Identification of Highly Selective and Potent Histone Deacetylase 3 Inhibitors Using Click Chemistry-Based Combinatorial Fragment Assembly

To find histone deacetylase 3 (HDAC3)-selective inhibitors, a series of 504 candidates was assembled using “click chemistry”, by reacting nine alkynes bearing a zinc-binding group with 56 azide building blocks in the presence of Cu(I) catalyst. Screening of the 504-member triazole library against HDAC3 and other HDAC isozymes led to the identification of potent and selective HDAC3 inhibitors T247 and T326. These compounds showed potent HDAC3 inhibition with submicromolar IC50s, whereas they did not strongly inhibit other isozymes. Compounds T247 and T326 also induced a dose-dependent selective increase of NF-κB acetylation in human colon cancer HCT116 cells, indicating selective inhibition of HDAC3 in the cells. In addition, these HDAC3-selective inhibitors induced growth inhibition of cancer cells, and activated HIV gene expression in latent HIV-infected cells. These findings indicate that HDAC3-selective inhibitors are promising candidates for anticancer drugs and antiviral agents. This work also suggests the usefulness of the click chemistry approach to find isozyme-selective HDAC inhibitors.


Introduction
Histone protein complexes associate with DNA to form higherorder structures called chromatin. Approximately 150 base pairs of DNA are wrapped twice around an octamer of histones to form a nucleosome, the basic unit of chromatin. Core histones with Nterminal tails extending from the compact nucleosomal core particles can be acetylated or deacetylated at the epsilon position of lysine residues, thereby modifying histone-DNA and histonenon-histone protein interactions. The acetylation status of histone and non-histone proteins is controlled by two enzyme classes with opposing activities; histone acetyltransferases and histone deacetylases (HDACs) [1][2][3]. HDACs are hydrolases that modulate epigenetic gene expression through deacetylation of the N-acetyl lysine residues of histone and non-histone proteins. There are currently 18 known HDACs that are organized into four classes: class I HDACs (HDAC1, HDAC2, HDAC3, and HDAC8) and class IV HDAC (HDAC11) which are mainly localized to the nucleus; class II HDACs (HDAC4, HDAC5, HDAC6, HDAC7, HDAC9, and HDAC10) which shuttle between the nucleus and the cytoplasm; and class III HDACs (sirtuin 1-7), whose cellular localizations include various organelles [4]. Class I, II, IV HDACs are zinc-dependent enzymes, whereas class III HDACs are NAD +dependent enzymes [5][6][7][8].
Among the HDAC family members, HDAC3 is unique in that it is expressed in the nucleus, cytoplasm, or membrane, and it deacetylates histone and non-histone proteins such as NF-kB, myocyte enhancer factor 2, and Src kinase [9][10][11][12][13][14][15][16]. Furthermore, recent studies have indicated that HDAC3 is associated with several diseases including cancer, inflammation, and neurodegenerative disorders [17][18][19][20]. Therefore, HDAC3-selective inhibitors are of great interest not only as tools for probing the biological functions of HDAC3, but also as candidate therapeutic agents with potentially few side effects.
Although many efforts have been directed to the discovery of potent and selective HDAC inhibitors by numerous academic groups, as well as pharmaceutical companies, only a few HDAC3selective inhibitors have been reported [4] [21][22][23][24][25][26]. For example, HDAC3 is selectively inhibited by compounds 1 and 2 ( Figure 1) [27][28], but their HDAC3-inhibitory activity and selectivity are insufficient for their development as candidate therapeutic agents. In addition, while this research was carried out, RGFP966, a novel HDAC3-selective inhibitor, was reported, although the details of the inhibitor are unclear [29]. Therefore, there is still a need to find HDAC3 inhibitors that are more potent and selective than compounds 1 and 2.
We recently described the identification of potent HDAC8selective inhibitors from a triazole compound library generated by the use of Cu(I)-catalyzed azide-alkyne cycloaddition (CuAAC), a representative reaction in click chemistry [30][31][32][33]. Our results indicated that the click chemistry approach is useful for the discovery of isozyme-selective HDAC inhibitors. Following these findings, we performed a further click chemistry approach, seeking to find HDAC3-selective inhibitors more potent and selective than compounds 1 and 2. We describe here the rapid identification of potent and selective HDAC3 inhibitors via the use of click chemistry to generate a library of HDAC inhibitor candidates.

