Research ArticleTuberculosis

Metformin as adjunct antituberculosis therapy

See allHide authors and affiliations

Science Translational Medicine  19 Nov 2014:
Vol. 6, Issue 263, pp. 263ra159
DOI: 10.1126/scitranslmed.3009885


The global burden of tuberculosis (TB) morbidity and mortality remains immense. A potential new approach to TB therapy is to augment protective host immune responses. We report that the antidiabetic drug metformin (MET) reduces the intracellular growth of Mycobacterium tuberculosis (Mtb) in an AMPK (adenosine monophosphate–activated protein kinase)–dependent manner. MET controls the growth of drug-resistant Mtb strains, increases production of mitochondrial reactive oxygen species, and facilitates phagosome-lysosome fusion. In Mtb-infected mice, use of MET ameliorated lung pathology, reduced chronic inflammation, and enhanced the specific immune response and the efficacy of conventional TB drugs. Moreover, in two separate human cohorts, MET treatment was associated with improved control of Mtb infection and decreased disease severity. Collectively, these data indicate that MET is a promising candidate host-adjunctive therapy for improving the effective treatment of TB.


Mycobacterium tuberculosis (Mtb) is the etiological agent of tuberculosis (TB), a disease that continues to be a leading killer worldwide, with an estimated global mortality of 1.43 million people every year (1). Although current treatment of drug-susceptible TB is effective, the therapy regimen is lengthy (6 to 9 months), and problems with drug toxicity and increasing multidrug resistance (MDR) urge the discovery/development of new drugs and therapeutic strategies (2, 3). The global need for new effective therapies has led to a resurgence in efforts to identify additional anti-Mtb agents, several of which are now being evaluated in clinical trials (4). However, conventional pathogen-targeted strategies suffer from the serious disadvantage of fostering microbial resistance. To circumvent this problem, a new paradigm in drug discovery that involves therapeutic modulation of host cell responses to improve pathogen eradication (57) has emerged. These “host-targeted” adjunct therapeutic strategies are less likely to engender microbial resistance than direct targeting of the pathogen with conventional drugs (6). We therefore tested whether existing approved drugs with defined effects on host cell functions could be repurposed for effective and durable anti-TB therapy. The validation of this approach would have the advantage of requiring shorter clinical trials than those usually needed for new drugs (4).

An effective host immune system is a crucial factor for both the control of Mtb growth and its containment as latent TB infection (LTBI). The success of Mtb in infecting the host cells and maintaining long-term persistent infection is associated with the ability of bacilli to evade host innate as well as adaptive immune responses (810). The host cell innate antimicrobial arsenal includes production of reactive oxygen species (ROS) and reactive nitrogen species (RNS), as well as the capacity to destroy intracellular pathogens using the phagosomal machinery or autophagy pathway. Autophagy is required for the effective control of intracellular pathogens (11, 12) and is regulated by mammalian target of rapamycin (mTOR) complex 1 (13), a serine/threonine kinase, and adenosine monophosphate–activated protein kinase (AMPK) (13, 14). Accordingly, perturbations in the autophagy network and AMPK signaling have previously been associated with Mtb virulence (15, 16). We therefore screened the capacity of various U.S. Food and Drug Administration (FDA)–approved mTOR-independent autophagy activators (17) and AMPK modulators for their ability to control intracellular mycobacterial growth. We discovered that the AMPK-activating antidiabetic drug metformin (MET; 1,1-dimethylbiguanide) inhibits the intracellular growth of Mtb, restricts disease immunopathology, and enhances the efficacy of conventional anti-TB drugs.


MET restricts mycobacterial growth by inducing mitochondrial ROS production

A total of 13 autophagy and AMPK-activating drugs were tested for the ability to control the intracellular growth of M. bovis bacillus Calmette-Guérin (BCG) in the human monocytic cell line THP-1 using a colony-forming unit (CFU) assay. Eight of these compounds significantly inhibited the intracellular growth of BCG (fig. S1, A and B, and table S1). None of these eight compounds displayed direct antimycobacterial effects in an in vitro minimum inhibitory concentration (MIC) assay (fig. S1, C to E), suggesting that these compounds restricted BCG growth via effects on host cell–BCG interactions with minimal direct effect, if any. In these experiments, rapamycin, a known autophagy inducer, was used as a positive control. In subsequent studies, we focused on evaluating the therapeutic potential of the AMPK-activating agent MET because this drug is already being used extensively in the clinic for the treatment of type 2 diabetes mellitus (DM), and the effects of the drug on cellular functions including AMPK signaling are known (18, 19). In our experiments, MET treatment inhibited the growth of intracellular BCG (fig. S2A) as well as the H37Rv strain of Mtb (Fig. 1A) in a dose-dependent fashion in THP-1 and human monocyte-derived macrophages (hMDMs), respectively. MET inhibition of mycobacterial growth was observed within the first 24 hours, and no further significant reduction in mycobacterial survival was observed at later time points (Fig. 1B and fig. S2B). This effect of MET on mycobacterial growth was abolished in the cells where AMPK was either genetically inactivated (Fig. 1C) or chemically inhibited (Fig. 1D). MET also restricted the intracellular replication of MDR strains of Mtb (Fig. 1E). In our experiments, MET exposure induced AMPK phosphorylation, consistent with the known effects of the drug on kinase activation (fig. S2C) (18). The relevance of AMPK in inhibition of mycobacterial growth was further confirmed by using 5-aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR), another AMPK activator (fig. S2, D and E, and table S1) (20). We did not observe a direct effect of MET on Mtb growth (fig. S3).

Fig. 1. MET inhibits the intracellular growth of mycobacteria by inducing mROS.

