Sox17 is required for vascular growth and arterial differentiation

We then investigated whether Sox17 is required for correct vascular development and for arterial endothelial cell specification. To this end, we first crossed Sox17^flox/flox mice with Tie2-Cre mice to produce Sox17^ECKO mice. Inactivation of Sox17 induced in utero lethality in 100% of the embryos, within the E10.5 and E12.5 developmental stages.

As shown in Supplementary Fig. S1E,F, these Sox17^ECKO mice embryos showed major alterations in vascular remodelling and the development of large arteries was largely absent in all of the organs examined. The intersomitic vessels showed defective patterning, and the vasculature in the yolk sac failed to undergo remodelling, thus missing the correct arterial/venous differentiation (see Supplementary Fig. S1E for the yolk sac, and Supplementary Fig. S1F for the head vasculature and the intersomitic vessels). These data are consistent with previously shown vascular alterations in Sox17 null embryos22.

To investigate the role of Sox17 in vascular development at the postnatal stages, we crossed the Sox17^flox/flox mice with Cdh5(PAC)-Cre-ER^T2 mice30, to produce Sox17^iECKO mice and we induced recombination by tamoxifen injection at the P1 postnatal stage.

At P5, the endothelial cells of the Sox17-deficient vessels showed a small, but non-significant, increase in the number of tip cells in the vascular growing area at the periphery of the retina (Fig. 2a,b; quantification in Fig. 2c,d). This was further enhanced at P9 (Fig. 2b), and up to P12 (quantification in Fig. 2d). In addition, the growing front of the retina of these Sox17^iECKO mice showed a higher vascular density due to the active, multidirectional hypersprouting of the vasculature (Fig. 2a–d). The advancing of the vasculature across the vitreal surface was also slightly, and significantly, reduced at P9 (Fig. 2b; quantification in Fig. 2c,d). Through a closer analysis of the sprouting vessels we observed that the endothelial cells at the stalk of Sox17^iECKO mice (Fig. 2e, right panels and Supplementary Fig. S2C) maintained a highly dynamic formation of filopodia like if they did not receive the inhibitory Notch signal by the tip cells. The remarkable multidirectional hypersprouting of the vessels at the periphery of the retina vasculature in these Sox17^iECKO mice is shown at higher magnification in Fig. 2f, right panels and in Supplementary Movie 1 (see also Supplementary Fig. S2A for lower magnification). As sprouting tip cells are in multiple layers the overall staining of Dll4 appears more intense (see Fig. 2g).

Open in a separate window

Figure 2

Deletion of Sox17 alters vascular development.

(a,b) Isolectin B4 (IB4) staining of P5, and P9 retinas from control and Sox17^iECKO newborn mice shows marked defects in vascular development. Higher magnification of the angiogenic front (boxed areas) are shown and tip cells are indicated by yellow dots. (c,d) Quantification of vascular parameters in retinas from control and Sox17^iECKO mice. Binarized images of retinas (P9) from control and Sox17^iECKO (c, left and right, respectively), as an examples of the quantification performed in (d). The vascular front density was measured in the areas highlighted in red, while vascular progression was measured as the average distance covered by the growing vessels (see green arrows in c). The quantification of the tip cells per length, the vascular front density, the vascular progression and the number of filopodia per tip cell are shown in (d). Retinas from control and Sox17^iECKO (blue and red columns, respectively) at P5, P9 and P12 are shown. Data are means ±s.d. of at least five mice per group. (e) Detailed analysis of vascular sprouts stained with Isolectin B4 in control and Sox17^iECKO retinas. Tip cells nuclei are highlighted by yellow circles (see also Supplementary Fig. S2C and Methods for details) and filopodia protrusions by yellow dots (middle panels). Higher magnification of red boxed areas are shown on the right panels where tip cells nuclei and corresponding filopodia are colours coded. (f) High magnification of the angiogenic front show multidirectional hyper sprouting in the Sox17^iECKO vasculature. Tip cells growing above the vascular plane are indicated by arrows; filopodia are also indicated as yellow dots. Bottom panels show depth-coded stacks (see Methods for details). Tip cells indicated by arrows are located above vascular plexus (green). (g) Higher magnification of Isolectin B4 (IB4, white) and Dll4 (green) staining at the vascular front shows high expression of Dll4 in Sox17^iECKO hyper sprouting retinas (lower panels, yellow arrowheads). Control tip cells are shown in the upper panel. Scale bar, 500μm (top panels a–c); Scale bar, 100μm (bottom panels a,b,e); Scale bar, 25μm (magnifications of e–g). *P<0.05; **P<0.01, two-tailed t-test assuming unequal variances.

