The E. coli transcription factor GrlA is regulated by subcellular compartmentalization and activated in response to mechanical stimuli

Significance

EHEC bacteria have been implicated in numerous outbreaks of serious foodborne illnesses. EHEC activates virulence factors that promote gut colonization upon host ingestion by relying upon the sensation of mechanical stimuli created when bacteria adhere to intestinal tissue and when fluid flows past the bacterial membrane following adhesion. We found that the EHEC regulator GrlA activates virulence genes in response to mechanosensation. In planktonic bacteria, GrlA is membrane bound and physically separated from its DNA targets. When EHEC is mechanically stimulated, GrlA is released from the membrane into the cytoplasm, where it can bind and activate virulence gene promoters. This regulatory mechanism allows EHEC to rapidly adapt to the highly dynamic mechanics of the host intestine and successfully colonize.

Abstract

Enterohemorrhagic Escherichia coli (EHEC) is a foodborne pathogen that colonizes the gastrointestinal tract and has evolved intricate mechanisms to sense and respond to the host environment. Upon the sensation of chemical and physical cues specific to the host’s intestinal environment, locus of enterocyte effacement (LEE)-encoded virulence genes are activated and promote intestinal colonization. The LEE transcriptional activator GrlA mediates EHEC’s response to mechanical cues characteristic of the intestinal niche, including adhesive force that results from bacterial adherence to epithelial cells and fluid shear that results from intestinal motility and transit. GrlA expression and release from its inhibitor GrlR was not sufficient to induce virulence gene transcription; mechanical stimuli were required for GrlA activation. The exact mechanism of GrlA activation, however, remained unknown. We isolated GrlA mutants that activate LEE transcription, independent of applied mechanical stimuli. In nonstimulated EHEC, wild-type GrlA associates with cardiolipin membrane domains via a patch of basic C-terminal residues, and this membrane sequestration is disrupted in EHEC that expresses constitutively active GrlA mutants. GrlA transitions from an inactive, membrane-associated state and relocalizes to the cytoplasm in response to mechanical stimuli, allowing GrlA to bind and activate the LEE1 promoter. GrlA expression and its relocalization in response to mechanical stimuli are required for optimal virulence regulation and colonization of the host intestinal tract during infection. These data suggest a posttranslational regulatory mechanism of the mechanosensor GrlA, whereby virulence gene expression can be rapidly fine-tuned in response to the highly dynamic spatiotemporal mechanical profile of the gastrointestinal tract.

The ability to sense and adapt to specific environmental niches is critical to any organism’s survival and, ultimately, to the biological success of the species. For decades, the bacterial capacity to perceive and respond to chemical cues has been subject to intense study. Their ability to sense mechanical stimuli, however, is a relatively recent discovery and remains understudied. While chemical sensing is driven by ligand–ligand receptor interactions, in the case of physical cues the exact mechanisms of signal perception and transduction are much more obscure.

Our knowledge about bacterial mechanosensing and responses to mechanical stimuli is limited. One example of such a response is seen in instances of severe osmotic shocks, to which bacteria respond by opening gated, mechanosensitive channels in the cytoplasmic membrane to allow uptake or efflux of cytoplasmic solutes (1, 2). These channels respond to changes in lipid tension within the cytoplasmic membrane and are conserved across all kingdoms, including higher eukaryotes (3). In contrast, very little is known about how bacteria perceive more subtle mechanical stimuli, such as those encountered within the host environment. Several recent studies, including one by our research group, have highlighted dynamic changes in the mechanical environment that bacteria encounter upon their transition between environmental and host-associated lifestyles as an essential cue for expression of virulence factors (4–6). However, detailed understanding of the mechanism that links mechanosensing to changes in gene regulation and consequently, adaptation of bacterial physiology to the host environment, is still lacking.

We recently determined that the human diarrheal pathogen enterohemorrhagic Escherichia coli O157:H7 (EHEC) responds during adherence to host cells and exposure to shear forces physiologically similar to those prevalent in the intestinal environment by up-regulating genes in its locus of enterocyte effacement (LEE). The LEE encodes a type III secretion system (T3SS) that translocates effector proteins into the host cell cytoplasm, where they modulate host cellular signaling to facilitate bacterial colonization of host cells, immune modulation, and bacterial persistence (7). Most notably, T3SS proteins promote cytoskeletal rearrangement, pedestal formation, and stable anchoring of the bacterium to the host cell. Their effects, however, are more wide ranging, and the complete repertoire of effectors and their biochemical activities are subject to ongoing studies. The LEE pathogenicity island encompasses over 40 open reading frames (ORFs) that are organized into five major transcriptional units. This genetic region was acquired through horizontal transfer, a process by which bacteria acquire DNA from different strains or species. Outside of the host, expression of the pathogenicity island is muted by global H-NS-mediated silencing because its costly to produce gene products are not necessary for survival (8).

