|Home | About | Journals | Submit | Contact Us | Français|
A hallmark of inflammasome activation is the ASC speck, a micrometre-sized structure formed by the inflammasome adaptor protein ASC (apoptosis-associated speck-like protein containing a CARD), which consists of a pyrin domain (PYD) and a caspase recruitment domain (CARD). Here we show that assembly of the ASC speck involves oligomerization of ASCPYD into filaments and cross-linking of these filaments by ASCCARD. ASC mutants with a non-functional CARD only assemble filaments but not specks, and moreover disrupt endogenous specks in primary macrophages. Systematic site-directed mutagenesis of ASCPYD is used to identify oligomerization-deficient ASC mutants and demonstrate that ASC speck formation is required for efficient processing of IL-1β, but dispensable for gasdermin-D cleavage and pyroptosis induction. Our results suggest that the oligomerization of ASC creates a multitude of potential caspase-1 activation sites, thus serving as a signal amplification mechanism for inflammasome-mediated cytokine production.
Detection of pathogens by the innate immune system relies on germline-encoded pattern recognition receptors (PRRs), which recognize a variety of pathogen-derived molecules, known as pathogen-associated molecular patterns (PAMPs), and host-derived danger signals, known as danger-associated molecular patterns1. Although most PRRs initiate transcriptional responses, such as the expression of cytokines2,3, a subset of cytosolic PRRs promote the assembly of inflammasome complexes and subsequent activation of the cysteine protease caspase-1 (refs 4, 5). To date, only members of the NOD-like receptor (NLR) family, the PYHIN protein family and PYRIN were shown to assemble inflammasomes in response to different cytosolic PAMPs or danger-associated molecular patterns5,6. A unifying feature of all these receptors is either a pyrin domain (PYD) or a caspase recruitment domain (CARD), both of which belong to the death-fold domain superfamily. On the basis of the presence of these domains, the receptors can be classified as PYD-containing receptors (NLRP3: NLR family, PYD containing 3; AIM2: absent in myeloma 2; pyrin) or CARD-containing receptors (NLRC4: NLR family, CARD containing 4; NLRP1: NLR family, PYD containing 1). After receptor activation and oligomerization, these domains recruit the adaptor protein ASC (apoptosis-associated speck-like protein containing a CARD) and pro-caspase-1 into the complex through homotypic domain–domain interactions. Within the inflammasome, pro-caspase-1 is activated by dimerization and auto-proteolysis, and the proteolytically active hetero-tetramer is released7. Several substrates of caspase-1 have been identified, among them the pro-forms of the cytokines interleukin (IL)-1β and IL-18 (refs 7, 8, 9). Another consequence of inflammasome activation is the induction of a pro-inflammatory cell death called pyroptosis10. Pyroptosis is driven by the amino-terminal cleavage fragment of gasdermin-D, a protein cleaved by caspase-1, and results in the lysis of the host cell and subsequent release of cytoplasmic content, among them processed IL-1β and IL-18 (refs 11, 12).
Recent studies have started to elucidate the role of the inflammasome adaptor protein ASC13,14,15, a bipartite protein composed of a PYD and a CARD, also known as PYCARD16,17. ASC has previously been regarded as a simple adaptor protein that links PYD-containing receptors to the CARD-containing caspase-1, via homotypic PYD–PYD and CARD–CARD interactions18. However, its role appears to be more complex, as on receptor activation ASC also forms the so-called ASC speck, a macromolecular form of the inflammasome with a diameter of ~1–2μm19,20. Formation of the ASC speck is independent of caspase-1 activity, but requires the oligomerization of ASC into large insoluble aggregates20,21. These ASC aggregates are stable and have even been shown to be released into the extracellular space, after pyroptosis induction, where they can trigger prolonged inflammasome activation in phagocytic cells22,23. Analyses of stimulated emission depletion microscopy images and electron micrographs suggest an irregular, filamentous shape22,23. Consistently, we and others have shown that ASC oligomerizes into long helical filaments via its PYD13,14,15. Despite the advances regarding the atomic structure of the ASC filament, whether the ASC speck is the functional unit of the inflammasome is unclear. Several inflammasome receptors (for example, NLRC4 and mouse Nlrp1b) contain a CARD instead of a PYD and are able to recruit caspase-1 even in the absence of ASC. Indeed, we and others have shown that ASC is not required for cell death induction by these receptors24,25,26,27,28. Nonetheless, activation of CARD-containing receptors results in ASC speck formation in wild-type (WT) cells, but how their CARD initiates ASC filaments is unknown. Furthermore, it is unclear whether ASC oligomerization itself is required for inflammasome signalling. Previous mutagenesis studies of ASC could not address the role of ASC oligomerization in signalling, as they were based on overexpression of ASC mutants in HEK293 cells, which is prone to artefact generation and lacks the means to study the effects on downstream signalling13,29,30.
Here we use retroviral expression of ASC mutants in Asc-deficient immortalized mouse macrophages, to address the mechanism of ASC speck formation and the functional relevance of ASC filament formation for inflammasome signalling. Our results reveal an architectural role for the ASCCARD, showing that it is required to link individual ASC filaments towards forming the dense ASC speck. Furthermore, we show that ASC bridging molecules are necessary to allow CARD-containing receptors the initiation of ASC filament formation. Finally, we use site-directed mutagenesis of the ASCPYD–ASCPYD interaction interfaces to identify mutations that disrupt ASC filament formation without affecting the interaction to receptor PYDs. Surprisingly, such ASC mutants are still able to initiate caspase-1-dependent gasdermin-D maturation and subsequent cell death, but lack the ability to form filaments, to assemble ASC specks and to process IL-1β, thus uncoupling the two major downstream signalling pathways. In conclusion, our data show that the ASC speck has a function in inflammasome signalling and support a model in which the rapid oligomerization of ASC via its PYD creates a multitude of potential caspase-1 activation sites, thus serving as a signal amplification mechanism for inflammasome signalling.
