Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 www.chemmedchem.org Structure-Based Virtual Screening and in vitro and in vivo Analyses Revealed Potent Methyltransferase G9a Inhibitors as Prospective Anti-Alzheimer’s Agents Aina Bellver-Sanchis+,[a] Bhanwar Singh Choudhary+,[b, c] Júlia Companys-Alemany,[a] Sukanya,[b] Pedro A. Ávila-López,[d] Antón Leandro Martínez Rodríguez,[e] Jose Manuel Brea Floriani,[e] Ruchi Malik,[b] Mercè Pallàs,[a] Belén Pérez,[f] and Christian Griñán-Ferré*[a] G9a is a lysine methyltransferase able to di-methylate lysine 9 function range was safe and recovered age-dependent paraly- of histone H3, promoting the repression of genes involved in sis. Likewise, we demonstrated that the best candidates direct learning and memory. Novel strategies based on synthesizing target G9a by reducing H3 K9me2 in the CL2006 strain. Further epigenetic drugs could regulate gene expression through characterization of these compounds involved the assessment histone post-translational modifications and effectively treat of the blood-brain barrier-permeability and impact on amyloid- neurodegenerative diseases, like Alzheimer’s disease (AD). Here, β aggregation, showing promising results. Thus, we present a potential G9a inhibitors were identified using a structure-based G9a inhibitor candidate, F, with a novel and potent structure, virtual screening against G9a, followed by in vitro and in vivo providing both leads in G9a inhibitor design and demonstrating screenings. First, screening methods with the AD transgenic their participation in reducing AD pathology. Caenorhabditis elegans strain CL2006, showed that the toxicity/ Introduction chemical modifications of the amino acid residues of proteins, leading to changes in gene expression. This process is mainly Epigenetic modifications have an important impact on regulat- mediated by epigenetic “writer” enzymes, such as histone ing the expression and transcription of genes involved in many methyltransferases (HMTs), which are responsible for adding human diseases, such as cancer,[1] addiction,[2] psychiatric[3] and chemical methyl groups in specific residues (lysine (K) or neurodegenerative disorders.[4] One of the most relevant arginine (R)), and as a consequence, leading to active or epigenetic mechanisms to control gene expression is chromatin repressive marks. In general, methylation of H3 K9 and H3 K27 remodeling.[5] This is catalyzed by different enzymes and is are typical hallmarks of transcription repression, whereas meth- responsible for the dynamic modification of chromatin struc- ylation of H3 K4, H3 K36, and H3 K7 are considered active ture, enabling or restricting the access of the transcription marks. Hence, dysregulation of methylation at specific histone machinery to condensed genomic DNA. An essential process sites has been implicated in many human diseases, highlighting responsible for the alteration of chromatin architecture is the the potential therapeutic value of HMTs as targets.[6] post-translational modification (PTM) of histones. PTMs are [a] A. Bellver-Sanchis,+ J. Companys-Alemany, Prof. M. Pallàs, Dr. C. Griñán- [d] Dr. P. A. Ávila-López Ferré Department of Biochemistry and Molecular Genetics Pharmacology Section, Department of Pharmacology, Toxicology and Feinberg School of Medicine Therapeutic Chemistry Northwestern University, Chicago, Illinois (USA) Faculty of Pharmacy and Food Sciences [e] Dr. A. L. Martínez Rodríguez, Dr. J. M. Brea Floriani Institute of Neuroscience, Universitat de Barcelona (NeuroUB) Centro de Investigación en Medicina Molecular y Enfermedades Crónicas Av. Joan XXIII 27–31, 08028 Barcelona (Spain) (CIMUS) E-mail: christian.grinan@ub.edu Departamento de Farmacología, Farmacia y Tecnología Farmacéutica [b] B. Singh Choudhary,+ Sukanya, Dr. R. Malik Universidad de Santiago de Compostela Department of Pharmacy 15782 Santiago de Compostela (Spain) Central University of Rajasthan [f] Dr. B. Pérez Bandarsindari, Ajmer, 305817 (India) Department of Pharmacology, Therapeutic and Toxicology [c] B. Singh Choudhary+ Universitat Autònoma de Barcelona, 08193 Barcelona (Spain) Department of Pharmaceutical Chemistry and Quality Assurance [+] These authors contributed equally to this work. Shree S. K. Patel College of Pharmaceutical Education and Research Supporting information for this article is available on the WWW under Ganpat University https://doi.org/10.1002/cmdc.202200002 Mehsana, Gujarat 384012 (India) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. ChemMedChem 2022, 17, e202200002 (1 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 In particular, G9a methyltransferase, also known as euchro- (determined by dividing the half-maximal effective concentra- matic histone methyltransferase 2 (EHMT2), is an HMT respon- tion (EC50) value of observed toxicity by the half-maximal sible for mono- and di-methylate lysine 9 on histone H3 inhibitory concentration (IC50) value of the functional potency), (H3 K9me1 and H3 K9me2).[7] Consequently, these repressive so to overcome its cytotoxicity, more specific and potent marks are capable of regulating gene expression.[8] Until now, molecules need to be discovered.[26,27] Crystallographic studies G9a has been widely explored as an anti-cancer and an anti- based on BIX01294 identified other inhibitory compounds to malarial target.[9–11] However, interestingly, recent evidence improve the therapeutic range. Further optimization of potentially suggests that G9a is a target in neurodegenerative UNC0638 led to potent and selective G9a inhibitors with a disorders, such as Alzheimer’s disease (AD), since its epigenetic better toxicity/function ratio than BIX01294 (> 100). However, marks are related to the repression of transcription in genes these two and the other published compounds were less involved in learning and memory formation, contributing to effective in in vitro and in vivo studies and had poor blood-brain cognitive impairment.