Design of Group IIA Secreted/Synovial Phospholipase A2 Inhibitors: An Oxadiazolone Derivative Suppresses Chondrocyte Prostaglandin E2 Secretion

Group IIA secreted/synovial phospholipase A2 (GIIAPLA2) is an enzyme involved in the synthesis of eicosanoids such as prostaglandin E2 (PGE2), the main eicosanoid contributing to pain and inflammation in rheumatic diseases. We designed, by molecular modeling, 7 novel analogs of 3-{4-[5(indol-1-yl)pentoxy]benzyl}-4H-1,2,4-oxadiazol-5-one, denoted C1, an inhibitor of the GIIAPLA2 enzyme. We report the results of molecular dynamics studies of the complexes between these derivatives and GIIAPLA2, along with their chemical synthesis and results from PLA2 inhibition tests. Modeling predicted some derivatives to display greater GIIAPLA2 affinities than did C1, and such predictions were confirmed by in vitro PLA2 enzymatic tests. Compound C8, endowed with the most favorable energy balance, was shown experimentally to be the strongest GIIAPLA2 inhibitor. Moreover, it displayed an anti-inflammatory activity on rabbit articular chondrocytes, as shown by its capacity to inhibit IL-1β-stimulated PGE2 secretion in these cells. Interestingly, it did not modify the COX-1 to COX-2 ratio. C8 is therefore a potential candidate for anti-inflammatory therapy in joints.


Introduction
Inflammation is a multi-faceted process involving numerous enzymes, such as phospholipases A 2 (PLA 2 s) and cyclo-oxygenases (COXs) [1]. PLA 2 s catalyze the hydrolysis of cell-membrane glycerophospholipids at the sn-2 position leading to the generation of free fatty acids such as arachidonic acid. The later is subsequently metabolized into potent pro-inflammatory mediators such as eicosanoids (e.g. prostaglandin E 2 [PGE 2 ]) through a pathway involving COX-1 and COX-2 in part [2]. PGE 2 is the main eicosanoid contributing to pain and inflammation in rheumatic diseases [3,4]. Nonsteroidal anti-inflammatory drugs (NSAIDs) reduce the production of PGE 2 , which leads to a significant improvement in rheumatic symptoms. However, these drugs exhibit gastrointestinal toxicity mainly because of a marked decrease in COX-1 activity [4] and renal and blood pressure toxicities mainly because of a decrease in COX-2 activity. COX-1 is constitutively expressed in most tissues and appears to be responsible for maintaining normal physiological function. How-ever, COX-2 is absent in most tissues under normal resting conditions but is induced in inflamed tissues and is responsible for increased PGE 2 production. This activation has motivated the development of selective COX-2 inhibitors. However, these inhibitors also have severe side effects such as myocardial infarction [5,6]. Overcoming this problem could involve the development of novel anti-inflammatory agents to efficiently inhibit the PLA 2 -dependent production of COX substrates without impairing the balance between COX-1 and COX-2.
PLA 2 s represent a growing family of enzymes of two main categories, intracellular and secreted. Among the 10 human secreted PLA 2 s (sPLA 2 s) known to date, the most studied is the non-pancreatic Group IIA, denoted GIIAPLA 2 , because of its involvement in the pathogenesis of many inflammatory diseases (for a review, see [7]). GIIAPLA 2 was originally purified from the synovial fluid of patients with rheumatoid arthritis [8,9,10]. The number of rheumatoid arthritis-affected joints and the presence of destructive erosion correlate with the amount of GIIAPLA 2 in the serum of patients [11]. Moreover, GIIAPLA 2 induces an inflammatory response when injected in rabbit joints [12] and exacerbates rat adjuvant arthitis after intradermal injection [13].
The systemic implication of sPLA 2 s in inflammation has prompted a number of research groups to develop selective inhibitors of different types of these enzymes. Some potent candidates have been evaluated in phase II clinical trials. Surprisingly, no effect was observed when such inhibitors were used to treat patients with sepsis or rheumatoid arthritis [14,15]. This failure could be due to the complexity of the inflammation process and the existence of compensatory pathways. However, these molecules have been tested only in high-level systemic inflammatory diseases, not in low-level inflammatory diseases such as atherosclerosis, diabetes, Alzheimer's, and osteoarthritis. Varespladib, a sPLA 2 inhibitor, was recently found to reduce atherosclerosis in apolipoprotein-E-null mice [16]. Thus, the efficacy of sPLA 2 inhibitors in these low-level inflammatory diseases should be re-examined.
We have developed various selective inhibitors of sPLA 2 s [17,18,19,20,21]. Previously, we reported on the computerassisted design and synthesis of a series of novel oxadiazolone derivatives that were shown to exhibit potent inhibitory properties against GIIAPLA 2 [20]. In this series, a Ca(II)-binding oxadiazolone ring was connected through a polymethylene chain of varying lengths to an indole ring, which has been shown to be involved in apolar and cation-p interactions with GIIAPLA 2 residues. The optimal length of the linker was found to encompass 5 methylenes, and the corresponding compound, (3-{4-[5-(indol-1-yl)pentoxy]-benzyl}-4H-1,2,4-oxadiazol-5-one), is denoted C1 in the present study. In the current work, the indole moiety was replaced by other aromatic groups, which gave rise to compounds C2 to C8. Using molecular modeling, we computed and ranked energy balances for the binding of these inhibitors to GIIAPLA 2 . The inhibitory potencies of C2 to C8 against GIIAPLA 2 was analyzed by enzymatic assay, and the anti-inflammatory activity of the most potent compound, C8, was evaluated in IL-1b-treated articular chondrocytes.

