Surface forces in lungs. II. Microstructural mechanics and lung stability

Recent lung microstructural models describing interactions between alveolar surface tension (gamma) and forces in structural elements of the alveolar duct predict that the component of lung recoil pressure due to gamma (P gamma) is proportional to gamma/V1/3, where V is the total lung volume. This relation is tested against experimental data obtained from pressure-volume measurements of excised rabbit lungs with different constant values of gamma. It is found that for values of gamma less than approximately 18 dyn/cm the data generally agree with the model predictions. With higher values of gamma, a mismatch between the data and predictions first occurs at low and high volumes and then spreads over the entire volume range. The mismatch at the lower volumes coincides with the appearance of nonuniformities of lung expansion. The nonuniformities are characterized by a coexistence of under- and overexpanded regions of the parenchyma referred to as a mixture of phases. These nonuniformities, as well as a pressure-volume curve with a shape similar to the shape of measured curves, are predicted from an analysis of lung stability. Results of this work indicate that if the lung expands uniformly, P gamma proportional to gamma/V1/3 is a good approximation over a wide range of volumes. The stability analysis indicates that the equilibrium configurations of the lung parenchyma when gamma is independent of interfacial area and elevated above normal values are nonuniform states of expansion, characterizable as a mixture of phases. This result confirms that a dependence of gamma on surface area is normally required to achieve stable, uniform states of lung expansion.     

Effect of long-term anemia and retransfusion on central circulation during exercise

The bulk modulus and the shear modulus describe the capacity of material to resist a change in volume and a change of shape, respectively. The values of these elastic coefficients for air-filled lung parenchyma suggest that there is a qualitative difference between the mechanisms by which the parenchyma resists expansion and shear deformation; the bulk modulus changes roughly exponentially with the transpulmonary pressure, whereas the shear modulus is nearly a constant fraction of the transpulmonary pressure for a wide range of volumes. The bulk modulus is approximately 6.5 times as large as the shear modulus. In recent microstructural modeling of lung parenchyma, these mechanisms have been pictured as being similar to the mechanisms by which an open cell liquid foam resists deformations. In this paper, we report values for the bulk moduli and the shear moduli of normal air-filled rabbit lungs and of air-filled lungs in which alveolar surface tension is maintained constant at 16 dyn/cm. Elevating surface tension above normal physiological values causes the bulk modulus to decrease and the shear modulus to increase. Furthermore, the bulk modulus is found to be sensitive to a dependence of surface tension on surface area, but the shear modulus is not. These results agree qualitatively with the predictions of the model, but there are quantitative differences between the data and the model.     

Surface biophysics of the surface monolayer theory is incompatible with regional lung function.

The surface monolayer theory of Clements was tested on open surface films of calf lung surfactant extract in a leak-free vertical film surface balance in which alveolar area (A) changes in each lung zone were simulated in accordance with the theory. We found that: 1) physiologically necessary low surface tension (gamma), < 4 dyn/cm, was sustained only by continuous film compression ("expiration"); 2) compression from A equivalent to total lung capacity to functional residual capacity produced fleeting gamma reduction in all zones and quick reversal to high gamma with A changes that simulated tidal volume (VT) breathing at both 14 (adult) and 40 (neonatal) cpm; 3) phase differences between gamma and A axes of VT loops that indicate mixed surface film composition may be attributable to film inertia and viscoelasticity; 4) estimated alveolar retraction pressure due to gamma (P gamma) exceeds "net" transpulmonary pressure, i.e., favors alveolar collapse, under virtually all conditions of the theory in all zones; 5) return to transient, fleeting low gamma in successive VT cycles was determined by the inherent difference in compression and decompression rates, which results in exhaustion of available A in very few cycles; 6) the "sigh", which restores stable low gamma according to the theory, actually produced unstable high gamma during virtually all phases of the maneuver. In contrast, closed bubble films of the surfactant were structurally stable and produce stable near 0 gamma and P gamma.     

