The zebrafish NLRP3 inflammasome has functional roles in ASC-dependent interleukin-1β maturation and gasdermin E–mediated pyroptosis

Abstract

The NLR family pyrin domain containing 3 (NLRP3) inflammasome is one of the best-characterized inflammasomes in humans and other mammals. However, knowledge about the NLRP3 inflammasome in nonmammalian species remains limited. Here, we report the molecular and functional identification of an NLRP3 homolog (DrNLRP3) in a zebrafish (Danio rerio) model. We found that DrNLRP3's overall structural architecture was shared with mammalian NLRP3s. It initiates a classical inflammasome assembly for zebrafish inflammatory caspase (DrCaspase-A/-B) activation and interleukin 1β (DrIL-1β) maturation in an apoptosis-associated speck-like protein containing a caspase-recruitment domain (ASC)-dependent manner, in which DrNLRP3 organizes DrASC into a filament that recruits DrCaspase-A/-B by homotypic pyrin domain (PYD)–PYD interactions. DrCaspase-A/-B activation in the DrNLRP3 inflammasome occurred in two steps, with DrCaspase-A being activated first and DrCaspase-B second. DrNLRP3 also directly activated full-length DrCaspase-B and elicited cell pyroptosis in a gasdermin E (GSDME)-dependent but ASC-independent manner. These two events were tightly coordinated by DrNLRP3 to ensure efficient IL-1β secretion for the initiation of host innate immunity. By knocking down DrNLRP3 in zebrafish embryos and generating a DrASC-knockout (DrASC^−/−) fish clone, we characterized the function of the DrNLRP3 inflammasome in anti-bacterial immunity in vivo. The results of our study disclosed the origin of the NLRP3 inflammasome in teleost fish, providing a cross-species understanding of the evolutionary history of inflammasomes. Our findings also indicate that the NLRP3 inflammasome may coordinate inflammatory cytokine processing and secretion through a GSDME-mediated pyroptotic pathway, uncovering a previously unrecognized regulatory function of NLRP3 in both inflammation and cell pyroptosis.

Introduction

Inflammasomes are cytosolic multiprotein complexes that assemble in response to exogenous microbial invasions and endogenous damage signals (1, 2). Inflammasomes act as a central platform to activate inflammatory caspases such as Caspase-1, which processes the proinflammatory cytokines interleukin 1β (IL-1β)³ and IL-18 for their maturation, and hydrolyze gasdermin D to generate an N-terminal fragment to induce membrane perforation, cytokine release, and cell pyroptosis (3, 4). Among numerous inflammasomes identified in mammals, NLRP3 inflammasome is the most extensively studied because of its crucial functional roles in innate immunity and pathogenesis of various diseases (5, 6). NLRP3 is a family member of nucleotide-binding domain and leucine-rich repeat (LRR)-containing proteins (NLRs) (7). The mammalian NLRP3 is structurally characterized by the presence of an N-terminal pyrin domain (PYD), a central adenosine triphosphatase (ATPase) domain known as NACHT, a domain associated with NACHT in fish and other vertebrates (FISNA) (8), and a C-terminal LRR domain (9). Once activated by a spectrum of stimuli, NLRP3 combines with the ASC (apoptosis-associated speck-like protein containing a caspase-recruitment domain) adaptor protein to form a speck-shaped inflammasome, which recruits and activates Caspase-1 for IL-1β maturation (10, 11). Despite numerous studies on mammals, the occurrence of NLRP3 inflammasome in ancient vertebrates remains poorly understood. The identification of NLRP3 inflammasome in lower vertebrates, particularly in primitive teleost fish, will contribute to cross-species understanding of NLRP3-mediated biology throughout vertebrate evolution.

Several previous studies identified two proinflammatory caspases, namely DrCaspase-A (Caspy) and DrCaspase-B (Caspy2), from zebrafish (12). These two caspases engage in proIL-1β cleavage with different specificities in an ASC-dependent manner (13). The zebrafish proIL-1β (proDrIL-1β) has two unique cleavage sites instead of one conserved Caspase-1 autoproteolytic site as observed in mammals. proDrIL-1β is first cleaved by DrCaspase-A into a partially processed 20-kDa intermediate form at the Asp¹⁰⁴ residue, and then cleaved again into a fully processed 18-kDa mature form by DrCaspase-B at the Asp¹²² residue (13, 14). The catalytic domain of DrCaspase-A/B shares the highest homology with that of human Caspase-1/5, with a sequence similarity of 54/57% (15). In humans, Caspase-4/5 seems to function similarly to mouse Caspase-11 as shown by their abilities to cleave gasdermin D (GSDMD) and thereby induce cell pyroptosis after they are activated by intracellular lipopolysaccharide (LPS) (16). Thus, DrCaspase-A and DrCaspase-B are likely the orthologs of Caspase-1 and Caspase-4/5/11 in mammals (12, 15, 17). Recently, a canonical NLRP1 inflammasome has also been identified from zebrafish; it can initiate DrCaspase-A/-B activation in a sequential recruitment manner (18). In addition, DrCaspase-B (caspy2) also senses cytosolic LPS through its pyrin domain, which mediates pyroptosis in zebrafish fibroblasts via a Caspase-4/5/11-like activity as observed in mammalian noncanonical inflammasomes (19). In mammals, multiple gasdermin proteins, including GSDMD, GSDMA, GSDMB, GSDMC, and GSDME, were found to induce pyroptosis; among them, GSDMD is selectively cleaved by Caspase-4/5/11 to liberate an N-terminal effector fragment from the C-terminal inhibitory domain (16, 20, 21). The N-terminal fragment oligomerizes in the cell membrane to form a 10–16-nm diameter pore through which mature-formed IL-1β and IL-18 are secreted (16, 22). With the accumulation of membrane pores, cells ultimately undergo membrane rupture and pyroptosis (23, 24). In zebrafish, however, only two gasdermin E homologs (DrGSDMEa/b, also known as DFNA5a/b) were identified to exert a pore-forming effect by their activated N-terminal domains; and DrGSDMEa that possesses a Caspase-3 cleavage motif induces pyroptosis after chemotherapy drugs are administered (25–27). Thus, whether DrGSDMEa/b acts as a substrate cleaved by DrCaspase-A/B to induce pyroptosis as mammalian GSDME did remains to be elucidated. In this study, we identified an NLRP3 inflammasome from zebrafish and revealed its unique characteristics in DrCaspase-A/B activation, DrIL-1β maturation, DrGSDMEa/b cleavage, and pyroptotic cell death. Our results showed the molecular and functional characterizations of an early NLRP3 inflammasome in an ancient vertebrate, which provides a cross-species understanding of the evolutionary history of NLRP3 inflammasome from teleost fish to mammals.

Results

Molecular characterization of DrNLRP3 and DrGSDME

Through a systematic search in Ensemble by BLASTN, one NLRP3 (DrNLRP3) and two DrGSDMEa/b candidate genes were retrieved from the zebrafish genome database. The DrNLRP3 gene is located within a 10.0-kb genomic fragment on zebrafish chromosome 1 with eight exons and seven introns. The DrGSDMEa/b gene is distributed within a 15.2/13.8-kb genomic fragment on chromosome 19/16 with 9/9 exons and 8/8 introns. Genes adjacent to DrNLRP3 and DrGSDMEa/b loci share an overall conserved chromosome synteny with those of humans and other mammalian species (Fig. 1, A and G). The cDNA of DrNLRP3 consists of a 64-bp 5′-untranslated region (5′-UTR), a 3243-bp 3′-UTR, and a 3345-bp open reading frame (ORF) that encodes 1114 amino acids (GenBank^TM accession no. MN088121). The cDNA of DrGSDMEa/b contains a 296/90-bp 5′-UTR, a 184/841-bp 3′-UTR, and a 1566/1419-bp ORF that encodes 521/472 amino acids (GenBank^TM accession no. XM_005170077.3/NM_001001947.1).

Figure 1.

Molecular and structural identification of DrNLRP3 and DrGSDME. A, gene synteny and chromosomal location analysis of genes adjacent to NLRP3 loci on human chromosome 1 (top) and zebrafish chromosome 1 (bottom). Arrows indicate gene orientation. B, schematic of the domain architecture of HsNLRP3 and DrNLRP3. C, DrNLRP3 and HsNLRP3 tertiary structures predicted by SWISS-MODEL with crystal structures of NACHT and LRR (PDB code 4kxf.3.A) as models. D, DrNLRP3 domain architecture and tertiary structure modeled by I-TASSER. The top five threading templates are 5irmA, 6npva, 4kxfK, 6b5bA, and 6eqoA. E, schematic of the domain architecture of HsGSDME and DrGSDMEa/b. The predicted cleaved sites in DrGSDMEa and DrGSDMEb are ²⁵³SEVD²⁵⁶ and ²⁴⁴FEID²⁴⁷, respectively. F, tertiary structures of full-length HsGSDME and DrGSDMEa/b protein predicted by SWISS-MODEL with crystal GSDMA3 structures (PDB code 5b5r.1.A) as model. G, gene synteny and chromosomal location analysis of genes adjacent to HsGSDME and DrGSDMEa/b loci on human chromosome 1 (top) and zebrafish chromosome 19/16 (bottom). Arrows indicate gene orientation. H and I, phylogenetic analysis of the relationship of NLRP3 and GSDME between fish and other species. The phylogenetic tree was constructed by MEGA (version 5.0) by using the maximum likelihood method. Each node reliability was estimated by bootstrapping with 2000 replications. J, quantitative RT-PCR analysis of the expression patterns of DrNLRP3 and DrGSDMEa/b genes in adult zebrafish tissues. K, expression patterns of DrNLRP3 and DrGSDMEa/b genes in zebrafish embryos at different developmental stages. The relative expression levels of the genes was calculated by the 2^−ΔCt method with β-actin for normalization. Each data point shows the mean ± S.D. with three replicates representative of three independent experiments.