Molecular modeling
We previously reported that one of the essential interactions between C1 and the target GIIAPLA 2 is Ca(II) bidentate chelation by the oxadiazolone moiety in its anionic form [20]. Because C2-C8 are structurally similar to C1 (Fig. 1), docking was performed upon first anchoring the oxadiazolone ring in the same position as compound C1, followed by energy minimization and molecular dynamics. As was observed for compound C1, the lowest-energy frames of C2-C8/enzyme complexes are stabilized by p-p and cation-p interactions involving His6, Phe23, and Phe63 on the one hand and Arg7 and Arg33 of GIIAPLA 2 on the other. Table 1 lists the energy values corresponding to the lowest-energy frames from molecular dynamics.
In vitro inhibition of enzymatic activity of sPLA 2 s by C1-C8 The compounds C1-C8 were submitted to fluorimetric assay to determine their inhibitory potencies and selectivity towards human GIIAPLA 2 (hGIIAPLA 2 ) versus porcine group IB PLA 2 (pGIB-PLA 2 ) ( Table 2). GIBPLA 2 is an enzyme of the same family as GIIAPLA 2 (sPLA 2 ) but is mainly involved in digestion of dietary phospholipids and is secreted by the pancreas [23]. Lipophilicity parameters, log P, of these products are calculated by use of Rekker's fragmental data [24] ( Table 2). The molecules C1-C8 are specific inhibitors of hGIIAPLA 2 because none inhibited pGIBPLA 2 at the highest concentration tested (100 mM). Such selectivity implies that C1-C8 should not interfere with the digestion process.
The experimentally measured IC 50 s for hGIIAPLA 2 ( Table 2) are associated with the final energy balances, denoted dE 2 in Table 1. The ranking of C1-C8 in terms of IC 50 is the same as that of the dE 2 magnitudes. In C2, the second phenyl ring is substituted with the ether O in the ortho position and in C6 in the para position. Both IC 50 and dE 2 values show C2 to have a significantly enhanced affinity for PLA 2 as compared with C6, even though both are iso-lipophilic (Tables 1 and 2). In C2, the biphenyl group has favorable van der Waals interactions with both Phe23 and Val30 of the enzyme, but in C6, the interactions are limited to Phe23. Such interactions could be further optimized, as when the biphenyl ring was replaced by phenantrene in C8. The lowest-energy complex is now stabilized by an enhanced overlap of this ring with Phe63 (Fig. 2). However, the lipophilicity increases in parallel, which could possibly limit the bioavailability of C8. We found C8 indeed endowed with the most favorable dE 2 value (Table 1), which was experimentally associated with the lowest IC 50 value (0.62 mM vs. 5 mM for C1).
At the other extreme, replacing the C1 indole ring by the smaller and less electron-rich phenyl ring, as in C7, resulted in a reduction of 10.3 kcal/mol in dE 2 value. Thus, C7 can be predicted to have the least inhibitory potency in the series. This finding was confirmed by experimentation showing C7 to have the highest IC 50 value (35 mM vs. 5 mM for C1).
Similar to C1, compounds C3-C5 have a bicyclic ring, whereas C3 possesses a benzo-1,3-thiazole instead of an indole ring. C4 and C5 have a chlorine and a methoxy substituent, respectively, in position 5 of the indole. In C3-C5, the aromatic rings interact simultaneously with His6, Arg7, and Val3, as was previously observed for C1 [20]. The difference in activity between C4 and C5 could be explained by additional electrostatic and/or van der Waals interactions contributed by methoxy substitution. C3 has anti-hGIIAPLA 2 activity close to that of C1, along with substantially reduced lipophilicity (2.88 vs. 3.81 for C1).
Thus, in the C1-C8 series, C8 has the most favorable dE 2 value and the lowest IC 50 on human GIIAPLA 2 activity, as evaluated by enzymatic assay. On the bases of the IC 50 values we focused our cellular assays on the most potent compound C8, the sole compound with a sub-micromolar activity. We thus chose to evaluate the cytotoxicity and anti-inflammatory activity of C8 in primary cultured rabbit articular chondrocytes treated with the pro-inflammatory cytokine IL-1b, which is known to play a key role in rheumatic diseases such as osteoarthritis (for reviews see [25,26]). Chondrocyte is the unique cell type in joints, and the cell model we chose is widely used to study the effect of inflammatory stress on joint cells.

