RUNX2 promotes fibrosis via alveolar-to-pathological fibroblast transition

Abstract

A hallmark of pulmonary fibrosis is aberrant activation of lung fibroblasts into pathological fibroblasts producing excessive extracellular matrix (ECM)^1–3. Thus, identification of key regulator(s) driving the generation of pathological fibroblasts can inform effective countermeasures against disease progression. In this study, we show that Leptin Receptor (Lepr)⁺ fibroblasts arising during alveologenesis include signal peptide, CUB domain, EGF-like 2 (Scube2)⁺ alveolar fibroblasts as a major constituent significantly contributing to collagen triple helix repeat containing 1 (Cthrc1)⁺ Periostin (Postn)⁺ pathological fibroblasts in two mouse models of pulmonary fibrosis. Genetic ablation of the Postn⁺ pathological fibroblasts attenuates fibrosis. Comprehensive analysis of single cell RNA sequencing (scRNA-seq) and single cell Assay for Transposase-Accessible Chromatin sequencing (scATAC-seq) reveal Runt related transcription factor 2 (RUNX2) as a key regulator of fibrotic genes. Consistently, conditional deletion of Runx2 with Lepr-CreERT2 or Scube2-CreERT2 reduces the generation of pathological fibroblasts, ECM deposition, and pulmonary fibrosis. Therefore, Lepr⁺ cells which include Scube2⁺ alveolar fibroblasts are a key source of pathological fibroblasts, and targeting Runx2 provides a potential treatment option for pulmonary fibrosis.

Pulmonary fibrosis is characterized by the overwhelming presence of aberrant activated fibroblasts with excessive deposition of extracellular matrix proteins (e.g. collagens), rendering a progressively dysfunctional lung^1–4. Multiple sources of fibrotic fibroblasts have been associated with pulmonary fibrosis including Gli1⁺ stromal cells, Tbx4⁺ resident fibroblasts, Axin2⁺ myofibrogenic progenitor cells and Pdgfra⁺Adrp⁺ lipofibroblasts^5–8. More recently, scRNA-seq analysis suggests that fibrotic changes are associated with the presence of Cthrc1⁺ pathological fibroblasts in both mouse models and human idiopathic pulmonary fibrosis (IPF)⁹. RNA velocity analysis suggests that these Cthrc1⁺ cells are derived from alveolar fibroblasts⁹ which are essential for maintaining normal alveolar architecture^10,11. However, cell-fate mapping evidence is lacking. Moreover, although multiple signaling pathways^12–14, including TGF-β and PDGF have clearly been connected to pulmonary fibrosis, the key downstream regulator(s) remain to be identified, in part due to the uncertain origins of these pathological fibroblasts. Thus, investigation of the major source(s) of pathological fibroblasts is critical for uncovering the underlying molecular mechanisms and developing effective therapies.

Recently, Lepr has been shown to mark a subpopulation of stromal cells in the bone marrow^15,16. Upon myelofibrosis induced by the ectopic expression of the glycoprotein hormone, thrombopoietin, Lepr-Cre labeled stromal cells become a major contributor to myofibroblasts¹⁶. Here, we found that Lepr is expressed in murine lung mesenchymal cells through unbiased scRNA-seq analysis. We generated a novel Lepr-CreERT2 allele and used it along with the existing Lepr-Cre mouse line to trace Lepr⁺ lung mesenchymal cells that arise during neonatal alveologenesis. Intriguingly, the majority of labeled cells are Scube2⁺ alveolar fibroblasts and they give rise to pathological fibroblasts (Cthrc1⁺ Postn⁺) upon bleomycin challenge as revealed by cell-fate mapping and scRNA-seq analysis. Moreover, genetic ablation of Postn⁺ cells attenuates pulmonary fibrotic changes. Further computational analysis combined with scATAC-seq identified the transcription factor RUNX2 as a key regulator promoting the generation of pathological fibroblasts. Conditional deletion of Runx2 with Lepr-CreERT2 or Scube2-CreERT2 blocks the generation of pathological fibroblasts and ECM deposition. Together, our findings support the concept that alveolar fibroblasts, one major constituent of Lepr⁺ fibroblasts serve as a key source of pathological fibroblasts, the generation of which is dependent on RUNX2.

Lepr⁺ mesenchymal cells contribute to pathological fibroblasts

Reanalysis of scRNA-seq data of developing mouse lungs¹⁷ revealed that Lepr transcripts were detected in multiple mesenchymal cells including alveolar fibroblasts, secondary crest myofibroblasts, adventitial fibroblasts and pericytes, as well as endothelial cells (Fig. 1a, b and Extended Data Fig. 1a-e). The expression level of Lepr gradually increased during postnatal alveologenesis and was abundant at postnatal day 14 (P14) but not embryonic day 18.5 (E18.5) (Fig. 1c). In line with this finding, very few tdTomato⁺ (tdT⁺) cells were observed in the lungs of Lepr-Cre; R26-tdtomato(R26tdT) mice at P0 (Extended Data Fig. 1f). In contrast, we observed an increased presence of tdT⁺ cells in the lung parenchyma at P10, and tdT⁺ mesenchymal cells were observed throughout the alveoli at P30 and remained at P60 (Extended Data Fig. 1f). Consistent with scRNA-seq data, very few tdT⁺ α-SMA⁺ cells (< 0.1% of tdT⁺ α-SMA⁺ cells) were also present in the peri-bronchiolar and peri-vascular areas, suggesting that a small number of airway and vascular smooth cells expressed Lepr (Extended Data Fig. 1g). We generated a novel Lepr-CreERT2 knock-in mouse strain to confirm the derivatives of Lepr⁺ cells (Extended Data Fig. 1h). Consistently, tdT⁺ cells were observed in the lungs of P14 and P20 Lepr-CreERT2;R26tdT mice that received a single dose of Tamoxifen (Tmx) at P7 and P14, respectively (Extended Data Fig. 1i, j). When Tmx-containing chow was used to label Lepr⁺ cell derivatives in the adult lung, we observed the extensive presence of tdT⁺ cells in the lung parenchyma (Extended Data Fig. 1k). Of note is that 75% of tdT⁺ cells were in close contact with alveolar epithelial type 2 (AT2) cells in the alveoli, suggestive of alveolar fibroblasts (Extended Data Fig. 1k). Consistently, reanalysis of scRNA-seq data^9,18,19 from adult murine lungs showed that Lepr and the alveolar fibroblast marker Scube2 were colocalized in alveolar fibroblasts (Extended Data Fig. 1l-p). Moreover, in situ hybridization with a Scube2 probe showed 63.54±1.36% tdT⁺ cells expressed Scube2 (Extended Data Fig. 1q, r), while 53.09±1.07% Scube2⁺ cells expressed tdTomato (Extended Data Fig. 1s), indicating that Lepr-CreERT2 mouse strain efficiently labels alveolar fibroblasts.

Fig. 1. Lepr-expressing lung mesenchymal cells generate pathological fibroblasts during pulmonary fibrosis.

a, UMAP plot showing multiple mesenchymal cell populations in the mouse lungs at E18.5, P0, P3, P7 and P14. b, Expression of Lepr in lung mesenchymal cells. c, Increased expression of Lepr in lung mesenchymal cells during alveologenesis. d, Schematic depicting the treatment of Lepr-Cre;R26tdT mice with bleomycin and representative images of α-SMA⁺ tdTomato (tdT⁺) cells in lung tissues. e, Quantification showing increased α-SMA⁺tdT⁺ cells following bleomycin treatment (Saline: n = 6, Bleomycin: n = 6). f, Representative images showing immunostaining of Cthrc1 and tdT in lung tissues. g, Quantification analysis showing the enrichment of Cthrc1⁺tdT⁺ pathological fibroblasts after bleomycin challenge (Saline: n = 6, Bleomycin: n = 6). h, Schematic diagram showing the use of tamoxifen (Tmx) containing chow and bleomycin challenge of Lepr-CreERT2;R26tdT mice, and representative images of α-SMA⁺tdT⁺ cells in lung tissues. i, Quantification showing increased α-SMA⁺tdT⁺ cells following bleomycin treatment (Saline: n = 6, Bleomycin: n = 6). j, Representative images showing immunostaining of Cthrc1 and tdT in lung tissues. k, Quantification analysis showing the increased Cthrc1⁺tdT⁺ pathological fibroblasts after bleomycin challenge (Saline: n = 6, Bleomycin: n = 6). Data are mean±SEM. Data are representative of at least three independent experiments. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction (e, g, i, k). Scale bars: 100 µm (20 µm in magnified views).

