Nitric oxide controls tuberculosis immunopathology by inhibiting NLRP3 inflammasome-dependent IL-1β processing

Adaptive immunity influences IL-1 production

To investigate the potential role of adaptive immunity in regulating innate inflammatory pathways, we examined the mechanisms driving TB pathogenesis in mice lacking mature T and B lymphocytes (Rag2^−/−), IFN-γ, (IFN-g^−/−) or iNOS (iNos^−/−). Following Mtb infection, each of these mutant mouse strains rapidly developed a qualitatively different disease than immunocompetent animals (C57BL/6). The disease was characterized by extensive tissue necrosis, neutrophil infiltration, and relatively high bacterial burdens (Fig. 1a-c). This severe pathology was associated with a distinct cytokine signature (Fig1. d-e dominated by IL-1α, IL-1β, and several monocyte/granulocyte chemo-attractants known to be induced by IL-1, e.g. CCL2 (MCP-1), CCL3 (MIP1α, CCL4 (MIP1β). While lung homogenates of iNOS-deficient animals were found to contain relatively high levels of the immature 31kDa precursor of IL-1β, the accumulation of the mature active form of the cytokine (p17) was even more marked. The ratio of mature/immature cytokine was increased sevenfold in iNOS-deficient animals (Fig. 1f), suggesting that IL-1β processing was enhanced. Consistent with the apparent increase in IL-1β maturation, the active form of caspase-1 (assessed by the p10 fragment) was only detectable in the lungs of iNos^−/− mice.

While Rag2^−/−, IFN-g^−/−, and iNos^−/− animals consistently produced relatively high amounts of mature IL-1β and related chemokines, the levels of many other cytokines characteristic of both innate and adaptive immune responses were either reduced or unchanged (Fig. 1d). This relatively specific alteration in cytokine profile suggested that IFN-γ-dependent NO might serve a specific role in modulating IL-1β responses. However, the immunodeficient mice lacking this important antimicrobial pathway harbored 30-200 fold more bacteria in their lungs at the time of analysis (Fig. 1c), and we could not exclude the possibility that this increased bacterial burden caused the observed differences in cytokine levels and pathology.

NO directly modulates IL-1 responses and immunopathology

To more directly determine if NO regulated IL-1β and inflammation, we developed an infection model that employs a 16S rRNA mutant of Mtb (strain 18b), which only replicates in the presence of streptomycin ^{21, 22}. The ability to exogenously control bacterial replication allowed the maintenance of bacterial numbers at the same level in both wild type and iNos^−/− animals throughout the infection. To colonize the lungs with a physiologically relevant burden of these relatively slow-growing bacteria, mice were inoculated either intratracheally or intravenously. After infection, streptomycin was administered for fourteen days before replication was arrested by drug withdrawal (Fig. 2a). Twenty-two days later, all mice harbored comparable numbers of Mtb in their lungs and spleens (Fig.2b). However, the disease observed in these two mouse strains was profoundly different, and the iNos^−/− animals infected by the intratracheal route became moribund by day 36. Increased local inflammatory pathology was apparent in iNos^−/− animals regardless of the route of infection, as the tissue architecture of both the lungs and spleens was disrupted by leukocyte infiltration. These mice also suffered from significant wasting, consistent with an increased systemic inflammatory response (Fig. 2c, d).

The cellular composition and cytokine profile of lesions in iNos^−/− animals were distinct from their wild type counterparts and consistent with an IL-1-mediated inflammatory disease. Using flow cytometry, we found that the airways and dissociated lung tissue of the iNos^−/− mice contained increased numbers of neutrophils (CD11b⁺ Gr1^hi), while T cell infiltration (CD3⁺) increased to a lesser degree. Increased neutrophil recruitment was confirmed by quantifying myeloperoxidase (MPO) activity in lung homogenates (Fig. 2e,f). This granulocytic infiltration was associated with an IL-1β-dominated cytokine signature similar to that observed in animals infected with wild type bacteria, and the active proteolytic fragment of IL-1β was only found in lung lysates of iNos^−/− mice (Fig. 2g and h). In contrast to IL-1β and the related chemokines, the production of a distinct proinflammatory mediator, IL-6, was correlated with bacterial burden and not NO production per se (Fig. 1d and 2g). Together, these data suggested that NO modulates immunopathology independent of its role in restricting Mtb growth, and that IFN-γ-stimulated NO plays a specific role in regulating the expression and maturation of IL-1β.

