Naproxen is one of the original non-steroidal anti-inflammatory drugs (NSAIDs) that target the cyclooxygenase (COX) enzymes, yet the molecular basis of its interaction with COX enzymes has not been well established. Through a structure-activity-guided approach, Lawrence Marnett and colleagues uncover interactions that dictate naproxen binding to COX-2 and distinguish it from other NSAIDs that bind the COX enzymes.

There are at least four structural classes of molecules that characterize NSAIDs. Naproxen is a relatively simple molecule, containing three functional groups: an arylpropionic acid (carboxylate), a p-methoxy group and an α–methyl group on a naphthyl scaffold. As a drug, naproxen inhibits both COX-1 and COX-2 with comparable IC50s and exhibits gastrointestinal side effects, but reduced cardiovascular side effects relative to other selective as well as non-selective COX inhibitors.

To probe the interaction between naproxen and the COX enzymes, Marnett and colleagues mutated residues R120 and Y355 of murine COX-2 (mCOX-2), as these residues have been identified to mediate the canonical interaction of the arylcarboxylic acid family of NSAIDs. These mutations suggest that the carboxylate group of naproxen binds in the canonical orientation, coordinated to these two residues. To verify this and to identify other naproxen interactions with mCOX-2, the authors solved the co-crystal structure at 1.7 Å resolution.

This high resolution structure enabled visualization of solvent, ion and detergent molecules not observed in lower resolution structures of the COX enzymes, as well as subtle differences from existing co-crystal structures with COX-2. The naproxen-mCOX-2 complex also revealed a single orientation of naproxen within the COX-2 active site. Beyond the hydrogen-bonding interactions of the naproxen carboxylate group to R120 and Y355 predicted from the mutagenesis data, all other interactions are van der Waals contacts. W387 is an important residue at the top of the mCOX-2 active site. That naproxen was unable to inhibit a W387F mutant, whereas other NSAIDs like diclofenac and indomethacin could inhibit enzymatic activity suggests that an interaction between naproxen and this residue is specific to this NSAID.

Turning to mutagenesis of the naproxen molecule, the authors substituted the α–methyl group with hydrogen, ethyl or dimethyl substituents and found a loss of enzymatic potency with each of them, indicating that an (S)-methyl group at this position is critical. The co-structure shows that this group inserts into a small hydrophobic cleft below V349, which seems to anchor naproxen within the active site and reinforce the canonical binding orientation. A similar structure-activity relationship analysis of the naproxen p-methoxy group suggested the importance of this constituent at the top of the enzyme active site channel. Interestingly, p-ethyl and p-methylthio substitutions at this position still allowed inhibition of mCOX-2, but these compounds could not inhibit ovine COX-1.

Also surprising was that both of these analogs inhibited the W387F COX-2 mutant as effectively as they inhibited the wild-type enzyme.

To follow up on these observations of COX-2 selectivity, the authors solved a second co-structure, this time between the p-methylthio naproxen derivative and mCOX-2. Compared to the naproxen co-structure, there was a shift of the carboxylate tails of the compounds and a new interaction between the naphthyl backbone of the p-methylthio naproxen, while many of the other interactions remain the same in the two co-structures.

These results define the contribution of the COX protein and naproxen atoms to the affinity and suggest that there is little tolerance for mutation in either component. This could tend to complicate any attempts to modify naproxen to eliminate the unwanted gastrointestinal side effects of the drug.