The molecular mechanisms of oxygen-sensing in human ductus arteriosus smooth muscle cells: A comprehensive transcriptome profile reveals a central role for mitochondria

Abstract

The ductus arteriosus (DA) connects the fetal pulmonary artery and aorta, diverting placentally oxygenated blood from the developing lungs to the systemic circulation. The DA constricts in response to increases in oxygen (O₂) with the first breaths, resulting in functional DA closure, with anatomic closure occurring within the first days of life. Failure of DA closure results in persistent patent ductus arteriosus (PDA), a common complication of extreme preterm birth. The DA’s response to O₂, though modulated by the endothelium, is intrinsic to the DA smooth muscle cells (DASMC). DA constriction is mediated by mitochondrial-derived reactive oxygen species, which increase in proportion to arterial partial pressure of oxygen (PaO₂). The resulting redox changes inhibit voltage-gated potassium channels (Kv) leading to cell depolarization, calcium influx and DASMC constriction. To date, there has not been an unbiased assessment of the human DA O₂-sensors using transcriptomics, nor are there known molecular mechanisms which characterize DA closure. DASMCs were isolated from DAs obtained from 10 term infants at the time of congenital heart surgery. Cells were purified by flow cytometry, negatively sorting using CD90 and CD31 to eliminate fibroblasts or endothelial cells, respectively. The purity of the DASMC population was confirmed by positive staining for α-smooth muscle actin, smoothelin B and caldesmon. Cells were grown for 96 h in hypoxia (2.5% O₂) or normoxia (19% O₂) and confocal imaging with Cal-520 was used to determine oxygen responsiveness. An oxygen-induced increase in intracellular calcium of 18.1% ± 4.4% and SMC constriction (−27% ± 1.5% shortening) occurred in all cell lines within five minutes. RNA sequencing of the cells grown in hypoxia and normoxia revealed significant regulation of 1344 genes (corrected p < 0.05). We examined these genes using Gene Ontology (GO). This unbiased assessment of altered gene expression indicated significant enrichment of the following GOterms: mitochondria, cellular respiration and transcription. The top regulated biologic process was generation of precursor metabolites and energy. The top regulated cellular component was mitochondrial matrix. The top regulated molecular function was transcription coactivator activity. Multiple members of the NADH-ubiquinone oxidoreductase (NDUF) family are upregulated in human DASMC (hDASMC) following normoxia. Several of our differentially regulated transcripts are encoded by genes that have been associated with genetic syndromes that have an increased incidence of PDA (Crebb binding protein and Histone Acetyltransferase P300). This first examination of the effects of O₂ on human DA transcriptomics supports a putative role for mitochondria as oxygen sensors.

1. Introduction

In utero the ductus arteriosus (DA) connects the aorta to the pulmonary artery, ensuring that placentally oxygenated blood is directed away from the developing lungs to the aorta [1]. High pulmonary vascular resistance and relatively low systemic vascular resistance facilitates the fetal shunting of blood from the pulmonary to the systemic circulation through the widely patent DA [2,3]. With a newborn’s first breath, the DA rapidly constricts in response to increased arterial partial pressure of oxygen (PaO₂) as the circulatory system quickly transitions from placental oxygenation to oxygenation by neonatal respiration. Initial DA constriction in response to oxygen elicits functional closure, beginning within minutes of a rise in PaO₂ [4]. Subsequent to functional closure, the DA undergoes remodelling leading to anatomical closure [5] eventually creating a fibrous structure, the ligamentum arteriosum. Failure of DA closure creates a persistent patent DA (PDA). PDA is one of the most common complications of preterm birth, accounting for approximately 5–10% of all congenital heart disease [6]. Although the incidence of PDA is fairly low among full-term infants (57 per 100,000), preterm infants are particularly susceptible, with more than 55% of infants with a birth weight < 1000 g being affected [4,7]. PDA is associated with serious adverse outcomes if not treated, including heart failure, necrotizing enterocolitis, intraventricular hemorrhage, renal dysfunction and chronic lung disease [7]. Therefore, understanding the physiology of DA closure becomes crucial in the development of therapeutic interventions in addition to nonsteroidal inflammatory drugs (NSAIDs), percutaneous DA closure devices, or surgery.

Currently, treatments for PDA include invasive surgical intervention and percutaneous interventions [8] in infants who fail a trial of NSAIDs (neonates <32 week’s gestational age) [9–11]. Intervention to close the patent DA includes risks of complications like recurrent laryngeal nerve damage whilst the use of NSAIDs may cause gastrointestinal, renal and cerebral side-effects [12–15]. Therefore, the development of alternative PDA treatments is necessary; and the understanding of the physiomolecular pathways in the functional and anatomical closure of DA is a useful, early step that should accelerate the discovery of molecules relevant to the pathogenesis and therapy of PDA.

Functional and anatomical closure of the DA is primarily mediated by exposure to oxygen [16]. With the transition from hypoxia in the fetal environment (PaO₂ < 40 mmHg) to normoxia in neonatal life (PaO₂ 80–100 mmHg), multiple pathways act to permit or enhance DA constriction. Withdrawal of endogenous vasodilatory prostanoids, like prostaglandin E, promotes DA constriction [17]. There is also increased endothelial-mediated, oxygen-induced production of the vasoconstrictor endothelin-1, which induces release of intracellular calcium stores resulting in sustained constriction [18]. While the endothelium plays an important role in DA constriction and is the target of common therapies to close the DA (prostaglandin synthesis inhibitors) or to maintain DA patency (intravenous prostaglandin E), the DA constricts in response to increased oxygen tension in the absence of endothelium [19]. Moreover, oxygen-induced DA constriction in human DAs persists in the presence of prostaglandin synthesis inhibitors, endothelin converting enzyme inhibitors and endothelin receptor antagonists [20]. This suggests the DA oxygen-sensing pathway is intrinsic to the DASMC. Our current understanding of the DASMC’s intrinsic oxygen response is that oxygen induces mitochondrial reactive oxygen species (ROS) production by increasing mitochondrial fission [21]. These mitochondrial derived ROS inhibit voltage-gated potassium channels (Kv channels) [20], including Kv1.5 and Kv2.1 [22]. Kv channel inhibition depolarizes the cell, causing voltage dependant L-type calcium channels to open, inducing calcium influx followed by SMC constriction [19]. The L-type calcium channel is also directly oxygen sensitive in the DASMC and developmental impairment of this mechanism contributes to patency of the preterm DA [23]. While the physiological mechanisms of oxygen-induced DA constriction are well described, the molecular pathways underlying these mechanisms remain unclear. Moreover, there has never been an unbiased look at mechanisms of DA oxygen sensing using multi-omics, specifically transcriptomics (i.e. the measurement of RNA produced by expressed genes).