Elastic properties of air- and liquid-filled lung parenchyma

Pressure-volume measurements and the punch indentation test are used to obtain the bulk modulus (kappa) and the shear modulus (mu) of lung parenchyma of air- and liquid-filled rabbit lungs. Plots of kappa and mu vs. transpulmonary pressure obtained from these measurements indicate that there is very little difference between the elastic behavior of the air- and liquid-filled lung, suggesting that the mechanism of resisting deformation in both cases is similar. On the other hand, from plots of kappa and mu vs. lung volume, it appears that the elastic moduli are higher in the air-filled lung than in the liquid-filled lung at the same volume. These differences, referred to as kappa gamma and mu gamma, as well as the difference in transpulmonary pressures (P gamma), are presumably due to the additional elastic recoil of the air-filled lung provided by alveolar surface tension (gamma). No conclusion could be reached about the shape of the kappa gamma vs. P gamma curve. However, the mu gamma vs. P gamma relationship appears to be approximately linear, with a slope of approximately 0.5. This result agrees qualitatively with the model (T. A. Wilson and H. Bachofen, J. Appl. Physiol. 52: 1064-1070, 1982) in which the part of the parenchyma that provides P gamma is pictured as mechanically analogous to an open cell liquid foam, having mu gamma = 0.4P gamma (J. Appl. Mech. Trans. ASME 51: 229-231, 1984), but it is statistically significant only at high lung volumes.     

Biophysical characterization and modeling of lung surfactant components.

The present study characterizes the dynamic interfacial properties of calf lung surfactant (CLS) and samples reconstituted in a stepwise fashion from phospholipid (PL), hydrophobic apoprotein (HA), surfactant apoprotein A (SP-A), and neutral lipid fractions. Dipalmitoylphosphatidylcholine (DPPC), the major PL component of surfactant, was examined for comparison. Surface tension was measured over a range of oscillation frequencies (1-100 cycles/min) and bulk phase concentrations (0.01-1 mg/ml) by using a pulsating bubble surfactometer. Distinct differences in behavior were seen between samples. These differences were interpreted by using a previously validated model of surfactant adsorption kinetics that describes function in terms of 1) adsorption rate coefficient (k1), 2) desorption rate coefficient (k2), 3) minimum equilibrium surface tension (gamma*), 4) minimum surface tension at film collapse (gammamin), and 5) change in surface tension with interfacial area for gamma < gamma* (m2). Results show that DPPC and PL have k1 and k2 values several orders of magnitude lower than CLS. PL had a gammamin of 19-20 dyn/cm, significantly greater than CLS (nearly zero). Addition of the HA to PL restored dynamic interfacial behavior to nearly that of CLS. However, m2 remained at a reduced level. Addition of the SP-A to PL + HA restored m2 to a level similar to that of CLS. No further improvement in function occurred with the addition of the neutral lipid. These results support prior studies that show addition of HA to the PL markedly increases adsorption and film stability. However, SP-A is required to completely normalize dynamic behavior.     

Viscoelastic behavior of a lung alveolar duct model

A study is conducted into the oscillatory behavior of a finite element model of an alveolar duct. Its load-bearing components consist of a network of elastin and collagen fibers and surface tension acting over the air-liquid interfaces. The tissue is simulated using a visco-elastic model involving nonlinear quasi-static stress-strain behavior combined with a reduced relaxation function. The surface tension force is simulated with a time- and area-dependent model of surfactant behavior. The model was used to simulate lung parenchyma under three surface tension cases: air-filled, liquid-filled, and lavaged with 3-dimenthyl siloxane, which has a constant surface tension of 16 dyn/cm. The dynamic elastance (Edyn) and tissue resistance (Rti) were computed for sinusoidal tidal volume oscillations over a range of frequencies from 0.16-2.0 Hz. A comparison of the variation of Edyn and Rti with frequency between the model and published experimental data showed good qualitative agreement. Little difference was found in the model between Rti for the air-filled and lavaged models; in contrast, published data revealed a significantly higher value of Rti in the lavaged lung. The absence of a significant increase in Rti for the lavaged model can be attributed to only minor changes in the individual fiber bundle resistances with changes in their configuration. The surface tension was found to make an important contribution to both Edyn and Rti in the air-filled duct model. It was also found to amplify any existing tissue dissipative properties, despite exhibiting none itself over the small tidal volume cycles examined.     

