Since MBD2 is a miscellaneous participant in various biological phenomena and disorders, that links epigenetic methylation marks and gene expression regulation, it has arisen as a potent target for multi-purpose. In current study, we aimed to inspect novel drug candidates targeting MBD2MBD through docking–based virtual screening and drug repurposing approaches. In this respect, we generated a comprehensive library including over 15,000 non-annotated compounds and biologically identified drugs, from ZINC15 database. Certain of ligands have been listed in Supplementary Table S1 as hitter ligands by filtering according to resulted binding affinity and physicochemical properties such as complexity, molecular weight, tPSA and also convenience to be purchasable or synthesizable for de facto attempts. Here, we have realized that numerous hit compounds (see Supplementary Table S1) and also unlisted ones consists of remarkable groups such as pyridine, quinoxaline, imidazole, isoquinoline, which are incorporated either in nucleosides and nucleotide analogs, or at least related structures to directly nucleotides or DNA-related constructs 44–46. For instance, quinoxaline and isoquinoline derivatives have been demonstrated to bind and act as antagonists of cyclic nucleotide-related enzymes such as cyclic nucleotide phosphodiesterase 47 and P2X7 nucleotide receptor 48,49, suggesting these groups might be attractive for DNA– or nucleotide–recognizing domains. Furthermore, some of screened chemicals were purine– or pyrimidine–containing molecules (ex. CID3100583, CID136707272, CID135623722, CID124903520, CID136732748, TB6:RiboPyrimidoP) or straightly nucleotide/nucleoside derivatives (ex. Regadenoson, Wyosine, 8-Methylamino-Guanosine, Thiosangivamycin, Adenosine 5'-monophosphoramide, 5,6-Trimethyleneuridine, Fludarabine Base, 8-Aminoguanosine, Tubercidin, Hypoxanthosine, GHMcytosine, 6-Mercapto-7-Methylguanosine, 8-Chloroguanosine, Isovaleriansaures Coffein, 2-Methylformycin, 5'-N-Methylcarboxamidoadenosine, etc.), documented in Supplementary Table S1. Additionally, these compounds highly contain methyl groups. Indeed, it has been quite interesting that a considerable number of particularly adenosine and guanosine derivatives have become prominent as a consequence of drug screening. It was reported that residues in MBD region of MeCP2 and MBD2 exhibited binding affinity to mCAT, mCAC, mCAT, mCGG patterns and surprisingly unmethylated TG-dinucleotides 31,50. Thereby, such interplay preferences of MBD might reflect why these compounds were recorded as hit molecules amongst the whole library. Then top ten of non-annotated and top five annotated molecules were proceeded for further docking analysis. Principally, CID3100583 (ZINC3109386) and 8,8-Ethylenebistheophylline (ZINC8612354) urged us for the exhaustive investigation due to the highest binding energy and the lowest Ki (Table 1). At this point, chemical contexts of both substances have been intriguing since CID3100583 contains two pyrimidine–based rings and Ebis-Theophylline is composed of two interconnected druggable Theophylline (Dimethylxanthine) molecules, a methylated xanthine analog which antagonizes human A1, A2a, A2b Adenosine Receptors as in the treatment of airways diseases and vasodilation 51–53. Together, similarities between actual recognition target of MBD2MBD (various nucleotide patterns above) and top hitters in the context of chemical and structural features have been speculated as possible reason underlying complementarity and fine docking.
Actually, all ligands have been demonstrated interacting with similar residues (Fig. 1). CID3100583 was prone to anchor MBD2MBD via 9 hydrogen bonds with amino acids Val177, Arg188, Lys190, Ser204, Phe206, Asp207, Phe208, and Arg209 (Table 2) and besides through van der Waals forces (Asp176, Tyr178, Ser189 and Phe208), pi-Alkyl (Val177), pi-Anion (Asp207) interactions (Table 3 and Fig. 2a-c). In case of Ebis-Theophylline, residues Val177, Phe208, Arg209 have participated in total 5 hydrogen bonds (Table 2) and van der Waals interactions between Arg166, Ser189, Pro191; pi-Alkyl interactions with Tyr178, Arg188, Lys190, Arg209; and pi-Anion interactions with Asp176 (Table 3 and Fig. 2d-e), signifying both ligands have been stabilized by cementing highly reactive surroundings. In this respect, Liu et al. reported MBD2 bound to DNA via constitution of H-bonding between Arg166, Asp176, Tyr178, Arg188 within MBD2MBD and methylated-Cytosine (mC) along with adjacent Thymine or Guanine bases in the context of mCGG, mCAG, mCAT, mCC, mCT di-/tri-nucleotide patterns 31. Along with pi-Anion interactions between referred amino acids and aromatic bases above, alternate residues such as Lys167, Ser168, Ser171, Lys186, Ser189, Lys190, Arg209, interacted with sugar-phosphate backbone of DNA. Buchmuller et al. also attributed regions 164–177 and 186–195 in vicinity to DNA duplex 27. In other studies, Lys174 and Tyr178 have been reported as critical amino acids to selectively recognize and bind methylated CpG probes 50,54. Favorably, all of these stated amino acids have been detected highly conserved between MBD1, MBD3, MBD4 and MeCP2 in human (see Supplementary Fig. S4) and MBD2 in different organisms via multiple sequence alignment. Residual counterparts in both MBD1 (Val20, Arg22, Tyr34, Arg44 and Ser45) 55 and MeCP2 (Arg106, Arg133, Trp155, Thr158) 31,56 have been involved in methylated-DNA recognition and interaction. These highlight the fact that ligands achieved in our study, especially CID3100583 and 8,8-Ethylenebistheophylline could actually interfere DNA–binding by selectively invading of those indispensable residues (Fig. 2) for the usual function of related domain. Moreover, our in silico prediction also yielded in a convergence between putative druggable pockets and ligand–anchored pockets (see Supplementary Fig. S3 and Table 2–3), again supporting the importance of docking sites of screened compounds, all together (Fig. 2 and see Supplementary Fig. S2). In this regard, CID3136570 (ZINC3151175) as one of the patented compound by Nelson et al., was quite outstanding for our consideration. Therein, it was invented to inhibit DNA-binding activity of both MBD2MBD and MeCP2MBD with relatively better IC50 of 0.67 µM and 0.61 µM, respectively 57. Because binding action remained unclear, it led us to survey its docking residues and energy on MBD2MBD in a similar manner. As a reference, binding energy resulted in our docking analysis was − 7.61 kcal/mol, even lower than assumptions for suggested binders in present study (Table 1). More interestingly, we detected Val177, Arg188, Arg209 in H-bonding while Arg166, Tyr178, Lys190, Phe208, Arg209 facilitated in hydrophobic interactions (see Supplementary Fig. S5), thus compatibly appreciating the value and expected potential of our outcome in a prospective real assay.
