Tripartite-motif family protein 35-28 regulates microglia development by preventing necrotic death of microglial precursors in zebrafish

Microglia are tissue-resident macrophages in the central nervous system (CNS) that play essential roles in the regulation of CNS development and homeostasis. Yet, the genetic networks governing microglia development remain incompletely defined. Here, we report the identification and characterization of a microglia-defective zebrafish mutant wulonghkz12 (wulhkz12) isolated from an ethylnitrosourea (ENU)-based genetic screen. We show that wulhkz12 mutants harbors a missense point mutation in the gene region encoding the PRY/SPRY domain of the tripartite-motif family protein 35-28 (trim35-28) gene. Time-lapse imaging revealed that the loss of Trim35-28 function causes lytic necrosis of microglial precursors/peripheral macrophages, as indicated by cytoplasmic swelling and membrane rupture of these precursors and accompanied by neutrophil infiltration and systemic inflammation. Intriguingly, the lytic necrosis of microglial precursors in trim35-28–deficient mutants appeared to depend neither on the canonical pyroptotic nor necroptotic pathways, as inhibition of the key component in each pathway could not rescue the microglia phenotype in trim35-28–deficient mutants. Finally, results from tissue-specific rescue experiments suggested that Trim35-28 acts cell-autonomously in the survival of microglial precursors. Taken together, the findings of our study reveal Trim35-28 as a regulatory protein essential for microglia development.

