Endothelial to mesenchymal transition during neonatal hyperoxia-induced pulmonary hypertension

Abstract

Bronchopulmonary dysplasia (BPD), a chronic lung disease in premature infants, results from mechanical ventilation and hyperoxia, amongst other factors. Although most BPD survivors can be weaned from supplemental oxygen, many show evidence of cardiovascular sequelae in adulthood, including pulmonary hypertension and pulmonary vascular remodeling. Endothelial–mesenchymal transition (EndoMT) plays an important role in mediating vascular remodeling in idiopathic pulmonary arterial hypertension. Whether hyperoxic exposure, a known mediator of BPD in rodent models, causes EndoMT resulting in vascular remodeling and pulmonary hypertension remains unclear. We hypothesized that neonatal hyperoxic exposure causes EndoMT, leading to the development of pulmonary hypertension in adulthood. To test this hypothesis, newborn mice were exposed to hyperoxia and then allowed to recover in room air until adulthood. Neonatal hyperoxic exposure gradually caused pulmonary vascular and right ventricle remodeling as well as pulmonary hypertension. Male mice were more susceptible to developing pulmonary hypertension compared to female mice, when exposed to hyperoxia as newborns. Hyperoxic exposure induced EndoMT in mouse lungs as well as in cultured lung microvascular endothelial cells (LMVECs) isolated from neonatal mice and human fetal donors. This was augmented in cultured LMVECs from male donors compared to those from female donors. Using primary mouse LMVECs, hyperoxic exposure increased phosphorylation of both Smad2 and Smad3, but reduced Smad7 protein levels. Treatment with a selective TGF-β inhibitor SB431542 blocked hyperoxia-induced EndoMT in vitro. Altogether, we show that neonatal hyperoxic exposure caused vascular remodeling and pulmonary hypertension in adulthood. This was associated with increased EndoMT. These novel observations provide mechanisms underlying hyperoxia-induced vascular remodeling and potential approaches to prevent BPD-associated pulmonary hypertension by targeting EndoMT.

Introduction

With advances in neonatal and perinatal care, premature infants can survive after an extremely short gestation (>22 weeks). Unfortunately, the mechanical ventilation and supplemental oxygen used to save them can also impair the growth of their pulmonary microvasculature and distal alveoli. This results in continued dependency on supplemental oxygen beyond 36 weeks corrected gestational age, referred to as bronchopulmonary dysplasia (BPD) [1]. This condition affects 10 000–15 000 premature infants annually in the USA. The pathology of BPD is characterized by alveolar simplification and dysregulated vascular development [1]. Although most BPD survivors eventually are weaned off oxygen, they may show evidence of pulmonary dysfunction and cardiovascular sequelae (e.g. pulmonary hypertension) as adolescents and adults [2–11]. Indeed, 30% of infants with moderate to severe BPD develop pulmonary hypertension [7,9,10,12]. However, the mechanisms underlying BPD-associated pulmonary hypertension are not fully understood.

Pulmonary vascular and right ventricle remodeling are common pathological features of pulmonary hypertension. Vascular remodeling is characterized by increased expression of smooth muscle cell-specific biomarkers. This results from proliferation and migration of vascular smooth muscle cells or originates from endothelial cells (ECs). Endothelial–mesenchymal transition (EndoMT) is a biological process whereby ECs progressively lose EC-specific biomarkers and gain a mesenchymal or myofibroblastic phenotype. It is observed in patients with idiopathic pulmonary arterial hypertension [13–15]. Pharmacological inhibition of EndoMT attenuates the progression of pulmonary arterial hypertension in rodents [14,16]. This suggests that EndoMT plays important roles in the pathogenesis of pulmonary arterial hypertension. However, whether EndoMT participates in the pathogenesis of BPD-associated pulmonary hypertension remains unclear.

At birth, mouse lungs are structurally similar to the lungs of human neonates born at 30–34 weeks of gestation, when the lung is in the saccular phase of development. Hyperoxic exposure in neonatal mice can be used to model lung injury in premature infants with BPD. Allowing for air recovery in these mice is commonly used to investigate persistence of pulmonary and cardiovascular sequelae. Therefore, we utilized neonatal mice to test the hypothesis that a 3-day neonatal hyperoxic exposure induces EndoMT, leading to the development of pulmonary hypertension in adulthood. Since male gender is associated with poorer outcomes in BPD and pulmonary hypertension [17–19], sex-specific differences in neonatal hyperoxia-induced pulmonary hypertension and EndoMT were assessed in mice and in cultured lung microvascular ECs (LMVECs) isolated from human donors at 18–22 weeks of gestation, respectively. Finally, we employed primary LMVECs isolated from neonatal mice to study potential mechanisms underlying hyperoxia-induced EndoMT.

