NATURE MEDICINE | REVIEW Subject terms: Chronic inflammation Immunological disorders Inflammasome Inflammatory diseases

The inflammasomes are innate immune system receptors and sensors that regulate the activation of caspase-1 and induce inflammation in response to infectious microbes and molecules derived from host proteins. They have been implicated in a host of inflammatory disorders. Recent developments have greatly enhanced our understanding of the molecular mechanisms by which different inflammasomes are activated. Additionally, increasing evidence in mouse models, supported by human data, strongly implicates an involvement of the inflammasome in the initiation or progression of diseases with a high impact on public health, such as metabolic disorders and neurodegenerative diseases. Finally, recent developments pointing toward promising therapeutics that target inflammasome activity in inflammatory diseases have been reported. This review will focus on these three areas of inflammasome research. Introduction Inflammation is a protective immune response mounted by the evolutionarily conserved innate immune system in response to harmful stimuli, such as pathogens, dead cells or irritants, and is tightly regulated by the host. Insufficient inflammation can lead to persistent infection of pathogens, while excessive inflammation can cause chronic or systemic inflammatory diseases. Innate immune function depends upon the recognition of pathogen-associated molecular patterns (PAMPs), derived from invading pathogens, and danger-associated molecular patterns (DAMPs), induced as a result of endogenous stress, by germline-encoded pattern-recognition receptors (PRRs). Activation of PRRs by PAMPs or DAMPs triggers downstream signaling cascades and leads to production of type I interferons (interferon-α and interferon-β) and proinflammatory cytokines. Of note, DAMP-triggered inflammation, which is particularly important in inflammatory diseases, is termed sterile inflammation when it occurs in the absence of any foreign pathogens1. Activation of the inflammasome is a key function mediated by the innate immune system, and recent advances have greatly increased our understanding of the macromolecular activation of inflammasomes. Several families of PRRs are important components in the inflammasome complex, including the nucleotide-binding domain, leucine-rich repeat containing proteins (also known as NOD-like receptors, NLRs) and the absent in melanoma 2 (AIM)-like receptors (ALRs) in both mice and humans2. Upon sensing certain stimuli, the relevant NLR or AIM2 can oligomerize to be a caspase-1–activating scaffold. Active caspase-1 subsequently functions to cleave the proinflammatory IL-1 family of cytokines into their bioactive forms, IL-1β and IL-18, and cause pyroptosis, a type of inflammatory cell death3, 4. Inflammasomes have been linked to a variety of autoinflammatory and autoimmune diseases, including neurodegenerative diseases (multiple sclerosis, Alzheimer's disease and Parkinson's disease) and metabolic disorders (atherosclerosis, type 2 diabetes and obesity)4. In the initiation of inflammatory disease, inflammasomes play either causative or contributing roles, and also exaggerate the pathology in response to host-derived factors. This review will focus on the current understanding of inflammasome activation; on the roles of inflammasomes in several prevalent diseases that are increasingly recognized as having an inflammatory contribution, such as neurodegenerative diseases and metabolic disorders; and on advances in potential therapies targeting inflammasomes. Mechanisms of inflammasome activation General principles of inflammasome activation. Recent developments in our understanding of the mechanisms of inflammasome activation have been expertly reviewed in depth4, 5, 6, 7, 8. Here, however, we give a brief overview of recent advances in the mechanisms of inflammasome activation in order to best explain their link with disease. Inflammasomes are multimeric protein complexes that assemble in the cytosol after sensing PAMPs or DAMPs7, 9. Although there are fundamental differences between inflammasomes dependent upon stimuli, in