Design, synthesis, and biological evaluation of novel carbazole derivatives as potent DNMT1 inhibitors with reasonable PK properties

Abstract The DNA methyltransferases (DNMTs) were found in mammals to maintain DNA methylation. Among them, DNMT1 was the first identified, and it is an attractive target for tumour chemotherapy. DC_05 and DC_517 have been reported in our previous work, which is non-nucleoside DNMT1 inhibitor with low micromolar IC50 values and significant selectivity towards other S-adenosyl-L-methionine (SAM)-dependent protein methyltransferases. In this study, through a process of similarity-based analog searching, a series of DNMT1 inhibitors were designed, synthesized, and evaluated as anticancer agents. SAR studies were conducted based on enzymatic assays. And most of the compounds showed strong inhibitory activity on human DNMT1, especially WK-23 displayed a good inhibitory effect on human DNMT1 with an IC50 value of 5.0 µM. Importantly, the pharmacokinetic (PK) profile of WK-23 was obtained with quite satisfying oral bioavailability and elimination half-life. Taken together, WK-23 is worth developing as DNMT1-selective therapy for the treatment of malignant tumour.


Introduction
Epigenetic modification, like DNA methylation, plays a major role in the expression of genetic information. Methylation of DNA at C-5 of cytosine is one of the most studied modifications of the mammalian genome 1 . DNMTs (DNA methyltransferases), which include DNMT1, DNMT3A, and DNMT3B, have been identified in humans 2 . DNMT1 is the most abundant among the three and is responsible for the maintenance of CpG methylation patterns in mammals with hemimethylated CpG dinucleotides serving as preferred substrates 3 . DNMT1 plays a significant role in the structural modification of chromosomes and the regulation of gene expression 1 . The methylation of the 5-carbon on cytosine residues (5mC) in CpG dinucleotides was the first described covalent modification of DNA and is one of the most extensively characterised modifications of chromatin, and thus, DNMT inhibitors have become useful tools for treating cancers 4 .
So far, two types of DNMT1 inhibitors have been thus found, namely nucleoside analogs and non-nucleoside analogs. Nucleoside analogs, such as the U.S. Food and Drug Administration (FDA) approved 5-azacytidine and 5-aza-2 0 -deoxycytidine 5 , or 2-pyrimidone-1-b-D -riboside (zebularine) 6 , or the dinucleotide derivative SGI-110 7 , exert their effects by incorporation into DNA inducing substantial DNA methylation inhibition and reactivation of hypermethylated genes. However, these drugs are unstable, show low specificity, and have obvious toxic side effects 8 . Therefore, specific concern has been given about nonnucleosides. As shown in Figure 1, various non-nucleoside analogs have been reported, including the followed natural compounds, such as genistein 9 , (-)-epigallocatechin 3-O-gallate (EGCG) 10 , and curcumin 11 ; repurposed drugs, such as the antihypertensive drug hydralazine 12 , procainamide 13 , and 7b 14 ; novel inhibitors, such as phthalimido-L -tryptophan (RG108) 15 , the quinolone derivative SGI-1027 16 , and the recently reported selective DNMT1 inhibitor GSK3482364 17 , and three small molecules identify from a DNMT focussed library including CSC027480404, CSC026286840, and CSC027694519 18 . Another recent research identified two 3bromo-3-nitroflavanones 3b and 4a as potent non-nucleoside DNMT inhibitors with good activity and stability targeting DNA methylation 19 . Compared with nucleoside analogs, non-nucleoside analogs are less likely to be incorporated into DNA and hence provide a relatively safe method to target DNA methylation.
There are two non-nucleoside analogs DC_05 and DC_517 have been reported as specific and highly potent inhibitors of DNMT1 via biochemical and cellular assays 20 . With an aim to improve PK properties, and decreased the side effect and toxicity, we designed and synthesised a series of derivatives based on the structures of DC_05 and DC_517. According to the DNMT1 enzyme inhibition assays, we obtained WK-23 as a specific DNMT1 inhibitor with excellent inhibitory activity (IC 50 ¼ 5.0 mM). In the further in vivo pharmacokinetic (PK) study, WK-23 displayed a good plasma exposure and an acceptable oral bioavailability of F% ¼ 37.1.

