A high-throughput screen against the diabetic drug target fructose-1,6-bisphosphatase (FBPase) identified a small thiol with a low micromolar IC50. FBPase is a homo-tetramer, and when the crystal structure of the HTS hit bound to the protein was solved, the researchers found that the thiols had oxidized to form disulfides: two dimers bound to each tetramer, with one phenyl sulfonylurea bound in each of four binding sites. Subsequent analysis of the screening hit revealed it to be contaminated with about 5% of the disulfide, which had presumably formed by air oxidation; this was the source of the inhibition in the HTS assay. In other words, the screen had identified linked fragments.
With this structure in hand, the researchers synthesized a series of dimeric molecules connected by various linkers; they also replaced the aniline with meta-substituted phenyl moieties. Clear linker-length dependence was observed, with the shortest linkers showing no activity, while the molecule with the six-carbon linker shown has an IC50 of 17 nM. Crystallography revealed that this molecule binds in a similar manner as the disulfide, although the linker itself showed some disorder. Gratifyingly, when the molecule was cut in half, the resulting monomer was found to bind much less tightly, with lower ligand efficiency. The dimeric molecule binds with a free energy of 10.6 kcal/mol, almost 2 kcal/mol more than would be predicted by simply adding the binding energies of the monomers.
Of course, the molecule is far from a drug (although it does show an impressive 100% oral bioavailability in mice). Nonetheless, it illustrates one of the key advantages of fragment-based ligand discovery. Fragments linked together can really be more than the sum of their parts.