CsrA Enhances Cyclic-di-GMP Biosynthesis and Yersinia pestis Biofilm Blockage of the Flea Foregut by Alleviating Hfq-Dependent Repression of the hmsT mRNA

ABSTRACT Plague-causing Yersinia pestis is transmitted through regurgitation when it forms a biofilm-mediated blockage in the foregut of its flea vector. This biofilm is composed of an extracellular polysaccharide substance (EPS) produced when cyclic-di-GMP (c-di-GMP) levels are elevated. The Y. pestis diguanylate cyclase enzymes HmsD and HmsT synthesize c-di-GMP. HmsD is required for biofilm blockage formation but contributes minimally to in vitro biofilms. HmsT, however, is necessary for in vitro biofilms and contributes to intermediate rates of biofilm blockage. C-di-GMP synthesis is regulated at the transcriptional and posttranscriptional levels. In this, the global RNA chaperone, Hfq, posttranscriptionally represses hmsT mRNA translation. How c-di-GMP levels and biofilm blockage formation is modulated by nutritional stimuli encountered in the flea gut is unknown. Here, the RNA-binding regulator protein CsrA, which controls c-di-GMP-mediated biofilm formation and central carbon metabolism responses in many Gammaproteobacteria, was assessed for its role in Y. pestis biofilm formation. We determined that CsrA was required for markedly greater c-di-GMP and EPS levels when Y. pestis was cultivated on alternative sugars implicated in flea biofilm blockage metabolism. Our assays, composed of mobility shifts, quantification of mRNA translation, stability, and abundance, and epistasis analyses of a csrA hfq double mutant strain substantiated that CsrA represses hfq mRNA translation, thereby alleviating Hfq-dependent repression of hmsT mRNA translation. Additionally, a csrA mutant exhibited intermediately reduced biofilm blockage rates, resembling an hmsT mutant. Hence, we reveal CsrA-mediated control of c-di-GMP synthesis in Y. pestis as a tiered, posttranscriptional regulatory process that enhances biofilm blockage-mediated transmission from fleas.

sources (e.g., K-gluconate) versus the primary carbon source glucose (44). The alternative sugars ribose and galactose appear to be primarily catabolized by Y. pestis during flea blockage (4,5). Hence, we tested if biofilm EPS formation in a csrA mutant is more drastically impaired in these biologically relevant sugars versus glucose. An assay based on the specificity of Congo red (CR) dye to polysaccharides (17,45) was used to allow direct comparison of EPS production from different carbon sources through normalization by bacterial biomass. The assay media were HIB, a rich routine culture medium, and the chemically defined medium TMH (46), supplemented with either glucose (TMH-glu), ribose (TMH-rib), or galactose (TMH-gal).
A csrA mutant (DcsrA) generated previously (44) in the avirulent epidemic KIM61 strain background, with an isogenic wild-type (WT) parent strain and a cis-complemented csrA mutant strain (DcsrA::csrA), were tested for EPS production. EPS production by the WT strain in HIB was .10-fold less than that in TMH medium regardless of the carbon source (Fig. 1A). EPS production in the WT strain was significantly lower in TMH-glu versus TMH-gal and TMH-rib. Under all conditions, a DhmsR strain, used as a negative control as it is unable to produce EPS, displayed little to no CR binding. Compared to the WT strain, the DcsrA strain had significantly reduced EPS levels under all conditions, a phenotype that was restored in the DcsrA::csrA strain. Mean reduction in EPS levels for the DcsrA strain relative to the WT strain were 24% and 40% in HIB and TMH-glu and 87% and 81% in TMH-rib and TMH-gal, respectively.
To assess if EPS levels in the DcsrA strain correlated with intracellular c-di-GMP pools, we quantified c-di-GMP in strains grown in TMH-glu, TMH-gal, and TMH-rib. The c-di-GMP levels of the WT strain ranged between 7.6 and 10.4, 63.4 to 107.8, and 63.6 to 117.8 pmol/mg cell weight in TMH-glu, TMH-gal, and TMH-rib, respectively (Fig. 1B). Under all conditions, a DhmsT strain, dysfunctional in c-di-GMP synthesis (19), produced little to no c-di-GMP, as expected (18,19,22). Compared to the WT strain, the DcsrA strain had 2.7-, 8.2-, and 42-fold mean reduction in c-di-GMP levels in TMH-glu, TMH-gal, and TMH-rib, respectively. The DcsrA::csrA strain exhibited c-di-GMP levels within ranges displayed by the WT strain. Therefore, biofilm and c-di-GMP production in TMH-gal and TMH-rib was highly dependent on functional CsrA. However, TMH-gal was selected for the next experiments to allow for comparison with published studies (47).
