CsrA-Mediated Translational Activation of the hmsE mRNA Enhances HmsD-Dependent C-di-GMP-Enabled Biofilm Production in Yersinia pestis

ABSTRACT The plague bacterium, Yersinia pestis, forms a biofilm-mediated blockage in the flea foregut that enhances its transmission by fleabite. Biofilm formation is positively controlled by cyclic di-GMP (c-di-GMP), which is synthesized by the diguanylate cyclases (DGC), HmsD and HmsT. While HmsD primarily promotes biofilm-mediated blockage of fleas, HmsT plays a more minor role in this process. HmsD is a component of the HmsCDE tripartite signaling system. HmsC and HmsE posttranslationally inhibit or activate HmsD, respectively. HmsT-dependent c-di-GMP levels and biofilm formation are positively regulated by the RNA-binding protein CsrA. In this study we determined whether CsrA positively regulates HmsD-dependent biofilm formation through interactions with the hmsE mRNA. Gel mobility shift assays determined that CsrA binds specifically to the hmsE transcript. RNase T1 footprint assays identified a single CsrA binding site and CsrA-induced structural changes in the hmsE leader region. Translational activation of the hmsE mRNA was confirmed in vivo using plasmid-encoded inducible translational fusion reporters and by HmsE protein expression studies. Furthermore, mutation of the CsrA binding site in the hmsE transcript significantly reduced HmsD-dependent biofilm formation. These results suggest that CsrA binding leads to structural changes in the hmsE mRNA that enhance its translation to enable increased HmsD-dependent biofilm formation. Given the requisite function of HmsD in biofilm-mediated flea blockage, this CsrA-dependent increase in HmsD activity underscores that complex and conditionally defined modulation of c-di-GMP synthesis within the flea gut is required for Y. pestis transmission. IMPORTANCE Mutations enhancing c-di-GMP biosynthesis drove the evolution of Y. pestis to flea-borne transmissibility. c-di-GMP-dependent biofilm-mediated blockage of the flea foregut enables regurgitative transmission of Y. pestis by fleabite. The Y. pestis diguanylate cyclases (DGC), HmsT and HmsD, which synthesize c-di-GMP, play significant roles in transmission. Several regulatory proteins involved in environmental sensing, as well as signal transduction and response regulation, tightly control DGC function. An example is CsrA, a global posttranscriptional regulator that modulates carbon metabolism and biofilm formation. CsrA integrates alternative carbon usage metabolism cues to activate c-di-GMP biosynthesis through HmsT. Here, we demonstrated that CsrA additionally activates hmsE translation to promote c-di-GMP biosynthesis through HmsD. This emphasizes that a highly evolved regulatory network controls c-di-GMP synthesis and Y. pestis transmission.

transcript containing GGA2 and GGA3 (2168 to 267 relative to the hmsE ATG start codon) shifted as a single distinct band with increasing concentrations of purified CsrA-His 6 , indicating CsrA binding ( Fig. 2A). An apparent K d of 138 6 20 nM CsrA (Fig.  2B) was determined by a nonlinear least-squares analysis, revealing that CsrA had a relatively low binding affinity for the hmsE transcript. To determine the specificity of the CsrA-hmsE RNA interaction, competitive gel shifts were performed using unlabeled specific (hmsE) and nonspecific (phoB) competitor RNAs in 10-, 100-, and 1,000-fold excess of radiolabeled hmsE transcript. At a concentration of 200 nM CsrA-His 6 , unlabeled hmsE RNA competed with the radiolabeled hmsE RNA for CsrA binding, while the nonspecific competitor phoB RNA did not, indicating that the CsrA-hmsE RNA interaction was specific (Fig. 2C).
