Gene expression control by selective RNA processing and stabilization in bacteria


Correspondence: Francis Repoila, INRA, UMR1319 Micalis, F78350 Jouy-en-Josas, France. Tel.: +(33) 134 65 20 69; fax: +(33) 134 65 20 65; e-mail:


RNA maturation is a key event regulating genes at post-transcriptional level. In bacteria, it is employed to adjust the amounts of proteins and functional RNAs, often in response to environmental constraints. During the process of RNA maturation, enzymes and factors that would otherwise promote RNA degradation convert a labile RNA into a stable and biologically functional molecule.


The expression of genetic information operates through consecutive steps ranging from transcription to post-transcriptional events, generally ending in protein synthesis. Each of these steps is regulated to adjust the expression of each gene to physiological needs imposed by the environment and the organism development. In bacteria, in contrast to eukaryotes, these steps are coupled in space and time, and their direct input, that is, RNA, is less stable; the average half-life of a bacterial messenger RNA (mRNA) is in a range of minutes vs. hours in eukaryotes (Belasco, 2010; Silva et al., 2011). Although the importance of post-transcriptional regulatory processes affecting RNA has been established for a few decades, they have been rather overlooked in bacteria. Nevertheless, several investigations in the mid-80s provided an essential base. For example, studies on the processing of T4 phage gene, 32 mRNA were instrumental in the identification of RNase E as the major endoribonuclease (Gorski et al., 1985; Mudd et al., 1988), and inverted repeats within repetitive extragenic palindromic sequences (REPs) were demonstrated to play an active role in RNA stability and affect the differential expression occurring in polycistronic operons (Newbury et al., 1987ab). In recent years, a series of discoveries has underlined the significance of the phenomenon. For instance, (1) many small RNAs (sRNAs) mediate regulations via imperfect RNA–RNA pairing and affect mRNA processing and stability (Repoila & Darfeuille, 2009; Wagner, 2009; Caron et al., 2010); (2) many genes express antisense transcripts (asRNAs) and may be subject to their effect via modulation of the mRNA stability and translational efficacy (Lasa et al., 2011); and (3) players in mRNA turnover can be targeted by antimicrobial agents highlighting the vital character of RNA processing and stability (Olson et al., 2011).

In bacteria, RNA processing cleaves RNA molecules into shorter fragments, and RNA degradation leads to oligonucleotides and ribonucleotides monomers. RNA degradation is a constitutive phenomenon affecting each transcript. It involves specific interplays of endo- and exoribonucleases and results in shutting off gene activity. However, under precise circumstances and with the involvement of specific regulatory actors (RNAs and proteins), RNA processing stabilizes RNA and becomes a crucial element controlling gene expression by modulating the fate, the abundance and the activity of transcripts (Arraiano et al., 2010; Caron et al., 2010; Condon & Bechhofer, 2011). To avoid confusion, we will employ the term (1) ‘RNA maturation’ to refer to a specific regulatory cleavage, which usually occurs within the 5′ transcript region and changes a transcript to a biologically active RNA molecule, and (2) ‘RNA degradation’ for a cleavage triggering RNA decay. For example, it is well-known that ribosomal and transfer RNAs (rRNAs and tRNAs) are synthesized as prefunctional transcripts and acquire their biological activity through maturation. Although these molecules are quite stable, they are also degraded (Deutscher, 2009).

Remarkably, distantly related bacteria such as Escherichia coli and Bacillus subtilis have converged on a similar macromolecular complex, named ‘degradosome’, for much of their RNA processing. However, the components of the degradosome may differ between species, and even be absent in some bacteria (Carpousis, 2007; Gorna et al., 2012; Lehnik-Habrink et al., 2012). The fate of each piece of a cleaved RNA molecule does not depend on a specific processing enzyme, a particular cleavage site or in the cleavage process, but rather in each end product. In addition, regulators such as proteins, metabolites, and sRNAs can guide endoribonuclease complexes to their targeted RNA sequences stimulating a regulatory cleavage rather than RNA degradation (Fig. 1; see below).

Figure 1.

