NOD2 triggers an interleukin-32–dependent human dendritic cell program in leprosy

Identification of NOD2- and TLR2/1-induced functional pathways

To determine the specific and shared immune responses triggered by distinct PRRs recognizing a microbial pathogen, we activated human peripheral blood monocytes from five healthy donors with two defined mycobacterial ligands: MDP, which activates cytoplasmic NOD2 (NOD2 ligand, NOD2L) and the mycobacterial 19-kDa triacylated lipopeptide, which activates cell surface TLR2/1 (TLR2/1 ligand, TLR2/1L). The concentration of NOD2L (1 µg ml⁻¹) and TLR2/1L (1 µg ml⁻¹) used for these studies was determined by dose titration of the ligands, initially measuring IL-12 p40 release and defensin β4 mRNA as the immunologic readout (data not shown). To identify molecular pathways induced by these different ligands, we collected cells at 0, 6 and 24 h and isolated mRNAs to obtain gene expression profiles using Affymetrix microarrays. Genes with differential expression between the two stimuli were identified by fold change and Student’s t test P value thresholds (Supplementary Figs. 1 and 2).

We defined four main classes of genes on the basis of comparisons between activated monocytes and control samples grown in medium: (i) all genes induced by NOD2L versus medium control (NOD2L > medium, 1.5-fold, P < 0.05), (ii) all genes induced by TLR2/1L versus medium control (TLR2L > medium, 1.5-fold, P < 0.05), (iii) genes that were induced only by NOD2L but not by TLR2/1L (NOD2L > TLR2L: NOD2L / medium > 1.5 and TLR2/1L / medium < 1.5) and (iv) genes that were induced only by TLR2/1L but not by NOD2L (TLR2L > NOD2L: TLR2/1L / medium > 1.5 and NOD2L / medium < 1.5). There were 3,388 common gene probes induced by NOD2L and TLR2/1L. NOD2L and TLR2/1L also activated distinct gene sets, 1,482 genes specifically by NOD2L and 1,100 specifically by TLR2/1L (Supplementary Fig. 3 and Supplementary Tables 1 and 2).

We performed enrichment analysis using Ingenuity pathways analysis (IPA) of the gene expression data to identify the main canonical pathways associated with the NOD2L- and TLR2/1L-induced gene sets. The genes induced in monocytes were evaluated according to their statistical association with each of 340 canonical pathways. For the gene sets induced by NOD2L but not TLR2/1L, the DC-related lipid antigen presentation by CD1 pathway showed the fourth highest statistical association of the 340 canonical pathways (Fig. 1a and Supplementary Table 3; Benjamini-Hochberg corrected P value = 0.05). Furthermore, for the gene set induced by NOD2L versus medium, the DC-specific maturation pathway showed the third strongest association of the 340 canonical pathways (Fig. 1a and Supplementary Table 4). The association of NOD2L-induced genes with canonical pathways specific for DC function prompted a comparison of all four listed canonical DC pathways in the Ingenuity database. Both the lipid antigen presentation by CD1 and antigen presentation pathways were specifically associated with the NOD2L-induced gene sets as compared to the TLR2/1L-induced gene sets. In contrast, the two other canonical DC pathways, DC maturation and DC crosstalk, were significantly associated with both NOD2 and TLR2/1 activation (Fig. 1b). The lipid antigen presentation by CD1 pathway includes the group I CD1 antigen presentation molecules CD1A, CD1B and CD1C, all preferentially upregulated by NOD2L compared to TLR2/1L activation of monocytes, with the ratio of NOD2L versus TLR2/1L induction being more than 2 s.d. from the mean for each (Fig. 1c).

Activation of monocytes by NOD2L induces CD1b⁺ DCs

Given that TLR2/1L induces both CD1b⁺ DCs and CD209⁺ macrophages²⁴ and our bioinformatics finding that NOD2 preferentially triggers DC functional pathways, we compared the capacity of NOD2 versus TLR2/1 activation to trigger DC and macrophage differentiation. We activated peripheral blood monocytes with either TLR2/1L or NOD2L for 48 h and, subsequently, measured CD1b⁺ and CD209⁺ cells by flow cytometry. Although CD209 is expressed on DCs derived by treatment of monocytes with GM-CSF plus IL-4, it is not expressed on DCs derived by GM-CSF treatment alone²⁴. CD209 is expressed on macrophages, not DCs, in a variety of human tissues and diseases²⁴. In addition, TLR2/1-induced CD209⁺ cells were previously shown to express macrophage but not DC markers²⁴. Because low levels of CD1b are induced in monocytes cultured in medium containing 10% human serum, we performed additional experiments using fetal calf serum, with which there is minimal background induction of CD1b²⁴. The frequency of CD1b⁺ DCs was twofold greater from NOD2L versus TLR2/1L activation (Fig. 2a,b), whereas CD209⁺ macrophages were induced only by TLR2/1L (Fig. 2a,b) across a wide range of ligand concentrations (Supplementary Fig. 4). The expression of CD1a and CD1c was also greater on NOD2L- versus TLR2/1L-activated monocytes, consistent with the microarray results (data not shown). The percentages of CD40⁺, human leukocyte antigen-ABC⁺ (HLA-ABC⁺), HLA-DR⁺, CD86⁺ and CD80⁺ cells were similar (Supplementary Fig. 5). However, the mean fluorescence intensities (MFIs) of HLA-ABC, HLA-DR and CD86 were significantly greater on the NOD2L- versus the TLR2/1L-induced CD1b⁺ cells, as well as on resting monocytes (Fig. 2c), and they were consistent when comparing two different doses of each ligand (Supplementary Fig. 6).

