A 3D epithelial–mesenchymal co-culture model of human bronchial tissue recapitulates multiple features of airway tissue remodeling by TGF-β1 treatment

Human fetal lung fibroblasts (IMR-90) were obtained from the American Type Culture Collection (Manassas, VA, USA) and grown in minimum essential medium (MEM) (Life Technologies, Carlsbad, CA, USA) with 10% fetal bovine serum (FBS; MP Biomedicals, Santa Ana, CA, USA). Normal HBECs (Lonza, Basel, Switzerland) were grown in Airway Epithelial Cell Growth Medium with SupplementPack (PromoCell, Heidelberg, Germany).

The methodological details for the 3D co-culture of IMR-90 cells and HBECs were described in our previous paper [14]. Cellmatrix type I-A (Nitta Gelatin, Osaka, Japan), 10 × MEM and reconstitution buffer (Nitta Gelatin) were mixed with 8:1:1 by volume ratios and applied to cell culture insert (10.5 mm diameter, 1.0-μm pore size, BD Biosciences, Franklin Lakes, NJ, USA) in 100 μL aliquots to prepare base layer. The base layer was gelled by placing in an incubator at 37 °C with a 5% CO₂ atmosphere for more than an hour. IMR-90 cells (approximately 2.5 × 10⁶ cells/mL in FBS), Cellmatrix type I-A, Cellmatrix type I-P (Nitta Gelatin), 10 × MEM, and reconstitution buffer were mixed with 1:4:4:1:1 by volume ratios and poured onto the base layer in 250 μL aliquots to prepare the collagen-embedded fibroblast layer. The fibroblast layer was gelled by placing in an incubator at 37 °C with a 5% CO₂ atmosphere for more than an hour. After 2 days of cultivation with MEM containing 10% FBS, HBECs suspended in Airway Epithelial Cell Growth Medium (approximately 3.0 × 10⁵ cells/mL) were seeded onto the collagen layer to prepare the co-culture model, and cultured under submerged conditions until reaching a semi-confluent state. The fibroblast mono-culture model was prepared without seeding of HBECs. The ALI culture was then initiated to induce mucociliary differentiation. The mono-culture model was also cultivated under ALI conditions. PneumaCult-ALI medium (Stemcell Technologies, Vancouver, BC, Canada) was supplemented with heparin (Stemcell Technologies) and hydrocortisone (Stemcell Technologies) according to the manufacturer’s instructions. GM6001 (30 nM, Sigma-Aldrich, St. Louis, MO, USA) and 1% FBS were also added to prepare the ALI culture medium. The apical and basolateral media were removed, and 600 μL of ALI culture medium were added to the bottom well. Stimulation of TGF-β signaling with TGF-β1 (R&D Systems, Minneapolis, MN, USA) and inhibition of TGF-β signaling with a TGF-β receptor type I blocker (SB525334; Wako Pure Chemical Industries, Osaka, Japan) commenced on the first day of ALI culturing. The ALI culture was maintained for 21 days. Images of each collagen gel were obtained, and gel contraction was analyzed with ImageJ software (National Institutes of Health, Bethesda, MD, USA). Data are expressed as the percentage to the initial gel area.

