Pulmonary fibrosis requires cell-autonomous mesenchymal fibroblast growth factor (FGF) signaling

Idiopathic pulmonary fibrosis (IPF) is characterized by progressive pulmonary scarring, decline in lung function, and often results in death within 3–5 five years after diagnosis. Fibroblast growth factor (FGF) signaling has been implicated in the pathogenesis of IPF; however, the mechanism through which FGF signaling contributes to pulmonary fibrosis remains unclear. We hypothesized that FGF receptor (FGFR) signaling in fibroblasts is required for the fibrotic response to bleomycin. To test this, mice with mesenchyme-specific tamoxifen-inducible inactivation of FGF receptors 1, 2, and 3 (Col1α2-CreER; TCKO mice) were lineage labeled and administered intratracheal bleomycin. Lungs were collected for histologic analysis, whole lung RNA and protein, and dissociated for flow cytometry and FACS. Bleomycin-treated Col1α2-CreER; TCKO mice have decreased pulmonary fibrosis, collagen production, and fewer α-smooth muscle actin-positive (αSMA+) myofibroblasts compared with controls. Freshly isolated Col1α2-CreER; TCKO mesenchymal cells from bleomycin-treated mice have decreased collagen expression compared with wild type mesenchymal cells. Furthermore, lineage labeled FGFR-deficient fibroblasts have decreased enrichment in fibrotic areas and decreased proliferation. These data identify a cell autonomous requirement for mesenchymal FGFR signaling in the development of pulmonary fibrosis, and for the enrichment of the Col1α2-CreER-positive (Col1α2+) mesenchymal lineage in fibrotic tissue following bleomycin exposure. We conclude that mesenchymal FGF signaling is required for the development of pulmonary fibrosis, and that therapeutic strategies aimed directly at mesenchymal FGF signaling could be beneficial in the treatment of IPF.

Pulmonary fibrosis is characterized by progressive scarring of the lung, with abnormal production of extracellular matrix by activated fibroblasts, or myofibroblasts (1). In idiopathic pul-monary fibrosis (IPF), 2 the cause of fibroblast activation and matrix deposition is unknown, although it is suggested to be a result of aberrant healing from an injury of unknown origin (2). IPF affects ϳ50,000 patients annually in the United States, and the median survival is 3-5 years after diagnosis (1). Recently approved novel antifibrotic agents Nintedanib (Ofev) and Pirfenidone (Esbriet) decrease the rate of decline in lung function and improve survival in IPF (3,4); however, overall outcomes remain poor.
Fibroblast growth factors (FGFs) are implicated in the pathogenesis of pulmonary fibrosis. The receptor tyrosine kinase activity of FGF receptors (FGFRs) is inhibited by Nintedanib (5,6), which decreases bleomycin-induced fibrosis in mice (7,8). Inhibition of FGF signaling using a soluble ectodomain of FGFR2c decreased TGF-␤1 induced primary lung fibroblast proliferation and differentiation in vitro, as well as bleomycininduced fibrosis in vivo (9). This ectodomain inhibits multiple FGFs known to bind to the IIIc splice variant of FGFR2, including FGFs 1,2,4,6,8,9,17, and 18 (10). A recent study demonstrated altered FGFR1 and FGF1 expression in IPF lungs, and suggested that FGF signaling is critical for fibroblast migration in pulmonary fibrosis (11). Additionally, administration of a specific inhibitor of FGFR1 (NP603) inhibits carbon tetrachloride-induced hepatic fibrosis in rats (12). Although these studies have demonstrated the importance of FGF signaling in the pathogenesis of pulmonary fibrosis, the cellular mechanisms involved remain elusive. In particular, it is not known which cell type(s) are critical targets for FGFs in the pathogenesis of pulmonary fibrosis in vivo.
To test whether FGF signaling in lung mesenchyme and fibroblasts is required for the pathogenesis of pulmonary fibrosis in vivo, we have generated mice with a mesenchyme-specific inducible knock-out of FGF receptors 1, 2, and 3 using mice with tamoxifen-inducible Cre recombinase driven by the promoter for procollagen I␣2 (Col1␣2-CreER) (19 -21). Simultaneous deletion of multiple FGF receptors was performed due to overlapping receptor specificity of many FGFs, including FGF1, FGF2, and FGF9 (22), as well as simultaneous expression of multiple FGFRs in lung mesenchyme in vivo.
In this report we demonstrate that the Col1␣2ϩ lineage is enriched in fibrotic tissue following bleomycin treatment, and that cell-autonomous FGFR signaling is required for this lineage enrichment. Deletion of FGFRs in lung mesenchyme decreases collagen expression and the development of pulmonary fibrosis in response to bleomycin. Our data suggests that inhibition of FGFR signaling pathways in fibroblasts slows the enrichment of lung mesenchyme into injured areas, providing a mechanistic rationale for the use of FGF inhibitors to slow the progression of pulmonary fibrosis.
Fgfr1 is expressed in mesenchymal cells and fibroblasts (22). In IPF, expression of FGFRs 1-3 is increased in myofibroblasts (11). To determine the potential for redundant function of Fgfr2, Fgfr3, and Fgfr4 in mesenchymal cells and fibroblasts, we measured Fgfr expression in freshly isolated (non-cultured) Col1␣2ϩ mesenchymal cells. Analysis of Fgfr expression showed that Fgfr1 Col1␣2-CreER; ROSA26 mTmG/ϩ mice were treated with tamoxifen at P21 and lungs were collected for frozen sections and imaged with direct fluorescence microscopy (A and B). Arrows indicate primary sites of Col1␣2ϩ peribronchiolar, perivascular, and interstitial cells. Frozen sections were stained for PDGFR␣ (C), Periosin (D), ␣SMA (E), and pro-SPC (F), and microscopy was performed for co-localization with GFP. Tamoxifen-treated Col1␣2-CreER; ROSA26 mTmG/ϩ mice were given a single dose of intratracheal bleomycin, and lungs were analyzed after 14 (G) and 21 days (H) for GFP fluorescence. Co-immunofluorescence for ␣SMA and GFP (I) was performed on 21-day postbleomycin lung sections.
When Col1␣2ϩ mesenchymal cells were cultured on collagen-coated dishes, Fgfr1 remained the most abundant FGFR, however, expression of Fgfr2 and Fgfr3 was significantly diminished by passage number 2, and there was little measurable Fgfr4 expression (Fig. 2J). These data indicate that to completely inhibit FGF signaling in lung mesenchyme in vivo, deletion of multiple FGFRs is necessary, as they are capable of compensatory signaling if only Fgfr1 is deleted.
To test whether mesenchyme-specific deletion of FGFRs is required for bleomycin-induced expansion of myofibroblasts, lungs from Col1␣2-CreER; TCKO and Col1␣2-CreER control mice treated with intratracheal bleomycin were enzymatically