Transcriptomics can provide a powerful and unbiased insight into the differentially regulated pathways of systems in response to a biological challenge [24]. Previous transcriptomic studies have been performed on the DA, however this research has often focused on preterm and term DA gene expression in comparison with the aorta [25–28]. A recent meta-analysis of the data from these studies compared differentially regulated genes between the human DA and aorta [29]. No global gene expression studies in either pre-clinical models or in human cells have examined changes in DA gene expression induced by oxygen. In this study we have purified and then cultured DASMCs from 10 human donors under hypoxic conditions that mimic fetal PaO₂. After carefully ensuring a pure DASMC population we confirmed the DASMCs remained oxygen sensitive by demonstrating an acute rise in intracellular calcium upon exposure to normoxia which was accompanied by DASMC constriction. We then performed RNA sequencing, which revealed the transcriptomic signature of hypoxic and normoxic DA. The data show that the most oxygen-regulated biologic process involved generation of precursor metabolites and energy whilst the most oxygen regulated cellular component was mitochondrial matrix. This first examination of the effects of O₂ on human DA transcriptomics supports an important role for mitochondria as oxygen sensors.

2. Materials and methods

2.1. Human studies

Ethics approval was obtained at each university where this research was performed. Human DA samples were isolated during the course of congenital heart surgery at either University of Chicago (IRB number A3523–01) or University of Nebraska (IRB number 100–11-EP). Ethics approval from Queen’s University Health Sciences and Affiliated Teaching Hospitals Research Ethics Board (HSREB) was obtained for the continued use of the human cell lines in ongoing research (TRAQ #6007784).

2.2. Isolation and culture of human DASMCs (hDASMC)

Primary cell lines were previously established from the surgically harvested DAs following excision as previously described [21]. DAs were harvested and the endothelium and adventitia were removed. The DA media was then minced and cells were grown in culture. Ten hDASMC cell lines were used: five from the University of Chicago and five from the University of Nebraska. Demographics on sex, age at time of surgery, diagnosis, and treatment were collected where available. All cell culture reagents were obtained from Gibco (Carlsbad, California USA). hDASMCs were grown in M231 smooth muscle cell growth media supplemented with 10% FBS, 5% smooth muscle growth supplement, 1% L-glutamine, 1% penicillin/streptomycin, and ciprofloxacin HCl (10 μg/mL). hDASMCs were used within the first five passages in culture and were diligently maintained in hypoxia (2.5% O₂, 5% CO₂, balance N₂) until protocols called for exposure to normoxia (19.6% O₂, 5% CO₂, balance N₂).

2.3. Purification of hDASMC

To ensure that only DASMCs were present in culture we first performed a negative flow sorting experiment to exclude fibroblasts (CD90⁺) and endothelial cells (CD31⁺). Briefly, cultured cells were trypsinized (0.25% Trypsin-EDTA, Gibco), washed with PBS and resuspended in 100 μL hypoxic PBS with 4% FBS containing PE anti-CD90/Thy1 (Abcam ab33694 clone MRC OX-7, Cambridge, MA, USA; 5 μL per 10⁶ cells) and APC anti-CD31 (BioLegend cat#102409 clone 390, San Diego, CA, USA; 1.25 μL per 10⁶ cells [0.2 mg/mL]) or Brilliant Violet 421 anti-CD31 (BioLegend cat#10243, clone 390; 5 μL per 10⁶ cells). Cells were incubated for 30 min at 37 °C in a hypoxic incubator (2.5% O₂, 5% CO₂, balance N₂). Single staining and unstained controls were used to set up sorting and compensation parameters. Excess antibodies were washed away using warm hypoxic PBS, and cells were promptly sorted to only include CD90⁻CD31⁻ cells using the SH800S cell sorter (Sony Biotechnology, San Jose, CA, USA). Sorted hDASMCs were collected in supplemented M231 smooth muscle cell growth media and immediately plated and cultured in hypoxic conditions. Purified cell lines were expanded and grown in hypoxic conditions prior to any further experiments.

2.4. Flow cytometric assessment of SMC phenotype

Putative purified hDASMCs (CD90⁻CD31⁻ cells) were trypsinized, washed, resuspended in PBS and counted. 10⁶ cells were washed with PBS and incubated with a viability stain (cat#L34975; ThermoFisher Scientific, Mississauga, ON, Canada; 1:1000 in 500 μL) for 30 min at 4 °C, then washed with PBS. Samples were fixed in 4% paraformaldehyde for 15 min at 4 °C, washed using PBS with 0.05% Tween^® 20 (PBS-T; Fisher Scientific cat#BP337–500, Whitby, ON, Canada) and 1% Bovine Serum Albumin (BSA; Sigma-Aldrich, Oakville ON, Canada), permeabilized with Triton^® X-100 (0.5%, Fisher Scientific cat#BP151–100) for 15 min at 4 °C, then washed with the PBS-T/BSA. Nonspecific binding was blocked using PBS with 1% BSA and 2% FBS for 30 min at room temperature. hDASMCs were washed, resuspended in 100 μL of the PBS-T/BSA containing Alexa-Fluor 488 anti-α-smooth muscle actin (clone 1A4, cat#53–9760–82; ThermoFisher Scientific, Mississauga, ON, Canada; 1 μL [0.5 mg/mL]), human smoothelin B Alexa-Fluor^® 405-conjugated antibody (cat#IC8278Vl; R&D systems, Bio-Techne, Oakville, ON, Canada; 3 μL), and PE-conjugated human Caldesmon antibody (clone REA1120, cat#130–119–344; Miltenyi Biotech, Auburn, CA, USA; 2 μL), incubating for 1 h at 4 °C. Excess antibody was washed with PBS-T, and the expression of alpha-smooth muscle actin, smoothelin B, and caldesmon were measured using the SH800S cell sorter/cytometer. Debris was excluded using side and forward scatter and laser settings were defined using unstained samples. Fluorescence minus one (FMO) controls were used to ensure accurate gating for analysis. Cells were analyzed using FlowJo software (FlowJo ^™, Becton, Dickinson & Company, USA).

