GM-CSF targeted immunomodulation affects host response to M. tuberculosis infection

Host directed immunomodulation represents potential new adjuvant therapies in infectious diseases such as tuberculosis. Major cytokines like TNFα exert a multifold role in host control of mycobacterial infections. GM-CSF and its receptor are over-expressed during acute M. tuberculosis infection and we asked how GM-CSF neutralization might affect host response, both in immunocompetent and in immunocompromised TNFα-deficient mice. GM-CSF neutralizing antibodies, at a dose effectively preventing acute lung inflammation, did not affect M. tuberculosis bacterial burden, but increased the number of granuloma in wild-type mice. We next assessed whether GM-CSF neutralization might affect the control of M. tuberculosis by isoniazid/rifampicin chemotherapy. GM-CSF neutralization compromised the bacterial control under sub-optimal isoniazid/rifampicin treatment in TNFα-deficient mice, leading to exacerbated lung inflammation with necrotic granulomatous structures and high numbers of intracellular M. tuberculosis bacilli. In vitro, GM-CSF neutralization promoted M2 anti-inflammatory phenotype in M. bovis BCG infected macrophages, with reduced mycobactericidal NO production and higher intracellular M. bovis BCG burden. Thus, GM-CSF pathway overexpression during acute M. tuberculosis infection contributes to an efficient M1 response, and interfering with GM-CSF pathway in the course of infection may impair the host inflammatory response against M. tuberculosis.

is essential for granuloma formation and maintenance, and experimental host directed adjunctive therapies combining TNFα neutralization with chemotherapy reported enhanced M. tuberculosis bacterial clearance [13][14][15] . However, neutralizing TNF is a risky measure to take in M. tuberculosis infected individuals. Here we hypothesized that GM-CSF immunomodulation could contribute to host defense against M. tuberculosis.
Indeed, a new, pivotal role of GM-CSF at the T cell -myeloid cell interface has recently emerged, and targeting the crosstalk between T cells and myeloid cells through GM-CSF seems promising in immunopathology and chronic tissue inflammation 16 . Beside its role in homeostasis, GM-CSF is considered as a key mediator in tissue inflammation, notably in chronic obstructive pulmonary disease (COPD), allergy and asthma. Neutralization of GM-CSF or its receptors are in development for indications including cancer or severe inflammatory diseases such as rheumatoid arthritis 17,18 . However, the role of GM-CSF during bacterial infections like TB is poorly understood. GM-CSF is induced after in vitro M. tuberculosis infection of THP-1 cells via PI3-K/MEK1/p38 MAPK signaling pathway 19 . Additionally, GM-CSF contributes to proinflammatory macrophage M1 polarization 20 . Indeed, GM-CSF induces monocyte and macrophage production of IL-6, IL-8, G-CSF, M-CSF, TNFα and IL-1β, although less potently than LPS 20 . On the other hand, GM-CSF seemed to control TNFα and IFNγ expression, since lung cells from GM-CSF deficient mice produced less TNFα and IFNγ after exposure to M. tuberculosis culture filtrate proteins than wild-type cells 21 .
It was proposed that GM-CSF controls granuloma formation in M. tuberculosis infection since GM-CSF deficient mice exhibited granuloma disruption 22 . However, while GM-CSF is essential for maintaining the ability of pulmonary alveolar macrophages to clear surfactant lipids and proteins from the lung surface, as evidenced in GM-CSF deficient mice 23 , it is uncertain whether GM-CSF has a direct effect on granuloma formation. Specific GM-CSF expression either in lung epithelial cells or in the hematopoietic compartment partially restored the control of the M. tuberculosis infection by GM-CSF deficient mice 22,24 .
Here, we show that GM-CSF and its receptor are overexpressed in the lung during acute M. tuberculosis infection in vivo and we tested how neutralizing pharmacologically GM-CSF host response might affect lung inflammation and granuloma formation, and interfere with the action of antibiotic treatment in M. tuberculosis infection. GM-CSF neutralizing antibodies that efficiently prevented sterile lung inflammation had no significant effect on bacterial burden and host response to M. tuberculosis infection in wild-type mice, while promoting granuloma formation. We then tested GM-CSF neutralization under sub-optimal antibiotic chemotherapy in TNFα-deficient mice. We propose that GM-CSF overexpression during acute M. tuberculosis infection contributes to an efficient M1 response, and neutralizing GM-CSF might be detrimental to the host inflammatory response to M. tuberculosis infection.