At homeostasis we did not detect apparent differences in the migration, production of extracellular matrix (ECM) and response to TGF-β1 stimulation between Lepr-cre-labeled fibroblasts and non-labeled lung fibroblasts (Extended Data Fig. 2a-d). We asked whether the derivatives of Lepr⁺ cells contribute to pathological fibroblasts during fibrosis. We employed two mouse models to address this issue, bleomycin- and silica-induced pulmonary fibrosis. Adult Lepr-Cre; R26tdT mice were treated with bleomycin and lungs were examined 14 days post injury (Fig. 1d). Bleomycin induced heterogeneous injuries and fibrosis in the lungs^20,21, and tdT⁺ cells were significantly expanded in the areas showing severe damage (Fig. 1d-g and Extended Data Fig. 2e-h). tdT⁺ cells were positive for the myofibroblast marker α-SMA and the pathological fibroblast marker Cthrc1 (Fig.1d, f). tdT⁺ cells also expressed Collagen I (Extended Data Fig. 2e), one of the major components of ECM deposited by pathological fibroblasts^9,22. Enrichment of tdT⁺α-SMA⁺ fibroblasts following bleomycin treatment was further confirmed by FACS analysis (2.83 ± 0.39% Vs 11.74 ± 0.95%, P = 0.001051) (Extended Data Fig. 2i-k). We also challenged Lepr-CreERT2; R26tdT mice with bleomycin after feeding the mice with Tmx-containing chow (Fig. 1h). Remarkably, an extensive number of tdT⁺ cells expressing α-SMA and Cthrc1 were detected in the fibrotic foci following bleomycin challenge, confirming that Lepr⁺ cell derivatives expanded when fibrosis occurred (Fig. 1h-k). We then used the silica-induced fibrosis model to further test the contribution of Lepr⁺ cell derivatives to pulmonary fibrosis in both Lepr-Cre; R26tdT and Lepr-CreERT2; R26tdT mice (Extended Data Fig. 2l-q). Silica treatment led to extensive fibrosis surrounding the terminal airways (Extended Data Fig. 2l, o), and tdT⁺α-SMA⁺ cells were abundant in the fibrotic foci surrounding the terminal bronchioles in the lungs of both Lepr-Cre; R26tdT and Lepr-CreERT2; R26tdT mice (Extended Data Fig. 2m, n, p, q). Together these findings demonstrate that Lepr⁺ mesenchymal cells arising during alveologenesis become a key source of pathological fibroblasts contributing to pulmonary fibrosis.

Lepr⁺ alveolar fibroblasts generate pathological fibroblasts

We next aimed to define the fibroblast subpopulations derived from Lepr⁺ cells. First, we sorted tdT⁺ cells from the lungs of Lepr-Cre; R26tdT mice treated with bleomycin or saline and performed scRNA-seq analysis (Extended Data Fig. 3a). tdTomato transcripts were detected in all cells (Extended Data Fig. 3b). Although the majority of the Lepr-Cre-labeled cells were mesenchymal cells, minor populations of epithelium, endothelium and immune cells were also identified (Extended Data Fig. 3c-e). Along this line, Lepr has been shown to be expressed in immune cells in the bone marrow¹⁵. To better characterize Lepr⁺ lung mesenchymal cells and their derivatives in the adult lungs, Lepr-CreERT2; R26tdT mice were fed with Tmx-containing chow before bleomycin challenge (Fig. 2a). tdT⁺ lung mesenchymal cells (EpCAM^-CD45^-CD31^-) were FACS-sorted from the lungs following saline or bleomycin treatment and then analyzed by scRNA-seq (Extended Data Fig. 4a-e). In the saline-challenged lungs, alveolar fibroblasts and adventitial fibroblasts stood out as two major subpopulations along with minor populations including peribronchial fibroblasts and pericytes (Extended Data Fig. 4c-d). As expected, bleomycin treatment induced fibrotic changes in the lungs, and the transcript levels of fibrotic genes including Acta2, Col1a1, Col3a1, Tnc and Fn1 were significantly elevated (Extended Data Fig. 4f).

Fig. 2. scRNA-seq analysis reveals Lepr-CreERT2 labeled pathological fibroblasts induced by bleomycin treatment.

a, Schematic diagram for scRNA-seq experimental design. b, UMAP plot showing four different Lepr-CreERT2 labeled lung fibroblast populations. c, Dot plot showing the representative markers for each fibroblast population. d, Red dots showing the cells from saline-treated lung tissues and blue dots showing the cells from bleomycin-treated lung tissues. e, Frequency of each fibroblast population in saline and bleomycin-treated lungs. f, UMAP plot showing the score of alveolar fibroblast signature. g, UMAP plot showing the expression of Scube2 in alveolar fibroblasts. h, UMAP plot showing the score of pathological fibroblast signature. i-j, Trajectory analysis with Monocle2 showing cells ordered by pseudotime (i) and cell type (j). k, Volcano plot showing the differentially expressed genes (DEGs) between alveolar fibroblasts and pathological fibroblasts. l, The top ten signaling pathways associated with gene enriched in pathological fibroblasts. m, UMAP plot showing the score of TGF-β signaling pathway signature.

Re-clustering of Lepr-Cre or Lepr-CreERT2 labeled fibroblast populations revealed four subpopulations including alveolar fibroblasts, adventitial fibroblasts, pathological fibroblasts and proliferating fibroblasts (Fig. 2b-e, Extended Data Fig. 3f-i and Extended Data Fig. 4g). Alveolar fibroblasts were identified with the alveolar fibroblast signature including the expression of Scube2, Npnt and Inmt as previously reported^9,23 (Fig. 2c, f, g, Extended Data Fig. 3h, j-l and Extended Data Fig. 4h, i). In addition, adventitial fibroblasts were also identified with the reported adventitial fibroblast signature including the expression of Pi16 and Dcn^9,23 (Extended Data Fig. 3h, m, n and Extended Data Fig. 4j-l), and immunostaining confirmed the tdT⁺ cells expressed Pi16 in the adventitial cuff area (Extended Data Fig. 3o and Extended Data Fig. 4m) where adventitial fibroblasts are by definition located^9,24,25. At homeostasis Lepr-Cre or Lepr-CreERT2 labeled fibroblasts were predominantly alveolar and adventitial fibroblasts (Fig. 2e and Extended Data Fig. 3i). Upon bleomycin challenge both fibroblast populations were reduced with alveolar fibroblasts being more pronounced (Fig. 2e and Extended Data Fig. 3i), accompanied by the overwhelming emergence of pathological fibroblasts and proliferating fibroblasts which are associated with increased expression of the cell cycle regulators (e.g. Ccna2, Cdk1) (Fig. 2c, e, Extended Data Fig. 3h, i, Supplementary Table 1 and Supplementary Table 2).

Pathological fibroblasts were recently defined by the expression of fibrotic signature genes including Cthrc1 and Postn^9,23. Significantly, the bleomycin-induced fibroblast population derived from Lepr-Cre or Lepr-CreERT2 labeled cells was also characterized by these genes (Fig. 2c and Extended Data Fig. 3h, p, q). Immunostaining confirmed that Lepr-Cre or Lepr-CreERT2 labeled cells in fibrotic foci expressed Cthrc1 (Fig. 1f, j) and Postn (Extended Data Fig. 3r and Extended Data Fig. 4n). Trajectory analysis suggested that Lepr-Cre or Lepr-CreERT2 labeled alveolar fibroblasts give rise to pathological fibroblasts (Fig. 2i, j, Extended Data Fig. 3s-u and Extended Data Fig. 5a-b), which is consistent with recent studies^9,26. Differential gene expression analysis confirmed the enrichment of Scube2, Inmt and Npnt in Lepr-CreERT2 labeled alveolar fibroblasts (Fig. 2k). By contrast, fibrosis-associated genes including Col1a1 and Tnc along with Cthrc1 and Postn were highly expressed in pathological fibroblasts (Fig. 2k). Pathway analysis indicated that ECM-associated pathways including ECM-receptor interaction, ECM organization and collagen biosynthesis were activated in pathological fibroblasts (Fig. 2l and Extended Data Fig. 5c, d). In line with these findings, Tgfβ1 transcript and TGF-β signaling pathway signature were enriched in Lepr-CreERT2 labeled pathological fibroblasts (Fig. 2m and Extended Data Fig. 5e), suggesting activation of the fibrotic program following bleomycin challenge. Notably, the level of Lepr transcript was significantly decreased upon bleomycin treatment (Extended data Fig. 3v and Extended data Fig. 5f), suggesting that Lepr is not activated during fibrosis.

Given Lepr-Cre/Lepr-CreERT2 labeled minor pericyte population (Extended Data Fig. 3c and Extended Data Fig. 4c) and the controversial contribution of pericytes to pathological fibroblasts (myofibroblasts) during lung fibrosis^27,28, we generated a novel Higd1b-CreERT2 knock-in mouse strain to specifically fate map pericytes in the lung and heart where HIG1 domain family member 1B (Higd1b) is exclusively expressed²⁹ (Extended Data Fig. 6a-e). We challenged adult Higd1b-CreERT2;R26tdT mice with bleomycin after three doses of Tmx injection. A few tdT⁺ cells co-expressing α-SMA were observed in fibrotic foci (Extended Data Fig. 6f). However, tdT⁺ cells rarely expressed Cthrc1 (Extended Data Fig. 6g, h), Collagen I and Ki67 (Extended Data Fig. 6i, j), suggesting that Higd1b-CreERT2 labeled pericytes contribute to α-SMA⁺ cells surrounding endothelial cells, but they rarely generate pathological fibroblasts. That being said, we can not completely rule out the possibility that a minor subpopulation of pericytes that do not express Higd1b contribute to fibrosis (Extended Data Fig. 6a, b). Taken together, our scRNA-seq data confirmed Lepr-Cre and Lepr-CreERT2 labeled fibroblasts include alveolar fibroblasts as a major constituent. Upon injury these labeled cells were converted into pathological fibroblasts expressing Cthrc1 and Postn. In consideration of the recent findings^9,26, Lepr⁺ fibroblasts likely represent a key contributor of pathological fibroblasts.

