The AIM2 inflammasome is critical for innate immunity to Francisella tularensis

Abstract

Francisella tularensis, the causative agent of tularemia, infects host macrophages, which triggers production of the proinflammatory cytokines interleukin 1β (IL-1β) and IL-18. We elucidate here how host macrophages recognize F. tularensis and elicit this proinflammatory response. Using mice deficient in the DNA-sensing inflammasome component AIM2, we demonstrate here that AIM2 is required for sensing F. tularensis. AIM2-deficient mice were extremely susceptible to F. tularensis infection, with greater mortality and bacterial burden than that of wild-type mice. Caspase-1 activation, IL-1β secretion and cell death were absent in Aim2^−/− macrophages in response to F. tularensis infection or the presence of cytoplasmic DNA. Our study identifies AIM2 as a crucial sensor of F. tularensis infection and provides genetic proof of its critical role in host innate immunity to intracellular pathogens.

Main

The innate immune system is an evolutionary conserved first line of defense against invading organisms. It is now well established that the innate immune system uses molecular pattern recognition to detect infection and mount an immune response to eliminate the pathogen and infected cells. Specialized receptors collectively called pathogen- or pattern-recognition receptors (PRRs) are present on the cytoplasmic or endosomal membranes or in the cell cytosol^1,2. PRRs sense highly conserved molecular components of invading organisms known as pathogen-associated molecular patterns.

The PRR AIM2 (absent in melanoma 2) recognizes cytoplasmic double-stranded DNA through its HIN-200 domain^3,4,5,6. AIM2 assembles a large inflammasome complex with cytosolic DNA; the complex is visible by confocal microscopy³. The fully assembled AIM2 inflammasome then recruits and subsequently oligomerizes the caspase-1-activating adaptor protein ASC, which activates caspase-1, leading to production of the mature cytokines interleukin 18 (IL-18) and IL-1β and subsequent cell death^3,7,8.

Despite such advances in the understanding of how cytoplasmic DNA is recognized by the AIM2 inflammasome, little is known about the precise role of the AIM2 inflammasome in the innate immune defense against intracellular microbial and viral pathogens. Of particular interest is the facultative intracellular Gram-negative bacteria Francisella tularensis, the causative agent of tularemia. Invasion of the cytosol by F. tularensis leads to the production of type I interferon as well as activation of caspase-1 and cell death^9,10. F. tularensis activates the interferon response by a mechanism that is independent of signaling by plasma-membrane or endosomal Toll-like receptors, cytosolic RNA helicases RIG-I or Mda5 or biosensors Nod1 or Nod2 but requires signaling by the IRF3 transcription factor^10,11. It has been proposed that the host cell may sense cytosolic F. tularensis DNA via unidentified cytosolic DNA sensors, leading to the induction of type I interferons as well as activation of caspase-1 and cell-death pathways during F. tularensis infection^10,11.

We describe here the generation of AIM2-deficient mice to investigate whether AIM2 has an important role in the host proinflammatory innate immune response to F. tularensis specifically and to cytosolic DNA in general. Our results show that AIM2-deficient macrophages were defective in caspase-1 activation, IL-1β secretion and cell death in response to cytosolic DNA or infection with F. tularensis. Although the AIM2 deficiency did not affect the transcriptional type I interferon response to F. tularensis infection, AIM2-deficient mice were more susceptible to the lethal effect of F. tularensis than were wild-type mice. Our results thus suggest that the type I interferon response is not sufficient on its own for protection against F. tularensis infection and that the AIM2 inflammasome activity is required for full innate immunity to this pathogen. Our work also indicates that among all known PRRs, AIM2 is uniquely required for the proinflammatory innate immune response to F. tularensis.

Results

AIM2 is important for caspase-1 activation by DNA

To investigate the precise role of AIM2 in the host innate immune defense against dangerous cytosolic DNA produced by intracellular viral and microbial pathogens, we generated AIM2-deficient mice by gene-trap technology^12,13 (Supplementary Fig. 1a,b). Immunoblot analysis of spleen and bone marrow samples from an Aim2^−/− mouse and its Aim2^+/+ littermate showed the presence of full-length AIM2 protein in the Aim2^+/+ mouse but not in the Aim2^−/− littermate (Fig. 1a, top), which indicated that insertion of the gene-trap vector in Aim2 resulted in disruption of this gene. Heterozygous Aim2^+/− mice had lower expression of AIM2 in the spleen than did Aim2^+/+ mice (Fig. 1a, bottom), which indicated that the expression of AIM2 was dependent on gene dose. No Aim2^−/− mice had any obvious phenotypic abnormalities and all were morphologically indistinguishable from their Aim2^+/− or wild-type littermates (Supplementary Fig. 1c), which indicated that AIM2 deficiency did not have any apparent adverse effects on mouse development.

Figure 1: Disruption of mouse Aim2 abolishes activation of the inflammasome by cytoplasmic DNA and vaccinia virus.

(a) Immunoblot analysis of the expression of AIM2 in spleens and BMDMs from Aim2^+/+ and Aim2^−/− littermates (top) and in spleens from Aim2^−/−, Aim2^+/+ and Aim2^+/− mice (bottom), assessed with an antibody specific for mouse AIM2. β-actin serves as a loading control. (b) Immunoblot (IB) analysis of mouse procaspase-1, caspase-1 (p20 subunit) and/or AIM2 in culture supernatants (Sup) and lysates (Lys) of Aim2^+/+ and Aim2^−/− BMDMs left untreated (None) or transfected with the synthetic DNA poly(dA:dT) or plasmid DNA (pcDNA), or treated for 5 h with LPS (500 ng/ml) followed by nigericin (2.5 μM) for 45 min (LPS + nig), assessed with monoclonal antibody to mouse caspase-1 p20 (anti-p20). (c) Release of LDH into culture supernatants of the BMDMs described in b, presented relative to the total cellular LDH content. *P < 0.05 and **P < 0.01, Aim2^+/+ versus Aim2^−/− (Student's t-test). (d) Immunoblot analysis of mouse procaspase-1 and caspase-1 in culture supernatants and lysates of mouse Aim2^−/− and Aim2^+/+ BMDMs infected for 18 h with vaccinia virus (multiplicity of infection (MOI), above lanes). Data are representative of at least three experiments (mean and s.d. in c).

