Temporally and regionally distinct morphogenetic processes govern zebrafish caudal fin blood vessel network expansion

All organs and tissues need to be vascularized during embryonic development. Work over recent decades has revealed a remarkable diversity in terms of anatomy and genetic make-up of organ-specific blood vessels, tailoring them to the needs of the organ they vascularize (Aird, 2007a,b; Augustin and Koh, 2017; Eelen et al., 2020). This is supported by single cell sequencing of mouse tissues revealing that EC heterogeneity can be more strongly correlated with different tissues instead of blood vessel types (Kalucka et al., 2020; Paik et al., 2020). Prominent examples of tissues containing endothelial cells (ECs) with specific properties are the brain (Langen et al., 2019) and the liver (Koch et al., 2021). Lymphatics also possess tissue-specific properties (Petrova and Koh, 2018). Species-specific vascular structures to meet habitat and lifestyle challenges also exist. Tuna fish have evolved special vascular beds that coordinate heat exchange between a large set of arterioles and venules, which is necessary for maintaining optimal muscle temperature and swimming performance in colder waters (Stevens et al., 1974). Dedicated lymphatic structures can provide hydraulic control of fin shapes (Pavlov et al., 2017). In addition, proper organ development relies on the exchange of signalling molecules between ECs and their surrounding cells in a process referred to as angiocrine signalling (Gomez-Salinero et al., 2021; Rafii et al., 2016). For example, in the central nervous system, specialized ECs initially organize neurogenesis (Tata and Ruhrberg, 2018) followed by blood-brain barrier formation (Andreone et al., 2015). A similar crosstalk instructs bone formation (Sivan et al., 2019). Therefore, intricate interactions between the endothelium and other tissue cell types are instrumental for proper organogenesis and tissue function.

Despite these insights, our understanding of the morphogenetic principles that govern the formation of organ- and tissue-specific blood vessels over time and how they might be intertwined with organ function is limited. We also do not know whether regenerating tissues re-use developmental programs (Goldman and Poss, 2020). As a first step towards an understanding of these mechanisms, we decided to study the stepwise assembly of the zebrafish caudal fin vasculature. Previous studies have shown that this vasculature forms in a stereotypical pattern, where each fin ray harbours a central artery that is located within the bone of the ray and is flanked by two lateral veins located on either side of this bone (Bump et al., 2022; Huang et al., 2003; Kametani et al., 2015; Paulissen et al., 2022; Xu et al., 2014). Imaging studies performed during fin regeneration have elucidated that pre-existing veins give rise to regenerating arteries in a process that depends on cxcr4a (Xu et al., 2014) and Notch signalling (Kametani et al., 2015). Elegant lineage tracing studies conducted in regenerating fins similarly provide evidence that arterial and venous ECs share a common lineage precursor cell (Tu and Johnson, 2011). Although a vascular plexus is present at the tips of regenerating fin vessels, this plexus is either absent or greatly reduced in normal growing fins (Huang et al., 2009; Xu et al., 2014), suggesting that distinct processes might pattern regenerating adult blood vessels in comparison with blood vessels in developing fish. A recent report has provided a detailed analysis of the formation of the caudal fin vasculature and examined the interplay between different cell types during this process. These studies showed that osteoblasts are required for normal fin blood vessel outgrowth (Bump et al., 2022). We have previously shown that the pro-angiogenic chemokine cxcl12a is expressed in central fin regions and can accumulate in the bone (Xu et al., 2014). Thus, structural features of the fin can affect local blood vessel patterning.

Fin blood vessel formation requires vascular endothelial growth factor (Vegf) signalling (Bayliss et al., 2006; Bump et al., 2022). Fins lacking blood vessels did not show overt patterning defects in bone or nerve fibres but failed to grow after the initial stages of morphogenesis (Bump et al., 2022). This was more pronounced during fin regeneration (Bayliss et al., 2006). Therefore, it remains to be determined whether blood vessels instruct fin morphogenesis in a subtle manner, leading to later defects in growth, or whether these defects are caused by a lack of nutrients and oxygen. Recent results further reveal surprising differences in the formation of blood vessels among different types of zebrafish fins. Das et al. showed that lymphatics are precursor cells for anal fin blood vessels (Das et al., 2022), while Paulissen et al. showed that the vasculature in pectoral fins is derived from pre-existing blood vessels (Paulissen et al., 2022).

To better understand the processes governing the step-by-step assembly of an organ-specific vascular bed, we set out to examine blood vessel formation throughout caudal fin development. Our findings reveal that initial blood vessel growth occurs through a reiterative set of angiogenic sprouts emanating from veins and giving rise to veins and arteries in a Vegf, Notch and cxcr4a-dependent manner. A later occurring set of vein-derived sprouts exclusively generates veins, establishing a three-vessel pattern. This is further elaborated through inter-vessel connections, which are formed through sprouting and intussusceptive angiogenesis. We also detected a separate vessel network that formed a plexus at the base of the fin and could be distinguished in terms of vessel diameters and transgene expression levels. This network formed through sprouting of blind-ended lumenized vessels. Together, our results reveal a marked variety in the morphogenetic principles occurring at different anatomical locations and developmental stages of fin growth, while generating distinct fin blood vessel types. They also lay the groundwork for further examination of their functional properties and the molecules controlling their formation.