Evaluation of the cytotoxicity of C8 on articular chondrocytes
We assessed the viability of the chondrocytes by MTT assay to evaluate the cytotoxic effects of C8 on these cells. Chondrocytes were treated for 20 h with 1 ng/mL IL-1b alone or 1 h after the addition of C8 at 0.31 to 9.92 mM, which corresponded to 0.5-to 5-fold the IC 50 of C8 on human GIIAPLA 2 activity (Table 2). Three different culture medium compositions were used: DMEM alone, or supplemented with 0.1% BSA or 2% FCS. IL-1b had no cytotoxic effects as compared with the untreated control condition for the three culture media tested (Fig. 3). In chondrocytes cultured in DMEM alone but with IL-1b, C8 had no cytotoxic effects at

Effect of C8 on IL-1b-stimulated PGE 2 secretion in articular chondrocytes
We tested the effect of C8 on the IL-1b-stimulated secretion of PGE 2 in chondrocytes. PGE 2 synthesis takes place mainly in response to cell activation by IL-1b, and its generation accounts for many of the actions induced by this cytokine [27]. In vitro, IL-1b induces the expression of COX-2 by chondrocytes, which results in increased PGE 2 production [28]. PGE 2 release thus represents a powerful IL-1b-and PLA 2 -dependent inflammatory marker in our cell model. Chondrocytes were treated for 20 h with IL-1b alone or 1 h after the addition of C8. As expected, IL-1b significantly stimulated PGE 2 secretion by chondrocytes in the three different culture media: 23.3-, 18.3-and 2.8-fold induction as compared with untreated control conditions, in DMEM alone or supplemented with 0.1% BSA or 2% FCS, respectively (Fig. 4). In chondrocytes treated with IL-1b, C8 had a strong and statistically significant inhibitory effect on PGE 2 secretion at all concentrations tested: from 0.31 to 1.24 mM in DMEM alone or from 0.31 to 4.96 mM in DMEM supplemented with 0.1% BSA or 2% FCS (Fig. 4). In DMEM alone, at  Table 3). The anti-IL-1b inhibitory effect of C8 at 0.31 mM increases when IL-1b concentration decreases. The inhibitory effect of C8 at 4.96 mM does not change when IL-1b concentration decreases. This is probably due to the fact that at 4.96 mM, the inhibitory effect of C8 on IL-1b-induced PGE 2 production is maximal. A parallel cellular test was performed on the compound C1 whose IC 50 is 5 mM (Table 2) and we observed that a 8 mM dose of C1, corresponding to 1.6-fold the IC 50 of C1 on human GIIAPLA 2 activity, does not decrease the stimulated PGE 2 secretion by IL-1b at 1 ng/mL (data not shown). Thus, C8, but not C1, decreases the IL- 1b-stimulated PGE 2 secretion in a dose-dependent manner in the three culture medium compositions used.

