Menaquinone depletion resensitises bedaquiline-resistant tuberculosis

Abstract

Tuberculosis remains a leading cause of global mortality, and rising bedaquiline resistance threatens the effectiveness of current drug-resistant treatment regimens. Bedaquiline resistance typically arises through mutations in Rv0678 that upregulate drug efflux and confer cross-resistance to multiple drug classes. Here, we identify and optimise a chemical series targeting MenG, a central enzyme in the menaquinone biosynthesis pathway, yielding potent bactericidal inhibitors with in vivo efficacy. Strikingly, MenG inhibition restored bedaquiline susceptibility in efflux-mediated resistant strains, an effect confirmed in vivo where combination therapy achieved a 99.8% reduction in bacterial burden compared with bedaquiline alone. Potentiation also extended to pretomanid and other key agents. Disruption of upstream menaquinone and shikimate pathway enzymes produced similar resensitisation, establishing these pathways as tractable targets for restoring drug susceptibility in Mycobacterium tuberculosis. These findings provide a novel strategy to overcome bedaquiline resistance and strengthen future regimens for efflux-mediated drug-resistant tuberculosis.

Tuberculosis (TB) remains one of the leading causes of death from an infectious disease, responsible for an estimated 1.23 million deaths in 2024 (1), and global control efforts are increasingly threatened by the rise of drug-resistant TB (DR-TB). The introduction of bedaquiline marked a significant advance in TB therapy as the first drug with a novel mechanism of action in decades (2). Incorporating bedaquiline into DR-TB treatment regimens enabled significant treatment shortening, exemplified by the six-month all-oral BPaL regimen (bedaquiline, pretomanid and linezolid) (3, 4). These regimens have transformed DR-TB care, delivering the highest treatment success rates seen in decades, reaching 71% in recent reports (5). However, clinical resistance to bedaquiline is emerging (6), with rates exceeding 14% in some countries (7), threatening to erode the gains achieved in DR-TB management and highlighting the urgent need for strategies that preserve or restore its efficacy.

Bedaquiline inhibits the ATP synthase c-subunit (AtpE) of Mycobacterium tuberculosis, the causative agent of TB, impairing its energy metabolism and depleting intracellular ATP (8–10). Resistance arises predominantly through mutations disrupting Rv0678 (MmpR5), which encodes the repressor of the MmpS5–MmpL5 efflux pump, resulting in increased drug efflux (11). Bedaquiline’s long terminal half-life generates prolonged periods of subtherapeutic exposure, potentially favouring selection of such efflux-mediated mutants. Importantly, clinical Rv0678-mediated resistance cannot be completely overcome by dose escalation (12, 13) and confers cross-resistance to other anti-TB agents, including clofazimine and macozinone (14). Although the full clinical impact remains to be defined, Rv0678 variants are increasingly reported worldwide, including lineages predating the clinical introduction of bedaquiline (15).

Here, we sought to identify and optimise a compound series capable of preventing or overcoming bedaquiline resistance. This effort led to the discovery of two lead compounds, JNJ-6887 and JNJ-1866, the latter exhibiting promising efficacy in a mouse infection model. Mechanistic studies identified MenG, a key enzyme in the menaquinone biosynthesis pathway, as the molecular target. In M. tuberculosis, menaquinone-9 (MK-9) mediates electron transfer along the electron transport chain (ETC) and is essential for oxidative phosphorylation during both active and latent infection (16–18). Building on this discovery, we demonstrate that dual inhibition of MenG and ATP synthase enhances bactericidal activity and restores bedaquiline susceptibility in a clinically relevant efflux-mediated mutant. Importantly, this resensitisation was confirmed in vivo in a mouse model of infection. Potentiation also extended to other compounds, including pretomanid, suggesting a broader opportunity for combination strategies.

The upstream shikimate pathway provides the precursors required for menaquinone biosynthesis (19), functionally linking the two metabolic pathways. Inhibition of other enzymes in the menaquinone and shikimate pathways produced identical resensitisation effects in an efflux-based resistant strain. Similar to ethionamide boosters (20, 21), these findings reveal a novel strategy to restore drug susceptibility, but with the added advantage of intrinsic bactericidal activity and broad potentiation of other TB drugs. Together, these findings establish MenG and the wider shikimate-menaquinone superpathway as tractable resistance-breaking targets and demonstrate that metabolic co-targeting offers a powerful strategy to counter emerging efflux-mediated resistance and strengthen future DR-TB treatment regimens.