(A) Mtb survival after 24 hours in hMDM cells treated with different doses of MET expressed as percentage of Mtb CFU in untreated cells. (B) Mtb survival after treatment of hMDM cells with 2 mM MET. (C) Mtb survival after 24 hours in control (NT) and AMPK knockdown (AMPK−/−) THP-1 cells not treated (UNT) or treated with 2 mM MET (M). (D) Mtb survival in THP-1 cells not treated or treated with 2 mM MET in the presence of compound C (C). (E) Survival of MDR strains of Mtb after 72 hours in THP-1 cells treated with 100 μM INH (I) and different concentrations of MET. The name of each strain is indicated. (F) Fluorescence intensity of MitoSOX-stained THP-1 cells. M. bovis BCG-infected THP-1 cells were treated for 4 hours with 2 mM MET (orange line) or 1 μM rotenone (red line). Blue line, infected cells; green line, uninfected cells; black line, unstained cells. (G) Composite data from four to five independent experiments conducted as in (E), showing MitoSOX mean fluorescence intensity (MFI) in uninfected (C) and infected cells treated with MET or rotenone (R), presented as fold change relative to untreated infected cells (UNT). (H) Mtb-infected hMDMs, as in (A), were treated with 2 mM MET in the presence or absence of ROS scavengers NAC (N) and GSH (G). (I) hMDMs were infected with Alexa Fluor 488 (AF488)–labeled BCG (BCG-AF) and treated with 2 mM MET in the presence or absence of NAC and 25 μM LTR for 4 hours before fixation. UNT, untreated infected cells; M, MET-treated cells; M + N, MET- and NAC-treated cells. Starved infected cells (ST) were positive controls. Scale bar, 10 μm. (J) Quantification of LTR-positive BCG-AF from (I). P = 0.0047, χ2 test. Data in (A) to (D) and (G) to (H) are representative of three to five independent experiments. Values are expressed as means ± SEM. P values are provided in table S11, two-tailed Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001.

AMPK may also regulate cell growth and important cellular innate antimicrobicidal functions (14). To determine the mechanism(s) by which MET restricted intracellular replication of mycobacteria, we next evaluated the MET-treated infected cells for ROS production, phagosome-lysosome fusion, autophagy, and apoptosis induction. MET treatment selectively induced the production of mitochondrial ROS (mROS) within just 4 hours of drug exposure (Fig. 1F and fig. S4A), whereas cytoplasmic ROS (cROS) were not elicited at this early time point (fig. S4B). The production of mROS could be due to MET-mediated inhibition of mitochondrial complex I (NADH dehydrogenase) activity (21, 22). We used the known complex I inhibitor rotenone (ROT) to confirm the efficient detection of induced mROS and cROS in our assays (Fig. 1F and fig. S4B). MET treatment overcame mycobacteria suppression of mROS synthesis (Fig. 1G) and enhanced the production of mROS in a time- and dose-dependent manner (fig. S4, A and C). Addition of the ROS-scavenging agents N-acetylcysteine (NAC) and glutathione (GSH) to our cultures abolished the ability of MET to induce mROS (fig. S5, A and B) and restored mycobacterial growth in MET-treated cells (Fig. 1H and fig. S5C). Next, we assessed the effects of MET on phagosome-lysosome fusion in cells infected with fluorescent BCG and stained with the lysosomal marker LysoTracker Red (LTR). MET treatment for either 4 or 16 hours led to a significant increase in fluorescent BCG accumulation in LTR-labeled compartments, whereas addition of ROS scavengers abolished these effects (Fig. 1, I and J, and table S2). MET treatment induced autophagy as assessed by up-regulation of lipidated form of light chain 3 (LC3B; fig. S6A); however, blocking autophagy by 3-methyladenine did not abolish MET-mediated restriction of mycobacterial growth (fig. S6B). These data suggested that mROS produced early during MET treatment is a key mechanism by which the drug restricts the intracellular mycobacterial growth in vitro. Mitochondrial derived ROS can promote dissipation of mitochondrial membrane potential, release of cytochrome c into the cytoplasmic milieu, and initiation of intrinsic apoptotic pathways leading to cell death (23). These phenomena were not observed in our experiments either at early time points (4 hours) or after prolonged MET exposure (24 and 48 hours), indicating that the drug’s mechanism of action is not causing cell death of host cells (fig. S7, A to D).

MET enhances the efficacy of conventional anti-TB drugs

We next evaluated the in vivo efficacy of MET treatment in a mouse model of acute and chronic TB (24). Mtb-infected mice were administered MET alone or MET in combination with either isoniazid (INH) or ethionamide (ETH), starting on day 7 or 42 after infection. In five different acute model experiments, mice treated with MET (500 mg/kg) alone had reduced bacillary load in both lung and spleen (Fig. 2, A to E, and fig. S8, A and C). This dose is equivalent to a MET dose of 2430 mg/day for a 60-kg human ( (25) and lower than the maximum daily dose for MET therapy in diabetic patients (3000 mg/day;; Furthermore, MET administration enhanced the efficacy of the conventional first-line anti-TB drug INH, as demonstrated by decreased bacillary load in the lungs of mice cotreated with INH + MET when compared with mice that received INH alone (Fig. 2C and fig. S8, A and B). Indeed, in some experiments, we were unable to detect any CFU in the lungs of mice that received combined INH + MET therapy (table S3). We next evaluated the efficacy of MET adjunctive therapy when used in combination with the second-line anti-TB drug ETH. In an acute model, Mtb-infected mice cotreated with ETH + MET exhibited decreased CFU in lung and spleen when compared with mice that received ETH alone (Fig. 2, D and E, and fig. S8, C and D). The efficacy of MET was further evaluated in the chronic model of Mtb infection. In this experiment, mice treated with MET or INH + MET starting from day 42 after infection showed decreased bacillary load in the lungs compared with untreated or INH-treated mice, respectively (Fig. 2F). Together, these experiments suggest that MET represents a potential adjunct drug that can enhance Mtb clearance by conventional treatment.

Fig. 2. Efficacy of MET monotherapy and combination therapy in a mouse model of TB.

(A) Mtb-infected mice were treated with MET [250 mg/kg (M250) or 500 mg/kg (M500)] starting 7 days after infection. Bacillary loads (CFU) were enumerated in the lungs on days 1, 7, 21, and 35 after infection. UNT, untreated infected mice. (B) Spleen bacillary load in Mtb-infected mice treated with MET as described in (A). (C) Efficacy of INH monotherapy (5 mg/kg, I5) and INH + M500 therapy in the lungs of Mtb-infected mice. (D) Efficacy of ETH monotherapy (15 mg/kg, E15) and ETH + M500 therapy in the lungs of Mtb-infected mice. (E) Spleen bacillary load in Mtb-infected mice treated as described in (D). (F) Mtb-infected mice were treated daily with MET (250 mg/kg, M250), INH (10 mg/kg, I10), and I10 + M250 in drinking water starting 42 days after infection. Bacillary loads (CFU) were enumerated in the lungs on days 1, 21, 42, and 100 after infection. In these experiments, each group per time point consisted of four to nine mice. Data are expressed as means ± SEM. P values are provided in table S12, two-tailed Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001.

