Structure, Activation and Regulation of NLRP3 and AIM2 Inflammasomes

The inflammasome is a three-component (sensor, adaptor, and effector) filamentous signaling platform that shields from multiple pathogenic infections by stimulating the proteolytical maturation of proinflammatory cytokines and pyroptotic cell death. The signaling process initiates with the detection of endogenous and/or external danger signals by specific sensors, followed by the nucleation and polymerization from sensor to downstream adaptor and then to the effector, caspase-1. Aberrant activation of inflammasomes promotes autoinflammatory diseases, cancer, neurodegeneration, and cardiometabolic disorders. Therefore, an equitable level of regulation is required to maintain the equilibrium between inflammasome activation and inhibition. Recent advancement in the structural and mechanistic understanding of inflammasome assembly potentiates the emergence of novel therapeutics against inflammasome-regulated diseases. In this review, we have comprehensively discussed the recent and updated insights into the structure of inflammasome components, their activation, interaction, mechanism of regulation, and finally, the formation of densely packed filamentous inflammasome complex that exists as micron-sized punctum in the cells and mediates the immune responses.


Introduction
Pathogen-associated molecular patterns (PAMPs) present in invading microbes and danger-associated molecular patterns (DAMPs) resulting from cellular insults are recognized by pathogen recognition receptors (PRRs). This recognition process results in the activation of a cytosolic supramolecular protein complex known as the inflammasome [1,2], which acts as a signaling platform and initiates an inflammatory response by triggering the production of proinflammatory cytokines (interleukin-1β (IL-1β) and interleukin-18 (IL-18)) [3]. Inflammasomes are divided into two categories, i.e., canonical inflammasomes, which activate caspase-1, and noncanonical inflammasomes that trigger activation of caspase-11 and caspase-4/5 in mouse and human, respectively [4][5][6]. The key components of canonical inflammasomes involve three classes of molecules, i.e., sensor, adaptor, and effector. These components tend to assemble via homotypic interactions between Death Domains (e.g., CARD-CARD (Caspase-activation and recruitment domain) and PYD-PYD (Pyrin domain)). In the presence of external stimuli or specific ligands, sensor proteins (ALRs (AIM2-like receptors) and NLRs (NOD-like receptors)) activate, oligomerize and nucleate the adaptor protein ASC (Apoptosis-associated speck-like protein containing a caspase-activation and recruitment domain (CARD)) through PYD-PYD interactions. ASC in turn recruits the effector protein procaspase-1 mediated by CARD-CARD homotypic recognition, thus serving as an activation platform of caspase-1.

General Information on NLRP3
NLRP3 is expressed in myeloid cells, muscle cells, neurons, and endocrine cells [98]. In the resting state, it exists as an autoinhibited form that becomes activated upon stimulation and assembles into a large, micrometer-size cytosolic complex. In macrophages, NLRP3 becomes functional in a two-step process that includes priming and activation. In the priming process, PRRs such as Toll-like receptors (TLRs), or NODs and cytokines such 2. NLRP3

General Information on NLRP3
NLRP3 is expressed in myeloid cells, muscle cells, neurons, and endocrine cells [98]. In the resting state, it exists as an autoinhibited form that becomes activated upon stimulation and assembles into a large, micrometer-size cytosolic complex. In macrophages, NLRP3 becomes functional in a two-step process that includes priming and activation. In the priming process, PRRs such as Toll-like receptors (TLRs), or NODs and cytokines such as TNF-α trigger the activation of the transcription factor NF-κB, thus affecting the expression of the inflammasome components NLRP3, caspase-1, and pro-IL-1β [99,100].

Role of the NLRP3 Inflammasome in COVID-19
COVID-19 (Coronavirus disease 2019), caused by the SARS-Cov-2 (Severe acute respiratory syndrome coronavirus 2) virus, was declared a pandemic by the WHO (World health organization), and as of 16 December 2020, approx. 73.5 million cases with 1.64 million deaths have been reported worldwide [104][105][106]. The pandemic has been associated with severe social and economic consequences. An upsurge of IL-1β, IL-18, and LDH (Lactate dehydrogenase) has been reported in sera of COVID-19 patients, which hints the involvement of the inflammasome network [107][108][109]. Similarly, a recent study conducted on COVID-19 patients reveals the activation of the NLRP3 inflammasome [110]. In this study, microscopic analysis in combination with luminescent assays show the formation of NLRP3 and ASC puncta, caspase-1 activation, and IL-1β secretion in PBMCs (Peripheral blood mononuclear cells) of COVID-19 patients during the disease and in postmortem lung tissues [110]. In addition, the level of Casp1p20 and IL-18 in COVID-19 patients has been shown as an important marker for determining disease severity [110].

Structural Details of NLRP3
NLRP3 is composed of N-terminal PYD, central NACHT domain, and C-terminal LRR domain ( Figure 1A). The structure of the PYD domain of human NLRP3 (NLRP3 PYD ) was determined by X-ray crystallography and solution-state NMR ( Figure 1B,C) [111,112]. Both techniques reveal an overall similar architecture of the PYD domain with six helices (α1-α6) and five connecting loops, which is also analogous to the six-helix motif observed for the PYDs of NLRP1, NLRP4, NLRP7, NLRP10, and NLRP12 ( Figure 1B) [111][112][113][114][115][116][117]. Among these, NLRP4 and NLRP10 show higher structural similarity with NLRP3 as compared to the other PYDs. However, pairwise structural alignments show that the orientation and length of the helices slightly differ in the NALP3 PYD structure. The six helices adopt a canonical anti-parallel helical-bundle fold tightly packed by a central hydrophobic core made up of helices α1 (L10, A11, Y13, L14), α2 (F25, L29), α4 (L54, A55, M58), α5 (I74, F75), and α6 (A87). In addition, NLRP3 PYD accommodates a second hydrophobic surface formed by F32, I39-P42, L57, and F61 residues [111], which stabilizes the shorter α3 helix by anchoring it to helix α2. Analysis of NLRP3 PYD electrostatic surface shows a charge distribution that might be involved in interactions with ASC PYD or with other members of the Death Domain (DD) superfamily. In addition, structural homology and structure-based sequence alignment establish the presence of a conserved surface exposing a hydrophobic cluster (I39 and P40-P42 at α2-α3 loop and L57 and F61 at α4 helix) that could be responsible for inflammasome assembly and thus, caspase-1 activation [111]. The three-dimensional structures of NLRP3 PYD and ASC PYD , as well as studies on their interactions, reveal that both electrostatic and hydrophobic interactions play important roles. In addition, there are several surfaces in the PYDs of the two proteins available for the interaction, which potentially enhance oligomerization. Conserved residues in NLRP3, C8 (α1 helix), and C108 (loop connecting PYD and NACHT) form a unique disulfide bond that might be involved in ROS (Reactive oxygen species) signaling and NLRP3 inflammasome activation [111]. Based on X-ray crystallographic studies, it has been speculated that the formation of the disulfide bond relieves the autoinhibitory state of NLRP3 upon activation by ROS. NLRP3 PYD structures obtained from X-ray crystallography and solution-state NMR exhibit a high degree of resemblance (RMSD 1.66 Å); however, helices α2, α3, and α4 show slight differences between the crystal and solution structures ( Figure 1C). Residue level calculation of B-factors obtained from X-ray crystallography suggests that NLRP3 PYD is highly compact in agreement with the NMR studies [112]. In particular, helices α2, α3, α4, and α5, and the two hydrophobic cores of the NLRP3 PYD are more rigid as reflected by the lower B-factor values, whereas the α2-α3 loop and C-terminus are more flexible. REVIEW 6 α4 show slight differences between the crystal and solution structures ( Figure 1C). due level calculation of B-factors obtained from X-ray crystallography suggests NLRP3 PYD is highly compact in agreement with the NMR studies [112]. In particular ices α2, α3, α4, and α5, and the two hydrophobic cores of the NLRP3 PYD are more rig reflected by the lower B-factor values, whereas the α2-α3 loop and C-terminus are m flexible.  Figure 7A of [112]; (C) superposition of the crystallographic dimeric structure of NLRP3 PYD (green, 3QF2 [111]) onto the monomeric NMR structure (blue); the dimeric interface is shown in the right panel. These superposed images are adapted from Figure 7B of [112].

Oligomerization of NLRP3 PYD
The crystal structure of NLRP3 PYD reveals the presence of a symmetric dimer, whereas NMR, MALS (Multiangle light scattering), and SEC (Size-exclusion chromatography) show the coexistence of monomeric with higher-order oligomeric forms of NLRP3 PYD in solution. The existence of monomer or dimer depends on the concentrations of NLRP3 PYD and salt, as well as pH [111,112]. The crystal structure of the NLRP3 PYD ( Figure 1C) dimer shows that residues E28 (α2), D29 (α2), and R41 (α3) in one protomer interact with the equivalent residues on the other protomer, which might result in repulsive interactions in the interface, thus suggesting that dimer structure is possibly an artifact from crystal packing [112]. The physiological relevance of the PYD dimer is still unclear. Some reports suggest that a PYD trimer, but not a dimer, can activate the inflammasome [119]; whereas others have found PYD dimer formation via a disulfide bond (e.g., NLRP12 PYD ) [120].
Sedimentation equilibrium and velocity results from analytical ultracentrifugation experiments and NMR studies indicate that NLRP3 PYD forms concentration-dependent oligomers [112]. At low protein concentration and acidic pH (0.2 mM, 303 K), NLRP3 PYD exists as a monomer that self-associates into higher-order oligomers at higher concentrations (0.6 mM, 303 K). The monomer-oligomer transition is drastically influenced by the presence of salt and temperature: Both the presence of salt and temperature increase favor oligomer formation even at low protein concentration (0.2 mM, 303 K, 100 mM NaCl). These results led to the conclusion that hydrophobic interactions play an important role in NLRP3 PYD self-association.