In vitro studies with bone marrow–derived macrophages (BMDMs) from Aim2^−/− and Aim2^+/+ littermates showed that liposome-delivered DNA activated caspase-1 and induced cell death in wild-type Aim2^+/+ BMDMs but not in Aim2^−/− BMDMs (Fig. 1b,c), which indicated that AIM2 is indeed critical for sensing cytoplasmic DNA and for the activation of caspase-1. The AIM2 deficiency had no effect on caspase-1 activation by the PRR NLRP3 stimulus of lipopolysaccharide (LPS) plus ATP or LPS plus the pore-forming toxin nigericin¹⁴ or the PRR NLRP1 stimulus of anthrax lethal factor¹⁵ (Fig. 1b and Supplementary Fig. 2a,b). Reconstitution of the Aim2^−/− macrophages with green fluorescent protein (GFP)-tagged AIM2 by retroviral transduction restored their responsiveness to transfected indocarbocyanine (Cy3)-labeled DNA, as shown by oligomerization of AIM2-GFP and the induction of pyroptotic cell death, which is dependent on caspase-1 activation (Supplementary Fig. 3a). There was no morphological evidence of pyroptotic death in the Aim2^−/− macrophages expressing GFP alone (Supplementary Fig. 3b).

Next we investigated the role of AIM2 in the innate immunity of macrophages to cytosolic DNA produced by infection with vaccinia virus. In contrast to wild-type Aim2^+/+ BMDMs, Aim2^−/− BMDMs had no caspase-1 processing at 18 h after infection with vaccinia virus (Fig. 1d), which provides definitive proof that the AIM2 inflammasome is indeed critical for detection of vaccinia virus infection, as demonstrated before by experiments using knockdown by small interfering RNA⁴.

AIM2 is important for caspase-1 activation by F. tularensis

F. tularensis is a highly infectious bacterium that replicates in the cytoplasm of infected cells^16,17, which leads to activation of an IRF3-dependent and Toll-like receptor–independent type I interferon response, as well as an NLRP3-independent but ASC-dependent inflammasome that causes cell death^9,10. Such observations suggest that cytosolic DNA produced by F. tularensis during its escape from the phagosome might be the common ligand that activates both the type I interferon response and possibly the AIM2 inflammasome pathways. To test that possibility, we analyzed inflammasome activation and cell death by measuring the release of lactate dehydrogenase (LDH) from macrophages from wild-type and Aim2^−/− mice after infection with F. tularensis subspecies novicida (called 'F. tularensis' here). Indeed, processing of caspase-1, secretion of IL-1β and release of LDH were completely absent in Aim2^−/− macrophages at 6 h after infection with F. tularensis (Fig. 2a,b), even at a high multiplicity of infection (Fig. 2a). Processing of caspase-1 and secretion of IL-1β were also absent in Aim2^−/− macrophages at 24 h after infection, although there was LDH release at this time point (Supplementary Fig. 4a,b), probably due to caspase-1-independent cell death, as reported before for caspase-1-deficient macrophages¹¹. In contrast, we found normal processing of caspase-1, secretion of IL-1β and release of LDH in wild-type macrophages (Fig. 2a,b and Supplementary Fig. 4a,b). Processing of caspase-1 in response to F. tularensis infection or transfected DNA was also unimpaired in heterozygous Aim2^+/− macrophages, although it was less than that in wild-type macrophages because of the smaller amount of AIM2 in the Aim2^+/− macrophages (Supplementary Fig. 5a,b). Consistent with a critical role for AIM2 in the pyroptotic cell-death pathway, morphological features of pyroptosis, including plasma membrane swelling and nuclear condensation, were very obvious in F. tularensis–infected Aim2^+/+ macrophages but not in Aim2^−/− macrophages (Fig. 2c). Infection with Salmonella typhimurium, which specifically activates the NLRC4 (Ipaf) inflammasome¹⁸, resulted in normal processing of caspase-1 in both wild-type and Aim2^−/− macrophages (Supplementary Fig. 4c), which indicated that AIM2 is not involved in the proinflammatory response to S. typhimurium infection. Consistent with published findings¹⁰, the activation of caspase-1 by F. tularensis infection was dependent on ASC but not on NLRP3 (Fig. 2d). Together our results indicate that AIM2 is critical for the proinflammatory and cell-death responses to F. tularensis infection but is dispensable for these processes in response to S. typhimurium infection.

Figure 2: AIM2 is required for F. tularensis–induced activation of the inflammasome.

(a) Immunoblot analysis of mouse procaspase-1, caspase-1, IL-1β, AIM2 and/or pro-IL-1β in culture supernatants and lysates of mouse Aim2^−/− and Aim2^+/+ macrophages left untreated or infected for 6 h with F. tularensis (FN; MOI in parentheses above lanes) or treated with LPS and nigericin as described in Figure 1b. (b) Release of LDH into culture supernatants of the macrophages in a. *P < 0.05, **P < 0.01 and ***P < 0.005, Aim2^+/+ versus Aim2^−/− (Student's t-test). (c) Confocal live-cell microsopy of Aim2^−/− and Aim2^+/+ BMDMs left uninfected (UI) or infected for 6 h with F. tularensis; nuclei were stained with Hoechst stain (blue). Images are merged differential interference contrast and Hoechst channels. Original magnification, ×40. (d) Immunoblot analysis of mouse procaspase-1, caspase-1, ASC, AIM2 and/or NLRP3 in culture supernatants and lysates of mouse wild-type, ASC-deficient (Pycard^−/−; called 'Asc^−/−' here) and Nlrp3^−/− macrophages infected with F. tularensis for 6 h or for 24 h (far right; MOI in parentheses above lanes). (e) Immunoblot analysis of mouse procaspase-1, caspase-1, IL-1β, pyrin and/or pro-IL-1β in culture supernatants and lysates of mouse pyrin-deficient (Mefv^−/−) and pyrin-sufficient (Mefv^+/+) macrophages infected for 6 h with F. tularensis (MOI in parentheses above lanes) or treated with LPS and nigericin as described in a. (f) Release of LDH into culture supernatants of the macrophages in e. Data are representative of at least three experiments (mean and s.d. in b,f).