After fixation in 4% paraformaldehyde at 4 °C on ALI culture day 21, bronchial tissue samples were embedded in paraffin and 5-μm sections were prepared using a microtome. Sections were deparaffinized and subjected to hematoxylin and eosin staining or immunostaining. Immunostaining was performed with Polink-2 Plus (GBI Labs, Bothell, WA, USA) using the following antibodies: Anti-vimentin antibody (1:250; ab92547; Abcam, Cambridge, UK), anti-alpha-smooth muscle actin (α-SMA) antibody (1:1000; ab5694; Abcam), anti-E-cadherin antibody (1:250; ab40772; Abcam), anti-acetylated α-tubulin antibody (1:250; ab24610, Abcam), anti-MUC5AC antibody (1:250; ab3649; Abcam), anti-cytokeratin (CK) 5 antibody (1:100; ab52635; Abcam), anti-fibronectin antibody (1:250; ab2413; Abcam), and anti-tenascin-C antibody (1:250; ab108930; Abcam). Sections were subjected to heat-induced antigen retrieval in a 10-mM sodium citrate buffer (pH 6.0) for α-SMA, E-cadherin, acetylated α-tubulin, MUC5AC, CK5 and fibronectin staining at approximately 95 °C for 30 min. EDTA (1 mM, pH 8.0) was used for vimentin and tenascin-C staining. Image analysis of immunostained sections was conducted with ImageJ software. Three sections were prepared in each experimental condition, and the area of the epithelial layer and mesenchymal collagen layer was measured. Colour Deconvolution plugin [18] was used for diaminobenzidine (DAB) and hematoxylin stain separation, and the DAB-positive areas in the epithelial and mesenchymal layers were measured. The results are expressed as the percentage of DAB-positive area in each layer.

Culture medium collected on ALI culture day 21 was subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis in 7.5% acrylamide gels containing 0.9 mg/mL gelatin. After electrophoresis, gels were washed twice (30 min and 45 min) in wash buffer (0.5% Triton X-100, 2.5 mM Tris-HCl, 150 mM NaCl) and incubated for 19 h in incubation buffer (2.5 mM Tris-HCl, 20 mM NaCl, 10 mM CaCl₂), then stained with 0.1% Coomassie blue. Images were obtained with ImageQuant LAS 4000 (GE Healthcare, Little Chalfont, UK), and signal densities were quantified using ImageQuant TL (GE Healthcare). Data are expressed as the fold change to the band density in the control.

The concentrations of TIMPs secreted into the culture medium were analyzed with a Bio-Plex Pro Human TIMP Panel (Bio-Rad, Hercules, CA, USA) using the Bio-Plex system (Bio-Rad) according to the manufacturer’s instructions. Culture medium collected on ALI culture day 21 was diluted at 1:10,000 for TIMP-1 analysis or 1:100 for TIMP-2 analysis. The concentrations of TIMP-3 and TIMP-4 were lower than detection limits, although the culture medium was analyzed without dilution.

Total RNA was isolated from tissues using RNeasy (Qiagen, Hilden, Germany), and RNA quality was analyzed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The RNA integrity number of the samples was ≥7.6. cDNA was synthesized with a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Waltham, MA, USA). The gene expression profile was analyzed with the Human Extracellular Matrix and Adhesion Molecules RT 2 Profiler PCR Array (PAHS-013, SABiosciences, Frederick, MD, USA) on an ABI 7900 PCR system (Applied Biosystems).

With the exception of the PCR array results, data are presented as the means and standard deviations of triplicate inserts. Multiple-comparisons tests were used in the analysis of the results obtained with co-culture model. Bartlett’s test was used to confirm homogeneity of variances from multiple groups. A parametric one-way analysis of variance followed by Dunnett’s test was performed to detect statistically significant differences. Results were considered significant at p < 0.05. For the analysis of the fibroblast mono-culture model, student’s t-test was used and the results were considered to be significant at p < 0.05. All statistical analyses except for those of PCR array data were performed using Ekuseru-Toukei (SSRI Co., Ltd., Tokyo, Japan). The PCR array data were analyzed using RT 2 Profiler PCR Array Data Analysis version 3.5, provided by SABiosciences. The results are presented as the means and 95% confidence intervals of the triplicate inserts. Student’s t-test was used; the results were considered significant at p < 0.05.