CsrA promotes translation of the hmsT mRNA. CsrA alters translation rates of mRNAs encoding enzymes for c-di-GMP synthesis or degradation, thereby altering c-di-GMP and EPS production in other bacteria (25,35,42,43). Willias et al. (44) proposed that CsrA targets the hmsP and/or hmsT mRNAs to reduce c-di-GMP levels in the DcsrA strain and identified putative CsrA binding sites in the 59 UTRs of these transcripts. Our experiments described above support this idea, since we showed reduced c-di-GMP levels in the DcsrA strain.
To determine if translation of the hmsP and hmsT mRNAs are CsrA dependent, we constructed posttranscriptional green fluorescent protein (GFP) fusion reporters. The 59 UTR plus predicted CsrA binding motifs (44) were engineered in-frame to the coding sequence (CDS) of gfpmut3.1 ( Fig. 2A) and an inducible promoter, P tetO . The use of the P tetO promoter was intended to uncouple transcription from control by growth phases, environmental signals, or transcription factors. The 59 UTR of flhDC was used as a positive control because it is a validated target of CsrA in Y. pseudotuberculosis (48) and shares 100% nucleotide identity with the 59 UTR of Y. pestis flhDC mRNA. CsrA is identical between Y. pestis and Y. pseudotuberculosis; thus, Y. pestis CsrA was expected to bind the flhDC transcript. The 59 UTR of housekeeping gene gyrB that lacks CsrA binding motifs was used as a negative control.
GFP reporter fusion constructs were transformed into the WT and DcsrA strains. Strains were grown in TMH-gal and fluorescence recorded at 3 h postinduction, when the greatest GFP induction was achieved for each construct (data not shown). As expected, GFP expression between the WT and DcsrA strains from the gyrB-gfp reporter was comparable but was significantly reduced in the DcsrA strain with the flhDC-gfp reporter (Fig. 2B). No significant difference in GFP expression between the WT and DcsrA strains was noted for the hmsP-gfp reporter. However, significant reduction in GFP expression occurred in the hmsT-gfp reporter in the DcsrA strain compared to the WT strain, suggesting that the hmsT mRNA translation was CsrA dependent.
Next, to determine if decreased hmsT mRNA translational levels in the DcsrA strain are a result of decreased hmsT mRNA levels, we evaluated steady-state levels of the hmsT mRNA in the WT and DcsrA strains using reverse transcription-quantitative PCR (RT-qPCR). We included an evaluation of the hmsP mRNA as a negative control, as hmsP mRNA translational levels were unaffected by CsrA. The relative steadystate mRNA levels of the hmsT mRNA were significantly lower in the DcsrA strain than the WT strain, while the hmsP mRNA levels were similar between these strains (Fig. 2C). These data suggested that hmsT mRNA abundance was CsrA dependent. FIG 1 CsrA is required for EPS production and c-di-GMP synthesis. (A) Congo red (CR) binding assays were used to quantify EPS production of strains cultured in HIB or TMH supplemented with 0.2% of glucose (TMH-glu), galactose (TMH-gal), or ribose (TMH-rib). Error bars represent the means 6 standard errors of the means (SEM) of bound CR samples from 3 to 6 independent experiments. (B) C-di-GMP was extracted from strains grown in TMH-glu, TMH-gal, or TMH-rib. Means of two independent experiments are shown. Statistical significance was determined using one-way analysis of variance (ANOVA) with Dunnett's multiple comparisons posttest for each medium type (black lines) or to compare the WT strain across media (gray lines) (*, P , 0.05; ***, P, 0.0005; ****, P , 0.0001; n.s., not significant).
CsrA does not bind directly to the hmsT mRNA. To determine if CsrA regulation of the hmsT mRNA results from direct binding of CsrA, RNA electrophoretic mobility shift assays (REMSAs) were conducted. A transcript of the hmsT mRNA containing the same region as that in the hmsT-gfp reporter fusion was used to examine the interaction of the hmsT mRNA 59 UTR and CsrA. The hns mRNA that does not bind to CsrA was used as a negative control, and the 59 UTR of flhDC served as a positive control (48). As previously reported (48), the labeled hns probe did not shift with increasing concentrations of CsrA-His 6 (Fig. 2D). As expected, a shift occurred for the labeled flhDC probe at increasing concentrations of CsrA-His 6 (Fig. 2E). No shift was seen for the labeled hmsT probe with increasing concentrations of CsrA-His 6 ( Fig. 2F), indicating that CsrA was not able to bind directly to the hmsT mRNA. These results strongly suggested that CsrA indirectly regulated hmsT mRNA translation.
