Lung endothelial cells regulate pulmonary fibrosis through FOXF1/R-Ras signaling – Nature Communications

Identification of transcription factors that regulate re-programming of normal EC into fibrosis-associated EC in IPF lungs

To identify transcriptional regulators that control re-programming of normal EC into fibrosis-associated EC, we performed single cell RNA-sequencing (10X genomics) to compare the transcriptome of endothelial cells in donor and idiopathic pulmonary fibrosis (IPF) lungs (Cincinnati Children’s Hospital Medical Center (CCHMC) datasets, Supplementary Fig. S1a, b). We also downloaded and analyzed the published scRNA-seq data¹⁹ from donor and IPF lungs (Northwestern University (NW) datasets, Supplementary Fig. S1c, d). Endothelial cell in both datasets were identified as cells expressing Pecam1 (CD31), Cdh5 (VE cadherin) and lacking Ptprc (CD45) mRNAs, consistent with published studies²⁰. Cross-comparison of two scRNA-seq datasets identified several top differentially expressed endothelial transcription factors. Among universally downregulated transcription factors in IPF endothelial cells were FOXF1, SMAD6, HIF3A, CREB5, TBX3 and GATA2 (Fig. 1a). Universally upregulated transcription factors included ETV6, LEF1, EGR3, MEOX1 and SMAD1 (Fig. 1a). For the follow-up studies, we decided to focus on FOXF1 transcription factor because it’s expression is highly enriched in lung endothelial cells compared to endothelial cells of other organs^18,21 and because the FOXF1 loss-of-function gene mutations are linked to ACDMPV, a fatal congenital lung disease with significant fibrotic remodeling¹³. To verify that identified decrease in FOXF1 is not limited to two datasets, we have also analyzed the publicly available datasets from Yale University and Vanderbilt University^6,7. We have confirmed the decreased levels of FOXF1 mRNA in endothelial cells of patients with IPF compared to donor controls (Supplementary Fig. S2a). We sought to determine whether downregulation of FOXF1 in fibrosis-associated EC is essential for lung fibrogenesis.

Expression of FOXF1 is decreased in endothelial cells within fibrotic lesions of human IPF lungs

Using immunostaining with FOXF1, CD31 and αSMA antibodies, we found that FOXF1 protein was undetectable in EC within fibrotic lesions of IPF lungs identified as αSMA-positive regions (Fig. 1b and Supplementary Fig. S2b). While in human donor lungs, FOXF1 protein was detected in approximately 80% of endothelial cells, only 5-10% of endothelial cells expressed FOXF1 in IPF lungs (Fig. 1c). We further verified by qRT-PCR that FOXF1 mRNA was decreased in FACS-sorted EC from IPF lungs compared to FACS-sorted EC from donor lungs (Fig. 1d). Thus, both mRNA and protein levels of FOXF1 are decreased in endothelial cells of IPF lungs.

To identify FOXF1-expressing lung endothelial cells in scRNA-seq CCHMC dataset, endothelial cells from donor and IPF lungs were visualized using uniform manifold approximation and projection (UMAP) after samples integration with Harmony²² (Fig. 1e–g). Both FOXF1 mRNA levels and the total number of FOXF1-expressing EC were decreased in IPF lungs compared to donor lungs (Fig. 1h). To demonstrate that these results are not limited to CCHMC scRNA-seq dataset, we analyzed the publicly available NW scRNA-seq dataset containing endothelial cells from IPF and donor lungs (GSE 122960¹⁹). Consistent with the results from CCHMC dataset, both the percentage of FOXF1-positive endothelial cells and FOXF1 mRNA levels were decreased in IPF lungs from NW dataset (Fig. 1i–l).

Based on gene expression signatures, pulmonary endothelial cells from both CCHMC and NW datasets were subdivided into five sub-clusters: arterial, general capillary (gCAP), alveolar capillary (aCAP), venous and lymphatic ECs (Fig. 1e, i, Supplementary Fig. S1a–d). Genes known to be selectively expressed in different sub-types of endothelial cells were used to annotate these sub-clusters^23,24. The arterial EC sub-cluster was identified as cells expressing EFNB2, GJA5, DKK2 and HEY1, whereas the venous EC sub-cluster included cells expressing ACKR1, CLU and VWF (Supplementary Fig. S1a, b). The gCAP sub-cluster was enriched in GPIHBP1 and SLC6A4 transcripts, aCAP was enriched in EDNRB, APLN and HPGD, and lymphatic EC sub-cluster expressed PROX1, CCL21 and PDPN (Supplementary Fig. S1a–d). FOXF1 was detected in arterial, venous, and capillary sub-clusters, with the highest expression of FOXF1 in aCAPs and gCAPs (Fig. 1m, n). FOXF1 was undetectable in lymphatic EC (Fig. 1m, n). Also, FOXF1 was not expressed in the COL15+ endothelial cells that were identified in the IPF lungs, but not in the control donor lungs (Supplementary Fig. S2c), confirming the published data⁶. In both scRNA-seq datasets, FOXF1 expression was decreased in capillary, arterial and venous sub-clusters of IPF lungs compared to donor lungs (Fig. 1m–o). Consistent with scRNA-seq data, RNA in situ hybridization showed that both the percentage of FOXF1-positive EC and expression levels of FOXF1 transcripts were decreased in capillary, arterial and venous sub-clusters of IPF lungs compared to donor lungs (Supplementary Fig. S3).

