|Home | About | Journals | Submit | Contact Us | Français|
Although enterocytes are capable of innate immune responses, the intestinal epithelium is normally tolerant to commensal bacteria. To elucidate the mechanisms of tolerance, we examined effect of pre-exposure to LPS on activation of p38, c-Jun and NF-kB in enterocytes by several inflammatory and stress stimuli. Shortly after the initial LPS challenge, enterocytes become tolerant to re-stimulation with LPS or CpG DNA, but not with IL-17 or UV. The state of tolerance, which lasts 20-26 h, temporally coincides with LPS-induced expression of the anti-inflammatory ubiquitin-editing enzyme A20. Small interfering RNA (siRNA) silencing of A20 prevents tolerance, whereas ectopic expression of A20 blocks responses to LPS and CpG DNA, but not to IL-17 or UV. A20 levels in the epithelium of the small intestine are low at birth and following gut decontamination with antibiotics, but high under conditions of bacterial colonization. In the small intestine of adult rodents, A20 prominently localizes to the luminal interface of villus enterocytes. Lower parts of the crypts display relatively low levels of A20, but relatively high levels of phospho-p38. Gut decontamination with antibiotics reduces the levels of both A20 and phospho-p38. Together with the fact that A20-deficient mice develop severe intestinal inflammation, our results indicate that induction of A20 plays a key role in the tolerance of the intestinal epithelium to TLR ligands and bacteria.
Intestinal epithelial cells express TLR and produce a variety of inflammatory factors in response to stimulation with TLR ligands, which is important for the gut homeostasis (1-8). Despite its potential sensitivity and exposure to high concentrations of bacteria, the intestinal epithelium is refractory to induction of inflammation by commensals under normal conditions. Being sensitive at birth, the epithelium becomes unresponsive, or tolerant, shortly upon initial stimulation with colonizing bacteria (9). As bacterial colonization persists, so does the tolerant state, which likely accounts for the lack of dramatic inflammatory reaction to commensals.
Several mechanisms that inhibit TLR signaling in enterocytes have been identified. Decreased expression of TLR4 (10, 11), or its co-receptor MD-2 (12, 13) has been proposed as a mechanism of hyporesponsiveness to LPS in the intestine. MD-2 is degraded by trypsin, which may suppress responses to LPS in the small intestine (14). Tolerance to TLR2 and TLR4 ligands in colonocytes has been reported to result from induction of TLR-interacting inhibitory protein Tollip (15). However, LPS does not induce Tollip in m-ICcl2 enterocytes of the small intestinal origin (9), and lack of intestinal inflammation in Tollip-/- mice (16) argues against pivotal role of Tollip in the intestinal tolerance to bacteria. Tolerance to LPS in m-ICcl2 enterocytes is associated with post-translational inhibition of the interleukin receptor-associated kinase IRAK-1, a key mediator of TLR signaling (9). Definitive mechanisms of tolerance to TLR signaling in the gut remain elusive (17).
A20 is a Zn finger protein whose gene is an early target of the pro-inflammatory transcription factor NF-κB (18, 19). A20 inhibits activation of NF-κB via inflammatory cytokine receptors (19-27), TLR (28, 29), and nucleotide-binding oligomerization domain-containing receptors NOD2 (30) by its two ubiquitin-editing activities, N-terminal deubiquitinase that removes lysine (K)63-linked polyubiquitin chains, and carboxy-terminal ubiquitin ligase that facilitates target protein degradation via attachment of K48-linked polyubiquitin chains (31, 32). These two activities cooperatively down regulate the key K63 polyubiquitination-dependent mediators of inflammatory signaling, TNFα receptor-associated factor 6 (TRAF6) (33, 34) and receptor-interacting protein kinase (RIP) (35). As A20-/- mice develop severe intestinal inflammation early in life (36, 37), it was suggested that A20 is important for the inhibition of innate immune responses in the gut (17). A role of A20 in the regulation of intestinal inflammation is further suggested by its inhibitory effect on TLR2-mediated production of IL-8 in enterocytes (38).
To gain insight into the regulation of innate immune responses in the small intestine, we examined tolerance to LPS in several enterocyte cell lines and in the native small intestinal epithelium. Here we report that LPS-induced expression of A20 is necessary and sufficient for the development of hyporesponsiveness to repeated stimulation with LPS, that A20 prominently localizes to the luminal interface of villus enterocytes in adult rodents, and that A20 levels in the intestinal epithelium positively correlate with the bacterial load. These findings point to the key role of A20 in the development of intestinal tolerance to the commensal bacteria.