The arterial/venous differentiation appeared altered in the absence of Sox17. In these Sox17^iECKO pups, in the absence of Sox17 the arteries acquired the expression of endomucin, a selective marker of large veins (Fig. 3a,b compare upper with lower panels; quantification Fig. 3f) and upregulated Eph4 (Fig. 3c). Conversely, Jag1 and Dll4 were significantly reduced (Fig. 3d,e). RT–PCR analysis of freshly isolated endothelial cells from the mutant mice (Fig. 6a and Fig. 3g, red columns) also showed that other arterial markers such as Dll4, Notch4 and CXCR4 were reduced and venous markers as COUP-TFII was increased.

Open in a separate window

Figure 3

Altered arterial and venous differentiation in the absence of Sox17.

Analysis of the vasculature of retina, diaphragm and small intestine at P9. (a) Whole-mount immunofluorescence for Isolectin B4 (IB4, white) and endomucin (green) of control and Sox17^iECKO retinas. Endomucin is expressed in control veins (V). Arrowheads indicate endomucin-positive arteries (A) in Sox17^iECKO mouse. (b) Staining for VE-cadherin (VEC, white) and endomucin (green) of vessels of control and Sox17^iECKO mice diaphragm. Arrowheads indicate an artery endomucin positive in Sox17^iECKO mouse. Quantification of data in (a,b) is shown in panel f (see below). (c) Immunostaining for Isolectin B4 (IB4, white) and EphB4 (green) of control and Sox17^iECKO retinas. The arrowheads indicate the EphB4 staining in the arteries of Sox17^iECKO mouse. (d) Isolectin B4 (IB4, white) and Jag1 (green) staining of the retina vasculature in control and Sox17^iECKO mice. The arrowheads point to decreased Jag1 staining in Sox17^iECKO mouse. (e) Immunofluorescence for VE-cadherin (VEC, white) and Dll4 (green) staining in the vasculature of the intestine in control and Sox17^iECKO mice. Arterial Dll4 staining is strongly reduced (see arrowheads) in Sox17^iECKO mice. (f) Quantification of endomucin-positive arteries in control (blue) and Sox17^iECKO (red) retinas. Data are expressed as percentages of total vessels ±s.d. (n>20). (g) qRT–PCR analysis of control and Sox17^iECKO lungs derived endothelial cells (red columns) and cultured venous endothelial cells infected with Sox17 (blue columns). Modifications induced in the expression of arterial (CXCR4, ephrinB2) and venous markers (COUP-TFII) are reported. The RNA level obtained for control was set to 1 and the ratio for Sox17^iECKO versus control or venous endothelial cells infected with Sox17 versus control is shown for each gene. Standard deviation of control values did not exceed 10% of the means. Scale bar, 100μm; **P<0.01; *P<0.05, two-tailed t-test assuming unequal variances. (h) Confocal analysis of cryosections stained with PECAM (green) and DAPI (blue) of control and Sox17^ECKO embryos (E10.5) revealed the presence of arteriovenous shunts (arrowheads) leading to a bypass circulation. Scale bar, 200μm.

Furthermore, detailed analysis of cryosections of Sox17^ECKO embryos at E10.5 showed fusion between the cardinal vein and the dorsal aorta (DA) (Fig. 3h) leading to the formation of a bypass circulation from the DA to the sinus venosus region of the heart. The presence of this type of shunts is typical of conditions were arterial–venous identity is lost14.