Post-Translational Modifications
Mutational studies uncovered that the LRR domain is not autoinhibitory because an LRR deleted variant demonstrates proinflammatory function similar to full-length (FL) NLRP3 in response to external stimuli [121]. These findings propose that the LRR domain is neither imperative for the assembly of active NLRP3 inflammasome nor crucial for the stabilization of the NLRP3 inactive state. It has been suggested that the importance of the LRR domain in NLRP3 activation could be related to: (1) Regulation by post-translational modifications due to the presence of ubiquitination and phosphorylation sites (K689 and Y861 in the human orthologue, and K687 and Y859 in the mouse orthologue) [121][122][123][124]; (2) the binding to NEK7, which is crucial for NLRP3 inflammasome activation [118]; and (3) the induction of T H 2 cell (T helper type 2) responses [125]. T H 2 cells facilitate adaptive immune responses against microorganisms and allergens by producing several interleukins (IL-4, IL-5, IL-9, IL-10, IL-13, and IL-25 (IL-17E)). The release of cytokines is further associated with antibody secretion, eosinophil/basophil activation, and various antiinflammatory responses in order to provide phagocyte-independent protective responses. It is noteworthy that NLRP3 is also involved in the regulation of gene expression in T H 2 cells in an inflammasome-independent manner. The transcription factor function of NLRP3 has been shown to be key for T H 2 cell polarization. NLRP3 binds to DNA in T H 2 cells at the Il4 promoter region and interacts with transcription factor IRF4 (Interferon regulatory factor) to regulate IL-4 synthesis. These results suggest that subcellular localization regulates NLRP3 functions by facilitating inflammasome assembly when expressed in the cytoplasm and promoting its transcriptional functions when localized in the nucleus.
Sequential phosphorylation-dephosphorylation events are also prerequisites for NLRP3 inflammasome activation. Several post-translational modification sites have been identified in human NLRP3 such as S5 (S3 in mouse), S198 (S194 in mouse), S295 (S291 in mouse), Y861 (Y859 in mouse), etc. [122,124,126,127]. Dephosphorylation by protein phosphatase 2 A (PP2A) at S5 located in the N-terminus PYD regulates the interaction of NLRP3 and ASC [122]. It was found that a phosphomimetic mutant of S5 that creates a negative charge completely attenuates NLRP3 activation by disturbing the interaction between NLRP3 and ASC [122]. Similar results were obtained from the phosphorylation of S295 by protein kinase A (PKA) and protein kinase D (PKD) [127,128]. In addition, dephosphorylation of Y861 by PTPN22 allows NLRP3 activation and subsequent IL-1β secretion [124]. The absence of PTPN22 in cells results in increased NLRP3 phosphorylation, which abolishes inflammasome assembly and reduces IL-1β secretion [124]. Another kinase, Jun N-terminal kinase1 (JNK1), has been shown to phosphorylate S198, providing a critical priming signal for NLRP3 self-association [126]. The phosphorylation/dephosphorylation interplay has been associated with the cryopyrin-associated periodic syndrome (CAPS), as it has been shown to be coupled to impaired phosphorylation at the S194 site in mouse [126]. Hence, the inhibition of NLRP3 phosphorylation/dephosphorylation processes is a potential pharmaceutical target for the treatment of NLRP3-associated diseases [126].

NEK7 Mediated Activation of NLRP3
NIMA (Never In Mitosis gene A)-related kinase 7 or mitotic Ser/Thr kinase NEK7 regulates NLRP3 activation [118]. Cryo-EM structural analysis of the complex formed between NLRP3 without the PYD, the NEK7 C-lobe, and ADP bound to the NBD of NLRP3, reveals that the C-terminal lobe of NEK7 interacts with the NBD, HD2, and LRR regions of NLRP3 (Figure 2A). In addition, the dissociation constant of the complex between NLRP3 and NEK7 was determined to be, K d 78.9 ± 38.5 nM. The binding between NLRP3 and NEK7 involves two interfaces: The first and second half of the NEK7 C-lobe interact with the LRR and NACHT domains (NBD and HD2), respectively. Residues Q129, R131, and R136 of NEK7 interact with the LRR domain, whereas residues D261, E265, and E266 interact with HD2, and D290, K293, and R294 interact with the NBD. Mutagenesis studies showed that both interfaces are required for NLRPE3-NEK7 complex formation. The substitution of G755 in LRR for the amino acids A or R leads to enhanced interaction between LRR and NEK7 [136]. In contrast, phosphorylation of Y859, located in the LRR domain, causes steric hindrance and charge repulsion, thus aborting the interaction with NEK7 [124].
A computational model of the oligomeric assembly of the NLRP3-NEK7 complex was generated to understand the NLRP3 activation mechanism using the structure of a full-length NLRC4 oligomer as a template [118]. It has been reported that the inactive form of NLRC4 undergoes a 90 • rotation of the NBD-HD1 module with respect to the WHD-HD2-LRR module, which generates an active NLRC4 conformation ( Figure 2B) [137,138]. The conformation of the NBD-HD1-WHD module of NLRP3 is similar to that of the inactive form of NLRC4; therefore, the active oligomeric structure of NLRP3 has been modeled from its inactive form by using the NLRC4 activation mechanism ( Figure 2C) [118].
LRR and NEK7 [136]. In contrast, phosphorylation of Y859, located in the LRR domain, causes steric hindrance and charge repulsion, thus aborting the interaction with NEK7 [124].
A computational model of the oligomeric assembly of the NLRP3-NEK7 complex was generated to understand the NLRP3 activation mechanism using the structure of a full-length NLRC4 oligomer as a template [118]. It has been reported that the inactive form of NLRC4 undergoes a 90° rotation of the NBD-HD1 module with respect to the WHD-HD2-LRR module, which generates an active NLRC4 conformation ( Figure 2B) [137,138]. The conformation of the NBD-HD1-WHD module of NLRP3 is similar to that of the inactive form of NLRC4; therefore, the active oligomeric structure of NLRP3 has been modeled from its inactive form by using the NLRC4 activation mechanism ( Figure 2C) [118].  [118]. The ribbon diagram is generated with PyMOL molecular graphics software; (B) structural organization and activation of NLRC4. This image is adapted from Figure 2C of [137]; (C) modelling of NLRP3 active conformation and NLRP3 dimer formation in association with NEK7. This image is adapted from Figure 5A,B of [118].

Role of Caspase-8 in Inflammasome Activation
Caspase-8 plays an important role in the regulation of inflammatory responses by direct cleavage of pro-inflammatory cytokines into their mature forms, and by activating the NLRP3 inflammasome [139,140]. Caspase-8 consists of a death effector domain (DED) at the N-terminus, and p18 and p10 catalytic subunits at the C-terminus. Cell studies showed that in the absence of caspase-1, pro-IL-1β processing is mediated by caspase-8 in response to LPS [141], and to a wide range of stimuli, such as activation of TLR4 in bone marrow-derived dendritic cells (BMDCs) [142] and Fas death receptor. Ligation to the bacterial and fungal C-type lectin receptor, dectin-1, triggers caspase-8 activation via CARD9-Bcl-10-MALT1 complex. Other stimuli include endoplasmic reticulum stress, chemotherapeutic agents, inhibition of c-FLIP (FLICE-like inhibitory protein), and histone  [118]. The ribbon diagram is generated with PyMOL molecular graphics software; (B) structural organization and activation of NLRC4. This image is adapted from Figure 2C of [137]; (C) modelling of NLRP3 active conformation and NLRP3 dimer formation in association with NEK7. This image is adapted from Figure 5A,B of [118].

AIM2
AIM2 is a cytosolic dsDNA sensor that is responsible for downstream signaling to the adaptor protein ASC in response to the presence of bacterial and viral DNA. AIM2 belongs to the PYHIN family (pyrin + HIN) and consists of an N-terminal PYD (1-87) and a C-terminal HIN domain (138-343) connected through a long linker [161,162] (Figure 3A). AIM2 interacts with ASC via PYD-PYD homotypic interaction and the HIN domain binds to dsDNA in a sequence-independent manner. In addition, AIM2 also heterodimerizes with other members of the PYHIN family such as p202, IFI16, and MNDA (Myeloid cell nuclear differentiation antigen) [69,[163][164][165]. AIM2 interaction with ASC further activates procaspase-1 leading to pyroptosis, whereas the interaction with p202 inhibits AIM2mediated inflammatory responses [68,[166][167][168].

AIM2 PYD Domain Structure
The crystal structure of AIM2 PYD reveals the six-helix bundle conformation characte istic of the DD superfamily [161] (Figure 3B). It shares structural homology with PYDs NLRP3 and ASC with RMSD values of 1.8 Å and 1.6 Å, respectively [161]. Howev AIM2 PYD exhibits a short and highly dynamic α3 helix and long α6 helix with the mo variable sequence among the known PYDs. Largely buried and highly conserved K26 re idue located at α2 helix buttresses α3 helix via hydrogen bonding with L40 and A43 of t interconnecting loop, thereby stabilizing the α3 helix. The overall surface of AIM2 PYD  [162]. The ribbon diagrams are generated with PyMOL molecular graphics software.

