Sphingosine-1-phosphate Signaling Promotes Critical Migratory Events in Vasculogenesis*

Here we have investigated the role of sphingosine-1-phosphate (S1P) signaling in the process of vasculogenesis in the mouse embryo. At stages preceding the formation of blood vessels (7.5–8 dpc) in the embryo proper, yolk sac, and allantois, the S1P receptor S1P2 is expressed in conjunction with S1P1 and/or S1P3. Additionally, sphingosine kinase-2 (SK2), an enzyme that catalyzes the formation of S1P, is expressed in these tissues throughout periods of vasculogenesis. Using the cultured mouse allantois explant model of blood vessel formation, we found that vasculogenesis was dependent on S1P signaling. We showed that S1P could replace the ability of serum to promote vasculogenesis in cultured allantois explants. Instead of small poorly reticulated clusters of rounded endothelial cells that formed under serum-free conditions, S1P promoted the formation of elongated endothelial cells that arranged into expansive branched networks of capillary-like vessels. These effects could not be reproduced by vascular endothelial growth factor or basic fibroblast growth factor administration. The ability of S1P to promote blood vessel formation was not due to effects on cell survival or on changes in numbers of endothelial cells (Flk1+/PECAM+), angioblasts (Flk1+/PECAM-), or undifferentiated mesodermal cells (Flk1-/PECAM-). The S1P effect on blood vessel formation was attributed to it promoting migratory activities of angioblasts and early endothelial cells required for the expansion of vascular networks. Together, our findings suggest that migratory events critical to the de novo formation of blood vessels are under the influence of S1P, possibly synthesized via the action of SK2, with signaling mediated by S1P receptors that include S1P1, S1P2, and S1P3.

Sphingosine-1-phosphate (S1P) 1 is a bioactive sphingolipid generated by the action of sphingosine kinase on sphingosine. S1P signaling is mediated by a family of G protein-coupled receptors, of which the prototypical member of the family is S1P 1 (1). S1P 1 was initially described as a protein whose expression was up-regulated during endothelial cell differentiation and angiogenesis (2). Subsequent to the identification of S1P 1 , four other structurally similar proteins (i.e. S1P 2 -S1P 5 ) were identified as S1P receptors (3)(4)(5)(6). All of the S1P receptors mediate signaling via heterotrimeric G protein family members (7). S1P 1 couples stringently to the G i family of heterotrimeric G proteins (e.g. G i , G i1 , G i3 , G o , and G z ), whereas S1P 2 and S1P 3 couple to the G i , G q , and G 13 G protein families (7). In endothelial cells, multiple interconnected S1P signaling pathways have been described that work through S1P 1 and S1P 3 to induce morphogenesis (i.e. in vitro capillary network formation) (8). Furthermore, S1P binding to S1P 1 induces signaling through G i and in turn activates Ras, ERK-2, and a cell survival pathway (8). S1P 1 can also act through an unknown G protein to activate Rac and stimulate cortical actin formation, activate Rho, and as a result stimulate the formation of stress fibers and the assembly of adherens junctions (8). Alternatively, S1P binding to S1P 3 can induce signaling through G 13 or G q to activate Rho and thus promote the formation of stress fibers and adherens junctions (8). S1P 2 signals via the same G proteins as does S1P 3 but has inhibitory effects on cell motility (9,10). S1P 2 can also activate adenylate cyclase, c-Jun Nterminal kinase, and p38 mitogen-activated protein kinases via an unknown mechanism (11). S1P 4 and S1P 5 display a more restricted pattern of expression than other S1P receptors with S1P 4 mainly expressed in hematopoietic cells (12), and S1P 5 is expressed in neuronal cells (6,13). S1P 4 couples to G i and G 12/13 and regulates cell shape and motility (14). S1P 5 couples via G i and G 12 and regulates cell rounding and neurite retraction (6,13).
Considerable evidence exists to indicate that S1P signaling influences an array of vascular cell behaviors. For instance, S1P promotes endothelial cell proliferation (15), migration (16), chemotaxis (17), cytoskeletal reorganization (8), adherens junction assembly (1,8), and endothelial tube/lumen formation (18,19). In addition, the defective vascular maturation observed in mice deficient in the S1P receptor, S1P 1 , highlights a fundamental role for S1P signaling associated with smooth muscle cell investment of nascent vessels (20). Furthermore, mice deficient in the ␣-subunit of G 13 (G␣ 13 ) display impaired vascular development (21). Whereas the collective data support a role for S1P signaling in angiogenic and maturation phases of neovascularization, little is known as to whether S1P signaling is important for vasculogenesis, the process of de novo formation of blood vessels from mesodermal progenitor cells. Herein, we establish that S1P receptors and sphingosine kinase are expressed in prevascularized embryonic tissues and throughout the stages of vasculogenesis, thus supporting a role for S1P signaling in de novo blood vessel formation. Furthermore, using the murine allantois explant culture model of vasculogenesis (22) we demonstrated a requirement for S1P signaling in vasculogenesis.
