The adaptor molecule CARD9 is essential for tuberculosis control

Aerosol infection with virulent strain MTB H37Rv caused rapid death of Card9^−/− mice (Fig. 1 A). Diseased animals gradually lost weight (Fig. S1) and started to succumb at 26 d post infection (p.i.). Over a wide range of the bacterial inoculum (100–450 CFUs) we observed a remarkable susceptibility of Card9^−/− mice, which did not survive beyond 34 d p.i. in any of the experimental settings.

Lung bacterial burdens were significantly higher in Card9^−/− mice (Fig. 1 B), and Ziehl-Neelsen staining of infected lungs confirmed CFU results (Fig. 1 C). Histological examination revealed profound tissue damage in susceptible animals (Fig. 1, D–F), characterized by acute pneumonia with accelerated accumulation of inflammatory cells. Parenchyma appeared disorganized with multiple necrotic foci and lacked typical granulomatous appearance. Large infiltrates were often surrounded by emphysematous tissue and rarely by functional alveoli. Spleen pathology did not significantly differ between mouse strains (Fig. S2). Because the lung architecture was massively altered, we assessed impact of Card9 deficiency on cell death. iNOS expression was determined as a major antimicrobial effector molecule in mice. In accordance with the results obtained for NO serum levels (Fig. S2), iNOS immunostaining revealed no apparent differences between WT and KO animals. The more intense TUNEL signals in the Card9^−/− mice indicated differential apoptotic cell death. We also detected a different TUNEL staining pattern. In WT tissues, apoptotic signals were nuclear, whereas extracellular, interstitial TUNEL-positive material preponderated in Card9^−/− lungs, indicative of secondary necrosis. Thus, increased cell death events in inflamed lungs of Card9^−/− mice correlated with TB susceptibility.

To define the mechanisms responsible for the high susceptibility to TB in Card9^−/− mice, we generated BM-derived macrophages (BMDMs) and BM-derived DCs (BMDCs) and infected them with MTB H37Rv. Card9^−/− BMDMs produced similar NO levels as WT cells (Fig. 2 C), and were neither impaired in internalization (Fig. 2 B) nor in killing of MTB after IFN-γ activation (Fig. 2 A). Strong evidence for increased cell death in lungs of Card9^−/− mice prompted us to determine whether deficient BMDMs showed accelerated cell death after MTB encounter. Nucleosome enrichment in postnuclear fractions did not differ, suggesting similar frequencies of apoptotic cell death in WT and KO cells after MTB infection (Fig. 2 D). Clearance of apoptotic cells by Card9^−/− BMDMs was also not impaired (Fig. S3).

To monitor cell activation status we measured cytokine levels in culture supernatants. Card9^−/− BMDMs were impaired in synthesis of the proinflammatory cytokines TNF, IL-6, and CCL5 and produced less IL-12 compared with WT cells (Fig. 2 E). Moreover, they were defective in IL-1β release (Fig. 2 F). A similar response pattern of infected BMDCs (depicted for IL-1β and TNF) and lower IL-12p40 abundance was recorded (Fig. S4). We hesitate to draw conclusions on the biological meaning of IL-12 levels in DC supernatants because of the high background caused by MACS separation of the cells. However, we do assume that CARD9 is critical for adequate activation of innate immune cells after MTB recognition.

CARD9 is involved in multiple signaling pathways, downstream of different classes of PRRs (Ishii et al., 2008). To characterize receptors upstream of CARD9 involved in APC responses to MTB, we performed experiments using heat-killed MTB or blocking phagocytosis. The differential activation after cytochalasin treatment or after recognition of heat-killed MTB confirmed our experiments with live MTB (Fig. 3 A). We conclude that the CARD9 pathway is important for sensing MTB lipoglycan compounds through cell surface PRRs.

Because the relevance of TLRs in TB is well established (Reiling et al., 2008), we focused on dectin-1/Syk–dependent macrophage activation by MTB. We treated cells with either 500 µg/ml laminarin or 1–10 µM piceatannol 1 h before infection. Both laminarin (β-glucan, thus blocking dectin-1) and piceatanol (syk inhibitor) specifically blocked TNF and IL-6 synthesis, leaving MTB internalization unaffected (Fig. S5 A). Thus, dectin-1 signaling was involved in proinflammatory responses against MTB (Fig. 3 B). Next, we stimulated BMDMs and DCs with the following MTB-derived molecules: peptidoglycan (PG), mycolyl-arabinogalactan (AG) PG (mAGP), manLam, and trehalose dimycolate (TDM). We measured levels of TNF, IL-10 (Fig. 3 C), IL-12, and IL-6 24 h after stimulation. Scant cell activation was observed for manLam and TDM (unpublished data). This observation is in agreement with studies that investigated TDM/TDB activity on myeloid cells and reported minimal cytokine release at 24 h, but substantial production at later timepoints (Werninghaus et al., 2009). Clear differences were measured for mAGP (Fig. 3 C and Fig. S5, B and C). The mAGP-triggered cytokine synthesis in macrophages depends on NOD2, an intracellular sensor for PG (Fortune et al., 2004; Gandotra et al., 2007). Mycolic acid methyl esters (MAMEs) and AG, which are constituents of mAGP, induced scant TNF release (unpublished data). Overall, our data suggest that the CARD9 adaptor collects signals from various PRRs, including dectin-1, and NOD2 during MTB infection, reemphasizing its critical role in control of innate immunity.

