Fgf10-positive cells represent a progenitor cell population during lung development and postnatally

The adult mouse lung consists of at least 40-60 different types of cells that are organized in a highly sophisticated 3D structure inside the thoracic cavity (Crapo et al., 1982; McQualter and Bertoncello, 2012). These cell types can be broadly classified into epithelial cells (populating the airways) and mesenchymal cells (populating the surrounding extracellular matrix). Along the proximodistal axis, the mouse lung is rich with stem/progenitor cells that are capable of self-renewal, multipotent differentiation and pollutant resistance.

Whereas the lineage tree of the lung epithelium has been thoroughly studied during embryonic development and postnatally (Giangreco et al., 2002; Hong et al., 2004; Kim et al., 2005; Rawlins et al., 2009; Rock et al., 2009; Stripp et al., 1995), the lineage tree of the postnatal lung mesenchyme is poorly understood. This part of the lung consists of a wide array of cell types such as smooth muscle cells, endothelial cells, chondrocytes, nerve cells, lipofibroblasts, myofibroblasts, lymphatic cells and others.

Most of our knowledge regarding the lineage tree of the lung mesenchyme has been achieved by studying embryonic development. For example, our research group has previously used a fibroblast growth factor 10 (Fgf10)-lacZ transgenic line (originally reported as Mlcv1v-nLacZ-24) to demonstrate that Fgf10-expressing cells are progenitors for parabronchial smooth muscle cells (PBSMCs) in the distal lung during early embryonic development (Mailleux et al., 2005). Shan et al., on the other hand, identified a population of mesenchymal cells, initially located near the lung hilum, as progenitors for airway smooth muscle cells (Shan et al., 2008). Using Wilms tumor 1 homolog (Wt1)-Cre transgenic line, Que et al. have suggested that the mesothelium is a source of vascular - but not airway - smooth muscle cells (Que et al., 2008). On the contrary, Greif et al. have shown that vascular smooth muscle cells (VSMCs) arise from Pdgfrb-positive mesenchymal cells rather than Wt1-positive mesothelial cells (Greif et al., 2012). More recently, Dixit et al. utilized Wt1^Cre-ERT2/+ mice to demonstrate that Wt1-positive cells are indeed progenitors for airway and VSMCs as well as desmin⁺ fibroblasts (Dixit et al., 2013). Fetal liver kinase 1 (Flk-1; Kdr - Mouse Genome Informatics), another term for vascular endothelial growth factor receptor 2 (Vegfr2), has been identified as the earliest marker for endothelial progenitors (angioblasts) that respond to Vegfa and give rise to the capillary plexus surrounding the airways (Del Moral et al., 2006; Kappel et al., 1999; Yamaguchi et al., 1993). Using a prospero homeobox 1 (Prox1)^Cre-ERT2 knock-in line, Srinivasan et al. have shown that lympthatic cells arise from Prox1-positive progenitor cells (Srinivasan et al., 2007). Finally, Langsdorf et al. used receptor tyrosine kinase (Ret)^EGFP knock-in line to transiently trace neuronal progenitors during embryonic development and show that nerve cells originate from the neural crest (Langsdorf et al., 2011).

One of the early markers of the lung mesenchyme is Fgf10. During early lung development, Fgf10 is expressed in the distal (submesothelial) mesenchyme and it acts on the opposite epithelium expressing Fgfr2b to maintain the epithelial cells in a progenitor-like state and induce branching and migration (Bellusci et al., 1997; Lü et al., 2005; Park et al., 1998). Gain of function of Fgf10 during development leads to epithelial progenitor state arrest and distalization of the lung (Nyeng et al., 2008; Volckaert et al., 2013), whereas loss of function of Fgf10 results in branching simplification and decreased numbers of epithelial progenitors. Although the primary target of Fgf10 is the epithelium, severe mesenchymal abnormalities are also observed when Fgf10 signaling is attenuated (Ramasamy et al., 2007).

We have previously generated a novel Fgf10^Cre-ERT2/+ (or Fgf10^iCre/+) knock-in mouse line that allows specific labeling of Fgf10-expressing cells, which can be lineage-traced temporally and spatially (El Agha et al., 2012). Using this line, combined with tomato^flox Cre-reporter line, we performed extensive lineage tracing of Fgf10-positive cells during lung development and postnatal homeostasis. We used a combination of time-lapse imaging, immunofluorescence, fluorescence-activated cell sorting (FACS) analysis and sorting, analysis of cytospin-prepared cells and reverse transcription polymerase chain reaction (RT-PCR) on isolated cells to characterize the progeny of Fgf10-expressing cells labeled at distinct time points throughout lung development and postnatally. Our data indicate that Fgf10-positive cells represent a pool of progenitors for multiple lineages in the embryonic and postnatal lung mesenchyme.