To compare the functional capacity of NOD2L- versus TLR2/1L-differentiated CD1b⁺ DCs, we performed standard MHC class I and MHC class II antigen presentation assays. NOD2L- and TLR2/1L-induced CD1b⁺ DCs were generated and then isolated by immunomagnetic sorting with CD1b-specific monoclonal antibody (mAb). The sorted cells were ≥ 95% CD1b⁺ and reflected the phenotypic pattern of unsorted CD1b⁺ cells (Supplementary Fig. 7). NOD2L-induced CD1b⁺ DCs were more potent antigen-presenting cells than TLR2/1L-induced CD1b⁺ DCs in terms of presentation of tetanus toxoid to CD8⁺ T cells, as assessed by both ³H-thymidine incorporation and interferon-γ (IFN-γ) production (Fig. 2d). These differences were consistent at a tenfold higher concentration of each ligand (Supplementary Fig. 8). The potential of these immature DCs to present and process antigen via MHC class II was assessed using a CD4⁺ T cell clone (BCD4.9) that recognizes the M. leprae GroES protein and a defined peptide spanning amino acids 28–39 in an HLA-DR15–restricted manner²⁵. We isolated NOD2L- and TLR2/1L-induced CD1b⁺ DCs from an HLA-DR15–matched healthy donor by immunomagnetic selection. NOD2L-induced CD1b⁺ DCs were more potent antigen-presenting cells than TLR2/1L-induced CD1b⁺ DCs, in terms of processing and presentation of M. leprae GroES protein and the specific M. leprae GroES peptide to the T cell clone (Fig. 2e,f).

The mechanism of NOD2L-induced DC differentiation

Given that GM-CSF is known to be a potent inducer of DC differentiation, we investigated the differential ability of NOD2L versus TLR2/1L to induce expression of GM-CSF and its receptor. TLR2/1 activation of monocytes induced GM-CSF mRNA and protein (Fig. 3a) and the GM-CSF receptor α chain GM-CSFRA mRNA (Fig. 3b) more strongly than did NOD2L activation. The GM-CSF receptor β chain GM-CSFRB mRNA was equally upregulated by both ligands (Fig. 3b). TLR2/1L-induced DC differentiation was completely and significantly (P = 0.01) (Supplementary Fig. 9) blocked by the addition of neutralizing GM-CSF–specific mAbs, as previously reported²⁴. In contrast, NOD2L-induced DC differentiation was only partially blocked by neutralization of GM-CSF and was not significantly different compared to isotype control–treated cells (Supplementary Fig. 9). Therefore, the mechanism by which NOD2L, as compared to TLR2/1L, more potently induces monocytes to differentiate into DCs does not seem to be related to GM-CSF release or receptor expression.

To determine the mechanism by which NOD2L induces DC differentiation, we performed IPA to identify transcripts encoding secreted proteins differentially induced in NOD2L- versus TLR2/1L-activated monocytes. We found nine candidate genes to be induced more by NOD2L than by TLR2/1L and classified to encode secreted proteins in activated monocytes (Fig. 3c). Further analysis of these candidate genes indicated that one gene, IL32, encodes a secreted protein reported to have direct action on monocytes leading to the induction of proinflammatory cytokines including tumor necrosis factor-α, IL-1β, IL-6, IL-8 and chemokines²⁶. Furthermore, two of the candidate genes, IL18 and IL32, encode secreted proteins that are part of a common pathway reported to be involved in the host response to mycobacteria^27,28. IL-32 has been shown to contribute to the pathogenesis of several infectious^29,30, autoimmune³¹and immunoregulatory disorders, including inflammatory bowel disease³² and cancer³³. The human IL-32 receptor has not been identified, and no rodent orthologs of IL-32 have been reported.

