1-(Piperidin-3-yl)thymine amides as inhibitors of M. tuberculosis thymidylate kinase

Abstract A series of readily accessible 1-(piperidin-3-yl)thymine amides was designed, synthesised and evaluated as Mycobacterium tuberculosis TMPK (MtbTMPK) inhibitors. In line with the modelling results, most inhibitors showed reasonable MtbTMPK inhibitory activity. Compounds 4b and 4i were slightly more potent than the parent compound 3. Moreover, contrary to the latter, amide analogue 4g was active against the avirulent M. tuberculosis H37Ra strain (MIC50=35 µM). This finding opens avenues for future modifications.


Introduction
Tuberculosis (TB) is a severe air transmitted infectious disease mainly caused by Mycobacterium tuberculosis (Mtb) and causes significant morbidity and mortality worldwide 1 . It is estimated that 10.4 million people fell ill with TB in 2017, and 1.6 million people succumbed the infection 2 . Standard treatment of drug-sensitive TB normally lasts 6 months, including a 2-month course with at least three first-line drugs and a 4-month course with at least Isoniazid (INH) and Rifampicin (RIF), resulting in poor treatment compliance and a high financial burden. Moreover, the increasing number of multidrug-resistant TB (MDR-TB), extensively drugresistant TB (XDR-TB) and rifampicin-resistant TB (RR-TB) makes the treatment even more challenging 3,4 . New drugs acting on yet unexplored targets/pathways are desirable to achieve worldwide TB control.
Thymidine monophosphate kinase (TMPK), an enzyme at the junction of the de novo (involving thymidylate synthase) and salvage pathway, is responsible for the conversion of thymidine monophosphate (TMP) to thymidine diphosphate (TDP) 5,6 . Further phosphorylation leads to the formation of thymidine triphosphate (TTP), which is indispensable for DNA synthesis 7 . Mycobacterium tuberculosis TMPK (MtbTMPK) shows low (22%) sequence identity with the human isozyme and has a unique catalytic mechanism 8,9 , further supporting it as an attractive target for developing new anti-TB drugs. Both industrial and academic efforts have afforded several potent MtbTMPK inhibitors in the past two decennia 10-19 , including thymidine-like and non-nucleoside inhibitors (Figure 1) [20][21][22][23][24] .
Using the Gram-positive bacterial thymidylate kinase inhibitor 1 (Figure 2) disclosed by Astra Zeneca 7 as a starting point, our research group is trying to identify potent non-nucleoside MtbTMPK inhibitors 25 . These efforts resulted in the identification of the regiomeric racemates 2 and 3, of which the latter showed the best MtbTMPK inhibitory activity and was therefore selected for further optimisation. In this study, we synthesised a series of D-ring substituted analogues of 3 and evaluated these as MtbTMPK inhibitors (Figure 2).

Chemistry
All synthetic compounds mentioned in this study were visualised under UV light at 254 nm or stained by corresponding reagents, and purified by column chromatography (CC) on a Reveleris X2 (Grace) automated flash unit. 1 H and 13 C NMR spectral data were obtained from a Varian Mercury 300/75 MHz spectrometer at 300 K using TMS as an internal standard. Structural assignment was confirmed with the assistance of COSY, HSQC and HMBC. High-resolution mass spectrometry was performed on a Waters LCT Premier XETM time of flight (TOF) mass spectrometer equipped with a standard electrospray ionisation (ESI) and modular LockSpray TM interface. The purity of the tested compounds was determined by LC-MS analysis (Waters Alliance 2695 XE separation module).

Molecular modelling
The modelling was conducted using AutoDock vina and AutodockTools-1.5.6 26 with publicly available X-ray structure of the MtbTMPK (PDB entry 1G3U 9 and 5NQ5 27 ). All PDB files of ligand were generated in ChemDraw 3 D 16.0 after being minimised the energy. Centred on MtbTMPK active site PHE70 CE2, the prepared PDBQT files of ligands and receptors were docked using a grid spacing of 0.375 and 60 Â 60 Â 60 number of grid points. Each ligand was docked three times in autodock vina by Lamarckian 4.2 method, affording total 60 possible conformations. All vinadock results were viewed in Chimaera, further analysed and validated through the predicted free energy in combination with possible interactions in LigPlus.