Although ASC is the main structural component of the ASC speck20,24, conflicting reports have implicated either the ASCPYD or the ASCCARD in speck assembly13,31,32. To address the role of the ASCPYD and the ASCCARD in speck formation and downstream signalling, we transduced immortalized Asc−/− murine bone marrow-derived macrophages (BMDMs) with fluorescently tagged full-length ASC (ASCFL) or ASCPYD and ASCCARD alone (Fig. 1a). As overexpression of ASCFL can result in aggregation even in the absence of activated receptor19, we selected individual clonal lines that did not display any autoactivation when unstimulated (Supplementary Fig. 1a). To test the functionality of these constructs, cells were primed with lipopolysaccharide (LPS) and transfected with the synthetic DNA analogue poly(deoxyadenylic-deoxythymidylic) acid (poly(dA:dT)), an activator of AIM2 (refs 33, 34). mCherry-tagged ASCFL, expressed in Asc−/− cells, promoted pyroptosis (as measured by the release of lactate dehydrogenase (LDH)) and IL-1β release (measured by enzyme-linked immunosorbent assay (ELISA)), to levels seen in immortalized WT macrophage controls (Fig. 1b). We also observed comparable levels of IL-1β/-18 and caspase-1 processing and release by western blotting (Supplementary Fig. 1b). In contrast, neither the ASCPYD nor ASCCARD restored cell death, IL-1β/-18 secretion or caspase-1 cleavage on AIM2 activation (Fig. 1b and Supplementary Fig. 1b). Similarly, expression of the individual domains did not restore inflammasome signalling in response to extracellular ATP, an NLRP3 activator35 (Supplementary Fig. 1c,d).
Heterologous expression of ASCPYD and ASCCARD or CARDs of mitochondrial antiviral-signalling protein (MAVS), retinoic acid-inducible gene 1 (RIG-I) and pro-caspase-1 results in the formation of filaments of varying length in different cell types and in vitro13,36,37,38,39,40. However, it is possible that such ASCPYD or ASCCARD filaments are overexpression artefacts, as ASCFL normally forms a distinct dense speck structure20. To assess the ability of ASCPYD and ASCCARD to form macromolecular assemblies, we examined the above described cell lines by microscopy, following inflammasome stimulation. As expected, activation of AIM2 or NLRP3 induced the formation of perinuclear specks in cells expressing mCherry-tagged ASCFL (Fig. 1c and Supplementary Fig. 1e,f). In line with the reported ability of ASC to form filaments13,15, specks formed by ASCFL had a dense core with emanating filaments (Fig. 1c). Instead of forming similarly dense specks, the ASCPYD assembled into filaments on the engagement of NLRP3 or AIM2 (Fig. 1c). Of note, filaments formed after NLRP3 or AIM2 activation differed in their appearance, possibly due to distinct modes of how NLR and PYHIN family members assemble complexes and initiate ASC filaments13,41. Remarkably, these filaments had varying lengths and widths, with some appearing clearly thicker than 90Å, the reported diameter of an ASCPYD filament13,15. Unlike reported before, we observed that the ASCCARD was not able to form macromolecular assemblies (Fig. 1c and Supplementary Fig. 1e,f). Taken together, these data suggest that although both domains of ASC are necessary for signalling, only the ASCPYD forms filaments on inflammasome activation. This is consistent with reports by us and others showing that full-length ASC forms helical filaments in vitro via the ASCPYD (refs 13, 15), whereas the ASCCARD is exposed on the surface and not involved in ASC filament formation15. As straight filaments are formed by ASCPYD and compact specks are formed by ASCFL, our data also suggest that the ASCCARD must contribute significantly to the macroscopic structure of the ASC speck.
To test whether ASCCARD is required to form ASC specks, we generated immortalized Asc−/− macrophage lines expressing ASCD130R or ASCD134R (Fig. 2a and Supplementary Fig. 2a), two mutations in surface-exposed residues of the ASCCARD that abrogate interaction of the ASCCARD with the pro-caspase-1CARD (ref. 31). As expected, Asc−/− macrophages expressing ASCD130R or ASCD134R did not activate caspase-1, release mature IL-1β or undergo pyroptosis in response to DNA transfection or ATP treatment (Fig. 2b and Supplementary Fig. 2b–d). ASCD130R and ASCD134R formed long filaments rather than dense ASC specks in response to AIM2 and NLRP3 activation (Fig. 2c,d and Supplementary Fig. 2e,f), similar to the ASCPYD alone, thus supporting our hypothesis that a functional ASCCARD is required to assemble ASCPYD filaments into a speck. To confirm that these mutations not only disrupt ASCCARD–pro-caspase-1CARD interactions but also ASCCARD–ASCCARD interactions, we expressed enhanced green fluorescent protein (eGFP)-tagged ASCCARD in the presence of reconstituted inflammasomes, formed by co-transfection of AIM2 with mCherry-tagged ASCFL, ASCPYD, ASCD130R or ASCD134R in HEK293T cells. Co-localization and co-immunoprecipitation assays demonstrated that GFP-ASCCARD was only able to interact with WT ASC and not with ASCD130R or ASCD134R, confirming that the two mutations blocked ASCCARD–ASCCARD interaction (Supplementary Fig. 2g,h).
If the ASCCARD organizes ASC filaments into dense specks, we speculated that increasing levels of ASCCARD mutants would disrupt specks formed by WT ASC, resulting in larger but less dense ASC specks. Therefore, we retrovirally transduced various ratios of mCherry-tagged WT and mCherry-tagged ASCD130R into primary WT C57BL/6 macrophages (containing endogenous ASC) and measured median speck size on the activation of AIM2 by poly(dA:dT) transfection or NLRP3 by Nigericin, a pore-forming toxin35. Indeed, increasing levels of ASCD130R resulted in a significant increase in ASC speck size (Fig. 2e,f). As expected, microscopy analysis of these large ASC specks showed that they were also less dense and had a filamentous nature (Fig. 2g). Taken together, these results support a model in which the ASC speck observed in WT cells is composed of individual ASC filaments that are formed via their PYDs and cross-linked to each other via their CARDs (Supplementary Fig. 3a).
