blf and the drl cluster synergistically regulate cell fate commitment during zebrafish primitive hematopoiesis

Hematopoiesis in vertebrates, from zebrafish to humans, is an evolutionarily conserved program producing all cellular blood components (Jagannathan-Bogdan and Zon, 2013; Dzierzak and Bigas, 2018; Elsaid et al., 2020). During hematopoiesis, progenitor cells undergo progressive specification and bifurcate into divergent lineages. How blood cell fate divergence is precisely orchestrated has been a topic of much interest because this knowledge is vital to improve our understanding of various blood disorders and to develop new therapeutic strategies.

It is now commonly accepted that hematopoietic cell generation involves bipotent mesodermal precursor cells, the hemangioblasts, which give rise to both hematopoietic and endothelial cells (De Bruijn, 2014). Evidence for the presence of hemangioblasts has been provided by in vivo analyses of both mouse and zebrafish embryos (Huber et al., 2004; Vogeli et al., 2006). Transcription factors, including Tal1, Lmo2, Gata2 and Fli1a, have been suggested to form a recursively wired gene-regulatory circuit to control the emergence of hemogenic endothelium during the early embryonic stages (Pimanda et al., 2007; Lancrin et al., 2009; Carroll and North, 2014; Menegatti et al., 2019). Recent studies have revealed that the initiation of hematopoietic differentiation of hemangioblasts relies on epigenetic regulators, such as Kdm1a (Takeuchi et al., 2015), but it is still largely unknown how presumptive hematopoietic cells shut down endothelial programming to commit to a hematopoietic fate.

In vertebrates, myeloid cells include a diverse repertoire of leukocytes that play essential roles in the immune response, embryogenesis and tissue regeneration (Lavin et al., 2015; Wynn and Vannella, 2016; Watanabe et al., 2019). Based on differences in morphology and biological functions, myeloid cells are classified into two major lineages: granulocytes (neutrophils, basophils and eosinophils), and monocytes/macrophages. In the developing embryo, both lineages are thought to derive from a common myeloid-restricted population termed the neutrophil-macrophage progenitors (NMPs) (Rosenbauer and Tenen, 2007). Recent ontogeny studies have suggested that most tissue-resident macrophages seed the tissues during embryonic development and maintain their pools in the majority of tissues over time (Gomez Perdiguero et al., 2015; Wu and Hirschi, 2020). During zebrafish embryogenesis, myelopoiesis occurs as three distinct waves in a spatially restricted manner. From 12 h post-fertilization (hpf), a population of bipotential myeloid progenitors (marked by spi1b expression) emerges from the anterior lateral plate mesoderm. Under the control of a highly conserved genetic program, these cells give rise to two major lineages: neutrophils and macrophages (Rosenbauer and Tenen, 2007; Le Guyader et al., 2008; Xu et al., 2012). From around 30 hpf, the intermediate hematopoiesis occurring in the posterior blood island (PBI), produces committed erythromyeloid progenitors (EMPs), which are capable of proliferation and contribute to the erythroid and myeloid lineages (Bertrand et al., 2007; Da'as et al., 2012). Finally, during the definitive wave, hematopoietic stem cells (HSCs) arise from the ventral wall of the aorta then migrate to the caudal hematopoietic tissue where they seed and divide giving rise to all hematopoietic lineages, including myeloid cells (Davidson and Zon, 2004). Previous studies have suggested that granulocyte versus monocyte/macrophage fate determination is primarily governed by the interplay between lineage-specific transcription factors/co-factors and epigenetic modifiers (such as Spi1b, C/EBPα, Irf8, Runx1 and Irf2bp2) and the relative abundances of these determinants at specific developmental stages (Dahl et al., 2003; Li et al., 2011; Jin et al., 2012; Wang et al., 2020). In this way, the molecular mechanisms of NMP cell fate choices are beginning to be understood, although many pieces of this puzzle are still missing.

