A defect in mitochondrial activity contributes to many diseases. We have shown that monolayers of the human colonic T84 epithelial cell line exposed to dinitrophenol (DNP, uncouples oxidative phosphorylation) and nonpathogenic Escherichia coli (E. coli) (strain HB101) display decreased barrier function. Here the impact of DNP on macrophage activity and the effect of TNF-α, DNP, and E. coli on epithelial permeability were assessed. DNP treatment of the human THP-1 macrophage cell line resulted in reduced ATP synthesis, and, although hyporesponsive to LPS, the metabolically stressed macrophages produced IL-1β, IL-6, and TNF-α. Given the role of TNF-α in inflammatory bowel disease (IBD) and the association between increased permeability and IBD, recombinant TNF-α (10 ng/ml) was added to the DNP (0.1 mM) + E. coli (106 colony-forming units), and this resulted in a significantly greater loss of T84 epithelial barrier function than that elicited by DNP + E. coli. This increased epithelial permeability was not due to epithelial death, and the enhanced E. coli translocation was reduced by pharmacological inhibitors of NF-κβ signaling (pyrrolidine dithiocarbamate, NF-κβ essential modifier-binding peptide, BAY 11–7082, and the proteosome inhibitor, MG132). In contrast, the drop in transepithelial electrical resistance was unaffected by the inhibitors of NF-κβ. Thus, as an integrative model system, our findings support the induction of a positive feedback loop that can severely impair epithelial barrier function and, as such, could contribute to existing inflammation or trigger relapses in IBD. Thus metabolically stressed epithelia display increased permeability in the presence of viable nonpathogenic E. coli that is exaggerated by TNF-α released by activated immune cells, such as macrophages, that retain this ability even if they themselves are experiencing a degree of metabolic stress.
- bacterial translocation
- T84 cells
a number of putative causative agents/mechanisms have been proposed for the inflammatory bowel diseases (IBDs): genetic predisposition, infection, dysregulated immune reactions, triggers from the commensal flora, and a defect in epithelial permeability (44). Abnormal mitochondria have been observed in epithelial cells in tissues resected from patients with Crohn's disease (28), indicating metabolic stress (i.e., ATP depletion) in these individuals. Similarly, reductions in mitochondria acetoacetyl CoA thiolase have been shown in the epithelium of patients with ulcerative colitis, which would result in reduced ATP synthesis from butyrate, and hence these cells would experience a degree of metabolic stress (36). Metabolic stress can arise as a consequence of infection, ischemia, and inflammation itself (8, 14). Thus we hypothesized that metabolic stress would impinge upon the normal interaction between enterocytes and the commensal flora, resulting in an increase in epithelial permeability that could predispose individuals to gut dysfunction and mucosal inflammation. Support for this hypothesis comes from analyses of kidney-derived epithelia showing that ATP depletion affects the arrangement of tight junction proteins with a consequent increase in paracellular permeability (2, 5). Our analyses revealed that uncoupling oxidative phosphorylation (i.e., inducing metabolic stress) in epithelial monolayers by treatment with dinitrophenol (DNP) in the presence of viable, noninvasive, nonpathogenic Escherichia coli (E. coli) (strain HB101) resulted in increased paracellular and transcellular permeability (27, 28).
With the emergence of the concept of immunophysiology, it is clear that the function of the intestinal epithelium is affected by many immune cell types (30). The macrophage produces a plethora of messengers, including tumor necrosis factor (TNF)-α. Resident intestinal macrophages have reduced expression of CD14 (the LPS coreceptor) (40) and tend to be less susceptible to activation by bacterial products (41). In contrast, monocytes are sensitive to bacterial products, and it has been postulated that it is monocytes recruited into tissues that are responsible for local pathology (36, 37). For instance, the permeability characteristics of T84 epithelial cell monolayers are unaffected by coculture with LPS-activated lamina propria mononuclear cells (LPMC) isolated from normal human intestine, whereas LPMC from patients with Crohn's disease and monocytes from healthy individuals and patients with Crohn's disease evoke significant increases in epithelial permeability (50). This loss of epithelial integrity was inhibited by neutralizing antibodies to TNF-α. Although TNF-α can directly increase paracellular permeability across selected renal and enteric epithelial cell lines (25, 38), by itself TNF-α has minimal direct effects on the barrier properties of T84 cell monolayers (4, 23), indicating that TNF-α was either activating LPMC and/or working in concert with other mediators released from the LPMC to increase the permeability to T84 cell monolayers (50).
