Adjuvanticity of a synthetic cord factor analogue for subunit Mycobacterium tuberculosis vaccination requires FcRγ–Syk–Card9–dependent innate immune activation

To characterize the adjuvant activity of TDB, we first immunized mice with H1 together with TDB or the TLR9 ligand CpG oligodeoxynucleotide (ODN) 1826, and investigated the priming of Th-1 and -17 T cell responses by restimulating lymph node cells with H1 in vitro. Both TDB and CpG induced a robust Th-1 T cell response characterized by profound antigen-specific IFN-γ production (Fig. 1 A). However, only TDB, and not CpG, also activated a strong H1-specific Th-17 T cell response, indicating that this glycolipid has an adjuvant activity that differs fundamentally from that of a classical TLR agonist (Fig. 1 A). Because the immunostimulatory capacity of TDB presumably depends on direct effects on the innate immune system, we next analyzed the ability of TDB to activate APCs in vitro.

As an initial readout for cell activation, we used nitric oxide (NO) production from BM-derived macrophages (BMMs; Fig. 1 B). TDB, as well as the cord factor TDM, induced substantial amounts of NO, especially after priming with IFN-γ (Fig. 1 B). However, compared with CpG, the kinetics of TDB-mediated NO production was delayed (Fig. 1 C). To investigate the differences in the effects of TDB and CpG on myeloid cell activation more generally, we next measured global changes in gene expression over time. As expected, both stimuli induced substantial transcriptional reprogramming, again with an overall slower response to TDB (Fig. 1 D). We also observed qualitative differences in gene expression, as a hierarchical clustering revealed TDB-selective (cluster C) or CpG-selective gene induction (cluster D) in addition to common down- and up-regulated target genes (cluster A, B1) and genes induced early by CpG and later by TDB (cluster B2). Selected examples of TDB-induced genes were confirmed at the protein level, validating the microarray data (Fig. S2, available at http://www.jem.org/cgi/content/full/jem.20081445/DC1). We then focused on differentially expressed cytokines that can direct Th cell differentiation (Fig. 1, E–G). IL-12p40 was strongly induced by CpG but not by TDB (Fig. 1 E), whereas IL-6 and IL-1β were expressed transiently after CpG and delayed after TDB treatment (Fig. 1, F and G). Notably, TDB, but not CpG, triggered a significant release of IL-1β into the culture supernatant (Fig. 1 F). We also analyzed the ability of TDB to activate DCs as the most potent APC. We found that TDB caused up-regulation of MHC-II and of the costimulatory molecules CD40, CD80, and CD86 on the cell surface (Fig. 1 H) and triggered cytokine and NO release in DCs in a manner similar to that in BMMs (Fig. S3). Thus, the adjuvant TDB potently activates APCs, in which it triggered a different expression program compared with TLR stimulation and presumably activates a distinct intracellular signaling pathway.

Diverse pattern recognition receptor families use different proximal signaling proteins to mediate myeloid cell activation. Whereas most TLRs signal via the adaptor protein MyD88 (3), certain ITAM receptors or ITAM-coupled receptors like C-type lectins use the kinase Syk for downstream signaling (14). To study whether any of these pathways are required for TDB-mediated cell activation we used MyD88^−/− and Syk^−/− BMMs (Fig. 2, A and B). As expected, control stimulation with TLR agonists resulted in normal production of NO in Syk^−/− cells, but was ineffective in MyD88^−/− BMMs. In contrast, TDB and TDM induced NO production in MyD88^−/− cells to a similar extent as in wild-type cells, but did not stimulate NO synthesis in the absence of Syk. Furthermore, the TDB-stimulated expression and production of IL-6 by Myd88^−/− BMMs was largely intact (Fig. S4, available at http://www.jem.org/cgi/content/full/jem.20081445/DC1), whereas Syk^−/− BMMs failed to produce IL-6 and -1β (Fig. S5). As the small molecule Syk inhibitor Piceatannol blocked NO release induced by TDB and TDM from BMMs in a dose-dependent fashion (Fig. 2 C and Fig. S6), the enzymatic function of Syk is essential for the Myd88-independent innate immune cell activation in response to these glycolipids in vitro. Interestingly, Geisel et al. reported that Myd88 is required for cord factor–induced inflammation in vivo (9). Further, we have recently found that Myd88 is required for the Th1-directing adjuvant effect of TDB in vivo (15). Because TLR2, TLR3, TLR4, and TLR7 were not required (15), the phenotype of Myd88 knockout mice in vivo may indicate a defect in the glycolipid-induced effector phase, e.g., downstream of the IL-1 receptor.