Microglia are tissue-resident macrophages in the central nervous system (CNS) that play essential roles in the regulation of CNS development and homeostasis. Yet, the genetic networks governing microglia development remain incompletely defined. Here, we report the identification and characterization of a microglia-defective zebrafish mutant wulong hkz12 (wul hkz12 ) isolated from an ethylnitrosourea (ENU)-based genetic screen. We show that wul hkz12 mutants harbors a missense point mutation in the gene region encoding the PRY/SPRY domain of the tripartite-motif family protein 35-28 (trim35-28) gene. Time-lapse imaging revealed that the loss of Trim35-28 function causes lytic necrosis of microglial precursors/peripheral macrophages, as indicated by cytoplasmic swelling and membrane rupture of these precursors and accompanied by neutrophil infiltration and systemic inflammation. Intriguingly, the lytic necrosis of microglial precursors in trim35-28-deficient mutants appeared to depend neither on the canonical pyroptotic nor necroptotic pathways, as inhibition of the key component in each pathway could not rescue the microglia phenotype in trim35-28-deficient mutants. Finally, results from tissue-specific rescue experiments suggested that Trim35-28 acts cell-autonomously in the survival of microglial precursors. Taken together, the findings of our study reveal Trim35-28 as a regulatory protein essential for microglia development.
Microglia are the central nervous system (CNS)-resident macrophages, and they play essential roles in regulating the homeostasis of the CNS. As the key immune cells in the CNS, microglia, upon brain infection or brain injury, are quickly activated and migrate to the target site and remove infectious agents and cellular debris, leading to the elimination and resolving of detrimental neuronal inflammation (1)(2)(3). In addition to functioning as scavengers, recent studies have indicated that microglia also actively participate in the regulation of synaptic pruning and neuronal activity (4)(5)(6). It is therefore not surprising that the aberrant function of microglia has been implicated to be closely associated with the onset and progression of many neurodegenerative disorders, such as Alzheimer's disease, Parkinson's disease, and multiple sclerosis (7)(8)(9). Unlike the neuroectodermal origin of other cell types in the CNS, microglia are of hematopoietic origin (8)(9)(10). Fate mapping studies have revealed that microglia in mice are derived from hematopoietic precursors generated in a single origin, the yolk sac (YS) (11)(12)(13), whereas microglia in zebrafish arise from two sources, the rostral blood island and the aorta-gonad mesonephros (14). Despite the difference in terms of their origins, the developmental regulation of microglia appears to be evolutionarily conserved between fish and mice, as microglia in both organisms undergo similar developmental steps that are governed by a highly conserved repertoire of regulators, including Pu.1 (14)(15)(16), Irf8 (17,18), and Csf1r (19,20). Yet, the molecular networks controlling the formation of microglia remain incompletely defined. Although recent advance of the CRISPR-Cas9 system has made reverse genetic study much feasible in zebrafish model (21), an unbiased forward genetic study with this tiny organism will still provide a useful tool to identify new players involving in various biological processes.
For a long time, apoptosis has been thought to be the only form of programmed cell death governed by well-defined intrinsic molecular pathways programs (22). However, emerging evidences have suggested that necrosis, although initially described as accidental cell death, can be programed by defined signaling pathways, including the receptor-interacting serine/ threonine-protein kinases (RIPKs)-dependent necroptosis and inflammasome-mediated pyroptosis (23-25). Apoptosis and necrosis can be easily distinguished based on their morphological changes of the dying cells. Although the morphological changes of apoptosis involve chromosome fragmentation, membrane blebbing, and shrinkage of cell volume, necrosis, either accidental or programed, display a common morphological feature: cytoplasmic swelling due to early loss of membrane integrity (23-25). Another conspicuous difference between apoptosis and necrosis is their distinct immunological consequences. Through packaging the cellular contents within This article contains supporting information. * For correspondence: Wenqing Zhang, mczhangwq@scut.edu.cn; Zilong Wen, zilong@ust.hk.
membrane-bound cell apoptotic bodies, apoptosis does not trigger inflammation. On the other hand, necrosis is a lytic cell death that causes the release of cellular contents and the exposure of pathogen-associated molecular patterns and dangerassociated molecular patterns, resulting in the induction of inflammation (23, 25,26). Despite numerus elegant studies showing that apoptosis and anti-apoptosis mechanisms play an essential role in cell lineage development and organ formation (27,28), the role of necrosis and anti-necrosis mechanisms in cell and organ development remains less explored. The TRIM proteins comprise a large family of proteins defined by the presence of the RBCC motif, which has a tandem organization of a RING domain, one or two B-box domains and a coiled-coil domain (29)(30)(31). The RING domain of the TRIM family proteins possesses E3 ubiquitin ligase activity, which has been shown to be crucial for mediating their effects on innate immunity. The B-box domains are zinc-binding motifs facilitating the recognition of the viral and intracellular proteins during innate response to viral infection. And the coiled-coil domain mediates homomeric and heteromeric interactions among TRIM family members and with other proteins to form high molecular complexes essential for the generation of innate immunity. TRIM35, a member in the class IV of TRIM family protein (32), is structurally characterized by the association of RBCC motif with the C-terminal PRYSPRY domain (also known as B30.2 domain). Although TRIM35 was initially cloned as a tumor suppressor gene (33) and a downstream factor of M-CSF signaling involving in apoptosis (34), a recent study has revealed that TRIM35 appears to inhibit the TLR7/9mediated type I interferon production via ubiquitination and degradation of IRF7 (35). In zebrafish, 37 members of Trim35 paralogues have been identified (36), and little is known about their functions in vivo. In this study, we described the isolation and characterization a microglia-defective zebrafish mutant wul hkz12 and uncovered the role of Trim35-28, a member of Trim35 orthologues, as an anti-necrosis factor during microglia development.

