Quoc-Tuan Do1 Ccile Lamy2 Isabelle Renimel2 Nancy Sauvan2 Patrice Andr2 Franck Himbert1 Luc Morin-Allory3 Philippe

Bernard1

Reverse Pharmacognosy: Identifying Biological Properties for Plants by Means of their Molecule Constituents: Application to Meranzin
Original Paper

Abstract Reverse pharmacognosy aims at finding biological targets for natural compounds by virtual or real screening and identifying natural resources that contain the active molecules. We report herein a study focused on the identification of biological properties of meranzin, a major component isolated from Limnocitrus littoralis (Miq.) Swingle. Selnergy, an in silico biological profiling software, was used to identify putative binding targets of meranzin. Among the 400 screened proteins, 3 targets were selected: COX1, COX2 and PPARg. Binding tests were realised for these 3 protein candidates, as well as two negative controls. The predictions made by Selnergy were consistent with the experimental results, meaning that these 3 targets can be modulated

Key words Inverse docking Limnocitrus littoralis Rutaceae cyclooxygenase meranzin Selnergy reverse pharmacognosy Supporting information available online at http://www.thieme-connect.de/ejournals/toc/plantamedica

1235

Introduction Pharmacognosy is the study of the pharmacochemistry of natural raw materials, mainly, but not exclusively extracted from plants [1], for pharmaceutical, dietary and cosmetic purposes. It leads to bioactive molecules after extraction, purification, characterisation and bioassays. After more than a century of research in pharmacognosy, a huge amount of data has been acquired on plants and on their molecules. Nevertheless, in most of the cases, little is known about the biological properties of these natural products. In order to exploit these data and to find biological properties for these natural

products, we have introduced and explored the usefulness of the concept of reverse pharmacognosy (RPn) [2]. RPn is similar to reverse pharmacology [3] as small molecules are used as probes to evaluate their effects on a biological system, but differs from reverse pharmacology by its final goal. RPn aims at finding applications for natural molecules and their sources (mostly plants), but does not focus only on identifying new targets or new biological routes. RPn allies chemoinformatics tools and traditional knowledge in the search for plants aimed at the development of botanicals, pharmaceuticals or cosmetics. In a first step, the biological properties of a molecule are screened in silico and/or in vitro, then, in a second step, it is pos-

Affiliation GREENPHARMA S. A., Orlans, France 2 GIE LVMH RECHERCHE, 45804 Saint Jean de Braye cedex, France 3 Institut de Chimie Organique et Analytique, Orlans, France
1

Correspondence Dr Philippe Bernard GREENPHARMA S. A. 3 alle du Titane 45100 Orlans France Phone: +33-2-3825-9980 Fax: +33-2-3825-9965 E-mail: philippe.bernard@greenpharma.com Received June 22, 2007 Revised July 16, 2007 Accepted July 19, 2007 Bibliography Planta Med 2007; 73: 12351240  Georg Thieme Verlag KG Stuttgart New York DOI 10.1055/s-2007-990216 Published online September 13, 2007 ISSN 0032-0943

Downloaded by: Jadhavpur University. Copyrighted material.

by an extract containing this compound in a suitable concentration. These results demonstrate that reverse pharmacognosy and its inverse docking component is a powerful tool to identify biological properties for natural molecules and hence for plants containing these compounds.