Enzyme Assays
Most HDAC inhibitors reported so far fit a three-motif pharmacophoric model, namely, a zinc-binding group (ZBG), a linker, and a cap group [21][22][23][24][25][26]. For instance, vorinostat (3) (Figure 2) [34] [35], a clinically used HDAC inhibitor, consists of hydroxamic acid (ZBG), which chelates the zinc ion in the active site, anilide (cap), which interacts with amino acid residues on the rim of the active site, and alkyl chain (linker), which connects the cap group and ZBG with an appropriate separation. Based on the typical HDAC inhibitor structure, we previously designed a library of candidate HDAC inhibitors in which the cap group and the ZBG are connected by a triazole-containing linker (Figure 2), and we identified potent HDAC8-selective inhibitors through screening of the library [30]. Following these findings, we expanded the library by the design and preparation of new alkynes with a ZBG and azides with a cap structure to find potent and selective HDAC3 inhibitors. For the preparation of the triazole library in this work, we designed and synthesized three alkynes Ak1-Ak3 with o-aminoanilide as the ZBG and 14 azides Az1-Az14 with an aromatic cap structure as building blocks for HDAC inhibitor candidate synthesis via CuAAC reaction. In designing alkynes  Ak1-Ak3, o-aminoanilide was selected as the ZBG because oaminoanilides tend to inhibit Class I HDACs [4]. Azides Az1-Az14 bearing an aromatic ring were expected to interact with aromatic amino acid residues such as Tyr and Phe which form the HDAC3 active pocket [36].
The routes used for the synthesis of compounds Az1-Az14, and Ak1-Ak3, which were prepared for this study, are shown in Figures 3,4,5,6. Figure 3 shows the preparation of aryl azides Az1-Az5, Az7, and Az11. The coupling reaction of aryl iodides 4-10 with sodium azide was carried out in the presence of CuI/Lproline catalyst to provide aryl azides Az1-Az5, Az7, and Az11 in 37-95% yield [37]. The routes for the synthesis of aryl azides Az6, Az8-Az10, and Az12 are illustrated in Figure 4. Treatment of anilines 11-15 with NaNO 2 under acidic conditions, followed by NaN 3 addition, yielded the desired aryl azides Az6, Az8-Az10, and Az12. The preparation of alkyl azides Az13 and Az14 is shown in Figure 5. Chlorides 16 and 17 were allowed to react with NaN 3 to afford alkyl azides Az13 and Az14. Figure 6 shows the preparation of alkynes Ak1-Ak3 bearing an o-aminoanilide moiety. Condensation of phenylenediamine 21 with the appropriate carboxylic acid chloride 18-20 gave o-aminoanilide derivatives Ak1-Ak3.
In all cases, disappearance of the alkynes and generation of the triazoles were confirmed by TLC. The generated triazolecontaining HDAC inhibitor candidates T1-T504 are shown in Figure 7.
These triazole compounds could be screened for HDACinhibitory activity without further purification [30] [39][40][41][42][43][44]. Since our final goal in this work is to identify compounds that selectively inhibit HDAC3 in cells, it is desirable to carry out in vitro enzyme assays in conditions similar to cellular environments. Because HDAC3 forms a complex with NCOR1 in cells [45], we used HDAC3/NCOR1 complex in in vitro HDAC3 assay. In addition, it is more important to find inhibitors that discriminate HDAC3 from HDAC1 and HDAC2 in cells. Therefore, as a primary in vitro screening for HDAC3 selectivity, we used total HDACs from HeLa nuclear extracts, in which the combined deacetylase activity of HDAC1 and HDAC2 is much higher than the activity of HDAC3 [46]. Initially, o-aminoanilides T1-T336 (10 mM) and hydroxamates T337-T504 (1 mM) were tested for inhibitory activity against HDAC3. In our HDAC3 assay, the IC 50 values of compounds 1-3 were 19 mM, .100 mM, and 0.27 mM, respectively. We therefore used compound 1 and vorinostat (3) as reference compounds in this assay. As shown in Figure 7, 59 oaminoanilides inhibited HDAC3 deacetylase activity by more than 90% at 10 mM, and 48 hydroxamates showed more than 60% HDAC3 inhibition at 1 mM. Next, we evaluated these 107     o-Aminoanilides inhibiting more than 90% of HDAC3 activity and hydroxamates inhibiting more than 60% of HDAC3 activity are indicated in red.
