An RNA trap helps bacteria get the most out of chitosugars

Authors


*E-mail vogel@mpiib-berlin.mpg.de; Tel. (+49) 30 28460 265; Fax (+49) 30 28460 244.

Summary

Small regulatory RNAs (sRNAs) are well known to command bacterial protein synthesis by modulating the translation and decay of target mRNAs. Most sRNAs are specifically regulated by a cognate transcription factor under certain growth or stress conditions. Investigations of the conserved Hfq-dependent MicM sRNA in Escherichia coli (article by Poul Valentin-Hansen and colleagues in this issue of Molecular Microbiology) and in Salmonella have unravelled a novel type of gene regulation in which the chitobiose operon mRNA acts as an RNA trap to degrade the constitutively expressed MicM sRNA, thereby alleviating MicM-mediated repression of the synthesis of the YbfM porin that is required for chitosugar uptake. The results suggest that ‘target’ mRNAs might be both prey and also predators of sRNAs.

Bacteria are known to produce large numbers of small RNAs (sRNAs), many of which act as post-transcriptional regulators by base-pairing to mRNAs (Waters and Storz, 2009). Since the serendipitous discovery of MicF sRNA as a post-transcriptional repressor of OmpF porin synthesis, the functional characterization of Escherichia coli and Salmonella sRNAs has shown that almost a third of them repress the synthesis of porins and other outer membrane proteins (Guillier et al., 2006; Vogel and Papenfort, 2006). The ∼84 nucleotide MicM sRNA is the latest arrival in the field of enterobacterial porin regulators. The micM gene (also known as sroB or rybC) was discovered in the ybaK-ybaP intergenic region (IGR) in genome-wide screens for new sRNAs in E. coli (Vogel et al., 2003; Zhang et al., 2003). In both E. coli and Salmonella, MicM was also observed to be tightly associated with the bacterial sRNA chaperone, Hfq (Zhang et al., 2003; Sittka et al., 2008), indicating that it acts on trans-encoded target mRNAs by short antisense pairing (Waters and Storz, 2009).

In an earlier issue of Molecular Microbiology this year, Poul Valentin-Hansen and colleagues identified the ybfM mRNA encoding a porin of, at the time, unknown specificity as a major target of MicM (Rasmussen et al., 2009). Like many other porin regulators (Guillier et al., 2006; Vogel and Papenfort, 2006), MicM repressed ybfM mRNA by Hfq-dependent pairing with the ribosome binding site (RBS) (Rasmussen et al., 2009), inhibiting translational initiation and, thereby, destabilizing this messenger (Fig. 1A, left part). On the face of it, MicM/ybfM seemed to be quite an ordinary case of post-transcriptional porin repression.

Figure 1.

Regulatory systems involving RNA trapping and mimicry.
A. RNA trapping mechanism for conditional unleashing of YbfM chitoporin synthesis from sRNA-mediated repression. Left: When inducing chitosugars are present in low concentrations, MicM sRNA and ybfM mRNA (as part of the ybfMN operon) are constitutively expressed, but Hfq-dependent pairing of MicM to the ribosome binding site inhibits synthesis of the YbfM chitoporin. RNase E degrades the untranslated ybfM messenger, whereas MicM, being refractory to coupled decay, accumulates to high intracellular levels. Right: High concentrations of the inducer further increase ybfM transcription and also turn on transcription of the chb operon. Basepairing of MicM is redirected to its binding site in the chbBC intergenic region of the polycistronic chb mRNA. The slightly longer interaction engages residues of the protective 3′ hairpin of MicM, rendering the sRNA more susceptible to ribonucleases. Thus, MicM sRNA is trapped through target mimicry by chb mRNA. Consequently, MicM is consumed, and the ensuing drop in its cellular concentrations alleviates the repression of YbfM synthesis. Stabilization of the ybfM mRNA further boosts chitoporin production, increasing the influx of chitosugars into the cell to be metabolized by products of the chb operon mRNA.
B. Target mimicry to sequester the repressor microRNA, miR-399, in the Pi deprivation response in plants (Franco-Zorrilla et al., 2007). Left: Upon Pi deprivation, miR-399 is activated and targets the PHO2 mRNA with perfect complementarity, which promotes decay of this messenger. Right: The non-coding IPS1 transcript is induced under the same condition. miR-399 recognizes a non-cleavable target site on IPS1, and this target mimic effectively sequesters the cellular population of miR399 to suppress PHO2 mRNA, which can now be translated.
C. Two seemingly homologous E. coli sRNAs form a regulatory RNA cascade in which processing and therefore inactivation of GlmZ sRNA are inhibited by the highly similar GlmY sRNA, presumably by RNA mimicry. GlmZ can then basepair to remove a translational block of the glmS mRNA messenger, thus activating GlmS synthesis. See Reichenbach et al. (2008) and Urban and Vogel (2008) for more details of the GlmYZ cascade.

