Structural Simplification of Bedaquiline: the Discovery of 3‐(4‐(N,N‐Dimethylaminomethyl)phenyl)quinoline‐Derived Antitubercular Lead Compounds

Abstract Bedaquiline (BDQ) is a novel and highly potent last‐line antituberculosis drug that was approved by the US FDA in 2013. Owing to its stereo‐structural complexity, chemical synthesis and compound optimization are rather difficult and expensive. This study describes the structural simplification of bedaquiline while preserving antitubercular activity. The compound's structure was split into fragments and reassembled in various combinations while replacing the two chiral carbon atoms with an achiral linkage instead. Four series of analogues were designed; these candidates retained their potent antitubercular activity at sub‐microgram per mL concentrations against both sensitive and multidrug‐resistant (MDR) Mycobacterium tuberculosis strains. Six out of the top nine MIC‐ranked candidates were found to inhibit mycobacterial ATP synthesis activity with IC50 values between 20 and 40 μm, one had IC50>66 μm, and two showed no inhibition, despite their antitubercular activity. These results provide a basis for the development of chemically less complex, lower‐cost bedaquiline derivatives and describe the identification of two derivatives with antitubercular activity against non‐ATP synthase related targets.


Introduction
Tuberculosis (TB) is as erious threat to human health, [1][2][3] causing 1.5 milliond eaths in 2013. [4,5] The emergence and transmission of multidrug-resistant (MDR) and extensively drug-resistant (XDR) TB strains is ap articular challenge for global TB preventiona nd treatment. [3] Bedaquiline (1)i st he first novel US Food and Drug Administration (FDA)-approved anti-TB drug for the treatment of MDR-TB cases in the past 40 years. [6][7][8][9][10][11] In vitro studies demonstrated potent inhibition of mycobacterial growth by bedaquiline, against both drug-sensitive and drugresistantm ycobacteria, with am inimum inhibitory concentration (MIC) of 0.06 mgmL À1 . [9] Bedaquiline also exhibitse xcellent clinical efficacy for the treatment of TB patients,p articularly those with MDR-TB infections. [9] It hasaremarkably long halflife, which was attributed to ah igh logP value (7.52) and cationic, amphiphilic properties resultingi nt issue accumulation. [8,12] Adverse side effects of bedaquiline, such as phospholipidosis and cardiovascular risks, may relate to these molecular features. In particular,a nN-desmethyl metabolite ("M2") was reported to be more toxic but less bactericidal. [8,13] At the molecular level the compound was shown to inhibit the mycobacterial F 1 F o ATPs ynthase by binding to its membrane-embedded F o rotor ring, ar ing-shaped assembly of identicalc -subunit copies. [6,14] Here, the half-maximali nhibitory concentration (IC 50 )o fb edaquiline, efficientlyi nhibitingt he mycobacterial ATPs ynthase, was reported to be remarkably low (25 nm). Furthermore, the mode of action of bedaquiline is highly target specific, [14] as the IC 50 values for human,b ovine, and mousemitochondrial ATPs ynthasesw ere found to be 20 000-fold higher. [10,14,15] Besides these obviousa ssets of bedaquiline, its chemical complexity in harboring two adjacent chiral centers, makes the chemicals ynthesis of this new anti-TB drug laborious and costly in the production process. [11,16] Many ongoing research programsa re focused on the optimizationo fb edaquiline to decreasei ts structuralc omplexity while maintaining its antitubercular activity. [17][18][19][20][21][22][23][24] Still, new bedaquiline analoguesa re requiredt odeliver potentialnew leads. Bedaquiline (BDQ) is an ovel and highly potent last-line antituberculosis drug that was approved by the US FDA in 2013. Owing to its stereo-structural complexity,c hemical synthesis and compound optimization are ratherd ifficult and expensive. This study describes the structural simplification of bedaquiline while preserving antitubercular activity.T he compound's structure was split into fragments andr eassembled in variousc ombinations while replacing the two chiral carbon atoms with an achiral linkage instead. Four series of analogues were designed;t hese candidates retained their potent antitubercular activity at sub-microgramp er mL concentrations against both sensitivea nd multidrug-resistant (MDR) Mycobacterium tuberculosis strains. Six out of the top nine MIC-ranked candidates were found to inhibit mycobacterialA TP synthesis activity with IC 50 values between 20 and 40 mm,o ne had IC 50 > 66 mm,a nd two showed no inhibition, despitet heir antitubercular activity. These resultsp rovide ab asis for the development of chemically less complex, lower-cost bedaquiline derivatives and describe the identification of two derivatives with antitubercular activity against non-ATP synthase relatedt argets.
The aim of this study was therefore to simplify the chemical synthesis of bedaquiline to work toward new,a lternative leads. Once these simplified scaffolds are achieved, they will be used to improvep harmacokinetic properties to decreaset he adverse side effects of bedaquiline. Based on the previously reported relationship of structure and antitubercular activity of variousd iarylquinoline analogues, [9] we dissected bedaquiline into variousf ragments and reorganized them to obtain novel antitubercular agents with simplified scaffolds. The MIC values of the new compounds were determined, and the best nine candidates were tested for their inhibitory activities against ATPs ynthesis inhibition in mycobacteria.

