Evidence for Widespread Gene Control Function by the ydaO Riboswitch Candidate

Abstract

Nearly all representatives of experimentally validated riboswitch classes in bacteria control the expression of genes for the transport or synthesis of key metabolic compounds. Recent findings have revealed that some riboswitches also regulate genes involved in physiological changes, virulence, and stress responses. Many novel RNA motifs are being identified by using bioinformatics algorithms that search for conserved sequence and structural features located in intergenic regions. Some of these RNAs are likely to function as riboswitches for metabolites or signaling compounds, and confirmation of this function would reveal the basis of the genetic control of new regulons. Herein we describe the analysis of the ydaO riboswitch candidate, which represents one of the most widespread candidates remaining to be validated. These RNAs are common in Gram-positive bacteria, and their genomic associations with diverse genes suggest that they sense a compound that signals broader physiological changes. We determined that the ydaO motif exhibits sequence- and structure-dependent gene control, and reporter assays indicate that its natural ligand is present even when cells are grown in defined media. A transposon-mediated knockout screen resulted in mutants with a dysregulated expression of genes controlled by the RNA motif. The mutations disrupt genes that drastically modulate energy-generating pathways, suggesting that the intracellular concentration of the ligand sensed by the ydaO motif is altered under these stress conditions.

RNA structures termed riboswitches are used in all three domains of life to regulate gene expression through the direct binding of small-molecule metabolites by RNA (3, 7). Riboswitches are composed of an aptamer domain, which is responsible for metabolite binding, and an expression platform, which reads out ligand-induced structural changes to regulate gene expression (27). In eubacteria, where the vast majority of examples are observed, riboswitches are almost exclusively found in 5′ untranslated regions (UTRs) of mRNAs. Furthermore, most riboswitches use either an intrinsic transcription terminator stem or an anti-Shine-Dalgarno (SD) stem to regulate the transcription or translation, respectively, of the downstream open reading frame (ORF).

Although aptamer domains are strikingly well conserved within a functional class even among distantly related organisms, expression platform features can vary widely, and the most common regulatory structures are not always used. For example, riboswitches that sense glucosamine-6-phosphate (or glmS ribozymes) use ligand binding to promote regulation by mRNA self-cleavage (42). In addition, a class I S-adenosylmethionine (SAM)-sensing riboswitch (SAM-I) in Clostridium acetobutylicum controls the transcription of an antisense transcript to regulate the ubiG operon (2). Although in this example, the direct result of the riboswitch is the control of antisense transcription, the mechanism of gene regulation occurs via transcriptional interference from RNA polymerase proceeding on opposite strands, as it was observed previously that the antisense RNA could not regulate gene expression when provided in trans (2). More recently, a SAM-I riboswitch was identified in Listeria monocytogenes, which, in addition to standard transcription termination-based regulation, acts as a small noncoding RNA to regulate PrfA expression in trans (18). Although additional diversity in aptamer and expression platform features undoubtedly exists, newfound structured RNA motifs are attractive riboswitch candidates if they exhibit genomic locations and structural features common to known riboswitches.

Riboswitch aptamers remain highly conserved through evolution because each one must preserve a selective binding pocket for its target metabolite. Structural analyses of riboswitch aptamers have revealed the mechanisms by which riboswitches display such high selectivity for their ligands (28, 29). For example, structural probing of the guanine-sensing riboswitch with various guanine analogs revealed that much of the guanine molecule participates in hydrogen bonding (19). Indeed, atomic-resolution structure models of this aptamer confirm that the ligand makes extensive contacts with highly conserved core nucleotides of the aptamer (5, 30). Often, the highly conserved nucleotides that are characteristic of riboswitch aptamers can be exploited by bioinformatics searches to identify new candidate riboswitches (4, 6). More recently, bioinformatics pipelines that also take advantage of the covariation inherent in the conservation of secondary structures to identify new candidate riboswitches have been developed (4, 6, 38, 40).

The identification of a motif through bioinformatics analysis does not always provide sufficient evidence to conclude that the RNA controls gene expression or to reveal possible ligands for candidate riboswitches. While the genetic context of some candidate riboswitches often implies a ligand, such as the genes controlled by the putative molybdenum cofactor (MoCo) riboswitch (26) and the cyclic di-GMP (c-di-GMP) riboswitch (34), other genetic contexts may include a large number of uncharacterized genes or too diverse an array of genes to infer a single ligand. For these reasons, a closer analysis of these candidate RNAs is necessary. The ydaO (yuaA) motif (4) is a typical example of a structured RNA, revealed by bioinformatics analysis, that has all of the hallmarks of riboswitch function. However, the available information on genetic distribution does not readily reveal its likely function. In the current study, we further investigate the characteristics of the ydaO motif and discuss the implications of our data for the RNA's role in regulating cellular processes.

