INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

THE CLEFT between vascular endothelial cells is widely accepted as the primary site of convective water flux and diffusive flux of small to intermediate size solutes. The primary barrier within the cleft is thought to be the tight junction. Tight junctions appear as focal occlusions of the paracellular pathway in thin-section electron micrographs. In freeze-fracture replicas the tight junctions are seen as rows or strands of intramembrane particles that apparently encircle the cell. However, evidence now indicates that peripheral vascular tight junctions are a discontinuous barrier, interrupted by open segments called strand gaps (1, 6, 31). In frog mesentery capillaries, these gaps are an average 150 nm long and occur about every 2-4 µm around the endothelial cell periphery (1). Whereas very little is known about direct regulation of tight-junction structure, several reports indicate that intracellular adenosine 3',5'-cyclic monophosphate (cAMP) induces decreased paracellular permeability (21, 27, 34, 36), and this may be correlated with an increase in tight-junction number or complexity (10, 39).

Many previous studies have found that increased intracellular cAMP can block the inflammatory response in a variety of experimental models. Since early transmission electron microscopy studies, it has been widely believed that endothelial cells respond to several inflammatory stimuli (e.g., histamine, serotonin, and thrombin) with a loss of endothelial-to-endothelial adhesion and cell retraction, inducing the formation of ~0.1-µm wide inflammatory gaps (23). The most widely accepted model of gap formation, based on the well-understood model of vascular smooth muscle contraction, suggests that an actin/myosin dependent increase in cell tension produces a centripetal force on the cytoskeleton, drawing the peripheral regions away from bordering cells (40). This Ca²⁺-dependent mechanism is thought to depend on myosin light-chain kinase and to be regulated by cAMP-dependent protein kinase A (PKA) phosphorylation of myosin light-chain kinase. The PKA phosphorylation of myosin light-chain kinase reduces the activity of the kinase and decreases tension development.

Although the mechanism is not definite, numerous studies confirm that elevated cAMP modulates the inflammatory response of endothelial cells. Simultaneous stimulation of adenylate cyclase (either by activating prostaglandin receptors or by stimulating adenylate cyclase directly with forskolin) and phosphodiesterase (PDE) inhibition with zardaverine (a strong inhibitor of PDE type III and type IV) was highly effective in inhibiting permeability increases induced by H₂O₂ in isolated rabbit lung (29). Another lung study found that a variety of agents that increase cAMP (dibutyryl-cAMP, isoproterenol, prostaglandin E₁, and rolipram) were effective to protect against permeability increases induced by ischemia-reperfusion injury (4). In canine forelimb, permeability increases to both fluid flow (edema) and protein extravasation (lymph protein) that were induced by histamine were diminished by either isoproterenol or norepinephrine (14). In a recent whole organ investigation, the ₂-agonist formoterol effectively reduced the extravasation of albumin and the high molecular weight tracer monastral blue that was inducible by either substance P or vagal nerve stimulation in tracheas of intact rats (3). Finally, we note that a previous study from our group also confirms the blocking effect of a cAMP analog on inflammatory stimulation (16). Solutions containing ATP (10 µM) stimulate a 5- to 10-fold increase in hydraulic conductivity (L_p) in both capillaries and postcapillary venules of frog and hamster mesentery. This characteristic transient inflammatory response was nearly abolished by 15 min pretreatment with 8-bromoadenosine 3',5'-cyclic monophosphate (2 mM). Thus a wide variety of whole organ and in situ observations is consistent with the hypothesis that increased intracellular cAMP concentration has an inhibitory effect on inflammatory permeability increases. However, there have been no previous reports from in vivo studies that increased cAMP can reduce permeability from the control state.

Several studies have shown that increased cAMP concentration can block the effects of inflammatory agents on cultured endothelial cell monolayers. Forskolin, cholera toxin, and prostaglandin E₁, all stimulators of intracellular cAMP concentration, effectively blocked the permeability-increasing effects of thrombin and hemolysin on endothelial monolayers (35). Similarly, inhibition of cAMP-specific PDE by rolipram or zardaverine was effective alone and in combination with adenylyl cyclase stimulation to block the permeability-stimulating effects of thrombin and hemolysin (35) as well as H₂O₂ (36). Elevation of cAMP also blocks monolayer permeability stimulated by histamine (30) and phorbol myristate acetate (13).

