The airway is the path to the lungs during respiration, composed of the trachea and bronchus, with 23 branches extending from the trachea to the terminal sacs. As the lumen of each branch of the alveolar duct becomes thinner, the epithelium gradually changes from pseudolaminated to single-layer ciliated columnar epithelium, where the numbers of goblet cells, glands, and cartilage decrease, and the amount of smooth muscle increases. Diastolic smooth muscle contraction can change the diameter of the airway, regulating the volume of air in the alveoli. The airways communicate with the external environment, and airway epithelial cells serve as the first line of defense against particles, pathogens, and toxins (25, 26). Smooth muscle spasms and lumen narrowing cause dyspnea in some pathological conditions, such as asthma and COPD.

Alveoli are the basic functional units of the lungs and the main component of air exchange. In adults, there are approximately 700 million bilateral alveoli, accounting for 90% of the whole lung (27); each one is a physiological unit for gas exchange between air–blood interface, composed of alveolar type I and II cells. Alveolar type I cells, covering approximately 95% of the entire alveolar surface, contribute to gas exchange with the associated endothelium. Whereas, alveolar type II cells are characterized by secretory organelles participating in innate immune responses and producing surfactants to maintain surface tension. Individual alveoli maintain good elasticity, with volumes increasing by approximately 15% during respiration. Adjacent alveoli are interdependent and interconnected through small pores. When one alveolus collapses, the tension of the surrounding alveolar walls increases, limiting further alveolar collapse and increasing stability via interdependence. The respiratory membrane, also known as the air–blood barrier, is important in maintaining the basic structure and microenvironment of the lungs (28, 29). The main cell components of the respiratory membrane are type I and type II alveolar epithelial cells, endothelial cells, and fibroblasts (29, 30). Lung diseases that thicken the respiratory membrane or increase the diffusion distance, such as atelectasis, emphysema, lobectomy, and capillary closure and obstruction, can slow the rate and amount of gas diffusion, reduce the respiratory membrane diffusion area, and affect pulmonary ventilation.

The alveolar interstitium is the space between the alveolar epithelium and capillary endothelial basal layer, including a variety of cells and ECM networks, such as elastic and bundled collagen fibers (31), which impart compliance and elasticity to the lungs, respectively, affecting the amount of air inhaled and the ease with which the lung retracts to its resting position after inflation (32, 33). Elastic fibers exhibit a linear stress-strain relationship over a large range facilitating lung parenchyma elastic retraction and stabilization (34, 35).

As a key component of the alveolar air–liquid barrier, the membrane structure in the lung-on-a-chip has undergone three stages of development ( Figure 2A ). Early models used Transwell membranes (36). Costa A et al. (37) and Bengalli R et al. (38) grew alveolar epithelium (NCI-H441) and pulmonary microvascular endothelium (HPMUC-ST1.6R) cells on both sides of a Transwell membrane (39). The simple alveolar respiratory membrane model was used to evaluate the translocation of nanoparticles in biofilm structures and pneumonia induced by ZnO nanoparticles. However, Transwell membranes cannot be combined with microfluidic devices and lack the flexibility required for bionic design. Organosilicone-based polydimethylsiloxane (PDMS) and some thermoplastics are increasingly being used as membrane materials for lung-on-a-chips (40). PDMS is the most studied and representative membrane for ventilation and nutrient exchange (19, 39, 41). It is transparent, oxygen-permeable, stretchable, and flexible allowing the precise imitation of the alveolar dynamic mechanical deformation caused by breathing (13, 42). Other polymer films have also been widely applied and are easy to manufacture, flexible, and cost effective (43). Guan et al. (44) used polycarbonate (PC) as the membrane structure of an air–liquid interface. Air–liquid exchange membranes have also been built using polyester (PET) (45) and polymethylmethacrylate (PMMA) (9, 46). High-throughput lung organ chips with the same basic structure and environment but different cell types can integrate multi-group studies (45).

