A bionic controllable strain membrane for cell stretching at air–liquid interface inspired by papercutting

Lung diseases associated with alveoli, such as acute respiratory distress syndrome, have posed a long-term threat to human health. However, an in vitro model capable of simulating different deformations of the alveoli and a suitable material for mimicking basement membrane are currently lacking. Here, we present an innovative biomimetic controllable strain membrane (BCSM) at an air–liquid interface (ALI) to reconstruct alveolar respiration. The BCSM consists of a high-precision three-dimensional printing melt-electrowritten polycaprolactone (PCL) mesh, coated with a hydrogel substrate—to simulate the important functions (such as stiffness, porosity, wettability, and ALI) of alveolar microenvironments, and seeded pulmonary epithelial cells and vascular endothelial cells on either side, respectively. Inspired by papercutting, the BCSM was fabricated in the plane while it operated in three dimensions. A series of the topological structure of the BCSM was designed to control various local-area strain, mimicking alveolar varied deformation. Lopinavir/ritonavir could reduce Lamin A expression under over-stretch condition, which might be effective in preventing ventilator-induced lung injury. The biomimetic lung-unit model with BCSM has broader application prospects in alveoli-related research in the future, such as in drug toxicology and metabolism.


Introduction
The lung is the main organ of air exchange in human beings. Lung diseases such as asthma [1] and the Corona Virus Disease 2019 (COVID−19) [2] seriously affect human health. Currently, most newly developed drugs for lung diseases have been studied in animal models [3]. However, research data from animals do not always translate to humans due to the physiological differences of different species [4,5]. A lung model in vitro can provide an effective way to coordinate animal models, which helps to yield data on drug action and translate to human better [6,7]. In other words, the in vitro lung model helps us to better understand and address clinical programs. Therefore, in vitro pulmonary model is urgently needed as a tool for optimizing respiratory disease research [8].
The main functions of the lung are to supply oxygen and remove carbon dioxide. The alveolar septum contains capillary network, elastic fibers and fibroblasts, lung macrophages, and other cells ( figure 1(a)). The capillary network adheres to the alveolar epithelium for air exchange [9,10], while the elastic fibers enhance alveolar resilience during breathing. The complexity is further compounded by the fact that the lung is a highly dynamic organ and its constituent tissues and cells are subjected to various mechanical forces, including cyclic deformation (strain), hydrostatic pressure, and shear stress [11]. Although how alveoli change volume remains controversial, substantial evidence suggests that a combined mechanism including isotropic expansion and shape change is the most typical mode for alveoli to expand in vivo, which means that the alveoli have varied deformation [12].
One of the main challenges in developing complex biomimetic in vitro lung models is simulating the basement membrane [13]. When developing in vitro lung models, the main focus has traditionally been on the 'cell adhesion' of the matrix, which may overlook the characteristics of a biomimetic matrix. For example, in vitro lung models, poly(dimethylsiloxane) (PDMS) membranes are widely used due to their mechanical elasticity [14][15][16][17]. PDMS is chemically inert and has suitable mechanical elasticity [18]. However, PDMS is hydrophobic and cannot adjust stiffness, which is not suitable for cell culture. Although PDMS can enhance wettability and cell adhesion by pre-conditioning (e.g. coating with extracellular matrix (ECM) proteins), the cells often aggregate or dislodge from the PDMS surface due to protein dissociation [19]. Polyethylene terephthalate (PET) and polyester (PS) membranes have also been used to simulate alveolus and airway-on-chips [15,20]. But PET/PS membranes are also stiff, limiting their application to mimic dynamic breathing of the lung. In addition, simulating the different strain on the cells of the lung is also a great challenge. Engineered lung models that recreate in vivo tissue architecture could enhance their ability to mimic different mechanical force to cells of lung tissues that are important for its function [11]. Therefore, there is an urgent need to establish a biomimetic basement membrane with suitable materials (with the ability to regulate local-area strain, air-blood barrier, substance exchange, suitable elasticity) for studying the physiological phenomenon and rapidly developing new therapies.
In this study, we introduce a functional BCSM (figure 1(b)) with pulmonary epithelial cells, and vascular endothelial (VE) cells for cell stretching. Papercutting can be designed on a piece of paper on a two-dimensional (2D) level. The paper forms complex three-dimensional (3D) structures under external stimulation (figure 1(c)). By applying the same logic to our BCSM (figure 1(d)), we designed the BCSM containing a PCL mesh and a Gelatin methacryloyl (GelMA) coating, that mimics the elastic fibers and native basement membrane, respectively. GelMA has cell-conducive and mechanoelastic properties [21][22][23][24] that mimic the microenvironment of the alveolar basement membrane. A PCL mesh structure can be adjusted by high-resolution electrohydrodynamic (EHD) 3D printing (figure 1(e)), which offers superior control over fiber diameter, high stability, and accurate control. The PCL mesh could be designed with different structures using structural topology with circles and arcs of different curvatures for regulating local-area strain of the BCSM. The circle element remains almost unchanged, but the arc element stretches when the PCL mesh changes from a static condition to a stretched condition (figure 1(d)). Our biomimetic lung unit model can accurately control the BCSM to 'breathe' by opposing pressure from controlled pumps (figure 1(f)). This in vitro model offers a potential opportunity to develop specialized, in vitro, human-pulmonary-disease models that could advance drug development.

