Idiosyncratic adverse drug reactions (IADRs) represent an important human health problem, yet animal models for preclinical prediction of these reactions are lacking. Recent evidence in animals suggests that some IADRs arise from drug interaction with an inflammatory episode that renders the liver sensitive to injury. Diclofenac (DCLF) is one of those drugs for which the clinical use is limited by idiosyncratic liver injury. We tested the hypothesis that modest inflammation triggered in rats by a small dose of lipopolysaccharide (LPS) renders a nonhepatotoxic dose of DCLF injurious to liver. Cotreatment of rats with nonhepatotoxic doses of LPS and DCLF resulted in elevated serum alanine aminotransferase activity and liver histopathologic changes 6 h after DCLF administration. Neither LPS nor DCLF alone had such an effect. Gene array analysis of livers revealed a unique gene expression pattern in the LPS/DCLF-cotreated group compared with groups given either agent alone. Antiserum-induced neutrophil (PMN) depletion in LPS/DCLF-cotreated rats protected against liver injury, demonstrating a role for PMNs in the pathogenesis of this LPS/DCLF interaction. Gut sterilization of LPS/DCLF-treated rats did not protect against liver injury. In contrast, gut sterilization did attenuate liver injury caused by a large, hepatotoxic dose of DCLF, suggesting that hepatotoxicity induced by large doses of DCLF is caused in part by its ability to increase intestinal permeability to endotoxin or other bacterial products. These results demonstrate that inflammation-DCLF interaction precipitates hepatotoxicity in rats and raise the possibility of creating animal models that predict human IADRs.
Idiosyncratic adverse drug reactions (IADRs) remain a challenging human health problem (Boelsterli, 2003; Kaplowitz, 2005). They are one of the leading causes for drug development failures and removal of medicines from the market. A well known example is troglitazone, which was removed from the market by the Food and Drug Administration in 2000, because of its potential to cause idiosyncratic liver injury (Chojkier, 2005). IADRs appear to be independent of dose, and the onset of injury varies relative to the onset of drug treatment. Unfortunately, the mechanisms of idiosyncratic reactions are poorly understood despite the large number of drugs associated with these reactions. One of the common targets of idiosyncratic toxicity is liver. Animal models to predict these adverse reactions are lacking. Therefore, it is of interest to establish models that mimic human IADRs. In addition, mechanistic studies of such models could provide biomarkers for IADRs or strategies to prevent them.
There are two conventional hypotheses to explain IADRs. One is that the reactions occur as a consequence of drug metabolism polymorphisms, which result in different levels of toxic drug metabolites among patients (Williams and Park, 2003). The other one argues that they arise from a specific immune response to a hapten formed by a drug or its metabolites (Pirmohamed et al., 2002). However, convincing evidence for these hypotheses is lacking for the majority of drugs associated with idiosyncratic toxicity. It is equally plausible that other unrecognized events render tissues susceptible to toxicity during drug therapy.
Results of several studies in experimental animals indicate that modest inflammation, triggered by a small dose of lipopolysaccharide (LPS), augments hepatotoxicity induced by several classes of xenobiotic agents. For example, coadministration of nonhepatotoxic doses of LPS and trovafloxacin (TVX), a quinolone antibiotic associated with hepatic IADRs, results in liver damage that resembles human TVX idiosyncrasy. By contrast, cotreatment with LPS and levofloxacin, a quinolone antibiotic without idiosyncratic liability, did not produce liver injury at a dose equipotent to that of TVX (Waring et al., 2005). Episodes of modest, subclinical inflammation are commonplace in people. Because they occur irregularly and may go unnoticed during drug therapy, their interaction with drugs could explain the erratic temporal and dose relationships that characterize IADRs (Roth et al., 2003).
Nonsteroidal anti-inflammatory drugs (NSAIDs) comprise another class of drugs, some of which cause hepatic IADRs. For example, diclofenac (DCLF) has caused rare but sometimes serious hepatotoxicity in humans (Boelsterli, 2003). Although the apparent incidence of severe DCLF-induced hepatic adverse reactions is quite low (from one to two cases per million prescriptions or six to 18 cases/100,000 persons per year), the large number of patients treated with DCLF makes the absolute number of cases impressive (Walker, 1997). In addition, cases of severe injury leading to liver transplantation comprise a large proportion of the reported cases of DCLF-induced hepatotoxicity (Lewis, 2003). The scarceness of liver biopsies from patients and the diverse histopathological presentations in available samples make it difficult to draw clues about mechanisms from human liver pathology alone (Zimmerman, 1999). Thus, the pathogenesis of this low-incidence/high-severity DCLF hepatotoxicity is largely unknown. Several mechanisms have been proposed, including formation of reactive drug metabolites, oxidative stress (Cantoni et al., 2003), mitochondrial injury (Masubuchi et al., 2002), and immune-mediated hypersensitivity (Greaves et al., 2001). Experimental data supporting such mechanisms are incomplete, especially at the in vivo level (Boelsterli, 2003; Cantoni et al., 2003).
