Host cell attachment elicits posttranscriptional regulation in infecting enteropathogenic bacteria

See allHide authors and affiliations

Science  17 Feb 2017:
Vol. 355, Issue 6326, pp. 735-739
DOI: 10.1126/science.aah4886

Touchdown for gut pathogen virulence

Escherichia coli is transformed from a commensal organism into a pathogen by acquisition of genetic elements called pathogenicity islands (PAIs). Katsowich et al. investigated how the PAI virulence genes of enteropathogenic E. coli (EPEC) respond when the bacterium attaches to a host gut cell. EPEC first sticks to the host by means of pili and then uses a PAI-encoded type 3 secretion system (T3SS) to inject multiple effectors into the host cell. But not all virulence mediators are injected. For example, CesT, a bacterial chaperone, delivers virulence effectors into the T3SS apparatus. Then, within the bacterial cytoplasm, it interacts with a gene repressor called CsrA, which reprograms bacterial gene expression to help the bacteria to adapt to epithelial cell–associated life.

Science, this issue p. 735


The mechanisms by which pathogens sense the host and respond by remodeling gene expression are poorly understood. Enteropathogenic Escherichia coli (EPEC), the cause of severe intestinal infection, employs a type III secretion system (T3SS) to inject effector proteins into intestinal epithelial cells. These effectors subvert host cell processes to promote bacterial colonization. We show that the T3SS also functions to sense the host cell and to trigger in response posttranscriptional remodeling of gene expression in the bacteria. We further show that upon effector injection, the effector-bound chaperone (CesT), which remains in the EPEC cytoplasm, antagonizes the posttranscriptional regulator CsrA. The CesT-CsrA interaction provokes reprogramming of expression of virulence and metabolic genes. This regulation is likely required for the pathogen’s adaptation to life on the epithelium surface.

Whereas most Escherichia coli strains are commensal, some clones evolve into pathogens through the acquisition of virulence genes carried on pathogenicity islands, prophages, and plasmids. These pathogens are classified into different virotypes, each of which causes a distinct disease (1). Two of these virotypes, enteropathogenic and enterohemorrhagic E. coli (EPEC and EHEC), as well as the murine-specific pathogen Citrobacter rodentium (CR), share a common chromosomal region of ~35,000 base pairs (bp), termed the locus of enterocyte effacement (LEE), which contains a cluster of operons that encode for a type III secretion system (T3SS) and related proteins (2). These pathogens attach to the apical surface of host intestinal epithelium and employ the T3SS to inject (translocate) into the host cell a battery of up to 40 effector proteins, which are encoded by genes localized to the LEE and other pathogenicity islands (2). The injected effectors subvert host cell processes and induce attaching and effacing (AE) lesions, which are characterized by effacement of the brush border microvilli and formation of actin structures (termed “pedestals”) beneath the attached bacteria (2, 3).

Each of the EPEC effectors is translocated at a typical efficiency (4), but what determines this efficiency is not clear. For most effectors, we found a direct correlation between translocation efficiency and effector concentration in the bacteria and, thus, a similar translocation/expression ratio (Fig. 1A and fig. S1). In contrast, the NleA/EspI effector exhibited a particularly high translocation/expression ratio (Fig. 1A) [NleA is critical for virulence (5, 6) and has been reported to inhibit the host secretory and inflammasome pathways (7, 8)]. Because translocation was performed by attached bacteria and expression was determined in unattached infecting bacteria (fig. S1A), we wondered whether NleA is expressed exclusively by bacteria attached to the host cells. To test this hypothesis, we infected HeLa cells with either an EPEC strain containing chromosomal nleA translationally fused to gfp (green fluorescent protein), which encodes a hybrid NleA-GFP protein (EPECnleA-gfp), or a strain containing chromosomal nleA transcriptionally fused to the gfp gene, which is preceded by a ribosomal binding site (RBS) and forms a bicistronic operon that coexpresses NleA and GFP (EPECnleA-RBSgfp). Because EPEC attach to the host cell primarily as microcolonies, we monitored GFP production by scoring for attached and unattached microcolonies. Notably, whereas EPECnleA-RBSgfp transcribed gfp regardless of attachment to host cells, only attached EPECnleA-gfp showed high GFP levels (Fig. 1, B to D), indicating that NleA production is induced upon attachment to the host cell by a posttranscriptional process. Accordingly, within a given attached microcolony of EPECnleA-gfp, only bacteria directly attached to the host cell produced NleA-GFP (Fig. 1D). Furthermore, the NleA-GFP–producing EPEC were frequently associated with an actin pedestal (Fig. 1D), which points toward intimate attachment (3). Similar results were observed using the closely related EHEC (fig. S2), demonstrating that contact-induced NleA production is common in AE pathogens.

