Anda di halaman 1dari 47

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

The Structure and Function of the Endothelial Glycocalyx Layer


Sheldon Weinbaum,1,2 John M. Tarbell,1 and Edward R. Damiano3
1

Department of Biomedical Engineering and 2 Department of Mechanical Engineering, The City College of New York, New York, NY 10031; email: weinbaum@ccny.cuny.edu, tarbell@ccny.cuny.edu

Department of Biomedical Engineering, Boston University, Boston, Massachusetts 02215; email: edamiano@bu.edu

Annu. Rev. Biomed. Eng. 2007. 9:12167 First published online as a Review in Advance on March 20, 2007 The Annual Review of Biomedical Engineering is online at bioeng.annualreviews.org This articles doi: 10.1146/annurev.bioeng.9.060906.151959 Copyright c 2007 by Annual Reviews. All rights reserved 1523-9829/07/0815-0121$20.00

Key Words
mechanical and biochemical properties of glycocalyx, revised Starling principle, cellular interactions with glycocalyx, inammatory response, mechanotransduction

Abstract
Over the past decade, since it was rst observed in vivo, there has been an explosion in interest in the thin (500 nm), gel-like endothelial glycocalyx layer (EGL) that coats the luminal surface of blood vessels. In this review, we examine the mechanical and biochemical properties of the EGL and the latest studies on the interactions of this layer with red and white blood cells. This includes its deformation owing to uid shear stress, its penetration by leukocyte microvilli, and its restorative response after the passage of a white cell in a tightly tting capillary. We also examine recently discovered functions of the EGL in modulating the oncotic forces that regulate the exchange of water in microvessels and the role of the EGL in transducing uid shear stress into the intracellular cytoskeleton of endothelial cells, in the initiation of intracellular signaling, and in the inammatory response cascade.

121

Contents
1. INTRODUCTION AND OVERVIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Early Indications of the EGL in Microvessels In Vivo . . . . . . . . . . . . . . . . The EGL in Large Vessels and Atherogenesis . . . . . . . . . . . . . . . . . . . . . . . . EM Observations Prior to 2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Direct Intravital Microscopic Evidence for the Full Extent of the EGL In Vivo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The EGL and the Starling Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. STRUCTURE AND COMPOSITION OF THE EGL . . . . . . . . . . . . . . Biochemical Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recent EM Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Relation between Biochemical Composition and EM Observations . . . . 3. MECHANICAL AND ELECTROCHEMICAL PROPERTIES . . . . . EGL Thickness Using Microviscometric Analysis and -PIV . . . . . . . . . Elastic Properties of EGL Fibers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electrochemical Properties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. MODELS FOR STRUCTURAL INTEGRITY AND RESTORATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrodynamic Models for Flow in the EGL . . . . . . . . . . . . . . . . . . . . . . . . . Models for the Restoring Mechanisms of the EGL . . . . . . . . . . . . . . . . . . . 5. CELLULAR INTERACTIONS WITH THE EGL . . . . . . . . . . . . . . . . . Red Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . White Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. PHYSIOLOGICAL FUNCTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Permeability and the Revised Starling Principle . . . . . . . . . . . . . . . . . . . . . . Mechanotransduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Inammatory Response and IschemiaReperfusion Injury . . . . . . . . . . . . 7. UNRESOLVED ISSUES AND FUTURE DIRECTIONS . . . . . . . . . . 122 123 123 124 124 125 127 127 130 131 132 132 134 135 137 137 138 143 143 144 146 146 149 155 158

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

1. INTRODUCTION AND OVERVIEW


It is now well recognized that the luminal surface of the endothelial cells (ECs) that line our vasculature is coated with a glycocalyx of membrane-bound macromolecules comprised of sulfated proteoglycans, hyaluronan, glycoproteins, and plasma proteins that adhere to this surface matrix. A similar coating is observed on the apical surface of epithelial cells that line many of our internal organs, but this review focuses only on the endothelial glycocalyx layer (EGL). We examine the remarkable properties of this hydrated gel-like structure and the three principal functions that it plays as the interface between the ECs and the owing blood with its plasma and cellular components. In particular, we explore its function as a modulator of permeability in the transcapillary exchange of water; as a mechanotransducer of uid shear stress (FSS) to the endothelial cytoskeleton, including the resulting biochemical responses;

EC: endothelial cell EGL: endothelial glycocalyx layer FSS: uid shear stress

122

Weinbaum

Tarbell

Damiano

and as a regulator of red and white blood cell (WBC) interactions, with emphasis on the inammatory response. Much of the early literature on the EGL is summarized in an excellent review by Pries et al. (1). Therefore, we provide only a brief overview of the literature prior to 2000 in this Introduction.

WBC: white blood cell EM: electron microscopy

Early Indications of the EGL in Microvessels In Vivo


The concept that a thin endocapillary layer might cover the entire endothelial surface was rst proposed in the 1940s by Danielli (2) and Chambers & Zweifach (3), and was subsequently reexamined by Copley (4) who suggested that the layer was an immobile sheet of plasma and macromolecules. However, this layer evaded observation by light and electron microscopy (EM) until 1966, when Luft (5), using ruthenium red staining, detected a thin layer ( 20 nm thick) in rat intestinal mucosa. For reasons that were not clear at the time, this layer was far thinner than the much thicker EGL proposed by Klitzman & Duling (6) to account for the low capillary tube hematocrits (i.e., the instantaneous volume fraction of red cells resident in the capillary) that they observed in skeletal-muscle capillaries. Evidence that the macromolecules of the EGL might interfere with ow in a much thicker plasma layer near the capillary wall was rst reported by Desjardins & Duling (7). After enzyme treatment targeted at cleaving specic proteoglycan molecules within the EGL, they observed up to a twofold increase in capillary tube hematocrit. In addition to these observations, evidence that the EGL contributes to microvascular ow resistance was provided by Pries et al. (8). Combining network simulations with measurements of blood ow in large-scale microvascular networks, Pries et al. (8) concluded that the resistance to blood ow in microvessels up to 30 m in diameter was dramatically higher than in glass tubes of the same diameter. In a subsequent study, Pries et al. (9) found that the resistance to blood ow in microvascular networks decreased markedly after enzyme treatment to remove the EGL. These studies and those of Duling and coworkers suggested that the EGL could serve to retard plasma ow near the vessel wall, which in turn would result in enhanced resistance to blood ow and lower capillary tube hematocrits in vivo than in smooth glass tubes. These experimental studies were accompanied by a series of theoretical models of increasing sophistication that described the axisymmetric single-le motion of spherical particles and red cells in a cylindrical capillary with a porous matrix on its luminal surface (1013).

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

The EGL in Large Vessels and Atherogenesis


Although early studies on the EGL were limited largely to microvessels, the association of altered EGL characteristics with atherosclerosis-prone locations in arteries was recognized in the early 1980s. Lewis et al. (14) observed the coronary arteries of White Carneau pigeons and noted that the EGL, as assessed by ruthenium red staining, was thinnest in areas with high disease predilection, and that upon cholesterol challenge, the EGL thickness was reduced in all arterial zones. This idea of association between EGL abundance and arterial disease has been revisited recently by van den Berg et al. (15), who showed that the EGL thickness in the disease-prone

www.annualreviews.org Endothelial Glycocalyx Layer

123

LDL: low-density lipoprotein

sinus region of the mouse internal carotid artery is signicantly less than in the nearby common carotid artery that is spared of disease. They also reported that the EGL is diminished upon systemic atherogenic challenge by a high-fat, high-cholesterol diet. These observations are consistent with previous studies demonstrating rapid shedding of the EGL from the endothelial surface upon acute stimulation with elevated plasma levels of oxidized low-density lipoproteins (LDL) or acute exposure to inammatory agents and FSS-induced synthesis of EGL components (reviewed in Reference 16).

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

EM Observations Prior to 2000


As noted above, the rst visualization of the EGL by EM used the cationic dye ruthenium red that binds to acidic mucopolysaccharides and generates electron density in the presence of osmium tetroxide (5). Subsequent studies by Baldwin & Winlove (17) and Clough & Moftt (18) used gold colloids and immunoperoxidase labeling. Adamson & Clough (19) demonstrated, using a large charged marker protein, cationized ferritin, to demarcate the edge of the EGL in frog microvessels, that, in the absence of plasma proteins, the EGL would collapse, presumably owing to elimination of intramolecular interactions with plasma proteins, and that its undisturbed thickness was several times greater than the 20 nm observed with ruthenium red. All of these methods suffer from dehydration artifacts associated with aqueous xatives that likely dissolve all but the protein cores of proteoglycans. A method developed to preserve water-soluble structures using uorocarbons as nonaqueous carriers of osmium tertroxide was applied to microvessels to obviate some of these limitations by Sims & Horne (20). Further elaborations of the uorocarbon-glutaraldehyde xation methods by Rostgaard & Qvortrup (21) revealed a lamentous brush-like surface coating on capillary walls with a layer thickness of <50 nm, suggesting a cleavage of more supercial matrix structures. All of the foregoing EM studies suggested an EGL with a thickness of less than 100 nm. None of these studies shed light on the possible organization of the EGL and its possible relationship to the F-actin scaffold beneath the apical membrane that was rst reported in Squire et al. (22), as described later in Figure 2.

Direct Intravital Microscopic Evidence for the Full Extent of the EGL In Vivo
What was missing from in vivo experiments prior to 1996 was a direct in vivo measurement of the thickness of either a positively labeled layer of macromolecular constituents of the EGL or an exclusion zone to red cells and uorescently labeled macromolecules. The rst important breakthrough along these lines came with the dye-exclusion technique developed by Vink & Duling (23). Using a 70 kDa FITCdextran plasma tracer, which they showed was sterically excluded by the EGL, they were able to provide the rst estimate of the in vivo thickness of the layer in capillaries. Comparing the anatomical diameter of hamster cremaster-muscle capillaries, visualized under brighteld illumination, with measurements of the functional diameter available to WBCs, red cells, and uorescent macromolecules, they concluded

124

Weinbaum

Tarbell

Damiano

that the thickness of the EGL was 0.40.5 m, which is 15%20% of the radius of the smallest capillaries in the microcirculation. This estimate of the in vivo thickness of the EGL is four to ve times greater than previous estimates derived from EM studies, which likely signicantly underestimated the value owing either to the dehydration of the extracellular matrix or the cleavage of its outer matrix components that accompanies tissue xation. This discrepancy was a catalyst for much of the work that has followed on the estimation of EGL thickness and its function as a barrier in cellular interactions. Beyond providing the rst estimate of the in vivo thickness of the EGL, the Vink & Duling experiments were signicant in several other important ways. In particular, they showed that except at velocities <20 m/s, red blood cells do not invade the region occupied by the EGL in tightly tting capillaries, whereas the much stiffer and larger WBCs crush the layer. However, when ow is arrested, red cells also appear to have this ability because they ll nearly the entire capillary lumen. From fundamental uid mechanical considerations, and based on our understanding of red cell mechanics, these data provided quantitative insight into important mechanical properties of the EGL, as discussed in further detail in Sections 24. Finally, Vink & Duling showed that exposing FITC-dextran perfused capillaries to epiuorescent illumination for 5 min resulted in signicant degradation of the EGL and that this degradation was effectively blocked by administering superoxide dismutase (SOD) and catalase. (SOD is an enzyme that catalyzes the dismutation of the superoxide anion radical into oxygen and hydrogen peroxide, whereas catalase is an enzyme that catalyzes the decomposition of hydrogen peroxide into water and oxygen.) This supports the hypothesis that damage to the EGL owing to light-dye treatment might be mediated by oxygen-derived free radicals. Further support for this hypothesis was demonstrated when Vink et al. (24) showed that oxidized LDL caused a similar degradation of the EGL, whereas native LDL did not. While it is known that heparan sulfate proteoglycans in the EGL bind SOD (25), these results suggest an important role for the EGL in scavenging free radicals from the blood. Such a role could have important implications for cardiovascular health and disease ranging from inammation to atherosclerosis. Although elegant, the dye-exclusion technique developed by Duling and coworkers does not provide adequate resolution in vessels larger than 1215 m in diameter (26) because of the increased out-of-focus light and optical difculties associated with the uorescent dye column near the vessel wall. In an effort to test whether the EGL extended into microvessels beyond the capillary regime, a new technique has been developed that uses high-resolution, near-wall, intravital uorescent microparticle image velocimetry (-PIV) to examine the velocity prole near the vessel wall in postcapillary venules of the mouse cremaster muscle. These studies (2729) are discussed in Section 3.

SOD: superoxide dismutase -PIV: microparticle image velocimetry

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

The EGL and the Starling Principle


For more than a century, investigators interested in uid exchange across microvesssels have widely assumed that the oncotic forces that drive this ow are determined

www.annualreviews.org Endothelial Glycocalyx Layer

125

by the difference in plasma protein concentration between the vessel lumen and tissue. The key role of the EGL in this exchange was not appreciated. The fact that the Starling principle may not have been correctly applied was highlighted at the 1996 Starling conference celebrating the one-hundredth anniversary of Starlings pioneering paper outlining his hypothesis for the ltration and absorption of water in capillaries and the formation of lymph (30). At this meeting, Levick presented a provocative paper based on his earlier review (31), which showed that when tissue oncotic pressures were carefully measured using the latest experimental methods, the widely accepted classical Starling force balance was violated in all tissues except the kidney and the intestinal mucosa (tissues whose main function is venous reabsorption), and there was no reabsorption on the venous side of capillaries, contrary to long-accepted views. Without venous reabsorption, one could not properly account for the low whole-body lymph ows measured in vivo. These observations were further supported by the experiments of Michel & Phillips (32) on single isolated perfused frog microvessels. They showed that when capillary pressures were suddenly dropped, there would initially be a short transient period of reabsorption, which would quickly decay, followed by a very low level of ltration in the new steady state. These observations clearly did not satisfy the classical application of the Starling principle. Subsequent to the 1996 Starling conference, Michel (30) and Weinbaum (33) independently proposed a revised Starling hypothesis, now referred to as the MichelWeinbaum model (34), which proposes that the primary molecular sieve for plasma proteins is the EGL and that the Starling forces are determined, not by the global difference in oncotic pressure between lumen and tissue, but by the local difference in protein concentration across the EGL alone. This revised hypothesis is described more fully in Hu & Weinbaum (35), where a detailed three-dimensional model of the EGL and the interendothelial cleft is developed. This revised model has since been experimentally conrmed (see Section 6) in frog and mouse microvessels (36, 37) and in EC monolayers in culture (38). As summarized in this introduction, much of the research related to the EGL prior to 2000 has been focused on hydrodynamic problems examining the hematocrit defect and hydrodynamic resistance of microvascular networks, ultrastructural problems related to measuring the thickness of the EGL and its uniformity, theoretical models describing the axisymmetric motion and deformation of red cells in capillaries with an undeformed EGL, and a revisiting of the classical Starling principle. Since 2000, major new research directions have evolved as the multifaceted functions of the EGL have become more apparent. In particular, these new studies of the EGL have focused on (a) its detailed structure and biochemical composition, including its mechanical and electrochemical properties, (b) its structural integrity and response to deformation, (c) its interactions with red blood cells and WBCs, (d ) its role as the primary mechanotransducer of FSS, (e) its role in the inammatory response cascade and ischemiareperfusion injury, and ( f ) experiments to conrm the revised Starling principle. These are the major themes of the present review.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

