Introduction

Microglia, the crucial immune cells residing in the central nervous system (CNS), have a profound impact on the maintenance of CNS homeostasis, as they not only function as scavengers to facilitate the removal of invading pathogens or damaged tissues, but also actively regulate neurogenesis, synaptic pruning and neuronal activity (Colonna & Butovsky, 2017; Gomez-Nicola & Perry, 2015). It is therefore not surprising that the dysfunction of microglia is closely linked with the occurrence and progression of many neurodegenerative disorders (Glass et al, 2010; Li et al, 2012; Prinz et al, 2011), thus making it a promising therapeutic target in the prevention and treatment of these diseases.

In mice, fate-mapping methods revealed that unlike other glia cells with neuroectoderm origin, microglia in adult originate from yolk-sac (YS) derived primitive macrophages, which colonize the developing brain rudiment during early embryogenesis (Ginhoux et al, 2010; Schulz et al, 2012). Whether microglia in adult mammals were contributed by monocyte-derived microglia precursors had long been a debate until the establishment of parabiosis assay (Ajami et al, 2007). It is now well-accepted that under physiological condition microglia in mice are maintained locally by self-renewal throughout the lifespan. In accordance with this concept, microglia population was found to be rapidly reconstituted by resident cells after genetic or pharmacological ablation of microglia (Bruttger et al, 2015; Elmore et al, 2014). However, when estimating the physiological turnover rate of microglia via different methods such as ³H-thymidine/BrdU/EdU incorporation, Ki67 staining or two photon live imaging, different studies yielded distinct or even contradictory results i.e., the estimated microglia turnover rate ranging from 0.13 to 0.8% (Askew et al, 2017; Füger et al, 2017; Lawson et al, 1992; Tay et al, 2017), which indicated that microglia in adult mice (∼2 years lifespan) were either long-lived without replenishment or replenished for several times during the whole lifespan of the animals. Nonetheless, these studies have reached the consensus that in adult mice all microglia have the proliferation capability and there are no putative microglia precursors existing to proliferate asymmetrically and continuously to contribute to microglia pool. Moreover, microglia proliferation had been shown to be spatially and temporally coupled with apoptosis to maintain the homeostasis of their population (Askew et al., 2017): which indicates that not only proliferating microglia locate near the dying microglia, but also newborn microglia manifest a higher frequency of apoptosis. However, whether this phenomenon is for the quality control of microglia or has any other physiological relevance remained unclear. In addition, as a fraction of newborn microglia underwent apoptosis and did not contribute to microglia homeostasis, previously calculated proliferation rate might potentially over-estimate the true turnover rate of microglia. Collectively, although these studies give some insights into the understanding of microglia turnover and maintenance, our knowledge about the underlying cellular and molecular mechanisms are still limited.

PU.1/SPI1, the founding member of the spleen focus forming virus proviral integration oncogene (SPI) subfamily of the E-twenty six (ETS) family transcription factor, is one of the best-characterized gene governing the development of macrophage lineage (McKercher et al, 1996; Scott et al, 1994). Disruption of PU.1 in mice causes early lethality of the embryos and complete absence of macrophages (McKercher et al., 1996; Scott et al., 1994), including microglia in brain (Kierdorf et al, 2013). Interestingly, the PU.1 expression level was found to remain at a high level in microglia after their terminal differentiation (Walton et al, 2000), implying that PU.1 also regulates the cellular behaviors and functions of microglia in the sophisticated CNS environment. Indeed, Chromatin immunoprecipitation (ChIP)-Seq analysis with microglia cell line BV2 cells revealed thousands of PU.1 targets (Satoh et al, 2014). Accordingly, silencing of PU.1 by siRNA in mixed human glial culture led to the reduced viability of microglia and decreased phagocytosis of amyloid-beta (Aβ42) (Smith et al, 2013). However, as a previous study reported that inactivating either 1 copy or 2 copies of PU.1 by fusing PU.1 with cytoplasm-retaining estrogen receptor (ER) had no effect on the viability of alveolar macrophages (Karpurapu et al, 2011), whether PU.1 indeed regulated the survival and maintenance of microglia in the in-vivo conditions remained to be further investigated. Moreover, human SNPs that altered PU.1 expression level were reported to be associated with Alzheimer’s disease (AD) pathogenesis (Cao et al, 2022; Huang et al, 2017). A recent study showed that conditional inactivation or pharmacological inhibition of PU.1 in AD mouse model ameliorated neuroinflammation, prevented microgliosis and improved cognitive performance (Ralvenius et al, 2023). Although reduction of microglia number was considered as a result of the amelioration of neuroinflammation, whether PU.1 directly regulated microglia survival and maintenance to prevent the propagation of neuroinflammation in diseased condition deserved to be carefully revisited. SPI-B, another member in the mammalian SPI subfamily, has been shown to be dispensable for the development of the macrophage lineage (Su et al, 1997). Yet, its involvement in the regulation of microglia turnover and maintenance remains to be investigated.

