Many of the major organ systems formed during mammalian development result from inductive interactions between epithelial cells of ectodermal or endodermal origin and a loosely packed mesenchyme of mesodermal origin (Wessells, 1977; Hogan, 1999; Perl and Whitsett, 1999; Cardoso, 2000). Both morphogenesis and differentiation of lung endoderm appear to be controlled by its associated mesenchyme (Alescio and Cassini, 1962; Wessells, 1970; Hilfer et al., 1985). The removal of lung mesenchyme, with or without replacement by tracheal mesenchyme, results in the failure of the epithelium to branch (Wessells, 1970). In contrast, bronchial mesenchyme induces the formation of branches in normally nonbranching tracheal epithelium (Alescio and Cassini, 1962). The stimulatory effect of lung mesenchyme can be transmitted across a porous filter, indicating that it is mediated by soluble signaling molecules (Taderera, 1967).
The signals for branching can be separated from those governing growth. Platelet-derived growth factor (PDGF) -B is required for proper growth but not for the branching of the epithelium (Souza et al., 1994). In contrast, specific roles in branching morphogenesis have been suggested for fibroblast growth factor-10 (FGF10; Bellusci et al., 1997), FGF7 (Post et al., 1996), FGF1 (Nogawa and Ito, 1995), hepatocyte growth factor/scatter factor (Ohmichi et al., 1998), PDGF-A (Souza et al., 1995), epidermal growth factor (Warburton et al., 1992; Seth et al., 1993), and the extracellular matrix protein laminin (Schuger et al., 1991). Of these, FGF10 represents the best candidate for an essential branching signal. First, it is expressed in patches in the mesenchyme preceding new branch formation, and the purified protein stimulates both the proliferation and chemotaxis of isolated lung epithelium (Bellusci et al., 1997; Park et al., 1998; Weaver et al., 2000). Second, expression of a dominant negative form of its receptor, the FGF receptor-2 isoform IIIb (FGFR2IIIb), in embryonic lung epithelium completely abolished branching morphogenesis, without preventing epithelial growth (Peters et al., 1994). Finally, mice that carry null mutations in the structural gene of either Fgf10 or Fgfr2 essentially fail to develop lungs (Min et al., 1998; Arman et al., 1999; Sekine et al., 1999; de Moerlooze et al., 2000).
Once a new lung branch forms, its outgrowth is limited by signals from the epithelium. The distal tip epithelium expresses elevated levels of sonic hedgehog (Shh), mouse Sprouty2 (mSpry-2), and Bone Morphogenetic Protein 4 (Bmp4; Bellusci et al., 1996; Tefft et al., 1999). mSpry-2 is the murine homolog of the Drosophila gene sprouty, which encodes an antagonist of FGF signaling (Hacohen et al., 1998). Ectopic overexpression of mSpry-2 results in a reduction of branching, while reducing the levels of mSpry-2 by treating lungs with antisense oligonucleotides results in an enhancement of branching (Tefft et al., 1999; Mailleux et al., 2001). BMP4 also acts to inhibit the proliferation, chemotaxis, and budding of the epithelium (Weaver et al., 2000), whereas SHH signaling appears to inhibit Fgf10 expression (Bellusci et al., 1997; Pepicelli et al., 1998). At least two of these inhibitors, mSpry-2 and Bmp4, are induced in response to FGF10 signaling (Weaver et al., 2000; Mailleux et al., 2001), which indicates that FGF signaling sets up a negative feedback loop that controls the extent of branch outgrowth.
