Bioengineering the Blood-gas Barrier

Abstract

The pulmonary blood-gas barrier represents a remarkable feat of engineering. It achieves the exquisite thinness needed for gas exchange by diffusion, the strength to withstand the stresses and strains of repetitive and changing ventilation, and the ability to actively maintain itself under varied demands. Understanding the design principles of this barrier is essential to understanding a variety of lung diseases, and to successfully regenerating or artificially recapitulating the barrier ex vivo. Many classical studies helped to elucidate the unique structure and morphology of the mammalian blood-gas barrier, and ongoing investigations have helped to refine these descriptions and to understand the biological aspects of blood-gas barrier function and regulation. This article reviews the key features of the blood-gas barrier that enable achievement of the necessary design criteria and describes the mechanical environment to which the barrier is exposed. It then focuses on the biological and mechanical components of the barrier that preserve integrity during homeostasis, but which may be compromised in certain pathophysiological states, leading to disease. Finally, this article summarizes recent key advances in efforts to engineer the blood-gas barrier ex vivo, using the platforms of lung-on-a-chip and tissue-engineered whole lungs.

Introduction

The blood-gas barrier, otherwise known as the alveolar-capillary barrier, is the key functional element of the lung, serving as the site of oxygen and carbon dioxide exchange between the distal airspaces and the pulmonary vasculature. The past decade has seen the rise of two parallel strategies for the ex vivo engineering of this barrier: lung-on-a-chip, which provides a micro-scale platform for recapitulating key aspects of pulmonary biology for uses such as drug testing; and whole lung engineering, the goal of which is to grow whole organs to supplement the pool of organs available for lung transplantation. While the primary function of the alveolar-like barriers in these two experimental systems is ostensibly the same—gas exchange by diffusion—the design criteria for each are very different. While gas exchange may be accomplished simply by the separation of air and liquid media by a thin barrier permissive of diffusion, the in vivo environment of the native lung (and by extension that of a transplanted engineered lung), and the need to oxygenate an entire organism, impose further engineering constraints. The principle design criteria for recapitulating the gas exchange functions of the native lung are (i) blood-gas separation, (ii) vast alveolar and capillary surface areas, (iii) thinness to permit gas diffusion, (iv) strength to withstand repetitive and dynamic transpulmonary as well as transmural pressures, (v) compliance to enable effective ventilation and to maintain alveolar patency, and (vi) active maintenance of an air-liquid interface (Table 1).

Table 1.

Blood-Gas Barrier Design Criteria of the Native Lung

Blood-gas separation

Vast alveolar and capillary surface areas

Thinness compatible with gas diffusion

Strength to withstand repetitive inflation/deflation, and to receive the whole of the cardiac output

Compliance to enable effective ventilation, and to maintain alveolar patency

Active maintenance of an air-liquid interface

It is this complexity that we explore in this article since any effort at biological engineering of the blood-gas barrier must consider not only what the barrier looks like, and what it must accomplish via its structure, but also what it does, and how it is maintained. In the first section, we discuss the initial formation of the barrier during alveolarization, detail the structural features and cellular components comprising its mature form, and describe the micromechanical forces to which the barrier is subjected. In the second section, we explore the features of the alveolar-capillary membrane—both cellular and extracellular—that maintain barrier function in homeostasis and consider several disease states in which the protective mechanisms of the barrier are overcome. Finally, we review recent advances in blood-gas barrier engineering within the platforms of lung-on-a-chip and whole lung decellularization-recellularization.

Design Principles of the Blood-Gas Barrier

The first step in any endeavor to engineer the blood-gas barrier should be an examination of its structure and its component parts, to understand what the native barrier is. The dynamic nature of the lung, as a result of ventilation, also necessitates a discussion of the mechanical environment to which the barrier is subjected in vivo.

In this section, we will first discuss key features in the development of the mammalian blood-gas barrier, since the transformations that occur during alveolar development in utero and postnatally give insight into which features of this barrier are critically important for its function. We will then detail the morphology and morphometry of the mammalian blood-gas barrier, including its unique vascular bed and constituent cellular compartments, with an emphasis on the architectural features of the lung that enable it to achieve the extreme thinness, enormous surface area, and close blood-gas approximation compatible with providing blood oxygenation by diffusion. We will then discuss the surface and tissue forces that define the micromechanics of the lung parenchyma, and how these forces affect alveolar structure and gas exchange.

Development of the blood-gas barrier

During lung development, the blood-gas barrier undergoes a dramatic series of changes in tissue form and structure. While the bulk of blood-gas barrier maturation takes place after birth in most mammals, including humans (32, 275), the lungs in these species at birth are functionally mature (i.e., capable of gas exchange), but not structurally mature (32). This observation begs the question: what are the features distinguishing the functional, immature barrier, and what features are missing at the early stages of life? What are the changes that take place to achieve maturity of structure? Much of the information known about the morphologic and morphometric changes that occur during lung development we owe to studies performed in rat, since the key events and patterns in the process of rat lung maturation closely parallel those in human, though they occur on a different time scale (275, 276).

Fetal and postnatal lung development proceeds through four overlapping but sequential stages: the pseudoglandular stage [5–17 gestational weeks in human, embryonic days 15–19 (E15–19) in rat], the canalicular stage (16–26 gestational weeks in human, E19–E21 in rat), the saccular stage (24 gestational weeks to late fetal period in human, E21-postnatal day 4 (P4) in rat), and alveolarization (late fetal period to at least childhood in human, P4 to at least P21 in rat) (Figure 1) (35, 229, 242). Some authors have made the distinction of an additional stage overlapping with alveolarization: that of microvascular maturation (195, 229, 275). Note that there is still no scientific consensus as to the time-point marking the end of alveolarization and the onset of “normal” organ growth, that is, the growth of the lung proportional to growth of the organism, without addition of new alveoli. Dunnill, in the 1960s, concluded that alveolarization largely finishes by 8 years old in human (57), but Burri, Zeltner, and colleagues in the 1980s revised this estimate to suggest that the formation of new alveoli largely stops by 1 to 2 years of age (32, 275). More recently, however, Schittny et al., using high-resolution synchrotron radiation X-ray tomographic microscopy to study new septa and associated capillary morphology, proposed that alveolarization continues until about 18 years of age, with half of the alveolar septa formed between age 3 and 18. In fact, they claim that “new alveoli may be formed in principle at any time and at any location inside the lung parenchyma” (195). This statement has significant implications for any efforts at lung regeneration.

Figure 1.

Timeline of early lung development and associated milestones in blood-gas barrier maturation. Blue trapezoids illustrate the magnitudes of morphometric changes during alveolarization and microvascular maturation in the rat. E, embryonic day. P, postnatal day. Gest wks, gestational weeks.

While the bulk of the gas-exchange surface is formed during alveolarization and microvascular maturation, aspects of barrier function begin to be developed during the fetal period, well before the nascent lung must be capable of gas exchange (Figure 1). Briefly, during the pseudoglandular stage, the primitive airway and acinar tree are formed via epithelial branching morphogenesis. Even at this stage, there is evidence of both epithelial and capillary endothelial tight junctions (32, 197). During the canalicular stage, the epithelial tree and its associated vasculature expand. This expansion is particularly notable in the parenchymal regions, where continued development and growth of the capillary network brings it in closer association with the prospective alveolar epithelium, and where the first sites of a thin epithelial-endothelial barrier arise. Next, in the saccular stage, airspaces further enlarge through interstitial thinning. In addition, there is development and maturation of the alveolar epithelial type II cell (AEC2) surfactant synthesis and secretion machinery. By the end of the saccular stage, the branching structure of the lung has achieved its mature form; no new acini (the functional unit of the lung, consisting of the gas-exchanging structures distal to a terminal conducting bronchiole) will form (17). Also by this point, the lung has reached a stage of functional maturity that may be compatible with survival ex utero, based on the presence of airspace walls capable of gas exchange. However, there is a high incidence of infant respiratory distress syndrome (RDS, formerly called hyaline membrane disease) in premature infants who are born during the saccular stage of lung development. This makes clear the fact that mature and sufficient pulmonary surfactant, the lack of which is the primary defect in RDS (8, 45, 172), is a requisite for normal lung function (32, 242). Of course, all of these stages of lung development, until the moment of birth, occur in a fluid-filled organ.

What does the lung look like at full-term, at the moment that it is required to provide gas exchange across the new blood-gas barrier? The P1 neonatal rat lung (corresponding to the late fetal human lung) is devoid of true alveoli, instead consisting of large, smooth-walled saccules (Figure 2A). A double layer of capillaries weaves through the thick primary septa surrounded by a highly cellular interstitium that bridges the layers of epithelial air sac lining. Notably, the actual sites of gas diffusion consist of a mature-appearing air-blood barrier, with epithelium and endothelium merely separated by a fused basement membrane (BM) (34). Between P1 and P4, the lung airspaces expand together with the chest wall. But beginning at P4 (Figure 2B), which is the onset of alveolarization, the lung undergoes an “explosive” (34) increase in alveolar and capillary surface areas, as selective proliferation of endothelial cells and fibroblasts leads to the lifting off of new secondary septa (Figure 2C). The formation of secondary septa effectively subdivides existing airspaces into smaller, more numerous alveoli. This burst of alveolar formation is most notable through the end of the second postnatal week (Figure 2D). During the third postnatal week, secondary septation continues even as cellular proliferation slows, leading to alveolar wall thinning as alveolar surface area increases. This thinning is augmented by simultaneous remodeling of the microvasculature into a single-layered capillary mesh, which occurs via fusion of double capillaries, as well as by preferential growth of one of the two microvascular meshes over the other. Additional capillaries are also added within the existing network at this time by intussusceptional growth (39). Thus, by P21 in the rat (Figure 2E), the lung has achieved an appearance similar to that of the adult lung (Figure 2F). At P21, thin septa comprising a mostly acellular interstitium weave between a dense meshwork of capillaries that bulge into the alveolar spaces (34, 115). Notably, throughout these dramatic postnatal changes, the diffusing capacity of the lung is not impaired at any point (36).

Figure 2.

Hematoxylin and eosin (H&E) staining of native rat lung fixed by formalin inflation at postnatal days (A) 1, (B) 4, (C) 7, (D) 14, (E) 21, and (F) 60. Scale bars, 50 μm.

The result of these substantial restructuring processes is a maximization of blood-gas barrier surface area (both alveolar and capillary) with a simultaneous minimization of membrane thickness, thereby achieving a structure that is optimized for gas exchange by diffusion. Quantitatively, the effectiveness of airspace subdivision and microvascular remodeling is evidenced by the fact that from P4 to P21 in the rat, the alveolar and capillary surface areas increase by the 1.6^th and 1.7^th power of lung volume, respectively. This increase in surface area is much greater than would be expected for proportional growth, in which case surface area would increase by the 0.67^th power of lung volume. At the same time, the average thickness of the blood-gas barrier decreases by more than 50% (most notably due to interstitial thinning). The diffusing capacity of the lung D_L (see the section titled “Fick’s law modeling of gas exchange”) increases proportionally to body weight from P1 to P21, at which point the increase in D_L slows (Figure 1) (36). A necessary consequence of these changes is that the barrier between blood and air becomes increasingly fragile. To bolster the mechanics of the thin BM encircling each alveolus, the postnatal period is accompanied by a dramatic increase in the quantity of both elastin and collagen (155).

Structure of the blood-gas barrier

The primary function of the lung is to oxygenate the blood and to allow removal of carbon dioxide from the systemic circulation. To achieve this, the lung spreads the blood into a very fine layer of intermeshed capillaries within the alveolar walls, allowing extraordinarily close proximity between perfused blood and inspired gas. In the human, the airways develop into tightly packed alveoli approximately 200 μm in diameter (163) which collectively resemble a polyhedral foam (46, 77) (Figure 3A). The walls of this super-structure are intermeshed with a highly dense capillary bed containing blood flowing from the pulmonary arterioles to the pulmonary venules. This setup allows the compression of nearly 70 m² of gas-exchanging surface into a total volume of only 5 to 6 liters (70).

Figure 3.

Scales of blood-gas barrier organization. Top of each panel, native lung image. Bottom of each panel, schematic. Tops of panels: (A) Scanning electron micrograph of distal lung parenchyma in mouse. (B) Light micrograph of alveolar septa in a saline-filled rabbit lung. (C) Transmission electron micrograph of an alveolar capillary in monkey. (D) Enlarged detail by transmission electron micrograph of the gas exchange surface. Scale bars in lower panels representative of human lung. (A) Reused, with permission, from Bastacky J and Goerke J, 1992 (18), copyright the American Physiological Society; (B) Reused, with permission, from Gil J, et al., 1979 (80), copyright the American Physiological Society; (C,D) Reused, with permission, from Weibel ER, 1970 (246), with permission from Elsevier.

The air-contacting surface of alveolar tissue is entirely coated with an alveolar lining layer (ALL) consisting of a surfactant-rich surface film and a lower aqueous hypophase, which sits on top of the alveolar epithelium (19, 116). This combination of surfactant and water hypophase imparts a number of remarkable properties to lung tissue and is essential for barrier maintenance and proper lung function. In addition to the capillary bed, the alveolar septa separating individual alveoli contain extensive fibrous matrix components and interstitial cell types that impart the bulk of the tissue mechanical characteristics. Between the interstitium and the ALL resides a continuous epithelial layer made up of alveolar type I (AEC1) and AEC2 cells.

This spatial arrangement of interfacial components means that oxygen passing from the alveolar compartment to a red blood cell (RBC) must pass, at a minimum, through the ALL, an epithelial cell, some amount of extracellular matrix, an endothelial cell, blood plasma, and the RBC membrane (77, 81, 146) (Figure 4). Since capillaries almost always bulge to one side of the septal mid-plane (252) (Figure 3B), two forms of the blood-gas barrier exist in roughly equal measure: a “thin” version which contains only epithelium, BM, and endothelium, and a “thick” version which also includes fibrous matrix and supporting cells (Figures 3C and 3D) (77). Since this arrangement has been argued to allow a threefold reduction in the effective mean diffusion thickness of the barrier while retaining the same amount of structural support (244), it has been postulated that this arrangement, found also in the capillaries of fish gills and bird lungs, represents a balance between structural integrity and respiratory efficiency (136). The thinnest portion of the blood-gas barrier (in human) is estimated to be between 0.2 and 0.4 μm thick (77). The harmonic mean of barrier thickness, which takes into account both forms of the barrier and is considered most important for estimates of diffusion capacity (246, 252), is only 0.62 μm thick in the human (77). It is even thinner in ultra-high metabolism mammals such as the shrew, where it measures only 0.33 μm (78).

Figure 4.

Layers of the blood-gas barrier.

Fick’s law modeling of gas exchange

The blood-gas barrier must be both very thin and remarkably strong (136). The diffusion capacity of the lungs, which measures the ability for gas to be transferred across the barrier from air to RBC, is inversely proportional to the thickness of the blood-gas barrier as governed by Fick’s law for diffusion. Weibel developed landmark calculations for morphometric descriptions of total pulmonary diffusion capacity (246).

The total membrane diffusion capacity of the lung D_M (which excludes diffusion into the RBC itself), must take into account the gas conductance of the alveolar lining/surface layer (D_s), the conductance of the tissue layer (D_t), and the conductance of the plasma layer (D_p). Since these interfacial components of the barrier are essentially resistors laid in series, with resistance being the reciprocal of conductance, the relationship governing the total membrane conductance (D_M) is thus

$$\frac{1}{D_{\text{M}}}=\frac{1}{D_{\text{s}}}+\frac{1}{D_{\text{t}}}+\frac{1}{D_{\text{p}}}$$ (1)

Note that due to the thinness of the ALL [which adds only about 3% to the barrier thickness, at about 500Å (81)], the contribution of the lining layer may be considered negligible in normal homeostasis but may be nonnegligible in edematous states, when the thickness of the lining may increase. Weibel showed that the diffusion capacity of each of these layers is given by

$$D=\frac{\stackrel{.}{V_{o_2}}}{△P_{o_2}}=K\cdot S\cdot \frac{1}{{\tau }_h}$$ (2)

where V̇_O2 is the rate of flow of oxygen across the layer, ΔP_O2 is the partial pressure gradient for oxygen (O₂) across the layer, K is the permeation coefficient for oxygen in the material making up the layer, S is the surface area, and τ_h is the harmonic mean barrier thickness. τ_h may be calculated by the equation

$$\frac{1}{{\tau }_{\text{h}}}=\frac{1}{n}\sum _{i=1}^n\frac{1}{{\tau }_i}$$ (3)

where τ_i is the thickness of the i^th element of n total elements that make up the total lung barrier.

