Identification of the KDM2/7 Histone Lysine Demethylase Subfamily Inhibitor and its Antiproliferative Activity

Histone Nε-methyl lysine demethylases KDM2/7 have been identified as potential targets for cancer therapies. On the basis of the crystal structure of KDM7B, we designed and prepared a series of hydroxamate analogues bearing an alkyl chain. Enzyme assays revealed that compound 9 potently inhibits KDM2A, KDM7A, and KDM7B, with IC50s of 6.8, 0.2, and 1.2 μM, respectively. While inhibitors of KDM4s did not show any effect on cancer cells tested, the KDM2/7-subfamily inhibitor 9 exerted antiproliferative activity, indicating the potential for KDM2/7 inhibitors as anticancer agents.


■ INTRODUCTION
N ε -Methylation of lysine residues on histone tails is a pivotal "epigenetic mark" that is involved in defining both transcriptionally active and inactive chromatin. 1 For instance, methylation at lysine 4 of histone 3 (H3K4) is associated with actively transcribed gene loci, whereas methylation at H3K9 and H3K27 leads to reduction in transcription. 2 Methylation of histone lysine residues is reversible and is mediated by histone lysine methyltransferases and histone lysine demethylases (KDMs). 3 KDMs fall into two classes as defined by their structure and mechanisms. 3 One class comprises the KDM1s, which are homologues of the flavin-containing amine oxidases. 3a,b The other class comprises Jumonji C (JmjC)-domain containing demethylases, which belong to the family of Fe(II)/αketoglutarate-dependent oxygenases. 3c,d Human JmjC-domain containing demethylases can be divided into five subfamilies, i.e., the KDM2/7, KDM3, KDM4, KDM5, and KDM6 subfamilies, according to sequence and structural similarities. 4 Several JmjC-domain containing demethylases such as KDM2, KDM4, KDM5, and KDM7 have been implicated in tumorigenesis. 5 For example, it has been reported that KDM7B (also known as JHDM1F, PHF8, KIAA1111) is associated with proliferation of prostate cancer cells and osteosarcoma cells. 6 Therefore, KDM2/7-subfamily inhibitors are of interest, both as tools for probing the biological functions of KDM2/7 subfamily and also as candidate anticancer agents. Several types of KDM inhibitors have been found so far by us and others, 5a,7 including N-oxalylglycine (NOG, 1), 2,4-pyridinedicarboxylic acid (PCA, 2), NCDM-32 (3), daminozide (4), and GSK-J1 (5) (Chart 1). Among these, daminozide (4) has been reported as a KDM2/7-subfamily selective inhibitor. Daminozide (4) was once widely used as a plant growth retardant but later withdrawn because of genotoxic concerns. 8 The genotoxicity of daminozide (4) is proposed to be derived from its 1,1dimethylhydrazine structure. 9 Therefore, it is desirable to identify novel KDM2/7-inhibitors without the 1,1-dimethylhydrazine structure for use as functional probes and for evaluation in cancer therapies. Here we describe the identification of a novel KDM2/7 subfamily inhibitor that shows antiproliferative activity.

■ CHEMISTRY
The routes used for the synthesis of compounds 7−13 (Chart 2), which were prepared for this study, are shown in Schemes 1 and 2. Scheme 1 shows the preparation of compounds 8−13. Michael addition of O-benzylhydroxylamine to tert-butyl acrylate 14 afforded amine 15. 9-Bromononanoic acid sodium salt 16 was converted to alkyl carboxylic acids 19−22 by treatment with an alkyl magnesium bromide 17 and dilithium tetrachlorocuprate. 9-Phenylnonanoic acid 23 was obtained by the treatment of 16 with phenyl magnesium bromide and ferric acetylacetonate in the presence of tetramethylethylenediamine and hexamethylenetetramine in THF. Carboxylic acids 18−23 were treated with amine 15 in the presence of EDCI and HOBt to give amides 24−29. The benzyl group of compounds 24−29 was removed by hydrogenation to give hydroxamates 30−35. Removal of the tert-butyl group of 30−35 using hydrochloric acid afforded the desired compounds 8−13.
