Functional drug screening reveals anticonvulsants as enhancers of mTOR-independent autophagic killing of Mycobacterium tuberculosis through inositol depletion

Abstract

Mycobacterium tuberculosis (MTB) remains a major challenge to global health made worse by the spread of multidrug resistance. We therefore examined whether stimulating intracellular killing of mycobacteria through pharmacological enhancement of macroautophagy might provide a novel therapeutic strategy. Despite the resistance of MTB to killing by basal autophagy, cell-based screening of FDA-approved drugs revealed two anticonvulsants, carbamazepine and valproic acid, that were able to stimulate autophagic killing of intracellular M. tuberculosis within primary human macrophages at concentrations achievable in humans. Using a zebrafish model, we show that carbamazepine can stimulate autophagy in vivo and enhance clearance of M. marinum, while in mice infected with a highly virulent multidrug-resistant MTB strain, carbamazepine treatment reduced bacterial burden, improved lung pathology and stimulated adaptive immunity. We show that carbamazepine induces antimicrobial autophagy through a novel, evolutionarily conserved, mTOR-independent pathway controlled by cellular depletion of myo-inositol. While strain-specific differences in susceptibility to in vivo carbamazepine treatment may exist, autophagy enhancement by repurposed drugs provides an easily implementable potential therapy for the treatment of multidrug-resistant mycobacterial infection.

Keywords: autophagy; multidrug‐resistant; myo‐inositol; tuberculosis.

Figure 1

Anticonvulsants stimulate killing of mycobacteria in vitro

1. Screening drugs for effects on survival of intracellular mycobacteria in macrophages. RAW 264.7 cells were infected with a luminescent strain of M. bovis BCG (BCG-lux) for 1 h, washed and treated for 24 h with vehicle alone (white), known autophagy enhancers (interferon-γ (IFNγ), 200 ng/ml and rapamycin, 200 nM; black), known mTOR-independent autophagy inhibitor (forskolin; 24 μM; blue), carbamazepine (CBZ, 50 μM; red), valproic acid (VPA, 3 mM; green) and other examples of hits from a large screen of compounds enhancing intracellular killing of mycobacteria (lithium, 10 mM; rilmenidine, 1 μM; clonidine, 1 μM; calpeptin, 50 μM; grey). P-values, unpaired Student's t-test (n ≥ 6) (compared to vehicle alone): IFNγ 0.03; rapamycin 0.003, CBZ 0.001, VPA 0.001, lithium 0.001, rilmenidine 0.001; clonidine 0.02; calpeptin 0.01. Inset: Correlation between measurements of colony-forming units (CFU) and luminescence (RLU) for cultures of M. bovis BCG-lux as previously described (Kampmann et al, 2000).

2. Effects on intracellular survival of M. bovis BCG-lux of treatment with varying concentrations of CBZ (red) or VPA (green). P-values, unpaired Student's t-test (n = 6) (compared to vehicle alone) for CBZ 30 μM: 9 × 10⁻⁷, 40 μM: 1.9 × 10⁻⁹, 50 μM: 1.3 × 10⁻⁹, 100 μM: 1.6 × 10⁻⁸, 200 μM: 2.7 × 10⁻⁹, VPA 0.5 mM: 6.9 × 10⁻⁸, 1 mM: 1.6 × 10⁻⁹, 3 mM: 4.3 × 10⁻¹⁰, 4 mM: 4.2 × 10⁻⁹, 10 mM: 1.1 × 10⁻⁹, 20 mM: 4.2 × 10⁻¹⁰. These compounds had no effect on cell-free mycobacterial viability (grey circles).

3. Anticonvulsants enhance intracellular killing of mycobacteria within human alveolar macrophages. Alveolar macrophages, obtained from the broncho-alveolar lavage fluid of three individuals, were infected with M. bovis BCG-lux and then treated with carbamazepine (CBZ, 50 μM; red), valproic acid (VPA, 3 mM; green) or vehicle alone (Con; black). Viable intracellular mycobacteria were determined after 24 h of treatment. Unpaired Student's t-test (n = 3) (compared to vehicle alone): CBZ 0.011; VPA 0.0009.