wul hkz12 mutants are defective in microglia and microglial precursors
To uncover novel factors involving in microglia development, we carried out an ENU-based forward genetic screen searching for microglia-defective zebrafish mutants (19). A mutant designated as wul hkz12 (wulong is the name of a Chinese tea) was identified because of the lack of neutral red (NR) staining in the developing brain at 3 days post-fertilization (dpf) and 6 dpf (Fig. 1A). Because of the nature of NR staining, the absence of NR signals in the brain of wul hkz12 mutants could be caused by the impairment of the lysosomal function of microglia or the loss of microglia population. To distinguish these two possibilities, we performed whole-mount in situ hybridization (WISH) with the microglial marker apolipoprotein Eb (apoeb) and the pan-leukocyte marker lymphocyte cytosolic protein 1 (Lcp1). Results showed that, whereas apoeb 1 and Lcp1 1 cells were abundantly present in the brains of control siblings, these cells were nearly absent in wul hkz12 mutants (Fig. 1B).
These data suggest that the lack of NR staining in wul hkz12 mutants is likely attributed to the loss of microglia.
Interestingly, we noticed that, although microglia were largely absent in the brains of wul hkz12 mutants, a significant number of Lcp1 1 leukocytes accumulated in the periphery (data not shown). To define the nature of the peripheral Lcp1 1 cells, we monitored the formation of peripheral macrophages and neutrophils at 4 dpf, both of which are known to be Lcp1 1 (37), with lineage-specific markers. Results showed that the number of peripheral macrophages was significantly reduced in wul hkz12 mutants (data not shown). In contrast, Sudan Black B (SB) staining, a dye that preferentially stains neutrophils (38), revealed a robust increase of neutrophils in the mutants (Fig.  1C). Notably, the increased neutrophils in wul hkz12 mutants were found to be scattered in the tissues, such as yolk sac and trunk regions, where normally devoid of neutrophils in WT siblings (Fig. 1C). Moreover, neutrophils often formed clusters in the caudal hematopoietic tissue (CHT) in wul hkz12 mutants (Fig. 1C). These data indicate that the wul hkz12 mutants are lack of microglia and peripheral macrophages but have an expanded neutrophil phenotype in the periphery.

trim35-28 is mutated in wul hkz12 mutants
To identify the gene mutated in wul hkz12 mutants, we carried out positional cloning analysis. The wul mutation was mapped to a 200-kb region on chromosome 16 flanked by two simple sequence length polymorphism (SSLP) markers, SSLP65 and SSLP82 ( Fig. 2A). DNA sequencing of the coding regions of the 10 candidate genes within this 200-kb region revealed a T to A missense mutation in the sixth exon of the trim35-28 gene, leading to a single amino acid substitution from isoleucine to asparagine at position 454 in the PRYSPRY (B30.2) domain ( Fig. 2B).
To confirm that the I454N mutation in the trim35-28 locus is responsible for the mutant phenotype, we designed a trim35-28 morpholino (MO) to test whether knocking down Trim35-28 expression in WT zebrafish would produce a phenotype similar to that of wul hkz12 mutants. As expected, results showed that trim35-28 MO knockdown caused a severe reduction of microglia in zebrafish (,5 microglia, n = 63/88; 5;15 microglia, n = 20/88 in morphants; .20 microglia in control embryos, n = 69/69) (Fig. 2C). We then performed the rescue experiment with in vitro synthesized WT or I454N trim35-28 mRNA. Consistent with the MO knockdown assay, we showed that the injection of WT but not I454N mutant trim35-28 mRNA could partially rescue the microglia number in wul hkz12 mutants (Fig. 2, D and E). Collectively, these data demonstrate that trim35-28 is indeed the causative gene responsible for the loss of microglia in wul hkz12 mutants.
We were next keen to investigate how this I454N mutation disrupts the function of Trim35-28 protein. We first aligned the protein sequences of the zebrafish Trim35-28 PRY/SPRY domain with four known PRYSPRY domains sharing the highest degree of similarity with Trim25-28 from the Protein Data Bank (PDB). Results indicated that the PRY/SPRY domains had a highly conserved secondary structure (Fig. S1A), indicating that the PRY/SPRY domain of Trim35-28 is likely to adopt a similar folding pattern with the canonical PRY/SPRY domains. We then took the human PYRIN PRY/SPRY domain (PDB ID 2wl1) (39) as a core structure to model the structure of the Trim35-28 PRY/SPRY domain (Fig. S1C). We found that Ile-454 in Trim35-28 appeared to correspond to Ile-755 in human PYRIN (Fig. S1B, indicated by red arrow), which locates in a scaffolding b-strand (b-strand 12) (39) and is buried in a hydrophobic core (Fig. S1C, blue dashed circle). The two loops (loop 2 and loop 5 in PYRIN) covering this hydrophobic core are the key components of the pocket essential for target recognition (40). The importance of this binding pocket is further documented by the observation that several mutations in the Familial Mediterranean fever patients are found to locate on the wall or near the rim of the target-binding pocket (39, 40) (Fig. S1C, labeled with red letters). Given the fact that the I454N mutation had no obvious effect on the stability and subcellular location of the proteins (Fig. S1, D and E), we believe that the substitution of hydrophobic isoleucine to hydrophilic asparagine at residue 454 might cause a conformational change of these two loops, resulting in the disruption of this protein function.

