Fresh layers of RNA-mediated regulation in Gram-positive bacteria.

Bacterial regulatory RNAs have been defined as diverse classes of cis and trans elements that may intervene at each step of gene expression, from RNA and protein synthesis to degradation. Here, we report on a few examples from Gram-positive bacteria that extend the definition of regulatory RNAs to include 5' and 3' UTRs that also act as cis and trans regulators. New examples unveil the existence of cis and trans acting regulatory RNAs on a single molecule. Also, we highlight data showing that a key RNA chaperone in Enterobacteriaceae, Hfq, does not fulfill the same role in Gram-positive Firmicutes.


Fresh layers of RNA-mediated regulation in Gram-positive bacteria
Philippe Bouloc 1

and Francis Repoila 2
Bacterial regulatory RNAs have been defined as diverse classes of cis and trans elements that may intervene at each step of gene expression, from RNA and protein synthesis to degradation. Here, we report on a few examples from Grampositive bacteria that extend the definition of regulatory RNAs to include 5 0 and 3 0 UTRs that also act as cis and trans regulators. New examples unveil the existence of cis and trans acting regulatory RNAs on a single molecule. Also, we highlight data showing that a key RNA chaperone in Enterobacteriaceae, Hfq, does not fulfill the same role in Gram-positive Firmicutes.