MATERIALS AND METHODS

Chemicals and oligonucleotide synthesis.

Sodium succinate and sodium fumarate were purchased from Sigma. Oligonucleotides for cloning were synthesized by the W. M. Keck Foundation Biotechnology Resource Center, Yale University. Oligonucleotides for quantitative PCR were designed by using the Primer Express program (Applied Biosystems) and synthesized by Sigma-Genosys.

Plasmids and reporter constructs.

Transcriptional fusion plasmids pDG1661 and pDG1728, in addition to mini-Tn10-containing plasmid pHV1249, were obtained from the Bacillus Genome Stock Center (Ohio State University). The dual-reporter plasmid was generated by PCR amplification of the Kan^r gene from plasmid pMK3 using primers ydaO6 (5′-GGGAAAAGGTGGTGAACTACTATGAGAATAGTGAATGGACCAATAAT-3′) and ydaO7 (5′-CCAGGATCCTCAAAATGGTATGCGTTTTGAC-3′). A subsequent amplification with primers ydaO5 (5′-GGCAAGCTTTTGATACACTAATGCTTTTATATAGGGAAAAGGTGGTGAACTACT-3′) and ydaO7 appended the ribosome binding site (RBS) of Bacillus subtilis spoVG to the Kan^r gene. The gene was then inserted as a HindIII/BamHI fragment into the pDG1728 reporter plasmid to yield pDG1728-spokan.

To generate the wild-type ydaO reporter, nucleotides −426 to +30 relative to the B. subtilis ydaO ORF were PCR amplified as an EcoRI/BamHI fragment by using primers ydaO3 (5′-AACAGAATTCGATTTTAGCCTCTGTTTTTTTATTTTTGGTAAGTAAA-3′) and ydaO4 (5′-TTTGGATCCAATCAAAAAACGTTTGATTGAATGATACAT-3′) and subsequently ligated into pDG1661. Mutations of the aptamer were made by using the QuikChange XL mutagenesis kit (Stratagene) and the appropriate primers (see Table S2 in the supplemental material). Additionally, wild-type and M3 ydaO fragments were amplified from the appropriate pDG1661 constructs by using primers ydaO3 and ydaO16 (5′-CGGTAAGCTTAATCAAAAAACGTTTGATTGAATGATACAT-3′) to generate EcoRI/HindIII fragments for ligation into pDG1728-spokan. All constructs were confirmed by DNA sequencing (W. M. Keck Foundation Biotechnology Resource Laboratory, Yale University). Constructs were transformed into B. subtilis 1A1 (Bacillus Genome Stock Center, Columbus, OH) according to previously described methods (43).

Reporter assays.

B. subtilis cells were inoculated from a tryptose blood agar base (TBAB) plate and grown overnight in 2×YT medium (16 g/liter tryptone, 10 g/liter yeast extract, 5 g/liter NaCl) with 5 μg/ml chloramphenicol. Cells were then diluted 1:20 in either 2×YT or modified glucose minimal medium (MGMM) (43), and β-galactosidase activity was measured as previously described (19).

Transposon mutagenesis.

B. subtilis cultures containing the wild-type ydaO motif controlling the dual reporter were transformed with pHV1249 as previously described (12). Transformants were grown at 28°C overnight and selected with 5 μg/ml chloramphenicol. Colonies were inoculated into LB containing both 5 μg/ml chloramphenicol and 0.5 μg/ml erythromycin and grown overnight at 28°C. Cultures were diluted 1:100 into 2 ml LB containing 5 μg/ml chloramphenicol and incubated 2.5 h at 28°C. A total of 1.5 ml of culture was added to 15 ml of LB containing 5 μg/ml chloramphenicol and grown to an optical density at 600 nm (OD₆₀₀) of ∼0.1. The culture was transferred to 50°C and incubated for 2 h. One milliliter of the culture was plated onto TBAB containing 10 μg/ml kanamycin, 3 μg/ml chloramphenicol, and X-gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside). Colonies were selected after 2 days to allow the blue color of LacZ⁺ colonies to develop.