To further characterize the role of CARD9 in TB in vivo, we performed a detailed analysis of innate and adaptive cell responses in the lungs and spleens of infected mice. Aerosol infection with a low inoculum (70 CFUs) did not reveal apparent differences in T cell subpopulations recruited to the lung at day 14 p.i. Disease progression did not alter the ratio of CD4⁺/CD8⁺ cells in Card9^−/− versus WT mice (Fig. 4 A). Moreover, the frequencies of PepA tetramer-positive CD8⁺ T cells (Fig. 4 B) and FoxP3⁺ T lymphocytes (unpublished data) did not differ in Card9^−/− and WT mice. Next, we determined T cell functions. Cytokine production (IL-2, IFN-γ, TNF, and IL-17) was not altered consistently (Fig. 4 B and Fig. S6). Moreover, in preliminary experiments CARD9 KO and WT mice mounted similar delayed-type hypersensitivity (DTH) responses and protective immunity in response to BCG vaccination (Fig. 4, C and D). Thus, vaccination-induced acquired immunity overcame CARD9-dependent innate immune deficiency. We conclude that CARD9 defects affect innate immune rather then acquired T cell responses in TB.

The Card9 deficiency caused marked accumulation of neutrophils in the lungs of infected mice (Fig. 5 A). The frequency of DCs and lung macrophages, as well as the expression level of a maturation marker expressed by APC (CD86), in infected lungs did not differ between the mouse strains (Fig. 5 B). Yet, myeloperoxidase (MPO) staining demonstrated that granulocytes accumulated in nest-like structures within large inflammatory foci (Fig. 5 G). The accelerated lung recruitment of polymorphonuclear neutrophils (PMNs) was paralleled by aggravated disease, suggesting a critical role of neutrophils in disease progression to fatal outcome.

We attempted to define triggers of neutrophilic inflammation during MTB infection and to verify the relevance of disbalanced cytokine responses as observed in vitro. To this end, we determined chemokine and cytokine concentrations in lung homogenates and serum. 2 wk p.i., we barely detected any differences (Fig. 5 C), but obvious changes appeared by 3 wk p.i. (Fig. 5 D). CCL5 concentration was reduced in the lungs of Card9^−/− mice. This deficiency alone could not explain the heightened lethality of Card9^−/− animals to TB because CCR5 has multiple ligands (including MIP-1α and MIP-1β) and there is substantial redundancy in the chemokine system (Algood and Flynn, 2004; Badewa et al., 2005). Strikingly, MCP-1 and cytokines involved in neutrophil differentiation (G-CSF) and recruitment (keritinocyte-derived chemokine [KC]) were markedly elevated in susceptible mice. Moreover, we detected increased concentrations of granulocyte-derived MPO in sera of Card9^−/− mice (Fig. 5 E). In absence of CARD9, PMN transcriptional responses were detected as early as 14 d p.i., at a time-point where bacterial burdens did not differ (Fig. 5 F). Thus, we consider systemic inflammatory disease followed by tissue damage a result of augmented granulocyte differentiation in the BM by G-CSF and accelerated recruitment of these neutrophils to the lung by KC.

We interrogated whether adaptor deficiency in hematopoietic cells was the primary cause of uncontrolled inflammation and lethality. We generated Card9^−/−/WT bone-marrow chimera mice and challenged them with virulent MTB. Transfer of CARD9 KO cells into irradiated WT mice rendered recipients susceptible to TB, demonstrating a critical role of hematopoietic rather than radio-resistant cells in control of TB in absence of CARD9 (Fig. 6 A). Because of increased PMN recruitment to lungs of Card9^−/− mice we compared expression of CARD9 in PMN versus other myeloid cells and alveolar epithelial cells (AECs). PMN abundantly expressed CARD9 transcripts in steady-state, whereas BMDCs and lung macrophages were similar in terms of adaptor transcriptional activity (Fig. 6 B). AECs showed negligible transcriptional activity. After MTB encounter, CARD9^−/− PMN released less TNF compared with WT controls, but apparently did not support mycobacterial growth (Fig. 6, C and D). PMN isolated from KO mice had lost their propensity to secrete IL-10 when challenged with MTB (Fig. 6 E). We conclude that CARD9-deficient animals fail to fine-tune lung inflammation by means of IL-10.