The familial Mediterranean fever protein pyrin, which can also form an inflammasome complex with caspase-1 and ASC^19,20, might be involved in the activation of caspase-1 by F. tularensis in human macrophages²¹. To test this possibility in mouse macrophages, we infected peritoneal macrophages derived from wild-type or pyrin-deficient mice²² with F. tularensis or stimulated them with the NLRP3 stimulus of LPS plus nigericin (Fig. 2e,f). Despite notable upregulation of pyrin protein in wild-type macrophages by these stimuli, there was no notable difference between wild-type and pyrin-deficient macrophages in the processing of caspase-1, secretion of IL-1β or release of LDH. These results indicate that pyrin does not have a role in the proinflammatory and cell-death responses to F. tularensis infection in mice or negatively regulate the activity of ASC or the inflammasome, as proposed before^22,23,24.

Mechanism of activation of AIM2 by F. tularensis

Engagement of AIM2 by cytoplasmic DNA leads to formation of the oligomeric ASC 'pyroptosome' (a macromolecular aggregate of ASC), which induces pyroptotic cell death by activating caspase-1 (refs. 3,8). To investigate whether F. tularensis infection induces AIM2-dependent formation of the ASC pyroptosome, we analyzed the presence of oligomeric ASC in cell pellets of F. tularensis–infected wild-type and Aim2^−/− macrophages. As expected, F. tularensis infection induced the formation of ASC pyroptosomes in wild-type macrophages but not in Aim2^−/− macrophages (Fig. 3a). Similarly, liposome-delivered cytoplasmic DNA induced the formation of ASC pyroptosomes in wild-type but not Aim2^−/− macrophages (Fig. 3a). In contrast, LPS plus nigericin, which activates the NLRP3 inflammasome, induced the formation of ASC pyroptosomes in both wild-type and Aim2^−/− macrophages (Fig. 3a). Collectively, our data indicate that F. tularensis infection induces pyroptotic cell death via AIM2-mediated formation of ASC pyroptosomes.

Figure 3: Role of the ASC pyroptosome, potassium depletion, actin polymerization and lysosomal acidification in activation of the AIM2 inflammasome by F. tularensis.

(a) Immunoblot analysis of ASC pyroptosomes in Aim2^+/+ and Aim2^−/− BMDMs transfected with poly(dA:dT), infected for 6 h with F. tularensis (MOI, 500), or treated with LPS and nigericin as described in Figure 1b; pellets of whole-cell lysates centrifuged at 3,800g followed by crosslinking with disuccinimidyl suberate (Pell), as well as cell lysates and culture supernatants, were hybridized with anti–mouse ASC, anti–mouse AIM2 and anti-caspase-1. (b) Immunoblot analysis of ASC pyroptosomes in Aim2^+/+ BMDMs left uninfected or infected for 6 h with F. tularensis (MOI, 500) in the presence of increasing concentrations of KCl in the culture medium, then fractionated and analyzed as in a. (c,d) Immunoblot analysis of mouse procaspase-1, caspase-1, AIM2 and/or pro-IL-1β in culture supernatants and lysates of Aim2^+/+ BMDMs infected for 6 h with F. tularensis (MOI, above lanes) in the presence (+ cytoD) or absence of cytochalasin D (c) or infected for 6 h with the F. tularensis (MOI, 250) in the presence of vehicle (+ veh), bafilomycin (+ baf) or NH₄Cl (+ AC; d). (e,f) Release of LDH into culture supernatants of the BMDMs in c (e) and d (f). (e) *P < 0.001, with versus without cytochalasin D (Student's t-test); (f) *P < 0.05, **P < 0.01 and ***P < 0.001, F. tularensis versus vehicle (Student's t-test). Data are representative of two (a–d) or three (e,f) experiments (mean and s.d. in e,f).

To gain further insight into the signaling pathway by which F. tularensis infection activates the AIM2 inflammasome, we investigated the role of cytoplasmic potassium efflux, which is critical for formation of the ASC pyroptosome and recruitment of caspase-1 to ASC oligomers⁸. Inhibition of potassium efflux by increasing the potassium concentration in the culture medium completely blocked the formation of ASC pyroptosomes, activation of caspase-1 and secretion of IL-1β induced by F. tularensis infection (Fig. 3b), which indicated that depletion of intracellular potassium is required for assembly of the AIM2 inflammasome and formation of the ASC pyroptosome. Activation of caspase-1 and cell death by F. tularensis infection was blocked by the actin-oligomerization inhibitor cytochalasin D and the endosomal-acidification inhibitors bafilomycin A and NH₄Cl (Fig. 3c–f), which indicated that bacterial internalization and lysosomal acidification are required for recognition of F. tularensis infection and activation of the AIM2 inflammasome. Inhibition of F. tularensis–induced caspase-1 activation by bafilomycin was not due to inhibition of bacterial uptake or replication, as we found more intracellular bacteria in bafilomycin-treated cells than in untreated control cells at 3 h after infection (Supplementary Fig. 6). The greater bacterial count in bafilomycin-treated cells is consistent with published observations linking phagosomal acidification to the killing and degradation of phagocytosed bacteria²⁵.

