Unexpected Enhancement of HDACs Inhibition by MeS Substitution at C-2 Position of Fluoro Largazole

Given our previous finding that fluorination at the C18 position of largazole showed reasonably good tolerance towards inhibitory activity and selectivity of histone deacetylases (HDACs), further modification on the valine residue in the fluoro-largazole’s macrocyclic moiety with S-Me l-Cysteine or Glycine residue was performed. While the Glycine-modified fluoro analog showed poor activity, the S-Me l-Cysteine-modified analog emerged to be a very potent HDAC inhibitor. Unlike all previously reported C2-modified compounds in the largazole family (including our recent fluoro-largazole analogs) where replacement of the Val residue has failed to provide any potency improvement, the S-Me l-Cysteine-modified analog displayed significantly enhanced (five–nine-fold) inhibition of all the tested HDACs while maintaining the selectivity of HDAC1 over HDAC6, as compared to largazole thiol. A molecular modeling study provided rational explanation and structural evidence for the enhanced inhibitory activity. This new finding will aid the design of novel potent HDAC inhibitors.

Mar. Drugs 2020, 18 Largazole, as a potent and selective Class I HDAC inhibitor, was discovered in 2008 by Leusch et al. [8,9]. Due to its superior anticancer properties, this marine natural product has attracted widespread attention in the medicinal chemistry community [10][11][12][13][14][15]. In order to improve the inhibitory activity and selectivity of HDACs, great efforts have been made to modify the macrocyclic moiety, the hydrophobic linker and the warhead of largazole. Thus far, almost all the modifications on the linker and the warhead cause obvious losses in activity against HDACs, possibly due to preventing the thiol group to efficiently penetrate into the narrow channel containing the Zn 2+ ion of HDACs and to coordinate it with the optimal geometry. In contrast, the modifications on the macrocyclic moiety seem to be allowable to some extent. Among them, certain variations in valine residue (Val) at the C2 position of largazole may maintain the HDACs inhibition despite with a little sacrifice in its potency or isoenzyme selectivity (Figure 1). For example, Hong and Leusch's group reported that replacement of Val with Ala decreased the activity by three-fold in growth inhibition of HCT-116 [16]. Subsequently, they introduced aromatic (Phe, Tyr), acidic (Asp) or basic (His) amino acid residues at that position, but all the analogs exhibited reduced activity in HDAC1 [17]. Based on the hypothesis of potentially hydrophobic interactions between Val and the surrounding residues of HDAC1 (Tyr196 and Leu263), Jiang et al. replaced Val with Leu, Phe and Tyr, reporting that Tyr had about five-fold higher GI 50 values of HCT116 and A549 cells compared to largazole despite with an increasing selectivity over normal cells [18]. Ganesan's and Williams' groups reported that replacement with Gly or Pro resulted in a significantly decreased inhibition against HDACs, respectively [19,20]. Recently, Hong's group investigated again the effect of the replacement of Val with Phe, Tyr, Asp or His on both the activity and selectivity of the class I HDACs, and the observed His substitution showed comparable activity, and a slight selectivity towards HDAC1 over HDAC2-3. This slight selectivity was hypothesized to be resulted from a possible hydrogen-bond ability of His. However, when they replaced Val with a set of residues tagged with the terminal amine or amides, the results suggested that the hydrogen bonding interaction did not play an essential role in HDAC inhibition [21]. All these works indicated that the structural optimization at the C2 position appeared not to improve the potency of HDACs inhibition.
Mar. Drugs 2020, 18, 344 2 of 12 Largazole, as a potent and selective Class I HDAC inhibitor, was discovered in 2008 by Leusch et al. [8,9]. Due to its superior anticancer properties, this marine natural product has attracted widespread attention in the medicinal chemistry community [10][11][12][13][14][15]. In order to improve the inhibitory activity and selectivity of HDACs, great efforts have been made to modify the macrocyclic moiety, the hydrophobic linker and the warhead of largazole. Thus far, almost all the modifications on the linker and the warhead cause obvious losses in activity against HDACs, possibly due to preventing the thiol group to efficiently penetrate into the narrow channel containing the Zn 2+ ion of HDACs and to coordinate it with the optimal geometry. In contrast, the modifications on the macrocyclic moiety seem to be allowable to some extent. Among them, certain variations in valine residue (Val) at the C2 position of largazole may maintain the HDACs inhibition despite with a little sacrifice in its potency or isoenzyme selectivity (Figure 1). For example, Hong and Leusch's group reported that replacement of Val with Ala decreased the activity by three-fold in growth inhibition of HCT-116 [16]. Subsequently, they introduced aromatic (Phe, Tyr), acidic (Asp) or basic (His) amino acid residues at that position, but all the analogs exhibited reduced activity in HDAC1 [17]. Based on the hypothesis of potentially hydrophobic interactions between Val and the surrounding residues of HDAC1 (Tyr196 and Leu263), Jiang et al. replaced Val with Leu, Phe and Tyr, reporting that Tyr had about five-fold higher GI50 values of HCT116 and A549 cells compared to largazole despite with an increasing selectivity over normal cells [18]. Ganesan's and Williams' groups reported that replacement with Gly or Pro resulted in a significantly decreased inhibition against HDACs, respectively [19,20]. Recently, Hong's group investigated again the effect of the replacement of Val with Phe, Tyr, Asp or His on both the activity and selectivity of the class I HDACs, and the observed His substitution showed comparable activity, and a slight selectivity towards HDAC1 over HDAC2-3. This slight selectivity was hypothesized to be resulted from a possible hydrogen-bond ability of His. However, when they replaced Val with a set of residues tagged with the terminal amine or amides, the results suggested that the hydrogen bonding interaction did not play an essential role in HDAC inhibition [21]. All these works indicated that the structural optimization at the C2 position appeared not to improve the potency of HDACs inhibition. Recently, we discovered that the modification on the linker of largazole was allowable in some cases (Figure 1) [22][23][24]. For example, fixing fluorine at C18 position of the linker could keep almost the same inhibitory activity and selectivity of HDACs. Subsequently, the replacement of Val with Phe gave slightly reduced inhibition of HDACs, similar to the previous observations reported for largazole [23]. As our research continued, we discovered to our surprise that the replacement of Val with S-Me Cys significantly enhanced the inhibition of HDACs, whereas the replacement with Gly significantly decreased the activity. With an expanded research program under way based on this observation, we would like to communicate here our preliminary results and provide rational explanation for the unexpected results through a molecular modeling study. To the best of our knowledge, the S-Me Cys Recently, we discovered that the modification on the linker of largazole was allowable in some cases ( Figure 1) [22][23][24]. For example, fixing fluorine at C18 position of the linker could keep almost the same inhibitory activity and selectivity of HDACs. Subsequently, the replacement of Val with Phe gave slightly reduced inhibition of HDACs, similar to the previous observations reported for largazole [23]. As our research continued, we discovered to our surprise that the replacement of Val with S-Me Cys significantly enhanced the inhibition of HDACs, whereas the replacement with Gly significantly decreased the activity. With an expanded research program under way based on this observation, we would like to communicate here our preliminary results and provide rational explanation for the unexpected results through a molecular modeling study. To the best of our knowledge, the S-Me Cys modification is the first reported example where replacement of the Val residue has led to significant Mar. Drugs 2020, 18, 344 3 of 12 enhancement in HDACs inhibition. We believe this insight may enable the design of highly potent HDAC inhibitors.

