Design, Synthesis, and Biological Evaluation of an Allosteric Inhibitor of HSET that Targets Cancer Cells with Supernumerary Centrosomes

Summary Centrosomes associate with spindle poles; thus, the presence of two centrosomes promotes bipolar spindle assembly in normal cells. Cancer cells often contain supernumerary centrosomes, and to avoid multipolar mitosis and cell death, these are clustered into two poles by the microtubule motor protein HSET. We report the discovery of an allosteric inhibitor of HSET, CW069, which we designed using a methodology on an interface of chemistry and biology. Using this approach, we explored millions of compounds in silico and utilized convergent syntheses. Only compound CW069 showed marked activity against HSET in vitro. The inhibitor induced multipolar mitoses only in cells containing supernumerary centrosomes. CW069 therefore constitutes a valuable tool for probing HSET function and, by reducing the growth of cells containing supernumerary centrosomes, paves the way for new cancer therapeutics.


SUPPLEMENTARY EXPERIMENTAL PROCEDURES
Computational Model Building. In order to generate potential binding mode hypotheses of CW069 a two--step ligand--based alignment approach has been used.
Compound 3 was first aligned onto a set of known KSP inhibitors and consecutively CW069 onto the resulting superimposed conformation of 3.
The initial conformations of CW069 and 3 were created using CORINA (Sadowski and Gasteiger, 1993). Based on fingerprint similarity, 7 KSP compounds (ligands binding to the allosteric pockets of PDB 2FL2, 2FL6, 2PG2, 2Q2Y, 2Q2Z, 2UYM and 3CJO) were selected and their corresponding binding sites were superimposed using Relibase (Hendlich et al., 2003). The resulting superimposition of the seven ligands was used as the target to overlay compound 3 onto. As compound 3 contains several rotatable bonds, a flexible ligand superimposition method was used to generate alignment hypotheses (Korb et al., 2010). The HSET protein structure (PDB 2REP) was  (Eswar et al., 2006).

Optimization of Protein--CW069 Complexes.
Optimization of each Mg 2+ .ADP-complexed protein in the presence of compound CW069 was carried out using a multi--step protocol. Initially, the conformation of CW069 was constrained, and several thousand steps of adopted basis Newton--Raphson (ABNR) minimization were performed until protein--ligand interaction energies reached convergence.
Subsequently, the constraints on the ligand were released, and several further thousand ABNR steps were carried out. Following convergence, and a final, short steepest descent minimization step, the protein--ligand interaction energy and its component contributions were recorded. Each protein--ligand complex was optimized in the CHARMM program (Brooks et al., 2009), using the CHARMM27/CMAP protein parameters, and the CHARMM CGenFF general force field for compound CW069 (Vanommeslaeghe et al., 2010). Non--bonded interactions were treated with an atom--based cutoff of 1.4 nm. Electrostatics used a distance-dependent dielectric, with a shifted potential going to zero beyond 1.2 nm. Van--der--Waals Forces were switched off from 1.0 nm to 1.2 nm.

