Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Apr 11;12(4):e0174410.
doi: 10.1371/journal.pone.0174410. eCollection 2017.

In Silico Design of the First DNA-independent Mechanism-Based Inhibitor of Mammalian DNA Methyltransferase Dnmt1

Free PMC article

In Silico Design of the First DNA-independent Mechanism-Based Inhibitor of Mammalian DNA Methyltransferase Dnmt1

Vedran Miletić et al. PLoS One. .
Free PMC article


Background: We use our earlier experimental studies of the catalytic mechanism of DNA methyltransferases to prepare in silico a family of novel mechanism-based inhibitors of human Dnmt1. Highly specific inhibitors of DNA methylation can be used for analysis of human epigenome and for the creation of iPS cells.

Results: We describe a set of adenosyl-1-methyl-pyrimidin-2-one derivatives as novel mechanism-based inhibitors of mammalian DNA methyltransferase Dnmt1. The inhibitors have been designed to bind simultaneously in the active site and the cofactor site and thus act as transition-state analogues. Molecular dynamics studies showed that the lead compound can form between 6 to 9 binding interactions with Dnmt1. QM/MM analysis showed that the upon binding to Dnmt1 the inhibitor can form a covalent adduct with active site Cys1226 and thus act as a mechanism-based suicide-inhibitor. The inhibitor can target DNA-bond and DNA-free form of Dnmt1, however the suicide-inhibition step is more likely to happen when DNA is bound to Dnmt1. The validity of presented analysis is described in detail using 69 modifications in the lead compound structure. In total 18 of the presented 69 modifications can be used to prepare a family of highly specific inhibitors that can differentiate even between closely related enzymes such as Dnmt1 and Dnmt3a DNA methyltransferases.

Conclusions: Presented results can be used for preparation of some highly specific and potent inhibitors of mammalian DNA methylation with specific pharmacological properties.

Conflict of interest statement

Competing Interests: ZMS is the recipient of a paid consultant for Jiva Pharmaceuticals. The presented in silico studies have been used as a foundation for our organic synthesis efforts that have given several highly potent inhibitors and an exclusive contract with a biopharmaceutical company, and we also have intention to file a patent for this inhibitor. There are no further patents, products in development or marketed products to declare. This does not alter our adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors.


