1. Cell Biology
  2. Developmental Biology

Decoding the IGF1 signaling gene regulatory network behind alveologenesis from a mouse model of bronchopulmonary dysplasia

  1. Feng Gao  Is a corresponding author
  2. Changgong Li
  3. Susan M Smith
  4. Neil Peinado
  5. Golenaz Kohbodi
  6. Evelyn Tran
  7. Yong-Hwee Eddie Loh
  8. Wei Li
  9. Zea Borok
  10. Parviz Minoo  Is a corresponding author
  1. Division of Neonatology, Department of Pediatrics, University of Southern California, United States
  2. Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, United States
  3. Department of Biochemistry and Molecular Medicine, Keck School of Medicine, University of Southern California, United States
  4. Norris Medical Library, University of Southern California, United States
  5. Department of Nephrology, Jiangsu Provincial Hospital of Traditional Chinese Medicine, China
  6. Division of Pulmonary, Critical Care and Sleep Medicine, Department of Medicine, University of California, San Diego, United States
  7. Hastings Center for Pulmonary Research, Keck School of Medicine, University of Southern California, United States
https://doi.org/10.7554/eLife.77522
Share this article
Cite this article
  1. Feng Gao
  2. Changgong Li
  3. Susan M Smith
  4. Neil Peinado
  5. Golenaz Kohbodi
  6. Evelyn Tran
  7. Yong-Hwee Eddie Loh
  8. Wei Li
  9. Zea Borok
  10. Parviz Minoo
(2022)
Decoding the IGF1 signaling gene regulatory network behind alveologenesis from a mouse model of bronchopulmonary dysplasia
eLife 11:e77522.
https://doi.org/10.7554/eLife.77522
  2. Download BibTeX
  3. Download .RIS

Abstract

Lung development is precisely controlled by underlying gene regulatory networks (GRN). Disruption of genes in the network can interrupt normal development and cause diseases such as bronchopulmonary dysplasia (BPD) – a chronic lung disease in preterm infants with morbid and sometimes lethal consequences characterized by lung immaturity and reduced alveolarization. Here, we generated a transgenic mouse exhibiting a moderate severity BPD phenotype by blocking IGF1 signaling in secondary crest myofibroblasts (SCMF) at the onset of alveologenesis. Using approaches mirroring the construction of the model GRN in sea urchin’s development, we constructed the IGF1 signaling network underlying alveologenesis using this mouse model that phenocopies BPD. The constructed GRN, consisting of 43 genes, provides a bird’s eye view of how the genes downstream of IGF1 are regulatorily connected. The GRN also reveals a mechanistic interpretation of how the effects of IGF1 signaling are transduced within SCMF from its specification genes to its effector genes and then from SCMF to its neighboring alveolar epithelial cells with WNT5A and FGF10 signaling as the bridge. Consistently, blocking WNT5A signaling in mice phenocopies BPD as inferred by the network. A comparative study on human samples suggests that a GRN of similar components and wiring underlies human BPD. Our network view of alveologenesis is transforming our perspective to understand and treat BPD. This new perspective calls for the construction of the full signaling GRN underlying alveologenesis, upon which targeted therapies for this neonatal chronic lung disease can be viably developed.

Editor's evaluation

This is an important paper describing the role of Igf1 and a corresponding 47-gene network in alveologenesis and bronchopulmonary dysplasia. The authors provide compelling computational and experimental evidence of their findings in a mouse model. The authors also provide solid evidence that these processes are recapitulated in human tissue, however there are caveats around the set of control samples. This paper will be of interest to those interested in lung development and disease.

https://doi.org/10.7554/eLife.77522.sa0

Introduction

Development is precisely controlled by the genetic program encoded in the genome and is governed by genetic interactions (Davidson et al., 2002; Levine and Davidson, 2005). Elucidating the network of interactions among genes that govern morphogenesis through development is one of the core challenges in contemporary functional genomics research (Przybyla and Gilbert, 2021). These networks are known as developmental gene regulatory networks (GRN) and are key to understanding the developmental processes with integrative details and mechanistic perspectives. Over the past few decades, great advances have been made in decoding these networks in classical model systems (i.e. Dequéant and Pourquié, 2008; Longabaugh et al., 2017; Olson, 2006; Satou et al., 2009; Sauka-Spengler and Bronner-Fraser, 2008), as well as in forming in-depth understandings of the general design principles of these networks in development and evolution (Carré et al., 2017; Davidson, 2010; Erwin and Davidson, 2009; Gao and Davidson, 2008; Lim et al., 2013; Peter and Davidson, 2009; Royo et al., 2011). Highlighted among them is the sea urchin developmental GRN, the most comprehensive and authenticated network constructed to date (Peter and Davidson, 2017).

Disruption of any gene in the network could interrupt normal development and conceivably cause diseases. Some diseases are caused by single gene disorders while others are complex and multifactorial such as bronchopulmonary dysplasia (BPD), a common cause of morbidity and mortality in preterm infants characterized by arrested alveolar development with varied severities within the subjects’ lungs (Jain and Bancalari, 2014; Jobe, 1999; Short et al., 2007). Both prenatal insults and postnatal injury increase the risk of BPD. The multifactorial etiology of BPD has made the development of therapies a unique challenge, and currently, no effective treatment exists to prevent or cure this debilitating disease.

In recent clinical trials, therapy with recombinant human IGF1 protein showed initial promise for BPD, but significant limitations did not allow further clinical trials (Ley et al., 2019; Seedorf et al., 2020). One impediment to further development of IGF1 as a therapy is the lack of comprehensive information regarding the precise role of the IGF1 signaling pathway in alveologenesis.

Alveolar development is a highly complex process that is driven by multiple signaling pathways including transforming growth factor beta (TGFB), fibroblast growth factor (FGF), sonic hedgehog (SHH), wingless/integrated (WNT), platelet-derived growth factor (PDGF), vascular endothelial growth factor (VEGF), hepatocyte growth factor (HGF), NOTCH, bone morphogenetic protein (BMP), and insulin like growth factor 1 (IGF1) (Juul et al., 2020; Nabhan et al., 2018; Tsao et al., 2016; Verheyden and Sun, 2020; Wu and Tang, 2021; Zepp and Morrisey, 2019). IGF1 and IGF1R expression are found throughout fetal lung development and fluctuate at different stages. Both IGF1 and IGF1R expression are significantly reduced in BPD lungs (Hellström et al., 2016; Löfqvist et al., 2012; Yılmaz et al., 2017), suggesting a potential role in the pathogenesis of BPD. The function of IGF1 and IGF1R has been examined in their respective constitutive and conditional knockout mice, but their specific function during alveologenesis has not been examined (Epaud et al., 2012; López et al., 2016; López et al., 2015). A recent study found interruption in IGF1 signaling compromised mechanosignaling and interrupted alveologenesis (He et al., 2021).

In our study here, we found Igf1 and Igf1r are primarily expressed in secondary crest myofibroblasts (SCMF) in postnatal mice lungs. As a result, we interrupted IGF1 signaling in SCMF by the inactivation of Igf1r at the onset of alveologenesis in postnatal day 2 (PN2) mouse neonates and analyzed the resulting phenotypes. Inactivation of Igf1r resulted in mutant lungs with simplified and immature alveolar structure resembling that of human BPD at a moderate severity (Short et al., 2007). We reasoned that the phenotype caused by the inactivation of Igf1r in SCMF reflects interruption in the genetic program downstream of IGF1 signaling that drives the normal functions of SCMF in the process of alveologenesis. Alterations in SCMF may further impact specification/differentiation of other key cell types, particularly the alveolar epithelial cells. We aim to decode the genes and their interactions behind this underlying genetic program.

Taking advantage of the high throughput next generation sequencing, people have been trying to build GRNs by computational analysis using wild-type gene expression data (i.e. Jia et al., 2017; Xu et al., 2012). Currently, the most reliable way is still to build GRN experimentally on data from perturbation analyses.

Genes we particularly focus on are the regulatory genes. These genes decide the outcome of a GRN as they regulate and control other genes’ expression forming a regulatory circuitry of varied hierarchy with structural and cellular genes as their terminal targets (Davidson, 2010; Erwin and Davidson, 2009). Their functions are defined by their logical control over the circuitry’s operation, and their synergetic biological effects are manifested by the effector genes under the circuitry (Peter and Davidson, 2009; Peter et al., 2012).

We followed the protocol similar to that used in the construction of the sea urchin GRN (Materna and Oliveri, 2008), regulatory genes were selected at the transcriptomic scale from the LungMAP database (https://www.lungmap.net/) where their expression during alveologenesis was reported; the cellular expression patterns of these genes were then annotated on lungMAP scRNAseq data for the screen of SCMF genes; the regulatory interactions among these genes were subsequently examined from in vivo and in vitro perturbations; and cellular communications were finally determined by secretome-receptome analysis. Combining this data, we constructed the IGF1 signaling GRN underlying alveologenesis from this mouse model of human BPD phenocopy.

Our GRN work on alveologenesis represents a transformative view of how to understand and perhaps even design future preventive and therapeutic strategies for BPD treatment.

Results

Postnatal expression of Igf1 and Igf1r in the lung

To define the expression pattern of Igf1 and Igf1r during early postnatal lung development, we performed RT-PCR using total lung RNA from embryonic day E18 to PN30. The analysis showed Igf1 is expressed dynamically during development, peaking at the onset of alveologenesis and progressively decreasing thereafter (Figure 1A). The expression pattern of Igf1r was biphasic, initially displaying an overlap with Igf1 in early alveologenesis with a subsequent peak occurring on PN21 (Figure 1B). Expression profiling of Igf1 and Igf1r from the entire lung and within myofibroblasts during lung development was calculated based on data from the LungMAP (Figure 1—figure supplement 1A) and from the latest scRNAseq dataset (Figure 1—figure supplement 1B, Negretti et al., 2021). Similar patterns of their expression were observed when compared to our data.

Figure 1 with 2 supplements see all
Download asset Open asset
Temporal and spatial expression of Igf1 and Igf1r during neonatal lung development in mice.

(A, B) Temporal expression of Igf1 (A) and Igf1r (B) from embryonic day 18 (E18) to postnatal day 30 (PN30), quantified by RT-PCR and normalized to Gapdh. RNA used was collected from the whole lung, and data was presented as box plots with each stage represented by at least five lungs. (C, D) Spatial localization of mRNA for Igf1 (C) and Igf1r (D) in PN7 lungs as detected by RNAscope and their overlapped expression with Pdgfra. Outlined area on the left is magnified on the right. Scale bars: 20 um under the whole view and 10 um under the magnified view.

Alveologenesis is characterized by secondary septa/crest formation, and the alveolar myofibroblasts are recognized as the driving force behind it. SCMF was derived from this concept and is broadly adopted in independent publications (i.e. Boström et al., 1996; Li et al., 2015; Li et al., 2018; Sun et al., 2022; Zepp et al., 2021).

The definition, specification, and function of SCMFs have not been systematically characterized. Presently, a consensus marker for these cells hasn’t been established although a list of markers, such as Acta2 (Kugler et al., 2017), Stc1 (Zepp et al., 2021), Tagln (Li et al., 2018), Fgf18 (Hagan et al., 2020), and Pdgfra Li et al., 2015 have been suggested from previous work. To help resolve this issue, we compiled a comprehensive list of these markers suggested from literature and examined their expression across different mesenchymal cell types as clustered on the two latest scRNAseq datasets (Figure 1—figure supplement 1C&Negretti et al., 2021, Zepp et al., 2021). Pdgfra was revealed as a good SCMF marker agreed by both datasets (Figure 1—figure supplement 1C&F). RNAscope showed the overlapped localization between Igf1/Igf1r and Pdgfra (Figure 1C&D), and the majority of Pdgfra+ cells were Igf1+ and Igf1r+ (Figure 1—figure supplement 2), indicating SCMF as a principal cellular site of Igf1 and Igf1r expression in lungs undergoing alveologenesis.

Mesodermal-specific inactivation of Igf1 and Igf1r

As SCMF are derived from the lung mesenchyme early in lung development and Twist2 (Dermo1) is activated at the onset of mesodermal lineage specification (Li et al., 2008), we used Twist2Cre to inactivate Igf1 and Igf1r separately, specifically within mesodermal lineages, and examined the mutant lung phenotype during embryonic and postnatal development (Figure 2—figure supplement 1A).

Loss of Igf1 or Igf1r in mesodermal progenitors using floxed alleles of these genes decreased their mRNA by approximately twofolds (Figure 2—figure supplement 1B). Consistent with previous reports (Epaud et al., 2012; López et al., 2015), reduction in body weight was observed in our Igf1r mutant pups (Figure 2—figure supplement 1C and D).

The Twist2cre;Igf1rflox/flox mutant displayed clearly visible lung defects: thickened saccular walls at E18; dilated sacculi, and severe reductions in the number of secondary crests at PN14 with the latter extending into PN30 (Figure 2—figure supplement 1E). Similar defects of less severity were observed in the Twist2cre;Igf1flox/flox mutant (Figure 2—figure supplement 1F). However, due to the timing of Twist2cre activation, it is difficult to determine whether the impact of Igf1r on alveologenesis occurs postnatally or is a carryover impact of events occurring prior to onset of alveologenesis.

Postnatal inactivation of Igf1r in SCMF profoundly arrests alveologenesis

The SHH targeted (i.e. Gli1+) fibroblasts have been rigorously examined through lineage tracing using Gli1CreERT2;Rosa26mTmG in our lab (i.e. Li et al., 2015; Li et al., 2019). Cell lineage analysis shows the SHH signaling targets different mesenchymal cell lineages through lung development.