Materials and methods

Cell culture

Primary LMVECs were isolated from neonatal mice (3–5 days old) and characterized as described previously [20]. These cells were cultured in dishes/plates precoated with 30 μg/ml human fibronectin using VascuLife^® EnGS-Mv medium, which was changed every 24 h. Primary LMVECs at passages 3–5 were used for experiments. Similarly, we also isolated LMVECs from hyperoxia-exposed mice for measuring Vim and Acta2 gene expression. Mouse fetal lung EC lines (MFLM-91U cells) were purchased from Seven Hills Bioreagents (Cincinnati, OH, USA), and grown in Ultra-Culture medium (BioWhittaker) [20]. We verified that there was no mycoplasma contamination. Human LMVECs, which were isolated from donors (four males and five females) at 18–22 weeks gestation in the canalicular stage of lung development, were purchased from ScienCell Research Laboratories (Carlsbad, CA, USA; Cat. # 3000). Characteristics of these cells are shown in Table 1. Human LMVECs were maintained in EC medium (Cat. # 1001; ScienCell Research Laboratories) containing 5% FBS, EC growth supplement, and antibiotics at 37 °C in 5% CO₂, and were used for experiments at passages 3–7. Cells were incubated with SB431542 (1 μm) for 24 h during the air recovery phase.

Table 1.

Characteristics of human donors for LMVECs.

Lot #	Sex	Gestation weeks
16008	Male	18
16021	Male	18
10885	Male	22
10899	Male	22
15900	Female	18
17777	Female	19
15902	Female	18
17800	Female	19
17779	Female	19

Hyperoxic exposure

Cells at 70–80% confluence were exposed to hyperoxia (95% O₂ and 5% CO₂) or air (21% O₂ and 5% CO₂) for 24 h, followed by normoxia (21% O₂ and 5% CO₂) for 24 h [20]. Culture media were changed every 24 h. Newborn C57BL/6J mice (< 12 h old) along with their mothers were exposed to room air or hyperoxia (>95% O₂) for 72 h in an A-chamber (BioSpherix, Parish, NY, USA) [20]. The dams were switched every 24 h between room air and hyperoxia to avoid injury. Some pups were allowed to recover in room air until postnatal day (pnd) 14, pnd60 or pnd120. There were no deaths in mice due to hyperoxic toxicity. All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Brown University.

Echocardiographic measurement

Under continuous inhalation of 1.5–2.0% isoflurane, mice were placed on a heated platform to maintain body temperature at 37 °C. Transthoracic echocardiography was performed using a 40 MHz linear-array transducer (Vevo 2100; FUJIFILM VisualSonics, Toronto, Canada) [21]. Tricuspid annular plane systolic excursion (TAPSE) was measured by aligning an M-mode cursor parallel with the right ventricle (RV) free wall as it meets the tricuspid annulus from the RV apical four-chamber view. TAPSE was determined by measuring the excursion of the tricuspid annulus from its highest position to the peak descent during ventricular systole. Two-dimensional Doppler and M-mode recordings were obtained to measure left ventricle (LV) end-systolic diameter, LV end-diastolic diameter, LV end-systolic volume, LV end-diastolic volume, LV stroke volume, LV ejection fraction, LV cardiac output, LV mass, LV mass corrected, LV end-systolic anterior wall thickness, LV end-diastolic anterior wall thickness, LV end-systolic posterior wall thickness, and LV end-diastolic posterior wall thickness.

In vivo hemodynamics

Right ventricle systolic pressure (RVSP), left ventricle systolic pressure (LVSP), and heart rate were measured using an opened-chest technique [21]. Mice were anesthetized with isoflurane inhalation (1.5–2.0%). A high-fidelity pressure-volume 1.0-Fr catheter (PVR-1030, Millar utilizing LabChart 8; AD Instruments, Colorado Springs, CO, USA) was inserted into the LV apex, and pressure measurements were recorded for 30–60 s. Subsequently, the same catheter was inserted into the apex of the RV, and pressure measurements were recorded for another 30–60 s.

Fulton index assessment

Following euthanasia with an overdose of ketamine and xylazine, lungs and hearts were harvested and dissected. RV and LV plus septum were weighed, and the Fulton index (weight of RV/[weight of LV + septum]) was calculated.

Immunohistochemistry

Pulmonary vessel thickness was calculated from sections immunostained for α-smooth muscle actin (α-SMA) [21]. In brief, mouse lungs were inflated with 1% low melt agarose at a pressure of 25 cm H₂O and fixed with 4% neutral buffered paraformaldehyde. Fixed lungs were embedded in paraffin and then sectioned at 5 μm thickness using a rotary microtome (MICROM International GmbH, Dreieich, Germany). Immunostaining for α-SMA was performed on lung sections by incubating slides with an α-SMA monoclonal antibody (supplementary material, Table S1) overnight at 4 °C, followed by a secondary antibody (Southern Biotech, Birmingham, AL, USA). A Zeiss Axiovert 200M Fluorescence Microscope was used to obtain digital images of the immunostaining. For each vessel, external diameter and two wall thicknesses were measured along two different axes based on α-SMA staining. Wall thickness was calculated as (external diameter − inner diameter)/external diameter × 100%, which was performed along two perpendicular axes. The two axis measurements were averaged together to determine the final wall thickness. A total of five or six vessels per mouse were utilized for calculating vascular wall thickness.

