FCAS was described in 1940 by Kile and Rusk who studied a unique family with affected members, equally male and female across five generations, exhibiting recurrent episodes of urticarial-like rash, limb pain, and fever following generalized cold exposure.¹² The proband described a lifelong history of burning, rather than pruritic, erythema beginning 30 min after going outside in cold damp weather, followed by fever to 101–103 °F, and joint stiffness lasting 6–8 h following mild cold exposure or 24 h following more extended cold exposures. Symptoms could be induced by immersion of her arm in cold water and were resistant to therapies for allergic urticaria commonly used at the time. Additional families with similar phenotypes and elevations in acute phase reactants were reported over the next 6 decades, variably described as cold hypersensitivity, familial cold urticaria (FCU), and finally FCAS^13,14 in an attempt to differentiate this chronic, inherited, systemic inflammatory disorder from the more common acquired cold urticaria.¹⁵

While the cold-induced urticaria uniquely defined the family described by Kile and Rusk, additional descriptions of families with apparent autosomal dominantly inherited urticarial disorders began to appear in the literature. In 1962, Muckle and Wells reported a family with similar features of non-infectious episodes of urticaria-like rash, limb pain, and fever with rigors, which they described as “aguey bouts” which lasted approximately 36 h.¹⁶ In contrast to FCAS patients, this family developed progressive bilateral sensorineural hearing loss beginning in childhood, and end stage renal disease secondary to AA amyloidosis beginning in adulthood, with laboratory evaluations showing anemia, increased serum immunoglobulins and elevated inflammatory markers.¹⁶ Similar to FCAS patients, these patients failed to respond to therapy with antihistamines or daily steroids. Muckle and Wells noted the similarities of recurrent skin rash and amyloidosis between MWS patients and FMF patients,¹⁷ in what may be one of the first attempts to phenotypically link autoinflammatory syndromes.

Chronic infantile neurologic cutaneous articular (CINCA) syndrome was first described in 1980 by Prieur who reported three unrelated children presenting in the neonatal period with a chronic urticarial-like rash, neutrophilia and significant involvement of the large joints. Described as “bread crumb appearance” on radiograph, the inflammatory joint disease involved the epiphyses and patella and resulted in physical deformation and disability.¹⁸ Extensive neurologic disease including seizures secondary to chronic sterile meningitis increased intracranial pressure and developmental delay were also described.¹⁸ Progressive deafness and visual impairment further added to the morbidity for this disorder. Initially termed chronic meningo-cutaneo-articular syndrome in children, other patients with similar phenotypes were subsequently described, and the disorder was referred to as either CINCA or NOMID.^19–21

Since the initial description of the NLRP3 gene in 2001, over 250 variants have been described worldwide.^25–28 There appears to be significant genotype–phenotype correlation in that certain mutations are often associated with specific disease spectrum severity, although different mutations affecting the same amino acid can cause different diseases along the CAPS spectrum (Figure 2).⁹ One complication in the literature is that two different methionines have been used as the initiating amino acid. The NLRP3 sequence from the originally used methionine starts with the amino acid sequence MASTR (e.g., NP_004886.2 and NP_001230062.1). The NLRP3 sequence has also been redescribed as beginning with an upstream in-frame methionine so that the protein starts with the sequence MKMASTR (e.g., NP_004886.3); however, this upstream methionine is not universally conserved in mammals. This ambiguity means that the same mutation in different publications can be given a position that differs by 2, for example p.L359W and p.L361W refer to the same mutation. In this review, we exclusively use the original nomenclature; hence, we will name the above mutation p.L359W.

The majority of patients in the CAPS spectrum have autosomal dominant NLRP3 missense mutations; however, there are mutation-negative patients with classic CAPS symptoms and pathology. At least some of these patients can be explained by the presence of somatic mutations that were not detected by standard sequencing, with some patients ultimately having mutations in NLRP3, and a few other patients with CAPS phenotype were determined to have NLRC4 mutations.^29–33 We have also identified a promoter variant in one NLRP3 mutation-negative patient associated with increased NLRP3 expression,³⁴ providing a potential novel disease mechanism. An extensive pathway-targeted candidate gene approach¹² and further linkage studies have not yet identified additional disease genes.