Chemistry
All anhydrous reactions were performed under a nitrogen atmosphere. The reaction progress was monitored by thin layer-chromatography (TLC) using silica gel F254 plates. Melting points (uncorrected) were determined on an XRC-1 micro melting point apparatus. Infra-red spectra (IR) were recorded on a PerkinElmer Spectrum Two FT-IR instrument. High-resolution mass spectra (HRMS) were taken on a Thermo-Fisher LTQ Orbitrap XL instrument. The 1 H and 13 C NMR (nuclear magnetic resonance spectra) experiments were performed by Bruker AM-600 spectrometer using TMS (tetramethylsilane) as the internal standard. Column chromatography was run on 200-300 mesh silica gel from Qingdao Ocean Chemicals (Qingdao, Shandong, China). Unless otherwise indicated, all materials were obtained from commercially available sources and used without further purification.

2-Chloro-N-(4-fluorophenyl) aniline (a2)
To a mixture of NaO t Bu (24.00 g, 250.0 mmol), [HP t Bu 3 ][BF 4 ] (1.02 g, 3.5 mmol) and Pd(OAc) 2 (560.0 mg, 2.5 mmol) in toluene (200 ml) and stirred to disperse it well. Then added 2-chloraniline (6.40 g, 50.0 mmol) (a1) and 4-bromofluorobenzene (8.75 g, 50.0 mmol). After the addition was completed, the temperature was raised to 110 C and refluxed for 4 h under a nitrogen atmosphere. After completion of the reaction as indicated by TLC (PE/EA ¼ 20:1) (petroleum ether/ethyl acetate), cool to room temperature, quenched by water (100 ml), extracted with EtOAc (ethyl acetate) (200 ml Â 2). The organic layers were combined and washed with brine, dried with anhydrous Na 2 SO 4 , and evaporated in vacuo to obtain the crude products of 12.60 g. The crude products were further purified by column chromatography with an eluting system of petroleum ether (100%) to give a2 as a light-yellow liquid ( 2 (510.0 mg, 2.3 mmol) in 1,4-dioxane (250 ml) and stirred to disperse it well. Then added a2 (10.07 g, 45.5 mmol). After the addition was completed, the temperature was raised to 110 C and refluxed for 4 h under a nitrogen atmosphere. After completion of the reaction as indicated by TLC (PE/EA ¼ 20:1), cool to room temperature and quenched by water (100 ml), extracted with EtOAc (200 ml Â 2). The organic layers were combined and washed with brine, dried with anhydrous Na 2 SO 4 , and evaporated in vacuo to give the grey black crude products 10.20 g. The crude products were further purified by column chromatography with an eluting system of PE/EA (100:1-20:1). The residue after concentration in vacuo was triturated with petroleum ether, the mixture was filtered to obtain a white solid a3 3.80 g in 45% yield.  2.2.4. General procedure for intermediates a4 and b4 2.2.4.1. 3-Fluoro-9-(oxiran-2-ylmethyl)-9H-carbazole (a4). At ambient temperature, a3 (2.40 g, 13.0 mmol) was dissolved in DMF (50 ml), followed by the addition of KOH (0.80 g, 14.3 mmol). After 5 min stirring under 0 C, epichlorohydrin (2.40 g, 26.0 mmol) was added dropwise. Then the mixture was warmed to room temperature and stirred for another 3 h. After a3 was completely consumed, the solution was poured into 20 ml water, and extracted with EtOAc (50 ml Â 2). The organic layers were combined and washed with brine, dried with anhydrous Na 2 SO 4 , and evaporated in vacuo to give 3.80 g of black liquid crude product. The crude products were further purified by column chromatography with an eluting system of PE/EA (80:1-20:1) to give off brown solid a4 2.2.5. General procedure for intermediates d1, d2, d3 2.2.5.1. 1,3-Bis(3-fluoro-9H-carbazol-9-yl)propan-2-ol (d1). To solution of a4 (0.90 g, 3.7 mmol) and a3 (0.69 g, 3.7 mmol) in acetone 8 ml was added KOH (0.42 g, 7.5 mmol) and anhydrous Na 2 SO 4 (0.53 g, 3.7 mmol). The resultant mixture was stirred at room temperature overnight. After TLC indicated the total conversion, water was added to quench the reaction and the mixture was washed with EtOAc (20 ml Â 2). The organic layer was dried over anhydrous Na 2 SO 4 and concentrated in vacuo. Flash column chromatography utilising PE/EA (10:1) as the eluting afforded d1.
Intermediates d5 and d6 were prepared like that described for d4.
IR  . Powder KOH was added to a 9H-b-carboline solution in N, N-dimethylformamide (DMF) at ambient temperature and stirred for 30 min until dissolved. Epichlorohydrin was added via syringe, and the mixture was stirred at room temperature. Upon completion, the solution was partitioned between EtOAc and H 2 O. The crude products were used for the next step without further purification. A mixture of crude products (200.0 mg) and tryptamine (aromatic amines) (284.0 mg, 1.8 mmol) dissolved in 5 ml EtOH, was introduced into a 10 ml sealed tube. The mixture was stirred at 60 C. Upon completion, the mixture was treated with EtOAc and water. The organic layer was washed with saturated aqueous NaCl, dried over anhydrous Na 2 SO 4 , and concentrated in vacuo. Finally purified by preparative thin layer chromatography to afford 1

