Research ArticleAngiogenesis

The endothelial adaptor molecule TSAd is required for VEGF-induced angiogenic sprouting through junctional c-Src activation

See allHide authors and affiliations

Science Signaling  19 Jul 2016:
Vol. 9, Issue 437, pp. ra72
DOI: 10.1126/scisignal.aad9256

Disconnecting endothelial cells for new blood vessels

The endothelial cells lining blood vessels are linked together by adherens junctions, where VE-cadherin protein complexes must come apart so that endothelial cells can migrate and proliferate to form new blood vessels. This process is triggered by activation of the receptor VEGFR2, which stimulates the kinase c-Src. Gordon et al. showed that the adaptor protein TSAd linked these two signaling molecules in the developing trachea. TSAd recruited active c-Src to adherens junctions, which resulted in the breakdown of VE-cadherin complexes and enabled the rearrangement of endothelial cells to form a new blood vessel sprout. Because TSAd was required for blood vessel formation in developing trachea, but not in the developing retina, TSAd could be targeted to prevent abnormal vascular growth in a tissue-specific manner.


Activation of vascular endothelial growth factor (VEGF) receptor 2 (VEGFR2) by VEGF binding is critical for vascular morphogenesis. In addition, VEGF disrupts the endothelial barrier by triggering the phosphorylation and turnover of the junctional molecule VE-cadherin, a process mediated by the VEGFR2 downstream effectors T cell–specific adaptor (TSAd) and the tyrosine kinase c-Src. We investigated whether the VEGFR2-TSAd–c-Src pathway was required for angiogenic sprouting. Indeed, Tsad-deficient embryoid bodies failed to sprout in response to VEGF. Tsad-deficient mice displayed impaired angiogenesis specifically during tracheal vessel development, but not during retinal vasculogenesis, and in VEGF-loaded Matrigel plugs, but not in those loaded with FGF. The SH2 and proline-rich domains of TSAd bridged VEGFR2 and c-Src, and this bridging was critical for the localization of activated c-Src to endothelial junctions and elongation of the growing sprout, but not for selection of the tip cell. These results revealed that vascular sprouting and permeability are both controlled through the VEGFR2-TSAd–c-Src signaling pathway in a subset of tissues, which may be useful in developing strategies to control tissue-specific pathological angiogenesis.


Vascular endothelial growth factor (VEGF) was originally classified as a vascular permeability factor (1) and has since been shown to regulate vascular morphogenesis, sprouting angiogenesis, and transient or chronic blood vessel permeability in physiological and pathological conditions (2, 3). Indeed, binding of VEGF to VEGF receptor 2 (VEGFR2) initiates signals regulating most, if not all, aspects of endothelial cell biology (4).

VEGFR2 contains various tyrosine residues within its intracellular domain that become phosphorylated upon VEGF stimulation, including Tyr949, Tyr1052, Tyr1057, Tyr1173, and Tyr1212 (corresponding to the numbering in the mouse VEGFR2). The interaction partners and in vivo role of these individual phosphorylation sites are now being unraveled. The phosphorylated Tyr1173 residue binds several signal transducers, including phospholipase C–γ (PLC-γ), Shc-like adapter (Sck), and Src homology 2 domain containing protein b (Shb) (4). Mice expressing a Y1173F VEGFR2 mutant show arrested endothelial expansion and die at E8.5 (5). The phosphorylated Tyr949 (Tyr951 in human) residue in VEGFR2 mediates binding of the T cell–specific adaptor (TSAd), which is essential for endothelial cell actin reorganization and cell migration through regulation of c-Src activity (6). In contrast, TSAd is dispensable for endothelial cell proliferation (6). In vitro, TSAd binds to the phosphorylated Tyr949 residue in VEGFR2 through its SH2 domain and to c-Src through its proline (Pro)–rich domain (7). This interaction leads to phosphorylation and activation of c-Src at cell junctions in cultured endothelial cells. Upon global loss of Tsad, c-Src activation is reduced, resulting in a failure of endothelial junctions to rearrange and loss of VEGF-induced permeability (7).

The Src family of cytoplasmic tyrosine kinases includes members c-Src, Yes, and Fyn, all of which are expressed in endothelial cells (8). In particular, c-Src has been implicated in the regulation of vascular permeability (9) by phosphorylating and promoting the turnover of VE-cadherin, resulting in the opening of paracellular junctions. However, c-Src in venous endothelial cells is constitutively phosphorylated as a consequence of ligand-independent, flow-mediated activation (10), and therefore, the exact role of c-Src in the permeability process remains unclear. Major substrates for the kinase activity of c-Src are involved in processes regulating cytoskeletal rearrangements, cell-cell adhesion, and cell-matrix adhesion (11). Thus, c-Src and most likely additional Src family members, such as Yes, are vital for the regulation of dynamic processes of the endothelium.

Sprouting angiogenesis occurs by endothelial cells undergoing coordinated cell behavior, involving a highly migratory, leading “tip cell” and trailing “stalk cells,” which form the lumen of the growing sprout (12, 13). This process is rapid and reversible, with cells actively shuffling along the stalk, competing for the tip cell position under the control of Notch signaling (14). VE-cadherin turnover, which is controlled by Notch and VEGF receptors, is a critical mediator of sprouting angiogenesis during development and in settings of pathologically high VEGF (14, 15). Differential adhesion between endothelial cells dependent on VE-cadherin–positive adherens junctions allows cells to dynamically rearrange and interchange positions to form tip and stalk cell patterns required for sprouting angiogenesis (15). Thus, functional remodeling of adherens junctions, mediated by VEGF/VEGFR2, is implicated in both vascular permeability and angiogenesis, suggesting that these two physiological processes may be controlled by a common signaling pathway.

Here, we sought to determine the cell-autonomous role and mechanism of action of TSAd in sprouting angiogenesis. We found that TSAd was required for VEGF-induced angiogenic sprouting in the tracheal but not retinal vasculature through regulation of sprout elongation, whereas tip cell selection was unaffected. TSAd function was exerted through regulation of c-Src activity at endothelial junctions. Once activated at junctions and in close proximity to VE-cadherin, c-Src acted to mediate VE-cadherin phosphorylation and turnover, a process that is required for sprout formation.


TSAd is required for VEGF-induced sprouting

We have previously shown that upon VEGF stimulation, TSAd forms a transient complex with VEGFR2, c-Src, and VE-cadherin, a process that is critical for junctional integrity and vascular permeability (6, 7). To determine whether TSAd is involved in angiogenesis induced by VEGF/VEGFR2 signaling (16), we used a three-dimensional (3D) sprouting assay, in which collagen-embedded aggregates of differentiating embryonic stem (ES) cells derived from blastocysts, embryoid bodies (EBs), form angiogenic sprouts in response to VEGF (17). Upon loss of Tsad, VEGF-treated EBs showed a reduction in endothelial sprout formation (Fig. 1, A and B). Reintroduction of TSAd by lentiviral transduction of a full-length Tsad construct fully rescued the sprouting ability of Tsad−/− EBs. There was no significant change in the sprouting capacity of Tsad-transduced wild-type EBs (Fig. 1, A and B). This finding confirmed that upon loss of TSAd, VEGF-induced angiogenesis was impaired, a process that could be reversed with reintroduction of TSAd.