126

Weinbaum

Tarbell

Damiano

2. STRUCTURE AND COMPOSITION OF THE EGL Biochemical Composition


The surface of ECs is decorated with a wide variety of membrane-bound macromolecules, which constitute the EGL. From glycoproteins bearing acidic oligosaccharides and terminal sialic acids (SAs) to proteoglycans along with their associated glycosaminoglycan (GAG) side chains, the polyanionic nature of its constituents imparts to it a net negative charge. Under physiological conditions, an extended endothelial surface layer arises from the association of components of the EGL with blood-borne molecules (1, 19). Plasma proteins, enzymes, enzyme inhibitors, growth factors, and cytokines, through cationic sites in their structure, as well as cationic amino acids, cations, and water, all associate with this matrix of biopolyelectrolytes (39, 40). An additional level in the complexity of this biological structure arises from its dynamic nature. The interactions between GAGs and proteins are highly dependent on the conditions of their local microenvironment, such as cation content and concentration, and pH (4145). Furthermore, ECs actively regulate the content and physicochemical properties of GAGs on their surface by having high rates of continuous metabolic turnover that allow adaptation to changes in the local environment (4648). GAGs are linear polydisperse heteropolysaccharides, characterized by distinct disaccharide unit repeats (49). Specic combinations of these give rise to different GAG families, such as the heparan sulfate (HS), chondroitin/dermatan sulfate (CS), and hyaluronic acid or hyaluronan (HA) found on ECs (50). Proteoglycans are proteins that contain specic sites where sulfated GAGs are covalently attached (49). HS and CS chains vary between 50 and 150 disaccharide units and have an average molecular weight of approximately 30 kDa (51). Sulfated GAGs form extended helical coils, whose conformation depends on the local patterns of sulfation, the exibility of the monosaccharides involved, and the degree of intramolecular electrostatic interactions (52). Local ionic strength and pH strongly inuence the level of extensibility of a GAG chain, and it appears that this is maximal in a NaCl solution having physiological concentration, as demonstrated recently by Seog et al. (53). Under these conditions, GAGs are considered to extend to approximately 80% of their contour length, so a chain containing 100 disaccharide units would correspond to 80 nm (1, 53). Recently, it was demonstrated that in small arteries of the rat mesentery, ionic strength is a major determinant of the overall state of the EGL, which can move from a collapsed to an extended state as ionic strength decreases (54). The association of GAGs with proteins impacts their structure. Adamson and Clough (19) demonstrated how, in the absence of plasma proteins, addition of a solution of a large charged marker protein (cationized ferritin, 440 kDa, 12 nm) reveals a collapsed EGL in frog mesenteric capillaries. Addition of a 2% frog plasma solution was enough to signicantly lift the layer of ferritin above the endothelial surface, whereas a 5% albumin solution had a less striking effect (19). The preferential behavior that the EGL displayed for the less oncotic native plasma over albumin alone makes it evident that specic interactions are essential to the physiological structure of the EGL.
SA: sialic acid GAG: glycosaminoglycan HS: heparan sulfate CS: chondroitin sulfate HA: hyaluronic acid (hyaluronan)

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

127

GPI: glycosylphosphatidylinositol

The GAGs commonly associated with the vasculature are HS, CS, and HA, with levels varying among cell types. The most prominent on the surface of ECs are HS, accounting for 50%90% of the total GAG pool, the rest being comprised of CS and HA (50). Owing to structural analogy with heparin, HS and associated proteoglycans have been the most extensively studied, with attention recently shifting toward their ability to function as signal transduction molecules (55, 56). The transmembrane syndecans, the membrane-bound glypicans, and the basement matrix-associated perlecans are the three major protein core families of HS proteoglycans found on ECs (57). Syndecans-1 (33 kDa), -2 (22 kDa), and -4 (22 kDa), which are expressed on ECs, have three GAG attachment sites, close to their N terminus and distal to the apical surface, substituted primarily, but not exclusively, by HS (57, 58). Syndecan-1 contains two additional sites that are close to the membrane and reserved for CS (59). On the other side of the plasma membrane, their cytoplasmic tails associate with the cytoskeleton and assist in its organization, through molecules such as tubulin, dynamin, and -actinin (60, 61). Active participation in signaling stems from the phosphorylation of certain intracytoplasmic residues, which act as switches controlling the oligomerization state and altering the binding properties of syndecans (55, 56, 60). Of the glypicans, glypican-1 (64 kDa) is the only one expressed on ECs (57). Close to the membrane, its three to four GAG attachment sites are exclusively substituted with HS (62). Glypican-1 is bound directly to the plasma membrane through a C-terminal glycosylphosphatidylinositol (GPI) anchor (62). Most importantly, the GPI anchor localizes this proteoglycan to lipid rafts, which are cholesterol- and sphingolipid-rich membranous domains involved in vesicular transport and cell signaling (6264). Caveolae can be considered a subset of lipid rafts, which arise from the incorporation of protein caveolin-1, a cholesterol carrier, into the membrane, where they may form characteristic cave-like structures (100 nm) that are supported by the cytoskeleton (65). In contrast with CS and HS, HA is a much longer disaccharide polymer, on the order of 1000 kDa, which is synthesized on the cell surface and is not covalently attached to a core protein (66). It is not sulfated, but obtains its negative charge from carboxyl groups that endow it with exceptional hydration properties (66). HA weaves into the EGL through its interaction with surface HA receptors, such as the transmembrane CD44, and CS chains (26). CD44 contains two GAGs, either CS or HS, and localizes along with HA in caveolae, where it has various functional interactions (67). Completing the picture, glycoproteins with short, branched oligosaccharides attached to their core are also found on the surface of ECs. These oligosaccharides are capped by SA, the nine-carbon monosaccharides that contribute to the net negative charge of the EGL, through their ionization at physiological pH (68). Many important receptors on the cell surface, such as selectins, integrins, and members of the immunoglobulin superfamily, have oligosaccharides attached to them and are classied as glycoproteins (1). An illustration that integrates all of the components of the EGL described above is shown in Figure 1 (adapted from Reference 69).

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

128

Weinbaum

Tarbell

Damiano

Syndecans Shedding CD44 Sialic acids


Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

Glycoprotein

Glypican

c
K+, Na+, Ca++, L-arginine Albumin, bFGF, LPL, polycationic peptides Cholesterol and glycosphingolipids Caveolin-1 K+, Na+, Ca++, L-arginine channels Hyaluronic acid Heparan sulfate Chondroitin sulfate

Figure 1 Representation of proteoglycans and glycoproteins on the surface of ECs. Caveolin-1 associates with regions high in cholesterol and sphingolipids in the membrane (darker circles, left) and forms cave-like structures, caveolae (right). Glypicans, along with their HS chains localize in these regions. Transmembrane syndecans are shown to cluster in the outer edge of caveolae. Besides HS, syndecans also contain CS, further down the core protein. A glycoprotein with its short oligosaccharide branched chains and their associated SA caps are displayed in the middle part of the gure. HA is a very long GAG that weaves into the glycocalyx and binds with CD44. Transmembrane CD44 can have CS, HS, and oligosaccharides attached to it and it localizes in caveolae. Plasma proteins, along with cations and cationic amino acids (red circles), are known to associate with GAGs. (a) The cytoplasmic domains of syndecans can associate with linker molecules, which connect them to cytoskeletal elements. (b) Oligomerization of syndecans helps them make direct associations with intracellular signaling effectors. (c) A series of molecules involved in eNOS signaling localize in caveolae. (This illustration is not drawn to scale and a few GAG attachments were omitted for simplicity.)

www.annualreviews.org Endothelial Glycocalyx Layer

129

Recent EM Observations
ACC: actin cortical cytoskeleton

Little was known about the ultrastructural organization of the EGL and its relation to the underlying F-actin cytoskeleton until the pioneering study of Squire et al. (22). Using computed autocorrelation functions and Fourier transforms of EM images obtained from both new (22) and previous studies (70) of frog mesentery capillaries, these authors were able to identify for the rst time the quasi-periodic structure of the EGL and the anchoring foci that appear to emanate from the underlying actin cortical cytoskeleton (ACC). The computer-enhanced images showed that the EGL is a three-dimensional brous network with 1012-nm-diameter focal scattering centers that have a characteristic spacing of 20 nm in all directions. Freeze-fracture replicas from rare sections where the fracture plane passed parallel and close to the endothelial surface showed that the anchoring foci formed a clearly dened hexagonal array with a spacing of 100 nm. This spacing is very similar to the spacing of the truncated bush-like structures that were seen by Rostgaard & Qvortrop (21) using a uorocarbon oxygen xation technique, which preserved the portion of the EGL that was close to the EC surface. Using these combined studies, Squire et al. (22) proposed an ultrastructural model for the EGL and its linkage to the underlying ACC. A modied sketch of this ultrastructural model, adapted from Weinbaum et al. (71), is shown in Figure 2a, where the basic organization of the bush-like core protein structures and the underlying ACC are sketched showing the spacing and size of various components. This basic ultrastructural model was then converted into the idealized mathematical model with hexagonal symmetry shown in Figure 2b. Weinbaum et al. (71) use this idealized model to predict the ow in the EGL, the drag forces and bending moments on the core proteins, and the resulting local deformation of the ACC. More recently, Zhang et al. (72) have used this same idealized model to develop a theory for the osmotic ow in the EGL and its reection coefcient. Another important EM study of the EGL came with the work of van den Berg et al. (73), who used a perfusion xation technique followed by Alcian blue staining to stabilize anionic carbohydrate structures on the vessel wall of rat ventricular myocardial capillaries. After xation, capillaries were sectioned and analyzed with EM. The luminal surface of normal capillaries were coated with discrete hairy, bush-like structures that were uniformly distributed over the luminal surface. The observed colocalization of certain lectins to these structures is suggestive of the structures having a saccharine nature rather than merely being crystallization artifacts associated with the xation procedure. The mean thickness of the surface coat in normal capillaries measured between 0.2 and 0.5 m, whereas in capillaries exposed to 1 h of hyaluronidase treatment, the mean thickness measured between 0.1 and 0.2 m. It is noteworthy that while hyaluronidase appeared to have no effect on EC thickness, the interstitial space between capillaries and their surrounding tissue increased from 0.28 m to 0.46 m after hyaluronidase treatment. The authors hypothesize that this increase in the pericapillary space after degradation of the EGL with hyaluronidase is suggestive of a protective role for the EGL (and hyaluronan in particular) in preventing myocardial edema.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

130

Weinbaum

Tarbell

Damiano

Glycocalyx bush structure 20 nm

100 nm 20 nm

150~400 nm

Nucleus
Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

Junctional complex Integrins Actin stress fibers

-actinin

Extracellular matrix Cortical cytoskeleton

b
Actin filaments

Periodic bush structure

Core protein Spacing: 20 nm Diameter: 10-12 nm Cytoskeletal foci Spacing: 100 nm

Figure 2 Structural model for the EGL. (a) Sketch of the arrangement of core proteins in the EGL and its anchorage to the underlying actin cortical cytoskeleton. (b) En face view of the idealized mathematical model in Weinbaum et al. (71) showing the hexagonal arrangement of core proteins and cluster foci. Adapted from Squire et al. (22) and gure 1 in Weinbaum et al. (71).

Relation between Biochemical Composition and EM Observations


One of the most important areas of future study is to create a bridge between the considerable base of knowledge that has been accumulated on the biochemical composition of the EGL (shown in Figure 1) and the structural organization observed

www.annualreviews.org Endothelial Glycocalyx Layer

131

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

in EM (shown in Figure 2). At present, it appears that the EGL may be organized into two layers: an inner region of several tens of nanometers near the apical membrane surface, seen, for example, in the EM observations of Rostgaard & Qvortrop (21), and an outer layer up to 0.5 m thick, which contains the extended core proteins. At the interface between these regions there could be a layer of HA because the membrane-bound molecules, CD44 and CS, have attachment sites for HA and CS has attachment sites for the various core proteins. The fact that HA lies close to the apical surface can be inferred from the fact that Henry & Duling (26) have shown that two hours of hyaluronidase treatment produces only minor changes in permeability and EGL thickness, suggesting that HA is in a deeper protected region near the membrane surface. In contrast, heparinase III, an enzyme that cleaves HS proteoglycans (74), has a dramatic effect on the ECs ability to produce NO (74) and transduce FSS to the actin cytoskeleton (75) (as further discussed in Section 6).

3. MECHANICAL AND ELECTROCHEMICAL PROPERTIES EGL Thickness Using Microviscometric Analysis and -PIV
As mentioned in the Introduction, above, the dye exclusion technique developed by Duling and coworkers (23, 26, 76, 77) was inadequate to study the EGL in vivo in microvessels larger than 1215 m in diameter (26) and it provided no information about the velocity proles near the vessel wall and within the EGL. Smith et al. (27) addressed these limitations by using intravital near-wall -PIV (see Figure 3ac).
Figure 3 (a) Typical dual-ash bright-eld images showing a double exposure of one microsphere ( 0.5 m diameter) near the vessel wall (top image) and near the vessel centerline (bottom image) of a mouse cremaster-muscle venule in vivo. The dual images of the microsphere (encircled in white upstream and black downstream) are separated in time by the double-ash interval. (b) Fluorescent intravital -PIV data in the plasma-rich region of a mouse cremaster-muscle venule (33 m diameter) obtained from images similar to those in (a). Notice that a linear extrapolation of the velocity prole leads to a negative intercept at the vessel wall. A highly nonlinear velocity distribution through the EGL is required to satisfy the no-slip condition at the vessel wall. (c) Hydrodynamically effective EGL thickness (mean and standard deviation) before (10 vessels) and after light-dye treatment (10 vessels), assuming different values of hydraulic resistivity, K, based on near-wall -PIV data similar to (b). (d ) Full-eld intravital -PIV data t to the velocity distribution over the cross-section of a control vessel using nonlinear regression analysis. (e) Using microviscometric analysis, the EGL thickness is estimated from the minimum in the least-squares error, E, in the t to -PIV data (similar to the t shown in d ) before and after light-dye treatment to degrade the EGL. For a particular value of K, an iterative search for the corresponding value of EGL thickness, Ra, that corresponds to the optimal t to the -PIV data in a least-squares sense (i.e. the local minimum for each curve in e) is taken to be the hydrodynamically effective EGL thickness. ( f ) Hydrodynamically effective EGL thickness (mean and standard deviation) before (10 vessels) and after light-dye treatment (9 vessels) assuming different values of K based on full-eld -PIV data similar to (d ) and least-squares error analyses similar to (e). Adapted from Smith et al. (27) and Damiano et al. (29).

132

Weinbaum

Tarbell

Damiano

a b
Distance from vessel wall (m)

1.2 0.8 0.4 0.0 400

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

400

800

Velocity, vz (m/s)

c
Estimated layer thickness (m)
0.5 0.4 0.3 0.2 0.1 0.0 Light-dye treatment Control

d
2000 1500

vz (m/s)

vz (m/s)

1000 500 0

80 40 0

1010

109

108
4

Hydraulic resistivity, K (dyn-s/cm )

8 10 r (m)

12

14

16

e
1.0 0.8 K K = 1010 dyn-s/cm4 1.0

f
0.7 0.6 Light-dye treatment Control

Estimated layer thickness (m)

0.9 0.8 0.7 0.4 0.5 0.6

0.5 0.4 0.3 0.2 0.1 0.0 1010 109

0.6 0.4 0.0

Control

Light-dye treatment 0.1 0.2 0.3

Estimated layer thickness (m)

Hydraulic resistivity, K (dyn-s/cm4)

www.annualreviews.org Endothelial Glycocalyx Layer

133

Their technique involves obtaining measurements of the instantaneous translational speeds and radial positions of FITC-labeled microspheres using dual-ash epi-illumination in an optical section through the median plane of the vessel. The uid-particle translational speeds are inferred from these measurements and a detailed three-dimensional analysis of the local uid dynamics in the vicinity of the vascular endothelium and its EGL (78). Using -PIV data in vessels before and after light-dye treatment to degrade the layer revealed a strongly exponential rather than linear velocity distribution very near the vessel wall in control vessels (see Figure 3c). These results show a signicant effect of the layer on near-wall microuidics and provide the rst direct estimates of the effective hydrodynamic thickness of the EGL in vivo. This thickness was estimated to be between 0.3 and 0.35 m (assuming a hydraulic resistivity, K, of the EGL that is greater than 1010 dyn-s/cm4 ; see Figure 3c). In the initial analysis of Smith et al. (27), the authors assumed a uniform shear eld in the plasma-rich zone (a zone which extends approximately 2 m from the vessel wall), and only considered particle tracers within the plasma-rich zone. In a subsequent study, Damiano et al. (29) show that this assumption leads to a lowerbound estimate of the effective hydrodynamic thickness of the EGL. The full-eld analysis of the uid dynamics over the entire cross section of the microvessel provided by Damiano et al. (29), which was validated in glass tubes and in microvessels in vivo (28), led to a slightly more accurate estimate of the EGL thickness than could be obtained solely from near-wall -PIV data. By taking a full-eld approach, and using -PIV data over the entire cross section (see Figure 3d ), Damiano et al. (29) were able to account for the variation in shear rate throughout the plasma-rich region near the vessel wall, which they found increased monotonically in absolute value with increasing radial position up to the EGL interface. As such, the maximum shear rate, which occurs at the EGL interface, slightly exceeds the mean shear rate within the plasma-rich region used by Smith et al. (27). Extrapolation based on the true velocity gradient at the EGL interface leads to a greater negative intercept at the vessel wall and hence a greater (and more accurate) EGL thickness estimate than does a linear extrapolation based on the mean shear rate in the plasma-rich region. Taking this approach, which they referred to as microviscometry (28), Damiano et al. (29) showed that in mouse cremaster-muscle venules (2040 m diameter), the mean hydrodynamically effective EGL thickness was 0.5 m in control vessels and 0.2 m after light-dye treatment to degrade the EGL (see Figure 3e, f ). The estimates of Damiano et al. (29) are approximately 40%50% greater than the estimates of the effective EGL thickness that Smith et al. (27) found in the control and light-dye treated vessels that they analyzed using near-wall -PIV applied only to the plasma-rich layer.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

