Burning the Candle at Both Ends: Have Exoribonucleases Driven Divergence of Regulatory RNA Mechanisms in Bacteria?

ABSTRACT Regulatory RNAs have emerged as ubiquitous gene regulators in all bacterial species studied to date. The combination of sequence-specific RNA interactions and malleable RNA structure has allowed regulatory RNA to adopt different mechanisms of gene regulation in a diversity of genetic backgrounds. In the model Gammaproteobacteria Escherichia coli and Salmonella, the regulatory RNA chaperone Hfq appears to play a global role in gene regulation, directly controlling ∼20 to 25% of the entire transcriptome. While the model Firmicutes Bacillus subtilis and Staphylococcus aureus encode a Hfq homologue, its role has been significantly depreciated. These bacteria also have marked differences in RNA turnover. E. coli and Salmonella degrade RNA through internal endonucleolytic and 3′→5′ exonucleolytic cleavage that appears to allow transient accumulation of mRNA 3′ UTR cleavage fragments that contain stabilizing 3′ structures. In contrast, B. subtilis and S. aureus are able to exonucleolytically attack internally cleaved RNA from both the 5′ and 3′ ends, efficiently degrading mRNA 3′ UTR fragments. Here, we propose that the lack of 5′→3′ exoribonuclease activity in Gammaproteobacteria has allowed the accumulation of mRNA 3′ UTR ends as the “default” setting. This in turn may have provided a larger pool of unconstrained RNA sequences that has fueled the expansion of Hfq function and small RNA (sRNA) regulation in E. coli and Salmonella. Conversely, the exoribonuclease RNase J may be a significant barrier to the evolution of 3′ UTR sRNAs in B. subtilis and S. aureus that has limited the pool of RNA ligands available to Hfq and other sRNA chaperones, depreciating their function in these model Firmicutes.

ProQ (20). In fact, ProQ appears to generally be absent in Firmicutes (21,22). In Listeria monocytogenes, it appears that Hfq contributes to pathogenesis and has a role in certain stresses such as osmotic and amino acid-limiting conditions (23); however, an hfq deletion showed no major sRNA expression changes (24). In S. aureus, the expression and role of Hfq appear to be strain specific, and the deletion does not seem to have the highly pleiotropic effects seen in E. coli hfq mutants (25,26). The exception to the rule may be Clostridioides difficile where deletion of hfq affects expression of 224 genes (5% of genes, compared with 785 [18%] of genes in Salmonella Typhimurium [27]) and has pleiotropic effects on sporulation, growth, morphology, and stress responses (28). Hfq binds and stabilizes a subset of sRNAs in C. difficile (28,29), and, importantly for the discussion below, recent Hfq RNA immunoprecipitation sequencing (RIP-seq) experiments have identified 18 39 UTR-encoded sRNAs, including five type II 39 UTR sRNAs (29). The C. difficile transcriptome may encode between 42 and 251 regulatory sRNAs (29,30), and the relative proportion of the total sRNA repertoire that is generated from 39 UTRs is unclear, but it seems that the role of Hfq has been expanded compared to other Firmicutes. Like B. subtilis and S. aureus, C. difficile encodes RNase J1 (49.8% and 52.1% amino acid identity to S. aureus and B. subtilis, respectively) and is expected to have 59!39 exoribonuclease activity.
In the section below, we propose that there exists a continuum of regulatory 39 UTR independence and that the 59!39 exoribonuclease activity of RNase J in B. subtilis and S. aureus may be a barrier to evolution of independent regulatory 39 UTR sRNAs along this continuum.

A CONTINUUM OF REGULATORY sRNA EVOLUTION
Multiple pathways likely exist for the evolution of regulatory sRNAs within the transcriptome (31)(32)(33)(34)(35)(36)(37)(38). However, for the evolution of any regulatory RNA species, the first steps are transcription and stabilization. Without both, there would be limited opportunity for interactions with target RNAs and for gaining a foothold on the ladder to positive selection. Small RNAs in E. coli largely appear in ancestral genomes before their cognate target mRNA binding sites, suggesting that sRNAs are first produced and then drive evolution of target mRNAs (39). Regulatory RNA species are suggested to have low expression levels that increase as the sRNA becomes integrated into the host regulatory network (35). Pervasive transcription has been suggested as a source of regulatory RNAs, and this occurs in most bacterial genomes but is limited by H-NS, RNase III, and Rho terminator (40). RNA surveillance within the cell also prevents accumulation of aberrant transcripts that lack stabilizing features like a structured 39 or 59 end. The most abundant stable RNA species within the cell that are not subject to the evolutionary constraints exerted by CDS or RNA structure (e.g., rRNA, tRNA, and transfer-messenger RNA [tmRNA]) are the UTRs of mRNAs. The 39 UTRs of mRNAs have been proposed to be a "playground" for sRNA evolution and may serve as a major reservoir of unconstrained RNA sequence for the evolution of regulatory RNA (41,42).