Ablation of Postn⁺ fibroblasts attenuates pulmonary fibrosis

Our scRNA-seq analysis demonstrated that pathological fibroblasts were characterized by the expression of Postn and Cthrc1 which were rarely expressed in normal lung fibroblasts (Fig. 3a-c). Consistently, cell-fate mapping with the knockin Postn-MerCreMer (Postn-CreER) mouse line³⁰ confirmed that very few mesenchymal cells were labeled in the saline-challenged adult lungs (Fig. 3d). By contrast, numerous tdT⁺ cells expressing the myofibroblast marker α-SMA, the pathological fibroblast marker Cthrc1 and ECM component Collagen I were present in the fibrotic area following bleomycin challenge (tdT⁺ cells: 0.00 ± 0.00 per mm² Vs 319.00 ± 12.87 per mm², P < 0.0001) (Fig. 3d-g). Re-analysis of published scRNA-seq data³¹ also showed the high expression of Postn in pathological fibroblasts upon silica exposure (Extended Data Fig. 7a-d). We further validated this finding with the silica injury model where prominent accumulation of lineage-labeled fibroblasts (tdT⁺ α-SMA⁺) was observed in the fibrotic areas (tdT⁺ cells: 0.00 ± 0.00 per mm² Vs 335.80 ± 16.65 per mm², P < 0.0001) (Extended Data Fig. 7e, f). Given the abundant presence of Postn⁺ pathological fibroblasts during fibrosis, we predicted that ablation of these cells attenuates fibrotic changes. Postn-CreER;R26-DTA mice were treated with bleomycin followed by Tmx injection (Fig. 3h). Pulmonary fibrosis was significantly attenuated upon Tmx injection, as determined by the quantification of fibrotic areas and hydroxyproline content (Fibrotic area: 23.28 ± 2.59% Vs 11.90 ± 1.83%, P = 0.002197, Hydroxyproline content: 1.90 ± 0.11 Vs 1.43 ± 0.13 µg/mg lung, P = 0.010615) (Fig. 3h-k), without apparent effects on the pulmonary vasculature (Extended Data Fig. 7g). These findings demonstrated that Postn is switched on during the conversion of Lepr⁺ lung fibroblasts into pathological fibroblasts and that ablation of Postn⁺ cells reduces pulmonary fibrosis.

Fig. 3. Ablation of Postn⁺ pathological fibroblasts attenuates pulmonary fibrosis.

a-b, UMAP plot showing the expression of Postn (a) and Cthrc1 (b) in pathological fibroblasts from the lungs of Lepr-CreERT2;R26tdT mice challenged with saline or bleomycin. c, Enriched expression of Postn and Cthrc1 in pathological and proliferating fibroblasts. d, Schematic diagram for Tmx injection and bleomycin challenge of Postn-CreER;R26tdT mice, and representative immunostaining images of α-SMA and tdT in lung tissues. e-f, Representative images showing Cthrc1⁺ tdT⁺ cells (e) and Collagen I⁺ tdT⁺ cells (f). g, Quantification analysis showing increased tdT⁺ cells upon bleomycin challenge (Saline: n = 6, Bleomycin: n = 6). h, Schematic diagram for Tmx injection and bleomycin challenge of Postn-CreER;R26-DTA mice, and representative H&E staining images. i, Quantification analysis showing decreased fibrotic areas in the lungs of Postn-CreER;R26-DTA mice (Postn-CreER;R26-tdT: n = 10, Postn-CreER;R26-DTA/tdT: n = 15). j, Representative images of Picro-Sirius Red staining. k, Reduced hydroxyproline content in Postn-CreER;R26-DTA mice following bleomycin challenge (Postn-CreER;R26-tdT: n = 10, Postn-CreER;R26-DTA/tdT: n = 14). Data are mean±SEM. Data are representative of at least three independent experiments. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction (g, i, k). Scale bars: d-f: 100 µm (20 µm in magnified views), h and j: 1 mm (h, 200 µm in magnified views, j, 100 µm in magnified views).

Runx2 in alveolar-to-pathological fibroblast conversion

The Enrichr program has been instrumental for enrichment analysis and identification of key transcription factors contributing to tissue regeneration and disease^32–34. We therefore employed this program to predict transcription factor(s) that regulate the conversion of alveolar fibroblasts to pathological fibroblasts. Among multiple transcription factors RUNX2 was predicted to regulate the most fibrosis-associated genes (Fig. 4a). Importantly, our scRNA-seq analysis of both Lepr-Cre and Lepr-CreERT2 labeled cells showed that the Runx2 transcripts were dramatically elevated and highly enriched in the pathological fibroblast population (Fig. 4b and Extended Data Fig. 8a-c). In parallel, we performed scATAC-seq to profile chromatin accessibility. Mesenchymal cells (EpCAM^-CD45^-CD31^-) were sorted from the lungs of mice treated with bleomycin or saline and subjected to scATAC-seq (Extended Data Fig. 8d). Similar to our scRNA-seq analysis of Lepr-Cre or Lepr-CreER labeled populations, multiple fibroblast subpopulations were identified (Extended Data Fig. 8e-g). Again, the pathological fibroblast population was only present in bleomycin-treated lungs (Extended Data Fig. 8e, f). ChromVAR³⁵ analysis revealed that the RUNX2-binding motif was preferentially enriched in pathological fibroblasts (Fig. 4c). Consistently, the chromatin accessibility at the Runx2 locus was highly enriched in pathological fibroblasts (Fig. 4d). Additionally, the transcript level of Runx2 was increased in pathological fibroblasts of bleomycin-induced lungs as shown by reanalysis of published scRNA-seq datasets^9,19,36,37 (Extended Data Fig. 8h-l). The findings were further confirmed by qRT-PCR analysis and the extensive presence of Runx2⁺tdT⁺ cells in the lungs of Lepr-Cre; R26tdT mice following bleomycin challenge (Fig. 4e, f and Extended Data Fig. 8m). Taken together, these data suggest that RUNX2 functions as a key transcription factor regulating the generation of pathological fibroblasts. Notably, our reanalysis of a scRNA-seq dataset³⁸ revealed that Runx2 was co-expressed with Cthrc1 in an extensive number of active hepatic stellate cells during carbon tetrachloride (CCl₄)-induced liver fibrosis (Extended data Fig. 8n-q).

Fig. 4. Runx2 insufficiency blocks the generation of pathological fibroblasts and attenuates lung fibrosis.

a, The top eight transcription factors predicted to regulate the fibrotic genes enriched in pathological fibroblasts. b, scRNA-seq UMAP plot showing Runx2 expression in lung mesenchymal cells of Lepr-CreERT2;R26tdT mice challenged with saline or bleomycin. c, scATAC-seq UMAP plot showing the RUNX2 motif MA0511.2. d, scATAC-seq UMAP plot showing Runx2 chromatin accessibility. e, Runx2⁺tdT⁺ cells (arrows) in lungs treated with saline or bleomycin. f, Quantification of Runx2⁺ tdT⁺ cells (Saline: n = 6, Bleomycin: n = 6). g, Schematic of experimental design. h, Representative H&E staining images. i, Quantification of fibrotic areas in the lungs of control (n = 8) and Lepr-CreERT2;Runx2^f/f;R26tdT mice (n = 8). j, Hydroxyproline content in the lungs of control (n = 9) and Lepr-CreERT2;Runx2^f/f;R26tdT mice (n = 9) following bleomycin challenge. k, Expression of Cthrc1 and tdT in the lungs. l, Quantification of Cthrc1⁺ cells in the lungs of control (n = 6) and Lepr-CreERT2;Runx2^f/f;R26tdT mice (n = 6). m, Schematic of experimental design. n, Representative H&E staining images. o, Quantification of fibrotic areas in the lungs of control (n = 8) and Scube2-CreERT2;Runx2^f/f;R26tdT mice (n = 8). p, Hydroxyproline content in the bleomycin-treated lungs of control (n = 10) and Scube2-CreERT2;Runx2^f/f;R26tdT (n = 10) mice. q, Expression of Cthrc1 and tdT in the lungs. r, Quantification of Cthrc1⁺ cells in the lungs of control (n = 6) and Scube2-CreERT2;Runx2^f/f;R26tdT (n = 6). Data are representative of at least three independent experiments. Data are mean±SEM. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction (f, i, j, l, o, p, r). Scale bars: e, k and q: 50 µm (20 µm in magnified views), h and n: 1 mm (200 µm in magnified views).