In this issue of Molecular Microbiology, the Valentin-Hansen laboratory reports an exciting turn in the MicM story. They have discovered an RNA trapping mechanism in which MicM itself becomes the target of a functionally related mRNA in order to derepress the ybfM target mRNA (Overgaard et al., 2009). The study was motivated by several intriguing and unique observations. Unlike many stress-responsive sRNA genes, the micM gene is almost constitutively expressed (Vogel et al., 2003) and does not show evidence for a conserved transcription factor binding site in its promoter region (Mandin and Gottesman, 2009; Rasmussen et al., 2009). The abundant MicM sRNA seemed to be in permanent excess over its target, because the ybfM mRNA was hardly detectable under standard growth conditions (Sittka et al., 2008; Overgaard et al., 2009; Rasmussen et al., 2009). In addition, a seminal paper by the Gottesman lab (Masséet al., 2003) had shown that several Hfq-associated sRNAs using the same RBS pairing mechanism as MicM were degraded along with their target mRNAs by the major endoribonuclease, RNase E, suggesting that sRNAs act stoichiometrically. However, overexpression of ybfM RNA did not destabilize MicM, arguing against a coupled degradation in this case. Collectively, how would the repression ever be overcome to permit YbfM synthesis?

A genetic screen for factors that could prevent MicM from repressing ybfM provided the surprising answer. Aside from two enzymes with established functions in RNA metabolism, poly(A) polymerase (Kushner, 2004) and the 5′ end pyrophosphohydrolase, RppH (Deana et al., 2008), ectopic expression of two genomic regions increased the activity of a chromosomal ybfM reporter construct. One was the 5′ UTR of ybfM itself, proving that MicM could be out-competed by an excess of target RNA. Intriguingly, the other was a non-coding chromosomal fragment derived from the first IGR of the chbBCARFG operon. The chitobiose operon encodes the components of a chitobiose/-triose phosphotransferase system, along with the ChbR regulator protein (Plumbridge and Pellegrini, 2004). Notably, the active chbBC IGR fragment exhibits short complementarity to almost exactly the same region in MicM that pairs with ybfM, indicating a competing MicM–chbBC interaction (Fig. 1A, right part). However, despite the similar RNA pairing, the activities of the two regions differ radically: unlike the ybfM pairing, interaction with chbBC triggers irreversible decay of MicM, and, thus, can clear the repressor. In other words, the chBC IGR transcript can trap the MicM sRNA to unlock YbfM synthesis.

An independent study by Lionello Bossi and colleagues in Salmonella (Figueroa-Bossi et al., 2009) not only corroborates the above findings but also sheds more light on both the biological meaning and the mechanism of the RNA trap. Their work has identified YbfM as an OprD-like chitoporin that is essential for Salmonella growth on chitotriose and chitobiose as the single nitrogen and carbon source. Importantly, these chitosugars strongly induce the chbBCARFG operon, thereby activating the RNA trap and enabling the synthesis of a porin for their own uptake. The Salmonella study also provides an experimentally validated explanation as to why MicM is degraded along with the chbBC IGR but not the ybfM transcript. The MicM–chbBC interaction is slightly longer, and engages two nucleotides of the 3′ terminal stem-loop of MicM, whose partial opening seems to render MicM more susceptible for 3′ exonucleolytic decay. In addition, RNase E was implicated in two steps of this circuit; it both facilitates the MicM-mediated destruction of ybfM mRNA and mediates processing of the chbBCARFG operon mRNA around the MicM pairing region.

Taken together, the combined results of the two studies (Figueroa-Bossi et al., 2009; Overgaard et al., 2009) suggest a novel mechanism of gene regulation by which one mRNA activates another via the elimination of an inhibitory sRNA. Both ybfM and the chitobiose operon are transcriptionally induced under the same condition, i.e. by the presence of chitosugars (Figueroa-Bossi et al., 2009), but only the concomitant removal of MicM fully unleashes the synthesis of the chitoporin by removing the sRNA block, too. This provides E. coli and Salmonella, which do not secret chitinases under all conditions (Francetic et al., 2000), with a positive feed-forward loop to get the most out of surrounding chitin-derived sugars.

Although the identified regulation makes much biological sense, further studies will be needed to understand fully the logic of the RNA trap. It remains to be established if and how protein synthesis from the chb operon is ensured when its polycistronic messenger interacts with MicM. The study on Salmonella from Bossi's lab shows that the trapping of MicM is accompanied by processing of the chb operon mRNA, liberating the chbBC RNA as a ∼300 nt species whose exact termini are yet to be defined (Figueroa-Bossi et al., 2009). Can the processed chb operon mRNA still make all of the proteins necessary for chitosugar metabolism, or does it produce them in altered ratios as seen in the sRNA-regulated galETKM or glmUS mRNAs (Møller et al., 2002; Urban et al., 2007)? Altered synthesis of individual proteins from the chb operon could add another layer of complexity to this physiological circuit.

