3.1 Incorporation of co-culture of two immortalised lung cell lines (A549 and BEAS-2B) and dynamic flow improved the integrity of the cellular barrier in the lung models
First, we attempted to optimise the culture condition (dynamic vs static) using two immortalised lung cell lines (A549 and BEAS-2B). The dynamic condition was created using the microfluidic 3D system – CNBio’s PhysioMimix™. This system allows the continuous perfusion of media with the flow rate at 0.5 µL/s in the basolateral compartments of the inserts. For static condition, the cell cultures were maintained in normal 24 well plates without any media perfusion.
The complexity was incorporated to recapitulate the cell-cell interaction in human lungs. To increase the complexity, BEAS-2B and A549 cells were selected for the co-culture model. To identify which cells should be seeded on the apical and basal side, two co-culture models were tested: BEAS2B-A549 (BEAS-2B cells were seeded on the apical side, A549 cells were seeded on the basal side) and A549-BEAS2B (A549 cells were seeded on the apical side, BEAS-2B cells were seeded on the basal side).
Four models from simple to complex condition were tested. They were: (i) single culture of BEAS-2B in static condition (static 10000 BEAS-2B); (ii) co-culture of BEAS-2B (apical side) and A549 (basal side) in static condition (static 10000 BEAS2B-A549); (iii) co-culture of BEAS-2B (apical side) and A549 (basal side) in dynamic condition (CNBio 10000 BEAS2B-A549); (iv) co-culture of A549 (apical side) and BEAS-2B (basal side) in dynamic condition (CNBio 10000 A549-BEAS2B) (Fig. 2). The morphology of cells in these models was taken every two days to monitor the growth of cells. After 11 days, all four models formed confluent layer of cells (Fig. 2A). Regarding the integrity of the cellular barrier, CNBio 10000 A549-BEAS2B had the highest electrical resistance value among four models (~ 200 Ω·cm2), followed by CNBio 10000 BEAS2B-A549 (~ 100 Ω·cm2), and static 10000 BEAS-2B and static 10000 BEAS2B-A549 (~ 50 Ω·cm2) (Fig. 2B). Overall, the static culture had a lower electrical resistance value when compared with the dynamic models (CNBio models), which indicated that the dynamic flow improved the cellular barrier integrity.
The importance of selecting which cell line should be seeded in the basal or apical side of the insert was shown in the comparison between dynamic models CNBio BEAS2B-A549 and CNBio A549-BEAS2B. The electrical resistance value of CNBio A549-BEAS2B was higher (around 300 Ω·cm2) than CNBio BEAS2B-A549 (~ 100 Ω·cm2), which was consistent with the lowest permeability of FITC-dextran in CNBio BEAS2B-A549 (Fig. 2C). Therefore, the location where the cells should be seeded is important in the development of lung models. There was no statistical significance between all the models and negative control, which confirms their low permeability and proper barrier function.
To investigate whether CNBio A549-BEAS2B can recapitulate the tight junctions in human lungs, the ZO-1 – an intracellular protein of the tight junction complex – was used as a marker. Data demonstrated that the CNBio A549-BEAS2B model had disrupted and fragmented tight junctions (indicated in red colour in Fig. 2D), which indicated that this model cannot form the appropriate tight junctions.
3.2 Replacing the immortalised lung cell lines (BEAS-2B, A549) with the primary lung cell lines (NHBE, NHLF and COPD-NHBE) improved the cellular barrier and developed important structures such as the tight junctions, ciliated cells, and goblet cells
Since the most promising CNBio A549-BEAS2B model could not form proper tight junctions, the next step was to optimise the cell lines. To optimise the cell lines, human bronchial epithelial cells (NHBE) and human lung fibroblasts (NHLF) were used in this study. NHBE and NHLF were used to mimic the human lung environment and reconstitute the human respiratory mucosa [26]. Fibroblasts can support the epithelial cell function by promoting proliferation and differentiation, modulating mucin secretion, and triggering a correct spatial distribution [26]. Hence, fibroblasts can contribute to an appropriate assembling of the bronchial epithelium and aid to maintain the mucociliary phenotype for a long duration.
For optimisation purposes, four models were developed: static single culture of NHBE (static NHBE), static co-culture of NHBE and NHLF (static NHBE-NHLF), dynamic single culture of NHBE (CNBio NHBE) and dynamic co-culture of NHBE and NHLF (CNBio NHBE-NHLF). The morphology of these models after 25 days is presented in Fig. 3. All models showed confluent layers of cells, however, the phenotype of cells in dynamic models (both single and co-culture models) was more heterogenous (as shown in red and blue circles). This could be due to the flow in perfusion system, which might impact on the growth of cells.
