Novel Vorinostat Analogues with Improved HDAC Inhibition, Stronger Cytotoxic Effect and Higher Selectivity Against Leukemias and Lymphomas

Histone deacetylase (HDAC) inhibitors are class of drugs used in the cancer treatment. Here, we developed a library of 19 analogues of Vorinostat, an HDAC inhibitor used in lymphomas treatment. In Vorinostat, we replaced the hydrophobic phenyl group with various tricyclic ‘caps’ possessing a central, eight-membered, heterocyclic ring, and investigated the HDAC activity and cytotoxic effect on the cancer and normal cell lines. We found that three out of the 19 compounds, based on dibenzo[b,f]azocin-6(5H)-one, 11,12-dihydrodibenzo[b,f]azocin-6(5H)-one and benzo[b]naphtho[2,3-f][1,5]diazocine-6,14(5H,13H)-dione scaffolds, showed better HDACs inhibition than the referenced Vorinostat. In leukemic cell line MV4-11 and in lymphoma cell line – Daudi three compounds showed lower IC50 values than Vorinostat. These compounds had higher activity and selectivity against MV4-11 and Daudi cell lines than reference Vorinostat. We also observed a strong correlation between HDACs inhibition and the cytotoxic effect. Cell lines derived from solid tumors: A549 (lung carcinoma) and MCF-7 (breast adenocarcinoma) as well as reference Balb/3T3 (normal murine fibroblasts) were less susceptible to compounds tested. Developed derivatives show superior properties than Vorinostat, thus they are applicable as selective agents for leukemia and lymphoma treatment.

For years our research group has been working on the design and synthesis of various mono-and polycyclic dilactam derivatives [89][90][91][92][93][94][95][96][97][98] with potential biological activity. The studies resulted in the discovery of tricyclic benzodiazepines exhibiting selective antileukemic effect [92][93][94]. These compounds could be treated as structural analogues of antitumor antibiotic Anthramycin. Recently, we focused on the development of novel synthetic methods leading to asymmetrically substituted tricylic lactam and dilacam compounds with central, eight-membered heterocyclic rings [95][96][97]. Such structures were used by us for the development of novel analogues of tricyclic drugs, which exhibited significant affinity to H1 receptors [98]. We envisioned then, that tricyclic heterocycles with two outer benzene rings and central, azocine or diazocine ring 7 ( Figure 2) could be useful scaffolds in the design of novel analogues of Vorinostat, HDACs inhibitor used for lymphoma treatment [99]. We decided to replace the phenyl group in Vorinostat with various tricyclic 'caps' and investigate the HDACs activity as well as potency and selectivity of the cytotoxic effect tested on the cancer and non-cancer cell lines.

Chemistry
Commercially available chemicals were of reagent grade and used as received. The reactions were monitored by thin layer chromatography (TLC), using silica gel plates (Kieselgel 60F254, E. Merck, Darmstadt, Germany). Column chromatography was performed on silica gel 60 M (0.040-0.063 mm, E. Merck, Darmstadt, Germany). Melting points are uncorrected and were measured on a Büchi (New Castle, DE, USA) Melting Point B-540 apparatus. All 1 H and 13 C NMR spectra were registered with a Bruker Avance III spectrometer operating at 500. 13 and 125.77 MHz for 1 H and 13 C, respectively, and equipped with a 5 mm probe head with Z-gradient coils. The experiments were performed using pulse programs from standard Bruker library for samples dissolved in CDCl3, DMSO-d6 or MeOH-d4. In each case spectra were calibrated at residual solvent resonances. High resolution mass spectra were performed by the Laboratory of Mass Spectrometry, Institute of Biochemistry and Biophysics PAS, on a LTQ Orbitrap Velos instrument, Thermo Scientific (Waltham, MA, USA). Synthetic procedures, physicochemical properties and spectra related to synthesized compounds are included in the supplementary file.

Histone Deacetylase Activity Assay (HDACs activity)
In order to measure histone deacetylase activity, Histone Deacetylase Activity Assay (HDACAA) (Sigma) was used (Catalog Number CS1010). The HDACAA kit is based on a two-step enzymatic reaction. The substrate for the reaction is a substituted peptide with an acetylated lysine residue and a bound fluorescent group. The first step of the reaction is deacetylation of the acetylated lysine side chain by the HDAC containing sample (HeLa cell extract). The second step is the cleavage of the deacetylated substrate with the developer solution and the release of the free highly fluorescent group. The experiment was carried out according to the manufacturer's protocol. The following concentrations were used: 0.1, 0.5, 1 and 2 µM. All calculations were done using Origin 9.0 software.

