Role of the endothelial surface layer in neutrophil recruitment

Review of how ESL has pro-adhesive activating and anti-adhesive effects on leukocyte adhesion.

Abstract

Neutrophil recruitment in most tissues is limited to postcapillary venules, where E- and P-selectins are inducibly expressed by venular endothelial cells. These molecules support neutrophil rolling via binding of PSGL-1 and other ligands on neutrophils. Selectins extend ≤38 nm above the endothelial plasma membrane, and PSGL-1 extends to 50 nm above the neutrophil plasma membrane. However, endothelial cells are covered with an ESL composed of glycosaminoglycans that is ≥500 nm thick and has measurable resistance against compression. The neutrophil surface is also covered with a surface layer. These surface layers would be expected to completely shield adhesion molecules; thus, neutrophils should not be able to roll and adhere. However, in the cremaster muscle and in many other models investigated using intravital microscopy, neutrophils clearly roll, and their rolling is easily and quickly induced. This conundrum was thought to be resolved by the observation that the induction of selectins is accompanied by ESL shedding; however, ESL shedding only partially reduces the ESL thickness (to 200 nm) and thus is insufficient to expose adhesion molecules. In addition to its antiadhesive functions, the ESL also presents neutrophil arrest-inducing chemokines. ESL heparan sulfate can also bind L-selectin expressed by the neutrophils, which contributes to rolling and arrest. We conclude that ESL has both proadhesive and antiadhesive functions. However, most previous studies considered either only the proadhesive or only the antiadhesive effects of the ESL. An integrated model for the role of the ESL in neutrophil rolling, arrest, and transmigration is needed.

Introduction

Several in-depth reviews have provided a detailed view of the ESL [1–3]. The present review focused on the aspects of ESL structure and function related to neutrophil adhesion and recruitment.

MOLECULAR COMPOSITION, THICKNESS, AND MECHANICAL PROPERTIES OF THE ESL

Composition

Syndecans and glypicans are directly membrane-bound proteoglycans [4] that each carry multiple heparan sulfate chains. Syndecans can also carry chondroitin sulfate chains. In addition to these sulfated GAGs, the ESL contains hyaluronan, which is anchored by both CD44 glycoprotein and the hyaluronic acid synthase enzyme. The sulfate groups of heparan sulfate and chondroitin sulfate and the carboxyl groups of the uronic acids in the sulfated GAGs and hyaluronan provide a negative charge on the cell surface. Endothelial cells mainly express 3 types of syndecans, syndecan-1, -2, and -4; synovial and brain endothelial cells also express syndecan-3 [5, 6] and only 1 member of the glypican family, glypican-1. The extracellular domain of syndecan-1, the largest syndecan proteoglycan core protein, is 232 amino acids. In contrast, the glycpican-1 core protein is 502 amino acids long. The syndecan core proteins are believed to be extended polypeptides; thus, their maximum length is limited to <100 nm. The glypicans have a more complex tertiary structure and thus have a more compact structure [7]. Heparan and chondroitin sulfate chains are ∼100 disaccharides long and have an ∼80 nm contour length [8]. Hyaluronic acid molecules can be up to 10 μm long [9] but are typically coiled up.

In mouse mesentery and glomerular capillaries, ESL glypicans were labeled with fluorescently labeled WGA lectin and imaged using confocal microscopy [10], showing an ∼500-nm-thick surface layer. This result is difficult to interpret, because WGA binds to N-acetylglucosamine or sialic acid on glycoproteins and not to hyaluronan. This was clearly shown in experiments with cultured chondrocytes, in which WGA-AF555 labeled the chondrocyte cell membrane but did not label the thick hyaluronan-rich pericellular coat [11]. Nevertheless, several in vivo studies applying nonchemical ESL detection methods have shown a thickness of ∼500 nm for ESL (see Thickness and mechanical properties of ESL). Considering that the largest membrane-bound proteoglycans are ∼100 nm tall, it is reasonable to divide the ESL into a 100-nm-thick foundation near the endothelial plasma membrane (also called the endothelial glycocalyx), which is dominated by membrane-bound proteoglycans, and into a 400-nm-thick adsorbed superficial layer (Fig. 1) composed of hyaluronic acid, secreted proteoglycans (e.g., syndecan-1 [12], versican [13], serglycin [14], perlecan [15], agrin [15], biglycan [13]), and bound plasma proteins (e.g., albumin, fibrinogen, blood coagulation factors, enzymes) [1, 2].

Figure 1. Neutrophil rolling on endothelial cells.

(A) Under nonstimulated conditions (left) endothelial cells present an even cell surface coated with complete, 500-nm-thick ESL (green field above the endothelial surface), of which the main components are the membrane bound and soluble proteoglycans (their core proteins are indicated with green lines). ESL anchors chemokines (blue symbols). Neutrophils show 300-nm-tall microvilli coated with NSL (green-blue above the neutrophil surface), adhesion molecules (red symbols), and chemokine receptors (blue curvy symbols). ESL and NSL are negatively charged; thus, charge repulsion keeps these surfaces apart. On inflammatory stimuli (middle), endothelial cells present selectin and immunoglobulin superfamily adhesion molecules (red symbols on the endothelial cell) on their luminal surface, project microvilli into the vessel lumen, and partially shed the ESL into the vessel lumen (residual thickness 200 nm). The shed ESL is still about 4 times taller than the adhesion molecules, suggesting that the ESL can shield these adhesion molecules even after shedding. A situation similar to the one indicated by the dashed rectangle is shown enlarged on panel B. During rolling (right), adhesion molecules engage, and the NSL and ESL likely invade each other. Because both are negatively charged, this comes at an energy cost and causes repulsion. ESL-bound chemokines can likely bind the leukocyte chemokine receptor only during rolling. Scale bar, 500 nm. (B) Molecular resolution view. Neutrophil surface and stimulated endothelial surface under inflammatory conditions (partial shedding of ESL) at higher magnification. The various adhesion molecules (P-selectin, E-selectin, ICAM-1, VCAM-1 on the endothelial cell and PSGL-1, L-selectin, β₂-integrin on the neutrophil, red to purple) do not reach out further than 50 nm from the cell surface (layer 1). The IL-8 receptors CXCR1 and CXCR2 are heptahelical transmembrane receptors (blue) shown on the neutrophil. The next layer (layer 2) contains directly membrane-bound proteoglycans (syndecan-1 [SDC-1] and -4 [SDC-4] and glypican-1 [GPC-1] are shown as examples; green) and hyaluronan (yellow) anchored by hyaluronan synthase or CD44 (shades of yellow). The endothelial cell surface is further covered by secreted proteoglycans (agrin and serglycin are shown as examples; green) or by shed membrane-bound proteoglycans (shed SDC-4 is shown as an example) (layer 3), which accounts for the outer 100 nm of partially shed ESL. These proteoglycans are decorated mainly by heparan sulfate (green) or chondroitin sulfate (light green). The ESL may contain more heparan sulfate than chondroitin sulfate and the neutrophil surface layer more chondroitin sulfate than heparan sulfate. The ESL is known to harbor bound chemokines (e.g., IL-8; blue) and various plasma proteins (gray). Scale bar, 100 nm.