CsrA binds to one site in the hmsE leader. Footprint assays with a transcript of hmsE containing GGA1, GGA2, and GGA3 were used to identify CsrA binding sites in the hmsE leader region (Fig. 3). Cleavage of single-stranded G residues in the absence and presence of CsrA was probed with RNase T1. Cleavage of the second G residue of the GGA2 motif (G71), as well as G74 just downstream, was less efficient, indicating that CsrA was bound to the RNA at this site ( Fig. 3A and 3C). No CsrA-dependent protection was observed for GGA1 and GGA3, implying that GGA2 is the sole binding site for CsrA in the hmsE transcript (Fig. 3A). In addition, the G residues at positions 125, 133, 136, and 142 were cleaved less efficiently in a CsrA-dependent manner ( Fig. 3A and 3B), implying that structural changes in the mRNA occurred upon CsrA binding.
CsrA activates hmsE translation by binding to GGA2. Green fluorescent protein (GFP) translational fusion reporters were used to determine if translation of hmsE is CsrA dependent. The 59 leader and N-terminal coding sequence (CDS) of hmsE was fused in frame to the gfpmut3.1 gene. Expression of this fusion was driven by the inducible promoter, PtetO. A csrA mutant strain (DcsrA) and the isogenic KIM61 parental strain (wild type [WT]) transformed with the hmsE-gfp translational fusion were grown to the exponential phase in TMH-gal medium, and fluorescence was recorded at 3 h postinduction with anhydrotetracycline (ATc). A significant reduction in GFP expression was observed in the DcsrA strain containing the hmsE-gfp reporter (Fig. 4A) compared to the WT strain, suggesting that translation of hmsE was positively regulated by CsrA.
To determine if activation of hmsE-GFP fusion expression was dependent on GGA2, a GG!CC mutation was introduced into the GGA2 motif and referred to as DBS2. In the WT strain, the hmsE-gfp DBS2 reporter showed significantly lower GFP expression levels than the WT reporter ( Fig. 4B), suggesting that BS2 was involved in translational activation of hmsE. Next, to determine if CsrA-mediated activation of hmsE expression impacted HmsE protein levels, the WT and DcsrA strains were engineered with a 3ÂFlag epitope sequence fused in-frame to the 39 end of the hmsE gene in the chromosome. These strains, named WT HmsE-Flag and DcsrA HmsE-Flag , were grown to exponential phase in TMHgal medium and the level of HmsE-Flag protein was assessed by immunoblotting of equivalently loaded samples. The HmsE-Flag protein mean band intensity was 2.0 6 0.3-fold greater for the WT HmsE-Flag strain than for the DcsrA HmsE-Flag strain (Fig. 4C), indicating that HmsE protein levels were increased in the presence of CsrA.
Since translational fusions monitor expression at the level of transcription and translation, we determined whether the steady-state mRNA levels differed in WT and DcsrA strains when grown to the early and mid-exponential phase. No difference in steady-state levels of the hmsE transcript was observed in the WT and the DcsrA strains (Fig. 4D), indicating that transcript abundance of hmsE was unaffected by CsrA activity. Taken together with the in vitro binding results, our expression studies indicate that bound CsrA activates translation of hmsE.
Predicted small open reading frames (ORFs) encoded in the hmsE leader do not affect HmsE expression. We identified two ORFs in the hmsE leader region near GGA2 (Fig. 1). One was a 30-amino acid ORF (designated uORF30; Fig. 1) with an ATG start codon that overlapped GGA2 and a UAG stop codon that overlapped the hmsE SD sequence. The SD sequence of the second ORF (designated uORF29; Fig. 1) appeared to overlap GGA2 with its UAG stop codon directly adjacent to the hmsE SD sequence. We reasoned that expression of these ORFs may obstruct ribosome binding to the hmsE SD sequence to reduce HmsE expression. Therefore, to determine if the putative uORF30 and uORF29 are translated, inducible GFP translational fusion plasmid constructs of uORF30 and uORF29 were generated as pBAD::uORF30-gfp and pBAD::uORF29-gfp. In addition, constructs with a stop codon replacing the second codon were also made. These constructs were named pBAD::uORF30STOP-gfp and pBAD::uORF29STOP-gfp, respectively. Similarly constructed pBAD::hmsE-gfp and pBAD::hmsESTOP-gfp served as positive controls. Expression levels of these fusions were examined following induction with arabinose. The control pBAD:: hmsESTOP-gfp construct did not produce GFP upon induction in E. coli. No difference in expression between pBAD::uORF30-gfp and pBAD::uORF30STOP-gfp or between pBAD:: uORF29-gfp and pBAD::uORF29STOP-gfp was noted in E. coli or Y. pestis, suggesting that these proteins were not expressed under these conditions (Fig. 5).