Immunofluorescence

Lung sections were deparaffinized and subjected to heat-mediated antigen retrieval in a citrate buffer solution (Vector Laboratories, Burlingame, CA, USA), then stained overnight at 4 °C with primary antibodies against vWF, vimentin, and α-SMA (supplementary material, Table S1). After incubation with secondary antibodies for 1 h at room temperature, sections were mounted in hard-set mounting medium containing DAPI (Vector Laboratories) and allowed to set overnight. Lung vessels with an outer diameter of ≤50 μm or >50 μm exhibiting luminal co-localization of vimentin/vWF or α-SMA/vWF were quantified in a total of six to ten vessels of 3–5 images per animal using a Zeiss Axiovert 200M Fluorescence Microscope. The data are expressed as the average of the percentage of vessels positive for EndoMT per mouse. The number of cells co-expressing vimentin/vWF or α-SMA/vWF was also counted in each vessel with an outer diameter of ≤50 μm. These experiments were carried out in a blinded manner.

Evaluating lung vascularization using a microCT

Pulmonary vascular morphometry (i.e. pulmonary artery branches) was determined and the number of microvessels (<300 μm diameter) was quantified using a Scanco Medical MicroCT 40 system (Scanco USA Inc, Southeastern, PA, USA) [22]. In brief, after ketamine/xylazine injection, the RV was cannulated. Heparinized saline was used to flush the lungs before infusion with a warm solution of 1% low-melting-point agarose and 30% barium (clinical radiology grade) into the pulmonary artery. The lungs were then fixed with 10% paraformaldehyde and the pulmonary vasculature was visualized using microCT. The number of microvessels <100 μm and ≥100 μm diameter was integrated and quantified using the Scanco software.

Determination of protein levels

Cells were homogenized using RIPA buffer. Protein samples (20–30 μg) were separated on a NuPAGE™ 4–12% Bis-Tris protein gel (Invitrogen, Carlsbad, CA, USA), and separated proteins were electroblotted onto nitrocellulose membranes. The membranes were blocked for 1 h at room temperature with 5% BSA, and then probed with primary antibodies against vWF, Pecam1 (CD31), vimentin, α-SMA, CD44, Snai2, and calnexin to determine the corresponding proteins. The primary antibodies used are listed in supplementary material, Table S1. Protein levels were measured using secondary antibodies in 5% BSA in PBS containing 0.1% Tween (v/v) 20 for 1 h linked to horseradish peroxidase (Vector Laboratories), and bound complexes were detected and assessed using a ChemiDoc™ Touch Imaging System (Bio-Rad Laboratories Inc, Hercules, CA, USA) using an enhanced chemiluminescence method (Millipore). Equal loading of the samples was determined by quantification of proteins as well as by reprobing membranes for the reference control calnexin or β-actin.

Measurement of steady-state mRNA levels

Total RNA was extracted using TRIzol reagent, and purified using the RNeasy miniprep kit (Qiagen, Valencia, CA, USA). RNA concentrations were measured spectrophotometrically (NanoDrop™ One Microvolume UV–Vis Spectrophotometer; Thermo Fisher Scientific, Wilmington, DE, USA). Four hundred nanograms of total RNA was used for reverse transcription using Taqman^® Reverse Transcription Reagents (Thermo Fisher Scientific). One microliter of cDNA was used for real-time qPCR reactions in a 7300 Real-Time PCR System (Applied Biosystems, Framingham, MA, USA). All Taqman gene probes were purchased from Thermo Fisher Scientific (supplementary material, Table S1). Gene expression was normalized to 18S rRNA levels. Relative RNA abundance was quantified by the comparative 2^−ΔΔCt method.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 7 (GraphPad Inc, San Diego, CA, USA). Results were expressed as mean ± SEM. A t-test was used for detecting statistical significance of the differences between means of two groups after checking the normality of data. The statistical significance of the differences among groups was evaluated by using one-way ANOVA for overall significance, followed by a Tukey–Kramer test. Statistical significance was accepted if p < 0.05.

