Mark1 regulates distal airspace expansion through type I pneumocyte flattening in lung development

In the early stages of fetal lung development [embryonic day (E)9.5–E16.5], the epithelium forms a branched bronchial tree, while the distal ends of bronchioles transform into saccular structures in the later stages (E17.5–E18.5) (Morrisey and Hogan, 2010). Alveolar epithelial cells consist mainly of two types of pneumocytes: flattened type I alveolar epithelial cells (AECI), which mediate gas exchange, and cuboidal type II epithelial cells (AECII), which secrete surfactant proteins to prevent alveoli collapsing. In fetal lung development, they are derived from common progenitors that show a bi-potency to generate both types of pneumocytes (Desai et al., 2014; Treutlein et al., 2014).

Mechanisms of branching morphogenesis by growth factor signaling and physical cues have been well studied (Varner and Nelson, 2017). However, mechanistic insights into distal lung sacculation have been less clear. It has recently been reported that a loss of Wnt signaling results in the reduction of apical constriction, and thereby contributes to distal lung sacculation (Fumoto et al., 2017). The regulation of capillary density by trinucleotide repeats-containing 6c (TNRC6C), an effector for micro (mi)RNA-mediated mRNA repression, is involved in distal lung sacculation through modulating transforming growth factor (TGF-β) signaling (Guo et al., 2017). Epigenetic changes mediated by histone deacetylase 3 (HDAC3) are required for alveolar cell remodeling through the regulation of miRNA and TGF-β signaling (Wang et al., 2016). Follistatin-like 1 (Fstl1), an antagonist against bone morphogenetic protein, is required for elastin deposition in the mesenchyme during lung sacculation and alveologenesis (Geng et al., 2013). Thus, although genetic, epigenetic and miRNA pathways for distal lung sacculation have been described, to date it is unclear how changes in the cell shape of AECI are regulated and involved in distal lung sacculation.

Map-microtubule affinity regulating kinase (Mark) family kinases are AMP-activated protein kinase (AMPK)-related kinases, also known as Par-1 family kinases, that regulate microtubule dynamics and cellular polarity (Bright et al., 2009). Mark family members share catalytic and variable C-terminal spacer domains (Timm et al., 2003). Microtubule-associated proteins (MAPs), doublecortin and PSD-95 (also known as DLG4), which are substrates of Mark family kinases, are involved in neuronal functions (Hayashi et al., 2012; Wu et al., 2012). Recently, Mark family kinases have also been recognized as regulators of the Hippo pathway. Mark1 acts as a downstream molecule of the tumor suppressor LKB1 (also known as STK11), forms a complex with components of the Hippo pathway, and mediates Yap activity through LKB1 (Lizcano et al., 2004; Mohseni et al., 2014). Mark3 forms a complex with MST1 or MST2 (MST1/2; also known as STK4 and STK3, respectively) and DLG5, and inhibits MST1/2 kinase activity (Kwan et al., 2016). Mark4 attenuates complex formation between MST1/2–SAV complex and LATS1, thereby repressing cell proliferation in breast cancer cells (Heidary Arash et al., 2017). Thus, Mark family kinases may be involved in various signaling pathways but their specific roles in vivo are unclear.

Recently, Mark1 was found to be involved in lung branching morphogenesis in vitro; however Mark1-knockout (KO) mice did not show a phenotype during the pseudoglandular stage in vivo (Fumoto et al., 2017). However, Mark1 is also expressed in the mesenchyme but its physiological role has not been addressed in the later stages of lung development. Here, we showed that Mark1 is expressed in fetal lung fibroblasts and is involved in distal lung sacculation through type I collagen assembly.

