We have previously shown that circulating levels of HO-1 are increased in pulmonary TB patients compared to uninfected individuals and those with LTBI in both adults and children (27, 28). In these studies, patients were recruited at the time of disease diagnosis and for this reason we did not have data on their history of TB infection. In order to temporally examine the induction of HO-1 following Mtb infection, we quantified its expression in three distinct animal models. Using a recently published model of lung cavitation in rabbits (33, 34), we observed that HO-1 protein expression increased in lung areas with granulomatous lesions compared to those with normal tissue (Figure 1A). HO-1 expression was further augmented in the cavity walls (Figure 1A). Notably, the level of HO-1 expression reflected bacterial burden in the different lung areas in infected animals (Figure 1A). Pulmonary infection of rabbits with Mtb was also associated with a significant increase in HO-1 plasma levels (Figure 1A). These results indicate that circulating levels of HO-1 reflect the induction of HO-1 protein expression at the infection site. In addition, they show that HO-1 expression correlates with the bacterial burden in different areas of the infected lung.

We next examined HO-1 expression in lungs and in serum in a mouse model of pulmonary TB (35), before and after a chemotherapy regimen that closely mimics that used in humans. In agreement with a recent study from our group (30), we found that HO-1 expression in both lungs and serum significantly increased following aerosol infection with Mtb (Figure 1B). Initiation of ATT reduced HO-1 induction, with concentrations substantially falling in the serum of mice receiving the antimicrobials compared to untreated animals (Figure 1B). Interestingly, although greatly reduced, significant levels of HO-1 expression remained relative to the levels displayed by uninfected control mice perhaps due to the presence of persistent bacteria.

Experimental infection of NHP has been described to induce a pulmonary disease that more accurately resembles human TB than that occurring in murine and rabbit models (36). We therefore measured HO-1 levels in plasma of cynomolgus macaques bronchoscopically infected with Mtb and treated with rifampin or isoniazid when active TB disease was established (37). In this NHP model, HO-1 levels significantly increased as early as 4 weeks postinfection and remained elevated following development of active TB (Figure 1C). Treatment with either rifampin or isoniazid was associated with gradual decreases in HO-1 concentrations (Figure 1C). At the end of therapy, treated animals exhibited HO-1 levels indistinguishable from those detected at preinfection. These findings in multiple animal models strongly support the association of HO-1 expression levels with TB infection and disease as well as the potential utility of the enzyme as biomarker of successful anti-TB therapy.

Our previous clinical data suggested that measuring HO-1 expression in patients undergoing ATT could potentially discriminate different treatment outcomes (27). To more rigorously test this hypothesis, we examined HO-1 concentrations in plasma from a well-characterized cohort of 130 South African patients with active pulmonary TB from whom serial samples were collected. HO-1 concentrations gradually diminished during treatment (Figure 2A). Notably, while levels of the enzyme consistently decreased in individuals who were successfully treated, they did not significantly change in patients who exhibited unfavorable outcomes (e.g., treatment failure or TB relapse at the end of therapy) (Figure 2B). Plasma concentrations of HO-1 were similar between patients who exhibited TB relapse or treatment failure during the study follow-up irrespective of the time point (Figure 2C), although the small number of patients in these subgroups of unfavorable outcomes preclude a definite conclusion. Similarly, prior to therapy, HO-1 levels were indistinguishable between patients who were successfully treated and those who were not (Figure 2B). Nevertheless, by week 20 of therapy, patients with unfavorable outcomes exhibited HO-1 levels substantially higher than those successfully treated (Figure 2B). These findings demonstrate that pretreatment levels of HO-1 cannot be used to predict the outcome of TB therapy, but the extent of HO-1 expression measured at the end of therapy does reflect outcome. To quantitatively assess the accuracy of HO-1 levels at week 20 of therapy in discriminating successful treatment from failure or relapse, we performed receiver operator characteristics (ROC) curve analysis. We confirmed that HO-1 levels at the end of anti-TB treatment exhibited high accuracy in classifying treatment outcomes (area under the curve: 0.81, sensitivity: 90%, specificity 67.4%, p = 0.003; Figure S1 in Supplementary Material).

