Identification of a myofibroblast differentiation program during neonatal lung development

ABSTRACT

Alveologenesis is the final stage of lung development in which the internal surface area of the lung is increased to facilitate efficient gas exchange in the mature organism. The first phase of alveologenesis involves the formation of septal ridges (secondary septae) and the second phase involves thinning of the alveolar septa. Within secondary septa, mesenchymal cells include a transient population of alveolar myofibroblasts (MyoFBs) and a stable but poorly described population of lipid-rich cells that have been referred to as lipofibroblasts or matrix fibroblasts (MatFBs). Using a unique Fgf18^CreER lineage trace mouse line, cell sorting, single-cell RNA sequencing and primary cell culture, we have identified multiple subtypes of mesenchymal cells in the neonatal lung, including an immature progenitor cell that gives rise to mature MyoFB. We also show that the endogenous and targeted ROSA26 locus serves as a sensitive reporter for MyoFB maturation. These studies identify a MyoFB differentiation program that is distinct from other mesenchymal cell types and increases the known repertoire of mesenchymal cell types in the neonatal lung.

Highlighted Article: During primary alveologenesis, alveolar myofibroblasts make up a proliferative mesenchymal lineage that is distinct from matrix fibroblasts and mature through several stages to populate emerging secondary septa.

INTRODUCTION

Alveologenesis is the final stage of lung development where the gas exchange surface area of the lung is increased by subdivision of alveolar saccules followed by thinning of the secondary septae (Rippa et al., 2021). This postnatal developmental process is essential to forming a fully functional adult lung. Impaired alveologenesis is a major cause of morbidity in extremely preterm infants and often results in respiratory distress syndrome, which can lead to bronchopulmonary dysplasia (BPD), one of the most significant chronic morbidities affecting nearly half of extremely preterm infants born between 22 and 28 weeks of gestation (Bancalari and Jain, 2019; Shah et al., 2012; Stoll et al., 2015; Whitsett and Weaver, 2015).

Alveologenesis occurs in several phases (Amy et al., 1977; Boström et al., 2002; Mund et al., 2008; Rippa et al., 2021; Vila Ellis and Chen, 2021). In the mouse, an initial expansion of alveolar surface area occurs between embryonic day 16.5 (E16.5) and postnatal day 2 (P2) in the absence of alveolar myofibroblasts (MyoFBs), and is referred to as the initial or saccular phase. This is followed by the classical or first phase from P3 to ∼P14, in which MyoFBs populate septal ridges, also referred to as secondary crests or secondary septae, that partition the alveolar saccule and thus expand the surface area of the lung. The second phase of classical alveologenesis occurs from ∼P15 to ∼P36 in the absence of MyoFBs. During this phase, alveolar size and number continue to increase, and, importantly, the septal walls become thinner to facilitate efficient gas exchange.

Septal ridges contain multiple mesenchymal cell types, capillaries and an elastin rich extracellular matrix (ECM), and are covered by alveolar type 1 (AT1) and alveolar type 2 (AT2) cells (Branchfield et al., 2016; Guo et al., 2023; Yang et al., 2016). Of the mesenchymal cells within the septal ridge, MyoFBs (also called secondary crest myofibroblasts) produce high levels of elastin, and, together with their intrinsic contractile properties, are essential for secondary septation (Endale et al., 2017; Guo et al., 2023; Li et al., 2019, 2015; Lindahl et al., 1997; Narvaez Del Pilar et al., 2022; Sun et al., 2022; Vila Ellis and Chen, 2021; Zepp et al., 2021a). The other major mesenchymal cell type is the matrix fibroblast (MatFB), which contains a large subpopulation of lipid-rich cells (lipofibroblasts) and other poorly defined mesenchymal cell types (Riccetti et al., 2020; Rippa et al., 2021). The function of lipofibroblasts is poorly understood (Park et al., 2019), but they may function to support surfactant and phospholipid synthesis by adjacent AT2 cells (McGowan and Torday, 1997). MatFBs also express FGF10, which signals through FGFR2b and is necessary for survival of AT2 cells in adult lung (Dorry et al., 2020; Liberti et al., 2021; Yuan et al., 2019). It is hypothesized that FGF10 is also necessary to maintain AT2 cells during alveologenesis (Chao et al., 2016; El Agha et al., 2014).

The MyoFBs are a transient cell population that functions during the first phase of alveologenesis. Genes that are highly expressed in MyoFBs, such as Acta2 (αSMA), are greatly reduced by the end of the first phase (∼P14) suggesting that MyoFBs are lost or, alternatively, differentiate into another cell type, following their functional role in alveologenesis (Branchfield et al., 2016; McGowan et al., 2008; Riccetti et al., 2020; Yamada et al., 2005). Additionally, genetic lineage tracing studies show that MyoFBs are lost by the end of the first phase and are largely absent in adult lung (Duong et al., 2022; Gao et al., 2022; Hagan et al., 2020; Li et al., 2015; Narvaez Del Pilar et al., 2022; Zepp et al., 2021a).

Several genetic tools have been developed to label and lineage trace mesenchymal cells in the mouse lung (Riccetti et al., 2020). Pdgfra^EGFP [Pdgfra^{tm11(EGFP)Sor/J}] expresses a nuclear localized H2BEGFP (GFP) gene in a broad range of mesenchymal cells under the control of the endogenous Pdgfra locus (Hamilton et al., 2003). In the neonatal lung, the level of GFP fluorescence intensity, as a proxy for Pdgfra expression, varies, with low levels (Pdgfra-GFP^Low) in MatFBs and high levels (Pdgfra-GFP^High) in MyoFBs (Endale et al., 2017; Kimani et al., 2009; McGowan and McCoy, 2014; Zepp et al., 2021a). Similarly, the adult lung also contains Pdgfra-GFP^Low and Pdgfra-GFP^High populations, although in the adult, Pdgfra-GFP^High cells have a MatFB gene signature, whereas Pdgfra-GFP^Low cells have a MyoFB gene signature (Green et al., 2016). In the neonatal lung, the different intensities of GFP fluorescence allow sorting dissociated lung cells to isolate MatFBs and MyoFBs.

Fibroblast growth factor 18 (Fgf18) is expressed in MyoFBs and AT1 cells but not in MatFBs (Hagan et al., 2020; McGowan and McCoy, 2015; Ruiz-Camp and Morty, 2015). Consistent with this, activation of FGF18(TDT) (Fgf18^CreER;ROSA^TDTomato) from P5-P8 labeled nearly all MyoFB and most AT1 cells at P9 with TDTomato (Hagan et al., 2019; Hagan et al., 2020). By P21, most lineage-traced MyoFBs were no longer detected; however, lineage-labeled AT1 cells persisted.

Here, we have used a combination of spatial localization, cell sorting, single-cell RNA sequencing (scRNA-seq) and primary culture of sorted mesenchymal cells to identify a spectrum of mesenchymal cell types in the mouse neonatal lung. This study shows a separation of MyoFB and MatFB lineages in the neonatal lung, and identifies a progression from immature progenitors to mature MyoFBs in sequence data, immunostaining patterns and in in vitro cell culture. We have also uncovered an unanticipated regulation of the endogenous and targeted ROSA26 locus [Gt(ROSA)26Sor] as a sensitive reporter for MyoFB maturation. These studies increase the known repertoire of mesenchymal cell types in the neonatal lung and provide a neonatal time point for comparison with embryonic and adult lung mesenchymal cell subtypes, with implications for activation of developmental programs that could function during lung injury and repair.

RESULTS

Localization of myofibroblasts and matrix fibroblasts relative to the elastin lung scaffold

Fgf18^CreERT2;ROSA^TDTomato;Pdgfra^EGFP mice label MyoFBs, AT1 cells and mesothelial cells in red (Fgf18-TDTomato lineage), and MyoFB and MatFB in green in the neonatal mouse lung (Endale et al., 2017; Hagan et al., 2020; Li et al., 2018; McGowan et al., 2008). Pdgfra^EGFP labels MatFBs with lower GFP intensity and MyoFBs with higher GFP intensity (Kimani et al., 2009; McGowan and McCoy, 2014). To localize these cells relative to the P7 neonatal elastin lung scaffold, we used optically cleared lungs and Alexa Fluor 633 Hydrazide, which selectively stains elastin in the alveolar entry rings and septal ridges (Fig. 1A-D) (Shen et al., 2012). Confocal imaging showed a dense elastin matrix in the mesothelial surface (top 5 µm) and elastin bundles outlining alveolar entry rings in the underlying tissue. Pdgfra^EGFP (GFP) labeled the nucleus of cells that were distributed throughout the lung, but was excluded from the mesothelium (Fig. 1E,F). Induction of the FGF18(TDT) (Fgf18^CreERT2;ROSA^TDTomato) lineage trace with tamoxifen from P2 to P5 showed abundant TDTomato-labeled cells throughout the lung and mesothelium (Fig. 1A,E-G). Merged color channels clearly identify at least four distinct cell types: MatFBs [green, PDGFRA(GFP)^Low], MyoFBs [green, PDGFRA(GFP)^High; yellow, PDGFRA(GFP)^High+FGF18(TDT)], AT1 cells and mesothelial cells [red, FGF18(TDT)]. High-magnification images show MyoFBs in close proximity to the elastin bundles and immature FGF18(TDT)-negative MyoFBs (see below), and MatFBs towards the base of septal ridges (Fig. 1G,H, Movie 1). This pattern of cellular localization is consistent with numerous studies using a variety of cellular markers that label MyoFBs and MatFBs (Branchfield et al., 2016; Endale et al., 2017; Hagan et al., 2020; Kimani et al., 2009; McGowan et al., 2008; McGowan and McCoy, 2014; Riccetti et al., 2020).

