Oxygenation as a driving factor in epithelial differentiation at the air–liquid interface

Abstract

Culture at the air–liquid interface is broadly accepted as necessary for differentiation of cultured epithelial cells towards an in vivo-like phenotype. However, air–liquid interface cultures are expensive, laborious and challenging to scale for increased throughput applications. Deconstructing the microenvironmental parameters that drive these differentiation processes could circumvent these limitations, and here we hypothesize that reduced oxygenation due to diffusion limitations in liquid media limits differentiation in submerged cultures; and that this phenotype can be rescued by recreating normoxic conditions at the epithelial monolayer, even under submerged conditions. Guided by computational models, hyperoxygenation of atmospheric conditions was applied to manipulate oxygenation at the monolayer surface. The impact of this rescue condition was confirmed by assessing protein expression of hypoxia-sensitive markers. Differentiation of primary human bronchial epithelial cells isolated from healthy patients was then assessed in air–liquid interface, submerged and hyperoxygenated submerged culture conditions. Markers of differentiation, including epithelial layer thickness, tight junction formation, ciliated surface area and functional capacity for mucociliary clearance, were assessed and found to improve significantly in hyperoxygenated submerged cultures, beyond standard air–liquid interface or submerged culture conditions. These results demonstrate that an air–liquid interface is not necessary to produce highly differentiated epithelial structures, and that increased availability of oxygen and nutrient media can be leveraged as important strategies to improve epithelial differentiation for applications in respiratory toxicology and therapeutic development.

Insight Box

In this paper, we challenge the widely accepted assumption that an air–liquid interface is essential for the differentiation of human primary bronchial epithelial cell cultures. Air–liquid and submerged cells are compared, and oxygen levels are manipulated. We show that bronchial epithelial cells submerged in medium with a high oxygen concentration display improved differentiation profiles compared to those cultured using conventional methods. These findings provide new insight into airway epithelial development and encourage the design of simpler and scalable respiratory culture platforms that avoid challenges associated with culture at an air–liquid interface.

INTRODUCTION

Inhalable drug and respiratory toxicity assays are often performed on human bronchial epithelial cells (HBECs). These primary cells are obtained from bronchial explants and expanded in growth medium whilst in an undifferentiated state [1, 2] and are then differentiated for several weeks at an air–liquid interface (ALI). This standard culture technique results in a polarized pseudostratified epithelium similar to in vivo tissues, consisting of closely packed ciliated cells, mucus-producing goblet cells and other columnar cells. Once differentiated, coordinated beating of apical cilia enable the epithelial cultures to clear mucus [3], an important function of the human airway epithelium [4] and a desirable feature in cultures used for inhalable particle toxicity studies.

Track-etched polycarbonate filters, such as those in commercially available Transwell® culture systems, are commonly used as ALI platforms in epithelial differentiation studies due to their handleability, defined porosity and ubiquity in standardized differentiation protocols [5]. Airway epithelial cells are typically seeded on the upper surface of filter inserts and cells are fed basolaterally with culture medium. Contact with air on one side and nutrients on the other imposes polarity and partially mimics in vivo conditions of mature adult airways. However, filter-based ALI culture platforms are expensive, labor-intensive and difficult to scale up for increased-throughput applications. Moreover, the filters hamper cell visibility during brightfield imaging, making it difficult to study live cells. Despite these disadvantages, the general assumption that ALI is essential for epithelial differentiation has made this culture practice a de facto standard in the field. Recently, Gerovac et al. [6] re-evaluated the validity of this widely accepted practice and reported that a thin layer of culture media covering the apical side of airway cells may further improve their differentiation. Their findings suggest that epithelial differentiation in submerged cultures is possible, and that the associated increase in availability of nutrients and other soluble factors may further enhance robust differentiation. However, the specific microenvironmental cues that trigger cell differentiation under ALI conditions remain unclear, and understanding these signals could lead to the development of simpler culture platforms that are more amenable to robust differentiation and higher-throughput assays.

