Regulation of human feto-placental endothelial barrier integrity by vascular endothelial growth factors: competitive interplay between VEGF-A₁₆₅a, VEGF-A₁₆₅b, PIGF and VE-cadherin

The human placenta is a fetal organ which allows oxygen and selective nutrient uptake from the mother to the fetus, whilst acting as a discriminatory barrier. It does this by having a unique architecture, specific to human and macaque monkeys, where the placental microvessels, encased in a single layer of syncytiotrophoblast, lie bathed in maternal blood (haemomonochorial). fetal blood enters these vessels through the umbilical arteries and returns replenished via the umbilical vein. Both the thin outer syncytial lining and the fetal endothelium act as resistance in series to transport of hydrophilic solutes from maternal to fetal blood [1]. The feto-placental endothelial barrier integrity is therefore of critical importance to fetal growth and well-being.

The placental endothelium is continuous, with well-defined cell–cell junctions that restrict movement of large hydrophilic molecules (>65 kDa) across the paracellular cleft [2]. Adherens junctions (AJs) are the major regulators of paracellular permeability in the placental capillaries with the transmembrane adhesion molecule – vascular endothelial (VE)-cadherin being the key player [3–5]. Phosphorylation of VE-cadherin leads to breakage of homophilic binding, loss of anchorage to peri-junctional actin and translocation from AJ domains with accompanying increases in paracellular cleft dimensions and increased paracellular permeability [6]. VEGF-A, via stimulation of VEGFR2, has been shown to increase phosphorylation of VE-cadherin at Tyr-685 and Tyr-731 facilitating increased solute permeability and extravasation of cells [7].

The human placenta expresses the pro-angiogenic/pro-permeability VEGF-A₁₆₅a, anti-angiogenic VEGF-A₁₆₅b, placental growth factor (PIGF) and their receptors, VEGFR2 (KDR), VEGFR1 (Flt-1), neuropilin-1 and soluble Flt-1 (sFlt-1). There is differential expression of the growth factors during gestation and in complicated pregnancies. VEGF-A levels are highest in the first trimester during de novo synthesis of placental vessels [8,9]). The in situ location also alters with gestation; in the last trimester VEGF-A is found predominantly in the terminal villi which house dilated fetal capillary loops involved in materno-fetal exchange [10]. Elevated levels of VEGF-A, loss of junctional VE-cadherin and increased vascular leakage have been reported in pregnancies complicated by maternal diabetes ([11,12]) whilst trophoblast derived factors from pre-eclamptic placenta have been shown to diminish barrier function and alter VE-cadherin distribution in vitro [13]. The anti-angiogenic splice variant VEGF-A₁₆₅b has been shown to be present in human term placentae as a small part of the total VEGF expression. Interestingly, there is a further down-regulation of this in pre-eclamptic placenta [14]. Whether this splice-variant can affect placental vascular permeability is not known.

PIGF, a homologue of VEGF [15], rises steadily until the second trimester of pregnancy – a period of maximal vessel maturation and then begins to fall [16]. Lower PIGF levels in maternal circulation have been correlated with small-for-gestational-age babies in normal pregnancies [17,18]. A down-regulation of syncytiotrophoblast PlGF was found in pre-eclamptic placenta [19]. Its function in the placenta remains to be shown experimentally. Of the various isoforms of PIGF, PIGF-2 is thought to enhance VEGF-induced permeability, whilst PIGF promotes VEGF-induced angiogenesis in some models and antagonizes in others [20,15]).

The interplay of VEGF-A₁₆₅a, VEGF-A₁₆₅b and PIGF depend largely on their interactions with their receptors and co-receptors. VEGF-A₁₆₅a is able to bind to both VEGFR1 and R2 but acts mainly via the latter [21]. Interestingly, neuropilin-1 can complex with VEGFR1 and R2 and has been shown to potentiate VEGFR2 activation [22]. VEGF-A₁₆₅b binds to both VEGFR1 and R2 receptors but is a weaker agonist for VEGFR2 [23]. VEGF-A₁₆₅b differs from VEGF-A₁₆₅a only in the last six amino acids, the residues necessary for interaction with neuropilin-1, [24–26] and this lack of co-receptor binding may be behind its function as a partial receptor agonist. PlGF only binds to VEGFR1 but has been shown to propagate kinase cascades for VE-cadherin autophosphorylation [27]. The competitive interplay between the VEGF splice variants and PIGF in regulation of human placental endothelial junctional integrity requires elucidation.