Chemistry
In our previous paper [23], we synthesized the two fluoroolefin analogs of largazole with Val and Phe residue at C2 position. The key macrolactamization was chosen at N2 position, but the yields were poor (less than 31% yield). As an alternative, the key macrolactamization for the two new fluoroolefin analogs in this study was arranged at N14 position with an intend to improve the yield of the key step. The evolved synthetic route is depicted in Scheme 1.
The key fluoro fragment (8) was prepared from acrolein via four steps according to our previous method [23]. Alcoholysis of 8 with TMSCH 2 CH 2 OH in the presence of DMAP provided 9 in good yield (67%). Condensation of 9 with S-Me Fmoc-l-Cystine or Fmoc-Glycine in the presence of EDCI, DMAP and DIEPA afforded ester 10c and 10d (yield: 75% for 10c). Removal of Fmoc group of 10c or 10d with Et 2 NH in CH 2 Cl 2 and subsequent condensation with the acid (11) resulted in the linear precursors (12c, 84%; 12d, 41% based on 9). Deprotection of Boc group of 10c or 10d with CF 3 CO 2 H in CH 2 Cl 2 and subsequent release of free acid provided the linear depsipeptides (13c and 13d) exposed at the N-and C-terminus, respectively. The depsipeptides were then subjected to the optimal cyclization condition (HATU/HOAT/DIPEA in anhydrous CH 2 Cl 2 solution with a diluted concentration of about 0.001 M). The macrolactamization yields were 48% for 14c and 40% for 14d in two steps, respectively. Compared to our previous cyclization at N2 position, cyclizing at N14 position afforded apparently improved yield. Following our previously reported procedure [24], the free thiol 15c and 15d were obtained through deprotection of Trt group from 14c and 14d, respectively. The subsequent acylation with n-C 7 H 15 COCl under the standard condition led to the final fluoro analogs 16c and 16d (yields: 57%, 40%), respectively.
Mar. Drugs 2020, 18, 344 3 of 12 modification is the first reported example where replacement of the Val residue has led to significant enhancement in HDACs inhibition. We believe this insight may enable the design of highly potent HDAC inhibitors.