Increased GM-CSF expression after in vivo airway M. tuberculosis infection. We first verified that
GM-CSF is regulated in the airways in vivo after inflammatory challenge or M. tuberculosis infection. Indeed, GM-CSF levels are increased in both bronchoalveolar lavage fluid (BAL) and lungs 24 h after LPS exposure (1 µg LPS i.t. or i.n.; Fig. 1A,B), as expected 25 , and in the lungs 1 month post M. tuberculosis infection in WT mice, although this was transient and essentially no GM-CSF was detected 3 months post-infection (Fig. 1C). Interestingly, there was also a strong upregulation of GM-CSFR β subunit genes (Csf2rb and Csf2rb2) in the lungs 28 days post-infection (Fig. 1E,F) in WT mice. Immunocompromised individuals are extremely susceptible to mycobacterial infections, and we examined here the response of TNFα-deficient mice as a model of immunodeficient mice. The production of GM-CSF and the upregulation GM-CSFR β subunit gene Csf2rb2 were exacerbated in the lung of highly susceptible TNFα-deficient mice, 1 month post-M. tuberculosis infection (Fig. 1D,F). Thus, both GM-CSF and its receptor are over-expressed during acute M. tuberculosis infection.

GM-CSF blockade inhibits cell recruitment after acute lung inflammation. Tuberculous granuloma
comprising macrophages, dendritic cells, lymphocytes and neutrophils are keys for the host response to M. tuberculosis 26,27 . We next assessed the contribution of GM-CSF in inflammatory cell recruitment to the lungs, by using anti-murine GM-CSF neutralizing antibodies. We first verified that anti GM-CSF neutralizing antibodies impairs acute lung inflammation and inflammatory cell recruitment in the airways after LPS exposure (1 µg/mouse i.n. and i.t). Administration of anti GM-CSF antibodies (200 µg per mouse i.p.) strongly reduced neutrophil recruitment in the bronchoalveolar lavage (BAL; Fig. 2A), as reported 28,29 , while lymphocytes were slightly increased and macrophage counts were not affected (Fig. 2B,C). GM-CSF neutralization also reduced lung CXCL1, TNFα, IL-1β and MPO response to LPS exposure ( Fig. 2D-G), together with the markers of tissue inflammation MMP9 and TIMP1 in BAL (Fig. 2H,I). Histologically, GM-CSF neutralization impaired inflammatory cell recruitment and reduced emphysema after airway inflammatory challenge (Fig. 2J). Thus, systemic GM-CSF neutralization prevented acute lung inflammation, proinflammatory cytokine and chemokine release, and neutrophil recruitment in the airways.
Limited effect of GM-CSF neutralization on acute M. tuberculosis infection. In view of the dual role of the inflammatory response in host control of tuberculosis, both deleterious but also indispensable for bacterial control 12 , we next asked how GM-CSF neutralization might affect the host response to M. tuberculosis. We addressed this question in wild-type C57Bl/6 mice, largely resistant to M. tuberculosis infection alike humans, and in susceptible TNFα −/− mice 30 , more representative of immunocompromised conditions. GM-CSF neutralizing monoclonal antibodies (MAB) or IgG2b isotype control were administered (200 µg i.p.) on day 14 and 21 after infection with M. tuberculosis H37Rv (1000 ± 200 CFU/mouse i.n.). After a transient decrease in bodyweight following anti GM-CSF MAB administration, the susceptible TNFα −/− mice showed a strong deterioration of their general status and had to be killed by day 26 ± 2, while WT mice were less affected (Fig. 3A,C). Indeed, lung bacterial loads were controlled in WT mice and were not affected by anti GM-CSF administration, as seen on day Effect of GM-CSF neutralization on M. tuberculosis induced lung inflammation. We next assessed the effect of GM-CSF neutralization on the lung inflammatory response, cell recruitment, and granuloma formation one month after M. tuberculosis infection. Macroscopically, WT mice receiving IgG2b isotype control showed few nodules that appeared larger in WT mice receiving αGM-CSF antibody (Fig. 4A). TNFα −/− mice receiving isotype control displayed many diffuse nodules that were exacerbated in TNFα −/− mice treated with αGM-CSF antibody (Fig. 4B). Microscopically, the number of granuloma present in the lung of WT mice was increased after GM-CSF neutralization, as compared to IgG2b isotype treated controls (Fig. 4A). TNFα deficient mice developed large confluent granulomatous lesions, associated with necrotic areas, which were increased after anti GM-CSF MAB treatment (Fig. 4B). Furthermore, Ziehl-Neelsen coloration of acid-alcohol resistant M. tuberculosis bacilli, detected very few M. tuberculosis bacilli in WT mice receiving IgG2b isotype