Enterotoxigenic Escherichia coli heat-labile toxin drives enteropathic changes in small intestinal epithelia

Abstract

Enterotoxigenic E. coli (ETEC) produce heat-labile (LT) and/or heat-stable (ST) enterotoxins, and commonly cause diarrhea in resource-poor regions. ETEC have been linked repeatedly to sequelae in children including enteropathy, malnutrition, and growth impairment. Although cellular actions of ETEC enterotoxins leading to diarrhea are well-established, their contributions to sequelae remain unclear. LT increases cellular cAMP to activate protein kinase A (PKA) that phosphorylates ion channels driving intestinal export of salt and water resulting in diarrhea. As PKA also modulates transcription of many genes, we interrogated transcriptional profiles of LT-treated intestinal epithelia. Here we show that LT significantly alters intestinal epithelial gene expression directing biogenesis of the brush border, the major site for nutrient absorption, suppresses transcription factors HNF4 and SMAD4 critical to enterocyte differentiation, and profoundly disrupts microvillus architecture and essential nutrient transport. In addition, ETEC-challenged neonatal mice exhibit substantial brush border derangement that is prevented by maternal vaccination with LT. Finally, mice repeatedly challenged with toxigenic ETEC exhibit impaired growth recapitulating the multiplicative impact of recurring ETEC infections in children. These findings highlight impacts of ETEC enterotoxins beyond acute diarrheal illness and may inform approaches to prevent major sequelae of these common infections including malnutrition that impact millions of children.

Fig. 1. Heat-labile toxin modulates expression of multiple genes in intestinal epithelia.

Model at left depicts the E. coli heat-labile toxin based on PDB structure 1LTS with the A1 subunit in blue, the A2 region in yellow, and pentameric B subunit in green. The E211K mutation of mLT is in the active site of the A1 subunit. a Scatterplot of RNAseq data right depicts differential expression profiles of Caco-2 cells following exposure to a heat-labile toxin (n = 2) relative to untreated cells (n = 2) and cells treated with the biologically inactive mLT (n = 2). (Because expression profiles of untreated and mLT-treated cells were virtually identical, their combined expression profiles totaling n = 4 replicates are compared here to LT-treated cells). b RNA-seq data from polarized small intestinal ileal enteroids treated with LT (n = 3) compared to control untreated (n = 3) cells. Differentially expressed genes were identified by DESeq2.

Fig. 2. Heat-labile toxin modulates multiple genes involved in microvillus assembly.

a Diagram at the top (adapted from ref. 30) depicts molecules involved in key elements of microvillus development. b Heatmaps of RNA-seq data obtained following treatment of Caco-2 intestinal cells (left) with mLT (n = 2 biologically independent samples) or LT (n = 2) relative to untreated cells (n = 2); and ileal enteroids (right) treated with LT (n = 3) relative to control untreated cells (n = 3). Comparisons were made with DESeq2. Bars indicate absolute log₂ fold change values + SE. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 10⁻⁴, and *****p ≤ 10⁻⁵.

Fig. 3. Heat-labile toxin impairs the effective formation of small intestinal microvilli.

a Quantitative RT-PCR data from n = 9 biologically independent samples for selected microvillus genes comparing untreated (−) and LT-treated (+) ileal epithelial cells (enteroid line 235D). b Villin production is suppressed in small intestinal enteroids following treatment (t + 18 h) with LT (100 µg/ml). Shown are representative confocal images obtained showing membrane (CellMask, blue), nuclei (white), villin (gold), and merged image. The graph at right shows apical villin geometric mean fluorescence intensity data relative to the corresponding cytoplasmic signal. Each symbol (n = 12) represents a unique region of interest. For a, b ****p < 0.0001, ***p < 0.001 by Mann–Whitney two-tailed testing. c TEM images of small intestinal microvilli following treatment LT (right) compared to control untreated cells (left). The graph at right shows the length of microvilli when enteroids (n = 2 biologically independent samples) are treated before (pre) and after (post) differentiation on polarized ileal cells ****<0.0001 by ANOVA (Kruskal–Wallis, nonparametric testing).

Fig. 4. Heat-labile toxin alters the transcription of multiple brush border SLC genes.

a Heatmap indicating key SLC genes modulated by heat-labile toxin (LT) compared to enzymatically inactive E112K LT mutant (mLT), or untreated (ø) Caco-2 cells (left) and human small intestinal (ileal) enteroids (Hu235D, right). Bars indicate absolute log₂ fold change values + SE. Comparisons were made with DESeq2. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 10⁻⁴, *****p ≤ 10⁻⁵. Real-time qRT-PCR confirming LT-mediated modulation of genes in ileal (Hu235D) enteroids encoding b the major thiamine transporter SLC19A3 and c the SP1 cis-regulatory element. Data reflect two independent experiments with two replicates each. d Uptake of [³H]-thiamine by Hu235D cells is impaired following LT treatment. Data presented in b–d are from two independent experiments with n = 3 replicates each. (*<0.05, **<0.01 by Mann–Whitney two-tailed comparisons).

