The phenotype of the human materno-fetal endothelial barrier: molecular occupancy of paracellular junctions dictate permeability and angiogenic plasticity

Abstract

In vitro models predict that molecular occupancy of endothelial junctions may regulate both barrier function and angiogenesis. Whether this is true in human vascular beds undergoing physiological angiogenesis has not been shown. This review presents data which demonstrate there are two distinct junctional phenotypes, ‘activated’ and ‘stable’, present in the vascular tree of the human placenta taken from two distinct highly angiogenic gestational periods (first and last trimester). Stability is conferred by the presence of occludin in tight junctions and plakoglobin in adherens junctions. Their localization may be influenced by vascular endothelial growth factor and angiopoietins 1 and 2 that have a similar temporal and site-specific differential expression. The junctional phenotypes are reversible, as shown in studies with endothelial cells isolated from placental microvessels and grown in the presence/absence of cAMP-enhancing agents. Reductions in protein levels and loss of junctional localization of adhesion molecules result in increased permeability to macromolecules, whilst up-regulation and re-targeting of these molecules inhibit cell proliferation and increase transendothelial resistance. These studies suggest junctional adhesion molecules can regulate physiological angiogenesis and vascular re-modelling. Moreover, the activated junctional phenotype of placental microvessels allows them to participate in increased growth and proliferation. This junctional immaturity appears to be at the expense of barrier function resulting in sites of maximal materno-fetal solute exchange.

Introduction

Throughout gestation, the human placenta provides a barrier regulating the transfer of materials between the mother and the developing fetus. The placental interface consists of two major tissue layers, the syncytiotrophoblast, which is bathed on its microvillous surface with maternal blood flowing through the intervillous space, and the endothelium of the fetal circulation. This review will concentrate on the endothelial layer which has distinct paracellular junctional entities implicated in regulation of vascular permeability.

The placental fetal vascular system develops by de novo synthesis of blood vessels in the first trimester. Re-modelling and growth of the vascular plexus is thought to occur in the second trimester, whilst angiogenesis or formation of terminal villous capillaries from the existing vasculature is the key event of the last trimester of pregnancy, when fetal demand is at its greatest. At term, it is the areas of the terminal villi that are believed to be responsible for nutrient and gas exchange. The newly synthesized capillaries present in these villi form a looped network with sinusoidal dilations near the inflexions of the loops. The external surfaces of these dilations are closely applied to the basal surface of the syncytiotrophoblast and form areas of presumed least resistance to transplacental diffusion of molecules (Kaufmann et al. 1979). The cellular/molecular mechanisms behind formation and development of this vascular bed and regulation of its barrier property remain neglected.

The paracellular cleft of human placental vessels

As addressed in the review by J. A. Firth in this issue (Firth 2002), the two important physical entities in endothelial paracellular clefts are adherens junctions and tight junctions which, whilst ensuring endothelial cell–cell adhesion, act as resistance in series to intercellular solute transport. Capillaries in the terminal villi of the human placenta possess numerous paracellular tight junctions (TJs) which are different from epithelial junctions (and blood–brain barrier) in showing no positional preference (luminal or abluminal) and being interspersed with adherens junctions along the paracellular cleft. At these TJ appositions there is a 4-nm separation between the outer electron-dense components of the phospholipid bilayers of adjoining endothelial cells (Leach & Firth, 1992). These structural characteristics are not unique to human placental capillaries – rat myocardial, frog mesenteric and guinea-pig placental capillaries, i.e. systemic non-brain continuous capillaries all show this divergence from the epithelial tight junction (Firth et al. 1983; Ward et al. 1988; Adamson & Michel, 1993). Electron microscopical studies of term placenta have shown that the ultrastructural organization of paracellular clefts are remarkably similar throughout the placental vascular tree (Leach et al. 2000). Newly formed vessels in the first trimester were also shown to possess tight junctional appositions, reminiscent of that in term placenta (Leach et al. 2002). Paracellular structural organization alone (as visualized by transmission electron microscopy) appears not to reveal the rationale behind the functional differences of exchange capillaries from large conduit vessels.