MET reduces TB-induced tissue pathology and enhances immune response

To significantly improve current TB therapy, anti-Mtb drugs should also promote resolution of tissue pathology in addition to accelerating bacillary clearance (4). Lungs and spleens from Mtb-infected mice treated with MET were smaller than those of untreated mice at 35 days after infection (Fig. 3, A and B). As expected, mice treated with the conventional anti-TB drug INH or ETH exhibited a clear reduction in organ size and tissue lesions, and combination therapy of MET with INH or ETH further reduced tissue pathology. Histopathological evaluation of the infected lungs of untreated control mice revealed diffuse coalescent lung lesions with large numbers of infiltrating macrophages and lymphocytes and scattered intracellular acid-fast bacilli (AFB) (Fig. 3C). In contrast, MET treatment was associated with reduced numbers of AFB and increased lymphocyte infiltration of the infected tissues (Fig. 3C), which has previously been linked with improved Mtb control in mice (2628). Mice treated with INH alone had few residual lesions in the lungs with areas of only increased cellularity. No granulomas were observed in the lungs of mice treated with INH + MET, with some areas of the lungs appearing completely normal. Morphometric analysis of the histological sections indicated that the percentage of total area of lung tissue involved in TB pathology at 35 days after infection was reduced in MET-treated mice compared with untreated control animals (3.8% versus 9.9%, respectively; Fig. 3D) and that combination therapy with MET and INH further reduced areas of lung tissue damage compared to INH-alone treatment (0.13% versus 0.70%, respectively; Fig. 3D). We next evaluated MET effects on T helper cell 1 (TH1) immune responses in lung tissues because TH1 immunity participates in controlling Mtb infection (29). MET-treated Mtb-infected mice showed a trend of larger CD4+ and CD8+ T cell numbers in the lung compared to untreated mice (fig. S9, A to C). In Mtb-infected mice, MET treatment was associated with an increased percentage and number of mycobacteria-specific interferon-γ (IFN-γ)–secreting CD8+ T cells compared with untreated control animals (Fig. 3, E and F). The number, but not the percentage, of IFN-γ–secreting CD4+ T cells also showed an increasing trend in the lungs of Mtb-infected mice upon MET therapy (fig. S9, D and E). An increase of lung IFN-γ–secreting CD8+, but not CD4+, cells was also observed in uninfected mice treated with MET for 3 weeks (fig. S10). These data suggest that MET treatment promotes lung accumulation of both CD8+ T cells and CD4+ T cells independently of infection and that this effect may also contribute to Mtb infection control.

Fig. 3. MET reduces lung tissue pathology and enhances immune responses in mice.

(A) Gross appearance of mouse lungs and spleens on day 35 after infection from Mtb-infected mice treated or not treated with MET [250 mg/kg (M250) or 500 mg/kg (M500)] and INH (10 mg/kg, I10) or cotreated with I10 and MET. UNT, untreated infected mice. (B) Gross appearance of mouse lungs and spleen on day 35 after infection from Mtb-infected animals treated or not treated with M500 and ETH (10 mg/kg, E10) or cotreated with E10 and M500. (C) Light micrographs of hematoxylin and eosin (H&E) staining of lung sections as described in (A). Scale bar, 500 μm (4×); 50 μm (40×). (D) Morphometric analysis of lung sections shown in (C) indicating the proportion of lung area involved in disease pathology. Bar represents the median. (E) Total lung cells from Mtb-infected mice treated or not treated with M500 were stimulated ex vivo with Mtb. Dot plots show the proportion of IFN-γ–producing CD3+CD8+ lung T cells isolated on day 35 after infection. (F) Percentage data of four to five mice per group as mentioned in (E). (G) Data in (F) represented as number of CD8+ IFN-γ+ in the lung. Data in (D) and (F) are expressed as means ± SD. (E) and (F) are representative of two experiments. *P = 0.036, **P = 0.006, and ***P = 0.0007, two-tailed Student’s t test.

MET reduces inflammatory response

Because MET therapy reduced tissue pathology and enhanced host immune responses in Mtb-infected mice, we next sought to evaluate the possible beneficial effects of the drug on chronic inflammation associated with TB disease. A genome-wide transcriptional analysis using total RNA isolated from Mtb-infected mouse lungs treated or not treated with MET (500 mg/kg) was performed. At 35 days after infection, 1580 genes were differentially expressed in response to infection (fig. S11A and table S4). Treatment of infected mice with MET affected the expression of 353 genes as compared to no treatment (fig. S11A and table S5). Among these genes, 97% (343 of 353) overlapped with those whose expression was affected by Mtb infection (fig. S11B). MET reversed the expression pattern of most of these genes (331 of 343; fig. S11B). Ingenuity pathway analysis (IPA) of respective gene sets indicated that the host cell genes up-regulated by infection and down-regulated by MET were mediators of inflammatory response including interferon signaling, which is found elevated in TB patients and is normalized by anti-TB therapy (30) (fig. S11, B and C, and tables S6 and S7). To further clarify the effects of MET treatment on host cell function during infection, we performed parametric analysis of gene set enrichment (PAGE) (31) on differentially expressed genes. We found that 45 of the 48 pathways modulated by Mtb infection tended toward normalization after MET treatment (Fig. 4A and tables S6 and S7). Many of these pathways, such as activation of liver X receptor, pathogen recognition receptor, and interferon regulatory factors, are associated with inflammatory responses. Consistent with these data, MET treatment reduced the expression of inflammatory associated genes such as IL-1β, TNF-α, IL-6, MCP-1, CXCL5, and CXCL10 in the mouse lung (Fig. 4B and tables S4 and S5) and spleen (Fig. 4C). In conclusion, the observed beneficial effects of MET on pathology in the Mtb-infected animals appeared to be associated with drug attenuation of mycobacteria-induced inflammation.