To further assess whether neutrophilic hyperreactivity and subsequent tissue damage accounted for lethality of MTB infected Card9^−/− mice, we designed rescue experiments. Two different strategies were followed. First, we used the cholesterol-lowering agent lovastatin to accelerate removal of apoptotic cells in the lung (Morimoto et al., 2006). Impaired removal of apoptotic cells causes secondary necrosis, thus triggering tissue damage (Kono and Rock, 2008). Second, we modulated inflammatory responses by either neutralizing G-CSF or depleting PMNs with anti Gr-1 monoclonal antibody (mAb). Treatment with lovastatin did not improve survival. In contrast, granulocyte-limiting approaches prolonged survival of MTB-infected Card9^−/− mice (Fig. 7 A). Although bacterial burdens were not significantly altered after treatment at 21 (G-CSF neutralization, preliminary data) or 25 d p.i. (PMN depletion; Fig. S7), survival was prolonged and accompanied by ameliorated lung inflammation, in 60–70% of the treated animals, and fewer MPO-positive cells were recruited to the lungs (Fig. 7, B and C). Remarkably, both treatments reduced circulating levels of MPO and granulocytes (Fig. 7, D and E). Thus, limiting granulocyte infiltration of the lung dampened inflammation and prolonged survival of susceptible mice.

Next, we interrogated whether uncontrolled inflammation could be brought to an end by anti-TB therapy. In preliminary experiments, we used the first-line TB drugs isoniazid (INH) and rifampicin (RIF) to reduce bacterial burden (Fig. 8, A–C). Despite diminished susceptibility, there were 30% death cases during the first 10 d when chemotherapy was started 3 wk p.i. (Fig. 8 A). When drug treatment started 2 wk p.i., however, KO animals controlled TB infection (Fig. 8 D). MTB burdens in lung at the end of antibiotic treatment and 2 wk thereafter were more affected by INH/RIF in Card9^−/− compared with WT mice (Fig. 8 E). We assume that reduced immune pressure in absence of CARD9 allowed for more metabolically active, replicating MTB with higher susceptibility to drug treatment. Collectively, mechanisms that fine-tune host inflammation and those controlling bacterial replication are both needed to overcome susceptibility to TB in the absence of CARD9.

The first events after MTB has reached the alveolar space of the host lung include its recognition by alveolar macrophages via PRRs. They recruit adaptor molecules that transduce signals culminating in transcriptional responses that determine the type of the ensuing immune response. The adaptor molecule CARD9 is expressed in myeloid cells (Bertin et al., 2000) and in lung tissue (Hsu et al., 2007). Here, we demonstrate that: (a) CARD9 adaptor was central for sampling of signals from various PRRs involved in MTB recognition; (b) CARD9 was essential for control of pulmonary TB; (c) uncontrolled MTB replication and excessive lung inflammation resulted in lethal outcome of MTB infection in Card9^−/− mice. We conclude that CARD9-dependent tight control of, and by, PMN are essential for innate immunity in TB. Our data thus emphasize the double-edged role of innate immunity in protection and pathology in TB.

Disruption of TLRs impairs long-term control of TB (Hölscher et al., 2008) and deficiencies in IL-1R (Fremond et al., 2007), MyD88 (Fremond et al., 2004), or TIR8 (Garlanda et al., 2007) result in fatal acute TB disease. Animals devoid of TIR8, a negative regulator of TLR signaling, or D6, a decoy and scavenger receptor for inflammatory CC chemokines, suffer from uncontrolled systemic inflammation, although they normally control bacterial replication (Garlanda et al., 2007; Di Liberto et al., 2008). These findings indicate that control of both bacterial growth and inflammation are decisive determinants of disease outcome. MTB-infected Card9^−/− mice suffered concomitantly from higher MTB burdens in lungs and a systemic inflammatory response associated with profound lung pathology. Pyogenic pneumonia in Card9^−/− mice resembled lesions observed in MTB-infected MyD88 and IL-1R KO mice (Fremond et al., 2004, 2007) or mice deficient for Ipr1 gene within the sst1 locus (Pan et al., 2005). Despite excessive cell death in the lung, Card9^−/− BMDMs were equally prone to apoptosis as their WT counterparts in contrast to the sst1-susceptible macrophages (Pan et al., 2005), suggesting that CARD9 is not directly involved in apoptosis control. In infected Card9^−/− mice, increased apoptosis in lungs was not compensated by cellular replacement, similar to some extent to mice deficient in the IL-1–MyD88 pathway (Fremond et al., 2004, 2007).

