An in vitro dual model of mycobacterial granulomas to investigate the molecular interactions between mycobacteria and human host cells

Authors

  • Marie-Pierre Puissegur,

    1. Department of Molecular Mechanisms of Mycobacterial Infections, Institut de Pharmacologie et de Biologie Structurale, CNRS UMR5089, 205, route de Narbonne, 31077 Toulouse, France.
    Search for more papers by this author
  • Catherine Botanch,

    1. Department of Molecular Mechanisms of Mycobacterial Infections, Institut de Pharmacologie et de Biologie Structurale, CNRS UMR5089, 205, route de Narbonne, 31077 Toulouse, France.
    Search for more papers by this author
  • Jean-Luc Duteyrat,

    1. Electron Microscopy Department, Rangueil Hospital Medical School, University of Toulouse, 31403 Toulouse, France.
    Search for more papers by this author
  • Georges Delsol,

    1. Department of Anatomy and Pathological Cytology, Purpan Hospital, 31300 Toulouse, France.
    Search for more papers by this author
  • Claude Caratero,

    1. Electron Microscopy Department, Rangueil Hospital Medical School, University of Toulouse, 31403 Toulouse, France.
    Search for more papers by this author
  • Frédéric Altare

    Corresponding author
    1. Department of Molecular Mechanisms of Mycobacterial Infections, Institut de Pharmacologie et de Biologie Structurale, CNRS UMR5089, 205, route de Narbonne, 31077 Toulouse, France.
    Search for more papers by this author

*E-mail frederic.altare@ipbs.fr; Tel. (+33) 5 61 17 54 63; Fax (+33) 5 61 17 59 94.

Summary

In the majority of individuals infected with Mycobacterium tuberculosis, the bacilli cause a long-term asymptomatic infection called latent tuberculosis, a state during which the bacilli reside within granulomas. Latently infected individuals have around 10% risk of progression to clinical disease at a later stage. Determining the state of the mycobacteria and the host cells during this latent phase, i.e. within the granulomas, would greatly improve our understanding of the physiopathology of tuberculosis, and thus enable the development of new therapeutic means to treat the one-third of the world's population who are latently infected. We have developed an in vitro model of human mycobacterial granulomas, enabling the cellular and molecular analysis of the very first steps in the host granulomatous response to either mycobacterial compounds or live mycobacterial species. In vitro mycobacterial granulomas mimic natural granulomas very well, with the progressive recruitment of macrophages around live bacilli or mycobacterial antigen-coated beads, their differentiation into multinucleated giant cells and epithelioid cells, and the final recruitment of a ring of activated lymphocytes. Besides morphological similarities, in vitro granulomas also functionally resemble natural ones, with the development of intense cellular co-operation and intracellular mycobactericidal activities.

Introduction

Tuberculosis is one of the major causes of mortality in the world with around 2–3 million deaths every year (Dye et al., 1999). This disease is caused by Mycobacterium tuberculosis, an acid-fast bacillus transmitted primarily via the respiratory route. The inhaled bacilli arrive in lung alveoli where they are thought to enter and replicate within alveolar macrophages. Cytokine and chemokine production by the infected lung macrophages triggers an inflammatory response leading to the recruitment of blood and tissue macrophages and lymphocytes at the infectious site. This cellular accumulation around the bacilli, called tuberculous granuloma, is the histological hallmark of tuberculosis. Granulomas mainly contain blood-derived macrophages, epithelioid cells (differentiated macrophages), multinucleated giant cells, also called giant Langhans cells (fused macrophages), surrounded by a rim of T lymphocytes (Saunders and Cooper, 2000; Toossi and Ellner, 2001). Their major function is now generally considered to be the containment of the infection to a localized area, thus avoiding bacterial spread to surrounding healthy tissues and to other organs, but also to concentrate the immune response to a limited infectious area (Saunders et al., 1999).

