INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

IN RECENT YEARS, THERE HAS been considerable interest in the possibility that pulmonary capillary stress failure is an important contributing factor to the development of a number of forms of pulmonary edema, such as neurogenic and high-altitude pulmonary edema (15, 24, 25). Experimentally, extreme elevations in pulmonary vascular pressure have been shown to increase the capillary filtration coefficient (K_fc), decrease the osmotic reflection coefficient (_d), and produce discontinuities in the capillary endothelium and alveolar epithelium (5, 8, 13, 17, 21, 23, 26). It is not known, however, how the above physiological changes in barrier function relate to the observed ultrastructural derangements because K_fc, _d, and ultrastructure have not been correlated.

To address this, we examined whether a transient extreme increase in capillary transmural pressure (Ptm) in a canine isolated perfused left lower lung lobe (LLL) preparation would produce changes in K_fc and _d and alterations in the ultrastructure of the alveolar blood-gas barrier. We selected a transient period of hypertension for study because pulmonary vascular pressures may be elevated for only short periods in individuals with neurogenic pulmonary edema (15), and it has been suggested that such increases in pressure may leave the pulmonary vascular bed with an increased permeability (15). A capillary Ptm of 80 Torr was selected for study because transient exposure of the canine pulmonary vasculature to Ptms of this magnitude has been shown to leave the vasculature with an increased permeability as measured by a decreased _d (5, 13) and frequently an increased K_fc (21). We hypothesized that the evaluation of capillary ultrastructure in this setting would provide insight into the role played by the development of blood-gas barrier discontinuities in producing the observed changes in permeability coefficients. For example, the presence of endothelial and/or epithelial gaps would suggest an ultrastructural correlate for the changes in _d and K_fc, whereas the absence of an increased number of gaps would suggest that the permeability defect was of a much smaller magnitude (i.e., of a size that was beyond the resolution of electron microscopy to detect).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Isolated perfused LLL preparation. The isolated perfused LLL preparation has been previously described in detail (12). Briefly, 12 dogs [weight 21.8 ± 2.9 (SD) kg] of mixed breed and sex were anesthetized with pentobarbital sodium (30 mg/kg iv). The animals were heparinized (1,000 U/kg) and then exsanguinated through a large-bore femoral arterial catheter. The first 200 ml of blood were used to prime the perfusion system. An additional 200 ml were collected and stored for supplementation of the perfusion volume as needed. A left thoracotomy was performed; the LLL [46.6 ± 8.4 (SD) g] was removed; and the airway, LLL artery, and LLL vein were cannulated with plastic cannulas. The lobe was placed on a weighing basket that was suspended from a force transducer (Grass Instruments, Quincy, MA) inside a heated (37°C) and humidified Plexiglas chamber. The lobe was perfused through the arterial cannula with blood pumped (Masterflex Pump, Cole-Parmer Instruments, Chicago, IL) from a venous reservoir. A heat exchanger (to maintain the blood temperature at 37°C) and a bubble trap were incorporated into the perfusion system upstream of the LLL. The venous drainage emptied into the reservoir. The average blood flow was 453 ± 49 (SD) ml/min. Pulmonary vascular pressures were measured using Gould pressure transducers and referenced to the top of the LLL. Baseline LLL arterial (Pa) and venous (Pv) pressures averaged 10.6 ± 1.2 and 2.1 ± 0.4 Torr, respectively. The latter pressure was set by adjusting the height of the reservoir. The LLL was ventilated with a humidified gas mixture (dry gas pressures: 15.5% O₂, 5.5% CO₂, and 79.0% N₂). The average tidal volume was 180 ± 35 ml, and the respiratory rate was 9 breaths/min. Peak inspiratory pressure was 7.9 ± 2.2 Torr, and end-expiratory pressure was set at an average 0.8 ± 0.3 Torr by means of a water-overflow system. Blood gases under control conditions were PO₂ of 109.7 ± 21.0 Torr, PCO₂ of 39.4 ± 2.1 Torr, and pH of 7.38 (range 7.36-7.41).

