In situ click chemistry generation of cyclooxygenase-2 inhibitors

Cyclooxygenase-2 isozyme is a promising anti-inflammatory drug target, and overexpression of this enzyme is also associated with several cancers and neurodegenerative diseases. The amino-acid sequence and structural similarity between inducible cyclooxygenase-2 and housekeeping cyclooxygenase-1 isoforms present a significant challenge to design selective cyclooxygenase-2 inhibitors. Herein, we describe the use of the cyclooxygenase-2 active site as a reaction vessel for the in situ generation of its own highly specific inhibitors. Multi-component competitive-binding studies confirmed that the cyclooxygenase-2 isozyme can judiciously select most appropriate chemical building blocks from a pool of chemicals to build its own highly potent inhibitor. Herein, with the use of kinetic target-guided synthesis, also termed as in situ click chemistry, we describe the discovery of two highly potent and selective cyclooxygenase-2 isozyme inhibitors. The in vivo anti-inflammatory activity of these two novel small molecules is significantly higher than that of widely used selective cyclooxygenase-2 inhibitors.

C yclooxygenase (COX) enzymes catalyze the metabolic conversion of arachidonic acid to prostanoids including prostaglandins (PGs), prostacyclin, and thromboxane, which play important roles in human physiology and various pathological conditions [1][2][3][4] . Despite several known side effects like myocardial infarction and atherothrombotic events, drugs aimed at COXs inhibition is a billion dollar industry, inspiring scientists to search constantly for novel COX inhibitors. COX exists in three isoforms: cyclooxygenase-1, 2, and 3 (COX-1, COX-2, and COX-3) [5][6][7] . COX-1 and COX-2 isoforms are of primary interest, as they are involved in physiological as well as pathological processes. COX-1 is a constitutively expressed house-keeping isozyme responsible for the basal production of essential PGs 8 . These PGs mediates homoeostatic functions in the gastrointestinal and cardiovascular system. COX-3 (a splice variant of COX-1) is expressed only in specific parts of the brain and spinal cord and its exact functions are still unclear 9 . In contrast, COX-2 isozyme is expressed at very low levels under normal conditions. However, COX-2 expression is rapidly upregulated in the immediate response to diverse pro-inflammatory and pathogenic stimuli. There is accumulating evidence for the critical involvement of COX-2 in various pathologies that include inflammation 10,11 , cancer [12][13][14] , neurodegenerative diseases 15 and multidrug resistance 16 . Therefore, beyond their traditional use as anti-inflammatory agents, COX-2 inhibitors have recently been used for molecular imaging [17][18][19] and therapy [20][21][22] of cancer. Hence, the development of selective COX-2 inhibitors as antiinflammatory and anti-tumor drugs is a major direction in academic research and pharmaceutical industry [23][24][25] . Traditional nonsteroidal anti-inflammatory drugs (NSAIDs) (aspirin, ibuprofen, naproxen) inhibit both COX-1 and COX-2 isoforms; and their use is limited due to associated ulcerogenic and gastrointestinal side effects. Discovered in the late 1990's, COX-2 selective inhibitors (the COXIBs: celecoxib, rofecoxib) are diarylheterocycles possessing a SO 2 NH 2 or SO 2 Me group as COX-2 pharmacophore, which exert similar anti-inflammatory and antipyretic properties as traditional NSAIDs but are devoid of gastrointestinal toxicity 4 . However, COXIBs are also under scrutiny since several studies have demonstrated that chronic use of COXIBs can elevate the risk of myocardial infarction and other thrombotic events by stalling the biosynthesis of antiaggregatory prostacyclin (PGI 2 ) while leaving the biosynthesis of pro-thrombotic thromboxane A 2 (TxA 2 ) unaffected [26][27][28][29] . As a result, COXIBs such as rofecoxib and valdecoxib were withdrawn from the market thereby leaving a demand for the synthesis and screening of novel COX-2 inhibitors.
