We compiled a list of 16 sRNAs (Table 1), outside of CsrB, CsrC, and McaS. The set of potential CsrA-binding sRNAs in Table 1 encompasses the diverse activity of known sRNAs in E. coli. These sRNAs were compiled and reanalyzed from previous data showing sRNA-CsrA interactions; these include CsrA CLIP-Seq (Potts el al., 2017), HITS-CLIP (Sowa et al., 2017) and differentially expression data in E. coli (Sowa et al., 2017) (described in Materials and Methods). Eleven of the 16 sRNAs are trans-acting sRNAs that have at least one confirmed mRNA target (Table 1; Supplementary Table S1) and many are associated with specific stress responses or metabolic pathways (Supplementary Table S2). As a note, three of these sRNAs do not contain a GGA binding motif and may be interacting through degenerate GGN-like sequences. To identify putative binding sites, we applied a previously developed biophysical model of CsrA-RNA interaction (Leistra et al., 2018). A full computational workflow for using the model is in the Materials and Methods. While some of these sRNAs have been tested for CsrA binding in vitro (summarized in Table 1), potential in vivo regulatory interactions with CsrA have not been considered, excluding McaS, CsrB, and CsrC. McaS and CsrB are included as positive controls because of their previously established sponge activity for CsrA (Dubey et al., 2005; Jorgensen et al., 2013). McaS is considered especially relevant as it contains 4 copies of the GGA motif which is on the same order of magnitude as the other sRNAs of interest; this is in contrast to CsrB, which contains 22 copies of the hairpin-contained GGA motif and is therefore assumed to bind CsrA with higher affinity.

However, it is important to note, that the sequence and structural features of the motif are not strict requirements for CsrA binding as degenerate GGA-like sequences may contribute to CsrA-RNA binding and regulation in some target mRNAs. For instance, CsrA binds at non-stem loop GGA sequences in cstA in vitro (Dubey et al., 2003) and degenerate GGA-like sequences are thought to contribute to CsrA-RNA binding and regulation of some targets (Kulkarni et al., 2014; Leistra et al., 2018; Mercante et al., 2009; X. Wang et al., 2005; H. Yakhnin et al., 2011).

We first assessed whether the 16 identified sRNAs bound CsrA in vitro using electrophoretic mobility shift assays (EMSA). CsrA was purified as described in the Materials and Methods and purity was assessed by SDS-PAGE (Supplementary Figure S1). For these experiments, 10 nM of P³² - radiolabeled sRNAs were incubated with 3 µM or 6 µM of CsrA, which produced a CsrA:sRNA molar ratio of 300:1 and 600:1, respectively. These ratios were selected based on the reported binding affinities for Spot42, GadY, GcvB, and MicL presumed to be physiologically relevant via CsrA CLIP-Seq data (Potts et al., 2017). Additionally, all binding assays are performed in large excess of yeast total RNA, which has been previously shown to inhibit non-specific association of CsrA to labeled sRNA (A. V Yakhnin et al., 2012).

Across all conditions, super-shifted complexes indicating CsrA binding were detected upon addition of CsrA for 14 of 16 sRNAs (Figure 2; Supplementary Figure S2). Clear in vitro CsrA binding not yet been reported in the literature was detected for several sRNAs: SgrS, ChiX, RprA, DsrA, and SibA (Figures 2A, B, D, E; Supplementary Figure S2D). In addition to these five novel interactions, the sRNAs Spot 42, GadY, and McaS bound CsrA (Figures 2G–I) as has been previously shown, which gives confidence to our EMSA conditions generating true novel sRNA-CsrA interactions. We also observed CsrA binding to MicL, RnpB, GlmZ and SibD at the 6 µM concentration of CsrA, which we believe to be sufficient evidence of in vitro CsrA interaction (Supplementary Figures S2A, B, C, E1). It is worth noting that stronger in vitro CsrA binding has been previously detected for MicL (Potts et al., 2017); however, this prior study utilized the full-length sequence (MicL-long), while we tested the short, 5′-processed MicL-S sequence (referred to here as MicL for simplicity). The latter is thought to be the functionally active form of the sRNA (Guo et al., 2014). In contrast, CsrA binding was not detected for the GcvB and SibC sRNAs (Supplementary Figures S2F, G). Full EMSA images are included in Supplementary Figure S3.

Recently, co-immunoprecipitation study of Hfq in E. coli demonstrated that CsrA co-precipitates with Hfq in an RNA-dependent manner (Caillet et al., 2019). To qualitatively assess if sRNAs from this set can bind both CsrA and Hfq in vitro and/or if Hfq influenced CsrA-sRNA binding, we tested binding of all the 16 sRNAs with CsrA (6 µM) in the presence of different Hfq concentrations (0.375 µM or 1.5 µM). These concentrations were selected to mimic estimated ratios natively occurring between these two proteins in vivo (Materials and Methods). Importantly, we found that the presence of Hfq affected the binding of many sRNAs to CsrA, likely forming ternary complexes of CsrA-sRNA-Hfq (Figures 2A–E, H, I). These results agree with prior in vitro observation of ternary CsrA-sRNA-Hfq complexes for ChiX, RprA, and McaS (Jorgensen et al., 2013; Peng, 2014 and Dr. S Woodson, personal communication, August 2018). The complete summary of the EMSA results is compiled in Table 2. In total, we demonstrate 10 novel, and 14 total sRNA-CsrA interactions in vitro.