IRF3 is important for AIM2 activation by F. tularensis

Activation of the inflammasome by F. tularensis infection requires an intact type I interferon response for efficient inflammasome activation¹⁰. Consistent with that, F. tularensis infection of Irf3^−/− macrophages, which are defective in the secretion of type I interferons in response to cytosolic DNA, resulted in less-efficient activation of the AIM2 inflammasome than that in wild-type macrophages (Fig. 4a). In contrast, liposome-delivered DNA induced similar activation of the AIM2 inflammasome in both wild-type and Irf3^−/− macrophages (Fig. 4a). We observed similar expression of AIM2 in wild-type and Irf3^−/− macrophages (Fig. 4a), which ruled out the possibility that Irf3^−/− macrophages have lower expression of AIM2. To determine if type I interferon signaling through its receptor, IFNAR1, could restore AIM2 activation induced by F. tularensis infection in the Irf3^−/− macrophages, we treated these macrophages with interferon-β (IFN-β) at the time of infection and assayed caspase-1 activation at 6 h after infection. Simultaneous treatment of Irf3^−/− macrophages with both IFN-β and F. tularensis restored F. tularensis–induced caspase-1 activation in these cells (Fig. 4b). Concomitant treatment with IFN-β at the time of infection slightly enhanced F. tularensis–induced activation of caspase-1 in wild-type Irf3^+/+ macrophages (Fig. 4b). Restoration of F. tularensis–induced activation of caspase-1 in Irf3^−/− macrophages required concurrent treatment with IFN-β and F. tularensis, as we observed no restoration when we treated cells with IFN-β at 2 h after infection (Supplementary Fig. 7a, left). We observed no activation of caspase-1 after IFN-β treatment alone (Supplementary Fig. 7a, right). Additionally, concurrent treatment of Aim2^−/− macrophages with IFN-β and F. tularensis did not restore F. tularensis–induced caspase-1 activation in these cells (Supplementary Fig. 7b), which indicated that IFN-β is not sufficient for F. tularensis–induced caspase-1 activation in the absence of AIM2.

Figure 4: IRF3 signaling is required for activation of the AIM2 inflammasome by F. tularensis but not by liposome-delivered DNA.

(a,b) Immunoblot analysis of mouse procaspase-1, caspase-1 and/or AIM2 in culture supernatants and lysates of mouse Irf3^−/− and Irf3^+/+ macrophages infected for 6 h with F. tularensis (MOI, in parentheses above lanes), treated with LPS and nigericin as described in Figure 1b, or transfected with poly(dA:dT) (a), or infected with F. tularensis (MOI, 250) in the presence or absence of IFN-β (b). (c) Immunoblot analysis of mouse procaspase-1, caspase-1 and/or AIM2 in culture supernatants and lysates of Ifnar1^−/− and Ifnar1^+/+ macrophages left untreated or treated for 2 h with IFN-β alone or followed by infection for 6 h with F. tularensis (MOI, in parentheses above lanes). (d) Enzyme-linked immunosorbent assay of IFN-β in culture supernatants of Aim2^−/− and Aim2^+/+ macrophages left uninfected or infected for 6 h with F. tularensis (MOI, 250). *P < 0.05 and **P < 0.005 (Student's t-test). (e) Immunoblot analysis of mouse STAT1 phosphorylated at Tyr701 (p-STAT1), total STAT1 and AIM2 in lysates of mouse Aim2^−/− and Aim2^+/+ macrophages left uninfected or infected with F. tularensis (MOI, 250). (f) Enzyme-linked immunosorbent assay of IFN-β in culture supernatants of Aim2^+/+ macrophages left untreated or infected for 6 h with F. tularensis (MOI, 250) in the presence (+ bafilo) or absence of bafilomycin (50 nM). *P < 0.01 (Student's t-test). Data are representative of two (a,c) or three (b,d–f) experiments (mean and s.d. in d,f).

To further investigate the role of type I interferons in the mechanism of activation of AIM2 by F. tularensis infection, we primed wild-type and Irf3^−/− macrophages with IFN-β 2 h before infection with F. tularensis and measured caspase-1 activation at various times after infection (Supplementary Fig. 8). In the absence of IFN-β priming, caspase-1 processing in wild-type macrophages was barely detectable at 3 h after infection. In contrast, we observed maximum caspase-1 processing at 3 h after infection in IFN-β-primed wild-type macrophages. Although unprimed Irf3^−/− macrophages showed less-efficient caspase-1 activation than that of wild-type macrophages at 3 h and 5 h after infection, they showed caspase-1 activation similar to that of wild-type macrophages after IFN-β priming. These results indicate that priming of macrophages with IFN-β accelerates and enhances activation of the AIM2 inflammasome by F. tularensis infection not only in IRF3-deficient macrophages but also in wild-type macrophages. They also suggest that the delay in caspase-1 activation observed in unprimed wild-type macrophages might have been due to the lack of type I interferons in the initial stages of F. tularensis infection. Like IRF3-deficient macrophages, IFNAR1-deficient macrophages were also defective in activation of caspase-1 in response to F. tularensis infection¹⁰ (Fig. 4c). However, this defect was not corrected by IFN-β priming, which indicated that signaling by IFNAR1 is required for efficient activation of the AIM2 inflammasome by F. tularensis infection.

The impaired AIM2 inflammasome activation in Ifnar1^−/− macrophages was not caused by a lack of intracellular bacterial replication. On the contrary, the number of intracellular F. tularensis was notably higher in Ifnar1^−/− macrophages than in wild-type macrophages at 5 h after infection and reached to almost 50-fold higher at 24 h after infection (Supplementary Fig. 9). These results are consistent with a published report¹⁰. Collectively, these observations indicate that type I interferon signaling acts upstream of the AIM2 inflammasome, perhaps to enhance the killing and lysis of F. tularensis in the phagosome to generate the cytosolic DNA that activates the AIM2 inflammasome (model, Supplementary Fig. 10, and discussed below).

AIM2 is not critical for type I interferon production

To rule out the possibility that the defective activation of the inflammasome in Aim2^−/− macrophages was due to defective type I interferon production or signaling, we quantified secreted IFN-β in culture media of wild-type and Aim2^−/− macrophages infected with F. tularensis and also determined the status of phosphorylation of the transcription factor STAT1 in lysates of these cells (Fig. 4d,e). Our results showed that IFN-β production was actually higher in Aim2^−/− macrophages than in wild-type macrophages in response to F. tularensis infection (Fig. 4d). However, there was no substantial difference between F. tularensis–infected wild-type and Aim2^−/− macrophages in STAT1 phosphorylation (Fig. 4e). These results indicate that AIM2 is not critical for type I interferon production or signaling and that AIM2 deficiency does not negatively affect the type I interferon response to infection with F. tularensis. In fact, AIM2 deficiency seemed to enhance IFN-β production in response to such infection. That result is consistent with the observation that small interfering RNA–mediated knockdown of AIM2 potentiates IFN-β production in response to transfected cytosolic DNA⁴. The production of IFN-β induced by F. tularensis infection was completely blocked by bafilomycin A (Fig. 4f), which indicated that lysosomal acidification is not only required for activation of the AIM2 inflammasome but also necessary for F. tularensis–induced production of type I interferon.