Based on the observation that a number of drugs that have the ability to cause IADRs in human patients cause liver injury in LPS-treated rats and are not hepatotoxic in naive rats, we tested the hypothesis that a modest inflammatory episode in rats increases the sensitivity of the liver to hepatotoxic effects of DCLF. We found that pretreatment of rats with a small, nonhepatotoxic dose of LPS renders a nontoxic dose of DCLF injurious to liver. Furthermore, we determined whether gene expression profiles could distinguish a LPS/DCLF interaction that causes hepatotoxicity from the nontoxic treatments with LPS or DCLF alone and provide clues about the mode of action.
Materials and Methods
Experimental Design. Rats received humane care according to the criteria outlined in the Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences, and procedures were approved by the Michigan State University Committee on Animal Use and Care. Male Sprague-Dawley rats (Crl:CD (SD)IGS BR; Charles River, Portage, MI) weighing 250 to 350 g were used for these studies. Animals were fed standard chow (Rodent chow/Tek 8640; Harlan Teklad, Madison, WI) and allowed access to water ad libitum. They were allowed to acclimate for 1 week in a 12-h light/dark cycle before use.
Rats were fasted for 16 h and then given a nonhepatotoxic dose of LPS [29 × 106 endotoxin units (EU)/kg; catalog number L-2880, lot 072K4095; Sigma-Aldrich, St. Louis, MO] or sterile saline, i.v. This activity was determined using a QCL Chromogenic LAL endpoint assay from Cambrex (East Rutherford, NJ). Two hours later, they were given 20 mg/kg DCLF (Sigma-Aldrich) or sterile saline i.p. For the high-dose DCLF study, 100 mg/kg DCLF was given without LPS administration. Rats remained fasted and were sacrificed 6 h after DCLF treatment for evaluation of liver injury, histopathology, and gene expression. They were anesthetized with sodium pentobarbital (50 mg/kg i.p.). Blood was collected from the dorsal aorta, and a portion of the blood was placed into a tube containing sodium citrate (final concentration, 0.38%) for collection of plasma. Another portion was allowed to clot at room temperature, and serum was collected and stored at –20°C until use. Representative (3–4-mm) slices of the ventral portion of the left lateral liver lobe were collected and fixed in 10% neutral-buffered formalin. A portion of the right medial lobe of the liver was flash-frozen in liquid nitrogen for subsequent gene expression analysis.
PMN Depletion. PMNs were depleted by intravenous administration of 0.25 ml of rabbit anti-rat PMN serum (anti-PMN serum; Intercell Technologies, Jupiter, FL), diluted 1:1 with sterile saline, 16 h before administration of LPS. Normal rabbit serum (normal serum) was administered to some animals as a control. Previous studies in which anti-PMN serum was administered to rats demonstrated a selective depletion of PMNs (Snipes et al., 1995; Luyendyk et al., 2005).
Gut Sterilization. Gut sterilization was achieved by treating rats daily with polymyxin B (150 mg/kg) and neomycin (450 mg/kg) orally for 4 days before LPS/DCLF treatment or high-dose DCLF alone. This treatment has been shown to completely abolish the Gram-negative bacterial growth in rat fecal culture and to reduce plasma endotoxin concentration after chronic ethanol treatment (Adachi et al., 1995).
Bacterial Culture in Fecal and Liver Homogenates. Fecal and liver bacterial cultures were performed 6 h after DCLF or saline treatment. Rat fecal pellets (400 mg) were collected directly from the anus and homogenized in 4 ml of trypticase soy broth containing 15% glycerol. A 10-μl fecal slurry was plated on MacConkey plates (Becton, Dickinson Co., Sparks, MD). For the liver bacterial culture, 100 mg of the liver from the left lateral lobe was collected in 0.5 ml of trypticase soy broth containing 15% glycerol and homogenized. Liver homogenates were diluted serially to 1:106, and 100 μl of each dilution was plated on MacConkey plates. The culture plates were incubated for 24 h at 37°C in an atmosphere of 95% air and 5% CO2. The presence or absence of bacterial colonies was determined for the fecal homogenates. The colonies of bacteria were counted for each dilution of liver homogenate and averaged for each animal.