Results

Hit-to-lead optimisation of GSK517A

To identify new inhibitors with activity against DR-TB, we focused on GSK517A (hereafter JNJ-8833; Fig. 1a), a chemically tractable hit compound previously identified in a high-throughput screen (22) and selected for structure-activity relationship (SAR)-guided optimisation to improve its activity and chemical characteristics. Medicinal chemistry optimisation of JNJ-8833 produced a bactericidal analogue, JNJ-6887, with a 25-fold improvement in potency (Fig. 1a–b). Bactericidal activity was further confirmed by single-cell microscopy, which enables direct observation of individual bacterial fates and a clearer distinction between bacteriostatic and bactericidal effects (23, 24) (Fig. 1c; Fig. S1; Supplementary Video 1).

Fig. 1: Hit-to-lead optimisation identifies potent, metabolically stable analogues –

a, Compound optimisation from GSK517A (JNJ-8833) (22) leading to JNJ-6887 (~25-fold increased potency) and JNJ-1866, which shows improved metabolic stability. MIC₉₀ = minimum inhibitory concentration, 90% inhibition; Cl_int = mouse microsomal clearance (in μL min⁻¹ mg⁻¹; Table S1). b, Time-kill kinetics assay based on colony-forming units (CFU) with 0.13–3.33 μM (2–60x MIC₉₀) JNJ-6887 compared with DMSO control and 0.5 μM (5x MIC₉₀) bedaquiline (BDQ). n = 3 biological replicates. c, Single-cell imaging showing bactericidal activity of JNJ-6887. Representative image series of M. tuberculosis expressing TdTomato grown in a microfluidic device and exposed to 1 μM JNJ-6887 between 76 and 267 h. Scale bar, 5 μm. See also Fig. S1; Movie S1. d, In vivo efficacy of JNJ-1866 in mice treated for 12 days (twice daily oral administration; 150 mg kg⁻¹; white) compared with untreated (vehicle; black) at day 21 post-infection. Dotted line indicates CFU at treatment start. n = 6 mice. Significance was calculated with a two-sided unpaired t-test. For panels b and d, data shown are mean ± s.d.

However, JNJ-6887 showed poor metabolic stability in liver microsomes, predicting poor exposure and rapid clearance. Subsequent compound optimisation efforts therefore focused on improving metabolic stability while retaining bactericidal activity, leading to the synthesis of JNJ-1866, which exhibited an 8-fold reduction in potency compared with JNJ-6887 but showed markedly improved metabolic stability (Fig. 1a; Table S1–2). In vitro absorption, distribution, metabolism and excretion (ADME) and in vitro safety profiling, including mitochondrial toxicity, revealed an acceptable safety and physicochemical profile for JNJ-1866 (Table S3), supporting progression to in vivo pharmacokinetic (PK) evaluation. In mouse PK studies, JNJ-1866 demonstrated superior exposure and bioavailability compared with JNJ-6887 (Fig. S2; Table S4–5). On this basis, JNJ-1866 was advanced to efficacy testing in an acute mouse model of infection, where mice are infected for 7 days with M. tuberculosis H37Rv before treatment starts (Fig. S3). Oral administration of 150 mg kg⁻¹ JNJ-1866 for 12 days resulted in a significant 1.35-log reduction in lung bacterial burden relative to the end-of-study vehicle control, demonstrating in vivo proof-of-principle (Fig. 1d). These findings provided a strong foundation for further exploration of this chemical series.

Inhibition of menaquinone biosynthesis

To identify the molecular target of this compound series, we selected M. tuberculosis mutants resistant to JNJ-6887 and JNJ-1866. The isolated independent colonies conferred 26-133-fold resistance and were cross-resistant to other series analogues (Table S6), while retaining susceptibility to first- and second-line TB drugs and late-stage clinical candidates, indicating a possible novel mode of action or resistance (Table S6). JNJ-6887 and JNJ-1866 both retained activity across drug-sensitive and -resistant clinical isolates (25), comparable to their activity against the laboratory-adapted H37Rv (Table S7).

Whole genome sequencing revealed single-nucleotide polymorphisms in Rv0558, which encodes the membrane-associated methyltransferase MenG (also annotated as UbiE/MenH; Table S6). MenG is an S-adenosylmethionine (SAM)-dependent methyltransferase that converts demethylmenaquinone-9 (DMK-9) to MK-9, a reaction essential for M. tuberculosis growth (Fig. S4) (26–28). Additional resistance-selection experiments with series analogues identified further mutations. Most mutations resided within or adjacent to the predicted UbiE/COQ5 (Coenzyme Q5) domains but did not impair growth (Fig. S5). The role of these mutations in resistance was validated using a recombinant strain expressing an additional menG copy harbouring a C146R substitution, which conferred resistance to MenG inhibitors while retaining susceptibility to a reference compound (Fig. 2a; Fig. S6). Analysis of a large clinical isolate diversity database (29) revealed a limited presence of MenG-based resistance-conferring mutations; only one isolate (<0.002% of total) carried the D25G variant (Fig. S5), suggesting that such mutations are rare but tolerated in clinical populations.