Membrane structures have some common characteristics. The pore size of the respiratory membrane is 1–10 μm, which not only ensures the exchange of nutrients and protein signal interactions of the epithelial-endothelial cells but also prevents the leakage of cells from both sides. Furthermore, to improve biocompatibility, porous polymer surfaces are typically modified with ECM-like protein materials (e.g., collagen, gelatin, and bovine fibrin). Recent fabrication techniques, such as 3D printing and bioprinting, may be valuable in achieving customized lung-on-a-chip membranes with improved biomimesis (43, 47, 48). Yang et al. (49) used PLGA nanofiber membranes as a substrate to form a uniformly sized porous network and found that different membrane thicknesses and PLGA concentrations affect the membrane permeability and ion diffusion. The printed fiber film can be adjusted to meet the design specifications and 3D-bioprinted lung models show a higher rate of proliferation and longer culture time (28 days) compared with 2D models (47).

The microporous polymer membrane acts as a barrier between the epithelial and endothelial cells, ensuring nutrient exchange and protein signal interactions between the epithelial-endothelial cells and also preventing the leakage of cells from each side. Further, the development of the lung-on-a-chip benefits from improved fabrication techniques that precisely control cell arrangement. Bioprinting technology enables the automated deposition of cells and biomaterials in 3D for highly controlled and customized production of tissue models. In constructing an alveolar model with bioinks (e.g., ECM-like hydrogel), fibroblast cells can be deposited in position with high precision. In addition, growth factors or cell inhibitors can be embedded in hydrogels to influence cell growth.

The arrangement of alveolar cells is a basic functional unit in the lung-on-a-chip. The typical cells found at the alveolar-capillary interface can be divided into four types: type I and II alveolar epithelial cells (A549, HPAEpiC, and NCI-H441) (43, 47, 50), microvascular endothelial cells (HUVEC, HUC-5A, HPMEC-ST1.6R, HPMEC, and Ea.hy926) (15, 41, 43, 51, 52), interstitial fibroblasts (HFL1 and MRC5) (41, 45), and other cells associated with disease models (THP-1, PMBC, and phLFs) (42, 43, 50). Endothelium/stromal cells/epithelium is the most common model arrangement, which can also involve immune cells to stimulate lung infection in vitro (53, 54). The interactions between the endothelial, epithelial, and immune cells provide quantitative data on unknown parameters in pulmonary infection using real-time live-cell imaging (55, 56). However, the cells used in these models grew on a semi-permeable membrane surface without a uniform arrangement, and a more biomimetic model was needed. Kang et al. (20) used inkjet printing to achieve a high-resolution arrangement of four cell types into a three-layer structure: pulmonary microvascular endothelial cells (HUC-5A) were printed on the bottom layer, pulmonary fibroblasts (MRC5) on the middle layer, and alveolar epithelial cells of type I and II on the top layer in an orderly manner, forming a respiratory membrane structure. The bionic respiratory membrane thickness was only 10 µm. To simulate the alveolar sac, DiHuang et al. (57) created an inverse opal structure using methylacrylate. The inverse opal has a vesicle and micropore connection structure highly similar to that of a human alveolar and forms a functional monolayer epithelium when filled with primary human alveolar epithelial cells. This vesicle-like structure resembles the physiological structure in the human body more closely than the previously studied planar structure ( Figure 2B ).

Cell type selection is important to the design of the lung on-a-chip. Cell lines and primary cells are both used; for example, the vascularized lung tumor-on-a-chip model consists of primary human umbilical vein endothelial cells (HUVECs) at passages 3–5, primary normal human lung fibroblasts (NHLFs) at passages 3–5, and A549 (human lung adenocarcinoma) cells (15). In a model of SARS-CoV-2 induced lung injury, the alveolar-capillary barrier is composed of human alveolar epithelial, vascular endothelial, and immune cells. In the model, the epithelial cells used are the immortalized human alveolar epithelial cell line (HPAEpiC), the vascular endothelial cells are the human lung microvasculature cell line (HULEC-5a), and the immune cells are primary isolated peripheral human blood mononuclear cells (24). Whether to use primary cells or cell lines has been extensively discussed in previous review articles (58, 59). In general, primary cells possess characteristics similar to the phenotype in the native environment, which is preferable in lung-on-a-chip models. However, it is difficult to maintain the functionality of primary cells over an extended period of culture time. Cell lines have been widely utilized due to their facile handling and growth, although they are limited by exhibiting similar functions in the original lung and airway.

