All animals were housed in a UK Home Office guideline-approved facility, and sample isolation from animals was performed by a license holder in accordance with the UK Home Office regulations under the Animals (Scientific Procedures) Act 1986. The strategies for generating VEGF^120/120 mice (deletion of exons 6 and 7), VEGF^164/164 mice (deletion of exon 6), and VEGF^188/188 mice (containing all exons, except for exon 6b) have been described by other groups previously (Carmeliet et al., 1999; Stalmans et al., 2002).

Mouse endothelial cells were isolated from lungs of VEGF^164/164, VEGF^188/188, or littermate wild-type mice at the age of 8–12 weeks old, as previously described (Dong et al., 1997; Tang et al., 2004; Branco-Price et al., 2012). VEGF^120/120 endothelial cells were isolated from lungs of E17.5 embryos due to the occasional early birth from VEGF^120/+ female of 18 dpc with the high rate of perinatal deaths of those pups. WT cells from E17.5 were also collected and characterized for proliferation and gene expression. Recombinant mouse VEGF164 protein (R & D Systems, Minneapolis, MN, USA) was added as the supplement for the VEGF^120/120 endothelial cell culture for at least 7 days after the isolation to rescue the suppressed proliferation. Cells were cultured in endothelial cell growth medium, consisting of low glucose DMEM:F12 with 1% penicillin/streptomycin, 1% nonessential amino acids, 2 mM sodium pyruvate, 20 mM HEPES. Twenty percent fetal bovine serum (FBS) (Gibco) and 75 μg/mL endothelial mitogen (Alfa Aesar, Ward Hill, MA, USA) were added for the complete growth medium.

The expression and localization of VE-cadherin, β-catenin, and Caspase-3/7 were visualized using immunofluorescence and confocal microscopy (Leica CTR 6500). Cells were cultured on the slide chamber coated with 0.1% gelatin (Sigma-Aldrich), grown to confluence, and serum starved overnight, and then put in hypoxia (1% oxygen, for up to 48 h) or kept in normal oxygen tension for the same length of time. Cells were washed and fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.5% Triton-X100 for 5 min, and blocked with 10% goat serum (Sigma-Aldrich), or 10% donkey serum (Sigma-Aldrich), or 3% bovine serum albumin (BSA) (Life Technologies) in phosphate buffered saline (PBS) for 1 h, and incubated with the following primary antibodies; anti-VE-cadherin (1:50, Santa Cruz Biotechnology), anti- β-catenin (1:50, BD Transduction Laboratories). Alexa Fluor 488- or 568-conjugated secondary antibody (Life Technologies), and FITC-conjugated secondary antibodies (Santa Cruz Biotechnologies) were used at 1:100. Caspase-3/7 Green Flow Cytometry Assay kit (Thermo Fisher) and NucRed 647 (Life technologies) was used for Caspase-3/7 staining and counterstaining of nucleus, according to the manufacture's protocol.

Each VEGF isoform cell colony was seeded at density of 2.0 × 10⁴/mL in a 24-well plate with complete culture medium (containing 20% FBS) or reduced serum (containing 2% FBS), or 0.1% BSA containing basal medium. For the analysis of cell proliferation under hypoxia, hypoxic treatment was started 24 h after plating. Data were collected from three wells per each genotype, each isolated from three mice.

To assess cell migration, wound-healing assay was performed as previously described (Yamamoto et al., 2010). The cells were plated on 10 ng/mL fibronectin (Biomedical technologies)-coated 6-well plates. The monolayer of endothelial cells was scratched manually with a plastic pipette tip. After washing with PBS, the wounded monolayers of cells were allowed to heal for 24 h under hypoxia (1% O₂) in the serum-free culture medium. The length of the wounds was measured and expressed as the relative folds to the migration distance of the wild type.