The co-culture human bronchial tissue model with an epithelial cell layer and collagen-embedded fibroblast layer was stimulated with TGF-β1 for 21 days under ALI culture conditions (Fig. 1a). Collagen gel contraction was observed in the untreated control on ALI culture days 7, 14, and 21 (Fig. 1b). Gel contraction was significantly enhanced by stimulation with 4 or 10 ng/mL TGF-β1, compared with contraction in the the untreated control at each time point (p < 0.05). The collagen gel was clearly detached from the wall of the cell culture insert on ALI culture day 21 following stimulation with 4 or 10 ng/mL TGF-β1 (Fig. 1a, arrowheads). When the co-culture model was stimulated with 10 ng/mL TGF-β1 with simultaneous inhibition of TGF-β receptor type I by 5 μM SB525334, gel contraction was not detectable (Fig. 1b).

To further understand the effects of TGF-β1 on our co-culture model, we performed hematoxylin and eosin staining on histological sections of tissues collected on ALI culture day 21. We found effects of TGF-β1 on the cellular morphology and increased numbers of elongated HBECs (Fig. 2a). In accordance with this increase, a concentration-dependent decrease in the thickness of the epithelial layer was observed following stimulation with TGF-β1 (Fig. 2a). This decrease was prevented by the addition of 5 μM SB525334. In the mesenchymal collagen layer, we found an increase in heavily stained fibroblasts (arrowheads in Fig. 2a). Next, we performed immunostaining to characterize the changes observed in the sections stained with hematoxylin and eosin. We stained for vimentin as a mesenchymal marker, and found a significant increase in vimentin-positive fibroblasts in the mesenchymal layer following stimulation with 4 or 10 ng/mL TGF-β1 (p < 0.05) (Fig. 2a and b). Vimentin expression was also confirmed by TGF-β1 stimulation in the basal cells of the epithelial layer (Fig. 2a, arrows). The percentage of the vimentin-positive area in the epithelial layer increased significantly following stimulation with 4 or 10 ng/mL TGF-β1 (p < 0.05) (Fig. 2c). Immunostaining was also performed with an anti-α-SMA antibody to confirm the presence of myofibroblasts. Intracellular expression of α-SMA in fibroblasts increased significantly following TGF-β1 stimulation (0.4, 4, or 10 ng/mL; p < 0.05) (Fig. 2a and d). This increase was suppressed by blocking TGF-β1 signaling with 5 μM SB525334. We also performed immunostaining with an anti-E-cadherin antibody to confirm the epithelial status. E-cadherin is adhesion molecule that plays a major role in the maintenance of epithelial intercellular junctions. In contrast with the increased vimentin expression (Fig. 2a and c), we observed a decrease of E-cadherin in the epithelium following stimulation with TGF-β1 (Fig. 2a and e).

We also analyzed the effect of TGF-β1 on the differentiation of bronchial epithelial cells. We used acetylated α-tubulin as a ciliated cell marker, MUC5AC as a goblet cell marker, and CK5 as a basal cell marker. These differentiation markers were found to be expressed under control conditions (Additional file 1: Figure S1), following stimulation with TGF-β1, and even when blocking TGF-β1 signaling by 5 μM SB525334 (Additional file 1: Figure S1).

In addition to collagen gel contraction, altered ECM homeostasis is a factor in the airway remodeling process. We therefore analyzed the expression levels of proteolytic MMP-2 and MMP-9 in the culture medium of co-culture model collected on ALI culture day 21 by gelatin zymography. Pro-MMP-9 (92 kDa), pro-MMP-2 (72 kDa), and active MMP-2 (62 kDa) were detected (see co-culture data in Additional file 2: Figure S2); expression levels of each MMP were quantified as the fold change to the control (Fig. 3a–c). Expression of pro-MMP-9 in the co-culture model did not change after stimulation with TGF-β1 (Fig. 3a). However, pro-MMP-9 expression was suppressed significantly when TGF-β1 signaling was blocked with 5 μM SB525334 (p < 0.05). Expression of pro-MMP-2 and active MMP-2 in the co-culture model increased in a concentration-dependent manner following stimulation with TGF-β1, and significant increases were detected following stimulation with 10 ng/mL TGF-β1 (p < 0.05) (Fig. 3b and c). This increase in expression was blocked by the addition of 5 μM SB525334.