Fig. 1 NleA is produced by EPEC attached to host cells in a specific manner.

(A) HeLa cells were infected with enteropathogenic E. coli (EPEC) strains containing blaM translationally fused to various effector genes, expressed from their native chromosomal promoters. The levels of the effector-BlaM hybrid proteins in the infecting bacteria (abundance) and the efficiency of their translocation into the host cells (translocation) were determined. The translocation/abundance ratios are shown, and the respective effectors are indicated. Assays were carried out at least in quadruplicates. Error bars represent SD. (B to D) HeLa cells were infected with EPECnleA-gfp and EPECnleA-RBSgfp, fixed, and stained with phalloidin rhodamine (actin, red); images were then recorded. (B) Scoring of attached or free EPEC microcolonies, with or without GFP signals (n > 100 microcolonies for each experiment). Error bars indicate SD of three biological repeats. (C and D) Representative images and enlargements of regions defined by black frames. Arrows indicate bacteria attached to the glass slide (green) or to the host cell (yellow). Scale bars, 10 μm.

The above data indicate that nleA mRNA contains a cis element that represses translation and that this repression is weakened upon attachment to the host cell. To characterize this cis element, we generated plasmids containing EPEC nleA and its regulatory region transcriptionally or translationally fused to gfp. These plasmids were termed pnleA-RBSgfp and pnleA-gfp, respectively (Fig. 2A). EPEC and E. coli K12 containing pnleA-RBSgfp produced GFP, but those strains containing pnleA-gfp did not produce NleA-GFP (Fig. 2B and fig. S3, A and B). Furthermore, EPEC containing pnleA-gfp produced NleA-GFP specifically upon attachment to host cells (fig. S3C). These results show that the pnleA-gfp and pnleA-RBSgfp plasmids contain the cis element that is required for contact-induced NleA production. To determine the function of this cis element, we either replaced the nleA promoter (PnleA) with the tet promoter (Ptet), deleted the nleA open reading frame (ORF), or replaced the native 5′ untranslated region (5′ UTR) with a synthetic one (Fig. 2A). Only replacing the 5′ UTR caused derepression of GFP (or NleA-GFP) production (Fig. 2C and fig. S4B), indicating that the nleA native 5′ UTR is sufficient to repress translation of the downstream gene. We noted that this 5′ UTR sequence contains four putative binding sites for the translation inhibitor CsrA (fig. S5), with the most downstream site overlapping the RBS, an arrangement typical of CsrA-repressed genes (911). Notably, the nleA 5′ UTR, including three of the putative CsrA binding sites, is conserved across all of the EPEC, EHEC, and CR strains we examined (fig. S6). Deletion of 50 bp, containing the two central putative CsrA binding sites (Fig. 2A and fig. S5), resulted in strong derepression of NleA-GFP production (Fig. 2D). Reciprocally, in EPEC and E. coli K12 lacking CsrA, the native nleA 5′ UTR no longer mediated repression of translation of the downstream gene (Fig. 2, E and F). Finally, gel mobility shift assays indicate that CsrA specifically binds two sites in nleA 5′ UTR mRNA, of which one overlaps with the RBS (Fig. 2G and figs. S5 and S7). These results indicate that binding of CsrA to the nleA 5′ UTR represses nleA translation.

