Conditional BCL-2 Expression in Fibroblasts Promotes Persistent Pulmonary Fibrosis which is Reversible by Therapeutic BCL-2 Inhibition

Abstract

Idiopathic pulmonary fibrosis (IPF) is a progressive, fatal lung disease that develops in response to chronic epithelial injury. Unlike injury-induced homeostatic lung repair during which fibroblasts undergo apoptosis and clearance, the lungs of IPF patients continue to accumulate apoptosis-resistant, pro-fibrotic, extracellular matrix-producing fibroblasts. Here, we show that prevention of PDGFRα⁺ fibroblast apoptosis by conditional BCL-2 expression leads to the emergence and persistence of senescent, pro-fibrotic fibroblasts along with enduring, pathologic fibrotic lung remodeling. Additionally, spatial transcriptomic studies of human IPF lungs confirmed the presence of senescent, BCL-2 expressing α-smooth muscle actin⁺ myofibroblasts in fibrotic regions. Of translational significance, selective BCL-2 inhibition with ABT-199 in fibrotic mice re-engaged the apoptotic pathway in fibroblasts, reduced senescence, and promoted fibrosis resolution and lung regeneration. Our findings suggest that sustained BCL-2 expression in fibroblasts prevents homeostatic lung repair, drives persistent fibrosis and is a therapeutically relevant target to reverse persistent pulmonary fibrosis.

Idiopathic pulmonary fibrosis is a fatal human disease driven by the accumulation of apoptosis-resistant fibroblasts that impede homeostatic lung repair. Here, the authors show that elevated BCL-2 expression in fibroblasts drives their survival and senescence prolonging fibrosis in mice, while BCL-2 inhibition reverses persistent fibrosis.

Introduction

Idiopathic pulmonary fibrosis (IPF) is an interstitial lung disease in which progressive scarring of the alveolar-capillary units leads to eventual respiratory failure¹. With a median survival of ~3–5 years following diagnosis and limited availability of effective therapies, the prognosis for IPF patients is bleak^2,3. IPF is thought to develop following injury to the alveolar and terminal bronchiolar epithelium followed by an aberrant repair process which leads to the abundant accumulation of pro-fibrotic fibroblasts and excessive deposition of collagen and other fibrotic matrix components^1,4. Unlike normal homeostatic repair⁵ where fibroblasts undergo apoptosis and clearance, fibroblasts develop resistance to apoptosis induction, gain a senescent phenotype and survive in fibroblastic foci in IPF patients^6–8. Understanding the fundamental mechanisms that lead to fibroblast resistance to apoptosis and the subsequent acquisition of a senescent phenotype could thereby provide insights into targeted therapeutic opportunities to reverse this key pathogenic event.

Previous studies have shown that both down-regulation of the cell surface death receptor, Fas^9–11, and expression of the anti-apoptotic proteins PTPN13, XIAP and BCL-2 contribute to apoptosis resistance in cultured human fibrotic lung fibroblasts^12–18. Among these, BCL-2 has been shown to block apoptosis by partnering with pro-apoptotic mitochondrial activators and pore formers preventing them from activating the mitochondrial apoptotic pathway^19,20. Increased levels of BCL-2 have been found in lung fibroblasts from IPF patients and in mouse models of pulmonary fibrosis^12,21,22. However, the mechanism by which fibroblast resistance to apoptosis and senescence contribute to persistent fibrotic lung disease remains largely unknown. To address this knowledge gap, we generated transgenic mice in which human BCL-2 can be conditionally expressed in fibroblasts. Herein, we demonstrate that BCL-2 expression in fibroblasts prevents lung repair and regeneration in bleomycin-instilled mice and leads to fibroblast persistence, acquisition of a senescent phenotype, augmentation of pro-fibrotic gene expression, abnormal lung remodeling and persistent pulmonary fibrosis. Of significance to the treatment of patients with IPF, our pre-clinical studies in mice show that BCL-2 inhibition with the FDA-approved BCL-2 BH3-mimetic drug ABT-199 (Venetoclax^TM) reversed fibroblast senescence, re-engaged mitochondrial apoptosis pathways, promoted fibrosis resolution and improved lung architecture and function.

Results

Conditional BCL-2 expression in fibroblasts inhibits apoptosis and promotes their persistence in bleomycin-injured lungs

We engineered mice in which human BCL-2 and the lineage tracer mCerulean3 cyan fluorescent protein (CFP) can be conditionally expressed through Cre-mediated recombination (Fig. 1a–c). We refer to this strain as CFP/BCL-2stop^fl/fl. We bred PDGFRα-CreERT2;CFP/BCL-2stop^fl/+ mice, which are referred to as PDGFRα-CFP/BCL-2⁺ mice following tamoxifen treatment. Lung fibroblasts from PDGFRα-CFP/BCL-2⁺ mice exhibit robust expression of predominantly mitochondrial BCL-2 and cytoplasmic CFP (Fig. 1b–d) and resisted apoptosis induction by both the mitochondrial and death receptor pathways (Fig. 1e,f). PDGFRα-CFP/BCL-2⁺ mice exhibit colocalization of BCL-2 and CFP in PDGFRα⁺ and PDGFRα⁺/β⁺ fibroblasts located in alveolar walls and septae (Fig. 1g and Supplementary Fig. 1). CFP was not detected in PDGFRβ⁺ (i.e., PDGFRα^-) pericytes, pro-SPC alveolar type II epithelial cells (AEC2), CCSP⁺ club cells, CD45⁺ alveolar macrophages or CD31⁺ endothelial cells, and had minimal expression in αSMA⁺ parabronchial or arterial smooth muscle (Fig. 1g and Supplementary Fig. 1). Using the flow cytometry strategy shown in Fig. 1h, CFP was abundantly detected in PDGFRα⁺ and PDGFRα⁺/β⁺ fibroblasts (Fig. 1i), but not in endothelial cells, epithelial cells, hematopoietic cells or PDGFRβ⁺ pericytes (Fig. 1i,j). CFP was not expressed in the absence of tamoxifen treatment (Fig. 1i,j). Thus, BCL-2 is specifically expressed in lung fibroblasts and blocks fibroblast apoptosis in PDGFRα-CFP/BCL-2⁺ mice.

Fig. 1. Lineage specific conditional expression of BCL-2 in mesenchymal cells.

a Schematic of the human BCL-2 construct and CFP/BCL-2stop^fl/+ mouse. b Primary lung fibroblasts from PDGFRα-CFP/BCL-2⁺ mice express hBCL-2 that co-localizes with Mitotracker Red (63X) (n = 3 experimental replicates). Additional images show overlays of phalloidin and CFP (40X). By Western Blot, (c) fibroblasts from PDGFRα-CFP/BCL-2⁺ mice express CFP (detected with anti-GFP antibody) and BCL-2 which is located within (d) the mitochondrial fraction (n = 3 experimental replicates). Caspase 3/7 activity in CFP/BCL-2^- and CFP/BCL-2⁺ fibroblasts after treatment with (e) staurosporine (n = 9 experimental replicates, +/−SD, 2-tailed t test with Welch’s correction, ***p < 0.001) or (f) Jo2 +/- TNF-α/IFNγ (n = 3 experimental replicates, +/-SD, 2-tailed t test with Welch’s correction, **p < 0.01). g endogenous CFP expression (teal) in naïve lungs of PDGFRα-CFP/BCL-2⁺ mice co-immunostained BCL-2 (red) or with lineage cell markers (red or yellow) for PDGFRα, PDGFRβ, αSMA, proSPC, CCSP, CD45 and CD31 (20x). h Flow cytometry gating of PDGFRα and PDGFRβ populations from lineage negative cells (CD45^-/CD31^-/EpCAM^-). Broad PDGFRα⁺ gating (pink gate) can be further sub populated into PDGFRα ⁺ only (CFP⁺, orange gate) and PDGFRα⁺/β⁺ (CFP⁺, black gate). PDGFRβ⁺ only pericytes (CFP^-, red gate) are PDGFRα^-. i Geometric mean of CFP expression in lineage sorted lung cell populations (n = 4 mice/group, +/-SEM, 2-tailed t test with Welch’s correction, ***p < 0.001) and (j) flow cytometry histograms. CO = corn oil, Tamox = tamoxifen. Source data for c, d, e, f and i are provided as a Source Data file.