Fig. 4. MET reduces inflammatory gene activation.

(A) Comparison of gene sets (functional pathways) significantly altered in the lung cells of Mtb-infected mice (I) versus uninfected mice (U) and MET (500 mg/kg)–treated infected mice (M) versus infected mice (I) at 35 days after infection. Heat map displays z-score normalized pathway enrichment values from PAGE analysis. The up-regulated (red) and down-regulated (green) pathways are shown. “*” indicates pathways that were significantly altered in infected mice upon MET treatment. P values are in table S7. (B) Absolute expression values of the inflammatory genes in the lung cells as described in (A). *P < 0.05; **P < 0.01; ***P < 0.001. (C) mRNA levels in spleen cells from mice described in (A). Data are expressed as means ± SEM. Each group consisted of four to five mice. *P < 0.05; ***P < 0.001. The experiments were performed twice, and one representative data set is shown.

MET therapy is associated with decreased TB severity and improved clinical outcome in patients

We next assessed whether the potent anti-Mtb effects of MET treatment observed in our in vitro and in vivo assays might also be associated with positive effects in human patients with TB disease. Because MET is an approved antidiabetic drug in routine clinical use, we evaluated the effects of MET treatment by retrospective study of data collected from all TB patients with diabetes as a comorbidity seen at the TB Control Unit of Singapore. Of the total cohort of 1695 TB patients, 296 patients had DM at the time of the clinical assessment for TB (a subgroup known to be highly susceptible to Mtb infection and confirmed here to have increased TB disease severity; P < 0.001) (32). Of this TB + DM patient cohort, 109 were being treated for diabetes with MET and 164 with other antidiabetic drugs such as sulfonylurea or were newly diagnosed for DM (“Non-MET”) at the time they visited the clinic. There was no significant difference between the two groups by sex, ethnicity, body mass index, smoking history, diabetic control, and the AFB smear status at baseline (table S8). Patients on MET therapy were more likely to be older (age >60 years) compared to those on other therapies (table S8). Within the TB + DM patients, individuals of the MET group had fewer pulmonary cavities than the Non-MET group [odds ratio (OR), 0.6; 95% confidence interval (CI), 0.36 to 0.97; P = 0.041; Fig. 5A] as assessed by chest x-ray (CXR). In a multivariate analysis, it was found that TB + DM patients who were receiving MET were older than those who were on a Non-MET regimen (log OR, 0.02; 95% CI, 0.002 to 0.044; z score = 2.173; P = 0.029). When CXRs were graded according to standard radiological classification, the lower proportion of TB + DM patients on MET had far advanced disease compared to those who were in the Non-MET group (37.8% versus 44.5%; OR, 1.307; 95% CI, 0.77 to 2.2; P = 0.3539; table S8). Because TB + DM patients exhibit increased risk of death (32), we next compared mortality rates during TB treatment between the MET (n = 102) and Non-MET (n = 153) groups. Mortality was 3% among patients who received MET compared to 10% among patients in the Non-MET group (OR, 0.29; CI, 0.14 to 0.95; P = 0.039; Fig. 5B). Multivariate analysis showed that the TB + DM patients in the MET group are less likely to die (log OR, −1.75; 95% CI, −3.14 to −0.36; z score = −2.481; P = 0.013) and were also older (log OR, 0.12; 95% CI, 0.073 to 0.18; z score = 4.518; P = 6.23 × 10−6) than Non-MET patients.

Fig. 5. Effect of MET in TB patients with DM as comorbidity.

(A) Percentages of TB patients with DM as comorbidity showing detectable pulmonary cavities determined by CXR at the time of first visit to TB clinic. MET-receiving (M) and alternative drug–receiving (Non-M) TB + DM patients. *P = 0.041; OR, 0.6; 95% CI, 0.36 to 0.97. (B) Mortality among M and Non-M TB + DM patients during anti-TB therapy. *P = 0.039; OR, 0.29; CI, 0.14 to 0.95.

MET treatment is associated with reduced incidence of latent TB

We next analyzed prospective data from a second cohort of DM patients and assessed any potential effects of MET on the incidence of LTBI. Of the total 220 DM patients, 62 had LTBI (DM + LTBI patients; table S9), as determined by T-SPOT.TB test. In a multivariate stratified analysis, we confirmed that MET therapy was associated with reduced T-SPOT reactivity (25.6%) when compared with patients treated with alternative drugs (Non-MET T-SPOT reactivity, 42.4%; P < 0.05; OR, 0.44; CI, 0.20 to 0.95; Fig. 6A), thus implying that MET lowers the chance of being T-SPOT–positive (LTBI). MET-treated DM + LTBI patients had larger numbers of IFN-γ–secreting T cells specific for the Mtb CFP-10 antigen than Non-MET DM + LTBI patients (table S9). Furthermore, a larger number of MET-treated than Non-MET–treated DM patients showed T cell reactivity to both CFP-10 and ESAT-6 antigens (35% versus 14%, respectively; Fig. 6B). Together, these findings suggest that MET treatment is associated with enhanced Mtb-specific T cell immune response that may protect against latent TB.

Fig. 6. Reduced incidence of LTBI in MET-treated DM patients.

(A) Frequency of LTBI among DM patients receiving MET (M) or alternative drugs (Non-M), as determined by T-SPOT IFN-γ test. The frequency of T-SPOT positivity was reduced in the MET-treated group. *P = 0.041; OR, 0.44; CI, 0.20 to 0.95, multiple logistic regression test. (B) Scatterplot showing the number of ELISpot forming units (SFU) in individual DM patient in response to the Mtb ESAT-6 and CFP-10 antigens in the blood as measured by T-SPOT IFN-γ test. Vertical and horizontal dotted lines represent the cutoff of ≥5 SFU per 2.5 × 105 cells. MET-treated (M) patients, n = 48; patients treated by Non-MET drugs (Non-M), n = 14.