Fig. 1 TSAd is necessary for VEGF-induced sprouting.

(A) 3D EB sprouting assay with wild-type (WT) and Tsad−/− ES cells transfected with either control (empty vector) or full-length TSAd lentivirus. Endothelial cells were stained with CD31. Scale bars, 500 μm. (B) Quantification of CD31-positive pixels in (A) normalized to WT empty vector; mean + SEM. Statistical significance was assessed using the Mann-Whitney test, *P < 0.05; n = 5 EBs per infection condition. (C) Matrigel plugs loaded with different growth factors (VEGF, FGF2, or combination of both) of WT and Tsad−/− mice. Dashed line shows the border of the Matrigel (Mg). Mt, mouse tissue. Scale bars, 100 μm. (D) Quantification of (C), normalized to WT; mean + SEM. Statistical significance was assessed using the Mann-Whitney test, *P < 0.05; n = 9 (WT VEGF), n = 10 (Tsad−/− VEGF), n = 7 (WT and Tsad−/− FGF2), and n = 8 (WT and Tsad−/− VEGF + FGF2); n is number of mice.

To corroborate a role for TSAd in angiogenic sprouting in vivo, Matrigel plugs containing VEGF and/or fibroblast growth factor 2 (FGF2) were implanted into wild-type or Tsad−/− mice, which lack TSAd in all cell types. These mice are healthy and fertile when unchallenged (18). Matrigel plugs containing VEGF showed a significant decrease in vascularization in mice lacking TSAd compared to wild-type. In contrast, the CD31-positive area in Matrigel plugs containing FGF2 alone or VEGF and FGF2 was similar in Tsad−/− and wild-type mice (Fig. 1, C and D). These data revealed that TSAd was specifically required for VEGF-induced sprouting angiogenesis in vivo.

The retinal vasculature develops in an ordered manner during early postnatal stages by radial outgrowth from the optic nerve. This process is mediated by balanced signaling through VEGF/VEGFR2 and Delta-like 4/Notch (19) as well as BMP/TGF-β/ALK signaling (20, 21). Unexpectedly, considering the loss of angiogenic sprouting in the TSAd-deficient conditions described above, we did not observe any differences in the number of cells with filopodia at the vascular front, branch points, outgrowth, or overall vascularization of Tsad−/− retinas at postnatal day 5 (P5) compared to wild-type littermates (fig. S1). Together, these data suggested that TSAd was only required for sprouting angiogenesis in certain VEGF-dependent settings.

Endothelial TSAd is required for VEGF-induced sprouting and developmental angiogenesis

In vivo, TSAd is present in certain immune cells as well as in endothelial cells (2225). To exclude the possibility that biological compensation in the global Tsad−/− knockout mice concealed a role for TSAd during retinal angiogenesis, we generated inducible, endothelial-specific Tsad knockout mice (Tsadfl/fl;Cdh5-CreERT2) (fig. S2A) (26, 27). Tsadfl/fl;Cdh5-CreERT2 mice are fertile with no apparent health defects. We observed an 80% decrease in TSAd protein abundance in lung lysates from Tsadfl/fl;Cdh5-CreERT2 mice compared to that from wild-type Tsadfl/fl mice after administration with tamoxifen (fig. S2, B and C). No change was detected in the phosphorylation of VEGFR2 Tyr1173 after administration of VEGF by tail vein injection (fig. S2B).

To explore whether endothelial TSAd was required for VEGF-induced angiogenesis, we implanted Matrigel plugs into wild-type and Tsadfl/fl;mTmG;Cdh5-CreERT2 mice. Cre efficiency (and therefore recombination of the Tsad floxed allele) in vessels invading the Matrigel was confirmed by expression of green fluorescent protein (GFP) after recombination of the mTmG reporter allele (28). Mirroring our findings in the global Tsad−/− knockout mice (Fig. 1, C and D), Matrigel plugs containing VEGF, but not FGF2, showed a significant decrease in vascularization in Tsadfl/fl;mTmG;Cdh5-CreERT2 mice compared to wild-type mice (Fig. 2, A and B).

Fig. 2 Endothelial TSAd is necessary for VEGF-induced and peripheral organ angiogenesis.

(A) Matrigel plugs loaded with growth factors (VEGF or FGF2) of Tsadfl/fl (WT) and Tsadfl/fl;mTmG;Cdh5-CreERT2 mice. Scale bars, 100 μm. (B) Quantification of CD31-positive pixels in (A); mean + SEM. Statistical significance was assessed using the Mann-Whitney test, *P < 0.05; n = 9 (WT VEGF), n = 7 (Tsadfl/fl;mTmG;Cdh5-CreERT2 VEGF), n = 10 (WT FGF2), and n = 11 (Tsadfl/fl;mTmG;Cdh5-CreERT2 FGF2); n is number of mice. (C) Tracheas from Tsadfl/fl or Tsadfl/fl;Cdh5-CreERT2 littermates. Scale bars, 100 μm. (D) Quantification of (C); mean + SEM. Statistical significance was assessed using the Mann-Whitney test, **P < 0.0025; n = 8 (Tsadfl/fl) and n = 12 (Tsadfl/fl;Cdh5-CreERT2); n is number of mice. (E) High-magnification images of the sprouting front of tracheas. Scale bars, 50 μm. (F) Heat map of VE-cadherin morphology of tracheas in (E). (G) Quantification of (F). The VE-cadherin morphology in each patch was manually classified using a scale from active (red, serrated line, and bright interior) to inactive (blue, straight line, and dark interior). Mean ± SEM. Statistical significance was assessed using the Mann-Whitney test (highly active junctions, *P < 0.05) and χ2 contingency test for trend (**P < 0.005); n = 4 mice. (H) Quantification of vessel width in (E); mean + SEM. n > 3 mice. Statistical significance was assessed using the Mann-Whitney test, *P < 0.05.

To examine whether TSAd is required in endothelial cells during developmental angiogenesis, we used the trachea as a model for endothelial sprouting in peripheral organs. The airway vasculature is remodeled rapidly after birth, with blood vessels sprouting in a VEGF-dependent manner over the tracheal cartilage rings up to P7, to form an ordered, ladder-like pattern (29). We induced endothelial deletion of Tsad after birth (P1), and then analyzed the tracheal vasculature at P5, at which point we observed significantly fewer capillaries crossing the tracheal cartilage rings in Tsadfl/fl;Cdh5-CreERT2 mice compared to that in wild-type littermates (Fig. 2, C and D). These vessels were blood and not lymphatic in nature by virtue of their morphology and lack of LYVE-1 (fig. S3). Because adherence junction activity is important for developmental angiogenesis, we analyzed VE-cadherin morphology using established image analysis software (15) in the actively sprouting areas of the trachea at the edge of the cartilage rings (Fig. 2, E and F). The VE-cadherin pattern of individual junctions was rated either as “active,” having a serrated and vesicular appearance, or “inactive,” with a straight and less vesicular appearance. Wild-type mice had a significant increase in the prevalence of highly active adherens junctions compared to Tsadfl/fl;Cdh5-CreERT2 mice and an overall increase in active patches over inactive patches (Fig. 2G). A decrease in endothelial heterogeneity and a subsequent equalization of junctional activity among neighboring cells drive vessel expansion instead of branching (30), and in agreement, we observed an increase in vessel width at the sprouting front of Tsadfl/fl;Cdh5-CreERT2 tracheas (Fig. 2H). Together, these results demonstrate that endothelial TSAd is required for VEGF-induced developmental angiogenesis in peripheral organs such as the trachea in a manner correlating with adherens junction morphology.

