Arterial-Venous Segregation by Selective Cell Sprouting: An Alternative Mode of Blood Vessel Formation

See allHide authors and affiliations

Science  09 Oct 2009:
Vol. 326, Issue 5950, pp. 294-298
DOI: 10.1126/science.1178577


Blood vessels form de novo (vasculogenesis) or upon sprouting of capillaries from preexisting vessels (angiogenesis). With high-resolution imaging of zebrafish vascular development, we uncovered a third mode of blood vessel formation whereby the first embryonic artery and vein, two unconnected blood vessels, arise from a common precursor vessel. The first embryonic vein formed by selective sprouting of progenitor cells from the precursor vessel, followed by vessel segregation. These processes were regulated by the ligand EphrinB2 and its receptor EphB4, which are expressed in arterial-fated and venous-fated progenitors, respectively, and interact to orient the direction of progenitor migration. Thus, directional control of progenitor migration drives arterial-venous segregation and generation of separate parallel vessels from a single precursor vessel, a process essential for vascular development.

During early stages of vertebrate embryogenesis, coordinated sorting and segregation of arterial- and venous-fated angioblasts into distinct networks of arteries and veins is essential to establish a functional vasculature. Recent studies in mouse and zebrafish have elucidated key roles for a number of signaling pathways and transcriptional regulators in arterial-venous specification (1, 2). However, we still lack a basic mechanistic understanding of how mixed populations of specified arterial-venous cells coordinate their behavior to segregate and form distinct vessels.

During zebrafish vascular development, angioblasts migrate from the lateral plate mesoderm to the midline (3, 4) and eventually give rise to the first embryonic artery (dorsal aorta, DA) and vein (caudal vein, CV). Notochord-derived Sonic hedgehog, which induces the expression of vascular endothelial growth factor a (Vegfa) in the ventral somites, is essential for angioblast differentiation (5). Vegfa-induced activation of Notch signaling (5, 6), as well as other factors (2, 79), subsequently promotes arterial specification in a subset of angioblasts (4), before arterial-venous segregation. To investigate mechanisms of arterial-venous angioblast sorting and segregation, we analyzed vascular development with high temporal resolution in Tg(kdrl:GFP)s843 (4) embryos (10). These transgenic embryos express green fluorescent protein (GFP) in the endothelial lineage, enabling the tracking of angioblasts during early stages of vascular development. GFP-positive angioblasts coalesced at the midline to form a single vascular cord by the 21-somite stage (19.5 hours postfertilization; hpf; Fig. 1, A and B, and fig. S1). Subsequent cord remodeling led to the formation of a lumenized DA primordium by 22 hpf. From 21 to 23 hpf, a subpopulation of angioblasts sprouted ventrally from the DA primordium and connected (anastomosed) with adjacent cells to form the CV primordium by 24 hpf (Fig. 1, A to C; figs. S1, S2, and S4; and movies S1 and S2). Hence, the first artery and vein share a common vessel primordium. Dorsal intersegmental vessel (ISV) sprouting occurred at regular intervals along the DA and was initiated later than ventral sprouting (23 hpf; Fig. 1B), suggesting that distinct mechanisms govern dorsal versus ventral sprouting behaviors. Thus, the DA forms by classical vasculogenesis, whereas formation of the CV involves an alternative mechanism whereby selective sprouting of venous-fated angioblasts, subsequent sprout termination, and cell-cell segregation allow distinct arterial and venous vessels to form. Hence, unlike angiogenesis in which new capillaries form a continuous network with the original vessel, two unconnected blood vessels can derive from a common group of cells.

Fig. 1

The CV forms by selective angioblast sprouting. (A to E) Mid-trunk transverse sections [(A) and (D)], lateral views [(B) and (E)], or lateral time-lapse [(C) and movie S1] of Tg(kdrl:GFP)s843 [(A) to (C)] or Tg(kdrl:GFP)s843;Tg(gata1:dsRed)sd2 [(D) and (E)] embryos. Angioblasts coalesce and remodel to form the DA by 22 hpf [white and red brackets; (A) to (C)]. Between 21 and 24 hpf, venous angioblasts sprout ventrally from the DA [arrowheads and dotted lines, (A) to (C)] and contribute to the CV primordium [yellow and blue brackets, (A) to (C)]. ISVs sprout dorsally from 23 hpf [arrows, (B)]. By 25 hpf, venous angioblasts surround Tg(gata1:dsRed) sd2-positive erythrocytes [asterisks, (A) and (D); red cells, (D) and (E)]. Erythrocyte displacement typically clears the CV lumen by 26 hpf [(A), (D), and (E)] but not in tnnt2 MO–injected embryos (E). Scale bars indicate 35 μm.

