Pulmonary ischemia/reperfusion injury: A quantitative study of structure and function in isolated heart-lungs of the rat

Rat Isolated Heart-Lung Model

The rat isolated-heart lung model has been described in detail elsewhere (Fukuse et al., 1995). Briefly, the heart-lung block was suspended in a warmed, humidified chamber at an angle of 45° in order to mimic the in situ position of the lungs relative to gravitation and to achieve homogeneous perfusion and ventilation. The temperature in the chamber was kept at 37°C by means of waterjacketed vessels connected to a warming pump (water thermostat type VTS 13c; Radiometer Ltd., Copenhagen, Denmark). The perfusate used was a Krebs-Henseleit buffer containing washed bovine red blood cells (KHrb) at a hematocrit of 38–40%, which has been shown to improve functional characteristics of isolated lungs, and to minimize perfusion dependent pulmonary edema in comparison with a cell-free KH-buffer (Fukuse et al., 1995). Leukocytes were removed by means of a leukocyte removal filter (RC100E, Pall Europe Ltd., Portsmouth, UK). Absence of leukocyte sequestration was confirmed histologically. The perfusate reaching the coronary arteries via the aorta was oxygenated with a mixture of 95% O₂ and 5% CO₂. The perfusate of the pulmonary circuit reached the lung via the right atrium after passage of a second oxygenator (Monolyth integrated membrane lung, Sorin Biomedical Ltd., Saluggia, Italy) gassed with 95% N₂ and 5% O₂ to obtain a deoxygenated pulmonary arterial inflow. From a preload pool, perfusion of the right atrium was performed at a hydrostatic pressure of 5 cm H₂O, and the right ventricle of the working heart provided pulsatile flow of the perfusate through the lung. The opening of the left atrial cannula was positioned to ensure a constant left atrial pressure (P_la) of 2 cm H₂O. The lungs were ventilated with room air at a tidal volume of 4 ml and a rate of 40 breaths per min. A positive end-expiratory pressure of 3 cm H₂O was maintained.

Measurement of Hemodynamic and Respiratory Parameters

Pulmonary artery pressure (P_pa) was measured via a pressure transducer and monitor (Servomed, Hellige Ltd., Hamburg, Germany). P_la was constant at 2 cm H₂O (see above). Cardiac output (CO) was determined by collecting pulmonary venous outflow. P_pa [mm Hg], P_la [cm H₂O], and CO [ml× min⁻¹] were used to calculate pulmonary vascular resistance (PVR) according to the standard equation PVR = (Ppa - P_la) × 80 / CO [dynes × sec × cm^-5] (Fukuse et al., 1995). Blood gases were determined at intervals of 10 min. P_O2 and P_CO2 of the deoxygenated perfusate of the preload pool were defined as arterial Pa_O2 and Pa_CO2. P_O2 and P_CO2 of the perfusate collected from the left atrium as venous Pv_O2 and Pv_CO2. To assess the lungs' capability for gas exchange, ΔP_O2 and ΔP_CO2 were defined as Pv_O2 − Pa_O2 and Pv_CO2 − Pa_CO2, respectively. Peak inspiratory pressure (PIP) was measured by means of the small animal respirator (Mod. No. 4601, Rhema Corp., Hofheim, Germany) and registered every 10 min. After 40 min of reperfusion, left and right lungs were separated for determination of the W/D-ratio and for fixation and subsequent histological and ultrastructural analysis, respectively.

Fixation

In all the control and experimental animals, fixation of the left lung was performed by vascular perfusion via the pulmonary artery at a hydrostatic pressure of 15 cm H₂O for only about 2 min in order to minimize potential fixation dependent fluid shifts. The quality of structural preservation did not differ from rat lungs, which were fixed by vascular perfusion for 10 min in the course of another study (Uhlig et al., 1995). During fixation, the airway pressure was adjusted to 12 cm H₂O. The fixative used was a mixture of 1.5% glutaraldehyde and 1.5% paraformaldehyde in 0.1 M cacodylate buffer at pH 7.35. The osmolality of the cacodylate buffer, also used in any further step, was adjusted to isotonicity, i.e., to 300 mOsm/kg. This resulted in a total osmolality of the primary aldehyde-containing fixative of about 760 mOsm/kg. At the end of perfusion fixation, main left bronchus and pulmonary artery were tightly clamped to prevent any leakage of air or edema fluid and to allow stiffening of the lungs during storage in cold fixative for 18–24 hours prior to sampling. Thus, collapse of the lung due to disconnection of the trachea from the spirometer and deformation of tissue blocks during sampling, the only disadvantages of this type of fixation method (Bachofen et al., 1982: method II), could be avoided.