On the basis of these findings, the present study addresses two questions: would macrophages under metabolic stress retain an ability to respond to LPS and produce cytokines; and, would the addition of TNF-α enhance the increase in epithelial barrier permeability caused by DNP + E. coli? The data support a positive feedback-loop hypothesis, in which metabolic stress allows the increased passage of bacteria across the epithelium to activate immune cells, including macrophages, which may also be metabolically stressed, but retains an ability to produce TNF-α. The TNF-α then exerts an additive effect with DNP + bacteria, further diminishing epithelial barrier function that can contribute to the initiation of inflammatory disease or exaggerate ongoing disease.
MATERIALS AND METHODS
Cell Culture and Reagents
The human monocytic cell line, THP-1 (American Type Culture Collection, Manassas, VA), was maintained in suspension culture in RPMI-1640 medium supplemented with 2% (vol/vol) penicillin-streptomycin (Pen-Strep), 36 μM N-2-hydroxethyl-piperazine-N′-2-ethanesulfonic acid (HEPES) (Invitrogen Life Technologies, Burlington, ON, Canada), and 10% (wt/vol) fetal bovine serum (FBS) (CanSera, Toronto, ON, Canada). For differentiation into a macrophage-like phenotype, THP-1 cells were seeded onto six-well sterile plastic culture plates (VWR, Mississauga, ON, Canada) at 250,000 cells/well (unless stated otherwise) and treated with phorbol myristate acetate (PMA, 10 nM; Sigma, St. Louis, MO) (47). Seventy-two h later, medium was replaced with PMA-free medium (37°C) and the experimental conditions established after an additional 24 h of culture (see below).
The human colon-derived T84 crypt-like cell line was used as a model epithelium throughout this study. T84 cells (passages 39–63) were seeded at a density of 106 cells/ml onto semipermeable filters (3-μm pore size, 1.1 cm2 surface area; Costar, Cambridge, MA) and cultured at 37°C in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium, supplemented with 2% Pen-Strep, 1.5% HEPES, and 10% FBS until transepithelial electrical resistance (TER) was ≥750 Ω/cm2 (typically 5–7 days, medium changed every 1–2 days) (23). Epithelial monolayers used here had starting TER of 753–4,400 Ω/cm2 (see figure legends).
Confluent T84 cell monolayers were either 1) cocultured with THP-1 cells (plated into the basal compartment of the culture well), which had been treated with PMA ± the metabolic inhibitor DNP (0.1 mM, Sigma) for 72 h, or 2) exposed to various combinations of DNP, TNF-α (10 ng/ml; R&D Systems, Minneapolis, MN), E. coli [strain HB101, 106 colony-forming units (cfu) from stationary phase of growth], and the inhibitors of NF-κβ activation, pyrrolidine dithiocarbamate (PDTC; 50 μM), NF-κβ essential modifier (NEMO)-binding domain binding peptide (1 μM), BAY 11–7082 (10 μM), and the proteosome inhibitor MG-132 (1 μM) (all Calbiochem, San Diego, CA). All reagents were added to the apical side of the epithelial monolayer, except for TNF-α, which was added into the basolateral compartment of the culture well.
Measurement to Assess Metabolic Stress
THP-1 cells (25,000) were seeded into 96-well plates and differentiated with PMA. Cells were treated with DNP (0.1 mM) for 6 h, rinsed, and culture medium replaced with 100 μl of 0.4 mg/ml 3-(4,5-dimethyl-thiazol-2-yl)-2,5-diphenyltetrazolium bromide salt (MTT) (Sigma). After 4 h at 37°C, the medium was aspirated and replaced with 100 μl acidic isopropanol (0.04N HCl in absolute isopropanol), and absorbance read at 570 nm (with background subtraction at 630 nm) (7). Higher optical density readings indicate greater MTT cleavage by the mitochondria and hence reflect greater mitochondrial activity.
