PTEN mediates Notch-dependent stalk cell arrest in angiogenesis

Coordinated activity of VEGF and Notch signals guides the endothelial cell (EC) specification into tip and stalk cells during angiogenesis. Notch activation in stalk cells leads to proliferation arrest via an unknown mechanism. By using gain- and loss-of-function gene-targeting approaches, here we show that PTEN is crucial for blocking stalk cell proliferation downstream of Notch, and this is critical for mouse vessel development. Endothelial deletion of PTEN results in vascular hyperplasia due to a failure to mediate Notch-induced proliferation arrest. Conversely, overexpression of PTEN reduces vascular density and abrogates the increase in EC proliferation induced by Notch blockade. PTEN is a lipid/protein phosphatase that also has nuclear phosphatase-independent functions. We show that both the catalytic and non-catalytic APC/C-Fzr1/Cdh1-mediated activities of PTEN are required for stalk cells' proliferative arrest. These findings define a Notch–PTEN signalling axis as an orchestrator of vessel density and implicate the PTEN-APC/C-Fzr1/Cdh1 hub in angiogenesis.

V essel sprouting is a central mechanism of blood vessel growth 1,2 and it relies on the induction of specialized endothelial cell (EC) populations, each accounting for distinct functions. At the very front of the sprouts, tip cells provide guidance and migrate towards gradients of vascular endothelial growth factor (VEGF)-A, but rarely proliferate [2][3][4] . Instead, trailing stalk cells located at the base of the sprout proliferate, establish adherent and tight junctions and form the vascular lumen 1,2,5 .
The tip cell phenotype is usually associated with high levels of Delta-like 4 (Dll4), which activate Notch in neighbouring stalk cells, preventing them from becoming a new tip cell. Notch signalling is initiated by receptor-ligand recognition between adjacent cells. This interaction results in two sequential proteolytic events that release the Notch intracellular domain (NICD). Subsequently, NICD translocates to the nucleus, where it forms a complex with the transcriptional factor Rbpj/Cbf1 and the Mastermind-like proteins to drive target gene expression 6,7 . Activation of Notch in ECs leads to cell cycle arrest both in vitro 8 and in vivo [9][10][11] . However, it is still unclear how Notch exerts its negative effects on EC proliferation, and the transcriptional programme that triggers stalk cell function is not understood 2,5,12 . Furthermore, it is not clear how stalk cells are ultimately released from this arrest to provide sufficient cell numbers for the sprout to elongate and stabilize.
PTEN (phosphate and tensin homologue deleted on chromosome TEN) is a dual lipid/protein phosphatase, which is often underexpressed in cancer [13][14][15] . The main activity of PTEN is to dephosphorylate the lipid phosphatidylinositol 3,4,5-trisphosphate (PtdIns (3,4,5)P 3 ) at the 3-position, thereby counterbalancing class I phosphoinositide 3-kinase (PI3K) signalling that mediates growth, cell division, survival, migration and metabolism 13,[16][17][18] . Genetic studies in mouse and zebrafish point to a restrictive role of PTEN in angiogenesis. Mice lacking PTEN specifically in ECs exhibit cardiac failure and severe haemorrhages due to defects of the myocardial wall and impaired mural cell coverage of blood vessels 19 . Mutant zebrafish embryos lacking functional PTEN show enhanced angiogenesis 20 ; whether this is due to a cell-autonomous effect of PTEN in ECs or is simply a consequence of increased VEGF levels is unclear. Importantly, the specific functions of PTEN in endothelial behaviour and vascular patterning remain unknown.
In most cells and tissues, PTEN localizes to the cytoplasm and the nucleus 13,15 . There is evidence to suggest that PTEN has nuclear, non-lipid phosphatase-dependent functions 21,22 . Interestingly, PTEN localization is cell cycle-dependent, with higher levels of nuclear PTEN during the G0-G1 phase than during the S phase 23,24 . This is in line with the observation that nuclear PTEN negatively regulates cell cycle progression 13,22 . Indeed, in late mitosis and G1, nuclear PTEN enhances the E3 ligase activity of APC/C by facilitating the association of APC/C with its activator Fzr1/Cdh1 (encoded by the Fzr1 gene), with no requirement of its phosphatase activity 22 . The APC/C-Fzr1/Cdh1 complex controls G1 progression by targeting several proteins for degradation, including mitotic cyclins (Cyclin-A), mitotic kinases (Aurora Kinase A (Aurora A) and Polo-like kinase 1 (Plk1)), proteins involved in chromosome segregation and DNA replication (Geminin; ref. 25). Despite the large body of molecular evidence, the role and relevance of nuclear PTEN in physiology is poorly understood.
