Paxillin in mechanosensitive spatial control of lung endothelial cell regeneration

Abstract

Mechanical forces altered after unilateral pneumonectomy (PNX) control post-PNX lung growth. Here, we demonstrate that post-PNX endothelial regeneration is stimulated at the peripheral region of the mouse lung, requiring the focal adhesion (FA) protein, paxillin, while inhibition of mechanical tension attenuates the effects. Paxillin mediates stretching-induced YAP1-TEAD1 signaling, which stimulates GATA2 activity on angiogenic factor angiopoietin-2 (ANGPT2) transcription in endothelial cells (ECs), dictating post-PNX endothelial regeneration at the peripheral region. Deleting endothelial paxillin suppresses expression of GATA2 in the specific EC subtype, capillary type 1 ECs (CAP1s) following PNX. Extracellular matrix protein, collagen VI that impacts cell mechanical responses is expressed more at the peripheral region in the post-PNX mouse lungs, which drives paxillin expression, leading to EC regeneration. Mechanosensitive paxillin signaling in ECs mediates spatial control of post-PNX endothelial regeneration.

Introduction

The formation of the vascular system is one of the most important processes in organ development and regeneration^1–3. Tissue-specific endothelium establishes vascular niches³, secretes angiocrine factors⁴, and promotes angiogenesis and regenerative growth after organ resection (e.g., lung⁵, liver⁶). We and others have reported that inhibition of angiogenesis suppresses regenerative lung growth after pneumonectomy (PNX)^5,7–11. Mechanical forces (e.g., cell density^12,13, extracellular matrix (ECM) stiffness¹⁴, flow¹⁵, pressure¹¹, stretch¹⁶) control angiogenesis³, and alteration of mechanical environment is one of the important initiating factors for post-PNX lung growth^3,11,17–23. Increases in pulmonary artery (PA) pressure and shear stress in the remaining lungs stimulate angiogenesis and induce post-PNX lung growth^11,21,23. Stretching forces associated with mechanical deformation and expansion of remaining alveoli after resection of the lung lobe also increase after PNX^17,20,22, which alters actin cytoskeleton structures and stimulates cell proliferation²⁰, resulting in lung growth after PNX^8,20–22. The levels of stretching and tissue deformation are higher at the peripheral regions of the post-PNX lung lobes compared to proximal regions^17–19, which correlates with EC proliferation¹¹. The signaling mechanism by which stretching forces control spatial changes in EC regeneration after PNX remains unclear.

ECMs and their mechanical properties (e.g., stiffness) control angiogenesis^13,14,24 and lung vascular development²⁵. The ECMs provide structural and mechanical integrity to maintain cell mechanics and tissue homeostasis. The ECM impacts cell mechanical responses, mechanotransduction, and mechanosensing, which contributes to tissue remodeling²⁶. We have reported that ECM protein, collagen VI (COL6), contributes to tissue development and tumor progression^13,27,28. Focal adhesion (FA) protein, paxillin, senses mechanical forces^29–31 and controls cell proliferation, migration, and tissue remodeling^32–39. We have demonstrated that paxillin senses mechanical forces and regulates the directional motility of ECs to organize angiogenesis in the neonatal retina^30,39. We have also reported that alveolar epithelial paxillin mediates postnatal lung alveolar morphogenesis and surfactant homeostasis⁴⁰. Thus, paxillin may mediate post-PNX EC regeneration in the lungs, in which stretching forces are altered.

Here, we have demonstrated that spatial changes in the expression patterns of ECMs, including COL6, alter the response to stretching after PNX and stimulate angiogenesis at the peripheral region through paxillin signaling. Paxillin senses stretching forces and mediates increases in the activity of mechanosensitive transcription factor GATA2 to stimulate angiogenic factor ANGPT2 expression in ECs and EC proliferation during post-PNX lung growth at the peripheral region. Modulation of ECMs that potentially target paxillin or associated EC signaling could lead to the ECM-based strategy for lung regeneration/repair.

Results

Stretching forces induced after PNX stimulate EC proliferation and angiogenic signaling at the peripheral region of the lung

To investigate the effect of mechanical tension on post-PNX EC regeneration, we performed left PNX or in combination with prosthesis insertion, in which a silicone prosthesis that imitates the shape and size of the left lung lobe was inserted into the resected area at the time of PNX. The remaining lung size and the ratio of the weight of the right lung lobe to mouse body weight (BW) increased 7 days after left PNX, while prosthesis insertion inhibited post-PNX lung growth (Fig. 1a and Supplementary Fig. 1a). Lung wet-to-dry weight (W/D) ratio⁴¹ did not change, suggesting that post-PNX lung growth was not due to water accumulation (Supplementary Fig. 1a). Stretching strain of the lung lobes changes greater at the peripheral region after PNX compared to the proximal region^17–19. CD31⁺ vessel area, the number of SPB⁺ AT2 cells as well as Ki67⁺ proliferating ERG⁺ ECs increased at the peripheral region of the right lung lobe after left PNX, while prosthesis insertion suppressed the effects (Fig. 1b and Supplementary Fig. 1b). The number of proliferating ECs also increased at the proximal region; however, the increased levels were lower than those at the peripheral region (Supplementary Fig. 2a). FACS analysis of BrdU⁺ ECs confirmed that BrdU⁺ CD31⁺ EC populations at the peripheral region of the lungs increased after PNX, which was inhibited by prosthesis insertion (Fig. 1c and Supplementary Fig. 2b).

Fig. 1. Increases in mechanical tension after PNX are necessary for EC regeneration.

a The ratio of the weight of right lung lobes to mouse body weight (BW) 7 days after PNX or in combination with prosthesis insertion (n = 6, mean ± s.e.m., *p < 0.05). b CD31⁺ blood vessels, SPB⁺ AT2 cells, and DAPI (top) and Ki67⁺ ERG⁺ ECs and DAPI (bottom) at the peripheral region of right lung lobe 7 days after PNX or in combination with prosthesis insertion (insert: higher magnification image). Scale bars, 25 μm. Quantification of vessel area (left), SPB⁺ AT2 cell number (middle), and Ki67⁺ ERG⁺ EC number (right, n = 6, mean ± s.e.m., *p < 0.05). c BrdU⁺ CD31⁺ ECs at the peripheral region of right lung lobes 7 days after PNX or in combination with prosthesis insertion (n = 5, mean ± s.e.m., *p < 0.05). d ERG⁺ ECs, ANGPT2 expression and DAPI (top) and CD31⁺ blood vessels, TIE2 expression, and DAPI (bottom) at the peripheral region of right lung lobe 7 days after PNX or in combination with prosthesis insertion (insert: higher magnification image). Scale bars, 25 μm. Quantification of ANGPT2 and TIE2 expression (n = 6, mean ± s.e.m., *p < 0.05).

Stretching forces stimulate angiogenesis by changing angiogenic factor expression¹⁶. Expression of angiogenic factor ANGPT2 and its receptor TIE2 (IHC) increased at the peripheral region of the post-PNX mouse lungs, while prosthesis insertion inhibited the effects (Fig. 1d). The protein levels of ANGPT2 and TIE2 (IHC) also increased at the proximal region after PNX; however, the increased levels were lower than those at the peripheral region (Supplementary Fig. 2a). We also confirmed that the protein levels of ANGPT2 (ELISA) and TIE2 (IB) increased in ECs (CD31⁺ cells) isolated from post-PNX mouse lungs, while prosthesis insertion attenuated the effects (Supplementary Fig. 2c–e).

To directly examine the effects of mechanical strain, we applied uniaxial 10% cyclic stretch to ECs adherent to flexible ECM substrates. Stretched ECs aligned perpendicular to the axis of the applied strain¹⁶ and the levels of Angpt2 and TIE2 increased in the stretched human lung microvascular ECs (L-HMVECs) (Fig. 2a, b) and ECs isolated from mouse lungs (Supplementary Fig. 2f). GATA2 senses ECM stiffness and controls angiogenesis¹⁴. Consistent with the response to stiffness, the levels of GATA2 increased in the stretched L-HMVECs (Fig. 2a, b). IHC analysis confirmed the co-expression of ANGPT2 and GATA2, and the levels increased more at the peripheral region after PNX; in contrast, the effects were suppressed by prosthesis insertion (Fig. 2c and Supplementary Fig. 2g). ANGPT2 has GATA binding motifs on their promoter site⁴². GATA2 bound to the ANGPT2 promoter region more in the stretched L-HMVECs compared to those under a static condition; control IgG immunoprecipitated these DNAs at a significantly lower level (ChIP-PCR, Fig. 2d). GATA2 knockdown decreased ANGPT2 expression in the stretched L-HMVECs (Fig. 2e), suggesting that GATA2 mediates stretching-induced ANGPT2 expression in ECs.

