Allosteric inhibition of Aurora-A kinase by a synthetic vNAR domain

Abstract

The vast majority of clinically approved protein kinase inhibitors target the ATP-binding pocket directly. Consequently, many inhibitors have broad selectivity profiles and most have significant off-target effects. Allosteric inhibitors are generally more selective, but are difficult to identify because allosteric binding sites are often unknown or poorly characterized. Aurora-A is activated through binding of TPX2 to an allosteric site on the kinase catalytic domain, and this knowledge could be exploited to generate an inhibitor. Here, we generated an allosteric inhibitor of Aurora-A kinase based on a synthetic, vNAR single domain scaffold, vNAR-D01. Biochemical studies and a crystal structure of the Aurora-A/vNAR-D01 complex show that the vNAR domain overlaps with the TPX2 binding site. In contrast with the binding of TPX2, which stabilizes an active conformation of the kinase, binding of the vNAR domain stabilizes an inactive conformation, in which the αC-helix is distorted, the canonical Lys-Glu salt bridge is broken and the regulatory (R-) spine is disrupted by an additional hydrophobic side chain from the activation loop. These studies illustrate how single domain antibodies can be used to characterize the regulatory mechanisms of kinases and provide a rational basis for structure-guided design of allosteric Aurora-A kinase inhibitors.

Keywords: antibody-assisted drug discovery; biochemistry; protein kinase; structural biology.

Figure 1.

vNAR-D01 is an Aurora-A inhibitor that competes with TPX2. (a) Surface plasmon resonance binding assays between Aurora-A KD CA-Avi and vNAR-D01. The kinase was immobilized on Biacore Sensor SA chips at 550, 350 and 250 RU and interacted with 0.01–50 µM vNAR-D01. Maximum responses were plotted against vNAR-D01 concentration and fitted to a one-site specific binding equation (solid lines) in Prism6 (GraphPad) to calculate binding affinities. (b) Co-precipitation assay between the Aurora-A KD CA/vNAR-D01 complex or His₆-Aurora-A KD CA and GST-TPX2^1–43. The complex and Aurora-A were immobilized on Nickel Sepharose beads using the His₆-tag on the vNAR domain and kinase, respectively. GST was used as a binding control. (c) Co-precipitation assay between GST-Aurora-A KD DN and vNAR-D01 and His₆-TPX2^1–43. In total, 2 µM GST-Aurora-A KD DN was immobilized on Glutathione Sepharose 4B beads and incubated with 5 µM vNAR-D01 and 0, 1, 2, 5, 10, 20 and 50 µM His₆-TPX2 (black triangle). GST was used as a binding control. (d) In vitro kinase activity assay of Aurora-A KD in the presence of vNAR-D01. MBP was used as a generic kinase substrate. Reactions were analysed by SDS-PAGE (top left panel) and incorporation of radioisotope resolved by autoradiography (bottom left panel). Incorporation of radioisotope was measured by scintillation counting (right). Error bars represent the standard error for two independent reactions. ** = p < 0.01, *** = p < 0.001 and **** = p < 0.0001 using one-way ANOVA with Dunnett's post hoc test compared with the kinase only reaction. (e) In vitro kinase activity curves of Aurora-A KD in the presence of WT and mutant vNAR-D01 proteins. The kinase activity of Aurora-A KD was measured by the incorporation of radioisotope into the generic kinase substrate, MBP by scintillation counting in the presence of 0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50 and 100 µM vNAR-D01. Data were normalized to % kinase activity using the Aurora-A KD only reaction as 100% and plotted against vNAR-D01 concentration (right). Data were fitted to a log(inhibitor) versus response—variable slope in Prism6 (GraphPad) to calculate IC50s (right, solid line). n.s. = no significant inhibition observed.

Figure 2.

Crystal structure of Aurora-A/vNAR-D01 complex. (a) Cartoon representation of the complex structure (crystal form 1). Aurora-A is coloured teal and vNAR-D01 is coloured orange. (b) Structures of CDK2/Cyclin-A (PDB 1FIN) and Aurora-A/TPX2 (PDB 1OL5) complexes. The αC-helix is marked with a black rectangle in panels (a) and (b). (c) The interacting regions of the Aurora-A and vNAR-D01 are shown as contrasting colours on the individual proteins (yellow and green, respectively).

Figure 3.

Details of the molecular recognition in the Aurora-A/vNAR-D01 complex. (a) Key interactions are shown in the three panels. Aurora-A is coloured teal and vNAR-D01 is coloured orange. (b) Co-precipitation assay between GST-Aurora-A KD DN and WT, and mutant vNAR-D01 constructs. GST-Aurora-A KD DN was immobilized on Glutathione Sepharose 4B beads and then incubated with vNAR-D01 proteins. GST was used as a binding control.

Figure 4.

The mechanism of allosteric inhibition of Aurora-A by vNAR-D01 is antagonistic to TPX2 activation. (a) Aurora-A/vNAR-D01 complex viewed along the αC-helix. (b) Aurora-A/TPX2 complex, equivalent view to that shown in (a). (c) Superposed structures of Aurora-A/vNAR-D01 (teal/orange) and Aurora-A/TPX2 (blue/red) complexes viewed with the αC-helix running from left to right. Note that the binding site of the CDR3 loop of vNAR-D01 on Aurora-A overlaps with the binding site of TPX2 residues Tyr8 and Tyr10. (d) Schematic of the structural basis by which vNAR-D01 and TPX2 stabilize distinct conformations of Aurora-A through binding at the same site. Key residues are shown as single-letter notation. Canonical R-spine residues are shown as green hexagons and the additional residue that joins the R-spine in inactive Aurora-A is shown as a light blue hexagon.

Figure 5.

Aurora-A in complex with vNAR-D01 adopts a DFG-up conformation. (a) Superposed structures of Aurora-A in complex with vNAR-D01/ADP (teal), TPX2/ADP (dark blue; PDB 1OL5) and MLN8054 (magenta; PDB 2WTV). (b) Magnified view of Phe275 and Glu181. (c) Schematic to show how the DFG-up conformation disrupts the Lys-Glu salt bridge. The activation loop is shown as a grey line, with the position of Phe275 marked with a hexagon labelled ‘F’. In the active conformation of Aurora-A (dark blue, top image), a salt bridge is formed between Lys162 (marked with a triangle labelled ‘K’) and Glu181 (shown as a Y-shaped appendage on the αC-helix). Distortion of the αC-helix by vNAR-D01 (orange, central image) breaks the Lys-Glu salt bridge and creates a hydrophobic pocket for Phe275. A similar configuration of the αC-helix and Phe275 is observed in the structure of Aurora-A bound to MLN8054 (magenta, lower image), which induces a rearrangement of the activation loop. (d) Superposed structures of Aurora-A in complex with vNAR-D01/ADP (teal) and adenosine (lilac).