In the majority of cases, M. tuberculosis establishes latency, a period during which infected individuals do not develop clinical disease. The bacteria remain in a kind of dormant stage, from which they can reactivate, in 5–10% of infected individuals, at a later stage (Parrish et al., 1998; Glickman and Jacobs, 2001). Latency can thus be considered as an equilibrium state between the host immune response and the bacillus. The host develops a strong primary immune response, culminating with the formation of granulomatous lesions, which are very effective, at least in restraining M. tuberculosis, but do not allow complete eradication of the bacilli. On its part, the mycobacteria appears to have elaborated a series of evasion strategies, as it happens to persist within the granuloma from which it is able to reactivate and give rise to clinical disease (Sandor et al., 2003). The complex balance between latency and reactivation seems to be the key element of M. tuberculosis pathogenesis, and the granulomatous lesion appears to be the battlefield where the interplay between the host immune response and the bacillus takes place.

Our limited understanding of the physiopathology of tuberculosis has probably played a large part in the lack of an effective treatment strategy for this disease. Given the importance of the granulomatous reaction for the outcome of M. tuberculosis infection, deciphering the molecular physiology of tuberculous granulomas would thus be an important step towards therapy improvement. In order to achieve this goal, animal models of granulomatous inflammation have been developed (Kunkel et al., 1998). Some have been performed in the mouse model and consisted of the analysis of tissue granulomas induced by either M. tuberculosis infection or intravenous challenge with mycobacterial antigen-coated beads (Sandor et al., 2003). More recently, zebrafish embryos were used as a model of Mycobacterium marinum-induced granuloma formation, exploiting the optical transparency of the embryos to study the granulomatous reaction (Davis et al., 2002). The major limit of such in vivo models is that only highly differentiated granulomas are accessible, even as early as 2 weeks after infection, thus preventing all studies of the very first molecular and cellular processes involved in the development of the granulomatous structure (Roach et al., 1999). In addition to these technical limits, the use of animal models always implies species differences, which preclude exact replication to human corollary disease.

In order to investigate the molecular and cellular interactions occurring during human granulomatous response, and especially within the very first steps in granuloma formation, we have developed an in vitro model of human mycobacterial granulomas. In this model, in vitro granulomas are induced by either (i) mycobacterial antigen-coated artificial beads or (ii) live mycobacteria. This dual in vitro model, respectively, enables analysis of (i) the influence of isolated mycobacterial compounds on the first stages of the granulomatous response and (ii) the molecular and cellular interplay between different live mycobacterial species and human granuloma cells.

Results

Kinetics of the cellular aggregation around PPD-coated Sepharose beads

The first step in the development of mycobacterial granulomas in vitro was to induce the cellular recruitment of peripheral blood mononuclear cells (PBMC) around artificial beads coated with mycobacterial compounds. Such antigen-coated beads have already been used in animal models to induce in vivo granuloma formation. Actually, Sepharose beads covalently coated with partially purified protein derivative of Mycobacterium tuberculosis (PPD) were delivered in rodents intravenously presensitized with PPD (Chensue et al., 1997). Based on these experiments, we first used cyanogen bromide (CNBr)-activated Sepharose beads coated with PPD to induce in vitro granulomas. PBMCs from a healthy donor were incubated with PPD-coated Sepharose beads for 1 week. The cellular recruitment around the beads was followed by light microscopy, in comparison with control beads coated with a single amino acid (glycine), used to block free CNBr sites of control beads. As shown in Fig. 1, the recruitment of a cellular monolayer occurred around PPD beads between days 3 and 5. This cellular accumulation kept on growing to reach a multilayer, granuloma-like stage at day 8, whereas control beads did not recruit any cells at this stage. The experiment was repeated 10 times with a conserved time course of around 5 days for monolayer accumulation and 8–9 days for a complex structure (not shown).

Figure 1.

Kinetics of PPD-induced granuloma formation. Sepharose beads (200 beads per well) coated with either glycine (glycine beads) or PPD (PPD beads) were incubated with 106 PBMCs from a healthy donor, as indicated. The granulomatous reactions are shown at days 0, 3, 5 and 8 of incubation (×400).