Reverse Transcription and PCR of RNA Isolated from Mouse Embryos, Yolk Sacs, and Allantoides-Embryos (7.5-9.5 days post coitum (dpc)) were removed from timed pregnant CD1 mice (Harlan, Indianapolis, IN) and placed into 4°C Dulbecco's phosphate-buffered saline (dPBS, Cellgro, Herndon, VA). Yolk sacs and allantoides were dissected away from embryos, and each was placed into RNA Later (Ambion, Inc., Austin, TX). Total RNA was extracted using RNA Stat-60 (Tel-Test, Inc.) from 110 7.5-8.0 dpc allantoides, 50 8 -8.4 dpc allantoides, 130 8.5-9.0 dpc allantoides, 30 9.5-10.0 allantoides, 60 7.8 -8.0 dpc yolk sacs, 15 8 -8.4 dpc yolk sacs, 30 8.5-9.0 dpc yolk sacs, 30 9.5-10.0 dpc yolk sacs, 60 7.8 -8.0 dpc embryos, 15 8 -8.4 dpc embryos, 30 8.5-9.0 dpc embryos, and 30 9.5-10.0 dpc embryos (all of embryos were free of extraembryonic membranes). Reverse transcription of total mRNA using random hexamer oligodeoxynucleotide primers and cDNA synthesis was performed using a SuperScript First-Strand synthesis System from Invitrogen. cDNAs were diluted 1:5 with deionized water, and 1-l aliquots were used as template in PCR with TrueFidelity DNA polymerase (0.08 units/l, Continental Laboratory Products, San Diego, CA), dNTPs (400 M each, Invitrogen), primers (0.4 M each), and 1ϫ TrueFidelity PCR buffer with MgCl 2 (0.8 mM) in total reaction volumes of 25 l. The computer program Oligo 6.1 (Molecular Biology Insights, Inc., Cascade, CO) was used for the design of the primers for mouse transcripts evaluated in this study. The accession numbers for each cDNA sequence and PCR primer sequences used herein are shown in Table I. Primer pairs and cDNA templates were tested over a range of amplification cycles to determine an optimal cycle number for exponential phase of production. Annealing temperatures and cycles of amplification for each primer pair are indicated in Table I. PCR was performed in 0.2 ml of thin wall polypropylene tubes in a MJ Research Dyad thermal cycler (Waltham, MA). PCR products were separated on 1% Agarose 1000 (Invitrogen) gels with 0.5ϫ Tris borate-EDTA buffer in the presence of 5 g/ml ethidium bromide.
Immunostaining of Cultured Allantois Explants-Explants were fixed in 2% paraformaldehyde by adding 0.8 ml of 3% paraformaldehyde, dPBS, 0.01% sodium azide (dPBSA) (Cellgro, Herndon, VA) to culture medium and incubating for 20 min at room temperature. The fixed explants were washed in dPBSA and then permeabilized in dPBSA, 0.02% Triton X-100 for 30 min at room temperature. Permeabilized explants were treated with blocking solution (3% bovine serum albumin, dPBSA) and then incubated for 1.5 h with antibodies to PECAM or Flk1, diluted to 20 g/ml in 3% bovine serum albumin and dPBSA. Fluorochrome-conjugated anti-IgGs (Jackson ImmunoResearch Laboratories, Inc.) were added at 10 g/ml and incubated for 1.5 h. In some cases, explants were also counterstained with either 0.5 ng/ml Hoescht 33342 stain from Molecular Probes (Eugene, OR) or 5 M Draq5 (Biostatus Limited, Shepshed, United Kingdom). The antibody-labeled explants were mounted under a number 1 coverslip using anti-bleaching mounting medium.
Confocal Microscopy and Image Processing-Immunolabeled allantois explants were examined using either a Bio-Rad MRC-1024 laserscanning confocal microscope or a Leica DMR light/epifluorescence microscope equipped with a SPOT-RT Slider camera (Vashaw Scientific, Norcross, GA). For laser-scanning confocal microscope, optical sections of the cultured explants were collected along the Z axis and collapsed into a single focal plane using the manufacturer's software to produce a single virtual image. Adobe Photoshop 7 (Adobe Systems, San Jose, CA) was used to compile montages of images of allantoic explants captured on the Leica DMR microscope and to measure the diameters of mesothelial layers and vascular networks.