Lung delivery of bleomycin in mice leads to acute epithelial injury followed by peak fibroblast accumulation and fibrosis at 2–3 weeks. The peak of fibrosis coincides with a wave of fibroblast apoptosis followed by fibrosis resolution and lung regeneration by 6–8 weeks^11,23. To investigate the consequences of BCL-2 expression on fibroblast apoptosis in vivo, PDGFRα-CFP/BCL-2^fl/+ mice were instilled with bleomycin or saline, treated with tamoxifen or corn oil (Fig. 2a) and apoptotic lung fibroblasts were identified by staining for PDGFRα, PDGFRβ and TUNEL at 3 weeks. Figure 2b,c shows that TUNEL⁺ fibroblasts were detected in bleomycin-instilled PDGFRα-CFP/BCL-2^- mice, but were significantly reduced (p < 0.01) in the fibrotic lungs of PDGFRα-CFP/BCL-2⁺ mice. TUNEL⁺ staining was negligible in the lungs of saline-instilled PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- mice.

Fig. 2. Conditional expression of BCL-2 in mesenchymal cells is sufficient to prevent their spontaneous apoptosis.

a Schematic of in vivo tamoxifen or corn oil treatment of PDGFRα-CFP/BCL-2^fl/+ mice. b Immunofluorescent staining of PDGFRα (magenta), PDGFRβ (green) and TUNEL (white) in PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ mice 3 weeks after bleomycin. c Quantification of TUNEL⁺ cells per high power field. Data is mean +/− SEM, n = 5. *p < 0.05, **p < 0.01, ***p < 0.001. Brown–Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons. Upper panels 20X, lower panels 40X. Source data for c is provided as a Source Data file.

PDGFRα⁺ fibroblasts and PDGFRβ⁺ pericytes have been implicated in the development of pulmonary fibrosis^24–27. To determine if BCL-2 expression led to increased lung mesenchymal cell numbers, we quantified PDGFRα⁺ fibroblasts, PDGFRα⁺/β⁺ fibroblasts and PDGFRβ⁺ pericytes in bleomycin- and saline-instilled PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- mice for up to 12 weeks (Fig. 3a). Figure 3b–d and Supplementary Fig. 2 show that PDGFRα⁺ fibroblasts, PDGFRα⁺/β⁺ fibroblasts and PDGFRβ⁺ pericytes were increased (p < 0.001) to similar extents in bleomycin-instilled mice of both genotypes at 3 weeks. However, while the numbers of PDGFRα⁺ fibroblasts, PDGFRα⁺/β⁺ fibroblasts and PDGFRβ⁺ pericytes returned to baseline by 6 weeks in PDGFRα-CFP/BCL-2^- mice, they remained significantly elevated for up to 12 weeks in PDGFRα-CFP/BCL-2⁺ mice (Fig. 3b–d). No differences in fibroblast or pericyte accumulation were observed when comparing PDGFRα-CFP/BCL-2^- and wild type C57Bl/6 mice (Supplementary Fig. 3a, b). The frequency of CFP expression in PDGFRα⁺ and PDGFRα⁺/β⁺ lung fibroblasts was >95% in saline and bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice and <10% in PDGFRβ⁺ pericytes (Fig. 3e–g). The observed deep penetrance of CFP in both PDGFRα⁺ and PDGFRα⁺/β⁺ fibroblasts suggests that these subsets were derived from PDGFRα⁺ cells.

Fig. 3. Conditional expression of BCL-2 in PDGFRα⁺ fibroblasts results in their persistence after fibrosis induction.

a Schematic of in vivo tamoxifen or corn oil treatment and time course of harvest in PDGFRα-CFP/BCL-2^fl/+ mice. b–d Quantification of PDGFRα ⁺ and PDGFRα⁺/β⁺ fibroblasts and PDGFRβ ⁺ pericytes over time. e–g Frequency of CFP ⁺ PDGFRα ⁺ and PDGFRα⁺/β⁺ fibroblasts and PDGFRβ ⁺ pericytes in PDGFRα-CFP/BCL-2⁺ mice after saline or bleomycin treatment. (h-i upper panels) CFP (teal), PDGFRα (magenta) and PDGFRβ (green) immunofluorescent staining of lungs over time after bleomycin in PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ mice. h–i lower panels) human BCL-2 (green) immunofluorescent staining of lungs over time after bleomycin in PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ mice. PDGFRα-CFP/BCL-2⁺ saline n = 4 mice/time point. PDGFRα-CFP/BCL-2⁺ bleomycin n = 5 mice/time point. PDGFRα-CFP/BCL-2^- saline n = 6 mice 0wk, 4 mice 3, 6, 12 wk, 5 mice 9 wk. PDGFRα-CFP/BCL-2^- bleomycin n = 5 mice/time point. Data is mean +/- SEM, Brown–Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons *p < 0.05, **p < 0.01, ***p < 0.001 compared to saline. #p < 0.05 9wk PDGFRα-CFP/BCL-2⁺ compared to other 9-week animals. ##p < 0.01, 6- and 12-week PDGFRα-CFP/BCL-2⁺ compared to other 6- and 12-week animals respectively. Upper panels 20X, lower panels 40X. Source data for b–g are provided as a Source Data file.

To determine the spatial localization of PDGFRα⁺ and PDGFRβ⁺ cells within fibrotic areas, we co-immunostained lung sections for PDGFRα, PDGFRβ and CFP. Additional sections were stained for human BCL-2. An expansion of both PDGFRα⁺ and PDGFRα⁺/β⁺ fibroblasts was observed to be coincidental with CFP staining. PDGFRβ⁺ (i.e., CFP-) pericytes were also noted within fibrotic areas in bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice at 3 weeks and persisted for up to 12 weeks (Fig. 3h, Supplementary Fig. 4). BCL-2 staining was also observed in fibrotic PDGFRα-CFP/BCL-2⁺ mice. As expected, neither CFP nor human BCL-2 were detected in the lungs of PDGFRα-CFP/BCL-2^- mice and the expansion of PDGFRα⁺ fibroblasts and PDGFRβ⁺ pericytes was only seen at 3 weeks post bleomycin (Fig. 3i). By 6 weeks, PDGFRα⁺ fibroblasts and PDGFRβ⁺ pericytes were detected in resolving fibrotic areas and by 9–12 weeks were once again observed in the alveolar walls of bleomycin-instilled PDGFRα-CFP/BCL-2^- mice, similar to naïve mice (Fig. 3i).

Conditional BCL-2 expression in fibroblasts leads to persistent pulmonary fibrosis and fibrotic remodeling

We next determined if the BCL-2 driven fibroblast persistence in bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice prevented lung repair and regeneration leading to sustained fibrosis and architectural remodeling. Assessment of lung hydroxyproline levels (Fig. 4a) revealed that bleomycin-instilled PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- mice developed similar levels of fibrosis at 3 weeks (p = 0.185). However, while fibrosis had largely resolved in PDGFRα-CFP/BCL-2^- mice at 6 weeks, it persisted in PDGFRα-CFP/BCL-2⁺ mice for up to 12 weeks (p < 0.01) (Fig. 4a). Histologic examination of Masson’s trichrome stained sections (Fig. 4b) from bleomycin-instilled PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- mice indicated that they developed similar airway distortion and alveolar collapse with patchy fibrosis, septal thickening and increased inflammation at 3 weeks. However, whereas lung regeneration was evident in PDGFRα-CFP/BCL-2^- mice at 6 weeks, PDGFRα-CFP/BCL-2⁺ mice exhibited non-resolving lung fibrosis, sustained pathologic airway distortion and architectural collapse, and areas of focal bronchiolization and cysts (Fig. 4b).