Syk can use the myeloid cell–specific adaptor protein Card9 for the activation of macrophages or DCs in response to various signals (4–6). We therefore used Card9^−/− BMMs to test whether APC activation by TDB requires this adaptor protein. Similar to Syk^−/− BMMs, cells lacking Card9 failed to produce NO in response to TDB (Fig. S7, available at http://www.jem.org/cgi/content/full/jem.20081445/DC1) and to up-regulate expression of iNOS, IL-6, or IL-1β, whereas the CpG-mediated expression of these molecules was largely unaffected (Fig. 3 A). Gene expression profiling was used to determine whether Card9 is required for all transcriptional responses to TDB stimulation (Fig. 3 B). The β-glucan Curdlan, which is a ligand of the C-type lectin receptor Dectin-1 that is known to signal via Card9 (5), and TDB induced very similar genome-wide transcriptional responses that were distinct from the pattern of CpG-induced activation. In the absence of Card9, responses to TDB and Curdlan were almost completely abrogated, whereas CpG-triggered gene expression was unaffected. To further study the function of Card9 in TDB-induced APC activation, we also analyzed DCs and found that the release of IL-1β, IL-6, and TNF-α, as well as nuclear translocation of the transcription factor NF-κB, was triggered by TDB in a strictly Card9-dependent manner (Fig. S8 and Fig. S9). In contrast to BMMs (Fig. 1 E), DCs also produced IL-12p40 after stimulation with TDB or Curdlan, as well as IL-12p70 and -23, which are all dependent on Card9 (Fig. 3 C). Notably, LPS was a far stronger inducer of these Th differentiation factors than TDB (Fig. 3 C). Curdlan was comparably weak in IL-12 induction, in agreement with a previous study (5). It is known that Card9 can couple to the downstream adaptor and scaffold proteins Bcl10 and Malt1 (4). Using TNF-α production by DCs as readout for Bcl10 and Malt1 signaling, we found an additional requirement for these molecules in the response to TDB and TDM (Fig. 3 D). Thus, these experiments demonstrate on a genetic and transcriptional level that TDB and TDM engage a specific innate pathway that depends on Syk kinase activity and the adaptor proteins Card9, Bcl10, and Malt1.

As the β-glucan receptor Dectin-1 has already been implicated in APC responses to whole mycobacteria (16–18), we tested its role in recognition of the mycobacterial cord factor TDM or TDB by APC. Using IFN-γ–primed Dectin-1^−/− BMMs, we found the expected defect in NO production after stimulation with Curdlan, but the response to TDB and TDM, as well as to the TLR stimuli CpG and LPS, was not affected (Fig. 3 E). In addition, the expression of Nos2 and IL-1β after stimulation with TDB, TDM, and LPS was independent of Dectin-1 (Fig. S10, available at http://www.jem.org/cgi/content/full/jem.20081445/DC1). Thus, Dectin-1 is not required for TDB-/TDM-induced macrophage activation. In contrast to Dectin-1, which possesses an ITAM-like motif in its cytoplasmic region, a large number of Syk-activating myeloid cell receptors are associated with one of the ITAM-bearing adaptor proteins Dap12 or FcRγ (encoded by the Fcer1g gene). To gain further insight into the signaling requirements of APC activation by TDB and TDM, we used Dap12^−/− and Fcer1g^−/− BMMs (Fig. 3, F and G). Expression of Nos2 was independent of Dap12 in response to TLR ligands, Curdlan, and TDB/TDM. In contrast, in Fcer1g^−/−, BMM expression of Nos2 and production of NO were specifically abrogated after stimulation with TDB or TDM, whereas TLR-triggered and Dectin-1–induced activation was intact. The specific requirement for FcRγ, but not Dap12, was also found for TDB-/TDM-induced expression of IL-6 (Fig. S11). Therefore, the ITAM-bearing adaptor FcRγ is critical for linking TDB and TDM recognition to the Syk–Card9 signaling pathway.