Exacerbated pathology depends on IL-1 and ASC

The coincident increase in both IL-1β and neutrophil infiltration in iNOS-deficient mice suggested that this cytokine might play a causal role in the observed immunopathology. To test this hypothesis, we infected wild type and IL-1 receptor-deficient mice (Il1r1^−/−) with Mtb and quantified the influx of neutrophils into the lung following treatment with the iNOS inhibitor, aminoguanidine hemisulfate (AGHS). To ensure that differential bacterial growth did not influence inflammation, the streptomycin-dependent infection model was used. Similar to our observations in iNos^−/− mice, AGHS treatment increased the recruitment of neutrophils into the lung, as assessed by both flow cytometry and MPO assay. This increase was abrogated in Il1r1^−/− mice, demonstrating that the inflammatory response seen in iNos^−/− mice depended upon IL-1 signaling (Fig. 3a), and not upon related cytokines such as IL-18.

IL-1β maturation appeared to be altered by NO (Fig. 1f), but the protease(s) responsible for this event in vivo remained unclear. Caspase-1-containing inflammasome complexes are absolutely required for the maturation of IL-1β in Mtb-infected macrophages in vitro^{23, 24}. However, the cell types producing IL-1β in vivo might be distinct²⁵ and inflammasome-independent mechanisms appear to dominate in immunocompetent mice chronically infected with Mtb²⁶. To determine if the inflammasome was required for the increased IL-1β activity we observed upon iNOS inhibition, we assessed the role of this complex in Mtb-induced inflammation. Consistent with previous reports, we could discern no significant effect of ASC deletion in mice infected with streptomycin-dependent Mtb. However, the increased neutrophil recruitment observed upon AGHS treatment was reversed in Asc^−/− animals (Fig. 3b). This phenotype was very similar to that seen in Il1r1^−/− mice. Thus, the immunopathology observed following iNOS inhibition was caused by inflammasome-dependent IL-1.

IFN-γ suppresses IL-1β in Mtb-infected macrophages

We investigated the mechanisms by which IFN-γ and NO regulated IL-1β in Mtb-infected macrophages. Consistent with previous reports^{23, 24}, IL-1β production in response to Mtb infection depended on the NLRP3-ASC-caspase-1 complex and is likely secondary to membrane damage mediated by the bacterial ESX1 secretion system (Supplementary Fig. 1a,b). As predicted by our in vivo observations, the pretreatment of Mtb-infected macrophages with IFN-γ specifically inhibited IL-1β release, but had no effect on the production of an unrelated cytokine, TNF. This suppression required signaling through the IFN-γ receptor 1, was independent of IFN-γ produced by the macrophages (Fig. 4a,b), and required IFN-γ stimulation prior to infection (Supplementary Fig. 1c). IFN-γ pretreatment did not affect bacterial uptake or reduce the viability of the macrophages (Supplementary Fig. 2a,b), suggesting a direct effect on IL-1β production.

By a number of criteria, IFN-γ appeared to suppress the inflammasome-dependent maturation of IL-1 and not the expression of the cytokine’s proform. By immunoblot, IFN-γ did not affect the abundance of pro-IL-1β in cell lysates, but specifically inhibited the processing and release of the mature cytokine into the supernatant (Fig. 4c). Furthermore, suppression by IFN-γ was equally apparent whether pro-IL1β expression was induced by Mtb alone or by prestimulation with either the TLR2 ligand, Pam3CSK4, the TLR4 ligand, lipopolysaccharide (LPS), or the TLR3 ligand, poly (I:C) (Fig. 4a, d). Finally, release of, IL-1α, which is regulated by inflammasome-dependent caspase-1 signaling ²⁷, was similarly suppressed by IFN-γ (Fig. 4d), further implying that IFN-γ modulates inflammasome activity.