Upon host ingestion, changes in temperature and recognition of specific chemical cues, in addition to mechanical stimuli specific to the host intestine, alert EHEC to a transition in environment. In response, the pathogen adjusts its gene expression profile as it passes through the gastrointestinal (GI) tract, where it initiates expression of the LEE and colonization in a highly site-specific manner. The T3SS facilitates colonization of both reservoir animals and human hosts and affects disease severity (9, 10). Although we showed that the LEE-encoded master regulator Ler and the global regulator of Ler (GrlA) are both required for activation of the LEE in response to mechanical stimuli (4), the precise mechanism of activation is not understood. Thus, we set out to study the molecular mechanism of LEE activation in response to mechanical cues upstream of Ler.

Results and Discussion

Mutations in the C Terminus of the Mechanosensitive Transcriptional Activator GrlA Induce the ler Promoter in the Absence of Mechanical Stimuli.

The LEE-encoded transcriptional activator GrlA is required for induction of ler by mechanical stimuli, such as attachment of EHEC to epithelial cells or to a rigid surface and exposure to flow (4). In contrast, growth of EHEC on a soft surface, such as agar plates, does not provide sufficient mechanical stimulation to activate ler. We exploited this property to identify mutations in GrlA that would increase ler activity using EHEC expressing a ler:gfp reporter. EHEC O157:H7 EDL933 containing the wild-type (WT) allele of grlA did not induce ler:gfp. In contrast, strains with either grlA:ML1 (K111A R114A) or grlA:ML2 (R132A K135A) allelic replacements on the chromosome induced robust expression of ler:gfp upon growth on agar plates (Fig. 1B). GFP fluorescence intensity in grlA:ML1 and grlA:ML2 backgrounds was similar to that of a wild-type strain carrying a ler99T:gfp reporter (Fig. 1 B, Top), where ler transcription is genetically uncoupled from GrlA induction (11). Strains containing promoterless reporter controls (U9:gfp) showed no GFP fluorescence under experimental conditions (Fig. 1B).

Fig. 1.

Mutation of key residues in the GrlA C terminus induce the ler promoter in the absence of mechanical stimuli. (A) Schematic of GrlA, with DNA/GrlR-binding helix-turn-helix (HTH) motif and res. R53/R54 and R65/K66 mutated in the GrlR-binding deficient mutants MR1 and MR2, highlighted in purple; and res. K111/R114 and R132/K135 mutated in the ler activating mutants ML1 and ML2, highlighted in orange. (B) EHEC with GrlA WT, ML1, or ML2 alleles on the chromosome and carrying plasmid reporters ler:gfp, promoterless U9:gfp, or the GrlA-independent, constitutively active ler99T:gfp. Plates were imaged, using top lighting (Left), to visualize bacterial growth and green fluorescence (Right) to visualize transcriptional activity. (C) EDL933 cotransformed with plasmids that contain GrlA WT, ML1, or ML2 and plasmid reporters ler:lacZ, U9:lacZ, or ler99T:lacZ. β-Galactosidase activity was measured as a readout for transcription. (D) MG1655 cotransformed with plasmids as in C; β-galactosidase activity was measured as in C. For C and D, values are means ± SEM (n = 3). p < 0.0001 (****), p < 0.05 (*), P ≥ 0.05 not significant (ns) compared to GrlA WT.

Next, wild-type GrlA, GrlA ML1, or GrlA ML2 were overexpressed in either EDL933 + ler:lacZ (Fig. 1C) or in the nonpathogenic E. coli strain MG1655 + ler:lacZ (Fig. 1D), and ler transcription was measured as β-galactosidase activity. Overexpression of wild-type GrlA promoted basal levels of ler transcription in both strains in the absence of mechanical stimuli, in agreement with previous studies (4, 11, 12). In contrast, overexpression of GrlA ML1 or GrlA ML2 mutants in both EHEC and non-EHEC backgrounds induced significantly higher levels of ler transcription than did wild-type GrlA in the absence of mechanical cues (Fig. 1 C and D). Transcription of ler:lacZ in the presence of either GrlA mutant was still lower than that of the constitutively active control (ler99T:lacZ), which is GrlA independent, and remained unchanged in the presence of either wild-type or mutant GrlA (Fig. 1 C and D). Promoterless controls (U9:lacZ) showed low activity throughout and were unaffected by GrlA overexpression (Fig. 1 C and D). These data suggest that point mutations of basic residues in the C terminus of GrlA allow the protein to induce the ler promoter in the absence of mechanical stimuli. These GrlA mutants are dominant over wild-type GrlA and are able to activate ler transcription independent of strain background. Of note, the region mapping to mutations ML1 and ML2 (spanning GrlA C-terminal residues [res.] K111–K135) is distinct from the helix-turn-helix motif responsible for ler promoter binding (GrlA res. 36 to 66, Fig. 1A).

The Transcriptional Activator GrlA Binds Acidic Membrane Phospholipids, and Point Mutations in the GrlA C Terminus Disrupt Lipid Binding.