Identification of new host-directed therapies that can improve clinical outcomes for TB patients has been highlighted as a priority by the World Health Organization, leading to the specific recommendation that studies of immunomodulatory agents for adjunct treatment of TB should be fast-tracked (33, 34). Such therapeutic approaches should potentially (i) augment the efficacy of antibiotic-based treatment, (ii) shorten treatment regimens for drug-susceptible and/or drug-resistant Mtb infections, (iii) reduce the immunopathology associated with TB, and (iv) promote development of immunological memory that protects against relapse. A number of in vitro and animal studies have attempted to address these issues by repurposing existing drugs, but mixed outcomes were reported (4, 3537). Some of these drugs, such as phosphodiesterase 4 inhibitors, alter tissue inflammation associated with Mtb infection, and the anti-inflammatory effects were considered beneficial (26, 38). Despite the studies in the field, it remains to be clarified how protective immunity and pathological immunity in Mtb infection are differentiated, and whether these two aspects of immune response can be separated (34). Our data suggest that they can be independently modulated with beneficial effects during Mtb infection because MET enhances Mtb-specific host immunity, reduces inflammation, promotes disease resolution, and improves TB treatment outcome.

At the cellular level, MET differentially affects immune response and inflammation through different mechanisms. The protective effect is mediated by increased host cell production of mROS and increased acidification of mycobacterial phagosome. Indeed, mROS produced upon mitochondrial recruitment to phagosomes is instrumental in the killing of intracellular bacteria by macrophages (39). The anti-inflammatory effect is mediated by activation of AMPK, a negative regulator of inflammation (40). MET treatment also promotes the expansion of Mtb-specific IFN-γ–secreting T cells in the lungs of infected mice. The expansion of lung IFN-γ–secreting CD8+ T cells in uninfected mice indicates that MET has an impact on the lung immune cells independently of current infections. These latter effects are consistent with MET-induced expansion of CD8+ memory T cells (41), another known consequence of AMPK activation (42). Both CD4+ and cytotoxic T cells are key to controlling primary Mtb infection in human subjects (29, 4345). Notably, our data also showed increased numbers of Mtb-specific T cells in MET-treated diabetic patients with LTBI.

MET therapy combined with standard TB treatment regimens was associated with beneficial consequences on clinical outcomes in active TB disease, as observed in the retrospective study presented here. Furthermore, MET treatment exerted favorable effects on LTBI in diabetic patients. Several issues remain to be addressed before introduction of MET in TB therapy. First, although our retrospective clinical data are very encouraging, they might be subjected to confounders, which are difficult to precisely identify. Second, the exact dose of MET to be used in the adjunctive therapy remains to be evaluated in TB patients. Third, it has to be investigated whether integration of MET with current therapeutic treatments may also improve prophylactic control of TB and reduce duration of treatment. All these issues require accurate prospective clinical trials to properly test the efficacy of MET as adjunct TB therapy in different environmental conditions.


Study design

The objective of this study was to identify FDA-approved drugs that can be repurposed to modulate host immune system and capable of controlling the growth of intracellular Mtb. First, we screened 13 autophagy and AMPK-activating drugs for their effect on the intracellular growth of M. bovis BCG, a surrogate for Mtb, in vitro in human cells. These experiments led to the identification of MET as the most appropriate candidate. Next, detailed mechanistic studies were performed to elucidate the mechanism of MET in controlling Mtb growth in vitro and in mouse models of Mtb infection. MET was then evaluated as adjunctive treatment together with standard TB therapeutic drugs in mouse Mtb infection models. Mice were infected and randomized in different groups 1 day before the start of treatment. Finally, to reveal the possible effect of MET in TB patients, retrospective data were mined from two independent cohorts (TB and diabetic cohorts).


The following drugs were used: MET (Sigma, no. D150959), INH (Sigma, no. I3377), ETH (Sigma, no. E6005), streptomycin sulfate (Sigma, no. S9137), Bay K8644 (Sigma, no. B112), R-(+)-Bay K8644 (Sigma, no. B132), S-(−)-Bay K8644 (Sigma, no. B133), FPL 64176 (Sigma, no. F131), 2′, 5′-dideoxy-adenosine (Merck, no. 288104), loperamide hydrochloride (Sigma, no. L4762), nimodipine (Sigma, no. N149), nitrendipine (Sigma, no. N144), KT5720 (Merck, no. 420323), rapamycin (Sigma, no. R8781), verapamil hydrochloride (Sigma, no. V4629), and amiodarone hydrochloride (Sigma, no. A8423).

Cell culture

THP-1 cells (human monocytic cell line) were maintained in RPMI 1640 supplemented with 10% fetal bovine serum (FBS), 1% sodium pyruvate, 1% l-glutamine, 1% kanamycin, and 1% nonessential amino acids.

Preparation of hMDMs

Total blood was processed to isolate peripheral blood mononuclear cells (PBMCs) with a Ficoll gradient. Monocytes were enriched from PBMCs by CD14 magnetic selection (Miltenyi-Biotec beads). The purified CD14+ monocytes were resuspended in RPMI 1640 with 10% FBS, 1% penicillin, 2 mM l-glutamine, and human macrophage colony-stimulating factor (M-CSF) (100 ng/ml) (R&D Systems) and incubated in six-well plates at 37°C in 5% CO2 for 6 days. The medium was changed on the 4th day. On day 6, differentiated macrophages were collected after a gentle wash, centrifuged, plated in fresh medium without human M-CSF, and cultured overnight. On day 7, cells were washed and used.

Bacterial strains and growth conditions

Mtb H37Rv and M. bovis BCG strains were grown in Middlebrook 7H9 broth (BBL Microbiology Systems) supplemented with albumin-dextrose-catalase (ADC) (Difco Laboratories) and 0.05% Tween 80 at 37°C for 5 to 7 days to an absorbance (A600) of 0.4 to 0.5. After this time, mycobacterial cells were pelleted, resuspended in fresh 7H9 broth with 20% glycerol, and stored at −80°C. One vial of the stock was thawed to enumerate CFU per milliliter. On the day of infection, the cells were thawed, washed, and sonicated before being used in experiments. Drug-resistant strains (table S10) were maintained at the Public Health Research Institute, Newark, NJ.