Introduction
Bacterial survival results from the capacity of the cell to coordinate the expression of genes in response to changing environmental conditions, thus allowing it to adapt and gain novel niches. Besides many proteins, RNAmediated controls are crucial actors modulating gene expression in bacteria [1]. RNA-mediated regulation involves (i) a target element, either RNA or a protein, and (ii) an RNA-based element made up of a sequence intrinsic to the RNA target (cis-regulatory RNA) or a transcript encoded elsewhere in the genome (trans-regulatory RNA). Cis-regulatory RNAs are usually located at 5 0 ends of messenger RNAs (mRNA). They fold in alternative structures that respond to physicochemical cues (e.g. temperature, pH) [2] or ligands (e.g., vitamins, nucleotides, ions, tRNA, or amino acids). These latter regulatory RNAs are usually called 'riboswitches' [3]. Trans-regulatory RNAs are encoded at loci that are distinct from their targets. They can bind to proteins, thereby affecting their activity, or pair to mRNAs, changing their translation efficiency, stability or termination.
Seminal publications in the last 20 years have defined these diverse classes of regulatory RNAs and unveiled their modes of action. However, recent studies in Firmicutes add 'fresh layers' to the diversity of RNA-mediated processes. This review highlights recent discoveries of diverse mechanisms whereby RNA mediates gene regulation in Gram positive bacteria. 5 0 UTR, not only the beginning; 3 0 UTR, not only the end; CDS, not only the coding sequence A few regulatory RNAs containing an open reading frame, such as SgrS in enteric bacteria [4], SR1 in Bacillus subtilis [5], or RNAIII in Staphylococcus aureus [6], have been described. In these cases, the expressed proteins are small and tend to be seen as peculiarities of regulatory RNAs. More recently, additional mRNAs encoding transcriptional regulators were demonstrated to act as regulatory RNAs via their 5 0 or 3 0 untranslated regions (UTRs), suggesting that the duality 'protein-encoding and regulation' in RNA is more frequent than anticipated.
Bacterial 5 0 untranslated regions (5 0 UTR) of mRNAs differ in length from a few to hundreds of nucleotides, and are central to numerous cis-regulatory functions. These regions may also sequester regulatory RNAs [7][8][9][10] or act as a trans-acting factor [11].
Streptococcus mutans is a Gram-positive oral commensal causative agent of dental cavities in humans, producing glucan-binding proteins (Gbp). One of them, GbpC, contributes to biofilm formation, cariogenesis, bacteremia, and infective endocarditis. Its transcription is stimulated by an unusual mechanism involving the irvA transcriptional repressor gene [12 ]. Surprisingly, the IrvA protein is not the regulatory element required for gbpC expression; the process depends solely on an irvA mRNA/ gbpC mRNA interaction, and only the irvA mRNA 5'UTR is required. Interestingly, the gbpC mRNA target region is not its 5 0 UTR, but the coding sequence ( Figure 1a). The gbpC mRNA is susceptible to RNase J2-mediated degradation but its pairing with irvA mRNA promotes its stabilization and consequently its translation ( Figure 1a). Expression of irvA is controlled by IrvR, a self-cleaving LexA-like regulator [13], and environmental stresses, including the sweetener xylitol, trigger gbpC expression by affecting its activation regulatory cascade.   (a) Regulatory 5 0 UTR. In S. mutans, expression of glucan-binding protein C is enhanced by irvA mRNA. irvA mRNA pairs to the coding sequence of gbpC mRNA and stabilizes it by preventing RNase J2 degradation [12 ]. (b) Regulatory 3 0 UTR. icaR mRNA translation in S. aureus is autocontrolled. icaR encodes a transcriptional repressor of exopolysaccharide synthesis. Depending on environmental conditions, icaR mRNA is translated (left part of panel) or not (right part). Repression of icaR translation is due to pairing of the 3 0 UTR to the 5 0 UTR of the mRNA. The reaction can be intra-or inter-molecular. The formed RNA duplex is recognized and cleaved by RNase III [15]. (c) RNA controls RNA to control RNA synthesis. Utilization of ethanolamine (EA) as a source of carbon and nitrogen involves the eut operon (symbolized by the first ORF, eutG), Demonstration of the dual function of the irvA gene highlights the possibility that many 5 0 UTR may contribute to regulation independently of their associated coding sequence.
The Hfq protein is a key RNA chaperone in Enterobacteriaceae involved in the vast majority of RNA duplexes formed between a trans-regulatory RNA and its mRNA target. The recent discovery of mRNA 3 0 regions enriched by co-immunoprecipitation with Hfq in Salmonella and Escherichia coli suggested that these sequences were functionally active in an Hfq-dependent manner (reviewed in [14]). These 3 0 RNA regions could originate (i) from internal promoters to mRNA genes (Type I) or (ii) from mRNA processing (Type II). In both cases, they are a source of regulatory RNAs that end at mRNA terminators [15]. The interactions between 3 0 RNA ends and Hfq are documented in Enterobacteriaceae and suggest a regulatory role for these RNA sequences. However, it is not the case in Gram-positive bacteria where the function of Hfq remains elusive (see below). Nevertheless, tiling-arrays and RNAseq data revealed an unexpected high density of transcription, with 3 0 UTRs that are long and sometimes conserved [16][17][18]19 ]. 3 0 UTRs may also participate in RNA-mediated control: about one-third of characterized staphylococcal mRNAs have long 3 0 UTR (>100 nt), underscoring their potential importance. Recently, the function of one of these extensions was tackled [20 ]. The icaR mRNA is about 1000 nt long and encodes a translational regulator from the TetR family (IcaR), which down-regulates synthesis of a main biofilm exopolysaccharide. The icaR 390 nt long 3 0 UTR pairs with the icaR 5 0 end over about 40 nt. This 3 0 UTR includes a C-rich region that interacts with the icaR Shine-Dalgarno (SD) sequence. Deletion of the 3 0 UTR leads to increased levels of IcaR due to stabilization and improved translation of icaR mRNA. Interestingly, authors favor the hypothesis that the UTR acts in cis, bridging the 5 0 and 3 0 ends and leading to a circular mRNA structure with a stem-loop (Figure 1b). These results are reminiscent of the intermolecular interactions of the hok/ sok mRNA of plasmid R1 modulating toxin translation [21]. While 3 0 UTR sequences controlling RNA stability and translation in eukaryotic mRNAs are widely acknowledged, icaR expression control by a 3 0 -5 0 intramolecular mRNA pairing is thus far an unconventional mode of regulation in bacteria.