Elastic Properties of EGL Fibers


In the elastohydrodynamic theory for the structural integrity of the EGL, the important parameter for determining how the proteoglycan core proteins transmit FSS from the free surface of the EGL to the ACC is the exural rigidity, EI, of the core proteins of the bush-like structures shown in Figure 2. There are no direct

134

Weinbaum

Tarbell

Damiano

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

measurements of EI for the core proteins of EGL proteoglycans analogous to the experiments that have been performed for F-actin, collagen brils, and microtubules (7981). Weinbaum et al. (71) determine EI using an indirect approach in which they predict the measured characteristic time for the time-dependent restoration of the EGL after it has been crushed by the passage of a WBC in a tightly tting capillary, as rst reported in Vink et al. (82). The model in Weinbaum et al. (71) treats the core proteins as elastic bers that are rmly attached by linker molecules to the more rigid actin cortical network beneath the membrane surface, much like bamboo shoots are interconnected by a network of roots beneath the surface of the ground. The model uses small deection theory to predict the long-time nal decay of the bers elastic recoil and neglects the initial short-lived, large deformation of the core protein bers after the passage of the WBC. The hydrodynamic interaction between the bers is considered, but the motion of the entrained uid is neglected. This linearized model predicts that EI is 700 pN nm2 , about 1/20 of the measured value of EI of an F-actin lament, 1.3 103 pN nm2 (80). Because the diameter of a core protein is approximately one-half that of an actin lament, and the moment of inertia, I, of the cross section varies as the fourth power of the radius, this is a reasonable prediction if values of E are comparable. More recently, Han et al. (83) have developed a more sophisticated large deformation model for elastica that predicts the time-dependent changes in shape of the core proteins from their initial large deformation to their nal restoration after the passage of the white cell. This more realistic model predicts that EI is 490 pN nm2 , surprisingly close to the highly idealized linear elastic model rst proposed in Weinbaum et al. (71). This large deformation theory is described in more detail in Section 4.

Electrochemical Properties
Motivated by results reported by Vink & Duling (76) on the clearance rates of charged and neutral plasma tracers from the EGL, Stace & Damiano (84) developed an electrochemical model of the EGL to gain a quantitative understanding of the transport of negatively charged molecules through the layer. This analysis also provides important new insights into the electrochemical composition of the layer. In particular, one is interested in devising a method for quantifying the xed-charge density of the EGL in vivo. The Stace & Damiano model consists of a quaternary mixture with solid, water, cationic, and anionic constituents in which xed negative charges are bound to the solid matrix. Their model is consistent with the triphasic theory developed by Lai et al. (85) in the absence of an elastic-restoring capability by the ber matrix and in the limit of small solid-volume fractions. The interaction of the xed negative charges with cations and anions in the blood leads to a charge distribution in equilibrium that achieves spatial charge neutrality globally but with local imbalances in charge that give rise to a double layer at the EGLplasma interface. Consideration of the equilibrium conguration in the model of Stace & Damiano (84) shows that polyanionic plasma tracers would be partially excluded by the EGL in equilibrium by virtue of their charge. It also quantitatively predicts the degree to which this exclusion would occur as a function of tracer valence, EGL xed-charge density, and

www.annualreviews.org Endothelial Glycocalyx Layer

135

the ionic strength of blood plasma. This exclusion would result in a reduction in uorescence intensity within the EGL relative to luminal intensity levels, and could therefore be detected using epiuorescence illumination in vivo. Thus, if the valance of a polyanionic plasma tracer were known, then a quantitative measure of the diminution of uorescence intensity in the EGL relative to the luminal intensity level would provide the information necessary to estimate the xed-charge density of the EGL in vivo. Vink et al. (86) compared the intensity distributions of neutral dextran plasma tracers (40 kDa, Texas red) with polyanionic uorescein-labled dextran tracers (40 kDa, FITC dextran sulfate) over the cross section of mouse cremaster-muscle venules in vivo. They showed a 30%50% attenuation in the anionic tracer intensity relative to that of the neutral tracer within 0.5 microns from the vessel wall. According to the electrochemical model of Stace & Damiano (84), such an exclusion corresponds to an EGL xed-charge density of between 0.7 and 1.3 mEq/l, which is approximately 1% of the ionic strength of blood plasma. The electrochemical model of Stace & Damiano (84) was later extended by Damiano & Stace (87) to investigate the effect of the EGL xed-charge density on charge-mediated mechanical recovery of the layer after deformation induced by a passing leukocyte in a capillary (see Section 4). In another analysis that investigated the electrochemical transport of charged molecules through the microvascular wall, Fu et al. (88) extended their earlier partition model of the interendothelial cleft to include a negatively charged EGL at the cleft entrance. Their analysis examined solute transport under steady-state conditions and did not consider transients. From their analysis, they estimated that the EGL xed-charge density was 2535 mEq/l. To make this estimate, they considered four different xed-charge density distributions across the EGL and used the results of Adamson et al. (89), wherein the permeability of frog mesenteric capillaries to a positively charged globular protein was found to be twice that of a negatively charged globular protein having nearly the same molecular weight. This estimate is 30-fold greater than the previous xed-charge density estimate of Stace & Damiano (84) and Damiano & Stace (87). While model assumptions of the Stace & Damiano model are slightly different from those of Fu et al. (88), the main source of this discrepancy may lie in the fact that the estimate of Fu et al. was based on the permeability measurements from frog mesenteric capillaries, whereas the estimate of Stace & Damiano was based on experiments in hamster cremaster-muscle capillaries. It is certainly possible that both estimates are valid in the context of their respective applications. However, it is noteworthy that estimates of EGL xed-charge density in mammalian microvessels must be consistent with the observations of Vink & Duling (23) and Han et al. (83). In particular, the EGL must be sufciently soft so as to become completely compressed by red cells brought to rest in 5-m-diameter capillaries when ow is occluded and yet at the same time be capable of restoring its equilibrium dimensions within 1 s after compression by a leukocyte (see Section 4). According to the mechano-electrochemical model of the EGL of Damiano & Stace (87), who were able to use the calculations of single-le red-cell motion in EGL-lined capillaries presented by Secomb et al. (13) to place bounds on the EGL xed-charge density,

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

136

Weinbaum

Tarbell

Damiano

an EGL xed-charge density greater than 1 mEq/l would not be consistent with the observations of Vink & Duling (23) and Han et al. (83). This need not necessarily be the case, however, for the EGL in amphibian capillaries.

4. MODELS FOR STRUCTURAL INTEGRITY AND RESTORATION


It is clear from intravital imaging and immunouorescence studies that the thickness of the EGL is nearly uniform and changes little during the passage of a red cell once the cell is moving at a fast enough velocity to lift off of the layer. In marked contrast, the experiments in Vink et al. (82) and Han et al. (83) show that the EGL can be crushed to as much as 20% of its undeformed thickness by the passage of a WBC through a capillary. As noted above, the structure is both light sensitive and easily compressible, as evidenced by the fact that a red cell brought to rest will expand to ll nearly the entire lumen of the capillary (23). Because the membrane of the red cell is highly exible, it might seem surprising that the ber structure of the EGL can both so easily collapse and yet be able to restore itself to its undeformed thickness typically within 0.5 s after the passage of a WBC. Three different biophysical models will be presented in this section to describe this paradoxical behavior. A second fundamental paradox is that all three models predict that the FSS at the edge of the EGL is greatly attenuated by the ber matrix, with the result that the FSS at the level of the EC membrane is vanishingly small. Thus, for each model a critical question that needs to be addressed is how FSS is transmitted to the intracellular cytoskeleton. This is one of the central issues in Section 6 on mechanotransduction.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

Hydrodynamic Models for Flow in the EGL


Among the earliest analytical attempts to investigate ow through the EGL was the work of Barry et al. (90). Their work considered steady and unsteady ow of a Newtonian uid in a channel lined with a poroelastic wall layer. Wang & Parker (91) applied mixture theory and two-dimensional lubrication theory to the problem of a sphere falling through a quiescent uid in a cylindrical tube lined with a deformable porous wall layer. This was followed by Damiano et al. (10), who obtained analytical solutions for axisymmetric pressure-driven ow of rigid close-tting particles in a cylindrical tube with a poroelastic layer lining its luminal surface. In all of these studies, ow through the EGL was modeled using the equations of binary mixture theory where the solid volume fraction was taken to be small. In the limit of a vanishing solid volume fraction, these equations degenerate to the Brinkman equation for a solenoidal uid velocity eld given by v K v = p v = 0, (1) where v is the uid velocity vector, p is the pressure eld, and the parameters and K denote the uid dynamic viscosity and hydraulic resistivity, respectively. K is related to the commonly used Darcy permeability, k, by K = /k.

www.annualreviews.org Endothelial Glycocalyx Layer

137

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

The important dimensionless quantity for characterizing drag through the EGL is 2 = /(h2 K ), where h is the mean effective thickness of the layer. This quantity characterizes the ratio of viscous drag forces associated with uid-velocity gradients in the EGL to permeation-induced viscous drag forces associated with the uid motion relative to the solid constituent of the EGL. For small values of , permeation-induced viscous drag forces dominate and balance the pressure gradient throughout the EGL, except very near the plasma interface where uid velocity gradients become large. Thus, for small values of , a Darcy ow prevails away from the plasma interface. In particular, the quantity (/K )1/2 = h is the characteristic exponential decay length (with respect to distance into the layer) of the axial velocity at the plasmaEGL interface, i.e., it characterizes the depth of penetration of axial ow through the EGL. Because it is likely that the solid volume fraction s 1 (92), it is reasonable to take the dynamic viscosity of the layer to be equal to that of blood plasma (0.01 dyns/cm2 ). Estimates of the hydraulic resistivity, K, of the EGL, based on the recovery time of the layer in capillaries in the wake of a passing WBC, place it between 1010 and 1011 dyn-s/cm4 (71, 87, 93), which puts 2 between 104 and 103 for an EGL thickness of 0.5 microns. It is noteworthy that a value of K > 1010 dyn-s/cm4 so severely attenuates ow throughout the EGL that FSS on the EC surface is effectively eliminated. This has important implications for understanding mechanisms of EC mechanotransduction, which is further discussed in Section 6.

Models for the Restoring Mechanisms of the EGL


Whereas there has been a general consensus as to how to best mathematically describe uid-dynamical drag through the EGL, a variety of hypotheses have emerged as to the origins of the restoring force mechanism that prevents the deformation of the layer owing to FSS in the physiological range and allows its recovery after compression by a passing WBC. Most prominently these include the oncotic models of Secomb and coworkers (12, 13, 94), the elastohydrodynamic models of Weinbaum and coworkers (71, 83), and the mechano-electrochemical model of Damiano and coworkers (87, 93). Oncotic model. Pries et al. (9) rst introduced the concept that even a minute difference between the concentration of plasma proteins adsorbed to the EGL and that of free plasma proteins in the capillary lumen would be sufcient to generate an oncotic (or colloid osmotic) pressure that would be sufcient to exclude red blood cells from the EGL region near the vessel wall under standard ow conditions in capillaries. This concept was then applied in Secomb et al. (12) to develop a model for single-le red blood cell motion in an EGL-lined capillary. They argued that the source of the restoration capability of a compressed EGL was not mechanical (elastic) but chemical and suggested that, in equilibrium, the excess oncotic pressure within the EGL relative to the plasma would be balanced by tension in the EGL matrix. Compression of the layer would result in a further increase in this pressure differential and a relaxation of the tension in the EGL matrix. Upon recovery of the layer back to its equilibrium conguration (after the externally applied load is removed), tension would mount

138

Weinbaum

Tarbell

Damiano

within this matrix until a balance is reestablished between macromolecular tension within the EGL matrix and the oncotic pressure within the layer. In their initial model for the single-le axisymmetric motion of deformable red cells in an EGL-lined capillary, Secomb et al. (12) assume that the red cells do not invade the EGL and that the restoring forces in the red-cell membrane are assumed to arise from membrane shear elasticity and isotropic membrane tension, but the bending elasticity of the membrane is neglected. The ow through the EGL is modeled using the momentum equation given by Damiano et al. (10), which is a special case of the Brinkman equation described in the previous subsection when ow in the capillary is axisymmetric and unidirectional. The pressure in the lubrication layer between the red cell and the vessel wall combines with the oncotic pressure imposed by the EGL to determine the magnitude of the normal component of the traction vector acting on the red-cell membrane. Their calculations assumed an oncotic pressure in equilibrium of 200 dyn/cm2 , which is approximately 8% of the typical oncotic pressure of the plasma (2500 dyn/cm2 or 25 cm H2 O). In a subsequent study, Secomb et al. (13) rened their previous analysis to include a more realistic model of the red-cell membrane (one which included the effect of bending elasticity) and considered the problem of the red cell invading the EGL as the ow rate in the capillary vanishes. Results of this analysis showed that an oncotic pressure of only 20 dyn/cm2 in equilibrium was large enough to exclude red blood cells (but not WBCs) from the EGL, and yet was small enough to allow the gap between the red cell and the capillary wall to approach zero thickness when the velocity vanished, as had been observed experimentally (23). At rst glance, this model for oncotic tension in the EGL may seem hard to reconcile with the models for the oncotic pressure in the EGL that are presented in Section 6 describing the revised Starling principle. In the case of the latter models, the EGL acts as the primary molecular sieve for plasma proteins, which results in a substantial drop in oncotic pressure across the EGL (the magnitude of which depends on the luminal pressure or ltration rate). At normal ltration rates in capillaries, the oncotic pressure across the EGL drops from approximately 25 cm H2 O (2500 dyn/ cm2 ) to 10 cm H2 O (1000 dyn/cm2 ). This oncotic pressure drop far exceeds the 20 dyn/cm2 oncotic pressure increase of the adsorbed proteins estimated by Secomb et al. (13), thus placing the EGL in a state of compression rather than tension. This difculty can be explained if the 100-nm-thick layer observed by Squire et al. (22) using EM (where a quasi-periodic structure was observed) functioned as the molecular sieve for plasma proteins and this was coated with a signicantly thicker and less organized structure that functioned as the effective hydrodynamic layer that is measured in Smith et al. (27) and Damiano et al. (29). This outer layer is often referred to as the endothelial surface layer, in contrast to the EGL where the oncotic sieving is assumed to occur. A small oncotic pressure of 20 dyn/cm2 would be sufcient to keep this outer layer hydrated and provide the lift force for red cells during the pop-out phenomenon modeled by Feng & Weinbaum (95) and Secomb et al. (13). Viewed from this perspective, the much larger oncotic pressures in the EGL could be resisted by the elastic properties of the core proteins in the elastohydrodynamic model that is presented next.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