Results

Neonatal hyperoxic exposure causes pulmonary vascular remodeling and impairs lung vascularization in adult mice

Pulmonary vascular remodeling contributes to increased pulmonary vascular pressures by increasing pulmonary vascular resistance [23]. Thus, we performed α-SMA immunohistochemistry to determine pulmonary vascular remodeling by morphometric analysis in mice exposed to hyperoxia as neonates. As shown in Figure 1A, neonatal hyperoxic exposure increased pulmonary vascular muscularization as indicated by α-SMA-positive vessels at pnd60 and pnd120. We then quantified vascular wall thickness in pulmonary arteries with <100 μm or ≥100 μm outer diameter in mice exposed to hyperoxia as neonates. As shown in Figure 1B, the vascular wall thickness was increased in pulmonary arteries of <100 μm at pnd60, which was further augmented at pnd120. However, the thickness of vascular wall in pulmonary arteries of ≥100 μm outer diameter was not significantly changed by neonatal hyperoxic exposure at either pnd60 or pnd20 (Figure 1C). At pnd120, the number of pulmonary arteries, detected by microCT, was significantly reduced in mice exposed to hyperoxia for 3 days as newborns compared with air-exposed controls (Figure 1D). These results demonstrate that neonatal hyperoxic exposure causes pulmonary vascular remodeling and reduces vascularization in adulthood.

Figure 1.

Neonatal hyperoxic exposure causes pulmonary vascular remodeling and impairs vascularization in adulthood. Neonatal mice (<12 h old) were exposed to air or hyperoxia (>95% O₂) for 3 days and allowed to recover in room air until pnd60 (A–C) or pnd120 (A–D). (A) Immunohistochemistry for α-SMA in lung sections at pnd60 and pnd120. Arrows denote vessels positive for α-SMA. Bar size = 50 μm. Wall thickness was calculated in pulmonary arteries with <100 μm (B) and ≥100 μm (C) outer diameter based on α-SMA staining. (D) MicroCT was performed to determine lung vascular branches, and the number of pulmonary arteries with <100 μm and ≥100 μm diameter was counted at pnd120. N = 5–11, *p < 0.05, **p < 0.01, ***p < 0.001 versus corresponding air group; ^†p < 0.05 versus O₂ group at pnd60.

Neonatal hyperoxic exposure causes RV remodeling and pulmonary hypertension in adult mice

We next determined whether neonatal hyperoxic exposure causes RV remodeling and pulmonary hypertension in adulthood. As shown in Figure 2A, a 3-day hyperoxic exposure in newborns increased the Fulton index, as indicated by increased weight ratio of RV to (LV plus septum) in mice at pnd60, and this was further augmented at pnd120. Neonatal hyperoxic exposure did not increase the levels of RVSP at pnd60 (Figure 2B). Compared with pnd60, the levels of RVSP were significantly increased at pnd120 in mice exposed to hyperoxia as neonates (Figure 2B). Neither heart rate nor body weight was altered at pnd120 in mice exposed to hyperoxia as neonates (Figure 2C,D). Altogether, these data demonstrate that neonatal hyperoxic exposure causes RV remodeling and pulmonary hypertension in adulthood.

Figure 2.

Neonatal hyperoxic exposure causes RV remodeling and pulmonary hypertension in adulthood, which is further augmented in male mice. C57BL/6J neonatal mice (<12 h old) were exposed to air or hyperoxia (>95% O₂) for 3 days, and then allowed to recover in room air until pnd60 (A, B) or pnd120 (A–G). (A) Fulton index was calculated as the weight ratio of RV to (LV plus septum). RVSP (B) and heart rate (C) were recorded using an opened-chest technique with a high-fidelity pressure-volume 1.0-Fr catheter. (D) Body weight was also recorded. (E) Immunohistochemistry for α-SMA in lung sections compared between male and female mice to measure wall thickness in pulmonary arteries with <100 μm outer diameter. (F) Fulton index was calculated as the weight ratio of RV to (LV plus septum) for male and female mice. (G) RVSP was recorded for male and female mice using an opened-chest technique with a high-fidelity pressure-volume 1.0-Fr catheter. Mean ± SEM. N = 5–12. *p < 0.05, **p < 0.01, ***p < 0.001 versus corresponding air group; ^†p < 0.05 versus O₂ group at pnd60 or O₂ group in male mice.

Male mice are more susceptible to developing neonatal hyperoxia-induced pulmonary vascular and RV remodeling as well as pulmonary hypertension

We and others have shown that male mice are more susceptible to developing neonatal hyperoxia-induced lung injury [24,25]. Therefore, we compared sex differences in neonatal hyperoxia-induced pulmonary vascular and RV remodeling, and pulmonary hypertension. As shown in Figure 2E–G, neonatal hyperoxic exposure increased vascular wall thickness, Fulton index, and RVSP in both male and female mice at pnd120. Compared with female mice, neonatal hyperoxia-induced pulmonary vascular and RV remodeling as well as RVSP were further augmented in male mice at pnd120 (Figure 2E–G). These results suggest that male mice are more susceptible to developing neonatal hyperoxia-induced pulmonary hypertension.

