A Novel Small RNA Promotes Motility and Virulence of Enterohemorrhagic Escherichia coli O157:H7 in Response to Ammonium

The process by which bacteria sense environmental cues to regulate their virulence is complex. Several studies have focused on regulating the expression of the locus of enterocyte effacement pathogenicity island in the typical gut pathogenic bacterium, O157.

uremic syndrome in humans (1). O157 usually infects the human colon through contaminated food and water at a low infectious dose (50 CFU) (2). Since no vaccine is available against O157 infections, effective treatments are urgently required to combat these infections (3). O157 typically colonizes the human large intestine (4), and successful O157 colonization is characterized by the formation of attaching and effacing (AE) lesions in the host epithelium (5). These lesions lead to the rearrangement of the actin cytoskeleton and effacement of microvilli, which results in the form of pedestallike structures and intimate attachment of O157 to the enterocytes (6). The locus of enterocyte effacement (LEE) pathogenicity island mainly encodes genes whose products cause AE lesions (7). The LEE contains five polycistronic operons (LEE1 to LEE5) that express the master LEE regulator, Ler, type III secretion system (T3SS), adhesins (such as intimin and its receptor [Tir]), and other effector proteins (8). In addition to T3SS and effectors encoded by LEE, other adhesins also participate in the initial attachment of bacteria to eukaryotic cells (9).
Flagella are the locomotive organelles important for bacterial pathogenesis and contribute to the initial breakdown and penetration through the mucus layer by bacterial pathogens (10). Moreover, flagella may directly regulate bacterial adhesion and colonization at the sites of infection (11). EHEC flagella directly interact with mucin in the mucus layer (12). O157 flagella also act as adhesins, directly contributing to adhesion to bovine intestinal epithelium in the early stages of colonization (13). Flagella also contribute to the pathogenesis of EHEC O113:H21 infections by promoting the invasion of the intestinal epithelium (14). More than 60 flagellum-related genes have been arranged in hierarchical order in three classes (14). Class 1 contains the master operon, flhDC, the expression of which is needed for transcribing class 2 and class 3 operons (15). Class 2 contains eight operons that encode components for constructing the hook-basal body complex, including the sigma factor fliA, to directly regulate flagellar genes (16). Class 3 encodes components for filament assembly and motor function, including the flagellin gene, fliC. FlhB is a class 2 constituent membrane protein of the basal body required for flagellar appendant synthesis and hook-length control involved in substrate specificity switch (17,18). Recently, FlhB was found to regulate flagellar gene expression in Listeria monocytogenes (17).
Small RNAs (sRNAs) are regulatory, noncoding, rapid-acting molecules whose synthesis does not require much energy consumption and are involved in many physiological processes of bacteria (19,20). sRNAs positively or negatively regulate target genes by base pairing with specific mRNAs and often need the sRNA chaperone, Smlike homohexameric ring protein, Hfq, to function (21,22). For bacterial pathogens, sRNAs play a vital role in their rapid adaptation to the host environment and regulating virulence gene expression (22). Several virulence-related sRNAs have been identified in O157 (23)(24)(25)(26). For example, sRNA, DicF, senses low oxygen concentrations in the environment to induce expression of LEE genes to promote the formation of AE lesions and virulence through direct binding with the transcriptional activator, pchA mRNA (23). In addition, several sRNAs (DsrA, Arl, Esr41, Spot42, sRNA56, sRNA103, sRNA350, GlmY, and GlmZ) regulate LEE operons in O157 (26)(27)(28)(29). However, the exact mechanisms by which most sRNAs in O157 regulate downstream genes and their roles in bacterial virulence remain unclear.
Bacterial two-component systems (TCSs) sense various microenvironmental cues, which are transferred from the cytoplasmic membrane to the cytoplasm to activate several processes (30). TCSs contain a sensor histidine kinase and its cognate DNAbinding response regulator (RR) (31). The conformational change of the membrane histidine kinase delivers the phosphate to the aspartic residue on the RR to activate a domain that modulates gene expression (32). Previous studies have identified several TCSs that are essential for the virulence of O157 (33). For example, CpxA/CpxR TCS senses the neurotransmitter, serotonin, and downregulates O157 LEE gene expression (33). FusK/R TCS senses fucose produced by the commensal Bacteroides thetaiotaomicron to promote O157 colonization (34). QseE/QseF TCS senses epinephrine and sulfate to regulate the effector protein synthesis, EspFu (35). It is likely that more TCSs are involved in the virulence regulatory network of O157; further investigation is required to confirm and clarify their roles. Here, we aimed to investigate the role of an sRNA, EsrF, and the signaling pathway and the mechanism involved in the motility and virulence of O157.

RESULTS
Identification and characterization of EsrF in O157. By analyzing the transcriptomic data of the O157 strain EDL933 during infection of HeLa cells (36), we predicted one potential sRNA in the intergenic region between ntrB (glnL) and glnA and called it EsrF. The transcriptomic data showed that EsrF expression is significantly upregulated by 4.43-fold when infecting HeLa cells compared to its expression in free cultures in Dulbecco modified Eagle medium (DMEM). We inferred that EsrF is associated with O157 virulence.