The ESL is understood to be a highly dynamic structure. Its composition and physical properties are determined by the balance between ESL synthesis and shedding. Both of these processes are regulated by the endothelial cell and the blood plasma via hydrolytic enzymes and oxidative stress. A recent study found that cultured HUVECs retain only 10% of the proteoglycans on the cell surface and secrete the rest into the culture medium [15]. Serum albumin is known to be essential in maintaining ESL properties [16, 17], and serum GAGs (chondroitin sulfate, hyaluronan) were shown to help ESL recovery [18, 19]. The ESL itself harbors several enzymes, including superoxide dismutase, which might protect the ESL from oxidative stress, and others that facilitate ESL shedding (e.g., matrix metalloproteinases, which can cleave proteins and some proteoglycans).

The GAG components of the ESL are known to bind and immobilize the most important chemokines for neutrophil arrest and transmigration, CXCL1, 2, and 8 [20, 21]. These chemokines are bound to GAGs via electrostatic interactions between the GAG’s negatively charged residues and the chemokine’s positively charged arginine and lysine residues. The chemokine docking site on the GAG chain is formed by a specific arrangement of sulfated saccharides [22–24]. In vivo, ESL-bound chemokines can be mobilized by injecting soluble heparin, which displaces ESL-bound chemokines and makes them appear in the blood stream and disappear from the endothelial surface [25].

Thickness and mechanical properties of ESL

Most studies have suggested the presence of ∼500-nm-thick ESL in the microvessels of various tissues. The ESL, despite its porous structure, blocks plasma flow [26], excluding fluorescent dextran. The ESL also prevents flowing RBCs from approaching the endothelial wall.

Various techniques were applied to estimate the ESL thickness in different microvascular beds. Microhemodynamic measurements have shown that intraluminal heparinase digestion of the hamster cremaster microvascular network increases the amount of RBCs within capillaries without changes in the apparent vascular diameter [27]. These early experiments suggested the presence of a heparan sulfate-rich transparent layer on the vessel wall, which is stiff enough to limit the approach of RBCs to the vessel wall. The first definitive measurement of ESL thickness was reported in hamster mesenterial microvessels [28]. Fluorescent high-molecular-weight dextran was excluded from a 500-nm-wide zone above the endothelium, suggesting a 500-nm-thick ESL. During this measurement, the vessel diameter, measured using bright-field microscopy, was compared with the diameter indicated by fluorescent dextran. The bright-field diameter was consistently larger than the area accessible to the fluorescent dextran, and the difference between the 2 values was interpreted as evidence that these large molecules cannot penetrate the ESL. In the rat mesenteric microvasculature, heparinase treatment was found to leave the vessel diameter unchanged but to reduce the network’s resistance to plasma flow [29]. The extent of the resistance reduction suggested the presence of a 0.5-µm-thick heparan sulfate-rich ESL impermeable to plasma flow. In mouse cremaster venules and arterioles, microparticle image velocimetry indicated a 500-nm-thick, nonmoving fluid layer near the endothelium. During this measurement, the flow profile within a microvessel was reconstructed from the convective velocity and the radial position of the perfused fluorescent tracer particles. The velocity of the near-wall microbeads was much lower than it should be in the absence of an ESL, showing that essentially no or very little flow is present in the first 500 nm from the endothelial surface [30]. In mouse lung microvasculature, dextran exclusion suggests that the ESL might be 1.5 μm thick [31]. However, lung capillaries are not straight, and their cross-section is not circular; thus, it is difficult to determine the location of the vessel wall border.

Some evidence suggests that the ESL could be thicker in large arteries. The ESL was reported to be 2.1–2.5 μm thick using confocal microscopy in isolated, fixed, rat and mouse aorta labeled with an antibody to heparan sulfate [32]. A 4.5-μm-thick ESL was detected using 2-photon microscopy in isolated, live WGA-lectin-labeled mouse carotid artery [33]. Some experiments conducted on cultured endothelial cells of different origin also showed a few-micrometer-thick ESL. A rapid freezing/freeze substitution EM technique showed an 11-μm-thick ESL on cultured bovine aortic endothelial cells and an ∼5-μm-thick ESL on a cultured rat fat pad endothelial cell line [34]. In contrast, in the same study, conventional EM indicated an ESL thinner than 100 nm. In another study using fluorescent correlation spectroscopy, fluorescently labeled albumin was found to accumulate 1–2 μm above the surface of cultured bovine lung microvascular endothelial cells. Confocal microscopy images of fixed, anti-heparan sulfate antibody, or hyaluronan-binding protein-labeled endothelial samples corroborated this result, revealing a 2- to 3-μm-thick ESL [35]. Another study found an ∼2-μm-thick ESL using confocal microscopy on a fixed and antibody- or lectin-labeled cultured rat fat pad endothelial cell line [36].

