The Transport of Albumin: A Critique of the Vesicular System in Transendothelial Transport 1 . 2 C. CHARLES MICHEL
Introduction Serum albumin plays a central role in microvascular exchange. It binds both specifically and nonspecifically to the surfaces of endothelial cells (1-6), reducing their permeability to fluid and macromolecules (7-10). As a carrier of smaller molecules, it determines the availability of fatty acids and certain amino acids, steroids, drugs, etc. for microvascular exchange (11). Albumin is also responsible for a large fraction of the plasma oncotic pressure so that its permeation through microvascular walls and accumulation in the interstitium is a major determinant of the distribution of fluid between the blood and the tissues (12). Despite its importance, the mechanism of albumin transport through microvascular walls remains controversiaL Although some suggest that serum albumin crosses through the extensivevesicular system of endothelium by transcytosis (13), others believe that the arguments against vesicular transport are compelling, and they assert that albumin (together with other macromolecules) is carried by convection through a small number of relativelylarge pores, which probably lie between the endothelial cells (14). These conflicting points of vieware based on very different lines of evidence. In assessing this evidence, I shall consider two sequential questions. The first is whether there is a balance of evidence for any transport of macromolecules through microvascular walls via the endothelial vesicular system.Assuming there is such a balance, the second question is whether transport via the vesicular system makes a significant contribution to total exchangeof macromolecules between the blood and the tissues. Before considering these general questions, however, some studies that implicate the vesicular system in the transendothelial transport of serum albumin will be briefly reviewed. Ultrastructural Evidence for Transendothelial Transport of Serum Albumin Even though many studies over the past 30 yr have shown that electron opaque tracers circulating in the blood can label the vesicles of endothelial cells, direct evidence for serum albumin in vesicles has been reported only relatively recently. Schneeberger and Hamelin (1)used an immunoperoxidase method to detect native serum albumin in rat pulmonary capillaries. Albumin was normally present in the endothelial vesicular system, but when the animal's blood was exchanged for a fluorocarbon perfusate with progressively lower concentrations of albumin, the density of reaction product in the vesicles diminished, indi532
SUMMARY Although albumin circulating in the plasma can exchange with the contents of the extensive vesicUlar system of microvascular endothelium, the mechanism and quantitative significance of vesicular transport remains unclear. It has recently been shown that transport of electron opaque tracers can occur via the vesicular system, but the detailed ultrastructure is inconsistent with the transcytotic shuttling of single vesicles. Although most physiologic measurements show a major degree of convective coupling in the microvascular permeation of perfused tissues, recent reports suggest that convective coupling is low when albumin escapes from the intact circulation. Because albumin clearance in these stUdies Is consistent with values calculated from ferritin·labeling of vestcles in perfused microvessels, the possibility remains that macromolecular transendothelial transport occurs through transient communications within the vesicular system. AM REV RESPIR DIS 1992; 146:S32-S36
eating that most of the albumin had been washed out ofthem. Although this work was designed more to demonstrate the presence of a slowly exchanging layer of albumin on the endothelial cell surface, a later study by Milici and coworkers (15)was directed at albumin transport. Using a double antibody immunogold technique to demonstrate bovine serum albumin (BSA) in ultrathin frozen sections of mouse myocardial capillaries, Milici and coworkers (15) found that, after a few minutes perfusion with a BSA-containing solution prior to fixation, BSA could be detected throughout the vesicular system of the microvascular endothelium and within the pericapillary spaces. Occasional areas of accumulation of albumin in the pericapillary spaces were associated with labeled vesicles and not with intercellular clefts. Because albumin was not detected in clefts on the abluminal side of the tight junction, Milici and coworkers (15) concluded that transport via the clefts was negligible and that the principal transendothelial pathway for albumin was through the vesicles. Other work has shown that gold albumin complexes bind within the luminal vesicles and appear much more slowly at the abluminal surface (6, 15), whereas glycosylated albumin binds to both the luminal endothelial surface and within the vesicles, apparently permeating microvascular walls more quickly than native albumin (16, 17). Possible Mechanisms of Vesicular Transport Three hypotheses have been proposed at different times to account for transendothelial vesicular transport. These are illustrated in figure 1. The oldest and best known is the "shuttle" or "ferryboat" hypothesis (18), which is analogous to the transcytosis of "coated vesicles" in certain epithelial cells. According to this hypothesis, single caveolae at the luminal or abluminal surfaces "bud off' to become free vesicles within the cytoplasm be-
