Respiratory Physiology on a Chip

Our current understanding of respiratory physiology and pathophysiological mechanisms of lung diseases is often limited by challenges in developing in vitro models faithful to the respiratory environment, both in cellular structure and physiological function. The recent establishment and adaptation of microfluidic-based in vitro devices (μFIVDs) of lung airways have enabled a wide range of developments in modern respiratory physiology. In this paper, we address recent efforts over the past decade aimed at advancing in vitro models of lung structure and airways using microfluidic technology and discuss their applications. We specifically focus on μFIVDs covering four major areas of respiratory physiology, namely, artificial lungs (AL), the air-liquid interface (ALI), liquid plugs and cellular injury, and the alveolar-capillary barrier (ACB).


Introduction
Recreating realistic features of the lung environment within an experimental model system is among the great challenges of modern respiratory physiology. e experimental task at hand is rendered prohibitively difficult due to limitations in controlling physiological functions, maintaining differentiation and expression of tissue-speci�c cellular functions, preserving the homeostatic cellular microenvironment, and integrating alveolar-capillary barrier, to name a few challenges. Although in vivo and ex vivo approaches have been successful at uncovering various aspects of pulmonary physiology, major concerns revolve around the source and supply of human lung tissue [1]. Recently, micro�uidic devices have shown tremendous potential in developing alternative approaches to in vitro models for biological and pharmaceutical applications [2]. In particular, micro�uidicbased in vitro devices ( FIVDs) offer several advantages over conventional in vitro models, and, as a result, this sophisticated miniaturized technology has been adapted to in vitro lung modules as well [2]. e speci�c advantages and drawbacks of FIVDs for respiratory physiology applications are summarized in Table 1. For a more detailed discussion on other types of FIVDs, readers are invited to consult pertinent reviews on the topic [3][4][5].
e aim of the present paper is to address recent progress undertaken in respiratory physiology using micro�uidic technology. We discuss, in particular, recent FIVDs developed over the past decade that mimic highly intricate lung structures (Figure 1(a)) and their ensuing complex physiological functions. Conventional approaches and their limitations are also brie�y discussed. Here, we consider four highly relevant areas of modern respiratory physiology (Figure 1(b)) and highlight current efforts being put forward to advanced in vitro methods for understanding pathophysiology of human lung injury and diseases.

�. �rti�cial �ungs
e human lung is a remarkable organ optimized for respiratory gas exchange. Until present day, man-made engineered devices have generally failed to achieve the extraordinary efficiency of this delicate structure. e alveolar-capillary interface brings air and blood into intimate contact ( Figure  1(a)), and the total area available for gas exchange is comparable to the size of a tennis court, that is, on the order of 100 m 2 [6]. e extremely dense arrangement of the alveolarcapillary structure creates a surface area to blood volume ratio of 300 cm −1 [7], which maximizes the efficiency of  [7]. Insufficiency or disorders in pulmonary functions, along with cardiopulmonary interventions, o�en require arti�cial respiratory support. Moreover, a need for an advanced arti�cial lung (AL) arises as a bridge-to-transplant, that is, during the waiting period prior to lung replacement surgery. Indeed, patients with severe respiratory disabilities that require whole lung transplant must usually wait for over a year until a donor is found [9]. For such cases, a long-term solution is needed: a stable and permanent AL that can improve life quality and reduce mortality while awaiting transplant [10,11]. Arti�cial lungs are designed to compensate for insuf-�cient respiratory functions of the natural lungs. In other words, these medical devices are required to oxygenate blood and simultaneously remove carbon dioxide. Traditionally, ALs are constructed from cylindrical bundles of hydrophobic hollow �ber membranes packaged in an extracorporal device. For a detailed discussion on the basic principles and traditional engineering approaches to ALs, the reader is referred to detailed reviews on the subject [7,11]. Brie�y, oxygen �ows on the luminal (inner) side of the bundles, while blood �ows through the device, that is, around the �bers. Under such con�guration, diffusion-driven gas exchange occurs: oxygen propagates down a partial pressure gradient into the blood stream while carbon dioxide transfers into the �ber lumen.