The zebrafish bloody fingers (blf) gene encodes a zinc finger family member with multiple sequential C2H2 zinc fingers spanning almost the entire length of the protein. Although blf is specifically expressed in hematopoietic tissues and blood cells (Sumanas et al., 2005), its function in hematopoiesis has not been explored. It has been noticed that blf displays sequence homology and an overlapping pattern of expression with draculin (drl), draculin-like 1 (drll.1), draculin-like 2 (drll.2) and draculin-like 3 (drll.3), four genes clustered consecutively on chromosome 5 with very high sequence homology (Herbomel et al., 1999; Pimtong et al., 2014) (hereafter referred to as the drl cluster). Multiple adjacent C2H2 motifs are known to confer DNA-binding activity (Wolfe et al., 2000), suggesting a role for blf and drl cluster genes as transcriptional modulators. Knockout of drl in zebrafish led to a reduction in expression of primitive erythrocyte marker genes (Kobayashi et al., 2020); taking into account the functional redundancy that may exist between blf and drl paralogs, a more sophisticated strategy is necessary to achieve a comprehensive understanding of the function of this group of genes in hematopoiesis.

In the current study, we provide in vivo evidence demonstrating that primary erythropoiesis and myelopoiesis are synergistically regulated by blf and drl cluster genes. Simultaneous depletion of blf and the drl cluster led to severe erythrocyte aplasia and excessive macrophage formation at the expense of neutrophil development. Integrative analysis of transcriptome and genome-wide binding data revealed that blf and drl cluster genes are responsible for constraining the expression of vasculogenesis-promoting genes in the intermediate cell mass and monocytopoiesis-promoting genes in the rostral blood island. These results indicate that blf and the drl cluster play important dual roles in cell fate commitment during hematopoiesis: they promote primitive erythropoiesis in hemangioblasts and confer a neutrophil fate on NMPs. Furthermore, we performed a rescue screen and found that Zfp932 is a potential mammalian functional equivalent to zebrafish blf and drl cluster genes. Our data provide insight into the molecular mechanisms directing primitive erythropoiesis and myelopoiesis.

In a Tol2 transposon-mediated gene trapping screen to search for regulators of zebrafish embryogenesis (Hou et al., 2017), we identified the RP2-527 line, in which the RFP reporter shows distinct expression patterns in hematopoietic tissues and blood cells (Fig. S1A). 5′ RACE was used to identify the gene trapped in the RP2-527 line and sequencing results indicated that the gene-trapping element was integrated within the second intron of the blf locus (Fig. S1B). Zebrafish Blf protein contains 15 zinc-finger domains and the insertion resulted in a transcript that encodes a fusion protein containing the first N-terminal 15 amino acids of Blf. Because this fusion protein lacked all zinc-finger domains, the allele that we identified from the RP2-527 line should act as a true null allele. Although previous studies showed that knockdown of blf in zebrafish embryos led to defects of morphogenetic movements during neurulation (Sumanas et al., 2005), we did not observe deformity in blf^−/− embryos or significant effects on overall survival and fecundity of blf^−/− fish (Fig. S1D). It has been noted that blf and drl cluster genes may play redundant roles in development because of their high sequence homology and overlapping expression pattern (Sumanas et al., 2005). In order to elucidate the function of this group of genes in development, we used the CRISPR/Cas9 strategy to delete simultaneously four drl cluster genes. Multiple sgRNAs targeting 5′ and 3′ ends of the drl cluster were designed (Fig. S2A) and injected into one-cell-stage zebrafish embryos with Cas9 protein. Among the alleles obtained through this procedure, one bearing a 49,188-bp deletion in the drl cluster was chosen to confirm the absence of expression of drl, drll.1, drll.2 and drll.3 (Fig. S2A,B). This allele was assigned as ‘drl 4KO’ and used in the subsequent analysis. Embryos from multiple drl 4KO heterozygous incrosses were collected and no evident developmental defect was observed in drl 4KO homozygous animals (Fig. S2C).

In order to determine whether blf and the drl cluster synergistically regulate hematopoiesis, blf^−/−; drl 4KO embryos were generated and examined. We first tested whether depletion of blf and the drl cluster affects primitive erythropoiesis. Whole-mount RNA in situ hybridization (WISH) and qPCR revealed that the mature erythrocyte marker hbae1 was slightly downregulated in drl 4KO embryos and almost absent in blf^−/−; drl 4KO samples (Fig. 1A,B). Furthermore, in wild-type, blf^−/− and drl 4KO embryos, hemoglobin-positive cells were found after circulation commenced, but o-dianisidine staining signal was absent in blf^−/−; drl 4KO embryos (Fig. 1A). These results indicated that blf and drl cluster genes, likely acting in a redundant fashion, regulate primitive erythropoiesis.