Owing to the genetic amenability and the feasibility for live-imaging and lineage tracing, zebrafish has been adopted as an ideal model for the study of microglia development and function (Hughes & Appel, 2020; Iyer et al, 2022; Li et al., 2012; Peri & Nüsslein-Volhard, 2008). Zebrafish microglia are highly conserved with their mammalian counterparts in terms of morphology, molecular signatures, development-regulatory genetic networks as well as functions (Xu et al, 2015; Yu et al, 2017). Different from the single origin of mouse microglia, zebrafish microglia were shown to arise from two distinct sources i.e., the rostral blood (RBI), equivalent of mouse YS for primitive myelopoiesis, and the aorta-gonad-mesonephros (AGM) region for hematopoietic stem cell (HSC) initiation. The RBI microglia colonize the developing brain during early embryogenesis, but are gradually replenished by AGM adult microglia (Xu et al., 2015). Further study revealed that this dynamic shift of microglia pool is controlled by Il34-Csf1ra signaling-dictated cell competition (Yu et al, 2023). Yet, when the microglia pool is established in adult zebrafish, the molecular mechanisms underlying their turnover and maintenance remain incompletely understood. Moreover, although Pu.1 (also called Spi1b) and its paralogue Spi-b (also called Spi1a), the zebrafish orthologues of mammalian PU.1 and SPI-B, were shown to be indispensable for microglia early development of both origins in zebrafish (Xu et al., 2015; Yu et al., 2017), their functions in the survival and maintenance of microglia in adult animals have not been established.

In the present study, via the condition knockout strategy to generate mosaic microglia population in adult zebrafish and mice, we revealed that the turnover and maintenance of microglia is evolutionarily conserved and regulated by Pu.1/Spi1 dosage from teleost to mammals.

Results

Adult microglia in zebrafish undergo rapid turnover and random proliferation to replenish and maintain their pool

Previous studies have shown that there are two major subtypes of myeloid cells, i.e., ccl34b.1⁺ microglia and ccl34b.1^-ccl19a.1⁺ brain-resident dendritic cells (DCs) in adult zebrafish brain (Wu et al, 2020; Zhou et al, 2023). To monitor the turnover and maintenance of microglia in adult zebrafish brain, we employed microglia-specific TgBAC(ccl34b.1:eGFP) transgenic reporter line (Wu et al., 2020) and performed EdU pulse-chase experiment, a similar strategy with that in mouse to determine microglia turnover (Askew et al., 2017; Tay et al., 2017). As a single dose of EdU injection might under-estimate the microglia turnover rate, we intraperitoneally injected EdU into adult TgBAC(ccl34b.1:eGFP) fish for either 1, 3 or 5 consecutive days. We then collected the fish brain at 1 day post injection (dpi) for cryosection and co-staining of eGFP, Lcp1 (a pan leukocyte marker expressed in both microglia and DCs), EdU and DAPI, in which ccl34b.1-eGFP⁺Lcp1⁺ cells are microglia and ccl34b.1-eGFP^-Lcp1⁺ cells represent DCs (Fig. 1A). Quantification results showed that, while a single dose of EdU pulse labeled 1.892±0.234% microglia, the EdU incorporate rate for microglia after 3 or 5 doses of EdU pulses increased almost linearly (Fig. 1B, green bars). Based on these data, the estimated microglia turnover rate (daily) is ∼2.6% (total EdU incorporation rate divided by the times of EdU-pulses) (Fig. 1C, green bars). On the other hand, the EdU incorporate rate for the DCs was much lower (Fig. 1B-C, red bars), suggesting that the DCs in the brain are relatively steady with a very low turnover rate.

Fig. 1.

Previously, via the EdU/BrdU dual-pulse assay, Tay et al. have shown that microglia in adult mouse brain proliferate randomly during homeostasis and no putative microglia precursors have been identified accordingly (Tay et al., 2017). To clarify whether there might be putative microglial precursors that continuously proliferate to contribute to microglia turnover in zebrafish, we employed a similar strategy, in which adult wild-type (WT) fish were injected with a single dose of EdU followed with five daily BrdU pulses to determine the BrdU proliferation rate of EdU⁺ and EdU^- microglia (Fig. 1D, upper panel). Quantification result showed that the BrdU incorporation rates were comparable between EdU⁺ and EdU^- microglia pools (Fig. 1D-E). These data strongly suggest that, similar to mammals (Tay et al., 2017), microglia in adult zebrafish randomly undergo proliferation and there are no well-defined microglial precursors that continuously proliferate to replenish and maintain the microglia pool.

Generation and characterization of the visible conditional knockout allele pu.1^KI

To elucidate the molecular mechanisms underlying microglia turnover and maintenance, we focused on the Ets family transcription factor PU.1/SPI1, which had been shown to be a master regulator for macrophage/microglia development in fish and mammals (McKercher et al., 1996; Scott et al., 1994; Yu et al., 2017). Because macrophages/microglia development is completely blocked at early development in pu.1-null zebrafish mutants (Yu et al., 2017), we decided to generate a visible pu.1 conditional knockout allele pu.1^KIto study the role of Pu.1 in microglia maintenance via the Non-homologous end jointing (NHEJ)-mediated knock-in method (Fig. 2A and Fig. S1A) (Li et al, 2015). As shown in Fig. 2A, the knock-in DNA fragment consisted of loxP, pu.1 exon 4 to 6 followed by the self-cleaving P2A sequence-conjugated eGFP, and a splicing-acceptor site SpA2 followed by the P2A-DsRed. In addition, the 3’ downstream sequence of pu.1 gene (gray box) was added to the 3’ end of each fluorescent protein to ensure efficient termination of RNA transcription and proper expression of the fluorescent proteins.

Fig. 2.

