Research ArticleVASCULAR BIOLOGY

FOXF1 maintains endothelial barrier function and prevents edema after lung injury

See allHide authors and affiliations

Sci. Signal.  19 Apr 2016:
Vol. 9, Issue 424, pp. ra40
DOI: 10.1126/scisignal.aad1899

Improving endothelial barrier in the lung

Acute lung injury decreases the ability of the endothelial cells lining pulmonary blood vessels to be an effective barrier, resulting in the accumulation of fluid in the lungs (a condition called pulmonary edema) and inflammation. Cai et al. found that, in adult lung endothelial cells, the transcription factor FOXF1 transcriptionally activated a gene encoding the receptor for S1P, a lipid mediator that enhances the barrier function of endothelial cells. Adult mice that lacked one Foxf1 allele in lung endothelial cells were more likely to develop pulmonary edema and die after acute lung injury, outcomes that were reversed by administration of S1P. Thus, therapies that increase the activity of FOXF1 or S1P signaling could be used to decrease the complications that arise after acute lung injury, which can require hospitalization and can be fatal.

Abstract

Multiple signaling pathways, structural proteins, and transcription factors are involved in the regulation of endothelial barrier function. The forkhead protein FOXF1 is a key transcriptional regulator of embryonic lung development, and we used a conditional knockout approach to examine the role of FOXF1 in adult lung homeostasis, injury, and repair. Tamoxifen-regulated deletion of both Foxf1 alleles in endothelial cells of adult mice (Pdgfb-iCreER/Foxf1−/−) caused lung inflammation and edema, leading to respiratory insufficiency and death. Deletion of a single Foxf1 allele made heterozygous Pdgfb-iCreER/Foxf1+/− mice more susceptible to acute lung injury. FOXF1 abundance was decreased in pulmonary endothelial cells of human patients with acute lung injury. Gene expression analysis of pulmonary endothelial cells with homozygous FOXF1 deletion indicated reduced expression of genes critical for maintenance and regulation of adherens junctions. FOXF1 knockdown in vitro and in vivo disrupted adherens junctions, enhanced lung endothelial permeability, and increased the abundance of the mRNA and protein for sphingosine 1-phosphate receptor 1 (S1PR1), a key regulator of endothelial barrier function. Chromatin immunoprecipitation and luciferase reporter assays demonstrated that FOXF1 directly bound to and induced the transcriptional activity of the S1pr1 promoter. Pharmacological administration of S1P to injured Pdgfb-iCreER/Foxf1+/− mice restored endothelial barrier function, decreased lung edema, and improved survival. Thus, FOXF1 promotes normal lung homeostasis and repair, in part, by enhancing endothelial barrier function through activation of the S1P/S1PR1 signaling pathway.

INTRODUCTION

Endothelial barrier function is critical for maintenance of normal alveolar homeostasis and lung repair after injury [reviewed in (1)]. Pulmonary endothelial cells play a key role in maintaining the endothelial barrier and mediate the exchange of gases, water, and macromolecules between blood and surrounding alveolar tissue. Endothelial cells secrete various cytokines and chemokines that control hemostasis and immunologic and inflammatory events, regulate vascular tone, and promote interactions with inflammatory cells and neighboring vascular cell types. Acute lung injury (ALI) occurs in response to various insults, including pneumonia, sepsis, trauma, and mechanical ventilation, and is characterized by increased endothelial permeability, pulmonary edema, alveolar injury, and recruitment of inflammatory cells to the lung (2). Acute respiratory distress syndrome (ARDS) is a life-threatening complication of ALI with a mortality rate of more than 35%, which accounts for about 75,000 deaths and 3.5 million hospital days per year in the United States (2, 3). Given the lack of major improvements in the clinical management of ALI and ARDS, there is a compelling need for innovative molecular approaches to complement existing ALI and ARDS therapies.

Endothelial barrier function is regulated by various factors. Thrombin, bradykinin, histamine, reactive oxygen species (ROS), vascular endothelial growth factor (VEGF), tumor necrosis factor–α (TNF-α), and endotoxin increase endothelial permeability in cultured pulmonary endothelial cells and animals [reviewed in (1, 4)]. In contrast, sphingosine 1-phosphate (S1P) and angiopoietin-1 (Ang-1) stabilize the endothelial barrier and decrease endothelial permeability (1). S1P is a lysophospholipid mediator that circulates in the blood and maintains endothelial barrier function primarily through the S1PR1 receptor (4, 5). Activation of S1PR1 by S1P substantially enhances the endothelial cell barrier through RacGTPase-dependent cortical actin rearrangement (6). S1pr1−/− mice exhibit embryonic lethality due to impaired vasculogenesis (7), implicating the S1pr1 gene in the development of blood vessels. In the adult lung, pharmacological inhibition of S1PR1 enhances pulmonary capillary leakage (8). In zebrafish and mice, S1PR1 restricts angiogenic sprouting and ectopic vessel branch formation by stabilization of the VE-cadherin protein (9). VE-cadherin is a key structural component of adherens junctions, cellular structures that play an important role in the maintenance of vascular integrity and lung repair after injury (1, 4). S1P/S1PR1 signaling stabilizes VE-cadherin and is required for proper maintenance and regulation of adherens junctions in endothelial cells.

The transcription factor forkhead box F1 (FOXF1) is a critical mediator of angiogenesis during lung embryonic development (1013). Heterozygous deletions and point mutations in the Foxf1 gene locus are found in patients with alveolar capillary dysplasia with misalignment of pulmonary veins (ACD/MPV), a rare congenital disorder of neonates and infants characterized by severe defects in the development of the alveolar capillary network, lung hypoplasia, malposition of pulmonary veins, and hypoxemia. Because of the severity of developmental defects and progressive respiratory insufficiency in ACD/MPV infants, mortality usually occurs within the first month of life. Global deletion of Foxf1 in mice is embryonically lethal (12). Mice heterozygous for the Foxf1 null allele (Foxf1+/−) exhibit alveolar capillary dysplasia and a variety of developmental defects in the lung, liver, and intestine (11, 13–16). Endothelial deletion of Foxf1 in utero (Tie2-Cre Foxf1−/− and Pdgfb-iCreER/Foxf1−/− embryos) produces an embryonic lethal phenotype due to impaired VEGF signaling in the yolk sac, placenta, and lung (10), indicating a critical role for FOXF1 during embryogenesis.