It was recently shown that Mark1 is induced by Wnt signaling in the epithelium during the early stages of lung development (Fumoto et al., 2017). However, at E14.5, an in vivo phenotype of the Mark1 KO lung was not obvious probably due to a redundancy of function among Mark family members (Fumoto et al., 2017). In the later stages of lung development, the expression of Mark1 mRNA was reduced in the epithelium and increased in the mesenchyme (Fig. 1A). In our previous study with immunostaining, Mark1 protein levels were indeed decreased in the epithelium of the E17.5 lung compared with what was found in the E14.5 lung (Fumoto et al., 2017). In contrast, Mark1 was induced by Wnt activation in cultured fibroblasts isolated from an E18.5 lung (Fig. S1A). The specific expression of Mark1 was confirmed in data from the LungGENS database (Du et al., 2015), revealing that Mark1, but not Mark2 and Mark3, is expressed in myofibroblasts at E18.5 (Fig. 1B). The expression of Mark4 is not shown in the same database. We also isolated E18.5 embryonic lung cells by fluorescence activated cell sorting (FACS), and found that Mark1 was highly expressed in platelet-derived growth factor receptor α (Pdgfrα)-high (Pdgfrα^high) and -low (Pdgfrα^low) cells (Fig. S1B). The expression of Mark1 in these fibroblasts was also confirmed by immunohistochemistry (Fig. 1C), suggesting that multiple fibroblast subsets express Mark1 .Acta2, a myofibroblast marker, was found to be more highly expressed in Pdgfra^high cells (Fig. S1B).

To examine the role of Mark1 expressed in fibroblasts, three guide (g)RNAs, which targeted three independent loci of Mark1, and Cas9 mRNA were injected into one-cell embryos (Fig. S2A) (Sunagawa et al., 2016). Each of these target sites was efficiently mutated in embryos at E17.5 (Fig. S2B). When sequenced, the genomic region between gRNA1 and gRNA3 target sites in the Mark1 locus was mostly deleted (Fig. 1D,E). Immunofluorescence analysis showed that the expression of Mark1 protein disappeared in the Mark1 KO lung from E17.5 (Fig. S2C). The expression of AECI (Pdpn and Aqp5) and AECII markers (Sftpc and Sftpb) in whole lung or isolated lung epithelial cells was not changed in the Mark1 KO lungs (Fig. 1F). Histologically, in wild-type embryos, distal lung saccules, which were composed of flattened AECI and cuboidal AECII alveolar epithelial cells, were clearly formed near distal lung airspace (Fig. 1G; Fig. S2D). However, in Mark1 KO embryos, distal lung sacculation was impaired and the flattening of epithelial cells was rarely observed (Fig. 1G; Fig. S2D). AECI that express Pdpn but not proSP-C, and AECII that express proSP-C but not Pdpn were present in the distal regions of airways in both wild-type and Mark1 KO lungs (Fig. 1G). Taken together, these results suggest that Mark1 regulates distal lung sacculation independently from pneumocyte differentiation.

The observation of the lungs from Mark1 KO mice suggested that impaired distal lung sacculating resulted from defects in lumen expansion in the distal airway associated with AECI flattening. Epithelial tubules generate an expanded lumen through complex processes including cellular polarization, trafficking and apoptosis (Datta et al., 2011; Martín-Belmonte et al., 2008). These processes were previously confirmed in an in vitro epithelial cell model using MDCK cells, where these cells mimicked lumen generation and expansion in vivo (Bedzhov and Zernicka-Goetz, 2014; Rodríguez-Fraticelli et al., 2015). Therefore, to obtain novel insights into how Mark1-expressing fibroblasts are involved in lumen expansion and AECI flattening during development, we developed in vitro distal lung sacculation models by using a culture of fetal lung alveolar organoids in the presence of embryonic fibroblasts.