general, canonical inflammasomes serve as a scaffold to recruit the inactive zymogen pro-caspase-1 (Figs. 1 and 2). Oligomerization of pro-caspase-1 proteins induces their autoproteolytic cleavage into active caspase-1 (ref. 10). Active caspase-1 is a cysteine-dependent protease that cleaves the precursor cytokines pro-IL-1β and pro-IL18, generating the biologically active cytokines IL-1β and IL-18, respectively11, 12, 13. Active caspase-1 is also able to induce an inflammatory form of cell death known as pyroptosis5, 6, 7. Figure 1: Mechanisms of NLRP3 inflammasome activation. NLRP3 must be primed before activation. Priming involves two distinct steps. First, an NF-κB–activating stimulus, such as LPS binding to TLR4, induces elevated expression of NLRP3 (as well as IL1B), which leads to increased expression of NLRP3 protein. Additionally, priming immediately licenses NLRP3 by inducing its deubiquitination. The adaptor protein ASC must become linearly ubiquitinated and phosphorylated for inflammasome assembly to occur. After priming, canonical NLRP3 inflammasome activation requires a second, distinct signal to activate NLRP3 and lead to the formation of the NLRP3 inflammasome complex. The most commonly accepted activating stimuli for NLRP3 include relocalization of NLRP3 to the mitochondria, the sensation of mitochondrial factors released into the cytosol (mitochondrial ROS, mitochondrial DNA, or cardiolipin), potassium efflux through ion channels, and cathepsin release following destabilization of lysosomal membranes. Recent studies have determined that activated NLRP3 nucleates ASC into prion-like filaments through PYD-PYD interactions. Pro-caspase-1 filaments subsequently form off of the ASC filaments through CARD-CARD interactions, allowing autoproteolytic activation of pro-caspase-1. Inset shows domain arrangement of the NLRP3 inflammasome components. Pro-caspase-1 and caspase-1 domains are simplified for clarity, the CARD domain is actually removed by cleavage, and two heterodimers form with the p20 and p10 effector domains (p20/10). Figure 2: Mechanisms of NLRC4, AIM2 and noncanonical NLRP3 inflammasome activation. (a) NLRC4 inflammasome agonists such as the bacterial needle protein bind directly to regions within the NACHT domains of the NAIP subfamily of proteins. hNAIP and mNAIP1 bind needle protein, mNAIP2 binds rod protein, and both mNAIP5 and mNAIP6 bind flagellin. Ligand-bound NAIP proteins then oligomerize with NLRC4 to form a caspase-1–activating inflammasome. Though NLRC4 can directly oligomerize with caspase-1 through CARD-CARD interactions, ASC is required for caspase-1 activation by the NLRC4 inflammasome, possibly through the formation of prionlike filaments (blue) by ASC. However, ASC is dispensable for the induction of pyroptosis. Inset shows domain arrangement of NLRC4 inflammasome components. NAIP proteins have three N-terminal BIR domains. hNAIP, human NAIP; mNAIP, mouse NAIP. (b) The mechanism of AIM2 inflammasome activation is well defined. The HIN domain of AIM2 directly binds cytosolic dsDNA, displacing the PYD and relieving autoinhibition. This allows oligomerization of AIM2 PYD with ASC PYD, converting ASC into its prion form. Prion-like filaments of pro-caspase-1 (violet) are then able to form off of the ASC filaments, inducing caspase-1 activation. Inset shows domain arrangement of AIM2 inflammasome components. (c) Studies have determined that mouse pro-caspase-11 (mPro-caspase-11) and human pro-caspases-4 and -5 (hPro-caspase-4/5) can directly bind intracellular LPS and activate a noncanonical NLRP3 inflammasome. This induces oligomerization of these pro-caspases, leading to their proximity-induced activation. This is sufficient for the induction of pyroptosis but not for the processing of pro-IL-1β. However, active mCaspase11 and hCaspase-4 can promote full assembly and activation of the NLRP3 inflammasome following a priming signal. Inflammasome names denote the protein forming the scaffold. Most inflammasomes are formed with one or two NLR family members, and NLRC4 requires interaction with an NLR member of the NAIP subfamily of proteins6, 14 (Figs. 1 and 2a). However, non-NLR proteins such as