Molecular Dynamics Simulations of Inhibitor--Free Motor Domains. The ionizable
groups of each protein model were assigned their most probable charged states at neutral pH. Each protein in complex with Mg 2+ .ADP was solvated with TIP3P water and a ∼0.1 M concentration of NaCl in a truncated octahedral periodic box of dimension ∼9--10 nm. Overlapping solvent molecules were removed, resulting in systems containing ∼60,000--80,000 atoms. At each stage of system setup, steepest descent energy minimization was performed to relax the protein geometry and to remove steric clashes between protein, cofactor, solvent and ions. Each system was equilibrated over 2 ns, during which position restraints, applied to all non--hydrogen protein atoms except modeled loop regions, were gradually removed to relax the protein structure and solvent. A total of 1 µs of production simulation sampling was generated for the three protein systems, each split into ten sets of independent 100 ns trajectories to ensure efficient exploration of conformational space. All simulations were performed using GROMACS (Hess et al., 2008) version 4.5 (Bjelkmar et al., 2010). The protein was treated using the CHARMM22/CMAP force field (MacKerell et al., 1998). Equations of motion were integrated using the leapfrog method with a 2 fs time step, and the LINCS algorithm was used to constrain bond lengths (Hess et al., 1997). Electrostatic interactions were computed using the Particle--Mesh--Ewald (PME) algorithm (Essmann et al., 1995) and the real--space sum was cut off at 12 Å. Van--der--Waals Interactions were switched off between 10 Å and 12 Å and the neighbor list was updated every 10 steps. Simulations were performed using conditions of constant temperature (300 K) and pressure (1 atm) via the Bussi thermostat (Bussi et al., 2007), and isotropic pressure--coupling using the Parrinello--Rahman barostat (Parrinello and Rahman, 1981) with a coupling constant of 1 ps.
Clustering analysis was performed using GROMACS. Visual analysis was performed using VMD (Humphrey et al., 1996). was a three--step process in which the key peptide--bond formation ( Figure S1C) was achieved using the coupling agent propylphosphonic anhydride solution (T3P).
However, poor yields were routinely observed for this reaction (26%), possibly due to an undesired side--reaction of the carboxylic acid moiety. The deprotection and reductive amination steps progressed well however (78% and 81%, respectively).
The reductive amination proceeds through formation of an imine species followed by subsequent reduction by NaCNBH 3 , as described by Grenga (Grenga et al., 2009).
Because our original synthetic route proved inefficient, a variety of alternative reaction conditions were explored to afford the desired peptide product in good yield. A Buchwald--Hartwig C--N bond--formation was eventually chosen as the most effective reaction in this case (main text Figure 1C). Good conversion to the peptide product (90%) was only achieved using these specific reaction conditions. A screen identified Pd 2 (dba) 3 /Xantphos as the best catalyst--ligand combination, with conversion increasing in parallel with catalytic loading.

Colony Forming Unit -Granulocyte Macrophage (CFU -GM) Assay
IMDM media (StemCell Technologies, Grenoble, France) and MethoCult media (H4025 optimum without EPO, StemCell Technologies, Grenoble, France) were thawed overnight at 4 °C. The MethoCult media was then shaken vigorously and allowed to settle before use. The cells used for the assay were prepared as follows.
Primary adult human bone marrow mononuclear cells, isolated from the posterior iliac crest (catalogue number ABM007F, StemCell Technologies, Grenoble, France), were thawed at 37 °C and suspended in 13 mL IMEM media. The suspension was centrifuged at 300 g for 10 mins, supernatant was removed, and cells were resuspended. The cell number and viability were then counted. At the same time, compound treatments were prepared as follows. In order to dose in triplicate, 4.4 mL of each compound concentration was prepared in media, with final DMSO concentrations of 0.2%. 4 mL MethoCult media was then added to each falcon tube using a syringe and blunt--end needle. The tubes were then subjected to vortex in order to thoroughly mix the contents, and a 2 × 10 5 cell / mL suspension was prepared. 0.4 mL of the cell suspension were then transfered to each falcon tube containing the MethoCult media and compound, giving the 1:10 v/v ratio critical for optimal cell growth. Each tube was subjected to a vortex to mix contents thoroughly and allowed to stand for 5 mins, to allow any bubbles to rise to the top. Using a syringe and a blunt--end needle, 1.1 mL of the final cell suspension was transferred into each well of a 6--well plate. The sample plates were transferred into one 15 cm plate with a 35 mm plate filled with water in the center. The assay plate was stored at 37 °C for 14 days before being scored for colony formation. Western Blotting. Cells were harvested whilst growing in exponential phase by scraping, and lysed in RIPA lysis buffer (50 mM Tris HCl pH 8.0, 150 mM NaCl, 1% NP--