Fig. 2. Stretching forces increase GATA2 and ANGPT2 expression in ECs.

a Phase contrast images of L-HMVEC alignment under stretching (top). Scale bars, 40 μm. Immunofluorescence (IF) micrographs showing ANGPT2, GATA2, and DAPI in L-HMVECs under stretching (bottom). Scale bars, 10 μm. Quantification of ANGPT2 and GATA2 expression (n = 5, mean ± s.e.m., *p < 0.05). b mRNA and protein levels of ANGPT2 (left, 2nd) and mRNA levels of TIE2 (3rd) and GATA2 (right) in L-HMVECs under stretching (n = 4-5, mean ± s.e.m., *p < 0.05). c IF micrographs showing GATA2 and ANGPT2 expression and DAPI at the peripheral region of sham vs. post-PNX (7 days) mouse lungs, or in combination with prosthesis insertion (insert: higher magnification image). Scale bars, 20 μm. Quantification of GATA2 expression (n = 5, mean ± s.e.m., *p < 0.05). d ChIP analysis showing the IP levels of ANGPT2 promoters co-immunoprecipitating (co-IP) with control IgG or GATA2 antibody in stretched L-HMVECs (n = 3, mean ± s.e.m., *p < 0.05). e mRNA levels of GATA2 and ANGPT2 in L-HMVECs treated with GATA2 siRNA under stretching (n = 4, mean ± s.e.m., *p < 0.05).

Mechanical strain altered by PNX induces expression of paxillin in the mouse lungs

We have demonstrated that paxillin senses mechanical forces to control directional cell motility^30,39 and regulates neonatal retinal angiogenesis^30,39 as well as postnatal lung alveolar morphogenesis⁴⁰. Thus, we next investigated whether paxillin mediates post-PNX EC regeneration in the lungs, in which stretching forces are altered. Paxillin expression significantly increased at the peripheral region after PNX, while prosthesis insertion inhibited the effects (Fig. 3a). The levels of paxillin also increased at the proximal region after PNX; however, the levels were significantly lower than those at the peripheral region (Fig. 3a).

Fig. 3. Paxillin mediates post-PNX EC regeneration in the mouse lungs.

a IF micrographs showing ERG⁺ ECs, paxillin expression, and DAPI at the peripheral or proximal region of right lung lobe 7 days after PNX or in combination with prosthesis insertion (insert: higher magnification image). Scale bars, 20 μm. Quantification of paxillin expression (n = 5, mean ± s.e.m., *p < 0.05). b mRNA levels of paxillin in ECs isolated from Pxn^fl/fl or Pxn^iΔEC mouse lungs 7 days after PNX (n = 5–6, mean ± s.e.m., *p < 0.05). c The ratio of the weight of right lung lobes to mouse BW in Pxn^fl/fl or Pxn^iΔEC mice 7 days after PNX (n = 6, mean ± s.e.m., *p < 0.05). d IF micrographs showing ERG⁺ ECs, ANGPT2 expression, and DAPI (top) and CD31⁺ ECs, TIE2 expression, and DAPI (bottom) at the peripheral region of Pxn^fl/fl or Pxn^iΔEC mouse lung right lobe 7 days after PNX (insert: higher magnification image). Scale bars, 20 μm. Quantification of ERG⁺ EC number, ANGPT2 and TIE2 expression (n = 5–6, mean ± s.e.m., *p < 0.05).

To examine the effects of paxillin knockdown in ECs, we developed Pxn^fl/fl-Cdh5(PAC)-Cre^ERT2 (Pxn^iΔEC) mice, in which Pxn^fl/fl mice (obtained from Dr. Christopher Turner, SUNY Upstate Medical University, Jackson Laboratory, stock # 35946)^33,36 were crossed with Cdh5(PAC)-Cre^ERT2 mice (obtained from Dr. Ralf Adams, Max Planck Institute)⁴³, an inducible cre deleter under the control of Cdh5 promoter. Paxillin expression in ECs was suppressed (Fig. 3b and Supplementary Fig. 3a), and post-PNX increases in lung growth, ERG⁺ EC density, and expression of ANGPT2 and TIE2 were inhibited in tamoxifen-induced Pxn^iΔEC mouse lungs and isolated lung ECs (Fig. 3c, d and Supplementary Fig. 3b).

In cultured ECs, the levels of paxillin were also higher in the stretched L-HMVECs and in ECs isolated from the post-PNX mouse lungs (Figs. 3b and 4a, b). Paxillin knockdown in human ECs decreased ANGPT2 promoter activity (HUVECs, Supplementary Fig. 3c) and cyclic stretch-induced GATA2 expression (L-HMVECs, Fig. 4c). Binding of GATA2 to the Angpt2 promoter region was also suppressed in tamoxifen-induced Pxn^iΔEC mouse lung ECs under stretching (Fig. 4d). We also analyzed the effects of EC paxillin knockdown on cellular phenotypes; (i) DNA synthesis was significantly suppressed in ECs and alveolar epithelial cells (EPCAM⁺ cells) isolated from Pxn^iΔEC mouse lungs (Supplementary Fig. 3d); (ii) consistent with the earlier report in vitro⁴⁴, paxillin knockdown did not significantly disrupt EC junctional integrity in the mouse lungs (Supplementary Fig. 3a); (iii) although lamellar structure of lamellar body (LB) did not significantly change in Pxn^iΔEC mouse lungs compared to control Pxn^fl/fl mice (TEM, Supplementary Fig. 3a), the levels of surfactant protein were significantly lower in the Pxn^iΔEC mouse lungs (ELISA, Supplementary Fig. 3e).

Fig. 4. Paxillin mediates stretching-induced GATA2 expression in ECs.

a IF micrographs showing paxillin expression and DAPI in L-HMVECs under stretching (left). Scale bars, 30 μm. Quantification of paxillin expression (right, n = 6, mean ± s.e.m., *p < 0.05). b mRNA levels of paxillin in L-HMVECs under stretching (n = 4, mean ± s.e.m., *p < 0.05). c mRNA levels of paxillin and GATA2 in L-HMVECs treated with paxillin siRNA under stretching (n = 4, mean ± s.e.m., *p < 0.05). d ChIP-PCR analysis showing the IP levels of Angpt2 promoter co-IP with control IgG or GATA2 antibody in the stretched ECs isolated from Pxn^fl/fl or Pxn^iΔEC mouse lungs (n = 3, mean ± s.e.m., *p < 0.05).

Biological processes related to angiogenesis are altered in bulk RNA sequencing (RNA-seq) analysis of post-PNX mouse lung ECs

We next used our published bulk RNA-seq data (GSE179227)⁴⁵ of sham-operated (control) and post-PNX mouse lung ECs (7 days), which detected 1156 genes (923 upregulated and 233 downregulated) being significantly differentially expressed. The top 50 significant Gene Ontology biological processes (GO-BP) terms included cell cycle process, chromosome segregation, extracellular matrix organization, extracellular structure organization, and cytoskeleton organization, which contribute to angiogenesis and tissue morphogenesis (Supplementary Fig. 4). These results indicate that genes controlling cell proliferation and vascular morphogenesis are altered in lung ECs following PNX.

Single-cell RNA sequencing (scRNA-seq) analysis of EC subpopulations and gene expression in post-PNX mouse lungs

Recent scRNA-seq analysis has demonstrated that lung ECs have different EC subpopulations, including arterials, capillary type 2 ECs (CAP2s), capillary type 1 ECs (CAP1s), and venous^46–48. The scRNA-seq analysis of whole lungs or peripheral (peripheral regions of each lobe (1.5 mm from the edge) were horizontally separated) vs. proximal regions of the lungs (pooled n = 4 per condition, Supplementary Fig. 5a) revealed that the percentage and the number of proliferating ECs that express cell-cycle markers (e.g., Mki67, Top2a, Cdk1)⁴⁹ significantly increased after PNX (Fig. 5a, and Supplementary Figs. 5b, c and 6a–c). When we further analyzed the proliferating EC cluster, while it included CAP1 and other EC types including arterial, venous, and CAP2 (red boxes), these EC subclusters expressed CAP1 markers (blue boxes) during post-PNX lung growth (Supplementary Fig. 6d, e), indicating that these CAP1 marker expressing subclusters behave like EC progenitors and potentially differentiate into other EC subtypes as previously reported^46–48. Thus, we combined the proliferating progenitor like CAP1 marker expressing EC cluster (called as “proliferating ECs”) and non-proliferating CAP1 cluster as “CAP1” for the subsequent analysis of the gene expression profiles. The number of BrdU⁺ Aplnr⁺ proliferating ECs increased at the peripheral regions of the post-PNX mouse lungs (Fig. 5b). Paxillin expression was significantly increased in CAP1s in the post-PNX mouse lungs (Fig. 5c). We confirmed the results using RNA in situ hybridization (RNA-ISH)/immunohistochemistry (IHC); the levels of paxillin and phosphorylation levels of paxillin (Tyr 118), which controls cellular function, in the Aplnr⁺ CAP1s significantly increased after PNX, and the increased levels were more at the peripheral regions compared to those at the proximal regions (Fig. 5d and Supplementary Fig. 7a). In the scRNA-seq data, Angpt2 expression was also significantly increased in the CAP1s at the peripheral region after PNX (Supplementary Fig. 7b), which was confirmed by RNA-ISH/IHC (Supplementary Fig. 7c). Gata2 expression was also significantly higher in the CAP1s after PNX at the peripheral regions but not at the proximal regions (Fig. 5e), which was confirmed using RNA-ISH/IHC analysis (Fig. 5f and Supplementary Fig. 7d).

