AMP-activated Protein Kinase Mediates the Interferon-γ-induced Decrease in Intestinal Epithelial Barrier Function*

Impaired epithelial barrier function plays a crucial role in the pathogenesis of inflammatory bowel disease. Elevated levels of the pro-inflammatory cytokine, interferon-γ (IFNγ), are believed to be prominently involved in the pathogenesis of Crohn disease. Treatment of T84 intestinal epithelial cells with IFNγ severely impairs their barrier properties measured as transepithelial electrical resistance (TER) or permeability and reduces the expression of tight junction proteins such as occludin and zonula occludens-1 (ZO-1). However, little is known about the signaling events that are involved. The cellular energy sensor, AMP-activated protein kinase (AMPK), is activated in response to cellular stress, as occurs during inflammation. The aim of this study was to investigate a possible role for AMPK in mediating IFNγ-induced effects on the intestinal epithelial barrier. We found that IFNγ activates AMPK by phosphorylation, independent of intracellular energy levels. Inhibition of AMPK prevents, at least in part, the IFNγ-induced decrease in TER. Furthermore, AMPK knockdown prevented the increased epithelial permeability, the decreased TER, and the decrease in occludin and ZO-1 caused by IFNγ treatment of T84 cells. However, AMPK activity alone was not sufficient to cause alterations in epithelial barrier function. These data show a novel role for AMPK, in concert with other signals induced by IFNγ, in mediating reduced epithelial barrier function in a cell model of chronic intestinal inflammation. These findings may implicate AMPK in the pathogenesis of chronic intestinal inflammatory conditions, such as inflammatory bowel disease.

Inflammatory bowel disease (IBD) 2 consists of two major subgroups, ulcerative colitis and Crohn disease (CD). A complex cascade of genetic, immunological, and bacterial factors contributes to IBD pathogenesis (1). In the healthy intestine, the epithelial barrier separates the luminal bacterial microbiota and other aspects of the external environment from cells of the mucosal immune system. In CD in particular, an impaired epithelial barrier (2,3) leads to increased exposure of the immune system to commensal bacteria. Along with possible genetic defects in bacterial sensing, this might contribute to a dysregulated immune response leading to further epithelial damage and active episodes of IBD (4). Epithelial barrier dysfunction in CD is characterized by alterations in intercellular tight junctions (5), as well as by an excessive loss of water and salt into the lumen. An important immunological marker in CD is the existence of excessively high levels of the pro-inflammatory cytokine, interferon gamma (IFN␥) (6).
IFN␥ treatment of intestinal epithelial cell monolayers severely compromises their barrier integrity. Most importantly from a functional perspective, IFN␥ causes a decrease in transepithelial electrical resistance (TER) and increases epithelial permeability (7,8). These defects closely resemble observations in CD, where there is a disruption of intercellular tight junctional complexes. This effect is due to disruption of the apical actin cytoskeleton in conjunction with decreased expression, as well as increased internalization, of important tight junction proteins such as occludin and zonula occludens-1 (ZO-1) (8 -11). Conversely, induction of epithelial apoptosis by IFN␥ is believed to contribute little to barrier dysfunction (12). IFN␥ also induces further alterations in epithelial function that include reduced expression of various ion transporters and associated decreases in epithelial ion transport (13,14). Despite the influence of IFN␥ on a number of epithelial functions, relatively little is known about intracellular signaling mechanisms mediating its effects following receptor activation. Recent studies demonstrated the involvement of phosphatidylinositol 3Ј-kinase (PI3K) in mediating IFN␥-induced effects on epithelial barrier function (11,15). However, this is unlikely to be the only regulatory pathway involved. Indeed, increased expression of receptors for tumor necrosis factor core family members, such as the tumor necrosis factor receptor and LIGHT (homologous to lymphotoxin, shows inducible expression and competes with herpes simplex virus glycoprotein D for herpes virus entry mediator (HVEM), a receptor expressed by T lymphocytes), can also occur in response to IFN␥ and lead to changes in intestinal barrier function (16 -18).
The effects of IFN␥ in intestinal epithelial cells resemble, at least in part, those of the cellular energy sensor, AMP-activated protein kinase (AMPK). Upon activation, AMPK restores intracellular ATP levels by stimulating energy-producing pathways, such as glucose uptake (19) and glycolysis, while inhibiting energy-consuming pathways, such as the synthesis of fatty acids or triglycerides (20,21). In the intestine, energy-consuming processes include epithelial ion transport, and, indeed, AMPK has been shown to decrease intestinal ATP-consuming ion transport as well as the synthesis of various proteins (22,23). Moreover, it has previously been demonstrated that ion transport processes are suppressed in intestinal biopsies from IBD patients (24 -26).