An important difference between cell culture studies and in vivo microvascular observations is that increased intracellular cAMP has been reported to reduce the basal permeability of cultured endothelial monolayers, whereas there are no reports of cAMP reducing baseline permeability in vivo. Forskolin, cholera toxin, or isoproterenol treatment of endothelial cell monolayers was shown to reversibly increase the transendothelial electrical resistance and significantly reduce passage of the macromolecule horseradish peroxidase (21, 34). Similarly, high electrical resistance monolayers of bovine brain endothelial cells were established when cAMP levels were increased by the PDE type IV inhibitor RO-20-1724 or by addition of cell-permeant forms of cAMP (27). These latter studies provide indirect evidence that the structure of the paracellular pathway is modified by high levels of intracellular cAMP.

There are very few reports of a relationship between cAMP and tight-junction structure. In one analysis of epithelium, the mean number of strands in necturus gallbladder epithelial cells increased from 8 to 10 and paracellular electrical resistance increased when monolayers were treated with a cell-permeant form of cAMP (10). In a recent endothelial study, cultured brain microvascular endothelial cells were found to retain tight-junction complexity when the cells were treated with forskolin and the PDE inhibitor RO-20-1724 (39). Proteins recognized as part of the tight-junction complex now include ZO-1, ZO-2, cingulin, 7H6 antigen, and occludin (2, 12). Occludin, the only known integral membrane protein, is hypothesized to form homotypic adhesions that occlude the paracellular pathway. Occludin, ZO-1, and ZO-2 are subject to serine-threonine phosphorylation and therefore are potential targets for cAMP-dependent protein kinase regulation (28).

In the present study we investigated the effects of cAMP stimulation on L_p and morphology of the paracellular pathway in continuous capillaries of frog mesentery. Stimulation of adenylate cyclase and inhibition of cAMP-specific PDE, either alone or in combination, were effective in reducing baseline permeability as measured by L_p. Vessels that responded strongly to combined treatment also exhibited an increase in the mean number of junction strands. These results suggest that the number of junction strands is regulated by intracellular levels of cAMP and that endothelium can modulate local vascular permeability via this mechanism. This study reports significant reductions in baseline microvascular L_p and is the first to demonstrate an induced change in the mean number of microvascular endothelial cell tight junctions.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Animals and video microscopy. All experiments were carried out on male leopard frogs (Rana pipiens, 2.5-3 in.) obtained from J. M. Hazen (Vermont). Animals were pithed, and the gut, which was exposed through a small lateral incision, was gently arranged around a transparent pillar enabling transillumination of the mesentery. The surface of the mesentery was continuously bathed in frog Ringer solution and maintained at 18-22°C. The circulation remained intact, and only animals with rapidly flowing blood, without white cell margination or adhesion, were used. Single capillaries with straight nonbranched segments of at least 700 µm were chosen for the study. Only one vessel was used in each mesentery. Both true capillaries and venular capillaries, generally 15-30 µm in diameter, were represented in the study groups. The vessels were observed with an upright microscope using a ×10 long-working distance lens. Experiments were recorded on videotape for playback and analysis. Further details have been published (24).

Solutions. Frog Ringer solution, used for suffusion of the mesentery and as the basis for all perfusion solutions, was composed of (in mM) 111 NaCl, 2.4 KCl, 1.0 MgSO₄, 1.1 CaCl₂, 0.195 NaHCO₃, 5.5 glucose, and 5 N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid (HEPES), and Na HEPES. Bovine serum albumin (10 mg/ml; BSA; Sigma A4378) was added to Ringer solution to prepare all perfusates. The pH was balanced to 7.4 by adjusting the ratio of HEPES acid to base. Rolipram (supplied by Eisai London Laboratories), a selective inhibitor of the cAMP-specific PDE IV, was prepared as a 50 mM stock solution in ethanol. Forskolin (Biomol) was prepared as a 25 mM stock solution in ethanol. These stocks were kept at 20°C and not used if older than 1 mo. Isoproterenol stock was 0.2 mg/ml isoproterenol hydrochloride injection. Appropriate control solutions contained ethanol as a vehicle.

