The sRNA DicF integrates oxygen sensing to enhance enterohemorrhagic Escherichia coli virulence via distinctive RNA control mechanisms

Significance

Bacteria sense host signals to regulate gene expression and establish infection. Oxygen availability varies within different niches of the gastrointestinal tract, suggesting that oxygen may be an important cue. We demonstrate that the small RNA DicF is a key factor in the ability of enterohemorrhagic Escherichia coli O157:H7 (EHEC) to sense the low oxygen environment of the colon to enhance virulence, through PchA. Mechanistically, DicF disrupts intramolecular interactions that normally inhibit PchA expression. Although commensal E. coli encode one dicF gene, EHEC acquired three additional dicF copies during its evolution, suggesting that oxygen sensing and virulence regulation through DicF provides EHEC with an important strategy to rapidly amplify virulence specifically within its host colonization niche.

Abstract

To establish infection, enteric pathogens integrate environmental cues to navigate the gastrointestinal tract (GIT) and precisely control expression of virulence determinants. During passage through the GIT, pathogens encounter relatively high levels of oxygen in the small intestine before transit to the oxygen-limited environment of the colon. However, how bacterial pathogens sense oxygen availability and coordinate expression of virulence traits is not resolved. Here, we demonstrate that enterohemorrhagic Escherichia coli O157:H7 (EHEC) regulates virulence via the oxygen-responsive small RNA DicF. Under oxygen-limited conditions, DicF enhances global expression of the EHEC type three secretion system, which is a key virulence factor required for host colonization, through the transcriptional activator PchA. Mechanistically, the pchA coding sequence (CDS) base pairs with the 5′ untranslated region of the mRNA to sequester the ribosome binding site (RBS) and inhibit translation. DicF disrupts pchA cis-interactions by binding to the pchA CDS, thereby unmasking the pchA RBS and promoting PchA expression. These findings uncover a feed-forward regulatory pathway that involves distinctive mechanisms of RNA-based regulation and that provides spatiotemporal control of EHEC virulence.

Host- and microbiota-dependent metabolic and chemical reactions shape the environmental landscape of the gastrointestinal tract (GIT), including distribution of microbes (1). Invading bacterial pathogens navigate microenvironments within the GIT to effectively compete with the microbiota for nutrients and coordinate virulence gene expression (2). Molecular oxygen plays a major role in establishment of bacterial communities in the gut (3, 4). Oxygen diffuses from the intestinal tissue into the GIT. In the colon, oxygen is readily consumed by the resident microbiota that reside close to the mucosal interface (3). This generates oxygen gradients in which the lumen is anaerobic and niches more proximal to the epithelial border are microaerobic. In contrast, the small intestine harbors significantly lower numbers of bacteria, and oxygen is not entirely consumed (5). These data support a model in which, during transit through the GIT, pathogens encounter a relatively oxygenated environment within the small intestine before progressing to the oxygen-limited environment of the colon. Therefore, sensing oxygen availability is a key strategy for pathogens to gauge their location within the host and effectively deploy their virulence arsenals (6); however, it is not fully understood how pathogens respond to oxygen levels to regulate virulence.

Enterohemorrhagic Escherichia coli O157:H7 (EHEC) is a food-borne pathogen that colonizes the colon and causes major outbreaks of bloody diarrhea and hemolytic uremic syndrome (HUS) (7). EHEC encodes several important virulence factors, including Shiga toxin that causes HUS (8) and the locus of enterocyte effacement (LEE) pathogenicity island. The LEE-encoded genes are required for attaching and effacing (AE) lesion formation on enterocytes (9). The LEE is comprised of five major operons that encode a type three secretion system (T3SS) and effectors (7, 10). The LEE-encoded ler gene encodes the master regulator of the LEE (11). EHEC uses the T3SS to translocate LEE- and non-LEE encoded effectors to hijack the host machinery, culminating in AE lesion formation, which is required for host colonization and overall pathogenesis (12).

The very low infectious dose of EHEC (as low as 50 colony forming units) is a major factor contributing to outbreaks (7) and suggests that EHEC has evolved mechanisms to efficiently regulate traits important for host colonization. Indeed, ler is a hub of transcriptional regulation that is responsive to numerous signals, such as metabolites and hormones (13, 14). Besides transcription factors, the RNA chaperone Hfq also modulates Ler expression (15), suggesting that RNA-based regulation is central to controlling global LEE expression. Whereas RNA regulatory mechanisms that control expression of specific T3SS apparatus proteins have been described (e.g., ref. 16), in-depth mechanistic insights into how RNA regulation affects global LEE expression and the consequence(s) to T3SS expression are lacking.

