An organ-on-chip model of pulmonary arterial hypertension identifies a BMPR2-SOX17-prostacyclin signalling axis

Abstract

Pulmonary arterial hypertension (PAH) is an unmet clinical need. The lack of models of human disease is a key obstacle to drug development. We present a biomimetic model of pulmonary arterial endothelial-smooth muscle cell interactions in PAH, combining natural and induced bone morphogenetic protein receptor 2 (BMPR2) dysfunction with hypoxia to induce smooth muscle activation and proliferation, which is responsive to drug treatment. BMPR2- and oxygenation-specific changes in endothelial and smooth muscle gene expression, consistent with observations made in genomic and biochemical studies of PAH, enable insights into underlying disease pathways and mechanisms of drug response. The model captures key changes in the pulmonary endothelial phenotype that are essential for the induction of SMC remodelling, including a BMPR2-SOX17-prostacyclin signalling axis and offers an easily accessible approach for researchers to study pulmonary vascular remodelling and advance drug development in PAH.

Fig. 1. Pulmonary artery-on-a-chip reproduces the functionality of a blood vessel.

a Schematic diagram of the pulmonary artery-on-a-chip. b Z-section of endothelial-smooth muscle interface, fluorescent confocal imaging. HPAECs (green: VE-cadherin), HPASMCs (red: α-smooth muscle actin). Bar = 10 µm. c An image and a schematic diagram of the flow system set up comprising a peristaltic pump, media reservoirs and chips, perfused with culture medium at 6 dynes/cm². d Fluid–structure interaction (velocity profile) in the endothelial channel, COMSOL modelling. e Phase contract (left) and fluorescent microscopy (right) images of HPAECs grown in microfluidic channels under static and flow conditions, as indicated. F-actin: red, VE-cadherin: green. Bar = 10 µm. f Endothelial cell elongation; n = 4 individual chips per time point. Error bars indicate mean ± SEM of a one-way ANOVA with a Tukey’s post-hoc correction test. ****P < 0.0001. g HPAEC and HPASMC phenotype under static conditions and underflow (48 h) in PA-on-a-chip; fluorescent microscopy, F-actin (red) and VE-cadherin (green); Bar = 10 µm. h, i mRNA expression of endothelial differentiation markers, PECAM-1 and KLF2 in cells treated, as indicated; qPCR. ***P < 0.001; Unpaired t-test. n = 3. j Effects of thrombin (1 U/mL) on endothelial barrier function in HPAECs co-cultured with HPASMCs in pulmonary artery-on-a-chip or in transwell dishes. Passage of FITC-dextran (1 mg/mL; 1 h) from top to bottom channel was used as a measure of endothelial permeability. Cells from three different biological donors grown 3–4 chips/transwells per treatment group, were used; n = 10–12. Error bars indicate mean ± SEM of a one-way ANOVA with a Tukey’s post-hoc correction test. **P < 0.01; ****P < 0.0001, comparisons, as indicated.

Fig. 2. Two-hit model of PAH in pulmonary artery-on-a-chip.

a Schematic diagram of pulmonary vascular remodelling in PAH. b α-SMA (red) and vWF (green) staining in healthy and PAH lungs. Bar = 50 µm. c BMPR2 mRNA expression in HPAECs treated AdCTRLshRNA with Ad-BMPR2-shRNA-GFP. n = 3; **P < 0.01; unpaired t-test. d Effect of two-hit (hypoxia and BMPR2 knockdown, 24 h) on HPAEC permeability, measured as the passage of 40 kDa FITC-dextran (1 mg/mL) from top to bottom channel; n = 9 individual chips. e Effect of hypoxia and endothelial BMPR2 knockdown on HPASMC proliferation (24 h, EdU assay). Cells were untreated or treated with 10 µM Imatinib mesylate for 24 h. n = 4–5 individual chips. In d, e error bars are mean ± SEM; one-way ANOVA with Tukey post-hoc correction test. Volcano plots show differentially expressed genes (DEG) in f HPAECs treated with BMPR2shRNA, g HPAECs in the two-hit PAH model (BMPR2 shRNA and hypoxia), h HPASMCs in the two-hit PAH model. Downregulated genes are in blue and upregulated genes are in red; P < 0.05, 0.25-fold cut-off. n = 4.

Fig. 3. PAH pathways in the two-hit disease model.

Top DEG associated with cardiovascular diseases were grouped into categories and visualised as transcripts per million (TPM) heatmaps. Colour-coded pathways are shown underneath the heatmaps, with key gene symbols enlarged in a HPAECs and b HPASMCs. The genotype status (control and BMPR2 knockdown) and oxygen status (normoxia or hypoxia) are shown at the top of the heatmap in different colours, as indicated. Changes in gene expression are also colour coded, with blue denoting a lower relative gene expression and red denoting a higher relative gene expression. Each column represents 1 experimental repeat (n = 4/treatment group).

