Endothelial versus pronephron fate decision is modulated by the transcription factors Cloche/Npas4l, Tal1, and Lmo2

Abstract

Endothelial specification is a key event during embryogenesis; however, when, and how, endothelial cells separate from other lineages is poorly understood. In zebrafish, Npas4l is indispensable for endothelial specification by inducing the expression of the transcription factor genes etsrp, tal1, and lmo2. We generated a knock-in reporter in zebrafish npas4l to visualize endothelial progenitors and their derivatives in wild-type and mutant embryos. Unexpectedly, we find that in npas4l mutants, npas4l reporter–expressing cells contribute to the pronephron tubules. Single-cell transcriptomics and live imaging of the early lateral plate mesoderm in wild-type embryos indeed reveals coexpression of endothelial and pronephron markers, a finding confirmed by creERT2-based lineage tracing. Increased contribution of npas4l reporter–expressing cells to pronephron tubules is also observed in tal1 and lmo2 mutants and is reversed in npas4l mutants injected with tal1 mRNA. Together, these data reveal that Npas4l/Tal1/Lmo2 regulate the fate decision between the endothelial and pronephron lineages.

The transcription factor Npas4l and its transcriptional effectors Tal1/Lmo2 regulate the endothelial versus pronephron fate.

INTRODUCTION

Vascular development includes complex morphogenetic and cell differentiation processes, starting with the specification of the endothelial progenitors during late gastrulation, and their subsequent assembly into the major axial vessels (1, 2). Several model systems have been used to study vascular development including the zebrafish, which offers unique advantages for live imaging (3, 4) and is resilient against cardiovascular defects at embryonic and early larval stages (5).

Endothelial origins have been mapped to the lateral plate mesoderm (LPM), a tissue that arises on the ventrolateral sides of the vertebrate embryo and gives rise to a multitude of cell types including pronephron, endothelial, blood, and cardiac progenitors (6–9). Fate mapping experiments in zebrafish have shown that single cells labeled at 40 to 50% epiboly stages [~5 hours post-fertilization (hpf)] can give rise to both blood and endothelial cells, providing evidence for the so-called hemangioblast (10, 11). Pronephron, endothelial, and hematopoietic progenitor populations overlap at the three-somite stage (11 hpf) but start to form distinct bilateral stripes of cells along the anterior-posterior axis at early- to midsomitogenesis stages (~10 to 14 hpf) (12). Pronephron progenitors, which express pax2a and pax8, are located between the more lateral hand2-expressing cells, which give rise to smooth muscle and mesothelial cells (4, 13), and the more medial endothelial and hematopoietic progenitors, which express etsrp (a.k.a. etv2; https://zfin.org/ZDB-GENE-050622-14), tal1 (a.k.a. scl; https://zfin.org/ZDB-GENE-980526-501), and lmo2 (2). Starting at around 14 hpf, endothelial and hematopoietic progenitors move along the ventral side of the forming somites toward the midline, where they coalesce into a vascular cord that remodels into two major axial vessels, the dorsal aorta (DA) and posterior cardinal vein (PCV) (14, 15). Despite our growing understanding of cell fate determination within the LPM, the underlying cellular and molecular mechanisms remain incompletely understood.

In zebrafish, the earliest known event in endothelial specification is the expression of the basic helix-loop-helix–Per-Arnt-Sim (bHLH-PAS) transcription factor gene cloche/npas4l (16). cloche/npas4l mutants lack most endothelial and hematopoietic cells (17–19). Mechanistically, Npas4l positively regulates the expression of several target genes including the transcription factor genes etsrp, tal1, and lmo2 (20, 21), thereby promoting the formation of endothelial and hematopoietic cells. Etsrp (Erythroblast Transformation-Specific)/Etv2 is an ETS transcription factor required for endothelial and myeloid specification (22, 23). In zebrafish etsrp morphants and mutants, etsrp-expressing cells, which differentiate into endothelial cells in wild-type animals, acquire alternative fates including cardiac and skeletal muscle (24, 25). In mice lacking the bHLH transcription factor TAL1 (T-cell acute lymphocytic leukemia protein 1), vascular plexus remodeling and hematopoietic development are both impaired (26, 27); and in explant cultures of Tal1 mutant mice, some endothelial progenitors contribute to spontaneously beating cardiac foci (28). Similar to what is observed in Tal1 mutant mice, zebrafish tal1 mutants lack the expression of the erythroid marker gata1a and erythrocytes (29), and they exhibit vascular defects (30). Endothelial and blood cell differentiation in npas4l mutants can be partially rescued by global overexpression of zebrafish tal1 as assessed by in situ hybridization (29). LIM domain only 2 (LMO2) is a LIM domain transcription factor required for hematopoietic development, and both zebrafish and mouse LMO2 loss-of-function models exhibit defective erythroid development (31, 32). Zebrafish lmo2 mutants also display vascular defects that have been linked to reduced endothelial cell migration (33, 34). Mechanistically, LMO2 is part of transcription factor complexes involving TAL1 (31, 35), where it has been reported to function as a scaffold (36). While the importance of these four transcription factors in endothelial and hematopoietic cell specification has been established, their role in separating these lineages from other LPM-derived cell populations remains unclear.

Here, we trace npas4l-expressing cells in wild-type zebrafish embryos and in npas4l, etsrp, tal1, and lmo2 mutants. We find that npas4l reporter–expressing cells in npas4l mutants exhibit two distinct cell fate changes: (i) a contribution to pronephron tubules, which we also observed in tal1 and lmo2 mutants, and (ii) a contribution to skeletal muscle, which we, and others (25), also observed in etsrp mutants. Single-cell RNA sequencing (scRNA-seq) analysis complemented by live imaging reveals the existence of a population of LPM cells that coexpresses endothelial and pronephron markers, potentially representing multilineage progenitors. Lineage tracing shows that pax2a-expressing cells contribute to endothelial cells as well as the pronephron tubules, further supporting their multilineage potential. Together, these data indicate that a transcriptional network downstream of Npas4l promotes endothelial development at the expense of pronephron and somite fates, with Tal1 and Lmo2 contributing to the block toward pronephron fates and Etsrp to the block toward somite fates.

RESULTS

A knock-in reporter for npas4l expression visualizes early endothelial development

Zebrafish npas4l mutants lack most endothelial and blood cells (17), and npas4l function is required cell-autonomously in the endothelial lineage for its specification (19). However, what happens to npas4l-expressing cells in npas4l mutants is unclear; these cells could undergo apoptosis, differentiate into other lineages, or arrest in their differentiation process. To distinguish between these possibilities, we generated a reporter line to visualize and track npas4l-expressing cells. We generated an npas4l-p2a-Gal4-VP16 reporter line (npas4l^bns313; Fig. 1A and fig. S1A) by inserting a p2a-Gal4-VP16 cassette (37) in the 3′ end of the endogenous npas4l coding sequence. Homozygous npas4l^bns313 animals develop normally and are viable (Fig. 1B), indicating that npas4l function is not affected by this insertion.

Fig. 1. Generation and use of a knock-in reporter to trace npas4l-expressing cells in wild-type and mutant embryos.

(A) An npas4l-p2a-Gal4-VP16 allele (bns313) was generated by knock-in at the 3′ end of the npas4l coding sequence. Subsequently, a knockout derivative of this allele, one lacking Npas4l function, was generated by mutating the DNA binding domain–encoding region. (B and B′) Use of a Tg(UAS:GFP) background to reveal the lack of intersomitic vessels (ISVs) in npas4l ^bns423 mutants; (B) shows an npas4l^+/+ embryo from an intercross of npas4l^bns313 hets, and (B′) shows an npas4l^−/− embryo from an intercross of npas4l^bns423 hets. (C and C′) npas4l reporter expression in the vasculature of npas4l^+/− (left) and npas4l^−/− (right) embryos. npas4l reporter expression marks endothelial cells in npas4l heterozygous embryos but skeletal muscle (M) and a ventral row of round cells in npas4l mutants (arrowheads). DA, dorsal aorta; PCV, posterior cardinal vein. Scale bars, 200 μm (B) and 50 μm (C).

