Crohn’s Disease Pathobiont Adherent-Invasive E coli Disrupts Epithelial Mitochondrial Networks With Implications for Gut Permeability

Abstract

Background & Aims

Adherent-invasive Escherichia coli are implicated in inflammatory bowel disease, and mitochondrial dysfunction has been observed in biopsy specimens from patients with inflammatory bowel disease. As a novel aspect of adherent-invasive E coli–epithelial interaction, we hypothesized that E coli (strain LF82) would elicit substantial disruption of epithelial mitochondrial form and function.

Methods

Monolayers of human colon-derived epithelial cell lines were exposed to E coli–LF82 or commensal E coli and RNA sequence analysis, mitochondrial function (adenosine triphosphate synthesis) and dynamics (mitochondrial network imaging, immunoblotting for fission and fusion proteins), and epithelial permeability (transepithelial resistance, flux of fluorescein isothiocyanate–dextran and bacteria) were assessed.

Results

E coli–LF82 significantly affected epithelial expression of ∼8600 genes, many relating to mitochondrial function. E coli–LF82–infected epithelia showed swollen mitochondria, reduced mitochondrial membrane potential and adenosine triphosphate, and fragmentation of the mitochondrial network: events not observed with dead E coli–LF82, medium from bacterial cultures, or control E coli. Treatment with Mitochondrial Division Inhibitor 1 (Mdivi1, inhibits dynamin-related peptide 1, guanosine triphosphatase principally responsible for mitochondrial fission) or P110 (prevents dynamin-related peptide 1 binding to mitochondrial fission 1 protein) partially reduced E coli–LF82–induced mitochondrial fragmentation in the short term. E coli–LF82–infected epithelia showed loss of the long isoform of optic atrophy factor 1, which mediates mitochondrial fusion. Mitochondrial Division Inhibitor 1 reduced the magnitude of E coli–LF82–induced increased transepithelial flux of fluorescein isothiocyanate dextran. By 8 hours after infection, increased cytosolic cytochrome C and DNA fragmentation were apparent without evidence of caspase-3 or apoptosis inducing factor activation.

Conclusions

Epithelial mitochondrial fragmentation caused by E coli–LF82 could be targeted to maintain cellular homeostasis and mitigate infection-induced loss of epithelial barrier function. Data have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO series accession numbers GSE154121 and GSE154122 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE154121).

Graphical abstract

Summary.

Disruption of the mitochondrial network is uncovered as a novel aspect of adherent-invasive Escherichia coli (strain LF82: implicated in inflammatory bowel disease) interaction with epithelial cells; the inhibition of which reduced the magnitude of the bacteria-evoked increases in epithelial permeability.

Studies in animal models and analysis of patients with inflammatory bowel disease (IBD) highlight the importance of the gut microbiota in the etiology and pathophysiology of enteric inflammation.¹ Tissues from patients with IBD often have increased numbers of Escherichia coli compared with those from healthy individuals,² and a subgroup therein, the adherent-invasive E coli (AIEC), has emerged as pathobionts.^3,4 Attaching via pili and long polar fimbriae,^5,6 AIEC can cause reactive oxygen species (ROS) and proinflammatory cytokine production by epithelial cells^7,8 and increase epithelial permeability.⁹ AIECs have enhanced survival within macrophages,¹⁰ and can exaggerate the severity of murine colitis.^11,12 A comprehensive knowledge of the mechanism(s) by which AIECs participate in IBD is lacking. Intriguingly, the description of the reference strain of AIECs, LF82, showed infected enterocytes with swollen mitochondria that lacked cristae,³ which are cardinal features of dysfunctional mitochondria.

Mitochondrial disease can present with gastrointestinal symptoms,¹³ and perturbed mitochondrial form and function is an aspect of enteric inflammation.¹⁴ Colonic biopsy specimens from patients with IBD can have swollen epithelial mitochondria, reduced adenosine triphosphate (ATP), an altered mitochondrial proteome, and increased sensitivity to uncoupling of oxidative phosphorylation.14, 15, 16, 17 Furthermore, mutation in the IRGM1 gene, which is associated with mitochondria, is a susceptibility trait for IBD,¹⁸ as are mutations in the OCTN2 gene, which facilitates carnitine uptake for fatty acid β-oxidation and ATP synthesis by mitochondria.¹⁹ Cell culture and murine studies have defined mechanisms whereby dysfunctional mitochondria contribute to enteric inflammatory disease, such as increased epithelial paracellular permeability, internalization of commensal bacteria, and increased ROS production.^20,21

Typically depicted as kidney-bean–shaped organelles, mitochondria actually exist in a network that is constantly remodeling by fission and fusion processes dictated in large part by the guanosine triphosphatases (GTPases), dynamin-related protein 1 (Drp1), and mitofusins 1 and 2 and the long isoform of optic atrophy factor 1 (OPA1-L), respectively.²² Fusion facilitates homogeneity of molecules throughout the network, subcellular targeting of ATP, and compensation for mutated mitochondrial DNA. Mitochondrial fission is necessary for cell division, mitophagy, and apoptosis. Excessive fragmentation of the mitochondrial network has been described in neurodegenerative disease, cardiac ischemia–reperfusion injury, kidney disease, and metabolic disease: all of which can have an inflammatory component.23, 24, 25 Systemic delivery of P110, a molecule designed to reduce mitochondrial fission by preventing Drp1–mitochondrial fission 1 protein (Fis1) interaction, reduced the severity of chemical-induced colitis in mice.²⁶

The mitochondrial network can be disrupted significantly in cultured cells infected with microbial pathogens,²⁷ although the functional significance of this is poorly understood. Thus, juxtaposing the concepts of a bacterial etiology for IBD and mitochondrial dysfunction as a contributor to enteric pathophysiology, we sought to determine if AIEC affected epithelial mitochondrial dynamics. E coli–LF82, unlike commensal E coli, caused fragmentation of the epithelial mitochondrial network: at early time points after infection (2–4 h) this was reduced by inhibition of Drp1 activity, while loss of OPA1-L occurred by 16 hours after infection. The reduction in barrier function that accompanied infection was ameliorated in part by Mitochondrial Division Inhibitor 1 (Mdivi1), a pharmacologic that selectively inhibits Drp1 activity. Thus, fragmentation of the mitochondrial network is shown as a novel aspect of E coli–LF82 epithelial interaction with implications for barrier function that could be relevant to AIEC triggering initiation or disease relapse in IBD.