Trim35-28 is cell-autonomously required for microglia development
Having shown that trim35-28 is the causative gene mutated in wul hkz12 mutants, we next asked whether the loss of microglia phenotype and the neutrophil expansion phenotype in the mutants was caused by a cell-autonomous or noncell-autonomous mechanism. To test that, we first examined the temporal-spatial expression expressions of trim35-28 during early zebrafish development. WISH analysis showed that trim35-28 transcripts were ubiquitously expressed, with a relative enrichment in the developing brain and pharyngeal arches (Fig. S2). The ubiquitous expression of trim35-28 raises the possibility that the microglia and neutrophil phenotype in wul hkz12 mutants may not be necessarily caused by a cell-autonomous mechanism. To clarify this issue, we generated two transgenic lines, Tg(mpeg1:trim35-28) and Tg(lyz:trim35-28), in which the expression of WT trim35-28 was under the control of the macrophage-specific gene 1 (mpeg1) promoter (41) and the neutrophil-specific lysozyme C (lyz) promoter (42), respectively (Fig.  3A). These transgenic lines were outcrossed with wul hkz12 mutant fish to test whether the restoration of WT trim35-28 expression in a lineage-specific manner could rescue the microglia and neutrophil defect in wul hkz12 mutants. Results showed that forced expression of trim35-28 in the mutants with the mpeg1 promoter could fully rescue the microglia phenotype (Fig. 3, B and C). Interestingly, the neutrophil phenotype was also rescued in this macrophage-specific trim35-28 expression Tg(mpeg1:trim35-28);wul hkz12 line (Fig. 3, D and E). In contrast, both phenotypes could not be restored in the neutrophil-specific trim35-28 expression Tg(lyz:trim35-28);wul hkz12 line (Fig.  3, B-E). Taken together, these results indicate that trim35-28 is cell-autonomously required for the development of microglia and their precursors and the expansion of neutrophils in wul hkz12 mutants is a secondary effect caused by the defect of microglia.
Accelerated necrotic cell death of microglial precursors in periphery accounts for the loss of microglia and expansion of neutrophils in wul hkz12 mutants To dissect the cellular mechanism underlying the loss of microglia in wul hkz12 mutants, we examined the formation of microglial precursors: peripheral macrophages at early developmental stages. WISH of microfibril-associated protein 4 (mfap4) expression revealed that peripheral macrophages were specified and distributed properly in wul hkz12 mutants at 36 hpf ( Fig. 4A, upper panel). However, by 2.5 dpf, the macrophage number in mutant embryos was significantly reduced (Fig. 4, A,  lower panel, and B), suggesting that the loss of microglia in wul hkz12 mutants is largely due to the impairment of microglial precursors in the periphery prior to their invasion and coloni-zation of the brain. To define which cellular mechanism, accelerated cell death or impaired proliferation, caused the reduction of microglial precursors in wul hkz12 mutants, we performed terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling (TUNEL) staining and bromodeoxyuridine (BrdU) Regulation of microglia development by a trim35 orthologue incorporation assay. As shown in Fig. 4, C and D, compared with the rare cell death in siblings, massive TUNEL signals were detected in wul hkz12 mutants and the majority of those signals co-localized with macrophage-specific marker mpeg1-GFP, suggesting that microglial precursors undergo accelerated cell death in the mutants. In contrast, BrdU incorporation assay showed a comparable proliferation rate of microglial precursors between the mutants and siblings (Fig. S3, A and B). Taken together, these results demonstrate that the loss of microglia in wul hkz12 mutants is largely ascribed to the excessive death of microglial precursors in the periphery.
To define the nature of the death of the mutant microglial precursors, we performed time-lapse imaging to monitor the morphological and behavioral changes of the microglial precursors in the mutants and siblings from 2 to 3 dpf, at which these precursors were shown to actively undergo cell death. We found that a substantial number of mpeg1 1 cells in wul hkz12 mutants underwent cell death as indicated by their fragmentation (Fig. 5, A and B). Notably, unlike the typical morphological changes (cytoplasmic shrinkage and membrane blebbing) in the apoptotic macrophages induced by metronidazole treat-ment of Tg(mpeg1:Gal4;uas:nfsB-mCherry) embryos ( Fig. S4A and Movie S1) (43)(44)(45), the dying mpeg1 1 cells in wul hkz12 mutants displayed distinctive morphology and behavior including retraction of processes and increase of cytoplasm volume (Fig. 5A, a'1-a'2; a''1-a''4, and Movie S2) followed by rapid cell membrane rupture and fragmentation into multiple small pieces (Fig. 5A, a'3-a'7 and a''5-a''7, and Movie S2). To prove that the death of microglial precursors in wul hkz12 mutants was indeed not apoptosis, we examined the generation of activated Caspase 3, a well-known apoptosis marker, by immunostaining with specific antibody. Results showed that, whereas the activated Caspase 3 was robustly detected in the apoptotic neurons in the brains of 2.5 dpf wul hkz12 mutants (Fig. S4B), it was completely absence in the mfap4-GFP 1 microglial precursors (Fig.  S4C). These data indicate that the death of mpeg1 1 cells in wul hkz12 mutants is not apoptosis but rather is necrosis. Indeed, the death of mpeg1 1 cells in wul hkz12 mutants was always coupled with the infiltration of neutrophils (Fig. 5B, b6-b7, dashed circles, and Movie S3) and the expression of pro-inflammatory cytokines, such as il-1b, tnfa, and il-6 ( Fig. 5, C and D), as well as the macrophage activation marker irg1 (Fig. S5)  microglial precursors. Altogether, these results indicate that the death of microglial precursors in wul hkz12 mutants resembles the characteristics of lytic necrosis, including cytoplasm swelling, membrane rupture, and induction of inflammation (23, 25).

