CsrA enters Hfq's territory: Regulation of a base‐pairing small RNA

Post‐transcriptional regulatory networks in Gammaproteobacteria are to a large extent built around the two globally acting RNA‐binding proteins (RBPs) CsrA and Hfq. Both RBPs interact with small regulatory RNAs (sRNAs), but the functional outcomes of these interactions are generally distinct. Whereas Hfq both stabilizes sRNAs and promotes their base‐pairing to target mRNAs, the sRNAs bound by CsrA act as sequestering molecules that titrate the RBP away from its mRNA targets. In this issue of Molecular Microbiology, Lai et al. reveal that CsrA interacts with the Hfq‐associated and base‐pairing sRNA Spot 42. In this case, CsrA increases Spot 42 stability by masking a cleavage site for endoribonuclease RNase E, thereby promoting Spot 42‐dependent regulation of srlA mRNA. Interestingly, the effect of CsrA on srlA expression is two‐fold. In addition to affecting Spot 42‐dependent regulation, CsrA directly inhibits translation of SrlM, an activator of srlA transcription. Together, this study reveals a new function for CsrA and indicates more intricate connections between the CsrA and Hfq networks than previously anticipated. Several recent studies have identified additional RBPs that interact with sRNAs. With new RBP identification methods at hand, it will be intriguing to see how many more sRNA‐binding proteins will be uncovered.

act as sequestering molecules that titrate the RBP away from its mRNA targets. In this issue of Molecular Microbiology, Lai et al. reveal that CsrA interacts with the Hfqassociated and base-pairing sRNA Spot 42. In this case, CsrA increases Spot 42 stability by masking a cleavage site for endoribonuclease RNase E, thereby promoting Spot 42-dependent regulation of srlA mRNA. Interestingly, the effect of CsrA on srlA expression is two-fold. In addition to affecting Spot 42-dependent regulation, CsrA directly inhibits translation of SrlM, an activator of srlA transcription. Together, this study reveals a new function for CsrA and indicates more intricate connections between the CsrA and Hfq networks than previously anticipated. Several recent studies have identified additional RBPs that interact with sRNAs. With new RBP identification methods at hand, it will be intriguing to see how many more sRNA-binding proteins will be uncovered.

K E Y W O R D S
CsrA, Hfq, post-transcriptional regulation, RBP, RNase E, spot 42 destabilization of the mRNA (for a recent review see Romeo & Babitzke, 2018). CsrA is a small (14 kD) homo-dimeric RBP that preferably binds to GGA-containing motifs located in unstructured regions of target RNA molecules. Targets of CsrA include mRNAs encoding proteins involved in carbohydrate metabolism, biofilm formation, motility, and virulence (for reviews see Kusmierek & Dersch, 2018;Pourciau et al., 2020;Romeo & Babitzke, 2018).
The regulatory activity of CsrA is controlled by sRNAs such as CsrB (Liu et al., 1997) and CsrC (Weilbacher et al., 2003), which carry multiple high affinity sites that titrate CsrA away from its lower affinity mRNA targets. CsrB/C levels are increased by various different nutritional stresses or metabolites, and decreased in the presence of preferred carbon sources such as glucose.
Accordingly, CsrA is primarily active during exponential growth where it represses genes involved in stationary phase growth and stress responses.
The homohexameric Hfq protein accommodates at least four different RNA-binding surfaces. Although Hfq has been proposed to carry out a number of cellular functions, it is best known for facilitation of the imperfect base-pairing between sRNAs and their mRNA targets, which in turn leads to increased or decreased translation and/or stability (Hör et al., 2020). One of the first identified targets of Hfq was the 109 nt long sRNA Spot 42 ( Figure 1a) (Møller et al., 2002). The expression of Spot 42 is activated by glucose and inhibited by the cAMP-CRP complex. Spot 42 appears to mainly target transcripts from genes encoding transporters and metabolic enzymes involved in utilization of non-glucose carbon sources (Beisel & Storz, 2011). This includes the srlA gene, encoding a component of a d-sorbitol-specific phosphotransferase system. Spot 42 base-pairs to, and negatively regulates translation of the srlA mRNA in an Hfqdependent manner (Figure 1b).