Three different founders were obtained. PCR and deep sequencing analysis of F1 embryos confirmed the correct integration of donor DNA fragment in the pu.1 locus (Fig. S1B-C). Since the knock-in pu.1^KI allele would generate intact Pu.1 and eGFP concurrently (Fig. 2A), we predicted the expression pattern of eGFP, which presumably represents the expression of the knock-in Pu.1, should completely overlap with the endogenous Pu.1. Indeed, co-staining of eGFP and Pu.1 antibodies revealed a perfect overlap of the eGFP⁺ cells and Pu.1⁺ cells in pu.1^KI/+ embryos (Fig. 2B), suggesting that the pu.1^KI allele recapitulates the expression of endogenous pu.1. Further studies suggested that pu.1^KI-eGFP efficiently labels embryonic microglia, and both microglia and DCs in adult, as revealed by the co-localization of pu.1^KI-eGFP and macrophage-specific reporter line Tg(mpeg1:LRLG) at 3 dpf, and by the co-staining of eGFP and Lcp1 antibodies on the brain sections of adult pu.1^KI/+ fish respectively (Fig. 2C and D). To further confirm functional intactness of the pu.1^KI allele, we crossed the pu.1^KI fish with the pu.1^Δ839 mutants to examine if the pu.1^KIallele is capable of rescuing the microglia phenotype. Indeed, quantification of microglia number revealed no difference between the pu.1^Δ839/KI and pu.1^Δ839/+ embryos (Fig. 2E and F), indicating that the pu.1^KI allele functionally recapitulates the endogenous pu.1.

To test whether the Cre mediated excision occurs correctly in the pu.1^KI allele, we injected Cre mRNA into the pu.1^KI embryos at one-cell stage. The injected embryos were raised to adulthood (pu.1^CKO fish) and crossed with the pu.1^KI fish to generate pu.1^CKO/KI fish. As illustrated in Fig. 2A, the Cre-mediated excision should remove the E4-6-P2A-eGFP cassette and allow the proper splicing of the splicing-acceptor site SpA2 and joining of the P2A-DsRed with exon 3 (Fig. 2A), leading to the induction of DsRed expression. Indeed, fluorescent imaging of the pu.1^CKO/KI embryos revealed robust DsRed (pu.1^CKO allele) signals perfectly overlapping with the eGFP (pu.1^KI allele) signals (Fig 2G), suggesting the correct DNA excision and RNA splicing in the pu.1^CKO allele. This conclusion was further validated by sequencing analysis of the pu.1^CKO transcripts (Fig. 2H). Since the endogenous pu.1 exons were not replaced, but instead they were separated far away from the upstream exons by donor plasmid, we performed quantitative RT-PCR analysis of pu.1^CKO/CKO embryos to determine the possible expression of the endogenous pu.1 exons in pu.1^CKOallele. The result showed that the relative expression of endogenous pu.1 exons is around 1∼2% in comparison with that of DsRed expression (Fig. S1D), suggesting that expression of pu.1 is largely abolished in pu.1^CKO allele. To phenotypically characterize pu.1^CKO allele, and to avoid incomplete inactivation of Pu.1 function in downstream microglia tracing experiments, which may occur when using two pu.1^KI alleles due to the lower CreER recombination efficiency in zebrafish compared to mice, we outcrossed the pu.1^CKO fish with the pu.1^Δ839 null mutants. As anticipated, the development of embryonic microglia was completely blocked in the pu.1^CKO/Δ839 embryos (Fig. 2I and J), a phenotype recapitulating the defect of the pu.1^Δ839 homozygous mutants (Yu et al., 2017). These data demonstrate that the pu.1^KI allele works very well and serves as a useful tool to explore the role of Pu.1 in adult microglia survival and maintenance.

Pu.1 and Spi-b regulate the turnover and maintenance of microglia in a dosage-dependent manner

To investigate the role of Pu.1 in the survival and maintenance of microglia in adult zebrafish, we crossed the pu.1^KI fish with the leukocyte-specific Tg(coro1a:CreER) line and the pu.1^Δ839 mutants to generate pu.1^KI/+;Tg(coro1a:CreER) and pu.1^KI/Δ839;Tg(coro1a:CreER) fish. We then intraperitoneally injected 3 doses of 4-OHT into these adult fish and collected the brain samples at 2 dpi and 8 dpi for the quantification of microglia number after cryo-section and eGFP/DsRed immuno-staining (Fig. 3A, upper panel). As shown in Fig. 3A (lower panel), DsRed⁺ cells were readily detected in the brains of both pu.1^KI/+;Tg(coro1a:CreER) and pu.1^KI/Δ839;Tg(coro1a:CreER) fish at 8 dpi. Quantification of the number and proportion of DsRed⁺ cells at 2 dpi and 8 dpi revealed no difference between pu.1^KI/Δ839;Tg(coro1a:CreER) mutant fish and pu.1^KI/+;Tg(coro1a:CreER) control siblings (Fig. 3B-C). In accordance with this observation, we also found that the number and proportion of microglia (ccl34b.1-eGFP⁺ Lcp1⁺) and DCs (ccl34b.1-eGFP^-Lcp1⁺) were comparable between pu.1^Δ839/Δ839;TgBAC(ccl34b.1:eGFP) null mutants and pu.1^Δ839/+;TgBAC(ccl34b.1:eGFP) siblings (Fig. S2). These data suggest that Pu.1 is not required for the survival of microglia and DCs. We reasoned that the survival of pu.1-deficient microglia is likely attributed to the presence of spi-b (the paralogue of pu.1 in zebrafish, also called spi1a), which has been shown to play a compensatory role for Pu.1 in microglia early development (Yu et al., 2017). Indeed, conditional inactivation of Pu.1 in the spi-b^Δ232/Δ232 null mutants led to a rapid depletion of microglia by apoptosis within several days (Fig. S3 and Fig. S4), indicating that Pu.1 and Spi-b act redundantly to regulate the survival of microglia.

Fig. 3.