2.5. Proliferation of cells grown in hypoxia and normoxia

Proliferation of hDASMCs was measured using the xCELLigence^® Real-Time Cell Analysis (RTCA) DP Instrument (Agilent, Santa Clara, CA, USA). Five randomly selected hDASMC cell lines were grown in hypoxia prior to the proliferation experiment. The cells were then trypsinized, resuspended in M231 supplemented media, counted and plated in duplicate at a density of 5000 cells per well in a total volume of 200 μL in an RTCA CIM Plate 16 (ref#5665817001; Agilent, Santa Clara, CA, USA). The instrument was kept in a cell incubator and a plate with M231 supplemented media was used to calibrate the instrument. Recordings were taken every 15 min for 96 h, pausing recording approximately two days after plating to change the media. The experiment was repeated with the same cell lines, freshly plated in either hypoxia (2.5% O₂, 5% CO₂, balance N₂) or normoxia (19.6% O₂, 5% CO₂, balance N₂). Duplicate wells of each cell line in each oxygen condition were averaged and the hDASMC proliferation is demonstrated by the change in cell index every 24 h, where cell index is a unitless measure reporting the impedance of electron flow caused by adherent cells and is calculated as: cell index = (impedance at time n – impedance in the absence of cells)/nominal impedance value.

2.6. Intracellular calcium indicator Cal-520 AM imaging

One day prior to imaging, hDASMCs were plated into 35 mm glass bottom dishes (No. 1.5 uncoated γ-irradiated, P35G-1.5–14-C MatTek Corporation, Ashland, MA, USA) in M231 supplemented media at a density of 5 × 10⁵ cells per dish. Imaging experiments were conducted in an oxygen-sensing assay buffer (modified Krebs solution: 115 mM NaCl, 5.4 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 25 mM NaHCO₃, 10 mM HEPES, 10 mM D-glucose). hDASMCs were loaded with 10 μM Cal-520 AM (Abcam ab171868) mixed with 0.02% Pluronic F-127 (Invitrogen P6866) in O₂-sensing assay buffer and incubated in the dark at 37 °C and 2.5% oxygen (5% CO₂, balance nitrogen) for 30 min. Following dye loading, cells were washed with assay buffer and incubated at 37 °C and 2.5% oxygen for 15 min prior to imaging. The assay was then conducted in 1 mL O₂-sensing assay buffer, modulating pO₂ using an OkoLab stage-top microscope incubator (OkoLab Bold Line, Pozzuoli, Italy). Live cell imaging was performed using a Leica TCS SP8 X confocal microscope (Leica Microsystems, Wetzlar, Germany). Cells were imaged during hypoxia (3% O₂, 5% CO₂, balance N₂) or normoxia (room air). Care was taken to avoid oxygen except when a normoxic challenge was administered. pO₂ and pH were confirmed using an ABL90 FLEX blood-gas analyzer (Radiometer, Mississauga, ON, Canada). Cells were imaged for 15 min in hypoxia, followed by 15 min in normoxia, and a further 15 min returning to hypoxia, capturing one frame every 15 s. The dynamic range of the dye (positive control) was established using 10 μM ionomycin and 20 mM EGTA to establish maximum and minimum intensity values. The oxygen-induced calcium response was measured as a percentage increase from hypoxic baseline. Cell constriction was also measured to determine oxygen-responsiveness. The length of ten randomly selected cells were measured during the hypoxic baseline and during the 15 min of normoxia. The cell constriction is reported as percentage decrease in cell length relative to the hypoxic baseline.

2.7. RNA isolation and sequencing of purified hDASMC cells under hypoxic or normoxic environmental conditions

Ten oxygen-responsive hDASMC cell lines were grown for 96 h in either hypoxia (2.5% O₂, 5% CO₂, balance N₂) or normoxia (19.6% O₂, 5% CO₂, balance N₂). The cells were then collected and stored at −80 °C prior to RNA isolation. Cells were lysed using 600–1000 μL TRI Reagent^® (Sigma-Aldrich, cat# T9424–100ML, Lot # MKCK9023). RNA extraction was processed in bulk using Zymo Direct-zol RNA MiniPrep (cat# R2050, Zymo Research, California, USA), according to manufacturer’s protocols, and then quantified using Qubit spectrophotometer (Thermo Fisher Scientific, Waitham, Massachusetts, USA). Libraries were prepared using a QuantSeq 3′ FWD mRNA-Seq Library Prep Kit for Illumina (Lexogen, Austria) with an input concentration of 450 ng. We used 16 cycles of final PCR amplification. Individual libraries were pooled at a concentration of 35fM each and the pooled samples cleaned using Lexogen bead purification ready for sequencing using the Illumina NextSeq 550 using the Mid V2.5 chemistry.

2.8. RNAsequencing bioinformatics

All bioinformatics were performed either on the Centre for Advanced Computing (CAC) Frontenac server, or on local machines running R (R Foundation for Statistical Computing, Vienna, Austria). Briefly, Fastq files were assembled from bcl2 output without lane split, and subject to quality control per-sample with fastqc and across samples using multiqc. Low quality reads and adapters were trimmed from reads using BBDUK (https://jgi.doe.gov/) [30], and aligned to the human genome (HG38) using STAR [31]. Reads were indexed prior to counting with HTSeqcount [32]. Within the R environment, count files from each sample were assembled into condition-wise groups and analyzed using DESeq2 [33]. Functional analysis was performed using Cluster Profiler [34] on a list of differentially regulated genes that satisfied corrected p-value (padj) <0.05. Functional groups were resolved in terms of GO domains: Cellular Component (CC), Biological Process (BP) and Molecular Function (MF), with enriched terms satisfying a Benjamini Hochberg corrected p-value cutoff <0.05.