Results

Transcriptomics Indicates Altered Mitochondrial Function in E coli–LF82–Infected Epithelia

T84 epithelia exposed to E coli–LF82 or E coli–F18 showed massive changes in gene transcription after 4 and 16 hours (P < .05) (Figure 1A), with principal component analysis showing clustering of the experimental groups indicative of different transcriptomic programs (Figure 1B). The majority of gene expression changes were small (<2-fold, and statistically significant) and there was considerable overlap in the genes affected by infection; however, gene pathway analysis showed differences in the epithelial response to the 2 strains of E coli that were statistically significant (Figure 1C). Specifically, gene pathways relating to cell viability, cellular stress, cell cycle, antigen handling and response to pathogens, signal transduction, cell-to-cell communication, and mitochondrial function were time-dependently differentially regulated by E coli–LF82 compared with epithelia exposed to E coli–F18 (Figure 1C). E coli–LF82 infection (10⁸ colony-forming unit [CFU]/mL, 4 h) affected 1265 genes and 105 signaling pathways that were not affected by E coli–F18 (Figure 2A and B). With a lower E coli inoculum and longer exposure (10⁴ CFU/mL, 16 h), 964 genes and 69 pathways were uniquely affected by E coli–LF82 (Figure 2C and D) in T84 epithelia. Changes in mitochondrial function were noteworthy (eg, the master transcriptional regulator of mitochondrial biogenesis, peroxisome proliferator-activated receptor γ coactivator 1α [PGC-1α], was reduced approximately 2-fold by exposure to E coli–LF82 [4 h, 10⁸ CFU/mL], but not E coli–F18).

Figure 1.

RNA sequence analysis shows significant differences in the epithelial cell response to E coli strains. T84 epithelia were treated with adherent-invasive E coli–LF82 or commensal E coli–F18. (A) The number of significant gene expression changes. (B) Principle component (PC) analysis shows obvious grouping of the epithelia by treatment, and (C) pathway analysis shows distinct E coli–evoked changes in epithelial gene expression as a function of time and dose of inoculum.

Figure 2.

Distribution of statistically significant changes in (A and C) gene expression and (B and D) gene pathway analysis in T84 epithelia treated with E coli–LF82 or E coli–F18 (dose and time shown above graphs) (red, number of gene expressions increased; blue, number of gene expressions decreased; numbers above bars indicate the total number of genes and pathways altered in the infected epithelia; dots and connections under each bar indicate that the expressions were either unique to that group or shared by the groups).

E coli–LF82 Induces Mitochondrial Dysfunction

Immunostaining showed significant epithelial-associated E coli–LF82 but not E coli–HB101 (Figure 3A). Transmission electron microscopy showed mitochondrial ultrastructural abnormalities in epithelia exposed to E coli–LF82 but not E coli–HB101 (Figure 3B): the mitochondria were electron-lucent, swollen, and lacked cristae. In accordance with these data, E coli–LF82 evoked a loss of mitochondrial membrane potential similar to that evoked by carbonyl cyanide m-chlorophenyl hydrazone (CCCP) (Figure 3C). After 4 hours of infection with E coli–LF82, but not E coli–F18, epithelial ATP levels were reduced by approximately 20% (12–24 epithelia, 6 experiments; data not shown), with greater reductions in ATP occurring 16 hours after exposure to E coli–LF82 (Figure 3D).

Figure 3.

Adherent-invasive E coli induces mitochondrial dysfunction in human colonic epithelium. (A) Immunostaining shows significant association of E coli–LF82 with T84 epithelia in contrast to epithelia treated with the commensal E coli–HB101 (representative images; n = 3 monolayers). (B) Transmission electron microscopy shows swollen and electron-lucent mitochondria in T84 cells infected with E coli–LF82 (n = 4). e, E coli; n, nucleus; pc, paracellular space. Arrows indicate mitochondria. (C) Quantification of tetramethylrhodamine ethyl ester (TMRE) dye average integrated fluorescence intensity, as an indicator of mitochondrial membrane potential, showed a significant reduction by 3.5 and 8.5 hours of exposure to E coli–LF82, 10⁸ and 10⁴ CFU/mL, respectively, but not E coli–HB101 (10⁴ CFU/mL, 8.5 h) (4 μmol/L CCCP was used as a positive control). ∗P < .05 Kruskal–Wallis, followed by the Dunn multiple comparison test compared with control, n = 5–16. (D) T84 cell levels of ATP are reduced after treatment with the pathobiont E coli–LF82 (∗P < .05, t test, n = 8 and 13 epithelial preparations) (data are means ± SEM).

Infection With E coli–LF82 Causes Mitochondrial Fragmentation

Assessment of the T84-epithelial mitochondrial network after exposure to E coli–LF82 (2–4 h 10⁸ CFU/mL or 16 h 10⁴ CFU/mL) showed significant fragmentation that consistently exceeded the negligible or minor changes evoked by either of the commensal E coli strains (Figure 4A–C). Loss of fused mitochondrial networks and increased fragmented networks occurred as early as 2 hours after exposure to E coli–LF82 (10⁸ CFU/mL), with almost complete disruption by 16 hours after treatment with 10⁴ CFU/mL. E coli–LF82 also evoked mitochondrial fragmentation in human colon-derived Caco2 and HT-29 epithelial cell lines (data not shown). In contrast, neither dead E coli–LF82 nor spent medium from E coli (cultured alone or with enterocytes) caused appreciable mitochondrial fragmentation in T84 cells (Figure 4D). Although infection with wild-type E coli–LF82 consistently led to significant fragmentation of the epithelial mitochondrial network, cells infected with E coli–LF82^ΔFimH had mitochondrial networks similar to uninfected cells (Figure 4E). Another AIEC, NRG857c, which is less invasive in T84 cells than E coli–LF82, evoked a small increase in mitochondrial fragmentation that was not statistically different from uninfected cells (Figure 5). Growth curves showed negligible differences between the E coli strains and growth was unaffected by any of the drugs used herein. T84 epithelia infected with enteropathogenic E coli (strain E2348; 10⁸ CFU/mL, 4 h) showed a significant increase in the number of cells with a fragmented mitochondrial network (6 epithelial preparations from 2 experiments; data not shown).