There was a window of time in the early postnatal stage during which the derived green fluorescent protein positive (GFP+)cells were observed primarily localized to the secondary septa and secondarily to parabronchial and perivascular smooth muscle fibers (Li et al., 2015). When Tamoxifen was titrated down to a certain dosage, the smooth muscle fibers were no longer labeled by green fluorescent protein (GFP; Figure 2—figure supplement 2C). This very specific regimen was employed in our current paper. Consistent with our observation, it was found that Gli1 is predominantly expressed by proliferative SCMF/myofibroblast cells as calculated from the two largest scRNAseq datasets recently published (Figure 1—figure supplement 1D and G; Negretti et al., 2021; Zepp et al., 2021). Nonetheless, though our experimental scheme on using Gli1 as a driver has demonstrated its high degree of specificity for SCMF, we acknowledge that its off-target effects in other cell types haven’t been fully excluded.

Using Gli1CreERT2, we inactivated floxed alleles of Igf1r on PN2 at the onset of alveologenesis (Figure 2A). Inactivation of Igf1r was validated by genotyping (Figure 2—figure supplement 2A) and its downregulation verified by RT-PCR with RNA from both entire lung and fluorescence-activated cell sorting (FACS)-isolated SCMF (Figure 2—figure supplement 2B). A successful recombination was assessed by Cre-induced GFP (Figure 2—figure supplement 2C). The mutant mice were slightly runted compared to the controls (Figure 2—figure supplement 2D).

Figure 2 with 2 supplements see all
Download asset Open asset
Postnatal inactivation of Igf1r from lung secondary crest myofibroblast cells.

(A) Schematic of the experimental protocol. (B) Hematoxylin and eosin (H&E) staining of lung sections from control (a–c) and Gli1CreERT2 mutant (d-f) mice and their morphometric measurements by mean linear intercept (MLI) (g–i) and peri/area ratio (j–l) at postnatal day 7 (PN7), PN14, and PN30. See Figure 2—figure supplement 1 for the definition and calculation of these indices. (C) Immunostaining of lung sections from control (a–d) and mutant (e–h) mice for elastin (a,e), ACTA2 (b,f), AQP5/HOPX (c,g), and SFTPC/PDPN (d,h), and the comparison of the number of AT1 (i) and AT2 (j) cells between the control and mutant. Quantitative data was presented as mean values +/-SD with data for each experiemntal group collected from eight lobes from four different lungs. p-Value: * stands for 0.05–0.01, ** for 0.01–0.001, *** for <0.001, n.s. for not significant. The same designation is used throughout the paper. Scale bar: 100 um in B and 25 um in C.

Histology of multiple Gli1CreERT2;Igf1rflox/flox lungs at the timepoints PN7, PN14, and PN30, revealed a phenotype of profoundly arrested alveolar formation as measured by the mean linear intercept (MLI; Figure 2B). In addition, ImageJ analysis showed decreased perimeter to area ratio of airspace (peri/area) as well as the number of airspaces per unit of area (# of airspace/area) – all consistent with the alveolar hypoplasia phenotype in the mutant lungs (Figure 2B, Figure 2—figure supplement 2E,F). The largest deviation between the controls and mutants in these measurements occurred at PN14, which marks the midpoint in the alveologenesis phase. Still, the severity of hypoplasia here is eclipsed by that seen in Gli1CreERT2;Tgfbrflox/flox induced BPD phenocopies (Gao et al., 2022).

Immunohistochemical analysis revealed no significant gross changes between the control and mutant lungs in either proliferation or apoptosis (Figure 2—figure supplement 2G). Similarly, examination of specific markers for various lung alveolar cell lineages, including the mesenchyme (ELN, TPM1), myofibroblasts (ACTA2, PDGFRA), endothelial cells (EMCN), lipofibroblasts (ADRP), epithelial cells (NKX2.1, SFTPC, HOPX), and pericytes (NG2) did not reveal significant differences (Figure 2C, Figure 2—figure supplement 2H). Nonetheless, it was observed in the mutant lungs that the deposition of elastin seemed aggregated at the secondary crest tips, and the number of AT1 and AT2 cells were reduced (Figure 2C). The altered elastin deposition has been reported in previous studies (He et al., 2021; Li et al., 2019), but whether this alteration is the cause or the consequence of the impaired alveolar formation remains unknown.

To examine the genetic changes in Gli1CreERT2;Igf1rflox/flox mutant lungs, we tested two selected groups of genes: lung signature (Figure 2—figure supplement 2I) and angiogenesis (Figure 2—figure supplement 2J) genes. The analysis showed significant alterations in a few genes including Fgf10, Igf1r, and Pdgfra in the mutant lungs. It is noteworthy that the RNA used in the test was from whole lung tissue, while inactivation of Igf1r by Gli1CreERT2 was only targeted to SCMF. To determine the cell-specific impacts of Igf1r inactivation, we characterized gene expression in FACS-sorted cells in the following studies.

Identification of SCMF genes altered in mouse lungs of BPD phenotype

Both GFP+ (SCMF) and Tomato+ (non-SCMF) cells were isolated by FACS from lungs dissected from Gli1CreERT2;Rosa26mTmG (control) and Gli1CreERT2;Rosa26mTmG;Igf1rflox/flox (mutant) mice. Using the sorted cells, we compared the gene expression between them to identify genes enriched within SCMFs, those altered within SCMFs, and those altered within non-SCMFs in the mutant lungs (Figure 3A, Figure 3—figure supplement 1A).

Figure 3 with 1 supplement see all
Download asset Open asset
Identification of secondary crest myofibroblasts (SCMF) genes altered in Gli1CreERT2;Igf1rflox/flox mutant lungs.

(A) Schematic of the experimental protocol. (B) RT-PCR data from the selected genes displaying their enrichment in SCMF and alterations in the mutant. Genes marked with the orange line: regulatory genes selected from LungMap database; genes with the green line: common SCMF markers; genes with the blue line: non-SCMF genes. Data was presnted as dot plots with the measurements from three lungs. Red circles: data points meeting the cutoff criteria described in the text; empty circles: data points failing the cutoff criteria. This designation is used throughout the manuscript. (C) Spatial expression of Foxd1-Tomato/ELN (a–c), SOX8/ELN (d-f), and TBX2/ELN (g–i) in the alveolar compartment of postnatal day 14 (PN14) lungs as detected by immunostaining. Specific antibodies were used for SOX8, TBX2, and ELN. RFP antibody was used for Foxd1-Tomato on lungs dissected from Foxd1GFPCreERT2;CAGTomato mice. Scale bar: 25 um for all images. (D) Spatial expression of Wnt5a/Pdgfra (a) and Fgf10/Pdgfra (b) in the alveolar compartment of PN14 lungs as detected by RNAscope. Outlined area on the left is magnified on the right. Scale bars: 20 um under the whole view and 10 um under the magnified view. (E) Biotapestry network illustration of altered SCMF genes and their connections to IGF1 signaling within SCMF. The source of IGF1 can be both autocrine and paracrine. Genes (nodes) are shown in the territories (colored boxes) in which they are expressed. Edges show regulation by the originating upstream factor and are either positive (arrow) or repressive (bar). Signalings across cell membranes are indicated as double arrow heads. DAPI: 4′,6-diamidino-2-phenylindole; FACS: fluorescence-activated cell sorting.

Figure 3—source data 1

Table of genes with their RT-PCR data showing their level of expression in secondary crest myofibroblasts (SCMF), enrichment in SCMF, and differential expression between control and mutant lungs, and annotation of their cellular expression in the lung based on LungMAP scRNAseq data.

Cells in red: data points which meet the cutoff criteria designated in the paper or the cell type where the gene is annotated as expressing.

https://cdn.elifesciences.org/articles/77522/elife-77522-fig3-data1-v3.xlsx
Figure 3—source data 2

List of the primers used in the paper.

Primers are for Mus musculus by default and for Homo sapiens when denoted with Hs.

https://cdn.elifesciences.org/articles/77522/elife-77522-fig3-data2-v3.xlsx

Our analysis didn’t focus on the components of the pathway itself, which are self-conserved and usually tissue-independent (Pires-daSilva and Sommer, 2003) but was instead directed at identifying genes downstream of the pathway, especially the regulatory genes through which the developmental GRN is specified (Erwin and Davidson, 2009). A total of 47 genes known to encode signaling molecules and transcription factors were screened from the LungMAP transcriptomic database (https://www.lungmap.net/), where they were indicated to be expressed from SCMF between PN3 and PN14 in the mouse developmental window (Figure 3—source data 1). Expression of the selected genes, together with 15 known SCMF cell markers and 3 non-SCMF genes (negative control), was examined and compared by quantitative RT-PCR using RNA from the sorted cells.

Three criteria were applied to identify IGF1 signaling targets within SCMF: (1) functionally active in SCMF with its delta cycle threshold (deltaCT) ≤9 (relative to Gapdh in Ctrl_GFP, Figure 3—source data 1), equivalent to ≥10 copies of transcripts per cell (copies of Gapdh transcripts per cell based on Barber et al., 2005); (2) highly enriched in SCMF with fold change (FC) ≥10, p≤0.05 (Ctrl_GFP vs Ctrl_Tomato, Figure 3B left column, Figure 3—source data 1); (3) significantly altered with FC ≥2, p≤0.05 (Mut_GFP vs Ctrl_GFP, Figure 3B right column, Figure 3—source data 1).

Nineteen genes were identified that met all three criteria (genes highlighted red in the first column of Figure 3—source data 1). Spatial localization of a selected number of these genes – including Foxd1, Sox8, Tbx2, Bmper, Actc1, Wnt5a, and Fgf10 – in SCMF was validated by immunohistochemistry (IHC) (Figure 3C, Figure 3—figure supplement 1B) or RNAscope (Figure 3D). Consistently, Wnt5a was recently reported as a signature gene of SCMF (Negretti et al., 2021), and data analysis on the two latest scRNAseq datasets revealed Wnt5a is dominantly expressed from SCMF/myofibroblast along with smooth muscle cells (Figure 1—figure supplement 1E and H).

Regulation of the latter 19 genes by Igf1 signaling was illustrated on Biotapestry (Longabaugh, 2012) as displayed in Figure 3E. Notedly, all connections were drawn directly from the signaling, whereas the regulation may happen indirectly through the crossregulation among these genes (to be examined below).

The constructed link map points to IGF1 signaling as a positive regulator for 17 out of the 19 putative downstream genes (Figure 3E). Based on their molecular and cellular functions, the 19 genes can be stratified into 2 separate groups. One group comprised regulatory genes encoding transcription factors (Gli1, Foxd1, Tbx2, Sox8, Twist2, Rxra) and signaling molecules (Fgf10, Wnt5a, Bmper, Cyr61, Pdgfra). The second group comprised Acta2, Actc1, Bgn, Des, Eln, Cnn1, Fn1, and Tnc, representing genes that encode structural and cellular molecules. Based on the GRN’s hierarchical design (Erwin and Davidson, 2009), it is most likely that IGF1 signaling first targets the regulatory genes, referred to as SCMF specifiers, whose expression subsequently targets the downstream structural and cellular genes in SCMF (SCMF effectors).

Crossregulation of altered SCMF genes in the mouse BPD phenocopy model

Within the inventory of altered SCMF genes, as defined above, is transcription factor Foxd1, one of the most reduced regulatory genes from our RT-PCR measurements, and two growth factor signaling molecules, Fgf10 and Wnt5a, both known to function in postnatal lung development (Chao et al., 2016; Li et al., 2020; Zhang et al., 2020). Secretome-receptome computation using the Fantom algorithm (Ramilowski et al., 2015) revealed the cognate receptors, fibroblast growth factor receptor 1, 3 and 4 (FGFR1/3/4) and receptor tyrosine kinase like orphan receptor 1 and 2 (ROR1/2), involve in FGF10 and WNT5a signaling transduction within SCMF (Figure 4B). The results were validated on bulk RNAseq (Figure 4—figure supplement 1A) and the latest scRNAseq datasets (Figure 4—figure supplement 1B,C; Figure 4—source data 1). WNT5A is a highly evolutionary conserved non-canonical Wnt ligand. In spite of being identified as a non-canonical ligand, it can, under certain circumstances, signal through canonical Wnt signaling pathways directly or indirectly (i.e. Mikels and Nusse, 2006). Current analysis is focused on the signaling’s non-canonical aspect and does not exclude other possibilities.

Figure 4 with 2 supplements see all
Download asset Open asset
Crossregulation of the altered secondary crest myofibroblasts (SCMF) genes.

(A) Schematic of the experimental protocol. (B) Flow chart of the secretome-receptome analysis among SCMF, AT1, and AT2 cells. The ligands and receptors were identified from the following RNAseq datasets: GSE126457 for SCMF (Li et al., 2019), GSE182886 for AT2, and GSE106960 for AT1 (Wang et al., 2018b). (C) RT-PCR data from the altered SCMF genes demonstrating their response to the treatments by Infigratinib, siWnt5a, and siFoxd1. Data was presnted as dot plots with the measurements from five treatments. (D) Biotapestry network illustration of the crossregulation among the altered SCMF genes.

Figure 4—source data 1

List of the ligands and receptors used for secretome-receptome analysis.

https://cdn.elifesciences.org/articles/77522/elife-77522-fig4-data1-v3.xlsx
Figure 4—source data 2

List of the inhibitors and their concentrations used in cell culture treatment.

https://cdn.elifesciences.org/articles/77522/elife-77522-fig4-data2-v3.xlsx

To investigate the possibility of crossregulation of these genes on the other altered SCMF genes, GFP+ SCMFs were isolated by FACS from postnatal Gli1CreERT2;Rosa26mTmG lungs and cultured in vitro (Figure 4A). Cultures were treated with inhibitors which target the gene or pathway of interest. The inhibitor’s dose was determined from the literature and a series of testing (Figure 4—source data 2), and inhibition was validated by examining its target genes (Figure 4—figure supplement 2A). Blocking IGF1 signaling using the IGF1R inhibitor (PPP at 4 uM as in Chen et al., 2017; Jin et al., 2018; Wang et al., 2019) led to altered expression of all genes – except Gli1 – in the same direction as observed in vivo (Figure 4—figure supplement 2B), indicating that isolated cells in vitro recapitulate the observed in vivo findings. Expression level of Gli1 was too low to be reliably detected in culture, likely due to absence of HH signaling which in vivo is provided exclusively by the lung epithelium.