DrNLRP3 and DrGSDMEa/b proteins share similar domain and tertiary architectures with mammalian counterparts (Fig. 1, B and C). DrNLRP3 contains an N-terminal pyrin domain (PYD), a central nucleotide-binding domain (NACHT), a domain associated with NACHT in fish and other vertebrates (FISNA), a series of leucine-rich repeats (LRR), and a unique C-terminal B30.2 domain (Figs. 1B and Fig. S1A). The PYD domain comprises five antiparallel α-helices and one conservative truncated helix. The NACHT domain possesses Walker A (ATP/GTPase-specific P-loop) and Walker B (Mg²⁺-binding site) motifs. Eight LRRs with conserved LXXLXLXXN/CXL motifs form a well-defined “horseshoe”-shaped scaffold in which the B30.2 domain is located without any complete α-helix or β-sheet architecture (Fig. S1, A and D). DrGSDMEa/b consists of an N-terminal domain with the most conservative β-sheets (β1–β11) and α-helices (α1–α4), an aspartic acid–cleaved site in a long-loop structure, and a C-terminal domain with seven helices (α5–α11) (Fig. 1, E and F). The C terminus harbors conserved hydrophobic residues, such as Ile²⁹⁶/Leu²⁹⁴, Ala⁴²⁰/Ala²⁶⁹, and Leu⁴³¹/Leu²⁸⁰ on α5, α8, and α9 helices, to form nonpolar surfaces that interact with the N-terminal domain for an auto-inhibitory regulation (Fig. S1B). Key pore-forming residues, such as Lys⁷/Lys⁷, Lys³⁹/Lys³⁹, Lys⁵⁰/Lys⁵⁰, Thr¹⁰¹/Thr⁹⁹, Arg¹³⁶/Arg¹³⁴, and Ser¹⁴³/Ser¹⁴¹ also exist in the N-terminal domain, which are conserved from fish to mammals. Phylogenetic analysis shows that DrNLRP3 and DrGSDMEa/b are clustered to their homologs with a high bootstrap probability (Fig. 1, H and I). In addition, DrNLRP3 and DrGSDMEa/b are extensively expressed in zebrafish adult tissues such as head, kidney, spleen, gill, and intestine and in embryos (Fig. 1, J and K).

DrNLRP3 initiates DrCaspase-A/-B activation in different manners

The functional role of DrNLRP3 was first examined by its initiation activity for DrCaspase-A/-B in HEK293T cells that naturally have minimal expression of inflammasome components. As expected, DrNLRP3 and DrASC coexpression significantly augmented (p < 0.01) the DrCaspase-A/-B activity in a DrNLRP3 dose-dependent manner (Fig. 2, A and B). However, the DrNLRP3 and DrCaspase-A/-B coexpression in the absence of DrASC dramatically promoted (p < 0.001) DrCaspase-B but not DrCaspase-A activation, with a maximum up-regulation of up to 180% (Fig. 2B). The DrNLRP3-ΔPYD and DrNLRP3-ΔNACHT (lacking both NACHT and FISNA) mutant proteins significantly impaired (p < 0.01 or p < 0.001) DrCaspase-A and DrCaspase-B activation. The DrNLRP3-ΔLRR mutants restrained DrCaspase-A but not DrCaspase-B activation. Conversely, DrNLRP3-ΔB30.2 did not influence either DrCaspase-A or DrCaspase-B (Fig. 2C). Western blot analysis showed the self-cleavage of 45/47-kDa pro-DrCaspase-A/-B into a 35-kDa hydrolytic product (p35) if pro-DrCaspase-A/-B was coexpressed with DrNLRP3 and DrASC (Fig. 2, D and E). By contrast, no p35 was detected without the coexpression of either DrNLRP3 or DrASC. The DrNLRP3 mutants that lacked PYD, NACHT-FISNA, and LRR domains failed to induce pro-DrCaspase-A/-B hydrolyzation (Fig. 2F). Coimmunoprecipitation (co-IP) assay revealed the protein–protein interaction among DrNLRP3, DrASC, and DrCaspase-A/-B (Fig. 2, G–I). The DrNLRP3-ΔPYD and DrASC-ΔPYD mutants lacked such an ability, whereas DrASC–ΔCARD maintained the activity (Fig. 2, H and I). Overall, DrNLRP3 could activate pro-DrCaspase-A/-B by self-hydrolyzation in a DrASC-dependent manner and initiate pro-DrCaspase-B activation in a DrASC-independent manner without undergoing self-hydrolyzation. The pro-DrCaspase-A/-B self-hydrolyzation depends on the association between DrNLRP3 and DrASC via the PYD–PYD homotypic interaction in which NACHT and LRR domains were included.

Figure 2.

DrNLRP3 involvement in the activation of DrCaspase-A and DrCaspase-B. A, DrNLRP3 and DrASC activate DrCaspase-A detected by the specific Ac-YVAD-AFC fluorescent substrate. Each data point shows the mean ± S.D. with three replicates. *, p < 0.05. B, DrNLRP3 and DrASC activate DrCaspase-B detected by specific Ac-WEHD-AFC fluorescent substrate. Each data point shows the mean ± S.D. with three replicates (*, p < 0.05; **, p < 0.01; ***, p < 0.001). C, DrNLRP3 and its domain-lacking mutants simultaneously activate DrCaspase-A and DrCaspase-B detected by Ac-YVAD-AFC or Ac-WEHD-AFC fluorescent substrates. Data are representative of three independent experiments as mean ± S.D. (*, p < 0.05; ***, p < 0.001). D and E, Western blotting assay of DrCaspase-A (D) and DrCaspase-B (E) auto-hydrolyzation when coexpressed with DrNLRP3 and DrASC. F, Western blotting assay of the DrCaspase-A and DrCaspase-B auto-hydrolyzation when coexpressed with DrNLRP3 mutants and DrASC. G, IP assay shows the DrNLRP3 interaction with DrASC through the PYD domain. HEK293T cells were transfected with pCMV-Tag2B–DrNLRP3/DrNLRP3-ΔPYD and pCMV-HA-DrASC for 48 h. Cell lysates were immunoprecipitated with rabbit anti-Flag Ab and analyzed by Western blotting by using mouse anti-Flag or anti-HA against DrNLRP3 or DrASC, respectively (top panel). Expression of the transfected plasmids was analyzed with anti-Flag or anti-HA Ab in the whole-cell lysates (bottom panels). H, coIP assay shows the protein–protein interactions among DrCaspase-A (CasA-5DA), DrASC (DrASC-ΔPYD/ΔCARD), and DrNLRP3 (WT). I, coIP assay reveals the protein–protein interactions among DrCaspase-B (CasB-4DA), DrASC (DrASC-ΔPYD/ΔCARD), and DrNLRP3 (WT). The results are representative of three independent experiments. Caspase activity was detected and expressed as the fold induction over the control as described under “Experimental procedures.”

DrNLRP3 triggers DrASC nucleation and inflammasome formation

A DrASC nucleation (i.e. speck formation) assay was performed in HEK293T cells to observe whether DrNLRP3 could organize an inflammasome. When DrNLRP3 or DrASC was expressed alone in cells, the fluorescence was a weak signal that diffused throughout the cell (Fig. 3A and Fig. S2, A and B). With the DrNLRP3 and DrASC coexpression, DrNLRP3 associates DrASC into a speck structure with a size of 1.84 ± 0.55 μm in diameter (Fig. 3, B and C). In some cases, it was observed that DrNLRP3 formed an outer ring around DrASC (Fig. S2C). This speck organization was also observed in zebrafish ZF4 cells (Fig. S2D). The DrNLRP3-ΔPYD, DrNLRP3-ΔNACHT, and DrNLRP3-ΔLRR mutants blocked the speck organization, whereas the DrNLRP3-ΔB30.2 did not have an influence (Fig. 3D and Fig. S2E). The expression of truncated DrASC-ΔPYD or DrASC-ΔCARD induced filament (ASC^CARD or ASC^PYD) formation instead of speck organization (Fig. 3E). DrASC^PYD rather than the DrASC^CARD filament could be colocalized with DrNLRP3 (Fig. 3, F and G). Functionally, only DrNLRP3 and DrASC coexpressed in cells could activate DrCaspase-A (Fig. 3H). This activity was competitively inhibited by introducing DrASC-ΔPYD but not DrASC-ΔCARD in a ratio-dependent manner (Fig. 3I). Thus, the DrNLRP3-DrASC inflammasome formation might be started by a linker DrASC, which was associated with the DrNLRP3 disk via the PYD–PYD interaction and recruited another DrASC via the CARD–CARD interaction to form a DrASC filament. The DrASC filament was composed of a DrASC^CARD core and DrASC^PYD cluster. The latter finally recruited DrCaspase-A/-B via the PYD–PYD association (Fig. S3, A and B). From the tertiary structures of ASC^PYD and ASC^CARD cores, six PYDs or four CARDs interacted with each other by PYD–PYD or CARD–CARD homotypic interaction to form a circular helix in one layer of the filament core (Fig. S3C). This phenomenon explained that the DrASC^PYD filament was thicker than the DrASC^CARD filament under a confocal microscope (Fig. 3, E–G). With the alignment of DrASC, MmASC, and HsASC, some surface electrostatic amino acids, including Glu¹³, Lys²¹, Arg³⁸, Arg⁴¹, Asp⁴⁸, Asp⁵¹, and Asp⁵⁴ in ASC^PYD and Arg¹²⁵, Glu¹³⁰, Asp¹³⁴, Tyr¹⁴⁶, Arg¹⁵⁰, Arg¹⁶⁰, and Asp¹⁹¹ in ASC^CARD that perform important roles in mammalian PYD or CARD fibrillation, are also highly conserved in zebrafish (Fig. S3, C and D). This condition suggested that the mechanisms underlying the PYD–PYD and CARD–CARD homotypic interaction are conserved from fish to mammals.