Here, we examined FOXF1 function in quiescent adult lung where angiogenesis is inactive but FOXF1 is abundant. Deletion of both Foxf1 alleles in endothelial cells in adult Pdgfb-iCreER/Foxf1−/− mice caused uniform mortality due to severe pulmonary edema and increased lung inflammation, whereas deletion of only one Foxf1 allele was sufficient to increase the susceptibility of heterozygous Pdgfb-iCreER/Foxf1+/− mice to ALI. In vitro and in vivo studies demonstrated that FOXF1 regulates endothelial barrier function through transcriptional activation of the S1pr1 promoter. Thus, FOXF1 is a key transcriptional regulator of endothelial cells and is essential for normal lung homeostasis and lung repair after injury.

RESULTS

Conditional deletion of Foxf1 from endothelial cells causes respiratory insufficiency and lethality in adult mice

To determine the role of FOXF1 in the adult lung, we generated conditional knockout mice with an inducible deletion of Foxf1 from endothelial cells. These mice contained the Pdgfb-iCreER transgene and two Foxf1-floxed alleles with a LoxP-flanked exon 1, which encodes the DNA binding domain of the FOXF1 protein (Pdgfb-iCreER/Foxf1−/− or endFoxf1−/− mice) (fig. S1A). Tamoxifen (Tam) was given to adult mice to induce Cre-mediated deletion of Foxf1-floxed alleles in endothelial cells (fig. S1B). Two weeks after Tam treatment, endFoxf1−/− mice had visible difficulties with breathing. Histopathology of endFoxf1−/− lungs revealed thickening of the alveolar septum, lung congestion, and severe inflammation in alveolar regions and pulmonary airways (Fig. 1A). Pulmonary hemorrhage and fibrosis subsequently developed in Tam-treated endFoxf1−/− mice, causing uniform mortality (Fig. 1, A and B). There were no visible histological abnormalities in the hearts, livers, or kidneys of endFoxf1−/− mice (fig. S2A). Evans blue dye did not accumulate in the brain of endFoxf1−/− mice, indicating normal blood brain barrier function (fig. S2A). All control groups of mice, including Tam-treated Foxf1fl/fl mice, vehicle-treated Pdgfb-iCreER/Foxf1fl/fl mice, and Tam-treated Pdgfb-iCreER mice, had normal lung histology (Fig. 1A and fig. S1D). To assess pulmonary edema, we measured the wet/dry weight ratio of lung tissue and the concentration of protein in bronchoalveolar lavage fluid (BALF) and found that both were significantly increased in Tam-treated endFoxf1−/− mice compared to Tam-treated Foxf1fl/fl controls (Fig. 1, C and D). Airway compliance was reduced in endFoxf1−/− mice, whereas airway resistance and respiratory elasticity were increased (Fig. 1E), a finding consistent with increased stiffness of the lung tissue due to inflammation and edema. Thus, deletion of Foxf1 from endothelial cells of adult mice caused severe respiratory insufficiency, pulmonary edema, and mortality.

Fig. 1 Deletion of Foxf1 from endothelial cells causes respiratory insufficiency and lethality in adult mice.

(A) H&E staining of lung paraffin sections from Tam-treated endFoxf1−/− mice shows lung congestion, infiltration of inflammatory cells from blood vessels (V) into the lung tissue, occlusion of pulmonary bronchioles (Br), and pulmonary hemorrhage (shown with arrowheads). Tam was given 15 days before the lung harvest. Trichrome staining shows fibrosis in Tam-treated endFoxf1−/− lungs. Scale bars, 50 μm. Images are representative of five mice per genotype. (B) Kaplan-Meier survival analysis shows uniform mortality in Tam-treated endFoxf1−/− mice (n = 20 mice per genotype). (C) Tam-treated endFoxf1−/− lungs were visibly edematous (left panel) (images are representative of five mice per genotype) and had a higher lung wet/dry weight ratio (right panel) (n = 5 mice per genotype). (D) Protein concentration in BALF was increased in Tam-treated endFoxf1−/− mice (n = 5 mice per genotype). (E) Tam-treated endFoxf1−/− mice exhibited decreased airway compliance, but airway resistance and respiratory elasticity were increased (n = 6 mice per genotype). *P < 0.05; **P < 0.01.

Foxf1 deletion from endothelial cells causes accumulation of macrophages in lung tissue

To assess both the efficiency and the specificity of Foxf1 deletion by the Pdgfb-iCreER transgene, we harvested endFoxf1−/− lungs at different time points after Tam treatment. Three days after the last Tam administration, Foxf1 mRNA in endFoxf−/− lungs was reduced by 85%, and this reduction was maintained until 17 days after treatment (fig. S3A). Consistent with mRNA data, FOXF1 staining was reduced in endFoxf1−/− lungs compared to controls (fig. S4, A to C). Because the Pdgfb-iCreER transgenic construct contains enhanced green fluorescent protein (eGFP) (17), we used flow cytometry to identify the cell population(s) targeted by the Cre transgene. In endFoxf1−/− lungs, eGFP was only detected in the CD45CD31(Pecam-1)+ endothelial cell population but not in the nonendothelial (CD45CD31) or hematopoietic CD45+ cell subsets (fig. S3B). Thus, Pdgfb-iCreER targets endothelial cells in the adult lung.