Factors governing the fate of a messenger RNA. The behavior of primary transcripts relies on the action of ribonucleases (violet) and other regulatory factors that guide RNAs to degradation or to maturation into functional RNA products according to the individual case. Blue lines symbolize RNAs; Pacman cartoons and scissors symbolize endoribonucleolytic and exoribonucleolytic activities, respectively. sRNA, small RNA; mRNA, messenger RNA; RBS, ribosome binding site; AUG, translation initiation codon.

In this review, we report recent and significant examples of RNA maturation processes turning a silent or weakly active RNA molecule into a fully functional transcript by highlighting the mechanisms and regulators involved.

Common and crucial actors to RNA maturation and degradation

One of the major modes initiating RNA maturation and RNA decay is accomplished by the degradosome, a multiprotein complex centered on an endoribonuclease enzyme that cleaves single-stranded RNA (ssRNAs) substrates. To date, degradosomes have been characterized only in a few bacteria. In addition to the endoribonuclease, the degradosome usually consists of three other major components providing it with (1) a processive exoribonuclease activity, (2) a RNA helicase function, (3) an activity most likely coordinated with the carbon metabolism via a glycolytic (enolase, phosphofructokinase) or a Krebs cycle enzyme (aconitase) (Carpousis, 2007; Gorna et al., 2012; Lehnik-Habrink et al., 2012). Moreover, degradosomes seem to be membrane associated (Khemici et al., 2008; Lehnik-Habrink et al., 2011ab). Evolutionarily unrelated proteins can ensure analogous functions rendering RNA processes slightly different in various species (Morita et al., 2005; Commichau et al., 2009; Ikeda et al., 2011; Prevost et al., 2011; Roux et al., 2011). In Enterobacteriaceae (e.g. Ecoli) and related Gram-negative species, RNase E is the endoribonuclease of the degradosome; however, in the Gram-positive phylum of Firmicutes (e.g. Bsubtilis, Staphylococcus aureus) where RNase E is absent, the unrelated endoribonuclease RNase Y replaces it (Fig. 2) (Lehnik-Habrink et al., 2011ab; Roux et al., 2011). Although RNase E and RNase Y can degrade native and processed transcripts, they are sensitive to the phosphorylation status of the 5′ RNA end (Mackie, 1998; Shahbabian et al., 2009). Their action is rather limited on primary transcripts that bear a 5′ triphosphate group and is greatly enhanced once the RNA pyrophosphohydrolase, RppH, has converted the 5′ triphosphate RNA end into a 5′ monophosphate (Celesnik et al., 2007; Richards et al., 2011). Nevertheless, endoribonucleolytic cleavages often take place within the 5′ mRNA regions generating new 5′-P extremities which (1) stimulate further cleavage by RNase E and RNase Y, and (2) break apart the ribosome binding site (RBS) from the remaining mRNA, exposing the latter to further endoribonucleolytic degradation after clearing of translating ribosomes.

Figure 2.

RNA decay pathways in Escherichia coli and Bacillus subtilis. Pyrophosphate removal from the 5′ end of primary transcripts is catalyzed by the pyrophosphohydrolase RppH. This first step of RNA decay and the double-stranded-dependent internal cleavage catalyzed by RNAse III are common to E. coli and B. subtilis. Endoribonucleolytic cleavage of mRNA is performed by 5′ monophosphate-dependent endoribonucleases (RNase E in E. coli; RNase Y in B. subtilis), and the decay of cleaved RNAs is mediated exclusively by 3′ to 5′ exoribonucleases in E. coli (PNPase, RNase R, and RNase II) or by both 3′ to 5′ exoribonucleases and the 5′ to 3′ exo- and endoribonucleolytic activities of RNase J in B. subtilis. Oligoribonucleotides are finally processed to mononucleotides by oligoribonucleases.

In Gram-negative bacteria, PNPase (polynucleotide phosphorylase) ensures the processive exoribonuclease activity from 3′ to 5′ to degrade transcripts, yet external enzymes to the degradosome may assist PNPase (e.g. RNase R, RNase II, and Poly(A) polymerase) for a complete degradation (for detailed review see Arraiano et al., 2010). In Firmicutes, PNPase is present, but RNase J (made up of two paralogous enzymes, J1 and J2) is also a component of the degradosome (Commichau et al., 2009). RNase J, which has also a ssRNA endoribonuclease activity, accomplishes a processive RNA 5′ to 3′ degradation, a unique feature reported to date in bacteria and supposed previously to be restricted to eukaryotes (Britton et al., 2007; Mathy et al., 2007; Li de la Sierra-Gallay et al., 2008). Because of this capacity for 5′ to 3′ RNA hydrolysis, degradation in Firmicutes could be initiated from the 5′ end via RNase J, or from the 3′ end of the original transcript via PNPase action (a pathway common with Gram-negative bacteria), depending on the mode of initiation of the decay (Fig. 2) (Condon, 2007).