Northern blotting was performed to confirm whether EsrF is an sRNA transcribed in O157 and to determine its size. Briefly, we used an EsrF-specific probe to detect O157 WT (wild type), DesrF (esrF mutant), and DesrF1PesrF (esrF complement strain) strains, with 5S rRNA as the internal control. The results revealed an RNA band that was approximately 80 to 90 nucleotides (nt) in length in WT and DesrF1PesrF strains, but not in the DesrF mutant, indicating that EsrF was transcribed from the reverse strand of the EDL933 genome (Fig. 1A).
We then performed the 59 and 39 rapid amplification of cDNA ends (RACE) assay to identify the transcription start and termination sites of EsrF. The results show that EsrF was exactly 85 nt in length, which is consistent with the Northern blotting results. Further, the transcription start site (11) of EsrF coding sequence (CDS) was mapped to T 4939965 in the EDL933 genome, while a Rho-independent terminator was mapped to T 4940049 ( Fig. 1B; see also Fig. S1B in the supplemental material). Using RNAfold (37), we also predicted the secondary structure of EsrF to have one hairpin, two bulges, and three internal loops (Fig. 1C).
Moreover, the transcription start site, "T," of esrF is the nucleotide adjacent to the glnA CDS stop codon (Fig. 1B). Hence, EsrF is a 39 untranslated region (UTR) type sRNA that may contain the full glnA 39 UTR. To investigate whether EsrF is independently transcribed or cotranscribed and cleaved from the full-length glnA transcript, we performed several analyses. First, the transcriptome data suggest that the putative unique promoter region of esrF is located 235 to 210 bp upstream of the esrF transcription start site ( Fig. 1B; see also Fig. S1A). Second, after deletion of the glnA promoter in O157 WT and subsequent Northern blotting, no difference was observed in EsrF abundance between DglnAp and WT strains (Fig. 1D). Third, considering that cotranscribed sRNAs processed from primary transcripts lead to the conversion of 59 triphosphorylated RNAs to 59 monophosphorylated RNAs (38), we treated total RNA samples of O157 WT with the 59 monophosphate-dependent terminator exonuclease (TEX), which degrades processed transcripts while sparing primary transcripts (39). Northern blotting was subsequently performed, which revealed that the EsrF transcript is resistant to TEX treatment (Fig. 1D). Fourth, we introduced a promoter-less plasmid (pJV300) containing esrF, with its predicted promoter region, into the DesrF mutant, which, according to Northern blotting results, lead to the overexpression of EsrF (Fig. 1A). Taken together, these results indicate that EsrF is an independently transcribed 39 UTR sRNA, the expression of which is regulated by its promoter and does not require glnA expression.
Hfq is a well-known sRNA chaperone that binds to both sRNAs and their target mRNAs to promote their base pairing, while protecting sRNAs from degradation by cellular nucleosidases. The expression of Hfq-binding sRNAs is generally reduced in Gramnegative bacteria hfq deletion mutants. We therefore examined the effect of hfq deletion on EsrF abundance by performing Northern blotting. Results for which demonstrate the presence of EsrF RNA in WT and Dhfq1Phfq strains, which was absent from the Dhfq strain (Fig. 1E), indicating that hfq promotes EsrF abundance.
Ultraviolet (UV) cross-linking has been commonly employed to determine whether a specific protein interacts directly with a specific RNA in living cells (40). Specifically, UV-induced RNA-protein cross-linking and high-throughput sequencing has been WT, DesrF, and DesrF1PesrF strains; 5S rRNA was used as a loading control. esrF with its putative promoter was cloned into a high-copy-number plasmid pJV300, and the abundance of EsrF in DesrF1pJV300-esrF was determined by Northern blotting. (B) The position of the esrF sequence in the O157 EDL933 genome is shown. The 235 and 210 boxes of the putative promoter region, transcription start site, flhB mRNA binding site, and Rho-independent termination site are depicted in green, blue, purple, and red, respectively. The palindromic sequence before poly(T) in Rho-independent termination site are demarcated with a dashed box. (C) Predicted secondary structure of EsrF sRNA determined using RNAfold (Institute for Theoretical Chemistry, University of Vienna [http://rna.tbi.univie.ac.at/]) and depicted as a color gradient from green to red, representing 59 to 39 ends of EsrF sRNA, respectively. (D) EsrF abundance in WT, glnA promoter mutant (DglnAp), and WT incubated with 59 monophosphate-dependent terminator exonuclease (WT1TEX) as determined by Northern blotting. (E) Northern blotting was performed with a specific probe directed against EsrF sRNA in O157 WT, Dhfq, and Dhfq1Phfq strains; 5S rRNA was used as a loading control. The data are presented as means 6 the SD (n = 3; ns, no significance; *, P # 0.05; **, P # 0.01; ***, P # 0.001). All P values were calculated using a Student t test. applied to identify transcriptome-wide targets of Hfq binding in O157 (41). By analyzing this published raw data, we found that Hfq-cross-linked reads cover the entire sequence of esrF (see Fig. S1C), suggesting that Hfq may directly bind to the "body" of EsrF sRNA and contribute to its stability.
EsrF promotes O157 virulence in vitro and in vivo. To investigate the impact of EsrF sRNA on the adhesion of O157 to host cells, we performed a HeLa cell adhesion assay. To avoid the influence of motility on bacterial adhesion, the bacteria were first centrifuged onto the HeLa cells before infection. The results show that the DesrF mutant adhered to cells at much lower levels (2.3-and 2.1-fold) than did the WT and DesrF1PesrF strains ( Fig. 2A).