2.9. Quantitative PCR (qPCR) validation

Six randomly selected hDASMC cell lines were woken from storage in liquid nitrogen and grown in hypoxia until confluent. Cells were then trypsinized, resuspended in M231 media, and counted. Each cell line was plated at a density of 150,000 cells per well in duplicate per oxygen condition in a 6-well cell culture dish. Cells were grown for 96 h in either hypoxia (2.5% O₂, 5% CO₂, balance N₂) or normoxia (19.6% O₂, 5% CO₂, balance N₂), following which they were collected and lysed in 700 μL TRI Reagent^® (Sigma-Aldrich, cat# T9424–100ML, Lot # MKCK9023). RNA was isolated using Zymo Direct-zol RNA MiniPrep (cat# R2050, Zymo Research, California, USA), according to manufacturer’s protocols, and quantified using the DropSense 16 (Trinean, Pleasanton, CA, USA) according to manufacturer’s protocols. Complementary DNA (cDNA) was synthesized using qScript^™ cDNA Supermix (Quantabio, Beverly, MA, USA) according to manufacturer’s protocol, with 240 ng RNA input per sample. cDNA was then quantified using the DropSense 16, and all samples were adjusted to the lowest concentration.

qPCR was conducted using TaqMan^™ probes (ThermoFisher Scientific, Mississauga, ON, CA) and PerfeCTa Fastmix II (Quantabio, Beverly, MA, USA) according to manufacturer’s protocol, using the Quant-Studio^™ 3–96-well 0.2 mL Block (cat#A28567, ThermoFisher Scientific, Mississauga, ON, CA). We validated a selection of genes that were differentially regulated on our transcriptomic analysis. These genes were chosen either because they were the most differentially regulated or were differentially regulated genes that were considered to be of functional importance to the proposed mechanism of DA constriction. All probes for target genes used FAM as the reporter and NFQ-MGB as the quencher. All samples were also run with Eukaryotic 18S rRNA Endogenous Control (cat#4319413E, ThermoFisher Scientific, Mississauga, ON, CA) with the reporter VIC and quencher NFQ-MGB, a housekeeping gene used as an internal control in all experiments. The genes tested were the following: NADH-ubiquinone oxidoreductase (NDUF), NDUFS2 (Hs00190020_m1), NDUFS5 (Hs02578754_g1), NDUFS7 (Hs01086219_m1), NDUFS8 (Hs00159597_m1), NDUFA9 (Hs00245308_m1), NDUFA13 (Hs00363071_m1), HIF1A (Hs00153153_m1), PSAT1 (Hs01107691_g1), EGLN3 (Hs00222966_m1), and APLN (Hs00175572_m1). Samples were run in triplicate, averaging the C_T values, with a no template control well. The delta C_T (ΔC_T), difference between the target gene and the housekeeping gene (18S rRNA), was calculated for all targets and differences between ΔC_T in hypoxia and normoxia were compared using two-tailed unpaired t-tests. Gene expression data from qPCR validation is all reported as log2 fold change [35].

2.10. Statistics and data analysis

Data are presented as mean ± standard error mean (SEM) unless otherwise stated. Descriptive statistics were performed to check for equal variance between groups and the Shapiro-Wilk test was used to test for normal distribution. All data analysis was conducted using Prism 9 (GraphPad Software, LLC, San Diego, CA, USA). The Kruskal-Wallis test with Dunn’s correction for multiple comparisons was used to compare the percent increase in calcium signal across all cell lines. Paired t-tests with false-discovery rate correction method of two-stage step-up Benjamini, Krieger and Yekutieli was used to test the calcium response of each cell line, comparing the Cal-520 signal in normoxia vs hypoxia. The same tests were used to compare cell length in hypoxia vs normoxia to examine cell constriction. Multiple unpaired t-tests with Welch correction and false-discovery rate correction method of two-stage step-up Benjamini, Krieger, and Yekutieli was used to compare proliferation data of hDASMCs grown in hypoxia vs normoxia. To examine the proliferation of cells grown in hypoxia and normoxia, RM one-way ANOVA with Geisser-Greenhouse correction and post-hoc testing of Tukey’s multiple comparisons test was used, with the comparison of interest being the cell index on day 1 and day 4, examining the different oxygen conditions separately.

3. Results

Ten hDASMC lines were established from neonates undergoing heart surgery (6 males, 2 females, 2 unknown). Sex information was not available for two of the patients and was assigned based on gene expression of the gene encoding Xist, which regulates X inactivation in females (Supplementary Fig. 1). Known demographics are reported in Fig. 1A which include sex, age at time of surgery and diagnosis; with some missing data points.

Fig. 1.

Human Ductus Arteriosus cell lines were enriched for smooth muscle cells.

A. Two female patients, six male patients and two patients with undisclosed sex* were used in this study. B. Unstained sample used for gating CD31 (endothelial cells) and CD90 (fibroblasts) double negative population. C. hDASMCs were negatively sorted using antibodies for CD31 and CD90. Representative plot shown. D. All hDASMC cell lines used for transcriptomics experiments were highly enriched SMCs when sorted using SMA antibody. E. Five randomly selected hDASMC cell lines were confirmed to be proliferative, with a significant increase in cell number between day 1 and day 4, both when grown for 96 h in hypoxia (p = 0.040) and when grown for 96 h in normoxia (p = 0.037). There was no significant difference in cell index between hypoxia and normoxia at any time point.

VSD = ventricular septal defect, CHF/PH = Congestive heart failure/pulmonary hypertension, HLHS = hypoplastic left heart syndrome, D-TGA = dextro-transposition of the great arteries, PFO = patent foramen ovale, TGA = transposition of the great arteries. *sex of these patients were determined to be male, a consequence of Xist gene expression from the RNAseq data (Supplementary File 1).

3.1. Isolated and purified hDASMCs respond to normoxia indicating a conserved oxygen-sensor

Purified hDASMC lines (Fig. 1B and C) were expanded. The SMC phenotype was confirmed by the expression of α-smooth muscle actin using flow cytometry (Fig. 1D). The cells were demonstrably proliferative both in hypoxia and normoxia, with a significant increase in cell index between day 1 and 4, and no significant difference in cell index between hypoxia and normoxia at any time point, though there was a trend suggesting the cells grow more slowly in normoxia (Fig. 1E).