The ribonuclease III (RNase III) is another major nuclease that belongs to a universally conserved endoribonuclease family. RNase III specifically binds to double-stranded RNAs (dsRNAs) and cleaves at specific locations (MacRae & Doudna, 2007). In bacteria, RNase III generates 5′ monophosphate and 3′ hydroxyl termini, each strand of the dsRNA having two nucleotides 3′ overhanging due to a staggered cleavage. In addition to RNase E and RNase J, RNase III is an essential player in rRNA maturation, and in certain bacterial species, the three nucleases coexist and participate together in specific and coordinated RNA maturations without functional redundancy (MacRae & Doudna, 2007; Taverniti et al., 2011).

Several other bacterial endoribonucleases have been discovered. These nucleases are encoded by the chromosome (e.g. RNase G, RNase P, RNase Z) (Arraiano et al., 2010), toxin–antitoxin (TA) modules (e.g. RnlB, RelE, MqsR, MazF, VapC) (Yamaguchi & Inouye, 2009), or phages (e.g. RegB) (Sanson & Uzan, 1995). Although their active role in gene expression control and cell physiology has been well established, they do not appear to be involved in the recent examples selected for this review (excepted MazF; see below); therefore, they will not be presented in further detail.

Factors governing the fate of cleaved RNAs

Although not totally understood, the fate of each RNA molecule produced by cleavage follows from the concerted or the independent action of intrinsic features of the RNA substrate and trans regulatory factors (e.g. proteins, translating ribosomes, metabolites, and sRNAs; Fig. 1). Intrinsic features mainly rely on local sequences and/or the folding of the RNA substrate. Endoribonucleases act on ss- or dsRNAs, by recognizing sequences and/or secondary structures, but environmental modifications can induce local structural changes in the RNA folding, and these novel conformations are no longer recognized as substrate. For instance, RNase E is a ssRNA-specific enzyme cutting preferentially A/U-rich regions near a hairpin structure; conversions from ss- to dsRNA of the cleavage site region or modifications of the hairpin can abolish the RNase E activity (Mackie, 2013). Similarly, PNPase and RNase J process ssRNAs but are blocked by hairpin-loop structures. Thus, under specific physiological conditions, a local switch from ss- to dsRNA turns off degradation (RNA processing) and provides a stable RNA (RNA maturation) (Condon, 2003, 2010). Most generally, trans acting factors appear to guide the action of endoribonucleases to their targets or in contrast to ‘hide’ the latter from cleavage (Marujo et al., 2003; Prevost et al., 2011; Stazic et al., 2011; Bandyra et al., 2012). For example, in E. coli, the degradosome interacts with transcription and translation machineries, as well as with two key players of the bacterial post-transcriptional regulation, Hfq and sRNAs (Morita et al., 2005; Vogel & Luisi, 2011; Tsai et al., 2012). Hfq is a RNA chaperone providing favorable environments for RNA/RNA interactions and protects partners from degradation or, in contrast, induces their degradation (Vogel & Luisi, 2011; Sobrero & Valverde, 2012). Many sRNAs act by pairing to mRNAs, and RNA duplexes formed increase or decrease the translation efficacy and/or the stability of the mRNA. In Gram-negative bacteria, numerous cases of sRNA/mRNA duplexes associated with Hfq are substrate for RNase E and RNase III leading to RNA decay (Repoila & Darfeuille, 2009; Waters & Storz, 2009). Nonetheless, besides the 5′-, 3′-rRNA maturation (Deutscher, 2009), several examples have been described where the cleavage process does not trigger RNA decay but instead stabilizes the transcript that becomes fully functional.

