The zebrafish ETS transcription factor Fli1b functions upstream of Scl/Tal1 during embryonic hematopoiesis

The earliest vascular endothelial and hematopoietic cell lineages in different vertebrates are thought to originate from a common precursor, the hemangioblast (Lacaud and Kouskoff, 2017). In mammalian embryos, the putative hemangioblasts emerge in the yolk blood islands and give rise to primitive erythrocytes, megakaryocytes and macrophages, as well as vascular endothelial cells (Biben et al., 2023). To date, the molecular mechanisms that regulate hemangioblast differentiation into the earliest hematopoietic and endothelial lineages are still incompletely understood.

Although it is difficult to study the earliest steps in blood and vascular differentiation in mammalian embryos, zebrafish has emerged as an advantageous embryonic model. The molecular mechanisms regulating hematovascular development are highly conserved among zebrafish and mammalian embryos. Blood and vascular progenitors in zebrafish embryos emerge during early somitogenesis stages from the lateral plate mesoderm (LPM), which is considered to be equivalent to mammalian blood islands. The anterior LPM (ALPM) gives rise to cranial vasculature and the primitive myeloid lineages, which include macrophages and neutrophils, whereas the posterior LPM (PLPM) gives rise to erythroid cells as well as trunk and tail vasculature (Davidson and Zon, 2004; Warga et al., 2009). Cell lineage-tracing studies have indicated that many (but not all) endothelial and hematopoietic cells in zebrafish share a common progenitor, the hemangioblast (Vogeli et al., 2006). However, these lineages separate very early and, to the best of our knowledge, cells that co-express vascular endothelial and hematopoietic markers have not been observed in vivo during later stages of embryogenesis, with the exception of the hemogenic endothelium that gives rise to hematopoietic stem cells (HSCs) (Bertrand et al., 2010; Kissa and Herbomel, 2010).

Several ETS transcription factors have been implicated in regulating distinct steps during hematopoietic and endothelial differentiation. The ETS transcription factor friend leukemia integration 1 (Fli1) functions near the top of the transcriptional cascade and is involved in regulating both vascular endothelial and hematopoietic differentiation (Ben-David et al., 2022). Fli1-deficient murine embryos die due to hemorrhages around embryonic day (E) 11.5. They show deficient vascular endothelial differentiation as well as defects in hematopoiesis, mainly in the block of megakaryocyte differentiation (Abedin et al., 2014; Hart et al., 2000; Spyropoulos et al., 2000). However, the role of Fli1 in the formation of primitive erythroid and myeloid lineages is less clear. Erythrocytes appear to be present at E8.5 and E9.5 in Fli1 mutant embryos. Although greatly reduced fetal liver hematopoiesis has been reported, this could be a consequence of hemorrhages and HSCs failing to repopulate the fetal liver (Spyropoulos et al., 2000). Overexpression of Fli1 in EWS-Fli1 fusion protein inhibits erythroid differentiation (Pereira et al., 1999; Tamir et al., 1999). Low levels of Fli1 expression in myelo-erythroid progenitors are thought to promote erythropoiesis, whereas high levels promote megakaryocyte differentiation (Ben-David et al., 2022). In contrast, inhibition of Fli1 function in frog (Xenopus laevis) embryos using morpholinos (MOs) results in the inhibition of hemangioblast differentiation and the downregulation of primitive and definitive hematopoiesis (Liu et al., 2008). However, these results have not been validated in genetic mutants.

Due to genomic duplication, there are two homologs of mammalian Fli1 present in zebrafish, fli1a (also known as fli1) and fli1b (also known as fli1rs), both of which are expressed in vascular endothelial cells (Brown et al., 2000; Craig et al., 2015). In addition, fli1a is expressed in hematopoietic progenitors (Brown et al., 2000). Overexpression of Fli1a and transactivating domain fusion protein was found to cause expansion of hematopoietic and vascular markers in zebrafish embryos (Liu et al., 2008). MO knockdown of Fli1a together with Erg or other ETS transcription factors caused defects in angiogenesis (Liu and Patient, 2008; Pham et al., 2007). However, no hematopoietic defects have been previously reported in Fli1a-deficient embryos, and previous MO knockdown results have not been validated using genetic mutants.

We have previously reported characterization of the zebrafish fli1b mutant, which was isolated in a transposon-mediated insertional mutagenesis screen (Craig et al., 2015). fli1b homozygous mutants were viable and did not show any obvious defects in vascular development. However, they exhibited a partial redundancy with another ETS transcription factor, Etv2 (also known as Etsrp), which functions as a master regulator of vasculogenesis (Sumanas and Lin, 2006). etv2 mutants showed loss of early vascular differentiation, yet vasculogenesis partially recovered during the 24-72 h post fertilization (hpf) stages (Craig et al., 2015). This recovery was completely absent in etv2; fli1b double-deficient embryos, illustrating a partial redundancy between these two factors (Craig et al., 2015). However, the role of fli1b in hematopoiesis has not been previously investigated.

In this study, we analyzed hematopoietic defects in fli1b mutant embryos. We show that primitive erythropoiesis and HSC formation were greatly reduced in fli1b mutant embryos. In addition, formation of primitive myeloid progenitors – macrophages and neutrophils – was significantly diminished in fli1b mutants. In contrast, fli1b mutant embryos showed expansion of myeloid cells derived from an alternative hematopoietic site, the endocardium. Intriguingly, these cells co-expressed both myeloid and vascular endothelial markers in fli1b mutants, suggesting that they are present in a hemangioblast-like state. These results identify new roles for fli1b in regulating hematopoietic differentiation and will be important for our understanding of the molecular mechanisms and transcriptional pathways that regulate hemangioblast and vascular differentiation during development and regeneration.