Measurement of _d and K_fc. We used the hematocrit-protein double-indicator technique to measure _d (14, 16, 20, 27, 28). With this approach, the relative increases in hematocrit and plasma protein concentration resulting from transvascular fluid flow are used to calculate the solvent drag reflection coefficient (_f) as follows (27)
where C₁ and C₂ are the initial and final plasma protein concentrations, and H₁ and H₂ are the initial and final hematocrits. At high rates of filtration (i.e., at rates where solute flux occurs predominantly via convection), the value of _f approaches that of _d (9). Hematocrit was measured by the microhematocrit technique, and plasma protein concentration was determined by refractometry (American Optical, Buffalo, NY). Plasma protein concentrations and hematocrits were corrected for perfusion pump-induced hemolysis using the increase in plasma hemoglobin concentration as an index of hemolysis as previously described (16). Plasma hemoglobin concentration was determined using the method of Blackney and Dinwoodie (4).

Experimental protocol. Two groups of LLLs were studied, a high-pressure group (n = 6) in which Pv was transiently raised to 85.2 ± 10.2 (SD) Torr by tightening a screw clamp around the venous tubing and then returned to baseline values and a control group (n = 6) in which Pv was not transiently elevated (Fig. 1). Elevation of Pv in the high-pressure group caused Pa and Pc to rise to 87.3 ± 9.6 and 86.3 ± 9.8 Torr, respectively. The peak Ptm (calculated using the mean airway pressure as the pressure surrounding the capillaries) was 80.3 ± 8.5 Torr. The total duration of Pv elevation (from the earliest increase in pressure to the time when the clamp was completely reopened) was 3.8 ± 0.4 min. After this time, Pv was set at an average 21.9 ± 4.0 Torr in both groups of LLLs by elevating the reservoir. This caused Pa and Pc to increase to 27.9 ± 4.4 and 24.9 ± 4.1 Torr, respectively. Ptm (for both groups) averaged 19.1 ± 4.7 Torr under these conditions. [Inasmuch as Pv was not transiently increased in the control group, the highest Ptm attained in this group was that (20.7 ± 6.1 Torr) which occurred during the measurement of K_fc and _d.] K_fc was determined from recordings of weight gain as described in Measurement of _d and K_fc. Blood samples were drawn for analysis of hematocrit and plasma protein and hemoglobin concentration immediately before the Pv was raised and after the filtered fraction had reached ~10%. This required an average 29 min for the high-pressure group and 70 min for the control group.

View larger version (14K): [in this window] [in a new window]   Fig. 1.   Experimental protocol for isolated perfused left lower lung lobe experiments. See text for description. Pa, left lower lung lobe arterial pressure; Pv, left lower lung lobe venous pressure; Kfc, capillary filtration coefficient; d, osmotic reflection coefficient.

Tissue perfusion fixation and sampling. After measurement of K_fc and _d, ventilation was stopped, and the lobes were inflated to a constant pressure (4.4 ± 0.5 Torr) for the entire fixation procedure. They were perfused with saline solution [11.06 g NaCl/l (350 mosM) and 10 U heparin/ml] until the outflow appeared clear of red blood cells (3-5 min) and then fixative (phosphate-buffered 2.5% glutaraldehyde, total osmolarity 500 mosM; pH adjusted to 7.4) for 10 min. For each lung, both saline and fixative perfusion were carried out at the same vascular pressure at which the K_fc and _d measurements had been made, i.e., a mean Pc of 24.9 ± 4.1 Torr for both groups. Although Pc was identical during the periods in which the permeability coefficients were measured and the lungs fixed, the capillary Ptms differed slightly during these periods because of the different ventilatory regimens. During fixation, Ptm was constant at an average 20.5 ± 4.2 Torr, whereas when the lobes were ventilated (during the K_fc and _d determinations), it oscillated slightly around a mean value of 19.1 Torr. All pressures were continuously measured during the process of fixation by transducers attached to the cannulas.