mRNA 39 UTRs THAT ACT IN CIS
cis-acting regulatory 39 UTRs are well-documented in eukaryotes and modulate the expression of the upstream CDS. A limited number of regulatory mRNA 39 UTRs have also been identified in bacteria (i.e., that are not independent or processed transcripts). The simplest arrangement of a regulatory UTR and target mRNA is found in S. aureus where the 39 UTR of icaR mRNA loops on itself (or between icaR mRNAs) and base pairs to the ribosome-binding site (RBS) of its own 59 UTR to block translation and promote RNase III-dependent degradation (43). In this arrangement, the mRNA UTRs act in cis, and a single transcript serves as both regulatory RNA and target RNA. A similar regulatory interaction, with the opposite regulatory effect, has been described in B. subtilis, where an interaction between the 59 UTR and 39 UTR of hbs mRNA occludes an RNase Y cleavage site in the 59 UTR and stabilizes the mRNA (44). The relative simplicity of this regulation suggests that more examples of cis-acting regulatory mRNA 39 UTRs may exist and control gene regulation in Gram-positive Firmicutes. This is supported by the observation that 39 UTRs are more variable than CDSs when species within the same genus are compared, and variation in the 39 UTR appears to be partly responsible for the differences in expression levels of orthologous coding sequences (45).

mRNA 39 UTRs THAT ACT IN TRANS
A slightly more complex variation where mRNA 39 UTRs act in trans is found in the Gram-positive bacterium Listeria monocytogenes. The mRNA 39 UTR of listeriolysin O encoded by hly base pairs with the 59 UTR of the listeriolysin O chaperone mRNA prsA2 (46). This mRNA-mRNA base-pairing between the hly 39 UTR and prsA2 59 UTR blocks RNase J1 exonucleolytic attack of the prsA2 59 end, stabilizing the chaperone transcript and listeriolysin O protein (46). The hly 39 UTR is an elegant example of a dual-function mRNA that provides coherent regulatory connections between functionally related mRNA UTRs. Many mRNA 39 UTRs in Gram-positive bacteria are long, and in S. aureus, more than 30% of mRNA 39 UTRs are greater than 100 nucleotides (nt) (43) (compared to 15% in E. coli [47]), suggesting that trans-acting regulatory 39 UTRs could be a widespread mechanism of regulation in Firmicutes. An advantage is that trans-acting regulatory 39 UTRs like the hly 39 UTR are protected from the 59!39 exoribonucleolytic processing of RNase J (to the extent that the mRNA is protected).

PROCESSED 39 UTRs THAT ACT IN TRANS (TYPE II 39 UTR sRNAs)
Many examples have now been described where the dual regulatory and coding functions of mRNAs are separated into distinct RNA species through processing or independent transcription of regulatory 39 UTRs. In S. aureus, the regulatory 39 UTRderived sRNA RsaC is generated by endoribonucleolytic cleavage of the polycistronic mntABC-rsaC transcript by RNase III (48). RsaC is, in effect, the long 39 UTR of the mntABC mRNA (encoding a manganese transporter) and while they share a transcriptional activation signal (repressed in the presence of Mn 21 by MntR), RNase III cleavage separates these transcripts so that they have independent fates within the cell. RsaC retains mRNA targets that are functionally coherent with the mntABC-rsaC operon, repressing the Mn 21 -dependent superoxide dismutase (SodA), and other metal-dependent pathways. RNase III cleavage of mntABC-rsaC generates a free 59 end that should render RsaC highly susceptible to RNase J exonucleolytic attack. However, a notable feature of RsaC is the 25-nt stem-loop that sequesters the 59 end in a duplex preventing exoribonuclease attack and stabilizes the 39 UTR-derived sRNA (48).
