Activity‐Directed Synthesis of Inhibitors of the p53/hDM2 Protein–Protein Interaction

Abstract Protein–protein interactions (PPIs) provide a rich source of potential targets for drug discovery and biomedical science research. However, the identification of structural‐diverse starting points for discovery of PPI inhibitors remains a significant challenge. Activity‐directed synthesis (ADS), a function‐driven discovery approach, was harnessed in the discovery of the p53/hDM2 PPI. Over two rounds of ADS, 346 microscale reactions were performed, with prioritisation on the basis of the activity of the resulting product mixtures. Four distinct and novel series of PPI inhibitors were discovered that, through biophysical characterisation, were shown to have promising ligand efficiencies. It was thus shown that ADS can facilitate ligand discovery for a target that does not have a defined small‐molecule binding site, and can provide distinctive starting points for the discovery of PPI inhibitors.


General Experimental
Solvents were removed under reduced pressure using a Büchi rotary evaporator with a Vacuubrand PC2001 Vario diaphragm pump, or under N2 blowdown at 40 o C. Dry solvents and reagents were purchased from commercial suppliers and used without further purification. Rhodium catalysts were purchased from Sigma-Aldrich and used as supplied. Flash column chromatography was carried out using silica gel 60 (35-70 μm particles) supplied by Merck. Thin-layer chromatography was conducted with Macherey-Nagel Polygram SIL G/UV254 0.2mm silica gel 60 with fluorescent indicator plates.
Analytical LC-MS was performed using a system comprising an Ultimate3000 HPLC instrument with a Brucker Amazon Speed MS detector with electrospray ionisation. The system ran with a positive and negative switching mode and UV diode array detector using a Phenomenex Kinetex C18 (50 mm × 2.1 mm × 2.6 μm) column and gradient elution with two binary solvent systems: MeCN/H2O or MeCN/H2O plus 0.1% formic acid. Accurate mass spectrometry was performed using electrospray ionisation on a Bruker MaXis Impact spectrometer. Chemical shifts are quoted in parts per million (ppm) and coupling constants are given in Hz. Splitting patterns have been abbreviated as follows: s (singlet), d (doublet), dd (doublet of doublets), t (triplet), q (quartet) and m (multiplet). NMR data is reported in the format: ppm (number of protons, splitting pattern, coupling constant). Infrared spectra were recorded on a Bruker Alpha ATR FR-IR spectrometer; absorptions are reported in wavenumber (cm -1 ).
The human homologue of MDM2 is referred to as hDM2 throughout. The p53/hDM2 fluorescence anisotropy assay was assembled and performed as described by Wilson et. al. 1 hDM2 protein (residues 17 to 125, with L33E mutation; referred to as hDM217-125) and p5315-31-fluorescein (Ac-SQETFSDLWKLLPENNVC(Flu)-NH2) peptide tracer, in which the fluorophore was linked to the Cterminal cysteine thiol through a maleimide, were used for all biological screening.
Total product concentration was used to standardise the effective screening concentrations for the high-throughput screening of reaction mixtures and is defined as the concentration of the limiting reactant in each well before the reaction took place (here, the diazo reactant).

Synthesis of diazo compounds
CAUTION: All diazo compounds (excluding those isolated as solid material) described below appear to be volatile at room temperature under reduced pressure. Gradual loss of mass was observed when left under high vacuum. Diazo compounds are potentially explosive on contact and should be treated with caution, although no adverse advents occurred during this study. Compounds D1, D4, D5, D7, D8 and 2-diazo-1-(pyrrolidin-1-yl)ethenone were synthesised as previously reported. 2 Enantiomerically pure compounds (S)-D4 ([ ] 20 = 35) and (R)-D4 ([ ] 20 = +28) were prepared by the same procedure as racemic D4. Compound D2 was prepared as previously reported. 3 Compound D8 was prepared as previously reported. 4

General procedure for the scale-up of ADS hits, A
A crimp vial (10 or 20 mL) was sequentially charged with solutions of Rhodium(II) catalyst in DCM (240 µL, 12.5 mM) and co-substrate in DCM (240 µL, 6.25 M) and stirred. A solution of diazo in DCM (240 µL, 1.25 M) was added and the vial capped. After 24 hours 900 mg of Quadrapure TU resin was added, followed by a further 720 µL DCM. After a further 24 hours the resin was removed by filtration and the solvent evaporated under reduced pressure to yield the crude reaction product.

General procedure for the scale-up of ADS hits/analogues using a syringe pump, B
A 20 mL vial was charged with Rhodium(II) catalyst (1 mol%) and degassed under N2 atmosphere, followed by the addition of co-substrate (6.25 M) in DCM. Diazo (1.25 M) in DCM was then added dropwise to the stirred solution over 6 hours using a syringe pump. After 24 hours Quadrapure TU TM resin was then added and the reaction left for a further 24 hours. The resin was then removed by filtration and the solvent removed under reduced pressure to give a crude product.

