Synthesis, Structure–Activity Relationship, and Mechanistic Studies of Aminoquinazolinones Displaying Antimycobacterial Activity

Phenotypic whole-cell screening against Mycobacterium tuberculosis (Mtb) in glycerol–alanine–salts supplemented with Tween 80 and iron (GASTE-Fe) media led to the identification of a 2-aminoquinazolinone hit compound, sulfone 1 which was optimized for solubility by replacing the sulfone moiety with a sulfoxide 2. The synthesis and structure–activity relationship (SAR) studies identified several compounds with potent antimycobacterial activity, which were metabolically stable and noncytotoxic. Compound 2 displayed favorable in vitro properties and was therefore selected for in vivo pharmacokinetic (PK) studies where it was found to be extensively metabolized to the sulfone 1. Both derivatives exhibited promising PK parameters; however, when 2 was evaluated for in vivo efficacy in an acute TB infection mouse model, it was found to be inactive. In order to understand the in vitro and in vivo discrepancy, compound 2 was subsequently retested in vitro using different Mtb strains cultured in different media. This revealed that activity was only observed in media containing glycerol and led to the hypothesis that glycerol was not used as a primary carbon source by Mtb in the mouse lungs, as has previously been observed. Support for this hypothesis was provided by spontaneous-resistant mutant generation and whole genome sequencing studies, which revealed mutations mapping to glycerol metabolizing genes indicating that the 2-aminoquinazolinones kill Mtb in vitro via a glycerol-dependent mechanism of action.

T uberculosis (TB) is an infectious disease, caused by the bacillus Mycobacterium tuberculosis (Mtb), and is the leading cause of death worldwide from a single infectious agent. According to the World Health Organization (WHO), approximately 10 million people acquired TB in 2018 and 1.2 million deaths occurred globally. 1 Treatments for drugsusceptible TB have been available for 60 years, with the current regimen comprising an initial two month intensive phase with four frontline drugs (isoniazid, rifampicin, ethambutol, and pyrazinamide), followed by a four month continuation phase using isoniazid and rifampicin. However, resistant strains have emerged over the years possibly as a result of poor patient compliance due to the extensive drug regimen, prolonged treatment time, socioeconomic factors, toxic effects, and the tendency to be asymptomatic before safe completion of the prescribed course. 2,3 In the past 15 years, several new classes of anti-TB drugs have emerged in the developmental pipeline including three new agents. Bedaquiline and delamanid received accelerated regulatory approval in 2012 and 2014, respectively, and are being progressed into phase III clinical trials for the treatment of multidrug resistant TB (MDR-TB). 2−4 Most recently, pretomanid was approved as part of a six month, three-drug regimen consisting of bedaquiline, pretonamid, and linezolid (BPaL) for the treatment of people with extensively drug resistant TB (XDR-TB). 5 Although there are a number of new chemical entities at various phases of clinical development, 6 the high attrition rate requires a continuous feed of new chemical matter into the drug discovery pipeline. The development of new TB treatments involves numerous challenges as the TB research community needs to focus on many aspects, e.g., shorter and simpler regimen, less side effects, effectiveness against drug resistant TB, reduced drug− drug interactions for human immunodeficiency virus (HIV)coinfected patients, and affordable cost for patients in developing countries where the majority of TB cases exist. 7,8 2-Aminoquinazolin-4(3H)-ones are derived from the quinazolinone class of compounds, which have been extensively explored in a wide range of therapeutic areas, whereas literature references on 2-aminoquinazolinones are relatively scarce. Secondary and tertiary 2-amino analogues have been found to have pharmacological relevance, with examples of antifungal, 9 anti-inflammatory, anticancer, 10,11 antibacterial, 12,13 antihypertensive, 14 antidepressant, 15 and antimycobacterial activity. 16,17 On the other hand, primary 2aminoquinazolinone derivatives have not yet been commercially developed but have shown a variety of biological activities as antimalarial, 18 antiviral, 19,20 and anticancer 21 agents. Although there is no known antimycobacterial activity, they appear as a valuable novel scaffold to explore as a potential anti-TB chemotype.