The arterial coverage with αSMA, NG2 and Desmin-positive cells was significantly reduced and/or altered in these arteries (Fig. 4a,d). The lumen of the veins was frequently enlarged (Fig. 4a, lower panel) and vascular lacunae were seen at the venous front (Supplementary Fig. S2B, compare upper with lower panels).

Open in a separate window

Figure 4

Impairment of mural cells coverage in the absence of Sox17.

(a–d) Analysis of the control and Sox17^iECKO retinal vasculature by whole-mount staining for Isolectin B4, αSMA, NG2 and Desmin. (a,b) Isolectin B4 (IB4, white) and αSMA (green) staining. Arrowheads indicate αSMA-negative arteries in Sox17^iECKO mice. Quantification of αSMA fluorescence intensity in arteries of control and Sox17^iECKO (blue and red, respectively) is shown in (b). Data are expressed as mean fluorescence ±s.d. (n>3 mice per group) *P<0.01, two-tailed t-test assuming unequal variances. (c) Isolectin B4 (IB4, white) and NG2 (green) staining. (d) Isolectin B4 (IB4, white) and Desmin (green) staining. Higher magnification of the insets is shown in the right panels. Arrowheads indicate regions of incomplete pericyte coverage in Sox17^iECKO mice. These cells appear poorly organized around the vessel wall leaving relatively large, uncovered areas. Scale bar, 25μm in magnifications and 100μm elsewhere.

Sox17 is upstream of Notch signalling

The observed phenotype of these Sox17^ECKO embryos and of the Sox17^iECKO retina vasculature was very close to that of the embryos and retina vasculature of Notch-signalling-deficient mice15,16,17. It was therefore important to understand the relationships between Notch and Sox17 signalling in arterial specification.

We first tested the possibility that Notch is upstream of Sox17 expression. Treatment of newborn mice with the Notch signalling inhibitor N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT) did not change the Sox17 expression of freshly isolated lung endothelial cells (Fig. 5a), while, as expected, DAPT inhibited the expression of the Notch-dependent transcription factors Hes, Hey1 and Hey2. DAPT treatment of the pups did not modify Sox17 expression in the vessels of the retina (Fig. 5b) confirming the RT–PCR data obtained from freshly isolated endothelial cells from DAPT-treated pups (Fig. 5a). Sox17 expression was also not changed in Rbpj^iECKO pups retina (Fig. 5c) and mRNA extracts from lungs (Fig. 5d). In both model systems, the Notch target genes where, instead, downregulated as expected (Fig. 5a,e). Consistently, also in Dll4^iECKO (Fig. 5f) Sox17 staining remained unchanged.

Open in a separate window

Figure 5

Sox17 is upstream of Notch signalling.

(a) qRT–PCR analysis of freshly isolated endothelial cells from lungs of vehicle or DAPT-treated pups (P6). DAPT does not affect the expression of Sox17, while inhibits the expression of Hes, Hey1 and Hey2. Data are expressed as fold change versus vehicle and are means ±s.d. from cells obtained from at least four mice per group. The s.d. of the values obtained in vehicle-treated mice did not exceed 10% of the mean. (b) Retina whole-mount immunostaining for Isolectin B4 (IB4, green) and Sox17 (red) after vehicle (left) or DAPT (right) treated pups (P6). Inhibition of Notch signalling does not significantly change Sox17 expression (arrowheads). (c) Whole-mount immunofluorescence for Isolectin B4 (IB4, green), and Sox17 (red) of P6 control and Rbpj^iECKOretinas. The genetic ablation of Rbpj does not change the expression of Sox17 (arrowheads). (d,e) qRT–PCR analysis of Rbpj^iECKO P6 mouse lungs. The genetic ablation of Rbpj does not change the expression of Sox17 (d), while it inhibits the expression of Hes, Hey1, Dll4, Jag1, Notch1 and Notch4. Data are expressed as fold change versus controls and are means ±s.d. from at least three mice per group. The s.d. of the values obtained in control animals did not exceed 10% of the mean. (f) Whole-mount immunofluorescence for Isolectin B4 (IB4, green), and Sox17 (red) of P6 control and Dll4^iECKOretinas. The genetic ablation of Dll4 does not change the expression of Sox17 (right panels, arrowheads). (g) Isolectin B4 (IB4, green) and Sox17 (red) staining of P6 retinas from control (ROSA^{Notch IC}-CRE) and the endothelial-cell-specific gain-of-function mutant of Notch signaling (ROSA^{Notch IC}+CRE). Activation of Notch signaling does not significantly change Sox17 expression (arrowheads). The upregulation and ectopic expression of αSMA (white) and Dll4 (red) in veins is shown in (h), see arrowheads. Columns are means ±s.d. of triplicates from a representative experiment. *P<0.05; **P<0.01, two-tailed t-test assuming unequal variances. Scale bar, 100μm.