EpCAM-positive (EpCAM⁺) epithelial cells and Pdgfrα-positive (Pdgfrα⁺) fibroblasts were independently collected by magnet-assisted cell sorting (MACS). Although co-cultures of organoids and fibroblasts have been reported previously (Barkauskas et al., 2013), the viability of fetal fibroblasts in Matrigel, which contains laminin, entactin and type IV collagen, was poor in our hands (data not shown). Therefore, Pdgfrα⁺ fibroblasts were first seeded on type I collagen-coated dishes. Matrigel was then solidified on fibroblasts and finally single epithelial cells suspended in culture medium were seeded on the solidified Matrigel (Fig. 2A,B). When co-cultured with fibroblasts, epithelial cells formed single-layered alveolar organoids, in which they expressed either one of AECI (Pdpn) or AECII (proSP-C) markers, or both (Fig. 2C). Organoids co-cultured with fibroblasts expressed alveolar markers more markedly than epithelial cells cultured alone (Fig. 2D). This is consistent with previous reports showing that Pdgfrα⁺ fibroblasts are required for alveolar differentiation (Zepp et al., 2017). In contrast, the level of the club cell marker Scgb1a1 was decreased, suggesting that alveolar organoids including both AECI and AECII were predominantly grown in this condition. The culture medium was supplemented with epidermal growth factor (EGF), noggin, R-spondin1 (Rspo1) and fibroblast growth factor 10 (FGF10). FGF10 was indispensable for this culture and other factors, especially noggin, were required for efficient differentiation into alveolar epithelial cells (Fig. 2E). Therefore, alveolar organoids are formed from isolated fetal lung epithelial cells when they are co-cultured with Pdgfrα⁺ fibroblasts. Furthermore, Ki-67-positive fibroblasts increased when they were co-cultured with epithelial cells in the presence of culture medium containing noggin (Fig. 2F), suggesting that epithelial cells also benefit from fibroblast proliferation.

The roles of Mark1 in fibroblasts on distal lung sacculation and pneumocyte differentiation were examined by the use of a co-culture system. The lumen expansion of organoids co-cultured with Mark1 KO fibroblasts was decreased when compared with organoids co-cultured with wild-type fibroblasts (Fig. 3A). In addition to CRISPR-Cas9-mediated KO, the same phenotypes were confirmed in organoids co-cultured with fibroblasts with shRNA-mediated Mark1 knockdown (KD) (Fig. 3B). The expression of markers of pneumocyte differentiation was not altered (Fig. 3A,B). The size of Mark1 KO organoids co-cultured with wild-type fibroblasts was unchanged (Fig. 3C), suggesting that Mark1 in fibroblasts is required for organoid lumen expansion but not for pneumocyte differentiation. When co-cultured with wild-type or Mark1 KO fibroblasts, the percentages of organoids including Ki-67-positive cells were 25.4% and 23.33%, respectively (Fig. 3Di). Few Ki-67-positive cells were noted in single organoids co-cultured with wild-type or Mark1 KO fibroblasts (1.17 or 1.43 cells/organoid, respectively; Fig. 3Dii). Significant changes in the number of epithelial cells in a cross-sectional area of organoids, irrespective of co-culture with wild-type or Mark1 KO fibroblasts, did not occur (Fig. 3Diii), suggesting that the decreased lumen expansion in organoids co-cultured with Mark1 KO fibroblasts was not a result of a loss of proliferative ability in epithelial cells.

As Mark1 is expressed in fibroblasts, we hypothesized that Mark1 is involved in distal lung sacculation indirectly through the expression of secreted proteins. To gain insight into the molecular basis of Mark1-mediated distal lung sacculation, a set of genes expressed in myofibroblasts were obtained from the LungGENS database. Genes encoding secreted proteins other than extracellular matrix proteins were first chosen, and genes with a reported function in the lung were further selected (Fig. 4A). Of these, we found that the expression of Hhip and Igf1 was decreased in Mark1 KD or KO fibroblasts co-cultured with epithelial cells (Fig. 4B). The expression of Hhip, a Hedgehog (Hh)-interacting protein, was consistently decreased in Mark1 KO lung (Fig. 4B). Hhip is known as one of the Hh signaling target genes, and Hh signaling is involved in branching morphogenesis through the differentiation of fibroblasts (Bellusci et al., 1997; Chuang et al., 2003; Pepicelli et al., 1998).