F. tularensis DNA induces AIM2 oligomerization

Cytoplasmic DNA activates AIM2 by directly binding to its HIN-200 domain, which leads to its oligomerization. We visualized this process in live cells expressing GFP-tagged AIM2 and Cy3-labeled DNA³ (Fig. 5a). Notably, we also visualized the cytosolic DNA bound to AIM2 in the oligomeric AIM2-DNA complex with the DNA-specific fluorescent stain Hoechst 33342 (Fig. 5a). To formally demonstrate that AIM2 senses cytosolic DNA produced by F. tularensis infection, we infected Nlrp3^−/−-AIM2-GFP-N1 macrophages (an Nlrp3^−/− cell line stably expressing AIM2 tagged with GFP at the C terminus)³ with F. tularensis and subsequently stained the infected cells with Hoechst dye. We used Nlrp3^−/−-AIM2-GFP-N1 macrophages in these experiments to rule out the possibility of involvement of NLRP3 in the response to cytosolic DNA or F. tularensis infection. AIM2 was evenly distributed in the cytoplasm and nuclei of uninfected cells (Fig. 5b). Infection of cells with F. tularensis resulted in AIM2 oligomerization (Fig. 5b and Supplementary Fig. 11), as shown by the clustering of AIM2-GFP. These AIM2-GFP clusters were also visible with the vital DNA-specific Hoechst stain, which indicated that AIM2 is clustered by binding to cytosolic DNA. The infected macrophages that contained the AIM2-DNA clusters showed distinct features of pyroptotic cell death.

Figure 5: Cytoplasmic DNA secreted by F. tularensis induces AIM2 oligomerization.

(a) Confocal live-cell microscopy of Nlrp3^−/−-AIM2-GFP-N1 BMDMs left untransfected or transfected with Cy3-labeled DNA. Arrow (right) indicates staining of clustered cytoplasmic DNA with the DNA-specific blue Hoechst dye. (b) Confocal live-cell microscopy of Nlrp3^−/−-AIM2-GFP-N1 BMDMs left uninfected or infected for 6 h with F. tularensis and then stained with Hoechst dye. Arrow (bottom row) indicates staining of the AIM2-GFP cluster with the DNA-specific blue Hoechst dye. (c) Confocal live-cell microscopy of Nlrp3^−/−-AIM2-GFP-N1 BMDMs left uninfected or infected for 6 h with Hoechst-labeled F. tularensis and then fixed on coverslips. Arrow (bottom row) indicates the Hoechst-labeled F. tularensis cytoplasmic DNA. Additional images, Supplementary Figure 11. DIC, differential interference contrast. Original magnification, ×40. Data are representative of at least three (a,b) or two (c) experiments.

To provide direct evidence that the cytosolic DNA responsible for the clustering of AIM2-GFP in those cells was derived from F. tularensis, we prestained F. tularensis DNA with Hoechst stain before using this F. tularensis to infect Nlrp3^−/−-AIM2-GFP-N1 macrophages. As expected, infection of these cells with the Hoechst-labeled F. tularensis resulted in clustering of AIM2-GFP around the Hoechst-labeled DNA (Fig. 5c and Supplementary Fig. 11b). Because only F. tularensis DNA was labeled with Hoechst stain, these results provide direct evidence that F. tularensis DNA is responsible for oligomerizing AIM2. Collectively, our results indicate that infection of macrophages with F. tularensis generates cytosolic DNA that specifically binds to AIM2 and induces AIM2 oligomerization, which leads to the activation of caspase-1 and cell death (model, Supplementary Fig. 10).

Innate immunity to F. tularensis infection requires AIM2

The results presented above indicated that AIM2 is important for the proinflammatory and cell-death responses to F. tularensis infection in vitro by a mechanism that involves the recognition of cytoplasmic DNA produced by this intracellular pathogen. To assess the precise role of AIM2 in the host innate immune defense against F. tularensis infection, we infected wild-type and Aim2^−/− mice subcutaneously with live F. tularensis. This resulted in much higher rate of mortality in Aim2^−/− mice than in wild-type mice (Fig. 6a). The mortality rate of Aim2^−/− mice on day 5 was 100%, compared with just 33% for wild-type mice. The remaining wild-type mice survived beyond 20 d after infection. Consistent with those results, the bacterial burdens in tissues of Aim2^−/− mice were much higher than those in wild-type mice at 48 h after infection (Fig. 6b). These results indicate that the AIM2 inflammasome has a crucial role in the host defense against F. tularensis infection, probably by decreasing the bacterial burden in tissues, thereby preventing systemic infection. Consistent with the critical role of AIM2 in the production of caspase-1-generated cytokines, the serum concentration of IL-18 was much higher in wild-type mice than in Aim2^−/− mice at 24 h after infection (Fig. 6c). Together these results indicate that the lack of AIM2 inflammasome activity increases the virulence of F. tularensis in mice because of defective activation of caspase-1, which is critical for the production of the caspase-1-generated cytokines such as IL-1β and IL-18 and the induction of pyroptotic death of the infected macrophages.

Figure 6: The AIM2 inflammasome is critical for innate immunity to F. tularensis infection.

(a) Survival of Aim2^+/+ mice (n = 9) and Aim2^−/− mice (n = 9) injected subcutaneously with F. tularensis (1.5 × 10⁵ colony-forming units (CFU)) and monitored over a period of 3 weeks; 66% of Aim2^+/+ survived beyond 3 weeks after infection. (b) Bacterial burden in livers and spleens obtained from mice 48 h after subcutaneous infection with F. tularensis (1.5 × 10⁵ CFU); organs were homogenized and dilutions were plated on cystine heart agar plates. (c) Enzyme-linked immunosorbent assay of IL-18 in serum from Aim2^+/+ mice (n = 3) and Aim2^−/− mice (n = 3) at 1 d after subcutaneous infection with F. tularensis. In b,c, each symbol represents an individual mouse; small horizontal lines indicate the mean. P values, Kaplan-Meier log-rank test (a) or Student's t-test (b,c). Data are representative of two experiments.