4. Enhanced killing of intracellular M. tuberculosis (H37Rv) within primary human macrophages by treatment with CBZ (50 μM; red), VPA (3 mM; green) and rapamycin (200 nM; black), compared to control (vehicle alone; white). Unpaired Student's t-test (n = 6) (compared to vehicle alone): rapamycin 48 h 0.00002, rapamycin 72 h 0.00005, CBZ 48 h 0.00021, CBZ 72 h 0.00022, VPA 48 h 0.0002, VPA 72 h 0.00049.

5. In vivo induction of autophagy by CBZ (50 μM; red) and rapamycin (RAP; 1 μM; blue) compared to vehicle control (white) monitored in zebrafish expressing fluorescent ATG8 co-treated with chloroquine to delay degradation of autophagosomes.

6. Wild-type zebrafish were injected with a red fluorescently tagged M. marinum strain M into the yolk sac circulation valley at 28 hpf. Larvae were imaged at 120 hpf by confocal microscopy and the total mycobacteria-associated fluorescence quantified using Volocity® software. Data expressed as mean ± SEM (n ≥ 13 fish performed as 3 independent experiments). P-values, unpaired Student's t-test (compared to vehicle alone): 0.0035.

7. Mice infected via aerosol with a highly virulent clinical strain of multidrug-resistant M. tuberculosis CSU 87 were treated from day 20 post-infection with carbamazepine (CBZ, 50 μg/kg i.p. daily), rifampicin/isoniazid (RIF/INH) or vehicle control (n = 5 per time point per group). CBZ treatment for 30 days resulted in (G) significantly less viable bacteria detected in lung and spleen, (H) reduced inflammatory pulmonary infiltrates compared to RIF/INH-treated or control animals and (I) decreased lung lesion scores. Unpaired Student's t-test (n = 5) (compared to vehicle alone): RIF/INH 0.007; CBZ 0.037.

Data information: *P < 0.05; **P < 0.005.

Figure 2

Carbamazepine promotes innate and adaptive immunity during mycobacterial infection Primary human macrophages infected with M. bovis BCG were treated with carbamazepine (CBZ, 50 μM) or rapamycin (Rap; 200 nM). IL-8 (black) and TNFα (grey) were measured in supernatants collected 24 h after infection.

1. Levels of IL-8 (black) and TNFα (grey) released by M. bovis BCG-infected primary human macrophages were autophagy dependent, measured following siRNA knock-down of the critical autophagy protein, ATG12, or control siRNA (control). P-values, unpaired Student's t-test (n = 3) ATG12 siRNA vs siControl: TNFα 0.0006; IL-8 0.0012.

2. Carbamazepine increases pro-inflammatory cytokine secretion by mycobacteria-infected human macrophages. P-values, unpaired Student's t-test (n = 5) (compared to vehicle alone): CBZ TNFα 0.00048, IL-8 0.0017; Rap TNFα 0.001, IL-8 0.0089.

3. Analysis (by flow cytometry) of intracellular cytokines (IL-27, IL-12, TNFα) and MHC class II surface expression for dendritic cells (CD11c⁺) and macrophages (CD11b⁺) from lung, spleen and draining lymph nodes of mice that were uninfected and untreated (grey), uninfected and CBZ-treated (orange), infected with multidrug-resistant M. tuberculosis (MDR-TB) and untreated (black), MDR-TB infected and treated with rifampicin and isoniazid (RIF/INH; blue) or MDR-TB infected and treated with carbamazepine (CBZ, red).

Data Information: * P < 0.05; ** P < 0.005. All experiments were carried out at least in triplicate and on at least 2 separate occasions.

Figure 3

Anticonvulsants stimulate autophagosome formation in vitro

1. The anticonvulsants carbamazepine and valproic acid enhance autophagic clearance of cytosolic substrate. We monitored the clearance of the known autophagy substrate (A53T) α-synuclein in stable inducible PC12 cells by Western blot. The A53T α-synuclein transgene was induced with doxycycline for 48 h and then switched off (by antibiotic removal) before cells were treated with carbamazepine (CBZ, 50 μM; red), valproic acid (VPA, 3 mM; green), rapamycin (Rap, 200 nM; black), forskolin (FSK, 10 μM; blue) or vehicle alone (DMSO; white) for a further 24 h. P-values, paired Student's t-test (n = 3) (compared to vehicle alone): Rap 0.0011; CBZ 0.0017; VPA 0.02; FSK 0.005.

