Loss of Fas-signaling in fibroblasts impairs homeostatic fibrosis resolution and promotes persistent pulmonary fibrosis

Idiopathic pulmonary fibrosis (IPF) is a progressive, irreversible fibrotic disease of the distal lung alveoli that culminates in respiratory failure and reduced lifespan. Unlike normal lung repair in response to injury, IPF is associated with the accumulation and persistence of fibroblasts and myofibroblasts and continued production of collagen and other extracellular matrix (ECM) components. Prior in vitro studies have led to the hypothesis that the development of resistance to Fas-induced apoptosis by lung fibroblasts and myofibroblasts contibributes to their accumulation in the distal lung tissues of IPF patients. Here, we test this hypothesis in vivo in the resolving model of bleomycin-induced pulmonary fibrosis in mice. Using genetic loss-of-function approaches to inhibit Fas signaling in fibroblasts, novel flow cytometry strategies to quantify lung fibroblast subsets and transcriptional profiling of lung fibroblasts by bulk and single cell RNA-sequencing, we show that Fas is necessary for lung fibroblast apoptosis during homeostatic resolution of bleomycin-induced pulmonary fibrosis in vivo. Furthermore, we show that loss of Fas signaling leads to the persistence and continued pro-fibrotic functions of lung fibroblasts. Our studies provide novel insights into the mechanisms that contribute to fibroblast survival, persistence and continued ECM deposition in the context of IPF and how failure to undergo Fas-induced apoptosis prevents fibrosis resolution.


INTRODUCTION
Idiopathic pulmonary fibrosis (IPF), a progessive fibrotic lung disorder of unacceptably high morbidity and mortality, develops in response to initial alveolar epithelial injury and results in an aberrant repair response in which resident lung interstitial and perivascular fibroblasts proliferate, differentiate into myofibroblasts and persist within the collapsed alveoli and septal interstitium (1,2). These fibroblasts and myofibroblasts (collectively referred to herein as "pro-fibrotic fibroblasts"), continue to produce excessive amounts of extrcellular matrix (ECM), which in turn leads to relentless fibrosis, progressive declines in gas exchange and eventual respiratory failure (1,2). Acute respiratory distress syndrome (ARDS) patients also exhibit high morbidity and mortality resulting from initial alveolar epithelial injury (3). Likewise, they also accumulate profibrotic fibroblasts that secrete collagen and ECM components within the injured alveoli early in their clinical course (4)(5)(6). However, while both conditions share some similarities in the development of the early response to injury, the clinical outcomes of these two disorders are substantially different. IPF patients develop a persistent and progressive fibrosis considered by many to be irreversible (1). By contrast, while ARDS patients experience high initial mortality, the early fibroproliferative response can resolve and lung structure and function can return towards normality in some, though not all, survivors (3,7,8).
Little is known about the factors that distinguish these different outcomes in IPF patients and ARDS survivors. In general, fibrosis is considered to be detrimental and is thought to lead to poor clinical outcomes. However, the early fibrosis that develops in ARDS patients could be viewed as a beneficial response in which the provisional ECM scaffold produced by proliferating pro-fibrotic fibroblasts supports the regenerating alveolar epithelium and endothelium. As alveolar structure and function are restored, the majority of the pro-fibrotic fibroblasts undergo apoptosis and clearance (9,10). During this period of tissue regeneration, macrophages and other cells are thought to degrade and phagocytose the fibrotic provisional ECM scaffold through an integrated 4 process that we define herein as "homeostatic fibrosis resolution". In contrast to this trophic regenerative process, the persistent and progressive pulmonary fibrosis that defines IPF patients can be viewed as a failure in homeostatic fibrosis resolution. We hypothesize that a failure in profibrotic lung fibroblast apoptosis and clearance plays a fundamentally important role in impeding homeostatic fibrosis resolution in IPF.
Previous studies by ourselves and others have provided insights into the mechanisms that control lung fibroblast apoptosis. In particular, susceptibility to apoptosis induction by the death receptor, Fas, has been shown to play a key role (11)(12)(13)(14)(15)(16). Fas is expressed by lung fibroblasts and once a pre-defined threshold has been exceeded, Fas ligation induces apoptosis (12,17). In contrast, lung fibroblasts from IPF patients exhibit variable resistance to Fas-induced apoptosis (13,15) due to down-regulation of Fas expression (14,17,18) and increased expression of multiple anti-apoptotic genes, including Bcl-2, XIAP, cFLIPL and PTPN13 (11,(19)(20)(21). Together, these in vitro studies have fueled the concept that the acquired resistance of lung fibroblasts to Fas-induced apoptosis promotes pro-fibrotic fibroblast accumulation in the persistently fibrotic lungs of IPF patients (11,15,16,20,22,23). However, until now, this concept has not been directly addressed in vivo.
To address this knowledge gap, we genetically deleted Fas in fibroblasts and investigated the functional consequences on the normally resolving model of bleomycin-induced fibrosis.
Intratracheal instillation of a single dose of bleomycin induces inflammatory epithelial injury followed by a resolving fibrotic response accompanied by regeneration of the distal lung architecture (9,24). Thus, the bleomycin model recapitulates acute lung injury and early fibrosis followed by fibrosis resolution and lung regeneration (25), and represents an authentic model of homeostatic fibrosis resolution at later time points. As we will show, loss of Fas signaling in profibrotic fibroblasts impairs homeostatic fibrosis resolution and results in fibroblast persistence in 5 lung tissues along with persistent pulmonary fibrosis, replicating a central feature of IPF.
Furthermore, Col1a1-and aSMA-reporter mice and bulk and single cell RNA-sequencing assessments of purified lung fibroblasts showed that Fas deficiency in fibroblasts results in preserved patterns of increased ECM and fibrotic gene expression in the persistently fibrotic lungs. We conclude that Fas-induced fibroblast apoptosis plays a necessary role in fibroblast elimination during homeostatic fibroblast resolution, and that resistance to Fas-mediated apoptosis promotes persistent pulmonary fibrosis.

