The term inflammasome was coined by Tschopp and co-workers1 in 2002 to describe a high-molecular-weight complex present in the cytosol of stimulated immune cells that mediates the activation of inflammatory caspases1. Since this seminal report, the field has expanded substantially, and multiple distinct inflammasomes have been identified, with the assembly of each being dictated by a unique pattern-recognition receptor (PRR) in response to pathogen-associated molecular patterns (PAMPs) or endogenous danger signals in the cytosol of the host cell. Recognition of the inflammatory ligand results in sensor activation, oligomerization and the recruitment of an adaptor protein known as ASC, which consists of two death-fold domains: a pyrin domain (PYD) and a caspase recruitment domain (CARD). These domains allow ASC to bridge the upstream inflammasome sensor molecule to caspase 1. Proximity-induced autoprocessing results in the formation of the catalytically active protease caspase 1, which initiates downstream responses, including the release of interleukin-1β (IL-1β) and IL-18, and induces pyroptosis, which is a lytic form of cell death.

Inflammasomes have been recognized for their crucial role in host defence against pathogens2, but dysregulated inflammasome activation is linked to the development of cancer and autoimmune, metabolic and neurodegenerative diseases. Therefore, the tight control of inflammasome assembly and signalling is crucial to allow the immune system to initiate antimicrobial and inflammatory responses while avoiding overt tissue damage.

Recent studies have started to unravel the upstream signalling events that result in the engagement of distinct inflammasome-forming receptors and have also started to shed light on the quaternary structure of these large multiprotein complexes. Furthermore, the identification of the pyroptosis mediator gasdermin D established a new paradigm for understanding pyroptosis and other forms of programmed cell death3,4. In this Review, we discuss the latest insights into the activation and assembly mechanisms of inflammasomes, the structure of distinct inflammasome complexes and the mechanisms that regulate inflammasome signalling on a molecular level.

Signals that lead to receptor activation

To date, five receptor proteins have been confirmed to assemble inflammasomes, including the nucleotide-binding oligomerization domain (NOD), leucine-rich repeat (LRR)-containing protein (NLR) family members NLRP1, NLRP3 and NLRC4, as well as the proteins absent in melanoma 2 (AIM2) and pyrin (Figs 1,2). This set of so-called canonical inflammasomes is complemented by the non-canonical pathway, which targets caspase 11 in mice and caspase 4 and/or caspase 5 in human cells (Fig. 3). The ligands and activation mechanisms of these pathways are well characterized and are discussed in more detail below. There may be other less well-characterized pathways, as NLRP6, NLRP7, NLRP12, retinoic acid-inducible gene I (RIG-I; also known as DDX58) and interferon-γ (IFNγ)-inducible protein 16 (IFI16) have been reported to promote caspase 1 activation2,5.

Figure 1: The NLR inflammasomes.
figure 1

a | Mouse NLRP1B and rat NLRP1 respond to Bacillus anthracis lethal factor and Toxoplasma gondii infection6,7,11. Cleavage of NLRP1B or NLRP1 by lethal factor results in the removal of an amino-terminal portion and receptor activation7,9. Autoprocessing within the function-to-find domain (FIIND) is also required for activation10. b | Potassium efflux is a common event that is associated with a diverse array of NLRP3 stimuli13. NLRP3 activation also requires NIMA-related kinase 7 (NEK7), which binds to the NLRP3 leucine-rich repeats (LRRs) and is required for its oligomerization14,15,16. c | NLR family, apoptosis inhibitory proteins (NAIPs) function as direct receptors for bacterial flagellin and the type 3 secretion system (T3SS) needle and rod subunits21,22. Ligand binding results in NAIP activation, allowing it to recruit and activate NLRC4. Activated receptors recruit the bipartite adaptor protein ASC, which consists of a pyrin domain (PYD) and a caspase recruitment domain (CARD), through homotypic PYD–PYD (NLRP3) or CARD–CARD interactions (NLRP1B and NLRC4). The CARD of ASC is necessary to recruit pro-caspase 1 into the complex, although CARD-containing receptors (NLRP1B and NLRC4) might also directly recruit pro-caspase 1. Following its proximity-induced autoproteolytic activation within the complex, caspase 1 processes pro-interleukin-1β (pro-IL-1β) and pro-IL-18 to their mature forms and cleaves gasdermin D3,4,84. The N-terminal fragment of gasdermin D drives pyroptosis, which is a lytic type of cell death, and allows the release of mature IL-1β and IL-18 from the cell. mtDNA, mitochondrial DNA; NOD, nucleotide-binding oligomerization domain; P2RX7, P2X purinoceptor 7; ROS, reactive oxygen species.

PowerPoint slide

Figure 2: The pyrin and AIM2 inflammasomes.
figure 2

a | Pyrin detects the inactivation of RHOA by bacterial toxins or effector proteins44. These toxins target the Switch-I region of RHOA and inactivate it via the glucosylation (by Clostridium difficile toxin B), adenylylation (by Vibrio parahaemolyticus VopS and Histophilus somni IbpA), ADP-ribosylation (by Clostridium botulinum C3 toxin) and deamidation (by Burkholderia cenocepacia type 6 secretion system (T6SS) effector or effectors) of critical residues. Human pyrin has a B30.2 domain that is not found in mouse pyrin but that responds to RHOA inactivation similarly to mouse pyrin. b | Absent in melanoma 2 (AIM2) detects the presence of double-stranded host or microbial DNA in the cytosol29,30,31. Murine cytomegalovirus (MCMV), vaccinia virus and several intracellular bacteria activate AIM2, among them Francisella tularensis subsp. novicida. Recognition of F. novicida DNA requires bacteriolysis, which is facilitated by guanylate-binding proteins (GBPs)117,120. The recruitment of pro-caspase 1 to activated pyrin and AIM2 relies on the bipartite adaptor protein ASC. Following its proximity-induced autoproteolytic activation within the complex, caspase 1 processes pro-interleukin-1β (pro-IL-1β) and pro-IL-18 to their mature forms and cleaves gasdermin D3,4,84. The amino-terminal fragment of gasdermin D drives pyroptosis, which is a lytic type of cell death, and allows the release of mature IL-1β and IL-18 from the cell. CARD, caspase recruitment domain; CC, coiled-coil; PYD, pyrin domain.