Fig. 2: Target deconvolution identifies menaquinone biosynthesis enzyme, MenG –

a, Dose-response curves for JNJ-6887 against WT M. tuberculosis (black), M. tuberculosis carrying a chromosomally integrated, inducible MenG-C146R expression plasmid (WT::C146R) with (red) and without (white) 500 ng mL⁻¹ anhydrotetracycline (ATc) induction. n = 2 technical replicates. Representative of two independent experiments. b, Measurement of menaquinone levels in culture after 4 days in the presence (white) or absence (black) of 3.33 μM JNJ-6887, showing an increase in DMK-9 (MenG substrate) and a decrease in MK-9 (MenG product) and downstream MK-9(II-H₂). n = 5 technical replicates. c, Direct inhibition of membrane-isolated MenG protein with JNJ-6887 in a dose-dependent manner. IC₅₀ = 1.1 nM. Representative figure from three independent experiments. d, Dose-response curves of JNJ-6887 following CRISPRi-mediated “low” menG knockdown showing >20-fold increased inhibitor susceptibility upon induction. MenG knockdown was achieved in an ATc-dependent manner (0-100 ng mL⁻¹). n = 3 technical replicates. Representative of two independent experiments shown. e, Bactericidal activity of a combination of CRISPRi-mediated “low” MenG knockdown with 100 ng mL⁻¹ ATc and 3.33 μM JNJ-6887 over 28 days. Rifampicin (RIF; 14.58 μM) and isoniazid (INH; 5.8 μM) (grey) used as positive kill control. LoD: limit of detection. n = 3 biological replicates. For panels a, b, d, e, data shown are mean ± s.d.

In the absence of an M. tuberculosis MenG crystal structure, we generated an AlphaFold model and docked the SAM substrate based on the SAM-bound S. cerevisiae COQ5 structure (30–32). This revealed a putative ligand-binding pocket adjacent to resistance-conferring mutations (D25, N28, S32), with compounds likely obstructing DMK-9 binding (Fig. S7). Predicted interactions included hydrogen bonds between the hydrazine and R121, hydrophobic binding of the bicyclic ring with F17 and L185, and pi-pi stacking of the phenyl or isothiazole ring with F148 and W39. Reduced conformational flexibility of aromatic substituents (33) may explain the increased potency and activity shifts seen with JNJ-6887 and JNJ-1866 (Fig. S7; Table S6).

To evaluate target inhibition in a whole-cell context, we measured the abundance of menaquinone species after 4 days of JNJ-6887 exposure. Treatment resulted in an accumulation of the MenG substrate, DMK-9, and a corresponding decrease in MK-9 and β-dihydromenaquinone-9 (MK-9(II-H₂)), the principal form of menaquinone in mycobacteria (34), consistent with MenG inhibition (Fig. 2b; Fig. S4, 8). Humans lack the menaquinone biosynthetic pathway, but they do obtain MK-4 (vitamin K₂) from the diet for unrelated physiological functions. We therefore assessed whether exogenous MK-4 could restore bacterial growth. JNJ-6887 retained full bactericidal activity, even at MK-4 concentrations >50-fold above plasma levels (35), indicating limited potential for host rescue (Fig. S9).

Direct target engagement was confirmed by measuring radiolabelled methylation of DMK-8 to MK-8 in MenG-containing M. tuberculosis membrane preparations (36). We observed dose-dependent inhibition of enzymatic activity by JNJ-6887 (IC₅₀ = 1.3 ± 0.4 nM) and JNJ-1866 (IC₅₀ = 12 ± 2.8 nM), correlating with whole cell potency (Fig. 2c). Using a validated CRISPRi system (26), a menG “low” transcript knockdown sensitised M. tuberculosis >20-fold to JNJ-6887 in an ATc-dependent manner, further confirming on-target activity (Fig. 2d; Fig. S10). Finally, given mixed reports regarding the essentiality of the menaquinone pathway and its apparent low vulnerability in genome-wide essentiality screens (26, 37), we investigated whether sustained inhibition could achieve killing below the limit of detection. A time-course assay with 3.33 μM (60x MIC₉₀) JNJ-6887 resulted in bactericidal activity below the limit of detection (Fig. 2e). Notably, combining treatment with CRISPRi-mediated menG “low” knockdown did not markedly accelerate killing.