On the basis of structural similarity shared with a previously discovered antimalarial 2-aminopyridine series, 22 2-aminoquinazolinones were synthesized for evaluation against the human malaria parasite Plasmodium falciparum. Subsequent phenotypic whole-cell cross-screening against Mtb identified the 2-aminoquinazolinone 1 as an attractive hit compound to follow up on despite its low solubility. During preliminary studies toward improving the solubility of 1, the sulfone substituent was converted to the sulfoxide 2, which displayed improved aqueous solubility while maintaining potency ( Figure 1). As a result, medicinal chemistry efforts were subsequently pursued in order to explore the antimycobacterial potential of this 2-aminoquinazolinone series.
Herein, we present the synthesis, structure−activity relationship, in vitro absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties, and in vivo pharmacokinetic (PK) properties of this series. In vivo efficacy and subsequent mechanistic studies are also presented.

■ RESULTS AND DISCUSSION
Chemistry. Structural modifications were explored around the aminoquinazolinone core with the aim of improving potency and aqueous solubility at pH 7.4. Four main points of diversity were established at positions 2, 3, 6, and 7 of the scaffold as indicated in Figure 2. R 1 and R 2 groups were investigated to identify appropriately substituted ring structures with various electronic, steric, and hydrophobicity properties. The addition of heteroatoms or water solubilizing groups was envisaged to reduce lipophilicity and, therefore, improve aqueous solubility. Additionally, the strategy of increasing sp 3 character through saturation of aromatic rings was employed to disrupt the crystal packing of molecules. The position of the substituent on the left-hand side (LHS) from the 6-to the 7-position of the molecule was also explored to determine which of the LHS positions was optimal for activity. Modifications were also made to the 2-position to understand whether or not the 2-amino group was essential for potency.
The synthesis of compounds 1, 2, and 7−36 was achieved following a relatively straightforward five-step synthetic route adapted from Leivers et al., starting with the commercially available 2-aminoiodobenzoic acid (Scheme 1). 19 Briefly, a cyclization reaction with the appropriately substituted phenyl isothiocyanate afforded intermediate 3. Chlorination of the thioxo intermediate to 4 was obtained using phosphorus oxychloride (POCl 3 ) in the presence of phosphorus pentachloride (PCl 5 ). The 2-amino group was introduced via a 4methoxybenzylamine intermediate (5) followed by deprotection of the PMB group to afford 6. Finally, a Suzuki crosscoupling reaction with the appropriate boronic acids yielded the desired compounds 1, 2, and 7−36.
The procedure to obtain the analogue without the 2-amino substituent was adapted from Gungor et al. 23 (Scheme 2). This approach involved refluxing 2-amino-5-iodobenzoic acid with 4-(trifluoromethyl)aniline in the presence of triethoxymethane to afford the iodo intermediate 37. A subsequent Suzukicoupling reaction with (3-(methylsulfonyl)phenyl)boronic acid yielded the desired derivative devoid of the 2-amino group 38.
Modifications to the 2-amino substituent were achieved using a similar synthetic route to Scheme 1, whereby the appropriate amine was installed by substituting the chloro group of intermediate 4a under basic conditions. Thereafter, the final target compounds 40−43 were obtained using the Suzuki-coupling conditions described previously (Scheme 3).
In Vitro Antimycobacterial Activity. Target compounds were initially evaluated for their in vitro antimycobacterial activity against the drug susceptible Mtb H37Rv strain, under replicating conditions using GASTE-Fe (glycerol−alanine− salts supplemented with Tween 80 and iron) culture media. The minimum inhibitory concentration (MIC 90 ) was read 14 days after incubation. Additionally, compounds were screened for in vitro cytotoxicity against a mammalian cell line, Chinese Hamster Ovarian (CHO). Initial structure−activity relationship (SAR) exploration was performed on the LHS with variations on the 6-position while keeping the 2-amino-  quinazolinone core unchanged and the right-hand side (RHS) constant as a para-trifluoromethyl phenyl group (Table 1).