Biological section
2.3.1. DNMT1 inhibition assays 2.3.1.1. ELISA DNMT1 activity assay. All the compounds were first screened using an ELISA EpiQuik DNA methyltransferase (DNMT) activity/inhibitor assay kit (Epigentek). To measure the effects of the compounds on human DNMT1 activity, 200 nM purified DNMT1 was incubated with 50 mM and 100 mM of the different compounds and S-adenosylmethionine in the DNMT assay buffer in the assay plate at 37 C for 2 h 36 . Next, every sample was incubated with the capture and detection antibody, followed by incubation with developer solution for 10 min at room temperature. The absorbance was measured at 450 nm using a POLARstar Omega microplate reader (BMG). S-Adenosylhomocysteine (AdoHcy) was used as a positive control. Methylation inhibition assays for EZH2, LSD, and G9a were performed in modified Tris, pH 9.0, buffer using AlphaLisa technology. An amount of 10 mL of the reaction system contained a corresponding concentration of SAM (Sigma) (EZH2 50 mM, LSD 50 mM, and G9a 50 mM), which was the Km value in each enzymatic reaction, plus 100 nM biotinylated peptide H3 (1-21) (synthesis by GLChina) and the relevant enzyme concentration (0.03 nM EZH2, 0.03 nM LSD, and 0.03 nM G9a). The proteins were preincubated with various compound concentrations for 15 min at room temperature before the substrate and SAM were added. After 60 min of incubation at room temperature, acceptor and donor AlphaLisa beads were added according to the manufacturer's recommendations. The signals were read in Alpha mode with an EnSpire multimode plate reader (PerkinElmer). IC 50 values were derived by fitting the data for the inhibition percentage to a dose-response curve by non-linear regression in GraphPad Prism 5.0 20 .

2.3.1.2.
Radioactive methylation assay. The DNMT1 radioactive methylation inhibition assays were performed in 30 mL reactions containing 0.1 mM adenosyl-L-methionine S-[methyl-3 H] ( 3 H-SAM, 15 Ci/mmol, PerkinElmer), 0.25 mg/mL poly (dI-dC)Ápoly (dI-dC) (Sigma), 40 nM DNMT1 in 50 mM Tris-HCl, pH 8.0, 1 mM DTT, 5% glycerol, and 100 mg/mL BSA. The proteins were preincubated with a range of compound concentrations for 15 min at room temperature before adding the substrate and [ 3 H] SAM. After 60 min of incubation at 37 C, the reaction systems were transferred to a MultiScreen HTS filter plate (Millipore), and the plate was washed 3 times with doubly distilled H 2 O via a vacuum. The radioactivity was determined by liquid scintillation counting (MicroBeta, PerkinElmer). IC 50 values were derived by fitting the data for the inhibition percentage to a dose-response curve by non-linear regression in GraphPad Prism 5.0.

Cell lines and culture conditions
The human lung cancer cells (A549) and human colon cancer cells (HCT116) were cultured in an RPMI-1640 medium containing 10% FBS, 100 U/mL streptomycin, and 100 U/mL penicillin at 37 C in a humidified atmosphere with 5% CO 2 . All compounds were dissolved in DMSO and stock solutions were stored at À20 C. Reagents were freshly diluted to the marked concentrations with a culture medium before use. DMSO concentration in experimental conditions never exceeded 0.1% (v/v). All cell lines were provided by the cell laboratory, School of pharmaceutical sciences, Guangzhou Medical University.