Fig. 2 CsrA and the nleA 5′ UTR act together to repress NleA expression.

(A) Plasmids containing nleA and its regulatory region fused to gfp are shown. These include transcription (a) or translation (b) fusions, replacement of the nleA promoter with the tet promoter (c and d), deletion of the nleA ORF (e), replacement of the native 5′ UTR with a synthetic one (f), and internal deletion within the 5′ UTR (g). Black and green lines represent the nleA and tet regulatory regions, respectively, and the arrows indicate promoters. Red and blue lines represent the nleA and synthetic 5′ UTR, respectively. Plasmid names (and numbers) are indicated. (B to E) Bacteria cultures grown under infection conditions were extracted, and GFP expression levels were analyzed by Western blots. The tested bacteria included EPEC, E. coli K12 (strain MC1061), and their respective ΔcsrA mutants containing the plasmids described in (A), as indicated above the lanes (a to g). Locations of GFP and NleA-GFP are denoted by arrows and NleA-GFP degradation products [(C) to (E)] by arrowheads. In (C), Tet was added to activate the tet promoter. (E) wt, wild type. (F) NleA expression by the EPEC ΔcsrA mutant with or without a plasmid expressing CsrA via the isopropyl-β-d-thiogalactopyranoside (IPTG)–regulated promoter (pCsrA-SBP). The presence of the plasmid and IPTG concentrations are indicated. (G) Binding of CsrA to the nleA 5′ UTR. Labeled RNA (0.6 nM) were incubated with serial 1:2 dilutions of CsrA (starting with 12 μM) and tested by gel mobility shift assays. The tested RNAs include a negative control (#4 RNA), putative CsrA binding sites in the nleA 5′ UTR (#1 and #3 RNA), and a positive control (PC RNA) (figs. S5 and S7).

Given that NleA is expressed only upon attachment, we predicted that an EPEC surface component senses the host cell and triggers a signaling pathway that ultimately antagonizes NleA repression by CsrA. We tested the involvement of three surface elements that are exclusive to EPEC: the T3SS; the outer membrane adhesion intimin (3); and the bundle-forming pili (BFP), which strongly enhances T3SS activity by promoting EPEC attachment (12). We used EPECnleA-gfp as a parental strain to construct mutants, each lacking one of the above surface elements, and tested them for contact-dependent NleA-GFP production. Only T3SS deficient mutants (i.e., ΔescV, ΔescN, and ΔespB) lost the ability to produce NleA-GFP upon attachment (table S1), demonstrating that translocation of effectors through the T3SS is required for host sensing. Effector translocation by EPEC is facilitated by two T3SS-dedicated chaperons, CesF and CesT, each of which is required for translocation of a different subset of effectors (1316). We found that the EPECnleA-gfp ΔcesF mutant exhibited the wild-type phenotype, whereas the equivalent ΔcesT mutant lost the ability to induce NleA-GFP production upon attachment (Fig. 3A). Next, to test whether activation of the T3SS is sufficient for NleA regulation, even in the absence of a host cell, we used the EPEC mutants ΔsepL and ΔsepD, which exhibit constitutive T3SS activity (17, 18). The mutants displayed a marked increase in NleA levels, in both bacterial and secreted fractions, regardless of host cell attachment (Fig. 3B, left panel). To verify that NleA production by these mutants was stimulated by uncontrolled T3SS activity rather than by direct involvement of SepL or SepD, we examined the NleA levels in double mutants deficient in sepD and lacking T3SS or CesT (i.e., ΔsepD, ΔescV; ΔsepD, ΔescN; or ΔsepD, and ΔcesT). All double mutants lost the capacity to overproduce NleA (Fig. 3B, central and right panels), suggesting that activation of the T3SS, which normally occurs only upon contact with the host cell, is the initial event that triggers nleA production.

Fig. 3 T3SS and CesT are required for contact-induced NleA production.