Additional analysis revealed that within the entire study population, HO-1 plasma concentrations did not differ between patients presenting with cavitary or non-cavitary disease (Figure 2D). Patients with non-cavitary disease exhibited a gradual decrease in HO-1 levels upon treatment initiation, while those with cavitary disease displayed significant changes in HO-1 concentrations only at the end of therapy. Prior to therapy, circulating concentrations of HO-1 were not different between patients classified as having mild or extensive radiographic disease (Figure 2E). Moreover, treatment-induced reductions in HO-1 levels were independent of the radiographic extent (Figure 2E).

In our patient cohort, 75 (57.7%) patients were coinfected with HIV-1 (Table 1). Of those, 48 (64%) were antiretroviral therapy (ART) naïve. Of note, HO-1 concentrations in plasma were significantly higher in HIV-coinfected than in HIV-1-uninfected patients before treatment initiation (Figure 2F). Although HO-1 levels substantially decreased upon treatment initiation irrespective of HIV status, HIV-coinfected patients had higher values than those of uninfected subjects at the end of TB therapy (Figure 2F). These results suggest that HIV-1 coinfection may influence the regulation of HO-1 expression in TB patients. We next tested whether HIV-1-driven immunosuppression is associated with increased HO-1 expression in TB patients. In TB/HIV coinfected patients, plasma levels of HO-1 negatively correlated with CD4⁺ T-cell counts (Figure 3A) before anti-TB treatment. Individuals who were highly immunosuppressed (CD4⁺ T-cell count <100 cells/μL) exhibited increased HO-1 concentrations compared to those who were not (Figure 3B). Patients who were on ART at week 0 of ATT had lower HO-1 levels than those who were ART-naïve (Figure 3C). In addition, HO-1 levels positively correlated with HIV-1 viral load (Figure 3D). These findings suggest that HIV-1 coinfection causes further increases in HO-1 expression in TB patients.

The results presented above demonstrate that HO-1 is induced upon Mtb infection in different experimental models in vivo as well as in patients with active disease. To further delineate the molecular mechanisms to induce HO-1 in the context of Mtb infection, we employed an in vitro system using mouse and human macrophages infected with Mtb H37Rv strain (10). We chose macrophages because this cell type is known to be an important source of HO-1 in several disease models including human and murine mycobacterial infections (29). In other HO-1 induction systems, expression of the enzyme is thought to require translocation to the nucleus of the transcription factor known as nuclear factor erythroid-derived 2-like 2 (NRF-2) (14). We found that this is also true for HO-1 induction by Mtb infection as bone marrow-derived macrophages (BMDMs) from mice genetically lacking NRF-2 were unable to produce HO-1 following bacterial exposure (Figures 4A,B). In our in vitro system, the results obtained with HO-1 protein quantification reflected the degree of HMOX1 messenger RNA (mRNA) expression (Figure S2 in Supplementary Material).

Several signaling pathways have been described to be associated with HO-1 expression, including classical pattern-recognition pathways (38), interleukin-10 (IL-10) (39), and type I interferon (40). Remarkably, induction of both nuclear NRF-2 translocation and HO-1 protein expression was not affected in Mtb-infected BMDM from mice lacking MyD88, TRIF, IL-10, interferon alpha and beta receptor subunit 1 (IFNAR), or ferritin heavy chain (FTH1) (Figures 4C–L). These observations reveal that HO-1 induction driven by Mtb infection of macrophages involves a mechanism that is independent of numerous signaling pathways involved in molecular recognition of the pathogen and innate immune responses classically linked to TB pathogenesis.