When crossing the npas4l^bns313 line to a fluorescent UAS reporter line, we observed reporter expression in presumptive endothelial progenitors as early as 10 hpf (tailbud stage; fig. S1B). At 24 hpf, endothelial cells and circulating cells express the npas4l reporter (Fig. 1B). At this stage, we also observed npas4l reporter expression in yolk syncytial layer cells and in a few skeletal muscle, pronephron tubule, and myocardial cells (Fig. 1B and fig. S2). npas4l reporter expression in circulating cells diminished over time and was not detectable beyond 48 hpf; however, npas4l expression in endothelial cells could still be observed at 120 hpf. Although npas4l expression peaks before the end of gastrulation and is barely observed after 24 hpf (16), the stability of Gal4-VP16 and green fluorescent protein (GFP) sustains the fluorescence labeling of npas4l-expressing cells for at least the first 48 hours of development.

npas4l reporter–expressing cells contribute to pronephron and skeletal muscle lineages in npas4l^−/− embryos

Next, we mutated the DNA binding domain–encoding region of npas4l in the npas4l^bns313 allele and recovered a +36–base pair (bp) in-frame indel allele that retains the Gal4 reporter expression but lacks Npas4l function (Fig. 1A and fig. S3A). Homozygous mutant embryos for this allele (npas4l^bns423) lack most endothelial and blood cells, phenocopying other strong npas4l mutant alleles including m39 (17, 19) and s5 (16). In addition, the npas4l^bns423 allele fails to complement the strong npas4l mutant allele bns297 (fig. S3B) (21).

While exhibiting a strongly decreased number of endothelial and blood cells, homozygous npas4l^bns423 mutants also display ectopic npas4l reporter expression in skeletal muscle cells (Fig. 1, B and C) and in a ventrolateral population of round cells (Fig. 2A). As the morphology and anatomical position of these round cells suggested that they might be pronephron tubule cells, we immunostained transverse sections at 20 hpf and observed that they indeed express the pronephron transcription factor Pax2a (Fig. 2B). Analyses at later stages show that these cells contribute to the glomeruli (fig. S4A) and the mature pronephron tubules (fig. S4B). While the pronephron tubules in npas4l^−/− embryos are clearly enlarged, the glomeruli appear to form as in wild type, as previously described (38).

Fig. 2. npas4l reporter–expressing cells are observed in pronephron tubules and skeletal muscle in npas4l mutants.

(A) npas4l reporter expression in ventral cells of 24 hpf npas4l^+/− (left) and npas4l^−/− (right) embryos. The main axial vessels (DA and PCV) are visible in npas4l^+/− embryos, while only the more ventrally located cells are detected in npas4l^−/− embryos (arrowheads). (B) Transverse sections of 20 hpf npas4l^+/− (left) and npas4l^−/− (right) embryos. The endothelial progenitors (outlined by yellow dotted lines) in npas4l mutants fail to reach the midline unlike those in wild-type siblings; they also remain round. The round ventrolateral npas4l reporter–expressing cells in npas4l^−/− embryos also express Pax2a and are part of the pronephron tubules. (C to E) Numbers of npas4l reporter/Pax2a double-positive cells (C), all Pax2a-positive cells (D), and npas4l reporter–expressing skeletal muscle cells (E) per field of view (319.45-μm-long area over the yolk extension). Cells were counted in three-dimensional (3D) confocal lateral views of npas4l^+/+, npas4l^+/−, and npas4l^−/− embryos at 24 hpf. Data are represented as individual data points, median, interquartile range, and extremes excluding outliers. P values were calculated by Poisson regression and adjusted to counteract the multiple testing problem. Data represent a subset of the data shown in Fig. 6 (B to D). DA, dorsal aorta; PCV, posterior cardinal vein. Scale bars, 50 μm (A) and 10 μm (B).

To quantify these changes, we immunostained 24 hpf npas4l^+/+, npas4l^+/−, and npas4l^−/− embryos for npas4l reporter expression and Pax2a protein and counted the number of npas4l/Pax2a double-positive pronephron tubule cells (Fig. 2C), Pax2a single-positive pronephron tubule cells (Fig. 2D), and npas4l reporter–expressing skeletal muscle cells (Fig. 2E) in all three genotypes. Although we consistently found a low number of npas4l reporter–expressing pronephron tubule and skeletal muscle cells even in npas4l^+/+ embryos, these numbers increased significantly in npas4l mutants, from a median of 14 to 66.5 Pax2a⁺ pronephron tubule cells and from a median of 36 to 76.5 skeletal muscle cells (Fig. 2, C to E). We did not observe a clear Pax2a expression phenotype at 12 (fig. S5) or 14 (fig. S6) hpf, suggesting that the 24 hpf pronephron tubule phenotype is not due to an enlarged Pax2a expression domain at early somite stages. To test whether this pronephron tubule phenotype was an artifact of this newly generated npas4l mutant allele or reporter, we analyzed two established npas4l alleles, the m39 large deletion allele (fig. S7A) and the s5 point mutation allele (fig. S7B), at 24 hpf and observed enlarged Pax2a expression domains in these mutants as well. From these observations, we speculate that the npas4l/Pax2a double-positive cells are derived from endothelial progenitors that, in the absence of Npas4l function, contribute to the pronephron lineage. These experiments also revealed a significant increase of npas4l expression in pronephron tubule (median = 21) and skeletal muscle (median = 62) cells in npas4l^+/− embryos (Fig. 2, C to E), suggesting a previously undescribed phenotype in these heterozygous embryos.

Together, these data indicate that npas4l mutants develop distinct cell specification phenotypes. The contribution of npas4l reporter–expressing cells to skeletal muscle in npas4l mutants is in agreement with previous findings regarding etsrp reporter expression and the abnormal differentiation of endothelial progenitors to skeletal muscle in etsrp mutants (25). In contrast, the contribution of npas4l reporter–expressing cells to the pronephron tubules has not been reported before.

Transcriptional profile of npas4l reporter–expressing cells in npas4l^+/− and npas4l^−/− embryos

To investigate whether npas4l reporter–expressing cells contributing to pronephron tubules and skeletal muscle in npas4l^−/− embryos exhibit a global change in transcription indicative of a change in cell fate, we compared the transcriptomic profile of npas4l reporter–expressing cells (5207 npas4l^+/− cells and 5262 npas4l^−/− cells) sorted from the trunks of 20 hpf embryos through scRNA-seq (Fig. 3A). By testing parameters of dimensional reduction, we determined that clustering into eight subgroups reflected the best representation of the data (Fig. 3B). We annotated clusters by cell-specific marker gene expression (Fig. 3, B and E, and tables S1 and S2), aided by published in situ hybridization data (ZFIN) and previous scRNA-seq data at similar embryonic stages (39). To investigate the relationship within and between clusters, we performed velocity analysis, and the resulting data (Fig. 3C) indicated changes in the contribution of npas4l reporter–expressing cells to the pronephron mesoderm (cluster 6) and skeletal muscle (clusters 2 and 3). The size of these three clusters was greatly expanded in npas4l^−/− embryos compared with npas4l^+/− siblings (Fig. 3D).

Fig. 3. Transcriptomic profiling of npas4l reporter–expressing cells in npas4l^+/− and npas4l^−/− embryos.

(A) Schematic of the experimental protocol. 5207 and 5262 npas4l reporter–expressing cells from the posterior half of 20 hpf npas4l^+/− and npas4l^−/− embryos, respectively, were sequenced using the 10x Genomics sequencing platform. FACS, fluorescence-activated cell sorting. (B) Uniform manifold approximation and projection (UMAP) representation of the data clustered by the Leiden algorithm. Numbers between parentheses indicate the percentage of heterozygous/mutant cells contributing to the cluster. (C) Velocity analysis done by comparing pre-mRNA and mRNA levels to infer the relation between the cells in a cluster. (D) Cluster composition per genotype. Graphical representation of the numbers displayed in (B). The main clusters increasing in npas4l^−/− compared with npas4l^+/− siblings are paraxial mesoderm (cluster 3), skeletal muscle (cluster 2), and intermediate mesoderm (cluster 6). (E) Dot plot with the top five markers separating the clusters. Pseudobulk expression data and marker genes used to separate the clusters can be found in tables S1 and S2, respectively. Differential gene expression in pseudo-bulk analysis can be found in table S3. The full and annotated dataset can be explored at https://bioinformatics.mpi-bn.mpg.de/20hpf_npas4l_expressing_cell or downloaded as an annotated data file using the link https://figshare.com/s/e6bdd5be14c7085d606c.

