Introduction

It has been well established that organ-specific blood vessels exhibit a large diversity of specialized endothelial cells. However, their developmental origin has been debatable. Previous studies have suggested that the progenitors of internal organ vasculature in mammalian embryos form by differentiation de novo, or vasculogenesis.^1–4 Different studies have demonstrated that the endocardium, and endoderm or mesothelium may contribute to the organ vasculature in mammals and avian embryos.^5–7 In contrast, several studies have argued that the posterior cardinal vein (PCV) serves as the major source of organ-specific vascular endothelial cells in zebrafish embryos.^8–10 According to this model, individual cells from the PCV migrate at 30-48 hours post fertilization (hpf) and coalesce into the intestinal vessels, which include the supraintestinal artery (SIA) and the subintestinal vein (SIV). These vessels subsequently undergo remodeling and ultimately provide blood supply for the internal organs, including the pancreas and the liver.¹⁰ However, this model in zebrafish contrasts with the de novo vasculogenesis model, suggested in mammalian embryos. In addition, it is unclear if the PCV is the only or even the main source of internal organ-specific vascular progenitors.

In zebrafish, similar to mammalian embryos, the earliest vascular progenitors originate from the lateral plate mesoderm (LPM) during late gastrulation and somitogenesis, and coalesce into the major axial vessels, the dorsal aorta (DA) and the PCV.^11–14 This process is thought to be largely complete by the 24 hpf stage, when blood circulation is first initiated. We have recently demonstrated that between 24 and 48 hpf a separate group of late-forming vascular progenitors emerges from the LPM along the yolk extension at the site termed the secondary vascular field (SVF).¹⁵ SVF cells can uniquely incorporate into the PCV after blood circulation has already been established, and show high expression of vascular progenitor markers, including etv2/etsrp, tal1/scl, npas4l, while they are negative for the expression of more differentiated endothelial markers such as kdrl or fli1a. Previous time-lapse imaging using etv2 reporter lines demonstrated that SVF cells can contribute to the intestinal vasculature, including the SIA and SIV.¹⁵ However, the ultimate fate of SVF cells has not been clear due to the absence of genetic tools to perform the lineage tracing. Furthermore, mechanisms regulating the emergence of SVF cells and their contribution to vasculature remain largely unknown.

To identify the molecular signature of SVF cells, we have previously performed transcriptomic analysis of the zebrafish trunk region at 30 hpf.¹⁶ This analysis identified some SVF marker genes, which have not been previously associated with vascular development, including junctional adhesion molecule 2b (jam2b).¹⁵ Vertebrate JAM homologs have been implicated in diverse processes including leukocyte-endothelial interaction and transendothelial migration, maintenance of hematopoietic stem cells (HSCs), vascular development, and others.^17,18 A related zebrafish paralog jam2a has been previously implicated in HSC formation.¹⁹ However, the role of jam2b in vascular development has not been described.

Here we analyzed the role of zebrafish jam2b in the formation of intestinal vasculature and the emergence of SVF cells. We demonstrate that jam2b is expressed along the lateral plate mesoderm in the region that includes SVF cells. By using genetic lineage tracing and time-lapse imaging we demonstrate that jam2b-positive cells contribute to multiple tissues, including the mesothelium, pericardium, fin buds and venous and intestinal vasculature. To study the role of Jam2 homologs in vascular development, we have generated genetic mutants in jam2b and a related homolog jam2a. Double jam2a; jam2b maternal-zygotic mutants show a reduced number of SVF cells, which correlates with the defects in the formation of the intestinal and venous vasculature, and display reduced regenerative potential after endothelial cell ablation. We further demonstrate that the transcription factor hand2 functions upstream of jam2b and is required for the formation of SVF cells. Altogether, our results identify a new role for Jam2 signaling and Hand2 function in the emergence of organ-specific vascular progenitors.

Results

Jam2b expression analysis

During the previous transcriptomic analysis, we have identified jam2b as one of the markers for SVF cells, which are late-forming vascular progenitors that emerge along the yolk extension at 24-48 hpf.¹⁵ As we and others have previously demonstrated, jam2b expression is localized to the ventrolateral region of the LPM along the anterior and trunk regions during 20 somite – 48 hpf stages.^15,20,21 It is bilaterally expressed along the yolk extension during 24-48 hpf, where SVF cells originate (Suppl. Fig. S1). We have previously demonstrated that jam2b and the vascular progenitor marker etv2/etsrp expression in the SVF cells partially overlaps along the medial portion of jam2b expression domain within the mid-trunk region.¹⁵