Figure 3.

Aggregation of DrASC-dependent DrNLRP3 inflammasome. A, transient transfection of pCMV-Myc-DrASC or pCMV-Tag2B-DrNLRP3 in HEK293T cells, and diffuse fluorescent signals were detected in the cells. B, DrNLRP3 and DrASC coexpression in HEK293T cells. Flag-DrNLRP3 signal (red) and Myc-DrASC signal (green) accumulate in the same speck. C, confocal microscopy image of DrNLRP3-DrASC speck in ZF4 cells transfected with pCMV-Myc-DrASC and pCMV-Tag2B-DrNLRP3 by electroporation. D, statistics of DrASC speck-forming rates induced by DrNLRP3 and its mutants. More than 100 cells with DrASC speck were counted in each experimental group to quantify DrNLRP3-dependent DrASC nucleation. Fig. S2E shows the original immunofluorescence images. Data are representative of three independent experiments as mean ± S.D. (**, p < 0.01). E, immunofluorescence examination of DrASC^PYD/CARD filament when transiently transfected by Myc-tagged DrASC-PYD (DrASC-ΛCARD) or DrASC-CARD (DrASC-ΛPYD) in HEK293T cells. F and G, with the DrNLRP3 and DrASC coexpression, DrNLRP3 was colocalized with the DrASC^PYD filament (F) but not with the DrASC ^CARD filament (G). H, DrNLRP3 coexpressed with DrASC-ΛCARD or DrASC-ΛPYD cannot activate DrCaspase-A when being detected by the specific Ac-YVAD-AFC. Each data point shows the mean ± S.D. with three replicates (*, p < 0.05). I, DrASC-ΛPYD but not DrASC-ΛCARD interacts with the linker ASC and inhibits the DrCaspase-A activation by DrNLRP3 and DrASC. Images were captured under a laser-scanning confocal microscopy (Zeiss LSM-710; original magnification, ×630; scale bars represent 5 or 10 μm). The results are representative of three independent experiments as mean ± S.D. (*, p < 0.05).

DrNLRP3-DrASC inflammasome recruits DrCaspase-A/-B in a sequential manner

Immunofluorescence results revealed perfect DrCaspase-A/-B colocalization with the DrASC speck, suggesting DrNLRP3-DrASC inflammasome as a platform for DrCaspase-A/-B recruitment (Fig. 4, A and B). Nevertheless, DrCaspase-A and DrCaspase-B are almost independently localized in an inflammasome, and only one inflammasome can be assembled in one cell (Fig. 4C). The percentage of DrCaspase-A–associating cells (∼20%) was higher than that of DrCaspase-B–associating cells (∼7%). By introducing two chimera bPYD–CasA and aPYD–CasB caspases in which the PYD of DrCaspase-A and DrCaspase-B was replaced by each other, these two caspases were still separately colocalized with the DrASC speck, whereas the percentage of aPYD–CasB-associating cells (∼28%) exceeded that of bPYD–CasA-associating cells (∼6%). This finding indicated that the privilege of DrCaspase-A was deprived by DrCaspase-B when their PYD domains were exchanged (Fig. 4D). Given that DrASC PYD shared a higher similarity (82.92%) to DrCaspase-A PYD compared with that of DrCaspase-B (55.46%) (Fig. S1C), the priority of DrCaspase-A into DrNLRP3-DrASC inflammasome might be determined by the high degree of similarity, which provided strong hydrophobic (from Leu¹⁶, Leu²¹, Ile⁴⁹, Val⁵⁷, and Ile⁷⁵) and surface charge (from Arg²², Lys²³, Glu⁴³, and Asp⁵⁰) effects on the homotypic PYD–PYD interaction. The fluorescence recovery after photobleaching (FRAP) assay revealed that DrCaspase-A/-B displayed fluorescence recovery in foci within the inflammasome in ∼150 s after photobleaching (Fig. 4, E, F, and H). This outcome supported the dynamic recruitment of DrCaspase-A/-B into the inflammasome. By contrast, DrNLRP3 and DrASC did not show fluorescence recovery (Fig. 4, G and H). This observation indicated that DrNLRP3-DrASC inflammasome maintained a stable structure that remained unmoved once it is organized. Collectively, DrCaspase-A and DrCaspase-B were dynamically and sequentially recruited into the DrNLRP3-DrASC inflammasome, with preference for DrCaspase-A, followed by a replacement of DrCaspase-B after DrCaspase-A was released from the inflammasome.

Figure 4.

Recruitment of DrCaspase-A/-B into DrNLRP3-DrASC inflammasome in a sequential manner. A and B, DrNLRP3-HA, DrASC-Myc, and DrCaspase-A-Flag (A) or DrCaspase-B-Flag (B) coexpression in HEK293T cells elicited the DrCaspase-A/-B colocalization with the DrASC speck. C, DrNLRP3-HA, DrASC-HA, DrCaspase-A-Myc, and DrCaspase-B-Flag coexpression in HEK293T cells elicited the formation of DrCaspase-B (white arrowheads) or DrCaspase-A (white arrows) specks in the cells. D, coexpression of DrNLRP3-HA, DrASC-HA, aPYD–CasB-Flag, and bPYD–CasA-Myc in HEK293T cells elicited the formation of aPYDCasB (white arrowheads) or bPYDCasA (white arrows) specks in the cells. E–G, FRAP of DrNLRP3 inflammasome. Bleaching was performed after HEK293T cells stably expressing DrNLRP3–GFP, DrASC–GFP, DrCaspase-A–RFP, or DrCaspase-B–RFP. Time-lapse micrographs of DrNLRP3–DrASC–DrCaspase-A (E), DrNLRP3–DrASC–DrCaspase-B (F), and DrNLRP3–DrASC (G) punctum formation after bleaching. Arrows indicate punctum. Scale bar, 5 or 10 μm. These images are representative of at least 10 photobleached cells mentioned previously. H, fluorescence intensities of DrNLRP3–DrASC and DrCaspase-A/-B specks over the time course of 5 min after bleaching. Each data point shows the mean ± S.D. with at least three replicates.

DrNLRP3 contributes to proDrIL-1β maturation in a DrASC-dependent manner

The above results showed that DrNLRP3 acted as an initiator that organized a DrASC-dependent DrNLRP3 inflammasome, leading to the self-cleavage and activation of DrCaspase-A/-B. This phenomenon might further contribute to proDrIL-1β maturation. For clarification, DrNLRP3, DrASC, DrCaspase-A/-B, and proDrIL-1β were coexpressed in HEK293T cells in different combinations, and proDrIL-1β maturation was determined through Western blot analysis. As expected with the DrNLRP3, DrASC, and DrCaspase-A/-B coexpressions, proDrIL-1β (31 kDa) was cleaved into an 18-kDa mature form, accompanied by the pro-DrCaspase-A/-B (45/47 kDa) activation through self-cleavage into a p35 product (Fig. 5A). However, proDrIL-1β was partially processed into a 20-kDa product (the first cleavage product at the Asp¹⁰⁴ residue of proDrIL-1β) by the activated DrCaspase-A alone (Fig. 5B). This product coexisted with several other mid-forms (25–30 kDa) known as the cleavage products of some other proteases existing in cells, including neutrophil elastase, proteinase 3, cathepsins G/D, granzyme A, and matrix metalloproteinases. By contrast, the activated DrCaspase-B alone failed in the proDrIL-1β cleavage (Fig. 5C). In the absence of DrNLRP3 and/or DrASC, no any mature/partial forms of proDrIL-1β (18/20 kDa) and DrCaspase-A/-B (35 kDa) were detected. These findings indicated that DrNLRP3-DrASC inflammasome contributed to pro-DrCaspase-A/-B activation through self-hydrolyzation, which leads to proDrIL-1β maturation. Interestingly, the DrIL-1β maturation could be enhanced when DrNLRP1 was introduced into the DrNLRP3 inflammasome (Fig. 5A). In this case, a chimera inflammasome with DrNLRP3 and DrNLRP1 was detected, which perfectly encircled the ASC speck (Fig. S4, A–D). Functionally, DrNLRP3 was found to preferentially activate DrCaspase-B, and DrNLRP1 preferred DrCaspase-A (Fig. S4, E and F). When DrNLRP3 associates with DrNLRP1, the DrCaspase-A and DrCaspase-B activation was enhanced synchronously (Fig. S4, G and H). This functional compensation might contribute to the enhancement of DrIL-1β maturation.

Figure 5.

DrNLRP3 contribution to the proDrIL-1β maturation in a DrASC-dependent manner. HEK293T cells were transfected with a pcDNA3.1–DrIL-1β construct alone or with pCMV–DrNLRP3, pCMV–DrNLRP1, pCMV–DrASC, pCMV–DrCaspase-A, and pCMV–DrCaspase-B. At 24 h post-transfection, immunoblot analysis was performed on the cell lysates with mouse anti-Flag or anti-Myc monoclonal Ab. A, ProDrIL-1β cleavage triggered by activated DrCaspase-A and DrCaspase-B. B, ProDrIL-1β cleavage triggered by DrNLRP3–DrASC-activated DrCaspase-A alone. C, ProDrIL-1β cleavage triggered by DrNLRP3–DrASC-activated DrCaspase-B alone. Blots were re-probed for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a loading control. The results are representative of three independent experiments, as described under “Experimental procedures.” Bar charts under A–C showed the relative density of the cleavage product of DrIL-1β in the blots. Each data point shows the mean ± S.D. with three replicates.