Unfortunately, such designs generally exhibit long diffusion distances (i.e., a membrane thickness is typically 10-30 m) together with a low surface area to blood volume ratio (∼ 28 cm −1 ) and hence fail to replicate features of the native lung. e result, as expected, is a low gas exchange rate that hardly matches the requirements for resting metabolic needs, even when supplying 100% oxygen gas through the �bers. Other limitations also jeopardize the efficiency of �ber-based designs. To begin, although �bers are hydrophobic, plasma leakage occurs over time. In turn, protein and phospholipidic clots are created within the device leading to a decrease in gas exchange efficiency [12]. In addition, blood �ow pathways within some AL designs are nonphysiological; they feature relatively high shear forces upon �owing blood that are hard to predict or control. ese forces can cause damage to red blood cells (RBCs) such as platelet activation, stimulate in�ammatory responses, and give rise to blood clotting, all of which result in higher morbidity and mortality rates [10,12]. Finally, thrombus formation requires anticoagulant administration which can cause bleeding and electrolyterelated imbalance that call for additional medical care [12]. Overall, the leading disadvantages of a traditional arti�cial lung include (Table 1) (i) low gas exchange rates due to geometry restrictions of the porous membranes, (ii) short lifespan as a result of materials used in the devices, and (iii) restricted mobility for patients with the requirement for 100% oxygen gas delivered through pressured gas containers.
In recent years, FIVDs have become attractive platforms in an effort to design an arti�cial lung assist ( Figure 2). In particular, micro�uidic systems yield compact devices with micron-size channel diameters and larger surface area to volume ratio, approximating more closely features of the natural lung. Moreover, micro�uidic systems enable culturing cellular monolayers within the microchannels to create a biocompatible environment for blood �ow. Since endothelial cells are known to secrete anticoagulant factors in vivo, coating microchannels with endothelium might improve biocompatibility and diminish thrombosis occurrence, ultimately reducing the necessity for systemic anticoagulant administration. e feasibility of including such endothelial monolayers in micro�uidic AL devices was recently demonstrated [13]. Microfabricated systems also allow for easier prediction and control over �ow parameters, including shear stresses created during blood �ow within the device. Indeed, computational �uid dynamics (CFDs) are commonly used to carefully design the delicate architecture of a micro�uidic AL device [9,10,12,14]. �ne of the �rst groups to utilize FIVDs towards developing an AL was Burgess et al. [13] who fabricated a small three-dimensional (3D) module of poly(dimethylsiloxane) (PDMS). PDMS, a widely used polymer for micro�uidics [15], is a favorable material for ALs. Among its advantages are its high permeability to gas, satisfactory performance with blood-contact applications, and cost effectiveness [12]. Each module contained up to six alternating PDMS layers (<100 m thick) of either blood microchannels or gas pathways stacked on top of each other. is micro�uidic device featured straight channels for oxygen and blood pathways and provided a surface area to blood volume ratio of 1000 cm −1 . However, in order to achieve the gas exchange rate required for resting conditions, around 600,000 F 2: An example of a FIVD of an arti�cial lung (AL). A micro�uidic device mimicking physiological functions of the native lung. Air and blood inlets are shown and branched architecture of microchannels can be noted. Reproduced from Potkay et al. [9] with permission from e Royal Society of Chemistry. channels would have to be incorporated within a single device. Another possible reason for rather low gas exchange rates stems from the large diffusion distances; since gas exchange in the device occurs through the alternating layers, thickness of the layers is a limiting factor for gas permeance. Indeed, gas permeability was shown to decrease with increasing layer thickness [12,13]. In general, permeability of the membrane for O 2 and CO 2 remains critically important, as is impermeability to �uids. To satisfy both requirements, ongoing efforts are invested to optimize and improve gas exchange membranes in micro�uidic ALs [16].
By mimicking the highly branched pattern of vascular networks, micro�uidic lung assist devices can potentially provide more efficient gas exchange and support the metabolic needs of a patient. Hoganson et al. [10] emphasized the importance of utilizing such architectures, compared to straight channels, to enlarge the surface area available and maximize the gas exchange rate. Blood �ow was introduced through a dense, branching network of micro�uidic channels with a thin gas-permeable membrane separating the channels from an air chamber, representing the alveolar space. Simultaneously, the device architecture was optimized to achieve controllable, low shear stresses within the capillary subunits and designed to eliminate plasma leakage by using only silicone surfaces to come in contact with the blood over the entire wetted area. By examining different blood �ow conditions, it was found that a loss in efficiency (i.e., rate of oxygen exchange) occurs when �ow rate is increased. Such results are consistent with a previous study [17], showing that full saturation of hemoglobin is 4 times longer within a 25 m wide capillary, compared to a 10 m wide one. Considering the fact that Hoganson et al. conducted their experiments in 200 m wide channels, a loss in efficiency was to be expected. Overall, minimization of channel diameters in parallel to increased membrane permeability can signi�cantly improve gas exchange efficiency within micro�uidic lung assists.
