Cutting Edge: TNF-α Mediates Sensitization to ATP and Silica via the NLRP3 Inflammasome in the Absence of Microbial Stimulation¹

The innate immune system provides the first line of defense against invading pathogens but is also involved in the development of inflammatory responses that occur in a sterile environment. Dendritic cells (DCs)³ and macrophages contribute to the innate immune response by sensing microbial structures or danger signals that induce the production of proinflammatory molecules and the activation of adaptive immune responses (1). In infection, the induction of inflammatory responses relies on the activation of host pattern-recognition receptors including TLRs and Nod-like receptors (NLRs) that recognize conserved and unique microbial structures (2, 3). The mechanism by which inflammation is activated and maintained in a sterile environment is less understood (1, 4). Members of the NLR family of proteins have been recently linked to the recognition of microbial and danger signals and the activation of inflammatory responses (5, 6). In response to microbial stimuli, certain NLR family members, including Nlrc4 and Nlrp3, assemble a large multiprotein complex called the “inflammasome” that induces the processing of procaspase-1 into an enzymatically active enzyme composed of two p20 and p10 chains (6). Subsequently, active caspase-1 processes pro-IL-1β into the biologically active IL-1β molecule, which acts synergistically with other cytokines in the orchestration of inflammatory responses (6). Nlrp3 (NLR family, pyrin domain containing 3; also called cryopyrin and Nalp3) is critical for caspase-1 activation and secretion of IL-1β in macrophages stimulated with several microbial molecules, ATP, urate crystals, silica, or asbestos particles (6). The importance of Nlrp3 in inflammatory homeostasis is underscored by the observation that mutations of this NLR are associated with the development of familial autoinflammatory syndromes (7). The activation of caspase-1 via Nlrp3 requires a signal provided by a microbial ligand such as LPS (6). In addition, a second signal that includes ATP or particulate matter is required for caspase-1 activation induced via the Nlrp3 inflammation (6). Thus, it is unclear how the Nlrp3 inflammasome contributes to inflammation under sterile conditions. In the present study, we identified a mechanism involving TNF-α and IL-1 that promotes activation of the Nlrp3 inflammasome in the absence of microbial stimulation.

Casp-1^−/−, Nlrp3 ^−/−, and Asc^−/− in a C57BL6 background have been described (8). TLR4^−/− and MyD88^−/−/Trif^−/− mice were provided by Dr. S. Akira (Osaka University, Suita, Japan). TnfrI/II^−/− and C57BL6 mice were purchased from The Jackson Laboratory. All animal studies were approved by the University of Michigan Committee on Use and Care of Animals (Ann Arbor, MI).

ATP was from Sigma-Aldrich. Ultrapure LPS from Escherichia coli, Pam3CSK (bacterial lipoprotein), polyinosinic:polycytidylic acid, and CpG were from InvivoGen. Recombinant TNF-α was purchased from R&D Systems or Peprotech with identical results. Endotoxin contamination in TNF-α preparations was <0.1 ng/μg of protein. Recombinant soluble CD40 ligand (CD40L) and IFN-γ were from Peprotech. Agonistic anti-Fas Ab (clone Jo2) was from BD Pharmingen. IL-1α, IL-1β, and receptor activator of NF-κB ligand (RANKL) were from R&D Systems. Salmonella enterica serovar Typhimurium strain SL1344 was a gift of D. Monack, Stanford University (Stanford, CA). Infection of macrophages with S. typhimurium was performed as described (9).

Bone marrow-derived macrophages (BMDMs) were isolated as previously described (9). For the differentiation of DCs, bone marrow cells were cultured with GM-CSF (10 ng/ml), with fresh GM-CSF added on days 3 and 5. After 6–7 days, >90% of the floating cells were CD11c⁺ and were used as DCs. Human PBMCs were purified from the venous blood of healthy volunteers by density centrifugation over Ficoll-Paque (Pharmacia Biotech). Monocytes were then enriched by adherence to plastic dishes.

Preparation of cell extracts and immunoblotting has been described (10). Membranes were probed with Abs anti-caspase-1, a gift from Dr. P. Vandanabeele (Ghent University, Ghent, Belgium). Abs against IκBα, phospho-IκBα, ERK, and phospho-ERK were from Cell Signaling. Amounts of IL-1β, TNF-α, and IL-6 were measured by ELISA (R & D Systems).

Statistical significance between groups was determined by two-tailed Student’s t test. Differences were considered significant when p < 0.01.

