Regulation of fibroblast lipid storage and myofibroblast phenotypes during alveolar septation in mice

Reagents

Mice

Mice bearing the PDGF-Rα-GFP construct have been described (27, 28). Production and nuclear localization of GFP is under the control of the endogenous pdgf-rα promoter (Fig. 1A). GFP expression in the PDGF-Rα-GFP mice spatially and temporally recapitulates endogenous pdgf-rα expression (16). The mice used in this study carried one pdgfrα-GFP allele (that does not encode for active PDGF-Rα) and one functional pdgf-rα allele. These heterozygous mice are phenotypically identical to wild-type (GFP⁻) mice, except for nuclear GFP, which enables their identification (16). Mice with a targeted deletion of LoxP-pdgfrα (Fig. 1C) have also been described (37). The DNA coding Cre-recombinase was inserted into exon 1 of tagln (TG) and mediates Cre-recombination postnatally but not in the embryo (64). Transgelin is expressed in pulmonary myofibroblasts, pericytes, and smooth muscle cells. The PDGFRα-GFP mice were bred with mice possessing the transgelin-driven Cre-recombinase B6.129S6-Tagln^tm2(cre)Yec/J (Jackson Labs 00678) and the LoxP-flanked dTomato reporter, which bears a LoxP-flanked stop codon B6.Cg-Gt(ROSA)26Sor^{tm14(CAG−tdTomato)Hze}/J (Jackson Labs 007914). After Cre-mediated excision of the stop codon, these mice demonstrate the nuclear GFP marker in cells expressing pdgfrα and the dTomato cytoplasmic marker in cells expressing Tagln. Triple heterozygote PDGFR⁻FP^+/−;TGCre^+/−;R26dTom^+/− mice were used to distinguish myofibroblastic Tagln-bearing PDGFR⁻FP⁺ LF from PDGFR⁻FP⁺ LF, which do not express Tagln.

Analysis of Images Obtained Using Confocal Microscopy

The Sox9- or MRTF-A-containing cells in the alveolar walls were classified as follows: nuclear GFP⁺ (pdgf-rα expressing, green), PoPo3⁺ (all nuclei, red), Sox9 or MRTF-A (AlexaFluor 647, pseudocolor blue) in cells without GFP (PoPo3, but without GFP), and Sox9 or MRTF-A in cells containing GFP and PoPo3. With the use of uniform segmentation criteria, the pixel regions contained 1) both red and blue, but not green, (nuclei that were Sox9⁺ and GFP⁻), 2) red, blue, and green (nuclei that were Sox9⁺ and GFP⁺), and 3) all PoPo3⁺ nuclei (red). The nuclei of cells that were Sox9⁺, GFP⁻, or Sox9⁺, GFP⁺ were enumerated and expressed relative to the total of pdgfrα-expressing alveolar cells (GFP⁺) or nonexpressing alveolar cells (GFP⁻). Because MRTF-A was located in the cytoplasm, a circle with a diameter of 10 μm was placed over each nucleus. The cell was classified as MRTF-A⁺ if the blue pixels met the segmentation criteria and were within the nuclear-centered circle. The percentages of nuclei in Sox9, GFP⁺ or Sox9, GFP⁻ groups were calculated for each field, and a mean was determined for each mouse. The mean data from four mice were combined to obtain a mean and SE for each age.

Statistical Methods

Data were expressed as means ± SE of the number of different mice that were used or the number of different experiments that were performed using cell cultures (46). ANOVA (either 1, 2, 3 way as indicated in the legends for Figs. 1–10) was performed using Systat (Chicago, IL) and Student's t-test using Microsoft Excel. Additional details describing how the data from different images or image stacks were combined appear in the method for a particular stereological procedure. Post hoc tests are described in the legends for Figs. 1–10. Values of P < 0.05 were considered significant.

