Abstract
The therapeutic potential of drugs that block the induction of cyclooxygenase-2 has been emphasized. When two 4-trifluoromethyl salicylate derivatives [2-acetoxy-4-trifluoromethyl-benzoic acid (triflusal) and its deacetylated metabolite 2-hydroxy-4-trifluoromethylbenzoic acid (HTB)] were compared with aspirin and sodium salicylate as cyclooxygenase-2 (COX-2) inhibitors, we observed that in bacterial lipopolysaccharide-activated human blood, triflusal, aspirin, and HTB, but not sodium salicylate, inhibited COX-2-mediated prostaglandin E2 (PGE2) production (IC50 = 0.16, 0.18, 0.39, and >10 mM, respectively). However, only triflusal and aspirin inhibited purified COX-2 enzyme. To test this apparent discrepancy, we realized that HTB and triflusal (but neither aspirin nor salicylate) produced a concentration-dependent inhibition of COX-2 protein expression in peripheral human mononuclear cells. This observation was further confirmed in a rat air pouch model in vivo, in which both aspirin and triflusal inhibited PGE2 production (ID50 = 18.9 and 11.4 mg/kg p.o., respectively) but only triflusal-treated animals showed a decrease in COX-2 expression. This different behavior may be, at least in part, due to the ability of HTB and triflusal to block the activation of the transcription factor nuclear factor-κB to a higher extent than aspirin and sodium salicylate. Thus, in addition to inhibiting the COX-2 activity at therapeutic concentrations, triflusal is able to block through its metabolite HTB the expression of new enzyme, and hence the resumption of PGE2synthesis. Triflusal and HTB may exert beneficial effects in processes in which de novo COX-2 expression is involved and, in a broader sense, in pathological situations in which genes under nuclear factor-κB control are up-regulated.
Cyclooxygenase (COX) is the rate-limiting enzyme in the conversion of arachidonic acid to prostaglandin (PG)H2, the precursor of a wide group of biologically active mediators such as PGE2, prostacyclin, and thromboxane A2. Two isoforms of this enzyme have been identified (Smith et al., 1996): COX-1, a constitutively expressed enzyme, and COX-2, an isoform induced in response to many proinflammatory agents, as well as mitogens and tumor promoters. COX-2 is expressed in activated macrophages, monocytes, and several other cell types (Mitchell et al., 1995) and has been identified in chronic inflammatory conditions in vivo (Vane et al., 1994). Moreover, recent discoveries have shown that COX-2 seems implicated in either physiological processes such as ovulation (Sirois and Dore, 1997) and delivery (Sawdy et al., 1997; Lim et al., 1997) or pathological states, such as colonic cancer, Alzheimer’s disease, heart failure, and even hypertension (Levy, 1997; Oka and Takashima, 1997; Hartner et al., 1998; Wong et al., 1998). In this scenario, there has been great interest in the role or roles of COX-2 and the usefulness of drugs that selectively block this isoenzyme (Frölich, 1997; see alsoPennisi, 1998).
The first anti-inflammatory synthetic drug, which remains the most popular worldwide, is aspirin. Despite this, the mechanism by which aspirin and salicylate derivatives exert their therapeutic effects, far from being established, remains open to debate. In the early 1970s,Vane (1971) demonstrated that sodium salicylate, aspirin, and other commonly used NSAIDs inhibited cyclooxygenase. From then on, numerous studies have supported the hypothesis that this is how salicylates work (Mitchell et al., 1997). However, there are some doubt as to the molecular basis for the therapeutic properties of salicylate derivatives, and due to recent descriptions of the surprising effects of these drugs on some central cell regulatory events and gene regulation processes, the mechanism of action of salicylates, in both the inflammatory response and the atherothrombotic processes, remains under discussion (Frölich, 1997).
Salicylates have been shown to inhibit the mitogen-activated protein kinase cascade (Schwenger et al., 1996) and the expression of several key enzymes as type IIA phospholipase A2(Vervoordeldonk et al., 1996) or the inducible isoform of NO synthase (Farivar and Brecher, 1996). A molecular mechanism by which salicylates could exert part of these effects was proposed by Kopp and Ghosh (1994)when they reported that salicylates inhibited the transcription factor nuclear factor-κB (NF-κB) activation through the stabilization of IκB. Because NF-κB regulates the expression of many genes, this mechanism of action was rapidly implicated in the effects of salicylates on inflammation (Pierce et al., 1996), atherogenesis (Weber et al., 1995), or neuroprotection (Grilli et al., 1996).