Development of compounds that selectively inhibit COX-2 over COX-1 is a substantial challenge as both isoforms share similar cellular expression locations, molecular weight, and amino-acid composition. In addition, the both isoforms share more than 60% sequence homology and their three-dimensional structures are almost superimposable. However, the key difference between the COX-1 and COX-2 isozyme active site is the exchange of isoleucine in COX-1 for valine in COX-2 at positions 434 and 523. The difference in the amino-acid sequence make the COX-2 substrate-binding site more flexible and approximately 25% larger by creating a distinct secondary-binding pocket 3,30 . Many COX-2 selective inhibitors explicitly bind to this secondary-binding pocket (lined by H90, R513, and V523) resulting in the specific inhibition of COX-2 activity. Another important region in the COX-2 active site is the hydrophobic pocket (lined by W387, Y385, F518, F381, L352), and a recent mutational study described the involvement of hydrophobic pocket residues in the proper positioning of fatty acid substrates for oxygenation 31 . Therefore, highly potent and selective COX-2 inhibitors should possess a pharmacophore which can selectively bind in the secondary pocket and deliver sufficient steric bulk to block the hydrophobic channel of COX-2.
Here, we deviated from conventional drug discovery approaches involving the laborious synthesis and screening of a range of compounds, and envisioned to explore the utility of in situ click chemistry for the discovery of specific and high-affinity COX-2 inhibitors. Click chemistry 32,33 , including 1,3-dipolar cycloaddition between alkyne and azide (Huisgen cycloaddition), have attracted much attention because of their remarkable efficiency, simplicity, and its ability to be employed for the synthesis of a wide range of compounds such as molecular imaging agents 34,35 and drugs 36 , protein modification 37,38 , DNA and RNA targeting 39,40 , and glycan imaging 41,42 . Considering its versatility, click chemistry has found numerous synthesis applications not only performed in traditional reaction vessels, but also in living systems. Kinetic target-guided synthesis (TGS), also termed as in situ click chemistry is an innovative synthesis process where a biological target assembles its own inhibitor through targetguided selection of appropriate building blocks 43,44 . TGS was elegantly used for the preparation of high affinity inhibitors of enzymes like acetylcholine esterase [45][46][47] , HIV protease 48 , bovine carbonic anhydrase II 49,50 , protein tyrosine phosphatases 51 , metalloproteases 52,53 , nicotinic acetylcholine receptors 54 , and chitinase 55 .
Herein, we demonstrate the use of the COX-2-binding site as a reaction vessel for generating its own highly potent and selective inhibitors.We designed and synthesized a range of pyrazole-based azide building blocks (5, 14, 27, and 31) and a collection of corresponding triazole-containing biheterocyclic compounds (7-12, 16-25, 28, 29, 32, and 33). After screening for their COX-1/COX-2 inhibitory potency, various 5-azido-pyrazoles (5, 14, 27, and 31) and aryl acetylenes as click chemistry building blocks were incubated in pairs with human recombinant COX-2 isozyme to test the capability of COX-2 for assembling its own highly potent inhibitors. Identification of compounds 18 and 21 as highly potent and selective COX-2 inhibitors demonstrated the feasibility of the in situ click chemistry approach. Compounds 18 and 21 displayed a superior in vivo anti-inflammatory activity profile compare to clinically used anti-inflammatory drugs. Multi-component competitive-binding studies confirmed that COX-2 can selectively group most appropriate building blocks from a pool of compounds to construct highly potent COX-2 inhibitors. The thermodynamic-binding signatures calculated from isothermal titration calorimetry (ITC) confirmed that binding of click chemistry building block 5-azido-pyrazole (14) to COX-2 involves a favorable change in free energy (ΔG = −36.20 kJ mol −1 ), which was mainly based on H-bonding and van der Waals interactions. Moreover, comprehensive computational analysis including structure activity relationship (SAR) and molecular docking indicated that the size and type of the COX-2 pharmacophore, and the orientation of the clickable building blocks inside the binding site of the target protein collectively contribute to the in situ construction of highly potent and selective COX-2 inhibitor.

Results
Design of clickable building blocks. Based on the structural features of the COX-2 active site, we concluded that suitable clickable building blocks should meet two criteria: (i) at least one of the building blocks (in our case, the azide component) should possess a SO 2 Me COX-2 pharmacophore to facilitate its tight binding into the secondary-binding pocket of the COX-2 isozyme; (ii) the azide component should have a proper size and orientation that will not interfere with the entry of the alkyne component into the COX-2 active site, allowing the in situ click chemistry formation of a potent and selective inhibitor large enough to block the hydrophobic channel of the COX-2 isozyme (Fig. 1).