Fig. 5. EC subpopulations and gene expression profiles alter in the post-PNX mouse lungs.

a UMAP dimension reduction analysis of five distinct EC clusters from sham and post-PNX mouse lungs (left). Each dot represents a single cell. Percentage of each EC subpopulation at the total, peripheral, or proximal region of sham vs. post-PNX mouse lungs (right). Table showing the p-value of each EC subpopulation in post-PNX mouse lungs compared to sham-operated mouse lungs. b RNA-ISH/IHC images showing Aplnr⁺ CAP1s (RNA-ISH), BrdU staining, and DAPI (IHC) at the peripheral region of sham vs. post-PNX mouse lungs (insert: higher magnification image). Scale bars, 20 μm. Percentage of BrdU⁺ Aplnr⁺ CAP1s in the mouse lungs (n = 7, mean ± s.e.m., *p < 0.05). c Violin plots showing the paxillin expression in the peripheral region (1.5 mm from the edge) of sham vs. post-PNX mouse lung EC subpopulations (red*; increased, black*; decreased). d RNA-ISH/IHC images showing Aplnr⁺ CAP1s (RNA-ISH), paxillin expression, and DAPI (IHC) at the peripheral and proximal regions of C57BL6 mouse lung right lobe 7 days after PNX (insert: higher magnification image). Scale bars, 20 μm. Quantification of paxillin expression (n = 5, mean ± s.e.m., *p < 0.05). e Violin plots showing the expression of Gata2 in the peripheral or proximal region of sham vs. post-PNX mouse lung EC subpopulations (red*; increased, black*; decreased). f RNA-ISH/IHC images showing Aplnr⁺ CAP1s (RNA-ISH), GATA2 expression, and DAPI (IHC) at the peripheral region of sham vs. post-PNX Pxn^fl/fl or Pxn^iΔEC mouse lungs (insert: higher magnification image). Scale bars, 20 μm. Quantification of GATA2 expression (n = 6, mean ± s.e.m., *p < 0.05).

To investigate the role of EC paxillin in the post-PNX gene expression profiles, we also performed scRNA-seq of Pxn^fl/fl or Pxn^iΔEC mouse lungs (sham vs. PNX). Post-PNX increases in the percentage and the number of proliferating ECs were suppressed in the Pxn^iΔEC mouse lungs (Fig. 6a and Supplementary Fig. 7e). Post-PNX increased levels of Gata2, Angpt2, and Tie2 in CAP1s were also lower in Pxn^iΔEC mouse lungs (Fig. 6b). RNA-ISH/IHC analysis confirmed that post-PNX increases in the levels of GATA2, ANGPT2, and TIE2 in the mouse lung Aplnr⁺ CAP1s at the peripheral regions were attenuated in the Pxn^iΔEC mouse lungs (Figs. 5f and 6c).

Fig. 6. EC subpopulations and gene expression profiles in the post-PNX Pxn^fl/fl or Pxn^iΔEC mouse lungs.

a Bar Plots showing the percentage of EC subpopulations in the sham vs. post-PNX Pxn^fl/fl or Pxn^iΔEC mouse lungs. b Dot plots showing average log fold change of Tie2, Tead1, Gata2, and Angpt2 after PNX in the Pxn^fl/fl or Pxn^iΔEC mouse lung EC subpopulations. c RNA-ISH/IHC images showing Aplnr⁺ CAP1s (RNA-ISH), ANGPT2 expression and DAPI (IHC, top), Aplnr⁺ CAP1s (RNA-ISH), TIE2 expression and DAPI (IHC, 2nd), Aplnr⁺ CAP1s (RNA-ISH), TEAD1 expression and DAPI (IHC, 3rd), and Aplnr⁺ CAP1s (RNA-ISH), YAP1 expression and DAPI (IHC, bottom) at the peripheral region of sham vs. post-PNX Pxn^fl/fl or Pxn^iΔEC mouse lungs (insert: higher magnification image). Scale bars, 20 μm. Quantification of ANGPT2, TIE2, TEAD1, and YAP1 expression (n = 5–6, mean ± s.e.m., *p < 0.05). d Circos plots showing the strength and direction of ligand-receptor interactions in the sham vs. post-PNX Pxn^fl/fl or Pxn^iΔEC mouse lung ECs and epithelial cells analyzed using CellChat. e Dot plots showing expression of Yap1, Tead1, and Hippo-related genes in the sham vs. post-PNX mouse lung EC subpopulations.

To further determine the effects of paxillin knockdown on communication between ECs and alveolar epithelial cells after PNX, we conducted a ligand-receptor analysis (CellChat) (Fig. 6d and Supplementary Fig. 8a). The strength of ligand-receptor interaction between epithelial clusters (AT1, AT2, AT1/AT2, ligand) and EC subtypes (proliferating ECs, CAP2, CAP1, arterial, venous, receptor) was increased in the Pxn^fl/fl mouse lungs after PNX; in contrast, these effects were inhibited or attenuated in Pxn^iΔEC mouse lungs (Fig. 6d). The strength of ligand-receptor interaction between epithelial cell clusters (AT1, AT2, AT1/AT2, ligand) and EC subtypes (receptor) was higher in sham-operated Pxn^iΔEC mouse lungs compared to Pxn^fl/fl mice; in contrast, these effects were inhibited or attenuated following PNX (Fig. 6d). Since EC-derived angiocrine factors contribute to alveolar epithelial cell behaviors and lung diseases^3,50–52, we also examined post-PNX changes in the interaction of EC derived ligands with receptors in alveolar epithelial cells and the effects of paxillin knockdown. Ligand-receptor interaction between proliferating ECs (ligand) and AT1 and AT1/AT2 transient epithelial cell cluster, which contributes to lung injury repair⁵³ (receptor), increased in post-PNX Pxn^fl/fl mice, which was decreased in post-PNX Pxn^iΔEC mouse lungs (Fig. 6d). Reciprocal interaction of ECs and alveolar epithelial cells controls post-PNX regenerative lung growth.

We also examined the specific signaling interactions using CellChat analysis (Supplementary Fig. 8b). In the AT1 (ligand) and EC (receptor) interaction: Sema3a-Nrp1/Plxnas signaling interactions were lost, and the levels of Vegfa-Vegfr1, 2 signaling were suppressed in the post-PNX Pxn^iΔEC mouse lungs. In the AT1/AT2 (ligand) and EC (receptor) interaction: multiple angiogenic signaling interactions, including Vegfa-Vegfr1, 2, Thbs-Cd47/Cd36, Sema3s-Nrps/Plxns, Fgf1-Fgfrs, and Efna1-Ephas signaling interactions were lost in post-PNX Pxn^iΔEC mouse lungs. We also examined angiocrine signaling (Supplementary Fig. 8b). In the proliferating EC (ligand) and AT1/AT2 (receptor) interaction: Tnfsf12-Tnfsf12a, Sema4s-Nrp1/Plxns, and Nectin signaling interactions were lost in the post-PNX Pxn^iΔEC mouse lungs. In the proliferating EC (ligand) and AT2 (receptor) interaction, the strength of which was higher in the post-PNX Pxn^iΔEC mouse lungs compared to Pxn^fl/fl mouse lungs (Fig. 6d): Sema3s-Nrps/Plxns signaling interactions were higher in the post-PNX Pxn^iΔEC mouse lungs. Since SEMA3 signaling inhibits angiogenesis, while SEMA4 signaling rather stimulates angiogenesis^54,55, switching the SEMA signaling in the different cluster interactions may be involved in the effects of paxillin knockdown on post-PNX lung vascular regeneration in Pxn^iΔEC mouse lungs.

It has been reported that CAP1s differentiate into CAP2s during lung development^46–48. To examine the involvement of CAP1s in post-PNX lung growth, we performed velocity analysis. Similar to lung development, CAP1 population exhibited directional flow toward CAP2 and arterials after PNX in Pxn^fl/fl mouse lungs, whereas this transition was inhibited in Pxn^iΔEC mouse lungs (Supplementary Fig. 9a). Recently, it has been demonstrated that venous ECs can be regenerative and generate capillary ECs after lung injury (i.e., influenza and hyperoxia injury)⁵¹. However, we detected no velocity-directed transitions from the venous cluster to other EC subclusters after PNX in Pxn^fl/fl nor Pxn^iΔEC mouse lungs (Supplementary Fig. 9a). These results suggest that venous cluster is not the major pathway during post-PNX lung growth; yet, further investigation using lineage tracing analysis will be necessary to determine the contribution of venous ECs to post-PNX lung growth.

TEAD1 mediates paxillin-ANGPT2 signaling in the post-PNX mouse lung ECs

We then examined the mechanism by which paxillin controls GATA2 expression and activity. We have reported that mechanosensitive YAP1-TEAD1 signaling in ECs is necessary for post-PNX lung growth^10,11. Thus, we investigated whether paxillin controls GATA2 activity on transcription of Angpt2 through YAP1-TEAD1 signaling. In our scRNA-seq analysis, Tead1 expression significantly increased in the CAP1s after PNX (Fig. 6e); in contrast, post-PNX increased levels of Tead1 in CAP1s were lower in Pxn^iΔEC mouse lungs (Fig. 6b). In the RNA-ISH/IHC analysis, the levels of TEAD1 expression and nuclear YAP1 were significantly increased in the post-PNX mouse lung Aplnr⁺ CAP1s at the peripheral region, while the effects were inhibited in the Pxn^iΔEC mouse lungs (Fig. 6c). YAP1 activity (nuclear YAP1) was upregulated (Fig. 7a, b) and the levels of TEAD1 were increased in the stretched ECs (Fig. 7c). Paxillin knockdown suppressed the expression of YAP1 and Tead1 in ECs under stretching (Fig. 7b, d). Tead1 knockdown inhibited Gata2 and Angpt2 expression (Fig. 7e, f) and binding of GATA2 to the Angpt2 promoter region in ECs (Fig. 7g). YAP1 S127A active mutant restored Gata2 and Angpt2 expression in ECs isolated from Pxn^iΔEC mouse lungs (Fig. 7h). These results suggest that YAP1-TEAD1 mediates paxillin-GATA2-ANGPT2 signaling in lung ECs.