Although the signaling pathways downstream of FGF10 are beginning to be elucidated, the upstream signals that activate Fgf10 expression are not yet understood. Recent studies in Xenopus have demonstrated that Fgf gene expression can be activated by the products of the T-box genes, the T-domain protein transcription factors (reviewed in Papaioannou and Silver, 1998, and Smith, 1999). Both Xbra and Veg-T have been shown to activate eFGF expression (Schulte-Merker and Smith, 1995; Casey et al., 1998; Kofron et al., 1999). Furthermore, Xbra and eFGF form an autoregulatory loop during Xenopus mesoderm formation (Isaacs et al., 1994; Tada et al., 1997). Because several members of the T-box gene family are expressed at high levels in the lung mesenchyme during organogenesis (Chapman et al., 1996), we hypothesized a possible role for these genes in the regulation of Fgf10 expression in the lung. We used an antisense oligonucleotide inhibition strategy to analyze the effects of decreasing T-box gene expression in cultured mouse lungs. The inhibition of two members of the T-box family, Tbx4 and Tbx5, resulted in a decrease in the level of Fgf10 expression in the lung mesenchyme. We also observed a decrease in the degree of branching morphogenesis, consistent with the proposal that FGF10 stimulates the branching of lung epithelium.
Effects of Inhibition of T-box Gene Expression on Lung Branching
Members of two T-box gene subfamilies (Tbx2/Tbx3 and Tbx4/Tbx5) are expressed in embryonic lung mesenchyme (Chapman et al., 1996). To test whether the expression of these genes is essential to mesenchymal function in branching morphogenesis, we established an in vitro culture system with lung buds isolated from 11.5 days postcoitum (dpc) mouse embryos (Fig. 1). We artificially reduced the levels of T-box gene transcripts by treatment with antisense oligodeoxynucleotides (AS ODNs); such ODNs have been shown previously to penetrate embryonic lung mesenchyme and epithelium without nonspecifically inhibiting growth and branching (Souza et al., 1994). As the large number of T-box genes expressed in the lung raised the possibility of functional redundancy, we took advantage of the high degree of sequence identity between members of the subfamilies in the putative DNA binding domain (Agulnik et al., 1996) to design AS ODNs that should hybridize to and block the function of multiple transcripts.
The developmental effects of this specific reduction in Tbx4 and Tbx5 transcript products can be seen in Figures 1 and 2. Lungs exposed to a sense-control ODN (4/5BoxS) exhibit the formation of additional buds on the main bronchi within 1 day (Fig. 1A) followed by secondary bifurcations at the tips by the second day (Fig. 1B,C) to form a highly branched structure. Lungs cultured without ODNs develop secondary bifurcations within 1 day, so that the lungs treated with the control ODNs develop normally, but with a 1-day delay. In contrast, treatment with an ODN directed against a sequence within the T-box that is identical in Tbx4 and Tbx5 (4/5BoxAS) eliminated the development of new epithelial branches (Fig. 1D–F). The number of secondary branches in 4/5BoxAS ODN-treated lung buds is unchanged after 2 days of culture (Fig. 2B vs. Fig. 2A). A similar inhibition of branching was obtained with a second AS ODN directed against a different sequence shared by Tbx4 and Tbx5 (data not shown). In contrast, lung buds cultured with equal amounts of three different control ODNs continued to branch normally (Fig. 2C and data not shown). This last result rules out the unlikely possibility that the inhibition might be due to cross-hybridization between AS ODNs and transcripts from genes other than Tbx4 and Tbx5. Furthermore, treatment with 4/5BoxAS ODN caused a profound reduction in the levels of both of these transcripts, as determined by reverse transcriptase-polymerase chain reaction (RT-PCR; Fig. 3). In contrast, the more distantly related Tbx2 transcript was largely unaffected.