Lung Volume Determination

After cold storage in fixative, the volume of the fixed lung (V, L) was determined by fluid displacement in an isotonic NaCl solution containing 10% dextran (osmolality: 310 mOsm/kg; specific weight: 1.045g/ml). Volume determination was performed immediately before sampling of the tissue blocks as recommended (Michel and Cruz-Orive, 1988). Because storage of up to 5 days in the fixative did not result in a significant alteration of fixed lung volume (A. Fehrenbach, Göttingen, unpublished observations), we can exclude that differences observed in lung volumes were due to the duration of storage in the aldehyde fixative.

Sampling

Systematic random sampling was performed according to Michel and Cruz-Orive (1988). Briefly, the fixed left lung was embedded in 2% agar-agar dissolved in distilled water. By means of a tissue slicer, the organ was cut from apex to basis into 3 mm thick slices, which were laid down apical face up. Then, the collection of slices was covered with a transparent sheet with 11×11 10-mm spaced points. Tissue blocks of approximately 3×3×3 mm³ in size were only taken from those sites where a test point, which determined the lower left edge of any block, fell on the section area of a lung slice. Thus, 5–11 blocks were obtained from each lung for LM and TEM. From the remaining lung slices, samples were collected for SEM. From the three control and three experimental lungs fixed for subsequent lectin histochemistry, every third lung slice was collected in toto for embedding in paraffin.

Tissue Processing

Samples to be studied by SEM were processed according to a modified OTOTO (osmium tetroxide − thiocarbohydrazide − osmium tetroxide) method described earlier (Uhlig et al., 1995). Briefly, tissue blocks were rinsed in 0.1 M cacodylate buffer and postfixed for 2 hr with cacodylate buffered 1% osmium tetroxide. Then, they were rinsed in distilled water, transferred to 1% aqueous thiocarbohydrazide for 1 hr, washed with distilled water, treated with 1% aqueous osmium tetroxide for 2 hr, and rinsed again with distilled water. These steps were repeated twice. After dehydration through a graded series of ethanols, critical point drying was performed and the specimens could then directly be observed by means of a Zeiss DSM 960 without additional metal-sputtering.

For LM and TEM, the tissue samples were processed as described earlier (Fehrenbach et al., 1995). Processing was standardized using an automated tissue processor with temperature control and cooling device (Histomat, Bio-med, Theres, Germany). Briefly, after 6 rinses (5 min each) in 0.1 M cacodylate buffer (300 mOsm/kg), samples were post-fixed in 1% OsO₄ in 0.1 M cacodylate buffer (300 mOsm/kg), washed again in the same buffer (4 rinses, 5 min each), rinsed in distilled water (2 rinses, 5 min each), transferred to half-saturated, aqueous uranyl acetate for en bloc-staining (overnight). All steps were performed at 8°C. After washing in distilled water, specimens were dehydrated through a graded series of ethanol (30%, 50%, 70%, 90%, 3 × 100%), transferred to Araldite via propylene oxide and a 1:1-mixture of propylene oxide/Araldite, embedded in Araldite, and polymerized at 60°C for 3 days. For histological studies, 5–10 semi-thin sections of three systematic random samples per lung were stained with Methylen blue and analyzed by means of LM (Laborlux 11; Leitz, Wetzlar, Germany). For ultrastructural studies, two or three tissue blocks per lung were selected from the sample collection by random numbers. From each block five ultrathin sections were cut, collected on Formvar-coated 1x2 mm slot grids, and counter stained with lead citrate by means of an Ultrostainer (Leica, Hamburg, Germany). From these five ultrathin sections the technically best one was chosen for subsequent stereological analysis. Qualitative TEM studies and systematic quadrate sub-sampling of micrographs for stereological analysis were performed by means of an EM 10 A (Zeiss, Oberkochen, Germany).