These data suggest that at 20 hpf, posterior npas4l reporter–expressing cells in npas4l mutants express LPM markers, as well as markers of pax2a-expressing LPM and pax3a-expressing paraxial mesoderm. Notably, the fli1a-expressing cells present in the posterior region of npas4l mutants (fig. S8) express genes associated with both pronephron and paraxial mesoderm, but not genes associated with hematopoietic development (differential gene expression analysis shown in table S3), indicating that unlike in npas4l^+/− embryos, these fli1a-expressing cells lack tal1 expression and thus hematopoietic potential. In addition, npas4l transcripts were only detected in the least differentiated cells of each of these lineages, indicating that npas4l is transiently activated in early progenitors and then quickly turned off as they differentiate (fig. S8), consistent with earlier observations (16, 21).

Pax2a is coexpressed with early endothelial and blood markers in a subset of the LPM

On the basis of the identity of npas4l-expressing cells in npas4l wild-type and mutant zebrafish, we hypothesized that some cells in the LPM coexpress markers associated with endothelial and pronephron progenitors. To test this hypothesis, we first assessed the expression of the pronephron gene pax2a and hematoendothelial reporters. pax2a mRNA and Pax2a protein are first detectable starting at 10 hpf, and their expression partially overlapped with that of a fli1a:GFP reporter in bilateral mesodermal stripes when the latter first became detectable at 12 hpf (Fig. 4A). We confirmed these observations by three-dimensional (3D) imaging of the expression of the drl:mCherry pan-LPM reporter and the tal1:EGFP hematoendothelial reporter together with Pax2a immunostaining (movie S1), as well as by live imaging of pax2a:EGFP; lmo2:dsRed embryos (movie S3). This analysis revealed the coexpression of the tal1 reporter and Pax2a in a subset of cells within the drl reporter–expressing LPM stripes. Together, these data indicate that pax2a is expressed in a subset of cells that, by marker expression, are endothelial and hematopoietic progenitors.

Fig. 4. fli1a and pax2a/Pax2a expression overlap during early stages of wild-type development and largely segregate before 14 hpf.

Confocal imaging reveals overlapping expression of fli1a:GFP and Pax2a (A), which decreases by 14 hpf (B), and is not detected at 18 hpf (C). Images of individual 0.74 μm optical sections (A′) clearly suggest colocalization. Yellow arrowheads point to double-positive cells. (D) Mapping of 10 to 18 hpf mesodermal cells from a single-cell time course dataset (39) suggests coexpression of endothelial markers and pax2a at 10 hpf, which decreases over time. PSM, presomitic mesoderm. (E) UMAP plots showing the expression of npas4l, pax2a, fli1a, tal1, lmo2, and etsrp; arrowheads point to the 10 hpf fli1a/pax2a double-positive cluster. Relevant clusters have been color-coded in the left plot in (E) and boxed with dotted lines in (D). “Pronephric duct” = pronephron tubule. Scale bars, 50 μm (A), 10 μm (A′), and 20 μm (B and C).

To further test the hypothesis that some cells in the LPM coexpress endothelial and pronephron markers, we carried out an analysis that relies on endogenous gene expression rather than on transgenic reporters. Fate mapping experiments have shown that some cells in gastrula stage embryos can give rise to endothelium, blood, and pronephron cells (11). Moreover, the coexpression of pronephron markers together with endothelial and hematopoietic markers in tailbud stage LPM has been documented in several zebrafish scRNA-seq datasets (4, 39). To investigate the time point at which the pronephron and endothelial lineages separate, we used the large zebrafish scRNA-seq dataset published by Wagner et al. (39), which provides a single-cell time course from pregastrulation stages to 24 hpf combined with a barcoding approach, thereby allowing lineage reconstruction. We extracted the 10, 14, and 18 hpf mesodermal cells and searched for coexpression of endothelial and pronephron markers. The expression of several early endothelial markers and of pax2a is displayed in a dot plot (Fig. 4D) and a uniform manifold approximation and projection (UMAP) plot (Fig. 4E). At 10 hpf, no distinct endothelial or pronephron clusters are observed but only one that coexpresses endothelial and pronephron markers. At the individual cell level, 44% (32 of 73) of all cells in this cluster coexpress fli1a and pax2a (table S4). It is important to note though that lowly expressed genes are often missed in droplet-based single-cell methods like 10x Genomics, potentially leading to an underestimate of the number of single- and/or double-positive cells. This fli1a/pax2a transcriptional overlap decreases over time as the endothelial and pronephron cells differentiate (table S4). To quantify these observations, we correlated fli1a expression with that of several pronephron markers in all mesodermal cells at 10, 14, and 18 hpf. This analysis indicates that fli1a expression indeed significantly correlates with that of several pronephron markers including pax2a (40), osr1 (41), foxj1a (42), and hnf1bb (43) (Table 1 and table S5). This correlation decreases significantly over time; osr1 and foxj1a stop showing a correlation by 14 hpf, and no significant correlation of fli1a expression with that of any pronephron markers is observed at 18 hpf.

Table 1. Endothelial and pronephron gene expression overlaps at 10 hpf, and this overlap decreases by 14 hpf and is no longer observed at 18 hpf.

Spearman correlation coefficient (ρ) between the expression of fli1a and the expression of four pronephron genes: pax2a, osr1, foxj1a, and hnf1bb; significant correlation is observed at 10 but not 18 hpf. A full list of gene expression correlations to fli1a expression and changes between the time points can be found in table S5. P values were adjusted to counteract the multiple testing problem.

Correlation of fli1a to:	ρ/P value at 10 hpf	ρ/P value at 14 hpf	ρ/P value at 18 hpf	Δρ/P value, 10 versus 14 hpf	Δρ/P value, 10 versus 18 hpf
pax2a	0.54/< 0.0001	0.38/< 0.0001	0.03/> 0.99	−0.16/0.00019	−0.51/< 0.0001
osr1	0.41/< 0.0001	−0.06/> 0.99	−0.07/> 0.99	−0.46/< 0.0001	−0.48/< 0.0001
foxj1a	0.50/< 0.0001	0.17/0.32	−0.03/> 0.99	−0.33/< 0.0001	−0.54/< 0.0001
hnf1bb	0.44/< 0.0001	0.27/< 0.0001	−0.02/> 0.99	−0.17/> 0.99	−0.46/< 0.0001

Table 2. Quantification of cell types observed after pax2a:CreERT2 induced recombination.

CNS, central nervous system; EC, endothelial cell.

pax2a:creERT2 x hsp70l:Switch + 4-OHT @ 10 hpf
GFP+ lineage-traced structure	Observed number of larvae with this structure
Kidney cells	20/20
CNS	20/20
Spinal cord neurons	20/20
Endothelial cells (for average number of ECs, see Fig. 5D)	19/20
Circulating blood cells	4/20