Although we observed no apparent defects in retinal angiogenesis of constitutive Tsad−/− mice (fig. S1), these mice display a mild decrease in the activity of retinal adherence junctions (15). To further explore the role of endothelial TSAd in retinal sprouting, we induced endothelial deletion of Tsad after birth (P1), followed by analysis of the retinal vasculature at P5. Similar to Tsad−/− mice, we did not detect any changes in the number of tip cells, branch points, outgrowth, or overall vascularization of Tsadfl/fl;Cdh5-CreERT2 retinas compared to wild-type littermate retinas (Fig. 3, A to F). Moreover, in contrast to the findings on Tsad−/− junctional morphology (15), we did not observe any significant changes in the adherens junction morphology of Tsadfl/fl;Cdh5-CreERT2 retinas (Fig. 3, G and H). The differences observed in Tsad−/− and Tsadfl/fl;Cdh5-CreERT2 retinas may be due to an incomplete deletion of Tsad in the retinas of Tsadfl/fl;Cdh5-CreERT2 mice. Alternatively, the decrease in adherens junction activity observed in Tsad−/− retinas may be due to a nonautonomous loss of TSAd in another cell compartment. Tsad expression is detectable by microarray analysis of the whole retina during development (31); however, analysis of published single-cell RNA sequencing databases (32, 33) did not reveal Tsad expression in endothelial cells (defined by the expression of Cdh5, which encodes VE-cadherin) (fig. S4). Together, these results demonstrated that TSAd was dispensable for developmental retinal angiogenesis, yet required for angiogenic sprouting of the postnatal tracheal vasculature.

Fig. 3 Endothelial TSAd is not required for developmental retinal angiogenesis.

(A and B) Retinas of mice injected with 100 μg of tamoxifen at P1 and P2 and sacrificed at P5. Scale bars, 500 μm (A) and 100 μm (B). (C to F) No changes were observed in the number of tip cells (C), vascular area (D), outgrowth from the optic nerve (E), or number of branch points (F) in Tsadfl/fl;Cdh5-CreERT2 mice compared to WT littermates. n = 10 (Tsadfl/fl) and n = 6 (Tsadfl/fl;Cdh5-CreERT2); n is number of retinas analyzed. Mean ± SD; statistical significance was assessed using the Mann-Whitney test. n.s., not significant. (G) Heat map of VE-cadherin morphology of retinas. (H) Quantification of (G). The VE-cadherin morphology in each patch was hand-classified with adherens junctions classified from active (red, serrated line, and bright interior) to inactive (blue, straight line, and dark interior). Mean ± SEM. Statistical significance was assessed using the χ2 contingency test for trend (not significant). n = 8 (Tsadfl/fl) and n = 9 (Tsadfl/fl;Cdh5-CreERT2); n is number of retinas analyzed. IB4, isolectin-B4.

TSAd is not required for tip cell selection

Blood vessel sprouts are formed by coordinated, dynamic endothelial cell behavior, with some cells forming filopodial protrusions and assuming the tip position (tip cells) and neighboring cells contributing to the stability and elongation of the new vessel (stalk cells) (12). To investigate the effect of TSAd on tip cell behavior and sprout formation, we used a chimeric human umbilical cord endothelial cell (HUVEC) VEGF-induced sprouting assay in a 3D fibrin gel. If loss of TSAd rendered cells unable to reach the tip, we expected to see more TSAd-deficient cells in the stalk cell position. We observed no change in tip/stalk cell distribution in cultures composed of a mixture of control small interfering RNA (siRNA) and Tsad siRNA–treated HUVECs (Fig. 4, A to D) and confirmed that the silencing of Tsad by siRNA was efficient by Western blot (fig. S2D). Although there was no change in tip/stalk cell specification, when 50% (Fig. 4B) or 100% (Fig. 4C) of cells were transfected with Tsad siRNA, there was a significant decrease in the ability of HUVECs to form sprouts (Fig. 4E), confirming a role for TSAd in angiogenic sprouting.

Fig. 4 TSAd is not required for tip cell specification.

(A to C) Sprout formation in 100% control HUVECs [scramble (sc) siRNA/sc siRNA], 50% TSAd-deficient HUVECs (sc siRNA/Tsad siRNA), and 100% TSAd-deficient HUVECs (Tsad siRNA/Tsad siRNA). Cells treated with siRNA were dyed green or red as indicated. Representative images of sprouts 4 days after embedding in gel. Scale bars, 200 μm. (D) Quantification of the percentage of siRNA-transfected cells at the tip position. Mean ± SD; n = 3 experiments, >100 sprouts per transfection condition analyzed. Statistical significance was assessed using the χ2 test (not significant). (E) Quantification of sprout formation in (A) to (C). n = 3 experiments. Mean + SEM. Statistical significance was assessed using the Kruskal-Wallis test, **P < 0.01. (F) Analysis of tip cell position in retinal and tracheal blood vessels from tamoxifen-injected mice. Recombined cells expressed GFP. Scale bars, 100 μm (left) and 50 μm (right). (G and H) Quantification of recombined TSAd-deficient cells at the tip, normalized to the overall contribution of TSAd-deficient cells to the endothelium. Mean ± SD; statistical significance was assessed using the χ2 test. n = 8 retinas per genotype; n = 4 mice per genotype for the tracheal analysis.

Low-dose tamoxifen injection in conditional Tsadfl/fl;Cdh5-CreERT2 mice crossed with mTmG reporter mice (28) produced recombination of the Tsad floxed allele as well as the mTmG reporter allele in a subset of GFP-positive cells, thus creating a chimeric situation. In mice where about 80% of cells were GFP-positive and had therefore experienced Cre activity, there was no significant difference in the retina (Fig. 4, F and G) or the trachea (Fig. 4, F and H) in the ability of GFP-positive, TSAdlow cells to reach the tip position compared to GFP-negative, TSAdhigh cells. Thus, despite a decrease in capillary sprouting in the trachea (Fig. 2), TSAd-deficient cells could still reach the tip position in this tissue. Furthermore, we did not observe any change in Dll4 immunostaining at the vascular front of TSAd-deficient retinas (fig. S5), suggesting that VEGFR2-TSAd signaling does not regulate Delta/Notch activity. These findings demonstrated that TSAd was not required for tip/stalk cell selection in vivo. Although tip cell selection under the control of Delta/Notch and BMP/TGF-β/ALK signaling is essential for sprouting angiogenesis, our results show that additional mechanisms governed by VEGF and TSAd are also required for sprout formation.