Discussion

Biochemical evidence has linked AIM2 to the recognition of and innate immune response to cytosolic DNA^3,4,5,6. The recognition of cytosolic DNA by AIM2 activates the nontranscriptional inflammatory caspase-1 pathway in macrophages, which leads to production of the potent inflammatory cytokines IL-1β and IL-18 and cell death. We have now provided genetic evidence that AIM2 is a critical PRR uniquely involved in the recognition and innate immune defense against infection with the potentially lethal intracellular pathogen F. tularensis.

Several proteins of the NLR family are involved in the activation of the inflammatory caspase-1 pathway in macrophages in response to pathogenic infection or products^26,27,28. For example, NLRC4 (Ipaf) is a PRR involved in the detection of intracellular infection with S. typhimurium and activation of the caspase-1 pathway in response to this pathogen¹⁸. S. typhimurium flagellin is the suggested molecular component responsible for NLRC4 activation^29,30. Another PRR, NLRP1, is involved in caspase-1 activation in response to stimulation with anthrax lethal toxin¹⁵. In contrast, NLRP3 (cryopyrin) is involved in caspase-1 activation after infection with a broad range of bacterial, viral or fungal pathogens and in response to various stress stimuli^28,31. Nevertheless, the mechanism by which NLRP3 recognizes such a broad range of pathogens and stimuli is still unclear.

None of the NLRs mentioned above or other known NLRs, however, are involved in sensing F. tularensis infection. Although macrophages from AIM2-deficient mice are not defective in sensing NLRC4, NLRP1 and NLRP3 stimuli, as shown by their normal caspase-1 activation response to S. typhimurium, anthrax lethal toxin, or LPS plus ATP or nigericin, or monosodium urate, they were clearly unable to activate caspase-1 in response to F. tularensis infection. This suggests that F. tularensis may have developed mechanisms to evade detection of their non–nucleic acid pathogen-associated molecular patterns by the host cell inflammasome components such as NLRC4, NLRP1 or NLRP3 and probably by other as-yet-uncharacterized inflammasomes. Nevertheless, several studies have suggested that F. tularensis infection activates Toll-like receptor 2 signaling via the adaptor MyD88, leading to transcription factor NF-κB–dependent transcriptional induction of proinflammatory cytokines^32,33,34. However, this response does not seem to be sufficient for activation of the NLRP3 inflammasome, because Aim2^−/− macrophages were completely defective in caspase-1 activation in response to infection with F. tularensis.

Phagosomal disruption occurs during the escape of F. tularensis from the phagosome into the cytosol^35,36,37. This may be necessary for the activation of caspase-1 and production of type I interferon induced by F. tularensis, because mutant strains that cannot disrupt the phagosome and escape into the cytosol are unable to induce type I interferons or activate caspase-1 (refs. 10,38). Despite published observations of activation of the NLRP3 inflammasome by phagosomal disruption^26,39,40, phagosomal disruption during F. tularensis escape did not lead to activation of the NLRP3 inflammasome, as shown by the absence of caspase-1 activation in F. tularensis–infected AIM2-deficient macrophages. This defect in caspase-1 activation cannot be attributed to a defect in NLRP3 signaling, because these macrophages had normal amounts of caspase-1 activation in response to specific NLRP3 stimuli such as LPS plus ATP or nigericin. Additionally, this defect is probably not due to a specific defect in F. tularensis–induced phagosomal disruption, as infection of the AIM2-deficient macrophages with F. tularensis led to even more production of type I interferon than that of wild-type macrophages, which indicates normal phagosomal disruption and escape. Together these observations suggest that phagosomal disruption generally does not induce activation of the NLRP3 inflammasome and it is likely that a more specific stimulus or additional stimuli are required. Alternatively, F. tularensis might have the means to inhibit activation of the NLRP3 inflammasome during its escape from the phagosome.

Activation of the AIM2 inflammasome requires direct interaction of AIM2 with DNA³. This interaction leads to oligomerization of AIM2 and the formation of a large AIM2-DNA complex visible by confocal microscopy³. Our results have shown that infection of macrophages with F. tularensis also induced the formation of a large AIM2-DNA complex, which indicates that F. tularensis infection delivers DNA into the cytosol that is then recognized by AIM2. The exact mechanism by which F. tularensis delivers its DNA into the cytosol is unclear at present. Nevertheless, our results suggest that the DNA that activates AIM2, and probably type I interferon signaling, is produced by breakdown and digestion of killed F. tularensis by phagosomal enzymes. In support of that conclusion, bafilomycin, which can prevent the killing and degradation of bacteria in the phagosome by inhibiting the activity of vacuolar-type ATPases that mediate lumen acidification^25,41, completely inhibited the activation of the AIM2 inflammasome and production of IFN-β induced by F. tularensis infection in wild-type macrophages. Given that bafilomycin does not prevent F. tularensis phagosomal permeabilization and escape into the cytosol³⁵, phagosomal escape by live F. tularensis and its replication in the cytosol is probably not the signal that activates AIM2 or the type I interferon response. Instead, the killing and degradation of F. tularensis and the subsequent phagosomal disruption triggered by the escape of live F. tularensis is probably responsible for the delivery of F. tularensis DNA into the cytosol for recognition by AIM2 and type I interferon signaling pathways. As a consequence, we expect that more virulent strains of F. tularensis such as F. tularensis tularensis, which infects humans, must have evolved mechanisms that protect them from phagosomal lysis, perhaps by altering the rate of acidification and maturation of their phagosomes³⁶. Therefore, these strains would be detected less efficiently by AIM2 because of less phagosomal release of killed bacterial DNA into the cytosol and thus would have a better chance of intracellular replication and a greater ability to cause systemic infection.