DrNLRP3 organizes DrCaspase-B to activate DrGSDMEa/b in a DrASC-independent manner

Although the DrNLRP3-DrASC inflammasome contributed to DrIL-1β maturation in cells, a minimal amount of mature DrIL-1β was detected in the supernatant out of the cells. This phenomenon might be attributed to the absence of GSDMD/GSDME mediated membrane perforation, which was essential for cell pyroptosis and IL-1β secretion. Thus, DrGSDMEa/b was introduced in the study. As expected, DrGSDMEa/b coexpression with DrNLRP3, DrASC, and DrCaspase-A/-B significantly induced pyroptosis, as determined by cytotoxicity accompanied with lactate dehydrogenase (LDH) release and pyroptotic morphology (Fig. 6, A and B). DrASC and DrCaspase-A removal from the components did not impair pyroptosis, whereas DrCaspase-B withdrawal significantly inhibited pyroptosis. Strikingly, the cells that expressed DrNLRP3, DrCaspase-B, and DrGSDMEa/b induced significant pyroptosis, in which DrGSDMEb had a higher activity (>50%) than DrGSDMEa (Fig. 6A). In this case, DrGSDMEa/b was cleaved into an N-terminal product by full-length DrCaspase-B without undergoing self-cleavage in the absence of DrASC. This phenomenon was evidenced by the observation that a quaternary mutant DrCaspase-B (Caspase-B–4DA) also cleaved DrGSDMEa/b into an N-terminal product (Fig. 6C). This Caspase-B–4DA was generated by using alanine to substitute the four potential autocleavage sites (Asp^{130, 137, 308, 314}) to prevent DrCaspase-B from self-cleavage. The full-length DrCaspase-B activity was also examined using DrCaspase-B fluorescent substrate and DrGSDMEa/b-mediated pyroptosis (Fig. 6, D and E). In addition, the DrCaspase-B P20 and P10 combination showed weak DrGSDMEb cleavage and pyroptosis (Fig. 6, F and G). These findings indicated that the association of DrNLRP3 with full-length DrCaspase-B was sufficient for DrGSDMEa/b activation to elicit pyroptosis. DrNLRP3 PYD shared an 89.33% similarity with DrCaspase-B PYD, and this value was higher than those of DrASC (55.46%) and DrCaspase-A (49.76%) (Fig. S1C). Consequently, DrNLRP3 was supposed to preferentially associate with DrCaspase-B without intervention of DrASC and DrCaspase-A through a more homotypic PYD–PYD interaction. As a support, the chimera DrCaspase-B (aPYD–CasB), whose PYD was replaced with that of DrCaspase-A, weakly interacted with DrNLRP3 and triggered DrGSDMEb-mediated pyroptosis (Fig. 6H).

Figure 6.

DrNLRP3 activates DrCaspase-B to cleave DrGSDMEa/b in a DrASC-independent manner. A, HEK293T cells were transfected with different combinations of plasmids of DrNLRP3, DrASC, DrCaspase-A/-B, and DrGSDMEa/b. Supernatants from the indicated cells were analyzed for cell death, as measured by LDH release. B, images were taken after 4.5 μm PI were added to the indicated cells. The dyed cells indicate the loss of plasma membrane integrity and exhibit pyroptotic-like features. Scale bar, 10/100 μm. C, DrGSDMEa/b cleavage by DrNLRP3 and DrCaspase-B (or the mutant CasB-4DA) was analyzed by immunoblotting. D, DrNLRP3 activating DrCaspase-B or CasB-4DA was detected by specific Ac-WEHD-AFC fluorescent substrate. E, supernatants from the indicated cells were analyzed for cell death, as measured by the LDH release. F, supernatants from the indicated cells coexpressed with different combinations of DrNLRP3 and DrCaspase-B (including CaspB-P35, CaspB-P20, and CaspB-P10) were analyzed for cell death, as measured by LDH release. G, DrGSDMEb cleavage by DrNLRP3 and DrCaspase-B in different combinations of CaspB-P35, CaspB-P20, and CaspB-P10 was analyzed by immunoblotting. H, supernatants from the HEK293T cells transfected with PCMV-GSDMEb/Caspase-B/aPYD–CasB/NLRP3 were analyzed for cell death, as measured by LDH release. I, protein–protein interactions between DrNLRP3 and DrCaspase-B but not their PYD-lacking mutants. J, DrNLRP3 and mutants activate DrCaspase-B detected by specific Ac-WEHD-AFC fluorescent substrate. K, supernatants from the indicated cells were analyzed for cell death, as measured by the LDH release. All the above results are representative of at least three independent experiments, and error bars denote the S.D. of triplicate wells. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

In multiple alignment analysis, three basic amino acids (Lys^{37, 44} and Arg⁷⁰) and six acidic amino acids (Asp^{6, 45, 48, 51, 60} and Glu⁴¹) are conserved in PYDs of DrNLRP3 and DrCaspase-A. Furthermore, 11 basic amino acids (Lys^{17, 28, 39, 44, 73} and Arg^{21, 37, 46, 70, 77, 80}) and 11 acidic amino acids (Asp^{6, 45, 48, 51, 60} and Glu^{9, 13, 15, 39, 41, 82}) are conserved in PYDs of DrNLRP3 and DrCaspase-B (Fig. S1D). The ratio of negatively-charged residues to positively charged residues in PYDs is 1:2 for DrNLRP3 and DrCaspase-A but 1:1 for DrNLRP3 and DrCaspase-B. This result means that isoelectric repulsion hardly exists between the PYDs of DrNLRP3 and DrCaspase-B compared with that of DrNLRP3 and DrCaspase-A. This phenomenon explains a strong interaction between DrNLRP3 and DrCaspase-B. Structurally, the PYD of DrNLRP3 is much more similar to that of DrCaspase-B rather than DrCaspase-A (Fig. S1E). This condition is beneficial to the association of DrNLRP3 with DrCaspase-B through the PYD–PYD interaction. Indeed, Co-IP assay clearly shows the interaction between DrNLRP3 and DrCaspase-B but not between DrNLRP3-ΔPYD and DrCaspase-B-ΔPYD mutants (Fig. 6I). With the lack of PYD and NACHT domains, DrNLRP3 cannot interact and activate DrCaspase-B, which decreases the DrGSDMEb-mediated cytotoxicity, whereas the lack of LRR and B30.2 domains does not affect cytotoxicity (Fig. 6, J and K). The DrNLRP3-ΔLRR mutant inhibits DrASC nucleation and pro-DrCaspase-A/-B hydrolyzation but maintains the activation DrCaspase-B in full length (Figs. 2, C and F, and 3D), further demonstrating that DrNLRP3 organizes full-length DrCaspase-B to activate DrGSDMEa/b in a DrASC-independent manner.

DrNLRP3 coordinates DrIL-1β maturation and cell pyroptosis

Given that DrIL-1β maturation occurs in a DrNLRP3-DrASC–dependent manner, whereas DrGSDMEa/b triggers DrIL-1β release and cytotoxicity/pyroptosis through a DrASC-independent way, the tight coupling of these two events was further explored. DrCaspaseA-5DA and DrCaspaseB-4DA mutants that lack autocleavage sites were introduced to exclude the DrNLRP3-DrASC–dependent reaction that depends on DrCaspase-A/-B autocleavage but retains DrASC-independent reaction that is independent of DrCaspaseB cleavage. When cells were coexpressed with a combination of DrNLRP3, DrASC, DrCaspase-A, DrCaspase-B, proDrIL-1β, and DrGSDMEb, a complete proDrIL-1β cleavage product (18 kDa) occurred, accompanied by the DrIL-1β secretion and LDH release (Fig. 7). When DrCaspase-B was replaced by CaspaseB-4DA in the combination, the proDrIL-1β was only cleaved into the 20-kDa–mediated form that was merely processed by the activated DrCaspase-A in a DrNLRP3-DrASC–dependent manner (Fig. 7A). This process declined the DrIL-1β secretion but did not influence the LDH release, suggesting that DrGSDMEb-mediated perforation occurred in a DrCaspaseB noncleavage manner that is DrASC-independent (Fig. 7, B and C). This finding means that the DrASC-dependent DrIL-1β maturation and DrASC-independent DrIL-1β secretion and cell pyroptosis were coordinated by DrNLRP3. The results from other combinations provide further support for this conclusion. For example, when DrCaspase-A was replaced by CaspaseA-5DA, DrIL-1β could not be cleaved and released, although cytotoxicity still appeared. A similar result was also observed when DrCaspase-A and DrCaspase-B were replaced by CaspaseA-5DA and CaspaseB-4DA together or with the DrASC removed from the combination. Without the DrNLRP3 expression, no detectable cytotoxicity, DrIL-1β maturation, and secretion were observed. Meanwhile, the absence of DrGSDMEb expression did not influence the DrIL-1β maturation but markedly decreased the cytotoxicity and DrIL-1β release. A schematic diagram for the two types of DrNLRP3 inflammasome is presented in Fig. 8.

Figure 7.