Fig. 7. TEAD1 mediates paxillin signaling after PNX and under stretching.

a IF micrographs showing YAP1 and DAPI in the mouse lung ECs under stretching. Scale bars, 10 μm. Quantification of nuclear YAP1 under stretching (n = 5, mean ± s.e.m., *p < 0.05). b IF micrographs showing YAP1 and DAPI in L-HMVECs treated with control or paxillin siRNA under stretching. Scale bars, 10 μm. Quantification of YAP1 expression and nuclear YAP1 under stretching (n = 6, mean ± s.e.m., *p < 0.05). c Immunoblots showing TEAD1 and β-actin protein levels (left) and mRNA levels of Tead1 (right) in ECs isolated from mouse lungs under stretching (n = 4, mean ± s.e.m., *p < 0.05). d Tead1 mRNA levels in the Pxn^fl/fl or Pxn^iΔEC mouse lung ECs under stretching (n = 4, mean ± s.e.m., *p < 0.05). e mRNA levels of Tead1 and Gata2 in the mouse lung ECs treated with Tead1 siRNA (n = 5, mean ± s.e.m., *p < 0.05). f mRNA levels of Angpt2 in the mouse lung ECs treated with Tead1 siRNA under stretching (n = 5, mean ± s.e.m., *p < 0.05). g ChIP-PCR analysis showing GATA2 co-IP levels of Angpt2 promoters in ECs treated with Tead1 siRNA (n = 3, mean ± s.e.m., *p < 0.05). h mRNA levels of Gata2 and Angpt2 in Pxn^fl/fl or Pxn^iΔEC mouse lung ECs treated with YAP1S127A mutant (n = 4–5, mean ± s.e.m., *p < 0.05). i IF images showing CD31⁺ blood vessels, IPO7 expression, and DAPI at the peripheral region of Pxn^fl/fl or Pxn^iΔEC mouse lung right lobe 7 days after PNX (insert: higher magnification image). Scale bars, 20 μm. Quantification of IPO7 expression (n = 5, mean ± s.e.m., *p < 0.05). j IF micrographs showing YAP1 and DAPI in L-HMVECs treated with control or IPO7 siRNA under stretching. Scale bars, 10 μm. Quantification of nuclear YAP1 (left; n = 5, mean ± s.e.m., *p < 0.05) and IPO7 mRNA expression (right; n = 4, mean ± s.e.m., *p < 0.05).

Recently, it has been reported that importin7 (IPO7), which transports proteins from the cytoplasm into the cell nucleus, mediates ECM stiffness-dependent YAP nuclear translocation⁵⁶. Therefore, we investigated whether IPO7 mediates stretching-induced YAP1 nuclear localization in ECs. IPO7 expression increased in the post-PNX Pxn^fl/fl mouse lung ECs and stretched L-HMVECs; in contrast, the effects were suppressed in post-PNX Pxn^iΔEC mouse lungs and L-HMVECs treated with paxillin siRNA (Fig. 7i and Supplementary Fig. 9b). siRNA-based IPO7 knockdown inhibited YAP1 nuclear localization in stretched L-HMVECs (Fig. 7j), suggesting that IPO7 mediates stretching-dependent nuclear localization of YAP1.

Post-PNX changes in ECMs control EC regeneration in the gel implanted on the mouse lungs

The ECM impacts cell mechanical responses, which contribute to tissue remodeling²⁶. Thus, we examined the local differential changes in the ECM proteins after PNX that may change the mechanical environment and regulate EC regeneration^{13,14,25,27,57}. We isolated ECMs from peripheral and proximal regions of sham vs. post-PNX mouse lungs (three replicates/each). Proteomics (LC/MS/MS) analysis identified distinct expression patterns between sham vs. post-PNX mouse lungs (Fig. 8a and Supplementary Fig. 10a). Among these up and downregulated genes at the peripheral region after PNX, 26 genes were detected as significantly differentially expressed genes in bulk RNA-seq analysis of post-PNX mouse lung ECs (Hspg2, Cbr2, Col1a1, Col3a1, Col4a1, Col4a3, Col4a4, Col6a1, Col6a2, Cyp4b1, Fbn1, Fom1, Fn1, Nid1, Nid2, Sptbn1, Cltc, Lama3, Lamb1, Lamb3, Lamc1, Myh9, Ehd4, Msn, Myo1c, Cd36). The differentially expressed proteins at the peripheral region in the post-PNX mouse lungs compared to sham-operated lungs were categorized by CC/BP GO terms that were classified as ECM, FA, cytoskeletal remodeling/tissue morphogenesis, and gene regulation/cell signaling, contributing to vascular formation and cell-cell communications (FDR < 0.05, Supplementary Fig. 10b). We also compared the expression of ECM components at the peripheral vs. proximal region of the sham-operated and post-PNX mouse lungs (Supplementary Fig. 10c). While significantly differentially expressed proteins of peripheral vs. proximal region of the sham-operated mouse lungs were more related to ECM (e.g., collagen trimer, laminin complex, basement membrane), those in the post-PNX mouse lungs were more related to cytoskeleton remodeling/tissue morphogenesis and cellular responses (e.g., response to mechanical stimulus, response to amino acid stimulus, blood vessel development) (FDR < 0.05, Supplementary Fig. 10d).

Fig. 8. ECM protein expression profiles in the post-PNX mouse lungs and the effects on EC recruitment in the gel implanted on the mouse lungs.

a Volcano plots of the lung ECM samples isolated from peripheral region of sham (control) vs. post-PNX mice. Blue: down-regulated proteins, Red: up-regulated proteins. Grey: proteins less than a cutoff threshold of Total Spectrum Counts. b mRNA levels of Angpt2 and paxillin in mouse lung ECs cultured on ECMs isolated from peripheral or proximal region of sham vs. post-PNX mouse lungs (n = 4, mean ± s.e.m., *p < 0.05). c EdU⁺ Pxn^fl/fl or Pxn^iΔEC mouse lung ECs cultured on the lung ECMs isolated from proximal or peripheral region of sham vs. post-PNX mice (n = 5, mean ± s.e.m., *p < 0.05). d IF micrographs of COL6 in the proximal vs. peripheral region of sham vs. post-PNX mouse lungs (7 days). Scale bars, 20 μm. Quantification of COL6 expression (n = 5, mean±s.e.m., *p < 0.05). e IF micrographs of CD31⁺ blood vessels, SPB⁺ AT2 cell recruitment, and DAPI (top) and CD31⁺ blood vessels, paxillin expression, and DAPI (bottom) in the fibrin gel supplemented with ECM proteins collected from proximal vs. peripheral region of sham vs. post-PNX mouse lungs or proximal region of sham-operated mouse lungs mixed with COL6 and implanted on the C57BL6 mouse lungs for 7 days. Scale bars, 25 μm. Integrated fluorescent density of CD31, SPB and paxillin (n = 5, mean±s.e.m., *p < 0.05).

When mouse lung ECs were cultured on the ECMs extracted from different regions of sham vs. PNX mouse lungs, the levels of paxillin and Angpt2 and DNA synthesis of mouse lung ECs were significantly higher in ECs cultured on post-PNX mouse lung ECMs compared to those on sham-operated lung ECMs; the levels of gene expression and DNA synthesis were higher when cultured on peripheral region of post-PNX mouse lung ECMs compared to those on ECMs from proximal region (Fig. 8b, c). The effects on DNA synthesis were inhibited when Pxn^iΔEC mouse lung ECs were cultured on the post-PNX mouse lung ECMs (Fig. 8c and Supplementary Fig. 11a). Actin stress fiber formation was also stimulated when ECs were cultured on the peripheral region of post-PNX mouse lung ECMs (Supplementary Fig. 11b).

We have reported that collagen III (COL3) and VI are mechanosensitive ECMs during organ development^27,28. Collagen VI (COL6) controls the phosphorylation of paxillin⁵⁸. In our proteomics data, COL3 and 6 significantly increased in the peripheral regions after PNX (Fig. 8a). IHC analysis confirmed that COL6 was significantly increased at the peripheral region of the lungs after PNX (Fig. 8d). In contrast, COL3 increased both at proximal and peripheral regions (Supplementary Fig. 11c).

To confirm the effects of post-PNX mouse lung ECMs on EC regeneration in the lungs, we implanted fibrin gel supplemented with ECM proteins (3 mg/gel) extracted from the proximal vs. peripheral region of decellularized sham-operated or post-PNX mouse lungs. Recruitment of CD31⁺ ECs and SPB⁺ AT2 cells from the host lungs and expression of paxillin and ANGPT2 were significantly stimulated in the gel containing ECM proteins extracted from the peripheral region of the post-PNX mouse lungs compared to that supplemented with ECM proteins from sham-operated mouse lungs or those from the proximal regions (Fig. 8e and Supplementary Fig. 11d). We also implanted the ECM gel collected from the proximal region of the sham-operated mouse lungs in combination with supplementation of COL6. Supplementation of COL6 stimulated EC and AT2 cell recruitment from the host lung and increased paxillin expression (Fig. 8e).