fore fusing with the opposite surface. Providing the cavity of the luminal and abluminal caveolae can equilibrate with the plasma and the perivascular fluid respectively, transport of solutes should occur down concentration gradients across the endothelium. There are two versions of how the caveolae form. The first suggests that each caveola arises de novo by invagination of the plasmalemmal membrane and is absorbed back into the membrane at the end of its cycle. The second views the vesicles as more permanent structures and suggests that after a vesicle has fused with a surface it forms a relatively stable caveola, which may then bud off again to form a free cytoplasmic vesicle. The second mechanism is sometimes referred to as the fusion-fission hypothesis (figure IB). It suggests that vesiclesdo not move from one endothelial surface to the other but merely far enough to fuse and communicate transiently with their immediate neighbors so as to allow intermixing of the vesicular contents (19). After one, two, or three fusions of this kind, a molecule that had initially diffused into a caveola at the luminal surface would be able to diffuse out of a caveola at the abluminal surface. A variant of this hypothesis is that vesicles that appear to be fused to each other in clusters and to luminal or abluminal caveolae are separated from each other by diaphragms that cyclicallyform and reform, allowingtransient communication between them. Thus, once again, a molecule entering a caveola at the luminal surface should be able to diffuse to the abluminal surface
1 From the Department of Physiology & Biophysics, St. Mary's Hospital Medical School, Imperial College of Science, Technology & Medicine, London, United Kingdom. 2 Correspondence and requests for reprints should be addressed to C. Charles Michel, Department of Physiology and Biophysics,St. Mary's Hospital Medical School, Imperial College of Science, Technology and Medicine, London W2 lPG, UK.
533
VESICULAR TRANSPORT OF ALBUMIN
•••
• ••
~
Fig. 1. Three hypotheses of vesicular transport. A. TransCy10sis or the vesicular shuttle. B.The fusion-fission hypothesis. C. Vesicular channels. Each hypothesis is shown as a sequence of events, though in the case of C, stable channels are considered to be a possibility.
i5»
•••
•••
•••
• •• • •
»s:»: ~
through the short-lived periods when the intervesicular diaphragms are missing or are permeable. The third hypothesis is that the vesiclesare permanently fused structures and form channels from the luminal tb the abluminal surfaces through their interconnections (20) (figure lC). Thus, they represent "pores" allowing diffusion and convection through the endothelium. A variant of this hypothesis would be that each pore or channel is transient but that on average there are a constant number of channels per unit area of endothelium. These three hypotheses should be distinguishable by experiment. The shuttle or transcytosis mechanism implies that the majority of vesicles are separate entities that move as single units from one surface of the cell to the other. The amount of tracer carried by a vesicle should not change once it has budded off to become a free cytoplasmic vesicle; and although the movements of the vesicles might be passive, i.e.,driven by Brownian motion (21), net transport would not be expected to be sensitiveto net fluid movements or transendothelial differencesin hydrostatic pressure. The processesof fusion and separation ofvesic1es at the plasmalemmal membranes, however, would be expected to be far more sensitive to tissue temperature than passive processes such as fluid flow or diffusion in aqueous solution. The fusion-fission hypothesis predicts there should be many fused groupings of vesicles and that the concentration of a tracer should be lower in labeled cytoplasmic and abluminal vesicles than in labeled luminal vesicles. It is unlikely that transport might be sensitive to luminal hydrostatic pressure though
it would be expected to be sensitive to temperature. If some vesiclesare fused to form transient or permanent channels, transport through them would be expected to be sensitive to transendothelial differences in hydrostatic pressure and be proportional to net fluid movements. Thus, one might favor the hypothesis that vesiclesare fused to form fluidconducting channels, if it could be shown that albumin transport through microvascular walls is proportional to net fluid filtration. This conclusion would be reinforced if, when tissue temperature was lowered, albumin transport was reduced in proportion to the reduction of fluid filtration and no more. Of course, physiologic findings of this kind would also be consistent with occasional pathways through the intercellular clefts where convective coupling of albumin transport might occur. Independent ultrastructural evidence for the existence of vesicular channels would be necessary to establish that the vesicles had any role at all in microvascular transport. Ultrastructural Evidence for and against Vesicular Transport
In the early 1980s the concept of vesicular transport was challenged on ultrastructural
Fig. 2. Three-dimensional reconstruction of a segment of endothelium from a microvessel in frog skeletal muscle. From reference 38, with permission.