To test whether fli1b is required for primitive hematopoiesis, we analyzed the expression of hematopoietic markers in fli1b^−/− mutant and wild-type (wt) control fli1a:GFP embryos using whole-mount in situ hybridization (WISH). The expression of the erythroid markers gata1a (hereafter gata1) and hbbe3 in wt embryos was present in the intermediate cell mass region above the yolk extension at 22 hpf (Fig. 1A,C). In fli1b^−/− mutants, the expression of gata1 was strongly reduced in 85.5% of embryos, whereas hbbe3 expression was reduced in 60.7% of embryos at 22 hpf (Fig. 1A-D). The transcription factor stem cell leukemia (scl, also known as tal1) is considered a master regulator of early hematopoiesis and functions upstream of gata1 and hbbe3 (Dooley et al., 2005; Juarez et al., 2005; Patterson et al., 2005). scl expression in the intermediate cell mass was reduced in 80.7% of fli1b^−/− embryos at 22 hpf, although the reduction was not as significant as that of gata1 and hbbe3 (Fig. 1E,F). To confirm WISH results, we performed quantitative PCR (qPCR) for gata1, hbbe3 and scl using wt and fli1b^−/− embryos at the 20-somite stage. Expression of both gata1 and hbbe3 was significantly reduced in fli1b^−/− zebrafish embryos, whereas scl expression was not strongly affected (Fig. 1G-I), possibly due to its early expression in both red blood cell (RBC), vascular endothelial cell and neuronal cell populations. As expected, fli1b mRNA expression was greatly reduced in fli1b^−/− embryos based on qPCR analysis (Fig. 1J). We then performed heme staining using o-dianisidine to evaluate the presence of RBCs in fli1b^−/− zebrafish embryos at 2 and 3 days post fertilization (dpf). Heme staining in fli1b^−/− embryos was significantly reduced at both time points compared to wt control fli1a:GFP embryos (Fig. 1K-P). Altogether, these results indicate that primitive erythropoiesis is diminished in fli1b mutant embryos.

To determine whether fli1b mutants also showed defects in definitive hematopoiesis, WISH was performed at 28 hpf for the hematopoietic stem and progenitor cell (HSPC) markers runx1 and cmyb (also known as myb). Expression of both runx1 and cmyb in HSPCs, located along the wall of the dorsal aorta, was substantially reduced in fli1b^−/− embryos compared to that in wt controls (Fig. 1Q-T). These results argue that both primitive and definitive hematopoiesis are deficient in fli1b^−/− embryos.

We have previously demonstrated that fli1b is expressed in vascular endothelial cells (Craig et al., 2015). To clarify whether fli1b is also expressed in hematopoietic cells or their progenitors, we analyzed co-expression of the erythroid progenitor marker gata1 and fli1b in vascular endothelial kdrl:GFP embryos at the 15-somite and 22 hpf stages using the hybridization chain reaction (HCR) (Fig. 2A-F). However, no overlap between gata1 expression in RBCs and fli1b expression was observed. In contrast, kdrl:GFP and fli1b expression overlapped extensively in vascular endothelial cells at 22 hpf (Fig. 2D-F). Notably, fli1b expression only had a partial overlap with kdrl:GFP expression at the 15-somite stage, likely due to strong fli1b expression in vascular endothelial progenitors (angioblasts), which are negative for kdrl:GFP (Fig. 2A-C). In addition, we analyzed fli1b and scl co-expression in live fli1b^+/−; scl:dsRed embryos. Because of the gene-trap insertion in the fli1b locus, GFP expression labels fli1b-positive cells in heterozygous fli1b^+/− embryos, whereas scl:dsRed expression is apparent in blood and endothelial cells. At the 22 hpf stage, there was no apparent overlap between GFP and scl:dsRed expression in blood cells or their progenitors, whereas GFP and scl:dsRed expression overlapped in vascular endothelial cells (Fig. 2G-I). Similarly, GFP and kdrl:mCherry expression overlapped in vascular endothelial cells in fli1b^+/−; kdrl:mCherry embryos at the 22 hpf stage (Fig. 2J-L). Altogether, these results confirm that fli1b expression is restricted to vascular endothelial cells and their progenitors and is largely absent from erythroid cells.

scl is one of the earliest factors expressed in hematopoietic progenitors. It functions upstream of gata1 in inducing erythroid differentiation. In addition, scl is expressed in vascular endothelial progenitors and is required for endothelial differentiation (Dooley et al., 2005; Juarez et al., 2005; Patterson et al., 2005). To test whether scl expression was affected in vascular endothelial or hematopoietic progenitors in fli1b mutant embryos, we analyzed its expression in fli1b^+/− and fli1b^−/− zebrafish embryos at the 15-somite stage. At least two distinct scl⁺ cell populations could be identified in the trunk region of fli1b^+/− embryos at this stage: low scl, high GFP (scl^low GFP⁺) cells, which presumably correspond to vascular endothelial cells, and high scl single-positive cells (scl⁺ GFP⁻), which are presumptive hematopoietic progenitors (Fig. 3A). Notably, at this stage, hematopoietic (scl⁺ GFP⁻) cells were located bilaterally, whereas most endothelial cells were either positioned at the midline or were in the process of migrating towards the midline (Fig. 3A). Intriguingly, the number of hematopoietic (scl⁺ GFP⁻) cells was significantly reduced in fli1b^−/− embryos. In contrast, the number of scl⁺ GFP⁺ cells was increased in fli1b^−/− mutants (Fig. 3B-F). However, these double-positive cells did not migrate to the midline, where vascular endothelial cells were located, but were positioned bilaterally, similar to hematopoietic progenitors in fli1b^+/− embryos (Fig. 3B-D). These results suggest that some cells have a mixed identity and co-express hematopoietic and vascular endothelial markers in fli1b^−/− mutants.

