Fragments vs PAK1 – allosterically

Some of our recent posts have discussed the use of fragment-based approaches to discover selective new chemical probes. Continuing this theme, a paper from Alexei Karpov and colleagues at Novartis in ACS Med. Chem. Lett. describes a success story against the kinase PAK1.

The six p21-activated kinases (PAKs) have been implicated in a variety of indications, from cancer to neurodegenerative diseases. Unfortunately, most reported inhibitors are not sufficiently selective to elucidate the biology. The researchers were particularly interested in PAK1, against which they performed a fragment screen (no details of methods or libraries are provided). One of the more interesting hits was fragment 1. Although a bit chunky (24 heavy atoms, MW = 328 Da) with only modest ligand efficiency, it is structurally unusual for a kinase binder.

In fact, crystallography revealed that the fragment binds not in the active site at all, but rather in an allosteric site adjacent to the ATP-binding site, with the so-called DFG-loop in the “out” conformation. Not surprisingly, this molecule was more active against the inactive form of the enzyme. Despite its binding mode, it did appear to be ATP-competitive, perhaps because the DFG-out conformation of the protein is not able to bind ATP.

Aficionados of GPCRs will not be surprised to learn that fragment 1 – a dibenzodiazepine – bound several with high affinity, but the researchers used structural information both to ablate this off-target activity as well as improve affinity for PAK1, resulting ultimately in compound 3. This molecule was completely selective for PAK1 when tested at 10 µM concentration against a panel of 442 kinases, as well as against a panel of 22 other potential off-targets. Unexpectedly, it was even >50-fold selective against the closely related PAK2. It also displayed good permeability and acceptable solubility, though the stability could be improved.

Compound 3 blocked PAK1 autophosphorylation in cells, but was only modestly effective at blocking proliferation of a pancreatic cancer cell line with high levels of PAK1. As it turns out, all known cancer cell lines that are dependent on PAK1 also seem to use PAK2, so perhaps this chemical probe is too selective. Nevertheless, it will be useful to help disentangle the overlapping roles of these two kinases.

More broadly, this is a nice illustration of the selectivity achievable with allosteric kinase inhibitors. Indeed, as we recently noted in a comment to a post earlier this year, Novartis has put at least one allosteric kinase inhibitor into the clinic.