Treatment with 4/5BoxAS ODN completely abrogated new lung branching in more than 20 independent experiments comprising a total of 138 AS ODN-treated lungs. To quantify the degree of inhibition, we counted the number of terminal buds of the left primordia at initiation of culture and after 48 hr for one series of experiments comprising 30 control and 52 experimental lungs. During the culture period, the number of terminal buds in the control ODN-treated cultures increased fourfold (from 2.3 ± 0.8 to 7.7 ± 3.8), whereas the number of terminal buds in the AS ODN-treated cultures was essentially unchanged (2.2 ± 0.4 vs. 1.9 ± 0.7). Occasionally, very early branches that were detectable at the outset regress during culture with AS ODNs. Similar experiments were also performed with AS ODNs directed against either Tbx4 or Tbx5 transcripts individually. Both classes of AS ODNs reduced branching relative to control ODN-treated lung buds (Fig. 2D,E). However, in both cases, the inhibition was not complete, suggesting an overlap in function between these two closely related genes. In contrast, clear evidence of new branch initiation was observed with an AS ODN directed against the T-box region of Tbx2 and Tbx3 (Fig. 2F), which are also expressed in lung mesenchyme (Fig. 7F, Chapman et al., 1996). Taken together, these data suggest that Tbx4 and Tbx5 are required for epithelial branching in the mammalian lung.
Effects of Inhibition of Tbx4 and Tbx5 Expression of Fgf10
Lung epithelial branching has been shown to be dependent on FGF10 signaling from the subjacent mesenchyme. To determine whether the inhibition of lung branching was the result of interfering with the production of the normal mesenchymal signals, we analyzed the expression of Fgf10 by semi-quantitative RT-PCR and in situ hybridization. We observed that 4/5BoxAS ODN treatment caused a dramatic reduction in the levels of Fgf10 mRNA (Fig. 4). The level of Fgf10 expression observed in the treated lungs by PCR analysis of the undiluted cDNA was roughly comparable to the level observed at a 32-fold dilution of cDNA from control ODN-treated cultures. Fgf10 expression in the antisense-treated lung samples became undetectable after only a fourfold dilution. In contrast, the levels of RNA for ribosomal S16 protein were roughly comparable in all samples.
Similar results were obtained by in situ hybridization (Fig. 5). Although the in situ hybridizations were carried out under identical conditions, we consistently failed to detect hybridization with the Fgf10 probe to the 4/5BoxAS ODN-treated lungs (Fig. 5A). Fgf10 expression was readily apparent in the mesenchyme of the control ODN-treated lungs (Fig. 5B), although it did not appear as localized as in uncultured lungs (see Fig. 7B,G). This finding may have been a result of the culture conditions in which the lung is forced to grow flat on a Nucleopore filter. We also detected a loss of epithelial Shh expression by in situ hybridization, consistent with the lack of new terminal bud formation (Fig. 5C). This reduction in Shh mRNA levels was not detectable by RT-PCR, probably because Shh is also expressed at low levels throughout the epithelium (data not shown). These data both indicate that the expression of Fgf10 is dependent on Tbx4 and Tbx5.
To confirm that the observed inhibition of branching is the result of a reduction in the level of FGF10 signaling, we supplemented 4/5BoxAS-treated cultures with beads soaked with FGF10. The addition of a localized source of FGF10 was able to restore the appearance of terminal budding in 78% of the lungs (11 of 14, Fig. 6A–C). bovine serum albumin (BSA) -coated beads failed to stimulate branch formation in 4/5BoxAS-treated lungs (0 of 9, Figure 6D–F).
Expression of Tbx Genes in the Developing Lung
Previous experiments have shown that Tbx4 and Tbx5 were expressed in lung mesenchyme, but no study has yet examined their distribution in detail. We examined the expression pattern of Tbx5 by both whole-mount and section in situ hybridization and compared it with Fgf10 expression. We found that Tbx5 is expressed throughout the lung mesenchyme, with the highest levels found adjacent to the epithelial branches (Fig. 7A,D). Tbx4 was expressed throughout mouse lung and tracheal mesenchymes (Fig. 7E), as previously reported (Chapman et al., 1996). Fgf10 is expressed in patches in the mesenchyme (Fig. 7B,G, arrows) as previously observed (Bellusci et al., 1997). The expression of Tbx4 and Tbx5 overlap the Fgf10 expression domains but are much more widespread (Fig. 7). Thus, Tbx5 and Tbx4 are required for Fgf10 expression but are probably not sufficient to account for its localized pattern.