To investigate caudal fin vascular morphogenesis in detail, we used a combination of different transgenic zebrafish lines with enriched transgene expression in either arterial or venous ECs. To visualize the arterial and venous vasculature simultaneously, we imaged Tg(-0.8flt1:RFP)^hu5333; (flt4:citrine)^hu7135 double transgenic embryos. In these, arterial ECs more strongly express RFP, as we had also previously observed in the adult fin (Xu et al., 2014), whereas venous ECs more strongly express citrine (van Impel et al., 2014). We examined vascular growth starting at 3 days post-fertilization (dpf) (Fig. 1A). We observed a dorsally located vein and a ventrally located artery that connected to the dorsal aorta (connection marked in yellow; Fig. 1B,B′). By 1 week post fertilization (wpf), we detected the presence of a new circulatory loop of axial vessels extending posteriorly (Fig. 1C, bracket). Furthermore, a new blood vessel sprouted from this loop towards the ventral region of the growing fin fold (Fig. 1C, arrow). Additional branching was initiated from these sprouts, resulting in the extensive expansion of the vascular tree by 2 wpf (Fig. 1D), as also previously observed by Bump et al. (Bump et al., 2022). Subsequently, these vessels organized into a radial pattern, where each fin ray was inundated by an artery and a vein, which came to lie next to each other, resulting in a two-vessel pattern (Fig. 1E,E′). By 4 wpf, the vasculature had matured into a three-vessel structure, where a central artery was flanked by two veins (Fig. 1F,F′). To corroborate our findings concerning the identity of arterial and venous caudal fin ECs, we combined Tg(-0.8flt1:RFP)^hu5333 animals with Tg(cxcr4a:YFP)^mu104 animals, showing arterial enriched YFP expression (Fig. 1G-K′) (Xu et al., 2014). We found a broad overlap between flt1:RFP expression with cxcr4a:YFP expression, with YFP expression appearing more artery restricted (Fig. 1J′,K′). To characterize vein ECs within the caudal fin, we used Tg(dab2:EGFP)^ncv67Tg (Shin et al., 2019). Dab2 shows higher expression in venous ECs in zebrafish (Kim et al., 2012; Sprague et al., 2006). Accordingly, we found enriched EGFP expression in caudal fin vein cells (Fig. S1A-E′). The same was true for Tg(mrc1a:EGFP)^y251, a gene known to be enriched in veins and lymphatics (Jung et al., 2019) (Fig. S1F-H). Curiously, another vein and lymphatic-enriched transgene, Tg(lyve1:dsred)^cq27Tg (Chen et al., 2019), showed expression within the posterior cardinal vein (PCV) but was not expressed within ECs of the caudal fin (Fig. S2), suggesting that their gene expression profile differs from that of other veins in the zebrafish embryo. We further analysed the location of caudal fin arteries and veins with respect to the nascent bone using the osteoblast marker Tg(sp7:EGFP)^b1212Tg (DeLaurier et al., 2010). We find that in each fin ray, two bone hemi-rays enclose a medially located artery (Fig. S3), as previously shown (Bump et al., 2022; Xu et al., 2014). An open space between the two hemi-rays allows the artery to connect to laterally located veins (Fig. S3F, yellow arrow). In summary, the vasculature of the caudal fin matures from an initial loop, through which each artery drains into a single vein, into a configuration, where each artery drains into two veins (Fig. 1L).

To understand the mechanism through which the fin vasculature expands, we performed live imaging for 16 h starting at 52 hpf using Tg(fli1a:nEGFP)^y7; Tg(-0.8flt1:RFP)^hu5333 double transgenic embryos (Fig. 2A, Fig. S4A,B, Movies 1-3). We could distinguish individual ECs that initially sprouted from the PCV (cells 1 and 2, 0 h time point). Eventually, ECs at the leading edge of the sprout acquired arterial identity, as evidenced by onset of Tg(-0.8flt1:RFP)^hu5333 expression (Fig. 2A, cell 1 at 3 h time point and cells 1 and 2 at 10 h time point, arrowheads). These ECs then migrated in the reverse direction along the existing vasculature until they reached the DA, where they anastomosed (Fig. 2A, 16 h time point). To corroborate our findings, we imaged the nascent fin vasculature at different time points in Tg(fli1:LIFEACT-EGFP)^mu240; Tg(-0.8flt1:RFP)^hu5333 double transgenic animals. We observed the onset of RFP expression in ECs that had just started to turn in their migration direction (Fig. S4E-H, red arrowheads). Subsequently, flt1:RFP-expressing ECs connected to the caudal artery, while new cells positioned at the distal sprout tip started to express flt1:RFP (Fig. S4I,I′, red arrowheads). Thus, although the caudal vein elongated in a proximal to distal manner, the caudal artery extended through the continuous addition of previously distally located cells that initially were part of the vein sprout.