Fibroblast-specific FGFR signaling in pulmonary fibrosis
compared with controls, GFP immunohistochemistry in areas with similar amounts of fibrosis were compared. The Col1␣2ϩ lineage had decreased enrichment in areas of fibrosis following bleomycin in TCKO mice (Fig. 11, A-C). Additionally, the overall abundance of GFPϩ cells, as a percentage of total lung single cells detected using flow cytometry, was decreased in TCKO mice compared with controls (Fig. 11D). The overall abundance of GFPϩ cells did not significantly increase following bleomycin compared with PBS controls (not shown), consistent with prior reports (24). We then assessed proliferation in response to bleomycin in TCKO mice by EdU labeling prior to lung collection, enzymatic dissociation, and cell sorting. Col1␣2-CreER; TCKO mice have decreased numbers of EdUϩ cells within the overall single cell population and in the Lin-neg mesenchymal population (Fig. 11, E and F), demonstrating decreased proliferation in mesenchyme after bleomycininduced injury. We then measured growth in cultures of Lin-neg, GFPϩ cells isolated from Col1␣2-CreER; TCKO; ROSA26 mTmG/ϩ and control Col1␣2-CreER; ROSA26 mTmG/ϩ mice 21 days after bleomycin. Compared with controls, Col1␣2-CreER; TCKO; ROSA26 mTmG/ϩ mesenchymal cells demonstrate decreased growth over a 9-day period (Fig. 11G). This demonstrates that FGFR signaling is required for proliferation and expansion of a pre-existing mesenchymal population that contributes to bleomycin-induced pulmonary fibrosis.