Flow Cytometry-Allantoides from 8.5 dpc embryos were cultured for 18 h in Dulbecco's modified Eagle's medium or in this medium containing either 10% serum or 1 M S1P. Cultured allantois explants were dissociated in 0.5 ml of 1ϫ trypsin-EDTA at 37°C for 5 min. Trypsin was inactivated by the addition of 1 ml of serum-containing medium. Cells were pelleted by centrifugation at 80 -100 ϫ g for 10 min and then resuspended in 0.5 ml of serum-containing medium and incubated at 37°C for 40 min to allow cells to recover. The cells were pelleted by centrifugation as before and resuspended in 300 l of dPBS containing 1% bovine serum albumin, 0.01% sodium azide, and 10 ng/ml DNase (FACS buffer) and kept on ice for the remainder of the experiment. The cell suspension was filtered through a cell strainer (BD Biosciences), and cells were counted on a hemocytometer. Aliquots (100 l) of the cells at 1 ϫ 10 3 cells/l were incubated with anti-Flk1 IgG conjugated to phycoerythrin and anti-CD31/PECAM IgG conjugated to fluorescein isothiocyanate (each conjugate at 50 g IgG/ml) and incubated for 40 min on ice in the dark. Following incubation, 1 ml of FACS buffer was added to each tube and the cells were pelleted by centrifugation as above. The cells were resuspended in 1 ml of FACS buffer containing 0.1 g/ml propidium iodide. The cells were once again pelleted by centrifugation and resuspended in 250 l of FACS buffer. Cell fluorescence was quantified on a FACSCalibur (BD Biosciences) flow cytometer, and the data were analyzed using CellQuest software (BD Biosciences). Statistical Analysis-Microsoft Excel (Redmond, WA) was used to perform a one-way ANOVA to identify statistical differences in allantois explant measurements as well as in flow cytometry data. Kaleidagraph (version 3.6.2, Synergy Software, Reading PA) was used to determine the statistical significance of differences using Tukey's post hoc comparisons. The ␣ was set to 0.05 for all of the statistical analyses. All of the charts were plotted using Kaleidagraph, and the error bars represent confidence intervals at 95%.

RESULTS
S1P 1 , S1P 2 , and S1P 3 Expression Coincide with the Process of Vasculogenesis-RT-PCR was used to determine the temporal pattern of expression S1P receptors (S1P 1 -S1P 5 ) during mouse embryonic development. S1P receptor expression was evaluated in the embryo proper and two extraembryonic tissues (the allantois and the yolk sac) at stages during which these tissues were undergoing vasculogenesis. The level of VE-cadherin mRNA expression was used as a measure of blood vessel formation in each tissue. VE-cadherin expression was first detected at 8.5-9.0 dpc in the embryo proper, yolk sac, and allantois ( Fig. 1). In the prevascularized allantois (i.e. 7.5-8.0 dpc prior to the expression of VE-cadherin mRNA), S1P 1 , S1P 2 , and S1P 3 transcripts were expressed (Fig. 1). The expression of these three receptors persisted in the allantois during stages when nascent blood vessels first appeared (8.25 dpc) (22,23) and through periods in which angiogenesis and blood vessel maturation occurred (8.5-10 dpc). S1P 1 and S1P 3 mRNA expression showed a marked increase in expression during this later period. S1P 4 and S1P 5 transcripts were not detected in the allantois at any stage examined, suggesting that these receptors are not involved in vasculogenesis. In the yolk sac, prior to the expression of VE-cadherin (7.8 -8.4 dpc), S1P 1 , S1P 2 , S1P 3 , and S1P 5 transcripts were expressed (Fig. 1). In the prevascularized embryo proper (7.8 -8.4 dpc), the expression of S1P 2 , S1P 3 , S1P 4 , and S1P 5 was evident but little or no S1P 1 was detectable ( Fig. 1). Fig. 2 summarizes the temporal pattern of expression of each of the S1P receptor mRNAs in the tissues examined. Based on the collective findings, S1P 1 , S1P 2 , S1P 3 but not S1P 4 , and S1P 5 are expressed during periods of vasculogenesis (i.e. Ͻ8.4 dpc). Only S1P 2 appeared to be expressed during earliest stages of vasculogenesis, irrespective of the site. However, S1P 2 expression in prevascularized tissues was accompanied by the expression of either S1P 1 or S1P 3 or both. The expression of S1P 1 , S1P 2 , and S1P 3 receptors was also found to persist through later developmental stages during periods in which existing blood vessels become further reticulated and mature (8.5-10 dpc).
