The ETS transcription factor Spi2 regulates hematopoietic cell development in zebrafish

Vertebrate blood development comprises primitive and definitive hematopoiesis, with definitive hematopoiesis generating hematopoietic stem and progenitor cells (HSPCs) capable of self-renewal and differentiation into all hematopoietic lineages (Kondo et al., 2003; Jagannathan-Bogdan and Zon, 2013). The definitive HSPCs arise from hemogenic endothelial cells (HECs) in the floor of the dorsal aorta (DA) in the aorta-gonad-mesonephros (AGM) through endothelial-to-hematopoietic transition (EHT) (Bertrand et al., 2010; Boisset et al., 2010; Kissa and Herbomel, 2010). After emergence from the AGM, HSPCs migrate to the intermediate hematopoietic tissues, the fetal liver in mammals or the caudal hematopoietic tissue (CHT) in zebrafish (Murayama et al., 2006; Jin et al., 2007), and finally colonize the adult hematopoietic tissues, the bone marrow in mammals or the kidney marrow (KM) in zebrafish (Jagannathan-Bogdan and Zon, 2013).

The E26 transformation-specific or E-twenty-six (ETS) genes encode a superfamily of transcription factors containing a conserved DNA-binding ETS domain (Sharrocks, 2001; Sizemore et al., 2017). In mammals, the ETS family consists of 27-28 members that are involved in many cellular processes, including hematopoiesis (Sizemore et al., 2017). For example, the ETS variant 2 (Etv2) is a key regulator for hemogenic angioblast formation and HEC specification (Liu and Patient, 2008; Sumanas and Choi, 2016; Zhao et al., 2022). Likewise, the SPI-subfamily, including PU.1, SPIB and SPIC, are involved in myeloid fate commitment and B cell development (Scott et al., 1994; Su et al., 1997; Kohyama et al., 2009; Willis et al., 2017). In zebrafish, 31 ETS genes have been identified (Liu and Patient, 2008), and some are zebrafish-specific, presumably resulting from a genome duplication event in ray-fin phylogeny (Postlethwait et al., 2000). These duplicated genes may take the partitioning function of the ancestral gene or acquire a new adaptive function (Force et al., 1999). Hence, understanding zebrafish-specific ETS genes may provide evolutionary insights into gene subfunctionalization and neofunctionalization.

In this study, we reported the identification and characterization of a ray-finned fish-specific ETS transcription factor Spi2 and its role in hematopoietic cell development.

To investigate the mechanism governing the specification of HECs and conventional endothelial cells (cECs), we have previously isolated HECs and cECs from the floor and roof wall of the DA and performed single cell RNA-sequencing (scRNA-seq) (Zhao et al., 2022). Differentially expressed gene (DEG) analysis revealed that spi2 was highly expressed in the HECs (Fig. 1A). The Ensembl genome browser showed that spi2 was found exclusively in ray-finned fishes but not in birds and mammals (Fig. 1B). Phylogenic tree analysis showed that spi2 represented a separated evolutionary branch different from other paralogs which have corresponding orthologs in mammals (Fig. 1C). Protein sequence alignment revealed that Spi2 shared 34%∼40% similarity with other paralogs in overall protein sequences and 55%∼61% similarity in the ETS domain (Fig. 1D; Fig. S1A,B). To confirm the spi2 expression pattern, we performed whole-mount in situ hybridization (WISH). Results showed that spi2 was first detected in the AGM at ∼24 h post-fertilization (hpf) and later was evident in both AGM and CHT at 2 days post-fertilization (dpf) (Fig. 1E; Fig. S1C). Fluorescence in situ hybridization (FISH) further confirmed that spi2 was expressed in 70.0±8.9% (mean±s.d.) of the DA floor flk1:GFP⁺ cells at 2 dpf (Fig. 1F,G), where HECs reside (Bertrand et al., 2010; Kissa and Herbomel, 2010; Zhao et al., 2022), and later was also expressed in the subsets (31.2±5.9%) of cd41:eGFP⁺ cells in the CHT at 3 dpf, which presumably represent definitive HSPCs and thrombocytes (Lin et al., 2005; Bertrand et al., 2008) (Fig. 1F,G). These observations indicate that spi2 is specifically expressed in HECs and HSPCs during zebrafish early development.