Fig. 1.

High-resolution imaging of elastin and lineage-labeled cells in the postnatal day 7 mouse lung. (A) Experimental plan for imaging Fgf18^CreER;ROSA^TDTomato;Pdgfra^EGFP cleared lung tissue using CLARITY. (B-H) XY- (C-E,G) and Z- (B,F,H) plane maximum intensity projection (MIP) images of elastin (Alexa Fluor 633 hydrazide-stained lung, white), Fgf18^CreER;ROSA^TDTomato (red) and Pdgfra^EGFP (green) fluorescent proteins. (B) Z-plane image showing a dense elastin matrix on the lung surface (arrow). (C) XY-plane image showing the elastin matrix 18-38 µm below the lung surface. (D) XY-plane image showing the elastin matrix in the mesothelium (top 0-5 µm). (E,F) Low-magnification images showing XY- (E) and Z- (F) plane images of Fgf18 lineage (red) and Pdgfra^EGFP expression (green). Arrows in F indicate Fgf18 lineage-labeled cells in the mesothelium. (G,H) High-magnification images showing the spatial relationship of labeled cells. AT1 cells (white arrowheads), MyoFBs (red arrows) and MatFBs (white arrows). Representative images from at least three mice are shown.

Identification of mesenchymal cell types in the neonatal lung

To identify the repertoire of mesenchymal cells in the neonatal lung, Fgf18^CreERT2/+;ROSA^TDTomato/+;Pdgfra^EGFP/+ mice were induced with tamoxifen from P2 to P5, and lungs were dissociated for cell sorting on P7 (Figs 1A and 2A). FACS identified four groups of cells (T, C1, C2 and C3) (Fig. 2A). Group T (TDTomato⁺, GFP⁻) is enriched in AT1, smooth muscle cells and mesothelial cells, as shown by increased relative levels of podoplanin (Pdpn), alpha smooth muscle actin (Acta2) and Wilms tumor 1 homolog (Wt1) expression, respectively (Fig. S1). Sorted C1 (TDTomato⁻, GFP^low), C2 (TDTomato⁻, GFP^High) and C3 (TDTomato⁺, GFP^High) cells were individually subjected to scRNA-seq (Fig. 2A,B). After filtering data, preliminary clustering and removal of a small number of contaminating hematopoietic and AT2 cells (109 from Fgf18^CreERT2/+ samples), we analyzed 1284, 3347 and 1511 cells from samples C1, C2 and C3, respectively (Table S1A). The sequence data, pooled along with a similar dataset from Fgf18 conditional knockout mice (Fgf18^CreERT2/f), was used for cluster analysis and cell type identification. All data presented here are from lineage-traced lungs that are heterozygous for Fgf18 (Fgf18^CreERT2/+).

Fig. 2.

Identification of mesenchymal cell populations in the postnatal day 7 mouse lung. (A) Whole lung from P7 Fgf18^CreER;ROSA^TDTomato;Pdgfra^EGFP mice (treated with tamoxifen from P2 to P5) were dissociated and sorted for GFP and TDTomato. Four cell types were collected: T (TDTomato⁺, GFP⁻), C1 (TDTomato⁻, GFP^Low), C2 (TDTomato⁻, GFP^High) and C3 (TDTomato⁺, GFP^High). C1, C2 and C3 cells were subjected to scRNA-seq. (B) Cell clustering identified four primary cell subtypes (MyoFB, MatFB, DivFB and AdvFB) that could be further subclustered as indicated. (C) Dotplot showing genes highly selective for expression in the four primary cell types. (D) Dotplot showing genes selective for expression in the 10 subclusters. (E) UMAP showing genes that are selective for MyoFBs and MatFBs. (F) UMAP showing genes selective for MyoFB subclusters. (G) UMAP showing genes selective for AdvFB and MatFB subclusters. (H) UMAP showing genes selective for DivFB subclusters.

The identified cell clusters clearly segregated into MyoFB, MatFB, adventitial (AdvFB) and proliferating (DivFB) cell clusters (Fig. 2B; Tables S1A, S2). These clusters could be distinguished by markers such as Myh11 and Zfp536, which were highly specific for MyoFBs; Npnt and Nebl, which were highly specific for MatFBs; and Dcn and Ebf1, which were highly specific for AdvFBs (Fig. 2C-G). Tcf21 and Meox2 mark MatFBs but are also present in AdvFBs (Fig. 2C-E). Plin2 (Adrp), a marker of lipofibroblasts, is expressed at low levels in most cells, but enriched in MatFBs (Fig. 2D; Tables S1B, S2).

Clusters enriched for proliferating cells (DivFBs) were also identified, and the majority of these cells shared markers with MyoFBs (Fig. 2C; Tables S1A, S2). Proliferating cells could be separated into S-phase, mixed S/G2M-phase and G2M-phase clusters using cell cycle scoring (Seurat) (Fig. 2B-D,H; Fig. S2, Table S1B) (Tirosh et al., 2016). Subclusters of proliferating cells could be distinguished by markers such as Ung, H1f5, Cenpa and Top2a (Fig. 2C,D,H). Analysis of proliferating cells grouped by the original cell sorting sample showed that C2 cells (TDTomato⁻, GFP^High) were enriched for DivFBs compared with C1 and C3 samples (Fig. S2).

Identification of myofibroblast subtypes

Within the MyoFB group (Myh11+, Tagln+), we could identify three types of cells that we designated as immature (MyoFB-imr), intermediate (MyoFB-int) and mature (MyoFB-mat) myofibroblasts (Fig. 2B-D,F; Tables S1B, S2). MyoFB-imr cells were identified by high expression of Zmat4, Tafa1, Lef1, Ubash3b and Piezo2, and high levels of Pdgfra and Pdgfra^EGFP (Fig. 2D,F; Fig. S3). MyoFB-mat cells were enriched for Aspn, Cdh4, Col14a1, Fgf9, Fgf18, Gja1, Hhip and Lgr6 (Fig. 2D,F; Fig. S3). MyoFB-mat cells show similar gene expression to Cdh4/Hhip/Lgr6 ductal myofibroblasts and MyoFB-imr cells appear similar to PDGFRA-GFP^High alveolar myofibroblasts identified by Narvaez Del Pilar et al. (2022) (Fig. S3A, Table S2). Additionally, we found that a subset of MyoFB-mat cells expressed high levels of Scx and Fgf9 (Fig. 2D,F; Fig. S3B). We also identified an intermediate cell population (MyoFB-int) characterized by higher levels of Prss35 and Igfbp5 (Fig. 2D,F). We posit that these cells represent a transitional state between immature and mature MyoFBs. Prss35 is a profibrotic matrisomal protein that regulates cell proliferation in response to hyperosmotic stress (Sanger et al., 2023).

Immunohistochemical staining was used to analyze the location of markers that identify immature and mature MyoFB cell types (Fig. 3; Fig. S4). MyoFB-mat cells within septal ridges could be identified that co-express FGF18(TDT), αSMA and Cx43 (Gja1) (Fig. 3A). FGF9 expression was analyzed in the lung of Fgf9^βGal mice (Huh et al., 2015) using an antibody to β-galactosidase. This showed colocalization of FGF9(βGal), Cx43 and αSMA in MyoFB-mat cells within septal ridges (Fig. 3B). Scx lineage cells, identified by immunostaining SCX(TDT) mice (Scx-Cre;ROSA^TDTomato) (Sugimoto et al., 2013) for TDTomato, showed colocalization with αSMA and Cx43 in MyoFB-mat cells (Fig. 3C). The expression of FGF9 and FGF18 in MyoFB-mat cells suggests the possibility of a protein gradient extending from the septal ridge tip to the septal base. CD34 is a transmembrane glycoprotein that is expressed in mesenchymal, muscle-derived and hematopoietic stem cells (Bose and Shenoy, 2016; Diaz-Flores et al., 2014; He et al., 2013; Sidney et al., 2014). In MyoFBs, Cd34 is expressed at high levels in MyoFB-imr cells and at low levels in MyoFB-mat cells (Fig. S3B). Consistent with expression in progenitor cells, CD34⁺, Cx43⁺, TAGLN⁺ MyoFB-imr and CD34⁻, Cx43⁺, TAGLN⁺ MyoFB-mat cells were both identified in developing alveoli (Fig. 3D). Similarly, FGF18(TDT)⁻, Cx43⁺, CD34⁺ MyoFB-imr cells were distinct from FGF18(TDT)⁺, Cx43⁺, CD34⁻ MyoFB-mat cells (Fig. 3E). Lef1 is another marker preferentially expressed in MyoFB-imr cells (Fig. 2D; Fig. S3B). Co-immunostaining for FGF9(βGal) and LEF1 identified LEF1⁺, FGF9(βGal)⁻ MyoFB-imr cells, LEF1⁺, FGF9(βGal)⁺ MyoFB-int cells and LEF1⁻, FGF9(βGal)⁺ MyoFB-mat cells (Fig. S4A). Co-immunostaining with Cx43 identified Cx43⁺, LEF1⁺ MyoFB-int cells and Cx43⁻, LEF1⁺ MyoFB-imr cells (Fig. S4B). Consistent with increased proliferation of MyoFB-imr, SCX(TDT)⁻ cells co-immunostained for LEF1 and MKI67, whereas septal tips contained SCX(TDT)⁺, MKI67⁻, LEF1⁻ MyoFB-mat cells (Fig. S4C).