139

Elastohydrodynamic model. In the elastohydrodynamic model, the core protein bers have a nite exural rigidity, EI, which provides (a) for their ability to withstand collapse owing to the decrease in oncotic pressure required for them to function as a molecular sieve for plasma proteins and (b) for their ability to withstand FSS in the physiological range without substantially deforming. This structural resistance to deformation in response to FSS is observed in intravital microscopy studies (23, 26, 77) and microviscometric studies (28, 29) of EGL thickness. The linear elastohydrodynamic model of Weinbaum et al. (71), described in Section 3 to predict EI, treats only the nal portion of the restoration of the EGL after its initial nite compression by the passage of a WBC. As noted previously, the predicted value of EI was 700 pN nm2 . A much more sophisticated nonlinear elastohydrodynamic model is presented in Han et al. (83) that uses large deformation theory for elastica and a modied Brinkman equation to describe the local relative motion of the bers and the uid. The analysis also takes into account the uid that is imbibed as the EGL expands and uses local hydrodynamic resistance coefcients to describe the local time-dependent hydrodynamic forces parallel and perpendicular to the ber motion. The experimental part of the paper greatly expands on the brief abstract of Vink et al. (82), in which results were rst reported for the time-dependent restoration of the EGL after the passage of the WBC. The nonlinear model in Han et al. (see Figure 4a) predicts that there are two phases for the ber recoil after the passage of a WBC, an initial phase for large compressions, where the EGL is less than 0.36 ho (for an undeformed thickness, ho ) and the ends of the core proteins overlap and are parallel to the endothelial membrane, and a second phase where the bers assume a shape that is close to the solution for an elastic bar with linearly distributed loading. As seen in Figure 4a, the transition from phase I to phase II occurs at t = 0.041 s if the EGL is initially compressed to a thickness of 0.2 ho . Figure 4b compares the predictions of the elastrohydrodynamic model in Han et al. (solid curve) with the experimental data for the time-dependent restoration of the EGL (solid circles) and the predictions of the mechano-electrochemical model of Damiano & Stace (87) (open triangles), which is discussed in the next subsection. The predictions of Damiano & Stace are rescaled by a factor of 4900 and shifted to the right by 0.01 because the starting point of their calculation is h/ho = 0.5, where ho = 400 nm. The predictions of the two models are too close to determine which model has greater validity. The principal difference in behavior is that the elastohydrodynamic model does not asymptotically approach the undeformed height for long time, as is the case for the mechano-electrochemical model, but instead it abruptly reaches h/ho = 1.0 at t/ = 0.05, where = 14 s is a dimensional scaling time related to the exural rigidity EI. The surprisingly close agreement in the predictions for EI of the linear and nonlinear elastica models (700 versus 490 pN-nm2 ) is due to the fact that the large deformation portion of the recovery is relatively short lived. Mechano-electrochemical model. Damiano & Stace (87) proposed a model for the structural integrity of the EGL that consists of a mixture of electrostatically charged macromolecules hydrated in an electrolytic uid. Their mechano-electrochemical

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

140

Weinbaum

Tarbell

Damiano

1.0
0.70 s 0.60 s

1.5

0.8
0.36 s

0.6

1.0 Phase II

y/ho
0.4

0.16 s

h/ho
0.5

0.041 s 0.014 s 0s

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

0.2

Phase I

0.2

0.4

0.6

0.8

1.0

0.05

0.10

0.15

x/ho

t/
Model predictions (Reference 83) Experimental results (Reference 83) Model predictions (Reference 87)

Figure 4 (a) The time-dependent change in shape of the core protein bers. y/ho and x/ho are scaled vertical and horizontal coordinates of the ber position, and ho , the undisturbed ber-layer thickness, is also ber length. Note the change from phase I to II occurs at 0.041 s. (b) Comparison of model predictions in Han et al. (83) (solid line) and the predictions (open triangles) derived from gure 2b in Damiano & Stace (87) with the experimental results in Han et al. (83) (solid dots). The ESL thickness is normalized by 400 nm, the undisturbed ESL thickness, ho (b). Adapted from gures 6 and 7A in Han et al. (83).

model considers one-dimensional nite deformations of the layer in the absence of an axial ow in the vessel. As in the electrochemical model of Stace & Damiano (84) described in Section 3, Damiano & Stace (87) model the EGL as an isotropic quaternary mixture. The constitutive law for the ux of the solid and ionic constituents is modeled using the Nernst-Planck equation, where it is assumed that these constituents have negligible volume fractions, and the viscosity of the uid constituent is also negligible. According to their analysis, EGL recovery after deformation is driven by an electrochemical potential gradient, which consists of electrostatic and chemicalgradient components. The former accounts for the interactions of the xed-bound charges on the matrix with the counter ions in solution, whereas the latter arises as a result of the higher concentration of bound GAGs, glycoproteins, and proteoglycans in the EGL relative to the concentration of corresponding blood-borne constituents in the lumen. Leukocyte-induced deformations of the EGL result in disturbances away from a nearly electroneutral equilibrium environment. In the wake of the leukocyte, uid ux into the compressed layer, driven by gradients in the electrochemical potentials of the ions and macromolecular matrix, results in rehydration of the EGL

www.annualreviews.org Endothelial Glycocalyx Layer

141

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

and a restoration of its equilibrium dimensions. It is assumed that the matrix has no elastic restoring capability under compression but can support a tensile stress that balances the electrochemical potential gradient in the equilibrium conguration. The two most important parameters in this model for determining the recovery time of the EGL after compression by a leukocyte are the dimensionless xed charge density, 0 , and the dimensionless macromolecular matrix concentration, F (the dimensional counterparts of 0 and F are nondimensionalized with the concentration of mobile cations in blood). The relative size of these two quantities determines which is the more dominant physical mechanism. Asymptotic and numerical analyses of their model reveal that the 90% recovery time, , is given in terms of 0 , F, and the dimensionless equilibrium thickness of the EGL (see Figure 5). As is evident in Figure 5, the model predicts that the response of the EGL falls into two distinct regimes. When 0 2 F, the recovery time is essentially independent of 0 and the chemical-potential gradient is the dominant restoring force. This mechanism is similar to the oncotic model of Secomb and coworkers (12, 13, 94), described above, except that the chemical potential gradient is derived not from adsorbed plasma proteins but from GAGs bound to the EGL. On the other hand, when 0 2 F, the electrostatic potential gradient dominates and the recovery time varies in proportion to 0 2 . Based on the 1 s recovery time of the EGL, and the value of K on the order of 5 1010 dyn-s/cm4 ,

100

F = 10-7

Finite difference solution Asymptotic equation

Recovery time, (s)

10

F = 1.22 x 10-6 F = 10-5 F = 10-4

0.1

0.01

10-5

10-4

10-3

10-2

10-1

0
Figure 5 The 90% recovery time as predicted by the mechano-electrochemical model of Damiano & Stace (87) as a function of the dimensionless xed-charge density, 0 , for different values of the dimensionless macromolecular matrix concentration, F. Results correspond to a hydraulic resistivity, K = 1011 dyn-s/cm4 , a cation concentration of 0.14 M (which is equal to the concentration of mobile cations in blood), a capillary radius of 2.5 m, and an EGL thickness of 0.5 m. Notice that for all values of F < 105 , a 1 s recovery time corresponds to 0 = 0.01, or equivalently, a dimensional xed-charge density of the EGL of 1 mEq/l. Adapted from gure 4 in Damiano & Stace (87).

142

Weinbaum

Tarbell

Damiano

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

a xed-charge density of the EGL of 1 mEq/l would result in the magnitude of the stress traction exerted by the layer on passing red cells of 2030 dyn/cm2 , which is consistent with the magnitude necessary to account for the exclusion of red cells by the EGL observed in skeletal muscle capillaries in vivo (13). This corresponds to approximately one xed-bound charge on the EGL for every 100 ions in blood. It is noteworthy that such a charge density would result in a voltage differential between the undeformed EGL and the capillary lumen of 0.1 mV. In a subsequent study, Damiano & Stace (93) developed a detailed model of the uid dynamics and EGL deformation in the wake region behind a leukocyte moving through a capillary. Their analysis details the radial and axial velocity components arising from an axisymmetric pressure-driven ow of a linearly viscous uid within the wake region and throughout the EGL. The analysis establishes the necessary relationships to connect the results of the one-dimensional mechano-electrochemical model of Damiano & Stace (87) with the results of a more realistic approximation of the ow regime arising in the wake of a WBC passing through a capillary.

5. CELLULAR INTERACTIONS WITH THE EGL


The early theoretical studies mentioned in the Introduction on the deformation of a moving red cell in a capillary either neglected the presence of the EGL or described the simpler case where the red cell did not enter the EGL. The experimental studies of Vink & Duling (23) clearly show that a red cell brought to rest will expand to ll the entire lumen of the capillary, thus crushing the EGL, and that when motion resumes, the cell will exhibit a pop-out phenomenon in which the cell will rise through the EGL as its velocity is increased until it reaches a critical velocity, at which time the cell will lift off of the EGL and create a thin lubricating layer between it and the EGL. In the case of the WBC, both its free interaction with the EGL and its tethered rolling in the presence of the EGL are of interest. To address these interactions, one needs to calculate the forces generated on the WBC microvilli tips for both types of encounters with the EGL as well as the penetration resistance of the layer. Both of these problems are examined in this section.

Red Cells
The rst realistic attempts to analyze the single-le motion of red blood cells in EGL-lined capillaries rst appeared in Damiano (11) and Secomb et al. (12) (as described in Section 4), who adopted similar approaches to elucidate the effect of the EGL on capillary resistance and capillary tube hematocrit. The apparent viscosity and capillary tube hematocrit predicted by these models showed strong sensitivity to the presence of an EGL, and predicted that a 0.5-m-thick hydrated macromolecular layer lining the capillary wall could result in a marked increase in capillary resistance and a signicant reduction in capillary tube hematocrit relative to model predictions of blood ow through smooth-walled glass tubes. To maintain continuity of mass, the retarded ow through the EGL demands greater red-cell clearances and results in more elongated red cell shapes in capillaries than in smooth glass tubes of the same

www.annualreviews.org Endothelial Glycocalyx Layer

143

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

diameter. Both models were valid only in a high-velocity limit (>1000 m/s), where red cell deformation is at a maximum and red cell shapes are nearly independent of ow velocity (96). However, in skeletal muscle capillaries, red cell velocities are often in the range of 100 to 200 m/s in steady ow and can drop to zero during intermittent ow conditions. For these situations, restoring forces in the red cell membrane are of the same order as FSS and pressure forces at the cell surface. This results in broader and shorter cell shapes with decreasing velocity. Secomb et al. (13) addressed this limitation by rening the red cell model, as in Secomb et al. (96), to include the effect of bending elasticity in the red cell membrane (see Section 4). With this generalization, they were able to consider red cell velocities down to 1 m/s. In addition, Secomb et al. (13) were able to model the so-called pop-out phenomenon, rst analyzed in Feng & Weinbaum (95), by considering transient changes in red cell shape as a function of velocity for a step increase in driving pressure across a red cell that is initially nearly stationary. Feng & Weinbaum (95) identied an important uid-dynamical lift force that will arise within the porous EGL matrix owing to the axial ow of plasma through the layer. To generate this lift force, a conning boundary or bearing surface (e.g., the red cell membrane) translating at a small angle relative to the vessel wall, such as might be provided by a passing red blood cell in a capillary, is required to prevent uid leakage out of the layer into the free luminal space of the vessel. Feng & Weinbaum draw an intriguing analogy between a human skiing on snow powder and a red cell gliding on the EGL. They show that the dimensionless parameter, h/ k (where h is the EGL thickness and k is the Darcy permeability), describing the uid motions in both porous layers is of the same order of magnitude, although a human and red cell differ in mass by 15 orders of magnitude. Such conditions can certainly arise in capillaries, particularly upon the resumption of red cell motion after a ow-cessation event. In light of this idea, and by virtue of the relatively well-understood mechanical properties of red blood cells, Secomb et al. (13) were able to provide a quantitative description of the ow-dependent exclusion of red cells from the EGL, which placed approximate bounds on important mechanical properties of the EGL.

White Cells
The adhesion of circulating leukocytes to vascular ECs during inammation and immune surveillance is initiated by the adhesive tethering of L-, P-, and E-selectins and 4 integrins to ligands that are located either on the tips of the WBC microvilli or the EC surface. This has stimulated an extensive literature on the mechanics and kinetics of tethered rolling of WBCs on coated glass substrates in which a host of ligand-receptor interactions have been studied, starting with the early work of Lawrence & Springer (97). Similarly, there is an extensive literature on the tethered rolling of WBCs on ECs grown in culture media where the state of the EGL is either unknown or not expected to be preserved. To our knowledge, there are no experiments that have explored the role of the EGL in regulating adhesive tethering or as a modulator of free rolling without adhesion. This is surprising because we shall see in Section 6 that the EGL plays a central role in mechanotransduction and can

144

Weinbaum

Tarbell

Damiano

have a profound inuence on the cytoskeletal and biochemical responses of cultured EC monolayers to FSS. In contrast, the role of the EGL in leukocyte adhesion is an uncharted area of research. A critical question in the case of WBC rolling is whether the microvilli of a freerolling WBC will be able to penetrate the EGL and initiate tethered rolling attachments. The quantitative feasibility of this occurring has only been examined theoretically (98). Tethered rolling is frequently observed in vivo in postcapillary venules, where one would anticipate that an intact EGL exists, but this rolling always seems to be initiated at the entrance to a postcapillary venule where a tightly tting WBC has just emerged from a capillary (99). At this location, the microvilli have been in close contact with the EC surface, the WBC having crushed the EGL during its passage through the capillary (82, 83). Because the length of WBC microvilli is typically 0.3 to 0.7 m (100), and thus, comparable to the thickness of the EGL (23, 29), the ability of a WBC to roll freely through microvessels needs to be questioned. Cell rolling would be quickly curtailed if the microvilli functioned as human legs trying to cross a eld of fresh fallen snow whose depth was the same as their length. Because there is no evidence that WBCs have the equivalent of snowshoes, one wonders whether a WBC can tip-toe across the EGL much like a Jesus Christ lizard runs across water. For the case of a free-rolling WBC on a cultured endothelial monolayer, one needs to consider sedimentation owing to weak gravitational forces. For a WBC whose specic gravity is 1.1, this sedimentation force is 0.3 pN. However, Zhao et al. (98) show that this weak gravitational force can be amplied 20- to 100-fold at the microvilli tips owing to viscous forces generated in the thin lubricating layer between the body of the WBC and the rolling surface. In vitro WBCs appear to roll as if they were displaced by a distance of 500 nm from solid boundaries (101, 102), suggesting that the microvilli behave as stiff protuberances on the time-scale of rolling contact, which is typically approximately 0.2 ms at a FSS of 5.0 dyn/cm2 . This stiffness is supported by the experiments of Shao et al. (103), who show that the viscoelastic response of a microvillus under stretch is 0.77 s, at least three orders of magnitude longer than the very brief rolling contact times. Early studies on the free rolling of WBCs as a precursor to tethered rolling on solid substrates examined the rolling interaction of single microvilli (101, 104, 105) but not the response of multiple microvilli, which typically exceed 20 in a cross-section (100). The latter was explored for the rst time in Zhao et al. (98). These authors predicted that for equal length microvilli, the contact force would increase from 2 pN at a FSS of 1 dyn/cm2 to 5 pN at a FSS of 10 dyn/cm2 . For the small population (5%10%) of long microvilli (>0.5 m) in a heterogeneous microvilli array, this force would increase from 8 to 30 pN at these same FSS. Note that the contact forces do not scale linearly with the FSS owing to the nonlinear coupling that arises from the sedimentation between contacts. When the WBC experiences tethered rolling in vivo or in vitro, the weak gravitational forces are insignicant because Zhao et al. show that the normal forces generated at the microvilli tips are at least an order of magnitude greater than they are during free rolling. At a FSS of only 1 dyn/cm2 , normal contact forces are already of the order of 100 pN and deep penetration of the EGL is easily achieved even at these relatively low FSSs, as discussed below.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

145

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

To determine the depth of penetration of the microvilli, Feng et al. (106) have developed a theory for the motion of a sphere in a Brinkman medium perpendicular to a planar wall as an idealized model for the normal penetration of the microvilli tips. The microvilli tips are treated spheres (typically 0.1 m in diameter). The diffusional resistance of 40-nm-diameter gold particles attached to lipids in the membrane was measured by Lee et al. (107). The theory in Feng et al. is used to scale the resistance of a 0.1-m-diameter sphere to this measured resistance. Feng et al. estimated that a microvillus would penetrate the EGL at a velocity of 6 m/s under the action of a 1 pN force. Taking account of the nonlinear change in contact time with FSS, Zhao et al. predict that the depth of penetration at a FSS of 10 dyn/cm2 would be 1.0 nm for equal-length microvilli and 20 nm for individual longer microvilli of heterogeneous length. The short penetration depth is due to the very short tip contact times, which are 0.1 ms at this shear stress. However, for a FSS of only 1 dyn/cm2 , the penetration depth could be as much as 200 nm for the longer microvilli owing to the longer contact time. This suggests that at least on the arterial side of the circulation, where the FSS exceeds 10 dyn/cm2 , the WBC may, in fact, behave like a Jesus Christ lizard during free rolling, even in the presence of weak gravitational forces. This could explain why tethered rolling is seldom seen on the arterial side of the circulation except in regions of very low FSS, such as in recirculation zones. In sharp contrast, the long microvilli on a WBC experiencing tethered rolling in vivo, where the normal penetration force is at least an order of magnitude greater at the same FSS, would easily penetrate the entire thickness of the EGL and form new tethering attachments. It is for this reason that once tethered rolling is initiated at the entrance to a postcapillary venule, such rolling can often be sustained for the entire length of the venule.