grounds in a series of papers by Bundgaard and coworkers (22, 23) and Frekjaer-Jensen (24). This group provided evidence (initially from tannic acid staining and subsequently from serial ultrathin sections and serial reconstruction) that revealed that endothelial cell vesicles were arranged in fused clusters. The cavities of vesicleswithin a cluster communicated with each other and through that of a caveola with the extracellular space at either the luminal or the abluminal surface of the endothelium (figure 2). The serial section studies revealed that there were few, if any, vesicles existing as single entities in the endothelial cell cytoplasm. By using rapid freezing techniques, Frekjaer-Jensen (25)has been able to show that the fused structures are not artifacts of chemical fixation. He has also argued that the full extent of the interconnections between vesicular units can be demonstrated only when the serial sections are no thicker than 15 nm. If thicker sections are used, connections between vesicles can be missed. One very interesting feature to emerge from rapidly frozen tissue is the closenesswith which some vesiclescan approach each other without complete fusion, suggesting at times that their cavities are separated by a singleunit membrane. The earlier work ofthe Copenhagen group led Crone (26) to argue that since there were no free vesicles there could be no vesicular transport. Indeed, the simple picture of transcytosisdid appear to be inconsistent with their findings, and since also they had failed to find transendothelial channels of fused vesicles, they suggested that vesicles played no role in transendothelial transport. The ultrastructural picture described by the Copenhagen group was, however, consistent with the fusion-fission model. This hypothesis had emerged from studies carried out in my laboratory (19,27,28) on the rates of ferritin labeling of endothelial vesiclesin single microperfused frog mesenteric capillaries. We had found that the concentration of ferritin in labeled luminal caveolae was greater than its concentration in labeled cytoplasmic vesicles. Weappreciated that this finding was inconsistent with the shuttle hypothesis and proposed the fusion- fission model as an alternative (19). Our experiments yielded several other features of importance. First, there was a major degree of volume exclusion of ferritin within the vesicular cavities (19, 27). Second, while 20070 of the luminal caveolae labeled within 1 s, labeling of the remaining 80% increased very slowly if ferritin entered the open necks of these vesicles by diffusion (19,27). Third, entry into 80% of the appar-
534
C. CHARLES MICHEL
TABLE 1 MICROVASCULAR PERMEABILITY TO SERUM ALBUMIN
Microvascular Bed
Fig. 3. Electronmicrograph of endothelial cell (E) of rete mirabile of the eel after intravascular injection of terbium chloride. A highly electron-dense patch of terbium is seen in two abluminal caveolae and in the adjacent interstitial space between the endothelial cell and the pericyte (P). N is the nucleus of the pericyte and C is a perinuclear cisterna. Bar; 0.1 urn. From reference 29, with permission.