Sphingosine Kinase-2 Is Expressed during the Earliest Stages in Allantois Vasculogenesis-RT-PCR was also used to investigate the expression of sphingosine kinase-1 and -2 in the developing mouse embryo, yolk sac, and the allantois. In the prevascularized embryo proper, yolk sac, and allantois, sphingosine kinase-2 (SK2) was expressed from the earliest stage examined (7.5-8 dpc) and the expression persisted through to the latest stages examined (9.5-10 dpc) (Fig. 1). The level of the SK2 amplicon appeared to increase coincident with the expression of VE-cadherin in each of the three tissues ( Fig. 1). By contrast, the expression of SK1 transcripts was not detected in the allantois at any of the stages examined. In the yolk sac, SK1 expression was not detected prior to the expression of FIG. 1. RT-PCR analysis of the expression of S1P receptors and sphingosine kinase mRNAs in the embryo proper, yolk sac, and allantois at different stages of development. PCR was performed using cDNA templates prepared from 7.5-9.5 dpc mouse allantoides, yolk sacs, and embryos free of extraembryonic membranes (Embryo Proper). Primers were based on mouse sequences encoding the indicated S1P receptors, SK1, SK2, VE-cadherin (VE-cad), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) ␤-actin and collagen type I, ␣1 (Col1a1). Col1␣-1 primers were designed from sequences on either side of a short intronic sequence and thus yielded a larger amplicon (488 bp) in the event of contamination by genomic DNA. Each cDNA template was tested for genomic contamination via 40 cycles of PCR with Col1␣-1 primers. Conditions for S1P 4 primers were optimized using cDNA prepared from adult mouse lung RNA. VE-cadherin. In the embryo proper, SK1 was expressed at the earliest stages (7.8 -8 dpc) but its expression was absent at 8 -8.4 dpc and then reappeared at 8.5-10 dpc (Fig. 1). Since control reactions indicate that there was genomic DNA contamination of the 7.8 -8 dpc embryo proper cDNA template, the apparent expression of SK1 at this early stage is equivocal. Taken together, the findings indicate that SK2 but not SK1 expression coincides with the earliest stages of vasculogenesis, irrespective of the site. The absence of expression of SK1 in the developing allantois, a tissue in which vasculogenesis is a key event, suggests that SK1 is not required for vasculogenesis. Furthermore, the findings also suggest that S1P expressed in embryonic tissues via the action of SK2 may be influencing the early events in the process of vasculogenesis. S1P Promotes Vasculogenesis in Cultured 7.8 dpc Allantois Explants-To investigate the role of S1P signaling in vasculogenesis, we evaluated the effects of exogenously added S1P on the process of de novo blood vessel formation in the cultured allantois explant model (22,24). Because serum is a rich source of S1P (25), we first evaluated the effect of culturing the explants in serum-free conditions. As shown in Fig. 3, 7.8 dpc allantois explants cultured in the absence of serum failed to form an expansive network of blood vessels as detected using antibodies to PECAM, a protein expressed by endothelial cells (Fig. 3D). Instead of the branched network of capillary-like vessels typically observed when the explants are cultured in serum-containing medium (Fig. 3, A-C), only small poorly reticulated clusters of rounded PECAM-positive cells formed in the absence of serum (Fig. 3, D-F). Often clusters of PECAMpositive cells were observed with few cell protrusions evident (Fig. 3, E and F, arrows). By contrast, in 7.8 dpc allantois explants cultured in serum-free medium containing S1P, there was a marked difference in the pattern of PECAM-positive cells (Fig. 3, G-I) as compared with explants cultured in the absence of serum. Networks of interconnected endothelial cells (i.e. reactive with PECAM antibodies) were evident in explants cultured in the presence of S1P (Fig. 3, G-I). Morphologically, PECAM-positive cells formed in the presence of S1P appeared less well spread than the ones formed in the presence of serum (Fig. 3, C versus I). Because endothelial cells (PECAM-positive cells) were formed in the 7.8 dpc allantois cultured in the absence of serum, the formation of the endothelial cell lineage does not appear to be dependent on S1P signaling. S1P Promotes Expansion of Vascular Networks in Cultured 8.5 dpc Allantois Explants-We next evaluated the influence of exogenous S1P on blood vessel formation in 8.5 dpc allantois explants. At 8.5 dpc, the allantois was vascularized and undergoing extensive neovascularization (22). When explanted, 8.5 dpc allantoides were cultured in the presence of 10% fetal bovine serum and a highly branched network of PECAM-positive blood vessels was apparent after 24 h (Fig. 4A). However, in the absence of serum, PECAM-positive endothelial cells remained in the central portion of the explant and displayed extremely limited degree of reticulated network formation (Fig.  4B). By contrast, in serum-free medium containing S1P (Fig.  4C), vascular network formation was comparable with that observed in explants cultured in serum-containing medium. Not only did S1P promote network branching to a level equivalent to that observed in explants cultured the presence of serum but also the overall area of vascular networks formed in the presence of S1P was similar to the area of vascular networks that formed in serum-containing medium. S1P was also found to be effective at promoting allantois blood vessel formation at concentrations ranging from 33 nM to 1 M (Fig. 4, C-F) with only a modest qualitative difference in the vasculatures formed over the range of 111 nM-1 M. Even at the lowest S1P concentration tested (33 nM), vascular network formation was discernibly better than the network formation that was apparent in the absence of serum (Fig. 4F). In addition, we also found that sphingosine, the precursor of S1P, was capable of promoting blood vessel formation in 8.5 dpc explant cultures (Fig. 4, G  and H). It was observed that 5 M sphingosine in serum-free medium elicited blood vessel formation to a similar extent as 1 M S1P. By contrast, 1 M sphingosine was only marginally effective. The high concentration of sphingosine that is re-quired to elicit a similar response as S1P may reflect limitations in its availability to SK for phosphorylation.