Signaling by IFN-β through its receptor, IFNAR1, is required for efficient caspase-1 activation by F. tularensis^10,26. IFNAR1- or IRF3-deficient macrophages, which have impaired IFN-β production, showed less efficient activation of caspase-1 by F. tularensis than did wild-type macrophages^10,26. Our data showed that the lower caspase-1 activation by F. tularensis in Irf3^−/− or Ifnar1^−/− macrophages was not due to lower AIM2 expression, because these macrophages had AIM2 expression similar to that of wild-type macrophages. Additionally, Irf3^−/− and Ifnar1^−/− macrophages⁴² show normal caspase-1 activation in response to transfected DNA, which rules out the possibility that IFN-β signaling is required for transcriptional induction of a host factor critical for caspase-1 activation by the AIM2 inflammasome. Although F. tularensis and transfected DNA both activate the AIM2 inflammasome, IFN-β signaling might be required for efficient killing of F. tularensis in the phagosome and/or phagosomal permeablization and release of F. tularensis DNA into the cytosol. Indeed, Ifnar1^−/− macrophages contained more bacteria at 5–24 h after infection than did wild-type macrophages, probably because of defective phagosomal bactericidal activity of the Ifnar1^−/− macrophages. Consequently, these macrophages might produce less cytosolic DNA, which could provide a rationale for the observed defective AIM2 inflammasome activation in these macrophages.

On the basis of the information presently available about the life cycle of F. tularensis in macrophages^10,26,43 and our own data, we propose a two-step mechanism by which infection of macrophages with F. tularensis activates the AIM2 inflammasome. Shortly after F. tularensis enters the phagosome, the phagosome is rapidly acidified¹⁷. Acidification causes lysis of some of the ingested bacteria and release of bacterial DNA into the lumen of the phagosome. During phagosomal escape of live F. tularensis, which occurs as early as 1 h after infection^16,17, the phagosome is ruptured, which releases both live F. tularensis and undigested DNA of killed F. tularensis into the cytosol. This amount of cytosolic DNA might not be sufficient for AIM2 activation but is probably sufficient for the activation of an as-yet-unknown DNA sensor, which in turn activates IRF3, leading to the production of type I interferons such as IFN-β. Bafilomycin treatment completely inhibited F. tularensis–induced production of IFN-β, which supports this scenario. IFN-β binds to IFNAR1, resulting in the activation of this signaling pathway. This initial IFN-β signaling process acts like a positive feedback loop, perhaps to increase phagosomal acidification and/or the bactericidal activity of the phagosomal enzymes, thereby enhancing the release of more DNA from killed F. tularensis into the cytosol. This is supported by the observations that prior priming of wild-type or Irf3^−/− macrophages before infection with F. tularensis or infection of these macrophages with F. tularensis together with IFN-β cotreatment accelerates and enhances AIM2 inflammasome activation, whereas inhibiting phagosomal acidification results in much less activation of the AIM2 inflammasome. In the second step of this process, the higher concentration of cytosolic DNA, as a result of IFN-β-induced phagosomal disruption, leads to full AIM2 activation through the induction of its oligomerization. The oligomerized AIM2-DNA complex serves as a molecular platform for the recruitment of ASC and facilitate its oligomerization into the large ASC pyroptosome⁸. Consistent with that, infection with F. tularensis or transfection with cytosolic DNA induced formation of the ASC pyroptosome in wild-type but not Aim2^−/− macrophages and was completely blocked by ∼120–180 mM extracellular KCl. It is likely that the ASC pyroptosome then activates procaspase-1, as described before⁸, leading to pyroptotic cell death and production of the proinflammatory cytokines IL-1β and IL-18.

The phenotype of F. tularensis–challenged Aim2^−/− mice provides further support for the idea that AIM2 is critical for sensing F. tularensis infection and activation of the ASC–caspase-1 pathway not only in isolated macrophages but also in a whole-animal model. Indeed, this phenotype is reminiscent of the phenotype of F. tularensis–challenged caspase-1-deficient or ASC-deficient mice⁹. In these mouse strains, as well as in Aim2^−/− mice, F. tularensis infection is associated with more death due to a greater bacterial burden and systemic infection⁹. The greater bacterial burden in these mice is most probably due to less death of F. tularensis–infected macrophage and less proinflammatory cytokine production. F. tularensis–induced macrophage death shares features of the pyroptotic cell death induced by other pathogenic bacteria⁹. However, F. tularensis–induced macrophage death is dependent on AIM2 as well as caspase-1 and ASC⁹, which provides further proof that AIM2, ASC and caspase-1 function in the same signaling pathway that recognizes F. tularensis infection. In conclusion, our data have provided genetic and biochemical evidence that activation of the AIM2 inflammasome represents a crucial innate immune defense against F. tularensis infection. Future studies with the AIM2-deficient mice should clarify its role in the innate immune response to other intracellular microbial and viral pathogens as well as its involvement in nucleic acid–dependant autoimmune diseases such as systemic lupus erythematosus.

Methods

Generation of AIM2-deficient mice.

AIM2-deficient mice were generated by the gene-trap method¹³ from an embryonic stem cell clone obtained from the International Gene Trap Consortium¹². All mice were used according to protocols approved by Institutional Animal Care and Use Committee, Thomas Jefferson University. Details are in the Supplementary Methods.

Antibodies and reagents.

Polyclonal antibody to AIM2 was raised in rabbits (Invitrogen) against a mixture of recombinant mouse and human AIM2 proteins prepared in the Alnemri laboratory. This antibody can detect endogenous mouse and human AIM2 proteins. Polyclonal antibody to mouse pyrin was raised in rabbits (Invitrogen) against a truncated recombinant mouse pyrin protein (residues 1–349) prepared in the Alnemri laboratory. Monoclonal antibody 3ZD (anti-IL-1β) was obtained from the National Cancer Institute Preclinical Repository, Biological Resources Branch. Other antibodies have been described^8,20,31: polyclonal antibody to mouse NLRP3 pyrin domain (from the Alnemri laboratory), polyclonal antibody to mouse ASC (from J. Sagara) and monoclonal antibody to mouse caspase-1 p20 (from J. Yuan). Antibody to STAT1 phosphorylated at Tyr701 (9171S) and antibody to mouse STAT1 (9172) were from Cell Signaling. ATP, nigericin, poly(dA:dT) sodium salt, bafilomycin A and cytochalasin D were from Sigma-Aldrich. Ultrapure LPS was from Invivogen. Anthrax lethal factor and protective antigen were from List Biological Laboratories. Disuccinimidyl suberate was from Thermo Fisher Scientific. The CytoTox 96 LDH-release assay kit was from Promega.