pax2a-expressing cells contribute to endothelium and blood

We next performed lineage labeling of pax2a-expressing cells to determine whether endothelial and hematopoietic cells can arise from this population. In zebrafish embryos, pax2a is prominently expressed in the otic placode, hindbrain, and spinal cord neurons, as well as in the pronephron progenitors (44). These kidney progenitors derive from a medial subset of the LPM, at times referred to as the intermediate mesoderm (4, 44, 45). To determine the fate of pax2a-expressing cells, we generated a pax2a:creERT2^tud53 knock-in line, in which the CreERT2-encoding sequence was inserted at the translational start codon of pax2a (Fig. 5A). Expression of the resulting CreERT2 transgene faithfully recapitulates endogenous pax2a expression (fig. S9) and enables 4-OH-tamoxifen (4-OHT)–dependent Cre activation in pax2a-expressing cells to recombine loxP-based reporter transgenes, resulting in the permanent labeling of pax2a-expressing cells and their descendants (46, 47). Crossing pax2a:creERT2 to hsp70l:loxP-STOP-loxP-EGFP (hsp70l:Switch; Fig. 5A) (48) and treating the embryos with 4-OHT between 10 hpf (one- to two-somite stage) and 24 hpf, we found enhanced GFP (EGFP) expression (i.e., lineage labeling) at 72 hpf in the descendant structures of the pax2a-expressing progenitors (49), including the diencephalon, rhombomeres 3 and 5, the otic-epibranchial progenitor domain, and podocytes, and along the entire pronephron tubule (fig. S10A). In 19 of 20 embryos exhibiting recombination, we also observed EGFP expression in endothelial cells of the intersomitic vessels (ISVs), dorsal aorta (DA), cardinal vein, and caudal hematopoietic territory (Fig. 5, B and C, Table 2). In 4 of 20 larvae, we also observed EGFP-expressing circulating cells that are likely erythrocytes (movie S2). Quantification of the data revealed that 4-OHT–induced recombination during early somitogenesis (starting at ~10 hpf) resulted in a median of 7.5 endothelial cells per larva exhibiting EGFP, whereas recombination during late somitogenesis (starting at ~16 hpf) resulted in a median of 4.5 endothelial cells per larva (Fig. 5, C and D). We also observed endothelial EGFP expression when combining pax2a:creERT2 with the endothelium-specific lmo2:loxP-dsRED-loxP-EGFP switch line (Fig. 5, E and F) (50), further showing that pax2a-expressing cells can contribute to the endothelium.

Fig. 5. Lineage tracing of pax2a-expressing cells labels endothelial cells.

(A) Schematic representation of the pax2a:cre:ERT2 knock-in line. pax2a:creERT2 zebrafish were crossed with hsp70l:Switch zebrafish, and the resulting embryos were treated with 4-OHT from 10 to 24 hpf or 16 to 36 hpf and heat-shocked at 72 hpf for 1 hour. TS, transcriptional start. Several pax2a:creERT2 switched EGFP-positive endothelial cells (arrows) can be observed in zebrafish treated with 4-OHT from 10 (B) or 16 (C) hpf. Arrowheads point to pronephron cells, and asterisks indicate spinal cord neurons. Recombined circulating blood cells can be seen in movie S2. 5′UTR, 5′ untranslated region. (D) Quantification shows a decrease in the number of recombined endothelial cells in the animals treated starting at 16 hpf compared with those treated starting at 10 hpf. (E) Schematic representation of the pax2a:creERT2-mediated recombination of an lmo2:Switch line. (F) Recombined endothelial cells (arrow) are observed in the trunk and tail of lmo2:Switch larvae. Scale bars, 500 μm.

Etsrp, Tal1, and Lmo2 regulate distinct aspects of endothelial development

Npas4l promotes the expression of the three transcription factor genes etsrp, tal1, and lmo2 (21). We hypothesized that the reduction in expression of these transcription factors in npas4l mutants plays a role in the contribution of npas4l reporter–expressing cells to the pronephron tubules and skeletal muscle. To test this hypothesis, we generated mutant lines for these three genes (Fig. 6A and fig. S11) and analyzed their phenotype in endothelial reporter lines and in the npas4l reporter background. These newly generated mutant alleles exhibit all previously published phenotypes (23, 25, 30, 33, 51), including the vascular defects (fig. S12). We counted the number of npas4l/Pax2a double-positive cells in the presumptive pronephron tubules of npas4l, etsrp, tal1, and lmo2 mutants at 24 hpf and observed an increase in npas4l, tal1, and lmo2 mutants but not in etsrp mutants (Fig. 6, B and E, and fig. S13). We observed the same upward trend in the overall number of Pax2a-expressing cells in the presumptive pronephron tubules of npas4l, tal1, and lmo2 mutants, indicating a physiological change rather than a marker misexpression (Fig. 6C). In addition, we also observed that some npas4l reporter–expressing cells in the PCV of lmo2 mutants were Pax2a positive (fig. S14), indicating a potential transition from an endothelial toward a pronephron fate. Conversely, the number of npas4l reporter–expressing skeletal muscle cells was increased in npas4l and etsrp mutants but not in tal1 or lmo2 mutants (Fig. 6D). As an interesting side observation, double mutants for npas4l/etsrp and etsrp/tal1 do not develop any endothelial cells as indicated by the lack of fli1a:EGFP expression (fig. S15). To our knowledge, this is the first time that a vertebrate model completely devoid of endothelial cells has been generated. Together, these data indicate that the contribution of npas4l-expressing cells to pronephron tubules is blocked by the npas4l targets/effectors tal1 and lmo2, whereas their contribution to skeletal muscle is blocked by etsrp.

Fig. 6. Endothelial progenitors exhibit distinct defects in npas4l, etsrp, tal1, and lmo2 mutants.

(A) Generation of etsrp, tal1, and lmo2 mutants containing in-frame or out-of-frame indels. aa, amino acid. (B to D) Numbers of (B) npas4l reporter and Pax2a double-positive cells, (C) all Pax2a-expressing cells, and (D) npas4l reporter–expressing skeletal muscle cells, per field of view (319.45 μm long area over the yolk extension). Cells were counted in 3D confocal lateral views at 24 hpf. Data are represented as individual data points, median, interquartile range, and extremes excluding outliers. P values were calculated by Poisson regression and adjusted to counteract the multiple testing problem. A subset of these data is also shown in Fig. 2 (C to E). (E) Transverse sections of npas4l^+/−, npas4l^−/−, etsrp^−/−, tal1^−/−, and lmo2^−/− embryos at 24 hpf; endothelial progenitors and endothelial cells outlined by yellow dotted lines. Major axial vessels form in npas4l^+/− embryos, while in npas4l^−/− embryos, endothelial progenitors fail to migrate or to express the fli1a:GFP reporter. In etsrp^−/− embryos, endothelial progenitors migrate but fail to form vessels. In tal1^−/− and lmo2^−/− embryos, endothelial progenitor migration is partly impaired; the endothelial progenitors that do reach the midline differentiate and express fli1a:GFP. Scale bars, 20 μm.

tal1 mRNA injections rescue endothelial development in npas4l^−/− embryos in an Etsrp-dependent manner

To better understand the relationship between Npas4l and its effectors, we tested whether injecting etsrp, tal1, or lmo2 mRNA would rescue different aspects of the npas4l mutant phenotype. While injections of etsrp or lmo2 mRNA did not have a noticeable effect on endothelial development in npas4l mutants, injections of tal1 mRNA were sufficient to significantly restore ISV formation (Fig. 7, A, B, and F). As Npas4l has several distinct transcriptional effectors, such a strong rescue of ISV formation by just one of them was unexpected. In these rescued mutants, we did not detect the bilateral population of npas4l reporter–expressing pronephron tubule cells, but the number of ectopic npas4l reporter–expressing muscle cells was not reduced compared with uninjected npas4l mutants (Fig. 7F). In addition, the finding that etsrp or lmo2 mRNA injections did not rescue npas4l mutants is consistent with the fact that etsrp and lmo2, but not tal1, are expressed in endothelial progenitors in npas4l^−/− embryos (fig. S15B). Also, tal1 mRNA injections into etsrp mutants did not rescue ISV formation, indicating that the tal1-mediated rescue of the npas4l mutant phenotype (which itself includes the loss of etsrp expression) depends on Etsrp function (Fig. 7C). These data also suggest that tal1 promotes etsrp expression. To test this hypothesis, we used reverse transcription quantitative polymerase chain reaction (RT-qPCR) to measure tal1 and etsrp mRNA levels at 10 hpf after npas4l, etsrp, tal1, and lmo2 mRNA injection into wild-type embryos (Fig. 7, D and E). We found that tal1 could only be induced by Npas4l at this stage, while etsrp could be induced by both Npas4l and Tal1. Together, these data show that the pronephron tubule contribution of npas4l reporter–expressing cells in npas4l mutants can be blocked by injection of tal1 mRNA, which also leads to an increase in endothelial cells. These experiments provide further evidence for a close relationship between the endothelial and pronephron lineages and indicate that Tal1 and Lmo2 modulate the endothelial versus pronephron fate decision downstream of Npas4l (Fig. 8).

Fig. 7. tal1 overexpression rescues endothelial development in npas4l mutants in an Etsrp-dependent manner.