FOXF1 is decreased in endothelial cells within fibrotic lesions of mouse bleomycin-injured lungs

We next examined expression of FOXF1 in endothelial cells in a mouse model of pulmonary fibrosis. Three weekly intratracheal (IT) injections of bleomycin were used to induce lung fibrosis in wild type mice (Supplementary Fig. S4a). Bleomycin administration caused a time-dependent increase in collagen depositions as quantified by Sircol assay (Fig. 2a) and confirmed by H&E and Sirius red/Fast green staining (Supplementary Fig. S4b). Increased accumulation of inflammatory cells in bleomycin-treated lungs was shown using flow cytometry analysis for CD45⁺ cells (Supplementary Fig. S4c). Next, we FACS-sorted CD31⁺/CD45⁻ lung endothelial cells at different time points after bleomycin treatment and measured endothelial Foxf1 expression by qRT-PCR. A decrease in endothelial Foxf1 mRNA was observed as early as day 3 after the first bleomycin treatment (Fig. 2b). The percentage of FOXF1-positive ECs in lung tissue was also decreased as early as day 3 after the first bleomycin treatment as shown by co-localization of FOXF1 with endothelial-specific ERG transcription factor (Supplementary Fig. S4d). Immunostaining for FOXF1 and CD31 showed decreased FOXF1 protein expression in endothelial cells within fibrotic lesions identified as αSMA-positive areas (Fig. 2d, right panel). Consistent with human IPF lungs (Fig. 1b, c), the number of FOXF1⁺/CD31⁺ endothelial cells was decreased in murine fibrotic lungs (Fig. 2c).

a Time-dependent accumulation of collagen in murine lungs after chronic bleomycin injury is quantified using Sircol collagen assay. Wild type mice were treated with three weekly IT injections of bleomycin to induce lung fibrosis (Day0, 3, 6, 10, 17, n = 4 mice per group; Day 13, n = 6 mice). b Time-dependent decrease of Foxf1 mRNA is shown in FACS-sorted lung endothelial cells during lung fibrogenesis using qRT-PCR (Day 0, 6,10, n = 3; Day 3, n = 4; Day13 n = 7; Day 17, n = 8, mice per group). c Co-localization studies show decreased FOXF1 in endothelial cells and decreased number of FOXF1⁺ endothelial cells within lung fibrotic foci at day 21 after bleomycin treatment. Mouse normal lungs (n = 5) and bleomycin-treated lungs (n = 5) were stained with antibodies against FOXF1 (red), CD31 (white) and αSMA (green). DAPI (blue) was used to visualize the nuclei. Bar = 25 µm. d Percent of FOXF1⁺/CD31⁺ double positive cells among CD31⁺ cells were counted in 5 random fields and presented as mean ± SD (n = 5 mice per group). ***p < 0.001. The integrated projection of lung EC from normal and bleomycin-treated fibrotic lungs (e, colored by cell type. f, yellow- normal lungs; blue- fibrotic lungs). Single cell RNA-seq was performed using pooled normal controls (n = 4) and bleomycin-treated (n = 6) mouse lungs. g UMAP plots show Foxf1 expression in endothelial cells of normal and fibrotic mouse lungs after Z-score normalization. h Violin plots show decreased expression of Foxf1 mRNA in ECs from fibrotic lungs. i Both Foxf1 mRNA expression and the frequency of Foxf1-positive cells are decreased in venous, aCap, gCap, and arterial endothelial cells from fibrotic lungs. Foxf1 is not expressed in lymphatic endothelial cells. Foxf1 expression is log normalized. Data presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. For two group comparisons, T-test (two-tailed) analyses were performed. For more than two group, statistical significance was determined by one-way ANOVA followed by Dunnett’s test. Source data are provided as a Source Data file.

Next, we performed a single cell RNA-sequencing of bleomycin-treated and control mouse lungs. Foxf1-expressing endothelial cells were visualized using UMAP (Fig. 2e–g). Consistent with human IPF lungs, bleomycin-treated mouse lungs had reduced numbers of Foxf1-positive endothelial cells and decreased expression of Foxf1 mRNA in these cells (Fig. 2h). Mouse lung endothelial cells were further subdivided into arterial, venous, aCAP, gCAP and lymphatic sub-clusters (Supplementary Fig. S5 and Fig. 2e). Endothelial cells expressing Foxf1 transcript were present in arterial, venous, aCAP and gCAP sub-clusters, but were completely absent in lymphatic cells (Fig. 2i). RNA in situ hybridization showed that both the percent of FOXF1-positive endothelial cells as well as the expression levels of FOXF1 transcripts were decreased in arterial, venous, aCAP and gCAP sub-clusters of bleomycin-treated lungs compared to control lungs (Supplementary Fig. S6). Altogether, FOXF1 expression is decreased in most EC types in human and mouse fibrotic lungs.

Deletion of FOXF1 in endothelial cells accelerates pulmonary fibrosis

To determine the role of FOXF1 in endothelial cells during pulmonary fibrosis, we used mice in which the Foxf1 gene was specifically deleted in endothelial cells using Pdgfb-CreER transgene, which does not target other cell types in the adult lung^18,25. Since homozygous (Pdgfb-CreER /Foxf1^fl/fl mice or endFoxf1^−/−) mice developed lung edema and respiratory insufficiency¹⁸, we used heterozygous endFoxf1^+/− mice for lung fibrosis studies (Supplementary Fig. S7a). Naïve heterozygous Tam-treated endFoxf1^+/− mice were phenotypically normal, had normal lung histology and alveolar capillary density as shown by CD31 staining (Supplementary Fig. S7b and¹⁸). Since the Pdgfb-Cre transgene contain GFP¹⁸, we used flow cytometry to demonstrate that the transgene targets only CD45⁻CD31⁺ (endothelial), but CD45⁺CD31⁻ (hematopoietic) or CD45⁻CD31⁻ (non-endothelial, non-hematopoietic) cell types in the lung (Supplementary Fig. S7c, d). The scRNA-seq analysis showed co-expression of endothelial Foxf1 and Pdgfb mRNA transcripts in the lung EC (Supplementary Fig. S7f).