RNA maturation and gene activation

The genetic information in bacteria is generally organized in operon structures that generate polycistronic transcripts containing different open reading frames (ORFs) that encode various proteins. To match specific physiological requirements, the level of each protein needs to be adjusted. Among possible mechanisms, RNA cleavage permits the separation of ORFs and confers different fates to each RNA portion (upstream and downstream of the cleavage site). In E. coli for instance, the dnaG operon expresses constitutively the rpsU-dnaG-rpoD polycistronic mRNA that encodes the S21 ribosomal protein, the primase and the vegetative sigma factor σ70, respectively. This mRNA is cleaved by RNase E between dnaG and rpoD, resulting in the selective increase in decay rate of the primase part and in a significant difference in protein levels expressed in the cell; ~ 50–100 copies of DnaG and ~ 3000 copies of σ70 (Burton et al., 1983; Yajnik & Godson, 1993). Similarly in B. subtilis, the gapA operon is mainly transcribed under a bicistronic mRNA encoding the negative regulator of its own transcription, CggR, and GapA, the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase, respectively. The differential needs for each protein are adjusted by an RNase Y-dependent cleavage occurring upstream of the gapA encoding sequence. The gapA transcript is relatively stable (> 3 min) as its 5′ end bears a stem-loop structure protecting it from RNase J degradation; in contrast, the cggR RNA is very unstable (< 30 s) (Ludwig et al., 2001; Condon, 2003).

In bacteria, glmS encodes the glucosamine-6-phosphate synthase, an essential enzyme of amino sugar metabolism. The glmS RNA is expressed from an operon structure, and its 5′ untranslated region (5′UTR) undergoes a cleavage that leads to an activation in E. coli and closely related species (activation by RNA maturation) (Urban & Vogel, 2008). In contrast, the sequence found in B. subtilis and relatives provokes RNA degradation (repression by an autocatalytic cleavage) (Winkler et al., 2004). In E. coli, glmS is transcribed on a bicistronic mRNA, downstream glmU that encodes an essential uridyltransferase. RNase E cleaves at the translation stop codon of glmU mRNA generating two mRNAs, glmU and glmS. glmU mRNA is rapidly degraded, and glmS mRNA is stabilized and efficiently translated (Fig. 3a). Indeed, the 161-nucleotide-long 5′ UTR of glmS mRNA folds in a stem loop that sequesters the ribosome binding site (RBS) and hampers translation. In the presence of Hfq, the sRNA GlmZ pairs to the 5′ portion of the glmS stem loop. The formation of GlmZ/glmS mRNA duplex releases the RBS and enables ribosomes to translate glmS mRNA. As a consequence, glmS mRNA is stabilized due to the formation of the ribonucleoprotein complex Hfq/GlmZ/glmS mRNA and the presence of translating ribosomes covering the mRNA (Kalamorz et al., 2007; Urban & Vogel, 2008). In this example, the RNA maturation process and its activator effect on glmS mRNA relies on the folding of the 5′UTR and the formation of a duplex GlmZ/glmS mRNA, both protecting the glmS transcript from further degradation after the RNase E–dependent cleavage.

Figure 3.

Gene activation by RNA maturation. (a) Maturation of glmU-glmS bicistronic operon in Escherichia coli. The cleavage by RNase E initiates the decay of glmU and the formation of a stabilizing complex between the 5′-UTR of glmS and the GlmZ sRNA allowing efficient synthesis of the glucosamine-6-phosphate synthase. (b) Maturation of the collagenase colA mRNA in Clostridium perfringens. The endoribonucleolytic cleavage of colA mRNA results in the formation of a structured 5′ end that protects colA from subsequent degradation and ensures a high level of collagenase synthesis. Green or blue lines symbolized sRNA and mRNA, respectively. The RNA chaperone Hfq is in violet.

A similar regulatory system based on RNA maturation sRNA-dependent has also been described in Clostridium perfringens, a Firmicute. In this species, the expression of the collagenase, a major virulence factor, is enhanced by the pairing of its encoding mRNA, colA, with the sRNA VR-RNA (Obana et al., 2010). The 5′UTR of colA mRNA folds in a hairpin that occludes the RBS and impairs translation. This secondary structure is targeted by VR-RNA that pairs to the sequence of colA involved in the sequestration of the RBS (Fig. 3b). Remarkably, and in addition to rendering the colA 5′UTR efficient for translation, the formation of the VR-RNA/colA duplex exposes a ssRNA cleavage site on colA mRNA between the RBS and the duplex formed with VR-RNA. When the RNA cleavage occurs, it generates a novel and translatable form of colA mRNA that bears a 5′ terminal hairpin loop protecting the transcript from further degradation and renders accessible to ribosomes the Shine-Dalgarno sequence (SD) recognized by the 16S rRNA (Fig. 3b). The processed colA mRNA is significantly more stable than the uncut version due to the absence of the 5′ RNA portion cleaved on the primary form. Data on this regulatory system indicate that another source of stabilization is provided by initiating ribosomes but not the translation process per se (Obana et al., 2010).