Generation and characterization of jam2b^Gt(2A-Gal4) knock-in reporter line

To analyze jam2b+ cells in live embryos, we used CRISPR / Cas9-mediated non-homologous recombination approach ²² to insert Gal4 transcriptional activator into the genomic locus of jam2b (Suppl. Fig. S2A). Similar to the endogenous jam2b mRNA expression, jam2b^{Gt(2A-Gal4)ci55};UAS:GFP (further abbreviated as jam2b^Gt(2A-Gal4);UAS:GFP) expression was observed bilaterally in the lateral plate mesoderm along the yolk and yolk extension at 24 hpf (Suppl. Fig. S2B; Fig. 1A,B). At 3 and 5 dpf jam2b^Gt(2A-Gal4);UAS:GFP expression was observed in the cell layer surrounding the yolk and yolk extension, which presumptively corresponds to the mesothelium (Fig. 1C-F). GFP expression was also observed in a subset of muscle cells within the skeletal and craniofacial muscle, and in the cells at the periphery of the eye (Fig. 1C-F). GFP expression in the trunk portion of the LPM partially overlapped with etv2 expression in SVF cells at 30 hpf (Fig. 1G-I). Interestingly, GFP-positive cells were observed in the vasculature between 30 hpf and 3 dpf (Fig. 1J-R). The majority of GFP+ cells were located in the PCV, while a few cells were found within the ISV and SIA, and in less frequent cases contribution to the DA and parachordal lymphangioblasts (PLs) was observed (Suppl. Fig. S3). In addition, some GFP cells were located in the thoracic duct at 4 dpf, suggesting that SVF cells contribute to lymphatic vasculature (Fig. 1S-U). Due to very bright GFP expression over the yolk, we could not reliably analyze GFP cell presence in the SIV vessel, which is also positioned over the yolk and yolk extension. Overall, these findings suggest that jam2b+ cells contribute to the vasculature, including the PCV, ISVs, intestinal vasculature and lymphatics. These findings are consistent with our previously reported contribution of SVF cells to the PCV and intestinal vasculature ¹⁵.

Figure 1.

Jam2b+ cells are derived from the lateral plate mesoderm and incorporate into functional vasculature

To visualize directly the contribution of jam2b+ cells to vasculature, we performed time-lapse imaging of jam2b^Gt(2A-Gal4);UAS:GFP; Tg(kdrl:mCherry) embryos starting at 24 hpf. This imaging showed that a subset of GFP-positive cells migrated dorsally away from the LPM region, integrated into the PCV, proliferated and initiated kdrl:mCherry expression as they differentiated into vascular endothelial cells (Fig. 2A-F and Movie 1). This data shows that mesenchymal cells can contribute to and incorporate into the functional vasculature, even after blood circulation has initiated, in agreement with our previous observations using etv2^Gt(2A-Venus) reporter line.¹⁵

Figure 2.

In an attempt to identify the origin of SVF cells, we analyzed jam2b^Gt(2A-Gal4);UAS:GFP expression at the 18-somite stage. Similar to jam2b mRNA expression, GFP-positive cells were located bilaterally in the lateral plate mesoderm, mostly lateral and ventral to the emerging etv2+ vascular progenitors (Fig. 2G-I). There was no apparent overlap between jam2b^Gt(2A-Gal4);UAS:GFP and etv2 expression at this stage. This domain largely overlapped with the expression domain of prrx1a (Fig. 2J-L), which labels the lateral portion of the LPM.²³ To analyze the behavior of these cells, we performed time-lapse imaging of jam2b^Gt(2A-Gal4);UAS:GFP embryos starting at the 18-somite stage. GFP-positive cells remained in the same LPM region from the 18-somite stage to 24-30 hpf stages (Movie 2). As the embryo continued to undergo convergent extension, the same LPM region became bilaterally located along the yolk extension. Thus, jam2b-expressing cells observed at 24-30 hpf along the yolk extension are derived from the ventrolateral portion of the LPM.

Genetic lineage tracing of jam2b+ cells

To perform genetic lineage tracing of jam2b-expressing cells, we generated UAS:Cre and UAS:CreERT2 zebrafish transgenic lines using Tol2 transgenesis. Subsequently they were crossed to jam2b^Gt(2A-Gal4); actb2:loxP-BFP-loxP-dsRed; fli1:GFP adults, which ubiquitously express floxed BFP under actb2 promoter, in combination with the jam2b^Gt(2A-Gal4 knock-in driver and vascular endothelial fli1:GFP reporter (Fig. 3A). This approach permanently labels jam2b-derived cells, which switch their color from BFP to dsRed due to Cre-LoxP recombination. The embryos were then analyzed at 3 dpf by confocal microscopy, revealing switched dsRed-positive cells in the presumptive mesothelial layer surrounding the yolk and yolk extension, a subset of skeletal muscle cells and vasculature (Fig. 3B,E). In addition, both epicardial and myocardial layers within the heart were labeled by dsRed fluorescence (Fig. 3C,D). An embryo negative for jam2b^Gt(2A-Gal4) expression was included for autofluorescence control (Fig. 3H). The majority of labeled vascular endothelial cells were positioned in the SIA and PCV, while a few cells were positioned in the DA, PL, and ISVs (Fig. 3F,G,I-K). Most ISVs, which showed dsRed expression, were venous, and very little contribution to arterial ISVs was observed (Fig. 3F,G,K). Due to intense dsRed fluorescence in the mesothelial layer along the yolk and yolk extension we were unable to reliably determine the contribution of jam2b+ cells to the SIV. To determine when jam2b+ cells contribute to vasculature, UAS:CreERT2 line was crossed to jam2b^Gt(2A-Gal4); actb2:loxP-BFP-loxP-dsRed; fli1:GFP adults and 4-hydroxitamoxifen (4-OHT) was added at different time points to induce Cre-LoxP recombination. 4-OHT addition at 7 hpf mostly in labeled SIA with a smaller fraction of cells present in the SIV, PCV and venous ISVs (Fig. 3L). Only cells in the SIA were labeled when 4-OHT was added at 22 hpf. Addition of 4-OHT at 7 hpf and subsequent washing out at 22 hpf resulted in the labeling of SIA and a small number of cells within the PCV (Fig. 3L). These results argue that jam2b+ cells contribute to vasculature both before and after 22 hpf. However, contribution after 22 hpf was largely limited to the intestinal vasculature, which supports our findings that SVF cells show major contribution to the SIA and SIV.