THP-1 cells were seeded in six-well plates (50,000/well), differentiated with PMA for 72 h, rested for 24 h, and then subjected to the various experimental conditions. Subsequently, cells were washed twice with PBS and lysed with 0.5% (wt/vol) trichloroacetic acid (TCA, Sigma). Cells were scraped from the culture plates with a sterile disposable cell scraper (Corning, Cornell, NY), the pH adjusted to ∼7.2 by the addition of 1 M Tris base buffer, and cell debris cleared from the extract by a 10-min centrifugation at 10,000 g. ATP levels were immediately measured by using a commercial ATP determination kit (Invitrogen). Briefly, 20 μl of cell lysate supernatants were added to a 96-well plate (Corning), 80 μl of reaction solution (reaction buffer, 1 mM DTT, 0.5 mM D-Luciferase, and 1.25 μg/ml firefly luciferase) was injected into each sample, and the resulting luminescence measured for 10 s with a Wallac 1420 Victor3 96-well luminometer (PerkinElmer, Waltham, MA). Background luminescence was subtracted (blank wells containing TCA-Tris buffer) and ATP levels read off a standard curve (1 nM to 1 μM). The ATP values for each sample were normalized against total protein [determined by the Bio-Rad/Bradford microplate assay (Bio-Rad, Hercules, CA)] in each sample.
THP-1 cells (105/ml) ± DNP were exposed to LPS (1 μg/ml, Sigma) for 6 or 24 h, supernatants collected, and IL-1β, IL-6, and TNF-α levels measured by ELISA by using paired antibodies (DuoSet ELISA, R&D Systems) following the manufacturer's instructions. All samples were assessed in duplicate serial dilutions, and assays had detection limits of 9 pg/ml.
Assessment of mRNA expression.
Total RNA was extracted from THP-1 or T84 cells with the Trizol method, and cDNA was reverse transcribed from 1 μg of RNA with the iScript cDNA Synthesis kit (Bio-Rad, Mississauga, ON, Canada). Each cDNA sample was incubated in a 25-μl reaction containing 2 mM MgCl2, 0.2 μl/PCR reaction Platinum Taq DNA polymerase (Invitrogen), and 0.4 μM of nucleotide primers [Toll-like receptor (TLR)4 (gene bank no., NM_138554) forward CAACAAAGGTGGGAATGCTT and TLR4 reverse TGCCATTGAAAGCAACTCTG, 317bp; TNF-αRI (gene bank no., NM_001065) forward ACCAAGTGCCACAAAGGAAC and reverse CTGCAATTGAAGCACTGGAA, 263bp; TNF-αRII (gene bank no., BC052977) forward CTCAGGAGCATGGGGATAAA and reverse AGCCAGCCAGTCTGACATCT, 192bp; mannose receptor (gene bank no., NM_002438) forward GGCGGTGACCTCACAAGTAT and reverse ACGAAGCCATTTGGTAAACG, 168bp; arginase-1 (gene bank no., AK128314) forward AAAACCAAGTGGGAGCATTG and reverse CCACTTGTGGTTGTCAGTGG, 196bp; and β-actin (gene bank no., NM_007393) forward CCAGAGCAAGAGAGGTATCC and reverse CTGTGGTGGTGAAGCTGTAG; 437bp]. All primers were designed using the primer3 program (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). All samples underwent PCR under the following conditions: an initial denaturation at 95°C for 3 min, then 35 cycles of 95°C for 30 s, 55°C for 8 s, and 72°C for 20 s, followed by a final extension for 3 min at 72°C. The PCR products were electrophoresed on a 2% agarose gel containing 0.5 μg/ml ethidium bromide and visualized under UV light. Densitometry was performed on digitized images with the ImageJ software package provided by the National Institutes of Health (Bethesda, MD), and specific gene bands were normalized against that of the housekeeping gene, β-actin.
Assessment of epithelial permeability.
Three indices of epithelial barrier function were employed. First, TER, a measure of the predominantly paracellular flux of ions, was measured with a Millicell-ERS voltmeter and matched asymmetrical electrodes (Millipore, Bedford, MA) at the start, at specific intervals, and at the end of each experiment. Data are presented as the % change from starting TER values from each individual epithelial monolayer (47). Second, 100 μM of fluorescein-labeled dextran (3,000 Da, Invitrogen) was added to the apical side of T84 monolayers and 100 μl samples collected from the basolateral compartment of the culture well 4 and 24 h later. Fluorescence was detected in the Wallac 1420 Victor3 spectrophotometer by using excitation and emission wavelengths of 475 and 520 nM, respectively, and the amount of dextran that had crossed the epithelial layer was read off a standard curve. Third, transcytosis of the noninvasive E. coli was examined. Bacteria (106 cfu) were added to the apical aspect of the epithelial monolayers and samples retrieved from the basolateral culture compartment 24 h later and cultured on LB agar plates. Bacterial growth was enumerated on a semiquantitative, logarithmic five-point scale as described previously (24): 0 = no colonies; 1 = <10 colonies; 2 = 10–100 colonies; 3 = >100 countable colonies; 4 = >100 colonies, uncountable, but discernible colonies are evident; 5 = bacterial lawn.