Fas deletion in fibroblasts impairs lung homeostatic fibrosis resolution.
To investigate the functional role of Fas expression by fibroblasts in homeostatic fibrosis resolution, we deleted Fas in mesenchymal cells by breeding Fas fl/fl mice with Dermo1(Twist2)-Cre mice (26) to create Dermo1-Cre;Fas fl/fl mice (abbreviated as Dermo1-Cre;Fas -/-). To ensure that Fas was deleted, we isolated lung fibroblasts from wild type and Dermo1-Cre;Fas -/mice and assessed their level of cell surface Fas expression by flow cytometry, and their ability to undergo apoptosis: (i) following sensitization with TNF-a and IFN-a and Fas-ligation with agonistic anti-Fas antibody (Jo2), and (ii) in response to staurosporine, an activator of the intrinsic apoptosis pathway. In comparison to wild type lung fibroblasts, fibroblasts from Dermo1-Cre;Fas -/mice displayed minimal cell surface Fas basally or in response to stimulation with TNF-a and IFN-a ( Figure 1A) and were not susceptible to Fas-induced apoptosis ( Figure 1B). However, fibroblasts from both wild type and Dermo1-Cre;Fas -/mice showed similar levels of apoptosis induction in response to staurosporine ( Figure 1B). Together, these data indicate that in vivo genetic deletion of Fas in mesenchymal cells led to loss of cell surface Fas and corresponding apoptosis induction by the Fas-induced extrinsic apoptosis pathway, but not by staurosporine.
7 of fibrosis at 1.5 and 3 weeks. However, whereas bleomycin-instilled, Col1-CreERT2;Fas +/+ mice underwent homeostatic fibrosis resolution between 4.5 and 9 weeks, homeostatic fibrosis resolution was impaired in the lungs of bleomycin-instilled Col1-CreERT2;Fas -/mice, as reflected by persistently elevated lung hydroxyproline levels ( Figure 2B) and sustained fibrotic lung histology patterns ( Figure 2C,D). Wild type C57Bl/6 mice and Col1-CreERT2;Fas +/+ mice showed indistinguishable patterns of fibrosis development and homeostatic fibrosis resolution ((24) and not shown). Together, these data show that loss of Fas-signaling in Col1a1-expressing pro-fibrotic fibroblasts also led to impaired homeostatic fibrosis resolution and persistent pulmonary fibrosis.