Next, we simulated each ligand–protein complex under a force–field in order to test their structural dynamics and stability in a time course. Initially, RMSD along the whole trajectory for each MBD2MBD–ligand system was compared. As plotted in Fig. 3a and 4a, CID3100583, CID343482 and Ebis-Theophylline established a more stable dimensionality effects or conformational distributions over time, corresponding docking affinities (Table 1). Herein, both ligands (CID3100583 and Ebis-Theophylline) along with protein backbone, showed amenable fluctuations and equilibration timing concurrently with MBD2MBD receptor (Fig. 3b and 4b). As known, RMSD is not useful in practice to deeply analyze and distinguish local regions in more flexible or rigid 58. Consistently, amino acids participated in particularly H-bonding with both CID3100583 and Ebis-Theophylline, such as Val177, Arg188, Lys190, Ser204, Phe206, Asp207, Phe208, Arg209, remained non-fluctuating as revealed by RMSF and B-Factor graphics (Fig. 3c and 4c). Actually, regions between 160–166, 175–780, 186–188, coincide with strand and residues 190–198 corresponds to a helix31 (also see Supplementary Fig. S4). As we expected, diminished B-factor along these districts were corroborated with the consent that increased fluctuations and thus higher B-factor is predicted for irregular loop regions 59. More in particular, secondary structures including Val177, Arg188, Lys190 displayed more regularity than part of 160–166, suggesting durability contributed by ligand binding. Surprisingly, even though amino acids between Ser204 and Arg209 have been involved within an irregular loop region (see Supplementary Fig. S4), they possessed lower B-factor values (Fig. 3c and 4c). The magnitude of that reduction may presumably be due to immobilization of the loop region by extensive interactions with either CID3100583 or Ebis-Theophylline over there. As a matter of fact that interactions with CID3100583 were potently concentrated on residues between Ser204 and Arg209, hence ensuring a further decline even rather than strands or helices and corresponding zone in MBD2MBD–Ebis-Theophylline complex (Fig. 3c and 4c). Of note, RoG has moderately diminished at around frames 250 and 400 for CID3100583 and Ebis-Theophylline (Fig. 3d and 4d), respectively, postulating a compaction in both protein–ligand systems 39,40. Within this context, spectroscopic analysis revealed that upon methylated CpG binding, dynamic mobility and conformation of complex has been shifted under favor of DNA bending by MBD2MBD 60. Similarly, Liu et al. reported a compaction in methylated DNA–bound MeCP2MBD detected by atomic force microscopy (AFM) in parallel with decrease in RoG correlation 61. That may point the hypothesis that reduction in RoG and a parallel increment in Q(x) could be resulted from a further compaction and folding due to a configuring action of docked CID3100583 or Ebis-Theophylline by mimicking an ordinary MBD–DNA binding. In correlation, acquired energetics data also corroborated our docking results and conclusions so far. Both CID3100583 and Ebis-Theophylline have been ranked amongst the top ligands with the highest binding free energy (see Supplementary Table S2). Energy decomposition for CID3100583 showed Asp207, Arg209, Phe208 and Phe206 appeared in the top five in terms of contribution to overall ΔGBind (Fig. 5a), supporting rigidity in the loop region discussed above. Besides, per residue energy decomposition (Fig. 5) and interactions in docking analysis (Fig. 2) established a fine correspondence for both ligands. For instance, Phe208, one of the top van der Waals energy–decomposing residue for CID3100583 was also constituted in van der Waals interaction as a result of docking analysis while Val177 incorporated with Ebis-Theophylline by electrostatic forces via both approaches (Fig. 5b).
Consequently, current computational study has established prosperous and crosschecking evidence for certain ligands in a comprehensive manner. CID3100583 and 8,8-Ethylenebistheophylline have successfully been invented as novel and most likely MBD2 inhibitors in the light of effective binding affinity and negligible off-target probability. Though in silico approach was utilized within the scope of this paper, we have provided a consolidated basis for the next experimental studies testing their actual druggability. Likewise, a wide compound repertoire has also been gained to give insight into drug development or repurposing against multi-functional MBD2 target and still required to be evaluated. In conclusion, we rationally propose that CID3100583 and 8,8-Ethylenebistheophylline are highly versatile with capability to be notably used as anti-cancer drugs and agents in iPSC reprogramming cocktails by impeding MBD2 function in related pathways.