Because the expression analysis detected blf and drl cluster genes in hemogenic endothelial cells (Fig. S1A), we next explored the roles of these genes in definitive erythropoiesis. In situ hybridization showed that expression of the HSC marker genes runx1 and c-myb (myb) was abolished in blf^−/−; drl 4KO samples but not in blf^−/− or drl 4KO embryos (Fig. 1C). At 7 days post-fertilization (dpf), in contrast to robust o-dianisidine staining of wild-type, blf^−/− and drl 4KO larvae, only sparsely distributed erythrocytes could be observed in blf^−/−; drl 4KO samples (Fig. 1C).

Owing to severe erythrocyte aplasia, blf^−/−; drl 4KO fish exhibited substantial mortality during juvenile stages (Fig. 1D). Approximately 30% blf^−/−; drl 4KO homozygous mutants could be raised to adulthood; we applied microscopic observation and hematological analysis on these animals. The kidneys from blf^−/−; drl 4KO fish were moderately hypoplastic compared with those of controls, and Hematoxylin & Eosin staining revealed that the ‘escapers’ have an abnormal accumulation of hematopoietic cells between the renal tubules (Fig. S3A,B). This observation was corroborated by differential cell counting, which showed that precursor fractions were markedly increased in blf^−/−; drl 4KO zebrafish compared with controls (Fig. S3C). The peripheral blood from ‘escapers’ also displayed decreased red blood cell counts, although no appreciable defects of red blood cells were observed (Fig. 1E, Fig. S3D). Taken together, these results indicate that blf and drl cluster genes are required for normal development of hematopoietic stem cells.

Vertebrate primitive hematopoietic and vascular systems are derived from common precursors, the hemangioblasts (Orkin and Zon, 2008). The expression of blf and drl cluster genes in the lateral plate mesoderm suggests that the generation of hemangioblasts may be impaired by mutation of these genes. We therefore analyzed expression profiles of three hemangioblast markers, etsrp, lmo2 and npas4l, but no differences were observed between wild-type and blf^−/−; drl 4KO embryos (Fig. 2).

It is known that definitive hematopoiesis is associated with arteries (De Bruijn et al., 2002). To establish whether arterial development was affected in blf^−/−; drl 4KO embryos, we determined the expression of arterial marker genes and examined vascular development with alkaline phosphatase staining and the Tg(kdrl:EGFP)^s843 line. No obvious difference in these assays was observed between blf^−/−; drl 4KO and control embryos (Fig. S4). This result implies that the blf and drl cluster genes are dispensable for hemangioblast formation and vasculature development.

In zebrafish, primitive erythropoiesis takes place in the intermediate cell mass region and a sophisticated network of transcription factors and epigenetic regulators has been described as regulating this process (Wells and Steiner, 2022). To assess the state of the hematopoietic transcription network in blf^−/−; drl 4KO embryos, the expression of gata1, trim33 and tal1 was analyzed and demonstrated to be greatly reduced (Fig. 3A). By utilizing RFP expression from the blf gene-trapping allele as lineage tracer, we examined erythroid progenitors in live embryos. Confocal microscopy images showed that primitive erythroid progenitors displayed a uniformly round shape in control embryos. Conversely, erythroid progenitors in blf^−/−; drl 4KO embryos exhibited irregular shape and size, and RFP-positive puncta could be observed in these cells (Fig. 3B,C). These observations prompted us to examine the viability of erythroid progenitors in blf^−/−; drl 4KO embryos. TUNEL staining revealed that differentiating erythroid cells in both the intermediate cell mass and the erythro-myeloid progenitor population undergo enhanced programmed cell death in blf^−/−; drl 4KO embryos (Fig. 3D,E). All of these results indicate that blf and drl cluster genes are essential for the differentiation and survival of primitive erythrocytes.