IFN-γ specifically inhibits NLRP3 inflammasome activity

By treating macrophages with defined inflammasome activators, we found that the suppressive effect of IFN-γ was specific to the NLRP3 inflammasome. NLRP3-dependent IL-1β processing and release in response to the pore-forming toxin nigericin²⁸ was inhibited by IFN-γ pretreatment (Fig. 5a,b). In contrast, IFN-γ had no effect on IL-1β production in response to B-form dsDNA, poly (dA:dT), which depends on the distinct AIM2-containing inflammasome²⁹. As we observed in Mtb-infected macrophages, IFN-γ specifically suppressed IL-1β processing and did not affect TNF secretion induced upon LPS priming.

Bacteria with distinct cellular structures or pathogenic strategies trigger different inflammasome complexes. Those that disrupt cellular membranes generally act on the NLRP3 complex, and those that express flagellin (or similar proteins) have the capacity to trigger the NLRC4 inflammasome³⁰. To determine if IFN-γ suppression was NLRP3-specific in the context of these more complex stimulants, we infected macrophages with pathogens that engage different inflammasomes. Consistent with previous studies³¹, NLRP3 was the predominant NLR required for M. marinum-induced IL-1β production, and NLRC4 dominated the response to Salmonella typhimurium³² (Fig. 5c). Following IFN-γ treatment, only the NLRP3-dependent IL-1β production produced by M. marinum infection was efficiently suppressed (Fig. 5d). Thus, even in the context of genuine bacterial infection, IFN-γ specifically suppressed NLRP3-dependent responses.

IFN-γ stimulated NO suppresses IL-1β maturation

Our in vivo observations suggested that IFN-γ suppression of IL-1β processing might rely on NO production. Indeed, genetic deletion of iNos restored IL-1β processing and release in IFN-γ-treated macrophages infected with Mtb (Fig. 6a,b). Chemical inhibition of iNOS enzymatic activity with aminoguanidine³³ also abrogated the ability of IFN-γ to suppress IL-1β secretion (Fig. 6c), implicating NO as the mediator of this suppression. Similarly, the direct addition of NO donors, such as S-nitroso-N-acetyl-DL-penicillamine (SNAP) and S-nitrosoglutathione (GSNO), inhibited the processing and release of IL-1β as efficiently as IFN-γ (Fig. 6d,e). These donors produced a modestly lower total amount of nitrite as IFN-γ-treated macrophages over the time course of the experiment (Supplementary Fig. 3a, b) and did not alter the secretion of TNF, indicating that nontoxic doses were applied. Consistent with the freely diffusible nature of NO, wild type macrophages had the ability to suppress IL-1β production from co-cultured iNOS deficient cells in trans (Fig. 6f). We conclude that NO is necessary for IFN-γ to suppress IL-1β and sufficient to mediate this effect.

NO inhibits the assembly of the NLRP3 inflammasome

NO has the ability to regulate gene expression by inducing cGMP synthesis³⁴, suggesting that this compound might act by altering the abundance of inflammasome components. However, IFN-γ did not appreciably alter the expression levels of the relevant inflammasome proteins, NLRP3, ASC, caspase-1, and pro-IL-1β, in cell lysates (data not shown). Furthermore, the cell membrane permeable cGMP analog 8-bromo-cGMP (8-Br-cGMP) had no effect on IL-1β release when added to macrophage cultures before and during Mtb infection (Supplementary Fig. 4).

These observations suggested that that NO might act posttranslationally to regulate inflammasome function. To investigate this possibility, we reconstituted the relevant protein complexes in 293T cells. Co-expression of NLRP3 (or AIM2), ASC, caspase-1, and pro-IL-1β in this heterologous system resulted in spontaneous processing of both caspase-1 and IL-1β. The addition of an NO donor (SNAP) inhibited the NLRP3-dependent maturation of IL-1β in these transfected cells, but had no effect on the AIM2 inflammasome, implying that this compound might directly modulate the assembly or activity of the NLRP3 inflammasome (Fig. 7a).

