Inhibition of Tgfβ signaling by endogenous retinoic acid is essential for primary lung bud induction

During foregut development, activation of morphogenetic programs in the pre-patterned endoderm gives rise to a number of gut-derived organs, including the lung. In the developing mouse, the thyroid and liver primordia arise at around embryonic day (E) 8.5, whereas the lung and pancreatic primordia appear at E9.5 (Wells and Melton,1999).

Respiratory progenitors (lung and trachea) can be identified in the foregut by E9.0 as a group of endodermal cells posterior to the thyroid that expresses the transcription factor Nkx2.1 (also known as Titf1 - Mouse Genome Informatics) (Minoo et al.,1999). Subsequently, Fgf10, a fibroblast growth factor that is crucial for budding, is expressed locally in the mesoderm adjacent to these Nkx2.1-expressing endodermal cells to trigger Fgf receptor 2b (Fgfr2b)signaling, resulting in primary lung bud formation. Once primary lung buds form, epithelial tubules undergo extensive branching morphogenesis, which ultimately results in the formation of the bronchial tree and the future alveolar region of the lung (Cardoso and Lu, 2006; Shannon and Hyatt,2004).

Lung morphogenesis depends on complex interactions between local signals present in this prespecified foregut endoderm and signals from the adjacent mesoderm (Cardoso and Lu,2006; Shannon and Hyatt,2004). The mechanisms that control gene expression and cellular activities in the lung field of the foregut at the onset of lung development are still poorly understood. Several studies have implicated retinoic acid(RA) signaling as a key regulator of these functions during development of the foregut and its derivatives. Genetic deletion of retinaldehyde dehydrogenase 2(Raldh2; also known as Aldh1a2 - Mouse Genome Informatics),an enzyme essential for RA synthesis in the mouse embryo, results in multiple organ defects and death at around E10.5(Niederreither et al., 1999). RA signals through nuclear receptors Rars and Rxrs (each with isotypesα, β and γ), which are found as heterodimers bound to RA-responsive elements (RAREs) of target genes(Chambon, 1996). These receptors are expressed in the developing lung from its earliest stages(Mollard et al., 2000b);double-null mutant (Rara/Rarb or Rara/Rxrb) mice show several features previously described in vitamin A-deficient animals(Chambon, 1996; Clagett-Dame and DeLuca, 2002; Kastner et al., 1997; Mendelsohn et al., 1994; Wilson et al., 1953). Maternal deficiency of vitamin A results in dramatic abnormalities in the respiratory system of the embryo, which include tracheoesophageal fistula, lung hypoplasia and lung agenesis (Dickman et al.,1997).

In the developing lung, RA synthesis and utilization are most prominent when primary buds are emerging from the primitive foregut(Malpel et al., 2000). Treatment of the whole E8.5 mouse embryo or isolated E8.5 foregut explants with the pan-RA receptor (RAR) antagonist BMS493 completely abrogates development of the lung and the neighboring stomach(Desai et al., 2004; Mollard et al., 2000a). We found that this phenotype results from failure to induce Fgf10expression in the foregut mesoderm at the prospective lung field. The regulation of Fgf10 by RA occurs within a defined developmental window and is not seen in other foregut derivatives (such as thyroid and pancreas), where Fgf10 is also required for normal development(Desai et al., 2004). These observations have been confirmed in Raldh2^-/- mice and vitamin A-deficient rats (Desai et al.,2004; Desai et al.,2006; Wang et al.,2006). Furthermore, we have shown that RA is not required for the initiation of lung endodermal cell fate in the foregut(Desai et al., 2006).

These studies raised the intriguing possibility that at the onset of lung development, RA-responsive genes are selectively activated or repressed in the prospective lung field of the foregut to allow bud formation. It was not known which genes present in the developing foregut could be involved in this process. Here we addressed this problem using oligonucleotide microarray and functional analyses in the models of RA deficiency that we had previously characterized, and at a stage in which lung development is crucially dependent on the RA status of the embryo. We provide evidence of a novel regulatory mechanism implicating RA, transforming growth factor β (Tgfβ) and Fgf10 interactions in primary lung bud induction. Our results suggest that at the onset of lung development, endogenous RA controls Tgfβ signaling in the prospective lung field of the foregut to allow Fgf10 expression and induction of primary lung buds.