In vitro mycobacterial infection

Frozen mycobacteria (M. bovis BCG or Mtb H37Rv) were thawed, washed, and resuspended in antibiotic-free RPMI 1640 with 10% FBS and were used to infect THP-1 and hMDM cells in six-well plates with a multiplicity of infection (MOI) of 5. The infected cells were incubated at 37°C with 5% CO2 for 3 hours. After this time, cells were washed two times with antibiotic-free medium by centrifuging at 800 rpm. The infected cells were counted, seeded in triplicate, and either left untreated or treated with different compounds. At the indicated time points, cells were used for various assays. THP-1 cells were differentiated for 16 to 18 hours with 4 μM phorbol myristate acetate (Sigma, no. 8139) before infection with MDR strains.

In vitro determination of MIC

The MIC of different compounds was determined using microdilution method. The log-phase culture of M. bovis BCG (with A600, 0.4 to 0.5) was adjusted to an A600 of 0.02 by 7H9-ADC. Two hundred microliters of this was plated in wells of 96-well plates along with different concentrations of various compounds (table S1). The BCG cultures treated with different compounds were incubated at 37°C with 5% CO2 for 4 days, after which A600 was measured. Experiments were conducted three times in duplicate.

Enumeration of mycobacteria in infected cells

At predetermined time points after infection, the infected cells were washed once with phosphate-buffered saline (PBS) and then lysed with 200 μl of PBS with 1% SDS. Various dilutions of this lysate were plated on Middlebrook 7H11 agar supplemented with 10% oleic acid–albumin–dextrose–catalase (OADC, Difco Laboratories), in triplicate. Agar plates were incubated at 37°C for 3 weeks, after which colonies were counted visually. CFUs obtained from two or three dilutions were used to calculate the total number of CFU per milliliter. Data are presented as percentage of mycobacterial survival in compound-treated cells versus untreated cells.

Measurement of ROS production upon M. bovis BCG infection

Uninfected and M. bovis BCG–infected THP-1 and hMDM cells were treated with different concentrations of MET or 1 μM ROT. At predetermined time points, mycobacterium-infected cells were harvested, washed with Hanks’ balanced salt solution (HBSS), and resuspended in either HBSS with 10 μM MitoSOX (which measures mROS and mainly detects superoxide radical; Molecular Probes) or HBSS with 5 μM CM-H2DCFDA (which measures cROS and mainly detects H2O2; Molecular Probes). Cells were incubated at 37°C with 5% CO2 for 15 min, washed with HBSS, and resuspended in 200 μl of HBSS. The stained cells were then acquired using a flow cytometer (BD Canto, Becton Dickinson). Analysis was carried out using FlowJo software. In some experiments, ROS quenchers such as 10 mM NAC or 10 mM GSH were used. These scavengers were added to the cultures 30 min before the addition of MET.

Measurement of M. bovis BCG in lysosomes

THP-1 and hMDM cells were cultured in ibidi chambers and infected with AF488-labeled BCG. After infection, cells were cultured in the presence of 25 μM LTR (Invitrogen) in the presence or absence of 2 mM MET for either 4 or 16 hours. Then, cells were washed with PBS and fixed in 4% paraformaldehyde for 20 min at room temperature. Fixed cells were washed with PBS and mounted in FluorSave (Calbiochem). Fluorescence intensities of AF488-labeled BCG and LTR were analyzed using an Olympus FV1000 confocal microscope and used to calculate the number of bacteria colocalized with LTR together with the total number of intracellular bacteria. In total, 50 to 60 infected cells were counted for different analyses. In some experiments, ROS quenchers such as 10 mM NAC or 10 mM GSH were used. As a positive control, infected cells were left in starvation medium (Earle’s balanced salt solution, Sigma) for 4 or 16 hours.

Western blotting

Protein lysates from M. bovis BCG THP-1 cells either untreated or treated with 2 mM MET were obtained by lysis in radioimmunoprecipitation assay buffer (Sigma, no. R0278) with protease (Roche, no. 04693159001) and phosphatase (Roche, no. 04906837001) inhibitors at the indicated time points. A Micro BCA Protein Kit (Thermo Scientific, no. 23227) was used to measure protein levels, and equal amounts of proteins were electrophoresed on 12% tris-HCl gels (Mini-PROTEAN TGX gels, Bio-Rad, no. 456-8036) and transferred onto polyvinylidene difluoride membranes (Trans-Blot Turbo transfer pack, Bio-Rad, no. 170-4156). The membranes were developed using AMPKα (Cell Signaling Technology, no. 2603), phospho-AMPKα (Thr172, Cell Signaling Technology, no. 2535), LC3B (Cell Signaling Technology, no. 3868), and GAPDH (glyceraldehyde phosphate dehydrogenase) (Cell Signaling Technology, no. 2118) antibodies.

Mouse infections with Mtb

The in vivo antitubercular activity of MET was evaluated in an acute and chronic infection model of C57BL/6 mice (24). Six- to 8-week-old female C57BL/6 mice were infected with Mtb H37Rv, using a nose-only aerosolization system (CH Technologies). Three to four animals were sacrificed on day 1 to determine the number of bacteria implanted in the lungs. Seven or 42 days after infection, animals were randomly distributed into different groups before the start of treatment. Treatment was given by gavaging the mice with respective drugs once a day, 6 days a week. Mice were sacrificed at predetermined time points (2 or 4 weeks after initiation of treatment, that is, day 21 or 35 after infection, respectively), followed by harvesting of tissues and Mtb CFU enumeration. In total, six experiments were performed. In four experiments (experiments 1 to 3, and 6), MET was given either alone or in combination with INH. In two experiments (experiments 4 and 5), MET was given either alone or in combination with ETH. Because both MET and INH are water-soluble, in two experiments (experiments 3 and 6) mice were administered drugs in drinking water. All mice were housed in a biosafety level 3 (BSL3) laboratory and treated humanely using procedures described in animal care protocols. The study was approved by the Institutional Biosafety Committee and Institutional Animal Care and Use Committee of Biological Resource Council, Agency for Science, Technology and Research (A*STAR).