All animal experiments have been approved by CHLA Animal Care and Use Committee. Newborn rats were obtained from timed pregnant Sprague Dawley females (Harlan). C57Bl/6 mice were bred in-house; adult mice were used at 12-16 wk age. Selective gut decontamination in mice raised under specific pathogen-free conditions was achieved by supplementing drinking water with polymyxin B and neomycin at 0.4 mg/ml ea. and rifaximin at 0.2 mg/ml for 7 d. Full-term newborn rodents were obtained by Caesarean (C)-section, avoiding introduction of bacteria or pyrogens, and maintained at 32°C and 80% relative humidity. Segments of neonatal terminal ileum were gently flushed with DMEM+10% FCS and incubated in the same medium at 37°C and 10% CO2. Mucosal scrapings were obtained by gentle scraping of mucosal side of longitudinally opened small intestine with a blunt end of a sterile scalpel; 70-80% of cells in mucosal scrapings were enterocytes. IEC-6, IEC-18, SW480, HEK293 cells and E. cloacae 29004 were purchased from ATCC; RIE-1 cells were a gift from Dr. Hua Xu, University of Arizona. SW480 cells were grown in Leibowitz medium + 10% FCS in humidified air; other cell lines were grown in DMEM + 10% FCS in 10% CO2. IEC-6, IEC-18, and RIE-1 cells were used at passage 20-30. Abs were from the following sources: A20, US Biologicals (immunofluorescence, IF), Alexis (Western blot for human A20) Dr. Averil Ma, UCSF (Western blot for rodent A20); V5, Invitrogen; β-actin, Sigma; IkB, p38, phospho-p38, c-Jun, phospho-c-Jun, β-catenin, GAPDH, Cell Signaling; Tgn38/S-20, Muc2/H-300, HSP70/K-20, Santa Cruz Biotechnology; TRAF6, Upstate. Mouse recombinant IL-17 and LPS from E. coli 0127:B8 were from Peprotech and Sigma. LPS was used without additional purification. We have previously found that purification of this commercial preparation by DNAse treatment, Proteinase K digestion, and repeated phenol extraction did not change its ability to elicit dose-dependent responses in IEC-6 cells (39). Moreover, effects of 1 μg/ml LPS from Sigma on p38, c-Jun, and NF-kB in IEC-6 cells were completely abrogated by 20-min pre-incubation with 20 μg/ml Polymyxin B, or adsorption on Polymyxin B agarose (Sigma), which rules out contribution of TLR ligands other than LPS. Synthetic oligonucleotide TCGTCGTTTCGTCGTTTTGTCGTT was used as CpG DNA.
Total RNA was extracted and purified using Trizol reagent (Invitrogen) and converted into first strand cDNA using oligo-dT and Moloney reverse transcriptase. Equivalent amounts of the first strand reaction were used as template in real time PCR with pairs of primers listed in Table S1. Primers in each pair belong to different exons to distinguish between cDNA and chromosomal amplicons. Real time PCR (95°C 10 min followed by 40 cycles of 95°C 10 s, 55°C 1 s, 72°C 40 s) was performed using the FastStart DNA masterPLUS SYBR Green I kit on the Light Cycler 480 (Roche), and critical cycle was determined for each reaction. The amount of target sequence was deduced from a calibration curve and normalized to the amount of the housekeeping transcript (Rps11) in the same sample. Samples containing more than 5% genomic amplicons, as determined by melting curves, were not scored. IF of cultured cells or paraffin sections were performed as recommended by Ab manufacturers. Prior to fixation, intestinal segments were flushed with PBS + 7 mM DTT to remove surface mucus. Images were taken on BX51 microscope equipped with color camera, using Picture Frame software (Olympus). For comparisons, samples were processed on the same slide and photographed at the same camera settings; identical adjustments were applied to the images. Care was exercised to minimize photobleaching. Poly-A+ RNA for Northern blots was isolated using the Dynabeads kit (Dynal). RNA samples (2 μg/lane) were resolved on 1% agarose-glyoxal gel. Northern analysis was performed using the NorthernMax-Gly kit (Ambion) and 32P-labeled cDNA probes. RNA bands were quantified by densitometry of underexposed autoradiograms using GelDoc scanner and Quantity One software (Bio-Rad).