All four models (static NHBE, static NHBE-NHLF, CNBio NHBE, CNBio NHBE-NHLF) exhibited highly organised tight junctions with ZO-1 detected in a discrete and continuous localisation around the periphery of the cells (Fig. 3). This result suggested that using primary cell lines can improve the tight junctions between the cells in comparison to immortalised cell lines. Moreover, the dynamic CNBio models (single and co-culture) retained the structure of cells on the surface, whilst the cells in the static condition (for both single and co-culture) were damaged and disappeared after SEM processing (Fig. 3). This result suggested that the dynamic flow in CNBio can promote the growth and support the function of the cells in the apical side. In addition, SEM images showed that more cilia presented in the dynamic co-culture model (CNBio NHBE-NHLF) when compared to other models (static NHBE and CNBio NHBE) (Fig. 3). This indicates that NHLF can enhance and support the proliferation and differentiation of NHBE in the apical side.
The electrical resistance of all four models (in the range of 500 to 2000 Ω·cm2) (Fig. 3E) was higher than CNBio A549-BEAS2B model (~ 300 Ω·cm2) (Fig. 2B). In addition, low permeability of FITC-dextran was shown in all four models (Fig. 3F), which indicated the formation of a proper cellular barrier.
In addition, the histological cross-section images were consistent with SEM images showing more cilia in the dynamic culture than in the static culture (Fig. 3G). The histological cross-section of models showed cells form multilayers, suggesting that all the models were composed of fully differentiated airway epithelial cells (Fig. 3G). Especially, the expression of markers for cilia (acetylated α-tubulin) and goblet cells (MUC5B) was higher for dynamic cultures compared to static cultures, which suggested that the dynamic condition supported the differentiation and development of the models.
Taken together, these data suggest that the dynamic co-culture of primary lung cell lines (NHBE and NHLF) would be the most appropriate model to mimic the living healthy human lungs.
3.3 Developed healthy lung models displayed similar composition and structure as commercial healthy lung models
To validate the developed healthy lung models (CNBio NHBE-NHLF), the commercial small airway healthy model SmallAirTM-HF (SmallAirTM-HF healthy), which comprised human airway epithelial cells and fibroblasts from healthy donors, was used. Both models expressed discrete tight junction ZO-1. However, CNBio NHBE-NHLF expressed more continuous ZO-1 expression than SmallAirTM-HF healthy (Fig. 4A). Fully ciliated cells were shown in SEM images and multilayers of cells were shown in cross section histological images in both models (Fig. 4A). Moreover, CNBio NHBE-NHLF developed a defined cilia structure (acetylated α-tubulin – green) and goblet cells (MUC5B – red) when compared to SmallAirTM-HF healthy. This result suggested that these models can mimic the important features of the human lungs. Importantly, the electrical resistance of these models was high (around 1000–1400 Ω·cm2), in which the control SmallAirTM-HF healthy had a higher value (1400 Ω·cm2) than developed model CNBio NHBE-NHLF (1000 Ω·cm2) (Fig. 4B). These results confirmed that the structure of our developed model closely resembled the tissue architecture of in vivo airway epithelium.
3.4 Different coating substrates and different types of transwell inserts affected the cellular barrier integrity, the growth and differentiation of developed healthy lung models
Two common coating substrates – collagen and fibronectin – and two transwell membranes with different pore-density – Corning and CellQart – were used to optimise the most appropriate healthy lung model CNBio NHBE-NHLF. To monitor the cellular barrier integrity of these models, we used an impedance measurement system, which allows real-time readouts of the cellular barrier function of the models. Unlike the TEER measurement which occurs at one single frequency, impedance measurement allows the measurement of impedance across a wide range of frequencies. This impedance measurement provides not only the barrier function of cell layers (TEER) but also the process of cell growth and cell differentiation in the model (cell capacitance). The TEER value of the model is calculated based on the differences in height between the curve in the models (with cells) and control (without cells) (as shown in the black arrow in the left panels in Fig. 5). The increase in TEER value of the model shifts the plateau of the curve upwards. The capacitance of the model is indicated by the width of the curve of the models (as shown in the black arrow in the right panels in Fig. 5). The increase in capacitance of the model narrows the plateau of the curve.
