Failure of ulcer healing may be critically important to the development of serious gastrointestinal complications in patients on long-term NSAIDs.
Failure of ulcer healing may be critically important to the development of serious gastrointestinal complications in patients on long-term NSAIDs.
To determine the effect of indometacin, celecoxib, a cyclooxygenase-2-specific inhibitor, and nabumetone, a pro-drug, on ulcer healing rates in the rat.
Gastric ulcers were induced using a cryoprobe. An NSAID or a vehicle control was administered to groups of eight rats for 3 or 6 days (2 mg/kg indometacin, 9 mg/kg celecoxib or 40 mg/kg nabumetone). The ulcer area was measured and epithelial proliferation at the ulcer margins was measured histochemically. The effect of the drugs on intestinal prostaglandin levels was also assessed.
The mean ulcer sizes in the four groups on day 3 were comparable. On day 6, control animals and those receiving nabumetone showed significant ulcer healing (P < 0.02), while the mean ulcer sizes in the indometacin (P < 0.01) and celecoxib (P < 0.02) groups were significantly larger than those in the control group. Higher doses of nabumetone (160 mg/kg), however, impaired healing. Intestinal prostaglandins were reduced (P < 0.01) only in indometacin-treated animals. The epithelial proliferation index was significantly lower among indometacin- (P=0.02) and celecoxib-treated (P=0.03) animals compared to controls at day 3.
Celecoxib and indometacin both decreased the epithelial proliferative response and delayed healing of cryoprobe-induced gastric ulcers. In contrast, nabumetone impaired ulcer healing only at very high doses.
The administration of traditional NSAIDs is associated with a three-fold increase in the risk of development of significant gastrointestinal complications, depending on previous risk factors.1–3 Cross-sectional studies indicate that gastric or duodenal ulcers are present in 10–25% of patients on long-term NSAIDs.2, 4, 5 Most NSAID-induced ulcers heal spontaneously without treatment, with or without discontinuation of NSAID therapy.6 However, a small proportion of the ulcers produced by NSAIDs progress to clinically serious events, such as a perforation or a bleeding ulcer.2, 7 The critical factors in the pathogenesis of these serious events are unknown, but the particular type and dose of NSAID taken, the location of the ulcer in relation to vessels and the effect of NSAIDs on platelet function may be important, as well as whether there is adequate healing of ulcers.8–11
One of the primary mechanisms responsible for the gastrointestinal toxicity produced by NSAIDs is believed to be suppression of endogenous gastric prostaglandin synthesis. Another is the ‘topical’ effect, which may involve a surface membrane phospholipid interaction and/or an effect to uncouple mitochondrial oxidative phosphorylation.12–16 With the identification17 of two distinct isoforms of cyclooxygenase (COX) responsible for prostaglandin synthesis, COX-1 and COX-2, their roles in the pathogenesis of NSAID-induced gastroduodenal ulcers have been debated. In normal gastric tissue, high levels of COX-1 are expressed, whereas COX-2 expression is undetectable.18 It has been suggested that, because COX-1 participates in mucosal defence, the inhibition of COX-1 may be responsible for NSAID-induced gastrointestinal toxicity. However, recent observations have called into question the precise role of COX-1 inhibition in this damage. COX-1 knockout mice do not develop lesions spontaneously, but do develop ulcers in response to NSAID administration, even though they have no COX-1 to inhibit.19 Also, administration of the selective COX-1 inhibitor, SC-560, reduces gastric prostaglandins20 without gastrointestinal lesions, while dual inhibition of COX-1 and COX-2 leads to gastrointestinal lesions in rats21 and mice.22
A role for COX-2 inhibition is now emerging in adaptive cytoprotection and epithelial integrity.23, 24 More importantly, following ulcer induction, high levels of COX-2 are evident in macrophages and other inflammatory cells along ulcer margins.18, 24 Selective inhibition of COX-2 has been shown to impair ulcer healing.18, 24 Therefore, it is possible that an important factor leading to serious events is impaired ulcer healing, mediated by local inhibition of COX-2.