It is readily apparent that a greater harmonic mean thickness, or a decrease of the coefficient K, for any of the above layers, will decrease effective diffusion capacity. However, ultimately diffusion is limited by mechanical requirements: if a barrier is too thin, particularly on the structurally supportive collagen-dense “thick” side, then the alveolar structure would not be able to withstand relevant stresses, which would lead to mechanical failure and bleeding into the airways (136). It should also be noted that while membrane thickness is a major limiting factor for oxygen uptake, RBC resistance to oxygen diffusion is of a similar order of magnitude and must be taken into account for more accurate estimations (191, 251).

Cross-species analysis shows that D_M correlates directly with organism body weight, with larger mammals having lower oxygen conductance across the barrier (248). This correlates with the increased structural fortitude of lung tissue that is required as body weight increases and is physiologically acceptable due to the slower metabolism of larger mammals relative to their mass. Interestingly, barrier thickness does not appear to be a dynamically regulated variable in normal homeostasis: in studies of hypo- and hyperoxic lung development, the mean barrier thickness did not change even though the total lung diffusing capacity (D_L) shifted to compensate for the experimental conditions, increasing in the hypoxic environment and decreasing in the hyperoxic condition (33). In both cases, the shift in D_L was attributed to a change in lung volume without any statistically significant changes in parenchymal barrier thickness.

As this discussion should make clear, the ability of the alveolar blood-gas barrier to support gas diffusion is largely a function of its constituent materials and of its morphometry. Thus, any effort to engineer this barrier must take into account these properties of all barrier layers, as well as of the resulting composite membrane, in assessing its functional capacity for gas exchange.

Morphometry of the pulmonary vasculature

The pulmonary capillary system is unique within the body for its exposure to air and was extensively studied and modeled during the latter half of the 20^th century. The predominant model of this system, termed “sheet flow,” was pioneered in 1969 by Fung and Sobin (71). This model treats the pulmonary microvasculature as a network of tubular capillaries so densely interconnected that their fluid flow is best modeled as a continuous “sheet” of blood with interspersed arterial sources and venous sinks. This sheet is punctuated by “posts” made up by the interstitium woven in between the capillaries and is bounded on both sides by air-filled alveoli. The “sheet-flow” fluidic model has been validated experimentally numerous times, and we will cover the basic principles here. Understanding lung microvascular topology is essential for those seeking to understand disease states of alveolar permeability, and for those attempting to artificially engineer the pulmonary barrier.

Blood flowing through the pulmonary gas-exchanging circulation passes, in sequence, through the pulmonary arterial tree, the microvascular capillaries, and the pulmonary venous tree. The morphology of the pulmonary vascular tree has been intensely studied by Weibel (245, 249), Singhal et al. (201), Horsfield (104), Sobin et al. (204), Yen et al. (271, 272), Jiang et al. (107), and Huang et al. (107). Although the arteries and the veins share a general branching structure (107), the arteries are thicker walled (allowing them to withstand higher transmural pressures) and contain a larger number of smooth muscle cells (allowing them to regulate pulmonary blood flow). The arteries of the lung generally follow alongside the branching major airways, reduce to arterioles, and then release their blood into the pulmonary capillaries in the alveolar walls. The venules, conversely, remove blood from regions of alveolar tissue not directly supplied by arterioles. In a given tissue section, arterioles and venules generally cluster together, and the resulting pattern is beautifully documented in Sobin et al., showing isolated regions of alveoli closest to arterioles (“islands”) surrounded by a continuous swath of alveoli most closely associated with venules (a venous “lake”) (Figure 5A) (204). Although there is great regional variability (and interconnection of fluid flow paths), the average arteriole supplies blood to 24.5 alveoli, and the average venule receives blood from 17.8 alveoli (278). This leads to the conclusion that there are slightly more venules than arterioles, and reflects that a slightly larger area of tissue is composed of the venous “lake” than the arteriolar “islands.” Morphometric techniques estimate that the mean capillary length (RBC travel distance during gas exchange) is between 600 to 800 μm in dog and cat and 550 to 650 μm in rabbit, with a given flow path traversing approximately 5 to 7 alveoli (214). Later measurements refined the average capillary length in cat to 556 ± 286 μm (278).

Figure 5.

Alveolar-capillary morphology. (A) Arteriolar “islands” and venous “lakes” in cat lung. (B) Pulmonary capillary beds in frog lung (left) and in the alveolar walls of cat lung (right). AL, alveolar sac. l, line delineating interalveolar wall. P, Intercapillary connective tissue posts. (C) Schematic of capillary network between a pulmonary arteriole (left) and venule (right). (D) Schematic of the alveolar-capillary “sheet” with associated dimensions in cat. (E) Comparison of pulmonary capillary sheet width to the height of the Empire State Building. (A) Reprinted with slight modification from Sobin SS, et al. 1980 (204), with permission from Elsevier; (B, left) Reprinted from Maloney JE and Castle BL, 1969 (138), with permission from Elsevier; (B, right) Reprinted from Sobin SS, et al. 1980 (204), with permission from Elsevier.

The capillary bed connecting arterioles and venules is short and so densely interconnected as to make Poiseuillian flow modeling inappropriate (71, 251). A RBC entering the capillary bed from an arteriole has numerous pathways it might take to reach a venule, and the microvascular resistance in the lung is far less than if only a single path existed. Any given flow path from arteriole to venule contains many individual capillary segments (165). Interstitial “posts” interrupt the “sheet” of flowing blood and represent the negative space between capillaries (Figures 5B and 5C). This negative space is extremely minimal: blood-filled capillaries underlie approximately 91% of a given alveolar septum, leaving only 9% for interstitial space/tissue (207). From the perspective of an RBC, Fung visualized the interior of the capillary bed as the inside of an underground parking garage (70, 207), with large flat areas for blood flow punctuated by intermittent supporting pillars. A more alveolar-scale description might be a sheet of lace so finely interconnected that it comes to resemble a continuous fabric; the fibers are the capillaries while the gaps between are occupied by interstitial matrix. Both of these descriptions are two-dimensional (2-D) representations. In reality, this vascular sheet is contained within the three-dimensional (3-D) interalveolar walls and therefore, in mammals, has the macrostructure of a polyhedral foam made up of this fine lace-like structure. However, because of the scales of the sheet thickness and its effective length, the fluid mechanics are the same between the 2-D (flat) and 3-D (curved) formalizations.

Remarkably, the distance from arteriolar source to venous sink is only 556 μm (278), but the total alveolar area is 0.87 to 1.27 m² (Figure 5D) (270). Put another way, if unfurled and laid flat, the alveolar capillaries in a cat would be a narrow strip only 0.5 mm long (278) running from the arterioles to the venules. However, with all the capillaries in the cat lung laid next to each other, the 0.5 mm strip becomes approximately 800 m wide, or about as wide as the Empire State Building is tall (Figure 5E). Hence, the alveolar-capillary network represents an extraordinarily dense meshwork of very short capillaries.

Another key feature to understand is that the pulmonary capillary bed is both elastic and collapsible (206). When the pressure in a capillary exceeds the pressure in the neighboring alveoli (i.e., the capillary transmural pressure is positive), then the capillary is patent and the thickness of the capillary sheet at that location increases proportional to the transmural pressure. If the pressure in the alveoli is sufficiently larger than the pressure in the capillary sheet (i.e., transmural pressure is significantly negative), then the capillary collapses and no flow traverses that portion of the interalveolar septum. We say significantly because there is a narrow (~1 cmH₂O transmural pressure) range in which the alveolar pressure can exceed the capillary pressure while still maintaining capillary flow (72). In this case of slightly negative transmural capillary pressures, the downstream end of the capillary becomes a sluicing gate that is held open by capillary sheet wall tension (70). The net effect of these properties is that the entire capillary bed is rarely perfused all at once because during normal life some of the alveoli are pressurized to a greater extent than their surrounding capillary beds.

In the normoxic state, only a portion of the capillary bed is recruited and contains active blood flow; the advent of a hypoxic state in the organism causes increased recruitment of capillaries (235). In the physiologic setting, this effect is minimally dependent on an increase in cardiac output or venous resistance, but rather is dominated by hypoxia-mediated vasoconstriction of the pulmonary arteries. [Note that ex vivo, abrupt changes in venous backpressure may also increase local capillary recruitment (165).] Normally, blood is preferentially directed to the more dependent regions of the lung due to the effects of gravity, with the result that the upper regions of the lung contain relatively more nonperfused, collapsed capillaries. In hypoxia, generalized pulmonary artery vasoconstriction, and the resulting increase in mean pulmonary artery pressure, counteracts some of the effects of gravity, leading to an overall redistribution of blood toward those nondependent regions that are normally poorly perfused (236). The resulting increase in perfused alveolar surface area leads to an increase in lung diffusing capacity [Eq. (2)]. This is in contrast to local hypoxic vasoconstriction, which acts to redirect blood flow from a poorly ventilated region of the lung to minimize shunt fraction (153).

Thus, the elasticity and collapsibility of the pulmonary capillaries are key to the proper functioning of the lung, as the organ can adapt to radical changes in metabolic demand for gas exchange simply through increased recruitment of reserve capillaries.

Morphometry of interstitium and epithelia

As capillaries occupy some 90% of the alveolar septa, there is very little room for both epithelia and supporting mesenchymal/stromal cells. Interstitial cells and supporting matrix are confined to the “thick” side of the capillary membrane and to the “posts” in between capillaries. The epithelia, in contrast, exist on the air-contacting surface of the septal tissue and completely isolate the capillary bed and interstitium from the lining layer of the alveolus. However, the large AEC1s that cover the majority of the alveolar membrane surface are by no means flat, as they have often been depicted in textbooks.

It was realized in the late 19^th century by Elenz (62) and Koelliker (118) that there appeared to be large swaths of alveolar surface covered by nonnucleated cytoplasm, a conclusion inconsistent with modern understandings of biology. Early electron microscopy refuted this but found there indeed to be a disproportionately low number of AEC1 nuclei compared to their cytoplasmic coverage (131). Later measurements of cell number and size in human lungs (51) showed that a single AEC1 has a mean thickness of only 0.36 μm but a total cellular surface area of nearly 5100 μm², an astonishing value not found in any other tissue (250). To put this another way, AEC1s make up only 8.3% of the cells in the alveolus by nuclear counts, but they cover 93% of the surface of the alveolar wall (51). In a remarkable paper from 2015, Weibel synthesizes this and other information (247) to present a comprehensive view of AEC1s as topologically complex cells interwoven with the capillary bed and able to cover gas-exchange surface area on both sides of the alveolar septum (Figures 6A and 6B) (250). Rather than simple squamous cells, AEC1s are huge branching cells with multiple squamous plates connected by inter-capillary stalks (Figure 6C); a single AEC1 may crisscross the capillary mesh multiple times and form tight junctions with itself to wrap individual segments of the capillary network (250). This topology is particularly pronounced in high metabolism mammals such as the Etruscan shrew, in which lung parenchyma looks like floating capillaries wrapped only by thin interconnected layers of AEC1 cytoplasm which completely enwrap individual capillary segments. Such an arrangement allows multiple portions of isolated flat epithelial cytoplasm to be served by a single central nucleus, facilitates nutrient transfer, and maximizes the surface area able to be covered by a single epithelial cell. We emphasize this point because this visualization runs counter to many depictions in medical and engineering textbooks, and it is essential to understanding the true cellular structure of the blood-gas barrier.

Figure 6.

Alveolar epithelial type I cell morphology. (A) Scanning electron micrograph of the alveolar surface of a human lung, with an alveolar epithelial type I cell (AEC1) highlighted in yellow. (B) Drawing of an AEC1 (in yellow, with inner and outer cell membranes in red and green, respectively) branching to either side of an alveolar septum. (C) Alternative view of the AEC1 in (B) illustrating the cell’s structure of intercapillary stalks and multiple cytoplasmic plates on both sides of the septum. Ep1, type I epithelial cell, Ep2, type II epithelial cell. Reprinted from Weibel ER, 2015 (250) with permission of the American Thoracic Society. Copyright © 2019 American Thoracic Society. The American Journal of Respiratory and Critical Care Medicine is an official journal of the American Thoracic Society.

Mechanical environment of the blood-gas barrier

The extreme thinness of the mammalian blood-gas barrier, needed for efficient gas exchange, is at odds with the requirement that it must withstand the cyclic mechanical stresses and strains of ventilation. The average human lung will undergo more than 20,000 respiratory cycles just over the course of 1 day, with volume changes of up to 150%. Throughout, the blood-gas barrier must not only preserve the ever-critical separation between the air and blood compartments, but also must maintain efficient gas exchange under the diverse mechanical conditions present within the same organ. Tissue architecture and structure change temporally over the course of the breathing cycle, while large variations in perfusion and ventilation exist spatially throughout the organ.

It is worth here recalling the structure of the lung parenchyma. The alveoli are not, as is often schematized, like a bundle of small balloons, or a bunch of grapes, both of which imply a collection of small spheres connected by cylindrical airway openings (10, 73, 180). The implication of those models is that the alveoli are inherently unstable. By application of Laplace’s Law for a sphere,

$$P=\frac{2T}{r}$$ (4)

where P is the transpulmonary pressure, T is the surface tension, and r is the radius, and assuming constant surface tension at a given lung volume (199), smaller alveoli in this model would have a higher transpulmonary pressure than would larger alveoli. The consequence of this would be a tendency for smaller alveoli to collapse, with their contents flowing into larger alveoli (180). However, rather than independent air sacs, the alveoli better approximate a froth (180) or a foam (264), with a semi-interdependent network of individual polyhedral alveoli separated by nearly planar shared walls (Figure 3A), with the air sacs interconnected by pores of Kohn (49, 80, 145, 148). In this arrangement, there can be no pressure differential from one alveolus to the next and, in addition, the shared septal tissue serves as scaffolding that prevents localized collapse (145). Indeed, alveolar collapse is generally prevented, except perhaps at very low lung volumes when folds and pleats in septal tissue are present, thereby creating structural instability (10, 80). In addition to the septal walls, the airspaces are also buttressed by two other interconnected connective tissue structures: (i) the peripheral connective tissue system, which consists of the interseg-mental and interlobular septa extending from the pleura to the parenchyma; and (ii) the major axial connective tissue, which surrounds the airways and arteries and forms the strong, fibrous alveolar entrance rings of the alveolar ducts (264). Given all of this, it should not be surprising that the mechanics that dictate alveolar inflation and deflation, and which maintain distal airspace patency, are more complex than those governing a group of balloons in parallel.

Given this structural complexity, it is clear that there is no simple equation governing the lung surface area-volume relationship (80). Multiple alveolar geometries may also exist at a given volume, depending on the prior history of lung inflation and deflation (11, 230). This is exemplified by the prominent hysteresis of the lung pressure-volume curve (Figure 7), which is mirrored in the pressure-volume curve of a single alveolus (148, 173). However, when considering gas exchange across the blood-gas barrier, it is the micromechanical forces acting upon a single alveolus that are most relevant, rather than the gross mechanical forces that drive whole lung ventilation (Figure 8). We can generally think of surface forces and tissue forces as the primary micromechanical factors affecting alveolar structure. Surface forces in the alveolus result from the presence of the ALI lining each alveolus. Alveolar tissue forces result from lung inflation and the resultant stretching of parenchymal tissue. Circumferential capillary strain due to intravascular pressures may also be transmitted to the alveolar wall and contribute to tissue forces (261). The interrelated effects of these forces together determine the range of alveolar architectures.

Figure 7.

Lung pressure-volume curve.

Figure 8.

Schematic of micromechanical forces acting upon the alveolar septum.

Surface and tissue forces

Much of our understanding of the relative roles of surface and tissue forces in determining alveolar geometry has come from studies comparing the structure and behavior of airinflated, saline-inflated, and/or detergent-rinsed lungs. Saline inflation effectively brings the surface tension within the alveolus to zero by abolishing the air-liquid interface in the air sac. In contrast, detergent rinsing removes amphiphilic surfactant molecules that comprise the surface-active alveolar lining, resulting in a high surface tension (approximately 20 dynes/cm (20 mN/m) in one study) and a tendency for alveolar collapse (9). Therefore, both liquid inflation and detergent rinsing substantially alter the mechanics of the alveolus relative to that in normal life.

