Comsol Muscle Force Generation

Membrane

Pauline Zamprogno1, Simon Wüthrich1, Sven Achenbach 1, Giuditta Thoma1, Janick D. Stucki1,2, Nina Hobi1,2,

Nicole Schneider-Daum3, Claus-Michael Lehr

3, Hanno Huwer4, Thomas Geiser5, Ralph A. Schmid6 &

1,5,6✉

The air-blood barrier with its complex architecture and dynamic environment is difficult to mimic in vitro. Lung-on-a-chips enable mimicking the breathing movements using a thin, stretchable PDMS membrane. However, they fail to reproduce the characteristic alveoli network as well as the biochemical and physical properties of the alveolar basal membrane.

Here, we present a lung-on-a-chip, based on a biological, stretchable and biodegradable membrane made of collagen and elastin, that emulates an array of tiny alveoli with in vivo-like dimensions. This membrane outperforms PDMS in many ways: it does not absorb rhoda- mine-B, is biodegradable, is created by a simple method, and can easily be tuned to modify its thickness, composition and stiffness. The air-blood barrier is reconstituted using primary lung alveolar epithelial cells from patients and primary lung endothelial cells. Typical alveolar epithelial cell markers are expressed, while the barrier properties are preserved for up to 3 weeks.

Open

COMMUNICATIONS BIOLOGY | (2021) 4:168 | https://doi.org/10.1038/s42003-021-01695-0 | www.nature.com/commsbio

O

rgans-on-chips (OOCs) are emerging as predictive tissue modelling tools and as a credible alternative to animal testing. These micro-engineered cell-based systems pro- vide cells with an environment that closely resembles their native in vivo milieu1–3. Tissue models of physiologically healthy or pathological primary cells from patients have been established, and are robust enough to permit applications such as drug screening4–6.

Integrated

membrane in a microfluidic setting have been reported to model various barrier tissue interfaces, such as those of the lung alveoli, the brain and the gut7. By implementing a flexible polymeric membrane in such microfluidic systems, mechanical forces, such as those induced by breathing, could be reproduced8–10.

Although these systems represent a crucial advance in cell culture research, they are still far from mimicking the whole in vivo intri- cacy. An important in vivo feature that is not reproduced in air–blood barrier models is the array of tiny alveoli. Indeed, reported lung-on-a-chip from the first generation8,9 simulate a single alveolus with an epithelium area much larger than that of an alveolus in vivo.

Given the close relationship between lung microstructure, mechan- ical forces, alveolar epithelial phenotype and lung functions11,12, the emulation of the alveolar network would be of great benefit. A further limitation of those systems is the use of an artificial basal

Poly-

dimethylsiloxane (PDMS) film. Although PDMS has good elastic, optical and biocompatible properties, it can distort the biochemical microenvironment through high adsorption and absorption levels of small molecules13,14. In addition, PDMS differs in important ways from the molecular composition and intrinsic stiffness of the native extracellular matrix (ECM), which is known to affect cellular phe- notype and homoeostasis15,16. The complex ECM environment provides the structural basis for cellular growth, and influences

Other

traits17,18. The role of the ECM in tissue development and function is closely associated with its composition and properties19. The replacement of PDMS as culturing membrane with a material made of ECM molecules would therefore be a significant step towards emulating in vivo-like tissue barriers and functions.

Hydrogels receive a strong interest to recreate the chemical composition and structure of the native ECM in cell culture systems20,21. Their intrinsic properties, including mechanical features, chemical composition and porosity, make them ideal candidates to supersede PDMS membranes22. However, the creation of thin membranes made of ECM molecules, with stretchable properties to mimic the cyclic mechanical strain of the lung alveolar barrier, is technically challenging. Dunphy et al.23 developed a stretchable, soft collagen–elastin (CE) membrane for tissue engineering. However, with a thickness of about 1 mm, it was developed to evaluate the mechanical properties of the material and not to mimic the air–blood barrier. More recently, collagen vitrified membrane24 has been integrated into micro- fluidic devices for use as cell culture substrates25,26, but, to the best of our knowledge, no stretchable membranes have been reported so far.