Twenty-four-well, non-tissue culture-treated plates were coated with 10 ng/ml fibronectin (Biomedical technologies) for 2 h at 37°C. 1.0 × 10⁵ cells suspended in 200 μL serum free media were added to each well and incubated at 37°C for 30 min. The supernatant was removed and adherent cells were selected by washing with PBS. Three wells per group were incubated for 5 min with 0.1% crystal violet in 20% methanol and then washed three times with PBS. Crystal violet contained in the adhesive cells was extracted in 50% ethanol and quantified by the absorbance at 590 nm.

Endothelial cells were seeded at the density of 1 × 10⁵ cells into COSTAR transwell (6.5 mm diameter, 0.4 μm pore, Corning, cat no. 3413). Cells were grown for 4 days and then pre-incubated at 21 or 1% O₂ for 3 h. HRP-conjugated streptavidin (1 ×, Abcam) was then added to the top of the upper chamber; after 1 h, 100 μL of medium was collected from the lower chamber, and the level of HRP-streptavidin in the lower chamber was measured by mixing the 100 μL medium with equal volume of 3,3′,5,5′-tetramethylbenzidine (TMB) peroxidase enzyme immunoassay substrate (Abcam), incubated for 30 min before the addition of 50 μL of Stop Solution (Abcam). The absorbance of the solution was measured at 450 nm.

Transendothelial migration of GFP-labeled Lewis Lung Carcinoma cells (LLC^GFP) was analyzed, as previously described (Branco-Price et al., 2012) with some modifications. Endothelial cells were seeded on COSTAR transwell inserts (6.5 mm diameter, 8 μm pore; Corning, cat no. 3422) at cell density of 2.0 × 10⁴ per well, and grown for 5 days. LLC^GFP cells, pre-stained with Calcein-AM dye (BD Bioscience), were seeded on the endothelial monolayer and the inserts were incubated at 37°C for 24 h under hypoxia (1% O₂), with serum-free media on both upper and lower chambers. Migrated cells were counted in five random fields per insert at 100x magnification by using the fluorescent stereomicroscope (Leica M165FC).

In vitro tube formation on Matrigel was analyzed, as previously described (Tang et al., 2004), with some modifications: Growth Factor Reduced Matrigel (BD Biosciences) was applied at 60 μL/well in 96-well plates and incubated at 37°C for 30 min to allow hardening. 6.0 × 10³ primary lung endothelial cells medium containing 0.5% serum, were seeded on top of the Matrigel. Plates were incubated under normoxia or hypoxia (1% O₂) at 37°C for 9 h. Cells were stained with Calcein AM dye (BD Bioscience) at the end of the incubation, and parameters of detected networks were analyzed using Image J software (Angiogenesis Analyzer, created by Gilles Carpentier).

Culture medium collected from the primary endothelial cells at the time point of 48 h under hypoxia at 1% oxygen or under normoxia were analyzed using an NOA280i (Siever, GE Healthcare) according to the manufacturer's instructions. Readings were performed a minimum of three times for each of three wells.

Extracellular matrix was prepared from a culture dish, as previously described (Yamamoto et al., 2009) with the following modifications: Total cell lysates in 100 mm dishes had been collected in 500 μL RIPA buffer [10 mM Tris/HCl pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.5% Sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 1% TritonX-100] supplemented with protease inhibitors (Roche). Extracellular matrix remaining on the dish was extracted at 100°C for 5 min in 375 μL of LDS Sample buffer (1x, Invitrogen) after washing with PBS and RIPA buffer three times.

Total RNA was extracted using RNeasy Mini Kit (QIAGEN) and converted into cDNA from 0.5 μg or 1 μg of total RNA using Superscript III (Invitrogen) according to the manufacturer's protocol. cDNA was amplified in SYBR Green with an ABI Prism system (Applied Biosystems). Forward and reverse primers were as follows: integrin alpha V 5′-AGGCTGGAACT CAACTGCTC-3′, 5′-TTGGCCCGTC AATGTCGTAA-3′; integrin β 3 5′-GCCTTCGTG GACAAGCCTGT-3′, 5′-GGACAATGC CTGCCAGCCTT-3′; β-actin 5′-CCCAGAGCA AGAGAGG-3′, 5′-GTCCAGACGCAG GATG-3′. Results were normalized to β-actin mRNA levels.