The existence of surfactant was hypothesized as early as the 1950s when Pattle noticed the remarkable stability of bubbles that were derived from lung fluid. He then surmised that the lung must contain a surface-active lining with zero surface tension (171). Since then, numerous investigators have experimentally confirmed his observations. Both indirect methods—via surface tension measurements of lung extracts using Langmuir-Adam or Wilhelmy film balances, or via calculations of surface area-surface tension relationships from pressure-volume loops—as well as direct methods, via micropipette, have demonstrated that the surface tension within the alveolus does indeed drop to values below 10 dynes/cm at functional residual capacity (FRC). [FRC is equivalent to the resting volume of the lung after tidal volume exhalation and is approximately equal to 40% of total lung capacity (TLC)]. In fact, the more accurate micropipette measurements suggest that the surface tension may become near-negligible (≤1 dyne/cm) at FRC. On the other end of the spectrum, as lung volumes approach TLC, surface tension approaches values exceeding 30 dynes/cm. However, during the volumes of normal tidal breathing, surface tension is likely confined to less than approximately 20 dynes/cm (8, 11, 31, 46, 103, 117, 199, 200). As a comparison, the surface tension of blood plasma may reach 50 to 60 dynes/cm (8, 31).

Surface tension arises due to relatively stronger liquid-liquid than liquid-air attractive forces. This force imbalance results in an affinity for liquid molecules to join the bulk fluid volume and to pull away from an air-contacting surface. Thermodynamically, the result is to minimize the interfacial surface between (hydrophilic) liquid and air as much as possible (262). In the lung, surface tension at the hypophaseair interface acts tangentially to the plane of the alveolar wall, with a resultant net inward-directed force—tending to collapse the alveoli if surface tension is unopposed (Figure 8). Due to its dependency on surface area, surface tension increases as lung inflation increases, and then decreases again with decreasing lung volume (11, 31, 46, 103, 172, 199, 200). Surfactant, a complex mixture of lipids and proteins secreted by AEC2s, acts to decrease this surface tension by adsorbing at the interface of gas and liquid across the lining of the alveolus. Surfactant is able to perform this duty because of its bi-functional composition: hydrophilic portions of the mixture are water-facing, and hydrophobic portions of the mixture are air-facing. The presence of surfactant at the interfacial layer within the alveolus has the effect of decreasing the density of water molecules at the interface with air (91). Because the number of water molecules in contact with air is lower, the surface tension is lower.

Importantly, the surface tension within the lung varies at a given volume depending on whether that volume is reached by inflation or by deflation. This is due to the mechanical consequences of exhalation on the surface-active layer. As lung volume decreases, the surfactant layer becomes compressed, causing the nonsaturated phospholipid and protein constituents to be squeezed out of the mixture, and resulting in an increased concentration of dipalmitoylphosphatidylcholine (DPPC). DPPC is a phospholipid that is particularly important for reduction of surface tension to near-negligible levels (91). Thus, surface tension is lower on the deflation limb than on the inflation limb of the hysteretic pressure-volume curve at the same lung volume (Figure 7). Accordingly, alveolar surface areas are greater on the deflation limb than on the inflation limb (11). Surface tension contributes to lung hysteresis in another manner as well, by causing tissue distortion that in turn increases elastic recoil (10, 11, 46, 263). In contrast, in the absence of surface tension, as in a saline-filled lung, hysteresis is nearly absent from the lung pressure-volume curve, and the surface area at a given lung volume is largely independent of whether that volume was reached by inflation or deflation (Figure 7).

An important exception to the inward-directed vector of surface forces due to surface tension arises from the bulging of various structures—capillaries (11, 80, 230), cuboidal AEC2s (19), and alveolar macrophages (1, 10)—into the alveolar lumen, beneath a lining of hypophase. These inward-protruding vessels and cells create convex curvatures along the alveolar wall. This, in turn, means that the surface forces acting on these structures are compressive (Figure 8). This is illustrated by the flattening of capillaries at higher lung volumes that occurs exclusively in air-, but not saline-filled, lungs (recall that saline-filled lungs have no ALI and thus no surface tension) (11, 80, 86, 230).

Surface forces may also be significant enough to compress alveolar cells. At lung volumes near TLC, Ravasio and colleagues estimated that the net surface force may reach 3 kPa, which is on the order of the elastic modulus of AEC2 subcellular structures (188), and presumably, those of other cell types as well. Accordingly, at high lung volumes, alveolar macrophages can be seen flattened against the alveolar epithelium beneath the hypophase (1, 10). Finally, further evidence that interfacial (surface) forces are relevant at the cellular, and not just tissue, level, is corroborated by the fact that contact of AEC2s with an ALI also leads to significant changes in gene expression (100).

Tissue forces arise from literal stretching of the alveolar wall (Figure 8) and thus, like surface forces, also increase with increasing lung volume. However, expansion of the lung is not merely a consequence of tissue stretch but also of septal wall unfolding, which is, in fact, the primary means of surface area enlargement at lower volumes (9, 230). Recruitment of atelectatic alveoli—those where the septal wall may be folded over to the point where alveolar air volume is obliterated—plays a role as well (9). During normal ventilatory movements, the most significant increase in alveolar volumes, per change in transpulmonary pressure, occurs near FRC (148). The most significant increase in the epithelial BM surface area, on the other hand—due to stretching, and not just unfolding, of the BM—occurs near TLC (230). Because the lung has nonlinear mechanics and because the history of lung stretch impacts minute-to-minute volumes and pressures, estimates for the magnitude of this surface area increase vary depending on experimental protocol but have been calculated from the data of several groups to be 28% to 34% from resting volume to TLC [from calculations in Tschumperlin and Margulies (230) using theirs and Mercer and Crapo (147) data, and from equation 7 in (230) using data from Perlman and Bhattacharya (173) for the purposes of this paper].

As a general principle, pulmonary parenchymal geometry is determined primarily by surface forces at the low volumes near FRC. In contrast, at high lung volumes, parenchymal geometry is determined by tissue forces, particularly as volumes approach TLC (9, 11, 46, 148). At low lung volumes, tissue stretch is very low, and the luminal surface area of the lung is less than the surface area of the epithelial BM, due to folds that form in the septal tissue (Figure 9) (11, 230). As lung volume increases toward TLC, the alveolar walls become thinner and appear stretched, with fewer (though not absent) pleats of the septal walls. At the same time, the alveolar surface becomes increasingly smooth, in part due to the flattening of capillaries, which bulge into the airspaces at lower lung volumes but which are reduced to slits at higher lung volumes (Figures 10A and 10B) (11, 80, 86, 230). Thus, in addition to the compressive force components acting on structures that bulge into the alveolus, an important consequence of surface forces is a “molding” of the alveolar wall (10, 11), manifested by redundant tissue rearrangement at low lung volume and decreased capillary caliber at high lung volume. Notably, in the absence of these surface forces, as in a saline-filled lung, the septa do not pleat at low lungs volumes, and the alveolar surfaces remain bumpy even at high volumes due to persistently bulging capillaries (Figures 10C and 10D) (11, 80).

Figure 9.

Alveolar septal pleats. An alveolar septum is shown by transmission electron micrograph and by line drawing. Arrowheads, alveolar luminal surface. Dotted line, pleated epithelial basement membrane. C, capillaries. SLL, surface lining layer. EPII, alveolar epithelial type II cell. M, alveolar macrophage. Modified from image originally printed in Bachofen H, et al., 1987 (11), copyright the American Physiological Society.

Figure 10.

Morphological differences of air- and saline-filled lungs at different volumes. (A,B) Scanning electron micrographs of air-filled lungs fixed by vascular perfusion at 40% and 80% TLC, respectively. (C,D) Saline-filled lungs fixed by vascular perfusion at 40% and 80% TLC, respectively. Originally printed in Gil J, et al. 1979 (80), copyright the American Physiological Society.

It is noteworthy from this discussion that even at high lung volumes, not all of the septal tissue is stretched, as evidenced by pleats of BM on the thin side of alveolar walls. In other words, even at high lung volumes, not all components of the parenchyma are stretched equally. Tissue stress and strain may fall disproportionately on supporting structures of the peripheral and axial connective tissue, as well as on the fibrous “thick” portions of the alveolar septa, thereby protecting the relatively delicate “thin” gas-exchanging portions of the alveolar walls (10, 11, 264).

An additional force that has so far been neglected in this discussion is that of circumferential wall tension within the alveolar capillaries (Figure 8). This tension arises due to capillary transmural pressure, a relationship that can be described by Laplace’s equation for a cylinder:

$$T=\mathrm{Pr}$$ (5)

where T is the circumferential wall tension, P is the transmural pressure, and r is the radius of the vessel (261). Alveolar capillaries are unique in that they are essentially unsupported (206). These vessels sit within the alveolar septum with two thin cell processes and a fused BM on one side, and, on the “thick” side, interstitial cells and extracellular matrix (47). As such, although the circumferential tension arises directly in the capillary wall, it may be transmitted to surrounding constituents of the blood-gas barrier, including the interstitium, epithelial and endothelial BMs, and alveolar epithelium. While the circumferential tension is small even at pressures as high as 50 cmH₂O (approximately 25 dyn/cm for a 10 μm-diameter vessel), the resultant wall stress may achieve “astonishingly high” levels in the thin part of the blood-gas barrier. Circumferential wall stress S is given by

$$S=\frac{\mathrm{Pr}}{t}$$ (6)

where t is the thickness of the wall. For a thin portion of the blood-gas barrier that is only 0.3 μm thick (77), a wall tension of 25 dyn/cm translates into a wall stress of approximately8.2 × 10⁵ dyn/cm², comparable to the wall stress in the human aorta (261).

Zonation of micromechanics

In addition to the dynamic changes in blood-gas barrier architecture that occur over the course of the ventilatory cycle, there also exists significant spatial heterogeneity of both blood flow and ventilation. Blood flow in the lung is far from uniform (129), a direct result of hydrostatic gradients in the pulmonary arterial and venous circulation (86, 114). The elasticity of the pulmonary capillary bed means that the recruitment of a given capillary flow path is significantly influenced by local capillary-alveolar (transmural) pressure. Permutt et al. (174) and West et al. (257) first described in the 1960s a model by which the distribution of flow in the pulmonary circulation may be explained by the combined effects of the pulmonary arterial, pulmonary venous, and alveolar pressures acting on the capillaries. This model is generally discussed in terms of three “Zones” (257) (Figure 11A). When the capillary is at a lower pressure than is the alveolus, then the capillary is collapsed, contains no blood flow, and is said to exist in Zone 1. When the arterial pressure is greater than the alveolar pressure, but the venous pressure is very close to (or below) the alveolar pressure, then there is blood flow in the capillary but the outflow path may be narrowed, and the capillary is in Zone 2. When the capillary pressure throughout its length exceeds the alveolar pressure, then there is blood flow continuously and the capillary is in Zone 3 (Figures 11B and 11C). In-depth discussions and modeling of this concept may also be found in work by Fung (70) and Glazier et al. (86).

Figure 11.

Zonation of pulmonary blood flow. (A) Location of zones, with Zone 1 near the apex of the lung and Zone 3 near the base. (B) Relative pressures, (C) capillary profiles, and (D) capillary cross-sections in Zones 1 to 3. P_A, alveolar pressure. P_a, arterial pressure. P_v, venous pressure.

During normal homeostasis, capillaries in the lung exist in all three zones. In a standing person, the apex of the lung is generally under Zone 1 conditions, with very few capillaries perfused. The base of the lung is generally in Zone 3, while a mediastinal region above the left atrium is generally in Zone 2. The distribution of blood flow from Zone 1 to Zone 3 is thus gravity-dependent (70), with approximately ninefold higher blood flow at the base compared to the apex of the lung (256). However, it should be noted that gravity is not the only, or perhaps even primary, factor affecting perfusion heterogeneity in the lung (87, 114). The anatomical structure of the pulmonary vascular tree may affect the distribution of blood flow independent of gravity (87). In addition, the elasticity of the alveolar capillaries, as described in the sheet-flow model (71), is such that capillary width is linearly dependent on alveolar pressure (86). Higher alveolar pressures flatten the capillary into an oval or a slit, rather than a circular cross-section (Figure 11D). Translated to the sheet flow model, higher alveolar pressures cause a decrease in the width of the sheet of blood flow. Accordingly, in dog, the observed capillary widths at 10 cmH₂O alveolar pressure range from approximately 2 μm in Zone 2 to approximately 13 μm in Zone 3, with a mean capillary width of approximately 6.5 μm(86), which is large enough to permit easy flow of RBCs. Thus, capillary diameter, along with capillary recruitment and the number of RBCs per unit septal tissue, are directly correlated with capillary hydrostatic pressure and increase as you go down the lung. There is a limit to capillary elasticity however, as evidenced by the achievement of a maximum capillary distension at a capillary pressure of approximately 50 cmH₂O (86).

Interestingly, under Zone 1 conditions in which flow is supposedly zero, approximately 30% of the septal corners contain open vessels (86). In contrast, septal vessels in this zone are collapsed (124). This difference has been attributed to differential stresses acting upon these two types of vessels. Surface forces create a more negative interstitial pressure in the corners of alveoli that tends to increase the transmural pressure of corner vessels relative to vessels in the middle of the septa. This increase in transmural pressure, in turn, means that corner vessels are less likely to collapse, compared to septal vessels (120, 124).

Like pulmonary perfusion, ventilation, defined as a change in lung volume over resting volume, also increases going from the top to the bottom of the lung. However, this change is not as significant as that for perfusion; ventilation at the base of the lung is approximately 1.5 times higher than at the apex (16, 256). This gradient of ventilation is due to gravity as well as to the mechanical properties of the lung. Due to the weight of the lung within the chest, the intrapleural pressure is greater (i.e., less negative) closer to the diaphragm. Consequently, the base of the lung is at a smaller resting volume than is the apex. However, the base of the lung experiences a relatively larger increase in volume for a given change in transpulmonary pressure during breathing, as evidenced by the steeper (more compliant) shape of the lung pressure-volume curve at lower volumes (Figure 7). Thus the base of the lung, compared to the apex, starts relatively underinflated and experiences larger changes in volume during breathing, corresponding to a higher level of ventilation (262).

A major effect of gravity-dependent perfusion variability in the lung is a spatial and positional dependence of the alveolar ventilation/perfusion ratio (V/Q ratio) (256). The ratio of recruited alveolar-capillary density to alveolar surface area increases significantly from apex to base (38), which allows substantially more gas exchange to take place in alveoli with highly recruited capillary beds. Steady-state alveolar oxygen tension changes by more than 40 mmHg from apex to base, with O₂ uptake per unit lung increasing eightfold going from the top to the bottom of the lung in a standing human (254). It should be noted that this change in gas tension has a negligible effect on overall gas exchange, due to the shape of the oxygen dissociation curve and the fact that hemoglobin is largely saturated above a pO₂ of 90 mmHg (254).

Mechanical forces due to zonation, however, have a nonnegligible effect on membrane diffusion capacity. Bachofen et al. demonstrated that the increased overall diffusion capacity (D_L) of Zone 3 lung tissue as compared to Zone 2 tissue is a consequence of (i) capillary recruitment, (ii) blood volume increase, and (iii) an increase in the membrane diffusing capacity (D_M) in Zone 3 due to barrier thinning (12). The membrane diffusing capacity increases with greater capillary distension because the capillary walls stretch out to their full extent, bulging into the alveolar chamber. The net result is a very thin blood-gas barrier with a high diffusing capacity. Zone 3 tissue has the highest static blood/gas ratio, the largest perfusion/ventilation ratio, the highest blood flow rates, the lowest oxygen tension, and the thinnest barrier. These combined effects generate the greatest amount of total gas exchange per volume of lung tissue within Zone 3 areas of the organ.

Resilience of the Barrier: Homeostasis and Disease

During development, the blood-gas barrier undergoes a substantial restructuring designed to maximize surface area while simultaneously minimizing thickness—thereby achieving a structure optimized for gas exchange by diffusion. A necessary consequence of these changes is that the separation between blood and air—at its thinnest, a mere 0.3 μm of cell processes and BM—is inherently fragile, though as we will discuss, at the same time it is impressively robust.

