ER-stress mobilization of death-associated protein kinase-1–dependent xenophagy counteracts mitochondria stress–induced epithelial barrier dysfunction

The gut microbiome contributes to inflammatory bowel disease (IBD), in which bacteria can be present within the epithelium. Epithelial barrier function is decreased in IBD, and dysfunctional epithelial mitochondria and endoplasmic reticulum (ER) stress have been individually associated with IBD. We therefore hypothesized that the combination of ER and mitochondrial stresses significantly disrupt epithelial barrier function. Here, we treated human colonic biopsies, epithelial colonoids, and epithelial cells with an uncoupler of oxidative phosphorylation, dinitrophenol (DNP), with or without the ER stressor tunicamycin and assessed epithelial barrier function by monitoring internalization and translocation of commensal bacteria. We also examined barrier function and colitis in mice exposed to dextran sodium sulfate (DSS) or DNP and co-treated with DAPK6, an inhibitor of death-associated protein kinase 1 (DAPK1). Contrary to our hypothesis, induction of ER stress (i.e. the unfolded protein response) protected against decreased barrier function caused by the disruption of mitochondrial function. ER stress did not prevent DNP-driven uptake of bacteria; rather, specific mobilization of the ATF6 arm of ER stress and recruitment of DAPK1 resulted in enhanced autophagic killing (xenophagy) of bacteria. Of note, epithelia with a Crohn's disease–susceptibility mutation in the autophagy gene ATG16L1 exhibited less xenophagy. Systemic delivery of the DAPK1 inhibitor DAPK6 increased bacterial translocation in DSS- or DNP-treated mice. We conclude that promoting ER stress–ATF6–DAPK1 signaling in transporting enterocytes counters the transcellular passage of bacteria evoked by dysfunctional mitochondria, thereby reducing the potential for metabolic stress to reactivate or perpetuate inflammation.


DAPK1 inhibitor DAPK6 increased bacterial translocation in
DSS-or DNP-treated mice. We conclude that promoting ER stress-ATF6 -DAPK1 signaling in transporting enterocytes counters the transcellular passage of bacteria evoked by dysfunctional mitochondria, thereby reducing the potential for metabolic stress to reactivate or perpetuate inflammation.
Bacteria within the epithelium are a pro-inflammatory threat. While observed in inflammatory bowel disease (IBD) 3 (1-4), and putatively important in disease pathophysiology, the mechanism(s) underlying this barrier defect is unclear. Mutation in the organic cation transporter (OCTN2) gene is a susceptibility trait for IBD (5); this transporter facilitates carnitine uptake for fatty acid ␤-oxidation and ATP generation. Moreover, deranged mitochondrial ultrastructure, loss of mitochondrial proteins, lower levels of ATP, and increased susceptibility to uncouplers of oxidative phosphorylation occur in some patients with IBD (6). Similarly, suspected triggers of IBD reactivation can perturb mitochondrial function. For instance, nonsteroidal anti-inflammatory drugs (NSAID) can cause relapses in IBD (7), and their use damages mitochondria contributing to decreased enteric barrier function (8,9). Consequently, mitochondria are emerging as an organelle of interest in IBD and other enteropathies (10).
While idiopathic, the consensus is that uncontrolled immune reactivity against a component(s) of the gut microbiota in a genetically susceptible individual causes IBD (11). A defect in epithelial barrier function is implied in this hypothesis. With the exceptions of uptake by microfold cells or dendritic cell pseudopodia, material in the gut lumen enters the submucosa via the epithelial tight junctions (paracellular) or by crossing the enterocyte (transcellular) (12), and regulation of both pathways is energy-dependent. With that, it can be proposed that perturbed epithelial mitochondrial function would increase gut permeability to initiate disease or to precipitate relapses in IBD. Accordingly, mitochondrial damage induced by uncoupling oxidative phosphorylation with the H ϩ ionophore, dinitrophenol (DNP), to model the mitochondrial dysfunction observed in IBD resulted in increased internalization into and transcytosis of Escherichia coli across epithelial monolayers (13)(14)(15).
Mitochondria couple with the endoplasmic reticulum (ER), where reactive oxygen species (ROS) are important in protein folding (16). When improperly folded proteins accumulate in the ER, an ER stress program (or unfolded protein response (UPR)) occurs. Mice deficient in any of the three major arms of the ER stress response (inositol-requiring kinase/endonuclease 1 (IRE1), pancreatic ER kinase (PERK), and activating transcription factor 6 (ATF6)) are more susceptible to DSS-induced colitis (17,18). Furthermore, mutation in the X-box-binding protein-1 (XBP1) gene (downstream of IRE1) is a susceptibility trait for Crohn's disease (17), and increased ER stress can occur in the epithelium of patients with IBD (17,19,20). We hypothesized that uncoupling oxidative phosphorylation (i.e. mitochondrial dysfunction) combined with ER stress would severely compromise epithelial integrity.