Further work by Hoganson et al. [14] addressed some of the considerations for matching physiological �ows and preserving shear stresses within the physiological range, while simultaneously mimicking vascular anatomy. e resulting microchannel network design was based on �ow optimality rules (i.e., minimization of viscous resistance), known as the Hess-Murray law, and on mimicking an anatomically accurate architecture (in terms of vessel diameters, bifurcation angles, vessel length, and 1 : 1 ratio between height and width of the channel). Moreover, the device was fabricated using micromilling techniques, allowing for higher precision compared to traditional milling or photolithography methods. In spite of these advantages, signi�cantly higher efficiency of gas exchange was not established in this design relative to previous AL devices. However, in view of potential medical use, multilayered devices could be obtained by stacking several functional units (i.e., each unit consists of the blood �ow microchannels, an air chamber, and the gas transfer membrane). For such purposes, 115 layers would be required to achieve proper function in a patient; however, such number would drastically enlarge the arti�cial device and prevent its implantation within the chest cavity.
FIVDs discussed above all rely on 100% oxygen gas supply. Elimination of such requirement would enable more portability and implantability and help take a leap further towards a long-term suitable AL solution. To this end, a novel design was recently introduced to enable blood oxygenation using ambient air ( Figure 2) [9]. e device was designed to function with partial pressure of oxygen in air on the gas side; in parallel, �uidic pressure drop across the arti�cial capillaries was constrained to physiologic values to allow for further implementation in vivo. Similar to the abovementioned micro�uidic ALs, the device featured branching networks of air-and blood-�ow channels, separated by a thin (15 m) PDMS membrane. By minimizing both capillary height and membrane thickness, gas exchange efficiency for Scienti�ca 5 this AL was found to be comparable with both commercially available and other micro�uidic devices [14]. In contrast, since the objective of the study was to maximize surface area rather than to optimize �ow conditions, shear stresses within the device exceeded the physiological range and thus increased the risk for RBC damage and thrombus formation.
Without a doubt, the impact of FIVDs towards engineering an arti�cial lung is of high value. Although sig-ni�cant improvements are still required both in the design and manufacturing of ALs, micro�uidic devices are contributing toward future clinical applications of respiratory support.

In Vitro Models of the Pulmonary
Air-Liquid Interface e luminal surface of the lungs is populated with a con�uent, uninterrupted epithelial cell sheet that exists as a continuum across the airway tree, from the larynx down to the alveoli [18]. Airway epithelial cells are covered with an extracellular liquid lining layer, which in combination with air on the luminal side of airways, creates the air-liquid interface (ALI) (Figure 1(a)). While in the conducting airways of the bronchial tree, the liquid �lm consists of sol and gel layers surmounted by a surfactant layer, in the alveolar region, this multiphase �lm consists rather of a thin hypophase covered by phospholipid-rich surfactant [19]. e presence of surfactant dramatically reduces surface tension at the air-liquid interface and thus prevents the delicate tissue of the airways from collapsing during expiration and allows alveoli to reopen during the next inspiration. Additionally, lung surfactant �lms are also known to contribute to innate defense mechanisms [20]. e aqueous liquid layer lining is highly dynamic and is reported to vary from approximately 10 to12 m in the trachea, to 2.5 m in the bronchi and �nally 0.1 to 0.2 m in the peripheral airways [19]. For a detailed discussion on properties of the liquid lining layer, including its structure and function, we refer the reader to extended reviews on the topic [21][22][23][24][25]. Below, we elaborate on in vitro airway models mimicking physiological aspects of ALI. Traditionally, in vitro experimental setups have been conducted using isolated cells under submerged conditions [26][27][28]. In these studies, in vitro models are composed of cell populations fully immersed in a culture medium. However, in the speci�c case of pulmonary cells, these submerged in vitro conditions only loosely re�ect the actual physiological environment characterizing the alveolar epithelium ( Figure  1(a)): inside airways, epithelial cells are constantly exposed to ALI. In other words, realizing an experimental setup with cells cultured at an ALI is of critical importance to mimic correctly in vivo airway conditions. In particular, in vitro culture conditions where the apical side of a cellular monolayer is exposed to air are known to induce cellular differentiation and to cause cells to express morphological and secretory phenotypes matching those found in vivo [29]. Of particular importance is the secretion of pulmonary surfactant by type II alveolar epithelial cells [30]. In the case of particle-related drug delivery and cytotoxicity, past studies emphasize the critical role of ALI in determining the characteristics of particle submersion upon depositing on the epithelium [19,31]. For instance, composition, morphology, and physical-chemical properties of particles are signi�cantly altered when suspended in medium due to interactions with medium components.