| CONNEC TI ON S B E T WEEN THE C srA AND Hfq NE T WORK S
Although both CsrA and Hfq are globally acting RBPs, they have been considered to belong to distinct regulatory systems, both regarding their RNA target repertoires and their function. Generally, the Csr system is believed to exert its function by direct binding of CsrA to mRNA targets, while Hfq is viewed as a matchmaking protein that guides the vast majority of known sRNAs to base-pair with their targets. Along these lines of thinking, previous reports of sRNAs that interact with CsrA suggested these interactions to function through regulation of CsrA activity, akin to CsrB/C (Jorgensen et al., 2013;Parker et al., 2017;Stenum et al., 2021). On a global level, CLIP-seq experiments have suggested that CsrA interacts with a handful of Hfq-associated sRNAs in both Salmonella enterica (S. enterica) and E.
coli Potts et al., 2017), but it has remained unclear whether the interactions between CsrA and these additional sRNAs have any functional consequences.
In this issue of Molecular Microbiology, Lai et al. reveal that the sRNA Spot 42 interacts with both Hfq and CsrA in E. coli, and that both these interactions, independently of each other, affect expression of the Spot 42 target srlA (Lai et al., 2021). The Romeo and Babitzke labs had previously mapped a putative CsrA binding site in Spot 42 by CLIP-seq and verified the interaction in vitro by gel shift experiments, demonstrating that CsrA binds Spot 42 with high affinity (Potts et al., 2017). In the present study, this interaction is shown to be specific, as mutating the single GGA motif in Spot 42 substantially weakened CsrA binding. The authors then asked whether CsrA affects the cellular levels of Spot 42, and showed that CsrA promotes Spot 42 stability by protecting against RNase E-mediated decay; CsrA binding masks two RNase E cleavage sites which lie in direct vicinity of the GGA motif in Spot 42 ( Figure 1b). The authors next asked how CsrA affects Spot 42-dependent regulation of srlA expression. In the absence of either CsrA or Spot 42, the srlA steady-state levels increased. This is expected, since Spot 42 is a direct negative regulator of srlA (Beisel & Storz, 2011), and CsrA is a positive regulator of Spot 42 (Lai et al., 2021). Interestingly, in the absence of both regulators, the srlA mRNA levels were elevated beyond those observed in the single deletion strains, indicating that CsrA also affects srlA expression through a mechanism independent of Spot 42. This effect disappeared when srlA was expressed from a heterologous promoter, suggesting that CsrA indirectly represses srlA transcription. A previous transcriptomic study from the same group had shown that srlM mRNA, encoding a transcriptional activator of srlA, was elevated in the absence of CsrA (Potts et al., 2017).

Previous studies investigating the function of interactions between
CsrA and sRNAs have focused on how sRNAs can interfere with CsrA activity through sequestration, thereby counteracting regulation that CsrA would otherwise exert on mRNA targets. The classical CsrA antagonist RNAs CsrB and CsrC consist of multiple stem-loops each presenting a GGA motif for CsrA binding. That is, these sRNAs seem to have evolved to primarily, possibly exclusively, function as CsrA antagonists. In contrast, Spot 42 only contains one single GGA motif, suggesting it to be inefficient as a CsrA titrator. The additional sRNAs that have been identified by in vivo crosslinking follow the same pattern, each harboring between zero and four GGA sequences Potts et al., 2017), suggesting that binding of CsrA to these sRNAs also may serve other purposes than regulating CsrA activity.
The present study convincingly demonstrates that binding of CsrA protects Spot 42 from RNase E cleavage, but also indicates an additional possible functional outcome of this interaction. The Spot 42 RNA can be divided into two functional domains; the 5′  Figure 1a) that participate in base-pairing to mRNA targets (Beisel & Storz, 2011), while the 3′ region contains an Hfqbinding module consisting of two stem-loop structures (Møller et al., 2002). Interestingly, the single GGA sequence in Spot 42, which is required for CsrA binding, overlaps with base-pairing region III. Although not a topic of the current study, this opens up interesting avenues for future studies. Possibly, CsrA may, in addition to acting as a stabilizing factor for Spot 42, mask region III to restrict the target repertoire of this sRNA. In this light it will be interesting to see whether uncharacterized interactions between CsrA and sRNAs involve protection from RNase activity (Lai et al., 2021), masking of base-pairing regions (suggested in Lai et al., 2021), or even CsrA-dependent involvement in sRNA-mRNA duplex formation, as recently shown in Bacillus subtilis (Müller et al., 2019).