Discussion

Here, we have demonstrated that EC regeneration is stimulated at the peripheral region of the mouse lung after PNX, where ECM components are differentially expressed to change response to mechanical tension through paxillin signaling in ECs (Supplementary Fig. 12). Cellular stretching increased the expression of paxillin, which induced GATA2 activity on Angpt2 transcription by increasing YAP1-TEAD1 signaling in ECs. Proliferation of ECs significantly increased in the post-PNX mouse lungs, and paxillin, GATA2, ANGPT2, and TEAD1 expression was upregulated in CAP1s. There are a number of novelties in this study: (1) a new mechanism by which paxillin spatially controls EC proliferation during post-PNX lung growth through GATA2-ANGPT2-TIE2 signaling; (2) scRNA-seq analysis of spatial changes in the gene expression profiles of sham vs. PNX mouse lungs and identification of specific EC clusters; (3) previously unknown spatial changes in the ECM protein profiles altered in the post-PNX mouse lungs and the role of specific ECM (e.g., COL6) in EC regeneration; (4) involvement of YAP1-TEAD1 in a new mechanosensitive mechanism of paxillin-GATA2-ANGPT2 signaling in post-PNX lung ECs.

We targeted YAP1-TEAD1 signaling because (1) YAP1 controls ANGPT2¹² and TIE2 expression^10,11 to induce angiogenesis; (2) we have reported that YAP1 senses post-PNX increases in pulmonary artery (PA) pressure, stimulates TIE2 expression, and mediates post-PNX lung growth¹¹; (3) Paxillin and other FA protein FAK signaling are known to control YAP activity in cancer cells and mesenchymal stem cells^59–61. In our ChIP-PCR results, Tead1 knockdown inhibited GATA2 binding to Angpt2 promoter region (Fig. 7g) and it has been reported that TEAD1 binds to GATA2 and controls GATA2 transcriptional activity⁶², suggesting that YAP1-TEAD1 signaling controls GATA2 activity on transcription of Angpt2. It has been reported that IPO7 mediates ECM stiffness-dependent YAP nuclear translocation⁵⁶. In our results, IPO7 expression increased in the post-PNX or stretched ECs, which was suppressed by paxillin knockdown (Fig. 7i and Supplementary Fig. 9b). IPO7 knockdown inhibited YAP1 nuclear localization in stretched L-HMVECs (Fig. 7j), suggesting that IPO7 mediates stretching-dependent nuclear localization of YAP1. YAP1 also controls paxillin expression and activity in cancer cells⁶³ and interacts with other Hippo molecules, transcriptional co-activator with PDZ-binding motif (TAZ). Other mechanosensitive molecules may also be involved in the mechanism. Wnt co-receptor LRP5 senses ECM stiffness and controls TIE2 expression and modulates lung development²⁵, while non-canonical Wnt signaling controls paxillin ubiquitination during Xenopus laevis gastrulation⁶⁴. Another mechanosensitive transcription factor TWIST1 that interacts with another FA protein zyxin⁶⁵, increases in the post-PNX lung ECs⁶⁶, which may also mediate stretching dependent effects after PNX. Paxillin regulates expression/activity of other angiogenic genes such as neuropilin2³⁹, and induces PASMC proliferation by increasing PDGFB expression⁶⁷. Thus, paxillin may interact with other mechanosensitive and angiogenic genes^34,35 and convey micromechanical signals to control lung EC regeneration.

We used Pxn^iΔEC mice, in which paxillin expression was knocked down in ECs. Although the levels of paxillin in Pxn^iΔEC mouse lung ECs were significantly lower compared to Pxn^fl/fl mouse lung ECs in sham-operated lungs (p = 0.0439, Fig. 3b; p < 0.0001, Supplementary Fig. 3a), the knockdown levels of paxillin were not dynamic (Fig. 3b and Supplementary Fig. 3a). This may be because of the tamoxifen dose, timing and duration; we chose lower dose tamoxifen treatment (125 μg/mouse, 5 days)^11,40,68, because it has been reported that high dose tamoxifen treatment has adverse effects on angiogenesis in neonatal Cdh5(PAC)-Cre^ERT2 mouse retina⁶⁹. Even with the lower level of paxillin knockdown, post-PNX lung vascular regeneration was significantly inhibited in our Pxn^iΔEC mouse line, suggesting the indispensable role of endothelial paxillin in vascular regeneration and post-PNX lung growth.

Our proteomics data have revealed that ECM components and their functions are distinct between proximal and peripheral regions of the post-PNX lungs, which may contribute to the spatial changes in the cellular responses to stretching during post-PNX lung growth; differences in the increased/decreased levels and patterns of ECMs between peripheral and proximal region of the lungs after PNX may accelerate the response to stretching, and increase the expression of paxillin and angiogenic factors in ECs, leading to EC proliferation more at the peripheral regions of the lung. It is important to note that proteins altered at the peripheral regions after PNX were more involved in the tissue morphogenesis and cellular responses rather than only maintaining ECM structure (Supplementary Fig. 10d), inducing EC regeneration. Peripheral regions of the lung also have special chemical and mechanical properties: (1) contain less septal tissues⁷⁰, (2) associate with systemic (bronchial) and pulmonary vessels⁷¹, and (3) contain abundant stem cells⁷². It has been reported that differentiation of AT2 cells to AT1 cells is more prominent at the peripheral region of the lung after PNX and during injury repair⁵³, indirectly altering angiogenic factor expression and EC proliferation. While other mechanical forces, such as PA pressure, also increase after PNX¹¹, the increased levels are less at the capillary levels. Thus, paxillin may sense not only stretching forces but also other types of mechanical forces (e.g., pressure, flow, stiffness) after PNX in a spatiotemporal manner and control EC regeneration.

There is a significant heterogeneity of EC subpopulations in the lung^46–48, which may differentially sense different types of mechanical forces in a spatiotemporal manner during post-PNX lung growth. While post-PNX ERG⁺ EC proliferation was stimulated more at the peripheral region, proliferation of CAP1 marker⁺ proliferating ECs increased at both proximal and peripheral regions after PNX (Figs. 1b, 5a and Supplementary Figs. 1b, 2a and 6b, c). This may be because of the small cell number of the cluster in the scRNA-seq analysis. Given that ANGPT2 and TIE2 increased in Aplnr⁺ CAP1s at the peripheral regions after PNX (Fig. 6c and Supplementary Fig. 7c), ANGPT2 binds to TIE2 in CAP1s to stimulate post-PNX EC regeneration. GATA2 and ANGPT2 co-expressed in stretched ECs and at the peripheral region after PNX, but not all ECs nor at the proximal region (Fig. 2a, c and Supplementary Fig. 2g), suggesting that sensitivity to the stretching forces may be different between GATA2 expression and its transcriptional activity or GATA2 controls ANGPT2 expression in specific EC subpopulations, such as CAP1s, at the peripheral region after PNX (Fig. 5e, f, and Supplementary Fig. 7b, c). Paxillin and Gata2 expression was significantly higher in post-PNX CAP1 populations compared to sham operated mouse lungs (Fig. 5c, e). Although expression of paxillin that induces Gata2 expression was also higher in post-PNX proliferating EC populations, it was not statistically significant (p = 0.22684). This may be due to the relatively small number of cells in the cluster. We have reported that ECM stiffness controls GATA2 nuclear localization but not expression in human dermal ECs¹⁴; however, expression of GATA2 increased in stretched lung ECs and after PNX (Figs. 2a–c, e and 5e, f). This inconsistency may be because of the differences in the mechanical forces (stiffness vs. stretching) or the tissue-specific differences (dermal vs. lung ECs).

Regional differences in EC response may be secondary to other lung cells (e.g., epithelial cells, smooth muscle cells, immune cells), which may crosstalk with ECs and change EC behaviors during regenerative lung growth. These other lung cells also sense changes in the micromechanical environment after PNX, regulate paxillin expression to control cell proliferation and differentiation as well as to secrete angiogenic and other chemical factors, and indirectly control EC regeneration after PNX in a spatiotemporal manner. In fact, in the ligand-receptor analysis (CellChat), the strength of ligand-receptor interaction between epithelial clusters (AT1, AT2, AT1/AT2, ligand) and EC subtypes (proliferating ECs, CAP2, CAP1, arterial, venous, receptor) was increased in the Pxn^fl/fl mouse lungs after PNX; in contrast, these effects were inhibited or attenuated in Pxn^iΔEC mouse lungs (Fig. 6d). It has been reported that AT1 cells produce one of the major angiogenic factors VEGF⁴⁸, which binds to VEGFRs in ECs to control angiogenesis. AT2 cells also secrete angiogenic factors, which stimulates blood vessel formation and nutrient transport^3,73. Thus, ligand-receptor interaction between alveolar epithelial cells and ECs controls EC regeneration mediated by paxillin. On the contrary, paxillin in ECs controls AT2 cell behaviors (e.g., DNA synthesis, surfactant production, Supplementary Fig. 3d, e). It has been demonstrated that EC-derived angiocrine factors contribute to alveolar epithelial cell behaviors and lung diseases^3,50–52. Our scRNA-seq analysis demonstrated that strength of ligand-receptor interaction between proliferating ECs (ligand) and AT1 and AT1/AT2 transient epithelial cell cluster, which contributes to lung injury repair⁵³ (receptor) increased in post-PNX Pxn^fl/fl mice, which was decreased in post-PNX Pxn^iΔEC mouse lungs (Fig. 6d) and Sema4s-Nrp1/Plxns signaling interactions that stimulate angiogenesis were lost in post-PNX Pxn^iΔEC mouse lungs (Supplementary Fig. 8b). On the contrary, the strength of proliferating EC (ligand) and AT2 (receptor) interaction was higher in post-PNX Pxn^iΔEC mouse lungs compared to Pxn^fl/fl mouse lungs (Fig. 6d), in which Sema3s-Nrps/Plxns signaling interactions that inhibit angiogenesis, were higher in the post-PNX Pxn^iΔEC mouse lungs (Supplementary Fig. 8b). Reciprocal effects of paxillin in ECs and alveolar epithelial cells may lead to spatiotemporal control of post-PNX regenerative lung growth.