We have recently described that patients with active pulmonary TB exhibit elevated circulating levels of oxidation products and lipid peroxidation compared with individuals with LTBI or uninfected controls, as well as in human macrophages infected with Mtb in vitro, demonstrating an important role for oxidative stress in this disease (11). In this study, we asked whether the oxidative stress triggered by Mtb infection might play a role in driving HO-1 expression. Indeed, treatment of Mtb-infected human monocyte-derived macrophage cultures with Tempol, a potent antioxidant, caused a substantial reduction in HO-1 protein expression (Figure 5A, left panel).

In unstimulated conditions, NRF-2 has been described to interact with the protein Kelch-like ECH-associated protein 1 (Keap-1) in the cytosol of macrophages and other cell types. This interaction induces proteasome-mediated degradation of the former molecule. Upon stimulation, ROS is thought to interfere with the NRF-2/Keap-1 interaction, releasing the transcription factor, which then translocates to the nucleus to induce HO-1 (41, 42). To test if the ROS requirement for HO-1 induction in our in vitro system involves a similar pathway, we incubated cultures of infected human monocyte-derived macrophages treated with Tempol in the presence of TAT-14, a peptide comprising a TAT-conjugated NRF-2 sequence known to interfere with the NRF-2/Keap-1 interaction (43). Notably, HO-1 induction was restored when TAT-14 was added to the cultures of infected cells treated with Tempol (Figure 5A, right panel), without altering ROS production (Figure 5B). The restoration of HO-1 induction upon Mtb infection was reflected by increased activation of NRF-2 in nuclear extracts (Figure 5C). These findings demonstrate that ROS-mediated interference of NRF-2/Keap-1 interaction is a major pathway involved in the induction of HO-1 by Mtb in murine macrophages.

The NADPH oxidase system is a major source of ROS in both physiological and pathological conditions (44). To test if NADPH-derived ROS plays a role in Mtb-driven HO-1 induction, we infected BMDM from mice deficient in either one of the two main NADPH-oxidase subunits (p47Phox and gp91Phox), which lack fully functional NADPH-oxidase activity (45, 46). Importantly, Mtb-induced HO-1 expression was markedly diminished in the genetically deficient mice compared with the wild-type (WT) animals (Figure 5D). In these experiments, the p47Phox^−/− mice and their respective WT controls were from C57BL/6NTac genetic background, whereas the gp91Phox^−/− animals were from the C57BL/6J background. The latter genetic background has been associated with a spontaneous mutation that results in mitochondrial redox abnormalities (47), which could have influenced the HO-1 production in vitro. However, the pattern of HO-1 production in the experimental conditions was very similar to that observed in mice from the C57BL/6NTac genetic background. In addition, as observed with Tempol-treated human macrophages (Figure 5A), TAT-14 addition restored both HO-1 expression and NRF-2 activation (Figures 5E,F, respectively) to the NADPH oxidase-deficient macrophage cultures from both genetic backgrounds. These results support the hypothesis that Mtb drives HO-1 expression in infected macrophages by inducing NADPH-derived ROS, which inhibits NRF2/Keap1 interaction.

Many pathological effects observed in cells infected with Mtb are related to bacterial production of the virulence factor ESAT-6, a protein that is secreted into the host cell cytosol (48). We have previously reported that Mtb lacking ESAT-6 failed to induce robust HO-1 expression in both human and murine macrophages (10). Experiments were next performed to determine whether the role of ESAT-6 in HO-1 induction involves the oxidative stress-dependent pathway described above. We found that infection of human macrophages with an H37Rv mutant lacking ESAT-6 was associated with a decreased extracellular and intracellular ROS response (Figure 6A). This reduction in ROS production was linked to reduced concentrations of malondialdehyde (MDA), a marker of lipid peroxidation, and of 8-hydroxy-2′-deoxyguanosine (8-OH-dG), a readout of oxidative DNA damage (Figure 6B). We next delivered recombinant ESAT-6 protein fused with the N-terminal fragment of the lethal factor of Bacillus anthracis (10, 49) into the cytosol of macrophages infected with ESAT-6-deficient Mtb. Notably, delivery of ESAT-6 into macrophages restored ROS production (Figure 6A) and was further associated with augmented lipid peroxidation and oxidative DNA damage (Figure 6B), at levels comparable to that induced by the control WT H37Rv strain. As expected, the decreased HO-1 expression observed in macrophages infected with ESAT-6-deficient Mtb was associated with reduced NRF-2 activation in nuclear extracts (Figure 6C), and these defects were corrected when macrophage cultures were treated with TAT-14 (Figure 6C). Together, these findings support a link between ESAT-6 expression and ROS-induced activation of NRF-2 in the triggering of HO-1 production by Mtb in human macrophages.