[12–14] barrier (BBB) permeability and very high melting points. Thus, Small molecule inhibitors of G9a are valuable tools to study there is a need to explore new chemical scaffolds other than this protein’s biological functions and further explore the quinazolines and quinolines, which exhibit improved pharma- therapeutic hypotheses associated with them, for instance, in cokinetic properties. the context of neurodegeneration. Several G9a inhibitors have In brief, this work is based on the hypothesis that over- been synthesized and published in the last decade (Figure 1). expression of G9a and its repressive mark serve as a driver of Among the previously reported G9a inhibitors, quinazolines repression of genes involved in learning and memory. There- (1,2,3)[9,10] and quinolines (4)[15,16] are widely explored; others are fore, we explored an extensive chemical database (Chembridge benzodiazepines (5),[17] amino-indoles (6),[18] 6H-anthra[1,9- CNS MPO database) using structure-based virtual screening cd]isoxazol-6-one (7),[19] pseudodehydrocorydaline (8),[20] 2,4- against G9a, to provide potent inhibitors with novel chemical diamino-6-methylpyrimidines (9),[21] Chaetocin (10)[22–24] and scaffolds as a research tool to study the function of this protein Epidithiodiketopiperazine (11).[25] The first G9a inhibitor identi- in biological processes, mainly focused on AD. fied by high-throughput screening (HTS) was BIX01294 (Fig- ure 1), which resulted in active cell-based assay and reduced H3 K9methylation levels in several G9a target genes. However, this compound had a toxicity/function ratio lower than 6 Figure 1. Structures of previously reported G9a inhibitors (IC50 values are dependent on the assay conditions used). ChemMedChem 2022, 17, e202200002 (2 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 Results and Discussion Table 1. Dock score, MM-GBSA dG score, interacting residues, and G9a IC50 (nM) of selected screened molecules. Identification of novel G9a inhibitors through structure-based Compound Dock Score MM-GBSA dG G9a IC50 [nM] virtual screening UNC0638 5.049 70.31 9.4 A 10.346 106.68 > 100 Protein Data Bank (PDB) id 5TTF is selected for creating a B 12.553 116.98 9.4 docking model based on its 1.72 Å resolution, the sequence C 10.809 90.21 1.7 D 10.552 101.76 10.5 length of 283 amino acids, no Ramachandran outliers, and E 11.548 106.01 2.1 quinazoline derivative MS012 (Figure 1) as a co-crystallized F 10.202 95.02 1.8 compound. The crystal structure of G9a bound to MS012 (PDB G 10.487 93.42 5.8 H 10.331 96.43 2.5 ID: 5TTF) was retrieved from the RCSB portal and prepared I 11.253 91.44 3.9 using the protein preparation wizard of Maestro. The enrich- ment-based validation of the docking protocol was performed, showing the receiver operating characteristic (ROC) graph where docking results were plotted between valid positive rate reflects the probability of a higher ranking of randomly selected (sensitivity) and false-positive rate (1-specificity). The ROC graph active compounds over randomly selected decoy molecules. The results of Glide SP docking showed ROC values of 0.97 for 5TTF (Figure 2). The RMSD of co-crystallized and docked MS012 was found 1.165 Å (Figure S1a). A hierarchical stepwise virtual screening was performed using Glide HTVS, SP, and XP docking for the Chembridge database of 496644 compounds and XP docking of UNC0638 as a reference compound. The top 100 compounds were selected, based on docking scores from the central nervous system multiparameter optimization (CNS MPO) database of Chem- bridge Corporation, for further analysis. The docking scores of the top 100 compounds ranged from 13.592 kcal/mol to 10.105 kcal/mol. The Molecular Mechanics-Generalized Born Surface Area (MM-GBSA) dG binding affinity of these 100 compounds ranged from 116.98 kcal/mol to 51.09 kcal/mol. 9 molecules were selected based on the Glide XP dock score, MM-GBSA dG binding affinity, hydrogen bond mediated interaction with G9a active residues, i. e. Aspartate (Asp)1074, Asp1083, Leucine (Leu)1086, Asp1088, and hydrophobic inter- action with Tyrosine (Tyr)1154, and Phenylalanine Figure 2. ROC graphs of the docking results for assessing the reliability of (Phe)1158[9,28–30] and scaffold novelty (Figure 3a, 5, S5, Table 1 the docking protocol. and Table S1). These selected molecules were further subjected Figure 3. Structure of virtually screened and selected molecules. ChemMedChem 2022, 17, e202200002 (3 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 to Pan-assay interference compounds (PAINS) filter and all were In vivo efficacy studies in Caenorhabditis elegans (C. elegans) found to be non-PAINS.