Fig. 3.

Identification of immature and mature myofibroblast subtypes. (A) MyoFB-mat cells (red arrow) within the lung septal ridges are identified by co-staining for Cx43, αSMA and FGF18(TDT); MyoFB-imr cells (green arrow) only co-stain for αSMA and FGF18(TDT) but are negative for Cx43. AT1 cells (white arrowheads) are FGF18(TDT)⁺ and have AT1 cell morphology. Dashed outline indicates area shown at higher magnification on the right. (B) MyoFB-mat cells (red arrows) identified in Fgf9^βGal mice show colocalization of FGF9 (βGal), Cx43 and αSMA. (C) MyoFB-mat cells (red arrows) identified in SCX(TDT) mice show colocalization of TDT, Cx43 and αSMA. AT1 cells (white arrowheads) also contain SCX(TDT) and Cx43. Dashed outline indicates area shown at higher magnification on the right. (D) TAGLN (SM22) labels all MyoFBs. MyoFB-mat cells (red arrows) co-stain with Cx43 and do not contain CD34. MyoFB-imr cells (green arrows) are CD34⁺ and Cx43⁺. In Myo-imr cells, Cx43 expression is cytosolic and not membrane localized as in MyoFB-mat cells. Dashed outline indicates area shown at higher magnification on the right. (E) MyoFB-mat cells (red arrows) contain FGF18(TDT) and Cx43, but not CD34. MyoFB-imr cells (green arrows) are Cx43⁺, CD34⁺, but contain low levels of FGF18(TDT). Representative images from at least three mice are shown.

RNAseq data showed Pdgfra^EGFP (GFP) at higher levels in MyoFB-imr cells and lower levels in MyoFB-mat cells (Fig. S3A). The endogenous Pdgfra and Pdgfrb transcripts have a similar profile. However, FGF18(TDT) has the opposite pattern; low levels in MyoFB-imr cells and higher levels in MyoFB-mat cells (Fig. S3B). Co-immunostaining for FGF18(TDT), GFP and Cx43 identified FGF18(TDT)⁺, GFP^High, Cx43⁺ MyoFB-mat cells and FGF18(TDT)⁺, GFP^Low, Cx43⁺ MyoFB-int cells (Fig. S4D). To confirm the expression pattern of Pdgfra^EGFP (GFP), we compared its expression with endogenous PDGFRA and αSMA (Acta2). Co-immunostaining showed colocalization of both markers in compact MyoFB-imr cells and in larger MyoFB-mat cells (Fig. S4E).

GO analysis

To identify potential functions of MyoFBs, differentially expressed genes (DEGs) enriched in MyoFB-imr and MyoFB-mat clusters, as well as FGF18(TDT)-negative and -positive cells were subjected to GO classification. MyoFB-imr cells were associated with GO terms that suggest transition to the G1 phase of the cell cycle and with GO terms such as ‘cytoplasmic translation’ and ‘macromolecule biosynthetic process’, which indicate enhanced biosynthetic capacity, consistent with production of ECM proteins (Fig. S5A). The GO terms associated with MyoFB-imr cells were most similar to FGF18(TDT)⁻ MyoFBs (Fig. S5B). By contrast, MyoFB-mat cells expressed genes involved in organization of the ECM and, interestingly, are a source of secreted signaling molecules that might regulate cells in the local environment, including the regulation and patterning of angiogenesis (Fig. S5C). The GO terms associated with MyoFB-mat cells were most similar to FGF18(TDT)⁺ MyoFBs (Fig. S5D). These data suggest that FGF18(TDT) expression is an indicator of MyoFB maturation.

Identification of matrix fibroblast subtypes

Within the MatFB group, we could identify three large distinct clusters (Fig. 2B-D; Tables S1B, S2). MatFB-im,cuff cells include an immature MatFB and a population of bronchovascular cuff (BVC) adventitial-like fibroblasts (detailed below). This cluster of cells was identified by higher expression of Ifitm1 and Adh1 (Fig. 2B,D,G; Table S2). The other two MatFB clusters were most similar to previously defined alveolar fibroblast 1 (AF1) cells in adult mice, which are marked by expression of Tcf21 and Wnt2 (Sun et al., 2022) and more specifically by Bmper, Col13a1, Fat3 and Fgfr4 (Koenitzer et al., 2020). We designated these AF1 cells as MatFB (AF1-type 1) and MatFB (AF1-type 2). MatFB (AF1-type 1) cells express high levels of Wipf3 and Scube2; MatFB (AF1-type 2) cells express high levels of Wnt2, Slc27a6 and Inpp4b (Fig. 2B,D,G; Table S2).

The MatFB-im,cuff cluster is heterogeneous, as a subset of these cells specifically expresses Adh7 (Fig. 2D; Fig. S6A, Tables S1B, S2), a gene found previously to mark a unique population of cytokine-deficient AdvFB (Tsukui et al., 2020). These cells also express Dner and Twist2, but do not express Dcn, Ebf1, Ccl11 or Il33, genes expressed in classical AdvFB (Fig. 2C,D,G; Figs S6A and S7A, Table S2) (Dahlgren et al., 2019; Narvaez Del Pilar et al., 2022; Tsukui et al., 2020). Subclustering of MatFB-im,cuff cells identified three subclusters (Fig. S6B, Tables S1A,C, S2). Subcluster 0 was distinguished from subclusters 1 and 2 by expression of high levels of Nebl and Nav2 (which regulates cell migration) (Schmidt et al., 2009) and enrichment for ribosomal and ECM genes. These cells appear distinct from NEBL⁺ pathological MyoFBs seen in lungs with idiopathic pulmonary fibrosis (IPF), which also express high levels of MYO10 and RYR2 (Adams et al., 2020). Subcluster 0 cells were designated as immature synthetic cells, MatFB-0 (Synthetic) (see GO analysis section, Fig. S5E).

Subcluster 1 cells express Adh7, Col14a1, Dner, Pdgfrb and Twist2, and do not express Il33. These cells are most similar to adventitial or BVC fibroblasts identified in P7 and adult lung (Narvaez Del Pilar et al., 2022; Tsukui et al., 2020). We designate these as BVC-type 1 fibroblasts (Fig. S6, Tables S1C, S2). Subcluster 2 appears to contain stress response factors (e.g. Fos, Nr4a1 and Atf3) and is likely not a unique fibroblast population. The location of MatFB-im,cuff subclusters relative to all other clusters is shown in Fig. 5A.

Fig. 5.

Potential differentiation trajectories for lung mesenchymal cell populations determined by RNA velocity and pseudotime analysis. (A) UMAP showing a projection of subclusters of MatFB-im,cuff cells and AdvFBs. (B-E) Trajectory analysis of P7 MyoFBs. (B) UMAP projection of MyoFBs and DivFBs. (C) RNA velocity determined by scVelo indicating a trajectory from MyoFB-imr to MyoFB-mat cells with arrowheads indicating the direction of differentiation on the MyoFB UMAP projection. (D) Diffusion map for MyoFBs using the first two diffusion components with overlaid RNA velocity streams. (E) Slingshot calculated pseudotime projected on the same diffusion map with DivFBs designated as the starting cluster. (F-I) Trajectory analysis of MatFB subtypes excluding the BVC-type 1 subcluster. (F) UMAP projection of MatFB populations and putative precursors. (G) RNA velocity stream projected in UMAP space from scVelo. (H) RNA velocity projections on the MatFB diffusion map. (I) Slingshot-calculated pseudotime projected onto the MatFB diffusion map with the starting cluster designated as MatFB-0 (Synthetic). (J-M) Trajectory analysis of adventitial and BVC-type 1 fibroblasts. (J) UMAP projection of the relevant adventitial/cuff clusters. (K,L) RNA velocity depicted as arrow streams, as determined by scVelo projected onto UMAP (K) and diffusion map (L) coordinates. (M) Slingshot-calculated pseudotime projected onto the diffusion map with AdvFB-0 designated as the starting cluster.

Immunohistochemical staining was used to assess the location and colocalization of MEOX2, a transcription factor expressed in MatFBs and AdvFBs, but not in MyoFBs (Figs 2D and 4; Table S2) (Narvaez Del Pilar et al., 2022). In the distal lung, MEOX2 was expressed in cells that do not express FGF18(TDT) (Fig. 4A-C). Consistent with previous studies, MEOX2 colocalized with Pdgfra-GFP^Low expressing MatFB but not with Pdgfra-GFP and αSMA expressing MyoFBs (Fig. 4A,B). In the more proximal lung, MEOX2 was expressed in the BVC region, and not in FGF18(TDT)- and αSMA-expressing airway and vascular smooth muscle cells (Fig. 4C,D).

Fig. 4.