Several factors complicate the interpretation of these in vitro data. Several studies have shown differences between ESL in vivo and in vitro [19, 37]. The reason for this difference is unknown, but obvious possibilities include the lack of shear stress [4, 38] and surrounding cells [39] (e.g., pericytes or smooth muscle cells) and a fundamental difference in the composition of the cell-culturing medium, which often contains serum. Also, the EM results are strongly dependent on the fixation technique [40]. Confocal microscopy is potentially biased because of fixation artifacts. Also, it is hampered by a low resolution. The current evidence supports the notion that the ESL is about 500 nm thick in most vessels.

The ESL has been modeled as a porous layer with regular pore sizes on the basis of findings from tracer exclusion studies, which showed that smaller molecules penetrated the ESL faster than did larger ones and in which a cutoff value in the size of the molecules invading the layer was observed [18, 41–43]. The ESL also restricts RBCs from approaching the endothelial membrane. However, the interpretation of this result is not straightforward, because RBCs are pushed into the centerline by their interaction with the flowing plasma, even in the absence of an ESL. A significant “plasma gap” between the wall and RBCs is observed in glass capillaries with no ESL. Some experiments have suggested that in the case of a live capillary, no convective liquid flow is present in the gap between the RBC and the vessel wall [29, 30]. However, these observations are not plausible from a fluid mechanics viewpoint. Most likely, the apparent distance between the RBCs and the endothelial wall in vivo is generated by a combination of the hydrodynamic plasma gap and the ESL. This has been corroborated by the observation that the RBC-free gap is smaller in glass capillaries than in living capillaries of the same diameter [28]. Also, it is supported by a recent study, in which flowing RBCs were found to attain larger deformation in glass capillaries coated with a 350-nm-thick, brush-like, highly hydrated layer of neutrally charged polymers than in bare capillaries of the same size [44].

In a modeling study, RBCs were proposed to glide on top of the ESL, similar to how snowboarders glide on powder snow. This presumption is based on the resistance to compression resulting from the slow rate of squeezing liquid out of the porous ESL after rapid compression by RBCs [45]. This model is supported by the observation that on flow cessation, RBCs can eventually compress the ESL and can occupy the entire capillary lumen [28]. In capillaries, the ESL is not stiff enough to exclude leukocytes. In hamster mesentery capillaries, the diameter of the capillary accessible to high-molecular-weight dextran increased after passage of a leukocyte. In these experiments, the apparent vessel diameter remained the same, suggesting that leukocytes did not distend the vessel wall [28]. In hamster cremaster capillaries, the ESL compressed by leukocyte passage was found to regain its original thickness in 0.5–1 s [46]. After passage of leukocytes, a broader RBC column was observed in human sublingual capillaries, suggesting that the leukocytes locally compressed the ESL [47].

Although these measurements elegantly demonstrated the mechanical resistance of the ESL, they do not allow quantification of ESL rigidity. An estimate of ESL rigidity was provided by in vitro AFM measurements [48]. Probing the surface of cultured endothelial cells showed an ESL that was ∼400 nm thick and had a stiffness of ∼0.3 pN/nm, where pN indicates the force bending the AFM cantilever and nm indicates the indentation depth [48, 49]. Similarly, the ESL of a freshly harvested, cut open mouse aorta was found to be as thick and stiff as the ESL on cultured endothelial cells [48]. However, other studies suggest that the ESL differs in vivo and in vitro [19, 37].

The stiffness of the ESL is hypothesized to be maintained by the stiffness of the ESL core proteins [50] and GAGs; the elevated oncotic pressure inside the ESL, which is presumed to be provided by accumulation of albumin in the layer [51]; and the charge repulsion inside the layer, which is provided by the negatively charged GAGs of the layer [1, 3]. It is not clear whether a single component is most responsible for the stiffness. Recently, the hydrodynamically relevant ESL thickness in the cremaster vasculature was compared between wild-type and syndecan-1 knockout mice, and no difference was found [52]. Heparinase (degrades heparan sulfate) or hyaluronidase (degrades hyaluronan) treatment both reduced (but did not eliminate) the ESL thickness and stiffness detectable with AFM, and digestion with only heparinase or hyaluronidase was not as effective as digestion with both enzymes [48, 49, 53]. Although commercial heparanase preparations do not degrade cell surface syndecans [54], commercial preparations of heparinase and hyaluronidase often have significant protease activity.

A recent study has suggested that albumin regulates ESL thickness not directly but by delivering S1P to the endothelial cells [55]. Incubation of cultured endothelial cells with fetal calf serum-free medium reduced the heparan and chondroitin sulfate levels detectable on the cell surface using immunofluorescence microscopy. The same ESL loss was detected after incubation with full medium supplemented with an inhibitor of S1P1, one of the endothelial receptors for S1P. Supplementing the serum-free media with BSA prevented ESL loss. However, supplementation with a more pure, fatty acid-free bovine serum albumin was significantly less protective. In the absence of serum and BSA, a lower amount of S1P was found in the supernatant, and it was accompanied by lower endothelial S1P1 receptor activation and increased supernatant GAG and syndecan ectodomain levels and supernatant MMP activity. All these effects were preventable with addition of exogenous S1P. Supplementation of MMP9 or MMP13 inhibitors reduced the MMP activity and the amount of shed ESL elements. These findings suggest that albumin delivers S1P to the endothelial cells, which results in S1P1 receptor activation, which appears to prevent MMP activation via an unknown molecular mechanism. That albumin has no direct effect on the ESL is further supported by a study in which the shear modulus of hyaluronan solution was examined using optical trapping-based microrheology measurement [56]. The addition of albumin (or heparan sulfate) did not increase the shear modulus of hyaluronan solutions, indicating that these components did not bind hyaluronan. However, aggrecan, a known hyaluronan-binding proteoglycan, increased the hyaluronan shear modulus, as expected.