Our initial description of NLRP3 mutations in FCAS included a variant (p.V198M) that was not initially observed in a panel of healthy controls.⁶ However, additional sequencing revealed that this variant and two other variants (p.R488K and p.Q703K) that had been initially observed in CAPS phenotypes were also identified in healthy controls as well as patients with atypical clinical presentations. The higher frequency of these variants in different populations and the genetic data is not supportive of classifying them as pathogenic variants; however, there are clinical and functional data to suggest that they have an intermediate phenotype between pathogenic CAPS and healthy controls.³⁵ The phenotypic differences between the classic CAPS mutations and low-penetrance mutations provide insight into the unique role of NLRP3 in modulating the immune response. Subsequently, detailed evaluations uniting molecular experiments and structural biology have been instrumental in our understanding of how NLRP3 responds to an array of diverse stimuli and directs formation of the cytoplasmic multiprotein platform known as the inflammasome.

Infection by pathogens and other environmental cues give rise to pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs) that induce the innate immune system. These PAMPs and DAMPs are thought to trigger the assembly of NLR proteins into large molecular weight complexes, termed inflammasomes after a pivotal study by Jorg Tshopp’s group describing a lipopolysaccharide (LPS)-inducible assembly of NLRP1 with apoptosis related speck-like protein containing CARD (ASC), caspase-1, and caspase-5.³⁹ Inflammasome formation has been verified for multiple NLR proteins, including NLRC4, NLRP1, NLRP3, and NLRP7.^39–44 Inflammasome formation by NLRP3 appears to be driven by two signals: The first signal is “priming,” and the second signal is “activation”. Signal 1 leads to NF-κB-driven expression of genes encoding NLR proteins and proinflammatory cytokines, such as IL-1β and IL-18, as well as deubiquitination of the LRR domain by BRCC3,^45–47 the recognition of LPS by Toll-like receptor (TLR) proteins, and binding of proinflammatory cytokines to their cognate receptors. Once expressed, inactive NLRP3 appears to be stabilized in the cytosol through interactions with the SCF ubiquitin ligase subunit SGT1 and heat shock protein 90 (HSP90),⁴⁸ similar to the interaction for pathogen-sensing NLR proteins in plants.⁴⁹ Signal 2 includes a large variety of triggers for NLR protein assembly into inflammasomes and the maturation and release of the proinflammatory cytokines.

Assembled inflammasomes form 10- and 11-subunit disks comprised of the NACHT and LRR domains (for NLRP3, see Figure 4).^41–43 The N-terminal pyrin domains of NLRP inflammasomes can form 5-symmetric helical filaments alone,^42,50 and electron microscopy suggests that inflammasome assembly generates a 10-or 11-subunit NLRP pyrin domain proto-filament capable of nucleating a filament made up of the pyrin domains of the ASC adapter protein.⁵¹ ASC is comprised of two domains, an N-terminal pyrin domain and a C-terminal CARD domain, which recruits pro-caspase 1 through a homotypic interaction between the CARD domains of ASC and pro-caspase 1 (Figure 4).^51–53 Activation of pro-caspase 1 molecules is likely achieved through increased local concentration mediated by ASC binding followed by trans-cleavage and dimerization of the mature caspase 1 subunits. Mature caspase 1 then cleaves the pro-inflammatory cytokines, leading to their release into the extracellular matrix for signaling, and cleaves and activates gasdermin D (GSDMD). The inflammasome-triggered formation of ASC filaments serves as a signal amplification mechanism to drive the innate immune response.⁵⁴

The PAMPs and DAMPs that activate some NLR proteins have been identified; however, the mechanism of activation remains unclear for most of these proteins. NLRC4 responds to bacterial flagellins and the inner rod components of bacterial type III secretion systems; NLRP1 responds to dsRNA and dsDNA, associated with viral infections, and NLRP7 responds to microbial lipopeptides.^40,55,56 The precise mechanism by which these molecular patterns activate inflammasome formation remains poorly understood for most NLR proteins. NLRP1 assembly appears to be triggered by proteolytic cleavage of the N-terminus that is induced by a variety of factors, including viral proteases, viral replication, and ribotoxic stress, rather than by direct interaction with PAMPs and DAMPs.⁵⁷ In contrast, NLRP6 may form inflammasomes through a liquid–liquid phase separation after directly interacting with viral RNA molecules or lipoteichoic acid during infection by Gram-positive bacteria.⁵⁸ Similarly, NLRC4 utilizes an accessory subunit, one of the NLR family apoptosis inhibitory proteins (NAIPs, a single gene in humans but present in 6 copies in mice) that provides ligand specificity^56,59–61 and promotes inflammasome formation by nucleating a disk containing 1 NAIP protein and 10 NLRC4 proteins.⁴³ It has recently been suggested that NLRP3 also forms a non-canonical inflammasome that also contains NLRC4, the RNA helicase DDX17, ASC, and caspase-1 in response to the presence of short interspersed nuclear element (SINE) RNA molecules.⁶²