To identify mutants for further analysis, colonies were cultured in LB with 100 μg/ml spectinomycin. Two milliliters of cultures grown overnight was pelleted and resuspended in freshly made 3 mg/ml lysozyme in Tris-EDTA (TE) buffer. Suspensions were incubated for 5 min at 23°C, followed by three freeze-thaw cycles at −80°C. Total RNA was isolated from samples with Trizol (Invitrogen) extraction. Total RNA was incubated with RQ1 DNase (Stratagene) for 60 min at 37°C, followed by heat inactivation of the DNase. Five micrograms of RNA was used for reverse transcription using the SSII RT kit (Invitrogen). Approximately 30 ng of input RNA was used for quantitative PCR using PowerSYBR green master mix (Applied Biosystems). Transcripts analyzed were 16S, lacZ, the ydaO ORF, and the ktrA ORF. The following cycle conditions were used: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 s and 60°C for 1 min. The C_T (cycle threshold) values were calculated with ABI 7500 SDS software with automatic baseline and threshold determinations. RNA levels were normalized to 16S levels and finally compared to wild-type levels to yield relative transcript abundances.

In-line probing.

The 165 ydaO and 173 ktrA constructs (numbers reflect RNA length) were amplified from B. subtilis genomic DNA. Primers ydaO1 (5′-TAATACGACTCACTATAGGGAAAACAAATCGCTTAATC-3′) and ydaO2 (5′-CATCGAACACAGGTTTCATTC-3′) were used to amplify the 165 ydaO construct. Primers ydaO25 (5′-TAATACGACTCACTATAGGTTCAATTTCATACAGTCTCTTTTACCGC-3′) and ydaO26 (5′-GATAATAGACCTCCTCTCTGTATACGCC-3′) were used to generate the 173 ktrA construct. The 129 kamA construct was amplified from Moorella thermoacetica genomic DNA (gift of Scott Ragsdale) by using primers ydaO27 (5′-TAATACGACTCACTATAGGGAATAAGCGAGCGCTGAATCCC-3′) and ydaO28 (5′-CACCTCACCTTTCTACGCTTACG-3′). The 152 trkA construct was amplified from Bacillus halodurans genomic DNA (ATCC) by using primers ydaO29 (5′-TAATACGACTCACTATAGGGCACCTTTAATCGCTGAATCTCTATTG-3′) and ydaO30 (5′-CCTCTCTCGAATCAACGCTTAC-3′). All constructs were TOPO cloned and sequence confirmed before subsequent transcription. The PCR template (30 pmol) was transcribed with a solution containing 80 mM HEPES (pH 7.5 at 23°C), 24 mM MgCl₂, 2 mM spermidine, and 40 mM dithiothreitol (DTT) with T7 RNA polymerase. Transcription products were subjected to denaturing (8 M urea) polyacrylamide gel electrophoresis (PAGE) for purification. RNA was eluted from the gel in a solution containing 200 mM NaCl, 10 mM Tris-HCl (pH 7.5 at 23°C), and 1 mM EDTA (pH 8.0 at 23°C) and precipitated with ethanol. After PAGE purification, 10 pmol RNA was used for subsequent dephosphorylation by using Roche phosphatase, and the RNA was subsequently 5′ ³²P labeled by using T4 polynucleotide kinase (New England Biolabs) and [γ-³²P]ATP. The labeled RNA was purified by using denaturing 6% PAGE and resuspended in 20 μl TE buffer for storage. In-line probing reactions were carried out as previously described (25, 32).

Metabolomics analysis.

The wild-type reporter strain and mutants F and H were revived from storage at −80°C on TBAB with 100 μg/ml spectinomycin overnight at 37°C. Single colonies were inoculated overnight in 2×YT containing 100 μg/ml spectinomycin. The following day, cultures were diluted to an OD₆₀₀ of 0.05 and grown to an OD₆₀₀ of 1.5. One hundred sixty-seven milliliters of culture, or 2.5 × 10¹¹ cells, was pelleted and washed twice with phosphate-buffered saline (American Bioanalytical). Pellets were snap-frozen in liquid nitrogen and submitted to Metabolon (Research Triangle Park, NC) for metabolomics analysis as previously described (16).

RESULTS AND DISCUSSION

Revised consensus model from expanded ydaO motif hits.

