The Starling equation represents the movements of fluid in and out capillary vessels. It depends on capillary filtration, hydrostatic, and oncotic pressure – i.e., Starling forces.

The Starling equation reads as follows:
where:

In conditions such as inflammation and elevated intracapillary pressure, the forces and membrane parameters governing transendothelial flux enhance filtration and increase the interstitial accumulation of albumin. The increased oncotic pressure in the neuroretina reduces fluid absorption and leads to retinal edema.

Barrier properties of retinal blood vessels and the RPE are due mainly to the presence of complex tight junction networks between cells. Tight junction and adherens junctions are integral membrane structures connected to the actin cytoskeleton via different adaptor molecules. Tight junctions are constituted by occludins, claudins (particularly claudin 5), and junction-associated molecules (JAM) connected to PDZ domain-containing proteins (among which is zonula occludens-1) and associated with an atypical protein kinase responsible for the tightly regulated phosphorylation of junction proteins (e.g., protein kinase C zeta, PKCζ). Junction proteins are transmembrane adhesive molecules closely linked to the cytoskeleton and with polarization proteins in the RPE. Tight junction destabilization can result from alteration of phosphorylation enzyme activity (e.g., PKCζ in diabetes), reduction of tight junction protein expression (e.g., occludin in diabetes), alteration of the cytoskeleton (e.g., secondary to oxidative damage or activation of RhoA/ROCK1 pathway), calcium dynamics [8], cell loss or severe cell damage (e.g., in case of severe inflammatory processes), and degradation of tight junction molecules by activation of proteases [9]. During inflammation, the exact molecular mechanisms that lead to tight junction disruption remain imperfectly understood. The cross talk between microglia and endothelial cells could contribute to tight junction expression regulation [10], while actin-binding molecules could also control vascular permeability via various signaling mechanisms such as activation of small GTPases [11]. Several extracellular signals could also intervene through signaling pathways leading to phosphorylation of actin and/or junction proteins, leading to their displacement from the membrane to other subcellular compartments.

Mechanical stress can also contribute to tight junction rupture as observed in the RPE submitted to chronic pressure secondary to vascular or melanocytic tumors in the choroid or to choroidal vasodilation in central serous chorioretinopathy.

Soluble mediators inducing vascular and/or RPE permeability include cytokines, such us monocyte chemoattractant protein-1 (MCP-1), tumor necrosis factor alpha (TNF-α), interleukins (IL-1b, IL-8, IL-6), vascular endothelial growth factor (VEGF) family members, acute phase proteins, enzymes, plasma activation systems (contact system, complement factor system, coagulation factors, fibrinolysis factors), arachidonic acid metabolites, biogenic or vasoactive amines (histamine, serotonin), eosinophil granular proteins, neuropeptides, oxygen free radicals, and nitric oxide.

Besides alteration of the tight and adherens junction complexes, other abnormal vascular changes can lead to increased fluid entry, visualized by “leakage” of dye during fluorescein angiography. This is the case for retinal neovascularization proliferating at the surface of the retina, with immature and low parietal stabilization, aneurysmal dilation of retinal capillaries (leaky microaneurysms in diabetic retinopathy), and vascular telangiectasia associated with intense protein leakage (as observed in Type 1 idiopathic macular telangiectasia and Coats’ disease). Factors potentially increasing the vascular permeability include lower pericyte coverage, hemodynamic changes with focal occlusions and secondary endothelial alterations, and elevation of the intravascular pressure.

Factors inducing vascular abnormalization include ischemia through hypoxia-inducible factor 1-alpha (HIF-1a), VEGFA and placental growth factor (PGF), and oxidative stress through advanced glycation end products (AGE). In certain disorders, such as Type 2 idiopathic macular telangiectasia, they remain unknown. The role of microglial cells and RMG cells is now considered as important players in the development of retinal vascular diseases [12].

RPE dysfunction can contribute to fluid entry from the choroid into the subretinal space. This enhanced fluid entry does not strictly belong to the classical vasogenic mechanisms. Indeed, the RPE transports water from the subretinal space into the choroid without rupture of the RPE tight junctions. The important RPE absorption capacity is particularly obvious in case of retinal detachment. RPE transport of Cl⁻ and K⁺ is thought to drive transepithelial water transport. But in physiologic basal conditions, the Cl⁻ conductance is up to 70% of the total basolateral conductance. The transport rate of water through RPE is estimated between 1.4 and 11 μl/cm²/h [13]. Fluid absorption involves complex mechanisms operating in the apical and basolateral membranes of the RPE cells that involve Cl⁻ transport, Na^+/K^+-ATPase activity, and Ca2⁺-activated, volume-activated, and/or cAMP-activated ion channels. These mechanisms are differentially regulated under light or dark conditions and are influenced by the circadian rhythm. Ion absorption in the RPE is accompanied by water transport through aquaporins [5, 14–16]. Calcium channels in the RPE were shown to regulate VEGF expression, suggesting a potential link between RPE ion transport and VEGF-induced permeability [17].

In pathological conditions such as diabetic retinopathy, changes in aquaporin expression were shown at the level of RPE [18].

Subretinal fluid accumulation resulting from alteration of fluid and ion transport across the RPE has also been suggested in central serous chorioretinopathy (Fig. 10a), but whether such changes per se are able to induce subretinal fluid in the absence of RPE barrier disruption has not been demonstrated.

Retinal vein occlusion leads to an increased intravascular pressure, blood-retinal barrier breakdown, and vascular leakage (Fig. 7a). Inner retinal hypoxia is associated with increased VEGF levels through HIF-1α, nitric oxide, and pro-inflammatory cytokines that contribute to the inner blood-retinal barrier rupture [25]. Hypertension, frequently associated with retinal vein occlusion, aggravates ME by further increasing the intracapillary hydrostatic pressure in the Starling equation. In addition, secondary hypoxic alterations of RMG cells may also lead to cytotoxic edema [26]. In cases of retinal vein occlusion with associated ischemia, excitotoxicity due to glutamate excess induces intracellular neuronal edema secondary to cellular energy failure [27]. Subretinal fluid is present in about half of central retinal vein occlusions and indicates that outer retinal barrier breakdown contributes to ME formation [28]. Indeed VEGF release also acts on the RPE barrier function through the VEGF receptor 1 (Flt-1), whose expression is under HIF-1α regulation [29].

The pathogenesis of DME is complex and multifactorial. Before any microangiopathy is clinically observed, intraretinal local inflammation (i.e., neuroinflammation) causes neuronal damage [30, 31]. Specifically, activation of microglial cells contributes to the local release of nitric oxide, TNF-α, interleukins, and VEGF [32]. In physiologic conditions, microglia trafficking contributes to retinal homeostasis. Active clearance of microglial cells through RPE transcytosis was demonstrated in the rodent retina, which prevents subretinal accumulation of activated cells. With aging, this active clearance increases in order to compensate for enhanced microglial activation to age-related debris, while it decreases in case of diabetic retinopathy as a consequence of alteration of cytoskeleton plasticity [32, 33]. Accumulation of microglia in the diabetic retina was also demonstrated to occur in humans [34, 35].