control, that seemed more frequent in WT mice treated with αGM-CSF antibodies (Fig. 4A). TNFα deficient mice receiving isotype control harbored more M. tuberculosis bacilli in lung tissue as compared to WT, and the number of bacilli in the lesions of TNFα −/− mice further increased after treatment with αGM-CSF antibodies, as estimated from the Ziehl-Neelsen staining (Fig. 4B). Although the overall lung inflammation, evaluated using relative lung weight as a surrogate marker, was slightly reduced in WT mice after anti GM-CSF MAB administration, this was not observed in TNFα −/− mice ( Supplementary Fig. S1). Spleen and liver relative weights were also unaffected in WT mice receiving isotype control or αGM-CSF antibodies while they were increased in susceptible TNFα −/− mice after αGM-CSF antibody treatment as compared to isotype control. GM-CSF neutralization had no effect on the lung levels of IL-12p40 and IFNγ, two prominent cytokines involved in host control after M. tuberculosis infection, both in WT and TNFα deficient mice ( Supplementary Fig. S1). Thus, GM-CSF neutralization increased the extent of granulomatous lesions and their bacilli content after M. tuberculosis infection.  Fig. 5C). When using a sub-optimal dose of 10 mg/L of both INH/RIF that only partially reduced the bacterial load in the TNFα −/− mice (7 log 10 CFU/lung, Fig. 5B), TNFα −/− mice presented a body weight loss at day 24 post-infection (Fig. 5D) and the co-administration of anti GM-CSF MAB led to a significant increase in body weight loss (Fig. 5D) and bacterial load (Fig. 5E), as compared with isotype-treated TNFα −/− mice, on day 32 post-infection. The lung levels of IL12p40 (Fig. 5F) and IFNγ (Fig. 5G) were not affected by anti GM-CSF antibodies, as compared with isotype control, in TNFα −/− mice treated with INH/RIF 10 mg/L. Histologically, the sub-optimal antibiotic treatment limited the necrosis in the lung of TNFα −/− mice (Fig. 6A,C), as compared to untreated TNFα −/− mice (Fig. 4B). Co-administration of anti GM-CSF antibodies in antibiotic treated TNFα −/− mice yielded a massive infiltration with larger nodule formation and extended necrosis, together with a trend towards decreased free alveolar space and increased granulomatous lesions (Fig. 6A,C). Very interestingly, clusters of M. tuberculosis bacilli were visible in TNFα −/− mice treated with INH/RIF and their number increased after anti GM-CSF antibodies treatment, as compared with isotype control (Fig. 6B). We then verified the effect of anti GM-CSF treatment under a more effective antibacterial chemotherapy ( Supplementary Fig. S2). When using a 20 mg/L INH/RIF regimen that controlled M. tuberculosis infection and yielded a strong bacterial load reduction in TNFα −/− mice (down to 2 log 10 CFU/lung), GM-CSF neutralization did not impair bacterial control, nor pulmonary IL12p40 and IFNγ levels. Lung integrity was preserved under the 20 mg/L antibiotics dose, and anti GM-CSF antibody treatment did not induce morphological differences 32 days post-infection, although macroscopically larger nodules were observed. Thus, in susceptible TNFα −/− mice GM-CSF neutralization compromised the bacterial control by sub-optimal antibiotic treatment, leading to exacerbated lung inflammation, necrosis and intracellular bacterial load, while it did not affect host response under more effective antibacterial chemotherapy.  (Fig. 7A-D). Further, the overexpression of the M1 marker Nos2 induced by LPS stimulation or M. bovis BCG infection, was reduced after GM-CSF neutralization (Fig. 7E). Conversely, GM-CSF neutralization yielded the overexpression of all M2-related genes, including Arg1 and Ym1 that are downregulated after LPS stimulation and M. bovis BCG infection, and including also CD206 coding gene Mrc1 and the regulatory gene Socs3, that  or isotype control. Anti GM-CSF MAB treatment abolished the strong nitric oxide production by macrophages after M. bovis BCG-GFP in vitro infection, as compared to isotype control (Fig. 8A). Pro-inflammatory TNFα and IL-12p40 were also reduced in infected macrophages treated with anti GM-CSF MAB (Fig. 8B,C), while anti-inflammatory IL-10 was increased after GM-CSF neutralization (Fig. 8D). Interestingly, macrophages treated with anti GM-CSF MAB displayed an increased number of intracellular mycobacteria as compared to macrophages treated with isotype control (Fig. 8E-G). Thus, in vitro GM-CSF neutralization favors an anti-inflammatory profile in macrophages, abrogating the release of mycobactericidal nitric oxide, and leading to an increased intracellular presence of M. bovis BCG-GFP.