Biomimetic controllable strain membrane design and manufacture
The BCSM manufacturing process is shown in figure 1(a). EHD 3D printing is an additive layer molding method that produces high-resolution patterns by applying a high voltage between the nozzle and the substrate.
First, PCL mesh was printed as fibrous structure by highresolution EHD 3D printing on a glass plate. The motioncontrol program can precisely control the movement of the printing nozzle, and the PCL meshes of different configurations can be printed. Second, the alcohol was spread on the glass plate and the PCL mesh was removed from the glass plate. The PCL mesh was cleaned with deionized water and placed on a smooth baseplate. Third, a GelMA liquid with 0.5% w/v lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) drop was added to the PCL mesh. Fourth, the baseplate was revolved until the GelMA liquid drop spread out evenly. And using 405 nm UV light to irradiate GelMA for 30 s to make GelMA cross-linked. Finally, we removed the BCSM from the baseplate. A series of topological structures with circles and arcs of different curvatures were designed for mimicking alveolar diverse deformation (figure S1), and fabricated three representative structures of PCL mesh (figures 2(b), (c) and S2). There were different permutations and combinations of the two main elementary units (circles and arcs of different curvatures) which can cause different local-area strain under the pressure condition (figure 2(d)). SEM images of the PCL meshes showed that PCL fibrous stacked to form curves of different curvature (figure 2(c)). EHD 3D printing can print out the structure according to the configuration we designed. We introduced a computeraided modeling method to scrutinize the relationship between the surface strain of the BCSM and the different PCL mesh structures. A finite element model was developed for beamsolid coupling modeling inside the BCSM. The same pressure was applied on the BCSM surface. The results show that the local-area strain of the straight line showed less strain; the local-area strain of the arc with a large curvature line showed more strain; and the local-area strain of the arc with a small curvature line between the local-area strain of the arc with a large curvature and the local-area strain of the straight line (figures S1 and 2(b)).