View this table: [in this window] [in a new window]   Table 1.   Morphometric measurements of endothelial and epithelial breaks per millimeter and average break length in each sample from isolated perfused LLL preparations and d and Kfc measurements from whole lobes

Short-term reversibility of ultrastructural changes. To determine whether structural changes in the canine blood-gas barrier at high vascular pressure are rapidly reversible, we perfused the lungs of three additional dogs [22.3 ± 1.1 (SD) kg] with blood at high pressure (Ptm of 68 Torr), followed by saline and glutaraldehyde perfusions both at low pressure (Ptm of 17 Torr). The in situ perfusion lung preparation in these dogs (see experiments 13-15, Tables 4 and 5) was identical to that used in our laboratory's previous study of stress failure in dog lung (17). The lungs were first inflated to a positive pressure of 18 Torr and then deflated to 4-Torr transpulmonary pressure, which was maintained during the entire procedure. The perfusion circuit was primed with blood, and the lungs were perfused for 1 min with blood at 68-Torr Ptm, followed by saline-dextran (11.06 g NaCl/l, 350 mosM; 3% T-70 dextran; 1,000 U heparin/100 ml) at a Ptm of 17 Torr for 5 min. Then, perfusion with fixative (phosphate-buffered 2.5% glutaraldehyde with 3% T-70 dextran; total osmolarity 500 mosM; pH adjusted to 7.4) followed for 10 min at the same Ptm of 17 Torr. Sampling of the lower left lobe was identical to that described in Mathieu-Costello et al. (17), i.e., one slab, ~0.5 cm thick, was taken perpendicular to the cranial-caudal axis at about one-half the distance from the bottom of the lobe. A thin vertical slice was obtained from each slab by following the same procedure as described in Tissue perfusion fixation and sampling and completely cut into smaller blocks. The ultrastructural appearance of these lung lobes, in which the lungs were fixed at low pressure after being transiently exposed to high Ptm, was compared with that observed in our previous study in which the lungs were fixed at either baseline or elevated Ptm (17).

Tissue preparation and morphometry. All procedures were identical to those used in our laboratory's previous studies (8, 17, 23). Briefly, the samples were rinsed overnight in 0.1 M phosphate buffer adjusted to 350 mosM with NaCl, pH 7.4, and postfixed for 2 h in 1% osmium tetroxide in 0.125 M sodium cacodylate buffer adjusted to 350 mosM with NaCl (total osmolarity: 400 mosM, pH 7.4). They were then dehydrated in increasing concentrations (70-100%) of ethanol, rinsed in propylene oxide, and embedded in Araldite. Two blocks were selected randomly from each sample, and 1-µm-thick sections cut from each block using an LKB Ultratome III were stained with a 0.1% aqueous solution of toluidine blue and examined by light microscopy. Ultrathin sections (50-70 nm) were contrasted with uranyl acetate and bismuth subnitrate (22), and micrographs for morphometry were taken on 70-mm films with a Zeiss 10 electron microscope, with micrographs of a carbon grating replica (E. F. Fullam, Schenectady, NY) recorded for calibration on each film.

Statistics. The SE of the estimates of the number of breaks per unit endothelial and epithelial boundary length was calculated by applying formulas for the SE of ratio (6). The variability between micrographs at the sampling location studied was obtained by pooling the data of the micrographs from two blocks (total 56-61 micrographs per sample). The SE of the estimates of break length and blood-gas barrier thickness indicates the variability between individual measurements. Group means were compared with analysis of variance and Bonferroni post hoc test to discern differences between individual groups. Paired comparisons were made by Student's t-test for paired samples, or, in cases where the assumptions for parametric testing were violated, the Wilcoxon signed-rank test was used. Differences were taken as significant for P < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Permeability coefficients. The results of the K_fc and _d determinations are shown in Table 1 and Fig. 2. K_fc in the high-pressure group (0.249 ± 0.042 g · min¹ · Torr¹ · 100 g¹) was 361% higher (P < 0.01) than that observed in the control group (0.054 ± 0.009 g · min¹ · Torr¹ · 100 g¹), and _d (high-pressure group, 0.52 ± 0.07; control group, 0.85 ± 0.08) was significantly reduced (P < 0.01).

View larger version (10K): [in this window] [in a new window]   Fig. 2.   Peak pulmonary capillary transmural pressure (A) and permeability coefficients [osmotic reflection coefficient (B) and filtration coefficient (C)] in control and high-pressure groups. Values are means ± SE.*P < 0.01 compared with control group.

Light microscopy. Fluid was apparent in the airway cannula of one lung in the high-pressure group (animal 8, Tables 1 and 2) at the end of the experiment. Figures 3 and 4 show light micrographs of sections of lung parenchyma in control and high-pressure groups. Macroscopically clear areas (see Tissue perfusion fixation and sampling in METHODS) showed well-perfused and moderately distended capillaries (Fig. 3, A and C), whereas capillaries with densely packed red blood cells were found in macroscopically dark areas (Fig. 3, B and D) in both groups. In addition, alveolar edema was found in the high-pressure group, both in macroscopically clear (Fig. 4A) and dark areas (not shown) and in central (Fig. 4B) or peripheral portions of the lung (Fig. 4C).