Three weekly IT injections of bleomycin were used to induce lung fibrosis in Tam-treated endFoxf1^+/− and control Foxf1^fl/fl mice (Fig. 3a). Using FACS-sorted lung endothelial cells, we have shown that bleomycin treatment caused more profound decrease of Foxf1 mRNA in endFoxf1^+/− endothelial cells compared to control endothelial cells (Supplementary Fig. S8a). Decreased Foxf1 in endothelial cells was associated with more severe fibrosis in bleomycin-treated endFoxf1^+/− mice compared to bleomycin-treated control mice, as shown by quantifying collagen depositions in the lung tissue using Sircol assay (Fig. 3b). Compared to controls, bleomycin-treated endFoxf1^+/− mice had increased Ashcroft scores (Supplementary Fig. S8b, c), decreased body weights (Fig. 3c), reduced lung capacity and compliance (Fig. 3d, e), impaired lung mechanics (Supplementary Fig. S8d–j) and arterial oxygenation (Supplementary Fig. S8k). Lung fibrosis in endFoxf1^+/− mice developed faster and was more severe compared to control mice, as shown by Sirius red/fast green (Fig. 3f), Trichrome (Fig. 3g), and immunostaining for αSMA (Fig. 3h). Increased fibrotic depositions in bleomycin-treated endFoxf1^+/− lungs were also supported by increased Acta2, Col1a1, Col3a1, Ctgf, Cthrc1, Vim and Fn1 mRNAs (Fig. 3i). Thus, FOXF1 deficiency in murine endothelial cells exacerbates bleomycin-induced pulmonary fibrosis.

a Schematic diagram of bleomycin administration to induce lung fibrosis and tamoxifen (TAM) administration to delete Foxf1 in endFoxf1^+/− and control Foxf1^+/− mice. b Lung collagen was quantified at 21 days after bleomycin administration by Sircol assay (n = 4 mice per group). Data presented as mean ± SD. *p < 0.05. Mann–Whitney Two-tailed test were performed. c Increased body weight loss in endFoxf1^+/− mice is shown at different time points compared to control mice. n = 6 per group. d, e Decreased lung capacity and lung compliance in endFoxf1^+/− (n = 5) mice at day 21 after bleomycin administration shown using FlexiVent. Control, n = 7. ***p = 0.0004 (d). **p = 0.0031 (e). f–h Increased severity of lung fibrosis in endFoxf1^+/− mice is shown using (f) Sirius red/Fast green at Days 14 and 21 after bleomycin treatment (untreated control n = 5 and endFoxf1^+/− n = 4; bleomycin treated Day 14 control n = 6 and endFoxf1^+/− n = 5; Day 21 control n = 6 and endFoxf1^+/− n = 7) and (g) Trichrome staining (untreated control n = 6 and endFoxf1^+/− n = 8; Day 14 control n = 3 and endFoxf1^+/− n = 8; Day 21 control n = 4 and endFoxf1^+/− n = 5), as well as immunostaining for αSMA (green) (untreated control, n = 6; untreated endFoxf1^+/− n = 3; Day 14 control, n = 6; Day 14 endFoxf1^+/−, n = 4; Day 21 control, n = 3; Day 21 endFoxf1^+/−, n = 7) (h)^. Bar = 100 µm. Sirius Red binds to all types of collagens, whereas fast green stains non-collagenous proteins. i Endothelial deletion of Foxf1 increases mRNA levels of collagen genes in total lung RNA as shown by qRT-PCR. Total lung mRNA was extracted from control and bleomycin-treated endFoxf1^+/−mice on day 21. Actb mRNA was used for normalization. Control Acta2, n = 7; Control Col1a1, n = 6; Control Col3a, n = 5; Control Ctgf, n = 6; Control Cthrc1, n = 6; Control Vim, n = 7; Control Fn1, n = 6; endFoxf1^+/− Acta2, n = 5; endFoxf1^+/− Col1a1, n = 4; endFoxf1^+/− Col3a, n = 3; endFoxf1^+/− Ctgf, n = 4; endFoxf1^+/− Cthrc1, n = 4; endFoxf1^+/− Vim, n = 5; endFoxf1^+/− Fn1, n = 5. Data presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. T-test (two-tailed) analyses were performed. Source data are provided as a Source Data file.

FOXF1-deficient endothelial cells increase myofibroblast activation

To identify molecular mechanisms by which FOXF1-deficient ECs promote lung fibrogenesis, we performed bulk RNA-seq using FACS-sorted endothelial cells from bleomycin-treated control and endFoxf1^+/− lungs (Supplementary Fig. S9). Gene Set Enrichment Analysis (GSEA) of RNA-seq data showed that the most enriched functional categories in endFoxf1^+/− endothelial cell were wound healing, extracellular matrix organization, cell migration, blood vessel remodeling, inflammation, myeloid leukocyte migration, and RHO GTPase signaling (Supplementary Fig. S9a–d). Consistent with inactivation of one Foxf1 allele, Foxf1 mRNA was decreased 2.1-fold in ECs of bleomycin-treated endFoxf1^+/− lungs compared to control lungs (Supplementary Fig. S9a), a finding confirmed by qRT-PCR of FACS-sorted ECs (Fig. 4a). FOXF1-deficiency in endFoxf1^+/− ECs was associated with increased expression of pro-fibrotic and pro-inflammatory genes, including Il6, Tnfα, Ccl2, Cxcl1 and Thbs1 (Supplementary Fig. S9a), findings validated by qRT-PCR (Fig. 4a).