For lectin- and immunohistochemistry, lung slices were processed by routine paraffin embedding. The biotinylated form of the type I pneumocyte-specific marker BPL (Vector Laboratories, Burlingham, USA) and the mouse monoclonal epithelial cell-specific cytokeratin-antibody MNF116 (Dako, Hamburg, Germany) were demonstrated as described by Kasper and co-workers (1994). Briefly, deparaffinized sections were preincubated with 5% normal horse serum, blocked for endogenous peroxidase with hydrogen peroxide, incubated with biotinylated BPL (1:400 dilution) for 30 min followed by avidin-biotin-peroxidase complex (ABC) (Vector Laboratories). For immunohistochemistry, dewaxed sections were treated with 0.05% Pronase (15 min, 37°C) prior to incubation for 60 min at 37°C with MNF116 (1:1,500). An anti-mouse ABC-kit was used as detection system (Vector Laboratories). Sections were developed with 0.05% 3,3-diaminobenzidine tetrahydrochloride (Serva, Germany) and 0.03% hydrogen peroxide, and counterstained with haematoxylin. Phosphate buffered saline (PBS) was used as incubation and rinsing buffer. A control section, which was incubated with PBS only, was processed together with each lectin- or antibody-incubated section. The specificity of the lectin was tested by preincubation with the specific sugar residue N-acetyl galactosamine (EY Laboratories, San Mateo, USA) (Kasper et al., 1994).

Stereology

Stereological analysis was performed according to a cascade procedure to allow for the determination of the absolute volumes of the various compartments according to standard procedures (Weibel, 1990). Volume densities (Vv) of the structures of interest were recorded at four different levels (Table 1). All parameters were determined by means of point and intersection counting. Test fields and micrographs were collected according to systematic quadrate sub-sampling (Michel and Cruz-Orive, 1988).

Three blocks of tissue per lung were used for stereological analyses at levels 1–3. One technically good 0.5μm thick section per sample was examined according to the procedure of systematic quadrate sub-sampling (Michel and Cruz-Orive, 1988). Using a Leitz Laborlux 11 light microscope, the whole section was scanned by moving the object stage at uniform intervals of 0.5 mm along the axes. At each position, point counting was performed directly by means of an eye-piece containing an integration plate (100/25 points).

Two or three tissue blocks of each lung were used for stereological analysis by means of TEM (level 4). One technically good ultra-thin section per tissue block was examined according to the procedure of systematic quadrate sub-sampling (Michel and Cruz-Orive, 1988). The whole section was scanned by moving the object stage at a uniform distance determined by the stage coordinates. Whenever the center of the fluorescence screen of the microscope hit the alveolar septum, a micrograph was recorded at a primary magnification of 3,150 x. The final magnification after photographic reproduction was determined by means of a calibration grid to be 6,500 x. Calibration was performed during each photographic session. A multipurpose coherent test system was used for point and intersection counting (324 points, test line length d = 10 mm).

The following structural parameters were analyzed (Table 1). At the first level (LM, 100 x), a distinction was made between parenchyma and nonparenchyma. In the mean (± SD), 70.6 ± 23.3 fields of view per lung were analyzed, which resulted in a total count of 1,764 ± 583 points per lung with 224 ± 155 points falling on non-parenchyma. At the second level (LM, 160 x), non-parenchyma was studied to distinguish the lumen and the wall of airways (tunica mucosa and muscularis) and vessels (tunica intima and media), respectively, as well as their joint connective tissue sheet, the peribronchovascular space. A distinction was made between the peribronchovascular space occupied by extravasated erythrocytes and the fraction that was free of erythrocytes. In the mean, 54.2 fields of view per lung were analyzed, which resulted in a total count of 1,354 ± 928 points falling on non-parenchymal structures counted per lung. At the third level (LM, 1000 x), parenchyma was studied to distinguish the alveolar space including alveolar ducts (filled with air or with edema-fluid) and the alveolar septum (tissue and capillaries). In the mean, 45.5 ± 14.5 fields of view were analyzed per lung, which resulted in a total count of 1,129 ± 359 points per lung falling on parenchyma, with a mean of 262 ± 77 points falling on alveolar septum.