ALT Activity and Histopathology Assessment. Hepatic parenchymal cell injury was estimated by quantifying serum alanine aminotransferase (ALT) activity. ALT activity was determined spectrophotometrically using Infinity-ALT from Thermo Electron Corp. (Louisville, CO). Formalin-fixed liver samples were routinely processed and stained with hematoxylin and eosin (H&E). Slides were read by a pathologist (E. A. G. Blomme) without knowledge of treatment.
RNA Preparation. Frozen liver samples (approximately 100 mg of tissue per sample) were immediately added to 2 ml of TRIzol reagent per sample (Invitrogen Life Technologies, Carlsbad, CA) and homogenized using a Polytron 300D tissue grinder (Brinkman Instruments, Westbury, NY). One milliliter of the tissue homogenate was transferred to a microfuge tube, and total RNA was extracted with chloroform followed by nucleic acid precipitation with isopropanol. The pellet was washed with 80% ethanol and resuspended in molecular biology grade water. Nucleic acid concentration was determined spectrophotometrically at 260 nm (Smart-Spec; Bio-Rad Laboratories, Hercules, CA), and RNA integrity was evaluated using an Agilent bioanalyzer (model 2100; Agilent Technologies, Foster City, CA).
Gene Array Analysis. Microarray analysis was performed using the standard protocol provided by Affymetrix, Inc. (Santa Clara, CA). In brief, approximately 15 μg of total RNA was reversed-transcribed into cDNA using a Superscript II Double-Strand cDNA synthesis kit (Invitrogen Life Technologies) according to the manufacturer's instructions, with the exception that the primer used for the reverse transcription reaction was a modified T7 primer with 24 thymidines at the 5′ end (Affymetrix). The sequence was 5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-(dT)24–3′. cDNA was purified via phenol/chloroform/isoamylalcohol (Invitrogen Life Technologies) extraction and ethanol precipitation. The purified cDNA was resuspended in molecular biology grade water. After this procedure, biotin-labeled cRNA was synthesized according to the manufacturer's instructions from the cDNA using the Enzo RNA Transcript Labeling Kit (Affymetrix). The labeled cRNA was then purified using RNeasy kits (Qiagen, Valencia, CA). Subsequently, cRNA concentration and integrity were evaluated. Approximately 20 μg of cRNA was then fragmented in a solution of 40 mM Tris acetate, pH 8.1, 100 mM potassium acetate, and 30 mM magnesium acetate at 94°C for 35 min. Fragmented labeled cRNA was hybridized to an Affymetrix rat genome RAE230A array, which contains sequences corresponding to roughly 15,900 transcripts, at 45°C overnight using an Affymetrix Hybridization Oven 640. The array was subsequently washed and stained twice with strepavidin-phycoerythrin (Molecular Probes, Eugene, OR) using a Gene-Chip Fluidics Workstation 400 (Affymetrix). The array was then scanned using the Affymetrix GeneChip Scanner 3000.
The microarray scanned image and intensity files (.cel files) were imported into Rosetta Resolver gene expression analysis software version 4.0 (Rosetta Inpharmatics, Seattle, WA). Error models were applied, and ratios were built for each treatment array versus its respective vehicle control (pooled in silico). Using Rosetta Resolver, a p-value is calculated for every fold change using the Rosetta Resolver error model (Rajagopalan, 2003).
Statistical Filtering of Genes Changed after LPS/DCLF Treatment. Genes changed after treatment with LPS/DCLF, LPS/saline, or saline/DCLF were identified with saline/saline as baseline using the following criteria: P < 0.01 in at least three of four animals (groups LPS/saline and LPS/DCLF); P < 0.01 in at least four of five animals (group saline/DCLF). To identify the genes that changed after LPS/DCLF treatment relative to LPS/saline group or saline/DCLF group, a second filter was applied with the LPS/DCLF group as baseline using the same criteria as above.