[31] Keeping in mind that our main goal in this work is to provide new and potent chemical scaffolds to treat AD pathology, we In vitro G9a inhibition assay proposed to use the C. elegans model to evaluate their effect on amyloid-β (Aβ)-mediated neurodegeneration. First, food The percentages of G9a activity for the 10 selected molecules consumption by the C. elegans strain N2 (Wild-Type) (N2 (WT)) were determined at 100 μM S-adenosylmethionine (SAM) and was assessed by the variation in the optical density (OD595) of peptide substrate using AlphaLISA® technology. These percen- bacteria over time as a measure of compounds toxicity, tages were used to calculate the IC50. As a control, we used the expecting a normal development and well-being of the animals well-established G9a inhibitor, UNC0638. The IC50 against G9a in each dose tested. Although UNC0638 is a published of the UNC0638 was 9.4 nM, in accordance with reported compound described as a non-toxic compound in in vivo studies that have described an IC50 < 15 nM. No significant studies, we also used it as a control in the assay (Figure S3). The inhibition was observed by A, and it was therefore considered OD595 of all compounds at different concentrations was not as not active (IC50 > 100 nM). Meanwhile, D presented higher different from the respective OD595 of the negative control in a IC50 in comparison with either UNC0638 or the other com- statistically significant manner (Figure S3). Thus, all of them pounds (IC50 = 10.5 nM), and therefore these compounds were were classified as safe, as the OD595 decrease was parallel to excluded for further studies. Compound B reached an IC50 equal that of the negative control samples, and visual inspection to that of UNC0638 (IC50 = 9.4 nM). The rest of compounds were confirmed the average growth of the animals. at least 4-fold more potent as G9a inhibitors than UNC0638 (IC50 C = 1.7 nM; IC50 E = 2.1 nM; IC50 F = 1.8 nM; IC50 G = 5.8 nM; IC50 H = 2.5 nM; IC50 I = 3.9 nM) (Table 1, Figure S2 and Table S2). Locomotion assay One of the best-characterized transgenic AD strains in C. elegans In silico ADMET analysis is the CL2006 strain, which contains the Punc-54::Aβ1–42 trans- gene, showing a progressive adult-onset paralysis. Then, to Moreover, physicochemical descriptors and pharmaceutically establish a dose-response profile, the pharmacological impact relevant properties of the selected compounds were calculated of G9a inhibition was evaluated in a locomotion assay with this with Schrödinger QikProp,[32] an ADME program useful for C. elegans strain. We performed a motility assay where animals predicting the CNS-relevant drug properties. For instance, CNS were placed in the center of a seeded plate and allowed to activity was predicted using a scale, with the highest activity for explore the environment freely. After 1 min, the animals that those that scored + 2, as was the case for E, F, G, and I; did not cross 1 cm circumference were scored as motor otherwise, the lowest activity was presented by B, D, with 0 as defective. N2 (WT), and CL2006 strain were treated with their score. Aqueous solubility (logS) could significantly affect dimethyl sulfoxide (DMSO) 1 %, used as a vehicle, which should the compounds’ absorption and distribution. All selected not cause a relevant locomotor defective phenotype. As compounds presented a similar value, fitting the recommended expected, the percentage of locomotion defects in N2 (WT) range ( 6.5–0.5). Other properties calculated were total solvent animals was the lowest (~ 16 %); whereas, as a result of the age- accessible surface area (SASA), octanol/water partition coeffi- dependent paralysis, CL2006 strain showed a significant cient (logPo/w), apparent Caco-2 cell permeability (PCaco), percentage of defects in their motility (~ 60 %) (Figure S4a), brain/blood partition coefficient, apparent MDCK cell perme- verifying that locomotor behavior is statistically different ability (PMDCK), skin permeability (logKp), binding to human between the two strains of C. elegans allowing us to study the serum albumin (logKhsa), Van der Waals surface area of polar impact of G9a inhibitors after chronic treatment. The motor atoms (PSA) and the number of Lipinski’s rule of five violations. index for each condition was calculated using the motor Moreover, human oral absorption (HOA) and the percent of performance of Bristol (N2) (WT, 100 %) and the transgenic AD HOA (%HOA) were predicted, resulting high in all of them, strain (CL2006, 0 %) animals as reference. The well-establishing although the %HOA of E was not strictly superior to 80 %. inhibitor, UNC0638, was tested and used as a control compared Additionally, hydrogen bond acceptor and donor (HBA, and to the other selected compounds (Figure 4). HBD, respectively). A total of 3 compounds A (HBA = 9.2), G By correlating in silico and in vitro G9a activity results, it is (HBA = 8.5), and I (HBA = 8.5) were not able to pass the proved that interactions mediated by HBD, water bridge (WB), qualifying limit for the number of HBA. UNC0638 is can also not salt bridge (SB), and π-cation (π-cat) with the amino acid pass the qualifying limit for SASA, logKhsa, and Lipinski rule of residues Asp1074, Asp1083, Asp 1084, Leu1086, Asp1088, and 5. To balance these properties, 6 compounds that showed CNS Tyr1154 are essential for the potent activity of G9a inhibitors. drug properties were considered potential hit compounds as The pyrrolidine, morpholine, oxazepane, piperidine, piperazine, drug candidates (Table S3). imidazole, pyridine, pyrimidine, indole, etc. moieties were present in the screened moiety. Nitrogen atoms present in these moieties interacted with G9a through a hydrogen bond, salt bridge, and π-cation interactions. These moieties interacted ChemMedChem 2022, 17, e202200002 (4 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 amide group interacted with water bridge mediated hydrogen bond donor interaction with Asp1083 and Ser1084, NH of indole ring interacted through hydrogen bond donor inter- action with Asp1088 of F; such interaction is not present in other compounds, and possibly because of this interaction, F is the most potent compound with an IC50 of 1.8 nM (Figure 5c, Figure S2e). In the same way, the motor index of F decreased progressively with the doses tested. Effects of the best G9a