Regulatory RNA-mediated protection against RNA cleavage

Several sRNAs activate gene expression by pairing to an otherwise translation inhibitory stem loop on their cognate target mRNAs; a mechanism termed ‘anti-antisense’. For example, in Staphylococcus aureus, the regulatory RNA RNAIII pairs to and activates translation of the hla mRNA encoding the alpha-hemolysin; in E. coli, the sRNA RhyB activates the expression of shiA, that encodes the shikimate transport system (Frohlich & Vogel, 2009). This sRNA-dependent activation primarily affects the folding of the mRNA translation initiation region that enables translation. However, in a few cases, an additional role has been shown for the sRNA: it also blinds the degradosome, prevents the cleavage, and stabilizes its target. Thus, the sRNA-mediated protection results in the functional activation of targets similarly to the examples described in the previous section, and may be considered as an alternative to the RNA maturation-mediated activation of genes. For example, in E. coli, DsrA, RprA, and ArcZ sRNAs pair to and allow rpoS mRNA translation encoding σS, a major stress sigma factor. The degradation of the rpoS mRNA occurs via RNase E- and RNase III-mediated processes (Resch et al., 2008; McCullen et al., 2010; Vecerek et al., 2010). It was shown that the formation of the rpoS/DsrA or the rpoS/RprA duplexes not only renders the rpoS mRNA efficient for translation but also protects it against the RNase E-mediated degradation (McCullen et al., 2010). In a similar manner, the lysC riboswitch in E. coli controls the expression of the lysine-sensitive aspartokinase III, LysC, at translational level by adopting two exclusive conformations. In the absence of lysine, the lysC riboswitch folds in a structure where the RBS is accessible to ribosome and translation can proceed. This folding hides a RNase E cleavage site in the dsRNA stem. In contrast, when lysine is present, the riboswitch and the amino acid form a complex hiding the RBS (translation stops) and exposing the RNase E site allowing the entry of the degradasome and the subsequent RNA degradation. Thereby, the complex riboswitch/lysine shuts off lysC expression by blocking translation and turning the native mRNA into a substrate for the degradosome (Fig. 4a) (Caron et al., 2012).

Figure 4.

Regulatory RNA-mediated protection against RNA cleavage. (a) Lysine-dependent regulation of lysC mRNA by a riboswitch in Escherichia coli. Two conformations of the riboswitch exist depending on the availability of lysine in the cytoplasm and mediate translation (ON) or decay (OFF) of lysC mRNA. (b) Post-transcriptional regulation of the streptokinase mRNA by FasX sRNA in Streptococcus pyogenes.

A peculiar example of sRNA-mediated protection has been reported in Streptococcus pyogenes (Ramirez-Pena et al., 2010). The streptokinase (Ska) is a secreted key virulence factor of S. pyogenes under control of the sRNA FasX (Kreikemeyer et al., 2001). In absence of FasX, the abundance of ska mRNA encoding the streptokinase is extremely low, and its half-life is less than a minute. When FasX is present, the half-life of ska is increased to more than 7 min, due to the sRNA-mediated stabilization of the mRNA. A molecular analysis demonstrated that FasX pairs to the very 5′ end of ska mRNA, and this dsRNA protects the mRNA against degradation (Fig. 4b). Experimental evidences converge to suggest that the FasX-mediated stabilization is due to the absence of unpaired nucleotides at the very 5′ end of ska mRNA, possibly restricting the access of RppH, RNase Y, and RNase J (Ramirez-Pena et al., 2010).