Neonatal hyperoxic exposure does not alter LV function in adult mice

Early exposure to hyperoxia causes LV dysfunction in adult mice [26]. We thus utilized a high-fidelity pressure–volume 1.0-Fr catheter and an echocardiogram to record the hemodynamic parameters of LV in mice exposed to hyperoxia as neonates at pnd120. As shown in Table 2, LV mass, size, volume, and wall thickness were not altered in mice exposed to hyperoxia as neonates. In addition, neonatal hyperoxic exposure did not affect LVSP, LV ejection fraction or cardiac output in these mice (Table 2). Furthermore, the TAPSE and RV free wall thickness were not significantly changed in the mice exposed to hyperoxia as newborns compared with air-exposed controls (Table 2). There were no differences in LVSP, LV ejection fraction or cardiac output between male and female mice exposed to hyperoxia as neonates (data not shown). These results suggest that neonatal hyperoxic exposure does not influence LV function in adulthood.

Table 2.

Neonatal hyperoxic exposure does not alter LV function in mice at pnd120.

Parameter	Air	O2
LV end-systolic diameter, mm	2.46 ± 0.064	2.53 ± 0.19
LV end-diastolic diameter, mm	3.77 ± 0.032	3.86 ± 0.19
LV end-systolic volume, μl	22.0 ± 1.31	24.5 ± 4.33
LV end-diastolic volume, μl	61.5 ± 1.49	66.1 ± 7.55
LV stroke volume, μl	39.5 ± 0.61	41.7 ± 3.6
LV mass, mg	116.2 ± 4.10	122.7 ± 7.79
LV mass corrected, mg	93.0 ± 3.29	98.0 ± 6.19
LV end-systolic anterior wall thickness, mm	1.39 ± 0.079	1.38 ± 0.052
LV end-diastolic anterior wall thickness, mm	0.95 ± 0.073	0.89 ± 0.038
LV end-systolic posterior wall thickness, mm	1.16 ± 0.030	1.31 ± 0.071
LV end-diastolic posterior wall thickness, mm	0.74 ± 0.033	0.83 ± 0.034
LV systolic pressure, mmHg	518.7 ± 10.9	597.0 ± 33.5
LV ejection fraction, %	64.4 ± 1.35	64.3 ± 2.84
LV cardiac output, ml/min	20.9 ± 0.45	21.4 ± 1.64
TAPSE, mm	0.58 ± 0.043	0.51 ± 0.023
RV free wall thickness, mm	0.37 ± 0.049	0.39 ± 0.042

Mean ± SEM, N = 6. TAPSE, tricuspid annular plane systolic excursion.

Neonatal hyperoxic exposure causes EndoMT in mouse lungs

EndoMT plays an important role in the pathogenesis of idiopathic pulmonary arterial hypertension [13–15,27,28]. It is unclear whether neonatal hyperoxic exposure causes EndoMT in mouse lungs. To answer this question, we performed dual immunofluorescence for vWF and vimentin or α-SMA, biomarkers of ECs and mesenchymal cells, respectively, in mice exposed to hyperoxia as neonates. As shown in Figure 3A,B, a mixed phenotype of ECs and mesenchymal cells (vWF⁺/vimentin⁺ or vWF⁺/α-SMA⁺) was observed in hyperoxia-exposed mouse lungs at pnd60 and pnd120, whereas cells co-expressing vimentin/vWF or α-SMA/vWF were present at low levels in the air-exposed groups. Lung vessels with this mixed phenotype were present at higher levels at pnd60 compared with pnd14 and pnd120 in response to hyperoxic exposure (Figures 3C,D and 4A), suggesting that EndoMT peaked at pnd60. Approximately 35–40% of vessels with an outer diameter ≤50 μm co-expressed vimentin/vWF or α-SMA/vWF, while cells co-expressing vimentin/vWF or α-SMA/vWF were present in ~4% of vessels with an outer diameter greater than 50 μm at pnd60 after hyperoxic exposure (Figure 3C,D). Furthermore, levels of Vim (vimentin) and Acta2 (α-SMA) mRNA were increased in LMVECs isolated from neonatal hyperoxia-exposed mice at pnd14, while these genes were not changed in LMVECs isolated from mice exposed to hyperoxia without air recovery (pnd3) (Figure 4B). Taken together, these results suggest that EndoMT is mainly present in remodeled vessels during the development of neonatal hyperoxia-induced pulmonary hypertension.

Figure 3.