inhibitor candidates on the repressive mark H3 K9me2 Considering the results described heretofore, we selected four compounds to go deeper into their effects on AD pathology. Firstly, to demonstrate that these compounds directly target G9a, we quantified the levels of its repressive mark H3 K9me2 after each drug treatment (Figure 6). First, we compared the ratio H3 K9me2/H3 total levels in the N2 (WT) strain compared to the AD transgenic strain (CL2006). Interestingly, the CL2006 presented higher levels of H3 K9me2, confirming the link between this repressive mark and AD pathology. Regarding each compound, we observed that the ratio H3 K9me2/H3 total levels after each drug intervention were reduced, suggesting that our compounds target G9a directly (Figure 6b). Figure 4. a-d) Treatment with G9a inhibitors ameliorates motor dysfunction of CL2006 worms. For dose-response: Values represented are mean � Inhibition of Aβ aggregation in C. elegans Standard error of the mean (SEM); n = 3 with at least 90–100 worms in each group. Statistical analysis: One-Way ANOVA, followed by Dunnett post-hoc analysis. 0 μM Vs. UNC0638 treatment: ***p < 0.001; ****p < 0.0001. 0 μM Vs. Furthermore, we evaluated the impact after the pharmacolog- new G9a inhibitor treatment: #p < 0.05; ##p < 0.01; ###p < 0.001; ical inhibition of G9a treatment on the deposition of Aβ ####p < 0.0001. aggregates, as an important hallmark of AD, using Thioflavin-S (ThS) staining. The transgenic AD strain CL2006 is characterized by the expression of human Aβ1-42 under the control of a with Asp1083, Asp1084, Asp1086, and Asp1088 (Figure 5, Fig- muscle-specific promoter, making it suitable to answer this ure S5). Although some of these interactions were present in objective in our work. As we mentioned above, UNC0638 was compound B, the dose-response profile did not show an used as a positive control. Thus, as expected, UNC0638 reduced improved performance against the motor indexes of the control the Aβ deposition by approximately 25 %. For the four UNC0638 (Figure S4b). Moreover, as aforementioned, we proposed candidates from our tested library, all of them described that both IC50 were very similar, and it did not present showed the same tendency but in a statistically different way. better properties among the selected compounds, such as the Similarly, in UNC0638, the reduction in the number of deposits CNS activity, justifying its exclusion from the study. NH of amide by compound E was approximately 25 %. F reduced Aβ and 2° amine group in C, F, G, H, and I are equally important aggregation by approximately 40 % compared to the vehicle and responsible for interaction with Asp1083, Asp1086, and group, being more effective in reducing Aβ aggregation than Asp1088. In parallel fashion, E showed SB interaction with the UNC0638. Unfortunately, treatment with H could not alleviate Asp1088, WB interaction with the Asp1083, and HBD interaction the aggregation of Aβ plaques (3 %), although previous in silico with Leu1086 being essential interactions for potent inhibitory and in vitro results were promising (Figure 7). activity of G9a. The dose-response profile of I showed the lowest motor index in the maximum and minimum dose tested, and in the case of G the motor index decreased progressively as In vitro BBB penetration study the dose decreased (Figure S4c,d). However, due to G and I not passing the qualifying limit of all of the CNS drug properties Then, we performed the parallel artificial membrane permeation described above, these compounds were discarded. C, E, F, and assay (PAMPA) to predict BBB penetration (PAMPA-BBB), H motor index decreased progressively with the doses tested, obtaining important information for compound selection and ameliorating the paralysis presented by CL2006. Moreover, the synthesis guidance for desirable CNS properties. The selected motor indexes of each dose were not statistically different from compounds E, F, and H had in vitro permeability (Pe) values of the profile of the control UNC0638 (Figure 4a–d). The NH of 2.2 × 10 6 � 0.1 cm · s 1, 10.9 × 10 6 � 0.4 cm · s 1 and 17.4 × 10 6 ChemMedChem 2022, 17, e202200002 (5 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 Figure 5. 3D interactions diagram of a) E G9a, b) F G9a, and c) H G9a. Protein-ligand Interactions were denoted by a dotted line. Hydrogen bond: Yellow; Salt bridge: Pink; π-π stacking: blue; π-cation: Green. Table 2. Permeability (Pe 10 6 cm · s 1) in the PAMPA-BBB assay of 14 � 2.4 cm · s 1, respectively (Table 2). The compound E had lower commercial drugs and tested compounds and predictive penetration in the Pe compromising its permeability, being lower than the Pe value CNS. of the control UNC0638 (PAMPA BBB Pe = 6.2 × 10 6 cm · s 1). Compound Bibliography Compound Bibliography Nevertheless, the compounds F and H presented positive BBB value[33] value[33] permeation, being able to cross the BBB and reach their Verapamil 16.0 25.9 � 0.8 biological target in the CNS. Therefore, the compound F is a Testosterone 17.0 26.6 � 0.9 Corticosterone 5.1 6.7 � 0.1 promising tool for in vivo AD studies with mammalian models. Clonidine 5.3 6.5 � 0.06 Ofloxacin 0.8 0.1 � 0.09 Lomefloxacin 0.0 0.8 � 0.05 Progesterone 9.3 14.7 � 0.3 Evaluation of EHMT1 (GLP) activity by luminescence assay Promazine 8.8 13.8 � 0.4 Imipramine 13.0 16.8 � 0.3 Last, but not least, further characterization of the best Hydrocortisone 1.9 1.4 � 0.06 Piroxicam 2.5 2.0 � 0.08 candidates was performed assessing their selectivity. Thus, we Desipramine 12.0 17.8 � 0.2 tested them against EHMT1 (G9a-like protein, GLP). Most Cimetidine 0.0 0.7 � 0.02 remarkably, all of the new G9a