Impact of in vivo GM-CSF blockade on macrophage polarization after M. tuberculosis infection.
We next investigated the impact of GM-CSF neutralization on M1/M2 polarization in vivo. In a typical pulmonary M1 context after in vivo LPS administration (1 µg LPS i.t.), the upregulation of M1 related genes Nos2 and Il12b was reduced in the lung after GM-CSF neutralization (Supplementary Fig. S3). In contrast, M2 expression markers Arg1 and Il10 were not affected by GM-CSF neutralizing antibodies. After in vivo M. tuberculosis infection, Nos2 was overexpressed, while Arg1 expression was unchanged in WT and TNFα −/− mice treated with isotype control (Supplementary Fig. S3). In this context, GM-CSF neutralization did not affect the M1 or M2 macrophages related genes Nos2 and Arg1 in M. tuberculosis infected WT or TNFα −/− mice, nor did it affect M2 related genes Mrc1, Ym1 and Socs3 in TNFα −/− mice. Thus, GM-CSF neutralization reduced M1 polarization in an acute LPS-induced lung inflammation, but it did not affect M. tuberculosis induced M1 phenotype in vivo.

Discussion
Although the contribution of GM-CSF in granuloma formation after M. tuberculosis infection based on granuloma disruption in GM-CSF deficient mice 22 raised questions due to the defect of alveolar macrophage maturation and surfactant metabolism in these mice 23 , the role of GM-CSF in the control of M. tuberculosis infection is triggering new interest. Indeed, among patients with high anti GM-CSF autoantibody titers that had associated cryptococcal meningitis, one out of seven also developed M. tuberculosis infection 34 .
We document here that both GM-CSF and its receptor are overexpressed during acute M. tuberculosis infection, especially in highly susceptible, immunocompromised TNFα deficient mice. These results are in line with GM-CSF expression during M. tuberculosis infection of immunocompetent mice 24,35 . Beside cellular sources of GM-CSF such as alveolar macrophages and type II epithelial cells 22 , iNKT and γδ T cells can sustain early host response to M. tuberculosis in the absence of IFNγ 36 , and CD4 + T cells that take over later are sufficient to confer protection in a GM-CSF-depleted non-hematopoietic environment 24 . GM-CSF induces M. tuberculosis killing by macrophages, an activity requiring PPARγ expression 24 , however the exact mechanism for this effect is still unknown.
Given the double-edge role of the inflammatory response in host control of mycobacterial infection, indispensable to ensure bacterial control but potentially deleterious, we asked how GM-CSF neutralization might affect host response to M. tuberculosis. GM-CSF neutralizing antibodies, used at a dose that effectively prevents LPS-induced lung inflammation, exacerbated granuloma formation after M. tuberculosis infection and increased the number of granulomatous lesions containing large clusters of M. tuberculosis in the lung tissue of TNFα deficient mice, although this did not translate into overall increased numbers of CFU per lung.
Adjuvant immunotherapies under discussion for tuberculosis 11,12 include the use of anti TNFα antibodies to alter granuloma structure and favor the access of antibiotics to the mycobacteria within the granulomatous structure 13 . Here, we thus assessed whether GM-CSF neutralization might affect M. tuberculosis control by sub-optimal chemotherapy in susceptible TNFα-deficient mice. Sub-optimal isoniazid (INH) and rifampicin (RIF) bi-therapy partially contained the infection in terms of CFU titer per lung, number of M. tuberculosis clusters and the number of M. tuberculosis per cluster, and reduced necrosis in the lung tissue. GM-CSF neutralization actually compromised the bacterial control by sub-optimal INH/RIF treatment in TNFα −/− mice, leading to extended necrotic granulomatous lesions. However, under a higher dose chemotherapy that effectively reduced pulmonary bacterial load, the anti GM-CSF treatment did not compromise bacterial control. There was a transient body weight loss after GM-CSF neutralization in infected mice. Cachexia may result from systemic 'cytokine storm' as seen during a systemic inflammatory response syndrome. Indeed body weight loss did not seem an effect of the anti-GM-CSF MAB administration itself but occurred in conjunction with M. tuberculosis infection, as there was no body weight loss in the sterile LPS-induced lung inflammation model, nor in infected mice under INH/RIF antibiotics chemotherapy.