MtbTMPK assay
MtbTMPK was overexpressed in Escherichia coli and purified by a two-step procedure as previously described 8 . MtbTMPK activities of all compounds were measured using the spectrophotometric assay reported by Blondin et al 28 . All compounds were tested at varying concentration from 0.006 to 0.12 mM with constant concentrations of ATP (0.5 mM) and dTMP (0.05 mM). The studies were done at 30 C in a medium containing 50 mM Tris-HCl, 2 mM MgCl 2 , pH 7.4, 50 mM KCl, 1 mM phosphoenol pyruvate, 0.2 mM NADH and two units each of coupling enzymes (lactate dehydrogenase, pyruvate kinase and nucleoside diphosphate kinase). The absorbance at 334 nm was monitored in an Eppendorf ECOM 6122 photometer. IC 50 were calculated using Kaleidagraph as previously described 27 .

Whole cell activity against Mtb
MIC 50 against mycobacteria was evaluated by a 2-fold serial dilution of the tested compounds. The in vitro antimycobacterial assay was based on a method in which a luminescent M. tuberculosis H37Ra strain Lehmann & Neumann (ATCC 25177) transformed with pSMTB1 luciferase reporter plasmid is used. Each compound was dissolved in DMSO (Sigma-Aldrich) to prepare the stock solution at concentration of 10 mM. Serial dilutions were made in Middlebrook 7H9 broth and 10% OADC (oleic acid, albumin, dextrose, catalase). Volumes of 20 ll of the serial dilutions were added in triplicate to flat-bottomed 96-well plates. By thawing and dissolving a frozen Mycobacteria pellet in 7H9-10% OADC, the bacterial suspension was made. To eliminate clumps, the dissolved pellet was passed through a 5.0 lM filter (Millipore) and left for 1 h to recover at 37 C, 5% CO 2 . Subsequently, the suspension was diluted in complete 7H9 broth to obtain 50,000 Relative Light Units (RLU)/ml and a volume of 180 ll of bacteria was added to each well. After 7 days of incubation, the bacterial replication was analysed by luminometry. The bacterial suspension from each well was collected and transferred to a black 96-well plate to evade cross luminescence between wells. The luminescent signal was evoked by addition of the substrate for the bacterial luciferase, 1% n-decanal in ethanol to each well by the Discover multi-plate reader from Promega and the light emission in each well was measured. Using the plot of the % viability as a function of compound concentration, MIC 50 values were determined.

Chemistry
For the synthesis of the target compounds, two divergent synthesis routes were considered based on a disconnection of the ether bond between the C and D ring (Scheme 1).
The target compounds 4a-4k may be obtained by Ullmann coupling of either substituted phenol (5) with 3-iodobenzamide 6 (Method 1) or substituted iodobenzene (9) with phenol 10 (Method 2). We experienced that Method 1 failed to produce the coupling product in the case of electron-withdrawing substituents on the phenol ring. Except for analogue 4a, all analogues were prepared according to Method 2 as detailed Scheme 2.

