Novel Aryl Substituted Pyrazoles as Small Molecule Inhibitors of Cytochrome P450 CYP121A1: Synthesis and Antimycobacterial Evaluation

Three series of biarylpyrazole imidazole and triazoles are described, which vary in the linker between the biaryl pyrazole and imidazole/triazole group. The imidazole and triazole series with the short −CH2– linker displayed promising antimycobacterial activity, with the imidazole–CH2– series (7) showing low MIC values (6.25–25 μg/mL), which was also influenced by lipophilicity. Extending the linker to −C(O)NH(CH2)2– resulted in a loss of antimycobacterial activity. The binding affinity of the compounds with CYP121A1 was determined by UV–visible optical titrations with KD values of 2.63, 35.6, and 290 μM, respectively, for the tightest binding compounds 7e, 8b, and 13d from their respective series. Both binding affinity assays and docking studies of the CYP121A1 inhibitors suggest type II indirect binding through interstitial water molecules, with key binding residues Thr77, Val78, Val82, Val83, Met86, Ser237, Gln385, and Arg386, comparable with the binding interactions observed with fluconazole and the natural substrate dicyclotyrosine.


■ INTRODUCTION
Tuberculosis (TB) is endemic in numerous countries of the developing world, and serious issues are caused by infection with drug-(and multidrug-) resistant strains of the causative bacterium Mycobacterium tuberculosis (Mtb) and by the common occurrence of HIV coinfection as a result of the high susceptibility of immunocompromised individuals to TB infection. 1,2 The global impact of TB is evident from the World Health Organization data, which reported that, in 2015, 1.8 million people died from TB and approximately 2 billion people, a third of the world's population, were infected with latent TB. 2 This disease has a global impact, and the "deadly synergy" of coinfection with TB and HIV has serious life-limiting consequences for individuals who do not receive the relevant therapeutics in a timely manner. 1,2 Challenges associated with the current antibiotic therapy include drug intolerances and toxicities: pharmacokinetic drug−drug interactions, particularly with ART drugs in patients coinfected with HIV, and patient adherence given the lengthy treatment regimens. 2 The most effective drugs in clinical use have been isoniazid, the first oral mycobactericidal drug, and rifampicin. The combined used of rifampicin and pyrazinamide shortened TB chemotherapy to 9 and 6 months, respectively. 3,4 Drug-resistant strains of Mtb fall into three main categories: multidrug resistant (MDR) TB, a form of TB caused by bacteria that do not respond to at least isoniazid and rifampicin, the two most powerful first-line anti-TB drugs; extremely drug resistant (XDR) TB, a form of MDR tuberculosis that responds to even fewer available medicines, including the most effective second-line anti-TB drugs; and totally drug resistant (TDR) TB, which describes strains of Mtb that are resistant to all available firstand second-line anti-TB drugs. 1,2,5 MDR-TB is now a major problem globally and threatens the ability to treat and control TB infection. In 2015, an estimated 480 000 people developed MDR-TB, resulting in 190 000 deaths. 1 Treatment should be "individualized", based on drug susceptibility testing (DST). However, this is not always possible, and treatment is usually based on the pattern of drug resistance in the local geographical region. MDR-TB requires a longer treatment (18−24 months) with drugs that are more toxic and less efficacious. Worldwide, MDR-TB treatment success rates are as low as 50%. 5 XDR-TB requires third-line anti-TB drugs, which are expensive and are often associated with more severe side effects. The emergence of MDR-, XDR-, and TDR-TB strains has led to intensified research to identify new anti-TB drugs over the past decade.
Mycobacterium tuberculosis encodes 20 cytochrome P450 enzymes (CYPs or P450s). One of these is CYP121A1, which was shown to be essential for microbial viability. 6 CYP121A1 catalyzes the formation of an unusual intramolecular bond between carbon atoms in the ortho-positions of two tyrosine groups in the cyclodipeptide dicyclotyrosine (cYY), the natural substrate, to form the metabolite mycocyclosin. cYY interacts indirectly with the heme iron through water molecules ( Figure  1A). Interestingly, this type of indirect binding mode was also observed in the crystal structure of CYP121A1 bound to the azole antifungal drug fluconazole. 7 The crystal structure of the CYP121A1/fluconazole complex, PDB 2IJ7, exhibits dual fluconazole conformations (I and II). In the indirect conformation (I), a triazole nitrogen of fluconazole binds to the heme iron via a distal water molecule that remains coordinated to the heme iron. However, in the direct conformation (II), the fluconazole binds through a triazole nitrogen directly to the heme iron ( Figure 1B). While both binding modes are evident, the indirect binding (via water) mode appears predominant in this structure. 8 The natural substrate cYY binds to the CYP121A1 active site without inducing any significant change in the P450 structure, consistent with it being a good substrate for the enzyme. 6 Through crystallization of ligand complexes, it was discovered that iodopyrazole binds with its pyrazole group in a channel between P450 helices F and G and "stacked" between Trp182 and Phe168 side chains. The iodine atom protrudes toward the active site and interacts with hydrophobic amino acids. 7 In designing potential new inhibitors of CYP121A1, we noted that "Y-shaped" molecules might provide a good "fit" within the P450 active site. This may be the result of a high conformational energy barrier necessary for binding to CYP121A1.
Using this observation to help design novel CYP121A1 inhibitors, our research explored the CYP121A1 binding and antimycobacterial activity of various "Y-shaped" biaryl pyrazole imidazole or triazole derivatives ( Figure 2). Using different linkers between the hydrophobic biaryl pyrazole moiety (blue, Figure 2) and the heme binding azole group (red, Figure 2), the proximity of the heme binding moiety will be extended to determine if this has an effect on the indirect/direct binding of new ligands to the heme iron.