PowerPoint slide

Figure 3: The non-canonical inflammasome.
figure 3

The non-canonical inflammasome pathway results in caspase 11 activation in mice or caspase 4 and caspase 5 activation in humans49,50,51,52. Lipopolysaccharide (LPS) from intracellular bacteria reaches the cytosol and is directly bound by caspase 11, caspase 4 or caspase 5. Guanylate-binding proteins (GBPs) facilitate the recognition of LPS from vacuolar bacteria, such as Salmonella enterica subsp. enterica serovar Typhimurium119. LPS binding results in the oligomerization and activation of the caspases50, which then cleave gasdermin D to induce pyroptotic cell death3,4,84. Active caspase 11 induces non-canonical activation of NLRP3 possibly by processing pannexin 1 and causing potassium efflux (not shown)86. Assembly of the NLRP3 inflammasome results in caspase 1 activation and the processing of pro-interleukin-1β (pro-IL-1β) and pro-IL-18, which are then released by gasdermin D-induced pyroptosis. CARD, caspase recruitment domain; PYD, pyrin domain.

PowerPoint slide

The NLRP1 inflammasome. NLRP1 was the first NLR to be shown to form an inflammasome complex1. Humans only have one NLRP1 protein, which features an amino-terminal PYD, a NOD, LRRs, a function-to-find domain (FIIND) and a carboxy-terminal CARD. By contrast, multiple paralogues of NLRP1 are found in rodents, such as NLRP1A, NLRP1B and NLRP1C in mice, which share this domain organization but which lack a PYD. Five different alleles of mouse Nlrp1b exist and two are associated with responsiveness to lethal toxin of Bacillus anthracis, which is an A/B toxin that consists of the protective antigen, a cell-binding protein, oedema factor and lethal factor6. The protective antigen forms a channel that translocates the zinc metalloproteinase lethal factor into the host cytosol, where it inactivates immune signalling by cleaving mitogen-activated protein kinase (MAPK) kinases. The initial hypothesis that lethal factor might cleave and activate NLRP1B6 was confirmed by studies that showed that lethal factor cleaves NLRP1 from lethal toxin-responsive Fisher rats and that cleavage is required for NLRP1-induced caspase 1 activation7. Mouse NLRP1B is also activated by lethal factor-mediated cleavage but, unexpectedly, lethal factor cleaves both lethal toxin-responsive and lethal toxin-unresponsive NLRP1B isoforms, which indicates that activation requires an additional event8,9. NLRP1B autoprocessing within the FIIND might be such a licensing event, as autoprocessing was shown to be essential for NLRP1B activity and only lethal toxin-responsive alleles undergo autoproteolytic processing10. It is likely that lethal factor-mediated cleavage relieves an intramolecular autoinhibition that is conferred by the N terminus and/or that induces conformational changes that allow oligomerization of the receptor. Intriguingly, both rat NLRP1 and mouse NLRP1B confer resistance to Toxoplasma gondii infection but the mechanism of activation in this context might be different as no processing of NLRP1B was observed11. Nevertheless, the model that has emerged is that rodent NLRP1 proteins act as decoy substrates for the protease lethal factor in response to B. anthracis, an ingenious mechanism that was first identified in several plant R (resistance) gene products.

The NLRP3 inflammasome. NLRP3 (also known as cryopyrin) responds to a surprisingly diverse set of stimuli. These include crystalline and particulate matter (such as uric acid crystals, silica, asbestos and alum), extracellular ATP, pore-forming toxins, RNA–DNA hybrids, and several viral, bacterial, fungal and protozoan pathogens12. Activation of NLRP3 also requires priming by extracellular inflammatory stimuli, which results in the transcriptional induction of NLRP3 and controls post-translational modifications that license receptor activation. Given the wide variety of stimuli, it is likely that NLRP3 responds to a common cellular event that is triggered by these activators. However, despite years of research, no unified mechanism for NLRP3 has emerged and many different mechanisms have been proposed, including the release of oxidized mitochondrial DNA, production of reactive oxygen species and mitochondrial dysfunction, lysosomal destabilization, changes in intracellular calcium levels, the formation of large nonspecific membrane pores and potassium efflux (for a detailed review see Ref. 12). The meticulous re-examination of these mechanisms has shown that potassium release is associated with all NLRP3 activators and that low potassium medium alone is sufficient to trigger NLRP3 activation13. This suggests that a reduction in intracellular potassium levels is the downstream convergence point for NLRP3 stimuli, but whether NLRP3 directly senses this low level of potassium or whether another event correlates with low intracellular potassium concentrations remains to be determined.

New insights into the mechanism of NLRP3 activation might be provided by three recent reports that have identified an unexpected role for NIMA-related kinase 7 (NEK7) in this process14,15,16. These studies found that NEK7 is essential for NLRP3 activation in response to both canonical and non-canonical stimuli and that it acts downstream of potassium efflux. NEK7 seems to be a component of the NLRP3 complex, as it directly binds NLRP3 and controls NLRP3 oligomerization14,15. This interaction requires the catalytic domain of NEK7, but the kinase activity of NEK7 is dispensable for inflammasome activation. Notably, NEK7 was also required for caspase 1 activation by NLRP3(R258W), a gain-of-function mutation that causes Muckle–Wells syndrome14. Previous work has linked NEK7 to the formation of mitotic spindles and the separation of centrosomes. Interestingly, mitotic cells show diminished NLRP3–NEK7 interactions and reduced inflammasome activation compared with cells in interphase15, indicating that NEK7 acts as a cellular switch that enforces mutual exclusivity of the inflammasome response and cell division.