Scheme 2 shows the synthesis of compound 7. Michael addition of O-benzylhydroxylamine to methyl acrylate 36 afforded amine 37. 10-Methylundecanoic acid 38 was prepared using the procedure described for the synthesis of 19−22. Amine 37 was treated with 38 in the presence of EDCI and HOBt to yield compound 39. Removal of the benzyl group of compound 39 gave compound 7.

■ RESULTS AND DISCUSSION
With the aim of identifying potent KDM2/7 inhibitors, hydroxamate derivatives, which have been previously prepared in work on KDM4 inhibitors, 7d were screened for KDM7B-, KDM4A (also known as JMJD2A, JHDM3A)-, KDM4C (also known as JMJD2C, JHDM3C, GASC1)-, and KDM5A (also known as JARID1A, RBP2)-inhibitory activity. We selected KDM4A, KDM4C, and KDM5A for the primary inhibition selectivity studies because they are relatively similar to KDM7B in that they have a plant homeodomain. 4 Compounds 1−4 were evaluated as positive controls for inhibition. In our enzyme assays, compound 4 selectively inhibited KDM7B in preference to KDM4 and KDM5 as reported before (Table 1). 4 We also found that compounds 1 7a and 2 7b did not inhibit KDM7B (at 250 μM), and compound 2 showed selectivity for the KDM4-subfamily over KDM7B as for the reported KDM4 inhibitor 3 7d under the conditions tested. Among the tested compounds, compound 6 was found to inhibit KDM7B relatively selectively although its potency was moderate (IC 50 = 24 μM).
To improve the potency and selectivity of 6 for KDM7B, we investigated the binding mode of 6 with KDM7B by modeling ( Figure 1). 10 An inspection of the simulated KDM7B/ compound 6 complex indicates that the hydroxamate group of compound 6 coordinates to the Fe(II) in a bidentate manner, via its carbonyl and hydroxyl groups, and that its carboxylate group will bind to form hydrogen bonds/ electrostatic interactions with Asn 189, Thr 244, and Lys 264 at the KDM7B active site, in a similar manner to the C-5 carboxylate of 2-oxoglutarate. Most importantly, it is also suggested that the alkyl chain of compound 6 binds to a hydrophobic pocket formed by the side chains of Ile 191, Tyr 234, Leu 236, Phe 250, Val 255, Tyr 257, and Phe 359, which we propose is unique to KDM2 and KDM7 (Supporting Information Figures S1−S6).
On the basis of the structure of the modeled KDM7B/ compound 6 complex, we designed and synthesized compounds 8−13 (Table 2) in which the isopropyl group of compound 6 was replaced with various hydrophobic groups that might interact with hydrophobic amino acid residues more efficiently than compound 6. Compound 7 was prepared to investigate the importance of the carboxylic acid of the inhibitor. As shown in Table 2, the conversion of carboxylic acid (6) to a methyl ester (7) significantly reduced the KDM7B-inhibitory activity, supporting the proposed binding mode of compound 6. Compounds 8−13 were found to be KDM7B inhibitors. In particular, a pronounced inhibitory effect (IC 50 = 1.2 μM) was observed with the cyclopropyl-containing compound 9, which was 20-fold more active than 6 in KDM7B in MALDI assays. Docking of compound 9 in the active site of KDM7B suggested that the cyclopropyl group of compound 9 interacts with the phenyl-group of Phe 250 through CH−π or cyclopropyl−π interactions ( Figure 2). Furthermore, compound 9 inhibited KDM2A (also known as JHDM1A, FBXL11), KDM7A (also known as JHDM1D, KIAA1718), and KDM7B, demonstrating inhibition of all KDM2/7 subfamily members tested (Tables 2  and 3). Compound 9 displayed selectivity for KDM2/7 over KDM4A (IC 50 > 120 μM), KDM4C (IC 50 = 83 μM), KDM5A (IC 50 = 55 μM), and KDM6A (also known as UTX) (IC 50 > 100 μM); note however different assays conditions were used (see Experimental Section). Thus, compound 9 was the most potent KDM7B inhibitor identified in the enzyme assays. In addition, compound 13 showed comparatively high selectivity for KDM2A over KDM7B and the other isozymes tested. Because the molecular modeling suggest that the space around      KDM2A has a hydrophobic pocket more spacious than KDM7B (Supporting Information Figure S2), in which the phenyl ring can efficiently interact with hydrophobic amino acid residues of KDM2A. These may be the reason that compound 13 shows selectivity for KDM2A over KDM7B.