The role of ion concentrations on ESL stiffness was investigated in an ex vivo study in which isolated microvessels were perfused with physiologic salt solutions modified to have greater or less ionic strength by varying the sodium content. NaCl was replaced by Na₂SO₄ or glucose, and the pH of the solution was controlled by the addition of MOPS. The ESL thickness was assessed by tracking the access of 50 kDa, fluorescently labeled dextran to the labeled endothelial cells. A greater sodium concentration reduced the ESL thickness and a lower sodium concentration increased the ESL thickness compared with the isosmotic control [57]. An elevated amount of sodium is thought to reduce the ESL thickness by reducing the charge repulsion effects in the ESL by neutralizing the negative charges of the ESL. In a different study, several days of incubation with elevated sodium caused the ESL of the cultured endothelial cells to be stiffer and thinner when measured using AFM [49]. However, coincubation with the aldosterone blocker spironolactone prevented these effects, suggesting that the sodium-induced changes in the ESL under these extreme conditions were mediated not by sodium binding to the ESL but by sodium overload of the endothelial cells.

This relatively soft porous structure of the ESL poses a formidable hindrance to water flow, which was seen in hydrodynamic (see earlier Thickness and mechanical properties of ESL) and modeling studies [26]. Thus, the “wall” shear stress produced by the flowing blood exists mainly at the interface between the ESL and the lumen and is about zero at the endothelial plasma membrane [26, 30]. The absence of shear stress at the endothelial surface could shield adhesion molecule bonds from dispersive flow forces, although the torque and drag force imposed by the flow would still be present [58] because the neutrophil reaches far above the ESL into the lumen of the vessel.

NEUTROPHIL SURFACE LAYER

Human neutrophils express syndecan-1 [59, 60], syndecan-4, serglycin [61–63], CD44, and hyaluronan [64, 65]. However, the thickness or stiffness of the NSL is not well known. EM of neutrophils stained with lanthanum nitrate showed an ∼20-nm-thick surface layer on the neutrophils [66]. However, the surface layer data obtained with EM must be interpreted with caution, because different sample preparation methods can affect the ESL thickness and morphology. In a recent study, the surface layer of the adhered RBCs was probed with AFM and found to be ∼175 nm thick [67]. Based on these data, the NSL is probably thinner than the ESL. We chose a thickness of 100 nm for illustrative purposes only, because no good measurements were available (Fig. 1).

NEUTROPHIL RECRUITMENT

Neutrophils are the primary and essential defenders against bacterial, fungal, and viral infections [68]. Neutrophils are produced and stored in the bone marrow, lung, liver, and spleen. Under steady state conditions, only 1–2% of the total neutrophil pool is present in the blood stream [68, 69]. Neutrophils are recruited to sites of inflammation within <30 min of the insult and thus provide a rapid, first line of reaction to injury.

Neutrophil recruitment involves several consecutive steps. These steps (capture, rolling, slow rolling, arrest, crawling, and transendothelial migration) have been the subject of many excellent reviews [68–71] and are not elaborated in the present study.

PROPOSED ROLES OF ESL IN NEUTROPHIL RECRUITMENT

The shielding model

Several studies have shown that the ESL is ∼500 nm thick and can resist compression (see Thickness and mechanical properties of ESL). However, the endothelial adhesion molecules are <38 nm tall [58]. These observations suggest that the ESL can keep neutrophils away from endothelial adhesion molecules; thus, the ESL should have an antiadhesive effect (Fig. 1). In unbranched venules or arterioles, de novo adhesion of neutrophils is rarely observed. However, the barrier function of the ESL might be overcome by at least 2 mechanisms: mechanical compression of the ESL by neutrophils traversing capillaries that are smaller than their own diameter (Fig. 2) or shedding of the ESL by enzymes.

Figure 2. Neutrophil rolling in postcapillary venules.

Capillaries and venules are lined with ∼0.5-μm-thick ESL (green arrow). ESL is compressed mechanically in the capillaries by the deforming neutrophils, but after a short period, ESL regains its original thickness (red arrow). Neutrophils start to roll immediately after leaving the capillaries (black arrows) but not downstream.

Mechanical compression of the ESL.

Neutrophils pass through capillaries whose diameter is smaller than that of neutrophils. Neutrophils are deformable, in part because their plasma membrane surface area is about twice what is needed to enclose their volume [72]. Thus, neutrophils assume elongated sausage-like shapes during capillary passage. Rolling typically initiates where capillaries drain into postcapillary venules. Because the stiffness of neutrophils is higher than that of the ESL, it is plausible that the deformed neutrophils compress the ESL in the capillaries, possibly making it thin enough for the P-selectin-PSGL-1 bond to form [73]. Experimental evidence is available for this process. Restoration of the ESL after a neutrophil passage was measured using intravital microscopy, but the resolution was not sufficient to definitively confirm whether the “compressed” ESL was really thin enough to allow adhesion molecules to emerge from it [28]. If the ESL compression model is correct [73], neutrophil rolling would only be initiated at the entrance of capillaries into postcapillary venules but not in the straight part of larger venules. This is well supported by experimental evidence. In larger venules, neutrophil rolling is initiated through secondary capture [74], where an already adherent neutrophil supports the adhesion of a neutrophil from the blood stream in a process requiring PSGL-1 on the adherent neutrophil and L-selectin on the blood neutrophil. No new rolling is initiated in unbranched venules, except at the entrance points of capillaries. Once rolling is established, the bonds would persist, and the ESL would part similar to how wheat stalks part when a tractor comes through a field (Fig. 2).

The shielding model suggests that the endothelial adhesion molecules are shielded by the ESL. If they were not, the neutrophils would be brought into contact with them by the marginating forces acting in the blood stream. It was observed that neutrophils (leukocytes) tend to flow on the periphery of the vessel lumen and that this phenomenon is affected by blood flow velocity, blood plasma viscosity, and the presence and aggregation tendency of RBCs [75–77]. Because of their high deformability, the RBCs are pushed into the centerline of the blood flow. The more rigid neutrophils are pushed out to the periphery by the (overtaking) RBCs [78–80].

A recent modeling study found that the marginating forces acting on leukocytes are ≤130 pN in size [81], a magnitude comparable to the estimated forces the ESL can bear. During AFM studies, the ESL of the mouse aorta was found to require 0.3 pN for each 1 nm of indentation [48]. For probing the ESL, an AFM cantilever with a 10-μm polystyrene bead on its tip was used; the bead is necessary to increase the tapped surface and thus increase the signal coming from the ESL. A 500-nm-thick ESL with homogenous 0.3 pN/nm stiffness would suggest that the neutrophil would need to be pushed with ∼150 pN toward the vessel wall to reach the selectins, consistent with the postulated marginating force.