We have previously shown that the Fgf10^iCre/+; tomato^flox/+lineage-tracing tool allows permanent labeling of Fgf10-positive cells (as well as their descendants) and thus opening the way to study their fate in a time-controlled manner (El Agha et al., 2012). To further validate this line, Fgf10^iCre/+; tomato^flox/flox mice were crossed with the previously established Fgf10-lacZ reporter line (Kelly et al., 2001; Mailleux et al., 2005) and pregnant mice received a single intraperitoneal (IP) injection of tamoxifen at embryonic day (E) 11.5 and embryonic lungs were harvested at E13.5 (supplementary material Fig. S1). We reasoned that lineage-labeled cells from E11.5 are likely to retain Fgf10 expression at E13.5. Double immunostaining for red fluorescent protein (RFP; reporting for Fgf10-expressing cells at E11.5 as well as their descendants) and lacZ (reporting for total Fgf10-expressing cells at E13.5) showed that all lineage-labeled cells were indeed lacZ positive, indicating the absence of ectopic expression from the Fgf10^iCre locus and demonstrating that Fgf10-expressing cells are specifically targeted with this tool. Quantification revealed that 65.44±4.27% (n=3) of total lacZ-positive cells were also positive for the lineage label (RFP), indicating that this model allows the targeting of a subset of Fgf10-expressing cells. Like any other lineage-tracing tool, the level of recombination greatly depends on the dose of tamoxifen used. For this study, a tolerable dose of 0.1 mg/g of body weight was used.

Throughout the study, Fgf10^+/+; tomato^flox/+ littermates from tamoxifen-injected mice and/or Fgf10^iCre/+; tomato^flox/+ offspring from corn-oil-injected mice were used as controls and no recombination was observed in these offspring at the level of whole-mount fluorescence imaging, fluorescence microscopy of lung sections and FACS analysis, confirming the absence of leakiness in the Fgf10^iCre/+; tomato^flox/+ tool.

In order to study the behavior of Fgf10-positive cells during the pseudoglandular stage of lung development, pregnant mice carrying Fgf10^iCre/+; tomato^flox/+ embryos received a single IP injection of tamoxifen at E11.5 and embryos were harvested at E12.5. Lungs were cultured in an air-liquid interphase and brightfield/fluorescence time-lapse imaging was carried out for 72 hours (Fig. 1; supplementary material Movie 1; n=3). Lineage-labeled cells were originally detected as small populations in the distal mesenchyme and were mostly abundant in the accessory lobe (Fig. 1B-B′). During the first 24 hours of culture, tomato-positive cells amplified exponentially and populated other regions of the growing lung (Fig. 1D-D′,J). At the end of the three-day culture, lineage-labeled cells were highly abundant and dispersed throughout the lung mesenchyme (Fig. 1E-E′). After ex vivo culture, lungs were fixed, processed and stained for αSma (Acta2 - Mouse Genome Informatics) (Fig. 1F-I). An average of 81±0.2 (n=3) tomato-positive cells was detected per section (Fig. 1G,G′,K) and a subpopulation of Fgf10-positive cells (5.67±0.57 cells; n=3) appeared in the PBSMC layer (Fig. 1I,I′,K).

To study the commitment of Fgf10-positive cells to the smooth muscle lineage in vivo, pregnant mice carrying Fgf10^iCre/+; tomato^flox/+ embryos received a single IP injection of tamoxifen at E10.5 and embryonic lungs were harvested at E13.5, E15.5 and E18.5 (Fig. 2A-C). Because tomato-positive cells were mostly abundant in the accessory lobe, this lobe was used for immunostaining. αSma staining of Fgf10^iCre/+; tomato^flox/+ lungs revealed a subpopulation of tomato-positive cells in the PBSMC layer at all three developmental stages (Fig. 2G,G′,K,K′,O,O′). Among total tomato-positive cells, tomato-positive PBSMCs were significantly abundant at E13.5 (14.03±1.31% of total tomato-positive cells; n=3) (Fig. 2A) and E15.5 (18.47±3.75%; n=3) (Fig. 2B) but to a less extent at E18.5 (7.51±0.72%; n=3) (Fig. 2C). Another population of tomato-positive cells was identified in the VSMC compartment at E15.5 (3.44±0.96%; n=3) (Fig. 2B; supplementary material Fig. S2) and E18.5 (4.88±0.83%; n=3) (Fig. 2C; supplementary material Fig. S2). A third population of lineage-labeled cells was observed in the vicinity of parabronchial and VSMCs at all three stages. Initially, we thought that these cells could have a neuronal nature, as Fgf10 is known to be involved in neurogenesis (Haan et al., 2013; Hajihosseini et al., 2008). However, β-III Tubulin (tubulin, beta 3 class III - Mouse Genome Informatics) immunostaining did not show any overlap with the lineage label (data not shown).