Fibroblast-specific FGFR signaling in pulmonary fibrosis
tion. Furthermore, we showed that FGFR signaling is required for bleomycin-induced collagen expression in freshly isolated fibroblasts. Additionally, intact FGFR signaling in lung mesenchyme is required for lineage-specific expansion of fibrotic tissue in response to bleomycin.
FGF signaling has been implicated in the pathogenesis of pulmonary fibrosis, as non-selective inhibition of receptor tyrosine kinases (7) and nonspecific inhibition of FGF receptors (9, 25) decrease bleomycin-induced pulmonary fibrosis in rodents.
FGFR signaling (in addition to VEGFR and PDGFR) is inhibited by Nintedanib, which decreases the rate of decline in lung function in IPF patients (3,4). The mechanism by which inhibition of multiple receptor tyrosine kinases decreases the progression of pulmonary fibrosis remains unclear. Nintedanib reduces fibronectin and collagen expression, induces autophagy, and inhibits TGF␤ signaling in IPF lung fibroblasts (26). Our data demonstrates that there is a cell-autonomous role for FGFR signaling in pulmonary fibrosis that promotes expansion of

Fibroblast-specific FGFR signaling in pulmonary fibrosis
lung fibroblasts into injured areas within the lung. A recent report by MacKenzie et al. (11) described increased migration of fibroblasts in response to FGF1, further supporting our findings that FGF signaling is required for fibroblast expansion within fibrotic areas.
Our studies use bleomycin as a model agent to induce fibrosis. One percent of patients treated with bleomycin develop pulmonary fibrosis (27). In mice, bleomycin causes alveolar epithelial injury, inflammation, and fibroblast proliferation and differentiation into myofibroblasts. Although bleomycin does not cause lesions characteristic for IPF (fibroblastic foci and honeycombing) in the mouse lung, bleomycin remains a well established and useful model to study underlying mechanisms of alveolar epithelial cell injury and subsequent fibroblast activation and fibrosis (27,28,30,31).
Col1␣2-CreER mice have been used previously to inactivate genes in lung mesenchyme that are important for the development of pulmonary fibrosis, including Tert (32), smoothened (33), Tgf␤R1 (19), Notch1 (34), and CEBP␤ (20). It is important to note that this allele targets multiple mesenchymal populations, and it is difficult to assess which population within the Col1␣2ϩ lineage directly contributes to fibrotic areas in the lung. Furthermore, we have found that cultured lung fibroblasts generated from enzymatically dissociated Col1␣2-CreER; ROSA26 mTmG/ϩ whole lungs, without enrichment for GFPϩ cells via cell sorting, results in a heterogeneous population of cells, with only 50 -60% of the cells being GFPϩ (not shown). Incorporating the ROSA26 mT/mG Cre reporter allele with the Col1␣2-CreER allele allows for direct analysis of lineage labeled and genetically manipulated mesenchymal cells both in vivo and in vitro.
This study underscores the importance of studying fibroblasts in vivo, as we have found that Fgfr expression changes significantly once lung fibroblasts are placed in culture, even after a single passage. We demonstrate significant expression of all Fgfrs in freshly isolated lung mesenchyme; however, only Fgfr1 is expressed when isolated fibroblasts are cultured. Low levels of Fgfr3 and Fgfr4 expression in cultured lung fibroblasts has previously been reported (35). FGFRs 1-4 were also found to be expressed in myofibroblasts of fibroblastic foci in human IPF lung biopsies (11). The mechanism(s) responsible for this change in Fgfr expression is not known; however, altered cellu-

Fibroblast-specific FGFR signaling in pulmonary fibrosis
lar phenotypes and gene expression as a result of culture conditions, including substrate stiffness and coating of culture dishes have been well described (36).
Simultaneous deletion of multiple FGF receptors was performed due to 1) overlapping receptor specificity of many FGFs to FGFRs 1-3, including FGF1, FGF2, and FGF9 (22), and 2) simultaneous expression of multiple Fgfrs in lung mesenchyme in vivo (Fig. 2). Although mesenchymal cells with deletion of FGFRs 1-3 demonstrate a functional deficiency of FGF signaling, it remains possible that FGFR4 could further contribute to lung fibroblast pathophysiology. Targeting Fgfr4 in lung mesenchyme and fibroblasts is an area of future study. Additionally, the individual roles of FGFR1, FGFR2, and FGFR3 in lung mesenchyme following injury and in the development of pulmonary fibrosis is not known, and future studies will be needed to determine the relative contribution of individual FGFRs. Mice expressing a soluble dominant-negative FGFR2 in late gestation, but not postnatally, causes airspace enlargement and emphysema (37). FGFR3 may represent a significant FGFR in adult lung pathophysiology. FGFR3 and FGFR4 have been shown to be critical for early postnatal alveologenesis (35,38), and mice with combined deficiencies of Fgfr3 and Fgfr4 have excessive elastic fiber deposition; however, their importance in lung injury and pulmonary fibrosis are not well described. Recent reports have also implicated FGFR3 in the pathogenesis of lung cancer (39 -41).
Although mesenchyme-specific FGFR signaling is important for bleomycin-induced pulmonary fibrosis, the endogenous FGF ligand(s) that mediate this pathology remains unclear. FGF1 and FGF2 are capable of activating most FGF receptors, including FGFR1, the mesenchyme-specific FGFR2c spice variant, and FGFR3 (10,22). Inhibition of FGFR2c ligands by expressing a soluble FGFR2c ectodomain (25) decreases bleomycin-induced fibrosis, and whereas this implicates FGF signaling in the pathogenesis of fibrosis, it does not point to a single ligand, as FGFR2c interacts with multiple ligands, includ-