Whole-cell protein extracts of THP-1 or T84 cells were assessed for expression of cleaved caspase-3, as described elsewhere (33). Briefly, cells were lysed in 80–200 μl of ice-cold lysis buffer [20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM NaVO3, and complete EDTA-free protease inhibitor complex (Roche Diagnostics, Indianapolis, IN)], the extracts vortexed (20 s), placed on ice for 10 min, vortexed again (20 s), centrifuged at 13,000 rpm for 5 min at 4°C, and then stored at −80°C. After protein concentrations were determined by using the Bio-Rad/Bradford assay, 40 μg of protein were mixed with loading buffer [(2% wt/vol) sodium dodecyl sulfate, 50 mM Tris·HCl, 100 mM DTT, 1% (wt/vol) normophenol blue, 5% (vol/vol) glycerol] and run in a 10% polyacrylamide gel (100 V for 1 h), and the separated proteins electroblotted to polyvinylidene difluoride membranes (VWR). Membranes were blocked with 5% nonfat Carnation milk for 1 h at room temperature, followed by a 4°C overnight incubation with rabbit IgG against anti-cleaved caspase-3 (1:1,000; Cell Signaling Technology, Beverly, MA). Membranes were washed extensively, incubated with goat anti-rabbit IgG (1:5,000, 1 h at room temperature; Santa Cruz Biotechnology, Santa Cruz, CA), and, after being washed, immunoreactive proteins were visualized by enhanced chemiluminescence (Amersham Pharmacia, Piscataway, NJ) and exposed to Kodak XB film (Eastman Kodak, Rochester, NY).
Trypan blue assessment of epithelial viability.
T84 cells (3 × 105) were seeded into sterile six-well plates, grown to 70–80% (as gauged by phase-contrast microscopy), and then treated for 24 h with DNP + E. coli ± TNF-α. Cells were trypsinized from the culture wells stained with 0.01% Trypan blue, and the percentage of dead cells was calculated.
T84 cells (5 × 106) were suspended in 100 μl of Nucleofector II solution (AMAXA Nucleofector II kit, E-bioscience), into which was added 4 μg of plasmid-encoding pmaxGFP, or 1.07 μg of plasmid, continuing the luciferase gene, but no associated promoter (TAL) or 2.5 μg of a plasmid containing the luciferase gene under the control of an NF-κβ promoter region (TAL and NF-κB plasmids were from Promega; the pmaxGFP was provided in the AMAXA Nucleofector II kit as a positive control to gauge transfection efficiency), and 1 ml of Nucleofector solution R1. The cell sample was transferred into an AMAXA-certified cuvette, which was then placed into the AMAXA Nucleofector II and nucleofected with program T-005. Subsequently the 1 × 106 transfected cells (70–80% efficiency based on GFP expression) were seeded into 24-well plates and cultured for 2 days until ∼80% confluent. Following treatment with TNF-α (10 ng/ml for 4 h), luciferase activity was determined following the manufacturer's instructions for a commercial assay kit (Promega). Briefly, duplicate 20 μl of cell lysates were transferred to a 96-well Optiplate (PerkinElmer), reagents injected by an automated Wallac 1420 Victor3 flurospectro-luminometer system (PerkinElmer), and light emission read 10 s later.
Data presentation and analysis.
Data are presented as the means ± SE, where n values indicate the number of individual epithelial monolayers from a specified number of experiments. Statistical analysis was performed by using ANOVA followed by post hoc pair-wise comparisons with Student's t-test. Data from bacterial translocation studies were assessed using the Chi-squared test. A statistically significant difference was accepted at P < 0.05.
DNP causes metabolic stress in differentiated macrophages and reduces cytokine responses to LPS.
THP-1 cells exposed to 0.1 mM DNP, a dose we have used to evoke metabolic stress in epithelia (28), for 6 h displayed a significant decrease in their ability to cleave the MTT substrate, indicative of reduced mitochondrial activity (Fig. 1A). Corroborating these data, DNP-treated THP-1 cells also showed a time-dependent reduction in ATP content (Fig. 1B). This degree of metabolic stress was not associated with decreased THP-1 cell viability as assessed by exclusion of the vital dye, Trypan blue (data not shown), cleavage of caspase-3 (Fig. 1C), and measurement of total THP-1 protein obtained from cells attached to the culture dish: control = 1.55 ± 0.18 vs. DNP-treated (6 h) = 1.48 ± 0.23 μg protein/μl extract (n = 3). The synthesis and release of the proinflammatory cytokines, IL-1β, IL-6, and TNF-α, by THP-1 cells in response to LPS was reduced by DNP treatment, although the metabolically stressed cells still produced substantial amounts of cytokines (Table 1). This diminished responsiveness to LPS was not accompanied by a reduction in TLR-4 mRNA expression (data not shown).