The R 1 phenyl group was first altered to explore the impact of various electronic effects and water solubilizing groups on activity. Changing the methyl sulfone in 1 to the more lipophilic cyclopropyl sulfone (8) led to a loss of activity, as did the introduction of a methyl group at the ortho-position of the biaryl moiety (7). Amide analogues (9,10,11) were generally equipotent to the sulfone derivative. Potency was also maintained when an aliphatic chain was extended off the amide functionality as in 13 and 14. Various morpholino-type carboxamides were synthesized to improve aqueous solubility whereby the thiomorpholine analogue, 18, was found to be highly potent (0.656 μM). Amides in the meta-position in combination with polar, ionizable groups, such as pyridyl groups 12 and 17 resulted in a 10-and 15-fold loss in activity, respectively. Transformation to the carboxylic acid group 20 was not tolerated (MIC 90 = 112 μM).
Replacements of the phenyl group by heterocycles were also investigated. While the unsubstituted pyridine (22) was equipotent to the frontrunner compounds 1 and 2, the metasubstituted counterparts (23 and 24) displayed a loss in potency (MIC 90 >2 μM). Likewise, the pyrimidine analogue (21) was inactive. Exploration of electronic and lipophilic effects identified that electron withdrawing, lipophilic substituents (e.g., CF 3 and Cl groups as in 28 and 29, respectively) were highly beneficial for potency, whereas electron donating groups (e.g., OMe, tBu, and NMe 2 as in 25, 26, and 27, respectively) were not well-tolerated.
Subsequent SAR explorations focused on moving the LHS substituent from the 6-to the 7-position as well as varying the 2-amino moiety (Table 1). All 7-position derivatives (using side chains in the 6-position, which led to active compounds in order to have matched pairs) maintained potency (<2 μM), thus allowing variation on this position and broadening the SAR scope on the LHS. Thereafter, several analogues were . The majority of compounds in this series were found to be noncytotoxic except for the cyclopropyl sulfone (8), tertiary amide (10), and morpholino (16) substituted analogues, although cytotoxicity data always has to be evaluated critically when they are higher than the solubility data. As far as the SAR exploration on the RHS is concerned, diverse groups were investigated to replace the trifluoromethyl substituent, while keeping the 2-aminoquinazolinone core unchanged and keeping the LHS group as a phenyl sulfoxide at position 6 constant, so that matched pairs could be compared (Table 2). Initially, the exploration of the substitution pattern around the ring highlighted that the CF 3 substituent exhibited optimal potency when at the para-position compared to the ortho-position (34) and meta-position (48). Solubility improvement strategies by adding a pyridyl group (47) or introducing a saturated ring (57) maintained moderate potency (MIC 90 = 0.922 and 2.75 μM, respectively), whereas the addition of a benzylic methylene linker (58) led to a loss in potency. The replacement of the trifluoromethyl substituent by polar groups (SO 2 Me, OH, OMe, CN) was detrimental to activity, while lipophilic substituents (F, Me) allowed moderate potency to be retained. The unsubstituted phenyl ring analogue (35) was not tolerated (MIC 90 = 33 μM). Overall, the RHS SAR revealed that directly attached phenyl rings bearing lipophilic groups at the para-position were optimal for activity.
