Syntheses and Structure–Activity Relationships of N-Phenethyl-Quinazolin-4-yl-Amines as Potent Inhibitors of Cytochrome bd Oxidase in Mycobacterium tuberculosis

The development of cytochrome bd oxidase (cyt-bd) inhibitors are needed for comprehensive termination of energy production in Mycobacterium tuberculosis (Mtb) to treat tuberculosis infections. Herein, we report on the structure-activity-relationships (SAR) of 22 new N-phenethyl-quinazolin-4-yl-amines that target cyt-bd. Our focused set of compounds was synthesized and screened against three mycobacterial strains: Mycobacterium bovis BCG, Mycobacterium tuberculosis H37Rv and the clinical isolate Mycobacterium tuberculosis N0145 with and without the cytochrome bcc:aa3 inhibitor Q203 in an ATP depletion assay. Two compounds, 12a and 19a, were more active against all three strains than the naturally derived cyt-bd inhibitor aurachin D.

The N-phenethylquinazolin-4-amine class, exemplified by compound 3 (Figure 1), has an IC 50 of 11 µM against Mycobacterium bovis BCG and an IC 50 of 27 µM against Mycobacterium tuberculosis H37Rv as determined by our published ATP depletion assay [13]. This assay is purposefully run in the presence and absence of Q203 (4, a selective cyt-bcc:aa 3 inhibitor) to indicate whether a compound inhibits alone or synergizes due to the conditional essentiality of cyt-bd to maintain ATP homeostasis once cyt-bcc:aa 3 is selectively inhibited by Q203. Compound 3 was inactive (IC 50 > 25 µM) in the absence of Q203 which is an indication of on-target potency based upon assay design. Similarly, compound 3 was amenable to chemical modifications and structure-activity relationship (SAR) development. Herein, we report our initial design, synthesis, and activity assessment of various quinazoline compounds for activity against cyt-bd of Mycobacterium bovis BCG and Mycobacterium tuberculosis.

LC-MS method:
The liquid chromatography-mass spectrometry method was performed on an Agilent 1290 infinity coupled to Agilent 6538 Ultra High-Definition Quadrupole Time of Flight (UHD-QTOF) instrument. A separation was achieved by using reverse phase Waters Acuity UPLC HSS T3 1.8 µm (2.1 × 100mm) column from Waters (Milford, MA, USA). All solvents were purchased from Fischer Scientific LCMS Optima grade solvents. Water containing 0.1% formic acid was used as mobile phase A and acetonitrile containing 0.1% formic acid was used as mobile phase B. The injection volume was set at 1 µL. Samples were injected in a gradient of 95% mobile phase A and 5% mobile phase B in the initial condition to 5% mobile phase A and 95% mobile phase B in 9 min. The eluent was held at that composition for an additional 3 min and switched back to the initial condition at 12 min.
The MS data acquisition was performed from 50-1000 m/z at 1.0 spectra/sec scan rate. The source gas temperature was set at 350 °C with a flow of 8 l/min. The nebulizer gas was set at 55 psig. The capillary voltage was set at 3500 volts with fragmentor at 100, skimmer at 45 and octopole RF 500 volts. Prior to sample runs, the instrument was calibrated using Agilent low mass calibrant solution.
Data analysis: The data collected in Agilent LC-MS was analyzed using Agilent Mass Hunter software for HRMS calculation.
General procedure B for acid catalyzed S N Ar reaction for the preparation of compounds 3, 8a, 10a, 11a. In a sealed vial, desired 4-chloroquinazoline 16-26 (1 equiv.) and 2- [4-(trifluoromethoxy)phenyl]ethylamine (7, 1 equiv.) were dissolved in a 3:1 tetrahydrofuran: 2-propanol solution (4 mL). Next, was added 12 M HCl (~0.4 equiv.). The solution was heated at 70 °C for 24 h. The reaction was concentrated to dryness. The residue was dissolved in CH 2 Cl 2 and washed with saturated aqueous NaHCO 3 solution, water, and brine. The organic phase was collected, dried over sodium sulfate, filtered, and concentrated in vacuo. Crude material obtained was purified by either silica gel column chromatography with a gradient of CH 2 Cl 2 : ethyl acetate: solvent system (0-80%) or recrystallized from hot isopropanol or acetonitrile to afford the product.

