Targeting Latent Tuberculosis Using Hypoxia-Activated Prodrugs of Bedaquiline Analogues

Abstract

The increasing emergence of antimicrobial resistance (AMR) is reducing the effectiveness of existing antibiotics against bacterial infections, including tuberculosis (TB), a deadly disease caused by Mycobacterium tuberculosis (M.tb). Approximately 90% of infected individuals develop latent tuberculosis infection, for which current treatment options are limited. Latent tuberculosis infection is characterised by the formation of hypoxic granulomas, which restrict oxygen and drug penetration, highlighting the need for therapies effective under hypoxic conditions. In this study, a hypoxia‑activated prodrug (HAP) strategy was employed to improve selective drug delivery of promising anti-tuberculosis drugs bedaquiline, TBAJ-587 and TBAJ-876 (sorfequiline) to hypoxic TB granulomas. Thirty-six nitroimidazole-based hypoxia-activated prodrugs (HAPs) were successfully synthesized and evaluated for antitubercular activity. The HAPs exhibited potent in vitro activity under both oxic and hypoxic conditions, with consistently lower MIC values observed in the hypoxic assays. Assessment of hypoxic selectivity demonstrated enhanced trigger cleavage under hypoxic conditions, providing strong evidence for the hypoxia-selective activation of these compounds. Collectively, these findings highlight the therapeutic potential of hypoxia-mediated strategies for the treatment of latent tuberculosis, particularly within hypoxic niches that drive disease persistence and prolonged treatment regimens.

Tuberculosis (TB) is a serious airborne disease that is caused by the bacillus Mycobacterium tuberculosis (M.tb). According to the 2024 Global Tuberculosis Report by the World Health Organization (WHO), it was the leading infectious cause of death, with an estimated 10.8 million new cases and 1.25 million deaths per annum. Latent tuberculosis (LTB), in which M.tb develop into a nonreplicating persistent (NRP) state, is one of the main mechanisms of TB virulence and occurs in 90% of individuals infected with M.tb. According to recent reports, more than a quarter of the global population has LTB infection. LTB can survive asymptomatically in the host for decades, and only becomes active when the host becomes immunocompromised. The current treatment regimen for LTB involves the use of first-line drugs (rifampicin and isoniazid) over an extended duration of three to nine months achieving a cure rate of up to 90%. Incorrect drug use and poor treatment adherence can lead to multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) which are resistant to key first and second-line therapies. , Due to limited treatment options and high mortality, these drug-resistant forms of TB pose serious global health challenges and contribute to the development of antimicrobial resistance (AMR). Declared by the WHO as one of the top global public health threats, AMR lowers the efficacy of existing antibiotic agents against bacterial infections including TB, highlighting the urgent need for new anti-TB agents.

Bedaquiline 1 (BDQ 1, also marketed as Sirturo), a first-in-class diarylquinoline (DARQ) that acts through a novel mechanism, inhibiting the M.tb ATP synthase, received U.S. Food and Drug Administration (FDA) accelerated approval in 2012 for treatment of MDR- and XDR-TB (Figure ). , The use of BDQ in MDR/XDR-TB treatment has been shown to improve clinical outcomes and lower overall mortality.

1.

Structures of BDQ 1 and the next-generation analogues TBAJ-587 2 and TBAJ-876 3 (sorfequiline). Common structural features of the diarylquinolines, the bromoquinoline and dimethylamine moieties, are highlighted in green and pink, respectively.

Despite its therapeutic benefits, BDQ 1 has several significant safety concerns. These include a long terminal half-life, which may increase the risk of acquired resistance and phospholipidosis, as well as a human ether-a-go-go-related gene (hERG) channel liability that can lead to cardiac toxicity. Hepatotoxicity and clinically relevant drug–drug interactions with other tuberculosis medications have also been reported. ,

In response to the limitations of BDQ 1, the next-generation bedaquiline analogues TBAJ-587 2 and TBAJ-876 3 were developed. These compounds demonstrated improved pharmacological profiles, including greater potency, reduced lipophilicity, and lower cardiac risks compared with BDQ 1. Both compounds have now successfully completed phase I clinical trials. , TBAJ-876 3, recently assigned the generic name sorfequiline, is undergoing evaluation in a phase II clinical trial. Early results based on patient responses appear encouraging, and TBAJ-876 3 shows particular promise as a next-generation therapeutic candidate.

Although the bactericidal activity of BDQ 1 and improved TBAJ analogues against MDR-TB has been well-documented in earlier studies, emerging evidence suggests that BDQ 1 may have limited penetration into TB granulomas. , Granulomascentral to both latent TB infection and active diseaseare complex, dynamic structures characterized by a hypoxic microenvironment and a core of lipid-rich necrotic material known as caseum. M.tb can persist within the caseum in a dormant or NRP state, where it is shielded from adaptive immune responses. Moreover, the abnormal vasculature associated with granulomas slows blood flow and reduces metabolic activity, impairing the effective delivery of several frontline TB drugs, including BDQ 1. These structural and physiological barriers can hinder drug diffusion into the hypoxic caseous regions, complicating extravascular transport and contributing to treatment resistance in TB patients.