To further investigate whether Pu.1-deficiency might have a long-term effect on the turnover and maintenance of microglia, we traced pu.1-deficient DsRed⁺ cells to 1 month post injection (mpi), and 3 mpi after conditionally inactivating Pu.1 in adult pu.1^KI/+;Tg(coro1a:CreER) and pu.1^KI/Δ839;Tg(coro1a:CreER) fish (Fig. 3D, upper panel). Intriguingly, we found that the number and proportion of DsRed⁺ cells in the midbrains of the pu.1^KI/Δ839;Tg(coro1a:CreER) fish began to decline at 1 mpi and were further reduced, becoming significantly lower than that in pu.1^KI/+;Tg(coro1a:CreER) fish by 3 mpi (Fig.3E-F). Likewise, the number and proportion of DsRed⁺ cells in the retina and spinal cord were also significantly reduced in the pu.1^KI/Δ839;Tg(coro1a:CreER) fish by 3 mpi (Fig. S5). Since microglia are the major population of the coro1a⁺ cells in the brain and spinal cord and the exclusive population in the retina (Wu et al., 2020), the reduction of DsRed⁺ cells in the pu.1^KI/Δ839;Tg(coro1a:CreER) fish at 1 mpi and 3 mpi are mainly attributed to the loss of microglia. Based on the amoeboid and ramified morphology, microglia and DCs could be distinguished with about 90% accuracy (Fig. S6A-B). Thus, to further confirm above conclusion, we quantified microglia and DCs separately in 3-mpi pu.1^KI/+;Tg(coro1a:CreER) and pu.1^KI/Δ839;Tg(coro1a:CreER) fish. As anticipated, pu.1-deficient microglia were significantly reduced at 3 mpi (Fig. S6C-D). These results suggest that, although pu.1 is dispensable for the immediate survival of microglia (Fig. 3B-C), it is essential for the turnover and long-term maintenance of microglia in adult zebrafish, which could not be fully compensated by spi-b.

pu.1-deficient microglia in mosaic condition were eliminated by cell competition

Given the fact that pu.1-deficient microglia are gradually eliminated in mosaic condition (Fig. 3E-F), but their number remains relatively normal in pu.1^Δ839 null mutants (Yu et al., 2017), we reasoned that the elimination of the pu.1-deficient microglia in mosaic condition is likely attributed to the cell competition between pu.1-deficient DsRed⁺ microglia and their neighboring eGFP⁺ sibling cells harboring WT pu.1. To test this hypothesis, we compared the cell death and proliferation rates of microglia by TUNEL and BrdU incorporation assays in both conditions respectively. To determine the cell death and proliferation rates of pu.1-deficient microglia in mosaic condition, we treated the adult pu.1^KI/Δ839;Tg(coro1a:CreER) fish with 4-OHT to inactivate Pu.1. At 26 dpi, the 4-OHT-treated fish were then injected 5 doses (one dose daily) of BrdU and the brain samples were collected 1 day later for TUNEL and BrdU staining respectively (Fig. 4A). Quantification results showed that the percentage of TUNEL⁺DsRed⁺ microglia (pu.1-deficient cells) in the brains of the pu.1^KI/Δ839;Tg(coro1a:CreER) fish was significantly higher than that of TUNEL⁺eGFP⁺ cells (Fig. 4B). In this assay, the 4-OHT-treated pu.1^KI/+;Tg(coro1a:CreER) fish were included as the control, and as expected, the percentages of TUNEL⁺ cells in the DsRed⁺ and eGFP⁺ microglia pools in control fish were comparable (Fig. 4B). In parallel, we also found that the BrdU incorporation rates of pu.1-deficient microglia (DsRed⁺) in the 4-OHT-treated pu.1^KI/Δ839;Tg(coro1a:CreER) fish was significantly lower than that of eGFP⁺ cells (Fig. 4C), while the BrdU incorporation rates of DsRed⁺ cells and eGFP⁺ cells in the control pu.1^KI/+;Tg(coro1a:CreER) fish was comparable (Fig. 4C). These results indicate that conditional inactivation of Pu.1 in adult zebrafish leads to the impairment of cell proliferation and accelerated cell death of microglia in mosaic condition. To estimate the cell death and proliferation rates of microglia in pu.1-null mutants, we utilized a similar strategy via pulsing the adult pu.1^Δ839/+;TgBAC(ccl34b.1:eGFP) or pu.1^Δ839/Δ839;TgBAC(ccl34b.1:eGFP) fish with 5 doses of BrdU (daily) and collecting the brains samples at 1 day later for TUNEL and BrdU staining respectively (Fig. 4D). As shown in Fig. 4E, the percentage of TUNEL⁺ccl34b.1-eGFP⁺ microglia showed no difference between pu.1^Δ839/Δ839 mutants and pu.1^Δ839/+ siblings, demonstrating that the intrinsic viability of pu.1-deficient microglia in the null mutants was largely unaffected. Similarly, the proliferation capability (as indicated by BrdU incorporation rate) of pu.1-deficient microglia in the null mutants remained normal as well in comparison with control siblings (Fig. 4F). Collectively, these results strongly suggest that the chronic elimination of pu.1-deficient DsRed⁺ microglia is indeed caused by cell competition, which only occurs in the mosaic condition when the eGFP⁺ microglia harboring a single copy of pu.1 are present.

Fig. 4.