Neonatal hyperoxic exposure causes EndoMT in mouse lungs. C57BL/6J neonatal mice (<12 h old) were exposed to air or hyperoxia (>95% O₂) for 3 days, and then allowed to recover in room air until pnd14, pnd60, and pnd120. (A) Immunofluorescence was performed for co-staining of vWF and vimentin in lungs of mice exposed to hyperoxia as neonates. Representative images are shown of mouse lungs at pnd60 and pnd120. Arrows denote cells co-expressing vimentin and vWF. (B) Immunofluorescence was performed for co-staining of vWF and α-SMA in lungs of mice exposed to hyperoxia as neonates. Representative images are shown of mouse lungs at pnd60 and pnd120. Arrows denote cells co-expressing α-SMA and vWF. (C, D) Quantification of lung vessels with outer diameter ≤50 μm or >50 μm exhibiting luminal co-localization of (C) vimentin/vWF or (D) α-SMA/vWF at pnd14, pnd60, and pnd120 when these mice were exposed to hyperoxia as newborns. Bar size = 20 μm. Mean ± SEM, N = 5. *p < 0.05, ***p < 0.001 versus corresponding air group.

Figure 4.

Expression of vimentin and α-SMA was increased in lung ECs of mice exposed to hyperoxia as neonates. C57BL/6J neonatal mice (<12 h old) were exposed to air or hyperoxia (>95% O₂) for 3 days. Some mice were then allowed to recover in room air until pnd14, pnd60, and pnd120. (A) Immunofluorescence was performed for co-staining of vWF with vimentin or α-SMA in lungs of mice exposed to hyperoxia as neonates. The number of ECs positive for vimentin/vWF and α-SMA/vWF per vessel with outer diameter ≤50 μm was counted. (B) Expression of Vim and Acta2 genes in LMVECs, which were sorted from hyperoxia-exposed mice, was measured using RT-qPCR. Mean ± SEM, N = 5. *p < 0.05, **p < 0.01, ***p < 0.001 versus corresponding air group.

Hyperoxic exposure causes EndoMT in cultured primary mouse LMVECs

To further determine whether hyperoxic exposure induces mesenchymal transition in cultured ECs, we isolated LMVECs from neonatal mice (3–5 days old) and exposed them to hyperoxia for 24 h followed by air recovery for 24 h, as described previously [29]. As shown in Figure 5A, hyperoxic exposure followed by air recovery reduced Vwf (vWF) and Pecam1 mRNA levels, but increased those of Acta2, Vim, Snai2, and Cd44. Similarly, protein levels of vWF and Pecam1 were reduced, whereas levels of α-SMA, vimentin, Snai2, and CD44 proteins were significantly increased in MFLM-91U cells exposed to hyperoxia followed by air recovery (Figure 5B). Snai2 is a transcription factor that represses VE-cadherin transcription to allow EndoMT [30], while CD44 plays an essential role for the initiation and propagation of EndoMT by stabilizing a cysteine–glutamate antiporter [15]. This suggests that hyperoxic exposure followed by air recovery causes EndoMT in cultured lung ECs.

Figure 5.

Hyperoxic exposure causes EndoMT in cultured mouse lung ECs. (A) Primary mouse LMVECs and (B) MFLM-91U cells were exposed to hyperoxia for 24 h followed by normoxia for 24 h. (A) Expression of Vwf, Pecam1, Acta2, Vim, Cd44, and Snai2 genes was measured by RT-qPCR. (B) Levels of vWF, Pecam1, α-SMA, vimentin, CD44, and Snai2 proteins were measured by immunoblotting. N = 4–6. *p < 0.05, **p < 0.01, ***p < 0.001 versus air group.

Hyperoxic exposure-induced EndoMT is augmented in cultured human LMVECs from male donors

Although male mice were more susceptible to neonatal hyperoxic exposure-induced pulmonary hypertension, it is not clear whether sex differences in hyperoxia-induced EndoMT account for these differences. To answer this question, we utilized human LMVECs from male and female donors at 18–22 weeks of gestation, and exposed them to hyperoxia for 24 h followed by air recovery for 24 h. As shown in Figure 6A, hyperoxic exposure reduced VWF and PECAM1 mRNA levels in LMVECs from both male and female donors. Furthermore, expression of ACTA2 and VIM genes was significantly increased in hyperoxia-exposed LMVECs from both male and female donors (Figure 6A). In addition, we found that LMVECs from male donors showed further reduction in VWF and PECAM1 but had increased ACTA2 and VIM mRNA levels, compared with those from female donors in response to hyperoxic exposure (Figure 6A). These results indicate that lung ECs from male donors are more susceptible to developing hyperoxia-induced EndoMT.

Figure 6.