Figure 4.

Adherent-invasive E coli–induced mitochondrial fragmentation is dependent on FimH attachment/invasion. (A) Representative confocal images of T84 mitochondrial networks stained with MitoTracker and (B and C) network quantification. ∗P < .05 compared with control, ^#P < .05 compared with HB101; 2-way analysis of variance (ANOVA) followed by the Tukey multiple comparison test, n = 5–6 epithelial monolayers (20 cells/monolayer) from 3 experiments. (D) Representative images and quantification showing that neither dead E coli–LF82 or spent medium (Med.) from E coli causes mitochondrial fragmentation in T84 epithelia. ∗P < .05 compared with control, 2-way ANOVA followed by the Tukey multiple comparison test, n = 8–9 epithelial preparations (20 cells/monolayer) from 3 experiments. (E) Bacteria lacking the FimH molecule are less invasive than the wild-type E coli–LF82 and fail to significantly induce fragmentation of the T84 epithelial cell mitochondrial network (representative images match the quantification bars). ∗P < .05 compared with control, 2-way ANOVA followed by the Tukey multiple comparison test, n = 6–18 epithelial monolayers from 5 experiments; 4 μmol/L CCCP. Data are means ± SEM. Ctrl, control; wt, wild type.

Figure 5.

Degree of epithelial mitochondrial fragmentation parallels pathobiont invasiveness. T84 epithelia were infected with either the commensal E coli–HB101 or one of the adherent-invasive strains of E coli–LF82 or E coli–NRG857c. (A) Significantly (∗P < .05, 1-way analysis of variance [ANOVA], Tukey multiple comparison test) greater invasion by E coli–LF82, compared to (B) fragmentation of the mitochondrial network 2 hours after infection with E coli-NRG857c resembling that of uninected cells (data are means ± SEM; ∗P < .05 compared with control [Ctrl], 2-way ANOVA, Tukey multiple comparison test in panel B, n = 9 epithelial preparations [20 cells/monolayer] from 3 experiments; control, noninfected epithelia).

Fragmentation of mitochondria can be associated with increased ROS production²⁸ and E coli–LF82 has been shown to induce ROS production in epithelial cells.²⁹ Co-treatment of E coli–LF82–infected epithelia with the antioxidant vitamin C, or the mitochondrial-targeted antioxidant Mito-TEMPO, did not prevent the mitochondrial fragmentation (data not shown) (antioxidant capacity of vitamin C and mito-TEMPO was confirmed by reductions in measurable ROS after treatment with rotenone; data not shown).

Targeting Drp1 Limits E coli–LF82–Induced Epithelial Mitochondrial Fragmentation

Total Drp1 and activated ^ser616p-Drp1 levels in isolated mitochondria (data not shown) or whole-cell extracts were not altered consistently in T84 epithelia treated with E coli–LF82 (2–8 h 10⁸ CFU/mL; 8–16 h 10⁴ CFU/mL) (Figure 6A). Because of many post-translational modifications in Drp1,²² which were not assayed here, an absence of detectable increases in Drp1 does not negate a role for this GTPase in mitochondrial fragmentation. Subsequently, co-treatment with P110 or Mdivi1 was found to significantly decrease E coli–LF82–induced (10⁸ CFU/mL, 2–4 h) mitochondrial fragmentation (Figure 6B and C). Corroborating these data, epithelia in which Drp1 expression was reduced by small interfering RNA (siRNA) showed less E coli–LF82–induced mitochondrial fission after 2 hours of infection (Figure 6C, middle panel). However, neither P110 nor Mdivi1 prevented the degree of mitochondrial fragmentation observed after 6–8 hours of infection with 10⁸ CFU/mL (Figure 6C, lower panel) or 16 hours of treatment with 10⁴ CFU/mL E coli–LF82 (data not shown) (adding P110 every 2 hours also did not ameliorate the E coli–LF82–evoked mitochondrial fragmentation observed at later time points in the infection).

Figure 6.

Inhibition of Drp1 activity reduces mitochondrial fragmentation induced by adherent-invasive E coli–LF82. (A) Western blot and densitometry analysis showing that neither total nor phosphorylated Drp1 (Ser616 site) protein expression in whole-cell lysates is altered consistently in T84 epithelia infected with E coli–LF82. (B and C) Inhibition of Drp1 activity with P110 (10 μmol/L) or Mdivi1 (5 μmol/L) or siRNA (30 nmol/L) knockdown of Drp1 protein reduces the severity of E coli–LF82–evoked T84 epithelial cell mitochondrial fragmentation assessed at 2 and 4 hours after infection, but not at 6 or 8 hours after exposure to the viable bacteria. Representative images of MitoTracker-stained mitochondrial networks. Middle: ∗P < .05 compared with mock transfection uninfected; ^#P < .05 compared with mock transfection infected. Bottom: ∗P < .05 compared with control (ctrl), 2-way analysis of variance followed by the Tukey multiple comparison test; n = 4–11 epithelial monolayers (20 cells/monolayer), 2–4 experiments. con.siRNA, control siRNA; Sts, 20 μmol/L staurosporine. Data are means ± SEM.