The aim of the present study was therefore to investigate whether VEGF-A₁₆₅a, VEGF-A₁₆₅b and PlGF, singly or in combination can affect junctional occupancy of VE-cadherin and alter paracellular permeability of human placental/fetal endothelium. Term placenta and umbilical cord from healthy pregnancies were used for ex vivo placental perfusion experiments and for endothelial primary cell culture studies. Results obtained will advance understanding of the mechanisms employed by the human placenta for maintenance of the feto-placental endothelial barrier and how this could be perturbed in complicated pregnancies.

Term placentas and umbilical cords were obtained at elective caesarean section from normal pregnancies with informed patient consent and full ethical approval (REC Ref 14/SC/ 1194; NHS Heath Research Authority, U.K.), and permission from Nottingham University Hospitals, NHS Trust, U.K. The work described here has been carried out in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki).

All term placentas and umbilical cords were obtained from term (≥37 weeks), non-labouring women (Table 1) undergoing scheduled elective caesarean section delivery. Indications for caesarean section include maternal request, breech presentation, and previous caesarean section. Women with pre-existing conditions such as hypertension (>140/90 mmHg), proteinuria, diabetes, gestational diabetes, renal and cardiac disease, and other conditions that may compromise patient health were excluded. Smokers were excluded from the study.

Human umbilical vein endothelial cell (HUVECs) were isolated from freshly delivered umbilical cords and cultured on 1% gelatin-coated flasks in complete endothelial cell medium M199 (Gibco), supplemented with 20% fetal bovine serum (FBS), heparin sodium salt (50 μg/ml), endothelial cell growth factor supplement (50 μg/ml), and penicillin/streptomycin (100 U/100 mg/ml), under humidified conditions at 37°C and 5% CO₂/95% air. In all experiments, HUVECs were grown and used only up to the third passage.

Polyester membrane transwell 6.5 mm² inserts with a 0.4 μm pore diameter (Corning, U.K.) were coated with 1% gelatin in 0.1 M PBS. They were suspended in a 24-well cell culture plates and seeded with 5 × 10³ cells. Hundred microlitres of complete endothelial cell medium M199 was added to the upper compartment and 400 μl added to lower compartment to equalize hydrostatic fluid pressures. HUVECs were allowed to form a confluent monolayer before experimentation. There were four inbuilt experimental repeats (transwell assays) per experiment. The whole was repeated (×4) using HUVECS isolated from four different cords.

Medium in the upper chamber was replaced with 100 μl of dialysed (10 kDa dialysis tubing) fluorescein isothiocyanate-conjugated bovine serum albumin (FITC-BSA, 66 kDa; Thermo Fisher) in phenol-free M199 medium at a concentration of 1 mg/ml. After a 30-min equilibration, HUVECs were exposed to a single addition of recombinant human VEGF-A₁₆₅a (50 ng/ml), VEGF-A₁₆₅b (50 ng/ml) or PlGF (50 ng/ml). In further experiments, VEGF-A₁₆₅a (50 ng/ml) was followed by VEGF-A₁₆₅b (100 ng/ml), PlGF-2 (264-PG, R&D Systems 100 ng/ml) or both VEGF-A₁₆₅b (100 ng/ml) and PlGF-2 (100 ng/ml) in combination. Fifty microlitres of samples were collected from the lower compartment at 0, 30, 60, 90 and 120 min intervals, with the equivalent volume of fresh phenol-free M199 medium replenished after each sample collection. Samples were diluted in 50 μl of 0.1 M PBS and measured using a Thermo Fluoroskan Ascent F2 fluorescence plate reader at an emission/excitation wavelength of 495/520 nm. Concentrations of FITC–albumin in each sample were calculated via linear regression of a serial dilution series of the tracer. Total permeability of monolayers were calculated for each treatment based on the rate of solute flux (calculated in µg/min) over the first 30 min, the surface area of the monolayer, and the initial concentration difference between the upper and lower wells (1 mg/ml).

After experimentation, monolayers were immediately fixed in 1% PFA. Cells were permeabilized with 0.15% Triton X-100 followed by blocking with 4% normal human serum in 0.1 M PBS for 30 min at RT. Monolayers were incubated with mouse anti-human CD144 (ThermoFisher) (VE-cadherin; 5 μg/ml) overnight at 4°C and FITC-conjugated goat anti-mouse IgG (Sigma-Aldrich, 20 μg/ml) for 2 h at 37°C in the dark after requisite washes.