Previous studies in both murine BMDMs and RAW264.7 cells found that exposure to soluble recombinant ESAT-6 protein triggers HO-1 expression via induction of the microRNA 155 (miR-155) (50). We therefore tested the participation of miR-155 in our in vitro system by assessing NRF-2 nuclear translocation and HO-1 production in miR-155KO murine macrophages (Figure S3 in Supplementary Material). Our experiments revealed that Mtb-infected BMDMs from miR-155KO mice failed to induce HO-1 production while exhibiting NRF-2 quantification in nuclear extracts similar to that observed in WT cells. Such diminished HO-1 production detected in cultures from miR-155KO cells was not reverted by treatment with TAT-14 (Figure S3 in Supplementary Material). These findings suggest that miR-155 is critical to induce HO-1 production in Mtb-infected macrophages by a distinct pathway that does not involve interference with NRF-2/Keap-1 interaction.

Patients with GeneXpert Mtb/RIF-confirmed rifampin-susceptible pulmonary TB were recruited at the Ubuntu HIV/TB Clinic, Site B, Khayelitsha, South Africa, as part of a prospective cohort study. Detailed sociodemographic data, past TB treatment history, and comorbidity data were collected. At the study baseline, HIV testing (Abbott Architect HIV Ag/Ab Combo test), CD4⁺ T lymphocyte count, and HIV-1 viral load quantification were performed. Extensive disease on chest X-ray was performed by experts and noted as either involvement of >1 lung or involvement of ≥1 out of 3 (upper, mid, or lower) zones per lung. Antitubercular drugs were provided in a 4-drug fixed-dose combination supplied by the National Tuberculosis Control Programme [Rifafour e-275 (Sanofi-Aventis) or Ritib (Aspen, South Africa)]. Each tablet contained rifampin at 150 mg, isoniazid at 75 mg, pyrazinamide at 400 mg, and ethambutol at 275 mg. Weight band-based dosing was used in line with WHO guidelines (70). Patients weighing 38–55, >55–70, and >70 kg were given doses of 3, 4, and 5 tablets, respectively. Anti-TB drugs were administered 7 days/week, along with 25 mg pyridoxine. Clinical care remained the responsibility of the Site B TB clinic. Unfavorable treatment outcomes were defined as failure or relapse. Failure was any TB patient who was culture positive at 5 months or more after starting treatment. Relapse was defined as a TB patient who was declared cured or treatment completed by a physician, but who reports back to the health service and is now found to be sputum culture positive. EDTA plasma samples were collected before initiation of ATT as well as at month 2 and 5 of treatment. Samples were frozen at −80°C until use in the immunoassays.