We then used Tg(gata1a:DsRed)^sd2 fish to visualize red blood cells. The nascent vein sprout was lumenized from very early stages onwards, while still exclusively being connected to the venous circulation (Fig. S4J-L, yellow arrowheads indicate red blood cells). As the sprout lacked an outlet, we observed oscillating red blood cells (Movies 4,5). The sprout eventually fused with the pre-existing artery (Fig. S4M,M′). Furthermore, after completion of the arterial connection, the flow pattern within the sprout changed from being oscillatory to laminar. Therefore, this morphogenetic process generates a ‘figure 8’ configuration of tail and trunk arteries and veins (Fig. 3I, schematic).

Of interest, these morphogenetic processes strikingly resembled those that we previously observed during formation of the eye vasculature in embryonic zebrafish (Hasan et al., 2017). In that context, we showed that Vegf and Notch signalling upstream of cxcr4a were important for guiding vein-derived arterial sprouting. We therefore investigated whether these signalling molecules similarly control fin vascular expansion. Time-lapse analysis of artery formation in cxcr4a^um20 mutant zebrafish showed normal expression of the arterial marker flt1:RFP, but failure of these cells to migrate towards the more proximally located artery (Fig. 2B, Fig. S5A,B, Movies 6-8). Instead, we found arterial-fated cells accumulating at the distal tip of the tail fin sprout (Fig. 2B, compare 10 and 16 h time point with wild-type 10 and 16 h time point in Fig. 2A, compare also schematic drawings in Fig. 2B with schematic in Fig. 2A). Thus, as in the embryonic eye vasculature, cxcr4a signalling is instrumental in guiding the nascent artery sprout in the tail fin towards the dorsal aorta. This dependence of vein-artery sprouting on cxcr4a signalling persisted at later juvenile stages, as nascent arteries failed to locate towards the fin ray centre but formed next to laterally positioned veins (Fig. S5E-I).

To examine the onset of Notch pathway activation and cxcr4a expression within the growing sprout, we performed time-lapse imaging in triple transgenic animals. To monitor Notch pathway activation, we used the Notch indicator line Tg(Tp1:Venus-PEST)^s940 together with Tg(-0.8flt1:RFP)^hu5333 and Tg(fli1a:nEGFP)^y7. Time-lapse imaging revealed active Notch signalling in ECs that also expressed flt1:RFP at the tip of the reversing artery sprout (Fig. 2C, see also Movie 9). We further confirmed cxcr4a expression in arterial ECs by performing time-lapse imaging of triple transgenic Tg(-0.8flt1:RFP)^hu5333; Tg(cxcr4a:YFP)^mu104; Tg(fli1a:nEGFP)^y7 animals. These movies revealed an overlap between cxcr4a and flt1 expression (Fig. 2D, Movie 10). We furthermore detected cxcr4a expression in ECs with activated Notch signalling (Fig. S6A, Movie 11). Thus, like ECs in other angiogenic settings, vein cells populating the caudal fin in zebrafish larvae can differentiate into the arterial lineage, initiate Notch signalling and express cxcr4a.

We then functionally tested for the requirement of Vegf signalling in these processes. Mutants for either the Vegfa receptor kdrl (Fig. S6B,C) or the Vegfaa ligand (Fig. 3A,B) displayed normal outgrowth of the vein sprout into the tail but lacked arterial cells. We made similar observations in larvae treated with the Vegf inhibitor SU5416 (Fig. S6D-G). We then used this inhibitor and vegfaa^mu128 mutants to test for Notch pathway activation downstream of Vegfa. We observed absence of expression of the Notch pathway indicator line Tg(Tp1:H2B-mcherry)^s939 in inhibitor treated embryos (Fig. S6D,I) and vegfaa^mu128 mutants (Fig. 3C,D,G), demonstrating that Notch pathway activation requires Vegfa signalling in caudal fin ECs. We also observed a decrease in arterial ECs in vegfaa^mu128 mutants (Fig. 3H). Tg(cxcr4a:YFP)^mu104 expression was lost in vegfaa^mu128 mutants (Fig. 3E,F) and SU5416-treated embryos (Fig. S6F,G,J). We then assayed whether inhibiting Notch signalling through treating embryos with the γ-secretase inhibitor DAPT would influence cxcr4a expression. Consistent with our previous results in intersegmental blood vessels (Hasan et al., 2017), we find downregulation of cxcr4a expression within caudal fin ECs in DAPT-treated embryos (Fig. S6H,J). Furthermore, Vegfa signalling was important for caudal fin EC proliferation (Fig. S7). Together, these results indicate that the mechanisms guiding the formation of the initial artery-vein circulatory loop inundating the caudal fin are like those operating in other regions of the embryo. We now term this mode of angiogenic sprouting ‘vein-artery sprouting’ (Fig. 3I).

As EC proliferation contributed to only a subset of ECs in the caudal fin, we set out to determine additional sources of caudal fin ECs. To do so, we developed a transgenic line that expresses dendra2 in ECs [Tg(fli1:dendra2)]. Dendra2 emission can be switched from green to red using blue light (Gurskaya et al., 2006). We then photoconverted all ECs in the DA and PCV either before caudal fin formation (32 hpf time point) or shortly after caudal fin sprouting (52 hpf time point) and re-imaged the newly forming caudal fin sprout at 72 hpf. Surprisingly, early photoconversion labelled only a subset of caudal fin ECs (Fig. 4A-C′, quantified in Fig. 4G), while later photoconversion led to almost all caudal fin ECs exhibiting red fluorescence (Fig. 4D-F′, quantified in Fig. 4G). Thus, EC migration from either the PCV or DA contributed to only a subset of ECs to the caudal fin vasculature. We also labelled ECs exclusively at the tip of the newly forming caudal fin vasculature and re-imaged their position 24 h later (Fig. 4H-J). These cells contributed to the newly forming tail fin artery and increased in number, corroborating the findings from our time-lapse imaging and BrdU incorporation studies.

