NDST1-dependent heparan sulfate regulates BMP signaling and internalization in lung development

Lung epithelium and mesenchyme generate essential secreted proteins for each other and thus coordinate lung embryonic morphogenesis. Multiple factors, including bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs) and hedgehog are reportedly involved in lung formation. Inhibition of BMP signaling with the BMP antagonist, noggin (NOGG) or dominant-negative BMP receptor (dnAlk6) alters expression of FOXJ1, uteroglobin (UTER/CC10) and pulmonary surfactant-associated protein C (PSPC/SFTPC) and causes a severe reduction in distal lung epithelial cell types and an increase in proximal cell types (Weaver et al., 1999). Ectopic expression of gremlin (GREM1), another antagonist of BMP, results in disruption of the proximal-distal pattern in embryonic lung (Lu et al., 2001). In addition, deletion of BMPR1a or BMP4 in mouse lung epithelium leads to reduction in number of type II pneumocyte and a decrease in epithelial proliferation (Eblaghie et al., 2006). Misexpression of BMP4 also causes a decrease of type II cells and inhibition of epithelial proliferation, along with cell death in the mesenchyme during lung development (Bellusci et al., 1996). Abnormal septa in alveolar were observed in mice with deletions of FGFR3 and FGFR4 (Weinstein et al., 1998). Overexpression of Sonic hedgehog leads to an abundance of mesenchyme and loss of typical alveoli (Bellusci et al., 1997). Additional signals such as EGF, Wnt and TGFβ are also important in lung development.

A number of signaling pathways including BMP, FGF, hedgehog and Wnt depend on heparan sulfate proteoglycans (HSPGs) (Hacker et al., 2005; Lin, 2004). HSPGs are macromolecules composed of heparan sulfate glycosaminoglycan (GAG) side chains covalently bound to core proteins. Biosynthesis of heparan sulfate (HS) is initiated from a chain composed of repeated D-glucuronic acid (GlcUA) N-acetyl-D-glucosamine (GlcNAc) residues. The glucosaminoglycan chains first undergo N-deacetylation and N-sulfation of selected GlcNAc residues by GlcNAc N-deacetylase/N-sulfotransferase (NDST) (Lindahl et al., 1998; Salmivirta et al., 1996). The subsequent modifications, such as C-5 epimerization of GlcA to iduronic acid (IdoA) and O-sulfation at various positions, are dependent on the prior N-sulfation of GlcN units created by NDST (Esko and Selleck, 2002; Lindahl et al., 1998). NDST has a key role in the modification of the HS polysaccharide chain.

Genes encoding four known NDST isozymes, Ndst1-Ndst4, have been identified in mammals (Aikawa and Esko, 1999; Aikawa et al., 2001; Eriksson et al., 1994; Hashimoto et al., 1992; Kusche-Gullberg et al., 1998). Ndst1 and Ndst2 are expressed ubiquitously in both embryonic and adult mice, whereas Ndst3 and Ndst4 are mostly expressed during embryonic development (Ford-Perriss et al., 2002; Grobe et al., 2002; Miettinen et al., 1997; Yabe et al., 2005). Mice lacking Ndst2 have abnormal mast cells without properly sulphated heparin and mast-cell proteases (Forsberg et al., 1999; Humphries et al., 1999), and those lacking Ndst3 have no obvious phenotype (Grobe et al., 2002). However, disruption of Ndst1 results in severe malformations in lung, brain, cranial facial, lens, vascular, skeletal, and lacrimal gland development (Abramsson et al., 2007; Fan et al., 2000; Grobe et al., 2005; Hu et al., 2007; Pan et al., 2008; Pan et al., 2006; Ringvall et al., 2000). Thus, NDST1 is an essential NDST isozyme in mouse embryonic development.

Previously we reported that Ndst1 mutant mice developed atelectasis and respiratory distress and died shortly after birth (Fan et al., 2000). However, the detailed developmental defects and the underlying mechanisms of NDST1-depedent HS modulation of signaling pathways remain unclear. Here, we report that BMP signaling is affected in Ndst1 mutant lung, which could be one of the causes of defective lung development.