Fas deletion in fibroblasts promotes lung fibroblast persistence.
We next assessed the consequences of loss of Fas-signaling on lung fibroblast numbers.
To confirm these findings, we used enzymatic lung cell dispersal and multi-parameter flow cytometry to assess the consequences of loss of Fas-signaling on lung fibroblast numbers in Dermo1-Cre-expressing cells. Lung single cell suspensions were analyzed using the gating 8 strategy shown in Supplemental Figure 3A and Supplemental Table 1. From these data we quantified the number of lineage negative (Lin-) cells, which are throught to be primarily fibroblasts (28)(29)(30), by excluding EpCAM/CD326+ (epithelial), CD45+ (hematopoietic) and CD31+ (endothelial) positive populations. Dermo1-Cre;Fas +/+ and Dermo1-Cre;Fas -/mice were instilled with saline or bleomycin and the numbers of Lin-cells quantified for up to 9 weeks. Figure 3B shows that during the development of fibrosis between 1.5-3 weeks, the number of Lin-lung fibroblasts increased ~2.3-fold, with similar numbers being observed in both Dermo1-Cre;Fas +/+ and Dermo1-Cre;Fas -/mice. Within the Lin-fraction, approximately 33% expressed variable levels of the pan-fibroblast marker PDGFRa, while 66% were PDGFRa-. Figure 3B,C,D also shows that whereas the numbers of total lung Lin-cells, Lin-PDGFRa+ and Lin-PDGFRa-cells in Dermo1-Cre;Fas +/+ mice were progressively reduced by 6 and 9 weeks, their numbers remained elevated at 6 and 9 weeks in Dermo1-Cre;Fas -/mice.

Fas deletion in fibroblasts impairs their apoptosis during homeostatic fibrosis resolution.
We next investigated the effect of Fas-deficiency on lung fibroblast apoptosis in vivo. Dermo1-Cre;Fas +/+ and Dermo1-Cre;Fas -/mice were instilled with saline or bleomycin and lung tissues harvested from 1.5-9 weeks. Lung sections were then subjected to triple-label immunofluorescent staining for TUNEL, a-SMA and S100A4 ( Figure 5A) and TUNEL+ fibroblasts were quantified. Figure 5B shows minimal numbers of apoptotic TUNEL+ fibroblasts in naïve Dermo1-Cre;Fas +/+ . Increased numbers were first detected in bleomycin-instilled mice at 1.5 weeks, peaked at 3 weeks and declined back to naïve numbers at 6 weeks, confirming a previous report (10).
Following Fas deletion in fibroblasts in Dermo1-Cre;Fas -/mice, the number of TUNEL+ fibroblasts were not significantly elevated over baseline values at any time point, but when compared to Dermo1-Cre;Fas +/+ mice, were significantly reduced at 1.5 weeks (p=0.06), 3 weeks (p<0.05) and 4.5 weeks (p=0.08) ( Figure 5A, B). We confirmed these data in naïve and bleomycin-instilled Col1-CreERT2;Fas fl/fl mice following tamoxifen or corn oil injection between 0.5-3 weeks, with harvest and analysis for TUNEL+ cells for up to 9 weeks. Figure 5C,D shows that Col1-CreERT2;Fas +/+ mice also exhibited a peak in TUNEL+ apoptotic lung fibroblasts at 3 weeks, while TUNEL+ fibroblast numbers were not elevated above baseline at any time point in Col1-CreERT2;Fas -/mice (0=0.02). Taken together, these data show that Fas deletion in fibroblasts prevents homeostatic fibroblast apoptosis in bleomycin-injured fibrotic lung tissues.