Chemistry
In our previous paper [23], we synthesized the two fluoroolefin analogs of largazole with Val and Phe residue at C2 position. The key macrolactamization was chosen at N2 position, but the yields were poor (less than 31% yield). As an alternative, the key macrolactamization for the two new fluoroolefin analogs in this study was arranged at N14 position with an intend to improve the yield of the key step. The evolved synthetic route is depicted in Scheme 1.
The key fluoro fragment (8) was prepared from acrolein via four steps according to our previous method [23]. Alcoholysis of 8 with TMSCH2CH2OH in the presence of DMAP provided 9 in good yield (67%). Condensation of 9 with S-Me Fmoc-L-Cystine or Fmoc-Glycine in the presence of EDCI, DMAP and DIEPA afforded ester 10c and 10d (yield: 75% for 10c). Removal of Fmoc group of 10c or 10d with Et2NH in CH2Cl2 and subsequent condensation with the acid (11) resulted in the linear precursors (12c, 84%; 12d, 41% based on 9). Deprotection of Boc group of 10c or 10d with CF3CO2H in CH2Cl2 and subsequent release of free acid provided the linear depsipeptides (13c and 13d) exposed at the N-and C-terminus, respectively. The depsipeptides were then subjected to the optimal cyclization condition (HATU/HOAT/DIPEA in anhydrous CH2Cl2 solution with a diluted concentration of about 0.001 M). The macrolactamization yields were 48% for 14c and 40% for 14d in two steps, respectively. Compared to our previous cyclization at N2 position, cyclizing at N14 position afforded apparently improved yield. Following our previously reported procedure [24], the free thiol 15c and 15d were obtained through deprotection of Trt group from 14c and 14d, respectively. The subsequent acylation with n-C7H15COCl under the standard condition led to the final fluoro analogs 16c and 16d (yields: 57%, 40%), respectively. Scheme 1. Synthesis of the analogs of Largazole. The red R 1 is the different substituent and the green number is the atom code in the molecule. Scheme 1. Synthesis of the analogs of Largazole. The red R 1 is the different substituent and the green number is the atom code in the molecule.