The NAIP–NLRC4 inflammasome. NLRC4 was initially identified on the basis of its similarity to apoptotic-protease activating factor 1 (APAF1) and it was shown to assemble inflammasomes in response to bacterial flagellin17,18,19. Follow-up studies demonstrated that NLRC4 also responds to the rod and needle subunits of bacterial type 3 secretion systems (T3SSs)20,21. To detect these distinct ligands, mouse NLRC4 uses NLR family, apoptosis inhibitory proteins (NAIPs) as direct upstream receptors21,22. Binding of their cognate ligand — the T3SS rod protein by NAIP2, the T3SS needle protein by NAIP1, and flagellin by NAIP5 and NAIP6 — allows the NAIPs to interact with NLRC4 and initiate inflammasome assembly. Somewhat unexpectedly, ligand sensing is not conferred by the LRR, as was previously assumed, but by the NOD of mouse NAIPs23. Humans only have one NAIP, which reportedly responds to the T3SS needle subunit of Chromobacterium violaceum and other bacteria21, but does not respond to flagellin or the T3SS rod protein. However, a recent report showed that two isoforms of human NAIP exist and that the extended isoform NAIP*, which is expressed in primary human monocyte-derived macrophages, confers responsiveness to Salmonella flagellin24.

The AIM2 inflammasome. The existence of a dedicated cytosolic DNA receptor was postulated on the basis of the observation that microbial and host DNA induce caspase 1 activation in an ASC-dependent manner, but independently of NLRP3, Toll-like receptor (TLR) or interferon signalling25. Subsequently, several groups showed that this receptor is the PYHIN (pyrin and HIN domain-containing) family member AIM2 (Refs 26,27,28), which features an N-terminal PYD and a C-terminal HIN domain with tightly packed oligonucleotide or oligosaccharide binding folds. The generation of Aim2−/− mice confirmed the importance of AIM2 in coordinating host defence to infections with DNA viruses, such as mouse cytomegalovirus and vaccinia virus, and, surprisingly, also to infections with intracellular bacterial pathogens29,30,31. Follow-up studies have demonstrated that bacterial activation of AIM2 requires the lysis of bacteria, such as Francisella tularensis subsp. novicida or Listeria monocytogenes, in the cytosol31,32.

A link between AIM2 and several human diseases has been established. Increased AIM2 expression is associated with psoriasis, abdominal aortic aneurysm and systemic lupus erythematosus33,34,35. In the case of psoriasis, autoinflammation could be linked to AIM2-mediated recognition of self-DNA in the cytosol of keratinocytes34. By contrast, reduced AIM2 levels correlate with the development of prostate and colorectal cancer36,37, and it has consistently been shown that Aim2-deficient mice are hyper-susceptible to colonic cancer development38,39. These reports indicate that besides its role in host defence, AIM2 also has a major role in tumour progression, possibly by sensing self-DNA. Indeed, the release of self-DNA into the cytosol owing to transient nuclear envelope rupture has been reported for cancer cells, but whether AIM2 detects this event remains to be shown.

The pyrin inflammasome. The most recent addition to the group of inflammasome-forming receptors is pyrin (also known as marenostrin and TRIM20), which is encoded by the gene MEFV. Pyrin features a PYD, two B-boxes and a coiled-coil domain. Unlike mouse pyrin, human pyrin also contains a C-terminal B30.2 domain (also known as a SPRY/PRY domain). Mutations within the B30.2 domain of pyrin are associated with the autoinflammatory disease familial Mediterranean fever (FMF), and result in excessive caspase 1 activation and IL-1β release. Initially, it was postulated that pyrin functions as a negative regulator of other inflammasomes or IL-1β release, and that FMF is caused by loss-of-function mutations in the B30.2 domain40,41. However, Chae et al.42 reported that Mefv−/− macrophages responded normally to inflammasome activators and they proposed that FMF is instead caused by gain-of-function mutations in Mefv, as knock-in mice with mutant B30.2 alleles showed increased caspase 1 activity, which was ASC dependent but NLRP3 independent42. Another mutation, which results in the loss of a 14-3-3-binding motif at phosphorylated Ser242, has recently been reported to cause a different type of autoinflammatory disease that is known as pyrin-associated autoinflammation with neutrophilic dermatosis43.

The physiological function of pyrin in immunity was only recently uncovered when it was shown that bacterial toxins, such as Clostridium difficile toxin B and Clostridium botulinum C3 toxin, and effector proteins, such as VopS from Vibrio parahaemolyticus and IbpA from Histophilus somni, trigger inflammasome assembly via pyrin44. Interestingly, pyrin seems to specifically recognize the inactivation of RHOA by these proteins but does not recognize the inactivation of the RHO family members RAC and CDC42 and seems to respond irrespective of the type of modification. This might indicate that pyrin does not directly interact with RHOA, but that it senses a downstream event, perhaps related to the modulation of actin dynamics. A direct interaction of pyrin with actin has been reported45, and a recent study showed that mice homozygous for a polymorphic allele of WD repeat-containing protein 1 (Wdr1), which encodes a factor required for actin depolymerization, show pyrin-dependent autoinflammatory disease and thrombocytopaenia46. These results support the idea that pyrin detects a specific pathological disturbance of cytoskeleton dynamics. However, the biological relevance of the pyrin inflammasome for host defence has so far only been shown for Burkholderia cenocepacia, whereas C. difficile infections mainly result in NLRP3 inflammasome activation47.

Given that only human pyrin contains a B30.2 domain but that both mouse and human pyrin respond to RHOA modification, the role of the B30.2 domain in this process is unclear. Pyrin also belongs to the tripartite motif-containing (TRIM) protein family, although the signature RING domain has been replaced by the PYD. Several TRIMs participate in selective autophagy of protein cargo. Consistent with a role for pyrin in autophagy, a recent study found that it targets NLRP1, NLRP3 and caspase 1 for autophagic degradation, and that the B30.2 domain functions to recognize its cargo48. This study supports previous reports that pyrin acts as a regulator of inflammasome signalling, as FMF-associated mutations in the B30.2 domain reduced cargo binding and degradation.