Journal of Medicinal Chemistry
To investigate whether compound 9 is active as an inhibitor of KDM7A and KDM7B in cells, we performed a cellular assay with Western blot analysis. Because KDM7 is known to catalyze the demethylation of H3K27me2, 10,11 the methylation level of H3K27 in cells was analyzed. In this study, we used N2a cells because it has been reported that KDM7 is expressed in the cells. 11 As Figure 3 shows, the level of H3K27me2 was dose-dependently elevated in the presence of 9. Although the interpretation of changes in global histone N-methylation status can be complex, the elevation in the H3K27me2 levels is consistent with KDM7 inhibition. These results suggest that H3K27me2 demethylation is inhibited by compound 9 in cells, and it looks to be useful as a tool for probing the biological role of KDM7.
Although it has been reported that RNAi-mediated knockdown of KDM7B suppresses the growth of some cancer cells, 6 it remains unclear whether the demethylase function of KDM7B is responsible for the suppression because RNAimediated knockdown of KDM7B should cause loss of not only the demethylase function but also other functions of KDM7B, 12 including those relating to noncatalytic binding domains. Initially, we investigated the N2a cell growth inhibition activity of compound 9. N2a cell growth suppression by compound 9 was observed (GI 50 = 86 μM) ( Figure 4) at the concentration range in which distinct H3K27 methylation was detected on Western blot analysis (Figures 3). Thus, this may suggest the demethylase function of KDM7 is involved in this cell growth, however, more potent and selective compounds will be needed to fully elucidate this fact. Next, we carried out cell growth inhibition assays of compound 9 as well as prodrugs of 1, 2, and 3 using HeLa cells and KYSE150 cells (Table 3). It has also been reported that knockdown of KDM4C decreases cell proliferation, 7a however, the KDM4C inhibitors 2, 3 (Table 3), and their methyl ester prodrugs (GI 50 > 500 μM) did not show any effects on the growth inhibition of tested cancer cells although they are cell membrane permeable. 7d,13 These results suggest that the demethylase activity of KDM4C is not directly associated with cancer cell growth, at least in some cell types. On the other hand, cell growth suppression by the KDM2/7subfamily inhibitor 9 was observed for KYSE150 and HeLa cell lines ( Figure 5). Furthermore, compound 9 caused H3K27 methylation both in HeLa cells and in KYSE150 cells at the concentration range in which the cell growth inhibition was observed ( Figure 6). The data shown in Table 3 indicate that KDM2/7 inhibitors are worthy of evaluation as candidate anticancer agents.
A recent study reported that KDM7B activates the transcription of the E2F1 transcription factor in HeLa cells, which promotes cell cycle progression. 14 Because compound 9 decreases the growth of HeLa cells with H3K27me2 accumulation (Table 3; Figures 5 and 6), we examined whether compound 9 down-regulates the expression of E2F1 in HeLa cells by quantitative RT-PCR. As shown in Figure 7, compound 9 significantly decreases the mRNA level of E2F1 at 80 μM in    which the growth of HeLa cells was affected. These data suggest that the KDM7B-mediated regulation of E2F1 gene expression may be one of the mechanisms of growth regulation in some cancer cells.