Measurement of L_p. L_p was measured with the modified Landis technique, which allows estimation of fluid flow across the walls of single vessels during downstream occlusion (24). Hamster erythrocytes that were separated from plasma and free of all leukocytes were used as flow markers in all perfusates (~4 ml packed cells per 100 ml solution). We determined the axial velocity of the flow markers under controlled hydraulic pressure difference across the vessel walls. The initial transcapillary water flow per unit area of capillary wall (J_v/A) per unit pressure provided an estimate of L_p. We used a range of 20-40 cmH₂O intravascular hydrostatic pressure. The assumptions and limitations of the measurement have been discussed in detail (24). Briefly, in those experiments using multiple pressure determinations of fluid flow, L_p was determined as the slope of the regression of J_v/A against applied hydraulic pressure. For the majority of experiments we used a single pressure estimate of J_v/A and assumed that the net effective pressure determining fluid flow was equal to the applied hydraulic pressure minus 3 cmH₂O, the approximate oncotic pressure contributed by the BSA (18).

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Effects on L_p. Perfusion with solutions containing both rolipram (10 µM) and forskolin (5 µM) elicited a consistent decrease in L_p. Data from two vessels (Fig. 1, A and B) illustrate that the initial response was observed within 5 min and that a further decrease in L_p was observed for up to 55 min in some vessels. Vessels with normal-to-low control L_p responded to the treatment as well as vessels that had higher control L_p. One of the most striking features of the study is that some vessels responded with L_p falling to levels that were considerably below the typical range for frog mesenteric vessels. The data of Fig. 1B show one such response in which the L_p fell to below 0.3 × 10⁷ cm/ (s · cmH₂O) after ~25 min of treatment. For that vessel the response was an approximately 10-fold fall from control L_p, a very large proportional decrease.

View larger version (28K): [in this window] [in a new window]   Fig. 1.   Data from 3 individual experiments. A: data from a vessel with a control hydraulic conductivity (Lp) in upper part of normal range. This vessel responded to treatment with rolipram (10 µM) and forskolin (5 µM) with a gradually declining Lp over ~40 min. B: data from a vessel with a lower intial Lp. This vessel responded with an early fall to about one-half of control Lp and continued to decline over at least 20 min before apparently stabilizing near 15-20% of the control Lp. C: sham control reperfusion was one of many controls to establish that a gradually falling Lp, as in the test vessels, is not a feature common to artificially perfused vessels. All data are from single occlusions. BSA, Bovine serum albumin.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Representative data from a single vessel. Throughout the experiment there was a linear relationship between flow per unit area (Jv/A) and pressure. Initial perfusion with BSA control solution (open circles) over ~12 min established control Lp of ~6.1 × 107 cm/(s · cmH2O). Vessel was recannulated and perfused with test solution containing rolipram (10 µM) and forskolin (5 µM) for 30 min. Data labeled 12 min, collected during the 7- to 15-min period of test perfusion, yield a regression Lp of 5.1 × 107 cm/(s · cmH2O). Data labeled 30 min, collected during the 28- to 32-min period, give a regression Lp of 2.5 × 107 cm/(s · cmH2O).

View larger version (18K): [in this window] [in a new window]   Fig. 3.   Ratio of test Lp (Ltestp) to control Lp (Lcontrolp) as a function of time after perfusion was initiated with test solution. Perfusion with either rolipram (10 µM, Roli) alone or rolipram (10 µM) with forskolin (5 µM, Roli & Forsk) induced a rapid fall in Lp. Time 0 min represents control level. Values are means ± SE. * Difference from control group at same time; P < 0.05.

View larger version (16K): [in this window] [in a new window]   Fig. 4.   Effect of isoproterenol (Iso) over time on Ltestp/Lcontrolp. Perfusion with Iso (10 µM) alone or with Roli (10 µM) induced a rapid initial fall in mean Lp shown here as a ratio to control values. These agonist-stimulated responses were comparable to the responses induced by forskolin and rolipram both in time course and in magnitude.

