Right place, right time: localisation and assembly of the NLRP3 inflammasome

Mitochondrial dysfunction leads to NLRP3 inflammasome activation

Positioning of NLRP3 following activation on MAMs/mitochondria is thought to enable immediate recognition of mitochondrial damage, which has been proposed as a central event downstream of various NLRP3-activating stimuli. These damage signals include mtROS, mitochondrial DNA (mtDNA) and cardiolipin (Figure 1)^26,36,37.

Cellular ROS was initially shown to activate the NLRP3 inflammasome in an NAPDH-dependent manner; however, further studies showed that ROS generated by the mitochondria is sufficient to lead to NLRP3 activation^26,38,39. Specific inhibition of mtROS significantly diminishes NLRP3 inflammasome activation⁴⁰. Yu et al. demonstrated that mitochondrial damage following NLRP3 inflammasome activation occurs in a caspase-1–dependent manner, resulting in the release of mtROS⁴¹. Caspase-1 was demonstrated to amplify this response by inhibiting mitophagy, the process of removal of damaged mitochondria, which then contributed to pyroptosis induction⁴¹. Conversely, ROS production has been shown to be dispensable for NLRP3 inflammasome activation, and the significance of K⁺ efflux has been highlighted⁴², thereby suggesting these as independent NLRP3 activation mechanisms. How increases in mtROS production activate NLRP3 is still unclear, although TXNIP has been implicated by various studies^43–45. TXNIP is a negative regulator of the antioxidant thioredoxin (TRX) and is thought to dissociate from TRX in the presence of ROS, allowing TXNIP to bind to NLRP3. In agreement, NLRP3 inflammasome activation is impaired in TXNIP^-/- mice⁴⁶. The precise mechanism remains unresolved. Furthermore, mutations in mtDNA-encoded cytochrome b gene in patients with fibromyalgia are associated with mitochondrial dysfunction and increased ROS production, leading to enhanced caspase-1 cleavage, and IL-1β and IL-18 secretion in fibroblasts and serum respectively⁴⁷. Increases in ROS have previously been implicated in other inflammatory disorders^48,49.

In addition to mtROS, the release of mtDNA—in particular, oxidised mtDNA—in response to mitochondrial dysfunction activates the NLRP3 inflammasome^36,50,51. NLRP3 inflammasome priming via TLRs has been shown to induce mtROS production and, more recently, the production of mtDNA^52,53. The latter is dependent on MyD88/TRIF- and IRF-1-mediated upregulation of mitochondrial deoxyribonucleotide kinase, CMPK2, an enzyme with rate-limiting activity in mtDNA synthesis⁵³. The newly synthesised mtDNA (made during priming) was found to be necessary to generate oxidised mtDNA fragments when cells were subsequently exposed to an NLRP3-activating signal (Figure 1)⁵³. Of note, oxidised DNA from any cellular source can activate the NLRP3 inflammasome. In fact, mtDNA contributes to inflammation in diseases such as atherosclerosis and inflammatory kidney disorders, and the inhibition of its release by exposure to agents that maintain mitochondrial integrity prevents NLRP3 inflammasome activation^54–59. Contrary to the above, TLR-dependent priming also restricts inflammasome activation through efficient removal of damaged mitochondria. TLR-activated NF-κB pathway induces prolonged accumulation of the autophagic receptor p62/SQSTM1, which targets damaged mitochondria for clearance by mitophagy⁶⁰, thereby preventing NLRP3 over-activation⁶¹. Whether inhibition of mtDNA release or maintenance of mitochondrial integrity (or both) will translate into novel therapeutics for disorders in which NLRP3 inflammasome is dysregulated is yet to be seen.

A small number of studies have implicated mtDNA in the activation of other inflammasome complexes. Mitochondrial damage by Pseudomonas aeruginosa leads to the release of ROS and mtDNA; the latter was shown to bind and activate the NLRC4 inflammasome. In agreement, abolition of mtDNA by DNase I treatment resulted in reduced caspase-1 activation and IL-1β secretion following P. aeruginosa infection⁶². However, caspase-1 activation was not completely abolished by DNase I in Nlrc4^−/− macrophages, suggesting roles for other inflammasomes, and the authors attributed this to AIM2 inflammasome activation. Although NLRP3 is not thought to be involved during P. aeruginosa infection, a role for NLRP3 was not experimentally ruled out^62,63. Dang et al. also demonstrated that mtDNA, released as a result of loss of mitochondrial integrity by elevated cellular cholesterol levels, can activate the AIM2 inflammasome in macrophages lacking cholesterol-25-hydroxylase. In wild-type cells, production of 25-hydroxycholesterol (25-HC) by cholesterol-25-hydroxylase repressed cholesterol biosynthesis, thereby maintaining mitochondrial integrity⁶⁴. However, only a small redundant role for NLRP3 was found in the above study, although mtDNA is known to activate this inflammasome³⁶. As the AIM2 inflammasome can recognise both foreign and self-DNA, it is likely that it can also recognise mtDNA; however, further studies are required to confirm these findings. Whether mtDNA has the ability to directly activate multiple inflammasomes⁶⁵ or whether upstream stimuli direct differential activation of NLRP3 and AIM2 inflammasomes requires further clarification.