The mRNAs for two other Hh signaling target genes, glioma-associated oncogene homolog 1 (Gli1) and Patched 1 (Ptch1), were also decreased in Mark1 KO or KD fibroblasts co-cultured with epithelial cells, suggesting that Mark1 is required for the activation of Hh signaling (Fig. 4B,C). In addition, the level of Acta2, a hallmark of the activation of Hh signaling and fibroblast differentiation (Pepicelli et al., 1998), was decreased in Mark1 KO or KD fibroblasts (Fig. 4C). When activated, fibroblasts express extracellular matrix proteins and remodel the mesenchymal milieu in various contexts (Tomasek et al., 2002; Watsky et al., 2010). Col1a2, a component of type I collagen, was also decreased in Mark1 KO or KD fibroblasts (Fig. 4C). Taken together, these data suggest that Mark1 mediates Hh signaling and is required for fibroblast differentiation. Acta2 and Col1a2 were decreased in Gli1 KD fibroblasts but, in this case, the differentiation of epithelial cells in organoids was concomitantly decreased (Fig. 4D), suggesting that Mark1 is involved in Hh signaling-mediated fibroblast differentiation but not alveolar differentiation.

It has been shown that the Mark family kinases form a complex with centriole proteins and suppresses a cilia disassembly protein, thereby promoting ciliogenesis (Kuhns et al., 2013). The Hh receptor, Ptch1, is localized to cilia, and in the presence of Hh ligands, Ptch1 mediates the translocation of Gli family transcriptional factors to the nucleus (Briscoe and Thérond, 2013). When cilia in co-cultured fibroblasts were observed, the number of cilia was decreased and the length of cilia was shortened in Mark1 KO or KD fibroblasts (Fig. 4E,F). Kinesin family member 7 (Kif7) plays an important role in both cilia formation and Hh signaling (Goetz and Anderson, 2010; Reilly and Benmerah, 2019). As seen in Mark1 KO fibroblasts, Kif7 KD fibroblasts showed defects in cilia assembly and fibroblast activation without defects in alveolar differentiation (Fig. 4G). These results suggest that Mark1 is primarily required for cilia assembly, thereby controlling Hh signaling.

In E17.5 wild-type lung, type I collagen fibers were accumulated around distal lung saccules (Fig. 5A). It is of interest to note a tendency for type I collagen fibers to be aligned beneath flattened AECI (Fig. 5A). However, in the Mark1 KO lung, type I collagen fibers were faintly observed throughout the section and not accumulated beneath epithelial cells (Fig. 5A). In organoids co-cultured with wild-type fibroblasts, the shape of Pdpn⁺ AECI was horizontally flattened but AECI flattening was impaired when they were co-cultured with Mark1 KO fibroblasts (Fig. 5B,C). AECI that express Pdpn but not proSP-C were observed in organoids co-cultured with wild-type or Mark1 KO fibroblasts (Fig. S3A, left panels). There was no significant change in the abundance of AECI (Pdpn), AECII (proSP-C), or epithelial cells expressing both markers in organoids co-cultured with wild-type or Mark1 KO fibroblasts (Fig. S3A, right panel). Cell flattening was observed in Mark1 KO epithelial cells co-cultured with wild-type fibroblasts (Fig. 5B,C), suggesting that Mark1 in fibroblasts induces epithelial cell flattening. When type I collagen was exogenously added to Matrigel (1:1), the decreased size of organoids (Fig. 5D,E) and impaired cell flattening were rescued (Fig. 5F,G), suggesting that Mark1 is functionally required for type I collagen deposition, both in vivo and in vitro.

To examine whether adhesion to type I collagen is required for cell flattening, the LungGENS database was used to determine which type I collagen receptors are expressed in AECI. None of the integrin family members that function as type I collagen receptors were reported in AECI but discoidin domain receptor 1 (Ddr1) had been identified (Carafoli and Hohenester, 2013) (Fig. S3B). However, Ddr1 depletion had no impact on organoid size (Fig. S3C), suggesting that type I collagen-mediated adhesion signaling in AECI was not required. Collagen induces physical stiffness (Cox and Erler, 2011) and, indeed, type I collagen-mixed Matrigel was stiffer than Matrigel only (Fig. S3D), giving rise to the notion that stiffness may be involved in cell flattening.