CsrA-dependent hmsE mRNA translational activation promotes greater HmsDdependent biofilm production. Taking advantage of an arabinose-inducible plasmid construct, pBAD::hmsE, that contains the hmsE CDS plus 170 bp of upstream leader sequence, we next asked if disruption of BS2 impacted HmsD-dependent biofilm production. Previous work demonstrated that an hmsE overexpression construct containing 39 bp of upstream sequence increased HmsD-dependent c-di-GMP levels and biofilm production in WT and DhmsT strains (13). Likewise, the pBAD::hmsE -170 construct caused increased biofilm in the DhmsT strain following induction with 0.5% arabinose (Fig. 6A). A construct with a GG!CC substitution in the GGA motif of BS2 was generated and referred to as pBAD::hmsE -170 DBS2. The plasmids were transformed into the DhmsT strain, and biofilm assays were conducted to compare HmsD-dependent biofilm formation following induction with arabinose ( Fig. 6B). Compared to the DhmsT (pBAD::hmsE -170) strain, the DhmsT (pBAD::hmsE -170 DBS2) strain showed a significant ;2-fold decrease in biofilm formation (Fig. 6B). These results indicated that BS2 was needed to augment HmsD-dependent biofilm formation.

DISCUSSION
Loss of function mutations that enhanced c-di-GMP-mediated biofilm formation in the flea foregut to increase transmissibility delineates the emergence of Y. pestis from its ancestor Yersinia pseudotuberculosis (25). Optimization of biofilm formation through modulation of c-di-GMP in synchrony with adaptation to the nutritional environment of the flea is also required. This process is multifaceted with regulatory checkpoints at both transcriptional (26) and posttranscriptional levels (15,27). For example, the WT and DcsrA strains were grown to the early and mid-exponential phase in TMH-gal medium, and the steady-state abundance of hmsE transcripts was compared between strains. The mean 6 standard deviation of three independent experiments is presented. An unpaired t test was used to determine significance. carbon metabolism regulator CsrA transduces alternative sugar metabolism signals to promote biofilm production in Y. pestis by translationally inhibiting the hfq mRNA to relieve Hfq repression of HmsT-dependent c-di-GMP biosynthesis (15). The present work demonstrates that CsrA also binds to the 59 leader region of hmsE mRNA to activate its translation, which in turn leads to greater HmsD-dependent biofilm production. Because HmsD is the predominant DGC in the flea, CsrA-dependent regulation of c-di-GMP levels through HmsD may serve as an important fine-tuning mechanism for the development of a transmissible Y. pestis infection from its flea vector and spread of plague by fleabite (9,28).