CsrA binds specifically to the hfq mRNA. The mRNAs of known negative regulators of hmsT, RcsB or Hfq, may instead be targeted by CsrA. Indeed, CsrA orthologs of the plant pathogen Erwinia amylovora and E. coli repress mRNAs of Y. pestis orthologs of rcsB (33) and hfq (49), respectively. Both rcsB and hfq genes are part of polycistronic operons (33,(49)(50)(51). CsrA can target downstream genes within polycistronic operons by binding to CsrA binding motifs located in the 39 end of the upstream gene (52). Therefore, to determine if Y. pestis rcsB and hfq mRNAs are candidate mRNA targets of CsrA, a position matrix scan (44) was applied to the 59 UTR of rcsB and hfq mRNAs to identify putative CsrA binding sites. Up to 2250 bases upstream from the translational initiation codon were queried, in keeping with ranges for to-date validated CsrA targeted 59 UTRs (36,53,54). Two potential binding sites were found in the hfq transcript ( Fig. 3A; see also Fig. S1 in the supplemental material); one overlapped the Shine-Dalgarno (SD) sequence of the hfq mRNA (BS1) and the other occurred in a stem-loop, spanning the 2151 to 2159 nucleotide sequence located at the 39 end of the upstream gene, miaA (BS2). The rcsB leader region contained no potential CsrA binding sites (data not shown). To experimentally test if hfq mRNA translation was CsrA dependent, we first generated an hfq-gfp posttranscriptional fusion reporter containing both putative CsrA binding sites. The GFP reporter expression from the hfq-gfp reporter in TMH-gal medium was compared between the WT and DcsrA strains (Fig. 3B). Expression was significantly higher in the DcsrA strain, suggesting that CsrA repressed hfq mRNA translation.
To determine if CsrA binds directly to the hfq mRNA, REMSAs were conducted. A minor shift in the labeled hfq probe was noted at 75 nM CsrA-His 6 , but at 150 nM CsrA-His 6 the entire complex had shifted (Fig. 4A). Competitive binding assays verified that CsrA binding to the hfq mRNA was specific because, when competed with 2-and 10fold excess unlabeled hfq probe, the labeled hfq probe shifted only partially or not at all, respectively (Fig. 4B).
To determine if CsrA binds to the identified GGA sites, we generated labeled probes with CC substitutions for the GG nucleotides in the GGA motif for BS1 (hfq BS1 mutant), CCC substitutions for the GGA motif at BS2 (hfq BS2 mutant) alone, and the respective aforementioned substitutions at both sites (hfq BS1/BS2 mutant). All mutant probes showed reduced binding to CsrA-His 6 but at varied levels ( Fig. 4C to E). For the hfq BS1 mutant probe, a small effect on binding to CsrA-His 6 was noted by the presence of two complexes, one that had not shifted and the other that had shifted at $112 nM CsrA-His 6 (Fig. 4C). The hfq BS2 mutant probe (Fig. 4D) and hfq BS1/BS2 mutant probe Posttranscriptional fusion reporter plasmids composing the upstream sequence and the first 9 codons of hfq fused to gfpmut3.1 (gfp), and the P tetO promoter was used in the WT and DcsrA strains. Error bars represent mean 6 SEM relative RFU/OD 600 from four independent experiments.
( Fig. 4E) both exhibited a complete loss of binding to CsrA-His 6 , because no shift was noted. Y. pestis hfq mRNA can also be transcribed from a promoter immediately upstream of its translational start codon. In this case, the hfq transcript composes a 90nucleotide 59 UTR sequence with only BS1 present (50). Additionally, a shorter GFP reporter fusion construct representative of this shorter 59 UTR (2115 through 127 nucleotides in relation to the ATG start) had increased GFP expression in the DcsrA strain compared to that of the WT strain (Fig. S2). Thus, BS1 and BS2 are authentic CsrA binding sites.
CsrA blocks translation but does not accelerate the decay rate of the hfq mRNA. CsrA may decrease hfq mRNA translation by accelerating its decay rate, blocking its translation by preventing ribosome binding, or by facilitating premature transcription termination (55). To test if CsrA alters stability of the hfq mRNA, we determined the half-life of this transcript in the WT and DcsrA strains after addition of rifampin to prevent transcription initiation. The mean (6 standard deviation [SD]) half-life of the hfq mRNA ( Fig. 5A) was similar between the WT (12 6 2.0 min) and DcsrA (11.3 6 1.3 min) strains. Therefore, CsrA binding does not destabilize the hfq mRNA.