To address the impact of cell flattening on distal lung sacculation, a mathematical model was developed. It was recently proposed that amniotic fluid pressure promotes epithelial stretching, resulting in the differentiation of AECI (Li et al., 2018). Therefore, the functional relationship between amniotic fluid influx and collagen-mediated cell flattening in distal lung sacculation was considered. Organoids were modeled as cysts in two-dimensional space, in which apical and basal vertices were shared between adjacent cells, as previously described (Fig. 6A) (Fumoto et al., 2017). By applying outward force against the apical side, the effect of amniotic fluid influx was represented as an inside pressure and Kt denoted its magnitude. Cell shape change was described by adjusting the movement of apical and basal vertices, and Ka and Kb denoted adjusting the strength of apical and basal lengths, respectively (Fig. 6A). Furthermore, to evaluate the effect of the mesenchyme on distal lung sacculation, the mesenchyme, including cells and extracellular matrix, was represented by interacting green particles, in which the strength of the inter-particle interaction (Kr) and the mobility of the particle (rg) were adjusted. In the case of higher Ka and Kb values (Ka=Kb=70), the inside pressure did not induce cell flattening and model organoid expansion (Fig. 6B). However, it did so in the case of lower Ka and Kb values (Ka=Kb=14) (Fig. 6C), suggesting that cell flattening is controlled by a distinct force independently of the inside pressure. When adjusting apical and basal vertices close to each other in the absence of the inside pressure, cell flattening occurred but the entire morphology of the model organoid was distorted (Fig. 6D). In the presence of both the inside pressure and forced cell flattening, the morphology of the model organoid expanded without distortion (Fig. 6E), suggesting that cell flattening is regulated independently of the inside pressure, and that both the inside pressure and the cell flattening are required for organoid expansion.

Whether mesenchymal parameters influence distal lung sacculation in the presence of forced cell flattening and inside pressure was also examined. Firstly, the inter-particle interaction (Kr) was reduced as a representation of the reduced cell-to-substrate interaction due to a decreased level of collagen in the Mark1 KO lung. In this case, cyst expansion was slightly increased (Fig. 6F, Kr=0.5, compared to Kr=10), suggesting that these interactions were irrelevant to the Mark1 KO phenotype. When mesenchymal parameters were adjusted to represent the phenotype as seen in the Mark1 KO lung, it was unexpectedly found that expansion of the model organoid was reduced when mesenchymal motility (rg) was weakened (Fig. 6G, rg=0.01, compared to rg=1). These results predict that collagen may be required for distal lung sacculation, not only through cell flattening directly but also indirectly through mesenchymal motility.

Previously, we demonstrated that Mark1 is required for apical constriction in vitro, but the physiological functions of Mark1 were unclear in vivo (Fumoto et al., 2017). Here, we show that in the later stages of lung development, Mark1 was only expressed in fibroblasts, with Mark1 KO lung showing defects in distal lung sacculation. Mechanistically, Mark1 mediated primary cilia formation, thereby regulating Hh signaling, fibroblast activation and type I collagen expression. Defects in distal lung sacculation in the Mark1 KO lung may be a result of impaired AECI flattening due to the loss of type I collagen deposition. Notably, levels of pneumocyte markers were not changed in the Mark1 KO lung, suggesting that type I collagen assembly induces cell flattening independently of pneumocyte differentiation. In addition, although Ddr1, a type I collagen receptor, was expressed in AECI, it had no impact on organoid lumen expansion. The addition of type I collagen in Matrigel rescued the expansion of organoids and cell flattening, with Matrigel stiffness increased by the addition of type I collagen. The cell shape is known to become flattened on a stiff matrix (Buxboim et al., 2017), suggesting that alveolar fibroblasts may control appropriate physical stiffness to induce AECI flattening.