Cellular organization has become more refined and closer to the native physiological structure; however, some biomimetic deficiencies still persist. For example, type I and type II alveolar epithelial cells are unlikely to stay in one alveolar space; they can pass through the alveolar interval and the epithelial tissue lining structure, and contribute to the secretion function on both sides of the alveolar interval (29). Through the use of secretory organelles, type II alveolar epithelial cells secrete surfactants that maintain the surface tension. However, to our best knowledge, this level of structural and functional complexity has not been replicated in in vitro alveoli models to date.

Dynamic respiratory motion is also a key component in lung-on-a-chip construction (60). A typical breathing-mimic model consists of a 2D planar stretching surface controlled by a vacuum either side of the airway perpendicular to the membrane. Ingber et al. constructed porous membranes connecting both sides of the airflow chambers in a lung-on-a-chip using soft lithography and chemical etching. When a vacuum was applied, the constriction and deformation of the chambers induced stretching and deformation of the porous membrane attached to the epithelium/endothelial tissue. The porous film returned to its natural state when the vacuum was removed. The physiological parameters of respiration in alveoli are a cyclic strain frequency of 0.2 Hz and tensile strain of 10% (13). Compared with the Transwell static air–liquid culture, this vacuum method is a significant improvement in respiratory behavior simulation. However, plane stretching cannot completely simulate alveolar expansion, and an arc-shaped expansion movement is more bionic. Furthermore, the alveolar-like breathing motion has been gradually developed in the lung-on-a-chip, bringing the model closer to physiological pulmonary ventilation and alveoli breathing-induced stretching activity ( Figure 2C ).

The air–liquid interface separating the air chamber from the blood chamber serves as the structural foundation for pulmonary gas and nutrient exchange (31, 54, 55). The air and blood channels can be integrated into a lung-on-a-chip using microfluidic channels. Models containing microfluidic channels are dynamic bionics and closer to the human respiratory membrane than the traditional semi-permeable membranes ( Figure 2D ).

In the lungs, the blood flows through the pulmonary microvascular networks. The lung-on-a-chip design must take two factors into account. First, the shear stress of circulating blood on the endothelium as it flows through the vascular system and across the cell surface. Endothelial cells are subjected to a shear stress in the range of 1–10 dyn/cm² (56), which regulates cell behaviors including proliferation, differentiation, cell information interaction, and barrier formation (61). Second, the pulmonary microvascular network plays a significant role in the immune response. Neutrophils and other immune cells circulate in the blood and create vascularized immunity. The airflow channel in the lung-on-a-chip not only simulates respiration and communication with the outside environment but also contains different substances (e.g., small particles of PM 2.5 dust, viruses, and bacteria) to model the inhalation causing different lung diseases, such as asthma (62), COPD (63), influenza (64), pneumonia (25), and tuberculosis (65). Airflow and blood flow have been integrated into the same chip to reproduce both air and blood transport. Miller and Stevens (66) used projection stereolithography to develop 3D multivascular transport regimes. Oxygenation and human red blood cell flow during tidal ventilation have also been explored.

A nutrient supply is essential for the growth and functional expression of a variety of cells on the air–liquid interface. The basic medium is available during the culture of a lung-on-a-chip, and growth factors can also be used to meet the endothelial, epithelial, and immune cell needs. In particular, the patient’s serum can be used for the culture component in the blood flow channels. The lung-on-a-chip often contains different chambers and microfluidic channels; the medium can be confined to certain cell chambers and different chambers can be connected according to the cell needs.