Inhibition of lung regeneration/repair contributes to lung diseases^74,75. GO terms categories in the CAP1s after PNX seem to be different from those in lung injury models (LPS, hyperoxia^76,77) but partly overlapped (Supplementary Fig. 13). In post-PNX lung, genes related to cell cycle regulation and gene regulation were differentially expressed; in LPS-induced inflammatory lung injury, genes related to cell migration/tissue morphogenesis and gene regulation were differentially expressed; in hyperoxia-induced lung injury, genes related to metabolism and cell adhesion/migration were differentially expressed. Further analysis will identify specific EC populations and genes altered in lung injury, leading to the development of new therapeutic targets for lung diseases.

In summary, we have demonstrated that paxillin signaling mediates spatial control of post-PNX EC regeneration. ECM proteins spatially differentially expressed in the post-PNX mouse lungs change the response to mechanical stretching and drive EC proliferation after PNX at the peripheral region of the lungs by changing paxillin expression. Modulation of mechanical environment or paxillin signaling may improve the way for lung EC regeneration and potentially lead to the development of new strategies for lung diseases.

Methods

Materials

Anti-CD31 (#553370), -paxillin (#610619), and -TEAD1 (610922) antibodies were from BD Biosciences (San Jose, CA). Anti–surfactant protein-B (SPB, ab40876), -ERG (ab92513), -COL3 (ab7778), and -COL6 (ab6588) antibodies were from Abcam (Cambridge, MA). Anti-β-actin monoclonal antibody (A5316) was from Sigma (St. Louis, MO). Anti-TIE2 (SC324), -GATA2 (sc-515178X), and -YAP1 (101199) antibodies were from Santa Cruz Biotechnology (Dallas, TX). Anti-ANGPT2 (MAB098-100) antibody was from R&D Systems (Minneapolis, MN). Anti-Ki67 (14-5698-82), -GATA2 (PA1-100) antibodies were from ThermoFisher (Waltham, MA). Anti-phospho (Tyr 118) paxillin antibody (E9U9F) was from Cell Signaling (69363, Danvers, MA). Anti-IPO7 (28289) antibody was from Proteintech (Rosemont, IL). Human lung microvascular ECs (L-HMVECs) and human umbilical vascular ECs (HUVECs, Lonza) were cultured in ECM medium containing 5% FBS and growth factors (VEGF, bFGF, and PDGF, ScienCell, Carlsbad, CA).

Animals

Pxn^fl/fl mice (obtained from Dr. Christopher Turner, SUNY Upstate Medical University, Jackson Laboratory, stock # 35946)^33,36 were crossed with Cdh5(PAC)-Cre^ERT2 mice (obtained from Dr. Ralf Adams, Max Planck Institute)⁴³, an inducible Cre deleter under the control of Cdh promoter, to create mice with Cdh5-Cre-dependent conditional deletion of paxillin (Pxn^fl/fl-Cdh5(PAC)-Cre^ERT2 (Pxn^iΔEC)), in which Cre recombination is induced in ECs by administration of tamoxifen^10,11,40,68.

Molecular biological, and biochemical methods

Quantitative reverse transcription (qRT)-PCR was performed with the iScript reverse transcription and iTaq SYBR Green qPCR kit (BioRad, Hercules, CA) using the BioRad real time PCR system (BioRad). β2 microglobulin, cyclophilin, or Gapdh controlled for overall cDNA content. The primers used for human GATA2, human β2 microglobulin, human TIE2, mouse Tie2, mouse cyclophilin, and Gapdh were previously described^10,14. The primers used for mouse Angpt2 forward; TTAGCACAAAGGATTCGGACAAT, reverse; TTTTGTGGGTAGTACTGTCCATTCA, human ANGPT2 forward; CGCCGCTCGAATACGATGA, reverse; CGTCTGGTTCTGTACTGCATT, human paxillin forward; CTGGCGGACTTGGAGTCTAC, reverse; CTCCTCCGACAAGAACACAGG, mouse paxillin forward; GGCATCCCAGAAAATAACACTCC, reverse; GCCCTGCATCTTGAAATCTGA, mouse Tead1 forward; GAGCGACTCGGCAGATAAGC, reverse; CCACACGGCGGATAGATAGC, mouse Gata2 forward; GTCACTCTCCGAGAGCATGA, reverse; GCCTTCTGAGCAGGAGCGAG, human IPO7 forward; CCCCAACACCATTATCGAGGC, reverse; AGAGACTTGTGTGCTTCATTGAG. For ChIP assay, DNA from L-HMVECs or mouse lung CD31⁺ cells applied to uniaxial cyclic strain (0.5 Hz, 5%) for 16 hrs¹⁶ was immunoprecipitated with GATA2 antibody (Santa Cruze Biotechnology; sc-515178 X) or control immunoglobulin according to the manufacturer’s instructions (Thermo Fisher Scientific)^14,78. The GATA-binding region of mouse Angpt2 was amplified using primers, 5’-GGGGCGGAGCGGCTGGCTG-3’ and 5’-AATCACTTTCCTTCCCTTATATG-3’. The GATA-binding region of human ANGPT2 was amplified using primers, 5’-GGCTAGTGACCCCCTACAG-3’ and 5’-ATCCGAATCAATCACTTTCCTTTCC-3’. Protein levels of mouse (MANG20) and human ANGPT2 (DANG20) were measured using R&D Quantikine ELISA kit (R&D Systems, Minneapolis, MN). Protein levels of mouse surfactant protein C (MBS265452) were measured using ELISA kit from MyBioSource (San Diego, CA).

Gene knockdown and overexpression

Gene knockdown was performed using the RNA interference technique. siRNAs for human TEAD1 (5’-AUGGCCGAUUUGUAUACCGAA-3’ and 5’- UUCGGUAUACAAAUCGGCCAU-3’), mouse Tead1 (5’-CAAGACGUCAAGCCCUUUGUG-3’ and 5’- CACAAAGGGCUUGACGUCUUG-3’), and human IPO7 (5’- GGAAUCUGCUUACAGGUCA-3’ and 5’- UGACCUGUAAGCAGAUUCC-3’) were purchased from Sigma Genosys (St. Louis, MO). For mouse Gata2 knockdown, ON-TARGETplus mouse Gata2 (14461) siRNA SMARTpool was used (Horizon Discovery, Waterbeach, UK). siRNAs for human and mouse paxillin, as well as human GATA2, were reported previously^14,39. Cells were transfected using siLentFect (BioRad, Hercules, CA)^10,14. As a control, siRNA duplex with an irrelevant sequence (QIAGEN, Germantown, MD) was used.

The retroviral full-length pQCXIH-myc-YAP1 (human) and pQCXIHflag-YAP1-S127A (human) were gifts from Kunliang Guan (Addgene plasmid # 33091 and # 33092)⁷⁹ and were reported previously¹⁰. Generation of retroviral vectors was accomplished as reported¹⁴. Viral supernatants were collected starting 48 h after transfection for four consecutive times every 12 h, pooled, and filtered through a 0.45 μm filter. Viral supernatants were then concentrated 100-fold by ultracentrifugation in a Beckman centrifuge for 1.5 h at 16,500 rpm. Cells were incubated with viral stocks in the presence of 5 μg/ml polybrene (Sigma) for 3 days in the BSL2 facility, and transduction efficiency was confirmed using immunocytochemistry¹⁴. If the efficiency was <80–90%, second round of transduction was performed, which achieved 80–90% transduction efficiency¹⁰. pGL3-human ANGPT2 (-405- +65)⁴² constructs were generated by Genscript (Piscataway, NJ), transfected to HUVECs, and luciferase assay was performed as we reported¹⁴.