ER stress protects epithelium transcellular permeability perturbed by mitochondrial stress
Human colonic biopsies were mounted in Ussing chambers; bacteria were added to the luminal buffer, and biopsies were treated with DNP Ϯ the ER stressor tunicamycin (blocks protein glycosylation (21), which is often used to elicit ER stress in cell lines (19)). DNP promoted E. coli uptake; however, tunicamycin reduced this barrier defect (Fig. 1A). DNP treatment of colonic enteroid monolayers resulted in increased transcytosis of E. coli that was reduced by tunicamycin (Fig. 1B). Similarly, DNP-evoked increases in bacterial internalization into and translocation across the T84 (Fig. 1, C and D), Caco-2 (Fig. S1A), and HCT116 (Fig. S1B) human epithelial cell lines was significantly reduced by tunicamycin co-treatment. (Tunicamycin induction of ER stress was confirmed by GRP78 expression (Fig. S1C). Tunicamycin did not increase intracellular E. coli (Fig. S1D).) The NSAID indomethacin increased intracellularly viable E. coli in T84 cells (9), and this was reduced by tunicamycin (Fig. 1E). (HeLa cells provided similar data (Fig.  S2).) Treatment with other ER stressors, brefeldin A (blocks Golgi transit) or thapsigargin (Ca 2ϩ -ATPase blocker that perturbs ER Ca 2ϩ stores) (22), also reduced the number of viable intracellular bacteria after exposure to DNP (Fig. 1F). Epithelia treated with DNP display decreased TER (Fig. 1G) and increased FITC-dextran flux (Fig. 1H), and tunicamycin did not affect the increased paracellular permeability (Fig. 1, G and H). Respiratory rate (basal respiration, maximal respiration, and respiratory capacity) and ATP synthesis were reduced by DNP, and these bioenergetic parameters were not restored in tunicamycin co-treated cells (Fig. S3, A-C) (glycolysis was unaffected by DNP Ϯ tunicamycin (Fig. S3D)).
Analysis of the 16 h post-DNP ϩ E. coli Ϯ tunicamycin exposure revealed no significant increase in T84 cell apoptosis; however, there was an ϳ50% decrease in viability by 36 h post-DNP ϩ E. coli treatment that was significantly reduced by tunicamycin (Fig. S4).