Fas-deficiency in Col1a1-expressing cells leads to persistent Col1a1 and a-SMA promoter activity.
Col1-GFP mice express GFP under the control of the Col1a1 promoter (39); while SMA-RFP mice express RFP under the control of the a-SMA promoter (40). We used these mice to determine if the Fas-deficient fibroblasts that remain in persistently fibrotic lung tissues of bleomycin-instilled mice continue to functionally express these pro-fibrotic genes. Frozen lung sections from naïve mice revealed modest numbers of dim GFP+ fibroblastic interstitial cells, while bleomycin-instilled Col1-GFP mice exhibited increased numbers and intensity of GFP+ interstitial cells peaking at 3 weeks before returning towards baseline by 6-9 weeks ( Figure 6A).
To determine if the Fas-deficient fibroblasts that persist in fibrotic lung tissues continue to exhibit Col1a1 promoter activity, we bred Col1-CreERT2;Fas fl/fl mice with Col1-GFP mice to obtain We also determined if the Fas-deficient fibroblasts that persist in fibrotic lungs also continued to express a-SMA by breeding Col1-CreERT2;Fas fl/fl mice with SMA-RFP reporter mice using the same strategy as discussed above to generate Col1-CreERT2;Fas +/+ ;SMA-RFP mice and Col1-CreERT2;Fas -/-;SMA-RFP mice. Compared to Fas-sufficient Col1-CreERT2;Fas +/+ ;SMA-RFP 13 mice, lung tissues of Col1-CreERT2;Fas -/-; SMA-RFP mice displayed abundant RFP fluorescence in the fibroblastic cells in the persistently fibrotic distal lung tissues at 6 weeks ( Figure 7E).
Interestingly, and in contrast to the Col1-CreERT2;Fas -/-;Col1-GFP mice, flow cytometry of lung cell suspensions revealed significantly increased numbers of RFP+Lin-PRGFRa+ CD90+, but not of CD26+ and CD90-CD26-fibroblasts in the fibrotic lung tissues of Col1-CreERT2;Fas -/-;SMA-RFP mice ( Figure 7F,G,H). Taken together, these data suggest that the Fas-deficient fibroblasts that remain in fibrotic lung tissues of bleomycin-instilled mice at 6 weeks continue to express functional pro-fibrotic Col1a1-and aSMA-promoter activity.