To determine if NO altered inflammasome assembly, we evaluated the oligomerization of ASC that occurs during the formation of this complex³⁵. Intact inflammasomes were isolated from stimulated macrophages by differential centrifugation, cross-linked, and detected by immunoblotting. The addition of either IFN-γ or SNAP to these cells inhibited the formation of the ASC and caspase-1 oligomers that are induced by the NLRP3 stimulant nigericin. This inhibition of assembly correlated with a lack of caspase-1 maturation into the processed 35kDa form (Fig. 7b). In contrast, IFN-γ and NO had little effect on the AIM2-dependent oligomerization of ASC and caspase-1 in response to poly (dA:dT), and the subsequent maturation of caspase-1. Thus, in both macrophages and transfected 293T cells NO specifically inhibited the assembly and activity of the NLRP3 inflammasome complex.

NO regulates inflammasome activity via S-nitrosylation

In addition to its gene regulatory roles, NO also has the ability to directly alter protein function through the chemical modification of cysteine residues³⁶. To determine if the observed NO-dependent regulation of inflammasome activity was due to S-nitrosylation, we chemically reversed these modifications with ascorbate. This mild reductant abrogates the nitrosylation of thiols without affecting the ability to form disulfide bonds³⁷. Ascorbate treatment had no effect on the production of NO by macrophages, but reversed the IFN-γ-mediated inhibition of IL-1β release in a dose-dependent manner (Fig. 7c). The processing of both IL-1β and caspase-1 were fully restored by this treatment indicating that the IFN-γ and NO-mediated suppression of inflammasome activity relied on S-nitrosylation.

The posttranslational effect of NO on inflammasome assembly and activity implied that this protein complex might be directly modified by S-nitrosylation. Using the ‘biotin-switch’ method³⁸ we labeled S-nitrosocysteines in macrophage lysates with biotin and isolated the associated proteins with streptavidin. Upon IFN-γ stimulation of Mtb-infected macrophages, we detected iNOS-dependent nitrosylation of a variety of cellular proteins (Fig. 7d). This nitrosylated protein fraction contained NLRP3 and caspase-1, but not AIM2 or ASC.

The S-nitrosylation of NLRP3 and caspase-1 suggested that the modification of one or both of these proteins might be responsible for the inhibition of IL-1β processing. Indeed, nitrosylation of the caspase-1 active site thiol has been previously proposed to regulate the proteolytic activity of this enzyme³⁹. To identify the relevant cellular target(s) of NO, we developed a reconstituted IL-1β processing system in which NLRP3 or caspase-1 could be individually expressed and chemically modified in 293T cells. When these cell lysates (containing either NLRP3 or caspase-1) were incubated with lysates from cells transfected with the remaining inflammasome components, IL-1β processing was only observed if NLRP3, ASC, caspase-1 and pro-IL-1β were simultaneously present. This ability to reconstitute a functional inflammasome from separately produced components afforded the opportunity to assess the specific effects of nitrosylation on either NLRP3 or caspase-1. SNAP treatment of cells expressing the NLRP3 protein efficiently inhibited the IL-1β processing activity of mixed lysates, whereas treatment of caspase-1 expressing cells had no effect on IL-1β processing at any non-cytotoxic concentration (Fig. 7e). To further investigate the effect of NO on caspase-1 activity, we determined the effect of SNAP treatment on the low level of caspase-1 autoprocessing that is observed upon overexpression of this protein alone in 293T cells. At concentrations that inhibit NLRP3-dependent caspase-1 maturation, this compound had no effect on caspase-1 autoprocessing (Supplementary Fig. 5). Together, these observations indicated that nitrosylation of NLRP3, but not caspase-1, was sufficient to suppress inflammasome activity. We propose that nitric oxide-mediated inflammasome inhibition represents an important mechanism by which the adaptive immune response prevents persistent inflammatory tissue damage during chronic infection. (Supplementary Fig. 6).