The out-competition of pu.1-deficient microglia in mosaic condition is mediated by Tp53

To uncover the key downstream targets that mediate the out-competition of pu.1-deficient DsRed⁺ microglia in mosaic condition, we manually picked DsRed⁺ and eGFP⁺ microglia from the 4-OHT-treated pu.1^KI/Δ839;Tg(coro1a:CreER) fish at 3 mpi and performed transcriptomic analysis (Fig. 5A). To minimize the contamination of DCs, we manually picked 3∼5 DsRed⁺ or GFP⁺ cells from the brain suspension of 3-mpi pu.1^KI/Δ839;Tg(coro1a:CreER) fish in each tube for cDNA library construction and RNAseq. Samples with strong expression of microglia-related genes and faint DC-related genes were chosen for further analysis (Fig. S7A). Among the differential expressed genes (DEGs), the tumor suppressor gene tp53 was one of the top targeted genes robustly upregulated in pu.1-deficient DsRed⁺ microglia (Fig. 5B-C, Fig. S7B). Interestingly, recent studies have revealed that, in addition to regulating cell cycle arrest and apoptosis in DNA damage response (Ou & Schumacher, 2018; Vaddavalli & Schumacher, 2022), Tp53 is also involved in cell competition regulation in both hematopoietic and non-hemopoietic tissues (Bondar & Medzhitov, 2010; Zhang et al, 2017). Hence, we speculated that the cell competition-mediated chronic elimination of pu.1-deficient microglia was likely dependent on Tp53. To support this hypothesis, we conditional inactivated Pu.1 to generate mosaic microglia populations in the tp53-deficient mutant background and asked if loss of Tp53 could rescue the microglia phenotype (Fig. 5D, upper panel). Indeed, results showed that the number of pu.1-deficient DsRed⁺ microglia in the 4-OHT-treated pu.1^KI/Δ839;tp53^-/-;Tg(coro1a:CreER) fish was significantly restored compared with pu.1^KI/Δ839;Tg(coro1a:CreER) control fish (Fig. 5D-F). Collectively, these results demonstrate that the chronic elimination of pu.1-deficient microglia in mosaic condition is mediated, at least in part, by the elevation of Tp53 activity, which leads to the reduced proliferation and excessive death of these microglia.

Fig. 5.

Dosage-dependent regulation of microglia maintenance by PU.1 is evolutionarily conserved from zebrafish to mice

The above data have demonstrated that the long-term maintenance of microglia in adult zebrafish is regulated by Pu.1 in a dosage-dependent manner. We next wondered whether this mechanism is evolutionarily conserved from teleost to mammals. To address this issue, we first examined the expression of Pu.1 and Spi-b in adult murine microglia. Quantitative RT-PCR revealed that in contrast to the robust expression of Pu.1, Spi-b was barely detectable in microglia (Fig. S8A). We therefore only focused on Pu.1 and generated a Pu.1^Fl allele, in which the exon 2 of Pu.1 gene is flanked by loxP (Fig. S8B-C). The Pu.1^Fl mice were then crossed with the Cx3cr1^CreER-YFP strain (referred to as to Cx3cr1^CreERhereafter) (Yona et al, 2013) to generate Pu.1^Fl/+;Cx3cr1^CreERand Pu.1^Fl/Fl;Cx3cr1^CreER mice for conditional PU.1 inactivation study (Fig. 6A).

Fig. 6.

To investigate whether microglia maintenance in adult mice requires PU.1, we intraperitoneally injected tamoxifen (TAM) (Askew et al., 2017) into the Pu.1^Fl/+;Cx3cr1^CreER and Pu.1^Fl/Fl;Cx3cr1^CreERmice, and collected the brain samples at 7 dpi for cryo-section and IBA1 staining (Fig. 6A, upper panel). As expected, without TAM, the number of the IBA1⁺ microglia showed no difference in the brain of Pu.1^Fl/+;Cx3cr1^CreER and Pu.1^Fl/Fl;Cx3cr1^CreER mice (Fig. S8D-E). However, after TAM injection, the number of the IBA1⁺ microglia in Pu.1^Fl/Fl;Cx3cr1^CreER mice were drastically reduced compared to that in Pu.1^Fl/+;Cx3cr1^CreER mice (Fig. 6A-C), suggesting that PU.1 is indeed required for the survival of adult microglia, which is similar to the double depletion of pu.1 and spi-b in zebrafish (Fig. S4). To further explore whether partial loss of PU.1 activity could reduce the competitiveness of microglia and impairs the long-term maintenance of microglia in mice, we intraperitoneally injected sub-dosage of TAM into adult Pu.1^Fl/+;Cx3cr1^CreERmice to generate mosaic microglia populations for long-term tracing. To quantify the proportion of Pu.1^KO/+microglia and Pu.1^Fl/+ microglia, we sorted YFP⁺ microglia from TAM-injected Pu.1^Fl/+;Cx3cr1^CreER mice at 3 dpi and 3.5 mpi, the timeline similar to previous research in zebrafish. Then we amplified Pu.1^KO and Pu.1^Fl alleles in the gDNA pool by PCR for T-clone quantification (Fig. 6D). As shown in Fig. 6E-F, while the percentage of Pu.1^KO/+ microglia was comparable to Pu.1^Fl/+ microglia at 3 dpi, it was dramatically reduced by 3.5 mpi, suggesting that removal of a single copy of Pu.1 leads to the chronic elimination of Pu.1^KO/+ microglia, which appears to occur only in the presence of the WT Pu.1^Fl/+ microglia, as PU.1 heterozygous mice show no difference in the viability of tissue resident macrophages (Karpurapu et al., 2011). Collectively, these results indicate that the dosage-dependent cell competition-mediated regulation of microglia maintenance by SPI1/PU.1 is evolutionarily conserved from teleost to mammals.