On the other hand, an endothelial-cell-specific gain-of-function mutation of Notch signalling (ROSA^{Notch IC}; see Methods) did not significantly change Sox17 expression in the retina (Fig. 5g, lower panel). The phenotype of these mice has been previously published31 and we also report here, as a typical marker of the effect of active Notch signalling, the upregulation and ectopic expression of alpha smooth muscle actin and Dll4 in veins (Fig. 5h).

Taken together, these data do not support the hypothesis that Sox17 acts downstream of Notch signalling.

We then tested whether Sox17 can increase Notch signalling. We first isolated and cultured venous endothelial cells defined as CD45⁻/PECAM-1⁺/CXCR4⁻cells (see Methods and (ref. 32)), which did not show Sox17 expression. These cells presented low expression of arterial markers as compared with arterial endothelial cells (defined as CD45⁻/PECAM-1⁺/CXCR4⁺ cells) (Supplementary Fig. S3B).

We then infected venous endothelial cells with lentiviral vectors coding for Sox17 or with the transcriptionally inactive Sox18^RaOP (RaOP) mutant that can compete with SoxF for DNA binding23. As shown in Fig. 6b, there was a strong increase in NICD translocation to the nucleus only in the Sox17-infected cells (quantification in Fig. 6c). The transcriptional SoxF competitor RaOP alone was inactive, although it inhibited the effects of Sox17 on Notch signalling, supporting the concept of a specific transcriptional effect of Sox17.

Open in a separate window

Figure 6

Sox17 affects Notch signaling.

(a) qRT–PCR analysis of control and Sox17^iECKO lungs derived endothelial cells (P6) shows that several members of the Notch signaling pathway (Hes, Dll4 and Notch4) are downregulated by the absence of Sox17. Data are expressed as fold change versus controls and are means ±s.d. from at least three mice per group. The s.d. of the values obtained in control animals did not exceed 10% of the mean. (b) Vein endothelial cells which do not express Sox17 (see Supplementary Fig. S3A,B) were infected with lentiviral vectors coding for GFP, Sox17, Sox18^RaOP or both Sox17 and Sox18^RaOP in combination. Immunofluorescence analysis of confluent monolayers stained for Sox17 (green) and NICD (red). Nuclear NICD staining is significantly increased by Sox17 expression (white asterisks), while the Sox18^RaOP mutant is inactive. The Sox18^RaOP mutant reduced the effects of Sox17 on NICD nuclear staining when it is expressed in combination. Scale bar, 100μm. Quantification of nuclear fluorescence intensity is reported in (c). Data are means ±s.d. of three experiments. (d) qRT–PCR analysis of Sox17 infected venous endothelial cells (see above). Sox17 expression (red columns) upregulates several members of the Notch signaling pathway, as Hey1, Dll4, Dll1 and Notch4, and arterial markers, such as ephrinB2. Conversely, venous markers, EphB4 and COUP-TFII are downregulated. DAPT treatment (magenta columns) prevent the Sox17-induced upregulation of Hey1 and Dll4. Data are means ±s.d. of four experiments. (e–g) ChIP analysis of Sox17 interaction with the promoters of Notch4, Notch 1, Dll1 and Dll4. The positions of the putative Sox17-binding sites for all the genes analysed are indicated. Chromatin from cultured endothelial cells was immunoprecipitated with Sox17 antibody and qRT–PCR on the indicated regions was carried out. The levels of DNA are normalized to input. (e) ChIP analysis of venous cells infected with GFP (blue) or Sox17 (red); (f) ChIP analysis of arterial cells from lungs expressing endogenous levels of Sox17 (violet); (g) ChIP analysis of arterial cells of embryonic origins expressing endogenous levels of Sox17 (green). Columns are means ±s.d. of triplicates from a representative experiment. *P<0.05; **P<0.01, two-tailed t-test assuming unequal variances.