Raldh2-null mutants were characterized previously(Niederreither et al., 1999). Raldh2^-/- homozygous embryos were distinguished from their heterozygous and wild-type (WT) littermates by their distinct phenotypic features and by genotyping by PCR(Niederreither et al., 1999). Detection of RA signaling was achieved using the RARElacZ reporter mouse line. These mice carry the bacterial lacZ gene under the control of a heat shock protein promoter (Hsp68) and RAREs from the Rarb promoter (Rossant et al.,1991). X-Gal staining is used to visualize lacZexpression (Malpel et al.,2000). Fgf10lacZ reporter mice have been reported previously (Kelly et al.,2001). In this mutant, a myosin light chain-lacZ(Mlc1v-nlacZ-24) transgene was integrated upstream of Fgf10,which allows control of lacZ expression by Fgf10 regulatory sequences; X-Gal staining reveals sites of Fgf10 expression in the embryo, including the lung (Mailleux et al., 2005). For all experiments, conclusions were based on the evaluation of three independent specimens per culture condition.

The foregut culture system has been reported previously(Desai et al., 2004; Desai et al., 2006). Briefly,timed-pregnant mice were sacrificed at E8.5 and foreguts were isolated from the embryos (8- to 12-somite stage) in phosphate-buffered saline (PBS) using tungsten needles. Extra-embryonic tissues and dorsal structures were removed. Explants were cultured for 1-3 days on 6-well Transwell-Col dishes (Costar)containing 1.5 ml of BGJb medium (Gibco-BRL), 0.3 mg of vitamin C (Sigma) and 10% fetal calf serum (FCS, Gibco-BRL) with or without the specific modulators of RA or Tgfβ signaling (see below). Cultures were shielded from light and incubated at 37°C in 95% air and 5% CO₂. Media were changed daily. Under control conditions, lung buds, stomach and pancreas formed within 24 hours of culture. In some experiments, heparin beads soaked in 100 ng/ml FGF10 (R&D systems) or PBS buffer were grafted onto the foregut after 24 hours of culture.

BMS493 (Bristol Meyers Squibb), a pan-RAR antagonist, dissolved in BGJb medium (10^-6 M) was used to antagonize RAR-dependent signaling, as previously reported in this and other systems(Mollard et al., 2000b; Wendling et al., 2000). All-trans RA (Sigma) dissolved in BGJb medium (10^-7 M) was used to rescue RA signaling in Raldh2^-/- foreguts in culture. Foreguts were treated for 3 days with recombinant human TGFβ1 (5-20 ng/ml, R&D Systems) or a pan-specific TGFβ-blocking antibody (200μg/ml, R&D Systems) dissolved in BGJb medium to activate or inhibit Tgfβ signaling, respectively.

E8.5 WT (control and BMS493-treated, n=3 each) and Raldh2^-/- (control and RA-treated, n=3 each)foreguts were cultured for 24 hours. Total RNA was isolated using the RNeasy Kit (Qiagen), and subjected to amplification, labeling, and fragmentation according to Affymetrix's recommendations. cRNA was hybridized to Affymetrix's Mouse Genome 430 2.0 array chips. Three array chips were used per experimental condition. A single weighted mean expression level for each gene per condition along with a detection P value was calculated using Affymetrix Microarray Suite 5.0 software. Data from each array were scaled to the target intensity of 500 to normalize the results for inter-array comparisons. Quality control parameters of each chip met the acceptable criteria provided by Affymetrix. Genes with detection P values greater than 0.05(considered to be `absent') in all twelve chips were eliminated from the analysis. Gene expression profiles were compared (WT control versus BMS493; Raldh2^-/- control versus RA). The difference in expression of each gene was considered to be significant if the P value was lower than 0.05 (Cyber t-test, http://visitor.ics.uci.edu/genex/cybert). To further increase the specificity of the analysis, only genes whose expression levels were significantly changed in both comparisons were retained on the final list of potential RA targets.