IL-18 mRNA was induced greater than twofold after NOD2L activation of monocytes but not induced by TLR2/1L, whereas IL-18R mRNA was induced equally by both NOD2L and TLR2/1L (Fig. 3d). Additionally, IL-32 was more strongly induced in monocytes treated with NOD2L as compared to TLR2/1L, both at the mRNA and the protein levels (Fig. 3e). IL-32 mRNA was significantly induced in monocytes by NOD2L and TLR4L (lipopolysaccharide) but not by other ligands (Fig. 3f). Finally, infection of monocytes with live M. leprae at increasing multiplicities of infection induced IL-32 mRNA (Fig. 3g).

Next, we investigated the requirement for IL-32 in NOD2L-induced DC differentiation using siRNA knockdown. Transfection of siRNA targeting IL-32 mRNA (siIL32) into primary human monocytes knocked down IL-32 mRNA by > 90% in cultures containing medium, NOD2L and TLR2/1L (Fig. 3h). The ability of NOD2L to induce CD1b expression was blocked by >80% by siIL32, whereas, in contrast, the ability of TLR2/1L to induce CD1b expression, although at a lower level of induction, was not affected by siIL32 (Fig. 3h). Knockdown of IL-32 mRNA did not directly affect NOD2L-induced IL-6, IL-8 and IL-10 mRNA levels (Supplementary Fig. 10) but blocked induction of IL-32 and of CD1b⁺ DCs (Supplementary Fig. 11). NOD2L induction of CD1b expression was partially blocked by IL-18–specific mAb (Supplementary Fig. 12). These data indicate that IL-32 production is required for NOD2L- but not TLR2/1L-induced CD1b⁺ DC differentiation.

We next investigated whether IL-18 and IL-32 are sufficient to induce monocytes to differentiate into DCs. The addition of recombinant IL-32 to primary monocytes induced CD1b expression in a dose-dependent manner (Fig. 4a); however, IL-18 did not induce CD1b expression (Supplementary Fig. 13). IL-32–differentiated cells had a dendritic morphology similar to that of NOD2L– and GM-CSF–differentiated cells (Supplementary Fig. 14) and expressed CD1b whether cultured with 10% FCS or serum-free medium (Supplementary Fig. 15).

As both IL-32 and GM-CSF induce DC differentiation, we compared the phenotype and antigen presentation function of these two types of DCs. GM-CSF–derived DCs without the addition of IL-4 are immature DCs that more closely resemble circulating and tissue DCs by cell surface phenotype³⁴. Dose titration experiments revealed that treatment of monocytes with GM-CSF induced a greater percentage of CD1b⁺ cells as compared to treatment with IL-32 (Supplementary Fig. 16). The effect of the cytokines at optimal and supraoptimal doses indicated that both IL-32– and GM-CSF–derived CD1b⁺ DCs expressed similar percentages of the other DC markers (Supplementary Fig. 17). However, the MFIs of HLA-ABC and CD86 were significantly higher on IL-32–derived versus GM-CSF–derived CD1b⁺ DCs (Fig. 4b and Supplementary Figs. 18 and 19).

Given the differential expression of MHC class I on IL-32– versus GM-CSF–derived DCs, we investigated the potential of the different DCs to present antigen. We found IL-32–derived DCs to be more potent than GM-CSF– derived DCs in presenting tetanus toxoid and influenza peptide M1 to CD8⁺ T cells in terms of proliferation and IFN-γ release (Fig. 4c,d and Supplementary Fig. 20). We also verified the potency of IL-32–derived DCs in stimulating MHC class I–restricted T cell responses by tetramer labeling, using HLA-A*0201–influenza-M1 (GILGFVFTL) tetramers (Supplementary Fig. 21). Although IL-32–induced DCs were more potent than GM-CSF–induced DCs in presentation of antigen via MHC class I to CD8⁺ T cells, both IL-32– and GM-CSF–induced DCs were equally effective in presenting the M. leprae GroES protein and peptide via MHC class II to CD4⁺ T cells (Fig. 4e,f). This was consistent with the demonstration that MHC class I expression was higher on IL-32– versus GM-CSF–derived DCs, whereas MHC class II expression was equivalent (Fig. 4b). Together, these data indicate that IL-32 is sufficient to induce the differentiation of DCs with potent MHC class I antigen-presenting function. Our studies demonstrating a role for IL-32 in DC differentiation are in contrast to a previous study that found that IL-32 induced CD1a⁺ cells that were referred to as macrophages²⁶, although the analysis of DC markers in the study was limited.

Role of NOD2 and IL-32 in leprosy

To identify which of the NOD2L- and/or TLR2/1L-induced DC–related genes were relevant at the site of infection in human disease, we used an integrative bioinformatics approach. We compared the gene expression in activated monocytes to the gene expression profiles in the skin lesions of the different clinical forms of leprosy. The numbers of lymphocytes and macrophages in leprosy lesions is similar in the different clinical forms, and this has allowed comparison of T cell cytokine patterns in lesions by PCR⁶ as well as study of macrophage function using microarrays^35,36. The relative induction of all the genes in NOD2L- versus TLR2/1L-activated monocytes was compared to their differential expression at the site of disease in T-lep versus L-lep leprosy (Fig. 5a). Notably, the group I CD1 antigen presentation molecules CD1A, CD1B and CD1C clustered together, with a high relative expression in NOD2L- versus TLR2/1L-activated monocytes and in T-lep versus L-lep lesions (Fig. 5a). A randomly chosen gene would be unlikely to have the differential expression observed for CD1A, CD1B or CD1C in the two gene expression data sets (P = 0.004).