To confirm that the Foxf1 gene was deleted from pulmonary endothelial cells, we used fluorescence-activated cell sorting (FACS) to isolate CD45CD31+ endothelial and CD45CD326(Epcam-1)+ epithelial cell subsets from Tam-treated endFoxf1−/− and control lungs (fig. S3C). Purified endothelial cells contained endothelial cell–specific mRNAs (Cdh5, Flk1, and Pecam-1) but not those encoding epithelial or fibroblast markers, whereas epithelial cells selectively expressed mRNAs encoding epithelial markers, such as surfactant-associated proteins SP-A, SP-B, and SP-C (fig. S5). Foxf1 mRNA was reduced in FACS-sorted endothelial cells from the endFoxf1−/− lungs (fig. S3C), a finding consistent with efficient deletion of Foxf1 by Pdgfb-iCreER. Compared to endothelial cells, FACS-sorted epithelial cells contained low amounts of Foxf1 mRNA that were unchanged in endFoxf1−/− mice (fig. S3C). Furthermore, colocalization studies with endothelial cell–specific VE-cadherin (a membrane marker) and the transcription factor Erg (a nuclear marker) demonstrated that FOXF1 protein was present in the nuclei of alveolar endothelial cells and that Foxf1 deletion reduced FOXF1 staining (figs. S3D and S4D). Although FOXF1 was also detected in peribronchial smooth muscle cells, its abundance was unchanged in endFoxf1−/− mice (fig. S4B). FOXF1 was not detected in alveolar macrophages (fig. S1C). Together, these data indicate that after Tam treatment, Pdgfb-iCreER efficiently and specifically deletes Foxf1 from the endothelial cells of adult lungs.

Deletion of Foxf1 resulted in the progressive accumulation of macrophages in lung tissue as shown by hematoxylin and eosin (H&E) staining (Fig. 2A) and immunostaining for Mac-3 (Fig. 2, B and C). The number of macrophages was increased in BALF obtained from endFoxf1−/− mice compared to that obtained from Foxf1fl/fl controls (Fig. 2D). The numbers of lymphocytes and neutrophils in BALF were not altered in endFoxf1−/− mice (Fig. 2D). Foxf1 mRNA abundance was about 60-fold higher in FACS-sorted pulmonary endothelial cells compared to those isolated from the liver or kidney (fig. S2B), suggesting that the lung phenotype in FOXF1-deficient mice is due to high Foxf1 expression in pulmonary endothelial cells. Finally, transplantation of wild-type bone marrow to irradiated endFoxf1−/− mice did not influence lung inflammation or survival (fig. S6). Thus, FOXF1 inactivation in resident lung endothelial cells causes respiratory insufficiency and mortality.

Fig. 2 Deletion of Foxf1 increases lung inflammation.

(A) H&E staining shows accumulation of inflammatory cells in endFoxf1−/− lungs at different time points after Tam administration (n = 5 mice per genotype and time point). (B and C) Immunostaining with Mac-3 antibody shows increased numbers of macrophages after deletion of Foxf1. Mac-3–positive cells were counted in 10 random microscope fields of Tam-treated endFoxf1−/− and Foxf1fl/fl lungs (n = 5 mice per genotype). (D) Diff-Quik staining shows increased numbers of macrophages in BALF of Tam-treated endFoxf1−/− mice (n = 5 mice per genotype). Scale bar, 20 μm. *P < 0.05; **P < 0.01.

Foxf1 deletion from endothelial cells decreases the abundance of S1PR1

We compared genome-wide mRNA expression profiles between Tam-treated endFoxf1−/− and control Foxf1fl/fl lungs. RNA sequencing (RNA-seq) analysis identified 1137 genes with decreased expression and 1417 genes with increased expression after deletion of Foxf1 (Fig. 3A and table S1). Genes in the “biological adhesion” and “cell-substrate adhesion” functional categories showed the greatest suppression among genes with decreased expression in endFoxf1−/− lungs (Fig. 3B). Among the genes showing increased expression in endFoxf1−/− lungs, “leukocyte migration” and “inflammatory response” were the most enriched functional classes (Fig. 3B). Foxf1 mRNA was decreased 7.7-fold in endFoxf1−/− lungs (Fig. 3C), confirming the quantitative reverse transcription polymerase chain reaction (qRT-PCR) data (fig. S3A). The expression of IL6 [which encodes interleukin (IL)-6], Ccl2 [which encodes chemokine CC motif ligand (CCL) 2], Ccl7 (which encodes CCL7), IL33 (which encodes IL-33), and Sele (which encodes E-selectin) was increased in endFoxf1−/− lungs (Fig. 3C), findings consistent with increased lung inflammation. In contrast, endFoxf1−/− lungs exhibited decreased expression of genes critical for structural integrity and regulation of endothelial cell junctions, such as S1pr1 (which encodes S1PR1), Clnd5 (which encodes claudin-5), Ctnna1 (which encodes α-catenin1), Ctnnd1 (which encodes δ-catenin1), Vcl (which encodes vinculin), Jup (which encodes junction plakoglobin), Actn1 (which encodes α-actinin), Lima1 (which encodes LIM domain and actin-binding protein 1), and Cdh5 (which encodes VE-cadherin) (Fig. 3C). The expression of these seven genes was confirmed by qRT-PCR analysis of total lung RNA (Fig. 3D).

Fig. 3 Foxf1 deletion from endothelial cells decreases S1pr1, Clnd5, and Cdh5 mRNAs.

(A) Differentially expressed genes in Tam-treated endFoxf1−/− lungs were identified using RNA-seq analysis. Heat map analysis shows the two-dimensional clustering of mRNAs that were significantly altered in Tam-treated endFoxf1−/− lungs compared to Tam-treated Foxf1fl/fl controls. Samples from three mice per genotype were combined together before the RNA-seq analysis. (B) Biological processes influenced by endothelial Foxf1 deletion were identified using ToppGene Suite (http://toppgene.cchmc.org). The statistical significance of each biological process was presented using negative log2 transformation of P value. (C) mRNA changes and P values in Tam-treated endFoxf1−/− lungs are shown in the table. (D) RNA-seq data were confirmed by qRT-PCR of total lung RNA (n = 5 mice per genotype). mRNA abundance was normalized to β-actin mRNA. (E and F) FACS was used to isolate endothelial cells (CD45CD31+eGFP+) from endFoxf1−/− lungs. Endothelial cells from Tam-treated Pdgfb-iCreER/Foxf1+/+ mice were used as a control. qRT-PCR shows reduced Foxf1 and S1pr1 mRNAs in FACS-sorted endothelial cells from Tam-treated endFoxf1−/− lungs (n = 3 mice per genotype). *P < 0.05; ***P < 0.001. SSC-A, side-scatter area; FSC-A, forward-scatter area.