In this section, we will consider the biological and mechanical resilience of the blood-gas barrier: the cellular and structural features that maintain its integrity and ability for gas exchange in homeostasis. We will also consider three pathological settings—acute respiratory distress syndrome (ARDS), capillary stress failure, and ventilator-associated lung injury (VALI)—as examples in which these normally resilient features of the barrier are overcome. The case of ARDS, and the devastating failure of biological regulation of the barrier that accompanies and contributes to this disease, is particularly illustrative of the complexity of design principles that must be considered in any effort to engineer the blood-gas barrier in vitro or ex vivo.

Biological resilience of the blood-gas barrier

A remarkable feature of the alveolar blood-gas barrier is that it is neither purely structural in nature, nor fixed in its barrier properties, but rather is continuously maintained and regulated by the cells that comprise its primary components, in response to changing biological and mechanical conditions. In particular, the alveolar epithelium and capillary endothelium determine the permeability of the barrier to the passage of fluid and solutes, and the epithelium plays additional critical roles in clearing excess fluid from the airspaces, and in secreting surfactant. All of these functions are required for the successful maintenance of a (relatively) dry alveolar surface. Their derangement manifests first as interstitial and ultimately as alveolar edema, which in its severest forms may extinguish gas exchange and cause alveolar collapse and hypoxia.

In this section, we will first describe the manifestations of a failure of blood-gas separation—edema—as well as its general causes. We will then consider in turn the design features of the blood-gas barrier—both structural and cellular—that protect against edema to maintain homeostasis. Finally, we will discuss the consequences of loss of barrier integrity when these homeostatic mechanisms fail or are overcome, as in the example of ARDS.

Pulmonary edema

Pulmonary edema, or the accumulation of extravascular water or plasma in the pulmonary interstitial or alveolar spaces, is generally categorized into either hydrostatic or high permeability types (66, 213). However, more broadly, edema in the lung may arise in any setting in which the balance between fluid efflux from and return to the vasculature, whether by lymph flow or by alveolar fluid clearance (AFC), is disrupted.

Historically, the “hydrostatic” type of pulmonary edema has been discussed in terms of an imbalance in Starling forces. “Starling forces” are those that determine fluid movement across capillary walls—capillary and tissue hydrostatic and colloid osmotic (oncotic) pressures—with their precise relation to fluid flux given by Starling’s law (66, 223). For a pulmonary capillary, Starling’s law may be written as:

$$J_{\text{v}}=L_{\text{p}}\cdot S\left[P_{\text{c}}-P_{\text{is}}-\sigma \left({\pi }_{\text{c}}-{\pi }_{\text{is}}\right)\right]$$ (7)

where J_v is the volume flux of fluid across the capillary wall, with positive values indicating filtration out of the microvessel, and negative values meaning vascular reabsorption; L_p is the hydraulic conductivity; and S the surface area of the capillary membrane. The product of these last two variables is K_fc, which is the filtration coefficient of the capillary wall. P_c and P_is are the hydrostatic pressures within the capillary and interstitial space, respectively; σ is the osmotic reflection coefficient; and π_c and π_is are the capillary and interstitial oncotic pressures, respectively (223). In the normal lung, P_c, P_is, and π_is all favor fluid transudation out of the microvasculature, while π_c opposes filtration (Figure 12A) (45, 66). Note that P_is, the pressure in the interstitial space in the lung, is typically negative throughout the organ (95), with a typical value of approximately −10 cmH₂O in both the interstitium of the gas-exchanging units as well as in the peri-adventitial connective tissue surrounding the small vessels and airways (151). [This result, determined by micropuncture measurements in situ, contradicts the previously held concept that the peri-adventitial tissue is at a more negative pressure than the septal interstitium, thereby generating a pressure gradient from the alveolar septa toward these sites and by extension, the lymphatic vessels to which they connect (66, 223). However, some have also pointed out that these micropuncture measurements may overestimate the magnitude of the interstitial pressure (60). Nevertheless, that the interstitial pressure is negative is not disputed.] Notably, the net lumen-directed surface tension within the ALL contributes to the negative interstitial pressure surrounding the alveolar capillaries (3, 29, 45, 172). Critically, this driving force is reduced (made less negative) by the presence of surfactant (see more in the discussion of “Surfactant”).

Figure 12.

Starling’s law and alveolar protective mechanisms against edema. (A) Schematic of the alveolar septum in homeostasis. Block arrows at left depict typical relative magnitudes of the Starling forces P_c, P_is, π_c, and π_is across the alveolar-capillary wall in homeostasis. Dashed arrow at right depicts small amount of lymph flow from the perivascular spaces. (B) Altered Starling forces in the setting of increased hydrostatic pressure P_c, and the structural and biological protective mechanisms that may limit edema.

The net balance of the Starling forces in the steady state is considered to be close to equilibrium. The forces favoring filtration out of the capillaries—the capillary hydrostatic pressure, negative interstitial pressure, and interstitial oncotic pressure—nearly balance the single force favoring reabsorption—the capillary oncotic pressure, which is relatively large due to the presence of plasma proteins. The result of this balance is that in homeostasis the amount of microvascular filtration is small (45, 66, 223), particularly considering the enormous pulmonary capillary surface area (60). In addition, fluid production may, in fact, be limited to only a small portion of the septal capillaries (24).

Importantly, vascular filtration that may contribute to pulmonary edema occurs not only in the capillaries (where Starling forces act) but also in the macrovasculature—the pulmonary arteries and veins. It is important to note that the pulmonary capillaries are particularly restrictive in their permeability, both compared to systemic capillary beds and compared to extra-alveolar vessels, at baseline and in the setting of inflammation (60, 162). The septal capillaries appear to be so impermeable to solutes that the protective role of plasma proteins in generating oncotic pressure may be relatively less than in other vessels (60). This highlights one important functional difference between pulmonary microvessels and those in the rest of the body. In contrast, the larger arteries and veins in the lung are relatively leakier, such that in some settings of edema, fluid extravasation may be limited to extra-alveolar connective tissue. The radiological correlate to this phenomenon is the perivascular cuffing that can be seen on chest imaging in patients with pressure overload. Such edema comparatively spares the gas-exchange surfaces (60, 162). While this fluid accumulation thus does not directly impair gas diffusion, perivascular fluid in itself has been shown to decrease lung compliance (132) and may induce dyspnea, or feelings of shortness of breath (60).

There exist at least four additional mechanisms within the lung to limit alveolar edema (Figure 12B). To decrease edema that may result from an increase in capillary hydrostatic pressure, protective mechanisms include (i) a decrease in interstitial oncotic pressure, (ii) an increase in interstitial fluid (hydrostatic) pressure, (iii) an increase in lymph flow (223), and (iv) AFC (24, 60). It is the first of these, what Clements termed “osmotic feedback”—whereby dilution of solutes within the interstitial space, caused by capillary leak of ultrafiltrate, promotes an osmotic gradient favoring reabsorption—which has been shown to be particularly important in the lung (45, 223). In addition, the very low compliance of the septal interstitial space at low tissue volume means that even a small increase in septal interstitial fluid volume may lead to a significant increase in P_is [up to several cmH₂O greater than atmospheric, even in the setting of mild edema (47, 150)]. In addition to reducing the net force favoring microvascular filtration, the increase in interstitial pressure also creates a pressure gradient to drive excess fluid toward the higher-compliance peri-adventitial spaces and out of the peri-alveolar space (47, 66, 144, 150, 223). While lymphatic vessels are absent from the alveoli (60, 221), they do drain the perivascular and peribronchiolar spaces. Lymph flow may increase up to 25-fold in the setting of increased vascular pressure in a high-permeability injury model (192). For fluid within the alveoli themselves, vectorial ion transport by the alveolar epithelium can also drive fluid reabsorption (see more in the discussion of “Alveolar fluid clearance”).

In settings where these protective mechanisms are overcome, the initial site of excess fluid accumulation from the capillaries is within the septal interstitium, predominately within the thick portions of the alveolar wall (30, 66, 221, 231). At the very early stages, this thickening may not be detectable even by light microscopy (221). However, the thin gas-exchanging portions of the septa may be affected by even very mild edema, as has been demonstrated in a recent study involving quantification of the pattern of septal thickening. Conforti et al. fixed rabbit lungs in situ that had developed “mild” pulmonary edema in the setting of slow intravenous saline loading, and observed an approximately 25% increase (average 0.1 μm) in the thickness of the thin, and 35% increase (0.7 μm) in the thick, septal portions.

That this fluid accumulation may inhibit gas exchange is evidenced by the significant increase (approximately 50%) in the harmonic mean thickness of the alveolar septa. By Fick’s Law [Eq. (2)], an increase in septal thickness of this magnitude should cause a proportionate decrease in the rate of oxygen diffusion (47). With sufficient tissue swelling, fluid may next accumulate in the connective tissue around small airways and vessels, until the volume limits of these potential spaces are overcome. Subsequently, additional fluid may flood the alveoli themselves (213, 221). Due to this pattern of extravascular fluid accumulation, in addition to the relative protection of the pulmonary capillaries to leak, it is important to recognize that alveolar edema is, in fact, a late manifestation of a severe fluid imbalance within the gas-exchanging units (221).

Paracellular junctions

Pulmonary edema may also result when the permeability of the alveolar wall is abnormally increased, thereby facilitating the leak of plasma proteins and fluid. Except in cases of outright epithelial cell damage, such as may occur in capillary stress failure, this type of pulmonary edema is typically associated with increased permeability of the capillary endothelium (24). It has long been recognized that the epithelium provides the tightest barrier component of the alveolar wall to the transport of fluid and hydrophilic solutes (26, 59, 178, 198, 221, 224, 241). In contrast, the endothelium, while considered continuous, serves as a much less restrictive size- and charge-selective semi-permeable barrier. Endothelium nonetheless has an important “sieving” function to prevent the passage of both large and small plasma proteins (24, 26, 30). In fact, using experimentally measured values of the reflection coefficients for various hydrophilic substances in excised dog lungs, Taylor and Gaar calculated the average equivalent pore radii of the alveolar epithelium and pulmonary capillary endothelium to be 6 to 10 Å (0.6–1.0 nm) and 40 to 58 Å (4.0–5.8 nm), respectively (224). This demonstrates an approximately fivefold higher permeability of the endothelium as compared to the epithelium.

The difference in permeability and effective pore size between these cell layers may be attributed to their respective arrays of paracellular junctions. The barrier function of the epithelium may be attributed to tight junctions, the most critical component of which, both structurally and functionally, are the claudins. Claudins, in turn, are linked to the actin cytoskeleton via scaffold proteins such as zonula-occludens-1 (ZO-1) and ZO-2 and are stabilized by occludin. At least 14 claudins are expressed within the alveolar epithelium, with varying contributions to barrier resistance. Claudin-3 (cldn-3), cldn-4, and cldn-18 are the most highly expressed, with cldn-3 found predominately in AEC2s (123, 169, 196). There is accumulating evidence that cldn-4 may have a protective role in the lung beyond tight junction formation. Cldn-4 levels have been shown to correlate with the rate of AFC (190) and may predict decreased severity of lung injury in some settings (169, 196).

In contrast, only one-fifth of endothelial cell-cell junctions are tight junctions, which typically contain cldn-5 (26). Rather than tight junctions, the predominant paracellular junction of capillaries is the adherens junction vascular endothelial cadherin (VE-cad, also called cadherin-5). In addition, gap junction molecules [connexins (Cx) −37, −40, and −43 in the lung] form direct communication channels between the cytoplasm of adjacent endothelial cells, and may also play a role in regulating barrier permeability (24, 170). The foundational unit of the adherens junction is a cadherin, which is a single-chain transmembrane protein that binds homotypically in a calcium-dependent manner. Like tight junctions, adherens junctions are also linked to the actin cytoskeleton, via catenins (26). Notably, the reorganization or disruption of adherens junctions, in association with altered cytoskeletal dynamics (actin stress fiber formation, or microtubule disassembly), leads to the formation of interendothelial cell gaps and is the primary source of increased endothelial permeability in the setting of inflammation or mechanical injury (24, 26, 162). Importantly, intercellular gap formation is not a normal phenomenon for pulmonary capillary endothelial cells. This lung-specific feature of the microvascular endothelium enhances the tighter barrier properties of these cells in comparison with the endothelium of extra-alveolar vessels. In addition, gaps that do form in the pulmonary capillaries have a particular tendency for reversibility (162).

Studies from the 1960s and 1970s identified a certain changeability in endothelial permeability, supporting the idea of “labile” or “stretchable” pores. Pietra and colleagues perfused isolated lung preparations with low molecular weight proteins such as horseradish peroxidase (HRP, 40 kDa or about 25 Å in diameter) or hemoglobin (64.5 kDa, approximately 60 Å). By transmission electron microscopy, these particles were observed to remain in the capillaries at physiological pulmonary artery pressures, but the particles passed into the interstitium at elevated pressures. HRP required a lower pressure threshold for translocation of 30 mmHg, versus 40 mmHg for hemoglobin. But while these molecules could leave the microvasculature at 30 to 40 mmHg pressure, they could not enter the airways until pressures were sufficient to disrupt the epithelium—for example, 50 mmHg (178, 221). A decade later, Nicolaysen et al. identified an increased filtration rate, and hypothesized a concomitant increase in hydraulic conductivity, in isolated rabbit lungs perfused at pressures above 25 mmHg. The increase in permeability was, in some cases, reversible upon return to lower pressures (160). Even at pressures sufficient to cause stress failure, the endothelium and epithelium appear capable of closing some cell breaks within minutes (63).

Alveolar fluid clearance

In addition to serving as the primary gatekeepers of alveolar septum permeability, the alveolar epithelium plays a critical, active role in maintaining blood-gas separation, through the removal of fluid and solutes from the distal airspaces. This process is termed AFC. Matthay et al. (143) first hypothesized the existence of this property of the alveolar epithelium in 1982, upon observing that following instillation of isosmotic solutions into the lungs of sheep in situ, fluid was removed from the airspaces at an exponential rate, whereas protein removal was very slow. Today, it is clear that both AEC1s and AEC2s of the distal lung epithelium participate in the active vectorial transport of ions, particularly sodium, across the alveolar wall. Ion transport goes from the apical to the basolateral surface, thereby providing a driving force for water removal from the alveolar lumen via generation of an osmotic gradient (24, 142, 154). AFC is, in fact, the major means of fluid removal from the distal airspaces (154). The primary transporters involved in AFC are the apical amiloride-sensitive epithelial sodium channel (ENaC) and the basolateral Na-K-ATPase. Additional amiloride-insensitive sodium transporters and the cystic fibrosis transmembrane regulator (CFTR) chloride channel also participate in fluid clearance. In addition, the aquaporin-5 (AQP5) water channel (which is found only on AEC1s) is important for passive water reabsorption (Figure 12B) (24, 142, 154).

Multiple hormonal mediators regulate the processes of AFC. In particular, AFC is stimulated by nonspecific adrenergic β-agonists and β₂-agonists (24, 67, 134, 154), as well as by glucocorticoids and thyroid hormone (24, 134, 154). Catecholamines stimulate AFC via upregulation of ENaC, CFTR, and the Na-K-ATPase. The process of increased sodium transport across the epithelial layer appears to be at least in part mediated by CFTR, possibly via increased chloride uptake that in turn generates a favorable electro-chemical gradient for sodium entry into cells (154). Notably, the catecholamine surge that occurs at birth likely has a role in stimulating the massive AFC required to empty the alveoli of fluid, and to transition from a liquid- to an air-filled organ (142).

Surfactant

Pulmonary surfactant is most commonly discussed in terms of its critical role in maintaining lung compliance and preventing alveolar collapse. However, the existence of a surface-active alveolar lining was hypothesized in the 1950s based on the need for a substance within the alveolus to maintain not just alveolar stability, but also fluid balance (45, 46, 171, 172). It was similarly recognized at this time that the principle defect in infant RDS was a lack of surface-active material. When assayed by film balance, the minimum surface tensions of lung extracts from newborns weighing less than 1.1 to 1.2 kg, or in infants suffering from hyaline membrane disease, were noted to be 20 to 30 dynes/cm (8). In comparison, in the normal lung, surface tension drops to below 10 dynes/cm at resting volume and in fact may be approach zero (8, 11, 31, 46, 103, 117, 199, 200). Thus, the lungs of an infant with RDS are characterized by significantly higher levels of surface tension than those in a normal lung.