AIM2 (Fig. 2b) and pyrin can also form inflammasomes. NLRC4 can directly associate with caspase-1 through CARD-CARD interactions15. NLRs containing an N-terminal pyrin domain (PYD) have been shown to associate with apoptosis-associated speck-like protein containing a CARD (ASC) in order to recruit pro-caspase-1 to the inflammasome9, 16 (Fig. 1). Inflammasome activation occurs when the scaffold protein senses or binds its activating stimuli. How this occurs is starting to be clarified for certain inflammasome proteins6; prominent among these are the roles of ASC, AIM2, and NAIP and NLRC4. For example, AIM2 can directly bind its stimulus, double-stranded DNA (dsDNA)17. However, many questions remain regarding inflammasome activation. We will now briefly discuss the mechanism of activation of the best-characterized inflammasomes, an area in which major advances have been made. The readers can refer to recent reviews in which all of the NLR inflammasomes have been discussed5, 6, 7, including evidence supporting the existence of less-characterized inflammasomes such as NLRP6, NLRP7, NLRP12 and IFI16 inflammasomes. It should also be noted that although NLRP1, which has many genetic variants in mice and rats, forms well-defined inflammasomes in these rodent models, the activation of the single human NLRP1 paralog into an inflammasome is less well understood18. NLRP3 inflammasome. The NLRP3 inflammasome (Fig. 1) is activated in response to the widest array of stimuli, leading to the theory that the dissimilar agonists induce similar downstream events that are sensed by NLRP3 (refs. 8,19,20). The mechanisms of NLRP3 activation supported by the most studies include potassium efflux out of the cell, the generation of mitochondrial reactive oxygen species (ROS), the translocation of NLRP3 to the mitochondria, the release of mitochondrial DNA or cardiolipin, and the release of cathepsins into the cytosol after lysosomal destabilization6, 7, 8 (Fig. 1). However, not all of these events are induced by all NLRP3 agonists, so the precise mechanism of NLRP3 activation is still debated. Additionally, increases in intracellular calcium can activate the NLRP3 inflammasome21, 22, but this is also not a requirement of all NLRP3 agonists23. Though many published studies support the involvement of lysosomal cathepsins, proteases that degrade internalized proteins, in NLRP3 inflammasome activation, it is important to note that this is not without some controversy24. In most cell types, NLRP3 must be primed, and a prototypical example of such a priming event is the binding of lipopolysaccharide (LPS) to TLR4. Priming has long been known to increase cellular expression of NLRP3 through NF-κB signaling25. However, recent findings have shown that priming rapidly licenses mouse NLRP3 inflammasome activation by inducing the deubiquitination of NLRP3 independent of new protein synthesis, whereas inhibition of deubiquitination inhibits human NLRP3 activation26, 27. Once primed, NLRP3 can respond to its stimuli and assemble the NLRP3 inflammasome. Additionally, ASC must be linearly ubiquitinated for NLRP3 inflammasome assembly28. Current stimuli recognized as NLRP3 agonists that induce NLRP3 inflammasome formation include ATP, pore-forming toxins, crystalline substances, nucleic acids, hyaluronan, and fungal, bacterial or viral pathogens6, 7. These stimuli can be encountered during infection, either produced by pathogens or released by damaged host cells. Additionally, pathologic conditions in the body may promote the formation of these stimuli in the absence of infection; an example is the formation of inflammatory cholesterol crystals, as discussed in more detail later. Recent studies showed that the NLRP3 NBD oligomerizes the NLRP3 PYD, which serves as a scaffold to nucleate ASC proteins through PYD-PYD interactions29, 30. This causes ASC to convert to a prion-like form and generate long ASC filaments that are crucial to inflammasome activation. Pro-caspase-1 then interacts with ASC through CARD-CARD interactions and forms its own prion-like filaments that branch off of the ASC filaments. The close proximity of pro-caspase-1 proteins then induces autoproteolytic