Host-derived CEACAM-laden vesicles engage enterotoxigenic Escherichia coli for elimination and toxin neutralization

Abstract

Enterotoxigenic Escherichia coli (ETEC) cause hundreds of millions of diarrheal illnesses annually ranging from mildly symptomatic cases to severe, life-threatening cholera-like diarrhea. Although ETEC are associated with long-term sequelae including malnutrition, the acute diarrheal illness is largely self-limited. Recent studies indicate that in addition to causing diarrhea, the ETEC heat-labile toxin (LT) modulates the expression of many genes in intestinal epithelia, including carcinoembryonic cell adhesion molecules (CEACAMs) which ETEC exploit as receptors, enabling toxin delivery. Here, however, we demonstrate that LT also enhances the expression of CEACAMs on extracellular vesicles (EV) shed by intestinal epithelia and that CEACAM-laden EV increase in abundance during human infections, mitigate pathogen-host interactions, scavenge free ETEC toxins, and accelerate ETEC clearance from the gastrointestinal tract. Collectively, these findings indicate that CEACAMs play a multifaceted role in ETEC pathogen-host interactions, transiently favoring the pathogen, but ultimately contributing to innate responses that extinguish these common infections.

Keywords: cell adhesion molecules; diarrhea; enterotoxigenic Escherichia coli (ETEC); extracellular vesicles; host–pathogen interactions.

Fig. 1.

CEACAM expression alters kinetics of ETEC intestinal colonization (A) ETEC shed in stool following challenge of either C57BL/6NCrl control mice (n = 8) or CEABAC10 CEACAM-expressing mice (n = 6). Dashed lines connect geometric means. *<0.05, **<0.01 Mann–Whitney two-tailed nonparametric comparison between groups. (B) Proportion of mice remaining colonized (≥1 CFU/mg stool) based on fecal shedding data. Shown are combined results of two independent experiments with total of n = 17 control (C57BL/6NCrl) and n = 21 CEABAC10 mice. P = 0.0004 Log-rank (Mantel-Cox) comparison of survival curves.

Fig. 2.

ETEC–CEACAM interactions in the intestinal lumen of CEACAM-expressing transgenic mice. Shown are confocal microscopy images of (A) ETEC H10407 attached to small intestinal villus enterocytes (arrows) and to CEACAM+ material in the lumen. (B). ETEC in the lumen reside in a CEACAM+ matrix. (C) Small (~100 to 300 nm) CEACAM+ structures engage ETEC in the intestinal lumen. The inset shows ETEC surrounded by CEACAM + structures in the lumen. In each panel anti-CEA antibodies were used to identify CEACAMs (blue) and anti-O78 antibodies were used to identify ETEC H10407 (yellow, serotype O78). (D and E) CEACAM6-positive vesicle, and clusters of EV isolated from the ileum of the CEABAC10 mouse. (F) CEACAM6+ EV clustered on the surface of bacteria isolated from ileal lavage following H10407 challenge. (G) CEACAM6-positive EV isolated from CEABAC10 mouse feces shown by immunogold labeling of anti-CEACAM6 monoclonal (9A6). (H) Flow cytometry of GFP+ bacteria isolated from fecal resuspension supernatants following challenge of CEABAC10 mice with H10407(pGFPmut3.1), showing the proportion of GFP+ bacteria that colabeled with CEACAMs (blue) vs. those which remain unlabeled with CEACAMs (yellow). (I) Majority of ETEC shed in feces are eliminated in large clusters of CEACAMs. The panel represents individual Z-stack confocal image of fecal resuspension (GFP+ bacteria pseudocolored yellow).

Fig. 3.

CEACAM-laden EV block ETEC enterocyte engagement. (A) Clusters of CEACAM+ EV are interposed between the brush border of small intestinal enterocytes and ETEC. (B) CEACAM+ EV emerging at the surface of microvilli engage ETEC (Arrows in A and B indicate immunogold labeling of CEACAM6). (C) Concentrated supernatants (sn) from polarized small intestinal enteroid monolayers impair ETEC pathogen–host interaction. Columns at Left, Middle, and Right of the graph indicate no treatment (sn−), treatment with concentrated supernatant (sn+, CEACAM+), and after supernatant absorption of CEACAMs (sn+, CEACAM−). Shown are the results of two independent experiments. ****P < 0.0001 (Kruskal–Wallis). Horizontal lines indicate median and quartiles. Inset immunoblot indicates CEACAM6 before and after absorption. (D) EV purified from small intestinal enteroids inhibit attachment of ETEC to target small intestinal epithelia. Shown are the results of two replicate experiments. “−”= untreated ileal enteroids (n = 50, total); Data reflect bacteria per region of interest (ROI) for wells without (−) and with (+) exogenously added purified EV (n = 55 total). ****<0.0001 Mann–Whitney two-tailed nonparametric comparisons. (E) LT treatment of enteroids enhances CEACAM production on EV and EV-mediated inhibition of ETEC adhesion to target Caco-2 cells **** <0.0001, *0.02 (Kruskal–Wallis). Inset immunoblot shows impact of LT treatment on CEACAM6 expression on EV isolated after treatment for 24, 72 h. Intestinal alkaline phosphatase (IAP) is shown as a loading control. (F) EV isolated from murine intestine inhibit ETEC adhesion. Shown are confocal imaging data of ETEC adherent to Caco-2 cells in the absence of EV, EV isolated from CEABAC10 mice (CEACAM+), and parental C57BL6 mice (CEACAM−). ****<0.0001 (Kruskal–Wallis). Total of n = 100 fields from two independent experiments.