Runx2 insufficiency attenuates pulmonary fibrosis

We next tested whether Runx2 promotes the generation of pathological fibroblasts during pulmonary fibrosis. Lepr-Cre; Runx2^f/f and Lepr-CreERT2; Runx2^f/f mice were challenged with bleomycin. Notably, deletion of Runx2 with Lepr-Cre which also labeled bone marrow stromal cells, did not alter the lung morphology or the proportion of monocytes and neutrophils in the bone marrow at homeostasis (Extended Data Fig. 9a-e, Supplementary Fig. 1), although they both have been associated with the pathogenesis of pulmonary fibrosis^39,40. Remarkably, specific ablation of Runx2 in Lepr-Cre derived cells attenuated bleomycin-induced lung fibrosis as the fibrotic areas were reduced from 24.97 ± 2.24% to 10.96 ± 1.58% (P = 0.000145) (Extended Data Fig. 9f, g). In line with this finding, Lepr-Cre; Runx2^f/f mutant lungs exhibited reduced hydroxyproline content (1.67 ± 0.09 vs 1.30 ± 0.08 µg/mg lung, P = 0.008515) and α-SMA⁺ myofibroblasts were also decreased in mutants (17.25 ± 0.95% to 7.68 ± 0.57%, P < 0.0001) (Extended Data Fig. 9h-j). Loss of Runx2 in Lepr-CreERT2; Runx2^f/f mutant lungs similarly reduced fibrotic areas and hydroxyproline content (Fibrotic area: 23.97 ± 2.01% Vs 15.34 ± 1.25%, P = 0.003504, Hydroxyproline content: 1.68 ± 0.15 Vs 1.26 ± 0.11 µg/mg lung, P = 0.039141) (Fig. 4g-j and Extended Data Fig. 9k). In addition, α-SMA⁺ myofibroblasts and Cthrc1⁺ pathological fibroblasts were decreased in Lepr-CreERT2; Runx2^f/f mutant lungs (α-SMA⁺ myofibroblasts: 16.03 ± 0.85% Vs 10.10 ± 0.55%, P = 0.000301. Cthrc1⁺ pathological fibroblasts: 142.70 ± 8.78 Vs 85.67 ± 6.16 per field, P = 0.000492) (Fig. 4k, l and Extended Data Fig. 9l, m). Given that Scube2⁺ alveolar fibroblasts are the major constituent of Lepr⁺ cell derivatives, we asked whether deletion of Runx2 using Scube2-CreERT2 attenuates bleomycin-induced fibrosis (Fig. 4m). Significantly, fibrotic area and hydroxyproline content were reduced upon Runx2 deletion (Fibrotic area: 21.45 ± 1.84% Vs 11.92 ± 1.55%, P = 0.001504, Hydroxyproline content: 1.79 ± 0.13 Vs 1.07 ± 0.11 µg/mg lung, P = 0.000654) (Fig. 4m-p and Extended Data Fig. 9n). Moreover, α-SMA⁺ myofibroblasts and Cthrc1⁺ pathological fibroblasts were also dramatically reduced (α-SMA⁺ myofibroblasts: 16.88 ± 0.33% Vs 8.32 ± 0.65%, P < 0.0001. Cthrc1⁺ pathological fibroblasts: 139.70 ± 7.07 Vs 61.33 ± 7.17 per field, P < 0.0001) (Fig. 4q, r and Extended Data Fig. 9o, p). Together these findings support the conclusion that Runx2 is critical for the conversion of alveolar fibroblasts into pathological fibroblasts during pulmonary fibrosis.

We next reanalyzed published scRNA-seq datasets⁴¹ to determine whether RUNX2 expression is also increased in human IPF lungs. Four fibroblast populations, including alveolar fibroblasts, adventitial fibroblasts, pathological fibroblasts and proliferating fibroblasts were identified (Fig. 5a and Extended Data Fig. 10a, b). LEPR was expressed in alveolar fibroblasts and adventitial fibroblasts but less in pathological fibroblasts, whereas SCUBE2 was barely expressed in human alveolar fibroblasts (Extended Data Fig. 10c, d). RUNX2 was predominantly expressed by POSTN⁺CTHRC1⁺ pathological fibroblasts which also expressed high levels of ACTA2 and COL1A1 along with other ECM proteins (Fig. 5b-d and Extended Data Fig. 10e-g). Notably, pathological fibroblasts from human and murine lungs shared key marker genes including Cthrc1, Postn, Tnc, Acta2, Fn1, Sparc, Mmp14 and various collagens, although some genes were differentially expressed, possibly due to the cross-species differences (Extended Data Fig. 10h and Supplementary Table 3). Further analysis of scRNA-seq data confirmed the increased expression of RUNX2 transcripts in IPF pathological fibroblasts (Fig. 5e, f), which was in line with the bulk RNA-seq data published by two different groups^42,43 (Fig. 5g and Extended Data Fig. 10i). In addition, numerous RUNX2⁺α-SMA⁺ fibroblasts were observed in IPF lungs in contrast to normal lungs where RUNX2⁺ cells were rarely present (36.38 ± 6.57 per mm² Vs 2.20 ± 1.07 per mm², P = 0.00115) (Fig. 5h). Consistently, the expression level of RUNX2 was significantly elevated in primary human normal lung fibroblasts upon TGF-β1 treatment (Fig. 5i), and knockdown of RUNX2 with siRNA suppressed TGF-β1-induced pathological fibroblast differentiation and ECM production (Fig. 5j). We also isolated fibroblasts from IPF samples and tested whether knockdown of RUNX2 affected the expression of fibrotic genes. Significantly, RUNX2 knockdown led to a dramatic reduction in the transcript levels of ACTA2, FN1 and COL1A1 (Fig. 5k), suggesting that RUNX2 also plays a functional role during human pulmonary fibrosis.

Fig. 5. RUNX2 mediates ECM production in human IPF fibroblasts.

a, UMAP plot showing four different lung fibroblast populations in human normal and IPF lungs. b-c, UMAP plot showing expression of POSTN (b) and CTHRC1 (c) in the pathological fibroblast population. d, UMAP plot showing expression of RUNX2. e, Enriched expression of RUNX2 in pathological fibroblasts. f, Violin plot of scRNA-seq showing the significantly elevated RUNX2 transcripts in IPF lungs. g, The increased transcript levels of RUNX2 in IPF lungs revealed by re-analyzing the bulk RNA-seq dataset GSE124685 (Normal: n = 35, IPF: n = 49). h, Representative immunostaining of α-SMA and RUNX2 in normal human lungs (n = 5) and IPF lungs (n = 8). The arrows indicate α-SMA⁺ RUNX2⁺ cells. i, qRT-PCR analysis of RUNX2 in freshly isolated primary normal human lung fibroblasts treated with TGF-β1 or vehicles (Vehicle: n = 3, TGF-β1: n = 3). j, siRNA-mediated knockdown of RUNX2 impacts expression of fibrosis-associated genes in freshly isolated primary normal human lung fibroblasts treated with TGF-β1 or vehicles. The transcripts of ACTA2, CTHRC1, POSTN, CCN2, COL1A1, COL3A1, COL4A1 and FN1 were determined by qRT-PCR (n = 3 for each group). k, qRT-PCR analysis of human IPF lung fibroblasts transfected with RUNX2 siRNA or control siRNA (control siRNA: n = 3, RUNX2 siRNA: n = 3). l, Schematic diagrams showing that Lepr⁺Scube2⁺ alveolar fibroblasts contribute to Cthrc1⁺Postn⁺ pathological fibroblasts, and that elevated Runx2 regulates expression of fibrotic genes during disease progression. Data are representative of at least three independent experiments. Data are mean±SEM. Statistical analysis was performed using two-sided Wilcoxon Rank Sum test (f), unpaired two-tailed t-test with Welch's correction (g, i, k) and two-way ANOVA multiple comparisons test with Sidak’s correction (j). Scale bars: h: 50 µm (10 µm in magnified views).

Discussion

The obscure cell of origin for pulmonary fibrosis contributes to the uncertainty of the factors that drive initial fibrotic changes. Our comprehensive scRNA-seq analysis uncovered Lepr as a marker labeling the majority of lung fibroblasts that persisted in the adult lung. Cell-fate mapping confirmed that upon injuries Lepr⁺ cell derivatives, the majority of which were alveolar fibroblasts differentiated into Cthrc1⁺Postn⁺ pathological fibroblasts. Genetic ablation of pathological fibroblasts significantly attenuated pulmonary fibrosis. Our further analysis revealed RUNX2 as a key driver for pulmonary fibrosis, and inactivation of Runx2 blocked the transition of alveolar fibroblasts towards pathological fibroblasts (Fig. 5l).

Our study demonstrated that both Lepr-Cre and Lepr-CreERT2 alleles labeled lung fibroblasts starting at the neonatal stage. These labeled fibroblasts became a key source of pathological fibroblasts during pulmonary fibrosis in two different injury models. In line with a previous study on myelofibrosis¹⁶, Lepr⁺ stromal cells seem to be a key contributor to fibrosis in both lung and bone marrow. Our scRNA-seq analysis revealed that Cthrc1 and Postn were expressed in the pathological fibroblasts derived from Lepr-Cre or Lepr-CreERT2 labeled cells, suggesting that these Lepr⁺ fibroblasts are significant contributors of the pathological fibroblasts identified previously⁹. Consistently, we observed the expansion of Lepr-Cre or Lepr-CreERT2 labeled cells in bleomycin- and silica-treated lungs, and genetic ablation of Postn-CreER labeled pathological fibroblasts attenuated fibrosis. Previous studies showed that fibrosis is attenuated in the lungs of mutants lacking the Postn gene^44,45, suggesting that Postn plays a functional role during fibrotic changes. Of note, very recently Tatsuya et al. showed that Scube2-CreERT2 labeled cells differentiate into Postn⁺ Cthrc1⁺ pathological fibroblasts²⁶, similar to the Lepr-CreERT2 labeled cells described here. Lepr is broadly expressed in several mesenchymal cell types, and our Lepr-Cre and Lepr-CreER label a broad collection of cells with alveolar fibroblasts constituting a predominant proportion. In the future it will be interesting to explore whether Lepr^- mesenchymal cells also meaningfully contribute to pulmonary fibrosis.