LF Differing in pdgfrα Expression (GFP Fluorescence Intensity) Exhibit Divergent Phenotypes

Our objective was to determine whether variations in PDGFRα levels identify distinct populations of lung progenitor cells using GFP fluorescence intensity as a surrogate for PDGFRα expression. LF that had been isolated from PDGFRα-GFP mice were stained for CD140a (PDGFRα) without permeabilization and examined using FACS to correlate the intensity of GFP fluorescence with the abundance of cell surface PDGFRα. The results in Fig. 1A show that the abundance of CD140a positively correlated with GFP intensity, indicating that GFP accurately represents the PDGFRα gene product. When LF were permeabilized, 100% of the GFP⁺ LF contained CD140a (data not shown). Alternatively using FACS, we sorted LF from PDGFRα-GFP mice and observed that PDGFRα mRNA was more abundant in GFP^high than in GFP^low LF (Fig. 1A).

We first performed flow cytometry on LF that had been fixed immediately after isolation and stained for various antigens (Fig. 2A). Based on the intensity of GFP fluorescence, three subpopulations of LF were observed (GFP⁻, GFP^low, and GFP^high). Whereas the dot plots shown in Fig. 2A depict all adherent cells, for subsequent quantitative comparisons, CD45⁺ cells that had adhered to tissue culture plastic during selection (including macrophages and potentially CD45⁺ bone marrow fibrocytes) were excluded from the analyses. After gating on CD45-negative cells, which do not stain with APC-anti-CD45, we identified three populations based on GFP fluorescence intensity (R3: GFP⁻, R4: GFP^low, and R5: GFP^high; Fig. 2Bc). We observed differential distribution of Sca1⁺, CD34⁺, and Sca1⁺,CD34⁺ cells in the R3, R4, and R5 gates (Fig. 2, Bd-Bf). Specifically, fewer Sca1⁺ cells were observed in the GFP^low subpopulation (Fig. 2Be). The proportion of CD45⁺ cells in the isolated GFP⁻ and GFP⁺ cells was 49.0 ± 3.5 and 1.6 ± 0.6%, respectively (mean ± SE, n = 5). These data demonstrate that CD45⁻ LF contain three distinct populations of PDGFRα-expressing cells as determined by GFP fluorescence intensity.

We next compared differentiation and progenitor markers in the three LF populations differing in GFP fluorescence intensity. Markers for neutral lipid droplets (stain with LTR, Fig. 3A) or ADRP (perilipin 2; Fig. 3B) were more abundant in the GFP^low population. By contrast, a larger proportion of GFP^high than GFP^low cells contained α-SMA (Fig. 3C). Next, cell surface markers of mesenchymal progenitors (Sca1, CD34, and CD166) were assessed. Sca1⁺ LF were observed in both the GFP⁺ and GFP⁻ populations, although CD34⁺ LF were more prevalent among GFP⁺ LF (Fig. 3, D and E). A significantly larger fraction of the GFP^low population displayed CD166 (Fig. 3, E and F).

Next, we investigated whether PDGFRα expression can distinguish between proliferative and quiescent LF. We found that Ki-67, an estimator of cellular proliferation, was more abundant in GFP^high compared with GFP^low LF at both P4 (Fig. 4A) and P8 (Fig. 4B), with a higher proportion of Ki67⁺,GFP^high LF at P4 than at P8 (P < 0.01, 3-way ANOVA with Student-Newman-Keuls post hoc test). A converse expression pattern was observed for markers of quiescence. Specifically, p57kip2 (Fig. 4C) and G0S2 (Fig. 4D) were more abundant in the GFP^low than in the GFP^high LF populations at P8. There were no differences in G0S2 comparing P4 with P8 (at P4 15.5 ± 5.4%, mean ± SE of the GFP^low LF). We also observed that the progenitor cell transcription factor Tcf21 was more abundant in the GFP^low population at both P4 (data not shown) and P8 (Fig. 4E). Likewise, G0S2 and tcf21 mRNA levels were higher in GFP^low than in either the GFP⁻ or GFP^high populations (Fig. 4F). Thus, the balance tilts toward proliferation in PDGFRα-high LF, whereas it tilts toward quiescence in PDGFRα-low LF.