(A) HeLa cells were infected with EPECnleA-RBSgfp or EPECnleA-gfp containing ΔcesT or ΔcesF. The cells were then fixed, and images were recorded. Arrows indicate EPEC attached to the slide (green) or to host cells (yellow). Scale bar, 20 μm. (B) Western blots were used to test for NleA levels in the bacterial pellets (P) or in the medium [secreted (S)] produced by wild-type EPEC (wt) or mutants (ΔsepL, ΔsepD, ΔnleA, ΔescV, ΔescN, and ΔcesT) grown in Dulbecco’s modified Eagle’s medium (DMEM). Some mutants were supplemented with plasmids expressing the wild-type alleles of sepL and sepD (pSepL and pSepD, respectively). (C) EPEC (wt), EPECnleA-gfp, and a ΔcesT mutant, with or without a plasmid (pCesT), were grown in DMEM. IPTG was added to activate CesT expression. Western blots and antibodies to NleA and CesT (anti-NleA and anti-CesT, respectively) were used to test for levels of NleA, NleA-GFP, and CesT in the bacteria. (D) E. coli MC1061 (wt) or a ΔcsrA mutant containing pnleA-gfp or both pnleA-gfp and pCesT plasmids were grown in DMEM, with or without the addition of IPTG (to induce CesT production). Western blots and antibodies to GFP were used to test for the levels of NleA-GFP in the bacteria. (E) Lysates containing CsrA-SBP, GST, or GST-CesT were mixed as indicated and pulled down by glutathione agarose beads. The inputs and captured proteins were detected by Western blots using anti-GST or anti-SBP. IP, immunoprecipitation. (F) CesT antagonizes binding of CsrA to the nleA 5′ UTR. Labeled #3 RNA [0.6 nM (fig. S7)] was incubated with or without CsrA (5 μM) and with or without increasing concentrations of GST-CesT (5 to 20 μM) or GST (negative control, 25 μM) and subjected to gel mobility shift assays.

We next asked whether the T3SS, together with CesT, removes from the EPEC cytoplasm a protein that is required for the CsrA-dependent repression of nleA. However, extensive screening failed to identify such a protein (table S1). We therefore tested the alternative possibility that CesT, which upon translocation of effectors is liberated and remains in the EPEC cytoplasm, antagonizes CsrA. Accordingly, an increased CesT level was sufficient to stimulate NleA production posttranscriptionally, both in wild-type EPEC and E. coli K12/pnleA-gfp (Fig. 3C and fig. S8). We excluded the possibility that CesT is required merely for NleA stability (fig. S9, A and B) and show that in the absence of CsrA, CesT is no longer needed for the production of NleA-GFP (Fig. 3D). These results indicate that CesT antagonizes CsrA. To test whether CsrA and CesT interact, we used CsrA and CesT, fused to streptavidin binding peptide (SBP) and glutathione S-transferase (GST), respectively, in a pull-down assay, after validating that they are biologically active (Fig. 2F and fig. S9C). The results show that CsrA-SBP was specifically copurified with GST-CesT (Fig. 3E), demonstrating interaction between CesT and CsrA. Furthermore, gel-shift competition assays show that CesT specifically antagonizes binding of CsrA to the nleA 5′ UTR (Fig. 3F). These results show that upon contact with the host cell, first the T3SS is activated and then the CesT-bound effectors are translocated into the host, liberating CesT, which binds and antagonizes CsrA.

Whereas our data show that CsrA represses NleA expression, Bhatt et al. (19) reported that CsrA positively regulates expression of one of the LEE operons, termed LEE4, which encodes essential T3SS components (20, 21). Furthermore, during our study, we noticed that CsrA is required for BFP expression (Fig. 4, A and B). Thus, we predicted that by antagonizing CsrA, CesT should repress expression of proteins related to BFP (BfpA) and LEE4 (EspB). As predicted, elevated CesT concentrations negatively regulated BfpA and EspB production (Fig. 4C). These findings imply that the CsrA and T3SS activities regulate each other through an indirect negative-feedback loop (Fig. 4D). Notably, CesT binds to ~10 effectors including NleA (22). This NleA-CesT interaction might function as a secondary negative loop.