Importantly, the problem plaguing lungs without surfactant is not that the peak surface tension near TLC is elevated, but rather that the minimum surface tension upon deflation is too high (8, 172). The advantage of surfactant comes from its low surface compressibility. This means that as lung volume decreases with exhalation, a small decline in alveolar surface area is accompanied by a relatively large decrease in surface tension. This drop in surface tension, in turn, imparts stability to the pulmonary parenchyma during lung deflation, when the alveoli would otherwise have a tendency to collapse (9). (Recall that surface forces govern alveolar geometry and stability at low lung volumes, whereas tissue forces determine alveolar stability at near-maximum lung volumes.) The drop in surface tension due to surfactant also allows a relatively larger range of alveolar unit sizes to coexist (31, 46, 103). Because surfactant is absent, or nearly absent, in the lungs of infants with RDS, the alveoli have an increased tendency to collapse on exhalation. This, in turn, makes it harder to reopen the airspaces on inhalation, and significantly increases the work of breathing for affected infants.

The reduction in alveolar surface tension imparted by surfactant also reduces the magnitude of the interstitial hydrostatic pressure P_is, making it less negative. This, in turn, reduces the driving force tending to favor fluid transudation out of the capillaries. That this reduction plays a meaningful role in maintaining fluid balance in the lung has been demonstrated by experiments in which pulmonary edema was induced in in situ dog lungs solely by artificially increasing surface tension, either via high tidal volume ventilation from low FRC at low temperature or by detergent lavage. In both cases, an increase in surface tension alone, without evidence of a change in either capillary hydrostatic or oncotic pressure, was sufficient to cause fluid leakage in the lung (3, 29, 161).

Thus, with deficient or ineffective surfactant, there are consequent increases in surface forces at low and medium lung volumes. Two primary effects result: first, fewer alveoli stay open upon deflation toward FRC, leading to increased work being required to re-inflate the lung [particularly so given the low compliance of the inflation limb of the pressure-volume curve at low volumes (Figure 7)]. Secondly, fluid balance is altered, increasing the tendency for fluid transudation out of the vessels that may cause edema. In RDS, this pathophysiology manifests as hypoxemia and hypercapnia; increased work of breathing, evidenced by grunting and retractions; and atelectasis and hyaline membranes (due to protein-filled fluid transudation into the alveoli) on histology (8, 48, 220).

Notably, the secretion of surfactant by AEC2s, and the size of the surfactant pool in the alveoli, is highly regulated, so that surfactant is secreted when it is needed. For example, there is a dramatic increase in the size of the surfactant pool at birth, accompanying the significant increase in surface tension that comes with the transition from liquid- to air-filled gas exchange units. The amount of surfactant in the air sacs also increases in the setting of exercise and the accompanying increased ventilatory movements (266). Exocytosis of surfactant may be stimulated by stretch from increased tidal volume or respiratory rate, and by various endocrine mediators that act directly on AEC2 cells (133, 139, 265, 266). Interestingly, the “sensing” of stretch by AEC2s appears to be accomplished by AEC1s, which communicate this signal to AEC2s via calcium oscillations through gap junctions (7).

Acute respiratory distress syndrome

ARDS was first described in 1967 by Ashbaugh et al. In a case series of 12 patients, they described a syndrome of severe dyspnea, tachypnea, oxygen-refractory cyanosis, decreased lung compliance, and diffuse bilateral infiltrates on chest radiographs. Precipitating insults were not dependent upon vascular hydrostatic pressures, and included trauma and pancreatitis (6). While our understanding of the general pathogenesis of this syndrome has improved substantially since this initial report 50 years ago, the incidence, as well as morbidity and mortality of ARDS, remain extremely high. In a recent international study of 29,000 patients admitted to intensive care units over a 4-week period, ARDS accounted for 10% of the admissions and was associated with mortality rates approaching 50% in severe cases (22). Despite numerous randomized controlled clinical trials, treatment approaches to decrease mortality, beyond lung-protective ventilation strategies, remain elusive (65).

By its most recent definition using the Berlin criteria of 2012, ARDS is a syndrome of respiratory failure of acute (less than 1 week) onset that is not explained by heart failure or fluid overload. Patients demonstrate bilateral opacities on chest radiograph or computed tomography scan and display a Pa_O2∕Fi_O2 of 300 mmHg or less in the setting of at least 5 cmH₂O positive end-expiratory pressure (PEEP). Mild, moderate, and severe sub-categories of ARDS are determined by more specific Pa_O2∕Fi_O2 ranges (113). At its foundation, however, ARDS is a syndrome characterized by a profound failure of the blood-gas barrier, with resultant derangement of oxygen and carbon dioxide exchange.

With our improved understanding of the etiology of this syndrome has come an increasing recognition of the ways in which the pathogenesis and course of ARDS are linked to the biological mechanisms that maintain barrier integrity. Whatever the inciting insult, the early “exudative” phase of ARDS is characterized by immune-cell mediated injury to the epithelium, endothelium, and underlying BMs (141, 225). The resultant increased endothelial permeability, with loss of protein sieving, is the principal cause of the interstitial and alveolar edema of ARDS, and endothelial leakiness, in turn, begets more inflammation (24, 26).

At the same time, the barrier-protective functions of the alveolar epithelium are disrupted in ARDS, whether due to outright cell injury/death or to metabolic alterations. The very flooding of alveoli with protein-filled fluid often leads to impairment of the process for AFC. Pulmonary edema fluid has been demonstrated to reduce the expression of sodium and chloride transporters by AEC2s in vitro, with a concomitant decrease in fluid transport (128). While AFC may be preserved or even increased in the setting of acute lung injury, such as may occur with surges of endogenous catecholamines (154) or in cases where epithelial permeability is increased(67), in the setting of alveolar flooding, AFC is significantly impaired (154). Overall, AFC is decreased in multiple models of acute lung injury (154), and in the majority of patients with ARDS (243). Not surprisingly, higher AFC rates are associated with better clinical outcomes in ARDS patients (243). In addition, ARDS is associated with changes in surfactant composition, including altered phospholipid and surfactant protein profiles (90, 93, 96). Surfactant that is secreted in ARDS patients also has decreased activity due to inactivation by serum proteins, particularly polymerized fibrin, that leaks from the capillaries into the airspaces (93). Surfactant secretion itself may also be impaired if altered lung mechanics in the setting of edema inhibit the normal mechanosensing behavior of AEC1s (24). These changes in surfactant have detrimental consequences for both alveolar stability and fluid balance, and ultimately for patient outcomes.

Importantly, the lung injury of ARDS may be worsened due to trauma arising from mechanical ventilation. Ventilatory mechanical trauma may be via direct damage to the alveolar septa (with rupture leading to air leaks or edema). Alternatively, ventilator injury can lead to inflammation in the setting of high strain, thereby exacerbating the dysregulation of biological protective processes such as surfactant secretion and fluid clearance. Thus, “lung-protective” ventilation strategies are a critical part of ARDS management, and will be discussed below in the discussion of “Ventilator-associated lung injury.”

Mechanical resilience of the blood-gas barrier

Thus far we have considered the gross consequences of surface and tissue forces on the architecture and morphology of the blood-gas barrier, but we have not considered whether the integrity of the barrier is impacted by the forces acting upon it. In addition to biological regulatory mechanisms, preservation of blood-gas separation relies critically on the mechanical robustness of the barrier.

In this section, we will first consider the architectural features of the lung parenchyma that provide protection to the alveolar-capillary membrane. We will then discuss the extracellular matrix components that endow the barrier with its strength. Finally, we will consider two examples—capillary stress failure and VALI—of settings in which the lung may undergo mechanical damage and subsequent gas exchange failure.

Role of the alveolar duct

While not yet discussed extensively in this review, the alveolar duct plays an important role in the transduction of forces from the bulk organ into the alveolus. The alveolar ducts comprise the most distal branches of the airway tree, following from the respiratory bronchioles in humans and giving rise to alveolar sacs. In addition, as the most distal extension of the axial connective tissue system (which also surrounds the airways and arteries of the lung), the alveolar ducts also provide structural support to the thin gas-exchanging walls of the alveoli.

This role of the alveolar ducts was formalized in a theoretical model by Wilson and Bachofen in the 1980s (264). In their model, the duct is composed of intersecting, helical line elements of connective tissue. This helical matrix composition defines the structure of the duct and the alveolar entrance rings and importantly serves as the primary force-bearing element of the alveoli. The delicate alveolar walls merely “drape,” supported, upon this helix, and as a result experience relatively little tissue strain at volumes below 80% TLC. This is not to say that the alveoli do not increase in volume with breathing below 80% TLC. Rather, alveolar expansion at low lung volumes occurs primarily due to the unfolding of septal pleats, instead of by stretch of the septal tissue.

At lower lung volumes, the architecture of both the ducts and the adjoining alveoli are dictated by surface tension. The fibers of the helical line elements are stretched, in turn causing the ducts to widen. At the same time, the alveoli become shallower, with the net result of both of these changes being that alveolar surface area decreases. Importantly, in this range of normal breathing, although surface tension acts at the air-liquid interface, the alveolar walls themselves experience little tension due to tissue stretch (9, 10, 264). On the other hand, at lung volumes approaching TLC, tissue forces dominate the lung pressure-volume relationship. While the peripheral, axial, and septal tissue may all stretch at these large volumes (10), the alveolar septa may still be relatively spared. Experiments using serial section reconstruction to examine the changes in alveolar and alveolar duct geometry over the course of lung inflation and deflation revealed that the effects of tissue forces in the distal parenchyma at high lung volumes are largely confined to the connective tissue of the alveolar ducts. The ducts widen with increased tissue stretch, even as the surrounding alveolar dimensions remain largely constant (148).

These observations suggest that within the physiological range of lung volumes and airway distending pressures, the alveoli may not be at risk of damage from either surface forces or tissue stretch. However, this relative protection of the delicate septa, and in particular it seems, of the thin portions of the wall that serve as sites of gas exchange, does not preclude damage to the alveolar walls in certain pathophysiological or extreme physiological states.

Collagen and elastin

The overall integrity and biomechanical properties of the lung parenchyma are due to the underlying architecture of collagen and elastin fibers and of the matrix of proteoglycans (PGs, often referred to as the “ground substance”) in which they are embedded. In general, collagen is the most important constituent for strength, elastin provides elasticity over a wide range of transpulmonary pressures, and PGs provide resistance to tissue compression (219).

Several studies examining the ECM architecture of the lung have made observations that appear to support the Wilson and Bachofen model, describing a structure of the alveolar duct that is consistent with its presumed role as the stress-bearing element of the alveolar parenchyma. Mercer and Crapo, using serial section analysis of rat and human lungs, described a relatively higher concentration of both collagen and elastin close to the alveolar duct wall in both species, with “bands” of elastic fibers, partly interwoven with collagen, delineating the “mouths” of the alveoli (147). Toshima and colleagues analyzed lungs by scanning electron microscopy (SEM) that had been selectively digested of all but the collagen or elastin. They noted “condensations” of collagen and elastin around alveolar openings (228). Most recently, Wagner et al. analyzed decellularized precision-cut lung slices with SEM and described a helical protein structure defining the alveolar duct that takes the form of a thick elastin cable (237). The matrix of the alveolar septal walls, on the other hand, consists of a fine network of apparently randomly oriented and likely mechanically interconnected collagen and elastin fibers, with collagen being more predominant (147, 205, 228), although the relative amounts of collagen and elastin vary by species (147). Importantly, the alveolar-capillary barrier portions of the alveolar walls are themselves devoid of supporting connecting tissue other than the fused BM made of collagen IV, laminins, and other glycoproteins. However, the interstitium that makes up the inter-capillary posts, which span the width of the capillary between two alveolar epithelial linings, consists of an elastin-like core surrounded by a helical meshwork of collagen fibers that emanate from the epithelial and endothelial BMs (Figure 13). This organization, while providing mechanical stability, also allows for the high compliance of the microvascular sheet (206).

Figure 13.

Schematic of major extracellular matrix and interstitial components in the alveolar wall.

Overall, these collagen and elastin fibers forming the lung parenchymal ECM are thought to act as parallel mechanical elements, with the collagen behaving as a stiff “string” of fixed maximum length, and the elastin acting as a distensible Hookean “spring” (137). At low lung volumes, collagen fibers are wavy and incompletely extended (42, 147, 228), and thus it is the parallel elastic fibers that bear the stress within the organ (Figure 14A). At high volumes, the collagen fibers straighten (Figure 14B). By one estimate, maximum fiber extension is as much as 16% from FRC to TLC (147), which limits further tissue distension (76, 137, 147, 228). While this model is commonly held and can account for the nonlinear character of the lung stress-strain curve (Figure 14C) (137), it may provide an incomplete description of lung elasticity. There is evidence that collagen contributes to elasticity even in the range of normal tidal breathing, as demonstrated by the notable shift in the stress-strain curve at all strains in response to collagenase treatment of lung tissue strips (273). This observation may still be consistent with the string-spring model, but reveal the presence of collagen fibers whose stop lengths are reached at strains that are significantly less than those at TLC.

Figure 14.

Lung stress and strain. Lung collagen and elastin have been described to behave as a “string-spring” pair of mechanical elements. Collagen and elastin at low (A) and high (B) lung volumes. (C) Stress-strain curve for the lung.

Proteoglycans

PGs comprise an additional major component of the lung connective tissue. However, their role in lung mechanics is not well understood (219). PGs are made up of highly negatively charged glycosaminoglycans (GAGs) attached to a protein core, and surround collagen and elastic fibers. They additionally make up parts of the alveolar and capillary BMs, where PGs may contribute a charge component to the filtration barrier opposing particle passage from the alveoli or capillaries into the interstitial space (232). The compressive resistance of GAGs is due largely to electrostatic repulsion between the negatively charged molecules, which increases with increasing PG density/proximity (37). GAGs also attract and noncovalently bind water molecules, contributing to tissue hydration (42). The consequences of GAG properties for matrix mechanics and integrity have been studied in uniaxial mechanical testing of tissue strips. By altering the tonicity of a surrounding solution bath, or by enzymatically degrading specific GAGs, the impacts of PGs on tissue mechanics can be studied under controlled conditions. Hypertonicity (with an associated increase in the number of positive ions in the solution bath) diminishes the repulsive forces between GAGs, causing PGs to “collapse” and lose their volume-maintaining properties. This, in turn, shifts the stress-strain curve of the whole lung tissue to the right (i.e., reduced stiffness), indicating the contribution of GAGs to normal lung tissue stiffness (42). The stiffness component supplied by GAGs is primarily viscous: rapid changes in strain are opposed by the highly hydrogen-bonded macromolecules and their water molecule payload. Cavalcante et al. (42) developed a 2-D network model of the alveoli incorporating collagen and elastin fibers as well as PGs, which demonstrated a role for PGs in stabilizing the collagen/elastin network by altering the ability of fiber folding and stretching particularly at lower lung volumes. In addition, selective digestion of the GAGs chondroitin/dermatan sulfate, or of heparan sulfate, leads to an increase in energy dissipation, as evidenced by increased hysteresivity. This may be consistent with a role for GAGs in providing lubrication between adjacent ECM fibers (2). The potential importance of PGs for lung tissue integrity was demonstrated in another network model mimicking mechanical wall rupture associated with emphysema, in which stiffer PGs were able to slow the rate of tissue deterioration (222). Thus, PGs appear to interact with other key matrix constituents to contribute to bulk lung tissue stiffness and stability.

Collagen IV

Although collagen, elastin, and GAGs are the primary determinants of gross lung integrity and mechanics, it is thought that collagen IV, a major matrix constituent of the fused epithelial-endothelial BM (122, 255), is the primary source of strength of the thin portion of the blood-gas barrier (258, 261). Evidence for this conclusion comes in part from studies of capillary stress failure, in which high capillary wall stress due to elevated transmural pressures causes breakage of components of the alveolar septum (see more in the discussion of “Capillary stress failure”). The thickness of each of the three layers of the blood-gas barrier—epithelium, interstitium, and endothelium—correlates in different species with the transmural pressures at which failure occurs (25). However, it is true that the epithelium and endothelium rupture up to two times more frequently than their accompanying BMs (140, 231, 260), which suggests that it is the thickness of the interstitium (which includes the BM), and less so the cells, which confers additional strength to the alveolar wall. Supportive of this fact, the calculated wall stress in pulmonary capillaries may reach values so high as to exceed the maximum strength of all soft tissues except for collagen (260). Further evidence for a critical role of the BM comes from the work of Welling and colleagues. They perfused isolated intact and decellularized rabbit renal tubules and found that the epithelium contributed negligibly to tubule distensibility, thus implying that the BM must be the structure responsible for tubule mechanical properties. Their further observation that the Young’s modulus of the BM, 0.7 to 1.0 × 10⁸ dynes/cm², is comparable to that of tendon collagen, suggests that BM collagen IV may indeed be responsible for the structure’s tensile strength (253).