AIM2 PYD Domain Structure
The crystal structure of AIM2 PYD reveals the six-helix bundle conformation characteristic of the DD superfamily [161] (Figure 3B). It shares structural homology with PYDs of NLRP3 and ASC with RMSD values of 1.8 Å and 1.6 Å, respectively [161]. However, AIM2 PYD exhibits a short and highly dynamic α3 helix and long α6 helix with the most variable sequence among the known PYDs. Largely buried and highly conserved K26 residue located at α2 helix buttresses α3 helix via hydrogen bonding with L40 and A43 of the interconnecting loop, thereby stabilizing the α3 helix. The overall surface of AIM2 PYD is populated by charged residues, thus resulting in distinct electrostatic charge distribution. These amino acids include acidic residues such as E7 and D15 in α1 helix; D19, E20, E21, and D23 in α2 helix; and basic residues such as K64, R67, and K71 in α5 helix; and K79, R80, K85, K87, K90, K93, and K97 in α6 helix. Among these, K64 and K85 are conserved in the PYHIN family, and E20 and E21 are specific to AIM2. Residues such as F27 and F28 create a solvent-exposed hydrophobic patch that shares similarities with the DED surface involved in homotypic interactions, suggesting that they may contribute to AIM2 PYD self-association and AIM2-specific functions [161].
AIM2 PYD tends to form large, insoluble oligomers in solution, and thus poses significant challenges for biophysical studies. To overcome oligomerization, AIM2 PYD has been fused to an MBP (Maltose-binding protein) tag and a specific mutant, F27G, has been created to shift the monomer-oligomer equilibrium to the monomeric form [169]. The crystal structures of mouse AIM2 PYD (mAIM2 PYD ), wild-type (WT) human AIM2 PYD (hAIM2 PYD ), and hAIM2 PYD F27G mutant very similar, as expected for an amino acid sequence identity of 56%. However, the conformational arrangement of helices α2 and α3 shows significant differences [169,170]. Helix α3 of mAIM2 PYD is positioned adjacent to the N-terminus of helix α2, similarly to the reported conformation of these helices for hNLRP10 PYD [169,170]. In contrast, helix α3 is positioned adjacent to the C-terminus of helix α2 in hAIM2 PYD F27G mutant and shows intermediate orientation in the structure of WT hAIM2 PYD . Differences in chain flexibility have also been observed; for example, the α2-α3 helical region is relatively ordered, with an average B-factor of 18.7 Å 2 in hAIM2 PYD F27G mutant and of~120 Å 2 in wild type AIM2 PYD [169].

AIM2 PYD Self-Association
Isolated AIM2 PYD can self-associate and form filaments similar to ASC PYD . Homology modeling of the AIM2 PYD filament using the cryo-EM structure of the ASC PYD filament as a template in combination with negative stating (ns)-EM data revealed that the AIM2 PYD filament shows a three-fold symmetry arrangement of the PYD protomer structures [167,171]. EM analysis proposes that the AIM2 PYD filament serves as a structural template for ASC polymerization [167]. It has been reported that mAIM2 PYD maintains monomeric conformation at low pH (4.0) and low salt concentration (< 100 mM) due to repulsive electrostatic forces between positively charged molecules and by interfering with hydrophobic interactions, whereas high salt concentration promotes oligomerization through hydrophobic interaction of hydrophobic patches on the protein surface [170]. These results confirm that both electrostatic and hydrophobic interactions are necessary for AIM2 PYD polymerization, which was also observed for the interaction between the PYDs of ASC and NLRP3 and for ASC PYD self-association [121,169,171]. Isolated AIM2 HIN cannot form ordered macrostructures, thus pointing to a predominant role of Death Domains such as PYD in the formation of ordered polymers. In fact, AIM2 PYD self-association has been shown to suffice for inducing the assembly and activation of the inflammasome [169].
Cryo-EM structural studies of GFP-AIM2 PYD filament indicate that this truncated construct forms~200 nm to~1 µm long filaments with outer and inner diameters of~90 Å and~20 Å, respectively. In this filament structure, the GFP tag protrudes from a filament core formed by AIM2 PYD [172]. Modeling studies propose that type I, II, and III interactions characteristic of Death Domains (PYDs and CARDs) play important roles in the helical organization of the AIM2 PYD filament. In the type I interaction, residues of helices α1 and α4 (S3, K6, L10, L11, D31, and I46) located on the first protomer interact with residues on helices α2 and α3 (R24, F27, F28, and D31) located on the fourth protomer. In the type II interaction, residues of helix α4 (Q54 and N55) located on the first protomer interact with residues of helices α5 and α6 (N73, Y74, and L76) located on the sixth protomer. In type III interaction, residues of helix α3 (G38 and K39) of the first subunit interact with residues in helices α1 and α6 (D15, N16 and I17) of the third protomer. The type I interaction is mediated by hydrophobic contacts and dominates filament assembly. Thus, mutations of residues such as F27G/F27L (type Ib surface) and L10A/L11A (type Ia surface) inhibit AIM2 PYD self-association and promote the monomeric form. Other structural studies have used the MBP fused to AIM2 PYD to impede oligomerization. It has been found that residues L10 and L11 are located near the MBP in this construct, which agrees with the finding that these residues are involved in oligomerization via the type I interface [172].

AIM HIN Domain Structure
The HIN domain of AIM2, with~200 amino acids, comprises two tandem OB (oligonucleotide/oligosaccharide binding) folds connected through a long linker ( Figure 3C). Canonical OB folds contain five β-strands that fold into two sheets [162]. The proximal OB1 fold consists of β1-β5, among them β1, β4, and β5 split into two short strands (β and β'). Similarly, the distal OB2 fold (β1-β5) shows the splitting of β5 into two shorter strands. The linker connecting OB1 and OB2 is~30 residues long and is folded into two alphahelices. The two OB folds firmly interact with one another through conserved hydrophobic interactions. The HIN domains of AIM2, IFI16, and p202 are highly conserved and show an identical topological arrangement of OB folds [162,173].

AIM2 HIN:dsDNA Interaction
The crystal structure of the AIM2 HIN domain in complex with dsDNA derived from the Vaccinia virus was determined using X-ray crystallography [162] (Figure 3C). This structure reveals that the highly positively charged surface of HIN interacts with the sugarphosphate backbone of dsDNA mainly via electrostatic interactions. The N-terminus of the HIN domain is positioned far away from the DNA-binding surface, possibly facilitating the interaction between the N-terminal PYD of AIM2 with the adaptor protein ASC. The binding of the HIN domain to both the major and minor grooves of the DNA could explain AIM2-induced activation of the innate immune system in the presence of dsDNA, but not ssDNA [66,68,162]. Both OB folds and the connecting linker participate in DNA binding. Specifically, residues K160 (β1), K162 (β1), and K163 (β1-β1' loop) of the OB1 fold, as well as residues L267 (β1), N287 (β2), K309 (β4), R311 (β4), K335 (β5), and I337 (β5), of the OB2 fold and linker residues such as R244 (α2), G247 (α2), and E248, T249, and K251 located at α2-α3 loop, participate in the binding between AIM2 HIN domain and dsDNA mainly via hydrogen bonding, van der Waals interactions, and salt bridges. The crystal structure also shows the formation of bidentate hydrogen bonds between residue R311 and phosphate groups in the DNA backbone [162].
It has been reported that~80 bp of dsDNA is the minimum size required for the induction of IL-1β by AIM2 activation [162]. Each HIN domain occupies four DNA base pairs, hence~20 AIM2 HIN domains wrap around the 80 bp of dsDNA with an observed axis tilt of 35 • . Multiple sequence alignment suggests that most residues interacting with dsDNA are also conserved in IFI16 HIN and mouse AIM2 HIN domains. Site-directed mutagenesis studies indicate that mutations involving residues located on the interacting regions, such as the OB1-linker, the OB1-linker-OB2, and residue F165, lead to a diminished binding affinity of AIM2 HIN to dsDNA. These results on AIM2 HIN were corroborated by similar mutagenesis experiments conducted on full-length AIM2 (AIM2 FL ), which resulted in an impaired association of AIM2 with DNA and reduced IL-1β secretion [162]. Altogether, these studies suggest that an intact receptor binding surface is required for the association to dsDNA and the subsequent immune responses.

AIM2 PYD:HIN Interaction
Xiao et al. originally proposed that in the absence of a ligand, intramolecular interactions between the PYD and HIN domains in AIM2 retain the sensor in an autoinhibited state that prevents PYD-mediated oligomerization and suppresses HIN:DNA binding [162].
Docking analyses of crystal structures suggest that the negatively charged helix α2 of the PYD locates at the interface of the PYD:HIN interaction. In addition, ITC (Isothermal titration calorimetry) studies reveal that AIM2 PYD interacts with AIM2 HIN with a dissociation constant (K d ) of 23.5 µM. Furthermore, it has been shown that mutations of acidic residues located in helix α2 abolish the binding of the PYD and HIN domains [161]. These results confirm that the PYD-HIN interface is dominated by electrostatic interactions between negatively and positively charged residues in the PYD and HIN domains, respectively. The negatively charged surface of AIM2 PYD that participates in the interaction with AIM2 HIN is also involved in the binding to ASC PYD , hence ensuring downstream signaling to the adaptor ASC only after AIM2 is activated by dsDNA.