After 24 hours of culture, the heart of the foregut explant was separated from the foregut region and discarded. The protein extract from individual foregut explants was subjected to SDS-PAGE, blotted onto nitrocellulose,washed in TBST [Tris-buffered saline (TBS) with 0.1% Tween 20], blocked with 5% milk in TBST, and incubated with polyclonal antibody (1:400) to phosphorylated Smad2 (pSmad2, Cell Signaling) in TBST with 5% milk overnight. The Immun-Star HRP Chemiluminescent Kit (Bio-Rad) was used for signal development according to the manufacturer's instructions. Total Smad2 (tSmad2,Cell Signaling) was used for normalization.

Whole-mount in situ hybridization (WMISH) of explants and embryos was performed in a 96-well plate as previously described(Lu et al., 2004; Wertz and Herrmann, 2000). Briefly, digoxigenin (DIG)-labeled riboprobes (Maxiscript kit, Ambion) were generated and amplified from total embryonic cDNA (Col1a2, Ctgf, Tgfb2,Tgfbr1, Tgfbr2) or plasmids carrying cDNA for the genes of interest(Tgfb1, Nkx2.1, Sftpc, Fgf10, Tgfb3, Tgfbi). Specimens were rehydrated, digested with proteinase K (Boehringer Mannheim), prehybridized (1 hour, 70°C) in buffer containing 50% formamide, 5×SSC, 1% SDS, 50 mg/ml yeast RNA and heparin followed by overnight hybridization with DIG-labeled RNA probes, and another overnight incubation with anti-DIG alkaline phosphatase conjugate (Boehringer Mannheim) at 4°C. Signal was visualized with BM Purple substrate (Roche Diagnostics). Conclusions were based on the evaluation of at least three independent specimens per probe per condition.

Tissues were fixed in 4% paraformaldehyde in PBS overnight at 4°C,washed twice in TBS with 0.1% Triton X-100 (Sigma), and blocked for 1 hour in blocking buffer (1×TBS with 5% donkey sera, 0.1% Triton X-100). Samples were incubated with primary antibody in blocking buffer overnight [1:150 dilution of rabbit anti-mouse pSmad2 antibody (Cell Signaling), 1 μg/ml sheep anti-mouse Tgfbi antibody (R&D systems)] at 4°C, washed for 1 hour and incubated with secondary antibody overnight (1:750 donkey anti-rabbit antibody conjugated to AF488 and donkey anti-sheep antibody conjugated to AF598, both from Molecular Probes), washed five times for 1 hour each in blocking buffer, then stored in SlowFade Gold Antifade buffer (Molecular Probes) and photographed with a laser confocal microscope.

Mouse neonatal lung mesenchymal (MLg) cells were cultured in DMEM, 10% FCS with or without the specified modulator (all-trans RA or TGFβ1) for 8-24 hours (n=3 100-mm dish plates per condition). Total RNA was isolated(Trizol, Invitrogen), reverse transcribed (1 μg RNA) and amplified by real-time PCR (SYBR Green qPCR Kit, Applied Biosystems). A dissociation curve was used to determine the relative concentration of the single PCR product. 18S RNA was used for normalization.

Cell proliferation was assessed by expression of PCNA protein (PCNA Staining Kit, Zymed). Apoptosis was evaluated by TUNEL (ApopTag Plus,Chemicon). These assays were performed in paraffin sections (5 μm) of foregut explants according to the manufacturers' recommendations. Sections were counterstained with Methyl Green.

To identify RA-dependent pathways potentially involved in lung bud initiation, we generated global transcriptional profiles of foregut explants in which lung formation occurred (RA-sufficient) or was abrogated by disruption of RA signaling (RA-deficient). For this we used both the pharmacologic (BMS493) and the genetic (Raldh2^-/-) models of RA deficiency that we had previously characterized. In the pharmacologic model, RA signaling was antagonized in E8.5 (8- to 12-somite stage) foreguts by treatment with BMS493; this prevented lung bud formation(Desai et al., 2004). In the genetic model, Raldh2^-/- foreguts were similarly cultured in control medium (in which no lung forms) or in medium containing RA(10^-7 M), which allowed lung formation in these mutants(Desai et al., 2006).