In the C3HeB/FeJ mouse model, which develops hypoxic granulomas resembling those observed in human tuberculosis patients, BDQ 1 exhibited poor penetration into the necrotic caseum, resulting in a pronounced concentration gradient between the cellular and caseous regions of granulomas. Consistent with these findings, studies in a rabbit TB model demonstrated that BDQ 1 penetrates the caseum slowly and only achieves subtherapeutic concentrations within necrotic lesions. Although the next-generation diarylquinolines TBAJ-587 2 and TBAJ-876 3 display improved caseum penetration, suggesting a reduced risk of resistance emergence, recent studies have indicated that BDQ 1 and TBAJ-587 2 are not only potent inhibitors of M.tb ATP synthase but also of human ATP synthase, highlighting that TBAJ-587 2 reduces cardiac toxicity only partially, without resolving mitochondrial toxicity. These challenges emphasize potential safety and toxicity concerns and emphasize the need to develop improved TB agents that offer greater selectivity, reduced toxicity, and improved delivery and penetration into TB granulomas.

Drugs that specifically target LTBs are crucial in controlling the TB epidemic. A possible approach to overcome these issues and selectively deliver drugs to the TB granuloma is to use hypoxia-activated prodrugs (HAPs), which selectively release active drugs in low-oxygen regions. TB granulomas are reported to have an average oxygen partial pressure of 1.6 mmHg, and this level is 5-fold more hypoxic than typical median oxygen levels in malignant tumors in lungs of 10 mmHg. The HAP strategy has been explored extensively in anticancer drug therapy to increase the selectivity and reduce the toxicity. , Under hypoxia, M.tb has been shown to have reduced susceptibility to standard antitubercular drugs but increased susceptibility to nitroimidazole drugs, such as metronidazole. This is attributed to the activation of metronidazole under anaerobic conditions, which occurs through a one-electron reduction of its nitro group at a reduction potential of approximately −480 mV. Despite the limited understanding of the intracellular redox potential of mycobacteria, there have been reports of microbial electron transport chains having a reduction potential of −420 mV or below under anaerobic conditions. − Additionally, 4-nitroimidazoles have been reported to have reduction potentials of −450 mV and lower. The presence of a positive charge on the adjacent linker of the 4-nitroimidazole, for example, on quaternary ammonium salt compounds, brings the reduction potential up into range for hypoxia-selective reduction (−350 to −400 mV) and provides a rationale for incorporating a similar drug design strategy in this work.

In a recent report, nitroprodrugs of the antitubercular drug moxifloxacin were developed, targeting NRP M.tb with the overall aim of addressing phenotypic antimicrobial resistance in TB. The prodrugs were designed to undergo reduction by endogenous nitroreductases within hypoxic M.tb to release moxifloxacin and were observed to undergo nitroreductase-induced cleavage by Escherichia coli (E. coli) nitroreductase NfsB, a type I nitroreductase that uses oxygen-insensitive two-electron reduction. Additionally, to verify prodrug activation in mycobacteria, the lead prodrug was incubated with whole-cell lysate of Mycobacterium smegmatis and was found to yield moderate levels of moxifloxacin, indicating efficient enzymatic bioreduction of the prodrug into the active drug. While this work highlights the validity of implementing a prodrug approach to address challenges associated with drug resistance and delivery in the treatment of LTB, activation of the prodrugs did not occur via a hypoxia-selective pathway. Hence, this work sought to explore the application of a prodrug strategy to target LTB by leveraging hypoxia-selective activation.

To achieve this, we aimed to prepare a series of nitroimidazole quaternary ammonium salt prodrugs by combining diarylquinolines with a nitroimidazole moiety, which would undergo fragmentation via one-electron reduction to release the diarylquinoline in hypoxic regions of the LTB. BDQ 1 and its next-generation analogues have a dimethylamine side chain, which is an essential moiety for binding to M.tb ATP synthase. From the data available in literature reports and our previous work, − any modification at this dimethylamine position dramatically reduces the activity of diarylquinolines, making them inactive. We predicted that the addition of a nitroimidazole trigger to the dimethylamine side chain (as the point of attachment) would disrupt the binding to both human and M.tb ATP synthase, leading to deactivation of the prodrug until it is released in the desired hypoxic regions in its active form as the parent drug, resulting in reduction of potential side effects and unwanted toxicities (Scheme ).

1. Proposed Synthetic Design of Hypoxia-Activated Prodrugs.

As a pathogen, M.tb employs multiple mechanisms to resist antibiotics, including energy-dependent drug efflux systems. In addition, the highly lipophilic mycobacterial cell envelope and its markedly negative zeta potential limit drug permeability and intracellular accumulation. It was therefore hypothesized that this strategy could further enhance the pharmacokinetic properties of diarylquinolines by mitigating their high lipophilicity and poor solubility. In comparison with the parent compounds, the proposed HAPs exhibit reduced lipophilicity and improved solubilityphysicochemical characteristics that would be advantageous for TB granuloma penetration.

We report the synthesis and characterization of novel HAPs of BDQ 1, TBAJ-587 2, and TBAJ-876 3 (parent drugs/effectors). In vitro antitubercular activities of the HAPs and effectors against M.tb H37Rv were assessed on the microplate alamarBlue assay (MABA) and low-oxygen recovery assay (LORA) to evaluate activity against replicating M.tb and NRP M.tb under hypoxic conditions, respectively. A series of 4-nitroimidazole HAP analogues featuring electron-donating and electron-withdrawing groups at C-2 were prepared to provide prodrugs with varying reduction potentials. This work systematically elucidates how substituent groups influence HAP reduction potentials and thereby control efficient fragmentation into the active parent drugs. These structure–redox insights provide a foundation for optimizing prodrug activation within hypoxic TB granulomas, enhancing selectivity and therapeutic efficacy.