(A) npas4l and tal1 mRNA injections rescue ISV formation in npas4l mutants. (B) Quantification of rescue: mRNA injections of npas4l and tal1, but not of etsrp or lmo2, rescue the ISV defects in npas4l^−/− embryos. Data are represented as individual data points, median, interquartile range, and extremes excluding potential outliers. P values were calculated by Poisson regression and adjusted for multiple comparisons between six groups using the Bonferroni correction. (C) tal1 mRNA injections do not rescue the vascular phenotype in etsrp^−/− embryos. Data are represented as individual data points, median, interquartile range, and extremes excluding outliers. P values were calculated by Poisson regression. (D and E) etsrp mRNA injections do not increase tal1 mRNA levels at the tailbud stage, whereas tal1 mRNA injections lead to increased etsrp mRNA levels at the same stage. Data are represented as individual data points as well as means ± SD. P values were calculated by unpaired two-sample t test and adjusted for multiple comparisons between five groups using the Bonferroni correction. (F) Uninjected and tal1 mRNA-injected npas4l^+/− and npas4l^−/− embryos. Ventrolateral npas4l reporter–expressing cells (arrowheads) are not present in tal1 mRNA-injected npas4l^−/− embryos, while the contribution of npas4l reporter–expressing cells to skeletal muscle (M) does not appear to be blocked. Scale bars, 500 μm (A) and 50 μm (F). P values were adjusted to counteract the multiple testing problem.

DISCUSSION

The specification of endothelial cells is an early step in cardiovascular development, and the transcriptional effectors required for endothelial specification in zebrafish are induced by the bHLH-PAS transcription factor Npas4l (21). npas4l^−/−embryos lack most endothelium and blood (16, 17, 21); however, the fates of npas4l-expressing cells in the absence of Npas4l function remain unclear. Here, we generated npas4l knock-in reporter alleles to track npas4l reporter–expressing cells in wild-type and mutant embryos. We report that in npas4l^−/−, tal1^−/− and lmo2^−/− embryos, npas4l reporter–expressing cells contribute to the pronephron tubules, and that in npas4l^−/− and etsrp^−/− embryos, npas4l reporter–expressing cells contribute to skeletal muscle. These data indicate that Tal1/Lmo2 and Etsrp modulate endothelial development downstream of Npas4l in different ways. Building upon these initial observations, we provide evidence for a population of early LPM cells that coexpresses endothelial and pronephron markers. While most of these cells contribute to hematoendothelial or pronephron lineages in wild-type embryos, more of them commit to the pronephron lineage in the absence of Npas4l, Tal1, or Lmo2.

Tal1/Lmo2 and Etsrp drive distinct cell fate decisions

In tal1 and lmo2 mutants, npas4l reporter–expressing cells remain in a ventrolateral position and contribute to the pronephron tubules. In contrast, in etsrp mutants, npas4l reporter–expressing cells migrate to the midline, but most of them fail to acquire endothelial characteristics. npas4l mutants, in which etsrp, tal1, and lmo2 expression is strongly reduced, display an almost complete absence of endothelial cells. Notably, embryos lacking both Etsrp and Tal1 function do not develop any endothelial cells, further demonstrating that these transcription factors are indispensable for endothelial development.

The etsrp loss-of-function phenotypes that we observed are in line with those reported by Chestnut et al. (25) who applied a reporter-based strategy to determine the fate of etsrp-expressing cells in etsrp mutants; in the absence of Npas4l or Etsrp function, endothelial progenitors become skeletal muscle cells. In mouse, chicken, and zebrafish, the paraxial mesoderm has been shown to be an additional source of endothelial cells (52–55). In zebrafish, the paraxial mesoderm has also been reported to be a bipotent cell population that can give rise to hematopoietic stem and progenitor cells as well as muscle progenitors (53, 56). Together, these data point to a close relationship between endothelial cells and the paraxial mesoderm.

As an increased number of pronephron tubule cells has not been reported in tal1 or lmo2 mutants before, we focused most of our attention on this phenotype. Defects in endothelial lineage specification leading to an increase in the number of pronephron tubule cells have previously been observed in several other contexts including in hand2 mutants (57) and in tbx16 mutants (58). Both of these mutants, in fact, display a reduction in tal1 expression (57, 58). Conversely, the loss of the Osr1 transcription factor (59, 60), or overexpression of tal1 (61) or hand2 (57), causes a reduction in the size of the pax2a-expressing LPM territory and a concomitant increase in the number of endothelial cells. It is thus possible that the alteration in tal1 expression contributes to the observed pronephron tubule phenotypes in these models. Together, these data indicate that Tal1, Hand2, Osr1, and Tbx16 play a role in the fate decision between endothelial/blood cells and pronephron tubule cells. Looking more closely at the acquisition of a pronephron fate by npas4l reporter–expressing cells in npas4l mutants, we did not observe an obvious increase in Pax2a expression at 14 hpf. However, in wild-type embryos, npas4l and fli1a reporter–expressing cells at 14 hpf usually express a lower level of Pax2a than pronephron cells (Fig. 4, A to A′). The loss of Npas4l, Tal1, or Lmo2 function and subsequent loss of endothelial marker expression may not necessarily elevate Pax2a levels at early stages. Instead, it could allow cells to contribute to the pronephron lineage despite low initial pronephron marker expression. As Pax2a expression is maintained in differentiated cells, the high marker expression that we observed at 24 hpf could be associated with maturation and not differentiation.

The direct comparison of tal1 and etsrp mutants revealed distinct phenotypes, and the injection of etsrp mRNA failed to induce tal1 expression, indicating that in zebrafish, tal1 expression is not induced by Etsrp. These observations are different from data in mouse that indicate that Tal1 is a direct target of ETV2 (62). Also, mouse Etv2 mutants, unlike zebrafish etsrp mutants, fail to form Gata1a-expressing erythroid progenitors (63). The mouse Etv2 mutant phenotype is more similar to the zebrafish npas4l mutant phenotype than to the zebrafish etsrp mutant phenotype. These results support the hypothesis that in mammals, Etv2 acquired the functions of Npas4l, potentially contributing to its loss (64). Thus, the assumption of functional equivalence between zebrafish Etsrp and mouse Etv2 requires reconsideration.

Endothelial and pronephron cells can come from the same progenitor population

The phenotypes observed in npas4l, tal1, and lmo2 mutants indicate that endothelial progenitors can acquire a pronephron tubule fate, leading to the hypothesis that the endothelial and pronephron lineages are closely related. To test this hypothesis, we investigated the lineage relationship between endothelial and pronephron cells in wild-type embryos.

Although classical LPM models divide this structure into distinct fate territories, an overlap of pax2a, pax8, tal1, and lmo2 expression in the LPM has been previously reported in Xenopus laevis (65) and hinted at in zebrafish (4, 12). Warga et al. (11) documented the emergence of pronephron, endothelial, and hematopoietic cells in single-cell clones by fate mapping experiments in zebrafish, and Spanjaard et al. (66) expanded these observations through scRNA-seq combined with barcoding-mediated lineage tracing, also in zebrafish. The results by Warga et al. (11) suggest that the pronephron and hematoendothelial lineages start to segregate at midgastrulation stages (5 hpf). However, according to our analysis of a published scRNA-seq dataset of 10 hpf zebrafish embryos (39), no distinct endothelial or pronephron clusters, but only one cluster coexpressing endothelial and pronephron markers, can be observed at this stage.