Macrophage polarization during M. tuberculosis infection is still a matter of debate. Indeed, M1 macrophages seem to play a key role in bacterial killing and granuloma formation while M2 macrophages were found in non-granulomatous lung tissues, inhibiting M1 macrophage effects 37 . Therefore, in tuberculosis a strong regulation of M1/M2 macrophages balance is essential for the host defense. GM-CSF was recently recognized to induce M1 polarization, culminating in monocyte and macrophage IL-6, IL-8, G-CSF, M-CSF, TNFα and IL-1β production 20 . Here, we report that macrophages produce GM-CSF after M. bovis BCG infection and hypothesized that GM-CSF may contribute to the host response to mycobacteria infection by favoring macrophage M1 polarization. We show that in vitro GM-CSF neutralization induced a shift towards M2 macrophages after M. bovis BCG infection. Indeed, blocking GM-CSF triggered the overproduction of anti-inflammatory IL-10, while TNFα and IL-12, pro-inflammatory cytokines which play a key role for the host-response to tuberculosis, were reduced. Although GM-CSF increased the phagocytic capacity of alveolar macrophages in vitro 35 , we show here that anti GM-CSF treatment increased the intracellular bacterial load in macrophages in vitro but also in vivo. Thus, we propose that the anti-inflammatory milieu induced by GM-CSF neutralization impairs the ability of macrophages to eradicate mycobacteria. Indeed, while M2-related markers such as Arg1 and Ym1 genes or CD206 were overexpressed after GM-CSF neutralization, M. bovis BCG-induced M1 marker Nos2 was reduced, leading to a deficient production of nitric oxide in macrophages infected in vitro in the presence of anti-GM-CSF antibodies.
We next hypothesized that GM-CSF overexpression during acute infection contributes to an efficient M1 response against M. tuberculosis. Although, in vivo GM-CSF MAB treatment reduced M1 polarization after LPS-induced acute lung inflammation, GM-CSF neutralization did not influence the M1/M2 balance in the context of M. tuberculosis infection. Indeed, several in vivo factors such as extracellular matrix composition, cell maturation as well as cell adhesion could greatly influence macrophages polarization 20 . During M. tuberculosis infection, high levels of Th1-derived IFNγ, but also TNFα and IL-12 produced by macrophages, dendritic cells and neutrophils contribute to a complex milieu favoring M1 macrophage polarization 32  GM-CSF was not essential for M1 polarization during M. tuberculosis infection. Indeed, the complex M1 milieu was not affected by GM-CSF neutralization, with no M1/M2 switch, indicating that the main pathway inducing M1 macrophage polarization after M. tuberculosis in vivo infection is not GM-CSF-dependent but most likely IFNγ produced by Th1, NK or NKT cells [38][39][40] .
Anti-cytokine autoantibodies present at steady-state may cause susceptibility to infections 41 . In healthy individuals, anti-GM-CSF autoantibodies present at low levels may neutralize GM-CSF and therefore regulate GM-CSF mediated inflammation. In contrast, high levels of anti-GM-CSF autoantibodies have been associated with pulmonary alveolar proteinosis, and disseminated Crytococcus and Nocardia infections 41 . Higher levels of anti-GM-CSF autoantibodies, together with anti-IFNγ antibodies have been reported in patients with pulmonary disease due to non-tuberculous mycobacteria (NTM), relative to healthy controls, questioning whether these anti-cytokine autoantibodies might be a predisposing factor for pulmonary NTM disease 42 .