To deliver an ALI within a cellular environment, macroscopic in vitro models of airways have primarily relied on cell culture inserts in well plates, with membranes permeable to growth media (Figure 3(a)). When epithelial cells are cultured on such membranes, compartmentalization of the cultured cells to each side of the membrane is realized, leading to a separation between conditions set on the apical and basal side of the epithelial monolayer, respectively [30,32]. Such con�guration allows exposure of the upper surface to air (i.e., replicating an airway lumen in an alveolus), while continuously maintaining growth media at the lower surface (i.e., basal side) to support cell viability. Although the abovementioned experimental setups recreate successfully the essence of an ALI, they fail however to replicate the actual cellular microenvironment by neglecting physiological conditions applied on the epithelial monolayer, including respiratory �ows on the cells.
To address the limitations of reproducing an accurate pulmonary air-liquid environment with traditional macroscopic approaches (i.e., well plates, petri dishes, etc.), FIVDs have become increasingly popular in allowing both a precise control of the cellular microenvironment and simulating physiological �ow conditions. To the best of our knowledge, Huh et al. introduced the �rst design of a FIVD that replicates closely the alveolar air space and created the epithelial air-liquid interface [33]. In their study, two microchambers, separated by a porous membrane, were fabricated and small airway epithelial cells (SAECs) were seeded on the porous membrane (Figure 3(b)). is seminal design featured a sandwiched structure, allowing for air�ow in the upper chamber in parallel to providing constant perfusion with growth media on the basal side of the epithelial monolayer. In addition to exposure of cells to air leading to differentiation and expression of secretory and morphological phenotypes, the cellular monolayer was exposed to physiological �ows of both liquid (on the basal side of the monolayer) and air (on the apical side) in an effort to simulate breathing conditions. Results showed that the air-liquid interface induced SAECs differentiation by expressing Clara cell 10-kDa proteins (CC10), a known marker for differentiated and chemically functional SAECs. In contrast, cells maintained under constant perfusion did not express any CC10 over the period of culturing. Moreover, higher integrity of the cell monolayer was established due to tight junction formation, and it was signi�cantly higher for cells cultured at an ALI in comparison to cells cultured solely under liquid perfusion.
To create a stable and viable ALI over long durations (more than three weeks in vitro), parallel efforts have characterized at length growth requirements under constant perfusion for a monolayer of human alveolar epithelial cells (i.e., A549 cell line) in a micro�uidic platform [34]. is cell line is widely used as a model for human type II alveolar epithelial cells [35]. In particular, Nalayanda et al. [36] examined cell viability, monolayer integrity, and surfactant production by cells both under submerged conditions and for air-exposed cultures [36]. It was found that epithelial cells cultured at an ALI exhibit higher cell viability and integrity together with decreased surface tension in response to surfactant production, when compared to cells cultured under conventional media exposure. ese latter �ndings are in agreement with Huh et al. [33] as well as recent studies exhibiting increased production of surfactant [37] and higher transepithelial electrical resistance (TEER) [38] for type II epithelial cells maintained at an ALI. Furthermore, A459 cells were found to resist higher mechanical forces (i.e., �uid shear stresses) when cultured at an air-liquid interface. In a very recent study [38], phosphate buffer saline-based (PBS) liquid plugs were propagated upon both submerged and air-exposed epithelial monolayers, creating �uid shear stress upon the cells. Although the epithelial cells at the ALI did not maintain viability, they remained nevertheless attached to the culture substrate even aer propagating 10 liquid plugs over it. In contrast, cells cultured under liquid perfusion completely detached from the substrate aer no more than 3 propagating plugs. e above �ndings emphasize the signi�cance of replicating air-exposed culture conditions and recreating adequately the air-liquid interface to achieve physiologically-realistic conditions within in vitro models of small airways and alveoli.
FIVDs offer a viable strategy to integrate these biological constraints and simultaneously provide both anatomicallyand physiologically-realistic environments of the respiratory tract.

