In the classic Starling’s principle, the net fluid flux across the inter-endothelial junction is determined by the balance between hydrostatic and oncotic pressure gradients [7]. High hydrostatic pressure in the vascular lumen drives fluid to the interstitial space, whereas lower levels of proteins in the interstitial space compared to the plasma results in opposing forces for hydrostatic filtration. This can be expressed by the following equation:

where J_v/A is the filtration rate per area, L_p is the hydraulic conductance, P_c is the intravascular hydrostatic pressure, P_i is the interstitial hydrostatic pressure, σ is the reflection coefficient, π_c is the plasma oncotic pressure, and π_i is the interstitial oncotic pressure.

However, several quantitative studies revealed that actual fluid filtration differed from the amount predicted by the classic Starling’s principle. As hydrostatic pressure gradually decreases from the arteriolar to venular ends of the capillary, it was assumed that fluid is filtered to the interstitium along the arteriolar side of the capillaries and then reabsorbed at the venular side. This model has been refuted in later studies [8,9]. Revisions to Starling’s principle has been made, suggesting that the return of fluids to the circulation occurs exclusively via lymphatic drainage. In contrast to earlier experiments, which were conducted in models where the interstitial oncotic pressure was near zero [10], Levick [11] used accurate measures of interstitial oncotic pressure, which is almost half the plasma, and showed that the net filtration pressure predicted by the classic Starling’s equation was in far excess of the observed lymphatic flow. Moreover, when the interstitial oncotic pressure was raised, the increase in net fluid filtration was much smaller than expected [8]. These discrepancies indicated that interstitial protein concentration plays a minor role in the generation of the oncotic pressure gradient across the capillary wall. The introduction of the glycocalyx-cleft model helped to resolve the paradox, and Starling’s principle was revised to the following:

where π_sg is the subglycocalyx oncotic pressure.

The EG acts as a molecular sieve of plasma proteins [9,12], and filtered fluid accelerates through the narrow junctional breaks. If this downstream fluid velocity approaches the upstream diffusion velocity of interstitial proteins, a washout of the diffusing proteins can maintain the protein concentrations in the sub-glycocalyx area at a very low level. Consequently, the effective oncotic pressure gradient opposing hydrostatic filtration is greater than the difference in oncotic pressure between the plasma and interstitium, reducing the basal fluid filtration rate to approximately the normal lymphatic drainage rate [13]. Enzymatic removal of EG components results in increased hydraulic conductivity, protein flux, albumin excretion, and edema [14–16]. When the negative charge of the EG was removed, increased permeability to albumin or dextrans were observed [17,18]. This evidence indicates that EG is a major determinant of vascular permeability.

In addition to a passive barrier role, recent studies suggested that the EG may regulate vascular permeability by acting as a mechanotransducer of flow-mediated shear stress [19]. According to Starling’s principal, hydrostatic pressure and fluid flux should have a linear relationship if L_p is held constant. However, it has been shown that increases in hydrostatic pressure results in an increase in L_p, thus resulting in a non-linear elevation of net fluid flux [20–23]. The shear-induced increase in L_p involves the activation of endothelial nitric oxide synthase (eNOS) and resultant alterations of junctional proteins [24]. The heparanase treatment of bovine endothelial cells abolished the shear-induced increase in L_p [23,25], suggesting a non-Starling mechanistic role of EG in the regulation of vascular permeability.