COX-2 is an early gene expressed in response to many cytokines, and its transcriptional regulation is, at least in part, under the control of NF-κB (Newton et al., 1997). Thus, the inhibition of this transcription factor by salicylates could block COX-2 expression. However, several studies failed to demonstrate an inhibition of COX-2 expression by aspirin and other NSAIDs in various experimental models and species (O’Sullivan et al., 1993; Barrios-Rodiles et al., 1996). Furthermore, recent studies suggest that sodium salicylate exerts its anti-inflammatory action because it is a weak competitive inhibitor of arachidonic acid in the active site of the enzyme, rather than via its action on NF-κB activation (Loll et al., 1995; Mitchell et al., 1997). Another controversial aspect of the newly reported effects for aspirin and salicylate is the high concentration, higher than that obtained after therapeutic doses, needed to elicit significant effects (Kopp and Ghosh, 1994; Mitchell et al., 1997).
Nevertheless, the previous observations that salicylates inhibit the NF-κB-mediated gene expression prompted us to analyze the potential inhibition of the bacterial lipopolysaccharide (LPS)-induced COX-2 expression by triflusal and its deacetylated metabolite, HTB (Fig.1). Triflusal is an antiplatelet drug that despite its structural analogy to aspirin, exhibits different pharmacological and pharmacokinetic properties (McNeely and Goa, 1998). Here, we report that triflusal and HTB at therapeutic doses are potent inhibitors of NF-κB activation, de novo COX-2 synthesis, and PGE2 production in both human PBMCs and a rat model of inflammation in vivo.
Materials and Methods
Reagents.
Tissue culture media and fetal bovine serum were purchased from GIBCO BRL Life Technologies S.A. (Madrid, Spain). Nonidet P-40 was obtained from Boehringer Mannheim GmbH (Mannheim, Germany), and nitrocellulose membranes were purchased from Bio-Rad Laboratories S.A. (Barcelona, Spain). Bicinchoninic acid protein assay and molecular weight protein standards were obtained from Pierce Chemical (Rockford, IL). Enzyme-linked immunosorbent assay (ELISA) kits specific for PGE2 and thromboxane B2 were purchased from Amersham Ibérica S.A. (Madrid, Spain). Tris, toluene, Tween-20, and dimethyl sulfoxide were from Merck Quı́mica S.A. (Barcelona, Spain). 2-Acetoxy-4-trifluoromethylbezoic acid and 2-hydroxy-4-trifluoromethylbenzoic acid were from URIACH Laboratories (Barcelona, Spain). Arachidonic acid (peroxide free) was purchased from Cayman Chemical Co. (Ann Arbor, MI) and stored at −20°C. Reagents for electrophoresis were obtained from Amersham Pharmacia Biotech GmbH (Madrid, Spain). LPS from Escherichia coli serotype 026:B6 (LPS) and all the other nonspecified reagents were purchased from Sigma-Aldrich Quı́mica S.A. (Madrid, Spain).
Whole Human Blood Assays.
Human whole blood (heparinized, 10 U/ml) was obtained at the Hospital de Sant Pau (Barcelona, Spain) from donors whom had no apparent inflammatory conditions and had not taken any NSAIDs during the previous 2 weeks. To determine COX-2 activity, the blood was aliquoted (1 ml) and incubated with compound (10−5 to 5 × 10−3M) or vehicle (dimethyl sulfoxide, 0.5% final) in the absence or presence of LPS (10 μg/ml) for 24 h at 37°C by gentle shaking (Patrignani et al., 1994). After the incubation, plasma was separated by centrifugation (10 min at 1000g, 4°C) and kept at −70°C until assayed for PGE2 by specific ELISA.
Enzyme Activity Determinations.
Cyclooxygenase activity (COX-2 from sheep placenta; Cayman Chemical Co.) was determined by measuring spectrophotometrically the turnover of the chromogenN,N,N′,N′-tetramethyl-p-phenylenediamine (TMPD) during the reduction of PGG2 to PGH2. The reaction mixture (1 ml) contained 1 μM hematin and 2 mM phenol as cofactors in Tris·HCl buffer (100 mM, pH 8.1). The enzyme (30–40 units) was preincubated at 37°C with 5 μl of the vehicle alone (controls) or 5 μl of the test compounds in dimethyl sulfoxide for 1 to 10 min at 37°C. The reaction was started by the addition of freshly prepared arachidonic acid (100 or 10 μM) and TMPD (170 μM) and measured by the change in absorbance at 611 nm for 30 s. The initial rate (linear for approximately 15 s) was measured, and the nonspecific rate of oxidation in the absence of enzyme was subtracted before calculation of the percentage of inhibition.