Importance of AIM2 PYD in dsDNA Interaction and Oligomerization
The hypothesis of the autoinhibitory model was challenged by Sohn and colleagues [171]. Based on their studies, they proposed that AIM2 PYD does not participate in AIM2 autoinhibition, instead, it actively helps in DNA binding and concomitant self-association. We mentioned above that the fusion of MBP to the N-terminus of AIM2 FL interferes with PYD oligomerization. To interrogate whether the PYD has a role in dsDNA binding, fluorescence anisotropy experiments were conducted to compare the affinity of fluorescein amidite (FAM)-labeled 72-bp dsDNA with MBP-AIM2 FL , MBP-AIM2 HIN , untagged AIM2 FL , and untagged AIM2 HIN . The results show that MBP-AIM2 FL binds 2-fold tighter to dsDNA than MBP-AIM2 HIN , whereas untagged AIM2 FL binds at least 20-fold more tightly than MBP-tagged AIM2 variants in presence of 160 mM KCl. Another important finding from the Sohn group is that the isolated HIN domain can oligomerize upon dsDNA binding and thus assists in filament formation. Salt concentration-dependent binding reveals that AIM2 HIN oligomerizes on dsDNA in presence of 160 mM KCl, but fails to bind at 400 mM KCl. On the other hand, AIM2 FL binds to dsDNA even at this high salt concentration, which hints the involvement of PYD in dsDNA binding. Furthermore, mutations of residues L10, L11, and F27 involved in AIM2 PYD self-association and non-conservative mutations of AIM2 PYD acidic residues D19, E20, E21, and D23 impede the binding of AIM2 FL to dsDNA at 400 mM KCl, further supporting that the oligomerization of AIM2 PYD plays an important role in dsDNA binding. Although the effects of these mutations in the 3D-fold of the PYD were not tested, it would be expected that these residues facilitate the transformation of AIM2 from the autoinhibited conformation to the activated form that can bind DNA even at high salt concentration. Therefore, the results obtained with these mutants contradict the inhibitory role of the AIM2 PYD in DNA binding.
AIM2 FL needs to bind to a larger dsDNA size (~12 bp) as compare to the HIN domain alone (~8 bp), confirming the relevance of oligomerization to potentiate dsDNA binding. Furthermore, binding of AIM2 FL to dsDNA increases 1000-fold in the presence of 10-times longer DNA, indicating cooperativity between dsDNA size and AIM2 binding affinity. For the formation of AIM2 FL -dsDNA complex,~70 bp dsDNA and six molecules of AIM2 FL are needed to cross the binding threshold (lag phase), and~250-300 bp dsDNA and 24 AIM2 FL molecules are required for establishing an optimal oligomeric complex, as determined by fluorescence anisotropy competition binding assays using FAM-dsVACV72 (1.5 nM) and AIM2 FL (70 nM) against various fragments of dsDNA at 400 mM KCl. The data were fit to competition binding equation; 1/[1 + DN A competitor /IC 50 ) Hill constant [171]. This observation is further confirmed by monitoring the increase of IL-1β secretion with increasing dsDNA size. Overall, these results propose that dsDNA size acts as a 'molecular ruler' to regulate AIM2 inflammasome assembly in a switch-like mechanism of PYD oligomerization and dsDNA binding.
In the presence of dsDNA excess, AIM2 FL shows saturating or increased size-dependent FRET (Fluorescence resonance energy transfer) signals, but AIM2 HIN displays decreased FRET signals, indicating that AIM2 PYD is key for dsDNA binding and promotes oligomerization in presence of dsDNA excess [171]. Interestingly, ns-EM images illustrate that AIM2 FL is able to self-oligomerize in the absence of dsDNA forming "Brussels sproutlike" filaments at high concentration (≥500 nM) [171]. Oligomerized AIM2 PYD forms the filament core (~9 nm) and the HIN domains are observed at the periphery of this core, like Brussel sprouts. In contrast, filaments formed by AIM2 FL bound to dsDNA arẽ 25 nm wide. The participation of the PYD in filament assembly is critical, as isolated AIM2 HIN and MBP-AIM2 FL do not show any ordered filament formation in absence of dsDNA, and isolated AIM2 HIN displays random 'beads on a string'-like cluster upon dsDNA addition. Moreover, mutagenesis studies show that both PYD and HIN domains are required for the oligomerization of AIM2 FL in the presence or absence of dsDNA. FRET results reveal that the length of dsDNA regulates the assembly kinetics and lifetime of the dsDNA-AIM2 complex [174]. Modeling analysis based on cryo-EM and ns-EM observations propose that in the dsDNA-AIM2 HIN filament complex of~7.5 nm diameter, AIM2 HIN is wrapped around the dsDNA core and each HIN molecule interacts with six adjacent HIN molecules [172]. Such an arrangement of AIM2 HIN around the DNA core and long linker between both domains brings AIM2 PYDs into close proximity where they form short helical filaments proposed to run parallel to the DNA and to act as a platform for ASC PYD filament nucleation.
Altogether, these results propose that in the absence of cytosolic dsDNA, AIM2 is expressed at a very low basal concentration level, and is therefore unable to oligomerize and induce downstream signaling via ASC. Pathogenic attack facilitates rapid oligomerization due to invasion of dsDNA in the cytosol, which hikes AIM2 local concentration [171]. The size of dsDNA acts as a molecular ruler and governs the AIM2 assembly.

Negative Regulators of AIM2 Inflammasome Activation
Regulation of inflammasome assembly is imperative for maintaining cellular homeostasis. The mouse protein p202 has been reported to sequester cytoplasmic dsDNA and inhibit AIM2 activation [69]. p202 consists of two HIN domains and lacks the PYD, rendering it unable to recruit ASC. The binding of p202 to DNA and AIM2 is proposed to attain a balance between pathological DNA-induced inflammation and physiological host defense. The crystal structure of mouse p202-dsDNA complex reveals that the p202 HIN1 domain binds to DNA, whereas p202 HIN2 interacts with AIM2 [173]. Full-length p202 (p202 FL ) forms a tetramer in cells as well as in vitro purified protein solutions. p202 HIN2 first dimerizes in a parallel fashion using both OB folds (OB1-OB2 to OB1-OB2) and the formed dimers assemble into tetramers in a tail-to-tail orientation of the OB2 folds (OB2 to OB2) [166]. Although p202 HIN2 lacks DNA binding capability, tetramer formation serves as a platform for p202 HIN1 attachment to dsDNA, increasing the overall DNA binding affinity of p202 FL as compared to AIM2. p202 HIN1 shares structural similarity with mAIM2 HIN and IFI16 HIN2 , but shows different charge distribution and opposite orientation of the dsDNA binding surface. Such a difference in surface electrostatic potential is responsible for the antagonist activity of p202 [166].
Unlike AIM2, the linker connecting the two OB folds of p202 does not participate in DNA binding. Positively charged residues located at the N-terminus and loop between β1-β2 of the OB1 fold engage with the DNA minor groove [166]. In the OB2 fold, residues located in the loop connecting β1-β2 and the loop between β4-β5 interact with the dsDNA major groove. Structure-based mutagenesis studies propose that among these, OB1 Nterminal residues K48, and K53, and OB2 residue R224 are crucial for HIN:DNA interactions. Most of these residues interact with the backbone of DNA; however, K53 side chain was found to make two hydrogen bonds with DNA bases. p202 HIN2 interacts with AIM2 through a short sequence motif (MFHATVAT) conserved in both proteins and buried in the core of the HIN domains [175,176]. The protein region, MFHATVAT, is required for p202 dimerization and subsequent interaction with AIM2 [175]. It has been reported that p202 HIN2 does not block the DNA binding surface of AIM2; therefore, DNA binding affinity of AIM2 remains unaffected in the presence of p202 HIN2 [166]. Computational docking studies showed that the binding of AIM2 HIN domains with both ends of the p202 HIN2 tetramer creates a spatial separation between AIM2 PYDs , thus preventing ASC oligomerization [166]. Consequently, the knockdown of p202 increases the level of ASC and cross-linked ASC oligomers [166]. In this line, modeling studies propose that two adjacent mouse AIM2 (mAIM2) molecules bound to DNA are separated by less than 10 Å, thus generating AIM2 molecular crowding and favoring the interaction with ASC PYD and the subsequent activation of inflammasome assembly [173]. In contrast, p202 spans a larger dsDNA fragment and binds with higher affinity compared to AIM2. Therefore, when both p202 and AIM2 are present in equal amounts, the former competes with the latter for dsDNA binding and covers a larger surface area of dsDNA [173].

IFI16-β Mediated Regulation of AIM2 Inflammasome Activation
A novel human isoform of IFI16 designated as IFI16-β has been shown to selectively inhibit the formation and activation of AIM2 inflammasome assembly [177]. IFI16-β is ubiquitously expressed in various human cells and shows upsurge expression in leukocytes in case of viral infection. IFI16-β co-localizes with AIM2 in the cytoplasm and by sequestering cytoplasmic dsDNA, impedes its detection by AIM2. Analogously to p202, IFI16-β contains two HIN domains (HIN A and HIN B) and disrupts AIM2-ASC inflammasome activation by interacting with AIM2, competing with dsDNA binding as well as inhibiting AIM2 oligomerization [166,173,177]. Competition binding experiments suggest that IFI16-β binds with higher affinity to dsDNA than AIM2 because the IFI16-β-DNA complex shows a more prominent band in biotin-dsDNA pull-down assays as compared to AIM2 [177]. Altogether, dsDNA binding studies of p202 and IFI16-β indicate that proteins expressing two HIN domains bind to dsDNA more robustly than single HIN domain-containing proteins like AIM2 [166,173,177].