Effect of C8 on IL-1b-stimulated NO secretion in articular chondrocytes
We tested the effect of C8 on the IL-1b-stimulated secretion of NO in chondrocytes. NO is a mediator of immune and inflammatory responses. In vitro, IL-1b induces the expression of inducible NO synthase (iNOS) by chondrocytes, and consequently an increase in NO production [29]. NO secretion, evaluated by nitrite concentration in the cell culture medium, represents a reliable IL-1b-dependent and PLA 2 -independent inflammatory marker in our cell model. Chondrocytes were treated for 20 h with IL-1b alone or 1 h after the addition of C8. As expected, IL-1b  C8 did not significantly inhibit the IL-1b-stimulated nitrite secretion in chondrocytes cultured in DMEM medium alone or supplemented with 0.1% BSA (Fig. 5A, B). In DMEM supplemented with 2% FCS, C8 did not inhibit the IL-1bstimulated nitrite secretion at 0.31-, 0.62-, and 0.94-mM and slightly decreased by 17-, 19-, 21-% the IL-1b-induced nitrite production at 1.24-, 2.48-, and 4.96-mM, respectively (Fig. 5C). Thus, C8 did not inhibit the IL-1b-stimulated NO secretion in DMEM alone or supplemented with 0.1% BSA and slightly inhibited IL-1b-stimulated NO secretion in DMEM supplemented with 2% FCS.

Effect of C8 on COX-1, COX-2, and iNOS protein levels in articular chondrocytes
We evaluated the effect of C8 on COX-1, COX-2 and iNOS protein levels in chondrocytes treated with IL-1b. Chondrocytes were treated for 20 h with IL-1b alone or 1 h after the addition of C8 (0.31-1.24 mM) in DMEM, and protein extracts were examined by western blot analysis. As expected, COX-1 protein was detectable but COX-2 and iNOS proteins were undetectable in untreated control conditions (Fig. 6A). Moreover, IL-1b treatment induced the expression of COX-2 and iNOS proteins without affecting the level of COX-1 protein (Fig. 6A). In the presence of IL-1b, C8 did not alter the COX-1, COX-2 and iNOS protein levels (Fig. 6A). Consequently, the protein ratio of COX-1 to COX-2 was not modified by C8 (Fig. 6B).