In addition to relying on the anatomical location of the pronephron tubules and their progenitors, we used pax2a/Pax2a expression as a proxy for the pronephron lineage (40, 67, 68) and further validated our findings by costaining with mature pronephron tubule and glomerulus markers. Our imaging and expression analyses revealed that the expression of pax2a, an early expressed transcription factor gene involved in several steps of midbrain-hindbrain boundary and kidney formation (40, 67, 68), is also observed in hematoendothelial progenitor cells in the LPM during early somite stages. Consistent with these observations, we found pax2a:creERT2-labeled endothelial cells throughout somitogenesis, with the highest efficiency when the 4-OHT treatment started at 10 hpf. These lineage tracing, coexpression, and published fate mapping data, as well as the npas4l/tal1/lmo2 mutant phenotypes, suggest the existence of a progenitor population in the tailbud/early-somitogenesis LPM that can contribute to endothelial, hematopoietic, and pronephron cells. It is not clear though how many of the pronephron cells once expressed npas4l. Furthermore, because both the LPM (2, 9) and paraxial mesoderm (52–55) have been found to display high vasculogenic potential, it is not clear whether the mesoderm in general has high vasculogenic potential or whether there are specific progenitor cells for different types of mesoderm and endothelium. Specifically, it will be interesting to further investigate cells coexpressing endothelial and pronephron markers, or “renangioblasts,” in various developmental contexts. Recently, a distinct population of cells has been described to lie very close to the pronephron tubules, turn on endothelial markers, and integrate into the existing vasculature (69), and it will be worth further investigating the origin of these cells. Note also that endothelial cells have been reported to arise during the in vitro differentiation of embryonic stem and induced pluripotent stem cells (also known as iPS cells or iPSCs) into kidney structures (70) and that human metanephric mesenchymal cells can develop into hematopoietic stem cells when transplanted into sheep (71). Furthermore, Tal1 is expressed in the mouse embryonic kidney, especially between embryonic day 13 (E13) and E17 (72), and fate mapping experiments in wild-type mice have revealed a contribution of Osr1-expressing cells to the endothelium (73).

Together, our data, in combination with previous observations (4, 11, 12, 65, 66, 70–74) indicate that some endothelial and pronephron cells share a common progenitor pool. Detailed knowledge of lineage decisions and alternative fates during endothelial development is instrumental when generating endothelial cells in a therapeutic context or when generating organoids consisting of heterogeneous cell populations. Furthermore, kidney cells that dedifferentiate during tumorigenesis could become a source of endothelial cells. To extend our work in zebrafish, future investigation of Tal1/Lmo2 as potential modulators of the fate decision between the endothelial and pronephron tubule lineages in mammals is warranted.

MATERIALS AND METHODS

Zebrafish husbandry and lines

Zebrafish were kept according to Federation of European Laboratory Animal Science Associations guidelines (FELASA). Animal procedures performed at the Max Planck Institute for Heart and Lung Research conform to guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes and were approved by the Animal Protection Committee (Tierschutzkommission) of the Regierungspräsidium Darmstadt (references: B2/1041 and B2/1218). Embryos and larvae were raised under standard conditions.

Animal procedures performed at the University of Colorado Anschutz Medical Campus were approved by the Institutional Animal Care and Use Committee (IACUC) of University of Colorado Anschutz Medical Campus under animal protocol 00979. Zebrafish were housed and provided with veterinary care by the University of Colorado Anschutz Medical Campus animal facility. The University of Colorado policy provides information on the housing and care of zebrafish at CU Anschutz, per guidance from the National Institutes of Health (NIH) and the Office of Laboratory Animal Welfare (OLAW). Animal procedures performed at the TU Dresden were conducted according to the guidelines and under supervision of the Regierungspräsidium Dresden (permit: TVV 21/2018). All efforts were made to minimize animal suffering and the number of animals used. Lines used in this study include cloche^m39 (17), npas4l^bns297 (21), Tg(UAS:GFP)^nkuasgf1a (75), Tg(UAS:nfsB-mCh)^c264 (76), Tg(fli1a:GFP)^y1 (77), Tg(fli1a:nls-GFP)^y7 (78), Tg(wt1b:EGFP)^li1 (79), Tg(drl:mCherry)^zh705 (80), Tg(hsp70l:Switch)^zh701 (48), Tg(lmo2:LOXP-DsRed-LOXP-EGFP) (abbreviated as lmo2:dsRED) (50), Tg(tal1:EGFP)^sq1 (81), and Tg(pax2.1:EGFP)^e1 (44).

Genome editing for knock-ins and knockouts using CRISPR-Cas9 technology

N₂₀-NGG sequences in the region of interest were selected manually and checked for off-targets using the CRISPR design tool CHOPCHOP v2 (https://chopchop.cbu.uib.no/) on the zebrafish GRCz10 genomic assembly. For knockout generation, sgRNA (single guide RNA) templates were in vitro transcribed using the MEGAshortscript-T7 Kit (Ambion, Austin, Texas), followed by purification on an RNA-cleanup column (Biozym, Hessisch Oldendorf, Germany). The guide RNA (gRNA) used to generate the Gal4 knock-in reporter was ordered as crRNA/tracrRNA duplex (Integrated DNA Technologies, Coralville, Iowa) and assembled according to the user method provided by the Essner laboratory (https://sfvideo.blob.core.windows.net/sitefinity/docs/default-source/user-submitted-method/crispr-cas9-rnp-delivery-zebrafish-embryos-j-essnerc46b5a1532796e2eaa53ff00001c1b3c.pdf?sfvrsn=52123407_10). The resulting sgRNAs or gRNAs were coinjected with Cas9 mRNA (300 pg) in a 2-nl injection volume with 20% phenol red into zebrafish zygotes.

To generate the knock-in reporter npas4l^{Pt(npas4l-p2A-Gal4-VP16)bns313}, henceforth named npas4l⁺-reporter, the 3′ end of the npas4l coding sequence was targeted using the gRNA 5′-CCAGAGCCACTGCTGGACGA(−GGG)-3′ [synthetic crRNA (CRISPR RNA), Integrated DNA Technologies]. The transgenic donor was inserted after the codon for the endogenous Npas4l amino acid L625, thus truncating the Npas4l protein sequence by 22 amino acids. The pGTag-p2A-Gal4/VP16 donor was cloned as described in (37) using oligos 5′-GCGGtttCACACACTCTCACCACACCTCCTGCTGGGCACCAGAGCCACTGCTGGAc-3′ (sense) and 5′-ATCCgTCCAGCAGTGGCTCTGGTGCCCAGCAGGAGGTGTGGTGAGAGTGTGTGaaa-3′ (antisense) to assemble the 5′-homology arm and 5′-AAGCGAGGGCATCATCGACAGCATCCTCAGAGAGCTGGACAGCACACACACccc-3′ (sense) and 5′-CGGgggGTGTGTGTGCTGTCCAGCTCTCTGAGGATGCTGTCGATGATGCCCTCG-3′ (antisense) to assemble the 3′-homology arm. Efficiency of the gRNA was tested by high-resolution melt analysis (HRMA) (forward: 5′-CGCACGACACACACACTC-3′, reverse: 5′-TGTGCTGTCCAGCTCTCTGA-3′) on an Illumina-Eco machine (Illumina, San Diego, California) using the DyNAmo ColorFlash SYBR Green qPCR Mix (Thermo Fisher Scientific, Waltham, Massachusetts). Knock-in efficiency was estimated by injections into the Tg(UAS:GFP)^nkuasgfp1a background followed by screening for GFP expression.

Founders were screened by outcrossing the injected individuals and screening for GFP expression in at least 300 embryos. Precise 5′ insertion of the cassette in F₁ zebrafish was confirmed by PCR on gDNA and complementary DNA (cDNA) (forward: 5′-CTTGGTCCCTGCTGTGTTCT-3′, reverse: 5′-CAGTCTTTCTAGCCTTGATTCCAC-3′). PCR analysis for the vector backbone of the pGTag donor plasmid in the bns313 allele indicated vector concatemer insertions; the 3′ insertion sequence at the genomic DNA level has not yet been fully determined. However, the cDNA sequence from exon 1 to the end of the Gal4-reporter was verified by Sanger sequencing, and the distal gene tmem264 is still present in the bns313 allele (fig. S1).

The mutant version of the npas4l⁺ reporter npas4l^{Pt(+36bp-npas4l-p2A-Gal4-VP16)bns423}, henceforth named npas4^bns423-reporter, was generated by inducing in the npas4l^bns313 allele an in-frame indel in the domain encoding the DNA binding bHLH domain to retain the reading frame and reporter expression but abolish Npas4l transcriptional activity using the sgRNAs 5′-CCGCGCCTTAGATGCTCCTT-3′ and 5′-CTCCACACTCTTCCTGATGT-3′. Injected embryos and potential founders were screened by PCR (forward: 5′-CTCTGTTCTGCTGGTGATCTGC-3′, reverse: 5′-ATGCGTTGTGGATGCTCTCC-3′). We recovered a +36-bp insertion at the end of exon 2 that caused a strong phenotype, yet correct splicing as determined by PCR spanning the coding sequence from exon 1 to the Gal4 insertion.