Tunicamycin promotes killing of internalized bacteria
Reduced numbers of intracellular bacteria could reflect a reduction in extracellular bacteria; analysis at the end of each experiment revealed no significant differences in the extracellular bacteria numbers between the groups. The reduced number of viable intracellular E. coli in DNP ϩ tunicamycin co-treated epithelia could be due to uptake inhibition. Immunolocalization ( Fig. 2A) and flow cytometry (Fig. 2B) revealed tunicamycin did not affect the DNP-induced internalization of inert fluorescent beads, suggesting that inhibition of the rate of E. coli internalization is not the cause of the reduced viable bacteria in DNP ϩ tunicamycin-treated T84 epithelia. However, the DNP-induced increase in inert E. coli (K-12) fluorescent bioparticles was abrogated by co-treatment with tunicamycin (Fig. 2C), suggestive of degradation and loss of fluorescence.
E. coli ϩ DNP ϩ tunicamycin-treated epithelia displayed enhanced autophagy as gauged by increased expression of LC3-II and beclin-1 and LC3 immunostaining (Fig. 2, D and E). Postulating that autophagy would target the internalized bacteria (i.e. xenophagy), co-localization of autophagosomes and internalized bacteria was observed in DNP ϩ tunicamycintreated cells (Fig. 2E). Furthermore, tunicamycin antagonism of DNP-induced E. coli internalization was lost in cells lacking the autophagy protein ATG16L1 (Fig. 3, A and B). Also, tunicamycin inhibition of the DNP-induced increased viable intracellular E. coli was not apparent in colonoids from mice expressing the knocked-in variant of Atg16l1 T316A (homologous to the IBD susceptibility trait ATG16L1 T300A ) (Fig. 3, C and D). These findings suggest that ER stress ablated the DNP effect by mobilizing autophagy to kill the bacteria. In accordance, rapamycin (induces autophagy (23)) significantly reduced the number of viable intracellular bacteria and translocation across T84 cell monolayers (Fig. 4, A and B). Autophagy inhibition with 3-MA increased numbers of viable bacteria in tunicamycin-treated wildtype T84 cells (Fig. 4C). Moreover, rapamycin prevented the DNP-induced increase in live E. coli HS transcytosis across human colonic biopsies (Fig. 4D). Like tunicamycin, rapamycin did not affect the DNP-induced drop in TER or FITC-dextran flux across T84 epithelia (Fig. 4, E and F). Although the ability of tunicamycin to induce xenophagy was lost in ATG16L1 transgenic cells, tunicamycin reduced the number of viable bacteria in DNP-treated NOD2 Ϫ/Ϫ T84 cells (Fig. 4G).
These data are in accordance with the supposition by Adolph et al. (24) that the ER stress observed in patients with IBD could compensate for defective/reduced autophagy, and, we add, a reaction to increased transcellular transport of commensal bacteria.

Mitochondrial and ER stress and epithelial barrier ATF-6 mediates inhibition of the DNP-effect
ER stress triggers three major pathways: IRE1 that activates the mitogen-activated protein kinase, JNK, and splicing of XBP1 transcription factor; PERK that leads to eIF2␣ phosphorylation, which acts as a transcription factor, and ATF4 mobilization, another transcription factor; and ATF6 that when added to the luminal side, and translocation into the serosal buffer assessed 3 h later (n ϭ 7; *, p Ͻ 0.05 compared with control). B, mouse colonic enteroids were grown for 7 days, resuspended, and grown as monolayer, and then E. coli (strain HB101, 10 7 cfu) was added to the luminal surface Ϯ DNP (0.1 mM) Ϯ TM (10 g/ml), and bacterial translocation into the basal compartment of the culture well was determined 16 h later (insets, phase contrast images of enteroid and subsequent filter-grown monolayer (10-m section stained with hematoxylin and eosin)). C, monolayers of the human colon-derived T84 cells were grown to confluence (i.e. starting TER ϭ 1000 -1800 ohms⅐cm 2 ) or to ϳ70% confluence (by phase microscopy) on 12-well plates. Epithelia were treated with E. coli (strain HB101, 10 6 cfu) and DNP Ϯ TM. Bacterial translocation from the apical surface of the epithelium into the basal compartment of the culture well (D) and bacterial internalization (as % of extracellular E. coli) were assessed 16 h later (one representative experiment of six experiments). E, increased epithelial (T84 cells) internalization of E. coli induced by the NSAID, indomethacin (indo; 1 M; 16 h treatment), was reduced by TM co-treatment (n ϭ 4 epithelial monolayers).