a qRT-PCR shows increased expression of fibrosis-associated genes in FACS-sorted endothelial cells of endFoxf1^+/− lungs at day 21 after bleomycin administration. Actb mRNA was used for normalization. n = 3 mice per group. b Efficient inhibition of FOXF1 expression in shFOXF1-transfected HUVEC cells is shown with qRT-PCR. ACTB mRNA was used for normalization. n = 9 samples per group. c CM from FOXF1-deficient HUVECs increases fibroblast proliferation. CCD-19Lu fibroblasts were cultured in the presence of CM from control or FOXF1-deficient HUVECs. n = 3 samples per group. Day 3, **p = 00290; Day 4 ***p < 0.0001 by two-way ANOVA test. d Conditioned media (CM) from FOXF1-deficient HUVECs increases invasion of cultured CCD-19Lu fibroblasts. Human CCD-19Lu fibroblasts were seeded on the insert of transwell chamber coated with matrigel in the presence of CM from scrambled control (Scr-CM, n = 10) or FOXF1-deficient (shFOXF1-CM, n = 9) HUVECs. Graph represents average numbers of invaded cells per field. Bar = 200 μm. **p = 0.0019. e CCD-19Lu fibroblasts cultured in CM from FOXF1-deficient HUVECs had increased expression of pro-fibrotic genes compared to fibroblasts cultured in CM from control HUVECs as shown by qRT-PCR. HUVEC-Scr-CM ACTA2, n = 6; HUVEC-Scr-CM VIM, n = 6; HUVEC-Scr-CM FN1, n = 3; HUVEC-Scr-CM COL3A1, n = 4; HUVEC-shFOXF1-CM ACTA2, n = 5; HUVEC-shFOXF1-CM VIM, n = 6; HUVEC-shFOXF1-CM FN1, n = 3; HUVEC-shFOXF1-CM COL3A1, n = 5. f CM from FOXF1-deficient HUVECs had increased levels of pro-inflammatory mediators as determined by Proteome Profiler Human Cytokine Array (n = 2). g Inhibition of IL-6 and TNFα using blocking antibodies attenuated CCD-19Lu fibroblast invasion in the presence of CM from FOXF1-deficient HUVECs. Bar = 200 μm. n = 6 samples per group. Data presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001. CM conditioned medium. For two group comparisons, T-test (two-tailed) analyses were performed. Source data are provided as a Source Data file.

Next, we determined the effect of FOXF1 deficiency on secretion of profibrotic and proinflammatory mediators by endothelial cells in vitro. Human endothelial cells, HUVEC, were infected with lentiviruses carrying either FOXF1 shRNA or control shRNA. The shFOXF1-treated HUVECs exhibited decreased FOXF1 mRNA as shown by qRT-PCR (Fig. 4b), demonstrating approximately similar efficiency of FOXF1 inhibition compared to haploinsufficient endothelial cells from endFoxf1^+/− mice (Fig. 4a). Conditioned media (CM) from control and FOXF1-deficient HUVECs were collected. Invasion and proliferation of human CCD-19Lu lung fibroblasts were measured in the presence of CM in vitro. CM from FOXF1-deficient HUVECs increased invasion and proliferation of lung fibroblasts compared to CM from control cells (Fig. 4c, d). CM from FOXF1-deficient HUVECs also increased expression of ACT2, VIM, FN1 and COL3A1 in human CCD-19Lu fibroblasts (Fig. 4e). To verify that the results are not limited to HUVECs, we also used human pulmonary arterial endothelial cells (HPAEC) and human pulmonary microvascular endothelial cells (HPMEC) to confirm that CM from FOXF1-deficient HPAEC and HPMEC cells increased invasion, proliferation as well as expression of pro-fibrotic genes in lung fibroblasts (Supplementary Figs. S10a–d and S11a–d). Thus, FOXF1-deficient endothelial cells induced the pro-fibrotic phenotype in lung fibroblasts, possibly, through secretion of soluble mediators.

To identify the differentially changed soluble mediators in the CM from FOXF1-deficient HUVECs compared to control HUVECs, we used the Proteome Profiler Human Cytokine Array. CCL2, CXCL1, G-CSF, GM-SCF, ICAM-1 and IL-6 proteins were increased in CM from FOXF1-deficient HUVECs, whereas CXCL12, IL-8, IL-13 and IL-16 were unchanged (Fig. 4f). Since TNFα and IL-6 increase activation of fibroblasts during lung fibrosis²⁶, we used blocking antibodies to inhibit increased levels of IL-6 or TNFα in CM from FOXF1-deficient HUVECs (Fig. 4g). Inhibition of either IL-6 or TNFα significantly decreased migration of CCL-19LU fibroblasts into the bottom chamber of Transwells containing CM from FOXF1-deficient HUVECs (Fig. 4g). To verify that the results are not limited to HUVEC cells, we also used HPAEC cells and confirmed that inhibition of either IL-6 or TNFα significantly decreased migration of CCL-19LU fibroblasts into the bottom chamber of Transwells containing CM from FOXF1-deficient HPAEC (Supplementary Fig. S10e, f). Thus, FOXF1-deficient endothelial cells stimulate the migration of lung fibroblasts in vitro by secreting IL-6 and TNFα.

FOXF1 deficiency in endothelial cells increases the number of macrophages in the fibrotic lungs

Since numerous cytokines from the protein array, including IL-6, TNFα, CCL2 and CXCL1, regulate recruitment of macrophages during lung injury²⁶, we examined the number of macrophages in Foxf1-deficient lungs. At day 21 after bleomycin injury, accumulation of macrophages in fibrotic lesions of bleomycin-treated endFoxf1^+/− lungs was increased compared to control lungs as shown by immunostaining for F4/80 (Fig. 5a) and Mac3 (Supplementary Fig. S12a). We also used flow cytometry analysis to demonstrate that the number of macrophages (CD45⁺ CD11c^low/+ CD64⁺) was increased in bleomycin-injured endFoxf1^+/− lungs compared to control bleomycin-injured lungs (Fig. 5b and Supplementary Fig. S12b–c). Of note, no changes in the number of inflammatory cells, including macrophages, were reported in uninjured endFoxf1^+/− and control lungs¹⁸. We next determined whether increased secretion of pro-inflammatory mediators by FOXF1-deficient endothelial cells is important for macrophage migration in vitro. CM from control or FOXF1-deficient HUVECs was added to the bottom chambers of Transwells, and the migration of human macrophages from the upper chambers towards CM was assessed in the presence of blocking antibodies specific to CCL2, CXCL1, IL-6 or TNFα (Fig. 5c). Inhibition of these pro-inflammatory cytokines decreased migration of macrophages towards CM from FOXF1-deficient HUVECs compared to CM from control HUVECs (Fig. 5c and Supplementary Fig. S12d). Thus, FOXF1-deficient endothelial cells stimulate migration of macrophages by secreting multiple proinflammatory mediators.