Deficient IL-1 signaling accounts for most of the MyD88 KO phenotype because of a shared signaling pathway (Fremond et al., 2007). CARD9 does not share any molecules involved in signal transduction with either IL-1R or the MyD88 adaptor, but regulates IL-1β release at APC level. Its propensity to regulate inflammasome function was recently revealed during fungal infection (Gross et al., 2009). Accordingly, we exclude a major impact of MyD88 on CARD9 signaling. Instead, we consider more likely a regulatory loop via modulation of IL-1β release by CARD9 signaling. Consistent with our assumptions, sensing of MTB by dectin-1 and NOD2 is MyD88-independent. Our in vitro data point to a syk dependency of dectin-1 signaling, whereas NOD2 signaling involves stable ubiquitination of RIP2 (Yang et al., 2007). Card9^−/− mice up-regulated iNOS in inflamed lungs and Card9^−/− BMDMs controlled MTB growth in vitro and produced NO similar to WT cells. Moreover, LRG-47 expression after MTB infection was apparently not impaired in Card9^−/− mice (Fig. S8). Thus, increased susceptibility of Card9^−/− mice was not caused by defective production of these antimicrobial effector molecules. Yet, CARD9 was required for an efficient innate activation program in vitro and for antimycobacterial immune homeostasis in vivo. We observed reduced cytokine synthesis by KO cells after MTB encounter. Reduced TNF production could explain defective granuloma formation, at least in part (Bean et al., 1999; Chakravarty et al., 2008). Diminished IL-12 synthesis by macrophages was reproduced in DCs, but could not be detected in infected tissue. Because of similar lung levels of IL-12, critical for Th cell polarization (Khader et al., 2006), apparently unrestrained Th1 responses could develop in absence of CARD9 signaling during TB. Card9^−/− mice suffered from higher bacterial titers as early as 3 wk p.i. In parallel, the animals exhibited pyogranulomatous infiltrates rich in neutrophils in the lung, increased death, and impaired removal of dead cells from the lung parenchyma. Fatal disease progression in the absence of CARD9 signaling was thus accompanied by exacerbated inflammation induced by granulocytes.

The relevance of PRRs for initiating and shaping adaptive immunity is beyond doubt (Medzhitov, 2007; Dorhoi and Kaufmann, 2009). Strikingly, absent CARD9 signaling did not seem to affect T cell responses to MTB, suggesting it is dispensable for development of appropriate T cell responses in TB. Card9^−/− mice developed normal Th1 responses to MTB in vivo despite reduced IL-12 production by APC in vitro. Moreover, we did not detect any deficiency in Th17 response to MTB, although involvement of the syk-CARD9 pathway in differentiation of IL-17-producing CD4⁺ T cells has been described (LeibundGut-Landmann et al., 2007). Neutrophil-mediated defense against C. albicans induced by efficient Th17 responses is CARD9 dependent. Mycobacteria are decorated with various pathogen-associated molecular patterns that trigger multiple PRRs simultaneously, whereas C. albicans is rich in β-glucans, which are sensed by CLRs. Simultaneous engagement of an array of PRRs likely has a different outcome (Seimon et al., 2006; Trinchieri and Sher, 2007). Hence, multiple receptor recognition of MTB cannot be mimicked by single receptor–ligand interactions, such as synthetic TDM derivatives. Our Th17 data during MTB infection are consistent with a recent study by Werninghaus et al. (2009), although the two studies investigated different tissue sites. The differences in IFN-γ producers in the lung tissue was not confirmed in our study, although overall lung IFN-γ abundance did not differ (Fig. 5, C and D). Notably, both studies point to efficient T cell activation in absence of CARD9. Strikingly, despite using a similar dose of MTB laboratory strain, the CARD9 KO animals survived longer in this study (Werninghaus et al., 2009). This could be explained by differential virulence of the MTB stocks used. In any case intact T cell immunity failed to overcome the severe innate defects of Card9^−/− mice in MTB infection. On the contrary, protective T cell responses were efficiently stimulated by BCG vaccination. CARD9 KO mice not only developed DTH, but also protection against MTB. We conclude that in the absence of CARD9, innate inflammation cannot be controlled although apparently adequate T cell responses can develop.