Fig. 5. ETEC disrupts in vivo formation of small intestinal microvilli.

Timeline at the top depicts the challenge with ETEC or control nontoxigenic isolate or sham (PBS) challenge. a Quantitative PCR results for genes involved in brush border development in small intestinal samples obtained from infant mice (n = 9/group) 7 days after challenge with toxigenic ETEC (jf876), nontoxigenic ETEC (jf4763, LT⁻/ST⁻) PBS controls. Comparisons between data represent ANOVA, Kruskal–Wallis testing where ***p ≤ 0.001, **p ≤ 0.01, and *p ≤ 0.05. b Immunofluorescence images of small intestinal sections showing villin expression (blue), and nuclei (white). c Mean villin fluorescence intensity normalized per enterocyte (n = 4 mice/group) ****p ≤ 0.0001 by Mann–Whitney (two-tailed) nonparametric comparisons. d Representative transmission electron microscopy (TEM) images of the small intestinal brush border from mice challenged with toxin-negative (∆, left) and toxigenic ETEC (wt, right). e Microvillus length ****p ≤ 0.0001 by Mann–Whitney (two-tailed) nonparametric comparisons. Data represent geometric mean length from n = 5 mice per group in three independent experiments. f TEM images from mice challenged with jf570 (eltA::Km^R), sham PBS controls, or mice challenged with wild-type ETEC. g Length of microvilli (dashed horizontal lines represent geometric means). ****p < 0.0001 by Kruskal–Wallis.

Fig. 6. Maternal vaccination with LT mitigates microvillus disruption inneonatal mice.

Timeline depicts vaccination and challenge (top): maternal intranasal (i.n.) vaccinations with 10 µg LT/immunization (yellow arrows on days 0, 14, 28) and neonatal challenge at 3 days of age (green arrow) followed by sacrifice and tissue collection at 7 days post-infection (gray arrow). a Kinetic ELISA data of from triplicate samples of breast milk anti-LT (IgA, and IgG in n = 2 immunized dams and 1 un-immunized control). *<0.05 by Mann–Whitney two-tailed nonparametric testing. b Anti-LT antibodies in the gastric contents of neonatal mice at day 53. ****p < 0.0001 by Mann–Whitney two-tailed comparisons. c Representative transmission electron microscopy images of brush border microvilli from unvaccinated mice challenged with wild-type ETEC (left), vaccinated mice challenged with wild-type ETEC, and vaccinated un-challenged controls. d Microvillus lengths (based on image analysis of n = 5 mice group) ****<0.0001 Kruskal–Wallis comparisons.

Fig. 7. Enterotoxigenic E. coli heat-labile toxin impairs production of HNF4 nuclear receptors.

a Heatmap demonstrating the impact of LT on transcription of paralogous transcription factors HNF4α and HNF4γ in Caco-2 cells (left) and ileal enteroids, (Hu235D, right) ø untreated, mLT mutant LT. b qRT-PCR (TaqMan) data confirming decreased transcription of HNF4 transcription factors following treatment of enteroids with LT (n = 9 biologically independent samples). ****<0.0001 by Mann–Whitney two-tailed comparisons. c HNF4γ is decreased in nuclear fractions obtained from small intestinal enteroids following treatment with LT. Shown in the HNF4γ immunoblot are samples from four independent experiments, with the graph below-representing quantitation of signal intensity normalized to the lamin B1 nuclear protein (p = 0.0017, paired t-test, one-tailed). Bars indicate mean ± 95% confidence intervals. d HNF4 transcription Gaussia luciferase reporter assay showing a decrease in signal following treatment of TR104-transfected Caco-2 cells with LT. 2 experimental replicates (n = 15 samples total) ***p < 0.001 Wilcoxon matched pairs, one-tailed). e Confocal microscopy of representative Ileal sections from sham-challenged (PBS) left, and ETEC-infected mice (right). Immunofluorescence intensity of HNF4γ signal in sections from n = 5 control mice, and n = 6 ETEC-challenged mice. Membranes (yellow) were strained with CellMask orange (Thermo Fisher C10045), nuclei (blue) were stained with DAPI, and HNFγ immunostaining was represented in white. Each symbol represents a microscopic region of interest. Bars represent geographic means (p < 0.0001, Mann–Whitney two-tailed comparisons).