In fli1b^+/− embryos, bilaterally located hematopoietic cells showed strong scl expression, whereas endothelial cells had much lower scl expression (Fig. 3A). The hematopoietic cells (scl⁺ GFP⁻) showed greatly reduced intensity of scl expression in fli1b^−/− mutants. In contrast, bilaterally located scl⁺ GFP⁺ cells showed stronger scl expression in fli1b^−/− mutants compared to that in fli1b^+/− embryos (Fig. 3G,H). These results suggest that in fli1b^−/− mutants, (1) scl expression in hematopoietic progenitors is reduced, thus resulting in defective erythropoiesis; and (2) a subset of erythroid progenitors show endothelial fli1b expression, resulting in mixed lineage identity.

We subsequently tested the epistatic relationship of scl and fli1b in vascular endothelial and blood cells. To test whether scl can regulate fli1b expression in vascular endothelial cells, wt embryos were injected at the single-cell stage with a previously validated scl MO to knock down scl function. As reported previously, similar to scl genetic mutants (Bussmann et al., 2007; Dooley et al., 2005; Juarez et al., 2005; Patterson et al., 2005), erythroid-specific gata1 expression was strongly reduced in scl MO-injected embryos, compared to that in uninjected controls. Similarly, endothelial fli1b expression was also reduced in scl MO-injected embryos (Fig. 3I-K). These results suggest that scl is required to maintain fli1b expression in vascular endothelial cells.

To test whether hematopoietic defects observed in fli1b double mutants can be rescued by upregulation of scl expression, we microinjected in vitro-synthesized scl mRNA into control wt fli1a:GFP and fli1b^−/− embryos and assayed for gata1 expression at 22 hpf using HCR. Wt embryos injected with scl mRNA showed ectopic expansion of both gata1 and GFP expression (Fig. 3L-P). Injected fli1b^−/− embryos also showed increased gata1 expression compared to that in uninjected fli1b^−/− control embryos (Fig. 3L,Q-T). Thus, scl overexpression can partially rescue hematopoietic defects in fli1b^−/− embryos.

In addition to PLPM-derived erythrocytes, primitive hematopoietic lineages also include myeloid cells, such as macrophages and neutrophils, which are known to originate in the ALPM (Davidson and Zon, 2004; Warga et al., 2009). To test whether myeloid cells were affected in fli1b^−/− embryos, we analyzed expression of the pan-myeloid marker lcp1, the macrophage marker mpeg1.1 and the neutrophil marker lyz at 24 hpf. We noticed two populations of lcp1⁺ myeloid cells along the surface of the yolk of fli1b^+/− embryos. 41.9% of lcp1⁺ cells were positive for GFP expression, whereas the remaining 58.1% were negative for GFP (Fig. 4A,C). Interestingly, the number of GFP-negative (lcp1⁺ GFP⁻) cells was reduced in fli1b^−/− embryos, whereas the number of double-positive (lcp1⁺ GFP⁺) cells was significantly increased (Fig. 4B,C). Consequently, most lcp1⁺ cells (89.1%) were positive for GFP expression in fli1b mutants. In addition, GFP signals in myeloid cells were much brighter in fli1b^−/− embryos, and these cells were readily apparent at the surface of the yolk (Fig. 4B). Similar to lcp1, two populations of lyz⁺ and mpeg1.1⁺ cells were observed in fli1b^+/− embryos. 60.8% of lyz⁺ and 23.4% of mpeg1.1⁺ cells were positive for GFP expression (Fig. 4D,F,G,I). Similarly, the numbers of GFP-negative neutrophils (lyz⁺ GFP⁻) and macrophages (mpeg1.1⁺ GFP⁻) were greatly reduced, whereas the numbers of GFP-positive neutrophils and macrophages were significantly increased in fli1b^−/− embryos (Fig. 4E,F,H,I). Thus, the majority of lyz⁺ neutrophils (94.8%) and macrophages (86.5%) were positive for GFP expression in fli1b mutants at 24 hpf.

In wt embryos, the earliest myeloid cells originate from the ALPM region at approximately the 10- to 15-somite stages and are largely negative for endothelial marker expression (Gurung et al., 2024). To clarify how fli1b affects the emergence of myeloid cells, we analyzed expression of the myeloid progenitor marker spi1b (also known as pu.1) at the 15-somite stage. As reported previously (Gurung et al., 2024), the majority of spi1b⁺ cells were negative for GFP expression in fli1b^+/− embryos at this stage, whereas only a small fraction (10.4%) were positive for GFP (Fig. 5A,B,E). fli1b^−/− mutants showed a reduction in the number of spi1b⁺ GFP⁻ cells, whereas the small number of spi1b⁺ GFP⁺ cells was not changed (Fig. 5C-E).

