Fragments vs PKCθ, selectively

There are more than 500 protein kinases in the human genome, many of which have been tackled with fragments – sometimes all the way to the clinic. Within the universe of kinases, the dozen different isoforms of protein kinase C (PKC) provide an interesting challenge. For example, PKCθ is important in T cell signaling and thus has potential for treating autoimmune and inflammatory disease, but one needs to steer clear of isoforms important for heart function, such as PKCα. In two recent papers in J. Med. Chem., Dawn George and collaborators at AbbVie, WuXi, and Inventiva describe their efforts towards this goal.

The first paper starts as many fragment stories do: a high-throughput screen had come up largely empty. The AbbVie (née Abbott) team had previously constructed a collection of fragment-sized kinase hinge binders, and after screening ~250 of these at 300 µM in a fluorescence (TR-FRET) assay they selected compound 1 because of its structural novelty and ligand efficiency.

Modeling suggested multiple possible binding modes. Crystallography for PKCθ is difficult, but the researchers were able to obtain a structure of the compound bound to a different kinase, FAK. This suggested introducing a positively charged moiety to target an aspartic acid residue in PKCθ, leading to the more potent compound 15a. Additional optimization led ultimately to compound 41, which had moderate potency in cell-based assays, good pharmacokinetics, and 74-fold selectivity against PKCα. The compound also showed activity in a mouse arthritis model, but only at high doses, and was toxic at a slightly higher dose.

That’s where the second paper picks up. The researchers thought that by improving the pharmacokinetics they could lower the dose of the compound required, thereby reducing the potential for off-target effects. By now they had been able to obtain co-crystal structures of some of the more potent compounds bound to PKCθ, which confirmed the proposed binding mode and also gave additional ideas as to how to proceed. Extensive medicinal chemistry ultimately led to compound 17l, with low nanomolar biochemical and cell-based activity as well as good pharmacokinetics.

Unfortunately, this compound did not give stellar efficacy results in the mouse arthritis model. Also, this compound and several others appeared to be toxic to mice, with effects ranging from lethargy to seizures to death. The compounds were screened against panels of kinases and other receptors to try to find the source of these effects, but all to no avail; the compounds were fairly selective. This one-two punch of limited efficacy and unpredictable toxicity led to the termination of the PKCθ program.

These two papers reveal yet again that fragment-based lead discovery is often just the beginning of an arduous medicinal chemistry journey that can lead a long way from Valinor. The final destination here proved to be a dead end, but at least a useful one: it shows that PKCθ is certainly not a straightforward target for arthritis, or perhaps any indication. Kudos to the researchers for publishing this story so other scientists will not have to take the same journey. And at the very least, this compound is a useful probe for dissecting the biology of PKCθ.