6. PHYSIOLOGICAL FUNCTIONS Permeability and the Revised Starling Principle


The classic Starling equation describing the balance of hydrostatic and colloid osmotic (oncotic) forces across the capillary wall includes four forces, Jv /A = Lp [(Pc Pt ) (c t )], where Pc and Pt are the hydrostatic pressures in the capillary lumen and tissue, respectively; c and t are the corresponding lumen and tissue oncotic pressures; Jv /A is the ltration rate per unit area of capillary wall; is the reection coefcient to the plasma proteins; and Lp is the hydraulic conductivity. This relation has been tested on numerous occasions in both whole organs and isolated microvessels, but nearly always under conditions where the tissue protein concentration was low owing to wash-down after rapid ltration or washout in exposed superfused tissue. Rarely has the tissue been backloaded to examine the effect of t . Furthermore, as noted in the Introduction, considering the most accurate measurements of t , most tissues cannot sustain venous reabsorption and one cannot account for measured lymph ows (31). A resolution to this important paradox has been proposed by Michel (30) and Weinbaum (33). They hypothesized that the effective osmotic barrier is not the whole

146

Weinbaum

Tarbell

Damiano

capillary wall, as previously assumed, but just the luminal EGL, which acts as the molecular sieve for plasma proteins. Moreover, if there is a tight junction (TJ) strand in the cleft with infrequent breaks, back diffusion into the luminal side of the cleft owing to the convective ux of water through these breaks can be severely limited, especially at high ltration ows. For this reason, the tissue oncotic pressure t can differ greatly from the oncotic pressure (0) in the protected region between the luminal side of the TJ strand and the back side of the EGL. Similarly, because a large fraction of the hydraulic resistance is due to the TJ strand, the hydrostatic pressure behind the EGL, P(0), can differ greatly from the hydrostatic pressure Pt in the tissue.
Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

TJ: tight junction

b
Lf Ac x0 2h x2

pc

x1

2d 2D

L1 = x1 x0 Lc = 0.5 pc Ac-1

L2 = x2 x1

L = x2 x0

Surface glycocalyx, EGL

Pc x = -Lf = 150 nm y x=0 x = Lf = 67 nm

2d Region A Junction strand Cleft exit

L1 L2 x y=0

x = L = 411 nm y=D

2D Region B LB = 5 m

Tissue Space

Pt

Figure 6 Schematic of a microvessel (a) showing cross-section of an interendothelial cleft covered by EGL and (b) three-dimensional view of TJ strand in plane of cleft (c). (d ) Idealized mathematical model showing four regions: surface glycocalyx (EGL), cleft with TJ strand (region A), near-eld tissue space <5m (region B), and far-eld tissue space >5 m from cleft exit. Adapted from gure 1 in Reference 37.

www.annualreviews.org Endothelial Glycocalyx Layer

147

This situation is illustrated in Figure 6, which shows a capillary with a cleft and EGL in cross-section (Figure 6b) and a TJ strand with infrequent breaks of length 2d and average spacing 2D (Figure 6c). The idealized mathematical model used to describe this geometry is shown in Figure 6d where the dimensions are taken from the detailed measurements in Adamson et al. (37) for rat mesenteric microvessels. In the new Starling equation, rst proposed in Weinbaum (33) to describe the revised Starling principle, t is replaced by (0) and Pt by P(0) because the Starling force balance is applied just across the EGL. This new equation is much more difcult to apply than Equation 2 because (0) and P(0), which are evaluated at the cleft entrance behind the EGL, can vary signicantly along the length of the cleft because the streamlines for water ow and solute ux lines follow a curved path through the TJ breaks, as illustrated in Figure 6d. This is particularly signicant for uid streamlines because a large pressure drop can occur in the narrow protected channel in the cleft on the luminal side of the TJ strand. Both (0) and P(0) are unknown and must be determined by solving a coupled three-dimensional boundary value problem in which boundary and matching conditions are applied across three regions: the EGL where the revised Starling equation is applied, region A (the cleft), and region B (the tissue space). The rst detailed three-dimensional solutions for this model were presented in Hu & Weinbaum (35), predicting that at high-ltration rates the local Peclet number, Pe, at the TJ breaks would be greater than 1 and the concentration of proteins in the protected region behind the EGL would be nearly independent of the tissue concentration, t . As the lumen pressure decreased, Pe would decrease and back diffusion from the tissue would occur, allowing (0) to rise. The foregoing theoretical predictions were tested in two sets of experiments, one performed on frog mesenteric microvessels (36) and a second on rat mesenteric microvessels (37), where detailed electron microscopic reconstructions of the cleft were also obtained for input into the theoretical model. In both sets of experiments, the tissue was backloaded with albumin (by damaging the mesothelium at a distance of approximately 100 m from the perfused microvessel) and a superperfusate was introduced that was isotonic (50 mg/ml) with respect to the lumen. The time-dependent changes in tissue concentration were carefully measured using confocal microscopy. Experimental measurements of Jv /A were not performed until the lumen and tissue were observed to be isotonic. According to the classical Starling equation, there should be no oncotic force across the vessel wall. However, measurements of in frog microvessels when Jv /A = 0 showed that nearly the full lumen oncotic pressure was present despite the fact that confocal microscopy indicated that the interstitial albumin concentrations were isotonic to within a few microns of the vessel wall. Similarly, in rat microvessels, where the fractional length of the breaks in the TJ was larger, the measurements of were 70% of the lumen oncotic pressure. Both experimental results were accurately predicted by the three-dimensional theoretical model in Hu & Weinbaum (35), as sketched in Figure 6d. Even at relatively low lumen pressures, the model predicts and the experiments conrm that t and (0) can differ substantially. These surprising ndings show that there is a large asymmetry in the effect of the Starling forces at the lumen and tissue fronts owing to the nonlinear effects of convection on the solute transport through the breaks in the TJ strands. In

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

148

Weinbaum

Tarbell

Damiano

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

view of the highly unusual nature of these ndings, several featured perspectives have appeared in highlights from the physiological literature summarizing the importance of this new view of the Starling hypothesis (34, 108). This same behavior has been observed in cultured bovine aortic EC monolayers (38). The proposed structure of the EGL depicted in Figure 1 has also been a catalyst for a major extension from the classic study on the osmotic ow across membranes with long circular cylindrical pores (109) to a ber matrix layer (72). One of the widely used results of this classical analysis is the expression for the reection coefcient of a spherical solute in a porous membrane, = (1 )2 , where is the partition coefcient. This analysis, and the expression for , have been used extensively in many investigations of porous media that have applied the ber matrix theory of Curry & Michel (110). Intuitively, investigators have thought that because the ow between bers is Poiseuille-like, the analysis for a ber matrix layer would be equivalent to that for circular pores. However, Zhang et al. (72) show that there is a fundamental difference because the exclusion layer where the potential energy function describing the solute distribution is satised corresponds to the uid annulus surrounding the ber. Thus, the exclusion layer for the solute molecule corresponds to the exterior of a circular cylinder rather than the interior of a cylinder, as is the case for a circular pore. Each ber is surrounded by an equivalent uid annulus in which pressures and oncotic forces must be determined. The theory is developed for a hexagonal ber array corresponding to the ber geometry shown in Figure 1b. A new closed-form expression for is derived, which turns out to be a function of two dimensionless parameters, a/R and b/R, and not just . Here, a and b are the solute and ber radii and R is the radius of the effective uid annulus. The results differ substantially from those associated with circular pores because of the large difference in the shape of the boundary along which the osmotic force is generated.

Mechanotransduction
A major recent development is the growing recognition that the EGL serves a critical role in the transmission of FSS to the actin cytoskeleton and in the initiation of intracellular signaling. Although numerous studies have demonstrated that FSS stimulates intracellular biomolecular responses and vascular regulation, it was widely assumed that this signaling was initiated either at the base of the cell via focal adhesion complexes or by the direct action of FSS on proteins in the apical membrane. The theoretical models in Section 4 have led to a fundamental paradox because all three models for the structural integrity of the EGL predict that the uid ow within the layer is negligible and, consequently, the FSS at the level of the membrane is vanishingly small. Furthermore, recent experiments in which the structural integrity of the EGL is compromised clearly demonstrate that cytoskeletal reorganization and biochemical responses can be nearly entirely abolished if the EGL is not intact. The studies described in this section convincingly show that the EGL plays a vital role in mechanotransduction. Cytoskeletal reorganization in response to FSS. The basic premise in the elastohydrodynamic model for the EGL proposed in Weinbaum et al. (71) is that the

www.annualreviews.org Endothelial Glycocalyx Layer

149

core proteins in the bush-like structures in Figure 2 have a exural rigidity, EI, that is sufcient to withstand signicant deformation owing to FSS in the physiological range. If this is the case, the hydrodynamic drag on the tips of the core proteins at the edge of the EGL will be transmitted as a bending moment to the ACC beneath the apical membrane of the EC. This exural rigidity allows the EGL to both serve as a molecular sieve for plasma proteins and as an exquisitely designed transducer of FSS. As noted earlier, Weinbaum et al. (71) and Han et al. (83) predict that EI falls in the range of 500700 pN nm2 . For the upper value of EI, one nds that the maximum displacement of the tips of the bers will typically be less than 10 nm for a FSS of 10 dyn/cm2 . In marked contrast, this value of EI is inadequate to prevent the buckling of the core proteins during the arrest of red cell motion or the passage of a WBC. Weinbaum et al. (71) predict that the compressive elastic resistance of the core proteins is at least a factor of 20 smaller than the hydrodynamic forces that are required to drain the uid from the EGL. Likewise, both the oncotic model of Secomb and coworkers (12, 13) and the mechano-electrochemical model of Damiano & Stace (87) can account for the connection between the extracellular GAGs in the lumen and the EC cytoskeleton. This would likely occur through connections between the GAG layer and an underlying macromolecular sublayer, where the sublayer would contain core proteins with transmembrane and cytoplasmic domains that anchor the entire ESL to the EC cytoskeleton. Tension in the thick GAG layer, induced by hydrodynamic drag arising from plasma ow through the layer, would provide an effective means of transmitting FSS through the thin sublayer containing core proteins and into the EC cytoskeleton. Thus, the primary distinction between this mechanism of FSS transmission and that of the elastohydrodynamic model for the EGL proposed by Weinbaum et al. (71) resides in the manner in which the EGL is thought to bear FSS. In contrast to the theory of exural rigidity of the core proteins proposed by Weinbaum et al. (71), the oncotic and mechano-electochemical models describe the EGL as being in a state of axial tension, and through connections to underlying core proteins, the hydrodynamic drag force on EGL GAGs pulls, to a greater or lesser extent, depending on the prevailing ow conditions in the vessel, on cytoskeletal elements. In either case, these two mechanisms both predict that uid drag throughout the EGL is converted into mechanical stress in core proteins and cytoskeletal elements. This is in stark contrast to the conventionally held view that cytoskeletal stress arises from a skin-friction mechanism that acts directly on an (EGL-free) EC lipid bilayer exposed directly to the ow. Such a view is not consistent with our new hydrodynamic understanding of the EGL described earlier because uid drag so severely attenuates ow throughout the EGL that FSS on the surface of the EC is effectively eliminated. To demonstrate the quantitative feasibility of the notion that the EGL might serve as a mechanotransducer of FSS, Thi et al. (75) have performed a series of experiments to explore the role of the EGL in the reorganization of the actin cytoskeleton and junction and focal-adhesion-associated proteins in response to FSS when the EGL is either intact or compromised (by enzymatic degradation or by the absence of plasma proteins) (19). Florian et al. (74) had previously shown that even partial removal of the most abundant proteoglycan, heparin sulfate, by heparinase III would entirely abolish

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

150

Weinbaum

Tarbell

Damiano

the increase in NO owing to FSS, but would have little effect on PGI2 , which is believed to be associated with basal adhesions. In particular, the redistribution of F-actin, vinculin, ZO-1, paxillin, and C 43 in conuent rat fat pad ECs in response to 5 h of FSS in four different bathing media (DMEM, DMEM + 10% FBS, DMEM + 1% BSA, and DMEM + 1%BSA + Heparinase III) were examined. DMEM alone and DMEM with heparinase III were observed to substantially reduce the cell surface expression of heparin sulfate compared to perfusion media with either BSA or FBS. The most striking observations in these experiments were that with an intact EGL there was a severe disruption of the dense peripheral actin band (DPAB), a formation of stress bers, and a migration of vinculin to cell borders after exposure to FSS. While this has been seen in numerous previous studies, this reorganization was completely abolished when the integrity of the EGL was compromised. Similarly, there were disruptions of the tight junctions (indicated by ZO-1) and gap junctions (indicated by C 43) for the intact EGL and no changes in the distribution of either of these junction-associated proteins when the EGL was compromised. In marked contrast, the distribution of paxillin, a marker for focal adhesions and FAK, was unaffected by the integrity of the EGL. These results, which were obtained for a steady FSS of 10 dyn/cm2 , were substantially reduced for a FSS of 5 dyn/cm2 applied for the same duration. This unequivocally demonstrated that the EGL was a transducer of FSS and that there was a threshold FSS for the cytoskeletal reorganization to occur. The above results are explained in Thi et al. (75) in a bumper-car conceptual model to describe the role of the EGL in activating cytoskeletal reorganization (see Figure 7). In Figure 7a, which describes controls without FSS, the DPABs act as rubber bumpers, which are held in lateral register (same elevation) by the weak attractive forces of the VE cadherins at the level of the adherens junctions. This prevents the stiff DPABs from denting the more vulnerable parts of the EC, much like bumpers on a car. When the cells are exposed to FSS for a substantial duration, the integrated bending moment owing to the drag on all the core proteins at the apical surface produces a torque on the cell (clockwise for the direction of FSS shown in panel b) that would act to lower the front (right side) and lower the rear (left side) of the DPAB. The same rotation is occurring on the cells in front of and behind the central cell in the sketch causing a disjoining force on the cadherin linkages between cells. When this disjoining force exceeds the strength of the cadherin complexes, which has been measured in Baumgartner et al. (111) as 70120 pN, an unzipping of the adherens junction occurs. The mathematical model (appendix C of Reference 75) predicts a disjoining force at the level of the DPAB of 70 pN for a FSS of 10 dyn/cm2 . Thus, it is striking that when the FSS was reduced to 5 dyn/cm2 , and the disjoining force reduced to 35 pN, there was little or no cytoskeletal reorganization with an intact EGL. This model also appears to explain why cells orient in the direction of ow; by elongating their planform they are able to lengthen the lever arm of the DPAB, decrease the disjoining force on the adherens junction, and produce greater stability in response to ow. When this occurs a new DPAB will eventually form after a sufciently long duration of change in average FSS. As shown in Figure 7b, the disruption of the DPAB leads to a redistribution of F-actin and the formation of stress bers on the basal aspects of the cell. In the

DPAB: dense peripheral actin band

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

151

Figure 7 Sketch of the conceptual bumper-car model for the structural organization of the EC in response to uid shear stress. In its conuent control state (a), ECs display an intact dense peripheral actin band (DPAB) that is localized to the adherens junction, where it serves as the base for the actin cortical web (ACW). (b) Cytoskeletal reorganization in response to uid shear stress. Adapted from gures 4A,B in Thi et al. (75).