Rat A20 open reading frame was amplified using primers CACCATGGCTGAACAACTTCTTCCT and GGCGTACATCTGCTTGAACTG, Deep Vent DNA polymerase (New England Biolabs), Moloney reverse transcriptase, and RNA from LPS-treated IEC-6 cells. Resulting RT-PCR product was inserted into pcDNA3.1-V5His (Invitrogen) to yield pcDNA3-A20. The insert and junctions were sequenced to verify the absence of mutations. To generate A20 siRNA, the partial A20 cDNA fragment was amplified with the primers GGGTAATACGACTCACTATAGAAACCAACGGTGATGGAAACTGCC and GGGTAATACGACTCACTATAGAGATGCCGTTAAACGTCCGAGTG; the amplification product was used as template to direct T7 RNA polymerase-dependent synthesis of dsRNA, which was then cut into 21-bp siRNA duplexes using the ShortCut ribonuclease (New England Biolabs). Control siRNA was prepared similarly using the LEU2 template generated from yeast DNA with primers GGGTAATACGACTCACTATAGCACGTTGGTCAAGAAATCACAGCC and GGGTAATACGACTCACTATAGAACTTCTTCGGCGACAGCATCACC.
IEC-6 or IEC-18 cells grown overnight to 90% confluence were gently trypsinized, washed with RPMI + 10% FCS, collected by centrifugation at 100 × g for 5 min, and re-suspended at 3×107 cells/ml in the Nucleofector Solution V (Amaxa). 100 μl suspension aliquots were mixed with 10 μg plasmid DNA and electroporated using the T-030 protocol of Amaxa Nucleofector. Following 10 min recovery in RPMI + FCS at 37°C, cells were plated in DMEM + FCS. Stable transfectants were selected with 1 mg/ml G418. HEK293 cells were transfected using Lipofectamine 2000 (Invitrogen), as directed by the manufacturer. Near-confluent IEC-6 monolayers, mucosal scrapings, or longitudinally cut ileal segments from neonatal rats were transfected with siRNAs complexed with Lipofectamine 2000.
HEK293 cells transiently transfected with pcDNA3-A20 were collected in the hypotonic buffer (70 mM Na-PO4 pH 7.0, 1 mM PMSF) 40 h post transfection and disrupted in Dounce homogenizer on ice (at least 95% cells disrupted by Trypan Blue staining). Following 10 min centrifugation at 10,000 × g, the cleared lysate was adjusted to 10% glycerol, 0.3 M NaCl and His6-tagged A20 protein was purified by adsorption on the Talon metal affinity resin (Clontech), as recommended by the manufacturer. Purity and concentration of the resulting preparation was evaluated on Coomassie-stained polyacrylamide gel by comparison to a series of standard BSA loads.
Small intestine of an adult mouse was flushed with PBS and opened by a longitudinal cut. The mucosal layer was scraped off and homogenized in 20× volume of PBS + 1 mM PMSF. Following 15 min centrifugation at 10,000 × g, MgCl2 was added to the supernatant to 10 mM and nucleic acids were digested with 20 μg/ml ea. of RNAse A and DNAse I for 5 h at 37°C. The soluble mucus was purified from non-covalently attached proteins by 3 rounds of equilibrium CsCl gradient centrifugation (40). Fractions containing mucus were identified by spotting onto nitrocellulose and probing with the Muc2 Ab. The final preparation was dialyzed against PBS and protein concentration was determined using the Bio-Rad protein assay.
Cells were suspended in the hypotonic buffer and disrupted on ice in a Dounce homogenizer. The homogenate was centrifuged for 5 min at 100,000 × g, 15°C in the A-95 rotor of the Airfuge ultracentrifuge (Beckman Coulter). Equivalent amounts of supernatant (soluble fraction) and pellet (particulate fraction) were analyzed by Western blotting.
Quantitative data were expressed as mean ± standard deviation. Pairs of data were compared using the Mann-Whitney test, with differences considered significant at P>0.95.