Resensitisation of bedaquiline resistance

Inhibitors of menaquinone biosynthesis have previously been shown to synergise with bedaquiline (38, 39). Checkerboard assays confirmed synergy between JNJ-6887 and bedaquiline (Fig. S11). At 1-2× MIC₉₀ concentrations, monotherapies produced minimal reductions in CFU counts, whereas dual combination resulted in a ≥3-log reduction, demonstrating synergistic bactericidal activity (Fig. S11).

Given this synergy in a wild-type (WT) background, we hypothesised it might extend to bedaquiline-resistant strains. The most common clinical mechanism of bedaquiline resistance involves disruption of Rv0678, leading to upregulation of the MmpS5-MmpL5 efflux pump. To model this, we generated a bedaquiline-resistant strain (BDQR^Rv0678) with complete disruption of Rv0678 and a concomitant increase in MmpS5-MmpL5 expression (Fig. S12). Kill kinetics were assessed using JNJ-6887 and bedaquiline, alone and in combination, in both WT and BDQR^Rv0678 backgrounds. In the WT strain, monotherapy with BDQ (5x MIC₉₀) and combination both reduced CFU by ~2-log after 21 days compared with the starting inoculum (Fig. 3a; Fig. S13). In this background, bedaquiline appeared to drive most of the killing. In contrast, bedaquiline monotherapy had no effect in the BDQR^Rv0678 background, consistent with the resistance phenotype, while JNJ-6887 retained equivalent activity in both strains, indicating that the resistance mechanism did not affect MenG inhibition. Remarkably, combination treatment in the resistant background resulted in a ≥3-log reduction in CFU counts after 21 days, indicating that MenG inhibition not only resensitised the strain to bedaquiline but also retained its synergistic capacity (Fig. 3a; Fig. S13). Equivalent results were also obtained with JNJ-1866 in the BDQR^R0678 background, and JNJ-6887 produced the same effect in an alternative Rv0678-mediated resistance strain carrying a distinct resistance-conferring mutation (Fig. S14–15). In contrast, no resensitisation was observed with linezolid under the same experimental conditions (Fig. S13).

Fig. 3: Resensitisation of bedaquiline resistance via MenG inhibition –

a, Day-21 from a time-kill kinetics experiment of 0.5 μM BDQ (5x MIC₉₀ activity in the WT background) and 0.2 μM JNJ-6887 (4x MIC₉₀), alone and in combination, in WT and BDQR^Rv0678 backgrounds. n = 3 technical replicates. Representative of two independent experiments shown. b, Dose-response curves of bedaquiline with the CRISPRi-mediated menG “low” knockdown strain (−ATc; MIC₅₀ = 89.7 nM) compared with induction (+ATc; MIC₅₀ = 23.6 nM). n = 3 technical replicates. Representative of three independent experiments shown. c, Brief outline of the ETC in M. tuberculosis. Complex II: succinate dehydrogenase; Complex III: Cytochrome bc₁; Complex IV; Cytochrome bd oxidase; Complex V: ATP synthase. Expanded model of the ETC in Fig. S17. d, Oxygen consumption rate (OCR) assay in samples pre-treated with 660 nM JNJ-6887 for 3 days followed by the addition of 32 μM bedaquiline then 4 μM CCCP (carbonylcyanide m-chlorophenyl hydrazone). n = 4 biological replicates. For panels b-c, d, data shown are mean ± s.d.

To confirm that resensitisation was not due to an off-target effect of the compounds, we used genetic knockdown of menG as an orthogonal approach. A CRISPRi-mediated menG “low” transcriptional knockdown, selected to avoid lethality during the assay, led to a 4-fold increase in bedaquiline susceptibility in the BDQR^Rv0678 background, consistent with the effect observed using chemical inhibition (Fig. 3b; Fig. S10). Finally, to rule out the activation of known bedaquiline-resistance mechanisms, transcript profiling of the BDQR^Rv0678 strain treated with JNJ-6887 revealed no differential expression of genes associated with resistance (Fig. S16). This supports the conclusion that MenG inhibition is specifically responsible for restoring bedaquiline activity in resistant strains.