Here, we show that under low oxygen conditions, the small RNA (sRNA) DicF is expressed and plays an extensive role in modulating EHEC gene expression, including Shiga toxin and LEE expression. Mechanistically, DicF promotes T3SS expression through the Ler-transcriptional activator PchA. The pchA transcript contains a cis-acting regulatory element in which the coding sequence (CDS) base pairs to the 5′ untranslated region (5′ UTR). This interaction sequesters the Shine-Dalgarno (SD) site and inhibits translation. DicF relieves this interaction by binding to the pchA anti-SD site within the CDS to unmask the pchA SD site and promote PchA expression. These data reveal a feed-forward pathway involving new mechanisms of RNA-based regulation that spatiotemporally controls virulence in response to oxygen availability.

Results

DicF Is an Oxygen-Responsive sRNA That Globally Modulates EHEC Gene Expression.

In nonpathogenic E. coli strains, the Hfq-dependent sRNA DicF influences expression of genes encoding cell division and metabolic processes (17–22). Significantly, environmental cues that promoted DicF expression were not known, and these original studies relied on plasmid-based, heterologous expression of DicF. Recent work demonstrated that DicF is exquisitely stabilized under low oxygen conditions (21) that are reflective of the colon.

Under oxygen-limiting conditions, enolase bound to the degradosome causes changes in cellular localization of RNase E, from the cytoplasmic membrane to the cytoplasm. This redistribution results in decreased stability and activity of RNase E and concomitant stabilization of DicF (21, 23). Under aerobic conditions, this process is reversed (21), and thus DicF-dependent gene regulation is responsive to oxygen availability. We examined DicF expression in WT and ∆hfq EHEC strains grown aerobically or under microaerobic conditions. Consistent with previous findings in nonpathogenic E. coli, DicF expression in EHEC required Hfq and was only detected following growth under microaerobic conditions (Fig. 1A).

Fig. 1.

DicF is expressed under microaerobic conditions and promotes AE lesion formation. (A) Northern blot analysis of DicF in EHEC (WT or ∆hfq) grown under aerobic conditions or in EHEC (WT, ∆hfq, and ∆dicF1-4) grown under microaerobic conditions. 5S rRNA is the loading control. n = 2. (B) Sequence alignment of dicF in E. coli K-12 and the four dicF copies in EHEC 86–24. (C) qPCR of LEE genes in WT and ∆dicF1-4. n = 3. (D) FAS assay showing AE lesions on HeLa cells infected with WT or ∆dicF1-4. AE lesions are indicated by arrows. (E) Quantification of AE lesions on HeLa cells infected with WT or ∆dicF1-4. n = 243–337 HeLa cells. For C and E, error bars show the mean ± SD. *P ≤ 0.01; ***P ≤ 0.0001.

EHEC shares a core set of genes with nonpathogenic E. coli. DicF is conserved in the core E. coli genome. However, during its evolution EHEC acquired >1 Mb of distinct DNA, including three additional copies of dicF that are located within different, EHEC-specific pathogenicity islands (24, 25). One copy (named dicF1) shares 100% identity to nonpathogenic E. coli K-12 dicF, whereas the other alleles (arbitrarily labeled dicF2, dicF3, and dicF4) contain distinct sequence variations (Fig. 1B). Because EHEC has acquired and maintained multiple dicF copies, we hypothesized that DicF may be important for coordinating oxygen-dependent virulence responses. To investigate how EHEC sensing of environmental oxygen through DicF is linked with virulence expression, we generated a quadruple dicF deletion EHEC strain (∆dicF1-4, Fig. 1A). Of note, loss of dicF in EHEC did not impact bacterial growth or replication (SI Appendix, Fig. S1), indicating that the deletion of chromosomal dicF does not lead to nonspecific defects in fitness or replication. Subsequently, we compared the transcriptomes of three biological replicates of WT and the ∆dicF1-4 strains grown under microaerobic conditions in DMEM. More than 300 genes were differentially expressed in the ∆dicF1-4 strain compared with WT (SI Appendix, Fig. S2A) (26). Of these, we measured expression differences of genes carried in the core genome, including genes encoding metabolic enzymes (nar, adhE, tnaA), regulatory factors (hnr, csrB), and fimbriae (ecpR, yehD) (SI Appendix, Fig. S2B). Notably, we also measured differences in EHEC-specific genes, including stx2A that encodes Shiga toxin (SI Appendix, Fig. S2 B and C). Trans-complementation with plasmid-expressed dicF1 restored expression to near WT levels (SI Appendix, Fig. S2C). Moreover, all dicF alleles rescued expression of narL and hnr in the ∆dicF1-4 strain (SI Appendix, Fig. S2 D and E). These data revealed an extensive role for DicF under conditions that recapitulate EHEC virulence gene expression in vivo (27).