Here we report that endothelial PTEN regulates stalk cell proliferation during vessel development. Our data further identify PTEN as a key mediator of the antiproliferative responses of Notch. We show that Dll4/Notch signalling arrests stalk cell proliferation by inducing expression of PTEN to balance stalk cell numbers and coordinate patterning. On PTEN deletion, Notch signalling fails to arrest early stalk cells and result in defective sprout length and patterning. Our results strongly indicate that both catalytic and non-catalytic activities of PTEN contribute to this function, providing evidence for an important in vivo physiological function for the PTEN-APC/C-Fzr1/Cdh1/axis.

PTEN negatively regulates vascular density in angiogenesis.
To study the EC-autonomous role of PTEN in sprouting angiogenesis, we crossed Pten flox/flox mice with PdgfbiCreER T2 transgenic mice that express a tamoxifen-activatable Cre recombinase in ECs 26 (further referred to as PTEN iDEC/iDEC ) and assessed postnatal retinal angiogenesis. 4-hydroxytamoxifen (4-OHT) was administrated in vivo at postnatal day 1 (P1) and P2, followed by analysis of the retinal vasculature at different time points. Comparing whole-mount-stained retinas of control (Pten flox/flox ) to PTEN iDEC/iDEC mice at P5 revealed a mild increase in vessel width (Supplementary Fig. 1a-g). By P7, loss of PTEN resulted in excessive branching and substantially increased vessel width (Fig. 1a-d), a phenotype that was further exacerbated at P10 (Fig. 1h,i). PTEN iDEC/iDEC P7 retinas showed efficient recombination of the Cre-reporter R26-R and depletion of PTEN in the retinal endothelium ( Supplementary Fig. 1h,i), with an increase in staining for phosphoS6 (pS6), a marker of PI3K pathway activation ( Supplementary Fig. 1j). Isolated mouse lung ECs (mECs) from PTEN iDEC/iDEC mice confirmed that effective depletion of PTEN protein in mECs was achieved 96 h following 4-OHT administration ( Supplementary Fig. 1k,l). To further characterize the cell-autonomous role of PTEN in ECs, we therefore focused on the P7 time point. No differences in radial expansion (Fig. 1e) and in the number of sprouts per 100 mm of leading endothelial membrane were found in PTEN iDEC/iDEC when compared with control retinas (Fig. 1f). Instead, the length of the sprouts was significantly reduced in the PTEN iDEC/iDEC retinal vasculature compared with controls (Fig. 1g). The hyperplastic phenotype observed on PTEN loss was validated in an independent cellular system based on embryoid body (EB) formation, in which clusters of embryonic stem cells respond to VEGF by forming vascular tubes 27 . Compared with wild type (WT), PTEN null EBs showed increased sprout width and length (Supplementary Fig. 2a-d), with no differences in the number of sprouts ( Supplementary Fig. 2e).
Next, we sought to address whether regulated elevation in PTEN expression in vivo would oppose the phenotype induced by loss of PTEN. To this end, we used super-PTEN transgenic mice (PTEN TG ) 28 , a mouse model that allows moderate organismal elevation of PTEN levels (two-fold over WT littermates), including in the vasculature ( Supplementary Fig. 1m). PTEN TG retinas exhibited decreased vessel width and increased sprout length (Fig. 1j,k,m,p), with no changes in the number of branches (Fig. 1l) and sprouts (Fig. 1o). A slight reduction in radial expansion was also observed on moderate PTEN overexpression (Fig. 1n), similar to retinas from mice that are heterozygous for a kinase-dead p110a PI3K allele 29 . Neither loss nor gain of PTEN function resulted in changes in Dll4 or Notch target genes ( Supplementary Fig. 3a-c), further supporting that PTEN is not required for tip/stalk selection. Analysis of EphB4, EphrinB2 and Nr2f2 gene expression, key genes involved in arteriovenous differentiation, did also not reveal any obvious difference between control and loss and gain of PTEN function in ECs, suggesting that PTEN signalling does not play a predominant role in this process ( Supplementary Fig. 3d,e).