Mouse lung arises from the laryngotracheal groove at 9.5 days post coitum (d.p.c.). Terminal sacs and vascularization develop in the period of 16.5-17.5 d.p.c. After 17.5 d.p.c., the number of terminal sacs and vascularization increase and type I and type II cells differentiate (Warburton et al., 2000). Previous studies have demonstrated that Ndst1-null mice develop pulmonary hypoplasia and neonatal respiratory distress (Fan et al., 2000; Ringvall et al., 2000). To further characterize the phenotype of the mutants, histological examination of embryonic lung development in mutant mice was performed by hemotoxylin and eosin staining lung sections of mice at 16.5 and 18.5 d.p.c. At 16.5 d.p.c., the terminal sacs were less dilated in lungs of Ndst1 mutants than those in the wild type (Fig. 1A,B,E,F). The mesenchyma in 16.5 d.p.c. mutant lungs was also thicker than that in their wild-type littermates (Fig. 1A,B). At 18.5 d.p.c., mutant lungs exhibited less dilated sacs and thicker septa compared with wild-type lungs (Fig. 1C,D,G,H). Furthermore, BrdU labeling of 16.5 and 18.5 d.p.c. embryos indicated that mutant lungs had many more proliferative cells than normal littermates (Fig. 1I-K), consistent with the observation that mesenchyma and septa were thicker in mutant lungs than in wild-type lungs.

Although NDST1 is essential for lung morphogenesis, it is not the only NDST protein expressed during lung development. Lung samples from Ndst1 mutants and normal littermates were examined for expression levels of NDST genes by RT-PCR. At 17.5 d.p.c., all four NDST isozymes were expressed in normal lungs. In the Ndst1^–/– lung, expressions of Ndst2, Ndst3 and Ndst4 mRNA transcripts were upregulated (Fig. 1L), suggesting a potential redundant effect among the NDST enzymes. Because it is considered to be the less prevalent phenotype in Ndst2 or Ndst3 mutant mice (Forsberg et al., 1999; Grobe et al., 2002; Humphries et al., 1999), NDST1 might be the most important isoform in the NDST family. The phenotype observed in Ndst1^–/– mice might form as a result of lack of NDST1 function that cannot be compensated by other members of this enzyme family.

Neonatal respiratory distress observed in Ndst1 mutants might be caused by reduced production of surfactant proteins (Fan et al., 2000). Immunostaining analyses revealed a striking reduction in expression of two surfactant proteins, SFTPC and SFTPA in lungs of Ndst1-knockout mice. At 16.5 d.p.c., staining of SFTPC was specifically detected in the distal epithelium (Fig. 2C), whereas staining of SFTPA was detected throughout the epithelium, including proximal and distal parts, in lungs of wild-type mice (Fig. 2A). Significantly, both proteins were barely detectable in mutant lungs (Fig. 2B,D). Furthermore, analysis of the mutant lungs at 18.5 d.p.c. revealed a notable decrease of SFTPC-positive cells (Fig. 2E-G). In parallel, real-time RT-PCR assay indicated that mRNA levels of Sftpc, Sftpa and Sftpb were significantly reduced in mutant lung at 17.5 d.p.c. (Fig. 2Y). These observations imply that inactivation of Ndst1 leads to defective development of distal epithelium and immaturity of type II alveolar cells, which fail to produce surfactant proteins.

Type I alveolar cells were also found to be immature in Ndst1-null mice. From 17.5 d.p.c., type I alveolar cells arise from their precursor cells. Both cell types characteristically express aquaporin-5 (AQP5), a water channel protein (Krane et al., 1999; Kreda et al., 2001). Immunofluorescent examination of 16.5 and 18.5 d.p.c. mutant mouse lungs indicated that the number of AQP5-positive cells was significantly reduced in Ndst1-null mice (Fig. 2H-L). Consistently, the mRNA level of Aqp5 was also reduced in lungs of 17.5 d.p.c. mutant animals (Fig. 2Y).

The expression pattern of the proximal bronchiole epithelium marker CC10 (UTER/Clara Cell 10 protein), did not alter between Ndst1 mutant and control lungs (Fig. 2M,N). Immunofluorescent examination revealed that the number of CC10-immunoreactive cells did not change in mutant lungs (Fig. 2O-Q). However, the bronchioles lining with these cells were less dilated in mutant lungs than in normal littermates (Fig. 2O,P), suggesting that development of proximal bronchiole epithelium was affected by Ndst1 inactivation.