The Syk–Card9 pathway is a pattern recognition pathway that can couple to adaptive immunity (5). To investigate whether this innate pathway is responsible for the adjuvant activity of TDB in vivo, we immunized Card9^−/− animals with H1 and TDB or TDM. Card9-deficient mice were used as they have selective defects in myeloid cells but intact lymphocyte activation, in contrast to Syk^−/−, Bcl10^−/−, and Malt1^−/− animals (4, 19). Although the transient footpad swelling after injection of H1 in liposomes alone was comparable between wild-type and Card9^−/− mice, the inclusion of TDB (Fig. 4 A) or TDM (not depicted) into the liposomes caused an increase in footpad thickness that was completely abrogated in Card9-deficient mice.4 Moreover, TDB and TDM injection augmented draining lymph node cellularity in wild-type but not Card9^−/− mice, whereas the response to CpG was comparable between the two genotypes (Fig. 4 B). The TDM- or TDB-induced capacity of lymph node cells for IFN-γ production was reduced in Card9^−/− mice, indicating that Th-1 differentiation directed by these glycolipids depends on Card9 (Fig. 4 C). The development of a H1-specific IL-17 production conferred by TDB and TDM was completely abrogated, suggesting a crucial role of the Syk–Card9 pathway for TDB-/TDM-induced Th-17 differentiation (Fig. 4 C). Moreover, H1-specific IgG2a antibody levels were Card9 dependently enhanced by TDB (Fig. 4 D). Thus, using different readout systems for Th-1, Th-17, and antibody responses, we find that Card9 is essential to mediate the adjuvanticity of TDB and TDM in vivo.

To test whether Card9 signaling is also responsible for the immune protection after subunit vaccination with H1 plus TDB, which confers a similar reduction in pulmonary mycobacterial load as the live BCG vaccine (13), we challenged immunized mice with low-dose aerosol MTB infection and analyzed CFUs in the organs 6 wk later. As expected, compared with naive mice, vaccination with H1 in TDB-containing liposomes caused a significant reduction of mycobacterial load in the lungs of wild-type mice, whereas the liposomes alone had no significant effect (Fig. 4 E). In sharp contrast, immunization with TDB and H1 failed to reduce the number of mycobacterial CFUs in the lungs of Card9^−/− mice, demonstrating a key role of Card9-signaling in TDB-induced protective immunity against tuberculosis in vivo. This protective effect correlated with the Card9-dependent induction of antigen-specific Th-17 cells in the mediastinal lymph nodes after vaccination, but not with the frequency of H1-specific IFN-γ–producing CD4⁺ T cells (Fig. 4 F). Notably, lung mycobacterial load in nonimmunized Card9^−/− mice was higher than in wild types (P < 0.01), suggesting an additional role for the Card9 pathway in the natural resistance to MTB infection that deserves further detailed investigation and that could explain the substantially elevated numbers of IFN-γ–producing T cells in Card9^−/− mice (Fig. 4 F).

In conclusion, we report the molecular pathway for the adjuvanticity of TDM and TDB by demonstrating that these glycolipids specifically activate APCs via Syk–Card9–Bcl10–Malt1 signaling. The precise receptor(s) that recognize(s) TDB or TDM remain(s) to be determined. As we have excluded an essential role of Dectin-1, its contribution to the activation of APCs by whole mycobacteria (16–18) is likely caused by mycobacterial ligands distinct from TDM. Given the multitude of receptors that can activate Syk–Card9 in myeloid cells, our finding that the ITAM-bearing adaptor FcRγ, but not Dap12, is essential for activation of macrophages by TDB and TDM provides a starting point for the isolation of pattern recognition receptors for these mycobacterial glycolipids. Although Dap12-coupled molecules like TREMs, Sirpb, and PILR-b are unlikely to be involved, some C-type lectins (e.g., Mincle, Dectin-2, and DCAR), as well as other FcRγ-coupled proteins (e.g., OSCAR and PIRA), can be considered candidate receptors.