Post-Translational Modifications of AIM2
Very limited information is available on AIM2 post-translation modifications. However, it has been reported that TRIM11 (tripartite motif 11) acts as a negative regulator of the AIM2 inflammasome. TRIM11 binds AIM2 and undergoes poly-ubiquitination at K458, leading to the recruitment of autophagy cargo receptor p62, thus mediating the subsequent degradation of AIM2 [178,179]. In addition, studies conducted on mouse models of stroke and cultured primary microglia show elevated expression of HDAC3 (Histone deacetylases 3) linked to the regulation of the inflammatory process by activating the AIM2 inflammasome. RGFP966, a HDAC3 inhibitor, downregulates the AIM2 inflammasome by enhancing acetylation and inhibiting phosphorylation (at Y701 and S727) of STAT1 (Signal transducer and activator of transcription) in order to protect against ischemic brain injury [180].

ASC
ASC (PYCARD; PYD and CARD domain-containing or TMS; Target of Methylationinduced Silencing-1) is a~24 kD bifunctional cytosolic adaptor protein that consists of an N-terminal PYD (1-89) and a C-terminal CARD  connected by a 23 residuelong linker   [181][182][183] (Figure 4A). ASC expresses in the nucleus of epithelial and immune cells, and in response to inflammatory stimuli, is redistributed to the cytoplasm where it assembles into a compact micrometer-size perinuclear structure referred to as ASC speck or ASC foci [181,182,184]. The ASC speck colocalizes with the sensor and the procaspase-1 by homophilic interactions mediated by the PYD and CARD domains, thus forming the inflammasome, which serves as the platform for caspase activation and pyroptotic cell death [168,[185][186][187]. In addition to ASC FL , three other isoforms also exist: ASC-b, which also bears an N-terminal PYD and C-terminal CARD as with ASC FL , although connected by a short 3 amino acid-long linker; ASC-c, retaining the CARD, but only a partial PYD; and ASC-d, a 105-amino acid long polypeptide that only conserves residues 1-35 of the original ASC FL sequence [188]. These isoforms respond differently to inflammatory stimuli, exhibit irregularly shaped perinuclear aggregates and differential cellular expressions.  ASC PYD interacts with the PYDs of NLRP3 and AIM2, whereas ASC CARD interacts with the CARDs of procaspase-1 and NLRC4 via homotypic interactions [20,189]. The 3D NMR-solution structure of ASC reveals that the PYD and CARD domains form rigid structures with RMDS of 0.78 ± 0.07 and 0.79 ± 0.08 Å, respectively. The two Death Domains do not interact with one another based on Nuclear Overhauser data (NOE) and are connected by a linker that shows the residual secondary structure and fast local motion on the picosecond time scale [183,190]. NMR-based secondary chemical shift analysis and NOE data indicate that the linker adopts low populated extended structures analogous to polyproline II-like conformation [183]. Rotational correlation times (τc) derived from NMR relaxation experiments for ASC FL and the individual domains suggest that both domains reorient at different rates, but feel the drag from each other due to the presence of the linker [183].

Structural Details of ASC PYD and Its Self-Association
ASC PYD adopts the classic six-helical bundle motif typical of the DD-fold showing a long loop between helices α2 and α3, a unique feature commonly found in PYDs [167,189] ( Figure 4B). The electrostatic surface of ASC PYD is highly bipolar, with helices α1 and α4 containing mainly negatively charged residues, whereas helices α2, α3, and the connecting loop mostly accommodate positively charged residues [189]. Charge complementarity and the corresponding charge-charge interactions resulting from the bipolar distribution of the electrostatic surface are responsible for ASC PYD self-associations. Two oppositely charged surfaces of ASC PYD assemble back to back and self-associate with KD = 40 μM-100 μM using the dominant type I interaction mode, resulting in a buried surface area of 880 Å 2 . In the type I interaction for ASC PYD self-association, helices α1 (E13), α4 (D51), the Nterminus of helix α5, and the α3-α4 loop (D48) of one surface interact with helices α2 (K21), α3 (R41), and the C-terminus of helix α5 of the opposite surface [112,167,189]. ASC PYD displays a structural difference compared to other PYDs such as NALP1 PYD and NALP10 PYD , as the latter shows considerable variation in the length of helices α1 and α6,  [183]. The ribbon diagram is generated with PyMOL molecular graphics software. ASC PYD interacts with the PYDs of NLRP3 and AIM2, whereas ASC CARD interacts with the CARDs of procaspase-1 and NLRC4 via homotypic interactions [20,189]. The 3D NMR-solution structure of ASC reveals that the PYD and CARD domains form rigid structures with RMDS of 0.78 ± 0.07 and 0.79 ± 0.08 Å, respectively. The two Death Domains do not interact with one another based on Nuclear Overhauser data (NOE) and are connected by a linker that shows the residual secondary structure and fast local motion on the picosecond time scale [183,190]. NMR-based secondary chemical shift analysis and NOE data indicate that the linker adopts low populated extended structures analogous to polyproline II-like conformation [183]. Rotational correlation times (τ c ) derived from NMR relaxation experiments for ASC FL and the individual domains suggest that both domains reorient at different rates, but feel the drag from each other due to the presence of the linker [183].

Structural Details of ASC PYD and Its Self-Association
ASC PYD adopts the classic six-helical bundle motif typical of the DD-fold showing a long loop between helices α2 and α3, a unique feature commonly found in PYDs [167,189] ( Figure 4B). The electrostatic surface of ASC PYD is highly bipolar, with helices α1 and α4 containing mainly negatively charged residues, whereas helices α2, α3, and the connecting loop mostly accommodate positively charged residues [189]. Charge complementarity and the corresponding charge-charge interactions resulting from the bipolar distribution of the electrostatic surface are responsible for ASC PYD self-associations. Two oppositely charged surfaces of ASC PYD assemble back to back and self-associate with K D = 40-100 µM using the dominant type I interaction mode, resulting in a buried surface area of 880 Å 2 . In the type I interaction for ASC PYD self-association, helices α1 (E13), α4 (D51), the N-terminus of helix α5, and the α3-α4 loop (D48) of one surface interact with helices α2 (K21), α3 (R41), and the C-terminus of helix α5 of the opposite surface [112,167,189]. ASC PYD displays a structural difference compared to other PYDs such as NALP1 PYD and NALP10 PYD , as the latter shows considerable variation in the length of helices α1 and α6, and helix α3 is replaced by a disordered loop that may dictate their exclusive inflammatory functions [183].
Mutations of hydrophobic residues located in the PYD disrupt PYD-PYD filament formation, significantly increasing solubility at neutral pH while still retaining the monomeric folded conformation [190,191]. Similarly, NMR studies of ASC mutants in residues located in the type I and type III interfaces show complete or partially reduced ability of filament formation [167,190,192]. The L25A mutation in ASC PYD has been commonly used in structural and biophysics studies to avoid oligomerization. NMR-based chemical shift analysis showed that the L25A mutation causes structural perturbations around residues K24 and L45 located at the α2-α3 binding interface. Structural perturbation around K24 destabilizes the α3-helix, hence diminishing ASC PYD oligomerization by reducing the PYD binding ability. However, it is able to form dimers via the α1-α4 interface [112,190].
Cryo-EM studies revealed that ASC PYD subunits pack densely in a helical tube-like, three-fold symmetry structure with 53 • right-handed rotation and 14.0 Å axial rise consisting of six molecules per turn with inner and outer diameter of~20 Å and~90 Å, respectively [167,193]. NMR and cryo-EM based structural analysis of ASC PYD and its comparison with other members of the Death Domain superfamily propose the involvement of all three interaction types in the stabilization of the PYD filament; i.e., intra-strand type I and inter-strand type II and III interactions [167,190]. These observations were further supported by mutagenesis experiments [182]. The type II interaction mode with a buried surface area of 524 Å 2 involves contacts between helix α4 and the α4-α5 loop of one surface with the α5-α6 loop of the opposite surface. The type III interaction mode with a buried surface area of 360 Å 2 involves contacts between helices α2 and α3 of one surface and the α1-α4 loop on another surface [167,182]. NMR and analytical centrifugation studies show that ASC PYD polymer formation is favored in the presence of salt. Cryo-EM and solid-state NMR experiments conducted on mouse ASC PYD with 71.8% sequence similarity to human ASC PYD reveal very similar polymer structures [182,194].