Identification of adventitial and matrix fibroblasts. (A) MatFBs within the lung septal ridges contain low levels of Pdgfra^EGFP (GFP) and MEOX2 (white arrows). MEOX2 shows no overlap with FGF18(TDT). MyoFBs (red arrows) contain FGF18(TDT) and higher levels of Pdgfra^EGFP (GFP), but no MEOX2. Inset shows a higher magnification of the outlined area with only the green channel showing dim GFP expression in a MEOX2⁺ cell (white arrow). (B) Outline indicates the area shown at higher magnification on the right. MEOX2 (white arrows) in MatFBs does not colocalize with αSMA or FGF18(TDT) in MyoFB-mat cells (red arrows). (C,D) Analysis of the bronchovascular cuff region in the proximal lung. (C) MEOX2⁺ cells are distributed around the airways and vessels, and do not colocalize with FGF18(TDT) and αSMA in vascular and airway smooth muscle. (D) AdvFBs and/or MatFBs in tissue surrounding the airways and vessels contain MEOX2 and low levels of Pdgfra^EGFP (GFP) (white arrows). Representative images from at least three mice are shown.

GO analysis

MatFB-im,cuff subcluster 1 (BVC-type 1) cells and AdvFB both express Dner and Twist2 suggesting that these cell types are related (Fig. S6A, Table S2). For gene ontogeny analysis of MatFBs, we therefore excluded MatFB-im,cuff subcluster 1 (BVC-type 1) cells. MatFB DEGs were determined from a subset of cells that include MatFB-im,cuff subcluster 0, MatFBs (AF1-type 1) and MatFBs (AF1-type 2). MatFB-im,cuff subcluster 0 cells were associated with GO terms that include ‘cytoplasmic translation’, ‘gene expression’ and ‘cellular respiration’, which are characteristics of an immature matrix-producing cell. We refer to these cells as MatFB-0 (Synthetic) cells (Figs S5E and S6B). MatFB (AF1-type 1) cells are associated with GO terms that include ‘positive regulation of kinase activity’, ‘small GTPase mediated signal transduction’, ‘transmembrane RTK signaling pathway’ and ‘regulation of endothelial cell migration’, suggesting that these cells integrate multiple environmental signals and function to organize the local tissue environment (Fig. S5E). MatFB (AF1-type 2) cells are associated with GO terms that include ‘cell-substrate junction’, ‘focal adhesion’ and ‘cellular response to growth factor stimulus’, suggesting that these cells interact with adjacent cells through adhesion molecules and/or junctional proteins, and integrate intercellular signals (Fig. S5E).

Identification of adventitial fibroblast subtypes

The AdvFB group expresses high levels of Dcn, Col14a1 and Cthrc1 (Fig. 2D,G; Fig. S7, Tables S1B,D, S2. Within this group, five subclusters could be identified (Table S1A,D). Cthrc1 is expressed at highest levels in clusters 0, 1 and 2, suggesting that these cells may be a reservoir of cells that could contribute to fibrotic lung disease in the adult lung (Tsukui et al., 2020). These subclusters also express Ccl11 and Il33, which have been found in cells within the BVCs in adult lung (Fig. 2D; Fig. S7A, Tables S1D, S2) (Dahlgren et al., 2019; Tsukui et al., 2020). Subcluster 0 was enriched for expression of Boc, Hhip and Lsamp, and subcluster 1 was enriched for expression of Zfp36 and Icam1. We refer to these clusters as Boc/Hhip⁺ and Icam1⁺ AdvFB, respectively. Subcluster 2 was enriched for expression of Ebf2, Il33 and Pi16, and corresponds to a subtype of adult lung BVC fibroblasts (Tsukui et al., 2020). We designate these as BVC-type 2 fibroblasts (Fig. S7, Tables S1D, S2).

Subcluster 3 (vascular smooth muscle, pericyte) expressed Acta2, Myh11 and Tagln, markers of smooth muscle, but also contained a subpopulation of cells expressing Sost and Cbr2, which are similar to heart vascular mural cells (Muhl et al., 2020), Rgs5 and Cspg4, which mark pericytes (Tsukui et al., 2020), and Thsd4 (which encodes ADAMTSL6), an ECM protein associated with elastic tissues (Mougin et al., 2021). Subcluster 4 cells expressed lung stromal markers, Col8a1, Igf1, Meis2 and Pcsk5, identified by the Tabula Muris Consortium, but also specifically expressed Grm7 and Thbs4, which were not found in the lung by the Tabula Muris Consortium (Tabula Muris Consortium et al., 2018), suggesting a potentially unique lung adventitial cell type.

GO analysis

AdvFB DEGs were determined from a subset of cells that include AdvFB subclusters 0, 1 and 2, and BVC-type 1 cells derived from MatFB-im,cuff subcluster 1 (BVC-type 1) (Fig. S5F). All of these cells were associated with GO terms that include ‘collagen containing extracellular matrix’. AdvFB subcluster 0 (Boc/Hhip⁺) cells were most associated with GO terms that include ‘extracellular matrix organization’, ‘regulation of cell migration’ and ‘regulation of cell adhesion’. AdvFB subcluster 1 (Icam1⁺) cells were associated with GO terms that include ‘focal adhesion’ and ‘regulation of apoptotic process’, ‘regulation of cell population proliferation’ and ‘regulation of cell migration’. AdvFB subcluster 2 (BVC-type 2) cells may additionally regulate the extracellular environment, including angiogenesis. MatFB-im,cuff subcluster 1 (BVC-type 1) cells have GO terms that include ‘focal adhesion’, ‘extracellular matrix organization’ and ‘positive regulation cell population proliferation’.

Trajectory analysis predicts differentiation of distinct neonatal myofibroblast and matrix fibroblast progenitors

To evaluate potential lineage relationships within neonatal fibroblast populations, we employed multiple trajectory inference approaches. For RNA velocity (La Manno et al., 2018), ratios of spliced to unspliced transcripts for each gene are used to extrapolate temporal relationships among cells. In MyoFBs, the inferred trajectory supported a differentiation pathway from DivFBs to MyoFB-imr and ultimately to MyoFB-mat cells, as projected in UMAP space (Fig. 5A-C). Non-linear dimensional reduction by diffusion components results in diffusion maps (DMs) that often recapitulate biological differentiation pathways in scRNA-seq data (Haghverdi et al., 2015). Diffusion mapping on a subset of cells including all MyoFB populations also suggested a linear trajectory from MyoFB-imr to MyoFB-mat cells, further supported by projection of RNA velocity on DM coordinates. Finally, pseudotime analysis was performed using Slingshot (Street et al., 2018) with DivFBs defined as the starting cluster. This approach also identified MyoFB-mat cells as a differentiation endpoint.

For MatFBs, a similar analysis was performed on the AF1-type 1, AF1-type 2, Synthetic and Synthetic/Stress populations. Proposed trajectories according to all methods showed progression from Synthetic cells with AF1-type 1 and AF1-type 2 as separate endpoints (Fig. 5A,F-I). By diffusion map, Synthetic/Stress cells largely form a third trajectory away from the Synthetic cluster (Fig. 5H,I), suggesting that these cells are not an intermediate in the development of mature alveolar fibroblast populations. By our classification, AdvFB included BVC-type 1 (MatFB-1) and BVC-type 2 (AdvFB-2) cells along with AdvFB-0 and AdvFB-1 subclusters. AdvFB-3 and AdvFB-4 were considered unlikely to share a lineage relationship with these subclusters and were thus excluded from this analysis. Trajectory analyses of these cell types as a subset suggest that AdvFB-1 cells are a point of origin that could give rise to the other subtypes, and that BVC-type 1 and BVC-type 2 cells represent distinct lineages (Fig. 5L-M). No differentiation pathway linking MyoFB and MatFB was noted, suggesting that, at this stage of development, MyoFBs and MatFBs make up distinct lineages.

Intercellular communication

To evaluate intercellular signaling among the full complement of cell types in the developing alveolus, including our identified fibroblast subpopulations, we merged and integrated our annotated expression matrix with an existing lung scRNA-seq dataset from P7 mouse lung (Fig. 6A) (Zepp et al., 2021a,b) and performed ligand-receptor analysis with CellChat. MyoFBs are recipients of an array of extracellular signals, including SHH, PDGF, TGFβ, WNT and NOTCH (Fig. 6B-F), and also act as a source of secreted signaling molecules, including FGF18 and WNT5a (Fig. 6F,G). Differential signaling was noted among subpopulations of MatFBs and MyoFBs. Among MyoFBs, FGF9 production is limited to MyoFB-mat cells and FGF18 to MyoFB-int and MyoFB-mat cells (Fig. 6G). FGF9 and FGF18 could signal to FGFR3 and FGFR4 in MatFBs or to FGFRs 1-4 in most mesenchymal cell types (Fig. 6G and not shown). Among MatFBs, few interactions distinguished AF1-type 1 and AF1-type 2 cells (Fig. 6C). Signals derived from endothelial cells (aCAP), such as PDGFB, may regulate PDGFRB-expressing MyoFBs and AdvFBs (Fig. 6E).

Fig. 6.

Selected potential intercellular communication pathways between lung mesenchymal cells and other lung cell types identified using CellChat. (A) UMAP projection of combined Zepp et al. dataset and sorted mesenchymal cells after merge and reclustering, as performed in preparation for CellChat analysis. (B-G) CellChat chord diagrams for selected signaling pathways. Terminal arrow width corresponds to relative signal strength. (B) Sonic hedgehog (SHH) signaling via patched 1 (PTCH1) and patched 2 (PTCH2). (C) TGFB1 and TGFB2 signaling. (D) Notch signaling. (E) PDGF signaling. (F) WNT and non-canonical WNT signaling. (G) FGF signaling. The FGF10-FGFR2 diagram is curated to include only epithelial signaling.