Biology
It is well-known that largazole is a prodrug species for a beneficial cell permeability and its free thiol is indeed the activated species for the inhibition against HDACs [10]. Given the instability of the free thiol for storage, the in vitro cell assays were used as the initial screening by our well-established cell tests (A549, HCT116, MDA-MB-231 and SK-OV-3 cells), aiming at speedy identification of potent HDAC inhibitors. The preliminary results were shown in Table 1, using largazole as the positive control compound.
Compared with largazole, 16c displayed the obvious growth inhibition towards A549, HCT116 and SK-OV-3 cells with IC 50 values ranged with 1.41-3.13 µM, but not towards MDA-MB-231 (IC 50 : >10 µM). In contrast, 16d almost lost the growth inhibition towards most of those cells, and only maintained a slight growth inhibition towards HCT116 (IC 50 : 6.62 µM), indicating that 16d was a less potent HDAC inhibitor, which was consistent with previous reports [19,20], where both Ganesan's and Williams' groups demonstrated that the replacement of Val with Gly in the case of largazole resulted in a significantly decrease of inhibition against HDACs. To further confirm the growth inhibition of 16c in cellular assays, we tested 16c and largazole with other cells. As shown in Table 2, 16c displayed growth inhibition against Hela, Eca-109, Bel 7402 and U937 cells with high potencies that are quite identical to largazole. Based on the obvious growth inhibition of 16c in in vitro cell assays, we next performed in vitro enzyme assays with the correspond active species 15c, and the data were compared with those of our previous compounds. The results are shown in Table 3.
In sharp contrast to the cellular assays where 16c exhibited identical or slightly inferior activities as compared to largazole, in the enzymatic assays, compound 15c surprisingly displayed significantly increased inhibition towards HDACs, with the IC 50 values of HDAC 1, 2, 3, 8 and 6 being 0.27, 1.33, 2.33, 0.44 and 23.91 nM, respectively. It was five-nine-fold more potent towards all the tested HDACs when compared with largazole thiol. Moreover, the selectivity of HDAC1 over HDAC6 remained unchanged (selectivity: 89). Notably, we demonstrated earlier that the fluoro analogs (15a and 15b) exhibited slightly less activities than largazole thiol, which was consistent with all previous observations made with substitutions at C2 for largazole [16][17][18][19][20][21]. We wanted to compare 15c with similar largazole compounds. However, a literature search revealed no prior report on modification at C2 position with a sulphur-containing substituent. 15c appeared to be the first example of such compounds in the largazole family. The surprisingly enhanced activity of 15c might be attributed to the variation in Val at the C2 position, suggesting that the MeS group of 15c played an important role in the interaction between 15c and HDACs. This was confirmed subsequently by the results from a molecular modeling study. in the interaction between 15c and HDACs. This was confirmed subsequently by the results from a molecular modeling study.

Molecular Modeling Study
To gain some structural insight on the increased inhibitory effects of 15c on the tested HDACs, molecular docking was performed by using MOE 2019 with MMF94 force field. The crystal structures of HDACs were downloaded from the Protein DataBank (PDB, http://www.rcsb.org), and were used to investigate the binding modes of 15c with HDAC1, HDAC6 and HDAC8, respectively (PDB code: HDAC1, 5ICN; HDAC6, 5EDU; HDAC8, 4RN0).
The binding modes of largazole thiol in HDAC1, HDAC6 and HDAC8 indicated that the thiol side chain could coordinate to the catalytic Zn 2+ ion and the overall metal coordination geometry was nearly perfectly tetrahedral [25]. Addtionally, largazole thiol formed hydrogen bond interactions with ASP99 in HDAC1 and TYR306 in HDAC8, respectively (Figure 2A-C). For 15c, which was derived from largazole thiol, additional hydrogen bonds were formed between sulphur atom in the methylthio group and ASN95 in HDAC1 and SER568 in HDAC6, respectively ( Figure 2D,E). Interestingly, the introduction of MeS substitution resulted in orientation change of the ester bond at C1 position, and then made an oxygen atom of the carboxyl group form a new hydrogen bond with HIS180 of HDAC8 ( Figure 2F). These additional interactions are likely the factors that have led to significant improvement in 15c's potency towards HDACs. These findings from molecular modeling study confirmed our hypothesis about the key role that MeS group played, and provided telling structural evidence for the significantly increased inhibition of 15c towards HDACs.

Molecular Modeling Study
To gain some structural insight on the increased inhibitory effects of 15c on the tested HDACs, molecular docking was performed by using MOE 2019 with MMF94 force field. The crystal structures of HDACs were downloaded from the Protein DataBank (PDB, http://www.rcsb.org), and were used to investigate the binding modes of 15c with HDAC1, HDAC6 and HDAC8, respectively (PDB code: HDAC1, 5ICN; HDAC6, 5EDU; HDAC8, 4RN0).
The binding modes of largazole thiol in HDAC1, HDAC6 and HDAC8 indicated that the thiol side chain could coordinate to the catalytic Zn 2+ ion and the overall metal coordination geometry was nearly perfectly tetrahedral [25]. Addtionally, largazole thiol formed hydrogen bond interactions with ASP99 in HDAC1 and TYR306 in HDAC8, respectively (Figure 2A-C). For 15c, which was derived from largazole thiol, additional hydrogen bonds were formed between sulphur atom in the methylthio group and ASN95 in HDAC1 and SER568 in HDAC6, respectively ( Figure 2D,E). Interestingly, the introduction of MeS substitution resulted in orientation change of the ester bond at C1 position, and then made an oxygen atom of the carboxyl group form a new hydrogen bond with HIS180 of HDAC8 ( Figure 2F). These additional interactions are likely the factors that have led to significant improvement in 15c's potency towards HDACs. These findings from molecular modeling study confirmed our hypothesis about the key role that MeS group played, and provided telling structural evidence for the significantly increased inhibition of 15c towards HDACs.