Remarkably, the NLRP3 inflammasome is unique in that it is responsive to a wide range of activating agents, including uric acid crystals, alum, silica, cholesterol crystals, extracellular adenosine triphosphate (ATP), hyaluronan, amyloid-β fibrils, mitochondrial reactive oxygen species (ROS) generation, and lowered intracellular potassium concentrations such as by nigericin or gramicidin treatment.^63–66 It seems unlikely that NLRP3 can directly detect each of these DAMPs, and a unifying feature for these agents has been suggested to be increased potassium efflux from the cell.^67,68 Consistent with this, inhibition of potassium efflux by the mammalian ketone metabolite β-hydroxybutyrate suppresses NLRP3 activity.⁶⁹

Reduced potassium levels appear to be an indirect activator of NLRP3 assembly.⁷⁰ Several of the treatments shown to activate NLRP3, including extracellular ATP and potassium efflux triggered by nigericin or gramicidin, lead to the disassembly of the trans-Golgi network. NLRP3 is recruited to vesicles in this dispersed trans-Golgi network (dTGN) through the charged interaction between the phosphatidylinositol-4-phosphate lipid and the linker between the NLRP3 pyrin and NACHT domains and a region of the fish-specific NACHT associated (FISNA) subdomain (Figure 3).^71,72 It has been proposed that the inactive “cage” form of NLRP3 could be the state at which NLRP3 binds to the dTGN vesicles⁷¹; however, constitutively activated CAPS mutant proteins are also capable of spontaneously forming membrane-bound NLRP3 puncta,⁷⁰ suggesting that active forms can also bind to membranes. Remarkably, two potassium-efflux independent NLRP3 activators, imiquimod and CL097, also promote the formation of dTGN vesicles and NLRP3 conformational changes similar to potassium efflux^70,72; however, it is currently unclear why the requirement for potassium efflux for NLRP3 activation varies among different treatments that disrupt the trans-Golgi network.

Even after full or partial assembly of NLRP3 on dTGN vesicles, several lines of evidence indicate that NLRP3 activity is further controlled through subcellular localization. First, NLRP3 activation in mice requires binding of the serine/threonine protein kinase NEK7 (centrosomal NIMA-related kinase 7) by the NLRP3 LRR domain (Figures 4 and 5).^73–75 Importantly, in humans, this activation appears to be independent of the NEK7 kinase activity.⁷⁵ While NEK7 binding would be predicted to disrupt the inactive NLRP3 “cage” structures (Figure 4),^50,71,76 NEK7 is primarily localized at the microtubule organizing center (MTOC), and activated NLRP3 is moved by microtubule retrograde transport from the trans-Golgi network to the MTOC.^77,78 This transport is responsible for the formation of ASC specks, which are large helical fibrils that typically form at one site per cell and are convenient experimental markers for NLRP3 activation, as they can be visualized by fluorescence microscopy.⁷⁹ The requirements for the NLRP3 retrograde transport are reminiscent of the retrograde transport of aggresomes, which are large molecular aggregates that are resistant to proteosome degradation and are degraded by autophagy at the MTOC.^80,81 A relocalization requirement likely explains why partial, but not complete, defects in inflammasome activity are caused by defects in the dynein adapter histone deacetylase 6 (HDAC6), HDAC6 inhibitors, defects in microtubule-affinity reducing kinase 4 (MARK4), dynein disruption, and loss of the aggresome component vimentin.^77,82–91 Cargo recruited for transport by HDAC6 is typically ubiquitinated,⁹² and extensive studies have indicated that NLRP3 inflammasome components are regulated both positively and negatively by ubiquitination.⁹³ A relocalization requirement also explains why the microtubule depolymerizing agent colchicine is useful in treating symptoms of gout, which is an NLRP3-mediated disease; colchicine blocks both ASC speck formation and IL-1β release driven by monosodium urea crystals.^94–96

It is currently unclear what assembly state NLRP3 is in during transport to the MTOC; however, a small molecule inhibitor of NLRP3, MCC950/CRID3, which likely prevents the inactive-to-active conformational change in NLRP3 subunits and subsequent inflammasome disc assembly, blocks ASC speck formation,⁹⁷ suggesting that, at minimum, the active conformation of NLRP3 is required. Similarly, the NLRP3-ASC interaction appears to be required. In the presence of ASC, activated NLRP3 is localized to a single speck; however, in the absence of ASC, activated NLRP3 is distributed in dTGN-associated puncta.^70,86 This suggests that ASC is required for retrograde transport or for retention of NLRP3 at the MTOC, possibly through recruitment of ubiquitinated ASC by HDAC6. ASC is known to be modified by linear ubiquitin chains via the linear ubiquitin chain assembly complex (LUBEC), and two LUBEC subunits, HOIL-1 and SHARPIN, promote inflammasome activity.^98,99 ASC is also subject to modification by K63-linked ubiquitin chains via the TRAF3 and Peli1 E3 ubiquitin ligases.^56,99–101 Remarkably, TRAF3, Peli1, and HOIL-1 were shown to promote ASC speck formation;^56,99,100 however, it is not yet clear if these proteins act at the step of retrograde transport.

