Identification of small molecule inhibitors of the Aurora-A/TPX2 complex

Abstract

Aurora kinases are a family of cell division regulators that govern the correct assembly of a bipolar mitotic spindle and the fidelity of chromosome segregation. Their overexpression is associated with genomic instability and aneuploidy, and is frequently observed in cancer. Accordingly, competitive inhibitors targeting Aurora kinase activity at the ATP-binding site are being investigated for therapeutic purposes. Despite promising pre-clinical data, these molecules display moderate effects in clinical trials and incomplete selectivity, either against distinct family members, or other kinases. As an alternative approach, protein-protein interaction inhibitors targeting mitotic kinases and their activators can be exploited to achieve increased specificity of action. In this study, a virtual screening of small molecules led to the identification of 25 potential inhibitors of the interaction between Aurora-A and its activator TPX2. In vitro experiments confirmed that 4 hits bind Aurora-A in the low micromolar range and compete for TPX2 binding. Immunofluorescence assays showed that 2 compounds also yield lowered Aurora-A activity and spindle pole defects in cultured osteosarcoma cells. The identified protein-protein interaction inhibitors of the Aurora-A/TPX2 complex might represent lead compounds for further development towards pioneering anti-cancer drugs and provide the proof-of-concept for a new exploitable strategy to target mitotic kinases.

Keywords: Aurora-A kinase; TPX2; anti-cancer therapy; protein-protein interactions; small molecule inhibitors.

Figure 1. Analysis of the Aurora-A/TPX2 interaction interface and hot spots identification

Residues 7-11 of human TPX2 (sticks) bind at a shallow hydrophobic groove at the N-terminal lobe of Aurora-A (grey surface). Evolutionary conservation (ConsScore) and ΔΔG upon computational Alanine mutagenesis are reported. Among these residues, Tyr 8 (ΔΔG = 3.24 Kj/Mol) and Tyr 10 (ΔΔG = 3.42 Kj/Mol) were predicted as key residues for the thermodynamic stabilization of the complex.

Figure 2. Pharmacophore hypothesis of the Aurora-A/TPX2-7-11 interaction

Residues 7-11 of human TPX2 (balls-and-sticks) were identified as a hot spot of interaction and used to derive a PH (arrows for projections and centroids) at a shallow hydrophobic groove at the N-terminal lobe of Aurora-A (surface, colored by hydrophobicity). Green arrow, hydrogen-donor feature; red arrow, hydrogen-acceptor feature; blue arrow/centroid, hydrophobic/aromatic interaction. Exclusion volumes are not shown. This figure was rendered with LigandScout (Inte:Ligand).

Figure 3. A subset of compounds directly binds Aurora-A in vitro

Interaction of compounds with Aurora-A immobilized on a COOH5 sensorchip was measured by SPR experiments. Sensorgrams were obtained using Aurora-A as ligand and compounds (concentrations: 0-25 sec: 0.94 μM; 25-50 sec: 1.875 μM; 50-75 sec: 3.75 μM; 75-100 sec: 7.5 μM; 100-125 sec: 15 μM; 125-150 sec: 30 μM) as analytes. Panel A shows representative sensorgrams, for C02, C20, C22, C23 and C25; the increase in RU relative to baseline indicates complex formation upon injection of each compound, the decrease in RU represents dissociation of analytes from immobilized Aurora-A after injection of running buffer. Panel B shows the RU values upon injection of analytes at the indicated concentrations. Affinities, calculated from Scatchard plots from each compound, are reported in Table 1. Alisertib was used as positive control. Each sensorgram is the average of two experiments.

Figure 4. Competition experiments between GST-TPX2-1-43 and selected compounds for Aurora-A binding

(A) Sensorgrams of competition experiments carried out on COOH5 chips with Aurora-A immobilized at a level of about 200 RUs, by injecting 1 μM GST-TPX2-1-43 together with increasing concentrations of compounds. (B) Fitting of competition experiments data. The decrease of TPX2 binding upon injection of increasing concentrations of competing compounds is shown. IC50 values of tested compounds are shown in the table on the right; standard errors of best-fit parameters are indicated.

Figure 5. Predicted binding mode of compounds to Aurora-A

The hydrophobic groove at the N-terminal lobe of Aurora-A is shown as grey surface. The position of TPX2-7-11 is shown in transparent sticks. Atom coloring is the following: N blue, O red, S orange. C02 (A), C23 (B), C25 (C) and C20 (D).

Figure 6. Effects of treatment with the C20 and C23 inhibitors on mitotic spindle structure and Aurora-A activity in U2OS cells

(A) Top: schematic representation of the experimental protocol. The immunofluorescence panels (below) show a control spindle (left panel) and representative images of the observed defects in C20- or C23-treated cultures. Quantification is shown in the histograms on the right. About 300 prometa-metaphase (PM/M) cells per condition from 4-7 experiments were analyzed. (B) Representative immunofluorescence images of active p-Thr288-Aurora-A in control (DMSO) and C20- or C23-treated prometaphases. The distribution of signal intensity is quantified in the box plots on the right (n, at least 40 measured cells per condition, from 2 experiments). Fluorescence intensity is shown in arbitrary units (a.u.). *: p<0.05; **: p<0.01; *** p< 0.001; Student's t-test. Scale bars: 10 μm.