The collection of micrographs obtained by systematic quadrate sub-sampling was used to determine the arithmetic mean thickness of the total air-blood-barrier as well as of its components, epithelium, septal interstitium, and capillary endothelium at the fourth level (TEM, 6,500 x). In the mean, 20.0 ± 2.9 micrographs per lung were collected and analyzed. Using the multipurpose test point system described above, the mean arithmetic thicknesses (τ_b), which can be defined as the ratio of tissue volume per barrier surface, were determined by counting the points falling on the respective barrier (P_b) and the intersections (I_b) with its two bounding surfaces according to Weibel, 1990, page 224 (formula 38): τ_b = 2 × k₁ × d × P_b/I_b with constant k₁ = 0.5 and test line length d = 1.54 μm, characteristics of our test system at the given final magnification. In addition, the degree of cell damage to alveolar type I pneumocytes was quantitatively determined as the fraction of normal, swollen, and fragmented areas (covered with or free of edema-fluid, respectively) relative to the total surface of alveolar type I pneumocytes, i.e., the ratio of intersections with injured areas per total number of intersections counted (Velazquez et al., 1991). A mean of 249 ± 74 intersections was counted per lung. The different types of injury to the alveolar type I epithelium were defined as follows: normal, only if there were absolutely no alterations in fine structure; swollen, if there were small vacuoles or a clearing of the cytoplasmic ground substance with the cell membranes being continuous; fragmented, if the apical cell membrane was disrupted, the epithelial leaflet vesiculated, or the basal lamina completely free of any epithelial remnants.

Since ischemia/reperfusion-induced lung injury and edema formation involved several compartments of the lung at the same time, a stereological analysis based on the determination of volume densities alone could run into the “reference trap” (Braendgaard and Gundersen, 1986). For direct comparisons of experimental and control groups, we therefore followed Mayhew (1991), who pointed out that “emphasis is given to absolute quantities because these avoid dangers prevalent when comparing groups using relative data.” The absolute volumes were used to define compartments that actually responded by volume changes. The absolute volumes of the main pulmonary components were estimated by multiplication of the volume densities obtained at each level with the absolute volume of the left lung (V, L) determined by fluid displacement. Total interstitial edema volume was defined as the sum of the absolute volumes of alveolar septal interstitium (V int) and peribronchovascular space (V ps) minus the respective control mean volumes. The analysis of structure-function correlation was based on volumes relative to total lung volume, since comparisons of physiological variables are generally based on normalized rather than absolute values (Jones and Longworth, 1992).

Statistics

Data referring to individual lungs are given as discrete values obtained according to standard stereological formulas. Mean values are given ± SD unless stated otherwise. Differences between control and experimental animals were tested for significance by an unpaired t-test, given that normality and equal variance were not rejected at P< 0.05. Otherwise, the nonparametric Mann-Whitney rank sum test was performed. Correlation between variables was tested by means of the nonparametric Spearman rank order correlation. We point out that the multiple correlation tests were used in terms of “exploratory” rather than “decisive” data analysis. All statistical analyses and computation of regression curves were performed using the SigmaStat 2.0 and SigmaPlot 3.0 software programs (Jandel Scientific, Erkrath, Germany). P<0.05 was considered to be statistically significant.

Functional Parameters

Hemodynamic and respiratory parameters are given in Table 2. The W/D-ratio ranged between 4.7 and 11.4 as compared to 5.1 ± 0.6 in the control group (Fig. 1), values also reported by others for normal and edematous lungs (Fisher et al., 1980; Michel et al., 1986; Uhlig et al., 1995). In the mean, oxygenation (ΔP_O2) was significantly decreased, total pulmonary vascular resistance (PVR) and PIP were significantly increased in comparison to baseline data. However, the individual lungs exhibited a broad range of data from normal to pathologic. Correlation analysis (Table 3) indicated significant negative relationships between oxygenation (Δ PO₂) and W/D-ratio (P< 0.01), and ΔP_O2 and PIP (r=-0.63; P<0.003), respectively. The virtual incongruity of a weak negative correlation between W/D-ratio and PVR (r=-0.49; P<0.05) will resolve in view of the stereological data.