Vorinostat (3)  compounds for inhibitory activity against total HDACs from HeLa nuclear extracts, in which the deacetylase activity of HDAC1 and HDAC2 is much higher than that of HDAC3 [46]. While all of the hydroxamates displayed more than 70% inhibition of total HDACs at 1 mM (Figure 8), 11 o-aminoanilides showed less than 10% inhibition at 10 mM ( Figure 9) suggesting that these oaminoanilides exhibited HDAC3-selective inhibition. Further-more, we investigated the HDAC3-inhibitory activity of these 11 o-aminoanilides at 1 mM and 3 mM. Among them, T247 and T326 showed HDAC3 inhibition comparable to that of vorinostat (3) at both 1 mM and 3 mM (Table 1). These results indicated that T247 and T326 might be potent and selective HDAC3 inhibitors. Figure 10 illustrates the resynthesis of triazoles T247 and T326. Cu-catalyzed coupling of alkyne Ak5 with Az23 and Ak6 with  Az46 provided triazoles T247 and T326, respectively. The resynthesized compounds T247 and T326 were purified by column chromatography and recrystallization. The pure T247 and T326 were then examined for inhibitory effects on total HDACs, HDAC1, HDAC4, HDAC6, and HDAC8. The results of the enzyme assays are shown in Table 2. Compounds T247 and T326 displayed potent HDAC3-inhibitory activity, greater than that of compound 1 and comparable to that of vorinostat (3)

Molecular Modeling
The lowest energy conformation of T247, the most active HDAC3-selective inhibitor in this series, was obtained when it was docked into a model based on the crystal structure of HDAC3 (PDB code 4A69) [36], using the Molegro Virtual Docker software package. Inspection of the simulated HDAC3/T247 complex showed that the o-aminoanilide group coordinates to the Zn ion bidentately through its NH 2 and CO groups, and also forms two hydrogen bonds with His 134 and Gly 143 ( Figure 11). In addition, the phenyltriazole part of the inhibitor snugly fits the catalytic site. The phenyltriazole group of T247 lies in the hydrophobic tunnel formed by Phe 144, Phe 200, and Leu 266, where it can interact with the amino acid residues via hydrophobic interactions. There also appears to be a hydrophobic interaction of the thiophene ring of T247 with Pro 23 and Phe 144. The observed interactions between T247 and HDAC3 suggest the importance of the o-aminoanilide as a ZBG and a hydrogen-bondforming group for high potency. They also suggest the significance of the lipophilic aromatic rings of T247 for hydrophobic interactions. The triazole ring appears to orient the ZBG and hydrophobic group into appropriate geometry.

Cell-based Assays
To examine whether compounds T247 and T326 selectively inhibit HDAC3 in cells, we performed a cellular assay using western blot analysis. Since HDAC3 is known to catalyze the deacetylation of NF-kB [13][14], we initially examined the effects of the inhibitors on the acetylation levels of NF-kB in HCT116 cells. As we expected, T247 and T326 induced a dose-dependent increase of NF-kB acetylation, and their effect was greater than that of compound 1 and comparable to that of vorinostat (3) (Figure 12). Although T247 and T326 caused NF-kB acetylation, it has also been reported that NF-kB is deacetylated by HDAC1 and HDAC2 [47]. To examine whether T247 and T326 can distinguish HDAC3 from HDAC1 in cells, we next analyzed the effects of T247 and T326 on the acetylation levels p53, a substrate protein of HDAC1 [48]. As can be seen in Figure 12, while vorinostat (3), a non-selective HDAC inhibitor, induced nonselective acetylation of NF-kB and p53, the levels of acetylated p53 Table 1. HDAC3 inhibition in the presence of vorinostat (3), compound 1, and 11 o-aminoanilides at 1 mM and 3 mM. a

Conc.