We next investigated the in vivo functional implications of the in vitro confirmed CsrA-sRNA interactions. We consider two non-exclusive hypotheses: 1) the possibility that sRNAs are minor antagonists of CsrA similar to CsrB and CsrC and 2) the possibility that sRNA-CsrA binding regulates downstream sRNA-mRNA target interactions. As such, we initially pursued 12 of the 14 sRNAs that 1) bound CsrA in vitro and 2) had some evidence of CsrA impact in vivo (Summarized in Table 2). We excluded SibA and SibD as these sRNAs proved difficult to work with in vivo due to toxic peptide function [(Fozo et al., 2008; Brantl and Jahn, 2015) see Materials and Methods]. This ultimately left us with 10 total sRNAs to pursue in these assays: McaS, GadY, Spot42, ChiX, SgrS, MicL, RnpB, GlmZ, FnrS, and CyaR.

To evaluate if the found in vitro binding interactions could be directly recapitulated in vivo, we employed a plasmid-based fluorescence complementation system previously developed in our lab and previously shown to capture CsrA-CsrB interactions (Gelderman et al., 2015; Sowa et al., 2017; Mihailovic et al., 2021). Under the conditions tested, we were only able to detect a significant increase in GFP complementation (indicative of sRNA-CsrA proximity) for Spot42-CsrA and ChiX-CsrA, relative to the case of CsrA and a random sRNA sequence control (Supplementary Figure S4B). While allowing us to validate the new in vivo interaction between ChiX and CsrA, the YFP complementation signal was low even for the McaS positive control. This may be due to the large MS2 binding protein affecting RNA folding. Thus, these results suggest the need for a more sensitive in vivo approach to investigate the in vitro supported sRNA-CsrA interactions.

As observed previously for CsrB, CsrC, and McaS (Jorgensen et al., 2013; M. Y. Liu et al., 1997; Weilbacher et al., 2003), we anticipated that some of the novel sRNA-CsrA interactions would fall into our first hypothesis and act as a sponge sRNA and sequester CsrA in vivo. The sponging would then impact regulation of CsrA on its mRNA targets. To screen for this sponge activity of each sRNA, we adapted a plasmid-based screen previously utilized in our lab (Sowa et al., 2017; Leistra et al., 2018). In short, we induced expression of each sRNA with a constitutively expressed glgC-gfp translational fusion reporter both K-12 MG1655 wild type E. coli and a ΔcsrBCD csrA::kan strain (i.e., Csr system deletion strain). The glgC-gfp fusion was selected as CsrA binds and represses translation of the glgC mRNA and glgC is one of the most well-characterized mRNA targets of CsrA (Baker et al., 2002; Mercante et al., 2009; Sowa et al., 2017). As a negative control, a random 80 nt sequence absent of any GGA motifs was selected from the E. coli genome (Supplementary Tables S1–S4). In this system, if a specific sRNA sequesters CsrA (diluting its effective concentration), we expect to observe an increase in fluorescence (indicative of less repressed glgC-gfp expression) following sRNA induction, relative to the fluorescent output of the uninduced system.

Screening revealed that CsrB, McaS, SgrS, and FnrS all significantly increased fluorescence of the glgC-gfp translational fusion only in the presence of induced CsrA (Figures 3D, E, p-value <0.05). While CsrB and McaS are positive controls and were expected, SgrS and FnrS are novel findings. Induction of CyaR and ChiX increased glgC-gfp fluorescence in the presence of CsrA (Figure 3D, p-value <0.05 by paired t-test), they also do so in the absence of CsrA (Figure 3E, p-value <0.05 and p-value =0.06), suggesting that these sRNAs may regulate the translational reporter directly or indirectly via a non-CsrA factor. In fact, CyaR has been reported to directly base-pair with the glgC transcript in RIL-seq experiments (Melamed et al., 2020). While ChiX has not been shown to directly target glgC, IntaRNA (Busch et al., 2008) predicts a base-pairing region between the two RNAs with a predicted ΔG of −5.83 kcal/mol; this is on the same scale as the IntaRNA prediction for the known IntaRNA ChiX-chiP interaction (ΔG of −4.82 kcal/mol), suggesting a potential direct interaction between ChiP and the glgC transcript (Mann et al., 2017). Surprisingly, GadY does not exhibit sponge activity for CsrA in our assays, as previously reported (Parker et al., 2017). This discrepancy may be explained by the lower copy number, and thus greater sensitivity, of the CsrA target reporter in the previous work.

To further evaluate the hypothesis that SgrS and FnrS are two novel CsrA sponge sRNAs, we interrupted putative CsrA binding sites within these sRNAs expecting this would abrogate CsrA sponge activity. McaS was used as a positive control in these experiments, given its earlier 1) demonstration of CsrA sponge activity, 2) confirmation of CsrA binding sites, and 3) more similar number of GGA motifs (compared to CsrB). Likewise, CyaR was used as a negative control because we anticipated that disrupting putative CsrA binding sites would not impact the likely CsrA-independent observed increase in glgC-gfp fluorescence (Figures 3D, E). For the sRNAs that contain one or more instances of the conserved GGA binding motif, we performed GGA:CCA mutations as done in previous literature to interrupt CsrA-RNA binding (Dubey et al., 2005; Patterson-Fortin et al., 2013). It should be noted that alternative mutations were made to the GGA motif if the standard GGA:CCA mutation altered the predicted secondary structure of the sRNA (Materials and Methods). We disrupted each instance of the GGA motif in the SgrS and CyaR sRNAs (termed “GGA mutants”, Supplementary Table S4). For McaS, the two GGA motifs (of 4 total) that were shown to be CsrA binding sites in vitro and in vivo in a previous study (Jorgensen et al., 2013) were mutated (“dual GGA mutant,” Supplementary Tables S1–S4). For FnrS, which does not contain a GGA motif, GGN sequences within the most likely pair of putative CsrA binding sites predicted using the biophysical model (Supplementary Table S3) were mutated (“GGN mutant 1,” Supplementary Table S4).