maturation of pro-caspase-1 into active caspase-1. Additionally, increasing evidence has identified a crucial role for caspase-8 in inflammasome activation and pro-IL-1β processing. Caspase-8 is a proapoptotic protease that initiates the external apoptosis pathway in response to external stimuli, such as FasL and TNF, and protects against an inflammatory form of cell death termed necroptosis31. It is now also recognized that caspase-8 is required for both the transcriptional priming and activation of the canonical and noncanonical NLRP3 inflammasomes in mice in response to pathogenic stimuli and ligands stimulating various different TLRs32, 33, 34. Thus, inflammatory diseases in which TLR ligands are generated could lead to caspase-8–mediated NLRP3 priming or activation. Additionally, caspase-8 was shown to bind and localize to ASC specks, further suggesting that caspase-8 is an important component of inflammasome complexes35. However, the exact molecular mechanism by which caspase-8 promotes caspase-1 activation has yet to be elucidated. Importantly, caspase-8 also has an identified role in NLRC4 and AIM2 inflammasome activation35, 36 and has even been shown to directly promote pro-IL-1β processing in a noncanonical caspase-8 inflammasome induced by the binding of certain extracellular pathogens to dectin-1 (ref. 37). Notably, the exact role of caspase-8–mediated inflammasome activation is somewhat controversial38. NLRC4 inflammasome. In contrast to the diverse stimuli that activate NLRP3 inflammasomes, the NLRC4 inflammasome responds to a more limited set of stimuli. A major advance in our understanding of the NLRC4 inflammasome is the recognition that NLRC4 forms a complex with various NAIP proteins, and NLRC4-activating ligands are bound by these NAIP components rather than by NLRC4 (Fig. 2a). This raises the question of whether NLRC4 is a scaffolding protein and not a receptor14, 39. In mice, NAIP1 binds the bacterial type III secretory system (T3SS) needle protein40, 41, NAIP2 binds the bacterial T3SS rod protein42 and both NAIP5 and NAIP6 bind bacterial flagellin42, 43. T3SS is found in several Gram-negative bacteria and allows the bacteria to inject effector molecules into infected host cells. By contrast to mice, in humans only one NAIP protein has been characterized, and it was found to bind only the T3SS needle protein40, suggesting a far more restrictive repertoire of ligands for the NLRC4 inflammasome in human cells than for the NLRP3 inflammasome, which responds to a plethora of stimuli. Once NAIP proteins bind their ligands, they can oligomerize with NLRC4 and form a NAIP/NLRC4 inflammasome14. In order for NLRC4 to be activated, its autoinhibition must be relieved to allow oligomerization with NAIP proteins, but how this occurs is unclear14. However, two new gain-of-function mutations of NLRC4 have recently been identified in humans that cause severe spontaneous autoimmune syndrome, suggesting that the helical domain is responsible for this autoinhibition44, 45. Though some reports indicate that mouse NLRC4 must be phosphorylated before inflammasome activation46, 47, there are also conflicting reports indicating that phosphorylation is dispensable14. Though NLRC4 contains a CARD domain, ASC is required for maximal inflammasome activation7 (Fig. 2a). A possible explanation might be the formation of ASC filaments off of NLRC4, as there is evidence that the CARD domain can convert ASC to its prion-like form31. AIM2 inflammasome. The non-NLR AIM2 can also form a caspase-1–containing inflammasome, but, unlike the NLRs, the HIN-200 domain of AIM2 can directly bind its stimulus, cytosolic dsDNA, which may be encountered in the cytosol during pathogenic infection (Fig. 2b)17. The autoinhibitory conformation of AIM2 is created by interactions of its two domains and relieved by the sugar phosphate backbone of dsDNA48. DNA binding displaces the PYD domain48, freeing the PYD domain to recruit ASC to the complex17, 49. AIM2 cannot interact with ASC unless autoinhibition is relieved50, and thus AIM2 maintains itself in an inactive state until its ligand binds. Interestingly, AIM2 does not appear to recognize a specific sequence