Taken together, these data uncover a selective role for PTEN in angiogenesis, regulating vascular density and consequently vessel growth in vivo.
PTEN regulates endothelial stalk cell number. Previous data have shown that constitutive targeting of PTEN in ECs results in altered mural cell coverage 19 . Instead, immunostaining with desmin, a retinal pericyte marker 30 , did not reveal any obvious defect in mural cell coverage in PTEN iDEC/iDEC retinas compared with control, consistent with the lack of sprouting defects on PTEN loss (Supplementary Fig. 3f).
Analysis of PTEN iDEC/iDEC retinas stained with a nuclear endothelial marker (Erg) revealed increased EC numbers in the angiogenic vasculature (Fig. 2a,b). Conversely, elevated PTEN expression resulted in reduced EC numbers at the sprouting front (Fig. 2c,d). We sought to validate whether these differences relate to changes in EC proliferation. Surprisingly, no difference in the number of proliferative ECs located in the subfront retinal area,     behind the sprouting front, was found on either loss (Fig. 2e,f) or gain of PTEN function (Fig. 2h,i). This is unexpected, given that in the growing vasculature ECs with high turnover are located in this subfront area ( Supplementary Fig. 4a,b). To test whether PTEN regulated proliferation of ECs in other retinal locations, we focused on the first line of cells located at the sprouting front where proliferating cells are rarely observed ( Supplementary  Fig. 4a,b). Interestingly, a 40% increase in proliferation in PTEN iDEC/iDEC retinas (Fig. 2g) or 60% reduction in PTEN TG retinas (Fig. 2j) compared with control retinas was observed in ECs at the front. These data point towards a selective role of PTEN in restricting EC proliferation in cells located at the sprouting front.
PTEN executes Notch-dependent cell cycle arrest. Given that the impact of PTEN loss or overexpression in vivo on proliferation are restricted to the sprouting front that is highly Notch-dependent 9 , we hypothesized that a functional connection exists between these two signalling pathways. Our results suggest that PTEN is necessary for the growthsuppressive activity of Notch signalling in ECs and imply that the PTEN loss-of-function phenotype is the result of an impaired response to Dll4 stimulation. Because PTEN is required for the Notch-dependent regulation of endothelial proliferation, we tested whether PTEN expression is regulated by Notch. Bioinformatic analysis of the PTEN locus identified the presence of three Rbpj motifs that are conserved in both the human and mouse PTEN gene (Fig. 3g). We used chromatin immunoprecipitation (ChiP) analysis on human ECs to determine the recruitment of NICD protein to the PTEN promoter after 2 h of incubation with Dll4. PTEN promoter occupancy was determined using real-time quantitative PCR (qPCR) probes that amplify seven regions spanning from À 2,380 to À 590 relative to the transcription initiation site. Our analysis revealed NICD occupancy in the À 1,492 region, which contains one of the three predicted Rbpj-binding sites (Fig. 3g). We next validated the functional significance of Rbpj binding to the promoter in luciferase reporter assays. Indeed, analysis of PTEN promoter activity showed activation in response to Dll4 (Fig. 3h) and on overexpression of the intracellular domain of the Notch receptor (NICD; Fig. 3i). Importantly, enhanced PTEN promoter responsiveness to Dll4 was abrogated by the g-secretase inhibitor dibenzazepine (DBZ; Fig. 3i). Western blot experiments confirmed that Dll4 stimulation results in elevated protein levels of PTEN in human ECs and mECs (Fig. 3j,k). Using lungs as highly vascularized tissue, we validated that overactivation of Notch signalling by inhibiting the Notch ligand, Jagged 1 (Jag1) 11 , resulted in higher PTEN expression levels (Fig. 3l). Taken together, these results demonstrate that PTEN is a target gene of Notch signalling in ECs, which becomes induced by Dll4 stimulation.
