Dr Jennifer Miles

Talk: Using motif scaffolding to generate selective designed binders

Aurora-A kinase is a Ser/Thr kinase that has roles in the regulation of mitosis, ciliogenesis, and DNA repair, and is highly expressed in cancer tissues. Aurora-A is a key target for inhibitor design, but developing selective kinase inhibitors can be challenging due to the presence of over 470 kinases in the human genome. Given the new ability to design de novo proteins, we wondered whether we could design selective, high-affinity Aurora-A binders. 

We began by using the structure of N-Myc bound to Aurora-A as an initial model, using motif scaffolding in RFdiffusion to scaffold part of the N-Myc sequence. After generating over 3,800 designs, 2-3% were deemed as hits. These binders were all predicted to form alpha-helical bundles. From these, 17 were expressed, purified, and tested; those that compete with a fluorescent N-Myc peptide for binding were taken forward. 

We solved the structure of 9 of these designed binders in complex with the Aurora-A kinase domain and see that the binders all interact at the predicted site. Using ITC, we determined that several of the binders have low nanomolar Kd values for Aurora-A. This is a 1000x improvement in binding when compared to the original N-Myc sequence. 

The binders were also tested in an ADP-Glo-based assay, and all inhibit Aurora-A kinase activity to some extent, with the lowest measured IC50 of 12nM. The binders were also tested against the closely related kinase Aurora-B (71% sequence homology), and we observed very little inhibition, suggesting the binders are selective for Aurora-A. 

Most recently, we have been able to pull out Aurora-A from HeLa cells using these binders. We are no longer testing the effect of binders on mitosis and are using mass spec to determine what else they might interact with in vivo.