To determine if hfq mRNA translational repression results from CsrA binding and preventing translation, we performed in vitro cell-free translational assays. We initially utilized an mRNA template containing the full-length sequence of the hfq gene engineered with a Flag tag preceding the stop codon. However, similar to a previous report using a full-length E. coli hfq gene template in cell-free translational assays (49), multiple bands were observed on immunoblots due to incomplete denaturation of the hexameric Hfq protein. Hence, mRNA templates derived from the hfq-gfp and hmsT-gfp (negative control) reporters described above were used. Similar amounts of the HmsT-GFP protein were synthesized in the presence and absence of CsrA (Fig. 5B), as expected if CsrA did not directly bind and alter translational levels of the HmsT mRNA. However, translation of the hfq-gfp transcript was inhibited in the presence of CsrA (Fig. 5B), as reflected by the 2.5 (6SD, 0.67) greater levels of Hfq-GFP protein in the absence of CsrA relative to its presence. This outcome supported that CsrA binds directly to and prevents translation of the hfq mRNA.
CsrA represses hfq mRNA translation, facilitating derepression of hmsT mRNA translation. Y. pestis Hfq posttranscriptionally represses hmsT mRNA translation by accelerating its decay rate (22,23). Therefore, the hmsT mRNA half-life should be shorter in a DcsrA strain if CsrA no longer represses the hfq mRNA. To determine the mRNA stability of the hmsT mRNA in the DcsrA strain versus the WT strain, we determined the half-life of the hmsT mRNA as described above. As predicted, the mean 6 We reasoned further that if CsrA represses hfq mRNA translation to facilitate derepression of hmsT mRNA translation, then a DcsrA Dhfq strain should show reestablishment of hmsT mRNA levels and increased biofilm production. To determine if this occurs, we generated a DcsrA Dhfq double mutant strain. A complemented derivative thereof, DcsrA Dhfq (pLGhfq) strain, was also constructed by inserting the hfq gene and native promoter sequence on a low-copy-number plasmid, pLG338, to create pLGhfq. First, steady-state levels of the hmsT mRNA at log phase were evaluated in the DcsrA Dhfq and DcsrA Dhfq (pLGhfq) strains compared to the DcsrA and WT strains in TMHgal (Fig. 6B). As hypothesized, the DcsrA Dhfq strain exhibited restored levels of the hmsT mRNA comparable to the WT strain. Additionally, the DcsrA Dhfq (pLGhfq) complemented strain showed hmsT levels comparable to the DcsrA strain and significantly lower than that of the WT strain.
During routine lab culture, Y. pestis incurs spontaneous loss of a 102-kb locus, referred to as the pigmentation locus (56), or Pgm locus, named for the ability to form pigmented colonies on CR-supplemented agar (57). The hmsHFRS operon is located within the pgm locus and confers this phenotype. When culturing our DcsrA Dhfq strain on media that promote high biofilm production, we noted that the strain had a high propensity to form nonpigmented colonies, reflecting loss of EPS production. Therefore, to quantify EPS levels in these strains, CR assays were performed using LB medium that does not support high levels of biofilm production. Nonetheless, EPS levels in the DcsrA Dhfq strain were not different from those of the WT strain (Fig. 6C), and the DcsrA Dhfq (pLGhfq) complemented strain showed EPS levels comparable to those The WT and DcsrA strains were grown to log phase in TMH-gal, and RNA was isolated from samples taken at 0 min (prerifampin) and after the addition of 400 mg/ml rifampin. Relative hfq mRNA levels remaining at the indicated time points were quantified using RT-qPCR. The amount of hfq mRNA in each strain at 0 min relative to rifampin addition was set to 100%. The percent mRNA remaining thereafter was plotted versus time as semilogarithmic graphs. The mean 6 SEM percent mRNA from four independent experiments is shown. (B) In vitro translational assays were performed with the PURExpress kit using translational fusions transcripts of hfq-gfp and hmsT-gfp (negative control) expressed from a T7 promoter. The mean 6 SD fold change in HmsT-GFP and Hfq-GFP signal between samples in the presence or absence of CsrA was derived from three technical replicates of the immunodot blot. One representative dot blot is shown. of the DcsrA strain but significantly lower than that of the WT strain (Fig. 6C). Additionally, we analyzed the pigmentation phenotypes of the strains on LB agar supplemented with CR. Pigmentation phenotypes matched the quantitative CR binding assay data. Thus, as hypothesized, EPS and hmsT mRNA steady-state levels in the DcsrA Dhfq strain and DcsrA Dhfq (pLGhfq) complemented strain corresponded with that of the WT and DcsrA strains, respectively.