Junctional adhesion molecules and regulation of paracellular permeability

Recent studies using in vitro models have shown that the molecules present within cell–cell junctions influence junctional stability and permeability (Dejana, 1996; Furuse et al. 1998). At adherens junctions (AJs), the key transmembrane protein is vascular endothelial (VE) cadherin (Lampugnani et al. 1992). Ve-cadherin is clustered at AJs and mediates cell–cell adhesion through homophilic binding of extracellular domains. VE-cadherin is anchored to peri-junctional actin via three known catenins, β-catenin, γ-catenin (plakoglobin) and α-catenin. Beyond cell–cell adhesion, these molecules are important signal transduction ligands which influence diverse cellular processes including proliferation (reviewed in Conacci-Sorell et al. 2002) and endothelial cell survival (Carmeliet et al. 1999). Truncated VE-cadherin (without the cytoplasmic tail for catenin-binding) can still promote homotypic recognition and weak adhesion of VE-cadherin, but is unable to control paracellular permeability (Navarro et al. 1998). Endothelial monolayer permeability has been linked with the type of catenin present in AJs – in human umbilical vein cells (HUVEC), at early stages of confluency VE-cadherin is primarily linked to β–catenin, with barrier property at a minimum. At full confluence, β-catenin partially detaches from the junctional complex and is replaced by plakoglobin with a concomitant decrease in monolayer permeability to macromolecules (Lampugnani et al. 1995). This mechanism may be a feature of blood vessels under normal or pathological conditions.

In addition to its structural role in adherens junctions, studies from epithelial cells have shown that β-catenin can act as a transcription factor in the nucleus by serving as a coactivator of the lymphoid-enhancing factor (LEF)/TCF family of DNA-binding proteins. β-catenin-mediated transcription is activated by the Wnt signalling pathway, which is crucial during embryonic development. Recruitment of β-catenin to AJs antagonizes LEF/TCF transactivation, i.e. sequestering to junctions can inhibit proliferation (Conacci-Sorell et al. 2002). The same mechanisms may be true for endothelial proliferation and is under investigation in our laboratory.

In systemic microvessels, the discontinuous network of tight junctional strands minimizes the contribution of tight junctions in regulating physiological permeability; they influence the extent of paracellular pathways open for solute exchange (see review by J. A. Firth in this issue). Strand complexity may be influenced by molecules present therein (Furuse et al. 1998), whereby claudin 1 was found to be a component of continuous TJ strands, whilst claudin 2 was localized to discontinuous strands. A further claudin, claudin-5, which may be endothelial-specific, has been located to arteries but not capillaries or veins in the kidney (Morita et al. 1999). The role of occludin, one of the first transmembrane molecules located to tight junctions, is under renewed investigation: tight junctions have been found in occludin knockout cells (Saitou et al. 1998) whilst co-transfection of fibroblasts with occludin resulted in incorporation of occludin with claudin-1-based strands (Furuse et al. 1998). This has led the authors to hypothesize that occludin may be an accessory protein in some function of TJ strands; it may very well be a feature of well-differentiated tight junctions. It is certainly present in epithelial tight junctions and in the blood–brain barrier.

Molecular profile of endothelial junctions in the last trimester of pregnancy

As stated before, at term, the terminal villous capillaries of the human placenta are the sites of maximal materno-fetal exchange. These capillaries are highly angiogenic and are formed in the last trimester of pregnancy by looped angiogenesis from existing microvessels in the intermediate villi. Confocal microscopy studies have revealed that they posses adherens junctions which contain VE-cadherin, α- and β-catenin but lack plakoglobin (Leach et al. 2000; Figs 1 and 2). The similarity of these junctions with HUVEC monolayers at early stages of confluency allowed us to suggest they were immature junctions. The tight junctions present in these vessels also showed negative immunoreactivity to occludin and claudin-1 (Fig. 1), although in the same sections adjacent epithelial tight junctions showed strong immunoreactivity to both these molecules. Given the angiogenic nature of terminal villous capillaries and their predominance as sites of materno-fetal exchange, we suggest that these terminal villous capillaries may be defined as ones representing an ‘activated’ phenotype.

Fig. 1.

Confocal image showing double immuno-labelling of VE-cadherin and occludin in the last trimester placenta. VE-cadherin (green) can be seen in terminal villous capillary loops, whilst both VE-cadherin and occludin (yellow) can be seen in the centrally located profiles of large vessels.

Fig. 2.

Confocal images of last trimester placenta; (a) shows immunolocalization of β-catenin in large vessels and terminal villous capillary profiles; (b) shows absence of plakoglobin immunoreactivity in these same profiles. Positive immunoreactivity can be seen in the large vessel.

Systematic random sampling of immunostained vascular profiles revealed that there were a percentage (∼15%) of terminal villous capillaries that do not show immunoreactivity to junctional molecules (Leach et al. 2000). This population of vessels may represent ones where adhesion molecules are no longer clustered to AJs. Phospho-tyrosine studies show that about 20% of vessels in normal placentae have intense antiphosphotyrosine immunoreactivity at paracellular clefts (Babawale et al. 2000). VEGF, a potent angiogenic factor, has been shown to cause phosphorylation of VE-cadherin, β-catenin and disruption of cell–cell contacts (Esser et al. 1998), prerequisites to both intercellular gap formation and the proliferation switch seen in epithelial cell types (Conacci-Sorell et al. 2002). Furthermore, inflammatory mediators, such as histamine, have been shown to lead to loss of VE-cadherin and β-catenin from adherens junctions of perfused placental microvessels (Leach et al. 1995).