Selective mRNA and rRNA maturation for specialized translation

Protein synthesis is initiated by a multistep process involving disassembled ribosomal subunits (30S and 50S), initiation factors (IFs, i.e. IF1, IF2, and IF3), the initiator tRNA (fMet-tRNAfMet) and mRNAs (Laursen et al., 2005; Malys & McCarthy, 2011). For a canonical mRNA, that is, an mRNA carrying a 5′UTR, the first step consists of the formation of a binary complex made up by the 30S ribosomal subunit associated with the RBS of the mRNA, where the SD sequence base pairs with the anti-SD sequence at the 3′ tail of the 16S rRNA gene. In a second step, the 50S ribosomal subunit associates with the complex 30S mRNA, resulting in the 70S translation initiation complex. Besides this ‘classical’ view, a substantial number of leaderless mRNAs (lmRNAs) efficiently translated have been discovered in the three domains of life (Moll et al., 2002; Laursen et al., 2005; Malys & McCarthy, 2011). The translation initiation process is not yet totally elucidated, but lmRNAs are preferentially bound by 70S ribosomes without prior disassembly compared with the 30S subunit, and once formed, the 70S–lmRNA initiation complex is able to engage the elongation process without requiring IFs (Udagawa et al., 2004; Brock et al., 2008). Although certainly not dominant in bacterial species, in silico predictions and primary transcriptomes show that at least 15% of the genes would be transcribed as leaderless in certain genera; for example, Actinobacteria, Helicobacter, Xanthomonas, Deinococcus (Sharma et al., 2010; Zheng et al., 2011; Schmidtke et al., 2012). The significance of the coexistence of mRNAs and lmRNAs is currently not understood, but under certain circumstances, it may be used to respond to environmental changes. In E. coli, in response to environmental conditions triggering the toxic effect of MazF, an endoribonuclease belonging to the mazEF TA module family, certain mRNAs are specifically processed into lmRNAs and are selectively translated by specialized ribosomes (Amitai et al., 2009; Vesper et al., 2011). The MazF induction shuts off synthesis of most of proteins, and only 10% are specifically produced, including proteins provoking death in the majority of cells within a population and survival in a subpopulation (Amitai et al., 2009). MazF cleaves ssRNA at specific sequences (5′-ACA-3′) located very close upstream of the translation initiation codon, converting mRNAs with 5′ UTRs into lmRNAs (Fig. 5) (Zhang et al., 2003; Vesper et al., 2011). Within the bulk of RNAs targeted by MazF, the 16S rRNA is also processed in 30S subunits and 70S ribosomes. MazF cleaves at the 3′ terminus of the rRNA and removes the anti-SD sequence. This action of the endoribonuclease MazF generates a subpopulation of ‘stress ribosomes’ that no longer recognize mRNAs but instead specifically translate lmRNAs (Vesper et al., 2011). Thus, the MazF-mediated maturation, that switches RNAs to lmRNAs and 70S ribosomes to stress ribosomes, appears as a RNA maturation process dedicated to the translation of specific proteins in response to harmful conditions.

Figure 5.

MazF-mediated specialized translation. MazF toxin is sequestered by MazE antitoxin into a protein complex. Under stress conditions, MazE is proteolyzed, and the endoribonuclease MazF is released on its active form. Its action on specific mRNAs and on 16S rRNA genes results in the translation of leaderless mRNAs by specialized ribosomes lacking the anti-SD sequence (16S*).

Concluding remarks

RNA cleavages are crucial processes to up- or down-regulate gene expression. Usually, the same enzymatic set is involved in either RNA degradation or maturation, and the transcript fates lie in their structures which can hide or unmask RBSs and ribonuclease sites. The bacteria have astutely exploited the relationship between ribonuclease and RNA structure to adapt and survive particular growth conditions; it is likely that many other ingenious regulations remain to be uncovered. At present, the prediction of transcript fates based on primary sequence is not possible; genome-wide studies of transcript stability and cleavage sites, possibly using a RNA tagging methodology (Fouquier d'Herouel et al., 2011), will be instrumental to establish the RNA degradation/RNA maturation relationship in bacterial adaptation processes.


Work in the authors’ laboratories was supported by a grant from the Agence Nationale pour la Recherche (ANR-12-BSV6-0008-ReadRNA) and French governmental institutions CNRS and INRA. F.R. thanks Stéphane Aymerich for his support. We are grateful to Colin Tinsley for his help with the English language. We apologize to colleagues whose work has not been mentioned due to space limitations.