The genetic model used in this study revealed that only a minor proportion of Fgf10-positive cells commits to the smooth muscle lineage (Fig. 2). Fgf10 is known to be expressed by adipocyte precursors and Fgf10-null neonates suffer from impaired development of the white adipose tissue (Sakaue et al., 2002; Yamasaki et al., 1999). Thus, we decided to investigate whether Fgf10-positive cells give rise to an adipose-related cell lineage in the lung, the lipofibroblasts. Fgf10-positive cells were labeled during early pseudoglandular stage (E11.5) and lungs were harvested at E12.5, E15.5 and E18.5 (Fig. 3A-C). FACS analysis showed a drastic increase in the total number of RFP⁺ cells from E12.5 to E15.5 and then E18.5 (Fig. 3D). Our unpublished data show that lipofibroblasts start to emerge in the lung at E15.5; thus, αSma and Adrp (Plin2 - Mouse Genome Informatics) immunostaining was performed at E18.5. For convenience, the left lung lobe was used to carry out immunofluorescence staining and quantification (Fig. 3F). An average of 450±10.12 (n=3) tomato-positive cells was counted per sample and a minor proportion of RFP-positive cells stained for αSma around the epithelium (5.68±0.39%; n=3) (Fig. 3E,G) and within vascular walls (4.31±0.23%; n=3) (Fig. 3E,H). Similarly to Fgf10-positive cells labeled at E10.5, a population of αSma^- RFP⁺ cells was identified in the vicinity of PBSMCs (10.17±0.88%; n=3) (Fig. 3E,I) and VSMCs (3.92±0.72%; n=3) (Fig. 3E). These cells did not express the neuronal marker β-III tubulin (supplementary material Fig. S3). However, a significant population of RFP-positive cells stained for Adrp (29.96±5.17%; n=3) (Fig. 3E,J) and the remaining population of RFP-positive cells (45.97±4.2%; n=3) that did not stain for αSma or Adrp was characterized by its big size and the presence of filopodia (Fig. 3E,K). The lungs were also stained for other markers, such as Cd45 (Ptprc - Mouse Genome Informatics) and Cd31 (Pecam1 - Mouse Genome Informatics), but no overlap was observed between these markers and the RFP signal (supplementary material Fig. S3).

Fgf10-positive cells from E11.5 showed to be precursor cells for PBSMCs, VSMCs and lipofibroblasts at E18.5 (Fig. 3). However, the commitment of these early progenitors to the smooth muscle lineage showed a decline beyond E15.5 (Fig. 2). To investigate this observation further and to test whether lipofibroblast progenitors are restricted to early developmental stages, Fgf10-positive cells were labeled at E15.5 and embryonic lungs were harvested at E16.5 and E18.5 (Fig. 4). RFP expression was low at E16.5 (Fig. 4A) and was more pronounced at E18.5 (Fig. 4B). The left lung lobe was used to carry out immunofluorescence staining and quantification (Fig. 4D). An average of 499±71.04 (n=3) tomato-positive cells was counted per sample at E18.5 and αSma staining showed minimal overlap with the RFP signal in the parabronchial (0.26±0.03%; n=3) and vascular (0.75±0.34%; n=3) compartments (Fig. 4E). Significant numbers of αSma^- RFP⁺ cells around PBSMCs (10.33±1.65%; n=3) (Fig. 4E,F) and VSMCs (4.99±0.32%; n=3) (Fig. 4E,G) were detected and Adrp staining revealed a considerable proportion of Adrp⁺ RFP⁺ cells (39.98±8.28%; n=3) (Fig. 4E,H,I). Similarly to cells labeled at E11.5, αSma^-Adrp^- β-III Tub^- Cd45^- Cd31^- cells (43.69±8.32%; n=3) were also observed (Fig. 4E,J,K; supplementary material Fig. S3).