ently open luminal caveolaecould be inhibited by reducing the tissue temperature to 4 0 to 7 C (27, 28). Fourth, the rates of labeling of the luminal, cytoplasmic, and abluminal vesicles allowedestimates to be made of transport through the vesicular system (19, 28). Although our observations of progressive labeling of luminal cytoplasmic and abluminal vesiclesdid suggest transport through the vesicular system, it did not prove it. Ferritin could have been slowly penetrating the "cytoplasmic vesicles"via their interconnections with luminal caveolae and then labeling the abluminal vesiclesby diffusion along the abluminal surface after crossing the endothelium by a different route. Many studies of endothelial vesicularlabeling, however, have reported occasional abluminal caveolae that are heavily labeled with tracer (which had been administered intravascularly) and appear to be discharging their contents into a virtually tracer- free extravascular space. There are two possible interpretations of these electron micrographs. Either the tracer has been transported through the vesicular system or it has been transported via another route (i.e., an intercellular cleft) that lies just outside the plane of section. Using serial sections and serial reconstruction to analyze regions of capillaries in the rete mirabile of the eel, Wagner and Chen (29) have shown that localized deposits of electron opaque tracer can occur in the pericapil0
lary space in association with labeled vesicles but separate from intercellular clefts (figure 3). It would seem that this evidence allows no interpretation other than that the tracer crossed the endothelium via the vesicular system. As such it establishes that transport through the vesicular system does occur and so answers our first question. Furthermore, the ultrastructure and labeling characteristics of the vesicular system would suggest that the mechanism of transport is unlikely to be transcytosis. To assess whether transendothelial channels or the fusion-fission model can contribute significantly to net transport of albumin through microvascular walls, we must consider the functional estimates of albumin permeability. Measurements of Albumin Transport In Vivo
The transport of serum albumin from the circulating blood into the tissues has been estimated from plasma disappearance curves, from tissue accumulation, and from the flow and concentration of albumin in prenodal lymph. With one or two exceptions,the different techniques haveyieldedcomparable values for albumin clearance and albumin permeability. Furthermore, the values of albumin permeability for different microvascular beds reveal much less variation than might be anticipated from their widely differing perme-
Dog paw Dog intestine Dog heart Rat skeletal muscle Rat skeletal muscle Rat skeletal muscle Frog mesentery
Permeability (cm/s x 10")
References
4.7 2.9 2.9 6.9 0.67 0.916
39 39 39
5.5
40
35 37 36
abilities to small molecules and net fluid movements. By contrast, the albumin permeabilities of monolayers of cultured endothelium are two orders of magnitude greater than the in vivo values (30). Occasional large intercellular gaps in the monolayers (30) probably account for their high permeabilities and these studies will be considered no further. Some values for albumin permeability are summarized in table 1. One of the striking results to emerge from the studies on bloodlymph transport is the dependence of net protein clearance upon net filtration (as judged by local lymph flow) (31-34). This is seen for molecules as large as fibrinogen and dextran 110 as well as for albumin. Taking this as evidence for convective transport of macromolecules, the fluxes of solutes from plasma to lymph have been analyzed in terms of ultrafiltration and diffusion through microvascular walls via two populations of pores: a group of small pores with radii varying from 4 to 8 nm and a very small number of large pores with radii in the range of 20 to 25 nm (33). These analyses of blood-lymph transport depend on the assumption that after a change in filtration rate at the microvascular wall, a steady state is established between the composition of the filtrate entering the interstitium and the fluid leaving the interstitium to enter the lymphatics. Failure to reach a steady state after capillary filtration has been increased leads to an overestimate of the convectivecomponent of macromolecular transport. Those working in the field are aware of this difficulty, and they have attempted to minimize it by making serial estimates of lymph protein concentration in which no change between successivesamples can be detected. Such an approach is reasonable in some regional microvascular beds where the endothelium is fenestrated and microvascular hydraulic permeability is very high. It is much more questionable in microvascular beds wherethe endothelium is continuous and hydraulic permeability is low. It is of considerable importance, therefore, that evidence favoring convective microvascular transport of serum albumin has also been obtained by Rippe and coworkers (35) from measurements of tissue accumulation in skeletal muscle using the perfused hindlimb preparation of the rat; figure 4 is taken from this study and shows the dependence
535
VESICULAR TRANSPORT OF ALBUMIN
INITIAL ALBUMIN CLEARANCE,ml/mln 1li00g
.