Active expression of particular genes is required to sustain progenitor cells and allay differentiation. Sox9 delays the differentiation of mesenchymal progenitors into smooth muscle cells, whereas Dlk1 modulates differentiation into either adipocytes or myofibroblasts (31, 56, 65). There was a trend toward greater abundance of Dlk⁺ LF in the GFP^high population (P = 0.06, 2-way ANOVA, Student-Newman-Keuls post hoc test; Fig. 5A). Dlk1⁺, Sca1⁺ LF comprised a significantly larger proportion of the GFP^high compared with either the GFP⁻ or GFP^low populations (Fig. 5A, inset). A larger proportion of GFP⁺ LF (either high or low) contained Sox9 compared with GFP⁻ LF (Fig. 5B). We next gated on the CD45⁻, Sca1⁺ cells and found that a larger proportion of Sox9⁺ LF resided in the GFP^high compared with either the GFP⁻ or GFP^low populations (Fig. 5B, inset). Therefore, although Sca1⁺, CD45⁻ GFP-negative LF were observed, Sca1⁺ LF within the GFP^high population exhibited a higher abundance of cell surface Dlk1 and Sox9. Analysis of mRNA levels demonstrated that Sox9 mRNA expression was more abundant in GFP^high but not in GFP^low compared with GFP⁻ LF (P = 0.15, 2-way ANOVA; Fig. 5C). Splice-independent Dlk1 mRNA was higher in the GFP^high subpopulation than in GFP⁻ LF, whereas GFP^low LF had significantly less dlk1 mRNA compared with GFP^high LF and GFP⁻ LF (Fig. 5C). This shows that Sox9 and Dlk1 are predominantly found in the Sca1⁺, GFP^high cells, identifying the progenitor characteristics of this LF population.

PDGFRα is Required to Maintain Sox9 in LF

PDGFRα and Sox9 may coincidently mark the same LF subpopulation, or PDGFRα signaling may directly influence Sox9 gene expression. To examine these possibilities, we studied LF from mice bearing a targeted deletion of pdgfrα in cells that express tagln, a marker of smooth muscle-like cells (Fig. 1C). We compared lipid-rich and myofibroblastic cells using the lipophilic fluorescent dye LTR after gating out CD45⁺ LF. Although pdgfrα deletion had no impact on the proportion LTR⁺ LF (Fig. 6A, bars on right), pdgfrα deletion increased the proportion of LF with Dlk1 on the cell surface within LTR⁺, Sca1⁺, or CD34⁺ cells (Fig. 6A, bars on left). Whereas the proportion CD34⁺ LF was reduced in pdgfrα-deficient mice, both CD166⁺ and the double-positive CD166⁺ and Dlk1⁺ LF were elevated after pdgfrα deletion (Fig. 6, B and C), showing that features of the PDGFRα GFP^low population were preserved.

We then examined the consequences of transgelin-mediated pdgfrα deletion on the myofibroblast marker α-SMA, which predominated in the GFP^high myofibroblasts. In line with findings in Fig. 3, we observed a reduction in α-SMA + LF (Fig. 6D, bars on right). Similarly, deletion of pdgfrα reduced the proportion of α-SMA⁺, Sox9⁺ LF within both the CD45⁻ and the fraction that was both CD45⁻ and Sca1⁺ (Fig. 6D). To further examine effects of PDGFRα gene deletion on myofibroblasts, we examined periostin (postn), a matricellular protein produced by myofibroblasts and cardiac fibroblasts (53). Whereas total postn mRNA was unchanged following pdgfrα gene deletion, periostin transcripts lacking exon 21 were diminished (Fig. 6E), demonstrating that the splice isoforms are differentially regulated. Figure 6F compares expression of the splice-independent dlk1-T (total) and the dlk1-A and dlk1-C splice isoforms in MLF isolated from control and TGCre^+/−;PDGFRαF/F mice. Although dlk1-T was unchanged, dlk1-C was increased in mice bearing the gene deletion. The Dlk1-C isoform cannot be cleaved and therefore would be preferentially identified using FACS for which only the cell surface pool was analyzed. This would explain why Dlk1 was increased in the retained lipid-laden LF after pdgfrα deletion (Fig. 6A). By contrast, tcf21 mRNA (which was more abundant in the GFP^low, lipid-laden LF) and pparγ mRNA (induces adipocyte differentiation) increased after pdgfrα gene deletion (Fig. 6E). Therefore, tagln-mediated deletion of pdgfrα preferentially depleted myofibroblasts while maintaining the population of LF that assume a lipid storage phenotype.