MTB recognition by different TLRs (Quesniaux et al., 2004; Schnappinger et al., 2006) and NOD2 (Ferwerda et al., 2005; Gandotra et al., 2007; Yang et al., 2007) is well established. Moreover, evidence for a role of CLRs in MTB sensing has been obtained (Yadav and Schorey, 2006; Rothfuchs et al., 2007). The complexity of the MTB cell wall requires combinatorial recognition and engagement of different types of PRRs, as emphasized by our experiments. Further, our data establish a critical involvement of CARD9 signaling in cytokine secretion by BMDMs in response to MTB and evidence that MTB sensed by dectin-1 and NOD2 uses CARD9 as an adaptor molecule. Involvement of other ITAM-based molecules that signal via CARD9, including Mincle, a CLR member recently proposed to sense mycobacterial TDM (Ishikawa et al., 2009), cannot be excluded. Convergence of multiple PRR pathways on the CARD9 adaptor minimizes metabolic expenses for the host. On the other hand, convergence increases host vulnerability. Consistent with recent gene expression profiling results, demonstrating that CARD9 is modulated in human DCs early after MTB infection (Tailleux et al., 2008), our data emphasize the need to clarify whether polymorphisms in CARD9 influence susceptibility to MTB in humans. Recent genetic analysis extended the link between fungi detection and CARD9 signaling from the mouse to the human system. Homozygous CARD9 mutation was found to predispose humans to chronic mucocutaneous candidiasis (Glocker et al., 2009). Based on our observations in TB-infected naive and BCG-vaccinated Card9^−/− mice, and because of the fact that in TB-endemic regions vaccination programs are in place, genetic screening for an association between CARD9 polymorphisms and susceptibility to TB is challenging but necessary.

Lung damage in Card9^−/− mice was governed by augmented granulopoiesis, neutrophilic inflammation, and apoptotic cell death which progressed to secondary necrosis. Pathology was correlated with local production of G-CSF, KC, and MCP-1 and systemic release of G-CSF. These cytokines are produced by myeloid, endothelial, and epithelial cells and stimulate granulopoiesis and attract granulocytes to tissue sites. G-CSF is induced by various cytokines and bacterial products and stimulates differentiation, maturation, and functional activation of neutrophilic granulocytes. Evidence is accumulating that acute mobilization of PMNs from the BM is induced by simultaneous release and action of G-CSF and KC (Wengner et al., 2008). This cytokine millieu occurred in TB-infected mice in absence of CARD9 and induced both release of granulocytes into the periphery and their accelerated recruitment to the lung. Transcriptional regulation of KC was augmented early after infection and paralleled PMN-specific transcripts. Neutrophil-specific genes are actively transcribed during differentiation in BM under steady-state conditions (Borregaard et al., 2007). However, during infectious stress and high neutrophil demand immature granulocytes are released, and hence PMN transcripts can be detected in infected tissue sites. Recent studies revealed that granulopoietic emergency program depends on IL-1R signaling and that low IL-1β levels could induce late neutrophilic responses (Ueda et al., 2009). Although we did not detect significant in vivo IL-1β differences between the CARD9 KO mutants and congenic controls during TB, the potential role of the CARD9 adaptor in hematopoetic stem cell and progenitor proliferation remains to be established. Card9 deficiency reduces nuclear translocation of NF-κB (Gross et al., 2006), and KC comprises NF-κB–binding sites in its promoter (De Filippo et al., 2008). Reduced proinflammatory cytokine responses in MTB-infected macrophages in vitro support deregulated KC synthesis. Diminished IL-6 production could directly impair KC abundance because gp130-dependent STAT-3 signaling negatively regulates KC responses and PMN clearance (Fielding et al., 2008). Whether lung PMN kinetics is directly affected or whether the emergency granulopoietic program is distorted remains to be clarified. Beneficial effects of PMN depletion or G-CSF neutralization support pathological functions of the granulocytes during acute TB in the absence of CARD9. Importantly, previous studies on transient PMN depletion by anti–Gr-1 mAb in WT mice (Seiler et al., 2000; Keller et al., 2006) did not observe major effects on TB, as seen here for CARD9 KO mutants (Fig. 7 A). Although inflammatory monocytes are Gr-1⁺ (Geissmann et al., 2003) and mAb treatment could have affected this population, available data rule out a major involvement of these cells in TB. Mice deleted in CCL2 (Kipnis et al., 2003) or CCR2 (Scott and Flynn, 2002), which are required for tissue recruitment or egress from BM of inflammatory monocytes, respectively, controlled MTB infection efficiently. We conclude that exacerbated lung neutrophilic inflammation represents a hallmark of lung pathology in MTB-infected Card9^−/− mice. Although PMN depletion reduced inflammation and prolonged survival, it did not considerably impact on bacterial burdens in lung, suggesting that inflammation alone triggered lethality. However, reduction of MTB loads by drug treatment during polarized PMN inflammation rescued Card9^−/− mice from lethal outcome.