a Increased number of macrophages in the lungs of endFoxf1^+/− (n = 11) mice compared to control (n = 6) mice at day 21 after bleomycin administration is shown using immunostaining for F4/80 (red) and αSMA (green) (control, n = 6 mice). Nuclei are counterstained with DAPI (blue). Bar = 100 μm. Average numbers of F4/80-positive cells were quantified using 10 random microscope fields per lung and presented as mean ± SD. **p = 0.0013. b Flow cytometry analysis shows increased percentage of macrophages in the lungs of bleomycin-treated endFoxf1^+/− mice compared to control mice. Macrophages were identified as CD45⁺ CD64⁺ CD11c^low/+ (Untreated Control, n = 5; untreated endFoxf1^+/−, n = 6; bleomycin-treated Control and endFoxf1^+/−, n = 4). Data presented as mean ± SD. c Inhibition of CCL2, CXCL1, IL-6 and TNFα in CM from FOXF1-dificient HUVECs (shFOXF1-CM) using blocking antibodies attenuated microphage invasion in transwell assay. Human macrophages were incubated in the presence of CM from scrambled control (Scr-CM) or shFOXF1-transfected (shFOXF1-CM) HUVECs. Invaded cells were counted in 5 random microscope fields and presented as mean ± SD. n = 5 samples per group. *p < 0.05, **p < 0.01, ***p < 0.001 by Student’s T test (two-tailed). CM condition medium. Source data are provided as a Source Data file.

R-Ras is a direct transcriptional target of FOXF1

FACS-sorted endothelial cells from bleomycin-treated endFoxf1^+/− lungs exhibited decreased expression of Rras (Fig. 4a), a critical mediator of endothelial barrier function and vascular repair after injury^27,28. Consistent with decreased expression of Rras in endothelial cells, bleomycin-treated endFoxf1^+/− lungs displayed increased endothelial permeability as determined by Evans blue dye (Supplementary Fig. S13a). Since increased endothelial permeability contributes to lung fibrosis²⁹ and endFoxf1^+/− mice had exacerbated lung fibrosis after bleomycin injury (Fig. 3), we examined whether FOXF1 regulates R-RAS in pulmonary endothelial cells. Based on scRNA-seq data analysis, R-RAS mRNA was decreased in human IPF endothelial cells compared to donor endothelial cells (Fig. 6a). We also FACS-sorted endothelial cells from IPF lungs and used qRT-PCR to demonstrate that R-RAS mRNA was decreased in IPF endothelial cells, coinciding with decreased expression of FOXF1 mRNA (Fig. 6b). Immunostaining for R-RAS, FOXF1 and CD31 showed that FOXF1 co-localized with R-RAS in endothelial cells of donor lungs, indicating that both proteins are co-expressed in normal ECs (Fig. 6c). In contrast, neither R-RAS nor FOXF1 were detected in endothelial cells within fibrotic lesions of IPF lungs (Fig. 6c). In vitro shRNA-mediated knockdown of FOXF1 in HUVECs decreased expression of R-RAS (Fig. 6d). These findings were confirmed in HPAEC and HPMEC cells (Supplementary Fig. S14a, b). In agreement with human data, shRNA-mediated knockdown of Foxf1 in mouse MFLM 91U endothelial cells also decreased expression of Rras (Fig. 6e). Next, we used publicly available ChIP-seq dataset²⁵ to show that FOXF1 protein directly bound to the Rras promoter region in endothelial cells (Fig. 6f). The FOXF1-binding region in Rras promoter had H3K4me3 but not H3K27me3 marks (Fig. 6f), suggesting that FOXF1 transcriptionally activates Rras gene promoter. To test this hypothesis, the −762/+13 bp Rras promoter region, containing the FOXF1-binding site identified by ChIP-seq (Fig. 6f), was cloned into the pGL2 luciferase reporter plasmid (Fig. 6g, upper schematic diagram). In co-transfection experiments, CMV-Foxf1 expression vector increased transcriptional activity of the −762/+13 Rras promoter region compared to CMV-empty vector (Fig. 6g). Thus, Rras is a direct transcriptional target of FOXF1 in endothelial cells.

a Violin plots show decreased R-RAS mRNA in EC of IPF (n = 4 lungs) compared to donor (n = 8 lungs), GSE 122960 datasets. R-RAS expression was log normalized. b R-RAS mRNA was decreased in FACS-sorted ECs from IPF lungs (n = 6) compared to donor lungs (n = 3) as shown by qRT-PCR. Data presented as mean ± SD, *p = 0.238, Mann–Whitney Two-tailed test. c Immunostaining for R-RAS, FOXF1 and CD31 shows co-localization of FOXF1 and R-RAS in ECs of donor lungs (n = 5). Neither R-RAS nor FOXF1 are detected in ECs within IPF fibrotic lesions (n = 5). Nuclei are counterstained with DAPI. Bar = 20 μm. d qRT-PCR shows that shRNA-mediated knockdown of FOXF1 (shFOXF1) in HUVECs decreased R-RAS mRNA compared to control (Scr). ACTB mRNA was used for normalization (n = 3), **p < 0.01 and ***p < 0.001 by Student’s T test (two-tailed). e qRT-PCR shows that siRNA-mediated knockdown of Foxf1 in mouse MFLM-91U ECs decreased Rras mRNA. Actb mRNA was used for normalization (n = 3). f ChIP-seq shows direct binding of FOXF1 protein to Rras promoter region in MFLM-91U cells. Binding of FOXF1 to Rras promoter is associated with H3K4me3 marks but not H3K27me3 marks. g Schematic drawing of the pGL2-Rras-Luc construct with the −762/+13 bp Rras promoter region containing the FOXF1-binding site (top panel). In co-transfection experiments, CMV-FOXF1 expression vector increased transcriptional activity of the −762/+13 Rras promoter region compared to CMV-empty vector (bottom panel), n = 3, ***p < 0.001. h Overexpression of R-Ras decreased Ccl2 and TNFα mRNAs in mock-transfected cells and prevented upregulation of Ccl2 and TNFα in cells transfected with Foxf1-specific siRNA. MFLM-91U cells were transfected with non-targeting siRNA (siControl) or siFoxf1, and EEV-empty vector or EEV-Rras. Actb mRNA was used for normalization (siControl+EEV-empty Foxf1, n = 6; siControl+EEV-empty Rras, n = 6; siControl+EEV-empty Ccl2, n = 3; siControl+EEV-empty Tnf, n = 6; siControl+EEV-Rras Foxf1, n = 6; siControl+EEV-Rras Rras, n = 6; siControl+EEV-Rras Ccl2, n = 6; siControl+EEV-Rras Tnf, n = 3; siFoxf1+EEV-empty Foxf1, n = 6; siFoxf1+EEV-empty Rras, n = 3; siFoxf1+EEV-empty Ccl2, n = 3; siFoxf1+EEV-empty Tnf, n = 3; siFoxf1+EEV-Rras Foxf1, n = 6; siFoxf1l+EEV-Rras Rras, n = 3; siFoxf1l+EEV-Rras Ccl2, n = 6; siFoxf1+EEV-Rras Tnf, n = 3), *p < 0.05, **p < 0.01, ***p < 0.001 by Student’s t test (two tailed). Source data are provided as a Source Data file.