AMPK is usually activated in response to cellular stress that depletes intracellular ATP and elevates the AMP:ATP ratio (27,28). AMPK-activating conditions include oxidative stress (29), hypoxia (30), and hypoglycemia (31). Binding of AMP to AMPK causes an increase in activity of 5-fold or less (32). Further, binding of AMP to AMPK makes AMPK a better substrate for upstream kinase activation, resulting in phosphorylation of the catalytic ␣-subunit of AMPK on the Thr 172 residue and subsequently in a 50-to 100-fold activation of the enzyme (32). A number of upstream kinases for AMPK have been identified, with LKB1 (33,34) or calmodulin kinase II (35)(36)(37) being the most important and well studied. However, recent studies also indicate that PI3K can activate AMPK (38,39).
The goal of this study was to determine whether AMPK mediates IFN␥-induced alterations in intestinal epithelial barrier function. We found that IFN␥ activates AMPK in intestinal epithelial cells and AMPK inhibition prevents, at least in part, IFN␥-induced barrier dysfunction. Our data indicate a novel role for the cellular energy sensor, AMPK, in the regulation of intestinal epithelial barrier properties in a cell model of chronic inflammation. These findings may have implications for barrier function in the setting of chronic inflammatory processes, such as IBD.
Western Blotting-An aliquot of each lysate was mixed with an equal amount of 2ϫ gel loading buffer (50 mM Tris, pH 6.8, 2% SDS, 200 mM dithiothreitol, 40% glycerol, 0.2% bromphenol blue) and boiled for 4 min. Proteins were separated by SDS-PAGE and transferred onto polyvinylidene fluoride membranes (Millipore). Membranes were blocked with 1% blocking solution, and an appropriate concentration of primary antibody was added in 1% blocking buffer over night. Membranes were washed with Tris-buffered saline containing 1% Tween 20 (1% TBST) for 1 h, horseradish peroxidase-labeled secondary antimouse-or anti-rabbit-IgG-antibody (BD Biosciences) in 1% blocking solution (1:2500) was added for 30 min, and membranes were washed for 1 h with 1% TBST. Finally, immunoreactive proteins were detected using an enhanced chemiluminescence detection kit (Amersham Biosciences). Densitometric analysis of Western blots was performed by using Image software (National Institutes of Health).
ATP Assay-T 84 cells were seeded on 0.6-cm 2 semi-permeable filter supports for 8 -10 days. After treatment, cells were washed twice with warm Ringer's solution. Cellular ATP levels were assessed using the CellTiter-Glo Luminescence Cell Viability Assay (Promega) according to the manufacturer's instructions. Briefly, equal volumes of Ringer's and CellTiter-Glo reagent were added apically to lyse the cells. After 10-min shaking, cells were scraped off the filters, and luminescence was read with open range settings on a SpectraMax Fluorescence Microplate reader using the SoftMax Pro v5 software (both from Molecular Devices, Sunnyvale, CA). To normalize cellular ATP levels, the cellular double strand DNA content was analyzed using the Quant-it TM Pico Green dsDNA Assay Kit (Molecular Probes, Eugene, OR) according to the manufacturer's instructions.
TER-T 84 cells were seeded onto either 12-or 30-mm semipermeable filter supports (pore size, 0.45 m) and cultured for ϳ10 days until the TER was at least 900 Ohm.cm 2 . The TER of each monolayer was then assessed before and after treatment using a voltohmeter (WPI, Sarasota, FL) and companion electrodes (Millipore). To normalize the variation in the absolute TER values of each monolayer, the data are expressed as the percentage of pre-treatment TER values (15,42).
Electrophysiological Studies-After treatment, T 84 cell monolayers were mounted in Ussing chambers with a window area of 0.6 cm 2 and bathed in oxygenated (95% O 2 , 5% CO 2 ) Ringer's solution at 37°C. Using short-circuit current (I sc ) the monolayers were continuously voltage-clamped to zero potential difference. Cells were allowed to equilibrate for 10 min before baseline I sc and conductance were assessed. Changes in I sc (⌬I sc ) under these conditions are exclusively due to changes in electrogenic chloride secretion (43) and were measured for a period of 20 min after stimulation with carbachol (100 m, basolaterally). Conductance was calculated according to Ohm's law as mS/cm 2 .
Small Interfering RNA Transfection-2 ϫ 10 6 T 84 cells were seeded 3 days before transfection and grown to 50 -70% confluency in T75 flasks. Three different annealed Silencer predesigned siRNA oligonucleotides targeting AMPK ␣1 were obtained from Applied Biosystems (Foster City, CA). For transfection reactions, 100 pmol of each of the three gene-specific siRNA oligonucleotides were transfected into T 84 cells using the Amaxa nucleofector system (Amaxa Inc., Gaithersburg, MD) according to the manufacturer's instructions. After transfection, T 84 cells were cultured on filter membranes for 48 h before further treatment. A nonspecific control siRNA SMARTpool (100 pmol, Upstate Biotechnology/Dharmacon, Chicago, IL) was used as a negative control.