Alveolar surface forces and lung architecture

The entire alveolar surface is lined by a thin fluid continuum. As a consequence, surface forces at the air-liquid interface are operative, which in part are transmitted to the delicate lung tissue. Morphologic and morphometric analyses of lungs show that the alveolar surface forces exert a moulding effect on alveolar tissue elements. In particular, in lungs at low degrees of inflation, equivalent to the volume range of normal breathing, there is a derecruitment of alveolar surface area with increasing surface tensions which reflects equilibrium configurations of peripheral air spaces where the sum of tissue energy and surface energy is minimum. Thus, changes in surface tension alter the recoil pressure of the lung directly and indirectly by deforming lung tissue and hence changing tissue tensions. However, the interplay between tissue and surface forces is rather complex, and there is a marked volume dependence of the shaping influence of surface forces. With increasing lung volumes the tissue forces transmitted by the fiber scaffold of the lung become the predominant factor of alveolar micromechanics: at lung volumes of 80% total lung capacity or more, the alveolar surface area-volume relation is largely independent of surface tension. Most important, within the range of normal breathing, the surface tension, its variations and the associated variations in surface area are small. The moulding power of surface forces also affects the configuration of capillaries, and hence the microcirculation, of free cellular elements such as the alveolar macrophages beneath the surface lining layer, and of the surfaces of the peripheral airways. Still enigmatic is the coupling mechanism between the fluid continua of alveoli and airways which might also be of importance for alveolar clearance. As to the surface active lining layer of peripheral air spaces, which determines alveolar surface tension, its structure and structure-function relationship are still ill-defined owing to persisting problems of film preservation and fixation. Electron micrographs of alveolar tissue, of lining layers of captive bubbles, and scanning force micrographs of surfactant films transferred on mica plates reveal a complex structural pattern which precludes so far the formulation of an unequivocal hypothesis.     

A species comparison of alveolar size and surface forces

The independent roles of alveolar size and surface tension in relation to lung stability were investigated in 11 different mammalian species whose body weight ranged from 0.03 to 50 kg. This range in species provided a wide variation in subgross anatomy as well as a fourfold range in alveolar diameter. Alveolar diameter was estimated from the mean linear intercept (Lm) of fixed lungs. Quasi-static pressure-volume curves were determined in excised lungs and the percent volume remaining on deflation from total lung capacity at 30 cmH2O to 10 cmH2O (%V10) provided an index of deflation stability related to functional surfactant. Surface tension of lung extract was measured in the Wilhelmy balance, and the minimum surface tension measured provided an index of surface tension lowering capacity of surfactant. Relationships of %V10 with alveolar diameter and surface tension with alveolar diameter were examined for correlations. Our results indicated that despite a range in Lm between 31 and 133 micron (mouse to pig), %V10 did not change in proportion with Lm across species. Similarly, minimum surface tension was about the same (6.1 to 8.8 dyn/cm) across a threefold difference in alveolar diameter. These results suggest that a stable alveolar configuration is maintained by both surface and tissue forces in a complex manner yet to be analyzed.     

