PDGFRα and αSMA mark two distinct mesenchymal cell populations involved in parenchymal and vascular remodeling in pulmonary fibrosis.

X Valentina Biasin, X Slaven Crnkovic, Anita Sahu-Osen, Anna Birnhuber, X Elie El Agha, Katharina Sinn, Walter Klepetko, Andrea Olschewski, Saverio Bellusci, X Leigh M. Marsh, and Grazyna Kwapiszewska Division of Endocrinology and Diabetology, Department of Internal Medicine, Medical University of Graz, Graz, Austria; Ludwig Boltzmann Institute for Lung Vascular Research, Graz, Austria; Otto Loewi Research Center, Division of Physiology, Medical University of Graz, Graz, Austria; Excellence Cluster Cardio-Pulmonary System (ECCPS), Member of the German Center for Lung Research (DZL), Universities of Giessen and Marburg Lung Center (UGMLC), Justus Liebig University Giessen, Giessen, Germany; Division of Thoracic Surgery, Department of Surgery, Medical University of Vienna, Austria; and Experimental Anesthesiology, Department of Anesthesiology and Intensive Care Medicine, Medical University of Graz, Graz, Austria


INTRODUCTION
Pulmonary fibrosis is characterized by progressive scarring and stiffening of the lung, which lead to significant functional impairment and ultimately death (18,20). These processes are not only limited to the lung parenchyma but also affect the vascular bed. Vascular remodeling leads to lumen reduction and vessel occlusion (11), resulting in pulmonary hypertension (PH) and worsened patient survival (6,13).
Remodeling in both the parenchymal and vascular compartments is characterized by aberrant cellular proliferation, enhanced extracellular matrix (ECM) deposition, and increased tissue stiffness. Expansion of ␣-smooth muscle actin (␣SMA)expressing cells, termed myofibroblasts, is thought to be the major pathomechanism responsible for vascular and parenchymal remodeling. While resident ␣SMAϩ cells are indeed a major pathological cell type that expands and contributes to pulmonary vascular remodeling (7,28), neither vascular nor airway smooth muscle cells represent the major source of interstitial myofibroblasts in lung parenchymal remodeling (10). Furthermore, the simplistic paradigm of myofibroblasts as the single most important cell type in lung fibrosis has been questioned by a recent fate-mapping approach, which showed that ␣SMAϩ cells are not the main source of collagen production (29). After bleomycin-induced lung injury, not only myofibroblasts but also multiple resident cell populations expand and are involved in collagen production (27,32).
During late lung development in mouse, collagen-I-expressing platelet-derived growth factor receptor ␣ (PDGFR␣)ϩ cells are the source of lung alveolar myofibroblasts (21). However, in adult animals there are strong compartmental differences in PDGFR␣ and ␣SMA expression. We have recently demonstrated that PDGFR␣ expression marks a population of ␣SMA negative perivascular (adventitial) fibroblasts (7). Furthermore, after bleomycin challenge, interstitial PDGFR␣ϩ cells are ␣SMA negative, while perivascular and peribronchial PDGFR␣ϩ cells express variable amounts of ␣SMA (27). Cumulatively, these studies indicate that ␣SMAϩ and PDGFR␣ϩ cells might represent two major pathogenic mesenchymal populations, possessing distinct compartment-specific responses and behavior during fibrotic remodeling in the adult lung. As such, it is still unclear whether ␣SMAϩ cells in particular lung niches are derived from PDGFR␣ϩ cells and whether ␣SMAϩ cells are the sole contributor to collagen production in lung fibrosis.
In the current study, we investigated the cellular fates of PDGFR␣ϩ and ␣SMAϩ cells in two animal models displaying lung parenchymal and vascular remodeling with associated PH. Our genetic labeling strategy combined with immunofluorescence allowed investigating transdifferentiation events between the two populations. The contribution of both cell types was anatomically annotated and divided into parenchymal and vascular compartments in lung tissue samples from both animal models as well as human donor and end-stage idiopathic pulmonary fibrosis patients. This comprehensive methodological approach revealed distinct fates of PDGFR␣ϩ and ␣SMAϩ cells, their association with collagen production, and a compartment-specific contribution to lung fibrosis.

MATERIALS AND METHODS
Full experimental details are given in the online supplement at the following link: https://doi.org/10.5281/zenodo.3532795.