Oxygen response was measured in cells by recording the change in cytosolic calcium using the dynamic calcium sensitive dye Cal-520 AM in hDASMCs exposed to oxygen (pO₂ ~ 141 mmHg, pCO₂ ~ 26 mmHg, pH ~ 7,5) compared to hypoxia (pO₂ ~ 44 mmHg, pCO₂ ~ 28, pH ~7.48) (Fig. 2A). While there was variability in the calcium response among cell lines (Fig. 2B), the percentage increase above hypoxic baseline in response to oxygen was 18.1% ± 4.4% (Fig. 2C). All hDASMC cell lines had a significant (p < 0.001) rise in intracellular calcium (Fig. 2B) confirming the persistence of oxygen sensing despite cell culture. All hDASMC cell lines also had a significant (27% ± 1.5%, p < 0.05) oxygen-induced decrease in cell length relative to the hypoxic baseline (Fig. 2D and E). This further confirms that the hDASMCs retained their characteristic oxygen responses following purification and cell culture.

Fig. 2.

Ductus arteriosus smooth muscle cells (DASMC) are functionally responsive to oxygen.

A. Change in cytosolic calcium in response to oxygen was measured using confocal imaging with the dynamic Cal-520 AM dye, reflected in an increased fluorescent signal following oxygen exposure within the indicated region of interest (ROI). B. Normoxia increased Cal-520 fluorescent signal within each cell line, relative to hypoxia. Each patient DASMC cell line was responsive to oxygen, as determined by a significant increase in cytosolic calcium. *indicates p < 0.001C. Normoxia caused an 18% increase in cytosolic calcium signal versus the hypoxic baseline level. There was no significant difference in the percent change in cytosolic calcium between cell lines. D. Each DASMC cell line showed a significant decrease in cell length upon exposure to normoxia. *indicates p < 0.001, #indicates p < 0.005, $indicates p < 0.05. E. Normoxia caused a 27% ± 1.5% decrease in cell length relative to the hypoxic baseline.

3.2. Transcriptomics

RNAsequencing revealed 1344 unique transcripts that were robustly differentially regulated in hDASMCs in response to culture in oxygen for 96 h; 702 of these transcripts were up-regulated, and 642 downregulated (Fig. 3; Supplementary File 1). The most up-regulated gene was XIST, which was 6.09(log2) fold change higher in normoxic cells compared to hypoxic cells; the XIST data allowed for us to identify the sex of the two unknown donors as male: XIST acts to prevent gene dosing in females (XX) and was 1000-fold more highly expressed in the female donors than male donors (Supplementary Fig. 1), enabling clear identification of the sex of the unknown donors. A list of differentially regulated genes was evaluated by GO enrichment analysis to determine overarching functions of differentially regulated genes within one of three GO domains (Supplementary File 2). Within the biological process (BP; 226 enriched GOterms) domain, the most enriched GOterm was ‘generation of precursor metabolites and energy’ (GO:0006091; p-adjust = 1.71E⁻¹¹ 78-genes). Within the cellular component (CC; 43 enriched GOterms) domain, the most enriched GOterm was ‘Mitochondrial Matrix’ (GO:0005759; p-adjust 1.32E⁻¹⁵; 77-genes). Within the molecular function (MF; 29 enriched GOterms) domain, the most enriched GOterm was ‘transcription coactivator activity’ (GO:0003713; p-adjust 1.58E⁻⁴, 48-genes).

Fig. 3.

RNAsequencing reveals normoxia-induced differential gene expression in human DASMC.

Ductus arteriosus smooth muscle cell (DASMC) lines were studied after culture in hypoxia or normoxia for 96 h using RNAsequencing. 1344 genes were differentially regulated by 96-h of oxygen exposure (corrected p-value < 0.05).

The top 10 GOterms in each GOdomain are presented in Fig. 4. In order to determine the networks of genes involved with each of the most enriched GOterms, the data was organized so that the association between genes and functions within each of the GOterm domains could be visualized (Fig. 5A, B, C). The gene clouds for each of these domains demonstrates the expected association of genes with overlapping functions. For example, many of the genes involved in mitochondrial matrix are also involved in mitochondrial protein complex and mitochondrial respiratory chain within the cellular component domain (Fig. 4A). Likewise, within the biological process domain many of the genes are associated with cellular respiration and generation of precursor metabolites and energy. In the molecular function domain many genes are common to various NADH dehydrogenase activities (Fig. 4B). Many of the genes encoding subunits of Complex 1 of the electron transport chain were involved in the biological process and cellular component functional annotations.

Fig. 4.

Functional annotation of RNAseq data reveals enriched pathways in human DASMC after oxygen exposure.

The differentially regulated genes were placed into biological context through analysis of enriched Gene Ontology terms (GOterms). GOterms are resolved into Biological Process (BP), Cellular Component (CC) and Molecular Function (MF) in an unbiased manner. Data are coloured according to corrected p-value (p < 0.05), and the symbol sized according to GeneRatio (the number of genes within our data as a ratio of the total number of genes within that GOterm). The GOterm is then ranked by the number of genes involved in each GOterm, such that the most regulated pathways have the highest count on the X-axis. Note that most regulated pathways relate to mitochondria and metabolism. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5.

Network of genes in each GO pathway within each Domain.

Nodes for genes are coloured according to fold-change. Edges are coloured according to GOterm category. A. Differentially regulated genes involved in overlapping most enriched Gene Ontology terms (GOterms) within the Cellular Component Domain. There is a high clustering of genes involved in GOterms involved with mitochondrial function and the mitochondria’s electron transport chain. B. Differentially regulated genes involved in overlapping most enriched Gene Ontology terms (GOterms) within the Molecular Function Domain. There is a high clustering of genes involved in GOterms involved with ribonuclease activity and transcription. C. Differentially regulated genes involved in overlapping most enriched Gene Ontology terms (GOterms) within the Biological Process Domain. There is a high clustering of genes involved in GOterms involved with cellular respiration, energy production and nucleoside metabolism.

3.3. qPCR validation

Many of the genes encoding subunits of Complex I of the electron transport chain were involved in the biological process (Fig. 5C) and cellular component (Fig. 5A) functional annotations; in total, 21 ‘NDUF (NDAH:ubiquinone oxidoreductase)’ genes are significantly regulated in hDASMC between hypoxia and normoxia, all of which are upregulated with the exception of NDUFS2. The qPCR results confirmed the upregulation of a number of NDUF genes found in the transcriptomic results in independent samples (Table 1A). NDUFS7 was examined due to its role in the rotenone binding pocket [36], and while it was not significantly regulated in the sequencing or qPCR data, in both cases it was trending towards upregulation in normoxia in both experiments. While NDUFS2 was downregulated in our transcriptomic results, it was found to be upregulated by oxygen exposure in our qPCR validation. Regulation of HIF1A, EGLN3, PSAT1 and APLN were all validated in independent samples, (Table 1B).