Hence, to optimize these parameters we synthesized a range of 5-azido-pyrazole derivatives with variable sizes containing a COX-2 pharmacophore (SO 2 CH 3 ) and evaluated the influence of the chemical structure on the success of in situ click chemistry reaction. We selected a pyrazole motif as the central scaffold for the azide building blocks, as pyrazole-containing compounds are quite prevalent in many anti-inflammatory drugs. The design of click chemistry building blocks and target compounds is illustrated in Fig. 1, where the central pyrazole motif was adapted from approved drug celecoxib (1). The SO 2 CH 3 COX-2 pharmacophore is located at the para position of one of the phenyl ring which was introduced to facilitate binding of 5-azidopyrazole (14) into the COX-2 secondary-binding pocket. Extensive SAR data is available in the literature, which describes the importance of SO 2 CH 3 and SO 2 NH 2 groups as COX-2 pharmacophores for selective binding to the COX-2 isozyme. However, SO 2 NH 2 groups are also found in many drugs inhibiting members of the carbonic anhydrase family 56 . Substitution pattern on the alkyne building blocks selected for this study was based on structures frequently present in various NSAIDs.
Furthermore, the influence of the size and the role of the SO 2 CH 3 COX-2 pharmacophore of the 5-azido-pyrazole building blocks upon in situ click chemistry reaction was studied with two newly synthesized 5-azido-pyrazole building blocks 27 and 31 (Fig. 2). In contrast to 5-azido-pyrazole (14), the new building block 31 lacks the SO 2 CH 3 COX-2 pharmacophore. Upon incubating 5-azido-pyrazole (31) with alkynes (6c or 15a) in the presence of COX-2 isozyme, no triazole product formation was detected even after 36 h. This experiment demonstrated that (for library 3) the presence of the SO 2 CH 3 COX-2 pharmacophore is an important structural requirement for the appropriate positioning of 5-azido-pyrazole building block (14) inside the COX-2-binding site to achieve a suitable steric orientation of the azide building block needed for the COX-2 mediated in situ click chemistry reaction. Another set of experiments involved the incubation of phenyl group-containing 5-azido-pyrazoles 27 (with SO 2 CH 3 COX-2 pharmacophore) and 5 (without SO 2 CH 3 COX-2 pharmacophore) as examples for larger size azide building blocks with the alkyne building blocks (6c or 15a) in the presence of COX-2 isozyme. And in none of the reagent combinations, no in situ generation of expected triazole product was detected. These results led to the following important two conclusions: (i) the presence of the SO 2 CH 3 COX-2 pharmacophore is the defining factor to advance the in situ click chemistry reaction between azides and alkynes inside the COX-2-binding site; (ii) a larger sized (additional steric bulk)5-azido-pyrazole building block like compound (27) greatly reduces the possibility to achieve an in situ click chemistry reaction in the COX-2-binding site. The results of in situ click chemistry reaction with various azides (5, 14, 27, and 31) and alkynes (6a-f, 15a-e) in the presence of the COX-2 isozyme are summarized in Fig. 4.