Long and overlapping RNAs
Transcription of two opposite DNA strands generates RNAs that can associate by base pairing and contribute to antisense RNA (asRNA) regulation. In Listeria monocytogenes, the presence of unusually long 3 0 or 5 0 UTRs, extending over the opposite strands of adjacent genes, led to the concept of 'excludon', which defines the class of genomic loci encoding long asRNAs spanning divergent neighboring genes with related or mutually exclusive functions [22,23]. The mechanistics behind the excludon are not yet characterized. However, one actor in its regulation is likely the double-stranded RNA (dsRNA)-specific nuclease RNaseIII: a recent high-throughput study in E. coli using dsRNA-specific antibody (at least 40-mers) led to the identification of asRNAs associated with their cognate sense RNAs in an RNase III-dependent manner, supporting the existence of numerous functional complexes [24]. Observations in S. aureus also identify widespread antisense transcription and the role of RNase III in mediating degradation of overlapping transcripts [18,25].
Transcription termination is an obvious means of controlling antisense organization. Interestingly, it can be modulated by its 5 0 sequence either being a riboswitch or pairing to a trans-acting regulatory RNA. In both cases, they affect the activity of the transcription termination factor Rho [7,8,26]. Genome-wide alteration of termination may have a drastic impact; for instance, Rho is an essential protein in E. coli [27]. However, rho mutants in S. aureus and B. subtilis are viable. In B. subtilis, mRNA extensions can reach up to 12 kb, and Rho appears as a general inhibitor of antisense transcription [19 ]. Yet in its absence, Firmicutes survive even with extensive inappropriate antisense transcripts, therefore questioning why this aberrant transcription is not deleterious and consequently what is the impact of pervasive transcription in this phylum.

RNA controls RNA to control RNA synthesis
A key element of RNA-based regulation is their transcription, which is often conditionally controlled [19 ].
Recently, a new level of complexity was observed in Gram-positive bacteria, with a riboswitch-controlled trans-acting RNA. Ethanolamine (EA) is an abundant molecule of the gastrointestinal environment produced by hydrolysis of phosphatidylethanolamine, a crucial component of cell membranes. EA is used as a nitrogen source by many bacteria, but is also utilized as a carbon source by enteropathogens to outcompete the resident microbiota [28]. Among them, Firmicutes Enterococcus faecalis and L. monocytogenes employ a sophisticated RNA-mediated regulation strategy, combining riboswitch-termination, sRNA-termination and antitermination dependent processes to coordinate EA catabolism (Figure 1 Legend Continued) the two-component system eutVW and the sRNA-riboswitch eutX or rli55 (described in E. faecalis or L. monocytogenes, respectively; [29 ,30 ]), and requires vitamin B12 (B12). In the presence of EA, the anti-terminator 'ANTAR' protein (i.e., containing an RNA binding domain; [32]), EutV, is activated by phosphorylation, and binds preferentially to an RNA-sequestering sequence present at the 3 0 end of EutX/Rli55. Transcription of the eut operon is stopped by a transcriptional terminator upstream of eutG. When B12 is present, it binds to the riboswitch structure at the 5 0 end of EutX/Rli55, and transcription stops upstream of the RNA binding sequence recognized by the activated EutV. The latter binds to the transcription terminator upstream of eutG, and transcription of the eut operon may then proceed.
with nutrient availability [29 ,30 ]. In these species, EA utilization requires expression of the eut operon, comprising over a dozen open reading frames (ORFs), and a sRNA encoding gene, eutX in E. faecalis or rli55 in L. monocytogenes, which lies upstream of eut. Transcription of eutX/rli55 is controlled by a B12-riboswitch aptamer [31]. Full expression of eut is achieved in the presence of EA and vitamin B12, an essential cofactor for enzymes encoded by the operon. In E. faecalis, EA in the environment is sensed by a two component system comprising a histidine kinase, EutW, and its cognate response regulator EutV that belongs to the ANTAR RNA-binding protein family. ANTAR regulatory proteins act as transcription antiterminators [32]. When EA is absent, eut transcription stops at a terminator preceding the first ORF of the operon (eutG), and EutV is not phosphorylated and inactive; in this condition, eutX/rli55 produces a full length sRNA. In the presence of EA, EutV is phosphorylated and binds to EutX/Rli55. EutV is thereby inactive on eut expression as it is sequestered by the sRNA. When EA and B12 are both present, B12 binds to the aptamer of EutX/Rli55 and provokes formation of a transcription terminator upstream of the EutV binding sequence, generating a truncated sRNA that no longer sequesters the active response regulator. EutV then binds to the nascent eut operonic RNA and allows transcription of downstream ORFs (Figure 1c).