Human Lung Tissue
Lung tissue from idiopathic pulmonary fibrosis (IPF) patients undergoing lung transplantation were obtained from the Department of Surgery, Division of Thoracic Surgery, Medical University of Vienna, Austria, following written informed consent and approval by the institutional ethics committee (976/2010). Samples of downsized transplant donor lungs served as controls.

Immunofluorescence Staining and Analysis
Mouse lungs were perfused with PBS followed by inflation with optimal cutting temperature compound (Tissue Tek, Sakura, CA), fixed overnight in 1% paraformaldehyde followed by overnight dehydration in 30% sucrose and stored at Ϫ80°C. Cryo-sections were treated with ice-cold methanol/acetone, blocked with 2.5% horse serum (Vector Laboratories, Palo Alto, CA), and stained with antibodies against PDGFR␣ (1:1,000; Cell Signaling, Leiden, The Netherlands), ␣SMA-Cy3 (1:200, Sigma Aldrich), ␣SMA-FITC (1:100, Sigma Aldrich), desmin (1:100, R&D systems, Minneapolis MN), pro-collagen I (1:500, Southern Biotech, Birmingham AL), and von Willebrand factor (vWF) (1:100; Dako/Agilent, Santa Clara, CA). Sections were incubated with respective secondary antibodies conjugated with Alexa Fluor 488, 555, or 647 (1:500, all from ThermoFisher Scientific, Bonn, Germany) and mounted with DAPI containing mounting media (Vector Laboratories). PDGFR␣ signal was enhanced by the Tyramide Signal Amplification Kit (TSA-488, ThermoFisher Scientific) according to the manufacturer's instruction. For each sample 10 -15 images from random areas were acquired at ϫ40 magnification (Image size x: 225 m; y: 22 5m) under an LSM510 confocal microscope (Zeiss, Oberkochen, Germany). Vascular compartment was identified by characteristic morphological appearance (Supplemental Fig. S2; https://doi.org/ 10.5281/zenodo.3532795) and validated by vWF endothelial staining. Parenchymal regions were defined by absence of vascular structure and vWF staining within the imaged region. Manual counting of tdTomato-, AF488-, and AF647-positive cells was performed by two independent investigators from 3-10 XY panels from each sample. Lineage-traced cells containing obvious nuclei were counted. Brightfield picture image was merged to the immunofluorescence signal to help delineating the cell border; however, in unclear cases the signal/cells were not included in the analysis. Blinding of samples was not possible due to obvious morphological differences in the lung tissue.

RNA Sequencing
The left lung lobe from tdTomato mice was mechanically separated with two scalpels followed by incubation with dispase (50 U/mL, Corning, NY) for 1 h at 37°C to generate a single cell suspension. tdTomato-positive cells were sorted directly into RNA lysis buffer (Qiagen, Venlo, The Netherlands) using a FACSAria II cell sorter (BD Biosciences, San Jose, CA). RNA was isolated with the RNeasy micro kit (Qiagen). Library preparation using the SmartSeq2 protocol and sequencing on the Illumina HiSeq 3000/4000 platform was done by the Biomedical Sequencing Facility (CeMM, Vienna, Austria). Next-generation sequencing reads were aligned with the TopHat2 (v2.1.1) (12) splice junction mapper using the Bowtie2 short read aligner (v2.2.9) (16). Transcriptome assembly and differential expressing calling was performed with Cufflinks (v2.1.1) (30). A detailed analysis of the initial analysis can be found in the online supplement. Differentially regulated genes were ranked according to their log-fold-change and their significance (q-value). Prcomp was used to calculate the principal components; the first two principal components were plotted using the ggplot2 package. For generation of heatmaps, data were transformed to log2(FPKMϩ1). Gene enrichment analysis was performed using EnrichR (5,15). The data discussed in this publication have been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO) (8) and are accessible through GEO Series accession number GSE126205.