Here, we report about a lung-on-a-chip of the second gen- eration that mimics the following central aspects of the air–blood barrier: (1) an array of alveoli, with dimensions similar to those found in vivo and (2) a biological membrane made of proteins of the lung ECM, collagen and elastin, enabling the membrane to be (3) biodegradable and (4) stretchable. The creation of the biolo- gical membrane is based on a bottom-up approach fabrication technique. The membrane is formed by drop-casting a CE solu- tion onto a gold mesh, where it spreads and is maintained by surface tension. The thin gold mesh, with a pore size of 260 µm, is used as the scaffold supporting the array of 40 alveoli. The resulting membrane is stable and can be cultured on both sides for weeks. Its permeability further allows cells to be cultured at the air–liquid interface, and its elastic properties mimic the respiratory motions by mechanically stretching the cells. Results with primary human alveolar epithelial cells from patients cocultured with primary human lung endothelial cells demon- strate that the air–blood barrier functions can be maintained and used experimentally in a resilient and reproducible manner. This proto-physiological membrane opens the way to new lung-on-a- chip and OOC devices that enable the mimicry of biological barriers with a new level of analogy to whole organ systems.

Results

Production of a thin, biological and stretchable membrane. A simple process was used to create the thin biological membrane (Fig. 1). A drop of CE solution was pipetted onto a 2-mm dia- meter and 18-µm-thin gold mesh (Figs. 1c and 2a) made of an array of 40 regular hexagons, with sides of 130 µm separated by 30-µm-wide walls. Once pipetted onto the mesh, the CE drop was maintained on its top by surface tension forces. After a gelation step at 37 °C, the CE solution dries out at room temperature within 2 days. While water evaporates from the drop, surface tension forces and residual forces counteract gravity force

Fst

Fst

Fig. 1 Second-generation lung-on-a-chip: creation of the lung alveoli array. a Schematic of the respiratory tree-like structure ending with alveolar sacs (adapted from https://smart.servier.com/smart_image/ lungs-7/, https://smart.servier.com/smart_image/lungs-11/ and https:// smart.servier.com/smart_image/lungs/; Servier Medical Art by Servier; https://creativecommons.org/licenses/by/3.0/). b SEM picture of a slice of human lung parenchyma with tiny lung alveoli and their ultrathin air–blood barrier (courtesy of Prof. Dr. Peter Gehr, Institute of Anatomy, the CE membrane used in the second generation lung-on-a-chip. A thin gold mesh with an array of hexagonal pores of about 260 µm is used as a scaffold, on which a drop of collagen–elastin solution is pipetted. e–g The collagen–elastin gel forms a suspended thin membrane that can be stretched at the alveolar level by applying a negative pressure on the basolateral side of the membrane. f, g Type I (ATI) and type II (ATII) primary human lung alveolar epithelial cells are cocultured with lung endothelial cells on the thin collagen–elastin membrane. h Schematic of the force balance during the drying of the membrane. FST, FG and σo stand for surface tension force, gravity and residual stress, respectively.

Article

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-021-01695-0

2

COMMUNICATIONS BIOLOGY | (2021) 4:168 | https://doi.org/10.1038/s42003-021-01695-0 | www.nature.com/commsbio enabling the suspended membrane to form (Fig. 1h). During this step, the collagen molecules self-assembled into fibrils which, as the natural ECM, provide a structural support (Fig. 2i). The fibrils had an average size of 45.7 ± 18.9 nm (Supplementary Fig. 1).

Figure 2d, e illustrates the dried CE membrane with a thickness of only a few micrometres that is suspended on the hexagons array. Once dried, the membrane was integrated into a microfluidic chip, where it was sandwiched between two microfluidic parts, a top part in PDMS with an apical reservoir and a bottom part in polycarbonate that formed the basolateral chamber (Supple- mentary Fig. 2). The dried membranes are robust and can be stored for at least 3 weeks at room temperature. The membranes are rehydrated by submersion in cell culture medium for 2 h prior to cell seeding.

Properties of the CE membrane. The thickness of the membrane was evaluated using reflective light. With a CE ratio of 1:1, the thinnest membrane obtained had a thickness of 4.5 ± 0.8 µm for a pipetted CE solution volume of 0.8 µL mm−2 (Fig. 2c). When the pipetted volume was doubled (1.6 µL mm−2), the thickness of the membrane also doubled (8.8 ± 1 µm). A thickness of 11.5 ± 1.2 µm was obtained with 2.4 µL mm−2. Decreasing the elastin con- centration (2:1 ratio) resulted in a reduction of the membrane thickness (Supplementary Fig. 3), to the detriment of its viscoe- lastic properties (Fig. 3c). The membrane thickness was homo- geneous within each hexagon. Variation in membrane thickness across the array was less than 20% (Supplementary Fig. 4) with a pipetted volume of 1.6 µL mm−2. Confocal imaging (Fig. 2d) confirmed these findings.