Chemistry

Preparation of the HAPs proceeded using the Menshutkin reaction between 4-nitroimidazole bromides and the diarylquinoline core (Scheme ). The 4-nitroimidazole triggers were synthesized using previously described or modified methods, and the diarylquinolines were prepared using literature procedures.

2. General Synthetic Approach toward Diarylquinoline HAPs.

Six nitroimidazole triggers were targeted for synthesis (Table ). Hammett constants for the C-2-substituted nitroimidazoles were estimated using standard σ_p values reported for benzene derivatives. These compounds encompassed a spectrum of trigger groups: neutral (4), electron-withdrawing (5, 7, and 9), and electron-donating (6 and 9). The differing electronic properties of these substituents were expected to influence the rate at which the prodrugs fragmented to release the parent diarylquinoline molecules, thereby helping us identify the optimal electronic characteristics required for hypoxia-selective activation.

1. Nitroimidazole Triggers Prepared in This Work, with Variation at C-2, and Corresponding Approximated Hammett Constants.

compound	R	σp
4	H	0.00
5	Br	0.23
6	OMe	–0.27
7	CN	0.66
8	CONH2	0.36
9	SMe	0.00

Preparation of 4-Nitroimidazole Triggers

Triggers 4 and 5 were synthesized by following previously reported synthetic routes (Scheme ).

3. Optimized Syntheses of Nitroimidazole Triggers .

^a Reagents and conditions: (i) MeI, K₂CO₃, DMF, 42 h, 65%; (ii) NBS, MeCN, 1000 W, reflux, 5 h, 19%; (iii) Br₂, CHCl₃, 25 °C, 2 h, 58%; (iv) NBS, MeCN, 1000 W, reflux, 5 h, 64%; (v) NaOMe, MeOH, 130 °C, 3 h, 59%; (vi) NBS, MeCN, 0 °C 5 h, 24%; (vii) K₄[Fe­(CN)₆], CuI, NMI, 140 °C, 19 h, 47%; (viii) NBS, MeCN, reflux, 5 h, 64%; (ix) H₂SO₄, 65 °C, 1.5 h, 70%; (x) NaSMe, MeOH, 25 °C, 2 h, 78%; (xi) NBS, BPO, CHCl₃, 25 W, 19 h, 23%.

The synthesis of trigger 6 began with the preparation of intermediate nitroimidazole 13 via methoxide displacement of compound 12. Subsequent rapid radical bromination of intermediate 13 was attempted by irradiating the reaction with a 1000 W halide lamp for 10 min. However, this resulted in the formation of a complex mixture of products, with the desired product 6 being isolated in only 5% yield. Alternative in-solution bromination methods were optimized to successfully synthesize trigger 6 in 24% yield.

2-Cyano derivative 14 was prepared from intermediate 12 via a Cu­(I)-catalyzed cyanation using potassium hexacyanoferrate­(II) (K₄[Fe­(CN)₆]) in 1-methylimidazole (NMI) at 140 °C for 19 h and gave the desired compound in 47% yield. Next, nitroimidazole trigger 7 was successfully synthesized via a one-step NBS-mediated radical bromination of compound 14 under reflux for 5 h to afford the desired product in 64% yield. This synthetic pathway presented an improved, optimized route toward trigger 7, as it was accessed in two steps compared to the previously reported route of four steps in an improved overall yield of 30% from 21%.

Trigger 8 could be accessed via further elaboration of trigger 7 by acid hydration using sulfuric acid at 65 °C for 1.5 h to give the desired trigger 8 in 70% yield.

Access to novel trigger 9 could be achieved from intermediate 12 undergoing sulfide displacement at the C-2 bromide position using sodium thiomethoxide in methanol to afford 15 in 78% yield. After various attempts at the problematic radical bromination of 15 (Table S1), 9 could be accessed from optimized in-solution bromination in 23% yield.

Synthesis of the Diarylquinolines BDQ 1, TBAJ-587 2, and TBAJ-876 3

Diarylquinolines were accessed by LDA-mediated coupling of two key intermediates X and Y (Scheme ) following published reports. ,

4. General Scheme of Diarylquinoline Synthesis via Coupling of X and Y Units.

The coupling reactions toward BDQ 1, TBAJ-587 2, and TBAJ-876 3 are presented in Schemes , , and , respectively; diarylquinolines were obtained via similar coupling conditions between the corresponding X units (16, 18, and 20) and Y units (17 and 19).

5. Synthesis of BDQ 1 .

^a Reagents and conditions: (i) LDA, THF, −78 °C, 5 h, 33%.

6. Synthesis of TBAJ-587 2 .

^a Reagents and conditions: (i) LDA, THF, −78 °C, 5 h, 34%.

7. Synthesis of TBAJ-876 3 .

^a Reagents and conditions: (i) LDA, THF, −78 °C, 5 h, 46%.