To gain insight into intestinal tolerance to bacteria, we examined the activating phosphorylation of p38 MAPK in response to repeated stimulation with LPS in several enterocyte cell lines. LPS activates p38 MAPK in IEC-6, IEC-18, RIE-1, and SW480 cells at concentrations as low as 2 ng/ml; maximum activation requires 0.5-1 μg/ml. The activation peaks at 5-15 min and returns to the baseline in about 1 h (data not shown). 15-min pre-exposure to LPS inhibits p38 activation in response to re-stimulation with this TLR ligand, beginning 1 h after the initial challenge, in each of the four cell lines. In proliferating cultures, cells remain unresponsive to LPS for 20-28 h (Fig. 1A). However, in stationary cultures, the unresponsive state may last over 72 h (Fig. 1B). Our findings corroborate previously described tolerance to TLR ligands in enterocytes(9, 15, 41).
To test whether pre-treatment with LPS affects other intracellular mediators of TLR signaling, we examined activation of NF-kB and phosphorylation of the transcription factor c-Jun following re-stimulation with LPS at various times after the initial challenge. Pre-treatment of IEC-6 cells with LPS dramatically inhibits these responses. No marked degradation of IkB, increase in phospho-c-Jun (Fig. 2A) or increase in NF-kB DNA binding activity (Fig. S1) occurs in response to re-stimulation at 1-18 h after the initial challenge. Therefore, a brief initial challenge with LPS profoundly inhibits the subsequent LPS-induced activation of p38, c-Jun, and NF-kB.
We next asked whether desensitization to LPS directly affects p38, c-Jun, or NF-kB. If pre-treatment with LPS renders these targets refractory, one might expect their unresponsiveness to other inflammatory and non-inflammatory stimuli. To test this, we examined effects of pre-exposure to LPS on responses to UV (stress), IL-17 (inflammatory cytokine that acts independently of TLR signaling), and unmethylated CpG DNA (TLR9 ligand). Pre-exposure to LPS inhibits responses to LPS and CpG DNA, but not to UV or IL-17. Pre-exposure to CpG DNA inhibits responses to CpG DNA and LPS (Fig. 2B shows data for IEC-6 cells; similar data for IEC-18, RIE-1, and SW480 cells not shown). Normal sensitivity to inflammatory cytokines in enterocytes tolerized to LPS have been previously reported by others (15). Because tolerance to LPS does not affect activation of p38, c-Jun, and NF-kB by stimuli other than TLR ligands, desensitization likely targets common upstream step(s) in the TLR signaling cascade.
Development of tolerance to LPS may involve induction of negative, and/or repression of positive regulators of inflammation. To identify genes potentially implicated in the development of tolerance to LPS, we used real time RT-PCR to examine the time course of LPS-induced expression of 28 innate immune transcripts in IEC-6 cells (Table S1). A20, TIRAP/MAL, and ST2 were induced by LPS. TIRAP/MAL and ST2 were not studied further because the former is a positive regulator of inflammation, and induction of the latter takes over 8 h, whereas tolerance to LPS develops in ~1 h after LPS challenge. This report focuses on the role of A20 in the regulation of responses to LPS.
To confirm induction of A20 at the mRNA level, we employed Northern analysis. In IEC-6 cells, A20 mRNA is barely detectable under basal conditions, however, its levels surge over 100-fold at 1 h of treatment with LPS, or Gram-negative enteric bacteria Enterobacter cloacae (Fig. 3A). A20 mRNA levels subsequently decrease, yet they remain above the baseline for at least 17 h (Fig. 3A, B). LPS also induces the A20 mRNA in IEC-18 and RIE-1 cells (Fig. 3A). Inhibitor of RNA polymerase II α-amanitin blocks A20 induction (Fig. 3C), indicating the requirement of transcription. Neither UV, nor IL-17 induces A20 in IEC-6 cells (data not shown). Thus, A20 transcription is strongly but transiently induced by LPS or Gram-negative enteric bacteria.