Because S1P/S1PR1 signaling enhances endothelial barrier function (1, 4), reduced S1PR1 abundance could contribute to edema and inflammation in endFoxf1−/− lungs. Therefore, we focused our next studies on the regulation of S1pr1 gene expression by FOXF1. S1pr1 mRNA was decreased in Tam-treated endFoxf1−/− lungs (fig. S7A), correlating with reduced Foxf1 mRNA (fig. S3A). Western blot and immunostaining of lung sections showed decreased abundance of S1PR1 and one of its targets, VE-cadherin (figs. S4D and S7, B and C). To determine whether changes in S1PR1 occurred specifically in endothelial cells, FACS-sorted pulmonary endothelial cells were isolated at day 3 after Tam treatment, a time point before edema and inflammation in endFoxf1−/− lungs. Reduced Foxf1 mRNA abundance was associated with decreased S1pr1 in FACS-sorted endothelial cells (Fig. 3, E and F), suggesting that FOXF1 directly or indirectly regulates the expression of the S1pr1 gene.

Foxf1 deletion from endothelial cells increases endothelial permeability

Because S1P/S1PR1 signaling is essential for structural integrity and regulation of adherens junctions in pulmonary endothelial cells (1, 4), we assessed vascular permeability in Tam-treated endFoxf1−/− lungs. BALF was collected after the injection of Evans blue dye into endFoxf1−/− and control Foxf1fl/fl mice. Increased amounts of Evans blue dye were found in BALF from endFoxf1−/− mice at day 6 and day 15 after Tam administration (Fig. 4A, left panel). FOXF1-deficient lungs retained Evans blue dye after extensive perfusion of blood vessels with saline (Fig. 4A, right panel), indicating vascular leak into the lung tissue. These results show that deletion of Foxf1 from endothelial cells increased vascular permeability in the adult lung.

Fig. 4 Knockdown of FOXF1 increases endothelial permeability.

(A) Pulmonary vascular permeability was determined by Evans blue dye extravasation. The amounts of Evans blue dye in BALF of endFoxf1−/− mice were unchanged at day 3 but increased at days 6 and 15 after Tam administration (n = 5 mice per genotype and time point). (B) Foxf1, Cdh5, and S1pr1 mRNAs were decreased in FOXF1-depleted HUVECs as shown by qRT-PCR. Bcl2 and Bax mRNAs were unchanged. β-Actin mRNA was used for normalization (n = 3 independent cultures). (C) Colocalization studies show efficient depletion of FOXF1 (red) from the nuclei of HUVECs [stained with endothelial cell–specific marker Erg (green)] transduced with a lentivirus containing FOXF1-specific shRNA (Lenti-shFoxf1). Images are representative of three independent experiments. Scale bars, 20 μm. (D) Immunofluorescence staining of VE-cadherin (green) and F-actin (red) shows loss of VE-cadherin in the adherens junctions of FOXF1-depleted HUVECs (arrowheads). Images are representative of three independent experiments. Scale bars, 20 μm. DAPI, 4′,6-diamidino-2-phenylindole. (E) FOXF1 depletion by Lenti-shFoxf1 increased transendothelial albumin leak (representative of three independent experiments). (F) The number of apoptotic cells was not changed after depletion of FOXF1. Forty-eight hours after lentiviral transduction, HUVECs were stained with annexin V–APC and 7AAD. Percentages of necrotic (7AAD+annexin V) and apoptotic cells (annexin V+) were determined by flow cytometry (representative of three independent experiments). (G) FOXF1 depletion in HUVECs did not influence cell survival and metabolic activity as demonstrated by the WST-1 assay (representative of three independent experiments). *P < 0.05; ***P < 0.001.

The effect of FOXF1 inhibition on endothelial barrier function was also analyzed in vitro using cell monolayers of human umbilical vein endothelial cells (HUVECs). FOXF1 was knocked down in HUVECs by a lentivirus containing a FOXF1-specific short hairpin RNA (shRNA), which efficiently reduced Foxf1 mRNA (Fig. 4B) and decreased FOXF1 protein staining in nuclei of HUVECs (Fig. 4C). Similar to in vivo data, knockdown of FOXF1 reduced S1pr1 and Cdh5 mRNAs (Fig. 4B). Accumulation of VE-cadherin protein at adherens junctions was decreased in FOXF1-deficient cells (Fig. 4D). To measure endothelial permeability, Evans blue dye–labeled albumin was added to the top chamber containing a complete monolayer of HUVECs. The albumin leak to the lower chamber was subsequently measured. Knockdown of FOXF1 increased albumin leakage through the HUVEC monolayer (Fig. 4E), indicating that FOXF1 is important for endothelial barrier function. Because the endothelial barrier can be disrupted by apoptosis or necrosis of endothelial cells, we performed FACS analysis of FOXF1-depleted HUVECs using 7-aminoactinomycin D (7AAD) and annexin V. FOXF1 knockdown did not change the percentage of dead (7AAD+annexin V) or apoptotic (annexin V+) cells (Fig. 4F). The lack of changes in cell survival and metabolic activity was confirmed by colorimetric measurement of WST-1 reagent (Fig. 4G). The abundance of the mRNAs of the apoptotic markers Bcl2 and Bax was not altered after depletion of FOXF1 (Fig. 4B). Six days after Tam treatment, endFoxf1−/− lungs had no change in the number of either apoptotic endothelial cells (Pecam-1+annexin V+) or necrotic endothelial cells (Pecam-1+7AAD+annexin V) compared to Tam-treated Foxf1fl/fl controls (Fig. 5, A and B). However, at day 15 after Tam treatment, increased apoptosis was observed in endFoxf1−/− lungs (Fig. 5C). Despite increased apoptosis, cleaved caspase-3 was not detected in endothelial cells of Tam-treated endFoxf1−/− lungs (Fig. 5C). Thus, knockdown of FOXF1 in vitro and in vivo reduced S1PR1 abundance, decreased the accumulation of VE-cadherin in cell junctions, and impaired endothelial barrier function.