The ydaO motif was first identified through BLAST searches of the intergenic regions of 91 sequenced microbial genomes (4). This original search revealed only 15 examples and therefore provided little evidence to construct a consensus sequence and structural model or to evaluate possible riboswitch ligands. By examining the recent explosion of microbial and metagenomic sequence data being made available, we identified many additional ydaO representatives. Thus far, 580 unique examples of ydaO motifs have been found, representing almost a 40-fold expansion of the collection of previously published ydaO motifs. Additionally, newly developed bioinformatics pipelines that take advantage of the covariation inherent in the conservation of base-paired substructures of riboswitches (44) allow us to refine further the consensus structure (Fig. 1A). Extensive evidence of covariation throughout the stems of the motif supports a revised global architecture for the RNA consisting of two three-stem junctions, rather than the originally proposed four-stem junction. In-line probing (25, 32) of the RNA indicates that much of the structure is preorganized before it senses its putative cognate ligand (Fig. 1B and C), as is often observed with other riboswitches (e.g., see reference 19).

FIG. 1.

Consensus sequence and structure of the ydaO motif. (A) A phylogenetic analysis of 580 unique examples of the ydaO motif was conducted to determine the consensus sequence and secondary structure of the motif. In some examples for which no transcription terminator is found, the pseudoknot formed between the loop of P7 and the 3′ end of the molecule encompasses potential Shine-Dalgarno sequences. (B) In-line probing of the B. subtilis 165 ydaO RNA. NR, T1, and ⁻OH indicate RNAs treated with no reaction, partial digestion with RNase T1, and partial digestion with alkali, respectively. Pre indicates precursor RNAs. Select bands corresponding to RNase T1 cleavage (3′ of G residues) are labeled. (C) Sequence and structure of the 165 ydaO RNA depicting locations of high spontaneous cleavage revealed by in-line probing data in B. Nucleotides in gray are not a part of the consensus ydaO motif. Two guanosine residues, indicated with a lowercase g, were added to the 5′ end of the construct to facilitate transcription by T7 RNA polymerase and are not included in the numbering of the construct. Arrowheads bracket the nucleotide linkages that could be assessed by the data in B.

The ydaO motif segregates predominantly into Gram-positive bacteria. A closer inspection of the genes associated with this RNA domain reveals a myriad of genes related to cell wall metabolism as well as the transport and biosynthesis of known osmoprotectants (Fig. 2; for a complete list of genes, see Table S1 in the supplemental material). Almost half of the representatives are associated with genes responsible for the degradation of polysaccharides and peptide linkers that compose the peptidoglycan matrix (reviewed in reference 37). Additional hits are associated with transporters implicated in the cell's response to osmotic shock (14, 31), including the transport of potassium ions and compatible solutes. Notably, the majority of ydaO motifs associated with acetyltransferases are observed in the putative 3′ UTRs of these genes. These RNAs may utilize a novel expression platform architecture with a yet-unknown mechanism of regulation.

FIG. 2.

Genetic contexts of ydaO motifs. Genes associated with the ydaO motif were grouped into general categories as noted. For a complete list of associated genes, see Table S1 in the supplemental material.

The presence of a ydaO motif directly preceding the ktrAB operon in Bacillus subtilis, combined with evidence of impaired osmoadaption in ktrAB knockouts (11), led us to hypothesize that the RNA may sense a compound related to osmoregulation. However, transcriptome analysis of osmotically shocked B. subtilis did not reveal an increased level of expression of either the ydaO gene or the ktrAB operon (33). Curiously, the transcription of the ydaO and ktrA genes increases in the first 10 min following spore germination (13). A recent characterization of cell wall hydrolases downstream of ydaO motifs in Streptomyces coelicolor indicated that these genes are expressed throughout development (10). Taken together with the large fraction of ydaO motifs preceding genes for cell wall metabolism, we propose that the ydaO motif may function in crucial cellular processes requiring complex rearrangements of the cell wall, such as, but not limited to, osmoadaption or spore germination.

The ydaO motif is a gene control element.

Examination of nucleotides proximal to the ydaO motif in B. subtilis revealed a transcription terminator immediately downstream of the motif but not overlapping the aptamer. This architecture suggests that the ydaO motif functions in B. subtilis via transcription termination control. In contrast to B. subtilis, examples of the ydaO motif in S. coelicolor do not appear to regulate transcription, as transcriptional profiles of the seven cell wall hydrolases under the control of the ydaO motif display different developmental patterns (10). Although one cannot rule out promoter-driven differences in transcript abundance, a plausible explanation is that the ydaO motif regulates at the level of translation initiation in S. coelicolor. A closer examination of the various ydaO motifs in S. coelicolor reveals an overlap of the pseudoknot with the ribosomal binding site of these mRNAs (data not shown), and ydaO S. coelicolor may therefore utilize a translation regulation mechanism similar to that implicated for the class IV SAM riboswitch in this same organism (39).