Biomimetic lung unit model assembly and function
We use the characteristics of the hydrogel to achieve balance by balancing the gravity of the liquid and the surface tension of the liquid for mimicking air-blood barrier (figure 3(a)). We added phosphate buffered saline (PBS) onto the BCSM, and PBS droplet could pass through the BCSM but could not drip (figures 3(b) and S3). Moreover, a tissue underneath the BCSM could soak up the liquid on top of the BCSM ( figure  S4). Because the adsorption force of PBS by the tissue break balance between the gravity of PBS and the surface tension of PBS. The GelMA coating is also wettable and nanoporous, which prevents cells from migrating into/through the BCSM but allows for growth factors, sufficient exchange of nutrients, and cell-signaling molecules between pulmonary epithelial cells and VE cells.
The biomimetic lung unit model consists of two glass and two molded PDMS on opposite sides. The two sides of the BCSM are the liquid chamber and the air chamber, respectively (figures 3(c) and S5). The cell medium cannot drop to the air chamber because of surface tension and air pressure. The 'breathing' process of the lung unit model with different structures is presented schematically in figure 2(d). The BCSM rises with liquid going out and air in, which causes the air chamber to enlarge and the liquid chamber to become smaller. Then, the BCSM is restored to a flat state by liquid going in and airing out. The cyclic mechanical movements mimic human breathing out and in. From figure S6, we can see that the BCSM without the PCL structure exhibits a free expansion. The BCSM with configuration (i) is an approximate triangle from the side view, indicating that the hydrogel expansion is regulated by the PCL structure. The BCSM of configuration (ii) expands to a higher height than the BCSM of configuration (i), because the wave curvature of configuration (ii) is larger than that of configuration (i). The BCSM of configuration (ii) is still an approximate triangle when viewed from the side. The BCSM of configuration (iii) looks like an approximate trapezoid from the side view. Because the structure of configuration (iii) is a straight line in the middle area, the stretching of the hydrogel in this area is constrained. And the BCSM of configuration (iii) still has the same waves as configuration (ii) on both sides, so it expands like the BCSM of configuration (ii) on both sides. The results of these different configurations of BCSM are basically consistent with the simulation results, which can illustrate that our composite designed BCSM has good local strain control ability.
It should be noted that, unlike the traditional air-liquid interface (ALI) (cells get their nutrients from the medium on the bottom side, with the top side facing the air), we inverted the air chamber and the liquid chamber. We found that air bubbles can be generated in the liquid chamber possibly due to temperature or cyclic stretching induced shocks. If it is a traditional ALI culture model (that is, the liquid chamber is below), the air bubbles generated will directly contact the cells, which will affect the cell function and even cell viability. Benefiting from the BCSM has the characteristics of the balance of gravity and surface tension of liquids, we can realize the liquid chamber on the top and the air chamber on the bottom. Thus, the air bubbles above the liquid chamber that does not direct contact with the cells. Moreover, because of the air bubbles' small size, the bubbles also do not move around as the liquid moves back and forth, affecting 'breathing' (figure S7).

Selection and characterization of an optimum hydrogel material
Based on the paper of the properties of basement membrane of the alveoli, the average of Young's modulus of a single alveolar wall is 0.30 MPa [25]. We selected target values such as Young's modulus of ≈0.3 MPa, sufficient values for nutrient exchange, and good cell viability as the most relevant parameters for the optimization of coating material to better mimic the character of the basement membrane.
We used four coatings with different degrees of substitution and concentrations of GelMA to find the components that best match the target parameters. The four groups of materials are GM30 10% (GM to substitute GelMA, 30 to substitute the percentage of the degree of substitution, and 20% to substitute the concentration of GelMA-all subsequent texts are abbreviated in this way if there is no special explanation), GM90 15%, GM300 20%, and GM500 20%. The SEM results show that the greater the degree of substitution with a higher concentration, the smaller the pore area (figures 4(a) and (b)).
Subsequently, we measured Young's modulus of different groups (figures 4(c) and (d)). The Young's modulus of GM300 20% was the appropriate stiffness we needed. In order to assess the water absorption capacity of the GelMA, the dried GelMA were placed in deionized water, then their weights were detected samples. Figure 4(e) showed that GelMA had great water retention capacity and a high concentration of GelMA had more water retention. We approximate that the maximum linear rate of change (detail of calculation method is in figure S8) for the coating of GM300 20% with a PCL mesh structure in figure 2(b(ii)) is 11.6%; and the maximum linear rate of change of the coating of GM300 20% with a PCL mesh structure in figure 2(b(i)) is 7.7% (figure 4(f)). Moreover, a 1D fatigue test was conducted to measure the behavior of the membrane under continued cyclic stretch with an amplitude of 20% linear strain (the specific calculation method of linear strain to surface-area change (∆SA) was in figure S8). No hysteresis or creep was observed for the test period (6 h), indicating that the BCSM can endure cyclic mechanical stretch without any plastic deformation or rupture (figure S9).
Next, we investigated the permeation properties of macromolecules on BCSM. A macromolecule FITC-dextran (500 kDa) was used to permeate the BCSM with different hydrogel coating groups. In the beginning, the upper layer was full of FITC-dextran and the under layer was full of PBS; the BCSM was in between. The under layer became full of FITCdextran at approximately 30 min, 50 min, 3.5 h, and 4.5 h for different BCSMs with GM30 10%, GM90 15%, GM300 20%, and GM500 20%, respectively (figures 4(g) and (h)). The lower the material concentration and the degree of substitution, the faster the velocity of permeability. The molecular weight of 500 kDa can pass through our BCSM very well, which shows that our membrane has a good effect on nutrient exchange and other substances exchange.
Finally, we investigated the effect of the BCSM thickness on driving pressure and static permeability (figures 4(i) and (j)). The higher the material concentration and the degree of substitution, the greater the pressure to drive. The thickness of the BCSM also affects the driving pressure, and the increase of the thickness increases the driving pressure. As the thickness of the BCSM increases, it takes longer to penetrate from the upper layer to the lower layer. In addition, it was found from figure 4(l) that the permeation efficiency was slower at the beginning and at the end. We speculate that it is because the process of FITC-dextran through BCSM is relatively slow at the beginning. The slower permeation efficiency at the end may be due to the fact that the contents of the upper and lower layers are basically the same, and the substance exchange efficiency becomes slower. In order to obtain faster material penetration ability and lighter driving force, we choose 200 µm of BCSM. In addition, the water contact angle of all materials groups are around 10 • (figure S10), indicating that our materials are super hydrophilic and are suitable for cell culture.
The BCSM with GM300 20% and a thickness of 200 µm could penetrate macromolecules, was sufficient for nutrient exchange, and showed wettability, which satisfied our target values.