As this coupling reaction generates two chiral centers, diarylquinoline products were obtained as a mixture of (RS,SR) and (RR,SS) diastereomers in the case of BDQ 1 and TBAJ-876 3, while (RR,SS) and (RS,SR) diastereomers were attained for TBAJ-587 2. Silica column chromatography was used to purify and separate the crude diastereomeric mixtures into pure diastereomeric pairs. The (RS,SR) and (RR,SS) diastereomers are the high- and low-R _F compounds, respectively, for BDQ 1a and 1b and TBAJ-876 3a and 3b. For TBAJ-587 2a and 2b, the (RR,SS) and (RS,SR) diastereomers are the high- and low-R _F compounds, respectively. The most potent, desired isomer (the (R,S) isomer for BDQ 1 and TBAJ-876 3 and the (R,R) isomer for TBAJ-587 2) is present in the high-R _F diastereomeric mixture. −

Preparation of Hypoxia-Activated Prodrugs of Diarylquinolines

HAPs were prepared by coupling the high- and low-R _F diastereomers of diarylquinoline effectors with nitroimidazole triggers 4, 5, 6, 7, 8, and 9 via the Menshutkin reaction. Similar Menshutkin reactions have been reported in the literature; examples include the preparation of a series of quaternary ammonium iodide amphiphilic anthranilamides as antimicrobial peptide mimics with potential antibacterial activity against Staphylococcus aureus (S. aureus) and E. coli and the preparation of the multikinase inhibitor SN29966, which was later developed into the nitroimidazole HAP tarloxotinib bromide. ,,

After careful optimization of reaction conditions, 36 prodrugs 21–56 (Schemes and and Tables –) were successfully synthesized. The reaction was best carried out in NMP with 1.3–1.4 equiv of trigger and an optimized reaction time of 72 h. Prodrugs were purified by recrystallization from a mixture of dichloromethane and diisopropyl ether solutions (see pp 8–10 in the Supporting Information).

8. Preparation of BDQ Prodrugs 21–32 .

^a Reagents and conditions: (i) NMP, 25 °C, 72 h, 24–63% (see Table ).

9. Preparation of TBAJ-587 and TBAJ-876 Prodrugs 33–56 .

^a Reagents and conditions: (i) NMP, 25 °C, 72 h, 8–73% for TBAJ 587 2 HAPs and 9–53% for TBAJ-876 3 HAPs (see Tables and ).

2. Prodrugs 21–26 for BDQ 1a (High-R _F Diastereomer).

compound	R	yield (%)
21	H	53
22	Br	63
23	OMe	28
24	CN	25
25	CONH2	24
26	SMe	41

4. TBAJ-876 Prodrugs 45–56 .

effector	compound	R	yield (%)
TBAJ-876 high-R F 3a	45	H	44
	46	Br	27
	47	OMe	12
	48	CN	36
	49	CONH2	9
	50	SMe	47
TBAJ-876 low-R F 3b	51	H	45
	52	Br	47
	53	OMe	53
	54	CN	29
	55	CONH2	9
	56	SMe	28

3. TBAJ-587 Prodrugs 33–44 .

effector	compound	R	yield (%)
TBAJ-587 high-R F 2a	33	H	14
	34	Br	8
	35	OMe	30
	36	CN	44
	37	CONH2	12
	38	SMe	35
TBAJ-587 low-R F 2b	39	H	73
	40	Br	32
	41	OMe	33
	42	CN	22
	43	CONH2	15
	44	SMe	35

Unfortunately, prodrugs made from low-R _F effector diarylquinoline compounds (compounds 27–32, 39–44, and 51–56) exhibited significant solubility issues; notably, prodrugs of low-R _F effector BDQ 1b (compounds 27–32) displayed very low solubility, which prevented the acquisition of satisfactory NMR spectra of the prodrugs, hindering the determination of the true yield of the desired products.

Hence, as the most biologically desired/potent isomers are present in the high-R _F diastereomeric mixtures, − only prodrugs synthesized from the high-R _F diarylquinolines were carried forward for further analysis and biological testing.

Biology

Evaluation of in vitro Antimicrobial Activity

The antimicrobial evaluation of HAPs 21–26, 33–38, and 45–50 against a replicating strain of M.tb (M.tb H37Rv) was performed with the MABA using established protocols to measure the inhibition of actively growing M.tb under aerobic conditions following incubation for 18–24 h at 37 °C. The minimum inhibitory concentrations of the HAPs were quantified as MIC₉₀ and were measured by fluorimetry and compared to those of the untreated bacterial controls. The results showed moderate to good inhibitory values across all series (0.13–3.97 μg/mL, Table ), with higher MIC₉₀ values compared to those of the effector compounds and reference frontline compounds (rifampicin (RMP), linezolid (LIZ), pretomanid (PA), and the active isomer of BDQ 1).