Only a few examples of CsrA-dependent activation have been identified. For example, CsrA binds to and prevents RNase E-mediated cleavage of the flhDC (29) and csrB (30) RNAs. A CsrA-mediated translation activation mechanism has been elucidated only for E. coli ymdA (31). In this case, CsrA binds to two sites in the ymdA leader region, which destabilizes a hairpin structure that otherwise sequesters the ymdA SD sequence. Like the ymdA transcript, the hmsE mRNA also contains a GGA motif overlapping the SD sequence, but our binding assays indicated that CsrA does not bind to this region. Instead, a site further upstream from the hmsE SD sequence was identified as a FIG 5 Putative small protein-coding ORFs identified in the hmsE leader are not expressed. Arabinoseinducible GFP expression was examined in E. coli carrying plasmids pBAD::hmsE-gfp, pBAD::uORF30-gfp, pBAD:: uORF29-gfp, pBAD::hmsESTOP-gfp, pBAD::uORF30STOP-gfp, and pBAD::UORF29STOP-gfp grown in M9 medium (0.2% glycerol) with and without arabinose induction. Y. pestis strains carrying pBAD::uORF30-gfp, pBAD:: uORF29-gfp, pBAD::uORF30STOP-gfp, and pBAD::UORF29STOP-gfp were grown in TMH-gal medium with and without arabinose induction. The data are represented as the difference between the induced and uninduced RFU/OD 600 values. Error bars represent the mean 6 SEM of three independent experiments. Black and green curves denote E. coli strains, and gray curves denote Y. pestis strains. binding site (GGA2). Additionally, CsrA binding protects some G-residues downstream of GGA2 from RNase T1 digestion, most likely by causing RNA structural rearrangements. CsrA-dependent translational activation of hmsE was established using WT and BS2 mutant hmsE-gfp translational fusions and RNA abundance experiments. Protein expression studies substantiated these findings, as a higher level of HmsE was produced from its native locus in the WT strain than in an isogenic csrA mutant. Functional studies of HmsD-dependent biofilm formation using HmsE overexpression constructs containing WT or BS2 mutations served as further evidence for the role of CsrA-dependent activation of hmsE mRNA translation. Taken together, our results suggest a model for CsrA-dependent translational activation of hmsE mRNA occurring via a CsrA-induced structural change to the RNA in the hmsE leader region. Presumably, the restructured hmsE transcript is more accessible for ribosome binding.
Previous studies of the iraD locus of E. coli (32) determined that CsrA binds and represses translation of the short upstream ORF27 with which iraD is translationally coupled, leading to translational repression of iraD mRNA as well. Therefore, we investigated if two in silico predicted putative small ORFs, named uORF30 and uORF29, present in the hmsE leader sequence could impact translation of hmsE mRNA. Under the employed testing conditions, no expression was noted from either small ORF, suggesting that these ORFs may not code for small proteins.
Homologs of hmsCDE have been described in pathogenic Pseudomonas aeruginosa (yfiBNR) and Escherichia coli CT073 (yfiLRNB) species, where these loci are implicated in regulating c-di-GMP and biofilm production (33)(34)(35). Whether CsrA regulates expression of these genes in P. aeruginosa and E. coli is unknown. However, the leader region of the yfiBNR transcript in P. aeruginosa does not contain predicted CsrA binding sites (36).
The HmsD and HmsC proteins are proposed to be inner membrane-spanning proteins, with HmsC localized extensively, and HmsD partially, to the periplasm. HmsE is likely a peptidoglycan-associated lipoprotein-like outer membrane protein (12,13). Currently, the model for how the HmsCDE tripartite signaling system operates (Fig. 7) is that modulation of the HmsC and HmsE protein levels dictate the DGC activity of HmsD. Various environmental signals differentially modulate HmsC and HmsE levels (37). A flea-specific signal likely prompts HmsE protein increases because an hmsE mutant shows poor biofilm-mediated blockage of fleas but wild-type biofilm levels in vitro (13). A reducing environment in the periplasm facilitates decreases in HmsC, but not HmsE, levels (14,37). Consistent with the model, this study demonstrates that CsrA increases HmsE expression and enhancement of HmsD-dependent biofilm production; however, the flea-specific conditional determinants for this interaction remain elusive.