Characterization of PPD-induced granulomas

In order to analyse the structure of in vitro granulomas, PPD-coated beads were incubated with control PBMCs. The granulomas were collected at different time points in the cellular accumulation and submitted to scanning electron microscopy (SEM). As early as day 1 of incubation, the cellular gathering appeared to start (Fig. 2A), with the recruitment of monocyte-like cells, rapidly flattening over the whole bead surface (Fig. 2B). At day 4, lymphocyte-like cells (small round cells) were recruited. They did not directly bind the beads, but rather appeared to bind to previously attached monocyte-like cells (Fig. 2C). After day 5 of incubation, the recruited cells progressively covered the whole bead surface, and multilayers comprising macrophages and lymphocytes started to form (Fig. 2D). By this point in the kinetics, the morphology of the bound cells started to change, with important signs of cellular differentiation, as shown by the presence of numerous pseudopodia at the cell surfaces (Fig. 2E). At this stage, the cellular structure had fully covered the whole bead, therefore mimicking natural granulomatous structures, with large and differentiated monocyte/macrophage-like cells surrounded by lymphocyte-like cells (Fig. 2F).

Figure 2.

Scanning electron micrographs of PPD-induced granuloma. The granulomas were collected at different times of culture (A, 24 h; B, day 2; C, day 3; D, day 5; E, day 7; F, day 9), prepared and observed under a scanning electron microscope. Bars represent 5 µm (A, C and E) and 50 µm (B, D and F).

Characterization of the cell populations recruited within in vitro granulomas

To identify and characterize clearly the different cell types observed by the SEM analysis, PPD bead-induced day 9 structures were collected. The cells were separated from the beads, plated on glass slides with a cytospin and stained (Fig. 3). May–Grünwald–Giemsa (MGG) staining showed that day 9 structures mainly consist of activated macrophages with numerous large vesicles (Fig. 3A). Some cells even appeared multinucleated, thus resembling giant Langhans cells from natural granulomas (Fig. 3B). To confirm that these multinucleated cells corresponded to Langhans cells, immunostaining for the macrophage marker CD68 was performed. As shown in Fig. 3C, the multinucleated cells expressed the CD68 marker, thus confirming that they belong to the monocyte/macrophage lineage, and possibly represent an early stage of multinucleated giant cell (MGC) formation. Interestingly, some of the MGC were shown to be tightly surrounded by blastoid cells. To determine whether these strongly bound cells were lymphoblasts, we tested them for the expression of the CD3 molecule (Fig. 3D). As expected, all the surrounding cells appeared to be CD3 positive, thus indicating important MGC/lymphocyte interactions within these granulomatous structures in vitro.

Figure 3.

Histological analysis of PPD-induced granuloma. After 9 days of culture, the PPD-induced granuloma cells were plated on glass slides and stained by May–Grünwald–Giemsa stain (A and B) or antibodies to CD68 (C), CD3 (D) and CD163 (E and F).
A and B. Two views of the same May–Grünwald–Giemsa-stained slide.
A. Activated macrophages (Mf) and lymphocytes (Ly) are present within day 9 granulomas (×400).
B. Multinucleated giant cells (GC) are also observed on the same slide and appear to be tightly linked with surrounding macrophages (Mf) (×400).
C and D. A CD68-positive multinucleated cell is shown (GC), surrounded by CD3-positive cells (Ly) at a magnification of 400×.
E and F. Two views of the same slide stained for CD163. Activated macrophages (Mf) displaying intense CD163 expression are visible (E), as well as weakly stained epithelioid cells (EC) (E and F), lymphocytes (Ly) (F) and multinucleated giant cells (GC) that do not express CD163 (F).