sible to retrieve the plants containing the active compounds thanks to a plant/molecule relational database. Thus, reverse pharmacognosy is complementary to pharmacognosy. Two key components are essential for implementing reverse pharmacognosy: (i) a virtual screening tool such as Selnergy or a real screening platform; (ii) a database linking plants and molecules. Selnergy is an inverse docking tool and consists in the association of a docking sofware and a 3-dimensional (3D) protein target database. Instead of searching for compounds that can bind to a specific protein, Selnergy will identify putative targets that can interact with a molecule. Virtual screening (or docking compound database) is now a well-established method that appeared more than ten years ago [4]. It has also been applied successfully to natural compounds [5], [6], [7]. The potential of inverse docking was evaluated in 2001 by Chen et al. [8]. These authors used Dock [4] as the docking tool. This software unfortunately suffers from a lack of speed and accuracy so is not suitable for a systematic screening of a huge target database. Similar works have been published that used other docking tools [9]. Selnergy [10] has been developed with FlexX that is a balance between accuracy and speed. Moreover, we brought particular attention to the quality of the protein models we have included into our target database by implementing a selection procedure (see Materials and Methods). A visual analysis is finally applied to the best hits from Selnergy to discard false positives. Herein we report the study of meranzin (Fig. 1), a coumarin derivative characterised by an epoxide group. The goal is to find an application for this product and consequently an application for its source. Meranzin was isolated from Limnocitrus littoralis (Miq.) Swingle, Rutaceae. Little is known about the biological properties of meranzin. As this molecule can be purified in huge quantities from a readily available sources, a rapid industrialisation can be expected with a possible sustainable development with local population.

the target library if it fulfills the following conditions. From a set of molecules consisting of active ligands known for their target affinity and molecules randomly selected, the virtual screening on this target will rank the actives among the best scored structures [2]. 2. Molecule drawing: Meranzin was sketched in Sybyl. The threedimensional structure was calculated with Concord 4.0.4 [14] and minimised with Tripos force field with 25 Simplex [15] steps coupled with Powell method [16]. The maximum number of iterations was set to 1000 with a termination gradient of 0.05 kcal mol1 1. 3. Docking of meranzin: Meranzin was considered flexible in the docking procedure. A rigid core with a maximum of interaction groups (e. g., donors and acceptors of hydrogen bonds) is first selected, then placed inside the active site of a protein. After that, the flexible parts of the molecule are iteratively grown with respect to steric and electrostatic constraints of the protein pocket [13]. Dissimilar molecule poses are scored and ranked. By default, the 30 best scored placements are retained. 4. Hit selection procedure: A spatial fitting is taken into account to determine whether meranzin is correctly docked into a protein active site. This is achieved by comparing the distance between the centroid defined in the active site of the studied protein and the centroid defined in meranzin (distance centroid to centroid: dC-C). A meranzin placement with a dC-C > 4 is considered out of the protein active site, and the protein/ligand pair is discarded from further analysis. As Selnergy can propose several poses for the same molecule, the placement with the shortest dC-C distance is selected. A macro written in SPL and distributed by Tripos was implemented to automatise this procedure. Hit identification is based on an estimated interaction energy of meranzin with a target divided by the number of meranzin atoms (E1). This energy is referred as normalised interaction energy. This normalised energy is then compared with a reference ligand normalised interaction energy (ligand co-crystallised with the protein) referred as E2. E1 should be less or equal to E2 for a solution to be accepted. We used the interaction energy normalised by atom number because huge molecules have the tendency to be better scored by docking and scoring tools, thus biasing the results. Plant material and molecule Leaves of Limnocitrus littoralis (Miq.) Swingle were collected in the province of Binh Thuan, Vietnam, in July 2003 and identified by Prof. Tran Hop of University of Ho Chi Minh City. A voucher specimen (NCY008913) is deposited at the Conservatory and the Botanical Gardens of Nancy, France. The extraction, purification and characterisation of meranzin are described in the Supporting Information section. COX-1 and COX-2 binding assays The COX inhibitor screening assay (Catalogue No. 560131) by Cayman Chemical (Ann Arbor, MI, USA) was used to identify COX inhibitors and to assess the selectivity between COX-1 and COX-2 of the inhibitors. The protocol [17] was modified as deDownloaded by: Jadhavpur University. Copyrighted material.

1236

Original Paper

Materials and Methods Virtual screening 1. Selnergy implementation: Selnergy is implemented in the Sybyl 6.9 package [11] in SPL language and consists of a target library from crystallography (structures retrieved from Protein Data Bank [12] http://www.rcsb.org/pdb/ or from homology modelling. Selnergy is based on FlexX [13]. Homology modelling was performed with Biopolymer and Composer modules within the Sybyl package. A protein may be represented with several 3D structures (if available) in order to reflect protein flexibilities. So far, we have more than 400 proteins and 300 olfactory receptor structures. A target is considered as validated and added to
Fig. 1 The structure of meranzin.