Discussion
Nonsteroidal anti-inflammatory drugs (NSAIDs), which inhibit COX-1 and COX-2, and selective COX-2 inhibitors are currently used to reduce rheumatic symptoms. However, these drugs exhibit gastrointestinal, renal, blood pressure and cardiovascular toxicities. To overcome this problem, GIIAPLA 2 inhibitors could be developed to inhibit the production of COX substrates without impairing the balance between COX-1 and COX-2. We designed and synthesized 7 new oxadiazolone derivatives (C2 to C8) derived from C1. Using molecular modeling, we computed and ranked energy balances for the binding of these inhibitors to GIIAPLA 2 . The energy balances (Table 1) taking into account solvation effects show a correlation between dE 2 , the overall energy balance for binding, and the experimentally measured IC 50 for our novel compounds C1-C8. This finding should lend additional credence to our previous results [20], despite the approximations of the computational approach used in that study,  which allows for only single-point computations of Poisson-Boltzmann solvation energies for the most stable minima of the molecular dynamics procedure. Our useful predictions made with the present simplified energy potential may be due to the very local changes we made in the C1-C8 series. These bear on the series' sole terminal aromatic group and target a limited number of amino acids, so that the accuracy of the energy potential may be sufficient. We plan to study such energy balances with the polarizable molecular mechanics procedure SIBFA [30], which, along with the Langlet-Claverie methodology for Continuum solvation [31], was recently used to investigate the binding of inhibitors to metalloenzymes [32]. This study should also allow for considering changes on other parts of the drugs as well.
One possible unfavorable feature of C8 is its enhanced lipophilicity as compared with the other compounds. Nevertheless, this feature did not prevent the pharmacological efficiency of C8 in chondrocytes. Reduction in Log P could be anticipated by replacing phenantrene with heterocyclic analogs and/or substitution with hydrophilic groups. Such reductions were seen on passing from compound C1 with an indole ring to C3 with a benzothiazole. Nevertheless, the high lipophilicity of C8 should be an interesting option for its prospective clinical development, considering the possibility of local administration (intra-articular infiltration).
The toxicity and anti-inflammatory activity of C8 were evaluated in rabbit articular chondrocytes in primary culture. The toxicity of C8 was assessed by MTT, which allows an evaluation of the cell number and/or metabolic activity in cells. C8 (from 0.31 to 9.92 mM) did not decrease cell viability in culture medium supplemented with 0.1% BSA or 2% FCS but did (at 4.96 and 9.92 mM) in culture medium alone. This observation is probably due to the cells being weakened in the absence of BSA or FCS. We also observed, as expected, an increase in cell number and/or metabolic activity in response to IL-1b. This effect increases in the presence of C8, at non toxic doses, whatever the culture conditions. Thus, depending on the culture conditions or C8 doses, C8 increases or decreases cell number and/or metabolic activity. Moreover, C8 from 0.31 mM inhibited IL-1b-induced secretion of PGE 2 by chondrocytes, corresponding to half of the IC 50 on human GIIAPLA 2 activity evaluated in vitro by enzymatic assay. Therefore, C8 could be a potent anti-inflammatory drug in vivo. However, C8 did not inhibit IL-1b-induced NO secretion by chondrocytes cultured in DMEM alone or supplemented with 0.1% BSA and slightly inhibited IL-1b-stimulated NO secretion in DMEM supplemented with 2% FCS. These data suggest that the anti-inflammatory property of C8 in chondrocytes mainly depends on its capacity to inhibit PLA 2 activity.
COX-1 is involved in normal physiological functions, whereas COX-2 is involved in the inflammatory response. Antiinflammatory drugs such as NSAIDs and selective COX-2 inhibitors, used to treat rheumatic disease, have severe side effects owing to impairment in the balance between COX-1 and COX-2 [4,5,6]. Interestingly, the present work shows that the potent PLA 2 inhibitor C8 decreases PGE 2 production without impairing this balance. Consequently, C8 could be a useful candidate in developing new anti-inflammatory drugs lacking the side effects observed with NSAIDs and selective COX-2 inhibitors.
In summary, we report on the design, synthesis and testing of 7 C1 analogs that differ from C1 by indole substitution or by indole replacement by other aromatic rings, the largest being phenanthrene. Compounds C2-C8 show both inhibitory activity on secreted/synovial GIIAPLA 2 and selectivity as compared with GIBPLA 2 , a pancreatic enzyme involved in the digestion of dietary phospholipids. The order of interaction energies predicted by molecular modeling of these compounds is associated with their experimental IC 50 values with GIIAPLA 2 used as a target. The most promising compound is C8 in terms of computed energy balance for binding GIIAPLA 2 and experimental potency towards GIIAPLA 2 , namely one order of magnitude larger than that of C1. In addition, C8 is endowed with anti-inflammatory activity in articular chondrocytes by inhibiting IL-1b-stimulated PGE 2 secretion in these cells. Furthermore, it does not modify the ratio between the COX-1 and COX-2 isoenzymes. C8 is therefore an attractive candidate for anti-inflammatory therapy in joints. Experiments in animal models of rheumatic diseases are in progress in our laboratory.

Ethics Statements
Experimental protocols using rabbits complied with French legislation on animal experimentation and were approved by

Molecular modeling
Molecular modeling is described in Supporting Information S1.

Synthesis of oxadiazolone derivatives
Synthesis of compounds C1-C8 is described in Supporting Information S1.