TSAd bridges VEGFR2 and c-Src through its SH2 and C-terminal domains

Because TSAd was not required for tip/stalk cell specification, we investigated alternative downstream pathways required for sprouting angiogenesis. Analysis of the phosphopeptide-capturing capabilities of fusion proteins representing different VEGFR2, TSAd, and c-Src motifs has shown that TSAd binds to the phosphorylated Tyr949 residue in VEGFR2 through its SH2 domain, and to c-Src through its Pro-rich domain (7). This interaction correlates with phosphorylation and activation of c-Src at endothelial junctions in cultured endothelial cells (7). We asked whether these molecular interactions occurred also between intact proteins in endothelial cells undergoing sprouting angiogenesis in more complex models. For this purpose, we generated EBs expressing mutant constructs of TSAd either (i) with a disrupted SH2 domain, (ii) lacking the Pro-rich domain, or (iii) with potential phosphorylated Tyr sites eliminated by Tyr→Phe mutation (Fig. 5, A and B). We found that disruption of the SH2 domain (which would prevent binding to VEGFR2) or removal of the Pro-rich domain (which would prevent binding to c-Src) of TSAd significantly reduced EB sprouting, whereas a mild decrease was observed upon elimination of the phosphorylated Tyr sites (Fig. 5, A and C). The only construct that rescued the sprouting capacity of Tsad−/− EBs was the full-length TSAd (Fig. 1 and fig. S6); in contrast, we observed a marginal rescue with the phosphorylated Tyr–site mutant and no rescue with the SH2 domain and Pro-rich domain deletion mutants (fig. S6). Furthermore, Tsad−/− EBs transduced with the disrupted SH2 domain or Pro-rich domain dissociated in the matrix, possibly implying a role for TSAd in cell-cell adhesion during stem cell differentiation (fig. S6). The mechanism underlying the mild rescue by the Y- to F-mutated TSAd is unclear because the role of these tyrosine residues in potential downstream signaling is not understood. Immunoprecipitation analysis confirmed that the interaction between TSAd and VEGFR2 was lost upon disruption of TSAd’s SH2 domain (Fig. 5D). Together, these data demonstrate that TSAd acts as a bridge between VEGFR2 and c-Src, by binding to VEGFR2 through its SH2 domain, to promote sprouting angiogenesis.

Fig. 5 The SH2 and C-terminal domains of TSAd are necessary for bridging between VEGFR2 and Src.

(A) 3D EB sprouting assay with WT ES cells transfected with control lentivirus (empty vector) or different TSAd mutants. Endothelial cells were visualized with CD31. n = 19 (empty vector), n = 15 (SH2-disrupted), n = 16 (Δ Pro-rich), and n = 10 [phospho Tyr (pY)–disrupted)]; n is the number of EBs analyzed per infection condition. Scale bars, 500 μm. (B) Schematic presentation of different constructs in (A). (C) Quantification of CD31-positive pixels in (A). Mean + SEM; statistical significance was assessed using the Mann-Whitney test, **P < 0.005,***P < 0.001, and ***P < 0.0001. (D) Pull down of the different TSAd constructs from HEK 293T lysates with biotinylated kinase insert domain (KID) peptides and streptavidin beads. n = 2 experiments. IB, immunoblot.

TSAd promotes c-Src recruitment and activity at junctions to promote sprouting angiogenesis

To confirm that TSAd acted by mediating c-Src activation at endothelial junctions to promote sprouting angiogenesis, we stained EB sprouts for activated, phosphorylated Tyr416 c-Src. Although we did not find a marked decrease in the total staining intensity of phosphorylated Tyr416 in the sprouts of Tsad−/− EB (Fig. 6A), we observed a more diffuse staining pattern throughout the vessel and significantly reduced colocalization between phosphorylated Tyr416 c-Src and VE-cadherin at junctions (Fig. 6, A and B). Inhibition of total c-Src activity using the potent c-Src inhibitor Saracatinib (AZD0530) (34) replicated the phenotype observed upon loss of TSAd in EBs in a dose-dependent manner (Fig. 6, C and D). We observed a more diffuse and less intense phosphorylated Tyr416 c-Src staining pattern at junctions in sprouts of the inhibitor-treated EBs (Fig. 6E). Finally, Saracatinib treatment significantly reduced the capacity for sprouting in the HUVEC 3D fibrin assay compared to mock treatment (Fig. 6, F to H).

Fig. 6 TSAd is necessary to localize c-Src to junctions in sprouts.

(A) Individual sprouts of WT and Tsad−/− EBs. n = 12 (WT) and n = 12 (Tsad−/−); n is the number of EBs analyzed. EBs were costained for phosphorylated (p) Tyr416 c-Src (pSrc; green) and VE-cadherin (red) to visualize junctions. Scale bars, 20 μm. (B) Quantification of phosphorylated Tyr416 c-Src localized to VE-cadherin–positive junctions. Mean + SEM; statistical significance was assessed using the Mann-Whitney test, ***P < 0.001. (C) WT EBs treated with different concentrations of Src inhibitor Saracatinib (AZD0530). n = 13 [0.1% dimethyl sulfoxide (DMSO)], n = 11 (0.5 μM AZD), n = 14 (1.0 μM AZD), and n = 11 (2.0 μM AZD); n is the number of EBs analyzed. Endothelial cells were visualized with CD31. Scale bars, 500 μm. (D) Quantification of CD31-positive pixels in (C). Mean + SEM; statistical significance was assessed using the Mann-Whitney test, **P < 0.01, ***P < 0.001, ****P < 0.0001. (E) AZD inhibits Src activity in WT EB sprouts, as assessed by staining for pTyr416. n = 2 experiments. (F) Representative images of HUVEC sprouting assay after treatment with AZD. Scale bars, 200 μm. (G and H) Quantification of sprout length and the number of sprouts per bead. Mean ± SD; n = 3 experiments, >50 sprouts per condition analyzed. Statistical significance was assessed using the unpaired t test, *P < 0.05, ***P < 0.001.

We confirmed reduced accumulation of phosphorylated Tyr416 c-Src at endothelial junctions in cultures of endothelial cells isolated from lungs of Tsad−/− mice (Fig. 7, A and B). VE-cadherin’s role in vascular permeability and leukocyte extravasation is controlled by phosphorylation of several tyrosine residues, and Tyr658 and Tyr685 are of particular interest because they have been implicated downstream of VEGF and c-Src (10, 35). We therefore predicted these residues to be under the control of the VEGFR2-TSAd–c-Src pathway. Isolated endothelial cells from Tsad−/− mice displayed reduced phosphorylation of Tyr658 and Tyr685 in VE-cadherin at junctions (Fig. 7, C to F). Furthermore, lungs (containing about 50% endothelial cells) retrieved from Tsad−/− mice showed a modest decrease in phosphorylation of VE-cadherin Tyr658 and Tyr685, both basally and after 1 min of VEGF stimulation (Fig. 7, G and H). In summary, activation of c-Src at junctions, through interaction with the VEGFR2-TSAd complex, mediates VE-cadherin phosphorylation and internalization, which is necessary for sprout formation and angiogenesis.