HDAC3 inhibition (%)   3  1  T52  T199  T247  T251  T254  T261  T263  T266  T267  T318  T326   1 mM  83  9  55  59  89  75  55  75  74  73  80  77  86 3 mM 9 3 2 9 8 1 8 3 9 5 9 2 8 0 8 9 9 1 9 1 9 2 9 1 were not elevated in the presence of T247 and T326. These results indicate that T247 and T326 do not inhibit HDAC1 and selectively inhibit HDAC3 in the cells. In addition, T247 and T326 did not enhance the acetylation of a-tubulin, a substrate of HDAC6 [49] suggesting that T247 and T326 are HDAC3selective inhibitors in cell-based assays. Because it has been suggested that HDAC3 is highly expressed in human colon cancer cells and prostate cancer cells and is associated with the cancer cell growth [50][51], vorinostat (3), compound 1, T247, and T326 were tested in cell growthinhibition assays using human colon cancer HCT116 and prostate cancer PC-3 cell lines. The results are shown in Table 3. HDAC3selective inhibitors T247 and T326 showed clear growthinhibitory effects on both HCT116 and PC-3 cell lines. In particular, the cell growth-inhibitory activity of compound T247 and T326 was much greater than that of compound 1 and comparable to that of vorinostat (3). These results suggest that HDAC3-selective inhibitors might be useful in the treatment of colon cancers and prostate cancers.
We also examined the effects of T247 and T326 on latent HIVinfected cells, because it has been suggested that HDAC3 represses the transcription of HIV type 1 (HIV-1) genes in such cells [52]. HIV-1-infected OM10.1 cells were treated with 0.1 mM, 1 mM, and 10 mM compound 1, vorinostat (3), T247, and T326. Although compound 1, a weak HDAC3 inhibitor, did not show any activity, vorinostat (3), T247, and T326 significantly stimulated HIV-1 expression at 1 mM and/or 10 mM (Figure 13). Compound T326 was less active at 10 mM due to cytotoxicity. These data suggest that the combination of HDAC3-selective inhibitor and other anti-HIV agents may be useful in the treatment of HIV infection [53][54][55].
In summary, we have designed a 504-membered triazolecontaining HDAC inhibitor candidate library and prepared it by means of CuAAC reaction between nine alkynes and 56 azides. Two compounds, T247 and T326, were hit as HDAC3-selective inhibitors by screening of the 504 library compounds. Compounds T247 and T326 showed potent inhibition of HDAC3 with IC 50 values of 0.24 mM and 0.26 mM, respectively, but did not inhibit other HDAC isozymes even at 100 mM. The molecular modeling study of T247 with HDAC3 suggested the importance of the o-  aminoanilide as a ZBG and a hydrogen-bond-forming group, and of the lipophilic part having three aromatic rings for hydrophobic interactions. In cellular assays, T247 and T326 induced a selective increase of acetylated NF-kB, suggesting that they are cellularly active HDAC3-selective inhibitors. T247 and T326 also inhibited the growth of colon cancer HCT116 and prostate cancer PC-3 cell lines, and stimulated HIV-1 gene expression in latent HIV-1-infected OM10.1 cells. We believe that T247 and T326 are the most potent HDAC3-selective inhibitors reported so far. The findings presented here should provide a basis for constructing new tools to probe the biology of HDAC3 and for developing new strategies to treat cancer and HIV-1 infection. Many groups have ongoing research programs to find selective inhibitors of HDAC isozymes, however, there has been no reported isozyme-selective inhibitors of HDAC1, 2, 5, 7, 9, 10, and 11, although the isozymes have been reported to be crucial for biological events and be responsible for several disease states [4].