Coordination of DrNLRP3 inflammasome between mature-formed DrIL-1β secretion and cell pyroptosis. A, HEK293T cells were transfected with a pcDNA3.1-DrIL-1β construct with pCMV-DrNLRP3, pCMV-DrASC, pCMV-DrCaspase-A (pCMV-CasA-5DA), pCMV-DrCaspase-B (pCMV-CasB-4DA), and pCMV-DrGSDMEb. Immunoblot analysis was performed at 24 h post-transfection on the cell lysates to detect the proDrIL-1β cleavage triggered by DrNLRP3 inflammasome. B, supernatants from the indicated cells were analyzed for cell death, as measured by the LDH release. C, levels of soluble IL-1β in culture supernatants were analyzed by ELISA by using DrIL-1β polyclonal antibody. Each data point shows the mean ± S.D. with three replicates. **, p < 0.01; ***, p < 0.001.

Figure 8.

Schematic for the functional roles of DrNLRP3 inflammasome in DrASC-dependent and DrASC-independent manners.

In vivo examination of DrNLRP3 inflammasome

DrNLRP3 inflammasome was evaluated in vivo in the zebrafish embryo model. The DrNLRP3 inflammasome occurrence is optimized at 6 hpf embryos or 72 hpf larvae with Edwardsiella tarda immersion infection (10⁸ CFU/ml) for 1–4 h as determined by significantly increased DrCaspase-A/-B activity at those times (Fig. 9, A and B). DrNLRP3 overexpression alone in embryos with infection only augmented endogenous DrCaspase-B activation. However, DrNLRP3 coexpression with DrASC promoted both DrCaspase-A and DrCaspase-B activation, accompanied by an increased DrIL-1β maturation (Fig. 9C). These observations support the existence of DrASC-dependent and DrASC-independent types for DrCaspase-A/-B activation and DrIL-1β maturation. Structurally, DrNLRP3- and DrASC-colocalized specks were clearly observed in the embryos (Fig. 9D). DrNLRP3 knockdown significantly inhibited (p < 0.05) DrCaspase-A/-B activation upon E. tarda infection. This inhibition was rescued by MO-resistant mRNAs in a dose-dependent manner (Fig. 9, E, F, H, and I and Fig. S5A). Correspondingly, DrNLRP3 knockdown increased the mortality of embryos with E. tarda infection, which was restored by MO-resistant mRNAs (Fig. 9, G and J). In addition, DrCaspase-A/-B activation was abrogated in DrNLRP3 morphants stimulated by LPS and H₂O₂ but not by muramyl dipeptide (MDP), bacterial DNA, and ATP, although the latter exhibited stimulatory effects on DrCaspase-A/-B activation at varying degrees in the embryos without DrNLRP3 knockdown (Fig. 9, K and L). This finding suggests the contribution of LPS and cellular oxidation to DrNLRP3 inflammasome activation. Next, cell pyroptosis and DrIL-1β maturation, two downstream events of the DrNLRP3 inflammasome, were examined. The pyroptosis occurrence was optimized in 72 hpf embryos with E. tarda infection (10⁸ CFU/ml) for 4 h. Therein, the expression of DrNLRP3, DrCaspase-B, and DrGSDMEb genes but not apoptosis-associated genes (such as Fas/FasL and Bcl2) was significantly (p < 0.001) augmented, and considerable cells presented pyroptotic morphology (Figs. 9M and 10A). FCM analysis showed that the percentage of the PI-positive cells increased from 8.80 to 34.8% in the embryos upon infection. This value decreased to 19.7% in DrNLRP3 morphants. This result implied the involvement of DrNLRP3 in pyroptosis (Fig. 10B). However, DrIL-1β maturation was suppressed in DrCaspase-A and DrCaspase-B morphants (Fig. 10C), whereas cell pyroptosis declined in DrCaspase-B morphants (13.2%) but not in DrCaspase-A morphants (30.6%) (Fig. 10D). This result indicated that pyroptosis depends on DrCaspase-B rather than DrCaspase-A. Furthermore, a DrASC mutant was generated by CRISPR/Cas9, resulting in transcripts with a nonsense codon within the first exon (Fig. S5, B–D). DrCaspase-A/-B activation and DrIL-1β maturation were diminished in the DrASC knockout larvae (Fig. 10, E and F), whereas pyroptosis was not inhibited by the DrASC knockout (Fig. 10G). This phenomenon supported the notion that DrIL-1β maturation is DrASC-dependent, and cell pyroptosis is DrASC-independent.

Figure 9.

In vivo determination of DrNLRP3 inflammasome. A and B, fluorogenic substrate detection of the DrCaspase-A/-B activation in 6-hpf embryos (A) or 72 hpf larvae (B) after E. tarda infection at 10⁸ CFU/ml for 0–6 h. The sample number for each group was 20–100 zebrafish embryos or larvae, and each data point shows the mean ± S.D. with three replicates. C, DrNLRP3 and DrASC coexpression in vivo increased the DrCaspase-A/-B activation level and promoted the DrIL-1β maturation under E. tarda infection in 72-hpf larvae (n = 100). Data are representative of three independent experiments as mean ± S.D. (*, p < 0.05; **, p < 0.01; ***, p < 0.001). D, Flag-tagged DrNLRP3 and Myc-tagged DrASC coexpression in vivo triggers the DrNLRP3-DrASC speck nucleation in 72-hpf zebrafish larvae (n = 100). E and H, DrNLRP3 knockdown by DrNLRP3-MO decreased the DrCaspase-A/-B activation in 6-hpf embryos (n = 50) (E) or 72 hpf larvae (n = 50) (H) after E. tarda infection for 40 min or 4 h. Each data point shows the mean ± S.D. with three replicates (*, p < 0.05; **, p < 0.01). F and I, MO-resistant DrNLRP3 mRNA rescued the DrCaspase-A/-B activation in 6-hpf embryos (n = 50) (F) or 72-hpf larvae (n = 50) (I) after E. tarda infection for 40 min or 4 h. Each data point shows the mean ± S.D. with three replicates (*, p < 0.05; **, p < 0.01). G and J, RSRs of 6-hpf embryos (G) or 72-hpf larvae (J) after E. tarda infection at 10⁶ CFU/ml for 12 h. Zebrafish embryos were microinjected with standard MO (Control), DrNLRP3-MO (NLRP3-MO), or both with the corresponding mRNA (NLRP3-(MO+mRNA)). Mortality in each group was monitored during the 1-h period at one interval. The results are performed in triplicate with 100 embryos per group. K and L, evaluation of bacterial LPS, MDP, and DNA and cellular metabolites H₂O₂ and ATP for the DrNLRP3 inflammasome activation in 6-hpf embryos (n = 50) via in vivo knockdown assay. Each data point shows the mean ± S.D. with three replicates (***, p < 0.001). M, fold change of mRNA levels of the genes involved in pyroptosis and apoptosis in zebrafish 72-hpf larvae (n = 20) after E. tarda infection for 4 h. The fold change of the relative expression levels was calculated by the 2^−ΔΔCt method with β-actin for normalization. Data are representative of three independent experiments as mean ± S.D. (*, p < 0.05; **, p < 0.01; ***, p < 0.001).

Figure 10.

In vivo examination of DrNLRP3 inflammasome in DrASC-independent manner. A, images were taken after PI was added to the 72-hpf zebrafish larval cells infected by E. tarda (10⁸ CFU/ml) for 4 h. The dyed cells exhibit pyroptotic-like features. Scale bar, 10 μm. The sample number for each group was 20–100 zebrafish larvae, and each image is representative of three independent experiments. B, flow cytometry with PI detected the cell pyroptosis in 72-hpf zebrafish larvae (n = 100) with E. tarda immersion infection (10⁸ CFU/ml) for 4 h. C, DrCaspase-A (CasA-MO) or DrCaspase-B (CasB-MO) knockdown decreased the DrCaspase-A/-B activation and the DrIL-1β maturation in zebrafish embryos (n = 100) after E. tarda infection. Data are representative of three independent experiments as mean ± S.D. (**, p < 0.01). D, flow cytometry with PI detected the cell pyroptosis in CasA-MO or CasB-MO zebrafish larvae (n = 100) with E. tarda immersion infection (10⁸ CFU/ml) for 4 h. E and F, DrASC knockout (ASC-KO) decreased the DrCaspase-A/-B activation (E) and the DrIL-1β maturation (F) in zebrafish embryos (n = 100) after E. tarda infection. Data are representative of three independent experiments as mean ± S.D. (*, p < 0.05). G, flow cytometry with PI detected the cell pyroptosis in ASC-KO zebrafish larvae (n = 100) with E. tarda immersion infection (10⁸ CFU/ml) for 4 h.

Discussion

In this study, we have identified an NLRP3 homolog (DrNLRP3) from zebrafish, which shares an overall conservative chromosomal synteny, gene organization, protein domain architecture, and tertiary structure to mammalian NLRP3s, except for a B30.2 domain that is unique in fish (28). This DrNLRP3 can trigger the assembly of a classical inflammasome speck structure (DrNLRP3-DrASC inflammasome) in a DrASC-dependent manner, in which DrCaspase-A/-B was activated by self-cleavage into the p35/p10 subunits and further cleaves proDrIL-1β at the Asp¹⁰⁴/Asp¹²² residues to generate an 18-kDa mature-formed DrIL-1β. The DrNLRP3-DrASC inflammasome assembly and DrCaspase-A/-B autoproteolytic activation require the PYD, NACHT, and LRR domains of DrNLRP3 as observed in mammalian NLRP3s, in which the PYD is believed to recruit ASC, the NACHT bridges oligomerization, and LRR recognizes stimulus signals (29–32). Meanwhile, the B30.2 domain is not functional in inflammasome assembly and DrCaspase-A/-B activation. The DrASC nucleation is the key process of DrNLRP3-DrASC inflammasome formation, in which DrNLRP3 first associates with a linker DrASC, which further recruits other DrASCs for their elongation to form a DrASC filament. The PYD cluster of DrASC filament finally recruits the PYDs of DrCaspase-A/-B. The DrNLRP3-DrASC inflammasome recruits DrCaspase-A/-B in a two-step sequential manner, with a preference for DrCaspase-A and a subsequent choice for DrCaspase-B. This process ensures proDrIL-1β maturation that requires cleavage at the Asp¹⁰⁴/Asp¹²² sites by activated DrCaspase-A/-B in turn. This sequential activation is determined by the homotypic degree between the PYDs of DrASC and DrCaspases, in which the number of conservative hydrophobic and charged amino acid residues is included. Notably, the sequential activation of DrCaspase-A/-B is also observed in an NLRP1 inflammasome that we recently identified from zebrafish, suggesting its universal significance in fish (18). Interestingly, a synergistic effect was detected between DrNLRP3 and DrNLRP1 by forming a chimera inflammasome, which promoted DrCaspase-A/-B autoproteolytic activation and DrIL-1β maturation. The functional collaboration of different inflammasomes was also observed among AIM2, NLRC4, and NLRP3 in various cells (33–35). However, the details of these collaborations remain to be addressed.