Biological Assessments
Mycobacterium bovis BCG and Mycobacterium tuberculosis strains were routinely cultured in Middlebrook 7H9 medium (Becton Dickson and Company Limited, Franklin Lakes, NJ, USA) supplemented with 0.05% tween 80, 0.5% glycerol, bovine serum albumin fraction V, D-glucose, and NaCl. M. tuberculosis clinical isolate N0145 was a gift from Sebastien Gagneux (Swiss Tropical and Public Health Institute at the University of Basel). Prior to use, bacteria cells were pelleted and resuspended in medium without glycerol and adjusted to a cell density of circa 10 7 CFU/mL. The test compounds were tested in the presence or absence of Q203. Prior to the assay, Q203 or DMSO (solvent control) was added to the bacteria cultures. Q203 was used in excess (at 100 nM) to completely inhibit the function of the cytochrome bcc:aa 3 , thereby revealing the activity of cyt-bd in the assay. The Q203 ATP IC 50 for M. bovis BCG, M. tuberculosis H37Rv, and M. tuberculosis N0145 are 2.6, 1.0, and 2.5 nM, respectively (Supplementary Materials Figure S61). The test compounds were tested in eight concentrations, starting at 25 µM, in two-fold dilutions (0.2-25.0 µM) in the presence or absence of 100 nM Q203. 1 µL of test compounds of varying concentrations was added to each well of 96-well white plates, and 100 µL of bacterial culture was subsequently added. The assay plates were incubated at 37 °C for 15 h, after which the BacTiter-Glo™ (Promega, Madison, WI, USA) reagent was added. Following a 12-min incubation, the luminescence of each plate was measured using a BioTek Cytation 3 Cell Imaging Multiple mode reader. IC 50 values were determined using GraphPad Prism 9.

Results and Discussion
Our initial efforts were to probe the effect of alteration and modification of the phenethylaniline moiety around a fixed quinazoline core. This was done by syntheses of 10 compounds and screening against M. bovis BCG and Mtb strains (H37Rv and N0145) to assess activity against mycobacterial cyt-bd (Table 1). A representative compound from this set (7a) served as the foundation toward the design of a second set of 11 compounds that focused on modification of the quinazoline core and were subsequently screened for cyt-bd activity ( Table 2).
The second set of compounds (16a-26a) was prepared by reacting eight functionalized 4chloroquinazolines (16-26) with 4-(trifluoromethoxy)benzylamine (7) using standard S N Ar conditions (Scheme 2). As with the first set, the activity was determined against cyt-bd in M. bovis BCG and M. tuberculosis H37Rv by ATP readout ( Table 2).

