human serum albumin
Mice which lack the p50 subunit of NF-κB and are heterozygous for the p65 subunit (3X mice), are exquisitely sensitive to LPS-induced shock. Here, we demonstrate that prior to becoming moribund, 3X mice challenged with LPS develop a profound enteropathy. The enteropathy is characterized by defects in intestinal barrier function, increased epithelial apoptosis, and deregulated intestinal cytokine gene expression. The defect that sensitizes 3X mice to LPS-induced enteropathy is located within the innate immune compartment, as LPS induced similar findings in 3X mice lacking lymphocytes (3X/RAG). TNF-α depletion ameliorated the ability of LPS to induce pathology and TNF-α was able to independently induce similar findings, suggesting that TNF-α plays a critical role in the development of LPS-induced pathology in these mice. These data highlight that NF-κB subunits have essential functions in regulating intestinal homeostasis during acute inflammation.
The transcription factor NF-κB regulates expression of more than 500 genes involved in immune responses, inflammation, cell survival, and cell proliferation. In its active DNA-binding form NF-κB is a dimer, composed of members of the Rel family 1, 2. Five mammalian proteins of this family are known: p65, Rel B, c-Rel, NF-κB1 (p50 and its precursor p105), and NF-κB2 (p52 and its precursor p100). The NF-κB dimers are primarily retained in the cytoplasm by association with IκB. Phosphorylation and degradation of IκB releases NF-κB and enables its translocation to the nucleus, where it regulates inflammatory transcriptional programs.
Various combinations of NF-κB have different transcriptional abilities. While many heterodimers have the ability to promote transcription, it is now clear than certain subunits can also exhibit inhibitory activity as well. For example, p50 represses expression of IL-2, TNF-α, and IFN-dependent genes in different cell types including lymphocytes and macrophages 3–5. Further, we have shown that mice that lack the p50 subunit and are also heterozygous for p65 (3X mice) are sensitive to the development of microflora-induced colitis. Macrophages derived from these animals express higher levels of inflammatory cytokines in response to challenge 6. We and others have shown that mice lacking p50 are sensitive to LPS-induced shock and this is strongly aggravated in 3X mice 7, 8. Shock in these animals is characterized by increased levels of TNF-α and IFN-γ within the serum, and is inhibited by the depletion of TNF-α . Initial studies demonstrated that LPS induced the accumulation of moderate edema within the gastrointestinal tract, but no further pathological lesion was identified.
The development of enteropathy is strongly associated with septic shock. For example, clinical correlative studies have established that defective intestinal barrier function is frequently observed in septic patients 9, 10. Here, we characterize bowel pathology after LPS treatment of p50 and 3X mice. We demonstrate that LPS induces a profound enteropathy in NF-κB-deficient animals that is associated with dramatic defects in intestinal barrier function, deregulated intestinal inflammatory responses, and epithelial apoptosis. This pathology depends on the deregulated response to TNF-α in the absence of NF-κB subunits p50 and p65.
NF-κB-deficient mice demonstrate defects in intestinal permeability after LPS challenge
Our previous results showed that LPS challenge leads to vascular congestion and edema in 3X mice but not in WT mice 7. To further evaluate whether LPS induces vascular leak in 3X mice, the low-molecular-weight tracer Evans Blue was administered after LPS challenge. While little tracer was identified within the kidneys, livers, and spleens of both 3X and WT animals 4 h after LPS challenge, we found significantly higher levels of the tracer within the small intestine and colons of 3X mice than in WT mice (Fig. 1A). The tracer was found predominantly accumulated within the lumen of the small intestine, rather than within the intestinal wall (Fig. 1B).
To test the role of lymphocytes in this phenotype, we compared the ability of LPS to induce tracer accumulation within the intestinal lumen of RAG, p50/RAG, and 3X/RAG mice (Fig. 1C). We found that similar to mice with lymphocytes, there were significant elevations of tracer within the intestinal lumen of 3X/RAG mice, while little was observed within the intestinal lumen of control RAG mice. Interestingly, intermediate levels of tracer were found within the intestinal lumen of p50 and p50/RAG mice, suggesting that heterozygosity at the p65 locus significantly exacerbated the leakage of Evans Blue into the intestinal lumen (Fig. 1B). These results suggest that there is a defect within the innate immune compartment of NF-κB-deficient mice that allows leakage of low-molecular-weight compounds into the intestinal lumen.