Three-dimensional reconstruction of alveoli in the rat lung for pressure-volume relationships

To determine alveolar pressure-volume relationships, alveolar three-dimensional reconstructions were prepared from lungs fixed by vascular perfusion at various points on the pressure-volume curve. Lungs from male Sprague-Dawley rats were fixed by perfusion through the pulmonary artery following a pressure-volume maneuver to the desired pressure point on either the inflation or deflation curve. Tissue samples from lungs were serially sectioned for determination of the volume fraction of alveoli and alveolar ducts and reconstruction of alveoli. Alveoli from lungs fixed at 5 cmH2O on the deflation curve (approximating functional residual volume) had a volume of 173 X 10(3) microns3, a surface area of 11,529 microns2, a mouth opening diameter of 72.7 microns, and a mean caliper diameter of 91.8 micron (SE). Alveolar shape changes during deflation from total lung capacity to residual volume was first (30 to 10 cmH2O) associated with little change in the diameter of the alveoli (102.7 +/- 2.4 to 100.3 +/- 3.3 microns). In the range overlapping normal breathing (10 to 0 cmH2O) there was a substantial decrease in diameter (100.3 +/- 3.3 to 43.3 +/- 2.3 microns). These measurements and others made on the relative changes in the dimensions of the alveolus suggest that the elastic network, particularly around the alveolar ducts, are predominant in determining lung behavior near the volume expansion limits of the lung while the elastic and surface tension properties of the alveoli are predominant in the volume range around functional residual capacity.     

Surface properties of rat pulmonary surfactant studied with the captive bubble method: adsorption, hysteresis, stability.

Surface tension-area relations from pulmonary surfactant were obtained with a new apparatus that contains a leak free captive bubble of controllable size. Rat pulmonary surfactant was studied at phospholipid concentrations of 50, 200 and 400 micrograms/ml. At the highest concentration, adsorption was rapid, reaching surface tensions below 30 mN/m within 1 s, while at the lowest concentration, approximately 3 min were required. Upon a first quasi static or dynamic compression, stable surface tensions below 1 mN/m could be obtained by a film area reduction of approximately 50%. After three to four cycles the surface tension-area relations became stationary, and the tension fell from 25-30 to approximately 1 mN/m for a film area reduction of less than 20%. Hysteresis became negligible, provided the films were not collapsed by further area reduction. Under these conditions, the films could be cycled for more than 20 min without any noticeable loss in surface activity. After only three to four consecutive cycles, surfactant films exhibited the low surface tensions, collapse rates and compressibilities characteristic of alveolar surfaces in situ. Remarkably, surface tension and area are interrelated in the captive bubble which may promote low and stable surface tensions. If the surface tension of the captive bubble suddenly increases ('click') because of mechanical vibration or unstable surfactant, the bubble shape changes from flat to more spherical. The associated isovolumetric decrease in surface area prevents the surface tension from rising as much as it would have in a constant-area situation. This feedback mechanism may also have a favorable effect in stabilizing alveolar surface tension at low lung volumes.     

Surface activity following natural surfactant treatment in premature lambs

Premature lambs with respiratory failure [CO2 partial pressure (PCO2) greater than 70 Torr] were treated with 50 mg/kg 3H-labeled natural surfactant by tracheal instillation. Minimum surface tensions of sequential samples suctioned from the airways fell from 25 +/- 3 dyn/cm before treatment to 8 +/- 5 dyn/cm after treatment and again rose to 32 +/- 2 dyn/cm at death. Minimum surface tensions of alveolar wash samples taken at death were 27 +/- 4 dyn/cm, whereas surfactant fractions reisolated from the alveolar washes lowered surface tension to under 10 dyn/cm. The alveolar washes, surfactant reisolated from the alveolar washes, and natural surfactant had similar phospholipid compositions; however, the alveolar washes contained about 40 times more protein per micromole phosphatidylcholine. The natural surfactant used for treatment apparently was inactivated by an inhibitor of surfactant function. After intravenous injections of [14C]palmitic acid, labeled saturated phosphatidylcholine appeared on the airways, indicating endogenous synthesis and secretion. However, the specific activity of the 3H-labeled saturated phosphatidylcholine in the natural surfactant used for treatment decreased by only 30 +/- 4% in the alveolar wash; thus the treatment dose was not diluted to a large extent by endogenous pools.     