Public Data Set Analysis
Human. scRNA-Seq data from Reyfman et al. (25) was downloaded from GEO (GSE122960), and the raw count matrices in HDF5 format imported and analyzed in Seurat 3.1.2 (https://linkinghub.elsevier.com/ retrieve/pii/S0092867419305598). Four donor samples (GSM3489182, GSM3489185, GSM3489187, GSM3489189) and four IPF samples (GSM3489183, GSM3489184, GSM3489188, GSM3489190) were individually processed and normalized using the SCTransform (10a) function removing cells with Ͼ10% mitochondrial percentage.Samples were concatenated using SCTransform, and dimension reduction was performed by PCA and t-SNE using default parameters. Cells were clustered at a resolution of 0.4. Fibroblasts clusters were identified by fibroblast markers as identified in Ref. 25.
Mouse. scRNA-Seq data from Peyser et al. (23) was downloaded from GEO (GSE129605), and the Feature-Barcode Matrices were imported in and analyzed in Seurat. Samples were individually processed, removing cells with high mitochondrial percentage Ͼ5%, and data were normalized using default parameters. Samples were concatenated using a precomputed anchor set identified by the function FindIn-tegrationAnchors. Concatenated samples were then scaled to regress out differences in number of features per cell, and dimension reduction was performed by PCA and t-SNE using default parameters. Clustering was performed at a resolution of 0.3. Fibroblast clusters were annotated by cluster alignment against collected mouse data sets available at GEO (1) using SingleR.

Statistical Analysis
Statistical analysis was performed with GraphPad Prism 5 and bioinformatic analysis was performed with RStudio (https://www. rstudio.com) and R (www.r-project.org) (version number 3.4.1). Data are presented as mean with SD in all graphs. Statistical differences between the groups were determined by using two-way ANOVA with Bonferroni post hoc comparison test. P values Ͻ0.05 were considered significant.

␣SMA and PDGFR␣-expressing Cells Represent Distinct Subsets of Collagen-producing Fibroblasts in Human Lungs
To determine the relative contribution of ␣SMA and PDGFR␣ cells in parenchymal and vascular remodeling associated with lung fibrosis, multicolor immunofluorescent staining against PDGFR␣, ␣SMA, and vWF was performed. Increased numbers of both ␣SMAϩ (~2.5-fold) and PDGFR␣ϩ (~2-fold) cells were observed in the lung parenchyma of patients with IPF compared with donors ( Fig. 1, A and B). The mean percentage of myofibroblasts (␣SMAϩ cells) in the fibroblast cell pool of the parenchymal compartment (total number of ␣SMAϩ and PDGFR␣ϩ cells) was increased from~20% in the donors to~30% of the IPF patients (Fig. 1B). The vast majority of cells were positive for either ␣SMA or PDGFR␣, while the percentage of double ␣SMAϩ/PDGFR␣ϩ cells reached maximally 12% in IPF lungs (Fig. 1B).
In the vascular compartment of IPF patients, we observed an increase in both PDGFR␣ϩ (~3-fold) and ␣SMAϩ (~2-fold) cells (Fig. 1C). The percentage of double positive cells was low in donors (~5%) and increased (to~15%) in IPF remodeled vessels (Fig. 1C). The localization pattern was unchanged between donor and IPF vasculature; in both groups ␣SMAϩ cells were confined to vascular regions up to the external elastic lamina ( . Importantly, the investigation of an independent scRNA-Seq data set (GSE122960) analyzing human IPF and donor lungs (25) revealed that, within the fibroblast cluster, ACTA2-and PDGFRA-expressing cells delineate two distinct subclusters with minimal overlap (Fig. 1D). This finding corroborates and reinforces our observations. We next assessed the contribution of ␣SMAϩ and PDGFR␣ϩ cells to collagen production. In both compartments of donor and IPF patients, approximately half of ␣SMAϩ cells displayed intracellular costaining with collagen (COL1), while almost all PDGFR␣ϩ cells were COL1 positive (Fig. 2, A-C). Preferential expression of collagens or contractile components is traditionally used to further subclassify ␣SMAϩ cells into synthetic or contractile phenotype (24). We therefore addressed the expression of the contractile marker desmin (DES) in ␣SMAϩ and PDGFR␣ϩ cells. As shown in Fig. 2, D-F, DES almost exclusively colocalized with ␣SMAϩ cells. In IPF lungs the percentage of ␣SMAϩ/DESϩ cells increased compared with donors in both parenchymal and vascular compartment (Fig. 2, E and F).