The alveolar stage of lung development is known for the prevalence of alveolar myofibroblasts. These cells, in addition to lipofibroblasts, are believed to play a crucial role in secondary septa formation and alveolar maturation during postnatal alveolarization (O’Hare and Sheridan, 1970; Rehan and Torday, 2012; Rubin et al., 2004; Torday and Rehan, 2007; Vaccaro and Brody, 1978; Yamada et al., 2005). To determine whether Fgf10-positive cells contribute to these lineages postnatally, pregnant mice carrying Fgf10^iCre/+; tomato^flox/+ embryos were left to develop to term and then fed tamoxifen-containing pellets starting on postnatal day 2 (P2). At P14, pups were sacrificed, lungs were harvested, and the left lung lobe was used for immunostaining (Fig. 5). No tomato signal was detected in lungs from Fgf10^+/+; tomato^flox/+ pups (Fig. 5A,L) whereas the signal was abundant in their Fgf10^iCre/+; tomato^flox/+ counterparts (Fig. 5B), especially near the main bronchus (Fig. 5C). An average of 141±40 (n=3) tomato-positive cells was counted per Fgf10^iCre/+; tomato^flox/+ sample and αSma immunostaining revealed a subpopulation of lineage-labeled cells expressing αSma (19.70±1.88%; n=3) (Fig. 5D-H,M). These cells were located in the lung parenchyma and not at the tips of secondary septa. On the other hand, Adrp immunostaining showed that a major proportion of Fgf10-expressing cells gives rise to lipofibroblasts (67.74±1.27%; n=3) (Fig. 5I-K,M).

Fgf10-expressing cells were labeled at E11.5 or E15.5 and lungs were harvested at E18.5. Lungs were dissociated into single-cell suspensions and stained for Cd45, Cd31, Epcam, Pdgfra, Kit and Sca-1 (Ly6a - Mouse Genome Informatics). A FACSAria III cell sorter was used to analyze 50,000-100,000 cells per sample per time point. No tomato signal was detected in lungs from Fgf10^+/+; tomato^flox/+ embryos (Fig. 6A) or from Fgf10^iCre/+; tomato^flox/+ embryos exposed to corn oil instead of tamoxifen (data not shown). Whereas lineage-labeled cells from E11.5 accounted for ∼0.5% of the total lung, those from E15.5 accounted for ∼1.5% (Fig. 6B,C) and most of these cells were Cd45^- Cd31^- (data not shown). Almost all lineage-labeled cells from both time points expressed low levels of Sca-1 (Sca-1^low) at E18.5 (Fig. 6E,G) and a subpopulation of Fgf10-positive cells from E11.5 and E15.5 also expressed the mesenchymal marker Pdgfra at E18.5 (∼9.9% and ∼17.5%, respectively) (Fig. 6F,H). Interestingly, almost no tomato-positive cells from E11.5 and ∼18.1% from E15.5 stained for the pan-epithelial cell marker Epcam at E18.5 (Fig. 6E,G). Kit expression was very low at this stage (E18.5) but staining revealed ∼7.3% and ∼1.4% overlap with the tomato signal in E11.5 and E15.5 time points, respectively (Fig. 6F,H). RT-PCR and cytospin analysis of sorted Fgf10-positive cells (labeled at E15.5) showed high expression of Adrp (Fig. 6D), confirming the immunofluorescence data obtained from processed lung tissues (Figs 3 and 4). Histological analysis of E15.5 Fgf10^iCre/+; tomato^flox/+ lungs (labeled at E11.5) showed a subpopulation of RFP⁺ Pdgfra⁺ cells (Fig. 6I,J).

Fgf10-expressing cells, labeled postnatally, represented ∼0.1% of the total lung (a total of one million cells was analyzed) (Fig. 6L). ∼15.7% and ∼22.8% of total RFP-positive cells stained for Epcam and Sca-1, respectively (Fig. 6M). A subpopulation (accounting for ∼10.5% of total RFP-positive cells) exhibited the molecular signature of resident MSCs (Cd45^- Cd31^- Sca-1⁺) (Fig. 6N).