Fibroblast-specific FGFR signaling in pulmonary fibrosis
ing FGF1, FGF2, FGF8, FGF9, and FGF18 (10). We recently reported that Fgf2 knock-out mice have preserved fibrosis in response to bleomycin (42), which suggests that other FGF ligands are either required for bleomycin-induced fibrosis or compensate for the absence of FGF2. A combined knock-out of FGF1 and FGF2 has decreased carbon tetrachloride-induced hepatic fibrosis (43), but the response of these mice to intratracheal bleomycin is not known. FGF1 is increased in IPF lungs, and is capable of increasing fibroblast migration in vitro (11). FGF9 and FGF18 are increased in IPF, stimulate fibroblast migration, and decrease fibroblast apoptosis (44,45). Inhibition of endogenous FGFR2b ligands (FGF7 and FGF10) does not alter bleomycin-induced pulmonary fibrosis (46), suggesting that these ligands are not essential for fibrosis. As the FGF ligand(s) critical for bleomycin-induced fibrosis remain unclear, further examination of the critical FGFRs are likely to provide insight to the ligands involved, as FGF-FGFR specificity is well described (22). Furthermore, pharmacologic targeting of FGFRs may be more efficacious than targeting specific FGF ligands, particularly given overlap in ligand-receptor interactions and as multiple FGF ligands are likely involved in fibrosis. Furthermore, targeting FGFR tyrosine kinases may potentially be done in a cell-type specific manner. Continued studies using Col1␣2-CreER to inactivate individual Fgfrs will determine which receptor is most critical for bleomycin-induced pulmonary fibrosis, and will allow for improved therapeutics targeted against fibroblasts contributing to fibrotic diseases.
FGF signaling in other cell types may also contribute to the development of fibrosis. A recent report suggested that endothelial FGFR1 signaling is required for hepatic fibrosis (47), as endothelial-specific deletion of FGFR1 lessened liver fibrosis in response to bile duct ligation. The role of endothelium and angiogenesis, and in particular endothelial-specific FGF signaling, in pulmonary fibrosis is not well understood (48) and warrants further study.
We show that inactivation of FGFRs in mesenchymal cells decreases collagen expression in response to bleomycin and TGF-␤1. This appears to be a cell-autonomous effect, as collagen expression is altered in isolated Cre-expressing mesenchymal cells following bleomycin treatment. These data also support prior reports where FGF2 signaling is required for TGF-␤1 induced collagen expression in lung fibroblasts (14,15,17). Other studies have demonstrated an inhibitory effect of FGFs on pulmonary fibrosis and TGF-␤1-induced myofibroblast differentiation (44, 45, 49 -52); however, those studies use exogenous or overexpressed FGF ligands and no study to date has demonstrated an inhibitory role of endogenous FGF ligands on pulmonary fibrosis in vivo.
Our data also suggests that deletion of FGF receptors leads to a cell non-autonomous effect on mesenchymal cells that are not targeted by Col1␣2-CreER. Deletion of FGFRs in the Col1␣2ϩ lineage leads to alteration in the abundance of GFP-negative, ␣SMAϩ cells as well as expression of Col3␣1, Ctgf, Lox, and Pai1 in whole lung, but not GFPϩ (Col1␣2ϩ lineage) mesenchymal cells. Future studies will investigate the factor(s) involved in this interaction, as well as the cell types involved. It should be noted that our experiments inactivated FGFRs prior to lung injury, and future studies will also address whether inactivation of FGFRs in the Col1␣2ϩ lineage post-bleomycin has similar cell autonomous and non-autonomous effects. Additional studies are needed to determine the mechanistic role of FGFRs in multiple phases of fibroblast activation following injury and in response to pro-fibrotic growth factors such as TGF-␤1, including migration, metabolism, differentiation into myofibroblasts, and recovery from injury.
In summary, this study provides direct evidence that deletion of FGFRs in lung mesenchyme in adult mice decreases pulmonary fibrosis, reduces fibroblast-specific collagen expression, and decreases fibroblast enrichment in fibrotic areas in response to bleomycin. This study provides a mechanistic rationale for the use of FGF inhibitors to slow the progression of pulmonary fibrosis. Improved understanding of the contributions of FGF and other growth factor pathways to the development of fibrosis remains important, and future studies are needed to determine fibroblast-specific, FGFR-dependent pathways amenable to therapeutic intervention for improved therapy in IPF and other fibrotic diseases.