Traditional NSAIDs have well-documented ulcerogenic properties in both animal models and humans. Celecoxib is a selective COX-2 inhibitor that has been associated with a low incidence of endoscopically determined ulcers25 and a reduction in serious outcomes.26 Nabumetone, a non-acidic pro-drug, is less specific for COX-2, but has also been associated with a very low incidence of gastroduodenal ulcers27 and infrequent complications (perforation and bleeding).28 In this study, we compared the effects of a traditional non-selective NSAID (indometacin), a COX-2-specific NSAID (celecoxib) and a pro-drug (nabumetone) on ulcer healing in the cryoprobe-induced ulcer model.
The study assessed the effect of vehicle (1 mL dimethyl sulphoxide, 10%), indometacin (2 mg/kg, Sigma, Dorset, UK), celecoxib (9 mg/kg, synthesized by SmithKline Beecham Laboratories, Collegeville, PA, USA) or nabumetone (40 mg/kg, SmithKline Beecham Laboratories, Collegeville, PA, USA) on healing rates of cryoprobe-induced gastric ulcers in male Sprague–Dawley rats (200–250 g, Charles River, Kent, UK).
The precise drug dose administered for each experiment of this type is a compromise between giving doses that reflect their use in humans and the administration of comparable anti-inflammatory doses in animals. The 2 mg/kg dose of indometacin in this study was chosen with reference to the maximum recommended dose in humans (75 kg person receiving 150 mg/day). The dose of celecoxib (9 mg/kg) given to the rats in this study, by comparison, corresponds to 675 mg/day (maximum recommended dose in humans is 200 mg/day; however, at the time of this study, doses of 200–800 mg/day were being evaluated, with the highest dose being used in patients with familial adenomatous polyposis). The nabumetone (40 mg/kg) dose corresponds to a dose of 3000 mg/day in humans (maximum recommended dose in humans is 2000 mg/day).
In relation to their anti-inflammatory activity in a standardized carrageenan-induced rat paw oedema model, the oral doses producing 50% inhibition (ED50) of swelling for indometacin, celecoxib and nabumetone are 3, 7 and 30 mg/kg, respectively.29–31
Gastric ulcers were induced on day 0 using a cryoprobe, according to the method described previously.32, 33 Unfasted rats were anaesthetized by induction and maintenance with 3.5% halothane. The abdomen was opened through a 2-cm laparotomy incision. A 3-mm diameter cryoprobe, manufactured from steel, was cooled to equilibrium in liquid nitrogen for 5 min and applied to the serosal surface of the anterior wall of the stomach. The procedure was performed at two separate sites, 8 mm apart, in the gastric body. A single operator induced all ulcers. The laparotomy incision was closed in two layers and the animals recovered on a warmed mat with free access to food and water.
Twenty-four hours following ulcer induction, groups of eight rats were randomly allocated to receive one of the three drugs or vehicle administered by gavage in a single dose at 09.00 hours and thereafter each morning throughout the study. On the test day (either day 3 or day 6), animals were anaesthetized as above 4 h after the last dosing. A 4-cm segment of jejunum (5 cm distal to the ligament of Treitz) was snap frozen and processed for prostaglandin measurements. Stomachs were removed and opened along the lesser curvature and rinsed with distilled water. The ulcers were excised from the stomach with a 2-mm margin and placed on a glass slide with a cover slip. The ulcer area was measured with a 0.25-mm2 ruled light microscope eyepiece. Ulcer areas were measured by a single operator who was blind to the treatment the animal had received. Perforated ulcers were not analysed further.
In a follow-up experiment, eight rats were treated with nabumetone at 160 mg/kg, a four-fold increase in dose. The ulcer size and intestinal prostaglandins were assessed at day 6 only in these animals.