The role of neutrophils in TB remains controversial (Seiler et al., 2000; Eruslanov et al., 2005; Keller et al., 2006). Our data reveal that PMN abundantly transcribe CARD9, yet reveal a detrimental role of granulocytes in TB. Our observation is consistent with a recent study describing a previously unknown role of PMN in dampening inflammation via IL-10 (Zhang et al., 2009). We demonstrate that CARD9 is essential for IL-10 synthesis in MTB-infected PMN. Absent IL-10 in CARD9 KO animals abrogated negative feedback regulation of various cytokine and chemokine genes. Control by both MyD88 and CARD9 of IL-10 production in PMN during TB could explain the similar pathology of MTB-infected KO mice deficient in CARD9 or MyD88. Neutrophilic inflammation of the lung is followed by defective removal of apoptotic cells, termed efferocytosis. Studies on chronic lung inflammation emphasize an important role of early efferocytosis for maintenance of biological lung functions (Vandivier et al., 2002). Moreover, a detrimental role of so-called “frustrated” PMN, unable to meet the pathogen in inflamed tissue, has been described (Nathan, 1987). In MTB-infected Card9^−/− mice, collateral damage during secondary necrosis and systemic release of MPO became life threatening. Neutrophils undergoing necrosis leak enzymes and noxious agents that damage the surrounding tissue without contributing to MTB control. Notably, granulocyte-derived MPO causes cytotoxicity in alveolar macrophages (Grattendick et al., 2002). Increased levels of circulating MPO were recently reported in patients with active lung TB (Koziol-Montewka et al., 2004). As a corollary, we assume absence of a regulatory loop via lack of IL-10 and a circulus vitiosus with granulocytes initiating inflammation and ineffective efferocytosis boosting inflammatory responses. Collectively, we conclude that both fine-tuning of acute PMN inflammation and control of MTB replication in lung tissue are responsible for lethal TB in absence of CARD9 signaling. The molecular mechanisms underlying unleashed bacterial growth remain to be deciphered. Susceptibility is apparently independent from control by the acquired T cell response and can be overcome by vaccination. Thus, the key adaptor CARD9 plays an essential role in autonomous orchestration of innate immunity in TB.

The following polyclonal antibodies were used for immunostainings: MPO (Abcam), iNOS (LabVision), and goat anti–rabbit (Dianova). We used the following mAb for flow cytometric investigations: CD4 (RM4-5), IFN-γ (XMG-1.2), IL-2 (JES6-5H4), Ly6G/C (RB6-8C5), CD11b (M1/70; eBioscience), CD8α (YTS169), TNF (XT22), CD3 (145-2C11), CD28 (37.51), CD62L (MEL-14), CD25 (PC61), CD16/CD32 (2.4G2), F4/80 (CI:A3-1), and CD11c (N418; antibodies purified from hybridoma supernatants; American Type Culture Collection), IL-17 (TC11-18H10; BD Biosciences). PG, mAGP, AG, MAME, and manLam were derived from MTB (Colorado State University). LPS and zymosan were obtained from Invivogen; piceatannol and lovastatin were obtained from Calbiochem; TDM, cytochalasin, and laminarin were obtained from Sigma-Aldrich. DNaseI and collagenase D were obtained from Roche. CD11c beads were used for selection of differentiated BMDCs (Miltenyi Biotec).

WT (C57BL/6) and Card9^−/− mice (Gross et al., 2006) were bred at III. Medizinische Klinik, Klinikum rechts der Isar, Technische Universität München, Germany. The Card9^−/− mouse strain was backcrossed onto the genetic background of C57BL/6 for more than six generations. Mice were 8–12 wk of age at the beginning of the experiments, matched for age and sex, and kept under specific pathogen-free (SPF) conditions at the Max Planck Institute for Infection Biology in Berlin, Germany. Infection of mice was performed using a Glas-Col inhalation exposure system. The initial challenge dose was verified 24 h p.i. by plating complete lung homogenates onto Middlebrook 7H11 agar plates. At different time points, bacterial burdens were assessed by mechanical disruption of the organs in water/1% wt/vol BSA/1% vol/vol Tween 80 (WTA), and plating serial dilutions onto Middlebrook 7H11 agar plates. After 3 wk, MTB colonies were counted. For PMN depletion, mice were treated i.p. with anti-Gr-1 antibodies (Rb6-8C5) in PBS 200 µg/mouse/shot or with control rat IgG at days 8, 11, and 14 p.i. G-CSF neutralization was achieved by i.p. injection of 100 µg/mouse/shot anti–mouse G-CSF (R&D Systems) at days 10 and 14 p.i. Mice were treated with first-line TB drugs INH (0.1 g/liter) and RIF (0.1 g/liter; Fluka) in drinking water for 6 wk. Antibiotic treatment started at different time points p.i., as indicated in text or figure legend. Animal protocols were approved by Landesamt für Gesundheit und Soziales, Berlin, Germany.