The transcriptional activator GrlA can form a complex with GrlR, which has been shown to bind lipids in vitro (13). This led us to ask whether GrlA can directly bind lipids independently of GrlR. Purified, recombinant GST-GrlA was incubated with nitrocellulose strips spotted with serial dilutions of EDL933 membrane lipid extracts, and bound protein was detected using Western blotting. Wild-type GST-GrlA, but not GST-GrlA ML1, or GST-GrlA ML2, or GST alone, bound membrane lipid extracts in a concentration-dependent manner (Fig. 2A). To further characterize the binding specificity of GrlA, we performed lipid overlay assays with purified GST-GrlA proteins and membrane lipid strips, which were prespotted with 15 different lipid species (Fig. 2B). Wild-type GrlA showed a distinct binding specificity for acidic phospholipids (phosphatidylinositol 4-phosphate, phosphatidylinositol 4,5-bisphosphate, phosphatidylinositol [3,5]-trisphosphate, cardiolipin, phosphatidic acid, and phosphatidyl serine). In contrast, GrlA ML1 and GrlA ML2 mutants and GST alone did not show strong binding to membrane lipid strips (Fig. 2B). Of the lipid species bound by GrlA in the lipid overlay assays, cardiolipin is the most relevant, as it is the dominant acidic lipid in the E. coli membrane. Cardiolipin constitutes 5 to 10% of the E. coli phospholipidome depending on growth phase and media (14, 15) while phosphatidic acid is, at 0.2%, much less abundant (16). Phosphatidylinositols do not naturally occur in the E. coli membrane. As a quantitative assay of the interaction between GrlA and cardiolipin, we performed liposome cosedimentation using recombinant GST-GrlA proteins and increasing amounts of liposomes that contain phosphatidylcholine and cardiolipin (Fig. 2C). Wild-type GrlA bound cardiolipin-containing liposomes, but not liposomes that contain only phosphatidylcholine, in a lipid concentration-dependent manner. Neither GrlA ML1, GrlA ML2, nor GST alone bound liposomes that contain phosphatidylcholine and cardiolipin. These data led us to conclude that GrlA can associate with membrane lipids independent of GrlR by binding specifically to acidic lipid species, such as cardiolipin, through basic residues in its C-terminal portion (K111 and/or R114 as well as R132 and/or K135).

Fig. 2.

GrlA binds acidic membrane lipids, and C-terminal point mutations disrupt lipid binding. (A) Nitrocellulose membranes spotted with serial dilutions of EHEC membrane lipid extracts (1,000 to 0.32 µg/mL lipid) were incubated with 1 µM recombinant, purified GST-GrlA WT, ML1, ML2, or GST alone and bound proteins were detected using α-GST and α-mouse-horseradish peroxidase (HRP) and enhanced chemiluminescence (ECL) detection. n = 3. (B) Membrane lipid strips prespotted with 15 different lipid species (Left) were incubated with 5 µM GST-GrlA WT, ML1, ML2, or GST and bound proteins detected as in A. n = 3. (C) Purified GST-GrlA WT, ML1, ML2, or GST alone were incubated with 0 to 1 mg/mL of liposomes that contain phosphatidylcholine (PC) alone or a 4:1 molar ratio of PC: cardiolipin (CL). Loaded protein (load) or protein cosedimented with 0 to 1 mg/mL liposomes (lanes 2 to 7) were resolved by SDS/PAGE and Coomassie staining. Liposome-bound protein was quantified by densitometry and graphed as fraction bound relative to load as a function of liposome concentration (Right). Values are means ± SEM (n = 3).

Differences in Lipid Binding between Wild Type and GrlA Mutants In Vitro Are Recapitulated by Their Cellular Localization Pattern In Vivo.

We constructed GFP-GrlA fusions to test their subcellular localization in EHEC in vivo. Fusion of GFP to the N terminus of GrlA did not disrupt GrlA function in the context of ler induction: Expression of GFP-GrlA WT, ML1 or ML2 mutants, but not GFP alone, activated transcription of ler:lacZ in EHEC (see Fig. 5D), with β-galactosidase activities comparable to those achieved with tagless GrlA constructs (Fig. 1C). This is in line with Padavannil et al. (12), who showed that fusion of a bulky tag to the N terminus of GrlA does not inhibit ler promoter binding. Staining of planktonic EHEC that expresses GFP-GrlA with the cardiolipin-specific fluorescent dye 10-N-nonyl-acridine orange (NAO) (17) showed enrichment of GrlA in patches at the cell membrane and a high degree of colocalization between wild-type GFP-GrlA and NAO, which also formed a patchy pattern outlining the cell membrane (SI Appendix, Fig. S1A). In contrast to GFP-GrlA WT, GFP-GrlA ML1 and GFP-GrlA ML2 were localized throughout the cytoplasm. Their fluorescence intensity distributions through a lateral bacterial cross-section showed no strong correlation with the fluorescence distribution pattern of NAO (SI Appendix, Fig. S1 B and C), and their correlation with NAO was significantly different from that of GFP-GrlA WT (SI Appendix, Fig. S1E). These data suggest that in planktonic EHEC, wild-type GrlA is primarily localized at cardiolipin-rich regions of the membrane, reflecting the protein’s capacity to bind cardiolipin in vitro, while GrlA ML1 and ML2 mutants are localized to the cytoplasm, in line with their inability to bind membrane lipids in vitro (Fig. 2). These data support a correlation between the subcellular localization of GrlA variants and their ability to induce the ler promoter in the absence of force: Wild-type GrlA is preferentially bound by the membrane and is force sensitive, whereas GrlA mutants that are mostly localized to the cytoplasm activate the ler promoter in a force-independent manner.