Structural Parameters

Fluid accumulation resulted in a significantly higher (P<0.05) mean volume of the fixed left lung in the experimental group (2,298 ± 452 mm³) as compared with the control group (1,640 ± 293 mm³). A corresponding increase in the volume of the fixed lung has been reported to occur in hydrostatic pulmonary edema induced in isolated rabbit lungs (Wu et al., 1995). Wide variations were seen by LM and EM with respect to pulmonary edema and to the degree of damage to the air-blood-barrier. Accumulation of edema fluid was observed in different compartments of the lung: i) alveolar space and alveolar ducts (Fig. 2), ii) peribronchovascular space (Fig. 3), and iii) interstitium of the alveolar septum (Figs. 2b, c). In severely affected lungs, intraalveolar edema was accompanied by various structural defects as were extensive alveolar (Fig. 2a) and perivascular hemorrhage (Fig. 3b), extended damage of alveolar type 1 pneumocytes (Fig. 4), and large gaps in the alveolar epithelial boundary of perivascular cuffs (Fig. 5d, f). Edema fluid did not accumulate in the lumen of airways proximal to alveolar ducts. The absolute and relative volumes of the various compartments analyzed by stereology are shown in Tables 4 and 5. Looking at the estimates of the various edema accumulations, however, one has to take into account that the chemical fixation procedure used, which is a prerequisite for the stereological approach, might result in an underestimation of edema volumes (see discussion).

Interstitial vs. intraalveolar edema

Intraalveolar edema was by far the most significant type of edema in the ischemic/reperfused lungs. Equivalent amounts of total interstitial edema were present in all lungs, while differences between individual lungs largely resulted from intraalveolar edema (Fig. 6). The volume of intraalveolar edema fluid (V,ape) correlated positively with the volume of septal interstitium (r=0.53; P<0.025). The unexpected negative correlation (r=-0.63; P=0.005) of V,ape with the volume of free peribronchovascular space (V,psf) most probably resulted from displacement of edema fluid from the peribronchovascular into the alveolar space by extravasated erythrocytes. This is supported by the lack of correlation between V,ape and the volume of total peribronchovascular space (V,ps). Notably, the peribronchovascular space was increased at the expense of air-filled alveolar space (Figs. 3, 7) indicating that the expanding peribronchovascular cuffs encroach on the more compliant parenchymal airspace.

Injury to alveolar air-blood-barrier

TEM revealed a wide variation in the structural integrity of alveolar type I pneumocytes. Normally structured and edematously swollen cells were seen in direct vicinity to areas of fragmented cells or even denuded basal lamina (Fig. 4a, c). The surface fractions of epithelial areas showing the respective category of integrity are given in Table 6. In the experimental lungs, regions with fragmented type I pneumocytes amounted from only 0.8% up to 58.0% of the total type I pneumocyte surface (mean ± SD, 23.1 ± 16.8%). In the control lungs, fragmentation was limited to 4.6 ± 4.8 % of total type I cell surface, which was significantly different from the experimental group (P<0.02). Notably, fragmentation of type I pneumocytes in experimental lungs did not necessarily coincide with the barrier being covered by edema fluid (Fig. 4c). This is supported by the stereological finding (Table 7) that in the experimental lungs equivalent degrees of cellular injury were determined for edematous and edema-free regions. Notably, the extent of epithelial fragmentation did not unambiguously predict the extent of edema coverage of the alveolar epithelium (Fig. 8). While even small areas of fragmentation could result in extensive edema-coverage, a high degree of fragmentation was always associated with wide surface areas being covered by edema fluid. This could be further elucidated by lectin-histochemistry of sections of complete left lung slices. By means of BPL, a specific marker for type I pneumocytes (Kasper et al., 1994), large gaps were seen in the alveolar epithelium separating alveoli and perivascular cuffs (Fig. 5a, d). By immunohistochemistry using an established epithelial cell-specific marker (Kasper and Singh, 1995) the disintegration of the alveolar epithelium with a loss of cytokeratin into the intraalveolar edema fluid could predominantly be seen in alveoli close to pulmonary vessels (Fig. 5b, c, e, f). Ultrastructural analysis of the liquid-gas interphase (Fig. 4b, e) indicated that the surface-active lining layer was capable of limiting the accumulation of intraalveolar edema fluid to preexisting pools. Notably, tubular myelin, the surface-active form of alveolar surfactant sub-types, frequently appeared irregular and disintegrated in experimental lungs (Fig. 4d).