Figure 3.

Intestinal vasculature forms largely by new vasculogenesis

Previous studies have argued that the intestinal vasculature, including the SIA and SIV, are derived from the PCV ^8–10. In contrast, our current data suggest that jam2b-positive SVF cells make a major contribution to the intestinal vasculature. However, the number of intestinal vascular cells labeled by jam2b^Gt(2A-Gal4); UAS:Cre or UAS:CreERT2 recombination was rather small. This could be due to only a small contribution of SVF cells to the intestinal vasculature, or alternatively, due to low recombination efficiency observed in jam2b^Gt(2A-Gal4); UAS:Cre and UAS:CreERT2 lines. To clarify the contribution of new vasculogenesis towards the intestinal vasculature, we photoconverted all vascular endothelial cells in etv2:Kaede transgenic line at 24 hpf stage, which expresses green to red photoconvertible Kaede in all vascular endothelial cells,²⁴ and then analyzed the percentage of red and green labeled cells at 4 dpf in the intestinal vasculature over the yolk extension region, where SVF cells are observed. Notably, 78.5% of cells in the SIA and 80.5% cells in the SIV showed only green expression of newly synthesized Kaede, and had no red fluorescence of photoconverted protein, arguing that they originated after photoconversion at 24 hpf stage (Fig. 4). These data suggest that most intestinal progenitors in the mid and posterior trunk regions form by new vasculogenesis from SVF cells.

Figure 4.

Transcriptomic analysis of SVF cells

We have previously performed single-cell RNA sequencing (scRNA-seq) of the entire trunk region of zebrafish embryos in order to identify the transcriptional profile of SVF cells.^15,16 While this analysis identified a set of SVF marker genes that included etv2, tal1, lmo2 and jam2b, only a limited number of marker genes were identified with this approach, because SVF cells comprise only a small portion of cells in the entire trunk. Therefore, we used jam2b and etv2 reporter lines to perform additional transcriptomic analysis in order to characterize the profiles of SVF cells and their transition to vascular endothelial cells in greater detail. mCherry and Venus-positive cells were FACS-sorted from jam2b^Gt(2A-Gal4);UAS:mCherry-NTR; etv2^Gt(2A-Venus embryos at 36 hpf and subjected to scRNA-seq using the Chromium platform (10x Genomics). Bioinformatic analysis using Seurat 4.0 identified a total of 37 cell clusters, including the SVF cell cluster, five venous, three arterial clusters, and a capillary endothelial cluster (Fig. 5A-C, Suppl. Table 1). The SVF cell cluster (cluster #15) was identified based on top marker genes, which included npas4l, tal1, lmo2, and etv2 (Fig. 5D,E, Suppl. Table 1). Additional SVF top marker genes included gfra3, sox7, si:dkey-52l18.4, egfl7 and tmem88a (Suppl. Table 1). The key SVF marker genes, including etv2, tal1, lmo2, sox7, si:dkey-52l18.4 and egfl7 were also identified in our earlier study,¹⁵ arguing that the current approach has successfully identified the transcriptional signature of SVF cells.

Figure 5.

To identify transitional events during SVF cell differentiation into vascular endothelium, we performed cell lineage trajectory analysis using Monocle 3.²⁵ Cell lineage trajectory showed that SVF cells can directly transition into both venous and arterial clusters (Fig. 5F). Such differentiation trajectory agrees with experimental observations of SVF cells contributing both to venous (PCV, SIV) and arterial vasculature (SIA). Unfortunately, it is currently not possible to distinguish SIA or SIV-specific cells on the lineage trajectory from other types of arterial or venous cells due to the absence of specific markers for these cells.