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

transition to a new steady state, there is migration of vinculin to cell borders to establish temporary focal adhesions at the periphery of the cell so that a conuent monolayer is maintained. When the EGL is compromised, the FSS acts on the apical membrane directly and is transmitted by stress bers connecting the apical and basal surfaces to basal adhesion plaques. These stress bers bypass the underlying ACC. The latter can be thought of as a geodesic dome that is supported in its interior by a more exible actin network that is clearly seen in Satcher et al. (112). This dome-like structure rests at its edges on the much stiffer DPAB at its circumference, and for this reason, when the EGL is intact, the integrated torque serves to try to rotate the

152

Weinbaum

Tarbell

Damiano

DPAB. In sharp contrast, the total force acting on the basal integrin attachments is the same whether the FSS acts on the apical membrane or is transmitted across the apical membrane to the ACC. Thus, the distribution of paxillin, a marker for focal adhesions, is nearly the same whether the EGL is intact or degraded. Biomolecular response to FSS. The primary evidence that supports a major role for the EGL in mechanotransduction comes from experiments in which enzymes were used to selectively degrade specic components of the EGL, followed by a reassessment of function, or, as noted in the previous section, from using bathing solutions without plasma proteins where the EGL is compromised (19) and structural organization examined in response to FSS. Florian et al. (74) used the enzyme heparinase III to selectively degrade the HS component of bovine aortic endothelial cell GAGs in vitro and observed that the substantial production of NO induced over 3 h by steady or oscillatory FSS (20 or 10 15 dyn/cm2 ) could be completely inhibited by an enzyme dose that removed only 46% of the uorescence intensity associated with a HS antibody. The enzyme did not degrade CS and displayed negligible protease activity. It was also demonstrated that receptor-mediated NO induction by bradykinin and histamine were not affected by the enzymatic treatment, demonstrating that eNOS activity was not impaired directly by the enzyme. In an earlier study, the enzyme neuraminidase was used to remove SA residues from saline-perfused rabbit mesenteric arteries, and it was observed that ow-dependent vasodilation was abolished by a 30-min enzymatic pretreatment (113). Because owdependent vasodilation is mediated by NO release in many arteries, this study suggested that SA also contributes to FSS-induced production of NO. Similarly, Hecker et al. (114) showed that when intact segments of rabbit femoral arteries were pretreated with neuraminidase, FSS-induced NO production was inhibited. They also demonstrated that the same enzyme treatment had no effect on another hallmark response of ECs, the FSS-induced production of prostacyclin (PGI2 ) (115). This study illustrated the fact that there are multiple mechanisms of mechanotransduction and not a single mechanotransducer. In a more recent study, the enzyme hyaluronidase was used in isolated canine femoral arteries to degrade HA from the ESL layer, and a signicant inhibition of FSS-induced NO production was demonstrated (116). A new in vitro study using bovine aortic endothelial cells showed that the enzyme chondroitinase, employed to selectively degrade CS, did not inhibit the characteristic FSS-induced NO production, but treatment with either neuraminidase or hyaluronidase did (117). This study, and the earlier one by Florian et al. (74), illustrated the specicity of the ESL GAG components in mechanotransduction because removal of CS had no effect, whereas removal of HS and SA blocked the FSS-NO response. It was also shown that none of the three enzymes used had an inhibitory effect on FSS-induced PGI2 production, again suggesting multiple mechanisms of mechanotransduction. As described in the previous section, Thi et al. (75) exposed conuent monolayers of rat fat-pad ECs to 10 dyn/cm2 of steady FSS in a parallel plate ow chamber for ve hours and recorded the distribution of key structural proteins, such as F-actin, vinculin, paxillin, and ZO-1 using confocal microscopy. The experiments showed

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

153

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

a dramatic reorganization of the DPAB and the actin-associated linker molecule vinculin in response to FSS when the EGL was compromised by either removing plasma proteins or enzymatic partial removal of HS. These results imply that HS linked to transmembrane proteoglycans, such as the syndecans, mediates the FSSinduced reorganization of actin. One might infer from the results of Thi et al. (75) on actin reorganization and from Florian et al. (74) on NO production that experiments in protein-free media would not display the characteristic FSS-induced NO production associated with an intact EGL. This, however, is not the case, as several studies using cultured cells in protein-free media have shown the FSS-induced production of NO (118, 119). One way of interpreting these seemingly contradictory studies is to realize that when experiments are run in protein-free media, the hydraulic permeability of the EGL is increased as has been shown directly in transport experiments by Dull et al. (120) and Tarbell et al. (121) and the thickness of the ESL is decreased. This implies that a much greater fraction of the external FSS is transmitted to solid elements closer to the membrane, and possibly as FSS directly on the plasma membrane. One possible scenario that is consistent with this view is that, at least in part, FSS-induced NO production is mediated by eNOS in caveolae that may be stimulated through the solid matrix components of the EGL, such as the HS of glypicans near the membrane. While cytoskeletal reorganization appears to depend on transmembrane HS proteoglycans, the same distinction cannot be made for the NO signaling dependence on HS. Because NO transduction occurs regardless of where shear is felt, the caveolae-associated membrane-bound glypicans and the plasma membrane itself are likely sensors in protein-free media, whereas syndecans would appear to be more important in physiologic media-containing protein. The Distinction Between Mechanosensing and Mechanotransduction. Focusing on NO, we have considered events from the perspective of sensing force by the GAGs and transmission of force to the apical aspects of the cellthe plasma membrane or the cortical cytoskeleton. We turn now to mechanisms by which that force is transduced into a biomolecular response. HS proteoglycans can be linked to both the decentralized and centralized mechanisms of mechanotransduction put forth by Davies (122). In terms of the former, syndecans have an established association with the cytoskeleton (55), and through it can decentralize the signal, by spreading it to multiple sites within the cell (i.e., nucleus, organelles, focal adhesions, intercellular junctions). In terms of central transduction, both syndecans and glypicans interact with signaling proteins that can initiate cascades or directly inuence eNOS activation. For example, glypicans reside in caveolae along with a multitude of signaling molecules, including eNOS. Simply by location, glypicans can be involved in any one of a multitude of signal transduction pathways as elaborated in greater detail in Tarbell & Pahakis (69). HS may be involved in mechanotransduction through effects that are secondary to eNOS activation, but crucial to NO signaling. Nieuwdorp et al. (123) noted that lack of HS on the EC surface would result in loss of EC-SOD, an enzyme that catalyzes the reaction of superoxide to oxygen. This would result in a pro-oxidant state, damaging

154

Weinbaum

Tarbell

Damiano

both the EGL and eNOS, which would lead to decreased NO availability (124). Along the same lines is the hypothesis that HS plays a role in the availability of arginine close to the EC surface and its transporters. The supporting evidence for this comes from several different sources. First, HS displays high afnities for polycationic molecules, such as poly-L-arginines, and the binding sites of all their ligands involve arginine residues (41, 49, 63). Thus, HS proteoglycans may serve as a means of concentrating arginine close to the plasma membrane. In an older study, it was shown that HS proteoglycans adsorbed to a surface undergo a conformational change when exposed to ow, their core proteins unfolding from a random coil to an extended lament, and their HS chains elongating by 35% (125, 126). This was used to illustrate how Na+ ions bound to HS could be delivered by the stretched GAGs to their transporter channels; an analogous hypothesis can be made for the case of L-arginine. In sum, when the EGL is intact, FSS is transmitted to the cell through the core proteins of the EGL, and the specic connections of these proteins to the actin cytoskeleton (syndecans) and the plasma membrane (glypicans) mediate specic cell signaling (e.g., NO production, cytoskeletal reorganization). However, this stress is also distributed to other regions of the cell, most notably the intercellular junctions and the basal adhesion plaques, where transduction to intracellular biochemical signals also occurs (122, 127). When the EGL is degraded (e.g., by removal of protein) or actually removed (e.g., by the use of an enzyme such as heparanase), FSS is transmitted closer to the plasma membrane and apical signaling can proceed by mechanisms that differ from those associated with the transmembrane core proteins of the EGL. However, mechanical equilibrium principles illustrated in Figure 7b insure that the stresses delivered to the basal adhesion plaques would be the same with or without an intact EGL, as indicated by the response to paxillin discussed earlier, suggesting that signaling through these structures is independent of the integrity of the EGL. In other words, for a given FSS level, a basal adhesion plaque does not know if the cell has an intact EGL or not; it senses the same stress. In trying to identify the mechanisms by which the EGL mediates mechanotransduction, one confronts a multitude of possibilities: (a) cytoskeleton-associated decentralized signaling, (b) centralized direct association of the EGL with intracellular signaling molecules, and (c) regulation of local concentration gradients and transport of ions, amino acids, and growth factors. Taken together, these support the concept of the EGL as an orchestrator of mechanotransduction.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

Inammatory Response and IschemiaReperfusion Injury


The presence of a 500-nm-thick EGL in postcapillary venules (27, 28, 87) has important implications for leukocyte recruitment and the inammatory response cascade. In particular, we must reconsider conventionally held views about the mechanisms of leukocyte capture from the free stream in postcapillary venules that neglect the EGL. Rolling leukocytes attach to blood vessel walls through the selectin family of adhesion molecules, as noted in Section 5. These molecules typically extend 2050 nm above the plasma membrane (128). The distance between the ligands on the leukocyte microvilli and the adhesion receptor on the endothelial surface is very unlikely to exceed

www.annualreviews.org Endothelial Glycocalyx Layer

155

100 nm. Based on current estimates of the hydraulic resistivity and restoring forces of the layer (13, 84, 87, 95) and the predictions of the penetration depth of microvilli in Zhao et al. (98; see Section 5), even long leukocyte microvilli will be unable to penetrate the EGL, unless the WBC is already tethered. Thus, de novo attachment of leukocytes from the free stream (primary capture) is unlikely in vivo. Nevertheless, primary capture is the prevailing view of how leukocytes initially begin to roll in postcapillary venules. Indeed, experimental evidence shows that primary capture is exceedingly rare in inamed microvessels (129). When free-owing leukocytes attach to the walls of larger venules, which is also a rare event, this occurs mainly through secondary capture, i.e., leukocyte-leukocyte interaction (130). The gravitational amplication mechanism described in Zhao et al. (98; see Section 5) applies primarily to free-rolling leukocytes owing across EC monolayers in culture. In this mechanism there is a 20100-fold amplication of the very weak gravitational contact forces to typically 10 pN. As discussed earlier, forces of this magnitude are insufcient in vitro or in vivo to initiate selectin-mediated adhesion between the leukocyte and the endothelium except for very long microvilli at extremely low FSS where the contact time is greatly expanded. This is conrmed by the observation in vivo that tethered rolling does not preferentially begin on the bottom luminal surface of a postcapillary venule, but rather is uniformly distributed over the entire luminal surface (27, 93). The primary mechanism for tethered rolling predicted in Reference 98 is the large penetration force that is generated once an initial tethering event occurs. As noted in Section 5, the penetration force can increase to typically 100 pN for a FSS of as little as 1 dyn/cm2 a force that will easily allow deep EGL penetration. This implies that once tethered rolling is initiated it will readily continue. This same conclusion is reached in Reference 27. It has been observed that the vast majority of rolling leukocytes attach at the entrance of postcapillary venules in vivo (99, 131, 132), where deformed leukocytes exit from capillaries with a diameter smaller than the resting leukocyte diameter. The signicant mechanical deformation of the EGL that arises as leukocytes squeeze out of the conned lumen of capillaries and enter into postcapillary venules brings the leukocyte membrane in close apposition to the capillary wall so that initial interactions between selectin molecules and their counterligands can occur (27, 71). Previous work has shown that the EGL is highly compressed by passing leukocytes (23, 83), and in tightly tting capillaries, Han et al. (83) predict that this compression is close to 80%. There is some evidence that the EGL signicantly changes its properties under inammatory conditions (77, 133), which may facilitate sustained rolling and adhesion of leukocytes. In particular, cytokine-mediated activation of proteases either dwelling in the EGL or secreted by ECs and leukocytes may partially degrade the layer and thus provide a mechanism for localizing and chemically regulating leukocyte recruitment preferentially to regions of inamed vascularized tissue. In essence, then, a ubiquitous intact EGL might serve as a mechano-chemical barrier against leukocyte recruitment in uninamed tissue. If such localized degradation is a necessary precursor to promote tethered rolling of leukocytes as they emerge from capillaries (primary rolling), then this would represent a critical early step in the inammatory

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

156

Weinbaum

Tarbell

Damiano

response cascade. Therapeutic interventions targeted at inhibiting EGL shedding could then provide effective regulation of inammation. Thus, the mechanical compression of the EGL as leukocytes emerge from capillaries into postcapillary venules combined with cytokine-mediated localized enzymatic degradation of the layer provide a new paradigm by which leukocyte recruitment may occur in vivo. As stated earlier, the state of the EGL has not been reported in any of the studies of tethered rolling in parallel-plate ow chambers. Furthermore, many of these studies are not performed in plasma, where an intact EGL might be expected. Their objective has largely been to examine the kinetics of selectin attachment to elucidate the specicity of a variety of ligand interactions. Using the dye-exclusion method developed by Vink & Duling (23), Duling and coworkers investigated the mechanism by which ischemiareperfusion injury degrades the EGL in capillaries and postcapillary venules of the mouse cremaster muscle in vivo (134, 135). Platts et al. (134) showed that ischemiareperfusion injury rapidly modies the EGL and results in a 50% decrease in the size of the dye-exclusion zone in capillaries and postcapillary venules, whereas the red blood cell exclusion zone showed little change. Pretreatment with the A2A agonist ATL-146e caused a signicant attenuation in the effect of ischemiareperfusion injury on the EGL. Their results suggest that adenosine A2A receptor activation may play an important role in protecting the EGL from damage due to ischemiareperfusion injury and that the modication to the EGL that they observed in response to this form of injury may represent an important early step in the inammatory response to ischemiareperfusion. On the other hand, adenosine A3 receptor activation was found to cause a dose-dependent reduction in EGL thickness (136). Whereas at low concentrations, adenosine can inhibit EGL degradation and mast cell degranulation by activating the A2A receptor, at high concentrations, it can cause mast cell degranulation and lead to EGL degradation by activating the A3 receptor. In a subsequent study, Rubio-Gayosso et al. (135) show that EGL degradation resulting from ischemiareperfusion injury is mediated by reactive-oxygen species. They suggest that the reactive-oxygen species are generated by xanthine-oxidoreductase, an enzyme bound to the EGL, because inhibiting xanthine-oxidoreductase with allopurinol, or competing with it using heparin, essentially prevents the reduction of the dye-exclusion zone after ischemiareperfusion injury. Their results show that both heparin and hyaluronan have protective effects against ischemiareperfusion injury when their circulating concentration is increased before the injury. This supports their hypothesis that during ischemiareperfusion injury, xanthine-oxidoreductase might bind to the EGL through its heparin-binding domain. After reperfusion, EGL-bound xanthine-oxidoreductase would then localize the damaging effects of reactive-oxygen species production to the apical EGL. Finally, they show that this process may be dependent on the organization of hyaluronan in the EGL because intravenous administration of exogenous hyaluronan ameliorates the deleterious consequences of ischemiareperfusion injury. Thus, Rubio-Gayosso et al. hypothesize that reactive-oxygen species, produced as a consequence of ischemia reperfusion injury, are a signal and not the direct mediator of the decrease in the thickness of the dye-exclusion zone, as other treatments that produce reactive-oxygen species, such as UV light, also modied the red blood cell exclusion zone. Taken

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

157

together, the results of Duling and coworkers offer compelling evidence of a multitude of physiological and pathophysiological roles for the EGL during inammatory insults.