To determine whether high-molecular-weight components also leaked into the small intestine of 3X mice after LPS challenge, mice were challenged with LPS and then injected i.v. with FITC-labeled human serum albumin (HSA) as a tracer. Tracer levels were then monitored within the serum and small bowel lumen by Western blotting with an anti-FITC Ab. Abundant tracer was identified with the ileal lumen of 3X mice after LPS challenge but little tracer was observed within the ileal lumen of p50 or WT mice (Fig. 2A). Comparing the ratio of tracer identified within the lumen and serum confirmed these observations (Fig. 2B).
Similar to what we observed in animals that contained lymphocytes, there was little tracer found in the ileal lumen after LPS challenge of RAG mice and increased levels observed in 3X/RAG mice (Fig. 2C, D). However, unlike lymphocyte-sufficient p50-deficient mice, we did observe increased accumulation of tracer within the ileal lumen of p50/RAG mice. This suggests that the presence of lymphocytes may exert a mild inhibitory function on leakage of high-molecular-weight compounds after LPS challenge. Overall, the data indicated that NF-κB subunits prevent the leaking of high-molecular-weight material into the intestinal lumen after LPS challenge.
Elevated levels of apoptosis accompany the defects in intestinal barrier function
It is speculated that apoptosis-dependent mechanisms may contribute to abnormal mucosal permeability in patients with Crohn's disease 11–13. To determine whether morphological changes accompany the altered intestinal barrier function in 3X/RAG mice after LPS challenge, sections from small bowel were analyzed by H&E staining (Fig. 3A). Increased levels of epithelial apoptosis were evident within the intestinal crypts of the small bowel in 3X/RAG animals after LPS challenge. The finding was confirmed by immunohistochemical staining for activated caspase-3 (Fig. 3B). LPS induced epithelial apoptosis in the small intestine of 3X/RAG mice, moderate levels of apoptosis in p50/RAG animals, and only individual apoptotic cells in WT and RAG mice. To compare the degree of apoptosis between genotypes, we quantified the number of apoptotic cells per gland within the ileum for each group (Fig. 3C). The data showed that 3X/RAG mice had higher numbers of apoptotic cells per gland than p50/RAG mice, suggesting that heterozygosity at the p65 locus exacerbates apoptosis observed in the p50-deficient animals. The increased levels of apoptosis observed in 3X/RAG mice compared to p50/RAG mice correlate with the higher permeability observed in 3X/RAG mice. Similar results were obtained in mice that contained lymphocytes.
TNF-α is required for development of enteropathy in 3X/RAG mice
LPS challenge induces markedly elevated serum levels of TNF-α in 3X and p50 mice compared to WT mice 7. We found similar differences in mice on the RAG background. As it has been suggested that TNF-α modulates the state of endothelial gap junctions and promotes permeability 14, 15, we questioned whether TNF-α was responsible for the changes in intestinal permeability observed in 3X/RAG mice. Pretreatment of 3X/RAG mice with anti-TNF-α Ab prior to LPS challenge prevented the development of ileal leak and epithelial crypt apoptosis (Fig. 4A–C). This suggests that TNF-α is required for alterations of intestinal permeability and the development of crypt apoptosis after LPS challenge of 3X/RAG mice.
To determine whether TNF-α challenge of 3X/RAG mice is sufficient to induce alterations in intestinal homeostasis, RAG, p50/RAG, p65+/–/RAG, and 3X/RAG mice were challenged with TNF-α by i.p. injection. TNF-α challenge induced epithelial crypt apoptosis in both 3X/RAG mice and p50/RAG mice, while little was observed in RAG mice or p65+/–/RAG mice (Fig. 5A). Importantly, the number of apoptotic cells per gland were elevated in 3X/RAG tissues when compared to p50/RAG tissues (Fig. 5B). TNF-α challenge also resulted in elevated permeability for FITC-HSA in the ileum of 3X/RAG but not in p65+/–/RAG or RAG mice (Fig. 5C, D). The p50/RAG animals had an intermediate phenotype with some leak present. The data imply that TNF-α challenge induced enteropathy in p50/RAG and 3X/RAG, but not in p65+/–/RAG or RAG mice.