PTEN is required for Notch function in vivo. We sought to confirm the Notch/PTEN functional interaction in vivo. We took advantage of endothelial Jag1 inactivation, which results in reduced EC proliferation and decreased vascular branching due to overactivation of Notch signalling 11 . We hypothesized that, if the regulation of the angiogenic process by Notch requires the increase in PTEN function, loss of PTEN would prevent the phenotype of Jag1 deletion. We tested this hypothesis in inducible endothelial-cell-specific PTEN iDEC/iDEC ;Jag1 iDEC/iDEC double mutants. The vasculature of Jag1 iDEC/iDEC retinas showed reduced endothelial proliferation and reduced vessel width, confirming previous reports ( Fig. 4a-g and Supplementary  Fig. 5a) 11 . However, concomitant PTEN deletion abrogated the phenotype observed in Jag1 iDEC/iDEC retinas ( Fig. 4a-g), while the phenotype of PTEN loss remained unaffected by Jag1 deletion (the increase in endothelial proliferation at the spouting front). Next, we validated our hypothesis in the PTEN TG mice by blocking Notch activation with the g-secretase inhibitor DAPT (N-[N-(3,5-difluorophenacetly)-L-alanyl]-S-phenylglycine t-butyl ester) that strongly enhances angiogenesis partially due to increased EC proliferation 9 . Remarkably, DAPT-induced increase in vascular density at the angiogenic front was abolished by increased levels of PTEN ( Fig. 4h,j,m,n and Supplementary Fig. 5b). However, elevated levels of PTEN did not prevent the enhanced sprouting caused by DAPT ( Fig. 4i,k,l), further indicating that PTEN is not required for Notchdependent tip/stalk selection.

Dual function of PTEN in angiogenesis.
To gain insight into the biological mechanism underlying the role of PTEN in sprouting angiogenesis, we investigated the contribution of phosphatasedependent and -independent activities of PTEN at the organismal and cellular levels. We observed a compartmentalization of PTEN in the nucleus and cytoplasm in cultured control cells (Fig. 5a), whereas PTEN iDEC/iDEC mECs showed no nuclear staining with some residual positivity in the cytoplasm. PTEN depletion in mECs resulted in increased Akt phosphorylation and in accumulation of the E3 ubiquitin ligase APC/C-Fzr1/Cdh1 complex substrates Aurora A, Plk1 and Geminin (Fig. 5b). Control mECs treated with 4-OHT did not show any of the aforementioned changes (Fig. 5c). In contrast, mECs isolated from PTEN TG lungs showed reduced Akt phosphorylation and reduced accumulation of E3 ubiquitin ligase APC/C-Fzr1/Cdh1 complex substrates compared with WT cells (Fig. 5d).
To investigate whether increased levels of the APC/C-Fzr1/ Cdh1 targets were only a consequence of increased PI3K activity, we tested the impact of GDC-0941 (a pan-class I PI3K inhibitor that blocks p110a, p110b, p110d and p110g; ref. 31). While pretreatment with GDC-0941 abrogated Akt activation in PTEN iDEC/iDEC mECs (Fig. 5e), it did not modify the levels of Aurora A and Geminin (Fig. 5e), further corroborating that the function of PTEN promoting the APC/C-Fzr1/Cdh1 activity is independent of its ability to inhibit PI3K signalling through its lipid phosphatase activity 22 .
As our data indicate that the principal function of PTEN in sprouting angiogenesis is to regulate EC proliferation, we tested to what extent phosphatase-dependent and -independent activities of PTEN participate in this regulation. In vitro isolated PTEN iDEC/iDEC and PTEN TG mECs showed increased and decreased BrdU incorporation, respectively ( Fig. 5f and Supplementary Fig. 6a,b). Inhibition of PI3K activity and Aurora kinase, one of the main targets of APC/C-Fzr1/Cdh1 (ref. 22), partially rescued normal proliferation rate in PTEN iDEC/iDEC mECs ( Fig. 5f and Supplementary Fig. 6a). A synergistic effect was observed on pretreatment with both GDC-0941 and VX680, further implying a dual function of PTEN in this process. Next, we complemented PTEN-depleted ECs with either WT    . 5g and Supplementary  Fig. 6c) of PTEN null ECs, albeit most prominently seen with PTEN WT . All together, these data indicate that phosphatasedependent and -independent activities of PTEN are important to regulate EC proliferation.