Table 1.

Quantitative PCR of genes in the NDUF family and select up- and downregulated genes confirms RNAsequencing findings in new samples.

A.
Gene	Seq. Hx counts	Seq. Nx counts	Seq. p-value	qPCR Hx Log Fold	qPCR Nx Log Fold	qPCR p-value
NDUFS5	1510.88	2316.82	0.0003	1.56	5.95	0.018
NDUFA9	320.88	552.73	0.0015	1.21	5.17	0.004
NDUFA13	652.52	1079.52	0.0009	1.14	6.92	0.003
NDUFS8	367.51	546.29	0.024	1.20	8.53	0.003
NDUFS7	204.34	274.67	0.09	1.06	1.47	0.08
NDUFS2	790.31	554.94	0.01	1.11	1.79	0.03
B.
Gene	Seq. Hx counts	Seq. Nx counts	Seq. p-value	qPCR Hx Log Fold	qPCR Nx Log Fold	qPCR p-value
PSAT1	609.31	1890.79	0.0035	1.20	5.47	0.0002
HIF1a	770.97	2559.17	<0.0001	1.18	8.76	0.0005
EGLN3	72.33	6.04	<0.0001	1.36	0.13	0.003
APLN	176.20	50.20	0.0032	1.99	0.23	0.04

The tables indicate the mean sequencing (Seq.) counts for the 10 hDASMC samples in hypoxia (Hx) and normoxia (Nx). The p-value is adjusted for multiple test correction, and the mean log base 2-fold expression of 6 newly derived hDASMC samples grown for 96 h in hypoxia and normoxia from qPCR. A. Our data reveal multiple members of the NDUF family are upregulated in hDASMC following normoxia, as we confirmed via qPCR. NDUFS2 was found to be uniquely downregulated in normoxia in our sequencing data but was upregulated in the qPCR results. B. Sequencing findings were validated for two of the most upregulated genes (PSAT1 and HIF1A), as well as one of the most downregulated genes (APLN) and the only differentially prolyl hydroxylase gene, EGLN3.

4. Discussion

We have profiled the transcriptional signature of human ductus arteriosus within a pure population of hDASMCs under either hypoxic conditions (mimicking in utero conditions) and normoxic conditions (mimicking birth). Prior to transcriptomic assessment, we first ensured that these cells were pure SMCs by adopting an appropriate sorting strategy and confirming the SMC purity using antibodies against Smoothelin-B, which is a marker only found in contractile SMC [37], and h-Caldesmon, which is specific for fully differentiated SMC [38,39]. We then ensured that these cells retained a DASMC phenotype by demonstrating that they had an appropriate response to O₂ including an 18% increase in intracellular calcium and an associated 27% decrease in cell length. These functional data further confirm the purity of our lines since neither endothelial cells nor fibroblasts behave in this way. We were therefore confident that the cells we used for transcriptomics analysis were pure DASMCs that retained an appropriate phenotype in response to O₂.

RNAsequencing, which captured 3′ end transcripts, was then performed on extracted RNA from these hDASMCs. Our bioinformatic pipeline revealed the differential expression of 1344 genes in response to oxygen. This transcriptome wide data demonstrates that irrespective of patient characteristics (genetic background, sex, age at time of surgery, location of surgery or primary diagnosis which necessitated intervention) hDASMCs grown in hypoxic conditions (mimicking in utero physiology), respond to chronic normoxia (mimicking newborn physiology) with a robust regulation of genes that are involved in mitochondrial mechanisms, cellular respiration and energy homeostasis. These unbiased data fit well with the mitochondrial theory of vascular oxygen sensing, which has been previously reviewed [40].

Functional DA closure entails constriction of DASMCs and begins within 5-min of a rise in pO₂, acting to reduce the right-to-left shunt immediately post-partum [4]. The patency of the DA is actively maintained in utero by a variety of factors including the ongoing production of vasodilators that ensure DA patency. Although prostaglandin E₂ (PGE₂) is primarily responsible for maintaining DA patency during fetal life [41–43], DA PGE₂ sensitivity drops during gestation [44], in preparation for the impending closure of the DA. PGE₂ is required for patency, and NSAIDs results in DA vessel contraction [45]. The loss of PGE₂ vasodilation at birth is coupled with an active, oxygen-induced constriction of the DASMCs to achieve functional closure [46]. Functional closure is a necessary precursor for permanent anatomic closure [47]. PGE synthases prostaglandin E synthase (PTGES) and prostaglandin E synthase 2 (PTGES2) are enzymes responsible for the final step in PGE₂ synthesis, and so are critical for PGE₂ production. In our data, the gene encoding PTGES is downregulated by oxygen in hDASMCs, consistent with the well documented fall in PGE₂ levels at birth. The DA cells that we relied upon for our transcriptomic work are legacy samples taken from donors who have incomplete treatment records. While these donors may have been treated with prostaglandins as a result of their congenital heart defects, we do not anticipate that this significantly impacted their gene expression, primarily because these cells constrict in response to O₂ and any brief exposure to prostaglandins in vivo would have had attenuated the O₂ response and so would not invalidate our findings.

We also found upregulation of PTGES2 in oxygen exposed hDASMCs. PGES-1 is a membrane bound PGE synthase functionally coupled to the inducible cyclooxygenase (COX) isoform COX-2, and plays a role in delayed prostaglandin generation and is induced by proinflammatory stimuli [48]. PGES-2, on the other hand, is synthesized as a membrane bound protein but can also function as a cytosolic enzyme [48]. Additionally, PGES-2 couples with both COX-2 and the constitutive isoform COX-1 involved in immediate prostaglandin generation [49]. The role of PGES-2 in physiological and pathological conditions is not well understood. For example, one study found that PGES-2 deficient mice had no specific phenotype and no alteration in PGE₂ levels in several tissues [50].