1400 W and LY294002 were purchased from Sigma-Aldrich and Cell Signaling Technology, respectively. VEGF Mouse Elisa kit (Abcam, cat no. ab100751) was used for the quantitative analysis of VEGF in culture medium. Anti-VEGFR2 (D5B1, Cell Signaling Technology, 9698) antibody and Protein A/G agarose (Santa Cruz Biotechnology, sc-2003) were used for VEGFR2 immunoprecipitation. The detail of primary antibodies used for western blot or immunofluorescence analyses are as following; VEGF (P-20, Santa Cruz Biotechnologies, sc-1836), VEGFR2 (D5B1, Cell Signaling Technology, 9698), VE-cadherin (Santa Cruz Biotechnology, sc-6456), β-actin (A5316, Sigma-Aldrich), α-phosphotyrosine (4G10, Millipore, 05-1050), phospho-AKT (Ser 473, Cell Signaling Technology, 9271), phospho-AKT (Thr 308, Cell Signaling Technology, 13038), β-catenin (BD Transduction Laboratories, 610153), PECAM-1 (M-20, Santa Cruz Biotechnology, sc-1506), ICAM-1 (M-19, Santa Cruz Biotechnology, sc-1511), VCAM-1 (Abcam, ab174279).

Each experiment was performed isolating from at least two different animals per group and three technical replicates. The statistical significance was assessed by Student's t-test and expressed as average ± SEM. P < 0.05 was accepted as significant.

To investigate the differential roles of endogenous VEGF isoforms on endothelial cell (EC) viability under hypoxic stress, we purified ECs from mice expressing exclusively one of the isoforms. We started by monitoring the proliferation rates of cells expressing single isoforms, and our first observation was that EC expressing single VEGF120 (VEGF120 EC) displayed severely suppressed proliferation in all tested conditions (Figure 1A), for which reason it was not included in these panels. Characterization of growth was therefore carried out solely on VEGF164 EC and VEGF 188 EC, and cell proliferation was quantified over a range of serum concentrations. We see that autocrine expression of VEGF164 conferred a significant proliferative advantage; the maximal difference between VEGF164 EC and other cells was noted at 96 h of hypoxic culture (Figure 1A).

The proliferation of VEGF188 EC was stalled after 48 h without serum (Figure 1A, far right panels). Another measure of viability can be the rate of spontaneous apoptosis, which was visualized by Caspase-3/7 staining, and demonstrates that VEGF188 EC are more apoptotic than VEGF164 EC, and that this difference is exacerbated by hypoxia, where there is a five-fold higher frequency of apoptotic cells in VEGF188 than in VEGF164 EC populations (Figure 1B). This is consistent with previous knowledge that the mitogenic activity of VEGF relies on its ability to effectively bind heparin (Keyt et al., 1996; Ferrara, 2010) and, consequently, cell surfaces and receptors, and explains why EC expressing exclusively VEGF120, which exists only as a soluble and diffusible signal (Supplementary Figure 1) do not proliferate and are only partially sustained in culture with supplementation of a potent mitogenic isoform, VEGF164.