Simulation of alveolar respiration using biomimetic controllable stress membrane
The VE cells and pulmonary epithelial cells were cultured on the BCSM to research cellular protein change on the cycle mechanical stretch of the equipment ( figure 5(a)). The pulmonary epithelial cells and VE cells show good cytoactive (figure S10) and adhesion (figures 5(b) and S12) on the BCSM. The pulmonary epithelial cells produce vinculin which is beneficial for adhesion, cell stretch, and proliferation.
Yes-associated protein (Yap) has been identified as a sensor of mechanical activity and mediates cellular and transcriptional responses downstream of mechanical forces [26,27]. Through the PCL configuration adjustment of different structures, cells felt different mechanical stimuli, and then had different protein expressions. Configuration (ii) (in figure 2(b)) of PCL mesh is designed as a uniformly strain alveolar strain structure. Immunofluorescence showed no significant difference in Yap1 expression between the inner circle and the outer circle ( figure 5(c)). Configuration (iii) (in figure 2(b)) design of the inner area has a small strain and the outer area has a large strain. Immunofluorescence results showed that Yap1 expression was lower in the inner circle and higher in the outer circle ( figure 5(d)). Yap1 is also differentially expressed in the region where the straight lines meet the arc (figure S13). Our results show that different configurations can simulate alveoli with different strain states.
During normal human breathing, a healthy lung typically inflates at a frequency of ≈0.20 Hz (12-15 inhalationexhalation cycles per minute at rest), resulting in increased alveolar size and surface area. During vigorous exercise the respiratory frequency can increase to approximately 0.55 Hz (26-33 breaths per minute). The alveolus undergoes volume fluctuations which have been estimated to cause an 8%-25% ∆SA distension of the alveolar basement membrane [28]. However, mechanical ventilation causes alveolar overexpansion (about 25%-44% ∆SA) which can cause lung injury [29,30]. In order to effectively analyze the effect of stretching on cells at the same strain, we selected BCSMs with PCL meshes of Configuration (i) and Configuration (ii) (in figure 2(b)) to mimic one of the most typical modes for alveoli. We mimic alveolar respiration under the physiological condition (16% ∆SA) (used PCL meshes of Configuration (i)) and a nonphysiologic (36% ∆SA) (used PCL meshes of Configuration (ii)) over-stretch condition with different PCL structures of the BCSM. In order to average the respiratory frequency in different states (normal state and exercise state), we chose 0.33 Hz as our stretching frequency.
Other researchers have reported that mechanical stimulation can increase the expression of Lamin A [31][32][33]. Previous research has shown external forces are transmitted across the cytoskeleton to the nucleus, where they result in substantial deformation [34]. Lamin A is one of the important components of the nuclear envelope, and increased cytoskeletal tension results in decreased Lamin A phosphorylation and higher Lamin A expression levels. The different stretches were applied on the BCSM for 3.5 h to pulmonary epithelial cells ( figure 5(e)). The intensity of immunofluorescence was analyzed on different strain levels (figures 5(f)-(i)). Cells at a stretch of 36% ∆SA have approximately six times the intensity of immunofluorescence than 0% ∆SA; cells at a stretch of 36% ∆SA have approximately three times the intensity of immunofluorescence than 16% ∆SA. In addition, there was no significant change in cell viability before and after stretching (figure S14).