The mutant sRNAs were screened in the same plasmid assay. Induction of mutant FnrS, SgrS, and McaS mutant sRNAs significantly reduced fold increase of glgC-gfp fluorescence relative to the corresponding wild type sRNA (Figure 4A). Northern blotting analysis of these samples indicated both the wild type and mutant sRNAs are induced at comparable levels (Figures 4B–D). These data agree with a previous study of McaS, where mutations of the same GGA motifs significantly decreased its CsrA sponge activity, as measured by a pgaA-lacZ reporter (Jorgensen et al., 2013). Additionally, as expected, the CyaR mutant sRNA does not alter glgC-gfp fluorescence between the wild type and mutant sRNAs (Figure 4A), supporting the hypothesis that the observed effects of CyaR on glgC-gfp fluorescence (Figures 3D, E) do not involve direct CyaR-CsrA interactions. Lastly, altering the known GGA motif in Spot42 and expressing the mutant sRNA slightly alleviates the fold-repression effect observed when using the wild type Spot42 sRNA sequence, though that difference is not statistically significant (Figure 4A). This does point to the idea that the decrease in glgC-gfp fluorescence upon expression of Spot42 relies upon a direct CsrA-sRNA interaction.

The above data indicate that SgrS and FnrS can exert CsrA sponge activity similar to that of McaS in a plasmid-based system. However, the effects of these sRNAs on glgC-gfp fluorescence are relatively small (∼1.2-fold increases) (Figure 4A) compared to that observed for CsrB (∼9-fold increase) (Figure 3D). This drastic difference is expected, as CsrB contains 22 GGA CsrA binding motifs while McaS, Sgrs, and FnrS contain only 4 CsrA binding motifs at most (Summarized in Table 1). It is important to recognize that these results were collected using a plasmid-based expression system. To evaluate physiological sponge activity of SgrS, we took two approaches. First, we tracked flhD-gfp expression in the presence and absence of genomically-expressed SgrS as induced the presence of a-methyl glucoside (Walder and Vanderpool, 2007); flhD is a well characterized (stabilized) mRNA target of CsrA (Yakhnin et al., 2011). Therefore, the reduction in fluorescence when tracking expression of the flhD-gfp reporter in the presence of SgrS (relative to the same condition without SgrS), that we observed in Supplementary Figure S13B, indicated increased in SgrS-dependent CsrA sequestration (Figure 5; Figure 6; Supplementary Figure S7). To ensure these results depended on direct CsrA-SgrS interaction, we also conducted these experiments using a SgrS mutant which had its predicted CsrA binding site mutated and observed no change in flhD-gfp reporter signal (Supplementary Figure S13C). Collectively, these data supported the notion that SgrS acts as sponge sRNA of CsrA under physiologically relevant conditions. It is worth noting that the flhD-gfp fusion was expressed from a promoter previously used for mimicking physiological mRNA expression levels (Adamson and Lim, 2013). The presence and absence of SgrS was induced using growth conditions previously confirmed to alter native levels of this sRNA (Wadler and Vanderpool, 2007); we confirmed that SgrS expression was significantly upregulated under these conditions using RT-qPCR analysis (Supplementary Figure S13D). We could not confirm native FnrS sponge activity as native increase in FnrS expression requires anaerobic growth conditions, which we were not able to confirm by RT-qPCR in consistency with previous works (Durand and Storz, 2010). Conditions in which FnrS is expressed natively is an ongoing area of interest for the field and will be useful for evaluating the ability for FnrS to act as a sponge for CsrA; identification of physiological conditions where FnrS sequesters and titrates CsrA away from its targets remains an outstanding question from this work.

We next evaluated our second hypothesis: that CsrA remodels the downstream regulatory interactions of the sRNA regulatory network itself. While this work was in preparation, CsrA enhancing Spot 42-srlA mRNA repression was elucidated (Lai et al., 2022), setting the precedent that CsrA can remodel downstream sRNA networks. In light of these possibilities, we tested the impact of CsrA on sRNA-mRNA regulation for CyaR-yobF, MicL-lpp, Spot 42-fucP, Spot 42-ascF, SgrS-manX (Azam and Vanderpool, 2017), SgrS-ptsG, FnrS-folX, and McaS-csgD. This subset included all sRNAs with multiple confirmed mRNA targets that we identified to interact with CsrA in vitro and had prior evidence of affecting CsrA in vivo (Table 2). It should be noted that we attempted to select the mRNA targets most sensitive to each sRNA regulator in vivo according to literature, if it was characterized (e.g., ptsG for SgrS (Bobrovskyy et al., 2019)). We excluded testing ChiX as we were unsuccessful at constructing reporters for the ChiX-mRNAs citA and chiP to due to the fact the translational fusions were not fluorescent.