2. Anticonvulsants enhance autophagosome synthesis in primary human macrophages. In the presence of a saturating concentration of bafilomycin A₁ (400 nM) which blocks autophagosome–lysosome fusion, levels of LC3-II reflect autophagosome production and were quantified in human macrophages (from a different healthy individual in 3 separate experiments) treated with rapamycin (Rap, 200 nM; black), carbamazepine (CBZ, 50 μM; red), valproic acid (VPA, 3 mM; green) or vehicle alone (DMSO; white). P-values, paired Student's t-test (n = 3) (compared to vehicle alone): CBZ 0.026.

3. Increases in LC3-II levels (assessed by Western blot) in primary human macrophages infected with M. bovis BCG after 4 h treatment with rapamycin (Rap; 200 nM), valproic acid (VPA, 3 mM) and carbamazepine (CBZ, 50 μM) compared to controls. P-values, paired Student's t-test (n = 3) (compared to vehicle alone): Rap 0.02; VPA 0.004; CBZ 0.003.

4. Treatment of M. bovis BCG-infected human macrophages with CBZ (50 μM) or VPA (3 mM) increased autophagosome number, assessed by confocal microscopy of cells stained with LC3-specific antibody (green) with quantification of the number of autophagosomes (defined as LC3 +  vesicles ≥ 1 μm diameter) per cell shown below. P-values, unpaired Student's t-test (n = 3) (compared to vehicle alone): CBZ 0.001, VPA 0.003. Scale bar represents 10 μm.

5. In contrast to rapamycin (Rap), carbamazepine (CBZ) treatment of macrophages, while increasing LC3-II levels, does not alter mTOR-dependent signalling (monitored by changes in phosphorylation of S6 and p70S6 kinase).

6. Enhanced intracellular killing of mycobacteria by anticonvulsants is mediated through autophagy. SiRNA knock-down of ATG12 in primary human macrophages blocks autophagy leading to (i) reduced LC3-II levels and (ii) loss of CBZ- (50 μM) and VPA- (3 mM) induced enhancement of intracellular killing of M. bovis BCG. P-values, unpaired Student's t-test (n = 3) (compared to vehicle alone): Rap 0.002; VPA 0.001; CBZ 0.0001.

Data information: * P < 0.05; ** P < 0.005. All experiments were carried out at least in triplicate and on at least 3 separate occasions.

Figure 4

Carbamazepine triggers mTOR-independent autophagy through myo-inositol depletion

1. Effect on intracellular killing of luminescent mycobacteria in primary human macrophages of CBZ (50 μM), other dibenzazepines (opipramol, 100 ng/ml; lofepramine, 200 ng/ml; trimipramine, 50 ng/ml), mianserin (200 ng/ml) and the cardiac-specific SCN5A antagonist, flecainide (2 μg/ml). P-values, unpaired Student's t-test (n = ≥ 6) (compared to vehicle alone): CBZ 0.0001; opipramol 0.0005; lofepramine 0.0001; trimipramine 1 × 10⁻⁵.

2. CBZ inhibits Na⁺-dependent myo-inositol uptake by macrophages. Rates of [³H]myo-inositol accumulation within cells were determined in the presence (squares) or absence (circles) of extracellular Na⁺ during treatment with CBZ (50 μM; red) or vehicle alone (white). P-values, unpaired Student's t-test (n = 6) (compared to vehicle alone): CBZ 9.2 × 10⁻⁸.

3. Sodium-dependent myo-inositol uptake (at 5 min) by macrophages is reduced by siRNA knock-down of the Na⁺–inositol co-transporter SLC5A3 (SMIT-1) under which conditions co-treatment with CBZ (50 μM; red) caused no additional inhibition of inositol uptake.

4. Effect on intracellular killing of luminescent mycobacteria in primary human macrophages of (i) siRNA knock-down of the Na⁺–inositol co-transporter SLC5A3 (SMIT-1) (P-values, unpaired Student's t-test (n = 3) siSLC5A3 compared to siControl: 0.009; siControl + CBZ compared to vehicle alone: 0.0067) and (ii) incubation with excess extracellular myo-inositol with (red) or without (white) treatment with CBZ at 50 μM (n = 3).

