Tandem Riboswitch Architectures Exhibit Complex Gene Control Functions

See allHide authors and affiliations

Science  13 Oct 2006:
Vol. 314, Issue 5797, pp. 300-304
DOI: 10.1126/science.1130716


Riboswitches are structured RNAs typically located in the 5′ untranslated regions of bacterial mRNAs that bind metabolites and control gene expression. Most riboswitches sense one metabolite and function as simple genetic switches. However, we found that the 5′ region of the Bacillus clausii metE messenger RNA includes two riboswitches that respond to S-adenosylmethionine and coenzyme B12. This tandem arrangement yields a composite gene control system that functions as a two-input Boolean NOR logic gate. These findings and the discovery of additional tandem riboswitch architectures reveal how simple RNA elements can be assembled to make sophisticated genetic decisions without involving protein factors.

Many riboswitches are composed of two functional domains, an aptamer and an expression platform, that work in concert to selectively bind a target ligand and modulate gene expression (13). In nearly all examples identified in bacteria, binding of a single metabolite by the aptamer controls transcription termination or translation initiation via structural changes in the expression platform. Some riboswitches balance the kinetics of metabolite binding and RNA folding with the speed of RNA transcription to control gene expression appropriately (46). However, most metabolite-sensing RNAs are generally believed to be unsophisticated in structure and action compared with the diversity of gene control factors made of protein (7).

Some riboswitches are known to control gene expression by using complex structures and mechanisms. For example, glmS genes found in Bacillus subtilis and related bacteria are controlled by metabolite-responsive self-cleaving ribozymes (8, 9). glmS ribozymes appear to use their target metabolite, glucosamine-6-phosphate, as a cofactor to accelerate mRNA cleavage (10, 11). Another complex riboswitch class uses two aptamers in tandem to bind two glycine molecules cooperatively (12). Both glycine aptamers control gene expression by influencing the folding of a single intrinsic transcription terminator stem (13, 14). This tandem aptamer assembly functions with far greater responsiveness to changing concentrations of ligand than do simple riboswitches.

Using filtered covariance model searches (15, 16), we identified additional bacterial mRNAs that carry riboswitches or their substructures in tandem. The existence of these modular architectures implies that some bacteria are using RNAs to address more demanding gene control requirements. Of particular interest is the 5′ region of the metE gene from Bacillus clausii, which carries sequences and structures that conform to two known riboswitch classes that bind S-adenosylmethionine (SAM) (1719) and coenzyme B12 (adenosylcobalamin or AdoCbl) (20, 21) (Fig. 1A). This tandem riboswitch arrangement was detected in six different isolates of B. clausii with only modest nucleotide sequence differences [fig. S1, Supporting Online Material (SOM)], which indicates that the tandem SAM-AdoCbl riboswitch arrangement is important for this bacterial species.

Fig. 1.

Riboswitches reside adjacent to B. clausii genes encoding enzymes for the biosynthesis of methionine and SAM. (A) The metE, metH, and metK mRNAs carry riboswitches that are predicted to repress gene expression by transcription termination caused by formation of a stable RNA hairpin (light blue) followed by a run of six or more uridine nucleotides (Un) (13, 14). (B) Two different methionine synthase enzymes (MetE and MetH) independently convert homocysteine to methionine, and SAM synthase (MetK) converts methionine to SAM (22, 24, 25).

A rationale for the presence of two distinct riboswitches in the metE mRNA was established by examining the metabolic pathway for methionine biosynthesis (Fig. 1B and SOM). B. clausii can express two proteins (MetE and MetH) that independently catalyze the formation of methionine from homocysteine, as well as another protein (MetK) that converts methionine to SAM (Fig. 1B) (22). The mRNAs for all three proteins carry SAM riboswitches, which are extensively used by Bacillus and Clostridium bacteria to repress this pathway when SAM is plentiful (1719, 23). In contrast, only MetE expression should be reduced when AdoCbl is abundant (24, 25), because cells can more efficiently produce methionine and SAM by expressing MetH, which requires the coenzyme methylcobalamin (MeCbl, a derivative of AdoCbl) (2426).

The architecture of the B. clausii tandem SAM-AdoCbl riboswitch (Fig. 2A) suggests that the two ligands can independently repress metE. Each aptamer resides immediately upstream of a structure that resembles an intrinsic transcription terminator (13, 14), and nucleotides known to be essential for ligand binding in both aptamers overlap with potential anti-terminator stems. Thus, the metE tandem riboswitch arrangement is expected to function as a natural Boolean NOR logic gate (27), wherein either of two chemical inputs yields an output of gene repression.

Fig. 2.