Most inflammasome-forming receptors contain a PYD and initiate ASC filaments directly by homotypic PYD–PYD interactions. Notable exceptions are NLRC4 and mouse Nlrp1b, which contain a CARD and thus might recruit and activate caspase-1 directly42,43. Nevertheless, ASC promotes caspase-1 processing and efficient IL-1β/-18 release after NLRC4 and Nlrp1b activation26,44. ASC specks can be observed after NLRC4 activation, but how CARD-containing receptors initiate oligomerization of ASC is unknown. Theoretically, NLRC4 could initiate ASC oligomerization either by (i) NLRC4CARD–ASCCARD interaction and ASCCARD oligomerization, (ii) heterotypic interaction between related death-fold domains, that is, by a non-canonical NLRC4CARD–ASCPYD interaction or (iii) via bridging ASC molecules that would be linked with NLRC4 via CARD–CARD interactions and provide their free PYDs as seeds for ASCPYD filament formation. To determine which domain of ASC was necessary for ASC speck formation after NLRC4 activation, we infected the above described cell lines harbouring the single domains of ASC with log-phase Salmonella enterica serovar Typhimurium SL1344 (Salmonella Typhimurium), a robust activator of NLRC4 (refs 45, 46). Consistent with published reports24,27,47, immortalized Asc−/− BMDMs still induced cell death, whereas efficient IL-1β processing and release required ASC (Fig. 3a). Expression of ASCFL in Asc−/− cells restored IL-1β release, but neither the ASCPYD nor ASCCARD alone were able to functionally restore inflammasome activation (Fig. 3a). We next analysed whether the ASCPYD or ASCCARD would form specks or filaments after NLRC4 activation by microscopy. Although ASCFL formed specks, ASCPYD alone did not initiate filaments following activation of NLRC4 (Fig. 3b,c), which was in contrast to what we had observed for PYD-containing receptors (Fig. 1c). In addition, cells expressing ASCCARD alone did not form any specks or macromolecular filaments, again suggesting that the ASCCARD in the endogenous setting is not able to form filaments (Fig. 3b,c). These results exclude the possibilities that ASC specks form via CARD oligomerization after NLRC4 activation, or that a heterotypic NLRC4CARD–ASCPYD interaction initiates ASC filament formation. Instead, they suggest that one or several bridging molecules of ASC are required to initiate ASC filament formation after NLRC4 activation, and that in these conditions ASC oligomerization into filaments and specks also proceeds via the ASCPYD.
We also examined cells expressing ASCD130R and ASCD134R mutants, which are defective for CARD–CARD interactions (Supplementary Fig. 2g,h (ref. 31)). As expected, expression of these mutants in Asc−/− macrophages did not complement the release of mature IL-1β after NLRC4 activation (Fig. 3d). Furthermore, we did not observe the formation of filaments or specks in cells expressing ASCD130R or ASCD134R (Fig. 3e), confirming the requirement for a functional ASCCARD for ASC oligomerization by NLRC4. Finally, we also tested our notion that the ASCCARD is required to condense ASC filaments into a speck in the context of NLRC4 activation. Primary WT C57BL/6 BMDMs (containing endogenous ASC, thus enabling initiation of ASC oligomerization) were retrovirally transduced with varying ratios of mCherry-tagged ASCFL or ASCD130R and the speck diameter was measured on S. Typhimurium infection (Fig. 3f). The increase in speck size concurring with increasing concentrations of the ASCD130R mutant indicated that also in the NLRC4 inflammasome the ASCCARD condenses the ASC filaments into the speck. In conclusion, our results suggested a triple functional role for ASCCARD in the NLRC4 inflammasome: (1) mediating the interaction with pro-caspase-1 (refs 17, 31), (2) condensing ASCPYD filaments into the speck and (3) initiating ASC oligomerization through bridging ASC molecules (Supplementary Fig. 3b).
Having elucidated the general architecture of the ASC speck, we next investigated whether the ASC speck constitutes the active, signalling-competent inflammasome. Higher-order signalosomes were reported for other innate immune signalling pathways (Toll-like receptors, receptor-interacting serine/threonine-protein kinase 2 (RIP2K) and MAVS) and could promote signal amplification and digital all-or-nothing responses48, leading us to hypothesize that ASC oligomerization and filament formation might have a similar function for inflammasome signalling.
Within the ASCPYD filaments that form the backbone of ASC specks in human and murine cells (Supplementary Fig. 4a (refs 13, 15)), individual ASCPYDs interact with other six adjacent ASCPYDs through three asymmetric interfaces, types I–III (Fig. 4a). Mutations targeting these interfaces can abrogate ASC filament formation in vitro or in cell overexpression systems13,29,49. As these mutations also most probably abrogate the interaction between the receptorPYD and ASCPYD, and/or were only tested in an artificial overexpression system, a definite confirmation that ASC filament formation is required for inflammasome signalling is still missing50. Identification of mutations that abrogate ASC oligomerization but leave ASC–receptor interaction intact allowed us to characterize the role of ASC filaments in inflammasome signalling. Therefore, we generated a number of clonal immortalized Asc−/− macrophage lines expressing different mCherry-tagged ASC mutants targeting all three interaction interfaces (Fig. 4a and Supplementary Fig. 4b–d). Next, we assessed the effects of these mutations on ASC speck/filament formation on AIM2 activation by poly(dA:dT) transfection. Mutations targeting interface I (that is, K21A, K26A, D48N and D51R) completely abrogated the filament formation as determined by microscopy (Fig. 4b upper panel and Supplementary Fig. 4e). This is consistent with interface I being the most extensive interface and required for the propagation of single layers of the filament13,15. In contrast, mutations in interfaces II and III displayed a larger variability, ranging from no defect in ASC speck formation (Y36A, E62A (interface II) or P40A (interface III)), intermediate phenotypes (M76A (interface II) or L15A (interface III)) to complete abrogation in ASC speck formation (Y59A, Q79E, E80R (interface II) or E13R and R41E (interface III)) (Fig. 4c,d upper panels and Supplementary Fig. 4e). Taken together, these data suggest that the formation of ASC specks requires all three interfaces of ASCPYD.