Capillary stress failure

A consequence of the extreme thinness of the blood-gas barrier is that the small circumferential tensions in the capillary wall arising from pulmonary vascular perfusion pressure may translate into extraordinarily high capillary and alveolar wall stresses (261) (see discussion in the section titled “Surface and tissue forces”). Such high wall stress is capable of damaging both the cellular and noncellular components of the alveolar septum, and in so doing, causing stress failure of the blood-gas barrier, as has been described extensively by West and colleagues. In a study of rabbit lungs perfused in situ, transmural pressures above 30 cmH₂O (about 22 mmHg) resulted in thickening of the interstitium on the thick side of the blood-gas barrier, presumably due to accumulation of edema fluid extravasated from the capillaries. At pressures above 50 cmH₂O (~36 mmHg), breaks in the endothelium and the epithelium were evident by electron microscopy, with concomitant translocation of RBCs into the interstitium and/or airspaces (231). Transmural pressures above a certain threshold have also been shown to cause endothelial, epithelial, and BM disruption in other species including dog (140) and thoroughbred racehorses (260). In the latter, capillary stress failure was ultimately shown to be the cause of the long-recognized but unexplained exercise-induced pulmonary hemorrhage observed in these animals after races.

Notably, the threshold for barrier failure varies significantly among species, rising to approximately 90 cmH₂O (~66 mmHg) transmural pressure in dog (140), and estimated as approaching 130 cmH₂O (~95 mmHg) in the horse (260). These differences in strength appear to be at least partly related to the thickness of the blood-gas barrier, which is smallest in rabbit, intermediate in dog, and thickest in horse (25, 140). However, dog and horse lungs experience failure starting at higher pressures than would be predicted by extrapolating on the basis of barrier thickness alone, suggesting that there may be additional structural differences to account for the increased barrier strength in these species (140).

Circumferential tension is not the only force acting upon the microvessels surrounding the alveoli. The largely unsupported nature of the alveolar capillaries leads to a reciprocal transmission of force, whereby microvessels are exposed to alveolar surface tension and to alveolar wall stretch (Figure 8). The bulging of capillaries into the alveolar space means that, at large lung volumes, increased surface tension may produce a compressive force that flattens the capillaries in the septal wall. However, the fact that vessel flattening may occur at high transpulmonary pressure in the absence of high transmural pressure (i.e., without the bulging of capillaries) (86) suggests that stretching of the alveolar wall at large lung volumes may in and of itself provide lateral tension to the capillaries (69, 261). This tissue force can contribute to capillary stress failure: at a given transmural pressure, a higher transpulmonary pressure (and thus lung volume) is associated with an increased frequency of tissue breaks, even below the above-described transmural pressure threshold (69). In contrast, it has been suggested that the compressive force due to surface tension may actually be protective, counteracting the strain on capillaries that arises from high transmural pressures (261).

Given that the above instances of stress failure were either experimentally induced (140, 231) or a consequence of extreme exercise (260), it is reasonable to ask whether the human blood-gas barrier is at risk of capillary stress failure under physiological, or pathophysiological, conditions. Studies of bronchoalveolar lavage (BAL) samples from elite athletes after either submaximal (102) or maximal (101) exercise suggest that the integrity of the blood-gas barrier can be compromised during exercise. This was evidenced by a significant increase in RBCs and protein in BAL samples but only in the setting of maximal exercise. This fragility only at the upper limits of physiological pressures implies that the thinness of the human blood-gas barrier may have evolved to attain its minimum possible dimension (261), that is functional under almost all conditions of normal activity. However, certain pathophysiological conditions are also known to put the blood-gas barrier at risk of failure. Exposing the barrier to supraphysiologic stress, such as the elevated pulmonary artery pressures observed in neurogenic pulmonary edema and hypoxia-associated high altitude pulmonary edema, may lead to membrane breakdown and pulmonary edema. Heart diseases associated with increased left atrial pressure (e.g., mitral stenosis), or certain settings of mechanical ventilation, can cause edema as well. Lastly, inherent damage to the collagen IV of the BM, as in patients with anti-glomerular BM antibody (Goodpasture’s disease), can lead to functional breakdown of the blood-gas barrier in the lung (259).

Ventilator-associated lung injury

The discussion thus far of lung and alveolar pressure-volume behavior has been predicated on an assumption of uniform expansion and deflation, implying also a degree of alveolar interdependence. However, these descriptions belie the heterogeneity that exists within the normal lung in terms of the behavior of individual alveolar elements—such as capillary or septal segments—in response to applied stresses (173). In studies of capillary stress failure, the wide variability observed in the occurrence and frequency of BM, endothelial, and epithelial cell breaks has been demonstrated to be due more to individual capillary than to regional differences (69, 140). There is also a significant amount of variability in the stretching and folding behavior of individual septal segments even within a single alveolus. In one study, even at volumes near-maximal lung inflation, the stretch of individual septal segments was found to vary between 5% and 25% (42, 173).

A consequence of these heterogeneities is that the alveolar distending pressures are not, in fact, equal across the entire organ. In an idealized, uniform lung, the airspace distending pressure is everywhere equal to the transpulmonary pressure, which, in turn, determines lung volume. However, when nonuniformities are present—such as in an atelectatic or emphysematous region—the local distending pressure applied by adjacent septal tissue will differ from the transpulmonary pressure in such a way as to oppose the change in lung volume. Mead et al., in a classical theoretical study, calculated that this phenomenon could lead to a local outward pressure of 140 cmH₂O on a collapsed region of the lung, in response to a transpulmonary pressure of just 30 cmH₂O (145). Acting upon a reduced volume, this high pressure will lead to the development of an even higher stress. Thus, it is important to recognize that an externally applied pressure, such as in mechanical ventilation, may be transduced locally into very high pressures.

It is regional variation in the degree of tissue stress, with accompanying local over-distension, which is at the root of VALI. Sometimes called ventilator-induced lung injury (VILI), this is a nonspecific pattern of lung injury associated with mechanical ventilation (76, 203). That positive-pressure mechanical ventilation may cause lung damage has long been recognized (56, 203), but it is only recently that the spectrum of VALI pathophysiology has been understood to encompass more than air leaks and pneumothorax due to “barotrauma.” Rather, VALI is now understood to arise predominately not from barotrauma per se, but from “volutrauma,” ventilation with high volumes; as well as from “atelectrauma,” ventilation with low volumes. “Biotrauma,” which is the release of inflammatory cytokines due to nonphysiological tissue strain even in the absence of tissue rupture, is also a component of VALI (56, 76, 203). Atelectrauma arises due to the cyclic opening and closing of collapsed alveoli, as well as the elevated wall stresses arising from the ventilation of smaller volumes of atelectatic parenchyma, as described earlier (76, 203). Thus, VALI may manifest as air leaks due to high distending pressures and volumes but also may manifest as alveolar edema due to impaired fluid clearance or altered epithelial or capillary permeability. In addition, VALI may lead to lung hemorrhage in the setting of tissue damage, including capillary stress failure (56, 68, 105, 127, 203). Ultrastructurally, barotrauma may be evidenced by full-thickness tears in alveolar walls, which is another example of stress failure (105).

It is important to emphasize that mechanical ventilation is not inherently damaging, but rather that it may cause injury once a certain stress or strain threshold has been met (76). Stress, or the force acting over an oriented area, in this setting is equal to the transpulmonary pressure, except where elevated or reduced by local inhomogeneity. Barotrauma in the lung arises when this stress exceeds the tensile properties of collagen fibers, leading to tissue rupture (75). Strain, or the percent increase in length over the original length, may be considered as the ratio between the tidal volume and the FRC, where FRC is equivalent to the end-expiratory volume. Volutrauma occurs when cells are stretched excessively or abnormally and is not contingent upon tissue failure (75, 184). Though the cells do not bear the brunt of the mechanical load applied to the lung tissue, they may respond to the strain within the ECM to which they are attached, by rearranging their cytoskeletons (76). The idea that some minimum stress or strain might be required to cause lung injury was demonstrated in a recent study involving the mechanical ventilation of healthy pigs. Ventilator-induced lung edema was observed only when the strain exceeded 1.5 to 2.0, corresponding to a TV equal to 150% to 200% of the FRC. Notably, this strain is also the point where the stress-strain curve becomes nonlinear, which may, in turn, correspond to regional attainment of TLC, or maximal effective volume for a particular lung segment (Figure 14C) (184). A strain of 2.0 is approaching the upper volume limit for a normal human lung (76).

A maximal strain of 1.5 to 2.0 suggests a certain robustness of the blood-gas barrier to strain-induced injury. However, this stress/strain threshold may be lowered in certain pathological conditions, such as ARDS, where the membrane integrity is already impaired by inflammatory and infectious assaults on the lung parenchyma. The weakness in the alveolar barrier also increases the likelihood of VALI. To illustrate why this is so, the term “baby lung” has been coined to describe the lung affected by ARDS. As first appreciated on computed tomography scans, ARDS is characterized by heterogeneity within the lung, wherein the dependent regions of the lung are poorly aerated due to a “squeezing” out of air in the setting of edema fluid. In ARDS patients, some smaller, nondependent region is normally aerated, and in fact has normal compliance (75). If one considers the ARDS lung to be this “baby lung,” often equivalent in size to the lung of a 5- or 6-year old (75), it should be clear that the strain threshold of 1.5 to 2.0 will be reached upon mechanical ventilation with a much smaller-than-normal tidal volume, compared to a normal lung, due to a reduced volume at FRC.

Accordingly, numerous in vitro, animal, and clinical studies have found evidence to support a “lung-protective” strategy of ventilation in the setting of ARDS, although the ideal strategy remains an elusive target (203). Current strategies include (i) low tidal volume, so as to limit strain; (ii) moderate [at least 5 cmH₂O (225)] PEEP, to keep regions of lung opened throughout the ventilatory cycle that would otherwise collapse; and (iii) low transpulmonary pressure (often measured clinically through the imperfect surrogate of plateau pressure, which is the end-inspiratory pressure under zero flow), to avoid barotrauma (65, 68, 75, 76). Prone positioning is also gaining traction as a clinical strategy (75), since this position reduces the gravitational pressure gradient in the lungs (87) and creates a more uniform distribution of parenchyma (114), thereby leading to a more even stress/strain pattern across the organ. Significantly, prone positioning has been shown to reduce mortality in severe ARDS (Pa_O2∕Fi_O2 of 150 mmHg or less) (65). Many of the classic studies leading to these findings are detailed in an excellent review by Dreyfuss and Saumon (56) and so will not be discussed in more detail here. One landmark study to remark upon, however, is the ARDS Network trial that showed a nearly 9% absolute mortality benefit from using reduced tidal volume ventilation (6 mL/kg body weight vs. 12 mL/kg) in patients with ARDS (156).

Ex Vivo Engineering of the Blood-Gas Barrier

The “holy grail” of pulmonary blood-gas barrier engineering is the creation of a tissue-engineered whole organ suitable for transplantation. Lungs-on-a-chip are designed to recapitulate select key features of the blood-gas barrier at the microscopic scale. Extracorporeal membrane oxygenation (ECMO) circuits are designed to provide the requisite structural features—vast surface area, exquisite thinness—of a membrane responsible for gas exchange. In contrast, an engineered lung must achieve and maintain the ability for gas exchange within the biologically and mechanically dynamic environment of the thorax, and on a scale to meet the oxygenation needs of an entire organism. As such, the engineered whole lung blood-gas barrier has both the broadest and most stringent design criteria of the engineered blood-gas barriers considered here (Table 1) and will require strict and extensive characterization to ensure full functionality prior to any clinical translation (Table 2). We will address here two approaches to ex vivo biological pulmonary barrier engineering: (i) lung-on-a-chip and (ii) whole-lung regeneration.

Table 2.

Methodologies for Assessment of Blood-Gas Barrier Function in Engineered Lung

Design criteria	Specific barrier features	Methodologies for assessment	Relevant references
Blood-gas separation	Paracellular junctions	Ultrastructural assessment for tight junctions by TEM
		IHC, PCR, WB for cell-cell junction proteins Epithelium: ZO-1, Cldn-3,4,18 Endothelium: ZO-1, Cldn-5, VE-cad, Cx37,40,43
		Barrier assay(s) demonstrating resistance to particle and fluid translocation Further assay development is needed	(189, 194, 211)
Vast alveolar and capillary surface areas	Vast surface area (alveolar and capillary)	Morphometric analyses: mean linear intercept, alveolar and capillary surface areas	(106)
Thinness compatible with gas diffusion	Alveolar-capillary membrane thickness permissive of gas diffusion	Measurements of arithmetic and harmonic mean barrier thicknesses	(106, 252)
		Ultrastructural assessment for the presence of AEC1s by TEM
Strength to withstand repetitive inflation/deflation, and to receive the whole of the cardiac output	Intact collagen, elastin, and GAGs	Trichrome, Verhoeff-Van Gieson (VEG), and Alcian blue stains
		Mechanical stress-strain testing of tissue strips; correlation to mechanical models for estimation of stress/strain parameters and determination of UTS	(175, 218)
		Quantitative assays and proteomics	(99)
		Multiphoton microscopy-second harmonic generation for collagen imaging	(50)
	Intact basement membrane collagen IV	IHC for collagen IV
		Quantitative proteomics	(99)
		Stress failure perfusion assay, with TEM: Determination of critical pressure range for capillary stress failure	(231, 261)
Compliance to enable effective ventilation, and to maintain alveolar patency	Functional surfactant	IHC, PCR of tissue for surfactant protein expression
		Ultrastructural assessment for AEC2 lamellar bodies by TEM
		WB for surfactant protein B (SPB) processing intermediates	(21)
		BAL analysis for surfactant proteins and phospholipids	(112)
		BAL surface tension measurement	(64)
	Functional surfactant and intact elastin	Quasi-static pressure-volume curves	(218)
		Forced oscillation technique: assessment of respiratory impedance	(20, 167, 218)
Active maintenance of an air-liquid interface	Vectorial ion transport	IHC, PCR, WB for machinery for Na+ transport: ENaC and Na-K-ATPase subunits
		IHC, PCR, WB for Cldn-4
	Functional surfactant	See above
	Functional lymphatic clearance	Assay development is needed

BAL, bronchoalveolar lavage; IHC, immunohistochemistry; PCR, polymerase chain reaction; TEM, transmission electron microscopy; UTS, ultimate tensile strength; WB, Western blot.

Lung-on-a-chip

The last decade has seen the rise of “organ-on-a-chip” devices that allow robust visual and quantitative explorations of cell-cell coculture within biomimetic ex vivo platforms. There are many versions of a lung-on-a-chip, the most physiologic of which incorporate multiple cell types on either side of a thin, cyclically stretched membrane mimicking the native epithelial-endothelial BM. None of the published devices allow the close level of epithelial-endothelial interweaving that is intrinsic to native lung tissue; however, they do preserve a simplified cellular topology and organization that allows appropriate cell polarization. Further, and importantly, chip-based platforms generally allow direct light microscopy, quantitative electrical measurements, and the design and implementation of robustly modeled and controlled drug toxicity studies. Lung chips can be powerful tools for exploring basic mechanisms of lung cellular function and parsing otherwise complex cell-cell interaction systems.

From both an engineering and a basic science perspective, there is a need for inexpensive, biologically useful ex vivo platforms that allow complex coculture of multiple pulmonary cell types within a mechanical environment approximating that found within the lung. Alveolar tissue contains epithelia, endothelia, mesenchyme, and immune cells all interacting in a highly dynamic mechanical environment (4, 94). These interactions are known to be crucial to the regulation of cellular phenotype, including endothelial control of epithelial differentiation (186, 269), and epithelial-mesenchymal interactions (44, 89, 130, 274). It is, therefore, likely that robust barrier engineering efforts will need to incorporate a wide variety of cell types present in native lung tissue. Cellular crosstalk has been strongly implicated in the development of interstitial lung diseases (13), and there is a need for novel platforms such as lung-on-a-chip devices that can allow targeted exploration of specific cell-cell interaction mechanisms and biologically relevant drug studies.

