The type 6 secretion system immunity protein RhsFI has been repurposed for small RNA regulation in pathogenic Escherichia coli

Abstract

RNA-binding proteins (RBPs) play key roles in regulating virulence gene expression in pathogenic enterohaemorrhagic Escherichia coli (EHEC). To capture the RNA-binding proteome in EHEC, total RNA-associated protein purification (TRAPP) was employed and identified 443 RNA-associated proteins. These included 35 encoded within pathogenicity islands, with RNA-binding confirmed in vitro for four, including RhsFI. In vivo RNA targets for RhsFI included regulatory small RNAs (sRNAs) encoded in 3′ UTRs. The rhsF–rhsFI locus encodes toxic effector and immunity proteins of the type 6 secretion system (T6SS). T6SS expression is repressed in EHEC, but internal transcription initiation generates a transcript encoding a truncated toxin (RhsF_int) and downstream RhsFI. RhsF_int has structural homology to nucleases and modelling predicted that RhsF_int can dimerize with RhsFI, forming an extended RNA interaction face. Deletion of rhsF_intrhsFI alters the expression of multiple operons in which embedded sRNAs are direct RhsFI targets. Loss of RhsF_int–RhsFI significantly reduced adhesion to bovine cells in vitro, altered cell morphology and increased acid resistance. Our results indicate that the T6SS immunity protein, RhsFI, binds sRNAs in vivo and disrupts the sRNA regulatory network in EHEC to promote host colonization.

Graphical abstract.

Introduction

From transcription to degradation, RNAs are bound by RNA-binding proteins (RBPs) that control their function in the cell. Recent high throughput analyses of the RNA-binding proteome (RBPome) in eukaryotes and prokaryotes have revealed an extraordinary wealth of RNA-interacting proteins, suggesting rich and complex networks of interactions that determine RNA fate in the cell [1–6]. Canonical RNA-binding domains (RBDs) facilitate a large proportion of these interactions; however, RBPs often lack recognizable RBDs and can have dual “moonlighting” functions [7]. Proteins like aconitase, that have well described functions in metabolism but regulate gene expression through RNA interactions [8], are emblematic of these multi-functional RBPs.

Bacterial RBPs play key roles in controlling gene expression at the transcriptional and post-transcriptional levels. This extends to pathogenesis, where expression of virulence factors is frequently coordinated post-transcriptionally. The human pathogen Enterohaemorrhagic Escherichia coli (EHEC) has a mosaic genome where the majority of genes required for virulence have been horizontally acquired on pathogenicity islands (termed S-loops in strain Sakai) that are inserted into a core genome largely shared with commensal E. coli strains [9]. The small RNA (sRNA) chaperone Hfq controls gene expression within both the core and accessory genome including the major virulence factors, Shiga toxin and the type 3 secretion system (T3SS) [10–12]. ProQ has also been shown to bind sRNAs and contributes to virulence gene expression in avian pathogenic E. coli and Salmonella by affecting the stability of a subset of sRNAs [13, 14]. sRNAs that confer a fitness advantage to EHEC are encoded within the pathogenicity islands and regulate virulence factors post-transcriptionally [15–18]. Many, but not all, EHEC regulatory sRNAs are bound by Hfq, suggesting that additional RBPs may contribute to virulence in EHEC.

Among the horizontally acquired regions of the genome is a type 6 secretion system (T6SS) that functions as a molecular “harpoon” to inject toxic effectors into neighbouring cells [19]. Effector toxins are encoded within the T6SS locus (S-loop 14) and within seven orphan effector-immunity loci [20]. The T6SS of EHEC does not appear to defend the cell from bacterial competitors, but deletion of the T6SS significantly attenuates EHEC in a mouse model of infection [21]. T6SS expression is repressed by the DNA-binding protein H-NS and in a Δhns background, the T6SS effector KatN (encoding calatase) increases EHEC survival in cultured macrophages suggesting that the T6SS is used to manipulate host cells.

Bioinformatic analysis has identified seven orphan T6SS effectors in EHEC str. Sakai [20]. Six belong to the Rhs-family and are modular, consisting of a cocoon-like RHS domain that envelops a variable C-terminal toxin domain [22, 23]. The cocoon and toxin domains are separated by an autoprotease sequence that self-cleaves the toxin domain from the cocoon. Work in Salmonella indicates that the C-terminal toxin domain and downstream immunity protein of Rhs-family effectors can be expressed independent of the RHS domain, functioning as a toxin–antitoxin (TA) system in the bacterial cytoplasm [24].

While Hfq coordinates expression of a large number of virulence genes, we speculated that additional RBPs within the pathogenicity islands may contribute to EHEC gene regulation. To identify novel RBPs we used total RNA-associated protein purification (TRAPP) [5] to capture the RBPome of EHEC under virulence-inducing conditions. TRAPP enriched 443 proteins in EHEC including 35 encoded within pathogen-specific, horizontally acquired regions of the genome. The T6SS immunity protein RhsFI is the most highly enriched protein within the pathogen-specific regions, and we demonstrate that RhsFI binds regulatory sRNA. RhsFI both stabilizes and destabilizes sRNA targets and has pleiotropic effects on gene expression including genes for acid tolerance, maltose transport, and drug efflux, often through interactions with sRNA embedded within the operon. Deletion of rhsF rhsFI reduces adhesion to cultured cells. Our results point to a T6SS immunity protein that has been re-purposed for sRNA regulation and precipitates broad changes in the transcriptome to promote virulence in EHEC.

Results

Identification of RNA-binding proteins in enterohaemorrhagic E. coli

TRAPP is an organic extraction-based method of capturing the RBPome [5]. To identify novel RBPs in EHEC O157:H7 str. Sakai, TRAPP was performed on cultures grown in virulence inducing media (supplemented MEM–HEPES that induces expression of the T3SS required for host cell adhesion [25–28]). The protocol was performed in triplicate for both UV-crosslinked and non-UV-crosslinked samples. Aliquots were taken before nuclease treatment and separated on an agarose gel to confirm the presence RNA–protein complexes in the UV-crosslinked samples. Following nuclease treatment, extracted proteins were separated on a 4%–20% gradient polyacrylamide gel before MS analysis. Coomassie staining of the polyacrylamide gel showed enrichment of proteins in the UV-crosslinked samples (Supplementary Fig. 1). Each lane was divided into three fractions and gel fragments were subjected to in-gel digestion using trypsin before LC/MS–MS analysis. This identified 443 significantly enriched proteins ( ≥ 2-fold enriched relative to non-crosslinked, FDR < 0.05) (Fig. 1A and Supplementary Table 1). Enriched proteins included CsrA and Hfq, as well as other expected RBPs, such as the chaperone ProQ and the cold shock proteins CspA, CspC, CspD, and CspE (Fig. 1A).

Figure 1. TRAPP enriches RBPs in EHEC.

(A) Volcano plot of TRAPP enrichment (log₂[fold change]) and adjusted P-values (log₁₀[P_adj]). Select known RBPs are indicated in blue and labelled. Horizontal dashed line indicates P_adj = 0.05. Vertical dashed lines indicates log₂(FC) = 2. (B) Ontological terms associated with proteins statistically significantly enriched by TRAPP. (C) Cumulative distribution function comparing the distribution of TRAPP enrichment (log₂[fold change]) for proteins with a predicted RNA-binding domain (RBD, red) and without a predicted RBD (yellow). P-value calculated using a two-sided Kolmogorov–Smirnov test.