Next, we tested the differential contribution of each of these functions in vivo, by first analysing the retinas of PTEN iDEC/iDEC on inhibition of class I PI3K activity with GDC-0941. The hyperplasia induced by PTEN loss was rescued by inhibiting PtdIns(3,4,5)P 3 production (Fig. 6a-d and Supplementary  Fig. 7a) ARTICLE sprouting angiogenesis by inhibiting Aurora kinase. Strikingly, the phenotype of PTEN iDEC/iDEC was abrogated by Aurora kinase inhibition (Fig. 6e-h), which is consistent with the contribution of the phosphatase-independent activity of PTEN to sprouting angiogenesis in vivo. This was further corroborated by genetic conditional and inducible deletion of Fzr1 in ECs. Complete depletion of Fzr1/Cdh1 protein was achieved 48 h post incubation with 4-OHT ( Supplementary Fig. 7b,c). Therefore, pups were treated with 4-OHT at P5 and P6, followed by analysis of the retinal vasculature at P7. The retinas of Fzr1 iDEC/iDEC showed increased vascular density and reduced length of sprouts ( Fig. 6i-n) comparable to PTEN iDEC/iDEC . All together, these data suggest that, in ECs, there is a dual contribution of PTEN, counterbalancing the PI3K signalling pathway through its lipid phosphatase activity and facilitating the APC/C-Fzr1/Cdh1 complex activity, which has been reported to be independent of its catalytic function 22 .

Discussion
Here we report that PTEN in ECs is required in a cellautonomous and dose-dependent manner for the control of vascular density and vessel growth, but is dispensable for the regulation of the sprouting activity of tip cells. We show that endothelial PTEN restricts vascular growth by limiting stalk cell proliferation during sprouting angiogenesis. An important conclusion from this study is that PTEN is not required in all ECs to regulate proliferation in vivo, as shown in cultured ECs (our data and ref. 19). Instead, our results indicate that the consequence of PTEN loss or overexpression in vivo is restricted specifically to the zone of the vasculature that is highly Notch-dependent. Constitutive targeting of PTEN in ECs leads to embryonic lethality due to aberrant angiogenesis 19 . Although these studies have established that PTEN regulates EC proliferation, the analysis of vasculature in Tie2Cre-PTEN flox/flox embryos did  Activation of Notch leads to cell cycle arrest [8][9][10][11]34 . In this study, we identify PTEN as a critical mediator of Notch antiproliferative response in stalk cells. If PTEN is not expressed in ECs, stalk cells become insensitive to the antiproliferative signals of Notch and exhibit unrestricted expansion, hence perturbing sprout length and pattern and eventually resulting in profound hyperplasia. Interestingly, our results also reveal that stalk cells located further away from the a  ARTICLE front are insensitive to changes in PTEN expression. Given that these stalk cells at the subfront area are highly proliferative, our data support the existence of two biological states for stalk cells. A first state in which stalk cells must remain arrested to ensure the correct patterning of the sprout, and a second state in which cells enter the cell cycle to expand the plexus. These two states are likely the consequence of dynamic changes in Notch signalling, with high Notch activity in early nonproliferative stalk cells and low Notch activity in the late proliferative stalk cells. Our data predict a rise and fall in PTEN levels that will accompany the early quiescent and late proliferative phases, respectively. This is supported by the observation that, in WT cultured ECs, higher PTEN levels are seen 8 h post stimulation with Dll4 compared with 24 h post stimulation. Furthermore, co-staining of PTEN and 5-ethynyl-2 0 -deoxyuridine (Edu) in the growing vasculature showed that Edu-negative cells express higher levels of PTEN than Edu-positive cells, supporting the notion that PTEN protein levels rise to guarantee cell cycle arrest. Whether high PTEN cells correspond to high Notch signalling still needs to be determined. Moya et al. speculated that early and late stalk cell behaviours might be orchestrated by oscillation in Notch activity. The authors proposed that Id proteins, members of HLH proteins, govern these two states by releasing the negative autoregulatory loop of Hes1 (ref. 35). While our results are consistent with the idea of two states, they identify PTEN as the key mediator of early stalk cell function in response to Notch. Why and how Notch exerts a unique negative regulation in the endothelium while driving proliferation in virtually every other cell type and in cancer has been a mystery 5,36 . Our data show that PTEN negatively regulates cell cycle progression in ECs through conserved pathways. Critically, what our results illustrate is a novel interaction between Notch and PTEN in ECs. We find that Notch stimulates PTEN transcription in the endothelium, an effect that is required for Notch-mediated cell cycle arrest. Interestingly, in cell types where Notch stimulates cell cycle progression, PTEN is transcriptionally repressed by Notch/Hes [36][37][38] . The PTEN gene locus contains both Rbpj-and Hes-binding sites, suggesting that binding to one or another is what determines the final biological output.