The optical properties of the CE membrane were assessed by light spectrometry. The CE membrane performed better than a polyester (PET) membrane of standard Transwell inserts. The 10-µm-thin CE membrane absorbed about 10% of visible light, whereas a 10-µm-thin PET membrane with 0.4 µm pores used in inserts absorbed about 20% (Fig. 2b). This low absorbance level was also obtained for a 2:1 ratio CE membrane and for a collagen membrane (Supplementary Fig. 5). The excellent optical proper- ties of the CE membrane were qualitatively confirmed by text placed at the backside of the membrane that was easily read from the apical side (Fig. 2a).

The biodegradability of the membrane was evaluated by optical microscopy. Upon exposure to all the tested concentrations of MMP-8, a matrix metalloproteinase secreted by neutrophils, the CE membrane was totally degraded with no remaining CE leftovers (Fig. 2e–g). The degradation time was found to be inversely proportional to the MMP-8 concentration (Fig. 2h). At 50 U mL−1, the membrane was degraded in less than 1 h, while more than 6 h were required to degrade the membrane with MMP-8 at 5 U mL−1.

The

membrane were tested using exposure to rhodamine B, a molecule that has often been used to show the limitation of PDMS27,28. Compared with PDMS and with the PET membranes of similar thicknesses, the CE membrane absorbed much less

****

Fig. 2 Properties of the thin biological membrane. a Optical clarity of a 10-µm-thin CE membrane integrated in the gold mesh. Scale bar: 200 µm. b Comparison of the spectral absorbance of the CE membrane (n = 4) and of a polyester membrane (Transwell insert 0.4 µm pores sizes) (n = 3). c Characterization of the CE-membrane thickness in function of the collagen–elastin solution volume pipetted on top of the gold mesh (n = 4). d Cross- section of the CE membrane visualized via confocal microscopy. Scale bar: 20 µm. e Picture of an array of several hexagons with a CE membrane. Scale bar: 100 µm. f Local disruption (top left hexagon) of a membrane after being exposed to 10 U mL−1 MMP-8 in PBS+ during 1 h and stretched at −2 kPa. Scale bar: 100 µm. g Totally disrupted membrane after being exposed during 1 h to 50 U mL−1 of MMP-8 in PBS+. Scale bar: 100 µm. h CE-membrane degradation in function of the time and of the MMP-8 concentrations at 550 nm (n = 4). i SEM picture of the collagen and elastin fibers of the CE membrane. Scale bar: 500 nm. j Difference of rhodamine B (10 µM) absorption between a 10-µm-thin CE membrane (n = 13), a 10-µm-thin PDMS membrane (n = 6) and a polyester porous membrane (Transwell insert, 0.4 µm pores sizes) (n = 6). k Pictures of CE membrane (left) and a PDMS membrane (right) after being exposed to RhoB for 2 h. Scale bar: 200 µm. l Transport of FITC–sodium and RITC–dextran molecules across the CE membrane (n = 19).

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-021-01695-0

Article

COMMUNICATIONS BIOLOGY | (2021) 4:168 | https://doi.org/10.1038/s42003-021-01695-0 | www.nature.com/commsbio

3

rhodamine B. After 2 h of immersion in 10 µM rhodamine B, the number of fluorescent molecules ab-/adsorbed was about 90% lower in a 10-µm-thin CE membrane than in the PDMS and the PET membranes (p value < 0.0001) (Fig. 2j). The absorptions/ adsorptions of all polymeric membranes tested are higher than those of all biological membranes (Supplementary Fig. 6). This absorption difference is illustrated in Fig. 2k, which shows the PDMS and CE membranes after 2 h incubation time with rhodamine B. Using the same imaging setting parameters, the PDMS membrane absorbs more rhodamine B than the CE membrane.

The CE-membrane permeability was assessed by the apical- basolateral transport of two molecules with different molecular weights: FITC–sodium (0.4 kDa) and RITC–dextran (70 kDa).

After 4 h of incubation, 25.0% ± 3.5% of the smaller molecules and 12.3% ± 2.9% of the larger molecules were detected in the basolateral chamber (Fig. 2l). The permeability of the membrane was further tested by culturing cells at the apical side of the membrane at the air–liquid interface (see “The CE membrane, a good cell culture support”).