To confirm that the ydaO motif indeed regulates gene expression, the 5′ UTR of the B. subtilis ydaO gene was incorporated into a transcriptional fusion construct with the lacZ gene. Additionally, mutations were made to the motif to determine if the proposed secondary structures are required for regulation. Specifically, mutants M1 and M3 disrupt pairing within the P4a and P6 stems, respectively, while mutants M2 and M4 restore pairing with compensatory mutations (Fig. 3A). Mutant M1 displays only a modest disruption of wild-type reporter expression levels (Fig. 3B), most likely attributed to the fact that the P4a stem is distal to the conserved core, and its lack of importance in RNA function could explain why this substructure is observed for only a fraction of the ydaO motifs. Mutant M2, in contrast, actually appears to depress further the already low levels of expression observed for the wild-type reporter strain. The nucleotide changes may adversely affect various aspects of RNA switch function such as proper RNA folding, ligand binding kinetics, or mRNA stability to reduce expression compared to wild-type RNA. Mutants M3 and M4 display more robust disruption and restoration of wild-type expression levels, as would be anticipated if this architecture is important for gene control (Fig. 3B).

FIG. 3.

The ydaO motif is a structure-dependent gene control element. (A) Mutations made to the B. subtilis ydaO motif depicted in Fig. 1C are indicated on the secondary structure. (B) Plot of the levels of reporter gene expression for constructs incorporating the wild-type (WT) B. subtilis ydaO motif or various mutants as depicted in A. Data were generated with rich (2×YT) and minimal (MGMM) media.

Additional mutations were made to the terminal three nucleotides of stems P3 and P7. A disruption of either stem results in the largest increases in reporter expression levels observed (data not shown). However, compensatory mutations do not restore wild-type levels of expression, suggesting that the nucleotide identities of these base pairs are also critical for regulatory function, as implied by their sequence conservation. Combined with the low expression level of the wild-type reporter, mutation analyses suggest that this example of the ydaO motif functions as a genetic “off” switch, in which the formation of the ligand-RNA complex represses expression.

It seems likely that the ligand is either abundant in rich medium or produced internally, because reporter gene expression controlled by the wild-type ydaO motif is low. To distinguish between these possibilities, the reporter constructs were additionally tested with a defined medium. Many riboswitch-reporter fusions, such as those responding to guanine (19) and SAM (43), are expressed at high levels in defined medium, as the corresponding ligand is no longer abundant externally. Surprisingly, the wild-type ydaO motif construct displays slightly reduced reporter expression levels in a defined medium (Fig. 3B). Mutant constructs exhibit similar trends with both rich and defined media (Fig. 3B). Specifically, the expression of the M1 reporter is within 2-fold of wild-type levels under both conditions, while the M2 construct restores near-wild-type levels of expression. Similarly, mutant M3 displays increases in levels of reporter gene expression of about 5-fold in both rich and defined media, with mutant M4 restoring wild-type levels. In total, the mutation analysis with defined medium suggests that if the ydaO motif recognizes a small-molecule metabolite, that metabolite is biosynthesized under both circumstances. This feature has not yet been observed for characterized riboswitches.

Genetic screen for variants with impaired ydaO-mediated regulation.

If the ligand is indeed available intracellularly under normal growth conditions, we hypothesized that mutations of the B. subtilis genome that reduce internal levels of ligand by disrupting its synthesis could be found. By incorporating a second selectable marker, kanamycin resistance, into the ydaO-lacZ reporter fusion described above as a bicistronic message, we established a dual selection/screen method for obtaining potential mutant strains. Specifically, we subjected the dual-reporter strain to mini-Tn10 transposon mutagenesis (24). The temperature-sensitive origin of replication on the transposon-containing plasmid combined with the location of the transposase outside the transposable element allow single insertions into the chromosome. Additionally, the insertion of the chloramphenicol resistance cassette in the mini-Tn10 transposon allows subsequent mapping of the transposition event.