One of the main roles of EG is represented by the mechanotransduction of blood flow-mediated shear stress to the cytoskeleton, nucleus, and enzymatic reactions, which is essential for vascular homeostasis. The EG functions as a sensor of mechanical forces exerted on the endothelial surface [26], in concert with other sensors including G-protein-coupled receptors [27], stretch-sensitive ion channels [28], rheological properties of the plasma membrane [29,30], caveolar structures [31], and integrins and focal adhesion molecules [32]. These mechanosensors initiate intracellular signaling such as increases in cytosolic calcium, the activation of eNOS [33], Rho-family guanosine triphosphatases, tyrosine kinases, Ras, Erk, and JNK; the phosphorylation of platelet endothelial cell adhesion molecule-1 (PECAM-1) intracellular domain [34]; and cytoskeletal remodeling [35], especially at the sites of highest tension like cell-cell junctions [36]. Physiologically, shear stress on the endothelium activates eNOS via the phosphorylation of S617, S635, S1177/S1179 residues by calmodulin kinase, PI3K/Akt, and PKA pathways [37]. In addition, shear stress also induces sirtuin 1, a nicotinamide adenine dinucleotide (NAD)-dependent deacetylase, which enhances eNOS activity through deacetylation at Lys496 and Lys506 [38]. Physiological shear stress also induces transcription factors such as Kruppel-like factor-2 and nuclear factor-like-2, both of which are responsible for the induction of eNOS and antioxidant defense [39]. In contrast, in endothelial cells, perturbed blood flow induces the activation of the nuclear factor kappa B (NF-κB) pathway and activator protein-1, which participate in the acquisition of a pro-inflammatory phenotype by endothelial cells [40,41]. Defective mechanosensing and the resulting reduction in the shear-induced release of nitric oxide (NO) is associated with decreased eNOS activity. Direct pulling of the EG components and core proteins using atomic force microscopy results in a rapid Ca²⁺ influx and NO production [33,42], whereas enzymatic removal of EG [43], inhibition or knockdown of specific transient receptor potential (TRP) channels [44,45], or disruption of caveolae [46], all of which result in the impairment of flow-induced NO production.

NO has a crucial role in vascular homeostasis including the regulation of vascular tone, permeability, and inflammatory phenotype [47]. Decreased bioavailability and uncoupling of eNOS are associated with increased oxidative and nitrosative stress as well as endothelial dysfunction [48]. Of note, endothelial dysfunction induced by the disintegration of the EG and defective mechanotransduction may accelerate the further loss of the EG, leading to a vicious cycle of endothelial dysfunction. Decreased bioavailability of NO can increase exocytosis of lysosome-related organelles, which are responsible for the intrinsic mechanism of the EG degradation, via reduced S-nitrosylation of the N-ethylmaleimide-sensitive factor [49]. Furthermore, sirtuin 1 deficiency also affects vascular homeostasis. Recently, we reported that endothelium-specific sirtuin 1 knock-out mice have a significantly reduced whole body EG volume compared with wild-type mice [50]. Sirtuin 1-mediated deacetylation of Forkhead box O (FoxO1) DNA-binding protein is required for its active conformational change and enhanced cellular protection from oxidative stress [51,52]. Deficient sirtuin 1 expression or activity, which is associated with diverse vasculopathies [53], leads to susceptibility to oxidative stress. Oxidative stress activates disintegrin and metalloproteinase domain-containing protein 17 (ADAM-17) [54]. Activation of this enzyme contributes to cleavage of syndecan, which results in the shedding of EG [55]. Additionally, sirtuin 1 deficiency activates NF-κB, which induces transcriptional activation of heparanase, one of the target genes of this transcription factor [56,57]. Subsequently, increased heparanase activity leads to the enzymatic removal of heparan sulfate side chains from the EG and disintegration of ESL.

The EG shields the endothelium from interactions with circulating blood cells. The length of projections of cell adhesion molecules, including selectins (PECAM, vascular cell adhesion molecules (VCAMs), and ICAMs (intercellular adhesion molecules)) and integrins (CD11/CD18), are shorter than the thickness of the EG [58]. Thus, intact EG inhibits firm adhesion of leukocytes and platelets, whereas shedding of the EG facilitates their adhesion. In rodent cremaster muscle preparation, degradation of the EG by oxidized lipoproteins or tumor necrosis factor alpha (TNF-α) increased platelet or leukocyte-endothelial adhesion [3,59]. Similarly, chemotactic peptide fMLP (formyl-Met-Leu-Phe) degraded the EG and increased adhesion of leukocytes, while the matrix metalloprotease (MMP) inhibitor doxycycline attenuated the degradation of the EG and reversed the facilitated adhesion of leukocytes [4,60]. Knockout of syndecan-1 in mice resulted in increased adhesion and transmigration of leukocytes under pro-inflammatory stimuli as well as basal conditions [61].