Our scATAC-seq and computational analysis identified RUNX2 as a critical transcription regulator of fibrotic drivers. In support of this, deletion of Runx2 with Lepr-CreERT2 or Scube2-CreERT2 reduced the generation of pathological fibroblasts. RUNX2 mediates TGF-β signaling function during bone development^46,47, and overexpression of Runx2 leads to fibrotic changes in vascular smooth muscle cells with increased expression of Col1a1 and Col1a2⁴⁸. We found that RUNX2 is also important for the expression of fibrotic genes in fibrotic fibroblasts isolated from human IPF samples. Interestingly, a previous study shows that RUNX2 is increased in AT2 cells but reduced in fibroblasts of IPF samples as measured by microarray and immunofluorescence staining⁴⁹. The study also demonstrates that siRNA-mediated knockdown of RUNX2 promotes the expression of COL1A1 and ACTA2 in TGF-β1-stimulated fibroblasts⁴⁹. By contrast, our scRNA-seq results indicate that Runx2 is activated during the conversion of alveolar fibroblasts into pathological fibroblasts in both mouse models and IPF samples. Genetic deletion of Runx2 blocked the conversion and reduced lung fibrosis. Previously TBX4 and PU.1 have also been shown to regulate the differentiation of fibroblasts, contributing to pulmonary fibrosis^6,50. Interestingly, our scRNA-seq analysis showed that the level of Tbx4 transcript was similar between alveolar fibroblasts and pathological fibroblasts. Moreover, pathological fibroblasts (Cthrc1⁺Postn⁺) exhibited minimal expression of PU.1 (data not shown), suggesting that PU.1 acts in different cell populations during pulmonary fibrosis. Further elucidation of the relationship and function of each player during fibrosis should offer new opportunities to intervene in the aggressive progression towards end stage disease.

In summary, we demonstrated that Lepr⁺ cells arising at the early postnatal stage included alveolar fibroblasts as major derivatives. Cell-fate mapping combined with scRNA-seq analysis confirmed that Lepr⁺ fibroblasts became a major contributor to pathological fibroblasts during fibrosis. Deletion of Runx2 blocked the differentiation of alveolar fibroblasts into pathological fibroblasts, offering a potential therapeutic target to treat the deadly pulmonary fibrosis.

Methods

Human lung tissues collection

Lung tissues from healthy donors and IPF patients with end stage undergoing transplantation were obtained from the department of pathology at Columbia University Medical Center, under a protocol approved by the Columbia University Institutional Review Board (IRB: AAAS4094) and from Cedars Sinai Medical Center (IRB: Pro00032727). Informed consent was obtained from each donor or authorized representatives. All experiments using human lung tissues were performed in accordance with the approved protocol mentioned above.

Mice

Lepr-Cre⁵¹, Col1a1-EGFP⁵² and Runx2^flox/flox53 mouse strains have been previously described. Scube2-CreERT2 mouse strain was kindly shared by Dr. Dean Sheppard at UCSF. Lepr-CreERT2 and Higd1b-CreERT2 mouse strains were generated in the Que lab. See below for details. Postn-MerCreMer (#029645), Rosa26-tdtomato (#007914), Rosa26-DTA (#009669) and ACTB-Flpe (#005703) were purchased from The Jackson Laboratory. All mice used in the experiments were between the ages of 8 and 12 weeks and maintained on C57BL/6 background (both male and female). Mice were housed with a 12-hour light/dark cycle at 18–23 degree and 40–60% humidity in the animal facility at Columbia University Medical Center. To induce Cre recombinase activity in Higd1b-CreERT2 mouse strain, mice were intraperitoneally injected with 200 mg/kg body weight Tmx (Sigma, T5648) dissolved in sunflower oil. Lepr-CreERT2 and Scube2-CreERT2 mice were fed with Tmx-containing chow (0.5g/kg, inotivco, TD.130857) for two weeks to induce Cre recombinase activity followed by at least two-week washout time prior to exposure to bleomycin or silica. Sample sizes were determined empirically based on experience or previous published relevant studies. No blinding method was applied. Gender- and age-matched mice were randomly assigned to experiments. All mouse experiments and care were conducted in accordance with the procedures approved by the Institutional Animal Care and Use Committee at Columbia University (AABM6565).

Generation of Lepr^{CreERT2-IRES-EGFP} and Higd1b^{CreERT2-P2A-EGFP} mouse strains

To generate the Lepr^{IRES-CreERT2-P2A-EGFP} mouse strain, a targeting construct containing internal ribosome entry site (IRES), a cDNA encoding tamoxifen-inducible Cre recombinase (CreERT2), a P2A self-cleaving peptide sequence and EGFP followed by an FRT-flanked neomycin resistance cassette was generated and electroporated into KV1 (129-C57BL/6 hybrid) embryonic stem (ES) cells. The targeting vector was inserted immediate 3’ of the stop codon in the last exon of the Lepr gene by homologous recombination. Following G418 selection, the targeted ES clones were validated by PCR analysis and injected into C57BL/6N blastocysts to generate chimeras. The chimeras were crossed with ACTB-Flpe mice to identify germline transmission of the targeted allele and to remove the neomycin cassette. A similar strategy was used to generate the Higd1b^{CreERT2-P2A-EGFP} knockin mouse strain, but the targeting vector was inserted to replace the start codon in exon 2 of the Higd1b locus. The targeting construct for Higd1b^{CreERT2-P2A-EGFP} contains a cDNA encoding tamoxifen-inducible Cre recombinase (CreERT2), a P2A self-cleaving peptide sequence and EGFP followed by an FRT-flanked neomycin resistance cassette.

Lung injury mouse models

8-week-old C57BL/6 mice were used for lung injury models. Intratracheal administration of bleomycin or silica was performed as previously described^54,55. Briefly, mice were anaesthetized and 1.75 unit/kg bleomycin (Fresenius Kabi, USP) or 200 mg/kg silica suspension in PBS was delivered through intratracheal injection with a 30-Gauge needle. Lung tissues were harvested at the indicated time points.

Tissue preparation and histology

Mice were euthanized with isoflurane and lungs were inflated and fixed with 4% paraformaldehyde overnight. Lung tissues were dehydrated and processed as previously described^23,54. Sections at 7-µm were cut and collected for further histology staining and immunostaining. Hematoxylin and Eosin (H&E) staining was performed as previously described^23,54. Picro-Sirius Red staining was performed according to the instruction of commercial kit (VitroVivo Biotech, VB-3017). Leica Aperio AT2 Microscope slide scanner was used to obtain whole section images.

Hydroxyproline assay

The hydroxyproline content in lung tissues was measured by a commercial hydroxyproline assay kit (Cell Biolabs, STA-675) as previously described⁵⁶. In brief, the lung tissues were homogenized in distilled water, and the samples were mixed with 12 N hydrochloric acid and incubated for 24 hours at 95°C to hydrolyze the homogenized tissue. Following clarifying the hydrolyzed samples through a 0.45 um syringe filter, the samples were added to tubes and dried in an oven to remove the residual hydrochloric acid. After incubation with Chloramine T Mixture for 30 minutes at room temperature, the samples were incubated with Ehrlich’s Reagent for 45 minutes at 60°C. Following incubation at 4°C for 5 minutes, the samples were centrifuged at 6000g for 15 minutes at room temperature. The supernatants were transferred to microplate wells and the absorbance was measured on a microplate reader using 540–560 nm as the primary wavelength.

Immunofluorescence staining

Immunostaining was performed as previously described^23,54,57. In brief, paraffin sections were dewaxed and rehydrated through gradient ethanol. Antigen retrieval was performed with high-pressure heating in a commercial antigen unmasking solution (Vector Laboratory, H-3300) for two minutes. The sections were washed in PBS, permeabilized and blocked with blocking buffer (0.2% Triton X-100 and 5% normal donkey serum in PBS) for one hour at room temperature. The sections were incubated with primary antibodies anti-α-SMA (Santa Cruz, sc-32251, 1:200), anti-tdtomato (biorbyt, orb182397, 1:1000), anti-RFP (Rockland, 600–401-379, 1:500), anti-Collagen I (SouthernBiotech, 1310–01, 1:200), anti-Pi16 (R&D systems, AF4929, 5 µg/ml), anti-Cthrc1 (MaineHealth Institute for Research, Vli55, 1:250), anti-Postn (Abcam, ab215199, 1:200), anti-ERG (Abcam, ab92513, 1:200), anti-Endomucin (Santa Cruz, sc-65495, 1:200), anti-Runx2 (Cell Signaling Technology, 12556S, 1:200), anti-Ki67 (Cell Signaling Technology, 9129S, 1:200), anti-NG2 (Millipore Sigma, AB5320, 1:200), anti-ProSPC (Abcam, ab211326, 1:500) diluted in blocking buffer at 4°C overnight. Following extensive washing with PBS three times, the sections were incubated with Fluorophore-conjugated secondary antibodies for two hours at room temperature. DAPI was used to counterstain the nuclei. After washing with PBS, the sections were mounted using Fluoromount-G (SouthernBiotech, 0100–20). Zeiss LSM T-PMT confocal laser-scanning microscope was used for obtaining the images.

RNA in situ hybridization

RNA in situ hybridization was performed using the RNAscope multiplex fluorescent detection kit v2 (Advanced Cell Diagnostics, 323100) as previously described⁵⁸. Briefly, pre-baked paraffin sections were dewaxed and rehydrated, then treated with the hydrogen peroxide solution for 10 min at room temperature. After target retrieval for 15 min in 95–105°C solution, the sections were incubated with protease for 30 min at 40°C. Mm-Scube2 probe (Advanced Cell Diagnostics, 488141) were hybridized for two hours at 40°C, followed by signal amplification steps. Fluorophore (Akoya Biosciences, FP1487001KT) was incubated for 30 min at 40°C. The sections were incubated with DAPI for counterstaining and then mounted using Fluoromount-G solution. Zeiss LSM T-PMT confocal laser-scanning microscope was used for obtaining the images.