The cellular microenvironment is a pivotal regulator of cell function [7–9], and oxygenation has been shown to play a critical role in cellular processes. Differences in oxygenation at the cell monolayer surface may arise between ALI and submerged cultures due to the relatively slow diffusion of oxygen in aqueous solutions: in submerged culture, limited diffusion and steady consumption creates an oxygen-depletion zone at the monolayer surface [10]. In contrast, at an air–liquid interface, cells experience ‘normoxic’ oxygen levels, which we define as the concentration of oxygen-saturated fluid at ambient O₂ (0.21 mol/m³). We hypothesize that submersion-induced hypoxia of epithelial cells limits their differentiation, and that recreating normoxic-like conditions on the apical side can enhance differentiation, even in submerged conditions. To test this hypothesis, we computationally model oxygen diffusion in media layers, to appropriately select and create environmentally hyperoxic culture conditions, which creates normoxic conditions at the epithelial monolayer surface while submerged, similar to a standard ALI culture (Fig. 1). We then cultured HBECs from three healthy human donors in filter inserts under standard (air)/ALI, standard (air)/submerged and hyperoxic/submerged conditions. We confirmed barrier integrity via transepithelial electrical resistance (TEER) measurements, and physiological detection of reduced oxygen tension by staining for nuclear hypoxia-inducible factor 1-alpha (HIF-1α). HBEC differentiation was assessed by measuring epithelial layer thickness, ciliated surface area and ciliary functional clearance of particles. Our results demonstrate that an air–liquid interface is not essential to produce highly differentiated human bronchial epithelium and that increased availability of oxygen and soluble factors on the apical side can yield further improvements of epithelial differentiation in vitro.

Figure 1.

Schematic of bronchial epithelial cells cultured in the apical compartment of a porous, polyester Transwell® filter insert in three different experimental culture conditions. Both standard/ALI and hyperoxic/submerged cultured provide the cell monolayer with normoxic levels of oxygen, whereas the standard/submerged conditions produce a hypoxic environment for cells.

METHODS

Unless otherwise stated, all cell culture materials and supplies were purchased from Fisher Scientific (Ottawa, ON), and chemicals from Sigma Aldrich (Oakville, ON).

Oxygen diffusion gradient finite element simulations

The oxygen gradient present during submerged Transwell® culture was simulated using COMSOL Multiphysics software v4.2 (COMSOL, Burlington, MA). Medium on the apical side of a 0.33 cm² (24-well) Transwell® filter was represented by a cylinder of liquid with aqueous properties and a diameter of 6.5 mm and a height of 3 mm. The system was set to a temperature of 37°C. The diffusivity of oxygen in water at 37°C was set at 2.616 × 10⁻⁵ cm²/s based on literature evidence [11]. For the walls of the Transwell®, no oxygen transport was assumed through the polystyrene plastic of the filter. The interface between the cell culture medium and the HBECs was assumed to be a uniform oxygen sink, with an oxygen consumption rate of 6.5 amol/cell-s. This rate was based on the OCR of normal human bronchial epithelial cells, which span from amols to 10s of amols/cell-s [12], a total cell density of 1,250,000 cells/cm² based on our experimental data, and an estimate of the cell fraction with apical surface exposure based on our experimental staining data of junction proteins at the apical surface (~35%). Oxygen transport through the Transwell filter was considered negligible in the finite element models, as polyester Transwell filters have greatly reduced oxygen permeability [13], and a porosity of only ~0.5% [14], presenting a significant barrier to diffusion. The interface between the incubator air and the cell culture medium was assumed to be oxygen saturated and therefore contained 6.727 mg/l of dissolved oxygen. It was also assumed that the cell culture medium was initially equilibrated with ambient air at t = 0 s. The ambient oxygen concentration parameter was swept to determine the conditions under which the cell layer would be exposed to normoxic culture conditions. During standard/ALI culture conditions, the cell surface remained hydrated due to osmotic water flux through the cells and the secretion of mucus onto the cell surface. The cells were therefore in direct contact with a fluid layer that was equilibrated with 21% O₂ and 5% CO₂ incubator conditions. Henry’s law was used to calculate the equilibrium concentration of dissolved oxygen (DO) in this liquid, indicating that cells would be exposed to a DO concentration of 6.727 mg/l (0.210 mol/m³) under standard Transwell® culture conditions, which was the desired cell layer oxygen concentration during these simulations.

Human bronchial epithelial cell culture

Human bronchial epithelial cells (HBECs) from three normal lung donors known to establish ciliated cultures (BD00843, BD00972, BD00954) were obtained at passage 2 from the Primary Airway Cell Biobank at McGill University for these experiments. All tissue procurement and handling procedures followed CIHR guidelines and were approved by the Research Ethics Office (Institutional Review Board) of McGill University (#A08-M70-14B). Transwell® filters (0.33 cm²; 0.4 μm pore sizes) were functionalized with a 1:10 solution of Collagen type IV (10X, C7521, Sigma Aldrich), a major component of the airway basement membrane, in double-distilled water overnight before seeding. Cells were seeded at densities between 70 000 and 120 000 cells/filter. After seeding, they were cultured in a 50:50 mixture of Dulbecco’s Modified Eagle Medium and LHC basal medium supplemented with 0.87 μM insulin, 0.125 μM transferrin, 0.1 μM hydrocortisone, 0.01 μM triiodothyronine, 2.7 μM epinephrine, 0.5 ng/ml epidermal growth factor, 5 × 10⁻⁸ M retinoic acid, 0.5 μM phosphorylethanolamine, 0.5 μM ethanolamine, 0.5 mg/ml bovine serum albumin, 100 U/ml penicillin, 100 μg/ml streptomycin and 50 μg/ml gentamycin. Filters were incubated (21% O₂/5% CO₂; 37°C) in submerged conditions for the first 4 days in culture with 500 μl of medium in the basolateral Transwell® compartment and 100 μl of medium in the apical compartment, or until they reached confluence. Culture medium was replaced daily during this period with the biological safety cabinet lights off, given that the culture media is photodegradable. Once the cells had reached confluence, the apical medium was removed and replaced with 100 μl of media on cells that were designated for submerged culture. Plates were placed under controlled oxygenation conditions: either 21% O₂/5% CO₂ or 30% O₂/5% CO₂, using an incubator subchamber (Biospherix, NY, USA). The medium was changed three times per week.