The signaling of FGF10 and the expression of Wnt5a and Foxd1 were blocked in culture with the FGFR inhibitor (Infigratinib at 1 uM as in Manchado et al., 2016; Nakamura et al., 2015; Wong et al., 2018), siWnt5a (at 20 nM as in Nemoto et al., 2012; Sakisaka et al., 2015; Zhao et al., 2017), and siFoxd1(at 20 nM as in Li et al., 2021; Nakayama et al., 2015; Wu et al., 2018), respectively, and their effects on other SCMF genes were quantified by RT-PCR (Figure 4C). Identified genes with significant change are designated as targets of the gene/pathway perturbed, and their connections are plotted on the network (Figure 4D).

The network constructed on the new perturbation data has built connections between Wnt5a and certain fibroblast effector genes (i.e. Acta2, Eln) confirming its regulatory control on fibroblast growth and development. This finding in regard to this aspect of Wnt5a’s function resonates well with the clinical data from the study of some idiopathic pulmonary fibrosis (IPF) patients (Martin-Medina et al., 2018). The architecture of the new construct reveals the identity of three transcription factors FOXD1, TBX2, and SOX8 forming a presumed connection hub. They are targeted by all three signaling pathways within the network and with each other as demonstrated by FOXD1’s regulation of Tbx2 and Sox8, likely representing a core network subcircuit (Peter and Davidson, 2009) tasked to lockdown a regulatory state so that the signaling effect from the upstream can be stabilized.

Alveolar epithelial genes are affected through WNT5a and FGF10 signaling

As mentioned above, both AT1 and AT2 cells were reduced in Gli1CreERT2;Igf1rflox/flox mutant lungs (Figure 2C). To identify any genetic alterations within these cells, we analyzed our FACS-sorted Tomato+ cells (Figure 3A), in which the lung epithelial cells reside, for the expression of a broad list of known AT1/AT2 signature genes, including their canonical markers and the ones identified by scRNAseq (i.e. Treutlein et al., 2014; Nabhan et al., 2018; LungMAP). The comparison between the control and the mutant data identified the genes as significantly altered (Figure 5A).

Figure 5 with 1 supplement see all
Download asset Open asset
Epithelial genes affected in Gli1CreERT2;Igf1rflox/flox mutant lung and their connections to IGF1 signaling from secondary crest myofibroblasts (SCMF).

(A) RT-PCR data from selected AT1 and AT2 genes revealing their alteration in the mutant lung. Data was presnted as dot plots with the measurements from three lungs. (B) The IGF1 signaling gene regulatory networks during alveologenesis consisting of genes downstream of IGF1 signaling from three cell types (SCMF, AT2, and AT1) and the intracellular and intercellular regulatory connections among them.

Since our Gli1CreERT2;Igf1rflox/flox mutant was an SCMF targeted deletion, these changes must be a secondary consequence of intercellular crosscommunication. Within the altered SCMF gene list there are a total of three ligands: Wnt5a, Fgf10, and Cyr61 (Figure 4D). CYR61 is more commonly recognized as a matricellular protein (i.e. Lau, 2011) rather than a classical growth factor such as FGF10 and WNT5A. Secretome-receptome analysis indicates that FGF10 and WNT5a from SCMF communicate with alveolar epithelial cells through their cognate receptors, FGFR2 and ROR1 (Figure 4B, Figure 4—figure supplement 1).

The intercellular connections discovered above, plus the affected epithelial genes, were also added to the network as shown in Figure 5B. Multiple evidence has shown that IGF1 signaling can promote lung epithelial growth and development (Ghosh et al., 2013; Narasaraju et al., 2006; Wang et al., 2018b). Our work here reveals a nearly comprehensive look at the genetic pathway behind that.

WNT5a is required for alveolar formation as inferred by the IGF1 signaling GRN

Our effort so far has led to construction of the IGF1 signaling GRN during alveologenesis. The network offers first a bird’s-eye view of how the genes involved in IGF1 signaling are connected in the form of a molecular circuitry for alveologenesis, and then a mechanistic perception of how this circuitry is initially turned on in SCMF for the function of these cells and then advances to the neighboring alveolar epithelial cells to influence their cellular activities.

Under this mechanistic view, the position where a gene is located on the architecture of the network and the connections it has with other genes manifest its role and impact on the network’s operation and outcome, hence, its specific biological function in the process of alveolar formation. The network makes it possible to assess and predict a gene’s function before it is tested in vivo.

Mainly derived from our in vitro data, it was discovered WNT5a signaling on the current network is located immediately downstream of IGF1, and the signaling has connections with many IGF1 downstream genes (Figure 5B). This indicates WNT5A’s feed forward role on the transduction of IGF1 signaling and its biological effects. Indeed, when Wnt5a was inactivated postnatally in mice, the mutant lungs exhibited a BPD-like phenotype with arrested alveologenesis, similar to what was seen in the Igf1r mutant (Figure 5—figure supplement 1; Li et al., 2020). Inversely, the in vivo perturbation data collected from this Wnt5a mutant mouse model can be used to further define the regulatory connections where WNT5A is involved on the current IGF1 GRN.

A GRN of similar components and wiring underlies human BPD

The expression of the mouse IGF1 signaling GRN regulatory genes in SCMF was also examined in lung samples from postmortem human BPD samples (Figure 6—source data 1). The regulatory genes, on the upper hierarchy of the network, determine the network’s outcome. In comparison to non-BPD lungs, 9 of the 12 genes examined were altered, and 8 were altered in the same direction as they were defined in the mouse GRN (Figure 6). These findings indicate a genetic program of similar components and wiring underlying in human BPD.

Figure 6
Download asset Open asset
Regulatory genes from IGF1 signaling gene regulatory networks and their expression in human bronchopulmonary dysplasia (BPD) lungs.

Data was presnted as box plots with nonBPD represnted by three lungs and BPD by four lungs.

Figure 6—source data 1

List of clinical data for human neonatal lung samples.

https://cdn.elifesciences.org/articles/77522/elife-77522-fig6-data1-v3.xlsx

Discussion

With massive gene expression data available (e.g. the LungMAP database) in this postgenomic era, one pressing task is to identify the function of genes, in particular, how they interact with one another (Przybyla and Gilbert, 2021). With GRN analysis as our targeted approach, we have constructed the IGF1 signaling GRN underlying alveologenesis using a mouse model of BPD.

The application of the GRN approach in the lung field is novel as it hasn’t been previously reported. Although not all the genes on the network have been examined by perturbation and the links haven’t been determined as direct or indirect, the regulatory connections running along the constructed GRN are already able to provide potential mechanistic explanation as to how the effect of IGF1 signaling is transduced from one gene to a constellation of its downstream genes, and from one cell type to another. Indeed, blocking WNT5A signaling is confirmed to produce a mouse BPD-like phenotype that mimics Igf1r−/− lungs as inferred by the network (Figure 5—figure supplement 1; Li et al., 2020). Signaling by IGF1, FGF10, and WNT5A have long been recognized in having roles in alveologenesis (i.e. Chao et al., 2016; He et al., 2021; Li et al., 2020; Zhang et al., 2020). The present GRN reveals a full genetic program and the crosstalk among them. The fact that all these signaling pathways have connections on the specification, and effector genes of both alveolar mesenchymal and epithelial cells provide a direct causal link between the signaling and alveolar development.

Behind, the overall process of alveologenesis is a much larger signaling GRN where several different cell types are involved. Within the mesenchymal cell type only, it is known that blocking IGF1, WNT5a, PDGFA, and TGFB signaling leads to impairment in development of alveoli, though with varying severities (Gao et al., 2022; He et al., 2021; Li et al., 2020; Li et al., 2019; Zhang et al., 2020). The sum of these leads to the construction of the hierarchical regulatory connections of these pathways within SCMF (Figure 7). Once signaling pathways from alveolar epithelium, endothelium, and immune cells are included, a much larger and more comprehensive signaling GRN behind alveologenesis is expected to emerge.

Figure 7
Download asset Open asset
The hierarchical connections among the signaling pathways within secondary crest myofibroblasts (SCMF) during alveologenesis.

Clinically, BPD is caused predominantly by extrinsic factors which in the first place would interfere with normal cell’s extracellular activities including cell signaling and communication. From the GRN perspective, BPD is a developmental disease when the signaling GRN for alveologenesis is derailed and disrupted by these extrinsic factors (Figure 7).

The type and effect of these extrinsic factors must be taken into consideration when studying this disease. There was a reported increase in Wnt5A expression by mesenchymal cells from hyperoxia BPD models (Sucre et al., 2020)—a finding seemingly contradictory to our data on Wnt5a. The etiology of BPD is multifactorial in that it involves a plethora of factors such as lung immaturity, injury, inflammation, and genetic defects. In the injury model, the two opposite conditions, hyperoxia and hypoxia, can both cause lung injury and lead to BPD-like phenotypes. In our model, the genetic defect from the targeted genetic manipulation is the culprit behind the lungs of the BPD phenotype. It is expected that genes on the network may respond differently to these different conditions and challenges. The derailed gene expression on the network, either increased or decreased, can both disrupt the signaling GRN for alveologenesis and cause BPD.

The aforementioned network disruption may occur within different cell types, onto different signaling pathways, and upon different hierarchies of the network (Figure 7). From the network’s structural view, the disruption at the higher hierarchy has greater manifestation, thus leading to a higher severity of disease. TGFB signaling, atop the whole hierarchy of the signaling GRN on Figure 7, exhibits in its mutant mouse model the most severe BPD-like phenotype when compared to mutants from the other signaling pathways under it (Chao et al., 2017; Gao et al., 2022; He et al., 2021; Li et al., 2020; Li et al., 2019). Also, different signaling sittings at various hierarchical levels have different coverages (sum of genes under the signaling’s control) on the network. If disruption occurred beyond a signaling’s coverage, the treatment by targeting this signaling would be off target. IGF1 signaling is found to be in the middle hierarchical level of the constructed network (Figure 7). By targeting simply one such signaling and hoping to treat BPD altogether will surely not always work.

Our GRN view on alveologenesis presents a transformative viewing of BPD and could potentially help in designing novel strategies to prevent and treat it. As for prevention, the GRN suggests a study centered on the network’s periphery with its focus on cutting off the extrinsic factors and blocking their connections to the intrinsic alveologenesis GRN. For treatment, it is a study on the network itself focusing on recovering and rescuing the signaling pathways that have been disrupted. However, all these rely on the foundational construction of the whole alveologenesis GRN—which includes all major signaling pathways involved within and between each cell type and the network assembly of the hierarchical regulatory connections among them. Though the network construction is currently not experimentally accessible in human models, our work shows the alveologenesis GRN is well conserved between mouse and human lungs. With the mouse model, the whole signaling GRN behind alveologenesis can be decoded as we did in this paper. The impact of extrinsic factors/signaling pathways and their manipulation on the network can be modeled and tested as well. The insights collected can then be used to guide health delivery for BPD clinically. The general process would then be as follows: clinical patient exams and tests are used to determine the extrinsic factors; cell type gene expression data from bronchoalveolar lavage and/or biopsy are used to reveal genes altered within the alveolar compartments in patient lungs; implied extrinsic factors and altered genes are mapped onto the alveologenesis GRN where points of network disruptions can thus be defined; therapies proven successful from mouse modeling data can then be clinically pursued to recover/rescue the disruption points in the patient.

The GRN is traditionally a network built solely on chemical interactions (RNA, DNA, protein, and other chemicals). Having said that, however, it has been recognized that these interactions can also happen at mechanical and electrical level (i.e. Azeloglu and Iyengar, 2015). Mechanosignaling is of particular interest in the lung as the organ itself is born to function through its non-stopping mechanical movement. New technologies, including whole-genome clustered regularly interspaced short palindromic repeats (CRISPR) perturbation and screening, epigenomic profiling, high-resolution chromatin immunoprecipitation sequencing (ChIP-seq), spatial genomics, and cis-regulatory analysis, are sprouting with their potential use for a high-throughput GRN construction (i.e. Cusanovich et al., 2018; Eng et al., 2019; Sanson et al., 2018; Skene and Henikoff, 2015). Our journey to decode the alveologenesis GRN is also an adventure to construct the new generation GRNs with rising technologies.