Docking
To gain insight into their putative binding mode, putative analogues were docked in the target enzyme using the crystal structure of MtbTMK in complex with (S)-1 (PDB code 5NQ5) 27 . Docking indicated that the (R)-3 is better accommodated than its (S)-enantiomer ( Figure 3). The thymine moiety of (R)-3 adopts an identical pose as that of (S)-1 and is involved in two hydrogen bonds between its NH and C(4)O groups and the side chains of Asn100 and Arg74, as well as a p-p stacking interactions with Phe-70. By protruding out of the active site, the meta-piperidine ring forms a p-alkyl interaction with the aromatic side chain of Tyr 103. Importantly, in the (R)-enantiomer the oxygen atom of the carbonyl that connects ring B and ring C is expected to form a hydrogen bond with Arg-95 (Figure 3(A)), which justifies our choice to connect ring C to the piperidine ring B through an amide bond. Furthermore, through various hydrophobic interactions ring C and ring D contribute to the overall affinity. The putative binding mode of the (S)-enantiomer is quite different ( Figure  3(B)). Although a hydrogen bond may be formed between the oxygen atom of the carbonyl connecting ring B and ring C and Arg-107, the loss of the hydrogen bonds between N 3 and C 4 groups of thymine ring and the side chains of Asn100 and Arg7, as well as the p-p stacking interaction between pyrimidine ring and Phe-70 results in a less promising docking score.
(R)-3 was also docked starting from the co-crystal structure of the natural substrate TMP (PDB code 1G3U) 9 . The resulting docking pose of (R)-3 in MtbTMPK is superimposed with the one described above (Figure 4). The MtbTMPK conformation is highly similar except for the flexible Lid region, leading to slightly more encouraging docking score due to extra pronounced hydrophobic interactions exhibited by the C and D ring of (R)-3. Considering the envisioned analogues are highly related to the structure of (S)-1, the co-crystal structure of MtbTMPK in complex with (S)-1 was used for modelling.
Docking analysis starting from the 5NQ5 co-crystal structure further indicated that judicious substituents of the D-ring might lead to extra interactions with the enzyme. A trifluoromethyl substituent in meta position (as in 4k) may establish a hydrogen bond with Tyr 39 ( Figure 5). However, the consequent distortion of the D-ring leads to a less favourable docking score (Table 1).

Biological activity
Compounds 4a-4k were evaluated for their capacity to inhibit MtbTMPK and growth of an avirulent M. tuberculosis strain (Table 1). Most compounds exhibited low micromolar enzyme inhibitory activity and for most analogues the nature and the positions of the D-ring substituents have only negligible effect on the activity. This is in line with the docking scores. However, introduction of two trifluoromethyl groups in meta position of ring D (4k) significantly lowered the inhibitory potency. This may be attributed to the fact that the large trifluoromethyl groups hinder the tail of molecule to form a stacking interaction with the enzyme, or the tendency of trifluoromethyl to form a hydrogen    bond by directional change of last ring, as suggested by the docking analysis. Unfortunately, only analogue 4g with a 3,4-dichlorophenyl D-ring showed moderate antimycobacterial activity, while all other compounds were virtually inactive. To further validate the antimycobacterial activity, compounds with good enzyme affinity (4b, 4g and 4i) were screened on virulent strain (H37Rv). Nevertheless, none of them displayed any whole cell antitubercular activity (MIC > 50 lM).

Conclusion
Based on previous findings, we report herein the synthesis and biological activities of 1-(piperidin-3-yl)thymine amides as MtbTMPK inhibitors. Modelling was conducted to understand the putative binding mode, resulting in similar docking scores with lead compound 3 for most analogues except compound 4k bearing sterically demanding substituents on the D-ring. Further MtbTMPK inhibitory activities were in line with the modelling results. Compounds 4b and 4i were slightly more potent than the parent compound 3, while compound 4k was significantly less active. Unfortunately, most analogues failed to exhibit whole cell antitubercular activity, except for 4g, which demonstrated moderate cellular activity on avirulent strain. We hypothesise that the poor permeability/uptake of these inhibitors in M. tuberculosis may be the reason for the disappointing cellular activity. Attempts to improve mycobacterial uptake will be addressed in future work.

Disclosure statement
No conflict of interest was reported by the authors.

Funding
This work was supported by the China Scholarship Council (grant numbers 201607060021). -À8.7 28 ± 2 >64 a Compounds were tested on H37Ra strain, measured as the half-maximal inhibitory concentration (MIC 50 ). Using the plot of the % viability as a function of compound concentration, MIC 50 values were determined.