Do Q-T et al. Reverse Pharmacognosy: Identifying Planta Med 2007; 73: 1235 1240

scribed in the kit booklet accompanying the product. All measurements were done in duplicates. Elastase binding assays In our study, pancreatic elastase (Sigma Aldrich; Saint-QuentinFallavier, France) was used. The assay was based on the hydrolysis of N-succinyl-L-ala-L-ala-L-ala-p-nitroanilide, a synthetic substrate developed by Bieth et al. [18]. Hydrolysis of N-succinyl-Lala-L-ala-L-ala-p-nitroanilide by elastase at 25 8C and pH 8.0 could be followed by reading the absorbance at 410 nm. PPARg cell-based assays HEK293 cells (ACACC; Salisbury, UK) were transiently transfected with the PPAR responsive human CPT-1b promoter-luciferase reporter plasmid [19] with or without expression vectors harbouring human PPARg. Cells were co-transfected with a b-galactosidase expression vector to correct for possible differences in transfection efficiency. Rosiglitazone (Axxora LLC; San Diego, CA, USA) was used as a reference compound. Luciferase activity was normalised to b-galactosidase activity [19]. Prostaglandins production by keratinocytes Keratinocytes were grown in supplemented keratinocyte serumfree medium (KSFMc) (Gibco ref : 17 005-034 + 37 000-015; Grand Island, NY, USA). The keratinocytes were seeded in 96well microplates (10,000 cells per well). The first day of culture was considered as D0. After 24 hours (D1) of incubation, the medium was replaced by KSFM-c containing meranzin. 24 hours later, PGE2 released was measured in cultured human keratinocytes medium using ELISA tests (R&D Systems, Ref DE0100; Lille, France). The positive control, indomethacin (Sigma Aldrich) and meranzin (LVMH Recherche; Saint-Jean-de-Braye, France) were tested at 2mM. Cell viability A cell viability test was conducted with the 2,3-bis(2-methoxy4-nitro-5-sulfophenyl)-5-[(phenylamino)carbonyl]-2H-tetrazolium hydroxide (XTT) assay (after 24 hours of treatment) to verify that the treatments at the concentrations studied were not toxic for the keratinocytes. The Cell Proliferation Kit II (XTT test) was purchased from Roche Diagnosis (Mannheim, Germany). Supporting information The isolation procedure for meranzin and the physico-chemical data of this molecule are available as Supporting Information.

Table 1

In silico prediction of affinity by Selnergy. PDB id column contains Protein Data Bank codes that identifies a protein 3D structure
PDB id 1eqh 1fm9 1rbp 1oyn 1soj 4cox 1qkt 1ela 1cx2 3pgh 1hgi 1hgh 1e8z 1acj 1hvr 1m7q 4phv 4hmg 1rkp 1aq1 1agw Targets Cyclooxygenase 1 Peroxisome proliferators activated Rreceptor g Retinol binding protein Phosphodiesterase 4D Phosphodiesterase 3B Cyclooxygenase 2 Estrogen receptor b Elastase Cyclooxygenase 2 Cyclooxygenase 2 Hemagglutinin Hemagglutinin Phosphoinositide 3-kinase gamma Acetylcholinesterase HIV-1 protease p38 Map kinase HIV-1 protease Hemagglutinin Phosphodiesterase 5A Cyclin-dependent protein kinase 2 Fibroblast growth factor receptor 1 E1-E2 1.793 1.178 1.017 0.816 0.783 0.756 0.661 0.532 0.490 0.466 0.421 0.411 0.360 0.324 0.310 0.177 0.152 0.080 0.068 0.066

Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Original Paper

this ranking list with respect to enzyme families. PDE4D and PDE3B have been well scored and topped at position 4 and 3 (noteworthy, PDE5A is on the extended list at position 19). The other putative targets are oestrogen receptor b and elastase. Because the number of interaction sites can be high in a protein active site, docking software can sometimes place a ligand in a wrong pose. In order to discard false positives, we have to examine visually the docking results. We will detail below our analysis on the 10 best pairs in order to make a final ranking according to knowledge on proteins and their preferred binding modes. The pro and con arguments are listed. The final selections will be submitted to experimental validations. The active site of COX1 is a deep close pocket surrounded mostly by hydrophobic residues [20]. Meranzin is a hydrophobic molecule, it is docked into the active site and forms hydrophobic interactions with residues of COX1. The coumarin fragment is located in the cleft formed by MET113, VAL116, ALA527, LEU531, LEU534, VAL349, TYR355 and LEU359 (the residues are numbered according to [20]), while the epoxide part is at the vicinity of PHE381, TYR385, TRP387, TYR348, PHE518, LEU532, GLY526 and SER530. The oxygen of the meranzin epoxide may form a hydrogen bond with the hydroxy group of SER530 (Fig. 2). Meranzin is docked in a similar fashion in the COX2 active site, except that no hydrogen bond is detected with SER530 (Fig. 2). By overlaying COX1 and COX2 active sites and docked meranzin in the respective pocket as shown in Fig. 2, we find a low RMS value of 1.7, which denotes that the poses of meranzin in both sites are quite similar. The only change in the active sites of the two sub-

1237

Results and Discussions The pairs of molecule/protein with a negative E1-E2 are selected. Table 1 lists all the protein structures that fulfill this criterion. We will focus on the first 10 as these proteins are most likely to interact with meranzin. COX1 is the best scored protein. Interestingly, there are three structures of COX2 that are ranked within the top 10, in position 6, 9 and 10. This family of enzymes seems to have a high affinity to meranzin. The second protein selected is the nuclear receptor PPARg, and the third one the retinal binding protein, a transport protein. Phosphodiesterases (PDE) come in the second place of

Do Q-T et al. Reverse Pharmacognosy: Identifying Planta Med 2007; 73: 1235 1240

Downloaded by: Jadhavpur University. Copyrighted material.

0.222

types of COX is that ILE523 in COX1 is replaced by VAL523 in COX2. This substitution is enough to confer selectivity [21]. Thus, we cannot discriminate which docking is better in the light of the placement and the selectivity data. Both subtypes will be kept in our selection. Meranzin seems to be well docked into PPARg. The polar group of the coumarin is positioned close to four residues that form hydrogen and/or ionic interactions with the carboxylic group of some ligands [22]. The carboxyl group of meranzin may form hydrogen bonds with SER289, HIS323, HIS449 and TYR473 (numbering according to [22]). At the vicinity of these polar residues, a hydrophobic cleft can be found in which the hydrophobic moiety of meranzin can fit in. An argument in disfavour of meranzin is the large size of PPARg active site. By considering the co-crystallised ligand GI262570 and meranzin best docked pose, the overlapping volume represents 66 % of meranzin's volume and 31 % of that of GI262570. The total volume of meranzin corresponds to 47 % of that of GI262570 (Fig. 3). PPARg should be a good candidate although meranzin just occupied a small fraction of the binding pocket. The third putative protein retained by Selnergy is the retinol binding protein. The ligands of this transporter are retinoids [23] that consist of a large apolar part and a polar head. So the binding cleft is composed of hydrophobic residues. Retinol in the retinol binding protein is docked with its hydroxy group at the entrance of the cleft whereas the apolar part is buried into the site. Meranzin adopts a different binding pattern: the molecule's polar groups are deeply placed into the pocket. These data considered, retinol binding protein appears a less likely binding partner for our compound. Phosphodiesterases (PDE) were not retained after reviewing the docking results. The PDE site possesses a conserved glutamine in orientations which purportedly discriminate between cAMP and

