The role of plasmalemma vesicle-associated protein in pathological breakdown of blood–brain and blood–retinal barriers: potential novel therapeutic target for cerebral edema and diabetic macular edema

The formation and maintenance of a functional BBB and BRB is highly dependent on a complex interplay of multiple cell types, including endothelial cells, pericytes, astrocytes and neurons [3, 79, 80]. As such, the BBB and BRB is usually referred to as a function of the neurovascular unit. Nevertheless, endothelial cells form the direct physical barrier between the vascular lumen and the underlying tissue, and therefore their characteristic phenotype critically governs the restrictive nature of the BBB and BRB.

Endothelial cells of the BBB and BRB are characterized by the presence of an elaborate junctional network between the cells, a paucity and abluminal location of pinocytotic vesicles and absence of fenestrations [19, 20]. Under normal physiological conditions, these features of the BBB and BRB restrict the passage of potentially damaging molecules into the neuronal tissue, which is essential for proper function, in particular axonal electrical conductance [81]. However, the BBB and BRB are not impermeable barriers. There are two main routes via which molecules and solutes can cross the barrier, that is via paracellular and transcellular transport [82]. Paracellular transport occurs via the intercellular space between endothelial cells and is accomplished via dynamic regulation of the intercellular junctions such as tight junctions and adherens junctions. Transcellular transport refers to transport of molecules across endothelial cells. Transcellular transport is accomplished via transcellular diffusion, membrane transporters and vesicle-like structures termed caveolae. Outside the brain and eye, this latter type of transport occurs either via specific receptor-mediated transcytosis or non-specific fluid phase transcytosis. In the BBB and BRB, lipid-soluble molecules can passively diffuse across the barrier. In contrast, water-soluble molecules and solutes are actively transported across the barrier via receptor-mediated transcytosis via caveolae and solute transporters.

It is well established that both paracellular and transcellular transport plays a pivotal role in pathological BBB and BRB breakdown [3, 83]. Although the contribution of transcellular transport in vascular permeability has been greatly underestimated for years, this type of transport mechanism is considered to play a prominent role in edema formation. According to Starling’s rules, the net movement of fluid from capillaries is dependent on hydrostatic and osmotic pressure of the luminal and extraluminal compartments. Since macromolecules cannot easily diffuse across the membrane, we have previously hypothesized that transport of large molecules via the transcellular pathway is the main contributor to increased tissue osmotic pressure and subsequent edema formation in ocular conditions like DME [3].

VEGF has been shown to induce retinal permeability via caveolae-mediated transcytosis and not via paracellular transport [84]. In addition, intravitreal injections of VEGF in monkey eyes shifted the distribution of endothelial vesicles from an abluminal localization to a luminal localization, without clearly altering junctional integrity [85]. Moreover, caveolae are increased in numbers after VEGF stimulation of human retinal explants [28] and cultured bovine retinal endothelial cells [84]. Collectively, these findings strongly argue for a pivotal role of caveolae-mediated transcytosis in VEGF-induced BRB permeability. Although the endothelium of the BBB and BRB share highly similar characteristics, it has not been clearly established whether caveolae-mediated transcytosis also plays a central role in VEGF-induced BBB permeability. However, a recent study shows that ischemia-induced permeability in the brain is also associated with an increased number of caveolae in mouse endothelium [86], suggesting that similar mechanisms are involved.

In this review, the focus is on vasogenic edema formation which is hallmarked by a compromised barrier function of endothelial cells. However, it is important to note that neural edema can also develop as a result of aberrant ion transport while the barrier is intact. This type of edema formation, which is known as cytotoxic edema and/or ionic edema, has been extensively reviewed recently [87].

In non-barrier endothelium, PLVAP is expressed in fenestrated endothelium (with an exception of the so-called porous endothelia in the glomerulus and liver sinusoidal capillaries) and in the continuous endothelium of skin, muscle and lung (Table 1) [22, 27, 33, 57, 88]. The development of genetic knockout mice that lack PLVAP expression has provided insight how PLVAP modulates endothelial barrier integrity in non-barrier endothelium. Analysis of the protein content of blood plasma of PLVAP-knockout mice with a mixed BALB/c–C57Bl/6J–129Sv/J background showed that total plasma protein, albumin and albumin/globulin ratios are significantly reduced with minimal electrolyte imbalance in PLVAP-knockout mice compared to control littermates [57]. This reduction in blood plasma proteins was not caused by decreased protein production, enhanced catabolism or nephropathy. In contrast, significant leakage of plasma proteins as measured with the Evans Blue dye extravasation assay was observed in organs with fenestrated capillaries (intestine, kidney, pancreas), whereas only a minimal increase of leakage in organs with continuous endothelium (heart, muscle, lung) was detected [57]. Interestingly, vascular permeability was unaltered in the liver sinusoids and glomeruli of PLVAP-knockout mice, indicating that PLVAP deletion does not alter barrier integrity in the porous endothelial cells that are known to lack FDs [57]. Collectively, these findings suggest that hypoproteinemia and hypoalbuminemia in PLVAP-knockout mice are the result of increased vascular leakage in fenestrated endothelium [57].