Gene expression analysis showed that both blf and drl cluster genes were present in myelopoietic tissues, including the rostral blood island (Fig. S1A; Pimtong et al., 2014); this prompted us to study the role of these genes in myelopoiesis. We first assessed early myeloid cell specification using WISH and qPCR for spi1b. At the 10-somite stage (14 hpf), there was no notable difference in the expression of spi1b in blf^−/−, drl 4KO or blf^−/−; drl 4KO mutants compared with the wild type (Fig. S5A). To assess the effect of loss of blf and drl cluster genes on primitive granulopoiesis and monopoiesis, we performed WISH for myeloid marker genes. The data showed that in blf^−/−; drl 4KO mutants the number of differentiating neutrophils (marked by mpx and lyz expression) was greatly decreased, whereas the number of macrophages (distinguished by mpeg1 and mfap4.1 expression) was mildly increased (Fig. 4A). To investigate the possible mechanism, we examined cell numbers of the entire myeloid population. The overall number of cells positive for coro1a, a pan-myeloid marker identifying all myeloid subsets (Li et al., 2012), was comparable between blf^−/−; drl 4KO mutants and control embryos (Fig. 4A). To investigate further the role of blf and drl cluster genes in the fate decision of primitive myeloid progenitors, we examined the proportion of progenitors taking the path of granulopoiesis or monocytopoiesis. To do this, hematopoietic cells in the rostral blood island were labeled with uncaging photoactivatable fluorescein (Flu) at 20 hpf (Fig. S6), and the contribution to neutrophils or macrophages was determined by double staining against Flu and mRNA of marker genes at 72 hpf. The data showed that in control embryos the myeloid progenitors from the rostral blood island have a similar probability to differentiate into neutrophils or macrophages. By contrast, in blf^−/−; drl 4KO embryos the vast majority of myeloid progenitors give rise to macrophages and only very small fraction of them adopt the neutrophil fate (Fig. 4B). This supports the notion that imbalanced myeloid lineage development could be caused by expansion of the macrophage population at the expense of neutrophil development.

To evaluate the roles of blf and drl cluster genes in intermediate myelopoiesis, we examined the expression of gata1 and lmo2 at 30 hpf. Loss of blf and drl cluster genes abolished the expression of these two genes in the PBI, indicating that the formation of EMPs was severely defective (Fig. S5B). In agreement with this result, in the PBI of blf^−/−; drl 4KO mutants the number of cells positive for mpx or mpeg1 expression was also profoundly lower and the decreased neutrophil population was further confirmed by Sudan Black staining (Fig. S7). In further analysis of the definitive wave of myelopoiesis, WISH results showed that at 96 hpf the number of both mpx-positive neutrophils and mpeg1-positive macrophages in blf^−/−; drl 4KO mutants was lower than that in controls (Fig. S8).

Collectively, these data showed that the loss of blf and drl cluster genes leads to aberrant neutrophil versus macrophage fate decision during primitive hematopoiesis and drastic defects in intermediate and definitive myelopoiesis.

To confirm that the blf and drl cluster genes act in a redundant fashion, we tested the efficacy of blf and four drl cluster genes in rescuing defective hematopoiesis in blf^−/−; drl 4KO embryos. By microinjection, mRNA of individual genes was introduced back into blf^−/−; drl 4KO embryos and primary erythropoiesis and myelopoiesis were analyzed with qPCR, WISH and o-dianisidine staining. The results showed that all five genes could partially restore both erythropoiesis and granulopoiesis in blf^−/−; drl 4KO embryos, with drl displaying the highest potency (Fig. S9).

blf and drl cluster genes encode transcription factors with multiple sequential C2H2 zinc fingers, and we performed integrative analysis on their transcriptome and genome-wide binding profile to investigate how these proteins modulate the transcriptional program in hematopoietic progenitors. In order to determine how loss of blf and drl cluster genes changes the transcriptome landscape, hematopoietic progenitors were collected by fluorescence-activated cell sorting (FACS) by utilizing RFP expression from the blf gene-trapping allele as lineage tracer, and low input RNA sequencing (RNA-seq) was carried out (Fig. S10). In hematopoietic progenitors from blf^−/−; drl 4KO embryos, a total of 1516 genes, consisting of 375 with decreased and 1141 with increased expression, were significantly altered [log₂ (fold change)>2, false discovery rate (FDR)<0.001] (Table S1). The majority of the aberrantly expressed genes were upregulated rather than downregulated in the blf^−/−; drl 4KO samples, suggesting that blf and drl cluster genes may primarily act as transcription repressors (Fig. 5A). Gene ontology (GO) enrichment analysis revealed that genes deregulated in blf^−/−; drl 4KO samples were involved in erythrocyte and myeloid cell differentiation, vasculature development, Notch signaling, cell migration and metabolism of extracellular matrix (Fig. 5B). These findings were further confirmed by gene set enrichment analysis (GSEA), an independent, unbiased approach analyzing enrichment of functional groups, and a similar set of pathways was detected (Fig. 5C, Fig. S11). Moreover, qPCR of representative genes from enriched gene sets was applied to support the reliability of the expression data obtained by low-input RNA-seq (Fig. 5D).