A well-established dual-perfusion procedure was employed [1,2,9–12]. To summarize, immediately after caesarean delivery, the placenta was transferred to a 37°C chamber with the umbilical cord kept clamped until cannulation to prevent loss of blood and collapse of feto-placental vessels. Within 20 min of arrival of the placenta, a vein and an artery supplying the microvascular bed of a randomly chosen intact cotyledon were each cannulated with a nasogastric (5 mm) tube to establish the fetal circulation. The placenta was inverted and the cotyledon was clamped in a Perspex chamber to isolate the lobule from the rest of the placenta. The independent maternal circulation was simulated by inserting five 5 mm nasogastric tubes into the intervillous space through the basal plate of the cotyledon, and drained through an exit tube in the Perspex chamber. Fetal and maternal circulations were connected to peristaltic pumps providing a constant 20 ml/min flow to the maternal circulation and a 5 ml/min flow to the fetal circulation, replicating the physiologic flow rates seen in utero. Establishment of both circulations were completed within 30 min of delivery to minimize hypoxic damage to the placenta. Perfusion was abandoned if fetal venous outflow was less than arterial inflow. From the 20 intact placenta recruited for the perfusion experiments, 15 allowed full perfusion (25% failure rate).

A 20-min open circuit equilibration period with oxygenated Medium 199 (Sigma, Poole, U.K.), with added sodium bicarbonate (2.2 g/l), albumin (5 g/l), high molecular weight dextran (2000 M_r; 8 g/l) and heparin (5000 IU/l) (final pH 7.2–7.4), was performed to reverse any post-parturition hypoxic changes [1]. After equilibration, fetal oxygenation was discontinued and both the maternal and foetal circulations were closed. Perfusion pressures were monitored and accepted if foetal pressure was between 40–80 mmHg and maternal pressure was between 18–20 mmHg [2].

Recombinant human VEGF-A₁₆₅a (20 ng/ml), VEGF-A₁₆₅b (20 ng/ml) or vehicle were introduced into the fetal circulation and the lobules perfused for 30 min. In reversal experiments, VEGF-A₁₆₅a exposure was followed by a separate VEGF-A₁₆₅b (40 ng/ml) perfusion (vehicle only acted as control) for an additional 30 min. Three different placentas were used for each experimental exposure. Tetramethylrhodamine isothiocyanate (TRITC)-conjugated dextran (76 M_r; Sigma, Dorset, U.K.), 1 mg/ml) was added as a bolus into the fetal circulation for the last 10 min of all perfusions (single growth factors or sequential VEGF isoform perfusions). The lobules were then perfusion fixed with 1% paraformaldehyde in 0.1 M phosphate buffer saline (pH 7.3) for 30 min. The lobule was excised and 10 mm³ biopsies taken for a further immersion fixation (2 h). Biopsies were rinsed in PBS, cut into 5 mm³ pieces, frozen in nitrogen-cooled isopentane (Fisher Scientific UK Limited, Loughborough, U.K.) and stored at −80°C until required.

A minimum of six different blocks (randomly chosen) per placenta were cryo-sectioned and a minimum of six sections (5 μm thick) were taken from each block at different depths. Cryo-sections were air-dried and washed in 0.1 M PBS/BSA, before permeabilization with 0.15% Triton X-100 followed by blocking with 4% normal human serum in 0.1 M PBS for 30 min at room temperature. Sections were incubated overnight with mouse anti-human CD144 (VE-cadherin; 5 μg/ml) at 4°C and secondary antibodies as described before.

Transwell membranes and placental sections were visualized for expression of CD144 and TRITC-Dextran tracer using a Nikon LaboPhot-2 fluorescence microscope (Nikon, U.K.) and appropriate TRITC/FITC filters. Images (obtained by systematic random sampling of entire sections or membranes) from both channels were acquired using a Nikon Coolpix 995 camera (Nikon, U.K.).