We also tested the ability of another lysophospholipid, LPA, to promote blood vessel formation. As shown in Fig. 5, LPA at 333 nM (Fig. 5E) and 33 nM (Fig. 5F) was not able to improve either the vascular network branching or the extent of network expansion over that observed in explants cultured in the absence of serum. However, LPA at a concentration of 1 M elicited modest improvement in the qualitative aspects of the vascular networks (Fig. 5D), but the degree of vascular network formation was always less than that observed using 1 M S1P (Fig. 5C). Because LPA is known to bind to the S1P receptor, S1P 1 , with low affinity (K d ϭ 2.3 M) (26), we evaluated the effect of LPA at 6 M and observed a greater degree of blood vessel formation as compared with that achieved using 1 M LPA with clearly more blood vessels present than in explants cultured in the absence of serum (data not shown). Whether the observed micromolar dosage LPA effects are mediated by S1P 1 or some other LPA-binding receptor remains to be elucidated.
VEGF and bFGF Do Not Influence Allantois Vascular Network Expansion in a Manner Similar to S1P-Because VEGF is known to be important in the assembly of angioblasts into cordlike structures during vasculogenesis (22), we next evaluated the ability of VEGF to influence blood vessel assembly in 8.5 dpc allantois explants cultured in the absence of serum. We observed that blood vessel networks within explants cultured in serum-free medium containing VEGF (Fig. 6D) were not as expansive as those present in explants cultured in serum (Fig.  6A) or S1P-containing medium (Fig. 6E). Furthermore, the vascular networks formed in the serum-free medium containing VEGF had less avascular spaces and were much more densely packed with PECAM-positive cells (i.e. endothelial cells) (Fig. 6D) as compared with the networks formed in serum-free medium (Fig. 6C). When VEGF was combined with S1P, there was a marked expansion of the vascular network (Fig. 6F) as compared with explants cultured in serum-free medium containing VEGF (Fig. 6D). Furthermore, the vasculature that formed in the presence of both VEGF and S1P (Fig.  6F) displayed a reduction in the amount of avascular space as compared with networks formed in medium containing serum (Fig. 6A) or serum-free medium containing S1P (Fig. 6E). The findings indicate that S1P and VEGF have distinct effects on 8.5-dpc allantois vascular network formation. S1P appears to promote expansion of allantois vascular networks, whereas VEGF appears to increase the density of endothelial cells and promote vascular fusion. The ability of exogenously added VEGF to promote an increase in endothelial cell numbers and lead to the formation of fused sinusoidal vessels is well established (27)(28)(29)(30).
Recent findings suggest that S1P signaling might be involved with bFGF signaling (31,32). Therefore, we tested the effect of bFGF on the formation of blood vessels in allantois explants. We found that blood vessels formed in serum-free medium containing bFGF (Fig. 7D) were not discernibly different from those observed in 8.5 dpc explants cultured in serumfree medium alone (Fig. 7B). Overall, bFGF did not promote the vascular network expansion that was observed in S1P-treated explants (Fig. 7C).

S1P Does Not Influence Cell Survival in Cultured 8.5 dpc
Allantois Explants-The aforementioned findings indicated that S1P elicited qualitative effects on the blood vessels of cultured allantois explants. Specifically, it appeared that the S1P alone was capable of supporting the expansion of vascular networks in cultured 8.5 dpc explants to a similar extent as serum. Therefore, we assessed the extent to which S1P effects on vascular network expansion might be due to it influencing allantois cell growth. We found that 8.5 dpc allantois explants cultured in serum-free medium containing S1P did not have significantly different numbers of total cells as compared with explants cultured in serum-free medium (p ϭ 0.19) (Fig. 8A). In addition, S1P, which is known to be survival factor for certain types of cells (33), did not significantly alter the level of total cell death occurring in cultured explants as compared with the level observed in explants cultured in serum-containing medium or serum-free medium (p ϭ 0.84) (Fig. 8B).