Structural Details of ASC CARD and Its Self-Association
CARDs adopt a conserved six-helix bundle fold and exclusively exhibit helix α1 divided into two small fragments, α1a and α1b, connected by a hinge [183,195,196]. Apaf-1 (Apoptotic protease activating factor), NOD1, ICEBERG, and RAIDD (RIP-associated ICH1/CED3-homologous protein with a death domain) are structurally homologous proteins, and structural comparison of their CARDs ( Figure 5A) indicates differences in the length and orientation of the helices [183]. Among these, NOD1 CARD exhibits an extended long helix composed of helices α5 and α6 due to their close proximity [183]. Electrostatic surface analysis of these CARDs shows polarized distribution of basic and acidic surfaces, which dictates specific protein-protein interactions Figure 5B [197]. NMR experiments reported the absence of a fragmented helix α1, variability in length and orientation of the helices and evenly distributed charge in the electrostatic surface of ASC CARD [183] ( Figure 4B). ASC CARD self-oligomerizes with a dissociation constant of 50 µM [112,198]. This NMR study indicates that residues located in the turn preceding helix α1 and helices α2, α3, α5, and α6 are involved in the self-association of ASC CARD , and generate three contact regions involving; (1) the N-terminus of helix α1 and C-terminus of helix α6; (2) the C-terminus of helix α5 and the N-terminus of helix α6; and (3) helices α2 and α3 [190]. Negatively stained TEM images illustrate that ASC CARD assembles into two types/levels of filaments;~3.4 ± 0.5 nm wide filaments that self-assemble into~10 ± 0.5 nm wide bundles of filaments [190,198]. These NMR and TEM studies conclude that ASC CARD plays a key role in the structure and stabilization of ASC filaments [190].  [195] and iceberg (light blue) [199]. Superposition of RAIDD-CARD (pink) [200] and NOD-1-CARD (purple) [201]. Helix 1 (cylinder) in RAIDD-CARD, although not fragmented, is substantially bent and is shown as two cylinders; (B) electrostatic surface representation of CARDs in the same orientation as displayed in A. This figure is adapted from figure S4 of [183].
In addition, cryo-EM studies reveal that ASC CARD can also polymerize into a helical tube-like filament with a diameter of ~8 nm and 3.6 subunits per turn, stabilized by type I, II, and III interactions [190,198,202,203]. The most predominant type I interaction involves charge-charge contacts between helix α2 (R119) in the surface of one protomer, and helices α1 and α4 (E130, D134 and R160) of the adjacent protomer. Hydrophobic residues such as W169 and Y187 participate in type II and III interactions. The type III interaction is also dominated by charge-charge interactions between R160 of helix α4, and D143 and E144 of helix α3 [202]. In the case of CARD polymerization, mutations involving residues that participate in type I (R119D, N128A/E130R, and D134K), type II (W169G, Y187A, Y187K), and type III interactions (D143K/E144K and R160E) completely abolish filament formation [202]. Analogously, mutations of E130, W131, and D134 by alanine impede the ability to oligomerize and thus ASC foci formation. These mutants form instead short and thin filaments as compared to WT ASC CARD . In addition, non-conservative mutations of negatively charged residues such as E130, D134, D191, and E193 by arginine completely impede filament formation [190].

ASC Filament Formation
TEM analysis of the dimensions of filaments and filament bundles formed by ASC FL and the individual domains, PYD and CARD, indicate that both domains form an integral part of the ASC filament, thus elevating the role of ASC CARD in filament formation, which has not been recognized in different studies of the truncated protein carrying only the PYD domain. In high-resolution TEM images, it is possible to discern the presence of stacked rings with an average diameter of 5 ± 0.6 nm, close to the dimensions of the experimentally-derived model of human ASC FL dimer of 6 nm [198]. In addition, single-molecule FRET experiments also reported that both domains in ASC form fibrils in which the CARD folds back onto the PYD domain [204]. These results altogether show that the ASC dimer serves as a building block for ASC oligomerization, and both PYD and CARD do-  [195] and iceberg (light blue) [199]. Superposition of RAIDD-CARD (pink) [200] and NOD-1-CARD (purple) [201]. Helix 1 (cylinder) in RAIDD-CARD, although not fragmented, is substantially bent and is shown as two cylinders; (B) electrostatic surface representation of CARDs in the same orientation as displayed in A. This figure is adapted from Figure S4 of [183].
In addition, cryo-EM studies reveal that ASC CARD can also polymerize into a helical tube-like filament with a diameter of~8 nm and 3.6 subunits per turn, stabilized by type I, II, and III interactions [190,198,202,203]. The most predominant type I interaction involves charge-charge contacts between helix α2 (R119) in the surface of one protomer, and helices α1 and α4 (E130, D134 and R160) of the adjacent protomer. Hydrophobic residues such as W169 and Y187 participate in type II and III interactions. The type III interaction is also dominated by charge-charge interactions between R160 of helix α4, and D143 and E144 of helix α3 [202]. In the case of CARD polymerization, mutations involving residues that participate in type I (R119D, N128A/E130R, and D134K), type II (W169G, Y187A, Y187K), and type III interactions (D143K/E144K and R160E) completely abolish filament formation [202]. Analogously, mutations of E130, W131, and D134 by alanine impede the ability to oligomerize and thus ASC foci formation. These mutants form instead short and thin filaments as compared to WT ASC CARD . In addition, non-conservative mutations of negatively charged residues such as E130, D134, D191, and E193 by arginine completely impede filament formation [190].

ASC Filament Formation
TEM analysis of the dimensions of filaments and filament bundles formed by ASC FL and the individual domains, PYD and CARD, indicate that both domains form an integral part of the ASC filament, thus elevating the role of ASC CARD in filament formation, which has not been recognized in different studies of the truncated protein carrying only the PYD domain. In high-resolution TEM images, it is possible to discern the presence of stacked rings with an average diameter of 5 ± 0.6 nm, close to the dimensions of the experimentally-derived model of human ASC FL dimer of 6 nm [198]. In addition, singlemolecule FRET experiments also reported that both domains in ASC form fibrils in which the CARD folds back onto the PYD domain [204]. These results altogether show that the ASC dimer serves as a building block for ASC oligomerization, and both PYD and CARD domains are crucial for filament assembly. These latter studies proposed that speck formation has two levels of compaction; firstly, type I interactions mediates homophilic PYD-PYD and CARD-CARD binding, and secondly, type II and III interactions organize ASC into larger assemblies [182]. Computational and FRET studies suggest that ASC speck formation is not simply unspecific aggregation, but instead self-association follows an organized scaffold [182]. To confirm the involvement of both PYD and CARD domains in ASC speck formation, mutagenesis studies have been conducted. Single-point mutations in ASC (human) of residues important for the specific domain interactions such as E13A, E19A, K21A, K26A, R41A, D48A, D51A, L68A, L73A, in the PYD region, and M159A and R160A in the CARD, disrupt filament as well as ASC speck formation when present in ASC FL [112,182,191,192,205,206]. Double mutations (K26A-R160A and L68A-R160A) generated in both CARD and PYD domains result in an inability to form filaments as well as ASC specks [182]. These data suggest that speck formation is due to the individual homophilic interactions mediated by the PYD and CARD. Altogether mutational, dynamics and structural studies of human ASC propose that speck formation is based upon two levels of compaction: One level involving a main type of homophilic interaction between PYD-PYD and CARD-CARD, and a second level organized by other interaction modes (e.g., type II and III interactions) [182].
These results match previous NMR studies on the 3D structure determination of ASC and its proposed model for polymerization indicating that the PYD-PYD and CARD-CARD domains are positioned in a confined space so that they do not cause steric interference with the binding interface of each other [183]. Dynamics resulting from a slightly structured linker lead to a back-to-back orientation of the two domains that increase the accessible space to facilitate the interaction of both PYD and CARD with the PYD of the sensor and the CARD of procaspase-1 [183,190].

Regulation of ASC Mediated by ASC2
Humans encode for 10-13 kD single domain PYD-only proteins (POPs) such as POP1/ASC2, POP2 and POP3, and CARD-only proteins (COPs) such as Pseudo-ICE/COP, ICEBERG, INCA. ASC2 shares 63% sequence identity with ASC PYD [189,[207][208][209], interferes with PYD-PYD interactions of inflammatory proteins, and has been shown to be crucial for modulating NF-κB and pro-caspase-1 regulation [208]. ASC2 binds to ASC PYD with K D = 4.08 ± 0.52 µM and serves as a negative regulator of ASC polymerization [209]. The L25A mutant of human ASC PYD is capable of interacting with ASC2 (K D = 3.81 ± 0.8 µM) via the α1-α4 interface, which indicates that L25A mutation does not affect ASC PYD -ASC2 interaction at least in one of the possible interacting interfaces [209]. Site-directed mutagenesis of residues located in helices α2 and α3 of human ASC PYD (K21, L25, K26, P40, and R41) disrupts ASC PYD self-association without disturbing the hydrophobic pocket, thus indicating that ASC2 binding site on ASC PYD is different from the site of self-association [191,209]. ASC2 displays a different orientation of helices α2 and α4 and has a disordered α3 helix. ASC2 PYD and ASC PYD share similar 3D structures (RMSD = 1.5 Å) with comparable charge distributions across the surface [209]. NMR data indicate that the positively charged residues, K21 and R41, located on helices α2 and α3 of ASC2 interact via the type I interaction with the negatively charged residues D6, E13, D48, and D54 located on helices α1 and α4 of ASC PYD [209].

Post-Translational Modifications of ASC
Phosphorylation-dephosphorylation events are also important for ASC oligomerization and its activity. Tyrosine kinase-mediated phosphorylation of human ASC at Y60, Y137, and Y146 (Y144 in murine ASC) is necessary for inflammasome assembly and subsequent inflammatory response [210]. Similarly, the dephosphorylation of ASC tyrosine residues is also essential in the activation of the NLRP3 inflammasome. For example, it has been reported that the compound phenylarsine oxide (PAO), a tyrosine phosphatase (PTPase) inhibitor, suppresses ASC oligomerization and speck formation in LPS-primed human THP-1 cells by targeting the self-association nucleation step [210]. Differential ubiquitination of ASC by the K63 ubiquitin chain or linear ubiquitin plays an important role in ASC inflammasome activation. A ubiquitination enzyme complex LUBAC (linear ubiquitin chain assembly complex), which consists of HOIL-1, HOIP, and SHARPIN (Shank-associated RH domain-interacting protein) proteins, participates in linear ubiquitination of ASC via HOIL-1, HOIP E3 ligase activity [211,212] providing an activation signal. Another study reports that MAVS protein (mitochondrial antiviral signaling protein) recruits an E3 ligase, TRAF3 (TNF receptor-associated factor 3), that promotes ubiquitination of ASC at K174 position, which in turn increases ASC speck formation and secretion of IL-1β in response to viral infection [213]. In addition, it was found that a mitochondrial E3 ubiquitin ligase (Mul1 or MAPL or MULAN) abolishes inflammasome activation by K48-linked ubiquitination and subsequent proteasomal degradation of ASC [214].