Liquid Plugs and Cellular Injury in Pulmonary Airways
In the distal airways of the lung (e.g., <1-2 mm in diameter), physiological changes in the thin liquid �lm are known to cause acute lung injury and epithelium damage. e origins of such alterations arise from abnormal physical forces, due to mechanical ventilation or �uid shear stresses, as well as from surfactant dysfunction (i.e., disruption of surfactant metabolism) [39]. In particular, modi�cation in cellular functions is intimately linked to pathogenesis and structural remodeling of the lungs, a consequence of respiratory diseases including asthma, chronic obstructive pulmonary disease, cystic �brosis, and respiratory distress syndrome (RDS) [40][41][42][43].
As mentioned earlier, pulmonary surfactant in the alveolated region of the lungs reduces dramatically surface tension at the air-liquid interface, that is, with magnitudes found between near-zero values (at end-expiration) and approximately 25 mN/m (at end-inspiration), well below values for a simple air-water interface, that is, ∼70 mN/m [18]. However, insufficiency in surfactant production, or conversely surfactant dysfunction, can not only lead to respiratory impairment and airway collapse, it can also give rise to two-phase, airliquid instabilities, creating liquid plugs that occlude small airways and impedes gas exchange (Figure 1(a)) [44,45]. In particular, distension during inspiration of the airway surface can propel the plug further downstream into the airways, ultimately causing plug rupture and airway reopening.
e formation of liquid plugs in small airways can also result from clinical therapies such as mechanical ventilation (MV) and surfactant replacement therapy (SRT) [39]. Yet, these therapeutic strategies are widely used for restoring lung Scienti�ca 7 functions. Liquid plugs are also thought to be used as vehicles for targeted drug delivery [46]. Whether a consequence of therapeutic intervention (e.g., MV, SRT, and drug delivery) or physiological disorders (e.g., surfactant dysfunction or insufficiency), the propagation of liquid plugs along respiratory airways, in conjunction with the progression of airway reopening, are believed to produce localized, yet severe �uid mechanical stresses on the underlying epithelial surface [47]. Both experimental and theoretical studies based on in vivo animal and computational models, respectively, suggest that abnormal shear stresses developed as a result of �uid mechanical stresses during the airway reopening can potentially injure the epithelial cells and lead to a variety of respiratory complications [47][48][49][50][51].
To explore the fundamental �uid mechanics of liquid plug propagation and its damaging effects on respiratory airway walls, experimental investigations have been initiated through the development of in vitro airway models. ese efforts include several bench-top models mimicking physiologically-realistic airways and address speci�cally the characteristics of airway closure and reopening upon the passage of a liquid plug. For instance, thin-walled polyethylene tubes of different radii using lining �uids of different surface tensions and viscosities were used as a model airway to measure the relationship between airway opening velocity and the applied airway reopening pressure [50]. Further, airway models coated with �ewtonian lining �uids of constant viscosity were used to investigate a collapsed airway surrounded by parenchyma and characterize how tethering and �uid forces interact to affect the reopening pressure [52]. Eventually, various types of rigid straight and bifurcating tubes, either dry or prewetted, were fabricated to analyze the physical phenomena of liquid plug transport through lung airways [53]. In parallel to these experiments, several computational studies were also conducted to predict the behavior of liquid plugs in airways [49,51,54,55]. For example, a liquid-�lled �exible-walled channel, taken as an initial model of pulmonary airway reopening, was developed to investigate the progression of a �nger of air through airways closed with liquid plugs [50]. While the abovementioned studies have been successful at examining the physical phenomena characterizing the dynamics of liquid plug transport, they are, however, limited in uncovering the interaction between plugs and the underlying epithelial cell monolayers. ese interactions are known for example to give rise to surfacetension-induced epithelial damage during airway reopening [47].
Given such shortcomings, more advanced platforms of idealized model airways have recently surfaced, including a parallel-plate �ow chamber that features channels lined with monolayers of fetal rat lung epithelial cells (CCL-149) [47]. ere, the reopening of a collapsed airway was mimicked by the progression of a semi-in�nite air bubble in a narrow �uid-occluded channel of the chamber. is study revealed that the progression of a semi-in�nite bubble in a narrow channel lined with pulmonary epithelial cells induces signi�cant injury to the epithelial cell population and that the addition of pulmonary surfactant signi�cantly alleviates such cellular injury. In another work, the magnitude of pressure, rather than the duration of stress exposure, was found to constitute the major determinant in controlling the degree of cellular injury during airway reopening [56]. Although these seminal studies represent a major advance in investigating experimentally �uid shear stress-induced cellular injury, they come short of mimicking accurately the in vivo airway environment. To begin, one of the major differences lies in the lack of mechanical �exibility (i.e., lung tissue distensibility) featured in the experimental models. Moreover, such experimental models are devoid of a collagen substrate beneath the epithelial cells. is latter point is of importance since the pressure exerted during airway reopening could be buffered by collagen proteins, owing to its �exible nature, under in vivo conditions [57]. Finally, size and geometry of the model airways signi�cantly differ from actual in vivo airway anatomy. Altogether, these limitations and others have motivated researchers to explore new experimental strategies to integrate more physiologically-realistic airway environments.