The complex choreography required for the formation of active NLRP3 inflammasomes at the MTOC may have arisen due to the ability of NLRP3 to respond to a wide variety of inflammatory activators. Because the proximal signal for activation, likely disruption of the trans-Golgi network, is only indirectly tied to infection, preventing inappropriate activation and being able to rapidly silence the inflammasome to return to a basal state are potentially more crucial for NLRP3-mediated inflammatory responses than for inflammasomes triggered by direct interaction with PAMPs/DAMPs. Remarkably, both NLRP3 and pyrin inflammasomes localize to the MTOC, whereas absent in melanoma 2 (AIM2) inflammasomes do not.⁷⁷ Activation by MTOC recruitment and binding of the MTOC-resident protein NEK7 localizes activated NLRP3 inflammasomes where they could be rapidly degraded by autophagy. Consistent with this view, autophagosome formation is induced by many NLRP3 stimuli, autophagosomes partially colocalize with activated NLRP3 inflammasomes, loss of autophagosome function increases inflammasome activity, and K63-linked ubiquitin chain modification leads to autophagosome recruitment of ASC.¹⁰²

The cold specificity of NLRP3 activation in FCAS led to the development of ex vivo, in vitro and murine models to further study the role of NLRP3 and the CAPS phenotype (Table 2). Ex vivo, adherent monocytes derived from the peripheral blood of patients with FCAS demonstrate significant release of IL-1β with as short as 1 h of incubation at 32°C, suggesting the presence of pre-transcribed and pre-translated pro-IL-1β in the presence of mutations in NLRP3. Prolonged incubation at this reduced temperature leads to release of IL-1β, IL-6, and TNF-alpha, consistent with perpetuation of an inflammatory cascade. Cold-induced IL-1β release could be inhibited by caspase-1 inhibition, confirming a role for inflammasome activation in cold-related symptoms. In addition, anakinra failed to prevent early IL-1β release but significantly reduced the late-phase transcription and release of all cytokines, consistent with a pro-IL-1 autoinflammatory feedback loop.¹¹⁶ This ex vivo system of “intrinsic” NLRP3 activation has become a key model for pre-clinical studies of therapeutics targeting the NLRP3 inflammasome.

More recently, Karasawa and colleagues investigated the long-standing question of how cold induces NLRP3 activation in FCAS.¹¹⁷ Using transfected cell lines, the authors showed that cells expressing FCAS-associated mutations p.L353P and p.Y563N formed cryo-sensitive aggregates that act as a scaffold for inflammasome activation.¹¹⁷ Aggregate formation was dependent on caspase-1 mediated calcium influx, rather than potassium efflux, suggesting a possible divergence in inflammasome activation between wild-type and mutant NLRP3, which has implications for therapy. Additional studies have implicated the involvement of heat shock protein 70 (HSP70) in the mechanism of cold induction observed in FCAS-associated mutant NLRP3.¹¹⁸

Most CAPS mutations affect exon 4 of NLRP3 (previously called exon 3 prior to the identification of a short non-coding 5′ exon¹⁰³), which encodes the NACHT domain (Figure 3). Analysis of disease-causing NLRP3 missense mutations in the Infevers database²⁵ reveals seven protein regions (regions A–G) that are affected by many missense mutations in CAPS and other autoinflammatory diseases (Figure 2). In many cases, multiple missense mutations have been identified for the same residue in these regions.

The recent availability of active and inactive NLRP3 structures provides an opportunity to decipher the effects of these disease-causing mutations.^{42,50,76,119,120} Remarkably, the NACHT domains in the human and mouse NLRP3 “cage” assemblies do not contact each other.^50,71,76 None of the mutations in the NACHT domain nor those in other NLRP3 domains affect the protein–protein contacts in the inactive assembly. Similarly, the protein–protein interfaces in the active NLRP3 disk conformation are not affected by most of the mutations, excepting p.D211N and p.H213R (NACHT-NACHT interface) and p.E567G/K/Q, p.R777C, and p.R918Q (LRR-NEK7 interface).⁴² Thus, most of the mutations likely do not stabilize or destabilize the protein–protein interactions in either the active or inactive conformations.