In addition to the fusion of some recruited macrophages into multinucleated cells, we looked for the presence of epithelioid cells, as usually observed in natural granulomas. For this purpose, we measured the expression level of CD163, a cell surface marker reported to be highly expressed by macrophages, but only weakly expressed by epithelioid cells, and not by multinucleated giant cells (Seitzer et al., 2001). As shown in Fig. 3E and F, both high and low CD163-expressing cells can be found in the macrophage population collected around the beads, thus confirming the presence of epithelioid cells within in vitro granulomas.

In vitro granulomatous reaction is not PPD specific and is conserved between individuals

A granulomatous response can thus be obtained in vitro with human cells incubated with PPD-coated beads. However, if CNBr-activated beads represent a good means of binding proteic antigens, by the covalent link of NH2 polypeptides, they are perhaps not the best candidate beads to be used to analyse complex pools of mycobacterial antigens, especially as the mycobacterial envelope is well known for its high density of lipids (Daffe and Draper, 1998). In order to define the best bead type for the analysis of various mycobacterial antigens, i.e. own ability to bind lipids and proteins as efficiently, we compared Polyacrylamide, Sepharose and Latex beads. The beads were coated with known concentrations of either model proteins (PPD) or lipids (glyco- and phospholipids), and the efficiency of binding was evaluated by the quantification of unbound proteins and lipids remaining in the binding reaction buffer (not shown). CNBr-activated Sepharose beads were selected as a good compromise for the binding of both types of compounds and no intrinsic cellular recruitment activity.

To determine whether in vitro granulomas are a PPD-specific phenomenon, or if they can be induced around other mycobacterial antigens, we used Sepharose beads covered with homogenized BCG, referred to as total extract (TE) in the text. After a 9 day incubation period with PBMCs, multilayer granulomas appeared (Fig. 4A). These granulomas presented the same structure and cellular composition as PPD-induced granulomas, with activated macrophages, multinucleated macrophages and epithelioid cells, surrounded by a ring of lymphocytes (not shown). Interestingly, when used as alternative negative controls, neither Escherichia coli extract-coated nor BSA-coated beads induced any cellular recruitment (not shown).

Figure 4.

Granulomatous reaction induced around TE-coated beads at day 9.
A. PBMCs were incubated with glycine-coated or TE-coated beads. Compared with glycine beads, TE beads are covered with a multilayer cellular accumulation.
B. PBMCs from 12 unrelated healthy donors were incubated with TE-coated beads. Representative day 9 granulomatous reactions are shown (×400).

In addition, we investigated interindividual variability in the capacity to develop in vitro mycobacterial granulomas. We thus incubated TE-coated beads with PBMCs from 12 BCG-vaccinated healthy donors and followed the granulomatous reaction over 9 days. As shown in Fig. 4B, multilayer cellular accumulations, with differentiated cellular structures, formed for each tested individual. These granulomas were strictly comparable to PPD bead-induced structures, in terms of both quantitative and qualitative cellular recruitment, as well as for the kinetics of this recruitment (not shown) with comparable multilayer structures at day 9 of incubation. However, it is interesting to note that small discrepancies in the recruitment speed were observed, between individuals, during the first 2 days of reaction, yet these differences disappeared between days 3 and 5 (not shown).

Development and characterization of BCG-induced granulomas

In parallel experiments, we tried to see whether granulomatous structures were also able to form in vitro around live mycobacteria. To achieve this goal, we incubated PBMCs from a healthy donor with live BCG and followed cellular accumulation around the bacilli over a 10 day period. In order to be as close as possible to in vivo conditions of the mycobacterial–host cell interaction, where a very few bacilli progressively meet a large number of host cells, we decided to lower the bacteria–cell ratio to the minimum value necessary to induce granulomatous reactions. We thus defined that the multiplicity of infection (MOI) of 1 BCG for 10 PBMCs suited these criteria (not shown).