Altered gene expression in the small intestine of 3X/RAG mice
Our results suggest a marked alteration in the response of the small bowel of 3X/RAG mice to LPS or TNF-α challenges compared to the response observed in RAG mice. To determine whether these differences were reflected at the level of gene expression, microarray-based analysis was used to compare gene expression within the ileal tissue of 3X and WT animals (data not shown). Out of 14 000 screened, 130 genes were up-regulated more than fourfold in LPS-challenged 3X mice. The following functional categories could be distinguished: IFN-stimulated genes with apoptotic function (TRAIL, IRF-1, RNAseL, galectin 9, OAS-1, IFI-204, schlafen-1, USP-18), IFN-stimulated genes with host defense activity (IL-1β, IL-6, IP-10), and transcription factors (STAT-1, STAT-2, STAT-3, IRF-7, IGR-1). Interestingly, some of these genes were also up-regulated in the tissues from non-challenged mice.
To expand on this we used RT-PCR to examine expression of a set of IFN-inducible genes in RAG and 3X/RAG mice prior to or after stimulation with TNF. While TNF-α was a poor inducer of IL-6, IRF-1, and IP-10 expression within the small bowel of RAG mice, it was a strong inducer of IL-6 and IRF-1 in the small bowel of 3X/RAG and p50/RAG mice (Fig. 6 and data not shown). This observation suggests that the transcriptional response to TNF-α within the small bowel is deregulated in 3X/RAG mice. Furthermore, while there was little induction of IL-12 p40, IFN-β, IFI-204, and schlafen-1 gene expression in response to TNF-α in either RAG or 3X/RAG mice, these genes were expressed at higher levels in 3X/RAG mice than in RAG mice in the absence of stimulation. As this set of genes (including IFN-β itself) has been described as part of the IFN response, and IFN has been reported to regulate IL-6, IRF-1, and IP-10 expression as well, these data raise the possibility of an ongoing IFN-dependent immune response in these animals.
Here, we describe that the NF-κB subunits p50 and p65 play a critical role in protecting mice from LPS-induced enteropathy. Pathology observed in LPS-challenged NF-κB-deficient animals was characterized by defects in intestinal barrier function, apoptosis, and regulation of inflammatory gene expression. However, there was no evidence of infiltration of inflammatory cells on H&E staining, and an absence of neutrophil infiltration was confirmed by immunohistochemistry for Gr-1 (data not shown). TNF-α was both necessary and sufficient for the development of LPS-induced pathology in 3X/RAG mice. Treatment with anti-TNF-α Ab protected animals from defects in barrier function and epithelial apoptosis. This correlated with down-regulation of pro-inflammatory responses (data not shown). Further, TNF-α challenge was able to induce defects in barrier function, epithelial apoptosis, and elevated inflammatory responses in p50/RAG and 3X/RAG animals, but not in RAG mice, suggesting that the absence of NF-κB p50 and p65 subunits sensitize to TNF-α-mediated effects and lead to development of enteropathy.
In vivo experiments provided evidence that inhibition of TNF-α restores barrier function induced by anti-CD3 treatment 16, 17. It is possible that TNF-α may modify barrier function through induction of apoptosis. In vitro experiments demonstrated that apoptosis contributes as much as 50% to the TNF-α-induced permeability changes in epithelial monolayers 18–23. We found that TNF challenge of 3X/RAG mice resulted in massive epithelial crypt apoptosis, which correlated well with defects in barrier function. Therefore, it is likely that the epithelial apoptosis is a significant contributor to the TNF-α-induced defects in barrier function.
The role of the NF-κB subunit p65 in preventing TNF-α-induced apoptosis has been well studied. Mice that lack p65 die at embryonic day 14 due to TNF-α-dependent hepatocyte apoptosis, and murine embryonic fibroblasts lacking p65 are highly sensitive to TNF-α-induced apoptosis 24–26. In contrast to the role for p65 in protecting from TNF-α-induced apoptosis, there has been little evidence to support a role for p50 in protecting from TNF. Hepatocytes lacking both p50 and p65 undergo apoptosis at embryonic day 12.5, 1.5 days earlier than mice lacking p65 alone, suggesting some low-level redundancy between subunits 26. However, the data reported here underline that the absence of p50 is the primary factor that sensitizes mice to TNF-α-induced apoptosis of enterocytes and that this is exacerbated by heterozygosity at the p65 locus.