Micrographs were analysed using Adobe Photoshop 6.0 (Adobe systems, U.K.). A pre-determined electronic grid was placed over each image and both VE-cadherin junction integrity and evidence of TRITC-Dextran tracer leakage quantified using systematic random sampling and unbiased ‘forbidden line’ counting principle to ensure that no vascular profile and paracellular cleft were counted twice [11]. For sampling efficiency, a minimum of 200 vascular profiles were counted from images per perfused placenta; c 600 vascular profiles were analysed for each experimental condition. For HUVEC monolayers, junctional integrity was determined by counting the % of paracellular clefts showing uniform VE-cadherin staining as a continuous thin line. Gaps in cell–cell adhesion regions with VE-cadherin negative cell edges were also counted. The % of cell–cell overlap showing thicker bands of VE-cadherin localization was noted. For placental sections, the total percentage of vascular profiles showing disrupted junctional VE-cadherin (discontinuous staining or total loss) from paracellular clefts and associated tracer leakage (76 M_r Dextran-TRITC visualized as peri-vascular ablumenal fluorescent puncta or ‘hotspots’) was counted [11]. All experimental images from both HUVEC monolayers and placental sections were blinded to treatment regime before analyses.

High-attachment 96-well ELISA plates (Corning, U.K.) were coated with recombinant Flt-1 overnight, and co-incubated with biotinylated-VEGF-A₁₆₅b (EC75), followed by non-biotinylated-VEGF-A₁₆₅a (0–160 nM) or PlGF (0–160 nM). Differences in VEGF binding to Flt-1 (n=12) were measured using streptavidin–HRP to biotin interactions, using a Thermo Fluoroskan Ascent F2 fluorescence plate reader at an absorbance wavelength of 450 nm.

All statistical analysis was carried out using GraphPad Prism. Comparisons of means were made using one-way ANOVA with Bonferroni’s post hoc test.

Incubation of HUVEC monolayers with VEGF-A₁₆₅a resulted in an increase in FITC–albumin tracer mass leakage over time (Figure 1A) compared with vehicle exposed cells. In contrast, incubation with VEGF-A₁₆₅b in isolation showed no change in FITC–albumin tracer mass permeate compared with vehicle over time. Co-incubation of VEGF-A₁₆₅b following 30 min of VEGF-A₁₆₅a exposure resulted in the inhibition of VEGF-A₁₆₅a-mediated increase in FITC–albumin tracer mass leak (Figure 1A).

Exposure to PlGF alone also resulted in no change in FITC–albumin tracer mass permeate with time (Figure 1B). Addition of PlGF after 30 min exposure to VEGF-A₁₆₅a did not prevent the VEGF-A₁₆₅a-mediated effects (see Figure 1B). Addition of PlGF alongside VEGF-A₁₆₅b prevented VEGF-A₁₆₅b-induced inhibition of VEGF-A₁₆₅a-mediated increase in FITC–albumin tracer mass leak (see Figure 1C). The calculated permeability values to FITC–albumin tracer reflected these observations, whereby exposure to VEGF-A₁₆₅a, but not VEGF-A₁₆₅b or PlGF significantly increased HUVEC monolayer permeability (P<0.001, Figure 1D). Similarly, VEGF-A₁₆₅a-mediated permeability increase was inhibited by VEGF-A₁₆₅b (P<0.001) but not PlGF (P>0.05). PlGF co-incubation with VEGF-A₁₆₅b abolished the rescue of the latter from VEGF-A₁₆₅a dependent increases in permeability (P<0.001) (Figure 1D).

VE-cadherin immunostaining demonstrated the presence of positive junctional cell–cell overlap regions and thin abutting junctions with continuous VE-cadherin staining (Figure 2) in control (vehicle only) experiments. Monolayers treated with VEGF-A₁₆₅a revealed observable endothelial junctional disruption after exposure (Figure 2B) with a decrease in continuous thin junctions and increase in discontinuities or total loss of VE-cadherin staining in cell–cell margins (described as gaps). This was not seen when monolayers were treated with VEGF-A₁₆₅b (Figure 2C) or PlGF (Figure 2D). Quantification (systematic random sampling) of the percentage of each type of VE-cadherin positive regions (Figure 2H) showed that there was a statistically significant decrease in the number of thin junctions showing continuous VE-cadherin occupancy (P<0.0001) after VEGF-A₁₆₅a exposure. VE-cadherin discontinuity within paracellular clefts increased with a significant increase in observable gaps (P<0.001). No significant change in both percentage of disrupted junctions and % of continuous VE-cadherin junctions were found in VEGF-A₁₆₅b or PlGF experiments when compared with controls (P>0.05, Figure 2H).