The non-canonical inflammasome. The non-canonical pathway initiates the activation of caspase 11 in mouse cells and caspase 4 and caspase 5 in human cells in response to lipopolysaccharide (LPS; a component of the Gram-negative bacterial cell wall) in the host cell cytosol49,50,51,52. Upon activation, caspases 4, 5 and 11 initiate pyroptosis similarly to caspase 1, but they do not cleave pro-IL-1β or pro-IL-18 (Refs 49,53). The secretion of mature cytokines is nevertheless observed, as activation of these caspases triggers the assembly of a NLRP3 inflammasome49,54,55,56. Surprisingly, caspase 11-induced pyroptosis was shown to mediate LPS-induced lethality in a mouse model of endotoxaemia and Escherichia coli-induced septic shock, which had previously been attributed to TLR4 (Refs 49,51,52). This discrepancy was resolved by the observation that the induction of caspase 11 by injection of poly(I:C) rendered wild-type and Tlr4-deficient mice susceptible to LPS-induced lethality, which indicates that caspase 11 drives LPS-induced endotoxaemia, whereas TLR4 is only required to prime this response51,52. Therefore, host cells have developed two independent systems to recognize LPS in the extracellular and intracellular spaces.

By analogy to the assembly mechanisms of canonical inflammasomes, a CARD-containing receptor was expected to mediate LPS recognition and caspase 11 recruitment. However, surprisingly, a recent study has suggested that no receptor protein is required; the authors found that caspases 4, 5 and 11 can directly bind LPS, and that this results in the oligomerization and activation of these caspases50. Intriguingly, LPS binding was conferred by the caspase 11 CARD, but not by the closely related CARD of caspase 1. The structure of the CARD in complex with LPS is currently unknown and it will be interesting to determine whether there is structural resemblance to other LPS-binding proteins. If confirmed, this mechanism suggests a new activation mode for inflammatory caspases; however, it still remains to be shown whether additional factors are required to promote cytosolic LPS recognition, similarly to LPS-binding protein, CD14 and MD2, which facilitate the detection of extracellular LPS by TLR4.

Other inflammasome complexes. Several other types of inflammasome might also exist, with the best documented being the NLRP6 inflammasome and the IFI16 inflammasome. Evidence for the existence of an NLRP6 inflammasome derived from observations that mice deficient in Nlrp6, Asc or Casp1 featured an invasive dysbiotic microbiota that increases their susceptibility to chemically induced colitis and colitis-induced tumorigenesis57. This phenotype was linked to a deficiency in basal NLRP6-dependent IL-18 secretion from intestinal epithelial cells, leading to the hypothesis that NLRP6 assembles an inflammasome complex in response to as yet unknown signals. An investigation into the nature of these signals identified several microbiota-derived metabolites that modulate the NLRP6-dependent IL-18 production in colonic explants either positively (taurine) or negatively (spermine and histamine)58. Intriguingly, it was also found that a dysbiotic microbiota is characterized by reduced taurine production and increased production of spermine, and that IL-18 production was required for the expression of antimicrobial peptides, some of which were crucial to maintain microbial diversity. These findings exemplify how the crosstalk between the microbiota and the host maintains intestinal homeostasis and explain why abrogation of IL-18 can lead to commensal dysbiosis. Nevertheless, how NLRP6 controls IL-18 production still needs to be defined, as NLRP6 was also reported to regulate nuclear factor-κB (NF-κB) and MAPK signalling59, to regulate goblet cell mucus secretion60, and to participate in the recognition of viral RNA and the subsequent production of type I IFNs and type III IFNs61.

The human PYHIN family member IFI16 and its mouse orthologue IFI204 have emerged as crucial regulators of STING (stimulator of IFN genes)-dependent IFN production during viral and bacterial infections62,63. Conversely, IFI16 was shown to interact with ASC in the nucleus of Kaposi's sarcoma-associated virus (KHSV)-infected cells and to be required for KHSV-induced caspase 1 activation64. Another link between IFI16 and caspase 1 activation has been reported for HIV infections65,66. In these cases, quiescent 'bystander' CD4+ T cells, which are nonpermissive to HIV infection, accumulate incomplete HIV transcripts in the cytosol. IFI16 was shown to recognize these viral DNAs, leading to caspase 1-dependent death of CD4+ T cells, which is a principle driver of acquired immunodeficiency syndrome65,66. However, because IFI16-dependent IFN production might also enhance caspase 1 activation (see below), additional studies will be necessary to unequivocally prove the existence of an IFI16 inflammasome.

Inflammasome assembly and structure

The receptor complex. Whereas the signals that lead to receptor activation are well studied, insights into the structure of the inflammasome complex have only now begun to emerge (Fig. 4). Given the functional relationship between inflammasomes and the apoptosome, it was hypothesized that inflammasomes adopt a comparable wheel-shaped architecture. Negative stain and cryo-electron microscopy studies of purified flagellin–NAIP5–NLRC4 and PrgJ–NAIP2–NLRC4 inflammasomes confirmed this assumption and demonstrated that these complexes adopt a wheel- or disk-like architecture, with 10–12 spokes that correspond to the individual protomers67,68,69,70. Mechanistic insights into the assembly of these complexes were provided by the observation that a striking 90° hinge rotation accompanies NLRC4 activation68,69,71. This conformational change results in the formation of a new oligomerization surface that interacts with the next protomer in the complex and thus facilitates progressive oligomerization. Importantly, this finding might provide the structural basis for understanding gain-of-function mutations in NLRC4, which cause autoinflammation with recurrent macrophage activation syndrome and which map to this important hinge region72,73,74. Interestingly, NAIPs themselves are precluded from self-oligomerization and thus only a single NAIP is found per complex. These data support a model in which ligand binding activates the NAIPs, which allows them to recruit NLRC4 and to induce the conformational changes that mediate progressive NLRC4 oligomerization. As the studies used CARD-truncated NLRC4, it is currently not known how the CARD of NLRC4 is oriented within the structure and how the complex is connected to downstream signalling components such as ASC and caspase 1.