To investigate the blf and drl cluster regulatory network and associated mechanisms of action, we used cleavage under targets and tagmentation (CUT&Tag) technology to determine their genome-wide occupancy in hematopoietic progenitors. To recapitulate the physiological binding profile, mRNA coding Blf-3xFLAG or Drl-3xFLAG was injected into blf^−/−; drl 4KO embryos and hematopoietic progenitors were purified using RFP as a lineage tracer at 20 hpf. To reduce the number of embryos required, we used the CUT&Tag procedure, which has proven highly efficient for globally identifying binding sites in low-input samples (Kaya-Okur et al., 2020; Li et al., 2021). Following next-generation sequencing, analysis of the CUT&Tag data resulted in the identification of 10,267 and 6394 high-confidence binding sites for Blf and Drl, respectively (Table S2), which provided an estimate of the number of genomic positions occupied by Blf or Drl in hematopoietic progenitors at this stage of development. The peaks were enriched in gene promoter regions (demarcated here as ±3 kb from the TSS; Fig. 6A,B), but a significant portion of Blf and Drl binding sites also localized to distal intergenic regions, suggesting that they direct transcription regulation at both promoters and enhancers (Fig. 6B). There was extensive overlap (63.9%) between Blf and Drl chromatin occupancy, as well as substantial overlap (65.7%) between their associated genes (Fig. 6C), indicating that they are involved in an interwoven regulatory network. DNA-binding motifs for other hematopoietic transcription regulators, including CEBP, GATA factors and KLF factors, were enriched on the overlap peaks (Fig. S12), implying cooperation between Blf/Drl genes and these hematopoietic factors. To determine enriched DNA motifs at Drl occupancy sites, de novo motif analysis was performed with the HOMER tool and a set of potential Drl binding motifs was identified (Fig. 6D). To identify the direct targets and mode of action of Drl, we integrated the blf^−/−; drl 4KO RNA-seq and Drl CUT&Tag results with the binding and expression target analysis (BETA) pipeline (Wang et al., 2013). This showed that 1547 genes (Table S3) were directly regulated by Drl; the genes repressed by Drl showed higher regulatory potential scores than those that were activated, indicating that Drl mainly acts as a transcriptional repressor (Fig. 6E). GO analysis showed that genes repressed by Drl were enriched in processes such as vasculature development and myeloid leukocyte differentiation (Fig. 6F). CUT&Tag detected Drl situated at transcription regulatory regions of vasculogenesis-promoting genes, such as etsrp, fli1rs and aplnra as well as of monocytopoiesis-promoting genes, such as irf8, irf5 and bcl6ab (Fig. 6G,H, Fig. S13). WISH results showed ectopic expression of etsrp, an essential regulator of vasculogenesis, in the intermediate cell mass, and expression of irf8, a master driver of monocytopoiesis, was profoundly upregulated in the rostral blood island in blf^−/−; drl 4KO embryos (Figs 5D, 6G,H). To verify the transcription regulatory role of Drl on these genes, we overexpressed Drl in wild-type embryos by mRNA injection, and found a significant decrease in irf8 and etsrp expression (Fig. 6G,H). To examine whether the mis-regulated irf8 expression underpinned skewed myeloid lineage development, we employed knockdown of irf8 by morpholino injection in blf^−/−; drl 4KO embryos and observed partially restored expression of the neutrophil marker gene mpx (Fig. 6I).