MK-9 shuttles electrons along the ETC; therefore, MenG inhibition is expected to disrupt electron flow and reduce proton motive force (PMF), the electrochemical proton gradient that fuels ATP production (Fig. 3c; Fig. S17). Indeed, previous studies has demonstrated that MenG inhibition reduces oxygen consumption and impairs ATP synthesis (38). Efflux systems of the RND family, including MmpS5-MmpL5, are thought to be PMF-dependent (40). We therefore hypothesised that reduced PMF would impair efflux pump activity and, together with ETC disruption, account for synergy with ATP synthase inhibitors (complex V; bedaquiline) and the observed resensitisation phenotype. Although oxygen consumption rate (OCR) does not directly measure PMF, it does reflect changes in electron transport that influence proton pumping. Following 3-day exposure to 10x MIC₅₀ JNJ-6887, basal OCR was reduced by ~50% relative to untreated control (Fig. 3d). Addition of bedaquiline, previously shown to increase respiration rate, and CCCP, a classic uncoupler, highlighted the impairment of electron transfer through respiratory complexes caused by MenG inhibition. Short-term exposure to JNJ-6887 did not measurably alter membrane potential (Fig. S18), used here as a proxy for PMF-dependent efflux, suggesting that resensitisation is not explained by acute depolarisation. This is consistent with the delayed onset of activity observed with MenG inhibition (Fig. 2e). As a complementary approach, we tested whether combining bedaquiline with JNJ-6887 could restore the activity of JNJ-6887 in a MenG-resistant strain (Fig. S19). This combination produced a 5-log improvement in activity relative to either monotherapy, demonstrating that reciprocal resensitisation occurs for this drug combination. Finally, to investigate whether alternative respiratory pathways compensate for MenG inhibition, we assessed the activity of JNJ-6887 in a cytochrome bd (complex IV) knockout (bdKO) strain. Loss of cytochrome bd did not enhance sensitivity to JNJ-6887, suggesting that MenG inhibition is not compensated by this respiratory branch (Fig. S20). Taken together, these results demonstrate that MenG inhibitors resensitise M. tuberculosis BDQR^Rv0678 to bedaquiline through combined disruption of bioenergetic pathways and impairment of the PMF.

Enhancement of other TB therapies and in vivo bedaquiline resensitisation

We next assessed whether MenG inhibition could enhance the activity of other TB therapies and overcome resistance beyond bedaquiline. Macozinone and pretomanid both inhibit the formation of decaprenylphosphoryl-D-arabinose (DPA), an essential precursor for the mycobacterial cell wall, by disrupting DprE1 and DprE2, respectively (41, 42). DprE1 functions as a menaquinone-dependent dehydrogenase that facilitates FAD reoxidation (43), leading us to hypothesise that MenG inhibition would enhance the activity of both compounds (Fig. 4a). Indeed, time-kill kinetics showed that MenG inhibition enhanced the bactericidal activity of both drugs (Fig. 4b; Fig. S21). Despite its distinct mode of action, macozinone shares the same efflux-mediated resistance mechanism as bedaquiline (14); JNJ-6887 fully restored macozinone susceptibility in the BDQR^Rv0678 strain.

Fig. 4: MenG inhibition enhances key TB drugs and restores resistance in vivo –

a, The decaprenylphosphoryl-D-arabinose (DPA) pathway and its interactions with macozinone, pretomanid and JNJ-6887. Expanded explanation in Fig. S21. Decaprenylphosphoryl-D-ribose (DPR); decaprenylphosphoryl-2-keto-erythro-pentofuranose, DPX; flavin adenine dinucleotide, FAD; nicotinamide adenine dinucleotide phosphate, NADPH; menaquinol-9, MK-9H₂. b, Day 21 from time-kill kinetics assays of 1.5 nM macozinone (M) or 500 nM pretomanid (P) with 0.2 μM JNJ-6887 in BDQR^Rv0678. The dotted line indicates the starting inoculum. n = 3 technical replicates. 5 Representative of two independent experiments shown. Full dataset shown in Fig. S21. c, Day 21 from a time-kill kinetics assay of combination treatment. Bedaquiline (0.5 μM; B), pretomanid (7 μM; Pa) and JNJ-6887 (0.2 μM; 6887). The dotted line indicates the starting inoculum. n = 3 technical replicates. Representative of two independent experiments shown. Full dataset shown in Fig. S23. d, In vivo efficacy of MenG “low” CRISPRi in WT (left) and BDQR^Rv0678 (right) backgrounds, after 28-day treatment with induction with doxycycline included in the chow, with 6.25 mg kg⁻¹ (+) or 25 mg kg⁻¹ (++) bedaquiline (BDQ; once daily oral administration, 5/7 days). n = 5 mice. ATc/doxycycline positive conditions were pre-induced for 7 days prior to infection. Significance was calculated with one-way analysis of variance (ANOVA) with Šídák’s multiple comparisons. Full time-course is shown in Fig. S24. For panels b-d, data shown are mean ± s.d.