Transepithelial electrical resistance measurements

The transepithelial electrical resistance (TEER) of cells cultured on Transwell® filters was measured using an EVOM2 epithelial voltohmmeter and STX2 chopstick electrodes (World Precision Instruments, Sarasota, FL) every 2–3 days to assess barrier function. Prior to each use, the EVOM2 meter was calibrated using the 1000 Ω test resistor supplied with the instrument. The chopstick electrode was sterilized using 70% ethanol/water solution and then conditioned in the appropriate cell culture medium. Before measuring TEER, the cell culture medium was changed, and the apical side of the Transwell® filters was filled with 100 μl of medium for 0.33 cm² filters and 500 μl for 1.12 cm² filters. The resistance of a filter and cell culture medium without cells was measured and subtracted from measurements to obtain the transepithelial resistance [15].

Histology

Filters were sectioned to determine the thickness of the epithelial layer and identify ciliated cells. After 25 days in their respective culture conditions, HBEC filter cultures were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 15 min, washed three times with PBS for 5 min and embedded and stained with hematoxylin and eosin at the Rosalind & Morris Goodman Cancer Research Centre Histology Core Facility at McGill. See supplementary material for detailed histological methods.

Immunostaining and fluorescent imaging

Filters were immunostained for hypoxia-inducible factor 1-alpha (HIF-1α) and zonula occludens-1 (ZO-1) by indirect immunostaining, and Hoechst nuclear counterstains, using standard protocols. Briefly, cells were fixed using 4% paraformaldehyde in PBS for 15 min and washed with PBS three times for 5 min. A blocking buffer solution containing 5% goat serum and 0.3% Triton-X100 diluted in PBS was applied for 60 min as per the manufacturer’s protocol. The blocking solution was then replaced with solution containing the primary monoclonal antibodies anti-HIF-1α (clone D1S7W, 1:800; Cell Signaling Technology, Danvers MA) and anti-ZO-1 (clone ZO1-1A12, 1:50, Invitrogen ThermoFisher) in blocking solution. Filters were incubated at room temperature overnight and washed three times with PBS for 5 min each. Goat anti-mouse IgG conjugated with Alexa Fluor 488 (ab150113, 1:1000, Abcam) and goat anti-rabbit IgG conjugated with Texas Red (ab 6719, 1:1000, Abcam) in blocking buffer was pipetted onto the filters and incubated at room temperature for 1.5 h. After washing three times with PBS for 5 min each wash, nuclei were stained using a 1 μg/ml Hoechst 33258 (Invitrogen) in PBS for 1 h. Filters were removed from the polystyrene supports using a scalpel and tweezers and mounted onto glass slides using Fluoromount aqueous mounting medium (F4680, Sigma-Aldrich). Fluorescence imaging was performed on an Olympus IX73 Inverted microscope, and all image processing and analysis were performed in ImageJ (NIH).

Quantification of HIF-1α expression via image analysis

HIF-1α regulates cellular adaptation to low oxygen conditions [16] and its detection in the nucleus indicates that the cell is in a hypoxic state. The nuclear and HIF-1α images were flat-field corrected to compensate for any variation in illumination during image acquisition by generating a flat-field background image in ImageJ for each image using the Gaussian Blur function with a kernel size of 100 pixels. After measuring the mean gray value of these resulting images, a corrected image was created using the ‘Calculator Plus’ function with variables set as follows: i1: experimental image, i2: flat-field image, Operation: Divide, k1: mean gray value, k2 = 0. After all images had been corrected, a mask was created by auto-thresholding the stained nuclei and setting the background to white and the nuclei to black. The area fraction representing the nuclei was measured using the measure function on the software and the image calculator function was used to multiply the thresholded nuclear images with the HIF-1α-stained images to create a 32-bit image. The resulting image consisted of the HIF-1α stain only in the nuclear regions. The mean gray value for each image was then calculated using the measure function and normalized to the nuclear area fraction. These gray values were compared to determine any differences between the three culture conditions.