Microarray analysis was performed on RNA isolated from these samples (as described in Materials and methods). Modulation of RA signaling was confirmed by altered expression of known RA targets, such as Rarb, Hoxa1 and Hoxb1 (Chambon, 1996; Niederreither et al., 1999). A comprehensive description of these transcriptional profiles will be reported elsewhere. Analysis of these profiles using the EASE (Expression Analysis Systematic Explorer, http://david.abcc.ncifcrf.gov)software revealed a striking over-representation of `Tgfβ signaling pathway-related genes' upregulated in the RA-deficient foreguts(P=0.009). This list was expanded, as we searched for additional Tgfβ targets based on published reports(Table 1, description below).

To determine whether the upregulation of Tgfβ targets in the RA-deficient foreguts reflected an overall hyperactivation of Smad-mediated signaling by Tgfβ1-3, we assessed levels of phosphorylated and total Smad2 in all groups by western blotting. As anticipated, Smad2 phosphorylation was consistently higher in RA-deficient foreguts compared with their respective RA-sufficient controls (Fig. 1A). Thus, we concluded that disruption of RA signaling leads to an abnormal hyperactivity of the Tgfβ pathway in the developing foregut.

The Tgfβ targets upregulated in RA-deficient foreguts encode a diverse group of molecules, and include several mediators of the fibrogenic activities of Tgfβ [connective tissue growth factor (Ctgf), cysteine rich protein 61 (Cyr61), procollagen type I alpha 2 (Col1a2),procollagen type III alpha 1 (Col3a1), procollagen C-endopeptidase enhancer protein (Pcolce), tissue inhibitor of metalloproteinase 1 and 3 (Timp1 and Timp3, respectively)], secreted proteins[biglycan (Bgn), transforming growth factor β induced(Tgfbi), secreted acidic cysteine rich glycoprotein (Sparc),secreted phosphoprotein 1 (Spp1)], cell surface receptors [CD44 antigen (Cd44)], transcription factors [activating transcription factor 3 (Atf3)], colony stimulating factor (Csf1), as well as insulin-like growth factor binding protein 4 (Igfbp4) (see Table 1 for references).

Previous studies and an initial assessment of the expression pattern of these genes at the onset of lung development showed them to be predominantly transcribed in mesodermal tissues (Chuva de Sousa Lopes et al., 2004; Ferguson et al., 2003; Ponticos et al., 2004). The presence of these targets in the mesoderm, where Raldh2 and RARElacZ are also expressed(Malpel et al., 2000),suggests that RA and Tgfβ pathways interact locally in the foregut. Thus,we assessed expression of three representative Tgfβ targets, Tgfbi,Ctgf and Col1a2, in our system.

Tgfbi (also called βig-h3) is a 68 kDa secreted cell-adhesion molecule, known to bind collagen, fibronectin and sulfated glycosaminoglycans,which has been shown to be induced by TGFβ1 in various cells lines(Billings et al., 2002; Skonier et al., 1994). First,we validated Tgfbi as a read out of Tgfβ activation in lung mesoderm-derived cells (MLg) by showing a dose-dependent induction of Tgfbi at 8 hours with recombinant TGFβ1 treatment(Fig. 1B). Then, to investigate the RA-Tgfbi relationship in the foregut, we compared sites of Rar-dependent signaling and Tgfbi transcription by X-Gal staining of RARElacZ reporter mice and WMISH of Tgfbi, respectively. In wild-type (WT) control cultures, Tgfbi was localized to the foregut mesoderm associated with the proximal (stalk) region of lung primordium, where RA signaling was reported by lacZ expression(Fig. 1C-E).