Lytic necrosis of microglial precursors in wul hkz12 mutants is mediated by neither the canonical pyroptotic nor necroptotic signaling pathways
Previous studies have shown that the loss of function mutations in the nucleotide oligomerization domain-like receptor nlrc3-like cause the pyroptotic death of microglia and their precursors due to hyperactivation of the inflammasome pathway (46,47). As the phenotypes observed in wul hkz12 and nlrc3-like mutants are similar, we hypothesized that disruption of the Trim35-28 function might also lead to the aberrant activation of the inflammasome pathway, resulting in the pyroptotic death of microglia and peripheral macrophages. To test this hypothesis, we outcrossed wul hkz12 mutants with asc D31 fish, in which the key component of apoptosis-associated speck-like protein containing a caspase recruitment domain (Asc) is inactivated, thereby abolishing the activation of the canonical inflammasome pathway (47). Surprisingly, we found that the microglia number in wul hkz12 ;asc D31 double mutants was comparable with that in wul hkz12 single mutants (Fig. 6, A and B), suggesting that the death of microglial precursors in wul hkz12 mutants is independent of the canonical pyroptotic pathway. To further support this notion, we employed MOs to knock down the expression of gsdmea and gsdmeb, two zebrafish counterparts of mammalian gasdermin family genes essential for pyroptosis (47) and asked whether it would rescue the mutant phenotype. Consistent with the observation in wul hkz12 ;asc D31 double mutants, knocking down gsdmea or gsdmeb in wul hkz12 mutants failed to rescue the microglia phenotype (Fig. 6, C and D), indicating that the lytic necrosis of microglial precursors in wul hkz12 mutants is not a typical pyroptotic cell death. We therefore speculated that the lytic necrosis of microglial precursors in wul hkz12 mutants could be necroptosis. To test this hypothesis, we generated ripk3 D52 mutant fish, in which the function of Ripk3, a key component involving in necroptosis (48,49) is disrupted (Fig. S6). The mutant fish were outcrossed with wul hkz12 mutants to test whether the inactivation of Ripk3 could rescue the microglia phenotype in wul hkz12 mutants. Surprisingly, results showed that the loss of Ripk3 function could not rescue the microglia phenotype in wul hkz12 mutants (Fig. 6, E and F). Taken together, these results indicate that the lytic necrosis of microglial precursors in wul hkz12 mutants is likely not mediated by the canonical pyroptotic and necroptotic pathways.