Fig. 4. Conditional expression of BCL-2 results in persistent fibrosis after bleomycin.

a Hydroxyproline levels in the lungs over time after bleomycin in PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ mice. PDGFRα-CFP/BCL-2⁺ saline n = 4 mice/time point. PDGFRαCFP/BCL-2⁺ bleomycin n = 5 mice/time point. PDGFRα-CFP/BCL-2^- saline n = 6 mice 0wk, 4 mice 3, 6, 12 wk, 5 mice 9 wk. PDGFRα-CFP/BCL-2^- bleomycin n = 5 mice/time point. Data is mean +/− SEM, **p < 0.01, ***p < 0.001 3wk bleomycin compared to saline. **p < 0.01 and ***p < 0.01 PDGFRα-CFP/BCL-2 compared to controls. 2-tailed t test with Welch’s correction. b Representative Trichrome stained lung sections over time in PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2⁺ mice. Blue arrows indicate cyst formation. c Quantitation of CCSP ⁺KRT8⁺ and (d) SPC⁺ cells per field over the time course. e Identification of CCSP⁺ (magenta) and Krt8⁺ (green) transitional cells in saline or bleomycin at 0, 3, 6, 9 and 12 weeks after bleomycin in PDGFRα-CFP/BCL-2⁺ (upper) and PDGFRα-CFP/BCL-2^- (lower) lungs. f Identification of pro-SPC⁺ (magenta) alveolar epithelial cells in PDGFRα-CFP/BCL-2⁺ (upper) and PDGFRα-CFP/BCL-2^- (lower) lungs. Aperio scans 2x. Immunofluorescence upper panels 20X, lower panels 40X. Source data for a, c and d are provided as a Source Data file.

The development of lung fibrosis is thought to involve initial epithelial injury which in turn drives fibroblast accumulation and pro-fibrotic fibroblast activation. The generation of PDGFRα-CFP/BCL-2⁺ mice provided an opportunity to determine if prolonging the lifespan of pro-fibrotic fibroblasts through conditional BCL-2 expression promotes aberrant epithelial cell accumulation, cyst formation and parenchymal bronchiolarization²⁸. We observed that the areas of remodeled lung in PDGFRα-CFP/BCL-2⁺ mice exhibited a significant accumulation of CCSP⁺KRT8⁺ cells at 3 weeks, which persisted up to 12 weeks (Fig. 4c,e upper panels). Additionally, staining for pro-SPC revealed a significant increase in SPC⁺ AEC2 cells throughout the fibrotic areas also persisting for up to 12 weeks in contrast to PDGFRα-CFP/BCL2^- mice, where AEC2 cells were localized to normal alveoli (Fig. 4d,f). Bleomycin-instilled PDGFRα-CFP/BCL2^- mice also showed limited numbers of KRT8⁺ cells in fibrotic areas and were absent during lung regeneration between 6 and 12 weeks (Fig. 4c,e lower panels). Naïve and saline-instilled PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- mice showed no histological abnormalities (Supplementary Fig. 5a,b), while bleomycin-instilled PDGFRα-CFP/BCL-2^- and control C57Bl/6 mice show identical patterns of fibrosis development and resolution (Fig. 4b lower panels and Supplementary Fig. 5c).

To validate the findings in PDGFRα-CFP/BCL-2⁺ mice, we bred Col1a1-CreERT2;CFP/BCL-2stop^fl/+ mice to allow tamoxifen-inducible conditional expression of BCL-2 in fibroblasts¹¹. Using the bleomycin and tamoxifen application timeline shown in Fig. 3a, bleomycin-instilled Col1a1-CFP/BCL-2⁺ mice were found to phenocopy PDGFRα-CFP/BCL-2⁺ mice with persistently elevated numbers of PDGFRα⁺ fibroblasts and PDGFRβ ⁺ pericytes, sustained elevations in lung hydroxyproline and histologically evident fibrosis with focal areas of bronchiolization and cysts for up to 12 weeks (Supplementary Fig. 6). Col1a1-CFP/BCL-2^- mice displayed similar fibrosis development at 3 weeks followed by lung regeneration at 6 weeks (Supplementary Figs. 5c and 6). Thus, using two Cre-drivers, conditional expression of BCL-2 in fibroblasts: (i) prevented normal, homeostatic fibroblast apoptosis following bleomycin-induced lung injury, (ii) led to the prolonged increases in PDGFRα⁺ fibroblasts and PDGFRβ⁺ pericytes, (iii) impeded lung repair and regeneration, and (iv) resulted in persistent pulmonary fibrosis along with aberrant epithelial cell differentiation and architectural remodeling.

Apoptosis resistant BCL-2 expressing fibroblasts exhibit sustained pro-fibrotic and senescence-related transcriptomic programming

In view of the findings presented above, we hypothesized that BCL-2 might exert additional pressure on fibroblast pro-fibrotic programming, beyond simply blocking apoptosis. To test this hypothesis, we isolated lung fibroblasts from naïve PDGFRα-CFP/BCL-2⁺ mice and bleomycin-instilled PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ mice at 3 and 6 weeks and conducted bulk RNA-sequencing and analysis (Fig. 5). Gene Ontology (GO) enrichment analysis of differentially expressed genes (DEGs) at 3 weeks showed similar enrichment in proliferation regulation genes in both PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- mice compared to naïve mice (Supplementary Table 1). Other pathways that were enriched in both mouse strains at 3 weeks included ECM structure, collagen fibril organization, collagen-containing ECM, and ECM organization, together with their associated genes (e.g., Col1a1, Col1a2, Col5a1, Col8a1, Col12a1, Eln, Fn1, Postn and Cthrc1) (Fig. 5a–c and Supplementary Figure 7). These findings suggest overlap in the pro-fibrotic programming patterns of lung fibroblasts at 3 weeks, though this was enhanced in PDGFRα-CFP/BCL-2⁺ mice. However, a notable difference between the two genotypes was that fibroblasts from PDGFRα-CFP/BCL-2⁺ mice showed significant enrichment in biological processes associated with positive regulation of cellular senescence (including transcripts encoding Cdkn2a (p16^Ink4a), Glb1 (senescence-associated β-galactosidase (SA-β-gal)) and p53 signaling (Tp53) although there was no significant difference in Cdkn1a (p21^Waf1) expression between PDGFRα-CFP/BCL-2⁺ and PDGFRα-CFP/BCL-2^- fibroblasts at 3 weeks. There was also an increased signature for the negative regulation of apoptosis, compared to fibrotic fibroblasts from PDGFRα-CFP/BCL-2^- mice (Fig. 5a,d,e,f, and Supplementary Tables 1 and 4).

Fig. 5. Bulk sequencing reveals an enhanced pro-fibrotic signature, reduced regulation of apoptosis and development of a sustained senescent phenotype in BCL-2 conditional-expressing fibroblasts.

a Dot plot representation of enriched Gene Ontology (GO) terms 3 weeks after bleomycin from PDGFRα-CFP/BCL-2⁺ fibroblasts compared to PDGFRα-CFP/BCL-2^- fibroblasts. b Differential gene expression associated with ECM organization (c) Box-and-whisker plots of ECM and pro-fibrotic genes expressed during non-resolving fibrosis in PDGFRα-CFP/BCL-2⁺ fibroblasts. Differential gene expression associated with (d) senescence regulation and (e) p53 signaling. f Box-and-whisker plots of key senescence and p53 associated genes in non-resolving PDGFRα-CFP/BCL-2⁺ fibroblasts. g Dot plot representation of enriched GO terms 6 weeks after bleomycin from PDGFRα-CFP/BCL-2⁺ fibroblasts compared to PDGFRα-CFP/BCL-2^- fibroblasts. Differential gene expression associated with (h) ECM disassembly and (i) box-and-whisker plots of genes expressed during resolving fibrosis in PDGFRα-CFP/BCL-2^- fibroblasts. Differential gene expression associated with (j) regulation of apoptotic signaling and (k) regulation of Wnt signaling in naïve, PDGFRα-CFP/BCL-2^-and PDGFRα-CFP/BCL-2⁺ fibroblasts. l Box-and-whisker plots of genes expressed during Wnt signaling in naïve, PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ fibroblasts. Heat maps generated with significant differentially expressed genes identified from GO BP pathways. c, f, i, l Box-and-whisker plots (n = 4 biological replicates/group, mean; whiskers: min to max) are normalized TPM values with pAdj from DE analysis *p < 0.05, **p < 0.01, ***p < 0.01. ECM, extracellular matrix. Avg NE = Average normalized expression. N Overlap = number of DEGs significantly expressed in the pathway. Source data for a–l are provided as a Source Data file.