Given its specific expression pattern, we reasoned that establishing a spi2 reporter line would provide a useful tool for studying HEC and HSPC emergence. We therefore used bacteria artificial chromosome (BAC) clone CH211-45O10, a 200-kilobase (kb) pairs DNA fragment covering both the upstream and downstream sequences of the spi2 gene locus, to generate the spi2 reporter BAC(spi2:P2a-GFP) construct by replacing the stop codon with P2a-GFP sequences (Fig. 2A). The BAC(spi2:P2a-GFP) constructs were used to generate the spi2 reporter line. A TgBAC(spi2:P2a-GFP) line with stable GFP expression was successfully obtained. To characterize the TgBAC(spi2:P2a-GFP) line, we outcrossed it with Tg(flk1:mCherry) fish, in which HECs and cECs are labeled by mCherry (Bertrand et al., 2010). Examination of the TgBAC(spi2:P2a-GFP);Tg(flk1:mCherry) fish showed that the spi2-GFP⁺ cells were restricted to the floor of the DA at 30 hpf (Fig. 2B). Time-lapse imaging further showed that the spi2-GFP⁺ cells in the DA floor could undergo EHT (Fig. 2C), confirming their HEC identity. In parallel, we also outcrossed this reporter line with Tg(gata1a:DsRed) or Tg(coro1a:DsRed) fish, where erythrocytes or leukocytes are marked by DsRed, respectively (Traver et al., 2003; Xu et al., 2015). Examination of these double transgenic lines showed that the spi2-GFP signals were detected in gata1a:DsRed⁺ erythroid progenitors and coro1a:DsRed⁺ leukocyte progenitors in the CHT at 3 dpf (Fig. 2D; Fig. S2A). These observations indicate that the TgBAC(spi2:P2a-GFP) reporter line recapitulates endogenous spi2 expression.

To examine spi2 expression in adult hematopoiesis, we analyzed the KM of adult TgBAC(spi2:P2a-GFP) zebrafish using fluorescence-activated cell sorting (FACS). Results showed that spi2-GFP⁺ cells were predominantly detected in the myeloid (84.3±1.7%) and progenitor (HSPCs) (8.2±1.5%) populations (Fig. 2E). Similar observations were also found in the KM of adult TgBAC(spi2:P2a-GFP);Tg(coro1a:DsRed) double transgenic zebrafish, where GFP⁺DsRed⁺ cells were found (98.1±1.8%) in the myeloid and HSPC populations (Fig. S2B,C). These observations suggest that spi2 is predominantly expressed in HSPCs and myeloid lineages in adult hematopoiesis. To further support this conclusion, we isolated erythrocytes (globin:eGFP⁺ cells), T cells (lck:DsRed⁺ cells), thrombocytes (cd41:eGFP^high cells), HSPCs (cd41:eGFP^low cells) and myeloid cells (pu.1: GFP⁺ cells) from adult KM to examine the expression of endogenous spi2. qPCR analysis revealed that spi2 was detected exclusively in the pu.1: GFP⁺ and cd41:eGFP⁺ cells (Fig. 2F). Collectively, these data show that spi2 expression is restricted to HSPCs, myeloid cells and thrombocytes in adult hematopoiesis.

To explore the role of spi2 in hematopoiesis, we generated a spi2-deficient mutant, in which 546 base pairs (bp) of the spi2 genomic sequences were deleted (Fig. 3A), resulting in the synthesis of a truncated protein lacking the ETS domain (Fig. 3B). Further qPCR analysis showed that the truncated spi2 transcripts were significantly reduced in spi2 mutants (Fig. S3A), indicating that the spi2 mutation is likely to be a null allele.

