Most kinase inhibitors that have been tested in clinical trials, such as the successful anticancer drug imatinib (Gleevec), act by directly competing with ATP at the ATP-binding site of the kinase. Reporting in Nature Chemical Biology, Adrian and colleagues now describe a new class of compounds that potently inhibit the same kinase as imatinib — BCR–ABL — but through a novel allosteric, non-ATP-competitive mechanism. Furthermore, these compounds maintain potency against some clinically relevant imatinib-resistant BCR–ABL mutants.

Typically, to discover new kinase inhibitors, screens are performed against recombinant catalytic kinase domains. However, this approach tends to identify ATP-competitive inhibitors that are often members of well-explored compound classes. So, to address this issue, and search for compounds with new mechanisms of inhibition, the authors developed an assay in which the cytotoxicity of compounds against cells specifically transformed with BCR–ABL was compared with their isogenetic parental cell line.

From an initial screen of 50,000 compounds, those that showed good differential cytotoxicity were investigated further, and compounds that were well characterized as targeting the ATP-binding site were discarded. Of the remaining compounds, one class in particular showed excellent differential cytotoxicity and became the focus of further optimization efforts. These led to compounds such as GNF2, which showed comparable potency to imatinib in inhibiting BCR–ABL-expressing cells but, remarkably and in contrast to imatinib, did not inhibit BCR–ABL kinase activity in vitro.

This high cellular selectivity — coupled with the observations that GNF2 did not inhibit any of 63 kinases including BCR–ABL in biochemical assays and binding experiments that demonstrated that GNF2 specifically bound to ABL — suggested that GNF2 inhibited cellular BCR–ABL kinase activity through an allosteric non-ATP-competitive mechanism. On the basis of knowledge about the mechanisms regulating BCR–ABL and their experimental data, the authors proposed that GNF2 stabilizes the inactive conformation of BCR–ABL by mimicking the binding of the myristoylated amino terminus of normal ABL to a binding pocket located near the carboxyl terminus of the kinase. BCR–ABL is not myristoylated and so lacks this auto-inhibitory mechanism. In support of this proposal, mutations to the myristate binding pocket were shown to render BCR–ABL resistant to GNF2, but not to imatinib.

Binding to this pocket distant from the active site of the kinase is consistent with the excellent cellular and enzymatic selectivity of compounds such as GNF2. In addition, GNF2 and imatinib showed synergistic antiproliferative effects when combined, further supporting the idea that they bind to different sites on BCR–ABL.

Compounds such as GNF2 showed good pharmacokinetic behaviour, and so could represent a promising starting point for the development of new drugs with diminished off-target activity. The demonstration that the cellular activity of BCR–ABL can be pharmacologically modulated by non-ATP-competitive inhibitors in a highly selective fashion also suggests that such inhibitors could be found for other kinases of therapeutic interest for which selectivity is a key concern.