5. Antimicrobial Activities of HAPs against M.tb H37Rv Determined Using MABA and LORA.

		MIC90 (μg/mL)		DR
compound	R	MABA	LORA	MABA	LORA	RDR (%)
1a	BDQ high-R F	0.07	0.06	–	–	–
21	H	0.46	0.30	6.57	5.00	24
22	Br	0.57	0.30	8.14	5.00	39
23	OMe	0.55	0.27	7.86	4.50	43
24	CN	1.12	0.43	16.00	7.17	55
25	CONH2	0.59	0.53	8.43	8.83	–5
26	SMe	0.51	0.25	7.29	4.17	43
2a	TBAJ-587 high-R F	0.07	0.064	–	–	–
33	H	0.31	0.09	4.43	1.41	68
34	Br	0.51	0.06	7.29	0.94	87
35	OMe	0.33	0.13	4.71	2.03	57
36	CN	0.42	0.11	6.00	1.72	71
37	CONH2	0.31	0.15	4.43	2.34	47
38	SMe	0.31	0.10	4.43	1.56	65
3a	TBAJ-876 high-R F	0.04	0.008	–	–	–
45	H	0.26	0.053	6.50	6.63	–2
46	Br	0.24	0.052	6.00	6.50	–8
47	OMe	0.14	0.032	3.50	4.00	–14
48	CN	0.14	0.055	3.50	6.88	–96
49	CONH2	0.13	0.018	3.25	2.25	30
50	SMe	0.04	0.036	1.00	4.50	–350
RMP	–	0.008	0.52	–	–	–
LIZ	–	0.21	0.67	–	–	–
PA	–	0.16	0.14	–	–	–
BDQ	–	0.05	0.028	–	–	–

^aThe MIC₉₀ values for the effector of each series are shown in bold.

^bDR is the deactivation ratio, given by DR = (MIC₉₀ of HAP)/(MIC₉₀ of DARQ effector compound).

^cRDR is the reduction in DR, given by RDR = [(DR_MABA – DR_LORA)/DR_MABA] × 100%.

These results suggest that the HAPs were able to undergo partial activation even under aerobic conditions in the MABA, resulting in the partial retention of antibacterial activity. However, compared to the parent effector compounds, they showed a significant loss of efficacy across all series. There are two possible mechanistic pathways that could be responsible for this effect. The first possibility arises when quaternization fails to mask the dimethylamine moiety adequately. Previous structure–activity relationship work on BDQ 1 and crystal structure studies revealed that modification of the dimethylamine moiety proved to be detrimental to activity, as this was a key site that allowed the other sections of the compound to form hydrophobic interactions with ATP synthase. Hence, it is likely that modification of the dimethylamine moiety to form the nitroimidazole quaternary salts would have significantly disrupted the binding to the ATP synthase, which likely would have a detrimental effect on the antituberculosis activity. The other probable cause is that alternative non-hypoxia-mediated activation mechanisms of HAPs occur under aerobic conditions. Several nitroreductases present in M.tb function under both aerobic and hypoxic conditions (examples from the literature include NfnB, deazaflavin-dependent nitroreductase (Ddn), and Rv2466c), which could activate the prodrugs to release the active effectors, contributing to the observed activities in the MABA.

To measure the potency of HAPs against NRP M.tb (M.tb H37Rv) under hypoxic conditions, the LORA was conducted using reported protocols that mimic the hypoxic environment of latent M.tb. The LORA reaches <0.16% oxygen concentration, mimicking the hypoxic granuloma environment, which has oxygen concentrations of ∼0.22%. Following incubation in anaerobic conditions for 10 days and a 28 h aerobic recovery period at 37 °C, MIC₉₀ was measured by luminescence signals and compared to that of the bacterial control that was untreated. HAPs showed potent MIC₉₀ values ranging from 0.018 to 3.92 μg/mL (Table ), even showing lower MICs than some frontline compounds. The HAPs possessed higher MIC₉₀ values than the effector compounds. HAPs 21–26 generally had MIC₉₀ values similar to or lower than those of RMP (except for HAP 24) and LIZ. HAPs 33–38 had lower MIC₉₀ values than RMP, LIZ, and PA (except HAP 37), and prodrugs 45–50 had lower MIC₉₀ values than RMP, LIZ, and PA and MIC₉₀ values comparable to that of the active isomer of BDQ 1. Prodrug 49 exhibited the lowest MIC₉₀ of 0.018 μg/mL on the LORA across all effector series and appeared to be 1.7–3 times more potent than the other compounds in the same series.

Notably, PA, which is included as a reference compound in Table , shows a MABA MIC₉₀ of 0.16 μg/mL and a LORA MIC₉₀ of 0.14 μg/mL. As a relevant nitroimidazole control, pretomanid is bioactivated by the nitroreductase Ddn and has two possible mechanisms of action: one that operates under aerobic conditions (for replicating M.tb) and another that occurs under hypoxic conditions in NRP M.tb. As a result, the combination of the two pathways could be responsible for the effective inhibition observed in both MABA and LORA, resulting in comparable potencies across the two assays.

Overall, the HAPs exhibited antimicrobial activity with moderate to good retention of inhibitory activity against M.tb H37Rv. All HAPs exhibited lower MIC₉₀ values in the LORA than in the MABA, demonstrating greater potency under hypoxic conditions and a potential for prodrug activation.

While it was encouraging to observe potent MICs for HAPs in in vitro anti-TB assays, one of the main objectives of this work was to selectively activate and release the active effector in hypoxic regions. To assess hypoxia selectivity, the deactivation ratios (DRs) of the prodrugs were compared to those of the parent effector compounds in both MABA and LORA (Table ). This allowed for the calculation of the reduction in activity as a ratio of HAP MIC₉₀ to the MIC₉₀ value of the corresponding effector compound.