To understand the origin of tail fin ECs in more detail, we performed time-lapse imaging starting at 30 hpf of Tg(fli1:LIFEACT-EGFP)^mu240; Tg(fli1:H2B-mCherry)^uq37bh double transgenic embryos (Movies 12, 13). These movies uncovered newly emerging angioblasts at the tip of the caudal fin that migrated towards the tip of the fin sprout, where they fused with emigrating ECs (Fig. 4K-P). Thus, newly differentiating fli1-expressing angioblasts contribute caudal fin vasculature ECs. Future studies are needed to determine the amount of contribution from these different sources of ECs.

We followed the further expansion of the fin vasculature, which occurred in several distinct anatomical locations. To do so, we used Tg(-0.8flt1:RFP)^hu5333 fish in combination with Tg(flt4:citrine)^hu7135. In the distal fin, we observed continued vein-artery sprouting (Fig. 5A,A″,B, blue arrowhead). This mode of sprouting continued as the fin expanded (Fig. 5C,C′, Fig. S8). Thus, part of the fin vasculature forms through the repetitive addition of new artery-vein loops. We detected another, more proximally located, fin region that showed vascular expansion (Fig. 5A,A′,B, C′, red arrowheads). In this area, we did not observe the formation of artery-vein loops, but rather the formation of a plexus (Fig. 5D-F′). ECs within this expanding plexus expressed varying levels of both the flt1:RFP and flt4:citrine transgenes (Fig. 5D-F′). Therefore, we can distinguish at least two distinct mechanisms that control fin blood vessel formation.

The proximal plexus remodelled into two separate sets of arteries that ran on either side of the fish body at 3 wpf (Fig. 6A-A″). At this time point, another vascular plexus started to form at the base of the fanning fin vasculature, where a major vein and artery run in parallel in a dorsal to ventral fashion (Fig. 6B,B′). We named this vascular structure ‘interlaced plexus’ (IP) to accommodate its highly ramified morphology. We detected new vascular connections that ran next to the pre-existing artery-vein loops and extended from proximal into distal fin regions (Fig. 6C,C′, pseudo-coloured in red). They frequently displayed blind-ending tips that reached into more distal fin regions (Fig. 6C′, red arrowheads). We found connections between these vessels and the bone-enclosed arteries (BEAs) (Fig. 6D,D″, yellow arrowheads). They further lacked expression of the flt4:citrine transgene (Fig. 6D), had a larger diameter compared with BEAs (Fig. 6E,F) and showed lower levels of the arterial marker flt1:RFP (Fig. 6G). To characterize these vessels further, we used Tg(pdgfrb:gal4ff)^ncv24tg; Tg(UAS:GFP)^nkuasgfp1a; Tg(-0.8flt1:RFP)^hu5333 triple transgenic animals to label vascular mural cells (Fig. 6H,H′). Different types of mural cells associate with distinct blood vessels, with smooth muscle cells being found around arteries, while pericytes invest capillaries (Ando et al., 2021). This analysis revealed that vessels of the IP readily contained pdgfrb-expressing mural cells, while their diameters were larger than those of BEAs (Fig. 6F,H). We analysed mural cell occupancy in more detail and found acta2-expressing smooth muscle cells on the BEA, while acta2 expression was absent in mural cells of the IP (Fig. 6I,I′). Thus, although mural cell on both the BEA and the IP expressed pdgfrb, only mural cells on the BEA also expressed acta2. Analysis of blood flow patterns revealed a higher number of circulating red blood cells within the vessels of the IP when compared with BEAs as visualized through lower fluorescence intensity of injected qdots (Fig. 6J-J″, yellow arrowheads indicate blood cells) and bright-field movies of red blood cells (Movies 14-17). Together, our analysis reveals that fin vascular morphogenesis generates distinct types of blood vessels. An initial wave generates artery-vein loops that are positioned in relation to other morphological features, such as bone rays, followed by a second wave of angiogenic sprouting that generates less defined vascular plexuses. We can distinguish plexus vessels in terms of transgenic marker gene expression, diameters, differences in mural cell occupancy and blood flow profiles. These successive processes furthermore occur in several different anatomical locations.

At around 4 wpf, we could divide the tail fin vasculature into a dorsal and ventral lobe (Fig. 7A-B″). In the dorsal lobe, arteries of individual fin rays came to lie dorsally to their respective veins, while in the ventral lobe, the arteries ran ventrally. At this stage, we observed a third morphogenetic process that we termed ‘ancillary vein-sprouting’. This process led to the transformation of the initial artery-vein loops into triads of medially located arteries flanked by two veins (Fig. 7C). Ancillary vein sprouting initiated from pre-existing main veins at the centre of the fin ray (Fig. 7C, arrow). Connections from the main vein either crossed the BEA, giving rise to ancillary veins, or connected to the artery. Thus, in contrast to the initial angiogenic sprouting, where veins gave rise to arteries, ancillary vein sprouting exclusively generates veins from pre-existing veins.