Infection With E coli–LF82 Results in Loss of OPA1-L

Analysis of protein expression showed a time- and dose-dependent loss of OPA1-L in E coli–LF82–infected epithelia. Infection with the control bacteria, E coli–HB101 and E coli–F18, at high inoculum could reduce OPA1-L levels, but this was not a statistically significant consistent observation (Figure 7). With an inoculum of 10⁶ CFU/mL E coli–LF82, OPA1-L levels began to diminish by 10 hours after infection and were statistically significantly reduced by 16 hours after infection (Figure 7B). The loss of epithelial OPA1-L owing to infection with E coli–LF82 was time- and dose-dependent (Figure 7C).

Figure 7.

Infection with E coli–LF82 results in loss of OPA1-L. (A) Western blots showing the time- and concentration-dependent (CFU inoculum) loss of OPA1-L in E coli–LF82–infected T84 epithelia and (B and C) corresponding densitometry analysis. Data are means ± SEM, 4 μmol/L CCCP for 2 hours. ∗P < .05 compared with control (ctrl), 1-way analysis of variance, the Dunnett multiple comparison test. OPA1-S, Optic Atrophy Factor 1 - Short Form.

Mdivi1 Partially Rescues the Epithelial Barrier Defect Evoked by E coli–LF82

Exposure to E coli–LF82 reduces the barrier function of human colon-derived epithelial cell lines.³⁰ The decrease in transepithelial resistance (TER) that occurred by 8 hours after infection with E coli–LF82 (10⁸ CFU/mL) was not affected by either inhibitor of mitochondrial fragmentation, P110 or Mdivi1, or the pan-caspase inhibitor, Z-VAD (Figure 8A). In contrast, assessment of the transepithelial flux of larger molecules showed that in 4 experiments Mdivi1 significantly prevented the E coli–LF82–evoked flux of 70 kilodalton dextran (Figure 8B), and trended toward a reduced apical-to-basolateral passage of E coli across the epithelium (Figure 8C) (Mdivi1 and P110 alone did not affect epithelial barrier function; data not shown).

Figure 8.

E coli–LF82–induced decreased epithelial barrier function is reduced by co-treatment with an inhibitor of mitochondrial fission. (A) The E coli–induced (10⁸ CFU/mL) decrease in TER across filter-grown T84 monolayers was unaffected by co-treatment with Mdivi1 (5 μmol/L), P110 (10 μmol/L), or the pan-caspase inhibitor Z-VAD (100 μmol/L) (∗P < .05 compared with control [Ctrl], 2-way analysis of variance [ANOVA], Tukey comparison test, n = 12 epithelial preparations from 4 experiments; starting TER = 1000–2000 Ω × cm²). In contrast, (B) barrier assessment by 70 kilodalton dextran–fluorescein isothiocyanate (FITC) flux (∗P < .05 compared with ctrl, 2-way ANOVA, Tukey multiple comparison test, experiment number was considered as a separate variable rather than pooled data owing to large variations in fold change) and (C) transepithelial passage of the E coli were both reduced by Mdivi1 (t test). Data are means ± SEM.

E coli–LF82 Evokes Mitochondrial Release of Cytochrome C

Excessive mitochondrial fragmentation can precede apoptosis. Forty-six percent of cells exposed to staurosporine had cytochrome C in their cytoplasm (Figure 9A and B). With treatment of E coli–LF82 (6 h, 10⁸ CFU/mL), cytoplasmic cytochrome C was not different from control epithelium, whereas by 8 hours after infection approximately 40% of the infected cells were positive for cytosolic cytochrome C (some preparations showed approximately 80% positivity). P110 and Mdivi1 did not prevent the E coli–LF82–evoked mitochondrial release of cytochrome C (Figure 9B), consistent with their inability to block mitochondrial fragmentation at later time points in infection (Figure 9D). Caspase activation can occur after cytochrome C release. T84 epithelia treated with E coli–LF82 (10⁸ CFU/mL) 8–12 hours previously showed no evidence of cleaved caspase-3 in whole-cell protein extracts, and, in accordance with other studies,³¹ there was a general reduction in total caspase-3 in the treated cells (Figure 9C). Immunostaining did not show a consistent localization of apoptosis-inducing factor (AIF) from the mitochondria to the nucleus in response to E coli–LF82 (10⁸ CFU/mL, 6 h or 8 h), and release of interleukin (IL)18, as a surrogate marker of necroptosis, was not increased after exposure to E coli–LF82 (16 h, 10⁴ CFU/mL; n = 4 epithelial preparations; IL18 <2 pg/mL in control and infected T84 epithelia). However, the nuclei of E coli–LF82–infected T84 cells were pyknotic and gel electrophoresis showed substantial fragmentation of the DNA that was unaffected by co-treatment with either P110 or Mdivi1 (Figure 6D).

Figure 9.

E coli–LF82–infected T84 epithelia show increased cytosolic cytochrome C, absence of caspase-3 activation, and obvious DNA fragmentation. (A) Representative immunostaining of T84 epithelia shows increased cytosolic cytochrome C at 8 hours after infection that was not apparent at 6 hours. (B) Co-treatment with the fission inhibitors, P110 (10 μmol/L) or Mdivi1 (5 μmol/L) did not prevent the mitochondrial release of cytochrome C (∗P < .05 compared with ctrl; 1-way analysis of variance followed by the Dunnett multiple comparison test). Ctrl, control; Sts, staurosporine, 20 μmol/L. (C) Western blot analysis of whole-cell lysates showed a reduction in total caspase-3 protein and an absence of cleaved caspase-3 at 8–12 hours after infection with E coli–LF82 (10⁸ CFU/mL) (Sts was used as a positive control) (blot representative of 4 separate experiments). (D) DNA fragmentation was obvious at 8 and 16 hours after infection. M, Mdivi1; P, P110; n = 3; E, empty lane, no sample added. Data are means ± SEM. bp, basepair; DAPI, 4′,6-diamidino-2-phenylindole.

The ability of P110 and Mdivi1 to suppress E coli–LF82–induced epithelial mitochondrial fragmentation after 2–4 hours of infection but not more than 8 hours infection was intriguing. Killing extracellular bacteria with gentamicin and removal of E coli–LF82 after 4 hours, followed by epithelial assessment 4 hours later, showed significantly less mitochondrial fragmentation and cytoplasmic cytochrome C compared with cells continuously exposed to the bacteria for 8 hours (Figure 10).