Design strategy
Bedaquiline (1)c ontains two adjacent chiral carbon atoms that bridge three aryl rings and ad imethylaminoethyl moiety ( Figure 1). The structure and antitubercular activity relationship of diarylquinoline analogues as well as the importance and contribution of each fragment of bedaquiline to its antitubercular activity have been quantitatively evaluated. [9] It was shown that the activity of bedaquiline is closely related to the spatiald istribution of its segments. Some fragments, such as the quinolino and dimethylamino groups,a re necessaryt o maintain its activity,w hereas other groups such as the bromo and hydroxy groups are less important. Therefore, we split bedaquiline into the core quinoline moiety,t hree major fragments (phenyl ring, naphthylmethyl, and dimethylaminoethyl), and two minor fragments (bromoa nd hydroxy) (Figure1). We then reassembledt he quinoline core with three major fragmentst od esign four new series of scaffolds. Notably,p revious structure-activity relationship (SAR) studies of bedaquiline and its analogue 2 suggested that the naphthylmethyl group can be replaced with as ubstituted benzylr ing. [9] Thus,i no ur chemicalo ptimization, we also used the substituted benzyl ring insteado fanaphthylmethyl group to design some new compounds. In addition, the minor bromoa nd hydroxy groups were added to the final compounds, asd esired.
In series 1, fragments Aa nd Cw ere attached to the tertiary carbon atom at the 3-position of the quinoline moiety,a nd fragment Bw as attached to fragment A. In series 2, fragment-sA and Cw ere attached to the tertiary carbon at the 3-position, whilef ragmentB was attached to fragment C. In series 3, fragments Aa nd Bw ere attached to the tertiary carbon at the 3-position, while fragment Cw as moved down to the 2-oxyl position. Finally,i ns eries 4f ragment Aw as attached to the 3positiono ft he quinolinem oiety,f ragment Bw as attached to fragment A, andf ragment Cw as movedd own to the 2-oxyl position. From series 1t o4 ,w es ystematicallys implified the structure of bedaquiline to remove the two chiral carbonsa nd to understand the structurea ntimycobacterial activity relationship of these new analogues.
In series 1 ( Table 1), we designedc ompounds, in which fragment Bw as attached to fragment A. In total,e ight compounds were synthesized,w hich provided primary information about the SAR of the compounds in series 1. An antitubercular activity screening revealed that compound 6a has the most potent antitubercular activity (1 mgmL À1 )i nt his series of compounds, while the growth of mammalian cells was inhibited with CC 50 > 64 mgmL À1 .A ni dentical molecule, 10 a,h arboring ab romide, is lessa ctive, suggesting that it is not required to preserve antitubercular activity.I na ddition, relative to compound 10 a,c ompound 10 b has an additional methylene unit extension between fragment Ca nd the other segments. However, compound 10 b shows less potencya gainst the growth of mycobacterium. Thus,t he greater distance between fragment C and the main scaffold likely disfavors its antitubercular activity. Other analogues from this series with similar structures exhibited only weak activity against mycobacteria, indicatingt hat the arrangement of the fragments in series 1i snot suitable to achieve the desired antimycobacteriala ctivity.N ote that all candidates of this series still contain one chiral carbon, and racemic mixtures were used to generate CC 50 and MIC values.