Meanwhile, DrNLRP3 possesses the ability to directly recruit and activate DrCaspase-B, but not DrCaspase-A, without undergoing autoproteolysis and the help of DrASC. This full-length DrCaspase-B activated by DrNLRP3 alone contributes to the DrGSDMEa/b cleavage. This situation induces DrIL-1β secretion and cell pyroptosis. However, the full-length DrCaspase-B fails in proDrIL-1β cleavage, and the latter requires DrCaspase-A/-B autoproteolysis in the presence of DrASC. Thus, it is evident that DrNLRP3 participates in proDrIL-1β maturation and DrIL-1β secretion followed by cell pyroptosis in two different ways. Specifically, proDrIL-1β maturation requires DrNLRP3-DrASC inflammasome formation and depends on DrCaspase-A/-B autoproteolytic activation in a DrASC-dependent manner. However, DrIL-1β secretion and cell pyroptosis require the occurrence of a DrNLRP3–DrCaspase-B complex, in which DrCaspase-B was activated in a full-length form to cleave DrGSDMEa/b in a DrASC-independent manner. The mature-formed DrIL-1β can finally be released from the cell only when these two types are tightly coordinated. These conclusions were supported by introducing two noncleavable DrCaspase-A/-B mutants (CasA-5DA and CasB-4DA) into the experiment. DrGSDMEb possesses a better ability to elicit cell pyroptosis compared with DrGSDMEa. The LRR domain of DrNLRP3 necessary for DrASC nucleation and DrCaspase-A/-B autoproteolysis does not influence the full-length DrCaspase-B activation and DrGSDMEb-mediated pyroptosis. This phenomenon signifies the existence of a unique mechanism underlying full-length DrCaspase-B activation. This DrCaspase-B activation might require a conformational transition from an inactive to an active form by some unknown allosteric triggers. Similar results were also observed in several other inflammasomes, such as NLRP1, NLRC4, and Nlrp1b, in which Caspase-1/5 was activated without self-cleavage (36, 37). For example, by reconstituting Caspase-1 in cells with a noncleavable mutant, the noncleavable Caspase-1 can be fully activated in response to Nlrp1b activators (38). Consistent with this study, the ASC knockout macrophages severely impaired their ability to process and secrete mature IL-1β but promoted rapid cell death as WT macrophages upon bacterial infection. In this case, ASC-driven Caspase-1 autoproteolysis is differentially required for cytokine maturation and pyroptosis (36). These observations indicate the universal existence of the uncleaved caspase activation. Further study is therefore needed to clarify the mechanism behind this type of activation.

How mature-formed IL-1β is released from cells is a basic question because of the lack of a signal peptide at the N terminus of this molecule. This concern has remained unclear for a long time until the discovery of the GSDMD/E-mediated pore-forming membrane initiated by Caspase-4/5/11- and Caspase-3–dependent noncanonical inflammasomes in mammals (16, 20). This finding suggests a potential correlation between a noncanonical inflammasome for GSDMD/E activation and a canonical inflammasome for IL-1β maturation despite the poorly understood manual coupling of these two kinds of inflammasomes. Our study showed that NLRP3 inflammasome alone can perform IL-1β maturation and GSDME activation without the help of other inflammasomes. This dual functional performance of DrNLRP3 enables a coordinator between DrIL-1β–induced inflammation and DrGSDME-mediated pyroptosis. To date, numerous gasdermin family proteins, such as GSDMD, GSDMA, GSDMB, GSDMC, and GSDME, are activated by Caspase-4/5/11 and Caspase-3 noncanonical inflammasomes to induce pyroptosis in mammals (4). However, only two GSDMEa/b isoforms and a DFNB59 family member exist in zebrafish and other teleost fish. DrGSDMEa/b becomes the major candidate that mediates pyroptosis in fish because DFNB59 lacks an Asp cleavage site and cannot be cleaved into an effective N-terminal domain (4, 20). This finding means that GSDME is an ancient gasdermin member involved in pyroptosis during vertebrate evolution. Thus, teleost fish can become an attractive model population to understand the evolutionary correlation between gasdermin proteins and inflammatory caspases regulated by various noncanonical and/or canonical inflammasomes.

The DrNLRP3 inflammasome is finally examined in vivo by using the zebrafish embryo model. Typical DrNLRP3-DrASC inflammasome was detected in embryos, which triggers DrCaspase-A/-B activation, DrIL-1β maturation, and cell pyroptosis under E. tarda infection. These activities can be diminished by antisense MO-based DrNLRP3 knockdown and rescued by MO-resistant mRNAs. Meanwhile, the high mortality of DrNLRP3 morphants with E. tarda infection is restored by administering MO-resistant mRNAs. These findings verify the functional roles of DrNLRP3 inflammasome in antibacterial immunity in vivo. Furthermore, by generating a DrASC CRISPR mutant (DrASC^−/−) with a 2-bp deletion in zebrafish by using Cas9/gRNA technology, it was found that the DrASC^−/− mutant only abrogates the DrIL-1β maturation but maintains the ability of cell pyroptosis. This result confirms the existence of a DrASC-independent pyroptosis pathway in vivo. Mechanistically, LPS and H₂O₂ administration into the embryos significantly activates the DrNLRP3 inflammasome. This outcome suggests that pathogen-associated molecular patterns and cellular metabolic homeostasis (such as redox state) can elicit the DrNLRP3 inflammasome activation as observed in mammalian NLRP3 inflammasomes (39, 40). Interestingly, the mouse NLRP3 (MmNLRP3) can replace DrNLRP3 for DrASC-dependent inflammasome aggregation and DrASC-independent but DrCaspaseB-dependent pyroptotic cell death (Fig. 11). This result implies the conservation between fish and mammalian NLRP3 inflammasomes throughout vertebrate evolution.

Figure 11.

Functional substitution of mouse NLRP3 (MmNLRP3) to DrNLRP3 in DrASC nucleation and cell pyroptosis. A, DrASC and DrNLRP3 coexpression in HEK293T cells elicits DrASC speck aggregation. B, DrASC and MmNLRP3 coexpression instead of DrNLRP3 also elicits DrASC speck aggregation. C, speck-forming rates of DrNLRP3 or MmNLRP3 with DrASC. Data are representative of three independent experiments as mean ± S.D. (**, p < 0.01). D, DrNLRP3 or MmNLRP3 directly activates DrCaspase-B and triggers DrGSDMEb-dependent cell pyroptosis. Images were captured under a laser-scanning confocal microscopy (Zeiss LSM-710; original magnification, ×630, scale bars represent 50 or 10 μm). The DrCaspase-B activity was detected and expressed as the fold induction over the control as described under “Experimental procedures.” Each data point shows the mean ± S.D. with three replicates. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

In conclusion, our study demonstrates the origin of NLRP3 inflammasome in teleost fish and reveals its role in ASC-dependent IL-1β maturation and GSDME-mediated pyroptosis. This finding enriches our current knowledge on NLRP3 inflammasome biology. Given that the NLRP3 inflammasome is also closely-associated with various diseases, such as type II diabetes and inflammatory bowel disease (41, 42), teleost fish will become a new research model for these kinds of diseases.

Experimental procedures

Experimental fish and embryo

WT AB zebrafish (Danio rerio) with body lengths of 3–4 cm and weights of 0.5–1.0 g were maintained in circulating water at 28 °C under standard conditions. The fish that exhibited healthy appearance and activity were used for the study. Zebrafish embryos were collected at optimized hours post-fertilization (6–96 hpf). The experiments were conducted in accordance with legal regulations and ethical approval.

Molecular cloning

The Genome and Expressed Sequence Tags (EST) databases maintained by the National Center for Biotechnology Information (NCBI), the University of California Santa Cruz (UCSC), and Ensembl were used to predict and confirm NLRP3 and GSDME homologs in zebrafish, as described previously (43). Briefly, with the PYD and NACHT domain sequence of human NLRP3 as a query, a candidate DrNLRP3 gene named si:dkey-156m2.3 was retrieved from the zebrafish genome and EST databases by using Genscan and Ensemble (TBLASTN) programs. With the full-length sequences of human GSDME or gasdermin D (GSDMD) as queries, three candidate gasdermin members (i.e. DrGSDMEa, DrGSDMEb, and DrDFNB59) were predicted from the zebrafish genome database. Given that mammalian GSDME can trigger cell pyroptosis, but DFNB59 cannot, due to its inability to cleave the N-terminal structure by lacking an aspartic acid cleavage site, DrGSDMEa and DrGSDMEb were chosen for further identification. Total RNA was isolated from zebrafish embryos and tissues by using an RNAiso Plus kit (Takara Bio). The cDNAs of DrNLRP3 and DrGSDMEa/b were amplified by RT-PCR according to the homologous sequences predicted previously. The PCR products were purified and inserted into the pGEM-T easy vector (Promega) and sequenced on a 3730xl DNA analyzer (Applied Biosystems). The primers used in cloning are shown in Table S1.