Ultrastructural correlates of decreased L_p. Eight vessels from the group treated with both rolipram (10 µM) and forskolin (5 µM) were chosen for ultrastructural investigation (Table 1). These vessels all had a strong response to the treatment as evidenced by an average fourfold fall in L_p during the measurement period. The mean ratio of final L_p to initial L_p was 0.25 ± (SE) 0.05 (n = 7). Initial mean L_p was 3.3 ± (SE) 0.65 × 10⁷ cm/(s · cmH₂O), whereas final mean L_p decreased to 0.79 ± 0.21 × 10⁷ cm/(s · cmH₂O). For comparison we chose six vessels that were reperfused with the control solution and otherwise treated just as the experimental vessels. Control vessels perfused for comparable times showed no change in L_p. The mean initial L_p for the control group [4.6 ± (SE) 0.85 × 10⁷ cm/(s · cmH₂O)] was not different from the final L_p [4.4 ± (SE) 0.34 × 10⁷ cm/(s · cmH₂O); n = 6].

View larger version (128K): [in this window] [in a new window]   Fig. 5.   A and B: representative electron micrographs of clefts between endothelial cells from vessels perfused with rolipram (10 µM) and forskolin (5 µM). Tight junctions appeared as punctate "kisses" where outer leaflets of membrane bilayers come to closest contact (arrowheads). C: cleft from a control vessel. Except for an increase in mean number of tight junctions per cleft, test vessels were morphologically indistinguishable from control vessels despite a fourfold lower mean Lp. Small arrows with lines illustrate ends of contour measurement of cleft length. L, lumen of vessel; I, interstitium; bar 0.1 µm.

View larger version (21K): [in this window] [in a new window]   Fig. 6.   Distribution showing variation in number of tight junctions per endothelial cleft. Height of bars represents percentage of clefts having associated number of tight junctions. Data are pooled from all control experiments in top histogram and from all test experiments in bottom histogram. From 6 control vessels we examined 192 individual clefts. Of those, 78 clefts (41%) had only 1 tight junction. From the 8 test vessels, we examined 303 clefts, of which 66 (22%) had only 1 tight junction. Mean number of tight junctions per cleft was higher in test group [2.2 ± (SD) 0.4] than in control [1.7 ± (SD) 0.3] (P < 0.05, Mann-Whitney U ).

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

We report two new and important observations with this study. The first is that treatment with agents known to increase intracellular cAMP levels induces a decreased L_p of single capillaries. We found that mean L_p fell to 37% of control during 25 min of treatment with combined inhibition of PDE IV and stimulation of adenylate cyclase. Some individual vessels responded by decreasing to 10% of initial control L_p. Such large proportional decreases from control permeability for intact vessels have not been reported previously for treatment with cAMP-enhancing agents. Our second important observation was that the decreased permeability was associated with an increase in the mean number of tight-junction strands between endothelial cells. The latter observation is consistent with the hypothesis that the number of tight-junction strands is a major determinant of L_p in peripheral exchange vessels and that modulation of the tight junction is a means of permeability regulation. The specific intracellular signaling cascades that regulate permeability are not known. However, many molecules of the junctional region are potential targets.

The tight junction is a large macromolecular complex with numerous constituents. Several molecular components are now recognized (2). Proteins associated with the cytoplasmic side of the tight-junction complex include ZO-1, ZO-2, cingulin, and the 7H6 antigen. Only one tight-junction protein, occludin, has been demonstrated to have both cytoplasmic and extracellular domains (12). This integral membrane protein is hypothesized to form homotypic adhesions at the tight junction, perhaps using the peptide loops that have a relatively high degree of hydrophobicity. Either regulation of the conformation of occludin itself or the number of junction strands could determine permeability of tight junctions in both epithelia and endothelia.

Epithelial monolayers form characteristic adherence regions just basal to the tight-junction complex. These adherens junctions have been widely investigated, and numerous protein components are now known (32). Whereas endothelial monolayers do not form morphologically distinct adherens junctions, they do express many proteins homologous to those of epithelial adherens plaques. Furthermore, the adhesion and barrier properties of endothelial monolayers depend on expression and conformation of adherens proteins. Of particular importance are cadherins molecules that have homotypic recognition sites (15). Conformation and adhesion of the cadherins are dependent on cytoplasmic interaction with -catenin, -catenin, and plakoglobin. The adherens complex linkage to the actin cytoskeleton is mediated through -catenin, and the catenins in turn have specific interaction sites with actin and other elements of the cytoskeleton. A newly reported cytoplasmic component of adherens plaque has been found associated with a variety of epithelial as well as endothelial cells (17). Tyrosine phosphorylation of -catenin has been found to alter cadherin adhesiveness and increase permeability of brain endothelial cell monolayers (32). Tyrosine kinases such as src, yes, and lyn have been found associated with adherens complex proteins, as has the phosphotyrosine phosphatase PTPµ (5). ZO-1 and the recently described catenin p120 are also substrates for tyrosine phosphorylation and therefore potentially are control elements in the permeability regulatory pathway (33). The interaction between cadherin adhesion and tight-junction permeability is not known. However, regulation of the relative activities of these tyrosine kinases and phosphatases provides a potential means to regulate monolayer permeability through the control of adherens adhesion.