For the generation of the pax2a:CreERT2^tud53 knock-in transgenic line, genomic DNA from wild-type AB was used to generate the pax2a bait by PCR (Phusion Polymerase, Thermo Fisher Scientific) using primers pax2a-bait-for (5′-GACAACGTTGTAGGCTACTACTAATTAACGACAC-3′) and pax2a-bait-rev (5′-GATAATCGACTGAGGTCGCCGTCTCGCCT-3′). The amplified 904-bp bait fragment was cloned into a pCS2+ vector containing the zebrafish codon-usage-optimized CreERT2 sequence. The CMV promoter was later removed from the pCS2+ vector, and the construct was verified by sequencing. The sgRNA designed for the pax2a locus had the sequence 5′-GGGGGGATCTGGGAAGGAGG(−GGG)-3′. Preparation of sgRNA and Cas9 mRNA and injection into one-cell stage embryos were performed according to standard protocols. The injected embryos were monitored for the next 5 days, and approximately 100 embryos were raised to adulthood. To identify founders, 4- to 6-month-old F₀ fish were outcrossed with wild-type strains, and 50 embryos from each clutch were used to isolate genomic DNA. Subsequently, PCR was carried out using the primer pair pax2a-int-for (5′-GGGAAATCAACATAAAAACATCCGACATCAATACC-3′) and Cre-5′-rev (5′-TGACTTCATCGCTGGTAGCGTCC-3′). An 1155-bp amplicon indicated a correctly oriented knock-in at the targeted locus, and PCR products from individual embryos were verified by sequencing. The knock-in strain was maintained as an outcross to reduce the general effects of inbreeding.

For the generation of tal1 mutant alleles, we targeted the region encoding the DNA binding bHLH domain in the fourth exon using the sgRNA 5′-CAAGAACGAGATCCTGCGTC-3′. Injected embryos and potential founders were screened by HRMA (forward: 5′-CAAGAAACTCAGCAAGAACGAGATC-3′, reverse: 5′-GTCCTGGTCGTTGAGGAGCT-3′). We recovered the −6 bp (in-frame) deletion allele bns498 leading to the loss of R228 and L229 in the second helix of the HLH-motif and the −7 bp (out-of-frame) deletion allele bns497 leading to a premature termination codon after R228 that is predicted to truncate the protein sequence by 97 amino acids.

For the generation of etsrp mutant alleles, we targeted the region encoding the DNA binding ETS domain in the sixth exon using the sgRNA 5′-AAGTTGGACTGGTGATGGCT-3′. Injected embryos and potential founders were screened by HRMA (forward: 5′-AGCTCTGGCAGTTTCTGCTAG-3′, reverse: 5′-CTCAGCGGGATCTGACATTTTAAAC-3′). We recovered the −9 bp (in-frame) deletion allele bns426 leading to the loss of D265 G266 and W267 and the −4 bp (out-of-frame) deletion allele bns422 after G264 leading to an altered peptide sequence AGSLKCQIPLRWRSGGASVKTSLK* followed by a premature termination codon at position 289 that is predicted to truncate the protein sequence by 78 amino acids.

For the generation of lmo2 mutant alleles, we targeted the region encoding the beginning of the second LIM domain in the third exon using the sgRNA 5′-TTCCTGTGAAAAGAGGATCC-3′ with the aim to split the two conserved LIM domains of this scaffolding protein. Injected embryos and potential founders were screened by HRMA (forward: 5′-TCCTTTCAGACTGTTTGGTC-3′, reverse: 5′-GCACACGCATGGTCATTTCAAAG-3′). We recovered the −6 bp (in-frame) deletion allele bns500 leading to the replacement of I101 R102 A103 with T101 and the +2 bp (out-of-frame) indel allele bns499 after R100 leading to an altered peptide sequence TTGPLK* followed by a premature termination codon at position 107 that is predicted to truncate the protein sequence by 52 amino acids.

As predicted from the domain annotations, the lmo2 out-of-frame allele bns499 exhibits a stronger phenotype than the lmo2^bns500 in-frame allele does and was therefore used exclusively. The in-frame and out-of-frame alleles generated for tal1 and etsrp, however, were phenotypically indistinguishable. Therefore, we worked exclusively with the tal1^bns498 and etsrp^bns426 in-frame alleles to minimize the possibility of transcriptional adaptation (82).

npas4l alleles and genotype of the embryos analyzed

We crossed the bns423 allele to the bns297 allele to generate the npas4l mutant embryos shown in several figures (Figs. 1C, 2, 3, and 7 and figs. S3, S5, S6, S13, and S14) as they exhibit the null phenotype and can be genotyped by HRMA (21). Figures 1B (npas4l^−/−) and 6 as well as figs. S4 and S15 display embryos from intercrosses of bns423 hets. To generate data for the etsrp, tal1, and lmo2 mutants, we used a single bns313 reporter allele. The npas4l^+/+ embryos in Fig. 1B and figs. S1B and S2 are from intercrosses of bns313 hets.

scRNA-seq sample preparation and data analysis

Trunks from ~150 npas4l heterozygous (npas4l^bns423/+) and trans-heterozygous (npas4l^{bns423/bns297}) embryos were cut at the anterior end of the yolk extension at 20 hpf in DMEM/F10 + 5% fetal bovine serum + 0.01% tricaine on agarose-coated plates. The cells were dissociated using the Pierce Cardiomyocyte Dissociation Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions and sorted for GFP fluorescence using an FACSAria III sorter (BD Biosciences, San José, California) and DAPI (4′,6-diamidino-2-phenylindole) as an indicator for dead cells. The cell suspensions were counted with a Moxi cell counter (ORFLO Technologies, Ketchum, Idaho) and diluted according to the manufacturer’s instructions to obtain 5000 single-cell data points per sample. Each sample was run separately on one lane in a Chromium controller with Chromium Next GEM Single Cell 3′ Reagent Kits v3.1 (10x Genomics, Pleasanton, California).

scRNA-seq library preparation was done following standard protocol. Sequencing was done on a NextSeq 500 (Illumina), and raw reads were aligned against the zebrafish genome (DanRer11) and counted by StarSolo (https://github.com/alexdobin/STAR) followed by secondary analysis in the Annotated Data Format. Preprocessed counts were further analyzed using the Scanpy software (https://github.com/theislab/scanpy). Basic cell quality control was conducted by taking the number of detected genes and mitochondrial content into consideration. We removed only 16 cells in total that did not express between 1000 and 7000 genes or had a mitochondrial content of less than 6%. Furthermore, we filtered genes if they were detected in less than 30 cells (<0.3%). Raw counts per cell were normalized to the median count over all cells and transformed into log space to stabilize variance. We initially reduced dimensionality of the dataset using principal components analysis (PCA), retaining 50 principal components. Subsequent steps, like low-dimensional UMAP embedding and cell clustering via community detection, were based on the initial PCA. Final data visualization was done using the scVelo (https://github.com/theislab/scvelo) and cellxgene (https://github.com/chanzuckerberg/cellxgene) packages. A trajectory inference was calculated using partition-based graph abstraction and displayed as force-directed graph calculated using ForceAtlas2 as implemented in Scanpy.

For the reanalysis of the dataset originally published by Wagner et al. (39), we relied on an annotated data file mapped to the latest zebrafish genome assembly (GRCz11) that was provided by the authors and analyzed using Scanpy. After creating subsets of 10 to 18 hpf mesodermal cells using the original cell type and time point annotations provided in the dataset, we normalized the raw counts per cell to the median count over all cells and log-transformed the data. We reduced the dimensions to 50 principal components as described in the paragraph above and visualized the data in UMAP and dot plots. For the correlation analyses, we calculated a nonparametric Spearman correlation between a bait gene and all other genes over all single cells. For comparisons between different correlations, we performed a Fisher transformation.

mRNA synthesis and microinjections

We used the published pCS2-npas4l plasmid (Addgene plasmid no. 164654) to synthetize npas4l mRNA. Coding sequences of tal1, etsrp, and lmo2 were cloned into pCS2z vectors (Addgene plasmid no. 62214), deposited as Addgene plasmids no. 164655 to 164657, and in vitro transcribed using the SP6 mMessage mMachine Kit (Ambion). The mRNAs were injected in a 2-nl volume with 20% phenol red into zebrafish zygotes. Different amounts of mRNA were injected to determine effective doses with minimal embryonic lethality or deformation. The optimal concentrations for rescue experiments were 5 pg for npas4l, tal1, and etsrp and 25 pg for lmo2.