Compounds of series 2
Aniline (11)r eactedw ith hydrocinnamoyl chloride to afford 12, which wast hen treated with phosphoryl chloride and DMF at 100 8Ct op rovide 13 (Scheme 3). The reactiono f13 with sodium methoxide produced 14 via aS N 2r eaction. Next, 14 was activated by N-bromosuccinimide (NBS) to make 15.T he intermediate 15 was treated with substituted aniline and potassium carbonate at room temperature, or followed by ar eaction with variousa minoalkyl bromides to produce 16.T he synthesis of the derivatives required minor modifications (Supporting Information).
In series 2 ( Table 2), fragment Bw as attached to fragment C. The nitrogen or oxygen atoms located closet of ragment C were placedt om imic the hydroxy group of bedaquiline. A total of six compounds were synthesized in this series. The activity screening revealed that most of the compounds in series 2e xhibit moderate antitubercular activity.C ompound 16 e,h owever,w as nearly inactive, whichm ight have been caused by the very short distance between the terminal amine and fragment Co ft he molecule. In addition, we can also conclude from series 2t hat the bromide is not important to retain the antitubercular activity of this compound type. Furthermore, there is one chiral carbon left in the structures of all candidates in this series, and the racemic mixtures were used to generateC C 50 and MIC values. As none of the compounds from series 2e xhibited sufficient antimycobacterial activity in the very low-micromolar range, these compounds were not chosen for further chemicaloptimization.
In series 3 ( Table 3) fragments Aa nd Bw ere joined with am ethylene linker forming ap lanarc onfiguration. In addition fragment Cw as added to the quinoline core by an ether bond. This structure allowed us to remove the two chiral carbons. Accordingt ot his strategy,w es ynthesized compounds 22 and 24.C ompound 23 was synthesized for structure-activity comparison. An antitubercular evaluation showedt hat all three compounds (22)(23)(24)f rom this series exert ar easonably potent activity against mycobacteria. Thiss uggestst hat placing frag- ment Cinthe 2-methoxyl positionm aintains the antituberculara ctivity.T he comparison of compounds 23 and 24 further reveals that the planar configuration between fragments Aa nd Bd oes not affect the antituberculara ctivity.I ta lso indicatest hat the hydroxy group is not important for the antitubercular activity of this new scaffold. Moreover,n one of the compounds from series 3c ontains the bromide. Hence it does not affect the antitubercular activity of this series of compounds, in agreement with the SAR information from series 1and 2.

Compounds of series 4
The design of series 4c ompounds was based on the findingt hat the planar configuration of compound 24 (series 3) showed no negative effect on the antitubercular activity.T hus, we moved the phenyl ring (fragment A) to the double bond position between the quinoline and basic amino groups, but kept fragment Ca ttached to the 2-position of the quinoline. The synthesis of 3-aromatic analogues 30 is shown in Scheme5.T he synthesis of key substituted quinoline intermediates 26 started from substituted quinoline 25.T he reaction of 25 and N-iodosuccinimide gave 3-iodoquinoline 26,w hich was then treated with meta-chloroperoxybenzoic acid( mCPBA) to produce 27,a nd chlorinated with phosphoryl chloride to give 2-chloro-3-iodoquinoline 28.T arget compounds 30 were prepared via Suzuki-Miyaura coupling of compound 28 with the corresponding substituted boric acid or boronic acid pinacolc yclic ester to obtain 29. [27] Subsequently,t he chloro moiety was exchanged with benzyl alcohol, aromatic alcohol, phenol,t hiophenol, or aromatic amine by an ucleophilic substitution to obtain 30. In total, 14 compounds were synthesized for series 4( 30 a-30 n;T able 4), all of which contain no more chiral centers. Intriguingly,series 4compounds were generally potent against my-cobacteriaw ith MIC values of~1 mgmL À1 ,w hereas their CC 50 values were 10-fold highero na verage. Comparison of the structures and antitubercular activitieso fs eries 4c ompounds shows that the length of the terminal amine chain can be 1-2 carbons (compounds 30 a, 30 b)w hile retaining antitubercular activity.Further,the presence of the terminal basic nitrogen atom was essential to sustain the antitubercular activity (compounds 30 h, 30 m), whereas the bromogroup was not (compounds 30 e, 30 f).