We have recently demonstrated that the endocardium functions as a secondary hematopoietic site and gives rise to myeloid cells at the 24 to 26 hpf stage, which are initially positive for expression of the vascular endothelial/endocardial marker kdrl (Gurung et al., 2024). As a result, two populations of myeloid cells were distinguished, kdrl⁻ myeloid cells derived from the early ALPM and kdrl⁺ myeloid cells derived from the endocardium. This suggests that the GFP-positive myeloid cell population, which is expanded in fli1b mutants, could be derived from the endocardium. To test this, we analyzed spi1b expression at the surface of the yolk and in the endocardium in fli1b^+/− and fli1b^−/− embryos at the 24 hpf stage. Like other myeloid markers, the number of spi1b⁺ GFP⁺ cells at the surface of the yolk was increased, whereas the number of spi1b⁺ GFP⁻ cells was decreased in fli1b^−/− embryos (Fig. 5F-H). Importantly, there was also an increase in the number of double-positive cells (spi1b⁺ GFP⁺) in the endocardium in fli1b^−/− mutants, whereas no significant change in the number of single-positive cells (spi1b⁺ GFP⁻) was observed (Fig. 5I-K). This suggests that fli1b^−/− mutants show reduced primitive myelopoiesis in the ALPM and increased myelopoiesis in the endocardium.

To determine whether these two different myeloid cell populations were present at a later stage, we analyzed myeloid marker expression and GFP fluorescence in the cranial region of fli1b^+/− and fli1b^−/− embryos at 3 dpf. At this stage, a striking difference in GFP expression was apparent between fli1b^+/− and fli1b^−/− embryos. GFP expression was restricted to vascular endothelial cells in fli1b^+/− embryos, whereas there were multiple GFP⁺ macrophage-like cells with filopodia scattered in the tissues outside of the cranial vasculature in fli1b^−/− embryos (Fig. 6A-B′). Time-lapse imaging demonstrated that these GFP⁺ cells actively migrated in the brain tissue in a manner similar to microglia (Movie 1). To characterize the identity of these cells, HCR was performed at 3 dpf for expression of the pan-myeloid marker lcp1, the neutrophil marker lyz and the macrophage marker mpeg1.1. Out of all GFP⁺ microglia-like cells quantified in a selected cranial region of fli1b mutant embryos, 99.2% were positive for lcp1 (n=12 embryos), 100% were positive for mpeg1.1 (n=10 embryos) and 19% were positive for lyz (n=9 embryos) (Fig. 6C-K). This argues that most of these migratory GFP⁺ cells are macrophages, whereas a subset of them are neutrophils, and some may co-express both markers. Because these cells were positive for GFP expression, which is linked to the fli1b mutation, this suggested that these cells also co-express vascular endothelial markers. Therefore, we performed HCR analysis for expression of the endothelial marker kdrl and cldn5b. Strikingly, 53% of all microglia-like GFP⁺ cells in the cranial region of fli1b mutants were also positive for kdrl expression (n=9 embryos), whereas 23.5% were positive for cldn5b expression (n=10 embryos). Almost none or very few such cells were observed in fli1b^+/− embryos, in which kdrl and cldn5b expression was restricted to vascular endothelial cells (Fig. 6L-Q). These results argue that many myeloid cells co-express endothelial markers in fli1b^−/− embryos.

To characterize transcriptomic changes in different cell populations, we performed single-cell RNA-sequencing (RNA-seq) analysis. Homozygous fli1b^−/− mutant and sibling wt parents were in-crossed to generate populations of fli1b^−/− and wt embryos. All cells were disaggregated at the 23 to 24 hpf stage and subjected to single-cell RNA-seq profiling using the Chromium (10× Genomics) platform. After the initial filtering steps, 12,957 cells were identified in wt embryos and 15,271 cells were identified in fli1b mutant embryos. Unsupervised clustering of aggregated wt and fli1b cells identified 26 distinct clusters (Fig. 7A). We focused further analysis on vascular endothelial cell, RBC and myeloid cell clusters, which were identified based on expression of the marker genes cdh5, gata1a and spi1b, respectively (Fig. 7A,B). To identify changes in gene expression, we performed differential expression analysis between endothelial cell, RBC and myeloid cell populations in fli1b mutant and wt embryos. Differential expression analysis in RBCs identified approximately 20 significantly downregulated genes in fli1b mutants, which included the myeloid markers spi1b, lcp1 and coro1a, and RBC-specific markers, such as urod and alas2 (Table 1). Many other RBC markers, including the globin genes hbae4, hbae3, hbbe3 and hbbe2, as well as gata1a also showed greater than 2-fold reduction in fli1b mutants, although they did not reach statistical significance (P<0.05) (Table 1, Fig. 7E). It is known that spi1b and other myeloid markers are also expressed in early erythrocytes (Bennett et al., 2001; Herbomel et al., 1999; Lieschke et al., 2001), which explains why their downregulation was observed in the RBC cluster. Differential expression analysis in vascular endothelial cells identified 19 significantly downregulated genes, including plvapb and glula (Table 2). The top endothelial downregulated gene was fli1b itself, expression of which is known to be inhibited in fli1b mutants (Table 2, Fig. 7C) (Craig et al., 2015). Violin plots show that fli1b expression was high in both endothelial and myeloid cells in wt embryos, and very low in fli1b mutant embryos (Fig. 7C). However, most endothelial-specific genes, including kdrl, were not significantly affected in fli1b mutant embryos (Fig. 7E). Differential expression analysis between fli1b mutant and wt embryos in myeloid cells did not identify any significantly downregulated genes, and myeloid spi1b expression was not significantly affected (Fig. 7E). By analyzing the expression of selected marker genes, we noticed that some vascular endothelial genes, including kdrl and cdh5, showed expression in a subset of myeloid cell population in fli1b^−/− mutants, whereas they were absent from myeloid cells in wt embryos (Fig. 7E). This is consistent with our observations of endothelial marker expression in myeloid cells using HCR analysis. Single-cell RNA-seq is known to have limited sensitivity and to be prone to ‘dropouts’ of low-expressing genes (Stegle et al., 2015), which could be the reason why only few myeloid cells show endothelial cell marker expression in this dataset. Intriguingly, scl expression was increased in endothelial cell and myeloid cell populations in fli1b mutants and reduced in RBCs (Fig. 7D). Reduction of scl expression in RBCs correlates with reduced erythroid differentiation. Its increase in endothelial cells could indicate that endothelial cells have a mixed endothelial-hematopoietic identity or that there is an increased population of angioblasts, which have higher scl expression, compared to differentiated endothelial cells. In summary, single-cell transcriptomic analysis correlates with the results of HCR analysis and indicates reduced erythroid differentiation and mixed cell identities in myeloid cells of fli1b mutant embryos.