Considering that the most important features associated with O157 adhesion to host cells are the AE lesion and pedestal formation (23), fluorescein actin staining (FAS) was performed to observe them in HeLa cells infected by WT, DesrF, or DesrF1PesrF strains. The results revealed significantly reduced AE lesion formation in HeLa cells infected by the DesrF mutant (20%) compared to cells infected by WT (45%) or DesrF1PesrF strains (50%), with averages of 15.0, 4.8, and 13.3 pedestals per cell observed in cells infected by WT, DesrF, and DesrF1PesrF strains, respectively ( Fig. 2B, C, and D). Notably, WT, DesrF, and DesrF1PesrF strains grew at similar rates in vitro (Fig. 2E), indicating that the decreased cell adhesion of the DesrF mutant was not due to slower bacterial growth. Cumulatively, these results suggest that EsrF positively regulates O157 adhesion to host HeLa cells.
To further investigate the effect of EsrF sRNA on O157 pathogenicity in vivo, we performed a rabbit colon colonization experiment. Three-day-old New Zealand White infant rabbits were intragastrically inoculated with WT and DesrF strains, respectively, and the numbers of bacteria recovered from the colon homogenates were determined. The results showed that rabbit colons infected with DesrF had relatively fewer (10.8fold) bacteria than the colons infected with the WT strain (Fig. 2F). This result indicated that EsrF sRNA had a positive influence on O157 virulence in vivo.
flhB expression is positively regulated by EsrF. sRNAs influence the expression of their target genes by directly interacting with mRNA molecules (42). We predicted 53 putative target genes of EsrF sRNA using TargetRNA2 (43) (see Table S1A in the supplemental material). Comparative transcriptome sequencing was then performed to detect differences in global gene expression profiles between WT and DesrF strains, which identified a total of 244 genes (187 downregulated genes and 57 upregulated genes) as differentially transcribed (.2-fold) between the two strains (see Table S1B). Among these differentially transcribed genes, only flhB was also predicted by TargetRNA2 as a putative target of EsrF sRNA. FlhB is a flagellar export apparatus membrane protein known to induce effector transport (18); however, more recently, FlhB was also found to regulate other flagellar genes in Listeria monocytogenes (44,45).
Quantitative reverse transcription-PCR (qRT-PCR) assay results confirmed that flhB expression was downregulated by 2.5-and 2.3-fold in the DesrF mutant, respectively, compared to WT and DesrF1PesrF strains (Fig. 3A). Moreover, flhB expression was found to be downregulated by 2.0-and 2.1-fold in Dhfq and DesrF strains, respectively, compared to the WT (see Fig. S2A). To further confirm the activation of flhB by EsrF, we performed Northern blotting to detect flhB mRNA levels in WT, DesrF, and DesrF1PesrF strains. Significantly decreased flhB mRNA levels were detected in DesrF compared to WT, which were restored in the complemented strain ( Fig. 3B; see also Fig. S2B). The 18-nt interaction region between EsrF sRNA (CCGUUGAUUUAGGGCCGC) and flhB mRNA (AGCCUAAAUCCCGCC) 59 UTRs was predicted by TargetRNA2 (Fig. 3C).
RNA-RNA electrophoretic mobility shift assay (REMSA) was performed to verify the direct interaction between EsrF sRNA and flhB mRNA in vitro. EsrF, the EsrF complementary strand, and flhB mRNA were transcribed in vitro and purified. The REMSA reaction was performed with different concentrations of EsrF sRNA (0.5, 1, 2, and 4 mM) and a fixed amount of flhB mRNA (1 mM). EsrF sRNA-flhB mRNA complex bands were formed by both EsrF sRNA and flhB mRNA in 8% native polyacrylamide gel electrophoresis (PAGE), and the band intensity increased as the EsrF concentration increased from

(F) Colonization capacities of WT and
DesrF strains in rabbit colon. The data are presented as means 6 the SD (n = 3; ns, no significance; *, P # 0.05; **, P # 0.01; ***, P # 0.001). All P values were calculated using a Student t test. The 9-nucleotide crucial motif (GGGAUUUAG) for EsrF sRNA action was predicted by TargetRNA2 (Fig. 3C). To determine whether this motif represents the key sequence bound to flhB mRNA, we performed REMSA under the same reaction conditions, using a mutated EsrF (EsrF*) sRNA (the motif GGGAUUUAG was mutated to CUCCCTAAA) generated by in vitro transcription with flhB mRNA. We also performed REMSA using a mutated flhB (flhB*) mRNA (the motif CUAAAUCCC was mutated to GAUUUAGGG) with EsrF sRNA. The results showed no interaction between the EsrF* sRNA and flhB mRNA, or between EsrF sRNA and flhB* mRNA ( Fig. 3E and F), indicating that this motif is vital to the ability of EsrF sRNA to bind to flhB mRNA.
EsrF and flhB promote O157 motility. Considering that flagella are directly related to bacterial motility (11), a swimming motility assay was performed to determine whether esrF or flhB influence O157 motility. The growth radius was found to decrease significantly in the DesrF and DflhB mutants compared to the WT and the corresponding complement strains (  The data are presented as means 6 the SD (n = 3; ns, no significance; *, P # 0.05; **, P # 0.01; ***, P # 0.001). All P values were calculated using a Student t test.
grew at similar rates in Luria-Bertani (LB) medium (Fig. 2E), indicating that the decreased motility in DesrF and DflhB strains was not due to slower growth. Together, these results indicate that both esrF and flhB enhance O157 motility.
