People often wonder how selective fragments need to be. According to molecular complexity theory, the answer is “not very”. After all, it would be hard to get a decent hit rate with a library of just a few thousand fragments if they were too selective. In the case of kinases, experimental studies support this theory. Indeed, a single fragment has given rise to several drugs – one of which is approved. In a new paper in J. Med. Chem., William Shipe and colleagues at Merck demonstrate the utility of a non-selective fragment for another class of enzymes, phosphodiesterases (PDEs).
The human genome contains more than 50 different PDEs, which cleave phosphodiester bonds. PDE10A hydrolyzes cyclic guanosine monophosphate (cGMP) and cyclic adenosine monophosphate (cAMP) and is a potential target for schizophrenia. It has been pursued extensively, both with fragments (see for example here and here) as well as more traditional approaches.
The researchers started with a biochemical assay that screened each fragment at 200 µM; 60 of the 1600 tested gave > 80% inhibition. Nine of these were soaked into PDE10A crystals, producing seven structures, including compound 5, with impressive potency and ligand efficiency. Initial SAR by catalog led to the even more potent compound 6, which revealed that an amino group was tolerated and pointed nicely towards another pocket, offering a way for further elaboration.
Fragment growing from the amino group was accomplished through several rounds of parallel synthesis, with crystallography used to understand and optimize the binding interactions. Compound 9s showed particularly impressive low nanomolar potency, as well as at least 80-fold selectivity against nine other PDEs. In contrast, the initial fragment 5 was at most only 11-fold selective against any of the other PDEs.
Previous work with PDE10A had revealed another “selectivity pocket” nearby, and the researchers further grew their molecule towards this, leading ultimately to compound 15h, with low picomolar affinity and at least >5900-fold selectivity against nine other PDEs. The compound also showed functional activity in a rat model, though it suffered from suboptimal pharmacokinetic properties.
This is a beautiful illustration of the power of combining fragment screening, structure-based drug design, and parallel synthesis. The researchers were able to gain more than a million-fold improvement in potency and take a marginally selective fragment to a highly selective lead. Of course, there is still plenty of work to do, and it will be fun to watch this story unfold.