CsrA is required for the robust formation of biofilm-mediated blockage in fleas. Finally, to define the role of CsrA in biofilm-mediated flea foregut blockage, we compared cumulative blockage rates in cohorts of rat fleas infected with WT, DcsrA, and DcsrA::csrA strains over a 28-day period (Fig. 7A). Fleas infected with the DcsrA strain achieved significantly lower cumulative blockage rates of 16.8% (6SD, 3.2) relative to the WT strain-infected fleas, which achieved rates of 39.8% (6SD, 10.4). Similar to the WT strain-infected fleas, the DcsrA::csrA strain-infected fleas exhibited rates of 34.2% (6SD, 11.7). An analysis of temporal incidence of blockage showed that while fleas infected with the WT and the DcsrA::csrA strains attained peak blockage incidence at ;15 days postinfection (dpi), the DcsrA strain-infected fleas exhibited generally low blockage incidence, which was significant at 15 and 19 dpi (Fig. 7B). Flea bacterial loads and flea infection rates were not significantly different among the strains (Fig. 7C and D) despite being slightly lower in the DcsrA-infected fleas. Therefore, growth kinetics of the DcsrA strain in fleas was likely slightly lower, similar to that reported during in vitro growth (44). CsrA mutants in other bacterial species also show slow growth kinetics (28,58). Additionally, the bacterial loads and infection rates of the DcsrA strain resembled those reported for biofilm-deficient Y. pestis strains (7,(59)(60)(61). Biofilm is thought to maintain bacteria in aggregates that are not easily cleared through defecation after flea blood-feeding and digestion (reviewed in reference 3), accounting for lower bacterial number in strains with reduced biofilm levels.

DISCUSSION
Our work provides evidence for CsrA control of biofilm production in Y. pestis occurring through a tiered posttranscriptional regulatory mechanism (Fig. 8). In summary, cues from alternative carbon catabolism are transduced by CsrA to repress hfq mRNA Congo red (CR) binding assays were used to quantify EPS production in strains cultured in LB medium. A representative picture of strains grown for 48 h on LB containing CR is shown below. Error bars represent the mean 6 SD bound CR from two independent experiments. Statistical significance was determined using oneway ANOVA with a Dunnett's multiple-comparison posttest (**, P , 0.01; ***, P = 0.0005; n.s., not significant).
CsrA Promotes Y. pestis Biofilm-Mediated Flea Blockage ® translation, which, in turn, leads to translational derepression of the hmsT mRNA and a respective increase in intracellular c-di-GMP pools and biofilm production. Thus, CsrA is required to facilitate robust Y. pestis biofilm-mediated foregut blockage rates in rat fleas. Maintenance of normal blockage rates drives epizootic-scale plague transmission events (2, 62). Thus, an ;50% reduction in blockage rate, as seen for the csrA mutant, is synonymous with a compromised ability to maintain natural plague transmission cycles. Coincidently, an hmsT mutant exhibits a similar ;50% decrease in blockage rate (20,21), in agreement with the notion that csrA mutant blockage rates are due to the absence of CsrA-dependent enhancement of hmsT mRNA translation.
Using ribose-and galactose-supplemented media enabled our reiteration that Y. pestis EPS levels are significantly greater during culture on alternative rather than primary carbon sources (44,63). This aligns with observations that metabolic genes involved in uptake of alternative sugars, particularly pentose sugars, are strongly induced in blocked fleas (4, 5), and an intact pentose phosphate pathway is required for efficient flea foregut blockage (61). In addition, we demonstrated that severe impairment in EPS levels in the csrA mutant resulted from exacerbated defects in producing c-di-GMP on alternative carbon sources. CsrA is therefore critical for stimulating c-di-GMP synthesis to increase EPS levels coincident with alternative carbon metabolism. Improved growth and biofilm production on alternative carbon sources in Y. pestis is mediated by cyclic AMP (cAMP) and the cAMP receptor protein complex, cAMP-CRP, under glucose-limiting conditions (44,63). Thus, Y. pestis CsrA-dependent biofilm formation overlaps carbon catabolite repression enabling biofilm production. Interestingly, in Y. pseudotuberculosis, cAMP-CRP activates csrC but represses csrB transcription, thereby optimizing infection fitness to the nutritional status of the mammalian host (64). CsrB also negatively regulates Y. pseudotuberculosis HmsHFRSdependent biofilms (65). In E. coli, cAMP-CRP interfaces with the Csr system by inhibiting csrC and csrB expression and promoting CsrA-dependent biofilm inhibition (41,66). If and how Y. pestis orthologs of sensory ncRNAs CsrB and CsrC (67) contribute to finely adjust CsrA activity in response to physiological changes during flea infection is under our investigation currently.