Given that JNJ-6887 enhanced the activity of bedaquiline and pretomanid, both backbone drugs for DR-TB treatment, we next investigated whether JNJ-6887 could restore bedaquiline activity in the BDQR^Rv0678 background using human exposure-equivalent concentrations of bedaquiline and pretomanid. In contrast to WT, bedaquiline-pretomanid (BPa) was inactive in the BDQR^Rv0678 strain, whereas the addition of only 4x MIC₉₀ JNJ-6887, which had no activity as a monotherapy, fully restored BPa activity (Fig. 4c; Fig. S22). These findings support the inclusion of a MenG-targeting inhibitor in future DR-TB combination therapies without the need to determine efflux-based resistance, aligning with drug-susceptibility testing-independent treatment strategies envisioned by the PAN-TB framework (44).

To assess whether the resensitisation phenotype extended to more physiologically relevant conditions, we used a 3-day ex vivo THP-1 macrophage infection model with the BDQR^Rv0678 strain, which similarly demonstrated robust resensitisation to bedaquiline in the presence of JNJ-6887 (Fig. S23). Encouraged by these findings, we next investigated whether this effect translated in vivo. Mice were infected with either WT or BDQR^Rv0678 strains for 7 days and treated with a subcutaneous long-acting injectable (LAI) formulation of JNJ-1866, a strategy previously shown to enhance in vivo compound exposure levels (45) (Fig. S2; Table S6). In a WT background, 1,250 mg kg⁻¹ JNJ-1866 LAI clearly enhanced bedaquiline activity; however, in the BDQR^Rv0678 background, combination treatment showed only a non-significant trend towards resensitisation (Fig. S3, ²⁴). As an alternative approach, we used the menG “low” knockdown strains in mice (Fig. S3). In these models, we confirmed that bedaquiline had reduced activity in the BDQR^Rv0678 background compared with WT, while menG knockdown alone had minimal impact in either strain. Combining menG knockdown with 6.25 mg kg⁻¹ bedaquiline in the WT background enhanced activity of the combination, matching the synergy observed in vitro. Importantly, combination with 25 mg kg⁻¹ bedaquiline in the BDQR^Rv0678 background resulted in a 3-log reduction in bacterial load (Fig. 4d; Fig. S24), restoring levels comparable to WT activity. These findings provide clear in vivo proof-of-concept that MenG inhibition can resensitise resistant strains to bedaquiline. Finally, to complement this work, we used Mycobacterium marinum, one of the closest genetic relatives of M. tuberculosis and a pathogen that enables in vivo efficacy testing in its natural host, the zebrafish. Treatment with JNJ-1866 demonstrated clear in vivo proof-of-concept after 2 days. We subsequently generated an efflux-mediated (MMAR-1007; Rv0678-equivalent) bedaquiline-resistant M. marinum strain and confirmed that it conferred in vivo-relevant resistance to bedaquiline. Importantly, inhibition of MenG resensitised this strain to bedaquiline, mirroring the phenotype observed in M. tuberculosis (Fig. S25).

Pathway inhibition restores bedaquiline susceptibility

The shikimate biosynthetic pathway provides the chorismate precursor essential for the MK-9 biosynthesis, linking aromatic amino acid biosynthesis to respiratory energy generation (Fig. 5a). To determine whether resensitisation of bedaquiline activity was a general consequence of menaquinone biosynthesis disruption, we performed matched time-kill kinetics using CRISPRi-mediated “low” knockdown of menG, menE, aroG and aroK, targeting multiple steps across the menaquinone and upstream shikimate biosynthetic pathways. Experiments were carried out in both WT and BDQR^Rv0678 backgrounds (Fig. 5b–e; Fig. S10, ²⁶). Monotherapy with 0.5 μM bedaquiline had no impact on CFU in the BDQR^Rv0678 background. Similarly, individual knockdown of each gene produced only a modest reduction in CFU, consistent with a “low” knockdown phenotype. However, combining bedaquiline exposure with “low” knockdown led to a marked and reproducible reduction in CFU (Fig. 5b–e), confirming a resensitisation phenotype comparable to that observed with JNJ-6887 and suggesting that disruption of any gene in the two pathways can resensitise bedaquiline resistance.