Relations among alveolar surface tension, surface area, volume, and recoil pressure

For pulmonary structure-function analysis excised rabbit lungs were fixed by vascular perfusion at six points on the pressure-volume (P-V) curve, i.e. at 40, 80, and 100% of total lung capacity (TLC) on inflation, at 80 and 40% TLC on deflation, and at 80% TLC on reinflation. Before fixation alveolar surface tensions (gamma) were measured in individual alveoli over the entire P-V loop, using an improved microdroplet method. A maximal gamma of approximately 30 mN/m was measured at TLC, which decreased during lung deflation to about 1 mN/m at 40% TLC. Surface tensions were considerably higher on the inflation limb starting from zero pressure than on the deflation limb (gamma-V hysteresis). In contrast, the corresponding alveolar surface area-volume (SA-V) relationship did not form a complete hysteresis over the entire volume range. There was a considerable difference in SA between lungs inflated to 40% TLC (1.49 +/- 0.11 m2) and lungs deflated to 40% TLC (2.19 +/- 0.21 m2), but at 80% TLC the values of SA were essentially the same regardless of the volume history. The data indicate that the gamma-SA hysteresis is only in part accountable for the P-V hysteresis and that the determinative factors of alveolar geometry change with lung volume. At low lung volumes airspace dimensions appear to be governed by an interplay between surface and tissue forces. At higher lung volumes the tissue forces become predominant.     

Mechanics of edematous lungs.

Using the parenchymal marker technique, we measured pressure (P)-volume (P-V) curves of regions with volumes of approximately 1 cm3 in the dependent caudal lobes of oleic acid-injured dog lungs, during a very slow inflation from P = 0 to P = 30 cmH2O. The regional P-V curves are strongly sigmoidal. Regional volume, as a fraction of volume at total lung capacity, remains constant at 0.4-0.5 for airway P values from 0 to approximately 20 cmH2O and then increases rapidly, but continuously, to 1 at P = approximately 25 cmH2O. A model of parenchymal mechanics was modified to include the effects of elevated surface tension and fluid in the alveolar spaces. P-V curves calculated from the model are similar to the measured P-V curves. At lower lung volumes, P increases rapidly with lung volume as the air-fluid interface penetrates the mouth of the alveolus. At a value of P = approximately 20 cmH2O, the air-fluid interface is inside the alveolus and the lung is compliant, like an air-filled lung with constant surface tension. We conclude that the properties of the P-V curve of edematous lungs, particularly the knee in the P-V curve, are the result of the mechanics of parenchyma with constant surface tension and partially fluid-filled alveoli, not the result of abrupt opening of airways or atelectatic parenchyma.     

Respiratory distress and surfactant inhibition following vagotomy in rabbits

We used the model of bilateral cervical vagotomy of adult rabbits to cause respiratory failure characterized by pulmonary edema, decreased lung compliance, and atelectasis. We documented an 18-fold increase in radiolabeled albumin leak from the vascular space into alveolar washes of vagotomy vs. sham-operated rabbits (P less than 0.01). Despite a twofold increase in percent of prelabeled saturated phosphatidylcholine secreted (P less than 0.01), the alveolar wash saturated phosphatidylcholine pool sizes were not different. The minimum surface tensions were 19.6 +/- 2.5 vs. 9.4 +/- 2.2 dyn/cm for alveolar washes from vagotomy and control rabbits, respectively (P less than 0.01). The soluble proteins from alveolar washes inhibited the surface tension lowering properties of natural surfactant, whereas those from the control rabbits did not (P less than 0.01). When vagotomy rabbits in respiratory failure were treated with 50 mg natural surfactant lipid per kilogram arterial blood gas values and compliances improved relative to control rabbits. Vagotomy results in alveolar pulmonary edema, and surfactant dysfunction despite normal surfactant pool sizes and respiratory failure. A surfactant treatment can improve the respiratory failure.     