Our primary observation that treatment with cAMP-enhancing agents decreases L_p has not been investigated previously using an intact vasculature. One associated study from our group reported that in vessels treated with the membrane-permeant cAMP analog, 8-bromo-cAMP, the L_p tended to decrease for 10-15 min and then stabilize (16). The typical response was a decrease to 80% of control. Because those experiments were performed in the absence of PDE inhibition, it is not surprising that only a slight decrease was observed. Similarly, treatment of single frog mesentery vessels with norepinephrine (1 µM) elicited no change in L_p (19).

We tested the effects of isoproterenol to investigate whether -adrenergic receptor-mediated stimulation of cAMP levels could decrease L_p. Such data are more relevant to possible physiological regulation than direct adenylate cyclase stimulation or PDE inhibition. We found that 10 µM isoproterenol decreased L_p with a time course and magnitude very similar to that of the combined effects of forskolin and rolipram. Intriguingly, there was no further decrease in L_p when rolipram was also present. Although it is tempting to suggest that isoproterenol alone was slightly more effective than the combined effects of isoproterenol and rolipram, there was no statistical difference between the two groups in Fig. 4. The effectiveness of isoproterenol to improve the barrier properties of cultured endothelial cells was previously reported (21). The lack of an additive effect of rolipram contrasts with the reported synergistic effect of isoproterenol and rolipram in inhibiting the inflammatory effects of ischemia-reperfusion injury in lung (4).

Because our single-vessel preparation did not enable measurement of cAMP levels, we cannot know the amount by which, or even if, our treatments increased cAMP. However, several studies using endothelial cell culture have demonstrated that combined treatment with forskolin and rolipram produces readily measurable increases in the intracellular cAMP. Whereas treatment with 1 µM forskolin alone was found to double the cAMP content measured by radioimmunoassay, treatment with 10 µM rolipram in addition to 1 µM forskolin induced a fivefold increase in cAMP in pulmonary endothelial cells (35). Similarly, using human umbilical artery endothelial cells, we found that forskolin (25 µM) alone induced an immeasurable change in cAMP, 3-isobutyl-1-methylxanthine (1 µM, a nonselective PDE inhibitor) alone approximately doubled the cAMP content, and combined treatment with both drugs produced an 8- to 10-fold increase in cAMP content within 5 min (21). When used with bovine pulmonary artery endothelial cells, forskolin (5 µM) induced a 2.5-fold increase in cAMP content (34). A substantially higher responsiveness was observed in rat brain microvascular endothelium for which isoproterenol (1 µM) stimulated a 128-fold increase in cAMP within 6 min, which was further potentiated by a factor of four when cells were treated with 3-isobutyl-1-methylxanthine (0.2 µM) (37). This variety of responses seen under cultured cell conditions confirms that combined treatment with PDE inhibition and adenylate cyclase stimulation induces higher levels of cAMP than either treatment alone.

Nothing is known concerning the regulation of cAMP levels within microvascular endothelium of frog mesentery. Indeed, the endothelium, which forms the primary permeability barrier, has associated with it numerous pericytes, as well as nearby fibroblasts and nerves. Therefore, it is conceivable that part of the effect of the cAMP-enhancing agents is on a cell type other than endothelium. The present experiments confirm that both adenylate cyclase activation and PDE inhibition, alone or in combination, have significant effect on the morphology and transport across microvascular endothelium in vivo.