There is a close relationship between blood vessel and lung structural development. Immunofluorescent examination showed that the mutant lungs had no significant alteration in distributions of caveolin-1 (CAV1) and α-smooth muscle actin (SMA), which are molecular markers of blood vessels (Fig. 2R-V for caveolin-1; Fig. 2W,X for SMA). It appears that inactivation of Ndst1 does not affect the lung blood vessel development.

The embryonic lung in Grem1^–/– mice exhibits an abnormal `proximalized' phenotype, which is caused by BMP-signaling dysregulation (Michos et al., 2004). Thus, BMP signaling was examined in Ndst1^–/– lungs. Binding of BMPs to preformed heteromeric receptor complexes results in the phosphorylation of Smad proteins, and subsequent stimulation of expression of Id1 (Hollnagel et al., 1999; Nohe et al., 2002), Dlx5 (Holleville et al., 2003; Miyama et al., 1999) and Tbx1 (Bachiller et al., 2003). Phosphorylated Smad1 (Smad1-P) was highly upregulated in lungs of 16.5 d.p.c. mutant mice (Fig. 3A,B), and upregulation of ID1 protein was evident at 18.5 d.p.c. (Fig. 3E,F). Interestingly, overexpressed ID1 in mutant lungs seemed to localize not only in the nucleus, but also in the cytoplasm. A similar phenomenon was observed in small cell lung cancer with upregulated ID1, although the reason for this was not clear (Kamalian et al., 2008). mRNA levels of both Dlx5 and Tbx1 were also increased in lungs of Ndst1-deficient animals (Fig. 3I). All these data demonstrate that the BMP-signaling pathway is upregulated in Ndst1^–/– lungs. However, mRNA levels of Bmp2, Bmp4, Bmp5 and Bmp7 were not changed in Ndst1 mutant lungs (Fig. 3J), indicating that the upregulation of BMP signaling in the mutants was not caused by an increased BMPs.

To determine whether dysregulated BMP signaling caused the defective differentiation of type I and type II cells in Ndst1-null lungs, Noggin, a BMP antagonist, was applied to block BMP signaling. Similarly to wild-type lung explants (Fig. 4A-H), treatment of 15.5 d.p.c. Ndst1^–/– lungs with noggin resulted in significantly downregulated expression of Smad1-P protein (Fig. 4K,L), and upregulation of SFTPC (Fig. 4M,N) and AQP5 (Fig. 4O,P). Furthermore, BrdU labeling indicated that cell proliferation was decreased in the presence of noggin (Fig. 4I,J). These results demonstrate that block of BMP signaling could rescue the developmental failure in type I and type II cells, or their precursor cells, in mutant lungs. And it reinforced the idea that upregulation of BMP signaling contributes to the defective lung development in Ndst1 mutants. However, exogenous noggin inhibited the proliferation of both wild-type and mutant lungs (Fig. 4R), indicating that a physiological concentration of BMP is essential for cellular processes, including DNA synthesis and mitosis.

Since NDST1 catalyzes the first modification step in biosynthesis of HS and the HS structure in most basement membrane is affected in Ndst1^–/– mice, the HS chain in mutant lungs is therefore probably affected. Thus, endogenous HS in wild-type and mutant lungs was detected using an antibody that reacts with O-sulfated N-acetylated glucosamine residues of HSPGs (Fig. 5A-D). Wild-type lungs showed a strong signal, whereas no signal was detected in mutant lungs. This demonstrates that loss of NDST1 causes failure in the synthesis of normal HS chains, which is consistent with a previous report in the Ndst-1^–/– lens (Pan et al., 2006). BMPs are reportedly involved in lung morphogenesis, and bind to heparin. To investigate how the BMP-signaling pathway is regulated in Ndst1 mutant lungs, histochemical assays were performed to test the interaction between secreted proteins and endogenous HS in mutant lungs. At 18.5 d.p.c., the binding of BMP2 and BMP4 was less in mutant lungs than that in normal littermates (Fig. 5E-H), suggesting that the capacity of the secreted BMP proteins to bind to HS was decreased in Ndst1^–/– lungs. Pre-treatment of lung sections with heparitinase greatly reduced the binding in Ndst1-null mice and wild-type littermates (Fig. 5I,J), indicating that the binding of these secreted proteins is indeed HS dependent.