Although leukocyte rolling is primarily occurring in the postcapillary venules, it can be induced in the arterial system (i.e., in the mouse aorta [82], femoral artery [83], or cremaster arterioles [84]). Although the compression model can explain leukocyte rolling in postcapillary venules, it is not applicable to the arterial system, because no capillaries enter arteries. In the absence of other plausible mechanisms, ESL shedding could be of pivotal importance in regulating leukocyte rolling in the arterial system.

Shedding.

The ESL has been reported to be degraded by proteases (MMPs [85] and a disintegrin and metalloprotease domain [86]), heparanase (an endogenous heparan sulfate-cleaving enzyme) [87], or oxidative stress [88]. Only heparanase was directly tested in the context of neutrophil recruitment. Recombinant heparanase injected i.p. was found to initiate neutrophil rolling and adhesion in rat mesentery microvasculature, and heparanase was also found to augment neutrophil adhesion to cultured endothelial cells [89].

ESL shedding can be activated by inflammatory stimuli; TNF, LPS, thrombin, and formylated chemoattractant peptide (fMLP) were shown in different models to induce ESL shedding [31, 43, 48, 90–94]. These inflammatory mediators reduced the ESL thickness by one-half to two-thirds (Table 1). Although significant, this reduction in thickness is not sufficient to allow neutrophil adhesion, because the longest known adhesion molecule pair, P-selectin-PSGL-1, is only 75 nm long [95]. Challenging the cremaster vasculature with heparinase or TNF was shown to reduce ESL thickness and to increase the number of rolling and adhering leukocytes [18, 96]. LPS administration was shown to reduce ESL thickness and also to increase neutrophil recruitment into mouse lung capillaries [31]. In a series of in vivo studies, fMLP increased neutrophil adhesion, as expected. This was associated with reduced ESL thickness. Doxycycline, a broad inhibitor of MMPs, prevented the loss of ESL thickness, suggesting that MMPs could be involved in ESL shedding [92, 97, 98]. However, MMPs cleave syndecans; thus, the observation that staining with anti-heparan sulfate antibody was lost but staining for syndecan-1 was maintained is not consistent with this hypothesis. In the mesentery of syndecan-1 knockout mice, TNF stimulation resulted in much more leukocyte adhesion than in wild-type mice, but whether this observation resulted from the lack of endothelial or neutrophil syndecan-1 and whether the ESL was changed were not investigated [99].

Table 1.

ESL shedding

Organ	Species	Thickness before (nm)	Thickness after (nm)	Shedding agent	Reference
Lung	Mouse	1600	500	LPS or TNF	[31]
Aorta	Mouse	260	140	LPS	[48]
Endothelial cells	Human	250	140	LPS or TNF or thrombin	[48]
Cremaster	Hamster	400	100	TNF	[43]
Mesentery	Rat	460	330	fMLP	[93]

Inflammatory mediators and enzymes can have multiple effects on endothelial cells and leukocytes, in addition to shedding the ESL. TNF and LPS are known to strongly induce neutrophil adhesion via induction of endothelial adhesion molecules and chemokines [100–104]. In vitro, TNF was shown to upregulate endothelial syndecan-4 and hyaluronan production [105, 106] and to alter the sulfation pattern of heparan sulfate chains secreted by cultured endothelial cells [15]. Also, TNF induced secretion of the hyaluronan cross-linking molecule TNF-inducible gene 6 protein, which was found to condensate and rigidify a layer of hyaluronan polymers grafted to an artificial surface [107]. Furthermore, just as did native heparanase, point-mutated heparanase without catalytic activity facilitated lymphocyte adhesion to cultured endothelial cells [108]. A potential explanation for this is that heparan sulfate binding by heparanase can trigger intracellular signaling via syndecan-1 and -4 clustering [109].

In humans, severe inflammation-related conditions, in which increased neutrophil recruitment was also observed, were shown to be accompanied by elevated levels of ESL degradation products (heparan and chondroitin sulfate, hyaluronan, syndecan-1) in the blood plasma [110–114]. However, it is not known how this relates to the thickness or composition of the ESL in the microvessels relevant for neutrophil adhesion. Although it is clear that the ESL changes during inflammation, the exact consequences of these changes for leukocyte adhesion are unknown.

Microvilli.

The entire surface of resting leukocytes is decorated with ∼300-nm-tall and 100-nm-thick ridges of microvilli [115, 116]. Adhesion molecules (L-selectin, PSGL-1) were shown to concentrate to the microvilli of leukocytes [23, 116].

Leukocyte microvilli were suggested to help leukocyte penetration into the ESL by focusing the marginating forces acting on the leukocyte into a small surface area [23, 50]. However, according to a modeling study, because of the ESL’s viscoelastic property and the relatively fast blood flow velocity, the ESL might withstand poking by leukocytes microvilli [117]. Thus, in the presence of normal ESL, even the microvilli of neutrophils cannot reach the endothelial cell membrane. Inflammatory mediator-induced shedding never resulted in an ESL thinner than 100–200 nm (see Table 1). It could be that microvilli help to penetrate this reduced-thickness ESL.

Stimulation with chemokines can also induce protrusion formation in endothelial cells. Microvilli ∼200 nm wide and 1 μm long and filopodia 500 nm wide were observed in vitro on stimulated human endothelial cells and in vivo on microvessels of the chemokine-stimulated skin of a rabbit [118]. The endothelial microvilli were shown to present endothelial adhesion molecules and to be coated with heparan sulfate [118, 119]. The endothelial protrusions, just as do the leukocyte protrusions, might facilitate the neutrophil-endothelial contact formation in the presence of shed ESL.

Electrostatic repulsion model

The ESL and the NSL are both negatively charged owing to the large amount of sulfate and carboxyl groups of the heparan and chondroitin sulfates and the carboxyl groups of the hyaluronan. Accordingly, negatively charged molecules were found to penetrate slower into the ESL than were the neutrally or positively charged molecules [42].