Fig. 5: Drug resistance rescued by entire menaquinone biosynthesis pathway –

a, Schematic of the shikimate and menaquinone biosynthetic pathways in M. tuberculosis. b-e, Day 14 CFU counts from time-kill assays using CRISPRi-mediated “low” knockdown of (b) aroG, (c) aroK, (d) menE or (e) menG with 0.5 μM BDQ (10× MIC₉₀) in BDQR^Rv0678 background. Strains were pre-induced with 100 ng mL⁻¹ ATc for 5 days (aroG and aroK) or 7 days (menE and menG). DMSO (black); rifampicin (RIF; 14.58 μM) and isoniazid (INH; 5.8 μM) as positive kill control (dark grey). Representative of two independent experiments with three biological replicates. Time-kill kinetics data are shown in Fig. S26. f, LC-MS analysis showing MK-9(II-H₂) reduction in the aroK “high” CRISPRi strain after induction with ATc for 5 days. n = 4 biological replicates. Representative of two independent experiments. Significance was calculated with two-sided (Bonferroni–Dunn) Student’s t-test with Welch correction. g, Dose-response curves of pretomanid following aroK “high” transcript reduction in an ATc-dependent manner, showing 13-fold increase in inhibitor potency after 14 days. Representative of three independent experiments. Inset: isobologram using fractional inhibitory concentrations (FIC) of pretomanid (PMD) and aroK induction (ATc concentration). n = 3 independent experiments. h-i, In vivo efficacy of CRISPRi-mediated aroK “high” (h) and menE “low” (i) knockdown in a WT background, induced with doxycycline in chow, and in combination with 6.25 mg kg⁻¹ bedaquiline (BDQ; once daily oral administration, 5/7 days). ATc/doxycycline positive menE conditions were pre-induced for 7 days prior to infection. n = 5 mice. Significance was calculated with one-way analysis of variance (ANOVA) with Šídák’s multiple comparisons. Full time-course is shown in Fig. S28. For panels b-f, h-i, the data shown are mean ± s.d.

In the absence of validated inhibitors, CRISPRi-mediated “high” knockdowns of aroG and aroK were generated to facilitate combination studies and provide further genetic validation that these enzymes represent viable drug targets (Fig. S10). Silencing of aroG and aroK produced a robust bactericidal effect, reaching the limit of detection after 14 days, in agreement with published vulnerability scores for these enzymes (26) (Fig. S27). LC-MS analysis of the aroK “high” knockdown strain revealed increased shikimate, the substrate of AroK, and reduced MK-9(II-H₂) levels, consistent with its role upstream of the menaquinone biosynthesis pathway and supporting the hypothesis that the resensitisation phenotype is specifically due to inhibition of menaquinone biosynthesis (Fig. 5f; Fig. S27). We next used our aroK “high” knockdown strain to investigate potential enhancement of pretomanid activity (Fig. 5g). This resulted in a robust, ATc concentration-dependent 13-fold increase in susceptibility to pretomanid; conversion of these data into an isobologram demonstrated that pretomanid and ATc-dependent aroK knockdown act synergistically (Fig. 5g). Finally, in an acute mouse model, combining aroG “high”, aroK “high” or menE “low” knockdown with 6.25 mg kg⁻¹ bedaquiline in the WT background significantly enhanced bedaquiline activity compared with either monotherapy or knockdown alone after 21 days (Fig. 5h–i; Fig. S28), replicating the in vivo resensitisation phenotype seen with menG inhibition. These findings indicate that disruption of any component of menaquinone biosynthesis is sufficient to both enhance the activity of other compounds and overcome bedaquiline resistance, highlighting enzymes from both pathways as potential “resistance-breaking” drug targets that should be prioritised for future screening efforts.

Discussion

The emergence of bedaquiline resistance represents a growing threat to DR-TB treatment. Current global prevalence of bedaquiline resistance remains uncertain due to the absence of robust prospective surveillance data, but it appears to be increasing in some regions, with estimates ranging from 0.2% in China in 2022 (46) to 14% in Mozambique in 2021 (7). The MmpS5-MmpL5 efflux pump-mediated resistance mechanism also confers cross-resistance to clofazimine, benzothiazinones (including macozinone), Q203 and next-generation ATP synthase inhibitors (TBAJ-587 and TBAJ-876) (14). Clinically, baseline Rv0678 mutations in bedaquiline-naïve patients enrolled in early registrational trials did not affect initial culture conversion (12, 47, 48). However, lower conversion rates were observed among patients who acquired Rv0678 mutations compared with those who did not (12). Patients with bedaquiline-resistant TB also have poorer outcomes overall, with only 57% achieving treatment success compared to 72% among those with susceptible infections (49). The risk of poor outcomes is further increased when Rv0678 variants occur in patients previously exposed to bedaquiline or clofazimine. This risk is highest in pre–extensively drug-resistant TB (pre-XDR-TB) and extensively drug-resistant TB (XDR-TB) populations, particularly in the presence of fluoroquinolone resistance and when treatment regimens lack additional active drugs (50, 51). Currently, alternative regimens used when bedaquiline cannot be given are longer and not standardised, supported by weaker evidence of efficacy and associated with greater toxicity (52). As bedaquiline remains the cornerstone of DR-TB treatment, strategies that both limit the spread of resistance and overcome established resistance are urgently required.