Anatomic closure typically occurs within the first week of life and is triggered by functional closure. DA constriction produces hypoxia in the DA media, which initiates DA remodelling [47,51]. The hypoxia leads to cell death in the muscle media and the production of vascular endothelial growth factor (VEGF), which stimulates endothelial proliferation [47]. VEGF has been found in all vascular layers in post-mortem human DA [52]. In sheep DA, VEGF stimulation has been shown to increase vasa vasorum ingrowth and neointimal formation, which is characteristic of the ductus wall’s response to hypoxia [53]. Studies of baboon DA demonstrate a correlation between DA lumen constriction, intense hypoxia, DNA fragmentation and VEGF expression [47]. In our hDASMCs, we noticed a downregulation of VEGFA and VEGFB following 96-h of oxygen treatment, relative to the hypoxic cells. This is in contrast to qPCR studies, where an increase in VEGF expression by qPCR was seen between immature patent DA, and closed DA in preterm baboons [54], which may be a consequence of time point, or the heterogeneous cell population in whole DA, unlike our pure DASMC culture materials. The study by Levin et al. also showed a robust upregulation of hypoxia inducible factor 1 α (HIF1-α) in closed DA compared to fetal DA [54]; HIF1-α is the main transcriptional regulator in response to hypoxic stress, regulating other hypoxia responsive genes. Consistent with this, our dataset revealed an upregulation of HIF1-α in response to oxygen. Although HIF1-α is classically upregulated by hypoxia, many groups have reported that it can be activated by oxygen-independent mechanisms such as nitric oxide, ROS, and growth factors [55–58]. Our data also revealed a downregulation of the prolyl hydroxylase isoform PHD3, also known as EGLN3, in response to oxygen. Prolyl hydroxylases act to target the HIF-α subunits for proteasomal degradation [59], so the downregulation of PHD3 in normoxia could be contributing to the observed upregulation of HIF1A in response to oxygen. However, time series experiments will be required to validate and fully understand HIF1-α expression in the DA in response to oxygen.

Although transcriptomic analyses have been performed on the DA, most studies have been performed in animal models and have characterized transcriptome wide patterns in the DA compared to another tissue, usually the aorta. Most of these investigations have been performed in preclinical models [60,61], and only two studies were performed in human DA [29,62]. The study by Yarboro et al. compared gene expression of human DA and aorta, with subsequent overlap analysis comparing results with similarly designed rodent studies. They found that only 11 genes were differentially expressed between the DA and aorta when both human and rodent data were considered [29]. Their study suggests that the uniqueness of the DA (relative to the aorta) is defined by a small group of genes in humans and rodents, although it did not address the DA’s O₂ response pathways. The uniquely regulated DA genes in that study included: PTGER4, which encodes the prostanoid receptor EP4, TFAP2B, which encodes the transcription factor AP-2b, periostin (POSTN), and phosphodiesterase 1C (PDE1C). Using GO biologic processes pathway analysis they found that the predominant differences in gene expression between DA vs aorta were related to regulation of the extracellular matrix (ECM). The other human study characterized the expression profile of patent DA to stented DA, finding that most differentially regulated genes related to extracellular matrix synthesis which could relate to vascular remodelling and neointimal formation [62]. The only transcriptomic study to date that has examined effects of oxygen was that by Costa et al., which examined the transcriptomic effects of birth and oxygen on the whole DA of rodents, finding exposure to hyperoxia induced changes in metabolic genes, consistent with our findings in humans [63].

Several of our differentially regulated transcripts are encoded by genes that have been associated with genetic syndromes characterized by an increased incidence of PDA. For example, transcripts encoding both CREBBP (Crebb binding protein) and the EP300 (Histone Acetyltransferase P300) are downregulated by oxygen in our hDASMC samples, and their mutations have been linked to Rubinstein-Taybi syndrome Type 1 and Type 2 respectively, which presents with congenital heart syndromes, often including PDA. A polymorphism in the Mthfr gene (Methylenetetrahydrofolate reductase), a gene which is downregulated in our hDASMCs, has also been identified as a predictor of congenital heart defects and PDA [64]. Although we have no data regarding the genotype of the hDASMC donors, it is interesting to note that these genes are regulated in the normal physiological response to oxygen in these cells and participate in the pathophysiology of PDA, highlighting their potential importance in the oxygen response in terms of DA closure.

Other differentially regulated transcripts of interest in our dataset include PSAT1 and APLN. Phosphoserine aminotransferase-1 (PSAT1) is involved in amino acid biosynthesis [65] and expression can be induced by oxidative stress (via ATF4), as demonstrated in mouse fibroblasts [66]. In our dataset, PSAT1 was one of the genes most strongly upregulated by oxygen, which could potentially relate to the transcriptional changes that occur with chronic normoxia. Conversely, apelin (APLN) was one of the genes most strongly downregulated by oxygen exposure. APLN is involved in cell signalling and proliferation through activation of ERKs and the PI3K-Akt pathway [67]. ERKs are serine/threonine kinases that regulate transcription factor activity and gene expression [68], while the PI3K-Akt pathway is involved in cell metabolism, growth, proliferation and survival [69]. The downregulation of APLN seen in our transcriptomic and qPCR data could underlie the trend in decreased proliferation seen in our hDASMCs grown in normoxia, though the difference in proliferation was not significant. Interestingly, APLN has also been found to stimulate myosin light chain phosphorylation in vascular smooth muscle cells [70]. Our study focused on chronic normoxic exposure mimicking the timing of anatomical DA closure, however, further investigation on the function of APLN in DA closure and expression over time is required as APLN-mediated vasoconstriction could contribute to acute responses to normoxia.

We used GOontology enrichment to organize our data into overarching mechanisms, allowing us to demonstrate the enrichment of terms within three distinct domains. The most common GOterms within biological process domain related to nucleoside or ribonucleoside metabolic process. Nucleotide metabolism broadly includes de novo synthesis, catabolism, metabolite recycling, and excretion [71]. Nucleotides and their derivatives play many key roles in the cell: beyond the genetic material making up DNA (nucleotides) and RNA (ribonucleotides). Nucleotides can be the base in other molecules such as the energy molecule ATP, coenzymes like NAD or NADP, or signalling molecules such as cAMP [72,73]. Nucleotide biosynthesis plays a key role in regulating cell growth and cell transformation [74,75]. Mitochondrial nucleoside salvage has been shown to be critical in maintaining mitochondrial DNA copy number and mitochondrial function in an autosomal recessive neurometabolic disorder (encephalomyopathic mtDNA depletion syndrome) [76]. Nucleoside metabolism has not been well studied in the DA or other oxygen-sensing tissues. The precise role of nucleoside and ribonucleoside metabolic processes requires further investigation.