Enumeration of Mtb CFU in infected mice

The mycobacterial load in the lung and spleen of infected mice was quantified by plating tissue homogenates on Middlebrook 7H11 agar supplemented with OADC. Briefly, at predetermined time points after infection/after treatment, mice were euthanized. Lungs and spleen were aseptically excised, washed in PBS, and homogenized using PBS with 0.25% Tween 80 and magnetic cell sorter (MACS) tissue dissociator (Miltenyi Biotec). Various dilutions of tissue homogenates were plated on Middlebrook 7H11 agar plates in triplicate. Agar plates were incubated at 37°C for 3 weeks, after which colonies were counted visually. CFUs obtained from two or three dilutions were used to calculate the total number of CFU per tissue per mouse.

Isolation of mouse lung cells

On day 35 after infection, mice were euthanized; lungs were aseptically removed from the pulmonary cavity and gently homogenized using RPMI 1640 and MACS tissue dissociator. Lung cells were passed through a 40-μm cell strainer (BD Biosciences) to separate cell clumps. Red blood cells (RBCs) were lysed with RBC lysis buffer (Lonza), washed, and resuspended in RPMI 1640, and total cell numbers were counted.

Ex vivo stimulation of mouse lung cells

Mouse bone marrow–derived dendritic cells (50 × 105) were infected with Mtb (MOI 1). After 4 hours, freshly isolated mouse lung cells (0.5 × 106) were added, and culture was continued for an additional 16 hours. During the final 4 hours of culture, brefeldin A (1 μg/ml) and monensin (1 μM) were added. Cultured cells were washed and stained with allophycocyanin (APC)–Cy7–anti-CD8α (BioLegend, no. 100714), phycoerythrin (PE)–Cy7–anti-CD3 (BioLegend, no. 100220), Pacific Blue–anti-CD4 (BioLegend, no. 100428), and LIVE/DEAD Fixable Green Dead Cell Stain (Invitrogen, no. L23101). Cells were washed, fixed, and permeabilized using Cytofix/Cytoperm buffer (Cytofix/Cytoperm kit, BD Pharmingen). Permeabilized cells were washed twice with Perm/wash buffer, followed by intracellular staining with APC–anti–IFN-γ (BioLegend, no. 505810). Stained cells were acquired using MACSQuant (Miltenyi Biotec), and analysis was carried out using FlowJo software (Tree Star Inc.).

Assessment of lung immune cells from uninfected mice

Uninfected mice were treated daily with MET (250 mg/kg) for 3 weeks. Post-euthanization lung cells were isolated as described above, followed by stimulation with plate-bound anti-CD3 and anti-CD28 for 16 hours. During the final 4 hours of culture, brefeldin A (1 μg/ml) and monensin (1 μM) were added. Cultured cells were washed, stained, and acquired using BD LSR2 (Becton Dickinson). Analysis was carried out using FlowJo software (Tree Star Inc.).

Histology and morphometry

The upper right lobes of mouse lungs were fixed in 10% buffered formalin and paraffin-embedded. Sections were stained with H&E or Ziehl-Neelsen acid-fast stain for evaluation of pathology and photography. For morphometric analysis of mouse lungs, H&E-stained lung sections (n = 5 per group) were scanned with a PathScan Enabler IV scanner (Meyer Instruments). SigmaScan Pro 5 software was used to measure the extent of lung involvement as the percent of the lung sections occupied by lesions.

Microarray analysis on lung cells

Lung cells from uninfected mice and Mtb-infected mice treated or not treated with MET (500 mg/kg) (five biological replicates for each condition) were harvested and homogenized in mirVana lysis buffer (Life Technologies). RNA was extracted using mirVana miRNA Isolation Kit (Life Technologies). Uninfected cells were kept as controls. RNA quality was confirmed using the Agilent Bioanalyzer, and only samples with RNA integrity number >7 were processed. Biotinylated complementary RNA was prepared starting from 100 ng of total RNA according to the Epicentre TargetAmp Nano-g Biotin-aRNA Labeling Kit for Illumina system, followed by hybridization on Illumina mouse WG-6 v3 arrays. Raw expression data were extracted using GenomeStudio Gene Expression v1.9.0 and processed with quantile normalization (46, 47). Hierarchical clustering analysis with complete linkage algorithms was done with R (48). Heat maps were plotted using Spotfire (TIBCO Software Inc.; Differential expression analysis was performed using Linear Models for Microarray Data (LIMMA) (49). Pathway analysis was carried out through the use of IPA (Ingenuity Systems;, and PAGE analysis was carried out as described (31).

RNA isolation from spleen cells and quantitative polymerase chain reaction

Splenocytes were lysed and homogenized in TRIzol reagent (Invitrogen), and total RNA was isolated using RNeasy Mini Kit (Qiagen). For quantitative polymerase chain reaction (qPCR) analysis, 1 μg of total RNA was reverse-transcribed using iScript cDNA synthesis kit (Bio-Rad), and qPCR was performed in triplicate to quantify the levels of IL-1β, TNF-α, IFN-γ, and IL-6 genes. GAPDH was amplified as an internal control. The following primers were used: IL-6, CCTCCGACTTGTGAAGTAGT (forward) and TGCAAGAGACTTCCAGTTG (reverse); TNF-α, CCACGTCGTAGCAAACCACC (forward) and CCTTGTCCCTTGAAGAGAACC (reverse); IFN-γ, AGCGGCTGACTGAACTCAGATT (forward) and GTCACAGTTTTCAGCTGTATAGGG (reverse); IL-1β, AAAGCTCTCCACCTCAATGG (forward) and TCTTCTTTGGGTATTGCTTGG (reverse); GAPDH, TCGTCCCGTAGACAAAATGG (forward) and TTGAGGTCAATGAAGGGGTC (reverse).