Proximal jejunal, rather than gastric, samples were obtained for prostaglandin measurements, as we sought an index of COX-1 inhibition in tissue free of macroscopic disease. Furthermore, the prostaglandin measurements demand in situ freezing of the tissue, precluding simultaneous assessments of gastric prostaglandins and morphological data. If the intestine is not frozen in situ, spontaneous prostaglandin production from the tissue yields variable and inconsistent control values, depending on the time that the tissue is handled. The snap frozen jejunal samples were ground using a pestle and mortar while in liquid nitrogen, and 100 mg of the ground material was weighed into a pre-cooled Eppendorf tube for prostaglandin extraction. The samples remained frozen throughout. To the frozen samples, 0.5 mL of buffer was added (50 mM phosphate, pH 7.4, 1 mL ethylenediaminetetra-acetic acid, 90 μM indometacin (to act as a cyclooxygenase inhibitor)) and the mixture was vortexed for 30 s and incubated for 15 min at 4 °C. The mixture was then purified and prostaglandin E was extracted according to the Amersham PGE-2 ELISA protocol using 100 mg Amprep C18 minicolumns (Amersham International, Amersham, UK). Purified samples were evaporated to dryness under liquid nitrogen and reconstituted in 20 μL ethanol followed by 480 μL phosphate-buffered saline.
Prostaglandin E levels were determined in duplicate by radioimmunoassay using suitable dilutions of prostaglandin E antiserum (Sigma Chemical Co., Dorset, UK) titrated against tritiated prostaglandin E standards (Amersham International, Amersham, UK).34, 35 Assay sensitivities were 10 pg and the intra- and interassay coefficients of variation were 3% and 2%, respectively. As the prostaglandin E antisera did not distinguish between prostaglandin E1 and prostaglandin E2, the results are expressed as prostaglandin E. Scintillation counts were measured using a Packard 2200CA fitted with a Securia program.
Epithelial cell proliferation was measured using Ki-67 immunohistochemistry.36 Following the measurement of the ulcer area, the ulcer and adjacent tissue were fixed in buffered formalin for at least 24 h and then a representative block was taken from each ulcer through the widest diameter and embedded in paraffin wax. Paraffin sections (2 × 5 μm) were cut from each block. One was stained with haematoxylin and eosin to ensure that the ulcer had been correctly orientated and Ki-67 immunohistochemistry was performed on the other.
For each ulcer, the proliferative index was determined by the number of epithelial cells with positive nuclear staining for Ki-67 per 1000 epithelial cells. Tissue was stained with a Ki-67 antibody, which recognizes a nuclear antigen expressed in all stages of the cell cycle except G0. Sections (5 μm) were cut onto silane-coated slides and placed in a 56 °C oven overnight. Slides were dewaxed, placed in alcohol and endogenous peroxidase activity inhibited by hydrogen peroxide. A pressure cooking antigen retrieval system was used in which the slides were boiled in citrate buffer at maximum pressure for 2 min. The primary antibody, a Ki-67 polyclonal antibody (A0047; DAKO), was applied at a 1:50 dilution for 60 min at room temperature. The slides were then washed in buffered saline and incubated with the third layer, streptavidin–biotin complex, and developed with diaminobenzene. Finally, the slides were counterstained with Mayer’s haemulum, differentiated and mounted.
For each ulcer, the percentage of cells (proliferative index) adjacent to the ulcer which expressed Ki-67 was computed. This was performed by counting the number of positive epithelial cells per 1000 epithelial cells. A cell was deemed to be positive only if it showed positive nuclear staining.
Ulcer areas, prostaglandin E levels and epithelial proliferative index results were normally distributed when assessed by the Shapiro Wilks W-test. Data are therefore presented as mean values with standard errors. Statistical comparisons were made between NSAID groups on day 6 and between day 3 and day 6, using Student’s t-test. The chi-squared test was used to test for statistical significance in frequency tables.
Sixteen ulcers were produced in each group (two ulcers per rat). Ulcers that perforated were not included in the analysis for Ki-67 immunostaining and ulcer area. Four, nine, nine and four ulcers progressed to perforations in the control, indometacin, celecoxib and nabumetone groups, respectively (P=0.1). There was no significant difference (P > 0.5) in the number of perforations seen on days 3 and 6 (16 vs. 10).