The lung and plasma samples employed were from infected animals described in previously published papers (33, 34). In these studies, female New Zealand white rabbits, weighing 3–3.5 kg (Covance Research Products, Gaithersburg, MD, USA), were sensitized with five 0.2 ml subcutaneous injections of 1 × 10⁷ heat-killed Mycobacterium bovis, emulsified 1:1 in Freund’s adjuvant. Twenty-one days after sensitization, skin test reactivity was determined by injection of 5 IU of purified protein derivative (Tubersol; Sanofi-Aventis, Bridgewater, NJ, USA) and measured at 48 h. For infection, rabbits were anesthetized by intramuscular administration of ketamine (10 mg/kg) and xylazine (20 mg/kg) prior to endotracheal tube intubation. A 3.0-mm flexible Pentax FB-8V pediatric bronchoscope (Pentax, Montvale, NJ, USA) was guided into 1 lower lobe, and 400 µL of 7.5 × 10³ log-phase Mtb strain H37Rv was inoculated via catheter. At day 42 postinfection, EDTA plasma specimens were collected from the central ear artery. Rabbits were euthanized following anesthesia by intravenous injection of Euthasol (Virbac, Fort Worth, TX, USA). Different lung regions presenting with cavitation (cavity wall), granulomatous lesions, or normal lung tissue were dissected for protein analysis and snap frozen in liquid nitrogen as previously described (33, 34). In addition, 3 mm punch biopsies (six/tissue/animal) from each dissected lung region were weighed and then homogenized in phosphate-buffered saline (PBS) prior to Mtb CFU enumeration on 7H11 selective medium (33). The remaining lung lobe (>90% by mass) was weighed and then homogenized, using a Polytron homogenizer (Kinematica, CH), in 40 mL PBS prior to CFU enumeration by serial dilution on selective 7H11 medium.

C57BL/6J mice between 6 and 8 weeks of age were obtained from National Institute of Allergy and Infectious Diseases (NIAID) through a supply contract with Taconic Farms (Germantown, NY, USA). All mice were bred and housed at Association for the Assessment and Accreditation of Laboratory Animal Care accredited Biosafety level 2 and 3 facilities at the NIAID, NIH, following all the National Research Council Guide for the Care and Use of Laboratory Animals guidelines. All experiments using mice were approved by the NIAID Animal Care and Use Committee. Mice were infected with approximately 100 CFU of the H37Rv strain of Mtb using an aerosol chamber (Glas Col, Terre Haute, IN, USA).

To evaluate the effects of chemotherapy, mice infected for 28 days were treated with a cocktail containing the antibiotics rifampin (R) (10 mg/kg/mouse), isoniazid (H) (25 mg/kg/mouse), and pyrazinamide (Z) (150 mg/kg/mouse) (all from Sigma-Aldrich). After 60 days, treatment was changed to a cocktail consisting of RH alone. The drug stocks were prepared weekly and were orally administered by gavage 5 days per week. At the designated time points, mice were euthanized, blood was collected, and lungs were extracted and homogenized in PBS and stored at −80°C for later analysis.

The donors were WT B6.SJL (CD45.1/1) mice or animals genetically deficient in the HO-1 transcription factor NRF-2 (NRF2^−/−), myeloid differentiation primary response gene 88 (MyD88^−/−), TIR-domain-containing adapter-inducing interferon-β (TRIF^−/−), IL-10^−/−, IFNAR^−/−, FTH1^−/−. Cells from p47phox^−/− or gp91phox^−/− (two distinct subunits of the NADPH oxidase system) mice, which lack a fully functional NADPH-oxidase system, were also used. MyD88^−/−, TRIF^−/−, IL10^−/−, and IFNAR^−/− together with their respective WT controls were obtained from Taconic and p47phox^−/− (C57BL/6NTac background, Taconic) and gp91phox^−/− (C57BL/6J background, Jackson Lab) mice, were provided by Dr. Sharon Jackson (NIMHD, NIH). Bone marrow from FTH1^−/− and NRF2^−/− mice were provided by Dr. Miguel Soares (Instituto Gulbenkian de Ciências, Oeiras, Portugal). Bone marrow from miR-155KO mice were a gift from Dr. Stefan Muljo (NIAID, NIH). All mice were backcrossed for at least eight generations on a C57BL/6J or N background (indicated when exact genetic background was referred to original references or by investigator donating the animals for our use).