Human subjects

Patients from two independent cohorts (TB and diabetic cohorts) were included in this study. Patients of the TB cohort (n = 1695) visited the Tuberculosis Control Unit of Tan Tock Seng Hospital (TTSH) in year 2010. In this TB cohort, 296 had DM and were treated or not treated with MET. Clinical information of 273 TB + DM patients was available for analysis. Of these 273, 109 were treated with MET, 95 were treated with non-MET drugs such as sulfonylurea, and 69 had newly diagnosed DM. All 95 and 69 patients (total 164) who did not receive MET were included in the Non-MET group. With housing type (one- to two-room government housing, three- to four-room government housing, or five-room/private housing) as the surrogate for the economic status of the patient, there was no significant difference between the two groups by the housing types. Three senior physicians independently read the CXRs, and the final outcome was decided by at least two concurring opinions of the CXR with cavity of any size. CXR were also graded according to the standard radiological classification of mild, moderate, or far advanced disease (Diagnostic Standards and Classification of Tuberculosis, National Tuberculosis and Respiratory Disease Association, 1969 classification system). All the data were retrospectively collected from medical records and are described in table S8. Patients from the diabetic cohort (DM, n = 220) were recruited at a specialist diabetic clinic at TTSH. The T-SPOT.TB (Oxford Immunotec), an IFN-γ release assay, and chest radiographs were performed on the patients of this DM cohort. The DM patients had no clinical pulmonary evidence of TB. Of 220 DM patients, 187 were treated with MET and 33 with other drugs (Non-MET). The clinical characteristics of these patients are described in table S9. MET-treated patients in both cohorts received 100 to 3000 mg of MET daily and were on diabetic therapy ranging from 2 months to 20 years. Both the retrospective data mining exercise (TB cohort) and the prospective clinical study (DM cohort) were done with ethical approval obtained from the National Healthcare Group Domain Specific Review Board (“C” Domain) (ethics approval codes C/2012/01177 and C/2006/00339, respectively).


Statistical analysis of data was performed using Prism (GraphPad) and Stata software packages. Two-tailed Student’s t test was used for analysis of in vitro assays. Mann-Whitney test was used for analysis of in vivo assays. One-way analysis of variance (ANOVA) with Tukey test was used for analysis of morphometric data. χ2 test was used for analysis of microscopy data and patient cohort data. The Kaplan-Meier method using log-rank test was used to analyze TB + DM patient survival. Multiple logistic regression analysis of the TB + DM cohort was performed using the Analysis of Overdispersed Data (AOD) package in R v2.15.2/Bioconductor. For the baseline analysis, use of MET and other drugs (Non-MET) was selected as the response variable, and CXR data (both cavity and classification), sex, age, smear positivity, culture positivity, cough indication, and smoking were used as predictors. The clinical outcome analysis used the morbidity as the response variable, and sex, age, use of MET or Non-MET, and CXR cavity information were used as the predictor variables. Data were fitted to a binomial (logistic) model. Multiple logistic regression analysis of DM cohort was carried out using Stata.


Fig. S1. Inhibition of the intracellular growth of mycobacteria by FDA-approved compounds.

Fig. S2. MET inhibits the intracellular growth of mycobacteria in an AMPK-dependent manner.

Fig. S3. MET does not have direct effect on Mtb.

Fig. S4. MET induces mROS but not cROS in the early phase of infection.

Fig. S5. Scavenging ROS inhibits MET-mediated mROS production and antimycobacterial effects.

Fig. S6. Autophagy is not required for MET-mediated inhibition of mycobacterial growth.

Fig. S7. MET does not induces neither cell death and cytochrome c release in mycobacterium-infected cells.

Fig. S8. MET enhances INH and ETH efficacy in a mouse model of infection.

Fig. S9. Effect of MET on infiltration of CD4+ and CD8+ T cells in the lungs of Mtb-infected mice.

Fig. S10. MET enhances CD8+ T cell responses in uninfected mice.

Fig. S11. MET modulates the genes associated with inflammatory responses during mycobacterial infection.

Table S1. List of FDA-approved compounds used, indicating the class they belong, their action, clinical indication they are prescribed for and their effect on intracellular M. bovis BCG survival.

Table S2. MET induces phagosome-lysosome fusion in M. bovis BCG-infected hMDM and THP-1 cells.

Table S3. Proportion of mice with no detectable CFUs during INH therapy with and without MET after 4 weeks of treatment (that is, day 35 after infection) in two different experiments.

Table S4. Differentially expressed genes between lung cells of untreated Mtb-infected mice versus uninfected mice.

Table S5. Differentially expressed genes between lung cells of MET-treated versus untreated Mtb-infected mice.

Table S6. Enrichment of significant canonical pathways (IPA analysis) in differentially expressed genes between lung cells of untreated Mtb-infected mice versus uninfected mice.

Table S7. Enrichment of significant canonical pathways (IPA analysis) in differentially expressed genes between lung cells of MET-treated Mtb-infected mice versus untreated Mtb-infected mice.

Table S8. Clinical characteristics of patients with comorbidity of TB and DM.

Table S9. Clinical characteristics of 220 diabetic (DM) patients.

Table S10. Characteristics of MDR strains of Mtb used in the study.

Table S11. P values for Fig. 1.

Table S12. P values for Fig. 2.

References (50, 51)


  1. Acknowledgments: We thank J. A. Tay and J. Y. Siew for technical assistance and the personnel of the Animal BSL3 laboratory for assistance. We thank E. G. Koh for assistance in confocal microscopy. We also thank N. McCarthy of Insight Editing London for revising the manuscript. This research was supported by Singapore Immunology Network A*STAR. Author contributions: A.S. conceived the idea, designed the study, performed the experiments, and analyzed the data. L.J., P.K., B.P., and N.K. performed the experiments. L.T. performed pathohistological and morphometric analysis. J.C. and M.P. performed bioinformatic analysis. F.Z. performed microarray studies. G.S.H., M.K.-S.L., C.C., and Y.T.W. performed retrospective clinical analyses. M.P. performed the statistical analysis. M.K.-S.L., F.Z., M.P., L.T., C.C., B.K., G.K., and Y.T.W. revised the manuscript. G.D.L. conceived the experiments and oversaw the study and data analysis. A.S. and G.D.L. wrote the manuscript. All authors discussed results and commented on the manuscript. Competing interests: A.S. and G.D.L. have filed the patent with respect to the use of MET and other drugs identified in this work for controlling mycobacterial infection, PCT/SG2013/000388. Other authors declare no competing financial interests. Data and materials availability: Gene Expression Omnibus accession number of microarray data is GSE57275.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article