Transepithelial Permeability-Transepithelial permeability was assessed by measuring the flux of fluorescein isothiocyanate-dextran (FITC-dextran, 10 kDa, Sigma) across T 84 cell monolayers. Following treatment, T 84 cells were washed (3ϫ) with Ringer's solution and incubated in Ringer's solution for 30 min at 37°C to equilibrate. FITC-dextran (1 mg/ml) was added to the apical side of the monolayer. 1 h later, 100 l of the basolateral solution was removed, and fluorescence of FITCdextran in this compartment was detected using excitation at 490 nm and emission at 520 nm.
Confocal Microscopy-T 84 cells were seeded onto 12-mm Millicell-HA filters for 2 days after siRNA transfection and before stimulation. After treatment, cells were washed twice with PBS, fixed with 10% Formalin in PBS for 10 min at room temperature, washed (3ϫ) with PBS, and permeabilized with 0.3% Triton X-100 in PBS for 10 min. After three washes with PBS, cells were blocked with 20% donkey serum (Sigma) in PBS for 1 h and washed with PBS once. Mouse anti-ZO-1 antibody (1 g/ml) in 20 mg/ml bovine serum albumin in PBS was applied overnight. Cells were washed (3ϫ) with PBS and secondary Alexa-488-conjugated donkey anti-mouse antibody (excitation/emission maxima at 495/519 nm, Molecular Probes) was added in a 1:500 dilution in 20 mg/ml bovine serum albumin in PBS for 30 min at room temperature. Cells were washed (3ϫ) with PBS, incubated with Hoechst 33258 (excitation/emission maxima at 352/461 nm, Molecular Probes) in PBS (1:500) for 20 min at room temperature, and washed four times with PBS. Finally, the cells on the filter membrane were transferred onto glass slides and mounted with ProLong Gold Antifade Reagent (Molecular Probes). Confocal microscopy was performed using a Zeiss LSM 510 Laser Scanning Confocal system on a Zeiss Axioscope 2 Upright Microscope (Zeiss, Jena, Germany). Data were analyzed using the Zeiss LSM 5 Image Examiner software.
Statistical Analysis-Data are presented as means Ϯ S.E. for a series of n experiments. Data are expressed as raw data, arbitrary units or as a percentage of the respective control. Statistical analysis was performed by analysis of variance followed by Student-Newman-Keuls post hoc test unless otherwise noted. p values of Ͻ0.05 were considered significant.

IFN␥ Activates AMPK in Human T 84 Intestinal Epithelial
Cells-To study whether the cellular energy sensor, AMPK, participates in IFN␥-induced effects on epithelial barrier function, we initially investigated whether IFN␥ activates AMPK. Polarized monolayers of T 84 cells were treated with IFN␥ at a concentration of 1000 units/ml for 0.5-72 h. At this concentration, IFN␥ has been shown to exert significant effects on intestinal epithelial cell barrier properties (7,44,45). Furthermore, in a previous study by Sasaki et al. (46), cultured biopsies from patients with active CD were shown to release IFN␥ at levels as high as ϳ1000 units/ml/mg of tissue, although the mean level was 75 Ϯ 215 units/ml/mg of tissue. So, although the majority of our data were generated using 1000 units/ml (100 ng/ml) of IFN␥ (in contrast to a number of other studies that used 10 or 20 ng/ml IFN␥), the experimental use of this concentration is not unreasonable. IFN␥ treatment for 6 h caused maximal activation of AMPK in human T 84 intestinal epithelial cells as assessed by phosphorylation of Thr 172 of the AMPK catalytic ␣1-subunit, which is an established marker of AMPK activation (47). Representative Western blots showed that IFN␥ stimulated AMPK activation as early as 30 min after treatment; thereafter phosphorylation increased in a time-dependent manner reaching a peak at 6 h ( Fig. 1A). Thereafter AMPK phosphorylation declined in a time-dependent manner almost to control levels by 48 and 72 h of IFN␥ treatment (Fig. 1B). To assess concentration-dependent effects of IFN␥ on the phosphorylation of AMPK, we administered 10, 100, and 1000 units/ml IFN␥ for 6 h, the time point showing a maximal IFN␥ effect on AMPK phosphorylation. As shown in Fig. 1C, a concentration of 1000 units/ml IFN␥ caused the strongest increase in AMPK phosphorylation in our cell model, whereas reduced effects were seen with of 10 and 100 units/ml IFN␥. Furthermore, the pharmacological AMPK inhibitor, Compound C (CC, 50 M), which competes with intracellular ATP for binding to AMPK (48), prevented the IFN␥-induced activation of AMPK (Fig.  1D). In a separate and as yet unpublished study, we generated a concentration curve for the inhibitory effect of CC on AMPK activation and observed peak inhibition with a dose of 50 M CC (data not shown). Therefore, we used this inhibitor concentration for subsequent experiments. Of note, 50 M CC is at the lower end of the standard range for AMPK inhibition in epithe- IFN␥-induced AMPK Activation Is Independent of Changes in Intracellular ATP-Because AMPK is usually activated by declining levels of intracellular ATP, we next investigated if AMPK activation in response to IFN␥ was dependent on the intracellular ATP level. T 84 cells were treated with IFN␥ for 0.5-72 h. Treatment with H 2 O 2 (100 M; 30 min, bilaterally) was used as a positive control to deplete ATP (51). Interestingly, IFN␥ treatment had no significant effect on intracellular ATP at any time point. Although the cellular ATP level declined somewhat within 3 h of IFN␥ treatment, this effect was not significant and therefore unlikely to explain IFN␥-induced AMPK activation, which occurred as early as 30 min after addition of IFN␥. Additionally, long term treatment with IFN␥ for up to 72 h did not significantly alter the cellular ATP level, correlating with a lack of AMPK phosphorylation at these time points. In contrast, H 2 O 2 significantly reduced ATP levels within 30 min (Fig. 2). These data indicate that IFN␥ activates AMPK independent of changes in intracellular ATP.