The tandem SAM and AdoCbl riboswitches from the B. clausii metE mRNA independently bind their target metabolites. (A) Sequence and secondary structure models for riboswitches in the 5′-UTR of the B. clausii NRRL B-23342 metE mRNA. Nucleotides are numbered relative to the predicted natural transcription start site. The G residues at the 5′ terminus were included to aid efficiency of in vitro transcription by T7 RNA polymerase. Nucleotides not included in the 405 metE construct are boxed in gray. (B) In-line probing analysis (28) of 5′ 32P-labeled 405 metE RNA. NR, T1, and OH identify rows containing nonreacted RNA, or RNA subjected to partial digest with ribonuclease T1 (G-specific cleavage) or alkali, respectively. Remaining lanes were incubated in the absence (–) or presence (+) of 10 μM SAM or 100 μM AdoCbl as indicated. Numbered arrows identify products of spontaneous RNA cleavage (analyzed in figs. S2 and S3) whose levels are altered in response to ligand-induced structural modulation.

NOR logic function does not require that the two riboswitches influence each other. To confirm that the tandem riboswitches function independently, we examined the metabolite-binding characteristics of a 405-nucleotide construct (405 metE) derived from the 5′ untranslated region (UTR) of the B. clausii metE mRNA (Fig. 2). 405 metE was subjected to inline probing (28), which exploits the chemical instability of RNA to reveal ligand-induced changes in RNA structure (1). SAM causes substantial changes in spontaneous RNA cleavage amounts at locations residing only in the four-stem junction of the SAM aptamer (Fig. 2, A and B) (19). Likewise, AdoCbl causes modest structural changes exclusively in the AdoCbl aptamer (20, 21) [see also (fig. S2)].

Independent function was confirmed by establishing dissociation constant (KD) values for the binding of one ligand in the absence or presence of the other ligand (fig. S2). KD values for each aptamer do not change significantly when the adjacent aptamer is saturated with its ligand (figs. S3 and S4). Moreover, binding profiles are consistent with a one-to-one interaction between RNA and each ligand. Therefore, the tandem aptamers do not bind ligands cooperatively, as is observed with some glycine riboswitches (12).

To establish transcription termination as the mechanism of riboswitch action, we examined metabolite-mediated changes in the levels of terminated and full-length transcripts. Transcription in vitro of a DNA construct encompassing the tandem riboswitches from B. clausii metE revealed that both ligands cause an increase in transcription termination relative to full-length transcripts (fig. S5) only at the two sites predicted to function as intrinsic transcription terminators (Fig. 2A).

Modulation of transcription termination in vitro is modest compared with similar in vitro transcription termination assays conducted with other riboswitches (1719). Therefore, we conducted an analysis of mRNA transcript types in vivo to provide further evidence that SAM and AdoCbl induce transcription termination. Four sites within the metE RNA were targeted for qRT-PCR (quantitative reverse transcription–polymerase chain reaction) (29) with DNA primers that selectively amplify regions corresponding to the SAM (UTR-1) or AdoCbl (UTR-2) aptamers or to two different regions within the open reading frame (ORF-1 and ORF-2) (Fig. 3A). For quantification, we used samples of total RNA isolated from B. clausii cells grown in chemically defined medium (see SOM) alone or supplemented with either methionine or AdoCbl. Methionine supplementation is known to increase levels of SAM in bacteria (30) and has been used previously to regulate genes carrying SAM riboswitches (19).

Fig. 3.

Control of metE transcript length by tandem SAM and AdoCbl riboswitches from B. clausii. (A) Predicted metE transcript lengths and regions analyzed by qRT-PCR. FL designates full-length transcript (sections are not drawn to scale). (B) Representative qRT-PCR data for metE RNAs isolated from B. clausii grown with metabolite supplements as indicated. Amplification of region ORF-2 was monitored by detection of increasing SYBR Green fluorescence. A region of 16S RNA was amplified to confirm that equivalent amounts of RNA were used in each amplification. Some data points for 16S RNA quantification are obscured because of their similarity. (C) Plot of the relative amounts of B. clausii metE mRNAs of different lengths detected by qRT-PCR of four target sites using RNAs isolated under four growth conditions as indicated. A value of 1 reflects the level of mRNA measured when cells were grown without metabolite supplementation.

Selective amplification of ORF-2 of metE RNAs isolated under four growth conditions revealed substantial reductions in full-length transcripts when either compound was added to the medium, and an even greater decrease in full-length transcript abundance resulted when both compounds were added (Fig. 3B and fig. S6). These results suggest that both riboswitches control transcription termination and that both work independently.

A compilation of similar qRT-PCR data for the four amplified regions shows that there is a substantial shift from full-length to truncated transcript types under excess methionine or AdoCbl growth conditions (Fig. 3C and fig. S7). The progressive decrease in transcript abundance of RNAs that carry SAM, AdoCbl, and ORF regions is consistent with tandem riboswitches that act independently to terminate transcription on binding their corresponding ligands. Moreover, the vast majority of the changes in transcript abundance can be explained by riboswitch-mediated transcription termination without invoking other mechanisms such as transcription initiation or mRNA degradation.