Jam2b mutants undergo normal vascular development

To analyze the functional role of Jam2b in vascular development, we generated jam2b mutants using CRISPR/Cas9 mediated genome editing. Two different guide RNAs were used to delete the jam2b promoter region, and a stable mutant line was established (Suppl. Fig. S4A). As indicated by ISH analysis, jam2b-/- embryos showed no detectable jam2b expression in the LPM region, suggesting that the mutant may be a null or close to a null (Suppl. Fig. S4C,D). However, both zygotic and maternal zygotic jam2b-/- embryos appeared morphologically normal and did not show any apparent vascular defects based on kdrl:GFP expression analysis (data not shown). Jam2b homozygous mutants were viable as adults and did not exhibit any apparent defects. Etv2 expression in SVF cells was also not affected in maternal-zygotic jam2b mutant embryos (Suppl. Fig. S5A-C). We have previously demonstrated that SVF cells contribute to vascular recovery after endothelial cell ablation in etv2^Gt(2A-Gal4); UAS:mCherry-NTR embryos.¹⁵ We then tested if vascular recovery was affected in jam2b deficient embryos. Metronidazole (MTZ) was added at approximately 10 hpf stage and then washed out at 45 hpf stage in zygotic jam2b-/- and sibling control jam2b+/- embryos in etv2^Gt(2A-Gal4); UAS:mCherry-NTR; kdrl:GFP background. This resulted in nearly complete ablation of vascular endothelial cells. As we previously demonstrated, such treatment results in a complete ablation of all vascular endothelial cells except for SVF cells, which emerge later than other endothelial cells.¹⁵ Afterwards, the recovery was analyzed at 72 hpf, when the partial regeneration of vascular cords was apparent, which as our previous study suggested, were derived from SVF cells.¹⁵ However, no difference in vascular recovery was observed between jam2b-/- and jam2b+/- embryos (Suppl. Fig. S5D,E).

jam2a; jam2b mutants display defects in SVF cell formation and intestinal and venous vasculature

We then explored the possibility that jam2b may function redundantly with a related homolog jam2a. While jam2a is broadly expressed in the somitic muscle,^20,21,26 our scRNA-seq analysis indicated that it also showed significant expression in the SVF cell population, although it was not enriched among marker genes.¹⁵ We used CRISPR / Cas9 mutagenesis to remove the entire coding sequence of jam2a (Suppl. Fig. S4B,E,F). To expedite the generation of double mutant embryos, jam2a mutant generation was performed in jam2b-/- mutant background. Stable jam2a;jam2b mutant line was generated and used for further experiments. Zygotic jam2a-/- / maternal jam2b-/- mutant embryos, obtained by the in-cross of jam2a+/-;jam2b-/- adults did not show any apparent defects in vascular development or SVF cell formation (data not shown). Subsequently, maternal-zygotic (MZ) mutant embryos were obtained and analyzed for SVF cell formation and vascular defects. Interestingly, MZ jam2a-/-; jam2b-/- embryos showed significant reduction in the SVF cell number when compared to control double heterozygous embryos in the same background. The number of etv2, tal1 and npas4l-positive SVF cells was greatly reduced in MZ jam2a;jam2b mutant embryos, while no other morphological defects were apparent (Fig. 6). Overall vascular patterning appeared largely unaffected in MZ jam2a;jam2b mutant embryos based on kdrl:GFP fluorescence analysis at 3 dpf (Fig. 7A,B). However, MZ jam2a;jam2b mutant embryos showed greater incidence of incompletely extended ISVs (Fig. 7A-C). Venous ISVs, derived from the PCV, made 97% (33 out of 34) of such truncated ISVs, while only 3% were arterial. This suggested that the number of venous cells could be reduced, possibly due to reduced contribution of SVF cells to the PCV. Indeed, the number of ECs in the floor of the PCV was slightly yet significantly reduced in MZ jam2a; jam2b mutant embryos (Fig. 7D-F). However, the most pronounced defects were observed in the formation of the intestinal vasculature, in particular the SIA, which was underdeveloped and showed gaps or missing segments in MZ jam2a;jam2b mutants (Fig. 7G-I). In addition, a significant fraction of MZ jam2a;jam2b mutants showed gaps in the SIV (Fig. 7G-I).

Figure 6.

Figure 7.

Because our data suggested that jam2b+ cells may contribute to lymphatics, we analyzed the formation of the thoracic duct using prox1:RFP, fli1:GFP reporter embryos. Approximately 37% of MZ jam2a;jam2b mutants had gaps in the thoracic duct compared to 5% of control jam2a+/-;jam2b+/- embryos (Fig. 7J-L). These results argue that jam2a and jam2b function is required for the formation of the intestinal vasculature and lymphatics.

We then tested if vascular recovery after endothelial cell ablation was affected in MZ jam2a; jam2b mutant embryos in etv2^Gt(2A-Gal4; UAS:mCherry-NTR background. In order to ablate all vascular endothelial cells, MTZ was added at 6 hpf and then washed out at approximately 45 hpf. Subsequently, embryos were analyzed at 72 hpf by confocal imaging. A substantial recovery was observed in the control jam2a; jam2b double heterozygous embryos, while it was greatly reduced in MZ jam2a, jam2b mutant embryos (Fig. 7M-O). These results argue that the loss of SVF cells in MZ jam2a; jam2b mutants leads to defects in venous, intestinal and lymphatic vessel development, and is also associated with the reduced vascular recovery after EC ablation.