Recently, these results in mice were supported by two case studies of infants who had a homozygous non-sense mutation in the PLVAP gene, and died at 5 months and 15 days of age [89, 90]. As a consequence of this mutation, PLVAP was targeted for degradation by non-sense-mediated mRNA decay, leading to complete absence of SDs and FDs in endothelial cells [89]. Comparable to the phenotype observed in PLVAP-knockout mice, these patients suffered from severe protein-losing enteropathy which is a pathological condition characterized by excessive loss of plasma proteins in the gastrointestinal tract, and eventually death [89, 90].

Together, these observations strongly indicate that lack of PLVAP expression leads to leakage in vessels with fenestrated endothelium, in particular of proteins. Based on the localization of PLVAP in diaphragms, it is tempting to speculate that PLVAP regulates vascular permeability by providing a size limitation or filter function to fenestrae and caveolae, allowing the passage of water and solutes but preventing passage of macromolecules (Fig. 2). The ultrastructural images of Bearer et al. [54] showed that the maximum arc length between the radial fibrils is 5.46 nm on average. However, the distance between the radial fibrils may be larger than the calculated average distance since the electron microscopic technique that was used shows a larger width of the radial fibrils of 1–2 nm beyond its real size due to metal shadowing [54]. According to the assumed wheel-like incorporation of PLVAP in diaphragms with a distance between adjacent radial fibrils at the rim of approximately 6 nm [54, 60] and the physiological upper limit pore size of fenestrated endothelium and the underlying lamina basalis [57, 91], it has been postulated that lack of PLVAP fibrils in diaphragms results in vascular leakage of plasma proteins with molecular diameters between 6 and 30 nm, which includes all plasma proteins except large protein complexes or lipoprotein particles such as chylomicrons [57].

Although it has been assumed that SDs and FDs have a similar architecture and composition, several studies indicate that the different types of diaphragms have differential biochemical properties. While SDs lack anionic sites, FDs are characterized by high numbers of anionic sites [92]. This difference in charge results in an extra layer of selectivity of permeability, presumably leading to impermeability for anionic (plasma) proteins at FDs, whereas SDs may allow the passage of anionic molecules [92]. These different biochemical characteristics may be related to the glycocalyx covering the luminal side of these diaphragms, as proteases and heparinase can remove the anionic sites on FDs [93]. On the other hand, despite the fact that it is well established that the endothelial cell surface is negatively charged, the content of anionic sites on FDs appears to be considerably higher compared to that of the adjacent plasma membrane [91]. Therefore, it is tempting to speculate that PLVAP binds to different glycoproteins in different vascular beds, which may alter the chemical composition of the endothelial opening and selectivity towards molecules. However, it is also possible that other proteins besides PLVAP are present in diaphragms and this may differ in SDs and FDs.

The lack of increased permeability in the continuous endothelium of the lung of PLVAP-knockout mice in the study of Stan et al. [57], where caveolae are normally covered with SDs, made the authors conclude that the absence of PLVAP does not directly induce leakage of plasma proteins via caveolae. The apparent lack of a significant role of SDs in regulating protein extravasation is consistent with our recent observations, which show that basal vascular permeability for proteins is similar in the continuous endothelium of the dorsal skin of heterozygous Plvap mice as compared to wild-type mice [Van der Wijk et al., submitted]. The difference in relevance of PLVAP in SDs and FDs may be explained by the morphological differences of caveolae and fenestrae. Lack of diaphragms in fenestrae results in pores that directly connect the capillary lumen and the underlying tissue without any permselective barrier, allowing diffusion of macromolecules out of the circulation. In contrast, lack of SDs may modulate the entry of macromolecules into vesicles, but vesicular trafficking is still dependent on other proteins which include, among others, caveolin-1, dynamins and SNARE proteins that regulate caveolae formation, scission and fusion, respectively [94‐96]. Alternatively, lack of FDs may considerably diminish the natural repellence against anionic proteins, which may explain why the leakage in fenestrated endothelium is so prominent. On the other hand, in vitro findings suggest that knockdown of PLVAP expression in human umbilical vein endothelial cells leads to a twofold increase in trans-endothelial electrical resistance and decreased permeability for both 70 kDa tracers and 766 Da tracers [28]. This indicates that lack of PLVAP in non-barrier endothelium may induce barrier-like properties, which is accompanied by a very low rate of vesicular transcytosis and may possibly counteract increased entry of macromolecules in vesicles when caveolae lack SDs. It is evident that PLVAP as a structural component of diaphragms has an important filter function. Therefore, we hypothesize that vesicular transcytosis in PLVAP-knockout mice is reduced to prevent aberrant transport of molecules across the barrier, and thus permeability does not increase when SDs are not present in continuous endothelium.