Discussion

In the present study, to explore the role of Pu.1 in microglia survival and maintenance, we established a visible conditional knockout pu.1^KIallele in zebrafish via the non-homologous end joining (NHEJ)-mediated knock-in method. Although conditional inactivation of targeted genes has been reported by several independent studies (Li et al., 2015; Li et al, 2019; Shin et al, 2023), it has not been widely applied in zebrafish, possibly due to the difficulty or low efficiency to generate loxP-flanked alleles. Compared to homologous recombination (HR), NHEJ-mediated knock-in of reporter gene has been shown to display 10-fold higher efficiency (Auer et al, 2014; Li et al., 2015). In our case, the germline transmission to F1 generation for pu.1^KI allele is around 15% (Fig. S1). Meanwhile, in our investigation, we noted that the efficiency of Cre-mediated excision of loxP sites at the pu.1 locus in zebrafish (Fig. 3) appears much lower than that in mouse (Fig. 6). Although it remains to be further validated whether this observation is a general phenomenon, it does highlight the necessity and significance of chimera study of interested genes. In our pu.1^KIallele, the expression of DsRed directly indicates Cre-mediated excision and helps us distinguish the mutant cells from their sibling cells, which serves as a good example for chimera study in the future. Moreover, our knock-in allele can also be simply regarded as a loxP-flanked fluorescent reporter line, which not only facilitates the labeling of cells that express gene of interest, but also can be used in combination with the LEGO-IR system (Xu et al., 2015) to trace the fate of these cells. Thus, due to the high integration efficiency of donor DNA fragment and the multifaceted applications of the labeling strategy, we anticipated that our knock-in system would be a powerful tool and broadly be adopted in the fields of genetics, developmental biology and cell biology for the chimeric study of interested genes and fate mapping of interested cells in the future.

Via EdU pulse-chase labelling assay, we found that the daily turnover rate of microglia in adult zebrafish is about 2.5%, much higher than that reported in mice and human (Lopez-Atalaya et al, 2018). These disparities might be explained by several reasons. First, in our calculation of microglia turnover rate, we did not factor the unknown cell cycle length of zebrafish microglia. It is possible that the duration of microglia proliferation in zebrafish is much shorter than that in mice. If microglia proliferation takes less than 24 hours in zebrafish, then the calculated daily turnover rate of 2.5% may be overestimated. Second, it is well recognized that zebrafish manifest much stronger regeneration capability than mammals. As such, differentiated cells, like cardiomyocytes and hair cells in zebrafish are in general more proliferative than those in mammals (Chen et al, 2019; Marques et al, 2019). In this regard, microglia in zebrafish might be pre-coded both genetically and epigenetically to be more proliferative. Finally, microglia turnover could potentially be affected by the distinct efferocytotic burdens via the lysosome pathway, as lysosome has been known to be a convergent point for many biological process, including metabolism, cellular senescence, DNA repair, ER stress and even apoptosis (Franco-Juárez et al, 2022; Perera & Zoncu, 2016). Indeed, efferocytosis of apoptotic cells has previously been shown to elicit NOX2 assembly and promote ROS production (Yvan-Charvet et al, 2010), which in turn triggers cell death via induction of lysosome damage (Patra et al, 2023; Wang et al, 2018). The more regenerative environment in zebrafish might be accompanied by higher frequency of cell death, which could lead to the increase of efferocytotic burdens of microglia, thereby accelerating the turnover of microglia. Further studies will be required to clarify this issue.

Our results revealed that the survival and maintenance microglia are highly sensitive to the dosage of PU.1/SPI1 in both teleost and mammals. Interestingly, a previous study reported that hetero- or homozygosity of PU/ER(T) allele, in which the estrogen receptor (ER) ligand binding domain-G525R is fused with the full-length PU.1 to block its nucleus translocation, showed no effect on the viability of alveolar macrophages (Karpurapu et al., 2011). This contradictory result might potentially result from the leakiness of ER, which was incapable to fully retain PU.1 in cytoplasm. A recent work found that conditional inactivation of PU.1 by feeding the animals with tamoxifen-containing diet in the Alzheimer’s disease model CK-p25 mice led to a 33% reduction in microglia number at their young ages. Although the authors considered prevention of microgliosis as a result of the amelioration of neuroinflammation (Ralvenius et al., 2023), our data implies that reduction in microglia could also been interpreted by the direct involvement of PU.1 in microglia survival and maintenance, as our data clearly demonstrate that conditional ablation of the PU.1 in mice or Pu.1/Spi-b in zebrafish leads to the rapid depletion of microglia in the brain (Fig. 6; Fig. S4). As PU.1 is a master regulator that controls the expression of series of genes required for macrophage/microglia development (Satoh et al., 2014), rapid death of Pu.1-null microglia might probably be caused by the orchestration of multiplex factors, including the downregulation of colony-stimulating factor 1 receptor (Csf1r) (the essential regulator controlling the survival and maintenance of microglia in both mice and zebrafish) (Elmore et al., 2014; Yu et al., 2023), the loss of anti-apoptotic genes, such as BCL2A1 (Jenal et al, 2010), and absence of other genes that maintain the integrity of cells. Intriguingly, differing from the rapid death of microglia after complete inactivation of PU.1 or Pu.1/Spi-b, microglia with a partial loss of PU.1 or Pu.1 are chronically eliminated only in the mosaic condition via cell competition, in which the microglia with WT Pu.1 or a single copy of pu.1 are present (Fig. 3 and Fig. 6). Moreover, we further showed that the chronic elimination of pu.1-dificent microglia in zebrafish is mediated by Tp53-dependent cell competition, which has been shown to depend on the relative level of Tp53 in competing cells (Bondar & Medzhitov, 2010; Zhang et al., 2017). Previously, we have shown that the level of csf1ra, the zebrafish orthologue of mammalian Csf1r, is highly associated with the fitness of microglia and regulates the turnover and maintenance of microglia via cell-competition (Yu et al., 2023). During the process of investigation, we also wondered whether the cell competition between pu.1-deficient microglia and pu.1-sufficient microglia relied on Csf1ra. However, in contrast to the robust upregulation of p53 in pu.1-deficient microglia, the decrease of csf1ra was not observed (Fig. S9). In summary, our findings suggest that the dosage of PU.1/Pu.1 may serve as a checkpoint to determine the fitness of microglia. It will be of great interest to explore the upstream events that modulate PU.1/pu.1 expression during microglia turnover.