The genes expressed to a greater degree after LPS/DCLF treatment than after LPS/saline and saline/DCLF treatment were classified based on the relationship of their gene products to one or more functions. The classification was based on the gene function described in Entrez Gene site (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene). Furthermore, this list of genes was imported into Genomatix Bibliosphere 5.13 software (München, Germany). This yielded a list of biological process terms ranked by their probability of over-representation or under-representation in a particular gene list. Functional annotations were based on the Genomatix knowledge base. “Expected” refers to the numbers of gene expression changes expected by chance in the gene list. The “Z Score” ranks the deviation of the observed number of genes in one biological process category of the Genomatix database found in the imported list of genes from the numbers expected to occur by chance. That is, it ranks the probability that the genes in this category are over-represented or under-represented in the gene list.
Serum MIP-2 Concentration and Blood/Liver Leukocyte Evaluation. Serum MIP-2 concentrations were determined using a commercially available ELISA kit (Biosource International, Camarillo, CA). Total blood leukocytes were quantified using a Unopette white blood cell determination kit (Becton-Dickinson, Franklin Lakes, NJ) and a hemacytometer. Slides were prepared from whole blood and stained using the Hema 3 Staining System (Fisher Scientific, Middletown, VA), and differential counting was performed. Immunohistochemical staining for PMNs was performed on formalin-fixed liver sections as described previously (Yee et al., 2003). Hepatic PMN accumulation was evaluated by counting PMNs in 20 randomly selected, high-power fields (400×).
Statistical Analysis. All data are expressed as mean ± S.E.M. ALT activity, ELISA, and circulating and liver leukocyte data were analyzed by two-way ANOVA with Tukey's posthoc test. For Figs. 1 and 5 and Table 5, one-way ANOVA was applied. Data were transformed if they did not pass the normality and equal variance tests for ANOVA. For microarray analysis, error models were applied, and ratios were built for each treatment array versus its averaged respective vehicle control using the Rosetta Resolver system. Gene expression was considered significantly changed if the P value was less than or equal to 0.01. Agglomerative cluster analysis was performed using the average link heuristic criteria and the Euclidean distance metric for similarity measure.
Hepatotoxicity from DCLF/LPS Interaction. In a preliminary study, a large dose of DCLF caused liver injury within 6 h. Accordingly, in an initial dose-response study, rats were given DCLF (i.p.) at doses from 0 to 100 mg/kg, and serum ALT activity was evaluated 6 h after DCLF treatment. Doses up to 40 mg/kg did not cause elevation in serum ALT activity. Administration of DCLF at doses of 50 and 100 mg/kg caused a significant increase in ALT activity (Fig. 1), indicating hepatocellular injury.
To determine the influence of a nonhepatotoxic dose of LPS on DCLF hepatotoxicity, rats were pretreated with LPS (29 × 106 EU/kg i.v.) or sterile saline. Two hours later they were treated with DCLF (20 mg/kg i.p.) or saline. Serum ALT activity was evaluated 6 h later. Neither LPS nor DCLF when given alone altered ALT activity. However, cotreatment with LPS and DCLF caused a significant increase in serum ALT activity (Fig. 2).
Livers from both saline/saline-treated and saline/DCLF-treated rats had no or minimal histopathological changes (Table 1). LPS/saline treatment caused moderate leukocyte infiltration, occasional hepatocellular apoptosis, and modest parenchymal edema and hemorrhage. LPS/DCLF cotreatment enhanced hepatocellular apoptosis, parenchymal edema, and hemorrhage relative to LPS/saline treatment.
Microarray Analysis of Livers from Rats Treated with LPS/DCLF. Rat livers were subjected to RNA extraction and subsequent microarray analysis for global gene expression. Hierarchical clustering analysis revealed mostly treatment-related clusters (Fig. 3). Saline/DCLF treatment caused few gene expression changes. All rats treated with LPS irrespective of cotreatment clustered together. Within this major cluster, LPS/DCLF-treated rats formed a distinct subcluster. One LPS/saline-treated rat clustered with the LPS/DCLF-treated group. Interestingly, this animal had the largest serum ALT value in the LPS/saline group.
The number of gene expression changes after LPS/saline, saline/DCLF, and LPS/DCLF treatments were identified with saline/saline-treated rats as the baseline. The numbers of genes changed in expression are depicted as a Venn diagram (Fig. 4). A large number of gene expression changes occurred after either LPS/DCLF or LPS/saline treatment, but many more expression alterations occurred in the former group. A large number of gene expression changes occurred after both of these treatments (L∩LD intersection in Fig. 4). Numerous genes were differentially expressed specifically in the LPS/DCLF group (LD in Fig. 4). DCLF alone caused few changes in gene expression based on the criteria used.