CC Ameliorates the IFN␥-induced Decrease in TER across T 84 Monolayers-Long term treatment with IFN␥ decreases TER across T 84 cell monolayers. Because AMPK is activated by IFN␥, our next goal was to investigate whether AMPK inhibition influences IFN␥-induced effects on epithelial barrier function. First, we determined the time and concentration dependence of the IFN␥-induced alterations in TER across T 84 monolayers. Therefore we administered 10, 100, or 1000 units/ml IFN␥ for 24, 48, or 72 h. In keeping with previous studies (7), IFN␥ did not affect TER within 24 h of treatment, but decreased TER significantly by 48 h and even more so by 72 h (Fig. 3A). Interestingly, whereas the IFN␥ effect was shown to be clearly time-dependent, different doses of IFN␥ had similar effects on TER (Fig. 3A). Based on these observations, we chose the 72-h time point and a concentration of 1000 units/ml IFN␥ for further analysis, and treated T 84 monolayers with IFN␥ in the presence or absence of the AMPK inhibitor, CC (50 M, bilaterally) for 72 h. TER was assessed by a voltohmeter, and the findings were confirmed by measuring the conductance across monolayers mounted in Ussing chambers. IFN␥ treatment for 72 h caused a significant drop in TER, whereas CC had no effect on TER by itself (Fig. 3B). However, CC partially, and significantly, reversed the IFN␥-induced decrease in TER (Fig.  3B) and increase in conductance (Fig. 3C). However, AMPK inhibition did not affect the IFN␥-induced suppression of another parameter of colonic epithelial cell function, Ca 2ϩ -dependent chloride secretion, as stimulated by the muscarinic receptor agonist, carbachol (100 M, basolaterally) (Fig. 3D). These data indicate that AMPK activity is, at least in part, specifically required for the effects of IFN␥ on epithelial barrier function in T 84 cells.  OCTOBER 9, 2009 • VOLUME 284 • NUMBER 41

AMPK Knockdown Diminishes the IFN␥-induced Increase in
Epithelial Permeability-Mindful of the possible limitations of pharmacological inhibition of AMPK with respect to specificity, we also sought a molecular approach to confirm a role for AMPK in the ability of IFN␥ to reduce epithelial barrier function. T 84 cells were therefore transfected with either siRNA oligonucleotides targeting AMPK or with nonspecific siRNA sequences as a control. 48 h later, cells were treated for a further 72 h with IFN␥. As shown in Fig. 4A, AMPK-specific siRNA clearly decreased the expression of AMPK protein. Additionally, no nonspecific effects on cellular protein expression could be observed, as shown by equivalent levels of the nuclear envelope protein, lamin A/C, which was used as a loading control. To examine the functional correlate of these results, we first assessed the flux of FITC-labeled dextran (10 kDa) across T 84 monolayers. In control siRNA-transfected cells, IFN␥ significantly increased permeability to FITC-dextran, an observation that is in good accordance with previous findings by Watson et al. (8). However, knockdown of AMPK ameliorated this effect, at least in part. Although tracer flux was still increased in IFN␥treated, AMPK-deficient cells compared with cells not treated with IFN␥, it was significantly lower than the flux across IFN␥treated cells transfected with the nonspecific control siRNA (Fig. 4B). Furthermore, we also assessed if AMPK knockdown affected the IFN␥-induced decrease in TER across T 84 monolayers. Although IFN␥ treatment caused a significant decrease in TER in cells transfected with control siRNA, this effect was partially prevented in AMPK-deficient cells. Although IFN␥ still caused a detectable reduction in the TER across AMPKdeficient cells, this effect was significantly lower than in cells transfected with nonspecific siRNA (Fig. 4C). Therefore, these data confirm a role for AMPK in IFN␥-induced alterations in the barrier properties of the intestinal epithelium, and extend this not only to TER, but also to macromolecular permeability.