The activity of the Hh signaling pathway is regulated by dynamic intra-ciliary transport (Briscoe and Thérond, 2013; Pak and Segal, 2016). When Hh ligand binds and internalizes Ptch1, a receptor for Hh, Smoothened (Smo) is activated and moved into the primary cilium, where Smo activates a Gli transcription factor and induces its translocation into nucleus to induce Hh-specific gene expression. In this context, cilia are a pivotal structure that support Hh signaling. Indeed, it has been reported that Kif7 regulates cilium length and architecture, thereby controlling Hh signaling (Pak and Segal, 2016). In Mark1 KO fibroblasts, cilia were shortened and Hh signaling activity was decreased. Therefore, it is possible that one of the possible mechanism through which Mark1 regulates Hh signaling is through cilia formation. On the other hand, the cilia assembly is also altered by Pdgfr, Notch, and autophagy signaling (Pala et al., 2017). Especially Notch signaling modulates the response of the Hh signaling pathway by regulating trafficking of Smo to primary cilia (Kong et al., 2015). It is intriguing to speculate that these pathways contribute to cilia formation and Hh signaling; further detailed biochemical and cell biological studies are necessary.

Pdgfrα⁺ fibroblasts are required for lung growth in vivo (Boström et al., 2002), and alveolar organoids from the adult lung grow and differentiate when co-cultured with Pdgfrα⁺ cells (Barkauskas et al., 2013). Recently, distinct fibroblast subsets in either bronchiolar or alveolar regions were found to direct the bronchiolar or alveolar differentiation of epithelial cells, respectively (Lee et al., 2017). Furthermore, a subset of Axin2-positive (Axin2⁺) and Pdgfrα⁺ myofibrogenic progenitor cells functions as an alveolar niche, maintaining the growth and self-renewal of alveolar progenitors in vivo (Zepp et al., 2017). Consistent with what was found in these studies, when co-cultured with fetal Pdgfrα⁺ fibroblasts, alveolar organoids were generated and they contained differentiated AECs and bipotent progenitor-like cells. When organoids were co-cultured with Mark1 KO fibroblasts, (1) organoid expansion was decreased, (2) AECI flattening was impaired, and (3) levels of AEC marker expression were not changed. Therefore, the organoid phenotypes could recapitulate the phenotypes observed in Mark1 KO lung in that distal lung sacculation and AECI flattening were impaired without any defect in pneumocyte differentiation. In addition, this system has also some technical advantages. As organoids align two-dimensionally in this system, their size and morphology can be observed in the same plane for imaging, as previously described (Barkauskas et al., 2017; Rock et al., 2011). Furthermore, fibroblasts and organoids can be transduced with lentiviral vectors separately and dissociated by dispase treatment. Thus, this system would be widely applicable to any organoids from diverse tissues. Since Pdgfrβ⁺ cells also express Mark1, they might contribute to type I collagen production in the lung mesenchyme. We have isolated Pdgfrβ⁺ cells by using MACS, and the effect on organoid formation was analyzed by the same co-culture system used in this study. However, since, in our hands, the proliferative ability of Pdgfrβ⁺ cells in culture was low, their functional roles remain to be determined at present.

Fetal breathing movement is observed around E15–E16 (Buckley et al., 2005) and the surgical transection of fetal cervical spinal cord or drainage of lung fluid in fetal sheep suppresses lung maturation and growth (Kitterman, 1996; Liggins et al., 1981; Moessinger et al., 1990). Recently, it was found that the pressure of amniotic fluid influx by embryonic breathing-like movement mechanically induces the differentiation of AECI (Li et al., 2018). Given the fact that Mark1 is involved in neuronal functions (McDonald, 2014), it is also interesting to speculate that Mark1 may also play a role in distal lung sacculation by controlling the fetal breathing-like movement.

