dock the filtered SPECS library to the substrate binding pocket with Glide program
using Glide SP as scoring function. Afterward, the top-ranked 1,000 compounds
were clustered according to their structural similarity and interactions with key
residues like Phe327, Phe580, or Glu444 (human PRMT5 numbering). Among the
42 selected and tested hits, DC_P33, a low micromolar hit (IC 50 35.6 μM), could be
obtained. Based on DC_P33, a chemical fingerprint similarity search was executed
which could help in investigating the structure-activity relationship (SAR) of this
series. Although the analogs showed lower activity, they could guide the optimization studies that led to discovery of DC_C01 (24, Fig. 6) with an IC 50 of 2.8 μM. The
selectivity of DC_C01 was tested and confirmed against PRMT1, EZH2, and
DNMT3A. However, the optimized hit is still ~100 times less potent than the
reported reference inhibitor for PRMT5, EPZ015666, which shows an IC 50 of
22 nM (25, Fig. 6).
Regarding PKMTs, several studies using computer-based approaches have been
reported for G9a [111–115], a SET domain-containing lysine methyltransferase
protein also known as euchromatic histone-lysine N-methyltransferase
2 (EHMT2), and some of them are summarized here.
Kubicek et al. combined diverse computer-based methods in order to filter down
the Boehringer Ingelheim chemical compound library prior to a HTS assay for G9a
[111]. Using nine known small molecule inhibitors of arginine and lysine
methyltransferases [116], a similarity searching approach was carried out. Meanwhile for the structure-based approach, a homology model based on the crystal
structure of DIM-5 [117], a lysine methyltransferase with 30% identity in the SET
domain to G9a, was generated. Thereby, ligand- and structure-based pharmacophore
fingerprint and site-point [118] models were developed and used for filtering the
library. Other compounds were randomly selected, and a total of ~125,000 compounds were evaluated in HTS which resulted in seven confirmed hits with IC 50 in
the low micromolar range. Of note is BIX-01294 (26, Fig. 6), which specifically
inhibited G9a at 1.7 μM (IC 50 ), to a lesser extent the closely related GLP (G9a-like
protein) with an IC 50 of 38 μM and showed no activity against SUV39H1 and
PRMT1. In addition, the mode of action was investigated and indicated a
noncompetitive inhibition with the cofactor SAM. Later, the quinazoline-based
scaffold of BIX-01294 was further explored by Liu et al. with the aim to elucidate
the SAR and improve potency and selectivity [112]. A highly potent and selective
inhibitor, UNC0224 (27, Fig. 6) with an IC 50 of 15 nM, was thus identified, and the
first crystal structure of G9a with a small molecule inhibitor was subsequently
resolved.
Using the abovementioned G9a crystal structure and the ChemBridge CORE
library, Zhang et al. discovered benzoxazole and benzimidazole scaffolds as new
G9a inhibitors by means of structure-based VS and optimization steps. The study
resulted in compound GA001 (28, Fig. 6) with moderate G9a inhibition and cellular
activity (G9a IC 50 of 1.32 μM, MCF7 cells IC 50 of 5.73 μM) [113].
An integrated procedure combining pharmacophore- and docking-based VS,
similarity search, and hit optimization, which led to the discovery of micromolar
468
D. Robaa et al.
using Glide SP as scoring function. Afterward, the top-ranked 1,000 compounds
were clustered according to their structural similarity and interactions with key
residues like Phe327, Phe580, or Glu444 (human PRMT5 numbering). Among the
42 selected and tested hits, DC_P33, a low micromolar hit (IC 50 35.6 μM), could be
obtained. Based on DC_P33, a chemical fingerprint similarity search was executed
which could help in investigating the structure-activity relationship (SAR) of this
series. Although the analogs showed lower activity, they could guide the optimization studies that led to discovery of DC_C01 (24, Fig. 6) with an IC 50 of 2.8 μM. The
selectivity of DC_C01 was tested and confirmed against PRMT1, EZH2, and
DNMT3A. However, the optimized hit is still ~100 times less potent than the
reported reference inhibitor for PRMT5, EPZ015666, which shows an IC 50 of
22 nM (25, Fig. 6).
Regarding PKMTs, several studies using computer-based approaches have been
reported for G9a [111–115], a SET domain-containing lysine methyltransferase
protein also known as euchromatic histone-lysine N-methyltransferase
2 (EHMT2), and some of them are summarized here.
Kubicek et al. combined diverse computer-based methods in order to filter down
the Boehringer Ingelheim chemical compound library prior to a HTS assay for G9a
[111]. Using nine known small molecule inhibitors of arginine and lysine
methyltransferases [116], a similarity searching approach was carried out. Meanwhile for the structure-based approach, a homology model based on the crystal
structure of DIM-5 [117], a lysine methyltransferase with 30% identity in the SET
domain to G9a, was generated. Thereby, ligand- and structure-based pharmacophore
fingerprint and site-point [118] models were developed and used for filtering the
library. Other compounds were randomly selected, and a total of ~125,000 compounds were evaluated in HTS which resulted in seven confirmed hits with IC 50 in
the low micromolar range. Of note is BIX-01294 (26, Fig. 6), which specifically
inhibited G9a at 1.7 μM (IC 50 ), to a lesser extent the closely related GLP (G9a-like
protein) with an IC 50 of 38 μM and showed no activity against SUV39H1 and
PRMT1. In addition, the mode of action was investigated and indicated a
noncompetitive inhibition with the cofactor SAM. Later, the quinazoline-based
scaffold of BIX-01294 was further explored by Liu et al. with the aim to elucidate
the SAR and improve potency and selectivity [112]. A highly potent and selective
inhibitor, UNC0224 (27, Fig. 6) with an IC 50 of 15 nM, was thus identified, and the
first crystal structure of G9a with a small molecule inhibitor was subsequently
resolved.
Using the abovementioned G9a crystal structure and the ChemBridge CORE
library, Zhang et al. discovered benzoxazole and benzimidazole scaffolds as new
G9a inhibitors by means of structure-based VS and optimization steps. The study
resulted in compound GA001 (28, Fig. 6) with moderate G9a inhibition and cellular
activity (G9a IC 50 of 1.32 μM, MCF7 cells IC 50 of 5.73 μM) [113].
An integrated procedure combining pharmacophore- and docking-based VS,
similarity search, and hit optimization, which led to the discovery of micromolar
468
D. Robaa et al.