The impedance of the model was measured for a total of 24 days of culture, which was divided further into two stages: the increase of TEER value (from Day 3 to Day 15) (left panels in Fig. 5) and the increase of cell capacitance (from Day 13 to Day 24) (right panels in Fig. 5). Regardless of the coating substrate, the TEER value of the models grown on Corning inserts on Day 3 was higher than the control (without cells) (140 Ω·cm2 for collagen coated models and 20 Ω·cm2 for fibronectin coated models) (left panels in Fig. 5A, C). Whereas the TEER value of the models grown on CellQart inserts on Day 3 was similar to the control (without cells) (left panels in Fig. 5B, D). From Day 3 to Day 8, the TEER value of the CellQart models increased. Indeed, the TEER value of the CellQart models coated with collagen increased from 37 Ω·cm2 (Day 6) to 80 Ω·cm2 (Day 8) (left panels in Fig. 5B) and the TEER value of the CellQart models coated with fibronectin elevated from 63 Ω·cm2 (Day 6) to 121 Ω·cm2 (Day 8) (left panels in Fig. 5D). However, while the TEER value of the Corning models coated with collagen increased from 337 Ω·cm2 (Day 6) to 442 Ω·cm2 (Day 8) (left panels in Fig. 5A), the TEER value of the Corning models coated with fibronectin decreased from 232 Ω·cm2 (Day 6) to 187 Ω·cm2 (Day 8) (left panels in Fig. 5C). On Day 10, the TEER value of all models dropped. This could be due to the change in culture condition from the liquid phase to the air-liquid interface (ALI). Nevertheless, in the first total 10 days of culture, the TEER value of the Corning models was higher than the CellQart models, regardless of the coating substrate.
For the Corning models coated with collagen, the TEER value reduced from 339 Ω·cm2 on Day 10 to 198 Ω·cm2 on Day 15 (left panels in Fig. 5A). Meanwhile, the TEER value of the Corning models coated with fibronectin increased on Day 13 (219 Ω·cm2) and dropped on Day 15 (172 Ω·cm2) (left panels in Fig. 5C). There were also differences in the TEER value between the CellQart models coated with collagen and fibronectin. The TEER value of the CellQart models coated with collagen increased from 56 Ω·cm2 (Day 10) to 174 Ω·cm2 (Day 15) (left panels in Fig. 5B). Whereas the TEER value of the CellQart models coated with fibronectin increased from Day 10 (91 Ω·cm2) to Day 13 (200 Ω·cm2) and decreased from Day 13 to Day 15 (153 Ω·cm2) (left panels in Fig. 5D).
Regardless of the insert types, the models coated with collagen showed an increase in cell capacitance from Day 13 to Day 24 (right panels in Fig. 5A, B). Whereas the models coated with fibronectin showed an increase in cell capacitance from Day 13 to Day 20, a decrease from Day 20 to Day 22 and an increase from Day 22 to Day 24 (right panels in Fig. 5C, D). This result suggested that the collagen promoted better growth and differentiation of the models than fibronectin.
To assess topography of the membranes and to measure their mechanical properties, Atomic Force Microscopy (AFM) and nanoindentation was used. The topological image of the CellQart insert showed a uniform and highly porous structure, where porous of sizes ~ 0.5 µm was uniformly distributed (Fig. 5E). In contrast the topography of the Corning insert showed less porous and heterogenous structure. Number, size, and uniform distribution of pores in the CellQart insert guarantees continuous nutrients delivery to cells, as well as more physiological media and paracrine signal exchange between cells that are grown on both sides of the membrane. The structure of the Corning insert, on the other hand, characterises with lower number and heterogenous pores of small size, suggesting that there are limited media and nutrients exchange between two sides of the membrane. The nanoindentation results showed that the average Young’s modules of the CellQart and Corning inserts were 511 ± 482 kPa and 266 ± 315 kPa, respectively and statistical analysis showed no statistically significant differences in the stiffness between the CellQart and Corning inserts. Since the stiffness is one of the key parameters that modulates cell responses and differentiation [27], the biomechanical properties of the membranes was further modulated by functionalising their surfaces with ECM. Cumulatively, functionalised membranes provided biomechanical and biochemical cues to enable desired cell differentiation and growth, thus functionality of the models.