BMS493 disruption of RA signaling abolished RARElacZ expression and resulted in strong Tgfbi signals in the foregut mesoderm,particularly obvious where the lung bud and stomach failed to form(n=6) (Fig. 1G-I). Immunostaining of Tgfbi showed a marked increase of Tgfbi protein locally in the lung field; the broader domain of expression, compared with that of mRNA,suggested accumulation that is likely to be due to a longer half-life of Tgfbi protein. The stronger pSmad2 staining in BMS493-treated foreguts compared with controls, as revealed by confocal image analysis, further suggested an association between increased Tgfbi expression and hyperactivation of Tgfβ signaling (Fig. 1F,J). The negative spatial correlation between Tgfbi and RA-dependent signaling was further supported by comparing the expression of Tgfbi and RARElacZ in E9.0-9.5 embryos. In WT embryos, Tgfbi transcripts were almost undetectable in the RARElacZ-expressing region that encompasses most of the foregut(Fig. 2A,B; see also Fig. 2C, asterisk on the right). This contrasted with the abundant Tgfbi signals observed in the corresponding region of Raldh2^-/- mice in vivo(Fig. 2C, arrowhead on the left), or in Raldh2^-/- foreguts cultured without RA supplementation (Fig. 2D). Although in Raldh2^-/- explants Tgfbi signals were highly induced in both thyroid and lung fields, bud formation and Nkx2.1 expression were disrupted only in the lung field(Fig. 2D,E). Strikingly, Tgfbi signals were dramatically reduced in Raldh2^-/- foregut explants in which exogenous RA rescued budding and expression of Nkx2.1 in the lung field(Fig. 2F,G). Independent evidence that activation of RA signaling (by treatment with RA at 10^-7 M) inhibits Tgfbi expression in lung mesoderm-derived MLg cells (which also respond to TGFβ1) is provided in Fig. 2H. Tgfbi levels are not affected by BMS493 treatment in these cells, as they lack endogenous RA signaling (data not shown).

WMISH analysis of other Tgfβ targets, such as Ctgf and Col1a2, in these explants revealed a similar influence of RA status on their expression pattern. Ctgf is a CCN-family member (along with Cyr61)and an immediate early response gene product induced by Tgfβ(Leask and Abraham, 2003). Col1a2 is one of the subunits of type I collagen, which represents a major structural component of most connective tissues(Olsen et al., 2003). Ctgf signals could not be identified in the WT control foreguts(Fig. 3A), but were dramatically induced in the foregut mesoderm associated with the presumptive lung field of BMS493-treated WT and non-RA-supplemented Raldh2^-/- foreguts(Fig. 3B,C). Rescue of the lung bud in RA-treated Raldh2^-/- cultures was associated with a marked decline in Ctgf expression(Fig. 3D). Analysis of Col1a2 expression showed a similar behavior under the experimental conditions described above (Fig. 3E-H).

We reasoned that if hyperactivation of Tgfβ signaling in the foregut was responsible for the disruption of lung formation in RA-deficient explants,a similar effect should be observed by directly applying Tgfβ protein to RA-sufficient foreguts. TGFβ1 was chosen because of its ability to stimulate Smad2 phosphorylation (Fig. 1A, lane 1) and because of its overall abundance in mesoderm-derived tissues (Schmid et al.,1991). Foregut treatment with 5 ng/ml of recombinant TGFβ1 resulted in only a small endodermal pouch where lungs were expected to develop(n=3, data not shown). At 20 ng/ml, TGFβ1 consistently blocked induction of lung buds and stomach (n=15), reminiscent of the effect seen in BMS493-treated cultures (Fig. 4A,B) (Desai et al.,2004). Tgfβ activation was confirmed by widespread induction of Tgfbi mRNA and protein throughout the explant(Fig. 4C,D). Unlike BMS493,TGFβ1 affected other foregut-derived structures. For example, albumin gene expression, not altered by BMS493 treatment, was inhibited by TGFβ1(Desai et al., 2004) (data not shown). The disruption of lung development was confirmed by the absence of the endodermal markers Nkx2.1 and surfactant associated protein C(Sftpc) transcripts in the presumptive lung field of TGFβ1-treated foreguts (Fig. 4I-L). The presence of abundant proliferating cell nuclear antigen(PCNA) and control-like pattern of TUNEL staining in all layers suggested that TGFβ1 was not inducing toxic effects in these explants(Fig. 4E-H and data not shown).

We asked whether the effects above could be secondary to TGFβ1-mediated disruption of RA signaling. This was ruled out by experiments in which RARElacZ foreguts were shown to maintain strong lacZ signals after 3 hours and 72 hours of culture in TGFβ1-containing medium (Fig. 5A,B).