We also observed the establishment of new connections between arteries and veins. Here, ECs within veins started to express the flt1:RFP transgene (Fig. 7D) before connecting to arteries (Fig. 7E). Whether this increase in flt1:RFP expression reflects the initiation of an angiogenic program or the onset of arterial differentiation remains to be determined. At the most distal end of the fin, vein cells continued to generate arteries through ‘vein-artery sprouting’ (Fig. 7F, red arrowheads marking arterial ECs and blue arrowheads marking venous ECs). Together, at this developmental time point, four simultaneously occurring, distinct morphogenetic processes expand the caudal fin vasculature along the proximal-distal axis (interlaced plexus sprouting, vein arterialization, ancillary vein sprouting and vein-artery sprouting, summarized in Fig. 7G,H).

After the establishment of the initial fin vasculature and artery-vein connections, we observed another vein-derived expansion of the arterial network through intussusceptive angiogenesis (Fig. 8A-D). ECs at the outer edges of distally located veins expressed lower levels of the flt4:citrine transgene (Fig. 8C, blue arrowheads), while we detected higher flt1:RFP expression in these cells (Fig. 8C, red arrowheads). Of note, we detected cytoplasmic pillars, which are characteristic of intussusceptive angiogenesis (De Spiegelaere et al., 2012) (Fig. 8C′, demarcated by dashed red lines). Ultimately, as far as we can judge from analysing static images, flt1:RFP-expressing ECs detached from their vein of origin and coalesced into small diameter arteries that now ran adjacent to the pre-existing vein and connected to other artery-vein connections (Fig. 8D, red arrowheads, summary in Fig. 8G). Thus, intussusceptive angiogenesis is the fifth morphogenetic process we detected in the expanding fin vasculature. As we analysed only static images, it is currently unclear at which stage these ECs might change the expression of either flt4 or flt1 and initiate intussusceptive angiogenesis.

In more proximal areas, we observed the formation of a more intricate network of smaller diameter interconnections between arteries and veins, spanning the inter-ray tissue. Here, ECs at the tips of crossing sprouts more strongly expressed the flt4:citrine transgene when compared with the flt1:RFP transgene (Fig. 8E-E″, red arrowheads indicate flt1:RFP-expressing cells; blue arrowhead marks a flt4:citrine-expressing cell). Cells at the base of the sprouts either connected to a vein and expressed flt4:citrine or to an artery and expressed flt1:RFP (Fig. 8E-E″). In established connections and when moving from the arterial towards the venous side of the vascular tree, we observed a gradual transition from flt1:RFP transgene expression (Fig. 8F,F′, red arrowheads) to flt4:citrine transgene expression (Fig. 8F,F′, blue arrowheads). Thus, the later stages of fin vascular development are characterized by an elaboration of the initially laid down framework of alternating arteries and veins (Fig. 8G). Distally located vessels mainly establish arterial side branches that connect to veins within a given fin ray through intussusceptive angiogenesis, whereas more proximal fin regions generate additional connections through sprouting angiogenesis that span between fin rays (a summary of all morphogenetic processes is provided in Fig. 9).

During embryonic development, every organ of the body needs to be vascularized. This entails the tight coupling of tissue development with vascular growth (Augustin and Koh, 2017). Although many of the genetic players orchestrating vascular growth and blood vessel sprouting, such as Vegf, Notch, Wnt and BMP signalling have been identified (Adams and Alitalo, 2007; Eelen et al., 2020; Potente et al., 2011; Red-Horse and Siekmann, 2019; Schuermann et al., 2014), our insights into the mechanisms controlling the stepwise assembly of the vasculature of a given organ and how these might regulate vascular variability have remained limited. To investigate the different morphogenetic processes that are required to shape the adult vasculature of an organ and to understand how these processes might build upon each other, we decided to study the developing caudal fin vasculature of zebrafish. We chose this vascular bed because the fin is relatively two-dimensional and using transparent zebrafish allows the imaging of vascular network assembly at later developmental time points (Bump et al., 2022; Paulissen et al., 2022; Xu et al., 2014, 2015). Our analysis revealed that the initial vascular loop in the fin formed through angiogenic sprouting from the posterior cardinal vein (PCV). In this process, vein-derived ECs relied on VEGF to turn on arterial marker gene expression and change their direction of migration, ultimately turning towards the dorsal aorta in a Notch- and cxcr4a-dependent manner. This sequence of events that we now term vein-artery sprouting was reiterated during the subsequent outgrowth of the fin vasculature into the expanding fin fold, generating an artery-vein loop within each fin ray. Our observations are similar to earlier published data examining the outgrowth of the caudal fin vasculature (Bump et al., 2022; Paulissen et al., 2022).