The genes altered in expression after LPS/DCLF treatment were further filtered by their expression pattern. Normalization to LPS/DCLF as the baseline was used to identify the genes changed to a greater degree after LPS/DCLF treatment compared with LPS alone and DCLF alone. This group of genes was interesting because liver injury occurred in the LPS/DCLF group but not in the LPS alone or the DCLF alone group. Accordingly, genes changed to a greater degree in the LPS/DCLF group than in the LPS/saline and saline/DCLF groups are associated with liver injury. These genes were classified based on the relationship of their gene products to one or more biological functions (Table 2). Many genes related to inflammation, cell death/survival, stress response, hemostasis, or hypoxia were selectively up-regulated by LPS/DCLF treatment.
The genes changed to a greater degree after LPS/DCLF treatment compared with LPS alone or DCLF alone were imported into Genomatix Bibliosphere 5.13. This yielded a list of gene ontology biological process terms ranked by their probability of over-representation or under-representation on the gene list (Table 3).
Neutrophil-Related Gene Expression Changes after LPS/DCLF Treatment and Effects of Neutrophil Depletion on LPS/DCLF-Induced Hepatotoxicity. Transcripts for the genes encoding the neutrophil chemokines, such as MIP-2 and MIP-1α, and the adhesion molecule ICAM-1, were changed to a greater degree in the LPS/DCLF cotreated group compared with LPS or DCLF given alone (Fig. 5). Serum MIP-2 protein levels in both saline/saline and saline/DCLF groups were below the detection limit of the assay (Fig. 6). LPS/saline treatment increased serum MIP-2 protein concentration, whereas LPS/DCLF treatment caused a much greater increase (Fig. 6).
Both LPS/saline and LPS/DCLF treatment caused hepatic PMN accumulation (Tables 1 and 4). To investigate the requirement for neutrophils in LPS/DCLF hepatotoxicity, neutrophil numbers in liver were reduced by treating rats with rabbit anti-rat PMN serum 16 h before LPS treatment. This treatment selectively reduced circulating and hepatic PMNs (Table 4). The anti-PMN serum alone did not alter ALT activity but did attenuate LPS/DCLF hepatotoxicity, as reflected by reduction in serum ALT activity (Fig. 7). LPS/DCLF cotreatment reduced circulating lymphocytes (Table 4), similar to effects of LPS alone (data not shown).
Effects of Gut Sterilization on LPS/DCLF Hepatotoxicity. DCLF is known to cause gastrointestinal injury and bacterial translocation from intestine to liver (Kim et al., 2005). The observation in the gene array study that the rat with the largest serum ALT activity after LPS/saline treatment clustered with the LPS/DCLF-treated rats suggested that the LPS/DCLF interaction might be due to bacteria/endotoxin exposure that resulted from a DCLF-induced intestinal permeability increase, which added to the effects of exogenous LPS given before DCLF exposure. To test this hypothesis, rats were treated with the nonabsorbable antibiotics, polymyxin B, and neomycin orally for 4 days before LPS/DCLF treatment. This antibiotic treatment completely abolished the colony growth from homogenized rat fecal samples on MacConkey agar plates, which are selective for Gram-negative bacteria (data not shown).
Gut sterilization did not significantly affect LPS/DCLF hepatotoxicity (Fig. 8). This result suggested that the LPS/DCLF interaction is not due to DCLF enhancing bacteria/endotoxin translocation from intestine to liver.
Effects of Gut Sterilization on Liver Injury in Rats Treated with a Hepatotoxic Dose of DCLF. Large doses of DCLF have been shown to cause bacterial translocation from the intestine to liver (Seitz and Boelsterli, 1998; Kim et al., 2005). Consistent with this, DCLF (100 mg/kg i.p.) did increase the level of Gram-negative bacteria in the liver (Table 5). Although gut sterilization had no significant effect on LPS/DCLF hepatotoxicity, the possibility remained that LPS or bacterial translocation from intestine to liver plays an important role in the hepatotoxicity from a large, toxic dose of DCLF. To test this hypothesis, rats were treated with nonabsorbable antibiotics as above before they were given a toxic dose of DCLF (100 mg/kg i.p.). This antibiotic treatment attenuated the Gram-negative bacterial colonies detected in liver (Table 5), and it reduced the serum ALT activity after administration of a hepatotoxic dose of DCLF, indicating attenuation of liver injury (Fig. 9).