or structure of dsDNA but instead requires a dsDNA strand of at least 80 base pairs for optimal inflammasome activation48. Similar to NLRP3, oligomerized AIM2 nucleates ASC through PYD-PYD interactions and converts ASC to its prion form, leading to the development of long PYD-PYD ASC filaments29, 30. Recently, a noncanonical AIM2 inflammasome was shown to mediate protection against Francisella novicida 51. F. novicida infection is detected by cGAS and STING, inducing the expression of the transcription factor IRF1. IRF1 increases the expression of guanylate-binding proteins, which increase the intracellular killing of the bacterium. This releases dsDNA into the cytosol and induces AIM2 inflammasome activation. Noncanonical inflammasomes A developing area of interest in the inflammasome field is the noncanonical inflammasome formed by caspase-11 in mice (Fig. 2c). Caspase-11 was initially found to be important for the activation of caspase-1 and caspase-3 (ref. 52). Recently, it was shown to promote NLRP3 inflammasome activation to indirectly enhance processing of pro-IL-1β or pro-IL-18 (ref. 53). More remarkably, caspase-11 detects intracellular LPS and some intracellular bacteria, directly mediating cell death and IL-1α secretion, but not IL-1β secretion, in a mechanism independent of the traditional LPS receptor TLR4 (refs. 7,54,55). Though humans do not express caspase-11, recent studies indicate that caspase-4 and caspase-5 in human cells serve a similar function56, 57 (Fig. 2c). Notably, active caspase-4 can promote the activation of the primed NLRP3 inflammasome without a need for a canonical NLRP3 activating stimulus57. As caspase-11–deficient mice are known to be protected from endotoxic shock53, further study of the noncanonical inflammasome in human cells is of great interest. Mechanisms of inflammasome spreading. ASC has long been recognized to redistribute upon inflammasome activation from the nucleus to the cytosol and form a large perinuclear aggregate in cells58, 59. In a recent breakthrough, ASC specks were reported to be released by dying cells, leading to cleavage of extracellular pro-IL-1β and activating caspase-1 in macrophages internalizing the specks60. Importantly, as activation of all major inflammasomes is associated with speck formation59, this suggests that inflammasome activation propagates inflammation from cell to cell. The buildup of specks at sites of inflammation has serious implications for inflammatory diseases, as injection of purified ASC specks into mice in vivo has been shown to propagate inflammation 60. Additionally, phosphorylation of ASC was recently shown to be a key checkpoint in ASC speck formation. The kinases Syk and JNK, which activate in response to a vast array of stimuli and lead to the phosphorylation of many downstream targets, mediate phosphorylation of ASC upon NLRP3 inflammasome activation, and inhibition of these kinases prevented ASC speck formation and blocked caspase-1 activation61. Importantly, phosphorylation was dispensable for NLRP3 and ASC oligomerization. This suggests that phosphorylation of ASC may be necessary for ASC to switch to its prion-like form and form self-propagating filaments. This also suggests that kinase inhibition may have potential therapeutic use against inflammatory diseases in the absence of more targeted inhibitors. Inflammasomes in disease Here we focus on neurologic disorders and metabolic diseases, neither of which are traditionally considered to be inflammatory diseases but which are increasingly recognized as having an inflammatory component that contributes significantly to the disease process. Misfolded protein aggregates and aberrant accumulation of certain metabolites accompanying those diseases are endogenous DAMPs that have been proved to be direct activators of the NLRP3 inflammasome, which plays a critical role in the initiation and progress of those diseases. The inflammasome and multiple sclerosis. Multiple sclerosis (MS), one of the most common autoimmune inflammatory diseases, is characterized by myelin-reactive CD4+ T cells that infiltrate the central nervous system (CNS), attack oligodendrocytes and induce demyelination62. Demyelination