Hand2 functions upstream of jam2b in SVF cell formation

Recent work has identified that a bHLH transcription factor hand2, previously implicated in cardiac development, is also expressed in the lateral plate mesoderm along the yolk extension and labels mesothelial and pericardial progenitors, similar to jam2b.^27,28. Intriguingly, hand2-2A-Cre based lineage tracing showed contribution to the venous and intestinal vasculature,²⁹ similar to our results obtained using jam2b^Gt(2A-Gal4); UAS:Cre. To determine if both genes are co-expressed in the same cells, we performed fluorescent hybridization chain reaction (HCR) analysis at 24 hpf. Indeed, expression of hand2 and jam2b largely overlapped in the LPM region along the yolk extension (Fig. 8A-F). To test if hand2 may function upstream of jam2b, we analyzed jam2b expression in the previously generated hand2 mutants by in situ hybridization. Indeed, jam2b expression was absent in hand2 mutants from most of the LPM, including the SVF-forming region along the yolk extension (Fig. 8G-J). Only the anterior domain of jam2b expression adjacent to the yolk was not affected in hand2 mutants (Fig. 8G-J, white arrowheads). We then analyzed if hand2 function is required for etv2 expression in SVF cells. Indeed, hand2 mutants showed complete absence of SVF cells at 30 hpf (Fig. 8K,L). In addition, hand2 mutants showed reduced etv2 expression in the PCV, which is consistent with its previously demonstrated requirement in the formation of venous endothelium.³⁰ These results argue that hand2 functions upstream of jam2b in the specification of SVF cells within the trunk region of the LPM.

Figure 8.

Discussion

Our previous work has demonstrated that new etv2+ progenitors emerge from the trunk region of the LPM along the yolk extension and contribute to the intestinal and axial vasculature.¹⁵ Here we implicate jam2a and jam2b function in the emergence of SVF cells and their contribution to vasculature. Our results show that jam2b is expressed along the broad region of LPM, including the site where SVF cells emerge. Lineage tracing studies revealed that jam2b-expressing cells show major contribution to the intestinal vasculature, and a smaller contribution to the PCV and venous ISVs (Fig. 9). MZ jam2a;jam2b mutants show a reduced number of SVF cells and display defects in the intestinal and venous vessel development. We further show that hand2 transcription factor functions upstream of jam2b in the specification of SVF cells within the trunk region of the LPM.

Figure 9.

Previous studies have argued that internal organ-specific vessels, including the intestinal vessels, are derived from the PCV.^8–10 According to the recent model, individual cells migrate and assemble into the two vessels, SIA and SIV, which can further branch and give rise to interconnecting vessels, ICV.^8–10 Our recent study challenged this model and showed that SVF cells can contribute to intestinal vasculature.¹⁵ However, it has not been clear if SVF cells have only a minor contribution to the intestinal vasculature. Our current results suggest that SVF cells may be the main source of the intestinal vasculature, while the PCV only has a minor contribution. Etv2:Kaede photoconversion showed that most cells in the intestinal vasculature emerge after the 24 hpf stage and, therefore, they were not present in the PCV at the time of the photoconversion. Although jam2b^Gt(2A-Gal4); UAS:Cre based lineage tracing resulted in only a fraction of the PCV and intestinal vessels labeled with switched dsRed expression, this could be due to the relative inefficiency or mosaic expression of Cre. Based on our earlier time-lapse lineage tracing experiments, it appears that some SVF cells migrate directly to the SIA and SIV, while others integrate into the PCV first, and only afterwards exit the PCV and migrate to the intestinal vasculature.¹⁵ Thus, our model can be reconciled with the previous observations if some SVF cells first integrate into the PCV and subsequently contribute to the intestinal vasculature. It is also possible that the earlier studies focused largely on the anterior portion of the intestinal vasculature, which forms the plexus-like structure and is largely derived from the PCV. Our study focused on the mid-trunk region along the yolk extension, where SVF cells are positioned, which is posterior to the intestinal plexus, analyzed in the previous studies.

jam2b expression within the LPM is observed during somitogenesis stages. It is evident that this region gives rise to diverse cell types, including mesothelial, pectoral fin, pericardium and muscle. Similar contribution has also been demonstrated for hand2-positive LPM cells,²⁸ which largely overlap with jam2b expression. Only a small fraction of jam2b-positive cells contributes to the vascular endothelium. The cells fated to become endothelium are located at the edge of the medial portion of the jam2b expression domain in the trunk region, closest to the midline (Fig. 9). It is tempting to speculate that all jam2b cells are multipotent initially, yet they experience different signals based on their relative position. The cells closest to the midline in the trunk region become induced to differentiate into vascular endothelium. The inductive signal is currently unknown and will require further investigation. Interestingly, vegfaa is expressed in the region where intestinal vessels form.⁸.= Furthermore, vegfaa mutants show reduced number of SVF cells, which correlates with the defects in the intestinal vessel formation,^8,15 suggesting that vegfaa could be a candidate signal.