Ciliation analysis and validation

HBEC differentiation was assessed by quantifying the percentage of the surface area that was ciliated. After 18 days in standard/ALI, standard/submerged and hyperoxic/submerged conditions, live phase contrast videos (2 s, 20 ms exposure time) were taken of the cell surface and exported as image stacks. A z-projection with standard deviation was then applied to identify areas of ciliary movement. The bright areas produced from the z-projection revealed the areas with the greatest ciliary movement. The projection was then thresholded to outline the ciliated area and the area fraction (as a %) was quantified using the measure area fraction function of ImageJ (Supplementary Fig. S2).

Ciliation measurements were validated by histology using H&E-stained sections of BD00843 HBECs cultured on Transwell® filters in standard/ALI, standard/submerged and hyperoxic/submerged conditions for 25 days. The sections were imaged with brightfield microscopy (Olympus IX73 Inverted microscope) and the percentage of surface area that was ciliated was calculated by measuring the length of the filter and the ciliated surface (red lines, Fig. 5A) and calculating the ratio between them.

Figure 5.

HBEC histology-based ciliation analysis after culture for 25 days. (A) Representative histological cross section of BD00843 patient cells cultured in hyperoxic/submerged conditions. The red lines represent the ciliated surface area length measured and used to quantify ciliated area. (B) Percent ciliated surface of BD00843 donor cells cultured in three conditions and measured from cross-sectional histology slices (n = 3, mean ± SD, ^*P < 0.05 by one-way ANOVA with Holm-Sidak post hoc pairwise comparison). Cells cultured in hyperoxic/submerged conditions produced the highest ciliation rates.

Mucociliary transport

Ciliary function was analyzed by scattering fluorescent beads on the epithelial surface and measuring their transport. After 19 days in standard/ALI, standard/submerged and hyperoxic/submerged culture conditions, the media in the apical chamber of the Transwells® was aspirated and replaced with a 1:1000 dilution of fluorescent polystyrene microbeads (FluoSpheres™ Carboxylate-Modified Microspheres, 0.2 μm, red fluorescence (580 nm/605 nm), 2% solids) in PBS. Beads were allowed to settle onto the epithelial surface for 5 minutes before excess beads were removed by aspirating the media. Live epifluorescent video (image stacks, 2 s, 20 ms exposure time) were taken of the cell surface covered in fluorescent beads and exported as stacks. Files were uploaded into ImageJ and the 2D/3D particle tracking function in the MosaicSuite plugin [17] was used to track the coordinates of each bead in every frame (particle detection radius = 16 pixels, cut-off = 0.001, per/abs = 0.500, link range = 2, displacement = 30 pixels, Dynamics: Brownian). This set of coordinates was used to calculate average bead movement speed and cumulative bead displacement in the videos (Supplementary Fig. S2).

Statistical analysis

Statistical analyses of HBEC ciliation area, mucociliary functionality, TEER, epithelial thickness, cell area, nuclear area and HIF-1α expression for all three culture conditions were performed using JASP 0.8.1.0 (The JASP Team, Amsterdam, the Netherlands). Datasets were tested for homogeneity of variance before significance was determined using one-way ANOVA tests. Post hoc pairwise comparisons were performed using Tukey’s method.

RESULTS

Liquid layers reduce oxygen transport

We first verified that the general concept of liquid layers limiting oxygen transport to establish reduced oxygen conditions at the monolayer surface was reasonable, by performing proof of concept experiments with immortalized epithelial CFBE41o-expressing wild-type CFTR (CFTR-WT) cells cultured at 40, 60 and 80% ambient oxygen conditions in submerged or ALI cultures. To assess barrier integrity, TEER was measured throughout the culture. While 40% oxygenation consistently allowed increased TEER barriers, and 80% oxygenation killed cells and created negligible-TEER barriers in both model systems, the intermediate 60% oxygenation level killed cells in ALI culture but allowed them to establish healthy barrier resistances in submerged culture (Supplementary Fig. S1). This preliminary experiment demonstrates that a 3-mm liquid layer, as prescribed for culture in Transwell® filters, is sufficient to establish reduced oxygenation levels at the monolayer surface.

Finite element simulation of oxygen diffusion through media

We then designed a finite element simulation to model oxygen diffusion through the culture medium over a monolayer of submerged epithelia (Fig. 2), to estimate the incubator oxygen concentration needed to achieve normoxic conditions at the cell monolayer surface. The simulation assumed a 3-mm thick layer of culture media was present on the apical side of a 24-well Transwell® filter insert during HBEC submerged filter culture. This cylindrical model (Fig. 2A) included an oxygen sink at the bottom surface of the cylinder to represent cellular oxygen consumption and a constant, oxygen-saturated boundary layer at the upper surface, representing the interface between incubator air and culture medium. We made the unrealistic [18, 19] but necessary first-order approximation that cellular oxygen consumption does not change with oxygenation and assumed that the oxygen concentration in the medium was initially at equilibrium with ambient oxygen in the incubator.