Further, our data demonstrate that, while both, p50/RAG and 3X/RAG mice, did not display signs of hepatocellular apoptosis, abundant apoptosis is seen in the epithelial crypts of the small intestine. The small intestinal crypts are sites of active proliferation. Similar to bone marrow cells, the enterocytes within the crypts are extremely vulnerable to irradiation-induced damage. Exposure to γ-irradiation results in massive crypt cell apoptosis 27. These challenges reportedly require NF-κB signaling to exert protection from apoptosis 28. While the mechanisms of protection are not fully understood, it has been suggested that the increased apoptosis of IKK-β-deficient enterocytes is accompanied by elevated activation of p53 and decreased expression of anti-apoptotic bcl-2 family members. Specifically, the anti-apoptotic function of NF-κB p65 is described to be at a transcriptional level by regulating expression of cIAP or bcl family members 29.
Surprisingly, our analysis of LPS- or TNF-α-induced gene expression pattern did not reveal differences in NF-κB-dependent anti-apoptotic gene expression including cIAP-1, cIAP-2, bcl-xl, p53, or bcl-2 in the absence of p50 and/or p65 (data not shown). In addition, no differences in the mRNA levels for other bcl-2 family members including bcl/w, bcl-x, bax, bfl1, bak, or bad were observed when 3X/RAG and RAG tissues were compared (data not shown). While our inability to observe differences could simply be a result of an inability to detect differences in a small percentage of cells in total ileal RNA, an alternative hypothesis is that the protective role for p50 is fundamentally different from what has been observed previously for p65.
We observed that there was markedly higher expression of several IFN-dependent genes including IL-6 and IRF-1 in the ilea of 3X/RAG mice than in RAG mice after TNF-α challenge 30, 31. Interestingly, some of these differences may be the result of altered gene expression in epithelial cells themselves, as we detected higher expression of IL-6 in epithelium stripped from the ileum of challenged 3X mice than from the ileum of challenged control mice (data not shown). In addition, while we identified several IFN-inducible genes that were not induced by TNF-α (Fig. 6), there appeared to be small, but reproducible increases in the baseline expression of these genes in 3X/RAG mice compared to RAG mice, including IFN-β itself. These results suggest that there might be differences in baseline as well as induced levels of IFN-dependent genes in 3X mice.
Interestingly, previous reports have indicated that p50 can inhibit expression of certain genes after stimulation of fibroblasts with IFN 32, 33. It has been reported that p50-deficient mice are resistant to infection with EMCV and that this associated with elevated IFN-dependent gene expression 29. In addition, it was shown that p50-deficient fibroblasts are resistant to infection with influenza 33. Taken together these results suggest that p50 may be playing an important role in limiting the expression of IFN-inducible genes.
While the role of aberrant IFN-inducible gene expression in the increased sensitivity to TNF-α-induced apoptosis observed in enterocytes of 3X mice remains to be determined, it is tempting to speculate that the altered IFN responses might lead to enterocyte apoptosis. Recent studies by Takaoka et al.34 highlighted the significance of IFN-α/β in induction of p53 expression. The data showed increased concentrations of p53 protein after IFN-β stimulation. The enhanced accumulation of p53 in cells treated with IFN-β rendered them more sensitive to apoptosis induced by γ-ray irradiation. The other evidence is the demonstration that the apoptotic response to DNA-damaging agent 5-fluorouracil is enhanced by IFN-β treatment by 50%, suggesting that the anti-tumor activity of IFN-β is mediated at least partly by p53. In addition, the apoptosis elicited by γ-irradiation which targets specifically cells of the crypts is p53-dependent 35.
We have defined previously unanticipated roles for p50 and p65 in maintaining the integrity of bowel epithelium after inflammatory challenge. The NF-κB subunits p50 and p65 have critical functions in promoting epithelial survival and ensuring appropriate barrier function after LPS challenge. In addition, TNF-α is not only necessary, but also sufficient to induce enteropathy in the absence of p50 and p65.