Animal care and use
Mice were housed in a pathogen-free barrier facility and handled in accordance with standard protocols, animal welfare regulations, and the NIH guide for the Care and Use of Laboratory Animals. All procedures complied with the standards for the care and use of laboratory animals as stated in the Guide or the Care and Use of Laboratory Animals (NIH publication No. 85-23, revised 1996), and all protocols were approved by the Animal Studies Committee at Washington University School of Medicine and at the University of Chicago. ROSA26 mTmG/ϩ mice were collected by cell sorting and grown in culture. After 2 passages, cells were serum-starved overnight and treated with TGF-␤1 (2 ng/ml). RNA was collected from cells 48 h after treatment, and analyzed by quantitative RT-PCR for Col1␣1 (A). Data were normalized to Gapdh and expressed as fold-change from PBS-treated controls. B and C, total protein, collected after 1 h of treatment, was analyzed by Western blot analysis for phosphorylated Smad2 and tubulin. Densitometric analysis was performed on n ϭ 6 replicates, and pSmad2 was normalized to tubulin and expressed as fold-change from control (C). * indicates p Ͻ 0.05; ** indicates p Ͻ 0.005. ns, not significant.

Bleomycin-induced lung fibrosis
Mice were fed tamoxifen chow for 14 days starting at 3 weeks of age, and subsequently fed regular chow. Adult mice between 8 and 10 weeks of age were sedated, orally intubated, and administered a single dose of intratracheal bleomycin (1.2 units/kg) (Sigma) in sterile phosphate-buffered saline (PBS) or PBS alone. Mice were monitored daily and weight was measured twice weekly. Mice with Ͼ25% weight loss or significant respiratory distress were humanely euthanized and treated as a death. At various times after treatment, mice were euthanized with an overdose of a mixture containing ketamine and xylazine, and exsanguinated by cutting the abdominal aorta. Lungs were then slowly perfused with 5 ml of PBS via the right ventricle. The trachea was dissected and cannulated, and lungs were fixed via intra-tracheal inflation with 10% phosphate-buffered formalin at a pressure of 20 cm H 2 O for 10 min, followed by immersion in fresh fixative for 24 h at room temperature. Samples were then dehydrated in ethanol and xylene, embedded in paraffin, cut in 5-m sections, and stained with hematoxylin and eosin (H&E) and Masson's trichrome.

Whole slide scanning
Digital scanning of whole slides was performed using a Nanozoomer 2.0 HT digital slide scanner (Hamamatsu, Bridge-water, NJ), available through the Washington University Hope Center Alafi Neuroimaging Lab (NIH Shared Instrumentation Grant (S10 RR027552)), or using an Aperio ScanScope XT through the University of Chicago Human Tissue Resource Center.