Fibroblasts isolated from the fibrotic lungs of PDGFRα-CFP/BCL-2⁺ mice at 6 weeks showed continued enrichment in pro-fibrotic terms together with genes associated with positive regulation of cellular senescence and p53 signaling (Fig. 5g). By contrast, fibroblasts from the regenerating lungs of bleomycin-instilled PDGFRα-CFP/BCL-2^- mice at 6 weeks lacked enrichment in pro-fibrotic biological processes (Supplementary Fig. 7), but exhibited enhanced enrichment in pathways associated with ECM disassembly, wound repair (Wnt and Hippo signaling), cell migration, wound healing, epidermal cell differentiation and cell population proliferation compared to fibroblasts from PDGFRα-CFP/BCL-2⁺ mice (Fig. 5g,h,i,k,l and Supplementary Tables 2, 3 and 4). Additionally, PDGFRα-CFP/BCL-2^- fibroblasts, but not PDGFRα-CFP/BCL-2⁺ fibroblasts, were enriched for terms associated with the positive regulation of apoptotic signaling at 6 weeks compared to 3 weeks (Supplementary Table 2), consistent with the notion that these cells are capable of undergoing apoptosis in the absence of BCL-2 expression (Fig. 5j). Thus, whereas PDGFRα-CFP/BCL-2^- fibroblasts develop a transcriptomic program consistent with a beneficial role in lung repair and regeneration, conditional expression of BCL-2 in PDGFRα-CFP/BCL-2⁺ mice promoted lung fibroblast transcriptomic programs associated with fibrosis and senescence but had no enrichment in pathways associated with positive regulation of apoptosis signaling.

Further analysis of the cellular senescence GO terms identified enrichment of a senescence signature in PDGFRα-CFP/BCL-2⁺ fibroblasts compared to PDGFRα-CFP/BCL-2^- and naïve fibroblasts at 6 weeks (Fig. 6a). The composite senescence score (using 111 senescence associated genes²⁹) was increased in PDGFRα-CFP/BCL-2⁺ fibroblasts compared to PDGFRα-CFP/BCL-2^- fibroblasts (p = 0.06)(Fig. 6b). In view of this enrichment, we generated a network correlation plot between BCL-2 and the senescence genes. We identified a strong ( | r | ≥0.5) to very strong ( | r | ≥0.75) Pearson correlation between BCL-2 and 54 senescence genes (p ≤ 0.05) expressed in PDGFRα-CFP/BCL-2⁺ fibroblasts (Fig. 6c, Supplementary Table 5). We next investigated the location and abundance of senescent fibroblasts at 6 weeks by immunostaining for the markers p16^INK, p21^WAF and SA-β-gal, along with human BCL-2, PDGFRα and PDGFRβ (Fig. 6d–k). Expression of BCL-2, p16^INK, p21^WAF and SA-β-Gal in PDGFRα ⁺ cells were all significantly increased (p < 0.001) in the fibrotic areas of PDGFRα-CFP/BCL-2⁺ mice but were minimally detected in PDGFRα-CFP/BCL-2^- mice (Fig. 6d–k). Furthermore, all three senescence markers co-localized with BCL-2 (Fig. 6h–k).

Fig. 6. BCL-2 expressing fibroblasts develop a senescent phenotype.

a Heat maps of differential gene expression associated with regulation of cellular senescence in fibroblasts from PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ mice 6 weeks after bleomycin. b Senescence gene score (n = 4 mice/group, mean +/− SEM, *p < 0.01, Brown–Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons.) and (c) BCL-2 Pearson Correlation maps for 111 senescence associated genes in naïve, PDGFRα-CFP/BCL-2^-, PDGFRα-CFP/BCL-2⁺ fibroblasts. d–f Quantification of fibroblasts/field for p21, p16 and SA-β-gal from control and persistently fibrotic lungs from PDGFRα-CFP/BCL-2^-, PDGFRα-CFP/BCL-2⁺ mice (n = 4 mice/group, mean +/− SEM of 10 individual images/mouse graphed as scatter plot with bar mean +/− SEM, ***p < 0.001, 2-tailed t test with Welch’s correction.). Immunofluorescence images for (g) PDGFRα (magenta) and PDGFRβ (green), (h) human BCL-2 (green), (i) p21 (yellow), (j) p16 (magenta) and (k) SA-β-gal (white). SA-β-gal, senescence associated-β-galactosidase. White arrow heads indicate shared features in serial sections. Spatial transcriptomic analysis of IPF lung. l H&E section (m) spatial map of cellular populations (n) BCL-2 expression map (red, individual transcripts) overlayed with myofibroblasts (yellow, cell boundaries) and (o) 10 gene senescence module ENV score. [A-C dotted line callouts are enlarged areas to show cell boundary, BCL-2 expression and senescence ENV score overlays]. p Heat maps of differential pro-fibrotic genes and senescence genes between BCL-2⁺ myofibroblasts and BCL-2^- myofibroblasts from the spatial transcriptomic data. q Myofibroblast cell counts between control IPF and BCL-2⁺ myofibroblast⁺ counts from control and IPF spatial images (*p < 0.05 and ***p < 0.001, 2-tailed t test with Welch’s correction). r Quantification of α-SMA ⁺ myofibroblasts/field for BCL-2 in control and IPF lung sections (individual images graphed as scatter plot with bar mean +/- SEM, ***p < 0.001, 2-tailed t test with Welch’s correction). Immunofluorescent images of control and IPF lungs stained for (s) αSMA (magenta) and BCL-2 (green), (t) α-SMA (green) and p16 (magenta), (u) p21 (yellow), (v) SA-β-Gal (white). Images n = 5 subjects/group. Spatial transcript analysis n = 13 controls, n = 15 IPF Immunofluorescence panels 20X. 10 images/mouse for immunofluorescent quantitation. SA-β-gal, senescence associated-β-galactosidase. Cor coe = correlation coefficient. Source data are provided for a–f and p–r as a Source Data file.

To determine the clinical relevance of our findings in PDGFRα-CFP/BCL-2⁺ mice, we performed spatial transcriptomics on IPF and non-diseased lung tissues (Fig. 6l,m). Myofibroblasts (yellow = cell boundaries, cluster defined by COL1A2, ACTA2, IGFBP5,COL6A1, COL6A3, and IGFBP3) were present within fibrotic foci with co-localized expression of BCL2 (red = transcripts) (Fig. 6n). Areas with a comprehensive senescence score (based on 10 genes associated with cell senescence; ATF3, C11orf96, CDKN1A, DUSP1, FOS, GADD45B, IL6, JUN, JUNB, ZFP36) were spatially-associated with BCL-2⁺ myofibroblasts (Fig. 6o). BCL-2⁺ myofibroblasts had enriched expression of profibrotic and senescence-associated genes compared to BCL-2^- myofibroblasts (Fig. 6p) and were significantly increased in IPF lungs compared to control lungs (Fig. 6q). We immunostained and quantified human non-diseased and IPF lung tissue for (1) BCL-2 and αSMA⁺ myofibroblasts (Fig. 6r, s), which are the predominant apoptosis-resistant fibroblast subset in persistent human fibrotic lung diseases^8,30, and (2) senescence associated proteins (Fig. 6t–v). Minimal staining of Bcl-2, αSMA, p21^WAF, p16^INK, or SA-β-Gal was detected in non-diseased lung tissue (Fig. 6s–v). However, abundant p21^WAF, p16^INK, SA-β-Gal and Bcl-2 staining was observed in α-SMA⁺ fibroblasts in lung tissues from IPF patients (Fig. 6s–v). We confirmed this protein data by querying two publicly available scRNA-seq datasets, namely the IPF Cell Atlas (GEO:GSE135893³¹) and the Integrated Human Lung Cell Atlas³². We observed increased expression of senescence-associated genes CDKN1A, CDKN2A, GLB1 in fibrosis-associated mesenchymal cells, including αSMA⁺ myofibroblasts, in ILD lung tissues compared to non-diseased control subjects (Supplementary Fig. 8). Notably, BCL2 was expressed in all fibrotic mesenchymal cell populations, but at low levels in these end-stage IPF samples (Supplementary Fig. 8). Thus, conditional expression of BCL-2 in mouse lung fibroblasts resulted in the development of pro-fibrotic and senescent fibroblast phenotypes in the persistently fibrotic lungs of bleomycin-injured mice, which was also observed in BCL-2-expressing αSMA⁺ myofibroblasts in the lungs of IPF patients, reinforcing the translational significance of the mouse studies.