Table 6. Lung structure—quantitative study of cell injury

Stereological parameter	Control (n = 6)		Experimental (n = 19)
	Mean ± SD	Median	Mean ± SD	Median		Level of significancec
Ss ape, EP [%]a	7.0 ± 5.0	7.2	52.4 ± 15.0	49.5	P < 0.001
Ss np1, P1 [%]b	80.8 ± 13.5		56.8 ± 19.7		P < 0.02
Ss sp1, P1 [%]b	14.6 ± 12.6		20.0 ± 10.9		n.s.
Ss fp1, P1 [%]b	4.6 ± 4.8	3.0	23.1 ± 16.8	23.5	P < 0.02

• a Fraction of alveolar epithelial surface covered with edema (Ss ape, EP) was determined by LM at level 3.
• b Fractions of type I pneumocytes exhibiting normal (Ss np1, P1), swollen, (Ss sp1, P1) or fragmented (Ss fp1, P1) ultrastructure were determined by TEM at level 4.
• c Differences between groups were tested for significance (P) according to an unpaired t-test if normality and equal variance were not rejected at P < 0.05. Otherwise, the nonparametric Mann-Whitney rank sum test was performed, and the median values are given.

Table 7. Degree of alveolar epithelial injury in edematous versus edema-free alveoli

Stereological parametera	Type I cells, free of edema Mean ± SD	Type I cells, covered with edema Mean ± SD	Level of significanceb
Control (n = 6)
Ss np1, P1 [%]	84.1 ± 11.9	53.4 ± 34.8	P = 0.068
Ss sp1, P1 [%]	11.5 ± 9.9	41.9 ± 34.5	P = 0.065
Ss fp1, P1 [%]	4.4 ± 5.4	4.7 ± 6.9	P = 0.943
Experimental (n = 19)
Ss np1, P1 [%]	62.0 ± 23.1	59.8 ± 23.9	P = 0.771
Ss sp1, P1 [%]	17.6 ± 15.0	13.5 ± 10.5	P = 0.331
Ss fp1, P1 [%]	20.4 ± 18.1	26.8 ± 19.6	P = 0.305

• a Fractions of type I pneumocytes exhibiting normal (Ss np1, P1), swollen, (Ss sp1, P1) or fragmented (Ss fp1, P1) ultrastructure were determined by TEM at level 4.
• b Differences between groups were tested for significance (P) according to an unpaired t-test, because normality and equal variance were not rejected at P < 0.05.

Explorative analysis of structure-function correlation

Significant relationships based on the analysis of the individual data within the ischemia/reperfusion group were seen (Table 8). Total pulmonary edema, i.e., the sum of intraalveolar (Vv ape,L), peribronchovascular (Vv psf,L), and interstitial edema of alveolar septa (Vv int/L) correlated significantly with the W/D-ratio (r=0.61, P<0.01). Only intraalveolar edema alone showed a significant correlation with the W/D-ratio. Intraalveolar edema further correlated highly significant with oxygenation, with the volume of intraalveolar edema relative to the gas exchange region (Vv ape,P) showing the strongest correlation with ΔP_O2 (r=-0.82; P<0.001) (Fig. 9). Further, the more intraalveolar edema was present the higher was the PIP measured (r=0.65; P<0.005). The weak inverse relationship between PIP and peribronchovascular edema simply reflects the inverse relationship between peribronchovascular and intraalveolar edema. The same indirect relationship was seen with PVR, which showed a highly significant increase with peribronchovascular edema (r=0.64; P<0.005), but in turn showed a virtual decrease with increasing intraalveolar edema. This indirect relation was also reflected by the negative relationship of PVR with W/D-ratio (Table 2). While pulmonary gas exchange and intraalveolar edema did not correlate with epithelial injury, a negative correlation of alveolar epithelial integrity (Ss np1,P1) and PVR was noted (r=-0.58, P<0.01).