Hydroxyproline assay
Left lung samples were transferred into glass tubes and hydrolyzed with 200 l of 6 N HCl at 110°C for 48 h. The hydrolyzed samples were evaporated to dryness to remove excess HCl, reconstituted with 400 l of H 2 O, and filtered in 1.5-ml centrifuge tubes equipped with a 0.45-m semipermeable membrane filter. Samples were added to a 96-well microplate, chloramine T solution was added and the plate incubated at TCKO; ROSA26 mTmG/ϩ and control Col1␣2-CreER; ROSA26 mTmG/ϩ mice were administered a single dose of intratracheal bleomycin (1.2 units/kg) and lungs were collected 21 days post-bleomycin. Immunohistochemistry for GFP was performed (A and B) and the percentage of total fibrotic areas that were GFPϩ was analyzed using ImageJ (C). Whole lungs were dissociated and the percentage of total live single cells that were GFPϩ following bleomycin treatment were analyzed by flow cytometry (D). Proliferation was measured by injecting mice with EdU (50 mg/kg i.p.) prior to collection, and dissociated lungs were stained for CD45, CD31, and EpCam (Lin), fixed, permeabilized, and then stained for EdU incorporation. Stained cells were measured by flow cytometry, and EdU incorporation within all single cells (D) or Lin-negative cells (E) was analyzed. GFPϩ cells from bleomycin-treated Col1␣2-CreER; TCKO; ROSA26 mTmG/ϩ and control Col1␣2-CreER; ROSA26 mTmG/ϩ mice were sorted into 96-well plates (2,000 cells/well) and cultured for 48 h, 5 days, 7 days, and 9 days prior to measurement of total cellular DNA using a CyQuant assay (F). * indicates p Ͻ 0.05; ** indicates p Ͻ 0.005. ns, not significant.

Fibroblast-specific FGFR signaling in pulmonary fibrosis
room temperature for 20 min. 100 l of Erlich's reagent was added to each well and the plate was incubated at 65°C for 18 min. This method gives an orange-red color that is linear up to 6 g of hydroxyproline. A 550 nm was obtained, and compared with a hydroxyproline standard curve.

Immunohistochemistry and immunofluorescence
Five-m sections were prepared from paraffin-embedded tissues, and deparaffinized with xylene. Endogenous peroxidase was inhibited by incubation with 3% H 2 O 2 in methanol, and sections were then re-hydrated. Antigen retrieval was performed using a pressure cooker and citrate buffer (pH 6.0). Samples were incubated in PBS, and blocked in PBS ϩ 0.2% Triton X-100 ϩ 0.1% BSA ϩ 5% serum. Primary antibodies for GFP (GFP-1010, Aves Inc, Tigard, OR), ␣SMA (M0851, Dako North America, Carpinteria, CA), PDGFRa (sc-338, Santa Cruz Biotechnology, Dallas, TX), pro-SPC (ab90716, Abcam, Cambridge, UK), s100a4 (ab27957, Abcam), and periostin (ab14041, Abcam) were added to blocking buffer and slides were incubated overnight at 4°C. Slides were washed with PBS, and for immunohistochemistry studies a biotin-conjugated secondary antibody was added, followed by streptavidin-HRP. Colorimetric reaction was performed using DAB staining (Vector Labs, Burlingame, CA), and sections were counterstained with hematoxylin. Immunofluorescent imaging was performed using a Zeiss Apotome or a Marianas Tokogawa spinning disk confocal microscope, and image processing was performed using Zeiss Axioplan and ImageJ (NIH, Bethesda, MD).

Quantification of GFP immunohistochemistry
Paraffin sections from PBS-and bleomycin-treated mice were deparaffinized and stained with an anti-GFP antibody and colorimetric reaction was performed using DAB staining. 10 -15 non-overlapping ϫ10 fields containing areas of tissue fibrosis were analyzed using ImageJ software to determine the percentage of total fibrotic area that was stained for GFP.

RNA isolation and quantitative real-time PCR
RNA was isolated from whole lungs or whole left lungs via homogenization in TRIzol, and subsequent purification using RNeasy spin columns (74104, Qiagen, Valencia, CA), with oncolumn DNA digest, per the manufacturer's instructions. RNA concentration was determined utilizing a Nanodrop spectrophotometer. cDNA was made using the Bio-Rad iScript Reverse Transcription Supermix for RT-qPCR kit (170-8841, Bio-Rad). Quantitative RT-PCR was performed on an Applied Biosystems StepOne thermocycler using ABI TaqMan Fast Advaned Master Mix (number 4444557, Applied Biosystems, Foster City, CA) and TaqMan gene expression assays. All samples were normalized to Gapdh and then scaled relative to controls using the standard ⌬C t method. Data are reported as ⌬C t or fold-change relative to wild type PBStreated mice.