Fig. 5 Deletion of Foxf1 does not induce apoptosis of endothelial cells in the lung.

(A and B) Flow cytometry was used to identify apoptotic cells in Tam-treated Foxf1fl/fl and endFoxf1−/− lungs at day 6 after Tam treatment. CD31+ endothelial cells were costained with annexin V and 7AAD. Percentages of live (annexin V/7AAD), early apoptotic (annexin V+/7AAD), late apoptotic (annexin V+/7AAD+), and necrotic (dead) cells (annexin V/7AAD+) are shown on scatter plots (A) and histograms (B). Three mice were used in each group. (C) Immunostaining for cleaved caspase-3 shows the presence of apoptotic cells in endFoxf1−/− lungs at day 15 after Tam treatment. Apoptosis was not detected in endothelial cells (arrowheads) (n = 3 mice per genotype). Scale bars, 50 μm.

FOXF1 directly binds to and induces the transcriptional activities of the S1pr1 and Cdh5 promoters

Because FOXF1 deficiency was associated with decreased S1PR1 expression in vivo and in vitro, we investigated whether FOXF1 directly regulates the S1pr1 promoter. Four potential FOXF1 binding sites were identified in the −6.0-kb promoter region of the mouse S1pr1 gene. Chromatin immunoprecipitation (ChIP) was performed using mouse lung tissue to determine whether FOXF1 protein physically bound to the S1pr1 promoter. qPCR was used to identify specific DNA promoter fragments that interacted with FOXF1. FOXF1 specifically bound to the −3880/−3876–base pair (bp) and −5141/−4815-bp regions of the S1pr1 promoter that contain FOXF1 binding sites (Fig. 6A). FOXF1 did not bind to the Hoxa11 genomic DNA or the −2772/−2759-bp S1pr1 promoter region that lacks FOXF1 binding sites (Fig. 6A). In endothelial MFLM-91U cells, overexpressed FOXF1 transcriptionally activated a reporter containing the S1pr1 promoter fused to luciferase (Fig. 6B). Furthermore, FOXF1 directly bound to and transcriptionally activated the mouse VE-cadherin (Cdh5) promoter through the −788/−775-bp promoter region (Fig. 6, C and D). FOXF1 did not bind to the Cldn5 promoter in spite of the presence of two potential FOXF1 binding sites in the −2451/−2439-bp and −1818/−1806-bp Cldn5 promoter regions (Fig. 6E). These data indicate that S1pr1 and Cdh5 are direct transcriptional targets of FOXF1 in endothelial cells.

Fig. 6 FOXF1 directly binds to and induces transcriptional activity of the S1pr1 and Cdh5 promoters.

(A) Schematic drawing of the mouse S1pr1 promoter region with potential FOXF1 DNA binding sites (white boxes). FOXF1 protein binds to the −5141/−4815-bp and −3880/−3876-bp S1pr1 promoter regions, the former of which has three potential FOXF1 binding sites. FOXF1 binding to both S1pr1 promoter regions was decreased in Tam-treated endFoxf1−/− lungs. FOXF1 binding to genomic DNA was normalized to immunoglobulin G (IgG) control antibodies. FOXF1 did not bind to the −2772/−2759-bp S1pr1 promoter region or HoxA11 genomic DNA that lacked potential FOXF1 binding motifs (negative controls) (n = 3 mice per genotype). (B) Schematic drawing of the pGL2-mS1PR1pr-Luc construct containing the S1pr1 promoter region. The CMV-FOXF1 plasmid increases luciferase (Luc) activity of this reporter construct in MFLM-91U cells. CMV-empty plasmid was used as a negative control. Data are representative of three independent experiments. (C and D) Schematic drawings, ChIP, and Luc assay of the mouse VE-cadherin (Cdh5) promoter. FOXF1 binds to and activates the Cdh5 promoter through the −788/−775-bp promoter region (representative of three independent experiments). (E) Schematic drawing of the mouse Cldn5 promoter region. Locations of potential FOXF1 DNA binding sites are indicated (white boxes). FOXF1 does not bind Cldn5 promoter DNA as shown by ChIP (representative of three independent experiments). **P < 0.01.

Heterozygous endFoxf1+/− mice exhibit increased susceptibility to butylated hydroxytoluene–induced lung injury

We next examined FOXF1 requirements during lung injury and repair. Because homozygous endFoxf1−/− mice developed lung edema and inflammation without exogenous insults (Figs. 1 and 2), we used heterozygous endFoxf1+/− mice in which only one Foxf1 allele was deleted from endothelial cells. Tam-treated endFoxf1+/− mice were phenotypically normal and exhibited normal lung histology (Fig. 7A). After lung injury was induced by butylated hydroxytoluene (BHT), 85% of Tam-treated endFoxf1+/− mice, but none of the Tam-treated Foxf1+/fl controls, had respiratory insufficiency and mortality (Fig. 7B). Pulmonary inflammation was more severe in BHT-treated endFoxf1+/− mice compared to controls (Fig. 7, C to E). The amount of Evans blue dye was increased in the BALF and lung tissue of BHT-treated endFoxf1+/− mice (Fig. 7F). Although Foxf1 mRNA was decreased after BHT treatment in both groups of mice, these changes were more prominent in endFoxf1+/− lungs (Fig. 7C). S1pr1 mRNA was lower in BHT-treated endFoxf1+/− lungs (Fig. 7C), a finding consistent with the transcriptional activation of the S1pr1 promoter by FOXF1 (Fig. 6). Moreover, FOXF1 staining was decreased in pulmonary endothelial cells from patients with severe bacterial pneumonia (Fig. 7G), suggesting that FOXF1 plays a role in lung injury and repair in humans. Together, our results indicate that endothelial cell–specific deletion of only one Foxf1 allele was sufficient to increase susceptibility of adult mice to BHT-mediated lung injury.