AMPK Mediates IFN␥-induced Down-regulation of the Tight Junction Proteins Occludin and ZO-1-A well-described effect of IFN␥ that correlates with reduced epithelial barrier integrity is the down-regulation of the tight junction protein, occludin (8,9,11). Having shown that AMPK inhibition reverses the effects of IFN␥ on epithelial barrier function, we investigated a possible effect on tight junction proteins that contributes to the epithelial barrier. To assess the involvement of AMPK, we incubated T 84 cells with IFN␥ (1000 units/ml) and/or Compound C (50 M) for 72 h. In IFN␥-treated cells, occludin expression was significantly reduced, whereas Compound C treatment alone had no significant effect (Fig. 5). However, co-treatment with Compound C prevented the ability of IFN␥ to reduce occludin levels (Fig. 5).
To confirm and extend these findings, we assessed the expression of various tight junction proteins in IFN␥-treated cells that were also rendered AMPK-deficient. Occludin, ZO-1, and claudin-4 play important roles in maintaining epithelial barrier function, whereas increased levels of the pore-forming claudin-2 are associated with increased epithelial permeability (52). As already shown in Fig. 4A, AMPK-specific siRNA reduced AMPK protein expression, whereas the levels of lamin A/C were unaffected in either control siRNA or AMPK siRNAtransfected T 84 cells. Treatment with IFN␥ for 72 h caused a significant reduction in the protein expression of occludin and ZO-1 in cells transfected with nonspecific siRNA. However, knockdown of AMPK largely abrogated the effect of IFN␥ on occludin (Fig. 6A) or ZO-1 (Fig. 6B) expression. In contrast, neither the pore-forming protein, claudin-2, nor claudin-4 was affected by either IFN␥ or the knockdown of AMPK or the combination of these treatments (Fig. 6C). These data indicate that AMPK activity is, at least in part, required to mediate the effect of IFN␥ on the expression of occludin and ZO-1 in intestinal epithelial cells, but not of the two claudins examined.
Loss of AMPK Does Not Affect IFN␥-induced Internalization of the Tight Junction Protein ZO-1-IFN␥ has been demonstrated to cause internalization of various tight junction proteins (9,12). We therefore set out to investigate the involvement of AMPK in the regulation of IFN␥-induced internalization of ZO-1. We transfected T 84 cells with either nonspecific control or AMPK-specific siRNA and stimulated these cells for 72 h with 1000 units/ml IFN␥. As shown in Fig. 7, ZO-1 appeared in unstimulated control siRNA (Fig. 7, A and AЈ) as well as in AMPK siRNA-transfected cells (Fig. 7, B plus BЈ) to be mainly localized within the cellular membrane. In contrast, and in good agreement with our protein data (cf. Fig.  6), in IFN␥-treated control-siRNA cells, ZO-1 was less prominent. Additionally, ZO-1 was more prominent within the intracellular compartment of these cells than in the cell membrane, even showing aggregated, intracellular protein complexes and a heavily disturbed pattern (Fig. 7, C and CЈ). Similarly, in IFN␥treated AMPK-deficient cells, ZO-1 also appeared to be more prominent within the cytoplasmic cell compartment and removed from the cell membrane (Fig. 7, D and DЈ). Nevertheless, correlating with our protein data, the abundance of ZO-1 in IFN␥-treated AMPK knockdown cells was apparently greater than in IFN␥-treated control siRNA-transfected cells (Fig. 7, C, CЈ, D, and DЈ). The impression that the overall shape of the cells subjected to AMPK knockdown seems to be altered in comparison to AMPK-competent cells might be due to altered tight junction assembly in AMPK-deficient cells (50, 53,

54).
These data indicate disruption of the cell membrane and ZO-1 localization with possible internalization, in addition to decreased overall expression of ZO-1 as indicated by Western blotting (cf. Fig. 6) in response to IFN␥. Furthermore, they suggest that AMPK is likely not involved in regulating the internalization of ZO-1, at least in our T 84 cell model.
PI3K Inhibition Partially Prevents IFN␥-induced Activation of AMPK-Having shown that IFN␥ exerts its effects on intestinal epithelial barrier function, at least in part, via AMPK, we next sought to explain the connection between IFN␥ and AMPK. Recent studies showed that the PI3K inhibitor, LY294002, diminished some aspects of IFN␥-induced effects on the epithelial barrier (11,15). Therefore, we investigated if PI3K inhibition could affect IFN␥-induced activation of AMPK. We first confirmed that LY294002 inhibits the activity of PI3K by assessing the phosphorylation of Akt, a well described marker for PI3K activity (55,56). IFN␥ alone did not significantly affect Akt phosphorylation, thus mirroring previous findings of constitutive Akt phosphorylation in T 84 cells (15). However, incubation of T 84 cells with LY294002, as well as the combination of IFN␥ and LY294002, resulted in a dramatic decrease in Akt phosphorylation (Fig. 8A). We next investigated whether PI3K inhibition affected IFN␥-induced increases in AMPK phosphorylation. Treatment with LY294002 had no significant effect on AMPK phosphorylation by itself, but completely prevented the IFN␥-induced increase in AMPK phosphorylation (Fig. 8B). Interestingly, and in contrast to AMPK inhibition by Compound C, PI3K inhibition did not reduce AMPK phosphorylation below control levels (compare Figs. 1B and 8B). These data indicate that IFN␥ likely requires activation of PI3K signaling to facilitate AMPK activation.