Fig. 5.

Inhibition of GrlA by GrlR and by membrane sequestration are genetically separable. (A) Membranes spotted with EHEC membrane lipid extracts (1,000 to 0.32 µg/mL lipid and solvent blank) incubated with 1 µM GST-GrlA MR1, GST-GrlA MR2, or GST alone; bound protein detected using α-GST and α-mouse-HRP and ECL detection. n = 3. (B) Membrane lipid strips prespotted with 15 different lipid species (Left) were incubated with 5 µM GST-GrlA MR1 or GST-GrlA MR2, and bound proteins detected as in A. n = 3. (C) A totl of 5 µM GST-tagged GrlA WT, ML1, ML2, MR1, or MR2, or GST alone were incubated with 5 µM His-GrlR and proteins pulled down with GSH agarose beads visualized by Western blotting with α-GST (Top) and α-His (Bottom). n = 3. (D) EHEC were cotransformed with plasmids that express GFP-GrlA WT, ML1, ML2, MR1, MR2, or GFP alone, and ler:lacZ and β-galactosidase activity was measured. Means ± SEM (n = 3) are shown. p < 0.0001 (****), p < 0.005 (**), p < 0.05 (*). Representative maximum intensity projections of z-stack images of EHEC expressing GFP-GrlA MR1 (E–G) or GFP-GrlA MR2 (H–J) grown planktonically (E and H), attached to a microfluidic channel under static conditions (F and I), or exposed to fluid shear force (G and J). (Scale bars, 1 µm.) n = 3. (K) Percentage membrane localized GFP-GrlA MR1 and MR2 in planktonic (black), adherent static (light blue), and adherent + shear exposed (dark blue) EHEC. Quantification based on ≥100 cells/condition (N = 3). Means ± SEM are shown. Significance compared to planktonic cells; p < 0.0001 (****), p < 0.05 (*).

Wild-Type GrlA Is Sequestered at the Inner Membrane in Planktonic Cells and Relocalizes to the Cytoplasm in Response to Mechanical Stimuli.

The correlation between subcellular localization of GrlA derivatives and their ability to activate ler led us to hypothesize that wild-type GrlA activation by mechanical stimuli may involve a change in its subcellular localization: from its membrane-sequestered form, where it is unable to bind and activate the ler promoter, to a cytoplasmic form, where it can bind and activate the ler promoter. To address this hypothesis, we compared GFP-GrlA localization in planktonic EHEC (not exposed to mechanical stimuli), EHEC attached to polylysine-coated microfluidic channels under static conditions (exposed to adhesive force), and EHEC attached to channels under flow (exposed to both adhesive force and fluid shear) (Fig. 3M and SI Appendix, Fig. S2 for experimental setup). For the latter, nonpulsatile, laminar flow with a shear force of 10 dynes/cm² was applied, approximating the upper physiological limit of fluid shear force reached in the human intestinal tract during transit (18). In planktonic EHEC, wild-type GFP-GrlA protein showed a punctate pattern that outlined the cells (Fig. 3A), and wild-type GFP-GrlA was membrane sequestered in 85% of cells (SI Appendix, Fig. S2A). EHEC attachment to a microfluidic channel under static conditions led to a visible redistribution of wild-type GFP-GrlA (Fig. 3E), with 52% of cells containing primarily membrane-associated GrlA (SI Appendix, Fig. S2B). Exposure of EHEC to both flow and attachment led to a larger fraction of GFP-GrlA transitioning from the membrane to the cytoplasm (Fig. 3I), with 78% of cells containing primarily cytoplasmic GFP-GrlA (SI Appendix, Fig. S2C). Although the pattern of GFP-GrlA localization was easily discernible in planktonic and flow-exposed cells, where a large fraction was either localized to the membrane or cytoplasm, we performed analyses of the GFP intensity distribution through the cell cross section to account for bacteria containing mixed populations of membrane and cytoplasmic GrlA (SI Appendix, Fig. S3). In planktonic EHEC, wild-type GFP-GrlA showed a roughly bimodal distribution, with peaks at the cell edges (SI Appendix, Fig. S3A). Upon attachment to microfluidic channels, the GrlA intensity distribution gradually shifted away from the edges to spread across the cell volume and reached a nearly symmetrical unimodal distribution in cells exposed to both adhesive force and flow (SI Appendix, Fig. S3C). Distribution of wild-type GrlA in planktonic EHEC was significantly different from its distribution in cells exposed to adhesive force, or adhesive force and flow (SI Appendix, Fig. S3D). In contrast to GFP-GrlA WT, GFP-GrlA ML1/2 were distributed throughout the cytoplasm in planktonic cells (Fig. 3 B and C); only 12% and 4% of GFP-GrlA ML1 and GFP-GrlA ML2, respectively, localized to the EHEC membrane (Fig. 3N and SI Appendix, Fig. S2A). Following attachment to microfluidic channels either under static conditions or in the presence of flow, both GFP-GrlA ML1 (Fig. 3 F and J) and GFP-GrlA ML2 (Fig. 3 G and K) remained localized throughout the cytoplasm, and their distribution did not change compared to planktonic EHEC (Fig. 3N). As controls, we used the stain FM4-64, which evenly stained the membrane and remained membrane localized in all conditions (Fig. 3 D, H, and L and SI Appendix, Fig. S2). The control condition, GFP alone, exclusively showed cytoplasmic localization (Fig. 3O and SI Appendix, Fig. S2). Together, these data support the hypothesis that the activity of wild-type GrlA is regulated by subcellular compartmentalization. Exposure of EHEC to static adhesive force causes a fraction of GrlA to transition to the cytoplasm, where it becomes available to bind and activate the ler promoter. Additional exposure of adherent EHEC to flow increases the fraction of wild-type GrlA released from the membrane. While adherence alone is sufficient to trigger GrlA relocalization, the process is further reinforced by flow. These observations suggest that GrlA regulation responds to mechanical stimuli in a force-dependent manner. This is in contrast to the recently identified fro operon in Pseudomonas aeruginosa, a flow-regulated bacterial circuit that can sense shear rate, but is not force responsive (18).