TRAPP enriches for known RNA-binding domains and RNA-associated ontologies

Gene ontology (GO) analysis of our TRAPP data showed an enrichment for GO terms associated with RNA processes such as “RNA processing”, “RNA modification”, and “Translation”. The most significantly enriched term (P < 10⁻³⁵) was for “ribosome biogenesis”, consistent with recovery of bona fide RBPs (Fig. 1B).

To predict RBDs in the EHEC proteome and assess enrichment by TRAPP, we used the computational pipeline APRI-COT that was developed for sequence-based identification and characterization of RBDs and RBPs [29]. Of a total of 1388 proteins that were detected by LC-MS/MS, 795 proteins had a statistically significant difference (FDR < 0.05) in either the UV-crosslinked or non-crosslinked samples. Of these, 274 and 259 proteins contained a putative RBD based on CDD and InterProScan predictions, respectively. Comparing the cumulative distributions of TRAPP enrichment scores for proteins containing a predicted domain showed that proteins with a predicted RBD were more highly enriched by TRAPP (Fig. 1C).

We next compared our EHEC TRAPP data with published commensal E. coli K-12 RBPome datasets generated by orthogonal organic phase seperation (OOPS) and TRAPP [4, 5]. Of the enriched proteins, 77% were also detected in the commensal E. coli K-12 TRAPP dataset and 43% overlap with proteins recovered from at least three replicates of OOPS (Supplementary Fig. S2 and Supplementary Table S1). In the previous E. coli K-12 TRAPP dataset, significantly more proteins were enriched by crosslinking and we attribute this to inclusion of SILAC labelling in the previous experiments, which enhances the sensitivity in quantifying enrichment. Construction of EHEC auxotrophs suitable for SILAC labelling may increase the sensitivity of our TRAPP analysis. Despite the reduced dataset size, the results demonstrate a strong overlap between the EHEC TRAPP dataset and published RBPome datasets for commensal E. coli.

Confirmation of novel RNA-binding proteins in pathogenicity islands

Of the 443 significantly enriched proteins identified by TRAPP, 35 (8%) are encoded by genes located within the pathogenicity islands of EHEC str. Sakai (Supplementary Table S1). These included LsoA, an endoribonuclease encoded on the virulence-associated plasmid pOSAK1 [30, 31]. A further 16 proteins were associated with putative DNA-binding functions or DNA-binding domains, including the phage cI repressor and DNA methylases. Of the remaining 19 EHEC-specific proteins, 13 were uncharacterized, and 2 were effector proteins secreted by the T3SS. The haem-binding proteins ChuX, ChuW, and ChuS from the Chu iron acquisition system were also recovered, possibly through haem binding to silica [32].

To verify that EHEC-specific proteins enriched by TRAPP were bona fide RBPs, a polynucleotide kinase labelling assay (PNK assay) was employed. PNK labelling allows detection of RNAs crosslinked to purified proteins by radiola-belling the RNA 5′ end. We analysed the non-LEE encoded effector protein EspY2 and uncharacterized proteins ECs0604 (hereafter RhsFI) and ECs1762, which lack detectable RBDs, and ECs1163, a predicted RBP of unknown function encoded on the Stx2 phage. All of these were significantly enriched in TRAPP. We also included ECs1184 from the Stx2 phage, which is proposed to be required for stable maintenance of the Sp5 lysogen [33], but fell short of statistical significance (FDR =0.065).

Genes encoding the candidate proteins were cloned into the arabinose-inducible plasmid pBAD24 and tagged with a His6-TEV-FLAG (HTF) dual-affinity tag. These were transformed into EHEC str. Sakai to assess RNA binding using a PNK assay. Cultures were UV-crosslinked with 500 mJ of 254 nm UV-C, and proteins were purified using M2 anti-FLAG resin and a denaturing purification step using IMAC. The 5′ ends of crosslinked RNAs were radiolabelled with ³²P before elution from Ni-NTA resins and separation on 4%–12% gradient polyacrylamide gels. The four proteins that were significantly enriched by TRAPP were radiolabelled above background levels (Fig. 2 and Supplementary Fig. S3). ECs1184 was detected on the autoradiogram in MEM–HEPES grown EHEC str. Sakai, but the signal detected was similar to sfGFP, which was used as a negative control suggesting ECs1184 is not a strong RBP (Fig. 2). These results confirm that TRAPP enriches novel pathogen-specific RBPs in EHEC.

Figure 2. RNA end labelling of purified RNA–protein complexes.

Autoradiograms of purified, UV-crosslinked RNA–protein complexes end-labelled with ³²P using PNK. RBPs were overexpressed in cells grown in LB (A) or MEM-HEPES (B). Lanes corresponding to each protein are noted above.

RhsFI (ECs0604) is a T6SS immunity protein that binds regulatory RNAs

ECs0604 (RhsFI in Fig. 3A) is encoded within the EHEC pathogenicity island S-loop 44. The coding sequence (CDS) is positioned downstream of RhsF, a predicted Rhs-domain containing T6SS effector protein that is predicted to be injected into target cells [20]. Based on synteny with T6SS effector loci in other bacteria, ECs0604 has been designated RhsFI [20] (z0704 in EHEC str. EDL933) and is proposed to be the immunity protein that binds RhsF [20], potentially protecting the bacterium against effector activity (Fig. 3A).

Figure 3. UV-crosslinking of RhsFI indicates binding within 3′ UTRs.

(A) Schematic representation of the T6SS accessory effector-immunity locus located in S-loop 44. (B) RNA classes recovered by UV-crosslinking Hfq, RhsFI, or the untagged parental EHEC str. Sakai. Total reads recovered for each replicate experiment (indicated _1 and _2) are presented. (C) Venn diagram showing the number of unique genes with statistical significant (FDR < 0.05) RhsFI-binding sites in the CDS, 5′ UTR, 3′ UTR, or within sRNAs. Overlapping regions of the Venn diagram indicate that a binding site within an mRNA maps to multiple regions, or mRNA has multiple binding sites.

To identify target RNAs bound by RhsFI, we performed UV-crosslinking and high-throughput sequencing (CRAC) of our tagged, overexpression strain EHEC pBAD24::rhsFI-HTF and empty vector control (Fig. 3B). Significantly enriched binding peaks (FDR < 0.05) were found within 24 transcripts. Notably, RhsFI bound the rhsFI mRNA consistent with an auto-regulatory loop described for other post-transcriptional regulators [34–36]. Binding sites were mapped in mRNA CDSs [17], 3′ UTRs [13] and 5′ UTRs [7], and also in sRNAs [7] (Fig. 3C). Crosslinking-dependant deletions in CRAC sequencing reads indicate the precise site of RNA-protein crosslinking and were distributed within the CDSs and 3′ UTRs. Lower confidence binding sites (618 with FDR < 0.1 representing 342 genes) showed a higher proportion of CDS, and a similar enrichment of 3′ UTRs and sR-NAs (Supplementary Table S2 and Supplementary Fig. S4). We concluded that ORFs and 3′ UTRs are preferential targets for RhsFI binding (Supplementary Table S2).

RNA targets for RhsFI substantially overlapped with the sRNA chaperone Hfq: 41.2% of statistically significant RhsFI peaks (FDR < 0.1) overlapped statistically significant Hfq peaks (Supplementary Fig. S4). While there is no ProQ crosslinking data available for EHEC, several known ProQ targets were among the highest enriched sRNAs [37]. Highly enriched 3′ UTR binding sites for RhsFI included sRNAs; the malM 3′ UTR, RaiZ sRNA, and RybA encoding the small protein MntS (Fig. 4A–D).

Figure 4. RhsFI binds regulatory RNAs and the rhsFI mRNA.