Differences between DA and CV tube formation were further highlighted by contrasting modes of lumen formation. DA lumen formation involved hollowing of the vascular cord (11), whereas a functional CV lumen was generated differently (Fig. 1, A, D, and E; figs. S1, S2, S5, and S6; and movies S3 and S4). By using Tg(gata1:dsRed)sd2 embryos that express the fluorescent protein dsRed in erythrocytes (12) and selective plane illumination microscopy (13), we found that initial hollowing of the CV primordium was followed by rapid luminal expansion upon the invasion of erythrocytes positioned ventral to the DA (fig. S6 and movies S3 and S4). Shortly thereafter, the CV lumen was cleared upon the flow-dependent displacement of erythrocytes, a process that failed to occur in cardiac troponin t2 (tnnt2) morpholino oligonucleotide (MO)–injected embryos, which lack circulation (14) (Fig. 1E).

We hypothesized that arterial-venous segregation and generation of two separate vessels might be achieved by selective incorporation of venous-fated angioblasts into ventral sprouts while limiting the ventral migratory behavior of arterial-fated angioblasts. Vegfa was recently shown to differentially regulate angioblast sprouting behavior by promoting the expression of the Notch ligand delta-like 4 (dll4) (1517). Expression of Dll4 in ISV tip cells activates Notch signaling in adjacent cells, thereby limiting their dorsal sprouting behavior. Injection of embryos with vegfa MO revealed that Vegfa signaling also limits the ventral migratory behavior of angioblasts (Fig. 2, table S1, fig. S7, and movies S5 and S6). In vegfa MO–injected embryos, angioblasts initially coalesced to form a vascular cord but then migrated ventrally excessively, failed to segregate, and ultimately generated a single vascular tube encircling Tg(gata1:dsRed)sd2-positive erythrocytes. Hence, Vegfa-mediated arterial specification is required for the control and termination of angioblast ventral migration and maintenance of an intact DA. A similar phenotype was observed on injection of MO to plcγ1, which encodes a key downstream component of Vegfa signaling (6), and on inhibition of Vegfa receptor activation using SU5416 (Fig. 2, table S1, fig. S7, and movies S5 and S7). ISV sprouting was also disrupted in Vegfa-deficient embryos, consistent with opposing roles for Vegfa signaling in the induction of dorsal sprouting (6) and restraint of ventral sprouting. Vegfa influences dorsal sprouting behavior by indirectly activating Notch signaling (1517). Accordingly, N-[N-(3,5-difluorophenacetyl)-1-alanyl]-S-phenylglycine t-butyl ester (DAPT)–mediated inhibition of Notch signaling promoted excessive ventral sprouting (Fig. 2B). Moreover, excessive ventral migration was also observed on injection of MO to gridlock/hey2, which encodes a key transcriptional effector of notch signaling (7). Thus, termination of ventral sprouting is regulated by genes downstream of Vegfa and Notch signaling.

Fig. 2

Vegf signaling limits ventral sprouting. (A and B) Lateral views of Tg(kdrl:GFP)s843 embryos (A) or transverse sections of Tg(kdrl:GFP)s843;Tg(gata1:dsRed)sd2 embryos (B) injected with control, vegfa, plcγ1 or hey2 MO, or exposed to dimethyl sulfoxide (DMSO), SU5416, DAPT, or LY294002. vegfa or plcγ1 MO–injection blocks ISV sprouting (asterisks), promotes excessive ventral sprouting (arrowheads), leads to loss of the DA (white bracket) and generates a single venous tube (yellow bracket). SU5416, DAPT, or hey2 MO–injection also promote loss of the DA, whereas LY294002 blocks CV morphogenesis. Scale bars, 35 μm.