A further Gram-positive 39 UTR-derived sRNA has been described in Streptomyces coelicolor. The 39 UTR of sodF mRNA, encoding an Fe-containing superoxide dismutase, is processed to release the 90-nt s-SodF sRNA. s-SodF destabilizes the mRNA for Nicontaining superoxide dismutase (sodN) allowing coordination between these functionally related genes (49). Like S. aureus, S. coelicolor encodes RNase J, and to prevent rapid degradation, the 59 end of s-SodF also folds into a 20-nt stem-loop, sequestering the 59 from exonuclease attack (49). Protective 59 structures would be expected in 39 UTR-derived sRNAs of any bacterium with robust 59!39 exoribonucleolytic activity, which may form an evolutionary barrier for widespread evolution of the regulatory 39 UTR sRNAs in RNase J-encoding bacteria.
In contrast to Gram-positive bacteria, for which relatively few 39 UTR-derived sRNAs have been described so far, Gram-negative organisms appear to be replete with these regulatory sRNA species that are released from mRNAs by endonucleolytic cleavage (5,(50)(51)(52). In the Gram-negative pathogen Vibrio cholerae, the 39 UTR sRNA OppZ is encoded at the 39 end of the oppABCDF operon (encoding an oligopeptide transporter). RNase E cleavage after the oppF stop codon releases OppZ from the parent transcript. In a regulatory circuit that parallels the 59 UTR-39 UTR looping autoregulation of icaR in S. aureus, OppZ binds the upstream RBS of oppB to silence expression and control the cellular levels of the precursor oppBCDF transcript through a feedback loop (53). The OppZ regulon is narrow and appears to regulate only the oppBCDF operon. Other 39 UTR-derived sRNAs control expression of genes that are functionally linked to the protein encoded within the mRNA, indicating that 39 UTR-derived sRNAs can act to coordinate protein expression between functionally related mRNAs. A recent example is the 39 UTR sRNA narS, derived from the narK mRNA encoding a nitrate (NO 3 2 ) transporter in Salmonella. The narK mRNA is expressed during anaerobic respiration (54,55), and RNase E cleavage releases NarS from the mRNA (56). NarS negatively regulates the nitrite (NO 2 2 ) transporter nirC, which is located within the nirBDC-cysG operon and controls cytoplasmic nitrate levels during anaerobic growth (54). NarS is able to repress nirC through a perfect 14-nt interaction that blocks the RBS of nirC. Similar to OppZ, NarS appears to regulate a single target, nirC, allowing suboperonic coordination between narK and nirC mRNAs.

INDEPENDENTLY TRANSCRIBED 39 UTRs THAT ACT IN TRANS (TYPE I 39 UTR sRNAs)
Transcription from an internal promoter incorporates further regulatory independence and allows integration of new transcriptional regulatory signals. This should correlate with increasingly diverse regulons as the regulatory RNA becomes more integrated into the broader regulatory network (57). Notably, many 39 UTR sRNAs that are transcribed from an independent promoter are still processed by RNases to release the mature sRNA, and many require the processed 59 monophosphate for activity, presumably for activation of the RNase E sensor pocket (58). One example of independent transcription of an mRNA 39 UTR is DapZ in Salmonella. The independent gene-internal promoter of DapZ is located upstream of the stop codon of the essential lysine biosynthetic dapB gene. Transcription from the DapZ promoter is controlled by the transcriptional activator HilD (59). Overexpression of DapZ in Salmonella showed ;15 differentially expressed mRNAs (such as the glt operon and serA and cycA mRNAs), and further experiments confirmed that DapZ negatively regulates the synthesis of major ABC transporters through the repression of both dpp and opp mRNAs (59). Like type II sRNAs, DapZ also shows evidence of internal cleavage; however, these cuts appear to be nonproductive, as they remove the R1 seed that allows it to commandeer the GcvB regulon (59).
In the Gram-positive bacterium Lactococcus lactis, the 39 UTR of the arginine responsive transcriptional factor argR encodes a 66-nt 39 UTR sRNA termed ArgX, which is transcribed from an ArgR-responsive promoter within its own 39 UTR. Both ArgX and ArgR function in a negative feed-forward loop to repress the arc arginine catabolism operon during arginine limitation (60). ArgX does not have a structured 59 end to prevent RNase J exonucleolytic attack. In B. subtilis, RNase J activity is inhibited by the 59 triphosphate present on primary transcripts (61), and this may protect the unstructured ArgX 59 from degradation. These results suggest an alternative pathway for stabilization of 39 UTRs in bacteria with 59!39 exonuclease activity: acquisition of an internal promoter and a 59 triphosphate "cap."