Implementation of high-throughput chemistry for Activity-Directed Synthesis Reaction Arrays
Activity-Directed Synthesis reactions were carried out in 0.75 mL shell vials (Chemglass CV-2100-0830) equipped with a teflon-coated stir bar (Biotage 0.2-0.5 mL magnetic stir bar #355545) and sealed using either a Freeslate 96-well reaction block or a Sigma-Aldrich Kitalysis 24-well reaction block (Z742107 Aldrich). Prior to the assembly of each reaction array the following stock solutions were made: diazo reaction solvent (1.25 M); catalyst in THF (25 mM); and co-substrate in DCM (6.25 M). Each reaction vial was charged with catalyst stock (8 µL) and the solvent allowed to evaporate to dryness, then DCM (84 µL) was added and the reaction block placed on a magnetic stirring plate.
Each reaction vial was then sequentially charged with co-substrate stock (8 µL) and diazo stock (8 µL), then the plate sealed using a Teflon film and stirred. After 24 hours Quadrapure TU resin (30 mg) was added to each vial and left overnight to scavenge the catalyst. The solvent was then evaporated under a stream of nitrogen gas and the crude material dissolved in molecular biology grade DMSO (200 µL) to create a biological screening master stock (50 mM total product concentration) that was passed through a 96-well filter plate (Agilent Technologies: #200933-100) and stored at 20 o C.
An overnight starter culture (10 mL The quality and purity of the preparation was assessed by mass spec and circular dichroism spectroscopy. The activity of the protein was verified by testing its binding to fluorescently labelled p53 peptide in a fluorescence anisotropy assay ( Figure S1). For 15 N labeled protein the expression was carried out using the same method but using M9 minimal media supplemented with 15 NH4Cl as nitrogen source.
Results were collected using a Perkin-Elmer Envision 2103 Multilabel Reader using a 431 nm mirror, 480(104) nm excitation filter, and 535(208) and 535(209) nm emission filters after 2.5 or 24 hours of incubation at room temperature. Test well anisotropy values were then calculated using the blank corrected S and P channel values using the following formula: The fraction of bound tracer was calculated using the following formula: Where λ is the intensity of bound/unbound tracer (λ = Ibound/Iunbound) and r is anisotropy. Where y is the fraction bound of p5315-31 Flu multiplied by 54.5 (p5315-31 Flu concentration, nM), [FL] is p5315-31 Flu, and x is the concentration of hDM2. The observed Kd for the p5315-31 Flu:hDM2 binding was 180 ± 30 nM.

Binding of p5315-31 Flu to hDM2
A serial dilution of hDM2 (0.0006 µM to 20.75 µM, final concentration) was added to a fixed concentration of p5315-31 Flu (54.5 nM) and PBSA buffer (20 µL), to give a 60 µL total volume per assay well. Each dilution was performed in triplicate and the measured intensity of each well calculated using equation 1, then anisotropy was calculated using equation 2. The fraction bound of the tracer could also be determined using equation 3.
EC50 values were determined and curves were fit in Origin Pro 2019b using a non-linear curve fitting with the dose response fitting procedure (equation 5) and Levenberg Marquardt iteration algorithm.
Eq. 5: Where p is the Hill slope and EC50 is the concentration for half-maximal response from the baseline. Round 1 HTS at 20 µM total product concentration:

Reaction array 1 controls for individual reagents:
Round 2 HTS at 20 µM: Reaction array 2 controls for individual reagents:

Determining IC50 values for isolated compounds
Pure compounds (P1 -P7 and SI1 -SI4) were serially diluted in 100% DMSO (using 12 two-fold dilution steps) to achieve the correct effective concentrations ( IC50 values were determined and curves were fit in Origin Pro 2019b using a non-linear curve fitting with the dose response fitting procedure and Levenberg Marquardt iteration algorithm.

P5
Due to poor compound solubility a full dose-response curve could not be obtained.

SI3
Due to poor compound solubility a full dose-response curve could not be obtained.

P7
Due to poor compound solubility a full dose-response curve could not be obtained. Curve fitting failed and EC50 could not be determined.

LC-MS Analysis of Reaction Mixtures
All 154 reactions from the round 1 reaction array were analysed by LC-MS to investigate how many combinations had produced a desired product ( Figure S4). All samples were diluted to 1 mg/mL concentrations from the original 50 mM DMSO master stock, with respect to the initial diazo starting concentration. Reaction wells containing diazo and substrate were analysed for intermolecular product(s) and blank control wells were analysed for intramolecular product(s). Dark green squares indicate clear m/z for the desired product(s) and a clear corresponding UV peak(s). Light green squares indicate m/z for the desired product(s) and either weak or no corresponding UV peak(s). Blank squares indicate that no m/z was observed for the desired product(s).   Figure S6. Overlay of docked poses of MDM2 binders and Nutlin-3a (PDB ID: 4HG7)

Similarity analysis
1769 compounds with annotated activity towards hDM2 (referred to as MDM2 within ChEMBL) were obtained from the ChEMBL database (accessed: 16/01/2020). Subsequent processing removed duplicate molecules leaving 1314 compounds (see accompanying Excel spreadsheet) which were then used for further analysis. The Morgan molecular fingerprint was then computed for each molecule using RDKit 10 and the pairwise Tanimoto similarity scores calculated. Table S3. Tanimoto similarity analysis comparing ADS products P1-P6 to 1314 known hDM2 ligands from the ChEMBL database.