■ RESULTS AND DISCUSSION
Chemistry. The imidazole (7) or triazole (8) pyrazole compounds were obtained via a five-step synthetic route (Scheme 1) beginning with the preparation of the imines (3)

Journal of Medicinal Chemistry
Article by the reaction of acetophenones (1) with phenylhydrazine under acidic conditions. 9,10 The aldehydes (4) were prepared as previously described 9,10 by a Vilsmeier−Haack reaction of the imines (3). Reduction of the aldehydes (4) with NaBH 4 resulted in high yields of the corresponding alcohols (5), which were subsequently converted to the chlorides (6) by treatment with thionyl chloride in toluene at 115°C for 2 h. 11 The chlorides were generally found to be unstable, so they were prepared immediately before use in the following step. Reaction of the chlorides (6) with either the potassium salt of imidazole or triazole, prepared in situ by treatment of imidazole or triazole with potassium carbonate in acetonitrile at 45°C for 1 h, overnight at 70°C, gave the required final imidazole (7) and triazole (8) pyrazole derivatives (Scheme 1).
The final amides (13) were prepared by a carbonyldiimidazole (CDI) coupling reaction through the coupling of the carboxylic acids (9) with the free amine, generated in situ by the treatment of the aminoethylimidazole dihydrochloride salt (13) with triethylamine in DMF. The final amide derivatives (13) were obtained in low to moderate yields (Scheme 2).
CYP121A1 Ligand Binding Affinity. The CYP121A1 binding affinity (K D ) of the various compounds was determined by UV−vis optical titration. Results shown in Tables 1−3 reveal that most of the compounds show a red (type II) shift in the Soret peak position (Figure 3, for example 7b), indicating that, in the solution state, most of these compounds coordinate either (i) directly to the CYP121A1 heme iron or (ii) indirectly to the heme iron through interstitial water molecule(s). Another possibility is (iii) that a ligand binding distant from