Mice

C57BL/6 and Rag2^−/− (Model # RAGN12) mice were obtained from Taconic farms. IFN-g^−/− (B6.129S7-IFN-g^tm1Ts/J, stock#002827) and iNos^−/− (B6.129P2-Nos2^tm1Lau/J, stock#002609) mice were obtained from the Jackson Laboratory. IFN-gr1^−/− mice were a gift from Dr. Susan Swain (University of Massachusetts Medical School, UMMS). All other mutant mice were provided by Dr Kate Fitzgerald. Housing and experimentation was performed in accordance with the guidelines set forth by the department of animal medicine of UMMS and Institutional Animal Care and Use Committee.

Bacteria

Wild type M.tuberculosis (Mtb) was a phthiocerol dimycocerosate-positive strain of H37Rv, and was cultured in 7H9 medium. Strain 18b^{21, 22}, was provided by Dr Antonio Campos-Neto, and was cultured in the same media supplemented with 50 μg/mL of streptomycin sulfate (Sigma, St. Louis, MO). Salmonella enterica serovar typhimurium was provided by Dr. Beth McCormick, UMMS.

Mouse infection

For strain 18b, mycobacteria were suspended in phosphate-buffered saline (PBS)-Tween 80 (0.05%), dissociated by sonication, and delivered intravenously at ~1×10⁶ or intratracheally at ~3×10⁵ CFU per mouse. Mice were administered a daily dose of 2 mg of streptomycin sulfate for two weeks. For wild type H37Rv, iNos^−/− mice and their controls were challenged with ~200 CFU using an aerosol generation device (Glas-Col, Terre Haute, IN). Rag2^−/−, IFN-g^−/− mice and their controls were infected with ~10⁶ CFU of H37Rv via the intravenous route. All the mice were of C57BL/6 background. For aminoguanidine treatment, groups of mice were supplied with water supplemented with 2.5% aminoguanidine and 2.5% glucose 7 days prior to Mtb infection, and this was maintained throughout.

Myeloperoxidase (MPO) assay

After perfusion with PBS, lung tissue was homogenized in 1 ml of MPO buffer (50 mM K₂HPO₄, pH 5.4, 0.5% hexadecyl trimethyl ammonium bromide (HTAB), 10 mM EDTA). The homogenate was frozen and thawed twice. After centrifugation at 2500g for 15 minutes at 4°C, the supernatant was collected and serially diluted. 25μl of diluted supernatant was added to 25 μl of assay buffer (16.7 mg/ml of o-dianisidine dihydrochloride in 50 mM K₂HPO₄, pH 5.4) and 200 μl of development solution (30 μl of 31.1% H₂O₂ per 10 ml of 50 mM K₂HPO₄, pH 5.4). The change in OD₄₆₀ was measured between 0 and 25 minutes, and MPO activity was calculated as ΔOD/min per milligram of total protein in the lysates.

Flow Cytometry

Single cell suspensions were prepared from bronchoalveolar lavage and PBS-perfused lung tissue as described previously⁴⁹. Cells were stained with anti-CD3-PE (clone 145-2C11), anti-CD11b-Pacific Blue (clone M1/70) and anti-Ly6G-PE (RB6-8C5) antibodies, and analyzed with an LSR II flow cytometer.

Macrophage infections

Bone marrow from mouse femurs was differentiated into macrophages for 7 days in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 20% L929 cell conditioned medium, 10% fetal bovine serum, 2 mM L-glutamine and 1 mM sodium pyruvate. BMDMs were infected with Mtb at an MOI of 10 for 4 hours, washed with PBS, and cultured in serum free media for 18 hrs. For Mycobacterium marinum, BMDMs were infected for 4 hours (MOI = 10), washed twice, and incubated for overnight in serum-free medium containing gentamicin (50 μg/ml). For Salmonella enterica serovar typhimurium, BMDMs were infected for 1h, washed, incubated for one hour in medium-containing gentamicin (50 μg/ml), and incubated for an additional 4 hours with culture medium containing gentamicin (10 μg/ml). Cytokines were measured in the supernatants generated during the final incubation.