Single cell isolation and flow cytometry analysis

To obtain single cell suspension, lung tissues were dissected, washed with PBS and minced with a razor blade followed by digesting in a digestion buffer (2 mg/ml Collagenase-IV, 2 mg/ml Dispase II and 10 U/ml DNase I) for 30 min at 37°C. DMEM containing 10% fetal bovine serum (FBS) was added to stop digestion and the cells were filtered through 100-µm and 40-µm strainers. After centrifugation at 300g for 5 min, the cell pellet was resuspended in red blood cell lysis buffer (Sigma), incubated for 2 min at 37°C, followed by washing with HBSS containing 10% FBS and centrifuged at 300g for 5 min. Flow cytometry analysis was performed as previously described^23,54. In brief, cells were incubated with PE-Cy7-CD45 (Biolegend, 103114, 1:100), APC-EpCAM (Biolegend, 118214, 1:100) and BV711-CD31 (Biolegend, 102449, 1:100) antibodies in FACS buffer (5% FBS with 0.5mM EDTA in PBS) for one hour at 4°C, and then incubated with Live/Dead stain dye for 10 min at room temperature to exclude dead cells. After washing with FACS buffer, the cells were fixed and permeabilized by using Fixation/Permeabilization buffer and then incubated with Alexa Fluor 488-α-SMA antibody (eBioscience, 53–9760-82, 1:100) at 4°C for one hour. BD LSRII and FlowJo V10 software were used for obtaining data and analysis, respectively. Live CD45^-EpCAM^-CD31^- mesenchymal cells were gated for further analysis. For Single cell RNA sequencing (scRNA-seq), single cell suspension was applied to sort live tdTomato⁺ cells from Lepr-Cre;R26tdT mice or live tdTomato⁺CD45^-EpCAM^-CD31^- mesenchymal cells from Lepr-CreERT2;R26tdT mice by using BD Influx cell Sorter. For bone marrow monocytes and neutrophils analysis, the single cell suspensions were obtained from femur bone marrow as previously described^59,60. Bone marrow cells were incubated with eFluor 450-CD45 (eBioscience, 48–0451, 1:100), Alexa Fluor 700-Ly6G (Biolegend, 127622, 1:100), Alexa Fluor 488-Ly6C (Biolegend, 128022, 1:100) and PE-Cy7-CD11b (eBioscience, 25–0112, 1:100) antibodies in FACS buffer for one hour at 4°C. The flow cytometry data were collected on BD LSRII and analyzed using FlowJo V10 software. CD45⁺Ly6G^-CD11b⁺Ly6C⁺ monocytes and CD45⁺Ly6G⁺CD11b⁺Ly6C⁺ neutrophils were gated for analysis.

Single cell RNA sequencing analysis

Sorted tdTomato⁺ cells were loaded onto a Chromium Controller instrument (10X Genomics) in Single Cell Analysis Core of Genome Center at Columbia University. 10X Single Cell 3’ V2 and V3 chemistry kit were used to produce single-cell barcoded droplets and prepare libraries. Illumina NovaSeq 6000 instrument was used to sequence the resulting libraries and obtain the fastq files. Reads were aligned to a custom reference containing mouse genome GRCm38/mm10 with tdTomato-WPRE-polyA sequence and the unique molecular identifier (UMI) counts were obtained by Cell Ranger v3.1.0 and v7.1.0 software. We re-analyzed the publicly available scRNA-seq data of developing mouse lung¹⁷ (GSE160876 and GSE165063), normal adult mouse lungs^9,18,19 (GSE132771, GSE201698 and GSE211713), bleomycin-challenged mouse lungs^9,19,36,37 (GSE131800, GSE132771, GSE183545 and GSE201698), silica-exposed mouse lungs³¹ (GSE184854), CCl₄-treated mouse liver³⁸ (GSE171904) and human IPF patients⁴¹ (GSE136831). R package Seurat^61,62 v4.4.0 was used for further analysis by importing the raw counts matrices. Low-quality cells were excluded as determined by the number of gene transcripts, UMI counts and the percentage of mitochondrial transcripts. UMI counts normalization and variable features identification were performed by using Seurat NormalizeData and FindVariableFeatures functions. Two sample objects were integrated by identifying integration anchors using Seurat FindIntegrationAnchors and IntegrateData. Seurat merge and RunFastMNN functions were used to integrate multiple Seurat objects with batch correction. ScaleData, Principal Component Analysis (PCA) and non-linear dimensional reduction Uniform Manifold Approximation and Projection (UMAP) were applied to perform dimensional reduction. Cell clusters were identified using Seurat FindNeighbors and FindClusters functions. The Seurat FindAllMarkers function was used with the default parameters to identify the differentially expressed genes for each cluster. Cell clusters were manually annotated according to the marker genes in each cluster. Fibroblast clusters were selected for re-clustering and four specific clusters including alveolar fibroblasts, adventitial fibroblasts, pathological fibroblasts and proliferating fibroblasts were identified based on the reported marker genes. To obtain the module scores for feature expression, the average expression levels of gene signatures on a single-cell level were calculated by the AddModuleScore function. The differentially expressed genes (DEG) between “Alveolar fibroblast” and “Pathological fibroblast” were identified by the Seurat FindMarkers function. R packages Monocle2⁶³ and Monocle3⁶⁴ were used to perform pseudotime analysis, specifying “Alveolar fibroblast” as roots of the pseudotime. Enrichr program^33,65,66 was used to predict transcription factors and signaling pathway by importing the enriched genes in “Pathological fibroblast”.

Single cell Assay for Transposase-Accessible Chromatin sequencing (scATAC-seq) analysis

For scATAC-seq, 8-week-old C57BL/6 wildtype mice were intratracheally administered with 1 unit/kg bleomycin or saline after anesthesia with ketamine (100 mg/kg) and xylazine (10 mg/kg). At day 14, the lungs were perfused with 12 ml of cold DPBS (Life Technology) and then inflated with dissociation buffer (PRMI1640 (Thermo Scientific) with 10% FBS, 1mM HEPES (Life Technology), 1mM MgCl2(Life Technology), 1mM CaCl2 (Sigma-Aldrich), 0.525 mg/ml collagenase D (Roche), 5 unit/ml Dispase (Stemcell Technologies) and 0.05 mg/ml DNase I (Roche)). Minced lung was placed in dissociation buffer for 30 min at 37°C, and cell solution was filtered through 70μm and 40μm strainers. Ammonium-Chloride-Potassium (ACK) lysis was used to remove blood cells, and cell pellets were resuspended in FACS buffer. Antibodies used for flow cytometry include BV510-CD45 (Biolegend, 103138, 1:100), PE-Epcam (Biolegend, 118206, 1:100), PE-CD31 (Biolegend, 102408, 1:100), APC-PDGFRa (Biolegend, 135908, 1:100), and DAPI to label dead cells. 18,000 cells from each group were subjected to scATAC-seq following standard protocol by Center for Epigenomics at the University of California, San Diego. Sample preparation and library construction were performed as described previously⁶⁷. For data analysis, reads alignment (mouse genome mm10) and cell barcode demultiplex were conducted by using 10x Genomics Cell Ranger ATAC (version 2.1.0) with default settings. Quality control, TSS enrichment, data normalization, dimensional reduction and UMAP-based cell clustering were performed by Signac⁶⁸. Analysis of differentially accessible peaks was performed by FindAllMarkers. ChromVAR³⁵ was used to analyze differential TF binding motif activities between groups of cells. The expression features of marker genes, as well as the projection of motif enrichment scores, was profiled and visualized by R package ggplot2 (version 3.5.1).

Cell migration assay

5 × 10⁴ sorted tdT⁺EGFP⁺ and tdT^-EGFP⁺ lung fibroblasts without mycoplasma contamination in DMEM supplemented with 10% FBS were added into the top chamber of BioCoat Control cell insert. To trigger cell migration, 10 ng/ml PDGF-BB was added to the bottom chamber. After incubation for 24 hours in an incubator at 37°C with 5% CO₂, the cells were fixed with 4% PFA for 30 min at room temperature. The cells on the upper side of the insert membrane were removed by scrubbing with a cotton swab. DAPI was used for counterstaining. Images were obtained by EVOS M5000 system.

RNA extraction and quantitative reverse transcription-polymerase chain reaction (qRT-PCR)

Total RNA extraction was performed according to the instruction of the commercial RNA extraction kit (Qiagen, 74134). SuperScript III First-Strand SuperMix kit (Thermofisher, 18080400) was used to synthesize the first-strand cDNA from RNA. To quantify cDNA, iTaq Universal SYBR Green Supermix (Bio-rad, 1725122) was used according to the manufacturer’s instruction on QuantStudio 5 Real-Time PCR system. At least three technical and biological replicates were performed. The primer sequences used in this study are listed in Supplementary Table 4.

siRNA-mediated knockdown of RUNX2 in human normal lung fibroblasts treated with TGF-β1 and in human IPF fibroblasts

Human normal and IPF lung fibroblast isolation and culture were performed as previously described⁶⁹. The isolated human lung fibroblasts were tested to be negative for mycoplasma contamination. Human normal lung fibroblasts were transfected with RUNX2 siRNA (Santa Cruz, sc-37145) or a non-targeting control siRNA (Santa Cruz, sc-37007) using Lipofectamine RNAiMAX Reagent (Thermofisher, 13778) according to manufacturer’s instruction while being treated with 5 ng/ml TGF-β1 for 48 hours. Briefly, Lipofectamine RNAiMAX Reagent and siRNA were diluted in Opti-MEM medium. Diluted siRNA was added to Lipofectamine RNAiMAX Reagent with a 1:1 ratio before applying to the cells. The cells were incubated in an incubator at 37°C with 5% CO₂. Human IPF lung fibroblasts were also transfected with RUNX2 siRNA or a negative control siRNA Cells were incubated in the mixture for 48 hours and harvested for further analysis. The oligonucleotides sequences of siRNA are listed in Supplementary Table 5.