During embryonic vasculogenesis arterial and venous progenitors originate from the LPM during somitogenesis stages.^11–14 While the arterial progenitors originate close to the midline, venous progenitors emerge from the more lateral portion of the LPM.²⁴ As we show here, early venous progenitors emerge adjacent to jam2b-positive LPM cells, which are positioned further laterally from the venous progenitors. While jam2b and etv2 expression overlaps in later forming SVF cells, it does not appear to overlap in early endothelial progenitors, which was also corroborated in our scRNA-seq data.^15,23 We also did not observe jam2b^Gt(2A-Gal4); UAS:GFP expression in the PCV prior to 24 hpf stage, suggesting that jam2b may not be involved in early vasculogenesis. It appears that jam2b+ LPM cells maintain vasculogenic potential until later stages past 24 hpf.

The mechanism of how jam2a; jam2b mutants regulate SVF cell specification is unclear. A previous study has demonstrated that antibodies against murine Jam-B (Jam2) inhibited angiogenesis in vitro and interfered with the VEGFR2 signaling pathway.³¹ Our previous data indicate that VEGF signaling is required to upregulate etv2 expression during early vasculogenesis, and it is also required for SVF cell emergence.^15,32 It is possible that Jam2 signaling is required to promote VEGF signaling, which will have to be investigated in future studies.

Previous studies have demonstrated that hand2 function is required for the PCV development. npas4l and etv2 expression were reduced in the venous progenitors, arguing that hand2 functions upstream of these transcription factors.^30,33 We show here that SVF cells are also absent in hand2 mutants. It is possible that hand2 directly regulates npas4l and /or etv2 expression. Alternatively, hand2 may function through jam2b and indirectly be involved in SVF specification through VEGF or a different pathway.

In summary, this study demonstrates that jam2b-positive LPM serves as a major source of internal organ-specific vascular progenitors and identifies the functional requirement for jam2a and jam2b in promoting etv2 expression and vasculogenesis. Because analogous late-forming vascular progenitors have been also identified in the mammalian embryos,¹⁵ it is likely that the mechanisms regulating their emergence are also evolutionarily conserved. These results will be important to understand the molecular mechanisms involved in the formation of different vascular beds and will promote the development of new approaches aimed at vascular regeneration.

Materials and methods

Zebrafish lines

The following previously established lines were used in the study: wild-type AB (Zebrafish International Resource Center), Tg(kdrl:mCherry)^ci5,³⁴ Tg(fli1a:GFP)^y1,³⁵ hand2^s6,²⁷ etv2^{Gt(2A-Gal4)ci32};UAS:mCherry-NTR,³⁶ etv2^Gt(2A-Venus),¹⁵ Tg(actb2:loxP-BFP-loxP-dsRed)^sd27,³⁷ Tg(kdrl:GFP)^s843,¹⁴ TgBAC(prox1a:KalT4-UAS:uncTagRFP)^nim538 (abbreviated as prox1:RFP), Tg(lyve1b:dsRed)^nz101.³⁹

Generation of jam2b^Gt(2A-Gal4) knock-in line

To generate the jam2b^{Gt(2A-Gal4)ci55};UAS:GFP knock-in line (further abbreviated as jam2b^Gt(2A-Gal4); UAS:GFP), a gBlock Gene Fragment was made by gene synthesis (IDT) which included jam2b sgRNA site, the P2A sequence, Gal4 DNA binding domain and poly-A tail. This fragment was then A-tailed using dATP and Taq DNA Polymerase (NEB, cat. no. M0273) followed by TA cloning into TOPO dual vector using manufacturer’s instructions (ThermoFisher, cat. no. 450640). A mixture of 300 pg Cas9 mRNA and 50 pg jam2b-sgRNA-P2A-Gal4 plasmid DNA was injected into Tg(UAS:GFP)⁴⁰ embryos at 1-cell stage. Embryos showing any specific GFP expression were raised through adulthood and screened to identify a founder which produced progeny with a specific expression pattern.

Generation of jam2a and jam2b mutant lines

To generate jam2b^sfl2 line, two synthetic gRNAs (Synthego) targeting jam2b promoter region were co-injected with Cas9 protein (NEB, EnGen Spy Cas9 NLS, Cat No. M0646) into zebrafish embryos at 1-cell stage. The following sgRNA sequences were used: GCAATTCTGGAAGAAGTCAA and AATAAGCAGGAGAAGAGCAG. Successful deletion was confirmed using the following primers outside of the deletion site: jam2b_P3.4.5_F: GCATCACTCAAAGGCAAATG and jam2b_P1.2_R: ACAGTTCCCACCATTTCAGC. A single founder with 1,071 bp deletion, which encompasses jam2b proximal promoter region and a portion of exon 1, was identified and used to establish the mutant line. The genomic sequence flanking the deleted region is shown here: TATAGACCGGGCAGATTAGCCATTG………TCTTCTCCTGCTTATTTACAGTAAG.