The NLRP3 protein undergoes a large closed-to-open conformational change when transitioning between the inactive and active states (Figure 6A,B). This motion can be described as a rigid body hinge motion between the FISNA-N BD-HD1 N-terminal subdomains of the NACHT domain and the WH-HD2-LRR portions of the protein. Strikingly, the commonly mutated regions A, B, and C form a single surface on the NBD, consistent with the predictions we made 15 years ago,⁹ and interact in the inactive conformation with the commonly mutated regions F (WH subdomain) and G (HD2 subdomain) (Figure 6C). The interface delineated by the regions A–C, F and G in the inactive conformation is where the NLRP3 inhibitor MCC950/CRID3 binds.^50,71,121 This inhibitor is thought to function by stabilizing the inactive conformation by binding a pocket that is only present in the inactive conformation. This interface is lost in the active conformation, and many of the residues in regions A–C, F and G become more solvent exposed (Figure 6D). Similarly, the commonly mutated regions D and E are part of an interface of the HD1 domain with the HD2 and LRR domains that is present in the inactive conformation but undergoes substantial alteration in the active conformation (Figure 6C). Thus, a shared feature of each of the regions commonly mutated in autoinflammatory diseases (Figure 2) is that they lie at domain-domain interfaces that are lost or substantially changed in the conversion of inactive NLRP3 to active NLRP3 (Figure 6C,D).

To systematically explore all of the identified mutations and not just those at regions A–G, we identified domain–domain interface residues whose environment changed between the active and inactive conformation, using a local environmental change (LEC) score that we devised. This score is calculated by finding all atoms within a 5 Å cutoff in both conformations and calculating the average absolute difference between the distances in both conformations. As expected, the commonly mutated regions A–G were robustly identified by this analysis as well as some of the other mutated residues (Figure 6E). This analysis, however, indicated that there were some mutations associated with gain-of-function disease whose local environment was not substantially altered in the inactive-to-active conformational change (Figure 6E).

These data suggest the hypothesis that many of the mutations cause disease by disrupting the domain-domain interactions that stabilize the inactive state (Figure 6F). A shift in the equilibrium of the NLRP3 conformation to the active state predicts that the mutations would reduce or eliminate the need for a second activating signal such as potassium efflux. This hypothesis is consistent with multiple experimental results. First, CAPS mutations have reduced crosslinking to the MCC950/CRID3 inhibitor, which binds the inactive conformation.¹²² Second, in the absence of stimulation, the p.R260W, pA352V, and p.L353P mutant proteins form dispersed puncta that are similar to those formed by wild-type NLRP3 in the presence of stimulation.⁷⁰ Third, the p.R260W mutation did not require potassium efflux for activation,¹²³ but retained the requirement for NEK7 binding,⁷³ while other mutations such as p.L353P appear to be more dependent on calcium influx.¹¹⁷ Fourth, more missense mutations are known to cause severe disease than mild disease, which could be consistent with a mechanism involving disruption of the inactive conformation, instead of stabilization of the active conformation if this bias is not due to the increased likelihood of patients with severe phenotypes to undergo more extensive genetic analysis. In general, mutations giving rise to different disease severity (FCAS vs. MWS vs. NOMID) are not distinguished by their position on the NLRP3 protein; disease severity may be a consequence of how strongly individual mutations affect the NLRP3 conformational equilibria and affect the sensitivity to intracellular ion flux.

Some disease-causing mutations do not affect residues that directly lie at the interfaces that change during the conformational change (Figure 6E). These mutations may drive a gain-of-function phenotype through one of several mechanisms. Some of these mutations might destabilize the inactive state indirectly by affecting the local protein structure, which may include adjacent interface residues. Other mutations might locally disrupt the structure and recruit protein chaperones; the NLRP3-chaperone interactions have the potential to promote the active, open conformation due to steric collisions with NLRP3 domains in the inactive, closed conformation. Finally, the small number of mutations that lie at the protein–protein interfaces in the active conformation could function by directly stabilizing the active, signaling state.

Mice have been used for the study of immunity for decades, initially for in vivo experiments that cannot be performed in humans, and later to determine specific gene/protein function using recombinant technology allowing for the study of specific gene knockouts. However, since the beginning, disagreements have arisen between human and mouse immunologists as significant differences exist, particularly in lymphocyte biology, but also in innate immune responses.¹²⁵ Despite these differences in immune function and phenotype, mice remain extremely useful as models for human immune diseases, and this holds true for the study of NLRP3 and CAPS, with some caveats.