Surface activity in situ, in vivo, and in the captive bubble surfactometer

For studies of the mechanical effects of lung surfactants, the captive bubble surfactometer (CBS) combines the advantages of the continuous film of Pattle's bubbles with the feasibility of the Langmuir-Wilhelmy balance to produce surface tension-area hysteresis loops. The CBS allows the compression of films to very low and stable surface tensions of 1-2 mN/m. Such low and stable surface tensions are in line with results obtained from pressure-volume studies on excised lungs. In addition, the CBS is useful to test other essential physical properties of the surfactant system, including: (1) rapid film formation (within seconds) through adsorption from the hypophase; (2) low film compressibility with a fall in surface tension to very low (<2 mN/m) values during surface compression; and (3) effective replenishment of the surface film on expansion by the incorporation of surfactant material from material associated with the surface (the surface associated surfactant reservoir). Morphological observations of films fixed in situ or in vitro reveal frequently their multilayered structure, which is consistent with the concept of the surface reservoir. The deviation of the bubbles from a Laplacian shape at very low surface tension and the morphological observations suggest that the surfactant film cannot be considered a simple monolayer.     

Changes in quantity, composition, and surface activity of alveolar surfactant at birth

We hypothesized that when the lung makes the transition from the fluid- to the air-filled state at birth, there are changes in physical and functional properties of the alveolar surfactant. To test this hypothesis, newborn rabbits were killed at different times in the first 24 h of life, their lungs lavaged with ice-cold saline, and the lavage fluid subfractionated by differential centrifugation. The phospholipid and protein content and composition and the kinetics of surface tension lowering of the subfractions were examined. We found that with the onset of breathing, shifts occur in the distribution of surfactant subfractions as a surfactant apoprotein-free phospholipid fraction is generated. The ratio of rapidly sedimentable apoprotein-rich to slowly sedimentable, apoprotein-free fractions decreases from 31 at birth to 4 at 24 h of life. Concurrently, rates of surface tension lowering by the subfractions increase with time. The results suggest that the adult pattern of pool sizes and surface activity of alveolar surfactant is not present at birth but evolves slowly over the 1st day of life.     

Phosphatidylglycerol in lung surfactant. III. Possible modifier of surfactant function.

Lamellar bodies and alveolar lavage from adult mammalian lung contain unusually high concentrations of phosphatidylglycerol that could serve as a sensitive indicator of surfactant. Phosphatidylglycerol was absent and phosphatidylinositol was correspondingly prominent in surfactant from the preterm rabbit fetus. Phosphatidylglycerol rapidly appeared and phosphatidylinositol decreased following the delivery. Surfactant isolated from the prematurely born rabbit or from humans with respiratory distress syndrome never contained phosphatidylglycerol. Comparison between lamellar bodies from fetal and postnatal rabbits revealed remarkably similar composition except for the acidic phospholipids; however, the physico-chemical properties were different. The compressibility of the surface film (i.e. the ratio of the fractional decrease in surface area and the corresponding decrease in surface tension) at low surface tensions was higher with fetal than with postnatal surfactant, whereas the difference in minimum surface tensions was small. These data suggest that phosphatidylglycerol is not an essential component required for the formation of the complex, but it improves the properties of surfactant in stabilizing the alveoli.     

Changes of lung surfactant and pressure-volume curve in bleomycin-induced pulmonary fibrosis.