PI3K Inhibition Ameliorates the IFN␥-induced Decrease in TER across T 84 Monolayers-Having shown that PI3K inhibition prevented the IFN␥-induced rise in AMPK phosphorylation, we sought to correlate this finding with IFN␥-induced effects on intestinal epithelial barrier function. McKay et al. studied the effects of pharmacological PI3K inhibition on the response to 20 ng/ml IFN␥ for a 48-h treatment period (15),  5). B, AMPK inhibition by Compound C partially prevented the IFN␥-induced decrease in TER following co-treatment for 72 h (n ϭ 5). Data are expressed as a percentage of the control value. C, the IFN␥-induced increase in conductance across T 84 monolayers was partially diminished by Compound C (n ϭ 4). D, AMPK inhibition by Compound C was unable to restore IFN␥-induced inhibition of chloride secretion in response to the muscarinic receptor agonist, carbachol (100 M; n ϭ 4). Asterisks indicate significant differences versus the respective control (*, p Ͻ 0.05; **, p Ͻ 0.01; and ***, p Ͻ 0.001). # , indicates p Ͻ 0.05; ### , indicates p Ͻ 0.001 versus 72 h IFN␥ treatment of T 84 cells. ϩ , indicates p Ͻ 0.001 versus 24-h IFN␥ treatment of T 84 cells.
whereas Boivin et al. tested the role of PI3K activity in the effects of a lower concentration of IFN␥ (10 ng/ml) also with a 48-h treatment (11). For purposes of comparison with our previous experiments, we investigated if PI3K inhibition could also modulate effects of IFN␥ produced by a concentration of 1000 units/ml (100 ng/ml) for 72 h. As expected, IFN␥ alone significantly decreased TER across T 84 monolayers, whereas treatment with LY294002 (20 M) alone had no effect. PI3K inhibition by LY294002 did, however, significantly attenuate the IFN␥-induced decrease in TER (Fig. 8C). However, this effect was, in keeping with data on AMPK phosphorylation, clearly less pronounced than the effect of Compound C on AMPK activity and was also lower than the effect of LY294002 on IFN␥-induced TER alterations observed by Boivin et al. (11) and McKay et al. (15). Nevertheless, these findings confirm our biochemical data, indicating that AMPK activation by IFN␥ in T 84 cells occurs in the context of PI3K activation. To further define the ability of LY294002 to ameliorate the IFN␥-induced decrease in TER, we also performed time-and dose-response experiments using T 84 cells treated with either IFN␥, or LY294002, or a combination of both agents. Because our previous data have demonstrated that the IFN␥-induced effect on the TER is independent of the actual concentration of the cytokine, but critically dependent on the duration of cytokine treatment (cf. Fig. 3A), we treated T 84 cells for 48 or 72 h with 100 or 1000 units/ml (10 or 100 ng/ml) IFN␥ and/or LY294002. As shown in Fig. 8 (D and E), and similar to Fig. 8 (C and F), PI3K inhibition had a small, but consistently significant effect in ameliorating the IFN␥-induced decrease in TER across T 84 monolayers by 48-h treatment independent of the IFN␥ dose used (100 or 1000 units/ml), as well as by 72-h treatment (100 units/ml, data not shown, or 1000 units/ml). These data demonstrate that inhibition of PI3K is able to partially, but significantly, diminish the IFN␥-induced decrease in TER across T 84 cell monolayers, independent of the concentration or treatment duration of IFN␥. Furthermore, they suggest that additional factors, besides PI3K, are likely required for IFN␥-induced activation of AMPK, as well as altered barrier function, under these experimental conditions.

DISCUSSION
It is well described that the pro-inflammatory cytokine, IFN␥, compromises the intestinal epithelial barrier by causing internalization and reduced expression of tight junction proteins, resulting in increased transepithelial permeability and decreased TER. IFN␥ also produces inhibition of epithelial ion  transport (7)(8)(9)(10)(11)13). However, the intracellular signaling pathways that mediate the various effects of IFN␥ are poorly understood. Recent studies have shown an involvement of PI3K in mediating the effects of IFN␥ on epithelial barrier properties (11,15). In contrast, a major role for the most established mediator of IFN␥ signaling, the signal transducer and activator of transcription 1, in epithelial barrier defects has not been confirmed (15,39,57).