Journal of Medicinal Chemistry
Article the heme does not induce a heme spectral shift (but would block the active site).
Series 7 (imidazole derivatives) showed the highest affinity for CYP121A1, with compounds 7a−e having K D values <12 μM (K D = 2.63 ± 0.19, 4.16 ± 0.47, 7.42 ± 0.66, 5.36 ± 0.59, and 11.35 ± 0.94 μM, respectively) ( Table 1). The values are comparable with the K D values for the azole drugs ketoconazole (imidazole) and fluconazole (triazole), which are 3.3 ± 0.3 and 8.6 ± 0.2 μM, respectively. Compounds 7a and 7b induce the most extensive Soret absorbance shifts (416−423.5 nm), suggesting that the CYP121A1 heme iron is predominantly coordinated by a direct imidazole nitrogen bond in these cases. Compounds 7c−h show less extensive Soret red shifts for binding the CYP121A1 heme, suggesting that ligation modes in these cases may involve imidazole−nitrogen interactions with heme iron mediated through an interstitial water ligand and that these could occur to a greater (e.g., compounds 7f and 7h) or lesser (e.g., compounds 7c and 7g) extent. The smallest Soret shift occurs with compound 7h (416−418 nm), and it is likely that the indirect coordination though the retained distal water on the heme iron is predominant in this case. However, alternative active site binding poses made by series 7 ligands that do not impact the heme iron or its distal water ligand cannot be ruled out at this stage, and this phenomenon could also provide explanations for the less extensive Soret shifts induced by a number of the series 7 molecules.
Series 8 (triazole derivatives) has affinities ranging from K D = 35.6 ± 2.7 μM (compound 8b) to 193 ± 21 μM (compound 8f) ( Table 2). Series 13 (pyrazole amide derivatives) had weaker affinities, with moderate K D values between 290 and 400 μM (290 ± 27, 314 ± 43, and 394 ± 39 μM for compounds 13d, 13e, and 13a, respectively). No evidence of binding was observed for the halogenated derivatives of this compound class ( Table 3). As was also observed for series 7, the extent of the CYP121A1 heme Soret shift observed at apparent saturation with the series 8 compounds varied considerably. In series 8, a similar pattern of Soret shifts was observed to that seen with series 7, with peak shifts as small as 416−417.5 nm (compound 8f) and as large as 416−423 nm (compound 8b). Again, we infer that the most extensive absorbance changes occur for those ligands that show predominantly direct coordination of heme iron through a triazole nitrogen (i.e., compounds 8a and 8b), while less extensive Soret shifts (i.e., compounds 8d and 8f) are likely indicative of either an indirect (through the heme distal water) coordination of the CYP121A1 heme iron or a possible result from mixed binding poses that include modes that do not influence the heme spectrum. For series 13, ligand binding curves were obtained for the compounds 13a, 13d, and 13e. However, the extents of the Soret band shift were modest (416−417.5/418 nm). Once again, these data suggest either indirect coordination of heme iron through an imidazole nitrogen or a combination of this type of ligand binding mode with other pose(s) that do not impact the heme spectrum.
Azole antifungal drugs (clotrimazole, econazole, fluconazole, ketoconazole, and miconazole) bind tightly to Mtb CYP121A1, inducing a Soret peak shift between 421 and 423.5 nm. The K D values for clotrimazole, econazole, and miconazole are all <0.2 μM, indicating very tight binding, as determined from data fitting using the Morrison (quadratic) equation. 14 MIC Determination Against Mycobacterium tuberculosis. The derivatives were screened against M. tuberculosis H37Rv by the REMA (Resazurin Microtiter Assay) method. 15 In the imidazole (7) and triazole (8) pyrazole series, the imidazoles (7) were generally more inhibitory to M. tuberculosis growth and displayed a better activity than the triazoles (8) (Tables 1 and 2), with the 4-bromo (7f) and 4-iodo (7g) derivatives displaying a good activity (MIC 6.25 μg/mL). All of