Mice were s.c. inoculated with 10⁶ CFUs BCG Danish 1331 (SSI). At 3 wk after vaccination, BCG-vaccinated and control mice were i.d. challenged with 2 µg PPD into the footpad (SSI). Foot pad swelling was recorded 48 h after challenge using a dial gauge caliper.

WT and Card9^−/− mice at 8 wk of age were maintained on antibiotics (0.1 mg/ml Ciprofloxacin and 2 mg/ml Neomicin) 1 wk before irradiation. Irradiation was performed using a dose of 4 Gy, and animals were further kept for 4 wk on antibiotics in drinking water. After withdrawal of antibiotics, complete repopulation of the organs with immune cells was allowed for another 4 wk, and, subsequently, leukocytes were PCR typed to control efficacy of reconstitution. MTB aerosol infection was performed (as described in the Mice and MTB infection section) at 10 wk after sublethal whole-body irradiation and complete reconstitution with donor leukocytes.

BMDMs and BMDCs were obtained from tibia and femur bones and maintained in Dulbecco's Modified Eagle Medium containing 20% L929-cell supernatant, 10% heat-inactivated FCS, 5% heat-inactivated HS, 2 mM glutamine or RPMI supplemented with 10% heat-inactivated FCS and 200 U/ml recombinant mouse GM-CSF, respectively. Differentiated resting cells and cells pretreated with recombinant mouse IFN-γ (100 U/ml; Strathmann Biotech AG) were infected with MTB H37Rv at MOI 5 or treated with MTB-derived compounds. MTB-derived compounds were solubilized in PBS unless otherwise stated. PG and mAGP were sonified for better solubilization. TDM solution was prepared by repeatedly warming the solution in PBS containing 2% DMSO at 50°C, followed by repeated vortexing. PMN were isolated from BM as previously described (Allport et al., 2002). For in vitro infection experiments, MTB was grown in Middlebrook 7H9 broth (BD) supplemented with 0.05% glycerol and Tween 80, and 10% ADC enrichment (BD) to an early log phase and washed with PBS, and single bacteria were resuspended in culture media at desired concentrations. After a 4-h incubation, noninternalized bacteria were washed away and incubation continued as indicated in the results. For blocking experiments, we used laminarin (500 µg/ml) and piceatanol (1–10 µM). MTB was heat inactivated at 80°C for 40 min. Bacterial growth was assessed by [³H]Uracil incorporation (Flesch and Kaufmann, 1987). Concentration of TNF, IL-6, CCL5, IL-10, and IL-12p40 in cell culture supernatants was determined 24 h p.i. or after stimulation using commercial ELISA kits (R&D Systems). Apoptotic rates were measured using Cell Death Plus Detection kit ELISA (Roche) according to the manufacturer's instructions. Nitric oxide was determined using the Griess reagent (Sigma-Aldrich).

Sera and lung homogenates from MTB-infected mice were assayed for cytokine and chemokine levels using multiplex bead-based immunoassay kits (Bio-Plex Cytokine Assay; Bio-Rad Laboratories) according to manufacturer's instructions.

Lung tissue or, alternatively, cell pellets were homogenized in Trizol and mRNA was extracted using acidic phenol chloroform extraction. RNA integrity and the absolute amount of total RNA were verified with a Bioanalyzer 2100 (Agilent Technologies). Expression levels of Cxcl1, Ngp, Mpo, Ctsg, Irgm1, LRG-47, Card9, and housekeeping gene GAPDH were analyzed using QuantiTect primer-probes (QIAGEN). Fold changes were calculated with the ΔCT method using uninfected samples as calibrator for tissue analysis or normalized to GAPDH with the equation 1.8^{(GAPDH-CARD9)} × (1,000,000) to evaluate CARD9 expression in different cell types.

Enzyme levels in sera from MTB-infected mice were determined using a commercial ELISA, according to manufacturer's instructions (Hycult Biotechnology).