Structure and Activation of Caspase-1
Caspase-1 (ICE; interleukin 1β-converting enzyme) is an inflammatory initiator that belongs to the aspartate-specific cysteine protease family [215][216][217]. It is expressed as a 404 amino acid-long inactive monomeric form called procaspase-1 zymogen, which is converted into a catalytic active form by autoproteolysis upon proximity-induced association mediated by the macromolecular organization of the inflammasome or ASC pyroptosome [218][219][220][221]. Procaspase-1 consists of one prodomain (or propeptide) CARD (1-119) that interacts with upstream adaptor proteins and a catalytic domain consisting of subunits p20 (120-297) and p10 (317-404) [167,222,223]. Caspase-1 activation involves proteolytic removal of the N-terminal CARD and 19 residues of the interdomain linker (298-316) connecting the p20 and p10 subunits [218,224]. Although the catalytic residues C285 and H237 reside in the p20 subunit, both subunits are essential for the activity [225]. X-ray crystallographic studies revealed that a tetramer of two p20/p10 heterodimers is considered as the functional form [223,226]. In contrast, recent cellular studies have shown that caspase-1 FL (p46) and transient species p33/p10 are dominant in the initial response to inflammasome assembly [219]. According to these studies, caspase-1 FL (p46) is recruited to the inflammasome via CARD-CARD interaction and generates an active p46 dimer, which then is self-processed and enables the cleavage of the linker connecting the p20 and p10 subunits to generate p30/p10 active species. Subsequently, the separation of the CARD domain linker (CDL) from p33/p10 releases the unstable p20/p10 tetramer (catalytic domain) from the ASC-caspase-1 complex, leading to the formation of the caspase-1 active form, therefore triggering the inflammatory response [219]. Activated caspase-1 facilitates the maturation of pro-IL-1β and pro-IL-18 into their bioactive forms IL-1β and IL-18, respectively [227,228] (Figure 6). Cytokine IL-1β induces the proliferation, activation, and differentiation of immune cells and facilitates phagocytosis, degranulation, and oxidative burst activity [229,230]. IL-18 is an inducer of IFN-γ and is involved in the activation and differentiation of various T-cell populations [231,232]. In addition to cytokine production, caspase-1 also cleaves gasdermin D (GSDMD) into two subunits of approximately similar size: N-and C-terminal halves. The GSDMD Nterm creates pores in the plasma membrane that are involved in cytokine secretion and facilitates cell-death by pyroptosis [224,[233][234][235][236][237] ( Figure 6). It has been shown that before CDL cleavage, dimerized caspase-1 retains its catalytic activity towards pro-IL-1β, pro-IL-18, and pro-gasdermin D, but once the CDL is detached caspase-1 protease activity deteriorates [219]. Figure 6. Schematic representation of NLRP3 and AIM2 inflammasomes activation and assembly. Pathogen-associated molecular patterns such as LPS, crystalline/particulate ligands, K + efflux, and ROS trigger the activation of NLRP3. TLR initiates the activation and nuclear translocation of NF-κB, which increases the synthesis of NLRP3 and IL-1β and IL-18 cytokines. AIM2 detects viral and bacterial dsDNA in the cytosol. Assembly of NLRP3 or AIM2 with ASC and procaspase-1 leads to the proximity-induced autoproteolytic maturation of caspase-1, functionalization of IL-1β and IL-18, and pyroptosis cell death mediated by the N-terminal fragment of gasdermin D. Inhibitors of NLRP3, ASC, caspase-1, and AIM2 are shown blue. The reference sources corresponding to the negative-staining electron micrographs are shown in square brackets in each image: ASC FL adapted from Figure 6 of [198], filamentous NLRP3 PYD-NBD -ASC PYD binary complex adapted from Figure 2 of [167]; filament of AIM2 FL from Figure 6 of [171], AIM2 FL filament with 600 bp dsDNA adapted from Figure  3 (Copyright National Academy of Science) of [174], and His-GFP-caspase-1 CARD /ASC FL /AIM2 PYD ternary complex adapted from Figure 6 [167].
When separated from the rest of the protein, caspase-1 CARD is able to polymerize into left-handed helical-tube macrostructures comprising four subunits per turn and with inner and outer diameters of ~ 10 Å and ~ 80 Å, respectively [231,238]. The formation of caspase-1 CARD filament also involves the interaction between the three interfaces [231]. Caspase-1 CARD filament shares helical symmetry with MAVS CARD filament and Myddosome DD complex. Fluorescence polarization results suggest that caspase-1 CARD polymerization increases in the presence of ASC CARD or ASC FL [167]. Oligomerized ASC CARDs nucleate procaspase-1 through CARD-CARD interaction thus serves as a platform for polymerization, autocleavage, and caspase-1 activation [167,[239][240][241].

Negative Regulation of Caspase-1 Activation
CARD-only proteins (COPs) inhibit inflammasome assembly and cytokine activation [242][243][244]. These inhibitors include COP-1 (Pseudo-ICE/CARD16), INCA (CARD17), and ICEBERG (CARD18), and share high sequence identity with caspase-1 CARD, i.e., 92%, 81%, and 53%, respectively [199,231,244,245]. COP-1 and ICEBERG can self-associate and form filaments, whereas INCA harbors monomer conformation [231,246]. In vitro and in vivo experiments show that ICEBERG is involved in the negative feedback of caspase-1 activation, and therefore suppresses IL-1β secretion [199]. ICEBERG has the ability to inhibit the interaction between RIP2 (Receptor-interacting protein 2, also known as or RIPK2 Figure 6. Schematic representation of NLRP3 and AIM2 inflammasomes activation and assembly. Pathogen-associated molecular patterns such as LPS, crystalline/particulate ligands, K + efflux, and ROS trigger the activation of NLRP3. TLR initiates the activation and nuclear translocation of NF-κB, which increases the synthesis of NLRP3 and IL-1β and IL-18 cytokines. AIM2 detects viral and bacterial dsDNA in the cytosol. Assembly of NLRP3 or AIM2 with ASC and procaspase-1 leads to the proximity-induced autoproteolytic maturation of caspase-1, functionalization of IL-1β and IL-18, and pyroptosis cell death mediated by the N-terminal fragment of gasdermin D. Inhibitors of NLRP3, ASC, caspase-1, and AIM2 are shown blue. The reference sources corresponding to the negative-staining electron micrographs are shown in square brackets in each image: ASC FL adapted from Figure 6 of [198], filamentous NLRP3 PYD-NBD -ASC PYD binary complex adapted from Figure 2 of [167]; filament of AIM2 FL from Figure 6 of [171], AIM2 FL filament with 600 bp dsDNA adapted from Figure 3 (Copyright National Academy of Science) of [174], and His-GFP-caspase-1 CARD /ASC FL /AIM2 PYD ternary complex adapted from Figure 6 [167].
When separated from the rest of the protein, caspase-1 CARD is able to polymerize into left-handed helical-tube macrostructures comprising four subunits per turn and with inner and outer diameters of~10 Å and~80 Å, respectively [231,238]. The formation of caspase-1 CARD filament also involves the interaction between the three interfaces [231]. Caspase-1 CARD filament shares helical symmetry with MAVS CARD filament and Myddosome DD complex. Fluorescence polarization results suggest that caspase-1 CARD polymerization increases in the presence of ASC CARD or ASC FL [167]. Oligomerized ASC CARDs nucleate procaspase-1 through CARD-CARD interaction thus serves as a platform for polymerization, autocleavage, and caspase-1 activation [167,[239][240][241].
It was proposed that the negatively charged surface of ICEBERG exhibits competitive binding for the positively charged surface of caspase-1 CARD with the upstream activator RIP2, which has a negatively charged patch [199]. Another study found that ICEBERG is not able to interact with RIP2, but COP1 can do so [231,[246][247][248][249]. In vivo assays showed that both ICEBERG and COP1 hamper the binding of RIP2 to caspase-1 and reduce IL-1β expression by~80% and 100%, respectively [246]. Two mutations, D27G and R45D, were created in caspase-1 in order to mimic the polypeptide sequence of INCA and ICE-BERG [245]. These mutants were unable to activate NF-κB signaling due to loss of caspase-1 CARD-CARD interaction [245].
It has been reported as well that INCA inhibits caspase-1 CARD polymerization at nanomolar concentration even in the presence of ASC CARD . The mechanism proposed for this inhibition involves the capping of the growing caspase-1 CARD filament via CARD-CARD interaction, thus blocking the binding of upcoming caspase-1 molecules. This process would abrogate full polymerization of the caspase and subsequent autoactivation [231]. On the other hand, it was found that ICEBERG neither interacts with caspase-1 CARD nor inhibits NLRP3 inflammasome activation, caspase-1 oligomerization and its activity [231]. These findings suggest that cellular and/or environmental factors may influence ICEBERG-mediated inflammasome inhibition. Therefore, comprehensive studies are required to unravel the structural and functional mechanisms governing these inhibitory processes [231].