7. UNRESOLVED ISSUES AND FUTURE DIRECTIONS


We complete this review by highlighting some of the central unresolved issues and future research directions. A central challenge for future research is to understand the relationship between the biochemical composition of the EGL and its ultrastructure, currently revealed by EM with its attendant artifacts. One of the puzzling features is that the EGL is substantially thinner when observed in EM compared to functional in vivo measurements using intravital microscopy. This has suggested an inner and outer organization of key proteins with a cleavage plane 50100 nm from the apical membrane. The biochemical composition of the EGL is essential to understanding its enzymatic degradation. For example, why does partial removal of select components of the EGL (HS, SA, HA) completely block FSS-induced NO production? Is FSS-induced NO production mediated by force transmission through syndecans or glypicans, or alternatively by changes in the binding of key substrates (e.g., L-arginine) close to the plasma membrane? Why does the integrity of the EGL play a different role in PGI2 and NO production in response to FSS? What are the mechanisms via which reactive-oxygen species mediate damage to the EGL? From a biophysical standpoint the central unanswered questions relate to the mechanisms that provide for the structural integrity of the EGL in response to FSS and its restoration after cellular compression. Is this due to oncotic forces, inherent exural rigidity of its bers, electrochemical potential gradients, or some combination of these and possibly other mechanisms? These forces also determine the EGLs properties as a molecular sieve for plasma proteins. As seen in Zhang et al. (72), the oncotic forces generated in a ber matrix layer should differ greatly from those in a membrane with cylindrical pores. This still needs to be demonstrated experimentally. Although there is now strong evidence in support of the revised Starling principle for water exchange, the role of the EGL in transmitting FSS to the tight junction strands and in turn modulating small-solute permeability has not been explored. It is striking that literally hundreds of papers have been written on tethered rolling on cultured ECs and on substrates with a host of different ligands and receptors, but no study has attempted to determine the importance of an intact EGL in either the initiation or maintenance of tethered rolling. It is not known whether the EGL is ubiquitous throughout the microvasculature or if its distribution and relative uniformity are preferential to capillaries and postcapillary venules. This has important consequences for inammation, mechanotransduction, vascular resistance, atherosclerosis, and other vascular diseases. It is safe to say the status of the EGL is still in its infancy and we have only scratched the surface in determining its structure and function. However, the eld is beginning to draw considerable attention as the scientic advances and volume of literature written on this subject since 2000 far surpass cumulative contributions of the previous 35 years since Lufts clear demonstration of its existence.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

158

Weinbaum

Tarbell

Damiano

SUMMARY POINTS 1. The EGL is a complex, multicomponent, biochemical structure that functions exquisitely as a molecular sieve, a lubrication layer for red blood cell motion, and a sensor of ow-induced FSS. 2. When observed in EM, the EGL is <100 nm thick and has a quasi-periodic structure with focal scattering centers of 1012 nm in diameter and 20 nm spacing. This extracellular structure appears to be linked to an underlying actin cortical network that forms a hexagonal lattice with 100 nm spacing. 3. Two methods have been developed for interrogating the EGL using intravital microscopy: a uorescent dye exclusion technique and -PIV. These approaches have revealed that the EGL thickness in vivo in capillaries and venules (550 m in diameter) is 0.5 m thick, which is at least vefold thicker than most EM studies indicate. 4. Theoretical models predict that the uid velocity is greatly attenuated within the EGL and hence a negligible FSS acts on the EC membrane. Thus, all of the ow-induced FSS is transferred as a drag to matrix bers. 5. When the EGL is crushed by the passage of a WBC or during the start-up motion of a red cell in a tightly tting capillary, it is restored to full thickness in <1 s. Three models have been advanced for its restorative mechanism under these conditions: oncotic forces owing to adsorbed plasma proteins, exural rigidity of its core proteins, and electrochemical properties owing to its xed negative charges. 6. The EGL is now believed to be the primary molecular sieve for plasma proteins and, therefore, the origin of the oncotic forces that control transcapillary uid exchange. This has led to a revised Starling principle in which the Starling forces are applied locally across just the EGL, rather than globally across the entire endothelial layer as had been widely assumed. 7. When the EGL is intact, it serves as the primary sensor of FSS on ECs. The EGL transmits this stress to diverse sites within the cell (e.g., apical plasma membrane, actin cortical web, basal adhesion plaques, intercellular junctions) where mechanotransduction occurs and intracellular signaling is initiated. A bumper car model has been proposed for cytoskeletal reorganization. 8. The EGL is likely to play a critical role in the inammatory response cascade, as it can inhibit primary leukocyte capture and systemic leukocyte extravasation and can provide a physical barrier that can regulate and localize leukocyte recruitment to inamed vascularized tissue.

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

159

ACKNOWLEDGMENTS
This work was supported by NIH grants R01-HL44485 (to S.W.), R01-HL57093 (to J.M.T.), and R01-HL076499 (to E.R.D.). We thank Danielle Wu for help with citations.

LITERATURE CITED
1. Pries AR, Secomb TW, Gaehtgens P. 2000. The endothelial surface layer. Pugers Arch. 440:65366 2. Danielli JF. 1940. Capillary permeability and edema in the perfused frog. J. Cell Physiol. 98:10929 3. Chambers R, Zweifach BW. 1947. Intercellular cement of capillary permeability. Physiol. Rev. 27:43663 4. Copley AL. 1974. Hemorheological aspects of the endothelium-plasma interface. Microvasc. Res. 8:192212 5. Luft JH. 1966. Fine structures of capillary and endocapillary layer as revealed by ruthenium red. Fed. Proc. 25:177383 6. Klitzman B, Duling BR. 1979. Microvascular hematocrit and red cell ow in resting and contracting striated muscle. Am. J. Physiol. Heart Circ. Physiol. 237:H48190 7. Desjardins C, Duling BR. 1990. Heparinase treatment suggests a role for the endothelial cell glycocalyx in regulation of capillary hematocrit. Am. J. Physiol. Heart Circ. Physiol. 258:H64754 8. Pries AR, Secomb TW, Gessner T, Sperandio MB, Gross JF, Gaehtgens P. 1994. Resistance to blood ow in microvessels in vivo. Circ. Res. 75:90415 9. Pries AR, Secomb TW, Jacobs H, Sperandio M, Osterloh K, Gaehtgens P. 1997. Microvascular blood ow resistance: role of endothelial surface layer. Am. J. Physiol. Heart Circ. Physiol. 273:H227279 10. Damiano ER, Duling BR, Ley K, Skalak TC. 1996. Axisymmetric pressuredriven ow of rigid pellets through a cylindrical tube lined with a deformable porous wall layer. J. Fluid Mech. 314:16389 11. Damiano ER. 1998. The effect of the endothelial-cell glycocalyx on the motion of red blood cells through capillaries. Microvasc. Res. 55:7791 12. Secomb TW, Hsu R, Pries AR. 1998. A model for red blood cell motion in glycocalyx-lined capillaries. Am. J. Physiol. Heart Circ. Physiol. 274:H101622 13. Secomb TW, Hsu R, Pries AR. 2001. Motion of red blood cells in a capillary with an endothelial surface layer: effect of ow velocity. Am. J. Physiol. Heart Circ. Physiol. 281:H62936 14. Lewis JC, Taylor RG, Jones ND, St Clair RW, Cornhill JF. 1982. Endothelial surface characteristics in pigeon coronary artery atherosclerosis. I. Cellular alterations during the initial stages of dietary cholesterol challenge. Lab. Invest. 46:12338 15. van den Berg BM, Spaan JAE, Rolf TM, Vink H. 2006. Atherogenic region and diet diminish glycocalyx dimension and increase intima-to-media ratios at

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

13. First realistic attempt to model axisymmetric ow-dependent red blood cell shapes in an EGL-lined capillary.

160

Weinbaum

Tarbell

Damiano

16.

17.

18.
Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

19. 20. 21. 22.

23.

24.

25.

26. 27.

28.

29.

30. 31.

murine carotid artery bifurcation. Am. J. Physiol. Heart Circ. Physiol. 290:H915 20 Gouverneur M, Berg B, Nieuwdorp M, Stroes E, Vink H. 2006. Vasculoprotective properties of the endothelial glycocalyx: effects of uid shear stress. J. Intern. Med. 259:393400 Baldwin AL, Winlove CP. 1984. Effects of perfusate composition on binding of ruthenium red and gold colloid to glycocalyx of rabbit aortic endothelium. J. Histochem. Cytochem. 32:25966 Clough G, Moftt H. 1992. Immunoperoxidase labeling of albumin at the endothelial cell surface of frog mesenteric microvessels. Int. J. Microcirc. Clin. Exp. 11:34558 Adamson RH, Clough G. 1992. Plasma proteins modify the endothelial cell glycocalyx of frog mesenteric microvessels. J. Physiol. 445:47386 Sims DE, Horne MM. 1993. Non-aqueous xative preserves macromolecules on the endothelial cell surface: an in situ study. Eur. J. Morphol. 31:25155 Rostgaard J, Qvortrup K. 1997. Electron microscopic demonstrations of lamentous molecular sieve plugs in capillary fenestrae. Microvasc. Res. 53:113 Squire JM, Chew M, Nneji G, Neal C, Barry J, Michel C. 2001. Quasi-periodic substructure in the microvessel endothelial glycocalyx: a possible explanation for molecular ltering? J. Struct. Biol. 136:23955 Vink H, Duling BR. 1996. Identication of distinct luminal domains for macromolecules, erythrocytes, and leukocytes within mammalian capillaries. Circ. Res. 79:58189 Vink H, Constantinescu AA, Spaan JAE. 2000. Oxidized lipoproteins degrade the endothelial surface layer: implications for platelet-endothelial cell adhesion. Circulation 101:15002 Abrahamsson T, Brandt U, Marklund SL, Sjoqvist PO. 1992. Vascular bound recombinant extracellular superoxide dismutase type C protects against the detrimental effects of superoxide radicals on endothelium-dependent arterial relaxation. Circ. Res. 70:26471 Henry CB, Duling BR. 1999. Permeation of the luminal capillary glycocalyx is determined by hyaluronan. Am. J. Physiol. Heart Circ. Physiol. 277:H50814 Smith ML, Long DS, Damiano ER, Ley K. 2003. Near-wall micro-PIV reveals a hydrodynamically relevant endothelial surface layer in venules in vivo. Biophys. J. 85:63745 Long DS, Smith ML, Pries AR, Ley K, Damiano ER. 2004. Microviscometry reveals reduced blood viscosity and altered shear rate and shear stress proles in microvessels after hemodilution. Proc. Natl. Acad. Sci. USA 101:1006065 Damiano ER, Long DS, Smith ML. 2004. Estimation of viscosity proles using velocimetry data from parallel ows of linearly viscous uids. J. Fluid Mech. 512:119 Michel CC. 1997. Starling: the formulation of his hypothesis of microvascular uid exchange and its signicance after 100 years. Exp. Physiol. 82:130 Levick JR. 1991. Capillary ltration-absorption balance reconsidered in light of dynamic extravascular factors. Exp. Physiol. 76:82557

27. In vivo study shows direct effect of EGL drag on blood plasma and demonstrates an EGL thickness of 400 nm in venules.

www.annualreviews.org Endothelial Glycocalyx Layer

161

37. Presents key experiments on mammalian capillaries demonstrating validity of revised Starling principle and 3-D theoretical model to predict ow induced oncotic pressure difference across the EGL.

32. Michel CC, Phillips ME. 1987. Steady-state uid ltration at different capillary pressures in perfused frog mesenteric capillaries. J. Physiol. 388:42135 33. Weinbaum S. 1998. 1997 Whitaker Distinguished Lecture: models to solve mysteries in biomechanics at the cellular level; a new view of ber matrix layers. Ann. Biomed. Eng. 26:62743 34. Levick JR. 2004. Revision of the Starling principle: new views of tissue uid balance. J. Physiol. 557:704 35. Hu X, Weinbaum S. 1999. A new view of Starlings hypothesis at the microstructural level. Microvasc. Res. 58:281304 36. Hu X, Adamson RH, Liu B, Curry FE, Weinbaum S. 2000. Starling forces that oppose ltration after tissue oncotic pressure is increased. Am. J. Physiol. Heart Circ. Physiol. 279:H172436 37. Adamson RH, Lenz JF, Zhang X, Adamson GN, Weinbaum S, Curry FE. 2004. Oncotic pressures opposing ltration across nonfenestrated rat microvessels. J. Physiol. 557:889907 38. Pang Z, Tarbell JM. 2003. In vitro study of Starlings hypothesis in a cultured monolayer of bovine aortic endothelial cells. J. Vasc. Res. 40:35158 39. Berneld M, Gotte M, Park PW, Reizes O, Fitzgerald ML, et al. 1999. Functions of cell surface heparan sulfate proteoglycans. Annu. Rev. Biochem. 68:729 77 40. Osterloh K, Ewert U, Pries AR. 2002. Interaction of albumin with the endothelial cell surface. Am. J. Physiol. Heart Circ. Physiol. 283:H398405 41. Hileman RE, Fromm JR, Weiler JM, Linhardt RJ. 1998. Glycosaminoglycanprotein interactions: denition of consensus sites in glycosaminoglycan binding proteins. Bioessays 20:15667 42. Coombe DR, Kett WC. 2005. Heparan sulfate-protein interactions: therapeutic potential through structure-function insights. Cell Mol. Life Sci. 62:41024 43. McGee MP, Liang J. 2001. Regulation of glycosaminoglycan function by osmotic potentials. Measurement of water transfer during antithrombin activation by heparin. J. Biol. Chem. 276:4927582 44. Kijewska I, Hawlicka E. 2005. A new radiochemical method to investigate ion binding with polyelectrolytes. Carbohydr. Res. 340:118591 45. Kan M, Wang F, To B, Gabriel JL, McKeehan WL. 1996. Divalent cations and heparin/heparan sulfate cooperate to control assembly and activity of the broblast growth factor receptor complex. J. Biol. Chem. 271:2614348 46. Arisaka T, Mitsumata M, Kawasumi M, Tohjima T, Hirose S, Yoshida Y. 1995. Effects of shear stress on glycosaminoglycan synthesis in vascular endothelial cells. Ann. N.Y. Acad. Sci. 748:54354 47. Paka L, Kako Y, Obunike JC, Pillarisetti S. 1999. Apolipoprotein E containing high density lipoprotein stimulates endothelial production of heparan sulfate rich in biologically active heparin-like domains. A potential mechanism for the antiatherogenic actions of vascular apolipoprotein E. J. Biol. Chem. 274(8):481623 48. Vijayagopal P, Srinivasan SR, Dalferes ER Jr, Radhakrishnamurthy B, Berenson GS. 1988. Effect of low-density lipoproteins on the synthesis and secretion of proteoglycans by human endothelial cells in culture. Biochem. J. 255:63946

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

162

Weinbaum

Tarbell

Damiano

49. Jackson RL, Busch SJ, Cardin AD. 1991. Glycosaminoglycans: molecular properties, protein interactions, and role in physiological processes. Physiol. Rev. 71:481539 50. Oohira A, Wight TN, Bornstein P. 1983. Sulfated proteoglycans synthesized by vascular endothelial cells in culture. J. Biol. Chem. 258:201421 51. Camejo G, Hurt-Camejo E, Wiklund O, Bondjers G. 1998. Association of apo B lipoproteins with arterial proteoglycans: pathological signicance and molecular basis. Atherosclerosis 139:20522 52. Almond A, Sheehan JK. 2000. Glycosaminoglycan conformation: do aqueous molecular dynamics simulations agree with X-ray ber diffraction? Glycobiology 10:32938 53. Seog J, Dean D, Rolauffs B, Wu T, Genzer J, et al. 2005. Nanomechanics of opposing glycosaminoglycan macromolecules. J. Biomech. 38:178997 54. van Haaren PM, VanBavel E, Vink H, Spaan JAE. 2005. Charge modication of the endothelial surface layer modulates the permeability barrier of isolated rat mesenteric small arteries. Am. J. Physiol. Heart Circ. Physiol. 289:H25037 55. Tkachenko E, Rhodes JM, Simons M. 2005. Syndecans: new kids on the signaling block. Circ. Res. 96:488500 56. Zimmermann P, David G. 1999. The syndecans, tuners of transmembrane signaling. FASEB J. 13(Suppl.):S91100 57. Rosenberg RD, Shworak NW, Liu J, Schwartz JJ, Zhang L. 1997. Heparan sulfate proteoglycans of the cardiovascular system. Specic structures emerge but how is synthesis regulated? J. Clin. Invest. 99:206270 58. Halden Y, Rek A, Atzenhofer W, Szilak L, Wabnig A, Kungl AJ. 2004. Interleukin-8 binds to syndecan-2 on human endothelial cells. Biochem. J. 377:53338 59. Kokenyesi R, Berneld M. 1994. Core protein structure and sequence determine the site and presence of heparan sulfate and chondroitin sulfate on syndecan-1. J. Biol. Chem. 269:123049 60. Simons M, Horowitz A. 2001. Syndecan-4-mediated signaling. Cell Signal. 13:85562 61. Yoneda A, Couchman JR. 2003. Regulation of cytoskeletal organization by syndecan transmembrane proteoglycans. Matrix Biol. 22:2533 62. Fransson LA, Belting M, Cheng F, Jonsson M, Mani K, Sandgren S. 2004. Novel aspects of glypican glycobiology. Cell Mol. Life Sci. 61:101624 63. Belting M. 2003. Heparan sulfate proteoglycan as a plasma membrane carrier. Trends Biochem. Sci. 28:14551 64. Cheng F, Mani K, van den BJ, Ding K, Belting M, Fransson LA. 2002. Nitric oxide-dependent processing of heparan sulfate in recycling S-nitrosylated glypican-1 takes place in caveolin-1-containing endosomes. J. Biol. Chem. 277:4443139 65. van Deurs B, Roepstorff K, Hommelgaard AM, Sandvig K. 2003. Caveolae: anchored, multifunctional platforms in the lipid ocean. Trends Cell Biol. 13:92 100 66. Laurent TC, Fraser JR. 1992. Hyaluronan. FASEB J. 6:2397404

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

163

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

69. Recent comprehensive review of the role of the EGL in mechanotransduction by endothelial cells including an extensive description of the biochemical structure of the EGL. 71. Describes the role of EGL in mechanotransduction of FSS to intracellular cytoskeleton and rst predictions of elastic properties of EGL core proteins.