We also investigated the effect of compound 9 on cell cycle progression by FACS analysis. HeLa cells and KYSE150 cells incubated with 10 or 100 μM of compound 9 for 24 h showed a dose-dependent reduction in G 2 −M phase, whereas there was a dose-dependent increase in G 0 −G 1 phase ( Figure 8). These results revealed that HeLa cells and KYSE150 cells cultured with compound 9 arrested in the G 0 /G 1 phase of the cell cycle, which is consistent with the down-regulation of E2F1 by compound 9 (Figure 7). 15

■ CONCLUSIONS
We have identified a novel KDM2/7 subfamily inhibitor 9, which should be useful as a lead structure in the development of more potent and selective KDM2/7 inhibitors. Such inhibitors are candidates for anticancer agents as well as tools for studying the biological roles of KDM2/7 subfamily in cells.
Steps 2, 3, and 4: Preparation of Methyl 3-[Hydroxyl(10methylundecanoyl)amino]propanoate (7). Compound 7 was prepared from 37 (3.1 g, 15 mmol) using the procedure described favorable binding poses, which allowed the ligand conformation to be flexibly explored while holding the protein as a rigid structure during docking.
KDM4C Inhibition Assay. The KDM4C-inhibitory activity was measured using 0.6 mg/mL enzyme. The compounds were dissolved in DMSO. The final concentrations of DMSO in the reaction mixtures were less than 3.3%, and it was confirmed that 3.3% DMSO did not affect the KDM4C activity. Reaction with DMSO alone was also done as a control. Reaction mixtures (94.6 μL), containing all of the materials except H3K9me3 peptide and 2-oxoglutarate, were preincubated for 5 min. Then the reactions were started by the addition of 5.4 μL of a solution of 0.93 mM H3K9me3 peptide and 3.7 mM 2-oxoglutarate . The enzyme activity was determined as described above. The ratio of the enzyme activity measured in the presence of inhibitor to that of the control was plotted against log [Inhibitor]. To confirm that the reduction of the KDM4C activity by test compounds was not due to inhibition of the coupled enzyme FDH, we examined the effects of the test compounds on FDH activity. The reaction mixture (0.1 mL) contained 20 mM HEPES-KOH, pH 7.5, 50 μM formaldehyde, 1 mM 3-acetylpyridine adenine dinucleotide, 1 mM reduced glutathione, 0.1 mg/mL BSA, 0.1 mg/mL FDH, and a fixed concentration of 123 μM test compound. The FDH activity was measured by monitoring APADH formation as described above. The FDH activity in the presence of test compounds was similar to that in the absence of the compounds.
KDM4A Inhibition Assay. The KDM4A activity was measured by the FDH-coupled assay as described for KDM4C except that reactions were performed in a final volume of 30 μL in 384-well plate (Nunc) and a final concentration of the KDM4A was 0.37 mg/mL.
KDM5A Inhibition Assay. The KDM5A activity was measured by the FDH-coupled assay as described for KDM4C except that reactions were performed with H3K4me3 peptide in a final volume of 30 μL in 384-well plate (Nunc) and a final concentration of the KDM5A was 0.64 mg/mL.
KDM2A Inhibition Assay. The inhibitory activities of test compounds against KDM2A was assayed according to the method reported in ref 4.
KDM6A Inhibition Assay. The Epigenase JMJD3/UTX demethylase activity/inhibition assay kit (Epigentek Group Inc.) was used for KDM6A enzyme assay. The materials supplied with the kit, 100 μM of test compounds and KDM6A (human, recombinant, BPS Bioscience, Inc.) (300 ng/well) were added to wells coating trimethylated histone substrate, according to the supplier's protocol. The resulting mixtures were incubated at 37°C for 120 min. After enzyme reaction, each well was reacted with capture antibody for 60 min and with detection antibody for 30 min. Finally, developer solution and stop solution were added to wells in sequence and the absorbance (450 nm) in each well was measured with an ARVO X3 microplate reader. The KDM6A inhibition activity of test compounds was calculated from the absorbance readings.