Table 8. Structure-function correlation matrix (n = 19 experimental animals)

	Total pulmonary edema	Intraalveolar edema		Peribronchovascular edema	Alveolar septal edema		Cellular injury
	Vv ape, L + Vv psf, L	Vv ape, L	Ss ape, Ep	Vv psf, L	Vv int, L	τint	Ss np1, P1	Ss fp1, P1
	+ Vv int, L [%]	[%]	[%]	[%]	[%]	[μm]	[%]	[%]
W/D-ratio	r = 0.60 P < 0.01	r = 0.56 P < 0.02	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.
ΔPO2 [mm Hg]	r = −0.66 P = 0.002	r = −0.72 P < 0.001	r = −0.45 P < 0.05	n.s.	n.s.	n.s.	n.s.	n.s.
ΔPCO2 [mm Hg]	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.
PIP [cm H2O]	r = 0.56 P < 0.02	r = 0.65 P < 0.005	r = 0.54 P < 0.02	r = −0.49 P < 0.05	n.s.	n.s.	n.s.	n.s.
Ppa [mm Hg]	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.	n.s.
PVR [dyn × sec × cm−5]	n.s.	r = −0.55 P < 0.02	n.s.	r = 0.64 P < 0.005	n.s.	n.s.	r = −0.58 P < 0.01	n.s.

• ^a Correlation coefficients (r) with levels of significance (P) determined according to Spearman's rank order correlation.

Methodological Aspects

The rat lungs examined showed a wide range of slight to severe pulmonary edema as reflected by the W/D-ratio. On the basis of semi-quantitative data, Montaner and co-workers (1986) showed that both interstitial and intraalveolar edema may result in equivalent W/D-ratios. Our stereologic approach allowed to estimate the volumes of the respective compartments to quantitate the particular contributions of interstitial and intraalveolar edema. Several studies clearly showed that, suitable fixation procedures given, LM and TEM are appropriate techniques to demonstrate and quantify edema accumulations in the peribronchovascular space (Michel et al., 1986; 1987; Zwikler et al., 1994) as well as in the alveolar space (Bachofen et al., 1993; Wu et al., 1995). There is no doubt that cryofixation is the best means to retain fluid in lung samples (Bastacky et al., 1995; Luchtel et al., 1991). However, cryofixation suitable for EM investigation is limited to the region just below the pleural surface (Bastacky et al., 1995). Stereological studies investigating organ function require that a representative collection of tissue blocks is obtained from the whole organ to guarantee an unbiased sampling protocol (Mayhew, 1991). Therefore, chemical fixation has to be performed to ensure adequate structural preservation of all parts of the lung. Adequate preservation of edema accumulations can be achieved by means of sequential vascular perfusion (Bachofen et al., 1982: method IV) or sequential processing, as in our study, using a glutaraldehyde-containing primary fixative, and osmium tetroxide and uranyl acetate as post-fixatives (Bachofen et al., 1982: method II). This approach permits studying the same lung by means of LM, SEM, and TEM with all samples being collected by unbiased systematic random sampling based on the entire organ. On this basis only, the volume of edema fluid accumulations and the extent of cell injury to the air-blood-barrier could be estimated for the same lung.

One may argue that the exclusion of leukocytes from the blood perfusate in our model might have resulted in less severe ischemia/reperfusion effects, because leukocytes may also contribute to reperfusion injury. However, the contribution of leukocytes to the development of reperfusion injury is controversially discussed (Deeb et al., 1990; Seibert et al., 1993).