Fig. 7 TSAd is required for c-Src and VE-cadherin phosphorylation at junctions.

(A to F) Endothelial cells isolated from lungs of WT or Tsad−/− mice stimulated with VEGF (80 ng/ml) and stained for phosphorylated Tyr416 c-Src (pSrc416), phosphorylated Tyr658 VE-cadherin (pVEC658), or phosphorylated Tyr685 VE-cadherin (pVEC685) and total VE-cadherin. (B), (D), and (F) are quantitation of (A), (C), and (E), respectively. Scale bar, 20 μm. n = 3 mice for each genotype; one cell isolation and immunostaining for each condition per mouse. Mean + SEM; statistical significance was assessed using the unpaired t test, *P < 0.05, **P < 0.01. (G) WT or Tsad−/− mice stimulated with VEGF (5 μg) for 1 min and lung lysates tested for VEGFR2 and VE-cadherin phosphorylation. Each lane represents one mouse; n = 3 mice per genotype, n = 2 for each Western blot. (H) Quantitation of (G) phospho–VE-cadherin normalized to total VE-cadherin. n = 3 mice for each condition. PBS, phosphate-buffered saline; a.u. arbitrary units.


During VEGF-induced vascular permeability, TSAd exerts its function to disrupt endothelial cell junction integrity by mediating VEGF-induced c-Src activity at junctions, thereby promoting VE-cadherin internalization (7). In agreement, removing the binding site for TSAd’s SH2 domain, as is the case in the Vegfr2Y949F/Y949F mouse, results in loss of VEGF-induced adherens junction rearrangement (36). Here, we showed that VEGFR2-TSAd–c-Src activity at junctions was also required specifically for VEGF-induced sprouting angiogenesis. We propose that junctional inertia resulting from a block in VEGF-induced junctional c-Src activity impairs rearrangement of endothelial cells in the sprout, thereby restricting sprout elongation.

c-Src targets many substrates implicated in cell-cell adhesion and cytoskeletal organization (11), providing the mechanistic basis for how the VEGFR2-TSAd–c-Src pathway affects both angiogenic sprouting and vascular permeability. Opening of endothelial junctions may depend not only on c-Src regulation of adherens junctions but also on c-Src–mediated phosphorylation of cytoskeletal components, allowing withdrawal of the overlapping endothelial flaps at the junction. Moreover, vascular leakage has been suggested to be important during angiogenesis by forming a fibrin matrix, which is used as a scaffold for the growing angiogenic sprout (37). Whether the potential loss of such a scaffold contributes to the reduced angiogenic sprouting shown here has not been directly addressed.

There are two main subcellular pools of c-Src: at junctions (10) and at focal adhesions (7, 38). Eliminating TSAd by siRNA or in endothelial cells isolated from Tsad−/− mice does not interfere with c-Src activity at focal adhesions (7). It is conceivable that the different subcellular pools of c-Src are controlled by different pathways, leading ultimately to phosphorylation of distinct sets of c-Src substrates regulating cell-cell (junctions) or cell-matrix (focal adhesions) dynamics. Because of technical restraints, it was not possible to define the subcellular localization of c-Src protein in TSAd-deficient tissues in vivo. In vitro analyses of endothelial cells isolated from Tsad−/− mice showed a marked overall loss in c-Src phosphorylation on Tyr416 (Fig. 7, A and B), due to either a more general suppression in c-Src activation than specifically at endothelial junctions or changes in c-Src turnover as a consequence of interruption of the VEGFR2-TSAd–c-Src pathway.

Upon VEGF stimulation, c-Src disrupts a VEGFR2–VE-cadherin–β-catenin complex at junctions to induce vascular permeability (39). We hypothesize that c-Src activated through TSAd primarily acts on VE-cadherin, either directly or indirectly through the substrates Vav and p21-activated kinase (PAK) (40, 41), to mediate cell-cell junction turnover. In agreement, phosphorylation of both VE-cadherin Tyr658 and Tyr685 was decreased in Tsad−/− endothelial cells (Fig. 7). Whereas phosphorylation of Tyr685 is required for VEGF-induced vascular permeability (35), the precise downstream consequence of phosphorylation of these sites during sprouting angiogenesis remains unclear.

We have previously demonstrated that Tsad−/− mice have fewer active junctions in the retina and upon VEGF-induced permeability (7, 15), suggesting that TSAd is required for angiogenesis. However, we did not detect any abnormalities in the development of the retinal vasculature in Tsad−/− or Tsadfl/fl;Cdh5-CreERT2 mice, or any changes in VE-cadherin patterning in the retinas of Tsadfl/fl;Cdh5-CreERT2 mice, indicating that TSAd is dispensable for retinal angiogenesis. Vessels in Tsad−/− mice maintain a differential patchwork of VE-cadherin, despite having less active junctions overall (15). This differential adhesion still allows for cells to dynamically rearrange, migrate, and form sprouts, in part, through VEGF-independent stimuli. Analysis of VE-cadherin in the tracheas of Tsadfl/fl;Cdh5-CreERT2 mice revealed a decrease in overall turnover of VE-cadherin in sprouting vessels. This effect likely leads to an inhibition in the speed in which cells can move within the sprout, resulting in decreased angiogenesis in settings such as EBs, Matrigel plugs, and the trachea.

Both the retina and trachea develop in an ordered fashion characterized by extensive growth and remodeling after birth, a process that is dependent on VEGF and hypoxia (12, 29). Blockade of VEGFR2 signaling using the blocking antibody DC101 virtually halts capillary sprouting in the trachea (29), indicating the dependence of the VEGF-VEGFR2 pathway for angiogenesis in this tissue. We noted that the sprouting defect in the Matrigel plugs was specific to VEGF, whereas sprouting in the presence of FGF2 was intact. In contrast, various studies have indicated that retinal vascular sprouting is dependent not only on VEGF but also on multiple signaling pathways that integrate to control sprouting in a complex fashion [reviewed in (42)]. VEGFR3 is also required during retinal angiogenesis to stabilize vascular branches, although this process still requires VEGFR2 expression and Notch activity (43, 44). Whether TSAd can interact with VEGFR3 to mediate its downstream signaling remains unknown. It is possible that the differential requirement for VEGFR2 and VEGFR3 between vascular beds contributes to the lack of apparent of phenotype in the retinas of TSAd mutant mice. Although the retinal vessels of Tsadfl/fl;Cdh5-CreERT2 mice did not have a sprouting phenotype, there was a significant decrease in sprouting in several other tissues, notably in the trachea. This difference in regulation between the vascular beds in the retina and the trachea could reflect the higher integrity of the retinal vasculature, which is protected by the retinal blood barrier. Central nervous system (CNS) organs such as the retina are not only controlled by various integrating signaling pathways but also display differential apicobasal VEGFR localization compared to peripheral organs (45). The emerging question of whether angiogenesis is controlled by general mechanisms, or in an organ-specific manner with effects being restricted to certain tissues, remains largely unexplored. Our results indicate that suppressing the VEGFR2-TSAd–c-Src signaling axis specifically targets angiogenesis in peripheral organs such as the trachea while leaving CNS organs such as the retina largely unaffected. This specificity may be useful for specific targeting of abnormal vascular growth in pathologies associated with airway inflammation (4648).