In a separate experiment, pregnant mice carrying Fgf10^iCre/+; tomato^flox/+ embryos received a single IP injection of tamoxifen or corn oil at E11.5 and were left to develop to term. The pups from these mice were sacrificed at P30 and lungs were harvested for FACS analysis. Lungs from corn-oil-exposed pups revealed no tomato expression (Fig. 6S), whereas lineage-labeled cells from tamoxifen-exposed lungs appeared in the distal mesenchyme, especially at the tip of the accessory lobe (Fig. 6T,U). These cells accounted for ∼0.4% of total lung suspensions (Fig. 6O) and similarly to the cells labeled at E11.5 and harvested at E18.5, minimal overlap between tomato and Epcam was detected (Fig. 6P). Nevertheless, ∼11.1% of total lineage-labeled cells expressed Sca-1, half of which representing a subset of resident MSCs (Fig. 6P,Q). LipidTOX™ staining revealed that most tomato-positive cells (∼83.7%) contained neutral lipids at P30 (Fig. 6R).

In order to further characterize Epcam⁺ RFP⁺ cells, Fgf10-expressing cells were labeled at E15.5 and RFP⁺ cells were sorted at E18.5 and stained for Pro-Spc (Sftpc - Mouse Genome Informatics) and E-Cadherin (Cdh1 - Mouse Genome Informatics) (Fig. 7A,B). Interestingly, a subpopulation of RFP⁺ cells was Pro-Spc⁺ but E-Cad^- (Fig. 7A′,B′). A similar observation was made for lung sections where RFP⁺ cells could not be detected in the lung epithelium (E-Cad^-) (Fig. 7D,D′,F) and 20.23±0.89% (n=3) of total RFP⁺ cells was Pro-Spc⁺ (Fig. 7C,C′,F). Finally, RT-PCR on sorted cells (labeled at P2 and collected at P30) showed that Epcam⁺ RFP⁺ cells were positive for Spc, but negative for Nkx2.1 and E-Cad, whereas Epcam⁺ RFP^- cells were positive for Nkx2.1, E-Cad and Spc. Both populations showed lack of Fgf10 expression (Fig. 7E).

The embryonic lung mesenchyme consists of cells that belong to distinct lineages, including myogenic, adipogenic, chondrogenic, neuronal and other lineages. Most of these cell types arise from endogenous (or resident) progenitor populations whereas few (e.g. nerve cells) arise from exogenous sources (the neural crest). Cell-fate determination probably arises from a complex network of autocrine and paracrine signals leading to the proliferation and differentiation of lineage-restricted as well as multipotent progenitor cell populations.

Fgf10 is an early marker for the submesothelial mesenchyme (Bellusci et al., 1997) and its expression levels have shown to be crucial for the maintenance of both epithelial and mesenchymal progenitors (Ramasamy et al., 2007). The domain of Fgf10 expression in the embryonic lung has been thoroughly studied. Northern blotting and quantitative PCR (qPCR) data show that Fgf10 transcripts progressively accumulate in the embryonic lung between E11.5 and E18.5 (Bellusci et al., 1997; El Agha et al., 2012). In situ hybridization for Fgf10 and X-Gal staining of Fgf10-lacZ lungs show that Fgf10 expression is restricted to the tips of the distal mesenchyme until E14.5, after which the expression becomes dispersed throughout the mesenchyme (El Agha et al., 2012; Mailleux et al., 2005; Ramasamy et al., 2007; Tiozzo et al., 2012). Our lineage-tracing data reveal two waves of Fgf10 expression, each of which is characterized by a distinct pattern in terms of cell localization as well as differentiation potential (Fig. 8A). The first wave, consisting of Fgf10-positive cells labeled at E11.5 as well as their descendants, yields cells residing mostly in the distal mesenchyme at E18.5 (especially observed in the accessory lobe) and contributes to myogenic (PBSMCs and VSMCs) as well as adipogenic (lipofibroblasts) lineages. By contrast, the second wave, consisting of Fgf10-positive cells labeled at E15.5 as well as their descendants, yields cells that are dispersed throughout the mesenchyme at E18.5 and gives rise to adipogenic, but not myogenic, lineages.