Macrophages have been classified as classical (i.e., proinflammatory) or alternatively activated macrophages (AAMs), which are regarded as more anti-inflammatory, being involved in tissue resolution after damage (13). Considering the possibility that metabolically stressed macrophages become AAM-like, RT-PCR was performed for markers of these cells: specifically, arginase-1 and the mannose receptor (CD206). Arginase-1 or CD206 mRNA expression was not increased in DNP-treated THP-1 cells compared with control cells (n = 4, data not shown).
Coculture with E. coli-activated macrophages decreases epithelial barrier function.
T84 cell monolayers cocultured with either untreated THP-1 cells or THP-1 cells exposed to DNP for 6 h displayed no significant drop in TER over a 24-h coculture period (Fig. 2). We have previously shown that 24-h exposure to E. coli HB101 alone has minimal effects on the permeability characteristics of T84 monolayers (28, 31). However, when THP-1 cells were treated with E. coli ± DNP for 6 h and confluent filter-grown epithelial monolayers subsequently added to the culture well, the result was a significant drop in TER 24 h later [reduced to 30–40% of pretreatment TER values (Fig. 2)]. These data are in accordance with our previous studies where coculture with monocytes or macrophages activated by bacteria or bacteria products evoked increases in the permeability of T84 cell monolayers (49, 50).
Since macrophages, even metabolically stressed macrophages (Table 1), release a variety of mediators following activation, we opted to focus subsequent studies on TNF-α, a factor which has been implicated in IBD and has been shown to affect endothelial (11, 12), gut (38), and kidney (25, 26) epithelial permeability.
TNF-α enhances DNP + E. coli-induced increased permeability.
TNF-α is a major player in inflammatory disease and in IBD, where it has been implicated in the perturbation of gut barrier function (42, 45, 51). We have previously shown that TNF-α released from LPS-activated macrophages can contribute to reductions in epithelial barrier function (50). In vitro analyses reveal that TNF-α directly disrupts the barrier function of monolayers of human colonic HT-29 and Caco-2 epithelial cell lines (19, 38) but has little direct impact on T84 cell monolayer permeability (23) in the absence of IFN-γ (4, 46). RT-PCR analysis confirmed that T84 cells constitutively express mRNA for both TNF-RI and TNF-RII (Fig. 3, inset), the levels of which were unaltered after 6–24 h of exposure to DNP + E. coli ± TNF-α (data not shown). Given the increase in epithelial permeability caused by exposure to DNP + E. coli and TNF-α production by metabolically stressed macrophages, although at reduced amounts compared with nonstressed cells, we hypothesized that TNF-α would have additive effects with DNP + E. coli in reducing epithelial barrier function.
In response to E. coli + TNF-α, there was a small but statistically significant reduction in TER of ∼25% after 24 h (n = 28, Fig. 3A) [As in previous investigations, treatment with TNF-α alone had negligible effects on TER (data not shown).]. However, the effects of E. coli + TNF-α on epithelial permeability were similar to those evoked by E. coli + DNP, with the exception that the drop in TER caused by DNP + E. coli was statistically significant 12 h after exposure (Fig. 3A). The addition of TNF-α to the E. coli + DNP treatment regimen significantly enhanced the drop in TER compared with that evoked by TNF-α + E. coli or DNP + E. coli, at both 12 and 24 h posttreatment (Fig. 3A).
When E. coli HB101 was added to the apical aspect of confluent T84 cell monolayers, translocation into the basal chamber of the culture well was only observed in 8% of the preparations (Fig. 3B). In contrast, epithelial monolayers treated with either E. coli + TNF-α or E. coli + DNP resulted in bacterial translocation across ∼60% of the monolayers (and at 100% in some experiments). Although treatment with TNF-α + E. coli resulted in translocation scores of 1–3, a score of 5 was not recorded for any of the 12 treated monolayers. In contrast, translocation scores of 4 and 5 were commonplace in epithelial monolayers treated with DNP + E. coli. The increased permeability was exaggerated in monolayers exposed to the triple threat of E. coli + DNP + TNF-α (Fig. 3B).