Figure 2.

Finite element simulation of oxygen diffusion in liquid media layer. (A) Surface area plot of oxygen diffusion through apical liquid media layer during (i–iii) standard/submerged and (iv–vi) hyperoxic/submerged cultures at t = 0 s , t = 1000 s and t = 5000 s. Media layer is initially in equilibrium with 21% ambient oxygen conditions with a constant oxygen consumption rate by the cells. Green region represents normoxic conditions (0.21 mol O₂/m³), red region represents hyperoxic conditions and blue represents hypoxic conditions. At equilibrium (t > 5000 s), cells in standard/submerged conditions experience hypoxic conditions; cells in hyperoxic/submerged experience normoxic conditions. (B) Dissolved oxygen at the cellular layer (mol/m³) as a function of time for different incubator oxygen concentrations. Black dotted line represents normoxic conditions of 0.21 mol O₂/m³. (C) Multiparametric analysis of oxygen kinetics in submerged culture to identify optimal parameters for normoxic culture conditions. Various combination of cellular oxygen consumption rate, apical media height and dissolved oxygen at the air–liquid interface allow for normoxic oxygen concentrations at the cell layer. For example, if we assume cellular oxygen consumption rate is 7.4 × 10⁻¹² mol O₂/(s-cm²) and the height of the media layer above the cells is 3 mm, then we would have to set the incubator oxygen concentration to a value that would allow for the upper media surface to be equal to 0.3 mol O₂/m³. Using Henry’s law, we can determine that an incubator oxygen concentration of 30% O₂ would allow for a dissolved oxygen concentration of 0.3 mol O₂/m³ at the upper air-liquid surface, and thus normoxic oxygenation at the epithelial monolayer surface.

Based on this model, the system reached equilibrium after ~1.5 h (Fig. 2A) at which point HBECs were exposed to a final dissolved oxygen concentration of 0.12 mol/m³, approximately half that of the air–liquid interface under normoxic conditions (0.21 mol/m³). The simulation was validated against a standard analytical equation for diffusion in one dimension (Fick’s first law), and similar results were obtained, suggesting that for culture vessels of this size, spatial edge effects were negligible. To compensate for limited oxygen diffusion, the simulation was repeated using elevated atmospheric O₂ concentrations (Fig. 2A(iv–vi) and B). Based on this simulation, we estimated that 30% oxygenation would yield near-normoxic dissolved oxygen levels at the cell surface (Fig. 2A and B), and this was used in subsequent hyperoxic culture experiments.

A parametric sweep varying the apical liquid layer thickness, cellular oxygen consumption rate and incubator oxygen level was then performed to identify different combinations of these three variables that result in a normoxic dissolved oxygen concentration of 0.21 mol/m³ at the cellular level (Fig. 2C). While several combinations yielded a normoxic O₂ concentration at the cellular monolayer, we chose to set liquid layer height at 3 mm of medium above the cells as recommended by the Transwell® manufacturer (~100 μl per well) and used the corresponding 30% ambient oxygenation conditions in all subsequent experiments.

Epithelial monolayer integrity and cell morphology

Normal human bronchial epithelial cells from three donors (identified as BD00843, BD00954 and BD00972) were cultured on 0.33 cm² Transwell® culture filters in standard/ALI, standard/submerged and hyperoxic/submerged conditions based on our finite element simulations, for 25 days. Epithelial barrier integrity was evaluated through TEER and expression of the tight junction protein ZO-1. TEER measurements were similar within all three culture conditions (Fig. 3A) indicating that increased nutrient accessibility due to submersion does not cause HBEC cells to proliferate more rapidly or stack on top of each other, in contrast to our results using the CFTR-WT cell line (data not shown). Moreover, TEER measurements stabilized after 7 days at ~500 Ohms-cm² (Fig. 3A) and the cells stained brightly for ZO-1 in all three culture conditions (Fig. 3B) indicating a healthy, stable epithelial layer with tight junctions and consistent barrier integrity. No statistically significant differences in cell spread area were observed under the three conditions (Fig. 3; morphological data presented for donor BD00954, similar trends observed across other subjects), suggesting that each condition allows formation of an equally dense epithelium. Hence, the parameters of barrier integrity and cell spread area are unlikely to have any differential downstream effects in establishing cell differentiation differences between these conditions.

Figure 3.