Hyperoxic exposure-induced EndoMT is augmented in cultured human LMVECs from male donors, and hyperoxic exposure activates the TGF-β/Smad pathway resulting in EndoMT in cultured mouse LMVECs. (A) Human LMVECs from male and female donors were exposed to hyperoxia for 24 h followed by normoxia for 24 h (refers to O₂). Expression levels of VWF, PECAM1, ACTA2, and VIM mRNAs were measured using RT-qPCR. (B, C, F–H) Primary mouse LMVECs and (D, E) MFLM-91U cells were exposed to hyperoxia for 24 h followed by normoxia for 24 h (refers to O₂). (B–E) Immunoblotting was performed to assess the phosphorylation of Smad2 and Smad3, total Smad2/3, and calnexin. (B) Densitometry of phosphorylated Smad2 and Smad3 bands normalized to calnexin, whereas in E, total Smad2/3 content was used to normalize the densitometry of phosphorylated Smad2 and Smad3 bands. (F) Smad7 gene expression was measured by RT-qPCR. (G) Smad7 protein was measured by immunoblotting, with calnexin used as a reference control. (H) Cells were incubated with SB431542 (1 μm) for 24 h during the air recovery phase. Expression of Vwf, Pecam1, Vim, and Snai2 genes was measured using RT-qPCR. Mean ± SEM. N = 4 or 5. *p < 0.05, **p < 0.01, ***p < 0.001 versus air; ^†p < 0.05, ^††p < 0.01 versus O₂ or O₂/male donor group.

The transforming growth factor-β (TGF-β)/Smad pathway is involved in hyperoxia-induced EndoMT in cultured lung ECs

The TGF-β/Smad pathway is a critical signal for triggering EndoMT. This is characterized by increased phosphorylation of Smad2 and Smad3 [20]. Therefore, we first measured the phosphorylation of Smad2 (Ser465/467) and Smad3 (Ser423/425) in mouse LMVECs exposed to hyperoxia followed by air recovery. As shown in Figure 6B–E, hyperoxic exposure significantly increased the phosphorylation of both Smad2 and Smad3 in mouse LMVECs and MFLM-91U cells. Another member of the Smad family, Smad7, plays an essential role in negatively regulating TGF-β signaling. Thus, we measured Smad7 mRNA using RT-qPCR and protein levels using immunoblotting in cells exposed to hyperoxia followed by air recovery. As shown in Figure 6F,G, hyperoxic exposure reduced Smad7 protein but did not alter its mRNA level in mouse LMVECs. To further determine the role of the TGF-β pathway in hyperoxia-induced EndoMT, we incubated LMVECs with a specific TGF-β inhibitor, SB431542 (1 μm, 24 h), during the air recovery phase. As shown in Figure 6H, incubation with SB431542 protected against hyperoxia-induced EndoMT in mouse LMVECs. These results suggest that hyperoxic exposure activates the TGF-β/Smad pathway, leading to EndoMT.

Discussion

Here, we have shown that hyperoxic exposure of newborn mice caused pulmonary vascular and RV remodeling as well as pulmonary hypertension in adulthood. Furthermore, male mice exposed to hyperoxia as neonates were more susceptible to developing pulmonary hypertension than females. Neonatal hyperoxic exposure also induced EndoMT, which was associated with activation of the TGF-β/Smad pathway. Hence, hyperoxic exposure causes EndoMT, which may result in pulmonary vascular remodeling and pulmonary hypertension.

A variety of hyperoxic exposure strategies are utilized to mimic BPD-associated pulmonary hypertension in rodents [31–37]. In those studies, exposure of newborns to 70–90% O₂ for 1–2 weeks with or without room air recovery for 1–6 weeks had been employed. Those exposures are significantly longer than used in our model. We rationalized that since mouse lung alveolar formation begins at pnd4, a short exposure early in neonatal life may be sufficient to result in long-term sequelae and therefore exposed C57BL/6J newborn mice to hyperoxia for 3 days during the saccular stage of lung development. Mice were exposed to greater than 95% oxygen to mimic severe BPD, as pulmonary hypertension is observed in infants with moderate to severe BPD [12,38]. These mice gradually developed pulmonary vascular and RV remodeling as well as pulmonary hypertension by 4 months of age. Concurrently, alveolar simplification was also observed in adult mice after exposure to hyperoxia as neonates [39]. Therefore, neonatal hyperoxic exposure causes persistent injurious effects on lung alveolarization and vascular remodeling. Further investigation using lower concentrations of oxygen that are utilized clinically for premature babies would reveal whether neonatal hyperoxia causes a concentration-dependent development of pulmonary hypertension. There are many potential confounding factors, such as prematurity, mechanic ventilation or infection, that are known risk factors for BPD. It will be important to determine whether these are independently involved in the development of pulmonary hypertension. This will need to be determined in larger animals where these confounders can be modeled.

We noticed that neonatal hyperoxia caused pulmonary vascular thickness and RV remodeling at pnd60. This remodeling seems compensatory and adaptive, as the RVSP measured by an open-chest technique was not altered by hyperoxia at this time point. Although this technique is an invasive and time-consuming procedure, it is able to precisely measure RVSP. Further examination on RV wall stiffening and collagen deposition may reveal discrepancies between increased Fulton index (RV weight) and no changes in RV thickness in hyperoxia-exposed mice at pnd120.