PBMCs from a healthy donor were incubated with live mid-log phase BCG. The kinetics of cellular recruitment were analysed by SEM at different time points of the reaction. After 6 h of incubation, monocytes were shown to gather around the bacilli (Fig. 5A) and, within the first day of culture, they progressively wrapped round the BCG, which now appeared to be tightly bound to the cell surface (Fig. 5B and C). At day 3, the mycobacteria have been engulfed within the cellular structure, thus forming a compact cellular aggregate with clear signs of intense cellular activation, as shown by the high number of visible cell membrane rearrangements compared with previous stages (Fig. 5D). After 4–5 days, lymphocytes are recruited around the structures (Fig. 5E), which then keep on growing with continuous cellular recruitment until the incubation is stopped. At late stages, the granulomatous reaction nicely resembles natural granuloma with multiple-layer cell accumulation, where activated large macrophages can be seen in tight contact with lymphocytes (small round cells), as shown in Fig. 5F.

Figure 5.

Scanning electron microscopy analysis of BCG-induced granuloma. Approximately 105 BCG were incubated with 106 PBMCs for 9 days. The granulomas were collected at different times of incubation (A, 6 h; B and C, 24 h; D, day 3; E, day 5; F, day 9) and prepared for transmission electron microscopy. Bars represent 5 µm (A–E) and 50 µm (F).

To better assess the internal structure of these in vitro granulomas, day 9 BCG granulomas (Fig. 6A) were collected, fixed and embedded in an Epon–araldite resin. Transverse sections (0.5 µm) were then performed through the whole granulomas. In order to evaluate the interplay between macrophages, lymphocytes and BCG, a section collected in the middle region of the structure was analysed by transmission electron microscopy (Fig. 6B). As shown in Fig. 6B, activated lymphocytes (Ly) presenting tight contacts (arrows) with activated macrophages (Mf) and multinucleated cells (GC) are visible in the middle of a BCG-induced granuloma. The high number of granules within activated macrophages depicts an intense activation state of these cells, thus confirming a phenotype of epithelioid cells for these macrophages. Figure 6C (enlargement of inset C from Fig. 6B), shows numerous BCG-containing phagosomal compartments in a highly activated epithelioid cell, characterized by a high level of phagocytosis and by the presence of numerous mitochondria (Mt) in the surroundings of the phagosomes. The central region of in vitro BCG-induced granulomas thus presents with a site of intense interaction between lymphocyte, macrophage and mycobacteria. Numerous immunological synapses are present between both lymphocytes and macrophages, as well as an important mycobactericidal activity within activated macrophages.

Figure 6.

Transmission electron microscopy analysis of BCG-induced granuloma. After 9 days of culture, BCG-induced granulomas were collected under light microscopy (A, ×400), fixed and embedded in an Epon–araldite resin. Multiple transverse sections (0.5 µm) were obtained, stained and observed using an HU12A transmission electron microscope.
B. A section corresponding to the centre of the granulomatous structure is represented. A giant cell (GC) with two visible nuclei (N) appears in tight relation with surrounding lymphocytes (Ly) and macrophages (Mf) (×2000).
C. The enlargement of inset C from (B) shows multiple BCG-containing phagosomal vesicles, surrounded by numerous mitochondria (Mt) as a sign of intense cellular activation around the vesicles (×6000).

Thus, not only do in vitro granulomas reproduce the cellular structure of natural ones, but they present with similar functions as shown by the signs of bactericidal activities within macrophages in the middle of the structure, and similar intercellular communications with tightly linked, and thus intensely communicating, lymphocytes surrounding the structure.

Discussion

We have developed an in vitro human model enabling the formation of either mycobacteria-induced or mycobacterial antigen-induced granulomas. Both systems are able to trigger a cellular aggregation reminiscent of natural mycobacterial granulomas, in terms of morphology and cell differentiation. In contrast to previously used granuloma models, our dual model is original as it has been developed for the first time with mycobacterial compounds or live mycobacteria. Moreover, it has been developed in vitro and in the human model, thus enabling for the first time a very early analysis of human granulomatous response to mycobacteria. Other in vitro models had been developed previously to analyse the ability of Schistosome antigens (Doughty et al., 1984), live Nematodes (Seitzer et al., 2001) or live Candida (Heinemann et al., 1997) to induce granulomas formation; however, the formation of human mycobacterial granulomas has never been investigated to date using such an in vitro model.