Chemistry
The chemicals and reagents were purchased from Acros (Shanghai, China), Alfa Aesar (Shanghai, China), and National Chemical Reagent Group Co. Ltd., P. R. China (Shanghai, China), and used without further purification. Anhydrous solvents (THF, MeOH, DMF, CH2Cl2 and CH3CN) used in the reactions were dried and freshly distilled before use. All the reactions were carried out under Ar atmosphere, otherwise stated else. The progress of the reactions was monitored by TLC (silica-coated glass plates) and visualized under UV light, and by using iodine or phosphomolybdic acid. Melting points were measured on an SGW X-4 microscopy melting point apparatus without

Chemistry
The chemicals and reagents were purchased from Acros (Shanghai, China), Alfa Aesar (Shanghai, China), and National Chemical Reagent Group Co. Ltd., P. R. China (Shanghai, China), and used without further purification. Anhydrous solvents (THF, MeOH, DMF, CH 2 Cl 2 and CH 3 CN) used in the reactions were dried and freshly distilled before use. All the reactions were carried out under Ar atmosphere, otherwise stated else. The progress of the reactions was monitored by TLC (silica-coated glass plates) and visualized under UV light, and by using iodine or phosphomolybdic acid. Melting points were measured on an SGW X-4 microscopy melting point apparatus without correction. 1 H NMR and 13 C NMR spectra were recorded either on a 400 MHz Varian Instrument at 25 • C or 600 MHz Bruker Instrument at 25 • C, using TMS as an internal standard, respectively. Multiplicity is tabulated as s for singlet, d for doublet, dd for doublet of doublet, t for triplet, and m for multiplet. The original spectra of the relative compounds could be found in Supplementary Materials. HRMS spectra were recorded on Finnigan-Mat-95 mass spectrometer, equipped with ESI source. Largazole, its free thiol and key intermediates (compound 8 and 11) were prepared according to our previous method [22][23][24].

Molecular Modeling Study
The known crystal complexes of HDAC1, HDAC6, HDAC8 and their ligand (PDB code: 5ICN, 5EDU, 4RN0) were obtained from PDB (http://www.rcsb.org). Molecular docking simulations in the HDACs were run using the MOE 2019 (Molecular Operating Environment, Chemical Computing Group, Montreal, Quebec, Canada) due to its universality and very fast speed. Ligands were prepared with the ChemBio3D Ultra 14.0 (PerkinElmer, Waltham, MA, USA), followed by MM2 energy minimization. Protein structures were also prepared with the MOE, which could automatically add hydrogen atoms to proteins by explicitly considering the protonation state of histidine and optimize the force field. All crystal water, small ligands and cofactors except HEM were removed. After this step, the binding sites were deduced from the known crystal complexes and the ligands were docked to the prepared proteins through flexible docking mode. Top scoring function poses were selected as representative of the simulations and were displayed with Open-Source PyMOLTM 1.8X software (Schrödinger, Ltd, New York, NY, USA).

Conclusions
Given our previous finding that fluorination at the C18 position of largazole showed good tolerance towards the inhibitory activity and selectivity of HDACs, the current study investigated further modifications on the valine residue in its macrocyclic moiety with S-Me Cysteine or Glycine residue. While the Glycine-modified fluoro analog showed poor activity, the S-Me l-Cysteine-modified analog emerged to be a very potent HDAC inhibitor. Unlike all previously reported C2-modified compounds in the largazole family (including our recent fluoro-largazole analogs) where replacement of the Val residue has failed to provide any potency improvement, the S-Me l-Cysteine-modified analog displayed significantly enhanced (five-nine-fold) inhibition of all the tested HDACs while maintaining the selectivity of HDAC1 over HDAC6, as compared to largazole thiol. Molecular modeling study provided rational explanation and structural evidence for the enhanced activity. This new finding will aid the design of novel potent HDAC inhibitors. An expanded research program is currently under way to investigate largazole analogs bearing similar structural characteristics.