Bile Mediates Intestinal Pathology in Endotoxemia in Rats

Abstract

Intestinal pathology frequently accompanies experimental endotoxic shock and is mediated by proinflammatory cytokines. Our hypotheses are that hepatobiliary factors operating from the luminal side of the gut make a major contribution to this damage and that tumor necrosis factor alpha (TNF-α) is involved in the pathology. We treated rats with lipopolysaccharide (LPS) intravenously and found that external drainage of bile totally protected the gastrointestinal tract, macroscopically and microscopically, 4 h after LPS administration and dramatically improved survival of the animals for 48 h after LPS administration. The concentration of TNF-α in bile increased markedly after LPS administration and was over 30 times higher in bile than in serum. Tissue damage and the biliary TNF-α response were abrogated when animals were pretreated with gadolinium chloride to eliminate Kupffer cells. TNF-α infusion into the duodenal lumen caused intestinal damage similar to that elicited by intravenous LPS. In rats treated with LPS, survival was significantly increased during the first 36 h in animals given an infusion of anti-TNF-α antibody into the duodenum. These results demonstrate that in endotoxemia, intestinal damage is mediated by factors derived from the bile. The findings indicate that luminally acting TNF-α contributes to the intestinal damage.

Septic shock associated with infection by gram-negative bacteria is a common problem in hospitalized patients. Intestinal lesions, including hemorrhage and diarrhea, are a prominent feature of endotoxemia. Several bacterial species can cause such lesions, and lipopolysaccharide (LPS) is an important bacterial product that initiates these effects (19). Antibiotics are not highly effective, and this has driven investigations on the mechanisms of pathogenesis. The major features of LPS-mediated shock are now considered to be elicited by a range of endogenous proinflammatory mediators released in response to LPS. Most attention has focused on tumor necrosis factor-α (TNF-α) (35), interleukin 1 (IL-1) (25), and platelet-activating factor (9). How these agents might network to produce the various effects is not known. TNF-α has attracted particular attention because neutralizing antibodies to TNF-α inhibit the toxic effects of LPS, and administration of TNF-α alone elicits the features of endotoxemia (2, 34). Furthermore, in mice lacking functional genes for TNF-α or for the 55-kDa TNF receptor, endotoxic shock was attenuated (26, 30).

It is implicit in studies on the intestinal effects of proinflammatory cytokines that these agents are derived from the interstitial aspect of the epithelium (32). However, there are indications that biliary molecules (other than bile salts) have functional activity in the gut. For instance, epidermal growth factor has been shown to be present in bile (36) and to be functional from within the lumen (15). Also, hepatobiliary delivery of immunoglobulins has an important role in the protection of the gut (11, 18). Recently, we have reported that factors in bile regulate the expression of major histocompatibility complex class II molecules on intestinal epithelium (7).

Our hypothesis is that intestinal damage in endotoxic shock results from the action of luminal agents. Since LPS is removed from the blood by the liver (24), and given that Kupffer cells can produce the relevant cytokines (4) and that products synthesized in the liver are likely to appear in bile in at least trace amounts (22), we reasoned that LPS-induced hepatobiliary factors could directly interact with and cause damage to the gut. To test this idea, we have examined the effect of external biliary drainage on intestinal integrity. Rats were given a high dose of LPS with the capacity to induce severe intestinal damage, similar to that observed in sepsis. We report that external drainage of bile abolished the toxic effects of LPS on the intestine. Furthermore, TNF-α has been detected in normal bile (12, 29). We therefore hypothesized that secretion of increased quantities of TNF-α into the bile, and delivery to the duodenum, may contribute to the intestinal pathology in endotoxemia. In this study, we investigated bile for the presence of TNF-α bioactivity. We also describe the effects of administration of TNF-α into the duodenums of normal rats and the effects of instillation of anti-TNF-α antibodies into the duodenums of animals treated with LPS.

MATERIALS AND METHODS

Animals.

Conventionally raised male Australian Albino Wistar rats approximately 10 weeks old and weighing 300 to 320 g were used in all experiments and were obtained from Combined Universities Laboratory Animal Supply of the University of New South Wales. All experiments were approved by the Animal Care and Ethics Committee of the University of New South Wales.

Surgical procedures.