Linkage between apoptosis resistance and senescence in pro-fibrotic lung fibroblasts

We next investigated potential relationships between fibroblast persistence, fibrosis and the acquisition of a sustained senescence phenotype in 2 additional models of persistent lung fibrosis (i.e., 24 weeks after repetitive instillation of bleomycin³³ and 6 weeks after a single bleomycin instillation in Col1a1-CreERT2;Fas^−/− mice¹¹). Bulk sequencing of fibroblasts isolated from both models exhibited a similar senescence-associated gene enrichment signature to PDGFRα-CFP/BCL-2⁺ fibroblasts (Figs. 6a and  7a), including increased expression of senescence-associated genes Cdkn1a, Cdkn2a, the secretory-associated senescence phenotype (SASP) genes, Il6, IL1a, and Serpine1, downstream IL-6 signaling (Il6st) and control of cell growth (mTor, Pten) (Fig. 7a). Fibroblasts isolated after repetitive bleomycin also exhibited a significant increase in senescence score (p < 0.01) (Fig. 7b) and a 55% positive (p < 0.05) Bcl-2:senescence gene Pearson correlation in the network plots (Fig. 7c, Supplementary Table 6). Examination of lung fibroblasts from Col1a1-CreERT2;Fas^−/− mice also revealed an increase in senescence score trending towards significance (p = 0.08) (Fig. 7b) and a Bcl-2:senescence gene Pearson correlation of 29% (p < 0.05) (Fig. 7c, Supplementary Table 7). Fibroblasts from all 3 models exhibited a significant positive correlation between Bcl-2 and SASP associated genes (Fgf7, PDGFα, Il-6, MMPs, TIMPs and Nox4) and a significant negative correlation between Bcl-2 and cell cycle (Ccnb1, Ccnd1) and cytokeratin genes (Cdk1, Cdk2b) (Figs. 6a, 7a and Supplementary Fig. 10).

Fig. 7. Fibroblasts develop a sustained Bcl-2 and senescent phenotype during persistent fibrosis.

a Heat maps of differential gene expression associated with regulation of cellular senescence in fibroblasts from two non-resolving models of fibrosis; repetitive bleomycin and Col1a1-Fas^-/-. b Quantitative senescence Z-score and (n = 4 mice/group, mean +/- SEM, **p < 0.01, Brown-Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons.) c Bcl-2 Pearson Correlation maps for senescence associated genes (p < 0.05) in repetitive bleomycin and Col1a1-Fas^-/- fibroblasts. Immunofluorescent images and quantitation of fibroblasts/per field from control and persistently fibrotic lungs from repetitive bleomycin and Col1a1-Fas^-/- mice for (d) PDGFRα (magenta) and PDGFRβ (green), (e) murine Bcl-2 (green), (f) p21 (yellow), (g) p16 (magenta) and (h) SA-β-Gal (white). d–h 5 mice/group; individual images graphed as scatter plot with bar mean + /-SEM, *p < 0.05, ***p < 0.001 Brown–Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons. Immunofluorescence panels 20X. SA-β-gal, senescence associated-β-galactosidase. Avg NE = Average normalized expression. Cor coe = correlation coefficient. nsd = no significant difference. Source data are for a–c and e–h provided as a Source Data file. Data for Col1a1-Fas^-/- fibroblasts (GEO: GSE161648) were previously published ¹¹.

To determine the location and abundance of senescent fibroblasts from these additional models of persistent fibrosis, lung sections were stained for p21^WAF, p16^INK, SA-β-gal and Bcl-2 together with PDGFRα and PDGFRβ (Fig. 7d–h). There was significant expression of Bcl-2, p21^WAF, p16^INK and SA-β-Gal in PDGFRα⁺ and PDGFRβ⁺ fibrotic areas in the lungs of mice repetitively-instilled with bleomycin compared to saline instilled mice (p < 0.001, Fig. 7d and Supplementary Fig. 9). Col1a1-CreERT2;Fas^−/− mice also showed increased, though less robust, expression of senescence markers p21^WAF (p < 0.05), p16^INK (p = 0.057), and SA-β-gal (p = 0.0814) (Fig. 7d–h). Taken together, these data suggest that Bcl-2 augments fibroblast senescence in the persistently fibrotic lungs of PDGFRα-CFP/BCL-2⁺ mice, in C57Bl/6 mice after repetitively-instilled bleomycin and in fibroblasts in which apoptosis has been inhibited by Fas deletion.

Therapeutic inhibition of BCL-2 rescues mice with persistent pulmonary fibrosis and promotes fibrosis resolution

ABT-199 (Venetoclax^TM) is an orally available FDA-approved Bcl-2 BH3-mimetic that inhibits the interaction between Bcl-2 and BH3-only activators and sensitizers³⁴. To determine if ABT-199 reduces lung fibroblast persistence and resolves the persistent fibrosis in bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice, ABT-199 or vehicle was administered by oral gavage for 4 weeks, beginning 5 weeks after bleomycin instillation (Fig. 8a). Following treatment with ABT-199, the numbers of PDGFRα⁺ lung fibroblasts were significantly reduced (p < 0.001) compared to vehicle-treated mice (Fig. 8b). The numbers of PDGFRβ⁺ pericytes (Fig. 8c) were unaffected by treatment with ABT-199, possibly reflecting the acquisition of apoptosis resistance through an ABT-199 insensitive pathway. CD45⁺ leukocytes and EpCAM⁺ epithelial cells were also significantly reduced after treatment with ABT-199 (Supplementary Fig. 11). Analysis of lung hydroxyproline levels (Fig. 8d), micro-CT quantification of non-aerated lung volume (Fig. 8e,f,h) and measurements of arterial oxygen saturation (Fig. 8g) showed that treatment with ABT-199 led to a significant reduction in established fibrosis and improved arterial oxygenation (p < 0.01, p < 05 and p < 0.001 respectively). Lung histology analysis showed that treatment with ABT-199 also led to a loss of bronchiolization in the lung parenchyma and partially restored alveolar architecture (Fig. 8i), whereas vehicle treated mice had sustained pathologic airway distortion, alveolar collapse and bronchiolization (Fig. 8i). To determine if treatment with ABT-199 reduced the senescent phenotype of the remaining PDGFRα⁺ fibroblasts and PDGFRβ⁺ pericytes, we stained lung sections for p16^INK, p21^WAF and SA-β-gal. Following treatment with ABT-199, fibroblast expression of p16^INK and SA-β-gal were significantly reduced (p < 0.001), with no change in p21^WAF expression (p = 0.3896) compared to vehicle control (Fig. 8j,k). Taken together, these pre-clinical data suggest that pharmacologic targeting of BCL-2 leads to the loss of senescent, apoptosis-resistant pro-fibrotic PDGFRα⁺ fibroblasts, resolved persistent fibrosis and promoted the restoration of lung architecture with improved gas exchange.

Fig. 8. Therapeutic treatment with ABT-199 reverses persistent fibrosis driven by BCL-2 conditional expression.

a Schematic of fibrosis induction and therapeutic dosing schedule with ABT-199. b Hydroxyproline content of lungs at 9 weeks, (c–d) Quantification of PDGFRα ⁺ fibroblasts and PDGFRβ ⁺ pericytes. e representative axial images (red overlay indicates disease area) and (f) 3D reconstructions (non-aerated fibrotic lung in blue, aerated lung in gray). g arterial oxygen saturation and (h) quantified non-aerated lung volume measured by micro-CT. i Representative Trichrome stained lung sections and semi-quantitative histology scoring after vehicle or ABT-199 treatment. b–i n = 5 mice/group, time-course line graphs mean +/− SEM. *p < 0.05, **p < 0.01, ***p < 0.001, Brown–Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons). j–k Representative immunofluorescent staining and quantitation of fibroblasts/per field in vehicle and ABT-199 treated PDGFRα-CFP/BCL-2⁺ mice for p21 (yellow), p16 (magenta) and SA-β-gal (white). Aperio scans 2x, panels 10x. Immunofluorescence panels 20X. n = 5 mice/group; individual images graphed as scatter plot with bar mean + /−SEM, *p < 0.05, **p < 0.01, ***p < 0.001, Brown–Forsythe and Welch’s ANOVA with Dunnett correction for multiple comparisons. Source data for b–d, g, h and j are provided as a Source Data file.