Fig 1
Fig 1. Catalytic mechanism of DNA methyltransferase Dnmt1.
The catalytic mechanism of bacterial and mammalian cytosine-C5 DNA methyltransferase has been described in details in the last 40 years in different crystal structures, enzyme kinetics studies, enzyme substrate binding studies, and QM/MM studies [–21, 29, 34]. The catalysis consists of three main steps. The catalysis is initiated when the aromatic target base ring is posited in the active site cavity (step 1.1). In the case of human and mouse enzymes, the active site is composed of conserved Arg 1310, Arg 1312, Glu1266, and Cys 1126 residues (step 1.1). The enzyme can form unstable covalent adduct intermediate when conjugated aromatic bonds in the target base ring are positioned in an asymmetric active site between polar amino acids (steps 1.1 to 1.2). Kinetic and QM/MM studies showed that formation of the covalent adduct intermediate is a reversible step unless the reaction is driven forward by the rate-limiting irreversible methyltransfer step [19, 20, 29]. Following the methyltransfer step (steps 2.1 to 2.2), there is a reversible target base eliminations step (steps 3.1 to 3.2). We have decided to target this reversible step with a mechanism-based suicide-inhibitor. The structure 4.1 is supposed to mimic the transition from structures 2.1 to 2.2. Thus, when structure 4.1 is positioned in the active site it should trigger the formation of the covalent adduct shown as the structure 4.2. This is in essence, the same process as the transition from structure 3.2 to structure 3.1. The figures were drawn in ChemAxon Marvin 15.0.1 [39]
Fig 2
Fig 2. (A-B) Lead structure for the mechanism-based suicide-inhibitor of Dnmt1.
(A) The figure shows our first prototype for a mechanism-based inhibitor of Dnmt1. The structure is derived from the structure 4.1 in Fig 1 as a reasonable challenge for organic synthesis that has a reasonable agreement with the Lipinski's rules [26, 27]. The structure is designed to act as a transition state analogue and consists of four functional parts: adenine, ribose, tertiary amine linker and 1-methyl-pyrimidin-2-one as the target base ring. The adenine and ribose parts are selected to assure that the lead compound can bind to AdoHcys/AdoMet sites. Tetrahedral tertiary amine linker is chosen as a flexible structure that can mimic the transition state structure (Fig 1). 1-methyl-pyrmidine-2-one was chosen as the target base ring instead of cytosine, to avoid problems with deamination at carbon 4 and to decrease polar surface on the target base ring [19, 22]. (B) The present structure is in agreement with the Lipinski's rules except for its low LogD value. The relative molecular mass is 417.45 Da. There are 6 rotatable bonds, and a total of 58 bonds between 55 atoms. At pH 7.40 the polar surface area is 156.42 Å2 while Van der Walls surface area is 562.17. There are five H-bond donor sites and 13 acceptor sites. At pH 7.40, 94% of the ligand is protonated at the linker nitrogen and the compound is positively charged. The LogD value in protonated form is -2.63 and -1.88 in unprotonated form. To improve ADME properties the future modifications will be designed to increase the LogD value. This can be achieved by removing the positively charged nitrogen at the linker, and by removing heteroatoms at the ribose and adenine ring that give the highest negative contribution to the LogD value. The numbers on the structure show contribution to the total LogD for each atom. All parameters have been calculated using ChemAxon Marvin suite version 15.3.16 [39]
Fig 3
Fig 3. MM/MD simulations of binding interactions between our prototype inhibitor and Dnmt1 (PDB: 4DA4 [18]).
(A) The cross-eyed stereo image shows that our lead compound can bind to Dnmt1 in the same position as AdoHcy and the target cytosine in crystal structures of Dnmt1 (PDB: 4DA4 [18]). The image shows the lead compound (lime green), the interacting amino acids (cyan) and the crystal structure of AdoHcys (pink) and the target cytosine (pink) [41]. Hydrogen atoms are shown in white, oxygen in red, nitrogen in blue, and sulfur in yellow. The figure was prepared by superimposing one of the MM/MD simulation frames of Dnmt1-inhibitor complex to the crystal structure of Dnmt1 in complex with AdoHcys and the target cytosine (PDB: 4DA4 [18]). The table on the left gives a full list of all possible binding interactions that can be observed in MM/MD simulations. The most stable interactions are shown in bold. For clarity, the figure does not show binding interactions with Cys1191, Phe1145, Trp 1170, Arg 1310 or the surrounding solvent molecules. All of the interacting amino acids are conserved between mouse and human Dnmt1 [21].
Fig 4
Fig 4. (A-D) MM/MD simulations of binding interactions between our prototype inhibitor and Dnmt1 (PDB: 4DA4 [18]).
Dynamic binding interactions within Dnmt1-linhibtor complex depicted in Fig 3 can be analyzed by following changes in “hydrogen-bond” plots (A), “average-ligand-RMSD” plots (B), “Cys1226-carbon-6-distance” plots (C), and “inhibitor pulling force” plots (D). (A) In average 7±2 hydrogen bonds can be observed in Dnmt1-inhibitor complex in 100 nsec simulations (S1 Movie). (B) Distinct steps and peaks in RMSD plots represent different conformations of the ligand within the complex, while uniform RMSD plots represent limited ligand mobility and one dominant conformation in a stable complex ([37, 38] and S1 Movie). (C) Distinct steps and peaks in “Cys1226-carbon-6-distance” plot represent different swiveling motions by the target base in the active site cavity (S1 and S3 Movies). (D) Steered molecular dynamics protocol was used to calculate “inhibitor pulling force” plots and the corresponding changes in binding interactions (S2 Movie). Changes in the pulling force that correspond to different dissociation steps are labeled with numbers. The numbers indicate: (1) force required to initiate movement of the inhibitor in its binding site and the resulting opening of active site loop; (2) force required for full displacement of the target base; (3–4) force required for full dissociation of the ribose ring from Glu 1168 and Arg 1573 respectively; (5) force required for full dissociation of the adenine ring.
Fig 5
Fig 5. (A-B) Binding of the prototype inhibitor to its binding site groove on Dnmt1.
The two cross-eyed stereo images show the lead compound bound to mouse Dnmt1 as in Fig 3, except that the binding site is shown in the surface mode to highlight its shape, depth, and hydrophobic patches. (A) image shows surface hydrogens in white, carbons in black, nitrogens in blue, oxygens in red, and sulfur in yellow [41]. The colors can help in tracing the amino acids shown in Fig 3 on the groove surface. (B) the image shows hydrophobic surface in brown, and hydrophilic surface in blue [37, 38]. In both images, the prototype inhibitor is shown in stick model and colored green. Reader is looking at the binding site cleft from the position of DNA substrate, therefore methyl group on nitrogen 1 of the target base ring is positioned almost perpendicular to the plane of paper. The adenine ring is positioned in a tight hydrophobic surface pocket enclosed by Met1169, Asn1190 and Cys1191. The ribose ring is anchored on a wide ridge with its two OH group leaning atop of Glu1168. The linker is positioned in the widest section of binding site cleft just above AdoHcys/AdoMet binding cavity. The target base ring is positioned in the active site cavity and its carbon 6 is in Van der Waals contact with Cys1226.
Fig 6
Fig 6. Frontier molecular orbitals on the target base and on the active site cysteine 1226.
The LUMO orbitals on carbon 6 of the target base ring are perpendicular to the plane of the ring. The two HOMO orbitals on nucleophilic sulfur anion are perpendicular to each other and have almost identical energy. The reaction can start when one of the two HOMO orbitals on the sulfur atom comes in the plane of the target base ring (S3 Movie). The orbitals and the corresponding energies were with program GAMESS [42] using DFT B3LYP protocol and visualized in VMD [41]. The bond forming motions can be driven by the by the rotation of the target base ring around its hydrogen bonds axis with Arg 1312 (S3 Movie).
Fig 7
Fig 7. The cross-eyed stereo image shows the inhibitor bound to the crystal structure of Dnmt1 with the active site loop open (PDB code 3AV6, [32]).
[41]. The inhibitor is docked in its binding site just as in Figs 3 and 4, except that the active site is open (amino acids 1220 to 1236). All binding interactions are still present, except that the calculated “Cys1226-carbon-6-distance” is 9.2 Å. Thus the inhibitor can bind to Dnmt1 when its active site is in open and closed position, however mechanism-based suicide-inhibition can happen only when the active site loop is closed.
Fig 8
Fig 8. (A-E) The lead compound can bind to mouse Dnmt3a (PDB: 4U7P ref [18]).
(A) The cross-eyed stereo image shows our prototype inhibitor bound at the AdoMet site on Dnmt3a. Similar to the complex with Dnmt1, with Dnmt3a adenine part of the inhibitor is buried in a hydrophobic pocket, while the ribose is sitting atop of a ridge formed by Glu664. Hydrogen atoms are shown in white, carbon in gray, oxygen in red, nitrogen in blue, and sulfur in yellow [41]. (B) Stereo image shows inhibitor (lime) bound to Dnmt3 (cyan). The adenosyl part of inhibitor closely overlaps with corresponding parts of AdoHcys in the crystal structure (orange). Similar to Dnmt1 the ribose ring on Dnmt3a is positioned between Glu664 and Arg891. The target base ring is held in its position by π-π stacking interactions with Trp893. (C) Stereo image shows a comparison between the extended conformation observed in the binding site of Dnmt1 (gold), and the twisted conformation observed in the binding site of Dnmt3 (cyan). (D) The plot shows dynamic changes in the number of binding interactions within Dnmt3a-inhibitor complex in a 100 nsec long MD simulation. With Dnmt3a our prototype inhibitor forms 6 ± 2 binding interactions, what is in average at least one interaction less than with Dnmt1. (E) The plots show changes in pulling force and the corresponding decrease in the number of dynamic binding interactions within Dnmt3a-inhibitor complex. The two plots were calculated using steered molecular dynamics protocols in program GROMACS [31].