Journal of Medicinal Chemistry
Article the imidazoles (7), with the exception of the 4-nitrile derivative (7h, MIC 90 = 100 μg/mL), were significantly more active than the azole drug fluconazole. Compounds 7a−d, 7f, and 7g are more active than clotrimazole (MIC 90 = 20 μg/mL) ( Table 1). The "extended" pyrazole compounds 13a−e had little antimicrobial activity (MIC 90 ≥ 100 μg/mL) ( Table 3). Molecular Modeling and Crystallography. The Molecular Operating Environment (MOE) program 16 was used to perform molecular docking and was found to closely replicate the position and binding interactions of cYY and fluconazole, as observed in the crystal structures PDB 3G5H and PDB 2IJ7, respectively.
The imidazole (7) and triazole (8) compounds were all found to interact with the heme through an interstitial water molecule via hydrogen-bonding interactions with Ser237, Gln 385, and Arg386 on one side of the structure, while on the other side of the structure interactions are through hydrophobic amino acid residues including Thr77, Val78, Val82, Val83, and Met86 ( Figure 4A). The docking results were consistent with the type II binding indicated from the binding affinity assays. The only exception was the nitrile-substituted derivatives 7h and 8h. In both cases, the nitrile group interacted with Ala167 and Trp182 via interstitial water molecules, holding the compounds away from the active site with the imidazole/ triazole ring positioned away from the heme group ( Figure 4B).
The amide imidazole compounds (13) were positioned in a similar manner ( Figure 4C) to 7 and 8. However, whereas imidazole (7) and triazole (8) were consistently positioned within the active site, numerous conformations were observed for the amide imidazole series (13), possibly owing to their increased conformational flexibility, which may correlate with the moderate binding affinity observed.
Two compounds were successful cocrystallized with CYP121A1, 7b and 7e ( Figure 5). However, the crystallized structures did not correlate with the binding affinity results as the imidazole group was positioned away from the heme group, and in both cases, the compounds were located in positions too far away to form direct or indirect binding interactions with the heme iron. However, the observed binding poses of 7b and 7e are similar, and both clearly result in effective blockage of the active site and thus result in inhibition of binding of the cYY substrate. These data illustrate how different P450 ligand binding modes can be adopted by molecules bound to proteins in the solution state or in the more constrained environment of protein crystals. In the case of the compound series investigated in this study, there is clear evidence from structural, molecular modeling and UV−visible ligand titration studies for the adoption of binding modes that involve direct or indirect (via the distal water ligand) coordination of the heme iron by imidazole or triazole nitrogens, as well as for modes in which the bound ligands are distant from the heme and are unlikely to influence the CYP121A1 heme spectrum.

■ CONCLUSION
The data presented demonstrate the successful binding to CYP121A1 of a range of different compounds inspired by the structure of CYP121A1's natural substrate (the cyclic dipeptide cYY). Series 7 produced the most effective antimycobacterial agents with the tightest binding constants and with K D values The left-hand panel shows data from a compound 7b titration with CYP121A1 (∼4.7 μM) with the ligand-free spectrum as a thick black line, with the spectra following progressive additions of 7b as thin solid lines, and with the final near-saturated protein spectrum shown as a thick red line. The inset shows an overlaid difference spectra generated by the subtraction of the starting spectrum from each consecutive ligand-bound spectrum collected in the titration. The right-hand panel shows a plot of compound 7b-induced absorbance change, calculated as the difference between the peak and trough in the difference spectra in the left-hand panel, using the same wavelength pair (429 and 392 nm, respectively) throughout. Data were fitted using the Hill equation to give a 7b K D value of 4.16 ± 0.47 μM.

Journal of Medicinal Chemistry
Article for the eight compounds in this series in the range between 2.63 and 45.9 μM, with an average (mean) K D value of ∼13.7 μM. The series 7 compounds with the best MIC values were those with bromo (7f) and iodo (7g) substituents (MIC = 6.25 μg/mL); the Hansch analysis suggested that increased lipophilicity ( Figure 6) may correlate with improved antimycobacterial activity in this series, possibly through facilitating increased drug uptake across the Mtb lipid-rich cell wall ( Table  1). The 4-nitrile derivative 7h is a relatively weak binder (K D = 45.9 μM), a property which may be explained from molecular docking ( Figure 4B) where the nitrile interacts with Ala167 and Trp182 via interstitial water molecules, holding the compound at a distance too far from the active site to attain indirect or direct interaction with the heme.
From the antimicrobial MIC data, it is clear that extending the "linker" region of the pyrazole derivatives from −CH 2 − (series 7 and 8) to −C(O)NH−(CH 2 ) 2 − (series 13) does not translate to improved MIC values. Computational docking studies suggested that these extended compounds (series 13) were able to interact with the heme indirectly and fit within the hydrophobic CYP121A1 pocket. However, only moderate K D values between 290 and 400 μM were observed for compounds 13d, 13e, and 13a, respectively, with only partial type II heme shifts (416−417.5/418 nm). There was also no evidence of binding for halogenated derivatives of this compound class.