Quantification and statistical analysis

Whole slide digital images of lung lobes were used to quantify the fibrotic area of pulmonary fibrosis. For quantification of tdTomato⁺ cells, EGFP⁺ cells, α-SMA⁺ cells, Cthrc1⁺ cells and Runx2⁺ cells, at least 10 random fields (20x magnification or 1mm²) were captured and ImageJ v1.51 software was used to count the positive cells or measure the area of α-SMA⁺ cells as previously described⁷⁰. Each group included at least three replicates for all experiments. All data are presented as means±SEM using GraphPad Prism 8. Unpaired two-tailed t-test with Welch's correction and two-sided Wilcoxon Rank Sum test were used to determine statistical significance. For multiple comparisons, two-way ANOVA was used with Sidak’s correction. P < 0.05 or less were considered statistically significant.

Extended Data

Extended Data Fig. 1. Lepr⁺ mesenchymal cells in the developing and adult mouse lung.

a-b, UMAP plot showing the different cell populations (a) in the lungs at E18.5, P0, P3, P7 and P14 (b). Reanalysis of the datasets GSE160876 and GSE165063. c, UMAP plot showing expression of Lepr in the developing lungs. d, UMAP plot showing the integration of lung mesenchymal cells at E18.5, P0, P3, P7 and P14. e, Dot plot showing the representative markers for each mesenchymal cell population. f, tdTomato (tdT) expression in the lungs of Lepr-Cre;R26tdT mice at different postnatal stages. P: Postnatal. g, tdT and α-SMA expression in the lungs of Lepr-Cre;R26tdT mice at different postnatal stages. The arrows indicate rare tdT⁺ smooth muscle cells. aw: airway, bv: blood vessel. h, Schematic depicting the generation of Lepr-CreERT2 mouse strain. i-j, Schematic diagram for Tmx injection and the representative immunostaining image of tdT in the lungs of Lepr-CreERT2;R26tdT mice at P14 (i) and P20 (j). k, Schematic depicting the use of Tmx-containing chow to feed Lepr-CreERT2;R26tdT mice and representative immunostaining images of surfactant protein C (SPC) and tdT in lung tissues. l, UMAP plot showing different mesenchymal populations from the adult control lungs that integrate the datasets GSE132771, GSE201698 and GSE211713. m, Dot plot showing the representative markers for each mesenchymal population. n-o, UMAP plot showing expression level of Lepr (n) and Scube2 (o). p, Blended feature plots showing co-expression of Lepr and Scube2. q, Representative images of tdT immunostaining and Scube2 RNA in situ hybridization. r-s, Quantification analysis showing the percentage of Scube2⁺ tdT⁺ cells in total tdT⁺ cells (r, n = 5) and Scube2⁺ alveolar fibroblasts (s, n = 5). Data are representative of at least three independent experiments. Data are mean±SEM. Scale bars: 100 µm (20 µm in magnified views).

Extended Data Fig. 2. Lepr-Cre labeled cells contribute to pathological fibroblasts.

a, Representative images of migrated tdT⁺EGFP⁺ and tdT^-EGFP⁺ lung mesenchymal cells triggered by PDGF-BB. b, Quantification of the migrated cells (DAPI⁺) in a (tdT⁺EGFP⁺: n = 3, tdT^-EGFP⁺: n = 3). c, Relative expression of matrix genes in tdT⁺EGFP⁺ and tdT^-EGFP⁺ lung mesenchyme (qRT-PCR) (tdT⁺EGFP⁺: n = 3, tdT^-EGFP⁺: n = 3). d, Expression of Acta2, Col1a1 and Ccn2 in tdT⁺EGFP⁺ and tdT^-EGFP⁺ lung mesenchymal cells treated with/without TGF-β1 (qRT-PCR) (n = 3 for each group). e, tdT and EGFP expression in the lungs of Lepr-Cre;R26tdT;Col1a1-EGFP mice challenged with saline or bleomycin. f, Quantification of tdT⁺ cells (Saline: n = 6, Bleomycin: n = 6). g-h, Quantification of the proportion of α-SMA⁺tdT⁺ cells among tdT⁺ cells (g) and among α-SMA⁺ cells (h) (Saline: n = 6, Bleomycin: n = 6). i, Gating strategy to identify α-SMA⁺ tdT⁺ cells in live mesenchymal cells. j, FACS analysis of tdT⁺ and α-SMA⁺ cells isolated from the lung of Lepr-Cre;R26tdT mice treated with saline or bleomycin. k, Flow cytometric quantification for the percentage of α-SMA⁺ tdT⁺ cells in the lungs of Lepr-Cre;R26tdT mice treated with saline or bleomycin (Saline: n = 3, Bleomycin: n = 4). l, Experimental schematic and representative H&E images. m, α-SMA⁺ tdT⁺ cells in the lungs of Lepr-Cre;R26tdT mice exposed to saline or silica. n, Quantification of α-SMA⁺ tdT⁺ cells (Saline: n = 6, Silica: n = 6). o, Experimental schematic and representative H&E images. p, α-SMA and tdT expression in lung tissues. q, Quantification of α-SMA⁺ tdT⁺ cells in the lungs of Lepr-CreERT2; R26tdT mice following silica exposure (Saline: n = 6, Silica: n = 6). Data are representative of at least three independent experiments. Data are mean±SEM. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction (b, c, f, g, h, k, n, q) and two-way ANOVA multiple comparisons test with Sidak’s correction (d). Scale bars: l, o: 200 µm, a, e, m, p:100 µm (e, m, p: 20 µm in magnified views).

Extended Data Fig. 3. scRNA-seq reveals the contribution of Lepr-Cre labeled alveolar fibroblasts to pathological fibroblasts.

a, Schematic diagram for scRNA-seq experimental design. b, UMAP plot showing the transcript levels of tdTomato. c, UMAP plot showing different cell populations. d, Red dots showing the cells from saline-treated lungs and blue dots showing the cells from bleomycin-treated lungs. e, Dot plot showing the representative markers for each population. f, UMAP plot showing four different fibroblast populations. g, Red and blue dots showing the cells from saline- and bleomycin-treated lungs, respectively. h, Dot plot showing the representative markers for each fibroblast population. i, Frequency of each fibroblast population in saline and bleomycin-treated lungs. j, UMAP plot showing the score of alveolar fibroblast signature. k-l, UMAP plot showing the expression level of the alveolar fibroblast markers Npnt (k) and Scube2 (l). m, UMAP plot showing the adventitial fibroblast signature score. n, UMAP plot showing the expression level of the adventitial fibroblast marker Pi16. o, Representative immunostaining image of Pi16 and tdT in the adventitial cuffs. aw: airway. p, UMAP plot showing the pathological fibroblast signature score. q, UMAP plot showing the expression level of the pathological fibroblast marker Cthrc1. r, Representative immunostaining image of Postn and tdT in the lungs of Lepr-Cre;R26tdT mice treated with bleomycin. s, UMAP plot showing pseudotime analysis with Monocle3. t-u, Trajectory analysis with Monocle2 showing cells ordered by pseudotime (t) and cell type (u). v, UMAP plot showing the expression of Lepr in the mesenchymal cells of the lungs isolated from Lepr-Cre;R26tdT mice challenged with saline or bleomycin. Data are representative of at least three independent experiments. Scale bars: o and r: 100 µm (o: 20 µm in magnified views).

Extended Data Fig. 4. scRNA-seq analysis reveals the heterogeneous lung fibroblasts labeled by Lepr-CreERT2.

a, Gating strategy to sort tdT⁺ lung mesenchymal cells (CD45^-EpCAM^-CD31^-) from the lungs of Lepr-CreERT2;R26tdT mice treated with bleomycin or saline for scRNA-seq. b, tdTomato transcript levels are shown by UMAP plot. c, UMAP plot showing multiple Lepr-CreERT2 labeled cell populations. d, Red and blue dots showing the cells from saline- and bleomycin-treated lungs, respectively. e, Dot plot showing the representative markers for each population. f, Violin plot showing the significantly increased fibrosis-associated genes upon bleomycin challenge. g, 3D UMAP plot showing four different Lepr-CreERT2 labeled fibroblast populations. h-i, UMAP plot showing the expression level of the alveolar fibroblast markers Npnt (h) and Inmt (i). j, UMAP plot showing the adventitial fibroblast signature score. k-l, UMAP plot showing the expression level of the adventitial fibroblast markers Dcn (k) and Pi16 (l). m, Schematic depicting for the use of Tmx-containing chow to feed Lepr-CreERT2;R26tdT mice and representative immunostaining image of Pi16 and tdT in the adventitial cuffs. aw: airway. bv: blood vessel. n, Schematic depicting for the use of Tmx-containing chow and bleomycin challenge of Lepr-CreERT2;R26tdT mice, and representative images of Postn⁺ tdT⁺ pathological fibroblasts in lung tissues. Data are representative of at least three independent experiments. Statistical analysis was performed using two-sided Wilcoxon Rank Sum test (f). Scale bars: m and n: 100 µm (20 µm in magnified views).