Mitochondrial and ER stress and epithelial barrier
truncated serves as a transcription factor (25). Tunicamycin resulted in subtle but significant increase in spliced XBP1, ATF4, and ATF6␣ in T84 cells (Fig. 5A). Transient siRNA knockdown of ATF6␣, PERK, or IRE1 in T84 cells revealed that only the former blocked tunicamycin inhibition of DNPevoked increases in viable intracellular E. coli (Fig. 5B). Selective pharmacological inhibitors of JNK (p-JNK increased in the cells) (Fig. 5C), IRE1, and PERK did not affect tunicamycin abrogation of the DNP effect (Fig. 5, D and E), focusing attention on ATF6. The increase in autophagy (i.e. LC3-II protein expression) evoked by tunicamycin was absent in ATF6␣ siRNA-treated T84 cells (Fig. 5F).

DAP kinase-1 is needed for tunicamycin inhibition of DNP-evoked decreased barrier function
DAPK1 is mobilized downstream of ATF6 to regulate autophagy and apoptosis (26). T84 epithelia exposed to E. coli ϩ DNP ϩ tunicamycin displayed increased DAPK1 expression compared with cells treated with ATF6 siRNA (Fig. 6A) (PERK and IRE1 siRNA did not inhibit DAPK1 protein expression (Fig. S5)). Use of a selective DAPK1 inhibitor, designated DAPK6, blocked tunicamycin's ability to reduce DNP-evoked increases in viable intracellular E. coli by T84 cells (Fig. 6B) and the mobilization of autophagy (Fig. 6, C and D). Moreover,

Mitochondrial and ER stress and epithelial barrier
bacteria-autophagosome co-localization was completely lost in DAPK6 co-treated cells (Fig. 6D).
Likewise, knockdown of DAPK1 in T84 cells resulted in inhibition of the tunicamycin effect and increased viable intracellular E. coli numbers (Fig. 7A). Although the activation of ATF6 was not affected by DAPK1 siRNA in DNP ϩ E. coli ϩ tunicamycin-treated epithelia, the increased autophagy was lost, showing that the signal needed to up-regulate autophagy by ER stress was interrupted in the absence of DAPK1 (Fig. 7B). Cotreatment with DAPK1 siRNA and DAPK6 prevented the abil-ity of tunicamycin to block the effects of DNP, but to a degree that was not statistically different from either DAPK6 or DAPK1 siRNA alone (Fig. 7C). These data also support the specificity of DAPK6 in targeting DAPK1, an important observation in the context of the murine in vivo experiments presented below.
Mice treated with DAPK6 were more susceptible to DSSinduced colitis as gauged by disease activity score (Fig. 8A), colon length (Fig. 8B), and weight loss (Fig. 8C). DAPK6 treatment resulted in increased translocation of aerobic bacteria

Mitochondrial and ER stress and epithelial barrier
into the colonic mucosal, mesenteric lymph nodes (MLN), and the spleen (Fig. 8, D-F). Furthermore, intra-rectal delivery of DNP resulted in small increases in bacterial translocation that were aggravated by DAPK6 and alleviated by tunicamycin cotreatment (Fig. 8, G-I).