In 2006, several landmark papers were published by various groups^95,126,127 using Nlrp3 knockout mice. These studies revealed that NLRP3 was a unique and multifunctional protein with diverse activators including urate crystals, RNA, and toxins, with significant roles in several inflammatory and infectious diseases including gout, contact dermatitis, and Salmonella. These mice have continued to be studied over the last 17 years to expand our knowledge of the increasing number of roles for NLRP3 in the pathophysiology of diseases involving almost every organ. In fact, descriptions of conditions where NLRP3 does not play a role are the exception, rather than the rule.¹²⁸

To better understand the function of the LRR region, we developed a unique mouse model with deletion of only the C-terminal LRR domain of NLRP3 in collaboration with Regeneron Pharmaceuticals. We showed that the LRR domain is necessary for NLRP3 function in both in vivo models (i.e., gout) and in ex vivo responses to several known activators by comparing this LRR deletion mouse to a full Nlrp3 knockout.¹²⁹ In order to understand the role of NLRP3 in specific cells and tissues in different disease models, we also developed and published the first conditional Nlrp3 knockout mouse in collaboration with Ariel Feldstein, showing that loss of Nlrp3 expression in myeloid cells is crucial for the development of acute and chronic liver disease.¹³⁰ The availability of the conditional Nlrp3 knockout combined with increased accessibility of specific Cre mice has allowed for increased understanding of the role of NLRP3 in non-hematopoietic cells and in other immune cells outside of the myeloid compartment.¹³⁰

While deletion of Nlrp3 in mice established the role of NLRP3 in numerous mouse disease models, knockout mice did not provide direct information related to CAPS, where patients have gain-of-function mutations in NLRP3. Although we had access to CAPS patient cells and already had three effective therapies for patients, questions remained concerning disease mechanisms, particularly the response to cold in FCAS patients.¹³¹ Pre-clinical models to test novel therapies were also desirable. Therefore, in collaboration with Susannah Brydges and Dan Kastner, James Mueller and I generated three different conditional mutant Nlrp3 mouse models to cover the mild, moderate, and severe aspects of the CAPS spectrum and attempt to understand differences between the phenotypes observed with different mutations.¹¹⁴

The conditional nature of the mutations in these mice was fortuitous and allowed for a variety of useful experimental manipulations. Our initial approach was to recreate heterozygous mutations in all cells similar to most CAPS patients using a universal Cre, resulting in significant systemic inflammation, poor growth, and reduced survival in all of the mutant models. Laboratory evaluation showed neutrophilia in blood and elevation of many proinflammatory cytokines in serum including IL-1β, IL-18, IL-6, IL-8, and granulocyte-colony stimulating factor (G-CSF), but not TNF. Histologic analysis showed neutrophilia in skin, joints, conjunctiva, and the meninges, but not lungs, heart, kidneys, and gut consistent with the tissues known to be affected in CAPS. Neutrophilia was also observed in bone marrow, liver, and spleen, and muscle. Similar to CAPS patients, IL-1β and IL-6 protein expression was increased in affected skin.¹¹⁴ The primary difference between human and mouse CAPS was the overall increased severity of disease in these mice, with all of the mutations showing mortality ranging between birth and 6 weeks of life. However, the most unexpected observation was the reversal of the spectrum of severity in that the mice with the FCAS mutation (correlating to human p.L353P) were the most severely affected, and the mice with the NOMID mutation were the least severely affected, highlighting unexplained differences between humans and mice.^114,132 This curious pattern was extended with the later development of a mouse using the same construct and strategy but inserting a mutation with a more severe human NOMID phenotype. This mouse developed by Chiesa and colleagues was less severely affected than the NOMID mouse we generated, living long enough to observe the chronic sequela of systemic amyloidosis observed in some CAPS patients.¹³³ A similar reversed pattern of severity was also observed in our three mice bred to a tamoxifen-inducible Cre in which injecting these adult mice with tamoxifen resulted in systemic inflammation, rash, weight loss, and death.^134,135 While the mechanism of reversal in severity between humans and mice with mutations in NLRP3 is still unknown, the spectrum of severity in murine CAPS allow us to choose different models depending on the desired study or expected effect of a genetic or pharmacologic intervention.