In contrast, vein morphogenesis was strikingly disrupted by the broad range phosphoinositide 3-kinase (PI3K) inhibitor LY294002 (Fig. 2B). Although LY294002 treatment did not affect total angioblast numbers, suggesting no proliferation defects (table S2), ventral sprouting was disrupted, and significantly more cells were retained in the DA (Fig. 3, A to C), consistent with a role for PI3K in angioblast ventral sprouting [supporting online material (SOM) text S1]. Cell migration during angiogenic sprouting selectively requires signaling via the p110α isoform of PI3K downstream of Vegf receptor activation (18). Consequently, we tested the hypothesis that a similar p110α-dependent mechanism may control ventral sprouting and CV formation. We interrogated the group I PI3K family (p110α, p110β, p110δ, and p110γ) with isoform-selective inhibitors (19, 20) (table S3) to avoid the early developmental defects and compensatory interactions associated with genetically silencing individual isoforms (19, 21, 22). Consistent with a role for p110α, exposure of embryos to inhibitors at concentrations that inhibit p110α blocked vein morphogenesis (fig. S12 and SOM text S2). Application of LY294002 or the isoform selective inhibitor, AS605240, at concentrations that blocked ventral sprouting also disrupted dorsal sprouting (Fig. 3, A to C), indicating a general requirement for PI3K in arterial-venous sprouting.

Fig. 3

Flt4/PI3K signaling mediates venous angioblast migration. (A to C) Tg(kdrl:GFP)s843 embryos were exposed to DMSO, LY294002, or AS605240, processed for microscopy [23 hpf, (A)], and then scored for (B) the number of ventral sprouts (as a percentage of DMSO-treated embryos) or (C) the percentage of cells in the DA (white bracket, a in graph) or CV (yellow bracket, v in graph). PI3K inhibition blocks ISV sprouting (arrows), venous sprouting (asterisks), and angioblast ventral migration (arrowheads). Lateral images (D) and quantification (E) of donor-derived Tg(kdrl:GFP)s843 cells in wild-type hosts at 48 hpf. Fewer cells from flt4 MO–injected donors contribute to the CV versus controls. Excessive ventral migration of angioblasts from plcγ1 MO–injected donors is reduced on coinjection with flt4 MO. Scale bars, 35 μm. Error bars represent mean ± SEM. *P < 0.05 versus control, **P < 0.05 versus plcγ1 MO, Student’s t test.

A common role for p110α in dorsal and ventral angioblast sprouting indicated that similar signaling pathways regulate these processes. Vegfa promotes dorsal ISV sprouting on activation of its receptor, Flk1, and downstream p110α-mediated signaling (18). Restricted venous expression of the related Vegfc receptor gene flt4 (fig. S8) suggested a role for p110α in vein morphogenesis downstream of Flt4-mediated signaling. Although angioblast ventral sprouting and migration was significantly delayed in vegfc or flt4 MO–injected embryos (fig. S13 and table S4), possible functional redundancy with other Vegf signaling components or compensation by alternate mechanisms enabled vein formation at later time points. However, a role for Flt4 signaling in angioblast ventral sprouting was confirmed on transplantation of cells from Tg(kdrl:GFP)s843 donors into nontransgenic wild-type hosts. Significantly fewer flt4 MO–injected than control MO-injected, donor-derived cells contributed to the CV (Fig. 3, D and E, and table S5). Furthermore, the excessive contribution of plcγ1 MO–injected, donor-derived cells to the CV was blocked on flt4 MO–co-injection. Hence, Vegfa/Flk1 and Vegfc/Flt4 signaling regulate multidirectional arterial-venous sprouting. However, vegfc is not expressed ventral to the DA but in the DA itself (fig. S14), indicating that other mechanisms must orient the direction of venous angioblast migration ventrally.