Second round of optimization
Based on the resultso btained from series 4, we explored the effect of 2-aryl substituents and the spacer length of the terminal amine group in our new scaffold of 2-aryl-3-(dimethylamino-substituted)phenylquinoline on antitubercular activity ( Figure 2).
In the first group of compounds (Table5), we investigated the positiona tw hich the basic nitrogen containing side chain was attached to the phenyl ring as well as the lateral chain length. We assembled as mall library of 6-bromo-2-((naphth-1yl)methyl)oxy-3-substituted phenylquinoline analogues (32). The synthetic route is outlinedi nS cheme 6. The antitubercular screening assay showedt hat:1 )the para-substituted N,N-dimethylaminomethylphenyl ring has the mostp otent antitubercular activity for this scaffold,a nd 2) the terminal nitrogen atom is necessary for antitubercular activity.O ther substituents decreased the antitubercular activity of this scaffold significantly.C ompounds 32 a, 32 d,a nd 32 e showed antitubercular activities (MIC) of 0.43, 0.47, and 0.44 mgmL À1 ,a nd cytotoxicity (CC 50 )a gainst mammalianc ells of 13, 12, and 3.9 mgmL À1 ,r espectively.
Next, we substituted the naphthalen-1-ylmethanol ring with variousaryl groups that have different physicochemical properties, while keeping 4-(N,N-dimethylaminomethyl)phenyl group as the 3-substituent. As mall group of analogues 34 were synthesized (Scheme 7). Antitubercular screening showed activity of severald erivatives with MIC~0.4 mgmL À1 ( Table 6). The arylmethyloxyg roup at C2 of the quinolone was substituted with av ariety of aromatic groups having different physicochemical properties, such as a para-chlorophen-1-ylmethanol ring (34 i,    ). These substituents maintain the antitubercular activity,a nd may provideamethodf or optimizing the physicochemical properties of the final compounds.

Inhibition of ATPs ynthesis activity by selected compounds in mycobacteria
The design of the new antimycobacterial lead compounds was inspiredb yb edaquiline, which is known to inhibit mycobacterial ATPs ynthase with an IC 50 value in the low nanomolar range (Supporting Information Figure S1). [14] To determine if the new compoundss hare this property,n ine candidates with MIC values in the sub-microgram per mL range were selected (Table S1)a nd tested for their ability to inhibitt he ATPs ynthesis activity of Mycobacterium phlei inverted membrane vesicles (IMVs) ( Figure 3A). Ac oncentration range of 0-100 mm was assayed for each compound. The calculated activitiesw ere used to determine an IC 50 value for each of the candidates (Figure 3B). Six of the compounds (34 m, 32 h, 34 a, 34 f, 32 e,a nd 32 o)i nhibited ATPs ynthesis with IC 50 values between 20 and 40 mm ( Figure S2). One candidate (34 i)s howed as light inhibitory effect;h owever, the IC 50 value could not be calculated due to the low solubility of the compound. The best candidate was compound 34 m with an IC 50 value of 20.3 AE 1 mm ( Figure 3C)a nd an improved logP of 5.55 (Table S1), which may also be beneficial from ap harmacokinetic point of view and which may help decrease some of the adverse side effects of bedaquiline. [8,12] Remarkably,t wo compounds (32 d and 32 p) did not affect ATPs ynthesis at all, while still displaying potent antimycobacterial activity (MIC 90 :0 .47 and 0.51 mgmL À1 ,r espectively;F igure S2). As ac ontrol, inhibition by bedaquiline was assayed, exhibiting an IC 50 value < 10 nm ( Figure S1). All seven tested compounds,w hich inhibited ATP synthesis activity,s hare as tructuralf eature with bedaquiline, that is, the presenceo fadimethylamino (DMA) moiety.T his structuralp art of bedaquiline represents the most important structural feature in the interaction with the ATPs ynthase c-ring and is hence ad ecisive elemento ft he molecule's inhibitory power. [14] In the lead compounds described herein, the DMA moietiesa re all attached to longera ryl side chainst han the ethyl chain present in bedaquiline. Therefore, the subtle but exact structural placemento ft his critical moiety is impaired, which explainst he higher IC 50 values measured for these derivatives. In agreement with that notion are the two candidates (32 d and 32 p), which show no influence on ATP synthesis activity and lack this structural element completely. The fact that these compounds still show specific killinga ctivity on mycobacteria (Table S1) implies that they targeta nd affect other cellular processes or proteins in mycobacterial cells. Future work needs to be carriedo ut to identify these potentially interesting targets.