(A–D) RNA-seq datasets for (A) rhsF_int-rhsFI (showing from the internal transcription start site), (B) RaiZ sRNA, (C) the 3′ UTR sRNA of malM, and (D) RybA sRNA. Genomic coordinates and a schematic of the RNA is the shown on below (bent arrows indicate transcription start sites and pins indicate RNA 3′ ends). RNA-seq datasets are indicated on the y-axis: dRNA-seq (without terminator exonuclease [TEX] treatment, reads, light green; light green arrow indicates step corresponding to RaiZ 5′ triphosphate), and dRNA-seq (TEX treated, reads, green) indicates RNA 5′ ends and transcription start sites, Term-seq (reads [red]) indicates RNA 3′ ends, RhsFI_1 and RhsFI_2 UV-crosslinked reads (RPM [blue], deletions are indicated by bar graph and on the right-hand axis), Hfq UV-crossslinked reads (RPM [orange]), reads UV-crosslinked in the untagged control EHEC str. Sakai (RPM [gray]). (E and F) Dot hybridization assays for RhsFI interactions with (E) rhsFI mRNA, (F) RaiZ sRNA, (G) the 3′ UTR sRNA of malM, (H) RybA sRNA, and (I) the 3′ UTR of the S. aureus mRNA vigR. The fraction of radiolabelled RNA retained with RhsFI (bound to nitrocellulose) is indicated. Dissociation constants (K_d) for each binding assay is indicated above the plot. Plots are from representative data from a minimum of two replicate experiments.

To identify RNA sequences that may bind RhsFI we looked for k-mers that were enriched with the top 100 statistically significant binding peaks using pyMotif [38]. A logo was constructed from the top 20 k-mers using MEME and GLAM2 (Supplementary Fig. S4). Neither approach identified a definitive motif, but G-rich elements predominated among the top k-mers.

We conclude that RhsFI binds RNAs with preferences for G-rich motifs and for 3′ UTRs that generate regulatory sRNAs, many of which are also targets for Hfq and ProQ.

In vitro confirmation of RhsFI target RNA binding

We next looked to confirm RhsFI interactions with target RNAs in vitro. We constructed a C-terminally tagged RhsFI with His₆, purified the protein under native conditions using immobilised metal affinity chromatography (IMAC) and SEC, and confirmed the identity of the protein using LC-MS/MS (Supplementary Fig. S5). Our UV-crosslinking analysis indicated that RhsFI binds Hfq- and ProQ-dependant sRNAs. We therefore selected RNA targets shared with Hfq and ProQ, and as well as RNAs exclusively bound by RhsFI. RNA sequences were in vitro transcribed for the malM 3′ UTR (ProQ and RhsFI target), RaiZ sRNA (Hfq, ProQ, and RhsFI target), and RybA (ProQ and RhsFI target). The rhsFI mRNA, bound by RhsFI (but not Hfq), was also included. The 3′ UTR of the vigR regulatory mRNA [39] from Staphylococcus aureus was used as a negative control.

Our initial attempts to use electrophoretic mobility shift assays (EMSA) were unsuccessful as RNA–RhsFI complexes were retained in the well of the native polyacrylamide gel. SECS analysis indicated that the protein was monomeric in the presence of 10% glycerol (Supplementary Fig. S5) and this was included in all analyses. We note that the pI of the fusion is 8.8 suggesting that the RNA–RhsFI–His₆ complex is positively charged in our running buffer (TBE pH 8.4) preventing migration into the gel. To assess RhsFI–RNA interactions we therefore adopted a filter-binding assay. Radio-labelled in vitro transcribed (IVT) RNA targets were incubated with increasing RhsFI concentrations and retention of RhsFI–RNA complexes on nitrocellulose membranes was assessed by phosphor-imaging and densitometry. RhsFI bound RNA targets with a range of affinities. The sRNAs that associate with ProQ, RaiZ, and malM 3′ UTR had the highest affinity (K_d =85.1 and 90.8 nM, respectively), followed by rhsFI mRNA (K_d =109.4 nM), RybA (K_d =329.5 nM), and the negative control vigR 3′ UTR (K_d =749.1nM) (Fig. 4 and Supplementary Fig. S6). The dissociation constants for ProQ and Hfq are reported in the low nanomolar range (ProQ-malM 3′ UTR K_d =8.4 nM, Hfq-RaiZ K_d =15 nM [40, 41]) and are an order of magnitude lower than RhsFI indicating that it may be outcompeted by these global RNA chaperones at shared sites. We anticipate that RhsFI is in complex with the toxin RhsF_intin vivo and speculate that the dimer has improved RNA-binding properties (see below).

These results confirm our in vivo UV-crosslinking data and indicate that RhsFI is able to bind to Hfq- and ProQ-associated 3′ UTR-derived sRNAs in vivo and in vitro.

Predicted RhsFI structure suggests a novel RNA-binding domain

RhsFI does not contain a recognized RBD, or structural similarity to known RNA-binding proteins, as assessed using the DALI server [42] and TM-align [43], suggesting a novel mode of RNA interaction. Alphafoldv2.1.1 [44] predicted a high confidence structure for RhsFI: the majority of pLDDT scores are over 90, and PAE scores across the length of the protein indicate a highly defined structure (Supplementary Fig. S7). The model predicts that the N-terminal domain adopts a bent α helix followed by seven antiparallel β-sheets arranged around part of the α helical N-terminus (Fig. 5A). Calculation of surface electrostatics of the surface by APBS [45] a large, positively charged patch on the C-terminal side of the β-sheet as a potential RNA interaction site (Fig. 5B).

Figure 5. Structure prediction of RhsF_int and RhsFI reveals an extended positively charged surface.

(A) Alphafold structure prediction for RhsFI showing the backbone ribbon diagram superimposed in the surface shell. Confidence values for predicted structure are presented in Supplementary Fig. S7. (B) Surface electrostatic representation of RhsFI and rotated to show the large positively charge surface predicted to bind RNA (indicated in blue). (C) Alphafold-multimer prediction of the RhsF_int–RhsFI dimer showing the backbone ribbon diagram superimposed on the surface shell. Confidence values for predicted structure are presented in Supplementary Fig. S7. (D) RhsF_int–RhsFI dimer coloured by monomer (RhsF_int =blue, RhsFI =gold). (E) Surface electrostatics of the RhsF_int–RhsFI dimer rotated to show each side (left, middle) and top view (right) showing the large positively charged surface formed in the top cleft that spans both monomers. RNA is predicted to binding within the positively charged cleft (see Supplementary Fig. S7).

RhsF–RhsFI and RhsG–RhsGI resemble toxin-antitoxin systems that are expressed independently of the EHEC T6SS

T6SS immunity proteins generally only suppress the activity of a cognate toxin, so the mRNA and sRNA binding activity of RhsFI was unexpected. In EHEC, T6SS expression is repressed by the DNA-binding repressor H-NS [21], and co-ordinated repression of RhsFI would be anticipated. To assess whether RhsFI functions independently of the T6SS, we analysed RNA-seq and dRNA-seq data that identifies RNA 5′ ends (including promoters). This confirmed the almost complete repression of T6SS core genes encoded by S-loop 14 (ecs0212-ecs0245) (Supplementary Fig. S8). Within the T6SS effector locus located in S-loop 44, the transcription start site for full-length rhsF was similarly inactive. In contrast, robust transcription was apparently initiated from a promoter located within the 3′ region of rhsF, upstream of rhsFI (Fig. 6A).

Figure 6. rhsF_int-rhsFI are transcribed from an internal promoter independent of the T6SS.