We also tested whether COX-2 isozyme is capable of selecting the most suitable compounds from a combined mixture of different azides and alkynes building blocks for assembling the "best-fit inhibitor." For this purpose, each azide (5, 14, 27, and 31, 1 µl of 3 mM DMSO solution) was mixed with eleven alkynes (6a-6f and 15a-15e, 1 µl of 20 mM DMSO solution) in the presence of COX-2 isozyme (95 µl COX-2) in 1 µl of 1 M Tris-HCl, pH 8.0 and incubated for 24 h at room temperature. After incubation for 24 h, only expected triazole products (18,21) were detected, demonstrating that COX-2 isozyme can sense small structural differences of the building blocks enabling selection of suitable building blocks for the in situ click chemistry formation of high affinity and selective COX-2 inhibitors (Fig. 4). The amount of triazole product 21 formed seemed to be higher compared to triazole 18 (estimated from the mass response), although the yield of both triazole products (18,21) was lower (almost 50% reduced) compared to previously performed in situ click chemistry reaction when both building blocked were used as   In parallel to the COX-2 isozyme mediated in situ click chemistry reactions, similar experiments were also performed using COX-1 isozyme. However, no in situ click chemistry reaction to form the desired triazole products was not observed in any of the used reagent combinations. As demonstrated in a later in vitro COX inhibitory-binding assay, none of the compounds displayed inhibitory potencies in the studied concentration range towards COX-1 isozyme. Thus, in situ click chemistry can also be used for the identification of highly selective COX-2 inhibitors. Compound 5, 14, 27, and 31 were also studied for their stability in human and rat serum at 37°C, and no significant degradation was observed up to 30 h (see Supplementary Fig. 3). The stability of hit compounds (18 and 21) was tested, and the percentage of intact compound at various time points is given in Supplementary  Fig. 3.
Isothermal titration calorimetry. The success of identifying 5-azido-pyrazole (14) as suitable building block for in situ click chemistry reaction prompted us to determine the protein-ligandbinding stoichiometry (N), change in thermodynamic parameters, and association constant (K a ) for the binding of 5-azido-pyrazole (14) with pure COX-2 protein under physiological conditions using ITC. The determined binding stoichiometry for compound 14 (N = 1.02 ± 0.0259) indicated that one molecule of the 5-azido-pyrazole (14) was bound per COX-2 isozyme. The high association constant (K a = 2.20 × 10 −6 M −1 ) and favourable free energy change (ΔG < 0; ΔG = −36.20 kJ mol −1 ) clearly supports the assumption of tight binding of compound 14 to the COX-2 protein (Supplementary Table 1). The ITC results summarized in Fig. 5 demonstrate that protein-ligand interactions are thermodynamically favored and that they are mainly driven by the enthalpy factor. The binding of compound 14 to COX-2 isozyme involves an exothermic reaction (ΔH < 0; ΔH = −34.51 kJ mol −1 ) which is indicative of a combination of H-bonding, π-π stacking and ion-dipole interactions. Interestingly, docking studies have also supported the contribution of H-bonding in protein-ligand binding. The determined favorable small entropic change (TΔS > 0; TΔS = 1.69 kJ mol −1 ) also reflects the little contributions of electrostatic contact, conformational freedom, and solvent reorganization.
The relative inhibitory potency and selectivity profile were observed in the following order: 21 (R 1 = NH 2 ), >9, 18 (R 1 = F) >8, 17 (R 1 = Cl), which shows that the para substituent (R 1 ) on the C-1 phenyl ring of triazole is crucial for both COX-2 inhibitory potency and high COX-2 selectivity. However, compounds 23-25 containing heterocyclic or aliphatic (ibuprofen like) motifs on C1 position of triazole ring displayed only low inhibitory potencies towards COX-2. Comparison of COX-2 inhibitory potency profiles of compounds 28, 29, 32, and 33 with 7-12 and 16-25 confirmed the importance of the SO 2 CH 3 COX-2 pharmacophore as an important structural requirement for high COX-2 inhibition. Among all tested compounds, triazoles 18 and 21 were identified as the most potent and selective COX-2 inhibitors. Thus, obtained in vitro COX-1/2 enzyme inhibitory data evidently demonstrate that in situ click chemistry reaction can also be employed to identify highly potent and selective COX-2 inhibitors. However, compounds 8 (COX-2 IC 50 = 0.16 µM), 9 (COX-2 IC 50 = 0.14 µM), and 22 (COX-2 IC 50 = 0.22 µM) also showed appreciable COX-2 inhibitory potencies in the in vitro COX-binding assay albeit nor formation of compounds 8, 