Several antibodies neutralizing GM-CSF or its receptor are in development for indications such as severe inflammatory diseases, chronic myeloid leukemia or cancer, either humanized IgG1 (leuzilumab) or fully human IgG (gimsilumab, namilumab). Mavrilimumab, a human monoclonal antibody targeting GM-CSFRα, has been tested in subjects with rheumatoid arthritis 17,18 . Our data, using anti GM-CSF antibodies in a tuberculosis infectious model may be relevant for the neutralization of GM-CSF pathway in the clinic. Indeed, we show here that murine monoclonal antibodies to GM-CSF induce a modification of granuloma integrity in the lung of immunocompromised mice infected by M. tuberculosis. The data raise concerns about the effect that such GM-CSF neutralizing therapies may have on latent tuberculosis infection in patients suffering from chronic inflammatory diseases. Indeed, recent mavrilimumab long-term safety and efficacy phase IIb studies in rheumatoid arthritis patients identified upper respiratory tract infection, including 2 cases of active pulmonary tuberculosis infection (0.22/100 PY) in patients with no active or latent tuberculosis at screening 43,44 . Thus, neutralization or adjuvant host-directed therapies using antibodies to a major cytokine such as GM-CSF might compromise the balance of host responses to mycobacterial infection and host immune system should be modulated with care in future immunotherapies.

Materials and Methods
Mice. C57BL/6 (B6) WT mice and TNFα deficient mice (TNFα −/− ) 45 were bred in our specific pathogen free animal facility at CNRS (TAAM UPS44, Orleans, France). For experiments, adult (8)(9)(10)(11)(12) week old) animals were kept in ventilated cages or for the infectious protocols in biological isolation safety cabinet glove boxes in a biohazard animal unit. The infected mice were monitored every day for clinical status, weighted twice weekly and were sacrificed in accordance with ethical guidelines whenever reaching an endpoint such as 20% bodyweight loss. This study was carried out in accordance with the recommendations of the French Government's animal experiment regulations and the protocol was approved by the "Ethics Committee for Animal Experimentation of Antibodies. Mouse anti-mouse GM-CSF monoclonal antibodies (IgG1 Clone B2.6 or IgG2b Clone A7.39), obtained after immunization as described 46 , were administered (200 µg/mouse i.p.) on the indicated days. IgG1 (Clone MAB005, R&D) and IgG2b (Clone MPC-11, Bio X Cell) 47 were used as isotype controls, respectively, as indicated. Bronchoalveolar lavage (BAL) fluid was collected by canulating the trachea and washing the lungs 4 times with 0.5 mL of ice-cold PBS. After centrifugation at 1850 rpm for 10 min at 4 °C, the supernatant of the first lavage was stored at −80 °C for cytokine analysis. Cell pellets were recovered, pooled and counted. For differential counts, cell cytospins were stained with Diff-Quik Staining (Merz & Dade AG., Dudingen, Switzerland) and twice 200 cells scored.

Acute airway inflammation. LPS from
Infection. Aliquots of M. tuberculosis H37Rv kept frozen at −80 °C were thawed, diluted in sterile saline containing 0.05% Tween 20 and clumping disrupted by 50 repeated aspirations through 18, 20, 26 and 27 gauge needles (Omnican, Braun, Germany). Pulmonary infection with M. tuberculosis H37Rv was performed by delivering 1000 ± 200 CFU/mouse into the nasal cavities under ketamine-xylasine anesthesia as above. The bacterial load in the lungs was determined on day 1 post infection on control mice. Isoniazid (INH) and Rifampicin (RIF) were administered at the indicated concentration (1 to 100 mg/L) dissolved in drinking water from day 14 to 35 post-infection. For quantifying bacterial burden, lungs were weighted, defined aliquots homogenized in phosphate-buffered saline (PBS; Dispomix homogenizer, Medic Tools, Switzerland), and tenfold serial dilutions prepared in 0.05% Tween 20 containing 0.9% NaCl were plated in duplicates onto Middlebrook 7H11 (Difco) agar plates containing 10% OADC. After 3 weeks incubation at 37 °C colonies were enumerated and results expressed as log 10 CFU per organ.
Histopathological analysis. Lung tissues were fixed in 4% buffered formaldehyde overnight, paraffinembedded, and 3 µm sections were stained with haematoxylin and eosin (HE). Lung cellular infiltration, emphysema and necrosis were quantified using a semi-quantitative score with increasing severity of changes (0-5). The area surface of pulmonary granulomatous lesions were quantified on digitalized sections using a Zeiss Axio Scan Z1 Zeiss microscope and the ZEN software. Granulomatous lesions were determined in two lobes per mouse comprising a lung surface area of 25 to 35 mm 2 per lung section. Data are represented as percentage of granulomatous lesions corresponding to % area of lesions in mm 2 /total lung tissue in mm 2 . The number of bacterial clusters was counted in the lung lesions (0.5 mm²) and the average number of bacilli per cluster estimated in the same surface area of granulomatous lesions. Analyses were performed by two independent observers including a trained pathologist.