VEGF-A₁₆₅b (Figure 2E), but not PlGF (Figure 2F), was able to prevent VEGF-A₁₆₅a-induced disruption of VE-cadherin junctions after 2-h exposure. However, addition of PlGF in combination with VEGF-A₁₆₅b (Figure 2G) prevented the observed VEGF-A₁₆₅b-mediated inhibition of VEGF-A₁₆₅a-induced junctional disruption. Analysis of percentage junctional integrity (Figure 2I) revealed a significant decrease in percentage of gaps after VEGF-A₁₆₅b co-incubation (P<0.0001) and an increase in percentage of thin continuous VE-cadherin junctions (P<0.01). No significant changes in either percentage gap or continuous VE-cadherin AJ staining were observed after PlGF co-incubation (P>0.05). However, PlGF exposure in combination with VEGF-A₁₆₅b abolished the VEGF-A₁₆₅b inhibition of VEGF-A₁₆₅a-mediated effects (P>0.05) (see Figure 2I).

As the VEGF-A₁₆₅b inhibition of VEGF-A₁₆₅a mediated increased permeability was blocked by PlGF, which only binds VEGFR1 (flt-1), we hypothesized that this could occur if VEGF-A₁₆₅b was inhibiting VEGF-A₁₆₅a by signalling through VEGFR1. VEGF-A₁₆₅a and VEGF-A₁₆₅b compete for R1 [28], so we therefore determined the effect of PlGF on VEGF-A₁₆₅b binding to VEGFR1 using an Fc-VEGFR1 chimaeric protein. Incubation of Fc-VEGFR1 with un-labelled human recombinant VEGF-A₁₆₅b resulted in a concentration-dependent decrease in binding of biotinylated-VEGF-A₁₆₅b (EC75). Similarly, co-incubation with un-labelled PlGF resulted in an almost identical concentration-dependent decrease in biotinylated-VEGF-A₁₆₅b binding, with no significant difference in calculated IC₅₀ between the two proteins (P>0.05) (see Figure 3), thus suggesting that the inhibition of VEGF-A₁₆₅a-mediated permeability by VEGF-A₁₆₅b could also be through its actions on VEGFR1.

Using the more physiological dual-perfusion system, addition of growth factors to the fetal microcirculation of term placental lobules resulted in observable differences in extravasation of TRITC-dextran (75 M_r) tracer (see Figure 4, right panel). VEGF-A₁₆₅a perfusions for 30 min resulted in a significant increase in percentage of vascular profiles associated with tracer leakage (71.2 ± 13.8%, Figures 4 and 5) compared with control perfusions (30.2 ± 4.4%; P<0.01). In contrast, perfusions of VEGF-A₁₆₅b saw no significant increase in percentage of leaky vessels (P>0.05) compared with control perfusions. Addition of VEGF-A₁₆₅b into the closed fetal circulation after a 30 min VEGF-A₁₆₅a perfusion period resulted in altered tracer leakage profile from that obtained with VEGF-A₁₆₅a only perfusions. Vascular profiles (9.4 ± 1.2%) were now found to be associated with peri-vascular ‘hotspots’ (P<0.05). This was not due to temporal recovery, as addition of vehicle only for 30 min after VEGF-A₁₆₅a perfusion showed no significant decrease in profiles expressing hotspots (P>0.05). Subsequent immunohistochemical analyses of the same sections revealed that tracer leakage matched disruption of VE-cadherin AJs (loss of VE-cadherin from paracellular clefts) (Figure 4, left panel). VEGF-A₁₆₅a-perfusion of placental microvascular beds resulted in a 49.3% decrease in vascular profiles showing VE-cadherin at paracellular clefts (P<0.01), while perfusions of VEGF-A₁₆₅b did not significantly alter VE-cadherin positive vascular profiles (86.7 ± 0.8%) compared with controls (89.3 ± 2.3%) (P>0.05). Sequential addition of VEGF-A₁₆₅b to VEGF-A₁₆₅a perfusions resulted in a recovery of junctional integrity with 81 ± 5% of vascular profiles showing VE-cadherin at paracellular clefts (P<0.05; Figure 5).

These studies are the first to show that the anti-angiogenic VEGF-A₁₆₅b does not disturb the junctional occupancy of VE-cadherin or induce paracellular tracer leakage in the human placental microvascular bed and HUVEC monolayers. Indeed, VEGF-A₁₆₅b (at a 2-fold concentration) can block/reverse VEGF-A₁₆₅a-mediated increases in feto-placental endothelial permeability to macromolecules (75 kDa) and loss of VE-cadherin from AJs. In these experiments, PlGF did not affect endothelial integrity, VE-cadherin localization and permeability when added singly; however, it could prevent the rescue of VEGF-A₁₆₅a-mediated effects by VEGF-A₁₆₅b.