DicF Enhances EHEC Virulence.

The LEE pathogenicity island carries 41 genes that are mostly organized into five major operons (SI Appendix, Fig. S3A). LEE1 encodes Ler that activates expression of all of the LEE genes (11). LEE4 encodes EspA which forms the filament of the T3SS apparatus (28). The transcriptomic data revealed at least a twofold decrease in expression of 37 LEE genes in the ∆dicF1-4 strain compared with WT (SI Appendix, Fig. S3B). We further analyzed LEE transcripts by qPCR, confirming that LEE expression required DicF (Fig. 1C and SI Appendix, Fig. S3C). Furthermore, Western blot analysis confirmed that levels of EspA were decreased in the ∆dicF1-4 strain compared with WT EHEC (SI Appendix, Fig. S3 D and E).

Identical or nearly identical sRNAs may have redundant as well as nonredundant targets and cause differential regulation of a specific target (29). To test the contribution of the DicF copies to LEE expression, we measured EspA expression in the ∆dicF1-2, ∆dicF1-3, and ∆dicF1-4 strains. These data indicated that DicF promoted LEE expression in an additive manner, as the double dicF deletion (∆dicF1-2) resulted in less EspA expression compared with WT, which became further decreased in correlation with the number of dicF genes deleted (SI Appendix, Fig. S3 D and E). In agreement with the expression data, the ∆dicF1-4 strain was attenuated for AE lesion formation (Fig. 1 D and E). Together, these data revealed that DicF plays an important role in EHEC virulence.

DicF and PchA Function in a Feed-Forward Pathway to Regulate LEE Expression.

How does DicF promote LEE expression? Considering that nearly all of the LEE genes were decreased in expression in the ∆dicF1-4 strain, we reasoned that DicF directly modulated Ler expression or expression of a Ler-transcriptional regulator. Unbiased, in silico analysis predicted the pch genes as potential DicF targets. The Pch (PerC homolog) family of proteins are horizontally acquired transcriptional activators carried by pathogenic members of the Enterobactericeae (11). In enteropathogenic E. coli or EHEC, PerC or Pch, respectively, promotes transcription of ler, to activate expression of the T3SS (30–34). EHEC encodes three pch genes (pchA, pchB, and pchC) located within distinct pathogenicity islands (33). To examine whether pch is a regulatory target of DicF, we measured pch transcript levels in the WT and ∆dicF1-4 strains grown under microaerobic conditions. These data indicated that Pch expression required DicF, as pch mRNA levels were approximately threefold decreased in the ∆dicF1-4 strain compared with WT EHEC (Fig. 2A).

Fig. 2.

DicF and PchA function in a feed-forward pathway. (A) qPCR of pch in WT and ∆dicF1-4 EHEC. n = 3. (B) qPCR of pch in WT and ∆hfq after growth under aerobic or microaerobic conditions. 16S rRNA was used as the reference control. (C) Western blot of EspA in WT, ∆pchA, and ∆dicF1-4. DnaK is the loading control. (D) EspA quantification in WT, ∆pchA, and ∆dicF1-4 grown microaerobically. n = 5. (E) qPCR of espA in WT, ∆dicF1-3, ∆pchA, and ∆pchA∆dicF1-3. Significance are compared with WT or between the ∆pchA and ∆pchA∆dicF1-3 strains. n = 3. For A, B, D, and E, error bars show the mean ± SD. **P ≤ 0.001; ***P ≤ 0.0001; ns, P > 0.05.