To determine the role of specific immune cells in CAPS, we adopted two strategies. First, we bred our universal mutant mice to mice lacking specific cell types (i.e., T cells, Rag1 or mast cells, Kit^Wsh). The progeny of these crosses did not have altered phenotypes, suggesting that T cells and mast cells do not play significant roles in mouse CAPS. Second, we crossed mice whose NLRP3 mutant expression was Cre dependent with mice whose Cre expression was cell-specific, including lysozyme Cre (myeloid cells), Mcpt5 Cre (mast cells), and MRP8 Cre (neutrophils). Restriction of Nlrp3 expression to myeloid cells resulted in a similar phenotype to universal expression.¹¹⁴ This result suggested that myeloid cells are the primary drivers of CAPS phenotype and was consistent with observations in patients whose complete and typical CAPS phenotypes are due somatic mosaicism and expression of NLRP3 mutations primarily in myeloid cells. Similarly, breeding our Nlrp3 mutant mice to MRP8 Cre mice resulted in a phenotype similar to the lysozyme Cre mice suggesting an important role for neutrophils in mouse CAPS.^136,137 In contrast, breeding the mutant CAPS mouse to Mcpt5 Cre had no apparent phenotype indicating a minimal role for mast cells.¹³⁶ despite the urticarial nature of the CAPS rash and previous data suggesting a role for mast cells in another mouse model.¹³⁸

The discovery of NLRP3 as a platform for caspase 1 activation and the success of IL-1 targeted therapies in patients with CAPS pointed toward a major pathologic role for IL-1β, so it was natural that we would investigate the effect of blocking the IL-1 pathway downstream using genetic and pharmacologic approaches in murine models. Therefore, we bred our Nlrp3 mutant mice to IL-1β and IL-1R knockouts.^114,139 While there was some rescue, we still observed persistent rash and decreased growth and survival in the first few weeks of life in these mice. Interestingly, we observed some differences between IL-1β and IL-1R knockouts suggesting a possible role for IL-1α in mouse CAPS, which remains to be investigated. The MWS mutant IL-1β or IL-1R deficient mice that survived past 6 weeks had long-lasting scarring due to rash, but they had near normal growth and survival, more similar to human CAPS patients treated with IL-1 targeted therapy. We also treated mutant pups with a mouse rilonacept and showed partial rescue similar to the early rescue of the IL-1 targeted genetic approaches.¹¹⁴ We and others have also treated mutant adult inducible mice with recombinant IL-1RA (anakinra) with partial rescue of phenotype, but this was not as effective as an IL-1RA gene therapy in this model.¹⁴⁰

The identification of IL-1 independent mechanisms in the pathophysiology of murine CAPS directed our attention to other pro-inflammatory cytokines, beginning with IL-18, another inflammasome-associated cytokine requiring caspase 1 for cleavage and activation. Mutant mice treated with an IL-18 binding protein demonstrated significant improvement in rash, weight gain, survival, and liver pathology.¹⁴¹ Similarly, mutant Nlrp3 mice on an IL-18 or IL-18R deficient background revealed a strong phenotypic rescue in pups with complete resolution of rash, and normal growth and survival through the first 2 months of life in the MWS model. However, we did observe a later onset phenotype after 8 weeks of age characterized by weight loss, systemic and tissue inflammation, and ultimately reduced survival suggesting that at least in MWS mice, IL-18 drives the early CAPS phenotype and IL-1 drives the later CAPS phenotype. The logical next step was to generate an Nlrp3 mutant mouse on an IL-1β/IL-18 double knockout background to investigate a synergistic role for these cytokines, hypothesizing that the phenotypic rescue would be better than each cytokine alone. While our hypothesis was correct, to our surprise, we did identify IL-1/IL-18 independent mechanisms involved and demonstrated that these were at least partially driven by NLRP3-dependent pyroptotic cell death.¹³⁹

Using similar strategies, we also evaluated other pro-inflammatory cytokines besides those specifically regulated by NLRP3 including the downstream cytokines TNF, IL-17, and IL-6. Historically, treating CAPS patients with TNF targeted therapy did not have beneficial effects, but treating our Nlrp3 mutant mice with etanercept, a soluble TNF receptor, resulted in an unexpected rescue.¹⁴² Consistently, breeding the Nlrp3 mutant mice to TNF knockout mice also improved our previously reported primary outcomes in the pups including early tissue inflammation and fibrosis,^142,143 but similar to mutant IL-18 knockout mice, this rescue was short-lived as we observed the development of a late onset inflammatory phenotype. While IL-17 plays a major role in several inflammatory diseases and can be driven by IL-1β, we observed little change in the mouse CAPS phenotype when we bred our mutant mice on an IL-17a deficient background.¹⁴³ Similarly, IL-6, another well-known IL-1 driven pro-inflammatory cytokine had no obvious effect when deleted from the Nlrp3 mutant mice,¹⁴⁴ consistent with the lack of clinical effect when a patient with CAPS was treated with an IL-6 targeted therapy.¹⁴⁵