Isolation and Culture of Human Mononuclear Cells.
Human peripheral blood mononuclear cells (PBMCs) were isolated from whole human blood by density gradient centrifugation in Ficoll solution (d = 1.077 g/ml, Biochrom KG; Cultek S.L., Madrid, Spain). Then, 50-ml centrifuge tubes were filled with 15 ml of Ficoll. Heparinized (10 U/ml) human whole blood was diluted in equal parts with PBS, and 25 ml was carefully poured over the Ficoll solution. The tubes were then centrifuged for 20 min at 1200g, and the mononuclear cells (80–100% recovery) were collected from the whitish interphase between the plasma and the separating solution. Cells were washed twice with PBS (first cycle, 10 min at 300g; second cycle, 10 min at 200g) and then resuspended (3 × 106 cells/ml) with culture medium (RPMI-1640, supplemented with 10% heat-inactivated FCS).
Cells were placed in appropriate plastic culture plates, incubated for 1 to 2 h at 37°C in 5% CO2/air, and then treated with the test compounds for 24 h. Cell counts were performed with a Celltak α MEK-6108-K (Radiometer S.A., Madrid, Spain), and standard Wright-Giemsa staining was performed to examine by light microscopy whether the cells displayed the morphological features of viable human mononuclear cells. Monocytes compose an average of 10% of PBMCs. The remaining 90% of cells were mostly lymphocytes that do not normally express COX (Spencer et al., 1998). Before the 24-h incubation, cell viability exceeded 95% as determined with trypan blue. Cell viability at the end of the incubation was determined by trypan blue exclusion and with a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT) assay, as described later.
Immunoblot Assays.
Human PBMCs, stimulated or activated with LPS and in the presence or absence of the drugs tested, were transferred to centrifuge tubes and pelleted by centrifugation for 5 min at 1200g. The supernatants were collected and stored at −70°C for later determination of PGE2 by specific ELISA. Pelleted cells were washed twice with PBS, and an aliquot (0.5 ml) was used for the MTT assay. The remainder (2 ml; ∼5 × 106 cells) was recentrifuged (600g, 10 min), the PBS was removed, and the cells were resuspended by gentle shaking in 50 μl of lysis buffer (1% Nonidet-40, 1 mM EDTA in PBS) and incubated for 15 min in ice. The detergent-insoluble material was removed by centrifugation at 20,000g for 15 min at 4°C, and the protein concentration of the supernatants was determined by the bicinchoninic acid protein assay reagent. The cell extracts were then boiled (5 min) in a ratio of 1:1 with gel-loading buffer [50 mM Tris, 10% (w/v) SDS, 10% (v/v) glycerol, 50 μl/ml 2-mercaptoethanol, and 2 mg/ml bromophenol blue]. Samples were centrifuged at 10,000g for 2 min, loaded (50 μg/lane) onto a 4% stacking gel/7.5% separating, and subjected to electrophoresis (1 h at 200 V). The separated proteins were transferred to nitrocellulose membranes (Bio-Rad; 2 h at 100 V) and overnight blocking at 4°C with PBS buffer containing 0.1% (w/v) Tween-20 and 20% dried lowfat milk. The membranes were washed with PBS containing 0.05% Tween-20 and incubated 1 h in a goat polyclonal antibody raised against purified human COX-1 or COX-2 protein (Santa Cruz Biotechnology Inc., Santa Cruz, CA). The blots were washed three times (PBS, 0.05% Tween-20) and incubated for 30 min with a horseradish peroxidase-conjugated secondary antibody raised in rabbit against goat (Pierce). Nitrocellulose membranes were washed twice, and bound antibody was visualized by enhanced chemiluminescence (ECL Kit; Amersham Ibérica S.A.).
Viability Determination (MTT Assay) .