Current state of lung-on-a-chip technology

An early prototype for engineering the alveolar-capillary barrier in vitro came from Hermanns et al. (98) in 2004 (Figure 15A). This system placed A549 or NCI H441 epithelial cells on one side of a noncompliant porous plastic membrane opposite primary human pulmonary microvascular cells (HPMEC) on the other side and measured barrier function using trans-epithelial electrical resistance (TEER). This study set a precedent for using TEER in studies of ex vivo lung barrier engineering that has continued as an important metric to this day. Interestingly, Hermanns and colleagues found that TNF-α exposure reduced barrier function if, and only if, the platform was exposed to the cytokine on the basolateral (endothelial) side. No significant effect was detected if exposure was done apically to the epithelial cells. This study laid important foundations and acted as a proof-of-concept that cellular crosstalk and topology are important for successful investigation of cell signaling mechanisms.

Figure 15.

Evolution of lung-on-a-chip designs. (A) A rudimentary model of the alveolar blood-gas barrier can be made by plating the underside of a porous Transwell membrane with endothelium and the apical side with pulmonary epithelium. This allows intermittent TEER measurements and visualization of the cultured cells but not stretch or flow. (B) A pioneering design by Huh et al. applied cyclic stretch to an elastic porous silicone membrane, allowing modeling of alveolar ventilatory mechanics. This design also allows media flow, thereby mimicking capillary and airway shear dynamics. (C) A later chip-based design with integrated gold-plated electrodes allowing simultaneous real-time microscopy and TEER underflow dynamics. (D) A recent microimpedance tomography-based lung-on-a-chip design, which segregates all electrodes on one side of the modeled barrier and opens the apical side of this chip for direct visualization and/or intervention. Chips with integrated barrier-measuring electrodes are generally capable of producing more repeatable and cross-laboratory translatable results.

A major step forward was established by Huh et al. through the incorporation of both an ALI and cyclic strain into a lung-on-a-chip device (Figure 15B) (109). This landmark study described a device in which epithelium and endothelium were seeded on opposing sides of a flexible and porous polydimethylsiloxane (PDMS) membrane, with the ability to flow air or media independently on each side. Cells formed well-organized monolayers with extensively developed tight junctions, and after cellular confluence, the device was able to support airflow in the apical epithelial compartment for over 2 weeks without loss of cellular viability or differentiation. The application of an ALI was found to significantly increase barrier function by TEER measurements. The endothelial populations in this setting recapitulated a significant portion of their in vivo physiology, including the upregulation of ICAM1 in the setting of inflammation (TNF or bacterial exposure) and the ability to bind and recruit neutrophils from the capillary lumen and allow transmigration into the alveolar compartment. The application of cyclic stretch was not, in this study, found to affect barrier integrity under normal homeostasis. However, cyclic strain was synergistic with nanoparticle toxicity in the lung, causing greater endothelial activation and reactive oxygen species (ROS) generation than nanoparticle delivery without applied strain. Strain was also found to significantly enhance cellular uptake of alveolar nanoparticles and to cause a fourfold increase in alveolar-capillary nanoparticle transport, even though no difference was noted in the transport of albumin, suggesting that this effect was not due simply to disruption of cell-cell junctions. This finding was confirmed in a whole-organ mouse lung perfusion model, and appears to hold across platforms, since a later-developed diaphragm-style stretch device found that 10% linear strain significantly increased the permeability of the barrier to hydrophilic molecules (217).

The ability of Huh et al.’s chip to incorporate multiple parenchymal cell types, circulating cells, resident pathogens, endothelial shear exposure, parenchymal cyclic strain, and trans-barrier transport makes it a formidable model in the biological engineer’s toolbox. Although many lung-on-a-chip studies are able to generate robust results with simpler platforms that do not necessarily incorporate strain or shear stress, this model and others like it are arguably the gold standard for lung chip technology to this day, and can be argued to be the most biomimetic testing platforms.

There are, nonetheless, a number of nontrivial deficits in this foundational design. Huh and co-authors themselves note that the membrane they used is far thicker (10 μm) than the native BM (~0.63 μm), has a nonphysiologic pore structure, and does not generally allow direct cell-cell contact between the two monolayers. Further, the device employs PDMS, which can complicate drug studies, since the polymer is renowned for its absorption of small molecules and subsequent confounding of delivery and dosage responses (226, 240). Some of these issues could be mediated by the use of alternative membrane compositions and structure. Mondrinos et al. (152) published a fabrication technique to make membranes for lung chips consisting entirely of extracellular matrix, yielding 20 μm-thick membranes with 400 nm pores and Young’s moduli significantly lower than those of standard polymer membranes. However, it is also likely that such membranes are subject to cell-mediated remodeling. Yang et al. recently published an interesting microfluidic membrane made from electrospun poly(lacticco-glycolic acid) (PLGA) that is only 3 μm thick and has less than 1 μm pores, which more closely approximates the native BM and does not have the absorption problems of PDMS. They were successfully able to culture three types of pulmonary cells in tri-culture on this platform but did not incorporate cyclic stretch (268).

TEER measurements of barrier function

TEER is a powerful technique for assessing tight junction dynamics in ex vivo cell culture and as such can be used as a strong indicator of barrier integrity and function (210). TEER measures the ionic conductance across a membrane and its associated cell monolayers, which by necessity takes into account the trans- and paracellular pathways, as follows:

$$\frac{1}{R_{\text{Total}}}=\frac{1}{R_{\text{tc}}}+\frac{1}{R_{\text{pc}}}$$ (8)

As the transcellular resistance is usually quite high, the measurement of total resistance across the membrane becomes roughly equivalent to the resistance of the paracellular route. The resistance of this paracellular route is closely related to tight junction formation and organization, thereby making TEER an invaluable technique for noninvasively probing intercellular barrier function. TEER is generally reported in units of Ω*cm², which must be normalized to the membrane area of the devices in question to provide comparable measurements.

Despite the widespread utility of TEER measurements, it can be very difficult to compare experiments across platforms. Temperature, physical supports, variable electrode compositions, and geometrically determined current densities can all have nonnegligible effects on TEER measurements (5). Aligning TEER measurements across platforms requires inter-device modeling based on structural geometry, which can be prohibitively difficult (164). Further, complete cellular coverage is necessary for relevant measurements, since even a 0.4% gap in cell coverage can cause an 80% drop in TEER, making the method less useful outside of the context of full cellular confluence (164).

Certain novel devices have been engineered which allow real-time integration of TEER electrodes into the lung chips themselves. The design published by Walter et al. (239) allows both TEER and continuous microscopy, and was used to model the intestinal, lung, and blood-brain barriers (Figure 15C). Henry et al. have since published on a similar device. Neither design incorporates the cyclic stretch important for recapitulation of the lung microenvironment(97). In contrast, Mermoud et al. (149) demonstrated a novel device with a number of unique benefits by using microimpedance tomography to measure barrier (Figure 15D). Microimpedance tomography is a more advanced technique than TEER that requires fewer assumptions about the electrical behavior of the system being measured (61). The results of microimpedance tomography are generally held to be more accurate than TEER values (210). Unlike TEER-based chips, the pulmonary model of Mermoud and colleagues was able to incorporate cyclic stretch, since microimpedance tomography electrodes need only be placed on one side of the membrane. This adds the further advantage that the apical surface of the cellular bilayer remains open for investigator visualization and access (149).

Due to the difficulties of comparing TEER values from various platforms, it is challenging to benchmark an “ideal” TEER value for a fully functional barrier. Further, AEC1s are notoriously difficult to culture ex vivo, so it is at this point unknown what would be the TEER value of the AEC1-endothelial combined cellular barrier. However, certain studies can provide a rough idea of the appropriate order of magnitude. Human tracheal and bronchial epithelia from healthy donors, when cultured at an ALI, yielded TEER values of 700 to 1200 Ω*cm² (177). AEC2s cultured ex vivo maintained TEER values on the order of 1000 to 2000 Ω*cm². A comprehensive list of ex vivo pulmonary studies and their associated TEER values can be found in Table 3 of Srinivasan et al. (210). No studies have yet reported TEER values associated with fully differentiated primary AEC1s, although a recent study demonstrated TEER values above 1000 Ω*cm² for lentivirus-immortalized alveolar type-I-like cells (121). We stress that these values are not directly comparable, however, until they are normalized for platform-specific membrane area.

Chip-based cell-systems findings

Lung-on-a-chip platforms have allowed careful investigation of cell-cell crosstalk patterns relevant for barrier engineering. Luyts and colleagues showed that in a pulmonary epithelialendothelial coculture, the addition of a macrophage/monocyte population negatively impacted barrier function, reducing TEER values by 73% (135). Interestingly, macrophages activated with lipopolysaccharide (LPS) were significantly less destructive of measured barrier, still reducing the barrier of the construct but much less so than in the nonactivated case. Benam et al. re-created a “small-airway” on a chip using primary human airway cells from conducting airways and associated endothelial cells. The authors found that IL-13 delivery to the epithelial compartment recapitulated goblet cell hyperplasia, cytokine hypersecretion, and decreased ciliary function seen in asthma. Further, they were able to identify active synergistic effects of endothelial-epithelial crosstalk on associated cytokine secretion (23). Hu et al. found that IL-2 delivery to the capillary compartment induced edema, the magnitude of which was synergistically impacted by cyclic strain. Edema could be rescued by delivery of angiopoietin 1, which is known to stabilize endothelial tight junctions. Hu and colleagues then tested a novel drug candidate, GSK2193874, and found that it could prevent pulmonary edema formation, a finding that was then confirmed in vivo (108). Chip-based models have been used to investigate processes as diverse as fibrocyte extravasation in response to CXCL12 expression (185) and pulmonary epithelial-smooth muscle crosstalk (110).

The most advanced chip-based models to date incorporate multiple tissue types into single experiments so that organ-organ crosstalk effects can be investigated. An early demonstration of this concept, and the applicability to pulmonary engineering, showed that a human lung chip could be integrated with pockets of brain, liver, and bone tissue to study pulmonary metastasis (267). More recently developed microphysiological systems (MPSs) allow the linking of lung chips with liver and heart (202) and then gut, endometrial, pancreas, brain, skin, kidney, and muscle tissue (58), all while preserving the ability to measure barrier integrity using TEER. Edington et al. (58) performed advance pharmacokinetic modeling of diclofenac across all organ systems, providing a proof-of-principle for next-generation drug testing studies and showing that such methods could be used, in principle, to study the effects of delivered drugs or growth factors to influence pulmonary barrier in vitro or in vivo. MPS approaches are likely to yield more biologically relevant findings, particularly regarding delivered small molecules, since they allow cross-tissue metabolism and communication. As these systems become more complex, their findings become both more biologically relevant and also more difficult to interpret without careful controls, measurement, and modeling. A panel of experts on epithelial barriers, reviewing the state of the art of chip-based epithelial barrier engineering in 2015, noted that as platforms become more complicated they also become much more difficult to control. The panel noted that “careful phenotyping of cells (passage number, calibration of barrier properties, molecular markers) and donors (age, clinical characterization, smoking history) is essential, as is standardization of methodologies to achieve reproducibility of results between laboratories” (88).

Whole organ studies

The field of whole lung tissue engineering as it exists today began nearly a decade ago with two reports of engineered rat lungs that achieved short-term gas exchange in vivo (168, 176). These lungs were generated by seeding decellularized rat lung scaffolds with primary mixed fetal (168) or neonatal (176) rat lung cells into the airway compartment, and endothelial cells into the vascular compartment (Figures 16A–16C). The resulting constructs were cultured in bioreactors for several days under conditions of perfusion and ventilation (Figure 16D), and subsequently, the engineered left lungs were transplanted orthotopically into rat recipients (Figures 16E and 16F). The engineered lungs exchanged gas for several hours but ultimately exhibited gross and microscopic pulmonary edema (168) or alveolar hemorrhage (176). Thus, these landmark studies demonstrated proof of concept for the lung decellularization-recellularization paradigm, while revealing a profound challenge of this strategy: re-establishing air-blood barrier function from the starting platform of an inherently leaky scaffold.

Figure 16.

Overview of decellularization-recellularization paradigm for whole lung engineering. (A) Intact native lungs are decellularized via detergent rinsing, creating an extracellular matrix scaffold free of intact cells (B). (C) The scaffold is subsequently seeded with cells into the airway and/or vascular compartments. (D) The reseeded construct is transferred to a bioreactor for culture under physiologically relevant conditions. The resulting engineered organ undergoes (E) ex vivo characterization before final (F) orthotopic transplantation for in vivo functional assessment.

Major recent accomplishments in the field

This strategy of whole lung engineering continues to be actively pursued by several groups, and since the initial reports in 2010, several important steps forward have been made. The most significant of these has been improved strategies for organ revascularization, including endothelial cell seeding techniques (40, 54, 125, 189, 194, 211) and endothelial-mesenchymal coculture (54, 189). In terms of cell seeding, a key observation was that the leakiness of the decellularized lung scaffold creates a sieving effect when a cell suspension is delivered into the vasculature, wherein fluid quickly and preferentially leaks across the alveolar-capillary membrane, leaving the cells lodged in the initial segments of the vasculature (Figure 17A) (40, 189). Thus, improved distribution of endothelial cells within the capillary bed of the decellularized organ may be achieved by seeding cells via both the pulmonary artery and pulmonary veins (40, 125, 189) as well as by diluting the seeded cell suspension and increasing the initial hydrostatic pressure drop into the organ (40, 125). Pulsatile perfusion of the cell suspension during seeding also enhances cellular distribution within the organ (Figure 17B) (125), by increasing capillary recruitment (181).

Figure 17.

Endothelial cell seeding strategy for decellularized lung. The leaky nature of the decellularized scaffold leads to a “sieving effect” whereby fluid preferentially leaks across the capillary wall. (A) Cells clump in the early capillary segments due to fluid leakage. (B) Seeding a dilute cell suspension from both ends of the vasculature, under higher pressure and with pulsatile perfusion, leads to enhanced distribution of endothelial cells within the decellularized scaffold.

The other key advance in re-endothelialization of decellularized lung scaffolds, endothelial-stromal cell coculture, was first demonstrated by Ren et al. They showed that compared to endothelial cell culture alone, coculture of human umbilical vein endothelial cells (HUVECs) with human mesenchymal stromal cells (hMSCs), or human-induced pluripotent stem cell-derived endothelial cells (hiPSC-ECs) with pericytes (hiPSC-PCs), yielded significantly improved matrix coverage by endothelial cells (up to approximately 75% of native coverage), as well as improved barrier function as evidenced by a dextran perfusion assay. The HUVEC-hMSC engineered microvessels were perfusable for 3 days upon orthotopic transplantation, but no other in vivo functional metrics were described (189). In another impressive study exemplifying the benefits of mesenchymal coculture, Doi et al. seeded decellularized rat lung scaffolds with rat lung microvascular endothelial cells (RLMVECs) together with primary rat adipose-derived stromal cells (ASCs). The seeded ASCs appeared to differentiate toward a pericyte phenotype during lung culture and to promote the survival and quiescence of the RLMVECs at 16 days. Upon orthotopic transplantation, RLMVEC-ASC engineered constructs demonstrated resistance to the intra-alveolar hemorrhage seen in transplanted RLMVEC-only lungs, thereby providing evidence of vascular barrier function to cell-sized particles in the setting of endothelial-mesenchymal coculture (54).

An additional major advance in the field of whole lung engineering has been the translation and adaptation of rodent lung engineering strategies to human-sized lungs, with a vision toward future clinical application. Perfusion detergent-based decellularization protocols have been successfully scaled up to whole lungs of larger animals including pig (15, 83, 159, 183, 212, 234), nonhuman primate (14, 27, 212), and human (14, 83, 158, 159). Concomitantly, bioreactors capable of providing physiologically relevant perfusion and ventilation have been developed for these larger organs (43, 187).