9, and 22 was detected in the COX-2 mediated in situ click chemistry reaction. Based on the in vitro COX-2 inhibitory data, seventeen compounds (Table 2)   Molecular docking studies. To further support our experimental results and to investigate the mode of ligand-protein molecular interactions, all click chemistry building blocks (6, 14, 27, 31, 6a-6f, and 15a-15e) and final triazole compounds (7-12, 16-25, 28, 29, 32, and 33) were docked into the COX-2 active site. It was found that the para SO 2 CH 3 group in 5-azido-pyrazole compound (14) was deeply inserted into the secondary-binding pocket region of the COX-2 active site ( Fig. 7 and Supplementary  Fig. 6). The O-atoms of the SO 2 CH 3 group showed H-bonding interaction with one of the terminal amino group of R513 (distance O---NH 2 = 2.25 Å) and with the nitrogen atom of the imidazole ring of the H90 residue (distance O---N = 1.92 Å). Interestingly, the other 5-azido-pyrazoles (5, 27, and 31) were located away from the secondary pocket region, and did not indicate any H-bonding interactions ( Supplementary Fig. 7), except for compound 27 where one of the O-atoms of the SO 2 CH 3 group was H-bonded to the-NH 2 group of R120 (distance O---NH 2 = 1.96 Å). Alkynes possessing either a-F or -NH 2 group (6c and 15a) were positioned in the vicinity of azide group of compound 14 ( Fig. 5 and Supplementary Fig. 9), whereas the other three alkynes (6b, R 1 = Cl; 6d, R 1 = OCH 3 ; and 15b R 1 = CF 3 ) were clustered together near Y385; alkynes with para CH 3 or para C 2 H 5 groups and heterocyclic rings (6e, 6f, and 15c-15e) were also docked far away from the azide group of pyrazole compounds (6,14, 27, and 31). Molecular docking indicated that 5-azido-pyrazole (14) and alkynes (6c and 15a) were favorably oriented in such a way that the azide group (-N 3 ) is present in proximity of 4-ethynyl group (-C≡CH) of both alkynes; and the calculated distance between the terminal carbon atom of the ethynyl groups in 6c (R 1 = F) and 15a (R 1 = NH 2 ) and one of the nitrogen atoms of the azide group in compound 14 was 1.78 and 2.85 Å, respectively (Fig. 7). Therefore, the observed formation of in situ click chemistry products (18 and 21) can be attributed to the close and special orientation of reactive functional groups of the complementary click chemistry building blocks (14, 6c, and 15a) in the COX-2-binding site.
Further molecular docking studies with in situ click chemistry products (18,21) Fig. 9). In the case of compound 21, the O-atoms of the para SO 2 CH 3 group indicated H-bonding with Y355 (distance O---HO = 2.39 Å) which is considered as one of  Fig. 9). In contrast, upon docking of all azide compounds (5, 14, 27, and 31) and alkynes (6a-6f and 15a-15e) into the COX-1 active site, it was found that none of the azide compounds were located near any of the alkyne compounds, indicating less opportunity for an in situ click chemistry reaction in the COX-1 active site. The preference of the COX-2 isozyme for in situ click chemistry reaction over the COX-1 isozyme can be explained by considering the structural differences of both enzymes. The size of the COX-2 isozyme active site (volume = 394 Å 3 ) is about 25% larger than the COX-1 isozyme-binding site (volume = 316 Å 3 ). Moreover, the COX-2 isozyme possesses an additional secondary pocket region. Because of these two distinct structural differences, the COX-2 isozyme can conjointly hold two reactive/more bulky ligands more appropriately than the COX-1 isozyme. The energy of intermolecular interactions (E intermolecular ) between the ligand and the enzyme residues obtained after computational studies (docking) for all compounds within the COX-1 and COX-2 active site are summarized in Supplementary Table 2 and Supplementary Fig. 8. Interestingly, compound 18 and 21 occupy an orientation that is very similar to the binding mode of crystallized COX-2 inhibitor SC558 in the COX-2 isozyme (Fig. 8). Overall, the computational results suggest that similar to the orientation of parent azide click chemistry building block 14, the 4-methanesulfonyl-phenyl motif of in situ click chemistry hit compounds 18 and 21 occupy conformational spaces in the secondary-binding pocket region of the COX-2-binding site, and the fluoro/amino-phenyl ring portion is embedded into the hydrophobic pocket; H90, R120, Q192, Y355, Y 385, R513, and F518 are the key amino-acid residues, which considerably contributed in the strong protein-ligand interactions. All these observations agree with the experimental results and support the success of in situ click generation of 18 and 21 from clickable building blocks 14 (azide) and 6c/15a (alkyne) in the COX-2binding pocket.