Next, we examined whether R-RAS is involved in FOXF1-mediated regulation of CCL2, CXCL1, IL-6 and TNFα. We over-expressed R-Ras in FOXF1-deficient MFLM 91U endothelial cells in vitro (Fig. 6h and Supplementary Fig. S14a, b). Overexpression of R-Ras decreased Ccl2 and TNFα mRNAs in mock-transfected cells and prevented upregulation of Ccl2 and TNFα in cells transfected with Foxf1-specific siRNA (Fig. 6h). Overexpression of R-Ras did not prevent upregulation of Cxcl1 and IL-6 in FOXF1-deficient endothelial cells (Supplementary Fig. S14c). Altogether, our data suggest that FOXF1 in lung endothelium stimulates transcription of Rras, which inhibits expression of CCL2 and TNFα. FOXF1 inhibits CXCL1 and IL-6 independently of R-Ras.

Transgenic overexpression of FOXF1 in endothelial cells decreases lung fibrosis after bleomycin-induced injury

Since conditional deletion of FOXF1 in endothelial cells increased pulmonary fibrosis after chronic bleomycin injury (Fig. 3), we examined whether overexpression of FOXF1 in endothelial cells was sufficient to inhibit lung fibrosis. To test this hypothesis, we generated an inducible, EC-specific FOXF1 overexpression mouse model (Pdgfb-CreER^tg/+; LSL-rtTA^tg/+; TetO-Foxf1^tg/+ mice; abbreviated as endFoxf1^OE) (Supplementary Fig. S15a). Upon tamoxifen administration, Cre-mediated recombination of LoxP-floxed stop codon (LSL) results in rtTA expression in endothelial cells. In the presence of doxycycline, rtTA binds to and activates the TetO₇-CMV promoter driving expression of the mouse Foxf1 transgene (Supplementary Fig. S15a). Thus, combined administration of tamoxifen (Tam) and doxycycline (Dox) causes increased expression of FOXF1 in endothelial cells of endFoxf1^OE mice (Fig. 7a). FACS-sorted endothelial cells from Tam- and Dox-treated endFoxf1^OE lungs had a 4-fold increase in Foxf1 mRNA compared to controls (Fig. 7b). Without bleomycin injury, the Tam- and Dox-treated endFoxf1^OE mice had normal lung architecture and capillary density as shown by H&E staining and immunostaining for CD31 (PECAM-1) (Supplementary Fig. S15b). To induce pulmonary fibrosis, Tam- and Dox-treated endFoxf1^OE and single transgenic (control) mice were given three weekly IT injections of bleomycin (Fig. 7a). While bleomycin treatment decreased Foxf1 mRNA in FACS-sorted endothelial cells from both groups of mice, Foxf1 expression was higher in endFoxf1^OE ECs compared to control ECs (Fig. 7b). Transgenic overexpression of Foxf1 increased survival of mice after bleomycin injury with 95% of endFoxf1^OE mice surviving the injury, but only 20% of control single-transgenic littermates being alive 25 days after the last bleomycin treatment (Fig. 7c). Fibrotic lung remodeling was decreased in bleomycin-treated endFoxf1^OE mice compared to controls as demonstrated by Ashcroft score, Trichrome staining, Sirius red/Fast green staining and immunostaining for αSMA (Fig. 7d–h). Biochemical quantification of lung collagen amounts using Sircol assay was consistent with decreased fibrosis in endFoxf1^OE lungs (Fig. 7i). Moreover, endothelial over-expression of Foxf1 decreased recruitment of macrophages into bleomycin-treated lungs and decreased expression of pro-inflammatory genes in total lung RNA (Fig. 7j, k). Thus, overexpression of FOXF1 in endothelial cells prior to bleomycin injury attenuates pulmonary fibrosis and improves survival after bleomycin-induced lung injury.