Journal of Medicinal Chemistry
Article ■ EXPERIMENTAL SECTION General Procedures. All reagents and solvents were of general purpose or analytical grade and purchased from Sigma-Aldrich Ltd., Fisher Scientific, Fluka, and Acros. 1 H and 13 C NMR spectra were recorded with a Bruker Avance DPX500 spectrometer operating at 500 and 125 MHz, with Me 4 Si as an internal standard. Mass spectra (HRMS) were determined by the EPSRC mass spectrometry center (Swansea, UK). Elemental analysis was performed by MEDAC Ltd. (Chobham, Surrey, UK). Flash column chromatography was performed with silica gel 60 (230−400 mesh) (Merck), and TLC was carried out on precoated silica plates (kiesel gel 60 F 254 , BDH). Compounds were visualized by illumination under UV light (254 nm) or by the use of vanillin stain followed by heating. Melting points were determined on an electrothermal instrument and were uncorrected. All solvents were dried prior to use and stored over 4 Å molecular sieves, under nitrogen. All of the compounds were ≥95% pure.
General Method for the Preparation of Imidazole and Triazole Derivatives 7 and 8. To a stirred suspension of K 2 CO 3 (3 mmol) in dry acetonitrile (20 mL) was added imidazole (3 mmol) or triazole (3 mmol). The reaction mixture was refluxed at 45°C for 1 h. After cooling the mixture to room temperature, the chloromethyl pyrazole compound (6) (0.76 mmol) was added, and the reaction mixture refluxed at 70°C overnight. The solvent was evaporated under reduced pressure, and the resulting mixture was diluted with EtOAc (50 mL) and washed with H 2 O (3 × 20 mL). The EtOAc layer was dried (MgSO 4 ) and evaporated under reduced pressure to give the crude imidazole (7) or triazole (8), which was further purified by recrystallization.

Journal of Medicinal Chemistry
Article (typically 0.05−0.2 μL aliquots) from concentrated stock solutions to the protein in a 1 mL final volume. Spectral measurements were taken before the ligand addition and after the addition of each aliquot of ligand until no further spectral change occurred. Difference spectra at each stage in the titration were obtained by the subtraction of the initial ligand-free enzyme spectrum from the subsequent spectra collected after each addition of ligand. From the difference spectra, a pair of wavelengths were identified and defined as the absorbance maximum (A peak ) and minimum (A trough ). The overall absorbance change (ΔA max ) was calculated by subtracting the A trough value from the A peak value for each spectrum collected after a ligand addition. Graphs of ΔA max against [ligand] were plotted for each titration. Titrations were done in triplicate, and the final K D value presented was determined as the average value across the three sets. The K D values were determined by fitting the data using either a standard hyberbolic function (eq 1) or the Hill equation (eq 2) using the Origin software (OriginLab, Northampton, MA).
In eq 1 (the standard hyperbolic function, the Michaelis−Menten function adapted for ligand binding), A obs is the observed absorbance change at ligand concentration L, A max is the maximal absorbance change observed at the apparent ligand saturation, and K D is the dissociation constant for the binding of the ligand (the substrate concentration at which A obs = 0.5 × A max ).
In eq 2 (the sigmoidal Hill equation), A obs is the observed absorbance change at ligand concentration L, A max is the absorbance change at the apparent ligand saturation, K is the apparent dissociation constant, and n is the Hill coefficient, a value describing the apparent extent of cooperativity observed in ligand binding. Antimycobacterial Activity Assay. M. tuberculosis H 37 Rv was grown in 7H9 liquid medium with 10% Middlebrook OADC Growth Supplement enrichment (BBL/Becton-Dickinson, Sparks, MD, USA). The bacterial growth occurred at 37°C until the mid log phase was reached (OD 600 nm = 0.4−0.6). After this period, the bacterial suspensions were prepared as described below and REMA assays were performed. The anti-M. tuberculosis activity of the compounds