A web resource of the network and the data presented in this paper is made available to the public: https://sites.google.com/view/the-alveologenesis-grn/home, which we hope can be used as an online repository to promote any further studies and collaborations in this direction.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
AntibodyAnti-ACTA2 (rabbit polyclonal)AbcamCat#: AB5694IF(1:300)
AntibodyAnti-ACTC1 (mouse monoclonal)Santa CruzCat#: SC-58670IF(1:100)
AntibodyAnti-ADRP (rabbit monoclonal)AbcamCat#: AB108323IF(1:200)
AntibodyAnti-AQP5 (rabbit polyclonal)AlomoneCat#: AQP-005IF(1:100)
AntibodyAnti-BMPER (mouse monoclonal)Santa CruzCat#: SC-377502IF(1:200)
AntibodyAnti-CAS3 (rabbit monoclonal)Cell SignalingCat#: 9664IF(1:100)
AntibodyAnti-ELN (rabbit polyclonal)AbcamCat#: AB21600IF(1:200)
AntibodyAnti-EMCN (mouse polyclonal)R&D SystemsCat#: AF4666IF(1:100)
AntibodyAnti-GFP (mouse monoclonal)Santa CruzCat#: SC-9996IF(1:100)
AntibodyAnti-HOPX (rabbit polyclonal)Santa CruzCat#: SC-30216IF(1:50)
AntibodyAnti-KI67 (mouse polyclonal)R&D SystemsCat#: AF7649IF(1:50)
AntibodyAnti-NKX2.1 (mouse monoclonal)Seven HillsCat#: 8G7G3-1IF(1:50)
AntibodyAnti-NG2 (rabbit polyclonal)AbcamCat#: AB5320IF(1:100)
AntibodyAnti-PDGFRa (rabbit monoclonal)Cell SignalingCat#: 3174IF(1:50)
AntibodyAnti-PDPN (hamster monoclonal)Thermo FisherCat#: 14-5381-82IF(1:300)
AntibodyAnti-RFP (rabbit polyclonal)RocklandCat#: 600-401-379SIF(1:300)
AntibodyAnti-SFTPC (rabbit polyclonal)AbcamCat#: AB3786IF(1:200)
AntibodyAnti-SOX8 (mouse monoclonal)Santa CruzCat#: SC-374446IF(1:50)
AntibodyAnti-TBX2 (mouse monoclonal)Santa CruzCat#: SC-514291IF(1:50)
AntibodyAnti-TPM1 (mouse monoclonal)Sigma AldrichCat#: T2780IF(1:500)
Chemical compound, drugTamoxifenSigmaCat#: T56488 mg/ml
Chemical compound, drugInfigratinibSelleckchemCat#: S27881 uM
Chemical compound, drugPPPSelleckchemCat#: S76684 uM
Commercial assay, kitNext Ultra DNA Library Prep KitNew England BiolabsCat#: E7370
Commercial assay, kitRNAscope Multiplex
Fluorescent Reagent Kit V2		Advanced Cell DiagnosticsCat#: 323100
Sequence-based reagentsiFoxd1: siRNA to
Foxd1(SMARTpool)		DharmaconCat#: L-046204-00-000520 nM
Sequence-based reagentsiWnt5a: siRNA to
Wnt5a(SMARTpool)		DharmaconCat#: L-065584-01-000520 nM
Sequence-based reagentsiRNA non-targeting controlDharmaconCat#: D-001810-01-0520 nM
Sequence-based reagentRNAscope probe: Igf1Advanced Cell DiagnosticsCat#: 443901-C11:750
Sequence-based reagentRNAscope probe: Igf1rAdvanced Cell DiagnosticsCat#: 417561-C31:500
Sequence-based reagentRNAscope probe: PdgfraAdvanced Cell DiagnosticsCat#: 480661-C21:750
Sequence-based reagentRNAscope probe: Fgf10Advanced Cell DiagnosticsCat#: 446371-C11:750
Sequence-based reagentRNAscope probe: Wnt5aAdvanced Cell DiagnosticsCat#: 316791-C31:500
Sequence-based reagentRNAscope 3-plex
Positive Control Probe		Advanced Cell DiagnosticsCat#: 3208811:1500
Sequence-based reagentRNAscope 3-plex
Negative Control Probe		Advanced Cell DiagnosticsCat#: 3208711:1500
Strain, strain background (Mus musculus)Rosa26mTmGThe Jackson LaboratoryCat#: 007676
Strain, strain background (M. musculus)CAGTomatoThe Jackson LaboratoryCat#: 007914
Strain, strain background (M. musculus)Twist2CreThe Jackson LaboratoryCat#: 008712
Strain, strain background (M. musculus)Igf1flox/floxThe Jackson LaboratoryCat#: 016831
Strain, strain background (M. musculus)Igf1rflox/floxThe Jackson LaboratoryCat#: 012251
Strain, strain background (M. musculus)CAGCreERThe Jackson LaboratoryCat#: 004453
Strain, strain background (M. musculus)Wnt5aflox/floxKuruvilla LaboratoryN/A
Strain, strain background (M. musculus)Foxd1GFPCreERT2McMahon LaboratoryN/A
Strain, strain background (M. musculus)Gli1CreERT2The Jackson LaboratoryCat#: 007913
Software, algorithmImage JNIHhttps://imagej.nih.gov/ij/
Software, algorithmSTAR 2.5Dobin et al., 2013PMCID:PMC3530905;
https://github.com/alexdobin/STAR; Dobin, 2022
Software, algorithmJMP pro 15Statistical Discoveryhttps://www.jmp.com/en_us/software/predictive-analytics-software.html
Software, algorithmFantom5 Cell ConnectomeFANTOM5 projecthttps://fantom.gsc.riken.jp/5/suppl/Ramilowski_et_al_2015/vis/#/hive
Software, algorithmImarisBitPlanehttp://www.bitplane.com/imaris/imaris
Software, algorithmR 3.2R Projecthttps://www.r-project.org/
Software, algorithmLAS XLeicahttps://www.leica-microsystems.com/products/microscope-software/p/leica-las-x-ls/
Software, algorithmBioTapestryInstitute for Systems Biologyhttp://www.biotapestry.org/

Mouse breeding and genotyping

Request a detailed protocol

All animal studies were conducted strictly according to protocols approved by the University of Southern California (USC) Institutional Animal Care and Use Committee (Los Angeles, CA, USA). The mice were housed and maintained in pathogen-free conditions at constant room temperature (20–22°C), with a 12 hr light/dark cycle and free access to water and food. Twist2Cre, Gli1CreERT2, Rosa26mTmG, CAGTomato, Igf1flox/flox, Igf1rflox/flox, and CAGCreER mice were purchased from the Jackson Laboratory. Foxd1GFPCreERT2 mice were generated by McMahon lab at USC. SftpcGFP mice were generated by Wright lab at Duke University. Wnt5aflox/flox mice were provided by Kuruvilla lab at Johns Hopkins University.

Twist2cre;Igf1flox/flox and Twist2cre;Igf1rflox/flox mice were generated by breeding Twist2cre with Igf1flox/flox and Igf1rflox/flox, respectively.

Gli1CreERT2;Rosa26mTmG mice were generated by breeding Gli1CreERT2 and Rosa26mTmG mice.

Gli1CreERT2;Igf1rflox/flox mice were generated by breeding Gli1CreERT2 mice with the Igf1rflox/flox mice.

Rosa26mTmG;Igf1rflox/flox mice were generated by breeding Rosa26mTmG mice with the Igf1rflox/flox mice.

Gli1CreERT2;Rosa26mTmG;Igf1rflox/flox mice were generated by breeding Gli1CreERT2;Igf1rflox/flox mice with the Rosa26mTmG;Igf1rflox/flox mice.

Foxd1GFPCreERT2;CAGTomato mice were generated by breeding Foxd1GFPCreERT2 mice with the CAGTomato mice.

CACCreER;Wnt5aflox/flox mice were generated by breeding CAGCreER mice with the Wnt5aflox/flox mice.

Genotyping of the transgenic mice was performed by PCR with genomic DNA isolated from mouse tails. The forward (F) and reverse primers (R) for transgenic mouse genotyping are listed in Figure 3—source data 2.

Tamoxifen administration

Request a detailed protocol

A single dose of tamoxifen (8 mg/mL in peanut oil) was administered by oral gavage to neonates at PN2 (100 μg per pup) with a plastic feeding needle (Instech Laboratories, PA). Neonatal lungs were collected between PN7 and PN30 for morphological, immunohistochemical, cellular, and molecular biological analyses.

Mouse lung tissue

Request a detailed protocol

Mice were euthanized by CO2 inhalation at the time of tissue harvest. Chest cavity was exposed, and lungs cleared of blood by perfusion with cold PBS via the right ventricle. Lungs were inflated with 4% formaldehyde under constant 30 cm H2O pressure and allowed to fix overnight at 4°C. Tissue was dehydrated through a series of ethanol washes after which they were embedded in paraffin and sectioned.

Immunohistochemistry

Request a detailed protocol

H&E staining was performed as usual, and morphometric measurements were made using ImageJ. Immunofluorescent staining was performed as previously described using paraffin-embedded lung sections (Li et al., 2019). In brief, 5-μm tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval. After blocking with normal serum, the sections were probed with primary antibodies at 4°C overnight. Combinations of Alexa Fluor Plus secondary antibodies (Thermo Fisher Scientific) were applied for fluorescent detection of above specific primary antibodies. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole. Primary antibodies used and their sources are listed in the Key resources table below. Images were made with Leica DMi8 fluorescence microscope and processed with Leica LAS X and ImageJ.

RNAScope

Request a detailed protocol

Samples were fixed in 10% neutral buffered formalin, dehydrated with ethanol, and embedded in paraffin wax. 5 μm sections from paraffin blocks were processed using standard pretreatment conditions per the RNAscope multiplex fluorescent reagent kit version 2 (Advanced Cell Diagnostics) assay protocol. TSA-plus fluorescein, Cy3, and Cy5 fluorophores were used at different dilutions optimized for each probe. RNAScope probes used are listed in the Key resources table below. Images were made with Leica DMi8 fluorescence microscope and processed with Leica LAS X and ImageJ.

Mouse lung single-cell dissociation

Request a detailed protocol

Single-cell suspension was prepared as described in Adam et al., 2017 with all the procedures performed on ice or in cold room. Mice were euthanized, and the lungs were perfused with PBS as described above. The lungs were inflated with cold active protease solution (5 mM CaCl2, 10 mg/ml Bacillus licheniformis protease), dissected, and transferred to a petri dish where the heart, thymus, and trachea were removed. The lobes were minced using a razor blade. The minced tissue was then immersed in extra cold active protease solution for 10 min and triturated using a 1 ml pipette. This homogenate was transferred to a Miltenyi C-tube with 5 ml HBSS/DNase (Hank’s balanced salt buffer), and the Miltenyi gentleMACS (magnetic-activated cell sorting) lung program was run twice on GentleMACs dissociator. Subsequently, this suspension was passed through a 100 um strainer, pelleted at 300 g for 6 min, suspended in 2 ml RBC lysis buffer (BioLegend), and incubated for 2 min. At this point, 8 ml HBSS was added and centrifuged again. The pellet was suspended in HBSS, then filtered through a 30 um strainer. The suspension was pelleted again and finally suspended in MACS separation buffer (Miltenyi Biotec) with 2% FBS (fetal bovine serum) for FACS. Cell separation and viability were examined under the microscope and through Vi-CELL cell counter after staining with trypan blue.

Flow cytometry and cell sorting

Request a detailed protocol

FACS was performed on a BD (Becton, Dickinson and Company) FACS Aria II at stem cell flow cytometry core at the Keck School of Medicine of USC. The sorting was gated on viability, singlets, GFP, and/or Tomato. GFP+ and/or Tomato+ cells were collected as needed. For cell culture, cells were sorted in DMEM (Dulbecco's Modified Eagle Medium) containing 10% FBS. For RNA, cells of interest were collected in Trizol-LS reagent (ThermoFisher).

Bulk RNA-seq and scRNAseq data analyses

Request a detailed protocol

Lungs dissected from SftpcGFP mouse at PN14 were dissociated into single-cell suspension, and GFP+ cells were sorted and collected as described above. RNA was extracted using Qiagen RNeasy microkit and then submitted to the Millard and Muriel Jacobs Genetics and Genomics Laboratory at Caltech for sequencing, which was run at 50 bp, single end, and 30 million reading depth. The unaligned raw reads from aforementioned sequencing were processed on the PartekFlow platform. In brief, read alignment, and gene annotation and quantification, were based on mouse genome (mm10) and transcriptome (GENECODE genes-release 7). Tophat2 and upper quartile algorithms were used for mapping and normalization. The RNA-seq data have been deposited with GEO (Gene Expression Omnibus) under the accession number GSE182886.

The raw scRNAseq datasets (GSE160876 and GSE165063 from Negretti et al., 2021, GSE149563 from Zepp et al., 2021) were downloaded from the SRA (Sequence Read Archive) server. The ×10 Genomics’ Cell Ranger pipeline was used to demultiplex raw base call files into FASTQ (FAST-All formatted sequence and its quality data) files, perform alignment, filtering, barcode counting, and UMI (unique molecular identifier) counting, combine and normalize counts from multiple samples, generate feature-barcode matrices, run the dimensionality reduction, clustering, and gene expression analysis using parameters once they were provided in the original papers. Quality control was done additionally in Partek Flow to filter out cells with excess mitochondrial reads and possible doublets and remove batch effects. Loupe Browser was used for data visualization and analysis including cell clustering, cell counting, cell type classification, gene expression, and comparative analysis.

Neonatal lung myofibroblast culture and treatment

Request a detailed protocol

FACS sorted GFP+ cells from PN5 neonatal lungs of Gli1CreERT2;Rosa26mTmG mice were suspended in DMEM containing 10% FBS, plated in 24-well culture plates at 50,000 cells/well, and incubated at 37°C with 5% CO2 on day 1 after sorting. On day 2 after sorting, the attached myofibroblasts were washed with PBS and cultured in fresh medium with inhibitors or siRNAs as indicated in each experiment. The ON-TARGETplus SMARTpool siRNAs from Dharmacon were used, and the transfection was done using DharmaFECT transfection reagents. On day 4, the cells were collected for RNA analyses. The cells were authenticated for absence of contaminations.

Real-time quantitative RT-PCR

Request a detailed protocol

Neonatal mouse lung cells were collected from FACS or cell culture as described above. The RNA was isolated with Direct-zol RNA MiniPrep kit according to the manufacturer’s protocol (ZYMO Research). Following RNA purification, cDNA was generated using the SuperScript IV First-Strand Synthesis System (ThermoFisher). Expression of selected genes was quantified by quantitative real-time RT-PCR performed on a light cycler (Roche) or 7900HT fast real-time PCR system (Applied Biosystems) using SYBR green reagents (ThermoFisher). The deltaCT method was used to calculate relative ratios of a target gene mRNA in mutant lungs compared to littermate control lungs. Gapdh was used as the reference gene. Primers for each gene were designed on IDT (Integrated DNA Technologies) website, and the specificity of their amplification was verified by their melting curve. Sequences of the primers are listed in Figure 3—source data 2.