inhibitors showed very low, Norfloxacin 0.1 8.8 � 0.4 E 2.2 � 0.1 CNS + /- almost non-existent, specificity against EHMT1 compared to the F 10.9 � 0.4 CNS + well-established UNC0638 (Table 3). We found that E did not H 17.4 � 2.4 CNS + inhibit EHMT1 from to 0.01 μM 10 μM. In contrast, compounds F and H showed slight activity against EHMT1 at low doses. ChemMedChem 2022, 17, e202200002 (6 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 Figure 6. a-b) Representation of western blotting and quantification of the total H3 K9me/H3 ratio after treatments with the best candidates to inhibit G9a. Values represented are mean � Standard error of the mean (SEM); n = 3 with at least 200 worms in each group. Statistical analysis: One-Way ANOVA, followed by Dunnett post-hoc analysis: 0 μM Vs. UNC0638, E, and F treatment: *p < 0.05;**p < 0.01. Statistical analysis: Unpaired t-test: N2 (WT) Vs. 0 μM: *p < 0.05; 0 μM Vs. H: *p < 0.05. Figure 7. a) Representative images from each group tested. b) Quantification of Thioflavin S-positive particles in the head region of CL2006 strain. Values represented are mean � Standard error of the mean (SEM); n = 3 with 25 worms in each group. Statistical analysis: One-Way ANOVA, followed by Dunnett post-hoc analysis. Vehicle Vs. new G9a inhibitor treatment: **p < 0.001; ***p < 0.001; ****p < 0.0001. Table 3. Activity against of EHMT1 (GLP) measured by luminescence assay. Conclusion % Inhibition of EHMT1 Compound 0.01 μM 0.1 μM 1 μM 10 μM To sum up, we have identified new G9a inhibitors through in UNC0638 2_ � 1 41_ � 4 66_ � 4 84_ � 1 silico and in vitro screens, with different scaffolds to those E 2_ � 1 4_ � 2 3_ � 3 8_ � 3 F 26_ � 5 14_ � 3 4_ � 2 1_ � 2 previously published inhibitors, and more potent. E, F, and H H 14_ � 1 7_ � 1 4_ � 1 2_ � 1 were further studied in in vivo assays. Interestingly, our selected compounds were able to reduce H3 K9me2 levels similar to UNC0638, suggesting the G9a is the direct target for them. On the one hand, the effect on the impact of Aβ aggregates However, they presented a very low percentage, and further- reduction was similar for UNC0638 and E. Also, this compound more this trend was not observed at high doses. showed promising pharmacokinetics results; however, its pre- dicted BBB-permeability was lower than UNC0638, possibly because of decreased lipophilicity due to lower logPo/w, logBB ChemMedChem 2022, 17, e202200002 (7 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 value, and higher PSA (Table S3). On the other hand, H did not for study resulting in basic physicochemical properties and in silico show a positive effect related to Aβ deposition, albeit the in pharmacokinetic for CNS drugability.[32,40] silico and in vitro results were promising, achieving the best predicted BBB permeability to date published for a G9a Compound preparation inhibitor. Nevertheless, it is noteworthy that compound F reduced Aβ deposition by approximately 40 %, thereby suggest- G9a inhibitors were serially diluted between 5 nM and 0.001 nM in 100 % DMSO (Sigma, St. Louis, USA). Then, respective concentra- ing a promising role for this compound in AD pathology. tions were subsequently diluted in MilliQ purified water to reach a Strikingly, F has an IC50 = 1.8 nM (Figure S2e), and none of the final concentration ranging between 50 and 0.001 μM in 1 % DMSO previously reported G9a inhibitors have such nanomolar in well. efficacy. Besides, in comparison to the well-established G9a inhibitor (UNC0638), F exhibited a higher permeability to BBB, G9a homogenous assay as well as a higher selectivity for G9a, as it exhibited low activity against EHMT1. Therefore, these outcomes highlight an impor- G9a activity was measured using the G9a Homogeneous Assay Kit tant difference and reinforce the potential use in neurodegener- (Catalog #52051, BPS Bioscience, San Diego, CA, USA). The G9a Homogeneous Assay Kit was set up in AlphaLISA® format on an ative diseases of these novel G9a inhibitors. Hereinafter, future optiplate-384-wells. Briefly, a sample containing the G9a enzyme is perspective will be to address the biological activity of our best incubated with the biotinylated substrate plus inhibitor solution for candidate, F (5-chloro-N-((3S,4S)-1-ethyl-4-methoxypyrrolidin-3- two hours. Next, acceptor beads and primary antibody is added, yl)-1H-indole-2-carboxamide) and its biological differentiation followed by donor beads. Finally, the Alpha-counts were read using from other known G9a inhibitors. Then, our study will help an AlphaScreen microplate reader. The experiment was run per future research establish the structure-activity relationship (SAR) triplicate. of F analogs to G9a, improving potency and BBB permeability as well as reducing one of the main AD hallmarks the Aβ G9a activity ð%Þ ¼ ðC CbÞ=ðCt- CbÞ x 100 plaques. Ct = The chemiluminescence reading in the absence of the compound. The data set here is defined as 100 % activity in the absence of the compound. Experimental Section Cb = The chemiluminescence reading in absence of enzyme. The data set here is defined as 0 % activity in the absence of the Structure-based virtual screening enzyme. PDB id of G9a for developing docking model was selected based on C = The chemiluminescence reading in the presence of the resolution, sequence length, PDB validation, and co-crystallized compound. molecule. The crystal structure of G9a bound to MS012 (Figure 1) (PDB ID: 5TTF) was retrieved from the RCSB portal.[34] Refinement of protein was performed using a protein preparation wizard. The C. elegans strains, and maintenance active site was defined around the co-crystallized MS012 using the receptor grid generation module of Glide.