Molecular profile of endothelial junctions in new vessels from first trimester placentae

After the first week of development, the human embryo requires a system for exchange of nutrients and wastes beyond simple diffusion. To do this it establishes the utero-placental circulation, which allows the fetal blood (flowing through the placenta) to come close to the maternal blood (bathing the placental chorionic villi). Vasculogenesis of placental fetal vessels is thought to occur, in the first 4 weeks, from stem cells present in the embryoblast-derived mesencyme invading the chorionic villi (Demir et al. 1989). An effective vascular plexus is fully formed by end of the first trimester (12 weeks). Several studies in the 1980s, using both the macaque monkey (King, 1987) and first trimester human placenta (Demir et al. 1989), have proposed the ultrastructural sequence of events for vasculogenesis, i.e. mesenchymal stem cells differentiate into haemangioblasts, the innermost cells of which become haemopoetic whilst the abluminal become pericytes.

In a recent study, we have reported that the cell–cell junctions of developing placental vessels in the early human placenta possess a dynamic molecular phenotype very distinct from that seen in well-differentiated mature vessels (Leach et al. 2002). The paracellular clefts present in ‘pre-endothelial cells’ in haemangioblast clusters and formed vessels located at the perimeter of first trimester chorionic villi do not contain occludin or claudins 1 and 2. Furthermore, the adherens junctions of the peripheral vessels (Fig. 3) contain VE-cadherin, α- and β-catenin but not plakoglobin, i.e. similar molecular occupancy as that seen in AJs of terminal villous capillaries of full-term placenta. Thus, plakoglobin-deficient adherens junctions may be a necessary prerequisite for vascular re-modelling. In these early placentae, more centrally located vascular profiles with numerous endothelial–endothelial paracellular clefts were found to contain occludin, but not claudin 1 or plakoglobin (Fig. 3).

Fig. 3.

Immunofluorescent micrographs showing molecular occupancy of ‘dynamic’ junctions in the first trimester placental vessels; (a) shows localization of PECAM-1 to all vascular profiles; (b) shows that in the same section occludin is absent from the vascular profiles but present in the trophoblast layer; (c) shows β-catenin immunolocalization in vessels and trophoblast layers; (d) shows plakoglobin immunoreactivity is restricted to the trophoblast only.

Ultrastructural studies of the paracellular clefts in these vascular profiles did not reveal any structural differences that matched the differential expression of occludin. All clefts contained tight junctions interspersed within wide zones. Goniometric tilting studies revealed that these newly formed ‘tight junctions’ were typically not fused. This allowed us to hypothesize that in the human placenta, occludin is not necessary for formation of endothelial tight junctions but may have an accessory role providing stability or added adhesiveness to tight junctions of large vessels.

Zonula occludens 1 (ZO-1), the cytoplasmic linking molecule of tight junctions, was found in all placental endothelial junctions, regardless of position in vascular tree or gestational stage. This molecule may not be exclusively involved in anchorage and signalling of tight junctions; it may be involved in cadherin-based cell adhesion by working as a link between the cadherin/catenin complex and the actin-based cytoskeleton (Itoh et al. 1997). Alternately, or in addition, it may still be a component of tight junctions; as yet unknown molecules may be residents of the placental tight junctions immunonegative to antioccludin, claudin-1 or claudin-2.

Differential expression of growth factors

Vascular endothelial growth factor (VEGF) plays an essential role in vascular permeability (see Kevil et al. 1998; Bates et al. 2002), the formation of new vessels and in endothelial survival. The early human placenta does contain both VEGF and KDR (Clark et al. 1996) and, as stated before, VE-cadherin and β-catenin are the molecular occupants of AJs in these developing vessels. In situ hybridization studies have shown that angiopoietin-1 and -2, the growth factors implicated in vascular re-modelling, are expressed in the trophoblast of early placenta (Dunk et al. 2000). These authors have shown that the expression of angiopoetins is restricted to the perivascular stroma at term, with Ang-1 being specific to stroma supporting large blood vessels of placental stem villi (Dunk et al. 2000).

We have looked at the immunolocalization pattern of VEGF, angiopoietins 1 and 2 in first and last trimester placentae with confocal microscopy (Leach et al. 2002) and found a site-specific and temporal location that may explain the differential expression of junctional proteins seen in the placenta. As expected, VEGF was predominant in the first trimester placenta (the period of vasculogenesis and formation of vascular plexi). It was also located in the terminal villi of the full-term placenta; the latter is the site of newly formed exchange vessels (Fig. 4). VEGF can cause alteration in junctional permeability associated with phosphorylation of adherens junctional molecules, specifically VE-cadherin and β-catenin (McDonald, 1994; Esser et al. 1998). It is tempting to suggest that the same may be true in situ, i.e. local VEGF production may be influencing the expression of junctional adhesion molecules in the terminal villi.