Human neonatal lung samples

Request a detailed protocol

BPD and non-BPD postnatal human lung tissues were provided by the International Institute for the Advancement of Medicine and the National Disease Research Interchange and were classified exempt from human subject regulations per the University of Rochester Research Subjects Review Board protocol (RSRB00056775).

Secretome-receptome analyses

Request a detailed protocol

Cell-to-cell communications were predicted using a published Fantom5 Cell Connectome dataset linking ligands to their receptors (STAR methods) (Ramilowski et al., 2015). The ligands and receptors were identified from the following bulk RNAseq datasets: GSE126457 for SCMF (Li et al., 2019), GSE182886 for AT2 (submitted with this paper), GSE106960 for AT1 (Wang et al., 2018a), and the following scRNAseq datasets: GSE160876 (Schuler et al., 2021), GSE165063 (Negretti et al., 2021), and GSE149563 (Zepp et al., 2021). The IGF1 signaling GRN was drawn in BioTapestry software developed by Longabaugh, 2012.

Quantification and statistical analysis

Request a detailed protocol

In gene expression quantification using RT-PCR, at least three biological replicates (in different cases including lungs/FACS sorted cells/cultured cells) for each experimental group (Ctrl vs Mut, FACS-sorted cell lineage #1vs #2, treated vs untreated, BPD vs nonBPD) were used. Measurement for each biological replicate was repeated three times. The Ct (cycle threshold) was normalized to Gapdh, and the final result was presented as deltaCT or fold change. In morphometric quantification and cell counting, four lungs for each experimental group (Ctrl vs Mut) were used. Left lobe and right inferior lobe from each lung were targeted. Five images from each lobe after staining were analyzed for morphometric quantification (at ×10 magnification) and cell counting (at ×40 magnification). A two-tailed Student’s t-test was used for the comparison between two experimental groups, and a one-way ANOVA was used for multiple comparisons. Quantitative data are presented as mean values ± SD. Data were considered significant if p<0.05.

Data availability

Sequencing data generated for this study have been deposited in GEO under the accession code GSE182886. All other data generated or analyzed during this study are included in the manuscript and supporting files. Source Data files have been provided in Figure 3-Source Data 1&2, Figure 4-Source Data 1&2, and Figure 6-Source Data 1. An online repository of the network and the data presented in this paper has been made available to the public at https://sites.google.com/view/the-alveologenesis-grn/home.

The following data sets were generated
    1. Gao F
    2. Li C
    3. Minoo P
    (2021) NCBI Gene Expression Omnibus
    ID GSE182886. Decoding A Gene Regulatory Network Behind Bronchopulmonary Dysplasia.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE182886
The following previously published data sets were used
    1. Li C
    2. Lee MK
    3. Gao F
    4. Webster S
    5. Di H
    6. Duan J
    7. Yang C
    8. Bhopal N
    9. Pryhubar G
    10. Smith SM
    11. Borok Z
    12. Bellusci S
    13. Minoo P
    (2019) NCBI Gene Expression Omnibus
    ID GSE126457. The Secondary Crest Myofibroblast PDGFRa Controls Elastogenesis Pathway via a Secondary Tier of Signaling Networks During Alveogenesis.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE126457
    1. Wang Y
    2. Tang N
    3. Cai T
    (2018) NCBI Gene Expression Omnibus
    ID GSE106960. The single cell RNA seq of pulmonary alveolar epithelial cells.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106960
    1. Zepp JA
    2. Morley MP
    3. Morrisey EE
    (2021) NCBI Gene Expression Omnibus
    ID GSE149563. The genomic, epigenomic and biophysical cues controlling the emergence of the gas exchange niche in the lung.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149563

References

    1. Jain D
    2. Bancalari E
    (2014) Bronchopulmonary dysplasia: clinical perspective
    Birth Defects Research. Part A, Clinical and Molecular Teratology 100:134–144.
    https://doi.org/10.1002/bdra.23229
    1. Nakayama S
    2. Soejima K
    3. Yasuda H
    4. Yoda S
    5. Satomi R
    6. Ikemura S
    7. Terai H
    8. Sato T
    9. Yamaguchi N
    10. Hamamoto J
    11. Arai D
    12. Ishioka K
    13. Ohgino K
    14. Naoki K
    15. Betsuyaku T
    (2015)
    Foxd1 expression is associated with poor prognosis in non-small cell lung cancer
    Anticancer Research 35:261–268.

Decision letter

  1. Nicholas E Banovich
    Reviewing Editor; Translational Genomics Research Institute, United States
  2. Edward E Morrisey
    Senior Editor; University of Pennsylvania, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Decoding the IGF1 Signaling Gene Regulatory Network Behind Alveologenesis from A Mouse Model of Bronchopulmonary Dysplasia" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by Nicholas Banovich as the Reviewing Editor and Edward Morrisey as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers were in agreement that this manuscript had strong merit and could be suitable for eLife. In particular, there was enthusiasm for the overall novelty of the approach, with respect to translating publicly available data into meaningful biological insights. However, there were a number of major concerns that must be addressed for this manuscript to be considered for publication. Below I have attached complete reviews from the three reviewers. However, for this paper to be considered for publication the following points should be explicitly addressed.

1) All three reviewers noted problems with the RNAScope (and to a lesser degree the immunostaining). New RNAScope images with rigorous quantification should be provided.

2) Multiple reviewers noted that bulk RT-PCR is not sufficient to validate signals specific to particular cell populations. The authors should analyze their results in the context of existing single-cell datasets to assess the robustness of their claims.

3) Multiple reviewers commented on the specificity of Gli1 to SCMFs. This must either be demonstrated with additional KO experiments using are drivers with higher specificity to SCMFs, or alternatively, data needs to be generated to assess whether the phenotype is restricted to SCMFs rather than other mesenchymal cells types.

4) There we a number of questions around the results pertaining to WNT5A. These include questions about specificity to SCMFs, canonical and non-canonical Wnt signaling, and conflicting results with other recent publications. These issues should be addressed.

If you choose to revise and resubmit this manuscript, please address these concerns as well as those raised below.

Reviewer #1 (Recommendations for the authors):

While the conclusions are mostly well-supported, a few claims could be better supported.

1) Some of the immunostaining looks fairly low quality with possible non-nuclear signals from transcription factors (like Foxd1 in Figure 3c). Perhaps a few close-ups to allow better visualization would be helpful.

2) The authors propose that Wnt5a is signaling non-canonically to both the MFB and AT2 cells based on the identification of Ror1 and Ror2 expression. However, the canonical Wnt target gene Axin2 has been identified in a subset of AT2 stem cells by others, and Wnt5a is capable of both canonical and non-canonical Wnt signaling. So, how do the authors know that canonical signaling is not involved, either alone or in conjunction with non-canonical signaling? At least Axin2 levels could be measured upon Wnt5a silencing in MFBs, and perhaps qRT-PCR performed for Axin2 and other Wnt receptors that could mediate canonical signaling. If they cannot experimentally demonstrate the absence of canonical Wnt signaling in MFBs (and AT2 cells), perhaps they should update their GRN to indicate this possibility?

2) The ISH for Igf1 in Figure 1b looks almost ubiquitous, yet the GRN proposes only autocrine signaling. I don't think this hurts their conclusion since they deleted Igf1R in MFBs which should abrogate signaling from any source, but this should probably be addressed or their GRN revised to indicate potential paracrine signaling.

3) In Supplemental Figure 1g, at E18 there is no statistically significant difference in MLI upon deletion of Igf1r, however, the images presented (a and d) appear to show a major reduction in MLI. Is this image representative?

Reviewer #2 (Recommendations for the authors):

In this manuscript, Gao et al., claim that they have constructed a gene regulatory network underlying alveologenesis and its significance to bronchopulmonary dysplasia (BPD). Using RT-PCR and in situ hybridization, the authors claim that Igf1 and Igf1r are expressed in secondary crest myofibroblasts (SCMFs) and their loss of function using Gli1-creER results in alveolar simplification, a tissue level disorganization of alveoli that phenocopies BPD. Further, the authors investigate transcriptomic changes in mesenchymal and epithelial populations from control and Igf1r mutant lungs. For this, the authors developed a 47-gene panel that they claim to represent signaling modules within SCMFs and used this panel for RT-PCR analysis. These data are used to generate an interaction network to evaluate signaling partners, co-effectors mediated by IGF1 signaling in SCMFs, other fibroblasts and alveolar epithelial cells. Using this GRN, the authors concluded that Wnt5a is a key signaling molecule downstream of IGF1 signaling that regulates alveologenesis.

While the authors' claims are salient, some of the conclusions were previously shown by others. For example, a role for Wnt5a driven Ror/Vangl2 has already been shown to be a key mediator of alveologenesis, by virtue of the same signaling effectors identified in this study (Zhang 2020 eLife). Additionally, the genetic loss of function studies performed here are not specific to SCMFs and instead they target broader alveolar and airway fibroblasts. The construction of a gene regulatory network is a potentially exciting tool, but this requires additional perturbations to distinct nodes identified in this work. It would be of particular interest to determine whether there is any redundancy among these nodes and what are the downstream effectors that are specific to each node. While I recognize that this is outside the scope of this work, the authors need to demonstrate the significance of at least one such node.

1. Timing of Igf1r deletion. The authors administered tamoxifen to induce deletion of Igfr1r at PN2 and analyzed tissues at PN7, PN14, PN20. It would be valuable to administer tamoxifen at later stages to test the critical time point at which Igf1r signaling is essential for alveologenesis. For example, the authors may consider administering tamoxifen at PN5 and PN10.

2. In Figure-1, the authors used RNAScope to determine the expression pattern of IGF1 and claim that it is expressed in Pdgfra+ cells. However, it appears that Igf1 transcripts can be seen in other cells. Do the authors need to assess cellular sources of Igf1 transcripts in other cells? Authors could use recently published single-cell transcriptome datasets (e.g. Negretti et al., Development 2021) or from LungMAP to assess this.

3. The authors used Gli1-CreER to delete Igf1r in SCMFs. However, Gli1 has been shown to be expressed in peribronchiolar fibroblasts (Wang et al., JCI 2018) and alveolar lipofibroblasts (Hagan et al., 2020). The cited publication (Li et al., Stem Cells 2015) also shows Gli1-labeled cells around the proximal airways and not just in SCMFs. Therefore, the phenotypes observed in Igf1r KO, as well as all downstream RT-PCR studies, could be a result of loss of IGF1R in other cell types and not specific to just SCMFs. Recent studies have shown that FGF18 is specific to SCMFs (Hagan et al., 2019). Authors could use FGF18-creER line to delete Igf1r.

4. The authors performed RT-PCR for a large panel of genes. However, the cell populations used for these analyses should be more specific. For example, in Figure 3B the authors compared GFP+ vs GFP- bulk populations. I suggest the authors compare between GFP+ SCMFs compared to non-SCMF fibroblasts. Additionally, Figure 5A should be compared with sorted AT1 or AT2 populations as opposed to bulk GFP-negative cells. As the authors show in Figure 2, AT1 and AT2 numbers are reduced in knockout mice, and thus RT-PCR on the bulk "GFP-negative population" does not seem appropriate.

5. Criteria for the selection of gene panels used for RT-PCR: The authors need to provide a rationale for the 47 gene panel selected for RT-PCR, as opposed to numerous other signaling molecules and transcription factors? This is particularly important as the downstream GRN constructed is derived from differentially expressed genes among this pre-selected subset. A more unbiased approach (bulk RNA-Seq or single-cell RNA-Seq) would allow for interrogation of all differentially regulated genes, from which a more inclusive GRN could be constructed.

6. Figure 3C, D – validation of markers should also be done on Igf1 deleted lungs, not just wild type. Additionally, the images for FOXD1 do not show nuclear localization as expected for a transcription factor.

7. In Figure 4, the claim that the predicted GRNs involving FOXD1, TBX2, and SOX8 are key for SCMF identity/function. Current data are merely based on expression and prediction analysis and such claims require additional evidence from loss of function studies.

8. On a technical note, the authors should better characterize their SCMF cultures. Is this culture condition optimized for SCMF maintenance, and how similar are cultured cells to their in vivo counterparts?

9. Is Wnt5a expression specific to SCMFs? To determine the role of Wnt5a specific to SCMFs, authors need to use a specific creER driver line for loss of function studies similar to Comment-3?

Reviewer #3 (Recommendations for the authors):

1. The RNA ISH shown in Figure 1 is not convincing of overlap between Pdgfra and Igf1 as specific markers of SCMF. Is SCMF a distinct cell population? Are all Pdgfra cells SCMF?

2. At least two large single-cell transcriptomes of the developing mouse lung have been published in the last 12 months (Zepp et al., Cell Stem Cell 2021 and Negretti et al., Development 2021) and the examination of the expression of Igf1 and Igfr in specific myofibroblast populations over time should be explored in these publicly available datasets, rather than using whole lung qPCR.

3. While Gli1 is a previously published marker of SCMF, does this marker have specificity in the context of newer single-cell transcriptomic datasets?

4. There exist significant concerns about the rigor of this study, including a lack of information about the number of technical and biological replicates used. This information should be disclosed.

5. Moreover, whole lung qPCR is used on human lung as an attempt to validate these methods, however, there is no significant clinical data given about the patients from who this RNA was obtained (e.g., at what age did they die? from what cause? what gestational age were they born?). Perhaps FFPE human tissue blocks could be used with RNA ISH as a way to validate the qPCR findings.

6. The discussion ignores several recent papers about Wnt5A in chronic lung disease including IPF (A. Martins-Medina, AJRCCM 2018) and BPD (J Sucre, AJRCCM 2020), the later of which showed an increase in Wnt5A expression by mesenchymal sells with hyperoxia injury and in human BPD. How do the authors reconcile their seemingly opposite findings?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Decoding the IGF1 Signaling Gene Regulatory Network Behind Alveologenesis from A Mouse Model of Bronchopulmonary Dysplasia" for further consideration by eLife. Two of the original reviewers have assessed your updated manuscript, and while the manuscript has been improved there are some remaining issues that need to be addressed, as outlined below:

While Reviewer 2 felt you addressed all of their concerns, Reviewer 3 still has three concerns that must be addressed.