The Hfq mystery in Gram-positive bacteria
Hfq is a well-characterized RNA binding protein involved in post-transcriptional regulation (reviewed in [33]). In Enterobacteriaceae, Hfq stimulates base-pairing of transencoded RNAs with their RNA-partner molecules and consequently, its absence produces highly pleiotropic phenotypes.
However, within the Firmicutes phylum, Hfq is either not found (e.g. Lactobacillales) or when present, it seems to marginally affect adaptation and gene expression. The absence of Hfq in B. subtilis and S. aureus does not impact growth in nearly two thousand tested growth conditions [34,35], nor L. monocytogenes in standard laboratory growth conditions [36]. Nevertheless, in L. monocytogenes, Hfq does contribute to osmotic and ethanol stress adaptation [36]. Comparative transcriptome studies in these organisms with their respective hfq derivatives reveal moderate or no transcriptional variations as compared to Enterobacteriaceae [17,34,37,38]. Although Hfqs of these Grampositive species bind to RNAs [37,39,40], their contribution in sRNA-mediated regulation remains enigmatic. So far, no sRNA-dependent regulation in S. aureus and B. subtilis was shown to require Hfq (e.g. in S. aureus [41][42][43][44] and in B. subtilis [45][46][47][48][49]). In B. subtilis Hfq is required for ahrC mRNA translation, a target of SR1 sRNA; oddly, however, the stimulatory effect of Hfq is SR1-independent [46,47]. Within the class of Bacilli, the only documented sRNA/Hfq-dependent regulation concerns the L.
monocytogenes sRNA LhrA, which directly down-regulates expression of at least three genes and is protected from degradation by Hfq [50,51]. However, a transcriptome study indicated that in the tested conditions, the absence of Hfq did not impact the abundance of over 60 sRNAs (including LhrA) [17,52]. Consequently, as was seen in S. aureus and B. subtilis, sRNA stabilization by Hfq in Listeria is likely not a general trait. Complementation of either Salmonella or E. coli Dhfq mutants has often been obtained by hfq genes from other species (e.g. [53]); however Hfq from B subtilis or S. aureus does not compensate the absence of Hfq in Salmonella [34,54]. These results indicate that the ability of those Hfq proteins to stimulate sRNA/mRNA regulation is either lost or restricted to specific pairs.
As genomes of Firmicutes have a low GC content, it was suggested that RNA pairing properties would be different in this phylum and consequently Hfq would not be required for sRNA-mediated regulation [55]. However, recent results show that Hfq plays an essential role in Clostridium difficile [56 ] which is 29% GC. Depletion of Hfq drastically affects its growth and gene expression, and notably, alters the levels of sRNAs shown to bind Hfq [56 ]. Hfq from C. difficile restores phenotypes associated to the Dhfq mutation in E. coli [53], further supporting its functional role in sRNA regulation.
Hfq forms a homo-hexameric donut-like structure. A conserved arginine rich sequence (RRER) on the outer rim of Hfq is required for E. coli Hfq chaperone function [57]. This motif is not conserved in S. aureus Hfq, and the presence of just one arginine in B. subtilis and Listeria [57] may explain why Hfq is not associated with general sRNA regulations. In contrast, two arginines present in the Clostridium Hfq motif (RKER) may contribute to its activity.
The hfq gene is present in most, if not all, natural isolates of B subtilis, S aureus and Listeria. It was recently shown that Hfq is involved in B. subtilis stationary phase survival and therefore provides a selective advantage for this soil bacterium, which encounters starvation periods [34,38]. Similarly, Hfq contributes to the long-term survival of L. monocytogenes under amino-acid-limiting conditions, which could explain its contribution to pathogenicity [36]. Possibly, the observed hfq virulence-associated phenotype in S. aureus [37] could be related to adaptation to stationary phase. In contrast to the generalized activity of Hfq proteins in Enterobacteriaceae, the orthologs present in the Firmicutes appear to act on a restricted set of substrates, indicating a more specialized role.

Concluding remarks
Past discoveries on posttranscriptional regulation initially emerged from studies in E. coli and Salmonella. The new results from Gram-positive bacteria, some of which are highlighted here, bring to light a wide diversity of mechanisms that regulate bacterial physiology. Different mechanisms may have been selected to cope with specific sets of enzymes acting on RNA metabolism (e.g. RNases [58,59]), alternate composition (e.g. to accommodate low G + C genome composition) or other evolutionary constraints. Analysis of the distribution of short and long RNA fractions in several Firmicutes revealed the existence of a generalized antisense transcription with processing patterns that differ from those in Gram-negative bacteria [18]. Different phyla were shown to have distinct regulatory features, for example, with respect to sRNA Hfqdependency, Rho essentiality, and RNase components. We expect that novel alternative regulatory solutions established in Gram-positive bacteria remain to be found.