The reciprocal induction of the lung mesenchyme and epithelium involves the close integration of signaling pathways in these two tissues. Unlike most other epitheliomesenchymal interactions, the lung mesenchyme controls both the architecture of epithelial branching and the cytodifferentiation of the lung epithelial cells (Alescio and Cassini, 1962; Wessells, 1970; Hilfer et al., 1985; Shannon et al., 1998). Recent research has shown that several paracrine factors can regulate lung epithelial branching; however, the best candidate for the mesenchymal signal is FGF10. FGF is expressed in punctate regions of the lung mesenchyme at the time preceding new branch formation (Bellusci et al., 1997; Park et al., 1998; Weaver et al., 2000), and mice carrying null mutations in either Fgf10 or its receptor fail to develop lungs (Min et al., 1998; Arman et al., 1999; Sekine et al., 1999; de Moerlooze et al., 2000). The epithelium, in addition to inducing the mesenchymal signal, also limits its expression by the production of antagonists to FGF signaling such as SHH, mSprouty2, and BMP4 (Bellusci et al., 1996; Hacohen et al., 1998; Pepicelli et al., 1998; Tefft et al., 1999). The latter two antagonists appear to be induced by the mesenchymal FGF10 signal itself (Weaver et al., 2000; Mailleux et al., 2001), and SHH from the epithelium is thought to suppress Fgf10 expression in the mesenchyme (Bellusci et al., 1997; Pepicelli et al., 1998). Thus, FGF signaling establishes a negative feedback loop wherein the epithelium is able to limit the extent of branch outgrowth by blocking the effects of FGF10 signaling in the epithelium and preventing Fgf10 expression in the mesenchyme.
The present studies demonstrate that two of the major proteins coordinating this feedback loop are the T-box transcription factors Tbx4 and Tbx5. When 11.5 dpc embryonic lung rudiments are cultured in vitro for 2 days, numerous new branches are seen. However, treatment of these embryonic lung cultures with AS ODNs that eliminate Tbx4 and Tbx5 is found to suppress Fgf10 expression in the mesenchyme and to completely eliminate new lung branches. Moreover, local replacement of FGF10 to these treated lungs restores epithelial branching. AS ODNs to Tbx2 and Tbx3 fail to suppress lung branching, even though these transcripts are also expressed in the mesenchyme during lung development. These studies took advantage of homologous regions common to Tbx4 and Tbx5 genes that enable the production of AS ODNs that eliminate both gene transcripts. It is possible that Tbx4 and Tbx5 play functionally redundant roles in the lung; the inhibition of branching was not absolute using AS ODNs directed against only Tbx4 or Tbx5. Lung branching does not appear to be as sensitive to the level of Tbx gene expression as the heart and limb; mice and humans heterozygous for a nonfunctional TBX5 allele show forelimb and heart dysgenesis, but do not show lung anomalies (Bruneau et al., 2001; Ghosh et al., 2001).
Our studies also show that expression of neither Tbx4 or Tbx5 is sufficient to explain the punctate distribution of Fgf10 gene expression in the lung mesenchyme. Other transcription factors, including those of the Max, SP1, and Nkx families, have been shown to interact with Tbx genes (Hurlin et al., 1999; Harrison et al., 2000; Bruneau et al., 2001; Hiroi et al., 2001), and the combinatorial expression of several transcription factors may be required for the localization of the Fgf10 expression domain.
The control of FGF expression by T-box transcription factors may be a pathway common to other vertebrate induction fields. The Xenopus Brachyury and VegT proteins are critical in activating FGF signaling in the production of Xenopus mesoderm (Isaacs et al., 1994; Schulte-Merker and Smith, 1995; Kofron et al., 1999), and the Brachyury protein of Xenopus is known to bind to specific sites in the regulatory region of the eFgf gene (Casey et al., 1998). In mouse blastocysts, the expression of another T-box gene, eomesodermin, appears to be necessary for the synthesis of FGF and epithelial differentiation (Li et al., 2001). FGF signaling pathways may also regulate T-box genes (Hug et al., 1997; Griffin et al., 1998; Isaac et al., 2000), and thereby provide a mechanism for autocrine regulation.