Figure 4: Assembly of inflammasome complexes.
figure 4

a | The NAIP–NLRC4 inflammasome. Binding of their specific ligand activates mouse and human NLR family, apoptosis inhibitory proteins (NAIPs) and allows them to interact with and activate NLRC4 (Refs 21,22,23). This interaction results in the generation of a new nucleating surface, which in turn allows NLRC4 molecules to recruit and further activate NLRC4 molecules in a domino-like reaction68,69. The completed NAIP–NLRC4 complex is a multi-subunit disk-like structure that contains 9–11 molecules of NLRC4, but only one NAIP molecule68,69. Following receptor complex assembly, the caspase recruitment domains (CARDs) of NLRC4 cluster and recruit bridging ASC molecules to initiate ASC polymerization. A NAIP–NLRC4 complex can also directly initiate caspase 1 activation in the absence of ASC (not shown), although ASC recruitment seems to be the default situation. b | The absent in melanoma 2 (AIM2) inflammasome. DNA is bound in regular intervals by the HIN domain of AIM2, resulting in the clustering of the AIM2 pyrin domains (PYDs)75,76. This allows the PYDs of AIM2 to recruit ASC and to initiate ASC polymerization into filaments. ASC filaments aggregate to form a macromolecular assembly that is known as the ASC speck. The CARD of ASC, which is exposed on the surface of the ASC filaments recruits pro-caspase 1 and in turn initiates filaments that are formed by the CARD of caspase 1 and that are necessary for proximity-induced activation of the caspase79,80. LRR, leucine-rich repeat; NOD, nucleotide-binding oligomerization domain.

PowerPoint slide

A fundamentally different complex is assembled by AIM2, which, unlike NLRs, lacks a NOD that could mediate self-oligomerization. Therefore, cytosolic DNA, which is bound by the HIN domain of AIM2 at regular intervals, was proposed to provide an oligomerization template or scaffold75. A recent study confirmed that the AIM2HIN forms filaments with double-stranded DNA (dsDNA) but the full-length protein could not be examined, probably owing to AIM2PYD-induced aggregation76. Modelling based on the crystal structure of human AIM2HIN in complex with DNA suggests that every AIM2HIN occupies four base pairs and is in contact with six adjacent AIMHIN molecules75,76. The relatively long linker between the AIM2HIN and the AIM2PYD would then allow the PYDs from several AIM2 molecules to swing around the DNA core and oligomerize, thereby nucleating ASC polymerization.

The ASC speck. Following its assembly, the receptor complex recruits the adaptor ASC and pro-caspase 1 via homotypic PYD–PYD and CARD–CARD interactions. This correlates with the oligomerization of ASC into a single macromolecular aggregate that is known as the ASC speck. Speck formation is observed irrespective of which receptor is activated and ASC specks were even found to be released into the extracellular space, where they enhance inflammatory responses77,78. ASC specks appear to be formed by filaments of ASC77, which is consistent with cryo-electron microscopy and solid-state nuclear magnetic resonance analysis studies that show that human and mouse ASCs oligomerize through their PYDs into long helical filaments79,80. The ASCCARD is exposed on the surface of these filaments80 and acts as a recruitment point for pro-caspase 1. Surprisingly, it also nucleates filaments that are formed by the caspase 1CARD and might be necessary for proximity-induced activation of the caspase79. The clustering of ASC filaments into ASC specks also seems to be mediated by the ASCCARD, as filaments (but no specks) are observed in cells expressing only the ASCPYD or full-length ASC with an inactive CARD (M. S. Dick and B.P., unpublished observations).

Although the adaptor ASC is essential for signalling by PYD-containing receptors, CARD-containing receptors (such as NLRP1B and NLRC4) could in theory directly recruit and activate pro-caspase 1. In fact, wild-type and Asc−/− macrophages display comparable levels of pyroptosis in response to NLRC4 or NLRP1B activators19,81,82. Asc-deficient cells, however, release significantly reduced levels of IL-1β19,81,82, indicating that cytokine maturation is closely linked to ASC speck formation. Consistently, single point mutations in ASC that abrogate ASC oligomerization, but that leave the receptor–ASC interaction intact, resulted in strikingly reduced levels of speck formation, IL-1β maturation and caspase 1 processing after AIM2, pyrin or NLRP3 activation (M. S. Dick and P.B., unpublished observations). Intriguingly, however, gasdermin D processing and pyroptosis were not affected. The formation of ASC filaments thus serves as a signal amplification mechanism for inflammasome-mediated cytokine production by generating a multitude of potential caspase 1 activation sites. As a single ligand-bound NAIP is able to initiate the assembly of a NAIP–NLRC4 complex, which in turn initiates ASC filament formation and caspase 1 activation, inflammasomes can translate the detection of a miniscule amount of ligand into a robust cellular response.

Effector functions of inflammasomes

The use of gene-deficient mice has been essential in determining the role of inflammasomes in host defence and in mouse models of autoinflammatory and inflammatory diseases. These studies have highlighted the importance of inflammasome effector mechanisms, mainly pyroptosis and the release of the cytokines IL-1β and IL-18. However, until recently, the molecular basis of how inflammatory caspases induce these responses was unknown.