To dissect the role of spi2 in hematopoiesis, we first examined the formation of embryonic erythroid and myeloid cells in spi2-deficient mutants at 22 hpf and 30 hpf. WISH showed that the expression of erythrocyte-marker gata1a (Rhodes et al., 2005), macrophage-marker mfap4 (also known as mfap4.1) (Zakrzewska et al., 2010) and neutrophil-marker lyz (Hall et al., 2007) were all intact in spi2 mutants (Fig. S3B,C), indicating that spi2 is dispensable for primitive hematopoiesis. To explore the role of spi2 in definitive hematopoiesis, we outcrossed spi2 mutants with the Tg(flk1:eGFP) or Tg(flk1:NLS-Eos) fish, in which cECs and HECs are marked by GFP or NLS-Eos (Kissa and Herbomel, 2010; Zhao et al., 2022). Examination of the Tg(flk1:eGFP) transgenic zebrafish showed that the DA and HECs, as indicated by the GFP signal at 30 hpf and runx1 WISH at 28 hpf, respectively, were intact in spi2 mutants (Fig. 3C,D). Time-lapse imaging of Tg(flk1:NLS-Eos) transgenic zebrafish further revealed comparable EHT events in siblings (Movie 1) and spi2 mutants (Fig. S3D; Movie 2), indicating normal formation of HECs and HSPCs in spi2-deficient mutants. This conclusion was further validated by the FISH results showing that c-myb⁺ (myb⁺) cells in the AGM were comparable between spi2 mutants and siblings at 36 hpf (Fig. 3E,F). From 2 dpf onwards, however, the c-myb⁺ cells in the CHT, where HSPCs are known to undergo rapid expansion (Murayama et al., 2006; Jin et al., 2007), in the mutant embryos were significantly reduced (Fig. 3E,F), suggesting that the maintenance of HSPCs in the CHT is impaired in spi2 mutants. To further consolidate this conclusion, we outcrossed spi2 mutants with the Tg(cd41:eGFP) zebrafish and quantified the number of cd41:eGFP^low HSPCs (Kissa et al., 2008). Indeed, the number of the cd41:eGFP^low cells in the CHT was markedly reduced in the mutants (Fig. S3E,F). Collectively, these observations demonstrate that spi2 is dispensable for primitive hematopoiesis and the emergence of definitive HSPCs, but essential for the maintenance of definitive HSPCs in the CHT.

Previous studies have showed that definitive HSPCs are heterogeneous and are capable of differentiating into multiple lineages during early development (Jagannathan-Bogdan and Zon, 2013). To further define which hematopoietic lineages are impaired in spi2 mutants during early development, we performed hbae1 FISH (hbae1.1; erythrocytes), rag1 FISH (T cells), anti-lyz:DsRed antibody staining (neutrophils) and anti-mpeg1:DsRed antibody staining (mpeg1.1; macrophages) at 5 dpf or 15 dpf when definitive hematopoietic lineages become predominant (He et al., 2018). Results showed that spi2 mutants exhibited reduced neutrophils, macrophages and erythrocytes, but normal T cells (Fig. 3G,H), indicating that spi2-deficiency predominantly affects erythroid and myeloid HSPCs during early embryogenesis.

As spi2 transcripts are highly expressed in adult hematopoietic cells (Fig. 2E,F), we next investigated whether spi2 deficiency would affect adult hematopoiesis. To address this issue, whole KM cells from adult spi2 mutants and siblings were collected and subjected to FACS analysis. We found that the myeloid lineage, but not erythroid and lymphoid lineages, was significantly decreased in spi2 mutants, accompanied by a moderate increase of progenitor population (Fig. 3I,J), showing that adult myelopoiesis is also impaired in spi2 mutants. May-Grunwald-Giemsa (MGG) staining of whole KM cells further revealed that spi2 mutants had reduced numbers of neutrophils and macrophages, accompanied by a higher number of undifferentiated myeloblasts (Fig. 3K; Fig. S3G), indicating that spi2 is required for adult myeloid development, possibly at the stage of myeloblasts differentiating towards neutrophils and monocytes/macrophages.

To delineate the cellular basis underlying the HSPC phenotype in spi2 mutants, we performed terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay and anti-phospho-histone 3 (PH3) staining to monitor the cell death and proliferation of HSPCs in the CHT, respectively. TUNEL assay showed that the mutants and siblings had a comparable percentage of c-myb⁺TUNEL⁺ HSPCs (Fig. S4A). In contrast, a significant decrease in the percentage of c-myb⁺PH3⁺ HSPCs was observed in spi2 mutants (Fig. 4A,B), indicating that the reduction of HSPCs in spi2 mutants is largely attributed to the impaired proliferation of HSPCs in the CHT. Although the number of HSPCs in spi2 mutants showed slight recovery by 5 dpf (Fig. 3F), the number of HSPCs in the mutants remained significantly lower than that in siblings (Fig. 3H), consistent with the continuously lower proliferative rate of HSPCs at 5 dpf in the mutants (Fig. S4B).