Lineage tracing of Fgf10-positive cells from E10.5 showed that these cells contribute to the smooth muscle cell lineage when examined at E13.5, E15.5 and E18.5; however, the percentage of αSma⁺ RFP⁺ cells relative to total RFP⁺ cells showed a decline between E15.5 and E18.5. As the overall number of Fgf10-positive cells (labeled at E11.5) showed a drastic increase from E12.5 to E15.5 and then E18.5 as demonstrated by FACS analysis (also demonstrated by time-lapse imaging of E12.5 lung undergoing branching morphogenesis), we conclude that the drop in the number of αSma⁺ RFP⁺ cells relative to total RFP⁺ cells at E18.5 is likely to be due to the expansion of the αSma^- RFP⁺ population at the expense of the αSma⁺ RFP⁺ population. In other words, the limited αSma⁺ RFP⁺ fraction that forms between E10.5 and E15.5 gets diluted in the αSma^- RFP⁺ population, which continues to expand beyond E15.5. This is supported by our observation that αSma⁺ cells rarely proliferate when examined at E13.5, E15.5 and E18.5 (supplementary material Fig. S4). The proliferative capacity of Fgf10-positive cells was also demonstrated by time-lapse imaging of E18.5 lung explants that were induced with tamoxifen at E11.5 (supplementary material Movie 2) or E15.5 (data not shown). Moreover, primary cultures of E18.5 lung mesenchyme (induced with tamoxifen at E11.5 or E15.5) showed that ∼40% of RFP⁺ cells were undergoing mitosis during the 24 hours of culture (supplementary material Movie 3).

It was recently reported that a multipotent cardiopulmonary progenitor cell population (Wnt2⁺ Gli1⁺ Isl1⁺), originating from the second heart field (SHF), invades the lung, giving rise to airway and VSMCs in addition to other lineages (Peng et al., 2013). Our lineage-tracing system does not label any SHF cells as lineage-labeled cells could not be detected in the heart at any given time point, probably owing to the deletion of intronic sequences containing transcription factor binding sites for Isl1, Nkx2.5, Tbx5 and others (El Agha et al., 2012). Yet the possibility that a subset of Fgf10-positive cells arises from Isl1-positive cells coming from the SHF cannot be excluded.

FACS analysis revealed that the first wave of Fgf10 expression accounts for ∼0.5% of the total lung at E18.5, whereas the second wave accounts for ∼1.5% of the total E18.5 lung. This indicates that during embryonic lung development, not all Fgf10-expressing cells derive from Fgf10-positive mother cells; rather, the Fgf10-expressing domain expands by de novo induction of Fgf10 expression (Fig. 8B). However, clonal expansion of Fgf10-positive cells needs to be further examined using confetti or rainbow Cre-reporter mice in order to assess the degree of heterogeneity of these cells and to determine whether they possess unipotent or multipotent capabilities. Flow cytometry analysis also revealed that a subpopulation of early and late lineage-labeled cells acquires Pdgfra expression at later developmental stages. Surprisingly, a subpopulation of late Fgf10-positive progenitors (second wave) also acquires Epcam expression. Epcam⁺ RFP⁺ cells were further analyzed following FACS and they showed lack of Nkx2.1 and E-Cad expression, indicating that these cells do not represent a bona fide epithelial cell population. Yet they express significant levels of Pro-Spc, as shown by immunofluorescence (of lung sections and sorted cells) and RT-PCR, and appear to have lost Fgf10 expression between P2 and P30. To date, there is no evidence for mesenchymal-to-epithelial transition events during lung development, and our previous work with the Dermo1-Cre; R26R line suggests that mesenchymal cells do not integrate into the lung epithelium (De Langhe et al., 2008; Tiozzo et al., 2012). Similarly, we show here that the Fgf10-expressing lineage is purely mesenchymal by nature and labeled cells were never detected in the lung epithelium. Whether these cells could ‘completely’ differentiate into bona fide epithelial cells under specific conditions (acute injury, for example) remains to be established.