Histone Deacetylase 8 Activity Assay
In order to measure histone deacetylase 8 activity, Histone Deacetylase 8 Activity Assay (Sigma) was used. The HDAC8 acts with the supplied Developer, to deacetylate and then cleave the HDAC8 Substrate (R-H-K(Ac)-K(Ac)-AFC). The experiment was carried out according to the manufacturer's protocol. The following concentrations were used: 0.5, 1, 1.5, 2, 3 and 4 µM. All calculations were done using Origin 9.0 software.

Cytotoxicity Assay
Exponentially growing cells were seeded onto a 96-well plate at the density of 104 cells/well (MV4-11, Daudi, Balb/3T3), 0.75x104 cells/well (MCF-7) or 0.5x10 4 cells/well (A549), cultured for 24 h, and treated for 72 h with newly synthesized Vorinostat derivatives at concentrations of 50, 10, 2, 0.4, and 0.008 µM, or with medium alone as a control. Proliferation inhibition readings were made using MTT (MV4-11and Daudi) or SRB (MCF-7, A549 and Balb/3T3) method. SRB: 50 µL 50% cold trichloroacetic acid (TCA) solution was added to wells and incubated for 1 h. The plates were rinsed with distilled water, and after drying on a paper towel, 50 µL of a 0.1% solution of sulforodamine B in 1% acetic acid was added and incubated for 30 min at room temperature. Subsequently, the plates were washed with 1% acetic acid and after desiccation of excess acid, 150 µL of 10 mM TRIS was added. After another 30 min, the optical density of individual samples was read at 540 nm using a plate reader (Synergy H4 Hybrid Reader).
MTT: 20 µL MTT solution (5 mg/mL) was added to wells and plates were placed in an incubator. After 4 h of incubation at 37°C, 80 µL of lysis buffer (SDS, DMF and water) was added and incubation continued for 24 h. The optical density of individual samples was read at a wavelength of 570 nm using a plate reader (Synergy H4 Hybrid Reader).
All experiments were carried out at least three times with three replicates for each inhibitor concentration.

Compounds preparation
The appropriate ionization states at pH = 7.4 for all structures used in docking procedures were assigned using Epik software. 3D structures were generated in Ligprep software under default settings (force field used OPLS2005, retention of specified chiralities and generation of only one low energy ring conformation per ligand).

Docking protocol
All receptors were centered at zinc ion located in a binding pocket with the grid box size set to 25×25×25Å. Docking runs were performed in Glide software at the SP level under default settings (sampling nitrogen inversion, sampling ring conformations with energy window equal to 2.5 kcal·mol-1, penalizing nonplanar conformation of amides, up to 100 steps during energy minimization and performing post-docking optimization). Docking was carried up with one constraint -mandatory coordination of zinc ion.

Synthesis and HDAC inhibition
Since increasing the size of the hydrophobic 'cap' in the Vorinostat structure could have had an impact on the optimal length of the side chain terminated with hydroxamic acid, in the first part of our research, we decided to synthesize two homologous series of compounds and used two selected tricyclic 'caps': Previously obtained compounds 8 and 9 were treated with appropriate ω-bromoester, in the presence of sodium hydride resulting in intermediate products 10a-l. After chromatographic purification and isolation, esters 10a-l were treated with hydroxylamine hydrate which led to the final hydroxamic acids 7a-l.
A standard fluorimetric HDACs inhibition kit (Sigma-Aldrich) was to determine the inhibitory potency of novel Vorinostat analogues. It involves a two-step enzymatic reaction: deacetylation of the peptide acetylated lysine side chain by the HDACs containing HeLa cell extract followed by a cleavage of the deacetylated substrate by the developer solution and the release of the highly fluorescent group. We evaluated the efficacy of HDACs inhibition activity of 12 newly synthesized Vorinostat derivatives and observed a correlation between the side chain length and the HDACs inhibition activity ( Figure 3). The obtained results for Series 1 and 2 are presented in Table 1. In both series the compounds with five-carbon atom linkers (n=5) were the most active ones, additionally compound 7k demonstrated superior inhibitory activity in comparison to reference compound Vorinostat. Because the compounds with a five-carbon side chain turned out to be the most active, thus, in Series 3 ( Table 1, Scheme 2), only derivatives with five-carbon side chain were used.  In the next part of our research we decided to synthesize the compounds of Series 3 and to use seven tricyclic 'caps':  [95] (Scheme 2). Previously obtained compounds 11-17 were treated with ethyl 6-bromohexanoate in the presence of sodium hydride which resulted in the intermediate products 10m-t. After chromatographic purification and isolation, esters 10m-t were treated with hydroxylamine hydrate resulting in the final hydroxamic acids 7m-t. Using a standard fluorimetric HDAC inhibition kit (Sigma-Aldrich) to determine the inhibitory potency of the series 3 Vorinostat analogues, we observed that tested the compounds inhibited HDAC activity to a different degree, however, two of them, namely, 7p and 7t, showed even higher activity than the reference Vorinostat.
Among 19 newly synthesized Vorinostat analogues, IC50 was determined for the seven most active and promising derivatives: 7e, 7k, 7l, 7p-t ( Figure 4, Table 2   We performed in silico molecular docking which revealed that the HDAC8 homologue is the most promising target for compounds used: 7e, 7k, 7p and 7t (the most significant differences in the interaction pattern between active and inactive compounds along with Vorinostat, see Molecular Modelling section). For this reason, we evaluated the efficacy of the HDAC8 inhibition activity of 7e, 7k, 7p and 7t and Vorinostat as the reference (Table 2). We observed that all tested compounds exhibited a significant inhibitory effect on HDAC8. The lowest value, IC50 = 1.51±0.13 µM, was obtained for Vorinostat. All new compounds showed comparable, yet slightly lower effect with IC50 values in the range of 1.95±0.17 µM for 7t to 5.67±0.64 µM for 7e. Lower IC50 values obtained for the mixture of HDAC homologues, compared to the IC50 values for HDAC8 protein may indicate that other HDAC homologues found in the cell lysate are more susceptible to the tested compounds, and could show lower IC50, than HDAC8. An important factor that should also be taken into account are the different concentrations of individual HDAC homologues in the cell lysate influencing the total enzymatic activity of the mixture of HDAC homologues.