Figure 10.

E coli–LF82–evoked release of cytosolic cytochrome C and mitochondrial fragmentation are reduced by gentamicin. T84 epithelia were treated with E coli–LF82 and examined for (A) the presence of cytosolic cytochrome C (∗P < .05 compared with ctrl, ^#P < .05 compared with 8 hours LF82; 1-way analysis of variance [ANOVA], Tukey multiple comparison test) or (B) pattern of the mitochondrial network 8 hours later. Time-matched epithelia were treated with gentamicin (200 μg/mL, 1 h) 4 hours after the addition of E coli and then examined 4 or 8 hours later (ie, the 8- and 12-hour gentamicin groups) (∗P < .05 compared with ctrl, ^#P < .05 compared with 8 h LF82 + gentamicin; 2-way ANOVA followed by the Tukey multiple comparison test; n = 5–12 epithelial preparation from 3–4 experiments; Ctrl, control; Sts, 20 μmol/L staurosporine). Data are means ± SEM.

Discussion

The mechanisms responsible for impaired mitochondrial structure and function in biopsy specimens from patients with IBD³² are unknown. The concept of the pathobiont in which environmental conditions promote the emergence of pathogenic traits in a bacterium normally considered a commensal is intriguing.³³ By using reductionist models we uncovered that the Crohn’s disease–associated pathobiont, E coli–LF82, can cause massive disruption of the epithelial mitochondrial network, possibly via the sequential activation of Drp1 followed by loss of the profusion GTPase OPA1-L.

Exposure of T84 epithelia to E coli–LF82, E coli–F18, or E coli–HB101 significantly affected gene expression, underscoring the high degree of communication between enterocytes and bacteria. Considerable overlap occurred in the genes affected by exposure to pathobiont or commensal; however, the magnitude and direction of changes often differed, and pathway analysis showed significant differences in the effects of the bacterial strains. Notably, E coli–LF82 affected genes related to mitochondrial form/function and the decrease in PGC-1α (controls mitochondrial biogenesis) messenger RNA distinguished it from the commensal bacteria. Although bacterial-driven perturbation of mitochondrial function is common,34, 35, 36 we are unaware of other reports of microbes affecting PGC-1α expression, with the exception of the hepatitis C virus protein NS5A.³⁷

Epithelia infected with E coli–LF82, but not E coli–HB101, show reduced mitochondrial membrane potential (ΔΨ_m) that could account for the mitochondrial swelling and loss of cristae, and hence lower levels of ATP. Maintenance of ΔΨ_m is critical in mitochondrial function, and dissipation of the H⁺ gradient indicates malfunction that can trigger mitochondrial fission or cell death. Infection with E coli–LF82 evoked a time- and dose-dependent fragmentation of mitochondria in gut-derived epithelia. The bacterial pathogens Listeria monocytogenes, Shigella flexneri, Mycobacterium tuberculosis, and Vibrio cholerae all cause mitochondrial fragmentation,38, 39, 40, 41 and this often was accompanied by reduced ΔΨ_m.

Listeriolysin O, vacuolating cytotoxin E, early secreted antigen 6 kilodalton, and vacuolating cytotoxin E, have been implicated in mitochondrial fragmentation induced by L monocytogenes, V cholera, M tuberculosis, and Helicobacter pylori, respectively.^38,39,42,43 A soluble factor is unlikely to be critical to the effect of E coli–LF82 because medium from bacterial cultures (±T84 cells) did not evoke mitochondrial fragmentation (the possibility of a short-lived mediator contributing to mitochondrial fragmentation cannot be unequivocally ruled out). Elatrech et al²⁹ reported increased ROS production by E coli–LF82–infected T84 epithelia, and ROS can disrupt mitochondrial dynamics.²⁸ Vitamin C and Mito-TEMPO reduced rotenone-induced ROS by T84 cells, yet neither affected E coli–LF82–induced mitochondrial fragmentation. This suggests that ROS are not a major factor contributing to the mitochondrial fragmentation, and could, in fact, be produced as a consequence of the fragmentation and perturbed oxidative phosphorylation.

Lack of an effect of dead E coli–LF82 indicated that viable bacteria attaching to and/or invading the epithelium caused fragmentation of the mitochondrial network. In accordance, epithelia exposed to E coli–LF82 deficient in type 1 pili, which are critical for adhesion and invasion,⁹ had mitochondrial networks similar to noninfected epithelia. Furthermore, the AIEC strain NRG857c that was less invasive in T84 epithelia caused little mitochondrial fragmentation compared with E coli–LF82. E coli–LF82’s lack of toxins/virulence factors suggests that either attachment to the cell, the process of invasion, and/or the presence of intracellular bacteria triggers mitochondrial fragmentation. Our findings with enteropathogenic E coli (EPEC) suggest that attachment to the epithelial cell membrane could initiate a cascade of events leading to mitochondrial fragmentation. However, injection of EPEC effector molecules may be the critical factor affecting host cell mitochondrial dynamics. The EPEC protein E coli secreted protein F (EspF) contains a mitochondrial localization sequence and was found to lower ΔΨ_m; however, mitochondrial dynamics were not assessed in that study.³⁶ Consistent with the hypothesis that invasion of E coli–LF82 was critical for mitochondrial fragmentation, it has been suggested that intracellular mechanostimulation resulting from bacteria is sufficient to trigger mitochondrial fission via Drp1.⁴⁴ This could account for the slight increase in mitochondrial fragmentation in T84 cells exposed to E coli–HB101, some of which are internalized because of epithelial membrane turnover.⁴⁵