Our results further indicate that TDB and TDM constitute adjuvants with a now molecularly defined mode of action that is distinct from TLR-triggering adjuvants engaging Myd88 (CpG and IC31) (20–22) or Trif (Monophosphoryl Lipid A) (23). It is intriguing that the innate activation program induced by the mycobacterial cord factor and its analogue TDB instructs the development of the appropriate, protective Th-1 and -17 response. Consistent with recent findings that established the importance of Th-17 cells after subunit vaccination using a TDM-containing adjuvant (8), the degree of Th-17 cell induction correlated with protection to MTB-challenge in our experiments. According to the current concept of murine and human Th-17 differentiation (24, 25), TDB likely operates through the combined induction of IL-1β, -6, and -23 from APCs, whereas CpG may inhibit Th-17 differentiation through IL-12 and IFN-γ (5, 25). Our finding that the glycolipid adjuvants TDB and TDM mediate their protective effects via Syk–Card9 signaling provides a rational basis for the further development of efficient TB immunization strategies.

C57BL/6 mice were purchased from Harlan-Winkelmann. Breeding pairs for mice deficient in Myd88 were provided by S. Akira (Research Institute for Microbial Diseases, Osaka, Japan). Card9^−/−, Bcl10^−/−, and Malt1^−/− mice have been previously described (4, 19). All knockout mice used for immunization experiments were backcrossed to C57BL/6 for at least six generations. Syk^+/− mice were provided by V. Tybulewicz (National Institute for Medical Research, London, England, UK). Syk^−/− and Syk^+/+ fetal liver cells were used to generate BM radiation chimeras. Dap12^−/− and Fcer1g^−/− mice were used with permission from T. Takai (26, 27). Dectin1^−/− mice have been previously described (28). Animal experiments were approved by the Regierung von Oberbayern (AZ209.1/211-2531-33/05) and the Animal Research Ethics Board of the Ministry of Environment, Nature Protection and Agriculture, Kiel, Germany.

CpG ODN 1826 has been previously described (21) and was supplied in phosphorothioate form by Coley Pharmaceuticals. E. coli O55:B5 LPS was obtained from Sigma-Aldrich. TDB was purchased from Sigma-Aldrich or Avanti Polar Lipids. TDM was purchased from Sigma-Aldrich or Axxora. The purity of TDB and TDM was confirmed by mass spectrometry. For preparation of liposomes, dimethyldioctadecylammonium (DDA) bromide and TDB were purchased from Avanti Polar Lipids. The fusion protein of Ag85B and ESAT-6 (designated H1) was produced as a recombinant protein, as previously described (29). Piceatannol (Calbiochem) was solubilized in DMSO. Curdlan was purchased from Wako Chemicals. IFN-γ was purchased from Tebu-Bio.

Macrophages were differentiated from BM cells in cDMEM containing 10% L-cell conditioned medium for 6–7 d. After removal of nonadherent cells, macrophages were detached from Petri dishes using Accutase treatment, washed, counted, and replated at a density of 10⁶ cells/ml. Macrophages were rested overnight before stimulation, as indicated in the figure legends. Where indicated, cells were pretreated for 3 h with IFN-γ at a concentration of 10 ng/ml. For generation of BMDC, BM cells were cultured on Petri dishes for 7 d in cRPMI containing 20 ng/ml of GM-CSF (Peprotech). The cells were plated at a density of 5 × 10⁵ cells/ml and stimulated as indicated in the figure legends.

Stocks of the glycolipids TDB and TDM were prepared by solubilizing in PBS with 2% DMSO by alternating vortexing and heating to 50°C at a concentration of 2.5 mg/ml. Macrophages were stimulated using 1–50 µg/ml TDB or TDM as indicated in figure legends. In some experiments, the glycolipids were coated directly to the bottom of cell culture plates after the procedure described by Ozeki et al. (30). In brief, 2.5 mg/ml TDB or 1 mg/ml TDM were dissolved in isopropanol by heating to 60°C for 20 min, followed by vortexing for 5 min. The glycolipids were coated in a volume of 20 µl for 96-well plates and 40 µl for 48-well plates. Concentrations were calculated by dividing the amount of TDB/TDM in the wells by the culture medium volume.