Although CsrA appears to stimulate c-di-GMP levels through activation of HmsT and HmsD activity, the spatial and temporal contexts of both DGC activities during Y. pestis biofilm-mediated foregut blockage is unknown. Certainly, spatially localized cytosolic pools of c-di-GMP within the cell that are distinctly contributed to by HmsT or HmsD activity may need to be individually tuned to maximize biofilm production with changing environmental conditions. Such knowledge may explain the disparate roles of HmsT and HmsD during flea blockage. Nevertheless, in this study we show that like in other bacteria, CsrA controls cellular c-di-GMP pools through more than one mechanism, reiterating that CsrA is a significant mediator that coordinates biofilm production in accordance with the bacterial nutritional environment. This study also serves as further evidence that a highly evolved regulatory network in c-di-GMP synthesis supports the regurgitative transmission of Y. pestis by fleabite.

MATERIALS AND METHODS
Bacterial strains, growth conditions, and plasmids. Bacterial strains are listed in Table 1. The Y. pestis KIM61 parental strain was designated as the wild-type (WT) strain in this study. All Y. pestis strains were routinely cultured on heart infusion broth (HIB) and on Congo red heart infusion agar (38) to verify the biofilm status of the strains. Strains grown in broth cultures were aerated continuously at 26°C. To support robust levels of biofilm, the chemically defined TMH medium (39) supplemented with 0.2% galactose (TMH-gal) was used as previously described (15). E. coli strains were generally cultured in LB medium, except for translational fusion reporter studies where M9 minimal medium supplemented with 0.2% glycerol was used. The primers are listed in Table 2. All plasmid constructs were verified by DNA sequencing (Eurofins Genomics or Azenta Life Sciences).
Quantitative reverse transcription-PCR (qRT-PCR) assays. Early and mid-exponential phase TMHgal cultures of strains of interest were mixed with RNAprotect bacterial reagent (Qiagen). Cells were then harvested at room temperature and stored at 280°C until RNA isolation. RNA isolation and qRT-PCR were conducted as previously described (15,40).
GFP translation fusion reporter construction and assays. The GFP translation fusion reporter using the pMW078 plasmid with the anhydrotetracycline (ATc) inducible promoter was made and assayed as previously described (15). The upstream 59 leader and first 31 codons of the hmsE gene were amplified by PCR from the genomic DNA of the WT strain with primer pair p1033/p1035. The generated fragment was then spliced by overlap extension PCR (SOE-PCR) to the gfpmut3.1 sequence from pFU34 that was generated with primer pair p1036/p126 (41). The resulting amplification product was digested with EcoRI and cloned into pMWO78 (42) at the EcoRI/SmaI sites and named pMWO78::hmsE-gfp.
CsrA-His 6 purification. Purified CsrA-His 6 was obtained from the induced BL21lDE3 pLysS pET28A:: csrA-his 6 strain and subjected to expression and purification as previously described (15,43). Exceptions were that dialysis was accomplished using a 7,000 molecular weight cutoff (MWCO) Slide-A-Lyzer G2 dialysis cassette (Thermo Fisher Scientific) with 100 mM Tris-HCl, 100 mM NaCl, and 10% glycerol, pH 7.5, as the dialysis buffer, and a follow-up buffer exchange step was omitted.
RNA gel mobility shift assay. Quantitative gel mobility shift assays followed a published procedure (44). Three different-sized RNAs were synthesized with the RNAMaxx kit (Agilent Technologies) using PCR-generated DNA templates: 174 nucleotides [nt]; (2168 to 13 relative to the hmsE translational start codon) with all 3 GGA repeats, 104 nt (2168 to 267 relative to the hmsE translational start codon) with GGA2 and 3, and 91 nt (267 to 123 relative to the hmsE translational start codon) with GGA1. The template for PCR was the pBAD30::hmsE -170 plasmid. Gel-purified RNAs were dephosphorylated and then  (44). RNase T1 footprint assays. CsrA-hmsE RNA footprint assays followed a published procedure (44). The 174-nt RNA (2168 to 13 relative to the hmsE translational start codon) with all 3 GGA repeats was labeled as described above for the gel mobility shift assay. The reaction mixtures were identical to those in the gel shift assay except that the concentration of labeled RNA was raised to 2 nM, and 1 mg of acetylated bovine serum albumin (BSA) was added to each reaction mixture. Reaction mixtures were incubated for 30 min at 37°C to allow CsrA-RNA complex formation, and then RNase T1 (0.08 U) was added, and incubation was continued for 15 min at 37°C. Reactions were stopped by adding 10 mL of gel loading buffer (95% formamide, 0.025% sodium dodecyl sulfate [SDS], 20 mM EDTA, 0.025% bromophenol blue, 0.025% xylene cyanol). Samples were heated for 5 min at 90°C and then fractionated through 6% sequencing gels. Cleavage patterns were examined using a phosphorimager and quantified using semiautomated footprinting analysis software (SAFA) (45). Each number was normalized to the 95th percentile value of the column.