(A) RNA-seq data at the accessory T6SS effector-immunity locus, S-loop 44. Genomic coordinates and a schematic of the RNA is the shown on below (bent arrows = transcription start sites, blue bars = Rhs autocleavage motif, orange shading = region cloned for rhsF_int and rhsG_int). RNA-seq datasets are indicated on the y-axis: dRNA-seq (without terminator exonuclease [TEX] treatment, reads, light green]) and dRNA-seq (TEX treated, reads, green) indicate RNA 5′ ends). Total RNA-seq coverage (RPM, blue). Primary transcription start sites are indicated by bent arrows (below). (B) Spot dilution assays of RhsF_int and RhsG_int expressed in E. coli str. DH5α. Induction with IPTG is inducted (left) (C) Western blot analysis of the RhsF_int–FLAG expression. Arrow head indicates the band corresponding to RhsF_int–FLAG.

Rhs-family effectors are modular and consist of a RHS domain that forms a large cocoon that envelops a variable C-terminal toxin domain [22, 23] (Fig. 3A). The toxin domain is preceded by an autoprotease sequence that self-cleaves the toxin domain from the cocoon. The internal promoter is positioned upstream of the autoprotease site and transcribes the C-terminal toxin domain and rhsFI immunity gene (Fig. 6A). An additional internal transcription initiation site was found at a similar position in the downstream T6SS effector-immunity pair rhsG-rhsGI (ecs0603-ecs5380).

We hypothesized that internal transcription initiation of rhsF might allow expression of the C-terminal toxin domain and downstream RhsFI immunity protein. We cloned the toxin domain from the internal transcription start site to the stop codon (here termed rhsF_int) and placed it under the control of the P_LlacO-1 promoter. We similarly placed the internal transcription start site of rhsG under the control of P_LlacO-1 (here termed rhsG_int). RhsF_int and RhsG_int were transcribed from a P_LlacO-1 promoter, and toxicity was assessed using spot dilution assays (Fig. 6B). Transcription of rhsG_int from the internal transcriptional initiation site was inhibitory, consistent with expression of RhsG_int toxin domain from this sub-genic transcript.

Transcription of rhsF_int was not toxic in E. coli (Fig. 6C). We verified that RhsF_int was translated and expressed from the internal transcription start site using Western analysis on an rhsF_int-FLAG construct (Fig. 6C).

Collectively, our data indicate that sub-genic transcription initiation from within rhsF and rhsG uncouples the effectorimmunity pairs from T6SS expression. Transcription from the internal promoter allows synthesis of the C-terminal sub-domain and likely the downstream immunity protein.

RhsFint has structural homology to nucleases and the heterodimer has an extended positively charged surface

To identify the activity of the RhsF_int toxin domain we used Alphafoldv2.1.1 [44] to predict the structure and DALI ⁴⁰ to identify structural homologues. Homologues of the toxin domain included many endonucleases with the closest match being the deoxyribonuclease BpuSI (PDB accession: 3s1s). Alignment of the structures (DALI) and comparing vacuum electrostatics (APBS plugin in pyMol) indicated a conserved negatively charged patch corresponding to the metal ion binding site of BpuSI, suggesting that RhsF_int may be a metal ion-dependant nuclease. Our earlier results suggested that RhsF_int was not toxic when overexpressed and we examined the active site for mutations that may indicate inactivation of the nuclease. RhsF_int retains conserved acidic residues (D104 and D127) predicted to coordinate the manganese ion in BpuSI and related nuclease FokI [46] (Supplementary Data). A third catalytic lysine residue in BpuSI/FokI is not conserved in RhsF_int (T129), but we note this residue is well removed from the metal ion in several related structures (>6 Å), and not conserved in related structures solved without a metal ion. It is unclear from the related structures whether RhsF_int T129 would impact metal ion binding or nuclease activity (see Supplementary Data for further information).

Many immunity proteins and antitoxins bind their cognate toxin to inhibit activity. We used Alphafold-multimer to assess the potential interactions between RhsF_int and RhsFI. A dimer was predicted with high confidence (Fig. 5C and D, and Supplementary Fig. S7B). Notably, this extended the positively charged surface of RhsFI that is predicted to bind RNA (Fig. 5E, modelled with malM3′ UTR sRNA in Supplementary Fig. S7C and D, and discussed further in Supplementary Data). The positively charged cleft is distant from the predicted metal ion binding site suggesting RNA binding may be independent of nuclease activity. Across 20 predictions of malM3’UTR sRNA interactions with RhsF_int–RhsFI using Boltz [47], the sRNA was consistently positioned in the positively charged patch (Supplementary Fig. S7C and D, Supplementary Data). However, placement of the overall RNA backbone varied substantially. The consistent positioning of the RNA in this location strongly suggests this is likely to be at least a part of the RNA interaction interface. We propose that the RhsF_int–RhsFI heterodimer may have improved RNA-binding activity compared to RhsFI alone.

RhsFint–RhsFI affects the stability of a subset of regulatory sRNAs

The contribution of RhsF_int–RhsFI to gene regulation was assessed by RNA-seq in virulence-inducing media (MEM–HEPES), for EHEC, an isogenic ΔrhsF ΔrhsFI deletion, and the deletion mutant complemented with a plasmid expressing RhsF_int–RhsFI. In the deletion strain 89 genes were differentially regulated (log₂FC ≥ 2, P_adj ≥ 0.05) and 43 of these were restored to levels comparable to wild-type in the complemented strain (Fig. 7A and Supplementary Table S3). Among the top differentially expressed genes were the operons for the glutamate-dependant (GAD) acid stress response, the maltose utilization operon malK-lamB-malM, and several outer membrane transporters, including the pyruvate transporter cstA and multi drug efflux pumps arcB, mdtE, and marA. Notably, each of these operons contain sRNAs that were bound by RhsFI in CRAC (Fig. 7B–D). This suggests that the RhsF_int–RhsFI complex affected regulation of the operon through the associated sRNAs. However, we note that RhsFI also directly bound mRNAs within some of these operons eg: hdeAB within the GAD operons. The effect of RhsF_int–RhsFI on sRNA expression was assessed by northern blot analysis of the RhsFI targeted sRNAs RaiZ, RbsZ, GadY, StxS, and OmrA. The 3′ UTR sRNAs, RaiZ, and RbsZ, were depleted in the deletion background and accumulated in the complemented strain (Fig. 8A). In contrast, GadY, StxS, and OmrA accumulated in the deletion background and were restored to wild-type levels in the complemented strain (Fig. 8B). These data are consistent with upregulation of GAD operons that are positively regulated by GadY. We were not able to detect the malM 3′UTR sRNA by northern, but note that it is a ProQ target, as are RaiZ and RbsZ. Destabilization in the ΔrhsF ΔrhsFI background would be consistent with upregulation of the mal operon in our RNA-seq data.

Figure 7. RhsF_int–RhsFI controls gene expression through direct interactions with sRNAs embedded within operons.

(A) RNA-seq analysis of WT, ΔrhsF ΔrhsFI, and rhsF_int rhsFI complemented strains. Changes in gene expression for WT versus ΔrhsF ΔrhsFI and ΔrhsF ΔrhsFI versus complement are plotted on the x- and y-axis respectively (log₂[fold change]). Statistically significant difference in both experiments coloured blue (P_adj < 0.05) and regions of the plot with recipricol changes in ΔrhsF ΔrhsFI and the complemented strain are indicated in green. (B–D) RhsFI UV-crosslinking (CRAC) showing sites of direct contact for the malK-lamB-malM operon (B), GAD acid tolerance gene cluster (C), and stx1AB genes (D). sRNAs are coloured red in the graphics below each plot. For stx1AB, the region deleted is indicated by the Δ (see [79]).

Figure 8. RhsF_int–RhsFI affects accumulation of sRNA targets in EHEC.