Based on our data and previous reports (Ruppert et al., 1996; Fisher et al., 2006; Jiao et al., 2007), HS seems to have an inhibitory role in the BMP-signaling pathway. Thus HS-dependent BMP binding appears to be distinct from receptor-dependent BMP binding. To study the function of HS-dependent BMP binding to the cell surface, BMP internalization was monitored in lung epithelial cells. In cultured wild-type lung epithelial cells, heparitinase treatment significantly reduced BMP binding to the cell surface and its consequent internalization (Fig. 6A,C,M,O,U). This demonstrates that HS-dependent binding of BMP is essential for BMP internalization.

To test whether BMP receptors are involved in BMP internalization, noggin, which inhibits BMP signaling by binding to BMPs and preventing their interaction with receptors (Smith and Harland, 1992; Zimmerman et al., 1996), was applied to internalization assays. BMP2 binding to the cell surface and internalization in normal and Ndst1 mutant lung epithelium were not changed in the presence of noggin (Fig. 6E,G). It is conceivable that BMP internalization occurs mainly via binding to HS chains rather than binding to BMP receptors. In Ndst1^–/– lung cells, the binding of BMPs and consequent internalization was also markedly attenuated (Fig. 6B,D,F,H,U), indicating the necessity of NDST1-dependent HS modification during this process. Similar results were obtained with mesenchymal cells (Fig. 6Q-T).

It was unexpected that exogenous heparin could largely rescue BMP2 binding to the cell surface and consequent internalization in Ndst1^–/– mice (Fig. 6J,L,U), whereas its effect on wild-type cells was much less significant (Fig. 6I,K,U). It seems that not only cell surface HS, but also extracellular HS, participates in the internalization of BMP.

Since exogenous heparin could rescue the BMP binding and internalization, it might be also able to rescue the dysregulated BMP signalling and consequent morphogenetic defects in Ndst1^–/– lungs. Treatment of 15.5 d.p.c. wild-type and Ndst1^–/– lung explants with 10 μg/ml exogenous heparin, which has a higher content of N-sulfation than HS, reduced the expression of Smad1-P (Fig. 7A,B,G,H) and enhanced the expression of SFTPC and AQP5 (Fig. 7C-F,I-L). The findings confirm that impaired BMP signaling with N-sulfation or heparin can facilitate the differentiation of type I and type II alveolar cells.

Consistently, in wild-type lungs, BMP4 and heparin together led to the decreased expression of Smad1-P (Fig. 7M,N) and increased expression of SFTPC, compared with that in lungs treated with BMP4 alone (red in Fig. 7O,P), but did not affect the expression of caveolin-1 (Fig. 7O,P, green).

Western blot assays were also performed. Similarly to results of the histochemical assays, Ndst1^–/– lung explants displayed a much higher level of phosphorylation of Smad1 than wild-type lung explants, whereas treatment with noggin or heparin dramatically reduced this abnormal high level (Fig. 7Q). This confirmed the inhibitory function of heparin in BMP signaling.

Here, we found that loss of Ndst1 in lung results in defective BMP signaling even with the redundantly enhanced expression of Ndst2, Ndst3 and Ndst4, whereas FGF signalling, but not hedgehog signalling, was also affected in mutant lungs (supplementary material Figs S1 and S2). Ndst1-null mice display severe brain and facial defects, which might be consistent with impaired sonic hedgehog (Shh) and FGF interaction with mutant HS (Grobe et al., 2005) in some mutants. It is also possible that only part of NDST1's function could be compensated by other NDST isozymes in lung development. For instance, it is possible that the hedgehog signaling in lung was compensated. Although multiple abnormalities previously described in Ndst1^–/– mice were not observed in Ndst2 or Ndst3 mutant mice, it is likely that other isoforms of the NDST enzyme family might have compensated for the loss of Ndst2 or Ndst3. Considering that NDST1 modulates FGF signaling, but not BMP and Wnt during lens development (Abramsson et al., 2007; Pan et al., 2006), compensation between NDSTs seemed to be tissue dependent. Moreover, in lungs, we found that blood vessel formation was not affected by loss of NDST1 whereas differentiation of lung epithelium cells was. This implies that the compensation might even be cell-type dependent.