Several proof-of-concept studies have been published of the recellularization, and in some cases, implantation of human-scale engineered lungs. Most of these studies use either primary porcine (157) or human (82, 158, 277) cells to repopulate either porcine or human single lung scaffolds, and in general have demonstrated the successful adhesion and proliferation of both epithelial and endothelial cells. It remains an unresolved question in the field as to which cell type(s) may ultimately prove effective for regenerating lungs for transplantation, both because of the logistical challenges of acquiring sufficient cell numbers [upwards of 15 billion cells will be necessary to repopulate a full human lung scaffold, even with 4 population doublings after seeding (215, 216)], and because what constitutes a lung stem cell, and in which settings, is an ongoing and evolving field of inquiry (119, 126). Ideally, any seeded cells would be of autologous origin, so as to reduce the incidence of transplanted graft rejection. To this end, two groups have attempted the use of human-induced pluripotent stem cells (hiPSCs) to repopulate human lung scaffolds. Ren et al. used hiPSC-ECs and hiPSC-PCs to recellularize a human lung lobe, and demonstrated evidence of perfusable reseeded microvessels. However, even with approximately 400 million total cells seeded and after 6 days in culture, they calculated endothelial cell coverage of less than 5%, demonstrating the vast number of cells needed for complete recellularization on this scale (189). Ghaedi et al. reseeded a human right middle lobe with 1 billion total hiPSC-alveolar and airway epithelial progenitor cells, and demonstrated some, though not exclusive, site-specific epithelial marker expression. While epithelial coverage was not quantified, it is clear from histological images that even this cell number after 4 days of culture leaves the majority of the lung extracellular matrix bare (79). Overall, functional assessments of these clinical-scale constructs have been limited: one study demonstrated in vitro oxygenation of perfusate by a decellularized human lung lobe seeded with primary pulmonary endothelial cells and basal epithelial cells (82). Another report demonstrated that a HUVEC- and primary human basal cell-seeded porcine lung scaffold was capable of some gas exchange for 1 h in vivo (277).

One notable recent study by Nichols and colleagues involved an ambitious month-long multi-phase recellularization of decellularized porcine lungs using autologous cells, followed by orthotopic left lung transplantation. The caveat to this experiment was that for the implantation, the airway was anastomosed to the recipient, but not the pulmonary artery and veins. Therefore, though ventilation could proceed within the implanted organ, circulation was not present. The constructs were implanted for up to 2 months in vivo (157). Given the absence of a deoxygenated blood-gas interface in the implanted lungs, assessment of gas exchange capability and barrier function was not possible. Rather, the focus of the study was on evaluating the interaction of the engineered lung with the recipient, in terms of the development of collateral circulation to promote tissue survival at the surfaces of the implant. In addition, any mounting of an immune response or graft rejection by the host, and microbiome colonization of the initially sterile construct, were also evaluated. This study suggested that systemic collaterals might form by 2 weeks after implantation and lymphatic vessels by 1 month. The authors did not see evidence of immune cell infiltration, pro-inflammatory cytokines, or an upregulated T-cell response. Finally, they observed establishment of lung and tracheal microbiological milieu that approached that of a native organ by 2 weeks after implantation. This study thus demonstrated proof of concept for the implantation of an autologously re-seeded allogeneic lung scaffold into a large animal, without evidence of rejection, and with some integration of the engineered organ with the recipient. Ultimately though, any gains toward development of a functional bioengineered lung may not be assessed in this case.

This study exemplifies the increased consciousness toward the logistical and clinical translational challenges associated with generation and implantation of human-scaled engineered organs. While collaterals to the distal lung parenchyma should not be necessary for a construct with fully revascularized alveolar capillaries, the ability of a lung graft to develop collateral circulation supplying the bronchi is certainly relevant, in light of the fact that the bronchial circulation is typically not reconnected in the setting of lung transplantation. Indeed, the lack of bronchial circulation may contribute to early rejection and late lung graft dysfunction (227). It is also conceivable that, in the future, engineered lung transplantation could begin with a period of in situ maturation, without re-connection of the pulmonary vasculature. These experiments made use of allogeneic scaffolds seeded with autologous cells and did not demonstrate evidence of graft rejection, at least over the 2 months post-transplantation. Notably, in a first for the field, the authors also sought to reconstitute at least part of the engineered lung’s immune system, by seeding mononuclear leukocytes as well as macrophages into the acellular scaffold. However, it is unclear whether, and to what extent, these cells may have contributed to a functional immune system for the implant. While an important consideration given the direct exposure of the airway tree to inhaled microorganisms and viruses, including immune cells among the seeded cell types may also contribute to engineered lung tissue homeostasis (111).

An engineered lung made from an allogeneic scaffold that is seeded with autologous cells, as in this study, would likely be ideal in a human translational setting in terms of achieving immune tolerance. Human donor lungs that go unused, which outnumber those transplanted by fourfold (85), could potentially serve as engineered lung scaffolds. Unfortunately, conditions that currently preclude transplantation may also make decellularized scaffolds from these organs unusable: lung scaffolds derived from smokers and from those with chronic lung disease have been shown to negatively impact the behavior and viability of reseeded cells (28, 158, 193, 208, 233). The use of porcine lungs as xenograft scaffolds is an appealing alternative given their potentially unlimited availability. However, the presence of xenogeneic antigens such as the alpha 1,3, galactosyltransferase (α-gal) epitope would lead to hyperacute rejection upon whole organ transplantation in humans. This has prompted the development of transgenic pigs lacking this epitope (74). Alpha-gal knockout pig lungs showed some differences in residual proteins following decellularization, as compared to wild-type lungs, although the behavior of several types of human cells reseeded onto these scaffolds was similar in a recent proof-of-concept study (179). Alpha-gal epitopes do remain in wild type, but not knockout, porcine lungs following decellularization (179). However, the host immune response to subcutaneous implantation of these knock-out tissues in Old World nonhuman primates (those possessing, like human, intrinsic anti-gal antibodies) was similar over 8 weeks to the response to alpha-gal-containing scaffolds (212). Thus, since α-gal negative and α-gal positive porcine scaffolds elicit similar immune responses in nonhuman primates, it may be that additional xenogeneic antigens still serve as a barrier to making porcine lung scaffold transplantation feasible in humans. In addition, more work needs to be done to evaluate the effects of culturing human cells on xenogeneic decellularized scaffolds, since at least one study has demonstrated that species mismatch between cells and scaffold may be associated with reduced cell engraftment and proliferation, and upregulation of inflammatory markers in seeded endothelial cells (14).

A recently proposed alternative to complete lung decellularization is that of targeted epithelial decellularization and recellularization, which may be accomplished via delivery of decellularization solution to the airway compartment. In rodent (55), pig (166), and human (92) lungs, selective instillation of detergent into the airways, followed by epithelial cell seeding, has been shown to be feasible. This strategy has several advantages, including maintenance of a patent and complete vasculature for organ perfusion, as well as barrier function (together with the preserved alveolar-capillary BM). In addition, lobar or segmental repair is theoretically feasible, making this a potential means for adjunctive repair in lungs undergoing ex vivo lung perfusion prior to transplantation (92). However, it is unclear which lung pathologies might be conducive to repair by this approach, and further work is needed on this topic.

Barrier function in engineered lungs

While much progress has been made within the field of whole lung tissue engineering over the past decade, there has been surprisingly little attention paid to robustly building and characterizing the complex barrier functions of the engineered alveolar-capillary membrane itself. Rather, the focus as relates to “barrier function” has largely been on achieving gross separation of the air and blood compartments within engineered constructs, given that the most apparent mode of failure of engineered lungs upon implantation is alveolar hemorrhage and edema, which to date still limits in vivo gas exchange to just hours or days (54, 84, 125, 168, 176, 189, 209, 277). While achieving (i) blood-gas separation is certainly requisite for gas exchange in an engineered lung, the design of the barrier in this context must necessarily fulfill a range of other criteria, as has been discussed in the first sections of this review and which include (ii) vast alveolar and capillary surface areas, (iii) thinness to permit gas diffusion,(iv) strength to withstand repetitive and dynamic transpulmonary and transmural pressures, (v) compliance to enable effective ventilation and to prevent alveolar collapse/maintain alveolar patency, and (vi) active maintenance of an air-liquid interface (Table 1). Many of these individual criteria are in turn achieved only by the combination of several different design features of the lung. Thus, evaluating barrier function of whole engineered lung tissue must necessarily entail a wide range of assessments (Table 2).

It is generally presumed that the various detergent-decellularized whole lung scaffolds that are used as the starting platform for nearly all engineered lungs fulfill the criteria for sufficient surface area by maintaining the native lung architecture. In addition, it is presumed that preservation of the main extracellular matrix components via decellularization results in adequate initial strength and at least part of the requisite compliance for lung engineering. Histologically, decellularized lungs do retain the complex alveolar meshwork of native lungs, and alveolar size and number in decellularized organs have been found to be not significantly different from that of native lung (168, 209, 277). However, decellularization protocols that are successful at removing intact cells by current methods almost inevitably alter the composition of the underlying extracellular matrix (52). Histological and immunofluorescent staining generally suggests that collagen I is well preserved in decellularized lung scaffolds. But elastin, collagen IV, laminin, and fibronectin are retained only to varying degrees, depending on decellularization protocol. PGs are consistently significantly depleted, both by Alcian blue staining and by quantitative assays (15, 53, 175, 176, 182, 233, 238, 277). [Importantly, at least some of the PG depletion can be attributed to the loss of cellular PGs, and so may not constitute entirely loss from the matrix (219).] The consequences of this depletion of PGs have not been specifically investigated, but might include decreased resistance of the alveolar wall to passage of charged particles, and reduced stiffness. Assessments of the gross mechanical properties of decellularized lungs have demonstrated a consistent, decreased whole lung compliance across different decellularization protocols (53, 168, 176, 182), which may be attributed at least in part to the loss of surfactant. Ultimate tensile strength (UTS), defined as the stress at which tissue failure occurs under tension, also varies with decellularization regimen but in some cases is similar to that of native lung (15, 175, 176). Interestingly, one decellularization protocol that depleted 60% of elastin and greater than 90% of sulfated GAGs demonstrated mechanical properties not significantly different from those of native lung (175), suggesting that sufficient gross scaffold strength may be achieved even with significant matrix depletion.

Moving forward, it will be important to more definitively characterize the differences in extracellular matrix content between native and decellularized organs, and to determine the relevance of these differences, recognizing that introduction of cells to the scaffolds will further alter the extracellular matrix structure through both matrix breakdown and synthesis. With this information, ideally the field could come to a consensus in terms of decellularization protocol and thus starting platform for whole organ regeneration. Toward this goal, Hill et al. (99) recently developed a targeted proteomics approach for extracellular matrix that enables quantitation of those insoluble proteins that have commonly been discarded following tissue digestion, particularly the fibrillar collagens but also a substantial fraction of BM constituents. These techniques have been applied to compare the extracellular matrix of native lung and of acellular lungs produced by two decellularization protocols. In accordance with other studies (15, 175, 238), the choice of detergent(s) was found to have a substantial effect on matrix retention. However, it is encouraging that a “gentle” (here, sodium deoxycholate (SDC)/Triton X-100-based) protocol may retain near-native quantities of the majority of proteins tested, including, notably, PGs (41). It will also be important to more robustly test the macro-and micro-scale mechanical properties of decellularized and recellularized lungs, to ensure a safety factor with respect to the range of physiologically relevant forces (218). Tests of failure could include stress-strain testing of tissue strips to tissue fracture, or perfusion assays with transmission electron microscopy correlation to determine critical pressure ranges for stress failure of the alveolar capillaries (Table 2).

Achieving the other design criteria for this blood-gas barrier—blood-gas separation, thinness, the bulk of compliance, and active barrier maintenance—relies on effective recellularization, primarily of the epithelial and endothelial compartments. To date, rigorous characterization of the engineered capillary endothelium and alveolar epithelium in whole lung constructs has often been lacking, with an emphasis placed largely on demonstrating phenotypic marker expression and overall cellular engraftment, distribution, and morphology, without assessment of function. In addition, the largely endothelial-biased focus on barrier function, while leading to the significant advances in lung scaffold revascularization discussed above, neglects the critical functionalities provided by the alveolar epithelium, which include surfactant secretion and vectorial ion transport, in addition to serving as the tightest barrier component for blood-gas separation. Nevertheless, several of the re-endothelialization studies have examined multiple metrics of endothelial barrier or functionality, including cell coverage of the BM (54, 125, 189), cell junctional marker expression (125, 189, 194), antithrombotic markers or function (125, 189), vascular resistance (54, 189), and assays to assess particle leak out of the vasculature (189, 194, 211). Two comments are necessary here as relate to some of these published assessments of vascular barrier function. The first is that “endothelial cell coverage,” while providing an important metric for cellular distribution within the decellularized scaffold, should not be taken as indicating a concomitant level of barrier function. The same level of histologic cellular coverage across different reseeding conditions, such as in the RLMVEC and RLMVEC-ASC conditions in Doi et al. (54), does not necessarily correlate with similar in vivo blood-gas barrier function. Second, caution should be taken in interpreting the results of the published intravascular dextran permeability assays (189, 194, 211). These assays involve the perfusion of a dextran- or bovine serum albumin (BSA)-containing fluid through the pulmonary artery, with quantification of percent particle efflux via the pulmonary vein (as opposed to leak across the alveolar-capillary barrier and out the trachea or across the pleural surface) as a metric of vascular air-blood barrier function. However, for such assays, a positive backpressure is artificially created on the trachea, with the effect that the calculated values for intravascular retention are likely overestimations.

Successful recellularization of acellular scaffolds remains the major challenge for the field. Advancement will rely on careful and precise assessment of engineered constructs, with an eye toward achieving not just high cellularity and survival, but ultimately a tissue with all of the functionalities requisite for achieving sustainable gas exchange. Suggested assessment methodologies for each of the stated design criteria are outlined in Table 2, which is meant not to be comprehensive, but rather as a starting point for thinking about strategies to better assess engineered lungs and to determine readiness for in vivo implantation. In general, in vivo correlations to histologic assessments will prove most valuable, however, the continued challenge of gross failure of blood-gas separation may currently limit meaningful assessment of those other more subtle aspects of alveolar-capillary barrier function.

Conclusion

The pulmonary blood-gas barrier is more than simply a multilayered membrane separating blood and air. Rather, the barrier is a topographically complex, dynamically regulated structure that achieves gas exchange despite its delicate structure, and in the setting of numerous physical and biological stresses. Efforts to engineer the blood-gas barrier ex vivo must consider all or some of the associated design constraints.

Lungs-on-a-chip are evolving as increasingly valuable platforms for studying complex multi-cellular or multi-organ interactions, and for assessing perturbations thereof. Starting out from simpler coculture experiments, more recent chips have incorporated multiple cell types, mechanical cues (shear stress, cyclic strain, ALI), and challenges such as induction of inflammation. Further work remains to be done to enhance the bio-mimicry of these systems, including the choice of membrane and incorporation of alveolar cells such as AEC1s. Importantly, additional metrics for assessing barrier function in these systems, and comparing barrier between systems, are needed.

Since the first reports of engineered whole lungs were published nearly a decade ago, much progress has been made. Many of the initial technological challenges associated with this platform have been surmounted: decellularization of lungs from small and large animals, as well as from humans, is robust and repeatable. Cellular seeding strategies have been developed for the unique pulmonary vascular system. Bioreactors have become increasingly sophisticated so as to better support and monitor ex vivo organ growth. During the same time period, tools such as quantitative matrix proteomics and single-cell RNA sequencing have been developed, which greatly improve our ability to interrogate complex biological systems. Thus, the focus of engineering efforts may now turn to recapitulating within decellularized lung scaffolds the complexity of biological features comprising the native blood-gas barrier. In addition, particularly as efforts continue toward the engineering of human-sized lungs, there is a critical need to develop, and implement, strategies for rigorous assessment of engineered barrier function at both a whole-organ and microscopic scale.

Didactic Synopsis.

Major teaching points

1. Key structural features of the blood-gas barrier:

   1. Extreme thinness
   2. Large alveolar surface area
   3. High density of vessels within the pulmonary capillary “sheet”

2. Lungs have a complex pressure-volume relationship principally governed by tissue stretch and surface tension.

3. Biological aspects of blood-gas separation:

   1. Cell-cell junctions, particularly epithelial tight junctions
   2. Alveolar fluid clearance
   3. Surfactant, which aids fluid balance across the pulmonary capillaries

4. Collagen I, elastin, and proteoglycans provide the bulk of lung parenchymal structural integrity. Collagen IV is the primary source of alveolar-capillary membrane strength.

5. Lung-on-a-chip:

   1. Advances: Chips with multiple cell types, air-liquid interface, and cyclic strain; multi-organ chip systems
   2. Challenges: Physiologically appropriate membrane, resistance measurements comparable across platforms
6. Engineered whole lungs:

   1. Advances: Decellularization and culture of humansized lungs, improved endothelial recellularization
   2. Challenges: Achieving blood-gas separation, developing/implementing methodologies for assessing barrier function.