In summary, these results indicate that blf and drl cluster genes play dual roles in establishing normal transcription programs in different hematopoietic progenitors. In the intermediate cell mass, blf and drl cluster genes support the hematopoietic fate by constraining expression of vasculogenesis-promoting genes; in the rostral blood island, these genes sustain granulopoiesis by confining monocytopoiesis-promoting genes.

Given the conservation of the hematopoietic regulatory program in different vertebrates (Jagannathan-Bogdan and Zon, 2013), we speculated that mammalian functional ortholog(s) exist. We first deployed synteny analysis on both blf and drl cluster loci, but low synteny conservation (Fig. S14) prevented us from inferring orthologs by gene neighborhood. Multiple sequential C2H2 zinc fingers in the Blf and Drl cluster proteins presented a challenge in identifying orthologous genes with routine sequence similarity-based methods such as BLASTp. Therefore, we adopted a more sensitive, hidden Markov model (HMM)-based search method, HHpred, which is able to establish connections in remotely homologous, characterized proteins (Gabler et al., 2020). By conducting an iterative search with the Drl protein sequence against the human and mouse proteomes, a list of ortholog candidates was compiled. To screen the mammalian functional orthologs of zebrafish blf and drl cluster genes, mRNAs of the top 25 candidate genes on this list (Table S4), ranked by secondary structure scores, were prepared and microinjected into blf^−/−; drl 4KO embryos. Rescue efficacy was measured by mRNA levels of hbae1 and mpx. Among all the candidate genes tested, Zfp932 was able to partially restore erythropoiesis and granulopoiesis in blf^−/−; drl 4KO embryos (Fig. 7, Fig. S15), which indicated that Zfp932 is a potential mammalian functional ortholog to the zebrafish blf and drl cluster genes.

Lineage commitment of hematopoietic progenitor cells is governed by intricate gene regulatory networks. In the present study, we have presented in vivo evidence demonstrating that blf and drl cluster genes are important components of these programs. Our data revealed that blf and drl cluster genes act as crucial determinants of primitive erythropoiesis and myelopoiesis by constraining the expression of vasculogenesis-promoting and monocytopoiesis-promoting genes in two progenitor populations. The identification of these master regulators of hematopoietic fate enhances our understanding on this topic and may lead to improved strategies for the generation of hematopoietic cells in vivo and in vitro for regenerative medicine.

Besides defects of primitive hematopoiesis, our study shows that simultaneous loss of blf and drl cluster genes leads to disruption of HSC development and premature death at a juvenile stage. These results indicated that blf and drl cluster genes are also involved in definitive hematopoiesis. It is possible that blf and drl cluster genes act by a similar mechanism in definitive hematopoiesis: repressing endothelial genes to facilitate formation of HSCs. Further investigation is required to explore this possibility. We also observed that approximately 30% of blf^−/−; drl 4KO homozygous mutants were able to survive to adulthood and that these animals displayed mild anemia and nearly normal red blood cell differentiation. It would be interesting to examine how hematopoietic stem cells emerge in these ‘escapers’ and whether myeloid lineage cells form normally.

Intriguingly, our data indicated that blf and drl cluster genes play multifunctional roles in hematopoiesis by binding to gene regulatory regions of key endothelial and hematopoietic factors. These different roles may require a variety of partners and future studies will be needed to identify specific partners in different cell populations. Given that long-tandem, zinc-finger arrays may be functionally broken into several regions of fewer zinc-finger repeats (Emerson and Thomas, 2009), it is possible that blf and drl cluster factors bind a variety of motifs as a result of interacting with distinct partners.

In a previous study, Pimtong et al. reported that knockdown of the expression of drll.3 with morpholino injection could cause decreased primitive erythropoiesis and myelopoiesis (Pimtong et al., 2014). We also observed a mild erythropoiesis defect in drl 4KO embryos (Fig. 1A,B), indicating that drl cluster genes probably act in a dosage-dependent fashion. During the process of generating the drl 4KO allele, we also obtained two CRISPR alleles bearing mutants on drll.3; no overt hematopoietic defects were observed in the homozygous embryos of these alleles.