Organs were fixed with 4% wt/vol PFA for 24 h and embedded in paraffin. Sections (2 µm) were Giemsa stained. Mycobacteria in the lungs were visualized by acid-fast staining (TB Stain kit; BD) according to manufacturer's instructions. For immunohistochemistry, fixed and rehydrated tissue sections were subjected to antigen retrieval, blocking and exposed to primary antibody, followed by goat anti–rabbit AP secondary antibody. Alkaline phosphatase activity was visualized using fuchsin (Dako). Sections were counterstained with hematoxylin. TUNEL staining was performed according to the vendor's protocol (in situ detection kit TMRed; Roche).

Purification of cells from lungs and spleens of MTB-infected mice was performed as previously described (Kursar et al., 2007). APC/phagocyte influx to lungs was assed by measuring frequencies of cells positive for the following markers: CD-11b, Ly6G, CD-11c. FoxP3⁺ T cell frequencies were measured using the Foxp3 Staining Set (eBioscience). For intracellular cytokine staining (ICCS) cells were stimulated with polyclonal activators (anti-CD3/CD28 antibodies) or a pool of MTB-derived peptides (Ag85A_241-260 [QDAYNAGGGHNGVFDFPDSG], Ag85B_240-260 [FQDAYNAAGGHNAVFNFPPNG], Ag85A_261-280 [THSWEYWGAQLNAMKPDLQ], TB10.4_4-12 [IMYNYPAML], ESAT6_1-20 [MTEQQWNFAGIEAAASAIQG], and Mtb32/RV0125_309-318 [GAPINSATAM]). All peptides were obtained from JPT Peptide Technologies and used at 5 × 10⁻⁵ M concentration for 6 h in the presence of 25 µg/ml brefeldin A. Nonspecific background activation was determined by culturing cells in RPMI only. After washing and blocking unspecific antibody binding, we performed surface staining for CD8α, CD4, and intracellular staining for the following cytokines: TNF, IL-17, IL-2, and IFN-γ. Cells were analyzed using a FACSCanto II flow cytometer and FACSDiva software (BD).

PMN were isolated from buffy coats using the reversal method. Ficoll gradient separation was followed by Dextran sedimentation and osmotic lysis of remaining erythrocytes. Overnight apoptotic PMNs were co-cultured with BMDMs at a ratio 5:1 for 1 h. BMDM monolayers were washed extensively with PBS and fixed with 2.5% glutaraldehyde. Internalized apoptotic PMNs were visualized by adding MPO substrate o-dianisidine (Sigma-Aldrich) and microscopic examination.

GFP-expressing BCG microorganisms were grown in Middlebrook 7H9 broth supplemented with glycerol and 0.05% Tween 80, and 10% ADC enrichment medium to an early log phase, washed in PBS, and single bacteria were resuspended in culture media at desired concentrations. Differentiated BMDMs and BMDCs were incubated at different MOI with fluorescent bacteria for 2 h. Cells were washed extensively with PBS, detached and stained for specific cell markers. APCs were analyzed using a FACSCanto flow cytometer and FACSDiva software.

PRISM GraphPad software was used for statistical analysis. Bacterial titers were analyzed by the Mann-Whitney test, survival was assessed by Kaplan-Meier curves and log-rank test. Frequencies and numbers of APC, cytokine-positive T cells, concentrations of cytokines and chemokines were compared using a two-way ANOVA followed by Bonferoni's post-test. In vitro cytokine levels were compared using t-student test. Only P values ≤ 0.05 were considered statistically significant.

Fig. S1 shows that Card9^−/− mice gradually lost body weight during acute TB infection. Fig. S2 depicts spleen pathology and indicates that despite increased susceptibility to TB, Card9^−/− mice had serum NO levels comparable to WT controls. Fig. S3 shows apoptotic PMN clearance capacities of Card9^−/− and WT macrophages. Fig. S4 indicates impaired cytokine production in MTB-infected BMDCs from Card9^−/− mice. Fig. S5 shows that laminarin treatment did not interfere with mycobacterial uptake and indicates that Card9^−/− BMDMs and DCs were impaired in mAGP detection. Fig. S6 shows that Card9^−/− mice mounted apparently intact T cell responses after TB challenge. Fig. S7 indicates that PMN depletion did not impact on bacterial burdens, but improved clinical disease. Fig. S8 shows that Card9^−/− mice were not impaired in LRG-47 regulation in the lung during TB infection.

We thank M.L. Grossman and Diane Schad for help in preparing the manuscript. MTB-derived PG, mAGP, MAME, and AG were received as part of National Institutes of Health, National Institute of Allergy and Infectious Disease contract no. HHSN266200400091C, entitled “Tuberculosis Vaccine Testing and Research Materials,” which was awarded to Colorado State University.

This work received financial support from the German Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung) Kompetenznetzwerk “PathoGenoMikPlus” to S.H.E. Kaufmann.

The authors have no conflicting financial interests.