HBEC intercellular junctions and cell morphology in standard/ALI, standard/submerged and hyperoxic/submerged culture conditions. (A) TEER measurements of HBECs throughout the entire 25-day culture (n = 3 donors). Epithelial barrier integrity follows similar trends when HBECs were cultured in all three conditions. (B) ZO-1 tight junction stain (green) and nuclear stain (blue) of BD00843 cells after 25 days in culture. (C–E) Morphological and immunostaining analysis for representative BD00954 donor cells after 25 days in culture. N = 3 data points presented as mean ± SD, with ^*P < 0.05 by one-way ANOVA with Holm-Sidak post hoc pairwise comparisons. (C) Average cell area showed no significant difference between cell area in all three conditions indicating that cells formed equally dense epithelial layers. (D) Average nuclear area in hyperoxic/submerged conditions was statistically lower than cells cultured in standard/ALI and standard/submerged conditions. (E) Mean gray value of HIF-1α stain within the nuclei showed higher levels of hypoxia in standard/submerged conditions than in standard/ALI and hyperoxic/submerged, demonstrating that increased incubator oxygen concentration compensates for diffusion limitations in submerged cultures.

Interestingly, HBECs cultured in hyperoxic/submerged conditions had statistically smaller projected nuclear areas compared to standard/ALI and standard/submerged (Fig. 3D), suggesting that hyperoxic/submerged cells are laterally compacted. Histology of these cells (Fig. 4A) demonstrates a correspondingly larger cross-sectional nuclear thickness, which further supports this interpretation. To confirm that this nuclear phenotype correlates with HBEC sensitivity to oxygen, we assessed hypoxia-inducible transcription factor 1α (HIF-1α) expression in the nucleus. HIF-1α is known to be affected by cellular adaptation to hypoxic stress [20], and thus this marker is used as an indicator of media oxygenation at the cellular level. Culture in standard/submerged conditions resulted in significantly increased HIF-1α expression compared to culture in standard/ALI conditions, indicating increased hypoxia in submerged cultures. However, the low HIF-1α phenotype was rescued in hyperoxic/submerged culture, demonstrating that increasing the partial pressure of oxygen above submerged cultures was sufficient to restore normal oxygen levels at the cell surface despite submersion (Fig. 3E). These results confirm that even a thin layer of media causes an oxygen depletion zone that impacts cell function, and that this can be corrected by hyperoxygenation.

Figure 4.

HBEC thickness analysis after 25 days in culture. (A) H&E stained cross-sectional histology slices after culture in standard/ALI and hyperoxic/submerged conditions. (B) Epithelium layer thickness from histology cross-sectional slices for three donor patient cells after 25 days in culture (n = 3, mean ± SD, one-way ANOVA with Tukey’s post hoc pairwise comparison, ^**P < 0.001). Hyperoxic/submerged conditions produced thicker epithelial cell layers universally across all donors, whereas the other two conditions produced results that varied between donors.

Well-differentiated human bronchial epithelia form a pseudostratified layer approximately 25 μm in thickness [21–23] that consists of ciliated, goblet and columnar cells in contact with the lumen, and relatively undifferentiated basal cells in contact with the basement membrane. Ideally, HBECs cultured in vitro should recapitulate these dimensions measured in histological sections of bronchial tissue [21–23]. Epithelial thickness was measured using fixed HBEC sections (Fig. 4A). While tissue thickness from individual donors varied in their response to culture conditions, there was a significant increase in tissue thickness in the group of hyperoxic/submerged conditions over both standard/ALI and standard/submerged cultures (Fig. 4B). Interestingly, and contrary to our expectations, we found no significant difference in epithelial layer thickness between standard/ALI and standard/submerged culture. ALI is widely presumed to promote differentiation [5, 24, 25], but we found no evidence of this in our experiments. We do note that few studies publish submerged control experiments for this culture method. Furthermore, these differences may arise due to improvements in media formulation, or our use of low-passage primary human epithelial cells, which may behave differently than conventional cell lines and exhibit some degree of spontaneous differentiation, even in submerged conditions. Nevertheless, these data do suggest that the combination of rescued oxygenation and availability of nutrients in submerged cultures may enable more rapid HBEC differentiation than in conventional ALI or submerged cultures.

Ciliated area and assessment of function

Ciliated cells cover ~80% of the human airway epithelium and ciliation provides a convenient index of bronchial epithelial cell differentiation in vitro [26, 27]. We evaluated the ciliation of HBECs cultured for 25 days on Transwell® polyester filters in standard/ALI, standard/submerged and hyperoxic/submerged conditions by fixing and sectioning the cultures and analyzing them histologically (red lines, Fig. 5A). The standard/ALI culture method yielded a ciliated area of ~23%, significantly lower than hyperoxic/submerged conditions (~47%; Fig. 5B). To confirm that these ciliated areas were functional, we also developed a semi-automated, video-based protocol to quantify the area of motile cilia in live cultures (Supplementary Fig. S2). Both live and histological methods demonstrated similar results (Figs 5B and 6A). Ciliated area also increased between the second and third weeks of culture for all conditions (Fig. 6B), suggesting that extended culture beyond the time points studied here may bring ciliation levels closer to in vivo expectations. More importantly, cells exhibited more robust ciliation in hyperoxic/submerged conditions compared to standard/ALI conditions after both 18 and 25 days, further supporting our central hypothesis that increased oxygenation and nutrient availability is beneficial for rapid differentiation of epithelial cells.