We further investigated whether flagellar synthesis was inhibited in DesrF and DflhB strains by transmission electron microscopy (TEM). The results showed that approximately 75% of the WT and complemented cells possessed flagella, whereas approximately 85 and 90% of DesrF and DflhB cells were aflagellate ( Fig. 4B; see also Fig. S3B). These results indicated that esrF and flhB promoted flagellar synthesis, which resulted in enhanced bacterial motility.
Although FlhB is generally regarded as a class 2 flagellar biosynthetic protein (46), recently flhB was reported to regulate the expression of several flagellar genes, including flaA (O157 fliC homolog), fliY, fliS, motA, lmo0695, and lmo0698, in Listeria monocytogenes (17). Hence, we performed qRT-PCR to determine whether flhB regulates the expression of flagellar genes in O157. The expression of flhD, flhC, fliA, and fliC was significantly downregulated in the DflhB strain compared to WT and DflhB1PflhB strains (Fig. 4C). Furthermore, the expression of flhB, flhD, flhC, fliA, and fliC was also significantly downregulated in the DesrF mutant compared to WT and DesrF1PesrF strains (Fig. 3A). These results indicate that EsrF sRNA promotes flhB-mediated expression of flagellar genes in O157.
To determine whether EsrF influences the motility of O157 directly by regulating flhB expression, we constructed an DesrF DflhB double mutant. Swimming motility and TEM results showed that the growth radius and flagellar synthesis decreased to similar levels in the DflhB and DesrF DflhB mutants compared to those of the WT strain ( Fig. 4A and B). These data indicate that esrF contributed to the motility of O157 by directly inducing the expression of flhB ( Fig. 3 and 4).
flhB promotes O157 virulence in vitro and in vivo. The effect of flhB on O157 virulence was investigated using the HeLa cell adhesion assay and rabbit colon colonization experiments. The results showed that DflhB cell adhesion decreased significantly compared with that of WT (2.9-fold) and DflhB1PflhB (2.7-fold) strains (Fig. 5A). Moreover, the FAS results revealed significantly reduced AE lesion formation in HeLa cells infected with the DflhB mutant (15%) compared to WT (50%) and DflhB1PflhB (55%) strains, with averages of 18.3, 5.2, and 15.3 pedestals on each cell infected by WT, DflhB, and DflhB1PflhB strains, respectively (Fig. 5B to D). Notably, WT, DflhB, and DflhB1PflhB strains grew at similar rates in vitro (Fig. 2E), indicating that the decreased cell adhesion of DflhB was not due to slower bacterial growth. In addition, an infant rabbit model of O157 infection revealed that significantly fewer bacteria were recovered from rabbit colons infected with the DflhB mutant (16.7-fold) than with the WT strain (Fig. 5E), thereby validating the inference exerting by flhB on O157 virulence in vitro and in vivo.
We next sought to determine whether EsrF directly influences O157 virulence by regulating flhB expression. HeLa cell adhesion and FAS results showed no further significant decrease in cell adhesion or pedestal formation for the DesrF DflhB strain compared to the DflhB mutant (Fig. 5A to D). Moreover, competitive infection assays in rabbits using DflhB and DesrF DflhB mutants revealed that the competitive index (CI) for DesrF DflhB versus DflhB strains in the colonization of rabbit colon was 1.09, indicating a similar colonization ability (Fig. 5F). Together, these data indicate that esrF contributes to O157 virulence by directly inducing the expression of flhB (Fig. 2 and 5).
NtrC negatively regulates esrF expression. The esrF gene sequence is located within the glnALG operon of the O157 genome; the expressions of all genes within this operon are regulated by NtrC (47). NtrC is the RR of the NtrC/B TCS, which regulates ;100 nitrogen (N) starvation genes to balance bacterial N metabolism (47). According to the reported NtrC binding sites in the glnA 59 UTR (NR1 to NR5) (48), we searched the esrF upstream region and found one 13-bp potential NtrC binding site (59-TCACGGATGAAGC-39; positions 2107 to 295 from the esrF transcriptional start site [TSS]; see Fig. S4A). The NtrC-His 6 protein was subsequently purified and used in a protein-DNA promoter EMSA, along with an amplified DNA fragment of the esrF promoter region to verify whether the phosphorylated form of NtrC protein (NtrC-P) directly binds to the esrF promoter. Slowly migrating bands were observed for the esrF promoter with increasing NtrC protein concentrations (Fig. 6A), and the intensities of these bands gradually strengthened as the concentration of NtrC protein increased from 0.25 to 1 mM. Meanwhile, no migrating bands were observed for the negativecontrol rpoS promoter with increasing NtrC protein concentrations (Fig. 6B), thus verifying the direct binding between the NtrC protein to the esrF promoter in vitro.
Ammonium-Regulated sRNA Promotes Virulence in O157: H7 ® described above. To determine whether the motif was necessary for binding to NtrC, we performed EMSAs using a mutated esrF promoter (the motif CACGGATGAAGCGATCGA was mutated to ACGCCTAGAAGCGATCGA) under the same reaction conditions. The results showed no interaction between NtrC and the mutated esrF promoter (Fig. 6D), indicating that the motif is vital for the binding ability of NtrC to the esrF promoter.