The relationship between Tbx genes and Fgf expression focuses attention on developing systems where both have independently been implicated. Certain disorders involving Tbx genes affect organs where FGFs have been implicated in morphogenesis. For instance, several members of the FGF family play important roles in the induction and patterning of the limb (Niswander et al., 1993; Crossley et al., 1996; Ohuchi et al., 1997), a structure that is affected by mutations in both the Tbx3 and Tbx5 genes (Bamshad et al., 1997; Basson et al., 1999). Posterior mesoderm and tail development requires both FGFs and the products of the T and Tbx16/spadetail genes (Dobrovolskaia-Zavadskaia, 1927; Chesley, 1935; Mansour et al., 1993; Griffin et al., 1998; Xu et al., 1999). Fibroblast growth factors FGF4, -8, and -9 have been implicated as epithelial signals regulating mesenchymal gene expression and cell proliferation during tooth initiation, whereas FGF3 and FGF10 are mainly restricted to the dental papilla mesenchymal cells (Kettunen et al., 2000). Tooth morphogenesis is also affected by mutations in the Tbx3 and Tbx1 genes (Bamshad et al., 1997; Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001). FGF10 is required for the formation of the salivary glands, which are also missing in mice homozygous for mutations in the Tbx1 gene (Ohuchi et al., 2000; Jerome and Papaioannou, 2001). Similarly, FGFs are known to play crucial roles in the development of apocrine glands (Jackson et al., 1997; Fantl et al., 2000) and genitalia (Haraguchi et al., 2000), sites that are affected by mutations of Tbx3 (Bamshad et al., 1999). It is tempting to speculate that a relationship between Tbx proteins and FGF proteins may also exist in some of these sites of epitheliomesenchymal interactions.
In his review of lung morphogenesis, Minoo (2000) declared that, although T-box transcription factors were expressed in the developing lung, their relationship to FGF signaling had not yet been explored. Moreover, he continued, “An examination of whether and how any of the latter transcriptional regulators are involved in epithelial-mesenchymal interactions during lung development is needed to elucidate the key role of the FGF signaling pathway in driving lung morphogenesis.” Our results demonstrate that the T-box transcription factors Tbx4 and Tbx5 play a pivotal role in the activation of the FGF signaling pathway during mammalian lung morphogenesis. Moreover, this relationship between T-box transcription factors and FGFs may enable the coordinated regulation of signaling pathways in numerous epitheliomesenchymal interactions.
Lung Culture and Antisense Oligonucleotide Treatment
Embryos were isolated from Swiss-Webster (Taconic), CD-1 or CF-1 (Charles River) mice at 11.5 dpc. Noon on the day that the vaginal plug is detected is considered to be 0.5 dpc. Lung buds were isolated by manual dissection and cultured on Transwell Nucleopore membranes (Costar) in defined serum-free medium consisting of DMEM and Ham's F12 (Mediatech, 1:1) supplemented with 1 mg/ml BSA (Sigma) and 0.1 mg/ml human transferrin (Life Sciences, Serra et al., 1994) for 45 to 48 hr. Phosphothioate-derivatized ODNs were synthesized in both the control (RNA-like strand; S) and or antisense (template strand; AS) orientations, 5′-fluoresceinated and purified. ODNs were added at a final concentration of 15 μM (CF-1) or 30 μM (Swiss, CD-1) for the duration of the experiment. ODNs used were as follows: 4/5BoxS (5′fluorescein-AACAACCACCTGGACCC), 4/5BoxAS (5′fluorescein-GGGTCCAGGTGGTTGTT), 4/5AS2 (5′fluorescein-AACACGTGGGTGCAAAA), 4/5S2 (5′fluorescein-TTTTGCACCCACGTGTT), Tbx5AS (5′fluorescein-AGGAGACAAGTTGTC), Tbx4AS (5′fluorescein-AAGAGACCAACTGCC), 2/3S (5′fluorescein-AAGTACCAGCCGCGATT), and 2/3AS (5′fluorescein-AATCGCGGCTGGTACTT). All ODNs were directed against the T-box region. 4/5BoxAS and 4/5AS2 are complementary to sequences that are identical in Tbx4 and Tbx5, whereas Tbx4AS and Tbx5AS are complementary to unique sequences. 2/3AS is complementary to a sequence that is identical in Tbx2 and Tbx3.