VEGF exerts its effects by binding to multiple receptor tyrosine kinases, VEGF receptor 1, 2, and 3 (VEGFR1-3), and co-receptors such as Neuropilins 1 and 2. VEGFR1 is a regulator of VEGFR2 signaling capacity, whereas VEGFR3 is associated with lymphangiogenesis by VEGFC and D (Olsson et al., 2006). VEGFR2 is implicated in all aspects of normal and pathological vascular endothelial-cell biology, and its expression is restricted to endothelial cells (Olsson et al., 2006). It has also been shown that VEGFR2 expression increases in response to hypoxia via a hypoxia-inducible factor (HIF)-1α-driven, VEGF-mediated autocrine loop on EC (Tang et al., 2004). Protein levels of VEGFR2 in EC under normoxia were evaluated by western blot analysis of total cell lysates of EC isolated from each genotype, as shown in Figure 1C. We used fresh isolates of VEGF120 EC to investigate the levels and activation of VEGFR2 in these cells, and strikingly, EC expressing exclusively VEGF120 were also unable to stabilize VEGFR2, in agreement with their failure to divide. WT EC purified from E17.5 littermates proliferated normally (not shown) and expressed normal levels of VEGFR2 (Supplementary Figure 2). VEGF164 is found in conditioned medium and cell lysates, and VEGF188 almost exclusively bound to extracellular matrix (Supplementary Figures 1, 3). VEGFR2 is most abundant in normoxic VEGF164 EC, which is also the only genotype where hypoxia-dependent induction of VEGFR2 is observed. On the other hand, tyrosine phosphorylation of VEGFR2 (normalized to total VEGFR2 levels) was equivalent in both VEGF164 EC and VEGF188 EC under hypoxia (Figure 1D). This is an indication that autocrine VEGF164, but not VEGF188, is necessary for the hypoxia-induced up-regulation of VEGFR2, whereas the phosphorylation of VEGFR2 under hypoxia is not necessarily correlated with VEGFR2 levels. We do see an increase in P-VEGFR2 in WT cells in response to hypoxia, suggesting that both VEGF164 and VEGF188 are required for the increased activation levels of this receptor in response to hypoxia.

To investigate the distinct autocrine roles of each VEGF isoform on endothelial barrier function, morphology and expression levels of cell junctional architecture was assessed and compared in each culture. At endothelial adherens junctions, VEGFR2 coexists with β-catenin and vascular endothelial (VE)-cadherin, known to regulate the half-life and the internalization of VEGFR2 (Calera et al., 2004; Lampugnani et al., 2006). VEGF signaling itself also induces the endocytosis of VE-cadherin (Gavard and Gutkind, 2006). We see the highest levels of VE-cadherin in VEGF164 EC during the 48-h course of hypoxia, and the lowest in VEGF120 EC (Supplementary Figure 4). β-catenin levels, also induced by hypoxia, were elevated in both VEGF164 EC and VEGF188 EC, and maintained at high levels throughout the hypoxic treatment, whereas the induction of this protein in WT EC is transient, decreasing after the 8 h time point. Lowest levels of this protein were found in VEGF120 EC (Supplementary Figure 4).

The morphology of EC expressing individual VEGF isoforms was investigated by immunofluorescent (IF) staining of VE-cadherin and β-catenin (Figures 2A,B). VEGF164 EC and WT EC formed a cohesive and uniform monolayer. In contrast, VEGF188 EC displayed a more mesenchymal appearance with elongated, spindle-shaped cells with extended processes, correlated with little cell-cell association. Furthermore, VE-cadherin of VEGF188 EC was ectopically internalized instead of being localized to the cell surfaces (Figure 2A). In hypoxia, only WT EC showed increased signal for VE-cadherin, implying that hypoxia induction of VE-cadherin levels or increased network formation may require the concomitant expression of all VEGF isoforms (Figure 2A). β-Catenin IF also showed close longitudinal alignment along EC in VEGF164 EC, localizing to the cell membrane. Instead, in VEGF188 EC, β-catenin was seen in the nucleus (Figure 2B). In hypoxia, cytosolic β-catenin levels were increased in all EC lines, including WT (data not shown).

The hypoxia-driven increase in endothelial permeability was more pronounced in VEGF188 EC than in VEGF164 EC (Figure 2C), with WT EC showing the lowest paracellular permeability, an indication that all three VEGF isoforms are required for effective EC barrier function. PECAM-1, classically involved in several types of EC adhesion junctions, was most abundant in VEGF164 EC (Supplementary Figure 4). Similarly, VEGF188 EC monolayers were more permissive to tumor cell migration (LLC^GFP) than VEGF164 EC monolayers (Figure 2D).