Does cAMP close preexisting gaps? In a wide variety of experimental models, both in vivo and with cultured endothelial cells, increased intracellular cAMP has been used to reduce the high permeability associated with inflammatory gaps (3, 4, 13, 14, 29, 30, 35, 36). Therefore, it was conceivable that our treatments to increase intracellular cAMP decreased baseline permeability by closing preexisting inflammatory type gaps in the perfused mesentery microvessels. This suggestion is extremely unlikely because such gaps should have been seen in control vessels in the present study and numerous previous studies. We can estimate how frequent such gaps should have been from the following. In a study of heat-induced inflammatory gaps in frog mesentery, baseline L_p increased from 4 to about 10 × 10⁷ cm/(s · cmH₂O), while surface area for exchange increased from about 0.0024 to 0.0043 cm²/cm² (8). This implies that for every 0.001 cm²/cm² exchange area increment, L_p increased by 3.2 × 10⁷ cm/(s · cmH₂O). In the present experiments used for electron microscopic analysis, treatment with rolipram and forskolin for 15-25 min stimulated a roughly 3 × 10⁷ cm/(s · cmH₂O) fall in L_p. With the assumption that rolipram-forskolin treatment closes gaps having a conductance equal to those described by Clough and colleagues (8), we calculate a loss of exchange area of 0.0009 cm²/cm². If we assume that an average inflammatory gap is circular with a diameter of 0.1 µm, then ~7,500 such gaps should be found on the wall of a 1-mm long vessel in mesentery (microvessels typically 20 µm diameter). Put another way, in a ribbon of 25 thin sections (each 40-nm thick) cut transversely to the axis of a microvessel, there should be at least seven such gaps. In short, the hypothesis that rolipram and forskolin treatment closes preexisting gaps leads to the conclusion that such gaps should be easily found in our typical examination of thin sections. We closely examined over 13 µm of vessel length in a previous study of endothelial clefts (1) and found no gaps in the endothelial barrier. According to these figures, in that study there should have been about 100 gaps. Similarly, we have examined and surveyed several micrometers in this and other studies and have found no gaps in control vessels. From this we conclude that rolipram and forskolin do not decrease L_p by closing preexisting gaps.

Does cAMP increase cell-to-cell overlap? Numerous studies of cultured endothelial cells indicate that effects of cAMP stimulation include rearrangement of actin into peripheral bands, an increase in the total amount of filamentous actin, suppression of motility, and decreased permeability of the monolayer (20, 21, 25, 38). These changes are also associated in subconfluent cultures with spreading of cells on the substrate. Stimulation of spreading in vivo could result in an increase in cell-to-cell overlap with an associated increase in the length of the paracellular pathway. An increased path length would increase hydraulic resistance and therefore would account in part for the fall in L_p that we observed. However, our measurements of cell-to-cell overlap showed no difference between the control group and the group of vessels treated with rolipram and forskolin, despite a fourfold difference in L_p. In fact, the treated group, with the lower L_p, had a slightly lower mean path length (nonsignificant). Therefore, in contrast to cell culture studies, in vivo the action of cAMP does not appear to be related to endothelial cell spreading.

Regulation of tight-junction structure. Our second important observation was that the reduction in L_p was associated with an increase in the number of junction strands between endothelial cells in true and venular capillaries. There are no previous observations of tight-junction modulation for vascular endothelium in vivo. Extensive investigations of the tight-junction structure of highly organized vascular fields of diaphragm revealed that pericytic venules have a low number of tight-junction strands (1.8), whereas vessels of lower permeability, capillaries (3.1), and arterioles (3.7) have higher numbers of strands in freeze fracture specimens (31). However, the latter study represented a general correlation with known gradients of permeability, whereas specific permeabilities were neither measured nor modified. In a study of primary cultures of bovine brain endothelial cells and capillary fragments modulation of the complexity of tight junctions was shown by treatment with cAMP (39). In those experiments endothelium in culture lost tight-junction complexity with time after isolation. Treatment with forskolin and the PDE inhibitor RO-20-1724 enabled the cells to maintain an initial degree of tight-junction complexity (number of branch points per strand length in freeze fracture). Loss of complexity was due primarily to decrease in strand branching; no change in strand number was reported. Perhaps more importantly, the treatment was also effective in maintaining the P-face association of tight-junction particles. In the absence of treatment the particles switched to the E-face, suggesting that cytoplasmic anchorage of tight-junction particles was compromised in culture and that increased cAMP reversed the trend.