Distribution of Edema and Functional Significance

The stereological estimates of the volumes of all three edema compartments, i. e. alveolar space, peribronchovascular space, and interstitium of alveolar septa, were significantly increased after preservation, ischemia and reperfusion as compared with control lungs. The mean volumetric increase over control values of the interstitial space of alveolar septa was nearly identical to the increase of the peribronchovascular space, i. e. about 72 mm³ versus 82 mm³. With a mean volume of 448 mm³, intraalveolar edema, however, clearly exceeded the rise in peribronchovascular and alveolar septal edema by about 5- to 6-fold. As a consequence, pulmonary function was governed by the amount of intraalveolar edema as indicated by its significant correlation with W/D-ratio, PIP, and ΔP_O2. Our data not only support the general view of alveolar flooding being responsible for respiratory impairment (Michel et al., 1986; Montaner et al., 1986; Staub, 1983), but we could further develop a logarithmic relationship between intraalveolar edema and ΔP_O2 following ΔP_O2 = 96 - 60 × log₁₀(Vv ape,P) [mm Hg] (Fig. 9). This agrees well with physiological data of 83.5 - 96 mm Hg reported for normal rats of 555–297 g body weight (Boggs, 1992). We may infer from the equation developed that if intraalveolar edema occupies about 10% of the lung's gas exchange region, pulmonary oxygenation will only reach about 40 mm Hg, while with an edema fraction of 40%, oxygenation will drop to zero.

Peribronchovascular edema showed a significant correlation with PVR only. This, however, could not be explained by compression of the nonparenchymal vasculature. The absolute volumes of the lumina of vessels and of airways of the experimental lungs did not differ significantly from control values. There was also no decrease in absolute or relative volumes of their lumina with increasing peribronchovascular space. These observations are consistent with the findings of Michel and co-workers (1987) that neither vessels nor airways are compressed by interstitial peribronchovascular edema. These authors suggested that the expanding peribronchovascular cuffs encroach on the more compliant parenchyma rather than on the lumina of vessels or airways. This is supported by our stereological finding that the increase in peribronchovascular space per lung volume was achieved at the expense of alveolar space (Fig. 7). Hence, the significant correlation between peribronchovascular edema and PVR may rely on an indirect effect of the expanding peribronchovascular space on the capillary bed. This is in line with findings obtained from isolated perfused dog lung lobes, which showed that vascular resistance raised and the number of open capillaries decreased only after interstitial edema had developed, but that the mere presence of liquid in the alveoli had no effect (Muir et al., 1975).

Cell Injury and Edema Formation

As a consequence of ischemia/reperfusion, a significant increase in alveolar epithelial cell injury was seen in the experimental group compared with the control group, while in turn the surface fraction of normal type I pneumocytes decreased. In parallel, the amount of intraalveolar edema was increased in experimental lungs. These findings are in line with other studies comparing different experimental groups (Bachofen et al., 1993; Montaner et al., 1986; Wu et al., 1995). Our analysis of the individual data, however, revealed that the extent of epithelial injury did not unambiguously determine the amount of intraalveolar edema. While highly injured experimental lungs always showed extended coverage of the epithelium with edema, some lungs with minor epithelial fragmentation also developed considerable edema (Fig. 8). This may be explained by our lectin- and immunohistochemical findings of large gaps in the alveolar epithelium lining the perivascular cuffs. These lesions likely represent the route along which excess interstitial fluid pours into and drains the encompassing alveoli, and may further proceed via pores of Kohn to adjacent alveoli, irrespective of the integrity of their air-blood-barrier. Alternatively, this unexpected phenomenon may be explained by the pulmonary surfactant being involved. The epithelial lining layer, which due to the surface tension created contributes to the pressure gradient across the air-blood-barrier (Clements, 1961), appeared to be capable of limiting the accumulation of intraalveolar edema fluid to the preexisting pools, which is in line with observations reported by Bachofen and co-workers (1993). With the surface tension being increased in fluid filled alveoles, probably in consequence of the disintegrated tubular myelin (Fig. 4d), a pressure gradient may be created across the air-blood-barrier along which edema fluid drains into a preexisting pool through small epithelial gaps, rather than flowing to the opposite side of the barrier despite the presence of a large area of fragmentation. Hence, studies of the relationship between intraalveolar edema accumulation and pulmonary surfactant alterations are needed to further enlighten this problem.