After incubation at 37°C for 48 hr, the lung buds were harvested, photographed, quick-frozen in liquid nitrogen, and stored at −80°C for RNA analysis. The expression of signaling molecules and members of the Tbx gene family was assessed by PCR analysis (RT-PCR). Total RNA from control and AS ODN-treated cultures was isolated by using TRIZOL reagent (Sigma; Chomczynski, 1993), and the cDNA was prepared and PCR performed as previously described (Bollag et al., 1994). Additional PCR primers used were Fgf10 (5′-TGTTTTTTGTCCTCTCCTGGGAG and 5′-GGATACTGACACATTGTGCCTCAG; as per B. Hogan), Tbx4 (5′-TCACTGGATGCGGCAGTTGGTCTCT and 5′-CACGTGGGTGCAAAAGGCTGTGTTT), and Tbx5 (5′-CCACTGGATGCGACAACTTGTCTCC and 5′-GACGTGGGTTGCAAAACGCAGTGTTC). RNA for the S16 ribosomal protein was used as a positive control to compare levels of cDNA (Bollag et al., 1994). In some experiments, serial twofold dilutions of the input cDNA were used to estimate the magnitude of the differences between sense and antisense ODN-treated cultures.
Heparin-acrylic beads (Sigma) were manually fractionated, rinsed three times with PBS and soaked with 0.25 μg recombinant human FGF10 (Research Diagnostics, Inc.) as described by Weaver et al. (2000). FGF10-coated beads were washed with PBS and transferred adjacent to cultured lungs by using a mouth pipet. Control beads were soaked in 0.25 μg of BSA.
Whole-Mount In Situ Hybridization
Cultured lungs were fixed overnight in cold 4% paraformaldehyde in PBS, washed, dehydrated with methanol, and subjected to whole-mount in situ hybridization as described (Riddle et al., 1993) with digoxigenin-labeled RNA probes for mouse Fgf10 (Bellusci et al., 1997) and Shh (Chang et al., 1994). Polyvinyl alcohol was added to the detection solution to enhance the color reaction (Barth and Ivarie, 1994). Embryos were added to lung in situ hybridization experiments to control for probe integrity.
Section In Situ Hybridization
In situ hybridization to tissue sections was performed as described (Frohman et al., 1990; Reddy et al., 2001). Embryos were fixed overnight in paraformaldehyde, dehydrated, and embedded in paraffin wax. Eight-micron sections were hybridized with 35S-labeled riboprobes, washed, and dipped in NTB2 emulsion (Kodak). Slides were developed, counterstained with 2 μg/ml Hoechst 33258 (Sigma), mounted in Canada balsam, and photographed by using an MVI Darklite stage adapter.
We thank Gail Speakman and Saw Kyin of the Princeton University Synthesis & Sequencing Facility for oligonucleotide synthesis; Molly Weaver for advice on application of FGF-soaked beads; and Roni Bollag, Brigid Hogan, and Philip Beachy for providing probes. We thank Sarah Millar for critical reading of the manuscript and for assistance with the radioactive in situ hybridization experiments; members of the Cebra-Thomas laboratory for thoughtful discussions; and Lee Silver for advice during the initial stages of this work. This work was supported by Franklin & Marshall College (J.C.-T. and R.G.), the Department of Molecular Biology, Princeton University (J.C.-T. and G.K.L.), and Swarthmore College (J.C.-T., H.S., J.B., and S.F.G.).