We next investigated angiogenic potential in each type of EC by performing adhesion, migration, and network formation assays. Previous studies, which analyzed vascular development and angiogenesis of mouse embryos expressing only VEGF120, underscored the importance of matrix-binding isoforms, VEGF164 and/or VEGF188, for cardiovascular development (Carmeliet et al., 1999; Ruhrberg et al., 2002; Van Den Akker et al., 2007, 2008). The adhesion of EC toward the extracellular matrix protein fibronectin was significantly higher in VEGF164 EC and lower in VEGF188 EC, when compared to that of WT (Figure 2E). In scratch assays on fibronectin-coated plates, hypoxic VEGF188 EC showed decreased cell migration relative to WT and VEGF164 EC (Figure 2F). Integrin β3 subunit, a component of the binding structure to fibronectin, was up-regulated by hypoxia in VEGF164 EC (Supplementary Figure 5), suggesting that VEGF164 is more critically involved in cell adhesion and migration of EC than VEGF188.

To determine the role of specific isoforms in angiogenesis, each EC genotype was plated onto Growth Factor Reduced Matrigel and allowed to self-organize into capillary-like structures. The basal angiogenic potential in normoxia was higher in VEGF188 EC than in VEGF164 EC (Figures 3A,B), with VEGF188 EC extending with more frequent junctions and mesh-like structures in normoxia than VEGF164 EC (Figures 3A,B). In contrast, formation of capillary-like mesh structures, induced by low oxygen tension, was profound in VEGF164 EC but not in VEGF188 EC (Figures 3A,B), suggesting the differential importance of VEGF164 and VEGF188 on network maturation under hypoxia and normoxia, respectively, due to distinct molecular characteristics.

Nitric Oxide (NO) is a bioactive modulator of various types of cells leading to the regulation of vascular tone, immune response, and synaptic neurotransmission (Napoli et al., 2013). It is known that endogenously produced NO can increase endothelial cell permeability, and suppress cell proliferation through cell cycle arrest of EC (Tanner et al., 2000; Villalobo, 2006). Endothelial NO synthase (eNOS) is a fundamental determinant for physiological NO production in EC. eNOS is activated by the serine/threonine protein kinase Akt/PKB and is up-regulated by VEGF (Bouloumie et al., 1999; Dimmeler et al., 1999). We have shown that NO production by EC in hypoxia is significantly increased by activation of the inducible NOS isoform (iNOS), a target of HIF-1α (Branco-Price et al., 2012). To investigate the role of VEGF isoforms in hypoxia-induced NO production by EC, we measured accumulation of NO in EC-conditioned media. NO was detected in all types of EC under normoxia, and the hypoxia-induced NO production was significant only in VEGF188 EC and WT EC, whereas VEGF164 EC did not show any increments of NO by hypoxia (Figure 4A). Treatment with LY294002, an inhibitor of phosphatidylinositol-3-OH kinase (PI3K), attenuated the hypoxia-induced NO release from VEGF188 EC and WT EC, but had no effect on the NO production by VEGF164 cells (Figure 4A), indicating the possible relationship between hypoxia-induced NO production of EC and activated PI3K under autocrine VEGF188. As the downstream molecules of PI3K pathway, hypoxia-induced phosphorylation of Akt at Ser473 was highest in VEGF188 EC (Figure 4B). Phosphorylation of Akt at Thr308 was not promoted by hypoxia (data not shown). Treatment with 1400 W, an iNOS-specific inhibitor, cumulatively increased hypoxia-mediated VEGFR2 stability (Supplementary Figure 6). These data indicate that each VEGF isoform acts distinctly on hypoxia-dependent stabilization of VEGFR2, and the underlying process is, at least partially, a consequence of differential NO production, as a result of altered PI3K/Akt activation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.