Promoting immune inflammation after stretch
We profiled gene expression in cyclic mechanical stretch and static conditions (figure S14). A total of 1376 genes were significantly changed by mechanical stretch and static (with an adjusted P-value lower than 0.05). In addition, Lamin A gene (LMNA) was overexpressed in cells of a cyclic mechanical stretch compared to static cells (figure S15). Functional enrichment of differential transcriptomic profiles was performed using ingenuity pathway analysis software to determine predicted functional differences between genotypes. This analysis shows that several pathways are related to the inflammatory response between cells of cyclic mechanical stretch and static cells (figure S15). And the polymerase chain reaction data showed that IL-6 and IL-8 were significantly higher after 36% stretching than non-stretching (figure S16).

Construction of the alveolar model for drug screening
We propose a biomimetic alveolar air-blood barrier of the lung unit model with pulmonary epithelial cells and VE cells during breathing for drug screening (figure S17). Moreover, we demonstrated that VE cells can be functionalized (form the tube formation) on the BCSM. This shows that our material has a good potential to functionalize cells. The VE growth factor was used to form the tube formation of VE cells ( figure 6(a)), forming a functional layer of blood vessels. The total meshes area of tube formation is eight times more than the normal total meshes area ( figure 6(b)); the number of meshes of tube formation is 50 times more than the normal total meshes area ( figure 6(c)). In addition, we also constructed BCSM containing confluent VE cells and lung epithelial cells (figure S18). The transepithelial electrical resistance (TEER) of the cell-free BSCM membrane is 357 ± 26 Ω cm 2 and TEER of BSCM with VE cells and lung epithelial cells is 750 ± 83 Ω cm 2 . The apparent permeability (P app ) of cell-free BCSM under static conditions is 1.9 ± 0.59 × 10 −5 cm s −1 .
The apparent permeability of BCSM with VE cells and lung epithelial cells is 3.7 ± 0.67 × 10 −6 cm s −1 .
Mechanical ventilation is a life-support treatment that helps patients, especially those with acute respiratory distress syndrome, to breathe [35,36]. However, heterogeneous aeration produced by mechanical ventilation results in anisotropic inflation, exposing certain areas such as ALIs, especially in damaged alveoli, causing ventilator-induced lung injury (VILI) [30,37]. Blocking nuclear mechanosensation may represent a useful strategy to treat lung injury caused by overstretching under pathological conditions [38]. LaminA as one of the important proteins in nuclear sensing mechanics. Lopinavir/ritonavir is known to inhibit the expression of ZMPSTE24-ZMPSTE24 is a necessary enzyme for Lamin A synthesis, thereby reducing the expression of the protein Lamin A [38]. Therefore, lopinavir/ritonavir was used to treat VILI, and LaminA was used to judge the therapeutic effect.
In order to acquire information on how long the drug could reach the layer of pulmonary epithelial cells, we used rhodamine B to replace the drug (because their molecular weights are almost the same). Rhodamine B needs 10 min through the monolayer cells of the membrane, and 15 min to proceed through the bistratified cells of the membrane (figures S17 and S19). We constructed an in vitro disease model of mechanically ventilated lung injury followed by drug therapy ( figure 6(d)). First, we over-stretched (at 36% ∆SA) the cells for 3.5 h for mimicking VILI. Second, the different amounts of lopinavir and ritonavir (1/0.25 µg, 10/2.5 µg and 100/25 µg) were added to the liquid chamber, and then the drug was through the layer of VE cells and the BCSM. There was no significant difference in cell viability at doses of 100/25 µg of lopinavir/ritonavir and without drug (figure S20). Third, the drug reached the layer of pulmonary epithelial cells. Finally, we kept the situation static for 1.5 h (figure 6(e)). The results show that 100 µg of lopinavir and 25 µg of ritonavir significantly inhibit the expression of the protein Lamin A (figure 6(f)). The fluorescence intensity of the added drug group was concentrated at around 25 (figure 6(g)), and the fluorescence intensity of the control group was concentrated at around 60 ( figure 6(h)). However, the treatment effect of lopinavir and ritonavir (1/0.25 µg and 10/2.5 µg) in the other two groups was not good as that in the group of 100/25 µg during the rest period of 1.5 h (figures 6(i) and S21) The results suggest that protease inhibitors can inhibit Lamin A (figure 6(i)) and might be useful for reducing side effects associated with mechanical ventilation.