PDGF-AA Increases Sox9 in Cultured LF

We next examined whether the ligand PDGF-A stimulates Sox9 gene expression by quantifying Sox9 mRNA and protein in primary cultures of LF, which assume a myofibroblastic phenotype in culture. In parallel, we analyzed MRTF-A because it is inversely related to Sox9 gene expression (31). PDGF-A increased Sox9 mRNA (Fig. 7A) and protein (Fig. 7B) expression compared with untreated control LF, whereas TGF-β1 did not alter Sox9 although it did increase MRTF-A protein (Fig. 7C). Therefore, PDGFRα signaling enhances Sox9 expression but does not enhance the transition to myofibroblasts.

Sox9⁺ pdgfrα-Expressing LF Decrease, Whereas MRTF-A⁺ LF Increase, During Alveolar Septation

Our next step was to examine whether there was an inverse relationship between Sox9 and MRTF-A in PDGFRα-expressing alveolar cells. Because more α-SMA⁺ myofibroblasts accumulate between P8 and P12, this interval should encompass the transition to myofibroblasts (27). Sox9 and MRTF-A were assessed in cells bearing the PDGFRα-GFP marker, and cells lacking GFP served as an internal control for cells that do not express PDGFRα. However, because both epithelial and mesenchymal cells contributed to GFP⁻ cells in lung tissue, we could not specifically assess Sox9 in the GFP⁻ LF population as we had done using FACS. Consistent with a PDGFRα-mediated effect, we observed a decrease in Sox9 and a concomitant increase in MRTF-A at P12 relative to P8 within GFP⁺ but not GFP⁻ alveolar cells (Fig. 8). Sox9 immunostaining was observed in alveolar cells that did not express PDGFRα, of which many are likely to be epithelial cells (8). These data demonstrate that the transition of PDGFRα-GFP⁺ LF to myofibroblasts is accompanied by a decrease in Sox9.

Lipid-Laden, PDGFRα-GFP^low LF Diminish After P8

We next examined whether PDGFRα gene expression regressed after secondary septation (which is maximal before P15) and whether the kinetics of regression vary with the level of PDGFRα gene expression. For these studies, we marked myofibroblasts as those that express tagln (tagln-Cre activated tdTomato), and cells containing lipid droplets were marked with a blue fluorescent lipophilic dye. GFP levels serve as a surrogate for PDGFRα expression. Lipid droplets were not observed in cells that contained tagln-dTomato (Fig. 9A), although they were observed in GFP⁻ alveolar cells that did not express tagln-dTomato (data not shown). The tagln-dTomato⁻, GFP⁻ cells were excluded from analysis because this population likely contains alveolar type II cells in addition to LF. A higher proportion of alveolar cells that express both tagln-dTomato and PDGFRα-GFP^high was observed at P8 (Fig. 9B), but this proportion decreased with age. Stereological analysis of tagln-dTomato, PDGFRα-GFP^high (GFP^high,dTom⁺) cells demonstrated that at P8 83 ± 13% (mean ± SE, n = 4) of the cells were positioned nearer the alveolar duct, i.e., within the proximal 40% of the distance extending from the duct to the alveolar base. Whereas a large majority (∼4.5-fold more) of GFP^low LF contained lipid droplets at P8, GFP^high LF accounted for most of the lipid storage in GFP^high LF at P15 and older (Fig. 9C). Similar age-related relationships were observed when the data were expressed relative to lung volume as determined by displacement (data not shown). Therefore, PDGFRα and tagln are only transiently coexpressed when septation is most active, and the GFP^low lipid-laden cells are nearly completely absent at P15 when septation starts to diminish.