Epithelial death does not account for the increased permeability.
TNF-α-induced epithelial cytotoxicity could account for the increased epithelial permeability. However, using caspase-3 cleavage as an index of apoptosis, we found no evidence to support this supposition when cells were examined 3, 6, 12 (n = 3, data not shown), or 24 h posttreatment (Fig. 4). In addition, epithelial necrosis was assessed by the Trypan blue dye exclusion technique. The number of epithelial cells retrieved after the various experimental treatments was not different, and there were no significant differences in the percentage positivities, which were 2.8 ± 0.5, 2.7 ± 2.0, 1.5 ± 1.0, and 3.0 ± 0.6% for control, DNP + E. coli, TNF-α + E. coli, and DNP + E. coli + TNF-α, respectively (n = 6 replicates from 2 experiments; means ± SD; one-way ANOVA P = 0.26).
Pharmacological blockade of NF-κβ signaling ameliorates the increased epithelial permeability induced by DNP + E. coli ± TNF-α.
Since many effects of TNF-α are mediated via the transcription factor NF-κβ, we postulated that NF-κβ mobilization contributed to the enhanced reduction in epithelial barrier function. Initial studies confirmed NF-κβ activation in T84 cells in response to TNF-α as revealed by an approximately twofold increase in arbitrary luminescence units 4 h after treatment: control = 2,501 ± 291 vs. TNF-α (10 ng/ml) treatment = 4,282 ± 540 (means ± SE, 6 T84 preparations from 2 experiments). This magnitude of NF-κβ activation is similar to that described by Ye et al. (48). Three inhibitors of NF-κβ activation were employed (i.e., PDTC, NEMO-binding peptide, and BAY 11–7082), as well as the proteosome inhibitor MG-132. None of these pharmacological agents had a direct effect on TER, and they were neither bacteriocidal nor bacteriostatic at the doses used in coculture studies over a 24-h experimental period (data not shown). The use of these inhibitors of NF-κβ activation failed to ameliorate either the drop in TER evoked by DNP + E. coli + TNF-α, or that caused by exposure to DNP + E. coli (Table 2). In contrast, blockade of NF-κβ signaling significantly ameliorated TNF-α + E. coli ± DNP-induced increases in E. coli transcytosis across the epithelium. As shown in Fig. 5, PDTC when added to E. coli + TNF-α, E. coli + DNP, or E. coli + TNF-α + DNP resulted in a significant decrease in the number of bacterial translocation scores of ≥3 assigned to the epithelial monolayers compared with the appropriate positive control group. It should also be noted that, in this series of experiments, exposure to DNP + E. coli resulted in a translocation score of 5 in 0 of 8 epithelial preparations, whereas 3 of 8 monolayers displayed a bacterial translocation score of 5 as a consequence of exposure to DNP + E. coli + TNF-α. Table 3 presents data related to the translocation score of 5 (i.e., the maximum value at which the retrieved and cultured bacteria grow as a lawn). Fifty percent of epithelial monolayers treated with DNP + E. coli + TNF-α showed maximum bacterial translocation, whereas this was reduced to 16% by inclusion of an inhibitor of NF-κβ activity in the culture well. Bacterial translocation across DNP + E. coli-treated T84 monolayers was also significantly reduced by inhibitors of NF-κβ: 70% compared with 11% (Table 3). In additional, PDTC added to the culture well 2 or 4 h, but not 6 h, after E. coli + DNP reduced bacterial translocation (Fig. 6).
Finally, the effects of E. coli, DNP, and TNF-α ± PDTC on TER, the flux of dextran (mw 3,000 Da, considered a marker of paracellular permeability) and the transcytosis of E. coli were simultaneously compared. Consistent with data from other experiments in this study, exposure to E. coli, DNP, and TNF-α resulted in a significant drop in TER and increased E. coli transcytosis; the drop in TER was not affected by treating the cells with PDTC, whereas less bacterial transcytosis was observed (Table 4). Unexpectedly there was not a statistically significant increase in the transepithelial flux of the dextran across E. coli + DNP + TNF-α-treated monolayers, despite a considerable drop in TER. This suggests that, although paracellular permeability has increased, there is still a restriction to the size of molecule that can freely cross the epithelial layer via this pathway. This is an important finding with T84 cells that spontaneously establish monolayers with very high TER values, illustrating that a significant drop in TER (e.g., from 1,500 to 650 Ω/cm2, that is, a 60% drop) does not necessarily translate into a dramatic opening of the tight junction that would allow large inert molecules to pass freely. This maintenance of a paracellular barrier also argues against epithelial apoptosis as a major contributing factor to the reduction in TER.