By means of rescue screening, we identified Zfp932 as a potential mammalian functional equivalent to blf and drl cluster genes. This suggested that the mechanisms of hematopoietic cell fate commitment, discussed in the current study, may be conserved during evolution. Interestingly, repeated sequence motif scanning with RADAR (Heger and Holm, 2000) identified multiple CDQCGK motifs on Drl, Blf and Zfp932 sequences (Fig. S16). This suggested that they represent a subgroup of zinc-finger proteins and it will be of great interest to characterize the biochemical properties and functions of this motif.

Previous reports demonstrated that etsrp is a potent positive regulator of endothelial development in zebrafish (Sumanas and Lin, 2006; Liu and Patient, 2008), but ectopic expression of etsrp in the intermediate cell mass did not induce excess endothelialization in blf^−/−; drl 4KO homozygous embryos (Fig. 6G). One possible explanation for this observation is that the differentiation of hemangioblasts to hematopoietic cells is suspended, and that immature hemangioblasts possessing endothelial characteristics went into apoptosis.

All animal experiments were carried out in accordance with the approved guidelines of the Institutional Animal Care and Use Committee of the Nanjing University. All zebrafish lines were kept on the AB background.

Transcription of digoxigenin-labeled antisense RNA probes was performed using standard methods. WISH was carried out as previously described (Thisse and Thisse, 2008).

gRNAs were designed using a CRISPR/Cas9 target online predictor (https://cctop.cos.uni-heidelberg.de), and 50 pg of sgRNAs and 500 pg of Cas9 protein were co-injected into one-cell-stage embryos. Live embryo genotyping was carried out as previously described (Zhang et al., 2020) and one allele bearing a 49,188 bp deletion on the drl cluster was identified and used for subsequent experiments. Genotyping oligos used were: drl 4KO allele: 5′-GAAACTCAGAAAGGTTTGAAACAACCC-3′ and 5′-CTGAGTTCAACAATCGAAGCATATTAAAAACCAG-3′; drl cluster WT allele#1: 5′-GGACCGAGTATCAGTAGTATGCA-3′ and 5′-GTTAGCTGGTACCCACTTCTC-3′; drl cluster WT allele#2: 5′-GGTGTCTGAGGATGCACGAC-3′ and 5′-AAAGCAGAGAATTGTTAACCGGG-3′.

Based on a previously described protocol (Paffett-Lugassy and Zon, 2005), o-dianisidine staining was used to detect the presence of hemoglobin. Embryos were stained in the dark for 30 min at room temperature with a solution containing o-dianisidine (0.6 mg/ml), 0.01 M sodium acetate (pH 4.5), 0.65% H₂O₂ and 40% (vol/vol) ethanol. Once stained, embryos were washed with RO water and then fixed in 4% paraformaldehyde for at least 1 h at room temperature. Pigments were removed from fixed embryos by incubating in a solution of 0.8% KOH, 0.9% H₂O₂ and 0.1% Tween-20 for 30 min at room temperature. Based on previously described protocols (Le Guyader et al., 2008), Sudan Black (SB; Sigma-Aldrich) solution was used to treat the paraformaldehyde-fixed embryos for 20 min, washed extensively in 70% ethanol in water, then progressively rehydrated in PBS and 0.1% Tween-20 (PBT).

Total RNA was prepared using TRIzol (Invitrogen, 15596) and Direct-zol RNA Miniprep (Zymo Research, R2052). cDNA was synthesized with the PrimerScript RT kit (Takara Bio, RR047A). RT-qPCR reactions were performed on the Roche LightCycler system using the SYBR Green Master Mix (Takara Bio, RR420A). Melt curves were examined to ensure primer specificity. Primers used in RT-qPCR were designed to span exon–exon junctions and are listed in Table S5.

For TUNEL reactions, staining with anti-digoxigenin-fluorescein was performed according to the manual of In Situ Cell Death Detection Kit (Sigma-Aldrich, 11684795910). The samples were fixed for 4 h in 4% paraformaldehyde, then washed twice, 10 min each wash, in PBS. After washing in PBS, DNA breaks were elongated with terminal transferase and digoxigenin-dUTP solution.

Whole-mount imaging was performed using a Leica DFC320 camera on a Leica M205FA stereomicroscope. All confocal images were acquired using a Zeiss LSM880 confocal microscope.

