The analysis of the docking poses could suggest some key interactions between the
newly identified inhibitors and PRMT1. For example, a common feature of the
active inhibitors is the hydrogen bond of a basic or polar group with the acidic
residue Glu152 (human PRMT1 numbering) of the active site. Moreover, van der
Waals interactions between the inhibitors and several aromatic residues in the
binding pocket (Tyr47, Tyr156, Trp302, human PRMT1 numbering) were present.
Interestingly, the inhibitors showed a reduction of estrogen receptor activation by
estradiol in a dose-dependent manner. Based on the first positive outcomes and the
latest insight into the ligand-protein interactions, further research was conducted. In
the new study, Heinke et al. screened the ChemBridge database containing 328,000
molecules using a structure-based pharmacophore generated with LigandScout [107]
on the basis of the PRMT1-allantodapsone complex [108]. The hits were then
docked using the GOLD program and GoldScore as fitness function into the
substrate binding site, and the top-ranked compounds were filtered based on the
distances to the amino acids in the binding pocket which were already identified as
important binding partners (i.e., the aforementioned Glu152 and Tyr156). Nine
compounds among the top-ranked 100 compounds that fulfilled the distance constraints were manually selected and tested in vitro. This approach resulted in the
discovery of inhibitors 20, 21, and 22 (Fig. 6) with inhibitory effects down to 13 μM
[108]. Even though the new hits did not show improvement of the activity in
comparison with allantodapsone (IC 50 1.7 μM), they could help to prove the
importance of specific interactions and to demonstrate the ability of VS to significantly increase the hit rates.
Another successful application of computer-based methods that guided the discovery of several PRMT4 (also named CARM1) inhibitors was reported by de
Freitas et al. [109]. Their strategy encompassed a combination of ligand- and
structure-based VS approaches followed by rescoring steps. The starting point of
the study was the structural analysis of some available crystal structures of PRMTligand complexes, which led to the identification of a common feature present in the
inhibitors. Indeed, all five inhibitors under analysis show a basic amine tail that is
anchored in the PRMT substrate arginine-binding channel. Thus, they constructed a
PRMT-focused virtual library by filtering the ZINC database (~22 million compounds) with substructure searches focused on the two basic amine tails featured in
known PRMT inhibitors. This resulted in a virtual library of ~132,000 compounds
that was consequently subjected to docking studies by means of Glide (Glide SP
scoring function) using the tail structures as reference core. The top-ranked 1084
docking poses were then rescored using a MM-PBSA protocol. Out of 51 tested
compounds, 11 showed dose-response inhibition of PRMT4. Among these, a valuable hit with an IC 50 of 1.9 μM was selected for further investigation. As a result of
two rounds of structure-based optimization, the nanomolar lead 23 (IC 50 50 nM,
Fig. 6) was obtained. With the exception of PRMT6 (IC 50 5.2 μM), compound 23
showed selectivity against a panel of 21 human protein methyltransferases.
Recently, Ye et al. described a structure-based VS followed by a second round of
fingerprint similarity search and a final optimization step for another arginine
methyltransferase, PRMT5 [110]. A crystal structure of PRMT5 was employed to
Structure-Based Design of Epigenetic Inhibitors
467
newly identified inhibitors and PRMT1. For example, a common feature of the
active inhibitors is the hydrogen bond of a basic or polar group with the acidic
residue Glu152 (human PRMT1 numbering) of the active site. Moreover, van der
Waals interactions between the inhibitors and several aromatic residues in the
binding pocket (Tyr47, Tyr156, Trp302, human PRMT1 numbering) were present.
Interestingly, the inhibitors showed a reduction of estrogen receptor activation by
estradiol in a dose-dependent manner. Based on the first positive outcomes and the
latest insight into the ligand-protein interactions, further research was conducted. In
the new study, Heinke et al. screened the ChemBridge database containing 328,000
molecules using a structure-based pharmacophore generated with LigandScout [107]
on the basis of the PRMT1-allantodapsone complex [108]. The hits were then
docked using the GOLD program and GoldScore as fitness function into the
substrate binding site, and the top-ranked compounds were filtered based on the
distances to the amino acids in the binding pocket which were already identified as
important binding partners (i.e., the aforementioned Glu152 and Tyr156). Nine
compounds among the top-ranked 100 compounds that fulfilled the distance constraints were manually selected and tested in vitro. This approach resulted in the
discovery of inhibitors 20, 21, and 22 (Fig. 6) with inhibitory effects down to 13 μM
[108]. Even though the new hits did not show improvement of the activity in
comparison with allantodapsone (IC 50 1.7 μM), they could help to prove the
importance of specific interactions and to demonstrate the ability of VS to significantly increase the hit rates.
Another successful application of computer-based methods that guided the discovery of several PRMT4 (also named CARM1) inhibitors was reported by de
Freitas et al. [109]. Their strategy encompassed a combination of ligand- and
structure-based VS approaches followed by rescoring steps. The starting point of
the study was the structural analysis of some available crystal structures of PRMTligand complexes, which led to the identification of a common feature present in the
inhibitors. Indeed, all five inhibitors under analysis show a basic amine tail that is
anchored in the PRMT substrate arginine-binding channel. Thus, they constructed a
PRMT-focused virtual library by filtering the ZINC database (~22 million compounds) with substructure searches focused on the two basic amine tails featured in
known PRMT inhibitors. This resulted in a virtual library of ~132,000 compounds
that was consequently subjected to docking studies by means of Glide (Glide SP
scoring function) using the tail structures as reference core. The top-ranked 1084
docking poses were then rescored using a MM-PBSA protocol. Out of 51 tested
compounds, 11 showed dose-response inhibition of PRMT4. Among these, a valuable hit with an IC 50 of 1.9 μM was selected for further investigation. As a result of
two rounds of structure-based optimization, the nanomolar lead 23 (IC 50 50 nM,
Fig. 6) was obtained. With the exception of PRMT6 (IC 50 5.2 μM), compound 23
showed selectivity against a panel of 21 human protein methyltransferases.
Recently, Ye et al. described a structure-based VS followed by a second round of
fingerprint similarity search and a final optimization step for another arginine
methyltransferase, PRMT5 [110]. A crystal structure of PRMT5 was employed to
Structure-Based Design of Epigenetic Inhibitors
467