MTT assay
Cell viability was detected using an MTT assay kit. Cells were seeded into 96-well plates at a density of $1.0 Â 10 4 /well. 24 h later, sextuplicate wells were treated with media and new compounds at a fixed concentration (50 mM). After 24, 48, and 72 h, the drug-containing medium was replaced by a 100 mL fresh medium with 5 mg/mL MTT solution. After 4 h of incubation, the medium with MTT was removed, and 100 mL of DMSO was added to each well. The plates were gently agitated until the purple formazan crystals were dissolved, and the OD490 was determined using a microscope (Olympus BX53, Japan). The data were calculated and plotted as the percent viability compared to the control.

PK study
In vivo pharmacokinetic properties of WK-22, WK-23, WK-27, and DC_517 were performed by Medicilon Company, Shanghai, China. SPF-grade SD male rats (8 groups, n ¼ 3 rats per group) with a body weight of 230-260 g were purchased from Shanghai SIPPR-BK LAB Animal Ltd., Shanghai, China, and used for the pharmacokinetic analysis of the tested compounds. All animals were deprived of food overnight after the cannulation surgery. Subsequently, the tested compounds WK-23 and DC_517 were dissolved/suspended in 5% DMSO, 10% Solutol, and 85% water, and the tested compounds WK-22 and WK-27 were dissolved/suspended in 5% DMSO, 40% PEG400, and 55% saline for intravenous administration (i.v.) and oral administration (p.o.), respectively. A final dosage of 2.0 and 10.0 mg/kg rat of the formulated compounds was administered for i.v. and p.o. purposes, respectively, and the blood samples were taken at various time points during a 24 h period. At different time points, blood samples were collected from the femoral vein. The plasma samples were obtained after centrifugation (6800 g, 6 min, 2-8 C) and stored at À80 C until the assay. The AUC (area under concentration-time curve) was calculated through the trapezoidal rule with extrapolation to time infinity. The concentration of the compounds in the blood was analysed by LC-MS/MS (Shimadzu liquid chromatographic system and SCIEX Triple Quad 5500þ mass spectrometer, Applied Biosystems, Ontario, Canada). The T max , T 1/2 , and C max value was obtained through visual inspection of the plasma concentrationtime curve. The Vss value was generated from DAS 3.2.8 software. The F value of WK-23 was calculated with the formula:

Molecular docking
Molecular docking was performed by the Glide program packed in Maestro (Maestro, Schr€ odinger, LLC, New York, NY, 2020.). The crystal structure of human DNMT1 was used as a template (PDB code: 4WXX) and removed the water molecules 20 . The docking procedure was initiated by the optimisation of protein structure using the Protein preparation Wizard module. Then the inhibitor was optimised by the Ligand Preparation module to generate stereoisomers and protonation states. Extra precision (XP) mode was used to perform the molecular docking, and the final result was selected through the Glide score function.

Chemistry
The preparation routes of all target compounds are outlined in Scheme 1. First, we synthesised the key intermediates, and the general synthetic routes are illustrated in Scheme 1. Briefly, the commercially available 2-chloroaniline or 2-chloro-4-fluoroaniline reacted with bromofluorobenzene by the Buchwald-Hartwig coupling reaction and then the transition metal-catalyzed C-C coupling ring was constructed for 3 or 3, 6 substitution carbazole 21 . Fluorine-substituted carbazoles reacted with epichlorohydrin to get epoxy intermediates 22 . The general synthetic route of piperazine sidechain is illustrated in Scheme S1. Finally, the epoxy intermediates interacted with different amine compounds under basic conditions by nucleophilic substitution reaction to achieve 30 target compounds 23 . The structure of these compounds was confirmed by 1 H NMR, 13 C NMR, high-resolution mass spectra, and infra-red spectra.

DNMT1 inhibition assays
DNMT1 inhibition test was performed on 30 candidates to verify their biochemical activities. The EpiQuik DNA methyl-transferase (DNMT) activity/inhibitor assay kit (Epigentek) was used to identify the activity of synthetic compounds. It was found that 11 compounds could inhibit DNMT1 activity by >80% (Table 1, Figure 2), and these compounds had similar potency to that of DC_517 against DNMT1 at a concentration of 50 mM and 100 mM.