In this study, we demonstrate the requirement for zebrafish fli1b in regulating embryonic hematopoiesis, including the emergence of the primitive erythroid and myeloid lineages as well as the specification of HSPCs. Zebrafish fli1b mutant embryos showed reduced number of RBCs and diminished expression of erythroid-specific markers, including gata1 and hbae3. The transcription factor Scl is the master regulator of hematopoietic differentiation and functions upstream of gata1 during erythropoiesis. Previous studies have shown that zebrafish scl knockdown results in the inhibition of hematopoiesis and reduction of vasculogenesis (Dooley et al., 2005; Juarez et al., 2005; Patterson et al., 2005). Our results show that scl expression in erythroid progenitors was reduced in fli1b mutant embryos. Furthermore, induction of scl expression partially restored erythroid development in fli1b mutant embryos. These results argue that fli1b regulates primitive erythropoiesis through scl expression. These findings agree with previous studies in mouse embryoid bodies and fetal liver, which have demonstrated that Fli1 and Scl can positively regulate expression of each other (Pimanda et al., 2007).

Although fli1b mutants show strong downregulation of primitive erythropoiesis, fli1b appears to have little to no expression in erythroid cells, and its expression is restricted to vascular endothelial cells. It is possible that fli1b function is required non-cell-autonomously in vascular endothelial cells. However, we consider this possibility unlikely because fli1b is a transcription factor, which typically function cell-autonomously. In addition, RBCs appear to differentiate largely normally even in the absence of vascular endothelial cells, as observed in etv2 mutant or knockdown embryos (Sumanas and Lin, 2006). A more likely explanation is that fli1b regulates hematopoiesis at the hemangioblast stage, where it induces scl expression. However, fli1b expression is not maintained in RBCs and is restricted to endothelial cells at later stages (Fig. 8A).

Although fli1b mutants showed strong inhibition of erythroid differentiation, the phenotype was quite variable between different embryos. This could be due to redundancy with another ETS factor, such as fli1a, that is expressed in both vascular endothelial and hematopoietic cells (Brown et al., 2000). To date, no hematopoietic defects have been described in zebrafish fli1a knockdown embryos and no in-depth studies of fli1a mutant phenotype have been performed.

In addition to defective primitive hematopoiesis, fli1b mutants also display loss or reduction of HSPC formation, as evidenced by the reduced runx1 and cmyb expression. HSPCs emerge from the hemogenic endothelium within the dorsal aorta (Bertrand et al., 2010; Kissa and Herbomel, 2010). However, vasculogenesis and dorsal aorta formation do not appear to be affected in fli1b mutants, based on our previous studies (Craig et al., 2015). This suggests that fli1b mutants have a specific defect in hemogenic endothelium formation, possibly through misregulation of scl expression, which plays a critical role in HSPC specification (Zhen et al., 2013).

Our recent study has demonstrated that primitive myeloid cells emerge from two distinct sites in zebrafish embryos (Gurung et al., 2024). Although the earliest myeloid cells, which include macrophages and neutrophils, originate from the ALPM at the 10- to 15-somite stages, an independent population of myeloid cells emerges from the endocardium at the 22 to 28 hpf stages (Fig. 8B) (Gurung et al., 2024). These cells contribute mostly to the neutrophil lineage, although a small percentage can also contribute to the macrophage lineage. They can be distinguished by early expression of vascular endothelial/endocardial markers such as kdrl, which are absent from ALPM-derived myeloid cells (Gurung et al., 2024). However, this endothelial marker expression in wt embryos is transient and is no longer observed at 48 hpf or later stages in the endocardial-derived myeloid cells. Intriguingly, fli1b mutants exhibit opposing effects on the two populations of myeloid cells. GFP-negative cells, which likely correspond to the ALPM-derived myeloid cells, are reduced in 24 hpf embryos. This reduction of ALPM-derived myeloid cells was also confirmed by the analysis of spi1b expression at the 15-somite stage, which was strongly reduced in fli1b mutants. In contrast, the GFP⁺ population, which likely corresponds to the endocardial-derived myeloid cells, was significantly expanded in fli1b mutant embryos. This correlates with the higher number of spi1b-positive cells observed in the endocardium of fli1b mutants. The mechanism of increased myeloid cell production in the endocardium is not clear. It may occur as a compensatory response due to the reduction in ALPM-derived myeloid cells. A similar compensation by endocardial-derived myeloid cells was also observed in etv2 mutant embryos, which also exhibit loss of ALPM-derived myeloid cells (Gurung et al., 2024).