The GTPase Drp1 drives fission by binding to mitochondrial fission factor, mitochondrial dynamics proteins 49 and 51, or Fis1. Pharmacologic inhibition of Drp1 or protein knockdown reduced E coli–LF82–induced mitochondrial fragmentation. Similarly, Mdivi1 has been used to implicate Drp1 in S flexneri–induced⁴¹ and H pylori–induced epithelial mitochondrial fragmentation.⁴² Addition of Mdivi1 or P110 at the start of the E coli–LF82–epithelial co-culture reduced the degree of mitochondrial fragmentation observed 4 hours later, but not at 8–16 hours after infection. These data can be explained by continued replication of the bacteria overwhelming the epithelium or may indicate that mitochondrial fission in the latter stages of infection are Drp1-independent. With respect to the latter, HeLa cells lacking Drp1 showed significant mitochondrial fragmentation when challenged with listeriolysin O.⁴³

OPA1-L facilitates mitochondrial fusion to work in opposition to profission events. E coli–LF82–infected epithelia showed loss of OPA1-L and this lack of a critical profusion molecule would contribute to mitochondrial fragmentation. Similarly, mitochondrial fragmentation in Brucella abortus–infected HeLa cells was accompanied by loss of mitofusin proteins, without evidence of Drp1 involvement.⁴⁶ Thus, we hypothesize that E coli–LF82–induced fragmentation of epithelial mitochondria occurs via the following sequence: loss of ΔΨ_m leads to Drp1 activation, driving active mitochondrial fission that subsequently is enhanced by or superseded by reduced capacity to fuse mitochondria resulting from loss of OPA1-L.

Mitochondrial fragmentation may be a common feature of microbial infections,²⁷ whether it benefits the host or microbe is unclear. It was suggested that mitochondrial fragmentation caused by L monocytogenes and S flexneri protected the cell from subsequent infection and promoted cell-to-cell spread of the bacteria, respectively.^40,41 Contrarily, hyperfused mitochondrial networks were observed in human umbilical vein endothelial cells (HUVECs) and HeLa cells shortly after infection with Chlamydia trachamatis, and it was speculated that the increase in ATP production favored replication and growth of the bacteria.⁴⁷

AIEC-infected epithelia show reduced barrier function,^30,48 and although the Drp1 inhibitor, Mdivi1, did not affect E coli–LF82–induced reductions in TER (neither did the pan-caspase inhibitor Z-VAD, which blocks caspase-dependent apoptosis), it significantly reduced the concomitant transepithelial passage of fluorescein isothiocyanate dextran (70 kilodalton) and on average the passage of the bacteria was less. The latter are significant because they indicate that a substantial defect in epithelial barrier function evoked by infection with E coli–LF82 can be reduced significantly, at least in the short term, by an inhibitor of mitochondrial fission. Collectively, the data highlight the importance of mitochondrial function (ie, ΔΨ_m, ATP production, remodeling) in maintaining epithelial barrier function.^15,18

E coli–LF82–infected T84 epithelia were characterized by pyknotic nuclei and DNA disintegration indicative of cell death. Lack of IL18 synthesis or receptor-interacting serine/threonine-protein kinase 3 phosphorylation (personal observation) 8–16 hours after infection with E coli–LF82 suggested the enterocytes were not undergoing necroptosis. Reduced ΔΨ_m can foreshadow release of cytochrome C into the cytoplasm as a precursor to cell death, via caspase-3 activation, as shown in H pylori–induced epithelial apoptosis.⁴² Increased cytosolic cytochrome C occurred 8 hours after infection with E coli–LF82 (10⁸ CFU/mL). Neither Mdivi1 nor P110 ablated the cytochrome C release, and, furthermore, we found no evidence of caspase-3 activation in E coli–LF82–infected T84 epithelia, which showed a reduction in total caspase-3 protein, recapitulating data from Dunne et al.³¹ Similarly S flexneri–evoked HeLa cell death after loss of ΔΨ_m occurred independent of caspase activation.^41,49 Apoptosis-inducing factor has been implicated in H pylori– and EPEC-induced epithelial cell death^50,51; however, AIF recruitment to the nucleus in E coli–LF82–treated cells was not increased significantly compared with controls. Thus, although E coli–LF82-induced fragmentation of nuclear DNA indicates cell death, no data were obtained to suggest that the reduced epithelial viability was via caspase activation, AIF mobilization, or necroptosis. Although loss of ΔΨ_m and/or mitochondrial fragmentation can lead to cell death, it remains unclear if these are the underlying reasons for any E coli–LF82–induced decreases in cell viability.

Conclusions

In summary, epithelia exposed to viable E coli–LF82 show loss of mitochondrial function and substantial fragmentation of the mitochondrial network mediated by Drp1 and, potentially, loss of OPA1-L protein. Furthermore, the increased epithelial permeability caused by E coli–LF82 was partially restored by inhibition of mitochondrial fission, showing the importance of normal mitochondrial dynamics in the control of gut barrier function (Figure 11). Using pertinent human-derived epithelial cell lines in conjunction with a Crohn’s disease–relevant pathobiont, the data herein present a novel perspective on AIEC–host interaction, identifying mitochondria as a therapeutic target to alleviate some of the pathophysiology evoked by enteric pathobionts that could be relevant to a cohort of patients with IBD.

Figure 11.

Schematic of model showing how attachment and/or invasion of epithelial cells by adherent-invasive E coli (strain LF82 depicted) can lead to fragmentation of the mitochondrial network via DRP1 activation, and possibly loss of OPA1-L, and mitochondrial disruption as defined by movement of cytochrome C from mitochondria to the cytosol. Infection with E coli–LF82 was not accompanied by evidence of caspase-3 activation. The cell stress experienced results in reduced epithelial function and may have implications for cell activation (ie, alarmin release) and viability (red lines indicate pathway supported by data in the current study).