TNF-α induces IL-1β secretion (11). However, the mechanism involved remains largely unknown. To understand the link between TNF-α stimulation and IL-1β secretion, we initially assessed whether TNF-α can induce caspase-1 activation in mouse BMDMs and DCs. Stimulation of macrophages with TNF-α, CD40L, agonistic anti-Fas Ab, IFN-γ, or LPS alone did not induce caspase-1 activation as determined by immunoblotting with an Ab that recognizes the p20 subunit of mature caspase-1 (Fig. 1^,A). However, brief stimulation with ATP induced processing of procaspase-1 in macrophages pretreated with TNF-α or LPS, but no or minimal caspase-1 activation was observed with CD40L, an agonistic anti-Fas Ab, or IFN-γ (Fig. 1). Incubation of macrophages with CD40L and Fas Ab induced IκBα phosphorylation and degradation, indicating that they were stimulatory (supplemental Fig. 1).⁴ Likewise, stimulation with IL-1α or IL-1β, but not RANKL or PMA, induced robust caspase-1 activation in the presence of ATP (Fig. 1,B). Similar results were obtained with DCs (data not shown). Activation of caspase-1 by TNF-α was detected after incubation with ATP for 5 min and increased with longer ATP stimulation (supplemental Fig. 2). Furthermore, the levels of caspase-1 activation were augmented with increasing concentrations of TNF-α (supplemental Fig. 2). TNF-α also promoted caspase-1 activation in response to silica (Fig. 1,C), another stimulus that activates the Nlrp3 inflammasome. However, TNF-α did not enhance caspase-1 processing triggered by infection with S. enterica serovar Typhimurium (Fig. 1 C), an intracellular bacterium that induces caspase-1 activation through Nlrc4 (9). These results indicate that cytokines can induce caspase-1 activation in macrophages stimulated with ATP or silica in the absence of microbial stimulation.

Next, we determined the requirement of TNF-α for the induction of caspase-1 activation and IL-1β secretion in response to ATP. First, we compared the ability of TNF-α to induce IL-1β secretion in wild-type and mutant macrophages deficient in TNFR-I and TNFR-II, the surface receptors required for TNF-α signaling. Secretion of IL-1β induced by TNF-α and ATP was abolished in macrophages lacking TNFR-I and TNFR-II or Nlrp3 when compared with wild-type macrophages (Fig. 2, A and B). Consistently, caspase-1 activation induced by TNF-α and ATP was abrogated in macrophages lacking TNFR-I and TNFR-II, Nlrp3, or ASC (apoptotic speck protein containing a C-terminal caspase recruitment domain; Fig. 2 C and supplemental Fig. 3). Similarly, caspase-1 activation triggered by IL-1α and ATP was abolished in macrophages deficient in Nlrp3 or Asc (supplemental Fig. 3). The requirement of TNF receptors was specific in that caspase-1 activation and IL-1β secretion triggered by LPS and ATP was abolished in macrophages lacking Nlrp3 but not TNFR-I and TNFR-II (supplemental Fig. 4). Stimulation of the purinergic P2X7 receptor, an ATP-gated ion channel, is required for activation of the Nlrp3 inflammasome in response to LPS and ATP (12). We found that the P2X7 receptor was also required for caspase-1 activation triggered by ATP in TNF-α-stimulated macrophages (supplemental Fig. 5). Thus, activation of the inflammasome by TNF-α in response to ATP is mediated by Nlrp3 and requires the TNF and P2X7 receptors.

We next sought to determine the mechanism by which TNF-α sensitizes macrophages to the danger signal ATP. Given that a long exposure to TNF-α was required for the induction of caspase-1 activation (see below), we tested whether sensitization to caspase-1 activation by TNF-α was affected by treatment with cyclohexamide (CHX), a general inhibitor of protein translation. Incubation of macrophages with CHX blocked caspase-1 activation induced by TNF-α and ATP (Fig. 3,A). In addition, caspase-1 processing triggered by TNF-α and ATP was abrogated by treatment with BAY 11-7082 (Fig. 3,A), a drug that inhibits NF-κB activation by targeting the Iκ-B kinase complex (13). Furthermore, the activation of NF-κB and MAPKs induced by TNF-α was comparable in wild-type and Nlrp3-deficient macrophages (supplemental Fig. 6), indicating that TNF-α does not act by regulating NF-κB or MAPKs activation through Nlrp3. CHX did not affect caspase-1 activation broadly, in that it did not impair activation of caspase-1 induced by infection with the S. enterica serovar Typhimurium, which relies on the Nlrc4 inflammasome (supplemental Fig. 7). Initial experiments suggested that LPS triggers caspase-1 activation independently of TLR4 (14). However, recent studies showed that LPS can induce caspase-1 via a TLR4-Trif pathway (10). We reexamined this issue and found that, in agreement with the recent studies (10), LPS required TLR4 to promote caspase-1 activation in response to ATP (supplemental Fig. 8). Similar to TNF-α, the induction of caspase-1 by LPS and ATP was inhibited by CHX and BAY 11-7082 (supplemental Fig. 8). Consistently, activation of caspase-1 by LPS, synthetic lipopeptide (TLR2 ligand), polyinosinic:polycytidylic acid (TLR3 ligand), and CpG (TLR9 ligand) in response to ATP was abolished in macrophages deficient in MyD88 and Trif (Fig. 3 B). These results suggest that TNF-α and TLR ligands promote activation of the Nlrp3 inflammasome, at least in part, through gene transcription and NF-κB.