Animals were fasted overnight with free access to water. Bile duct cannulation (BDC) and occlusion were carried out under ether anesthesia as described by Lambert (17). Briefly, for BDC, a laparotomy was performed and a cannula (internal diameter, 0.4 mm; outside diameter, 0.8 mm; length, ∼20 cm) was inserted into the bile duct and secured with the other end passing out through the flank of the rat, allowing bile to be collected externally, to prevent bile from entering the gut. In some animals, bile was deviated to the ileum; briefly, following a laparotomy, one end of a cannula was inserted into the bile duct, while the other end was inserted into the ileum, approximately 15 cm proximal to the cecum. The abdominal incision was then sutured, and rats were held in restraining cages. Control animals were subjected to ether anesthesia and sham laparotomy. Rats from each treatment group and control rats were then injected intravenously (i.v.) with phenol-extracted LPS from Escherichia coli serotype 0111:B4 (catalog no. L2630; Sigma, St. Louis, Mo.). A dose of 15 mg/kg was used except in the survival studies, where the dose was 35 mg/kg. Sera were prepared from tail vein blood, and bile was collected from the BDC group, with both being stored at −20°C. At 4 h after LPS administration, rats were euthanized and the macroscopic appearance of the whole small intestine was assessed.

One- to two-centimeter segments of the duodenum, jejunum, and ileum from each rat were fixed in 10% buffered formalin and mounted in paraffin, and 5-μm sections were prepared and stained with Harris hematoxylin and eosin. In some experiments, rats were treated i.v. with 25 mg of gadolinium(III) chloride hexahydrate (Aldrich Chemicals, Milwaukee, Wis.) per kg 24 h prior to surgery. In preliminary experiments, the optimal dose and time of administration of gadolinium chloride were determined by assessment of loss of Kupffer cell labeling with India ink. In some experiments, survival after LPS injection was determined. Animals were observed hourly for 48 or 72 h following i.v. injection of LPS. Euthanasia was performed when animals became moribund as determined by hyperventilation or loss of righting reflex.

TNF-α infusion.

Human recombinant TNF-α (hrTNF-α) was kindly provided by Peptide Technology, Sydney, Australia. This preparation contained less than 150 pg of LPS per mg. Human TNF-α has been documented to be biologically active on rat cells (27). All hrTNF-α was diluted in pyrogen-free sterile saline (Astra Pharmaceuticals). Under anesthesia, a laparotomy was performed and the duodenal wall was pierced with a 23-gauge needle. A cannula of the same size used for BDC was then inserted into the duodenal lumen and secured in place with sutures. hrTNF-α solutions were infused into the duodenum at the rate of 1 ml per h for 4 h using a peristaltic pump (Pharmacia, Uppsala, Sweden). hrTNF-α concentrations were increased up to 2 h and then reduced, to reflect the pattern of the TNF-α concentration in bile following i.v. challenge with LPS. One group of rats was infused with hrTNF-α at the following times and concentrations: 0 to 30 min, 1 μg/ml; 30 to 60 min, 5 μg/ml; 60 to 120 min, 10 μg/ml; 120 to 180 min, 5 μg/ml; and 180 to 240 min, 1 μg/ml. The control group was infused with pyrogen-free sterile saline. The rats were killed at the end of the infusion. The macroscopic appearance of the whole small intestine was examined, and tissues were prepared for microscopic examination as described above.

Anti-TNF-α infusion.

Rabbit anti-mouse TNF-α antiserum was obtained from Genzyme (Cambridge, Mass.) and diluted in pyrogen-free sterile saline. This preparation has been shown to neutralize rat TNF (31). Normal rabbit serum was used as a control. A duodenal cannula was inserted as described above. Each rat was then injected with LPS i.v. Immediately afterwards, infusion of sera into the duodenum was commenced. Sera were infused at the rate of 1 ml per h for 4 h using a peristaltic pump. All rats were infused with the serum dilutions as follows: 0 to 30 min, 1:10 dilution; 30 to 60 min, 1:5; 60 to 120 min, 1:2; 120 to 180 min, 1:5; and 180 to 240 min, 1:10. Rats were euthanized at the end of the 4-h infusion. The macroscopic appearance of the whole small intestine was examined, and tissues were prepared for microscopic examination as described above.

Measurement of intestinal fluid volume.

The small intestine of each rat was tied at each end and freed from the abdominal tissues, and the contents were expressed by gentle squeezing from the duodenum to the ileum. The solid material, including the mucus, was removed by centrifugation (500 × g, 15 min), and the volume of clear fluid was measured.

Bioassay for TNF-α.