Here, we propose that dysregulation of BMP signaling pathway contributes to the defective morphogenesis in Ndst1-null lung. Consistently, similar abnormal septation of the lung airway epithelium was observed in embryonic lung when GREM1, an antagonist of BMP, was deficient (Michos et al., 2004). By contrast, overexpression of BMP4 in embryonic distal lung epithelium results in thicker mesenchyma than observed in control and in the `emphysematous' phenotype (Bellusci et al., 1996). Nog^–/– mice exhibit abnormal morphology with a malformed and truncated lobe (Weaver et al., 2003). There are several possible explanations for these discrepancies. First, endogenous GREM1 is expressed in proximal airway epithelium (Weaver et al., 1999), whereas noggin is normally expressed in the distal mesenchyme (Lu et al., 2001). By inhibition of BMP signaling, GREM1 and noggin might have different roles in spatial and temporal regulation of lung development. Thus, the phenotypes observed in these mice models result from deficiencies in partial epithelium or mesenchyme differentiation alone. Similarly, ectopic expression of BMP4 in distal epithelium had no obvious affect on the epithelium in the proximal airway (Bellusci et al., 1996). HS plays an upstream regulatory role in the BMP-signaling pathway, and thereby modulates the differentiation of all the epithelium and mesenchyme. To determine how HS does this, further conditional gene knockout studies are needed. In addition, depending on concentration, BMP promotes mesenchyme proliferation or death (Bellusci et al., 1996). In Bmp4 transgenic mice, a high level of BMP4 resulted in cell death and inhibition of cell proliferation. However, in our model and in GREM1-deficient mice, cell proliferation increased, probably because of the occurrence of endogenous BMP at proper concentrations. Finally, HS modulates not only BMP signaling, but also other important pathways during lung development, including the FGF-signaling pathway (supplementary material Fig. S1). Thus, the phenotype we report here might result from dysregulation of several signaling pathways.

Accordingly, we asked how NDST1-dependent HS was involved in the BMP-signaling pathway. It is reported that HS binds to the N-terminal of BMP-2 and modulates the function of BMP2 in chick limb bud assays (Ruppert et al., 1996). In C2C12 myoblast cells, blockade of cell surface HSPG sulfation or removal of the GAG chain enhanced BMP2 signaling and bioactivity, which could be attenuated by exogenous heparin (Jiao et al., 2007). In this study, we further proved that the binding of HS to BMP inhibits BMP signalling, rather than facilitates it, in lung development. In particular, NDST1-dependent modification is essential for the activity of HS in the regulation of BMP-signaling pathways. In addition to regulating BMP binding, HS also has a role in binding noggin to the cell surface. Therefore, reduced noggin binding in Ndst1 mutant lungs might also contribute to the hyperactivity of the BMP-signaling pathway (Viviano et al., 2004).

Recent studies have implied that cell surface HSPGs are also involved in cellular internalization of proteins (Belting, 2003; Jiao et al., 2007; Payne et al., 2007). Live-cell imaging indicates that HSPGs mediate cationic ligand internalization via a clathrin- and caveolin-independent, but flotillin- and dynamin-dependent pathway (Payne et al., 2007). Our data further demonstrate that the NDST1-dependent HS modifications, including N-deacetylation and N-sulfation, are required for BMP internalization. We also observed that HS-mediated BMP internalization might be independent of BMP receptors (Fig. 6C,G). More interestingly, we found that exogenous heparin could partly rescue the BMP internalization defects caused by loss of NDST1 (Fig. 6J,L). This implies that, not only cell-surface HS, but also extracellular heparin or HS, also contribute to BMP internalization, which has not been reported before.

Taken together, it is postulated that BMPs that bind to HS prefer to internalise, whereas BMPs that bind to receptors prefer to activate downstream signaling. The balance between BMP binding to HS and to receptors might control BMP signaling.

Compared with our previous work on Ndst1 mutant bone (Hu et al., 2007), the activity of HS in modulating signaling pathways seems variable in different tissues. First, unlike in embryonic lung, exogenous heparin might not be able to rescue the morphogenetic defects in Ndst1^–/– embryonic limb bone. Second, FGF1 binding to the cell surface is attenuated in mutant lung but not in mutant bone. It is conceivable that other factors participate in the process of HS binding to cytokines.