Discussion
In this report, we demonstrate that zebrafish trim35-28, a member of trim family, is essential for the formation of have classified teleost trim genes into two distinct subclasses. The first subclass includes the trim family genes containing 1 or 2 fish paralogues corresponding to the mammalian counterparts, and this family of trim genes is thought to carry out conserved functions across species. On the other hand, members of the second subclass are highly expanded trim genes, including the trim35 family, in which the C-terminal PRY/SPRY domains, which define their substrate specificity, are likely to be evolved under a positive selection mechanism (36). Hence, we speculate that this unique evolutionary pathway may endow each zebrafish trim35 gene with a specific function, thereby facilitating the adaption of zebrafish to their unique adequate milieu. Indeed, protein sequence alignment indicates that the PRY/SPRY domains of Trim35-28 share only 30-40% similarities with other Trim35 proteins and its mammalian counterparts (data not shown), suggesting that the functions of Trim35-28 may not be substituted by the other members of the Trim35 family and its mammalian counterpart. This idea is supported by the observations that trim35-28-deficient mutants die at around 10 dpf and forced expression of mammalian TRIM35 fails to rescue the mutant phenotype (data not shown). Thus, it would be interesting to explore the roles of other trim35 members as well as its mammalian counterpart.
Another intriguing finding is that inactivation of the key players in pyroptotic (Asc and Gasdermin) and necroptosis (Ripk3) pathway fails to rescue the death of microglial precursors/macrophages in wul hkz12 mutants, despite microglial precursors/macrophages in trim35-28-deficient mutants undergo typical characteristics of necrotic cell death. These results could be explained by several mechanisms. One Figure 5. The death of microglial precursors/macrophages in wul hkz12 mutants resembles necrotic cell death. A, time-lapse confocal imaging of a wul hkz12 ;Tg(mpeg1:LRLG) embryos from 2.5 dpf to 3.5 dpf. White arrows indicate the swelling and breakdown of two microglial precursors. B, time-lapse confocal imaging of a wul hkz12 ;Tg(mpeg1:LRLG);Tg(lyz:eGFP) embryo from 2.5 dpf to 3.5 dpf. White arrows indicate a dying microglial precursor and the dash white circle indicates the infiltration of neutrophils. C, WISH of il-1b expression and co-staining with lyz-GFP and Lcp1 antibody in 2.5 dpf siblings and wul hkz12 mutants. D, quantitative RT-PCR of il-1b, tnfa, il-6, il-10, ifn1, and tgfb1 expression in 3 dpf siblings and wul hkz12 mutants. Expression level of target genes was normalized with elf1a expression. Error bars represent S.D.
possible mechanism is the complementary effect caused by activation of the genetic compensation response. Recent studies have shown that premature termination codon-bearing mRNA could trigger genetic compensation response via Upf3a and COMPASS components (51,52). It is therefore reasonable to speculate that the asc D31 and ripk3 D52 mutations, both of which generate premature termination codonbearing mRNA, may activate the genetic compensation response pathways, resulting in the complementary effect on the blockage of pyroptotic and necroptosis pathway. However, our preliminary study showed that the elimination of the genetic compensation response pathway by MO knockdown of upf3a expression (47,48) could not rescue microglia defects in wul hkz12 ;asc D31 or wul hkz12 ;ripk3 D52 double mutants (data not shown), suggesting that activation of the genetic compensation response is likely not to be the mechanism. An alternative mechanism is that the trim35-28-deficient microglial precursors/macrophages undergo pyroptosis or necroptosis through an Asc-independent or a Ripk3idenpendent mechanism. Finally, the death of microglial precursors/macrophages in wul hkz12 mutants may represent another type of cell death, which has not been well characterized. Further in-depth analysis will be required to define the nature of the death of trim35-28-deficient microglial precursors/macrophages and the underlying molecular basis.