Sox17 infected venous endothelial cells upregulated members of the Notch signalling pathway, including Hey1, Dll4, Dll1 and Notch4, and arterial markers, such as CXCR4 and ephrinB2 (Fig. 6d, red columns and Fig. 3c blue columns). Conversely, venous markers, EphB4 and COUP-TFII were downregulated (Fig. 6d, red colums and Fig. 3g, blue columns). To investigate which of these markers were upregulated by Sox17 directly and not indirectly through Notch activation we incubated Sox17 expressing cells with DAPT. As reported in Fig. 6d (magenta columns), with the exception of Hes, Hey 1 and Dll4, which were downregulated by DAPT, Notch4 and Dll1 were upregulated by Sox17 also in the presence of Notch inhibition.

To investigate whether Sox17 could interact directly with the promoters of Notch4, Notch1, Dll1 and Dll4 we carried out chromatin immunoprecipitation assay (ChIP). These experiments were performed in venous endothelial cell infected with Sox17 (Fig. 6e), in arterial cells from adult lungs and arterial cells from embryo vasculature (Fig. 6 panel f and g, respectively). In all the three cell lines Sox17 could bind Notch4 promoter at high affinity (Fig. 6e–g, left panels). Five consensus sequences were identified in the Dll4 promoter region. Sox17 associates to these regions in the venous cells reconstituted with Sox17 (Fig. 6e, right panel). The binding to Dll4 promoter in the other two cell lines is weaker and occurs mainly in region B (Fig. 6f,g, right panels). Sox17 binding to Notch1 promoter, even if weak, occurs in venous endothelial cell infected with Sox17 and in arterial cells from adult lungs. Instead, Sox17 binding to Dll1 promoter is negligible (Fig. 6e–g, middle panels).

These data support the idea that Sox17 acts upstream of Notch signalling in its induction of arterial specification of endothelial cells. The action of this transcription factor appears to be quite complex directly interacting with the promoters of different members of the Notch pathway.

As VEGF is known to act upstream of the Notch pathway, we then tested whether Sox17 is downstream or upstream of VEGF signalling. As shown in Supplementary Fig. S3C, upregulation of Sox17 did not significantly change the expression of VEGF-A or its receptors, VEGFR-2 and VEGFR-3. On the other hand, VEGF activation of these cells did not increase Sox17 (Supplementary Fig. S3A).

Antibodies

The following antibodies were used: anti-VE-cadherin (5μgml⁻¹, BV13, eBioscience), anti-PECAM-1 (1:200, BD), anti-Sox17 (1:200, R&D Systems), anti-αSMA (1:400, FITC-conjugated, Sigma), anti-β-galactosidase (1:500, Abcam), anti-endomucin (1:200, Santa Cruz), anti-Dll4 (1:200, R&D Systems), anti-NICD (1:100, Abcam), anti-EphB4 (1:200, R&D Systems), anti-Jag1 (1:100, Santa Cruz), anti NG2 (1:200, Abcam), anti-desmin (1:200, Abcam) and isolectin-B4 (1:100, Vector Labs). The anti-Sox17 antibody specificity was confirmed by immunostaining of the mice retinas (see Supplementary Fig. S1G). Alexa Fluor 488, 555 and 647 donkey secondary antibodies were from Invitrogen.