Bellows et al. (22) described that the hmsT mRNA is posttranscriptionally repressed by Hfq binding to its 59 UTR, leading to accelerated transcript decay and decreases in transcript abundance of the hmsT mRNA. Hfq also posttranscriptionally regulates the hmsT mRNA through its AU-rich, long 39 UTR sequence (23). A yet-to-be-identified small ncRNA is predicted to facilitate Hfq-hmsT mRNA interactions (22,23). Consistent with these reports, we demonstrated that CsrA directly repressed hfq mRNA translation and that the hmsT mRNA incurred accelerated transcript decay. Epistatic analysis using a csrA hfq double mutant, in which we observed a restoration of hmsT mRNA and EPS levels to that exhibited by the wild-type strain, reinforced our findings. Notably, the csrA hfq double mutant exhibited slow growth kinetics relative to the other strains particularly when the Pgm locus was retained. However, when the Pgm locus was lost, the strain grew relatively faster (data not shown), suggesting that high biofilm production compromises growth fitness. Thus, CsrA in Y. pestis may play a role in optimizing growth fitness under nutritional conditions that promote costly biofilm production.
Hfq indirectly activates transcription of hmsP, and, together with posttranscriptional repression of the hmsT mRNA, causes decreases in biofilm and c-di-GMP levels in Y. pestis cultured on brain-heart infusion medium (22). However, in TMH-gal we noted that in an hfq mutant strain, the mRNA levels of hmsT were significantly greater than that of the wild-type strain, as expected, whereas hmsP transcript levels were similar FIG 8 Current model for the CsrA-dependent regulation of c-di-GMP synthesis and biofilm formation in Y. pestis. In Y. pestis, the HmsHFRS proteins synthesize and export EPS to outside the cell to propagate biofilm formation, which leads to biofilm-mediated blockage in fleas. CsrA transduces alternative carbon metabolism cues to enhance biofilm EPS production (green thick broken arrow). CsrA does this through indirect positive regulation of the c-di-GMP synthesis enzyme, HmsT. Hfq represses hmsT mRNA translation by promoting its mRNA decay, but CsrA binds to two binding sites in the 59UTR of hfq to inhibit its translation (solid red line), thereby alleviating repression of the hmsT mRNA translation by Hfq. How and under which physiological conditions ncRNAs CsrB and CsrC antagonize CsrA activity (solid blue line) in Y. pestis are yet to be determined.
CsrA Promotes Y. pestis Biofilm-Mediated Flea Blockage between the wild-type and hfq mutant strains (see Fig. S3 in the supplemental material). We also showed that CsrA does not affect hmsP mRNA steady-state or translational levels, thereby eliminating a role for CsrA-dependent regulation of the hmsP mRNA in biofilm formation under biologically relevant conditions.
In E. coli MG1655, CsrA facilitates repression of an hfq mRNA originating from a promoter immediately upstream of the hfq ATG start through occlusion of ribosome binding and not accelerated mRNA decay (49). Coincidently, the CsrA binding site at the Shine-Dalgarno site for hfq is identical between Y. pestis and E. coli (Fig. S4). However, unlike Y. pestis, E. coli MG1655 does not similarly encode other GGA motifs upstream of the hfq mRNA translational start codon (Fig. S4). In Y. pestis, hfq mRNA transcribed with the upstream miaA gene fully restores Hfq function to an hfq mutant (50), emphasizing the significance of this transcript for production of functional Hfq levels. CsrA repression of this miaA-hfq-containing transcript in Y. pestis likely transpires through the two CsrA binding sites or the binding site at the Shine-Dalgarno site when the hfq transcript is derived from the immediately upstream promoter. In this way, hfq mRNA translation is expected to be more robustly inhibited by CsrA. Furthermore, the Y. pestis hfq transcript mutated only at the Shine-Dalgarno site possessed only a minor inability to completely shift (Fig. 3C). This may reflect an insufficiency of this binding site to properly repress hfq translation of the miaA-hfq-containing transcript, hence the need for the second binding site.
In numerous pathogens, Hfq is critical for mammalian virulence (68,69), and in Y. pestis it is also essential for flea foregut blockage (47). Hfq regulates distinct ncRNA repertoires required for physiological fitness that are conditionally expressed during changing infection stages (68,69). The highly stable nature of the hfq mRNA and repression of hfq mRNA translation without mRNA decay revealed by our half-life studies conceivably preserves hfq mRNA levels for any future critical rapid redeployment of Hfq protein. CsrA-mediated alterations in the hfq mRNA chiefly occurred when the number of blocked rat fleas seen was at its peak. Whether, during this peak blockage stage, (i) the ncRNA that facilitates the Hfq-hmsT mRNA interaction is produced, requiring CsrA repression of the hfq mRNA and (ii) ncRNAs CsrB and CsrC are not expressed to prevent sequestration of CsrA are intriguing questions.