The Griess assay was used to determine the reaction products of NO in culture supernatants of stimulated macrophages. Levels of cytokines were measured by sandwich ELISA using antibody pairs from R&D Systems.

Total RNA was prepared from stimulated macrophages using RNeasy Mini kit (QIAGEN) and measured on a Nanodrop spectrophotometer. RNA samples used for microarray analysis were quality controlled for integrity of the RNA using Bio-Rad Experion Standard Sensitivity Chips (Bio-Rad Laboratories).

Affymetrix GeneChips were used for genome-wide transcriptome analysis of 1 µg total RNA per sample, following the manufacturer's instruction for labeling and hybridization of the samples. In the first experiment measuring the kinetics of changes induced by CpG or TDB in Fig. 1, samples were processed using the GeneChip Expression 3′ Amplification One-Cycle Target Labeling kit. Biotinylated cRNA was then hybridized on MOE430 2.0 GeneChips that were stained, washed, and scanned after standard procedures. In the second experiment, comparing wild-type and Card9^−/− macrophages after a 48-h stimulation with CpG, Curdlan, and TDB (Fig. 3), samples were processed with the Whole Transcript Assay kit and hybridized to Mouse Gene 1.0 ST GeneChips, as detailed by the manufacturer. CEL files were processed for global normalization and generation of expression values using the robust multi-array average algorithm. To select regulated genes, the data were filtered by excluding all probe sets that over all conditions showed <10-fold (Fig. 1) or <5-fold (Fig. 3) change and an absolute difference between maximum and minimum expression values of <200 (Fig. 1) or <100 (Fig. 3). The probe sets passing these criteria were z-score normalized and subjected to hierarchical clustering using Spotfire DecisionSite Functional Genomics software. Tables S1 and S2 (available at http://www.jem.org/cgi/content/full/jem.20081445/DC1) list the expression data for the probe sets shown in Fig. 1 D and Fig. 3 D, respectively. The Cel files for both experiments were submitted to the public repository Gene Expression Omnibus (Accession nos. GSE10530 and GSE10532).

Expression levels of the housekeeping gene HPRT, the cytokines IL-6 and -1β, and iNOS were analyzed using primer/probe combinations selected from the Roche Universal Probe Library and are available upon request. Fold changes were calculated with the ΔΔCT method using the untreated sample as a calibrator.

Groups of 3–6 mice were immunized subcutaneously by footpad injection 2 or 3 times, with a 2-wk interval between injections. All mice were immunized with 2 µg of the vaccine antigen H1 emulsified in 100 µl of each adjuvant (50 µl per foot), comprising either 250 µg DDA liposomes alone, 250 µg DDA + 50 µg TDB, or 250 µg DDA + 50 µg TDM. DDA was suspended in sterile distilled water at 2.5 mg/ml, heated to 80°C for 20 min with continuous stirring in order for liposomes to form, and left to cool at room temperature. A stable formulation of DDA/TDB was prepared by the lipid film hydration method, as previously described (12). TDM was suspended in sterile distilled water containing 2% dimethylsulfoxide to a concentration of 5 mg/ml by repeated passaging through a fine-tipped pipette, followed by vortexing. This step was repeated 3 times before freezing the solution at −20°C until use. Before immunization, H1 was mixed in Tris buffer and TDM was added and finally mixed with DDA. For immunization with CpG as adjuvant, H1 was mixed in PBS with the ODN such that one dose (per footpad) contained 2 µg H1 protein and 5 nmol CpG.

Footpad thickness was determined with a caliper before and at several time points after injection. Baseline values were subtracted individually for both feet of each mouse.