It is well-known that transcription factor Runx1 is essential for HEC fate specification, EHT and HSPC proliferation (Kissa and Herbomel, 2010; Eliades et al., 2016; Gao et al., 2018; Yzaguirre et al., 2018; Zhao et al., 2022). Double fluorescence in situ hybridization (dFISH) showed that spi2 and runx1 were largely overlapped in the floor of the DA (Fig. S4C). Based on the expression pattern of spi2 and runx1 (Bonkhofer et al., 2019) and the phenotype of spi2- and runx1-deficient mutants (Jin et al., 2009; Kissa and Herbomel, 2010), we speculated that spi2 would likely be a downstream target of runx1. To elucidate the epistatic relationship between spi2 and runx1, we probed the expression of spi2 in runx1 mutants (Jin et al., 2009) and showed that spi2 was absent in runx1 mutants (Fig. 4C), indicating that spi2 acts downstream of runx1. To further support this conclusion, we took advantage of Tg(flk1:runx1-P2a-GFP) fish (Zhao et al., 2022), in which runx1 expression is under control of the endothelial flk1 promoter, to test whether overexpressing runx1 could induce spi2 expression. Indeed, increased spi2 expression was observed in the floor of the DA in the runx1-overexpression embryos (Fig. 4D). To quantitatively compare the expression of spi2 in runx1-overexpressed and control embryos, we crossed Tg(flk1:mCherry) with Tg(flk1:runx1-P2a-GFP) fish and sorted mCherry⁺GFP⁺ (HECs and cECs with runx1 overexpression) and mCherry⁺GFP⁻ (HECs and cECs without runx1 overexpression) cells from the trunk region of Tg(flk1:mCherry);Tg(flk1:runx1-P2a-GFP) or Tg(flk1:mCherry) embryos at 28 hpf, respectively (Fig. S4D). qPCR analysis revealed upregulated spi2 expression in runx1-overexpressed mCherry⁺GFP⁺ endothelial population (Fig. 4E), demonstrating that spi2 indeed acts genetically downstream of runx1. However, forced expression of spi2 failed to rescue the formation of HSPCs in runx1 mutants as indicated by the lack of c-myb⁺ cells in the AGM at 32 hpf (Fig. 4F), consistent with the finding that spi2 is dispensable for HSPC emergence (Fig. 3E). To examine the role of runx1-spi2 axis in HSPC expansion in the CHT, we overexpressed runx1 by injecting the hsp70l-runx1-P2a-GFP DNA constructs into spi2 mutants and siblings, and we found that overexpressing runx1 led to a robust increase of c-myb⁺ HSPCs in the CHT in siblings, but this increase was reduced in spi2 mutants (Fig. 4G), suggesting that Spi2 works together with other Runx1 downstream factor(s) to regulate HSPC expansion in the CHT.

In conclusion, our study identified a previously uncharacterized ETS factor Spi2 in zebrafish and revealed its essential role in definitive hematopoiesis. Despite the lack of corresponding mammalian ortholog, Spi2 may take the partitioning function of Pu.1, a conserved master regulator of myeloid lineage development in fish and mammals (Scott et al., 1994; Burda et al., 2010; Yu et al., 2017). The specific expression of Spi2 in HECs and HSPCs and its role in HSPC maintenance indicate the adaptive neofunctionalization. It will be of interest to perform cross-species functional analysis of Spi2 to gain insights into the gene diversification.

Zebrafish were maintained according to standard protocols (Westerfield, 2000). Embryos used for experiments were kept at 28.5°C in 0.5× E2 medium containing methylene blue (egg water) and changed to 0.003% N-Phenylthiourea (P7629, Sigma-Aldrich) egg water at 1 dpf to block pigmentation. The following strains were used: AB, Tg(flk1:eGFP)s843 (Jin et al., 2005), Tg(cd41:eGFP) (Lin et al., 2005), Tg(coro1a:DsRed) (Xu et al., 2015), Tg(gata1:DsRed) (Traver et al., 2003), Tg(globin:eGFP) (Ganis et al., 2012), Tg(lck:DsRed) (Tian et al., 2017), Tg(mpeg1:DsRed) (He et al., 2018), TgBAC(c-myb:GFP) (Zhang et al., 2012), Tg(flk1:runx1-P2a-GFP)hkz041 (Zhao et al., 2022), Tg(flk1:NLS-Eos)ncv6 (Fukuhara et al., 2014), TgBAC(spi2:P2a-GFP)hkz046 and spi2^hkz18. This research was conducted according to the guidelines of the Animal and Plant Care Facility and approved by the Animal Ethics Committee of the Hong Kong University of Science and Technology.