KDM7A Inhibition Assay. The KDM7A assay was run using MALDI and K27me2 peptide as a substrate under the same conditions as used in ref 4.
Western Blot Analysis. N2a cells, HeLa cells, or KYSE150 cells (5 × 10 5 ) were treated for 24 h with inhibitors at the indicated concentrations in OPTI-MEM medium, then collected and extracted with SDS buffer. Protein concentrations of the lysates were determined using a Bradford protein assay kit (Bio-Rad Laboratories); equivalent amounts of proteins from each lysate were resolved in 4− 20% SDS-polyacrylamide gel and then transferred onto nitrocellulose membranes (Bio-Rad Laboratories). After having been blocked for 30 min with Tris-buffered saline (TBS) containing 3% skim milk, the transblotted membrane was incubated overnight at 4°C with dimethylated H3K27 antibody (CST) (1:1000 dilution) or H3 antibody (Abcam) (1:100000 dilution) in TBS containing 3% skim milk or TBS-Tween 20 (TBS-T). The membrane was probed with the primary antibody, then washed twice with TBS or TBS-T, incubated with goat antirabbit IgG-horseradish peroxidase conjugates (diluted 1:2500) for 1.5 h at room temperature, and again washed twice with TBS and once with TBS-T. The immunoblots were visualized by enhanced chemiluminescence.
MTT Assays. The cells were plated at initial densities of 5000 cells/ well (100 μL/well) in 96-well plates in RPMI 1640 with 10% fetal bovine serum and allowed to attach overnight. The cells were exposed to inhibitors for 48 h at 37°C in 5% CO 2 incubator. A solution (5 mg/ mL) of 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) (Sigma) was added (10 μL/well) and incubated with the cells for 3 h before solubilization buffer (0.04 mol/L HCl−2-propanol) was added (100 μL/well) onto the cultured cells. The solubilized dye was quantified by colorimetric reading at 560 nm using a reference wavelength of 750 nm. Absorbance values of control wells (C) and test wells (T) were measured. Moreover, absorbance of the test wells (T 0 ) was also measured at time 0 (addition of compounds). Using these measurements, cell growth inhibition (percentage of growth) by a test inhibitor at each concentration used was calculated as: % growth = 100[(T − T 0 )/(C − T 0 )], when T > T 0 and % growth = 100[(T − T 0 )/T], when T < T 0 . Computer analysis of the % growth values afforded the 50% growth inhibition parameter (GI 50 ). The GI 50 was calculated as 100[(T − T 0 )/(C − T 0 )] = 50.
RNA Isolation and Semi-qRT-PCR. HeLa cells were treated for 48 h with 0.238% DMSO or compound 9 at the concentration of 30 and 80 μM, respectively. Total RNA was isolated from HeLa cells using RNAzol (Molecular Research Center, Inc.) following the manufacturer's protocol. First-strand cDNA synthesis from total RNA was carried out using ReverTra Ace (TOYOBO). Resulting cDNA was then analyzed by semiquantitative PCR (semi-qPCR) using 2720 thermal cycler (Applied Biosystems). Primers are specific for genes tested, and their sequences are as follows: GAPDH Cycle parameters were 94°C for 2 min, followed by 28 (E2F1), 20 (GAPDH) cycles of 98°C for 10 s, 60°C for 30 s, and 68°C for 30 s, with a final extension at 68°C for 1 min.
FACS Analysis. Cells (5 × 10 5 ) were treated for 24 h with compound 9 at the indicated concentrations in RPMI 1640 with 10% fetal bovine serum, then harvested by trypsinization. The cells were collected by centrifugation, fixed with ice-cold 70% ethanol, washed with phosphate-buffered saline, and resuspended in 0.5 mL of phosphate-buffered saline containing propidium iodide (10 μg/mL) and RNase A (0.2 mg/mL). After a final incubation at 25°C for 30 min, the cells were analyzed using a JSAN flow cytometer (Bay Bioscience). A total of 30000 events were counted for each sample. Data were analyzed using FlowJo software (Tree Star).