To evaluate the effect of CsrA on sRNA-mRNA regulation, we adapted the fluorescence assay described previously (Figure 3A) to determine the impact of sRNA-CsrA interaction on sRNA-mRNA regulation. Briefly, the glgC 5′ UTR was replaced with the 5’ UTR of at least one confirmed mRNA target for each sRNA and constitutively expressed in the ΔcsrBCD csrA::kan strain. Both the sRNA and CsrA were induced on plasmids individually and concurrently to determine the impact of CsrA on sRNA-mRNA regulation.

After confirming that expected sRNA-mRNA repression was detectable for all pairs except CyaR-yobF (Supplementary Figure S7), we noticed that in some cases, the presence of CsrA affected the sRNA’s ability to regulate its cognate mRNA target. We first observed that McaS-csgD repression is significantly enhanced in the presence of CsrA via our plasmid-based system. While it has been shown that McaS sponges CsrA to affect expression of CsrA targets in a CsrA-dependent manner (e.g., pgaA) (Jorgensen et al., 2013), there has been very little work into understanding if CsrA affects the metabolic networks that McaS itself regulates. McaS has been shown to target the csgD and flhD mRNAs, which are master transcriptional activators of curli biogenesis and flagella synthesis (Thomason et al., 2012; Andreassen et al., 2018). From our reporter assay we observed csgD-gfp fluorescence is significantly lower in the presence of both McaS and CsrA, relative to McaS alone (Figure 5A, T-test p-value <0.05). Although csgD-gfp fluorescence increases upon induction of CsrA alone, this change is likely an indirect effect of expressing the global CsrA regulator. When each condition is compared to its respective CsrA null condition (+sRNA–CsrA to–sRNA–CsrA; +sRNA + CsrA to -sRNA + CsrA), repression of the csgD reporter significantly enhanced upon expression of McaS alongside CsrA (Figure 5A, T-test p-value <0.05).

To assess if direct interaction between CsrA and McaS was leading to the measured repression, a mutant Mcas sRNA was constructed with the 5′-most GGA CsrA binding site mutated (Supplementary Table S4). Using the same assay, enhanced repression in the presence of CsrA is significantly reduced when using the McaS mutant sRNA (Supplementary Figure S9A). While alleviation of repression is not fully achieved using the McaS mutant, this may be due to the fact that the second known GGA binding site for CsrA in McaS (Jorgensen et al., 2013) overlaps with the csgD binding site (Figure 5C) and thus not mutated. Importantly, these results suggest that CsrA is in fact remodeling the interaction between McaS and csgD via direct McaS-CsrA binding. This is the first time CsrA has been observed to remodel the network of one of its sponge sRNAs. Differential McaS sRNA accumulation between the presence and absence of CsrA was confirmed to be marginal via Northern blotting analysis (Supplementary Figure S9B). Since these results were measured under sRNA overexpression, we next sought to establish if we could recapitulate these results under more relevant physiological conditions.

Enhanced McaS-csgD repression at physiologically relevant conditions was additionally observed by assaying curli production, a phenotype governed by csgD expression, on Congo Red indicator plates. In accordance with previous work (Thomason et al., 2012), endogenous expression of csgD produces bright red colonies in both wild type and ΔmcaS E. coli strains, as endogenous McaS is likely not expressed under the curli-producing conditions (Figure 5B, Panels I and II). Expression of wild type McaS from a very low copy number plasmid (∼1.2 copies per cell, Material and Methods) produces opaque white colonies, indicating csgD repression (Figure 5B, Panel III–Lower Right). These opaque white colonies were also observed previously (Thomason et al., 2012). These data suggest that a functional McaS interaction with endogenous CsrA enhances McaS repression of csgD and subsequent curli formation. Importantly, expressing an McaS mutant sRNA, with a single CsrA binding site (Jorgenson et al., 2013) mutated, restores the red colony color observed when csgD is not repressed (Figure 5B, Panel IV–Lower Left). While these effects are mild, our results demonstrate that McaS-CsrA interactions play a role in regulating curli biosynthesis under the proper physiological conditions. While we are expressing McaS from a plasmid, due to the low copy number of the plasmid, McaS expression levels are at the same, or even lower, expression levels previously established for physiological conditions (Jorgenson et al., 2013). As a whole, this leads to the notion that the interaction between CsrA-McaS interactions may lead to downstream changes in regulatory outcomes for csgD regulation.

We designed this McaS mutant (5′ GGA:CCA; Supplementary Table S4) based on the hypothesis that direct CsrA binding at the 5′-most GGA in McaS is critical to enhanced csgD repression. Previous in vitro McaS-CsrA binding assays support this notion, as disrupting the 5′-most GGA of McaS impeded CsrA binding more than disrupting the GGA site that overlaps with the csgD binding site. Based on these data, we propose CsrA-McaS binding at the 5′-most GGA motif seeds a bridging interaction that disrupts a downstream hairpin structure containing the csgD binding site (Figure 5D). Such a McaS-CsrA complex could increase accessibility of the McaS binding site for csgD and enhance csgD repression. Alternatively, since McaS-CsrA binding at the 5′-most GGA motif directly overlaps a known McaS-flhD mRNA binding site (Figure 5C), it is plausible that McaS-CsrA interaction may bias partitioning of McaS among its mRNA targets towards csgD rather than flhD. This interaction is under further investigation.

CsrA binds Spot42 at two additional non-GGA sites and is required for Spot42 regulation of its mRNA target fucP under physiologically relevant conditions.