In accordance with DicF modulating oxygen-dependent responses in EHEC, we measured increased levels of pch mRNA in WT EHEC grown under microaerobic conditions compared with aerobic conditions, and this increase required Hfq (Fig. 2B). Moreover, EspA was only detected after growth under microaerobic conditions (Fig. 2 C and D), highlighting the importance of low oxygen availability as a signal for EHEC virulence expression. Although overexpression of any pch gene results in increased levels of LEE expression, PchA is the major contributor to LEE activation (33, 35). Therefore, to test how PchA contributes to oxygen-dependent LEE expression, we generated a pchA deletion EHEC strain (∆pchA). Significantly, EspA expression was decreased in the ∆pchA and ∆dicF1-4 strains compared with WT EHEC (Fig. 2 C and D and SI Appendix, Fig. S4), indicating that DicF and PchA are required for coordinating oxygen sensing and virulence responses.

Next, we investigated whether DicF- and Pch-dependent regulation of the LEE are functionally linked. For this assay, we generated a ∆pchA EHEC strain in which three dicF alleles were deleted (∆pchA ∆dicF1-3). As expected, we measured decreased espA expression in the ∆dicF1-3 and ∆pchA strains; however, no further decreases in espA transcript levels were measured in the ∆pchA ∆dicF1-3 strain compared with the ∆pchA strain (Fig. 2E). These findings demonstrated that DicF and PchA operate in the same pathway to promote LEE expression, with DicF being upstream of PchA.

DicF Base Pairs with the pchA CDS to Promote Expression.

To better understand DicF control of PchA expression, we used the program CopraRNA (36, 37) to identify predicted interaction sites. sRNAs usually bind to the 5′ UTR of the target mRNA over short regions, e.g., 7–12 nucleotides, with imperfect complementarity (38). Notably, DicF was predicted to interact with the pchA CDS through extensive base pairing (over 40 nucleotides) beginning at nucleotide +49 (based on the ATG site) (Fig. 3A). To test this predicted interaction, we performed RNA electrophoretic mobility shift assays (EMSAs) using in vitro transcribed and biotinylated DicF1 RNA. Upon addition of pchA transcript, we measured a shift in the labeled DicF RNA indicating direct base pairing (Fig. 3B). Moreover, mutation of six pchA nucleotides within the predicted DicF binding site (generating pchA^mutA RNA, Fig. 3A) resulted in diminished DicF–pchA RNA interaction (Fig. 3B). To further substantiate DicF base pairing with the pchA CDS, we generated point mutations in the seed region of DicF (creating DicF^mutA) (Fig. 3A) that are expected to decrease interactions with the pchA transcript. Then, we performed competition RNA EMSAs using labeled WT DicF and increasing amounts of unlabeled DicF or DicF^mutA transcript. Unlabeled DicF competed with labeled DicF for binding; however, unlabeled DicF^mutA showed decreased competition (SI Appendix, Fig. S5A). In the reciprocal experiment, unlabeled DicF^mutA effectively competed against labeled DicF^mutA for binding to the pchA^mutA transcript that harbors compensatory mutations, whereas unlabeled DicF did not compete for binding (SI Appendix, Fig. S5B).

Fig. 3.

DicF base pairs with the pchA CDS. (A) Predicted DicF-pchA RNA base pairing. Point mutations to generate the disrupted and compensatory alleles, DicF^mutA and pchA^mutA or DicF^mutB and pchA^mutB are shown. (B) EMSA of DicF and pchA, pchA^mutA, bla (2 µM), and ftsZ (2 µM) transcripts. The graph shows quantification of shifted DicF. (C) Western blot of PchA::FLAG in WT (carrying the pBAD24 vector), ∆dicF1-4 + pBAD24, ∆dicF1-4 + pdicF1, and ∆dicF1-4 + pdicF^mutB. DnaK is the loading control. (D) PchA::FLAG quantification in WT (carrying the pBAD24 vector), ∆dicF1-4 + pBAD24, ∆dicF1-4 + pdicF1, and ∆dicF1-4 + pdicF^mutB. n = 4. (E) Western blot of PchA^mutB::FLAG in WT (carrying the pBAD24 vector), ∆dicF1-4 + pBAD24, ∆dicF1-4 + pdicF1, and ∆dicF1-4 + pdicF^mutB. DnaK is the loading control. (F) PchA^mutB::FLAG quantification in WT (carrying the pBAD24 vector), ∆dicF1-4 + pBAD24, ∆dicF1-4 + pdicF1, and ∆dicF1-4 + pdicF^mutB. n = 4. For D and F, error bars show the mean ± SD. *P ≤ 0.01; **P ≤ 0.001; ns, P > 0.05.