The spi2 mutants were created using the CRISPR/Cas9 system (Chang et al., 2013). Briefly, two sgRNAs were designed using the CRISPR design tool (http://www.e-crisp.org/E-CRISP/) to target exon 3 (5′-3′: GCGAGGAATAAACACCCTAAAGG) and exon 4 (5′-3′: GGCTATAATTTCGCAAGGCCCGG) of the spi2 gene. The double strand DNA templates for spi2 gRNAs synthesis were PCR amplified from the pMD-19-T-gRNA scaffold vector (Chang et al., 2013) (primers used for amplification: gRNA FP1: 5′-TAATACGACTCACTATAGGGCGAGGAATAAACACCCTAAGTTTTAGAGCTAGAAATAGC-3′; gRNA FP2: 5′-TAATACGACTCACTATAGGGCTATAATTTCGCAAGGCCGTTTTAGAGCTAGAAATAGC-3′; gRNA RP: 5′-AGCACCGACTCGGTGCCACT-3′) and purified with the QIAquick PCR Purification Kit (28104, Qiagen). gRNAs were synthesized from the amplified double strand DNA templates using the MEGA-shortscript Kit (AM1354, Invitrogen) and purified using a MEGAclear Transcription Clean-Up Kit (AM1908, Invitrogen). Cas9 protein (20 µM, M0646M, New England Biolabs) and gRNAs (120 ng/µl) were co-injected into one-cell-stage wild-type (WT) embryos. The injected embryos were raised to adult and outcrossed with WT for mutation screening.

The BAC transgenic line TgBAC(spi2:P2a-GFP) was generated as previously described (Suster et al., 2011; Wu et al., 2020). The BAC clone CH211-45O10 was purchased from CHORI-211 library (BACPAC Resources Center). The modified spi2:P2a-GFP BAC constructs were extracted from the DH10B using the Nucleobond BAC 100 Kit (740579, Macherey-Nagel) and injected (30 ng/µl) together with transposase mRNA (50 ng/µl) into one-cell-stage WT embryos. The injected embryos were raised to adults and outcrossed with WT for transgene screening.

WISH was performed as previously described (Tian et al., 2017). AP-conjugated anti-digoxigenin antibody (11093274910, Roche, 1:2000 diluted in blocking buffer) and BM purple (11442074001, Roche) were used as the antibody and substrate for color reaction. dFISH assay was performed as previously described (Lauter et al., 2011). POD-conjugated anti-fluorescein antibody (11426346910, Roche, 1:2000 diluted in blocking buffer), POD-conjugated anti-digoxigenin antibody (11207733910, Roche, 1:2000 diluted in blocking buffer) and TSA-Cy3 or TSA-FITC (NEL741001KT, PerkinElmer) were used for color reaction. The RNA probes were in vitro synthesized with T7 RNA polymerase (Promega) with dig-RNA-labeling mix (11277073910, Roche) or fluorescein-RNA-labeling mix (11685619910, Roche). The probes sequences are listed below: runx1 probe (NM_131603.3; 385-2069); spi2 probe (XM_009300320.3; 182-841); hbae1 probe (NM_182940.3; 1-504); rag1 probe (NM_131389.1; 1430-2949); c-myb probe (NM_001309822.1; 192-3013). WISH images were captured by Stereo Discovery V20 microscope (Zeiss) with digital camera (DS-Ri2, Nikon). dFISH images were captured by SP8 confocal microscope (Leica) and analyzed by ImageJ (National Institutes of Health).

For cell proliferation detection, the PH3 staining was performed after FISH. Anti-PH3 antibody (sc-56739, Santa Cruz Biotechnology, 1:500) and Alexa 555-anti-mouse antibody (A-31570, Invitrogen, 1:400) were used as primary and secondary antibody, respectively. For cell apoptosis detection, TUNEL assay was performed. In brief, embryos were fixed with 4% paraformaldehyde (PFA) overnight at 4°C and dehydrated with methanol at −20°C. Then, the embryos were rehydrated with phosphate-buffered saline plus 0.1% Tween-20 detergent (PBST) for 20 min. After rehydration, the embryos were permeabilized with 25 µg/ml proteinase K for 15 min at 21°C and incubated with 45 µl labeling solution plus 5 µl enzyme solution (In Situ Cell Death Detection Kit, Fluorescein, Roche) at 37°C for 2 h. The embryos were washed three times with phosphate-buffered saline plus 0.1% Tween-20 detergent (PBST) for 5 min each and the images were examined by SP8 confocal microscope (Leica).