partially disrupts the communication of the nervous system, resulting in physical, mental and psychiatric challenges, among other problems. Presently, MS has no cure and shortens the lifespan of affected individuals by approximately 5–10 years63. Experimental autoimmune encephalomyelitis (EAE) is an animal model commonly used to mimic MS. To induce EAE, mice are immunized with the peptide myelin oligodendrocyte glycoprotein (MOG) emulsified in adjuvant, inducing infiltration of MOG-specific T cells and other inflammatory cells into the CNS64. Prior to the discovery of NLRs, the inflammasome products caspase-1, IL-1β and IL-18 had been shown to contribute to EAE progression. Casp1−/−, Il1a−/−, Il1b−/− and Il18−/− mice are resistant to EAE and show concomitant reductions in interferon (IFN)-γ and/or IL-17 levels65, 66, 67. It has recently been shown that Nlrp3 expression increases in the spinal cord during EAE progression and that Nlrp3-deficient mice have a dramatically delayed course and reduced severity of disease, accompanied by fewer infiltrating inflammatory cells and reduced astrogliosis 64, 68. In addition, a study using a cuprizone model of MS also showed that Nlrp3deficient mice had delayed demyelination and oligodendrocyte loss 69. Additionally, EAE mice show increased IL-18 levels compared with controls, and Il18-deficient mice phenocopy the reduced disease seen in Nlrp3-deficient mice, suggesting that NLRP3 functions through IL18 to promote EAE 64, 68. Despite these findings, the role of NLRP3 in EAE progression is complicated. Expression of Nlrp3 in antigen-presenting cells (APCs) was required to stimulate T helper type 1 (T 1) and T 17 cells to respond to brain autoantigen in one study64. Additionally, Nlrp3 and Asc (also known as Pycard) deficiency caused reduced expression of many chemokines and chemokine receptors, such as Ccr2 and Ccr6, in both APCs and T cells, reducing migration of T 1 and T 17 cells into the CNS of Nlrp3and Asc-deficient mice following EAE induction by MOG peptide immunization. However, direct delivery of CD4 + T cells from EAE-induced WT, Nlrp3−/− or Asc−/− mice into the brain and spinal cord of recipient Rag2−/− mice, which lack mature T cells, induced the same degree of disease68. In summary, although these results suggest that the NLRP3 inflammasome contributes to both T 1 and T 17 cell responses and migration during EAE, the function of the NLRP3 inflammasome is not an inherent function of T cells. In the clinic, peripheral blood mononuclear cells (PBMCs) from patients with relapsing-remitting MS had higher levels of NLRP3, IL-1β and caspase-1 than were found in PBMCs from healthy controls. Intriguingly, soluble factors secreted by human PBMCs upon NLRP3 activation skew the cytokine profile of CD4+ T cells toward a proinflammatory T 17 phenotype, supporting a link between MS and the NLRP3 inflammasome70. However, a role for NLRP3 and ASC in EAE is not found in all studies and varies with variations in the disease model. Aggressive immunization of mice with heat-killed mycobacteria (Mtb) was able to induce EAE even in the absence of NLRP3 or ASC, whereas lowerdose Mtb immunization required NLRP3 and ASC for EAE induction71. Another study found no difference in MOG-induced EAE disease between WT and Nlrp3-deficient mice. In the same study, ASC promoted EAE progression in an inflammasome-independent manner through a mechanism that involved maintaining CD4 + T cell survival. In agreement with this, Asc-deficient mice were even more resistant to EAE than were Casp1-deficient mice 72. Part of the difference in inflammasome dependency may be explained by recent findings showing that IFN-β inhibits IL-1β production by macrophages, and only NLRP3-dependent EAE is ameliorated by IFN-β treatment. This suggests that IFN-β may therapeutically inhibit the NLRP3 inflammasome–IL-1β–IL-18 axis in MS71. Though IFN-β has been used therapeutically for more than 15 years, one-third of MS patients fail to respond to IFN-β, reflecting heterogeneity in the disease. In addition to the NLRP3 inflammasome, a recent study using the pertussis toxin (PTX)-induced EAE model showed that TLR4 was required H H