Macrophage cell culture and stimulation.

Mouse bone marrow cells were isolated from mouse femurs in sterile DMEM and were cultured in six-well plates for 5–7 d in DMEM (GIBCO) supplemented with 10% (vol/vol) medium conditioned by L929 mouse fibroblasts, 10% (vol/vol) FBS and 1% (vol/vol) penicillin-streptomycin (15140; Gibco). Macrophages were infected for 18 h with vaccinia virus (WR strain) at various MOI values. In some experiments, differentiated macrophages were treated for 5 h with Ultrapure LPS (500 ng/ml) in serum-free Opti-MEM I medium, followed by ATP (5 mM) or nigericin (2.5 μM). Macrophages were treated with anthrax lethal toxin (protective antigen and lethal factor) at a concentration of 5 μg/ml.

Infection of macrophages with F. tularensis.

Wild-type F. tularensis (U112) was obtained from D.M. Monack and was grown overnight with shaking in tryptic soy broth (Difco Laboratories) supplemented with 0.2% (wt/vol) cysteine as described¹⁰. For analysis of the effect of AIM2 deficiency on F. tularensis–induced activation of caspase-1, Aim2^−/− and Aim2^+/+ macrophages were seeded in six-well plates at a density of 2 × 10⁶ cells per well in Opti-MEM I serum-free medium and were allowed to attach for 2 h. Cells were infected for 6 h with F. tularensis (at various MOI values) in 2 ml Opti-MEM I medium as described¹⁰. Culture supernatants and cells were separated and then processed for immunoblot analysis as described below. In the experiments involving IFN-β priming, cells were pretreated with 250 U/ml of mouse IFN-β at 2 h before infection. Cells were then infected with F. tularensis. After 1 h, gentamycin (100 μg/ml) was added, followed by incubation for 30 min. Cells were then washed with PBS and were further incubated in Opti-MEM containing a low concentration of gentamycin (2.5 μg/ml) for various periods of time until they were collected.

Immunoblot analysis.

Proteins were precipitated from cell culture supernatants by the addition of an equal volume of methanol and 0.25 volumes of chloroform as described^3,40. Samples were separated by 12.5% SDS-PAGE and were transferred onto nitrocellulose membranes. Blots were probed with rat monoclonal antibody to mouse caspase-1 p20 or monoclonal antibody to IL-1β (3ZD). Total cell lysates were mixed with SDS sample buffer and separated by 12.5% SDS-PAGE and then were analyzed by immunoblot with the appropriate antibodies as described above.

Assay of ASC pyroptosome formation in macrophages.

These experiments were done essentially as described^44,45 (additional description, Supplementary Methods).

Infection of mice with F. tularensis.

All mice were kept in specific pathogen–free conditions in filter-top cages at Thomas Jefferson University and experimental studies were in accordance with the Institutional Animal Care and Use Guidelines. Mice were provided with sterile water and food ad libitum. Pairs of Aim2^−/− and Aim2^+/+ littermates 8–10 weeks of age were inoculated subcutaneously with the appropriate dose of F. tularensis strain U112 in a volume of 0.05 ml. Each pair of mice were siblings derived from the same parents. Mice were monitored for signs of distress and/or illness and lethality three times daily for 3 weeks for the survival study. For the determination of bacterial burden in mouse tissues, six pairs of Aim2^−/− and Aim2^+/+ littermates (siblings) 10–12 weeks of age were inoculated subcutaneously with the appropriate dose of F. tularensis strain U112 in a volume of 0.05 ml. Spleens and livers were collected 48 h after infection and homogenized, and dilutions were plated on cystine heart agar plates and incubated for 24 h at 37 °C; CFU were counted and are presented as CFU per gram of tissue.

LDH-release assay.

Cell culture supernatants and cell pellets from treated macrophages were collected at the end of treatment and LDH activity was assayed with the CytoTox 96 LDH-release kit as described by the manufacturer's protocols (Promega).

Confocal microscopy.

In experiments with Cy3-labeled plasmid DNA, Nlrp3^−/−-AIM2-GFP-N1 or Aim2^−/−-AIM2-GFP-N1 macrophages were seeded on 35-mm glass-bottomed culture dishes (Mat Tek) and were allowed to attach for 24 h. The next day cells, were transfected for 2–3 h with Cy3-labeled plasmid DNA (0.5 μg/dish) using Lipofectamine 2000 and then were stained for 30 min with Hoechst 33342. Cells were then observed with a Zeiss LSM 510 Meta confocal microscope at the Kimmel Cancer Center core facility. In F. tularensis infection experiments, Nlrp3^−/−-AIM2-GFP-N1 macrophages seeded on 35-mm glass-bottomed culture dishes were infected for 6 h with unstained or Hoechst-stained F. tularensis (MOI, 200). Cells infected with unstained F. tularensis were then stained with Hoechst 33342 stain; cells infected with Hoechst-stained F. tularensis were not stained after infection. Cells were observed with a Zeiss LSM 510 Meta confocal microscope. GFP (green) was excited with a 488-nm argon laser; Hoechst 33342 (blue) was excited with a 405-nm diode laser; and Cy3 (red) was excited with a 543-nm He-Ne laser. The stable Nlrp3^−/−-AIM2-GFP-N1 or Aim2^−/−-AIM2-GFP-N1 cell lines were generated by retroviral transduction, followed by sorting with Flow Cytometry at the Kimmel Cancer Center core facility, as described³.

Statistics.

Samples were analyzed with Student's t-test.

Note: Supplementary information is available on the Nature Immunology website.