Following mutagenesis, we selected for mutations resulting in kanamycin resistance and screened for LacZ expression. For each candidate mutant, the ydaO motif was sequenced to discount aptamer mutations as the source of reporter dysregulation. However, because the reporters are expressed as a bicistronic message, there was a possibility that a mutation occurred that resulted in higher levels of expression from that locus irrespective of the ydaO motif. To ensure that we indeed isolated chromosomal mutations that globally affect the regulatory function of the ydaO motif, we subjected mutant strains to quantitative reverse transcription and PCR (qRT-PCR) by using the two native loci (the ydaO and ktrAB mRNAs) regulated by ydaO motifs as internal controls. Increased mRNA levels at all three loci would indicate a mutation of interest. Although many transposon mutants displayed a deregulation of reporter activity without additional effects on the native loci (Fig. 4A), mutants F and H showed increased levels of expression at all three loci.

FIG. 4.

Chromosomal mutations disrupt regulation by the ydaO motif. (A) Transposon mutant strains were analyzed by qRT-PCR to examine the transcript abundance of the dual reporter and two native loci regulated by the ydaO motif (ydaO and ktrA genes). The transcript abundance relative to that of a wild-type reporter strain (value of 1) is depicted. (B) The wild type and mutant H were subsequently grown in the presence of succinate and fumarate and analyzed for transcript abundance relative to that of the same strain without supplementation.

Mutants F and H were gene-disrupting insertions subsequently mapped to the ndh and menH genes, respectively. The product of ndh is the major NADH dehydrogenase in B. subtilis (9), while MenH directs the final methylation step in the production of menaquinone (15). While these are seemingly disparate pathways, the control of NADH levels in the cell and the use of menaquinone functionally overlap in both the oxidative phosphorylation and citric acid cycle pathways. Specifically, menaquinone is required for the oxidation of succinate to fumarate (17), a step in the citric acid cycle pathway that also participates in the electron transport chain. The oxidation of NADH is also required for electron transport, and elevated intracellular levels of NADH, a direct result of knocking out ndh (9), can downregulate the citric acid cycle. Both mutations affect energy flux and would therefore explain the reduced cell growth observed for both mutant strains compared to the wild-type reporter strains.

Because the menH::Tn10 mutation potentially represents a block in a specific step in the citric acid cycle, we hypothesized that bypassing the mutation via supplementation with fumarate may restore regulation by the ydaO motif. Indeed, as observed by qRT-PCR, supplementation with fumarate, but not succinate, produces a modest restoration of repression at all three loci under the control of the ydaO motif (Fig. 4B), suggesting that this mutation is responsible for the dysregulation of the ydaO motif. However, in-line probing of a construct consisting of the ydaO motif preceding the B. subtilis ktrAB operon (173 ktrA) with fumarate and other compounds from the citric acid cycle, as well as the menaquinone precursor menadione, did not reveal evidence of ligand binding (for a complete list of compounds tested, see Table S2 in the supplemental material). Even these results cannot be used to rule out these compounds as possible riboswitch ligands. For example, we cannot be sure that an RNA construct chosen for in-line probing assays is not missing nucleotides or may have additional nucleotides that prevent the proper function of the putative aptamer. To avoid a premature dismissal of potential ligands, many compounds were subjected to in-line probing analysis with several ydaO motifs from different species (see below).

A partial survey of the levels of metabolites present in wild-type B. subtilis cells compared to those in the F and H mutants was conducted (see Materials and Methods for details). This analysis revealed that both mutant strains have a number of key differences in metabolite concentrations compared to the wild type (see Table S3 in the supplemental material), which is consistent with the crucial roles performed by the disrupted genes. Given our hypothesis regarding the default presence of a putative riboswitch ligand, we were specifically interested in metabolites observed in lower abundances in both mutant strains. However, it is important to note that this metabolomics analysis is far from complete. For example, many fundamental metabolites, such as SAM, thiamine pyrophosphate, adenosylcobalamin, and oxoglutarate, were not identified in any strain. This does not mean that SAM or these other compounds are not present in these cells but likely reflects the fact that assigning compound identities to mass spectrum hits is incomplete even for long-known metabolites. Indeed, numerous mass spectrum hits, some with deficiencies in both mutant strains, remain unassigned in the data set. The problem of unassigned mass spectrum hits of course extends to compounds that may remain to be described for this bacterium. The latter issue highlights the possibility that the ydaO motif may bind a natural compound that has not been described in the literature to date.