A mathematical model showed that the pressure of amniotic fluid influx does not induce cell flattening when cell shape change is not feasible, suggesting that cell flattening is regulated as a distinct parameter. However, for sacculation, it was predicted that cell flattening is not sufficient, and that the inside pressure induced by amniotic fluid influx is required. Furthermore, when mesenchymal parameters were adjusted, it was unexpectedly revealed that lowered mesenchymal motility results in decreased sacculation. Given the fact that cell adhesiveness, motility and polarity are positively correlated with extracellular matrix stiffness (Ahmadzadeh et al., 2017; DiMilla et al., 1991; Koka et al., 2003; Wolf et al., 2013; Zaman et al., 2006), it is assumed that a decreased level of type I collagen results in impaired mesenchymal cell motility and therefore the mesenchyme may become an obstacle to distal lung sacculation. Thus, it is predicted that net mesenchymal motility may be required for efficient distal lung sacculation. Live imaging of the fetal lung mesenchyme would elucidate how mesenchymal motility are coordinated with distal lung sacculation.

Protocols used for all animal experiments in this study were approved by the Animal Research Committee of Osaka University, Japan (No. 21-048-1). All animal experiments were carried out according to guidelines for the care and use of experimental animals at Osaka University.

Dilutions used and full details of all antibodies used in this study are listed in Table S1. Mark1 (ab15437), proSP-C (ab3786), type I collagen (ab34710), and Ki-67 (ab16667) antibodies were purchased from Abcam (Cambridge, UK). Pdgfrα (R&D Systems, AF1062), Pdgfrβ (eBioscience, APB5), E-cadherin (BD Biosciences, 610181), podoplanin (Novus, 8.1.1), Zo-1 (Merck, R40.76) and acetylated tubulin (Sigma-Aldrich, T7451) antibodies were also used. Biotinylated EpCAM (118203) and Pdgfrα (135910) antibodies were from BioLegend. Streptavidin MicroBeads were from Miltenyi Biotec (130-048-102).

Mark1 mutant mice were generated using a triple CRISPR method as previously described (Fumoto et al., 2017; Sunagawa et al., 2016). F0 mutant mice at E17.5 were analyzed in Fig. S2. Adult Mark1 mutant mice had airspace enlargement (data not shown) but were viable adults. Therefore, Mark1 mutant mice were established and used throughout this study.

Quantitative real-time (RT)-PCR was performed according to the manufacturer's protocol. Primers used for this analysis are summarized in Table S2.

Lungs were dissected from E17.5 ICR embryos (Slc:ICR, originally from Charles River Laboratories), minced in Hank's buffered saline solution (HBSS; Thermo Fisher Scientific, Waltham, MA) and incubated in phosphate-buffered saline (PBS) containing 5 mM EDTA and 10% fetal bovine serum (FBS) for 15 min at 37°C. Lungs were then collected by centrifugation (800 g for 5 min) and incubated in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS, collagenase D (Sigma, COLLD-RO), dispase (Thermo Fisher Scientific, 17105041) and DNase I (Sigma, D4527) for 30 min at 37°C. Lung cells were dissociated by passing through a 1-ml disposable pipet and collected by centrifugation (800 g for 5 min). Cells were then incubated with red blood cell (RBC) lysis buffer (BioLegend, 420301) for 5 min at room temperature and washed once with FACS buffer [PBS containing 0.5% bovine serum albumin (BSA) and 2 mM EDTA]. EpCAM⁺ epithelial cells and Pdgfrα⁺ fibroblasts were isolated using magnet-assisted cell sorting (MACS) technology according to the manufacturer's protocol (Miltenyi Biotec). Briefly, total lung cells (10⁷ cells) were collected and incubated with biotinylated antibodies for 20 min on ice. Cells were then washed once and Streptavidin MicroBeads were added (Miltenyi Biotech, 130-048-102), and then the mixture was incubated for 15 min. Antibody-labeled cells were captured by using an LS column (Miltenyi Biotech, 130-042-401) and collected. Cells were frozen in Recovery™ cell culture freezing medium (Thermo Fisher Scientific, 12648010).