Fig. 4 The T3SS forms a negative-feedback loop mediated by the CsrA-CesT switch.

EPEC (wt) or the ΔcsrA mutant was used for infection of HeLa cells (A) or for protein extraction followed by Western blot analysis with anti-BfpA (B). Arrows in (A) indicate attached bacteria and induced actin pedestals. Top panels, phase-contrast images; bottom panels, phalloidin rhodamine staining (actin, red). Enlargements of the boxed areas are shown. Scale bar, 20 μm. (C) EPEC (wt) or the ΔcsrA mutant, with or without plasmids expressing CesT (pGST-CesT) or CsrA (pCsrA-SBP), were grown in DMEM in the presence or absence of IPTG to induce CsrA and CesT expression. Production of NleA, BfpA, EspB, and CesT was analyzed by Western blot with appropriate antibodies. (D) The T3SS negative-feedback loop. The T3SS eliminates effectors by secretion, liberating CesT to antagonize CsrA. This results in repression of CsrA-dependent T3SS proteins, followed by the attenuation of T3SS activity and reaccumulation of effectors, which resequesters CesT.

CsrA is known to mediate global gene regulation in E. coli K12 (11, 2325). A similar regulation is expected in EPEC, as its genome, including csrA, is mostly identical to that of E. coil K12 (26). As in E. coli K12, the EPEC csrA mutant exhibits marked glycogen accumulation (19), indicative of a fundamental change in carbon flow in the bacteria. To assess the extent of CsrA regulation in EPEC, we employed an algorithm that predicts which genes will be negatively regulated by CsrA (27). We identified 196 EPEC genes that are potentially repressed by CsrA, of which 150 are also found in E. coli K12 strains and 46 are exclusive to EPEC, located in various pathogenicity islands (fig. S10 and table S7). On the basis of this bioinformatics study, we selected three genes (glgC, ydeH, and pgaA), which are repressed by CsrA in K12, for further analysis (11, 25). Of these, glgC and ydeH are predicted to also be repressed by CsrA in EPEC, and pgaA is a K12-exclusive gene (table S7). We found that expression of CesT in E. coli K12, or EPEC, is sufficient to activate the expression of these genes (figs. S11 to S13). We further determined that, in EPEC, expression of glgC and ydeH is specifically activated upon contact with host cells (fig. S14). Taken together, our analyses show that, in attached EPEC, the CsrA-CesT-T3SS regulatory circuit influences the expression patterns of nleA, glgC, ydeH, BFP, LEE4, and probably other genes as well.

Here we show that the T3SS serves to sense attachment to the host cell and, as a result activates, the CsrA-CesT-T3SS regulatory circuit. Integration of our results with the broad knowledge of CsrA regulation in E. coli K12 points to possible massive remodeling of gene expression upon EPEC attachment (fig. S15). In this study, we used EPEC to analyze the link between the T3SS and gene expression and provide data supporting the premise that the same regulation occurs in EHEC and CR. We suggest that this remodeling of gene expression allows AE pathogens to colonize the intestinal epithelium surface and successfully survive and flourish in this niche, which is essentially free of microbiota (28). We predict that a similar link between T3SS activity and expression of metabolic and signaling genes exists in other pathogens.

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Tables S1 to S7

References (2950)

References and Notes

  1. Acknowledgments: We thank J. Kaper, S. Gruenheid, M. Donnenberg, and G. Frankel for providing antibodies and strains; R. Kulkarni for providing and running the CsrA predicting script; A. Winer, M. Elgradly-Weiss, T. Hershko-Shalev, and L. Argaman for technical help; and S. Ben Yehuda, G. Segal, and J. Sivaraman for reading the manuscript. The work was funded by a grant from the Israel Academy of Sciences and Humanities. N.K. is a recipient of a fellowship from the Carol and Leonard Berall Endowment. I.R. is an Etta Rosensohn Professor of Bacteriology. Plasmids and bacterial strains described in this paper are available from I.R. under a material transfer agreement with the Hebrew University of Jerusalem.
View Abstract

Navigate This Article