Physicochemical and In Vitro Microsomal Metabolic Stability Profiling. As mentioned previously, solubility was established as a critical parameter to optimize. Using the kinetic solubility assay, we aimed to achieve values >100 μM, but compounds with moderate solubility and potency were also progressed. To that effect, we looked at introducing hydrophilic groups as well as reducing overall aromaticity and planarity of the molecules (Table 1). For the latter, a methyl substituent was inserted at the ortho-position to the LHS sulfone, as exemplified by 7, in order to disrupt the planarity of the biaryl scaffold. This resulted in a 10-fold improvement in solubility (53.3 μM) with moderate potency (4.22 μM). Generally, derivatives from the R 1 SAR displayed poor to moderate solubility, except for 13 and 14 for which the addition of the aliphatic amide substituents drastically improved solubility (168 and 192 μM, respectively). The addition of heteroatoms as in 22 and 23 did not lead to improvements in solubility, except when the heterocycle contained an additional meta-substituent such as an amide (12) or methoxy (24) moiety (40 μM). Solubility for the morpholine-, thiomorpholine-, and piperazine-type derivatives (16,18,19) was moderately improved (46−56 μM).
Modifications to the 2-position of the core scaffold maintained good metabolic stability, but no significant improvements were made in terms of solubility (Table 1). Removal of the 2-amino group decreased the solubility (38 = <5 μM) of the sulfoxide derivative potentially due to the increase in lipophilicity (clogD = 4.25). The movement of the LHS group from position 6 to position 7 on the scaffold resulted in poor solubility, except for the primary amide derivative (33), which exhibited a 10-fold improvement compared to 9.
RHS modifications were found to be the most favorable to improve aqueous solubility with five analogues (34, 58, 36, 53, and 56) exhibiting moderate to high solubility (>100 μM) ( Table 2). The introduction of a pyridyl nitrogen (47) did not improve solubility as seen previously on the LHS; this could be due to the electron withdrawing nature of the trifluoromethyl substituent, which lowers the pK a and reduces the ionizability of the nitrogen. The movement of the trifluoromethyl substituent from the para-to the meta-position was shown to be beneficial for solubility (164 μM), however, a reduction in metabolic stability was observed. Similarly, the incorporation of a benzyl linker improved solubility, albeit detrimental to metabolic stability. Variations of the trifluoromethyl substituent with hydrophilic, electron donating groups such  Despite multiple analogues being synthesized and exhibiting favorable in vitro properties, the two frontrunner compounds remained the best overall compounds in terms of combined potency, solubility, metabolic stability, and cytotoxicity properties. Therefore, compound 2 and its active metabolite, 1, were selected for further in vivo PK studies.
In Vivo Pharmacokinetic Studies. In vivo PK studies were performed on compound 1 (sulfone) and compound 2 (sulfoxide) ( Table 3). All studies and procedures were conducted with prior approval of the animal ethics committee of the University of Cape Town (approval numbers 013/028 and 014/028) in accordance with the South African National Standard (SANS 10386:008) for the Care and Use of Animals for Scientific Purposes and guidelines from the Department of Health. The sulfone 1 was poorly absorbed achieving a bioavailability of 18%. Compound 2 had much higher oral bioavailability and, notably, was converted in vivo to the sulfone (1). This resulted in a higher clearance for the sulfoxide 2 (21.8 vs 1.3 mL/min/kg). Higher exposure of the sulfone was achieved when the sulfoxide was dosed and underwent metabolic conversion than when the sulfone was dosed directly (AUC 18430 vs 7249 min·μmol·L −1 ) as indicated in Figure 3.
In Vivo Tolerability and Efficacy Studies. The high exposure of the sulfone 1 achieved by dosing the sulfoxide 2 suggested that the sulfoxide could be used for in vivo efficacy experiments. Compound 2 was therefore dosed orally at 200 mg/kg to determine the exposure of the compound at the efficacious dose and to confirm that the exposure would be maintained long enough for efficacy. The free concentrations of compound 2 and its sulfone metabolite 1 are represented in Figure 4.
The free exposure was maintained above the MIC 90 of compound 1 for more than 24 h, and this dose was therefore projected to give in vivo efficacy.