008
»
Introduction Serum albumin plays a central role in microvascular exchange. It binds both specifically and nonspecifically to the surfaces of endothelial cells (1-6), reducing their permeability to fluid and macromolecules (7-10). As a carrier of smaller molecules, it determines the availability of fatty acids and certain amino acids, steroids, drugs, etc. for microvascular exchange (11). Albumin is also responsible for a large fraction of the plasma oncotic pressure so that its permeation through microvascular walls and accumulation in the interstitium is a major determinant of the distribution of fluid between the blood and the tissues (12). Despite its importance, the mechanism of albumin transport through microvascular walls remains controversiaL Although some suggest that serum albumin crosses through the extensivevesicular system of endothelium by transcytosis (13), others believe that the arguments against vesicular transport are compelling, and they assert that albumin (together with other macromolecules) is carried by convection through a small number of relativelylarge pores, which probably lie between the endothelial cells (14). These conflicting points of vieware based on very different lines of evidence. In assessing this evidence, I shall consider two sequential questions. The first is whether there is a balance of evidence for any transport of macromolecules through microvascular walls via the endothelial vesicular system.Assuming there is such a balance, the second question is whether transport via the vesicular system makes a significant contribution to total exchangeof macromolecules between the blood and the tissues. Before considering these general questions, however, some studies that implicate the vesicular system in the transendothelial transport of serum albumin will be briefly reviewed. Ultrastructural Evidence for Transendothelial Transport of Serum Albumin Even though many studies over the past 30 yr have shown that electron opaque tracers circulating in the blood can label the vesicles of endothelial cells, direct evidence for serum albumin in vesicles has been reported only relatively recently. Schneeberger and Hamelin (1)used an immunoperoxidase method to detect native serum albumin in rat pulmonary capillaries. Albumin was normally present in the endothelial vesicular system, but when the animal's blood was exchanged for a fluorocarbon perfusate with progressively lower concentrations of albumin, the density of reaction product in the vesicles diminished, indi532
SUMMARY Although albumin circulating in the plasma can exchange with the contents of the extensive vesicUlar system of microvascular endothelium, the mechanism and quantitative significance of vesicular transport remains unclear. It has recently been shown that transport of electron opaque tracers can occur via the vesicular system, but the detailed ultrastructure is inconsistent with the transcytotic shuttling of single vesicles. Although most physiologic measurements show a major degree of convective coupling in the microvascular permeation of perfused tissues, recent reports suggest that convective coupling is low when albumin escapes from the intact circulation. Because albumin clearance in these stUdies Is consistent with values calculated from ferritin·labeling of vestcles in perfused microvessels, the possibility remains that macromolecular transendothelial transport occurs through transient communications within the vesicular system. AM REV RESPIR DIS 1992; 146:S32-S36
eating that most of the albumin had been washed out ofthem. Although this work was designed more to demonstrate the presence of a slowly exchanging layer of albumin on the endothelial cell surface, a later study by Milici and coworkers (15)was directed at albumin transport. Using a double antibody immunogold technique to demonstrate bovine serum albumin (BSA) in ultrathin frozen sections of mouse myocardial capillaries, Milici and coworkers (15) found that, after a few minutes perfusion with a BSA-containing solution prior to fixation, BSA could be detected throughout the vesicular system of the microvascular endothelium and within the pericapillary spaces. Occasional areas of accumulation of albumin in the pericapillary spaces were associated with labeled vesicles and not with intercellular clefts. Because albumin was not detected in clefts on the abluminal side of the tight junction, Milici and coworkers (15) concluded that transport via the clefts was negligible and that the principal transendothelial pathway for albumin was through the vesicles. Other work has shown that gold albumin complexes bind within the luminal vesicles and appear much more slowly at the abluminal surface (6, 15), whereas glycosylated albumin binds to both the luminal endothelial surface and within the vesicles, apparently permeating microvascular walls more quickly than native albumin (16, 17). Possible Mechanisms of Vesicular Transport Three hypotheses have been proposed at different times to account for transendothelial vesicular transport. These are illustrated in figure 1. The oldest and best known is the "shuttle" or "ferryboat" hypothesis (18), which is analogous to the transcytosis of "coated vesicles" in certain epithelial cells. According to this hypothesis, single caveolae at the luminal or abluminal surfaces "bud off' to become free vesicles within the cytoplasm be-