We next tested how these mutations affect downstream inflammasome signalling by measuring IL-1β secretion (Fig. 4b–d middle panels) and pyroptosis (Fig. 4b–d lower panels). In line with the loss of ASC speck formation, interface I proved to be essential for induction of downstream inflammasome signalling (Fig. 4b). Macrophages expressing mutants in interface III of ASCPYD displayed similar phenotypes as interface I mutations: the ability to form ASC specks correlated with both IL-1β secretion and pyroptosis, that is, mutants either lost all inflammasome signalling (E13R and R41E), were partially affected (L15A) or not affected at all (P40A) (Fig. 4d). These results show that interface III is important for induction of downstream inflammasome signalling, albeit some mutations are tolerated.
Mutants in interface II displayed a differential signalling phenotype (Fig. 4c). Although some mutations had no or only a very small effect on inflammasome signalling (Y36A, E62A and M76A) or abrogated inflammasome signalling altogether (Q79E), we identified two mutations (Y59A and E80R) that retained the ability to induce cell death, while losing the ability to release mature IL-1β and to form ASC specks in response to AIM2 stimulation (Fig. 4c middle and lower panels, highlighted in grey). Of note, the ELISA assay displays a higher sensitivity for the mature form of IL-1β than the pro-form, thus pro-IL-1β, which is potentially released into the supernatant by pyroptosis, is poorly detected (Supplementary Fig. 5). To test whether the phenotype of the different mutants is specific to the AIM2 inflammasome, we analysed ASC speck formation, secretion of mature IL-1β and pyroptosis in response to NLRP3 stimulation by ATP treatment. Asc−/− macrophages expressing the ASC interface I, II and III mutants showed the same phenotype for NLRP3 stimulation as for AIM2 stimulation (Supplementary Fig. 6a–c). Importantly, also ASCY59A and ASCE80R mutations abrogated speck formation and IL-1β secretion but still promoted pyroptosis on NLRP3 stimulation (Supplementary Fig. 6b, highlighted in grey). Finally, we also examined IL-1β release and pyroptosis of macrophages expressing interface II mutants treated with Clostridium difficile Toxin B (TcdB). TcdB was recently described to activate the pyrin inflammasome, which senses modifications of Rho GTPases6. Again, Asc−/− macrophages expressing ASCY59A or ASCE80R were competent for cell death induction but could not efficiently release IL-1β (Supplementary Fig. 6d, highlighted in grey). In conclusion, our site-directed mutagenesis approach identified two mutations, Y59A and E80R, in the inflammasome adaptor ASC that uncouple pyroptosis from ASC speck formation and cytokine secretion.
As cell death was still observed in cells expressing ASCY59A and ASCE80R, we speculated that the mutant proteins retained the ability to interact with the PYD of the receptor, but lost the ability to oligomerize into ASC filaments. To investigate this, we reconstituted the AIM2 inflammasome by co-expressing V5-tagged AIM2 with ASCFL or the ASCK21A, ASCY59A or ASCE80R mutants in HEK293T cells. AIM2 interactors were then immunoprecipitated using an anti-V5 antibody and analysed by western blotting (Fig. 5a). Consistent with our notion, ASCY59A and ASCE80R retained the ability to interact with AIM2, similar to WT ASC. In contrast, ASCK21A did not interact with AIM2, possibly explaining the clear-cut phenotypes of interface I mutants (Fig. 4b and Supplementary Fig. 6b).
Given that ASCY59A and ASCE80R mutants still interacted with the receptor (Fig. 5a), but did not form any ASC filaments/specks (Fig. 4c and Supplementary Fig. 4e), we speculated that the mutations abolished or reduced the ability of ASC to organize itself into filaments following the recruitment to the activated receptor. Based on our previous observation that purified ASCPYD or ASCFL form filaments in vitro15, we established an assay to monitor the kinetics of this oligomerization process by dynamic light scattering and analysed the kinetics of filament formation for WT ASCPYD and the above-mentioned mutants (Fig. 5b,c and Supplementary Fig. 7a,b). Although WT PYD rapidly formed well-structured filaments, the K21A mutant was no longer able to oligomerize correctly and formed unspecific aggregates (Fig. 5d). Interestingly, E80R also lost the ability to oligomerize into ordered filaments, whereas the Y59A mutant formed well-ordered filaments, but at a much slower rate than the WT protein (Fig. 5b–d and Supplementary Fig. 7a,b). Thus, these results confirmed that the ASC–receptor interaction and the ASC filament formation can be uncoupled, and that the ability to rapidly form filaments/specks strongly correlates with the levels of cytokine production but not cell death.