Discussion
Increased numbers of bacteria, including E. coli, in the mucosa and gut epithelium can characterize IBD (1). Many lines of evidence point to dysregulated mitochondrial function in cohorts of patients with IBD (6, 10). By juxtaposing these two putative etiological factors in enteric inflammation, we showed increased E. coli internalization and transcytosis across epithelia treated with the hydrogen ionophore, DNP, that uncouples oxidative phosphorylation (15). Mitochondria do not function in isolation; they couple to the ER. Indeed, dysregulation of the UPR (or ER stress) can contribute to gut inflammation (27); data in support of this have accumulated mostly from analyses of Paneth (24) and goblet cells (28). The importance of mitochondria-ER interaction in the transporting enterocyte (the majority of the gut surface area) to host defense is poorly appreciated. The findings herein suggest that mobilization of the ATF6-arm of the UPR via DAPK1-dependent autophagy results in killing of bacteria taken into the cell because of disrupted mitochondrial function, effectively reducing the barrier defect elicited by metabolic stress. Support for this conclusion comes from the use of a variety of human-derived cell lines, human biopsies, primary mouse organoids, and murine models. Although each model has its pros and cons, the demonstration of the major phenomenon of this research in four model systems, adds, we believe, validity to the findings.
Many stimuli implicated in IBD reactivation can result in dysfunctional mitochondria (5). For instance, NSAID use can elicit disease relapse in some patients with IBD (29), and NSAID enteropathy is characterized by abnormal epithelial mitochondrial structure (8,9). Modeling mitochondrial dysfunction with DNP treatment provided evidence that targeted loss of mitochondrial function resulted in ROS-dependent internalization of bacteria by enteric epithelia (15).
An excessive UPR can lead to colitis (28), and an inability to mount a normal UPR response increases susceptibility to colitis (17). Here, the hypothesis that a UPR-ER stress response would exaggerate an epithelial barrier defect triggered by increased oxidative stress proved to be incorrect; reduced translocation and fewer viable intracellular bacteria occurred in epithelia treated with ER stressors. Also, the apoptosis that occurred with longer DNP ϩ E. coli exposure (36 h) was reduced by tunicamycin, which would preserve epithelial barrier function. ER stress occurs in response to pathogens: bacterial, viral, and fungi. The demonstration that provocation of an ER stress response to counter the effect of epithelial internalization of commensal bacteria caused by mitochondrial dysfunction underscores that this is an important aspect of innate immunity when challenged with intracellular microbes.
Autophagy is important in host defense, where, for example, protection against colitis by outer membrane vesicles from Bacteroides fragilis requires ATG16L1-competent dendritic cells (30). Tunicamycin did not affect internalization of inert beads into DNP-treated epithelia but enhanced autophagy markers suggesting that the reduction in viable intracellular bacteria was because of reduced survival. In accordance, the protective effect of tunicamycin was lost in epithelia lacking the autophagy protein, ATG16L1. The threonine-to-alanine (T300A) ATG16L1 variant that increases the risk for Crohn's disease results in degradation of the protein (31), reducing autophagic capacity and leaving the individual vulnerable to pathogenic bacteria (32,33). The equivalent mutation in the mouse (T316A) also targets Atg16l1 for destruction (31). Monolayers from colonoids of the Atg16l1 T316A knockin mouse not only had increased responses to DNP, as gauged by E. coli internalization but tunicamycin failed to reverse this event. These data support the view that the UPR via autophagy antagonizes the barrier defect evoked by perturbed mitochondrial function.
Given the significance of the colonic microbiota in IBD pathophysiology and that transporting enterocyte is the major constituent of the epithelium, we speculate that the impact of the ATG16L1 T300A mutation would be compounded in the context of reduced mitochondrial activity due to NSAID use or other stimuli (6). Furthermore, a cytoprotective effect of ATG16L1 in epithelial cells promoted mitochondrial homeostasis and prevented necroptosis (34). Collectively, these data suggest that the development of drugs that drive autophagy (35) or an appropriate UPR (ER stress) pathway would be of value in treating enteropathies (e.g. IBD) characterized by loss of epithelial barrier function, i.e. uptake and passage of commensal bacteria, pathobionts, and pathogens, due to perturbed mitochondrial function. In this context, the anti-colitic pro-drug thioguanine promoted autophagy in human colon-derived epithelial cell lines, limiting Salmonella typhimurium replication (36).
The balance of this mitochondria-ER-autophagy cross-talk would be central to cell health. Up-regulation of autophagy would prevent the effects of excessive ER stress, and in the short-term, it has the benefit of targeting internalized bacteria as a consequence of mitochondrial dysfunction. Contrarily, prolonged ER stress and/or autophagy would be detrimental and potentially of pathophysiological significance (37,38).
Essentially the UPR is a three-pronged pathway, in which any one arm can dominate (39), such as the mobilization of Ire1/ Xbp1 signaling following bacterial infection or accumulation of Ire1␣ aggregates in Atg16l1 ⌬IEC mice that may contribute to Crohn's disease-like ileitis (37). Molecular knockout and use of PERK, IRE1, and JNK-selective inhibitors indicated that tunicamycin abrogation of the DNP-induced epithelial barrier defect was dependent on the ATF6 arm of the UPR. Other stud-ies support a beneficial role for Atf6 in limiting enteric inflammation: mutation in a protease resulting in reduced Atf6 activation increased susceptibility to DSS-induced colitis (data on bacterial translocation were not presented) (40); Atf6 Ϫ/Ϫ mice display increased susceptibility to bacteria correlating with decreased autophagy in macrophages (41). Furthermore, IL-10