In vitro anti-TB activities against drug-resistant clinical isolatesofM. tuberculosis
To further explore the effect of selected compounds on drugresistantc linicali solates of M. tuberculosis and comparei tw ith the effect on the drug-sensitive H37Rvs train, we performed MIC determinationsw ith two representative compounds that showedA TP synthesis inhibitory activity (34 m and 32 e)a nd two candidates that showed only marginal and no effect on the ATPs ynthase, 34 i and 32 d,r espectively ( Figure 3C). Thei n vitro inhibitory activities of these four compounds were screened against two drug-resistantc linical isolates (isolates 12513 and 6133), whicha re both resistant to isoniazid (INH) and rifampicin (RFP). [28] The MIC values against the drug-sensitive H37Rv strain of M. tuberculosis werea lso determined again (

Conclusions
In this study we describe the chemical simplification of bedaquiline, while largely retainingt he potent antitubercular activity.I nspired by the SAR of bedaquiline, we designed as et of new lead compounds using af ragment-based approach. The bedaquiline molecule was first split into its various functional groupsa nd recombined in aw ay that decreased or completely removed the initial stereochemical complexity. The approach promises to significantly decreasep roduction costs and make it accessible for ab roader varietyo fc hemical laboratories and clinics. Initially,f our series of compounds were designed, synthesized, and evaluated for their antitubercular activity.A mong the initial four series of compounds, series 4, incorporating a2 -aryl moiety and a3 -(4-(N,N-dimethylamino)methyl)phenyl group, was found to contain the most potent analogues with MIC values < 1 mgmL À1 .F urther optimization of this scaffold, modifying the C2 and C6 positions of the quinoline, yieldeds everal potent inhibitors with MIC < 0.6 mgmL À1 .T he most potent candidates were selected and assayed for their capacity to inhibitA TP synthesis activity in mycobacterial membranes.O ur biochemicala nalysis reveals that some of the newly synthesized compounds showedadirect impact on mycobacterial ATPs ynthesis, with IC 50 values between 20 and 40 mm,while others did not affect ATPs ynthesis itself. The latter compounds are promising new candidates that are able to targeto ther cellular processes in mycobacterial cells. This study provides the basis for the development of novel, chemically simplified,l ow-cost bedaquiline derivativest of ight drug-resistant M. tuberculosis strains.

Experimental Section
Chemistry:A ll reagents and anhydrous solvents are commercially available and were used without further purification. Te mperatures are given in degrees Celsius (8C). NMR spectra were recorded on Bruker AVIIIHD spectrometers using TMS as internal standard. Chemical shifts (d)a re reported in ppm, and coupling constants (J) are given in hertz (Hz). The following multiplicity abbreviations are used:( s) singlet, (d) doublet, (t) triplet, (q) quartet, (m) multiplet, and (br) broad. ESI-HRMS data were measured on aT hermo Exactive Orbitrap Plus spectrometer.A ll reactions were monitored by TLC. Column chromatography was carried out with silica gel (200-300 mesh size). Flash column chromatography was performed on aB iotage Isolera One instrument. Purity was determined by LC-MS and NMR spectroscopy.A ll final compounds are > 95 %p ure. See the Supporting Information for detailed synthesis and characterization of all intermediates and final compounds.