Northern blot analysis of RaiZ (A), RbsZ (B), GadY (C), StxS (D), and OmrA (E) in WT, ΔrhsF ΔrhsFI, and rhsF_int rhsFI complemented strains. Histograms indicate the abundance of each sRNA normalized to 5S rRNA and expressed relative to WT.

These data indicate that RhsF_int–RhsFI has global effects on gene expression in EHEC often through interactions with sRNAs embedded within the operon. RhsF_int–RhsFI can both stabilize or destabilize sRNAs depending on the specific target sRNA, similar to the sRNA binding proteins ProQ and KhpAB [37, 40, 48, 49].

RhsFint–RhsFI contributes to cell adhesion and affects cell morphology

Our data suggest that RhsF-RhsFI functions as a global post-transcriptional regulator and indicated regulation of the GAD acid tolerance operons. In line with this result, the ΔrhsF ΔrhsFI strain had increased GAD resistance tolerance, which was reduced in the complemented background (Fig. 9A). This is consistent with RhsF_int–RhsFI destabilization of GadY sRNA that promotes GAD operon expression.

Figure 9. RhsF_int–RhsFI controls virulence phenotypes.

(A) GAD acid tolerance of WT, ΔrhsF ΔrhsFI, and rhsF_int rhsFI complemented strains. (B) Adherence of WT, ΔrhsF ΔrhsFI, and rhsF_int rhsFI complemented strains to EBL cells after 4 h. Asterisks indicate P < 0.05. (C) Maximum projections of 3D confocal z-stacks of WT, ΔrhsF ΔrhsFI, and rhsF_int rhsFI complemented strains attaching to EBL cells following fluorescent staining for O157, filamentous actin using Phalloidin and nuclei using Hoechst 33342. Fluorescent stains are indicated in the column titles and strains in the row titles. The scale bar (white line) represents 10 μm.

To understand whether RhsF_int–RhsFI affects colonization, we performed cell adhesion assays on embryonic bovine lung (EBL) cells (Fig. 9B and C). Ruminants (predominately cattle) are the natural reservoir of EHEC where it colonizes the terminal rectum of the gastrointestinal tract and is shed in faeces that contaminate foodstuffs, ultimately causing human infection [50]. EBL cells are an established model of T3S-dependant A/E lesion formation for EHEC [27, 51–53]. Deletion of rhsF–rhsFI resulted in 34.3-fold reduction in cell adhesion, and this was restored to wild-type levels by complementation with rhsF_int–rhsFI (Fig. 9B). Visualization of the adherent cells supported a reduction in cell adhesion in the mutant that was restored by complementation (Fig. 9C). The most striking cellular phenotype was cell shortening in the deletion mutant, whereas complementation with rhsF_int–rhsFI led to filamentation. Filamentation was also seen in cells grown in liquid media (data not shown), indicating that the filamentation is not dependant on cell contact.

We conclude that RhsF_int–RhsFI alters gene expression in EHEC affecting phenotypes linked to pathogenesis, including cell adhesion, acid tolerance, and cell morphology.

Discussion

RBPs play critical roles in gene expression and biophysical purification techniques coupled with mass spectrometry have revealed a rich and complex RBPome in bacterial cells [3–6]. Here, we used TRAPP to identify RBPs in the human pathogen EHEC and focussed on RBPs within pathogenicity islands, which encode the majority of virulence factors.

We discovered that a T6SS immunity protein, RhsFI, is expressed independently of the T6SS and binds to many Hfq- and ProQ-dependant regulatory sRNAs. RhsFI is transcribed from a promoter that is within the coding sequence of the up-stream Rhs-domain T6SS effector, RhsF. This sub-genic transcription initiation allows expression of the RhsF toxin sub-domain (here termed RhsF_int). We found that similar internal transcription initiation within RhsG produces a toxin (RhsG_int) that can kill EHEC, with the RhsG_int–RhsGI complex likely functioning as a type II TA-system. This is consistent with work in Salmonella demonstrating that T6SS Rhs-domain effector-immunity loci are transcribed from internal promoters and act as TA systems within the cell [24].

RhsF_int overexpression was not inhibitory, indicating that it has evolved an alternate function. It is likely a nuclease based on structural homology, although we cannot definitively determine if the active site is intact (see Supplementary Data). Alphafold-multimer predicts a strong interaction between RhsF_int and RhsFI consistent with many effector-immunity pairs and type II TA systems. Notably, association of RhsF_int extends the predicted RNA-binding surface of RhsFI, suggesting that the dimer may have improved RNA-binding activity. The RNA-binding surface is distant to the predicted RhsF_int active site, indicating that these may be independent activities of the complex. Loss of nuclease activity and retention of an independent RNA-binding function has been reported in yeast where the RNase II-family ribonuclease Ssd1 has been repurposed as a post-transcriptional regulator [54]. Collectively, our results suggest that the T6SS immunity protein RhsFI binds regulatory RNA in vivo and may use a large positively charged surface on the RhsF_int–RhsFI heterodimer to bind RNA.

UV-crosslinking of RhsFI to the transcriptome indicated that many binding sites were shared with the sRNA chaperones Hfq and ProQ including within 3′ UTR sRNAs. We were able to confirm RhsFI binding to a subset of these targets in vitro but note that the affinity of RhsFI for these targets is an order of magnitude lower than reported for Hfq or ProQ; however, the heterodimer extends the positively charged surface and we predict that it will have a higher affinity. Previous work has shown that ProQ binding can inhibit Hfq-dependant sRNA sponging [37] and many 3′ UTR sRNAs are stabilized by ProQ. Our results suggest that RhsFI (or the RhsF_int–RhsFI complex) binding affects the stability of a subset of sRNAs and we speculate that RhsF_int–RhsFI may function similar to ProQ or KhpB that block RNases from degrading the sRNA [37, 48, 55, 56]. It is possible that RhsFI acts in synergy with ProQ to stablize sRNAs. A common feature of sRNAs that are destabilized by RhsFI is that they are bound more strongly by Hfq than ProQ [37]. We speculate that RhsFI may destabilize sRNAs (like GadY, StxS, and OmrA) by occluding Hfq binding and regulation. Our data suggest that RhsFI has dual functions to enhance the stability of a subset of ProQ-bound sRNAs and to destabilize a subset of Hfq-bound sRNAs.

One target that was unique to RhsFI was the rhsFI mRNA where crosslinking was identified along the length of the transcript. RhsFI binding to rhsFI mRNA is consistent with auto-repression functions observed in other type II TA systems that repress transcription of the TA loci through DNA-binding domains within the antitoxin [57]. In the case of RhsFI, this repression likely occurs post-transcriptionally (rather than transcriptionally) through interactions with the rhsFI mRNA. Antitoxins from the type II TA systems MqsRA, YafQ-DinJ, and HipBA have also been found to repress gene transcription more broadly through direct binding to palindromic sequences within promoters including rpoS, relA, cspE, and cspD [58–61]. Our results suggest that RhsFI or the RhsF_int–RhsFI complex also has broader regulatory functions but acts the post-transcriptional level.

Deletion of rhsF and rhsFI results in differential regulation of 89 genes under virulence inducing conditions, and many of these operons contained sRNAs that were bound by RhsFI suggesting indirect regulation of the operon through the sRNA. Expression of the glutatmate dependent acid resistance operon was reduced, likely through destablization of GadY, and we demonstrate that the rhsF rhsFI mutant has a modest increase in acid tolerance. This result seems counter intuitive as acid tolerance is required during gastrointestinal colonization. However, the transcription factor GadE is a repressor of the T3SS [62] and several EHEC-specific regulators control T3SS indirectly through gadE [27, 63]. We speculate that destabilization of GadY and the GAD operon ultimately promotes virulence through repression of GadE and de-repression of T3SS. Several T3SS genes are repressed in the ΔrhsF ΔrhsFI background but were not restored to wild-type levels in the overexpression strain, possibly due to filamentation.