Bulk and single cell RNA-sequencing defines preserved pro-fibrotic fibroblast clusters and signatures in the absence of Fas-signaling.
To further characterize the fibroblasts that persist in fibrotic lung tissues in the absence of Fas signaling, we conducted bulk and single cell RNA-sequencing (scRNA-seq) on sorted Lin-cells isolated from bleomycin-instilled Col1-CreERT2;Fas -/and Col1-CreERT2;Fas +/+ mice at 3 and 6 weeks. We also sorted Lin-cells from naïve Col1-CreERT2;Fas fl/fl mice for comparison. Bulk RNAseq showed that compared to naïve mice, bleomycin-instilled mice of both genotypes showed an expected increase in pro-fibrotic gene expression and fibrosis-associated pathways at 3 weeks (not shown). Principal Component Analysis (PCA) of highly variable genes partitioned samples by time of harvest after bleomycin instillation along the PC1 axis (30% variance) and by genotypic Fas deficiency along the PC4 axis (5% variance) ( Figure 8A). There were no significant differences in pro-fibrotic gene signatures in Lin-cells isolated from bleomycin-instilled Col1-CreERT2;Fas -/and Col1-CreERT2;Fas +/+ mice at 3 weeks. Pathway analysis of 393 upregulated differentially expressed genes (DEGs) at 6 weeks revealed that the most significantly enriched pathways in Col1-CreERT2;Fas -/mice were associated with extracellular matrix organization  Table 2). The transcripts that contributed to these pathways included the typical pro-fibrotic genes Col1a1, Col5a1, Col6a1, Col7a1, Col8a1, Col11a1, Eln and Loxl2 ( Figure 8C). By contrast, Lin-cells from the Col1-CreERT2;Fas +/+ mice, which were undergoing homeostatic fibrosis resolution at 6 weeks, showed enrichment for regulation of cell migration (p=0.0079) and included genes involved in wound healing (Epb41l4b, Erbb3, Itgb3, Wnt7A and Nrg1), and Wnt signaling (Wnt10B, Fzd5, Wnt3A, Wnt7A, Sost, Rspo4, Wnt4) ( Figure 8B, Supplemental Figure 9 and Supplemental Table 2).
To investigate heterogeneity among lung fibroblast populations, we conducted scRNA-seq on sorted Lin-cells from naïve mice and bleomycin-instilled Col1-CreERT2;Fas -/and Col1-CreERT2;Fas +/+ mice at 3 and 6 weeks ( Figure 9A). Although our scRNA-seq was conducted on purified Lin-fibroblasts, minor clusters of microvascular and lymphatic endothelial cells, alveolar epithelial cells, hematopoietic cells and erythroid cells were also identified (Supplemental Figure   10A,B). To focus our analysis on fibroblast populations, we removed these non-fibroblastic cell clusters, resulting in an final dataset of cell populations highly expressing canonical fibroblast genes (PDGFRa, PDGFRb and Acta2, Figure 9A and Supplemental Figure 10C). With this strategy, we identified 7 major cell groupings comprising 11 cell clusters in naïve mice and fibrotic mice. (Figure 9A and Supplemental Figures 11 and 12).
Next, we compared the gene expression patterns among the Wnt2+ lipofibroblasts, Col14a1+ MANCs and pro-fibrotic fibroblast subsets in bleomycin-instilled Col1-CreERT2;Fas -/and Col1-CreERT2;Fas +/+ mice at 3 and 6 weeks. To verify efficient deletion of Fas in Col1-CreERT2;Fas -/mice, we compared the gene expression pattern of the deleted exon 9 between Col1-CreERT2;Fas -/and Col1-CreERT2;Fas +/+ mice at 3 and 6 weeks. Expression of exon 9 was significantly reduced (p<0.001) and the number of cells expressing exon 9 was decreased 4.8fold (from 39% to 8%) at 3 weeks and 2.3-fold (from 43% to 19%) at 6 weeks in Fas-deficient cells (Supplemental Figure 13). Exon 9 was deleted similarly in all clusters in Col1-CreERT2;Fas -/mice (not shown). Similar to the differential expression analysis seen by bulk RNA-seq, Fas-deficiency in fibroblasts had little impact on the induction of genes related to pro-fibrotic pathways at 3 weeks in any of the fibroblast subsets, consistent with our data showing that fibroblast Fas-deficiency had no effect on the development of fibrosis. By contrast, comparison of specific fibroblast populations between Col1-CreERT2;Fas -/and Col1-CreERT2;Fas +/+ mice at 6 weeks revealed fibrosis-associated DEGs in the pro-fibrotic fibroblasts (clusters 5 and 6; 40 genes) and the Wnt2+ lipofibroblasts (clusters 1 and 2; 95 genes), but not in the Col14a1+ MANCs (clusters 3 and 4).
Lastly, to provide insight into the potential progenitors and lineages leading to the pro-fibrotic fibroblast populations (clusters 5 and 6) that emerge with injury and fibrosis, we performed in silico trajectory analyses of our single cell data. We included the Wnt2+ lipofibroblast population (clusters 1 and 2) in this analysis as potential progenitors given their related expression profile, exemplified by the connected nature of clusters 1, 2, 5, and 6 in UMAP space. Further supporting cluster 1 as a progenitor for clusters 5 and 6, we found the frequency of this cluster decreased from 30% in naïve animals to 20% and 14% as fibrosis developed at 3 and 6 weeks respectively.
Pseudotime trajectory analysis inferred two major trajectories: (Lineage 1): proceeding from cluster 1 through cluster 5 to cluster 6, and (Lineage 2): proceeding from cluster 2 through cluster 5 to cluster 6 ( Figure 9F). Through clustering of pseudotime-associated genes, we computed the trajectories into early, mid, and late phases. The pseudotime-associated genes for the early and  Table 4). Similar pathways were enriched during the early and mid phases of the lineage 2, but also included genes involved in glutathione metabolism (Gpx3) and VEGF signaling (Hspb1). Late trajectory genes included matrix metalloproteinase genes (Mmp3, Timp3) and genes involved in IGF1 signaling (Igfbp3, Igfbp6) ( Figure 9F, Supplemental Figure 14 and Supplemental Table 4). Thus, pseudotime analysis suggests that the pro-fibrotic fibroblasts (clusters 5 and 6) develop from the Wnt+ lipofibroblasts (clusters 1 and 2) in naïve lungs. Taking into consideration both the bulk and scRNA-seq data, our findings suggest that once each fibroblast cluster has undergone its unique pro-fibrotic programming response, the pro-fibrotic gene expression profile remains largely unchanged as long as fibrosis persists.