Based on the reported role of Fgf10 in organogenesis(Min et al., 1998; Sekine et al., 1999), we proposed that the disruption of lung bud formation by TGFβ1 was due to loss of Fgf10 expression in the mesoderm. WMISH analysis confirmed that this was the case, as Fgf10 expression was almost abolished throughout the TGFβ1-treated foreguts(Fig. 5C,D). The effect did not seem to result from loss of Fgf10-expressing mesodermal cells, but rather from downregulation of Fgf10 expression in these cells. Indeed, treating MLg cells with TGFβ1 also resulted in marked downregulation of Fgf10 mRNA (Fig. 5E), in agreement with observations in other systems(Beer et al., 1997; Lebeche et al., 1999; Tomlinson et al., 2004). Interestingly, engrafting a heparin bead soaked in FGF10 protein onto TGFβ1-treated foreguts rescued bud outgrowth and Nkx2.1expression in the foregut endoderm, in spite of the widespread hyperactivation of Tgfβ signaling (Fig. 5F,G). This suggests that the effects of TGFβ1 in the prospective lung endoderm are likely to be secondary to the loss of Fgf10 in the mesoderm, because when Fgf10 is replaced exogenously,Fgfr2b signaling appears to stimulate lung gene expression and growth even in the context of Tgfβ pathway hyperactivation.

Although our functional assays indicated that Tgfβ gain-of-function is incompatible with lung formation, the question remained whether endogenous Tgfβ signaling was required for lung development. Analysis of pSmad2 in the developing embryo suggests that Tgfβ signaling is present diffusely in the foregut and at a very low level (de Sousa Lopes et al., 2003), in agreement with our findings(Fig. 1F). Expression of components of the Tgfβ pathway has been extensively reported in the developing foregut, although these studies have not focused specifically on the lung domain (Millan et al.,1991; Roelen et al.,1994). To address this issue, we performed a comprehensive expression pattern analysis of the Tgfβ subfamily ligands(Tgfb1-3) and receptors (Tgfbr1-2) both in vivo (E8.5-9.5 embryo and E12.5 lung) and in foregut cultures. At E8.5, both Tgfbr1and Tgfbr2 were expressed diffusely in all layers of the foregut(Fig. 6A,E). This diffuse pattern was also seen after 24 hours in culture, but by 72 hours both receptors were predominantly expressed in the foregut endoderm with lower signals in the mesoderm (Fig. 6B,C,F,G,L,M). Strong Tgfbr1/2 signals were detected later in the distal E12.5 lung in vivo(Fig. 6D,H). Interestingly,analysis of Fgf10 expression either by WMISH or by X-Gal staining of an Fgf10lacZ reporter mouse(Kelly et al., 2001; Mailleux et al., 2005) showed that sites where Fgf10 is induced in the lung field are also enriched in Tgfbr1/2 transcripts (Fig. 6I-M). This suggests that Fgf10-expressing mesodermal cells in the lung field are able to respond to Tgfβ ligands (either endogenous or exogenous) and activate Tgfβ signaling. Together with the observations from the previous section, the data further support the idea that Tgfβ signaling in the mesoderm influences Fgf10 expression. Which ligands are expressed in the lung field?Fig. 6N-V confirms previous reports showing that the foregut mesoderm expresses Tgfb1-3 at E8.5 and at 24 hours in culture. Later, Tgfb2 became more restricted to the epithelium (in the E12.5 lung, Fig. 6S); Tgfb3 expression could also be seen in the mesoderm and mesothelial surfaces (see pleura in Fig. 6V).

A number of studies have addressed the role of Tgfβ in vivo and in vitro in lung branching morphogenesis and differentiation. Although the lung does form in mice lacking Tgfb1, 2 or 3, uncertainties remain about how functional redundancy or maternal Tgfβ transfer influence the overall severity of the phenotype(Bartram and Speer, 2004; Cardoso and Lu, 2006; Letterio et al., 1994).