We also analyzed the concentrations of natural MMP inhibitor TIMPs in the co-culture medium on ALI culture day 21. Levels of TIMP-1 increased significantly following stimulation with 4 or 10 ng/mL TGF-β1 (p < 0.05) (Fig. 3d). Secretion of TIMP-1 following stimulation with 10 ng/mL TGF-β1 was suppressed to control levels by the addition of 5 μM SB525334. Secretion of TIMP-2 showed a similar trend, and increased significantly in response to TGF-β1, compared with levels in the control (p < 0.05) (Fig. 3e).

To confirm the role of epithelial-mesenchymal interaction in ECM homeostasis in the co-culture model, a mono-culture model with lung fibroblast cells was prepared. The fibroblast mono-culture model was stimulated with 10 ng/mL of TGF-β1, MMP secretion was analyzed with gelatin zymography. Unlike findings in the co-culture model, pro-MMP-9 expression was very low in the mono-culture even with TGF-β1 stimulation (see fibroblast mono-culture data in Additional file 2: Figure S2). Similarly to secretion in the co-culture model, pro-MMP-2 secretion in the mono-culture model increased significantly following stimulation with TGF-β1 (p < 0.05) (Fig. 4a). However, levels of active MMP-2 did not increase with TGF-β1 stimulation (Fig. 4b). Secretion of TIMPs in the mono-culture model stimulated with TGF-β1 was also analyzed; a significant increase in TIMP-1 and TIMP-2 levels was observed (p < 0.05) (Fig. 4c and d).

We analyzed the expression of ECM-related genes in the co-culture model on ALI culture day 21. Among the 84 genes analyzed, the expression of 33 genes increased significantly following stimulation with TGF-β1 (p < 0.05) (Fig. 5). These genes were categorized as follows: TGF-β and integrins (Fig. 5a), TIMPs and MMPs (Fig. 5b), adhesion molecules (Fig. 5c), collagens (Fig. 5d), and ECM glycoproteins and proteoglycans (Fig. 5e). The expression of most of these genes was upregulated in a concentration-dependent manner by TGF-β1 stimulation. These increases were suppressed to control levels or lower by the addition of 5 μM SB525334 (p < 0.05) (Fig. 5a–e). The top five genes upregulated following treatment with 10 ng/mL TGF-β1 in the co-culture model were FN1 (29.3-fold change; Fig. 5e), TNC (10.5-fold change; Fig. 5e), VCAN (7.7-fold change; Fig. 5e), MMP9 (7.38-fold change; Fig. 5b), and COL1A1 (6.8-fold change; Fig. 5d).

We also found that 14 of the 33 genes were upregulated in the fibroblast mono-culture model following stimulation with 10 ng/mL TGF-β1 (Fig. 6). However, the level of fold increase of several genes was lower in the mono-culture model compared with the co-culture model. For example, FN1 expression showed a 29.3-fold change in the co-culture model after stimulation with 10 ng/mL TGF-β1 (Fig. 5e), compared with only a 2.2-fold change in the mono-culture model (Fig. 6).

As FN1 and TNC expression was markedly upregulated (Fig. 5e), we analyzed the expression pattern of proteins encoded by these genes in the co-culture. The results revealed fibronectin expression mainly across the mesenchymal layer (Fig. 7). However, we also found fibronectin-positive basal cells in the epithelial layer following TGF-β1 stimulation (Fig. 7, arrowheads). Tenascin-C was also expressed across the mesenchymal layer, and the staining appeared to become denser with higher TGF-β1 concentrations. Strong expression of tenascin-C was observed in the sub-epithelial basement membrane of the co-culture model (Fig. 7, arrows). Expression decreased as TGF-β signaling was suppressed by the addition of 5 μM SB525334.