Materials and Methods

Zebrafish

Zebrafish were maintained according to standard protocol as described previously (Westerfield, 2000). AB wild-type, pu.1^Δ839 (Yu et al., 2017), spi-b^Δ232 (Yu et al., 2017), tp53^M214K (Berghmans et al, 2005), pu.1^KI (short for pu.1^KI-eGFP), pu.1^CKO (short for pu.1^KI-DsRed), TgBAC(ccl34b.1:eGFP) (Wu et al., 2020), Tg(coro1a:CreER^T2)^sz101tg, Tg(mpeg1:loxP-DsRedx-loxP-eGFP, shorted as LRLG)^hkz015t (Xu et al, 2016) were used in this study.

Mouse strains

Pu.1^Fl (No.T010980, purchased from GemPharmatech Co.,Ltd) and Cx3cr1^CreER-EYFP (Stock No. 021160) (Yona et al., 2013) were used in this study. All animal experiments and procedures were proved by Ethics Committee for the Welfare of Laboratory Animals in Shenzhen PKU-HKUST Medical Center.

Generation of pu.1^KI allele

Donor plasmid was modified based on the reported Th-KI plasmid (Li et al., 2015). In brief, DNA fragments containing pu.1 exon 4 to exon 6 and the 3’ downstream fragment were cloned from cDNA and genomic DNA libraries respectively and inserted into the plasmid containing P2A-eGFP, P2A-Dsred and loxP elements. sgRNA was designed in the intron 3∼4 of pu.1 with the online website Crisprscan (https://www.crisprscan.org/). It was transcribed by MEGAshortscript Kit (AM1354, Invitrogen) from PCR product which was amplified following reported protocol (Vejnar et al, 2016). Then Cas9 mRNA (700 ng/ul), sgRNA (70 ng/ul) and donor plasmid (15 ng/ul) were mixed and injected into zebrafish embryos at one-cell stage. Embryos with eGFP signal were selected at 3 dpf and raised to adulthood for germline transmission. Adult F0 fish were crossed with wild-type fish to generate F1 and eGFP⁺ F1 were selected. The 5’- and 3’-junctions of the donor plasmid integration site were amplified from the genomic DNA of F1 embryos and sent for sequencing. The primers used for 5’ junction are 5’-GATCTATCGACCACCAATGGAG-3’ and 5’-GCCATAGTGTGCATTCTCAGG-3’ and for 3’ junction are 5’-GTTGTAAAACGACGGCCAG-3’ and 5’-GAGTGTAGTGCTCATTCAAGC-3’.

Cryo-section

Adult fish were anesthetized on ice. The tissue was dissected and fixed in 4% PFA at 4 [overnight. After being washed with phosphate-buffered saline (PBS) at room temperature for 1 day, the tissues were dehydrated with 30% sucrose at 4 [overnight, then soaked in coagulating solution (optimal cutting temperature compound, OCT) and subjected to cryosection with 30-μm thickness.

EdU and BrdU incorporation assays, TUNEL staining, and immunostaining

EdU and BrdU incorporation assays were conducted as previously reported (Yu et al., 2023). In brief, EdU (A10044, Invitrogen) or BrdU (B5002, Sigma-Aldrich) was dissolved in PBS to the final concentration of 10 mg/ml. 5 ul EdU or BrdU was intraperitoneally (i.p.) injected into adult zebrafish each time. Brain samples were dissected at 1 day post injection and fixed in 4% PFA for cryosection.

EdU detection was conducted with (Click-iT™ EdU Cell Proliferation Kit for Imaging, Alexa Fluor™ 647 dye, Invitrogen, C10340) according to the manufacturer’s instructions. For BrdU detection, samples were treated with 2M HCl for half an hour and then stained with the BrdU antibody. TUNEL staining was performed with TUNEL BrightGreen Apoptosis Detection Kit (A112, Vazyme) according to the manufacturer’s instructions. Antibody staining was performed after EdU/BrdU and TUNEL detection. Primary antibodies used in this study were mouse anti-BrdU (11170376001, Roche), goat anti-GFP (ab6658, Abcam), rabbit anti-DsRed (632496, Clontech), rabbit anti-Lcp1 (Jin et al, 2009), rabbit anti-Pu.1 (Jin et al, 2012) and rabbit anti-IBA1 (019-19741, FUJIFILM Wako) antibodies.

Tamoxifen injection

4-Hydroxytamoxifen (4-OHT, H7904, Sigma) was dissolved in ethanol to final concentration of 10 mM. For adult fish, 1 ul 4-OHT was i.p. injected into each fish for 3 times. For adult mice, tamoxifen (T5648, Sigma) was dissolved in corn oil to final concentration of 20 mg/ml and i.p. injected into the mice with 35 mg/kg (sub-dosage) or 150 mg/kg (over-dosage).

Neutral red staining

Neutral red staining was performed as previously described (Herbomel et al, 2001).

Cy5-Annexin V treatment

Annexin V-Cy5 Reagent (1013, Biovision) solution was directly injected into brain ventricle of 3 dpf larva and immediately imaged by confocal microscope.

