolamine (Ea) or ethanolamine and B12 together. (G) Proposed model of Rli55-mediated regulation of eut expression in the presence of ethanolamine (Ea) alone or ethanolamine + B12 (Ea + B12).

To test if the ANTAR elements are involved in Rli55-mediated regulation, we complemented the Δrli55 strain with chromosomally integrated rli55 alleles carrying mutations in the ANTAR elements and examined which ones restored Rli55-mediated regulation of eut expression (Fig. 3, B and C, and fig. S5). A strain with an empty construct (Δrli55::Empty) could not prevent expression of the eut genes in ethanolamine alone (as in the parental Δrli55 strain), whereas a strain with a wild-type copy of rli55rli55::rli55) fully restored Rli55-mediated inhibition. However, a strain with a deletion in the riboswitch (Δrli55::Δribo) inhibited eut expression in all conditions, as the riboswitch can no longer terminate rli55 transcription in response to B12. In strain Δrli55::rli55ΔM1, wherein four uridine residues in the first ANTAR site were mutated to adenines (Fig. 3D, ΔM1), inhibition of eut expression by Rli55 was abolished in the presence of ethanolamine alone. In contrast, in strain Δrli55::rli55ΔM1/M2, where compensatory mutations were made to the opposite side of the ANTAR stem-loop (Fig. 3D, ΔM2), wild-type regulation of eut expression was restored. Mutation of the six nucleotides in the stem-loop of the second ANTAR element (Fig. 3D, ΔM3, Δrli55::rli55ΔM3) had no significant effect on Rli55-mediated regulation. Thus, the first ANTAR element is necessary and sufficient for Rli55-mediated regulation.

The long form of Rli55 containing an ANTAR element might bind and sequester EutV and so prevent it from activating expression of the eut genes in the presence of ethanolamine but absence of B12. When sufficient levels of B12 accumulate, B12 would bind the riboswitch, producing truncated Rli55 transcripts, which would lack an ANTAR element and be unable to sequester EutV. To examine this hypothesis, we constructed a strain with an additional copy of the eutV gene carrying a 2XFLAG-tag (EutVFLAG) and first showed that expression of EutVFLAG protein is regulated identically to the native eutV gene in response to ethanolamine and B12 (fig. S6A). We also constructed a strain with an additional eutV gene lacking a FLAG tag (EutVNOFLAG). Anti-FLAG immunoprecipitations of cell lysates from these two strains (fig. S6, B and C), followed by RNA-seq analysis (Fig. 3, E and F), showed that Rli55 is enriched by coimmunoprecipitation with EutVFLAG primarily when bacteria are grown in the presence of ethanolamine alone, although we saw no enrichment in a parallel immunoprecipitation with the EutVNOFLAG strain (Fig. 3E). In contrast, the ANTAR element upstream of the eutV gene (Fig. 3F) is enriched by coimmunoprecipitation of lysates from EutVFLAG bacteria, but not EutVNOFLAG bacteria, grown in the presence of ethanolamine and B12 together but not from lysates of bacteria grown in ethanolamine alone. To a lesser extent, the ANTAR-containing region upstream of eutA and the entire eutA-Q locus are enriched under the latter condition (fig. S7). These data support a model in which the majority of EutV is bound and sequestered by Rli55 in the presence of ethanolamine alone. Conversely, in the presence of ethanolamine and B12, the riboswitch produces short truncated Rli55 transcripts, which cannot bind EutV, and so allows EutV to bind eut mRNAs and to activate eut expression (Fig. 3G).

This riboregulatory mechanism coordinates expression of the ethanolamine utilization (eut) locus with the availability of B12, the essential cofactor for ethanolamine catabolism. Previously, ethanolamine utilization has been shown to be important after oral infection by Salmonella enterica serovar Typhimurium and enterohemorrhagic Escherichia coli (3, 4, 14); however, the contribution of ethanolamine utilization to L. monocytogenes pathogenesis in an intravenous mouse infection model suggests that ethanolamine utilization is important outside of the intestine and possibly in the intracellular environment. This study also extends the role of riboswitches in the regulation of noncoding RNAs (15, 16). Finally, our data show that Rli55 represents a new member of the small family of regulatory RNAs that function by sequestering a protein, which also includes the 6S and CsrB/C RNAs (17), and highlights a distinctive means of signal integration in bacterial gene regulation.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Tables S1 and S2


References and Notes

  1. Acknowledgments: We are grateful to D. Garsin and W. Winkler for their exchange of ideas and discussion of unpublished data. We also thank, N. King and J. Pederson for helpful and insightful discussions and Y. Chao and J. Vogel for assistance with the CLIP-seq protocol. This work was supported by grants to P.C., European Research Council advanced grant (233348), ANR (BACNET 09-BLAN-0024-02), ANR Investissement d’Avenir Programme (10-LABX-62), Fondation Le Roch, and Fondation Jeantet. P.C. is an International Senior Research Scholar of the Howard Hughes Medical Institute.
View Abstract

Navigate This Article