The calculated deactivation ratios under aerobic conditions (MABA) for HAPs 33–38 based on TBAJ-587 high-R _F 2a (Figure ) show that these HAPs were not readily activated/cleaved, and the resulting MABA MIC₉₀ values of 0.31–0.51 μg/mL reflect the overall lower cellular activities (4.42–7.29 times poorer activity compared to TBAJ-587 high-R _F 2a). Conversely, under low oxygen conditions (LORA), MIC₉₀ values of 0.06–0.15 μg/mL for HAPs 33–38 were observed (0.93–2.34 times poorer activity compared to TBAJ-587 high-R _F 2a). Under hypoxic LORA conditions, HAPs produced MIC₉₀ values comparable to those of effector 2a, suggesting activation under the assay conditions.

2.

Deactivation ratios of prodrugs 33–38.

In general, lower reduction in activity was observed in the LORA across most series compared to the MABA, with the exception of the TBAJ-876 high-R _F 3a series (compounds 45–50) (Figure S7). Compounds 45–50 exhibited low MIC₉₀ values across both the MABA and LORA (<0.3 μg/mL and <0.1 μg/mL, respectively); however, they did not exhibit lower deactivation ratios in the LORA, with deactivation ratios similar to those observed in the MABA. Given the very potent MABA and LORA MIC₉₀ values observed for TBAJ-876 high-RF 3a (0.04 μg/mL and 0.008 μg/mL, respectively), even minor experimental variability could substantially influence the calculated prodrug deactivation ratios, and this warrants further investigation (discussed further in the Conclusions).

Next, the effect of varying the C-2 substituent on the nitroimidazole trigger on effector activation was explored by calculation of the reduction in deactivation ratio (RDR), which allowed for a comparison of HAP MABA and LORA activities across the different triggers (Table ).

For the lead series (compounds 33–38), nitroimidazole triggers with an electron-withdrawing group (34, R = Br and 36, R = CN) or neutral group (33, R = H) at C-2 showed the best reduction of deactivation ratios between MABA and LORA, at 87%, 71%, and 68%, respectively. The same trend was observed in prodrugs bearing electron-withdrawing groups in the other two series: in the BDQ high-R _F series, compound 24 (R = CN) demonstrated the highest reduction in deactivation ratio of 55%, and in the TBAJ-876 high-R _F series, compound 49 (R= CONH₂) demonstrated 31% reduction. These results indicate that nitroimidazole triggers containing electron-withdrawing or neutral substituents may raise the reduction potential, enabling more facile reduction and promoting highly selective effector release under hypoxic conditions.

Cytotoxicity Evaluation

The mammalian cell toxicity of the HAPs was evaluated against Vero green monkey epithelial kidney cells (American Type Culture Collection CCL-81) and a tetrazolium dye assay using previously reported protocols. All compounds had IC₅₀ > 14 μg/mL, demonstrating low cytotoxicities of the HAPs against normal mammalian cells that were comparable to Vero IC₅₀ values of the effectors (13.39 μg/mL to >20 μg/mL).

Physicochemical Property Investigation of HAP 33 via Pulse and Steady-State Radiolysis

To further characterize the activation mechanism of the prodrugs, pulse radiolysis experiments were performed using the 4 MeV linear accelerator. Pulse radiolysis studies were conducted under anoxic conditions in nitrous oxide (N₂O) saturated solutions which contained sodium formate (0.1 M) following previously reported protocols. Under these conditions, the low-potential carbon dioxide radical anion (CO₂ ^•–) is exclusively produced (0.64 μmol J^–1), which undergoes fast electron transfer to the substrate to produce its radical anion.

Compound 33, bearing a neutral trigger (R = H), was analyzed as a representative HAP compound with the intention of obtaining a baseline measurement of prodrug activation. The one-electron reduction potential (E°′) was determined to be −440 ± 7 mV by establishing redox equilibrium with methylviologen (−447 mV), as described in the literature. This value is consistent with the range of E°′ reported for ferredoxin redox systems of anaerobic microbes (−430 mV to −460 mV). Additionally, several nitroreductases in M.tb have been reported to have a range of reduction potentials; including Rv3131. While the exact reduction potential of Rv3131 has not been reported in the literature, it has been found to activate metronidazole (reduction potential of −480 mV), suggesting that it possesses a reduction potential within this range.

The spectrum of the radical anion was observed to decay with first-order kinetics over a range of radiation doses (i.e., independent of radical anion concentrations, Figure ). Extrapolation back to zero dose yielded a value of 245 ± 34 s^–1, which is assumed to be the rate constant for the elimination of the effector from the radical anion (k _elim).

3.

Rate constants of the loss of radical anion absorption at 330 nm for increasing radiation dose in N₂O-saturated solutions at pH 7 (5 mM phosphate) containing 33 (100 mM) and sodium formate (0.1 M).

Additionally, compound 33 was completely converted into its effector 2a following stepwise one-electron reduction, as confirmed by mass spectrometry analysis (Scheme ).

10. Conversion of the Radical Anion of 33 into Effector 2a and Imidazole Radical 57 .

Compound 33 was also irradiated in a Cs-137 γ irradiator (1.98 Gy min^–1), and changes in its UV–vis absorption spectrum were followed. Stepwise changes in the UV–vis spectrum with accumulated radiation dose are shown in Figure . The isosbestic points remained constant as the starting spectrum (compound 33) underwent conversion into the final product spectrum (compound 2a). The absorption spectra changed with accumulated radiation dose, decreasing in the region circled in blue and increasing in the region circled in black, with isosbestic points being retained throughout. Complete conversion of the spectrum of 33 to that of 2a required 79 Gy.