Negative regulation of esrF by NtrC is released in response to the high ammonium environment in rabbit colon. Microenvironmental cues are sensed by various TCSs to regulate bacterial gene expression (30). For instance, the NtrC/B TCS can respond to a low ammonium ion concentration (3 mM) and regulate gene expression; however, it is inhibited under high ammonium ion concentrations (47). The average ammonium concentration in rabbit colon was found to be 24.1 mM (Fig. 7A), representing a high ammonium environment, which is in agreement with the previously reported concentration of 20.8 6 8.0 mM in rabbit cecum (49).
We, therefore, investigated whether ammonium concentrations influence EsrF and flhB expression in vitro. Northern blotting results found that the abundance of O157 EsrF sRNA significantly increased in the presence of a high ammonium (24 mM) concentration compared to a low ammonium (3 mM) concentration ( Fig. 7B; see also Fig. S4B). The qRT-PCR results further demonstrated that flhB expression increased significantly in DntrC compared to the WT or DntrC1PntrC strains in DMEM (Fig. 7C). Further, the qRT-PCR results found that O157 flhB expression significantly increased with increasing ammonium concentrations (3, 6, 12, and 24 mM), while the reverse trend was observed for ntrC (Fig. 7D). However, the expression of esrF and flhB of the DntrC strain exhibited no change between high-ammonium and low-ammonium conditions ( Fig. 7B and E). This result indicates that repression of esrF by NtrC becomes released in a high-ammonium environment, resulting in upregulation of flhB expression in vitro.
In addition, in vivo qRT-PCR results showed that the ntrC expression of O157 WT strains in the colon of infected rabbits decreased significantly compared to that in the LB-cultured sample, indicating that ntrC expression was repressed in the high ammonium environment in vivo (Fig. 7F). Furthermore, in the colon of an infected rabbit, the expression of esrF and flhB of the O157 strain WT was enhanced compared to that in the LB medium-cultured sample. However, no difference was observed in the expression of esrF or flhB of the DntrC strain in the colon of an infected rabbit and LB medium-cultured samples (see Fig. S4C). Together, these data indicate that O157 sensed the high ammonium concentrations in the rabbit colon and responded by turning off the negative regulation of esrF by NtrC, thus inducing the expression of esrF and flhB to promote its motility and virulence in vivo.
The expression of glnA, which is located upstream esrF in the genome, is also negatively regulated by NtrC (50). The qRT-PCR results showed that the expression of glnA in the DntrC mutant increased significantly compared to that in WT and DntrC1PntrC strains (see Fig. S5A). Moreover, the expression of glnA in O157 significantly increased with increasing concentrations of ammonium (3, 6, 12, and 24 mM; see Fig. S5B). However, no changes were observed in the expression of glnA in the DntrC mutant when exposed to different ammonium conditions (see Fig. S5C). The trend observed for glnA expression was the same as that for esrF under the regulation by NtrC in highammonium conditions, indicating that NtrC represses the expression of esrF and glnA through interacting with two unique independent promoters, respectively ( Fig. 1D; see also Fig. S5).

DISCUSSION
O157 senses different host environmental signals, such as butyrate, fucose, and pH, via different TCSs that help to regulate virulence factor expression and promote bacteria colonization in the colon (51). For instance, a low biotin signal activates BirA-mediated expression of LEE genes to promote O157 virulence (52). In fact, most of the signal transduction pathways regulate O157 virulence by influencing the expression of LEE genes. However, the mechanisms by which microenvironmental cues in the colon regulate other non-LEE virulence genes remain unknown. In this study, we revealed a regulatory signal transduction pathway in which O157 utilizes a novel sRNA, EsrF, to sense high ammonium concentrations in the host colon to enhance bacterial motility and virulence. Specifically, the high ammonium concentration in the colon caused the repression of EsrF by NtrC to be lost, resulting in high levels of EsrF expression. EsrF then directly bound the 59 UTR of flhB mRNA to promote flhB expression, which enhanced the expression of other flagellar genes to facilitate flagellar biosynthesis. This mechanism functions to significantly promote O157 motility and colonization in the colon.
Flagella play diverse roles in bacterial pathogenesis, such as in the migration to an optimal site in the host, colonization, and survival at the site of infection (53). They act as adhesins to promote host cell adhesion and microcolony formation (13). Flagella were identified as adhesins when a fliC mutation was found to reduce EHEC virulence (54). For pathogens, it is necessary to upregulate the expression of flagellar genes during the early stages of infection. Here, we discovered that EsrF promoted O157 motility and host cell adhesion during these early stages. FlhDC and FliA are well-known master flagellar transcriptional activators in E. coli that regulate other flagellar genes (16). In our study, we found that EsrF positively regulated flagellar gene expression via FlhB in O157, which upregulated the expression of flhD, flhC, fliA, and fliC. This observation is consistent with the findings of a previous report that FlhB can regulate flagellar gene expression in Listeria monocytogenes (17). However, the detailed mechanism of how FlhB regulates flagellar genes is unknown.