Mitochondrial and ER stress and epithelial barrier
can exert an anti-inflammatory effect via inhibition of TNF␣induced nuclear translocation of ATF6 (20).
ATF6 regulates DAPK1 synthesis (26), which can phosphorylate Beclin-1 to induce autophagy (42) and stimulate trafficking of Atg9-containing vesicles for autophagosome formation in response to starvation (43); dapk1 Ϫ/Ϫ fibroblasts display less autophagy following tunicamycin treatment (44). Although preliminary, a polymorphism in the DAPK1 gene is reported in some patients with Crohn's disease (45). Noting that ATF6 siRNA-treated epithelia had reduced DAPK1 expression, it was determined that tunicamycin failed to ablate the DNP-evoked increase in viable intracellular E. coli in epithelia treated with a DAPK1-selective inhibitor, DAPK6, or DAPK1 siRNA, correlating with reduced autophagy. Moreover, treatment of mice with 2.5% DSS resulted in bacterial translocation into the colonic mucosa and MLNs that was significantly increased by DAPK6 co-treatment. Similarly, bacterial translocation (i.e. barrier defect) evoked by intra-rectal DNP (which was reduced by tunicamycin co-treatment) was enhanced by systemic delivery of DAPK6.
There is an elegance in assessing a single variable in an experimental system; however, the coordinated activity of intracellu-  Male C57Bl/6 mice were treated with 2.5% (w/v) DSS (ϳ40 kDa) in their drinking water for 5 days (A-F). Additional mice received the DAPK1 inhibitor DAPK6 (20 g/kg, i.p. daily). Mice were euthanized, and disease activity was assessed in the DSS-treated mice, and bacterial translocation into the colonic mucosa, MLN, and spleen was determined in all mice. G-I, male C57Bl/6 mice were intra-rectally administered DNP (3 mM) for 24 h. Additional mice received DAPK6 (20 g/kg, i.p. daily) Ϯ TM (20 g/kg, i.p. daily). Mice were euthanized, and bacterial translocation into the colonic mucosa, MLN, and spleen was determined in all mice (mean Ϯ S.D.; * and #, p Ͻ 0.05 compared with control and DSS or DNP, respectively).