In addition to remodeling McaS-csgD repression, we observed that CsrA significantly impacts Spot 42-fucP regulation in our plasmid-based assay. Specifically, Spot 42-fucP repression was only observed with the fucP-gfp reporter when CsrA expression was induced while apparent significant fucP activation was detected with the fucP-gfp reporter in the absence of CsrA (Figure 6A). Given that for these overexpression reporter assays (Figure 6A), CsrA was expressed on medium copy number plasmids leading to component concentrations 5–15-fold greater than native levels as previously quantified in (Sowa et al., 2017), we next evaluated if CsrA-dependent regulation could be observed under more physiologically relevant conditions.

As a tool to test biological relevance, we required a Spot42 sRNA mutant with abrogated CsrA interactions. We thus mapped the sites on the Spot42 sRNA that contributed to CsrA binding. We first conducted RNase T1 enzymatic probing of CsrA-Spot42 binding (Supplementary Figure S10A) to identify additional potential CsrA binding sites on the Spot42 sRNA, beyond the previously identified one GGA binding site that overlaps one of the RNAse E cleavage sites (Lai et al., 2022). Importantly, Spot 42 only contains one GGA CsrA binding motif, so we anticipated that additional binding sites would correspond to degenerate GGA-like sequences. From our sequencing gel, we identified protection on the 3′ region of the Spot42 sRNA, starting at nucleotide 55, in accordance with the previously identified GGA binding motif (Supplementary Figure S10A). Within the rest of the 3′ end of Spot42, we identified a GGC and GGG motif, at nucleotides 61-63 and 97-99 respectively, that showed protection in the presence of CsrA. With the GGA, GGC, and GGG motifs as guides, we ran EMSAs to evaluate the sites of Spot 42-CsrA interaction. Briefly, mutations were made to eliminate the putative binding motifs without disrupting the predicted secondary structure of the Spot42 sRNA (Supplementary Figure S10D). After testing each of the mutations in combination, we observed an ∼ 1.8-fold reduction in Spot42-CsrA binding only when all three mutations were present in the sRNA (Supplementary Figures 10B, C). As such, we conclude CsrA binds Spot 42 using the established GGA site, as well as the two novel degenerate GGN sites on the 3’ end of Spot42. We anticipate that a preferred interaction between Spot 42 and one binding pocket of a CsrA dimer at the GGA site likely tethers a dimer to the sRNA and allows for an additional contact to be made at a lower affinity GGN site by the other binding pocket. We expect that the single stranded GGG site is more likely bound alongside the GGA site, given its greater accessibility than the hairpin contained GGC site (Supplementary Figure S10D). Such a model has been previously proposed for CsrA-RNA binding (Mercante et al., 2009) and likely also explains CsrA repression of the hfq transcript, which contains only a single GGA site but presents a CsrA-occluded AGA site in in vitro footprinting studies (Baker et al., 2007; Leistra, Gelderman, et al., 2018). Moreover, degenerate GGA-like sequences (e.g., AG-rich or GGN sequences) are thought to contribute to CsrA-RNA binding and regulation in some target mRNAs (i.e., glgC, csrA, and pgaA), although these targets also contain at least two instances of the preferred GGA CsrA binding site (Mercante et al., 2009; X. Wang et al., 2005; H. Yakhnin, Yakhnin, et al., 2011).

Using this information, we designed a strain with Spot42 containing all three mutations to abrogate Spot42-CsrA interactions in vivo, a Δspf::spf_triple_GGNs mutant (referred to as Spot42 Triple Mutant hereafter). To directly evaluate CsrA-Spot42 interactions upon the regulatory Spot42-fucP activity, we measured regulation of the fucP-gfp translational reporter expressed from a low copy number plasmid (Materials and Methods) in wild type E. coli, the Spot42 triple mutant strain and a Δspf strain. We induced endogenous Spot 42 transcription from the genome at an OD₆₀₀ of 0.2 by addition of 0.2% glucose to M9 media supplemented with 0.4% glycerol and casamino acids, alleviating CRP-Spot 42 transcriptional repression, as shown in previous literature (Beisel and Storz, 2011). From this assay, we observed that the Spot42 repression of the fucP-gfp reporter is significantly alleviated in the Spot42 Triple Mutant expressing strain relative to the wild type Spot42 strain (Figure 6B). Additionally, to confirm that the changes in regulation we observed was due to changes interaction and not due to changes in sRNA stability, we ran a northern blot and confirmed that both the wild type and Triple Mutant Spot42 sRNAs were expressed at the same level (Supplementary Figure S14). These results point to the fact that Spot42 repression of the fucP-gfp is in fact partially CsrA dependent under physiologically relevant conditions. Interestingly, we could not alleviate the fucP-gfp reporter signal to that of the Δspf in the Spot42 Triple Mutant strain. This is likely due to network interactions other than CsrA, such as residual Hfq-driven repression of the fucP-gfp target by Spot42 (Beisel and Storz, 2011).