Here, we show that IFN␥ activates the cellular energy sensor, AMPK, in intestinal epithelial cells and that inhibition of AMPK diminishes the detrimental effect of this inflammatory cytokine on epithelial barrier function. Previously, AMPK was thought to be activated predominantly in response to decreased levels of intracellular ATP (24). For example, the production of reactive oxygen species, such as hydrogen peroxide, during intestinal inflammation can lead to oxidative stress, reduced ATP levels, and consequent cellular dysfunction or damage (58). H 2 O 2 can also promote activation of AMPK (26). However, IFN␥ alone, which plays a crucial role in the pathogenesis of IBD, had not previously been reported to activate AMPK in vitro or in vivo. Additionally, our data suggest that IFN␥ acti-vates AMPK independent of changes in the levels of intracellular ATP.
Nevertheless, a recent study from Riboulet-Chavey et al. using pancreatic ␤-cells demonstrated that AMPK phosphorylation in glucose-pretreated cells was increased 48 h after administration of a cytokine mix consisting of tumor necrosis factor, interleukin-1␤, and IFN␥ (59). Furthermore, administration of the cytokine mix for 48 h caused a significant decrease in the cellular ATP level in glucose-pretreated cells. However, this study did not investigate the effects of IFN␥ alone on either the phosphorylation of AMPK or the cellular ATP level. Because we detected neither an increase in AMPK phosphorylation with 48-or 72-h IFN␥ treatment nor significant alterations in the cellular ATP level at these respective time points, the discrepancy between these observations could be due either to the different cell types used in the respective studies (intestinal epithelial cells versus pancreatic ␤-cells), to the glucose pretreatment, or to the additional cytokines used (tumor necrosis factor and/or interleukin-1␤).
Our studies suggest that AMPK-mediated alterations in barrier function and composition are dependent on the signaling context in which AMPK is activated, with IFN␥ altering the cytosolic milieu to conditions that reveal a role for AMPK in altering barrier function. Because the intestinal epithelium is subjected to small changes in intracellular ATP very frequently, we can speculate that it is an appropriate adaptation that AMPK activity alone would be insufficient to impair intestinal epithelial barrier function. From a physiological perspective, this makes sense as it would be extremely undesirable for every stimulus of AMPK, i.e. any event that causes a decrease in ATP levels, to result in barrier dysfunction in a tissue that is so frequently subjected to hypoxia. Interestingly, AMPK inhibition could not restore impaired carbachol-induced chloride secretion seen in cells treated with IFN␥, indicating that the cytokine exerts this latter effect independent of AMPK.
These findings suggest a novel role for AMPK as not only a cellular energy sensor, but also an important signal transducer for pro-inflammatory cytokines such as IFN␥. This finding may also have implications for chronic inflammatory conditions, such as IBD, which exhibit high levels of inflammatory cytokines. The exact molecular mechanism responsible for AMPK activation by IFN␥ will be the subject of future studies.
We demonstrated that inhibition of AMPK attenuated IFN␥-induced barrier dysfunction. AMPK inhibition prevented both the decline in TER and the increase in conductance across T 84 monolayers by ϳ50%. Using an AMPK knockdown approach, we also uncovered a role for AMPK in the regulation of IFN␥-induced epithelial permeability to macromolecules, such as 10-kDa FITC-dextran. The latter finding is of particular interest, because increased permeability to macromolecules, specifically bacterial components or products, is believed to contribute significantly to the pathogenesis of CD. We also investigated one or more possible mechanisms by which AMPK mediates IFN␥-induced changes in epithelial barrier function. A well established effect of chronic IFN␥ treatment is to alter the expression of tight junction proteins, such as occludin and ZO-1 (9 -11, 60). AMPK inhibition completely prevented IFN␥-induced down-regulation of occludin and partially restricted the reduction in ZO-1 expression. Thus we have identified a new signaling component mediating the effect of IFN␥ on the expression of tight junction proteins. The almost complete restoration of ZO-1 and occludin by AMPK knockdown was sufficient to partially ameliorate the IFN␥-induced decrease in TER and the increases in conductance and permeability. Our observed effects of amelioration of IFN␥-induced effects on occludin and permeability to 10-kDa FITC-dextran by AMPK knockdown, correlate with studies by Watson et al. (8) who demonstrated that IFN␥ increased epithelial permeability via increased frequency of a large macromolecule (10 kDa) pore, and this was associated with changes in occludin expression and phosphorylation. Interestingly, IFN␥ did not affect the expression of the pore-forming claudin-2 or of claudin-4, and, subsequently, loss of AMPK had no effect on their expression, either. Because AMPK inhibition or knockdown were only partially able to mitigate IFN␥-induced changes in epithelial barrier function, we speculate that there may also be additional AMPK-independent modifications of tight junction proteins, either through post-translational events or other transcriptional effects of IFN␥. Additionally, the effect of pharmacological AMPK inhibition by CC was greater than the effect of AMPK knockdown as regards amelioration of the IFN␥-induced decrease in TER. This finding correlates with the greater inhibition of AMPK phosphorylation, even below control levels, in IFN␥ and CC co-treated cells. Although these data support our evidence of a pre-existing "basal" level of AMPK phosphorylation in these cells, one might hypothesize, that CC, which has already been shown to have off-target effects in nonepithelial cell lines (61)(62)(63), or possibly to even cause AMPK knockdown itself, may also have AMPK-independent effects, that could contribute to an enhanced protective effect on epithelial barrier function versus AMPK knockdown. However, as stated above, the concentration of CC chosen for this study was based on its functional efficacy and was actually lower than used in similar studies of AMPK function in epithelial cell lines, including T 84 cells.