[35] The protocol resulting The WT C. elegans strain (N2), and the transgenic CL2006 (dvIs2 from a crystal structure (PDB ID: 5TTF) was validated using receiver [pCL12(unc-54/human Aβ peptide 1–42 minigene) + rol-6(su1006)]) operator characteristic (ROC) metrics and docking of co-crystallized provided by the C. elegans Genetic Center were used. Standard MS012 into the G9a (5TTF). The ROC analysis actives dataset was methods were used for culturing and observing C. elegans. N2 were merged with the Schrodinger decoy set containing 1000 drug-like propagated at 20 °C, while CL2006 worms were maintained at 16 °C decoy compounds.[36,37] The combined dataset was docked into the in a temperature-controlled incubator on solid nematode growth protein using the SP docking method in glide. ROC plots were medium (NGM) seeded with Escherichia coli (E. coli) OP50 strain as plotted using Enrichment Calculator in Maestro.[38] A Chembridge a food source. CNS MPO library of 496.644 compounds was downloaded from the Chembridge portal were prepared with LigPrep using OPLS 2005 forcefield.[35] The prepared Chembridge CNS MPO library was Food clearance assay screened using virtual screening workflow docking protocols (HTVS, N2 (WT) worms were subjected to a drug assay in liquid format. SP, and XP docking) of the Schrodinger software package on the Worms were grown with continuous shaking at 180 rpm at 20 °C for selected crystal structure (PDB ID- 5TTF) carrying forward 10 % each 7 days. Each well contained a final volume of 60 μL, comprising 25– from HTVS to SP and SP to XP and top 100 kept in XP docking. 30 animals in the larva 1 (L1) stage diluted in S-medium solution, These selected top 100 molecules in XP docking were subjected to G9a inhibitor at the appropriate dose, and OP50 inactivated by the binding affinity (dG) calculation using the Prime-MM-GBSA freeze-thaw cycles suspended in S-medium complete solution to a method. From these top 100 molecules, 9 molecules were selected final OD595 of 0.7 measured in the microplate reader. Tested based on the Glide XP docking score, dG of binding, nature of concentrations ranged between 1 μM to 0.001 μM. For control wells interaction with G9a active site key residues (Asp1074, Asp1083, were used DMSO 1 %(vehicle) and DMSO 5 % (toxic condition); and and Asp1088),[29,34,39] and scaffold novelty. for blank wells were used S-medium and S-medium complete only, without eggs or OP50, respectively. The effect of compounds on C. elegans physiology was monitored by the rate at which the OP50 In silico analysis of CNS drugability suspension was consumed, as a readout for C. elegans growth, QikProp module of the Schrodinger software package was used for survival, or fecundity. The OD595 was measured daily. Assays were determining the drug properties. The 9 best scoring hits were taken run in triplicates (n = 3), with a total of at least 150 animals tested per compound concentration. ChemMedChem 2022, 17, e202200002 (8 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 Locomotion assay (1 : 1000) (Epigentek, #A-4035), and anti-H3 total (1 : 1000) (Cell signaling, #9715). Locomotion assay of G9a inhibitors was assessed to obtain dose- response profile, evaluating the impact of the pharmacology treatment in motor dysfunction presented by the transgenic strains PAMPA-BBB assay CL2006. Nematodes were synchronized by alkaline hypochlorite treatment, and drug assay was performed as described above, To evaluate the brain penetration of the different compounds, a except for the OP50 bacteria OD595 adjusted to 0.9, and the parallel artificial membrane permeation assay for BBB was used, duration of the treatment was 5 days. N2 (WT), and the transgenic following the method described by Di et al..[33] The in vitro Pe of strain CL2006 were treated in liquid media for 4 days at 20 °C and fourteen commercial drugs through lipid extract of porcine brain 180 rpm, starting at the L1 stage. On day 5 of age, worms were membrane together with the test compounds were determined. transferred from the 96-well plates onto an unseeded NGM plate Commercial drugs and assayed compounds were tested using a for 45 minutes before starting the trial, allowing the plates to dry mixture of PBS:EtOH (70 : 30). Assay validation was made by and worms. Locomotion assays were performed in 30 mm NGM comparing the experimental permeability with the reported values plates, in which the whole surface of the plate was covered by of the commercial drugs by bibliography, and linear correlation OP50. 5 to 10 adult nematodes were placed in the center of a circle between experimental and reported permeability of the fourteen (with 1 cm of diameter) in the seeded 30 mm NGM plates. After commercial drugs using the parallel artificial membrane permeation 1 min, the number of animals remaining inside the circle was assay was evaluated (y = 1.572x–1.147; R2 = 0.9452). From this scored as a locomotor defect. Motor behavior assays were run in equation and taking into account the limits established by Di triplicates (n = 3), with a total of at least 100 animals tested per et al.