To test compound 2 for in vivo efficacy, we employed the BALB/c acute TB mouse infection model (see Supporting Information, Section 4.2). In this model, mice are infected by On the basis of the in vivo PK and tolerability studies, the lack of in vivo efficacy is unlikely to be accounted for by low exposure and/or toxicity issues. In addition, plasma concentrations in infected mice at 1 and 24 h were comparable to those in healthy mice at the same time points (Table S1). This suggests that PK was similar between healthy and infected animals. Further investigation into the discrepancy between in vitro and in vivo results was therefore undertaken to understand the in vitro/in vivo disconnect.
Mechanistic Studies. Compound 2 was tested in biology triage assays to assess the activity against promiscuous targets. It did not show hypersensitivity against a cytochrome-bd oxidase knockout mutant strain (Δcyd GASTE-Fe MIC 90 = 2.5 μM), which is hypersusceptible to compounds that inhibit the QcrB-containing cytochrome bc1 complex, thereby eliminating QcrB as a potential target of 2. 25,26 Additionally, 2 did not show a positive signal in two standard bioluminescence reporter assays: PiniB-LUX 27 detects modulation in iniB expression if compounds target Mtb cell-wall biosynthesis, and PrecA-LUX 27 detects modulation in recA expression, an indicator of genotoxic compounds. On the basis of this initial triage process, the 2-aminoquinazolinone scaffold was identified as a hit series with a potential novel mechanism of action (MOA).
However, to further understand the discrepancy between in vitro and in vivo results, compound 2 was subsequently retested in vitro using different Mtb strains cultured in different media. As shown in Table 4, 2 was active against both H37Rv and Erdman Mtb strains when grown in media containing glycerol as the carbon source (7H9, glycerol, ADC, and Tween 80). In contrast to these observations, 2 lost activity against both strains when cultured in glycerol-free media (7H12, casitone, palmitic acid, BSA, catalase, and Tween 80).
Additional in vitro assays using glucose as the carbon source were performed on 15 other aminoquinazolinone analogues. In all cases, no activity was observed in the absence of glycerol in the culture medium, thus suggesting that the glycerol metabolism played a key role in effectively inhibiting Mtb growth. We therefore hypothesized that the in vitro/in vivo disconnect was due to the fact that glycerol was not used as a primary carbon source by Mtb in the mouse lungs, as has previously been observed. 28−30 In order to test this hypothesis, MOA studies using spontaneous-resistant mutant generation combined with potential target identification using whole genome sequencing were performed with 2. The spontaneous-resistant mutants   Table 5. The GlpK gene (Rv3696) encodes for glycerol kinase, which is the first committing enzyme for glycerol metabolism in most bacteria. The second metabolite in the glycerol dissimilation pathway is dihydroxyacetone phosphate (DHAP), which is formed from glycerol-3-phosphate (G3P) by the enzyme glycerol-3-phosphate dehydrogenase encoded by the glpD1 and glpD2 genes. 28 Previous reports observed that frameshift mutations and SNPs in glpK were linked to resistance to Mtb growth inhibitors. 28−30 These studies showed that compounds with glycerol-specific activity promote the accumulation of G3P, resulting in the rapid depletion of adenosine triphosphate (ATP) utilized for phosphorylation of glycerol, ultimately leading to self-poisoning of Mtb. 28,29 The activity of these compounds was therefore found to be associated with the metabolism of glycerol contained in 7H9 media. 28−30 No such reports have, however, been observed for glpD2 mutants related to in vitro drug screening.
All considered together, the results from the MOA studies suggest that the aminoquinazolinone series potentially exert its inhibitory effect by dysregulating the glycerol dissimilation pathway, 31 leading to the toxic accumulation of sugar phosphates such as G3P and DHAP in Mtb. 28 When glycerol is not used as a carbon source by Mtb in vivo, compounds with such MOA do not show efficacy in the mouse model.