The viability of cells treated with any of the compounds tested was determined by assaying the ability of mitochondrial dehydrogenases to convert the soluble tetrazolium salt, MTT, into an insoluble purple formazan through cleavage of the tetrazolium ring. After 24-h incubation with vehicle or the drug tested, the cells were counted (Celltak α MEK-6108-K) and washed two times with PBS. Then, the cells were resuspended (half the original volume) in a medium with MTT (Sigma) at a concentration of 5 mg/ml and without phenol red or FBS. Cells were incubated for 4 h at 37°C in 5% CO2. At the end of the incubation, the medium was removed, the cells were washed once, and the converted purple dye was solubilized with dimethyl sulfoxide at the same volume. Absorbance of converted dye was measured at 550 nm with background subtraction at 630 nm in a microplate reader (Ceres UV900 HD; Bio-tek Instruments Inc., Winooski, VT).
Electrophoretic Mobility Shift Assay.
Human PBMCs were washed with ice-cold hypotonic lysis buffer (10 mM HEPES-KOH, pH 7.9, 10 mM KCl, 1.5 mM MgCl2, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 5 μg/ml aprotinin, 5 μg/ml leupeptin, and 0.6% Nonidet P-40). The cells were allowed to swell on ice for 10 min and vortexed vigorously for 10 s. Unbroken cells were eliminated by centrifugation (1000g, 10 min, 4°C), and the nuclei were collected by centrifugation at 15,000gfor 1 min in a microcentrifuge. The nuclear pellet was resuspended in high salt extraction buffer containing 25% glycerol and 0.5 M KCl, and the nuclear extract was obtained by pelleting for 30 min at 105,000g in a Optima TL ultracentrifuge (Beckman) using a TLA 100.2 rotor. Then, 22-mer double-stranded oligonucleotide probes containing NF-κB sequence were end-labeled with [γ-32P]ATP using T4 polynucleotide kinase and separated from the unincorporated label by minicolumn chromatography. The kilobase sequence used was sense, 5′-AGTTCAGGGGAATTTCCCAGGC-3′, and the complement, 5′-GCCTGGGAAATTCCCCTGAACT-3′. Nuclear protein (10 μg) was incubated for 20 min on ice with radiolabeled oligonucleotide probes (2–6 × 104 cpm) in a 25-μl reaction buffer containing 2 μg of poly(dI/dC), 10 mM Tris·HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 8% Ficoll, and 4% glycerol. Nucleoprotein-oligonucleotide complexes were resolved by electrophoresis in a 4% nondenaturing polyacrylamide gel in Tris-borate/EDTA buffer at 175 V for 3 h at 4°C. The gel was dried and autoradiographed with an intensifying screen at −80°C for 2 to 12 h. The specificity of the DNA-protein complex was confirmed by competition with a 300-fold molar excess of unlabeled nucleotide containing the consensus sequences. Quantification of the DNA-protein complex containing the NF-κB sequence was carried out by densitometric scanning using software of the series Discovery 3.1 from PDI-Pharmacia.
Air Pouch Model of Inflammation.
For determination of PGE2 in exudate and stomach, male Lewis rats (175–200 g) were used. Air cavities were produced by a s.c. injection of 20 ml of sterile air into the intrascapular area. Every 2 days, 10 ml of air was injected again into the cavity to keep the space open. Seven days after the first injection, 2 ml of a 1% solution of λ-carrageenan (Sigma) in saline was injected into the air pouch to produce an inflammatory reaction. The animals were sacrificed 24 h later, and the volume of exudate was measured. The type and number of cells present in the exudate were determined with a Coulter Counter and standard WrightGiemsa staining. Cells were pelleted by centrifugation at 1200g for 5 min at 4°C, and PGE2 was determined in the supernatant by specific ELISA. Cells were resuspended in 2 ml of saline and submitted to a hypotonic shock with 6 ml of distilled water for 20 s to selectively eliminate red blood cells. The isotonicity of the solution was restored with 2 ml of 3.5% (w/v) NaCl, and the cells were collected by centrifugation (1000g, 10 min, 4°C), resuspended in lysis buffer at 2 × 108cell/ml, and used for immunoblotting (see above).
Immediately after the exudates were collected, the stomachs were excised and frozen at −70°C. The day of analysis, stomachs were homogenized in 70% ethanol. The homogenates were centrifuged (1000g, 10 min, 4°C), and the supernatants were dried under N2 stream and resuspended in ELISA buffer for PGE2 determination.
Results
Effects of Salicylates on Whole Human Blood Assays.