We previously identified the same genetic program during eye blood vessel development in zebrafish embryos and in the mouse retina, describing a process through which vein-derived angiogenic ECs form new arteries (Hasan et al., 2017; Pitulescu et al., 2017; Xu et al., 2014). Similar vein-to-artery conversions have been described for the coronary arteries (Red-Horse et al., 2010; Su et al., 2018), in early mouse embryos (Hou et al., 2022) and additional studies in the mouse retina (Lee et al., 2021; Park et al., 2021). In the pectoral fin, two vein sprouts initially emerge from the common cardinal vein and fuse along the rim of the endoskeletal disc (Paulissen et al., 2022). Later, a new sprout grows from this vein loop towards the dorsal aorta. Thus, also in the pectoral fin, veins are the source for new ECs, followed by artery specification and migration of a new sprout towards a pre-existing artery. This sprouting depended on Notch signalling (Paulissen et al., 2022) and we speculate that sprouting towards the dorsal aorta will also depend on cxcr4a signalling. Together, these findings underscore the conserved nature of this mode of blood vessel formation that generates perfused artery-vein loops, in which arteries and veins can be distinguished by virtue of marker gene expression, diameter and blood flow (Red-Horse and Siekmann, 2019). However, this configuration of direct artery-vein loops allows arterial flow patterns to be directly transmitted into venous blood vessels, as we do not observe capillary beds at this time point in the fin. Therefore, the initial expansion of the fin vasculature generates vascular patterns consisting of what is traditionally viewed as arterio-venous shunts (Pries et al., 2010).

Curiously, we then observed a shift in the mode of blood vessel sprouting. New sprouts again emerged from pre-existing veins, but exclusively generated new veins instead of arteries in a process we call ancillary vein sprouting. These veins migrated in proximal and distal directions along the bone ray and either connected with a pre-existing vein or with the bone-enclosed artery in the most distal fin region, effectively transforming the direct artery-vein connections into a pattern, where one artery now drains into two veins. Therefore, this mode of angiogenic sprouting resolved the initially present arterio-venous shunts. A similar vessel pattern is also observed in pectoral fins (Paulissen et al., 2022). This switch in angiogenic sprouting behaviour thus appears to be a fundamental process to ensure appropriate flow patterns through a later expansion of the venous pole of the vasculature. It will be important in the future to understand how this switch in angiogenic sprouting is achieved and whether it also occurs in other vascular beds. Previous studies have implicated BMPs in controlling vein sprouting in the trunk of zebrafish embryos (Wiley et al., 2011). In this setting, a first wave of angiogenic sprouting generates intersegmental blood vessels that are connected to the dorsal aorta in a process dependent on Vegf and Notch signalling (Hogan and Schulte-Merker, 2017; Nasevicius et al., 2000; Siekmann and Lawson, 2007), whereas in this setting cxcr4a signalling is dispensable (Siekmann et al., 2009). Subsequently, secondary sprouts re-wire this initial artery-only network into arteries and veins (Geudens et al., 2019; Weijts et al., 2018). It is therefore likely that BMPs might also control vein-generating sprouting in the tail fin, as BMPs have similarly been implicated in vein formation in other contexts (Neal et al., 2019). BMP signalling is strongly influenced by blood flow (Baeyens et al., 2016a,b; Vion et al., 2018; Zhou et al., 2012). It will be of interest to investigate whether changing flow patterns in the tail fin might alter signalling within ECs to orchestrate the switch from angiogenic sprouting that generates arteries from veins to a program that exclusively generates veins. Together, our results demonstrate that two distinct angiogenic programs exist that allow the expansion of the venous pole of the vasculature after the initial establishment of artery-vein loops.

We observed an additional set of blood vessels that formed at the base of the fin through a process we term plexus sprouting. These could be distinguished in terms of their diameters, transgene expression, mural cell investment and blood flow patterns. It is currently unclear from which blood vessels these ECs emerge, as they do not show expression of the vein-enriched flt4:citrine transgene. They further express lower levels of the arterial-enriched flt:RFP1 transgene. Of interest, low levels of flt1:RFP expression were recently reported in lymphatic ECs that give rise to blood vessels in the anal fin (Das et al., 2022). It might therefore be conceivable that the cells of the interlaced plexus have a similar lymphatic origin, albeit the absence of flt4:citrine expression argues against this possibility. ECs forming veins in the pectoral and tail fin vasculature also express the lymphatic marker mrc1a:egfp (Paulissen et al., 2022). How these gene expression patterns reflect functional properties of the investigated blood vessels, and their origin remains to be determined. This is of particular interest, as a recent study showed that endoderm-derived vascular progenitors differentiate into ECs with a distinct gene expression signature forming the hematopoietic niche in the zebrafish tail (Nakajima et al., 2023). Of note, caudal fin ECs did not express lyve1, while the PCV showed robust lyve1 expression, suggesting that caudal fin ECs might constitute a separate EC population. Our finding that newly differentiating angioblasts contribute to the caudal fin vasculature underscores this notion. Other studies showed that similar vascular progenitors can contribute to the DA (Nguyen et al., 2014) or the PCV and the subintestinal vessels in zebrafish embryos (Metikala et al., 2022). Thus, de novo differentiating angioblasts might also supply ECs to blood vessels previously thought to arise solely through angiogenesis, which is traditionally defined as the sprouting of new blood vessels from pre-existing ones (Risau, 1997). These findings support decades-old studies performed in avian embryos concluding that new embryonic blood vessels likely form through a combination of angiogenesis and the addition of newly differentiating angioblasts (Noden, 1990).