Next, we functionally interrogated the importance of DicF interaction with the pchA mRNA CDS. For these experiments, pchA or pchA^mutB (Fig. 3A), including the native 5′ UTR, was fused to a FLAG tag and cloned into an IPTG-inducible pUCP24 vector to specifically assay posttranscriptional regulation. We examined PchA::FLAG or PchA^mutB::FLAG expression in the ∆dicF1-4 strain after trans-complementation with DicF1 or mutated DicF^mutB (Fig. 3 C–F). DicF1 complemented the ∆dicF1-4 strain by restoring PchA expression, whereas DicF^mutB did not restore expression (Fig. 3 C and D). Significantly, the DicF^mutB that contains compensatory mutations rescued PchA^mutB::FLAG expression in the ∆dicF1-4 strain (Fig. 3 E and F). Collectively, these data indicated that DicF binds directly and specifically to the pchA mRNA CDS to promote PchA expression.

DicF Disrupts an Anti-SD Structure between the pchA mRNA CDS and 5′ UTR to Promote Translation.

To date, only a handful of sRNAs bind deep within the CDS (>5 codons downstream of the start site) (39) of the target transcript to repress expression (40–43). For example, in Salmonella, the sRNA MicC binds the ompD mRNA CDS and recruits RNase E, leading to degradation (43). To provide mechanistic insights into DicF regulation of PchA, we first examined whether DicF functions in the opposite manner to promote target transcript stability. After microaerobic growth of the WT and ∆dicF1-4 strains, cultures were treated with rifampicin to halt further transcription. RNA samples were prepared from cells before and at indicated time points posttreatment. Chromosomal pch or plasmid-encoded pchA transcript abundance and stability was then determined by qPCR or Northern blot analyses, respectively. Both assays revealed that the pch(A) transcript was slightly more stable in the ∆dicF1-4 strain compared with WT EHEC (SI Appendix, Fig. S6 A and B). These data suggested that DicF does not promote PchA expression by enhancing stability.

Stem loop structures within the CDS of an mRNA transcript may influence translation (41). Therefore, we performed in silico analyses to assess whether the pchA transcript harbors stem loop structures that may impact translation. Intriguingly, these queries revealed that the pchA CDS contains an anti-SD sequence that interacts with the 5′ UTR and masks the pchA SD sequence (Fig. 4A and SI Appendix, Fig. S7A). To test this prediction, we probed the structures of pchA or of pchA^mutA RNA that harbors mutations predicted to relieve pchA cis-interactions and expose the SD sequence (Fig. 4C and SI Appendix, Fig. S7B). Comparison of cleavage patterns revealed guanine residues that were exposed in the pchA^mutA ribosome binding site (RBS) but which were protected by secondary structures in the pchA RNA (SI Appendix, Fig. S8 A and B).

Fig. 4.

DicF disrupts an anti-SD structure between the pchA mRNA CDS and 5′ UTR. (A) Predicted base pairing between the pchA mRNA CDS and 5′ UTR. (B) Schematic showing DicF interaction with the pchA mRNA anti-SD site. (C) Schematic showing the mutated nucleotides in the pchA^mutA transcript. (D) Western of PchA::FLAG or PchA^mutA::FLAG in WT and ∆dicF1-4. DnaK is the loading control. (E) PchA::FLAG or PchA^mutA::FLAG quantification in WT and ∆dicF1-4. n = 5. (F) Schematic showing the mutated nucleotides in the pchA^mutC transcript. (G) Western of PchA::FLAG or PchA^mutC::FLAG in WT and ∆dicF1-4. DnaK is the loading control. (H) PchA::FLAG or PchA^mutC::FLAG quantification in WT and ∆dicF1-4. n = 5. For E and H, data were normalized to PchA::FLAG expression in WT. Error bars show the mean ± SD. *P ≤ 0.01; ns, P > 0.05.