The potential mechanism underlying the increased flhB mRNA abundance induced by EsrF-flhB mRNA interaction is likely related to the loss of Rho termination in O157. In bacterial cells, the Rho factor mediates termination of transcription at specific sites, which can be relieved by the Rho-specific inhibitor bicyclomycin (BCM) (55). Certain sRNAs in bacteria can anneal within 59 UTRs to inhibit premature Rho termination and activate the expression of corresponding genes (27). Hence, we analyzed the sequencing reads in the flhB 59 UTR region from the published raw data generated by Sedlyarova et al. from an RNA-seq analysis on E. coli, with or without BCM treatment (27) (see Fig. S6A). The upregulated expression and released termination within flhB 59 UTRs in response to BCM treatment was found to occur between sites 0 and 100 (counted from flhB translational start site), indicating that Rho factor targets are located within flhB 59 UTRs. Moreover, our identified EsrF binding sites were determined to be located at flhB 59 UTR positions 282 to 297, within the Rho termination region. Besides, we analyzed the transcription of flhB open reading frame (ORF) in O157 WT, with or without BCM treatment, by qRT-PCR assays and found that flhB expression in O157 WT with BCM treatment exhibited a 2.2-fold increase compared to that in O157 WT without BCM treatment (see Fig. S6B). Taken together, these results indicate that the interaction between EsrF with the flhB 59 UTR may release the Rho-dependent termination within the flhB 59 UTR, leading to increased flhB expression.
Inorganic N, produced by colonic microflora and digested food, is the secondmost important element after carbon in organisms and is a crucial constituent of proteins, nucleic acids, and cell walls. Various types of inorganic and organic N are assimilated by mammals, including ammonium (NH 4 1 ), nitrate (NO 3 -), nitrite (NO 2 -), urea, purine, and amino acids. Ammonium, widely used in various food additives, is the most effective N source for bacteria, which is assimilated directly into important biosynthetic reactions (47). Bacterial N metabolism is mainly regulated by the NtrC/B TCS, which regulates approximately 100 genes, mainly for N assimilation under N starvation (56). During N limitation, the NtrB/C TCS positively modulates Pseudomonas aeruginosa virulence by producing rhamnolipids (57) and exopolyphosphatase (Ppx) (58,59). NtrCregulated exopolysaccharides are involved in biofilm formation and the pathogenic interaction of Vibrio vulnificus (60). However, NtrB/C has not been reported to be related to the virulence of pathogenic E. coli. In this study, we have demonstrated that NtrC bound directly to the esrF promoter, sensing high ammonium concentrations in the colon to induce O157 motility and host cell adhesion.
Genome sequence analysis for 226 representative E. coli strains revealed that esrF is highly conserved and ubiquitous in different E. coli strain pathotypes. Specifically, the strains containing esrF include EHEC strains O157:H7, O157 Sakai, O157 Xuzhou21, enteropathogenic E. coli (EPEC) strains O55:H7, neonatal meningitis-associated E. coli strain RS218, urinary pathogenic E. coli strain CFT073, avian pathogenic E. coli strain LF82, as well as other clinical isolates of E. coli strains (see Fig. S7A and Table S2 in the supplemental material). However, esrF is absent in nonpathogenic E. coli strains, such as K-12, indicating that EsrF is a horizontal transfer-acquired regulatory sRNA of pathogenic E. coli, while esrF may be associated with the evolution from nonpathogenic E. coli to pathogenic strains.
To investigate whether homologous EsrF in other flagellated E. coli strains also regulate flhB expression, we selected three representative strains (EHEC O26:H11, EPEC O55:H7, and EPEC O86:H34) and deleted the esrF orthologous genes to construct corresponding deletion strains. Compared to the corresponding WT strains, the flhB expression level of these mutants was significantly reduced (see Fig. S7B), indicating that the EsrF-mediated mechanism for flhB regulation may be present in various pathogenic E. coli strains.

MATERIALS AND METHODS
Bacterial strains, plasmids, and primers. The bacterial strains, plasmids, and primers used in this study are listed in Table S3 in the supplemental material. Mutant strains were generated by substitution of the chloramphenicol or kanamycin resistance genes in plasmid pKD3 or pKD4 with sRNA or the relevant genes, respectively, by using the l Red recombinase system (61). The complementary strains were constructed by cloning the ORF and upstream promoter sequence of corresponding genes into the vector, pBluescript II SK(1) (62). The pET-ntrC strain was generated by cloning the ntrC gene into the downstream region of the His tag element in plasmid pET-28a. All the resulting clones were verified by PCR amplification and DNA sequencing.
Bacterial growth and cell culture conditions. O157 and its derivatives were cultured in LB medium or DMEM to simulate in vivo colon conditions. IPTG (isopropyl-b-D-thiogalactopyranoside) and antibiotics were used whenever necessary. After adding overnight subcultures into 96-well plates, the growth curves of all strains were measured over 24 h by using a plate reader. Each experiment was independently performed three times. The HeLa (ATCC CCL-2) cell line was grown in DMEM supplemented with 10% fetal bovine serum (Gibco) and penicillin-streptomycin and incubated at 37°C in a 5% CO 2 -containing atmosphere.