The charge repulsion between the negatively charged ESL and NSL was proposed to prevent the neutrophil-endothelial interaction. In a seminal report in the cell adhesion field [120], Bell showed this to be the source of an ever-present repulsive force. In fact, receptor-mediated adhesion cannot be modeled without assuming such a repulsive force [121].

Some experimental support for the repulsion model comes from the observation that myeloperoxidase facilitates neutrophil adhesion. Because myeloperoxidase binds to heparan sulfate and might mask its negative charge, it is plausible that shielding the negative charge of the ESL could reduce the charge repulsion and thus promote leukocyte adhesion [122].

The “activating bed” model

The ESL is known to bind chemokines, some cytokines, and growth factors. Chemokines mutated to lack the ability to bind GAGs but capable of binding and activating chemokine receptors function in vitro but not in vivo [24]. This was best shown for CCL5, also known as RANTES [21]. Chemokine presentation was proposed to contribute to the establishment of chemokine gradients along the vessel wall by preventing chemokine washout by the blood flow. GAG binding also concentrates chemokines locally by providing a diffusion trap: chemokines produced by endothelial cells or cells in the tissue effectively can become trapped in the ESL [23]. GAG binding has also been proposed to affect the orientation and oligomerization of chemokines and to affect the interaction between the chemokines and their receptors [21, 23, 24, 123]. Because the ESL is so thick, a substantial portion of the rolling neutrophil is effectively bathed in a gel to which chemokines are bound (Fig. 1). Thus, many more of the G-protein-coupled receptors could be activated than would be the case if the neutrophils were to “tiptoe” on their microvilli on a flat surface. Consistent with this idea, a few experiments have shown that ESL shedding [124] or reduced sulfation of heparan sulfates [125] can reduce neutrophil arrest.

In a recent study, heparanase-overexpressing transgenic mice were examined. Heparanase overexpression was shown to shorten the core protein-bound heparan sulfate chains and to release fragmented heparan sulfate into the blood stream. In cremaster venules after CXCL-2 stimulation, the number and velocity of rolling neutrophils was slightly increased, and the number of extravasated neutrophils was significantly reduced in the heparanase-overexpressing mice. Also, the presentation of exogenous CXCL-2 was found to be reduced in the heparanase transgenic mice [124].

Another study demonstrated the importance of sulfation of the ESL heparan sulfate in neutrophil recruitment [125]. Endothelial NDST, an enzyme that initiates the sulfation of the heparan sulfate chains, was inactivated selectively in endothelial cells in mice. Systemic knockout of NDST-1 is lethal in mice, but endothelial inactivation resulted in impaired response in various inflammation models. NDST-1-deficient endothelial cells isolated from the NDST-1 null mice showed reduced heparan sulfation, reduced IL-8 chemokine presentation on the ESL, and impaired neutrophil arrest.

Another study showed that altering 2-O-sulfation of uronic acids in heparan sulfate can enhance inflammation and neutrophil recruitment [126]. In that study, the HS2ST enzyme was inactivated in endothelial cells. Inactivation of endothelial HS2ST resulted in reduced 2-O-sulfation, increased N-sulfation, and 6-O-sulfation of heparan sulfates. The rearrangement of sulfated sequences in the mutant imparted a gain-of-function phenotype in which the animals showed enhanced inflammation. IL-8 bound better to the ESL, and, accordingly, enhanced neutrophil arrest was observed.

These reports show that an altered sulfation pattern without ESL shedding could influence neutrophil recruitment. TNF and IL-1 stimulation of cultured endothelial cells has been shown to increase the amount of secreted proteoglycans containing more 2-O-sulfated and less 6-O-sulfated disaccharides, suggesting that chronic inflammation could alter the responsiveness of the ESL through changes in HS composition [15].

The adhesive surface model

Heparan sulfates are ligands of P- and L-selectins [22, 127–129]. This is corroborated by the observation that soluble heparan sulfate can attenuate leukocyte rolling and adhesion [130]. Modification of the heparan sulfate sulfation pattern was shown to affect neutrophil rolling; endothelial NDST-1 inactivation resulted in reduced ESL heparan sulfate sulfation and impaired L-selectin mediated neutrophil rolling. Inactivation of HS2ST yielded reduced 2-O-sulfation but increased 6-O-sulfation and enhanced L-selectin-mediated rolling [125, 126]. 6-O-sulfated N-acetylglucosamine residues in O-linked oligosaccharides on glycoproteins are well recognized L-selectin ligands, regulating lymphocyte arrest on high endothelial cells [131], but they are not expressed by other endothelial cells.

In addition to heparan sulfates, the second most abundant ESL GAG, hyaluronan, was also shown to be adhesive for neutrophils by binding CD44. CD44 is a transmembrane glycoprotein that exists in many isoforms generated by splice variants [132]. CD44 [64] is expressed on both the endothelial and leukocyte cell surface; thus, it is plausible that an “endothelial CD44-hyaluronan-leukocyte CD44” sandwich model could contribute to neutrophil arrest.

Several in vivo and in vitro experiments were done to investigate this hypothesis. The arrest of neutrophils on TNF-stimulated HUVEC monolayers was tested in dynamic and static adhesion assays. Pretreatment of HUVEC with CD44 blocking antibody or hyaluronidase attenuated neutrophil rolling and enhanced neutrophil adhesion [65, 133]. These findings are consistent with the study in which neutrophil adhesion to hamster cheek pouch venules was attenuated by i.v. hyaluronan administration [134], and increased levels of blood hyaluronan might have blocked both endothelial and neutrophil CD44. However, in the same study, hyaluronan pretreatment of activated HUVEC cultures attenuated neutrophil adhesion under static conditions. In the cremaster venules of CD44 knockout mice, the CXCL-2-induced neutrophil rolling was enhanced, but the adhesion was attenuated [64]. In wild-type mice, removal of endothelial hyaluronan with hyaluronidase did not significantly affect neutrophil rolling but did attenuate neutrophil adhesion. In the same study, neutrophils were found not to roll on a hyaluronan-coated glass surface, suggesting that the CD44-hyaluronan interaction was mediating neutrophil adhesion but not rolling [64]. Taken together, the CD44-hyaluronan interaction probably plays a role in neutrophil arrest, but this is only partly understood.