EHEC utilizes the T3SS to attach to gastrointestinal cells and form actin-rich pedestals. To understand whether the repurposed RhsF^int–RhsFI complex impacted pathogenesis, we performed cell adhesion assays. Deletion of rhsF rhsFI reduced attachment 30.4-fold and overexpression with rhsF_int rhsFI restored attachment to wild type levels. While the specific sRNA interactions involved remain to be defined, these results indicate that RhsF_int–RhsFI increases cell attachment and likely promotes host colonization.

The RBPome has expanded significantly through sequence-dependant and -independent enrichment techniques. Several proteins beyond Hfq and ProQ have now been identified that modulate sRNA stability to control cell physiology [6, 64–69]. sRNAs and newly identified sRNA-binding proteins play a role in virulence gene regulation. Our results indicate that horizontally acquired RBPs can modulate the sRNA regulatory network and may be an effective way to rapidly precipitate broad changes in the transcriptome to promote virulence gene expression.

Materials and methods

Strains, plasmids, and media

Bacterial strains, plasmids, and primers used for this chapter are listed in Supplementary Table S4. Unless otherwise indicated, E. coli was cultured in LB broth. Where indicated E. coli strains where cultured virulence-inducing MEM-HEPES (Sigma M7278) supplemented with 0.1% glucose and 250 nM Fe(NO₃)_3. When necessary, ampicillin (100 μg/ml), or kanamycin (50 μg/ml) were added to the growth media.

Construction of EHEC deletion strains

A gene deletion in rhsF rhsFI was constructed in EHEC str. Sakai stx(-) by allelic exchange. The flanking region of rhsF rhsFI, were amplified and joined together using splice overlap extension (SOE) PCR. The SOE product was then digested and ligated into pTOF25 using SalI and SmaI. A tetRA cassette was introduced into a NotI site within the SOE product. Allelic exchange was performed as described in [70]. Plasmid pCP20 was used to remove the tetracycline cassette leaving a ~ 35 bp scar. The deletion was verified by colony PCR and Sanger sequencing. The complementation vector pZE12::rhsF rhsFI contains the sequence from rhsF internal promoter transcription start to the rhsFI stop codon. The internal rhsF TSS is cloned into the TSS of P_LlacO-1 as described previously [71].

Purification, identification, and analysis of the EHEC RBPome using TRAPP

The TRAPP protocol for purification of the total RBPome of EHEC was done according to [5] with some modifications. A detailed method for preparing the EHEC TRAPP data is presented in Supplementary Methods. Briefly, overnight cultures of E. coli O157:H7 str. Sakai stx(-) were subcultured 1/100 into 800 ml of MEM-HEPES supplemented with 0.1% glucose and 250 nM Fe(NO₃)₃ and grown up to an OD₆₀₀ of 0.8. Cells were UV-crosslinked with 1800 mJ of UV-C in a W5 small diameter UV-crosslinker (UVO3) as described in [72]. Cell pellets were prepared for UV-treated and untreated samples in triplicate. To each pellet, 3 ml of a 1:1 mix between phenol (pH 8.0) and GTC-buffer (4 M guanidinium thiocyanate, 50 mM Tris–HCl pH 8.0, 10 mM EDTA, and 1% β-mercaptoethanol) were added with 1 volume of 0.5 mm zirconia beads, and cells were lysed in a FastPrep-24 5G.

Two milliliters of 1:1 phenol:GTC-buffer was to the lysate and centrifuged clear the lysate. The supernatants were pooled and the volume topped up to 10 ml with GTC-phenol. To each sample, 0.1 volume of sodium acetate (pH 4.0), 1 volume of absolute ethanol, 1 ml of silica beads (50% slurry in MilliQ), and an additional 500 μl of absolute ethanol. Beads were centrifuged and resuspended in 15 ml of wash buffer I (4 M guanidinium thiocyanate, 1 M sodium acetate pH 4.0, and 30% ethanol). Wash buffer was removed, and the wash was repeated twice more. Samples were washed three times with 10 ml of wash buffer II (100 mM NaCl, 50 mM Tris–HCl pH 6.4, and 80% ethanol). Beads were resuspended in 1.5 mL of wash buffer II, pelleted and the supernatant was discarded.

Nucleic acids and crosslinked RNPs were eluted by resuspending the silica beads in 500 μl of 10 mM Tris pH 8.0, and eluate is transferred into a clean 2 ml of microcentrifuge tube. This is repeated twice more for a total of 1.5 ml of eluate per sample. DNA and RNA was digested using Benzonase (Sigma–Aldrich, Cat. no.: E1014). To specifically digest RNA, 0.5 μl of RNase A/T1 was added and samples were incubated in a SpeedVac at 45°C overnight to dry the samples.

Dried samples were resuspended in 75 μl of Laemmli buffer and incubated at 100°C for 10 min. Samples were pooled and loaded onto a gradient 4%–20% Mini-PROTEAN TGX gel (Bio-Rad, Cat. no.: 4561095). The polyacrylamide gel was stained with Imperial protein stain (ThermoFisher, Cat. no.: 24615), according to the manufacturer’s protocol. Proteins were extracted from the gel as per [73].

LC-MS analysis was performed on an Orbitrap Fusion Lumos Tribrid Mass Spectrometer coupled to an Ultimate 3000 RSLCnano Systems. Peptide separation was done on a 50 cm EASY-Spray column at 50°C. Peptides were loaded onto the column at a flow rate of 0.3 μl/min, and eluted at a rate of 0.2 μl/min using the following gradient: 2%–40% mobile phase B (80% acetonitrile and 0.1% formic acid) for 136 min, then to 95% for 11 min. After an additional 5 min at 95% mobile phase B, flow was shifted to 2% until the end of the run. LC-MS was performed by Christos Spanos at the Wellcome Trust Centre for Cell Biology in Edinburgh, Scotland.

Processing of raw MS output was done on the MaxQuant software platform [74, 75]. The average ratio of UV+ to UV ™ LFQ intensity for each peptide was calculated. A Student’s t-test was used to calculate statistical significance, and false-discovery rates were calculated using the Benjamini–Hochberg method.

Polynucleotide kinase assay to confirm RNA-binding properties of candidate RBPs

PNK assays are a simplified version of our previously described CRAC protocol [17, 76]. A more detailed overview of this method can be found in the Supplementary Methods. Briefly, overnight cultures of Sakai stx- containing arabinose-inducible dual-affinity tagged RBP candidates were grown in LB broth or MEM–HEPES media supplemented with 0.1% glucose and 250 nM Fe(NO₃)₃ containing kanamycin and ampicillin. Cultures were grown at 37°C for 30 min at 200 rpm before adding 0.2% arabinose to induce protein expression. Incubation was continued until an OD₆₀₀ of 0.8 was reached. Cultures were UV-crosslinked and harvested.

Cells were lysed and clarified, then loaded onto an equilibrated M2 Anti-FLAG resin. Samples were incubated on M2 anti-FLAG resin for 2 h at 4°C. To elute the proteins from the resin, 20–30 units of TEV protease were added to each sample and incubated with rotation for 2 h at 18°C. The resin was gravity filtered through a Bio-Rad Micro Bio-Spin^® Chromatography column (Cat. no.: 7326204) to collect the eluate. To each sample, 0.15 units of RNace-It™ Ribonuclease Cocktail was added and incubated for exactly 7 min at 20°C. Following termination of the reaction, ribonuclease treated samples were loaded onto equilibrated Pierce Ni-NTA Magnetic Agarose Beads (ThermoFisher, Cat. no.: 78 605). Samples were incubated on the resin overnight at 4°C with rotation.