Combination therapy remains fundamental to TB control. Drugs with distinct modes of action can interact synergistically or antagonistically, producing effects greater or less than expected from the addition of monotherapies. Synergistic combinations could be particularly valuable clinically, as they may accelerate bacterial clearance, shorten treatment duration, reduce the risk of resistance and lessen adverse effects. Here, we show that MenG inhibition enhances the activity of bedaquiline, pretomanid and macozinone. This is consistent with earlier reports (38), demonstrating that MenG inhibition also synergises with the first-line drugs rifampicin and isoniazid. Previous work has likewise shown that MenA inhibition enhances the activity of clofazimine (39).

Crucially, the synergy observed with MenG inhibition resensitised strains with clinically relevant efflux-mediated bedaquiline resistance. Previous work with non-clinical compounds has shown that combinations containing chlorpromazine, an ETC inhibitor, can restore the efficacy of spectinomycin, a protein synthesis inhibitor, against M. tuberculosis strains carrying spectinomycin-resistance mutations, highlighting the potential of potentiating combinations to bypass resistance (53). Furthermore, although efflux pump inhibitors such as verapamil can also restore bedaquiline activity in vitro (Fig. S29), they lack bactericidal activity as monotherapies and do not restore bedaquiline efficacy in vivo (12, 54). Inhibiting menaquinone biosynthesis therefore provides a more effective strategy to reverse resistance.

Ongoing work exploring non-bactericidal “booster” molecules that enhance ethionamide activity further highlights the potential of agents that strengthen existing regimens (20, 21). Importantly, MenG inhibition is intrinsically bactericidal, offering a key advantage over these booster compounds. As a result, menaquinone biosynthesis inhibitors could contribute directly to bacterial clearance while simultaneously enhancing the activity of partner drugs. This dual action raises the possibility of meaningful treatment shortening. Our current compound series lacks high-level efficacy in a mouse model of infection; however, this broader principle is supported by work on cytochrome bc1 inhibitors, which exhibit limited or static bactericidal activity but show strong combinatorial effects with multiple drug classes (55).

The representative molecule from the current series, JNJ-1866, has served as a valuable tool compound for characterising the mode of action and assessing the relevance of MenG. However, further optimisation is required to improve drug-like properties and achieve a druggable candidate. Addressing its metabolic instability will be essential for defining the PK/PD relationship in animal models and for improving the translatability of these efficacy models to human settings.

Both the menaquinone and shikimate biosynthetic pathways are essential in mycobacteria (26, 56) but absent in humans, making them attractive therapeutic targets. M. tuberculosis relies on MK-9 as the principal electron carrier linking primary dehydrogenases to terminal oxidoreductases in the ETC. We demonstrated that MenG inhibition depletes menaquinone levels, disrupts electron transport and, when combined with drugs exported via the MmpL-MmpS5 efflux system, restores susceptibility to near WT levels. As both ATP synthase and menaquinone-dependent respiration are essential, dual inhibition may also enhance killing of persistent subpopulations, with potential to shorten treatment duration and reduce relapse risk.

Given the rising burden of DR-TB, prolonged treatment duration and increasing resistance rates, new therapeutic strategies are urgently needed. Our study provides clear in vivo proof-of-principle that inhibiting menaquinone biosynthesis can restore the efficacy of bedaquiline against efflux-mediated resistance. Since depletion of MK-9 appears to underlie this resensitisation, targeting any essential enzyme within the shikimate or menaquinone pathways, such as MenA, which has already been validated as druggable (57), may yield similar outcomes. Together, these findings confirm these pathways as critical targets in M. tuberculosis and, for the first time, demonstrate that their inhibition can reverse efflux-mediated bedaquiline resistance, potentially extending the clinical lifespan of key TB drugs and informing the rational design of next-generation combination regimens aimed at shortening therapy and preventing the spread of resistance.

Data, code, and materials availability:

All data generated or analysed during this study are included in this published article (and its supplementary information files).