Bioinformatics analysis

Genome assemblies and locations of NLRP3 and GSDME genes in human and zebrafish genomes were retrieved from the University of California at Santa Cruz (UCSC) genome bioinformatics website and Genome Data Viewer in the National Center for Biotechnology Information (NCBI). Gene organizations (intron/exon boundaries) were elucidated by comparing DrNLRP3 and DrGSDMEa/b cDNAs with genome sequences from UCSC, and figures were drawn using GeneMapper 2.5 (44). Full-length cDNAs of DrNLRP3 and DrGSDMEa/b were assembled using the CAP3 Sequence Assembly Program (45). Multiple alignments of DrNLRP3 and DrGSDMEa/b with their mammalian counterparts were analyzed by using the Clustal X program (version 2.0) (46). Phylogenetic trees were constructed by using MEGA 5.0 with the maximum likelihood method. The node values represent the percentage bootstrap confidence derived from 1000 replicates (47). The percentage of amino acid sequence identity was calculated using the MEGALIGN program from DNASTAR. The potential functional motifs in DrNLRP3 and DrGSDMEa/b proteins were predicted by using the Pfam 31.0 and Conserved Domains of NCBI online software with protein sequence (48). The domain structures of DrNLRP3 and DrGSDMEa/b were analyzed using SWISS-MODEL, and the whole-protein structure of DrNLRP3 was analyzed by using I-TASSER (49, 50). The tertiary structural figures were reviewed and colored in PyMOL software (51).

Plasmid constructions

The full-length encoding sequences of DrNLRP3 and DrGSDMEa/b and the partial encoding sequences of DrNLRP3 with the deletion of PYD, NACHT, LRR, and B30.2 domains were inserted into pCMV (Stratagene) or pcDNA3.1 (Invitrogen) vectors with Flag/Myc/HA-tags at the N/C termini. The mutant constructs were named pCMV-DrNLRP3-ΔPYD, pCMV-DrNLRP3-ΔNACHT, pCMV-DrNLRP3-ΔLRR, and pCMV-DrNLRP3-ΔB30.2. A quinary mutant of DrCaspase-B (named as CaspaseB-4DA) in which four aspartic acids (Asp¹³⁰, Asp¹³⁷, Asp³⁰⁸, and Asp³¹⁴) were substituted by alanines was constructed using a QuikSite-directed mutagenesis kit (Beyotime). DrASC, DrCaspase-A, DrCaspase-B, proDrIL-1β, and DrNLRP1 encoding constructs and their mutant constructs, including DrASC-ΔPYD, DrASC-ΔCARD, aPYD–CasB, bPYD–CasA, and CaspaseA-5DA, were generated in our previous study. The plasmids for transfection and microinjection were prepared free of endotoxin by using endo-free plasmid mini kit II (Omega Bio-tek). The primers used in cloning and construct generation are listed in Table S1.

Quantitative real-time PCR

The transcripts of DrNLRP3, DrGSDMEa/b, DrFas/DrFasL, and DrBcl2 genes in zebrafish tissues and embryos were analyzed via quantitative real-time PCR on a Mastercycler EpRealplex instrument (Eppendorf). In brief, all PCR experiments were performed in a total volume of 10 μl by using a SYBR Premix Ex Taq kit (Takara Bio). The reaction mixtures were incubated for 2 min at 95 °C, followed by 40 cycles of 15 s at 95 °C, 15 s at 60 °C, and 20 s at 72 °C. The relative expression levels were calculated using the 2^−ΔCt and 2^−ΔΔCt method with β-actin for normalization. Each PCR trial was run in triplicate parallel reactions and repeated three times. The primers are listed in Table S1, and the efficiency of these primers was checked.

Constitution of DrNLRP3 inflammasome in HEK293T cells

HEK293T cells were seeded into 6-well plates at 5 × 10⁵ per well in DMEM culture medium (HyClone) with 10% fetal bovine serum (Bovogen) at 37 °C in 5% CO₂. After 24 h, cells were transfected with plasmids expressing DrNLRP3 (400 ng), DrASC (200 ng), DrCaspase-A (200 ng), DrCaspase-B (200 ng), and proDrIL-1β (800 ng) by using polyethyleneimine (Invitrogen) (52). After another 24–48 h, cell lysates were prepared for Western blotting or caspase fluorogenic assay.

Caspase assay with fluorogenic substrates

HEK293T cells (one well in a 6-well plate) or zebrafish embryos (∼20 embryos) with transfection or microinjection were harvested and lysed with 100 μl of caspase cell lysis buffer (Enzo Life Sciences). Lysate protein (100 μg) was added to the caspase assay buffer (Enzo Life Sciences) containing 100 μm acetyl–Tyr–Val–Ala–Asp–amido-4-trifluoromethylcoumarin (Ac-YVAD-AFC, specific to DrCaspase-A) or Ac–Trp–Glu–His–Asp–AFC (Ac-WEHD-AFC, specific to DrCaspase-B) (Alexis, San Diego, CA) as described previously (12, 53). After incubation at 37 °C for 2 h, the cleavage of caspase-type–specific substrate emitted a fluorescent signal that was measured with excitation at 400 nm and emission at 505 nm on a Synergy H1 Hybrid Reader (BioTek Instruments). The activation level of DrCaspase-A/-B was calculated as follows: ((experimental group − control group)/control group) × 100%.

Western blot analysis

HEK293T cells or zebrafish embryos with designated treatments were treated with cell lysis buffer for Western blotting and immunoprecipitation (Beyotime) containing protease inhibitor mixture (Roche Applied Science). The proteins were separated through 12% SDS-PAGE and then transferred onto polyvinylidene difluoride transfer membranes (Millipore). The blots were blocked with 5% nonfat dry milk (BBI Life Sciences) and incubated with mouse anti-Flag (1:1000, catalog no. ab125243, Abcam, clone FG4R), anti-Myc (1:1000, catalog no. ab56, Abcam, clone 9E11), and anti-HA monoclonal Abs (1:1000, catalog no. ab1424, Abcam, clone 4C12) or rabbit anti-Flag (1:5000, catalog no. D110005, BBI Life Sciences) and anti-Myc (1:5000, catalog no. D110006, BBI Life Sciences) polyclonal Abs. The objective proteins were visualized with enhanced chemiluminescence reagents (GE Healthcare) by using a digital gel image analysis system (Tanon) after horseradish peroxidase–conjugated goat anti-rabbit (1:8000, catalog no. ab6721, Abcam) or mouse IgG Ab (1:8000, catalog no. ab205719, Abcam) was added. The protein strip was subjected to grayscale quantization by using ImageJ software.

Co-immunoprecipitation assay

Co-IP was performed to detect the interaction among DrNLRP3, DrASC, and DrCaspase-A/-B. In this process, HEK293T cells were plated in 10-cm dishes (Corning) and cotransfected with 6 μg of recombinant plasmids or an empty vector as a negative control. After 48 h, the cells were lysed with pre-cooling cell lysis buffer (Enzo Life Sciences). The lysates were incubated with mouse or rabbit Abs (1:200 dilution) at 4 °C overnight. The next day, the mixture of the cell lysates and the antibody were incubated with 50 μl of protein A-agarose beads (Roche Applied Science) for 4 h. The beads were washed three times with lysis buffer, and the obtained samples were analyzed with Western blotting assay. The expression of the transfected plasmids was also analyzed in the whole-cell lysates as an input control.

Immunofluorescence imaging of DrNLRP3-dependent DrASC nucleation

HEK293T cells (1 × 10⁵) were seeded on coverslips in a 24-well plate per well for 24 h. The cells were transfected with plasmids expressing DrNLRP3 (400 ng/ml), DrASC (100 ng/ml), and DrCaspases (100 ng/ml) (54). After 48 h, the cells were fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.1% Triton X-100, and blocked with 2% BSA at 37 °C for 1 h. The cells were then incubated with primary antibodies (rabbit anti-Myc along with mouse anti-Flag) at 4 °C overnight. After washing with phosphate-buffered saline (PBS), the cells were incubated with secondary FITC-conjugated anti-rabbit antibodies (1:100, catalog no. sc-2359, Santa Cruz Biotechnology) and AlexaFluor 594–conjugated anti-mouse (1:1000, catalog no. A-11005, Life Technologies, Inc.) according to the manufacturer's instructions. Finally, the cells were incubated with 0.1% 4′,6-diamidino-2-phenylindole (Invitrogen) for staining the nucleus. The images were captured under a two-photon laser-scanning confocal microscope (Zeiss LSM710, Germany) at ×200 and ×630 magnifications. Three-dimensional super-resolution images were captured using a three-dimensional structured illumination microscope with the N-SIM System (3D-SIM, Nikon) (55, 56). DrASC nucleation was quantified by calculating the speck-forming rate. More than 100 cells with immunofluorescence speck images were counted in each experimental group.

Electroporation for ZF4 cell transfection

Zebrafish ZF4 cells were cultured at 28 °C in a DMEM/F-12 mixture medium (HyClone) with 110 μg/ml sodium pyruvate (Corning) and 10% fetal bovine serum (Bovogen). The cell numbers were counted using a hemocytometer after being digested by 0.25% trypsin (Thermo Fisher Scientific). Approximately 2 × 10⁶ cells were suspended in 200 μl of DMEM/F-12 medium in a 0.4-cm electroporation cuvette and mixed with 20 μg of plasmid DNA. After electroporation (square wave pulses, 270 V, 25 ms, Bio-Rad MicroPulser electroporator), the cells were suspended into 1 ml of DMEM/F-12 medium and transferred onto two wells of a 24-well plate containing coverslips. ZF4 cells at 30 h post-transfection were fixed with 2% paraformaldehyde for immunofluorescence imaging of DrASC nucleation and DrNLRP3 inflammasome assembly as described above.