In summary, previous studies on the Ndst1-knockout mouse indicate that HS could regulate various secreted ligands in different tissue contexts. Therefore, the tissue-specific regulation of HS and secreted ligands might contribute to the observation that HS plays a vital role in different developmental tissues via various signaling pathways. In this study, we provide a mouse model for exploring the function of HS in BMP signaling. We point out the different roles of HS in regulating BMP signaling and BMP internalization, and thus present an explanation for the mechanistic involvement of NDST1-dependent HS in modulation of the BMP-signaling pathway during lung development,

In addition to insights into developmental regulation, our study also has important clinical implications. We found an inhibitory effect of heparin on the modulation of BMP signalling, which might provide an explanation for the clinical observations that heparin can improve outcome in small-cell lung cancer (SCLC) (Altinbas et al., 2004; Lebeau et al., 1994), in which BMP8 is overexpressed (Henderson et al., 2005).

The generation of the Ndst1-deficient mice and molecular examination by PCR to distinguish wild-type and mutant Ndst1 alleles have been reported previously (Fan et al., 2000). All mice used in this study were bred and maintained at Shanghai Institute of Biological Sciences under specific pathogen-free conditions in accordance with institutional guidelines.

Embryonic lungs were fixed overnight in 4% buffered formaldehyde at 4°C and embedded in paraffin for sectioning. For histological analysis, 5 μm sections were stained with hematoxylin and eosin, mounted in xylene-based medium and photographed. For immunohistochemistry, 5 μm tissue sections were pretreated with 10 mM sodium citrate buffer (pH 6.0) at 95°C, and then incubated overnight at 4°C with anti-CC10 (T-18), anti-SFTPA (N-19), anti-SFTPC (M-20), anti-AQP5 (G-19), anti-caveolin-1 (N-20), anti-Smad1-P (Ser463-Ser465), anti-Id1 (C-20), anti-Patched (G-19), anti-Gli1 (N-16) (all from Santa Cruz Biotechnology, Santa Cruz, CA), anti-SMA (Sigma, St Louis, MO) or anti-HepSS-1 (Seikagaku Corp., Tokyo, Japan). For enzymatic staining, biotinylated secondary antibodies and the ABC staining system (Santa Cruz) were applied. The images were captured on a cooled CCD camera (SPOT II, Diagnostic Instruments, Sterling Heights, MI) on an Olympus BX51 microscope. For immunofluorescence, secondary antibodies conjugated with appropriate fluorochrome, FITC, Cy3, Cy5 (Jackson ImmunoResearch, West Grove, PA), were used. Digital images were analyses by confocal laser-scanning microscope (Leica SP2 system) and processed using Adobe Photoshop. At least three embryos for each genotype were analyzed for each antibody.

Total RNA was isolated from lungs of 17.5 d.p.c. Ndst1^–/– and wild-type embryos using Trizol Reagent (Invitrogen, Paisley, UK), and then reverse transcribed with Super-scriptase (Invitrogen). The primers used in PCR assays were Ndst1 (forward, 5′-CTG CCC TGG CGT GCC TCC-3′; reverse, 5′-TGG GCC GTG TCA CAT AGA GCA GT-3′); Ndst3 (forward, 5′-TCA CAT GCA GCC CCA CCT CTT-3′; reverse, 5′-GCT CCC CTC CAT GAA TAC TCT TGT-3′); Ndst2, Ndst4 (Pan et al., 2006); Bmp2 (forward, 5′-TCT TCC GGG AAC AGA TAC AGG-3′; reverse, 5′-TCT CCT CTA AAT GGG CCA CTT-3′); Bmp4 (Zhang et al., 2007), β-actin (Actb) (forward, 5′-CTG GCT GGC CGG GAC CTG ACA-3′; reverse, 5′-ACC GCT CGT TGC CAA TAG TGA TGA-3′). Actb was used as an internal control for quantification.

Real-time PCR assays were performed on a DNA Engine Opticon 2 (MJ Research, Watertown, MA) using the DyNAmo SYBR Green qPCR kit (FinnzymesOy, Espoo, Finland). The data were expressed as relative mRNA (gene) copies, which were normalized to the expression level of Actb. The following primers were used: Aqp5 (forward, 5′-AGC CTT ATC CAT TGG CTT GTC-3′; reverse, 5′-TGA GAG GGG CTG AAC CGA T-3′); SftpA (Okubo and Hogan, 2004); SftpB (forward, 5′-ACG TCC TCT GGA AGC CTT CA-3′; reverse, 5′-TGT CTT CTT GGA GCC ACA ACA G-3′); SftpC (forward, 5′-ACC CTG TGT GGA GAG CTA CCA-3′; reverse, 5′-TTT GCG GAG GGT CTT TCC T-3′); Dlx5 (forward, 5′-GTC CCA AGC ATC CGA TCC G-3′; reverse, 5′-GCT TTG CCA TAA GAA GCA GAG G-3′); Tbx1 (forward, 5′-AGG CAG ACG AAT GTT CCC C-3′; reverse, 5′-GCT TGT CAT CTA CGG GCA CA-3′).