Heparinized (10 U/ml) whole human blood obtained from healthy donors was treated in vitro with sodium salicylate, HTB, triflusal, aspirin, and other NSAIDs. The COX-2 activity was measured as PGE2 production in blood incubated with LPS (10 μg/ml) for 24 h. Under unstimulated conditions, plasma PGE2 averaged 0.8 ± 0.5 ng/ml (mean ± S.D., n = 15). When the blood was incubated with LPS for 24 h, an substantial increase in PGE2production was observed (11.5 ± 2.3 ng/ml, n = 15). Sodium salicylate, HTB, triflusal, and aspirin inhibited LPS-induced PGE2 increase in a dose-dependent manner (Fig. 2). HTB was much more potent than salicylate, with IC50 values of 0.39 ± 0.09 and >10 mM, respectively. On the contrary, both aspirin and triflusal behaved as equipotent COX-2 inhibitors in this assay with similar IC50 values (Table1). The nonselective COX inhibitor indomethacin and the selective COX-2 inhibitor NS-398 totally abrogated PGE2 production with apparent IC50 values far lower than 1 μM (Table 1), proving that the PGE2 increase in blood incubated with LPS is produced mainly by a COX-2-dependent mechanism, as previously reported (Patrignani et al., 1994).
Effects of Salicylates on Purified COX-2 Enzyme.
To determine whether COX-2 enzyme activity was modified by sodium salicylate, aspirin, triflusal, or HTB, the effect of these compounds on the oxidation of arachidonic acid (100 and 10 μM) by the purified sheep COX-2 enzyme was examined. The results obtained with these compounds at a substrate concentration of 100 μM, as well as those obtained with indomethacin and NS-398, are compared in Table 1. Aspirin and triflusal showed a time-dependent profile of inhibition with a maximal effect after 10 min of preincubation with the enzyme. Neither sodium salicylate nor HTB at concentrations up to 10 mM significantly affected COX-2 activity independently of the time of preincubation (Table 1). When the lowest concentration (10 μM) of substrate was used, the profiles of aspirin, triflusal, HTB, and sodium salicylate were not modified (data not shown). As expected, both indomethacin and NS-398 completely blocked COX-2 activity, with apparent IC50 values of 0.98 and 0.042 μM, respectively (Table 1).
Induction of COX-2 Expression by LPS in Human PBMCs.
COX-2 was not detected by immunoblotting in untreated human PBMCs. Overnight incubation of these cells also failed to induce COX-2 expression. However, LPS (10 μg/ml) produced an up-regulation of COX-2 protein. To optimize conditions for LPS-induced COX-2 expression in cultured PBMCs, a dose- and time-response study was performed. As shown in Fig.3A, cells exposed overnight to LPS (0.1–50 μg/ml) showed a dose-dependent increase in COX-2 levels. On the other hand, COX-2 expression is detected as early as 2 h after stimulation with 10 μg/ml LPS and reached the maximum expression after 19 h of incubation (Fig. 3A). PGE2generation almost paralleled COX-2 protein expression (Fig. 3B). Modification of COX-1 expression was not detected in any of the conditions tested (data not shown).
Effects of Exposure to Salicylates on PGE2 Release and COX-2 Expression in Human PBMCs Stimulated by LPS.
Human PBMCs, either untreated or LPS activated, were assayed for PGE2 released into the medium. As shown in Fig.3B, untreated PBMCs released low levels of PGE2after 19-h incubation (3 ± 0.5 ng/ml), but exposure to 10 μg/ml LPS led to a dose-dependent increase in PGE2production with a ∼5-fold increase (15 ± 3 ng/ml). The COX-2-selective inhibitor NS-398 totally prevented the LPS-induced PGE2 production (data not shown), confirming that COX-2, and not COX-1, is involved in this process.
When incubated together with LPS, sodium salicylate or aspirin (at concentrations up to 5 mM) had no effect on the LPS-induced COX-2 expression (Figs. 4A and5A). On the other hand, aspirin, but not sodium salicylate, completely inhibited PGE2release (Figs. 4B and 5B). Similarly, the COX-2-selective inhibitor NS-398 totally inhibited PGE2 release but failed to prevent COX-2 expression at concentrations up to 50 μM (data not shown). However, at the same concentration range tested for the other salicylates, HTB and triflusal caused an inhibition of PGE2 release and COX-2 expression in a concentration-dependent manner (Figs. 4 and 5). Cell viability measured by trypan blue exclusion and with the MTT assay was more than 90% of the control regardless of the treatment used.