S1P Does Not Influence Allantois Endothelial Cell Numbers or Differentiation of Angioblasts to Form Endothelial Cells-
Flow cytometry was also used to evaluate the effect of S1P on the growth of specific subpopulations of 8.5 dpc allantois cells (i.e. undifferentiated mesodermal cells, angioblasts, and endothelial cells). As shown in Fig. 8C, the explants cultured in the absence of serum or in serum-free medium containing S1P had ϳ2-fold lower percentage of endothelial cells (i.e. Flk-1 ϩ /PE-CAM ϩ population) as compared with explants cultured in the presence of serum (p Ͻ 0.0001). Importantly, there was no significant difference in the number of endothelial cells in explants cultured in the absence of serum versus explants cultured in serum-free medium containing S1P (p ϭ 0.71) (Fig.  8C). Thus, the ability of S1P to promote blood vessel formation was not due to it acting to increase the number of endothelial cells over the levels observed in explants cultured in serum-free medium. Similarly, there was not a significant effect of S1P on the numbers of angioblasts (i.e. Flk-1 ϩ /PECAM Ϫ cells) (ANOVA p ϭ 0.24) (Fig. 8D) or undifferentiated mesodermal cells (i.e. Flk-1 Ϫ /PECAM Ϫ cells) (ANOVA, p ϭ 0.29) as compared with allantois explants cultured in serum-free medium or in the presence of serum (Fig. 8, D and E, respectively). These findings indicate that S1P does not influence the differentiation of allantois mesodermal cells to form angioblasts or angioblasts to form endothelial cells.

Quantification of the Effects of S1P on the Expansion of Cultured Allantoic Vascular Networks and Mesothelial Layers-To
quantify the effects of S1P on vascular network expansion, we measured the diameters of PECAM-positive vascular networks in 7.5-7.8 dpc allantoides cultured in serum-containing medium, serum-free medium, and serum-free medium containing S1P. As shown in Fig. 9G, the average diameters of the vascular networks formed in 7.5-7.8 dpc allantoides explants cultured in S1P-containing medium were 2.1-fold greater than those that formed in serum-free medium (p ϭ 0.0001). Additionally, we measured the influence of S1P on the expansion of the mesothelial layer, the epithelial layer that forms the surrounding sheath of the allantois. After using the Hoechst 33342 stain to label the nuclei (Fig.  9, A-C), the average diameter of the mesothelial discs formed in serum-free medium containing S1P was found to be only 1.36fold greater than the diameter for those explants grown in serum-free medium (p ϭ 0.004) (Fig. 9H). Finally, we measured the influence of S1P on the percentage of the mesothelial disc area that was occupied by the vascular network (i.e. (vascular network area/mesothelial disc area) ϫ 100). As a result, it was found that the vascular networks formed in 7.5-7.8 dpc allantois explants cultured in the presence of S1P occupied 61% mesothelial disc area (Fig. 9I). By contrast, the vascular networks formed in the explants cultured in serum-free medium occupied only 26% mesothelial area (Fig. 9I). In addition, vascular networks formed in the presence of serum occupied 65% mesothelial area (Fig. 9I). Thus, S1P promoted a 2.3-fold greater degree (p Ͻ 0.0001) of vascular network expansion relative to mesothelial disc expansion as compared with what occurred in serum-free medium. The effect of S1P on the ratio of vascular network area/mesothelial disc area was similar to that observed in serum-containing medium.
We next quantified the effects of S1P on vascular network and mesothelium expansion in 8.5 dpc allantoides cultured in serum-containing medium, serum-free medium, or serum-free medium containing S1P. As shown in Fig. 9J, the average diameters of the vascular networks formed in 8.5 dpc allantois explants cultured in S1P-containing medium were ϳ2-fold greater than those that formed in serum-free medium (p ϭ Ͻ0.001). Furthermore, the average vascular network diameter of the S1P-treated explants was not significantly different from that of explants cultured in serum-containing medium (p ϭ 0.92). The average diameter of the mesothelial discs formed in serum-free medium containing S1P was found to be ϳ1.4-fold greater than for those explants grown in serum-free medium (p ϭ Ͻ0.0001) and only slightly less (p ϭ 0.044) than for those explants cultured in serum-containing medium (Fig. 9K). As for the influence of S1P on the percentage of the mesothelial disc area that was occupied by the vascular network, it was found that vascular networks formed in the presence of S1P occupied 72.5% mesothelial disc area (Fig. 9L). By contrast, vascular networks formed in the explants cultured in serum-free medium occupied 40% mesothelial area (Fig. 9L). In addition, the vascular networks formed in the presence of serum occupied 59% mesothelial area (Fig. 9L). Thus, S1P increased the ratio of vascular network area/mesothelial disc area by nearly 50% over that observed in serum-free medium (p ϭ 0.0003) and 20% greater than serum-containing medium (p ϭ 0.016).