Cytokine and nitric oxide measurements

Cytokine concentrations were quantified using commercial ELISA kits (BD Opt EIA) or the Multi-analyte ELISArray kit from SA Biosciences (Cat No. MEM-005A). Nitric oxide was measured using the Griess reagent (Sigma, Cat No. G4410). Immunoblotting was conducted using anti-murine caspase-1 p10 (sc-514, Santa Cruz Biotechnology), anti-murine IL-1β (R&D systems, Minneapolis, MN), rabbit polyclonal anti-murine AIM2 (generously provided by Dr E. Alnemri), or anti-murine NLRP3 (Enzo-Life sciences). All samples were normalized for total protein content.

Detection of S-nitrosylated proteins

The biotin switch assay for detecting S-nitrosylated proteins was performed essentially as described³⁸. Biotinylated proteins were purified by streptavidin-agarose (Pierce). Bound proteins were recovered in elution buffer (20 mM HEPES, pH 7.7, 100 mM NaCl, 1 mM EDTA, and 100 mM of 2-mercaptoethanol) and visualized by immunoblotting.

ASC oligomerization assay

ASC oligomerization assay was performed as described with minor modifications³⁵. Briefly, immortalized BMDMs from C57BL/6 mice⁵⁰ were primed with LPS (200 ng/ml) and IFN-γ (200U/mL) for 3 hrs or treated with NO donor SNAP (500μM) for 30 min before the addition of inflammasome ligands. The cells were stimulated with poly(dA:dT) (1.5 μg/10⁶ cells; for 3 hrs) and nigericin (10 μM; for 30 min). The cytosolic fraction of these cells was centrifuged at 5000 rpm to pellet the inflammasome complexes, which were treated with 2mM of disuccinimidyl suberate (DSS). The cross-linked proteins were separated on 10% SDS-polyacrylamide gels and analyzed for ASC oligomerization by immunoblotting with a rat monoclonal antibody against ASC (kind gift from Dr. Gabriel Nunez, University of Michigan).

Inflammasome reconstitution in 293T cells

The inflammasome reconstitution assay shown in figure 7a was performed as described²⁹ with minor modifications. 2 × 10⁴ 293T cells were transfected with human ASC (20 ng), caspase1 (5 ng), guassia-luciferase FLAG-tagged IL-1β (80 ng), AIM2 (20 ng), and NLRP3 (20 ng) expression plasmids. After 2-4 h, the cells were treated with different concentrations of SNAP (125, 250 and 500 μM). Sixteen hours later, cells were lysed and analyzed by immunoblotting with anti-IL-1β (3ZD, National Cancer Institute, NIH) or anti-FLAG M2 antibodies (Sigma).

in vitro inflammasome reconstitution assay

For the experiments described in Figure 7e, independent cultures of 4× 10⁵ 293T cells were transfected with the indicated combinations of human ASC (150 ng), caspase1 (50ng), guassia-luciferase FLAG-tagged IL-1β (1γg) and NLRP3 (150ng) expression plasmids. After 4 hrs, the cells were treated with SNAP (125 and 250γM). Sixteen hours later, cells were lysed with cold lysis buffer (25mM Tris.Cl, pH, 7.4, 150mM NaCl, 1mM EDTA, 1% NP-40, 5% glycerol, 1mM nucuproine, 1mM PMSF) and snap frozen in liquid N₂. The lysates were mixed in different combinations, incubated at 30°C for 20 minutes to facilitate inflammasome activation, and analyzed by immunoblot with goat anti-mouse IL-1β (R&D systems), anti-mouse caspase-1 antibody (clone 5B10; eBioscience) or anti mouse-NLRP3 (Enzo life sciences).