Biles and sera were assayed by inhibition of WEHI 164 (clone 13) cell proliferation (6). Proliferation was measured by metabolism of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (21). Recombinant murine TNF-α (Genzyme) was used to generate a standard curve, and all samples were assayed in the presence and absence of a neutralizing polyclonal rabbit anti-mouse TNF-α antiserum (Genzyme) at a dilution of 1:10. Negative controls were wells in which cells remained untreated throughout the incubation period of the assays (20 h, 37°C). Dilution series were used to determine the values of TNF-α. The lower level of detection of the assay was 0.005 ng/ml.

Statistics.

Survival data were drawn as Kaplan-Meier plots, and probability (P) values were determined by the Breslow-Gehan-Wilcoxon test using Statview 3.5 software (Abacus Concepts, Berkeley, Calif.). Other data are shown as means ± standard errors (SEs). Significance was accepted at the 5% level.

RESULTS

Effect of external drainage of bile.

We investigated the effects of BDC and external drainage of the bile in rats treated with LPS. Rats were either sham operated or subjected to BDC and then were injected with a lethal dose of LPS i.v. Survival in the BDC group was highly significantly greater than that in the controls (P < 0.0001). None of the animals in the control group survived for more than 48 h, whereas in the BDC group, 10 of the 13 animals survived for this length of time (Fig. 1). In the BDC group, the first euthanasia was performed at 20 h, by which time only 5 of the 13 control animals were still alive.

FIG. 1.

Kaplan-Meier plots of the effect of BDC on survival. Animals were either sham operated or subjected to BDC and then were challenged with LPS at time zero. In the BDC group, saline was infused intraduodenally for the first 24 h. The bile duct was then occluded, and the saline infusion ceased. There were 13 animals in each group.

At 4 h after inoculation with LPS, macroscopic examination indicated no hyperemia or fluid accumulation in rats subjected to BDC. By contrast, in animals treated with LPS and not subjected to BDC, there was marked hyperemia and fluid accumulation within the small bowel (Fig. 2). In LPS-treated rats, bile duct occlusion had effects similar to those of BDC, with a substantial reduction in hyperemia and no fluid accumulation compared with rats in which biliary drainage was not interrupted (not shown). There was no hyperemia or fluid accumulation in normal and BDC animals not treated with LPS (not shown).

FIG. 2.

Effect of BDC on the macroscopic appearance of the small intestine. BDC-treated (left) and sham-operated (right) rats at 4 h after administration of 15 mg of LPS per kg i.v. are shown. The findings are representative of those from 15 separate experiments.

Experiments were performed to determine the effects of infusing bile from normal rats not treated with LPS into the duodenums of rats subjected to BDC. If the recipients were not treated with LPS, the transfer did not induce any intestinal toxicity. Therefore, the surgical procedures did not damage the intestine. If the BDC recipients of normal bile were treated with LPS, no intestinal damage was detected (not shown). This result indicates that the protective effect of BDC in LPS-treated rats was not due to the removal of components present in normal bile. In another experiment, BDC was performed in normal rats and bile from LPS-treated rats was infused into the duodenum. Marcoscopically, hyperemia and edema were observed in these animals (not shown).

The histological appearance in sham-operated control rats was normal (Fig. 3A), and similar observations were made for rats subjected to BDC but not treated with LPS (not shown). By contrast, LPS injection alone was followed by marked tissue damage 4 h later (Fig. 3B). There was neutrophil margination in the villi, the epithelial layer was severely damaged with superficial necrosis of villi and sloughing of epithelium, and some villi were completely absent. There was edema in the submucosa and lamina propria, with congested and dilated blood vessels. In animals treated with LPS, all the features of intestinal damage were greatly reduced if BDC was performed (Fig. 3C). There was no evidence of congestion, vasodilatation, or edema in the lamina propria, and the epithelial layer was intact. There was minimal congestion in the submucosa. Occasional villi in the BDC group were edematous at the tips, although clearly there was no major disturbance in the integrity of the tissues. Figure 3 shows microscopy of the jejunum, and similar results were obtained for the duodenum and ileum (not shown).

FIG. 3.

Effect of BDC on the histological appearance of the jejunum. Sections from a normal rat (A), an LPS-injected rat (B), and an LPS-injected rat after BDC (C) are shown (magnification, ×100). The data are representative of those from three separate experiments, and on each occasion there were five animals per group. In panel C there is poorly defined material in the lumen. This was observed in some BDC-treated animals when it was difficult to remove luminal contents from the specimen.