ROLES OF NSL IN NEUTROPHIL RECRUITMENT

Little and controversial evidence supports that NSL has an antiadhesive effect via a possible shielding mechanism. In a static adhesion assay, syndecan-1 knockout neutrophils showed enhanced adhesion to activated HUVEC monolayers [60], preventable via heparin administration. However, in the same study, syndecan-1 knockout neutrophils also showed stronger adherence to unstimulated HUVECs. The molecular background of these observations was not examined. In another study, decreased sulfation of NSL did not affect neutrophil trafficking in wild-type mice transplanted with NDST1 knockout bone marrow [125]. NSL’s antiadhesive effect is corroborated by the NSL’s negative charge, which contributes to the charge repulsion effect (see Electrostatic repulsion model). Some evidence has suggested that under certain circumstances the NSL can be proadhesive. In a dynamic adhesion assay, HUVEC pretreatment with blocking anti-CD44 antibody and hyaluronidase pretreatment of neutrophils both attenuated neutrophil rolling [65]. This suggests that hyaluronan of the NSL can bind to the endothelial CD44.

VARIATION OF ESL AND ITS ROLE IN NEUTROPHIL RECRUITMENT ACROSS DIFFERENT ORGANS

Endothelial cell properties and gene expression patterns vary among arterioles, capillaries, and venules [135, 136]. In addition, endothelial cells show organ specificity. One aspect of this is the anatomic appearance (continuous, fenestrated, or discontinuous) of the endothelial layer [137]. Other aspects are the nature and tightness of the interendothelial junctions and the molecules expressed on the endothelial cell surface. These differences are only beginning to be recognized, and knowledge is sparse. Recently, the transcriptome of high endothelial venules was reported [138]. These endothelial cells express a unique 6-sulfated N-acetylglucosamine-containing ligand for L-selectin called peripheral node addressin that is not expressed in other endothelial cells. It is likely that the molecular composition, thickness, and mechanical properties of the ESL vary among organs, between large and small vessels, between arterioles and venules, and depending on the inflammatory status. To begin to address the organ-specific differences, we review the current knowledge about the role of ESL in neutrophil recruitment into different organs.

Cremaster and mesentery vasculature

The cremaster and mesentery vessels are lined with a continuous layer of endothelial cells, which are coated with an ∼0.5-μm-thick ESL. Neutrophil rolling and arrest is primarily observed in the postcapillary venules and follows the classic adhesion cascade.

Neutrophil rolling is rapidly induced by the surgical trauma associated with tissue preparation and is in part due to the effects of mast cell-derived histamine [139] and tryptase [140] on venular endothelial cells. Experimental evidence supports the shielding effect of the ESL but also its importance in chemokine presentation and binding L-selectin. The CD44-hyaluronan interaction has relevance but requires additional elucidation.

Lung vasculature

The lung vasculature is paved with a continuous layer of endothelial cells, which was reported to be covered with a thick ESL (∼1.5 μm) [31], although this thickness is not certain and remains to be confirmed. In the lung, neutrophil extravasation occurs primarily through the capillaries [141], where leukocytes are wedged between the endothelial cells and where mechanical trapping of neutrophils was proposed to contribute to neutrophil extravasation. The involvement of selectins and integrins in neutrophil recruitment seems to be dependent on the inflammatory stimulus [141].

The ESL was shown to prevent neutrophil recruitment into the lung during ALI in mice. In that study, ALI was induced by i.v. administration of LPS. At 30 min after the injection, the lung ESL thickness was assessed with tracer exclusion intravital microscopy, which showed a decrease in ESL thickness from ∼1.6 to ∼0.5 μm on induction of ALI. This should not be sufficient to promote neutrophil adhesion. Adoptively transferred (i.v. injected) GFP-expressing neutrophils or ICAM-1 or VCAM-1 antibody-coated fluorescent microspheres were arrested in greater amounts in the lung vasculature. Forty-five minutes after LPS injection, no increase in the endothelial ICAM-1 or VCAM-1 expression level was observed. The proadhesive effects of LPS did not develop in mice cotreated with heparin (inhibitor of heparanase) or in transgenic mice lacking TNF-receptor-1 or heparanase. It was concluded that in this ALI model, LPS induced TNF release, which in turn activated pulmonary endothelial heparanase, resulting in ESL shedding and increased neutrophil recruitment into the lung. These findings are consistent with the higher expression level of heparinase detected in lung biopsy samples of human patients with nonpulmonary sepsis [31].

The antiadhesive role of ESL in the lung is further supported by the observation that in syndecan-4 knockout mice, intratracheal LPS stimulation leads to threefold more neutrophils in the bronchoalveolar lavage compared with wild-type mice [142]. The role of CD44 binding to hyaluronan in neutrophil recruitment into the lung is unclear [143–146].

Coronary system

The coronary vasculature is coated with a continuous layer of endothelial cells, which presents an ESL of ≥100 nm thickness according to EM studies [90], but the limitations of EM for assessing the ESL thickness have been addressed above [40]. In the coronary vasculature, neutrophils are recruited primarily in the postcapillary venules and through the classic recruitment cascade [147, 148].

The role of the ESL in coronary neutrophil recruitment was studied in the context of I/R. In a hamster Langendorf-heart preparation, coronary I/R resulted in elevated syndecan-1 levels in the outflow and reduced layer thickness on electron micrographs, suggesting ESL shedding. This was associated with increased adhesion of leukocytes to the endothelial surface, which was reflected by a decreased leukocyte count in the outflow [149]. Syndecan-1 overexpressing mice were found to develop smaller ischemic lesions and to recover from the postischemic trauma better [150].