Besides acting as a physical sieve that controls transcytosis of molecules in mature endothelium, diaphragms may also give structural stability to caveolae and/or fenestrae during embryogenesis. In a C57BL/6N background, PLVAP-knockout mice die before birth as a consequence of subcutaneous edema and hemorrhages [55]. Transmission electron microscopy revealed that loss of SDs in subcutaneous capillaries of PLVAP-deficient embryos leads to vessels with large openings that are covered by degranulated thrombocytes [55]. This indicates that PLVAP may provide mechanical strength to the capillaries during embryogenesis, preventing vessels to become damaged during the dramatic remodeling events of angiogenesis and thus to prevent hemorrhages [55]. This function of PLVAP seems plausible as lack of FDs results in aberrant morphological phenotypes of fenestrae with widths of 50–120 nm in PLVAP-knockout mice and 20–400 nm in an in vitro assay instead of the typical 60–80 nm [57, 97], indicating that PLVAP may also give mechanical support to fenestrae. Alternatively, the lack of vessel wall integrity as observed in PLVAP-knockout mice may also be the consequence of impaired angiogenesis, as PLVAP plays an important role in this process [36].

It seems not very likely that PLVAP gives structural stability to mature endothelium, since there are also caveolae and fenestrae known to lack diaphragms in mature endothelium such as that of the kidney glomeruli and liver sinusoids [57]. PLVAP is present in these endothelia during embryogenesis [97, 98], which could be related to its function in mechanical stabilization. On the other hand, this may imply that controlled permeability is important during embryonic development.

Taken together, these results indicate that PLVAP is a crucial regulator of vascular permeability in non-barrier endothelium in both embryos and adults. During embryogenesis, PLVAP may directly promote angiogenesis or give mechanical support to capillaries which prevents the formation of leaky vessels after the dramatic remodeling steps during angiogenesis. In the mature vascular system, PLVAP has a crucial gatekeeping function, allowing the entry of small molecules but limiting leakage of large plasma proteins.

PLVAP expression is absent in mature endothelium with BBB and BRB properties (Table 2) [23, 24]. However, PLVAP is widely expressed in endothelia in the brain and retina during embryogenesis or postnatal development [30, 68, 99, 100, Van der Wijk et al., submitted], which is likely linked with an important function during vascular development (Fig. 3). PLVAP expression is negatively correlated with the maturation of the vasculature and acquisition of a functional BBB and BRB [30, 68, 99, 100]. Correspondingly, PLVAP expression in the BBB and BRB is induced in pathological conditions associated with vascular leakage such as diabetic retinopathy [25, 37, 75], brain ischemia [35, 36] and brain tumors [22, 23, 35]. Collectively, these observations suggest that absence of PLVAP expression in barrier endothelium is essential for the formation and maintenance of the BBB and BRB.

This is in agreement with the recent observation that suppressed transcellular transport is essential for the maturation of the BBB and BRB [101‐103]. While immature and leaky retinal vessels have a functional tight junctional network that is similar to that of mature endothelium, a significant reduction in the number of vesicles governs maturation of the endothelium to form a functional BRB [101]. Further research is needed to establish the exact function of PLVAP in vascular development.