Potential modulation of other structures. Other permeability barriers are potentially altered by the treatments used to increase intracellular cAMP. Additional hydraulic resistance would develop if there were an increase in the number of cleft-spanning structures (22). However, very little is known about the size or density of such structures. Much more modeling has been done regarding both the resistance to hydraulic and solute transport of the luminal endothelial glycocalyx (9). A moderate increase in the density of the glycocalyx, alone or in combination with an increase in glycocalyx depth, would also predict a lower L_p. For example, with the use of the simplest model of the glycocalyx as a fiber matrix with all of the control hydraulic resistance in the glycocalyx, a fourfold decrease in L_p would be predicted by either a fourfold increase in the thickness of the glycocalyx or a twofold increase in the fiber volume for a constant layer thickness. There has been no investigation to test the effect of cAMP on the structure of the glycocalyx.

Correlation of tight-junction structure to fluid flow. The vessels chosen for detailed transmission electron microscopy analysis exhibited an approximate fourfold reduction in L_p correlated with an increase in the mean number of junction strands from 1.7 to 2.2 per cleft. It is important to ask whether this moderate increase in mean strand number could cause the proportionally much larger increase in hydraulic resistance. Previous serial section and reconstruction of tight junctions in frog mesentery capillaries indicated that BSA-perfused vessels typically have only one nearly continuous tight-junction strand between endothelial cells (1). The additional 0.7 of strand per cleft was often seen as short branches from the main strand or as unattached segments. The dominant continuous strand was found to be interrupted at ~3-µm intervals by 150-nm gaps that provided a hydrophilic pathway from the lumen to the interstitium. Such an arrangement provided more than adequate hydraulic conductance to account for the baseline L_p values in these vessels as predicted by independent hydrodynamic modeling (11, 26). We have not determined the spatial distribution of the additional strand (0.5 per cleft section, or equivalently, 0.5 µm for every 1 µm of cleft length) that was found in the cAMP-enhanced vessels analyzed by transmission electron microscopy. However, a simple analysis enables us to speculate that a small increase in the amount of strand material could modify L_p in a highly nonlinear relationship. One such simple model could be that the frequency of the 150-nm gaps is reduced in the cAMP-enhanced vessels. With the assumption that all hydraulic flow passes through the 150-nm strand gaps, which are distributed an average of every 3 µm, by "filling in" three of four of the gaps the L_p would decrease to 25% of control comparable to the cAMP effect in the test vessels analyzed by transmission electron microscopy. The minimum additional strand material to close three-fourths of the openings would be only 450 nm for every 12 µm of cleft length or ~0.04 µm of strand length in every 1 µm of cleft, taking the mean predicted number of strands from 1.7 per cleft to only 1.74 per cleft. The excess of induced strand length (0.46 µm per 1 µm of cleft length) could be distributed as additional short branches and unattached segments.

A second speculative model of the structural effect of elevated cAMP is the induction of a second strand with a geometrical arrangement similar to the first. This condition could be induced by rearrangement of the strand branches and unattached segments found in the control state with the additional induced strand segments. The result would be two parallel strands with a small additional amount of strand material as short branches and unattached segments. Both strands would have the 150-nm openings distributed approximately every 3 µm to provide a continuous path from lumen to interstitium. With the breaks in the respective strands randomly positioned with respect to one another, some arrangements would increase the tortuousity of the fluid flow path within the plane of the cleft. The increase in tortuosity greatly increases the expected hydraulic resistance in two ways. First, the water flow is confined in a narrow channel for up to 1 µm or more (dependent on the regularity of gap spacing), which would increase the mean path length from lumen to interstitium. Second, the flow is forced to sharply change direction, further dissipating energy out of proportion to path length. Preliminary analysis indicates that the combined effects could be adequate to account approximately for the fourfold fall in L_p (personal communication, S. Weinbaum and X. Hu).

The foregoing hypothetical strand arrangements could possibly account for the changes in hydraulic flow due to agents known to increase intracellular cAMP. Further tests of structural models will be enabled by measurement of the effects on solute permeability and by detailed investigations of the continuity and arrangement of tight junction strands using serial reconstruction.