Pyroptosis. The pro-inflammatory cell death that is induced by inflammatory caspases (caspase 1 and caspase 11 in mice or caspase 4 and caspase 5 in humans) was named pyroptosis, from the Greek 'pyro' (meaning fire or fever) and 'ptosis' (meaning to fall)83. Pyroptosis is morphologically distinct from apoptosis and is characterized by cell swelling, lysis and the release of cytoplasmic content, presumably as a result of the formation of membrane pores83. Pyroptosis induction was recently shown to require gasdermin D, which is a member of the enigmatic gasdermin protein family3,4,84 (Box 1). The generation of Gsdmd−/− mice confirmed the essential role of gasdermin D in pyroptosis induction in vitro and in vivo in a mouse model of endotoxaemia3. Gasdermin D is a substrate of inflammatory, but not apoptotic, caspases, and its cleavage results in the generation of an N-terminal fragment that drives pyroptotic cell death. How gasdermin D carries out pyroptosis and whether this is linked to the formation of a membrane pore are currently unknown.

Interestingly, although gasdermin D is essential for caspase 11-induced pyroptosis3,4, pyroptosis can be observed in Gsdmd−/− cells after prolonged caspase 1 activation, indicating that there are additional pro-pyroptotic caspase 1 substrates3. Alternatively, prolonged caspase 1 activation might result in secondary necrosis, as Gsdmd−/− cells activate apoptotic caspases after canonical inflammasome stimulation in an ASC-dependent manner84, which is consistent with the ability of ASC specks to activate caspase 8 in the absence of caspase 1 (Ref. 85). A recent study has suggested that caspase 11 also cleaves pannexin 1 and that the subsequent release of ATP and the activation of P2X purinoceptor 7 (P2RX7) is required for pyroptosis induction86. Because pannexin 1 cleavage also results in the release of intracellular potassium86, these results might explain previous findings that showed that caspase 11-driven NLRP3 activation is a cell-intrinsic mechanism3,56. However, the relevance of this pathway is uncertain, as pannexin 1 and P2RX7 only have a role at early rather than late time points after caspase 11 activation49,86, and previous work has shown that Panx1−/− mice remain susceptible to intraperitoneal challenge with LPS87.

Cytokine maturation and release. Caspase 1 was initially discovered as interleukin-converting enzyme (ICE), and processing of pro-IL-1β and IL-18 into their mature biologically active forms is still its most well-known effector function. IL-1β and IL-18 are important cytokines that promote inflammation and coordinate innate and adaptive immune responses (for a review see Ref. 88). Intriguingly, they both lack a signal sequence and are released in a manner that is independent of the endoplasmic reticulum and Golgi, which is usually referred to as unconventional secretion. However, despite 25 years of research into this mechanism, it is still not understood how IL-1β exits the cell89. Microvesicle shedding, exosomes, secretory autophagy and lysosomes have been proposed to release IL-1β, and evidence was brought forward both for and against each model. Other IL-1 family members also lack a signal sequence, but are generally regarded to be alarmins. Recent studies suggest that cell lysis is the main release mechanism, at least in macrophages, as Gsdmd−/− cells process IL-1β normally but fail to release it owing to an absence of pyroptosis4,84. Thus, IL-1β and IL-18 might be unique alarmins that require caspase-mediated cleavage to become biologically active. Nevertheless, other cell types might release IL-1β differently, and neutrophils, for example, release IL-1β without undergoing pyroptosis90.

Modulation of inflammasome signalling

Research over the past decade has not only uncovered the essential components of inflammasomes, but has also highlighted many regulatory mechanisms, including transcriptional and post-transcriptional control, post-translational modifications, and a number of proteins that regulate inflammasomes at the level of the receptors, ASC or the caspases. Insight into inflammasome regulation has also come from studies on bacterial or viral inflammasome inhibitors (reviewed in Ref. 2). We discuss below the endogenous regulatory mechanisms that act directly at the level of ligand recognition or complex assembly (Fig. 5; Table 1).

Figure 5: Regulation of inflammasome complex assembly and signalling.
figure 5

a | NLRC4 activation requires phosphorylation within the helical domain 2 (HD2) of the nucleotide-binding oligomerization domain (NOD) module (consisting of the NBD, HD1, WHD and HD2) and is mediated by protein kinase Cδ (PKCδ)103. b | NLRP3 priming involves pattern-recognition receptor (PRR)-induced upregulation of NLRP3 expression and NLRP3 deubiquitylation by BRCC3 (Ref. 106). NIMA-related kinase 7 (NEK7), an essential component of the NLRP3 complex, is necessary for mitotic progression and thus only available for NLRP3 binding during interphase14,15,16. Interferons (IFNs) negatively regulate NLRP3 through the induction of inducible nitric oxide synthase (iNOS) and nitric oxide (NO)-mediated nitrosylation116. Orphan nuclear receptor small heterodimer partner (SHP) can bind NLRP3 and negatively regulate inflammasome signalling, but the mechanism is unknown130. Pyrin domain (PYD)-only proteins (POPs) regulate NLRP3 inflammasome assembly by sequestering ASC (POP1) or by binding to the PYD of NLRP3 (POP2)98,99. c | IFN-activable protein 202 (p202) inhibits absent in melanoma 2 (AIM2) signalling by binding to DNA, which results in the spatial separation of the PYDs of AIM2 on the DNA, preventing ASC clustering and signal propagation28,102. POP3, an IFNγ-induced protein, directly binds to the PYD of AIM2 and blocks the recruitment of ASC101. d | IκB kinase-α (IKKα) retains ASC in the nucleus, thus controlling its availability113. POP1 binds to the PYD of ASC, thus preventing its recruitment to receptor PYDs and ASC polymerization98. Caspase 1 activation requires the phosphorylation of the caspase recruitment domain (CARD) of ASC by spleen tyrosine kinase (SYK) or JUN N-terminal kinase (JNK)110. Pro-caspase 1 availability, and thus indirectly its activity, is controlled by CARD16–18 and caspase 12, which bind and sequester pro-caspase 1 (Refs 91,92,93,94,96). NAIP, NLR family, apoptosis inhibitory protein; S. Typhimurium, Salmonella enterica subsp. enterica serovar Typhimurium.