FIVDs have shown great potential towards creating in vitro airway models that combine the generation and transport of liquid plugs with cell cultures, all integrated within a single platform [33,39]. Recently, a microfabricated lung airway model featuring two-phase, air-liquid micro�ows has demonstrated its potential for (i) engineering lung morphologies on a chip and (ii) generating physiological and pathological liquid plug �ows while (iii) modeling a physiologically-realistic ALI, as noted earlier [33]. Outcomes from this study con�rmed earlier �ndings that the pressure associated with airway reopening could be injurious to epithelial cells. However, the major �nding of this work lies in the critical importance of the �nal steps occurring during reopening events, that is, the progression of very thin liquid plugs and their rupture in promoting mechanical tissue injury. In the footsteps of Huh et al. [33], a micro�uidic lung airway model integrated with a computer-controlled on-chip plug generator and equipped with pressure sensors, solenoid valves and �ow meters, was recently fabricated to address the precise control and measurement capability of pressure levels in these microenvironments [38]. is device delivers a solution to obtain reproducible, well-de�ned liquid plugs, and ultimately enables the analysis of epithelial cell response to shear stress associated with liquid plug propagation under physiologic differential pressures. ese recent results indicate that knowledge of local, in vivo pressures can provide a better understanding of the �uid shear stress-induced lung injury. Furthermore, FIVDs have been utilized to investigate the effect of wall �exibility on the plug propagation and the resulting wall stresses in �exible microchannels that closely mimic human small airways [58].
In the past, the development of an in vitro network of the airway tree was thought to be a difficult task to perform using conventional "large-scale" approaches. However, FIVDs have opened the possibility of creating an airway tree network and generating air-liquid interfaces within an integrated single platform [33,36]. For instance, a simpli�ed micro�uidic model of pulmonary airway tree with �ve generations was recently developed to study the motion of liquid plugs arising at lung bifurcations [59]. While this design did not feature an epithelial cell monolayer, the airway tree model did characterize liquid dispersion along the airway branches. In parallel, a micro�uidic alveolar model was recently developed to expose cultured monolayers of alveolar epithelial cells (A549) to combinations of mechanical stresses and surface tension [60]. is study is the �rst of its kind to study the combined effects of �uid mechanical (i.e., liquid �lm motion) and solid mechanical stress (i.e., distension-contraction movements) on alveolar epithelial cells. �o extend the applicability of micro�uidic platforms, an airway model of wound-healing was designed to investigate the regeneration of a wounded lung epithelial cell layer exposed to hepatocyte growth factor [61]. Altogether, these recent micro�uidic-driven efforts have laid the foundation for recreating in vitro models of lung airways that can be useful for exploring liquid plug propagation, lung injury, and wound healing while circumventing many of the limitations seen in conventional large-scale systems.