In line with the observation that PTEN restricts stalk cell proliferation, endothelial gain and loss of PTEN proliferation phenotypes are reminiscent of gain and loss of Notch function in stalk cells [9][10][11] . However, in response to Notch signalling PTEN appears to only regulate EC proliferation while it is not required for tip and stalk specification. This is shown by the observation that Notch mutants not only show aberrant proliferation phenotypes in the nascent plexus but also sprouting defects 9,11 , while PTEN mutants only show vascular density defects. In the same line, increased levels of PTEN protect angiogenic ECs treated with DAPT from uncontrolled proliferation but fail to prevent excessive tip cell numbers. Conversely, a recent study has shown that inhibitors of the VEGFR3 kinase activity rescue the hypersprouting phenotype of Notch loss-of-function mutants, without reducing EC proliferation 39 . Taken together, these data suggest that Notch regulates tip cell numbers and stalk cell proliferation independently through VEGFR3 and PTEN pathways, respectively.
The predominant activity of PTEN is the dephosphorylation of PtdIns(3,4,5)P 3 and thus the counteraction of class I PI3K-mediated functions 13,15 . However, PTEN also exhibits PtdIns(3,4,5)P 3 -independent functions, including protein phosphatase 14 and non-catalytic activities 13,15 . In this context, PTEN can be found in the nucleus where it regulates DNA stability and cell cycle progression 22,40 . Several reports have highlighted the relevance of nuclear PTEN in disease 22,[41][42][43] . To date, the physiological relevance of nuclear PTEN in vivo remains elusive. Our results reveal that both lipid phosphatase-dependent and non-catalytic activities of PTEN regulate stalk cell proliferation during sprouting angiogenesis. Inhibition of class I PI3K activity with GDC-0941 or Aurora kinase with VX680 significantly abrogates the phenotype observed on PTEN loss. However, the observation that pretreatment with either GDC-0941 or VX680 is not able to completely rescue the hyperplasia phenotype of PTEN iDEC/iDEC retinas and cultured ECs indicates that both types of activities of PTEN are required to drive the PTEN response in angiogenesis. Furthermore, genetic deletion of Fzr1 in ECs recapitulates the phenotype observed on endothelial loss of PTEN, reinforcing the relevance of nuclear PTEN facilitating the APC/C-Fzr1/Cdh1 function. Taken together, our study provides in vivo evidence that nuclear PTEN is not only involved in disease such as cancer or cerebral ischaemia 22,[41][42][43] but is also critical to regulate a fundamental physiological process such as angiogenesis. We and others have previously shown that inhibition of class I PI3K isoform in vivo does not lead to blockade of EC proliferation 29,[44][45][46] . Although contradictory, these observations may reflect that PTEN principally regulates EC proliferation independently of its lipid phosphatase activity 22 . In line with this, our data also reveal that the regulation of APC/C-Fzr1/Cdh1 by PTEN seems to play a major role in response to Notch signalling in angiogenesis. This is shown by the altered APC/C-Fzr1/Cdh1 target expression under conditions of PTEN loss and Notch activation. This observation, together with the fact that Notch stimulation in ECs results in phosphorylation of Akt 47,48 , suggest that Notch stimulates PTEN nuclear translocation. These findings would be in agreement with the notion that higher nuclear PTEN levels are found during G0-G1 phase than during the S phase 23,24 . Further experiments are needed to elucidate how PTEN accumulates in the nucleus on Notch activation.