cGMP. This residue intervenes in the binding of the cyclic nucleotides and inhibitors [24]. In our case, meranzin did not interact with this crucial residue either in PDE4D or in PDE3B. For that reason, the affinity of meranzin to both PDE may be questionable. In the case of estrogen receptor b, the molecule is situated near but out of the active site. So we have discarded this target from our final selection. In the case of elastase, the co-crystallised ligand shares hydrogen bonds with VAL224, SER222 and possibly SER203. None of these interactions are observed for our compound, or other polar interactions [25]. It seems that the predicted binding of our molecule relies only on the apolar interaction. In the light of these data, we rejected elastase as a potential target. The rest of the targets ranked from 11 to 21 were also reviewed and none of them satisfied our binding mode analysis. We now present results from in vitro evaluations for the best ranked protein partners of meranzin, namely COX1, COX2 and PPARg. As negative controls, we tested meranzin with elastase and 5-lipoxygenase (LOX). Some chemical series can bind to COX and LOX as dual inhibitors [26]. So this enzyme was also added to the list of proteins to be evaluated although it was not selected by Selnergy. Table 2 shows the results of inhibition assays for COX1 and COX2. The experimental data nicely fit with our prediction for COX2. Meranzin inhibits COX2 in a dose-dependent manner with an inhibition of 56.2 % at 0.4 mM. Therefore the IC50 of meranzin on COX2 is in a submicromolar range, which is an outstanding activity for a natural compound. Meranzin is less active on COX1 than on COX2. Moreover, the inhibition on COX1 is not dose-dependent.

1238

Do Q-T et al. Reverse Pharmacognosy: Identifying Planta Med 2007; 73: 1235 1240

Downloaded by: Jadhavpur University. Copyrighted material.

Original Paper

Fig. 2 Superpossition of COX1 (in green) and COX2 (in white). The placements of merazin are colour-coded as follows: merazin in COX1 in green and in COX2 in white. The hashed line represents a hydrogen bond. Residues of interest such as VAL523 and ILE523 are highlighted in capped sticks. These amino acids are responsible for the selectivity between the subtypes. Colour codes for atoms: carbon in white or green, nitrogen in blue, oyxygen in red, sulphur in yellow and hydrogen in cyan.

Fig. 3 Merazin docked in the PPARg active site. It is represented in a green ball-and-stick fashion. The co-crystallised ligand is represented in magenta capped sticks. The volumes occupied by both molecules are highlighted by colour-coded volumes. Important residues of the protein are displayed in capped sticks. Hashed lines symbolise hydrogen bonding. Colour codes for atoms: carbon in white or green, nitrogen in blue, oyxygen in red, sulphur in yellow and hydrogen in cyan.

Table 2

Binding assays of meranzin on cyclooxygenase 1, cyclooxygenase 2 and 5-lipoxygenase. The inhibition values are the mean of 2 measures
COX1 COX2 % inhibition Meranzin (mM) 1 10 100 LOX % inhibition

intended to rank affinities of different molecules on the same target. To conclude, we have validated in vitro that meranzin can bind to COX1, COX2 and PPARg as predicted by Selnergy. Moreover, it inhibits prostaglandin release in human keratinocytes. Our results demonstrate that inverse docking can be useful in finding new applications for compounds and hence natural sources, e. g., plants that contain such compounds. Selnergy can accelerate this process of known target identification through its indexed target library. In the case of meranzin, the prediction was in good agreement with experimental data in respect to the interaction with COX1, PPARg and particularly with COX2. Our in silico tool is obviously beneficial for decision making in the process of finding new applications for molecules and their related sources. Nevertheless, it cannot predict selectivity for subtypes with very close active sites such as COX that differs by only one apolar residue. Moreover one has to bear in mind that simple models are used here to simulate complex biological processes so that experimental validations will still be necessary. With the fast evolution of analytical chemistry, information on the composition of natural material is becoming more and more accessible in quantity and in quality. Our approach can help to find new applications for plants based on Western medicinal concepts of molecules and protein targets, and, more focused on symptoms. It complements traditional cures which are more focused on the origin of disease. Bearing in mind that toxicity also constitutes a crucial issue, we hope, however, that our approach will widen access to treatments for populations that cannot afford Western medicine drugs.
1239