The specification of tip and stalk cells is regulated by Dll4/Notch signaling downstream of VEGF receptors (14). We did not observe any changes in Dll4 abundance in Tsadfl/fl;Cdh5-CreERT2 mice (fig. S5), and TSAd-deficient cells were still able to reach the tip position (Fig. 4) despite TSAd-null vessels displaying a decrease in sprout formation in vitro (Fig. 4, C and F) and in vivo (Fig. 2). Although often going hand in hand, disruption of tip cell specification does not always accompany abnormalities in sprout formation, or vice versa. For example, cells with a heterozygous mutation of neuropilin1, a key effector of tip cell selection downstream of Notch, display a severe decrease in their ability to reach the tip, yet these mice do not have any apparent defects in vascular morphogenesis (20). Conversely, here, we identified TSAd as a key effector of sprout morphogenesis that does not regulate tip cell selection or Delta/Notch signaling. Rather, TSAd mediated c-Src phosphorylation and internalization of VE-cadherin at junctions, allowing sprouts to convert into endothelial tubes, connect to other vessels, and form a functional vascular network.

In conclusion, we have identified the VEGFR2 Tyr949–TSAd–c-Src signaling axis as a regulator of VEGF-induced sprouting angiogenesis in vitro and in an organ-specific manner in vivo. Although TSAd does not control overall c-Src abundance, it is required to act as a scaffold between VEGFR2 and c-Src at endothelial cell junctions in sprouts, to bring active c-Src into close proximity to VE-cadherin and mediate adherence junction turnover critical for sprout formation and elongation.


Antibodies and growth factors

The following antibodies were used: rat anti-CD31 (BD Biosciences, 553370), Armenian hamster anti-CD31 (Thermo Scientific, MA3105), rabbit anti–phospho-VEGFR2 (Tyr1175) (Cell Signaling, 2478), rabbit anti-VEGFR2 (Cell Signaling, 2479), rabbit anti–phospho–c-Src (Tyr416) (Cell Signaling, 2101), rabbit anti–phospho–c-Src (Tyr418) (Invitrogen, 44660G), rabbit anti–phospho–VE-cadherin (Tyr658) [from E. Dejana (10)], rabbit anti–phospho–VE-cadherin (Tyr685) [from E. Dejana (10)], rabbit anti-Erg1/Erg2/Erg3 (Abcam, ab92513), rabbit anti-GFP (Santa Cruz Biotechnology, sc8334), rat anti–VE-cadherin (BD Biosciences, 555289), goat anti–VE-cadherin (R&D Systems, AF1002), rabbit anti–LYVE-1 (Millipore, AB2988), goat anti-Dll4 (R&D Systems, AF1389), sheep anti-SH2D2A/TSAd (R&D Systems, AF6265), sheep anti-SH2D2A/TSAd (R&D Systems, AF6020), rabbit anti–β-catenin (Millipore, AB19022), mouse anti–α-tubulin (Sigma-Aldrich, T9026), mouse anti–β-catenin (BD Transduction Laboratories, 610153), goat anti-fibrinogen (Nordic Immunology, GAM/Fbg/7S), and isolectin-B4 directly conjugated to Alexa 568 (Life Technologies/Thermo Scientific, I21412). Fluorescently labeled secondary antibodies were obtained from Invitrogen. Rabbit anti-TSAd antiserum was a gift from V. Shapiro (Mayo Clinic).

Recombinant VEGFA165 (100-20) was purchased from PeproTech EC Ltd. and used for in vitro analyses. Dog VEGFA164, which is 99% identical to mouse VEGFA, was used for in vivo analyses and was provided by K. Ballmer-Hofer (Paul Scherrer Institute). Recombinant VEGFR2 KID was provided by K. Ballmer-Hofer. Mouse FGF2 was obtained from PeproTech (450-33).

Transduction of lentiviruses

Murine Tsad (accession BC034847) complementary DNA (cDNA) was amplified by polymerase chain reaction (PCR) from IMAGE clone 3327113 with a 5′ Asc I site and a 3′ Nhe I site. Subsequently, it was cloned into pFUGIE (49) between the human ubiquitin promoter and the IRES-GFP. Mutant Tsad constructs were generated using QuickChange Multisite-directed mutagenesis (Agilent Technologies). The correct sequence was verified by sequencing. Lentivirus was produced in human embryonic kidney (HEK) 293T cells using the third-generation packaging system and concentrated by ultracentrifugation. Wild-type and Tsad−/− mouse ES cells were transduced with LvTSAd, empty vector, or mutated TSAd lentivirus. After transduction, cells were seeded sparsely on mouse embryonic fibroblast feeders, and GFP-positive colonies were picked and characterized.

ES cell establishment

C57BL/6 murine ES cells were established as previously described (50). Tsad−/− ES cell lines were established from Tsad−/− blastocysts by the Uppsala University Transgenic Facility (

EB culture

ES cells were induced to differentiate by omitting leukemia inhibitory factor (LIF) from the medium and aggregated in hanging drops to create EBs, as described previously (17). ES cells were cultured in Dulbecco’s minimum essential medium (DMEM)/glutamax (Gibco), 25 mM Hepes (Gibco), 1.2 mM sodium pyruvate (Gibco), and 19 mM monothioglycerol (Sigma), supplemented with 15% fetal bovine serum (FBS) and 1000 U/ml LIF (Chemicon). At day 0, 1200 cells were aggregated in hanging drops (20 μl) without LIF. At day 4, the EBs were placed on top of a polymerized collagen I gel composed of collagen I (1.5 mg/ml; Inamed Biomaterials) in Ham’s F12 medium (PromoCell), 6.26 mM NaOH, 20 mM Hepes, 0.117% NaCO3, and 1% Glutamax (Gibco/Invitrogen) and subsequently covered with a second layer of the collagen I mix. Two hours later, medium supplemented with VEGF-A165 (50 ng/ml) was added. Medium was changed every other day, and at day 14, EBs were fixed and processed for whole-mount immunofluorescence.

Immunofluorescence staining of EBs

EBs in collagen I gel were fixed for 30 min in 4% paraformaldehyde in PBS. After washing, samples were blocked and permeabilized for 2 hours in 3% bovine serum albumin (BSA) and 0.2% Triton X-100 in PBS, followed by sequential overnight incubation of primary and fluorescently labeled secondary antibodies. After washing, samples were kept in PBS at 4°C before imaging using a Zeiss LSM700 confocal microscope. Quantification was performed using ImageJ.