Endothelial cell culture and qRT–PCR

For the isolation of endothelial cells from embryos, E9.5 to E10.5 embryos were dissected and dissociated with collagenase type I (Roche, 1.5mgml⁻¹), DNase (Roche, 25μgml⁻¹) at 37°C for 1h and passed through a 40μm cell stainer18. Cells in this flow-through were collected by centrifugation and washed in 0.1% BSA in PBS. The venous endothelial cells were separated with the FACSAria cell-sorting system (BD Biosciences), with the combination of APC-conjugated anti-PECAM-1 (BD), PE-Cy7-conjugated anti-CD45 (eBioscience) and PE-conjugated anti-CXCR4 (eBioscience)32 antibodies (venous endothelial cells were defined as CD45⁻/PECAM-1⁺/CXCR4⁻). The endothelial cells were immortalized52 and lentivirally transduced with Sox17 and Sox18^RaOP cDNAs, subcloned into the lentiviral plasmid pRRLsin.PPTs.hCMV.GFPwpre under the transcription of CMV promoter23.

DMSO or an equal volume of DAPT (10μM) dissolved in DMSO was added to the culture cells and incubated for 24h. Cells were then lysate for RNA extraction.

For VEGF stimulation, the endothelial cells were grown to confluence, serum starved (MCDB, 1% BSA) overnight, and then incubated with 80ngml⁻¹ VEGF (Peprotech) for 8h. These cells were then lysed for RNA extraction.

Freshly isolated endothelial cells from lungs of Sox17^iECKO (P6) and DAPT-treated mouse pups (see above) were isolated using Dynabeads (Invitrogen) and processed for RNA extraction, as described below18. Total RNA from lungs of Rbpj^iECKO pups (P6) was isolated as described below40.

Human endothelial cells isolated from umbilical vein (HUVECs) by treatment with Collagenase (Roche, 0.1%, for 30min at 37°C) were routinely cultured53; HAEC were purchased from Life Technologies. Mesenteric arteries and veins were isolated from mice and immediately processed for RNA. All of the types of endothelial cells were grown, in MCDB-131 (GIBCO) with 20% serum.

For gene expression analysis, the total RNA was isolated with RNeasy Mini kits (QIAGEN) and 1μg total RNA was reverse transcribed with random hexamers (High-Capacity cDNA Archive kits; Applied Biosystems) following the manufacturer instructions. Micro fluidic cards (qPCR cards, Applied Biosystems) were created for array profiling of the transcripts of the genes described in Supplementary Table S1. The cDNAs were amplified using the TaqMan Gene Expression Assay (Applied Biosystems) in an ABI/Prism 7900 HT thermocycler. The relative expression differences represent the means of the results obtained from at least three-independent experiments.

ChIP assay

ChIP assay was carried out as described18,54; briefly, chromatin extracts containing 1mg of DNA fragments with an average size of 500bp were immunoprecipitated with 10μg Sox17 antibody overnight at 4°C in the presence of protein G magnetic beads (Invitrogen). Bound DNA fragments were eluted and amplified by standard PCR and qRT–PCR with the use of oligonucleotides flanking the assayed promoter regions. Primers used are listed in Supplementary Table S2.

For qRT–PCR analyses, DNA was diluted with specific primers (10μM each) to a final volume of 25μl in SYBR green Reaction Mix (Perkin Elmer).

Immunofluorescence

For cell immunofluorescence, confluent endothelial cells were fixed with 4% paraformaldehyde (PFA) in PBS, followed by permeabilization with 0.5% Triton X-100 for 5min. The blocking (1h), primary (2h) and secondary (1h) antibodies were in PBS with 2% BSA. The nuclei were visualized using DAPI.