To establish whether the accumulation of A20 mRNA in LPS-treated enterocytes is associated with the increased expression of A20 protein, we used Western blotting with A20 Abs. A20 protein is strongly induced in enterocytes after 1 h of LPS or CpG DNA treatment, and levels of A20 protein remain high within at least 4-8 h (Fig. 3D shows data for LPS- and CpG DNA-treated IEC-18 and LPS-treated SW480 cells; other data not shown). As an independent approach to evaluating A20 levels, we used IF with A20 Ab. The chicken A20 Ab from US Biologicals (the two other A20 Abs did not work in IF) specifically reacts with A20 in IEC-6 cells because staining is not observed with chicken pre-immune serum, and because pcDNA3-A20-transfected, but not vector-transfected cells display strong A20 signal (Fig. S2A). A20 IF intensity reaches plateau at 1 h of LPS treatment, remains high for at least 8 h, and somewhat decreases by 24 h (Fig. S2B), which corroborates Western blot results. The long-lasting increase in A20 protein levels following the surge of the A20 transcript suggested that this protein is stable. To assess the stability of A20 protein, IEC-6 cells transfected with pcDNA3-A20 were incubated with the protein synthesis inhibitor cycloheximide, and levels of the V5 epitope-tagged A20 protein were examined by Western blotting. t1/2 of A20-V5, as determined by cycloheximide chase, is about 2 days (Fig. 3E). In proliferating cells, the ratio of A20 to total protein is likely to decrease faster than in quiescent cells due to the ongoing protein synthesis after the cessation of A20 expression, which may explain faster recovery of LPS responses in proliferating cultures (Fig. 1A, B). LPS-induced expression and stability of A20 protein are consistent with its role in the development of long-lasting hyporesponsiveness to LPS.
From the images in Fig. S2 it appears that A20 is not homogenously distributed inside the cell, as expected of a cytosolic protein, but displays granular localization. To assess intracellular distribution of A20, we separated cell homogenate into soluble and particulate fractions by high speed centrifugation, and examined A20 content of each fraction by Western blotting. Successful fractionation was confirmed by probing fractions for glyceraldehyde phosphate dehydrogenase, a typical soluble protein, and β-catenin, a typical protein associated with the particulate fraction. About half of A20 in LPS-stimulated IEC-6 cells is associated with the particulate fraction (Fig. 4A). This supports recently reported association of A20 with intracellular membranes (42). A20 does not localize to Golgi bodies in IEC-6 cells (Fig. 4B). Although A20 IF appears distributed throughout the cell (Fig. S2 and and4B),4B), confocal microscopy reveals that this protein is largely excluded from the nuclei (Fig. 4C).
To examine causal relationship between A20 expression and tolerance to LPS, we evaluated responses to LPS under conditions of A20 ectopic expression or silencing. Transfection of IEC-6 cells with pcDNA3-A20, but not with the empty vector, dramatically decreases phosphorylation of p38 and c-Jun, as well as degradation of IkB in response to LPS or CpG DNA, but not UV or IL-17 (Fig. 5A), indicating that A20 expression is sufficient for the specific inhibition of TLR signaling. Because our transfectants expressed A20 at the level similar to that in LPS-stimulated cells (Fig. 5B), inhibition of LPS signaling was not due to massive overexpression of the A20 protein. By contrast, transfection with A20 siRNA, but not with the control LEU2 siRNA, abrogates establishment of hyporesponsiveness to LPS-induced p38 activation and IkB degradation within 1-16 h of the initial LPS challenge (Fig. 5C). Experiments with IEC-18 transfectants yielded similar results (data not shown). These results demonstrate that A20 expression is both necessary and sufficient for the tolerance to LPS in enterocyte cell lines.