Figure 6.

Video-based functional ciliation analysis. (A) Percent ciliated surface area of donor HBECs after 18 days (n = 3, mean ± SD, ^*P < 0.05 by one-way ANOVA with Tukey’s post hoc pairwise comparison, P = 0.013 between hyperoxic/submerged and standard/ALI), measured using video microscopy. (B) Percent ciliated surface of representative BD00843 donor cells cultured for 18 and 25 days (n = 3, mean ± SD, ^*P < 0.05 by two-way ANOVA with Holm-Sidak post hoc pairwise comparison), measured using live stream image capture and processing. (C) Average displacement speed of beads settled on ciliated surface of cells in standard/ALI, standard/submerged and hyperoxic/submerged conditions (n = 3, mean ± SD, ^*P < 0.05 by one-way ANOVA with Holm-Sidak post hoc pairwise comparison). Hyperoxic/submerged conditions resulted in the highest bead displacement speed. (D–G) Representative time lapse of fluorescently labeled beads on ciliated surfaces for cultures at standard/ALI conditions at t = 0 s, t = 20 ms, t = 40 ms and t = 60 ms. White dashed outlines represent the initial positions of each fluorescent bead.

Ciliary beating enables the mucociliary clearance of inhaled pollutants and pathogens from the lung by moving foreign substances up the respiratory tract where it can be expelled through the mouth or swallowed [28]. It is therefore important that cilia are motile and capable of transporting small particles. While many techniques have been developed to characterize ciliary activity, such as differential dynamic microscopy [29], spectral domain optical coherence phase microscopy [30] and Doppler-based optical coherence tomography [31], we performed a simple bead-movement assay (Fig. 6D–G) to evaluate mucociliary clearance capacity. Using this technique, we found that the mean ciliary beating speed of cells cultured in the hyperoxic/submerged condition (18.14 ± 3.71 μm/s) was significantly greater by a factor of 2, over cells in the standard/ALI (9.35 ± 2.84 μm/s) conditions (Fig. 6C). Speculatively, the reduction in cilial movement may be due to greater concentrations of viscous mucin in ALI cultures compared to submerged culture, where mucins dissolve into apical medium and are removed during media exchange. The cumulative displacement of each bead (Supplementary Fig. S3) demonstrates that there are more beads with a higher total displacement in the hyperoxic/submerged conditions than in the standard/submerged condition, where accumulation and dilution of secreted mucins would be comparable. It is hence likely that hypoxic conditions in standard/submerged culture alter the cells’ metabolic state and, therefore, affect cellular functions such as cilia beating frequency.

DISCUSSION

In this study, we asked whether reduced oxygenation at the surface of an epithelial monolayer limits the differentiation normally observed in epithelial cells cultured at an air–liquid interface. We experimentally demonstrate that, in our hands, normal primary human bronchial epithelial cells experience hypoxic osmotic tension under media layers as thin as 3 mm. To rescue this phenotype and ultimately take advantage of the benefits associated with submerged cultures such as scalability, ease of handling and increased nutrient availability, we developed a computationally supported approach to increase the partial pressure of oxygen and restore normoxic conditions at the submerged monolayer. Human bronchial epithelial cells from three healthy lung donors were cultured in our hyperoxic/submerged system and displayed improved differentiation characteristics including reduced projected nuclear area, increased epithelial thickness, increased ciliation area and increased functional mucociliary clearance capabilities. These results therefore challenge the common assumption that ALI is essential for HBEC differentiation, and taken together demonstrate that hyperoxic/submerged conditions produce more highly differentiated HBEC epithelium than conventional culture protocols.

Some limitations exist in interpreting the results of this study. First, our finite element model requires assumptions to be made regarding a static oxygen consumption rate [12] and the inherent variability in metabolic activity that must exist between donors. Although unrealistic, such assumptions are both necessary and reasonable, considering the lack of published experiments of cellular oxygen consumption rate, our experiments confirming relatively hypoxic HIF-1α cellular responses when submerged, and that this oxygen tension can be rescued under appropriately hyperoxygenated incubator growth conditions. These issues may be circumvented by directly measuring oxygen consumption rates at the cell surface, but this presents considerable challenges in obtaining a sufficiently sensitive measurement, with sufficient spatial resolution to measure oxygen at the cellular surface, without disrupting the cells or affecting the experiment. Some techniques have now been developed that may address this [32, 33] and such measurements should be considered in adapting this approach towards other cell types and studies.