Fig. 7 Pharmacological treatment with S1P protects heterozygous endFoxf1+/− mice from BHT-mediated lung injury.

(A) Deletion of one Foxf1 allele in adult heterozygous endFoxf1+/− mice did not influence lung structure as shown by H&E staining (left panels). Increased inflammation was found in Tam-treated endFoxf1+/− lungs after BHT administration (right panels). Images are representative of five mice per genotype. (B) Kaplan-Meier survival analysis shows that administration of S1P partially rescues mortality in BHT-treated endFoxf1+/− mice (n = 15 mice per genotype and treatment). (C) qRT-PCR shows that BHT injury decreased Foxf1 and S1pr1 mRNAs in Tam-treated endFoxf1+/− lungs compared to Tam-treated Foxf1+/fl controls. β-Actin mRNA was used for normalization (n = 5 mice per genotype and time point). (D and E) S1P decreases macrophage inflammation in BHT-treated endFoxf1+/− mice as shown by Mac-3 staining. Mac-3–positive cells were counted in 10 random microscope fields (n = 5 mice per genotype). (F) Increased vascular permeability in endFoxf1+/− mice after BHT-induced lung injury. Administration of S1P decreased vascular permeability in BHT-treated endFoxf1+/− mice. The amounts of Evans blue dye were measured in BALF (n = 5 mice per genotype and treatment). (G) FOXF1 is decreased in human patients with ALI. FOXF1 staining was performed using lung paraffin sections from patients with severe bacterial pneumonia (n = 7). Normal human lungs (patients without lung injury) were used as controls (n = 6). FOXF1 abundance is decreased in pulmonary endothelial cells from patients with ALI (arrowheads in inserts). Arrows show FOXF1-positive cells in alveolar region. Scale bars, 50 μm. Ar, artery. *P < 0.05.

Pharmacological treatment with S1P prevents mortality and reduces vascular permeability in BHT-treated endFoxf1+/− mice

To determine whether diminished S1P/S1PR1 signaling contributes to increased susceptibility of FOXF1-deficient mice to BHT lung injury, we used S1P to induce S1P/S1PR1 signaling in vivo. Two tail vein injections of S1P significantly decreased mortality in heterozygous endFoxf1+/− mice after BHT injury (Fig. 7B and fig. S8). Administration of S1P reduced the number of macrophages in BHT-treated endFoxf1+/− lungs to that seen in BHT-treated Foxf1+/fl controls (Fig. 7, D and E). S1P also decreased the accumulation of Evans blue dye in the BALF and lung tissue of BHT-treated endFoxf1+/− mice (Fig. 7F). Thus, S1P improved endothelial barrier function and reduced lung inflammation in BHT-injured endFoxf1+/− mice. S1P treatment did not increase the survival of homozygous endFoxf1−/− mice (fig. S9, A and B), most likely due to the contribution of other FOXF1 target genes to the endFoxf1−/− phenotype. Similar to BHT-induced lung injury, intratracheal administration of lipopolysaccharide (LPS) increased the number of inflammatory cells in the BALF and lung tissue of endFoxf1+/− mice (fig. S10, A and B). The S1PR1-selective agonist CYM5442 attenuated lung inflammation in LPS-treated endFoxf1+/− mice (fig. S10, A and B). Thus, S1pr1 is an important transcriptional target of FOXF1 in the adult lung, and the increased susceptibility of endFoxf1+/− mice to lung injury could be ameliorated by pharmacological administration of S1P or CYM5442.

DISCUSSION

S1P maintains endothelial barrier integrity by binding to the S1PR1 receptor, causing RacGTPase-dependent cortical actin rearrangement (1, 6). Although S1PR2 and S1PR3 are also present in endothelial cells, the barrier-enhancing property of S1P is mainly mediated through S1PR1. A stable knockdown of S1PR1 disrupts adherens junctions and decreases VE-cadherin abundance in HUVECs after LPS and TNF-α stimulation (18). S1P/S1PR1 stimulation inhibits vascular permeability and alveolar flooding in preclinical animal models of ALI (19), demonstrating that activation of the S1P/S1PR1 signaling pathway could be beneficial in patients with ALI and ARDS. Although both signal transducer and activator of transcription 3 (STAT3) and Krüppel-like factor 2 (KLF2) can transcriptionally activate the S1pr1 promoter in tumor cells and T lymphocytes (20, 21), neither STAT3 nor KLF2 is active in endothelial cells of quiescent lungs where S1pr1 is robustly expressed. Therefore, other transcriptional regulators must be involved in the maintenance of S1pr1 transcription during normal lung homeostasis. An important contribution of the present study is that FOXF1 induced S1pr1 transcription in both quiescent and injured lungs, suggesting that FOXF1 is a major regulator of S1P/S1PR1 signaling in endothelial cells. FOXF1 regulation of S1PR1 appears to be at the transcriptional level, as supported by ChIP analysis and luciferase reporter assays using the S1pr1 promoter. Genetic ablation of FOXF1 in purified endothelial cells caused a 50% reduction in S1pr1 mRNA. These data suggest that the S1pr1 gene is regulated by other endothelial transcription factors in addition to FOXF1.