Discussion

The mechanisms contributing to impaired homeostatic repair, persistent fibrosis and honeycomb cyst formation in IPF patients remain poorly understood. In this study, we show that conditional BCL-2 expression in PDGFRα and Col1a1 expressing fibroblasts: (i) promotes fibroblast senescence and resistance to apoptosis, (ii) prevents homeostatic lung regeneration, and (iii) leads to prolonged lung fibrosis and cyst formation in an otherwise resolving model of bleomycin-induced pulmonary fibrosis. Additionally, (iv) we show that therapeutic in vivo inhibition of BCL-2 with ABT-199 in PDGFRα-CFP/BCL-2⁺ mice resolved BCL-2-driven pro-fibrotic fibroblast persistence and fibrosis and promoted lung regeneration.

PDGFRα is preferentially expressed by alveolar fibroblasts which support AEC2 survival and proliferation, lipofibroblasts that transport and store neutral lipids, and matrix fibroblasts that synthesize ECM proteins^35,36. By contrast, PDGFRβ⁺ pericytes provide niche support to alveolar capillary endothelial cells and serve as sentinels of barrier disruption^37,38. Previous studies have shown that PDGFRα⁺ fibroblasts, PDGFRα⁺ /PDGFRβ ⁺ fibroblasts and PDGFRβ⁺ pericytes exhibit substantial pro-fibrotic transcriptomic re-programming as they adapt and contribute to fibrosis and repair following bleomycin-induced injury^{11,24,39–41}. Additionally, and in confirmation of previous studies^11,41, we observed the emergence of a minor PDGFRα⁺/PDGFRβ⁺ fibroblast subset from the PDGFRα⁺ population which was characterized by robust expression of pro-fibrotic pathways and genes (e.g., Col1a1, Fn1, Spp1, Cthrc1)^11,39,40.

Homeostatic lung regeneration following bleomycin-induced fibrosis is preceded by apoptosis and clearance of pro-fibrotic fibroblasts. Here, we show that conditional BCL-2 expression in PDGFRα⁺ fibroblasts prevented the wave of fibroblast apoptosis known to precede homeostatic lung regeneration^11,23, leading to fibroblast persistence in bleomycin-injured lungs. Lineage tracing confirmed that BCL-2 was expressed by >95% of PDGFRα⁺ fibroblasts and PDGFRα⁺/PDGFRβ⁺ cells. Intriguingly, PDGFRβ⁺ pericytes also persisted in the fibrotic lungs of PDGFRα-CFP/BCL-2⁺ mice, raising the question of how BCL-2-expressing PDGFRα⁺ fibroblasts contribute to this response. Recent lineage tracing studies in mice have suggested that PDGFRα⁺ alveolar fibroblasts can proliferate and trans-differentiate into PDGFRβ⁺ cells⁴¹. However, the minimal penetrance of CFP ( < 10%⁺) observed in PDGFRβ⁺ pericytes herein argues against this, potentially suggesting that BCL-2 expression in PDGFRα⁺ fibroblasts blocks this pathway in fibrotic lungs. We speculate that the PDGFRβ⁺ cells that persist in PDGFRα-CFP/BCL-2⁺ mice originate from PDGFRβ⁺ pericytes, whereas the PDGFRα⁺/PDGFRβ⁺ cells arise from PDGFRα⁺ adventitial fibroblasts^24,27. A potential mechanism could involve increased secretion of PDGF-B and/or PDGF-D by PDGFRα⁺ fibroblasts, which could support the survival of PDGFRβ-expressing cells in a paracrine fashion^42–44.

Key histopathologic features in IPF lung tissues include the presence of honeycomb cysts and the appearance of KRT8^hi transitional epithelial cells^29,45–49. Until recently, parenchymal bronchiolarization and cysts have been poorly reproduced in murine models of fibrosis. Instillation of a single dose of bleomycin in wild type mice results in the accumulation of small numbers of KRT8^hi epithelial cells³³. However, repetitive bleomycin instillation³³ leads to parenchymal bronchiolarization, the appearance of KRT8⁺/CCSP⁺ cysts, sustained fibrosis and the persistence of BCL-2 expressing pro-fibrotic fibroblasts³³. In the present study, instillation of a single dose of bleomycin in PDGFRα-CFP/BCL-2⁺ mice also led to substantial, sustained lung remodeling with cyst formation that appeared as early as 3 weeks, suggestive of a role of BCL-2 expressing fibroblasts in cyst development. Similar cysts were also observed in bleomycin-instilled Col1a1-CreERT2;Fas^−/− mice at 6 weeks²⁹. With a lack of models to study honeycomb cyst formation in mice, little is known about the mechanisms leading to their development in IPF patients. Thus, the development of cysts, coincident with persistent fibrosis in bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice, may represent a robust model for future mechanistic studies. Additionally, we observed a sustained increase in pro-SPC⁺ AEC2s in the lungs of bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice, a phenotype that contrasts with the loss of AEC2 cells observed in the lungs of IPF patients⁵⁰. Potential explanations for this difference between mice and humans include: (i) decreased aeration in fibrotic mice, which may give the impression of increased numbers of AEC2s in collapsed alveoli, and/or (ii) unavoidable temporal differences between the end-stage disease that characterizes explanted IPF lungs and the earlier progressive stage that characterizes the mouse model⁵¹.

Prior scRNA-seq studies have shown that during lung regeneration following injury, fibroblast transcriptomic programming transitions from pro-fibrotic genes and pathways to those involved in tissue repair and wound healing, e.g., Hippo and Wnt signaling, cell-cell adhesion and epithelial cell development¹¹. Fibroblasts isolated from the lungs of bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice, however, failed to transition towards homeostatic repair and regeneration transcriptomic profiles and remained locked in a robust pro-fibrotic programming pattern. How increased BCL-2 expression promotes this enhanced and sustained pro-fibrotic phenotype remains unclear though potential mechanisms include: (i) binding to, and sequestration of, Beclin-1⁵² which sustains pro-fibrotic fibroblast function through impaired autophagic degradation of ECM in vitro^53,54 and enhances renal fibrosis in vivo^55,56; (ii) augmentation of pro-survival and pro-fibrotic pathways, e.g., migration, invasion, and Sonic Hedgehog signaling^57,58, or as our data suggests, (iii) through the downstream development of fibroblast senescence, which promotes both a pro-fibrotic senescence-associated secretory phenotype (SASP) and prevents apoptosis^59,60.

Based on these emerging findings, we hypothesize that increased expression of BCL-2 prevents fibroblast apoptosis and enables them to persist within the fibrotic milieu. Through crosstalk with aberrant epithelial cells and the SASP, these apoptosis-resistant fibroblasts are subsequently licensed to acquire a stable senescent phenotype, which differs from the transient senescent phenotype observed in the resolving single dose bleomycin instillation model^61,62. The senescent transcriptional phenotype in lung fibroblasts of fibrotic PDGFRα-CFP/BCL-2⁺ mice was characterized by increased mRNA and protein expression of p16^INK, p21^WAF and SA-β-gal along with enrichment of pathway terms associated with senescence regulation and p53 signaling. This senescence signature was phenocopied in fibroblasts isolated from the persistently fibrotic lungs of mice repetitively-instilled with bleomycin and in mice with fibroblast specific Fas deletion^11,33. Transcriptomic analyses across species have also identified a shared senescence gene expression signature associated with IPF pathogenesis^31,63,64 with elevated levels of senescent fibroblasts and epithelial cells, indicating an important role for senescence in disease persistence and progression^65–68. Additionally, our spatial transcriptomic and immunostaining studies in IPF lung tissue revealed co-expression of senescence-associated genes and BCL-2 in αSMA⁺ myofibroblasts. Our data and other publicly available data indicate that BCL-2 transcript expression in fibroblasts isolated from end-stage IPF or ILD tissues is relatively low, while senescence gene transcript expression is high. Thus, we propose BCL-2 expression precedes and enables senescence initiation by preventing their apoptosis. Once the final common fibroblast senescence program has been initiated, fibroblast BCL-2 expression appears to be diminished, as measured in our spatial transcriptomics studies in end-stage IPF lung tissues. While our studies do not address the mechanisms underlying the development of the senescent fibroblast phenotype, they suggest that the persistence of a pro-fibrotic senescent fibroblast population may contribute to the aberrant epithelial remodeling associated with cyst formation through the production of the SASP^67,69,70, resulting in a positive fibroblast-epithelial crosstalk feedback loop recapitulating that observed in the lungs of IPF patients^{29,63,67,71,72}.