In Vitro Models of the Alveolar-Capillary Barrier
e functional efficiency of the lung is primarily characterized by the delicate structure that separates alveolar air from capillary blood (Figure 1(a)). is thin barrier (<1 m [8]) is comprised of three discrete layers: the alveolar epithelium, the capillary endothelium, and the basement membrane separating the two cellular monolayers [62]. Besides serving as a barricade between the air and blood side, the alveolarcapillary barrier (ACB) is critical for regulating various physiological functions including (i) gas exchange (oxygen and carbon dioxide), (ii) transport of solutes and proteins between capillary blood and alveolar air, (iii) alveolar �uid clearance, and (iv) liquid homeostasis [1,63]. Since the alveolar lumen surface is susceptible to microbial infection, injury, and in�ammation, the ACB also regulates defense mechanisms by controlling the movement of macrophages and lymphocytes from the interstitium and/or capillaries toward the alveolar lumen surface. Alternatively, the diffusive nature of the barrier may be exploited as well for targeted drug and gene delivery [64]. Barrier integrity is mainly dependent on the tightness of the alveolar epithelium and endothelium for the passively transported molecules [1]. Due to the difference in the composition of the interstitium, one side of the alveolar-capillary membrane is found to be thinner than the other, as evidenced by ultrastructural image analysis of the cross section of a capillary lying within the alveolar wall [62]. e thinner part is mainly involved in dynamic gas exchange processes, whereas the thicker part contributes to �uid �ux and higher resistance against mechanical hydrostatic pressure. Damage to the barrier, as a result of acute lung injury for instance, can cause alterations in epithelial and endothelial permeability [65,66], increased inhaled nanoparticle translocation to the systemic circulation [64] as well as pulmonary edema [62] and injury to epithelial cells [65]. With its large surface area and close juxtaposition with underlying capillaries, alveoli are considered the functional units of the lung [67]. Due to their small size (∼250 m) and the large surface to volume ratio available for air, alveoli are inherently unstable structures since surface tension forces at the ALI have a natural tendency to collapse the lung airspace [1]. e thin liquid lining �uid covering the alveolar epithelium isolates it from direct contact with air while reducing potentially harmful effects of surface tension. e underlying epithelium consists of cuboidal type II cells and squamous type I cells. An overwhelming portion of the alveolar surface (>90%) is covered by type I cells since their principal function is to constitute the thin barrier between blood and air. Although the number of type II cells is about twice the number of type I cells, type II cells make up for only about 7% of the alveolar surface. However, their major function is vital: [the secretion of surfactant, a lipoprotein-like substance, via exocytosis of secretory vesicles termed lamellar bodies (LBs).] Surfactant consists of glycerophospholipids (∼80%), with dipalmitoylphosphatidylcholine (DpPC) as the predominant component, cholesterol (∼10%), and proteins (∼10%) including mainly (SP)-A, SP-B, SP-C, and SP-D. Besides, type II cells are also actively involved in the metabolism of xenobiotics, transepithelial movement of water and ions, and the regeneration of type I cells during normal lung development or following lung injury. Alveolar epithelial cells form tight connections with each other by various types of cell-cell contacts, that is, tight junctions, adherence junctions, and others; the anchoring of such contacts is governed by several groups of proteins [1].
While the crucial role of the ACB in various physiological processes has been addressed quite extensively in the past using in vivo (e.g., animals and human models) approaches [68][69][70], developing an in vitro model system that can mimic the in vivo barrier conditions remains a challenging task. e need for developing such model platforms is motivated by the intricate and delicate nature of alveoli that creates hurdles in understanding cell-cell interactions between different cell types, as well as answering other cell-biology-related questions with reference to the physical barrier itself [1]. In addition, considerations for the ACB as a potential route for systemic delivery of therapeutic drugs, or alternatively as a route for inhaled toxic (nano)particulate matter to access the systemic circulation, call for in vitro ACB models to study quantitatively both (i) drug absorption phenomena and (ii) aerosol translocation processes across the ACB. Although ex vivo approaches (i.e., using excised lung tissues) have been considered useful models for examining lung injury, barrier permeability, and alveolar �uid clearance, the precise determination of cellular processes leading to barrier permeability, particle translocation, and drug absorption mechanisms lies widely out of reach using such methods. erefore, there is a need for convenient, reliable, and robust in vitro cell-based models that not only possess a realistic ACB anatomy, but can also facilitate the study of such physiological processes.
A growing number of cell lines derived from different parts of the airways are available for in vitro ACB studies [1]. Examples include human type II alveolar epithelial cells (A549), human bronchiolar epithelial cells (NCI H-441), rat type-I-like alveolar epithelial cells (R3/1), rat type-II alveolar epithelial cells (L-2), and mouse type-II alveolar epithelial cells (MLE-12 and 15). Among these, the A549 (American Type Culture Collection, ATCC CL-185) line is the most frequently used epithelial model that is immortalized and derived from human lung adenocarcinoma [35]. e morphology and biochemical features of this cell line resemble the characteristics of human pulmonary type II cells in situ. However, the major problem with A549 cells is their incapacity of forming functional tight junctions, that is, tight monolayers of polarized cells. erefore, primary cultures of lung epithelial cells, in conjunction with pulmonary microvascular endothelial cells (PMEC), are recommended in order to reconstruct a more complete alveolar-capillary barrier. Since there are limited sources of human lung tissues, as well as ethical considerations related to the use of human lung and pulmonary microvascular tissues, the majority of studies have been conducted on cells isolated from the lungs of animals, including rat, mouse, rabbit and pig [1]. However, the physiological relevance of the data stemming from such animal models may not be necessarily accurate when considering the speci�c human system, as the precise ACB structure and the actual pathway for physiological processes differ signi�cantly among species [71].