The observation that even in the absence of ASC oligomerization inflammasome activation still promotes cell death but not cytokine release (Fig. 4c, Supplementary Fig. 4e) was reminiscent of NLRC4 activation in Asc−/− macrophages24. In the absence of ASC, NLRC4 interacts with pro-caspase-1 to initiate cell death but not efficient caspase-1 autoprocessing and cytokine secretion. We therefore examined to what level ASCY59A and ASCE80R were able to promote caspase-1 autoprocessing on activation of the AIM2 inflammasome. In contrast to cells expressing ASCFL, macrophages expressing the mutant proteins displayed significantly reduced levels of caspase-1 processing, as determined by western blotting for the released p20 subunit (Fig. 6a). Furthermore, caspase-1 autoprocessing correlated with the amount of released bioactive IL-1β (Figs 4c and and6a),6a), but not with the induction of cell death (Fig. 4c) and the release of the danger signal high mobility group box 1 (HMGB1) into the cell supernatants (Fig. 6a), an alternative marker for cell lysis during inflammasome activation51. Similarly, results were obtained on activation of the NLRP3 inflammasome (Supplementary Fig. 8a). To exclude the possibility that the observed cell death is induced through caspase-1-independent pathways, we knocked out Casp1 by CRISPR-Cas9-mediated gene targeting52,53 in the cell lines expressing ASCFL, ASCY59A or ASCE80R. Western blotting for pro-caspase-1 confirmed successful targeting (Supplementary Fig. 8b). Casp1 deficiency significantly abrogated pyroptosis induction in cells expressing ASCFL, ASCY59A or ASCE80R (Fig. 6b), thus confirming that the cell death observed in the parental cell lines was caspase-1 dependent. Taken together, these data confirm that caspase-1 processing is not a prerequisite for induction of cell death, expanding previous findings from CARD-containing inflammasome receptors24 to PYD-containing receptors.
Recent reports showed that a major substrate of caspase-1 responsible for pyroptosis is gasdermin-D, and that Gsdmd-deficiency abrogates pyroptosis11,12. Caspase-1 cleaves full-length gasdermin-D, whereby the active N-terminal fragment (gasdermin-DNterm) is generated (Supplementary Fig. 8c,d), which induces pyroptotic cell death11,12. Consistently, CRISPR-Casp9-mediated Gsdmd knockout (Supplementary Fig. 8e) reduced cell death significantly in response to AIM2 and NLRP3 activation (Supplementary Fig. 8f,g). As ASCY59A and ASCE80R mutants induced cell death, we speculated that gasdermin-D is still processed in cells expressing these mutant forms of ASC. Indeed, although gasdermin-D was not processed in Asc−/− macrophages on stimulation of AIM2 or NLRP3, we observed gasdermin-D processing to its 30kDa gasdermin-DNterm fragment in Asc−/− macrophages expressing ASCFL, ASCY59A and ASCE80R (Fig. 6c and Supplementary Fig. 8h). Furthermore, processing of gasdermin-D on AIM2 stimulation is dependent on caspase-1 in cells harbouring oligomerization-deficient ASC mutants, as Casp1 deficiency abrogated gasdermin-D cleavage (Fig. 6c). These results indicated that gasdermin-DNterm causes the cell death observed in these mutations. As significant levels of cell death can be observed in Asc-deficient macrophages after NLRC4 activation24, but caspase-1 processing is reduced below detection levels, we next determined whether gasdermin-D is processed under these conditions. Indeed, we found that the induction of cell death in S. Typhimurium-infected WT, Asc−/− and Casp1−/−/Casp11−/− macrophages (Fig. 6d) correlated with detectable levels of processed gasdermin-D (Fig. 6e). Furthermore, CRISPR-Cas9-mediated Gsdmd knockout confirmed that gasdermin-D played an essential role in inducing pyroptosis during S. Typhimurium infections (Supplementary Fig. 8i). Taken together, these results confirm that even the smallest amounts of active caspase-1, as judged by the amount of processed caspase-1 p20 subunits, are sufficient to efficiently process gasdermin-D and trigger gasdermin-D-induced cell death. On the other hand, large amounts of processed, active caspase-1 are required to produce detectable amounts of mature, bioactive IL-1β as seen by the direct correlation of the level of caspase-1 processing with the levels of cytokine release. Importantly, the ability of ASC to form filaments and ASC specks correlates with caspase-1 activation and cytokine processing, thus supporting a model in which the rapid formation of ASC filaments acts as a signal amplification mechanism for inflammasomes, generating a multitude of caspase-1 activation sites and thus enabling the cells to rapidly mature IL-1β before the onset of pyroptosis.
The formation of higher-order signalling machineries, signalosomes, for transmission of receptor activation information to cellular responses is an emerging theme in signal transduction48. It is particularly important in innate immune signalling, where the signal generated by a few ligand–receptor interactions needs to trigger an appropriate cellular response. Formation of oligomers has been reported for different signalling adaptors, for example, B-cell lymphoma/leukemia 10 (Bcl10) or MAVS filaments, the Myddosome36,54,55 and others. However, the archetypical supramolecular assembly formed during inflammasome activation, the ASC speck, has remained relatively poorly understood. Here we present evidence that the formation of ASC speck by oligomerization of the inflammasome adaptor protein ASC acts as a signal amplification mechanism for inflammasomes (Fig. 7), as the rapid formation of ASCPYD filaments that expose ASCCARD on their surface creates a multitude of pro-caspase-1 recruitment and activation sites. We speculate that such a system might therefore be able to detect the smallest amounts of PAMPs in the host cell cytosol, as recent reports have shown that a single ligand molecule (flagellin, PrgJ) is sufficient to initiate assembly of a NLR family, apoptosis inhibitory protein (NAIP) and 10–12 NLRC4 proteins into a wheel-shaped receptor oligomer that acts as seed for ASC oligomerization39,56.
Our result reveals that this signal amplification mechanism only applies to cytokine maturation, whereas gasdermin-D-induced pyroptotic cell death could be observed even in the absence of ASC oligomerization. Why would cells need to amplify the receptor-generated signal to process IL-1β/-18? We speculate that a cell will inevitably progress towards pyroptosis, once a few molecules of caspase-1 have been activated, given the rapid kinetics by which gasdermin-D is processed57 and the strong toxicity of the gasdermin-DNterm fragment11,12. This fate leaves a limited time window to mature the pro-forms of IL-1β/-18 to their bioactive form. As processing of IL-1β/-18 occurs with slow kinetics, increasing the amount of active caspase-1 is a means to generate large amounts of bioactive IL-1β/-18 in the available time window. Consistent with this notion we have observed that gasdermin-D processing does not correlate with the overall level of active caspase-1. Furthermore, pyroptosis has been observed to occur within 30–60min in stimulated cells, whereas ASC speck formation was reported to proceed much faster23,58, with all cytosolic ASC being incorporated into a speck in <3min20,59. Thus, ASC speck formation might serve to activate sufficient amounts of caspase-1 to generate enough bioactive cytokines before the cell lyses and releases its intracellular content, including proteolytically matured cytokines.