The major defence mechanism used by the host to control M. tuberculosis infection is the formation of epithelioid granulomas. It was previously described that newly formed mycobacterial granulomas only consist of immature phagocytes surrounded by T lymphocytes, whereas mature granulomas contain specialized cell types such as multinucleated giant cells and epithelioid macrophages, thus forming the so-called epithelioid granulomas. The latter cell type displays tightly interdigitated cell membranes that link adjacent cells, thus walling off mycobacteria and preventing further dissemination (Bouley et al., 2001). In our model, within 9 days of culture, human PBMCs are recruited to form a typical epithelioid granuloma presenting with fused macrophages, appearing as multinucleated cells, and epithelioid macrophages tightly linked to surrounding macrophages and lymphocytes. These in vitro granulomas display very similar morphological characteristics and cellular differentiation levels to natural granulomas. They form by the early aggregation of immature phagocytes and lymphocytes but, later on, there is an important differentiation of these phagocytes into highly activated cells, presenting with large cell membrane rearrangements and intense intracellular activation, as shown by high phagocytosis and the presence of numerous organelles. All activation and differentiation stages are probably modulated by the tight links developed between the epithelioid cells and the surrounding recruited lymphocytes.

The classically accepted definition of human mycobacterial granulomas is cellular aggregation around mycobacteria. These multicellular structures mainly contain immature activated macrophages, but also mature activated macrophages with a larger cytoplasm and interdigitated membranes classically called ‘epithelioid cells’, and fused epithelioid cells forming multinucleated cells classically called Langhans cells or multinucleated giant cells (MGC). All these cell subtypes are surrounded by a ring of activated lymphocytes (Adams, 1974; Sandor et al., 2003). Our in vitro model of human mycobacterial granulomas enables the recruitment and maturation of all these cell types with, in addition to macrophages and lymphocytes, the presence of both epithelioid cells and MGC. The characterization of in vivo human mycobacterial granulomas being, to date, limited to histological descriptions, we think that having a comparable histological presentation, the cellular aggregations occurring in our in vitro model, induced by either live mycobacteria or mycobacterial compounds, can thus be considered as mycobacterial granulomas.

When using in vivo animal models of mycobacterial granulomas, several weeks of activation with coated beads or infection with live bacilli are necessary for the induced granulomas to be large enough to enable their localization in the animal tissues and thus further study, which precludes the analysis of the very early stages of the granulomatous reaction. In contrast, in vitro granulomas induced by live mycobacteria allow comparison of the granulomatous response to different mycobacterial species, or different bacterial mutants of the same species, at the very first steps in the immune response. Such in vitro studies will permit, on the one hand, the identification of key molecular effectors involved in the very first stages of the human immune response to a defined mycobacterial species and, on the other hand, the identification of mycobacterial virulence factors interfering with the granulomatous reaction, by a comparative study of mycobacterial mutants.

Another interesting use of our in vitro mycobacterial granuloma model is the characterization of candidates for the development of alternative vaccines for tuberculosis. Actually, the granulomatous reaction being considered as a necessary step in the antimycobacterial response (Altare et al., 1998a), it is tempting to suggest that a good vaccine candidate has to be a good granuloma inducer. It is important to note that the BCG vaccine, although displaying good granuloma-inducing ability, does not protect 100% of immunized individuals against tuberculosis. Therefore, the induction of granuloma formation may not be a sufficient way of inducing protection. Thus, in vitro granuloma will be very useful to (i) evaluate the granuloma-inducing activity of particular antigens or attenuated mutants and (ii) compare them with the BCG vaccine, and thus eventually identify, among candidate molecules able to trigger granuloma formation, more potent inducers of antimycobacterial response than the BCG vaccine. On the other hand, mycobacterial attenuated mutants, also classically proposed as live vaccine candidates, would thus be analysed with the in vitro model, first to check their attenuated virulence state and, hence, that this attenuation did not affect their ability to still induce a protective immune response.