Second, the TEER values obtained in this study decline for all culture conditions, from ~1300 to 500 Ohms-cm² over 7 days in culture. A wide range of TEER values have been reported for airway epithelial cells [34] and TEER is often low initially and increases to ~400 Ohms-cm² [26, 35]. We speculate that our high seeding density enabled the HBECs to cover the filter surface area rapidly, leading to high initial resistance measurements that was followed by differentiation and remodeling into ciliated and goblet cells, at which point proliferation arrest occurred and resistant declined to a stable value [36]. This may also be due to our use of primary human epithelial cells at low passage number, which may present different TEER measurements than those reported in immortalized cell lines.

Third, while our measurements of epithelial thickness are in good agreement with published experiments demonstrating that the trend for normal human bronchial epithelial cells to form thicker but less functional epithelial membranes when cultured under submerged conditions [37], the magnitude of this effect appears to be dependent on cell donor. While we were not able to observe consistent statistical improvements in differentiation with hyperoxygenation for each donor, our statistical analysis does reveal an aggregate improvement in thickness and ciliation with the hyperoxic/submerged culture model developed here. Most importantly, the lack of differences observed between submerged culture and ALI culture conditions requires further scrutiny. Recent studies do not typically present a submerged control condition for ALI cultures, likely because ALI is so well established as a method to promote differentiation [5, 24, 25]. As we observe spontaneous ciliation in submerged culture, these findings suggest that given modern media formulations, or our use of primary human cells in these experiments, the broad assumption that ALI is necessary for differentiation of all epithelial cultures should be revisited in a broader study.

In the broader context of better understanding how and why air–liquid interfaces improve differentiation, this work also identifies oxygenation as a critically important parameter, amongst the many microenvironmental changes associated with air–liquid interfaces. This idea is not without precedent. In vivo, human lung epithelia undergo rapid differentiation during fetal development [38] between the 16th and 26th week [39]. Covering airway epithelial cells with several millimeters of medium may simulate fetal development conditions in utero when the lungs are filled with liquid and undergoing rapid differentiation [40–42]. In contrast, early in vitro studies such as those by de Jong et al. [5] compared ciliogenesis of HBECs under submerged conditions grown on both tissue culture plastic and collagen membranes and also compared the results with cells cultured at the ALI on collagen membranes and on decellularized dermis. In both ALI systems, cells grew mature cilia within 21 days of culture, whereas mature cilia were not observed in submerged cultures even after 31 days. Kondo et al. [43] showed that dog tracheal epithelial cells cultured on cellulose ester membranes had improved electrical properties and a thicker, ciliated epithelium at the ALI compared to submerged conditions. Other studies have produced variable results [44, 45]. Perhaps ALI culture may no longer be needed with improved media formulations that have been developed in recent years [46–49], and our study supports recent findings that differentiation can be achieved and even improved upon in submerged culture conditions [6]. We speculate that increased availability of nutrients, epithelial-derived factors, and improved oxygenation in submerged cultures can ultimately promote better differentiation than conventional methods.

CONCLUSION

Respiratory and toxicology studies are often performed using well-differentiated human bronchial epithelial cells cultured at the air–liquid interface. However, ALI cultures are expensive, labor-intensive and difficult to use in live cell imaging applications. In this work, we deconstructed the role of the air–liquid interface and specifically examined how oxygen availability might influence differentiation of human bronchial epithelial cells. We demonstrate that an appropriately designed and computationally supported hyperoxygenation system can be used to create ALI-like oxygenation conditions at the surface of an epithelial monolayer, while maintaining cultures in submerged conditions. Our findings of reduced oxygen tension, increased epithelial thickness, increased ciliation and improved functional transport demonstrate that oxygenation is a critical parameter in the epithelial differentiation process, and that submerged cultures may also demonstrate improved differentiation when sufficient oxygen is available. These findings could therefore encourage the design of simpler and scalable lung culture platforms that avoid the handling difficulties and expenses associated with implementing an air–liquid interface.

Funding

This work was supported by the Fonds de Recherche du Quebec (FRQ)—Nature et technologies (grant no. 205292), the Natural Sciences and Engineering Research Council of Canada (Discovery RGPIN-2015-05512), and the Canada Research Chair in Advanced Cellular Microenvironments to C.M; and the Canadian Institutes of Health Research (grant # PJT-156183) and the Primary Airway Cell Biobank at McGill University from Cystic Fibrosis Canada (grant number 2997) to J.W.H.

Conflict of interest statement

None declared.