Alveolar myofibroblasts and lipofibroblasts are two cell lineages that are highly abundant during postnatal alveolarization. These cell types play important roles in the formation of functional alveoli in newborns. In this study, the contribution of Fgf10-positive cells to these two lineages was addressed. When labeled during the alveolar stage of lung development, most Fgf10-expressing cells gave rise to lipofibroblasts that are believed to represent a stem cell niche that signals to adjacent type 2 alveolar epithelial cells in a paracrine fashion (Barkauskas et al., 2013). Furthermore, a minor population of Fgf10-expressing cells gave rise to myofibroblasts residing in the parenchyma rather than in secondary septae. Alveolar myofibroblasts are believed to derive from Pdgfra-positive cells as Pdgfa-null newborns suffer from arrest in alveologenesis because of the absence of these cells (Boström et al., 1996; Lindahl et al., 1997). More recently, McGowan et al. have shown that Pdgfra expression directly correlated with αSma expression and inversely correlated with neutral lipid production (McGowan et al., 2008), indicating that Pdgfra-positive cells are progenitors for alveolar myofibroblasts rather than lipofibroblasts. Our FACS analysis shows that only ∼10% of Fgf10-expressing cells co-express Pdgfra postnatally (data not shown), which could explain our observation that Fgf10-expressing cells preferentially give rise to lipofibroblasts rather than alveolar myofibroblasts (Pdgfra⁺). Lipofibroblasts originate from multiple progenitor pools and the Fgf10-expressing pool contributes to a subset of lipofibroblasts in the lung (D.A.A., E.E.A. and S.B., unpublished). Although it has been suggested that lipofibroblasts serve as a source of alveolar myofibroblasts during alveologenesis (Perl and Gale, 2009), our data suggest that at least Fgf10-expressing lipofibroblasts do not undergo this transdifferentiation. However, as Fgfr2b ligands have shown to be required for the formation of alveolar myofibroblasts during homeostasis (Perl and Gale, 2009; Ramasamy et al., 2007) as well as regeneration after pneumonectomy (Chen et al., 2012), we speculate that paracrine signaling between Fgf10-positive cells and Pdgfra-positive progenitors is likely to be necessary for the formation of the alveolar myofibroblast lineage during alveologenesis.

Fgf10-positive cells were also examined during lung homeostasis in the adult mouse. Despite the low efficiency of the Fgf10^iCre line in labeling Fgf10-expressing cells postnatally (El Agha et al., 2012), FACS analysis of disaggregated adult lungs revealed that these cells constitute ∼0.1% of the whole lung. A subpopulation of Fgf10-expressing cells (∼10.5% of total Fgf10-expressing cells) exhibits the molecular signature of resident MSCs (Cd45^- Cd31^- Sca-1⁺) that were previously described by McQualter et al. (McQualter et al., 2009). In the latter study, the authors reported that the Cd45^- Cd31^- Sca-1⁺ fraction of the adult lung represents a mesenchymal population of resident progenitor cells that are also crucial for the growth of epithelial progenitors in vitro. Our data support the findings of McQualter et al. and provide direct evidence that Fgf10 expression identifies a subset of resident MSCs.

Finally, Fgf10-positive cells, labeled at E11.5, were retained after birth (P30), indicating that these cells have self-renewal capabilities. Nevertheless, the majority of these cells stained for the neutral lipid dye, indicating their engagement into the lipofibroblast phenotype.

In summary, the reported genetic model of lineage tracing shows that Fgf10-expressing cells represent a population of mesenchymal progenitors in the lung in vivo. These cells give rise to airway and VSMCs as well as lipofibroblasts during embryonic development. Neonatally, these cells contribute to the formation of lipofibroblasts and during adult life, they represent a subset of resident MSCs that might play a crucial role in regeneration after lung injury. Our observation that some descendants of Fgf10-expressing cells express Pro-Spc shows how little is known about lineage hierarchy in the lung mesenchyme. More studies need to be performed in order to better characterize the various mesenchymal lineages in the lung, by combining lineage-tracing tools with responder lines allowing genetic manipulation of key signaling pathways (e.g. FGF, Wnt, Hedgehog), and eventually implementing the attained knowledge in novel therapeutic strategies for treating interstitial lung diseases.

Fgf10^iCre knock-in mice were generated as previously described (El Agha et al., 2012) and tomato^flox reporter mice (B6;129S6-Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J) were purchased from Jackson Lab (stock number 007905). Fgf10-lacZ hypomorph/reporter mice were described previously (Kelly et al., 2001; Mailleux et al., 2005; Ramasamy et al., 2007). Mice were housed in a specific pathogen-free environment and E0.5 was assigned to the day when a vaginal plug was detected. Harvesting organs and tissues from wild-type and transgenic mice was approved by Justus Liebig University Giessen (approval numbers 405_M and 437_M, respectively) and animal experiments were approved by the Regierungspraesidium Giessen (GI 20/10 Nr. 38/2011 and 117/2012) and the Animal Research Committee at Childrens Hospital Los Angeles (CHLA; Protocol number 31-11) in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institute of Health. The approval identification for CHLA is AAALAC A3276-01.