Inhibition of NF-κB Activation by Salicylates in Human PBMCs Stimulated by LPS.
To determine the possible implication of the transcription factor NF-κB on the inhibition of COX-2 expression in PBMCs, we compared the effect of different concentrations of aspirin, salicylate, triflusal, and HTB on the activation of NF-κB in LPS-activated PBMCs. As shown in Fig. 6A, the incubation of PBMCs with LPS induced a rapid and prominent activation of NF-κB as judged from the appearance of κB-binding activity in the nuclear extracts. The mean increase of the binding activity elicited by LPS was 3.5 ± 0.6-fold (n = 4, Fig. 6B) and was more prominent than that produced by TNF-α, an agonist that has been used in several studies dedicated to assessment of the effect of salicylates on NF-κB activation. Figure 6C shows that incubation for 15 min with either sodium salicylate or HTB before the addition of 10 μg/ml LPS produced a significant reduction in κB-binding activity in nuclear extracts of these cells. Attempts to compare the potency of these compounds were carried out by densitometric scanning of the protein/oligonucleotide complexes in three independent experiments. Figure 6D shows that 3 mM HTB produced an inhibition of 85 ± 7% of κB-binding complexes, whereas the inhibition observed for the same concentration of sodium salicylate was 35 ± 12%. Similar experiments were carried out with aspirin and triflusal (electrophoretic mobility shift assay; not shown), and in keeping with previous findings in human umbilical vein endothelial cells (Y. Bayón, A. Alonso, and M. Sanchez-Crespo, unpublished observations), triflusal was also more potent than aspirin in the inhibition of κB-binding activity in PBMCs (87 ± 3% and 48 ± 5% of inhibition at 3 mM, respectively; n = 3).
Effects of Triflusal and Aspirin on COX-2 Expression and Activity in Infiltrating Cells in the Rat Air Pouch.
Carrageenan induced an outstanding prostaglandin release into the exudate fluid when it was administered into the intrascapular area of the control rats (72.3 ± 14.7 ng/ml). Vasodilation, edema, and significant increase in cellular influx were also observed after carrageenan injection. Polymorphonuclear leukocytes, monocyte/macrophages, and lymphocytes accounted for 86%, 12%, and 2% of the infiltrating cells, respectively.
The administration of triflusal (3–30 mg/kg p.o.) to rats (Fig.7) inhibited in a dose-dependent manner COX-2 protein expression in the cells present in the exudate and caused a decrease in PGE2 levels with an ID50 value of 11.4 mg/kg. Aspirin at the same doses did not exert any detectable effect on COX-2 expression, although it inhibited PGE2 release to the same extent than triflusal (ID50 = 18.9 mg/kg). In contrast, the inhibitory effect of triflusal on PGE2 production in stomachs was 6-fold lower than that of aspirin (ID50 = 13.9 and 2.3 mg/kg, respectively), as shown in Fig. 8.
Discussion
The search for selective COX-2 inhibitors started after the identification of this cytokine-inducible isoform of cyclooxygenase (Fu et al., 1990). So far, more than a dozen of these compounds have been described (Frölich, 1997), all of them behaving as competitive inhibitors of the enzyme. An alternative approach chosen by some research groups was to develop irreversible COX-2-selective inhibitors (Kalgutkar et al., 1998) that could prevent the resume of prostaglandin production once the drug plasma levels fall, in the same way as aspirin does with COX-1 and, to a lesser extent, with COX-2. Moreover, because of the fast response of COX-2 to induction by proinflammatory stimuli, another interesting approach could be to identify drugs that selectively block the expression of the enzyme (Pennisi, 1998).
In this context, the initial aim of the present study was to test the ability of two salicylate derivatives, triflusal and its deacetylated metabolite HTB, to inhibit COX-2-mediated PGE2 synthesis in several experimental models. As a result of this study, the mechanism underlying PGE2 synthesis inhibition by triflusal and HTB was elucidated, and it turned out to be clearly different than those of their counterparts aspirin and salicylate.
