Fragments vs PDE10A revisited

Independent teams have reported using fragments to identify structurally distinct inhibitors against a popular psychiatric target.

Last month Practical Fragments highlighted a paper from Merck describing researchers’ success in advancing a fragment to a potent selective inhibitor of PDE10A, a potential target for schizophrenia. The final molecule had picomolar activity but suffered from various shortcomings, and the post ended by stating that “there is still plenty of work to do, and it will be fun to watch this story unfold.” Well, we didn’t have to wait long: a new paper in Bioorg. Med. Chem. Lett by Izzat Raheem and Merck colleagues describes further optimization of this series – again using fragments.

The team started by making various changes to compound 15h (shown in the previous post), ultimately leading to compound 4. Although this had lower affinity, it had significantly improved solubility and pharmacokinetic properties. Unfortunately, although selective against other PDEs, it was less selective against a broader panel of off-targets and inhibited both CYP2C9 and CYP3A4. In fact, 1000 analogs (!) containing the central fragment also hit these two enzymes, suggesting the problem was inherent to this core.

At this point the researchers returned to their original fragment screen and recognized that compound 5 had a similar structure to the original fragment. Appending the two “arms” of compound 4 onto this core led to the compound called Pyp-1, with good potency, solubility, and >5800-fold selectivity against other PDEs. Importantly, this molecule did not show the CYP activity of the previous series, and also displayed good pharmacokinetic properties in rats, dogs, and rhesus monkeys. A rat toxicity study didn’t reveal any red flags, and the molecule showed good pharmacodynamic effects in several animal models. The researchers acknowledge that this is a crowded field, with at least 7 compounds having entered the clinic, but Pyp-1 looks promising; at the very least it is a worthy chemical probe.

Continuing the theme of PDE10A, a second paper in Bioorg. Med. Chem. Lett. by Jeffrey Varnes and Jeffrey Albert reports an earlier-stage program from AstraZeneca. In this case, the researchers used a fragment-assisted drug discovery approach, integrating fragment information with data from high-throughput screening.

A functional screen of 3000 fragments led to a fairly high hit rate, with 414 compounds having ligand efficiencies ≥ 0.3 kcal/mol per atom. Many of these were similar to previously described PDE10A inhibitors and were thus deprioritized. On the other hand, compounds 6 and 7 were rather unusual structurally.

A high-throughput screen was conducted at the same time, and this also generated a high hit rate: ~5%, or 11,000 compounds. Unlike the Merck group, the AstraZeneca researchers were unable to obtain crystal structures of their fragments bound to PDE10A, so instead they looked for HTS hits similar to fragments 6 and 7, resulting in 14 compounds. Most of these were false positives or contained unattractive functionalities, but compound 8 turned out to inhibit significantly better than either fragment. Further medicinal chemistry led to compound 12, which is both potent and structurally distinct from other PDE10A inhibitors.

Together these papers reveal how fragments can be exploited to develop quite different molecules against the same target. Although the Merck series is clearly more advanced, it is impressive that the AstraZeneca work was done in the absence of crystallographic support. And in both cases, medicinal chemistry played an essential role: Valinor may beckon, but it will have to wait.