We also investigated whether the metabolite-induced changes in relative abundance of transcript types corresponded to changes in protein expression. The complete B. clausii metE 5′ UTR was fused to a β-galactosidase reporter gene and integrated into the genome of B. subtilis. Reporter function was measured for cells grown in minimal medium, or in medium supplemented with methionine or AdoCbl (Fig. 4 and fig. S8). A construct carrying the wild-type tandem riboswitch yields a pattern of gene control that is consistent with NOR gate function. Some repression of the wild-type construct is expected in minimal medium, because the bacterial strains used for qRT-PCR and protein expression studies naturally synthesize methionine. Also, the AdoCbl concentration used for the in vivo assays was less than that needed for maximal repression to reduce interference of the compound's absorbance with the reporter signal (see SOM).

Fig. 4.

Control of reporter gene expression by the tandem riboswitch arrangement exhibits NOR logic function. Plot depicts the level of gene expression for a β-galactosidase reporter gene fused to the 5′ UTR of the B. clausii metE mRNA or to two mutants that disrupt essential structures in the aptamers for SAM (M1; nucleotides 30 and 31 changed to A and G, respectively) and AdoCbl (M2; nucleotides 250 through 252 changed to UGG, respectively). β-Galactosidase activity (Miller units) measured in the absence of excess metabolites yielded 5.4 (± 1.6) for the wild-type UTR, 85.6 (± 17.0) for M1, and 15.8 (± 4.7) for M2.

Mutations that disrupt the consensus sequence or secondary structure of the SAM aptamer (M1) or the AdoCbl aptamer (M2) cause insensitivity to the corresponding ligands but do not affect the function of the adjacent unaltered riboswitch. These data support our conclusion that the B. clausii metE 5′ UTR carries two simple riboswitches that together permit protein production to be controlled by two different metabolites. At SAM and AdoCbl concentrations either well above or below those needed to trigger riboswitch action, the composite system operates as a genetic switch with Boolean NOR logic (27). Overall, this tandem riboswitch arrangement provides a capability that is similar to that found in Escherichia coli metE gene control (31, 32). E. coli uses one protein factor to directly sense SAM and another protein factor that responds indirectly to AdoCbl (see SOM). The fact that these distantly related bacteria address the same complex gene control challenge with different polymers suggests that riboswitches are comparable to proteins in their function as metabolite-sensing gene regulators.

In addition to the NOR gate arrangement (Fig. 5A) described in this report and the cooperative glycine riboswitch (Fig. 5B) described previously (12), five other tandem riboswitch systems have been discovered. One new-found arrangement (Fig. 5C and fig. S9) includes a sequence and secondary structure that correspond to a guanine riboswitch aptamer (33) and a second element that is predicted (9) to function as a riboswitch aptamer for an unknown ligand. The architecture leads us to suggest that guanine binding would switch off expression and that the second ligand would switch on expression. Possible genetic outcomes are discussed in fig. S10.

Fig. 5.

Diverse tandem riboswitch configurations and functions. (A) Tandem SAM and AdoCbl riboswitches from the B. clausii metE mRNA described in this study. Transcription terminator stems are shaded light blue. (B) The common configuration of glycine (gly) riboswitches that carry two aptamers and one expression platform as observed with the B. subtilis gcvT mRNA (12). (C) Guanine (gua) aptamer (33) adjacent to a candidate riboswitch termed “ykkC” (9) with unknown ligand specificity as observed in the Moorella thermoacetica purE mRNA. (D) Tandem riboswitches for thiamine pyrophosphate (TPP) (34) found upstream of the Bacillus anthracis tenA gene, which is similar to a tandem TPP riboswitch arrangement reported previously (36). (E) Tandem AdoCbl riboswitches (20, 21) found upstream of the Desulfitobacterium hafniense DSY1435 mRNA. (F) A tandem arrangement of a candidate riboswitch termed “ykoK” (9) found upstream of the Bacillus halodurans mgtC gene. (G) Tandem arrangement of T-box RNAs (35) as present in the B. anthracis trpE mRNA. Ligands whose identities are unknown are represented by (?). Experimental evidence supporting the proposed biochemical and genetic functions is reported for the architectures depicted in (A) (this study) and (B) (12).

The most common tandem arrangement discovered involves the presence of two complete riboswitches that respond to the same compound (Fig. 5, D to F). Tandem TPP riboswitches (34) from Bacillus anthracis (fig. S11) likely form a composite genetic switch that enables greater dynamic range for gene control and greater responsiveness to changes in metabolite concentration compared with lone riboswitches. In addition, it has recently been reported (35) that T-box RNAs, which activate gene expression by binding nonaminoacylated transfer RNAs, frequently occur in tandem (Fig. 5G).

It seems likely that additional tandem riboswitch arrangements will be identified that further expand the complex functions of metabolite-sensing RNAs. However, the current collection of tandem riboswitches already provides opportunities to establish how RNAs can be assembled in modular fashion to form complex gene control elements without relying on protein genetic factors.

Supporting Online Material

Materials and Methods

Figs. S1 to S11


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article