Protein extraction and Western blotting
Protein was extracted from whole left lung or cultured fibroblasts via homogenization in RIPA buffer with freshly added 2% Protease Inhibitor Mixture (P8340, Sigma) and Phosphatase Inhibitor Mixture I and II (P2850 and P5726, Sigma). Protein concentration was determined utilizing a Pierce BCA assay kit (23225, Thermo Scientific, Rockford, IL). Total protein (50 -100 g) was separated on 4 -12% polyacrylamide gels (Bio-Rad) and transferred to PVDF membranes using a Trans-Blot Turbo transfer system (Bio-Rad). Membranes were blocked for 1 h at room temperature with gentle shaking in TBST (50 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Tween 20) containing 5% nonfat milk or 5% BSA, and then probed with primary antibodies against phospho-ERK1/2 (number 4376, Cell Signaling Technology, Danvers, MA), ERK1/2 (Cell Signaling number 9102), tubulin (Abcam number ab6046), and phospho-Smad2 (Cell Signaling number 3101) overnight at 4°C. After three rinses in TBST, membranes were incubated for 1 h at room temperature in horseradish peroxidase-linked secondary antibodies in TBST with 5% nonfat milk, rinsed again in TBST, and developed using SuperSignal West Femto Substrate (34096, Thermo Scientific). Protein bands were quantified using Image Lab (Bio-Rad), normalized to tubulin, and scaled relative to control samples set at a value of 1.

Flow cytometry and cell sorting
Mice were euthanized with an overdose of a mixture containing ketamine and xylazine, and exsanguinated by cutting the abdominal aorta. Lungs were then slowly perfused with 5 ml of cold PBS via the right ventricle. The trachea was dissected and cannulated, and lungs were inflated with DMEM digest media containing Liberase TM (Roche Diagnostics), hyaluronidase (Sigma), elastase (Worthington Biochemical, Lakewood, NJ), and DNase (Sigma). Inflated lungs were transferred to digest media and placed on ice until all mice were dissected. Lungs were then minced with razor blades and incubated in digest medium at 37°C for 30 min with agitation. Digest enzymes were inactivated with DMEM with 10% FBS, and samples were filtered with a 100-m strainer. Red blood cells were then lysed with ACK buffer (Thermo Scientific) and cells were resuspended in FACS buffer (PBS ϩ 3% FBS ϩ 1 mM EDTA). Cells were stained with BV-421-conjugated antibodies against CD45, CD31, and CD326 (EpCAM) (BioLegend, San Diego, CA). GFPϩ cells were sorted using a Sony Synergy (Washington University) or FACSAria II (University of Chicago), using a 100-m diameter nozzle. For culture, cells were sorted directly into complete media containing 10% FBS. For RNA isolation, between 50,000 and 100,000 cells were sorted directly into RLT lysis buffer with ␤-mercaptoethanol, and RNA was purified using an RNeasy Plus Micro Kit (Qiagen).

Lung fibroblast isolation and culture
GFP-labeled cells were sorted directly into culture media, and subsequently cultured on collagen-coated plates in DMEM with 10% FBS, penicillin/streptomycin, fungizone, L-glutamine, and HEPES. Cells were grown in a humidified incubator at 37°C with 5% CO 2 . Media was replaced every other day, and cultured fibroblasts were used for experiments between passages 2 and 5.

In vivo proliferation assay
Prior to collection, mice were given an intraperitoneal injection of EdU (50 mg/kg, Thermo Fisher). Mouse lungs were enzymatically dissociated as described above, and cells were stained with BV-421-conjugated antibodies against CD45, CD31, and CD326 (EpCAM) (BioLegend). Cells were then fixed, permeabilized, and stained for EdU using the Click-iT EdU assay (Thermo Fisher) according to the manufacturer's protocol. Flow cytometry was performed using a BD LSR II cytometer, and data analyzed using FlowJo.

In vitro proliferation assay
Mouse lungs were dissociated and stained with BV-421-conjugated antibodies against CD45, CD31, and CD326 (EpCAM) (BioLegend) as described, dead cells were excluded using 7-aminoactinomycin staining, and 2000 GFPϩ cells were directly sorted into individual wells of 96-well collagen-coated plates containing complete media using a FACSAria II sorter. Cells were grown as described above. At specified times, plates were washed with PBS and placed at Ϫ80°C. DNA content was measured using a CyQuant assay per manufacturer protocol.

Statistical analysis
Significant differences in mean values were calculated using paired Student's t tests or one-way analysis of variance. Survival analysis was performed using Kaplan-Meier analysis. A p value of less than 0.05 was considered to be significant.