In a whole human blood assay in vitro, plasma PGE2 levels increased 15-fold after 24-h incubation with 10 μg/ml LPS. Triflusal, aspirin, and HTB blocked this increase with apparent IC50 values (Table 1) well within plasma concentrations found in the therapeutic use of triflusal or HTB, which can approach 1 mM (McNeely and Goa, 1998). Sodium salicylate in the range of doses tested (up to 5 mM) failed to exert any effect on this LPS-mediated PGE2synthesis. On the other hand, only aspirin and triflusal inhibited the purified COX-2 isoenzyme activity in a time-dependent fashion, whereas HTB and sodium salicylate were ineffective. Lack of inhibition of COX-2 activity by sodium salicylate has been reported by others (Frölich, 1997). It has also been reported that some salicylates are competitive inhibitors of arachidonic acid for the active site of the COX-2 (Frölich, 1997; Mitchell et al., 1997), and thus the substrate concentration becomes a rate-limiting factor. To evaluate the influence of this factor, two different concentrations of arachidonate (10 and 100 μM) were used. Neither HTB nor salicylate had any inhibitory effect, even at the lowest level of arachidonic acid tested.
Thus, the inhibition of LPS-induced PGE2 release in human whole blood by HTB cannot be explained in terms of enzyme inhibition, and other explanations, such as a decrease in enzyme expression, had to be considered. Indeed, before the full establishment of the COX-1/COX-2 concept, Wu et al. (1991) showed that aspirin (and sodium salicylate) at concentrations as low as 60 nM partially inhibited the induction of COX in human umbilical vein endothelial cells (HUVECs) stimulated by interleukin-1. Moreover, the inhibition of COX expression was almost complete at an aspirin concentration of 2.7 mM (0.5 mg/ml). However, these striking results could not be reproduced by others using murine fibroblasts activated with phorbol ester (Kujubu and Herschman, 1992), rabbit alveolar macrophages stimulated with LPS (O’Sullivan et al., 1993), human macrophages stimulated with LPS (Barrios-Rodiles et al., 1996), the human pulmonary epithelial line A549 activated with interleukin-1β (Mitchell et al., 1997), or quiescent murine NIH 3T3 cells stimulated by the addition of 10% FCS (Spencer et al., 1998) or in an experimental system in which COX-2 expression in endothelial cells (HUVECs) was induced by coculture with tumor cells (Tsujii et al., 1998), despite the fact that much higher aspirin and/or sodium salicylate concentrations were tested. This disagreement could be due to a cell-specific regulation by salicylates of COX-2 expression, or it could reflect a stimulus-dependent response, as has been reported for the inhibition of tumor necrosis factor but not epidermal growth factor-induced mitogen-activated protein kinase activation by sodium salicylate (Schwenger et al., 1996). Nevertheless, because the human blood method we used is clearly dependent on LPS-induced de novo expression of COX-2 (Patrignani et al., 1994), we hypothesized that HTB and triflusal could exert their effect on PGE2 synthesis by suppressing COX-2 expression. This hypothesis was confirmed using human PBMCs, in which PGE2 release and COX-2 expression were dose and time dependently activated by LPS (Fig. 3). HTB produced a concentration-dependent inhibition of both the PGE2 release and the COX-2 expression. The inhibitory effect on LPS-induced COX-2 expression was evident at 0.1 mM and was complete at 3 mM. Unlike HTB, sodium salicylate, at concentrations up to 5 mM, had no effect on COX-2 expression. In addition, triflusal, but not aspirin, inhibited COX-2 expression to the same extent as HTB.
Provided that human COX-2 promoter has two putative NF-κB motifs and that NF-κB seems to be involved in the induction of COX-2 expression (Hwang et al., 1997; Newton et al., 1997), we wondered to what extent the effect of triflusal and HTB on COX-2 expression was mediated by the suppression of the LPS-inducible NF-κB activation. Our data show that the inhibitory effect of triflusal and HTB in NF-κB activation and COX-2 expression paralleled each other. Accordingly, NF-κB is clearly involved in the regulation of COX-2 expression in LPS-stimulated macrophages (D’Acquisto et al., 1997). However, the significant effect observed in NF-κB activation for aspirin and salicylate (Kopp and Ghosh, 1994; Weber et al., 1995) does not correspond with an observed similar effect on COX-2 expression in either this study or others (as mentioned). It agrees with a more complex regulation of COX-2 expression in which other transcription factors, as activator protein-1, NF-interleukin-6, or cAMP response element-binding protein, may be involved (Inoue et al., 1995, Miller et al., 1998).