Our methodology using click chemistry could be used to find not only HDAC3-and HDAC8-selective inhibitors, but also other isozyme-selective inhibitors. We believe that selective inhibitors against the HDAC isozymes will be discovered using this click

Chemistry
General. Melting points were determined using a Yanagimoto micro melting point apparatus or a Büchi 545 melting point apparatus and were left uncorrected. Proton nuclear magnetic resonance spectra ( 1 H NMR), carbon nuclear magnetic resonance spectra ( 13 C NMR) were recorded on a JEOL JNM-LA500, JEOL JNM-A500 or BRUKER AVANCE600 spectrometer in the indicated solvents. Chemical shifts (d) are reported in parts per million relative to the internal standard tetramethylsilane. Elemental analysis was performed with a Yanaco CHN CORDER NT-5 analyzer, and all values were within 60.4% of the calculated values. Fast atom bombardment (FAB) mass spectra were recorded on a JEOL JMS-SX102A mass spectrometer. GC-MS analyses were performed on a Shimadzu GCMS-QP2010. IR spectra were measured on a Shimadzu FTIR-8400S spectrometer. Reagents and solvents were purchased from Aldrich, Tokyo Kasei Kogyo, Wako Pure Chemical Industries, and Kanto Kagaku and used without purification. Flash column chromatography was performed using silica gel 60 (particle size 0.046-0.063 mm) supplied by Merck.

Western Blot Analysis
HCT116 human colon cancer cells were purchased from American Type Culture Collection (ATCC, Manassas, VA, U.S.A.) and cultured in McCoy's 5A culture medium containing penicillin and streptomycin, which was supplemented with fetal bovine serum as described in the ATCC instructions. HCT116 cells (1.0610 5 ) were treated for 8 h with 20 mM etoposide and samples at the indicated concentrations in McCoy's 5A 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 AnykD SDS-polyacrylamide gels and then transferred onto nitrocellulose membranes (Bio-Rad Laboratories). After having been blocked for 30 min with Tris-buffered saline (TBS) containing 3% skimmed milk, the transblotted membranes were incubated overnight at 4uC with acetyl NF-kB antibody (CST) (1:1000 dilution), NF-kB antibody (CST) (1:1000 dilution), acetyl a-tubulin antibody (Sigma) (1:2000 dilution), atubulin antibody (Sigma) (1:2000 dilution), acetyl p53 antibody (CST) (1:500 dilution) or p53 antibody (CALBIOCHEM) (1:500 dilution) in TBS containing 3% skimmed milk. The membrane was probed with the primary antibody, then washed twice with TBS, incubated with sheep anti-rabbit IgG-horseradish peroxidase conjugates (diluted 1:1000 for acetyl NF-kB, 1:2000 for NF-kB or 1:500 for acetyl p53) or donkey anti-mouse IgG-horseradish peroxidase conjugates (diluted 1:5000 for acetyl a-tubulin, 1:5000 for a-tubulin, or 1:500 for p53) for 1.5 h at room temperature, and again washed twice with TBS and once with TBS-Tween 20 (TBS-T). The immunoblots were visualized by enhanced chemiluminescence.
Cell growth inhibition assay. The cells were plated at the initial density of 5,000 cells/well (50 mL/well) in 96-well plates in medium culture and exposed to inhibitors for 48 h in an incubator at 37uC in 5% CO 2 in air. A solution (5 mg/mL) of 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was added (10 mL/well) and incubation was continued for 3 h. The solubilized dye was quantified by colorimetric reading at 570 nm. The absorbance values of control wells (C) and test wells (T) were measured. The 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: Viral p24 antigen assay. The p24 antigen level in the cell culture supernatant was measured by p24 antigen capture ELISA assay using a commercial kit (RETRO-TEK HIV-1 p24 Antigen ELISA kit; Zepto Metrix, Buffalo, NY, USA) according to the method reported in ref [54].
Molecular modeling. The X-ray structures of HDAC3 and HDAC8 (PDB code 4A69 and 1T64, respectively) were used as the target structures for docking. Protein preparation, receptor grid generation and ligand docking were performed using the Molegro Virtual Docker software package. Compound T247 was docked into the active site of the protein and was located in a position where the amino group of T247 can interact with the zinc ion. The standard precision mode of Molegro Virtual Docker was used to determine favorable binding poses, which allowed the ligand conformation to be flexibly explored while holding the protein as a rigid structure during docking.