fore fusing with the opposite surface. Providing the cavity of the luminal and abluminal caveolae can equilibrate with the plasma and the perivascular fluid respectively, transport of solutes should occur down concentration gradients across the endothelium. There are two versions of how the caveolae form. The first suggests that each caveola arises de novo by invagination of the plasmalemmal membrane and is absorbed back into the membrane at the end of its cycle. The second views the vesicles as more permanent structures and suggests that after a vesicle has fused with a surface it forms a relatively stable caveola, which may then bud off again to form a free cytoplasmic vesicle. The second mechanism is sometimes referred to as the fusion-fission hypothesis (figure IB). It suggests that vesiclesdo not move from one endothelial surface to the other but merely far enough to fuse and communicate transiently with their immediate neighbors so as to allow intermixing of the vesicular contents (19). After one, two, or three fusions of this kind, a molecule that had initially diffused into a caveola at the luminal surface would be able to diffuse out of a caveola at the abluminal surface. A variant of this hypothesis is that vesicles that appear to be fused to each other in clusters and to luminal or abluminal caveolae are separated from each other by diaphragms that cyclicallyform and reform, allowingtransient communication between them. Thus, once again, a molecule entering a caveola at the luminal surface should be able to diffuse to the abluminal surface
1 From the Department of Physiology & Biophysics, St. Mary's Hospital Medical School, Imperial College of Science, Technology & Medicine, London, United Kingdom. 2 Correspondence and requests for reprints should be addressed to C. Charles Michel, Department of Physiology and Biophysics,St. Mary's Hospital Medical School, Imperial College of Science, Technology and Medicine, London W2 lPG, UK.
533
VESICULAR TRANSPORT OF ALBUMIN
•••
• ••
~
Fig. 1. Three hypotheses of vesicular transport. A. TransCy10sis or the vesicular shuttle. B.The fusion-fission hypothesis. C. Vesicular channels. Each hypothesis is shown as a sequence of events, though in the case of C, stable channels are considered to be a possibility.
i5»
•••
•••
•••
• •• • •
»s:»: ~
through the short-lived periods when the intervesicular diaphragms are missing or are permeable. The third hypothesis is that the vesiclesare permanently fused structures and form channels from the luminal tb the abluminal surfaces through their interconnections (20) (figure lC). Thus, they represent "pores" allowing diffusion and convection through the endothelium. A variant of this hypothesis would be that each pore or channel is transient but that on average there are a constant number of channels per unit area of endothelium. These three hypotheses should be distinguishable by experiment. The shuttle or transcytosis mechanism implies that the majority of vesicles are separate entities that move as single units from one surface of the cell to the other. The amount of tracer carried by a vesicle should not change once it has budded off to become a free cytoplasmic vesicle; and although the movements of the vesicles might be passive, i.e.,driven by Brownian motion (21), net transport would not be expected to be sensitiveto net fluid movements or transendothelial differencesin hydrostatic pressure. The processesof fusion and separation ofvesic1es at the plasmalemmal membranes, however, would be expected to be far more sensitive to tissue temperature than passive processes such as fluid flow or diffusion in aqueous solution. The fusion-fission hypothesis predicts there should be many fused groupings of vesicles and that the concentration of a tracer should be lower in labeled cytoplasmic and abluminal vesicles than in labeled luminal vesicles. It is unlikely that transport might be sensitive to luminal hydrostatic pressure though
it would be expected to be sensitive to temperature. If some vesiclesare fused to form transient or permanent channels, transport through them would be expected to be sensitive to transendothelial differences in hydrostatic pressure and be proportional to net fluid movements. Thus, one might favor the hypothesis that vesiclesare fused to form fluidconducting channels, if it could be shown that albumin transport through microvascular walls is proportional to net fluid filtration. This conclusion would be reinforced if, when tissue temperature was lowered, albumin transport was reduced in proportion to the reduction of fluid filtration and no more. Of course, physiologic findings of this kind would also be consistent with occasional pathways through the intercellular clefts where convective coupling of albumin transport might occur. Independent ultrastructural evidence for the existence of vesicular channels would be necessary to establish that the vesicles had any role at all in microvascular transport. Ultrastructural Evidence for and against Vesicular Transport
In the early 1980s the concept of vesicular transport was challenged on ultrastructural
Fig. 2. Three-dimensional reconstruction of a segment of endothelium from a microvessel in frog skeletal muscle. From reference 38, with permission.