We investigated whether alveolar surface force increased and participated in the lung pressure-volume relationship in bleomycin-induced pulmonary fibrosis in hamsters and, if so, whether lung surfactant was hampered in the lungs. On the air-filled pressure-volume curve, decreases of lung volume from control level were significantly higher at 3-8 cmH2O pressure on day 10 than on day 30. Because the change of lung tissue elasticity evaluated from the saline-filled pressure-volume curve was equal for the 2 days, the higher decrease of air volume on day 10 was due primarily to contribution of alveolar surface force. Pressure differences between deflation limbs of air-filled and saline-filled pressure-volume curves, which represented net alveolar surface force, were significantly higher at any lung volume between 50 and 90% total lung capacity on day 10, but almost no significance was observed on day 30. Phospholipid concentration in bronchoalveolar lavage fluid significantly decreased on day 10 but had improved by day 30. Analysis of phospholipid species in purified lung surfactant showed decreased fractions of disaturated phosphatidylcholine and phosphatidylglycerol on day 10. Surface-active properties of the surfactant, measured by a modified Wilhelmy balance, were remarkably hampered on day 10, but most of them had improved by day 30. We consider that the quantitative and functional abnormalities of lung surfactant have a part in the aggravation of lung mechanics in the acute phase of pulmonary fibrosis.     

Theoretical modeling of the interaction between alveoli during inflation and deflation in normal and diseased lungs

Alveolar recruitment is a central strategy in the ventilation of patients with acute lung injury and other lung diseases associated with alveolar collapse and atelectasis. However, biomechanical insights into the opening and collapse of individual alveoli are still limited. A better understanding of alveolar recruitment and the interaction between alveoli in intact and injured lungs is of crucial relevance for the evaluation of the potential efficacy of ventilation strategies. We simulated human alveolar biomechanics in normal and injured lungs. We used a basic simulation model for the biomechanical behavior of virtual single alveoli to compute parameterized pressure–volume curves. Based on these curves, we analyzed the interaction and stability in a system composed of two alveoli. We introduced different values for surface tension and tissue properties to simulate different forms of lung injury. The data obtained predict that alveoli with identical properties can coexist with both different volumes and with equal volumes depending on the pressure. Alveoli in injured lungs with increased surface tension will collapse at normal breathing pressures. However, recruitment maneuvers and positive endexpiratory pressure can stabilize those alveoli, but coexisting unaffected alveoli might be overdistended. In injured alveoli with reduced compliance collapse is less likely, alveoli are expected to remain open, but with a smaller volume. Expanding them to normal size would overdistend coexisting unaffected alveoli. The present simulation model yields novel insights into the interaction between alveoli and may thus increase our understanding of the prospects of recruitment maneuvers in different forms of lung injury.     

Stress failure in pulmonary capillaries

In the mammalian lung, alveolar gas and blood are separated by an extremely thin membrane, despite the fact that mechanical failure could be catastrophic for gas exchange. We raised the pulmonary capillary pressure in anesthetized rabbits until stress failure occurred. At capillary transmural pressures greater than or equal to 40 mmHg, disruption of the capillary endothelium and alveolar epithelium was seen in some locations. The three principal forces acting on the capillary wall were analyzed. 1) Circumferential wall tension caused by the transmural pressure. This is approximately 25 dyn/cm (25 mN/m) at failure where the radius of curvature of the capillary is 5 microns. This tension is small, being comparable with the tension in the alveolar wall associated with lung elastic recoil. 2) Surface tension of the alveolar lining layer. This contributes support to the capillaries that bulge into the alveolar spaces at these high pressures. When protein leakage into the alveolar spaces occurs because of stress failure, the increase in surface tension caused by surfactant inhibition could be a powerful force preventing further failure. 3) Tension of the tissue elements in the alveolar wall associated with lung inflation. This may be negligible at normal lung volumes but considerable at high volumes. Whereas circumferential wall tension is low, capillary wall stress at failure is very high at approximately 8 x 10(5) dyn/cm2 (8 x 10(4) N/m2) where the thickness is only 0.3 microns. This is approximately the same as the wall stress of the normal aorta, which is predominantly composed of collagen and elastin. The strength of the thin part of the capillary wall is probably attributable to the collagen IV of the basement membranes. The safety factor is apparently small when the capillary pressure is raised during heavy exercise. Stress failure causes increased permeability with protein leakage, or frank hemorrhage, and probably has a role in several types of lung disease.