Meranzin (mM) 0.4 4 40

% inhibition

37.1 19.1 64.4

56.2 71.1 79.5

8.3 0.9 4.9

Original Paper

In the case of COX, Selnergy was able to detect activity at micromolar ranges for meranzin but failed to correctly rank its affinities for COX subtypes. A possible explanation is the structural similarity of the two active sites which differ by a single residue. We also verified that there is no significant activity of meranzin on LOX at 100 mM. We also measured the inhibition of prostaglandin release by human keratinocytes. The inhibition rate by meranzin at 2 mM is 25  0.9 % (n = 4) compared to the reference molecule indomethacin at 2 mM, for which the inhibition rate is 17  0.3 % (n = 6). Therefore, our molecule is slightly more active. Thus meranzin appears a potent inhibitor of the inflammatory process. The meranzin/PPARg interaction was validated in a cell-based assay with transfected PPAR. The reference compound was rosiglitazone. Meranzin activates PPARg in a dose-dependent manner (see Table 3). It is active above 10 mM. At 100 mM, its activity is comparable to that of rosiglitazone at 10 mM. We verified that meranzin is not active on pancreatic elastase at a concentration of 1 mM (I% = 2  0.01 %), compared to ellagic acid (our positive control) with an inhibition rate is 98  15 % at 1 mM on the same test. These results demonstrate the ability of Selnergy to select the most probable protein partners for a molecule within its top 2 % of the best scored proteins. This reduced number of hits allows for human supervisions to retain the most promising candidate and to discard false positive proteins. Ranking accurately the affinity proteins and ligands is a difficult challenge. This is due to the imprecision of current scoring functions which are

Acknowledgements Our warm thanks to Prof. T. Hop and to the Spectropole of Marseille for NMR analysis. Special thanks to Dr. M. van Bilsen and Dr. A. Gilde from University of Maastricht.

References
1

Table 3

PPARg cell-based assay. Measures are in triplicates. Activities are expressed relative to vehicle control (DMSO) arbitrarily set to 1.0
Activating rate 1.0  0.1 1.1  0.0 1.4  0.2 1.0  0.0 1.6  0.0

Meranzin (mM) 1 10 100 DMSO Rosiglitazone 10 mM

Bruneton J. Pharmacognosie, Phytochimie, Plantes Mdicinales. Paris: Lavoisier; 1993: viii. 2 Do QT, Bernard P. Pharmacognosy and reverse pharmacognosy : a new concept for accelerating natural drug discovery. IDrugs 2004; 7: 1017 27. 3 Harrigan GG, Brackett DJ, Boros LG. Medicinal chemistry, metabolic profiling and drug target discovery: a role for metabolic profiling in reverse pharmacology and chemical genetics. Mini Rev Med Chem 2005; 5: 13 20. 4 Shoichet BK, Stroud RM, Santi DV, Kuntz ID, Perry KM. Structure-based discovery of inhibitors of thymidylate synthase. Science. 1993; 259: 1445 50. 5 Rollinger JM, Haupt S, Stuppner H, Langer T. Combining ethnopharmacology and virtual screening for lead structure discovery: COX-inhibitors as application example. J Chem Inf Comput Sci 2004; 44: 480 8. 6 Liu H, Li Y, Song M, Tan X, Cheng F, Zheng S et al. Structure-based discovery of potassium channel blockers from natural products: virtual screening and electrophysiological assay testing. Chem Biol 2003; 10: 1103 13. 7 Bernard P, Scior T, Didier B, Hibert M, Berthon JY. Ethnopharmacology and bioinformatic combination for leads discovery: application to phospholipase A(2) inhibitors. Phytochemistry 2001; 58: 865 74.

Do Q-T et al. Reverse Pharmacognosy: Identifying Planta Med 2007; 73: 1235 1240

Downloaded by: Jadhavpur University. Copyrighted material.

1240

Do Q-T et al. Reverse Pharmacognosy: Identifying Planta Med 2007; 73: 1235 1240

Downloaded by: Jadhavpur University. Copyrighted material.