To generate jam2a^sfl3 line, two synthetic gRNAs (Synthego) were designed to delete jam2a coding sequence with the following sequences: TCTGTGTGCCTCTAGGTGAG and CAGAGCACAAACCCACGCAA. Both gRNAs were co-injected together with Cas9 protein into one-cell stage zebrafish embryos, homozygous for jam2b^sfl2 mutation. Successful deletion was confirmed using the following primers that flank the deletion site: jam2a-sg1_2_F: CAAACACCTGTGACCTTACAAAA and jam2a-sg1_2_R: TATCAATGGCAACACAGTGG. A founder was identified with the deletion of 15,771 bp, which includes the entire jam2a coding sequence, and used to establish the stable line for further studies. The genomic sequence flanking the deleted region is shown here: CCTACAGGTGGTTTGTCAAGCCCAC AAAGGTAGGGAGAACATGCAAACTC

Generation of UAS:Cre and UAS:CreERT2 lines

To generate UAS:Cre^sfl4 and UAS:CreERT2^sfl5lines, the gene fragment containing 5xUAS sequence, hbae3 5’and 3’UTR sequences, zebrafish Kozak sequence,⁴¹ NLS-Cre, and NLS-CreERT2 codon optimized for zebrafish expression using iCodon approach,⁴² SV40 polyA site, and attL4 and attL3 sequences specific to LR recombinase, was synthesized by gene synthesis and subcloned into pUC57 vector (Genscript). LR recombination into the Gateway Destination vector containing Tol2 transposase sites and lens specific alpha crystallin-dsRed sequence⁴³ was performed using LR recombinase (ThermoFisher). Zebrafish embryos were co-injected with 25-40pg of each construct together with 75-150 pg tol2 mRNA. Founders were screened for the presence of lens fluorescence in their progeny. Recombination activity was tested by mating to jam2b^Gt(2A-Gal4); Tg(actb2:BFP-loxP-dsRed-loxP) driver / reporter line. One of each lines, UAS:Cre and UAS:CreERT2, which showed good recombination activity, was selected and used for all subsequent experiments.

Confocal microscopy imaging and image processing

For confocal imaging, embryos were anesthetized in 0.016% Tricaine solution, mounted in 0.6% low melting point agarose and imaged using Nikon Eclipse or Nicon AX confocal microscope. Series of z-slices were acquired using NIS Elements software. Nikon AI Denoise function was used to reduce the image noise. Maximum intensity projection was generated using Fiji or Nikon Elements software. Images were adjusted using Levels function in Adobe Photoshop to enhance the brightness and contrast. In all cases, the same adjustments were performed in control and experimental embryo samples.

Hybridization chain reaction (HCR)

Whole mount fluorescent in situ hybridization was performed using the hybridization chain reaction version 3.0 as previously described.⁴⁴ Fluorescently labeled RNA probes for etv2, prrx1a, hand2 and jam2b were purchased from Molecular Instruments, Inc.

Whole mount in situ hybridization (ISH) was performed as previously described.⁴⁵ DIG-labeled antisense RNA probes for jam2b,¹⁵ npas4l,^15,46 etv2/etsrp,⁴⁷ scl/tal1⁴⁸ were synthesized as previously described. Following ISH, embryos were whole mounted in 0.6% low melting point agarose and imaged under light microscopy at 10X or 20X with Nikon Eclipse compound microscope. A series of z-slices were acquired using NIS Elements software to produce extended focus projected images.

4-hydroxitamoxifen (OHT) treatments

To induce tissue-specific recombination at a specific time point, jam2b^Gt(2A-Gal4);UAS:CreERT2 line was crossed with the Tg(actb2-loxP-BFP-loxP-dsRed; fli1:GFP) reporter line. Embryos were treated with 10 µM 4-OHT (Sigma-Aldrich) solution starting at 7 or 22 hpf stage. Alternatively, to analyze the timing requirement, embryos were treated at 7 hpf stage and subsequently washed at 22 hpf by rinsing a few times in embryo water. Confocal z-stacks were acquired at 3 dpf and analyzed with NIS Elements Software. Recombination was quantified by counting dsRed-positive cells within vasculature.

Endothelial cell (EC) count

To assess EC number in jam2a;jam2b double homozygous mutants and control heterozygous embryos, confocal images were analyzed using NIS Elements Imaging Software. An area consisting of 5 intersegmental vessels in the trunk region was selected, and ECs located at the bottom of the posterior cardinal vein (PCV) were manually counted.

Subintestinal vessel and thoracic duct (TD) measurements

To evaluate SIA, SIV and TD in jam2a; jam2b double homozygous mutants and control heterozygous embryos, confocal images were analyzed using NIS Elements 5.41.02 Imaging Software. SIA, SIV and TD were categorized into normal and abnormal groups according to morphological characteristics. Normal vessels were considered uniform and without any gaps throughout the whole embryo; everything else was considered abnormal.