Mouse embryos were labeled with BrdU (BrdU labeling and detection kit II, Roche, Germany) by intraperitoneal injection of 10 mM BrdU (1-2 ml per 100 g body weight) into pregnant females 1 hour before sacrifice. Cultured lung explants were treated with 100 μM BrdU for 1.5 hours before harvesting, and then were fixed in 95% ethanol at 4°C and embedded in paraffin. Antibody staining of embryo sections was carried out according to the manufacturer's instructions. The number of BrdU-positive nuclei and total cells in each field of vision was estimated from eight sections from three animals for each genotype.

The assays were performed on paraffin-embedded sections essentially as previously described for cryosections (Chang et al., 2000; Friedl et al., 1997). Briefly, after blocking, sections were incubated with 15 nM BMP2, 15 nM BMP4 and 30 nM FGF1 (all from R&D Systems, Wiesbaden, Germany), respectively. Then, sections were incubated with anti-BMP2, anti-BMP4, anti-FGF1 antibodies (all from R&D Systems) and stained using the ABC staining system (Santa Cruz).

Mouse embryonic lungs were cultured essentially as previously described (Dean et al., 2005; del Moral et al., 2006). Briefly, lungs were isolated from 15.5 d.p.c. crosses of Ndst1^+/– mice. They were placed in a six-well plate on an 8 μm Nucleopore membrane floating in 1 ml BGJ-B medium (GibcoBRL, Grand Island, NY) with antibiotic/antimycotic (Life Technologies, Paisley, UK) and 0.1% BSA, and were maintained at 37°C in a humidified 5% CO₂ incubator. Lung explants were cultured in medium supplemented with 10 μg/ml heparin (Sigma) or with 500 ng/ml noggin (R&D Systems) for 3 days and were compared with explants cultured in BGJ-B medium. The medium was changed every day.

Epithelial and mesenchymal cell cultures were obtained by differential adhesion as previously described (Lebeche et al., 1999). Briefly, whole lungs were dissected at 15.5 d.p.c. and digested with 0.4% dispase 0.8% Collagenase (37°C, 60 minutes; GibcoBRL) to give rise to single cells. The resulting filtered suspension was plated in 30 mm dishes and incubated at 37°C in a humidified 5% CO₂ incubator for 1 hour for differential adhesion. The supernatant containing epithelium cells was removed and centrifuged at 1000 r.p.m. for 10 minutes at room temperature. The cell pellet was resuspended in Dulbecco's modified Eagle's medium (DMEM) with 10% heat-inactivated fetal bovine serum, and plated in dishes. The mesenchymal cells attached to the dish were washed with PBS and cultured with fresh medium.

BMP2 internalization was performed mostly as previously described (Jiao et al., 2007). Briefly, BMP2 (12 μl of 400 ng/ml; R&D systems) was incubated with goat anti-BMP2 antibody (40 μl of 500 ng/ml; Santa Cruz) for 30 minutes at 37°C to form the BMP2-anti-BMP2 complex. Lung epithelial or mesenchymal cells, seeded on glass coverslips in 24-well plates, were incubated with BMP2-anti-BMP2 complex at 4°C for 30 minutes. After a wash with ice-cold serum-free DMEM, the cells were incubated at 37°C for 30 minutes. Cells were then incubated with FITC-conjugated secondary antibody (Jackson ImmunoResearch) and 4,6-diamidino-2-phenylindole (DAPI) and examined by confocal laser-scanning microscope (Leica SP2 system) after being fixed.

Lung explants were homogenized and lysed after treated with noggin and heparin. Then lysates were collected after brief concentration. Immunoblotting was performed as described previously (Huang et al., 2002) with primary antibodies against Smad1-P (Cell Signaling Technology), Smad1 (a kind gift from Yeguang Chen, Tsinghua University, Beijing, China) and β-actin (Sigma).

The Student's t-test was used to determine levels of difference between groups, and P values for significance were set to 0.05. Values for all measurements were expressed as the means ± s.d.