␣SMAϩ Cells Are a Separate Subpopulation of Fibroblasts with Compartment-specific Contribution to Remodeling
Our human data revealing limited overlap between ␣SMAϩ and PDGFR␣ϩ cells, including their divergent expression of collagen and desmin, indicated that these cell populations might represent two main and distinct fibroblast populations in the adult lung. We therefore employed a fate-mapping approach using transgenic mice to assess lineage hierarchy and possible transdifferentiation between ␣SMAand PDGFR␣expressing cells. We employed two murine models in which the parenchymal and vascular remodeling process is driven by two different underlying pathomechanisms: 1) bleomycin-induced lung fibrosis and 2) ectopic overexpression of Fra-2 (Fra-2 Tg). Both models accumulate ␣SMAϩ cells and collagen in parenchymal regions (Supplemental Fig. S3, A-D; https://doi.org/10.5281/zenodo.3532795), have increased coverage of distal pulmonary vessels by ␣SMAϩ cells with deposition of intra-/perivascular collagen (Supplemental Fig.  S4, A-D; https://doi.org/10.5281/zenodo.3532795), and have impaired lung function with restrictive phenotype (2)(3)(4).
In the parenchymal compartment of bleomycin-treated mice, the total number of Acta2-tdT cells increased (~5-fold) compared with the saline control group (Fig. 3, A and B). However, additional immunostaining against PDGFR␣ revealed that the vast majority of cells in fibrotic regions were PDGFR␣ϩ and that Acta2-tdT cells only represented~20% of the total fibroblast pool (total number of ␣SMAϩ and PDGFR␣ϩ cells) (Fig. 3, A and B). In contrast, the vascular compartment contained mostly lineage-labeled Acta2-tdT cells (Fig. 3C). Postbleomycin, despite a strong increase in number of PDGFR␣ϩ cells, the relative proportion of both cell types remained relatively constant in the parenchyma (Fig. 3B), suggesting an equal expansion of both cell population in bleomycin. In contrast, in the vasculature, the percentage of PDGFR␣ϩ cells increased compared with the percentage of Acta-tdTϩ cells (Fig. 3C). In both saline-and bleomycintreated mice only a limited percentage of cells (~5-8%) in either compartment coexpressed both markers (Fig. 3, B  and C).

PDGFR␣-expressing Cells without ␣SMA Expression Constitute the Bulk of Cells in Parenchymal Remodeling
The low percentage of double Acta2-tdT/PDGFR␣ϩ cells prompted us to validate these results by a reciprocal approach. Here the fate of PDGFR␣ϩ cells was followed using a second lineage-tracing tool, Pdgfra-CreERT2; tdTomato flox (Pdgfra-tdT) in the bleomycin and Fra-2 Tg models (Figs. 5 and 6). The colocalization of cytoplasmic tdTomato reporter and plasma membrane PDGFR␣ was considered positive only if both signals were present in the same cell (Supplemental Fig. S1, dashed lines and arrows; https://doi.org/10.5281/zenodo.3532795).
An expansion of Pdgfra-tdT cells was observed in the parenchymal regions of bleomycin compared with saline- treated mice (~1.5-fold increase; Fig. 5, A and B). No change in the percentage of double Pdgfra-tdT/␣SMAϩ cells was observed in the parenchyma of bleomycin-treated mice (~3%, Fig. 5B). In line with the PDGFR␣ immunostaining of the Acta2-tdT line (Fig. 4), lineage-labeled Pdgfra-tdT cells represented the majority of cells in fibrotic regions and only a minority in the vascular compartment of bleomycin mouse model (Fig. 5C). Pdgfra-tdT cells were mostly confined to the perivascular region and accounted for~12% of cells in bleomycin-treated mice (Fig. 5C). Double Pdgfra-tdT/␣SMAϩ cells were not detected in the vessels of saline mice, while in bleomycin-treated mice their percentage increased to 4% (Fig. 5C).
Fra-2 Tg mice displayed similar number of Pdgfra-tdT cells compared with littermate control mice (Fig. 6, A and  B), but the resulting ratio of Pdgfra-tdT to ␣SMAϩ cells decreased in the lung parenchyma, changing from 70% in controls to 40% in Fra-2 Tg mice (Fig. 6B). Again, the presence of Pdgfra-tdT/␣SMA double positive cells was a rare event (~3-5%) in Fra-2 Tg and littermate controls (Fig.  6B). In the lung vasculature the number of perivascular Pdgfra-tdT cells did not change significantly in Fra-2 Tg compared with control littermates (Fig. 6C). Again, we observed almost no double Pdgfra-tdT/␣SMAϩ cells in the vascular compartment of normal and remodeled vessels (Fig. 6C).