Although recent studies revealed that ASC specks are filamentous in nature22,23, consistent with recently reported structures of human and murine ASCPYD filaments13,15, it was so far unknown how these filaments assemble into a speck. Our mutagenesis studies indicates that the clustering of ASCPYD filaments and finally their condensation into a dense ASC speck structure is mediated by the ASCCARD, which, as we have shown15, is exposed on the surface of the ASCPYD filament. These results have uncovered a function for the ASCCARD in ASC speck formation, beyond its role as adaptor domain between receptor and caspase-1. How ASCCARDs condense ASCPYD filaments is unknown, but it can be assumed that either dimeric interactions between ASCCARDs or limited oligomerization of ASCCARD from different filaments could build lattices interconnecting the ASCPYD filaments. Interestingly, the same residues in ASCCARD that are required for ASC–pro-caspase-1 interaction are also needed for ASC speck formation. Notably, it has previously been shown that phosphorylation of the ASCCARD at Y144/Y146 is not only required for caspase-1 activation, but also for ASC speck formation60,61 Our model of the ASC speck, which includes interactions between the ASCCARD domains, now provides a structural rationale for these data. To which extent ASC filaments are still formed in the absence of ASCCARD phosphorylation remains to be determined.
Recently, a unified assembly mechanism was proposed for the PYD-containing receptors AIM2 and NLRP3: the receptors nucleate clusters of ASC through PYD–PYD interactions, which in turn nucleate caspase-1 filaments13. In analogy, it was proposed that the CARD-containing receptor NLRC4 directly nucleates caspase-1 filaments13,39, as in theory CARD-containing receptors do not require ASC for caspase-1 recruitment. Although direct recruitment of pro-caspase-1 to NLRC4 might indeed happen in Asc-deficient cells as we hypothesized before24, it does not seem to efficiently activate caspase-1, because we and others have reported that caspase-1 processing into its p20/p10 subunits and the levels of released mature cytokines are reduced. Furthermore, a large number of ASC specks can be observed in WT cells on NLRC4 activation24,25,26,27,28, indicating that ASC nevertheless plays an important role in NLRC4 signalling. Our results now indicate that the ASCCARD lacks the ability to form higher-order oligomers at physiological concentrations, and that even on NLRC4 activation ASC speck formation relies on the ASCPYD. Our data thus support a model in which CARD-containing receptors initiate ASC speck formation with the help of ASC bridging molecules, and that the PYDs of these ASC molecules nucleates an ASCPYD filament. Thus, the ASCCARD can have a triple function in inflammasome signalling on CARD-receptor activation.
In conclusion, our study gives insights into the architecture of the ASC speck and the mechanism by which ASC filaments assemble this structure. Furthermore, we provide experimental evidence that ASC filament formation serves as an amplification mechanism in inflammasome signalling, and that this amplification serves to generate sufficient mature cytokines before the onset of pyroptotic cell death. Nevertheless, important questions remain; the higher-resolution structure of the activated NLRC4 receptor has been recently reported39,56, yet how exactly such receptor complexes with C10–C12 stoichiometry initiate the 3-start ASCPYD filaments is still unknown. Furthermore, the structure of the whole ASC speck assembly, including receptor, ASC and caspase-1, is still lacking and it remains unknown whether other proteins participate in its formation. Finally, additional theoretical and experimental approaches will be necessary to understand if and how ASC speck formation might impart threshold responses, reduce biological noise and control temporal and spatial control of inflammasome signalling.
Immortalized BMDM cell lines were generated from bone marrow using infection with a v-myc/v-raf-expressing J2 retrovirus24,62. WT and Asc−/− immortalized murine BMDMs were cultured in DMEM (Sigma) supplemented with 10% FCS (Bioconcept) and 10% 3T3-MCSF supernatant, and incubated at 37°C with 5% CO2.
Murine AscFL and AscPYD were cloned with a carboxy-terminal mCherry tag and AscCARD with an N-terminal enhanced GFP tag into V48, a derivative of the replication-defective murine stem cell retroviral construct pMSCV2.2 (excision of the IRES-GFP by EcoRI digestion, gift from Thomas Henry). Mutations were introduced by SOE PCR using appropriate oligonucleotides (Supplementary Table 1) and cloned into V48. GP2 packaging cells were transfected with the individual vectors (9μg per 1.5 × 106 cells) and the lentviral envelope vector VSV-G (Addgene, 6μg per 1.5 × 106 cells), and the retroviral particles were used to transduce Asc−/− iBMDMs (106 cells). Seven days later, the transduced iBMDMs were sorted into single cells based on the mCherry or enhanced GFP expression (FACS) and grown up to clonal cell lines. Up to ten clones of each cell line were tested for inflammasome activation and the level of ASC expression was assessed by western blotting and a representative clone was selected for further analysis.
For transduction of primary bone marrow cells, retroviral particles were generated in Phoenix-Eco packaging cells and used to transduce WT C57BL/6 bone marrow cells after 48 and 72h of culture in medium with 10% 3T3-MCSF supernatant. Inflammasome stimulation was performed 4 days after the first transduction24.