grounds in a series of papers by Bundgaard and coworkers (22, 23) and Frekjaer-Jensen (24). This group provided evidence (initially from tannic acid staining and subsequently from serial ultrathin sections and serial reconstruction) that revealed that endothelial cell vesicles were arranged in fused clusters. The cavities of vesicleswithin a cluster communicated with each other and through that of a caveola with the extracellular space at either the luminal or the abluminal surface of the endothelium (figure 2). The serial section studies revealed that there were few, if any, vesicles existing as single entities in the endothelial cell cytoplasm. By using rapid freezing techniques, Frekjaer-Jensen (25)has been able to show that the fused structures are not artifacts of chemical fixation. He has also argued that the full extent of the interconnections between vesicular units can be demonstrated only when the serial sections are no thicker than 15 nm. If thicker sections are used, connections between vesicles can be missed. One very interesting feature to emerge from rapidly frozen tissue is the closenesswith which some vesiclescan approach each other without complete fusion, suggesting at times that their cavities are separated by a singleunit membrane. The earlier work ofthe Copenhagen group led Crone (26) to argue that since there were no free vesicles there could be no vesicular transport. Indeed, the simple picture of transcytosisdid appear to be inconsistent with their findings, and since also they had failed to find transendothelial channels of fused vesicles, they suggested that vesicles played no role in transendothelial transport. The ultrastructural picture described by the Copenhagen group was, however, consistent with the fusion-fission model. This hypothesis had emerged from studies carried out in my laboratory (19,27,28) on the rates of ferritin labeling of endothelial vesiclesin single microperfused frog mesenteric capillaries. We had found that the concentration of ferritin in labeled luminal caveolae was greater than its concentration in labeled cytoplasmic vesicles. Weappreciated that this finding was inconsistent with the shuttle hypothesis and proposed the fusion- fission model as an alternative (19). Our experiments yielded several other features of importance. First, there was a major degree of volume exclusion of ferritin within the vesicular cavities (19, 27). Second, while 20070 of the luminal caveolae labeled within 1 s, labeling of the remaining 80% increased very slowly if ferritin entered the open necks of these vesicles by diffusion (19,27). Third, entry into 80% of the appar-
534
C. CHARLES MICHEL
TABLE 1 MICROVASCULAR PERMEABILITY TO SERUM ALBUMIN
Microvascular Bed
Fig. 3. Electronmicrograph of endothelial cell (E) of rete mirabile of the eel after intravascular injection of terbium chloride. A highly electron-dense patch of terbium is seen in two abluminal caveolae and in the adjacent interstitial space between the endothelial cell and the pericyte (P). N is the nucleus of the pericyte and C is a perinuclear cisterna. Bar; 0.1 urn. From reference 29, with permission.