Methods

Cell Culture

Human colon-derived T84, Caco2, and HT-29 cell lines were maintained as described.⁵² Briefly, T84 cells (passages 40–120) were maintained in a 1:1 Dulbecco’s modified Eagle medium/F12 Ham medium (Sigma, MO) mixture containing HEPES (2 mmol/L; Sigma), L-glutamine (2.68 mmol/L; Gibco, MA), sodium pyruvate (0.6 mmol/L; Sigma), sodium bicarbonate (0.015%; Gibco), and penicillin-streptomycin (120 U/mL penicillin, 0.12 mg/mL streptomycin; Sigma) and 10% (vol/vol) fetal bovine serum. The HT-29 and Caco2 lines (passages 40–60) were cultured in Dulbecco’s modified Eagle medium and constituents similar to T84 epithelia with the addition of nonessential amino acids (1×; Sigma), and L-glutamine at 2.16 mmol/L, penicillin-streptomycin (108 U/mL penicillin, 0.11 mg/mL streptomycin), and 5% fetal bovine serum.

Commensal E coli (strains HB101, F18; used as controls interchangeably in all experiments to safeguard against any strain-specific effect) and AIEC (strains LF82, LF82^ΔFimH, and NRG857c; from Drs P. Sherman [University of Toronto], J.D. Söderholm [Linköping University], and B. Coombes [McMaster University], respectively) and EPEC (strain E2348) (from P. Sherman) were maintained as described.^15,30,53 Two infection paradigms were used: short-term, high-dose E coli (4 h, 10⁸ CFU/mL; multiplicity of infection, ∼100) and longer-term, low-dose (16 h, 10⁴ CFU/mL; multiplicity of infection, ∼0.01). Dead E coli–LF82 (10⁸ CFU/mL) were prepared by 2-hour fixation in 2.5% glutaraldehyde (Sigma). Spent medium was prepared by incubating E coli–LF82 (10⁴ CFU/mL) ± T84 cells for 16 hours, followed by centrifugation and 0.2-μm filtration of the supernatant.

The mitochondrial fission protein-1 (Fis1) inhibitor, P110 (designed to block Drp1–Fis1 binding, a gift from Dr D. Mochly-Rosen⁵⁴) was used as a 30-minute pretreatment at 10 μmol/L (bioactivity of P110 confirmed by its ability to reduce BAPTA [5 μmol/L, 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid] and CCCP [4 μmol/L] induced mitochondrial fragmentation; both Sigma). The selective inhibitor of Drp1 GTPase activity, Mdivi1 (5 μmol/L; Sigma), was applied concomitantly with E coli.⁵⁵ Staurosporine (20 μmol/L) was used as a positive control in some assays (Sigma). To test a role for ROS in the model, the general antioxidant, vitamin C (0.25 mmol/L), or the mitochondria-targeted antioxidant, Mito-TEMPO (10 μmol/L) (both Sigma), were used as a co-treatment with E coli.²¹

siRNA Knock-Down of Drp1

Drp1 expression was knocked-down using siRNA (Drp1 siRNA, sc-43732; Santa Cruz Biotechnology, TX; control nontargeted siRNA, sc-37007; Santa Cruz Biotechnology) and Lipofectamine 2000 (ThermoFisher Scientifc).⁵² Briefly, siRNA and Lipofectamine were prepared in Opti-MEM culture medium (Gibco) and, when combined with cells, would be 2 μL/well Lipofectamine, 30 nmol/L siRNA, 10⁵ T84 cells in 100 μL/well. As described in the manufacturer’s instructions, the siRNA mixture was added to the Lipofectamine mixture, mixed gently, and incubated for 5 minutes at room temperature and the appropriately adjusted concentration of cells was added, mixed gently, and 10⁵ cells/100 μL was seeded into each well of 8-well chamber slides (Lab-Tek; ThermoFisher Scientific). Medium was changed no later than 16 hours after seeding. Cells were stained with MitoTracker dye (see mitochondrial morphology section) 72 hours after transfection, treated with bacteria (2 h, 10⁸ CFU/mL), and imaged live. Efficiency of siRNA knockdown of Drp1 was confirmed by immunoblotting (Figure 12).

Figure 12.

(A) Representative Western blot and (B) accompanying summary densitometry showing the effectiveness of siRNA (30 nmol/L) against Drp1 to reduce Drp1 protein in T84 epithelial cells, control or infected with E coli LF82 (mock represents identically treated epithelia without exposure to siRNA and csiRNA is a control irrelevant siRNA sequence). Data are means ± SEM. Ctrl, control.

RNA Sequence Analysis

RNA sequencing was performed at the Centre for Health Genomics and Informatics at the University of Calgary, where samples were made into libraries using the Illumina TruSeq Stranded messenger RNA LT library preparation kit and were sequenced on a 75-cycle, high-output NextSeq 500 run. Briefly, transcripts were quantified with kallisto 0.43.1⁵⁶ using Homo sapiens GRCh38 (Ensembl release 90) complementary DNA, with sequence bias correction turned on and 50 bootstraps. The tximport package, 1.12.1, was used to import the kallisto results and aggregate transcript abundances to the gene level, and DESeq2 1.24.0⁵⁷ was used for differential expression. Infected epithelia were compared with noninfected control T84 cell monolayers using the Wald test and differentially expressed genes were selected based on a P value cut-off of .05. Enriched gene sets for both KEGG⁵⁸ and Reactome⁵⁹ pathways were identified with over-representation tests (P < .05) using clusterProfiler 3.10.0⁶⁰ and ReactomePA 1.26.0,⁶¹ respectively. Data were deposited in NCBI's Gene Expression Omnibus and are accessible through GEO series accession numbers: GSE154121 and GSE154122 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE154121).

E coli Localization

T84 cells were seeded on coverslips in 24-well plates and infected with E coli–LF82 or E coli–HB101 (4 h, 10⁸ CFU/mL), fixed with 4% paraformaldehyde for 10 minutes, washed, and permeabilized with 0.1% Triton X-100 in phosphate-buffered saline (Sigma, 2 × 5 min.) and blocked with 10% donkey serum (room temperature, 30 min). After addition of anti–E coli antibody (1:50, 30 min, 37ºC; Abcam, Cambridge, UK), washing (2 × 5 min), donkey anti-goat secondary Alexa Fluor–488 antibody (1:1000, 30 min, room temperature), cells were washed (2 × 5 min), stained with 4′,6-diamidino-2-phenylindole (DAPI, 1:1000, 5 min, room temperature), rinsed, mounted with Dako fluorescent mounting medium (Dako North America, Inc, Agilent Technologies, CA), and viewed on an Olympus (Tokyo, Japan) BX41 wide-field fluorescence microscope.