To investigate the role of endogenous Tgfβ signaling in lung formation, first we asked whether we could reliably prevent activation of signaling by all Tgfβ subfamily ligands in our foregut explants. For this, we cultured E8.5 foreguts in medium containing a pan-specific TGFβ-blocking antibody (referred to here as TAb) or control isotype-matched immunoglobulins. Changes in expression of pSmad2 in TAb-treated explants could not be detected, given that pSmad2 signals were already relatively low in untreated WT foreguts. To circumvent this problem,we used WMISH to assess expression of the Tgfβ targets Tgfbi,Col1a2 and Ctgf as `reporters' of Tgfβ activity in these cultures. Tgfbi and Col1a2 transcripts were significantly reduced in WT foregut explants treated with TAb (200 μg/ml)(Fig. 7A-D). Efficient inhibition of Tgfβ signaling was also suggested by the dramatic reduction in the Ctgf levels observed in Raldh2^-/- foreguts cultured in TAb-containing medium (Fig. 7E,F, also compare with WT in Fig. 3A). Analysis of TAb-treated WT foreguts showed that blocking endogenous Tgfβ signaling does not interfere with primary lung bud formation. This was in full agreement with results from Tgfb-knockout mouse models(Bartram and Speer, 2004), and further supported the idea that an excess, but not deficiency, of Tgfβfunction is deleterious to early lung bud morphogenesis.

Our data suggested that the increased activation of the Tgfβ pathway could play a major role in the abrogation of lung development of RA-deficient foreguts. If this is the case, suppressing the overactive Tgfβ signaling in the RA-deficient foreguts should lead to lung bud formation in both Raldh2^-/- and BMS493 models. To test this hypothesis, E8.5 foregut explants were cultured in media containing BMS493 alone, or BMS493 mixed with either isotype-control antibody (CAb) or TAb (200 μg/ml). Nkx2.1 expression was used to identify lung progenitor cells in the foregut endoderm and nascent lung buds. As previously reported, BMS493 cultures failed to induce lung buds or express proper levels of Nkx2.1 (n=13/13) (Desai et al., 2004) (Fig. 8A). By contrast, bud formation and high levels of Nkx2.1expression were consistently observed in BMS493 plus TAb-treated WT foreguts(n=13/13) (Fig. 8B). TAb also seemed to have rescued the formation of the stomach(n=13/13), which is suppressed by BMS493(Desai et al., 2004). Remarkably, suppression of Tgfβ signaling by TAb was also effective in rescuing bud formation and Nkx2.1 expression in the lung field of Raldh2^-/- foreguts (n=11/13)(Fig. 8C,D). Interestingly, in none of the TAb-supplemented cultures (either from the BMS493 or the Raldh2^-/- model) did the rescue of lung buds occur bilaterally. Whether the right or the left lung was rescued was not determined. Presumably, it was the right lung, because other RA-deficient models (Rara/Rarb-null mice, vitamin A-deficient rats, and Raldh2^-/- embryos rescued with low RA doses)characteristically show left lung agenesis and right lung hypoplasia(Mendelsohn et al., 1994; Wang et al., 2006; Wilson et al., 1953).

How could TAb prevent the total lung agenesis characteristic of the RA-deficient foreguts? WMISH assessment of Fgf10 expression confirmed that this gene is selectively downregulated in the lung field of foregut cultures treated with BMS493 alone, as previously reported(Desai et al., 2004). The local disruption of Fgf10 by BMS493 contrasted with the strong Fgf10 expression associated with the rescued lung bud in BMS493 plus TAb-treated foreguts (Fig. 8E,F). Nevertheless, the distribution of Fgf10 mRNA in the lung field in these cultures was overall more diffuse than in control foreguts (compare with Fig. 5C). This is likely to have contributed to the prevention of full rescue of the lung phenotype in BMS493 plus TAb-treated cultures. Endogenous RA may be crucial in establishing local gradients of signaling molecules, such as Fgf10, in the prospective lung field. Our data support a model in which endogenous RA maintains low levels of Tgfβ signaling in the lung field to allow proper expression of Fgf10 and initiation of lung bud morphogenesis (Fig. 9).