The first concern relates to the RNAscope data. While the quality is much improved, the original review asked for quantification of the RNAscope data. As this was not provided and the colocalization of Pdgfra and Igf1 observed in Figure 1 remains difficult to assess, please provide the quantitation requested or explain why such studies are technically unaddressable.

The second concern is around the specificity of Gli1. The Reviewer feels as if the point was not sufficiently addressed. Please see their comment.

The final concern is around the new human data. Again, please see the Reviewer's comments.

We believe that these issues can be addressed with revisions to the language in the manuscript.

The full comments are below.

Reviewer #2 (Recommendations for the authors):

The authors provided additional data and explanation (ex: SCMF specific markers etc.) to address my prior comments. I have no further comments.

Reviewer #3 (Recommendations for the authors):

In this revised manuscript, the authors have attempted to address many of the concerns raised by the reviewers. These efforts have addressed nearly all the issues raised by reviewers initially. That said, there remain some outstanding areas of concern, that if addressed, would significantly improve the manuscript.

RNAscope data in Figure 1 does not support the claims made by the authors. In Figure 1 C, in the representative image shown, there is not consistent colocalization of Pdgfra and Igf1 expression. Despite requests by reviewers, the RNAscope has not been quantified (expression levels and co-localization).

This reviewer continues to have concerns about the specificity of Gli1 for SCMFs. The authors note that Gli1 is a suitable marker when combined with other markers of "SCMF"s, but still need to address the specificity of Gli1 and the possibility of off-target effects in other cell types. We appreciate the rigor gained by the downloading and analysis of other recent scRNAseq datasets, but even this analysis suggests that Gli1 does not mark a unique subpopulation of mesenchyme and may be expressed in other cell types. To address this, the authors could revise the text to acknowledge that using Gli1 as a driver may result in off-target effects in other cell types.

We thank the authors for including additional details about the clinical data. Review of this has raised some additional concerns about how the human data should be interpreted. Two of the controls in the "non-BPD" group should really not be considered controls-the infants were born preterm (26 and 24 wks) but died at 28 and 26 wks, before a BPD diagnosis could be made (since BPD is diagnosed at 36 weeks corrected gestational age). Including these infants in the control group could significantly skew the data as the other control infants were born at term. How do the authors know that the differences in gene expression are not simply due to differences in term vs preterm gestation? A better comparison would be between infants of the same gestational age who did/did not develop BPD. The human BPD data has 4 subjects in control (with 2 of these controls being not true controls as noted above) and 5 in BPD. Given the high degree of variability in human subjects, this appears to be underpowered to detect significant differences between groups, especially since it is not clear if the data are corrected for multiple comparisons. In summary, this human data detracts from an otherwise high-quality manuscript, which provides novel insights into the developing lung. We recognize that obtaining the additional human samples required to expand the dataset is not possible, but feel that the conclusions drawn from the mouse experiments would be stronger without the inclusion of this data.

https://doi.org/10.7554/eLife.77522.sa1

Author response

The reviewers were in agreement that this manuscript had strong merit and could be suitable for eLife. In particular, there was enthusiasm for the overall novelty of the approach, with respect to translating publicly available data into meaningful biological insights. However, there were a number of major concerns that must be addressed for this manuscript to be considered for publication. Below I have attached complete reviews from the three reviewers. However, for this paper to be considered for publication the following points should be explicitly addressed.

We thank the editors and the reviewers for their recognition of the novelty and importance of deriving meaningful biological insight from the massive, publicly available data using the GRN approach.

Our overall goal is to construct a comprehensive developmental GRN behind alveologenesis. Using the IGF1 knockout mice with BPD phenocopy as an example, we built the GRN behind alveologenesis by employing concepts and methods mirrored from the sea urchin GRN’s construction. One primary goal of this paper is to serve as a proof of concept to bring the GRN study into the lung field.

(1) All three reviewers noted problems with the RNAScope (and to a lesser degree the immunostaining). New RNAScope images with rigorous quantification should be provided.

We have repeated these RNAScopes. New images, taken by confocal microscopy, are now provided in Figures 1 and 3 of the revised manuscript.

(2) Multiple reviewers noted that bulk RT-PCR is not sufficient to validate signals specific to particular cell populations. The authors should analyze their results in the context of existing single-cell datasets to assess the robustness of their claims.

The bulk RNAseq data used are all from FACS isolated cells which closely correspond to the cell populations used in our RT-PCR analyses—on which the network was constructed. We are happy to see more single-cell datasets coming out since our work. The two largest scRNAseq datasets from the latest publications (Schuler et al., 2021; Negretti et al., 2021; Zepp et al., 2021) were selected for the suggested analyses. The results of our analyses of the datasets are shown in Figure 4—figure supplement 1 and Figure 4-Source Data 1.

In brief, the raw scRNAseq datasets (GSE160876 from Schuler et al., 2021; GSE165063 from Negretti et al., 2021, GSE149563 from Zepp et al., 2021) were downloaded from the SRA server. The 10x Genomics’ Cell Ranger pipeline was used to demultiplex raw base call (BCL) files into FASTQ files, perform alignment, filtering, barcode counting, and UMI counting, combine and normalize counts from multiple samples, generate feature-barcode matrices, run the dimensionality reduction, clustering, and gene expression analysis using parameters once they were provided in the original papers. Quality control was done additionally in Partek Flow to filter out cells with excess mitochondrial reads and possible doublets and remove batch effects. Loupe Browser was used for data visualization and analysis including cell clustering, cell counting, cell type classification, gene expression and comparative analysis. For Secretome-Receptome analysis, signaling genes were collected respectively from AT1, AT2, and SCMF/Myofibroblast cell clusters; Cell to cell communications were predicted using a published Fantom5 Cell Connectome dataset linking ligands to their receptors (STAR Methods) (Ramilowski et al., 2015). The cellular communications between IGF1, WNT5A, and FGF10 ligands from SCMF and receptors from SCMF, AT2, and AT1 are confirmed to be consistent from bulk RNAseq and scRNAseq data analyses.

(3) Multiple reviewers commented on the specificity of Gli1 to SCMFs. This must either be demonstrated with additional KO experiments using are drivers with higher specificity to SCMFs, or alternatively, data needs to be generated to assess whether the phenotype is restricted to SCMFs rather than other mesenchymal cells types.

Alveologenesis is characterized by secondary septa/crest formation, and the Alveolar Myofibroblasts are recognized as the driving force behind it. Secondary Crest Myofibroblast was derived from this concept and is broadly adopted in independent publications (i.e. Bostrom et al., 1996; Li et al., 2015; Li et al., 2018; Sun et al., 2022; Zepp et al., 2021).

The SHH targeted (i.e. Gli1+) fibroblasts have been rigorously examined through lineage tracing using Gli1CreERT2;Rosa26mTmG in our lab (i.e. Li et al., 2015; Li et al., 2019). Cell lineage analysis shows SHH targets different mesenchymal cell lineages through lung development. There was a window of time in the early postnatal stage during which the derived GFP+ cells were observed primarily localized to the secondary septa, and secondarily to parabronchial and perivascular smooth muscle fibers (Li et al., 2015). When Tamoxifen was titrated down to a certain dosage, the smooth muscle fibers were not labeled by GFP (Figure 2—figure supplement 2C). This very specific regimen which demonstrates the specificity of the Gli1 to SCMFs was employed throughout our current manuscript.

Our analysis from the two largest scRNAseq datasets recently published, shows that consistent with our observations, Gli1 is predominantly expressed by proliferative SCMF/Myofibroblast cells (Figure 1—figure supplement 1D and G, Negretti et al., 2021; Zepp et al., 2021).

Presently, a consensus marker for SCMF hasn’t been established although a list of markers,such as Acta2 (Kugler et al., 2017), Stc1 (Zepp et al., 2021), Tagln (Li et al., 2018), Fgf18 (Hagan et al., 2020) and Pdgfra Li et al., 2015 have been suggested from previous work. To help resolve this issue, we compiled a comprehensive list of these markers suggested from literature and examined their expression across different mesenchymal cell types as clustered on the two latest scRNAseq datasets (Negretti et al., 2021; Zepp et al., 2021). Please see Figure 1—figure supplement 1C and F in revised manuscript. We found: 1. Not all markers are specific to SCMF; 2. Among some of the few good candidate markers agreed upon by both datasets are Pdgfra, Stc1, Gli1, Tgfbi, and P2ry14.

Taken together, we have shown that under conditions described above, in this manuscript, and in several published works (e.g. Li et al., 2015 and Li et al., 2019) cells targeted by Gli1CreERT2 are predominantly SCMFs which are also ACTA2/PDGFRA double positive and, importantly, localized to the secondary crests during alveologenesis. In line with our current understanding of SCMF, we opted to continue our original designation of SCMF to describe them. If an agreement cannot be reached on the using of this term here, we are happy to call them alveolar myofibroblasts or use the term “hedgehog-responsive Pdgfra+ fibroblasts,” as we did in a recent publication (Gao et al., 2022).

The revised text was highlighted in blue on the revised manuscript and can be found from line 128 to 140 on Page 5 and from line 161 to 171 on Page 6.

(4) There we a number of questions around the results pertaining to WNT5A. These include questions about specificity to SCMFs, canonical and non-canonical Wnt signaling, and conflicting results with other recent publications. These issues should be addressed.

These issues have been addressed in our Response to the individual reviewer’s specific concerns (Please see below).

If you choose to revise and resubmit this manuscript, please address these concerns as well as those raised below.

Reviewer #1 (Recommendations for the authors):

While the conclusions are mostly well-supported, a few claims could be better supported.

(1) Some of the immunostaining looks fairly low quality with possible non-nuclear signals from transcription factors (like Foxd1 in Figure 3c). Perhaps a few close-ups to allow better visualization would be helpful.

Foxd1 expression in Figure 3C is labeled by Tomato driven by Foxd1CreERT2, and Tomato is not restricted to nuclei. We have repeated the RNAScope on the other genes, and new images have been taken by Confocal and provided in Figures 1 and 3 of the revised manuscript. We hope this clarifies the situation and alleviates the Reviewer’s concern.

(2) The authors propose that Wnt5a is signaling non-canonically to both the MFB and AT2 cells based on the identification of Ror1 and Ror2 expression. However, the canonical Wnt target gene Axin2 has been identified in a subset of AT2 stem cells by others, and Wnt5a is capable of both canonical and non-canonical Wnt signaling. So, how do the authors know that canonical signaling is not involved, either alone or in conjunction with non-canonical signaling? At least Axin2 levels could be measured upon Wnt5a silencing in MFBs, and perhaps qRT-PCR performed for Axin2 and other Wnt receptors that could mediate canonical signaling. If they cannot experimentally demonstrate the absence of canonical Wnt signaling in MFBs (and AT2 cells), perhaps they should update their GRN to indicate this possibility?

We agree with the reviewer that WNT5a can signal through canonical WNT signaling directly or indirectly, and our work does not exclude this possibility in our model. In fact, our previous study has shown that the Ror receptor does activate canonical WNT signaling (Li et al., 2008). The constructed network only illustrates the data we analyzed within this paper at this time. We have revised the manuscript to include this possibility.

The revised text was highlighted in blue on the revised manuscript and can be found from line 255 to 259 on Page 8.

(2) The ISH for Igf1 in Figure 1b looks almost ubiquitous, yet the GRN proposes only autocrine signaling. I don't think this hurts their conclusion since they deleted Igf1R in MFBs which should abrogate signaling from any source, but this should probably be addressed or their GRN revised to indicate potential paracrine signaling.

The reviewer is correct. Following the suggestion, we have now added the paracrine signaling to the GRN (Figure 3E) in the revised manuscript. Thank you for pointing that out!

(3) In Supplemental Figure 1g, at E18 there is no statistically significant difference in MLI upon deletion of Igf1r, however, the images presented (a and d) appear to show a major reduction in MLI. Is this image representative?

By definition, MLI is the mean length of line segments on random test lines spanning the airspace between intersections of the line with the alveolar surface. The number of intercepts determines the value of MLI. Thicker primary septa is the main difference between the mutant and control while the number of saccules exhibits little variation. This may be one of the limitations of using MLI as a measurement. To address this issue, we used ImageJ and analyzed the Airspace Area in these lungs. The new index can better distinguish these structural differences and was appended to Figure 2—figure supplement 1g, in the revised manuscript.

Reviewer #2 (Recommendations for the authors):

In this manuscript, Gao et al., claim that they have constructed a gene regulatory network underlying alveologenesis and its significance to bronchopulmonary dysplasia (BPD). Using RT-PCR and in situ hybridization, the authors claim that Igf1 and Igf1r are expressed in secondary crest myofibroblasts (SCMFs) and their loss of function using Gli1-creER results in alveolar simplification, a tissue level disorganization of alveoli that phenocopies BPD. Further, the authors investigate transcriptomic changes in mesenchymal and epithelial populations from control and Igf1r mutant lungs. For this, the authors developed a 47-gene panel that they claim to represent signaling modules within SCMFs and used this panel for RT-PCR analysis. These data are used to generate an interaction network to evaluate signaling partners, co-effectors mediated by IGF1 signaling in SCMFs, other fibroblasts and alveolar epithelial cells. Using this GRN, the authors concluded that Wnt5a is a key signaling molecule downstream of IGF1 signaling that regulates alveologenesis.

