The results support the hybrid scaffold as a starting point for further anti-inflammatory lead optimization, but the animal experiment did not directly measure COX-2 inhibition or establish COX isoform selectivity. The investigation addressed a well-known limitation in nonsteroidal anti-inflammatory drug research: conventional NSAIDs reduce pain and inflammation by inhibiting cyclooxygenase enzymes, but inhibiting both COX-1 and COX-2 can contribute to gastrointestinal, renal and cardiovascular complications. Ketoprofen is a propionic-acid NSAID with analgesic and anti-inflammatory effects, although its nonselective inhibition of COX isoforms may contribute to the adverse effects associated with this class.

The researchers reasoned that adding a thiazolidin-4-one ring to the ketoprofen framework could change molecular volume, polarity, hydrogen-bonding capacity and three-dimensional orientation in a way that might improve compatibility with the larger, more flexible COX-2 binding channel. That structural hypothesis is linked to a consequential difference between the two main COX isoforms. COX-1 contains a bulkier isoleucine residue at position 523, whereas COX-2 contains the smaller valine at that position.

Together with other local differences, this substitution creates an additional side pocket within the COX-2 active-site channel that is less accessible in COX-1. Selective inhibitors can exploit this additional space by occupying the hydrophobic channel while directing bulky or properly oriented polar groups into the COX-2 side pocket. The investigators used the experimentally resolved structure of human COX-2 bound to rofecoxib, deposited as Protein Data Bank entry 5KIR, as the principal docking target.

They also examined whether the designed molecules could interact with residues including Arg120, Tyr355, Tyr385, Ser530, Glu524, Trp387, Phe518 and Val523. Eight hybrid designs were evaluated computationally, and four were selected for synthesis and biological testing. The synthetic route began with conversion of ketoprofen into its methyl ester, followed by hydrazide formation, condensation with substituted benzaldehydes to create four hydrazone intermediates, and thioglycolic-acid-mediated cyclization to produce the thiazolidin-4-one derivatives.

The final compounds carried para-nitro, para-chloro, para-hydroxy or para-dimethylamino aryl substituents and were designated AR1 through AR4. Ketoprofen methyl ester was obtained in 95% yield, ketoprofen hydrazide in 50%, the hydrazone intermediates in 50% to 70%, and the final thiazolidin-4-one derivatives in 65% to 84%. Infrared and nuclear magnetic resonance measurements supported the expected functional-group transformations, including new ring-related aliphatic signals and carbonyl features in the products.

Docking results were broadly favorable but depended on the scoring method. In the initial GOLD PLP fitness workflow, AR1 ranked highest among the designed compounds with a score of 81.10, followed by AR2 at 77.11, AR4 at 76.31 and AR3 at 74.24. Those four highest-ranking derivatives advanced to synthesis and biological evaluation.

The protocol reproduced the crystallographic ligand pose with a root-mean-square deviation of 0.276 angstroms across 33 matched atom pairs. Tyr355 repeatedly formed hydrogen-bonding contacts, while Arg120 anchored several compounds near the entrance to the COX-2 channel. Substituent-dependent contacts with His90, Leu352, Tyr385, Ser530, Glu524, Trp387, Phe357 and Phe518 changed the internal interaction patterns.

The comparative docking trial with COX-1 produced a provisional computational signal. Ketoprofen generated an acceptable predicted pose in the COX-1 structure 1EQG, whereas the hybrid derivatives and rofecoxib did not generate acceptable poses under the conditions used. This may be consistent with the bulkier hybrids fitting COX-2 more readily than COX-1, but the authors emphasized that failure to obtain a docked pose is not proof of biochemical selectivity.

Docking outcomes depend on receptor preparation, search parameters and scoring functions. Experimental COX-1 and COX-2 inhibition assays were therefore needed to determine whether the predicted isoform preference translated into true biochemical selectivity. The same caution applied to the molecular dynamics results, which were performed only for AR1 rather than for the entire series or the reference drugs.

In the AR1–COX-2 molecular dynamics simulation, the protein C-alpha root-mean-square deviation rose initially and then fluctuated around a stable plateau, while monitoring indicated that the global protein fold remained preserved. AR1 retained recurring interactions with Tyr385, Tyr355, Arg120 and Glu524, supporting a plausible association with the active site over the 100.102-nanosecond production trajectory. Ligand RMSD increased toward the end of the simulation, which was interpreted as conformational adjustment within the pocket rather than clear dissociation.