␣SMA and PDGFR␣ Cells Are Characterized by a Separate and Distinct Gene Expression Profile
Both our human data and lineage tracing experiments indicated that ␣SMAand PDGFR␣-expressing cells represent two largely separate populations in the normal and fibrotic adult lungs with apparently limited transdifferentiation capacity. We next expanded our investigations and performed RNA sequencing on tdTomato-positive cells sorted from the lungs of Acta2-tdT and Pdgfra-tdT mice (Fig. 7). Representative examples of gating strategy for Acta2-tdT and Pdgfra-tdT cells are shown in Supplemental Fig. S5 (https://doi.org/10.5281/zenodo.3532795). Principal component analysis (PCA) of the global gene expression profiles and heatmap analysis revealed that Acta2-tdT and Pdgfra-tdT cells indeed represent two distinct cellular clusters in normal adult murine lungs (Fig. 7, B and C). Gene ontology (GO) analysis of molecular functions revealed that Acta2-tdT cells were enriched in actin and tropomyosin binding, suggesting a predominance of the contractile phenotype in these cells, while Pdgfra-tdT cells in MAPK kinase binding and complement component and PDGF binding (Supplemental Table S1A; https://doi.org/10.5281/ zenodo.3532795). Although bleomycin challenge (Fig. 7D) induced significant gene expression changes in both Acta2-tdT and Pdgfra-tdT cell populations, it did not affect Acta2-tdT and Pdgfra-tdT distinct cellular clusters (Fig. 7, E and F, and Supplemental Table S1B; https://doi.org/10.5281/zenodo.3532795). As expected, bleomycin induced significant changes in both Acta2-tdT and Pdgfra-tdT cells compared with saline treatment (Fig. 7, G-I and J-L). Our GO molecular functions were unique for each population. RNA binding, translation factor activity, and integrin binding were among the 20 most enriched molecular functions in Acta2-tdT cells from bleomycin-treated mice (Supplemental Table S1C; https://doi.org/10.5281/zenodo.3532795). In the Pdgfra-tdT population, there was enrichment for endopeptidase activity and cholesterol transporter activity upon bleomycin treatment (Supplemental Table S1C; https://doi.org/10.5281/zenodo. 3532795). Importantly, the molecular function collagen binding(SupplementalTableS1C;https://doi.org/10.5281/zenodo. 3532795) was enriched in both populations upon bleomycin treatment, suggesting that both Acta2-tdT and Pdgfra-tdT cells are associated with collagen.

Murine Models Show Cellular Collagen Expression with Both ␣SMAand PDGFR␣-expressing Cells but Preferential Association of Desmin with ␣SMA-expressing Cells
Our human histological analysis and RNA-Seq results revealed shared collagen production, but skewed expression of contractile machinery proteins between ␣SMAand PDGFR␣expressing cells. In the final step, we therefore investigated the association of collagen and desmin marker expression with each of those cell populations in murine pulmonary fibrosis models.
In contrast to the human lungs where nearly all PDGFR␣ϩ cells were positive for collagen, in the bleomycin model onlỹ 10% of parenchymal and~30% of perivascular Pdgfra-tdTϩ cells costained with COL1 (Fig. 8, A-C). Moreover, bleomycin treatment showed~20% of parenchymal and~10% of perivascular double ␣SMA/COL1 ϩ cells (Fig. 8, B and C) Similar to the human lungs, DES expression was observed preferentially in ␣SMAϩ cells in bleomycin (Fig. 8D). Interestingly, while we observed an~20 -30% increase in ␣SMAϩ/DESϩ cells in the parenchyma of bleomycin, the percentage of double positive cells decreased by~30% in the vascular compartment of bleomycin compared with the saline controls (Fig. 8, E and F). These finding were further substantiated by analyzing an independent public scRNA-Seq data set (GSE129605) derived from lungs of saline-and bleomycin-treated mice (23). In the fibroblast cell pool of both bleomycin-and saline-treated mice, the expression of Acta2 and Pdgfra was limited to distinct cell subsets with limited overlap (Fig. 8G). Similar to the fluorescence approach, expression of collagen (Col1a1) expression was detected in both Acta2 and Pdgfra subsets, while desmin (Des) was mostly restricted to Acta2 cells (Fig. 8G).
In Fra-2 Tg mice, collagen deposition in the vascular compartment was more pronounced than in bleomycin-treated mice and was observed both peri-and intravascularly (Fig. 9).  Similar to the bleomycin model, only~10 -20% of parenchymal and~30 -40% of perivascular Pdgfra-tdTϩ cells costained with COL1 (Fig. 9, B and C) in Fra-2 Tg. The percentage of double ␣SMAϩ/COL1ϩ cells was, similar to the human lungs,~50% and~65% of in the parenchymal and vascular compartments, respectively (Fig. 9, B and C). DES expression was again almost exclusively observed in ␣SMAϩ cells (Fig.  9, D-F).