Construction of HmsE-33Flag-expressing strains. The WT HmsE-Flag and DcsrA HmsE-Flag strains expressing recombinant HmsE protein with a 3ÂFlag epitope tag at the C terminus (HmsE-Flag) were constructed by replacing the native chromosomal gene using a modification of the lambda red recombination system (46). To generate these strains, a 520-bp upstream (primers p1090/p1091) region and a 509-bp downstream (primers p1092/p1093) region of the 39 end of the hmsE gene were amplified separately. Splice-overlap PCR joined these fragments to another PCR amplicon containing a kanamycin resistance-encoding gene and the 3 Â Flag epitope tag sequence that was previously amplified (primers p1081/p1082) from pSUB11. The resultant product was transformed into WT and DcsrA competent cells expressing the lambda red recombinase. Kanamycin-resistant transformants were confirmed by PCR to contain the insertion. The kanamycin resistance cassette was excised as previously described (47), and DNA sequencing verified the intended gene replacement.
Anti-FLAG tag immunoblotting. For immunoblotting, the WT HmsE-Flag and DcsrA HmsE-Flag strains were cultured in TMH-gal medium, and cells were harvested during the exponential growth phase. Cells were suspended and boiled in 1Â Laemmli sample buffer for 5 min. Lysates were separated by SDS page and transferred to a 0.2-mm nitrocellulose membrane (Bio-Rad). HmsE-Flag protein was detected using monoclonal anti-FLAG M2 (Sigma) antibodies (1:20,000), affinity purified horseradish peroxidase (HRP)labeled goat anti-mouse (KPL) antibodies (1:10,000), and the SuperSignal West Femto substrate (Thermo Scientific). The No-Stain protein labeling reagent (Invitrogen) was used to normalize total protein loaded per lane of the gel. The HmsE-Flag protein was quantified by densitometry using the ChemiDoc software Image Lab 4.1 (Bio-Rad).
Arabinose-inducible GFP translational fusion assays. Overnight HIB cultures of Y. pestis strains harboring pBAD plasmid constructs were diluted 1:100 in TMH-gal. Overnight LB cultures of E. coli strains harboring pBAD plasmids were diluted 1:100 in N-minimal medium supplemented with 0.2% glycerol. The 1:100 dilutions were split in two with, one receiving 0.2% L-arabinose. Triplicate aliquots of 100 mL were added to a 96-well uClear black plate, and the plate was shaken in the Tecan Spark device for 16 h, during which GFP and optical density at 600 nm (OD 600 ) readings were taken every hour.
Biofilm assays. Overnight HIB cultures of the DhmsT Y. pestis strain harboring pBAD plasmid constructs were diluted 1:100 in TMH-gal medium and induced with 0.5% L-arabinose. Triplicate aliquots of 100 mL were added to a 96-well plate and the plate was shaken for 24 h at room temperature. Biofilm was quantified by safranin staining as previously described (48).
Statistical analysis. Details of statistical analysis using GraphPad Prism 9.4.1 are provided in the legends of Fig. 4 to 6.
Data availability. Data are available on reasonable request from the corresponding author.