Discussion
In recent years, many lung-organ-models increasingly mimic the core features of the lung [14,15,25,28,[39][40][41][42][43], including mechanical stretch and ALI conditions. Many biomimetic lung models use PDMS as a biomimetic basement membrane, but PDMS membranes cannot well simulate the main characteristics of lung cell-supporting extracellular matrix [44]. We used GelMA as the material for mimicking basement membrane, which has good biocompatibility for facilitating cell adhesion. GM300 20% has a stiffness closer to that of a single alveolar wall, and has sufficient values for nutrient exchange and good tensile property, making it very suitable as a biomimetic basement membrane.
Biomimetic basement membranes also have many developments in thickness, stiffness, permeability and bioactivity. Doryab et al developed a hybrid biphasic membrane (composed of gelatin and PCL) for an in vitro cell stretch model of the lung cultured under ALI conditions [45]. Taskin et al electrospun various ratios of six-arm star-shaped poly (ethylene oxide-propylene oxide) prepolymers with isocyanate end groups (NCO-sP [EO-stat-PO]) to stabilize gelatin as a bioactive component within nanofibrous hydrogel scaffolds [46]. Zamprogno et al used biofilms made of lung ECM, collagen and elastin proteins for stretch culture [39]. The materials used in these models all have great properties for biomimetic basement membrane. However, both these models and recent developments lack the ability to regulate local strain to mimic varied deformation of alveolus. We innovatively proposed to control the local strain design of GelMA coating with PCL mesh as the skeleton. Through the topology of simple elements (circles, arcs, and lines), it is possible to mimic the different deformability. More importantly, it provides an idea for composite material design to regulate local strain. We designed the biomimetic elasticity of the BCSM with PCL mesh by mimicking the physiological structure for enhancing tensile strength. We used the structural design to reduce the tension on the material itself, which mimics equally distributed elastic fibers to help the BCSM recover after stretch.
Our BSCM can pass through macromolecule (500 kDa), which shows that it has very good substance exchange performance. It can support the transmission of information between cells and the exchange of nutrients. And micronscale pores do not penetrate cells. The balance of liquid surface tension and gravity was used to construct an inverted ALI (traditional ALI model has a gas chamber on top and a liquid chamber on the bottom). This solves the problem of air bubbles contacting the cells in the sealed liquid chamber. Air bubbles created in our upper liquid chamber do not contact with the cells and do not affect the cell stretching.
The nuclear lamina is composed of several lamin-type proteins that include Lamin A as a key component [47,48]. The nuclear lamina acts as a scaffold for the nuclear envelope, and the reversible depolymerization of the nuclear lamina regulates the disintegration and reconstruction of the nuclear envelope [49,50]. The nuclear envelope has been identified as one of the key cellular mechano-sensing structures that responds to extracellular stiffness, decreasing its compliance [33,34,38]. Lamin A change occurs after mechanical stretch and reduce its compliance. Our results show that cyclic cell stretching promotes IL-8 gene expression and activation and contributes to an inflammatory response.
Our work has limitations that are worth discussing. The in vitro model may not be fully represent alveolar characteristics. For example, our membrane is still too thick (about 200 µm), and the model does not have a multilayer cell structure (like alveolar type-I and type-II cells). However, a biomimetic lung unit model with cellular biological properties, yielding cellculture models that can display 3D architecture, multicellular interactions, air-blood interfaces, fluid flow, and organ-level mechanical cues has broad research significance in the study of the human alveolar respiration mechanism and drug screening. The biomimetic lung unit model can add nanoparticles, viruses, etc. to the air layer and can add drugs, inflammatory cytokines, etc. to the medium (blood) layer, which can mimic different pathological models such as smoking and COVID−19. Future studies can focus on creating effective, multiple, biomimetic lung chips in series that will work for mimicking the multi-alveolus of the lung-something that can allow for the research of the correlation between alveoli in different pathological states and can mimic multi-cell interplay in the lung.