In contrast, it is relatively well-established that PLVAP plays a functional role in BBB and BRB breakdown in pathological conditions (Fig. 4). With the use of short hairpin RNA (shRNA)-mediated knockdown of PLVAP expression, it was shown that PLVAP inhibition reduces VEGF-induced loss of BRB integrity as shown in a trans-endothelial electrical resistance assay [28]. Moreover, inhibition of PLVAP expression significantly reduced VEGF-induced permeability of 70 kDa tracers but not of 766 Da tracers in an in vitro model of retinal leakage [28, 104]. Correspondingly, siRNA-targeting of PLVAP expression reduced vascular permeability in the oxygen-induced retinopathy (OIR) mouse model as shown by decreased extravasation of fluorescent tracers (70 kDa) and fluorescein angiography [28]. Large molecules (70 kDa) are expected to cross the endothelial barrier via transcellular transport, whereas small molecules (766 Da) cross the barrier via the paracellular pathway. Thus, these findings suggest that (VEGF-induced) leakage via the transcellular pathway is dependent on PLVAP expression. Correspondingly, shRNA-mediated knockdown of PLVAP expression did not prevent VEGF-induced alterations in endothelial junction integrity [28]. However, stress fiber formation was significantly reduced. Ultrastructural analysis revealed that knockdown of PLVAP expression in human retinal explants blocks VEGF-induced caveolae formation to basal levels [28], suggesting that PLVAP induces leakage by enabling formation of caveolae, probably with PLVAP-containing SDs. The observation that inhibition of PLVAP expression does not modulate basal caveolae numbers is in agreement with the study of Herrnberger et al. [55], that did not observe a reduction in the number of caveolae in endothelial cells of Plvap-deficient mice. As previously emphasized, vesicular trafficking is a complex process that requires the need of multiple proteins and molecules [94‐96], indicating that increased numbers of caveolae do not necessarily implicate increased transcytosis. However, VEGF has been shown to induce the expression of several important genes involved in the transcellular transport pathway [62] and promotes caveolae-mediated transcytosis in vivo [85], which strongly indicates that this argument is invalid during pathological vasogenic edema formation, as long as PLVAP is also expressed.

The exact molecular mechanisms by which PLVAP modulates caveolae formation or how the presence of caveolae with PLVAP-containing SDs allows barrier loss remain still unknown. Previous research indicates that VEGFR2 is localized in caveolae [77, 78]. Thus, PLVAP may modulate vascular permeability and caveolae formation via regulating the availability of VEGFR2 in cells, as a primary or secondary mechanism. Interestingly, PLVAP has been shown to interact with NRP-1, which is an important co-receptor in VEGFR2 signaling regulating VEGFR2 surface expression [45, 105]. An intriguing possibility is that PLVAP affects the stability of VEGFR2/NRP-1 complexes, and thus VEGFR2-mediated signaling. At the same time, PLVAP expression is induced in a VEGFR2-dependent manner [61]. Thus, this opens the interesting possibility of a positive feedback loop that promotes breakdown of the BBB and BRB. Further research is needed to elucidate how PLVAP regulates VEGF-dependent caveolae formation and increased permeability.

Although we have considered paracellular and transcellular transport as two independent transport mechanisms in this review so far, it is becoming clear that complex interdependency between these two pathways exist [82, 106]. In line with this insight, despite the direct association of PLVAP with transcellular transport, PLVAP appears to modulate paracellular transport as well. A direct indication that PLVAP plays a functional role in the paracellular pathway comes from the observation that PLVAP is able to regulate the number of fenestrae and caveolae in cells [28, 55, 58]. Although the exact molecular events behind the biogenesis of these structures is not known, it is evident that it requires reorganization of the cytoskeletal framework [97]. It would be interesting to determine whether PLVAP, with its intracellular connections to the cytoskeleton, affects the dynamic nature of the cytoskeleton, and subsequently intracellular gap formation. Interestingly, previous work has shown that inhibition of PLVAP expression in bovine retinal endothelial cells significantly reduced stress fiber formation after VEGF stimulation, indicating that PLVAP may regulate intracellular gap formation [28]. Although this function of PLVAP may operate via its suspected link to actin-binding proteins, it should be noted that knockdown of PLVAP expression had only limited effects on stress fiber formation in unstimulated cells [28]. This indicates that VEGF plays a central role in this process and that the role of PLVAP is indirect. A striking phenotype observed in these cells was that knockdown of PLVAP expression significantly blocked VEGF-induced caveolae formation, which led to the hypothesis that PLVAP regulates VEGFR2 availability [28]. Interestingly, VEGF-A has been shown to induce activation of RhoA, which is well known for its function in stress fiber formation and cellular contractility [107, 108]. Hence, PLVAP may modulates stress fiber formation via induction of VEGFR2-dependent Rho GTPase signaling in cells, which is an interesting avenue to be explored in future research. Moreover, in a recent study we have found that expression of tight junctions and adherens junctions is altered in the retina of heterozygous Plvap mice [Van der Wijk et al., submitted]. Although it is not known how this altered expression is translated into functional junctional characteristics, it indicates that PLVAP affects paracellular transport as well.

In short, PLVAP plays an important functional role in VEGF-induced BBB and BRB breakdown by regulating the number of caveolae with PLVAP-containing SDs. Although the sieving function of PLVAP during macromolecular transport in fenestrated endothelium is well established, PLVAP regulates vascular permeability in multiple and complex additional ways that remain incompletely understood.