Measurement of the volume of fluid in the small intestine revealed that 4 h after LPS, there was a marked increase compared with control animals (Fig. 4). In LPS-treated animals there was a significant reduction in fluid volume when bile was prevented from entering the gut, either by occlusion or cannulation of the bile duct (Fig. 4). To further implicate bile in these events, the bile ducts of LPS-treated rats were cannulated and the bile was deviated to the upper jejunum. Here the duodenum appeared normal, whereas hemorrhage and fluid accumulation occurred in the lower reaches of the intestine (not shown).

FIG. 4.

Volumes of clear fluid from the small intestine. Groups of rats were normal (bar a), LPS treated (bar b), normal with BDC (bar c), LPS treated with BDC (bar d), and LPS treated with bile duct occluded (bar e). Volumes are shown as means and SEs (n = 5).

TNF-α production in bile.

TNF-α in the bile and sera of normal and LPS-treated animals was measured by bioassay. TNF-α was detected in normal bile, and there was a great increase in biliary TNF-α after LPS treatment (Table 1). In the LPS-treated animals, the concentration of TNF-α in the bile was over 30-fold higher than that in the serum. Rats were treated with gadolinium chloride, which preferentially eliminates Kupffer cells (8). Gadolinium chloride at 25 mg/kg given 24 h prior to LPS markedly reduced the levels of serum and biliary TNF-α at 1.5 h after LPS administration (Table 1). Pretreatment with gadolinium chloride also protected the intestine from the damaging effects of LPS (Fig. 5). The toxic effects of LPS (Fig. 5A) were totally abrogated by gadolinium chloride (Fig. 5B).

TABLE 1.

TNF-α concentrations in sera and bile 1.5 h after administration of LPS (15 mg/kg) i.v.

Treatment	TNF-α concn (ng/ml)a in:
	Serum	Bile
None	<0.005 (10)	3.10 ± 1.3 (11)
LPS	36 ± 32.45 (7)	1,345 ± 616 (5)
Gadolinium chloride 24 h prior to LPS	<0.005 (3)	0.06 ± 0.06 (3)

^aResults are expressed as means ± SE. The number of animals in each group is shown in parentheses. 

FIG. 5.

Effect of gadolinium chloride on the jejunum. Rats were injected with LPS i.v. and pretreated without (A) and with (B) gadolinium chloride (magnification, ×100). The data are representative of those from three separate experiments, and on each occasion there were four animals per group.

TNF-α infusion into the intestine.

In rats infused with TNF-α intraluminally, there were marked changes to the intestine (Fig. 6A) which overall were similar to, although not as severe as, the changes in animals given LPS intravenously (Fig. 3B). There was loss of mucosal integrity, epithelial sloughing, and severe congestion and edema in the lamina propria (Fig. 6A). Compared with animals given LPS, the animals treated with infusion of TNF-α into the lumen had less damage in the submucosa (Fig. 6A compared with Fig. 3B), suggesting that TNF-α in the lumen is sufficient for damage to the mucosa but that systemic TNF-α may be required for damage to the lower layers of the intestinal wall. The duodenum, the site of TNF-α infusion, was most severely affected, and the ileum was less affected (not shown) than the jejunum (Fig. 6A). When saline was infused into the duodenum, there was no evidence of damage to the small intestine (Fig. 6B).

FIG. 6.

Effect of TNF-α on the jejunum. Rats were infused intraduodenally with TNF-α (A) or sterile saline (B) (magnification, ×100). The data are representative of those for five animals in each group.

Luminal administration of anti-TNF-α antibodies.

To investigate the role of intraluminal TNF-α in the toxicity of LPS to the intestine, the effect of intraduodenal instillation of anti-TNF-α antibodies on survival was assessed. Animals in the two control groups were either sham operated or treated with normal rabbit serum intraduodenally. The results in these two groups were very similar, and most animals did not survive beyond 36 h. By contrast, in the group treated with anti-TNF-α antibodies for the first 4 h after administration of LPS, survival was markedly improved in the first stages of the experiment, and few animals succumbed within the first 36 h (Fig. 7). At this time survival in the anti-TNF group was significantly greater than in the sham-operated group (P = 0.0252), the normal rabbit serum group (P = 0.0211), or both control groups combined (P = 0.0150). However the beneficial effect of anti-TNF antibody was less apparent at later time points. At the end of the experiment at 72 h, the improvement in survival of the anti-TNF group did not reach statistical significance (P = 0.0755 for the comparison of the anti-TNF group with both control groups combined).

FIG. 7.

Kaplan-Meier plots of the effect of intraduodenal anti-TNF-α antibodies on survival. LPS was administered at time zero to animals that had been either sham operated or infused intraduodenally with either normal rabbit serum or anti-TNF-α antibody. There were 12 animals in each group.