Radioactive methylation assay against DNMT1
To quantitatively analyse DNMT1 inhibitory effect, the H 3 -labelled radioactive methylation assay was conducted to measure the methyltransferase activity of DNMT1 at a range of concentrations for these compounds. As shown in Table 1, WK-22 (IC 50 ¼ 4.9 mM) and WK-23 (IC 50 ¼ 5.0 mM) showed similar inhibitory activity as DC_517 (IC 50 ¼ 2.3 mM), whereas the other compounds were less potent. This result helped us to uncover sufficient information for the study of hit optimisation and structure-activity relationships (SARs).

Structure-activity relationships investigation
All the carbazole derivatives summarised in Scheme 1, including twenty-one derivatives of DC_05 and nine derivatives of DC_517 were prepared and biologically evaluated for the DNMT1 inhibitory activity. The experimental data of compounds bearing monocarbazole or dicarbazole as DNMT1 inhibitors are displayed in Table 1, respectively. The structure-activity relationships of DC_05, DC_517, and its analogs were determined and investigated by comparing the DNMT1 inhibitory activity.    Accordingly, as shown in Figure 2, the catalytic site of DNMT1 is thereby includes two parts, namely, the cofactor binding site SAM pocket and the substrate binding site cytosine pocket. In our previous study, we conducted the molecular docking study and found that the compounds occupied the SAM pocket and cytosine pocket of DNMT1 (PDB code 4DA4) 20 .
WK-20 and WK-21 with a nitrogen heterocyclic replaced at region b, which showed lower inhibitory activity than DC_05 at 100 mM. It could be that the lone pair on the nitrogen is not tolerated in the pocket either because the environment is non-polar or because there is actually already an atom with a lone pair that leads to a repulsive effect.
The result of enzyme inhibitory activities showed that different amines at region c will lead to a different activity. When it comes to monocarbazole, the decreasing order of influence of amines in inhibition activity was tryptamine, phenethylamines, and aliphatic amines. It could be corresponding to the space structure of the cytosine pocket since the structures linked to the amines are too small to occupy the cytosine pocket completely. But WK-6, WK-7, WK-18, and WK-19 still showed a relatively potent activity, possibly because the derivatives form extra hydrogen bonds via its hydroxyl to bind to the amino acid residue of the cytosine pocket. Similarly, phenethylamines derivative WK-13 (IC 50 ¼ 19 mM), which has a hydrogen donor on the aromatic ring exhibited a better activity. The inhibition of WK-2 and WK-9 are inferiors to DC_05, which suggests a long linking chain is optimal for the distance of binding targets. This derivation is also consistent with the activity data of dicarbazole molecules. In addition, WK-26 lost its activity probably because the Boc-protecting group resulted in steric hindrance while WK-24 was also inactive perhaps because the cyclopropyl group reduces hydrophobic and weakens the binding with the cytosine pocket.

Methyltransferase enzymatic selectivity profiling
Based on the DNMT1 Inhibition Assays data, some derivatives showed a considerable DNMT1 inhibition activity. Monocarbazole molecules that exhibited better inhibition on DNMT1 than DC_05, or dicarbazole molecules that exhibited better inhibition than DC_517 at 50 mM and 100 mM, were chosen to promote the next step. In addition to DNMT1, there are many other methyltransferases that can bind with S-adenosyl-L -methionine (SAM) to facilitate transmethylation reactions 24,25 . To investigate the selectivity of these potent compounds for DNMT1, we then evaluated the inhibitory activities against DNMT1 and other important methyltransferases at the concentration of 50 mM, including DNMT3A/3L, DNMT3B/3L, and other SAM-dependent enzymes 1,2,26 , such as EZH2 27 , LSD1 28 , G9a (histone H3 lysine 9 methyltransferase) 29 . As shown in Figure 3, the enzymatic selectivity and inhibitory activity were measured at the concentration of 50 mM. Compounds WK-1, WK-12, and WK-13 displayed a strong activity on DNMT1 while showing hardly any inhibitory activities against DNMT3B/3L, EZH2, and G9a. Compounds WK-22, WK-23, and WK-27 displayed a similar activity targeting DNMT1 to that on DNMT3A/3L, DNMT3B/3L, and LSD1, while showed nearly no inhibitory activities against EZH2 and G9a. As shown in Table S1, WK-23 exhibited good selectivity on EZH2 and G9a (Relative selectivity index ¼ 0.85 and 0.89), while had a similar effect on DNMT 3 A/3L, DNMT 3B/3L, and LSD1 to that on DNMT1 (Relative selectivity index ¼ 0.06, 0.02 and 0.2). Here, the selectivity of nonnucleoside DNMT1 inhibitors is somewhat predictable because other methyltransferases mentioned above catalyse different substrates and share very low homology with DNMT1, even in their catalytic domain [30][31][32] . Subsequently, the screened compounds were evaluated in a dose-dependent assay against HCT116 (human colon cancer cell line) and A549 (human lung carcinoma cell line).