DIVERGENCE OF RNA SURVEILLANCE MACHINERY MAY EXPLAIN DIVERGENCE OF REGULATORY RNA MECHANISMS
Collectively, the progression from cis interactions between mRNA 59 and 39 UTRs, to trans interactions between mRNA UTRs, to cleavage of trans-acting regulatory 39 UTRs, and to independent gene-internal transcription of trans-acting 39 UTRs may represent a continuum of sRNA evolution (Fig. 1). While there are clearly other sources of stable RNA for the evolution of regulatory sRNAs (e.g., pervasive transcription, genome rearrangements, tRNA spacers, and anti-termination regulated promoters), we propose that the stepwise acquisition of 39 UTR sRNAs outlined here may serve as a major pathway for the evolution of regulatory sRNA, particularly in Gammaproteobacteria. If mRNA 39 UTRs are a substantial evolutionary source of sRNAs, it may explain why sRNAs and chaperones appear so different between Gram-positive and Gram-negative bacteria. Many Gram-negative organisms lack a 59!39 exoribonuclease and rely on endoribonucleolytic cleavage (such as RNase E) and 39!59 exoribonucleases to degrade RNA within the cell. In the model Gram-negative bacteria E. coli and Salmonella Typhimurium, RNase E is the major endoribonuclease and cleaves RNA on average every 175 nt (56), generating short RNA fragments that are degraded by exonucleolytic processing from the free 39 end by PNPase, RNase R, and RNase II. Protective RNA structures at the 39 end of RNA transcripts inhibit exonucleolytic attack and can lead to differential stability of genes within polycistronic transcripts or internal RNA fragments (47). It seems likely that the last RNase E cleavage fragment of most mRNAs would also have increased stability. This fragment is protected by an intrinsic terminator (or another 39 structure for Rho-terminated transcripts [47]), and these 39 fragments might be expected to have a slightly longer half-life in bacteria lacking 59!39 exonuclease activity. Analysis of RNA stability 200 nt before and after stop codons indicates that 39 UTRs in E. coli are more stable than the upstream coding sequence (Fig. 2A). In contrast, the 39 UTRs of S. aureus are less abundant than the upstream coding sequence even at steady state (Fig. 2B). This is supported by differential transcriptome sequencing (dRNA-seq) data that identify transcription start sites (TSS; triphosphorylated 59 ends) and processing sites (PS; monophosphorylated 59 ends). In E. coli, processing sites are abundantly detected at stop codons (Fig. 3A), likely reflecting the increased stability of 39 UTRs or 39 terminal mRNA cleavage fragments. Similar RNase E-dependent cleavage sites at stop codons has been described in S. Typhimurium (56). In S. aureus and Bacillus amyloliquefaciens, processing sites are less abundant at stop codons (relative to primary transcription at start sites), suggesting that independently transcribed or processed 39 UTRs are less abundant in these Gram-positive organisms (Fig. 3B to D).
In E. coli and Salmonella, it is plausible that this pool of 39 mRNA fragments provides a source of transiently stable, unconstrained RNA sequence space for the selection and evolution of regulatory RNA features. It follows that RNase J exoribonucleolytic activity in many Gram-positive organisms may pose a significant barrier to evolution of new 39 UTR regulatory RNAs. Without first acquiring stabilizing 59 structures, these 39 RNA fragments are rapidly degraded and would have limited opportunity to gain a foothold on the ladder to positive selection.