Hypervolemia can result in the shedding of the EG, which is induced by the release of ANP due to atrial stretch. When colloid solutions are administered to normovolemic patients, 60% of the infused solution was extravasated immediately, whereas almost the entire volume was maintained within the intravascular space in an acute normovolemic hemodilution setting [112,113]. The prophylactic administration of fluids upon the induction of general or neuraxial anesthesia has long been advocated to counteract hypotension due to an anesthesia-induced decrease in cardiac preload. However, this practice cannot be recommended considering the role of the EG in fluid management. Furthermore, preloading of fluids did not show consistent efficacy for decreasing the incidence of hypotension or requirement for vasopressors after neuraxial anesthesia [114,115]. Moreover, liberal perioperative fluid administration leading to a positive fluid balance has been associated with increased morbidity [116]. Perioperative degradation of the EG may provide one rationale for restrictive fluid administration by a goal-directed protocol; however, the clinical impact of liberal versus restrictive fluids needs more extensive discussion [117] and is beyond the scope of this review.

The EG can be stabilized by supplementation with albumin within the ESL [63]. The intercalation of albumin can provide protection against oxidative damage owing to an oxidizable sulfhydryl group of albumin, excessive attachment of leukocytes and platelets, and maintenance of shear-induced vasodilatation [63,118]. In addition, albumin carries sphingosine-1-phosphate, which inhibits MMP activity on the endothelium [119]. In an animal model for heart transplantations, the addition of histidine-tryptophan-ketoglutarate solution, which contains albumin reduced EG degradation [120]. In an experimental model of massive hemorrhage, resuscitation with plasma showed partial restoration of the EG, whereas lactated Ringer resuscitation did not [110]. These data suggest a possible role of albumin in the protection of the EG, although there is a paucity of direct clinical evidence.

A variety of pharmacological agents has been used to restore or prevent the degradation of the EG (Fig. 1). Agents that potentially interfere with EG degradation include antioxidants [121], antithrombin III [122], MMP inhibitor doxycycline [67], TNF-α analogue etanercept [123], NO donors [124], hydrocortisone [125], and volatile anesthetics [126,127]. However, these promising experimental study results have not yet been translated into clinical practice. Other methods to prevent EG degradation involve the supplementation of EG constituents such as heparan sulfate or hyaluronic acid [128]. Although accelerated restoration of the EG was achieved in vitro, heparan sulfate and hyaluronic acid could not reconstitute EG in vivo in a sepsis model [129]. The lack of effect in vivo can be explained by the overwhelming degradation process, which cannot be suppressed by the supplementation of EG components. Interestingly, we have reported a remarkable effect with sulodexide, an 8 : 2 mixture of fast-moving heparin and dermatan sulfate, on the restoration of the EG and survival benefit in a mouse model of severe sepsis [129]. The superior efficacy of sulodexide is thought to be due to its dual action, supplementation of EG constituents, and concomitant inhibition of degrading enzymes. Fast-moving heparin components of sulodexide, as a close mimic of heparan sulfate, are responsible for heparanase inhibition at basic clusters of heparanase in a competitive manner [130]. The dermatan sulfate fraction interacts with the active zinc binding site of the pro-MMP-9 molecule to block its conformational change into an active form [131]. It has also been reported that sulodexide attenuates the release of MMP-9 from leukocytes [132]. Clinically, Broekhuizen et al. [133] demonstrated that oral sulodexide administration for two months partially restored the thickness of the EG in patients with type 2 diabetes using sidestream dark field imaging of sublingual microcirculation. It is uncertain whether sulodexide would show a comparable effect in the perioperative setting. The required dose of sulodexide was estimated to be much higher to reconstitute the EG in a severe sepsis model [129]. Although sulodexide is known to exert minimal in vivo anti-coagulant effects due to its preferential uptake by the endothelium [134], care should be taken for its application in the perioperative setting because of a potential bleeding tendency from the administration of large doses. From this point of view, chemically modified, non-anticoagulant variants of heparin could be fascinating candidates for the protection of EG [135–137].