VE-cadherin is a key structural component of adherens junctions and critical for the maintenance and regulation of vascular integrity during embryonic development, lung repair, and alveolar homeostasis (1, 4). VE-cadherin controls vascular permeability and inhibits unrestrained vascular growth (22). Several barrier-destabilizing mediators, such as thrombin and VEGF, increase endothelial permeability through VE-cadherin endocytosis and degradation (1). Our data suggest that FOXF1 maintains endothelial barrier function in pulmonary endothelial cells, at least in part, by activating the Cdh5 promoter upstream of the gene encoding VE-cadherin. Furthermore, we showed that FOXF1 induces endothelial barrier function through transcriptional induction of the S1pr1 promoter, which, in turn, causes activation of S1P/S1PR1 signaling and stabilization of VE-cadherin in endothelial junctions. Pharmacological administration of S1P to BHT-injured endFoxf1+/− mice improved survival after BHT injury, providing direct support to our concept. Whereas S1P/S1PR1 signaling is important for endothelial barrier function, endothelial cell–specific inactivation of S1pr1 or inhibition of plasma S1P does not cause lethality in adult mice (2325). Here, we found that FOXF1 inactivation in endothelial cells disrupted the expression of multiple genes in addition to S1pr1 and Cdh5. These include genes encoding structural proteins of endothelial cell junctions and proinflammatory regulators. Therefore, it is likely that respiratory insufficiency and lethality in FOXF1-deficient mice occur as a result of simultaneous disruption in the expression of many endothelial genes.

We previously demonstrated that FOXF1 induces transcription of Vegf receptor genes in embryonic endothelial cells of the yolk sac (10). Surprisingly, we found no differences in Flk1 and Flt1 mRNAs in adult endothelial cells after inactivation of FOXF1 (table 1). These results suggest that FOXF1 has different functions in embryonic endothelial cells as compared to adult endothelial cells. It is possible that distinct coactivator proteins alter the transcriptional effects of FOXF1, depending on tissue specificity or developmental status. RNA-seq analysis of FOXF1-deficient mouse lungs revealed an increase in mRNAs encoding proinflammatory mediators, such as IL6, Ccl2, and IL33. Given the important role of CCL2 and IL-6 in macrophage recruitment during lung injury, these findings are consistent with the increased numbers of macrophages in the BALF and lung tissue of FOXF1-deficient mice. It is possible that macrophage-produced proinflammatory mediators further exacerbate lung edema and inflammation, contributing to mortality in endFoxf1−/− mice. Although apoptosis was increased in endFoxf1−/− lungs, most apoptotic cells were nonendothelial cells. These data suggest that the increased apoptosis was a consequence of edema and inflammation in FOXF1-deficient lungs. Because IL-6 and IL-33 increase endothelial permeability in cultured endothelial cells and in vivo (26, 27), high concentrations of these proinflammatory mediators can contribute to increased endothelial permeability in FOXF1-deficient mice.

In summary, FOXF1 is a critical transcriptional regulator of pulmonary endothelial cells and is required for maintenance of endothelial barrier function during normal alveolar homeostasis and lung repair after injury. Discovery of small-molecule compounds that activate or stabilize FOXF1 protein could provide novel treatments for patients with ALI and ARDS.

MATERIALS AND METHODS

Mice, experimental lung injury, and measurement of endothelial permeability

Foxf1fl/fl mouse line was previously generated and bred into the C57Bl/6 mouse background (10). Foxf1fl/fl mice were bred with Pdgfb-iCreER mice (17) to generate Pdgfb-iCreER/Foxf1fl/fl (abbreviated as endFoxf1−/−) and Pdgfb-iCreER/Foxf1+/fl (abbreviated as endFoxf1+/−) mice. Tam (3 mg; Sigma) was given as oral gavage. Evans blue dye (50 mg/kg) was injected intraperitoneally, and BALF was collected 4 hours after injection. The vasculature was perfused with saline to remove intravascular Evans blue dye. The concentration of Evans blue dye was determined by measuring the absorbance at 620 and 740 nm and normalized to a standard curve of known Evans blue dye concentrations. Lung injury was induced by a single intraperitoneal injection of BHT (3,5-di-tert-butyl-4-hydroxytoluene; 200 mg/kg per body weight; Sigma) diluted in corn oil as described (16) Mice were harvested at day 6 after BHT administration. Tam was given 5 days before BHT treatment. For the S1P rescue experiment, Tam-treated endFoxf1+/− and control mice were intravenously injected with S1P (20 nmol per mouse; Sigma) or vehicle (saline) on days 2 and 8 after BHT treatment. LPS (10 μg per mouse; catalog no. L3012-5G, Sigma) was given intratracheally on days 3, 5, and 7 after Tam administration. CYM5442 (10 mg/kg per body weight; Tocris) was given intraperitoneally on days 6, 7, 8, and 9 after Tam treatment, and mice were harvested on day 10. A flexiVent computer-controlled small animal ventilator was used to measure lung function as previously described (28, 29). All animal studies were approved by the Animal Care and Use Committee of Cincinnati Children’s Research Foundation.

Histology and immunostaining

Lung paraffin sections were used for H&E staining or immunohistochemistry as previously described (30, 31). The following antibodies were used for immunostaining: FOXF1 [1:1000; (10)], Mac-3 (1:1500; #553322, BD Pharmingen), Pecam-1 (1:500; #553370, BD Pharmingen), Erg-1/2/3 (1:200; #sc-353, Santa Cruz Biotechnology), β-catenin (1:300; #sc-7199 or #sc-1496, Santa Cruz Biotechnology), VE-cadherin (1:200, #sc-6458, Santa Cruz Biotechnology), and S1P1/EDG-1 (1:100; #MAB7089, R&D Systems). Antibody-antigen complexes were detected using biotinylated secondary antibody followed by avidin–biotin–horseradish peroxidase complex and 3,3′-diaminobenzidine substrate (Vector Labs), as previously described (16, 32, 33). Sections were counterstained with nuclear fast red (Vector Labs). For immunofluorescence imaging, secondary antibodies conjugated with Alexa Fluor 488 or Alexa Fluor 594 (Invitrogen/Molecular Probes) were used as described (34, 35). Cell nuclei were counterstained with DAPI. Rhodamine phalloidin (1:1000; #R415, Molecular Probes) was used to visualize F-actin. Images were obtained using a Zeiss AxioPlan 2 microscope.