The simulated system contained 55,873 atoms, including 15,632 water molecules, and ran at 300 kelvin under an NPT ensemble. Because only one candidate underwent dynamic simulation, the result provides support for AR1’s predicted binding mode but does not establish that it is dynamically more stable than the other derivatives or reference compounds. Predicted absorption, distribution, metabolism and excretion properties suggested that the designed series was generally within an oral drug-like range, although aqueous solubility emerged as a limitation.

All eight computationally assessed derivatives had molecular weights below 500 daltons, and predicted human oral absorption ranged from 87.473% to 100%. AR1, AR3, AR5, AR6 and AR8 had no predicted Lipinski Rule-of-Five violations, whereas AR2, AR4 and AR7 had one violation each. The compounds with violations were generally more lipophilic.

Most derivatives also had more negative QPlogS values than the two reference compounds, indicating lower predicted aqueous solubility despite favorable absorption estimates. AR3 combined no Rule-of-Five violation with a predicted oral absorption of 93.751%, while AR1 had a predicted absorption of 87.473% and no violation. The results identified solubility as a principal developability challenge, particularly for the more lipophilic analogues.

Anti-inflammatory activity was assessed with egg-white-induced paw edema in 36 female albino mice allocated equally to six groups: ketoprofen, AR1, AR2, AR3, AR4 and vehicle. The mean body weight was 20.34 grams. Treatments were delivered intraperitoneally 30 minutes before paw swelling was induced.

Ketoprofen was administered at 5 milligrams per kilogram, while the hybrids received molecular-weight-adjusted equimolar doses: 9.35 mg/kg for AR1, 9.14 mg/kg for AR2, 8.78 mg/kg for AR3 and 9.31 mg/kg for AR4. Analysis of baseline-adjusted paw-thickness changes from 60 to 240 minutes showed significant effects of treatment group, time and the treatment-by-time interaction. Ketoprofen significantly reduced edema versus vehicle at 60 minutes by an estimated 0.513 millimeters, with p = 0.028, but not at later time points.

The hybrids displayed a delayed pattern: AR3 and AR2 were significantly below vehicle at 120 minutes, with differences of 0.591 millimeters and 0.504 millimeters, respectively, while AR1, AR2 and AR3 remained significantly below vehicle at 180 and 240 minutes. AR3 emerged as the strongest biological candidate in the tested set. Its model-adjusted change in paw thickness was 0.257 ± 0.104 millimeters at 120 minutes, −0.071 ± 0.104 millimeters at 180 minutes and −0.251 ± 0.104 millimeters at 240 minutes, compared with 0.848 ± 0.123, 0.666 ± 0.123 and 0.421 ± 0.123 millimeters for vehicle.

It also produced lower edema than ketoprofen at 240 minutes by an estimated 0.436 millimeters after Bonferroni adjustment, with p = 0.031. AR4 was less convincing, showing greater edema than ketoprofen at 60 minutes by an estimated 0.495 millimeters, with p = 0.007, and no comparable late-phase advantage. The divergence between computational and biological rankings shows that docking scores alone cannot capture solubility, tissue exposure, distribution following intraperitoneal administration or the complexity of the inflammatory response.

The study therefore identifies AR3 as the best-supported initial lead and AR1 as a computationally promising, biologically active alternative, while calling for biochemical selectivity tests, expanded pharmacological and safety evaluation, pharmacokinetic profiling, solubility optimization and comparative molecular dynamics simulations.

Subject of Research: Computational and mouse evaluation of ketoprofen–thiazolidin-4-one hybrid anti-inflammatory candidates

Article Title: Design and synthesis of thiazolidin-4-one–ketoprofen hybrids as potential COX -2 inhibitors with anti-inflammatory activity

Article References: Mazin, A., Mahdi, M. F., & Kareem Khan, A. (2026). Design and synthesis of thiazolidin-4-one–ketoprofen hybrids as potential COX-2 inhibitors with anti-inflammatory activity. Pharmacia, 73, Article e194751. https://doi.org/10.3897/pharmacia.73.e194751

Image Credits: AI Generated

DOI: 10.3897/pharmacia.73.e194751

Keywords: ketoprofen hybrids, thiazolidin-4-one, COX-2, anti-inflammatory activity, mouse paw edema, molecular docking, molecular dynamics, ADME prediction, drug discovery, medicinal chemistry, Design, synthesis