Materials and methods
All experiments were approved by the Harvard Medical School Standing Committee on Animals. The WT (129S6/SvEv), p50–/– (p50), p65+/–;p50–/– (3X), RAG-2–/– (RAG), p50–/–;RAG-2–/– (p50/RAG), p65+/–;RAG-2–/–, and p50–/–;p65+/–;RAG-2–/– (3X/RAG) strains were described previously 36. All animals were backcrossed on the 129S6/SvEvTac background for at least six generations and were maintained in conditions free of known Helicobacter species.
LPS and TNF-α treatments
Mice were challenged with 2 mg/kg of LPS Escherichia coli 0111:B4 (Sigma, Saint Louis, MO) i.p. Segments (2 cm) of the ileal tissue, proximal to the cecum, were harvested for analysis 4 h after challenge. For the TNF-α treatment experiments, mice were treated with 100 ng/animal of recombinant murine TNF-α (Roche Applied Science, Mannheim, Germany) and were sacrificed 3 h post-treatment. Tissues were collected and paraffin-embedded, or snap-frozen in liquid nitrogen for subsequent RNA analysis.
TNF-α depletion experiments
Mice were injected with 200 μg anti-TNF-α Ab (clone XT-3; BioExpress, West Lebanon, NH) 1 h before LPS challenge. Four hours later animals were sacrificed; tissues were harvested, and paraffin-embedded for histology experiments or snap-frozen for RNA analysis.
Mice received 200 μL of a solution containing 0.15% Evans Blue (Alfa Aesar, Ward Hill, MA) mixed with 2 mg/mL ovalbumin (Sigma) by i.v. injection 30 min after the LPS injection. Four hours after the LPS challenge, organs including livers, spleens, kidneys, colons, and small intestinal segments were harvested. Evans Blue that had accumulated in the tissue was extracted for 3 days with 5 μL formamide (Acros Organics, Morris Plains, NJ) per mg of tissue, and then the dye concentration was determined by spectrophotometry at OD 610 37.
FITC-HSA permeability assays
Thirty minutes after LPS or TNF-α challenge, mice received 200 μg of FITC-HSA (Sigma) by i.v. injection. Total intestinal content and serum were collected 4 h after challenge. The intestinal content was mixed with a set volume of PBS containing protease inhibitors (Roche Diagnostics, Mannheim, Germany) for each mg of sample. Samples, consisting of 1 μL of serum or 4 μL of intestinal content, were fractionated by SDS-polyacrylamide gel electrophoresis on 4–15% Tris-HCl gradient gels (Bio-Rad, Hercules, CA), and analyzed by Western blotting. The blots were developed with rabbit anti-FITC polyclonal Ab (Zymed, San Francisco, CA) at 1:1200 dilution, followed by goat anti-rabbit polyclonal Ab (Santa Cruz Biotechnology, Santa Cruz, CA) at 1:1000. The densities of the bands were quantified by histogram analysis using Adobe Photoshop, and the ratios of the FITC-HSA bands in the lumen versus FITC-HSA bands in the serum were calculated.
Sections (4 µm) of formalin-fixed, paraffin-embedded intestinal samples were cut and stained for activated cleaved caspase-3. The anti-activated caspase-3 Ab (Cell Signaling Technology, Beverly, CA) was used at 1:150 dilution in PBS. The staining was visualized with Vectastain ABC kit (Vector Laboratories, Burlingame, CA). Sections were counterstained with 20% hematoxylin (Fischer, MA). Apoptotic index was calculated by determining the average number of apoptotic cells per gland. Ten adjacent glands in a section were used to determine the average number of apoptotic cells. Several consecutive sections from each animal were used to ensure consistency.
Total RNA (2–5 μg) were DNAse-treated (Invitrogen, Carlsbad, CA) and reverse-transcribed with TaqMan Reverse Transcription Reagents (Applied Biosystems, Foster City, CA). The levels of IL-6, IL-12 p40, IFN-β, IFI-204, IRF-1, and schlafen-1 were determined with TaqMan gene expression assays (Applied Biosystems) following the manufacturer's instructions, employing Prism Sequence Detection System 7700 (Applied Biosystems).
Apoptotic indexes were compared using the one-way ANOVA for parametric data (Prism 4 statistical analysis software package). Unpaired two-tailed t-test was used to compare inflammatory gene expression data. Differences were considered statistically significant when p<0.05.
This work was supported by National Institutes of Health grant A152267 (B.H.H.) and CCFA William and Shelby Modell Family Foundation Senior Research Grant (B.H.H.).