DCLF is well known to cause idiosyncratic hepatotoxicity (Boelsterli, 2003). In this study, we showed that pretreating rats with a small dose of LPS converts a dose of DCLF that is normally noninjurious to one that is hepatotoxic. This result suggests the possibility that human patients might become susceptible to DCLF hepatotoxicity during a modest inflammatory episode. This appears even more intriguing, because most patients take DCLF for inflammation-related diseases such as rheumatoid arthritis or osteoarthritis. In fact, osteoarthritis has been associated with increased risk for DCLF-induced liver injury (Banks et al., 1995).
Mechanisms of DCLF-induced liver injury remain incompletely understood. Mitochondrial injury leading to cell death has been proposed as a mode of hepatocellular injury, but by itself, this does not explain the occurrence of injury only in a minority of patients (Masubuchi et al., 2002; Gomez-Lechon et al., 2003). In addition, to our knowledge, no direct evidence of mitochondrial injury in vivo has been reported. Given our results, it would be interesting to see whether DCLF induces mitochondrial injury in vivo or in vitro in the presence of inflammatory mediators.
Whether inflammation in human patients acts as a risk factor for DCLF-induced idiosyncratic liver injury requires additional study. Interestingly, polymorphisms in genes encoding cytokines IL-4 and IL-10 have been identified in patients developing DCLF-induced liver injury (Aithal et al., 2004). These polymorphisms cause less IL-10 and more IL-4 production. Because IL-10 is an anti-inflammatory cytokine and IL-4 participates in B cell activation, these changes could enhance hepatotoxic interactions between DCLF and LPS or other inflammagens to which patients are exposed during drug therapy.
The potentiation of DCLF-induced hepatotoxicity by LPS might appear paradoxical, because DCLF acts as a NSAID to dampen inflammatory responses. The primary mechanism of action of DCLF in this respect is inhibition of the enzymes cyclooxygenase (COX). COX catalyzes production of prostaglandins (PGs) and thromboxanes (Tx) from arachidonic acid. PGs and Tx have diverse biological actions, either proinflammatory or anti-inflammatory. However, as a nonselective COX inhibitor (i.e., inhibits both isozymes COX-1 and COX-2), it inhibits the synthesis of cytoprotective and anti-inflammatory PGs produced by constitutive COX-1 activity as well as products of inducible COX-2, which are mostly proinflammatory. Moreover, the toxic interaction of DCLF with LPS might arise from inflammatory mediators (e.g., cytokines) that are not products of the COX pathway and are not inhibited by NSAIDs.
Analysis of hepatic gene expression revealed a unique pattern of gene expression for animals treated with LPS/DCLF. The genes expressed to a greater degree in the LPS/DCLF group than in LPS/saline and saline/DCLF groups were selected for further functional analysis because liver injury followed this pattern. Interestingly, many genes involved in inflammation, cell death/survival, and response to stress were expressed to a greater degree in the LPS/DCLF group, consistent with the hypothesis that DCLF exaggerates the inflammatory stress caused by LPS.
Neutrophil/immune cell chemotaxis was at the top of the biological process rankings of the GO functional analysis for genes expressed to a greater degree in the LPS/DCLF group, and transcripts representing several neutrophil chemokines and adhesion molecules were up-regulated to a great degree in this group, suggesting that PMNs might be involved in the pathogenesis. Serum MIP-2 concentration reflected the increase in the transcripts for this chemokine. The mechanism by which DCLF enhances MIP-2 protein expression remains unknown. There are reports that COX-2 inhibition enhances MCP-1/MIP-2 production in models of inflammation caused by radiation (Kyrkanides et al., 2002). In addition, it has been proposed that in the face of inhibition of COX, arachidonic acid is metabolized more extensively by lipoxygenase. In this regard, it is interesting that the lipoxygenase products derived from arachidonic acid, such as 15-hydroxyperoxyeicosatetraenoic acid (15-HPETE), can increase adhesion molecules and transmigration of neutrophils across endothelial cell barriers (Sultana et al., 1996). Thus, shifting of the generation of eicosanoids from the COX pathway to the lipoxygenase pathway by DCLF during inflammation (e.g., LPS exposure) could contribute to enhancing the expression of MIP-2 and subsequent PMN activation. Depletion of PMNs in LPS/DCLF-treated rats attenuated liver injury, indicating the involvement of PMNs in the pathogenesis. PMNs cause cytotoxicity through the release of reactive oxygen species and other toxic factors. It is interesting in this regard that the addition of noncytotoxic concentrations of peroxidase/H2O2 to hepatocyte cultures markedly increased DCLF cytotoxicity (Tafazoli et al., 2005). In fact, several novel reactive metabolites of DCLF formed by incubation of the drug with PMN-derived myeloperoxidase (Zuurbier et al., 1990; Miyamoto et al., 1997).