Upon visual inspection of Hoescht 33342-stained explants, it appeared that the number of nuclei in explants cultured in the absence of serum was less than the number of nuclei cultured in the presence of S1P (Fig. 9, B and C). Because the total numbers of cells in explants cultured in the presence or absence FIG. 9. S1P promotes expansion of allantoic vascular networks to a greater extent than expansion of the mesothelium. A-C, epifluorescent microscopy images of 8.5 dpc allantois explants cultured for 24 h in serum-containing medium (A), serum-free medium (B), or serum-free medium containing 1 M S1P (C) and stained with Hoechst 33342 stain (red nuclei) and labeled with PECAM antibodies and fluorescein isothiocyanate-conjugated secondary IgG. Panels D-F are laser-scanning confocal microscope images of central regions of 8.5 dpc allantois explants cultured in serum-containing medium (D), serum-free medium (E), and serum-free medium containing S1P (F) stained with Draq5 to reveal nuclei. For D-F, optical sectioning of entire allantois explants was performed and the complete Z series collapsed. To derive quantitative data pertaining to the effects of culture conditions on vascular network and mesothelium expansion, measurements (double arrowheaded lines) were made as indicated in panel B on the diameter of the Hoechst 33342-stained mesothelial disc and the diameter of the PECAM-stained vascular network. For each explant evaluated, a pair of diameter measurements were taken for the vascular network and averaged and the same was done for the mesothelium. G, H, and I represent data from 7.5-7.8 dpc allantois explant cultures. J, K, and L represent data from 8.5 dpc allantois explant cultures. For the values plotted in G-L, the experimental number (n) for each was 8 -12. of S1P were not significantly different (Fig. 8A), we sought an explanation for the apparent disparity. We found that the average height of 8.5 dpc allantois explants cultured in the absence of serum (80.2 Ϯ 8.35 m, n ϭ 20) was significantly greater (1.76-fold greater, p Ͻ 0.006) than those cultured in the presence of S1P (45.5 Ϯ 3.6 m, n ϭ 11). When explants were optically sectioned and when the entire Z series collapsed, it was evident that there were many more nuclei present in the central regions of explants cultured in the absence of serum as compared with those cultured in S1P or serum (Fig. 9, D-F).
Taken together, these findings (summarized in Table II) indicated that S1P promotes the expansion of the vascular network and mesothelial layers of cultured 7.5-7.8 and 8.5 dpc allantois explants. Furthermore, the fact that S1P promotes a greater level of vascular network expansion within the mesothelial sheath as compared with the expansion that occurs under control conditions suggests that S1P can enhance vascular cell motility to a greater extent than mesothelial cell motility.

DISCUSSION
The findings presented herein indicated that S1P signaling contributes to the process of murine vasculogenesis. First, it was shown that the S1P receptor S1P 2 was expressed in conjunction with S1P 1 and/or S1P 3 at stages preceding the formation of blood vessels (i.e. 7.5-8 dpc) in the embryo proper, yolk sac, and allantois. Furthermore, SK2, an enzyme that catalyzes the formation of S1P, was also expressed in the allantois, yolk sac, and embryo proper prior to the vascularization. Finally, we showed that exogenous S1P or sphingosine, but not VEGF or bFGF, could effectively replace the requirement of serum for promoting vasculogenesis in cultured allantois explants.
The process of de novo blood vessel assembly involves several key steps (22) that are diagrammatically depicted in Fig. 10. Major steps in the process include the birth of angioblasts from mesodermal progenitors and their differentiation into endothelial cells. Considering the fact that endothelial cells indeed formed in 7.5 dpc allantois explants cultured in serum-free medium, it is evident that the formation of both angioblasts and endothelial cells is not dependent on exogenously added S1P. Because SK2 is expressed in the 7.5-8.0 dpc allantois, we cannot rule out that, in the absence of exogenous S1P, endogenously expressed S1P contributes to the differentiation events required to generate endothelial cells. However, the fact that S1P treatment of cultured allantois explants did not alter angioblast or endothelial cell numbers supports the possibility that the lineage process is not driven by S1P signaling. By contrast, endothelial cell numbers increase in allantois explants treated with VEGF (data not shown), consistent with the well characterized mitogenic effects of VEGF on endothelial cells (27). Our observation that exogenous S1P did not lead to an increase in endothelial cell number is in contrast to the report that S1P can stimulate human umbilical vein and bovine aortic endothelial cell proliferation (15) and may highlight a difference in S1P responsiveness between early endothelial cells and cultured endothelial cells derived from mature blood vessels.