Fig. 3.

Wild-type GrlA is sequestered at the inner membrane in planktonic cells and relocalizes to the cytoplasm in response to mechanical stimuli. Representative maximum intensity projections of z-stack images of planktonic EHEC (A–D), or EHEC attached to a microfluidic channel under static conditions (E–H), or exposed to flow (I–L). EHEC expresses GFP fusions to GrlA WT (A, E, and I), GrlA ML1 (B, F, and J), GrlA ML2 (C, G, and K), GFP alone (O), or were strained with FM4-64FX membrane dye as a control (D, H, and L). (Scale bars, 1 µm.) (M) Setup of microfluidic experiment to visualize GFP localization following EHEC exposure to mechanical stimuli – adhesive force (light blue), or adhesive + fluid shear force (dark blue). (N) Percentage membrane-localized GFP in planktonic (black), adherent (light blue), and adherent + fluid shear-exposed (dark blue) EHEC. Quantification based on ≥100 cells/condition. Means ± SEM are shown. p < 0.0001 (****), P ≥ 0.05 not significant (ns).

The GrlA Inhibitor GrlR Associates with Acidic Membrane Lipids Independent of GrlA, and Its Localization Is Not Mechanoresponsive.

GrlA is regulated by several known mechanisms: Its transcription is subject to positive feedback through Ler (19), posttranscriptional regulation via Hfq (20), and posttranslational regulation by the inhibitor GrlR (21). This allows multiple environmental cues to converge on GrlA and titrate its levels and activity in the cell. GrlR forms a complex with GrlA. Because residues in GrlA that are critical for GrlR and DNA binding overlap (Fig. 1A), GrlR binding to GrlA competitively inhibits ler activation (12, 22). Previous work has suggested that GrlR binds lipids in vitro, so we further characterized GrlR interactions with EHEC membrane lipids. Recombinant, purified GST-GrlR bound to EDL933 membrane lipid extracts, albeit with slightly lower apparent affinity than GrlA (Fig. 4A). Further characterization of lipid-binding specificity of GrlR revealed that, like GrlA, purified GST-GrlR preferentially binds to acidic phospholipids, of which cardiolipin is the predominant species in E. coli membranes (Fig. 4B). GFP-GrlR fusions localized to the bacterial membrane of planktonic EHEC in vivo (Fig. 4C) and colocalized with the cardiolipin-specific fluorescent dye NAO (SI Appendix, Fig. S1D). Next, we tested whether exposure to mechanical stimuli would alter the distribution of GFP-GrlR across the cell. In planktonic EHEC, GFP-GrlR intensity showed a biomodal distribution with peaks toward the cell edges (SI Appendix, Fig. S3E), similar to wild-type GFP-GrlA (SI Appendix, Fig. S3A). In contrast to GrlA, exposure to adhesive force alone or in combination with flow (at a shear force of 10 dynes/cm² for 4 h) did not alter the localization pattern of GFP-GrlR (SI Appendix, Fig. S3 F–H). GrlR remained bound to the lipid membrane in mechanically stimulated EHEC (Fig. 4 D–F).

Fig. 4.

GrlR associates with membrane lipids in vitro and in vivo independent of GrlA, and its localization is not mechanoresponsive. (A) Nitrocellulose membranes spotted with serial dilutions of EHEC membrane lipid extracts (1,000 to 0.32 µg/mL lipid and solvent blank) were incubated with 1 µM recombinant, purified GST-GrlR, and bound protein detected using α-GST and α-mouse HRP and ECL detection. n = 3. (B) Membrane lipid strips prespotted with 15 different lipid species (Left) were incubated with 5 µM GST-GrlR and bound proteins detected as in A. n = 3. Representative maximum intensity projections of z-stack images of EHEC expressing GFP-GrlR grown planktonically (C), attached to a microfluidic channel under static conditions (D), or exposed to fluid shear force (E). (Scale bars, 1 µm.) (F) Comparison of % membrane localized GFP-GrlR in planktonic (black), adherent static (light blue), and adherent + shear exposed (dark blue) EHEC. Quantification based on ≥100 cells/condition (n = 3). Means ± SEM are shown. P ≥ 0.05 not significant (ns).