Lastly to understand if we could observe the CsrA-dependent regulation of the fucP-gfp reporter by Spot42 occurred under physiological conditions, we modified the reporter assay from Figure 6A such that it could be recapitulated at genomically relevant CsrA concentrations. As such, we chose to use a Δspf and ΔcsrA::kanΔspf strain of E. coli to replicate the plus and minus CsrA conditions and expressed Spot42 from the low copy plasmid mentioned earlier. This was chosen, as Spot42 cannot be reliably expressed from the genome in a ΔcsrA::kan strain (Beisel and Storz, 2011). Using this adapted assay, we observed that repression of the fucP-gfp reporter by Spot42 only occurred when CsrA was present. (Figure 6D). While repression of the fucP-gfp reporter was also observed in the absence of Spot42 when CsrA was present, in vitro Transcription-Translation (IVTT) assays failed to observe similar repressive effects in CsrA-only conditions (Supplementary Figure S11D). This suggests that the in vivo repressive effects are likely due to redundant regulation by other CsrA-regulated genes. There was no change in fucP-gfp reporter signal in the ΔcsrA::kan, regardless of Spot42 expression. Also to note, while it appears that CsrA is repressing fucP-gfp in the absence of Spot42 (Figure 6D – 1st and 3rd columns), this is due to the fact that the fluorescence values are normalized by growth (e.g., OD₆₀₀) and the -CsrA condition is derived from the ΔcsrA::kanΔspf strain, which exhibits poorer growth (e.g., a lower final OD₆₀₀). With these data in mind, we demonstrate that under physiological conditions, Spot42 repression of the fucP-gfp transcript is CsrA-dependent. Furthermore, IVTT results show significant repression of fucP by Spot42 with increasing concentrations of CsrA, supporting the fact that CsrA is needed for Spot42 regulation of fucP (Supplementary Figure S11D). Additionally, we observed in specific experimental conditions, such as without Hfq and CsrA regulation, Spot42 can activate translation of the fucP-gfp transcript (Supplementary Figures 11D, 12C). Spot42 can also activate the fucP-gfp transcript when the wild type fucP binding region on the Spot42 sRNA (Figure 6C–red outlined nucleotides) was mutated (Supplementary Figure S11B). Thus, there may be non-native or engineerable conditions in which fucP activation can occur. Lastly, while this work was being prepared a destabilizing effect of Spot42 in a csrA deletion strain of E. coli was published (Lai et al., 2022). This work is the second known demonstration of CsrA-dependent regulation of an mRNA target via Spot42.

All E. coli strains used in experiments are derivatives of K-12 MG1655 and are described in Supplementary Tables S1–S4. DH5α (NEB) and BL21 (DE3) (NEB) E. coli strains were used for cloning and protein purification, respectively. Routine cultures were performed in LB medium (Miller) (BD Biosciences) with 50 μg/mL kanamycin and/or 100 μg/mL carbenicillin as required. Starter cultures were grown from single colonies overnight in 5 mL LB (test tubes) at 37°C with 200 rpm orbital shaking. Cultures were diluted 1:100 in LB media for fluorescence assays and protein purification except where noted otherwise. Final concentrations of 1 mM isopropyl ß-D-1-thiogalactopyranoside (IPTG) and 100 ng/mL anhydrotetracycline (ATc) were used in cultures to induce plasmid expression. Congo red plate assays were performed on agar plates containing LB medium without salt (10 g/L yeast extract, 5 g/L tryptone, 0 g/L NaCl), 40 μg/mL Congo red dye, 20 μg/mL Coomassie Brillant Blue G 250 dye, 50 μg/mL kanamycin, and 1 mM IPTG (Bak et al., 2015). Kanamycin resistance cassettes were cured from sRNA::kan deletion strains (Supplementary Tables S1–S4) and confirmed by colony PCR prior to use as previously described (Cherepanov and Wackernagel, 1995; Datsenko and Wanner, 2000). Genomic expression conditions for SgrS and FnrS were adapted directly from previous established protocols for expressing SgrS (Wadler and Vanderpool, 2007 PNAS) and FnrS (Durand and Storz, 2010 Mol. Microbiol.).

Plasmids used in this study are documented in Supplementary Tables S1–S4. All plasmids were constructed by Gibson assembly or purchased from Genscript. Oligonucleotide primers for Gibson assembly were designed using the NEBuilder 2.0 web tool and are included in Supplementary Tables S2–S6. DNA oligonucleotides and double stranded DNA fragments (GBlocks) were purchased from Integrated DNA Technologies. Plasmids were verified by Sanger sequencing (University of Texas GSAF core). Details on cloning methods used for construction of pTriFC plasmids, pHL600 pLacO-sRNA pTetO-csrA plasmids, pHL1756 5′UTR-gfp reporter plasmids, pBTRK-pLacO-sRNA plasmids, and pBTRK-pLacO-fucP-gfp (CmR) plasmids are included in Supplementary Material. It should be noted that the pBTRK-pTrc-Empty plasmid, a generous gift from the lab of Dr. Brian Pfleger that contains a very low copy number pBBR-1 origin (∼1-3 copies per cell) (Youngquist et al., 2013; Hernández Lozada et al., 2018), was used as a parent plasmid for the pBTRK-pLacO plasmids constructed in this work.

A previously-demonstrated CRISPR-cas9 genome modification protocol was used to construct Δspf and GGA:GCA mutant spf K-12 MG1655 strains (Mehrer et al., 2018). It should be noted that the spf gene encodes the Spot 42 sRNA. This approach uses a two-plasmid system to achieve genomic deletion or insertion. The first plasmid, a generous gift from the lab of Dr. Brian Pfleger, provides kanamycin-resistant guide RNA (gRNA) expression under the constitutive J23119 promoter and is termed pgRNA. The second plasmid, pMP11, is a temperature sensitive, carbenicillin-resistant plasmid that contains 1) wild type cas9 from Streptococcus pyogenes under the J23119 constitutive promoter, 2) λ-red recombinase under the pBAD promoter, and 3) a gRNA specific to the pgRNA plasmid to enable its removal under a pTeto promoter.