The stretchability of the CE membrane was tested by applying a cyclic negative pressure to the basolateral chamber. The membranes of the 40 hexagons deflect simultaneously and homogeneously in three dimensions (Supplementary Movie 1 and Supplementary Fig. 7). Figure 3a shows a numerical simulation of the deflection of the membranes in the array of hexagons. For the 1:1 CE membrane, the applied radial strain reaches 4.9% ± 0.8% for a negative pressure of 1.0 kPa, and almost doubles when −2.0 kPa is applied. The deflection of the CE membrane was compared with the one of a 10-µm-thin PDMS membrane. It appears than the biological substrate deflected more than the synthetic material (Fig. 3c). At −2 kPa, the radial strain reaches 9.1% ± 2.5% for the CE membrane against only 2.7 ± 0.5% for the PDMS membrane (Fig. 3b). The gold mesh slightly deflected during the experiments, but this did not influence the

Hexagon

(Supplementary Movie 1). When lung alveolar epithelial cells were seeded onto the membrane, −4.0 kPa was needed to induce a 10% linear mechanical strain (Supplementary Fig. 8).

The tunability of the membrane was further investigated by changing the ratio of the proteins concentration. When the elastin or the collagen concentrations were decreased, the membrane became stiffer or softer, respectively, which resulted in smaller or larger deflection, respectively, and thus in smaller or larger linear strains (Fig. 3c and Supplementary Fig. 9). For example, at −1.5 kPa, the deflection of the CE membrane was 31 ± 4.4 µm for a 2:1 ratio, whereas it reached 38.6 ± 3.1 µm for a 1:1 ratio and 51.3 ± 3.9 µm for a ratio 1:3 (Fig. 3c). After being exposed to 10 U mL−1 MMP-8, the CE membranes deflected more than an untreated CE membrane. After 45 min of treatment, the membrane deflected at 35.6 ± 3.6 µm (−1 kPa), while the deflection reached 31.0 ± 2.3 µm without treatment (Fig. 3d). By increasing the pressure at −2 kPa, the deflection increased to 63.4 ± 3.0 µm (exposed to MMP-8), against only 42.2 ± 5.8 µm (no treatment). The deflec- tion increased in function of the MMP-8 exposure time (Supplementary Fig. 10). Localised disruption of the membrane could be observed after 1 h MMP-8-exposure (Fig. 2f).

The CE membrane, a good cell culture support. Human pri- mary alveolar epithelial cells (hAEpCs) and human lung micro- vascular endothelial cells (VeraVec) were successfully cultured on

A

Fig. 3 Membrane flexibility. a Numerical simulation of the deflection of the CE-membrane array. b Linear strain inside the CE membrane in function of the applied vacuum (n = 6). A 10-µm-thin PDMS membrane was taken as reference (n = 6). c Deflection of CE membranes of various compositions in function of an applied vacuum (n = 6 for CE membrane (1:1) and CE membrane (2:1) and n = 4 for CE membrane (1:3)). A 10-µm-thin PDMS membrane was taken as reference (n = 6). d Deflection of CE membrane exposed to MMP-8 in function of an applied vacuum (n = 6 for CE membrane (1:1) (CTRL) and n = 3 for treated membranes (MMP-8 (45 min))).

Article

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-021-01695-0

4

COMMUNICATIONS BIOLOGY | (2021) 4:168 | https://doi.org/10.1038/s42003-021-01695-0 | www.nature.com/commsbio each side of the membrane. Lung alveolar epithelial cells were also cultured at the air–liquid interface for several days. The cells were confluent and created a functional barrier (Fig. 4 and Supple- mentary Fig. 11). In these culture conditions, the nutrients diffuse from the basolateral to the apical side of the membrane. On the contrary to PDMS, the CE membrane does not require any preliminary coating for cells to adhere on the membrane. The cells tightly adhered and interacted to the collagen and the elastin as illustrated in the TEM picture of the membrane cross-section and by the observation of phosphorylated focal adhesion kinase (Supplementary Figs. 12 and 13). Lung epithelial cells seeded at high and low density spread and proliferated on the membrane (Fig. 5a and Supplementary Fig. 14). A difference in cellular surface area was observed between day 2 and day 8 between high and low seeding concentration (Supplementary Fig. 14). At high seeding density, cell confluence was reached at day 2. The cellular surface area remained at 1407 ± 160 µm2, whereas it increased to 2480 ± 136 µm2 at low seeding concentration. After 2 weeks, primary human lung alveolar epithelial cells were confluent showing nice cell–cell contacts (Fig. 5b, c). Primary human lung alveolar epithelial cells and primary human endothelial cells could both be cultured for at least 3 weeks (Supplementary Fig. 15).