Unfortunately, a great number of metabolic changes were expected for mutants F and H, given the fundamental roles that these disrupted pathways play in the cell. Some metabolites deficient in both mutant strains fit well with our initial hypothesis that the ydaO motif senses some aspect of the osmoadaptive response, including betaine and glutamate. These compounds function as the counter-ion to potassium that initially floods osmotically shocked cells. Not surprisingly, NAD⁺ was observed in lower abundances in the ndh::Tn10 strain, but because of a similar decrease in abundance in the menH::Tn10 strain, we were also interested in testing this compound as a possible ligand. Similarly, because our fumarate supplementation results implicate the citric acid cycle in maintaining genetic regulation by the ydaO motif, we were interested in changes in the abundances of citric acid cycle intermediates. Interestingly, the level of alanyl-alanine, a major component of the peptidoglycan (reviewed in reference 37), was also decreased in both mutants. Because of the large proportion of ydaO motifs associated with cell wall-remodeling genes, this compound was of particular interest as a potential ligand. However, none of these candidate metabolites induced structural changes in ydaO motifs from Bacillus halodurans (152 trkA) or Moorella thermoacetica (129 kamA) using in-line probing (data not shown).

Several lines of evidence support the hypothesis that ydaO motifs function as riboswitch aptamers. As described above, sequence and structural perturbations of a ydaO motif cause a dysregulation of gene expression. Mutation-compensation analysis indicates that the structure of the motif is critical for regulatory function, a finding that is in agreement with the natural covariation observed for the consensus sequence. While the putative aptamer domain displays a high level of conservation, evidence exists for multiple forms of expression platforms, which is another characteristic of known riboswitches. Analysis of the B. subtilis ydaO motif reveals a likely transcription terminator, while other examples of the ydaO motif appear to use ribosome binding site occlusion as the mechanism of regulation (10). However, the presence of ydaO motifs in the 3′ UTRs of some acetyltransferase genes may represent a novel form of regulation by riboswitches.

To date, the ydaO motif is the second most widely spread candidate riboswitch for which no ligand is determined (1). Recent evidence suggests a link between the most common candidate riboswitch, the yybP motif, and pH sensing in Gram-negative bacteria (23). Accordingly, the ydaO motif may soon become the most widespread riboswitch candidate lacking direct evidence of ligand-RNA interactions. It seems unlikely that the RNA serves solely as a binding site for a protein factor that functions as the direct sensor. A protein factor would relieve the demand for the RNA to retain sequence and structural features required to bind a metabolite. Moreover, RNA structures that are bound by metabolite-sensing protein factors tend to be less well conserved because both RNA and protein partners can coadapt through evolution.

Although no ligand has been determined for the ydaO motif, our inspection of the RNA highlights key aspects of the regulatory function of the RNA. The genes identified through transposon mutagenesis, menH and ndh, may not directly produce the ligand. Rather, it appears that the dysregulation of the ydaO motif observed for mutant strains is a secondary effect of the drastic nature of the mutations, as highlighted by the number of changes to the metabolome of these mutants. In other words, these mutants cause such widespread changes in metabolism that they offer little assistance in pinpointing possible ligands.

Many riboswitches bind either cofactors used in metabolic reactions, such as thiamine pyrophosphate (41) and adenosylcobalamin (22), or fundamental building blocks of the cell, as in the purine-sensing (19, 20) and amino acid-sensing (8, 21, 35) riboswitches. However, the recent discovery of a c-di-GMP riboswitch (34) perhaps mirrors the potential type of ligand sensed by the ydaO motif. Like the ydaO motif, the c-di-GMP riboswitch controls a number of genes involved in cellular processes, such as motility and virulence, although the specific genetic contexts of the c-di-GMP riboswitches segregate between disparate phyla. In contrast, the ydaO motif appears to regulate similar genes across evolutionarily diverse organisms.

Another possibility, based on the regulatory profile of the motif, is that the RNA utilizes a metal ion or a compound from an unrelated metabolic pathway. For example, the glycine riboswitch characterized for “Candidatus Pelagibacter ubique” (36) reads glycine levels to control energy flux through the citric acid cycle by regulating malate synthase expression. Thus, regulation by the ydaO motif may represent a novel connection between a known metabolite and the process of cell wall remodeling. Resolving this mystery should shed light on the interconnections between fundamental metabolic processes and cell integrity across a great diversity of bacteria.