A gRNA for deletion of the spf gene was designed using the CRISPR gRNA Design webtool from Atum (atum.bio). This sequence was cloned into the gRNA plasmid by Gibson Assembly (Supplementary Tables S2–S6). Additionally, a dsDNA gBlock was designed to contain the desired spf deletion (here, the last 7 nt of the sRNA) or mutation sequence plus 500 base pairs of sequence homologous to the genome upstream and downstream of the edit site. To avoid gRNA-directed cleavage of the dsDNA homology GBlock, we targeted the gRNA to the CDS of a protein-coding gene, yihA, 150 nucleotides 3′ of spf. Synonymous codon mutations were made to the gRNA binding site in the dsDNA homology GBlock to protect it from pgRNA-directed cleavage. As a result, the 500 nucleotide homology regions were defined to be 500 nucleotides 5′ of spf and 3’ of yihA (Sequence in Supplementary Tables S2–S6).

K-12 MG1655 E. coli containing the pMP11 plasmid was grown in 5 mL of LB overnight to saturation. The overnight culture was diluted 1:100 in 50 mL of SOB media (BD Biosciences) and grown at 30°C until an _OD600 of 0.2. The ?-red genes were then induced by addition of 1% w/v L-arabinose. At an OD₆₀₀ of 0.4–0.6, the culture was made electrocompetent according to common lab procedures. 50 ng of the gRNA plasmid and 25–50 ng of the dsDNA homology block was added to a single aliquot of electrocompetent cells, and subsequently transformed. Transformants were recovered in 1 mL of SOB media at 30 °C for at least 3 hours, plated onto LB agar plates containing kanamycin and carbenicillin, and grown at 30 °C. Gene deletions were confirmed via colony PCR (Supplementary Tables S2–S6) and sanger sequencing of the purified pcr fragment.

Upon verification of the spf deletion or mutation, the gRNA plasmid was cured by growing a single colony in 5 mL of LB containing carbenicillin and 0.2 ng/mL aTc overnight at 30 °C. Single colonies were of this culture were generated and streaked on carbenicillin versus kanamycin and carbenicillin LB agar plates to confirm pgRNA plasmid removal. The pMP11 plasmid was then cured by patching successful colonies onto LB agar only plates and growing overnight at 42°C. Removal of pMP11 was verified by streaking on LB agar only versus LB agar plates with carbenicillin.

RNA sequencing and downstream data analysis was previously performed for csrA::kan mutant, ΔcsrB ΔcsrC, and wild type (WT) K-12 MG1655 E. coli strains before and after glucose deprivation in M9 minimal media (Sowa et al., 2017). Briefly, biological triplicate cultures were grown to an OD₆₀₀ of 0.6 in M9 media with 0.2% glucose before centrifugation and resuspension in fresh M9 media lacking glucose. Samples were taken 10 min and 0 min prior and 30 min, 60 min, 180 min, and 300 min after glucose starvation. Aligned, counted reads from the previous work were reanalyzed by DESeq2 to include sRNAs and tRNAs (rRNA read counts were still excluded from analysis). As done in the prior work, all samples of the ΔcsrB ΔcsrC strain collected 300 min post stress were excluded from further analysis due to large variation amongst the replicas. Additionally, principal component analysis identified one sample, a replica of the of ΔcsrB ΔcsrC strain collected at time 0, that clustered far from the rest of the data. Data from this sample were also excluded from further analysis. Strain and time factors were paired in order to identify differentially expressed genes between the WT and csrA::kan strains and the WT and ΔcsrB ΔcsrC strains at each time point. sRNAs differentially expressed (log2 fold change >1 or < −1 and Wald test P-adjusted <0.05) between the WT and ΔcsrB ΔcsrC strains at two time points or more were considered as potentially interacting with the Csr system and selected for further study (Supplementary Tables S3–S7). It should be noted that by these criteria, no sRNAs were differentially expressed between the WT and csrA::kan strains. As levels of free CsrA are likely increased in the ΔcsrB ΔcsrC strain (Sowa et al., 2017), we considered differential expression between the WT and ΔcsrB ΔcsrC strains to indicate sRNAs likely impacted by CsrA.

The biophysical model estimates the free energy of CsrA binding to pairs of 5 nucleotide potential binding sites within an RNA of interest based on RNA sequence, structure, and inter-site spacing. Based on these predictions, we analyzed the 15 most likely, i.e., lowest free energy, CsrA-RNA binding conformations, to identify putative CsrA-RNA binding regions. Previously, we confirmed the utility of this model for correctly predicting CsrA binding sites for 6 mRNAs of 8 total that have documented CsrA footprints. Within this subset, the model also successfully captured CsrA footprints for 3 mRNAs that lack the preferred GGA motif. When the model was run for sRNAs, predicted CsrA binding sites overlap GGA motifs found within the sRNA sequences (Supplementary Tables S3, S4). They also highlight non-GGA containing regions that may contribute, albeit likely transiently, to sRNA-CsrA recognition and binding in these sRNAs. Importantly, predicted sRNA-CsrA binding sites in the FnrS and MicL sRNAs, which lack GGA motifs, highlight a distinct region in each sRNA that contains a bulk of the predicted binding sites. CsrA-sRNA binding sites were predicted as previously done for mRNAs (Leistra et al., 2018) and are further explained in the Supplementary Material.

CsrA was purified as previously published (Dubey et al., 2005) with minor modifications, as described in the Supplementary Material. Hfq was purified according to published methods (Santiago-Frangos et al., 2016) and generously gifted to us by the lab of Dr. Sarah Woodson.