74. First study to show the role of heparan sulfate proteoglycans as mechanosensors for FSS-induced nitric oxide production in endothelial cells in vitro. 75. Bumper-car model for cytoskeletal reorganization of F-actin and junctional proteins in response to FSS demonstrated that an intact EGL is essential for this reorganization.

83. Non-linear large deformation theory for the elasto-hydrodynamic restoration of the EGL after passage of a WBC with key experiments by Vink and coworkers.

67. Singleton PA, Bourguignon LY. 2004. CD44 interaction with ankyrin and IP3 receptor in lipid rafts promotes hyaluronan-mediated Ca2+ signaling leading to nitric oxide production and endothelial cell adhesion and proliferation. Exp. Cell Res. 295:10218 68. Varki A. 1997. Sialic acids as ligands in recognition phenomena. FASEB J. 11:24855 69. Tarbell JM, Pahakis MY. 2006. Mechanotransduction and the glycocalyx. J. Intern. Med. 259:33950 70. Clough G, Michel CC. 1988. Quantitative comparisons of hydraulic permeability and endothelial intercellular cleft dimensions in single frog capillaries. J. Physiol. 405:56376 71. Weinbaum S, Zhang X, Han Y, Vink H, Cowin SC. 2003. Mechanotransduction and ow across the endothelial glycocalyx. Proc. Natl. Acad. Sci. USA 100:798895 72. Zhang X, Curry FR, Weinbaum S. 2006. Mechanism of osmotic ow in a periodic ber array. Am. J. Physiol. Heart Circ. Physiol. 290:H84452 73. van den Berg BM, Vink H, Spaan JAE. 2003. The endothelial glycocalyx protects against myocardial edema. Circ. Res. 92:59294 74. Florian JA, Kosky JR, Ainslie K, Pang Z, Dull RO, Tarbell JM. 2003. Heparan sulfate proteoglycan is a mechanosensor on endothelial cells. Circ. Res. 93:e13642 75. Thi MM, Tarbell JM, Weinbaum S, Spray DC. 2004. The role of the glycocalyx in reorganization of the actin cytoskeleton under uid shear stress: a bumper-car model. Proc. Natl. Acad. Sci. USA 101:1648388 76. Vink H, Duling BR. 2000. Capillary endothelial surface layer selectively reduces plasma solute distribution volume. Am. J. Physiol. Heart Circ. Physiol. 278:H285 89 77. Henry CB, Duling BR. 2000. TNF-alpha increases entry of macromolecules into luminal endothelial cell glycocalyx. Am. J. Physiol. Heart Circ. Physiol. 279:H281523 78. Damiano ER, Long DS, El-Khatib FH, Stace TM. 2004. On the motion of a sphere in a Stokes ow parallel to a Brinkman half space. J. Fluid Mech. 500:75 101 79. Kishino A, Yanagida T. 1988. Force measurements by micromanipulation of a single actin lament by glass needles. Nature 334:7476 80. Dupuis DE, Guilford WH, Wu J, Warshaw DM. 1997. Actin lament mechanics in the laser trap. J. Muscle Res. Cell Motil. 18:1730 81. Gittes F, Mickey B, Nettleton J, Howard J. 1993. Flexural rigidity of microtubules and actin laments measured from thermal uctuations in shape. J. Cell Biol. 120:92334 82. Vink A, Warnier G, Brombacher F, Renauld JC. 1999. Interleukin 9-induced in vivo expansion of the B-1 lymphocyte population. J. Exp. Med. 189:141323 83. Han Y, Weinbaum S, Spaan JAE, Vink H. 2006. Large-deformation analysis of the elastic recoil of ber layers in a Brinkman medium with application to the endothelial glycocalyx. J. Fluid Mech. 554:21735

164

Weinbaum

Tarbell

Damiano

84. Stace TM, Damiano ER. 2001. An electrochemical model of the transport of charged molecules through the capillary glycocalyx. Biophys. J. 80:167090 85. Lai WM, Hou JS, Mow VC. 1991. A triphasic theory for the swelling and deformation behaviors of articular cartilage. J. Biomech. Eng. 113:24558 86. Vink H, Stace TM, Damiano ER. 2003. High resolution 3D intravital uorescence microscopy reveals partial exclusion of anionic tracers within a 1 micron thick capillary endothelial cell glycocalyx. FASEB J. 17:A70 87. Damiano ER, Stace TM. 2002. A mechano-electrochemical model of radial deformation of the capillary glycocalyx. Biophys. J. 82:115375 88. Fu BM, Chen B, Chen W. 2003. An electrodiffusion model for effects of surface glycocalyx layer on microvessel permeability. Am. J. Physiol. Heart Circ. Physiol. 284:H124050 89. Adamson RH, Huxley VH, Curry FE. 1988. Single capillary permeability to proteins having similar size but different charge. Am. J. Physiol. Heart Circ. Physiol. 254:H30412 90. Barry SI, Parker KH, Aldis GK. 1991. Fluid ow over a thin deformable porous layer. J. Appl. Math. Phys. 42:63348 91. Wang W, Parker KH. 1995. The effect of deformable porous surface layers on the motion of a sphere in a narrow cylindrical tube. J. Fluid Mech. 283:287305 92. Levick JR. 1987. Flow through interstitium and other brous matrices. Q. J. Exp. Physiol. 72:40937 93. Damiano ER, Stace TM. 2005. Flow and deformation of the capillary glycocalyx in the wake of a leukocyte. Phys. Fluids 17:03150917 94. Secomb TW, Hsu R, Pries AR. 2001. Effect of endothelial surface layer on transmission of uid shear stress to endothelial cells. Biorheology 38:14350 95. Feng J, Weinbaum S. 2000. Lubrication theory in highly compressible porous media: the mechanics of skiing, from red cells to humans. J. Fluid Mech. 422:281 317 96. Secomb TW, Skalak R, Ozkaya N, Gross JF. 1986. Flow of axisymmetric red blood cells in narrow capillaries. J. Fluid Mech. 163:40523 97. Lawrence MB, Springer TA. 1991. Leukocytes roll on a selectin at physiologic ow rates: distinction from and prerequisite for adhesion through integrins. Cell 65:85973 98. Zhao Y, Chien S, Weinbaum S. 2001. Dynamic contact forces on leukocyte microvilli and their penetration of the endothelial glycocalyx. Biophys. J. 80:112440 99. Ley K. 1996. Molecular mechanisms of leukocyte recruitment in the inammatory process. Cardiovasc. Res. 32:73342 100. Bruehl RE, Springer TA, Bainton DF. 1996. Quantitation of L-selectin distribution on human leukocyte microvilli by immunogold labeling and electron microscopy. J. Histochem. Cytochem. 44:83544 101. Tissot O, Pierres A, Foa C, Delaage M, Bongrand P. 1992. Motion of cells sedimenting on a solid surface in a laminar shear ow. Biophys. J. 61:20415 102. Tempelman LA, Hammer DA. 1994. Receptor-mediated binding of IgEsensitized rat basophilic leukemia cells to antigen-coated substrates under hydrodynamic ow. Biophys. J. 66:123143

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

87. Theoretical model to predict electrochemical potential gradient across EGL and mechanics of restoring-force mechanism during compression and recovery of the EGL.

98. Theoretical model to predict penetration of the EGL by leukocyte microvilli and the biophysical role of the EGL in the inammatory response.

www.annualreviews.org Endothelial Glycocalyx Layer

165

103. Shao JY, Ting-Beall HP, Hochmuth RM. 1998. Static and dynamic lengths of neutrophil microvilli. Proc. Natl. Acad. Sci. USA 95:6797802 104. Hammer DA, Apte SM. 1992. Simulation of cell rolling and adhesion on surfaces in shear ow: general results and analysis of selectin-mediated neutrophil adhesion. Biophys. J. 63:3557 105. Tozeren A, Ley K. 1992. How do selectins mediate leukocyte rolling in venules? Biophys. J. 63:7009 106. Feng J, Ganatos P, Weinbaum S. 1998. Motion of a sphere near planar conning boundaries in a Brinkman medium. J. Fluid Mech. 375:26596 107. Lee GM, Zhang F, Ishihara A, McNeil CL, Jacobson KA. 1993. Unconned lateral diffusion and an estimate of pericellular matrix viscosity revealed by measuring the mobility of gold-tagged lipids. J. Cell Biol. 120:2535 108. Verrico CD. 2004. Highlights from the literature. Adamson et al. (April 8, 2004) J. Physiol. Nominated by S. Sage, Chair, Editorial Board, J. Physiol., communicated by G. Clough. Physiology 19:16162 109. Anderson JL, Malone DM. 1974. Mechanism of osmotic ow in porous membranes. Biophys. J. 14:95782 110. Curry FE, Michel CC. 1980. A ber matrix model of capillary permeability. Microvasc. Res. 20:9699 111. Baumgartner W, Hinterdorfer P, Ness W, Raab A, Vestweber D, et al. 2000. Cadherin interaction probed by atomic force microscopy. Proc. Natl. Acad. Sci. USA 97:400510 112. Satcher R, Dewey CF Jr, Hartwig JH. 1997. Mechanical remodeling of the endothelial surface and actin cytoskeleton induced by uid ow. Microcirculation 4:43953 113. Pohl U, Herlan K, Huang A, Bassenge E. 1991. EDRF-mediated shear-induced dilation opposes myogenic vasoconstriction in small rabbit arteries. Am. J. Physiol. Heart Circ. Physiol. 261:H201623 114. Hecker M, Mulsch A, Bassenge E, Busse R. 1993. Vasoconstriction and increased ow: two principal mechanisms of shear stress-dependent endothelial autacoid release. Am. J. Physiol. Heart Circ. Physiol. 265:H82833 115. Frangos JA, Eskin SG, McIntire LV, Ives CL. 1985. Flow effects on prostacyclin production by cultured human endothelial cells. Science 227:147779 116. Mochizuki S, Vink H, Hiramatsu O, Kajita T, Shigeto F, et al. 2003. Role of hyaluronic acid glycosaminoglycans in shear-induced endothelium-derived nitric oxide release. Am. J. Physiol. Heart Circ. Physiol. 285:H72226 117. Pahakis MY, Kosky JR, Dull RO, Tarbell JM. 2007. The role of endothelial glycocalyx components in mechanotransduction of uid sheer stress. Biochem. Biophys. Res. Commun. 355:22833 118. Ye JF, Zheng XX, Xu LX. 2003. Real-time detection of nitric oxide in cultured rat aorta endothelial cells induced by shear stress. Sheng Wu Hua Xue Yu Sheng Wu Wu Li Xue Bao (Shanghai) 35:296300 119. Qiu W, Kass DA, Hu Q, Ziegelstein RC. 2001. Determinants of shear stressstimulated endothelial nitric oxide production assessed in real-time by 4,5diaminouorescein uorescence. Biochem. Biophys. Res. Commun. 286:32835

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

166

Weinbaum

Tarbell

Damiano

120. Dull RO, Jo H, Sill H, Hollis TM, Tarbell JM. 1991. The effect of varying albumin concentration and hydrostatic pressure on hydraulic conductivity and albumin permeability of cultured endothelial monolayers. Microvasc. Res. 41:390407 121. Tarbell JM, Lever MJ, Caro CG. 1988. The effect of varying albumin concentration of the hydraulic conductivity of the rabbit common carotid artery. Microvasc. Res. 35:20420 122. Davies PF. 1995. Flow-mediated endothelial mechanotransduction. Physiol. Rev. 75:51960 123. Nieuwdorp M, Meuwese MC, Vink H, Hoekstra JB, Kastelein JJ, Stroes ES. 2005. The endothelial glycocalyx: a potential barrier between health and vascular disease. Curr. Opin. Lipidol. 16:50711 124. Faraci FM, Didion SP. 2004. Vascular protection: superoxide dismutase isoforms in the vessel wall. Arterioscler. Thromb. Vasc. Biol. 24:136773 125. Siegel G, Malmsten M, Klussendorf D, Walter A, Schnalke F, Kauschmann A. 1996. Blood-ow sensing by anionic biopolymers. J. Auton. Nerv. Syst. 57:207 13 126. Siegel G, Walter A, Kauschmann A, Malmsten M, Buddecke E. 1996b. Anionic biopolymers as blood ow sensors. Biosens. Bioelectron. 11:28194 127. Resnick N, Yahav H, Shay-Salit A, Shushy M, Schubert S, et al. 2003. Fluid shear stress and the vascular endothelium: for better and for worse. Prog. Biophys. Mol. Biol. 81:17799 128. Springer TA. 1995. Trafc signals on endothelium for lymphocyte recirculation and leukocyte emigration. Annu. Rev. Physiol. 57:82772 129. Kunkel EJ, Chomas JE, Ley K. 1998. Role of primary and secondary capture for leukocyte accumulation in vivo. Circ. Res. 82:3038 130. Eriksson EE, Xie X, Werr J, Thoren P, Lindbom L. 2001. Importance of primary capture and L-selectin-dependent secondary capture in leukocyte accumulation in inammation and atherosclerosis in vivo. J. Exp. Med. 194:20518 131. Dunne JL, Ballantyne CM, Beaudet AL, Ley K. 2002. Control of leukocyte rolling velocity in TNF-alpha-induced inammation by LFA-1 and Mac-1. Blood 99:33641 132. Schmid-Schonbein GW, Usami S, Skalak R, Chien S. 1980. The interaction of leukocytes and erythrocytes in capillary and postcapillary vessels. Microvasc. Res. 19:4570 133. Mulivor AW, Lipowsky HH. 2004. Inammation- and ischemia-induced shedding of venular glycocalyx. Am. J. Physiol. Heart Circ. Physiol. 286:H167280 134. Platts SH, Linden J, Duling BR. 2003. Rapid modication of the glycocalyx caused by ischemiareperfusion is inhibited by adenosine A2A receptor activation. Am. J. Physiol. Heart Circ. Physiol. 290:H236067 135. Rubio-Gayosso I, Platts SH, Duling BR. 2006. Reactive oxygen species mediate modication of glycocalyx during ischemiareperfusion injury. Am. J. Physiol. Heart Circ. Physiol. 290:H224756 136. Platts SH, Duling BR. 2004. Adenosine A3 receptor activation modulates the capillary endothelial glycocalyx. Circ. Res. 94:7782

Annu. Rev. Biomed. Eng. 2007.9:121-167. Downloaded from www.annualreviews.org by Annual Reviews on 03/31/11. For personal use only.

www.annualreviews.org Endothelial Glycocalyx Layer

167

Anda mungkin juga menyukai