While the authors' claims are salient, some of the conclusions were previously shown by others. For example, a role for Wnt5a driven Ror/Vangl2 has already been shown to be a key mediator of alveologenesis, by virtue of the same signaling effectors identified in this study (Zhang 2020 eLife). Additionally, the genetic loss of function studies performed here are not specific to SCMFs and instead they target broader alveolar and airway fibroblasts. The construction of a gene regulatory network is a potentially exciting tool, but this requires additional perturbations to distinct nodes identified in this work. It would be of particular interest to determine whether there is any redundancy among these nodes and what are the downstream effectors that are specific to each node. While I recognize that this is outside the scope of this work, the authors need to demonstrate the significance of at least one such node.

1. Timing of Igf1r deletion. The authors administered tamoxifen to induce deletion of Igfr1r at PN2 and analyzed tissues at PN7, PN14, PN20. It would be valuable to administer tamoxifen at later stages to test the critical time point at which Igf1r signaling is essential for alveologenesis. For example, the authors may consider administering tamoxifen at PN5 and PN10.

The specific deletion of Igfr1r in this case depends on Gli1CreERT2. The specificity of Gli1 to SCMF cells has been discussed above in our response to General Comment #3. Administering tamoxifen at PN2 in practice was observed to have the highest efficiency in labeling the SCMFs. As a result, we chose to administer tamoxifen at PN2 to induce deletion of Igfr1r from these cells. Less severe phenotype is expected with later administration. We did try Tamoxifen at PN14, and as expected, there was no clear difference between the control and mutant lungs.

Author response image 1
Download asset Open asset
Postnatal inactivation of Wnt5a.

(A) Schematic of the experimental protocol. (B) HandE staining. (C) MLI. Scale bar: 100um.

2. In Figure-1, the authors used RNAScope to determine the expression pattern of IGF1 and claim that it is expressed in Pdgfra+ cells. However, it appears that Igf1 transcripts can be seen in other cells. Do the authors need to assess cellular sources of Igf1 transcripts in other cells? Authors could use recently published single-cell transcriptome datasets (e.g. Negretti et al., Development 2021) or from LungMAP to assess this.

The reviewer is correct in the sense that IGF1’s expression is broad. Our study was focused on one aspect of IGF1 signal transduction within one specific cell lineage—SCMFs. Gli1CreERT2 was used to delete Igf1 and Igf1r specifically from these cells. Because of its broader expression, deleting Igf1 ligands from a cell doesn’t necessarily block the signaling inside the cell if extracellular IGF1 from other sources is present. As a result, the deletion of Igf1r became the focus of our study in the paper. Igf1 expression from the entire lung and different types of mesenchymal cells was analyzed on suggested public datasets (Negretti et al., 2021; LungMAP; Zepp et al., 2021) and is shown in SFigures 1 A-C and F in the revised manuscript.

3. The authors used Gli1-CreER to delete Igf1r in SCMFs. However, Gli1 has been shown to be expressed in peribronchiolar fibroblasts (Wang et al., JCI 2018) and alveolar lipofibroblasts (Hagan et al., 2020). The cited publication (Li et al., Stem Cells 2015) also shows Gli1-labeled cells around the proximal airways and not just in SCMFs. Therefore, the phenotypes observed in Igf1r KO, as well as all downstream RT-PCR studies, could be a result of loss of IGF1R in other cell types and not specific to just SCMFs. Recent studies have shown that FGF18 is specific to SCMFs (Hagan et al., 2019). Authors could use FGF18-creER line to delete Igf1r.

Please see our response above to General Comment #3 for the specificity of Gli1 to SCMFs. As it was described in Hagan et al., 2020, Fgf18, aside from its expression in mesothelial, peribronchial, and perivascular cells, is primarily expressed not only in Alveolar Myofibroblast but also in AT1 cells. We hope this addresses the concern expressed by the Reviewer.

4. The authors performed RT-PCR for a large panel of genes. However, the cell populations used for these analyses should be more specific. For example, in Figure 3B the authors compared GFP+ vs GFP- bulk populations. I suggest the authors compare between GFP+ SCMFs compared to non-SCMF fibroblasts. Additionally, Figure 5A should be compared with sorted AT1 or AT2 populations as opposed to bulk GFP-negative cells. As the authors show in Figure 2, AT1 and AT2 numbers are reduced in knockout mice, and thus RT-PCR on the bulk "GFP-negative population" does not seem appropriate.

These are great suggestions. One way to separate these cell populations is to label them with their surface markers during cell sorting. Unfortunately, good markers are unavailable for us to use at this time, especially for lung fibroblasts and AT1s. To make the comparison possible between GFP+ SCMFs and non-SCMF fibroblasts, we need an antigen which can label all lung fibroblasts. As for isolating AT1s from bulk AT2/AT1epithelial cells, there is lack of a rigorously tested, specific surface antigen for FACS of AT1s. The reviewer is correct in that AT1 and AT2 numbers are reduced in our knockout mice. Our measurement and comparison on the bulk "GFP-negative population" more evidently reflected the difference of AT2 or AT1 cells as a whole population in the control and mutant lungs. Our recent paper reveals how the population effect has an impact on alveologenesis (Gao et al., 2022). Nonetheless, we agree it is better to compare these cells in a more direct and specific way as the reviewer suggested.

5. Criteria for the selection of gene panels used for RT-PCR: The authors need to provide a rationale for the 47 gene panel selected for RT-PCR, as opposed to numerous other signaling molecules and transcription factors? This is particularly important as the downstream GRN constructed is derived from differentially expressed genes among this pre-selected subset. A more unbiased approach (bulk RNA-Seq or single-cell RNA-Seq) would allow for interrogation of all differentially regulated genes, from which a more inclusive GRN could be constructed.

We apologize for this oversight on our part. In the present work, we used the concept and approaches mirroring the model of sea urchin GRN’s construction. The sea urchin GRN was built predominantly on RT-PCR data as the Next Generation Sequencing was not available at the time. To determine the candidate genes to be examined by RT-PCR in our work, we looked through genes on the LungMAP transcriptomic dataset available at the time and annotated their molecular function, temporal, and spatial expression (Figure 3-Source Data 1). The genes considered for analysis were: (1) transcription factors or signaling molecules; (2) actively expressed in the lung during PN3 to PN14; (3) highly enriched in lung myofibroblasts. 47 genes met these criteria best.

We acknowledge that when using RT-PCR there is always a limitation on the number of genes to be detected. Our present plans to make the GRN more comprehensive, includes exactly what the Reviewer suggested as more reliable and unbiased approaches (bulk RNA-Seq or single-cell RNA-Seq). We thank the Reviewer for this valuable suggestion.

6. Figure 3C, D – validation of markers should also be done on Igf1 deleted lungs, not just wild type. Additionally, the images for FOXD1 do not show nuclear localization as expected for a transcription factor.

The validation of some markers was done on Igf1 deleted lungs as shown in Figure 3—figure supplement 1B. In general, it is hard to observe any additional differences between the control and mutant lungs from staining alone other than the morphological and structural defects already known to us. The main goal of Figure 3C,D is to show where these markers are expressed in the wild type lung. Foxd1 expression in Figure 3C is labeled by Tomato driven behind Foxd1CreERT2, and Tomato is not restricted to nuclei.

7. In Figure 4, the claim that the predicted GRNs involving FOXD1, TBX2, and SOX8 are key for SCMF identity/function. Current data are merely based on expression and prediction analysis and such claims require additional evidence from loss of function studies.

The reviewer is correct from this perspective. One use of the constructed GRN is to make predictions, and its construction indeed, in a way, is a continuous back and forth process between testing and such predictions. As for the claim itself, we agree that it is largely a prediction at the current time.

8. On a technical note, the authors should better characterize their SCMF cultures. Is this culture condition optimized for SCMF maintenance, and how similar are cultured cells to their in vivo counterparts?

We agree with the Reviewer that it is always better to use the culture condition optimized to a specific cell type. Such conditions have not been established for SCMF. In our study, culture conditions for fibroblasts were used for SCMFs. The culture was checked under the microscope every day during the experiments. The growth and morphology of these cells with undulating membranes, multiple processes, and the occurrence of very few dead cells led us to believe the culture conditions were appropriate.

In support of that we examined and compared the response of SCMFs in culture to those in vivo, to IGF1R inhibitor. We found majority of genes tested in vitro were altered in the same direction as observed in vivo (Figure 4—figure supplement 1B), indicating that the cultured conditions are appropriate for maintaining the intrinsic characteristics of SCMFs.

9. Is Wnt5a expression specific to SCMFs? To determine the role of Wnt5a specific to SCMFs, authors need to use a specific creER driver line for loss of function studies similar to Comment-3?

Wnt5a is expressed in SCMF and smooth muscle cells as evidenced by the latest scRNAseq datasets (Figure 1—figure supplement 1EandH), LungMAP Drop-seq data (https://research.cchmc.org/pbge/lunggens/tools/quickview.html?geneid=Wnt5a), and our staining (Figure 3D). Wnt5a is inactivated ubiquitously via CAGCreER. Using Gli1CreERT2, we inactivated Ro1/Ro2—the predicted WNT5A receptors on SCMFs. Comparison of the Wnt5a inactivation mutants using CAGCreER vs Gli1CreERT2 shows identical phenotypes which further suggests that functional Wnt5a is predominantly expressed by targets of Gli1CreERT2, i.e. SCMF (Li et al., 2020). Please see our Response above to General Comment #3 for the specificity of Gli1 to SCMFs.

Reviewer #3 (Recommendations for the authors):

1. The RNA ISH shown in Figure 1 is not convincing of overlap between Pdgfra and Igf1 as specific markers of SCMF. Is SCMF a distinct cell population? Are all Pdgfra cells SCMF?

We have repeated the RNA ISH of Igf1. New images have been taken by confocal microscopy and provided in Figure 1C in the revised manuscript. While definitive criteria for the term SCMF have not been widely agreed upon, it is nevertheless accepted that at a minimum, they are Hedgehog-targeted-Acta2+ and Pdgfra+ fibroblasts, providing a transient niche indispensable for the alveologenesis process (i.e. Li et al., 2019; Li et al., 2018; Zepp et al., 2021).

Pdgfra is identified as a signature marker for SCMF based on the literatures above and the latest scRNAseq datasets (Figure 1—figure supplement 1C and F, Negretti et al., 2021; Zepp et al., 2021). Please refer to our Response above to General Comment #3 for more information about SCMF and its markers.

2. At least two large single-cell transcriptomes of the developing mouse lung have been published in the last 12 months (Zepp et al., Cell Stem Cell 2021 and Negretti et al., Development 2021) and the examination of the expression of Igf1 and Igfr in specific myofibroblast populations over time should be explored in these publicly available datasets, rather than using whole lung qPCR.

We thank the reviewer for this suggestion. We downloaded these datasets, recapitulated the analyses the papers performed, and examined our data and results in the context of these public data sources. The expression of Igf1 and Igfr from the entire lung and in myofibroblast cells through lung development is presented as Figure 1—figure supplement 1 A and B in the revised manuscript.

3. While Gli1 is a previously published marker of SCMF, does this marker have specificity in the context of newer single-cell transcriptomic datasets?

Gli1 is found as a good SCMF marker in the context of newer single-cell transcriptomic datasets suggested by the reviewer (Please see Figure 1—figure supplement 1DandG in the revised manuscript).

4. There exist significant concerns about the rigor of this study, including a lack of information about the number of technical and biological replicates used. This information should be disclosed.

We apologize for this oversight. The Information the Reviewer is asking for can now be found in Quantification and Statistical Analysis under the MATERIALS AND METHODS section. It is copied here for your reference:

“In gene expression quantification using RT-PCR, at least three biological replicates (in different cases including lungs/FACS sorted cells/cultured cells) for each experimental group (Ctrl vs Mut, FACS sorted cell lineage #1vs #2, treated vs untreated, BPD vs nonBPD) were used. Measurement for each biological replicate was repeated three times. The Ct (cycle threshold) was normalized to Gapdh, and the final result was presented as deltaCT or fold change. In morphometric quantification and cell counting, four lungs for each experimental group (Ctrl vs Mut) were used. Left lobe and right inferior lobe from each lung were targeted. Five images from each lobe after staining were analyzed for morphometric quantification (at 10x magnification) and cell counting (at 40x magnification). A two-tailed Student's T-Test was used for the comparison between two experimental groups and a one-way ANOVA was used for multiple comparisons. Quantitative data are presented as mean values +/-SD. Data were considered significant if p < 0.05.”

5. Moreover, whole lung qPCR is used on human lung as an attempt to validate these methods, however, there is no significant clinical data given about the patients from who this RNA was obtained (e.g., at what age did they die? from what cause? what gestational age were they born?). Perhaps FFPE human tissue blocks could be used with RNA ISH as a way to validate the qPCR findings.

The Human neonatal lung samples used in the paper were provided by the International Institute for the Advancement of Medicine and the National Disease Research Interchange and were classified exempt from human subject regulations per the University of Rochester Research Subjects Review Board protocol (RSRB00056775).

We thank the reviewer for pointing out the missing clinical data about these patients. This information is now presented in Figure 6-Source Data 1 in the revised manuscript.

We also thank the Reviewer for the suggestion to examine our qPCR findings with RNA ISH on FFPE human tissue blocks. We plan to consider this additional approach in further studies.

6. The discussion ignores several recent papers about Wnt5A in chronic lung disease including IPF (A. Martins-Medina, AJRCCM 2018) and BPD (J Sucre, AJRCCM 2020), the later of which showed an increase in Wnt5A expression by mesenchymal sells with hyperoxia injury and in human BPD. How do the authors reconcile their seemingly opposite findings?

The reviewer is correct. While we are aware of the two important works, lack of reference to them was simply an oversight on our part. We have now compared their data with ours and discussed them in the revised manuscript.

Our perturbation data has built connections—as shown on the network—between Wnt5a and fibroblast effector genes (i.e. Acta2, Eln), confirming its role in fibroblast growth and development. From this aspect of WNT5A’s function, this finding is consistent with the clinical data from the study of certain IPF patients (Martin-Medina et al., 2018).