a Schematic diagram shows bleomycin, tamoxifen (TAM) and doxycycline (DOX) treatments in endFoxf1^OE mice. b qRT-PCR shows that Foxf1 mRNA is increased in lung ECs isolated from untreated and bleomycin-treated endFoxf1^OE mice compared to control mice. Actb mRNA was used for normalization (Untreated Control, n = 6; Untreated endFoxf1^OE, n = 5; Bleomycin Control and endFoxf1^OE, n = 8 mice per group), ***p < 0.001. c Endothelial-specific overexpression of FOXF1 increases mouse survival after bleomycin injury (n = 10), **p = 0.001, Log-rank (Mantel–Cox) test. d Decreased fibrosis in lungs of bleomycin-treated endFoxf1^OE mice is shown by H&E, Trichrome, Sirius red/fast green staining and immunostaining for αSMA (green). Mouse endFoxf1^OE (n = 5) and control lungs (n = 5) were harvested at day 21 after bleomycin treatment. Bar = 100 μm. e–h Assessment of fibrotic lesions in endFoxf1^OE (n = 5) lungs is presented as Ashcroft score ***p < 0.0001 (e), the percent of Trichrome-positive **p = 0.0058 (f) Sirius red-positive. **p = 0.0011 (g), and αSMA-positive areas in the lungs (Control, n = 6), **p = 0.0013. Areas positive for collagen depositions were quantified in 10 random fields per lung using Nikon’s NIS-Elements AR software (h). i Lung collagen depositions were quantified by Sircol collagen assay. Left lung lobes from endFoxf1^OE (n = 4) and control mice (n = 4) were used. *p = 0.0286. t test (two tailed). j Decreased number of macrophages in the lungs of bleomycin-treated endFoxf1^OE mice (n = 5) is shown using immunofluorescent staining with anti-F4/80 antibodies (red) and anti-αSMA antibodies (green) at day 21 after bleomycin administration (Control, n = 6). Nuclei are counterstained with DAPI (blue). Bar = 100 μm. Average numbers of F4/80-positive cells were quantified using 10 random microscope fields per lung and presented as mean ± SD. **p = 0.0013. t test (two tailed). k Endothelial overexpression of Foxf1 decreases mRNA levels of pro-inflammatory genes in FACS sorted EC as shown by qRT-PCR. mRNA was extracted from control and bleomycin-treated endFoxf1^OE mice on day 21. Actb mRNA was used for normalization. Control, n = 4; endFoxf1^OE, n = 5. Data presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 by Student’s t test (two tailed). Source data are provided as a Source Data file.

Next, we assessed whether increasing endothelial FOXF1 in already established lung fibrosis will improve the outcomes. Overexpression of FOXF1 in endothelial cells at day 10 after bleomycin injury decreased body weight loss, improved mice survival, and decreased collagen depositions in endFoxf1^OE mice compared to bleomycin-injured control mice (Fig. 8a–i). Based on Flexivent measurements of lung mechanics, many functional parameters in endFoxf1^OE mice were improved, including increased lung compliance, total lung capacity and decreased tissue resistance (Supplementary Fig. S16a–h). Finally, overexpression of FOXF1 in endothelial cells at day 10 after bleomycin injury, decreased expression of pro-fibrotic genes in total lung RNA and inhibited recruitment of macrophages to the lung tissue (Fig. 8j, k). Altogether, overexpression of FOXF1 in endothelial cells either prior to or after bleomycin injury attenuates pulmonary fibrosis and improves mice survival after the injury.

a Schematic diagram shows bleomycin, tamoxifen (TAM) and doxycycline (DOX) treatment in endFoxf1^OE mice. b Endothelial-specific overexpression of FOXF1 decreases bodyweight loss in bleomycin-treated endFoxf1^OE mice (Control, n = 12; endFoxf1^OE, n = 12 for days 0 and 21, n = 4 for day 10, n = 6 for day 14), *p < 0.05, two-way ANOVA with the Geisser–Greenhouse correction. c Endothelial-specific overexpression of FOXF1 increases mouse survival after bleomycin injury (Control, n = 11; endFoxf1^OE, n = 17), *p < 0.05, Log-rank (Mantel–Cox) test. d Decreased collagen depositions in bleomycin-treated endFoxf1^OE lungs is shown by Hydroxyproline assay. Left lung lobes from endFoxf1^OE (n = 4) and control (n = 3) mice were used, **p = 0.0061, t test (two tailed). e Decreased fibrosis in bleomycin-treated endFoxf1^OE lungs is shown by H&E, Sirius red/fast green, Trichrome, and immunostaining for αSMA. Lungs were harvested at day 21 after bleomycin treatment (n = 8 per group). Bar = 100 µm. f–i Ashcroft score of endFoxf1^OE lungs (***p < 0.0001) is consistent with the percent of Sirius red-positive (***p < 0.0001), Trichrome-positive (**p = 0.0031) and αSMA-positive areas (***p < 0.0001). Areas positive for collagen depositions were quantified in 10 random fields per lung using Nikon NIE CIC Analysis Elements software. n = 8 mice per group. j Endothelial overexpression of Foxf1 at day 10 after bleomycin administration decreases mRNA levels of pro-fibrotic genes in total lung RNA as shown by qRT-PCR. Total lung mRNA was extracted from control and bleomycin-treated endFoxf1^OE mice on day 21. Actb mRNA was used for normalization (Control Col1a1 and Col3a1, n = 10; Control Vim, n = 9; Control Cthrc1, n = 9; Control Fn1, n = 8; endFoxf1^OE Col1a1, Col3a1, Vim and Cthrc1, n = 7; endFoxf1^OE Fn1, n = 8). k Decreased number of macrophages in the lungs of bleomycin-treated endFoxf1^OE mice is shown using immunofluorescent staining with anti-F4/80 antibodies (red) and anti-αSMA antibodies (green) at day 21 after bleomycin administration, ***p < 0.0001. Nuclei are counterstained with DAPI (blue). n = 8 mice per group. Data presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, Student’s t test (two tailed). Source data are provided as a Source Data file.

Nanoparticle delivery of non-integrating Foxf1 expression-vector into the lung endothelium attenuates pulmonary fibrosis