CsrA-dependent biofilm regulation can occur through multiple mechanisms in bacteria. For example, in E. coli, CsrA represses translation of mRNAs encoding DGCs (25), PgaABCD, a homolog to HmsHFRS (36), and the NhaR transcriptional regulator that activates pgaABCD transcription (70) but promotes translation of the ymdA mRNA involved in inhibiting biofilm production (54). Similarly, we predict that CsrA targets other mRNAs involved in biofilm formation in Y. pestis. One putative Y. pestis target that encodes a CsrA GGA binding motif in its 59 UTR is the hmsHFRS mRNA (44). Future studies will be needed to validate putative mRNA targets in Y. pestis. Nonetheless, indirect modulation of DGC mRNAs to promote biofilm production, as illustrated here for CsrA in Y. pestis, is seldom appreciated as part of the repertoire of CsrA-mediated molecular mechanisms of biofilm control. Notably, this contrasts with the better-known paradigmatic role of CsrA direct translational repression of DGC mRNAs to inhibit biofilm formation in other Gammaproteobacteria.

MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions. Bacterial strains and plasmids used in this study are listed in Table S1 in the supplemental material. The Y. pestis KIM61 (pCD1 2 ) strains were cultured on Congo red-heart infusion agar (57) to confirm the presence of the hmsHFRS locus prior to subsequent culturing. Strains were grown at 25°C with shaking unless otherwise stated. The chemically defined TMH medium was prepared as described previously (46). DNA sequencing verified all constructs.
Construction of Y. pestis mutant and complemented strains. Primers are listed in Table S2 under "Mutant and complementation strain construction." To generate the Y. pestis DcsrA::csrA complemented csrA mutant strain, the csrA gene, with flanking promoter, and terminator regions were cloned into pUC18R6KT-mini-Tn7T-Km (71) at the EcoRI sites to create pUC18R6KT-mini-Tn7T-Km-csrA. This plasmid was then used to transpose the csrA expression fragment into the glmS-pstS site of the previously generated csrA mutant strain (44) as previously described (71).
To create the DcsrA Dhfq mutant strain, the hfq gene in the csrA mutant strain was replaced by a kanamycin resistance cassette by homologous recombination, as reported in a previous study (47). This was done by PCR amplification of a fragment containing flanking regions of the hfq gene and kanamycin cassette from the hfq mutant (47) genomic DNA using hfq deletion primers (Table S2). The hfq gene and its up-and downstream sequences were PCR amplified (Table S2) and cloned into a low-copy-number vector, pLG338 (72), at the SmaI site to create plasmid pLGhfq. The DcsrA Dhfq strain was transformed with pLGhfq to generate the DcsrA Dhfq (pLGhfq) complemented strain.
CR binding assay. CR binding assays were performed as previously described, with minor modifications (17). CR was used at final concentrations of 0.03 ng/ml (HIB), 0.06 ng/ml (TMH-glu), 0.12 ng/ml (TMH-gal or TMH-rib), or 0.02 ng/ml (LB) to account for EPS production and incubated for 3 h (HIB and LB) or 1 h (TMH). The A 500 values of sample supernatants were subtracted from medium controls containing CR at their respective concentrations to calculate relative bound CR. The value for the WT strain under each condition was set to 1 for Fig. 1. These values then were multiplied by either 0.5, 1, or 4 for HIB, TMH-glu, and TMH-gal/rib to correct for the amount of CR added per medium condition.
C-di-GMP extraction and quantification. Strains were grown to late log phase. C-di-GMP was extracted from pelleted cells with extraction buffer (100 ml/48-mg cell pellet), and samples were neutralized and quantified using high-performance liquid chromatography (HPLC) as described previously (16,18,22).
Fusion reporter construction and assay. The upstream sequences and partial coding sequences were PCR amplified (primers are listed in Table S2 under "Translational fusion reporters") for flhDC, gyrB, hmsP, hmsT, and hfq. The generated fragments were then fused to amplified fragments of gfpmut3.1 from pFU34 (73) by splice overlap extension PCR (SOE-PCR) and then subjected to digestion with EcoRI and cloned into low-copy-number plasmid pMWO78 (74) at the EcoRI/SmaI sites. Plasmids generated were denoted pMWO78::5'UTR of interest-gfpmut3.1 and transformed into the WT and DcsrA strains (Table S1).
GFP reporter strains were grown to early log phase, split into separate flasks, and treated with anhydrotetracycline (ATc; 200 ng/ml) or vehicle. At 3 h postinduction, the numbers of relative fluorescent units (RFU; excitation, 475 nm; emission, 515 nm) were measured on a TECAN Spark plate reader with optical density at 600 nm (OD 600 ) values taken simultaneously. The medium blank RFU reading was subtracted from the culture RFU of samples and normalized to OD 600 values to account for bacterial growth. The difference of RFU/OD 600 values of induced samples from uninduced samples then was calculated.