The EphB4 receptor tyrosine kinase and its plasma-membrane-spanning ligand, EphrinB2 (EfnB2), demarcate venous and arterial domains, respectively (23). Studies of EphB4- and EfnB2-null mice have revealed roles for forward and reverse EphB4-EfnB2 signaling in vascular morphogenesis (2325); however, the exact function and mode of action of bidirectional EphB4-EfnB2 signaling remain unknown (1). Observations that zebrafish Efnb2a expression is restricted to the dorsal-most cells of the vascular cord (4) (fig. S15) and is tightly regulated by Vegfa (figs. S11 and S16) as well as Notch signaling (57) led us to hypothesize a role for Ephb4a-Efnb2a in the directional control of angioblast migration. Consistent with this hypothesis, the generation of distinct arterial and venous vessels was disrupted, and the PI3K-dependent ventral migration of angioblasts deregulated, in efnb2a or ephb4a MO–injected embryos (Fig. 4, A and B, fig. S17, and table S6). In contrast to vegfa MO–injected embryos, efnb2a and ephb4a MO–injected embryos still formed ISVs (Figs. 2A and 4, A and B), indicating that angioblasts migrate both dorsally and ventrally but fail to segregate. To elucidate the distinct functions of Efnb2a and Ephb4a in arterial-venous segregation, we transplanted cells from Tg(kdrl:GFP)s843 donors into nontransgenic wild-type hosts. efnb2a MO–injected donor-derived cells contributed to the CV at over twice the frequency of control cells (Fig. 4C, table S6, and figs. S18 and S19). In contrast, significantly fewer ephb4a MO–injected, donor-derived cells contributed to the CV compared to controls. Thus, Efnb2a limits the ventral migration of arterial angioblasts, whereas Ephb4a promotes the ventral migration of venous angioblasts, suggesting a role for bidirectional Ephb4a-Efnb2a–mediated repulsion between arterial and venous cells in the directional control of angioblast sprouting (fig. S19). Furthermore, donor cells overexpressing efnb2a or ephb4a preferentially contributed to ISVs, consistent with increased repulsion of arterial angioblasts away from ephb4a-expressing venous cells in response to elevated Efnb2a or a potential forced Ephb4a-Efnb2a cis-interaction in response to elevated Ephb4a (Fig. 4C, table S6, and figs. S18 and S19). Donor cells overexpressing reverse signaling–deficient Efnb2a (26) (ΔC-Efnb2a) were not retained in the DA (Fig. 4C, table S6, and figs. S18 and S19), confirming a role for reverse Ephb4a-Efnb2a signaling in the regulation of arterial-venous segregation. Hence, Vegfa-induced efnb2a expression and bidirectional Ephb4a-Efnb2a signaling selectively exclude arterial angioblasts from venous sprouts while promoting the ventral migration of venous angioblasts. These data support the concept that Notch, EfnB2, and EphB4 can regulate the size of developing arteries and veins and that endothelial cells from the DA may contribute to the formation of the CV in mouse (27). Thus, repulsive Ephb4a-Efnb2a signaling regulates the directional control of angioblast sprouting behavior.

Fig. 4

Ephb4a-Efnb2a signaling determines the direction of angioblast migration. (A and B) Lateral projections of Tg(kdrl:GFP)s843 embryos (A) or transverse sections of Tg(kdrl:GFP)s843;Tg(gata1:dsRed)sd2 embryos (B) injected with control, efnb2a, or ephb4a MO. Reduced Efnb2a expression promotes excessive ventral sprouting [(A), arrowheads], loss of the DA [(B), white bracket], and formation of a single vessel [(A) and (B), yellow bracket] without affecting initiation of ISV sprouting [(A), asterisks]. Similarly, posterior arterial-venous segregation is disrupted in ephb4a MO–injected embryos. (C) Quantification of donor-derived Tg(kdrl:GFP)s843 cells in wild-type hosts at 30 hpf. More cells from efnb2a MO–injected or ΔC-efnb2a-overexpressing donors contribute to the CV versus controls, whereas fewer cells from ephb4a MO–injected donors contribute to the CV. Overexpression of efnb2a or ephb4a enhances the number of donor cells contributing to ISVs. Scale bars, 35 μm. Error bars represent mean ± SEM (*P < 0.05, Student’s t test).

Previous work focused on identifying signals that modulate specification of arterial-venous angioblasts, but how these signals ultimately affect cell behavior to sort and segregate angioblasts into distinct vessels was unknown. Our data provide a cellular framework for arterial-venous tube morphogenesis whereby coordinated regulation of dorsal and ventral angioblast sprouting behaviors by Vegf, Notch, and Ephb4a-Efnb2a signaling efficiently fashions the embryonic vasculature (fig. S20 and SOM text S3) and shed light on how multidirectional arterial and venous sprouting can be achieved while maintaining the integrity of a precursor vessel. Most importantly, we have uncovered an alternative mode of vascular development that allows the segregation of cells into discrete arterial and venous vessels from one common precursor vessel.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S20

Tables S1 to S6

Movies S1 to S7

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank J. E. Cooke, S. Schulte-Merker, and H. Okamoto for reagents and D. Shin, P. Gut, and S. Woo for critical reading of the manuscript. S.P.H. is a Wellcome Trust Sir Henry Wellcome postdoctoral fellow. J.H. was supported by the Human Frontier Science Program Organization and is supported by a Cardiovascular Research Institute fellowship. This study was supported in part by grants from the NIH (HL54737) and Packard Foundation to D.Y.R.S.
View Abstract

Navigate This Article