Inhibitions of GrlA by GrlR and by Membrane Sequestration Are Genetically Separable Features.

The partial structure of GrlA (residues 1 to 106) in complex with its inhibitor GrlR has been solved previously and has revealed key residues of GrlA involved in GrlR binding (12). These are basic residues (R53/R54/R64/R65/K66) in the conserved helix-turn-helix (HTH) motif of GrlA (Fig. 1A), which is also required for DNA binding. Consequently, GrlA-GrlR binding inhibits activation of the ler promoter. It is unclear, however, if mutation of key residues in the GrlA HTH motif would also affect its membrane lipid association and thus, mechanoregulation. In addition, it is unclear whether mutation of the lipid-binding residues that we identified in the C-terminal region (res. K111/R114 in the case of GrlA ML1 and res. R132/K135 in the case of GrlA ML2, respectively) would have an impact on the interaction of GrlA with GrlR.

We constructed two versions of recombinant GrlA with mutations in the HTH motif: GrlA R53A R54A (termed GrlA MR1) and GrlA R65A K66A (termed GrlA MR2). These point mutants were previously determined to disrupt both GrlR and ler promoter binding (12). Both recombinant, purified GST-tagged proteins bound to EHEC membrane lipid extracts with affinities comparable to wild-type GrlA (Fig. 5A). When tested in lipid overlay assays, using membrane lipid strips, both GST-GrlA MR1 and GST-GrlA MR2 showed binding specificity for acidic membrane lipids similar to wild-type GrlA (Fig. 5B). Next, we expressed GFP-GrlA MR1 or GFP-GrlA MR2 in EHEC and imaged their localization in planktonic, attached, and attached cells exposed to flow (SI Appendix, Fig. S3). A large fraction of both GFP-GrlA MR1 and MR2 (91% and 93%, respectively) was bound by the membrane in the absence of mechanical stimuli (Fig. 5 E, H, and K and SI Appendix, Fig. S4A). Following bacterial attachment to microfluidic channels, the membrane-localized fraction of GFP-GrlA MR1 and MR2 was significantly reduced compared to planktonic EHEC (Fig. 5 F, I, and K and SI Appendix, Fig. S4B). Following bacterial attachment to channels and exposure to flow, both proteins relocalized to the cytoplasm (Fig. 5 G, J, and K and SI Appendix, Fig. S4C).

Next, we tested the ability of GST-tagged wild-type GrlA, the lipid-binding mutants ML1 and ML2, and the GrlR-binding mutants MR1 and MR2 to bind His-tagged GrlR (Fig. 5C). Wild-type GrlA, GrlA ML1, and ML2 were able to bind and pull down His-GrlR. In agreement with previously published results (12), neither GrlA MR1, GrlA MR2, nor GST alone bound to His-GrlR (Fig. 5C). Finally, we tested the extent to which overexpression of each of the GFP-GrlA variants promoted ler:lacZ transcription in the absence of mechanical stimuli using β-galactosidase assays in EHEC EDL933 (Fig. 5D). As previously observed (Fig. 1D), ler:lacZ was transcribed at a low level in the presence of wild-type GFP-GrlA. Expression of both GFP-GrlA ML1 and ML2 lipid-binding mutants significantly enhanced ler:lacZ transcription, while transcription in the presence of either GFP-GrlA MR1, GFP-GrlA MR2 (GrlR-binding mutants), or GFP alone was significantly lower than with wild-type GrlA (Fig. 5D). These data suggest that the GrlA residues required for GrlR/DNA binding and membrane binding are distinct and located in separate regions of GrlA (HTH motif, res. 39 to 66 versus C-terminal res. 111 to 135, respectively). Mutations that prevent GrlR binding and DNA binding/ler induction do not disrupt the ability of GrlA to respond to mechanical stimuli. Conversely, mutations that abolish GrlA’s function as a mechanosensor by rendering it unable to bind to the membrane do not disrupt its ability to bind to GrlR and/or DNA.

GrlA Relocalizes Rapidly in Response to Mechanical Stimuli.