5. Effect of depleting and adding excess myo-inositol on LC3-II mCherry autophagosome formation in RAW 264.7 macrophages. Myo-inositol depletion increases autophagosome formation (white), with no significant additional change with CBZ treatment (red). Excess myo-inositol further reduces autophagosome number (grey). P-values, unpaired Student's t-test (n = 3) (compared to myo-inositol-free): 0.2 μM 0.013, 200 μM 0.01; 0.2 μM compared 0.2 μM + CBZ: 0.023. Scale bar represents 40 μm.

Data information: * P < 0.05; ** P < 0.005. All experiments were carried out at least in triplicate and on at least 3 separate occasions.

Figure 5

Carbamazepine induces autophagy through decreased IP₃ signalling

1. Phospholipase C and IP₃ receptor regulate autophagic killing of mycobacteria in Dictyostelium. Wild-type (WT, black) and mutant Dictyostelium AX2 strains were grown axenically and incubated with M. abscessus-lux (Renna et al, 2011) at an MOI of 0.01:1. M. abscessus-lux with no Dictyostelium (white) were grown under identical conditions as a control. Viable mycobacteria were quantified by measuring luminescence at 2 h post-infection. Mutant strains tested: mipp1− (multiple inositol polyphosphate phosphatase null), I6KA− (inositol hexakisphosphate phosphate kinase null); plc− (phospholipase C null); IP3R− (inositol (1,4,5)-trisphosphate receptor null mutant iplA-). P-values, unpaired Student's t-test (n = 3) (compared to wild-type): PLC− 0.01; IP3R− 0.007.

2. Quantification by mass spectroscopy of levels of C18:0 PIP₂ species in macrophages following treatment with CBZ (red) or vehicle control (white). SA- C18:0/C20:4. P-values, unpaired Student's t-test (n = 3) (compared to vehicle alone): SA 6.6 × 10⁻⁹, 20:3 1.6 × 10⁻⁶, 18:2 3.5 × 10⁻⁸, 18:1 3.2 × 10⁻⁴.

3. Levels of basal IP₃ (1,4,5) in macrophages following treatment with CBZ (red), lithium (white) or vehicle alone (black) for 24 h (top) or 48 h (bottom). P-values, unpaired Student's t-test (n ≥ 3) (compared to vehicle alone): CBZ 24 h 0.00007; CBZ 24 h 0.0013; LiCl 24 h 0.00005; LiCl 48 h 0.04.

4. Resting levels of mitochondrial calcium (monitored by mt-Cameleon sensor expression) are reduced by CBZ treatment (red) compared to controls. (i) Representative calcium recordings of individual mitochondria from CBZ-treated (red) or control (black) cells, matched for similar fluorescence levels after ionomycin addition (F_iono; arrow) showing lower resting mitochondrial calcium levels after treatment with CBZ. (ii) Normalized mitochondrial calcium levels (F/F_iono) from cells treated with CBZ (red) or vehicle alone (white). P-values, unpaired Student's t-test (n ≥ 20) (compared to vehicle alone): 9.5 × 10⁻¹⁵.

5. CBZ treatment of macrophages results in (E) a decrease in cellular ATP and consequently (F) increased phosphorylation of AMP kinase (AMPK) and ULK1 (ULK). P-values, unpaired Student's t-test (n = 3) (compared to vehicle alone): 0.00003.

6. Model for mechanism of autophagy induction by CBZ. Inhibition of the Na⁺ channel SCN5A leads to reduced activity of the Na⁺–inositol co-transported SLC5A3 leading to a drop in cellular inositol levels. The subsequent reduction in PIP₂ (4,5) leads to a fall in basal IP₃ (1,4,5) and reduced mitochondrial Ca²⁺ levels. The resultant fall in cellular ATP activates AMP kinase and subsequently ULK1 to trigger autophagic disposal of mycobacterial killing.

Data information: * P < 0.05; ** P < 0.005. All experiments were carried out at least in triplicate and on at least three separate occasions.