ently open luminal caveolaecould be inhibited by reducing the tissue temperature to 4 0 to 7 C (27, 28). Fourth, the rates of labeling of the luminal, cytoplasmic, and abluminal vesicles allowedestimates to be made of transport through the vesicular system (19, 28). Although our observations of progressive labeling of luminal cytoplasmic and abluminal vesiclesdid suggest transport through the vesicular system, it did not prove it. Ferritin could have been slowly penetrating the "cytoplasmic vesicles"via their interconnections with luminal caveolae and then labeling the abluminal vesiclesby diffusion along the abluminal surface after crossing the endothelium by a different route. Many studies of endothelial vesicularlabeling, however, have reported occasional abluminal caveolae that are heavily labeled with tracer (which had been administered intravascularly) and appear to be discharging their contents into a virtually tracer- free extravascular space. There are two possible interpretations of these electron micrographs. Either the tracer has been transported through the vesicular system or it has been transported via another route (i.e., an intercellular cleft) that lies just outside the plane of section. Using serial sections and serial reconstruction to analyze regions of capillaries in the rete mirabile of the eel, Wagner and Chen (29) have shown that localized deposits of electron opaque tracer can occur in the pericapil0
lary space in association with labeled vesicles but separate from intercellular clefts (figure 3). It would seem that this evidence allows no interpretation other than that the tracer crossed the endothelium via the vesicular system. As such it establishes that transport through the vesicular system does occur and so answers our first question. Furthermore, the ultrastructure and labeling characteristics of the vesicular system would suggest that the mechanism of transport is unlikely to be transcytosis. To assess whether transendothelial channels or the fusion-fission model can contribute significantly to net transport of albumin through microvascular walls, we must consider the functional estimates of albumin permeability. Measurements of Albumin Transport In Vivo
The transport of serum albumin from the circulating blood into the tissues has been estimated from plasma disappearance curves, from tissue accumulation, and from the flow and concentration of albumin in prenodal lymph. With one or two exceptions,the different techniques haveyieldedcomparable values for albumin clearance and albumin permeability. Furthermore, the values of albumin permeability for different microvascular beds reveal much less variation than might be anticipated from their widely differing perme-
Dog paw Dog intestine Dog heart Rat skeletal muscle Rat skeletal muscle Rat skeletal muscle Frog mesentery
Permeability (cm/s x 10")
References
4.7 2.9 2.9 6.9 0.67 0.916
39 39 39
5.5
40
35 37 36
abilities to small molecules and net fluid movements. By contrast, the albumin permeabilities of monolayers of cultured endothelium are two orders of magnitude greater than the in vivo values (30). Occasional large intercellular gaps in the monolayers (30) probably account for their high permeabilities and these studies will be considered no further. Some values for albumin permeability are summarized in table 1. One of the striking results to emerge from the studies on bloodlymph transport is the dependence of net protein clearance upon net filtration (as judged by local lymph flow) (31-34). This is seen for molecules as large as fibrinogen and dextran 110 as well as for albumin. Taking this as evidence for convective transport of macromolecules, the fluxes of solutes from plasma to lymph have been analyzed in terms of ultrafiltration and diffusion through microvascular walls via two populations of pores: a group of small pores with radii varying from 4 to 8 nm and a very small number of large pores with radii in the range of 20 to 25 nm (33). These analyses of blood-lymph transport depend on the assumption that after a change in filtration rate at the microvascular wall, a steady state is established between the composition of the filtrate entering the interstitium and the fluid leaving the interstitium to enter the lymphatics. Failure to reach a steady state after capillary filtration has been increased leads to an overestimate of the convectivecomponent of macromolecular transport. Those working in the field are aware of this difficulty, and they have attempted to minimize it by making serial estimates of lymph protein concentration in which no change between successivesamples can be detected. Such an approach is reasonable in some regional microvascular beds where the endothelium is fenestrated and microvascular hydraulic permeability is very high. It is much more questionable in microvascular beds wherethe endothelium is continuous and hydraulic permeability is low. It is of considerable importance, therefore, that evidence favoring convective microvascular transport of serum albumin has also been obtained by Rippe and coworkers (35) from measurements of tissue accumulation in skeletal muscle using the perfused hindlimb preparation of the rat; figure 4 is taken from this study and shows the dependence
535
VESICULAR TRANSPORT OF ALBUMIN
INITIAL ALBUMIN CLEARANCE,ml/mln 1li00g
.
008
»