In contrast, perfusion of the coronary vasculature of explanted and cannulated guinea pig hearts with heparinase solution was found to increase the adhesion of isolated neutrophils perfused through the coronary system [151]. A clinical study of human patients with a myocardial infarct found that serum catecholamine and syndecan-1 levels, possibly derived from degraded ESL, correlated with myocardial infarct severity [152]. Adenoviral syndecan-4 overexpression in rats was found to be beneficial in postinfarct myocardial recovery [153]. Although in these studies the ESL thickness was not measured, the investigators suggest that the presence of ESL is protective during I/R.

Kidneys

In the kidney, neutrophil extravasation occurs in the glomerular capillaries and cortical venules. These vessels are lined with fenestrated endothelial cells, and a 500-nm-thick ESL has been reported to be present on them [10]. In these microvessels, neutrophil recruitment follows the classic cascade [141].

Enzymatic removal of heparan sulfate from TNF-activated glomerular endothelial cells attenuated the rolling and adhesion of a neutrophil-like cell line (32Dcl3) in dynamic and static adhesion assays [154]. Heparan sulfate antibodies yielded the same results [155]. These findings suggest that the proadhesive role of ESL could be relevant in the kidney.

This was further corroborated by an in vivo study conducted on endothelial-specific NDST-1 knockout mice. As discussed earlier, this mouse shows reduced sulfation of ESL heparan sulfates. In these mice, reduced neutrophil infiltration into nephritic kidney glomeruli was observed. Neutrophils rolled faster and adhered less to the monolayers of NDST-1 knockout glomerular endothelial cells cultured from these mice. Also, the binding of recombinant L-selectin and chemokines to these cells was attenuated, consistent with the involvement of endothelial NDST in forming L-selectin ligands and chemokine-immobilizing sulfation patterns [156].

The importance of the endothelial CD44 interaction with neutrophil hyaluronan was shown in renal neutrophil extravasation in an ischemia-reperfusion model [65]. Wild-type mice injected with CD44 blocking antibody and CD44 knockout mice were protected against kidney I/R. Adoptively transferred CD44^+/+ and CD44^−/− neutrophils both extravasated into kidney I/R of wild-type mice. After transplanting wild-type bone marrow into CD44 knockout mice, the recruitment of wild-type neutrophils was attenuated on I/R. These findings suggest that the hyaluronan presented by the neutrophil might bind endothelial CD44.

Liver

The capillaries of the liver vasculature form sinusoids, which are lined with a discontinuous and fenestrated endothelial layer. The presence of ESL components in liver microvessel has been reported [88, 157, 158]; however, the thickness of the ESL in the liver microvasculature was not measured. Neutrophil recruitment occurs mainly in the liver sinusoids and to a smaller extent in the postcapillary venules [159]. While in the postcapillary venules, neutrophils follow the classic recruitment cascade. In the sinusoids, they do not show selectin-mediated rolling. Selectins are not expressed by the liver sinusoid endothelial cells [141]. Mechanical trapping by neutrophil stiffening in response to inflammatory stimuli was proposed to contribute to neutrophil arrest in the sinusoids [157].

The hyaluronan-CD44 interaction was found to be important for neutrophil recruitment into the liver sinusoids. Hyaluronidase pretreatment attenuated LPS-induced neutrophil arrest in the liver sinusoids but not in the postcapillary venules. Pretreatment with CD44 blocking antibody or performing the experiment in CD44 knockout mice showed the same result. Experiments with bone marrow chimeras indicated that the neutrophil, and not the endothelial, CD44 is important for adhesion [158]. Similarly, administration of CD44 blocking antibody reduced the neutrophil extravasation into the I/R-challenged liver and reduced the liver damage on I/R [160].

Brain

The brain vasculature has a continuous layer of endothelial cells, which form the blood-brain barrier. EM showed a continuous ESL in these vessels [161, 162]; however, the validity of EM is debated, and the ESL thickness has not been measured. Syndecans, glypican-1, and CD44-presented hyaluronan were all detected on brain endothelial cells [6, 163]. Neutrophils are recruited according to the classic cascade with the involvement of some special adhesion molecules [68, 164]. Nothing is known about the role of ESL in neutrophil recruitment into the brain. Some observations were made in the retina microvasculature, which is known to have strong similarities with the brain microvasculature [165]. Syndecan-1 knockout mice showed enhanced leukocyte adhesion [99]. In wild-type mice transplanted with syndecan-1 knockout bone marrow, the same increase in leukocyte binding was observed, suggesting that leukocyte, but not endothelial, syndecan-1 has antiadhesive effects in this model. However, the type of adhered leukocyte was not characterized.

CONCLUSION

The ESL is an at least 500-nm-thick interface between the vessel wall and the flowing blood. The ESL is a key, but understudied, player in neutrophil recruitment. In most studies of neutrophil recruitment, the ESL has been completely ignored and not even discussed. The ESL has both pro- and antiadhesive effects. The proadhesive effects include chemokine presentation and L-selectin and CD44 ligand activity. The antiadhesive effects of the ESL can potentially be overcome by mechanical compression or enzymatic shedding of the ESL, or both. Some evidence has shown that the ESL might be thinner in cultured endothelial cells and thicker in large arteries, and that some organ differences might exist.

AUTHORSHIP

A.M., J.D.E., A.R.P., and K.L. wrote the text. A.M. and K.L. created the figures and table. K.L. supervised the project.

Glossary

AFM

atomic force microscopy

ALI

acute lung injury

EM

electron microscopy

ESL

endothelial surface layer

GAG

glycosaminoglycan

HS2ST

uronyl sulfate 2-O sulfotransferase

I/R

ischemia/reperfusion injury

MMP

matrix metalloproteinase

NDST

N-deacetylase N-sulfotransferase

NSL

neutrophil surface layer

PSGL-1

P-selectin glycoprotein ligand 1

S1P

sphingosine-1-phosphate

WGA

wheat germ agglutinin

ACKNOWLEDGMENTS

This work was supported by U.S. National Institutes of Health Grants HL115232 and HL78784 (National Heart, Lung, and Blood Institute) and DK91222 (National Institute of Diabetes and Digestive and Kidney Diseases) to K.L.

DISCLOSURES

The authors declare no competing financial interests.