Figure 1 shows the mean ulcer size and statistical comparisons for each NSAID treatment group by day. There was no significant (P ≥ 0.05) difference in ulcer size between the groups on day 3.
On day 6, the mean ulcer areas in the control and nabumetone (40 mg/kg) groups, but not in the indometacin or celecoxib groups, were significantly (P < 0.02) smaller compared to the mean ulcer size at day 3 in the respective groups. On day 6, the mean ulcer size in the indometacin (P < 0.01) and celecoxib (P < 0.02) groups was significantly larger than that in the control group. In contrast, there was no significant difference in mean ulcer size between the nabumetone (40 mg/kg) and control groups or between the indometacin and celecoxib groups at day 6. However, when given at a dose of 160 mg/kg (equivalent to 12 g per 75 kg in humans), nabumetone impaired healing in a manner similar to indometacin at 2 mg/kg and celecoxib at 9 mg/kg.
Expressed as a percentage of the mean ulcer size (which is only an estimate of healing as the same animals were not assessed at days 3 and 6) in animals studied at day 3 vs. day 6 for each treatment group, the changes in ulcer size were 45%, 19%, 23% and 35% for controls, indometacin, celecoxib and nabumetone (40 mg/kg), respectively.
Figure 2 shows that the prostaglandin E levels were significantly lower in indometacin-treated rats compared to controls on days 3 (P < 0.0001) and 6 (P < 0.01). In contrast, prostaglandin E levels did not decrease in the celecoxib- or nabumetone-treated (40 or 160 mg/kg) rats compared to controls.
Figure 3 shows that the epithelial proliferation was significantly lower among indometacin- (P < 0.02) and celecoxib-treated (P < 0.03) animals compared to controls at day 3. Nabumetone (40 mg/kg) did not lower epithelial proliferation significantly (P=0.5) compared to controls at day 3. At day 6, there were no significant differences between any of the treatment groups.
We have shown that indometacin and the selective COX-2 inhibitor celecoxib delay healing of cryoprobe-induced gastric ulcers over a 6-day period. The same was evident with the high dose of nabumetone (160 mg/kg). In contrast, at a dose of 40 mg/kg, nabumetone did not delay healing of cryoprobe-induced gastric ulcers. Inhibition of ulcer healing with both celecoxib and indometacin was associated with a decrease in proliferative response at day 3 that had normalized by day 6. Only indometacin was associated with significant decreases in intestinal prostaglandin E levels.
A number of factors have been implicated in the pathogenesis of the damage induced by NSAIDs to the gastrointestinal tract. Somasundaram et al.37 suggest that one of the main mechanisms is a ‘topical’ insult which involves the uncoupling of mitochondrial oxidative phosphorylation, thereby beginning a cascade of events that can result in cellular damage and sometimes death.37, 38 Immediate local prostaglandin synthesis via the constitutive COX-1 enzyme normally enlists multiple defensive mechanisms, such as enhanced blood flow, mucus production and bicarbonate secretion, that help to defend against these aggressive forces. However, NSAID-induced inhibition of COX-1 prevents this immediate defence mechanism. Once the cellular injury has occurred and progressed to erosions and ulcers, a second phase of defensive mechanisms commences. A number of reparative processes are initiated and implicated in the healing of ulcers. At the ulcer margin, COX-2 messenger RNA is markedly increased23, 39 and epithelial cell proliferation increases.40 The high levels of COX-2 messenger RNA during the acute stages of gastric ulcer formation are believed to be related to the repair processes,23 as the COX-2 activity is localized to reparative cells.39
NSAIDs can interfere in the process of gastric repair at a number of points. Most traditional NSAIDs, such as indometacin, are weak organic acids and can initiate direct topical cellular damage following ingestion. In addition, their ability to inhibit COX-1, at concentrations present in the gastric mucosa, prevents the immediate prostaglandin defence mechanisms. Selective COX-2 inhibitors can, in theory, interfere with gastric ulcers by preventing the second phase of inflammatory-mediated repair processes.