PTEN WT
The unique direction of the coupling of Notch and PTEN in the endothelium (Fig. 7), and the highly selective effects on the active vascular front raise the prospect that targeting this interaction and stimulation of PTEN signalling may be used therapeutically to render EC quiescence in aberrant tumour angiogenesis and in turn promote a normalization effect. Clinically, our results imply that stimulating both arms of PTEN function in ECs could render a more quiescence phenotype of highly proliferative tumour ECs 2,49 . However, inhibition of PI3K in the tumour stroma not only results in reduced EC proliferation but also in reduced vascular function 47 . It is thus tempting to speculate that promoting nuclear PTEN may offer more selectivity towards a tight control of EC proliferation. (ref. 26). To generate PdgfbiCreER T2 ; PTEN flox/flox ; Jag1 flox/flox (PTEN iDEC/iDEC ; Jag1 iDEC/iDEC ), PdgfbiCreER T2 ; PTEN flox/flox (PTEN iDEC/iDEC ) and PdgfbiCreER T2 ; Jag1 flox/flox (Jag1 iDEC/iDEC ) littermates, two different types of breeding were set up; PdgfbiCreER T2 ; PTEN flox/flox ; Jag1 flox/flox were interbred with PTEN flox/flox ; Jag1 flox/WT or with PTEN flox/WT ; Jag1 flox/flox . Cre activity and gene deletion were induced by intraperitoneal injection of 25 mg 4-OHT (10 mg ml À 1 ) in all pups of the litter, at P1 and P3 and retinas were collected at P7. To assess proliferating ECs, the pups were injected intraperitoneally with 60 ml of Edu (0.5 mg À 1 ml À 1 ) 2 h before being killed. Edu was dissolved in a 1:1 ratio DMSO:PBS.
Immunofluorescence. Eyes were fixed in 4% paraformaldehye (PFA) for 2 h at 4°C. For PTEN staining, mice were exsanguinated by transcardiac perfusion of PBS, followed by perfusion with 4% PFA before dissecting and continuing fixing the retinas with methanol at À 20°C. Samples were rehydrated for 30 min at room temperature (RT). After washing twice in PBS, retinas were permeabilized in PBS containing 1% bovine serum albumin (BSA) and 0.3% Triton X-100 overnight (ON) at 4°C, followed by incubation with primary antibodies (PTEN (Abcam;  1 Embryoid bodies. ES cells were cultured and EBs were generated, as previously described 27 . Briefly, ES cells were regularly cultured on a layer of irradiated DR4 mouse embryonic fibroblast in DMEM glutamax (Life Technologies, #61965-026) in the presence of 20% fetal bovine serum, HEPES (30 mM), sodium pyruvate (1.5 mM), monothioglycerol (1.5%) and leukaemia inhibitory factor (Chemicon#ESG1107, 123 units ml À 1 ). For vascular sprouting assays, cells were cultured for two passages without feeders, depleted of leukaemia inhibitory factor and left in suspension as hanging drops. Four days after, the formed EBs were transferred to a polymerized collagen I gel with the addition of 60 ng ml À 1 VEGF (Peprotech). The medium was changed on day 6 and every day thereafter. Overall, 70,000 WT ES cells and 10,000 PTEN À / À ES cells were plated to generate EB. PTEN À / À ES cells were provided in ref. 53.
Isolation and stimulation of mECs. Mouse lungs were digested with Dispase (Life Technologies, #17105-041; 4 units ml À 1 ) for 1 h at 37°C, followed by positive selection with antimouse vascular endothelial-cadherin (Pharmingen, #555289) antibody coated with magnetic beads (Dynal Biotech, #110-35). Cells were seeded on a 12-well plate, and were coated with gelatin (0.5%) in DMEM/F12 supplemented with 20% fetal calf serum and EC growth factor (PromoCell, #C30140). After the first passage, the cells were re-purified with vascular endothelial-cadherin antibody-coated magnetic beads. Cells were cultured until passage 6. To induce gene deletion, 4-OHT (5 mM) or vehicle (ethanol) was added to the cultured medium at P4 for 96 h and the medium was replaced every other day. For Dll4 stimulation, mouse Dll4 (500 mg ml À 1 ) was immobilized by coating culture dishes for 1 h at RT, followed by seeding mECs for 6, 8 or 24 h. Mouse and human Dll4 were used accordingly.