Precisely where RNase E, poly(A) polymerase and RppH act, and for which of these events Hfq is required, is also yet to be addressed. In addition, we do not know which nuclease gets rid of MicM once it is trapped. Moreover, the Gottesman laboratory has identified yet another mRNA that interacts with MicM, i.e. the dpiBA mRNA encoding a two-component signal transduction system (Mandin and Gottesman, 2009). Strikingly, genomic deletion of this target increases MicM levels (Mandin and Gottesman, 2009). Once the distinction of mRNA between effector and target of a sRNA is blurred, it is tempting to speculate that, aside from being a MicM target, the dpiBA mRNA serves as a further sink for MicM, especially given the putative physiological link between ybfM and dpiBA functions, as discussed (Figueroa-Bossi et al., 2009).

Notwithstanding these open questions, the current results have important implications for our understanding of bacterial gene regulation and for the growing field of sRNA biology. The RNA mimicry behind the MicM trap resembles recently discovered target mimicry for plant microRNAs (Franco-Zorrilla et al., 2007), a class of ∼22 nucleotide RNAs that acts as post-transcriptional repressors of eukaryotic mRNA. Specifically, during phosphate starvation, Arabidopsis miR-399 is induced to promote cleavage of PHO2 mRNA encoding a protein whose activity is not desirable in the early phase of the stress response (Fig. 1B). As miR-399 levels rise, this repressor itself becomes a target of the non-coding IPS1 transcript that prevents further decrease of PHO2 mRNA levels. The IPS1 RNA contains a very similar miR-399 interaction site to PHO2 mRNA, but two mismatches in the region required for coupled mRNA-miR cleavage prevent its consumption. Although IPS1 simply acts as a sponge that sequesters miR-399, the outcome for the primary target is strikingly similar to ybfM upregulation upon the trapping of MicM.

RNA mimicry might also underlie the activation of GlmS synthesis in E. coli by the conserved GlmYZ sRNAs (Reichenbach et al., 2008; Urban and Vogel, 2008). Here, GlmZ sRNA acts a direct translational activator by basepairing to glmS mRNA, whereas the highly similar GlmY sRNA acts indirectly by suppressing cleavage and therefore premature inactivation of GlmZ sRNA (Fig. 1C). As in the case of MicM, the regulatory RNA cascade of GlmY and GlmZ also has several entry points for Hfq, RNase E and poly(A) polymerase (Görke and Vogel, 2008).

The RNA trap to capture MicM illustrates the importance of controlled RNA degradation. With respect to sRNAs, the focus has been on the control of translational initiation, whereas mRNA decay has been regarded as adding robustness to regulatory events, especially if the regulatory sRNA is also degraded in the process (Masséet al., 2003; Morita et al., 2006). While this might also apply in part to MicM, the recognition of the decisive role of RNase E in sRNA-mediated regulation (Masséet al., 2003), of RNase E complexes with Hfq and sRNAs (Morita et al., 2005), and of RNase E-dependent mRNA silencing without translational repression (Pfeiffer et al., 2009) all indicate that programmed killing of RNA, be it mRNA or sRNA, might play a stronger role in sRNA regulatory circuits than appreciated. Likewise, we are beginning to understand better the cellular factors that ensure that sRNAs do not remain unbridled (Suzuki et al., 2006; Viegas and Arraiano, 2008).

Besides the control of outer membrane protein biogenesis, the regulation of metabolic sugar pathways has emerged as a major domain of bacterial sRNA action (Vanderpool, 2007; Görke and Vogel, 2008). Global transcriptome analyses of hfq mutants and microarray- or deep sequencing-based identification of Hfq-associated RNA in E. coli and Salmonella (Zhang et al., 2003; Guisbert et al., 2007; Sittka et al., 2008) suggest that Hfq targets not only mRNAs of many sugar transporters and metabolic enzymes but also transcription factors that coordinate the uptake and utilization of carbohydrates. Moreover, footprints of the MicM-ybfM-chb connection are clearly visible in these data sets. In Salmonella, for example, all of the three RNAs, including the chbBC IGR transcript, were highly selected by co-immunprecipitation with Hfq (Sittka et al., 2008), and ybfM is one of the most highly deregulated mRNAs in hfq deletion strains (Figueroa-Bossi et al., 2006; Sittka et al., 2008). Thus, inspection of the available global Hfq data could provide hints at additional RNA traps that operate beyond the well-studied transcriptional control of uptake and metabolism of sugars.

In summary, the exciting discovery from the Valentin-Hansen and Bossi laboratories should make reflect on how many of the mRNAs hitherto identified as sRNA targets are both prey and predators. The two studies are very timely, as they come at a moment of intense discussion about eukaryotic microRNAs, which have often been assumed to interact with hundreds of target mRNAs. A recent study has challenged this perception (Seitz, 2009) by arguing that many computationally identified miRNA targets might actually be competitive inhibitors of miRNA function. If so, the cellular RNA forest would be crowded with mRNAs that trap regulatory sRNAs the way MicM is ‘caught at its own game’ (Figueroa-Bossi et al., 2009).

Ancillary