Imaging

Fluorescent signals from live reporter lines or immunostaining were imaged under a Leica SP8 confocal microscope or a Zeiss LSM980 confocal microscope. The objectives HC PL APO 203/0.70 DRY (Leica) and Plan-Apochromat 20×/0.8 M27 (Zeiss) were used in this study.

cDNA preparation and RNA-seq

Cell suspensions were prepared as previously described (Yu et al., 2017). Cells of interest were manually picked under the fluorescent microscope (Nikon Ti-S) equipped with the micro-manipulator (NT-88-V3, Nikon). Subsequently, picked cells were transferred by the glass needle to lysis buffer (mainly including Triton X-100 and RNase inhibitor) for thorough vertex. Then the cell lysate was used for reverse-transcription and amplification with the Smart-Seq2 method (Picelli et al, 2014) to generate the cDNA library. cDNA was sent to Novogene Company for Illumina sequencing with an average depth of 6×10⁶ raw reads per sample. Sequence data were aligned to zebrafish genome by STAR (Spliced Transcripts Alignment to a Reference). Samples with strong expression of microglia-related genes and faint DC-related genes were chosen for further analysis.

Mouse microglia isolation

Adult mice brain was dissected and cut into small pieces. They were dissociated into single-cell suspensions with Adult Brain Dissociation Kit (130-107-677, Miltenyi Biotec) according to the manufacturer’s instructions. Then Myelin Removal Beads II (130-96-733, Miltenyi Biotec) were used to remove myelin debris from single-cell suspensions. YFP⁺ Microglia were sorted from suspensions by FACS Aria IIIu. The sorted microglia were lysed by proteinase K to extract DNA or TRIzol to extract RNA.

TA clone assay

Pu.1^Fl and Pu.1^KO alleles were amplified by PCR from the genomic DNA. The common forward primer was 5’-GCATCGCATTGTCTGAGTAGGT-3. The reverse primers were 5’-AAATCTGCCTGGGTGACCTTC-3’ for Pu.1^Fl and 5’-ACACAACGGGTTCTTCTGTTAG-3’ for Pu.1^KO. The PCR products for Pu.1 ^Fl and Pu.1 ^KO alleles were 248 bp and 129 bp respectively. PCR products were ligated to pUCm-T vector (B522211, Sangon) with Blunt/TA Ligase Master Mix (M0367S, NEB) according to the manufacturer’s instructions. The clones containing ligated plasmids were amplified with M13F (5’-GTTGTAAAACGACGGCCAG-3’) and M13R (5’-CAGGAAACAGCTATGAC-3’). The PCR products for Pu.1 ^Fl and Pu.1 ^KO alleles were 437 bp and 315 bp respectively. Clones of Pu.1^Fl and Pu.1^KO alleles were distinguished by the size of PCR products.

Quantitative PCR

cDNA was synthesized from extracted RNA with SuperScript™ III Reverse Transcriptase kit (18080093, Invitrogen). Quantitative PCR was performed with iTaq™ Universal SYBR® Green Supermix (1725121, Bio-rad) on a CFX96 Dx Instrument (Bio-rad) and analyzed using the ΔΔCt method. The following primers were used to determine the expression of DsRed and pu.1 from pu.1^CKOallele: DsRed, 5’-GATCTATCGACCACCAATGGAG-3’ (pu.1-161FP)/ 5’-GCCGTTCACGGAGCCCTCCAT-3’ (DsRed-72RP); pu.1, 5’-GATCTATCGACCACCAATGGAG-3’ (pu.1-161FP)/ 5’-CGCATGTAGTGACTGCACGC-3’ (pu.1-357RP). The following primers were used to determine the expression of Pu.1 and Spi-b in mouse microglia: Gapdh, 5’-TGTGTCCGTCGTGGATCTGA-3’/5’-TTGCTGTTGAAGTCGCAGGAG-3’; Pu.1, 5’-GAGGTGTCTGATGGAGAAGCTG-3’/5’-ACCCACCAGATGCTGTCCTTCA-3’; Spi-b, 5’-AGGAGTCTTCTACGACCTGGAC-3’/5’-GGAGTGGCTAAAGGCAGCAGTA-3’.

Quantification and Statistics

The number microglia in embryos were quantified manually, and then genotyped to prevent bias. For the experiment with adults, fish were genotyped first and then numbered without gender bias for downstream 4-OHT injection, brain sample collection, cryo-section and immunostaining experiments. For microglia number quantification, 3 representative slices from anterior, middle and posterior of the midbrain of numbered fish were chosen for further manual counting with the same criterial. For EdU/BrdU or TUNEL experiments, at least 6 representative slices from anterior, middle and posterior of the midbrain were chosen for calculation. The number of EdU/BrdU⁺ or TUNEL⁺ microglia/DCs from the slices were counted and then divided by the total number of microglia/DCs to calculate the ratio.

Adult mice were genotyped first and then only males were chosen for the downstream experiments. For statistics, cortex, hippocampus and thalamus on the slices of mouse brain were chosen for microglia density quantification. The number of microglia and region of interest were calculated with Imaris 9.9 software to determine the density.

Statistical analyses are performed with Graphpad Prism 9.5.1. F test was first used to check the variances of two groups. If the variances of two groups were not significantly different (P>0.05), Student’s t test was performed. If the variances of the two groups were significantly different (P<0.05), t test with Welch’s correction was performed. A result was considered significant if P <0.05. Values represent mean ± standard error of the mean.

Acknowledgements

We thank Dr. Keng Chen and Dr. Keyu Chen in Shenzhen Bay Laboratory for their assistance in the Fluorescent-activated cell sorting (FACS) of mouse microglia, and thank Prof. Jin Xu in South China University of Technology for his kind suggestions on this project.

Additional information

Funding

This work is supported by National Natural Science Foundation of China Grant (31801211), Natural Science Foundation of Guangdong Province Grant (2024A1515030122), Shenzhen Medical Academy of Research and Translation (SMART) 2023 Shenzhen Medical Research Funding (B2302034), and Research Grants Council of the Hong Kong Special Administrative Region Grants (16103920, AoE/M-09/12, T13-605/18-W, and T13-602/21-N).

Additional files

Supplemental Figure 1-9