Mitochondrial and ER stress and epithelial barrier
lar signaling cascades regulates cell and tissue function. By defining an ATF6 -to-DAPK1-to-autophagy (and apoptosis) pathway, we define how the UPR can preserve epithelial barrier function when faced with deranged mitochondrial function. The implications of this could be wide-ranging. In the context of IBD, the effect of reduced activity or the loss of ATG16L1 affects not only macrophages, goblet cells, and Paneth cells but the transporting enterocyte as well, negating the value of mobilizing ATF6 to counter the transcellular passage of microbes across metabolically-stressed epithelia. It will be important to address how the mitochondria-to-ATF6 -DAPK1-autophagy interaction is impacted by other molecules that control epithelia-microbe interaction (e.g. NOD proteins, NF␤, and the inflammasome (38,46,47)). For instance, cells devoid of NOD2 are more sensitive to DNP-induced barrier defects (48), and reduced epithelia autophagy due to knockout of the amino acid sensor GCN2 (a starvation model analogous to DNP exposure) was linked to ROS generation, inflammasome activation, and intestinal inflammation (47).
Epithelial-microbial interactions are at the nexus of IBD. Having defined a molecular pathway in the transporting enterocyte by which mobilization of the UPR abolishes the epithelial transcellular permeability defect (i.e. bacteria uptake and passage) in metabolically-stressed cells, we suggest that this is critical in gut homeostasis. Controlled ER stress, via autophagy, rescues the enterocyte from the loss of barrier function, as measured by bacterial transcytosis, caused by uncoupling oxidative phosphorylation. Reciprocally, the pathophysiological consequences of compromised autophagy or ER stress (both susceptibility traits for IBD) is compounded by perturbed mitochondrial function that drives the internalization of commensal bacteria by the epithelium, the enhanced survival of which could be a significant pro-inflammatory stimulus.

Human biopsies and murine colonoids
Human colonic biopsies were obtained from two healthy men and five women (mean age 78 years) at the Gastroenterology Department, University Hospital in Linköping. The Regional Committee of human ethics approved the study, and all subjects gave informed consent.
Mice with the Atg16l1 T316A knockin gene (from M. van Lookeren Champagne, Genentech, San Francisco) were housed in standard conditions under protocol AC13-0015 (Animal Care Committee, University of Calgary). Experiments conformed to Canadian guidelines on animals in research. Colonic crypts were isolated (49), cultured in IntestiCult TM organoid growth medium (Stemcell, BC Canada), and monolayers grown on filter supports (50). E. coli HB101 was grown overnight in Luria-Bertani (LB) broth, and 10 7 colony-forming units (cfu) were added to the luminal aspect of filter-grown epithelia, and bacterial translocation was assessed by sampling medium from the basal compartment of the culture well.

Ussing chambers studies
Human colonic biopsies were mounted in Ussing chambers (1.8-mm 2 surface area) (51). Live non-pathogenic GFP-labeled E. coli (HS strain; 10 8 cfu/ml) (52) were added to the mucosal side of the tissue with DNP (0.1 mM) Ϯ tunicamycin (10 g/ml) or rapamycin (100 nM). Additional biopsies from the same individual bathed in Krebs buffer only served as controls. All conditions were performed in duplicate biopsies. After 3 h, serosal buffers were collected, and fluorescence was measured at 488 nm in VICTOR TM X3 multileader plate reader (PerkinElmer Life Sciences, Sweden). Other biopsies were homogenized in 200 l of RIPA buffer, serially diluted, plated onto LB agar, cultured under aerobic conditions (ϳ16 h, 37°C), and colonyforming units counted.

Cell culture, bacterial translocation, and intracellular viable bacteria
The T84, CaCo2, and HCT116 human colonic epithelial cell lines and the cervical epithelial HeLa cell line were maintained as described (15) (T84 NOD2 Ϫ/Ϫ cells were used generated via shRNA (48)). Cells (10 6 /ml) were seeded into 12-well plates and used at ϳ70% confluence based on phase-contrast microscopy or onto 3.0-m porous filter supports and cultured for 7 days until electrically confluent (T84 Ͼ1000 ohms⅐cm 2 ; HT-29 Ͼ300 ohms⅐cm 2 ) (15). E. coli HB101 was grown overnight in LB broth, and 10 6 cfu were added to the epithelium (luminal aspect of filter-grown polarized preparations). The intracellular viable bacteria number was assessed by killing extracellular bacteria with 200 g/ml gentamicin (2 h), followed by enterocyte lysis (0.1% Triton X-100), and culturing serial dilutions of the cell extract. Bacterial translocation was assessed by sampling the medium from the basolateral compartment of filter-grown epithelia (15). All drugs were added simultaneously with the E. coli and were tested (at the doses used) for bacteriostatic and bactericidal effects.