Twenty-four plates were coated with type I collagen solution (Advanced BioMatrix, Carlsbad, CA; 5005) and 10⁵ Pdgfrα⁺ fibroblasts were seeded. At 12 h after seeding, 100% Matrigel (BD, 354230) was added to cover the wells entirely and these were allowed to solidify for 20 min at 37°C. Then, 10⁵ EpCAM⁺ epithelial cells were suspended in organoid medium [1:1 mix of DMEM and Ham's F-12 medium, containing 25 ng/ml EGF (R&D, 236-Eg), 50 ng/ml RSpo1 (R&D, 7150-RS), 50 ng/ml mouse noggin (R&D, 1967-NG), 50 ng/ml FGF10 (PeproTech, Rehovot, Israel; 100-26), 1× B-27 supplement (Thermo Fisher Scientific, 17504044), 1× N-2 supplement (Thermo Fisher Scientific, 17502048), 1× Glutamax (Thermo Fisher Scientific, 35050061) and 10 mM HEPES–NaOH pH 7.4] supplemented with 2% Matrigel and seeded on top of solidified Matrigel.

Immunostaining was performed as described previously with modifications (Fumoto et al., 2017). Briefly, freshly frozen embryonic lungs were sectioned at 8 µm, fixed in 4% paraformaldehyde (PFA), and blocked in blocking buffer (PBS containing 1% BSA and 0.3% Triton X-100). Samples were incubated with antibodies diluted in blocking buffer for 1 h or overnight, washed three times, and then stained for 1 h with secondary antibody diluted in blocking buffer. After washing, samples were covered with PBS containing 50% glycerol. When organoids were stained, they were isolated from Matrigel using dispase (BD Biosciences, San Jose, CA) for 30 min at 37°C and fixed with 4% PFA at room temperature for 20 min. After blocking, organoids were stained as described above. When fibroblasts were stained, they were seeded on type I collagen-coated coverslips. All images were observed using an LSM 880 laser scanning microscope (Carl Zeiss Microscopy Co., Ltd, Jena, Germany).

At 5 days after co-culture, organoids and Matrigel were removed by incubating with dispase. Then, fibroblasts remaining on coverslips were fixed with 4% PFA and stained with anti-acetylated tubulin antibody to visualize cilia. Cilia length was measured using ImageJ software and is presented in a beeswarm plot.

Lentivirus production and infection were performed as described previously (Fumoto et al., 2017). MISSION TRC small hairpin (sh)RNAs (shMark1, TRCN0000024173; shGli1: TRCN0000054884; shKif7, TRCN0000090438; shDdr1, TRCN0000023369; Sigma-Aldrich) were used. shLuc was constructed by cloning a cDNA oligo into a pLKO.1 vector (Addgene 10878). The target sequence is 5′-CGCTGAGTACTTCGAAATGTC-3′.

A small cyst-like geometry comprised of 22 cells was considered and no proliferation assumed. The optimal perimeter, and the basal and apical lengths were set as c̅ = 6, lb = 2, and la = 0.5. The cellular height was controlled so that the perimeter of the cell was sustained in the model. Therefore, large la and lb values result in cell flattening. In Fig. 6D, la was increased to 2.5 in 20,000 timesteps, and lb was increased to 3 in 10,000 timesteps.

Experiments were performed at least three times and results are expressed as the mean or mean±s.d. Statistical analysis was performed using a paired Student's t-test and Mann–Whitney U-test. For each quantification, at least 50 cells from at least three sections of lung epithelia were counted per experiment. P-values of less than 0.05 were considered statistically significant.

We thank Dr M. Morimoto (Riken, Japan) for constructive discussions, and also thank Mr Katashi Deguchi for his assistance in the force measurement in atomic force microscopy experiments shown in Fig. S3D.