DISCUSSION

In this study we report a novel involvement of bile in the pathology of gastrointestinal tract damage in endotoxemia. External drainage of bile or bile duct occlusion markedly reduced the intestinal effects of LPS and prolonged survival. These experiments indicate that after treatment with LPS, the bile contains substances that are capable of mediating intestinal damage, and external drainage protects the intestine from their toxic effects. We have recently reported similar observations in two other models in rats. One is a model of food allergy, where external drainage of bile abrogated the toxic effects that the allergen normally elicits on the intestine (5). The other is a model of Salmonella infection, in which external drainage of bile reduced the capacity of organisms to invade the liver and mesenteric lymph nodes after oral infection (10).

Obvious questions arise as to the nature and source of the factor(s) in bile which is responsible for the tissue damage. TNF-α is a likely candidate, as it has a major role in shock and tissue injury (35) and is an early mediator, acting synergistically with other factors (23). In this study we report that the ulcerative appearance of the gastrointestinal tract after systemic LPS challenge was similar to that after intraduodenal TNF-α infusion (Fig. 3 and 6) and that neutralizing anti-TNF-α antibodies infused into the duodenum prolonged survival after LPS administration (Fig. 7). Taken together with the studies on biliary drainage, these findings suggest that luminal TNF-α, derived from the bile, makes a major contribution to the intestinal damage in endotoxemia.

This conclusion is supported by the bioassay findings presented in Table 1. In the bioassay, the activity was inhibited by anti-TNF-α antiserum, providing evidence that the observed effects were indeed caused by TNF-α. However, the precise molecular nature of the activity detected in the bioassay is uncertain. When bile was analyzed for TNF-α by immunoblotting, several bands with higher molecular masses than the expected 17 kDa were identified, and specific bands at 17 kDa were not detected (not shown). The higher-molecular-weight bands may consist of TNF-α conjugated to other proteins. In a recent study of TNF-α in human bile, anti-TNF-α antibodies also detected bands of several different molecular sizes (1). Other factors in bile may contribute to intestinal pathology. We have observed elevated levels of IL-1α and IL-1β in bile after LPS challenge in rats (M. T. Wiseman, W. A. Sewell, and G. D. F. Jackson, unpublished observations).

There are several reasons to believe that TNF-α in bile is derived from hepatic synthesis rather than extracted from plasma. First, LPS is taken up by cells in the liver (24) and leads to production of TNF-α by these cells (4). Second, the concentration of TNF-α in bile was markedly reduced by treatment with gadolinium chloride (Table 1), an agent which preferentially inhibits Kupffer cell function (8). Finally, in LPS-treated animals, the concentration of TNF-α was much higher in bile than in serum (Table 1), again supporting the concept of production in the liver.

The degradative environment of the gut would require that molecules secreted into bile possess interesting survival properties. In this regard, TNF-α has been reported to be stable in a detergent environment (28), and we have found that recombinant TNF-α retains its bioactivity in bile for at least 4 h (not shown). Markedly elevated levels of TNF-α in the feces of patients with active inflammatory bowel diseases have been reported (3). The ability of TNF-α to induce adhesion molecules on intestinal epithelial cell lines (16) and to influence cytokine production and proliferation in intestinal cell lines (14, 20) suggests the presence of specific receptors on such cells.

TNF receptors are predominantly located on the basolateral aspect of intestinal epithelial cells (33), as are receptors for IL-1 (13). These findings raise the question of how luminal TNF-α could gain access to such receptors. A possible mechanism is based on the report that after LPS stimulation, macrophages release factors that increase intestinal epithelial permeability (37). Therefore, we propose that in endotoxemia, intestinal macrophages may increase epithelial permeability, allowing luminal TNF-α to reach receptors on the basolateral aspect of the epithelial cells.

In summary we emphasize that (i) the intestinal damage component of endotoxic shock can be separated from other features of the syndrome, (ii) biliary factors operating from the lumen of the gut are involved in at least the final stages of tissue damage, (iii) there is evidence suggesting that TNF-α is involved in the tissue damage, and (iv) means for assessing new treatments to control or prevent injury at this site have been demonstrated. Such findings are a useful model for delivery of neutralizing antibodies to patients with endotoxemia and high levels of TNF-α in their gut lumens. Further, we speculate that biliary factors are likely to be involved in the induction and/or continuing pathogenesis of other inflammatory disorders of the gastrointestinal tract.