DISCUSSION
Substantial progress has been made in understanding the mechanisms of alveolar epithelial injury and its role in driving fibroblast recruitment, proliferation and excessive ECM deposition in the distal lung. However, the mechanisms that distinguish pro-fibrotic fibroblast persistence and ECM accumulation in the context of progressive pulmonary fibrosis seen in IPF from the fibrosis resolution and lung regeneration that occurs in the setting of recovery from ARDS, remain unclear.
Here, we show that Fas deletion in fibroblasts inhibits fibroblast apoptosis, impedes homeostatic fibrosis resolution, maintains pro-fibrotic transcriptomic fibroblast gene expression programing and permits fibroblast persistence and enduring pulmonary fibrosis. Taken together, these findings suggest that Fas signaling plays a fundamentally important, physiologic role in the elimination of pro-fibrotic lung fibroblasts during fibrosis resolution in vivo. Furthermore, our results suggest that impaired Fas signaling through Fas down-regulation or by increased expression of anti-apoptotic proteins that inhibit pro-apoptotic Fas signaling (11,(17)(18)(19)(20), may lead to critical deviations in lung fibrosis outcome from resolution to persistence.
Fibroblast Fas deficiency profoundly inhibited both the apoptosis and elimination of lung fibroblasts during homeostatic fibrosis resolution. Using a novel flow cytometry strategy for pulmonary fibroblasts, based on previous studies describing CD90+, CD90-and CD26+ fibroblasts in lung and skin (35,51), we found that among Fas-deficient lung fibroblasts, CD26+ and CD90-CD26-fibroblasts remained the most persistent and elevated subsets at 6 and 9 weeks. CD90+ fibroblasts underwent a modest early expansion, but also remained significantly elevated at 6 and 9 weeks in the absence of Fas signaling. Furthermore, whereas apoptotic S100A4 + and aSMA + fibroblasts were detected in the lungs of bleomycin-instilled Dermo1-Cre;Fas +/+ and Col1-CreERT2;Fas +/+ mice at 3 and 4.5 weeks, they were not detected in Dermo1-Cre;Fas -/and Col1-CreERT2;Fas -/mice suggesting that Fas-signaling plays an essential role in fibroblast apoptosis during homeostatic fibrosis resolution. The lungs of naïve Dermo1-Cre;Fas -/mice were found to have basally elevated numbers of Lin-PDGFRa+ cells compared to naïve Dermo1-Cre;Fas +/+ mice, with the majority being CD26+. These findings suggest that Fas signaling may be of greater relevance to basal turnover of CD26+ fibroblasts compared to CD90+ and CD90-CD26-fibroblasts. However, we did not observe spontaneous lung or skin fibrosis in these mice or in aged (i.e. >1-year old) Dermo1-Cre;Fas -/mice (not shown). We also noted that fibroblast Fas-deficiency in Dermo1-Cre;Fas -/and Col1-CreERT2;Fas -/mice had no effect on the initial bleomycin-induced fibroblast expansion at 1.5 and 3 weeks. Previous in vitro studies have suggested that Fas-signaling during this period contributes to pro-fibrotic fibroblast proliferation via cFLIPL and TRAF2-induced NF-kB activation (20). Our use of genetic in vivo approaches to delete Fas in fibroblasts challenge this notion. Taken together, our results suggest that Fas signaling plays a necessary role in fibroblast apoptosis and elimination during homeostatic fibrosis resolution following bleomycin-induced lung injury and likely contributes to fibroblast turnover in naïve mice.
Much remains to be understood about the functions of the increasingly heterogeneous lung fibroblast subsets. We initially studied how Fas-signaling affects CD90+, CD26+ and CD90-CD26fibroblasts. CD90 is expressed by Tcf21-expressing lipofibroblasts and Col13a1 interstitial matrixproducing fibroblasts (52). In the context of bleomycin-induced pulmonary fibrosis, CD90-deficient 20 mice phenocopy the impaired homeostatic fibrosis resolution described herein, and lung fibroblasts from CD90-deficent mice are resistant to Fas-induced apoptosis (10,53). CD26 has been reported to be expressed on dermal fibroblasts that synthesize and deposit ECM during scar formation (35) and we have shown it to identify a CD90-fibroblast subset. We have also shown that CD90+, CD26+ and CD90-CD26-fibroblasts from bleomycin-instilled mice exhibit increased scRNA-seq provided additional insight. We identified two major fibroblast subsets in the lungs of naïve mice. "Wnt2+ lipofibroblasts" (clusters 1 and 2) displayed similarity to Wnt2+ (30) and Col13a1 fibroblasts (29), but were also enriched with transcripts encoding lipofibroblast genes including Plin2 (29,52). "Col14a1+ MANCs" were similar to Axin+PDGFRa+ and Col14a1+ fibroblasts (29,30). At the peak of fibrosis, Wnt2+ lipofibroblasts from both Fas-sufficient and Fasdeficient mice displayed increased expression of typical pro-fibrotic genes. In contrast, gene expression profiles in Col14a1+ MANCs from bleomycin-instilled mice did not display a gene expression profile associated with the development or maintenance of fibrosis and very few DEGs were found in comparision with lung fibroblasts from naïve mice, as previously reported (30). We also identified a third, novel, pro-fibrotic fibroblast population (clusters 5 and 6) that was present in the lungs of both Fas-sufficient and Fas-deficient bleomycin-instilled mice, but was absent from the lungs of naïve mice. Our data supports the conclusion that these populations arise from the Wn2t+ lipofibroblasts (clusters 1 and 2) present in naïve lungs. In silico trajectory analysis provided two possible differentiation paths initating with either cluster 1 or 2, and passing through cluster 5 to generate the pro-fibrotic cluster 6 cells.