Fig. 4.

EV scavenge and neutralize ETEC toxins. (A) EV contain ganglioside receptors for LT. Shown are anti-LT dot immunoblots demonstrating LT-binding to increasing amounts of immobilized BSA (negative control, Top) GM-1 ganglioside (positive control, Bottom) and EV. (B) Immobilized EV bind LT. Shown are kinetic ELISA in which EV bound to ELISA plates capture increasing amounts of LT. Summary of three independent experiments, ***0.0008, **0.0014 by Kruskal–Wallis comparisons to no LT control. (C) Molecular pulldown study using anti-CEACAM antibody coated protein G beads (bait) to pull down EV (prey), and bound LT. Following incubation with biotinylated LT-B (LT-B*), immunoblot was developed with avidin-HRP to detect bound toxin subunit. Immunoblots (Left panel) verify presence of IAP and CEACAM6 in EV input prey. Biotinylated LT-B is indicated in blots of pulldown and controls. (D) EV block LT-mediated activation of cAMP in target Caco-2 cells. Data reflect baseline-corrected values (raw data-baseline/baseline) and are from two independent experiments (n = 10 total replicates). Analysis by Kruskal–Wallis. (E) EV impede toxin delivery by ETEC. Caco-2 cAMP levels following infection with ETEC H10407 ± EV. Data are from two independent experiments (n = 8 total replicates). Analysis by Kruskal–Wallis. (F) Fractionation of GST–STh/EV complexes by SEC. Shown below the chromatogram are dot immunoblots for CEACAM6 and GST corresponding to individual fractions. Control fractions from GST–EV interactions are shown below. (G) Western immunoblot of EV-containing fractions from SEC demonstrating coelution of CEACAM6 and GST–STh. (H) EV compete with T84 intestinal epithelial cells for ST binding. Shown are the results of two independent confocal microscopy experiments, with symbols representing mean fluorescence intensity of GST–STh binding in individual fields. Vertical lines represent geometric means. *P = 0.02 (Mann–Whitney). (I) EV neutralize STh activation of cGMP. Shown are results of three independent experiments (n = 15 technical replicates). ****<0.0001, and ***n = 0.0004 by Kruskal–Wallis.

Fig. 5.

Comparative tandem mass tag spectrometry of EV from LT-treated (127C label) and control (ø, 126C) enteroids. Subset of proteins identified in EV: which were increased (Left) and unchanged or decreased (Right) following exposure of enteroids to LT (100 ng/mL overnight ~18 h).

Fig. 6.

human ETEC infection is associated with increases in fecal CEACAMs. (A) CEACAM-laden EV are shed in stool of children with acute ETEC diarrheal illness. Immunogold-labeled TEM image of EV isolated from diarrheal stool demonstrates detection of CD9 (larger 18 nm particles, white arrows), and CEACAMs (12 nm, black arrow). (B) Schematic (created with BioRender.com) of kinetic ELISA strategy to capture and detect CEACAM+ EV from human stool. (C) Graph depicts summary of two independent experiments performed on samples from five patients naturally infected with ETEC and five healthy controls (icddr,b in Dhaka), each with three technical replicates (total of n = 30 data points for each day). Day 1 = day of presentation to icddr,b. Negative control wells contain only buffer used in sample extraction. (D) Summary of two technical replicates in two separate experiments with samples obtained from human volunteers on the day prior to infection (d-1) and on days 7, 28 following challenge with ETEC H10407 (n = 17; 3 with mild-moderate diarrhea, 8 with severe diarrhea, and 6 with no diarrhea following challenge). Statistical analyses by Kruskal–Wallis nonparametric comparisons: ****P < 0.0001, ***P = 0.0002, **P = 0.002, *P = 0.0127.