Mouse lung cell isolation

Mouse lung ECs were isolated from mouse lungs using anti-CD31 conjugated magnetic beads as described previously^10,11,78. Mouse lung epithelial cells were isolated from mouse lungs using anti-EPCAM conjugated magnetic beads as previously reported with modifications⁸⁰. Briefly, we cut lung tissue into small pieces using small scissors and treated the tissue with collagenase A (5 ml, 1 mg/ml) for 30 min at 37 °C. The tissue suspension was filtered through a 40 μm cell strainer (Corning, Glendale, AZ) to remove the undigested cell clumps and separate single cells. Cells were centrifuged (1000 rpm, 5 min) at room temperature (RT) and the pellet was resuspended into 0.5 ml RBC Lysis Buffer (Sigma, 1 min, RT). The lysis reaction was stopped by adding 10 ml 10% FBS/DMEM, centrifuged (1000 rpm, 5 min, RT), and the pellet was resuspended into 0.5 ml 4% FBS/PBS with APC-conjugated anti-mouse EPCAM (1/100, Biolegend, San Diego, CA), incubated (20 min, on ice), and washed three times with 4% FBS/PBS. Cells were centrifuged (1000 rpm, 5 min, RT) and resuspended into 0.1 ml 4% FBS/PBS with anti-APC microbeads (Miltenyi Biotec, Somerville, MA), incubated (10 min, on ice), and washed three times with 4% FBS/PBS. The cells were then resuspended in 0.5 ml 4% FBS/PBS, and EPCAM⁺ lung epithelial cells were magnetically separated using MACS column (Miltenyi Biotec) according to the manufacturer’s instruction. To increase the purity of the magnetically separated fraction, the eluted fraction was enriched over a second new MACS column. Using this method, we obtained at least 5 × 10⁵ cells/mouse. FACS analysis confirmed that 85–96% of the FACS-sorted isolated cells express EC markers (CD31, Supplementary Fig. 2c)^10,11,78 and epithelial markers (EPCAM, Supplementary Fig. 3d). Paxillin mRNA expression was lower in CD31⁺ ECs isolated from Pxn^iΔEC mouse lungs compared with those from control Pxn^fl/fl mouse lungs under tamoxifen induction (125 μg/mouse, 5 days, Supplementary Fig. 3a and Fig. 3b). We used freshly isolated lung cells in this study. The proliferation of ECs in the mouse lungs after PNX or in combination with prosthesis insertion was analyzed by measuring the number of BrdU⁺ cells using FACS (BD Biosciences BrdU flow kit)¹⁰ or Ki67 staining¹¹.

Mechanical strain application

L-HMVECs or mouse lung CD31⁺ cells (1 × 10⁵) were seeded on the stretch chamber made of poly-dimethylsiloxane (As One International, Santa Clara, CA) coated with 0.5% gelatin 24 h before stretching. The chambers were then applied to uniaxial cyclic strain (0.5 Hz, 10%) for 16 h¹⁶. After stretching, total RNA and cell lysate were collected for subsequent qRT-PCR and biochemical analysis, respectively.

Unilateral PNX

The in vivo animal study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was reviewed and approved by the Animal Care and Use Committee of Medical College of Wisconsin (AUA5598). The mice were euthanized using CO₂ inhalation followed by cervical dislocation, which was approved by the protocol. C57BL6 (Jackson Laboratory, stock # 664), Pxn^fl/fl, and Pxn^iΔEC male mice were used for the study. Unilateral PNX was performed as described in refs. ^9–11,66. In brief, mice (8–12 week old, approximately 25 g BW) were anesthetized with Ketamine/Xylazine (IP), and intubated with a 21-gauge cannula and mechanically ventilated at 120 cycle/min with a tidal volume of 10 ml/kg using a rodent ventilator (MiniVent, Harvard Apparatus, Holliston, MA). After ensuring adequate anesthesia, a 1 cm incision was made through the skin, muscles above the left lung, along the intercostal space between the fourth and fifth ribs were cut, and thoracotomy was performed. A small retractor was placed to provide access to the thoracic cavity. The left lung was gently lifted through the incision, and a 5–0 silk suture was passed around the hilum and tied. The hilum was then transected distal to the tie. The remaining portions of the hilum and tie were returned to the thoracic cavity. To study the effects of mechanical tension after PNX, we inserted silicone prosthesis to replace the dissected lobe and fill the dead space. The mouse was extubated and observed for return of spontaneous respirations. Sham operated mice underwent tracheal intubation and thoracotomy without PNX. The weight of the right lobes was measured and normalized to body weight (BW). The peripheral regions of each lobe (1.5 mm from the edge) were horizontally separated using small scissors under the dissection microscope. Histological samples were prepared as previously reported, and morphological analysis of vascular structures was performed as described in refs. ^9–11,66. Fluorescent images were taken on a Nikon A1 confocal laser scanning microscope, and morphometric analysis was performed using ImageJ software^9–11,66. Integrated density (sum of the values of the pixels in each image) of the florescence signal was measured using ImageJ^9–11,66. The same imaging settings were used to acquire and analyze the images for comparison. Detection of oversaturation was done by using pixel saturation indicators.

Transmission Electron Microscopy (TEM)

Mouse lungs were collected and TEM samples were prepared as previously described⁴⁰. Freshly excised mouse lung pieces were immersed in a mixture 1-part 2.5% glut in 0.1 M cacodylate + 2-parts OsO₄ in 0.1 M sodium cacodylate buffer—mixed just before use for 45 min on ice. Tissues were then washed with distilled water 3 × 10 min and immersed in 0.25% aqueous uranyl acetate en bloc for 2 h. Tissues were again washed 3 × 10 min with distilled water, then dehydrated through graded methanol, acetonitrile washes, and infiltrated with EMBed 812 epoxy resin (14120; Electron Microscopy Sciences, Hatfield, PA). Ultrathin sections (70–80 nm) were cut with an Ultra 45° diamond knife (DiATOME, Hatfield, PA), collected onto copper grids (G200H-Cu; Electron Microscopy Sciences), and stained with uranyl acetate and lead citrate for contrast. Images were captured using a JEOL 1400 Flash transmission electron microscope (JEOL, Tokyo, Japan) equipped with a 2 K × 2 K ultrahigh resolution digital camera.

Bulk RNA sequencing data analysis of post-PNX mouse lung ECs

We reanalyzed our published bulk RNA-seq results available in NCBI Geo (GSE179227)⁴⁵. Briefly, ECs were isolated from C57BL6J mouse lungs 7 days after PNX and sham-operated mouse lungs (8-week-old, n = 3 per group, each n was pooled from 3 male mice) as described above. We used only male mice for bulk RNA-seq analysis to avoid hormonal effects. RNA was extracted using RNeasy mini kit (QIAGEN). Total RNA samples were submitted to the Institute for Systems Biology Molecular and Cell Core (Seattle, WA) for RNA sequencing. Library preparation was employed using the Illumina TruSeq Stranded mRNA kit. Sequencing was performed using the Illumina NextSeq500. Paired end sequencing was performed on a high output 150 cycle kit v2.5. The RNA sequencing reads were aligned to the mm10 reference genome. Differential gene expression analysis was performed using the raw count data and the DEseq2 pipeline to compare gene expression levels between ECs isolated from post-PNX mouse lungs (7 days) and sham-operated control lungs. Genes were considered significantly differentially expressed if absolute log2 fold changes are greater than 1 and false discovery rate (FDR, corrected by Benjamini–Hochberg procedure) is less than 0.05. As a result, we identified 1156 significantly differentially expressed genes; among these, 923 were upregulated, and 233 were downregulated. Pathway analysis was performed using over-representation analysis, in which 1156 significantly differentially expressed genes were tested for enrichment in GO-BP using Fisher’s exact test. Significantly enriched pathways were identified using an FDR cutoff of 0.05.

scRNA-seq and data analysis

After dissecting lungs (whole lungs, peripheral region (<1.5 mm from the edge), proximal region (>1.5 mm from the edge)) from sham-operated or post-PNX male mice (n = 4 per condition), lung cells were isolated as described above (Supplementary Fig. 5a). The cells (10⁶ cells/ml in 2% FBS/PBS) were loaded to 10X Chromium system (10X Genomics, San Francisco, CA) and run through the library preparation protocol following guidance from the Chromium Single Cell 3’ Reagent Kits v3.1 rev B (Dual index). The cDNA and libraries were quantified using High Sensitivity D5000 and D1000 TapeStation kits (Agilent Technologies, Santa Clara, CA), respectively. The libraries were sequenced via Illumina NovaSeq 6000 (Novogene, Sacramento, CA).

The scRNA-seq raw data from each sample were demultiplexed, aligned to mm10 mouse reference genome, and gene-level unique molecular identifier (UMI) count matrices were extracted and quantified using the 10X Genomics Cell Ranger pipeline (v 6.1.2, 10X Genomics). Further analysis with the filtered barcode matrix files was performed using the Seurat package (v4.0)⁸¹ in R (v4.1.0 or above). For initial quality control, cells expressing less than 200 genes or more than 7000–10,000 genes, as well as cells with more than 10–20% mitochondrial genome content, were removed. The downstream analysis on sham vs. post-PNX samples was performed following a data integration vignette on Seurat. The clustering of cells was done using the ‘FindClusters’ function in Seurat, which groups cells based on the Euclidean distance in the PCA and a non-linear dimensional reduction. The cluster identification was achieved by identifying the canonical cell type markers using the “FindConservedMarkers” in Seurat and comparing them to previously identified lung cell markers (Mouse Cell Atlas, LungMAP Consortium, Lung Endothelial Cell Atlas)^47,49,82 (Supplementary Figs. 5b and 6a). EC sub-clusters were identified using markers; CAP1s (Aplnr, Gpihbp1, Epas1), proliferating cells (Mki67, Top2a, Cdk1), CAP2s (Apln, Ednrb, Kdr), arterials (Hey1, Gja4, Sox17), and venous (Nr2f2, Ackr3, Selp)^46–49 (Supplementary Fig. 6a). EC clusters were extracted from the whole lung dataset to create new objects for further analysis using the “subset” function in Seurat. The data was presented in the form of Uniform Manifold Approximation and Projections (UMAPs), dot plots, violin plots, and circos plots. Four mice were pooled into one per each condition. For more thorough statistical analysis of the population changes following PNX, we used the dataset from whole lung, peripheral region of the lungs, and proximal region of the lungs (4 mice × 3 samples/condition) and calculated statistical significance of the percentage of each cluster in combined datasets using T-test.