[33] for BBB permeation, we established the ranges of perme- compound concentration. Motor index establishes a score where ability as compounds of high BBB permeation (CNS +): Pe treated animals showing the same motor defect as CL2006 have an (10 6 cm · s 1) > 5.14; compounds of low BBB permeation(CNS-): Pe index of 0 %. By contrast, when animals improve motor behavior (10 6 cm · s 1) < 1.99 and compounds of uncertain BBB comparable to the WT (N2) they get an index of 100 %. All the permeation(CNS + / ):5.14 Pe (10 6 cm · s 1) > 1.99. results shown in the figures are calculated using the following formula: EHMT1 (GLP) chemiluminescence assay kit Motor index ð%Þ ¼ ðLD CL2006vehicle LD CL2006drug Þ = The assay was carried out using EHMT1 (GLP) Chemiluminescent ðLD CL2006vehicle LD WTvehicle Þ � 100 Assay kit. Compounds were added on a 96-well plate precoated with histone substrate supplied by kit. 30 μL of master mixture SAM LD ¼ locomotion defective were added and the reaction was initiated adding 20 μL of diluted EHMT1 (GLP) enzyme. The reaction was incubated at room temper- ature (RT) for 1 hour. The plate was washed three times with 200 μl TBST buffer. 100 μL of Blocking buffer were added to every well Thioflavin-S staining Aβ aggregation and the plate was incubated at RT for 10 minutes with shaking. Supernatant was removed and 100 μl of Primary antibody were After 5 days of treatment, adult CL2006 C. elegans were fixed in 4 % added per well. The plate was incubated at RT for 1 hour with slow Paraformaldehyde/Phosphate-buffered saline (PBS) (pH 7.5), for shaking. The plate was washed three times with 200 μl TBST buffer. 24 hours at 4 °C. Then, worms were permeabilized in 5 % fresh β- 100 μL of Blocking buffer were added to every well and the plate mercaptoethanol, 1 % Triton X-100, 125 mm Tris (pH 7.5), at 37 °C was incubated at RT for 10 minutes with shaking. 100 μl of for another 24 hours. On the last day, nematodes were stained with Secondary HRP-labeled antibody were added per well and the plate 0.125 % Thioflavin-S (ThS) (Sigma, CAS# 1326–12-1) in 50 % ethanol was incubated at RT for 30 minutes with slow shaking. The plate (EtOH) for 2 min, destained in 50 % EtOH for 2 min, washed 3 times was washed three times with 200 μl TBST buffer. Finally, 100 μl of with PBS. To prepare the glass slide for microscopy, approximately substrate mix were added and the luminescence was measurement 10 μL volume was transferred on a droplet of Fluoromount G with integration time of 1 second in in Perkin Elmer EnSpire (Electron Microscopy Sciences, CAT#17984-25). Fluorescence im- Multimode plate reader. ages were acquired using a 20 Å ~ objective of a fluorescence microscope. Aβ in the head region of worms were quantified by counting the number of Th S positive spots using ImageJ and were Statistical analysis expressed as Aβ deposits/anterior area. Aβ aggregates were scored Data analysis was conducted using GraphPad Prism ver. 9 statistical by an investigator blinded to G9a inhibitors treatments. software. The statistical outliers were identified with Grubb’s test and excluded from the analysis. Group comparisons were carried Western blotting out with a One-way Analysis of variance (ANOVA), followed by the Dunnett post hoc test. In some cases, comparison between groups To determine the ratio of H3 K9me2/H3 total N2 (WT) and CL2006 was also performed by unpaired t-test or One-way ANOVA, animals were incubated in liquid culture with vehicle (DMSO 1 %), followed by Fisher’s LSD post hoc test. For the calculation of IC50, or 0.1 μM of G9a inhibitor in a 96-well plate format as already non-linear regressions were performed. Regarding food clearance mentioned. For chronic treatment, four-day-old animals were assay, results were compared non-linear regression model for collected with M9 buffer. Histone extraction was performed sigmoidal curves against the negative control. Values are shown as following the manufacturer’s instructions (EpiQuik Total Histone the mean � Standard error of the mean (SEM) of at least n = 3. Extraction HT Kit, EpiGentek, #OP-0007-192). The samples were Statistical significance was considered when p values were < 0.05. resolved in a 14 % SDS-gel, as previously described.[12] To capture chemiluminescence signals were used Amersham Imager 680 and Western blot quantifications were performed using ImageLab software (Bio-Rad). Immunoblots were probed with anti-H3 K9me2 ChemMedChem 2022, 17, e202200002 (9 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH Research Article ChemMedChem doi.org/10.1002/cmdc.202200002 Acknowledgements [16] O. Rabal, E. San José-Enériz, X. Agirre, J. A. Sánchez-Arias, I. de Miguel, R. Ordoñez, L. Garate, E. Miranda, E. Sáez, A. Vilas-Zornoza, J. Med. Chem. 2021, 64, 3392–3426. B.S.C. and S. are grateful to Council for Scientific and Industrial [17] C. Milite, A. Feoli, J. R. Horton, D. Rescigno, A. Cipriano, V. Pisapia, M. Research, India, for Senior Research Fellowship (Grant No. 09/ Viviano, G. Pepe, G. Amendola, E. Novellino, S. Cosconati, X. Cheng, S. Castellano, G. Sbardella, J. Med. Chem. 2019, 62, 2666–2689. 1131(0004)/2016-EMR-I and 09/1131(0026)/2019-EMR-I) and Cen- [18] R. F. Sweis, M. Pliushchev, P. J. Brown, J. Guo, F. Li, D. Maag, A. M. Petros, tral University of Rajasthan for providing computational facilities. N. B. Soni, C. Tse, M. Vedadi, ACS Med. Chem. Lett. 2014, 5, 205–209. A.B-S acknowledge the Agència de Gestió d’Ajuts Universitaris i de [19] W.-L. Chen, Z.-H. Wang, T.-T. Feng, D.-D. Li, C.-H. Wang, X.-L. Xu, X.-J. Zhang, Q.-D. You, X.-K. Guo, Bioorg. Med. Chem. 2016, 24, 6102–6108. Recerca (AGAUR) for her FI-SDUR fellowship (2021 FISDU 00182). [20] J. Chen, X. Lin, K. J. Park, K. R. Lee, H.-J. Park, J. Comput.