The gentamycin assay assessed bacterial internalization.¹⁵ T84 cells were grown to 50%–70% confluence as determined by phase-contrast microscopy, infected with E coli, a sample of the medium was collected, and then gentamycin (100 μg/mL, 1 h; Sigma) was added to the culture wells. After rinsing, epithelia were lysed with 1% Triton X-100, and the lysate was cultured for 18 hours at 37ºC on blood agar plates: CFUs were counted and data are presented as the percentage of internalization.

Mitochondrial Function

Mitochondrial membrane potential was assessed by seeding T84 epithelia (1 × 10⁵/well) into 96-well plates followed by washing and incubation in tetramethylrhodamine ethyl ester dye (75 nmol/L, 30 min, 37ºC; ThermoFisher). After rinsing, cells were treated with E coli, images were captured on the IncuCyte ZOOM microscope 30 minutes later at 20× magnification (Essen Biosciences, Sartorius, Gottingen, Germany), and analyzed using the IncuCyte ZOOM software. Loss of fluorescence indicates tetramethylrhodamine ethyl ester leakage from the mitochondria resulting from mitochondrial membrane depolarization. The H⁺ ionophore CCCP (4 μmol/L) was used as a positive control.

Intracellular ATP was assessed using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI) as applied to T84 cells ± E coli–LF82 cultured in 12-well plates and following the manufacturer’s instructions. Samples were collected into white-walled, 96-well plates (Griener Bio-One, KI, Austria), luciferase was measured in a Victor3V 1420 Multilabel Counter (PerkinElmer, Waltham, MA), and data were normalized to protein content in the samples.

Mitochondrial Form and Associated Proteins

T84 epithelia ± E coli were cultured on plastic plates, fixed in gluteraldehyde, processed and sectioned, and the ultrastructure was assessed on a Hitachi (Tokyo, Japan) H-7650 transmission electron microscope.

For mitochondrial network analysis, epithelial cells were seeded on 8-well chamber slides (Lab-Tek, 155409; ThermoFisher Scientific), and stained using the MitoTracker Red CMXRos probe (50 nmol/L, Invitrogen Detection Technologies, Molecular Probes, MA) for 30 minutes at 37ºC, washed, stained with Hoescht dye (1 mg/mL; ThermoFisher), washed, and treated with E coli. Live cell imaging was performed on a Leica (Wetzlar, Germany) DMI6000B Diskovery Flex spinning disk confocal microscope. Cells were chosen randomly based on nuclear staining, fluorescence then was switched to view MitoTracker dye, and mitochondrial morphology was quantified; 20 cells per epithelial preparation were assessed. Mitochondrial networks were classified as fused (interconnecting mitochondrial networks >50% of the cell), fragmented (≥80% spheric mitochondria), or intermediate (combination of fused and fragmented mitochondria, or short tubules not interconnected with each other).⁶²

Whole-cell protein extracts or isolated mitochondria (Mitochondria Isolation Kit for Tissue and Cultured Cells; BioChain, Newark, CA) were subjected to Western blot analysis. Primary antibodies were as follows: OPA1 (1:1000; Abcam), p-Drp1 (1:1000; Cell Signaling, MA), Drp1 (1:2000; Abcam), actin (1:1000; Santa Cruz), cleaved-caspase 3 (1:1000; Cell Signaling), total caspase 3 (1:1000; Cell Signalling), and heat-shock protein-60 (1:1000; Santa Cruz).

Epithelial Viability

Release of cytochrome C from mitochondria into the cytosol and subcellular localization of AIF were determined by fluorescence microscopy in a blinded fashion on paraformaldehyde-fixed epithelial monolayers ± E coli treatment. Antibodies used were rabbit anti–translocase of outer membrane 20 (1:100; Santa Cruz), Alexa Fluor–488 mouse anti–cytochrome C (1:50; BD Pharmingen, CA), and Alexa Fluor–568 goat anti-rabbit (1:1000; ThermoFisher). In other cells, AIF was visualized with a rabbit anti-human antibody (1:100, 4642; Cell Signaling) followed by a Cy3-conjugated anti-rabbit IgG (1:2000, 711-164-152; Jackson Immunoresearch, PA). Images were captured on a Nikon (Tokyo, Japan) A1R laser scanning confocal microscope and image analysis was performed using Ezcolocalisation and ImageJ (National Institutes of Health, Bethesda, MD) plugin (AIF and 4′,6-diamidino-2-phenylindole colocalization compared with Mander’s coefficient of colocalization).⁶³ In addition, DNA was extracted from T84 cells using TRIzol (Invitrogen), 2-μg samples of DNA and 1 kb-plus DNA marker (Invitrogen) were loaded onto a 2% agarose gel with 0.5 μg/mL bromide. After electrophoresis, the gel was examined on a UV transilluminator and images were captured.

Epithelial Barrier Function

TER, flux of fluorescein isothiocyanate dextran (size 70 kilodalton, concentration 200 μg/mL), and transepithelial passage of E coli across filter-grown (3-μm pore) T84 cell monolayers (starting TER, ≥1000 Ω × cm²) were performed as previously described.¹⁵

Data Presentation and Analysis

Data are presented as means ± SEM and statistical significance was accepted at P < .05, where n values are the number of epithelial preparations from multiple combined experiments. When comparing 2 groups, the Student t test was performed. When multiple groups were compared, a 1-way analysis of variance followed by either a Dunnett multiple comparison or Tukey multiple comparison post-test was performed. For nonparametric data, a Kruskal–Wallis followed by the Dunn multiple comparison post-test was performed. When analyzing mitochondrial morphology, data are indicated as different if a statistical difference was observed when comparing either the percentage fused or the percentage fragmented morphology between groups. Analyses were conducted using Prism software version 6.0 (GraphPad Software Inc, CA).