A number of disorders are associated with loss of mitochondrial function (22). Abnormal mitochondrial structure indicative of reduced ATP synthesis has been noted in epithelial cells in tissues from patients with Crohn's disease, other enteropathies, and in animal models of gut disease (9, 17, 28, 39, 43). A variety of stimuli can interfere with mitochondrial activity resulting in metabolic stress: infection, exposure to bacterial toxins, ischemia, inflammation, psychological stress, certain medications (8, 10, 14, 20), and, potentially, the loss of commensal bacteria that provide short-chain fatty acids as an energy source for colonocytes (21). Postulating that metabolic stress in the enteric epithelium could be a contributing factor to IBD, we found that epithelial monolayers exposed to a classic uncoupler of oxidative stress, DNP, displayed reduced barrier function but only when viable commensal E. coli were present, a pertinent observation in light of studies implicating the commensal microflora in the pathogenesis of IBD (28, 44). However, depending on the source of the metabolic stressor, other cells in the gut, besides the enterocyte, could be affected. Thus, by using a cell-culture model, we assessed the effects of DNP on THP-1 macrophages and the impact of TNF-α, which could originate from macrophages or other cells (e.g., T cells), + DNP + E. coli (HB101) on epithelial barrier function.
Using a model system analogous to that used here, peripheral blood mononuclear cells differentiated into adherent macrophages were found to increase epithelial permeability, particularly when activated with LPS (49). DNP treatment of THP-1 macrophages resulted in significant reduction of ATP production [events mimicked by the unrelated metabolic stressor and inflammatory mediator H2O2 (3)]. Despite this, the metabolically stressed THP-1 cells remained responsive to E. coli and were as effective as non-DNP-treated THP-1 cells in reducing epithelial TER. Similar findings have been reported for monolayers of Caco-2 epithelial cells, but, in contrast to our findings, data were presented in favor of macrophage-derived TNF-α, causing apoptosis in the Caco-2 cells (37).
AAMs participate in tissue restitution and recovery from injury (16). We postulated that DNP-stressed THP-1 cells could take on an AAM-like phenotype and that this would explain the reduced responses to LPS; RT-PCR analysis of markers indicative of AAM failed to support this notion. Also, analysis of TLR-4 mRNA revealed no significant differences between control and DNP-treated THP-1 cells. Thus the reduced cytokine production by DNP-treated macrophages in response to LPS reflects a shift in metabolic activity in the stressed cells. However, in the context of our experimental hypothesis (Fig. 7), it was important to demonstrate that, despite the levels being reduced, the DNP-treated THP-1 cells still produced TNF-α and other proinflammatory molecules (i.e., IL-1β and IL-6).
Since metabolically stressed epithelia are more permeable to bacteria and metabolically stressed THP-1 cells release proinflammatory cytokines, we hypothesized that this would create a positive feedback loop to exaggerate the loss of epithelial barrier function. Similar events in the human colon, where bacteria are omnipresent, could convert short-term, nondangerous increases in epithelial permeability into chronic, dramatic decreases in enteric barrier function that would be of pathophysiological significance.
Considering the immune components of such a feedback loop, we focused on TNF-α because neutralizing anti-TNF-α antibodies reduce the increased barrier function observed in patients with Crohn's disease (45, 51); neutralization of TNF-α in monocyte-epithelial cocultures ameliorates the epithelial barrier defect (49); and TNF-α can directly increase paracellular permeability in monolayers of pulmonary endothelial cells (i.e., CPAE), as well as renal (i.e., MDCK and LLC-PK1) and enteric (i.e., HT-29 and Caco2) epithelial cell lines. These increases in paracellular permeability have been presented as a normal physiological process that does not lead to significant gaps in the monolayer or, alternatively, as a consequence of TNF-α-induced apoptosis (11, 18, 26, 29, 38). Moreover, although TNF-α can directly stimulate T84 cells to produce cytokines (1, 32), we (23) and others (4) have not observed a direct effect of TNF-α on T84 monolayer permeability, unless administered with IFN-γ (46).