Sprouting assay

HUVECs (PromoCell) were cultured in EGM-2 BulletKit (Lonza), following the manufacturer’s manual. HUVECs were transfected with 25 pmol of siRNA for TSAd (Invitrogen; siTsad 3, 5′-­CCAGTACAGCCCAATCATCAA-­3′) or stealth RNA interference (RNAi)–negative control (Invitrogen) per six wells using 2.5 μl of RNAiMax (Invitrogen) according to the manufacturer’s instructions. After 24 hours, HUVECs were labeled with either PKH26 (red) or PKH67 (green) (Sigma-Aldrich) according to the manufacturer’s instructions, and a ratio of 1:1 cells were coated on cytodex 3 microcarrier beads (Sigma-Aldrich) for 24 hours. For Src inhibitor experiments, all cells were dyed with PKH67 and coated on beads. Beads were embedded in a fibrin gel [fibrinogen (2.5 mg/ml; Sigma-Aldrich) in EBM-2 (Lonza) supplemented with 2% FBS and aprotinin (50 mg/ml; Sigma-Aldrich)] with fibrinogen solution clotted with 1 U of thrombin (Sigma-Aldrich) for 30 min at 37°C. WI38 cells (25,000 cells per well), in EBM-2 supplemented with 2% FBS and VEGF (50 ng/ml), were then plated on top of the fibrin layer. Sprouts were imaged 3 to 4 days later, and images were acquired using a Zeiss LSM700 confocal microscope.

In vitro Src inhibition assay

The Src inhibitor Saracatinib (AZD0530) was a gift from AstraZeneca (Macclesfield). The drug was administered to EB cultures in complete DMEM (as above) containing either vehicle alone (0.1% DMSO) or DMSO/saracatinib (0.5, 1.0, or 2.0 μM) from day 4 to day 14. VEGF-A165 was added at 50 ng/ml as indicated. Medium with complements was replaced every other day, and EBs were harvested for immunofluorescence staining at day 14. For the HUVEC sprouting assay, the drug was administered 24 hours after HUVEC-coated microcarrier beads were embedded in the fibrin gel, as described above. Beads were imaged 48 hours later. The inhibitory effect of saracatinib on VEGF-induced angiogenesis was determined by fluorescence microscopy using a Zeiss LSM700 confocal microscope. Quantification was performed using ImageJ.

Pull-down assay

HEK 293T cells were transfected with 18 μg of TSAd mutant expression vector cDNA using polyethylenimine (PEI) and lysed in commercial radioimmunoprecipitation assay (RIPA) buffer (ProteinSimple). KID was biotinylated using FluoReporter Mini-Biotin-XX Protein Labeling Kit (Invitrogen, F-6347), following the manufacturer’s manual. Lysates were incubated with 5 pmol of biotinylated KID for 2 hours at 4°C and precipitated using Streptavidin Sepharose High Performance (GE Healthcare, 17-5113-01) at 4°C for 1 hour. Proteins were released by boiling in NuPAGE LDS Sample Buffer (Invitrogen, NP0007) and NuPAGE Sample Reducing Agent (Invitrogen, NP0009). Proteins were separated by NuPAGE Novex 4-12% Bis-Tris Gel (Invitrogen) and transferred to nitrocellulose (Hybond-C Extra, Amersham Biosciences), blocked in 5% milk/TBS, 0.2% Tween 20, followed by incubation with primary and horseradish peroxidase–conjugated secondary antibodies in block. Blots were developed using enhanced chemiluminesence (GE Healthcare) with ChemiDoc MP Imaging System (Bio-Rad Laboratories).

Immunoprecipitation and immunoblotting

Tissues were lysed in commercial RIPA buffer (ProteinSimple) with protease and phosphatase inhibitors (ProteinSimple) followed by immunoprecipitation for VEGFR2 and immunoblotting or immunoblotting of total lysates. Proteins were separated in NuPAGE Novex 4-12% Bis-Tris gels (Invitrogen). Further immunoblotting protocol details have been described previously (7, 36).


The retina endothelial cell gene expression values were analysed on the basis of the RNA sequencing data from a previously published study (32). There were 252 retina vascular endothelial cells reported by the study. The original sequence counts data were downloaded from the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) database (accession number GSE63472). The gene sequence counts for Tsad (Sh2d2a), Vegfr2 (Kdr), Cdh5 (VE-cadherin), and c-Src from each of the 252 vascular endothelial cell were extracted and plotted in the bar graph in the four panels, respectively. The endothelial expression profiles in different tissues were analyzed on the basis of the microarray data set from a previously published study (33). The array raw data were downloaded from the NCBI GEO database (accession number GSE47067) and normalized using the PLIER algorithm (Affymetrix Technical Note. Guide to Probe Logarithmic Intensity Error Estimation. The average expression values and SEs were calculated from the three replicated samples in each group.


Animals were propagated at the local animal facility under laminar airflow conditions with a 12-hour light/dark cycle at a temperature of 22° to 25°C. All animal work was approved by the Uppsala University board of animal experimentation. C57BL/6 wild-type mice were purchased from B&K Scanbur. Tsad−/− mice (18) in the C57BL/6 background (backcrossed 10 generations) were provided by J. A. Bluestone (University of Chicago). mTmG mice were obtained by from the Jackson Laboratory (28). Cdh5-CreERT2 mice were provided by R. Adams [Max Planck Institute (MPI), Münster] (26, 27).

To generate Tsad floxed mice, targeted ES cells were generate by injecting linearized and extracted vector plasmid (100 μg), which was electroporated into C57BL/6N ES cells. Homologous recombinants were selected by G418 and ganciclovir. ES clones were selected for PCR and Southern blot screening. The following primers were used for amplification and screening: 5′ end, GCCGAGTGTGCACCTCCTTTGAGAG and CACAACGGGTTCTTCTGTTAGTCC; 3′ end, ATGGTCTGAGCTCGCCATCAGTTC and CGTTCGCATCAACCTGTGGTTTGGAG. One clone positive for both 5′ and 3′ was selected for blastocyst injection and chimera generation. Chimeric mice were crossed with C57BL/6 mice and heterozygote mice from this crossing constituted F1, genotypes identified by PCR [forward, TTGACTTCCCTGCCGTGACAT; reverse, GTGGCTGGGTTTTGGTTTTAG; product, 533 versus 441 base pairs (bp) for wild-type]. Mice (F1) were delivered from the NiceMice, National Resource Center for Mutant Mice, Model Animal Research Center, China. The F1 mice were further backcrossed with C57BL/6N in the host facility and subsequently crossed with FLP deleter mice [B6.Cg-Tg(ACTFLPe)9205Dym/J; Jackson Laboratory]. Heterozygous FLP-deleted individuals were then bred to constitute the colony. The following were the primers used to identify successful excision of the LacZ-neomycin cassette: forward, AGGGTGGACACGGAAGAAAGA; reverse, GCCTTCAAAGCAATCGGTCTG; weight, 286 bp; FLP-deleted, 513 bp).