RNA isolation and qRT-PCR. To test the influence of virulence genes, strains were grown in DMEM up to the exponential phase (optical density at 600 nm [OD 600 ] ;0.6). To test the influence of ammonium concentrations on gene expression, strains were grown in M9 medium under low-ammonium (3 mM) and high-ammonium (24 mM) conditions. To quantify gene expression in vivo, the constitutively expressed red fluorescence protein (rfp) gene, inserted into the O157 chromosome, was used as the reference control for normalization. We also constructed a plasmid containing the esrF, flhB, and ntrC promoter transcriptionally fused to a green fluorescent protein (gfp) gene and introduced it into O157 cells with rfp. The expression of esrF, flhB, and ntrC was then indirectly determined by qRT-PCR analysis for the expression of gfp (to avoid interference of esrF, flhB, and ntrC homologs from intestinal commensal bacteria). The overnight cultures were separated into one sample for RNA extraction and another for oral gavage to a rabbit, the colon of which was dissected for RNA extraction. Total RNA was extracted using the TRIzol LS reagent (catalog no. 10296028; Invitrogen, Carlsbad, CA) and treated with RNase-free DNase I to eliminate genomic DNA contamination. qRT-PCR was performed as follows. Briefly, 2 mg of diluted, extracted RNA was converted to cDNA with Superscript IV VILO Master Mix (Invitrogen). Validated primers (see Table S3) and SYBR green were added to the cDNA, and the mixture was amplified using QuantStudio 5 (Applied Biosystems). Data were collected using QuantStudio Real-Time PCR Software v1.3, normalized to endogenous rpoA levels, and analyzed using the comparative critical threshold (C T ) method. Each experiment was performed in triplicate.
RNA-seq and other bioinformatic tools. Total RNA was purified by using an RNeasy minikit (Qiagen) and a Ribo-Zero rRNA removal kit (Epicentre Biotechnologies, RZNB1056). Libraries of RNA samples were generated by using an NEBNext R Ultra Directional RNA Library Prep kit for Illumina R (NEB). The libraries were sequenced using an Illumina HiSeq platform to generate paired-end reads. To search for potential sRNAs of O157, plot sequence reads were aligned to the reference sequence of O157 using Burrows-Wheeler Aligner software and SAMtools. TargetRNA2 was used to predict target genes of sRNA by searching the 59 regions (ca. nt 2120 to 120 relative to the start codon) of all O157 gene mRNAs for potential RNA duplex formation sequences (43).
Northern blotting. Northern blotting was performed using the DIG Northern Starter kit (Roche). RNA was separated on a 1.2% agarose gel containing 37% formaldehyde. The size of the RNA was determined by comparing it to the RNA Century-Plus Markers (Invitrogen). Gels were electroblotted onto Brightstar Plus nylon membranes (Applied Biosystems) and immobilized at 120°C for 30 min. Crosslinked membranes were prehybridized for 30 min in digoxigenin (DIG) Easy Hyb buffer. Single-stranded RNA probes were labeled with DIG added to fresh Easy Hyb buffer, and the blots were incubated with hybridization buffer overnight. After high-and low-stringency washes, the blots were further washed using the Wash and Block Buffer, and CDP-Star (Roche) was added as the substrate. Hybridization signals were visualized by using Amersham Imager 680, 5S rRNA being used as an internal control. ImageJ software was used to measure the band intensities. The specific DIG-labeled RNA probes used in the Northern blotting are presented in Table S3.
Exonuclease digestion of RNA. RNA samples were prepared from O157 WT grown overnight in LB. Treatment of RNA samples with terminator 5-monophosphate-dependent exonuclease (TEX; Epicentre Biotechnologies) was performed as described previously (38). Total RNA (10 mg/sample) samples were incubated with TEX in a final 20-ml reaction volume containing 2 ml of 10Â reaction buffer (500 mM Tris-HCl [pH 8.0], 20 mM MgCl 2 , 1 M NaCl), 1 ml of RNasin (40 U; Promega), and 1 ml of TEX for 1 h at 30°C. The reaction mixture was then purified and analyzed by Northern blotting as described above.
Ammonium-Regulated sRNA Promotes Virulence in O157: H7 ® 59 and 39 RACE assay. RACE was performed with the 59/39-RACE system (Invitrogen). For 59-RACE, 5 mg of RNA was reverse transcribed using an sRNA-specific antisense primer and SuperScript reverse transcriptase (Invitrogen). cDNA was then purified, dC tailed, and used as a template in a PCR with the Abridged Anchor Primer (AAP) and a nested gene-specific primer. 39-RACE was performed by ligating a poly(A) tail using a poly(A) polymerase tailing kit (Epicentre Biotechnologies) before reverse transcription. Specific cDNAs were then directly amplified by PCR using an anchor primer (AP) that targets the poly(A) tail region and a gene-specific primer that anneals to a region of known sRNA sequence. PCR products were cloned into the pEASY-T1 simple cloning vector (TransGen) before sequencing.
Cell adhesion assay. Subcultured bacteria were grown in DMEM until the OD 600 was 0.6; the OD 600 was then adjusted to 0.3 with DMEM. Next, bacterial cells (2 ml each well) were added to HeLa cell monolayers in 6-well plates at a multiplicity of infection (MOI) of 100. To avoid motility-related effects, all bacteria were centrifuged onto the surface of cells and incubated for 1 h at the early infection stage. The cells were then washed, lysed, and plated on LB agar plates. The adhesion rates were calculated as percentages of the number of bacteria recovered relative to the total bacteria inoculated. All assays were performed in triplicate.
Fluorescein actin staining. Fluorescence actin staining was performed as previously described (63). Human cervical adenocarcinoma (HeLa) cells were grown on coverslips to 60% confluence and infected by bacteria as described for the cell adhesion assay. After infection, the coverslips were washed and fixed with 4% formaldehyde, permeabilized with 0.2% Triton X-100, and treated with fluorescein isothiocyanate (FITC)-labeled phalloidin to visualize actin accumulation, followed by propidium iodide (PI) to visualize the DNA. The coverslips were mounted on slides and observed under a Leica TCS SP8 microscope. The AE lesions formed in each cell were enumerated for 50 HeLa cells; three slides were observed for each strain in each experiment.