DISCUSSION
Understanding the cellular heterogeneity accompanying parenchymal and vascular remodeling is an important step in the development of novel antifibrotic therapies. Previously, myofibroblasts (␣SMAϩ cells) were speculated to be the main source of collagen production in the lung, however, this paradigm has recently been questioned in mouse lung fibrosis (29).

L693 FIBROBLAST DIVERSITY IN SPECIFIC LUNG COMPARTMENTS
Furthermore, recent studies have given first insights into underlying differences in mesenchymal populations (25,34). It is becoming increasingly obvious that the mesenchymal cell pool is a highly heterogeneous mixture of cells, yet how much of this difference is due to predetermined cell fates and different lineage trajectories versus more plastic cell states and phenotypic modulation is currently unknown. The main finding of our study is that transcriptionally distinct cell populations marked by high expression of either ␣SMA or PDGFR␣ represent independent lineages in adult lungs with a characteristic contribution to parenchymal and vascular remodeling that is shared between murine models and human patients.
Using a binary lineage-tracing approach (Acta2-tdT and Pdgfra-tdT) in two distinct animal models of lung fibrosis (bleomycin and Fra-2 Tg), we show anatomically preferential localization of ␣SMAand PDGFR␣-expressing cells at baseline and in fibrosis. We identified a compartment-specific fibrotic response in which vascular remodeling is mostly represented by the expansion of resident ␣SMAϩ cells, with the contribution of PDGFR␣ϩ cells limited to the perivascular regions, and parenchymal remodeling consisting predominantly of PDGFR␣ϩ cells. Importantly, the fate-mapping approach supported the immunofluorescent histological analysis showing limited overlap between ␣SMA and PDGFR␣ expression. The relatively low number of lineage-labeled cells that are coexpressing both ␣SMA and PDGFR␣ indicate a low transdifferentiation potential between these two cell populations in adult lungs and imply that even during a fibrotic response, these two populations are maintained and behave as separate lineages. This data interpretation was further supported by the analysis of open access scRNA-Seq data set from human donor and IPF lungs deposited in the public domain by an independent research group (25). Additionally, our global gene expression analysis revealing divergent transcriptome profiles between the two populations further suggest not only separate lineages but specialized biological roles as well. Our results suggest that independent accumulation of these cell types is the main cellular mechanism underlying pathological lung remodeling in adulthood. This is in contrast to a recent study that used a tetracycline-inducible Pdgfra lineage tracer line and reported that up to 40% of myofibroblasts originate from Pdgfra lineage upon bleomycin injury (17). The reason for this discrepancy with our results is currently unknown and would require additional clarification but could be the result of efficiency and specificity of tamoxifen labeling and use of different myofibroblast markers (␣SMA versus SM22␣). Different lineage tracing mouse lines and strain could also affect the results as we have observed a lower number of labeled PDGFR␣ between saline and wild type littermate controls.
The observed differences between these two major populations on a cellular level could form a basis of variable disease progression and heterogeneity observed in the clinical setting. Indeed, a growing body of evidence indicates that diverse lung-resident mesenchymal populations contribute to pulmonary fibrosis (27,32,34). This is particularly mirrored by their specific set of genes upregulated in fibrotic condition. While ␣SMA ϩ cells upregulated mostly ECM genes, PDGFR␣ϩ cells were mostly enriched in inflammatory and ribosomal genes.