To gain an insight into the role of A20 in the small intestine, we examined A20 IF in the ileal sections from adult mice. The A20 Ab (US Biologicals) prominently decorates the luminal surface of the villi and large spots that were tentatively identified as goblet cell cups (Fig. 6A, left). The staining is specific because it is abrogated by substitution of the primary Ab with pre-immune serum, and by pre-incubation of the primary Ab with the excess of affinity-purified A20 protein (Fig. 6B). Staining of goblet cells raised the concern that the Ab might have also reacted with the intestinal mucus. Indeed, the staining is abrogated by pre-incubation of the Ab with the excess of purified intestinal mucus (Fig. 6B). Therefore, the A20 Ab has a spurious affinity to the mucus, which apparently accounts for the non-specific staining of goblet cell cups. To distinguish between A20 and mucus staining, we performed two-color IF for A20 and the predominant intestinal mucin Muc2. The Muc2 Ab decorates goblet cell cups (Fig. 6A, middle), which appear yellow/orange on the merged image due to co-staining with the A20 Ab (Fig. 6A, right). Although surface mucus was removed by rinsing with PBS + 7 mM DTT, it was possible that surface staining was due to Ab reactivity against residual mucin other than Muc2. To rule this out, we performed two-color staining for A20 and β-actin, the cytoskeletal protein that localizes to the apical sub-membrane of villus enterocytes. Co-localization with β-actin (Fig. 6C) indicated that the A20 signal is intracellular, which rules out Ab reactivity with surface mucus. Thus, pure green color on the merged image Fig. 6A identifies A20. Although A20 is concentrated in the apical sub-membrane in villus enterocytes, it localizes diffusely in the epithelium of upper parts of the crypts and crypt openings (Fig. 6A, arrowheads). A20 signal is largely absent from the lower parts of the crypts. In villus enterocytes of adult rodents, a sizeable fraction of A20 is found in the supranuclear spots, which were identified as Golgi bodies by co-localization with the Tgn38 marker (Fig. 7A). Localization of A20 to the apical sub-membrane and Golgi may indicate association of this protein with membrane-bound TLR signaling complexes. Since the A20 target TRAF6 is a known member of TLR signaling complexes that associates with membranes (43), we examined localization of A20 and TRAF6 using two color IF. The strongest TRAF6 signal was observed in the apical sub-membrane of villus enterocytes (Fig. 7B); somewhat weaker signal was found in supranuclear punctate structures, presumably secretory vesicles or other membranous organelles (Fig. S3). These results are consistent with patterns of intracellular TRAF6 localization that have been described previously (44-47). Sub-membrane TRAF6 co-localizes with A20 (Fig. 7B), which supports A20 recruitment to the apical membrane-bound TLR signaling complexes.
Having validated the A20 Ab for IF in intestinal sections, we set out to examine changes in A20 levels associated with ageing and bacterial colonization of the gut. Levels of A20 protein in the epithelium of rat ileum are low at birth (Fig. 8A, left; Fig. S4A, left), but increase by day 4, the time of emergence of significant bacterial population of the gut, and are high at 6 mo, the time by which the intestinal microbiota are fully established (Fig. 8A, left; Fig. S4A, middle and right). In the epithelium of 4 day old rats, A20 localizes diffusely and is present at high levels in the villi, but at relatively low levels in the nascent crypts (Fig. S4A, middle). By 6 mo of age, the characteristic localization similar to that seen in Fig. 6A can be observed (Fig. S4A, right). Similar patterns were found in the intestines of newborn, 3 day old, and 12 wk old mice (data not shown). Therefore, levels of the intestinal A20 increase postnatally and remain high through adulthood.
Increased epithelial expression of A20 in 4 day old rats suggested that this protein might be induced in vivo by LPS or other TLR ligands of colonizing bacteria. To test this, we examined LPS-induced expression of A20 in the naive neonatal intestine. Newborn rats obtained by C-section at term were kept under sterile conditions and orally gavaged with LPS solution or water. A20 protein and mRNA levels, respectively, were higher in the in the ileal epithelium of the group that received LPS (Fig. 8A, middle; Fig. S4B), indicating A20 induction by oral LPS. Because induction of A20 could have been due to systemic effects of luminally administered LPS, we also performed LPS treatment ex vivo. Rat ileal segments were excised immediately after birth and treated with or without LPS. Ex vivo treatment with LPS increases A20 protein and mRNA levels in the epithelium (Fig. 8A, right; Fig. S4C), demonstrating that induction of A20 by luminally administered LPS may be a local response that does not require systemic inflammation.
We next examined, using antibiotic treatment, whether reduction of bacterial load of the lumen decreases expression of A20. Gut decontamination with a mixture of antibiotics of limited oral bioavailability in drinking water markedly reduces A20 protein and mRNA levels in the ileal epithelium (Fig. 8B, C). Partial rather than complete abrogation of A20 expression is likely due to incomplete elimination of intestinal bacteria by the antibiotic treatment. These results show that A20 expression in the epithelium positively correlates with bacterial load of the lumen.
To assess the effect of gut decontamination on the inflammatory signaling, we examined effect of antibiotic treatment on phosphorylation of p38 in the intestine. In conventionally housed mice that received regular drinking water, moderate levels of phospho-p38 are present the epithelium of the upper crypts and lowermost parts of the villi. Somewhat lower levels are observed in lower crypts, villus cores, and muscularis layer; no phospho-p38 was detected in the most of the villus epithelium (Fig. 8D, upper). Antibiotic treatment markedly reduces the levels of phospho-p38 (Fig. 8B; Fig. 8D, middle). Phospho-p38 staining is specific because it is not observed upon substitution of the primary Ab with pre-immune serum (Fig. 8D, lower). Thus, p38 activation in the epithelium of adult small intestine is largely limited to the crypts, and, like A20 expression, it positively correlates with bacterial load of the lumen.