Functional properties of mesenchymal cell types in the neonatal lung

Functional properties of mesenchymal cell populations marked by Fgf18^CreERT2;ROSA^TDTomato and Pdgfra^EGFP were examined in in vitro cultures of sorted C1, C2 and C3 cells from mice induced with tamoxifen from P2 to P5 and harvested on P7 (Figs 2A and 7A). Consistent with the cell-sorting parameters, after 1 day of in vitro culture (DIV1), C1 and C2 cells expressed GFP but not TDTomato, whereas C3 cells expressed both GFP and TDTomato (Fig. 7B). Violin plots of GFP expression from scRNA-seq data showed GFP in all cell samples, with highest levels in C2 (Fig. 7C). Consistent with sorting and DIV1 culture, TDTomato was highly expressed in C3 with only trace amounts detected in C1 and C2 (Fig. 7D).

Fig. 7.

Gene expression patterns of sorted populations of lung mesenchymal cells. (A) Experimental scheme for lineage labeling of mesenchymal cell populations in Fgf18^CreER;ROSA^TDTomato;Pdgfra^EGFP mice. Tamoxifen injection at P2-P5. P7 lung tissue was then dissociated, sorted for GFP and TDTomato, and cultured. (B) Cultures of C1 (TDTomato⁻, GFP^low), C2 (TDTomato⁻, GFP^High) and C3 (TDTomato⁺, GFP^High) cells after 1 day in vitro culture (DIV1) were imaged [brightfield, Pdgfra^EGFP (GFP) and FGF18(TDT)]. (C,D) Violin plots grouped by sample, showing (C) highest GFP expression in C2 cells and lower expression in C1 and C3 cells, and (D) highest TDTomato expression in C3 and trace expression in C1 and C2 cells. (E) UMAP showing the cluster distribution of samples C1, C2 and C3 (see Table S1A for composition). (F) Heatmap showing genes enriched in samples C1, C2 and C3. (G) Violin plots showing expression of genes enriched in samples C1, C2 and C3. (H,I) Violin plots comparing the MyoFB expression of immature markers Pdgfra, Cd34 and Lef1 (H), and mature markers Gja1, Fgf9 and Scx (I) in C2 cells, TDTomato-negative cells (TDT⁻), TDTomato-positive cells (TDT⁺) and C3 cells.

Gene signatures that define the sorted C1, C2 and C3 cells are consistent with C1 containing MatFBs, a small subcluster of AdvFB (vascular smooth muscle) and a very small number of DivFB (Fig. 7E; Fig. S2). C1 cells were highly enriched for markers such as Meox2, Tcf21 and Nebl (Fig. 7F,G). Sample C2 cells were represented by all MyoFB clusters and were the predominant contributor to DivFB and AdvFB clusters (Fig. 7E; Fig. S2). C2 cells were enriched for MyoFB-imr markers, such as Vsnl1 and Piezo2 (Figs 2D and 7F,G). C3 cells contain primarily MyoFBs and a small proportion of DivFBs (Fig. 7E; Fig. S2). C3 cells were enriched for markers expressed in MyoFB-mat cells, such as Lgr6, Cdh4 and Eln (Fig. 7F,G). C2 and C3 were more related to each other and are distinguished from C1 by the expression of common markers, such as the transcription factor Zfp536, and MyoFB genes such at Tagln, Myh11 and Itga8. Although C2 and C3 cells show significant similarities, they are distinguished by expression of FGF18(TDT) and other genes such as Piezo2 (high in C2) and Col12a1 (high in C3) (Fig. 7D,F).

Further comparison of MyoFBs in C2 and C3 cells show that genes associated with less mature cells, such as Pdgfra, Cd34 and Lef1, were expressed at higher levels in C2 cells with a profile that was most similar to that of TDTomato⁻ cells (Fig. 7H; Table S1E). By contrast, C3 cells expressed these genes at lower levels with a profile that was most similar to TDTomato⁺ cells (Fig. 7H; Table S1E). Similarly, genes associated with more mature cells, such as Gja1, Fgf9 and Scx, were expressed at lower levels in C2 cells with a profile that was most similar to that of TDTomato⁻ cells. By contrast, C3 cells expressed these genes at relatively higher levels, with a profile that was most similar to TDTomato⁺ cells (Fig. 7I). Overlaying TDTomato expression onto the MyoFB UMAP shows relatively low expression in MyoFB-imr cells and higher expression in MyoFB-int and MyoFB-mat cells (Fig. S8).

Trajectory, diffusion and velocity analysis predicted that MyoFB-imr transition to MyoFB-int and finally to MyoFB-mat cells (Fig. 5A-E). To explore this prediction, C2 cells were maintained in culture and monitored for GFP and TDTomato expression (Fig. 8A,B). Although negative for TDTomato expression on DIV1, TDTomato expression became visible by DIV2 and increased by DIV4. Sorting C2 DIV4 cells identified TDTomato⁻,GFP^Low (C1*); TDTomato⁻,GFP^High (C2*) and TDTomato⁺,GFP^High (C3*) cells (Fig. 8C,D).

Fig. 8.

Spontaneous maturation of C2 cells in vitro is marked by expression of the endogenous and targeted ROSA26 locus. (A) Experimental schematic. Sorted C2 cells (TDTomato⁻, GFP^High) were placed in culture at P7. (B) Images of C2 cells at DIV1, DIV2 and DIV4. (C-F) Quantification of C2 cell maturation in the absence (C,D) or presence (E,F) of 4-OHT. (C,E) At DIV4, C2 cell cultures were sorted for GFP- and TDTomato-expressing cells, identifying three populations of cells: C1* (GFP^Low), C2* (GFP^High) and C3*(TDT⁺, GFP^High). (D,F) Quantification of the percentage of C1*, C2* and C3* cells at DIV0 (gray bars) and DIV4 (open bars) (D, n=3; F, n=4). (G) Violin plots showing expression of the endogenous ROSA26 long non-coding RNA [Gt(ROSA)26Sor], Fgf18 and transgenic Fgf18^CreERT2 (CreERT2) in C1, C2 and C3 cells. (H,I) Experimental schematic for sorting and culturing GFP^Low and GFP^High (C2+C3) cells from P7 Fgf18^CreERT2;ROSA^TDTomato;Pdgfra^EGFP lungs from mice that were not induced with tamoxifen. (J) C2 cells (GFP^High) cultured in the absence of 4-OHT until DIV5 do not express TDTomato. Adding 4-OHT to the culture medium on DIV5 results in abundant TDTomato⁺ cells on DIV6. A two-way ANOVA with Fisher's LSD test was used to test for differences; n=3 or 4; ns, not significant; *P<0.05, **P<0.001, ***P<0.0001.

Interestingly, the activation of TDTomato expression (derived from the ROSA^TDTomato reporter) in culture occurred in the absence of tamoxifen, suggesting that expression of the ROSA26 gene locus was itself regulated and reports maturation of immature MyoFBs. To control for this, we cultured C2 cells in the presence of 4-hydroxytamoxifen (4-OHT). However, even in the presence of 4-OHT, similar numbers of C1*, C2* and C3* cells were present at DIV4 (Fig. 8E,F), demonstrating that additional recombination at the ROSA26 locus is not occurring during culture.

Consistent with regulation of the ROSA26 locus, expression of the endogenous ROSA26 long non-coding RNA [Gt(ROSA)26Sor] in the scRNA-seq data was present at low levels in C1 and C2, and at high levels in C3 cells (Fig. 8G), similar to the expression of TDTomato in C1, C2 and C3 cells (Fig. 7D). Furthermore, expression of endogenous Fgf18 and transgenic Fgf18^CreERT2 was detected in C2 cells (Fig. 8G). The expression of CreERT2 in C2 cells suggests that in mice induced with tamoxifen, ROSA^TDTomato recombination had occurred but the ROSA26 allele and TDTomato were expressed at undetectable levels. To control for any leaky CreERT2 activity, GFP^High cells, which should contain the equivalent of both C2 and C3 cells, were sorted from Fgf18^CreERT2;ROSA^TDTomato;Pdgfra^EGFP mice that were not induced with tamoxifen (Fig. 8H,I). Sorted GFP^High cells, harvested at P7 and cultured for 5 days (Fig. 8J) continued to express GFP but did not express TDTomato. However, treatment of DIV5 cultures with 4-OHT for 24 h robustly activated TDTomato expression. Therefore, these cells must express Fgf18^CreERT2, CreER was not leaky and could be activated in vitro, and the ROSA26 locus was active in these cells after 5 days in culture.