Previous studies have shown that prolonged stimulation with TNF-α renders macrophages refractory to a successive stimulation with TNF-α or LPS (15). Therefore, we tested whether prolonged exposure of macrophages to TNF-α reduces the sensitivity to caspase-1 activation in response to ATP. In line with published results (15), the activation of NF-κB and MAPK in response to TNF-α was impaired in macrophages prestimulated with TNF-α for 24 h (supplemental Fig. 9). Similarly, macrophages stimulated with TNF-α or LPS for 24 h become refractory to LPS, synthetic bacterial lipopeptide (TLR2 agonist), and CpG (TLR9 agonist) as determined by reduced secretion of TNF-α and IL-6 when compared with the response observed in naive macrophages (supplemental Fig. 10). We asked next whether macrophages tolerized to TNF-α or LPS could respond to a danger signal such as ATP. Remarkably, secretion of IL-1β and caspase-1 activation was induced in macrophages stimulated with TNF-α for 6 h and further enhanced with 24 h of stimulation, both of which required Nlrp3 (Fig. 4, A and B). In contrast, stimulation of macrophages with LPS for 6 h induced IL-1β secretion and caspase-1 activation, but the cells were refractory to ATP stimulation after LPS treatment for 24 h (Fig. 4, C and D). We next tested whether these findings could be extended to human monocytes. In agreement with published results (16, 17), ATP potentiated the production of IL-1β when human monocytes were pretreated with LPS for 4 h but not 24 h (supplemental Fig. 11). Consistent with the mouse results, the secretion of IL-1β in response to ATP was enhanced ∼ 40-fold in monocytes prestimulated for 24 h with TNF-α but not LPS (supplemental Fig. 11). These results indicate that both human monocytes and mouse macrophages rendered insensitive to TNF-α by continuous stimulation with TNF-α remain responsive to ATP stimulation by producing IL-1β.

We show that stimulation with TNF-α or IL-1 is sufficient to trigger caspase-1 activation and IL-1β secretion in response to ATP, providing a mechanism for activation of the Nlrp3 inflammasome in the absence of infection. TNF-α induces the synthesis of pro-IL-1β (18), but our results indicate that it alone cannot trigger the activation of caspase-1, a step that is required for IL-1β secretion, in mouse macrophages. TNF-α required the P2X7 receptor to induce caspase-1 in response to ATP, which may act, at least in part, by inducing a K⁺ efflux (6). Like LPS, the activation of caspase-1 induced by TNF-α was suppressed by CHX and by an inhibitor of NF-κB activation, suggesting that TNF-α and LPS stimulation induces a factor or factors that contribute to the activation of the Nrlp3 inflammasome. A potential candidate factor is Nlrp3 itself, because its expression is induced by TNF-α and TLR ligands in human monocytes (19). Alternatively, TNF-α may promote the formation of endogenous ligand or ligands that could act as agonists of Nlrp3. Regardless of the mechanism, our results indicate that endogenous cytokines such as TNF-α and IL-1 can functionally substitute for microbial stimuli to activate the Nlrp3 inflammasome in response to ATP and silica.

Gain-of-function mutations in NLRP3 are associated with the development of autoinflammatory syndromes (7) that are in turn associated with constitutive activation of NLRP3 and inappropriate production of IL-1β (20). Monocytes from patients with NLRP3 mutations produce elevated levels of TNF-α subsequent to IL-1β secretion (21). Thus, TNF-α and IL-1β could participate in a positive feedback loop to augment the constitutive caspase-1 activation observed in patients with autoinflammatory syndromes. Similarly, the fibrillar peptide amyloid-β, which is associated with the pathogenesis of Alzheimer’s disease, requires LPS stimulation to induce activation of the Nlrp3 inflammasome (22). Because amyloid-β fibrils induce the release TNF-α (23), it is possible that this cytokine participates in activation of the Nlrp3 inflammasome in the brains of patients with Alzheimer’s disease. In this study we show that macrophages tolerized to TNF-α, but not to TLR ligands, retain full sensitivity to the danger signal ATP. Our results suggest that ATP is capable of breaking the tolerant state that results from long exposure to TNF-α through the activation of the Nlrp3 inflammasome. Thus, danger signals may play a role in maintaining inflammation in diseases characterized by chronic production of TNF-α and/or IL-1.

We thank Richard Flavell, Shizuo Akira, and Millenium Pharmaceuticals for generous supply of mutant mice, Joel Whitfield for technical support, and Sherry Koonse for excellent animal husbandry.

The authors have no financial conflict of interest.