Within the cell component GOdomain, the overarching functions relate to mitochondrial function, respiratory chain metabolism and cellular respiration. We have previously identified mitochondrial fission and an associated increase in oxidative metabolism within DASMCs as being a central mechanism behind the acute response of these cells to a highly oxygenated environment [21]. The increase in pO₂ that initiates functional closure is accompanied by an increase in reactive oxygen species. Our functional annotation of genes involved in the DASMC’s response to oxygen supports our prior findings that the mitochondria are critical to the appropriate and normal DASMC response at birth. These data suggest that the mitochondria may play a key role not only in functional but also permanent anatomic closure. We acknowledge that many of our differentially regulated transcripts are not exclusively annotated within a single GOterm, and that there is significant overlap between terms. Therefore, we have resolved our differentially regulated genes into networks in order to identify overlaps in the functions of individual transcripts and note the participation of multiple transcripts encoding Fe—S centres (NADH-ubiquinone oxidoreductase or “NDUFS”). While we highlight the finding from our data that the mitochondrial function is critical to the DASMC O₂ response, we understand that the response is likely to be dependent upon multiple mechanisms. For example, one of the predominant theories regarding the pathway mediating the action of oxygen on DA muscle purports that the oxygen effector is endothelin-1 (ET-1). ET-1 is a peptide formed in the endothelium, a potent vasoconstrictor proposed to be the effector leading to DASMC contraction following increased pO₂ [18,77]. This theory is supported by studies finding inhibition of the ET-1 pathway, predominantly by targeting the endothelin receptor ET_A, attenuates O₂-induced DA constriction [78,79]. ET_A is a G-protein coupled receptor, whose second messenger IP₃ induces the release of calcium from intracellular stores, namely the sarcoplasmic reticulum, producing vasoconstriction [18]. The sensor in this proposed mechanism is presumed to be CYP3A13 [80], a cytochrome P-450 (CYP) hemoprotein: a family of enzymes containing a heme group as a co-factor that function as monooxygenases [81]. Their role as sensor in the oxygen-induced ET-1 pathway is supported by the observation that CYP inhibitors attenuate oxygen induced DA constriction [82]. This scheme of CYP sensor and ET-1 effector is predicated on the idea that there is an uncharacterized CYP monooxygenase product that serves as a messenger from sensor to effector [80]. We searched our dataset for expression of any CYP isoforms: 41 CYP enzymes were expressed in our samples, none of which were significantly differentially expressed between the two oxygen conditions. Furthermore, the CYP3A13 isoform proposed to be the oxygen sensor by Baragatti et al. [80] was not expressed in our samples.

It is intriguing that in our data, sustained normoxia upregulates expression of many subunits of the mitochondrial electron transport chain Complex I (ETC-1), including genes encoding the iron sulfur clusters, such as NDUFS5 and NDUFS8, and genes encoding accessory proteins, such as NDUFA9 and NDUFA13. The ETC-1 subunit NDUFS2 (NADH:Ubiquinone Oxidoreductase Core Subunit S2) is notable as this 49 kDa protein serves as both the binding site for ubiquinone, during electron transport [83], and rotenone [84], a potent ETC-1 inhibitor. NDUFS2 has been identified by multiple groups as an oxygen sensor in both the carotid body and the pulmonary vasculature [85,86]. Fernandez-Aguera et al. first examined the effects of NDUFS2 knockout in the carotid body of mice. NDUFS2 deficient mice lost responsiveness to hypoxia (but not hypercapnia) and failed to exhibit a hypoxic hyperventilatory response. Further, carotid body glomus cells lacking NDUFS2 lacked oxygen-sensitive Kv channel responses to hypoxia and had no hypoxic membrane depolarization [85]. We have previously identified NDUFS2 as being an essential component of the mitochondrial oxygen sensor in the adult rat pulmonary vasculature in vitro and in vivo. Dunham-Snary et al. showed that human and rodent pulmonary artery smooth muscle cells (PASMC) lacking NDUFS2 did not experience a hypoxia-induced rise in intracellular calcium. This was confirmed in vivo via nebulization of NDUFS2 siRNA. Partial NDUFS2 knockdown inhibited hypoxic pulmonary vasoconstriction but did not elevate normoxic pulmonary artery pressure [86]. Unexpectedly, our dataset exhibited opposing results between the sequencing and qPCR results, with our sequencing revealing NDUFS2 downregulation in normoxia while there was an upregulation of NDUFS2 seen in the qPCR results. The discordant results may be the product of false discovery within the transcriptomic dataset. In support of NDUFS2 being upregulated all other NDUFS were upregulated by oxygen (both on transcriptomic measurement and upon confirmation by qPCR). While this discrepancy requires further investigation, which will be conducted at the proteomic level, the putative role of NDUFS2 widely seen in other oxygen sensitive tissues suggests this to be an important target for future studies. Despite the questions that remain regarding NDUFS2, the rise of many genes in the NDUF family in normoxia further highlights that not only are mitochondria important, but it is likely that ETC complex I is an important target for oxygen’s effects on human DASMCs.

5. Conclusions

In conclusion, we have identified the transcriptional signature of human DASMCs and identified their normal physiological response to chronic oxygen exposure, as would occur in the first 4 days following birth. This unbiased transcriptomic survey places mitochondria at the heart of the DASMC’s response to oxygen. We chose a time point that reflects a maximal impact of oxygen on these cells to establish the molecular signature of the hDASMC’s response to oxygen at the transition to anatomical closure. Future studies will provide insight into the molecular changes that lead to this point and help to better understand the differences between functional and anatomic closure in DASMC’s response to oxygen. Moving forward, we will build upon this primary dataset through molecular manipulation and/or intervention of preclinical models of PDA, aiming to unravel the mechanisms of oxygen sensing in this tissue. It will be important to integrate the observed transcriptomic changes in response to oxygen with a proteomic assessment of these normoxic versus hypoxic DASMCs.