While the above experiments suggest that GrlA changes its subcellular localization in response to mechanosensation, they do not provide insight into the kinetics of this transition relative to the mechanical stimulus. Thus, it is unclear if release of GrlA from the membrane is a direct response to mechanosensation, or its transition to the cytoplasm is a secondary effect and requires transcriptional changes that take place following mechanosensing. To address this, we set up timecourse experiments and imaged GFP-GrlA localization and ler transcription simultaneously. To image ler activity in single bacteria, we used a ratiometric probe consisting of an inducible ler:CFP reporter and a constitutive tet:mCherry reporter. We calibrated the upper and lower bounds of ler induction using two calibration strains: Expression of ler:CFP in EHEC EDL933 cells + GFP-GrlA ML1 allowed for maximum ler induction (CFP/mCherry ratio of 1.0). For the low end of ler induction (CFP/mCherry ratio of 0.01), we expressed a promoterless CFP reporter in EHEC EDL933 cells + GFP-GrlA ML1. Next, we expressed the ler:CFP/mCherry ratiometric probe in EHEC EDL933 + GFP-GrlA WT. GrlA localization and ler induction were imaged in attached bacteria over a period of 240 min. Although GrlA began to be released from the membrane within 10 min (Fig. 6 D and M) following mechanical stimulation (the earliest timepoint imaged), and its distribution was significantly different 20 min (Fig. 6 E and N) following adhesion compared to planktonic cells (Fig. 6 C and L), ler activity at 10 and 20 min postattachment was not significantly different from planktonic cells (Fig. 6K). Redistribution of GrlA continued from 30 to 240 min (Fig. 6 F–H and O–Q), while induction of ler above basal activity in planktonic cells was first seen at 30 min (Fig. 6F), peaked at 60 min, and plateaued through 240 min postadherence (Fig. 6K). Imaging and ratiometry of promoterless CFP reporters expressed in EHEC + GFP-GrlA WT and EHEC + GFP-GrlA ML1 attached to channels showed that, while the GrlA localization patterns were identical in those cells and in ler:CFP cells, their CFP/mCherry ratios were close to background (Fig. 6 I and J). Together, these data demonstrate that GrlA redistribution in response to mechanosensing occurs rapidly (within 10 min or less in some cells). The timescale on which GrlA responds to mechanical stimuli suggests that it responds directly to mechanical stimuli and does not require de novo protein synthesis as a result of transcriptional changes downstream of mechanosensation.

Fig. 6.

Relocalization of GrlA in response to mechanical stimuli occurs rapidly. EHEC expressing GFP-GrlA WT and a ratiometric probe (inducible ler:cfp + constitutive tet:mcherry) were grown planktonically (C) or attached to microfluidic channels for indicated times (D–H), fixed, and imaged. EHEC expressing GFP-GrlA ML1 + promoterless cfp (A) and ML1 + ler:cfp (B) were used for calibration. Attached EHEC expressing GrlA WT and GrlA ML1 + promotorless cfp (I and J) were used as additional controls for the timecourse. (K) Transcriptional activity was quantified from single cell CFP/mCherry ratios. Values are means, SEM (boxes), and min/max ratios (error bars) from ≥100 cells/condition (N = 3). p < 0.0001 (****), P ≥ 0.05 not significant (ns) compared to planktonic EHEC expressing GFP-GrlA WT. Mean GFP intensity distributions ± SEM for EHEC GFP-GrlA WT in planktonic EHEC (L) or cells attached for 10 to 240 min (M–Q), for 50 cells/condition. (Scale bars, 2 µm.)

GrlA and Its Ability to Respond to Mechanical Stimuli Are Required for Optimal Virulence Regulation and Colonization.

During infection, EHEC undergoes a drastic change in environment. As a result, virulence regulation is fine-tuned by many different chemical and physical cues that converge on the ler promoter. To understand the role of GrlA and its ability to respond to mechanical cues in the context of other environmental cues and regulators, we measured ler transcription in EHEC strains that carried wild-type or mutant grlA alleles, a grlA deletion, or a grlR deletion, during infection.

Initially, we looked at the ability of EDL933 that carries different grlA alleles to infect epithelial cells, using bacterial adherence and ler:gfp transcription as readouts. Following a 3-h infection of HeLa epithelial cells, EHEC strains that harbored either grlA:ML1 (Fig. 7B) or grlA:ML2 (Fig. 7C) mutations or a grlR deletion (Fig. 7D) attached to epithelial cells significantly more than wild-type bacteria (Fig. 7 A and F). In contrast, a EHEC ∆grlA strain (Fig. 7E) was hypoadherent compared to the wild-type strain (Fig. 7F). Both EHEC grlA:ML1 and EHEC grlA:ML2, as well as EHEC ∆grlR showed more ler:gfp transcription than the EHEC wild-type strain, while EHEC ∆grlA showed significantly less ler:gfp transcription (Fig. 7G).

Fig. 7.

GrlA expression and its ability to respond to mechanical stimuli are required for optimal virulence regulation and intestinal colonization. Representative maximum intensity projections of z-stack images of HeLa cells infected with EHEC that contain grlA:WT (A), grlA:ML1 (B), or grlA:ML2 (C) alleles on the chromosome, ΔgrlR (D), or ΔgrlA (E) deletion strains, transformed with the ler:gfp reporter plasmid. DNA (blue, row 1), attached EHEC (red, row 2), ler:gfp transcriptional activity (green, row 3), and merged channels (row 4). (Scale bars, 10 µm.) (F) Quantification of attached bacteria/cell. (G) Quantification of ler transcription. For F and G, values are means, SEM (boxes), and min/max values (error bars) (n = 3, total ≥100 cells/condition). (H) EHEC burden/fish following 20 h of colonization of 6 dpf zebrafish via foodborne infection. Values are individual data points (circles), means ± SEM (n = 18 to 20 fish/condition). p < 0.0001 (****), p < 0.001 (***), p < 0.005 (**), p < 0.05 (*). Mean burdens for mutants expressed as % of wild-type burden (blue, Top).