Surprisingly, GFP expression in myeloid cells was apparent in fli1b mutants at 3 dpf, which was largely absent in fli1b^+/− embryos. Because GFP expression is linked to fli1b mutation, this suggests that fli1b promoter activity, which is normally restricted to vascular endothelial cells at 3 dpf, was also pronounced in myeloid cells in fli1b mutants. Indeed, our results show that other vascular endothelial markers, such as kdrl and cldn5b, were expressed in the myeloid cells in fli1b mutants, whereas they were absent from the myeloid cells in fli1b^+/− embryos. Although endocardial-derived myeloid cells show residual expression of endothelial markers at 24 hpf, such expression is no longer apparent at 2-3 dpf (Gurung et al., 2024). Therefore, the endocardial origin cannot explain prolonged expression of GFP (linked to fli1b mutation) and vascular endothelial markers in fli1b mutants. It is possible that these cells are arrested in a hemangioblast-like state and, even though they commit to myeloid differentiation, they are unable to downregulate vascular endothelial markers. This suggests that fli1b is important for cells to transition from hemangioblast to hematopoietic lineages.

Global transcriptomic single-cell analysis confirmed the results obtained using HCR. It showed reduced expression of scl, gata1 and hbae3 and many other RBC markers in erythroid cells of fli1b mutants. Although overall myeloid spi1b expression was unaffected, there was a small but notable expression of vascular endothelial markers apparent in myeloid cells, consistent with HCR results. However, only a relatively small number of genes were significantly affected in fli1b mutants. This is likely due to the methodology used in this study, which involved a whole-embryo disaggregation strategy that resulted in identification of rather small populations of hematopoietic cells and, therefore, limited the statistical significance of observed changes.

Similar to zebrafish fli1b mutants, mouse Fli1 has been implicated in regulating myeloid differentiation and Spi1 expression (Masuya et al., 2005; Starck et al., 1999; Suzuki et al., 2013). Differently from the zebrafish fli1b mutants, primitive erythropoiesis does not appear to be greatly reduced in mouse Fli1 mutants (Spyropoulos et al., 2000). In contrast, Fli1 overexpression reduces erythroid differentiation in mammalian cells, suggesting that Fli1 negatively regulates erythropoiesis in mouse embryos (Pereira et al., 1999; Tamir et al., 1999). It is possible that mouse Fli1 mutant embryos exhibit slight defects in erythroid differentiation, which have not been observed or well characterized yet. However, it is also possible that there are species-specific differences between transcription factors. In mice, the related ETS transcription factor Etv2 plays a critical role in both hematopoiesis and vasculogenesis, and mouse Etv2 mutants show loss of hematopoiesis, including deficient erythroid differentiation (Ferdous et al., 2009; Lee et al., 2008). However, zebrafish etv2 knockdown embryos show only loss of endothelial and myeloid lineages, whereas erythropoiesis is largely unaffected (Sumanas et al., 2008; Sumanas and Lin, 2006). It is possible that zebrafish Fli1b has a requirement in erythropoiesis, which is equivalent to the role of Etv2 in murine embryos.

In summary, this study has identified new roles for the transcription factor Fli1b in regulating hematopoiesis. These results will promote our understanding of molecular mechanisms that regulate hematovascular differentiation during development and may also contribute to the development of improved strategies to generate hematopoietic stem and progenitor cells for regenerative medicine.

All studies were performed according to the animal protocols approved by the University of South Florida Institutional Animal Care and Use Committee. Male and female adults between 3 months and 2 years of age were used for mating to acquire embryos for experiments. Experiments utilizing embryos during the stages of somitogenesis were performed with incubation at 25°C. All other embryos used for analysis were incubated at 28.5°C in embryo medium; embryos analyzed past 24 hpf were also treated with 0.003% 1-phenyl-2-thiourea (PTU; Millipore Sigma) to prevent pigment formation. The previously established fli1b^tpl50Gt line (further labeled as fli1b^+/− or fli1b^−/−) was used in the study (Craig et al., 2015). The targeting vector contains Gal4 and UAS:GFP elements, and results in vascular endothelial GFP expression when inserted in fli1b locus (Craig et al., 2015). fli1b^−/− mutant embryos were obtained by the cross of homozygous fli1b^−/− parents, whereas control fli1b^+/− embryos were obtained by mating fli1b^+/− adults (which were siblings of fli1b^−/− adults) to wt AB line and selecting embryos with GFP fluorescence. In some experiments, wt Tg(fli1a:GFP)^y1 (Lawson and Weinstein, 2002) controls were used, as noted in the text or the figure legend. Other lines used included: Tg(kdrl:GFP)^s843 (Jin et al., 2005), Tg(kdrl:mCherry)^ci5 (Proulx et al., 2010) and TgPAC(scl:dsRed) (Zhen et al., 2013).