Analysis of the lung gene expression profiles in fibroblasts isolated from bleomycin-instilled
Col1-CreERT2;Fas +/+ mice at 6 weeks showed enrichment in pathways associated with regulation of cell migration and motility, wound healing, Wnt signaling and epithelial cell development, consistent with homeostatic fibrosis resolution. However, the Wnt2+ lipofibroblasts (clusters 1 and 2) and the pro-fibrotic fibroblasts (clusters 5 and 6) isolated from fibrotic bleomycin-instilled Col1-CreERT2;Fas -/mice continued to express pro-fibrotic gene expression patterns and pathways at 6 weeks, indicating that their pro-fibrotic programming had been preserved in the absence of Fassignaling.
Studies into the role of impaired Fas-signaling in tissue injury and fibrosis began to appear in the 1990s following the identification of a spontaneous inactivating Fas mutation in lpr mice (54).
These mice provided a new tool to address the consequences of Fas-inactivation in vivo.
However, they led to conflicting results when used to investigate the role of Fas in bleomycininduced acute lung injury and fibrosis (55,56). Loss of Fas signaling in lpr mice was also reported to reduce hepatocyte apoptosis and subsesuent hepatic fibrosis after bile duct ligation (57). A clear limitation of these early studies is that whole body Fas-inactivation does not allow for adequate modeling of the temporally distinct complex cellular interactions that are required to promote injury, fibrosis and repair where the impact of loss of Fas-signaling in alveolar epithelial cells may be quite different from that in fibroblasts or immune cells. The use of lineage-restricted and conditional Fas-inactivation in fibroblasts, as applied herein, circumvents this problem.
In summary, our data suggest that following initial expansion and pro-fibrotic programming of fibroblasts during the development of fibrosis, fibroblasts deficient in Fas maintain their pro-fibrotic transcriptomic programming in their persistent state. Therefore, autonomous Fas-induced apoptosis and clearance of pro-fibrotic fibroblasts during fibrosis resolution plays a necessary role in homeostatic fibrosis resolution, whereas impaired Fas-signaling is sufficient to maintain both their presence in persistently fibrotic lungs and their pro-fibrotic programming. We speculate that loss of the ability of fibroblasts to undergo Fas-induced apoptosis represents a decisive checkpoint at which the beneficial, homeostatic resolving fibrotic response might be diverted to the persistent and potentially more harmful progressive fibrosis seen in IPF patients.