| sRNA-B IND ING PROTEIN S B E YOND C srA AND Hfq
The current study demonstrates that CsrA counteracts RNase Edependent cleavage of Spot 42. This is in line with previous studies, in which CsrA was shown to protect both the flhDC mRNA, as well as its antagonist RNA CsrB, from RNase E cleavage (Vakulskas et al., 2016;Wei et al., 2001;Yakhnin et al., 2013). While it is too early to speculate about the extent to which CsrA exerts this role on other RNA targets, it is interesting to note that modulation of RNA stability seems to be a common function among bacterial sRNAbinding proteins.
A major example of this is the post-transcriptional regulator ProQ, a FinO-domain protein that acts as a global RBP in S. enterica and E. coli (Holmqvist et al., 2018;Melamed et al., 2020;Smirnov et al., 2016). As Hfq and CsrA, ProQ binds many mRNAs and sRNAs, and has been implicated in both RNA stabilization and sRNA-based regulation (Holmqvist et al., 2018;Melamed et al., 2020;Smirnov et al., 2017). However, for the majority of the ProQ targets, functional consequences are still largely unknown, and a full understanding of this protein and its effects on RNA remain to be determined. Additionally, several other FinO-domain proteins, including FopA from Salmonella (Gerovac et al., 2020) and RocC from Legionella pneumophila (Attaiech et al., 2016), have been identified and characterized as sRNA-binding proteins. Another recent example is the sRNA-binding protein Zea, which binds a variety of different RNAs in Listeria monocytogenes (Pagliuso et al., 2019). Zea is secreted into the cytosol of mam-

malian cells during infection, thereby affecting virulence traits.
Other examples of sRNA-binding proteins include the cold shock proteins CspC/E from Salmonella that protect RNA targets from RNase E activity (Michaux et al., 2017), and RapZ from E. coli, which acts as an adaptor protein for sRNA degradation by RNase E (reviewed in Khan & Görke, 2020).

| D ISCOVERY OF NE W RB P
Recent global studies in both E. coli and S. enterica have identified hundreds of proteins as potential RBPs, the majority of which were not previously known to bind RNA (Gerovac et al., 2020;Queiroz et al., 2019;Shchepachev et al., 2019;Urdaneta et al., 2019). This suggests that the investigated bacteria possess substantially more RBPs than anticipated. What are the functions of these new RBPs?
In addition to many uncharacterized proteins, several of these putative RBPs have known functions unrelated to RNA-binding. For instance, metabolic enzymes participating in the TCA cycle and glycolysis have repeatedly been found to interact with RNA, consistent with RBP identification results in eukaryotic systems (for a recent review, see Curtis & Jeffery, 2021). One example is the eukaryotic glyceraldehyde-3-phosphate dehydrogenase (GAPDH) that "moonlights" as an RBP; binding of GAPDH to RNA increases during low glycolysis conditions and affects both translation and stability. These findings helped formulate the RNA-enzymemetabolite (REM) hypothesis (Hentze & Preiss, 2010), which states that the RNA-binding activity of metabolic enzymes links the nutritional state of the cell to gene expression. One bacterial example congruent with the REM-hypothesis is the RNA-binding activity of the E. coli aconitases AcnA and AcnB, which catalyze the conversion of citrate to isocitrate in the TCA cycle. During low iron conditions their iron-sulfur clusters are disrupted, which enables them to bind and stabilize a number of transcripts (Tang & Guest, 1999). In this way the aconitases rapidly regulate gene expression as a result of changes in iron concentration.
Several additional hypotheses can be put forward to explain the remarkably high number of proteins with RNA-binding capacity: (a) The binding of RNA to protein could modulate the activity and/or stability of the protein, in a similar way that CsrB/C counteracts CsrA activity. (b) The RNA could drive localization of proteins or, vise-versa, a protein could drive localization of RNA. (c) Binding to RNA could provide a structural scaffold for proteins of related function and retain them in close proximity in the cell through liquid-liquid phase separation. This would be particularly beneficial for proteins working in cascades, for example where one enzyme is reliant on the product of another enzyme. It is possible that many of the newly found bacterial RBPs have fewer targets than the globally acting RBPs Hfq, CsrA, and ProQ, but it cannot be ruled out that more global post-transcriptional regulators still await discovery.

| CON CLUDING REMARK S
The seminal work on CsrA and Hfq during the last decades has been instrumental for our understanding of post-transcriptional networks built by RBPs and sRNAs as cornerstones of bacterial gene regulation. Decades of research on these networks notwithstanding, the present study shows that there is still much to learn, and that regulation by CsrA and Hfq may be more connected than previously thought. Research on these two regulators has laid the ground for the discovery of additional RBPs, such as ProQ. It is now clear that many more proteins than previously assumed can function as RPBs, and that their functions may be more diverse than expected.

ACK N OWLED G EM ENTS
We are grateful to E.G.H. Wagner for commenting on the manuscript. The work in the Holmqvist laboratory is supported by the Swedish Research Council (grant 2016-03656) and the Swedish Foundation for Strategic Research (grant ICA16-0021).

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.