Conclusion
Collectively, a biomimetic lung unit model with BCSM has an air-blood barrier, biomimetic elasticity, and controllable strain cell stretching for cellular mechanisms and drug screening. The BCSM surface strain were controlled by different PCL mesh structures for mimicking diverse alveolar deformation in vivo. A computer-aided modeling method and a tensile experiment were verified the effectivity of the structure, and different protein expressions also effectively verified our model. The BCSM with configuration (ii) had a maximum of 99% ∆SA, and no rupture occurred under 6 h cyclic stretching. The BCSM also had good material exchange capacity (up to 500 kda molecular weight substances can penetrate) and an ALI. The biomimetic lung unit model had great 'breathing' ability in confined space which can be used in the research drug treatment of diseases such as VILI. The protease inhibitors lopinavir/ritonavir can decrease Lamin A protein expression, suggesting that protease inhibitors might be effective in preventing VILI. This study provides an idea for a composite material that regulates local-area strain and can guide the design of in vitro alveolar models. It is believed that this work is helpful to visually observe the working state of the biomimetic alveoli and support the development of new approaches to drug development.

Experimental design
This study aimed to develop an adaptable technology platform for mimicking alveoli respiration by reconstructing a physiological microenvironment that would enable us to explore the physiology of the alveoli and for rapid drug screening. A computational model was developed to simulate the BCSM stretch, and the relationship between the surface strain of BCSM and the structure of PCL mesh was discussed. The purpose of our cells study was to determine whether our model has a microphysiological environment and to prove that our chip creates the near-physiological microenvironment. We engineered a biomimetic alveoli model to create a robust platform for the biologic mechanism and drug screening.

Device fabrication
The Poly(ε-caprolactone) (PCL: Sigma, Mn 80 000) meshes were printed by a high-resolution, 3D EHD printing device (Engineering for life (EFL), Suzhou, China) on 80 × 80 mm glass. We put the glass with the PCL meshes into alcohol for stripping the PCL meshes. We blow-dried the PCL meshes and added GelMA (EFL, Suzhou, China) liquid (contains 0.5% w/v LAP, EFL, Suzhou, China) on the PCL meshes. After using a 405 nm band light source to illuminate for 30 s, we took down the BSCM. For the PDMS (Sylgard, Dow Corning) interlayer, the PDMS base and curing agent were thoroughly mixed in a weight ratio of 10:1. This mixture was then degassed and poured onto a 3D-printed model containing the chamber features of the device to be fabricated. The molded PDMS was cured at 60 • C for 7 h. The molded PDMS was 4 mm thick and contained a circular chamber (22 mm in diameter). The molded glass had four, 2 mm wells designed to exert pressure for sealing.

Permeation characterization method
To measure the permeability in the apical-to-basolateral direction under static conditions. First, the biomimetic lung model was constructed; then, 1 ml of PBS and 1 ml FITC-Dextran (500 kDa, Sigma-Aldrich) were added to liquid chamber and air chamber, respectively. Samples from the air chamber were taken at 0, 10, 20, 30, 40, 50, 60, 90, 120, 150, 180, 210, 240, and 270 min. In order to keep the liquid level consistent (the analysis remains uniform at the same height), the samples were into a cell counting plate (Watson, Japan) and were analyzed (fluorescence intensity obtained for a given reference gain/pinhole) by fluorescence microscope (LSM880, Carl Zeiss; Oberkochen, Germany) and emission wavelengths of 490 and 525 nm, respectively. Fluorescence intensities were analyzed by ImageJ. The other penetration experiments in this article are basically operated in this way.

Water absorption capability
The water absorption property of the gels was determined by the mass method. Gels with the same volume were cured and placed in an oven for drying at 60 • C for 24 h. The dried gels were placed in deionized water at 20 • C, then taken out at different intervals, and placed on a stainless-steel net for 30 s to filter the surface water and weighed.