Immortalized BMDMs (seeded at 250000 cells per ml in 96-well plates) were primed for 4h with 100ngml−1 LPS O55:B5 (Invivogen). The NLRP3 inflammasome was triggered by addition of 5mM extracellular ATP (Sigma-Aldrich) for 60min. The AIM2 inflammasome was triggered by transfection of 1μgml−1 of the synthetic DNA analogue poly(dA:dT) (Invivogen) using Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocol, in OptiMEM (Gibco) for 3h. The NLRC4 inflammasome was triggered by infection of the cells with S. enterica serovar Typhimurium SL1344 at a multiplicity of infection of 10. The infection was synchronized by centrifugation and continued for 60min. The pyrin inflammasome was triggered by addition of 1μgml−1 (final concentration) of C. difficile toxin B (CdtB, Enzo Biotech) for 2.5h.
IL-1β release was measured by ELISA (eBiosciences). Cell death was quantified by measuring LDH release using the LDH Cytotoxicity detection kit (TaKaRa Clontech). To normalize for spontaneous cell lysis, the percentage of cell death was calculated as follows: [(LDH sample)−(LDH negative control)]/[(LDH positive control)−(LDH negative control)] × 100.
Cells were seeded on coverslips (150000 cells per coverslip) and treated as described above with the addition of 25μM Z-VAD-fmk (Bachem) to prevent detachment of pyroptotic cells and therefore loss of cells with ASC specks. Coverslips were fixed with 4% paraformaldehyde (PFA) (15min, 37°C, Alfa Aesar) and washed with PBS. Nuclei were stained with Hoechst 33342 (Life Technologies) for 10min and the slides mounted using Vectashield (Vector Laboratories). For quantifications of ASC aggregates (specks or filaments), ten random regions of interest were imaged at × 20 magnification (Leica DMI3000B inverted fluorescence microscope, HCX PL FLUOTAR objective, Leica DFC3000G camera and LAS AF Version 3 software) and the number of ASC aggregates and cells were counted. For representative images, the slides were imaged at × 63 magnification (Leica point scanning confocal ‘SP8', HC PL APO CS2 × 63 objective, Leica AF software version 3). For measurements of speck sizes, random regions of interest were images at × 63 magnification (Leica point scanning confocal ‘SP8') and the largest diameters of individual specks were measured using Fiji63.
For western blotting of supernatant and lysate samples, cells were seeded at 106 cells per well in six-well plates and treated as described above. Supernatants were precipitated with 10% trichloracetic acid, precipitates washed with acetone and resuspended in 40μl 1 × SDS–PAGE sample buffer and boiled at 95°C for 10min. Cells were lysed with 200μl 1 × RIPA (50mM Tris-HCl pH 7.5, 150mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% NP-40, 1 × protease inhibitor cocktail (Roche)) for 30min on ice. Lysates were re-suspended in 5 × SDS–PAGE sample buffer and boiled at 95°C for 10min64. For combined supernatant and lysates, samples were prepared as above, but supernatant precipitates were resuspended in lysate samples. Samples were run on 14% (supernatants or lysates) or 12% (combined SN+lysates) acrylamide gels (1h, 170V, 40mA per gel), transferred to polyvinylidene difluoride membranes (1h, 100V constant), blocked in 5% milk in Tris-buffered saline+Tween-20 (TBS-T) and incubated with primary antibodies in 5% BSA–TBS-T for 16h at 4°C or 2h at room temperature with agitation. Secondary antibodies were diluted 1:5,000 in 5% milk–TBS-T and incubated for 1h at room temperature. The membranes were developed using either LumiGLO (KPL) or LumiGLO Reserve (KPL). The following antibodies and dilutions were used: rat anti-Caspase-1 p20 (Genentech, 1:1,000; Supplementary Figs 1b,d and 2c,d) or mouse anti-Caspase 1 p20 (1:4,000, AG20B-0042, AdipoGen; Fig. 6a and Supplementary Fig. 8a), rat anti-ASC (Genentech, 1:2,000; Fig. 5a and Supplementary Fig. 2a,h) or rabbit anti-ASC (1:1,000, AG25B-006, Adipogen, all other figures), rabbit anti-IL-18 (5180R-100, Biovision, 1:1,000), goat anti-IL-1β (AF-401-NA, R&D, 1:1,000), mouse anti-GFP (632381, Clontech, 1:1,000), mouse anti-mCherry (ab125096, Abcam, 1:1,000), mouse anti-V5 (R960-25, Invitrogen, 1:1,000), rabbit anti-HMGB1 (GTX-101277, Genetex, 1:1,000), mouse anti-gasdermin-D (GSDMDC1 (A-7), Santa Cruz Biotechnology, 1:1,000; Supplementary Fig. 7), rabbit anti-gasdermin-D (1:2,000, G7422, Sigma, all other figures) and mouse anti-β-actin (A1987, Sigma, 1:1,000).
Co-immunoprecipitation was done according to a modified protocol published previously49. For assessment of ASC/AIM2 interactions, 3 × 800,000 HEK293T cells were transfected with 2μg per well AIM2-V5 and ASC-mCherry (ASCFL, ASCK21A, ASCY59A or ASCE80R) encoding plasmids at a ratio of 1:4 using linear polyethylenimine (PEI, Polysciences) at a ratio DNA:PEI of 1:4 (refs 65, 66). Twenty-four hours after transfection, the cells were washed twice with ice-cold 1 × PBS and lysed using HEK lysis buffer (20mM HEPES pH 7.4, 10mM KCl, 1mM EDTA, 0.1mM phenylmethylsulfonyl fluoride, 1mM Na3VO4, 5mM NaF, 0.5 % Nonidet P-40 and 1 × protease inhibitor cocktail (Roche)). Lysates were sonicated for 5 × 7s before removing debris and non-lysed cells by centrifugation (10,000g, 15min, 4°C). The samples were incubated with 1μg mouse anti-V5 antibody (R960-25, Invitrogen) or a control antibody (mouse anti-GFP, 632381, Clontech) with agitation for 16h at 4°C. For assessment of ASCCARD/ASCCARD interactions, 3 × 800000 HEK293T cells were transfected with 1μg per well ASC-mCherry (ASCFL, ASCPYD, ASCD130R or ASCD134R), 1μg per well ASCCARD-GFP and 0.2μg AIM2-V5, to initiate ASC/ASC interactions using linear PEI. Forty hours after transfection, the cells were washed, lysed and sonicated as described above. The samples were then incubated with 1μg mouse anti-mCherry antibody (ab125096, Abcam) or a control antibody (mouse anti-HA, MMS-101R-200, Covance). The samples were then incubated with agitation for 2h at 4°C with 25μl Pierce Protein A Plus Agarose bead slurry (22810, Thermo Scientific). Beads were washed three times with lysis buffer (centrifugation 1min 1,000g, 4°C), resuspended in 2 × SDS–PAGE sample buffer (30μl) and boiled at 95°C, before analysing the eluted proteins by western blotting as described above.