View larger version (104K): [in this window] [in a new window]   Fig. 3.   Light micrographs of portions of parenchyma in macroscopically clear (A, C) and macroscopically dark areas (B, D) in control (A, B) and high-pressure group (C, D). Note well perfused capillaries in macroscopically clear areas (A, C) and densely packed red blood cells in capillaries in macroscopically dark areas (B, D) in both groups. All micrographs are at the same magnification, and, for each group, micrographs of macroscopically clear and dark areas are from the same lungs.

View larger version (117K): [in this window] [in a new window]   Fig. 4.   Light micrographs of portions of parenchyma in macroscopically clear (A), central (B), and peripheral regions (C) of the lung in high-pressure group, showing alveolar edema in each region. Micrographs of central and peripheral regions are from the same lung.

Electron microscopy. Figure 5A shows the normal appearance of the blood-gas barrier in control lung, and Fig. 5, B and C, shows interstitial edema, red blood cells in the interstitium, and granular material in the alveolar space in a representative lung from the high-pressure group. Figure 6 shows examples of one of the very few endothelial (A) and epithelial (B) disruptions seen in lobes exposed to high pressure. Morphometric evaluation of the capillary endothelium and epithelium revealed the presence of few endothelial and epithelial breaks in some lobes in either group (Table 1). Break frequencies of 0.5, 0.9, 1.6, and 2.1 in those samples correspond to the finding of 1, 2, 3, and 5 disruptions (endothelial or epithelial; Table 1), respectively, out of 56-60 micrographs examined in each site from each animal. There was no difference in the incidence of either endothelial or epithelial breaks in macroscopically clear or dark areas of the lobe or peripheral and central regions of the lung where these comparisons were made (Table 1). When breaks were observed, they had the appearance of those previously described (Fig. 6, A and B). Out of the 12 (endothelial) and 3 (epithelial) disruptions found out of all samples, all had a continuous basement membrane, except for one endothelial break in the high-pressure group.

View larger version (108K): [in this window] [in a new window]   Fig. 5.   Electron micrographs showing normal appearance of the blood-gas barrier in control group (A) and interstitial edema (i; B and C), red blood cells in interstitium (*), and granular material in alveolar space (B) in high-pressure group. c, Capillary; a, alveolar space.

View larger version (81K): [in this window] [in a new window]   Fig. 6.   Electron micrographs of one of the very few endothelial (A; arrow) and epithelial disruptions (B; arrow) among all lobes exposed to high pressure. Note red blood cells in interstitium (*) or protruding into endothelial opening (; A).

Short-term reversibility of ultrastructure changes. After 1 min of perfusion at 68-Torr Ptm, followed by a reduction of Ptm to 17 Torr during the 5-min saline and 10-min glutaraldehyde-fixative perfusions, the number of endothelial and epithelial breaks were 2.8 ± 0.4 and 4.2 ± 0.6/mm, respectively (Table 4). These values were intermediate though not significantly different from the number of breaks previously found after continuous high Ptm (endothelium, 4.5 ± 2.3/mm; epithelium, 7.6 ± 4.7/mm) and low Ptm (endothelium, 0.4 ± 0.3/mm; epithelium, 1.8 ± 1.4/mm). This suggests that some of the breaks that had occurred during the 1-min perfusion with blood at 68 Torr rapidly reversed when the pressure was reduced to 17 Torr for the saline and fixative perfusions. Practically all the disruptions had a broken basement membrane (endothelium 100%; epithelium, 96 ± 4%). This contrasts with the results after continuous high or low pressure, where a continuous basement membrane was found along the majority of the breaks (endothelium, 70-80%; epithelium 70-90%), with no difference between pressures (17). It suggests that the breaks that closed were those associated with a continuous membrane. Because of the great intragroup variability in the occurrence of stress failure (17), there was no difference in break length or frequency between the three groups. However, the larger average break lengths in the present experiment (Table 4) suggest that breaks that closed were the shorter breaks.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The transient elevation of pulmonary capillary Ptm to an average 80.3 Torr in the isolated perfused LLL produced significant decreases in _d and increases in K_fc measured at a Ptm of 19.1 Torr. The reductions in _d were similar to those that our laboratory previously observed in dog lungs after a transient period of pulmonary hypertension (5, 13) and indicate that vascular protein permeability had increased. The increases in K_fc were also similar to those previously observed in dog lungs after vascular pressure elevation (21). Because K_fc was determined at the same elevated vascular pressure in both the control and high-pressure groups, the contribution of an increased surface area available for transvascular fluid flow to the estimate of K_fc should be the same for each group. Therefore, the significant increase (361%) in K_fc observed in the high-pressure group should reflect primarily an increase in hydraulic conductivity. The changes in _d and K_fc, plus the observation that the time required to achieve a filtration fraction of ~10% when Pv was elevated to determine _d was less in the high-pressure (29 min) compared with the control lobes (70 min), thus indicate that the transient period of pulmonary hypertension increased both the water and protein permeability of the lobar vasculature.