Significantly, the anti-SD sequence within the pchA RNA overlaps with the DicF base-pairing site (Fig. 4B). Therefore, we hypothesized that DicF disrupts anti-SD base pairing between the pchA CDS and 5′ UTR to promote translation. If our model is correct, point mutations that disrupt pchA interactions between the anti-SD site and the 5′ UTR would be expected to restore PchA expression to WT levels in the ∆dicF1-4 strain. To test this idea, we transformed WT and the ∆dicF1-4 strains with a plasmid encoding pchA, pchA^mutA, or pchA^mutC alleles. pchA^mutC carries distinct mutations from pchA^mutA that are also predicted to unmask the SD sequence (Fig. 4F and SI Appendix, Fig. S7D). In support of our model, although DicF was required for PchA expression, DicF was not required for robust expression of PchA^mutA or PchA^mutC (Fig. 4 D, E, G, and H). To ensure that the rescue of PchA expression was not due to nonspecific effects of the mutations, we also generated the pchA^mutD allele that is predicted to strengthen pchA cis-base pairing (SI Appendix, Figs. S7E and S9A). These mutations did not rescue PchA expression in the absence of DicF (SI Appendix, Fig. S9 B and C). Consistent with these findings, the pchA^mutB allele (shown in Fig. 3A and SI Appendix, Fig. S7C) does not unmask the RBS, and its expression requires DicF (Fig. 3 E and F). Altogether, these data substantiate our model, as although DicF was required for PchA expression, mutations that disrupted base pairing between the pchA anti-SD and 5′ UTR alleviated the requirement for DicF and resulted in robust PchA expression.

pchA mRNA Cis-Interactions Impact Translation Initiation.

Initiation is the rate-limiting step in translation. Secondary structures in the 5′ UTR are able to inhibit translation completely, whereas RNA duplexes within the CDS do not restrict the ability of the ribosome to efficiently translate mRNA (44). In the previous experiment, PchA^mutA and PchA^mutC expression was similar in WT and ∆dicF1-4 as well as to levels of PchA in WT (Fig. 4 D, E, G, and H). These data indicate that DicF interaction with the pchA CDS does not impair or enhance translation elongation and supports a role for DicF in disrupting intramolecular interactions between the pchA 5′ UTR and CDS that inhibit translation initiation. To investigate how cis-interactions within the pchA transcript impact translation initiation, we measured progression of reverse transcriptase on the pchA or pchA^mutA (in which the anti-SD structure is disrupted) transcript. Addition of ribosomes to the reactions resulted in more rapid inhibition of reverse transcriptase on the pchA^mutA transcript and corresponding decrease in full-length cDNA compared with the pchA transcript (Fig. 5 A and B), indicating that pchA cis-interactions limit efficiency of ribosome binding. To support this idea, we performed in vitro translation assays using pchA, pchA^mutA, or pchA^mutD (in which the anti-SD structure is strengthened) transcripts as templates. These assays demonstrated that disruption of the anti-SD structure in the pchA^mutA allele resulted in more rapid translation and accumulation of PchA^mutA compared with PchA or PchA^mutD (Fig. 5 C and D). These data revealed that pchA cis-interactions between the CDS and 5′ UTR control translation initiation.

Fig. 5.

pchA mRNA cis-interactions impact translation initiation. (A) Reverse transcription inhibition assay of pchA or pchA^mutA after incubation without or with ribosomes. The arrow indicates full-length cDNA. (B) Relative levels of full-length pchA or pchA^mutA cDNA after incubation without or with ribosomes. n = 3. (C) Western blot of in vitro translated PchA::FLAG, PchA^mutA::FLAG, or PchA^mutD::FLAG. In vitro translated DnaK is the reaction control. (D) Quantification of in vitro translated PchA::FLAG, PchA^mutA::FLAG, or PchA^mutD::FLAG. Data are shown relative to PchA::FLAG at 15 min. n = 3. Error bars show the mean ± SD. **P ≤ 0.001.