Two guide RNAs52 targeting exon 4 of Casp1 (5′-gagggcaagacgtgtacgag-3′ and 5′-cgagtggttgtattcattat-3′) and one guide RNA targeting exon 2 of Gsdmd (5′-ggtcaagaatgtgatcaagg-3′) were cloned into lentiCRISPRv2 harbouring a puromycin resistance cassette (Addgene53). These constructs were transfected into HEK293T cells using PEI (as described above) together with the lentiviral packaging vector PsPax2 (Addgene) and the lentviral envelope vector VSV-G (Addgene). Sixteen hours after transfection, medium was exchanged with macrophage medium and incubated at 37°C for 2 days to produce lentiviral particles. The lentiviral particles were used to transfect the immortalized macrophage cell lines (800,000 cells per well in 6-well plates) using polybrene (Merck) to favour virus attachment. Two days after viral transduction, the macrophages were expanded. Attached macrophages were then treated with 10μgml−1 puromycin (Gibco) for 6 days to select for successful lentiviral transduction. After puromycin selection, the cells were tested for successful knockout by western blotting.
The ASCPYD (residues 1–91) was cloned with a C-terminal His6 tag into the pET28a vector under the control of a T7 promoter. Protein expression was induced by isopropyl-β-D-thiogalactopyranosid addition in BL21(DE3) Escherichia coli at an OD600 of 0.8 for 4h at 37°C. Bacteria were harvested by centrifugation and resuspended in 50mM phosphate buffer pH 7.5, 300mM NaCl, with Complete protease inhibitor (Roche). Resuspended bacteria were incubated for 1h at room temperature with DNase I, sonicated on ice and centrifuged at 20,000g at 4°C for 30min. The pellet, including ASCPYD-containing inclusion bodies, was solubilized in 50mM phosphate buffer pH 7.5, 300mM NaCl, 6M guanidinium hydrochloride and centrifuged at 20,000g at 4°C for 30min. The supernatant was incubated for 2h at room temperature with pre-equilibrated Ni-NTA affinity resin (Thermo Scientific) and then passed through a plastic body column for gravity flow purification. The column was washed with 20 column volumes of solubilization buffer containing 20mM imidazole and eluted with 3 column volumes of solubilization buffer with 500mM imidazole. The pH of the elution fraction was decreased to 3.8 and dialysed against 50mM glycine buffer pH 3.8, 150mM NaCl. The protein was further purified on a pre-equilibrated Superdex 75 gel filtration column (GE Healthcare). This gel-filtration step removed traces of pre-existing aggregates and yielded highly pure, monomeric soluble form of ASCPYD. Samples were either used immediately or stored after flash-freezing in small aliquots in liquid N2.
Immediately before the experiments, samples of monomeric soluble ASCPYD were centrifuged at 20,000g at 4°C for 30min and filtered with 0.1μM filter (Millipore). The protein concentration was adjusted to 25μM by dilution from a higher-concentrated stock solution. Filament formation was triggered by rapid dilution to neutral pH. Thereby, 70μl of monomeric ASCPYD was mixed with 0.45μl of 2.75M NaOH solution to a reach the pH of 7.5. The solution was mixed at room temperature by careful pipetting, to avoid introduction of air bubbles, and immediately transferred to a quartz cuvette with 1cm path length. Between runs, cuvettes were carefully cleaned with 1M Hellmanex solution (Sigma-Aldrich) to avoid cross-seeding effects between sequential measurements. Filament growth was monitored by dynamic light scattering with a Malvern Zetasizer Nano ZS series instrument. The laser focal spot was positioned in the middle of the cuvette and maintained fixed for all the measurements. To maximize the intensity of the scattered light, the minimal attenuation level was used. Data were acquired in 60s intervals by averaging three runs of 20s, until a total time of 350min. Afterwards, the protein solution was blotted on EM grids, negatively stained and imaged with transmission electron microscopy to visualize filament formation.
Filament growth was modelled assuming pseudo-first-order kinetics where the filament propagation step occurs by the addition of monomers to the initial growth centre. The time-dependent growth signal I was fitted independently for each measurement by a single exponential function,
where I∞ corresponds to the signal at time t and at infinitive time, respectively, and kF is the first-order rate constant. Fits were done with nonlinear least-square minimization.
The data that support the findings of this study are available from the corresponding author upon request.
How to cite this article: Dick, M. S. et al. ASC filament formation serves as a signal amplification mechanism for inflammasomes. Nat. Commun. 7:11929 doi: 10.1038/ncomms11929 (2016).
Supplementary Figures 1-9, Supplementary Table 1 and Supplementary References.
We thank V. Dixit (Genentech) for the rat antibody to ASC and rat antibody to caspase-1 p20; Vesna Oliveri, Tim Sharpe, the Imaging Core Facility, FACS Core Facility, Biophysics Core Facility and the Microscopy Center of the Biozentrum, University of Basel, for technical assistance. Work was supported by the Swiss National Science Foundation (PP00P3_139120/1 to P.B).
Author contributions M.S.D., L.S., S.H. and P.B. designed research. M.S.D., L.S., S.R. and P.B. performed experiments. M.S.D., L.S., S.H. and P.B. analysed data and wrote the manuscript.