NLRP3 is recruited to and binds mitochondrial cardiolipin

Cardiolipin, a phospholipid localised to the inner mitochondrial membrane (IMM), has also been linked to NLRP3 inflammasome activation. Upon NLRP3 activation or in the presence of stimuli that destabilise mitochondria, cardiolipin redistributes to the outer leaflet of the mitochondrial membrane (OMM), where NLRP3 anchors through its leucine-rich repeat (LRR) domain^37,66 (Figure 1). Depletion of cardiolipin or abrogation of cardiolipin synthesis by culturing cells in the presence of palmitate (C16:0), a saturated long-chain fatty acid, blunted inflammasome activation in J774A.1 macrophages³⁷. Similarly, small interfering RNA (siRNA) knock-down of cardiolipin synthase reduced NLRP3 activation³⁷. Corroborating these findings, activation of the NLRP3 inflammasome, by HIV reverse transcriptase inhibitor abacavir, was abrogated following inhibition of cardiolipin synthase-1³⁷. Furthermore, cardiolipin-dependent NLRP3 inflammasome activation by the antibiotic linezolid, which induces mitochondrial toxicity, was ROS-independent³⁷. Thus, mitochondrial dysfunction, which induces ROS under certain conditions, and not ROS itself, may be responsible for NLRP3 inflammasome activation. Additionally, recent work from the group established that NLRP3 is recruited to the mitochondria at the priming stage and binds to cardiolipin on the OMM⁶⁷. Interestingly, the movement of cardiolipin to the OMM was found to be dependent on ROS produced during the NLRP3 priming step. Surprisingly, caspase-1 also binds cardiolipin at the mitochondria, while ASC was recruited only following NLRP3 activation⁶⁷. The binding of both caspase-1 and NLRP3 to mitochondrial cardiolipin validates mitochondria as a scaffold for inflammasome activation, thereby amplifying its activation by bringing molecules together (Figure 1). In line with this, other immune signalling pathways initiated at the mitochondria, including the MAVS complex, occur in a similar manner⁶⁸. Additionally, these studies identify a role for priming beyond the upregulation of NLRP3 and pro-IL-1β.

Endoplasmic reticulum stress triggers caspase-1 activation

NLRP3 has been shown to reside in the ER prior to its activation, and ER dysfunction has been reported to trigger inflammasome activation (Figure 1). The ER is a membrane-bound organelle that is critical for protein folding, assembly and modification in addition to being the site for lipid synthesis and Ca²⁺ homeostasis. Cellular stress leading to accumulation of misfolded or unfolded proteins induces an unfolded protein response (UPR) aimed at restoring ER homeostasis by regulating levels of transcription, translation, and protein folding⁶⁹. The UPR response activates several stress sensors located in the ER, including inositol-requiring protein 1 alpha (IRE1α), protein kinase R-like endoplasmic reticulum kinase (PERK), and activating transcription factor 6 (ATF6). These sensors subsequently regulate downstream cytosolic effectors and signalling pathways, including proteasomal degradation, autophagy, and antioxidant defence mechanisms. Prolonged ER stress induces an inflammatory response and cell damage, thereby triggering NLRP3 inflammasome activation^18,70–73. An initial study suggested that ER stress activated the NLRP3 inflammasome via a UPR-independent pathway, as short hairpin RNA (shRNA) knock-down of either Ire1α or Perk or macrophages obtained from Atf6α^−/− mice released comparable IL-1β following NLRP3 activation⁷⁴. However, the authors also noted that not all NLRP3 activators induced ER stress, indicating that this may be just one mechanism by which the NLRP3 inflammasome is activated. In agreement, Bronner et al. found that ER stress and IRE1α were not induced in mouse BMDMs with the canonical NLRP3 activator ATP, although they were induced during Brucella abortus infection¹⁸. During B. abortus infection, IRE1α-initiated mtROS production recruited NLRP3 to mitochondria. Mechanistically, IRE1α activated TXNIP and caspase-2, leading to truncation and activation of the mitochondrial protein Bid, which resulted in mitochondrial damage and release of mitochondrial damage-associated molecular patterns (DAMPs) that promoted NLRP3 activation¹⁸. Surprisingly, NLRP3 was required to activate caspase-2/Bid upstream of mitochondrial damage, suggesting a role for NLRP3 in initiating mitochondrial damage by a feed-forward loop. IRE1α and its activation of TXNIP have been implicated in NLRP3 inflammasome activation in other studies^19,75–78. Similarly, inhibition of the ER stress sensor PERK was shown to reduce caspase-1 activation and IL-1β secretion in J744.1 macrophages, although how PERK inhibition decreases NLRP3 activation was not determined⁷⁹. Targeting IRE1α to dampen ER stress–induced NLRP3 inflammasome activation has shown benefits in a wide variety of inflammatory conditions^{75,76,80–85}. These studies again suggest mitochondrial damage as the downstream mechanism by which ER stress initiates NLRP3 inflammasome formation. Additionally, this work may indicate that the ER is the site where NLRP3 activation is initiated before the inflammasome is assembled in the cytosol or at the mitochondria/MAMs.