The similarities between the observed induced changes in both fetal endothelial cell cultures and the more physiological and complex perfused microvascular bed are re-assuring. Placental microvessels showed a robust response to a 30 min perfusion of exogenous VEGF-A₁₆₅a, with loss of VE-cadherin and increased vascular leak of 76 kDa dextrans. These changes were not induced when microvessels were perfused with VEGF-A₁₆₅b; indeed this splice variant could reverse the VEGF-A₁₆₅a-induced changes both in the perfusion model and in transwell experiments.

The monolayer permeability results are similar to those recently described in lung pulmonary endothelial cells [29], where VEGF-A₁₆₅a but not VEGF-A₁₆₅b was shown to increase monolayer permeability and decrease junctional integrity. However, they did not investigate the effect of PlGF on these cells.

The observed inability of either PlGF or VEGF-A₁₆₅b to alter VE-cadherin occupancy and AJ integrity in human placental microvessels and fetal endothelial cells may be due to their lack of neuropilin-1 binding property [22,26,30]. Neuropilin-1 is critical for VEGF-mediated endothelial permeability in human pulmonary endothelial cells and for pulmonary vascular leaks in inducible lung-specific VEGF transgenic mice [22]. Stable transfection of the neuropilin-1–VEGFR2 complex in endothelial cells resulted in decreased transendothelial resistance in a dose-dependent fashion following addition of VEGF-A₁₆₅a; this was not seen for single transfections of VEGFR2 or neuropilin-1 alone. Moreover, VEGF-A₁₆₅b prevents the formation of neuropilin–VEGFR2 complexes by VEGF-A₁₆₅a [31]. The reduction in VEGF-A₁₆₅a-mediated enhanced permeability in our studies may be due to competitive binding of VEGF-A₁₆₅b to the VEGFR2 receptor [25]. KDR occupancy by VEGF-A₁₆₅b may be followed by internalization but not the subsequent neuropilin-1 dependent re-shuttling of KDR to the membrane surface [32]. This therefore would rule out further VEGF-A₁₆₅a binding and triggering of phosphorylation events that lead to translocation of VE-cadherin from AJ domains. The recovery of junctional integrity seen in the sequential VEGF-A₁₆₅a + VEGF-A₁₆₅b perfusions, but not in VEGF-A₁₆₅a + vehicle perfusions argues that VEGF-A₁₆₅b is acting as an active signalling inhibitor. Our data allow one to hypothesize that the relative contributions of the two different VEGF-A₁₆₅ isoforms may be an important driver behind the different feto-placental vascular permeability (and angiogenesis) observed for the different trimesters of pregnancy [9,10].

Plasma taken from pre-eclamptic mothers has been shown to induce transient increases in permeability in amphibian models, which was blocked by VEGF-A₁₆₅b specific neutralizing antibodies, and receptor tyrosine inhibitors at concentrations specific to VEGFR1 blockage [33]. This neutralizing antibody to VEGF-A₁₆₅b was shown to prevent the inhibition of VEGF-A₁₆₅b-mediated blockade of VEGF-A₁₆₅a-induced migration and cytoprotection of endothelial cells [34] and anti-angiogenesis in peripheral vascular disease [35]. This suggests that the pre-eclamptic plasma not only contained physiologically active VEGF-A₁₆₅b, but its action was incurred by altering the balance of VEGF-A₁₆₅b, PlGF and VEGF-A₁₆₅a, such that VEGF-A₁₆₅a was no longer able to induce the increase in permeability, potentially by binding heterodimers of VEGF-A₁₆₅b with either PlGF or VEGF-A₁₆₅a. Such heterodimers are theoretically possible but have not yet been clearly demonstrated.