Fluorescence recovery after photobleaching (FRAP) assay

FRAP assay was used to evaluate the dynamic recruitment of DrCaspase-A/-B into the DrNLRP3 inflammasome. Enhanced GFP–DrASC/DrNLRP3 and RFP–DrCaspase-A/-B fusion proteins were coexpressed in HEK293T cells, and FRAP was performed on an LSM710 Airyscan microscope with a 488-nm or 543-nm laser. The DrCaspase-A/-B in the inflammasome speck structure was fully or partially photobleached with 100% laser power. Time-lapse images were acquired over 45 cycles after bleaching at 3-s intervals (57–59). The fluorescence intensities of regions of interest (ROIs) were corrected by unbleached control regions and then normalized to prebleached intensities of the ROIs. Data from image analysis were graphed using GraphPad Prism 7.

Morphological examination for pyroptotic cell death

To examine the morphological changes during pyroptotic cell death, the images of transfected or stimulated cells were captured. 4.5 μm PI (Invitrogen) was added to the medium as an indicator of cell membrane integrity. The dyed cells exhibit the loss of plasma membrane integrity and exhibit pyroptotic-like features when captured under a two-photon laser-scanning confocal microscope (Zeiss LSM710, Germany) at room temperature (19).

Cytotoxicity assay

The cytotoxicity triggered by DrNLRP3 inflammasome was monitored by the release of LDH by using a CytoTox 96® nonradioactive cytotoxicity assay kit (Promega). After transfection by the plasmids encoding DrNLRP3, DrASC, DrCaspase-A, DrCaspase-B/CaspaseB-4DA, and DrGSDMEa/b for 48 h, the supernatant (50 μl) of the HEK293T cells was collected from each sample and mixed with the CytoTox 96® reagent (50 μl) for enzymatic assay. After a 30-min reaction in an opaque box, the stop solution (50 μl) was added to each sample. Absorbance (OD₄₉₀) was also detected by a Synergy H1 Hybrid Multi-Mode Microplate Reader (BioTek). The cytotoxicity (%) of the effector cells was calculated following the protocol previously described (60, 61).

Intracellular bacterial challenge assay

A bacterial challenge assay was performed in zebrafish embryos by immersion infection with E. tarda TL5m strain, an intracellular virulent pathogen for various aquatic animals, as described previously, to examine in vivo the DrCaspase-A/-B activation, DrIL-1β maturation, and pyroptosis. For this process, the embryos (6–72 hpf) were exposed to 1 × 10⁸ CFU/ml E. tarda in a 10-cm dish for 40 min to 4 h. Then, the embryos were collected to detect the activation of DrCaspases by Ac-YVAD/WEHD-AFC and the maturation of DrIL-1β by the rabbit anti-DrIL-1β polyclonal Ab prepared in our previous study (18, 62). DNA from E. tarda was extracted with a bacterial DNA preparation kit (Omega Bio-tek). DNA (200 pg per embryo), LPS (Escherichia coli O55:B5, Sigma, 2 ng per embryo), MDP (InvivoGen, 2 ng per embryo), and ATP (BBI Life Sciences, 20 ng per embryo) were microinjected into 6 hpf embryos. Meanwhile, 10 mm H₂O₂ was used for immersed stimulus. After 40 min, the embryos were collected, and the DrCaspase activation and the DrIL-1β maturation were detected.

Visualization of DrNLRP3 inflammasome in vivo

DrNLRP3 inflammasome was examined by its occurrence in zebrafish embryos under bacterial infection. For this process, pCMV-Tag2B-DrNLRP3 and pCMV-Myc-DrASC were co-microinjected into the one-cell–stage embryos at a concentration of 100 pg per embryo. At 72 hpf, the embryos were challenged with E. tarda (1 × 10⁸ CFU/ml) for 4 h, collected, and sliced into 6-μm–thick frozen sections by a freezing microtome (CM1950, Leica). For DrNLRP3 inflammasome visualization, immunofluorescent staining was performed on the sections as described above. Moreover, DrNLRP3 cooperation with DrNLRP1 in the inflammasome signaling was characterized in embryos by co-microinjecting with pEGFP-N1-DrNLRP3 and pDsRED-C1-DrNLRP1 expression vectors. The DrNLRP3 (green) and DrNLRP1 (red) colocalization in 72 hpf embryos was visualized via an inverted microscope (Zeiss Axiovert 40 CFL; Carl Zeiss, Jena, Germany).

MOs and capped mRNAs

The MOs against the mRNAs of DrNLRP3 (NLRP3-MO, 5′-CATCAACCTGTTCATGGCCTCCATTTTC-3′), DrCaspase-A (CaspaseA-MO, 5′-CCATGTTTAGCTCAGGGCGCTG-3′), and DrCaspase-B (CaspaseB-MO, 5′-AGCTGGGTAATATCCTCCATTTTCT-3′) and the standard control MO (Ctrl-MO, 5′-CTCTTACCTCAGTTACAATTTATA-3′) were designed and synthesized by Gene Tools (Philomath, OR). The MOs were dissolved with nuclease-free H₂O to 1 mm as stock solutions (63). For MO-resistant mRNA synthesis, DrNLRP3, DrCaspase-A, and DrCaspaseB cDNA sequences were constructed into the pcDNA3.1 vector by using the primers as shown in Table S1. The capped mRNAs were synthesized using the mMESSAGE kit (Ambion), purified with Mini Quick Spin RNA columns (Roche Applied Science), and solubilized in diethyl pyrocarbonate water. The one-cell–stage embryo was microinjected with DrNLRP3-MO (1.5–4.5 ng) or CaspaseA/B-MO (4.0 ng) for gene knockdown and MO-resistant mRNAs (200 pg) for the sake of rescue.

Evaluation of DrNLRP3 inflammasome in antibacterial immunity

The functional role of DrNLRP3 inflammasome in innate immunity was evaluated through its antibacterial activity in the zebrafish embryo model. For this evaluation, the DrNLRP3 knockdown, rescue, and control embryos were challenged with E. tarda (1 × 10⁸ CFU/ml) at 6 or 72 hpf. Mortality in each group was monitored during the 12-h period at one interval. The relative survival rate (RSR) was calculated using the following formula: RSR (%) = (survival rate of the infected group/survival rate of the mock PBS-administered control group) × 100%. Infection experiments were performed in triplicate with 100 embryos per group.

Generation of DrASC CRISPR knockout line

The gRNA for CRISPR-Cas9–based gene knockout was designed by a website program (CHOPCHOP: http://chopchop.cbu.uib.no). The target sequence of the first DrASC exon is GTGTTCACATCAAAAGACGCGG. The gRNA was prepared by in vitro transcription by using the MEGAscript^TM T7 high-yield transcription kit (Invitrogen) and purified using the MEGAclear^TM kit (Invitrogen). The mixture of 500 ng/μl TrueCut^TM Cas9 protein v2 (Invitrogen) and 300 ng/μl DrASC-KO gRNA was microinjected into one-cell–staged embryos. TheDrASC^−/− zebrafish homozygote was acquired after two generations (64–66).

Flow cytometry analysis for cell pyroptosis

The percentage of pyroptotic cells in zebrafish embryos was evaluated by a flow cytometer (FACSCalibur or FACSJazz, BD Biosciences). For this process, the 72 hpf WT embryos, DrNLRP3, the DrCaspase-A/-B knockdown embryos, and the DrASC knockout embryos were challenged with E. tarda (1 × 10⁸ CFU/ml) by immersion infection for 4 h. The embryonic cells were gently separated from the infected embryos (n = 100) digested with type IV collagenase (1.0 unit/ml, Sigma, at room temperature for 30 min) by filtration through a 40-μm strainer (Falcon, BD Biosciences) and centrifugation at 100 × g at 4 °C for 5 min, and then washed three times with PBS by centrifugation, and labeled by PI according to the manufacturer's protocol (Thermo Fisher Scientific). In FCM analysis, at least 10,000 cells were acquired from the gate for examination. FlowJo 7.6 software (BD Biosciences) was used for data processing (67, 68).

Statistical analysis

The data in the study were presented as the mean ± S.D. of each group. Statistical significance between experimental and control groups was assessed by two-tailed Student's t test and was considered at *, p < 0.05; **, p < 0.01 or ***, p < 0.001. The sample number for each group of zebrafish exceeded 20–100 embryos or larvae. More than 100 cells with immunofluorescence speck were counted for quantifying DrASC nucleation. All experiments were replicated at least three times.

Author contributions

J.-Y. L., Y.-Y. W., D.-D. F., and L.-X. X. data curation; J.-Y. L. and T. S. software; J.-Y. L., T. S., L.-X. X., and J.-Z. S. formal analysis; J.-Y. L., T. S., A.-F. L., L.-X. X., and J.-Z. S. validation; J.-Y. L. and Y.-Y. W. visualization; J.-Y. L. and L.-X. X. methodology; J.-Y. L. and Y.-Y. W. writing-original draft; D.-D. F., L.-X. X., and J.-Z. S. resources; A.-F. L., L.-X. X., and J.-Z. S. writing-review and editing; L.-X. X. and J.-Z. S. supervision; L.-X. X. and J.-Z. S. funding acquisition; L.-X. X. and J.-Z. S. project administration; J.-Z. S. conceptualization; J.-Z. S. investigation.