PowerPoint slide

Table 1 Endogenous regulators of inflammasomes

CARD- or PYD-containing regulators. Inflammasome assembly largely relies on death-fold domain-mediated interaction of the receptor with ASC, and on interaction of ASC with caspase 1. Modulating these interactions offers a possibility to control inflammasome signalling, mainly through so-called CARD-only proteins (COPs) and PYD-only proteins (POPs) that are found in humans and higher primates, but that are absent from mouse and rat genomes.

Three COPs — CARD16 (also known as COP and PSEUDO-ICE), CARD17 (also known as INCA) and CARD18 (also known as ICEBERG) — are found in the human genome and they are all highly similar to the CARD of caspase 1 (92%, 81% and 52% identity, respectively)91,92,93. COPs were proposed to act as negative regulators of inflammasomes by sequestering caspase 1, but they might also regulate other signalling pathways, as CARD16 and CARD18 also interact with receptor-interacting serine/threonine protein kinase 2 (RIPK2)91,92,93. The role of CARD16 as a negative regulator was challenged by a recent study that showed that it can enhance caspase 1 activation94. Therefore, the role of these proteins might be more complex than previously anticipated and is worth revisiting.

Caspase 12, a mostly uncharacterized member of the inflammatory caspases, was also shown to regulate inflammasome signalling. A polymorphism in the CASP12 gene results in the expression of a full-length (caspase 12L) and a truncated protein in humans. Caspase 12L is confined to populations of African descent and confers hypo-responsiveness to LPS-induced cytokine production in ex vivo whole blood95. Consistent with a role as a negative regulator, mouse caspase 12 interacts with caspase 1, and overexpression of caspase 12 reduces caspase 1 activity. Casp12 deficiency also conferred resistance to sepsis and increased bacterial clearance in vivo96; however, because the Casp12−/− mice used in the study lacked a functional allele of Casp11 (Ref. 97), additional studies with caspase 11-sufficient Casp12−/− animals are necessary to confirm this phenotype.

POPs are fairly well characterized and three of these proteins (POP1, POP2 and POP3) interfere with inflammasome signalling at the level of the PYD–PYD interaction. POP1 shows sequence similarity to the ASCPYD, and its overexpression reduces IL-1β release in response to canonical and non-canonical inflammasome activators98. Because POP1 binds ASC and prevents ASCPYD–NLRP3PYD interactions, it was hypothesized that POP1 blocks ASCPYD nucleation at the receptor. Consistently, POP1 gene knockdown increases NLRP3 inflammasome-mediated caspase 1 activation. Although mice do not have POPs, POP1 can be functional in mouse cells, as overexpression of POP1 reduces inflammasome activation, and POP1 knock-in mice show ameliorated disease outcomes in mouse models of LPS-induced peritonitis and cryopyrin-associated autoinflammatory syndrome98. As POP1 expression is induced by TLR or IL-1R signalling, it was proposed that POP1 provides a regulatory feedback loop that shuts down inflammasome signalling, but how this protein would disassemble an already-formed ASC speck is unclear. POP2, the second member of this protein class, resembles the PYDs of NLRP2 and NLRP3. Overexpression studies show that it can inhibit NLRP3 signalling, but its physiological role and regulation are not understood99,100. Type I IFNs control the expression of POP3, which shows sequence similarity to the PYD of AIM2 and other HIN-200 family members101. It interacts with these proteins, and knock down of Pop3 results in increased AIM2 inflammasome activation. Reduced AIM2-mediated antiviral immunity to murine cytomegalovirus (MCMV) infections was reported for knock-in mice that express POP3 (Ref. 101), which supports the conclusion that POP3 functions as a specific regulator of the DNA-induced response. Interestingly, IFN-activable protein 202 (p202, which is another HIN2 family member) also negatively regulates AIM2 activation, but through binding DNA and interacting with AIM2, which is thought to result in a spatial separation of the PYDs of AIM2, thus preventing ASC clustering28,102.

Post-translational modifications. Protein phosphorylation or ubiquitylation control the activity of inflammasome receptors and the adaptor ASC. For example, the phosphorylation of NLRC4 by protein kinase Cδ (PKCδ) has been shown to be important for NLRC4 activity during Salmonella infections103,104. Surprisingly, PKCδ was found to be dispensable for Shigella flexneri-induced NLRC4 activation in LPS-primed macrophages105. Although other kinases may also phosphorylate the same site of NLRC4, different priming conditions may account for the discrepancy104. Our recent findings have shown that Ser533 phosphorylation is mainly required for NLRC4-dependent caspase 1 activation in unprimed macrophages, and that priming induces NLRP3 expression, which associates with NLRC4 independently of Ser533 phosphorylation to enhance caspase 1 activation104. The critical residue (Ser533) maps to the helical domain 2 (HD2) of the NLRC4 NOD module, but if and how this controls NLRC4 self-oligomerization into wheel-like assemblies remains to be determined. Interestingly, the post-translational modification only occurred in the presence of NLRC4 activators, but whether this modification is linked to PRR-mediated recognition of Salmonella enterica subsp. enterica serovar Typhimurium is currently unknown. A role for priming — that is, the engagement of PRRs or cytokine receptors — has been firmly established in the activation mechanism of NLRP3. Priming was initially thought to result in the transcriptional upregulation of NLRP3 expression, but it was recently shown that it also licenses inflammasome signalling by initiating the deubiquitylation of NLRP3 by the Lys63-specific deubiquitinase BRCC3 (Refs 106,107). This step is essential for NLRP3 activity, involves the production of reactive oxygen species and occurs only after TLR stimulation106,107. Whether NLRP3 needs to be phosphorylated, similarly to NLRC4, remains to be determined, but it is worth noting that the double-stranded RNA-dependent protein kinase (PKR; also known as EIF2AK2) was proposed to regulate the activity of NLRP3, as well as NLRP1, NLRC4 and AIM2, in a manner that requires its kinase activity108. However, a second study did not find a role for PKR in inflammasome signalling109.