We have used similar genetic and pharmacologic approaches to tease apart upstream inflammasome pathways. Breeding our Nlrp3 mutant mice on an ASC-deficient background resulted in the only complete and long-lasting phenotypic rescue that we have identified to date.¹¹⁴ We have observed time-limited clinical improvement in our mice bred on caspase 1 knockout backgrounds, suggesting that several inflammasome independent pathways are driven by NLRP3. Finally, we and others have treated our mice with NLRP3 inhibitors demonstrating significant efficacy and providing the pre-clinical data for ongoing and planned clinical trials in CAPS patients.^97,146

The observation that cold exposure led to reproducible and easily observed inflammatory symptoms and laboratory changes in patients with FCAS led to an investigator-initiated trial in a small number of patients to determine whether pre-treatment with an anti-IL-1 therapeutic could prevent the NLRP3-specific cold-induced symptoms.¹¹⁶ At the time, the only approved drug in the IL-1 pathway was anakinra, a recombinant version of the naturally occurring IL-1R antagonist. This short acting injectable medicine was developed for the treatment of sepsis, but approved only for rheumatoid arthritis at the time. Self-appointed the “anakinra chicks,” three women with FCAS underwent purposeful cold exposure—45 min in a laboratory cold room—with and without pre-treatment with two doses of anakinra. Cold exposure prior to treatment led to the typical pattern of fever, rash, arthralgias, with rapid increases in WBC and serum IL-6 within hours. Symptoms failed to develop when patients were pre-treated for 24 h with anakinra. This somewhat unconventional approach demonstrating that IL-1 blockade could successfully reduce acute inflammatory symptoms in patients with gain-of-function NLRP3 mutations was similar to results observed in a more traditional study measuring chronic symptom and laboratory inflammation control in two patients with the more severe MWS.¹⁴⁷ These two proof of concept trials were the basis of a more complex and extremely successful clinical trial in patients with NOMID leading to ultimate approval by the FDA.¹⁴⁸

To date, three drugs targeting the IL-1 pathway are approved by the FDA and EMA for the treatment of CAPS spectrum disorders, namely anakinra, rilonacept, a recombinant IL-1R that binds to and inhibits IL-1α, IL-1β, and IL-1RA, and canakinumab, a human monoclonal antibody that binds specifically to IL-1β.^149,150 While each drug has different structure and mechanisms including the ability to block IL-1α and very different pharmacokinetics, they all have similar safety profiles and are well tolerated by patients. Decisions of which medicine to choose are often based on patient preference, availability, and costs/insurance limitations. Combined with their success in treating these rare disorders, the expansion of diseases implicating IL-1 in inflammatory disease pathogenesis has made this pathway a lucrative target for the pharmaceutical industry. Beyond inhibition of IL-1 and the IL-1R, development of other biologics and small molecule inhibitors targeting the different steps from NLRP3 assembly to IL-1 release is an active area, with CAPS patients often playing an integral part in the process (Table 3).

Despite the efficacy of IL-1 drugs in CAPS and other inflammatory disease, there has been a keen interest in other targets for a variety of reasons including finding oral available treatments and theoretical advantages of targeting specific mechanisms upstream of IL-1 and the IL-1R in order to preserve some IL-1 function through other inflammasomes. For example, specific inhibition of the NLRP3 pathway may reduce the risk of infections associated with more general IL-1 blockade.¹⁵¹ There was early interest in the blocking the enzymatic function of caspase 1 as a strategy to block IL-1 release resulting in inhibitors with reasonable specificity reaching later stages of development. We demonstrated that one of these compounds VX-765 was very effective at blocking ex vivo IL-1β release from FCAS patient cells.¹¹⁵ In vivo studies were initiated, but never completed nor reported. There has also been steady interest in the signaling pathways downstream of IL-1R such as MEK, IRAK4, and MK2 with initiation of pre-clinical studies in CAPS; however, the latter is the only published report.¹⁵² Other strategies to block the NLRP3 inflammasome pathway that are being studied and developed include ASC, NEK7, and gasdermin D but the largest focus has been on inhibiting NLRP3 directly beginning with the most prevalent NLRP3 targeting molecule initially called CRID3 (now known as MCC950).⁹⁷ While clinical development in this drug was halted years ago, it is a commonly used effective tool compound¹⁵³ and has been used to guide development of dozens of related and unrelated compounds currently in different stages of development. Many of these drugs have been studied in CAPS mice and patient cells,¹⁴⁶ and a few have been administered in vivo to CAPS patients.¹⁵⁴ In addition to small molecule inhibitors of NLRP3, we have also studied anti sense therapy as a novel approach to the treatment of CAPS and other NLRP3 driven diseases.¹⁵⁵ (Table 4).