3.5 Using diseased human bronchial/tracheal epithelial cells - COPD (DHBE) was effective in recapitulating the features observed in COPD patients
To develop the disease models, the diseased primary lung cell line (DHBE) was used. Similar to the healthy models, three disease models were established and compared: static single culture of DHBE (static DHBE), static co-culture of DHBE and NHLF (static DHBE-NHLF), and dynamic co-culture of DHBE and NHLF (CNBio DHBE-NHLF). All disease models were developed using traditional Corning inserts. A confluent monolayer of cells after 25 days of culture was shown in all disease models (Fig. 6A). However, a more heterogenous cell structure was developed in the CNBio DHBE-NHLF model (as indicated in the blue circles). This again highlighted the impact of the flow on the growth and morphology of cells. The electrical resistance of static DHBE-NHLF was highest (941 Ω·cm2), followed by CNBio DHBE-NHLF and static DHBE (~ 250 Ω·cm2) (Fig. 6B). Low permeability of FITC-dextran was shown in all disease models (Fig. 6C), which suggested that a proper cellular barrier was formed in all disease models. There was significant difference between positive control and CNBio DHBE-NHLF (Fig. 6C).
After the assessment of cellular barrier and permeability, the expression of tight junction ZO-1 in static DHBE, static DHBE-NHLF and CNBio DHBE-NHLF was evaluated. The expression of tight junctions in the single model (static) was different from co-culture models (static and dynamic) (Fig. 6A). The static single culture DHBE expressed discrete and continuous tight junctions, whereas the co-culture static DHBE-NHLF and CNBio DHBE-NHLF had discontinuous and altered localisation of proteins. The differences between the models were also revealed in SEM images (Fig. 6A). SEM images of the static DHBE and DHBE-NHLF presented a small number of cilia, while the representative SEM image of the apical side of the CNBio DHBE-NHLF model showed more presence of cilia (Fig. 6A). Despite the differences in tight junctions and morphology of cells on the apical sides of the models, the number of cilia in disease models (Fig. 6A) was significantly reduced in comparison to healthy models (Fig. 4). Additionally, the thickness of the disease models was smaller than the healthy models, which confirmed the impact of disease on the differentiation and growth of the models. Notably, the model grown under dynamic conditions exhibited higher expression of goblet cells when compared to the model grown in static conditions (Fig. 6D). Since the high expression of goblet cells is one of the features in COPD patients, this result suggested that the cells cultured under dynamic condition can recapitulate the hallmarks of diseased human lungs.
The MucilAirTM-HF derived from COPD patients (MucilAirTM-HF COPD) was used to compare against the developed disease model (CNBio DHBE-NHLF). Both models expressed a reduced numbers of cilia in comparison to healthy models as shown in SEM images and IF staining of acetylated α-tubulin in cross-section (Fig. 7A). However, there were differences in the expression of tight junctions and the composition presented in cross-section histology between the developed disease model (CNBio DHBE-NHLF) and control MucilAirTM-HF COPD model (Fig. 7A). While CNBio DHBE-NHLF expressed discontinuous tight junctions, control MucilAirTM-HF COPD had discrete, continuous tight junctions. The cross-section histological images of these models were also different. While CNBio DHBE-NHLF expressed overproduction and hypertrophy (increase in size) of goblet cells, MucilAirTM-HF COPD did not exhibit the hypertrophy of goblet cells. The hypertrophy of goblet cells was also observed in another static culture disease model (Fig. 6D). In addition, the electrical resistance value of the MucilAirTM-HF COPD model (around 400 Ω·cm2) was higher than the CNBio DHBE-NHLF (around 250 Ω·cm2) at the end of culture (Fig. 7B).
3.6 The level of pro-inflammatory cytokine IP-10 and IL-6 is significantly higher in the developed disease model than healthy model
For a better understanding of the inflammation process in the lung models, profiling of the cytokine produced by the healthy and disease models was performed. Chemoattractant cytokines (chemokines), pro- and anti-inflammatory cytokines can affect the host response to inflammation by mediating the activation of leukocytes. Therefore, the increase or decrease levels of these cytokines can be used as an early marker for lung inflammation in COPD [28]. Hence, the production of 27 cytokines was measured using Luminex bead-based assay for developed disease (CNBio DHBE-NHLF) and healthy (CNBio NHBE-NHLF) models (Fig. 8). Overall, the expression level of all cytokines in disease CNBio DHBE-NHLF was higher than healthy CNBio NHBE-NHLF. Importantly, the level of pro-inflammatory cytokine IP-10 and IL-6 increased significantly in disease CNBio DHBE-NHLF in comparison to healthy CNBio NHBE-NHLF (Fig. 8). This result suggested that the pro-inflammatory cytokine IP-10 and IL-6 could be early markers for COPD disease models.