Various characterization techniques have been used to analyze the lung-on-a-chip models. Cell-cell interactions and microenvironment and physical parameters can be detected to verify the physiological conditions and pathological progress in vivo. Most conventional techniques in cell and tissue biology can be used as off-line bioassays for the lung-on-a-chip, including fluorescence staining, western blotting, and PCR (20, 67). For example, immunofluorescence microscopic images using specific antibodies: the endothelial marker CD31 derived from HULEC-5a cells, the tight junction protein ZO-1, and hydrophilic surfactant SP-A secreted by type II alveolar cells. The alveolar barrier model was stained with hematoxylin & eosin and SARS-CoV-2 infection was predominantly identified in the epithelium layer by viral spike protein expression. RNA-seq analysis showed distinctive immune responses to SARS-CoV-2 infection in cocultured HPAEpiC cells and HULEC-5a cells on chips (24). To analyze the biochemical changes in the microenvironment of the lung-on-a-chip, chemokines and cytokines can be collected. For example, the medium is collected to compare the levels of IL-8, IL-6, IL-1 β, MCP-1, and GM-CSF secreted by hAECs in the chips (57). Chiu et al. constructed a signal amplification sensing film to detect the cytokeratin 19 fragment (68). The lung-on-a-chip can have built-in electrodes for TEER biosensors and micro-impedance tomography. Oxygen, temperature, and lung disease biomarker sensors also measure the physical parameters of the lung-on-a-chip (69, 70). For example, an organ- and disease-specific in vitro mini lung fibrosis model equipped with noninvasive real-time monitoring of cell mechanics has been introduced. The real-time measurement of cell/tissue stiffness and compliance is a clinical biomarker of the progression/attenuation of fibrosis upon drug treatment, which is confirmed for inhaled Nintedanib—an antifibrosis drug (43). Gao et al. designed a giant magnetoresistance multi biomarker immunoassay that can detect 12 kinds of tumors (71, 72).

Pneumonia is an inflammation of the lung parenchyma including the terminal airways, alveolar cavities, and interstitial spaces. This inflammation can cause endothelial and epithelial cell damage, apoptosis, respiratory barrier disruption, capillary dilation, leukocyte infiltration, and massive release of inflammatory factors.

Alveoli-on-a-chip can be used as a platform for pneumonia research by introducing pro-inflammatory factors (TNF-α) or stimulants (LPS/silica/zinc oxide nanoparticles) into the vascular microchannels to stimulate the airway; most of the irritants are inhaled through airway aerosolization to simulate exposure to biological species in the environment (13, 38, 73, 74). High-concentration ultrafine particles (UFP, less than 100 nm) in the environment are likely to cause respiratory system inflammation and negatively impact health. Fine particles enter the bloodstream via the air–liquid barrier and interact with immune cells, affecting the vascular endothelium and other tissues. Camatini et al. (38) investigated the toxicology of zinc oxide nanoparticles (nZnO) using a Transwell model and discovered that nZnO activates alveolar epithelial and endothelial cells to release inflammatory mediators (IL-6 and IL-8). Immune system monocytes (THP-1) can modulate the epithelial response to nZnO. However, the molecular mechanism of action is not known. Kooter et al. (75) used an exposure model to investigate transcriptomic responses in the epithelium of healthy and asthmatic airways exposed to different copper oxide nanoparticle aerosols.