Finally, we observed two distinct modes of blood vessel formation that expanded arterial collateral vessels in proximal and distal fin regions, respectively. In more distal regions, we observed intussusceptive angiogenesis in veins (De Spiegelaere et al., 2012). This process occurs through the insertion of tissue pillars within blood vessels that subsequently expand, leading to a splitting of a blood vessel into two individual vessels (Burri and Djonov, 2002). Intussusceptive angiogenesis frequently occurs after sprouting angiogenesis and has been observed in the caudal vein plexus of zebrafish embryos (Karthik et al., 2018). Despite the detailed description of the processes involved in intussusceptive angiogenesis, changes in gene expression patterns within ECs undergoing intussusceptive angiogenesis have not been detected at the cellular level. We now observe that individual ECs within veins express the arterial marker flt1:RFP during intussusceptive angiogenesis. In other animals, we find separate blood vessels made up of flt1:RFP-expressing ECs that connected to arterial side branches. As we only analysed static images, we cannot draw firm conclusions concerning the sequence of events occurring during intussusceptive angiogenesis. We can also only speculate as to the function of the flt1:RFP-expressing side branches. They might play a role during blood vessel regeneration to serve as collaterals that would restore blood flow after the main artery has been severed. We have previously observed that these interconnections were enlarged during artery regeneration in animals lacking mural cells (Leonard et al., 2022), with a role for vascular mural cells being described during intussusceptive angiogenesis (Burri et al., 2004). Further studies will be required to fully elucidate their role and what mechanisms control the observed changes in gene expression in ECs undergoing intussusceptive angiogenesis, with changes in haemodynamics playing a potential role (Styp-Rekowska et al., 2011).

In more proximal regions of the tail fin, we observed inter-ray sprouting angiogenesis from veins. Sprouts regularly crossed the space between two adjacent fin rays. They expressed varying levels of arterial and venous marker genes that either formed sharp boundaries or displayed more gradual changes from the arterial towards the venous pole of the vasculature, as also recently described for capillaries forming during mouse embryogenesis (Hou et al., 2022). Therefore, in proximal fin regions, small blood vessels resembling a capillary bed form at later stages of fin vascular expansion. As our studies solely rely on the observation of static images and differences in transgene expression patterns, we can currently only speculate on the origin of the ECs that give rise to new capillary-type side branches. Detailed genetic lineage-tracing or imaging studies will be necessary to investigate these processes in the future.

Together, our longitudinal analysis of the development of the zebrafish tail fin vasculature illustrates how blood vessels assemble in a stepwise manner to irrigate a mature organ. Our results furthermore show how different morphogenetic processes are either reiteratively used, as for the generation of arteries from veins, or build upon each other, as in the case of the expansion of the venous vasculature to transform arterio-venous shunts into a configuration where one artery drains into two veins. How these underlying principles organize the formation of blood vessels in other organs, as well as the signalling pathways controlling these steps, and the mechanisms through which they are being activated remains to be determined.

Zebrafish embryos were maintained in 1× E3 medium under recommended husbandry conditions (Westerfield, 1993). Embryos were transferred to tanks in the zebrafish facility and raised until further analysis. Previously described zebrafish lines were Tg(-0.8flt1:RFP)^hu5333 (Bussmann et al., 2010), TgBAC(pdgfrb:gal4ff)^ncv24 (Ando et al., 2016), Tg(Tp1bglob:VenusPEST)^s940 (Ninov et al., 2012), Tg(Tp1bglob:H2B-mCherry)^s939 (Ninov et al., 2012), Tg(fli1:LIFEACT-EGFP)^mu240 (Hamm et al., 2016), Tg(fli1a:nEGFP)^Y7 (Roman et al., 2002), Tg(gata1:dsRed)^sd2 (Traver et al., 2003), TgBAC(cxcr4a:YFP)^mu104 (Xu et al., 2014), Tg(flt4:citrine)^hu7135 (van Impel et al., 2014), Tg(kdrl:EGFP)^s843 (Jin et al., 2005), kdrl^hu5088 (Bussmann et al., 2010), vegfaa^mu128 (Lange et al., 2022), Tg(acta2:mCherry)^ca8Tg (Whitesell et al., 2014), TgBAC(dab2:EGFP)^ncv67Tg (Shin et al., 2019), Tg(mrc1a:EGFP)^y251Tg (Jung et al., 2019), Tg(-5.2lyve1b:DsRed)^cq27Tg (Chen et al., 2019), Tg(sp7:EGFP)^b1212Tg (DeLaurier et al., 2010) and Tg(fli1:EGFP)^y1 (Lawson and Weinstein, 2002). Some of the transgenic lines were kept in a casper (roy, nacre) double-mutant background (White et al., 2008) to increase clarity for imaging due to loss of pigment cells. All animal experiments were performed in compliance with the relevant laws and institutional guidelines and were either approved by local animal ethics committees of the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen or in accordance with protocols approved by the University of Pennsylvania Institutional Animal Care and Use (number 806819). Zebrafish veterinary care was performed under the supervision of the University Laboratory Animal Resources at the University of Pennsylvania.