To examine causal relationship between A20 induction and establishment of tolerance to LPS in the intestinal epithelium, we used siRNA transfection. Mucosal application of siRNA-cationic lipid complexes allows effective gene silencing in the epithelium in vivo (48, 49). We reasoned that this approach could be also applied ex vivo. Mucosal scrapings or segments from small intestines of term rat fetuses were transfected with A20 or LEU2 siRNAs complexed with Lipofectamine. Tissue samples were then pulsed with LPS, washed, incubated without LPS, and subjected to the second LPS challenge. Phosphorylation of p38 and levels of A20 at various time points were examined by Western blots, real time RT-PCR and IF. Because mucosal scrapings contain significant numbers of non-enterocyte cell types, Western blot results were interpreted in conjunction with immunofluorescence data. The latter identify enterocytes as the cell type where expression of A20 or phosphorylation of p38 occurs. Transfection with A20 siRNA, but not with LEU2 siRNA, prevents LPS-induced expression of A20 (Fig. 9; Fig. S5). Initial LPS challenge increases levels of phospho-p38 (Fig. 9), with bulk of the signal localized to the nuclei (Fig. S5). Samples transfected with A20 siRNA, but not LEU2 siRNA, displayed p38 activation following the second LPS challenge (Fig. 9, Fig. S5). Thus, A20 induction is required for the establishment of tolerance to LPS in the epithelium of the neonatal small intestine ex vivo.
Our findings provide multiple lines of evidence for the key role of the ubiquitin-editing enzyme A20 in the development of tolerance to TLR ligands in the small intestine. Unlike many other innate immune response proteins, A20 is induced by LPS and enteric bacteria in cultured enterocytes and in the neonatal ileal epithelium. LPS-induced A20 expression temporally coincides with the establishment of hyporesponsiveness to repeated stimulation with LPS. Ectopic expression of A20 dramatically decreases LPS-induced activation of p38, c-Jun, and NF-kB, whereas siRNA silencing of A20 prevents desensitization to LPS, indicating that A20 is necessary and sufficient for the development of tolerance to LPS. High levels of A20 attained by exposure to LPS or transfection with A20 plasmid block responses to LPS and CpG DNA, but not to IL-17 or UV, therefore A20 specifically inhibits TLR signaling. A20 levels are high in bacteria-colonized intestine, but low in naive intestinal epithelium; they are also high in bacteria-exposed villi, crypt openings, and upper crypts, but low in protected lower crypts. Moreover, A20 levels in the small intestine decrease following antibiotic treatment. Although our data do not rule out induction of A20 during postnatal development of the intestine or enterocyte differentiation, they argue for TLR ligand-induced expression. Prominent localization of A20 to the apical sub-membrane of villus enterocytes in conventionally housed adult rodents is consistent with the role of A20 in the suppression of TLR signaling at the luminal interface. The facts that A20-deficient mice develop severe gut inflammation early in life (36), and that this inflammation can be alleviated by antibiotics or knockout of the TLR signaling mediator myeloid differentiation factor MyD88 (37) further support the key role of A20 in the intestinal tolerance to the intestinal microbiota.
The critical contribution of A20 into the intestinal tolerance to LPS raises an important question about the relationship between A20 and other known inhibitors of TLR signaling in the gut. One possibility is that some of the mechanisms, for example post-translational down regulation of IRAK-1 (9), depend on A20. Indeed, IRAK-1 is regulated by K63 polyubiquitination (50), and thus may be an A20 target. Another possibility is that A20's role is limited to the small intestine, whereas the large intestine, with its higher bacterial burden, utilizes different mechanisms. Because Tollip is mostly expressed in the large intestine and is induced by LPS in colonic enterocyte cell lines (15), but not in the small-intestine-derived IEC-6 (this study), or m-ICcl2 (9), tolerance via induction of Tollip may be colon-specific. It is also possible that other negative regulators of inflammatory signaling are redundant with A20. Examples of such mechanisms could be down regulation of TLR in villus enterocytes, ligand-induced TLR internalization, NF-κB-mediated re-synthesis of IκB, or degradation of MD-2 by trypsin. Furthermore, intestinal tolerance to bacteria involves a variety of specialized innate immune cells with their specific regulatory mechanisms (51). It is not surprising that such critical function as tolerance to commensal bacteria is supported by multiple mechanisms.