Viruses, bacteria, and other pathogenic microorganisms are common pathogenic factors in pneumonia; primary influenza viral infection and bacterial coinfection significantly increase mortality. After infection, alveolar damage and massive macrophage infiltration cause inflammatory activation of the lung, which cannot be replicated in traditional models. Endothelial injury caused by primary influenza virus infection and Staphylococcus aureus co-infection was studied in a human alveolar model (76). Co-infection leads to a significant impairment of the endothelial barrier integrity, and immune cell inflammation resulting in lung injury. Moreover, the shear stress of the blood cell and macrophage flow strengthens the barrier function. To model influenza infection in the upper respiratory cortex, Jung et al. (20) constructed an ultrathin high-resolution 3D alveolar model by inkjet printing, and influenza A H1N1 virus (PR8) was blown into the epithelial airway side of the lung-on-a-chip. Both NCI-H441 (type II epithelium) and MRC5 (lung fibroblast) cells were extremely sensitive to influenza infection.

The COVID-19 pandemic poses a serious threat to human health. Research models that replicate organ-level physiology are critical for understanding the COVID-19 pathogenesis. Studies on SARS-CoV-2 virus-host interactions using the lung-on-a-chip support the effective diagnosis and treatment of COVID-19. COVID-19-induced pulmonary microvascular injury and immune response in the lung-on-a-chip demonstrate that viral infection causes endothelial injury accompanied by alveolar barrier damage, which is more severe in the presence of PBMC (42). The model revealed a complex crosstalk among alveolar epithelial, endothelial, and host immune responses that are not readily realized in cell and animal models. The lung-on-a-chip has also been used to evaluate the feasibility of antiviral therapy against SARS-CoV-2 in preclinical studies. Remdesivir is an antiviral compound against multiple RNA viruses and has been approved by the FDA (77, 78). After three days of administering remdesivir into the airway of a lung chip model infected with SARS-CoV-2, viral replication was inhibited, resulting in significant therapeutic effects and alleviation of the barrier disruption. This platform could be used to test drug candidates (anti-inflammatory cytokine inhibitors).

Lung cancer is the second most commonly diagnosed cancer and leading cause of cancer death in 2020, representing 11.4% of all diagnosed cancers and 18.0% of deaths. It is also the leading cause of cancer-related morbidity and mortality in men, followed by that of women (4). It is important to understand the basic biological characteristics of lung cancer, such as infinite proliferation, apoptosis resistance, and migratory movements, to develop treatment strategies. Lung-on-a-chip technology can enable tumors to grow, develop, and interact within their own microenvironment.

Yang et al. (49) established a PLGA-based lung-on-a-chip with cocultured human fetal lung fibroblasts (HFL1) and NSCLC (A549) to evaluate gefitinib sensitivity. Insulin-like growth factor (IGF-1) secreted by HFL1 cells mitigated the inhibitory effect of gefitinib on the EGFR signaling pathway by activating the PI3K/Akt signaling pathway, promoting tumor cell growth, and reducing sensitivity to gefitinib. Furthermore, it was discovered that A549 cells could cause endothelial cell apoptosis or death, followed by tumor invasion when HUVECs were introduced into this model.

Lung tumor-on-a-chip has also used isolated cells from primary lung cancers (CAFs and ECs) (15, 79) to accurately design and reproduce cell-cell communication in the lung tumor microenvironment (80). Isolated CAFs from lung adenocarcinomas were investigated for their effect on A549 cell migration (81). NSCLC cell lines in the lung tumor-on-a-chip are influenced by mechanical stretching (82) and blood fluid flow (78, 83) during lung cancer progression and drug response. The lung tumor-on-a-chip from the primary tumors of NSCLC patients under dynamic perfusion has been used to characterize tumor-immune interactions through autologous tumor-infiltrating lymphocytes (84) and can predict patient specificity for immune checkpoint blocking therapy (85). These studies investigated individual-specific tumor immunobiology and drug responses, which is a potential future study direction for lung-on-a-chip techniques (80).