As a result, several research groups have established in vitro cellular models using instead lung epithelial cells and endothelial cells of human origin. For instance, a coculture system of human distal lung epithelial cells (NCI H441) and human pulmonary microvascular endothelial cells (HPMEC) using a �lter membrane was developed to study cellular interactions of the ACB epithelium and endothelium in both pathogenesis and recovery from acute lung injury [72]. A comparable model system has also been used for predicting effects of novel drugs on healthy or in�amed tissues [64]. Similarly, to examine pathomechanisms occurring at the ACB during acute lung injury, a primary coculture system was established by cultivating HPMECs with primary isolated human type II alveolar epithelial cells (HATII) on opposite sides of a permeable �lter support [73]. Most recently, a model made of an epithelial cell line (H441) and an endothelial cell line (ISO-HAS-1) of human origin was used to examine the in�ammatory and cytotoxic responses of monodisperse amorphous silica nanoparticles (30 nm) at the ACB [74]. In spite of such progress, the abovementioned model systems do not incorporate yet other cell types such as macrophages, lymphocytes, dendritic cells, and polymorphonuclear cells that characterize a true ACB. Moreover, these platforms lack the ability to recreate the mechanical properties of the interface (e.g., wall distensibility).
In vitro cellular models based on micro�uidic technology have been envisioned to address the limitations of traditional coculture approaches and mimic the structural, functional, and mechanical properties of human ACB. For example, a multistage FIVD integrated with a suspended porous membrane was recently designed in order to recreate the alveolarcapillary membrane [75]. e functional characteristics of the device were validated by the culturing of human pulmonary endothelial (HMEC-1) and alveolar epithelial (A549) cell lines within the device. Although this work represents a step towards recreating a preliminary model of ACB, the speci�c model lacks the in vivo structure of ACB, such as coculturing pulmonary epithelial and endothelial cells on opposite sides of the membrane. Despite such limitations, the micro�uidic study characterized optimal growth conditions for alveolar epithelial cells cultured at ALI. Most recently, a more sophisticated micro�uidic-based model was designed to mimic the critical functional aspects of human ACB [37] and simultaneously implement wall expansion/contraction to replicate breathing movements. is micro�uidic solution is achieved by culturing human AECs and human PMECs in two closely apposed microchannels separated by a porous and �exible PDMS membrane. By incorporating lateral microchambers on both sides of the ACB, subatmospheric pressure-driven stretching was recreated upon applying vacuum. is model system reproduces complex integrated organ-level responses to bacteria and in�ammatory cytokines introduced into the alveolar space.
e pioneering micro�uidic ACB model of Huh et al. [37] was also found to be useful for nanotoxicological research. Namely, cytotoxicity effects of toxic nanoparticles exposed at ALI were addressed. e FIVD revealed that cyclic mechanical strain accentuates toxic and in�ammatory responses of the luminal surface to silica nanoparticles. Alveolar epithelial cells exposed to ultra�ne (12 nm) silica nanoparticles in the absence of mechanical strain showed little or no reactive oxygen species (ROS) production; in contrast, the intracellular formation of ROS was signi�cantly increased in the presence of mechanical distortion. Similarly, the intracellular levels of ROS in the underlying endothelium were also found to be signi�cantly elevated in the presence of cyclic strain. Moreover, mechanical strain enhances epithelial and endothelial uptake of nanoparticulates and stimulates their transport into the underlying microvascular channel (i.e., translocation).

Conclusions and Future Directions
Developing anatomically and physiologically accurate models of the human respiratory system is of high importance towards applications for lung assist development as well as cytotoxicity and lung injury studies at the alveolar level. For these purposes, micro�uidic systems hold great potential for use as a platform in in vitro pulmonary airway studies; FIVDs are able to successfully replicate the delicate structure of the native lung while incorporating critical respiratory functions. e novel technologies reviewed here have deeply contributed towards understanding cellular behavior and physiological processes at the alveolar level. While the use of FIVDs is rapidly rising in popularity, these platforms will need, in the future, to incorporate additional cell types (i.e., macrophages, dendritic cells, etc.) in an effort to mimic regulatory and immune processes characteristic of the alveolarcapillary barrier. In addition, although signi�cant research is still required before clinical applications of an arti�cial lung can be commercially available and replace current technologies, micro�uidic systems have shown feasibility and demonstrated excessive advancement in the efficiency of arti�cial respiratory support. Overall, respiratory physiology on a chip is a noteworthy technology towards research and development of clinical applications in human respiratory disciplines.