CsrA-His 6 expression and purification. The csrA gene was amplified (primers listed in Table S2 under "CsrA-his tag construct"), digested with NcoI/XhoI, and cloned into matching sites in pET28A (Novagen) to generate pET28::csrA-his 6 , which was transformed into the E. coli strain, Bl21lDE3 pLysS. Cultures were induced as previously described (75). Cell pellets were resuspended in protein buffer (100 mM Tris-HCl, 300 mM NaCl, pH 7.5, 20 mM imidazole), 26 U/ml Benzonase (Sigma), and one tablet of cOmplete Mini EDTA-free protease inhibitor cocktail (Roche) per 10 ml buffer and then lysed by sonication. Protein was purified from the supernatant using affinity chromatography as previously described (75). Relevant fractions were dialyzed with a 10,000 molecular weight cutoff (MWCO) Slide-A-Lyzer G2 dialysis cassette (Thermo Fisher Scientific) in dialysis buffer (protein buffer without imidazole). Buffer exchange was performed with 100 mM Tris-HCl, pH 7.5, using a Pall Microsep centrifugal device (3,000 MWCO). Concentrated protein was quantified using the DC protein assay (Bio-Rad). Before incubation with RNA probe, CsrA-His 6 was prepared using 10 mM Tris-HCl, 10% glycerol, pH 7.5. CsrA protein was diluted as previously described (76).
Construction of hfq mRNA probes with GGA site mutations. The nucleotide fragment containing a T7 promoter, the upstream region of hfq (bp 2177 through 74), and the GG-to-CC mutation in BS1 was commercially synthesized (Eurofins) and blunt-end cloned into pJet1.2 (Thermo Fisher Scientific) to create pJet1.2::hfq UTR BS1mut. To generate the BS2 and BS1/BS2 fragments, primers (Table S2 under "EMSA probes") containing the GGA-to-CCC mutation in BS2 or CC-to-GG restoration in BS1 were used with inverse PCR of the pJet1.2::hfq UTR BS1mut to generate mutated fragments for labeled probe generation.
REMSA. To generate biotin-labeled EMSA probes, PCR fragments were first generated from primers (Table S2 under "EMSA probes") where each forward primer contained a T7 promoter sequence. Fragments were gel purified and transcribed with the MegaShortScript T7 transcription kit (Invitrogen). Probes were purified using the RNA Clean and Concentrator-25 kit (Zymo) and their size confirmed by electrophoresis. RNA probes were then labeled using a Pierce RNA 39-end biotinylation kit and purified with the Oligo Clean and Concentrator kit (Zymo). Labeled RNA concentrations were determined using a Thermo Scientific NanoDrop.
CsrA-His 6 probe binding reactions were conducted in 10Â CsrA binding buffer (76) and incubated at 37°C for 30 min. For REMSAs shown in Fig. 4, binding reaction mixtures included 60 ng unlabeled yeast RNA (Invitrogen) and SUPERase-In (Invitrogen). CsrA-His 6 -probe complexes were electrophoresed on a 6% nondenaturing polyacrylamide gel, transblotted to a positively charged nylon membrane, and then UV cross-linked (120 mJ/cm 2 ). Biotinylated probes were detected using the chemiluminescent nucleic acid detection module (Thermo Fisher Scientific).
In vitro cell-free translation assay. The PURExpress kit (New England Biolabs) was used per the manufacturer's instructions. The reporter fusion constructs pMWO78::hfq-gfpmut3.1 or pMWO78::hmsT-gfpmut3.1, described above, were used as templates to generate hfq-gfp and hmsT-gfp mRNA transcripts with primers containing a T7 promoter sequence (Table S2 under "EMSA probes") using the MegaShortScript T7 transcription kit. mRNA transcripts were purified with the RNA Clean and Concentrator-25 kit (Zymo). Reaction mixtures contained 326 nM mRNA transcript with 6.2 mM CsrA-His 6 and were incubated at 37°C. GFP signal was detected by immunoblot using 1:20,000 rabbit anti-GFP (Invitrogen), 1:100,000 goat anti-rabbit horseradish peroxidase (Invitrogen), and the SuperSignal West Femto kit (Thermo Fisher Scientific) and quantified by densitometry on a ChemicDoc MP using Image Lab 4.1.
Flea infections. Cohorts of Xenopsylla cheopis fleas were artificially infected with Y. pestis strains, and infected flea maintenance, blockage, and CFU enumeration were performed as previously described (21,78). To account for growth defects in the DcsrA strain, strains were grown overnight at room temperature and then moved to 37°C at 5 h prior to infection. Studies with mice were performed in strict accordance with the U.S. National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals (79) and as approved by the Washington State University Institutional Animal Care and Use Committee.
Statistical analysis. Details of statistical analysis using GraphPad Prism version 8.1.1 are provided in the legends of Fig. 1 to 7.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.