Treatment of T84 monolayers with TNF-α + E. coli resulted in reductions in TER and increased bacterial transcytosis, although bacterial translocation scores of 5 were not recorded. This is a significant finding: TNF-α is now having an effect on epithelial barrier function in the presence of nonpathogenic bacteria. The inability of PDTC to block the TNF-α + E. coli-induced drop in TER, while reducing bacterial translocation, suggests that the bacteria may be passing transcellularly across the epithelial cells, a possibility that needs to be tested. In addition, these data complement studies in which biopsies from patients with Crohn's disease showed increases in epithelial endocytosis of the protein horseradish peroxidase after TNF-α exposure (42).
Moreover, we found that the considerable reduction in epithelial barrier function induced by exposure to DNP + E. coli is further enhanced by exposure to TNF-α. Vital dye exclusion and immunoblotting for caspase-3 revealed no evidence of increased necrosis or apoptosis in epithelial monolayers exposed to DNP + E. coli + TNF-α, and, although subtle changes in epithelial viability could contribute to the barrier defect, our data suggest that epithelial cell death is not a major factor in these studies. Next, NF-κβ, a major intracellular signaling pathway mobilized by TNF-α, was assessed as a mediator of the epithelial barrier defect. Use of three pharmacological inhibitors of NF-κβ signaling (which function by different mechanisms) revealed that the increased bacterial translocation, but not the drop in TER, evoked by DNP + E. coli ± TNF-α was due, at least in part, to NF-κβ activity (a possible antioxidant effect of PDTC cannot be unequivocally ruled out). This is not unexpected given the relationship between TNF-α and NF-κβ and the ability of other inhibitors of NF-κβ (e.g., curcumin and NF-κβ anti-sense oligonucleotides) to block TNF-α-induced increases in epithelial permeability (19, 48). What was initially more surprising was the fact that the inhibitors of NF-κβ activity reduced the increased E. coli transcytosis that occurred in response to DNP + E. coli treatment, although we had reported that IL-8 released from T84 cells cultured with E. coli + DNP was blocked by PDTC (28). In addition, many bacterial products, including LPS, mobilize NF-κβ in epithelia (15), and the loss of epithelial barrier function caused by exposure to the oxidant and metabolic stressor, hydrogen peroxide, is lessened by inhibition of NF-κβ activity (3). Moreover, the present study also shows that PDTC added to the culture 4 h after DNP + E. coli still blocks the bacterial transcytosis; this lag phase may represent the time required for the metabolic stress to impact the enterocyte.
Dissecting the mechanism of control of epithelial barrier function, the present study adds to a body of data illustrating that paracellular and transcellular permeability are differentially regulated. Indeed, in our earlier investigations we showed that epithelia under metabolic stress have increased internalization of noninvasive E. coli (28) and that inhibitors of endocytosis and microtubule activity significantly reduced bacterial translocation across T84 cell monolayers (27). We did not specifically address the route of bacterial transcytosis in the present study. Although we cannot unequivocally rule out that some bacteria could move paracellularly, the differential regulation of TER and bacterial transcytosis by inhibitors of NF-κβ is compatible with the bacteria being endocytosed by the stressed ± TNF-α-treated epithelial cells and crossing the monolayer via a transcellular route. Numerous investigations have assessed the regulation of paracellular permeability in health and disease, but significantly less information is available on the control of transcellular permeability, in general, and the ability of components of the commensal flora to cross the epithelium, in particular (6, 24). We suggest that a transcellular route of bacterial entry into the mucosa is likely to represent a major portal of entry that must be given equal, if not greater, consideration when evaluating epithelial barrier function and its contribution to local injury or systemic disease (e.g., sepsis).
In conclusion, we highlight three points. First, metabolic stress in the epithelium should be considered when assessing enteric disease pathogenesis or relapses in disease activity. Second, activation of NF-κβ may represent a common signal mediating increased transcytosis of bacteria across epithelial monolayers by proinflammatory cytokines, bacterial products (either directly or indirectly), and metabolic stress. Thus NF-κβ inhibitors are not only immunosuppressive, but part of their anti-inflammatory property could be the maintenance of epithelial barrier function. Third, in juxtaposing epithelium, bacteria, and immune cells (i.e., macrophages) as three of the prominent players in IBD, we have established an in vitro immunophysiological model system (Fig. 7) that allows testing of specific hypotheses related to the control and regulation of epithelial barrier function under physiological and putative pathophysiological conditions.
This work was funded by an operating grant from the Canadian Institutes of Health Research (MT-13421) to D. McKay, who is an Alberta Heritage Foundation for Medical Research Scientist. D. McKay and P. Sherman are recipients of Canada Research Chairs (Tier 1).
Suzanne Cho is thanked for her help with cytokine ELISA.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 the American Physiological Society