Differentially expressed genes were also identified comparing CAP1s (non-proliferating CAP1+proliferating ECs) from sham vs. post-PNX mouse lungs and CAP1s from sham vs. post-PNX Pxn^fl/fl vs. Pxn^iΔEC mouse lungs using model-based analysis of single-cell transcriptomics (MAST), which calculates the FDR using the Benjamini-Hochberg analysis and presented as an adjusted p-value⁸³. Although proliferating EC clusters included CAP1s and other EC types, such as Arterial, Venous, and CAP2 (red boxes), these EC subclusters expressed CAP1 markers during post-PNX lung growth (blue boxes), which potentially differentiates into other EC subtypes as previously reported^46–48 (Supplementary Fig. 6d, e). Thus, we combined the proliferating progenitor like CAP1 marker expressing EC cluster (proliferating ECs) and non-proliferating CAP1 cluster as CAP1 to analyze the gene expression profiles. The Seurat objects were normalized using the “LogNormalize” transformation to account for differences in sequencing depth between cells and the “ScaleData” function was implemented to make the expression values comparable across genes within each cell. The average log fold changes of Tie2, Tead1, Gata2, and Angpt2 were also extracted from the differentially expressed gene list and visualized in dot plots using ggplot2 R package (Fig. 6b).

CellChat package (v2.2.0) was used to predict and visualize ligand-receptor interactions among the ECs and epithelial (AT1, AT2, AT1/AT2) clusters from sham vs. post-PNX Pxn^fl/fl and Pxn^iΔEC mouse lungs using the mouse CellChatDB database. The probability of ligand–receptor interactions was computed based on gene expression levels in each cell group, and interaction strength was estimated using the 10% truncated mean method implemented in CellChat. Predicted interactions were visualized using Circos plots to illustrate the number and directionality (arrows) of ligand-receptor signaling between cell populations. Signaling pathways enriched in these ligand–receptor interactions were further compared between post-PNX Pxn^fl/fl and Pxn^iΔEC mouse lungs. Differential interaction networks were visualized using the function netVisual_diffInteraction with weight.scale = TRUE and a common edge.weight.max to ensure that edge widths were comparable across different comparisons.

Developmental trajectories were inferred by R package Monocle 3 (v1.3.7)⁸⁴, using reversed graph embedding framework to learn a principal graph from transcriptional similarities. Cells were then ordered along the learned trajectory to derive pseudotime. RNA velocity was inferred from spliced/unspliced counts with the software Velocyto (v0.17)⁸⁵ and Python package scVelo (v0.3.3)⁸⁵. RNA velocity vectors were projected onto the UMAP plot for visualization. ScRNA-seq raw data are available in NCBI Geo (GSE230548).

Differentially expressed genes were identified comparing CAP1s from sham vs. post-PNX mouse lungs from our data, control vs. endotoxin LPS induced injured mouse lungs (3 days, GSE148499), and normoxia (21% O₂) vs. hyperoxia (85% O₂) treated mouse lungs (GSE151974) from publicly available database^76,77 using MAST⁸³. GO analysis of these genes was performed via DAVID v 6.8 using the Functional Annotation Chart tool, and top 20 GO Terms were displayed and categorized.

ECM extraction and proteomics analysis

Decellularization of the lung tissues was performed as previously reported⁸⁶. Briefly, mouse lungs (three replicates/each) were collected 7 days after PNX and dissected into peripheral (<1.5 mm from the edge) and proximal (>1.5 mm from the edge) regions. The tissue was rinsed with 1X PBS for 30 min, washed with water, and mechanically homogenized in water until broken up. The homogenate was centrifuged (2500 × g, 5 min), and floating fat was removed. After repeating this process, the pellet was resuspended into 2.5 mM sodium deoxycholate/PBS and incubated for 2 h at room temperature (RT) with homogenizing the tissue with 18 G needle every 30 min. The homogenate was then strained over a sieve; all collected material was placed back into fresh 2.5 mM sodium deoxycholate/PBS and incubated for an additional 15 h (RT). The ECM was strained again and washed in 1X PBS for three times. The ECM was then extracted with SDT solution (4% SDS/0.1 M Tris-HCL pH 7.5/0.1 M DTT) by heating at 95 °C for 7 min. ECM solution was then sonicated and centrifuged (15,000 × g, 5 min). The extracted ECM was submitted to Washington University Mass Spectrometry Technology Access Center for proteomics (LC/MS/MS) analysis. Ten microliters of each sample (10 μg/μl) was loaded into the SDS-PAGE gel, and samples were purified following the “stacking gel” method. The gel bands were excised, and proteins were in-gel digested with trypsin using the optimized protocol. Digested peptides were desalted on C18 columns and then subjected to mass spec analysis. Data was searched against a Mus musculus database. Proteomics data analysis on three control (sham) and three PNX C57BL6 mouse lung ECM proteins from peripheral and proximal regions was performed using Scaffold 5.1.0 software. Total 1126 proteins were identified in the control and PNX sample replicates. A cutoff threshold of greater than or equal to eight Total Spectrum Counts was used, and four differential gene expression analyses using Fisher’s test were performed: (1) Sham: peripheral vs. proximal, (2) PNX: peripheral vs. proximal, (3) Peripheral: sham vs. PNX, and (4) Proximal: sham vs. PNX. Significantly differentially expressed proteins were defined as having a p-value of <0.05. These proteins underwent Cellular Component (CC) and Biological Processes (BP) GO analysis via the Functional Annotation Chart feature of the DAVID v6.8 software. The GO Terms were color-coded into groups such as ECM, FA, cytoskeleton remodeling/tissue morphogenesis, and gene regulation/cell signaling. Proteomics results are available via ProteomeXchange with identifier PXD038600.

RNA in situ hybridization (RNA-ISH)

RNAscope technology (Advanced Cell Diagnostics (ACD), Newark, CA) was used for RNA-ISH. The formalin-fixed paraffin-embedded (FFPE) mouse lung tissue samples were sectioned at 5 μm thick using a microtome, mounted on slides, baked for 1 h at 60 °C, and deparaffinized in xylene and 100% ethanol. For target retrieval, hydrogen peroxide (ACD 322381) was applied for 10 min at room temperature, followed by incubating at 98–102 °C for 15 min in 1x target retrieval reagent buffer (ACD 322001). Sections were then treated with Protease Plus (ACD 322381) at 40 °C for 30 min in hybridization oven. In situ detection of Aplnr was performed by hybridization with target probes (Mm-Aplnr (ACD 436171)) following the ACD instruction with RNAscope Multiplex Fluorescent Reagent Kit v2. For dual RNA-ISH/IHC, FFPE sections were first subjected to RNA-ISH, and followed by IHC, such as blocking (1% fetal bovine serum in TBS) and incubation with primary and secondary antibodies. Images of the stained slides were taken on a Nikon A1 confocal laser scanning microscope, and morphometric analysis was performed using ImageJ software as we reported^9–11,66.

Fibrin gel implantation on the mouse lung

Fibrin gel was fabricated as described^10,11,78,87. Briefly, we added thrombin (2.5 U/ml) to the fibrinogen solution (12.5 mg/ml), mixed well, supplemented the gel with ECM proteins (3 mg/gel) extracted from proximal or peripheral region of sham vs. post-PNX mouse lungs or with COL6 (2 μg/gel), incubated the mixture at 37 °C for 30 min until they solidified, and implanted on C57BL6 (8–12 week old) mouse lungs. For gel implantation on the mouse lungs, mice were mechanically ventilated, and thoracotomy was performed in the fifth left intercostal space^10,11,78,87. After thoracotomy, a small area of the left visceral pleura (0.5 mm²) was scraped using forceps, and the fabricated fibrin gel was implanted on the mouse lung surface using fibrin glue. For histological analysis, gels were fixed in 4% PFA overnight at 4 °C, followed by OCT embedding and cryosectioning. Fluorescent images were taken on a Nikon A1 confocal imaging system. Host mouse lung-derived ECs and SPB⁺ cells in the gel were evaluated in five different areas of the gel using ImageJ software^10,11,78,87 .

Statistical analysis

All phenotypic analysis was performed by masked observers unaware of the identity of experimental groups. Error bars (SEM) and p values were determined from the results of three or more independent experiments. Student’s t-test was used for statistical significance for two groups. For more than two groups, one-way ANOVA with a Turkey’s post-hoc analysis was conducted using GraphPad Prism 8. A p value of less than 0.05 was considered statistically significant.

Author contributions

Conceived and designed the experiments: T.M. and A.M. Performed the experiments: T.M., P.K., M.S., T.H., E.H., and A.M. Analyzed the data: T.M., P.K., M.S., T.H., E.H., X.M., C.W.L., S.M., and A.M. Contributed reagents/materials/analysis tools: T.M., P.K., M.S., T.H., E.H., C.T., S.M., and A.M. Wrote the paper: T.M. and A.M. All authors read and approved the manuscript.

Data availability

All data are available in the main text or the supplementary materials. The raw datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. scRNA-seq data have been made available on Gene Expression Omnibus database (GSE230548). Proteomics data have been made available on proteomeXchange (PXD038600).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41536-026-00476-9.