Cytotoxic activity and selectivity index
For the selection of the most promising cytotoxic agents, all 19 newly synthesized compounds, as well as the reference HDAC inhibitor -Vorinostat, were initially tested on two cancer cell lines: MV4-11 (biphenotypic B myelomonocytic leukemia) and Daudi (Burkitt's lymphoma) ( Table 3). Five compounds: 7k and 7p-t, exhibited IC50 below 1 µM, in the range of 0.093 µM (7t) to 0.692 µM (7s) on MV4-11 and in the range of 0.137 µM (7t) to 0.944 µM (7s) on Daudi. In the case of leukemic cell line MV4-11, three out of five compounds 7k, 7p and 7t showed lower IC50 values than Vorinostat (0.220, 0.200 and 0.093 µM, respectively, versus 0.636 µM). Again, with the lymphoma cell line -Daudi, 7k, 7p, and 7t showed lower (or comparable) IC50 values than Vorinostat (0.460, 0.318, and 0.137 µM, respectively, versus 0.493 µM). The most potent cytotoxic compounds, 7k and 7p-t, were further evaluated for their cytotoxic effect on two solid tumor cancer cell lines: A549 (lung carcinoma) and MCF-7 (breast adenocarcinoma). To determine the selectivity of the tested compounds, one reference cell line Balb/3T3 (mouse fibroblasts), derived from a non-cancerous cell line was also used. In the case of A549 cell line, two compounds: 7t and 7p showed lower IC50 values than Vorinostat (1.05, 1.21, respectively, versus 1.64 µM). Similarly, in the case of MCF-7 cell line two compounds: 7t and 7p showed lower (or comparable) IC50 values than Vorinostat (0.368, 0.661, respectively, versus 0.685 µM). Compounds 7p and 7t exhibited the strongest cytotoxic effect on cancer cell lines but the observed cytotoxicity also extended to the reference normal fibroblasts cell line. The two most active compounds 7t and 7p were also more toxic to Balb/3T3 than Vorinostat (0.69, 1.04, respectively, versus 1.42 µM). The obtained results showed a close correlation between HDAC inhibition and the cytotoxic effect of the tested compounds. The three most potent HDAC inhibitors: 7k, 7p, and 7t also showed the strongest cytotoxic effect on tested cell lines.
To determine the selectivity of tested compounds, we compared the cytotoxic effect observed for cancer cell lines (MV4-11, Daudi, A549, MCF7) and the reference line Balb/3T3. The selectivity indexes were calculated for the five most active compounds (7k, 7p-7t) and Vorinostat (Table 4). In all cases, the selectivity index of Vorinostat never exceeded the value of 3 and varied from 2.88 (Daudi) to 0.87 (A549). The highest selectivity indexes were obtained for compounds 7s (17.4 for MV4-11 and 12.75 for Daudi) and 7t (7.42 for MV4-11 and 5.05 for Daudi). The remaining compounds 7k, 7p and 7r (except of 7k and Daudi) also possessed higher selectivity indexes for MV4-11 and Daudi, than Vorinostat. We observed that while Vorinostat is slightly more selective for Daudi (lymphoma) than for MV4-11 (leukemia), our compounds exhibited better selectivity toward MV4-11 (leukemia) than for Daudi (lymphoma). In general, Vorinostat, as well as the newly synthesized compounds, exhibited relatively low selectivity toward solid tumor cancer cell lines (A549 and MCF-7) as compared to BALB/3T3 reference line.