The inflammasome adaptor ASC is phosphorylated in response to inflammasome stimuli at several residues that map to its CARD110. These phosphorylation events require the kinases spleen tyrosine kinase (SYK) and JUN N-terminal kinase (JNK) and are necessary for caspase 1 activation110. The kinase SYK is also known to promote NLRP3-dependent caspase 1 activation during infections with C. albicans, but it remains to be shown whether this is linked to its role in ASC phosphorylation111. Notably, the TAK1 (TGFβ-activated kinase 1; also known as MAP3K7)–JNK pathway was found to be activated by lysosomal rupture and to promote NLRP3 activation, which indicates that lysosomal rupture might not be an NLRP3 trigger but is instead required for the priming of the NLRP3–ASC pathway112. Another kinase, IκB kinase-α (IKKα), has been implicated in the negative regulation of ASC by controlling its subcellular localization in resting macrophages113, but it is unclear whether this involves ASC phosphorylation.

Regulation by IFNs. IFNs are an important group of cytokines that activate immune cells and initiate antimicrobial defences. The most well-studied classes are type I and type II IFNs, and both are known to control inflammasome signalling. Type I IFNs were shown to reduce the expression of pro-IL-1β and pro-IL-18, but also to repress NLRP1B and NLRP3 inflammasome activity114. Whereas the repression of NLRP1B activity relies on the induction of the anti-inflammatory cytokine IL-10 and signalling through signal transducer and activator of transcription 3 (STAT3), the molecular mechanism of NLRP3 repression is not fully understood, but it involves STAT1-dependent reduction of caspase 1 processing. This mechanism of inflammasome control is also important in vivo, as type I IFN induction by poly(I:C) pretreatment reduces the recruitment of inflammatory cells after alum injection in mice and also increases susceptibility to C. albicans114. During mycobacterial infections, type I and type II IFNs act together to fine-tune IL-1 production by inflammatory monocyte–macrophages and dendritic cells115. Type I IFNs inhibited IL-1 production by both subsets, whereas CD4+ T cell-derived IFNγ suppressed IL-1 expression selectively in inflammatory monocytes. How IFNγ modulates IL-1 production in this context is unknown, but T cell-derived IFNγ has been shown to inhibit the NLRP3 inflammasome in a mouse model of tuberculosis through inducible nitric oxide synthase, which resulted in the nitrosylation and inactivation of NLRP3 (Ref. 116).

Conversely, IFNs can also enhance inflammasome signalling during microbial infections. AIM2 activation during infections with F. novicida but not MCMV requires STING-dependent production of type I IFNs29,30,31,117. Similarly, efficient activation of caspase 11 during Gram-negative bacterial infections requires the TLR4–TRIF (TIR-domain-containing adaptor protein inducing IFNβ)–IRF3 (IFN-regulatory factor 3) axis53,118. IFN-induced signalling increases the expression of AIM2 and caspase 11, which partially accounts for the IFN requirement. However, inflammasome activation also requires a group of IFN-inducible GTPases, known as guanylate-binding proteins (GBPs), which are upregulated in an IFNAR–IRF1-dependent manner117,119,120. GBPs have so far mainly been shown to function in cell-autonomous immunity and they restrict the replication of intracellular bacterial and protozoan pathogens121. They control antimicrobial processes ranging from oxidative and autophagy-based defences, to the destabilization of pathogen-containing vacuoles and direct killing of the pathogen121,122. Among the 11 mouse GBPs, GBP2 and GBP5 were shown to be required for F. novicida-induced AIM2 activation and GBP2 was required for caspase 11 activation by Gram-negative bacteria, but not for inflammasome activation in response to the purified ligands (such as poly(dA:dT) and LPS)117,119,120. In line with their function in cell-autonomous immunity, we and others117,119,120 have proposed that GBP-mediated destabilization of pathogen-containing vacuoles or direct attack on the pathogen releases PAMPs into the cytosol that are then sensed by AIM2 or caspase 11. However, a direct role for GBPs in inflammasome assembly and signalling has also been described123,124. In this model, GBPs enhance inflammasome signalling by promoting NLRP3–ASC oligomerization or by acting further downstream at the level of pyroptosis induction. Although GBPs do not seem to be essential components of inflammasomes, understanding their role in cell-autonomous immunity and inflammasome signalling is an exciting field of future research.

Conclusions and future perspectives

The past couple of years have brought rapid progress in our understanding of inflammasome biology, going far beyond the basic concept that was introduced in the early 2000s. The molecular basis of ligand recognition has been determined for several receptors, and new paradigms for how the innate immune system senses microbial pathogens have been established through the identification and characterization of the pyrin inflammasome and the non-canonical inflammasome pathway. Furthermore, the molecular mechanisms that drive receptor oligomerization, complex assembly and signal propagation within the complex are emerging, and crucial players that mediate downstream signalling have been identified.

These advances have set the stage for the next era of inflammasome studies by allowing unprecedented understanding of inflammasome biology at a structural and biochemical level. Although this progress has raised many new questions, many old mysteries also remain unsolved. For example, the upstream events that control pyrin and NLRP3 activation are not understood, and a role for most of the human and mouse NLRs has not yet been identified. The advances ushered in by new genomic editing technologies such as CRISPR–Cas9, are already transforming immunology research and the field of inflammasome studies4,16, and will certainly lead to a rapid identification and characterization of such novel regulatory and signalling components.

But finally, and most importantly, how will this newly obtained knowledge be translated into treatments for inflammasome-associated infectious and inflammatory diseases? Recent studies have highlighted the potential of small molecules that block NLRP3 activation in attenuating inflammasome signalling in mouse models of cryopyrin-associated autoinflammatory syndrome and experimental autoimmune encephalomyelitis125,126. Therefore, in-depth understanding of the upstream mechanisms and the structural basis of signal recognition and complex assembly are likely to be instrumental in the identification of new therapeutic targets and could foster the development of new anti-inflammatory therapies for inflammasome-associated diseases based on a range of selective inhibitors of individual inflammasomes.