Chen YZ, Zhi DG. Ligand-protein inverse docking and its potential use in the computer search of protein targets of a small molecule. Proteins 2001; 43: 217 26. 9 Paul N, Kellenberger E, Bret G, Muller P, Rognan D. Recovering the true targets of specific ligands by virtual screening of the protein data bank. Proteins 2004; 54: 671 80. 10 Do QT, Renimel I, Andre P, Lugnier C, Muller CD, Bernard P. Reverse pharmacognosy: application of selnergy, a new tool for lead discovery. The example of epsilon-viniferin. Curr Drug Discov Technol 2005; 2: 161 7. 11 Tripos Inc; 1699 South Hanley Rd.; St. Louis, Missouri, 63144, USA.: 12 Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H et al. The protein data bank. Nucleic Acids Res 2000; 28: 235 42. 13 Rarey M, Kramer B, Lengauer T, Klebe G. A fast flexible docking method using an incremental construction algorithm. J Mol Biol 1996; 261: 470 89. 14 Pearlman RS. CONCORD 4.0.7A, Concord User's Manual,. distributed by Tripos Inc; St. Louis, MO.: 15 Press WH, Flannery BP, Teukolsky SA, Vetterling WT. Simplex. Numerical recipes in C, the art of scientific computing. Cambridge: University Press; 1988: 312 27. 16 Powell MJD. An efficient method for finding the minimum of a function of several variables without calculating derivatives. Comp J 1964; 7: 155 62. 17 Maclouf J, Grassi J, Pradelles P. Development of enzyme-immunoassay techniques for the measurement of eicosanoids. In: Walden TL, Hughes HN; editors. Prostaglandin and lipid metabolism in radiation injury New York: Plenum Press; 1987: 355 64. 18 Bieth J, Spiess B, Wermuth CG. The synthesis and analytical use of a highly sensitive and convenient substrate of elastase. Biochem Med 1974; 11: 350 7.

19

Gilde AJ, van der Lee KA, Willemsen PH, Chinetti G, van der Leij FR, van der Vusse GJ et al. Peroxisome proliferator-activated receptor (PPAR) alpha and PPARbeta/delta, but not PPARgamma, modulate the expression of genes involved in cardiac lipid metabolism. Circ Res 2003; 92: 518 24. 20 Garavito RM, Picot D, Loll PJ. Preliminary X-ray investigations into NSAID-binding to cyclooxygenase-1. Am J Ther 1995; 2: 611 5. 21 Gierse JK, McDonald JJ, Hauser SD, Rangwala SH, Koboldt CM, Seibert K. A single amino acid difference between cyclooxygenase-1 (COX-1) and -2 (COX-2) reverses the selectivity of COX-2 specific inhibitors. J Biol Chem 1996; 271: 15 810 4. 22 Xu HE, Lambert MH, Montana VG, Plunket KD, Moore LB, Collins JL et al. Structural determinants of ligand binding selectivity between the peroxisome proliferator-activated receptors. Proc Natl Acad Sci U S A 2001; 98: 13 919 24. 23 Zanotti G, Berni R. Plasma retinol-binding protein: structure and interactions with retinol, retinoids, and transthyretin. Vitam Horm 2004; 69: 271 95. 24 Zoraghi R, Corbin JD, Francis SH. Phosphodiesterase-5 Gln817 is critical for cGMP, vardenafil, or sildenafil affinity: its orientation impacts cGMP but not cAMP affinity. J Biol Chem 2006; 281: 5553 8. 25 Mattos C, Rasmussen B, Ding X, Petsko GA, Ringe D. Analogous inhibitors of elastase do not always bind analogously. Nat Struct Biol 1994; 1: 55 8. 26 Fiorucci S, Meli R, Bucci M, Cirino G. Dual inhibitors of cyclooxygenase and 5-lipoxygenase. A new avenue in anti-inflammatory therapy? Biochem Pharmacol 2001; 62: 1433 8. 27 Hito C, Furukawa H. Constituents of Murraya exotica L. Structure elucidation of new coumarins. Chem Pharm Bull 1987; 35: 4277 85.

Original Paper