Infant rabbit model of O157 infection. To prepare the inoculum, bacteria were grown overnight in LB broth, harvested by centrifugation, resuspended in sterile phosphate-buffered saline, and adjusted to a cell density of ;10 9 CFU ml 21 . Three-day-old New Zealand White rabbits were intragastrically inoculated with ;5 Â 10 8 CFU/90 g (body weight) of WT or mutant in a 0.5-ml inoculum, which was followed by the administration of 2.5 ml of sterile 0.85% saline solution to ensure delivery of the entire inoculum. Rabbits were euthanized 48 h postinfection. At necropsy, the intestinal tract from the duodenum to the anus was removed to collect samples for microbiologic analyses. Rabbit competition experiments were performed as follows: DflhB and DesrF DflhB mutant strains, both in logarithmic phase, were mixed at a 1:1 ratio for oral inoculation of rabbits, as described above. The CI value of the DflhB mutant versus the DesrF DflhB mutant was calculated. All experiments were performed in triplicate.
Motility assay and transmission electron microscopy. Overnight cultures were adjusted to an OD 600 of 0.2, and then 5 ml was stab inoculated onto 0.35% semi-LB-agar plates. The agar plates were incubated at 25°C for 12 h, and the diameter of the swimming zone around the inoculation site was measured. TEM was performed to observe the flagellum morphology on the surface of O157, as previously described (17,64). The strains were cultured in LB plates overnight. Single colonies were resuspended in 50 ml of medium; 10 ml was then dropped onto and absorbed for 5 min into carbon-stabilized Formvar supports on 200-mesh copper grids. Cells were then stained by submerging the grids for 5 min in 2% (wt/vol) uranyl acetate and imaged using a JEM-1400 Plus transmission electron microscope (JEOL) operating at 100 kV and fitted with a high-sensitivity real-time charge-coupled device camera. All strains were tested in triplicate, and the number of flagella were counted per 30 bacteria.
Purification of NtrC protein and protein-DNA EMSA. NtrC-6ÂHis protein was expressed using BL21(DE3)-containing pET-NtrC and purified by using a MagneHis protein purification system (Promega). Protein-DNA EMSAs were performed as described previously (65). EsrF promoter regions were amplified and purified. NtrC protein shift assays were performed by incubating EsrF promoter fragments (15 nM) at 37°C for 30 min with various concentrations of NtrC-6ÂHis protein (0 to 3 mM) in a 20-ml solution containing bandshift buffer (24 mM Tris-HCl [pH 7.5], 80 mM NaCl, 0.1 mM EDTA, 1 mM dithiothreitol [DTT]). In addition, 30 mM acetyl phosphate (AcP) was added as the donor to phosphorylate NtrC to NtrC-P. The samples were loaded on an 8% polyacrylamide gel. The DNA fragments were stained for 10 min with StarGreen (Genstar) and visualized by UV transillumination. ImageJ software was used to measure the band intensities.
Dye primer-based DNase I footprinting assay. The EsrF promoter (7.5 nM) was amplified with a forward primer (with a 6-FAM modification at the 59 end) and a reverse primer and then incubated with 1 mM NtrC and 1 mM bovine serum albumin (BSA) in bandshift buffer. The protein-DNA mixtures were then partially digested with 0.05 U of DNase I for 5 min at 25°C, quenched by using 0.25 mol/liter EDTA, and purified. Control samples were prepared with 2 mM BSA. All genotype samples were analyzed using the ABI 3730 DNA analyzer.
Ammonium concentration measurements. The colons from standard 3-day-old New Zealand White rabbits were collected, washed, and homogenized using a Tissueprep Lyser. The samples were pelleted, and the supernatants were collected. The ammonium concentrations of the serially diluted supernatants were measured by using a Quantofix ammonium test kit (Merck Millipore, UK) according to the manufacturer's instructions.
Phylogenetic analysis. Orthologous groups were identified using OrthoFinder (67) by which all nucleotide sequences were compared using a BLASTN all-against-all search with an E value cutoff of ,10 24 . Nucleotide sequences used to construct the phylogenetic tree were aligned in MAFFT (68), and a maximum-likelihood tree was constructed in PhyML and FigTreev1.4.3 based on the GTR model of nucleotide substitution with c-distributed rates among sites.
Ethics statement. All animal experiments were carried out according to the standards set forth in the Guide for the Care and Use of Laboratory Animals published by the Institute of Laboratory Animal Resources of the National Research Council (United States). The experimental protocols were approved by the Institutional Animal Care Committee at Nankai University. Every effort made was to minimize animal suffering and to reduce the number of animals used.
Statistical analysis. Statistical analysis was performed using MedCalc 15.6; figures were drawn by GraphPad Prism 8 and integrated by using Adobe Illustrator CC 2020. All data are expressed as means 6 the standard deviations (SD). Differences between two groups were evaluated using a two-tailed Student t test or a Mann-Whitney U test. P values of #0.05, 0.01, or 0.001 were considered statistically significant (*), highly significant (**), or extremely significant (***), respectively.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.