In the present study, we provide evidence of a novel mechanism implicating RA-Tgfβ-Fgf10 interactions in initiation of lung morphogenesis. Regulation of Tgfβ signaling by RA has been extensively reported, but data are overall conflicting. In some cell lines, RA increases production of Tgfβ1 and its receptors, leading to growth inhibition(Batova et al., 1992; Danielpour, 1996; Falk et al., 1991; Glick et al., 1989; Imai et al., 1997; Kojima and Rifkin, 1993). RA,however, has also been reported to inhibit endogenous Tgfβ signaling in the mesenchyme of the developing inner ear(Frenz and Liu, 2000) and palate (Yu and Xing, 2006) in mice. Disruption of Rxra in mice results in abnormal upregulation of Tgfb2 in cardiac tissue (Kubalak et al., 2002). A microarray analysis of vitamin A-deficient rat embryos revealed upregulation of some of the genes (such as procollagens)found in our study, suggestive of increased Tgfβ-dependent activity(Flentke et al., 2004). By contrast, in the yolk sac, disruption of RA signaling inhibits Tgfβ-dependent expression of fibronectin and integrin and disrupts visceral endoderm and vascular development(Bohnsack et al., 2004). The discordant data suggest that the way RA and Tgfβ interact is strongly influenced by the local milieu and thus depends on the particular system studied.

Our finding of negative regulation of the Tgfβ pathway by RA was supported by results from multiple approaches in two highly relevant models. These included identification of a number of Tgfβ targets differentially expressed in RA-deficient foreguts by microarray analysis, with further confirmation of their expression pattern, and changes in Smad2 phosphorylation. It was noteworthy that a number of Tgfβ targets involved in synthesis, binding or remodeling of extracellular matrix components, such as Ctgf, Cyr61, Col1a, Col3a1, Pcolce, Timp1 and Timp3, were upregulated in RA-deficient foreguts. We could only speculate, but not prove, that increased matrix deposition and presumably fibrosis could have a negative effect in epithelial-mesenchymal interactions and bud induction. Our study, however,strongly suggests that the underlying reason for disruption of lung formation due to hyperactive Tgfβ signaling is the Tgfβ-mediated downregulation of Fgf10.

Although under RA-deficient conditions our microarray analysis showed upregulation of the Tgfβ signaling transducing elements Tgfbr1and Smad4 (P<0.05 in both), we could not determine precisely how the Tgfβ pathway was influenced by endogenous RA. Presumably, an increase in the amount of Tgfβr1 could increase the number of functional receptor complexes to enhance signaling. Nevertheless, an equivalent increase in expression of Tgfβr2, the receptor that binds to and phosphorylates Tgfβr1 (Bartram and Speer, 2004), was not observed. Moreover, at the mRNA level, none of the Tgfβ ligands was upregulated by loss of RA signaling (data not shown). We considered the possibility that a RA-regulated mechanism leading to conversion of latent Tgfβ ligand into an active form could be enhanced in RA-deficient foreguts. This idea was attractive, as several genes associated with this function were differentially upregulated by RA deficiency in both models, such as annexin A2, cathepsin H, mannose-6-phosphate receptor, matrix metallopeptidase 9 and S100 calcium binding protein A10(Harpel et al., 1993; Krishnan et al., 2004; Ling et al., 2004; Lyons et al., 1988; Nunes et al., 1997; Rifkin et al., 1997; Taipale et al., 1994; Yu and Stamenkovic, 2000; Zhang et al., 2004) (F.C. and W.V.C., unpublished). Using an immunofluorescence protocol to detect latent and active TGFβ1 (Ewan et al.,2002), we could not identify significant differences in staining pattern that could account for the major changes described here. Demonstration of active Tgfβ ligand is technically challenging, even when Tgfβsignaling is evident, in part owing to the short half-life of the active Tgfβ ligand (Araya et al.,2006; Flanders et al.,2001; Wakefield et al.,1990). In our RA-deficient foreguts, it is possible that the activation of Tgfβ signaling would require only minimal increase in the active Tgfβ ligand, as at least one of the Tgfβ receptors is also being upregulated. Further studies will be required to clarify these issues.

We thank Pascal Dollé, Janet Rossant, Robert Kelly and Saverio Bellusci for providing us with the Raldh2^+/-, RARElacZ and Fgf10lacZ mice, respectively; Chris Zusi(Bristol Myers Squibb) for the pan-RAR antagonist BMS493; Parviz Minoo, Andrew McMahon, Nobuyuki Itoh and Harold Moses for cDNA clones; and Chun Li and Xiuzhi Tang for their excellent technical assistance. This work was supported by grants from NIH/NHLBI (F32 HL080846, R01 HL67129 and P01 HL47049) and the GlaxoSmithKline Pulmonary Fellowship Award.