In addition to being implicated in the activation of the NLRP3 inflammasome, ER stress is thought to play a role in the activation of the NLRP1 inflammasome. NLRP1 expression in HeLa cells is upregulated upon induction of ER stress by tunicamycin and thapsigargin, which inhibit the N-linked glycosylation of proteins and sarcoplasmic/endoplasmic reticulum Ca²⁺-ATPase (SERCA) respectively. Upregulation of Nlrp1 involved IRE1α and PERK, and siRNA knock-down of either Ire1α or Perk abrogated increase in NLRP1 expression⁸⁶. Consistent with this, studies have shown a link between ER stress and NLRP1 upregulation in leukaemia and cardiovascular injury models^87,88. Thus, as suggested by studies discussed below, ER seems to be a key subcellular site to regulate inflammasome activation.

Endoplasmic reticulum calcium homeostasis in inflammasome activation

The ER is also the site at which Ca²⁺ homeostasis occurs, and Ca²⁺ mobilisation has been implicated in NLRP3 inflammasome activation. Blockade of the ER-resident calcium channel IP3R led to reduced NLRP3 inflammasome activation in mouse macrophages^89,90. Other studies have challenged these claims, showing no role for Ca²⁺ and indicating that K⁺ efflux is more important^42,91. It can be expected that, as with other mechanisms of NLRP3 activation, Ca²⁺ mobilisation from the ER activates the NLRP3 inflammasomes only under specific conditions. Similarly, it is possible that Ca²⁺ mobilisation precedes ER stress or is a consequence of ER dysfunction and therefore happens to occur alongside ER stress–induced NLRP3 activation.

Endoplasmic reticulum cholesterol levels regulate NLRP3 activation

The ER not only is involved in lipid synthesis but also is the site at which cholesterol levels within the cell are sensed and regulated. Cellular cholesterol homeostasis is achieved by maintaining an equilibrium between de novo synthesis at the ER, exogenous cholesterol uptake in the form of low-density lipoprotein (LDL), and cholesterol efflux programs. LDL is endocytosed by LDL receptor and subsequently trafficked through the endosomal system to other subcellular compartments, including the ER, plasma membrane, and Golgi. Blockade of cholesterol efflux from the lysosome was recently demonstrated to inhibit NLRP3 inflammasome activation in mouse macrophages, an effect attributed to decreased cholesterol within the ER²¹. Similarly, inhibition of cholesterol biosynthesis in the ER by exposing cells grown in lipoprotein-deficient media to statins dampened NLRP3 inflammasome activation²¹. Additionally, the ER-localised cholesterol-sensing transcription factor sterol regulatory element-binding protein 2 (SREBP2) has been implicated in NLRP3 inflammasome activation⁹². The cholesterol metabolite, 25-HC, has also been shown to activate the NLRP3 inflammasome by induction of K⁺ efflux, mtROS production, and activation of the cholesterol transcriptional regulator liver X receptors (LXRs)⁹³. Thus, either cholesterol may be involved in directly influencing inflammasome complex formation or ER cholesterol content may support NLRP3 in achieving the necessary conformation. Regardless, these studies emphasise a critical role for ER in inflammasome activation.