Lung cancer metastasis is a complex physiological process, and lung-on-a-chip can simulate metastasis by integrating multi-organ chips. A multi-organ-on-a-chip consists of an upstream “lung” and three downstream “distal organs,” and it uses a multi-channel microfluidic chip to mimic the in vivo microenvironment in lung cancer metastasis. Bronchial epithelium, lung cancer cells, microvascular endothelial cells, monocytes, and fibroblasts grow on either side of the biofilm in the upstream “lung,” and astrocytes, osteocytes, and hepatocytes grow in distal compartments, mimicking lung cancer metastasis to the brain, bone, and liver. Furthermore, quantitative analysis is used to replicate lung cancer growth, invasion, and metastasis. Preclinical in vitro models should accommodate the interactions between tumors and immune cells, preferably including individual tumor cells harvested directly from patient biopsies as our understanding of the complexity of lung cancer development and metastasis is limited.

Asthma is characterized by reactive spasms of the small airways, and COPD by chronic damage to the small airways and alveoli. The main characteristics of asthma include chronic airway inflammation, airway hyperresponsiveness to various stimuli, reversible airflow restriction, and a series of structural changes in the airway with a prolonged disease course (86). COPD is the third leading cause of death worldwide, affecting 200 million people (5, 87, 88). COPD is usually caused by the accumulation of harmful particles or gases in the airways or alveoli, and inflammatory cells such as neutrophils, macrophages, and T lymphocytes are involved in the pathogenesis. Small airways are essential for delivering air to the lungs and excreting secretions, and they are the primary sites of exposure to environmental factors; therefore, they are closely involved in such diseases as COPD and asthma (25, 89).

To establish a small airway model that closely mimics the physiological microenvironment associated with COPD, cell culture with an air–liquid interface (ALI) is used. Based on an in vitro COPD model, Chen et al. (36) determined the differences in the expression and characteristics of the autophagic protein LC3B between human normal and COPD small airway epithelial cells. LC3B affects the differentiation of COPD cells into basal, secretory, mucous, and ciliated cells. The spreading patterns and morphology changes of the blood vessels in the airways affect airway remodeling, resulting in irreversible airway obstruction that aggravates asthma (90, 91). To better illuminate the functions of the blood vessels in asthma, Nam et al. (52) fabricated an airway model with 3D printing technology featuring an interface between tracheal epithelium and perfusable blood vessels. This asthma disease model demonstrated that it is possible to imitate the tissue infiltration of immune cells, which is the initiation of an active immune/inflammatory response in asthma patients. Furthermore, asthma enhances the sensitivity of the airways to nanoparticle aerosols (75), possibly as a combined result of a hyperactive airway and inefficient mucociliary clearance mechanisms.

Lung fibrosis is a chronic and fatal disease featuring fibroblast proliferation, abnormal ECM deposition, stiffening of lung tissue, and loss of lung function in the end (92, 93). In addition to using simple ALI models using Transwell membranes under static conditions, an increasing number of pulmonary fibrosis studies have focused on the lung-on-a-chip (13, 94–96). Felder et al. (97) developed a respirable lung chip to examine the effects of human liver growth factor (rhHGF) and physiological-cycle mechanical stretching. Cyclic mechanical stretching significantly hindered wound healing, while rhHGF could partially improve wound healing. These findings help elucidate the complex pathogenesis of lung fibrosis. Sundarakrishnan et al. fabricated 3D bioengineered pulmonary fibrotic (Eng-PF) tissues recreating the pathology of human fibroblastic foci (Hum-FF). This pulmonary fibrosis-on-a-chip incorporated different components to simulate various aspects of IPF, including epithelial injury with bleomycin and cellular recruitment by perfusion of cells through the hydrogel microchannel (98). A more recent study described a comprehensive organ/disease-specific model that recapitulated the key attributes of pulmonary fibrosis and the conditions during inhalation therapy (43). The pulmonary fibrosis chip used in the study not only imitated the microenvironment of alveolar cells but also allowed for real-time measurement of tissue stiffness or compliance, which are key parameters used for clinical diagnostics of the progression/attenuation of pulmonary fibrosis. With such results as enhanced tissue compliance and reduced collagen formation, this study demonstrated the effectiveness of aerosolized nintedanib, an FDA-approved antifibrotic drug, in the treatment of lung fibrosis.