-Aided Mol. Des. J.C-A acknowledge the Ministry of Economy and Competitiveness 2018, 32, 917–928. (MINECO) for her FPI fellowship (MDM-2017-0729-19-2). C. elegans [21] K. Katayama, K. Ishii, E. Tsuda, K. Yotsumoto, K. Hiramoto, M. Suzuki, I. Yasumatsu, W. Igarashi, M. Torihata, T. Ishiyama, Bioorg. Med. Chem. Lett. strains were provided by the Caenorhabditis Genetics Center 2020, 30, 127475. (CGC), which is funded by NIH Office of Research Infrastructure [22] D. Greiner, T. Bonaldi, R. Eskeland, E. Roemer, A. Imhof, Nat. Chem. Biol. Programs (P40 OD010440). We thank Gemma Triola from the 2005, 1, 143–145. [23] E. Iwasa, Y. Hamashima, S. Fujishiro, E. Higuchi, A. Ito, M. Yoshida, M. Institute of Advanced Chemistry of Catalonia (IQAC-CSIC) and her Sodeoka, J. Am. Chem. Soc. 2010, 132, 4078–4079. lab for offering the possibility of using the AlphaScreen microplate [24] F. L. Cherblanc, K. L. Chapman, J. Reid, A. J. Borg, S. Sundriyal, L. Alcazar- reader. Fuoli, E. Bignell, M. Demetriades, C. J. Schofield, P. A. DiMaggio Jr, J. Med. Chem. 2013, 56, 8616–8625. [25] S. Fujishiro, K. Dodo, E. Iwasa, Y. Teng, Y. Sohtome, Y. Hamashima, A. Ito, M. Yoshida, M. Sodeoka, Bioorg. Med. Chem. Lett. 2013, 23, 733–736. Conflict of Interest [26] F. Liu, X. Chen, A. Allali-Hassani, A. M. Quinn, G. A. Wasney, A. Dong, D. Barsyte, I. Kozieradzki, G. Senisterra, I. Chau, A. Siarheyeva, D. B. Kireev, A. Jadhav, J. M. Herold, S. V. Frye, C. H. Arrowsmith, P. J. Brown, A. The authors declare no conflict of interest. Simeonov, M. Vedadi, J. Jin, J. Med. Chem. 2009, 52, 7950–7953. [27] Y. Chang, T. Ganesh, J. R. Horton, A. Spannhoff, J. Liu, A. Sun, X. Zhang, M. T. Bedford, Y. Shinkai, J. P. Snyder, X. Cheng, J. Mol. Biol. 2010, 400, 1–7. Data Availability Statement [28] F. Liu, X. Chen, A. Allali-Hassani, A. M. Quinn, G. A. Wasney, A. Dong, D. Barsyte, I. Kozieradzki, G. Senisterra, I. Chau, A. Siarheyeva, D. B. Kireev, A. Jadhav, J. M. Herold, S. V. Frye, C. H. Arrowsmith, P. J. Brown, A. The data that support the findings of this study are available Simeonov, M. Vedadi, J. Jin, J. Med. Chem. 2009, 52, DOI 10.1021/ from the corresponding author upon reasonable request. jm901543 m. [29] M. Vedadi, D. Barsyte-Lovejoy, F. Liu, S. Rival-Gervier, A. Allali-Hassani, V. Labrie, T. J. Wigle, P. A. DiMaggio, G. A. Wasney, A. Siarheyeva, A. Dong, Keywords: Alzheimer’s disease · epigenetics · G9a W. Tempel, S.-C. Wang, X. Chen, I. Chau, T. J. Mangano, X. Huang, C. D. methyltransferase · amyloid-β · structure based virtual Simpson, S. G. Pattenden, J. L. Norris, D. B. Kireev, A. Tripathy, A. Edwards, B. L. Roth, W. P. Janzen, B. A. Garcia, A. Petronis, J. Ellis, P. J. screening Brown, S. V. Frye, C. H. Arrowsmith, J. Jin, Nat. Chem. Biol. 2011, 7, DOI 10.1038/nchembio.599. [30] M. Ramya Chandar Charles, H.-P. Hsieh, M. Selvaraj Coumar, J. Biomol. [1] Z. Zhao, A. Shilatifard, Genome Biol. 2019, 20, 1–16. Struct. Dyn. 2019, 37, DOI 10.1080/07391102.2018.1491422. [2] P. J. Hamilton, E. J. Nestler, Curr. Opin. Neurobiol. 2019, 59, 128–136. [31] J. B. Baell, G. A. Holloway, J. Med. Chem. 2010, 53, 2719–2740. [3] J. Richetto, U. Meyer, Biol. Psychiatry 2020. [32] A. K. Ghose, T. Herbertz, R. L. Hudkins, B. D. Dorsey, J. P. Mallamo, ACS [4] A. Berson, R. Nativio, S. L. Berger, N. M. Bonini, Trends Neurosci. 2018, 41, Chem. Neurosci. 2012, 3, DOI 10.1021/cn200100 h. 587–598. [33] L. Di, E. H. Kerns, K. Fan, O. J. McConnell, G. T. Carter, Eur. J. Med. Chem. [5] A. Portela, M. Esteller, Nat. Biotechnol. 2010, 28, 1057–1068. 2003, 38, 223–232. [6] K. P. Bhat, H. Ü. Kaniskan, J. Jin, O. Gozani, Nat. Rev. Drug Discovery [34] Y. Xiong, F. Li, N. Babault, A. Dong, H. Zeng, H. Wu, X. Chen, C. H. 2021, 20, 265–286. Arrowsmith, P. J. Brown, J. Liu, J. Med. Chem. 2017, 60, 1876–1891. [7] M. Tachibana, J. Ueda, M. Fukuda, N. Takeda, T. Ohta, H. Iwanari, T. [35] G. Madhavi Sastry, M. Adzhigirey, T. Day, R. Annabhimoju, W. Sherman, Sakihama, T. Kodama, T. Hamakubo, Y. Shinkai, Genes Dev. 2005, 19, J. Comput.-Aided Mol. Des. 2013, 27, DOI 10.1007/s10822-013-9644–8. 815–826. [36] R. A. Friesner, J. L. Banks, R. B. Murphy, T. A. Halgren, J. J. Klicic, D. T. [8] S. R. Shankar, A. G. Bahirvani, V. K. Rao, N. Bharathy, J. R. Ow, R. Taneja, Mainz, M. P. Repasky, E. H. Knoll, M. Shelley, J. K. Perry, D. E. Shaw, P. Epigenetics 2013, 8, 16–22. Francis, P. S. Shenkin, J. Med. Chem. 2004, 47, DOI 10.1021/jm0306430. [9] M. R. C. Charles, A. Dhayalan, H.-P. Hsieh, M. S. Coumar, Future Med. [37] T. A. Halgren, R. B. Murphy, R. A. Friesner, H. S. Beard, L. L. Frye, W. T. Chem. 2019, 11, 993–1014. Pollard, J. L. Banks, J. Med. Chem. 2004, 47, DOI 10.1021/jm030644 s. [10] H. Cao, L. Li, D. Yang, L. Zeng, X. Yewei, B. Yu, G. Liao, J. Chen, Eur. J. [38] J.-F. Truchon, C. I. Bayly, J. Chem. Inf. Model. 2007, 47, DOI 10.1021/ Med. Chem. 2019, 179, 537–546. ci600426e. [11] C. Poulard, L. M. Noureddine, L. Pruvost, M. Le Romancer, Life 2021, 11, [39] F. Liu, X. Chen, A. Allali-Hassani, A. M. Quinn, G. A. Wasney, A. Dong, D. 1082. Barsyte, I. Kozieradzki, G. Senisterra, I. Chau, J. Med. Chem. 2009, 52, [12] C. Griñán-Ferré, L. Marsal-García, A. Bellver-Sanchis, S. M. Kondengaden, 7950–7953. R. C. Turga, S. Vázquez, M. Pallàs, Aging (Albany NY) 2019, 11, 11591– [40] R. Malik, B. S. Choudhary, S. Srivastava, P. Mehta, M. Sharma, J. Biomol. 11608. Struct. Dyn. 2017, 35, DOI 10.1080/07391102.2016.1253503. [13] Y. Zheng, A. Liu, Z. J. Wang, Q. Cao, W. Wang, L. Lin, K. Ma, F. Zhang, J. Wei, E. Matas, J. Cheng, G. J. Chen, X. Wang, Z. Yan, Brain 2019, 142, 787–807. [14] J. L. T. Han, K. K. L. Pang, S. R. X. Ang, M. Sharma, S. Sajikumar, Transl. Neurodegener. 2021, 10, 1–3. Manuscript received: January 1, 2022 [15] O. Rabal, J. A. Sánchez-Arias, E. San Jose-Eneriz, X. Agirre, I. de Miguel, L. Revised manuscript received: April 11, 2022 Garate, E. Miranda, E. Saez, S. Roa, J. A. Martínez-Climent, J. Med. Chem. Accepted manuscript online: April 12, 2022 2018, 61, 6546–6573. Version of record online: May 19, 2022 ChemMedChem 2022, 17, e202200002 (10 of 10) © 2022 The Authors. ChemMedChem published by Wiley-VCH GmbH