CD1b⁺ DCs are potent antigen-presenting cells that induce an adaptive immune response in leprosy^24,37. Furthermore, the frequency of CD1b⁺ DCs at the site of disease correlates with clinical form of the disease, as it is higher in T-lep than in L-lep lesions⁸. Examination of the microarray data indicated that the expression of both CD1b and IL-32 mRNA was significantly greater in T-lep versus L-lep lesions (Fig. 5b). In contrast, the gene expression of the DC marker CD205 and the myeloid marker CD11b was similar in the two disease types (Fig. 5b). Consistent with the microarray data, we detected IL-32 protein by immunohistochemistry in granulomas in leprosy lesions, with a higher frequency of positive cells in T-lep as compared to L-lep lesions (Fig. 5c). NOD2-expressing cells were numerous in granulomas of T-lep lesions, whereas only a few NOD2⁺ cells are found in L-lep lesions (Fig. 5d). Isotype controls were consistently negative (Supplementary Fig. 22). Quantification using ImageJ software revealed IL-32⁺ cells were eightfold more frequent in T-lep versus L-lep lesions, and NOD2⁺ cells were greater than threefold more frequent in T-lep versus L-lep lesions (Fig. 5e).

We next used confocal laser microscopy to determine the relative localization of IL-32 in relation to macrophages and DCs. We detected IL-32 in proximity to CD68⁺ macrophages and CD1b⁺ DCs (Fig. 5f). In addition, NOD2⁺IL-32⁺ cells could be detected in T-lep lesions (Fig. 5f). However, this double immunolabeling shows only the relative locations of the various markers and cannot distinguish production from uptake. In summary, the data indicate that IL-32⁺ cells are more frequent in T-lep versus L-lep lesions.

Given the low expression of both IL-32 and CD1b in L-lep lesions, we assessed the functional capacity of NOD2 to trigger DC differentiation in monocytes from patients with L-lep. NOD2L activation of monocytes induced a higher frequency of CD1b⁺ DC expression in cells from normal donors versus patients with L-lep (Fig. 6a). In contrast, GM-CSF treatment induced similar numbers of CD1b⁺ DCs in monocytes from normal donors and patients with L-lep (Supplementary Fig. 23). The frequency of CD1b⁺ DCs induced by treatment of monocytes with NOD2L was equivalent in monocytes from patients with T-lep and normal donors (Supplementary Fig. 24).

One possible explanation for the altered responsiveness of monocytes from donors with L-lep to NOD2L could be the release of immunosuppressive factors by monocytes from these individuals. IL-10 is known to be differentially expressed in L-lep versus T-lep lesions⁶ and to inhibit immune responses to mycobacteria^10,38. Monocytes from patients with L-lep but not healthy controls spontaneously secreted IL-10 into the culture medium (Fig. 6b). The addition of recombinant IL-10 to monocytes from healthy donors inhibited NOD2L-induced CD1b expression (Fig. 6b). Conversely, the addition of IL-10–specific neutralizing antibodies to monocytes from donors with L-lep restored NOD2L-induced CD1b expression (Fig. 6b).

Additionally, we considered the possibility that NOD2L-mediated induction of IL-32 was defective in donors with L-lep. NOD2L induced IL-32 and IL-18 mRNA expression in normal monocytes, but the response was significantly diminished in monocytes from patients with L-lep (Fig. 6c), whereas GM-CSF mRNA was induced to similar levels in monocytes from both normal donors and patients with L-lep (Fig. 6c). Furthermore, NOD2L-mediated induction of GM-CSF and IL-1β proteins was statistically equivalent in normal donors and subjects with L-lep, although NOD2L-mediated induction of tumor necrosis factor-α was greater in monocytes from normal donors than patients with L-lep (Supplementary Fig. 25). Although NOD2L did not induce appreciable CD1b expression in monocytes from patients with L-lep (Fig. 6d), the addition of recombinant IL-32 alone to monocytes from subjects with L-lep induced CD1b expression to levels similar to those in monocytes from healthy controls. NOD2L and recombinant IL-32 acted synergistically in the induction of CD1b expression in patients with L-lep (Fig. 6d). In conclusion, we provide evidence that NOD2L induces an IL-32–dependent DC program in monocytes with relevance to innate and acquired immune responses in human infectious disease.