Rapid development of tolerance to LPS via induced expression of A20 appears to be enterocyte-specific. Indeed, development of tolerance to LPS in macrophages takes longer time and does not require A20 (28). In addition to its cytosolic localization, A20 localizes to late endocytic vesicles in HEK293, COS-7 and HeLa cells (42), whereas in adult villus enterocytes it localizes to apical sub-membrane and Golgi bodies. Golgi localization in villus enterocytes and association with particulate structures in IEC-6 cells may reflect inclusion of A20 into membrane-associated TLR signaling complexes. Because A20 lacks membrane-spanning domains or lipid modification sites, its association with membranes likely depends on interaction with membrane-tethered proteins. One such protein could be the A20 target TRAF6, which is associated with membranes in unstimulated cells (43). The unique physiology of A20 in the intestine may be relevant to the fact that enterocytes are the only cell type in the body that is continuously exposed to high concentrations of bacteria under normal conditions. Accordingly, in enterocytes A20 may facilitate establishment and maintenance of tolerance to commensal bacteria, whereas in other cell types it may blunt responses to inflammatory cytokines and invading pathogens.
In adult villus enterocytes, A20 localizes to apical sub-membrane and Golgi bodies. By contrast, enterocytes in the upper parts of the crypts display diffuse A20, which resembles localization in IEC-6 cells. Crypt-villus differences in A20 localization may be related to the establishment of apical-basolateral polarity during differentiation of crypt precursors into villus enterocytes. Interestingly, TLR4 also localizes diffusely in undifferentiated enterocytes, but is concentrated at the apical pole following differentiation (52). This observation is consistent with the proposed association of A20 with TLR signaling complexes. Diffuse localization of A20 in the epithelium of 3-4 day rodents may reflect incomplete epithelial differentiation at the early postnatal stages.
LPS-induced expression of A20, localization patterns of this protein along the crypt-villus axis, and sequestration of bacteria-dependent activation of p38 to upper crypts and crypt openings suggest a mechanism whereby the intestinal epithelium may sense commensal bacteria while avoiding dramatic inflammation. Because bacteria have limited access to lower parts of the crypts, emerging enterocyte precursors may be sensitive to TLR ligands. As these cells progress towards crypt openings, their exposure to luminal bacteria may activate TLR signaling, leading to production of inflammatory factors necessary for the gut homeostasis. However, concomitant induction of A20 would rapidly block TLR signaling. Consequentially, epithelial response to luminal bacteria may be benign because it is limited spatially to upper crypts and crypt openings, and temporally to the time needed to induce A20 or other negative regulators. Since t1/2 of the A20 protein is comparable to the life span of villus enterocytes, the latter maintain tolerance to LPS until they shed off at the villus tip. The continued renewal of the intestinal epithelium thus may allow both the limited homeostatic response to commensal bacteria and the suppression of TLR signaling at the luminal interface. Because the whole intestinal epithelium is naive with regard to TLR signaling at birth, the onset of bacterial colonization may involve a quasi-inflammatory episode, which is expected to resolve fast due to rapid down regulation of the innate immune responses in enterocytes. Dynamic bacteria-induced development of tolerance in initially sensitive epithelial cells, as opposed to constitutive unresponsiveness, is supported by the fact that enterocytes isolated from adult germ-free animals are sensitive to stimulation with bacteria (53).
Contribution of A20 into the establishment of dynamic epithelial tolerance to bacteria warrants investigation of the role this protein in the pathogenesis of the intestinal disorders characterized by abnormal hypersensitivity to the commensals.
We thank Dr. Averil Ma for A20 Ab, Dr. Hua Xu for RIE-1 cells, Drs. Timothy Billiar, Mitchell Fink, Wei Shi, David Warburton for critiques, and Kerstin Goth for expert technical assistance.
Disclosures: The authors have no financial conflict of interest.
1This work was supported by National Institutes of Health Grants 5RO1AI014032 and 5R01AI049473 to H.R.F.