An increase in Wnt5A expression by mesenchymal cells from hyperoxia BPD models was reported by Sucre et al., 2020. There are several potential causes for the seemingly conflicting findings between Dr. Sucre’s and ours. For example, etiology of BPD is multifactorial. It includes lung immaturity, hyperoxia, barotrauma and inflammation. Therefore, individual patient may develop BPD from different causes. In support of this, both hyperoxia and hypoxia, two opposite conditions, can both cause lung injury and lead to BPD-like phenotypes. There are also differences in the experimental approaches, which may account for the seemingly different observations that reflect the complexity of Wnt5a signaling and its associated GRN.

The revised text was highlighted in blue on the revised manuscript and can be found from line 278 to 282 on page 8 and from line 362 to 375 on Page 10.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #3 (Recommendations for the authors):

In this revised manuscript, the authors have attempted to address many of the concerns raised by the reviewers. These efforts have addressed nearly all the issues raised by reviewers initially. That said, there remain some outstanding areas of concern, that if addressed, would significantly improve the manuscript.

We deeply appreciate these critical points from reviewer #3 for the purpose to improve our manuscript.

RNAscope data in Figure 1 does not support the claims made by the authors. In Figure 1 C, in the representative image shown, there is not consistent colocalization of Pdgfra and Igf1 expression. Despite requests by reviewers, the RNAscope has not been quantified (expression levels and co-localization).

Cellular localization of Pdgfra, Igf1 and Igf1r from RNAscope was analyzed and quantified as presented in Figure 1—figure supplemental 2.

Legend for Figure Supplemental 2: Majority of Pdgfra+ cells are Igf1+ (A) and Igf1r+ (B). Monochrome DAPI staining was imported into ImageJ where the outline of each cell was traced. Under this outline, Cellular localization of Pdgfra, Igf1 and Igf1r was analyzed and quantified. 10 images from each group were used and data was presented as box plots. Scale bar: 20um for all images.

We didn’t quantify expression of Pdgfra, Igf1 and Igf1r at the cellular level from RNAscope as we have not made any conclusions in the paper which required such data directly. Also, our RNAscope was done on 5um tissue sections. Given a typical mammalian cell diameter at 10um, stained signals collected for each cell are actually from one segmented slice of the whole cell. The relative location of these cellular slices and their thickness may vary cell by cell on the tissue section, which makes their quantification not the reflection of the whole cell and their comparison to each other not so relevant.

The revised text was highlighted in blue on the revised manuscript and can be found from line 140 to 142 on Page 5 and from line 820 to 823 on Page 25.

This reviewer continues to have concerns about the specificity of Gli1 for SCMFs. The authors note that Gli1 is a suitable marker when combined with other markers of "SCMF"s, but still need to address the specificity of Gli1 and the possibility of off-target effects in other cell types. We appreciate the rigor gained by the downloading and analysis of other recent scRNAseq datasets, but even this analysis suggests that Gli1 does not mark a unique subpopulation of mesenchyme and may be expressed in other cell types. To address this, the authors could revise the text to acknowledge that using Gli1 as a driver may result in off-target effects in other cell types.

Nonetheless, though our experimental scheme on using Gli1 as a driver has demonstrated its high degree of specificity for SCMF, we acknowledge that its off-target effects in other cell types haven’t been fully excluded.

The revised text was highlighted in blue on the revised manuscript and can be found from line 173 to 175 on Page 6.

We thank the authors for including additional details about the clinical data. Review of this has raised some additional concerns about how the human data should be interpreted. Two of the controls in the "non-BPD" group should really not be considered controls-the infants were born preterm (26 and 24 wks) but died at 28 and 26 wks, before a BPD diagnosis could be made (since BPD is diagnosed at 36 weeks corrected gestational age). Including these infants in the control group could significantly skew the data as the other control infants were born at term. How do the authors know that the differences in gene expression are not simply due to differences in term vs preterm gestation? A better comparison would be between infants of the same gestational age who did/did not develop BPD. The human BPD data has 4 subjects in control (with 2 of these controls being not true controls as noted above) and 5 in BPD. Given the high degree of variability in human subjects, this appears to be underpowered to detect significant differences between groups, especially since it is not clear if the data are corrected for multiple comparisons. In summary, this human data detracts from an otherwise high-quality manuscript, which provides novel insights into the developing lung. We recognize that obtaining the additional human samples required to expand the dataset is not possible, but feel that the conclusions drawn from the mouse experiments would be stronger without the inclusion of this data.

The mark of 36 weeks corrected gestational age for BPD diagnosis is applied to live patients in clinics. The human samples used in the paper were collected from patients who unfortunately died. The diagnosis of these patients was confirmed by the Histopathological sections of their lungs where BPD is characterized by arrested alveologenesis and peripheral vascular dysmorphia.

https://doi.org/10.7554/eLife.77522.sa2

Article and author information

Author details

  1. Feng Gao

    Division of Neonatology, Department of Pediatrics, University of Southern California, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    fremontgao@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8764-1107

  2. Changgong Li

    Division of Neonatology, Department of Pediatrics, University of Southern California, Los Angeles, United States
    Contribution
    Funding acquisition, Investigation, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared

  3. Susan M Smith

    Division of Neonatology, Department of Pediatrics, University of Southern California, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared

  4. Neil Peinado

    Division of Neonatology, Department of Pediatrics, University of Southern California, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared

  5. Golenaz Kohbodi

    Division of Neonatology, Department of Pediatrics, University of Southern California, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared

  6. Evelyn Tran

    1. Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, Los Angeles, United States
    2. Department of Biochemistry and Molecular Medicine, Keck School of Medicine, University of Southern California, Los Angeles, United States
    Contribution
    Writing - review and editing
    Competing interests
    No competing interests declared

  7. Yong-Hwee Eddie Loh

    Norris Medical Library, University of Southern California, Los Angeles, United States
    Contribution
    Data curation, Software, Investigation
    Competing interests
    No competing interests declared

  8. Wei Li

    Department of Nephrology, Jiangsu Provincial Hospital of Traditional Chinese Medicine, Nanjing, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared

  9. Zea Borok

    Division of Pulmonary, Critical Care and Sleep Medicine, Department of Medicine, University of California, San Diego, San Diego, United States
    Contribution
    Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8673-8177

  10. Parviz Minoo

    1. Division of Neonatology, Department of Pediatrics, University of Southern California, Los Angeles, United States
    2. Hastings Center for Pulmonary Research, Keck School of Medicine, University of Southern California, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Project administration, Writing - review and editing
    For correspondence
    minoo@usc.edu
    Competing interests
    No competing interests declared

Funding

National Heart, Lung, and Blood Institute (5R01HL144932-04)

  • Changgong Li

National Heart, Lung, and Blood Institute (5R01HL143059-04)

  • Parviz Minoo

Hastings Foundation

  • Parviz Minoo
  • Zea Borok

National Institutes of Health (HL122764)

  • Changgong Li
  • Parviz Minoo

National Institutes of Health (R35 HL135747)

  • Zea Borok
  • Parviz Minoo

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Andrew P McMahon (Keck School of Medicine of USC) for providing the Foxd1GFPCreERT2;CAGTomato mice. We thank Dr. Gloria S Pryhuber (University of Rochester Medical Center) for providing the human BPD samples. We thank Arnold Sipos (Keck School of Medicine of USC) for help with imaging and Sean Gao (Arcadia High School/Duke University) for data analysis, editing, and the construction of the Alveologenesis GRN website. This work was supported by the National Institutes of Health [HL144932, HL122764 (C.L.& P.M.), HL143059 (P.M.), R35 HL135747 (Z.B.& P.M.)], the Hastings Foundation (P.M., Z.B.).

Ethics

Human subjects: BPD and non-BPD postnatal human lung tissues were provided by the International Institute for the Advancement of Medicine and the National Disease Research Interchange, and were classified exempt from human subject regulations per the University of Rochester Research Subjects Review Board protocol (RSRB00056775).

Senior Editor

  1. Edward E Morrisey, University of Pennsylvania, United States

Reviewing Editor

  1. Nicholas E Banovich, Translational Genomics Research Institute, United States

Version history

  1. Preprint posted: January 25, 2022 (view preprint)
  2. Received: February 2, 2022
  3. Accepted: October 7, 2022
  4. Accepted Manuscript published: October 10, 2022 (version 1)
  5. Accepted Manuscript updated: October 11, 2022 (version 2)
  6. Version of Record published: October 19, 2022 (version 3)

Copyright

© 2022, Gao et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 576
    Page views
  • 126
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Feng Gao
  2. Changgong Li
  3. Susan M Smith
  4. Neil Peinado
  5. Golenaz Kohbodi
  6. Evelyn Tran
  7. Yong-Hwee Eddie Loh
  8. Wei Li
  9. Zea Borok
  10. Parviz Minoo
(2022)
Decoding the IGF1 signaling gene regulatory network behind alveologenesis from a mouse model of bronchopulmonary dysplasia
eLife 11:e77522.
https://doi.org/10.7554/eLife.77522

Categories and tags

Research organisms

Further reading

    1. Cell Biology

    Rab12 is a regulator of LRRK2 and its activation by damaged lysosomes

    Xiang Wang, Vitaliy V Bondar ... Anastasia G Henry
    Short Report Updated

    Leucine-rich repeat kinase 2 (LRRK2) variants associated with Parkinson’s disease (PD) and Crohn’s disease lead to increased phosphorylation of its Rab substrates. While it has been recently shown that perturbations in cellular homeostasis including lysosomal damage can increase LRRK2 activity and localization to lysosomes, the molecular mechanisms by which LRRK2 activity is regulated have remained poorly defined. We performed a targeted siRNA screen to identify regulators of LRRK2 activity and identified Rab12 as a novel modulator of LRRK2-dependent phosphorylation of one of its substrates, Rab10. Using a combination of imaging and immunopurification methods to isolate lysosomes, we demonstrated that Rab12 is actively recruited to damaged lysosomes and leads to a local and LRRK2-dependent increase in Rab10 phosphorylation. PD-linked variants, including LRRK2 R1441G and VPS35 D620N, lead to increased recruitment of LRRK2 to the lysosome and a local elevation in lysosomal levels of pT73 Rab10. Together, these data suggest a conserved mechanism by which Rab12, in response to damage or expression of PD-associated variants, facilitates the recruitment of LRRK2 and phosphorylation of its Rab substrate(s) at the lysosome.

    1. Cell Biology

    Genome-wide screen reveals Rab12 GTPase as a critical activator of Parkinson’s disease-linked LRRK2 kinase

    Herschel S Dhekne, Francesca Tonelli ... Suzanne R Pfeffer
    Research Advance Updated

    Activating mutations in the leucine-rich repeat kinase 2 (LRRK2) cause Parkinson’s disease. LRRK2 phosphorylates a subset of Rab GTPases, particularly Rab10 and Rab8A, and we showed previously that these phosphoRabs play an important role in LRRK2 membrane recruitment and activation (Vides et al., 2022). To learn more about LRRK2 pathway regulation, we carried out an unbiased, CRISPR-based genome-wide screen to identify modifiers of cellular phosphoRab10 levels. A flow cytometry assay was developed to detect changes in phosphoRab10 levels in pools of mouse NIH-3T3 cells harboring unique CRISPR guide sequences. Multiple negative and positive regulators were identified; surprisingly, knockout of the Rab12 gene was especially effective in decreasing phosphoRab10 levels in multiple cell types and knockout mouse tissues. Rab-driven increases in phosphoRab10 were specific for Rab12, LRRK2-dependent and PPM1H phosphatase-reversible, and did not require Rab12 phosphorylation; they were seen with wild type and pathogenic G2019S and R1441C LRRK2. As expected for a protein that regulates LRRK2 activity, Rab12 also influenced primary cilia formation. AlphaFold modeling revealed a novel Rab12 binding site in the LRRK2 Armadillo domain, and we show that residues predicted to be essential for Rab12 interaction at this site influence phosphoRab10 and phosphoRab12 levels in a manner distinct from Rab29 activation of LRRK2. Our data show that Rab12 binding to a new site in the LRRK2 Armadillo domain activates LRRK2 kinase for Rab phosphorylation and could serve as a new therapeutic target for a novel class of LRRK2 inhibitors that do not target the kinase domain.

    1. Cell Biology

    Differential regulation of hair cell actin cytoskeleton mediated by SRF and MRTFB

    Ling-Yun Zhou, Chen-Xi Jin ... Hao Wu
    Research Article Updated

    The MRTF–SRF pathway has been extensively studied for its crucial role in driving the expression of a large number of genes involved in actin cytoskeleton of various cell types. However, the specific contribution of MRTF–SRF in hair cells remains unknown. In this study, we showed that hair cell-specific deletion of Srf or Mrtfb, but not Mrtfa, leads to similar defects in the development of stereocilia dimensions and the maintenance of cuticular plate integrity. We used fluorescence-activated cell sorting-based hair cell RNA-Seq analysis to investigate the mechanistic underpinnings of the changes observed in Srf and Mrtfb mutants, respectively. Interestingly, the transcriptome analysis revealed distinct profiles of genes regulated by Srf and Mrtfb, suggesting different transcriptional regulation mechanisms of actin cytoskeleton activities mediated by Srf and Mrtfb. Exogenous delivery of calponin 2 using Adeno-associated virus transduction in Srf mutants partially rescued the impairments of stereocilia dimensions and the F-actin intensity of cuticular plate, suggesting the involvement of Cnn2, as an Srf downstream target, in regulating the hair bundle morphology and cuticular plate actin cytoskeleton organization. Our study uncovers, for the first time, the unexpected differential transcriptional regulation of actin cytoskeleton mediated by Srf and Mrtfb in hair cells, and also demonstrates the critical role of SRF–CNN2 in modulating actin dynamics of the stereocilia and cuticular plate, providing new insights into the molecular mechanism underlying hair cell development and maintenance.