Evaluation of antitumor activity in cells
Collectively, compounds WK-1, WK-12, WK-13, WK-19, WK-22, and WK-23 showed significant DNMT1 inhibiting activity, and we concomitantly explored the anti-proliferative effect of these most interesting compounds on cancer cell lines. As shown in Table 3, we tested these molecules in HCT116 34 and A549 35 . Notably, WK-22 and WK-23 displayed the highest anti-proliferative effects in these two types of cancer cell lines, which is consistent with those of the DNMT1 inhibition assays. The results also demonstrated that WK-22 and WK-23 led to obvious dose-dependence and time-dependence anti-proliferation in HCT116 cancer cells ( Figure  4). Moreover, in comparison with A549, generally, these compounds are more sensitive to HCT116 cells. Regarding the Pharmacokinetics study, WK-23 exhibited an advantage in elimination half-life (T 1/2 ¼ 7.9 h) and oral bioavailability (F% ¼ 37.1 ± 1.7) over other compounds. Consequently, we further explored the binding mode of WK-23 to DNMT1.

Molecular docking
Molecular docking was performed to explore the binding mode of WK-23 to DNMT1. The docking procedure was validated by reproducing the SAH binding mode with a root-mean-square deviation (RMSD) of 0.965 Å ( Figure S1). According to our previous study, there is no obvious difference between the enantiomers of DC_517, which implies the racemic of WK-23 is applicable for molecular docking simulations with DNMT1 (PDB code 4WXX). As shown in Figure 5, the binding pattern of WK-23 with DNMT1 is similar to that of DC_517 20 . One carbazolyl of WK-23 occupied the SAM pocket, and the other carbazolyl forms cation À p interactions with R1574 and W1170. Besides, the amino group of WK-23 interacts with the main chain of F1145, and the hydroxyl group forms hydrogen bonds with E1168. These polar interactions further stabilise the binding of the inhibitor to DNMT1.

Conclusion
In summary, a series of novel carbazole-based derivatives were designed, synthesised, and evaluated for their biological activity. The structure-activity relationship of their anti-proliferative activity Figure 4. Effect of WK-23 and DC_517 on the viability of human colon cell lines and human lung cell lines. WK-23 has stronger concentration dependence and time dependence than DC_517, which inhibited the viability of HCT116 and A549 cells in the cell viability assay. Significance between groups was analysed by one-way analysis of variance (ANOVA) using IBM SPSS software. Ã p < 0.05. Values are the means of at least two independent experiments.
was explored. Among these compounds, WK-22 and WK-23 displayed appreciable human DNMT1 inhibitory activity in the micromolar range (IC 50 ¼ 4.9 mM and 5.0 mM). Simultaneously, both WK-22 and WK-23 has promising anti-proliferative effect on A549 and HCT116 cell lines. In further in vivo pharmacokinetic study, WK-23 displayed a better plasma exposure and prolonged elimination half-life (T 1/2 ¼ 7.9 h), especially the more acceptable oral bioavailability of (F% ¼ 37.1) than WK-22 (F% ¼ 27.0). Concomitantly, the molecule docking showed the binding pattern of WK-23 with DNMT1 is similar to that of DC_517, forming stable binding to DNMT1. In conclusion, due to its favourable biological performance, compound WK-23 warrants further assessment as a potential therapeutic agent for the treatment of human cancers.