The involvement of vascular remodeling in fibrotic lung disease is an underinvestigated field. We have previously shown extensive vascular remodeling in advanced IPF and specific gene expression patterns (11). However, cellular changes in the vascular compartment and the fate of different cell types during fibrosis was not addressed. In this study, we have demonstrated that vascular ␣SMAϩ cells originate from pre-existing smooth muscle cells applying two independent lung fibrosis models. This is further supported by the exclusive expression of the contractile marker DES on ␣SMAϩ cells. This is in line with our previous study where similar results were obtained by using different remodeling stimuli such as hypoxia or allergen challenge (7,19). Additionally, in both our murine models we could show that the percentage of ␣SMAϩ cells expressing DES decreased, while those positive for COL1 increased, suggesting either a shift from a contractile to a synthetic phenotype in ␣SMAϩ cells or expansion of the COL1ϩ/␣SMAϩ population in fibrotic conditions. In human IPF, COL1-and DES-positive ␣SMAϩ cells were both increased, suggesting a different dynamic of ␣SMA-mediated remodeling in mice and humans. It is also possible that the observed differences reflect different stages of the remodeling process, which is still ongoing in mice but at an end stage in the human IPF samples.
The contribution of ␣SMAϩ cells in pulmonary fibrosis has been extensively addressed; however, the source of parenchymal myofibroblasts remains elusive. A recent study reported axis inhibition protein 2 (Axin2)ϩ cells as the source of ␣SMA-expressing myofibroblasts upon injury (35), while double Axin2/PDGFR␣-positive cells represented a distinct lineage showing a proliferative response upon lung injury (35). Another study reported that glioma-associated oncogene homolog 1 (Gli1)-positive cells give rise to ␣SMAϩ cells, while only a minor percentage of Gli1ϩ cells retained PDGFR␣ expression in the lung (14). However, an independent study delineating the transcriptional consequences of hedgehog pathway activation in pulmonary fibroblasts revealed major effects on cell cycle genes, but no increase in contractile machinery, indicating a lack of transdifferentiation into myofibroblasts (22). Similarly, we demonstrate an expansion of PDGFR␣ϩ cells with minimal ␣SMA coexpression. This is in contrast to observations from lung development studies where PDGFR␣ϩ cells have been identified as the main source of interstitial ␣SMAϩ cells (alveolar myofibroblasts) (21), implying the importance of contextualizing the obtained finding. Possible explanations for these differences could be the loss of cell plasticity of a common progenitor cell in adulthood. While PDGFR␣ expression might be a shared  persists as a small cell population into adulthood. Irrespective of the origin of the double positive cells, our reciprocal fate mapping approach effectively excludes a significant contribution of this subpopulation to fibrotic remodeling. The overarching conclusion is that the appearance and expansion of parenchymal ␣SMAϩ cells (myofibroblasts) seem to be independent of PDGFR␣ϩ cells in adulthood, but rather controlled by different cellular mechanisms, possibly including expansion of resident alveolar myofibroblasts and upregulation of ␣SMA expression in other lineages.
Our results from lineage tracing and immunofluorescence staining in bleomycin and Fra-2 Tg animal models and in human IPF lungs revealed that both PDGFR␣ϩ and ␣SMAϩ cells exhibit collagen positivity. This finding was also confirmed in our bulk RNA-Seq data set and further supported by an independent data set derived from bleomycin-and salinetreated mice (23). The finding of collagen production being not exclusively produced by ␣SMAϩ cells but also by PDGFR␣ϩ cells is in line with a recent study showing that in bleomycininduced lung fibrosis, PDGFR␣ϩ cells are the main matrix producing fibroblasts while only a subset of ␣SMAϩ cells expressed collagen (34). We confirmed and expanded on these findings of the two cellular populations with compartment specificity in two animal models and most importantly in human samples. While we observed an involvement of PDGFR␣ϩ cells in collagen production during the active phase of lung fibrosis development, ␣SMAϩ cells have been shown to play an important role in the resolution phase of bleomycin-induced lung fibrosis (10), suggesting that the contribution of these two cell populations to the lung fibrosis process also depends on their different temporal appearance.
A key conclusion from the present study is that the biology not only of ␣SMAϩ cells, but also that of PDGFR␣ϩ cells should be incorporated into studies focusing on lung fibrosis. Future studies aimed deciphering the underlying molecular pathways governing the expansion and appearance of distinct fibroblast populations could offer novel and more efficient therapeutic targets for parenchymal and vascular remodeling.