Discussion
We have, for the first time, demonstrated the use of in situ click chemistry for the identification and synthesis of highly potent and selective COX-2 isozyme inhibitors. In situ click chemistry with COX-2 isozyme as molecular template was highly regioselective and products are exclusively formed as 1,4-regioisomers through [3,2]-cycloaddition reaction inside the COX-2-binding pocket. Our comprehensive SAR and computational studies suggest that size, presence of a SO 2 Me COX-2 pharmacophore, proper orientation of the azide click chemistry building block inside the  into the LC/MS instrument by following the procedure described above, except the ions are monitored for all possible masses. The cyclo addition products were identified by their molecular weights and by comparison of the retention times of authentic products prepared through Cu-catalyzed reactions. Control experiments using BSA (1 mg ml −1 ) in place of COX-2 isozyme and in the absence of COX-2 isozyme were run consecutively.
Isothermal titration calorimetry. The thermodynamics of the precursor 5-azidopyrazole (14) binding with pure human COX-2 protein studied with a high sensitivity ITC instrument (VP-ITC, MicroCal). All titrations were performed by following VP-ITC general procedure described in manual. Protein (20 µM) and ligand (200 µM) solution were prepared in PBS buffer, pH 7.4. Prior to each ITC run, all solutions were filtered using membrane filters (pore size 0.45 μm) and thoroughly stirred and degassed using Thermovac accessory for 20 min to remove any air bubbles at 25°C. In a typical experiment, aliquots (6 µl) of ligand solution (prepared at 10 times the COX-2 protein concentration) were injected with a computer-controlled stirrer-syringe into a reaction cell containing COX-2 protein.
Titrations were carried out with a stirring speed of 400 rpm and 240 s intervals between injections. All experiments were conducted at 25°C. Control experiments, including titration of the ligand into buffer alone or COX-2 protein into buffer alone, were carried to determine the heat due to dilution and subtracting during the original experiment. Calorimetric data analysis was performed with ORIGIN software provided by MicroCal. All results, including the binding constant (K a ), stoichiometry of binding (N), and thermodynamics of binding (ΔH and TΔS) were determined by fitting the experimental binding isotherms. Each ITC experiment was repeated three times and showed good reproducibility.
In vivo anti-inflammatory activity. The two lead compounds 18 and 21 were advanced for in vivo anti-inflammatory activity study and celecoxib (1) was used as reference drug. In vivo anti-inflammatory activity was measured using a carrageenan-induced rat paw edema assay. In brief, three to five male Sprague-Dawley rats, 8-11-weeks-old, weighing 180-200 g (Charles-River Canada) were used in each group. Animals were randomized into different treatment groups based on similar paw size and body weight. Test compounds 18 and 21 suspended in water containing 1% methyl cellulose were administered orally for a minimum of four different doses (0.3. 1, 5, 10 mg kg −1 ) 1 h prior to a 0.05 ml subcutaneous injection of fresh 1% carrageenan in 0.9% NaCl solution under the plantar skin of the hind paw. Control experiments were identical, except that the vehicle did not contain a test compound. The volume of the injected paw was measured at 0, 3, and 5 h using a UGO Basile 7141 Plethysmometer (series no. 43201), each value is mean of 10 measurements. A dose-response curve was constructed using GraphPad Prism 5.0 and ED 50 were calculated. No unusual change in behavior and toxic effects was noticed in all animals. In vivo anti-inflammatory assays were carried using a protocol approved by the Health Sciences Animal Welfare Committee, University of Alberta, Edmonton, Canada.
Molecular docking procedure. The detailed procedure for the molecular docking is provided in the Supplementary Information. Statistical analysis. Unless specified, all data were obtained from at least triplicate samples and represent at least three independent experiments and presented as mean ± s.e.m. Graphs were constructed using GraphPad Prism 4.0 (GraphPad Software). Where applicable, statistical differences were tested by unpaired Student's t-test and were considered significant for P < 0.05.
Data availability. The authors declare that all data supporting the findings of this study are available in the article and in the supplementary information file. Additional information are available from the corresponding author upon request.