Journal of Medicinal Chemistry
Article was determined by the REMA (Resazurin Microtiter Assay) method. 15 Stock solutions of the tested compounds (10 μg/mL) were prepared in DMSO and diluted in Middlebrook 7H9 broth supplemented with 10% OADC.
The microdilution of the compounds was performed in 96-well plates to obtain final compound concentration ranges 0.39−100 μg/ mL. Rifampicin in the concentration range between 0.004 and 1 μg/ mL was added as a control. Bacterial suspensions were prepared, and their turbidities were adjusted to match the optical density of the McFarland no. 1 standard. After a further dilution of 1:20 in the Middlebrook 7H9 broth supplemented with OADC, 100 μL of the inoculum was added to each well of the 96-well plate. Cultures were incubated for 7 days at 37°C, and 30 μL of 0.01% resazurin was added. Wells were read after 24 h for color change, and the fluorescence was measured (excitation/emission of 530/590 nm filters, respectively) in a microfluorimeter. The MIC was defined as the lowest concentration resulting in 90% inhibition of M. tuberculosis growth. The presented results are from two independent experiments.
Molecular Modeling and Docking. Docking studies were performed using the MOE 16 program and Mtb CYP121A1 cocrystallized with fluconazole (pdb 2IJ7). All minimizations were performed with MOE until a RMSD gradient of 0.01 kcal/mol/A with MMFF94 force field and partial charges were automatically calculated. The charge of the heme iron at a physiological pH was set to 3 + (geometry d2sp3) through the atom manager in MOE. The Alpha Triangle placement, which derives poses by random superposition of ligand atom triplets through alpha sphere dummies in the receptor site, was chosen to determine the poses. The London ΔG scoring function estimates the free energy of binding of the ligand from a given pose. Refinement of the results was done using the MMFF94 force field, and rescoring of the refined results using the London ΔG scoring function was applied. The output database dock file was created with different poses for each ligand and arranged according to the final score function (S), which is the score of the last stage that was not set to zero.
Crystallography Studies on CYP121A1. Untagged CYP121A1 protein and crystals were prepared as previously reported, with the following adaptations. 14 Crystals were prepared using a Mosquito pipetting robot (Molecular Dimensions, Newmarket, UK) in 800 nL drops with protein-to-mother liquor at a ratio of 1:1, by vapor diffusion in 1.5−2.1 M ammonium sulfate and 0.1 M sodium MES, or Cacodylate from pH 5.5−6.15. Co-crystals were prepared following incubation with 2 mM ligand prepared in DMSO. Protein solutions were centrifuged at 14,000 rpm for 20 min at 4°C immediately before crystallogenesis. Ligand soaks were also carried out either by directly dissolving solid ligand to saturation, or by the addition of a 2−5 mM ligand solution in DMSO to the mother liquor, and soaking was carried out for a minimum period of 24 h. Crystals were immersed in mother liquor supplemented with 10−30% oil as cryoprotectant, and cryoprotected and flash-cooled in liquid nitrogen. Data were collected on beamline i02 (wavelength 0.9795 Å) at the Diamond Light Source Facility (Harwell, UK). The diffraction data were reduced, scaled and merged using XDS or Xia2. 18,19 Structures were refined using PHENIX 20 with the native CYP121A1 structure (PDB 1N40) 14 as the starting model. Structural rebuilding and validation were performed with COOT, 21 Molprobity, 22 and PDB REDO. 23 ■ ASSOCIATED CONTENT

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jmedchem.7b01562.

Accession Codes
PDB codes for CYP121A with 7b and 7e bound are 5OPA and 5OP9, respectively. The authors will release the atomic coordinates and experimental data upon article acceptance for publication.