Alveolar Surface Area is Decreased and Compliance is Increased in TGCre^+/−;PDGFRαF/F Mice

Some TGCre^+/−;PDGFRαF/F mice with a less severe phenotype survived to adulthood. Their lungs showed (Fig. 10) a reduction in alveoli and alveolar ducts (A and B) and significantly greater static compliance (C). Elastic fibers (shown in the insets) were diminished. TGCre^+/−;PDGFRαF/F mice had a higher pressure-volume curve shape constant, lower dynamic elastance, increased hysteresivity (η = G/H), and a smaller alveolar surface area compared with control lungs (Table 1). These findings are consistent with a loss of α-SMA and elastin-producing myofibroblasts.

LF Bear Markers Characteristic Of Mesenchymal Progenitors

Current therapies for pulmonary emphysema or fibrosis reduce complications but do not allay the progression of these diseases, which involve alveolar destruction. Alveolar regeneration, activated to repair damaged alveoli and to restore the gas exchange surface area, requires restoration and repopulation of resident cells and/or an influx of circulating stem cells. During alveolar development, contractile interstitial fibroblasts control air entry and support the thin capillary-epithelial interface of alveoli. To explore how resident progenitors could be harnessed to restore alveoli, we studied how fibroblast subpopulations cooperate during alveolar development and have summarized their contrasting properties in Table 2. In other tissues (adipose, bone, heart, and skeletal muscle), CD45⁻ bi- or oligopotent PDGFRα⁺ mesenchymal cells assume adipogenic, myogenic, or osteogenic phenotypes and display markers of mesenchymal progenitor cells (Sca1, CD34, CD166) (25, 42, 58). Using mice that express GFP under the control of the PDGFRα promoter, we examined mesenchymal progenitor cell markers as well as the divergent proliferative and differentiation capabilities in PDGFRα-expressing and -nonexpressing LF in neonatal mice. As others have observed in lungs from adult mice, Sca1⁺ LF were identified in GFP⁻ LF not expressing pdgfrα, as well as in GFP⁺ LF (Fig. 3C) (38). Within the Sca1⁺ population, Dlk1⁺ and Sox9⁺ were more abundant among PDGFRα-GFP^high myofibroblasts (Fig. 5), consistent with their involvement in both skeletal and smooth muscle programming and differentiation (2, 58). A larger proportion of PDGFR-expressing than -nonexpressing LF exhibited the mesenchymal progenitor marker CD34 (Fig. 3E), and this fraction diminished when pdgfrα was deleted (Fig. 6B) (49). Although pdgfrα deletion did not reduce Sca1⁺ LF, the proportion of Sox9 + LF decreased, whereas cell surface-associated Dlk1⁺ LF increased within the Sca1⁺ population (Fig. 6, A and D). This differential effect suggests that downstream effectors such as Sox9 could be targeted to fine tune the regenerative function of fibroblasts.

Whereas Sca1⁺ cells were less abundant in GFP^low compared with GFP-negative LF (Fig. 3D), significantly more GFP^low cells exhibited CD166 expression (Fig. 3F). A larger fraction of the GFP^low than either GFP⁻ or GFP^high LF contained lipid droplets (Fig. 3A). This differs from the findings of McQualter and associates who observed that CD166-negative lung stromal cells were more likely to acquire lipid droplets in culture (39). Because both the level of CD166 expression and the phenotype change with culture of mesenchymal cells, our freshly isolated LF are not equivalent to cultured stromal cells from adult lungs, although cells from both sources were CD34⁺ and CD166⁺ (Fig. 3, E and F) (59). We found that LF bearing the progenitor markers Sca1 and CD34 reside within both the lipid-laden and myofibroblastic subpopulations, which is similar to others' observations in the adult lung and other organs (21, 49, 58).