Computational modeling
COMSOL Multiphysics (COMSOL Inc.) was used to computationally model the 3D-surface strain of the BCSM. The same pressure was applied to the different configurations of BCSM surface, driving the membrane to swell. The biomimetic membrane geometry was modeled as a fixed solid boundary with an average thickness of 300 µm encasing the PCL mesh of a 60 µm diameter.
Solid-beam multiple physical field models were employed to simulate the surface strain of the BCSM, where the beam represented the PCL mesh and the solid represented the hydrogel coating. We chose Euler-Bernoulli beam theory to simulate beam deformation because the mesh had a little twist. The Young's modulus and Poisson ratios measured by the experiment were 0.3 GPa, 0.325 MPa, and 0.46, 0.3, respectively.

Cell culture
Human bronchial epithelial cells (BEAS-2B) (American Type Culture Collection [ATCC] were maintained in a bronchial epithelial cell medium (BEpiCM, Sciencell) with a bronchial epithelial cell-growth supplement (BEpiCGS, Sciencell). Human umbilical-vein endothelial cells (HUVEC) were maintained in Dulbecco's modified Eagle's medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (Gibco) and contained 1% penicillin-streptomycin (Gibco). All culture processes were expanded in a humidified culture incubator maintained at 37 • C and 5% CO 2 . Prior to cell seeding, the model was sterilized by UV irradiation, and the BCSMs were embedded in the central culture channels. BEAS-2B cells and HUVEC cells were added at a density of 3 × 10 5 cells ml −1 and 1.8 × 10 5 cells ml −1 , respectively.
Pulmonary epithelial cells and culture medium were pumped into the air chamber of the biomimetic alveolar model. Then the pulmonary epithelial cells allowed to attach to the BSCM surface for 12 h at static conditions. The biomimetic lung unit model was inverted for pumping EC cells and culture into the blood chamber of the biomimetic lung unit model. The culture medium of the air chamber was extracted to form the biomimetic lung unit model.

Cell-stretch
The BSCM was cyclically stretched by the biomimetic breathing system under the condition of air-liquid interface culture. The biomimetic breathing system consisted of the bioreactor chamber, injection pump, and control system. The bioreactor chamber consisted of an air chamber and a liquid chamber; the membrane separated the two chambers for cell growth. The control system used an Arduino control panel (Arduino Uno R3) and Arduino IDE programming software for assisting the injection pump in accurately controlling the volume of gas and liquid. The specific process consists of instructing the control system to control the injection-pump cyclic motion for a certain volume. When the air is pumped into the air chamber, the same volume of culture medium is pumped out of the liquid chamber, which makes the BCSM inflated. Similarly, the air is drawn from the air chamber and the same volume of culture medium is pumped into the liquid chamber, causing the BCSM to drop. The air in the air chamber while the cell-culture medium pulls away from the liquid chamber by using the programmable injection pump, respectively. This process mimics breathing, with the circulating pulses of pressure induced by air and liquid in and out causing the BSCM (and the cells) to change shape. During cellstretch, the air-injection pump pushed volume V1 of air into the air chamber while the liquid-injection pump drew volume V1 of the culture medium from the liquid chamber. Assuming that the motion of membrane can be approximated as a spherical cap geometry, the corresponding membrane displacement (V 1 , ml) is related to the radius (r) and axial deflection of the membrane (∆h, cm).
To study the role of mechanical stretch on cell physiology, a cyclic mechanical stretching of 16% ∆SA and 36% ∆SA strain was applied to cells at 0.33 Hz to mimic the physiologic and non-physiologic conditions, respectively:

Drug testing
After the cell-stretch was carried out for 3.5 h at 36% ∆SA for simulating VILI, we drew the culture medium out from the liquid chamber and washed the liquid chamber three times using PBS. After that, we added different amount of lopinavir/ritonavir (1000/250 µg, 100/25 µg, 10/2.5 µg) with 3 ml culture medium to the liquid chamber for 1.5 h in the static condition. Cells were then digested from the BCSM using trypsin and centrifuged (3 min, 1200 rpm) for subsequent characterization.

Statistical analysis
All experiments were performed at least three times. The data were analyzed using GraphPad Prism 8 (GraphPad Software, USA). The comparison between the two groups was statistical analysis using the unpaired two-tailed Student's t-test. One-way ANOVA followed by Dunnett's multiple comparison test (vs. control group) or Tukey's multiple comparisons test (vs. every other group) was performed for comparison of multiple groups. Results with a P < 0.05 or smaller were considered to be significant.