CreERT2-based lineage tracing

Lineage tracing experiments were performed by crossing male pax2a:creERT2 with female hsp70l:Switch or lmo2:dsRED animals. Embryos were induced using preheated (65°C for 10 min) 4-OHT (H7904; Sigma-Aldrich H7904) at a final concentration of 10 μM in E3 embryo medium at the indicated time points. Embryos were washed in fresh E3 medium at 24 hpf and 1-phenyl-2-thiourea (Sigma-Aldrich) was added to a final concentration of 0.003% to inhibit pigment formation. To induce EGFP transcription in hsp70l:Switch embryos, the embryos were incubated at 37°C for 1 hour at 68 hpf and 4 hours before imaging. Larvae were imaged using a ZEISS LSM 880 confocal microscope or a ZEISS Z.1 light-sheet microscope (ZEISS W Plan-Apochromat 20×/0.5 numerical aperture objective). For imaging, 72 hpf larvae were treated with 0.016% ethyl 3-aminobenzoate methanesulfonate salt (Tricaine, Sigma-Aldrich) in E3 and mounted in 0.5% low melting point agarose (LMA; A1801-LM, Benchmark Scientific) for confocal microscopy and 1.0% LMA for light-sheet microscopy. Z-stack maximum projections were made using Fiji.

Histology and microsections

Embryos were fixed in 4% paraformaldehyde (PFA) solution overnight, washed 3× with phosphate-buffered saline (PBS) and embedded in gelatin. Briefly, the embryos were incubated at 4°C in 30% (w/v) sucrose in PBS overnight. The embryos were then incubated for 1 hour at 37°C in 7.5% (w/v) gelatin and mounted afterward in the same solution. The tissue blocks were frozen in liquid nitrogen–cooled isopentane and stored at −80°C. The blocks were sectioned using a CM3050S cryostat (Leica, Wetzlar, Germany) and the tissue sections stored at −20°C.

Immunostaining

Immunostaining was performed according to standard protocols, with the following parameters: For whole-mount stainings, embryos were fixed overnight in 4% PFA at 4°C. Early embryos (≤20 hpf) were dechorionated after fixation, older embryos were dechorionated before fixation. Embryos were then permeabilized for 3 min with Proteinase K (10 μg/ml) and blocked in PBS + 5% goat serum + 0.1% Triton X-100.

For the staining with the a6F antibody, embryos were fixed overnight in Dents fixative [80% MeOH and 20% dimethyl sulfoxide (DMSO)] at 4°C. After stepwise rehydration, embryos were blocked in PBS + 10% goat serum + 1% DMSO + 0.1% Tween 20.

drl:mCherry zebrafish were crossed with tal1:EGFP zebrafish, and the resulting embryos were dechorionated and fixed at the indicated time points with 4% formaldehyde, 0.1% Triton X-100 in PEM (0.1 M Pipes, 2 mM MgSO₄, and 1 mM EDTA) overnight at 4°C. The next day, embryos were washed in 1× PBS with 0.1% Triton X-100 and 0.1% bovine serum albumin, permeabilized in 1× PBS with 0.5% Triton X-100 for 30 min, and blocked in PBS + 1% goat serum and 0.1% Triton X-100.

For the staining of gelatin sections, the gelatin was removed from the slides by incubation in PBS for 10 min at 37°C. The sections were permeabilized for 10 min at room temperature in PBS + 0.5% Triton X-100 and blocked in PBS + 5% goat serum + 0.1% Triton X-100.

We used the following primary antibodies: chicken anti-GFP (1:500; Aves Labs, Tigard, Oregon), rabbit anti-dsRed (1:500; Takara Bio, Kusatsu, Japan), mouse anti-mCherry (1:500; Takara Bio), and rabbit anti-Pax2a (1:200; GeneTex, Irvine, California). The a6F monoclonal antibody (anti–Na⁺‑ and K⁺-dependent adenosine triphosphatase; 1:100) developed by D.M. Fambrough was obtained from the Developmental Studies Hybridoma Bank, created by the National Institute of Child Health and Human Development of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242. Alexa fluorophore–conjugated secondary antibodies (Thermo Fisher Scientific) were used in a 1:500 dilution. DAPI was added to the secondary antibody solution (final concentration of 1 μg/ml).

In situ hybridization

Probes corresponding to the full coding sequence of pax2a and CreERT2 were used. Probe synthesis and in situ hybridization were performed according to standard protocols.

Confocal microscopy imaging

Embryos were embedded in 1% low melting point agarose on their side. Living embryos were anesthetized with 0.01% tricaine before embedding and stayed under anesthesia during the procedure. All experiments on living embryos and larvae were nonrecovery experiments. For genotyping, the anesthetized embryos were taken out of the agarose, exposed to heat briefly, and lysed using 50 mM NaOH for 10 min at 95°C.

Confocal images were acquired using an LSM 800, LSM 880, or LSM 710 confocal microscope (ZEISS, Oberkochen, Germany). The images were acquired and processed using the ZenBlue software package. Only linear adjustments were used, and acquisition parameters were kept constant throughout the imaging whenever possible. The confocal microscopy data presented in this manuscript were not used for the quantification of fluorescence intensity.

For overview images of the whole embryo, a tile scan with a Plan-Apochromat 10×/0.45 DIC II objective (ZEISS) was performed and stitched with ESID (electronically switchable illumination and detection module) as the reference channel. Bright field–like images for this magnification were generated using an ESID channel and enhanced depth of focus. The channel was then added to the orthogonal projection of the fluorescence channels.

Images of the trunk region were acquired using an LD LCI Pln Apo 25× 0.8 W (ZEISS) or a C Apo 40×/1.1 W DICIII (ZEISS) lens on an LSM 800 observer or Pln Apo 40×/1.3 oil DIC M27 or 40×/1.2 immersion-corrected DIC M27 lens on an LSM 710. As an anatomical landmark, we kept the yolk extension in the field of view. For the presentation of the bright field–like ESID channel, a single plane was exported from the middle of the stack and added to the orthogonal projection of the fluorescence channel.

Sections were imaged using a Pln Apo 40×/1.4 Oil DIC II (ZEISS) lens and the Airyscan detector, followed by 2D Airyscan processing. Similar-looking transverse sections over the yolk extension were used, but the location along the anterior-posterior axis varies slightly between different sections because of the lack of precise anatomical landmarks.

Reverse transcription quantitative polymerase chain reaction

Total RNA was isolated from pools of 10 embryos at the tailbud stage using TRIzol (Thermo Fisher Scientific) and purified by isopropanol precipitation. First-strand synthesis was performed using the Maxima First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) with 1 μg of RNA template. A total of 0.5 μl of the resulting first-strand cDNA was used in a 10 μl RT-qPCR. qPCR was performed in technical duplicate using a CFX Connect Real-Time PCR system (Bio-Rad, Hercules, California) and fold changes calculated using the 2^−ΔΔCt method.

Quantification of ISV numbers

fli1a:GFP-positive ISVs were counted under a stereomicroscope at 48 hpf along the entire body axis. Controls and treated embryos were derived from the same clutch. All experiments were repeated at least twice.

Statistics

Statistical analyses were performed in R (count data) and python (qPCR and scRNA-seq data). Count data were fitted to a Poisson model using the “glm” library. For log-transformed qPCR data, normality was assumed and P values calculated by unpaired two-sample t test using the “scipy” package. Bonferroni correction was applied to adjust P values, when appropriate.

Supplementary Materials

This PDF file includes:

Figs. S1 to S16

Table S4

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S3, S5

Movies S1 to S3

View/request a protocol for this paper from Bio-protocol.