Retina immunostaining was carried out with littermates processed simultaneously under the same conditions. To analyse and quantify the retina vascular phenotype, the eyes (P5, P9 and P12) were fixed in 2% PFA for 2h at 4°C and dissected. After blocking/permeabilization in 1% BSA, 5% donkey serum and 0.5% Triton-X100 in PBS overnight, the retinas were washed in Pblec Buffer (1% Triton X-100, 1mM CaCl₂, 1mM MgCl₂ and 0.1mM MnCl₂ in PBS, pH 6.8) and incubated overnight at 4°C in Pblec buffer containing biotinylated isolectin B4 (IB4) and the primary antibodies. Then the retinas were incubated with fluorophore-conjugated antibodies and with Alexa Fluor 488/555/647 streptavidin (Invitrogen), and mounted with ProLong Gold (Invitrogen).

For whole-mount immunostaining18, embryos were dissected in PBS and fixed in 4% PFA overnight at 4°C. After blocking/permeabilization for 3h in PBS containing 0.3%TX-100 and 5% donkey serum, embryos were incubated overnight at 4°C with the primary antibodies followed by the species-specific Alexa Fluor-coupled secondary antibodies.

Cryostat sections (8–12μm thick) were cut using a cryotome (Leica), mounted on Superfrost glass slides and dried. Sections were permeabilized for 5min in PBS containing 0.5% TX-100, blocked in 2% BSA, 5% donkey serum and 0.05% TX-100 in PBS and incubated overnight at 4°C with the primary antibodies. Secondary detection was performed with Alexa Fluor-coupled secondary antibodies for 3h at room temperature. Speciments were mounted with Vectashield (Vector Labs).

Cerebella (P5) were dissected, fixed and embedded in 3% low melting point agarose and 100μm sections were cut on vibratome and immunostained.

After anaesthesia, mice were perfused for 2min with fixative (1% PFA in PBS, pH 7.4) from a cannula inserted through the left ventricle into the aorta55.

The aorta, diaphragm, urinary bladder and intestine were fixed in 1% PFA at pH 7.5 for 1h, antibodies were diluted in 5% donkey serum in PBS containing 0.3% Triton-X100. The stained tissues were mounted with Vectashield (Vector Labs).

Confocal microscopy was performed with Leica TCS SP2 and SP5 confocal microscopes. ImageJ (NIH) was used for the data analysis.

The figures were assembled using Adobe Photoshop and Adobe Illustrator. The only adjustments used in the preparation of the figures were for brightness and contrast. For comparison purposes, different sample images of the same antigen were acquired under constant acquisition settings.

Image analysis

Image analysis was performed using the ImageJ software. Depth-coded stacks were generated using the dedicated plug-in, which assigned a gradient of pseudo-colours that ranging from red to blue (see ‘Depth Index’ in Fig. 2), varying according to focal plane depth.

The mean fluorescence intensities in Figs 4, ​,6,6, and Supplementary Fig. S1 were determined as the mean intensities of the fluorescent pixels after segmentation, which was performed by assigning a threshold value (constant within each experiment) of fluorescence for every image.

The vessel front density in the retinas was measured as a percentage of pixels with a value greater than the threshold applied for image segmentation.

Tip cell nuclei highlighted in Fig. 2e derived from a plane-by-plane analysis of confocal stacks. However, in order to obtain a more comprehensive visualization of the complete sprouts only stack projection are shown. For clarity, an example of a single confocal plane with highlighted nuclei is presented in Supplementary Fig. S2C.

This work is supported by Grants from Fondation Leducq Transatlantic Network of Excellence, Associazione Italiana per la Ricerca sul Cancro (AIRC) and ‘Special Program Molecular Clinical Oncology 5 × 1,000’ to AGIMM (AIRC-Gruppo Italiano Malattie Mieloproliferative), the European Community (EUSTROKE-contract-202213; OPTISTEM-contract-223098; ENDOSTEM-HEALTH-2009-241440; JUSTBRAIN-HEALTH-2009-241861; ITN Vessels), the European Research Council (ERC) and CARIPLO Foundation. We are indebted with Dr S.J. Morrison and Dr D.A. Melton for generously sharing their genetically modified mouse strains.