NLRP3 translocation near the Golgi is required for inflammasome activation

Although roles for mitochondria and ER have been suggested by many labs in the last decade, a few recent studies have also implicated Golgi in inflammasome activation. The Golgi apparatus is important for the modification and transport of proteins and lipids within the cell. Zhang et al. first implicated the Golgi in NLRP3 inflammasome activation. Following treatment with NLRP3 activators, the levels of diacyl glycerol (DAG) at the Golgi increased and this coincided with the localisation of NLRP3 on MAMs adjacent to the Golgi²⁰. Disruption of Golgi integrity with brefeldin A reduced caspase-1 activation, IL-1β secretion and ASC speck formation following NLRP3 inflammasome activation, a result which was corroborated in a subsequent study⁹⁴. This effect was attributed to protein kinase D (PKD), a DAG effector, which phosphorylated NLRP3 at a conserved residue (Ser293) in its NBD domain, allowing its release from MAMs to form an inflammasome complex within the cytosol (Figure 2). Of note, PKD phosphorylated after NLRP3 self-oligomerised and PKD inactivation both retained self-oligomerised NLRP3 at MAMs and reduced NLRP3 inflammasome activation. Interestingly, PKD inhibition in cells from patients with cryopyrin-associated periodic syndrome (who exhibit spontaneous NLRP3 oligomerisation) also showed reduced NLRP3 activation²⁰. This added to previous studies that indicate that post-translational modifications can both positively and negatively affect NLRP3 activation^95,96. Furthermore, this study highlights cross-talk between Golgi and MAMs during NLRP3 inflammasome activation by means of DAG accumulation. In support of this, a recent study demonstrated recruitment of NLRP3 to the trans-Golgi network (TGN) through ionic bonding between NLRP3 and phosphatidylinositol-4-phosphate (PtdIns4P) in the membrane of the Golgi (Figure 2)²². The authors found that the TGN dispersed and formed vesicles in the perinuclear space that co-localised with overexpressed NLRP3 following inflammasome activation in HeLa cells. Similar results were shown in primary BMDMs lacking ASC; therefore, whether the full assembly of the inflammasome complex occurs at the TGN was not addressed. However, in that study, in contrast to previous studies, NLRP3 was not found to be associated with mitochondria, and instead the TGN, rather than the mitochondria, provided the scaffold for NLRP3 inflammasome activation²². Furthermore, translocation of NLRP3 from the ER to the Golgi, this time requiring SREBP and SREBP cleavage-activating protein (SCAP), has also been reported in primary mouse macrophages⁹². When cholesterol levels within the cell are low, the SCAP-SREBP2 complex translocates from the ER to the Golgi, where SREBP2 is cleaved into its active form by site-1 protease (S1P) and site-2 protease (S2P), and translocates to the nucleus to transcribe genes involved in cholesterol biosynthesis and uptake⁹⁷. The latter study demonstrated that NLRP3 partially co-localises with SCAP-SREBP2 and translocates to the Golgi, and that inhibition of SCAP-SREBP2 translocation abrogated NLRP3 inflammasome activation (Figure 2). Subcellular fractionation studies showed NLRP3 in the Golgi as well as in the mitochondrial fraction with SCAP. This correlates with findings from Zhang et al., who suggested that NLRP3 activation occurs in close proximity to both the Golgi and the mitochondria²⁰. Whether multiple mechanisms can independently initiate NLRP3 recruitment to the Golgi requires further work and so does the role of NLRP3 phosphorylation in inflammasome activation.

Figure 2. Emerging role of the Golgi apparatus in NLRP3 inflammasome activation.

NLRP3 localised at the endoplasmic reticulum has been shown to translocate to the Golgi via binding to the SCAP-SREBP2 complex, which regulates cholesterol homeostasis. SREBP2 is cleaved at the Golgi by S1P and S2P to yield the active nuclear form of SREBP2, which then translocates to the nucleus and upregulates genes involved in cholesterol synthesis. NLRP3 has also been shown to be recruited to the TGN through ionic bonds formed with PtdIns4P. The Golgi then acts as a scaffold for NLRP3 assembly. Finally, DAG accumulation at the Golgi recruits and activates PKD, which is required for NLRP3 phosphorylation at the mitochondria, leading to its release and activation at mitochondrial-associated membranes. ASC, apoptosis-associated speck-like protein containing a CARD; DAG, diacylglycerol; IL-1β, interleukin-1 beta; IL-18, interleukin-18; PKD, protein kinase D; PtdIns4p, phosphatidylinositol-4-phosphate; S1P, site-1 protease; S2P, site-2 protease; SREBP2, sterol regulatory element-binding protein 2; TGN, trans-Golgi network.