C1 cells (TDTomato⁻,GFP^low) are primarily composed of MatFB and AdvFB (Fig. 7E; Table S1A). Trajectory, diffusion and velocity analysis shows potential differentiation pathways within these two populations (Fig. 5A,F-I); however, there were no trajectories linking MatFBs/AdvFBs and MyoFBs. This is consistent with 3H-thymidine uptake studies demonstrating that lipid-containing interstitial cells (corresponding to MatFB) were distinct from non-lipid containing interstitial cells (corresponding to MyoFB) (Brody and Kaplan, 1983). To determine whether C1 cells have the capacity to generate TDTomato⁺ MyoFB, sorted C1 cells were maintained in culture for 4 days. At DIV4, C1 cells were larger and more spread out, the intensity of GFP fluorescence was decreased, but no TDTomato⁺ cells were observed (Fig. 9A). Cell sorting confirmed the stability of the C1 population cultured either without or with 4-OHT (Fig. 9B-E). The small percentage of TDTomato⁻,GFP^High C2* cells (11±7%) and TDTomato⁺,GFP^High C3* cells (5±5%) in 4OHT-treated cultures likely arise from proliferation of contaminating MyoFB progenitors (GFP^High) within the sorted C1 cells, as the sorting gates are adjacent to each other, but we cannot rule out differentiation of a GFP^Low MatFB or AdvFB into a GFP^High MyoFB. It remains possible that C1 MatFBs could differentiate into MyoFBs in response to growth factors expressed in epithelial or endothelial cells (e.g. FGF18, PDGFA, PDGFB, SHH and TGFB1) (Fig. 6), which are not present in our in vitro culture conditions. In adult lung, TGFB1 and PDGFA have been shown to promote MatFB differentiation to MyoFB-like cells (Green et al., 2016). However, lipofibroblasts lineage traced with Tcf21-mCrem do not label αSMA⁺ cells in adult lung (Park et al., 2019).

Fig. 9.

MatFBs and TDTomato-expressing MyoFBs remain stable in in vitro culture. (A) Images of C1 cells at DIV1 and DIV4. (B-E) Quantification of C1 cell maturation in the absence (B,C) or presence (D,E) of 4-OHT. (B,D) At DIV4, C1 cell cultures were sorted for GFP and TDTomato. Of the GFP-expressing cells, most retained the GFP^Low (C1*) phenotype. (C,E) Quantification of the percentage of C1*, C2* and C3* cells at DIV0 (grey bars) and DIV4 (open bars) (C, n=3; E, n=4). (F) Images of C3 cultures at DIV1 and DIV4. (G-J) Quantification of C3 cell maturation in the absence (G,H) or presence (I,J) of 4-OHT. (G,I) At DIV4, C3 cell cultures were sorted for GFP and TDTomato. Of the GFP-expressing cells, most retained the TDTomato⁺, GFP^High (C3*) phenotype. (H,J) Quantification of the percentage of C1*, C2* and C3* cells at DIV0 (gray bars) and DIV4 (open bars) (H, n=3; J, n=4). Two-way ANOVA with Fisher's LSD test; n=4; ns, not significant; *P<0.05, **P<0.001.

C3 cells (TDTomato⁺,GFP^High) are composed almost exclusively of MyoFBs and proliferating cells with a MyoFB gene signature (Fig. 7E; Fig. S2). After 4 days in culture, C3 cells increased in size and retained TDTomato expression (Fig. 9F). Cell sorting confirmed the stability of the C3 population (cultured with or without 4-OHT) showing that, within GFP-expressing cells, over 96% also expressed TDTomato (C3*) (Fig. 9G-J).

DISCUSSION

Alveologenesis is a unique stage of lung development during which MyoFBs, an essential cell that guides the formation of secondary septa, is a transient cell population that is present only during the primary phase of alveologenesis (Branchfield et al., 2016; Duong et al., 2022; Gao et al., 2022; Hagan et al., 2020; Li et al., 2015; McGowan et al., 2008; Narvaez Del Pilar et al., 2022; Riccetti et al., 2020; Yamada et al., 2005; Zepp et al., 2021a). In the mouse, MyoFBs are first observed at ∼P2-P3, at the initiation of alveologenesis, and most can no longer be detected by the end of primary alveologenesis, ∼P14-P15. By contrast, AdvFBs and MatFBs (including lipofibroblasts) are present during alveologenesis and persist in the adult lung. Although embryonic mesenchymal cells that give rise to lung mesenchymal lineages have been identified (Li et al., 2015, 2018; Moiseenko et al., 2017; Yamada et al., 2005), the progenitor cell that allows expansion of the MyoFB population during primary alveologenesis is not defined.

Myofibroblast lineage

The MyoFB has previously been identified and isolated based on its expression of high levels of GFP driven by the Pdgfra gene promoter (Endale et al., 2017; Kimani et al., 2009; McGowan and McCoy, 2014; Zepp et al., 2021a). In the Pdgfra-GFP^High MyoFBs, the expression of Fgf18 is greatly increased during the primary phase of alveologenesis in rodents and humans (Boucherat et al., 2007; Chailley-Heu et al., 2005; Franco-Montoya et al., 2011; McGowan and McCoy, 2015; Ruiz-Camp and Morty, 2015). This expression pattern was confirmed by lineage labeling with the FGF18(TDT) (Fgf18^CreER;ROSA^TDTomato) lineage reporter (Hagan et al., 2019, 2020).

In an effort to further characterize lung mesenchymal cell populations during alveologenesis, cells from dissociated P7 lung tissue were sorted based on expression of GFP (Pdgfra^EGFP) and TDTomato (Fgf18^CreER;ROSA^TDTomato). This configuration allowed separation of the Pdgfra-GFP^High cell population into two groups: TDTomato⁻,GFP^High (C2 cells) and TDTomato⁺,GFP^High (C3 cells). Single-cell sequencing revealed that both of these cell populations contained MyoFBs and that the TDTomato⁻ cells also contained AdvFBs. Interestingly, GFP^High,TDTomato⁻ cells were enriched for proliferating cells with a MyoFB gene signature, and the GFP^High,TDTomato⁺ cells were enriched for markers of mature MyoFBs. In culture, some of the GFP^High,TDTomato⁻ cells were able to produce GFP^High,TDTomato⁺ cells, suggesting that the initial C2 cell population contains a proliferating progenitor that gives rise to MyoFB-mat cells. Activation of TDTomato expression during this maturation process occurred in the absence of tamoxifen and thus indicates that upregulation of ROSA^TDTomato gene expression is not induced but is a regulated event that marks MyoFB maturation. Consistent with this, single-cell sequencing data shows both increased TDTomato and increased endogenous ROSA26 long non-coding RNA expression in C3 compared with C2 MyoFB.

Two types of MyoFB were identified in human postnatal day 1 lung (Duong et al., 2022). MyoFB_1 was thought to be a progenitor based on higher expression of PDGFRA, IGF1 and MEF2A, and MyoFB_2 was thought to represent a more mature population with elevated expression of ACTA2, MYH11 and MYOCD. MyoFB-imr cells have higher levels of Pdgfra and Mef2a, but lower levels of Igf1. MyoFB-mat cells have elevated Myocd, but no difference in Acta2 or Myh11 expression. These differences may reflect differences between mouse and human or differences in the relative developmental stage.

A unique feature of the MyoFB-mat cells is the expression of Gja1, Fgf9 and Scx. Immunostaining showed that these proteins or their lineage reporters are present in MyoFB in septal tips and alveolar ducts. The function of these genes in the neonatal lung is not known; however, in other tissues, both Gja1 and Scx can maintain a myofibroblast phenotype or promote a fibroblast to myofibroblast transition (Bagchi et al., 2016; Paw et al., 2017). The MyoFB-mat cells also express periostin (Postn), a direct transcriptional target of SCX (Nagalingam et al., 2022). The function of FGF9 in the neonatal lung is not known, but in the adult, it may inhibit human IPF fibroblast differentiation to myofibroblasts and promote a migratory phenotype (Joannes et al., 2016). We posit that, during alveologenesis, FGF9 and FGF18 could function to regulate migration or maturation of MyoFB-imr progenitor cells, which, relative to MyoFB-mat cells, express higher levels of Fgfr1 and Fgfr2. Unlike FGF18, FGF9 can also signal to lung epithelial cells via FGFR3 (Yin et al., 2013, 2016; Yin and Ornitz, 2020), and thus could coordinate MyoFB-mat-epithelial cell interactions.

MyoFB-imr cells likely contain a progenitor population that gives rise to MyoFB-int and MyoFB-mat cells (Fig. 10). Consistent with this model, MyoFB-imr cells are enriched for Cd34, Eng (endoglin, CD105), Lef1 (lymphoid enhancer factor 1), Stc1 (stanniocalcin 1), and Plcb1 (phospholipase C, beta 1), genes associated with a progenitor phenotype in multiple tissues and cell lineages (Hennrick et al., 2007; Merrill et al., 2001; Schelker et al., 2021; Schmidt-Ott and Barasch, 2008; Sidney et al., 2014; Song and Kim, 2022). In adult lung and kidney fibrosis, lineage tracing showed that pericytes could give rise to pathological myofibroblasts (Armulik et al., 2011; Barron et al., 2016; Humphreys et al., 2010; Hung et al., 2013); however, in the lung this remains controversial and one lineage tracing study showed that pericytes are excluded as a major source of pathological myofibroblasts (Rock et al., 2011). Interestingly, Pdgfrb, an accepted marker for pericytes, is enriched in MyoFB-imr cells compared with MyoFB-int and MyoFB-mat cells; however, MyoFB-imr cells express low levels of other pericyte markers, Cspg4 (Ng2) and Foxj1, suggesting that they are not closely related to pericytes.

Fig. 10.

A model of a secondary septa. Most proliferating mesenchymal cell progenitors feed into a MyoFB maturation program that results in mature MyoFB near the apex of the septal ridge. Several types of MatFB are generally located towards the base of secondary septa.