Tcf21 (Pod-1) is required during branching morphogenesis, but little is known about tcf21 gene expression during alveolar septation (44). In response to cardiac stress, Tcf21⁺ mesenchymal cells migrate from the epicardium into the myocardium, where resident cardiac fibroblasts mount fibrotic reactions (6, 52). We found that tcf21 gene expression was greater in PDGFRα-low, lipid-laden LF than in PDGFRα-high myofibroblastic LF (Fig. 4). Similarly, other investigators observed that PDGFRα expression is required to sustain epicardial fibroblasts and adipocytes from white fat, but not coronary artery smooth muscle cells or brown adipocytes (1, 52, 55).

The proliferation marker Ki-67 was lower, whereas the growth arrest proteins G0S2 and p57kip2 were higher, in PDGFRα-low than in PDGFRα-high LF, suggesting that the lipid-laden population becomes quiescent (Fig. 4). G0S2 is expressed in the lung and is abundant in adipose cells, and its expression increases during differentiation from preadipocytes to mature adipocytes. In these mature cells, G0S2 inhibits adipose triglyceride lipase and controls the size of cytoplasmic lipid droplets (11, 19, 20). Others have reported that the expression of G0S2 is diminished at P8 in lungs from mice bearing a compound deletion of both FGFR3 and FGFR4, which exhibit defective alveolarization (54). Concordant with our observation, growth arrest in PDGFRα-expressing oligodendrocyte precursors, smooth muscle cells, and cardiomyocytes is also regulated by p57kip2 (CDKN1c) (22, 24).

CD166⁺, Dlk1⁺ LF are Maintained When PDGFRα is Reduced

CD166, also termed activated leukocyte adhesion molecule, has been identified on adipocyte-derived mesenchymal stem cells, fetal lung cells, bronchial fibroblasts, and adipocyte or chondrocyte progenitors, particularly during proliferation and migration (15, 47). When pdgfrα was deleted, Dlk1-positive, CD166-positive, and Dlk1,CD166 double-positive LF were more prevalent within the Sca1⁺ or CD34⁺ subpopulations (Fig. 6, A and B). Others have shown that CD166 is predominantly found in progenitor populations and that Dlk1 can suppress differentiation, enabling Dlk1⁺ progenitors to accumulate (10, 56). Taken with our observations in pdgfrα-deficient mice, it is possible that CD166 and Dlk1 work in concert to regulate differentiation.

Dlk1 gene expression and function are regulated in a complex fashion variously involving paternal imprinting, alternative mRNA splicing, and extracellular protein cleavage (56). Thereby, the protein assumes divergent characteristics during development that have been best characterized in adipocytes and myocytes (2). Most notably, the two major splice variants termed here dlk1-A and dlk1-C involve inclusion (dlk1-A) or exclusion (dlk1-C) of a juxtamembrane extracellular TACE cleavage site, releasing a soluble peptide (Dlk1^S) that increases preadipocyte proliferation (41). In contrast, the uncleaved membrane-associated form (Dlk1^M) suppresses proliferation. Comparably in myocytes, the Dlk1^S form inhibits, whereas the Dlk1^M form promotes, myocyte fusion and myotube formation, and Dlk1-C mRNA and Dlk1^M predominate during postnatal muscle growth in the mouse (48). This complex regulation contributes to the paradoxical observation that Dlk1 promotes myogenesis during embryogenesis but inhibits myogenic repair in adult mice (2). To our knowledge, our findings using qRT-PCR are the first demonstration that PDGFRα influences alternative splicing of Dlk1 during development.

The PDGFRα-high population contained more α-SMA⁺ LF than the PDGFR-low population (Fig. 3C). Alveolar cells containing both PDGFRα- and tagln-driven dTomato localized near the alveolar entry rings, indicating that tagln-mediated PDGFRα deletion targets the distal septal tips where α-SMA is most abundant (32). In contrast, LF bearing markers of the PDGFRα-low subpopulation (LTR⁺, CD166⁺, and Tcf21⁺) were preserved. Within Sca1⁺ LF, the proportion of Sox9⁺ LF diminished, whereas Dlk⁺ LF increased, suggesting that PDGFRα deletion has a differential effect on LF populations bearing Sox9 or Dlk1. Observation of diminished postn and increased pparγ mRNA in the PDGFRα-deficient LF further supports a shift from the myofibroblastic toward a lipid storage phenotype.