TUNEL assay
Apoptotic cells were detected by TUNEL using ApopTag in situ apoptosis detection kits (EDM Millipore).

Epithelial paracellular permeability
TER across filter-grown T84 cell monolayers was recorded at time 0 and 16 h later using a voltmeter and matched electrodes (Millicell-ERS, Millipore, Bedford, MA). Data are presented as a percentage of pretreatment TER value.
Following treatment, FITC-dextran (80 g/ml, 4 kDa) in fresh culture medium was added to the apical surface of filtergrown T84 cell monolayers for 4 h at 37°C. Samples were collected from the basolateral well of the chamber, and fluores-cence was read at excitation (427 nm) and emission (536 nm) wavelengths.

Cell bioenergetics
The oxygen consumption rate and extracellular acidification rate were measured using a Seahorse XF-24 metabolic flux analyzer (Seahorse Bioscience, North Billerica, MA).

GFP-LC3 autophagy detection
HeLa cells expressing LC3-GFP (53) were used, and images were captured by confocal microscopy. ImageJ was used for unbiased LC3 puncta analysis in the mid-optical section of live cells. Cells were identified as a region with greater than 1 S.D. higher intensity MAP2 or HA stain than the mean intensity for that individual image. A punctum was identified as a region of LC3 stain that was more than 2 S.D. brighter than the individual image mean. All LC3 puncta larger than 0.1 m within the MAP2 or HA stain were included in the analysis (smaller puncta are outside the size range of phagophores/autophagosomes at 0.2-10 m).

ATG16L1 knockout
CRISPR-Cas9 methodology (54) was used to knockout ATG16L1 in the human HCT116 epithelial cell line. A gRNA target region (5Ј-ATGAGTATCCACATTGTCCT-3Ј) was identified using the crispr.mit.edu website (Zhang laboratory at MIT), cloned into pSpCas9(BB)-2A-Puro plasmid (Addgene plasmid 48139), and verified by sequencing. HCT116 cells were grown to 50% confluence in 100-mm tissue culture dishes and transfected with 2 g of pSpCas9(gRNA)-2A-Puro plasmid using 20 l of polyethyleneimine (Polysciences, Warrington, PA). Successfully transfected cells were selected 24 h later with 3 g/ml puromycin. Limiting dilution was used to create clonal cell lines, of which the identity and knockout of ATG16L1 were verified by sequencing and immunoblotting for ATG16L1.

In vivo bacterial translocation
Data from mouse models have limitations when extrapolating to humans; indeed, there is no model of IBD, but there are useful models of colitis for pre-clinical investigation. Bearing this in mind, colitis was induced in 7-9-week-old male C57Bl/6 mice (Charles River Laboratories, Senneville, Quebec, Canada) by 2.5% (w/v) DSS (ϳ40 kDa) in drinking water for 5 days (an often used model (15)) Ϯ daily DAPK6 (20 g/kg; i.p.). At necropsy, a disease activity score based on body weight, colon length, and the animal's macroscopic appearance, colon, and stool was calculated. Two-cm pieces of colon, the mesenteric lymph nodes, and spleen were excised, rinsed in gentamicin (200 g/ml), homogenized, serial dilution plated onto blood agar (37°C, 24 h), and cfu counted.

Statistical analysis
Data are means Ϯ S.D. Multiple group comparisons were performed by one-way analysis of variance, followed by pairwise post hoc tests, with p Ͻ 0.05 as a level of statistical significant difference.