Sequential neutral red and Sudan black B staining
NR staining (14) was conducted first and then the embryos were separated into two pools based on the microglia phenotype for subsequent SB staining (38).

Positional cloning
Positional cloning was performed as described previously (59). In brief, the wul hkz12 mutation was first mapped to linkage group 16 by two SSLP markers z10036 and z4670. Two closer Figure 6. Lytic necrosis of microglial precursors/macrophages in wul hkz12 mutants is independent of canonical pyroptotic or necroptotic signaling pathways. A, NR staining of 3 dpf siblings, asc D31 mutants, wul hkz12 mutants, and wul hkz12 ;asc D31 double mutants. B, quantification of NR positive microglia in A. n.s., not significant; p . 0.05. C, NR staining of 3 dpf siblings, wul hkz12 mutants, siblings injected gsdmea MO, wul hkz12 mutants injected gsdmea MO, siblings injected gsdmeb MO, wul hkz12 mutants injected gsdmeb MO, siblings injected with both gsdmea and gsdmeb MOs, and wul hkz12 mutants injected with both gsdmea and gsdmeb MOs. D, quantification of neutral red-positive microglia in C. p . 0.05. E, NR staining of 3 dpf siblings, ripk3 D52 mutants, wul hkz12 mutants, and wul hkz12 ;ripk3 D52 double mutants. F, quantification of NR positive microglia in E, p . 0.05. Error bars represent S.D. markers, z25218 and z17383 (one for each direction), were then chosen to screen embryos for fine mapping. By analyzing 2548 total meiosis events, the wul hkz12 mutation was narrowed down to a 200-kb region with two SSLP markers, BX571757.8-sslp65 (1 recombinant of 2548 total meiosis) and CR391910. 16-sslp82 (1 recombinant of 2528 total meiosis). Sequencing of the coding regions of candidate genes revealed a Thr to Ala missense mutation in the last exon of trim35-28. In vitro mRNA synthesis WT and mutant trim35-28 cDNA were cloned into PCS2(1) plasmid. The resulting constructs were linearized by NotI digestion and the trim35-28 mRNA was synthesized using mMESSAGE mMACHINE TM SP6 Transcription Kit (AM1340) according to the manufacturer's instructions.
In vitro synthesis of antisense RNA probes and WISH Antisense RNA probes and WISH were performed according to standard protocol (53).

Generation of transgenic lines
The 4-kb mpeg1 promoter (41) and 2.4-kb lyz promoter (42) was cloned into the pBluescript II SK(1) vector containing two arms of Tol2 sequences to drive trim35-28 expression in macrophages and neutrophils, respectively. The -4kbmpeg1:trim35-28 or -2.4kblyz: trim35-28 constructs, together with transpose mRNA, was injected into 1-cell stage fertilized embryos. The injected embryos were then raised to adult for germline transmission screening.
Acridine orange, BrdU incorporation, TUNEL staining and Caspase 3 immunostaining Acridine orange (AO) staining was performed according to manufacturer's instruction. BrdU incorporation assay and TUNEL staining were performed as previously reported (60). Caspase 3 immunostaining was performed as previously described (61).

Time-lapse imaging and images
Time-lapse imaging was performed under Leica SP8 confocal microscopy with 203 HC PL APO 203/0.70 DRY objective according to the protocol described previously (61). Images of fluorescent signals and immunostaining were captured by Leica SP8 confocal microscope and Zeiss LSM 710 confocal microscope.

Statistics
Statistics were analyzed using the two-tailed Student's t test. Results were considered significant if p , 0.05. Values represent mean 6 S.D.

Data availability
All data presented in this paper are contained within the article.
Acknowledgments-We thank Dr. Hae-Chul Park for providing the Tg(uas:nfsB-mCherry) line.  Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.