In any case, HTB and triflusal are clearly more potent inhibitors of NF-κB than aspirin and sodium salicylate, showing a complete inhibition of κB binding to the nuclear fraction at concentrations of 3 mM. Recently, the effect of both aspirin and salicylate on NF-κB activation has been mechanistically related to the selective inhibition of IκB kinase-β (Yin et al., 1998) by interfering with ATP binding to the enzyme. It seems likely that a similar or related mechanism could explain the effect of both triflusal and HTB.
Next, we investigated whether the inhibition of COX-2 expression observed in the human PBMCs treated with triflusal or HTB in vitro can also be seen in a model of inflammation in vivo after treatment with triflusal at therapeutic doses. The rat s.c. air pouch model was used because it is characterized by the release of prostaglandins due to a rapid induction of COX-2 mRNA and protein in exudate cells (Masferrer et al., 1994). The administration of triflusal (3–30 mg/kg p.o.) blocked COX-2 expression in the cells present in the exudate and produced a dose-dependent decrease in PGE2synthesis. Aspirin at similar doses did not exert any detectable effect on COX-2 expression, although it inhibited PGE2to the same extent as triflusal. Thus, in vivo aspirin effect could be mainly attributed to enzyme inhibition, whereas COX-2 expression blockade better accounts for the inhibition of PGE2 release by triflusal. Moreover, triflusal showed 6-fold less inhibitory effect than aspirin on PGE2 production in the stomach, which is mainly due to COX-1. Thus, triflusal is 10 times more selective for COX-2 over COX-1 than aspirin (COX-1/COX-2 ID50 ratio of 1.2 for triflusal and 0.12 for aspirin) in this model, and provided that COX-1 has cytoprotective effects in the gastric mucosa (Masferrer et al., 1994), this may explain the better gastric tolerance and lower incidence of bleeding events with triflusal compared with aspirin in the clinical use of both drugs (McNeely and Goa, 1998).
In conclusion, the data presented here show that triflusal and its main metabolite HTB are able to inhibit NF-κB activation and COX-2 expression in both human mononuclear cells in vitro and rat inflammatory cells in vivo. The concentrations required to elicit these effects are in keeping with the plasma concentrations achieved in therapeutic use (McNeely and Goa, 1998). Neither aspirin nor sodium salicylate shows significant effects on COX-2 expression in any of the experimental models tested. Thus, the introduction of the trifluoromethyl group in the 4-position of salicylates confers new properties to the molecule. To our knowledge, this is the first demonstration of an inhibition of COX-2 expression by salicylates at therapeutic concentration ranges. Moreover, because HTB does not affect COX-1 activity (De la Cruz et al., 1992), it becomes the first NSAID that selectively blocks the expression of the inducible isoform of COX, which makes it a possible and promising tool for knocking out this enzyme. These findings have several implications. First, triflusal and HTB may become useful tools in the investigation of the mechanism of action of salicylates in the diverse processes in which this family of drugs has been implicated. Second, because COX-2 expression and/or NF-κB activations play a role in a broad variety of pathological conditions, including inflammation, colorectal cancer, Alzheimer’s disease, and stroke (Pennisi, 1998; Rustgi, 1998), the ability of triflusal and HTB to inhibit NF-κB activation/COX-2 expression at therapeutic concentrations may be useful in these diseases.
Acknowledgments
We thank Dr. Mariano Sanchez-Crespo from CSIC (Faculty de Medicina, Valladolid) for his helpful discussion and review of the manuscript and Assumpta Oliveras for her extremely appreciated technical assistance.
Footnotes
-
Send reprint requests to: Dr. Javier Forn, Department of Pharmacology, Uriach Research Center, Degá Bahı́, 59-67, 08026 Barcelona, Spain. E-mail rd{at}uriach.com
- Abbreviations:
- COX-1
- cyclooxygenase-1
- COX-2
- cyclooxygenase-2
- ELISA
- enzyme-linked immunosorbent assay
- HTB
- 2-hydroxy-4-trifluoromethylbenzoic acid
- MTT
- 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium
- HUVEC
- human umbilical vein endothelial cell
- LPS
- bacterial lipopolysaccharide
- NF-κB
- nuclear factor-κB
- NSAID
- nonsteroidal anti-inflammatory drug
- PBMC
- peripheral blood mononuclear cell
- PG
- prostaglandin
- Received November 2, 1998.
- Accepted January 19, 1999.
- The American Society for Pharmacology and Experimental Therapeutics