Fig. 4.

Confocal image of last trimester placenta. The angiogenic growth factor, VEGF, is localized predominantly to the terminal villous branches of the placental villous system.

There were also differential immuno-localization of Ang-1 and -2 in the placentae studied, Ang-1 being a feature of large quiescent vessels (in chorionic stem villi) whilst Ang-2 was also present in terminal villi of the full-term placenta. The terminal villous specificity of Ang-2 is not surprising given its role in vascular remodelling from existing vascular beds (Gale & Yancopoulos, 1999). Thurston (2002) clearly demonstrates that Ang-1 plays a major role in reducing vascular leakage and promoting enlargement of existing vessels in overexpressed Ang-1 mice. The differential expression of the Ang-1 in placental stem vessels with stable junctional phenotype implicates angiopoetins in regulation of this phenotype.

Reversal of placental junctional phenotypes: in vitro manipulations

Isolation of endothelial cells (HPMEC) from last trimester placental microvessels shows that, in culture, these cells can be manipulated to exhibit the two different junctional phenotypes seen in situ and these can be linked to monolayer permeability, transendothelial resistance and proliferation index (Dye et al. 2001). Confluent cells exposed to endothelial-conditioned growth supplement (which contains acidic fibroblast growth factor and heparin) have a proliferative response similar or higher than those grown in control media. They possess poorly organized cell–cell junctions with β-catenin, rather than plakoglobin, being the predominant molecule of AJs (protein levels and immunoreactivity at junctions). ECGS had little effect on transendothelial resistance (TER) but increased the macromolecular permeability of these monolayers to 20-kDa dextran. This strengthens the hypothesis that in systemic vessels it is the adherens junction, not TJs, which regulates paracellular permeability.

In primary cell cultures of HPMEC, plakoglobin can be targetted back to the AJs by growing the cells in the presence of cAMP-enhancing agents (Dye et al. 2001). In the latter study, cAMP-treated cells were quiescent and possessed extended paracellular junctions, resembling the placental endothelium in situ. cAMP led to up-regulation of junctional expression of VE-cadherin and increased the levels of the TJ molecules occludin and ZO-1. These cells showed a steady increase in TER over 48 h which was further increased to reach 1.7 times control TER after a second treatment given at 48 h (Dye et al. 2001). A small but significant decrease in macromolecular permeability was observed. These differences correlated with intensity of junctional phosphotyrosine, being lowest with cAMP treatment. The expression levels of junctional components and their tyrosine phosphorylation may play an important role in dynamic regulation of endothelial cell–cell junctions. Proliferation of HPMEC was inhibited by elevated cAMP. Thus the cAMP-treated HPMEC display a stable phenotype reminiscent of that seen in mature vessels of the human placenta.

cAMP signalling has been associated both with endothelial proliferation and differentiation (Davison & Karasek, 1981; Rubin et al. 1991; Satoh et al. 1996). Primary cultures of human placental microvascular cells show distinct long-term responses to elevation of cAMP. The mechanisms behind these changes are as yet unknown, but inhibition of myosin light-chain kinase (MLCK) may be mediated by cAMP (Garcia & Schaphorst, 1995; Alexander & Elrod, 2002). The striking effects on the TJ markers in HPMEC produced by cAMP suggest that cAMP increases the formation or stability of TJ. cAMP could increase the expression of TJ markers by stabilizing TJ complexes, since it can rapidly increase the assembly of TJ strands (Adamson et al. 1998). It is also possible that cAMP increases the synthesis of TJ molecules, as described for the TJ component 7H6 in lung endothelial cells (Satoh et al. 1996). Effects of cAMP may also result from inhibition of tyrosine kinase receptor signalling, or src family kinases associated with junctional complexes, rather than from cytoskeletal reorganization. Physiological activation of adenylate cyclase by agents which signal through G_s-coupled receptors (e.g. adenosine, serotonin, α-adrenergic agonists and PGI₂) can decrease endothelial permeability (Langeler & van Hinsbergh, 1991; Baluk & McDonald, 1994; van Hinsbergh & van Nieuw Amerongen, 2002).

In summary, the human placental vascular bed, with its capacity for continual growth and re-modelling and its site-specific exchange areas, provides a useful model to study cellular mechanisms involved in physiological regulation of permeability and angiogenesis. The differential expression of junctional adhesion molecules in the vascular tree and the capacity of these molecules to redistribute when insulted with exogenous agents suggest they play an important role in placental endothelial function. We hypothesize that the ‘activated’ junctional phenotype lacking plakoglobin, occludin and claudin-1 may be the preferred phenotype of new or continually re-modelling human blood vessels. This phenotype also predisposes vessels to be ones capable of maintaining maximal solute exchange.