For embryonic trunk tissues dissection, embryos at 28 hpf stages were pooled together (each biological replicate with 60∼100 embryos). The trunk tissues of embryos were dissected with a 27G needle and then homogenized by syringe with a 27G needle. The cell suspension was prepared as previously described (Wu et al., 2020) and subjected to FACS analysis (FACS Aria IIIu). For KM dissection in adulthood, the KM of adult fish was dissected and homogenized by grinding on a 40 μm nylon mesh in 5% fetal bovine serum (FBS) in phosphate-buffered saline (PBS) solution. The cell suspensions were then subjected to flow cytometry analysis. The cDNA was prepared from sorted cells (100∼500 cells per well; two or four technical replicates per sample) and amplified by Smart-seq2 protocol (Picelli et al., 2014). Real-time (RT)-qPCR was performed to quantify the spi2 transcripts. qPCR primers used in this study are: spi2 (fwd: 5′-ACATCATCCAGTGGCTCTGT-3′; rev: 5′-GGCAGTGACTTGACACGCT-3′); ef1a (fwd: 5′-TACTTCTCAGGCTGACTGTG-3; rev: 5′-ATCTTCTTGATGTATGCGCT-3′).

Zebrafish at 15 dpf were anesthetized on ice and fixed in 4% PFA at 4°C for at least 1 day. After washing with PBS, fixed samples were dehydrated with methanol at −20°C. After rehydration, the fish were permeabilized with 25 µg/ml proteinase K for 20 min at 21°C and blocked with blocking buffer (5% FBS in PBST) for 1 h. The anti-DsRed primary antibody (632496, Clontech) was diluted (1:400) in blocking buffer, and Alexa Fluor 555 anti-rabbit antibody (A31572, Invitrogen, 1:400 diluted in blocking buffer) was used as secondary antibody. After staining, samples were washed with PBST and imaged using a Leica SP8 confocal microscope. For MGG staining, the dissected KM was homogenized by grinding on a 40 μm nylon mesh in 5% FBS-PBS solution. The cell suspensions were smeared onto the slides by cytospin (Shandon, Thermo Fisher Scientific) and the slides were then air dried and subjected to MGG stain (Sigma-Aldrich), according to the manufacturer's instructions.

To generate the PCS2-spi2-P2a-GFP construct, spi2 full-length coding sequence was amplified from 2 dpf stage zebrafish embryo cDNA using primers (5′-3′): spi2-F, ATGACGGATCTCTCTGTCACAT; spi2-R, CGCTGCTTTCTTTGCACCACGCT. The P2a-GFP sequence was amplified from previous flk1-runx1-P2a-GFP construct (Zhao et al., 2022). The two types of PCR products were assembled into a PCS2+ vector, the sequence of which can be obtained from the Addgene website (https://www.addgene.org/vector-database/2295/). To generate the hsp70l-runx1-P2a-GFP construct, runx1-P2a-GFP sequence was amplified from flk1-runx1-P2a-GFP construct and cloned under hsp70l promoter (Halloran et al., 2000). The plasmids were injected into one-cell-stage embryos. For heat-shock experiments, embryos at 3 dpf were collected in a 50 ml Falcon tube containing 5 ml of E3 embryo medium and submerged in a 40°C water bath for 40 min.

Photoconversion was performed as previously described (Zhao et al., 2022). For time-lapse imaging, the embryos were subsequently mounted in 1% low-melting-point agarose in E3 embryo medium with 0.01% tricaine for anesthetization. Time-lapse imaging was carried out on an SP8 confocal microscope using a 20× objective (Leica). A heated microscope chamber (28.5°C) was used for recording time-lapse videos. Stacks were taken every 15 min with a step size of 3.0 to 5.0 μm. Confocal stacks and time-lapse videos were analyzed using ImageJ.

Embryos were anesthetized on ice and homogenized in Buffer RLT (RNeasy Mini Kit, Qiagen) using needles and syringes. The RNA was extracted using the RNeasy Mini Kit (74104, Qiagen) and reverse-transcribed with Superscript IV VILO Master Mix (11756050, Invitrogen).

All statistical analyses were performed using GraphPad Prism version 8. Paired or unpaired Student's t-tests were used to calculate the two-tailed P-value for comparisons. For multiple comparisons, significances were calculated using two-way analysis of variance (ANOVA), followed by Tukey's multiple comparisons test. Data are presented as mean±standard deviation (s.d.). Statistical significance is shown as follows: n.s., P>0.05; *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001.

We thank Dr Shuting Wu, Ms Kefan Cheng, Ms Thi Huong Trinh and Mr Liang Lou for their help in FACS analysis. We are grateful to Mr Shachuan Feng for the insightful discussion.