Molecular modelling
To define and elucidate the binding modes of the synthetized compounds, molecular docking of these compounds in the active site of histone deacetylases 1, 2, 3, 4, 6, 7 and 8 (HDACs) was performed. Among all HDACs' structures stored in PDB only structures with the co-crystalized ligand and binding pocket exposed to the solvent were taken into account. Structures which were co-crystalized with Vorinostat and had the best resolution were preferred. Finally, one crystal structure per HDAC was selected (Table 5) and prepared in Protein Preparation Wizard [100][101], under default settings (coordination of zinc ion was set as constraint, centered on ligand). The three dimensional structure, conformation and protonation states of the evaluated compounds were generated by LigPrep (at pH 7.4) and Epik [101][102][103]. Finally, Glide [101][102][103][104][105][106] was used for docking each compound to every protein crystal. Each pose was ranked according to docking score (the lower value, the better), and the best scored pose per compound was chosen for further analysis. The majority of the evaluated compounds were docked to six out of seven HDAC types. Only HDAC3 was unable to form protein-ligand complex with the synthetized compounds and only Vorinostat was docked to this crystal. Binding modes for these evaluated compounds in binding pockets of HDACs 1, 3 and 7 were both very similar for all docked compounds. Active compounds (7e, 7k, 7t and 7p) docked to the HDAC1 did not have optimized geometry of zinc ion coordination despite the fact that they had very low (below -8, see Table S1, SI file) values of the scoring function. Both, active and inactive compounds share the same interaction pattern, involving amino acid residues H180, G151 and some of them even F208. In the binding mode of Vorinostat the additional interaction with residue D101 can be observed, but there is no hydrogen bond with H180 ( Figure 5A). Docking results for HDAC3 structure allowed for distinguishing between active from inactive compounds using scoring function (four active compounds are in the five top scoring compounds; values below -6, see Table S1, SI file), but the interaction profile of active and inactive compounds were very similar. Almost all compounds interacted with the zinc ion and residue G151, while contacts with other residues were rarely formed. Nevertheless, the hydrophobic cap of inactive compounds (e.g. for 7i) had a somewhat different orientation in the binding pocket than for active ones.
( Figure 5B). The interaction pattern for Vorinostat with the HDAC3 binding pocket was extremely poor -the compound interacted only with the zinc ion. For HDAC7 binding poses did not allow for separating active from inactive compounds. Almost all compounds interacted with the zinc ion and residue G151, but there were no specific interactions observed for active compounds only ( Figure 5C). In contrast, the binding mode of active compounds was different than inactive compounds for HDACs 2, 6 and 8. In the case of HDAC2, active compounds created hydrogen bonds with residues Y306 and Y207. Moreover, they interacted via π-π stacking interaction with F208 ( Figure 5D). These interactions were not commonly observed neither for inactive compounds nor for Vorinostat. Similar observations can be seen for HDAC6. Active compounds, except commonly observed interactions with residues Y306 and H143, interacted with residue H180 and formed at least one hydrogen bond with an aromatic cluster, i.e. F207 and F208 ( Figure 5E). This interaction with F207 or F208 enriches the interaction profile which characterizes Vorinostat in HDAC6 binding mode. In HDAC8 crystal structures Vorinistat interacts only with zinc ion and residue H143 and this interaction profile was shared by inactive compounds. Active compounds showed additional interactions with Y306 and F207 (hydrogen bonds or π-π stacking, Figure 5F), which were not observed for the remaining compounds.

Conclusions
We synthesized 19 novel HDAC inhibitors based on the Vorinostat structure. The introduction of a larger tricyclic hydrophobic 'cap' to the structure of Vorinostat in the place of the phenyl group was beneficial for biological properties and allowed for the development of compounds with improved HDAC inhibition, stronger cytotoxic effect and higher selectivity against leukemia and lymphoma cell lines. We observed that the enlargement of the hydrophobic 'cap' from a single benzene ring to a heterocyclic three-ring system forced a shortening of the length of the linker connecting the hydrophobic group to the hydroxamic acid residue. We also observed that the biological properties of the tested compounds (HDAC inhibition and the resulting cytotoxic effect) strongly depend on the tricyclic core used. We concluded that for optimal biological properties, an appropriate balance between the size/type of the hydrophobic group and the length of the side chain, terminated in the hydroxamic acid group, is necessary. We also observed a strong correlation between HDAC inhibition and cytotoxicity, so we concluded that HDAC inhibition should be the main factor responsible for the observed biological properties of the developed compounds. The tested HDAC inhibitors exhibited a stronger and more selective cytotoxic effect against MV4-11 and Daudi, while the cell lines derived from solid tumors and mouse fibroblasts proved to be much less sensitive to our compounds. Thus, we expect that they can find application as selective compounds against leukemias and lymphomas.