Inducible gene deletion

Cre activity and gene deletion were induced by intraperitoneal injections to pups with 100 μg of tamoxifen (Sigma, T5648) at P1 and P2, and mice were sacrificed at P5. Tamoxifen (30 μg) was injected at P1 to induce mosaic deletion, and mice were sacrificed at P5. Cre activity was induced in adult mice by intraperitoneal injections of tamoxifen (2 mg/day for five consecutive days) in 6-week-old mice.


Eyes were removed and prefixed in 4% paraformaldehyde (PFA) for 20 min at room temperature. Tracheas were fixed in 4% PFA for 15 min at room temperature. After dissection, tissues were blocked overnight at 4°C in 1% FBS (Gibco), 3% BSA (Sigma), and 0.5% Triton X-100 (Sigma). Samples were incubated overnight with primary antibodies in blocking reagent, followed by washing and incubation with the appropriate secondary antibody for 2 hours at room temperature, and mounted in fluorescent mounting medium (DAKO). Cells were treated with mouse VEGFA164 (80 ng/ml) for 10 min before permeabilization with 3% PFA and 0.5% Triton X-100 in PBS for 3 min, followed by fixation in 3% PFA in PBS for 15 min. Antibodies were added in 5% BSA, 5% donkey serum in PBS. Images were acquired using a Zeiss LSM700 confocal microscope. For comparison purposes, different sample images of the same antigen were acquired under constant acquisition settings.

VE-cadherin patching image analysis

VE-cadherin junctional patterning was assessed using an unbiased software “patching” method (Matlab), as previously described and validated (15). 3D retinal and tracheal image stacks were processed for VE-cadherin morphology, lying within the vascular mask, into “patches” of 100 × 100 pixels on each slice image. Images were acquired with Zeiss LSM700 confocal microscope, ×63 numerical aperture, and 1.4 objectives. The VE-cadherin morphology in each patch was hand-classified on a scale from 1 being active (irregular/serrated morphology with diffuse/vesicular regions, red in output images) to 6 being “inactive” (straighter morphology with less vesicular staining, blue in heat map images).

Matrigel assay

Matrigel plugs contained dog VEGF164 (2 μg per plug) with S1P (1 μM; Avanti Polar Lipids, 860492P) or mouse FGF2 (0.4 μg per plug) with heparin (30 μg per plug; Sigma). DMEM with fatty acid–free BSA (0.5 mg/ml; Sigma) with growth factor was mixed 1:1 with high-concentration Matrigel (BD Bioscience, 354248) on ice. Six- to nine-week-old mice were injected with 300 μl of Matrigel–growth factor solution under the skin of the abdomen. After 7 days, mice were sacrificed and vasculature-fixed with 5 ml of 4% PFA by cardiac perfusion. Matrigel plugs were removed and fixed for an additional 1 hour in 4% PFA. After washing, samples were blocked for 3 hours in PBS with 3% BSA and 0.2% Triton X-100. Samples were incubated overnight with primary antibodies in blocking reagent at 4°C, followed by washing and incubation with secondary antibodies overnight at 4°C. Endothelial cells were visualized with CD31, and Cre recombination efficiency in Tsadfl/fl;mTmG;Cdh5-CreERT2 mice was determined by staining with GFP. After washing, samples were kept in PBS before imaging using a Zeiss LSM700 confocal microscope. Quantification was performed using ImageJ.

In vivo VEGF induction

Six- to eight-week-old mice were injected in the tail vein with PBS or dog VEGFA164 (5 μg) followed by circulation for 1 min before sacrifice and isolation of lungs. Lungs were stored at −80°C until protein extraction.

Isolation of lung endothelial cells

Endothelial cells were isolated as described previously (36). Mouse lungs were collected at P10, minced and digested in 10 ml of Dulbecco’s PBS medium containing 1 collagenase type I (2 mg/ml; Sigma), for 1 hour at 37°C with shaking, followed by filtration through a 70-mm disposable cell strainer (BD Falcon). Cells were centrifuged at 400g for 8 min at 4°C, suspended in cold PBS with 0.1% BSA, and incubated with anti-rat immunoglobulin G–coated magnetic beads (Dynabeads sheep anti-rat IgG, Invitrogen) precoupled with rat anti-mouse CD31 for 10 min, with gentle agitation. Beads were separated using a magnetic particle concentrator (Dynal MPC-S, Invitrogen). The beads were washed with PBS, and endothelial cells were suspended in EMV2 growth medium with the following supplements: 1 endothelial growth factor (5 ng/ml), hydrocortisone (0.2 mg/ml), VEGFA (0.5 ng/ml), basic fibroblast growth factor (10 ng/ml), insulin-like growth factor-1 (20 ng/ml; Promocell), and 5% penicillin/streptomycin (Sigma). The isolated lung endothelial cells were cultured on slides precoated with gelatin.


Fig. S1. TSAd is not required for developmental retinal angiogenesis.

Fig. S2. Generation of TSAd conditional knockout mice.

Fig. S3. TSAd-deficient tracheas have normal lymphatic vessel morphology.

Fig. S4. Expression profiles of Tsad and associated genes.

Fig. S5. TSAd-deficient cells express Dll4.

Fig. S6. The SH2 and C-terminal domains of TSAd are necessary for sprouting.


Acknowledgments: We are grateful to M. Hedlund for assistance with animal housing and Matrigel experiments, C. Wikner for assistance with mouse genotyping, and L. Koll for technical assistance. We thank I. Lonnstedt for assistance with statistics. We thank the SciLifeLab imaging facility, BioVis, Uppsala University for assistance with image acquisition and the Uppsala University Transgenic Facility for ES cell establishment. We thank R. Adams (MPI, Münster) for providing the Cdh5/CreERT2 mice. We thank J. A. Bluestone (University of Chicago, Chicago, IL) for providing the Tsad−/− mice. We thank AstraZeneca for providing the Src inhibitor Saracatinib (AZD0530). We thank V. Shapiro for the rabbit anti-TSAd antisera. We thank K. Ballmer-Hofer (Paul Scherrer Institute) for dog VEGFA164 and recombinant VEGFR2 KID. Funding: E.J.G. is supported by a Wenner-Gren Postdoctoral Fellowship. D.F. and N.P. are supported by the Gustav Adolf Johansson Foundation. L.C.-W. is supported by the Swedish Cancer Foundation (CAN 2013/661), the Worldwide Cancer Research (13-1295), and the Swedish Science Council (2015-03275). E.D. is supported by the Italian Association for Cancer Research (14471) and the European Research Council (268870). K.B. is supported by the Beth Israel Deaconess Medical Center. A.S. is supported by the Norwegian Research Council (214202) and the Norwegian Cancer Society (17561). L.C.-W., E.D., and K.B. are supported by the Knut and Alice Wallenberg Foundation project grant “Towards control of formation and resolution of edema by deciphering mechanisms of vascular leak and lymphatic function.” Author contributions: E.J.G., D.F., S.W., E.O.S., and L.C.-W. conceived and designed the study. E.J.G. and L.C.-W. wrote the paper. E.J.G., D.F., S.W., N.P., E.O.S., L.v.M., and L.H. performed the experiments. F.O., E.D., K.B., and A.S. provided essential experimental input and generated essential reagents. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article