These results highlight the importance of validating the relevance of in vitro assay conditions to reproduce the environment encountered by Mtb in the infected host. 28,30 The culture media currently used for sufficient Mtb growth were empirically devised several decades ago for optimal propagation of the bacterium in vitro and were not optimized for drug screening efforts. Consequently, these media were not developed to replicate the conditions encountered by Mtb in vivo, thus limiting the predictive value of the in vitro assays. There is no clear consensus thus far as to which in vitro methods best reflect the in vivo biology in order to effectively identify novel candidates that will inhibit Mtb growth in vivo. 28

■ CONCLUSION
This study identified a new class of aminoquinazolinone compounds active against Mtb through phenotypic whole-cell assays. Although these compounds exhibited excellent in vitro potency and favorable pharmacological properties, they were found to be inactive in an BALB/c mouse acute TB infection model. The discrepancy between in vitro activity and in vivo efficacy was elucidated through retesting in vitro using different Mtb strains cultured in different media, which revealed that activity was only observed in media containing glycerol. This was supported by spontaneous-resistant mutant generation and whole genome sequencing studies, which revealed mutations mapping to glycerol metabolizing genes, suggesting a major difference in carbon metabolism between bacteria growing in standard TB culture medium containing glycerol compared to those found in TB-infected lungs.
■ METHODS General Methods. All reagents and solvents were obtained from commercial sources and used as received. All dry solvents were obtained from an SP-1 standalone solvent purification system from LC Technology Solutions Inc. Reactions were monitored by thin layer chromatography (TLC) using Merck TLC silica gel 60 F 254 aluminum-backed precoated plates and were visualized by ultraviolet light at 254 nm. Purification of compounds was carried out by either column chromatography on silica gel 60 (Fluka), particle size of 0.063−0.2 mm (70− 230 mesh ASTM), as the stationary phase or by Waters' HPLC using an X-bridge C18 5 μm column (4.6 × 150 mm). Additional details are as follows: organic phase, 10 mM ammonium acetate (pH 3.7) in HPLC grade methanol; aqueous phase, 10 mM ammonium acetate (pH 3.7) in HPLC grade water; flow rate, 15.00 mL/min; detector, photodiode array (PDA). All target compounds and intermediates were characterized by 1 H NMR, 13 C NMR, and MS. NMR spectra were recorded on either a Varian Mercury-300 ( 1 H 300.1 MHz, 13 C 75.5 MHz) or Bruker-400 ( 1 H 400.2 MHz, 13 C 100.6 MHz) instrument using CDCl 3 , MeOH-d 4 , and DMSOd 6 as solvents. The 1 H NMR data are reported as follows: chemical shift in parts per million (δ) downfield of tetramethylsilane (TMS), multiplicity (s = singlet, bs = broad singlet, d = doublet, t = triplet, q = quartet, dd = doublets of doublets, dt = triplets of doublets, and m = multiplet), coupling constant (Hz), and integrated value. The 13 C NMR spectra were measured with complete proton decoupling. LC−MS/MS analysis was performed using an Agilent 1260 Infinity Binary Pump, Agilent 1260 Infinity Diode Array Detector (DAD), Agilent 1290 Infinity Column Compartment, Agilent 1260 Infinity Standard Autosampler, and an Agilent 6120 Quadrupole (Single) MS with an APCI/ ESI multimode ionization source. The purities were determined by Agilent LCMS/MS or Waters' HPLC using an X-bridge C18 5 μm column (4.6 × 150 mm). Additional details are as follows: organic phase, 10 mM ammonium acetate (pH 3.7) in HPLC grade methanol; aqueous phase, 10 mM ammonium acetate (pH 3.7) in HPLC grade water; flow rate, 1.20 mL/min; detector, photodiode array (PDA). The purities of all compounds were found to be >95%.
Additional details of the synthetic protocols and characterization of all intermediates and final compounds and the procedures used for the in vitro biological assays, ADMET, and in vivo studies (PDF)