Finally, we asked whether GrlA’s ability to be regulated by mechanical stimuli was important for colonization of the intestinal tract in vivo, using a previously established larval zebrafish model of foodborne EHEC infection (23). The zebrafish is a vertebrate model with similar intestinal physiology, developmental, and regulatory features to that of mammalian model systems (24, 25). In addition, it has the advantage of high statistical power due to large clutch sizes and ease of maintenance, as well as feasibility for intravital imaging of infections (26).

Conventionally colonized zebrafish (6 d postfertilization) were administered a dose of ∼7 × 10⁵ bacteria/fish, using the protozoan Paramecium caudatum as a vehicle, as described previously (27, 28). EHEC burden per fish was measured following 24 h postinfection, using dilution plating of homogenized samples on CHROMagar O157 to distinguish EHEC from conventional microbiota. Wild-type EHEC caused robust colonization of all larvae (n = 20), whereas both deletion strains (EHEC ∆grlA and EHEC ∆grlR) as well as both strains with allelic replacements of grlA, rendering them unresponsive to force (EHEC grlA:ML1 and grlA:ML2), had significantly decreased bacterial burdens (Fig. 6H). These data suggest that GrlA, and its ability to respond to mechanical stimuli, is required for optimal colonization and virulence regulation. The absence of GrlA renders EHEC hypoinfective both in vitro and in vivo. Mutations in GrlA that render the transcriptional regulator unresponsive to mechanical cues lead to overactivation of ler transcription, and a hyperadherent phenotype in vitro. Loss of the GrlA inhibitor GrlR, which is a force-independent regulator, causes a similar phenotype, albeit to a lesser extent. In vivo, dysregulation of GrlA and virulence genes downstream of GrlA leads to a colonization defect in the zebrafish intestinal tract, suggesting that the ability to appropriately integrate physical cues from the environment to modulate expression of virulence-related genes is essential for fitness and persistence within the host GI tract.

Conclusions

Over recent years, it has become apparent that the ability of microbes to perceive mechanical cues in their environment and integrate them to regulate virulence gene expression and broader aspects of their physiology is a common trait among many bacterial species. For example, both P. aeruginosa (5, 6, 29) and Caulobacter crescentus (30) use retraction of type IV pili on rigid surfaces as a cue to induce genes associated with a sessile lifestyle, in a manner that depends on the second messenger cyclic-di-GMP. Enzymatic turnover of the cyclic di-nucleotide allows rapid responses to changes in mechanical environment. In addition to pili, mechanosensing in C. crescentus also requires the flagellar motor as a sensor, while flagellar appendages are dispensable (31). Downstream activation of a cyclic-di-GMP dependent glycosyltransferase leads to holdfast production, a polysaccharide-based adhesin required for bacterial attachment and biofilm formation. Here, we describe an evolutionarily divergent strategy that allows rapid adaptation of EHEC to the highly dynamic mechanical properties of the gastrointestinal tract.

The mechanosensitive transcriptional activator GrlA is regulated by subcellular compartmentalization. In response to mechanical stimuli, GrlA is released from the inner membrane and relocalizes to the cytoplasm where it activates the virulence-specific promoter ler.

While we can conclude that GrlA is a mechanosensor (i.e., “a cellular component that modulates its biochemical activity in response to transmitted forces” and “induces a downstream mechanoresponse”) (32), the corresponding mechanotransmitter, i.e., the force-bearing structure upstream of GrlA, remains unknown. It is also unclear whether an additional mechanosensor acts upstream of GrlA. Previous work has reported that E. coli undergoes transient membrane depolarization followed by calcium influx in response to mechanical stimuli (33). It is conceivable that these oscillations in membrane potential and/or fluctuations in intracellular cation concentrations could affect the electrostatically driven interaction between C-terminal basic residues in GrlA and the acidic phospholipid cardiolipin we identified here as necessary and sufficient for GrlA’s sequestration at the inner membrane in planktonic EHEC.

Understanding the molecular details of mechanoregulation of EHEC virulence may also present an opportunity for exploiting steps along this pathway as a treatment strategy for EHEC infections. Antibiotic treatment of EHEC infections is associated with an increased risk of developing hemolytic uremic syndrome, a severe complication that can lead to acute kidney failure and death (34, 35). There remains a critical need for development of nonantibiotic treatments that successfully reduces EHEC virulence without increasing toxin-associated pathologies. Targeting key virulence regulators, such as GrlA, and perhaps even altering physiological aspects of the gut environment, such as fluid movement within the intestine during infection, may offer alternatives in EHEC infection treatment.

Materials and Methods

Bacterial Strains, Plasmids, and Growth Conditions.

A list of plasmids and strains can be found in SI Appendix, Tables S1 and S2. The EHEC wild-type strain used was serotype O157:H7 EDL933. Chromosomal point mutants and deletion strains were constructed, using gene doctoring, as previously described (36). Unless otherwise stated, bacterial cultures were grown in lysogeny broth at 37 °C for 16 h, diluted into Dulbecco's Modified Eagle Medium 1:100, and grown at 37 °C to an OD₆₀₀ of 0.4 (midlog phase).

Data Availability Statement.

All data and associated protocols are available in the manuscript and SI Appendix. Strains and plasmids are available upon request from the authors.