Since genetic overexpression of Foxf1 in endothelial cells attenuated pulmonary fibrosis, we next tested whether nanoparticle delivery of Foxf1 cDNA into endothelial cells can be a therapeutic approach to inhibit lung fibrosis. Mouse Foxf1 cDNA was cloned into non-integrating self-replicating episomal EEV vector to generate EEV-Foxf1 construct (Fig. 9a). Transfection of EEV-Foxf1 plasmid into Foxf1-negative HEK-293T cells increased the FOXF1 protein levels in vitro as shown by Western blot (Fig. 9b). To deliver EEV-Foxf1 plasmid to endothelial cells in vivo, we utilized the poly β-amino esters (PBAE) polymer^30,31, which formed stable nanoparticles in complex with plasmid DNA (Fig. 9c and Supplementary Fig. S17a, b). The hydrodynamic average diameter of the PBAE nanoparticles loaded with EEV-Foxf1 plasmid DNA (Nano-Foxf1) was 146.27 ± 9.54 nm (Supplementary Fig. S17c), whereas the average surface charge of the Nano-Foxf1 was 28.3 ± 1.71 mV (Supplementary Fig. S17d). In vitro treatment of HEK-293T cells with nanoparticles containing EEV-Foxf1 plasmid resulted in the efficient expression of red fluorescent protein (RFP) in vast majority of cells (Supplementary Fig. S17e). Next, fluorescently-labeled PBAE nanoparticles with either EEV-Foxf1 or EEV-Empty (control) plasmids were delivered to bleomycin-treated mice (Fig. 9d, e and Supplementary Fig. S17f, g) via tail vein on the same day as bleomycin administration. Using FACS analysis of enzymatically digested lung tissue, nanoparticles were detected in ~92% of lung endothelial cells (CD31⁺/CD45⁻) (Fig. 9e). Nanoparticles-mediated targeting of epithelial (CD326⁺/CD45⁻/CD31⁻), immune (CD45⁺/CD31⁻) and mesenchymal (CD31⁻/CD45⁻/CD326⁻) cells was ineffective (Fig. 9d, e). Nanoparticles were still detected in ~50–70% of pulmonary endothelial cells at day 28 after single nanoparticle delivery (Fig. 9e and Supplementary Fig. S17g). Administration of EEV-Foxf1-nanoparticles increased Foxf1 mRNA in FACS-sorted lung endothelial cells as demonstrated by qRT-PCR (Fig. 9f). Treatment with nanoparticles containing pEEV-Foxf1 vector on the same day as bleomycin injury significantly increased survival of mice as shown by Kaplan–Meier curve (Supplementary Fig. S18a) and decreased collagen depositions in the lung tissue as shown by H&E and Sirius red staining (Supplementary Fig. S18b).

a Diagram of the FOXF1 episomal plasmid (EEV-Foxf1). b Western blot shows increased FOXF1 protein after transfection of EEV-Foxf1 plasmid into FOXF1-negative HEK-293T cells. c Schematic of PBAE nanoparticles loaded with plasmid DNA. All elements of images were created using Autodesk 3ds Max 2020 (Autodesk, version 2020). d FACS analysis of mouse lungs show the presence of labeled nanoparticles in endothelial cells but not in other cell types at day 7 after I.V. administration (n = 3). e FACS analysis shows nanoparticles in lung ECs at different time-points after bleomycin injury (n = 5 mice per group). f Diagram shows nanoparticle I.V. delivery to mice at day 10 after bleomycin administration. g Kaplan–Meier analysis shows increased survival of bleomycin-injured mice after nanoparticle delivery of EEV-Foxf1 plasmid (Nano-Foxf1, n = 15) compared to EEV-Empty (Nano-Empty, n = 11), **p < 0.005, Log-rank (Mantel–Cox) test. h Sircol assay shows decreased collagen depositions in Nano-Foxf1-treated lungs (n = 4 mice per group), *p < 0.05, ***p < 0.001, Mann–Whitney Two-tailed test. i Foxf1 mRNA is increased in FACS–sorted lung ECs from Nano-Foxf1 treated mice (n = 7), *p < 0.05, **p < 0.005, Student’s t test (two tailed). j Decreased lung fibrosis in Nano-Foxf1-treated mice is shown with Sirius red/fast green, Masson’s trichrome, and immunostaining for αSMA. Decreased number of macrophages in Nano-Foxf1-treated lungs is shown by immunostaining for F4/80 and αSMA at day 21 after bleomycin administration (n = 6 mice per group). Bar = 100 μm. k–n Quantification of fibrotic lesions in mouse lungs is shown as the percent of Sirius red-positive (k), Trichrome-positive (l) and αSMA (m) areas. Areas positive for collagen depositions were quantified in 10 random microscope fields per lung (Nano-Empty, n = 7; Nano-Foxf1, n = 6). n Nanoparticle delivery of Foxf1 at day 10 after bleomycin administration decreases mRNA levels of pro-fibrotic genes in total lung RNA as shown by qRT-PCR. (Nano-Empty control, n = 8, and Nano-Foxf1-treated, n = 10, on day 21 after bleomycin administration. Actb mRNA was used for normalization. o Decreased Ashcroft score in nano-Foxf1 treated lungs (n = 10) compared to control nano-Empty (n = 8). Data presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 by Student’s t test (two tailed). Source data are provided as a Source Data file.

Next, we assessed the therapeutic efficacy of EEV-Foxf1-nanoparticles after lung injury. Mice were treated with EEV-Foxf1-nanoparticles at day 10 after bleomycin administration (Fig. 9f) when the fibrosis lung remodeling is already present (Fig. 2a). Nanoparticle delivery of Foxf1 into endothelial cells improved mice survival (Fig. 9g), prevented the loss of body weight after bleomycin injury (Supplementary Fig. S19a), and decreased lung collagen depositions quantified by Sircol assay (Fig. 9h). Decreased collagen amounts in EEV-Foxf1-treated lungs coincided with increased Foxf1 mRNA in FACS-sorted lung endothelial cells (Fig. 9i). Decreased fibrotic remodeling in EEV-Foxf1-treated lungs was detected and quantified based on Sirius red/Fast green staining, Trichrome staining, and immunostaining for αSMA (Fig. 9j–m). In addition, nanoparticle delivery of Foxf1 into endothelial cells at day 10 after bleomycin injury, reduced macrophage infiltration into fibrotic regions (Fig. 9j, right panels and Supplementary Fig. S19l), decreased expression of pro-fibrotic genes in total lung RNA (Fig. 9n), decreased Ashcroft score (Fig. 9o), improved lung compliance and total lung capacity in EEV-Foxf1-treated mice (Supplementary Fig. S19b–j) and arterial oxygenation (Supplementary Fig. S19k). Altogether, nanoparticle delivery of Foxf1 cDNA into endothelial cells on the day of bleomycin injury or during fibrotic stage after the injury inhibits pulmonary fibrosis and improves mice survival.