Pre-clinical studies showed that therapeutic BCL-2 inhibition with the selective BH3-mimetic, ABT-199, facilitated reengagement of the mitochondrial apoptotic pathway and thereby enabled elimination of senescent, apoptosis-resistant, pro-fibrotic fibroblasts, resolved established fibrosis and partially restored lung architecture in bleomycin-instilled PDGFRα-CFP/BCL-2⁺ mice. BH3-mimetics act by preventing the interaction between anti-apoptotic BCL-2 members and pro-apoptotic activators (BIM, BID) and pore formers (BAX and BAK) and have also been classified as senolytics^73–75. Thus, in the presence of sufficient activators and pore formers, the increased expression of BCL-2 in pro-fibrotic lung fibroblasts in IPF patients and persistently fibrotic mice^12,76 renders these cells apoptosis-resistant, yet uniquely sensitive to ABT-199. By contrast, cells with lower expression of BCL-2, such as uninvolved fibroblasts or PDGFRβ ⁺ pericytes, are less sensitive to ABT-199, explaining, in part, the selectivity of these drugs⁷⁷. Therapeutic interventions targeting BCL-2 expressing cells, including senescent fibroblasts, with BH3-mimetics in multiple mouse models have shown promising outcomes, offering translational potential for human IPF treatment strategies^{71,74,76,78–80}.

In conclusion, by integrating a custom-engineered mouse model with preclinical and human studies, we have delineated the role of fibroblast senescence and resistance to apoptosis in IPF and validated a targeted therapy of clinical relevance. Furthermore, the availability of CFP/BCL-2stop^fl/+ mice should facilitate the assessment of therapeutic interventions targeting human BCL-2, as well as being a useful tool to conditionally induce senescence and resistance to apoptosis in murine cell lineages in fibrosis and other disease settings.

Methods

Study approvals

All animal studies were approved by the National Jewish Health Institutional Animal Care and Use Committee (AS2741). Euthanasia was conducted in accordance with AVMA guidelines. All human lung tissue samples were de-identified and determined by the National Jewish Health and University of Colorado Anschutz Medical Campus Institutional Review Boards (IRBs) to constitute non–human subjects research; therefore, informed consent and additional IRB approval for this study were not required. Tissue procurement followed approved IRB protocols at both institutions. Healthy lungs were obtained through Donor Alliance, Inc. (Denver, CO). IPF lungs were obtained as explants during lung transplantation.

Mice

Generation of CFP/BCL-2stop^fl/fl mice

Mice capable of conditionally expressing human BCL-2 were generated by synthesizing a targeting vector in which a loxP-flanked transcriptional stop cassette was inserted downstream of a synthetic CAG enhancer. This was followed by the coding sequence of human BCL-2 fused to a P2A protease cleavage site⁸¹ followed by the mCerulean Blue (CFP)^82,83 to allow lineage tracing of BCL-2 expressing cells. B6N JM8.A3 embryonic stem cells⁸⁴ were electroporated with the ROSA26-STOP^fl BCL-2-2A-CFP targeting construct (Supplementary Fig. 12a), placed under antibiotic selection and the resulting clones were picked and expanded. gDNA was prepared for each clone and used for qPCR analysis by loss of allele (LOA)⁸⁵. Putative positive clones were then subject to in vitro transfection with a Cre expression plasmid, which led to the expected recombination-mediated excision of the floxed-stop-cassette and expression of eCFP in multiple clones. Blastocyst injection was then performed using B6J blastocysts and the functionally verified ES cell clones (untransfected). Chimeric mice of genotype Rosa26R-stop^fl/fl-CFP/BCL-2 (abbreviation: CFP/BCL-2stop^fl/fl) were generated and breeding with B6J wildtype mice established germline transmission, which was confirmed by genotypic analysis (Supplementary Fig. 12b).

CFP/BCL-2stop^fl/+ mice were crossed with mice expressing tamoxifen-inducible Cre recombinase driven by the PDGFRα (Jax strain: 18280 - B6N.Cg-Tg(Pdgfra-cre/ERT)467Dbe/J) and Col1a1 (Jax strain:016241 - B6.Cg-Tg (Col1a1-cre/ERT2)1Crm/J) (The Jackson Laboratory, Bar Harbor, ME) promoters. Recombination was induced by intraperitoneal injection of tamoxifen (0.25 mg/g body weight in corn oil, Sigma Aldrich, St. Louis, MO) or corn oil as a control starting at day 4 after bleomycin. Mice received 5 injections between day 4 and 21. Mice receiving corn oil or tamoxifen are referred to as PDGFRα-CFP/BCL-2^- or PDGFRα-CFP/BCL-2⁺ mice, respectively. All mice were maintained in a specific pathogen-free facility under 12-h light/dark cycles with ad libitum access to autoclaved water and standard chow (Diet 2919, Inotiv, West Lafayette, IN).

Primary fibroblasts

Primary mouse lung fibroblasts were sorted from healthy male PDGFRα-CFP/BCL-2^- and PDGFRα-CFP/BCL-2⁺ lungs on a FACSAria Fusion (BD Biosciences, San Jose, CA), as previously described^9,86. The cells were maintained in DMEM containing 10% (v/v) FBS and used between passages 3-8. Intrinsic CFP expression was confirmed by fluorescence microscopy (Fig. 1b). Fibroblast mitochondria were stained with Mitotracker Red CMXRos (M7512, Invitrogen/ThermoFisher, Waltham, MA) while actin filaments were stained with phalloidin (AF488, #A12379, Invitrogen/ThermoFisher). Stained cells were imaged on a Zeiss LSM 700 Confocal microscope with image acquisition using ZenBlack (Zeiss, White Plains, NY). Detection of BCL-2 (1:1000 anti-rabbit, Abcam #59348, Waltham, MA), GFP (AvesLab #GFP-1020, chicken, 1:1000, Davis, CA), β-actin (Millipore #MAB150, mouse, 1:1000, St. Louis, MO), GAPDH (Abcam #8245, mouse, 1:1000) and Tom70 (Cell Signaling #65675, rabbit, 1:1000, Danvers, MA) was done by Western blot analysis in the mitochondrial and cytosolic fractions, as described⁸⁷. Fibroblasts were sensitized with TNF-α (10 ng/mL) and IFN-γ (50 U/mL) for 48 h prior to stimulation with agonistic anti-Fas antibody (Jo2, 250 ng/mL BD Pharmingen, San Diego, CA) for 4 h or were incubated with 1 μM staurosporine (Sigma Aldrich) for 4 h. Active caspase-3/7 was measured using a Caspase-Glo kit (Promega, Madison, WI) as previously described^15,76.

Assessment of fibrotic lung disease

Pulmonary fibrosis was initiated by the intratracheal instillation of 50 μL of bleomycin (1.5U/kg, Amneal Biosciences, Bridgewater, NJ dissolved in sterile saline) to anesthetized male and female mice 8–10 weeks of age⁸⁸. Fibrosis was assessed by lung measurements of collagen in the upper right lobe (hydroxyproline assay)^11,88 and histology was evaluated by Masson’s trichrome staining of sections from the left lung⁸⁸. Images were taken on an upright Olympus BX51 or by Aperio Scanning (Leica Biosystems, Buffalo Grove, IL). Where specifically indicated C57Bl/6 mice (Jackson Laboratories, Bar Harbor, ME) were used as controls. For therapeutic intervention with ABT-199, mice were dosed with ABT-199 (100 mg/kg) by daily oral gavage for 28 days or vehicle (60% phosal 50 propylene glycol (PG), 30% polyethylene glycol 400, 10% ethanol), starting 5 weeks after fibrosis was established⁷⁸.