The 5′ ends of crosslinked RNA were radiolabelled and 5′ phosphorylated using T4 PNK (NEB, Cat. no.: M0201S) and 30 μCi ³²P-γATP (PerkinElmer, Cat. no.: BLU502A250UC). Proteins and their associated RNAs were eluted, precipitated, and run on a Bolt^™ 4%–12% Bis-Tris protein gel (Invitrogen, Cat. no.: NP04120BOX) in NuPAGE™ MOPS SDS running buffer (Invitrogen, Cat. no.: NP0001). Following the run, gel was exposed on a BAS-IP SR Phosphorimaging plate (GE Healthcare, Cat. no.: 28-9564-78).

RhsFI UV-crosslinking and sequencing (CRAC)

UV-crosslinking and sequencing (CRAC) was performed as previously described in [17, 76] with some modifications. This protocol is described in more detail in the Supplementary Methods. Briefly, pellets of UV-crosslinked or uncrosslinked E. coli O157:H7 str. Sakai stx(-) expressing arabinose-inducible RhsFI were prepared as described in the PNK assay. Samples were lysed and loaded onto Ni-NTA resin as previously described. Crosslinked RNA was 3′ dephosphorylated using Theromosensitive Alkaline Phosphatase (Promega M9910). Crosslinked RNA was 5′ radiola-belled and using 40 units of T4 PNK (NEB M0201S) and 3 μl of ³²P-γATP (PerkinElmer BLU502A250UC). This was followed by the consecutive ligation of 5′ and 3′ adapters. RNA–protein complexes were eluted, precipitated, run on a 4%–12% sodium dodecyl-sulfate polyacrylamide gel (SDS-PAGE) and imaged as previously described.

RNA–protein complexes were excised from the gel and protein was digested using Proteinase K. Gel pieces were filtered out by passing through a BioRad Micro Biospin column followed by RNA extraction using phenol:chloroform:isoamyl alcohol. RNA was reverse transcribed using SuperScript RT IV with the PE_reverse RT oligo (5′-CAGACGTGTGCTCTTCCGATCT-3′). Libraries were amplified using Phusion Hot Start Flex 2x Master Mix (NEB M0536S) containing 1.25 μM P5 5′oligo and barcoded 3′ reverse oligos in a 25 μl of reaction. PCR was run with an initial denaturation of 98°C for 30 s, followed by 24 cycles of a 98°C denaturation for 10 s, 54°C annealing for 10 s, and a 72°C extension for 1 min, followed by a 72°C final extension for 10 min. PCR reactions were run on a 2% MetaPhor agarose gel (Lonza Biosciences, 50180). Following this, 175–300 bp fragments were excised and purified using a MinElute Gel Extraction (Qiagen, 28604). Libraries were pooled, quantified on an Agilent Bioanalyzer and submitted for paired-end 50 bp sequencing using the Illumina NovaSeq system with Novogene.

Analysis of RhsFI CRAC dataset

Data was analysed as previously described by Tree et al., 2014; Waters et al., 2017; and Mediati et al., 2022. Briefly, paired-end reads were merged using “bbmerge” and demultiplexed using the pyBarcodeFilter.py script in “pyCRAC” [38, 77]. Reads were trimmed using “bbduk” with settings k = 3, mink =11, hdist =1, hdist2 =1, minlen =20, qtrim = rl, trimq =28, entropy =0.2, entropywindow =20 and entropyk =5 (Bushnell et al., 2017). PCR-duplicated sequences were removed using the pyFastqDuplicateRemover.py script in “pyCRAC” [38]. Identification of protein RNA-binding sites were identified using “pyCRAC”. All the above were performed using the Hyb-CRAC-R Snakemake workflow described in [39].

Purification of RhsFI

The coding sequence of ECs0604 (rhsFI) containing flanking 5′ Ndel and 3′ BamHI sites was synthesized as a gBLOCK and cloned into the pHis17 vector, which contains a N-terminal hexahisitdine tag (His₆), using sticky-end ligation. The insert was confirmed using Sanger sequencing and the pHis17::rhsFI-His₆ construct was transformed into chemically competent E. coli strain BL21-AI (ThermoFisher).

Escherichia coli BL21-AI containing the pHis17::rhsFI-His₆ construct was grown in 2 L of LB media at 37°C with 200 rpm shaking to an OD_600nm of 0.6. Expression of RhsFI-His₆ was induced with the addition of 1 mM isopropyl-β-D-1-thiogalactopyranoside and 0.2% arabinose and incubated for another 3 h at 200 rpm shaking. The cells were harvested by centrifugation at 5000 g for 20 min and resuspended in lysis buffer (50 mM Tris–HCl, 300 mM NaCI, and 1 mM DTT). Cell lysis was performed using the cell disruption system (Constant Systems) at 25 kPSI at 4°C and cellular debris was clarified by centrifugation at 48 000 g for 30 min at 4°C. RhsFI-His₆ was purified by immobilized metal affinity chromatography where the cleared lysate was loaded onto 2 × 5 ml HisTrap column (Cytiva) pre-equilibrated with lysis buffer and bound RhsFI-His₆ was eluted in lysis buffer supplemented with 1 M imidazole. The eluate was then pooled, concentrated, and further purified by size-exclusion chromatography using the HiPrep^™ 16/60 Sephacryl^® S-100 HR column pre-equilibrated with 20 mM Tris (pH 7.8), 150 mM NaCl, 0.1 mM dithiolthreitol [DTT], and 1 mM sodium azide. Peak fractions that yield purified RhsFI-His₆ were pooled and concentrated to 12 mg/ml.

Filter binding assay

The raiZ, rybA, malM 3′UTR, vigR 3′ UTR, and rhsFI RNA species were IVT using the HiScribe T7 RNA synthesis kit (NEB #E2050S) from PCR-generated DNA templates incorporating the T7 promoter (see Supplementary Table S4 for primers). IVT products were RQ1 DNase treated (Promega) for 30 min at 37°C, phenol–chloroform extracted, and ethanol precipitated, and then separated on a 5% polyacrylamide TBE-6M urea gel. Products were excised, crushed, and incubated in 600 μl of RNA gel elution buffer (10 mM magnesium acetate, 0.5 M ammonium acetate, and 1 mM EDTA) for 16 h at 4°C. RNA was extracted from the eluate using phenol-chloroform extraction and ethanol precipitation. Approximately 25 pM of RNA was dephosphorylated using quick calf intestinal alkaline phosphatase (CIP, Thermo), then extracted using phenol–chloroform and ethanol precipitation. The 5′ ends of the RNA were radiolabelled with γ³²P-ATP using T4 PNK (NEB). Radiolabelled products were separated from free nucleotides using a MicroSpin G-50 column (Cytiva). To assess ECs0604 protein binding, increasing amounts of the ECs0604 protein were incubated with 500 fmol of each ³² P-RNA species, with the addition of 250 nM yeast tRNA (Thermo) for 1 h at room temperature in 1× duplex buffer (40 mM Tris-acetate, 0.5 mM magnesium acetate, 100 mM NaCl, and 10% glycerol) in 25 μl of reactions. The reaction mixture was then vacuumed through a 0.2 μm nitrocellulose membrane (Thermo) and a Hybond-N + nylon membrane (Cytiva) that were pre-soaked in cold 1× duplex buffer and sealed in a bio-dot blot microfiltration apparatus (Bio-Rad). Membranes were washed with 500 μl of cold 1× duplex buffer, dried and imaged using a BAS-MP 2040 phosphorscreen on a FLA9500 Typhoon (GE Healthcare). Retention of ³²P-RNA in complex with RhsFI on the nitrocellulose membrane was assessed using K_d values where the K_d was obtained by fitting the data with nonlinear regression using the equation: RhsFI–RNA bound complex = B_max (RhsFI–RNA) × (RhsFI)/(K_d + RhsFI).