4.

UV–vis spectra of compound 33 (50 μM) showing stepwise changes in absorption spectrum with accumulated radiation dose under the same conditions as in Figure .

Together, these results demonstrate preliminary evidence that this class of HAPs undergoes hypoxia reduction to release the effector parent drug.

Kinetic Solubility Studies

The kinetic solubilities of representative HAP 33 and its parent effector 2a were determined in water and in 20% w/v hydroxypropyl cyclodextrin (HPCD) using established procedures (Table ; see the Supporting Information for experimental details). HAP 33 exhibited substantially higher aqueous solubility than effector 2a. In water, the kinetic solubility of HAP 33 was greater than 483.2 μg/mL, whereas effector 2a showed a solubility of only 31.34 μg/mL, representing at least a 15-fold improvement. A similar trend was observed in 20% HPCD solution, where HAP 33 remained highly soluble (>482.4 μg/mL) compared with effector 2a (88.5 μg/mL). These findings support the hypothesis that HAP analogues possess improved aqueous solubility relative to their parent effectors.

6. Kinetic Solubilities of Effector 2a and HAP 33 .

	kinetic solubility in H2O		kinetic solubility in 20% HPCD/H2O
compound	μM	μg/mL	μM	μg/mL
2a	51	31.34	144	88.5
33	>579	>483.2	>578	>482.4

Conclusions

Hypoxia is a constitutive feature of TB granulomas and is an emerging target for the treatment of LTBs. Hypoxia within granulomas is a result of poor extravascular diffusion of oxygen, and similar gradients for nutrients and therapeutic agents likely exist. This poses a challenge for drug design that may be surmounted by using a HAP approach. The HAP design should provide inert diffusible molecules that are selectively activated under hypoxia within the TB granuloma, leading to increased efficacy against LTB and an increased therapeutic index through reduced normal tissue toxicity.

In this work, such a prodrug strategy was applied to promising antitubercular drugs BDQ 1, TBAJ-587 2, and TBAJ-876 3 to optimize for LTB eradication. The strategy involved masking the parent drugs as 4-nitroimidazole prodrugs which could be selectively activated by fragmentation following one-electron reduction under hypoxic conditions. A series of 4-nitroimidazoles with varying electron densities were prepared and conjugated with the parent drugs to form 36 nitroimidazole HAPs (compounds 21–56). Inclusion of the 4-nitroimidazole moiety, in conjunction with the positive charge on the linker, results in reduction potentials within the range for one-electron enzymatic reduction and the subsequent release of the effectors. There is still a lack of understanding of the exact types and expression levels of reductases responsible for activating 4-nitroimidazole prodrugs in M.tb. Despite the extensive research efforts that are currently underway to identify and profile them, further insights and identification of such reductases would aid in elucidation of the possible mechanism of action and prodrug optimizations.

The antimicrobial biological evaluation of HAPs were carried out in two in vitro assays. The MABA focused on the replicating, oxygen-rich environments, while the LORA was conducted to evaluate/mimic the prodrug’s activity in the nonreplicating, hypoxic granuloma environments. The prodrugs showed good retention of inhibitory activity against M.tb H37Rv on both assays, with stronger potency and lower MIC₉₀ values observed in the LORA compared to the MABA. From the comparison of the deactivation ratios in the two assays, our lead set of prodrugs 33–38 (derived from the high-R _F TBAJ-587 2a effector) demonstrated the greatest selectivity for hypoxic reduction. The calculated deactivation ratios of HAPs 33–38 compared to effector 2a ranged from 4.4 to 7.3 μg/mL and 0.92–2.34 μg/mL in the MABA and LORA, respectively, indicating that these HAPs were not readily activated under aerobic conditions but appeared to cleave more efficiently under hypoxic conditions in vitro. For prodrugs 21–26 (derived from the high-R _F BDQ 1a effector), less pronounced hypoxia selectivity was observed compared to the lead series, but they still showed positive reduction in deactivation ratios. For prodrugs 45–50 (derived from the high-R _F TBAJ-876 3a effector), negative reductions in deactivation ratio were observed, meaning they performed worse in hypoxic conditions. This could be attributed to the discrepancy in the MIC₉₀ values obtained for the TBAJ-876 3a effector in this work, which came out to be 0.04 and 0.008 μg/mL in the MABA and LORA, respectively. In previous reported literature, MABA and LORA values of TBAJ-876 3a effector came to be 0.004 and 0.006 μg/mL, respectively, suggesting that the MABA MIC₉₀ value in this work is 10-fold less potent. For previously reported diarylquinoline analogues, MIC₉₀ values determined by the MABA and LORA were typically within a similar range. If the TBAJ-876 3a effector had shown comparable activity in both assays, this would have resulted in an improved hypoxia-selective reduction profile for this series.

It is important to note that the deactivation ratio considers only one of the many variables that could describe the selective activation of HAPs in hypoxic regions and may not generalize to in vivo models. The effect of hypoxia in TB granuloma and the subsequent response of HAPs in this environment, as well as on different types of granulomas, are excluded from this model; therefore, further in vivo studies in the C3HeB/FeJ mouse model or other similar models capable of simulating granulomatous hypoxia are required to investigate this further.