Flow cytometry

Single cell suspensions were obtained from perfused lungs by enzymatic dispersion with collagenase (1 mg/ml, Sigma Aldrich, C2139), elastase (4U/ml, Worthington, LS002274 Lakewood, NJ) and dispase (5U/ml, Worthington LS02104) incubation (30 min, 37 °C) followed by a 0.25% trypsin-EDTA (ThermoFisher, 2520056) + DNaseI (20 μl/10 ml, Worthington LS006331) digest (20 min, 37 °C)¹¹. Cells were stained with fluorescently-tagged monoclonal antibodies against CD45 (17-0451-82), CD326/EpCAM (17-5791-80), CD31 (541814, BD Biosciences, San Jose, CA), PDGFRα/CD140a (25-1401-82) and PDGFRβ/CD140b (ThermoFisher, Waltham, MA) at a 1:200 dilution. Cell analysis data were acquired with the LSRFortessa (BD Biosciences) and analyzed with FlowJo software (V10, Tree Star, Ashland, OR). Lineage (CD326/EpCAM, CD31, CD45) negative cells (Lin-) were gated, (Supplementary Fig. 13)¹¹, and assessed for PDGFRα and PDGFRβ staining. Lin- fibroblasts were found to be comprised of PDGFRα⁺, PDGFRα⁺/β+ and PDGFRβ⁺ cells.

Immunofluorescent staining

Formalin fixed, paraffin embedded sections, were deparafinized and rehydrated followed by antigen retrieval using citrate buffer. Non-specific binding was reduced by incubation in 10% mouse serum and horse IgG serum from the M.O.M immunodetection blocking kit (#BMK-2202, Vector Laboratories, Burlingame, CA). Primary antibodies for BCL-2 (BioRad #MCA1550, mouse, 1:100 Hercules, CA), Bcl-2 (Cell Signaling Technology # 3498, rabbit, 1:500), PDGFRα (Cell Signaling, #3174S rabbit, 1:100), PDGFRβ (R&D Systems #AF1032 goat, 1:100, Minneapolis, MN), GFP (which identifies CFP; AvesLab #GFP-1020, chicken, 1:200), αSMA (Sigma Aldrich #2547, mouse, 1:500), p16^INK (Abcam #ab211542 rabbit, 1:100), p21^WAF (Abcam #ab188224 rabbit, 1:250), β-galactosidase (Cell Signaling #27198 rabbit, 1:200), CCSP (Seven Hills Bioreagents #WRAB-3950 rabbit, 1:200, Cincinnati, OH), pro-SPC (Millipore #AB3786 rabbit, 1:500), Krt8 (DSHB #AB531826 rat, 1:100), CD45 (Invitrogen, 17-0451-82 rat, 1:100), CD31 (Invitrogen, #14-0311-82 rat, 1:100) were incubated overnight at 4^oC followed by incubation with fluorescently tagged goat anti mouse-A647 (A21236), donkey anti-rabbit-A555 (A31572), goat anti rat-A555 (A21434), or donkey anti goat-A647 (A21447) secondary antibodies at 1:100 dilution (Invitrogen, Carlsbad, CA). TUNEL staining to detect apoptotic cells was completed prior to antibody staining per manufacturer’s instructions (Promega, Madison, WI). TUNEL positive cells were counted and averaged from 10 images from each animal as described¹¹. Images were captured in fibrotic areas using PDGFRα and PDGFRβ positive staining as an indicator of fibroblast accumlation. TUNEL positive cells located in the airway epithelium were excluded from the analysis. Quantitative cell counts from CCSP, KRT8, BCL-2, p16^INK, p21^WAF, and β-Gal images were compiled using stereology grids and averaged from 10 images from each animal. Frozen 4% PFA fixed OCT (Fisher Scientific, Hampton, NH) sections were mounted with Fluoroshield Mounting Media containing DAPI (Vector Laboratories). Images were acquired on a Zeiss Axioplan 2 epi-fluorescence microscope and analyzed with Axiovision software (Zeiss, Jena, Germany).

Bulk RNA sequencing

Prior to cell sorting, single cell suspensions were enriched for lineage negative cells by incubating with CD45 (130-052-301), CD31 (130-097-418) and CD326 (130-105-958) MicroBeads and purified off of LS columns per manufacturer’s instructions (Miltenyi Biotech, Bergisch Gladbach, Germany/Auburn, CA). At least 300,000 live (DAPI-) lineage negative cells were collected per sample on the FACSAria Fusion (BD Biosciences). Purified cells were pelleted, lysed, and RNA was prepared, as described¹¹. Libraries were sequenced as barcoded-pooled samples and processed for next-generation sequencing (NovaSeq 6000 - Illumina platform).

Bulk RNA sequencing analysis

To improve downstream mapping quality, raw sequencing reads were trimmed using skewer with parameters (end-quality=15, mean-quality=25 and min=30). Trimmed reads were aligned to the mouse reference genome GRCm38 using Hisat2. Gene quantification was performed with htseq-count using GRCm38 Ensembl v84 GTF. Differential expression analysis and gene correlation between groups was conducted with the R (3.5.1) package DESeq2 (1.22.2). FDR less than or equal to 0.05 ( | r | ≥0.5) and a Pearson correlation equal to or greater than 0.7 ( | r | ≥0.75) were used as thresholds of significance. Pathway enrichment analysis was conducted with Enricher to identify Gene Ontology:Biological Processes (GO:BP) significantly genes by pair-wise comparison^89,90. The Col1a1-CreERT2;Fas^-/- mice data set has been deposited in the National Center for Biotechnology Information/Gene Expression Omnibus under accession number GSE161648¹¹, the PDGFRα-CFP/BCL-2⁺ mice are under accession number GSE307983 and the spatial transcriptomics data are  under accession number GSE282639.

Spatial transcriptomic analysis

We performed Nanostring CosMx Digital Spatial Profiler to obtain subcellular resolution spatial transcriptomic data on 15 IPF and 13 unaffected controls using methods reported in the “Supplemental Methods” and by Blumhagen et al.⁹¹. For this manuscript, we focused on BCL-2 expression in myofibroblasts and cell senescence. Following cell annotation, we applied the spatial correlation method InSituCor⁹² to identify modules of gene co-expression in which we focused on a module that comprised 10 genes associated with cell senescence (ATF3, C11orf96, CDKN1A, DUSP1, FOS, GADD45B, IL6, JUN, JUNB, ZFP36). In addition to spatial clustering, we compared expression between BCL-2⁺ myofibroblasts and BCL-2^- myofibroblasts using a linear mixed model on VST transformed counts adjusting for disease and including a random effect for subject.

Statistical overview

Data are presented as the mean ± SEM. Data were analyzed using GraphPad Prism software Version 10 (GraphPad, San Diego, CA). Differences between conditions at specific time points were examined using Student’s unpaired 2-tailed test with Welch’s correction with p < 0.05 considered to be significant. Box-and-whiskers plots show median, minimum and maximum values. Violin plots show median, upper and lower quartiles. Specific details about the replicates used where n =  individual animal replicates are available in the figure legends. All assays and measurements were taken from distinct samples. Immunostained sections were cut from distinct individual animals.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

E.F.R., T.S., J.M., B.L.E., N.J., J.A.W., D.G.F., R.G., S.C., S.S., J.C.C., and R.Z.B. conducted the experimental work. D.W.H.R. conceived the project. E.F.R., M.A.S., I.V.Y, R.L.Z., D.A.S., R.Z.B. J.P.B., R.M.T. and D.W.H.R. designed and planned experiments, analyzed the data and contributed to writing the manuscript. All authors reviewed the manuscript.

Peer review

Peer review information

Nature Communications thanks Victor Thannickal, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available in the main text and figures, the Supplementary Information and the Source Data file. Sequencing data are deposited into the Gene Expression Omnibus data base under accession numbers: GSE161648GSE307983GSE282639. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69865-4.