Structure-Activity Relationship Studies
The structure-activity-relationship (SAR) studies revealed that the cyt-bd of M. bovis BCG was more sensitive to inhibition than M. tuberculosis H37Rv (Table 1). This mirrors the trend observed previously and is indicative that there is a greater expression of cyt-bd in the lab-adapted M. tuberculosis H37Rv strain relative to the M. bovis BCG or "fresh" Mtb clinical isolate N0145 [13]. This is further supported by the activity of aurachin D [7], a known synergistic cyt-bd inhibitor of Mtb, which had greater activity against M. bovis BCG in our ATP assay relative to M. tuberculosis H37Rv (2.9 µM vs. 5.5 µM, respectively; Table   1). In addition, based on our assay, compounds 6a-15a are all specific inhibitors of cyt-bd as they show no activity against wild-type (WT) M. bovis BCG or M. tuberculosis H37Rv (Mtb-H37Rv), whereas the positive control bedaquiline (BDQ), a selective ATP synthase inhibitor, remains very potent against both strains (Table 1). Interestingly, aurachin D showed the expected synergy with Q203 against mycobacterial strains but was unexpectedly active against wild-type Mtb-H37Rv (Table 1). SAR trends revealed that the pendant aryl group can accommodate a variety of steric and electronic changes and retain good activity ( Table 1). Evaluation of the compounds with electron withdrawing substituents 6a (p-CF 3 ), 7a, (p-OCF 3 ), 8a (p-Cl), 9a (p-SF 5 ), indicated that 7a is the most active compound with IC 50 values of 2.1 µM against M. bovis BCG and 11 µM against Mtb H37Rv. We explored electron withdrawing groups at the metaand ortho-positions of the pendant aryl group by comparing compound sets 7a (para-OCF 3 ) and 13a (meta-OCF 3 ) and 6a (para-CF 3 ) and 15a (ortho-CF 3 ). We found that the meta-OCF 3 compound (13a) position was less active compared to the para-OCF 3 (17a). Whereas incorporation of substituents in the ortho-position gave slightly improved potency against M. bovis but equal activity (within error) in Mtb-H37Rv as demonstrated by comparing 6a and 15a. When evaluating the compounds with electron donating substituents, 10a (p-CH 3 ), 11a (p-OCH 3 ), and 12a (p-tert-butyl), 12a was found to be significantly more active with IC 50 values of 0.8 µM against M. bovis BCG and 5.8 µM against Mtb H37Rv. That is an improvement over thieno [3,2-d]pyrimidine (1) and activity on par with the optimized quinazoline ND-11992 (2). This is particularly interesting as it follows the Topliss aromatic substitution decision tree pattern [23] where one would first evaluate the phenyl (3), then para-chloro (8a) and if equally potent (as it is in our assays) then prepare the tert-butyl (12a) to derive the most potent compound. While the Topliss approach can often lead to compounds with improved activity it also directs the syntheses of compounds which often have metabolic liabilities (CH 3 , OCH 3 , t-butyl) or toxic pharmacophores (NH 2 , NO 2 , I). For this reason, we chose 7a which bears a para-trifluoromethoxy group (which is prevalent in various anti-TB drugs such as delamanid, pretomanid, and telacebec) as the basis for our second set of compounds (Table 2) because it possesses good activity (IC 50 ′s of 2.1 and 11 µM against BCG and Mtb-H37Rv, respectively).
The SAR trends for alteration of the quinazoline core suggest an interplay between both steric and electronic effects ( Table 2). For instance, the 6-position was probed with three substituents-fluoro (16a), methyl (17a) and methoxy (18a)-which resulted in a wide spectrum of activity. The 6-methyl analog (17a) was the most active against M. bovis BCG (IC 50 of 7.5 µM) but weakly active against Mtb H37Rv (IC 50 of 22 µM). Whereas the fluoro (16a) and methoxy (18a) analogs had diminished activity against M. bovis BCG (IC 50 of 13 µM and 27 µM, respectively) and both were >25 µM against Mtb H37Rv. The 7-position was explored with five different substituents-fluoro (19a), chloro (20a), bromo (21a), trifluoromethoxy (22a) and methoxy (23a)-resulting in only two active compounds (19a and 21a). The 7-fluoro analog (19a) displayed good activity against both mycobacterial strains (IC 50 of 0.8 µM and 7.6 µM; respectively) and the 7-bromo (21a) showed much better activity against M. bovis BCG than Mtb H37Rv (IC 50 of 7 µM and 20 µM; respectively). The most interesting SAR was observed with substitution of the quinazoline 2-position with methyl (24a) and cyclopropyl (25a) groups. Both compounds displayed good activity range (3-12 µM) with or without cyt-bcc:aa 3 inhibition, suggesting possible dual modes of action. Lastly, one compound (26a) bore both 7-fluoro and 2-methyl quinazoline substituents which biased activity back towards cyt-bd (IC 50 of 7 and 17 µM with Q203, and 28 µM and 24 µM without Q203; respectively) Our previous work had shown that the laboratory adapted Mtb H37Rv strain over expresses cyt-bd resulting in higher IC 50 values than those observed using a clinical Mtb isolate [24]. These higher IC 50 values render compound ranking more challenging. To gain greater insight on the activity of these compounds, they were re-screened against the clinical Mtb isolate N0145 strain. As observed previously with the thieno [3,2-d]pyrimidin-4-amine class of cyt-bd oxidase inhibitors [13], these compounds displayed greatly improved activity against the Mtb clinical isolate likely due to lower expression of cyt-bd within clinical isolates as compared to lab-adapted strains like H37Rv [12]. Two compounds, 7a and 12a, from the first set were more active than or comparable in activity to aurachin D (0.1 and 1.1 µM vs. 1.5 µM, respectively). While six additional compounds (6a, 9a, 10a, 13a-15a) had IC 50 values below 5 µM against N0145-Mtb. In the second set of compounds, only 19a had activity superior to aurachin D (0.2 µM vs. 1.5 µM, respectively) and four additional compounds (17a, 21a, 24a-26a) had activity below 5 µM. The two most potent compounds, 12a and 19a, had slightly improved activity relative to ND-11992 which was active in the murine infection model of tuberculosis. However, preliminary metabolic stability assessment of these compounds against human and rat microsomes indicate much more rapid metabolism than that of ND-11992 (data not shown).

Conclusions
The described N-phenethylquinazolin-4-amine class of compounds are mycobacterial cyt-bd inhibitors. Through focused SAR, the activity of the hit compound 3 (IC 50 = 6.9 µM against Mtb N0145) was improved with 14 of the 22 analogs (64%) and two, 12a and 19a, had sub-micromolar activity (0.1 and 0.2 µM against Mtb N0145, respectively). Future work includes identification of more efficacious compounds through additional SAR around the quinazoline core. These results will be reported in due course.