Reproduction of the lung alveolar barrier. The typical pheno- types of lung alveolar epithelial cells were investigated using TEM imaging and immunostaining. The characteristic morphologies of type I (ATI) and type II (ATII) lung alveolar epithelial cells—flat and elongated for ATI (Fig. 6b), small and cuboidal for ATII29 (Fig. 5f)—were recognisable by TEM imaging. Tight junctions, a further characteristic of lung alveolar epithelial cells, were clearly identifiable in Fig. 5d. Zonula occludens (ZO-1) were expressed along the cell borders at day 4 (Figs. 4 and 5e) and day 21 (Supplementary Fig. 15). The barrier formation and the expres- sion of tight junction markers has been checked for four different patients (Supplementary Fig. 16). Surfactant protein-C (SP-C) and lamellar bodies, both typical ATII markers, are shown in Fig. 5e, f.

The permeability of the CE membrane with a monolayer of lung alveolar epithelial cells was further assessed by testing the

Earlier

(FITC–sodium, RITC–dextran) (Fig. 5g). The experiment was performed between days 5 and 8 to guarantee the confluence of the epithelial layer. The transport properties of the membrane were significantly affected by the presence or absence of cells (Fig. 5h). For FITC–sodium and RITC–dextran molecules, 25.0% ± 3.5% and 12.3% ± 2.9%, respectively, were transported through the membrane without cells against 9.4% ± 3.3% and 2.1% ± 1.2%, respectively, with hAEpCs. This result was confirmed with cells from four patients.

To further reproduce the lung alveolar barrier, human lung microvascular endothelial cells were cocultured on the basolateral side of the membrane, with lung alveolar epithelial cells on the apical side. Both cell types reached confluence and populated the whole array (Fig. 6a). Figure 6b illustrates a close-up of the alveolar barrier, with the CE membrane sandwiched between the alveolar epithelium and the microvascular endothelium.

Discussion

The lung parenchyma comprises of a large number of tiny alveoli organized in a three-dimensional architecture. Thin alveolar walls made of capillary networks and connective tissue separate the alveoli and stabilise the parenchymal construction11,29. The breathing movements to which this environment is exposed to, induce a deformation of the alveolar airspaces and of the inter- alveolar septa. The resulting mechanical strain is spatially het- erogenous due to the variable thicknesses of the interstitial space and the complex parenchymal architecture11,30. This multifaceted and dynamic environment makes the lung alveolar unique and difficult to mimic in vitro. First-generation lung-on-a-chip devi- ces imitate the rhythmic mechanical strain of the alveolar barrier induced by breathing motions8,9. Although these systems allow investigation of the mechanobiology of the air–blood barrier for the first time, they are limited by the nature of the PDMS membrane they are made of. The main drawback of PDMS is that it is synthetic which limits its function and the ability to mimic physiological capacities. The ECM of the lung alveolar region has structural and mechanical cell substrate functions but beyond that the ECM is pivotal in determining normal cellular function and differentiation in health and dysregulation in disease15,31,32.

Another limitation of PDMS membranes is the absorption and adsorption of small molecules and the effect on the ECM as local reservoir of growth factors and bioactive molecules, which are not maintained by the microenvironment at physiological con- centrations, and therefore distort effects in the system. This is also a major concern for preclinical drug testing applications, as the effective drug concentration that cells are exposed to is difficult to evaluate13. A further drawback is the rather laborious and chal- lenging fabrication process of ultrathin and porous PDMS membranes9,33.

Lung-On-A-Chip

devices imperfectly reproduce the geometric dimensions of the native lung alveoli, as the surface of the culturing membrane creates a unique alveolus of non-physiological dimensions, rather Fig. 4 Immunostaining of primary human lung alveolar epithelial cells. hAEpC cultured on the hexagonal mesh with the CE membrane after 4 days and at air–liquid interface for 2 days with expression of adherent junction markers (E-Cadherin, red), tight junctions with zonula occludens-1 (ZO-1, green) and merged (Hoechst, blue; E-Cadherin, red; ZO-1, green). Scale bar: 100 µm.

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-021-01695-0

Article

COMMUNICATIONS BIOLOGY | (2021) 4:168 | https://doi.org/10.1038/s42003-021-01695-0 | www.nature.com/commsbio

Related Journal Articles & DOI Links

Selected peer-reviewed publications relevant to 12 Lead ECG Acquisition. Click the DOI to access the full paper (may require institutional access).

Why Choose Us?

Bangalore guidance for robotics, Spectre and autonomous systems projects.