WISH was performed as previously described (Jowett, 1999). RNA probes were synthesized using T3, T7 or SP6 RNA Polymerases (Promega). RNA probes used were for gata1 (Detrich et al., 1995), hbbe3 (Brownlie et al., 2003), scl (Liao et al., 1998), runx1 (Galloway et al., 2005) and cmyb (Thompson et al., 1998). Prior to WISH, embryos were fixed in BT-Fix (4% paraformaldehyde in 1× PBS) at 4°C overnight, then dehydrated through sequential ethanol series and stored at −20°C in 100% ethanol. For WISH, embryos were sequentially rehydrated and washed three times with PBT (1× PBS, 0.2% BSA, 0.2% Tween 20). Embryos fixed at 24 hpf and 3 dpf were digested with proteinase K (10 µg/ml; Thermo Fisher Scientific) for 5 min and 30 min, respectively, at room temperature. Following this, embryos were washed with PBT and refixed with BT-Fix for 20 min, then washed three times for 5 min with PBT at room temperature. Embryos were then incubated in prewarmed prehybridization buffer [prehyb: 50% formamide, 5× sodium citrate buffer ( SSC) 50 µg/ml heparin, 5 mM EDTA, 0.5 mg/ml torula yeast RNA (Millipore Sigma), 9.2 mM citric acid, 0.1% Tween 20] for 2 h at 65°C while shaking, incubated overnight with a probe solution in prehyb at 65°C, and subsequently washed with the following solutions at 65°C: 75% prehyb and 25% 2× SSC; 50% prehyb and 50% 2× SSC; 25% prehyb and 75% 2× SSC; 2× SSC; and twice with 0.2× SSC. Subsequent room temperature washes included: 75% 0.2× SSC and 25% PBT; 50% 0.2× SSC and 50% PBT; 25% 0.2× SSC and 75% PBT; and 100% PBT. After washes, embryos were incubated overnight in 1× PBT with 2% lamb serum and a 1:4000 dilution of anti-digoxigenin-AP antibody (Sigma-Aldrich, 11093274910). Embryos were subjected to six washes in PBT and were then incubated in AP buffer (100 mM NaCl, 50 mM MgCl₂, 100 mM Tris-Cl, pH 9.5, 0.1% Tween 20) containing 4-nitrotetrazolium blue chloride (NBT, 0.225 mg/ml; Millipore Sigma) and 5-bromo-4-chloro-3-indolyl phosphate disodium salt (BCIP, 0.175 mg/ml; Millipore Sigma). Staining was ended with additional washes in PBT, then embryos were incubated in BT-Fix for 2 h. Embryos were scored into groups and subsequently mounted in slide chambers using 0.6% low-melting agarose. Stained embryos following WISH were imaged using a 10× objective on an AxioImager Z1 compound microscope from Zeiss, equipped with an Axiocam ICC3 color camera, or a Nikon Eclipse Ni-E microscope.

Live embryos at 2 and 3 dpf were developed with o-dianisidine staining solution [6 mg o-dianisidine-chloride, 6 ml water, 33 µl 3 M sodium acetate (pH 5), 4 ml ethanol, 186 µl 35% hydrogen peroxide] in the dark for 1 h. Then, embryos were washed with PBST (1× PBS, 0.2% Tween 20), fixed with BT-Fix for 2 h at room temperature, and washed with PBST three times for 5 min.

Batches of 20-25 embryos were frozen on dry ice at the 20-somite stage. They were then homogenized in lysis buffer from the RNAqueous 4-PCR Kit (Thermo Fisher Scientific) with a 23-gauge needle, and extraction of RNA from embryos was performed using the same RNAqueous 4-PCR kit (Thermo Fisher Scientific). cDNA synthesis was performed using the SuperScript VILO cDNA Synthesis Kit or the Superscript III cDNA synthesis kit (Thermo Fisher Scientific). Quantitative real-time PCR was performed using PowerUp SYBR Green Master Mix (Applied Biosystems) in a Cielo Real-Time PCR System (Azure Biosystems). Quantification was performed using the relative standard curve method; controls were assigned a value of 1 and experimental values were computed by the software based on C_T values, relative to the control standard curve. For each experiment, two to three replicates were performed with duplicates in each run, and ef1α was used as an endogenous control. The relative quantity of cDNA for each sample was normalized to the value of ef1α. Data were analyzed using the Azure Cielo Manager Software (Azure Biosystems). Primer sequences are listed in Table S1.

FISH was performed using HCR (Choi et al., 2018). HCR v3.0 probes for gata1, fli1b, scl, lcp1, kdrl, lyz, mpeg1.1 and spi1b were obtained from Molecular Instruments, Los Angeles, CA, USA. HCR was performed using embryos fixed overnight with BT-Fix at 4°C, dehydrated in sequential ethanol series and stored at −20°C. Embryos were rehydrated and washed three times with PBT. A prehybridization step was done using hybridization buffer (30% formamide, 5× SSC, 9 mM citric acid at pH 6.0, 0.1% Tween 20, 50 μg/ml heparin, 1× Denhardt's solution and 10% dextran sulfate) for 30 min at 37°C. Following this, each probe (2 pmol) was combined with 500 µl of hybridization buffer and incubated overnight gyrating at 37°C. Following incubation, embryos were washed four times for 15 min with 30% formamide, 5× SSC, 9 mM citric acid at pH 6.0, 0.1% Tween 20 and 50 µg/ml heparin at 37°C. This was then followed with three 5 min washes with 5× SSCT (5× SSC and 0.1% Tween 20) at room temperature and a 30 min incubation in amplification buffer (5× SSC, 0.1% Tween 20 and 10% dextran sulfate) at room temperature. Hairpin probes (30 pmol each), fluorescently labeled through snap cooling of 3 µM stock solution, were added to the embryos in amplification buffer and incubated overnight at room temperature in the dark. Next, samples were washed five times with 5× SSCT. Embryos were mounted in 0.6% low-melting agarose and imaged using a Nikon A1R HD confocal microscope using a 20× objective.