In vitro measurement of mEC cell proliferation. Overall, 10 4 mECs were plated in a 24-well plate for 48 h; 2 h before the termination of the experiment, BrdU (10 mM) was added to the medium. For Ki67 staining, cells were plated for 24 h in Dll4-coated dishes. Cells were fixed in 4% PFA for 10 min at RT, permeabilized for 10 min with TBS-T (25 mM Tris HCl pH 7.4, 150 mM NaCl, 0.5% Triton X-100), blocked with TBS-T containing 2% BSA and incubated with primary antibodies BrdU (1:100) or Ki67 (1:50) at 4°C ON. The following day, cells were washed three times with TBS-T and incubated with Alexa-conjugated secondary antibodies for 2 h at RT. DAPI was added in the final wash. Specimens were mounted in Mowiol. Cells were visualized in a Nikon-80I microscope. For pharmacological inhibition of class I PI3K and Aurora kinase, GDC-0941 (1 mM) and VX680 (0.5 mM) were added, respectively, on plating.
Plasmids and transfections. pRK5-Myc-PTEN, C124S pEGFP-PTEN-wt and pEGFP-PTEN-K13,289E expressing human WT, lipid phosphatase-inactive and nuclear-excluded PTEN mutants, respectively, were provided in ref. 22. All three PTEN mutants were subcloned with an N-terminal yellow fluorescent protein into a modified lentiviral vector TRIPZ. Lentiviral particles were prepared by transfecting HEK293FT cells with the TRIPZ vector of interest and the packaging vectors psPAX, VSV-G and pTAT. Viral particles in the supernatant were concentrated with Lenti-X-concentrator (Clontech). mECs of P2 or P3 from PdgfbiCreER T2 ; PTEN flox/flox were infected with lentivirus expressing WT PTEN, PTEN (C124S) or PTEN (K13,289E) in the presence of viralplus transduction enhancer (Applied Biological Material #G698). For infection, mECs were plated at a density of 4 Â 10 4 per well of 12-well plate and infected with virus from 293FT cells 48 h after transfection. After 48 h post infection, mECs were re-plated and treated with 4-OHT (5 mM) to induce gene deletion for 72 h. Next, 10 4 mECs were plated in 24-well plate for 48 h in the presence of doxycycline (4 mM); 2 h before the termination of the experiment, BrdU (10 mM) was added to the medium. Cells were then fixed in 4% PFA for 10 min at RT, permeabilized for 10 min with TBS-T (25 mM Tris HCl pH 7.4, 150 mM NaCl, 0.5% Triton X-100), blocked with TBS-T containing 2% BSA and incubated with primary antibodies BrdU (1:100) at 4°C ON. The following day, cells were washed three times with TBS-T and incubated with Alexa-conjugated secondary antibodies for 2 h at RT. DAPI was added in the final wash. Specimens were mounted in Mowiol. Cells were visualized in a Nikon-80I microscope.
MTS viability assay. mECs were cultured in 96-well plate (2,000 cells per 100 ml culture medium per well) in the presence of the test compounds (GDC-0941 (1 mM) and VX680 (0.5 mM)) or the respective controls for 48 h, followed by MTS assay (Promega, #G5421).
ChiP assay. To analyse the binding sites for RBPJ located in the PTEN proximal promoter, we used the Genomatix software. For analysis, the gene bank sequence used was NG_007466.2, which contains the promoter sequence AF406618.1. Three putative RBPJ-binding sites located at À 1,914, À 1,492 and À 1,132 positions relative to transcription initiation site 37 were identified. ChiP assay was performed as previously described 55 . Briefly, chromatin was isolated from HUVECs stimulated for 2 h with vehicle or Dll4 (500 ng ml À 1 ). Crosslinked chromatin was sonicated for 10 min, to medium-sized powder particles, at 0.5-min intervals, with a Bioruptor (Diagenode) and precipitated with anti-NCID or control IgG. After crosslinkage reversal, DNA was used as a template for PCR. qPCR was performed with SYBR Green I Master (Roche, #04.887.352.001) in the LIghtCycler480 system. Primers used are described in Supplementary Table 1.