PIGF did not increase AJ disruption or permeability of the fetal human umbilical vein endothelial cells. PIGF-1, but not PIGF-2, has been shown to stabilize VE-cadherin junctions after activation with VEGF in bovine retinal endothelial cells and after intra-vitreal injections in mouse, during a critical window [36]. The isoform used here was PlGF-2, which binds heparin and is therefore able to signal through VEGFR1 with heparin, indicating that full signalling does require the heparin binding domains. The PlGF-mediated abolition of VEGF-A₁₆₅b induced reversal of VEGF-A₁₆₅a-induced permeability was surprising. However, recent studies by Ganta et al. [37] have shown that VEGF-A₁₆₅b inhibits VEGF-A₁₆₅a-mediated signalling in adipose endothelial cells through inhibiting VEGFR1-mediated activation of the signal transducer and activator of transcription 3 (STAT3) pathway. It is therefore possible that VEGF-A₁₆₅b binds VEGFR1, inhibiting STAT3 signalling. When PlGF binds VEGFR1, even if it does not inhibit STAT3 signalling (or stimulate it), it prevents VEGF-A₁₆₅b from inhibiting it, so VEGF-A₁₆₅a would then be at liberty to increase permeability. There are currently no studies identifying whether PlGF signals through STAT3 or how it affects VEGF-A₁₆₅a-mediated STAT3 signalling, but it would appear that such studies are warranted.

In our HUVEC studies, when PlGF was added in combination with VEGF-A₁₆₅b, there was an abolition of the VEGF-A₁₆₅b-induced reversal of VEGF-A₁₆₅a-induced permeability. Whilst it is well known that VEGF-A₁₆₅a and PIGF can compete for VEGF-R1 binding [38], in the present study we have also shown that incubation of Fc-VEGFR1 with un-labelled PlGF resulted in a concentration-dependent decrease in biotinylated-VEGF-A₁₆₅b binding. Thus, addition of PIGF may have resulted in release of both splice variants. The excess VEGF-A₁₆₅a would then be at liberty to increase permeability via KDR–neuropilin signalling. Hetero-dimerization of PIGF with VEGF-A₁₆₅b could further assist VEGF-A₁₆₅a activity. Further studies are needed to understand the complex interplay between VEGF-A/PIGF and their receptors/co-receptors. Elucidation of the multiple signalling pathways, including whether/where the VEGFR1-linked angiogenic signalling pathway interacts with the VEGFR2-linked phosphorylation events that lead to disruption of VE-cadherin junctions is needed to understand how placental/fetal barrier function is regulated.

In summary, we have shown that VEGF-A₁₆₅b is able to inhibit vascular permeability induced by VEGF-A₁₆₅a in vitro, and in human placenta ex vivo, and that this is interfered with by placental growth factor. These results suggests that alterations in the ratio of these growth factors during normal placental development and in complicated pregnancies such as pre-eclampsia and diabetes would influence VE-cadherin clustering in fetal vessels and therefore placental barrier function, given that both the fetal endothelium and the syncytiotrophoblast act as resistance in series to materno-fetal hydrophilic solute transport.

• The anti-angiogenic VEGF-A₁₆₅b isoform does not increase permeability in human placental microvessels or fetal endothelial cells.

• VEGF-A₁₆₅b isoform can interrupt VEGF-A₁₆₅a-induced permeability of feto-placental vessels.

• The interplay of the VEGF-A₁₆₅ isoforms with PIGF suggests that the ratio of these three factors may be important in determining the placental endothelial barrier and therefore fetal well-being in normal and complicated pregnancies.

This work was supported by the British Heart Foundation [ grant number PG/13/85/30536]. We would like to thank clinical staff and midwives at the Labour Ward, Queens Medical Centre, Nottingham University Hospital for the timely retrieval of placenta.

The authors declare that there are no competing interests associated with the manuscript.

V.P. performed most of the experiments, statistical analyses and was involved in writing the manuscript. L.L. and D.O.B. co-designed and co-directed the study and were involved in manuscript writing. L.L. assisted with the placental perfusions, directed analyses of the tracer leakage and VE-cadherin dynamics. All authors reviewed the manuscript.

AJ

adherens junction

FITC-BSA

fluorescein isothiocyanate-conjugated bovine serum albumin

Flt-1

Fms-like tyrosine kinase

HUVEC

human umbilical vascular endothelial cell

KDR

kinase insert domain receptor

PI

propidium iodide

PIGF

placental growth factor

PFA

para-formaldehyde

sFlt-1

soluble fms-like tyrosine kinase

STAT3

signal transducer and activator of transcription 3

TRITC

tetramethylrhodamine isothiocyanate

VE-cadherin

vascular endothelial-cadherin

VEGF

vascular endothelial growth factor

VEGFR1

vascular endothelial growth factor receptor-1

VEGFR2

vascular endothelial growth factor receptor-2