Extended Data Fig. 5. Lepr-CreERT2 labeled alveolar fibroblasts generate pathological fibroblasts.

a, The cell distribution of Lepr-CreERT2 labeled fibroblast clusters along with the color-coded pseudotime (upper panel). Heatmap showing dynamic gene expression in fibroblast clusters (lower panel). Representative genes (left) and enriched pathways (right) in alveolar fibroblasts and pathological fibroblasts are shown, respectively. b, The expression levels of the alveolar fibroblast makers Scube2, Npnt, Inmt and the pathological fibroblast markers Cthrc1, Postn along pseudotime analysis. c-d, The top ten GO terms (c) and KEGG pathways (d) associated with genes enriched in pathological fibroblasts. e, UMAP plot showing the enrichment of Tgfβ1 transcript in pathological fibroblasts. f, UMAP plot showing the expression of Lepr transcript in lung mesenchymal cells of Lepr-CreERT2;R26tdT mice challenged with saline or bleomycin.

Extended Data Fig. 6. A novel Higd1b-CreERT2 mouse strain specifically labels pericytes which rarely contribute to pathological fibroblasts during lung fibrosis.

a, UMAP plot showing different mesenchymal populations from the adult control lungs that integrate the datasets GSE132771, GSE201698 and GSE211713. b-c, UMAP plot showing the expression levels of Higd1b (b) and Cox4i2 (c) in pericytes. Of note, 88% pericytes expressed Higd1b. d, Schematic depicting the generation of Higd1b-CreERT2 mouse strain. e, Schematic diagram for Tmx injection of adult Higd1b-CreERT2;R26-tdT mice and representative images showing immunostaining of pericyte marker NG2 and tdT. f, Schematic diagram for Tmx injection and bleomycin challenge of adult Higd1b-CreERT2;R26-tdT mice and representative images showing immunostaining of α-SMA, ERG and tdT. g, Representative images of two examples showing immunostaining of Cthrc1 and tdT in the lungs of Higd1b-CreERT2;R26-tdT mice challenged with bleomycin. h, Quantification analysis showing the rare contribution of pericytes to Cthrc1⁺ pathological fibroblasts in the lungs of Higd1b-CreERT2;R26-tdT mice (n = 6). Data are representative of at least three independent experiments. Data are mean±SEM. i-j, Representative images showing immunostaining of tdT, Collagen I (i) and Ki67 (j) in the lungs of Higd1b-CreERT2;R26-tdT mice challenged with bleomycin. Scale bars: 100 µm (20 µm in magnified views).

Extended Data Fig. 7. Postn⁺ pathological fibroblasts are identified in silica-challenged lungs.

a, UMAP plot showing four different lung fibroblast populations in silica-treated lungs. The results were generated by re-analysis of the dataset GSE184854. b, Dot plot showing the representative markers for each fibroblast population. c-d, Postn expression shown by UMAP plot (c) and Violin Plot (d). e, Schematic diagram for Tmx injection and silica exposure of Postn-CreER;R26tdT mice and representative images showing immunostaining of α-SMA and tdT in lung tissues. f, Quantification analysis showing increased tdT⁺ cells upon silica exposure (Saline: n = 6, Silica: n = 6). g, Representative images showing immunostaining of α-SMA, ERG and Endomucin in lung tissues. Data are representative of at least three independent experiments. Data are mean±SEM. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction. Scale bars: 100 µm (20 µm in magnified views).

Extended Data Fig. 8. The TF-binding motif enrichment identified by scATAC-seq.

a, UMAP plot showing the expression of Runx2 transcripts in lung mesenchymal cells from Lepr-Cre;R26tdT mice challenged with saline or bleomycin. b, Violin plot showing the increased expression of Runx2 in the lungs of Lepr-CreERT2;R26tdT mice challenged with bleomycin. c, Violin plot showing the enrichment of Runx2 in pathological fibroblasts and proliferating fibroblasts of bleomycin-treated Lepr-CreERT2;R26tdT mouse lungs. d, Schematic depicting the experimental design for scATAC-seq. e, UMAP plot of scATAC-seq showing various lung mesenchymal populations. f, Blue and red dots showing the cells from saline- and bleomycin-treated lungs, respectively. g, Heatmap showing the representative markers for each lung mesenchymal population. h-i, UMAP plot showing various lung mesenchymal populations (h) from bleomycin-treated (blue dots) and control (red dots) lungs (i) that integrate the published scRNA-seq datasets GSE131800, GSE132771, GSE183545 and GSE201698. j, Dot plot showing the representative markers for each fibroblast population. k-l, UMAP plot (k) and Violin Plot (l) showing the enrichment of Runx2 transcripts in pathological fibroblasts. m, qRT-PCR analysis of Runx2 transcripts in FACs-sorted tdT⁺ cells isolated from the lungs of Lepr-Cre;R26tdT mice treated with bleomycin or saline (Saline: n = 3, Bleomycin: n = 3). n-o, UMAP plot showing the quiescent and active hepatic stellate cells (n) of control (red dots) and CCl₄-treated (blue dots) mice (o). The dataset GSE171904 was used for re-analysis. HSC: Hepatic stellate cell. p-q. UMAP plot showing the expression of Cthrc1 (p) and Runx2 (q) in hepatic stellate cells. Data are representative of at least three independent experiments. Data are mean±SEM. Statistical analysis was performed using two-sided Wilcoxon Rank Sum test (b) and unpaired two-tailed t-test with Welch's correction (m).

Extended Data Fig. 9. Conditional deletion of Runx2 attenuates bleomycin-induced lung fibrosis.

a,b, FACS analysis to quantify bone marrow monocytes among CD45⁺ cells (Lepr-Cre;Runx2^+/+: n = 3, Lepr-Cre;Runx2^f/f: n = 3). c,d, FACS analysis to quantify bone marrow neutrophils among CD45⁺ cells (Lepr-Cre;Runx2^+/+: n = 3, Lepr-Cre;Runx2^f/f: n = 3). e, Representative H&E staining images of untreated lungs from Lepr-Cre;Runx2^+/+ and Lepr-Cre;Runx2^f/f mice. f, Experimental schematic and representative H&E staining images of the lungs from Lepr-Cre;Runx2^+/+ and Lepr-Cre;Runx2^f/f mice after bleomycin challenge. g, Quantification of fibrotic areas in the lungs of Lepr-Cre;Runx2^+/+ (n = 9) and Lepr-Cre;Runx2^f/f mice (n = 9). h, Hydroxyproline content in the lungs of Lepr-Cre;Runx2^+/+ (n = 7) and Lepr-Cre;Runx2^f/f mice (n = 7) treated with bleomycin. i, α-SMA and SPC expression in the lungs of Lepr-Cre;Runx2^+/+ and Lepr-Cre;Runx2^f/f mice after bleomycin challenge. j, Quantification of α-SMA⁺ cells in the lungs of Lepr-Cre;Runx2^+/+ (n = 6) and Lepr-Cre;Runx2^f/f mice (n = 6). k, Experimental schematic and representative Picro-Sirius Red staining images. l, α-SMA and tdT expression in the lungs of Lepr-CreERT2;Runx2^+/+;R26tdT and Lepr-CreERT2;Runx2^f/f;R26tdT mice. m, Quantification of α-SMA⁺ cells in the lungs of Lepr-CreERT2;Runx2^+/+;R26tdT (n = 6) and Lepr-CreERT2;Runx2^f/f;R26tdT mice (n = 6). n, Experimental schematic and representative Picro-Sirius Red staining images. o, α-SMA and tdT expression in the lungs of Scube2-CreERT2;Runx2^+/+;R26tdT and Scube2-CreERT2;Runx2^f/f;R26tdT mice. p, Quantification of α-SMA⁺ cells in the lungs of Scube2-CreERT2;Runx2^+/+;R26tdT (n = 6) and Scube2-CreERT2;Runx2^f/f;R26tdT mice (n = 6). Data are representative of at least three independent experiments. Data are mean±SEM. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction. Scale bars: e, f, k and n:1 mm (e and f: 200 µm in magnified views, k and n: 100 µm in magnified views), i, l and o:100 µm (20 µm in magnified views).

Extended Data Fig. 10. Increased expression of RUNX2 in IPF lungs.

a, Separated UMAP plots showing fibroblasts from normal human lungs and IPF lungs. These results were generated through the re-analysis of the dataset GSE136831. b, Dot plot of scRNA-seq showing the representative markers for each fibroblast population. c-f, UMAP plot showing the expression levels of LEPR (c), SCUBE2 (d), COL1A1 (e) and ACTA2 (f). g, Violin plot of scRNA-seq showing the expression levels of ECM associated genes. h, Venn diagram showing the common and differential genes expressed in pathological fibroblasts of human IPF and bleomycin-treated mouse lungs. i, Increased transcript levels of RUNX2 in IPF lungs revealed by re-analyzing the bulk RNA seq dataset GSE134692 (Normal: n = 26, IPF: n = 46). Data are mean±SEM. Statistical analysis was performed using unpaired two-tailed t-test with Welch's correction.

Data availability

The scRNA-seq and scATAC-seq data generated in this study were deposited in Gene Expression Omnibus (GEO) (accession # GSE229523, GSE276546 and GSE278419). The publicly available scRNA-seq data for mouse lung development (accession # GSE160876 and GSE165063), normal adult mouse lungs (accession # GSE132771, GSE201698 and GSE211713), bleomycin-challenged mouse lungs (accession # GSE131800, GSE132771, GSE183545 and GSE201698), silica-exposed mouse lungs (accession # GSE184854), CCl₄-treated mouse liver (accession #GSE171904), human IPF patients (accession # GSE136831) and bulk RNA-seq data for human IPF patients (accession # GSE124685 and GSE134692) were used for analysis. Source data are provided with this paper.

Code availability

The codes used in this study for scRNA-seq and scATAC-seq analysis are available from the corresponding authors upon requests. No custom code was generated.