Discovery of selective activators of PRC2 mutant EED-I363M

Many common disease-causing mutations result in loss-of-function (LOF) of the proteins in which they occur. LOF mutations have proven recalcitrant to pharmacologic intervention, presenting a challenge for the development of targeted therapeutics. Polycomb repressive complex 2 (PRC2), which contains core subunits (EZH2, EED, and SUZ12), regulates gene activity by trimethylation of histone 3 lysine 27. The dysregulation of PRC2 catalytic activity by mutations has been implicated in cancer and other diseases. Among the mutations that cause PRC2 malfunction, an I363M LOF mutation of EED has been identified in myeloid disorders, where it prevents allosteric activation of EZH2 catalysis. We describe structure-based design and computational simulations of ligands created to ameliorate this LOF. Notably, these compounds selectively stimulate the catalytic activity of PRC2-EED-I363M over wildtype-PRC2. Overall, this work demonstrates the feasibility of developing targeted therapeutics for PRC2-EED-I363M that act as allosteric agonists, potentially correcting this LOF mutant phenotype.


S2
The following data were calculated from the molecular dynamics simulations: RMSDs: • EZH2 SRM helix (residues 143-153) upon SRM Cα alignment (Fig. S2C, Fig. S9) • EZH2 SRM helix upon EED Cα alignment (Fig. S1, Fig. S2D, Fig. S9); • EZH2 loop, residues 136-142, upon EED Cα alignment (Fig. S10); • Ligand RMSD (Fig. S14); Minimum distances (designated as Distance on figures): • Jarid2 R115 N + -EZH2 D136 O - (Fig. S2A); • Jarid2 R115 N + -EZH2 D140 O - (Fig. S2B); • Ligand's R mimic N + -EZH2 D140 O - (Fig. S11); • Ligand's R mimic N + -EED D362 O - (Fig. S12, Fig. S13); • Ligand's N-terminus N + -EED D362 O - (Fig. S15); • Ligand's N-terminus N + -EED W364 benzene ring (Fig. S16); The legends of the figures represent median values of the parameter (and median absolute deviation in the brackets) for each of the five independent simulation runs of each simulated system (PRC2 WT (WT) with Jarid2, PRC2 I363M (I363M) mutant with Jarid2, WT-3 (UNC6083), I363M-3, WT-2 (UNC5636), I363M-2). For clearer representation of the time series lines the Savitzky-Golay filter with window size 171 and polynomial order of 3 was used to remove high frequency noise from the data, and the full data is also plotted using transparent lines. Figure S1. (A) Previously reported EED allosteric inhibitors: UNC4859, UNC5114, and UNC5115 (1). (B) UNC5115 competes with Jarid2 in binding to EED leading to PRC2 inhibition (1). By lacking an arginine mimic, the ligand is unable to capture D140 which leads to SRM helix unfolding shown in our simulation studies (higher values of RMSD with respect to the crystal structure of Jarid2 bound active complex, PDB ID 5HYN (2)). The molecule possessing ˜1 µM EED binding affinity served as a starting structure for PRC2 activator design.   Fig. S12). Octahydroindole of 3 (UNC6083; inhibitor) prefers to lay 'flat' over the surface of EED protein and orient its arginine mimic towards and form a salt bridge with D362 of EED. Among the four simulated systems R mimic of compound 3 in wild type has the lowest tendency to form a salt bridge with D140 of EZH2 (SI Fig. S11), and the highest tendency to form a salt bridge with D362 of EED, which explains the experimental data indicating the highest degree of inhibition of the catalytic activity of the complex in WT-UNC6083 system. Since the simulations started from a hypothetical pose of the ligand in which its arginine mimic forms a salt bridge with EZH2 D140 and we used all collected data to plot these distributions, the distributions are skewed towards the higher values of the distance. However, the zoomed-in plot (insert) clearly indicates that the states possessing this salt bridge have started to be populated and the ranking order based on this distance distribution clearly explains the experimental findings of the inhibition of the complex. Additionally, this data explains the lowest RMSD of the ligand (SI Fig. S14) in WT-UNC6083 system-the formation of this salt bridge 'locks' the molecule on the surface of EED, which could also indicate a higher affinity of the ligand to wild type EED.

Analysis of products
Analytical LCMS (at 220 nm and 254 nm) was used to establish the purity of targeted compounds. All compounds that were evaluated in biochemical and biophysical assays had

FAM-UNC5636)
The alkynyl peptoid was synthesized on Fmoc Rink amide MBHA resin (500 mg, 0.43 mmol/g loading, Anaspec). The alkynyl peptoid was installed as the first residue by coupling on bromoacetic acid (1 M) with N,N-diisopropylcarbodiimide (1 M) for 30 minutes followed by 7 DMF washes. Next, the displacement reaction was conducted with propargylamine (1 M) in DMF for 2 hours followed by 7 DMF washes. Following installation of the peptoid residue, standard Fmoc synthesis (as described in the solid-phase peptide synthesis section above) was used for the remaining residues. The alkynyl peptoid was then used to generate the labeled peptide through click chemistry with azido-fluorescein (1 eq, Tenova Pharmaceuticals), using copper sulfate (10 eq) and ascorbic acid (10 eq) dissolved in 1 mL DMF. The mixture was first heated to 70°C while stirring for 10 mins, after which the reaction was allowed to cool to room temperature and stirred overnight prior to concentration under vacuum. The crude mixture were re-dissolved in 1:1 water:acetonitrile (1 mL), filtered, and purified via preparative HPLC. Formaldehyde (37 % w/v solution in water; 4.7 mL, 8 eq, 64 mmol) was added to the cloudy S17 solution while stirring at room temperature. After 5 min, acetic acid (2 mL, 5 eq, 40 mmol) and sodium cyanoborohydride (0.6 g, 2 eq, 16 mmol) were added to the reaction and the mixture was left to stir at room temperature for 2 hours. The reaction was concentrated in vacuo and redissolved in 1:1 water:acetonitrile (10 mL). The crude material was purified via reverse phase flash chromatography (H2O + 0.1% trifluoroacetic acid and acetonitrile) to yield product. The product was lyophilized to a hygroscopic, white powder and yielded 2.52 g (80%) of product.

S22
Supplementary Movies Movie S1. Interactions of compound 2 with WT PRC2. In WT PRC2, compound 2 is able to form interactions with the EZH2 loop, however, these contacts are significantly less frequent than in the mutant complex, thereby preventing effective stabilization of the SRM helix, which has been previously shown to be a key component of the EZH2 catalytically active state (see the main manuscript for details). The color scheme is the same as that in Figure 4.

Movie S2.
Interactions of compound 2 with PRC2 EED-I363M mutant. The most distinctive feature of the dynamic behavior of compound 2 in PRC2-EED-I363M is the persistent ionic interactions between the unmodified lysine side chain of the compound and D140 located in a flexible loop of EZH2 (residues 136-142), which drastically reduces the mobility of the loop, thus preserving the integrity of the adjacent SRM helix (see the main manuscript for details). The color scheme is the same as that in Figure 4.