A previously validated translation-blocking scl MO (5′-AATGCTCTTACCATCGTTGATTTCA-3′) obtained from Gene Tools (Philomath, OR, USA) was used (Dooley et al., 2005; Gurung et al., 2024). Embryos were injected into the yolk at the one-cell stage with 7.5 ng (per embryo) of this scl MO.

scl mRNA (Addgene, Watertown, MA, USA) was prepared by NotI linearization of the scl-pCS2+ expression vector (Mattonet et al., 2022), followed by transcription using SP6 mMessage Machine kit (Thermo Fisher Scientific). At the one-cell stage, 100 pg of scl mRNA was injected into zebrafish embryos. Embryos were then fixed at 22 hpf for FISH.

fli1b^−/− and wt embryos were obtained by the corresponding in-crosses of sibling parents that were either homozygous mutant (fli1b^−/−) or wt (fli1b^+/+). Approximately 40 embryos from each sample were manually dechorionated at approximately 23-24 hpf, then transferred into 1.5 ml tubes. Whole embryos were then dissociated into a single-cell suspension by cold protease tissue dissociation protocol (Potter and Potter, 2019). All steps were performed on ice. Embryo water was removed and deyolking buffer was added (55 mM NaCl, 1.8 mM KCl and 1.25 mM NaHCO₃ in 1× PBS). Embryos were pipetted up and down until the yolk was dissolved, then centrifuged at 300 g for 1 min. The supernatant was aspirated and embryos were resuspended in 0.5× Danieau buffer. This step was repeated and, after the supernatant was aspirated, Bacillus lichenformis enzyme mix [1× fetal bovine serum (FBS; Millipore Sigma) in PBS, 0.5 mM of EDTA, 125 U/ml DNase 1 (Thermo Fisher Scientific), 10 mg/ml Bacillus lichenformis (Millipore Sigma)] was prepared and added to embryos. Embryos were incubated on ice for 20 min and titrated 15 times every 2 min. Cell dissociation was monitored under microscope and, once completed, centrifugation was performed at 1200 g for 5 min. The supernatant was discarded and the pellet was resuspended in 10% FBS in PBS. Cells were then passed through 20-µm strainer, then washed with 500 µl of 10% FBS. Cells were then pelleted again at 1200 g for 5 min, washed once with 1 ml of 10% FBS, and centrifuged at 1200 g for 5 min. Cells were resuspended in 250 µl of 10% FBS, then counted using a hemocytometer. A suspension of approximately 16,000 cells (for each genotype) was loaded onto the 10× Genomics Single Cell 3′ chip at the University of South Florida Genomics Core. Samples were sequenced on an Illumina NextSeq 2000 instrument (Illumina, San Diego, CA, USA). The raw .fastq files obtained from the sequencing core were then mapped to the Danio rerio genome (Zv11) to generate single-cell feature counts using Cell Ranger version 8.0 (10× Genomics). Datasets from wt and fli1b mutant embryos were aggregated using the Cell Ranger ‘aggreg’ function. Two rounds of sequencing were performed for each sample, and the sequence files were aggregated in Cell Ranger for the final analysis. Cell clustering and data analysis were performed using Loupe Browser version 8.0 (10× Genomics). Cells were filtered based on unique molecular identifiers (2213-30159 range), expressed genes (799-4873) and mitochondrial unique molecular identifiers (<5%). This resulted in the identification of 28,228 cells (12,957 wt and 15,271 fli1b^−/− cells), which clustered into 26 individual clusters. The myeloid cell cluster was not identified in the initial clustering and was then subset manually based on marker gene expression using Loupe Browser.

Live or fixed embryos were mounted in 0.6% low-melting point agarose. 0.002% tricaine (Sigma-Aldrich, SKU#A5040) was added for live-embryo imaging. Imaging was performed on a Nikon Eclipse confocal microscope. Images were processed using the ‘denoiseAI’ function (NIS Elements, Nikon) to reduce the noise. Selected slices were used to create a maximum-intensity projection in NIS Elements AR software (Nikon). All image panels were created using Adobe Photoshop CC, and image levels were adjusted in Adobe Photoshop CC to optimize brightness and contrast. The same adjustments were performed on control and experimental embryos in all cases.

To quantify the fluorescence intensity of scl expression at the 15-somite stage (Fig. 3A-D), the average integrated density was measured using Fiji/ImageJ within small areas of ten selected representative cells in each embryo. The average background value was then subtracted for each embryo.

To quantify the intensity of fli1b expression in control and scl MO-injected embryos (Fig. 3I-K), the rectangle tool in Fiji was used to select the measured area that constituted most of the fli1b expression domain in the trunk region above the yolk extension. The width and length of the rectangle were kept constant between all embryos analyzed. The integrated density was measured and the background was subtracted. All measurements in scl MO-injected embryos were then normalized to the average values in control embryos.

Fluorescence intensity quantification in scl mRNA injection experiments (Fig. 3L-T) was performed using the custom shape tool in Fiji software to make a tight area encircling gata1 expression along the yolk extension of the embryo. The integrated intensity value was measured in the selected area. The background was measured using the rectangle tool and adjusted according to the area measured for gata1 expression. Then, it was subtracted from the integrated density to calculate the final value.

Sample size and number of replicates are indicated in the main figure or figure legends for each specific analysis. All experiments have been replicated at least twice. Calculation of mean±s.d., two-tailed unpaired Student's t-test, Fisher's test and plotting of graphs were performed in GraphPad Prism.

We thank Matthew Mercurio and the University of South Florida Genomics Core for their assistance with performing single-cell RNA-seq.