Similar articles

See all similar articles

Cited by 1 article


    1. Covic M, Karaca E, Lie D. Epigenetic regulation of neurogenesis in the adult hippocampus. Heredity. 2010;105(1):122–34. 10.1038/hdy.2010.27 - DOI - PubMed
    1. Romanoski CE, Glass CK, Stunnenberg HG, Wilson L, Almouzni G. Epigenomics: Roadmap for regulation. Nature. 2015;518(7539):314–6. 10.1038/518314a - DOI - PubMed
    1. Pennarossa G, Maffei S, Campagnol M, Tarantini L, Gandolfi F, Brevini TA. Brief demethylation step allows the conversion of adult human skin fibroblasts into insulin-secreting cells. Proceedings of the National Academy of Sciences. 2013;110(22):8948–53. - PMC - PubMed
    1. Khavari DA, Sen GL, Rinn JL. DNA methylation and epigenetic control of cellular differentiation. Cell Cycle. 2010;9(19):3880–3. - PubMed
    1. Sen GL, Reuter JA, Webster DE, Zhu L, Khavari PA. DNMT1 maintains progenitor function in self-renewing somatic tissue. Nature. 2010;463(7280):563–7. 10.1038/nature08683 - DOI - PMC - PubMed

Publication types

MeSH terms

Grant support

This work has been supported by the University of Rijeka project number 511-12; and the Croatian Ministry of Science. We gratefully acknowledge services of NVIDIA CUDA Teaching Center, and high performance computing facility at the Center for Advanced Computing and Modeling provided by the University of Rijeka. ZMS is recipient of funds from Croatian Science Foundation’s project number O-1505-2015, and a paid consultant for Jiva Pharmaceuticals. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.