Compounds from the lead series (33–38) also had MIC₉₀ values ranging from 0.31 to 0.51 μg/mL on the MABA and 0.06–0.15 μg/mL on the LORA; the values for the HAPs obtained on the LORA were observed to be more potent compared to those of the antitubercular drugs rifampicin and linezolid (0.52 and 0.67 μg/mL, respectively). All HAP compounds demonstrated low cytotoxicity in Vero mammalian cell assays, with IC₅₀ > 14 μg/mL. These values were comparable to those observed for the effectors, whose Vero IC₅₀ values ranged from 13.39 μg/mL to >20 μg/mL. However, as BDQ 1 and TBAJ-587 2 have been reported to be potent inhibitors of human ATP synthase, further studies are warranted to investigate the binding affinity of HAPs for human ATP synthase and to assess their potential mitochondrial toxicity.

Collective insights via data obtained through biological evaluation and initial pulse radiolysis studies suggest increased selectivity and activation within M.tb under hypoxic conditions. While the results of the pulse radiolysis studies cannot be extrapolated to all HAPs, they provide compelling evidence that prodrug 33 undergoes one-electron reduction under hypoxic conditions. Moreover, the calculated reductions in activity and deactivation ratios of the prodrugs offer insights into a possible correlation between electron-withdrawing and neutral substituents on the nitroimidazole triggers and the improved hypoxic selectivity of the prodrugs, which may create a favorable electronic environment for optimal prodrug activation. Although the deactivation ratio comparison provides some evidence of differential activation, it does not directly demonstrate prodrug cleavage under the desired hypoxic conditions.

Direct analytical detection of released effector in both hypoxic and normoxic M.tb cultures and comparison of these results is warranted for future studies.

We hypothesized that the HAPs would exhibit favorable pharmacokinetic properties, such as reduced lipophilicity and improved solubility, which have been identified as key factors for TB granuloma penetration, as reported by the Dartois group. The cited work describes a model of passive diffusion through the lipid-rich matrix of TB granuloma lesions that was developed using neutral antitubercular compounds. However, the permanent positive charge introduced by the key Menshutkin quaternization in our series is expected to diminish passive diffusion across this lipid-rich environment. Consequently, the applicability of the cited model to our compounds may be limited, as it does not account for this charge-dependent effect. Furthermore, the nitroimidazole trigger adds at least one aromatic ring to the prodrugs, and the model predicts that caseum binding is proportional to the sum of the drug’s lipophilicity (clogP) and its total aromatic ring count. However, the prodrugs exhibit reduced lipophilicity (consensus log P _o/w = 5.33 for 2a and 2.90 for 33, predicted using SwissADME) and improved aqueous solubility relative to their parent effectors (31.3 μg/mL for 2a and >483.2 μg/mL for 33 in water; Table ). Within the framework of this model, this reduced lipophilicity and increased solubility would be predicted to enhance penetration into TB granuloma lesions, potentially offsetting, at least in part, the reduced passive diffusion associated with the permanent positive charge introduced by Menshutkin quaternization and increase in aromatic ring count.

These promising results support the validity of the hypoxia-targeted prodrug strategy developed in this work. For future studies, HAP metabolic stability studies, , along with further in-depth biological studies including intramacrophage testing via macrophage infection models, and in vivo C3HeB/FeJ mouse models that mimic the pathology of human TB granuloma formation, will be pursued to establish a more comprehensive pharmacokinetic profile of diarylquinoline HAPs.

Safety Statement

No unexpected or unusually high safety hazards were encountered.

Glossary

Abbreviations

AMR

antimicrobial resistance

BDQ

bedaquiline

DARQ

diarylquinoline

Ddn

deazaflavin-dependent nitroreductase

E. coli

Escherichia coli

FDA

U.S. Food and Drug Administration

hERG

human ether-a-go-go-related gene

HAP

hypoxia-activated prodrug

LTB

latent tuberculosis

LIZ

linezolid

LORA

low-oxygen recovery assay

MABA

microplate alamarBlue assay

MDR-TB

multidrug-resistant tuberculosis

M.tb

Mycobacterium tuberculosis

NRP

nonreplicating persistent

PA

pretomanid

RIF

rifampicin

S. aureus

Staphylococcus aureus

TB

tuberculosis

WHO

World Health Organization

XDR-TB

extensively drug-resistant tuberculosis

E°′

one-electron reduction potential

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.6c00238.

• Experimental details for the syntheses of the nitroimidazoles (compounds 4–9), diarylquinolines (compounds 1–3), and prodrugs (compounds 21–26 and 33–56); characterization data (¹H NMR, ¹³C NMR and HRMS) of these compounds; experimental optimization details for the key Menshutkin reaction for prodrug synthesis; full ¹H NMR characterization and assignment of a representative compound (prodrug 34); graphs showing the deactivation ratios of prodrugs 21–26 and 45–50; experimental details for kinetic solubility studies (PDF)

C.G.K.: chemical syntheses and characterization of prodrugs, manuscript preparation. M.P.H. and C.B.C.: data analysis, manuscript review. R.F.A.: pulse radiolysis studies, manuscript review. S.K.: kinetic solubility studies, HPLC analysis. S.G.F.: in vitro biological data, manuscript review. J.J. and P.J.C: conceptualization, synthetic route design, manuscript preparation. All of the authors read and agreed to the published version of the manuscript.

The authors declare no competing financial interest.