In vitro PLA 2 assay
Fatty-acid free BSA and pancreatic PLA 2 were from Sigma. hGIIAPLA 2 was prepared as previously described [33]. The fluorescent substrate for PLA 2 assay, 1-hexadecanoyl-2-(10pyrenedecanoyl)-sn-glycero-3-phosphoglycerol, ammonium salt (b-py-C 10 -PG) was from Molecular Probes (Eugene). PLA 2 activity was evaluated as previously described [34] with b-py-C 10 -PG used as a substrate (2 mM final concentration). In a total volume of 1 mL, the standard reaction medium contained 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 1 mM EGTA, 2 mM b-py-C 10 -PG, 0.1% fatty-acid free BSA and 6 ng/mL pancreatic PLA 2 or 1 ng/mL hGIIAPLA 2 . The fluorescence (l ex = 342 nm and l em = 398 nm) of the enzymatic reaction medium was recorded for 3 min with use of a spectrofluorimeter LS 50 (Perkin-Elmer) equipped with a Xenon lamp. The reaction was initiated by the addition of CaCl 2 (10 mM, final concentration). The increase in fluorescence was continuously recorded for 1 min, and PLA 2 activity was calculated as previously described [34]. When used, the inhibitor was added to the reaction medium after introduction of BSA. The activity is expressed in micromoles of fluorescent b-py-C 10 -PG hydrolyzed per min. The standard error of the mean of three independent experiments was less than 10%, which allows for the determination of the IC 50 values (concentration of inhibitors producing 50% inhibition) of each compound.
Isolation and culture of chondrocytes from rabbit articular cartilage Articular chondrocytes were isolated from 5-week-old Fauve de Bourgogne female rabbits (CPA, Orleans, France) and cultured at the first passage in conditions avoiding cell dedifferentiation as previously described [35]. Cells were cultured at 37uC in 12-well plates in Ham's F-12 medium containing 10% FCS, 20 IU/mL penicillin, and 20 mg/mL streptomycin (all from Invitrogen) until nearly confluent. Then medium was replaced with DMEM (Invitrogen) containing 20 IU/mL penicillin, and 20 mg/mL streptomycin and, if necessary, 0.1% fatty acid free BSA (Sigma) or 2% FCS. At this time the C8 compound dissolved in DMSO (Sigma) was added to the medium (the amount of DMSO was kept at 1% (v/v) in all the wells). 1 h after the addition of C8, IL-1b (PeproTech) was added to the medium. Consequently, chondrocytes were incubated for 20 h with IL-1b and for 21 h with C8.

Evaluation of cell viability
At 18 h after the addition of IL-1b, 3-[4,5-dimethylthiazol-2yl]-2,5-diphenyl tetrazolium bromide (MTT; Sigma) was added to the cell culture medium at 0.5 mg/mL. Cells were incubated 2 more hours at 37uC. The medium was then removed, and DMSO was added to dissolve the formazan crystals. The absorbance of the resulting solution was spectrophotometrically measured at 570 and 690 nm (background). The value corresponding to absorbance 570nm -absorbance 690nm was directly proportional to the number and activity of the viable cells.
Determination of PGE 2 and nitrite concentrations in culture medium 20 h after the addition of IL-1b to the chondrocytes, culture media were collected, and aliquots were stored at 280uC until PGE 2 and nitrite quantification. PGE 2 concentration in culture media was determined by use of an enzyme immunoassay (EIA) kit (PGE 2 EIA Kit-monoclonal; Cayman Chemical). Nitrite concentration was determined by a spectrophotometry method based on the Griess reaction [36]. Briefly, 200 mL of culture medium or sodium nitrite (NaNO 2 , Merck) standard dilutions were mixed with 100 mL Griess reagent [0.5% (w/v) sulphanilic acid (Merck), 0.05% (w/v) N(1-naphtyl)ethylenediamine (Merck), 30% (v/v) acetic acid, 1.5 N HCl] and incubated for 10 min at 50uC. The absorbance was measured at 540 nm.
Preparation of whole-cell protein extracts, protein quantification and western blot analysis Immunocomplexes were detected by an enhanced chemiluminescence kit (Amersham Bioscience). The membrane was stripped by incubation in 0.2 M NaOH between successive immunodetections. Semi-quantitative scanning densitometry involved use of the ImageJ program (NIH, USA).

Statistical analysis
Results are expressed as means 6 SEM for the number of experiments indicated. Statistical analysis involved use of the Kruskal-Wallis test, then the ANOVA Fisher's test. A P,0.05 was considered statistically significant.

Supporting Information
Supporting Information S1 Materials and Methods in chemistry and molecular modeling.