RNA-seq analysis of gene expression

Overnight cultures of E. coli O157:H7 str. Sakai stx(-), ΔrhsF/rhsFI and ΔrhsF/rhsFI pZE12::rhsF/FI were sub cultured 1/100 in 25 ml of MEM–HEPES supplemented with 0.1% glucose and 250 nM Fe(NO₃)₃ and grown up to an OD₆₀₀ of 0.8. Complement was induced with 1 mM of IPTG at the time of subculture. Cultures were treated with stop solution (5% phenol to 95% ethanol solution) then pelleted and resuspended 1 ml of GTC buffer (4 M guani-dinium thiocyanate, 2% sarkosyl, 50 mM Tris pH 8.0, 10 mM EDTA, and 1% β-mercaptoethanol) to one part phenol. 0.25 volumes of 0.1 mm zirconia beads (Biospec products, Cat. no.: 1107910z) were added and incubated in a water bath at 65°C alternating samples on a vortex for 1 min for three rounds. Lysates were taken and vortexed with 350 μl of chloroform and 120 μl of 3M NaOAc pH 5.2 and spun down for phase separation with the aqueous layer transferred to 550 μl of 25:24:1 phenol:chloroform:isomyl alco-hol solution followed by another round of phase separation with 450 μl of chloroform. The aqueous phase was transferred into 100% ethanol with GlycoBlue (Thermofisher, Cat. no.: AM9515) and precipitated overnight at ™80°C. The RNA was spun down twice, washing with 70% ethanol before being air dried and resuspended in nuclease free water. The RNA was then treated with RQ1 RNase-Free DNase (Promega, Cat. no.: M6101) alongside the addition of RNasin Ribonuclease inhibitor (Promega, Cat. no.: N2511) for 30 min at 37°C then cleaned and concentrated using the Zymo RNA Clean up and Concentrator kit (Zymo, Cat. no.: R1019). Sample concentration and RINe score was assessed through a nanodrop and TapeStation (Agilent 4150).

Samples were sent to RNA sequencing at the Ramaciotti Centre for Genomics for NextSeq 1000 P1. The READemption v2.0.4 [78] pipeline was used to align reads to the E. coli O157:H7 str. Sakai genome (accession number: NC_002695.2) and calculate differentially expressed genes with the minimal Phred thread score was set at 30.

Northern blot analysis

Total RNA (3 μg) was loaded onto either a 10% or 6% TBE Urea (8 M) polyacrylamide gel Samples were run at 220 volts for 3–4 h then stained with SYBR safe. The gel was transferred onto a Nylon membrane at 30 V overnight and UV crosslinked. Probes were labelled with ³²P-ATP using a PNK reaction then cleaned up using a G-50 column, and 10 pmol of labelled probe was hybridized overnight at 42°C in hybridization buffer (Thermo, Cat. no.: AM8690). Membranes were washed three times with 2× SSPE + 0.1% SDS then exposed using a phosphor screen and imaged on a Typhoon FLA 9500.

Acid stress resistance

IPTG induced, overnight cultures of E. coli O157:H7 str. Sakai stx(-) pZE12::luc, ΔrhsF ΔrhsFI pZE12luc and ΔrhsF ΔrhsFI pZE12::rhsF/FI were sub cultured 1/50 dilution in MEM–HEPES with 1 mM IPTG and respective antibiotics were grown to and normalized to an OD₆₀₀ of 0.8. A 1 ml of aliquot was taken and pelleted and washed twice with phosphate-buffered saline (PBS). Cells were resuspended in 100 μl of M9 Minimal media (M9 salts from Sigma–Aldrich, Cat. no.: M6030, 0.4% Glucose, 2 mM MgSO₄, 0.1 mM CaCl₂), and half of the sample was aliquoted in M9 media containing 1 mM D,L-glutamic acid (Sigma–Aldrich, CAS# 617-65-2, C₅H₉NO₄), 1 mM IPTG, and respective antibiotics. The other half was aliquoted into pH 2.5 M9 media with 1 mM D,L-glutamic acid, 1 mM IPTG, and respective antibiotics; 100 μl was taken starting from t = 0 for every 30 min these aliquots were spun down and washed with PBS thrice, serially diluted then spot-plated onto LB plates overnight.

Cell adhesion assays

Cell adhesion assays and imaging are described in detail in the Supplementary Methods. Briefly, bovine EBL cells were seeded at ~10⁵ cells per well in 24-well plates and incubated overnight at 37°C with 5% CO₂. Bacterial strains, previously grown overnight in LB broth, were sub-cultured at a 1:10 dilution in MEM–HEPES (supplemented with 0.1% glucose and 250 nM FeNO₃) and incubated for ~3 h. The EBL cells were then infected with 100 μl of bacterial culture (MOI ~100). At 3 h post infection, the wells were washed twice with MEM–HEPES media. At 5 h post infection, the wells were washed twice with PBS.

To quantify adhesion by viable counts, infected monolayers were the treated with trypsin and the cells were gently pipetted or scraped with a scraper. A dilution series was performed and 10 μl was streaked onto an LB plate.

For microscopy slide preparation, 24-well plates with cov-erslips coated with collagen. After EBL cell infection, cells were fixed with 100 μl of 1% paraformaldehyde (PFA) and incubated for 10 min at room temperature. The wells were washed with PBS, and 200 μl of PBS was added and stored at 4°C.

For immunostaining, the cells were permeabilized using 200 μl of 0.1% Triton X-100. The cells were blocked by incubating with 200 μl of 1 mg/ml bovine serum albumin (BSA) in PBS. The primary antibody, anti-O157 rabbit antibody(1:50 dilution, Abcam, Cambridge, UK, Catalogue no.: Ab78822), was added at 200 μl per well and incubated for 1 h at room temperature. After incubation, the cells were washed three times with PBS, and the secondary antibody, goat anti-rabbit IgG (Alexa Fluor 568-labelled, 1:1000 dilution, Abcam Catalogue no.: Ab175471), was added at 200 μl per well and incubated for 1 h before washing with PBS.

For actin visualization, 200 μl of Flash Phalloidin Green 488 (BioLegend, Catalogue no.: 424201) was added to each well and incubated for 30 min at room temperature in the dark. The cells were then washed twice with PBS. For DNA visualization, 100 μl of Hoechst 33342 (1:2000 dilution) was added and incubated for 5–10 min at room temperature in the dark. The cells were washed three times with PBS.

Coverslips were inverted onto a drop of Prolong antifade mounting media cured. Imaging was performed at the Katherina Gaus Centre for Microscopy, UNSW, using a Zeiss LSM900 inverted confocal laser scanning microscope at a magnification of 63× under oil immersion. Confocal z-stacks were acquired and then presented as maximum projections using ImageJ software.

Supplementary Material

Supplementary data is available at NAR online.

Data availability

Sequencing data generating in this study are available at NCBI Gene Expression Omnibus (GEO) under the accession GSE291334. The reviewer token for this record is wtwleycelbenpqn. The RhsFI CRAC data can be visualized in the UCSC Genome Browser using the link: https://genome.ucsc.edu/s/jaitree/RhsFI_CRAC_data, or at https://treelabdata.com. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD063410.