Fig. 4. Model validation using cells from PAH patients with BMPR2 mutations.

a Schematic diagram of isolation and culture of patient ECFCs. b Permeability of ECFCs from healthy individuals and PAH patients with BMPR2 mutations cultured in PA-on-a-chip. c Proliferation of healthy and PAH ECFCs in PA-on-a-chip under normoxic or hypoxic conditions (2% O₂, 24 h), as indicated. d Proliferation of HPASMCs co-cultured with control or patient ECFCs under normoxic or hypoxic conditions. n = 4–5 biological donors, each assayed in a separate chip. Bars are means ± SEM; one-way ANOVA with a Tukey post-test; *p ≤ 0.05. Volcano plots show e DEG in PAH ECFCs with BMPR2 mutations vs. healthy controls in normoxia, f DEG in PAH ECFCs vs. healthy controls in hypoxia (ECFC PAH model), g DEG in HPASMCs in hypoxia. In e–g n = 5 biological donors, each in a separate chip.

Fig. 5. Disease pathways in control and patient-derived ECFC datasets.

Top DEG associated with cardiovascular diseases were grouped into categories and visualised as TPM heatmaps. Colour-coded pathways are shown underneath the heatmaps, with key gene symbols enlarged in a PAH ECFCs and b HPASMCs co-cultured with ECFCs. The genotype status (control and BMPR2 knockdown) and oxygen status (normoxia or hypoxia) are shown at the top of the heatmap in different colours, as indicated. Changes in gene expression are also colour coded, with blue denoting a lower relative gene expression and red denoting a higher relative gene expression. Each column represents 1 donor; n = 5 different biological donors/treatment groups.

Fig. 6. Comparative analysis of microfluidic PAH models and published PAH and IPAH gene datasets.

DEG datasets from the adenoviral two-hit models (PAH Model) and patient ECFC models (ECFC PAH) of PAH were compared against previously reported RNAseq studies and lists of genes known to be associated with PAH and IPAH. UpSet plots and heatmaps detail comparative analysis of a, b endothelial gene datasets and c, d smooth muscle cell datasets. In a, c the largest gene overlaps are highlighted in yellow. Heatmaps in b, d visualise the similarity of gene expression changes between the two microfluidic models of PAH, the two-hit PAH model and the ECFC PAH model; downregulated genes are in blue and upregulated genes are in red. RNAseq datasets used for comparative analysis were from PASMCs from IPAH lung transplants (database named here IPAH PASMCs); PAECs isolated from IPAH lung transplants (named here IPAH PAEC). Other comparative gene sets included genes with known PAH-associations (here named “known PAH2”) and the DisGENET public database  https://www.disgenet.org/search, where following gene sets were obtained: DisGENET PAH (https://www.disgenet.org/browser/0/1/0/C2973725/), (here named “known PAH1”); DisGENET IPAH (https://www.disgenet.org/browser/0/1/0/C3203102/) (here named “known IPAH1”).

Fig. 7. Endothelial SOX17 regulates PASMC proliferation in the two-hit PAH model.

a Representative confocal images of lung sections of wildtype and BMPR2 (C118W/+) mice with SOX17 (red) and vWF marking endothelium (green) and nuclei in blue (DAPI), as indicated; Bar = 50 µm. Arrows point to cell nuclei. Magnified images of boxed areas are shown in the top right corner. b, c Corresponding graphs showing SOX17 and vWF expression in mouse lung tissues, as indicated. d Confocal images showing localisation of SOX17 in BMPR2-deficient HPAECs (BMPR2 shRNA) and controls (control shRNA), as indicated. Arrows point to cell nuclei. Bar = 10 µm. e Volcano plot showing SOX17-induced DE proteins, with prostacyclin synthase (PTGIS) marked in pink and selected proteins of interest highlighted. f Effect of endothelial SOX17 overexpression on HPASMC proliferation in the two-hit model (BMPR2shRNA + hypoxia) model, with and without prostacyclin receptor inhibitor, RO1138452 (10 µM). g, h Effect of Ambrisentan and BRD4 inhibitor AZD5153 on HPASMC proliferation in two-hit models of PAH, utilising BMPR2-deficient HPAECs or PAH ECFCs, as indicated. In b, c n = 6/group. ***P < 0.0001, Student t-test; In e n = 4–6, *P < 0.05 comparison with normoxic control, #P < 0.05, comparison as indicated; one-way ANOVA with Tukey post-test. In g n = 4–8, in h n = 5–10.

Fig. 8. Microfluidic two-hit model of PAH.

AdBMPR2 shRNA-treated HPAECs and PAH ECFCs with disabling BMPR2 mutations were co-cultured with HPASMCs under hypoxic (2% O₂) conditions in PA-on-a-chip for 24 h. The diagram shows selected key PAH genes and pathways identified by transcriptomic profiling of endothelial and smooth muscle cells.