The functional importance of Hfq appears to be significantly expanded in many Gram-negative organisms, and we propose that this may be linked to the availability of "stable" 39 UTR degradation intermediates (Fig. 1). Hfq binds many sRNAs to a proximal RNA binding surface that recognizes the poly(U) tail of the intrinsic terminator. Hfq binding appears to be associated with transcripts that carry a slightly longer poly(U) tract, which provides some selectivity within a pool of hundreds of transcripts that utilize intrinsic termination (62)(63)(64). Extension of the poly(U) tail of 39 UTR fragments, and association with Hfq, may provide a secondary step on the ladder to functional sRNA. Hfq binding would further stabilize the 39 end of the 39 UTR fragment by occluding 39!59 exonucleases and allow selection of mRNA seed complementarity. This may be one of the reasons that Hfq function has been depreciated in many Gram-positive organisms that encode 59!39 exonucleases: as the final endonucleolytic cleavage fragment would not have increased stability by default, there may not exist a ready pool of 39 UTR fragments to positively select through stabilization and target annealing. In   FIG 1 Legend (Continued) cis-encoded 59 UTRs, trans-encoded 59 UTRs, and CDSs to control mRNA stability. The regulatory functions of many 39 UTRs appear to have been physically separated from the mRNA. For bacteria that lack 59!39 exoribonuclease activity, cleaved 39 UTRs will be stabilized by the intrinsic terminator alone (left branch). This may allow cleaved 39 fragments to acquire additional stabilizing features, like an extended poly(U) tail that recruits the match-making sRNA chaperone, Hfq. As the regulatory 39 UTR acquires more mRNA targets, additional regulatory inputs (like internal transcription start sites) may allow further separation of 39 UTR and mRNA functions. (Right branch) In bacteria that encode 59!39 exoribonuclease activity, mRNA 39 UTRs are efficiently degraded from the 59 end. These 39 UTRs must first acquire stabilizing 59 structures (stems that sequester the 59 end), or internal promoters that deposit protective 59 triphosphates, before they are stabilized.
Opinion/Hypothesis ® microorganisms that efficiently degrade 39 UTR fragments, Hfq may be deprived of an important source of RNA to select for advantageous sRNA-mRNA interactions, leading to the depreciation of Hfq function.
An analogous scenario has occurred in many Gram-positive Firmicutes where the function of Rho terminator may have been depreciated because the pioneering ribosome lags behind the elongating RNA polymerase, exposing the nascent transcript to premature Rho interactions and potentially toxic transcription termination (65). In contrast, in Gram-negative Proteobacteria, the pioneering ribosome remains closely associated with the elongating RNA polymerase and prevents pervasive Rho association and premature termination. For both Rho and Hfq, the availability of RNA targets in many Gram-positive organisms may have selected against their widespread incorporation into gene regulatory circuits (albeit an overabundance of targets for Rho and paucity of targets for Hfq). Opinion/Hypothesis ® CONCLUSIONS All bacteria appear to use regulatory RNA to control gene expression posttranscriptionally; however, it is clear that differences exist in the distribution and importance of RNA chaperones that facilitate sRNA-mRNA interactions. We propose that mRNA 39 UTRs are a major evolutionary source for regulatory sRNAs and highlight some of the potential intermediate stages of 39 UTR sRNA evolution. In addition, we propose that the lack of 59!39 exoribonucleases in E. coli and Salmonella allows transient stabilization of cleaved mRNA 39 UTRs, which has provided abundant raw materials for the evolution of 39 UTR sRNAs and the expansion of sRNA regulatory networks. This in turn has centralized the function of sRNA chaperones like Hfq in E. coli and Salmonella as the sRNA network has expanded. In Gram-positive Firmicutes that encode a 59!39 exoribonuclease (RNase J), cleaved 39 UTRs are rapidly degraded. In these bacteria, 39 UTR sRNAs must first acquire 59 stems or internal transcription start sites before they are stabilized. We propose that RNase J represents a major evolutionary barrier to the expansion of the 39 UTR sRNA network and has depreciated the function of the sRNA chaperones Hfq and ProQ.
Some testable predictions arise from the above. (i) As the relative importance of Hfq for global gene regulation is uncovered in more bacteria, the size of the Hfq regulon should be negatively correlated with the presence of a functional RNase J. (ii) Endoribonuclease cleavage of 39 UTR sRNAs should be less prevalent in RNase For JKD6009 data, reads were aligned using Novoalign and read counts were mapped using pyCRAC software (70). For all data sets, TSS and processing sites were identified using the tss_ps module of ANNOgesic (71) on default settings.
Opinion/Hypothesis ® J-encoding bacteria. These bacteria may make more widespread use of regulatory mRNA 39 UTRs (like hly mRNA in Listeria) that are not cleaved to protect the 59 end of the UTR. Evolution of internal transcription start sites should be the preferred mechanism of 39 UTR release, as this provides a protective 59 triphosphate.