Tamoxifen solution was prepared as previously described (El Agha et al., 2012) and pregnant females carrying Fgf10^iCre/+; tomato^flox/+ embryos received a single IP injection of 0.1 mg tamoxifen (Sigma, Schnelldorf, Germany) per gram of body weight. For continuous tamoxifen exposure, mice were fed tamoxifen-containing pellets (0.4 g of tamoxifen per kg of food) (Altromin, Lage, Germany).

Lungs were harvested in PBS and examined using Leica M205 FA fluorescent stereoscope (Leica Microsystems, Wetzlar, Germany) equipped with Leica DFC360 FX camera. The endogenous tomato signal was detected using the RFP channel. For time-lapse microscopy, intact E12.5 lungs or E18.5 lung explants were harvested in Dulbecco’s modified Eagle medium (DMEM) and cultured on Whatman Nuclepore membrane filters (GE Healthcare, Solingen, Germany) placed on 500 μl of DMEM with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Culture dishes with embryonic tissues were then transferred into the culture chamber (37°C; 5% CO₂) of Leica DMI6000 B live imaging microscope (Leica Microsystems) connected to Leica DFC365 FX camera where imaging was performed. RFP signal intensity was measured using ImageJ software (NIH).

Freshly isolated lungs were washed in PBS and then fixed in 4% paraformaldehyde (PFA) in PBS according to standard procedures. Then, tissues were embedded in paraffin and sectioned at 5 μm thickness. Slides were deparaffinized, blocked with 3% bovine serum albumin (BSA) and 0.4% Triton X-100 [in Tris-buffered saline (TBS)] at room temperature (RT) for 1 hour and then incubated with primary antibodies against αSma (Sigma; 1:100), Adipose differentiation-related protein (Adrp) (Abcam; 1:50), Pro-Spc (Seven Hills; 1:1000), E-Cadherin (BD; 1:200), β-III Tubulin (R&D; 1:50), Cd45 or Cd31 (Biolegend; 1:100) at RT for 1 hour or at 4°C overnight. After incubation with primary antibodies, slides were washed three times in TBST (TBS buffer + 0.1% Tween 20) for 5 minutes, incubated with secondary antibodies at RT for 1 hour and then washed three times in TBST before being mounted with Prolong Gold Anti-fade Reagent with DAPI (4′,6-diamidino-2-phenylindole; Invitrogen). Fluorescent images were acquired using Leica DM5500 B fluorescence microscope connected to Leica DFC360 FX camera and the endogenous tomato signal was detected using the RFP channel. Colocalization data were further confirmed by z stacks of multiple optical sections using Leica TCS SP5 X confocal microscope.

Lungs were isolated, washed with Hanks’ balanced salt solution (HBSS) and kept on ice. Then, they were cut into fine pieces using a sharp blade before being digested with 0.5% Collagenase Type IV in HBSS (Life Technologies, Invitrogen) at 37°C for 45 minutes. Single-cell suspensions were obtained by passing lung homogenates through 18, 21 and 24G syringes before being passed through 70 and 40 μm cell strainers (BD Biosciences, Heidelberg, Germany). Viable cell count was determined using Trypan Blue stain. Cells were then resuspended in cold staining solution (0.1% sodium azide/5% FCS/0.05% Triton X-100 in 1× PBS) containing antibodies (Biolegend, Fell, Germany) against Cd45 (FITC-conjugated; 1:100), Cd31 (FITC-conjugated; 1:100), Epcam (APC-Cy7-conjugated; 1:50), Sca-1 (Pacific Blue-conjugated; 1:100), Pdgfra (APC-conjugated; 1:100) and Kit (PE-Cy7-conjugated; 1:50) for 30 minutes on ice in the dark and were then washed with 0.1% sodium azide in PBS. LipidTOX staining was performed according to the manufacturer’s instructions (Invitrogen). FACS measurements and sorting were carried out using a FACSAria III cell sorter (BD Biosciences) and the endogenous tomato signal was detected using the PE channel. Gates were set according to unstained controls.

Quantitative data were assembled using GraphPad Prism software (GraphPad Software) and presented as average values ± s.e.m. Figures were assembled using Adobe Photoshop CS5.

We would like to thank Matthew Jones for his help in genotyping the mice used for these experiments, Jana Roskovius for contributing to the immunofluorescence protocol, Kerstin Goth for efficiently managing mouse colonies at the Justus Liebig University Giessen and Salma Dilai for her help with processing the movies. We also acknowledge Oleg Pak for his help with confocal microscopy. Finally, we would like to thank Jonathan Branch, Clarence Wigfall and Soula Danopoulos for their technical support at CHLA.