Our studies show that indometacin and celecoxib significantly delayed ulcer healing, whereas comparable doses of nabumetone (40 mg/kg) did not, and that this correlated with the effects of the drugs on epithelial proliferation at day 3. The epithelial proliferative index was inhibited on day 3 only by indometacin and celecoxib. Previous studies33, 39 have demonstrated inhibition in epithelial proliferative activity, as assessed by the bromodeoxyuridine technique, by both non-selective COX inhibitors and by selective COX-2 inhibitors. The maximal proliferative rate appears to be early in the healing phase,39 and may explain why a difference is not seen at 6 days. In addition, it may be that the Ki-67 technique is less sensitive than the bromodeoxyuridine technique.
Indometacin decreased intestinal prostaglandins significantly, while celecoxib and nabumetone did not, suggesting that COX-1 inhibition is not the main mechanism that impairs healing of these ulcers. Furthermore, it is unlikely that the ‘topical’ effect played a major role in the delayed healing, as celecoxib and nabumetone are both non-acidic and therefore do not have this effect.41 The reason why celecoxib impaired the healing of the ulcers, while nabumetone, given at comparable anti-inflammatory doses, did not is intriguing. One possibility is that nabumetone, being a pro-drug, only inhibited gastric COX-2 in the systemic circulation after its conversion to the active component, 6-methoxy-2-naphthylacetic acid, by the liver. Celecoxib, on the other hand, being an effective COX-2 inhibitor, may have inhibited gastric COX-2 activity profoundly during absorption, when a much higher local concentration of the drug may occur, than in the systemic circulation after absorption has taken place. Alternatively, our results could be explained by a lower bioavailability of nabumetone compared with indometacin and celecoxib. However, previous studies of the metabolism of nabumetone have demonstrated that approximately 70% of a 20-mg/kg intragastric dose of nabumetone appears as 6-methoxy-2-naphthylacetic acid in the plasma of rats, suggesting that nabumetone is sufficiently bioavailable in this study at the lower dose.42 However, with administration of a very high dose (160 mg/kg), ulcer healing was delayed. It may therefore be that, at this very high dose, local concentrations of 6-methoxy-2-naphthylacetic acid, derived from the systemic circulation, increase sufficiently to inhibit COX-2 and impair ulcer healing.
Lastly, it is conceivable that delayed healing is due to the inhibition of a postulated COX-3,43 or processes that are not a result of the effect of COX inhibition during drug absorption. In this context, it is interesting that conventional NSAIDs and selective COX-2 agents impair angiogenesis,44 which is thought to play an important role in healing. It is not entirely clear whether this inhibitory effect is mediated by COX-2,45, 46 as it is only partially abolished by high prostaglandin concentrations.47 Furthermore, NSAIDs and selective COX-2 inhibitors have an antitumour action, but it is interesting that this may not be related to the COX inhibitory effects of these drugs.48, 49
One implication of our observations, namely delayed healing of pre-existing lesions as opposed to ulcerogenic effects, may be relevant to Helicobacter pylori ulcers and other inflammation within the gastrointestinal tract, such as inflammatory bowel disease. Most of such data, however, comes from experimental animals which suggests an effect on angiogenesis and/or COX-2.18, 44, 50 That conventional NSAIDs interfere with the healing of gastroduodenal ulcers33, 51 and cause relapse of inflammatory bowel disease52, 53 in humans is not in question. The mounting evidence that selective COX-2 agents delay healing of gastrointestinal inflammation and ulcers in experimental animals provides sound justification for similar studies in humans.
In conclusion, the results of these studies in the cryoprobe ulcer healing model suggest that the administration of celecoxib, a selective COX-2 inhibitor (as well as indometacin, a dual COX inhibitor), may result in significant impairment of healing of established ulcers, even though it may not initiate new ulcer formation. In contrast, nabumetone, a non-acidic pro-drug that is converted to a metabolite with both COX-1 and COX-2 inhibitory characteristics, given in similar anti-inflammatory doses, is not associated with significant impaired healing.
This research was supported in part by SmithKline Beecham Pharmaceuticals.