The plasma samples analyzed were from animals utilized in a previously published study (37). In brief, 10 cynomolgus macaques (Macacca fascicularis) were infected with Mtb (Erdman strain) via bronchoscope, using an inoculum of 40–500 CFU (Table S1 in Supplementary Material). When the monkeys developed active TB (between 54 and 173 days postinfection; see Table S1 in Supplementary Material), they were treated daily with orally delivered isoniazid (N = 4) (15 mg/kg/dose) or rifampin (N = 6) (20 mg/kg/dose). Prior to treatment initiation and after 2 months of treatment, each macaque was given a PET/CT scan, using ¹⁸F-2 fluoro-deoxyglucose (FDG) as a probe, and animals were matched for disease pattern in each treatment group. Change in FDG avidity in lungs over treatment was reported in Ref. (37). Plasma samples obtained from these animals were stored at −80°C prior to analysis.

CD14⁺ column-purified human elutriated monocytes were obtained from peripheral blood of healthy donors from the NIH blood bank under Institutional Review Board-approved protocols of both the NIAID and the Department of Transfusion Medicine. Macrophages were generated by culturing monocytes in the presence of RPMI 1640 media containing 10% human AB serum and M-CSF 50 ng/mL (PeproTech, Rocky Hill, NJ, USA) for 7 days; fresh media with growth factor were added every 48 h, as previously described (71). This method of macrophage differentiation was chosen based on a recently published guideline (72). For infection experiments, cells were plated at a density of 10⁶ cells/well in 24-well plates in phenol and serum-free media (Opti-MEM; Life Technologies, Carlsbad, CA, USA).

Murine BMDMs were employed in a second group of experiments. Bone marrow cells were cultured in 30% L929 cell-conditioned medium for 7 days. An additional 10 mL of L929 cell-conditioned medium were added after 4 days of incubation. The resulting adherent macrophages were detached with cold PBS and seeded in 96-well plates at 10⁵ cells/well, containing serum and phenol-free medium at 37°C in 5% CO₂.

Cells were exposed to WT H37Rv or an ESAT-6-deficient H37Rv mutant (a gift from Dr. Volker Briken, University of Maryland, College Park, MD, USA) Mtb strain at the indicated multiplicity of infection for 3 h, washed to remove extracellular bacteria, and cultured in serum-free media for 24 h in the presence or absence of the indicated concentrations of Tempol (4-hydroxy-Tempo; 4-hydroxy-2,2,6,6-tetramethylpi-peridine-N-oxyl, Sigma-Aldrich) or a TAT-conjugated NRF-2 peptide (TAT-14) that interacts with the Keap-1/NRF-2 complex (TOCRIS, Bristol, UK). In some experiments, recombinant ESAT-6 was delivered into the cytosol of infected macrophages using a fusion protein with the N-terminal fragment of the lethal factor of B. anthracis as previously described (10). The anthrax-protective Ag used in this procedure was also prepared as described previously (73). The two proteins at the indicated doses were added to cell cultures 3 h after Mtb infection and incubated for 24 h. Cobalt (III) pro-toporphyrin IX dichloride (CoPPIX, Frontiers Scientific, Logan, UT, USA) was used as a positive control to induce HMOX1 mRNA expression in macrophages (Figure S2 in Supplementary Material). Culture supernatants were then collected, sterile filtered, and stored at −80°C until use. Whole cell extracts were prepared following the manufacturer’s instructions for the HO-1 ELISA kit (Enzo Life Sciences). Nuclear extracts were obtained using a kit from Active Motif (Carlsbad, CA, USA).

Rabbit and mouse lungs were perfused with PBS and homogenized in PBS containing Complete Ultra protease inhibitor cocktail (Roche, Basel, Switzerland) and 2 mM phenylmethylsulfonyl fluoride (Sigma-Aldrich). Rabbit HO-1 protein expression levels were quantified in plasma and lung homogenates (after normalization of their protein concentrations) using an ELISA kit (CUSABIO, College Park, MD, USA). HO-1 levels were measured in serum and lung homogenates of mice by ELISA (ADI-960-071, Enzo Life Sciences, Farmingdale, NY, USA) and in EDTA plasma samples from NHP using a human ELISA kit (ADI-EKS-800, Enzo Life Sciences). The same kit was used to measure HO-1 levels in plasma samples from TB patients as well as in whole cell extracts of the monocyte-derived macrophages studied in vitro.