Dendra2 was amplified from pCS2+dendra2 (a gift from Periklis Pantazis, Imperial College London, UK) using primers dendra2_attB1-GGGGACAAGTTTGTACAAAAAAGCAGGCTCCACCatgaacaccccgggaattaac and dendra2_attB2-GGGGACCACTTTGTACAAGAAAGCTGGGTATTAccacacctggctgggcagggg to generate gateway cloning (ThermoFisher Scientific) compatible PCR products (uppercase letters indicate attb sites and lowercase letters indicate dendra2 sequence). The PCR product was transferred into pDONR221 through a BP reaction and sequence verified. Dendra2 was then cloned into pTolfliepDest (Villefranc et al., 2007). Zebrafish embryos were injected with 30 ng of purified pTol2fliepdendra2Dest plasmid DNA together with 50 ng of Tol2 mRNA. Embryos showing robust endothelial expression of dendra2 were grown to adulthood and outcrossed to wild-type fish. F1 animals were used for the studies in this article.

Embryos (2-3 dpf) were embedded in 1% low melting point agarose containing 168 mg/l tricaine in a glass-bottomed dish. Embryos were imaged for the desired period of time with 20 min imaging intervals. A heated microscope stage set to 28.5°C was used to maintain the appropriate temperature. Larvae from 4 dpf to 4 wpf were anaesthetized in tricaine and later embedded in 1% low melting point agarose. For imaging of 1.5-month-old fish, caudal fins were amputated from anaesthetized fish and embedded in 1% low melting point agarose for imaging. Angiography was performed by injecting qdots 705 into the hearts of 4 wpf fish.

SU5416 (Millipore Sigma) was dissolved in DMSO and stored at −20°C. Embryos were dechorionated and treated with 2.5 μM of SU5416 or 0.1% DMSO in E3 media beginning at 48 hpf until 72 hpf. DAPT (Cell Signaling Technology) was dissolved in DMSO and stored at −20°C. Dechorionated embryos were treated with 100 μM of DAPT or 0.1% DMSO from 48-72 hpf. The drug solutions were refreshed every 12 h. Embryos were incubated in E3 media containing 0.003% phenylthiourea to inhibit pigmentation.

Images for blood flow videos were acquired by mounting an Alvium 1800 U-158 m/c [Sensor (Sony IMX273)] camera on the eye piece of a Leica Sp8 confocal microscope. The acquired images were processed in ImageJ (NIH) to generate movies of the same frame rate.

To perform BrdU pulse-chase experiment, dechorionated embryos were incubated in 10 mM of BrdU for 20 min in the cold room. After incubation the embryos were washed three times with E3 medium. BrdU labelled embryos were incubated at 28°C in E3 medium containing PTU until 72 hpf. At 72 hpf the embryos were fixed in 4% PFA overnight at 4°C. On the following day, embryos were washed two times with 1 ml of 100% methanol and stored in 100% methanol at −20°C overnight. Immunostaining was performed the following day as previously described (Nicoli et al., 2012), using primary BrdU monoclonal antibody (Invitrogen, MA3-071; 1:200) and primary EGFP antibody to detect ECs (1:200; Torrey Pines Biolabs, anti GFP rabbit polyclonal TP401). Secondary antibodies were Alexa Fluor 647 goat anti-mouse IgG for BrdU (Invitrogen, A21235; 1:500) and Alexa Fluor 488 goat anti-rabbit to detect EGFP (Invitrogen, A11008; 1:400).

Photoconversion was performed on a Zeiss LSM 780 confocal microscope. Anaesthetized larvae were embedded in 0.75% agarose containing 0.02% tricaine. The region to be photoconverted was marked using the regions tool in the ZEN software. Photoconversion was performed with 5% intensity of the 405 nm laser with 40 iterations, for 25 cycles with 2.5 ms intervals. After photoconversion, larvae were placed in E3 medium and reimaged at 72 hpf.

Fluorescent confocal images were acquired using a Zeiss LSM 780 (objective lens: 20× Plan Apo NA 0.80) or Leica SP8 (objective lens: HC PL Fluotar 20×/0.50 or HCX APO L 63×). All images shown are representative of the results obtained for each group and experiment.

Zeiss ZEN software or Leica LAS-X were used to stitch the acquired images. Imaris software (Oxford Instruments) was used to generate maximum intensity projections. 3D projections were made by creating surfaces in Imaris. ImageJ (NIH) was used to generate monochrome images of different fluorophores. Adobe Illustrator and Adobe Photoshop software were used to compile images and create schematics.

Nuclei numbers were manually counted using the Imaris software suite (Oxford Instruments) in blood vessels of approximately the same length. Diameters of interlaced plexus vessels and bone enclosed arteries was calculated by measuring at least 10 regions in segments of equal length. Fluorescence intensity was calculated using maximum intensity projections. Individual cells were marked, and the intensity values were obtained from ImageJ (NIH). Data were analysed using Prism 9 (Graphpad) and graphs were plotted with mean standard deviation (s.d.). P<0.05 was considered statistically significant.

We thank Reinhild Bussmann, Mona Finch-Stephen, Bill Vought and Nadine Greer for excellent fish care. We also thank Mike Pack and Zeenat Diwan for critical reading of the manuscript. We also acknowledge support from the Imaging Core facility of the Department of Cell and Developmental Biology of the University of Pennsylvania.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201030.reviewer-comments.pdf