DCLF, as other NSAIDS, is known to cause gastrointestinal injury and bacteria translocation from the intestine to liver (Banks et al., 1995; Seitz and Boelsterli, 1998; Kim et al., 2005). The resultant excessive bacteria/endotoxin exposure of liver could combine with the small dose of administered LPS to cause liver injury. This is not likely, because gut sterilization with nonabsorbable antibiotics did not dampen the hepatic damage caused by coadministration of LPS and DCLF. In contrast, gut sterilization did reduce liver injury caused by a larger, hepatotoxic dose of DCLF. It is possible that DCLF becomes hepatotoxic only in the presence of bacteria/endotoxin or other inflammagens and that the larger dose of DCLF is hepatotoxic because it causes translocation of LPS or bacteria from the intestine to liver.
LPS can increase intestinal permeability by inducing epithelial barrier dysfunction. This is reflected, for example, as an increase in dextran absorption (Han et al., 2004; Moreiz et al., 2005). In addition, LPS administration to rats can change drug bioavailability through alterations in expression and activity of transporters and drug-metabolizing enzymes in the intestinal epithelium (Kalitsky-Szirtes et al., 2004). Accordingly, it seems possible that LPS exposure could alter the toxicity of orally administered diclofenac by increasing its absorption from the gastrointestinal tract. This potential mode of action probably did not pertain to our results, because we administered the drug intraperitoneally. However, in human patients who typically take diclofenac orally, this could provide an additional mode by which LPS exposure could influence drug toxicity.
Diclofenac causes IADRs in people in a time frame varying from a few weeks to a year after onset of therapy with the drug (Boelsterli, 2003). If drug-inflammation interaction underlies some human IADRs, then a reaction would be expected to occur only in patients with an inflammatory episode occurring during drug therapy and having sufficient magnitude to precipitate a hepatotoxic interaction with the drug. Modest inflammatory episodes probably vary in both timing of onset and magnitude among and within people (Roth et al., 2003). Accordingly, this hypothesis of drug-inflammation interaction could explain the variable onset of diclofenac IADRs in human patients.
In summary, we established a potential model of DCLF-induced idiosyncratic liver injury by pretreating rats with a small dose of LPS before an otherwise nonhepatotoxic dose of DCLF. The molecular mechanisms of this inflammation-DCLF interaction remain unknown, but gene array analysis and PMN depletion studies indicated a role for PMNs. Gut sterilization of LPS/DCLF-treated rats failed to protect against liver injury, suggesting that this interaction is not due to excessive bacteria/endotoxin translocation from intestine to liver caused by DCLF. However, gut sterilization attenuated the liver injury caused by a large, toxic dose of DCLF, suggesting that high-dose DCLF hepatotoxicity is partially due to bacteria/endotoxin translocation caused by injuring the intestinal barrier.
We thank Dr. Vince Young and Dr. Linda Mansfield for the aid in microbiological evaluations.
This work was supported by Abbott Laboratories and National Institutes of Health Grant DK061315.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
ABBREVIATIONS: IADR, idiosyncratic adverse drug reactions; 15-HPETE, 15-hydroxyperoxyeicosatetraenoic acid; LPS, lipopolysaccharide; PMN, neutrophils; ALT, alanine aminotransferase; IL, interleukin; COX, cyclooxygenase; TVX, trovafloxacin; NSAID, nonsteroidal anti-inflammatory drugs; DCLF, diclofenac; MIP-2, macrophage inflammatory protein 2; MIP-1α, macrophage inflammatory protein-1α; ICAM-1, intercellular adhesion molecule-1; ANOVA, analysis of variance.
- Received June 30, 2006.
- Accepted September 20, 2006.
- The American Society for Pharmacology and Experimental Therapeutics