The mechanisms that regulate the disappearance of MyoFBs after primary alveologenesis are still elusive. In one study, in the absence of epithelial Wnt secretion, αSMA⁺ cells persisted within alveolar septa (Fang et al., 2022); however, this was attributed to increased endothelial to mesenchymal transition rather than to a failure to lose pre-existing MyoFBs. MyoFB-imr cells and, at lower levels, MyoFB-mat cells express the canonical WNT receptors LRP6 and FZD1. By contrast, expression of the non-canonical WNT ligand Wnt5a is increased in MyoFB-int and MyoFB-mat cells. WNT5a functions to suppress canonical WNT/β-catenin signaling and to induce caspase activity in differentiating embryonic stem cells (Bisson et al., 2015). Vsnl1 (visinin-like 1), which encodes a WNT/β-catenin-regulated calcium-sensing protein, is highly expressed in MyoFB-imr and MyoFB-int cells, and at low levels in MyoFB-mat cells. In colorectal cancer, VSNL1 functions to suppress apoptosis (Tage et al., 2023). By analogy, its downregulation in MyoFB-mat cells could contribute to their eventual loss. Future studies will be needed to determine whether WNT signaling directly regulates MyoFB maturation and whether loss of WNT signaling triggers MyoFB apoptosis at the end of primary alveologenesis.

Matrix and adventitial fibroblasts

Although we focus on the MyoFB lineage in this study, our scRNA-seq data revealed unique subclusters of MatFBs and AdvFBs, and cell culture shows that MatFBs (the predominant cell type in C1 cells) are stable and do not spontaneously convert to TDTomato⁺ cells. Bronchovascular bundles are conduits for conducting airways and associated vasculature, and contain adventitial (bronchovascular cuff) fibroblasts (Dahlgren et al., 2019; Narvaez Del Pilar et al., 2022; Sun et al., 2022; Travaglini et al., 2020; Tsukui et al., 2020). Subclustering of AdvFB and MatFB-im,cuff cells identified two subclusters that are related to bronchovascular cuff fibroblasts. We have termed these cells as Bronchovascular Cuff FB type 1 and type 2 (BVC-type 1, BVC-type 2). Both of these clusters express common markers, such as Dner, Itgbl1, Mecom, Serpinf1 and Twist2. These clusters are distinguished by expression of Adh7, Lgr5 and Wif1 in BVC-type 1 cells and expression of Col14a1 and Pi16 in BVC-type 2 cells. This expression pattern suggests that BVC-type 1 cells may have progenitor-like properties, as Lgr5 expression is a marker of stem cells and the WNT inhibitor WIF1 can regulate progenitor cell proliferation (Barker et al., 2007; Yu et al., 2022). Interestingly, Lee et al. have identified an adult Lgr5⁺ lung mesenchymal cell population that localizes to the alveolar region and promotes alveolar differentiation via activation of WNT signaling (Lee et al., 2017). BVC-type 2 cells, which express Col14a1, may have a greater role in producing ECM components. Col14a1⁺ cells are found in perivascular regions of the developing lung (Mižiková et al., 2022).

Meox2 is expressed in MatFBs and AdvFBs, and is excluded from smooth muscle. Consistent with published data (Narvaez Del Pilar et al., 2022), our immunostaining shows MEOX2⁺ cells localized to BVC regions adjacent to peribronchial and perivascular smooth muscle, and colocalizing with cells that express low levels of Pdgfra^EGFP. The expression pattern of Meox2 is similar to that of Tcf21, suggesting that these genes could be co-regulated.

Limitations

These studies have only analyzed mesenchymal cells that express Pdgfra^EGFP and lineage-traced Fgf18^CreERT2 cells at a single time point (P7) near the peak of alveologenesis. However, as alveologenesis occurs dynamically over time and likely progresses from proximal to distal as the lung grows, any time point should contain all the required cellular components. In this study, we sequenced only one sample for each sorted population. However, each sample was composed of pooled cells from three mice to minimize variability among animals. In addition to Pdgfra^EGFP cells, there are likely other mesenchymal cells present in the neonatal lung that do not express Pdgfra^EGFP. The identity of these Pdgfra^EGFP-negative cells will require high-resolution single-cell analysis of the whole lung or other lineage markers that can be used for cell sorting and sequencing. All cells analyzed in this study are heterozygous for Fgf18 and Pdgfra. Although no overt phenotypes have been associated with haploinsufficiency of these genes, given their importance for mesenchymal cell differentiation, we cannot rule out subtle effects on relative cluster size and quantitative levels of gene expression. Our in vitro cultures were carried out in 0.5×Matrigel, which is rich in growth factors. Although C1 cells appear stable under our culture conditions, it is possible that endothelial- or epithelial-derived growth factors that are not present in Matrigel could influence the differentiation or transdifferentiation of these cells. Finally, in Table S2 we have tried to cross reference the names given to the multitude of mesenchymal cell types that have been identified by many investigators both in the neonate and adult. We apologize if we have inadvertently missed some names given to mesenchymal cell types.

MATERIALS AND METHODS

Animals

All mice (Mus musculus) were housed in a pathogen-free barrier facility. All studies were performed under a protocol approved by the Institutional Animal Care at Washington University in Saint Louis (Approval No. 20190110). Mice of both sexes were used. Mice were maintained on a mixed FVB, C57BL/6J, 129X1/SvJ genetic background. Mouse strains used were: Fgf18^CreERT2 [Fgf18^{tm2.2(cre/ERT2)Dor}] (Hagan et al., 2019), ROSA^TDTomato [Gt(ROSA)26Sor^{tm9(CAG−TDTomato)Hze}/J] (Madisen et al., 2010), Pdgfra^EGFP [Pdgfra^{tm11(EGFP)Sor/J}] (Hamilton et al., 2003), Fgf9^βGal (Huh et al., 2015) and Scx^Cre (Sugimoto et al., 2013). Mice heterozygous for Fgf18^CreERT2/+, ROSA^TDTomato/+ and Pdgfra^EGFP/+ were used for lineage studies and fluorescence-activated cell sorting (FACS). The day of birth was assigned as P0.

Tamoxifen administration

Tamoxifen was administered to neonates by intraperitoneal injection at a dose of 150 µg/mouse/day on 4 sequential days beginning at P2. Tamoxifen (Sigma, T5648) was prepared at a concentration of 15 µg/µl dissolved in corn oil or sunflower seed oil.

Tissue preparation for CLARITY and immunohistochemistry

At designated collection times, mice were sacrificed with an overdose of a cocktail containing ketamine and xylazine, and perfused with PBS through the right ventricle. For immunohistochemistry, the lungs were fixed via intratracheal inflation with 4% paraformaldehyde (PFA) (Electron Microscopy Sciences, 15714-S) at a pressure of 20 cm H₂O. Lungs were post-fixed overnight at 4°C with gentle agitation. For immunohistochemistry, samples were washed three times in PBS for 20 min, cut into lobes, dehydrated in ethanol and xylene, embedded in paraffin wax, sectioned (5 μm) and stained with Hematoxylin and Eosin (H&E) or used for immunostaining.

For CLARITY, the lungs were fixed in 0.5 ml of 4% PFA by intratracheal injection with a 26 G needle. Lungs were post-fixed overnight at 4°C in 4% PFA with gentle agitation and then washed 5 times with PBS for 20 min at room temperature.

Antibodies and chemical reagents

The following chemical reagent and antibodies were used for immunofluorescence: Alexa Fluor 633 hydrazide (Elastin) (A30634, Invitrogen), goat anti-TDTomato (1:200, MBS448092, MyBioSource.com), rabbit anti-connexin 43 (Cx43) (1:200, SC-9059, Santa Cruz Biotechnology), mouse anti-smooth muscle actin (αSMA) (1:300, Clone 1A4, Dako), goat anti-TAGLN (1:200, ab10135, Abcam), chicken anti-β-galactosidase (1:200, ab9361, Abcam), FITC-conjugated mouse anti-CD34 (1:200, 11-0341-82, Invitrogen), mouse anti-LEF1 (1:100, SC-374522, Santa Cruz Biotechnology), rabbit anti-PDGFRA (1:100, sc-338, Santa Cruz Biotechnology), chicken anti-GFP (1:200, AB-2307313, Aveslab), rabbit anti-MEOX2 (1:200, NBP2-30647, Novus Biological), Alexa-488, 555, 594 or Alexa-647-conjugated donkey anti-goat, anti-rabbit, anti-mouse, anti-chicken secondary antibodies (all from Invitrogen, A11055, A21432, A21447, A21206, A21207, A31573, A21202, A31570, A21203, A31571 and A78952) and Alexa-488 or Alexa 594-conjugated donkey anti-chicken secondary antibodies (703-545-155 and 703-585-155 from Jackson ImmunoResearch).

Tissue clearing (CLARITY)

Samples were optically cleared using the SHIELD tissue preservation approach (Park et al., 2018) and subsequently delipidated using ETC active CLARITY. Briefly, 4% PFA-fixed tissues were immersed in SHIELD-off polyepoxy solution for 3 days at 4°C and subsequently polymerized by immersion in SHIELD-on solution at 37°C for 24 h. Preserved samples were delipidated in a 4% sodium dodecyl sulfate buffer for 24 h at 45°C, transferred to a SmartBatch+ electrophoretic tissue clearing (ETC) platform (LifeCanvas Technologies) and cleared for 30 h at 35V and 1250 mA at 42°C. After clearing, samples were washed in PBS for 24 h. Lungs were washed three times with PBS/0.2% Triton X-100 for 1 h at 37°C and stained for elastin using Alexa Fluor 633 hydrazide (Shen et al., 2012) (1:1000, Thermo Fisher Scientific, A30634) in PBS/0.2% Triton X-100 for 24 h at 37°C. Alexa Fluor 633 hydrazide selectively stains elastin. After staining, the lungs were washed four times with PBS/0.2% Triton X-100 for 2 h at 37°C, and then stored in PBS/0.2% Triton X-100 at 4°C. Before imaging, the refractive index of lungs were matched to 1.52 using Easy Index (LifeCanvas Technologies).