Interaction of ASC with Procaspase-1 and Formation of the NLRP3 Inflammasome
ASC CARD is an integral part of the ASC filament and participates in speck formation [183]. In addition, co-expression of ASC CARD with caspase-1 can also form foci similar to those formed by ASC FL [206]. Procaspase-1 CARD has one negatively and one positively charged surface oriented in opposite sides of the domain, which facilitates the type I interaction that is prominent in the recruitment of procaspase-1 CARD by ASC CARD . Mutational studies reported key residues responsible for the formation of foci: R10 (α1), D27 (α2), E41 and K42 (α3), R55 and D59 (α4) on caspase-1 CARD , and R125, E130, D134, Y137, E144, R160, and D191 on ASC CARD [206,245,252]. The mutation of caspase-1 CARD residues D27 and R55 completely interrupts ASC-caspase-1 signaling [206]. It is interesting to note that except for residues D143 and Y146, the rest of the mutants designed to perturb the type I interaction abolish foci formation. However, these ASC mutants can interact with caspase-1 CARD and show oligomeric assemblies, albeit lacking the ability to propagate active signaling platforms, which results in concomitant loss of IL-1β secretion [206]. Mutations of residues D143 and Y146 do not destabilize the ASC CARD structure and allow foci formation, thus reflecting that a network of side-chain interactions might stabilize ASC CARD -ASC CARD binding even in the presence of mutations in the interface [196,206,252]. These results point out the crucial role of ASC CARD in foci formation and ASC-dependent inflammasome signaling [196,206,252].
The type III interaction (R45) is essential for caspase-1 CARD auto-oligomerization and recruitment of RIP2 [245,252]. D27 (type I interaction) mutant of caspase-1 CARD does not compromise auto-oligomerization; however, it fails to activate NF-κB signaling [206,245]. Similarly, caspase-1 CARD R45 mutant can interact with ASC without affecting proteolytic activation, but fails to trigger NF-κB signaling [245]. These findings suggest the importance of two oppositely charged surfaces and the synergistic effect of R45 and D27 on RIP2mediated activation of NF-κB signaling [206,245,252].
Immunoprecipitation and EM studies of the NLRP3 inflammasome triggered by monosodium urate (MSU) crystals in THP-1 cells indicate the formation of filamentous structures that cluster into ball-of-yarn-like particles upon overnight incubation [167]. Similarly, expression of eGFP-ASC in COS-1 cells against anti-ASC primary antibodies reveal the formation of a densely packed gigantic perinuclear punctum (~1-2 µm) in each cell [167]. Analogously to the AIM2-PYD /ASC/Caspase-1 CARD assembly, a macro protein complex formed by NLRP3 PYD /ASC FL /Caspase-1 CARD also assembles into initial star-shaped structures [68,167,253]. Molecular modeling based on NMR data and supported by in-cell immunoprecipitation and in vitro reconstitution suggest that ASC CARD monomers could form 6-7 member ring structures via the type I interaction to which procaspase-1 CARDs could stack, amplifying the disk-like structure while leaving the ASC CARDs accessible for interaction through type I, II, and III interactions [252]. The stacking of ASC FL onto this ring creates an additional ring of PYDs below the CARDs. Overall, this arrangement would help PYD:PYD and CARD:CARD self-association and its interaction with the NLRP3 and AIM2 sensors, and procaspase-1 [252]. The systematic formation of NLRP3-inflammasome has been illustrated in Figure 6.

Interaction of AIM2 with ASC and Formation of the AIM2 Inflammasome
Upstream cytosolic dsDNA sensor AIM2 induces the polymerization of downstream ASC via homotypic PYD-PYD interactions, which further activate procaspase-1 [167,238]. Filaments formed by AIM2 PYD and ASC PYD exhibit similar helical symmetry, overall dimensions, and subunit organization [167,171,174].
EM results along with modeling studies show that the AIM2 PYD filament nucleates the polymerization of the ASC PYD filament [174], resulting in the localization of the latter at the end of former [167]. Thus, although ASC PYD can self-polymerize, the presence of AIM2 PYD enhances this process. Fluorescence polymerization experiments suggest that both AIM2 PYD and the complex between AIM2 FL and dsDNA participate in ASC PYD filament nucleation. Although it was found that the length of dsDNA regulates the selfassociation and propagation of AIM2 and ASC PYD polymerization, both of them can accelerate their assembly irrespectively of the presence or absence of dsDNA [174]. It has been suggested that AIM2 PYD -ASC PYD assembly generates a three-stage continuous signal amplification in which AIM2 PYD , ASC PYD , and AIM2 PYD -ASC PYD complex filaments are all persistent. In order to initiate AIM2 PYD -ASC PYD assembly, a critical concentration threshold is required. Both AIM2 PYD and AIM2 FL can induce ASC PYD polymerization in a concentration-dependent manner and such ability is augmented by 4-fold in the presence of dsDNA [174]. FRET assays reveal that pre-assembled ASC PYD polymers accelerate AIM2 PYD polymerization as well as the assembly of AIM2 FL on dsDNA, resulting in the regulation of the AIM2-ASC PYD assembly by positive feedback loops. The AIM2 PYD/FL -ASC PYD complex filament is highly stable and once formed, biochemical data suggest that it cannot be disassembled [174].
6.4. AIM2 Ternary Complex: AIM2 PYD :ASC FL :GFP-Casp1 CARD To reconstitute the ternary complex, His-MBP-AIM2 PYD , His-MBP-ASC FL , and His-GFP-caspase-1 CARD were mixed in a ratio of 1:1:3 and incubated with TEV protease to remove the MBP tags [167]. After purification of the ternary complex, AIM2 and ASC were subjected to immunogold labeling, whereas caspase-1 CARD was labeled with Ni-NTAnanogold conjugate. EM results reveal the formation of a star-shaped ternary complex in which AIM2 PYD could nucleate the short filaments of ASC FL [167]. ASC subunits localize at the center of the complex-forming short filaments or rings, whereas the concomitant polymerization of AIM PYD and ASC PYD form long filaments. These results point to a critical role of ASC CARD in the control of ASC supramolecular assemblies, suggesting distinctive structural features of the latter compared to ASC PYD [198]. In the ternary complex, caspase-1 CARD is observed along the arms of the stars, possibly polymerizing via interaction with ASC CARD [167]. The presence of the flexible linker connecting the PYD and CARD in ASC could facilitate the interaction with caspase-1 CARD [183]. The formation of the AIM2 and NLRP3 inflammasomes has been illustrated in Figure 6.

Concluding Remarks and Perspective
Inflammasomes provide host defense from pathogens by means of IL-1β and IL-18 maturation and secretion. The complex and high-order oligomeric nature of inflammasome components pose significant challenges for protein expression by recombinant methods and for successful purification. Moreover, size and shape heterogeneity deter detailed structural investigation. However, recent advancements have provided new insights into the structure, function, activation, and regulation of NLRP3 and AIM2 inflammasomes. Structural techniques, including NMR, X-ray crystallography, and cryo-EM, in association with biophysical, biochemical, and cell-based studies, have revealed high-resolution structures that help to understand the molecular mechanisms of ligand/receptor-driven conformational changes, the release of auto-inhibition, oligomerization, and complex assembly, nucleationinduced polymerization of ASC and caspase-1, and further downstream signaling. PTMs play important roles in the control of inflammasome activation. Hence, PTM dysfunction leads to autoimmune diseases resulting from chronic inflammasome activation.
In spite of recent progress in inflammasome research, several fundamental questions still remain unanswered. For instance, the ligand-induced activation mechanism of NLRP3 and its contribution in pyroptosis, which in turn can cause serious injury to vital organs, are still unclear [254]. Little is known about the exact mechanism of phosphorylation and ubiquitin-mediated controlled activation of inflammasome components. AIM2 can detect damaged or mislocalized self-DNA, which is released into the cytosol due to the loss of nuclear envelope integrity resulting from the perturbation of cellular homeostasis [255,256]. However, detailed information on how AIM2 regulates the detection of self as well as foreign DNA is lagging. Similarly, NLRP3 is also involved in self-DNA sensing, but the exact mechanism of self-DNA induced activation is not known [257,258]. In addition to NLRP3 activation, NEK7 also regulates microtubule dynamic and spindle assembly during the cell cycle [259]. How this protein limits NLRP3 activity without affecting AIM2 activation during cell cycle progression is unknown. Potential roles of AIM2 in inflammasomeindependent processes, such as neuronal morphology, anxiety, and memory of mice will uncover new functions of AIM2 [260]. Although remarkable progress has been made on the structure of inflammasome components and mechanism of ASC speck formation, more structural details are required to elucidate the whole ASC speck assembly consisting of receptor, the ASC adaptor, and caspase-1. Likewise, super-resolution microscopy studies at the cellular level are required to uncover the formation of endogenous specks and involvement of accessory proteins that contribute to speck size and organized shape in intact cells. In-depth investigation of these fundamental questions will open up new doors for the development of novel therapeutics, and faster and efficient anti-inflammatory therapies for the treatment of associated autoimmune and autoinflammatory diseases. Finally, the determination of specific structural components of SARS-CoV-2 involved in NLRP3 activation and decoding the subsequent downstream pathway that leads to cell death, will aid in the development of potential therapeutics for the treatment of COVID-19 in near future.