Discussion

We discovered that the sRNA DicF plays an essential role in integrating oxygen sensing and virulence regulation in EHEC. DicF disrupts intrinsic silencing mechanisms within the pchA transcript to promote PchA expression, which ultimately results in global expression of the LEE and AE lesion formation. These data suggest a model in which DicF-dependent regulation of PchA enables EHEC to precisely time deployment of its T3SS and effectors within the colon, the site of EHEC host colonization (SI Appendix, Fig. S10). Although oxygen is appreciated as an environmental signal that modulates EHEC virulence (45, 46), the underlying mechanisms are not fully understood, and the role of DicF in EHEC physiology and virulence was unknown. In addition to EHEC, other bacterial pathogens sense oxygen to coordinate virulence, including Shigella, enterotoxigenic E. coli, and Salmonella (47–49). In these examples, transcriptional adaptation through the regulatory factors FNR or ArcAB mediates changes in gene expression, including expression of sRNAs that modulate virulence (48). However, the ability to rapidly integrate this signal via RNA-based regulation may be an important and conserved strategy for bacterial pathogens to establish infection, and it is likely that further studies will uncover additional RNA-mediated mechanisms of oxygen sensing and virulence.

This work also provides insights into mechanisms of RNA-mediated regulation. Cis-RNA interactions are well recognized to play essential roles in bacterial physiology and virulence; however, the majority of cis-interactions involve interactions solely within the 5′ UTR. The prototypical example is represented by riboswitches in which binding to a metabolite results in structural changes within the 5′ UTR that inhibit gene expression (50). Bacteria have evolved to minimize binding between the CDS and respective SD sequences to promote efficient translation initiation and thus enhance fitness (51). Nevertheless, although much less common, long-distance cis-interactions between the CDS and SD sequences have been reported to influence expression of genes important for thermostresses or growth rate (52–54). In these examples, factors intrinsic to the mRNA or directly involved in its expression influence stability of the anti-SD structure and thus gene expression. Regulation of pchA expression via cis-interactions reveals that genes important to bacterial virulence are also regulated via anti-SD sequences within the CDS. Moreover, disruption of the anti-SD sequence also requires an external factor, the sRNA DicF.

We also established an important role for DicF in bacterial virulence and demonstrated that DicF controls PchA expression via a distinctive mechanism. In nonpathogenic E. coli, DicF negatively regulates the cell division gene ftsZ as well as the xylose uptake gene xylR and maltose transporter gene malX (17, 18). In EHEC, besides influencing LEE expression, DicF also modulated expression of Shiga toxin, revealing key functions in virulence regulation. Additionally, EHEC carries multiple copies of DicF, suggesting that DicF may be important to amplify bacterial virulence. In line with this idea, the other pch genes in EHEC, pchB and pchC, also promote Ler expression (33, 34). The DicF recognition sequence is conserved in pchA, pchB, and pchC, hence it is likely that DicF promotes expression of all pch transcripts to activate T3SS expression.

To date, only a handful of sRNAs are known to regulate targets by binding deep within the CDS, and these inhibit gene expression (40–43). DicF regulation of PchA expression is therefore unique in that base pairing deep within the pchA CDS promotes translation. Notably, the ribosome is an RNA helicase, and although this activity does not function efficiently during translation initiation, the ribosome is able to disrupt RNA duplexes during elongation (55). Thus, we propose a model in which DicF interaction with the pchA CDS promotes ribosome loading to the 5′ UTR and translation initiation. Subsequently, the ribosome displaces DicF during elongation. It is likely that the mechanism of DicF regulation will have broader implications for understanding sRNA functions in other bacteria. Indeed, although direct evidence is lacking, similar mechanisms of sRNA-dependent gene regulation have been suggested to occur in Pseudomonas aeruginosa and Clostridium acetobutylicum (56, 57). In summary, this work identifies an oxygen responsive feed-forward pathway and provides fundamental insights into RNA-mediated virulence regulation and environmental signaling in bacterial physiology and pathogenesis.

Materials and Methods

Bacterial Strains, Plasmids, and Growth Conditions.

Strains and plasmids are listed in SI Appendix, Tables S1 and S2; primers are listed in SI Appendix, Table S3. Unless indicated otherwise, bacteria were grown statically overnight in LB broth, diluted 1:100 in low-glucose DMEM (Invitrogen), and grown statically for 6 h at 37 °C, 5% CO₂ (microaerobic conditions). For aerobic growth conditions, cultures were grown shaking in DMEM to an OD₆₀₀ of 0.8 (late-logarithmic growth). Oxygen concentrations have been measured at >200 µmol O₂/L or <10 µmol O₂/L under aerobic or microaerobic conditions, respectively (45). Deletion strains were constructed using lambda red mutagenesis (58). Point mutations were generated using the NEB Q5 Site-Directed Mutagenesis Kit. Deletions and mutations were confirmed by Sanger sequencing.