Flow cytometry and measurement of apoptosis

Flow cytometry was performed using enzyme-digested lung tissue as previously described (29, 36, 37). Apoptotic endothelial cells were detected by the annexin V apoptosis detection kit APC (eBioscience). 7AAD (eBioscience) was used to label necrotic cells. Antibodies against CD31 (Pecam-1), CD45, and CD326 were purchased from eBioscience. Stained cells were analyzed using a BD FACSCanto III flow cytometer. The WST-1 reagent was purchased from Roche and used according to the manufacturer’s instructions.

Cell culture, qRT-PCR, and Western blot

The MISSION shRNA Lentiviral system (Sigma-Aldrich) was used to knock down FOXF1 in HUVECs (Lonza) according to the manufacturer’s instructions. For all experiments, HUVECs were used in passages 3 to 6. Evans blue dye–labeled albumin was used to measure permeability in confluent monolayers of HUVECs as previously described (38). Evans blue dye–labeled albumin leak was measured as a flux of albumin in nanograms per minute using a standard curve of known albumin concentrations. mRNAs of specific genes were measured by qRT-PCR using TaqMan probes (table S2) and the StepOnePlus Real-Time PCR system (Applied Biosystems) as described (31, 39). RNeasy Mini kit (Qiagen) was used to isolate the total RNA from endothelial cells. Western blot was performed using Pierce ECL Western blotting substrate (Thermo Scientific) as described (40, 41).

ChIP, dual-luciferase assay, and RNA-seq analysis

Lung tissue from Tam-treated endFoxf1−/− and control Foxf1fl/fl mice was cross-linked by formaldehyde. ChIP was performed using a FOXF1 antibody (10). Chromatin DNA fragments were 500 to 1000 bp in size. Reverse cross-linked ChIP DNA samples were subjected to qPCR using primers specific to promoter regions of murine S1pr1, Cdh5, and Cldn5 genes (table S3). Primers specific to the HoxA11 gene were used as a negative control. DNA binding was normalized to control ChIP DNA samples, which were immunoprecipitated using IgG control antibody (42). Luciferase reporters driven by the S1pr1 and Cdh5 promoters were generated by PCR amplification of mouse genomic DNA (primers are listed in table S3) followed by cloning of the DNA fragments into pGL2-Basic firefly luciferase plasmid (Promega) as described (43). FOXF1 mammalian expression construct (CMV-FOXF1) was previously described (44, 45). CMV-Renilla was used as an internal control to normalize transfection efficiency. Dual luciferase assay (Promega) was performed in MFLM-91U cells as described previously (32, 43). For RNA-seq analysis, RNA was isolated from endFoxf1−/− and control Foxf1fl/fl lungs at day 6 after Tam administration. Differentially expressed genes were identified using DESeq, and P value was calculated using the χ2 likelihood ratio test in Agilent GeneSpring GX suite (46). Genes with expression altered by a factor of 1.5 in endFoxf1−/− lungs were selected for gene set enrichment analysis using ToppGene Suite (47). The statistical significance of each bioprocess was calculated using negative log2 transformation of P value. Heat map was constructed on differentially expressed genes using JMP Genomics 6.0.

Statistical analysis

Statistical analysis was performed using Prism software. Student’s t test and one-way analysis of variance (ANOVA) were used to determine statistical significance. Right-skewed measurements were log-transformed to meet normality assumption before analyses. P < 0.05 was considered significant. Values for all measurements were expressed as means ± SEM. Kaplan-Meier estimate of survival distribution of two groups was also computed.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/9/424/ra40/DC1

Fig. S1. Breeding strategy, Tam administration protocol, and analysis of FOXF1 in pulmonary macrophages.

Fig. S2. Deletion of Foxf1 from endothelial cells does not influence the histological structure of the liver, heart, and kidney.

Fig. S3. FOXF1 is reduced in pulmonary endothelial cells of Tam-treated endFoxf1−/− mice.

Fig. S4. Tam treatment reduces the number of FOXF1-positive endothelial cells in endFoxf1−/− lungs.

Fig. S5. Cdh5, Flk1, and Pecam-1 mRNAs are present in FACS-sorted endothelial cells.

Fig. S6. Transplantation of bone marrow does not affect the lung phenotype of endFoxf1−/− mice.

Fig. S7. Deletion of Foxf1 from endothelial cells reduces the abundance of VE-cadherin and S1PR1 in lung tissue.

Fig. S8. Experimental protocol for the administration of S1P to BHT-treated heterozygous endFoxf1+/− mice.

Fig. S9. S1P does not prevent mortality in Tam-treated homozygous endFoxf1−/− mice.

Fig. S10. CYM5442 decreases lung inflammation in LPS-treated endFoxf1+/− mice.

Table S1. RNA-seq analysis of FOXF1-deficient lungs.

Table S2. TaqMan primers for qPCR.

Table S3. Primers used for the ChIP and dual-luciferase assays.

REFERENCES AND NOTES

Acknowledgments: We thank Y. Zhang for excellent technical assistance, N. Timchenko for helpful comments, M. Fruttiger for providing Pdgfb-iCreER mice, and L. Fei for assistance with statistical analyses. Funding: This work was supported by NIH grants HL84151 (V.V.K.), HL123490 (V.V.K.), and CA142724 (T.V.K.). Author contributions: Y.C. designed, performed, and analyzed the experiments; C.B., T.L., C.G., and Y.X. performed and analyzed the experiments; T.V.K. and V.V.K. conceived the research, analyzed the data, and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The RNA-seq data were deposited in the National Center for Biotechnology Information Gene Expression Omnibus database (accession number GSE78184) and can be accessed at www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=ahuxauewhhupxev&acc=GSE78184.
View Abstract

Navigate This Article