In any case, the use of such an in vitro model will above all enable an in-depth characterization of human granulomatous response, at both the cellular and molecular levels, thus potentially giving new tools to (i) better understand the human immune response to mycobacteria and (ii) improve this response and try to reach the complete control of mycobacterial infections. The development of live mycobacteria-induced in vitro granulomas perfectly complements in vivo animal experiments, enabling a precise analysis of the immune response developed within granulomatous structures, especially for the very first steps in this response.

Altogether, we have developed for the first time a dual model of in vitro mycobacterial granulomas in the human model. This dual model now represents a very useful tool for the in-depth deciphering of the complex molecular interplay between mycobacteria and host cells within granulomatous structures, which represents the very key point driving either the containment of the bacilli or the development of progressive disease. In combination with in vivo animal models, this dual human granuloma model will help to define new molecular targets for the development of better therapeutic or prophylactic means to control M. tuberculosis infection, but also infections by other pathogenic mycobacteria. This model would also be very useful for the study of other human diseases caused by intracellular bacteria known to induce a granulomatous reaction, as well as other conditions not even clearly related to an infective agent, but presenting with granulomatous reactions, such as Crohn's disease or sarcoidosis.

Experimental procedures

Blood samples and cell culture

Human blood samples were collected from BCG-vaccinated, PPD-reactive, non-tuberculous control individuals at the Etablissement Français du Sang, Toulouse, France, with informed consent obtained from each donor. All the analyses presented here were performed according to the principles expressed in the Helsinki Declaration. PBMCs were isolated from blood by density gradient centrifugation as described previously (Altare et al., 1998b), using Ficoll-Paque Plus (Amersham Pharmacia). For all experiments, the PBMCs were cultured in RPMI-1640 medium (Gibco) supplemented with 10% heat-inactivated pooled human AB serum (Sigma), containing (bead-induced granulomas) or not (live mycobacteria-induced granulomas) 300 U ml−1 penicillin, 0.3 mg ml−1 streptomycin (Gibco). This culture medium will be referred to as cell culture medium.

Mycobacterium bovis BCG growth

Mycobacterium bovis BCG (Pasteur strain) were a kind gift from M. Daffe (Department of Molecular Mechanisms of Mycobacterial Infections, CNRS UMR5089, Toulouse, France). They were grown in Middlebrook 7H9 medium supplemented with 10% albumin–dextrose–catalase (Difco) and 6% glycerol. The bacilli used for granuloma induction were grown in liquid suspension under moderate shaking for 7 days and recovered by centrifugation. The bacterial aggregates were removed by smooth agitation with 3-mm-diameter glass beads (Merck); the bacteria were then diluted in PBS and submitted to low-speed centrifugation. Individualized bacteria were counted in the supernatant, collected by centrifugation and diluted in cell culture medium without antibiotics at the appropriate concentration.

Preparation of BCG total extract (TE)

For antigen preparation, the mycobacteria were cultured as a surface pellicle. After 3–4 weeks culture, the bacterial biofilm was carefully collected and homogenized using an equivalent volume of 0.1-mm-diameter glass beads (Merck) and a bead beater system (Retch), as described previously (Ortalo-Magne et al., 1995). The supernatant was removed, filtrated, and the quantity of proteins in the total extract was determined by Bradford's method (Bradford, 1976) (Bio-Rad protein assay). Aliquots corresponding to 2 mg of proteins were lyophilized (CHRIST Alpha1-2) and stored at −20°C.