Given the above, we asked whether the increases in water and protein permeability were associated with changes in capillary or alveolar ultrastructure. We found that the reductions in _d and increases in K_fc were not associated with microscopical evidence of endothelial and epithelial layer discontinuities. Although endothelial and epithelial breaks were observed in some experiments, their incidence was small and did not increase in the high-pressure group. These data thus suggest that the increased transvascular water and protein flux observed in these lobes occurred through pathways of a size not resolvable at the level of the electron microscope. This conclusion is consistent with the results of our previous pore analysis (5) in which we found that the reduction in _d after a similar elevated Pv transient in an in situ canine lung lobe preparation was highly correlated with the fractional flow through a large-pore system having an average molecular radius of 354 Å.

One interpretation of these data is that the increased transvascular water and protein flux occurred through pathways unrelated to those that may be formed when endothelial gaps develop. Although this is certainly possible, these results do not rule out the possibility that the formation of endothelial gaps is an event required to increase vascular water and protein permeability when the pulmonary vasculature is exposed to high pressure. In this regard, in both the rabbit (8) and dog (Tables 4 and 5), fewer breaks were observed in lungs fixed for ultrastructural analysis after the pressure had been returned to normal after initially being elevated, compared with lungs fixed at elevated pressure. This difference suggests that many of the gaps may close once pressure is reduced. As in the rabbit (8), once the pressure initially set at a level known to cause stress failure of pulmonary capillaries in the dog lung (68 Torr) was reduced in dogs 13-15, 1) the number of endothelial and epithelial breaks decreased, 2) the breaks that closed were those initially small and associated with an intact basement membrane, and 3) there was a significant reduction in the widening of the interstitium caused by edema (Tables 4 and 5). These results indicate that the pressure-induced ultrastructural changes in the wall of pulmonary capillaries caused by stress failure in the dog are rapidly reversible. In experiments 7-12, Ptm was also raised to a level (80.3 ± 8.5 Torr) previously shown to induce gap formation (17). It is thus conceivable that after Pc was reduced, gaps that had developed during the period of elevated pressure could have closed to the point that the endothelial lining appeared to be continuous but leaving small pathways for water and protein flux that are not resolvable by the electron microscope after vascular perfusion-fixation of the unventilated lung at a capillary Ptm of 20.5 Torr.

Another potential explanation of our data is that the changes in barrier coefficients represented events occurring in extra-alveolar vessels (EAVs). In this regard, the results of a number of studies that have evaluated the filtration characteristics of pulmonary EAVs have indicated that a significant fraction of the extravasation of fluid in the lung may occur from vessels both up- and downstream of the capillaries (1, 2) and that in some forms of lung injury, EAV permeability may be increased (11). Inasmuch as we did not observe an increased incidence of gap formation in the capillaries [vessels that should be the most prone to stress failure due to their high wall stresses (26)] of the high-pressure group, it would not appear likely that stress failure occurred in the EAVs in this group of lobes unless EAVs exhibit a greater permeability response to a given level of mechanical stress than do the capillaries. Although an answer to the latter question is unknown, control of pulmonary endothelial cell barrier function has been shown to differ in conduit and microvascular endothelial cells. For example, cultured rat pulmonary arterial endothelial cells have been found to exhibit gaps and increased permeability when exposed to increases in internal calcium concentration, whereas rat pulmonary microvascular endothelial cells do not (10). Although we did not conduct an ultrastructural evaluation of EAVs in this study, we examined sections of the same samples at the light-microscopic level and did not note any greater tendency to observe extravasated erythrocytes in the vicinity of EAVs in the lobes of the high-pressure compared with the control group. Although these observations do not rule out the possibility that EAV permeability might have become increased to the point that water and protein more readily escaped the vasculature, we observed no evidence for stress failure leading to gap formation in these vessels.