Systemic hypertension is common in preterm infants with severe BPD [40,41]. A previous study has shown that neonatal hyperoxic exposure (85% O₂) for 2 weeks in mice caused LV dysfunction without alterations in LVSP at the age of 12–14 weeks [26]. In contrast, neonatal hyperoxic exposure between pnd3 and pnd10 increased systemic blood pressure in adult rats [42]. Our findings showed that neonatal hyperoxic exposure (>95% O₂) for 3 days in mice did not alter LV size, volume, cardiac output or systolic pressure at 4 months. These discrepancies may be due to differences in species, duration or concentration of exposure. Follow-up beyond 4 months with analysis of the histology of the heart and major vessels will be required to confirm whether a short hyperoxic exposure leads to long-term changes in cardiac function.

Compared with females, male mice are more susceptible to developing neonatal hyperoxia-induced lung injury [24,25,43]. This may be due to sex differences in the pulmonary transcriptome [44]. In addition, male sex is associated with increased severity of pulmonary hypertension in BPD patients [19]. This is corroborated by our findings that male mice were sensitive to developing neonatal hyperoxia-induced vascular remodeling and pulmonary hypertension. ECs from male donors were more susceptible to hyperoxia-induced EndoMT. This is in agreement with a previous report showing that hyperoxic exposure reduced PECAM1 but increased α-SMA protein levels in lung ECs from male donors [45]. The mechanisms underlying these findings remain unclear. Further research using lung tissues from patients with pulmonary hypertension resulting from BPD to detect EndoMT would enhance the translational relevance of our findings in this disease. Nevertheless, all our findings suggest that hyperoxia-induced EndoMT may contribute to the development of pulmonary vascular remodeling and pulmonary hypertension. Interestingly, EndoMT was not observed in mice exposed to hyperoxia without air recovery, suggesting that EndoMT is initiated during the air recovery phase. EndoMT can be detected only in a narrow time window by immunofluorescence when ECs express both endothelial and mesenchymal markers. Further studies using a cell-lineage tracing approach are required to confirm direct evidence of EndoMT in cells originating from ECs in mouse lungs after neonatal hyperoxic exposure. A previous report and our preliminary data show that lung EC proliferation was increased in premature infants with BPD [46] and in neonatal hyperoxia-exposed mice (data not shown). It is unclear whether hyperoxia induces proliferation of ECs or vascular smooth muscle cells, thereby contributing to the development of vascular remodeling and pulmonary hypertension.

Among the signaling networks responsible for EndoMT, the principal mediators are the TGF-β superfamily of proteins [47]. Canonically, TGF-β binds to the type II receptors, which recruit and activate the type I TGF-β receptor ALK5. The latter in turn phosphorylates Smad2 and Smad3, which form a complex with Smad4. This results in the transcription of mesenchymal cell genes in ECs, leading to EndoMT. To counter this, Smad7 antagonizes TGF-β signaling by binding to ALK5 and inhibiting the recruitment and phosphorylation of Smad2 and Smad3 [48]. Previous studies have shown that neonatal hyperoxic exposure spatiotemporally affects TGF-β signaling in mouse lungs [49–52]. We found that hyperoxic exposure increased the phosphorylation of both Smad2 and Samd3, and reduced Smad7 protein levels in cultured lung ECs. Blocking the TGF-β pathway using a specific inhibitor attenuated hyperoxia-induced EndoMT. A previous study showed that a TGF-β-neutralizing antibody protects against hyperoxia-induced alveolar and vascular simplification in neonatal mice [53]; therefore the TGF-β/Smad pathway plays a potential role in mediating hyperoxia-induced EndoMT and subsequent pulmonary hypertension. Nevertheless, the roles of hyperoxia-induced activation of the BMP and Wnt/β-catenin pathways in mediating EndoMT are not excluded [49,50]. Whether the lung TGF-β signal exhibits sex differences in mice exposed to hyperoxia remains unclear. Neonatal hyperoxic exposure also causes activation of Smad2 and Smad3 in fibroblasts [49], which may contribute to fibrotic responses in the lung. It is interesting to note that hyperoxic exposure did not alter the Smad7 mRNA level but reduced its protein level. It is possible that hyperoxic exposure causes Smad7 posttranslational modifications, including methylation and acetylation, leading to degradation [54,55]. Furthermore, neonatal hyperoxia causes enrichment of DNA methylation on the TGF-β signaling pathway [56,57], which may account for TGF-β activation even after months of air breathing.

In conclusion, neonatal hyperoxic exposure causes vascular and RV remodeling as well as pulmonary hypertension, especially in male mice. Hyperoxic exposure induced EndoMT in mouse lungs and cultured lung ECs, which is associated with an activation of the TGF-β/Smad pathway. Our novel findings demonstrate sex differences in neonatal hyperoxia-induced pulmonary hypertension and provide potential therapeutic approaches by targeting EndoMT and the TGF-β/Smad pathway to prevent pulmonary vascular remodeling in premature infants with BPD.