Interestingly, AMPK knockdown did not affect IFN␥-induced internalization of the tight junction protein ZO-1. A possible candidate for IFN␥-induced regulation of tight junction protein localization is myosin light chain kinase, which can also be activated in response to IFN␥ treatment and plays an important role in tight junction protein internalization as established by Turner and colleagues (64,65). However, a previous report using smooth muscle cells indicates that myosin light chain kinase is a substrate of, and is inactivated by, AMPK (66). Whether AMPK regulates myosin light chain kinase function in epithelial cells in response to IFN␥ or other stimuli has not been identified, but this avenue of investigation will likely yield findings of general interest in the field of tight junction regulation.
Peak AMPK activation in response to IFN␥ occurred at 6 h. This is in keeping with the observations of McKay et al. and Boivin et al. regarding the role of PI3K in IFN␥-induced barrier defects (11,15). In the McKay study, PI3K inhibition could only prevent IFN␥-induced effects on TER when LY294002 was administered within the first 6 h after IFN␥ treatment, whereas the Boivin study showed that IFN␥-induced PI3K activation occurs within 10 min after IFN␥ exposure. In contrast, a detectable effect of IFN␥ on epithelial barrier function did not occur until 48 h after treatment in both studies. Our data confirmed the involvement of PI3K in the signaling events that underlie IFN␥-induced barrier dysfunction. Interestingly, Akt, a common downstream target of PI3K signaling, is likely not involved based on the lack of an effect of IFN␥ on Akt phosphorylation (cf. Fig. 8A) (15). However, even though PI3K inhibition reduced IFN␥-stimulated AMPK phosphorylation to control The histogram shows the densitometric analysis of nine similar experiments. C, PI3K inhibition by LY294002 partially ameliorated the IFN␥-induced decrease in TER following co-treatment for 72 h. Data are expressed as percentage of control (n ϭ 9). D, co-treatment with the PI3K inhibitor, LY294002, restricted the IFN␥-induced (100 units/ml; 48 h) decrease in TER (n ϭ 6). Data are presented in Ohm/cm 2 . E, LY294002 diminished the IFN␥-induced (1000 units/ml) decrease in TER after co-treatment for 48 h (n ϭ 6). Data are presented in Ohm/cm 2 . F, PI3K inhibition by LY294002 ameliorated the IFN␥-induced (1000 units/ml) decrease in TER by co-treatment for 72 h (n ϭ 6). Data are presented in Ohm/cm 2 . Statistical analysis in D-F was performed by Student t test. Asterisks indicate significant differences versus the respective control (*, p Ͻ 0.05; ***, p Ͻ 0.001). ϩ , p Ͻ 0.05 versus 6 h IFN␥ treatment of T 84 cells. # , p Ͻ 0.05; ### , p Ͻ 0.001 versus treatment of T 84 cells with IFN␥ for 72 h. § , p Ͻ 0.05 versus treatment with IFN␥ for 48 h. levels, LY294002 only partially diminished the IFN␥-induced decrease in TER, and to a far lesser extent than specific AMPK inhibition by Compound C, or to levels observed in the studies of McKay et al. (15) and Boivin et al. (11). To investigate the mechanism for the different findings in our study compared with previous work (11,15), we also administered IFN␥ for 48 h, as the previous data were generated using this time point. However, in our cell model we did not detect an increased effect of PI3K inhibition on the IFN␥-induced effects on the TER at the different time points (48 versus 72 h) or with different concentrations of IFN␥ (100 versus 1000 units/ml). Therefore, our data suggest that additional mediators, supplementary to PI3K, may be necessary to display the full effect of IFN␥ on AMPK activity, and subsequently to diminish intestinal epithelial barrier function. The identification of these additional factors was beyond the scope of the present study. Nevertheless, we have identified a central role for AMPK as a mediator of IFN␥-induced signaling and downstream pathophysiological events.
In conclusion, our data indicate an important role for the cellular energy sensor, AMPK, in the regulation of IFN␥-induced changes in intestinal epithelial barrier function, independent of cellular energy status. In addition, besides its well established regulatory effects in energy homeostasis and cell survival, AMPK participates in regulating the integrity of the epithelial barrier. In a pathophysiological setting, these findings suggest a novel role for AMPK in mediating the consequences of chronic intestinal inflammation, and thus may have implications for chronic inflammatory processes in the intestine, such as CD.