In vitro granuloma formation

Antigen-coated bead-induced granuloma.  Two milligrams of TE (protein equivalent), purified protein derivative (PPD) from M. tuberculosis (Avantis Pasteur) or glycine (Amersham) was diluted in 5 ml of coating buffer and coupled to 100 mg of cyanogen bromide (CNBr)-activated Sepharose 4-B beads (Pharmacia) according to the manufacturer's instructions. The efficiency of the coating step was evaluated by comparison of the protein concentration in the reaction buffer before and after coating. Glycine (Gly) beads were always used as a negative control, whereas PPD beads were used as a positive control for in vitro mycobacterial granuloma reaction in experiments with TE-coated beads. Approximately 106 PBMCs, freshly isolated as described above, were plated in 24-well culture plates (Nunc) and incubated with 200 coated beads in a final volume of 1 ml of complete culture medium per well, for up to 15 days at 37°C in a 5% CO2 atmosphere. Cellular aggregation was followed every day under light microscopy.

BCG-induced granuloma.  The mycobacteria were grown for 7 days as described above, counted, diluted in 24-well culture plates to reach a final concentration of 105 bacteria ml−1. The granulomatous reactions were induced by adding 106 PBMCs (ratio bacteria–cell 1:10) and kept at 37°C in a 5% CO2 atmosphere for 15 days. Cellular aggregation was followed every day under light microscopy.

Electron microscopy

Scanning electron microscopy (SEM).  For SEM analysis, the granulomas were rapidly collected under light microscopy at different time points and prepared for SEM by fixing in 2% glutaraldehyde in 0.1% phosphate buffer for 4 h. After two washes in the same buffer, the granulomas were removed, dehydrated in a graded ethanol series, dried by critical point drying with EMSCOPE CPD 750 and coated with gold–palladium for 3 min at 100 Å min−1, and observed with a S450 scanning electron microscope (Hitachi) at an accelerating voltage of 15 kV.

Transmission electron microscopy.  Before microscopic analysis, the granulomas were treated as follows. Granulomas were fixed for 4 h at 4°C in 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4,  containing  6 mM  CaCl2.  After  an  extensive  wash  with the same buffer, granulomas were post-fixed for 1 h at room temperature with 1% OsO4 in the same buffer containing 1.5% K3Fe(CN)6. The samples were then treated for 12 h with 1% uranyl acetate aqueous, dehydrated and embedded in an Epon–araldite resin (Electron Microscopy Sciences). Thin sections of 0.5 µm (Ultracut Reichert Jung) of the resin-embedded granulomas were stained with 3% uranyl acetate and lead citrate and observed under a HU12A transmission electron microscope (Hitachi).

Immunohistochemistry

Granuloma cells were collected and plated on glass slides with a cytospin (Cytobuckets; Jouan) and submitted to May–Grünwald–Giemsa (MGG) and/or immunohistological staining as follows. After 10 min incubation in cold acetone, the slides were stained by MGG reagents (Sigma) according to the manufacturer's instructions. For alkaline phosphatase staining, LSAB-2 (labelled streptavidin biotin reagents; Dako) was used. Rapidly, specimens were fixed for 10 min in cold acetone, incubated with the corresponding primary mouse monoclonal antibody [CD3 (clone T3-4B5), CD68 (clone PG-M1) or CD163 (clone Ber-Mac3); Dako] for 15 min followed by three washes in 0.05 M Tris-HCl, pH 7.5. The slides were then incubated with rabbit anti-mouse Ig antiserum (Dako) for 15 min, washed in Tris-HCl and incubated with streptavidin–alkaline phosphatase complex for 15 min. The Fuchin chromogen was added for 15 min, the slides were then counterstained with haematoxylin (Dako), 1% NH4OH (Sigma) and mounted. The stained slides were observed by inverted microscopy (Nikon TE 300).

Acknowledgements

F.A. would like to thank Professor Alain Fischer for continuous support. We thank all the members of the laboratory of Molecular Physiology of Mycobacterial Granulomas for helpful discussion, and Drs Barbara Doughty (Texas University) and Ghislaine Daste (Purpan Hospital) for helpful technical advice. M.P.P. was supported by a grant from the Ministère de l’Education Nationale. This work was supported by grants from the Fondation pour la Recherche Medicale, Ministère de la Recherche (ACI jeune chercheur 2002) and CNRS.

Ancillary