Chemical cell ablation

To assess vascular tissue ablation and regeneration, jam2a;jam2b homozygous mutant zebrafish in kdrl:GFP background were crossed with jam2a;jam2b homozygous mutants carrying the etv2^Gt(2A-Gal4); UAS:NTR-mCherry transgene. As a control group, heterozygous embryos were obtained by crossing jam2a; jam2b homozygous mutants carrying the etv2^Gt(2A-Gal4); UAS:NTR-mCherry transgene with wild-type kdrl:GFP. Embryos were collected and treated with 10 mM metronidazole (MTZ) dissolved in 0.1 % DMSO at 6 or 10 hpf stages. At 2 dpf, embryos were dechorionated, thoroughly washed, and transferred to fresh embryo water to allow for recovery. Confocal imaging was performed at 3 dpf to assess vascular regeneration. NIS-Elements Imaging Software (version 5.41.02) was used to quantify tissue ablation and recovery. The total length of regenerated vessels was calculated by measuring and adding the lengths of all recovered vessels based on mCherry and GFP expression.

Statistical analysis

Data representing graphs were generated and statistical analyses (t-test and Fisher’s exact test) were performed using GraphPad Prism software.

Single-cell RNA-seq analysis

Transgenic etv2^Gt(2A-Venus) fish were crossed to jam2b^Gt(2A-Gal4);UAS:NTR-mCherry fish. Double-positive embryos expressing both Venus and mCherry fluorescent proteins were selected at 36 hpf. Embryos were dechorionated and immediately transferred into a 1.5 ml Eppendorf tube for the generation of single-cell suspension. Whole embryos were then dissociated into a single-cell suspension by cold protease tissue dissociation protocol ⁴⁹. Cells expressing Venus only, mCherry only, and both (Venus and mCherry) were sorted using a BD/FACSAria II flow cytometer cell sorter. After sorting, cells expressing both Venus and mCherry were equally divided and added into the Venus only and mCherry only populations manually. Suspensions of ∼10,000 single Venus and mCherry-positive cells were loaded separately onto the 10XGenomics Single Cell 3’ (V3 chemistry) chip. 12 cDNA amplification cycles were used to generate cDNA. Samples were sequenced on an Illumina NovaSeq 6000 instrument (Illumina, San Diego, CA) running an S2 flow cell with parameters as follows: Read1, 28 cycles; Index Read1: 8 cycles; Read 2: 91 cycles at the CCHMC DNA Sequencing core. The fastq files obtained from the sequencing core were then mapped to the Danio rerio genome (version zv11) to generate single-cell feature counts using Cell Ranger version 3.0.2. Counts were generated using fastq files from each of the populations individually. Utilizing R packages: Seurat,⁵⁰ ggplot2,⁵¹ dplyr,⁵² reshape2,⁵³ and tidyverse⁵⁴ the data was filtered for high quality cells with 200-3500 features and less than 5% mitochondrial gene content. The datasets were then integrated and the figure plots were made using custom written R code.^55–57 To find a stable cluster resolution we tested resolutions between 0-3 in 0.1 increments and decided on 1.6, revealing the SVF cells in cluster 15. We then selected clusters for trajectory analysis using the Monocle3 package. The UMAP, feature plots, and violin plots were made using Seurat, and the lineage plot was made using Monocle3.²⁵ Cluster annotation was done manually based on the top marker gene expression. In some cases, AI ChatGPT 5 assistance was used to help with the cell type prediction based on 40-60 top marker gene expression, which was subsequently verified manually based on published gene expression patterns. Single-cell RNA-seq dataset has been submitted to NCBI GEO database under accession number: GSE179926.

Acknowledgements

This research was supported by NIH R01 HL153005 to S.S., a supplemental postdoctoral diversity award R01 HL163161-S1 to R.D., AHA 19POST34400016 award to S.M. and AHA 19POST34450021 award to S.G. We thank Olivia Nester and Kanwal Jussa for their assistance with zebrafish experiments.

Additional information

Data availability

scRNA-seq data were deposited in NCBI GEO (accession number GSE179926). All data generated or analyzed during this study are included in the manuscript and supporting files.

Funding

HHS | National Institutes of Health (NIH) (HL153005)

• Saulius Sumanas

HHS | National Institutes of Health (NIH) (HL163161-S1)

• Saulius Sumanas

American Heart Association (AHA) (19POST34400016)

• Sanjeeva Metikala

American Heart Association (AHA) (19POST34450021)

• Suman Gurung

Additional files

Supplemental Table 1

Movie 1. >Migration of jam2b-expressing cells from the lateral plate mesoderm into the PCV. A time-lapse video of a jam2b^Gt(2A-Gal4);UAS:GFP;Tg(kdrl:mCherry) embryo shows GFP positive cells migrating dorsally, integrating into the PCV, proliferating and initiating kdrl:mCherry expression (white arrowheads) (see Fig. 2A-F). Imaging was started at 24 hpf. PCV, posterior cardinal vein; DA, dorsal aorta. Scale bar: 20 µm.

Movie 2. jam2b-expressing cells are derived from the lateral plate mesoderm. A time-lapse video of a jam2b^Gt(2A-Gal4);UAS:GFP;Tg(kdrl:mCherry) embryo showing from the 18-somite stage to 24 hpf. GFP positive cells can be seen in the ventrolateral LPM region at the 18-somite stage and remain in the same region until 24 hpf as the embryo undergoes convergent extension. PCV, posterior cardinal vein; DA, dorsal aorta. Scale bar: 50 µm.