Mouse strains
To enact lineage specific deletion of Fas, we bred mice that expressed endogenous Cre (Integrated DNA Technologies, Coralville, CA).

Primary Fibroblasts
Primary lung fibroblasts were isolated from healthy murine lungs as previously described (17,58). They were maintained in 10% DMEM and grown on plastic and used in experiments between passages 3-8.

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) to anesthetized mice, as previously described (24).
Fibrosis was assessed by lung measurements of collagen in the upper right lobe (hydroxyproline).
Briefly, lungs were homogenized in PBS and hydrolyzed overnight at 120 o C using 12M HCl.
Assessment of hydroxyproline was determined by the absorbance at 500nm on a microplate reader (Epoch2 BioTek Instruments, Winooski, Vermont). Histology was evaluated by Hematoxylin and Eosin (H&E) staining and Picrosirius red staining (PSR) of sections from the left lung as previously described (24). Images were taken on an upright Olympus BX51.

Flow cytometry and cell sorting
Single cell suspensions were obtained from perfused, enzymatically dispersed lungs. Briefly,

Bulk RNA sequencing analysis
To improve downstream mapping quality, raw sequencing reads were trimmed using skewer with parameters (end-quality=15, mean-quality=25, min=30). Trimmed reads were aligned to the mouse reference genome GRCm38 using Hisat2. Gene quantification was performed with htseqcount using GRCm38 ensembl v84 GTF. Differential expression analysis between groups was conducted with R package DESeq2. Pathway analsys was conducted with Enricher (59,60).

Single cell clustering and trajectory analysis
Initial processing of 10X scRNA-seq data, including cell demultiplexing, alignment to the mouse genome GRCm38, and UMI-based quantification was performed with Cell Ranger (version 3.0). To ensure that high quality cells were used for downstream analysis, we removed cells with fewer than 200 genes detected or cells with greater than 20% mitochondrial reads (Supplemental Figure 15). Additionally, to remove possible doublets, we remove cells with higher than 7500 genes detected. For gene filtering, we remove lowly expressed genes (detected in fewer than 4 cells). Using the above filtering, we have a dataset consisted of 12,043 cells and 37,887 genes.
After initial clustering and visualization, 4,215 cells were removed that we characterized as non- Using the previously computed UMAP components, trajectory analysis was performed using Slingshot that build lineages of cells that link cell clusters together by fitting a minimum spanning tree (MST) onto the selected clusters followed by the application of simultaneous principal curves to create trajectories. Clusters 1, 2, 5, and 6 were included in the trajectory analysis. Additional constraints were added by imposing cluster 1 and cluster 2 as the starting clusters and cluster 6 as the end cluster.

Statistical Overview
Time course data are presented as the mean ± SEM. Data were analyzed using GraphPad