At 7 dpf, 80 wild-type siblings and blf^−/−; drl 4KO animals were each put into a 3 l tank. Every 3 days, the number of living fish was counted until 60 dpf. A Kaplan–Meier curve was generated using Prism 7 (GraphPad) (Goel et al., 2010).

Lineage tracking was performed as described previously (Jin et al., 2007). Briefly, CMNB-caged-fluorescein (Invitrogen, F7103) was injected into one-cell-stage embryos then left to develop in the dark. At the 22-somite stage, cells at the anterior hematopoietic site were uncaged by focusing and confining 405-nm laser excitation. Successful activation of fluorescein was verified by local enhancement of green fluorescence. Uncaged embryos were incubated in the dark until fixation at 72 hpf for further analysis.

Two-color fluorescence staining was performed essentially as described previously (Jin et al., 2006). Briefly, embryos were first hybridized with a digoxigenin-labeled antisense RNA probe then incubated overnight with anti-fluorescein-POD, Fab fragments (Roche, 11426346910) and detected with Alexa Fluor 488 tyramide substrate (Invitrogen, B40953). The color reaction was then stopped by sequential washing with methanol/PBS Tween-20 (PBST) and 1% H₂O₂/methanol. Finally, the embryos were subjected to a second color staining with anti-digoxigenin-POD, Fab fragments (Roche, 11207733910) and Alexa Fluor 555 tyramide (Invitrogen, B40955) as a substrate.

The libraries were constructed with the NEBNext Single Cell/Low Input RNA Library Prep Kit for Illumina (NEB, 6420) and sequenced on the Illumina NovaSeq platform. HISAT2 V2.1.0 was used to map the sample sequencing reads to the GRCz11 reference genome. Gene expression counts were calculated using FeatureCounts v1.6.0. based on current Ensemble annotation. The R package DESeq2 was then employed to make differential gene expression calls.

CUT&Tag was performed in two biological replicates as originally described (Kaya-Okur et al., 2020) with the Hyperactive Universal CUT&Tag Assay Kit for Illumina (Vazyme, TD-903). Briefly, mRNA was injected into embryos at the one-cell stage, and RFP⁺ hematopoietic progenitors were collected by FACS at 20 hpf then immobilized on Concanavalin A-coated paramagnetic beads and permeabilized with digitonin. Cells were incubated overnight with the anti-FLAG antibody (Sigma-Aldrich, F3165). To increase the number of Protein A/G binding sites for each bound antibody, goat anti-mouse IgG H&L antibody (Abcam, ab6708) was incubated at room temperature for 1 h. Tethered cells were washed thoroughly to remove unbound antibody, and then incubated with pAG-Tn5 for 60 min. Tn5 transposase-mediated tagmentation was then initiated by the addition of MgCl₂. Tagmented DNA was extracted using magnetic beads then sequencing libraries were generated as previously described (Kaya-Okur et al., 2020), and sequenced on an Illumina HiSeq 2500 platform. Drl- and Blf-binding regions were identified by model-based analysis of ChIP-seq (MACS) (Zhang et al., 2008) with the parameter ‘macs2 callpeak -t C&T_flag.bam -c Input.bam -n C&T_flag_peaks -g 1.37e9 -f BAMPE -q 0.01’. The peaks were annotated using ChIPseeker (Yu et al., 2015). Each peak set was obtained by the intersection of two independent biological replicates.

irf8 splicing morpholino oligonucleotides (MO)^sp (5′-AATGTTTCGCTTACTTTGAAAATGG-3′) were synthesized based on a previous report (Li et al., 2011). Standard control morpholino was purchased from Gene Tools. One-cell-stage embryos were injected with 2 nl of morpholino solution at a concentration of 0.6 mM irf8 MO^sp.

Two-sided, unpaired Student’s t-tests were performed using GraphPad Prism 7 software and numerical data are presented as mean±s.e.m. Differences were considered significant if the probability value was P<0.05 and highly significant if the probability value was P<0.01. All experiments were carried out with at least three biological replicates. The numbers of animals used are described in the corresponding figure legends.

We thank all members of the Lou lab for valuable insight and suggestions. We also thank Dr Min Li from core facilities of Nanjing University for technical support.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200919.reviewer-comments.pdf