Based on our morphometric analyses, we concluded that exogenously added S1P mediates the expansion of nascent blood vessel networks in cultured allantois explants. In 7.5-7.8 dpc allantois explants cultured in the absence of S1P, it appears that there is a failure of either angioblasts or early endothelial cells to migrate. As a result, the vascular networks that form remain confined to a limited area with constituent cells displaying an abnormally high packing density. A role for S1P in promoting motility of angioblasts and early endothelial cells is consistent with its reported effects on the motility of cultured endothelial cells (16) and other types of cells (34). The importance of angioblast/early endothelial cell migration in the context of vasculogenesis has not been established. As a result of gastrulation, angioblasts first appear randomly distributed throughout the mesoderm (35). One model for the formation of vascular networks is that angioblasts merely extend processes to connect with other angioblasts and thus form polygonal arrangements. Another model, which is not necessarily mutually exclusive of the former model, is that angioblasts movement is required for the formation of a primary network. Our findings support the later model and indicate a role for S1P signaling in mediating angioblast/endothelial cell motility and extension. We believe that the observed expansion of vascular networks in response to S1P stimuli was the result of the outward migration of pioneering angioblasts/early endothelial cells that subsequently organize into a nascent vascular network. The fact that we observed isolated clusters of endothelial cells at the outer margins of the S1P-treated explants is evidence that the S1P-stimulated vascular network expansion was not due to angiogenic sprouting (Fig. 3G). A remarkable finding from our studies was the fact that VEGF was not able to mimic the effects of S1P on angioblast/ early endothelial cell motility and produce an expansive network. Indeed when explants were treated with VEGF alone, endothelial cells formed into a fused mass of small diameter. These findings demonstrated that S1P and VEGF have distinct effects on angioblast/early endothelial cell behaviors. Apparently, in the absence of S1P signaling, VEGF signaling was not sufficient to drive the migration of angioblasts/early endothelial cells. As a consequence of the failure of the cells to move and the increase in cell number due to the mitogenic effects of VEGF, the density-dependent phenomenon of vascular fusion occurred (36). Supporting the distinct roles of S1P and VEGF was the fact that when explants were cultured in the presence of both S1P and VEGF, the expected combinatorial phenotype was observed (i.e. expansion of the area of the vascular network and fusion of endothelial cells) (Fig. 6F).
Whereas our findings point to a role for S1P signaling in the process of vasculogenesis, the question is which of the S1P receptors acts as mediator(s)? Based on RT-PCR analysis, we have found that S1P 1 , S1P 2 , and S1P 3 , but not S1P 4 and S1P 5 , are expressed at sites of mouse embryo vasculogenesis at stages early enough for them to participate in vasculogenesis. However, evidence from targeted inactivation of S1P 1 , S1P 2 , and S1P 3 genes does not support a critical role for these recep-tors individually in the process of vasculogenesis (20,(37)(38)(39). Recently, double knock-out mice lacking S1P 1 and S1P 2 and triple knock-out (TKO) mice lacking S1P 1 , S1P 2 and S1P 3 have been generated (40). Both double knock-out and TKO embryos showed a 30% lethality with associated bleeding at 10.5 dpc. The TKO condition was more severe than that of the double knock-out as indicated by the fact that 100% lethality was reached at an earlier stage by TKO embryos (i.e. double knockout, 100% lethality at 14.5 dpc; TKO, 100% lethality at 11.5 dpc) (40). The vascular defects observed in these knock-out mice have been attributed to faulty angiogenesis (40). However, considering that vasculogenesis precedes angiogenesis, it is possible that defective vasculogenesis may be the underlying basis for observed vascular abnormalities in these knockouts. Additional support for the involvement of S1P signaling in vasculogenesis comes from targeted deletion of G␣ 13 , a G protein that couples to S1P 2 , S1P 3 , and S1P 4 (14). G␣ 13 deficiency results in embryonic lethality with embryos having defective vascular development (e.g. absence of yolk sac blood vessels) (21). Importantly, PECAM-positive endothelial cells formed in the yolk sac of G␣ 13 -null embryos but failed to assemble into vascular networks, suggestive of a defect in vasculogenesis. Cranial blood vessels that form via the process of vasculogenesis appeared enlarged and disorganized. Furthermore, the finding that fibroblasts from G␣ 13 -deficient embryos exhibit impaired motility (21) makes it plausible that defective angioblast and endothelial cell motility may contribute to the observed vascular abnormalities.
Recently, five orphan G protein-coupled receptors (i.e. Gpr3, Gpr6, Gpr12, Gpr61, and Gpr63) have been identified as S1P receptors (41)(42)(43). Rather little is known regarding these receptors. Pertussis toxin inhibits Gpr6, indicating that it couples to inhibitory G proteins (G i ) (41). Gpr6 also activates sphingosine kinase (41). Gpr3, Gpr6, Gpr12, and Gpr63 have been shown to be expressed by endothelial cells (i.e. human umbilical vein endothelial cells, human coronary artery endothelial cells, human pulmonary microvascular endothelial cells, and human pulmonary artery endothelial cells) as well as by smooth muscle cells (42,44). It remains to be determined whether any of these receptors might also be additional candidates for mediators of S1P signaling required for angioblast motility.