Protein levels of the transcription factor NRF-2 in nuclear extracts from mouse macrophages were measured by ELISA (CUSABIO). NRF-2 activation was determined in nuclear extracts using TransAM DNA-binding ELISA kits (Active Motif, Carlsbad, CA, USA). Lipid peroxidation in culture supernatants was quantified using a kit (Cayman Chemical, Ann Harbor, MI, USA), which measures the formation of MDA. DNA/RNA oxidation was quantitated in culture supernatants using an immunoassay from Cayman Chemical that detects all three oxidized guanine species: 8-hydroxy-2′-deoxyguanosine from DNA, 8-hydroxyguanosine from RNA, and 8-hydroxyguanine from either DNA or RNA. Results from this assay were expressed as the concentration of 8-OH-DG in each sample. Intracellular production of ROS in macrophage cultures was assessed by staining cells with the oxidative fluorescent dye probe, dihydroethidium (DHE) 5 mM (Invitrogen/Molecular Probes, Grand Island, NY, USA) for 30 min at 37°C in5% CO₂, and then analyzed by flow cytometry, as described previously (11). The cells used for ROS assay had been previously detached from the culture plates using trypsin (0.25%), washed, and then re-suspended in phenol and serum-free medium. Results were plotted as the mean fluorescence intensity of each of the experimental groups. Extracellular superoxide production was quantified as previously described (74) by adding hydroxylamine (0.5 mM) during cell culture, which converts superoxide into nitrite plus nitrate, which were reduced by VCl₃ treatment and quantified by Griess reaction (Sigma-Aldrich).

Total RNA was isolated from murine macrophages using the RNeasy Mini Kit, and residual DNA was digested using RNase-free DNase (both from QIAGEN, Valencia, CA, USA). The RNA samples were reverse transcribed using SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA). Gene expression was measured using SYBR Green-based real-time quantitative PCR, and 18S mRNA was used as the housekeeping gene. The following oligonucleotide primers were used: 18S, forward, 5′-CACGGCCGGTACAGTGAAAC-3′ and reverse, 5′-CCCGTCGGCATGTATTAGCT-3′; and HMOX1, forward, 5′-TCTCAGGGGGTCAGGTC-3′ and reverse, 5′-GGAGCGGTGTCTGGGATG-3′. Fold induction of HMOX1 gene expression was calculated using the DD threshold cycle method, normalizing mRNA levels for each sample to levels of 18S and comparing with mRNA levels in unstimulated cells.

The median values with interquartile ranges were used as measures of central tendency. For the in vitro experiments, bars represent mean and SD. The Mann–Whitney U test (for two groups) or the Kruskal–Wallis test with the Dunn’s multiple-comparison or non-parametric linear trend post hoc test (for more than two groups) were used to compare continuous variables in unmatched groups, whereas the Wilcoxon (for two groups) or Friedman (for more than two groups) matched-pairs tests were used in matched analysis, as indicated in table notes and figure legends. The Fisher’s exact test or chi-square was used to compare variables displayed as percentages. Spearman rank tests were used to assess correlations. ROC curves were used to test the power of HO-1 levels to distinguish cure (successful treatment) from treatment failure or TB relapse at week 20 of ATT. All comparisons were prespecified and two-tailed. Differences with p values below 0.05 after Holm’s adjustment for multiple comparisons were considered statistically significant. The statistical analyses were performed using GraphPad Prism 7.0 (GraphPad Software, La Jolla, CA, USA), STATA 11 (StataCorp, College Station, TX, USA), JMP 13.0 (SAS, Cary, NC, USA), and R 3.1.0 (R Foundation, Vienna, Austria) programs.