Confocal imaging and image analysis

High-resolution imaging datasets of cleared lungs were acquired on a Zeiss LSM 980 microscope using the Airyscan 2 SR-4Y mode with either a 40×/1.2 water immersion or a 63×/1.4 oil immersion objective. Raw Airyscan files were processed using Huygens Professional (Scientific Volume Imaging) and visualized in Imaris (Oxford Instruments). Thin section confocal images were acquired on a Zeiss LSM 980 microscope with a 40×/1.2 water immersion or a 63×/1.4 oil immersion objective and visualized in Imaris.

Flow cytometry

Lung cells were isolated as described previously (Hagan et al., 2020) with minor modifications. Fgf18^CreERT2/+;ROSA^TDTomato/+;Pdgfra^EGFP/+ pups were injected with tamoxifen (150 µg) on P2, P3, P4 and P5, and then anesthetized and sacrificed on P7 with an overdose of a cocktail containing ketamine and xylazine. The lungs were dissected and placed in PBS on ice, minced with a razor blade with 50 µl of dispase (5000 Caseinolytic Units, Discovery Labware, 354235) and then transferred to PBS containing 833 units/ml dispase, 0.01 mg/ml DNase I (Roche, 10104159001) and 2 mg/ml collagenase/dispase (Roche, 10269638) at 37°C for 40 min with gentle agitation. Digestion enzymes were then inactivated with FACS buffer (10% FBS and 1 mM EDTA in PBS), samples were sequentially filtered through a 70 µm cell strainer (Thermo Fisher Scientific, 22363548) and a 40 µm cell strainer (Thermo Fisher Scientific, 22363547), and then centrifuged at 800× g for 4 min. Red blood cells were lysed with ACK lysis buffer (Gibco, A10492-01) and cells were resuspended in FACS buffer. SYTOX blue (1 µl, Thermo Fisher Scientific, S34857) was added to each tube for live or dead cell sorting. Cells were sorted using a SY3200 ‘Synergy’ cell sorter (Sony Biotechnology) or a Beckman Coulter MoFlo cell sorter with a 100 µm nozzle directly into FACS buffer or culture media (1.5 ml) on ice. Four cell populations were collected: TDTomato⁺, GFP⁻; TDTomato⁺, GFP^High; TDTomato⁻, GFP^High; and TDTomato⁻, GFP^Low. Single channel controls (Fgf18^CreERT2;ROSA^TDTomato;Pdgfra^EGFP mice) were used to set compensation at the time of sorting and a fluorescence minus one control (ROSA^TDTomato lacking Fgf18^CreERT2) was used as a negative control. Plotted percentage of the four populations of cells was calculated relative to the total number of sorted cells. Freshly sorted cells were either placed into culture or submitted to the Washington University Genome Technology Access Center for scRNA-seq.

Cell culture

Sorted cells were resuspended in DMEM/F12 (Gibco, 11330-032) with 20% fetal bovine serum (Gibco, 26140-079) and 2% penicillin/streptomycin, and plated on Matrigel (1:2 dilution)-coated 24- or 48-well plates and incubated at 37°C in 5% CO₂. Media were changed daily during the course of the experiment. Cells were imaged using a Leica DM IL LED inverted microscope.

scRNA-seq

For single-cell sequencing, C1, C2 and C3 cells (Fig. 2A) were sorted from three individual mice. Each of the three samples (C1, C2 or C3) was pooled from the three mice and was used for droplet-based scRNA-seq using 10x Genomics 3′v3 kits at the Washington University Genome Technology Access Center (GTAC). Flow-sorted populations were loaded on individual lanes in a Chromium Controller as per the manufacturer's instructions. After their preparation according to the 10x Genomics protocol, each library was sequenced using an Illumina NovaSeq 6000 instrument. After demultiplexing of Illumina runs, FASTQ files were processed by the 10x Genomics Cell Ranger pipeline (version 7.0.0) with exonic and intronic reads incorporated in the final molecule counts.

Cell clustering and annotation

For initial clustering, cells from Fgf18 heterozygous (Fgf18^CreERT2/+;ROSA^TDTomato/+;Pdgfra^EGFP/+) P7 mice were clustered with cells obtained from Fgf18 conditional knockout (Fgf18^CreERT2/f;ROSA^TDTomato/+;Pdgfra^EGFP/+) P7 mice obtained with an identical flow-sorting approach. Expression matrices were input into Seurat for clustering analysis. Low quality cells were filtered by feature counts (>2300 genes/cell and <6500 genes/cell) and mitochondrial gene content (<10%). After merge, batch effect correction was performed with the Harmony package (Korsunsky et al., 2019). After Louvain clustering and UMAP visualization (using Harmony embeddings), cell type-specific markers were identified with Seurat's FindAllMarkers function, with log fold change set to 0.25. Annotation for broad cell types (myofibroblasts, matrix and adventitial fibroblasts, and dividing cells) was performed using known markers, and specific marker genes were identified for subtypes within these categories. Rare cells identified as undesired cell types (epithelium, endothelium and immune) based on known marker genes were removed after clustering. For all downstream analyses, a subset of Fgf18 heterozygous cells from the combined object was used.

Trajectory analysis of mesenchymal cells

For trajectory analysis, spliced and unspliced mRNAs for all genes in all cells were determined using Velocyto (https://github.com/velocyto-team/velocyto.py) using a cell-sorted BAM file for each sample as input. The resulting loom files were input to scVelo, merged in an anndata object and velocities were estimated using scv.tl.velocity. For visualization, dimensional reduction was performed using the standard scanpy (https://scanpy.readthedocs.io;v1.9.1) pipeline with PCA and UMAP, and cluster assignments annotated based on the Seurat analysis above. To calculate diffusion maps, we built the k-nearest neighbor graph (k=25) using the first 30 principal components in scanpy and used the scanpy.tl.diffmap function to build diffusion maps with an exponential kernel (method=‘umap’). For projection of velocities onto diffusion maps, DM coordinates were added to adata.obs in the appropriate subsetted adata objects. The R Bioconductor package slingshot (https://bioconductor.org/packages/release/bioc/html/slingshot.html) was used to calculate the lineage structure and pseudotime, which were projected onto the same diffusion maps.

CellChat analysis

For receptor-ligand analysis, a publicly available P7 mouse lung scRNA-seq filtered expression matrix was used (sample GSM4504963) (Zepp et al., 2021a,b). In Seurat, SCTransform normalization was applied to both our dataset (GSE261508) and GSM4504963. After initial clustering and annotation with Seurat, mesenchymal cell populations were removed from GSM4504963 data and then ‘merge’ was used to combined our annotated fibroblast populations (with dividing populations removed). PrepSCTFindmarkers was then used to create a normalized count matrix of values corrected for sequencing depth. The combined object was then re-clustered using the Seurat pipeline as above, and inputted into the CellChat package in R using default settings (Jin et al., 2021). For filterCommunication, min.cells was set to 10.

Reverse transcriptase quantitative PCR

RNA was isolated from sorted TDTomato⁺ cells according to manufacturer's instructions using the Arcturus PicoPure RNA Isolation Kit (Applied Biosystems, Thermo Fisher Scientific, 12204-01). P7 whole lung and adult heart RNA, isolated with Trizol Reagent (Ambion by Life Technologies, 15596026) served as positive and negative controls, respectively, for gene expression. cDNA was generated by reverse transcription of up to 9 ng RNA from each sample using iScript Reverse Transcription Supermix for RT-qPCR (BioRad, 1708841). mRNA expression was determined on a StepOnePlus Real-Time PCR System (Thermo Fisher, 4376600) using TaqMan Fast Advanced Master Mix (Thermo Fisher, 4444557) and TaqMan assay probes (Gapdh, Thermo Fisher, Mm99999915_g1;Pdpn, Thermo Fisher, Mm01348912_g1;Wt1, Thermo Fisher, Mm00460570_m1). Relative gene expression was calculated based on the average cycle (Ct) value of technical triplicates, normalized to Gapdh control, and reported as fold change (2^−ΔΔCT).

Statistical analysis

Data were analyzed with Prism software using analysis of variance (ANOVA). For quantification of gene expression by qPCR, three or four individual samples were measured and analyzed using one-way ANOVA with Tukey's multiple comparisons test. For quantification of sorted cell populations, three or four independent cultures derived from different animals were analyzed using two-way ANOVA with Fisher's least significant difference (LSD) test for multiple comparisons. The data are reported as means±s.d., and changes with P<0.05 were considered to be statistically significant.

Supplementary Material

Table S1. Genomic data.

A. Numbers of cells analyzed by single cell RNA sequencing. B. Differentially expressed genes (DEGs) in cell clusters shown in Figure 2B, D. C. DEGs in reclustered MatFB-im,cuff cells. D. DEGs in reclustered AdvFB cells. E. DEGs in FGF18(TDT) positive and negative MyoFB

Peer review history

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202659.reviewer-comments.pdf