4 wk after the first immunization, the draining lymph nodes of the mice were removed and passed through a 100-µm nylon cell strainer (BD). For restimulation, 3 × 10⁵ cells (in a total volume of 250 µl RPMI) were plated in triplicate in 96-well culture plates. H1 antigen was used at 1 µg/ml and 10 µg/ml final concentrations for restimulation. Wells containing medium only or 1 µg/ml ConA were used as negative and positive controls, respectively. Culture supernatants were harvested after 96 h incubation, and the amount of IFN-γ and IL-17 was determined by ELISA.

To evaluate the presence of H1-specific antibodies, mouse sera of the immunized mice were analyzed in triplicate in threefold dilutions (beginning at a 1/1,000 dilution). Reagents used for determining specific antibody titers were rabbit anti–mouse IgG1 and IgG2a^b (BD). Absorbance was read on an ELISA reader at 450 nm.

Before infection of experimental animals, stock solutions of MTB H37rv were diluted in sterile distilled water and pulmonary infection with 100 CFU/lung was performed using an inhalation exposure system (Glas-Col). 6 wk after infection, mice were sacrified to aseptically remove lungs and mediastinal lymph nodes. To determine bacterial loads, lungs were weighed and homogenized in PBS containing a proteinase inhibitor cocktail (Roche). 10-fold serial dilutions of organ homogenates were plated in duplicates onto Middlebrook 7H10 agar plates containing 10% OADC and incubated at 37°C for 19–21 d. Colonies on plates were enumerated and results expressed as log₁₀ CFU per organ. To analyze the frequency of Th-1 and -17 cells, CD4 T cells from mediastinal lymph nodes were enriched by magnetic cells sorting (Miltenyi). Detection of antigen-specific IFN-γ– and IL-17–producing CD4 T cells was conducted using ELISPOT assay kits from BD and R&D Systems, respectively. Doubling dilutions of CD4 T cells were incubated with mitomycin-D–inactivated (Sigma-Aldrich) splenocytes from uninfected wild-type mice and were restimulated with H1 at a concentration of 10 µg/ml in the presence of 10 U/ml recombinant mouse IL-2 (Peprotech). After development, spots were automatically enumerated using an ELISPOT reader (ELISpot 04 XL; AID Diagnostika), and the frequency of responding CD4 T cells was determined.

Fig. S1 gives the chemical structures of TDB and TDM. Fig. S2 depicts validation of microarray data. Fig. S3 shows production of TNF-α and NO by DCs in response to TDB and TDM. Myd88-independent expression and secretion of IL-6 by TDB-stimulated macrophages is shown in Fig. S4. Fig. S5 shows IL-6 and -1b ELISA data from Syk-deficient macrophages. Inhibition of TDB- and TDM-induced NO release by Piceatannol is depicted in Fig. S6. Fig. S7 depicts Card9-dependent NO production by macrophages. Fig. S8 shows the expression of cytokines in DCs in response to TDB. Fig. S9 is a NF-κB EMSA showing Card9-dependent activation in DCs. Fig. S10 gives qPCR data showing Dectin1-independent expression of iNOS, IL-1β, and IL-6 in response to TDB and TDM. Fig. S11 compares the requirement for Dap12 and Fcer1g in TDB-induced expression of IL-6.

Technical support by Barbara Bodendorfer and Angela Servatius is highly appreciated.

Work in the authors' laboratories is funded by grants from Deutsche Forschungsgemeinschaft (SFB-576 to J. Ruland, R. Lang, and H. Wagner; SFB 643 to F. Nimmerjahn and R. Lang; and HO 2145/3-1 to C. Hölscher), Max-Eder Programm der Deutschen Krebshilfe (J. Ruland), Bundesministerium für Bildung und Forschung (EXC306 to C. Hölscher), Bayerisches Genomforschungsnetz (F. Nimmerjahn), the Hungarian National Office for Research and Technology (grant NKFP-A1-2006-0069 to A. Mocsai), and the European Commission FP6 (TB-VAC grants to E.M. Agger, P. Andersen, H. Wagner, and R. Lang).

Potential conflict of interest: P. Andersen is a co-inventor of patents relating to cationic liposomes as vaccine adjuvants. The other authors have no competing financial interests.