A potential methodological concern relates to the possibility that the ultrastructural sampling did not detect existing endothelial and epithelial disruptions in the tissues. However, a similar random sampling of one tissue slab across lung lobes has previously allowed detection of the incidence of disruptions of the blood-gas barrier with increased Pc (17, 23). In those studies, a large variability in break frequencies was found between animals, which was not reduced by either increasing the number of tissue blocks sampled at a given site or sampling another lobe of the same lungs (17). Statistical analysis revealed that the source of variability in the incidence of stress failure was at the level of micrographs, i.e., individual capillaries, rather than between tissue samples (17). In this study, very few disruptions were found in all micrographs examined from all lungs, both in macroscopically dark or clear areas and in central or peripheral regions of the lobes in the high- and low-pressure groups, suggesting that the paucity of breaks was not a sampling artifact.

In previous stress-failure studies, in which the lungs were fixed shortly after the onset of blood perfusion, the lungs did not have a patchy appearance consisting of macroscopically clear and dark areas (8, 17, 23). The reason why there were areas of the lungs (i.e., dark areas) that could not be cleared of red blood cells in this study is not clear but probably relates to the extended perfusion time before fixation. It is conceivable that the isolated lung lobes (lacking a normal lymphatic drainage) developed discrete areas of interstitial edema, which might have produced an increase in capillary resistance impeding the washout of red blood cells during the fixation procedure. Suggestive of this possibility was the observation that the total thickness of the blood-gas barrier was significantly greater in samples taken from areas of the lung that did not clear after saline perfusion compared with those that did. Because there was also a tendency for the interstitial thickness to be larger in the dark areas, it is likely that the observed increases in total thickness resulted from edema.

The above observations thus suggest that there were regional differences in the degree of edema development in lobes that exhibited both macroscopically clear and dark areas. This possibility is consistent with observations of Bachofen et al. (3), who found that the edema was inhomogeneously distributed in rabbit lungs that had been exposed to a transient elevation in Pc. These investigators also observed striking differences in the density of proteinacious material of the edema fluid, which presumably resulted from differences in albumin concentration. Taken together, the results from both studies suggest that high vascular pressures may disrupt vascular integrity in a complicated inhomogenous fashion and suggest that the estimates of _d and K_fc represent weighted mean values that reflect the relative proportions of relatively low- and high-permeability vascular domains. In this regard, the observed changes in _d and K_fc could reflect either a small change in permeability of a relatively large surface area or the contribution of a relatively smaller number of high-permeability areas.

As indicated above, some proportion of the endothelial and epithelial breaks may close spontaneously after the elevated vascular pressure is reduced. Recent observations by Parker and Ivey (19) suggest that the maintenance of normal permeability during exposure to elevated vascular pressure or recovery after normal pressures are reattained might be subject to active control. In this regard, these investigators observed that a portion of the increase in K_fc that occurred in isolated perfused rat lungs exposed to elevated venous pressure could be prevented by the administration of isoproterenol, a -adrenergic agonist that appears to reduce conductance of cell junctional pathways. These investigators suggested that isoproterenol might also induce closure of transcellular breaks as well (19).

In conclusion, transient elevation of Ptm to an average 80.3 Torr in canine isolated perfused lung lobes with subsequent vascular permeability evaluation at an average Ptm of 19.1 Torr showed that vascular permeability had been increased by the initial pressure elevation but revealed no evidence of an increased incidence of either endothelial or epithelial discontinuities that could serve as a physical basis for the observed physiological changes. These observations suggest that the increased transvascular water and protein flux observed in the ventilated lung occurred through pathways not resolvable by electron microscopy after vascular perfusion-fixation at a transmural pressure of 20.5 Torr. These sites could either be areas that are independent of the development of discontinuities or could represent gaps that had not completely closed once the initial pressure elevation was reduced.