Sox9 is Involved in the Differentiation of PDGFRα-Expressing LF to Myofibroblasts

Our novel finding that pdgfrα and sox9 are coexpressed in alveolar myofibroblast progenitors parallels observations made in OPC and epicardial fibroblast precursors. PDGFRα and the SoxE proteins, Sox9 and/or Sox10, mark a population of NG2 chondroitin sulfate protein-bearing OPC in the periventricular zone of the brain and the spinal cord and regulate their progenitor state and migration (23). It remains unclear whether the Sox9 and PDGFRα coincidently mark the same cellular population or whether they interact to enhance proliferation and migration (14). PDGFRα is required to maintain Sox9 gene expression in embryonic epicardial cells, for migration into the myocardium, and for differentiation (52). Forced expression of Sox9 enables epicardial cells to overcome defects in migration and differentiation that are acquired after pdgfrα gene deletion (52). These data suggest that Sox9 acts downstream from PDGFRα.

Consistent with findings in cardiac fibroblasts and OPC, we observed that Sox9 was more abundant in Sca1⁺ GFP⁺ LF when PDGFRα gene expression was higher (Fig. 5B) (14, 52). PDGF-A, which exclusively signals through PDGFRα, increased Sox9 mRNA and protein levels, suggesting a pretranslational effect (Fig. 7, A and B). Important PDGFRα downstream signaling mediators include PI3K-Akt, p65Rel A, and MEK1/2-Erk, which have all been shown to regulate Sox9 gene expression (29). In addition, we observed a decrease in the numbers of PDGFRα-GFP⁺ alveolar cells containing Sox9 and a reciprocal increase in cells containing MRTF-A from P8 to P12 (Fig. 8), which is accompanied by an increase in α-SMA (32). There is a similar reciprocal relationship between Sox9 and myocardin in smooth muscle cells. Sox9 suspends smooth muscle cell differentiation by complexing with serum response factor (SRF) and interfering with the binding of SRF to CarG box elements or by complexing with the transcription factor TSHZ3 (31, 63).

PDGFRα and Modulation of LF Phenotypes During Alveolar Restoration

We observed that the lipid-storing PDGFRα-low population was abundant in neonates but virtually absent in adults where PDGFRα-high lipid-laden LF predominate (Fig. 9) (27, 32). However, following pneumonectomy in adult PDGFRα-GFP mice, Chen and associates observed that α-SMA was more abundant in the PDGFRα (GFP)-low subpopulation (9). One possible explanation is that pneumonectomy restores the PDGFRα-low LF subpopulation, which in that context contains α-SMA and participates in alveolar regeneration.

Our observations about the divergent characteristics of PDGFRα-expressing LF have potential therapeutic relevance to human diseases. In bronchopulmonary dysplasia, the transition from LF to myofibroblasts results in thickened alveolar septa and reduced oxygen diffusion in the alveolar walls via increased collagen deposition. Current salutary therapies, including positive pressure ventilation and supplemental oxygen, may further exacerbate these alveolar structural abnormalities. However, the PPARγ agonist rosiglitazone improves alveolar formation and reduces collagen (62). These findings suggest that directing fibroblasts from a myofibroblast to a lipid storage phenotype may normalize alveolar architecture during bronchopulmonary dysplasia. A similar strategy may benefit adults with pulmonary fibrosis based on data in cultured human lung fibroblasts in which pioglitazone, another PPARγ agonist, reduces collagen content and bleomycin-induced fibrosis (40). Focusing on specific targets of PDGF-A signaling, such as Sox9, may help select a restorative phenotype that avoids excessive fibrosis.