Primers were designed to amplify sRNAs (Supplementary Tables S2–S6) with an upstream T7 promoter from genomic K-12 MG1655 DNA (wild type sRNA sequences) or previously constructed plasmids (mutant sRNA sequences). Forward primers were designed to contain four random nucleotides (GACT) upstream of the promoter sequence (TAA​TAC​GAC​TCA​CTA​TAG​GGA​GA). PCR product sizes were verified on a 1% agarose gel prior to PCR clean up (DNA Clean and Concentrator-5, Zymo). A DNA fragment for transcription of lpp (lpp sequence preceded by aforementioned T7 transcription-enabling sequence) was manufactured by Integrated DNA Technologies (Supplementary Tables S2–S6).

Lpp was in vitro transcribed using the MEGA Script IVT Kit (Thermo Fisher Scientific) according to manufacturer instructions (with incubation time extended to 6 h as recommended for <500bp RNA products). sRNAs were transcribed with modifications to enable internal P-32 labeling. Specifically, 1.5 µL UTP [α-32P] (3000 Ci/mmol 10 mCi/mL, 500 μCi, PerkinElmer) was used to replace equivalent unlabeled UTP from the MEGA Script IVT kit. Following MEGA Script T7 DNase digestion, RNA recovery was performed using RNA Clean and Concentrator-5 (Zymo Research) according to manufacturer instructions. RNA samples were re-suspended in 20 µL of nuclease-free water following recovery and, for sRNAs only, free NTPs removed using DTR Gel Filtration Cartridges (EdgeBio) following manufacturer instructions. RNA concentration was measured via spectrophotometry, and, in the case of lpp, transcript quality validated on an 8% urea gel, run at 100 V for 3 h, and stained with Sybr Green II (Thermo Fisher Scientific).

EMSAs were performed largely according to the TBE CsrA-RNA gel shift protocol detailed in (A. V Yakhnin et al., 2012), with modifications to support the use of a different CsrA dilution buffer (Dubey et al., 2005). Twelve µL binding reactions were comprised of 10 nM denatured sRNA (3 min at 85°C), 1 µL 10X CsrA binding buffer (100 mM Tris-HCl pH 7.5, 100 mM MgCl₂, 1 M KCl), 4 µL CsrA dilution buffer (same as CsrA storage buffer: 10 mM Tris-HCl, 100 mM KCl, 10 mM MgCl₂, and 25% glycerol, pH 7.0), 11% glycerol, 20 mM DTT, 333 U/mL RNase Inhibitor, Murine (NEB), 1.5 μg/μL yeast total RNA, 0–6 μM CsrA, and 0, 0.375, 1.5, or 3 μM Hfq. These concentrations of Hfq were selected to mimic the expected relative in vivo expression ratios of CsrA to Hfq based on estimations of their in vivo concentrations: 0.4–10 µM hexameric Hfq per cell (Kajitani et al., 1994; Moon and Gottesman, 2011; Wagner, 2013) and 6–17 µM dimeric CsrA (Romeo et al., 2013). Importantly, each reaction contained an excess of total yeast RNA (Thermo Scientific) to exclude results due to non-specific binding. For these experiments, radiolabeled sRNAs (10 nM) were incubated with 3 µM or 6 µM concentrations of CsrA at 37°C for 30 min prior to loading and running on a 10% non-denaturing polyacrylamide gel (A. V Yakhnin et al., 2012) with 0.5X TBE running buffer (IBI Scientific, 10x composition: 89 mM Tris, 89 mM Boric Acid, 2 mM EDTA) at 170 V between 5 h and overnight (depending on sRNA) at 4°C. These concentrations of CsrA represent Protein:sRNA ratios of 300:1 and 600:1, respectively. These ratios were selected to screen for CsrA-sRNA binding based on CsrA binding affinities reported for Spot 42, GadY, GcvB, and MicL, interactions presumed to be relevant in vivo based on significant enrichment in CsrA CLIP-seq data (Potts et al., 2017). Briefly, these previous in vitro binding assays employed molar ratios of CsrA:sRNA ranging from 50:1 to 2,000:1 (5 nM CsrA:0.1 nM sRNA to 200 nM CsrA:0.1 nM sRNA) and determined dissociation constants in the range of 100:1 to 660:1 CsrA:sRNA (10 nM CsrA:0.1 nM sRNA to 66 nM CsrA:0.1 nM sRNA). While our assays scale up the total molar amount of each component, they test CsrA-sRNA binding at molar ratios of 300:1 and 600:1 CsrA:sRNA. These ratios are within the range of CsrA-sRNA binding affinity presumed to be physiologically relevant in (Potts et al., 2017). Additionally, all binding assays are performed in large excess of yeast total RNA (88 ng/μL, as in (A. V Yakhnin et al., 2012)) to inhibit non-specific association of CsrA to labeled sRNAs.

Gel exposure on a bioWORLD bioExposure cassette was phosphor-imaged at 1000 V using Typhoon FLA 700 (GE Health Life Science).

MicL-lpp-CsrA EMSA was performed with slight modifications. Larger reaction volumes (15 uL) were used to accommodate addition of the lpp 5’ UTR (80 or 240 nM final concentration) and relative molar amounts of sRNA and CsrA equivalent with the gels described above were included (8 nM denatured sRNA and 0, 2.4, or 4.8 μM CsrA). All other components of the binding reaction were scaled appropriately.

Fluorescence complementation assays to detect RNA-protein binding in vivo were conducted as previously described, with modifications as described in the Supplementary Material.