Lung hypoplasia and neonatal death in Fgf9-null mice identify this gene as an essential regulator of lung mesenchyme

Chicken lung graft experiments dating back to 1933 demonstrated that mesenchyme is essential for airway branching in embryonic lung. Subsequent studies showed that bronchial mesenchyme stimulates branching, while tracheal mesenchyme inhibits branching (Hilfer, 1996). Recent studies have identified fibroblast growth factor 10 (FGF10) as a mesenchymal signal essential for airway branching (Min et al., 1998; Sekine et al., 1999). We present evidence that an epithelial signal, FGF9, regulates airway branching by stimulating mesenchymal proliferation and modulating Fgf10 expression.

Lung embryogenesis begins in mice at embryonic day (E)9.5 with the outpocketing of two epithelial buds from the ventral foregut into the surrounding mesoderm. Later lung development is divided into several stages:

(1) During the early pseudoglandular stage (E9.5-E14.2), the lung epithelium branches to form the proximal conducting airways, which are lined by columnar epithelium. At this stage, the epithelial tubes are separated by abundant mesenchyme.

(2) During the late pseudoglandular and canalicular stages (E14.2-E16.6), further airway branching generates bronchial tubules with columnar epithelium, and, distally, acinar tubules with cuboidal epithelium. During this period, the epithelial cells progressively flatten, the proportion of mesenchyme to epithelial tissue decreases, and the terminal buds dilate.

(3) During the saccular stage (E16.6-P5 (postnatal day 5)), distal airspaces form. These airspaces are acinar tubule derivatives, including alveolar ducts, sacs and primitive alveolar structures.

(4) During the alveolar stage (beginning at P5 in mice), alveoli are formed by septation (Bellusci et al., 1997a; Hogan and Yingling, 1998; Quaggin et al., 1999; Ten Have-Opbroek, 1981; Thurlebeck, 1995; Weinstein et al., 1998).

FGF signaling is essential at several stages of mammalian lung development. Fgf10 and its receptor, Fgfr2b, are required for epithelial branching. Targeted deletion of either gene prevents branching, causing the trachea to terminate as a blind sac (De Moerlooze et al., 2000; Min et al., 1998; Sekine et al., 1999). Double Fgfr3/Fgfr4-null mice survive into adulthood and exhibit normal embryonic lung development, but lack postnatal formation of alveoli (Weinstein et al., 1998). Fgf7 is expressed in developing lung mesenchyme; but Fgf7^−/− mice exhibit normal lung morphology, indicating some functional redundancy among Fgf genes in developing lung (Guo et al., 1996). To date, no member of the Fgf family has been demonstrated to regulate lung mesenchyme development.

Fgf9 is one of at least 22 members of the Fgf family (Ornitz and Itoh, 2001). In situ hybridization on mouse embryos detects Fgf9 expression in skeletal myoblasts, motoneurons, limb apical ectoderm, ventricular myocardium, the mesenteric lining surrounding endoderm-derived organs, gut luminal epithelium and the inner ear (Colvin et al., 1999). At E10.5, Fgf9 is expressed in the visceral pleura lining the outside of the lung bud and in the epithelium of the developing bronchi. At E12.5 and E14.5 (data not shown), Fgf9 expression persists in the mesothelium of the visceral pleura, but is no longer detected in airway epithelium (Colvin et al., 1999). This lung expression pattern contrasts with expression of Fgf10 and Fgf7, which are restricted to lung mesenchyme, and suggests novel roles for Fgf9 in lung development.

To investigate the in vivo functions of Fgf9, we generated mice homozygous for a targeted disruption of Fgf9 (Colvin et al., 2001). Here we report lung hypoplasia and early postnatal death in Fgf9^−/− mice. All Fgf9^−/− lungs exhibit reduced mesenchyme and decreased airway branching. We further demonstrate that both mesenchymal proliferation and mesenchymal Fgf10 expression are reduced in Fgf9^−/− lung. These findings suggest a model in which epithelial FGF9 and mesenchymal FGF10 coordinately regulate epithelial airway branching and embryonic lung size.

The Fgf9 genomic locus was targeted in SM-1 embryonic stem (ES) cells using a targeting vector engineered to replace 500 bp of the genomic locus with a neomycin resistance cassette. Homologous recombination resulted in a deletion eliminating 224 bp of the 277 bp in the protein-coding region of exon 1 and extending into the first intron (Colvin et al., 2001). Expression from the targeted allele was assessed using RT-PCR of E18.5 brain RNA. The results indicated that no Fgf9 transcript joining exon 1 to either of the 2 known downstream exons was produced from the mutant allele (Colvin et al., 2001). Fgf9^−/− mice were maintained in a mixed 129/C57BL6 background. Fgf9 genotype was determined by PCR or by Southern blot. Southern blots of EcoRI-digested DNA were probed with probe 25.3. The PCR assay produced a 310 bp fragment from the genomic locus and a 234 bp fragment from the targeted allele (Colvin et al., 2001).

Tissues were fixed in 4% paraformaldehyde or 10% formalin, decalcified if necessary in Surgipath decalcifier II, dehydrated with gradient ethanol, placed in xylene and embedded in paraffin. Sections (5 μm) were cut and stained with Hematoxylin and Eosin. Computer imaging using Foster-Findley PC Image software was used to calculate epithelial and mesenchymal areas on sections of E13.5 lungs. Tissues for plastic sections and transmission electron microscopy were fixed in 2% gluteraldehyde and postfixed in osmium tetroxide before embedding.

Fgf9^+/− females that had been mated with Fgf9^+/− males were given an intraperitoneal injection of BrdU (120 μg/gram body weight; Sigma) and FrdU (12 μg/gram body weight; Sigma) 1 hour before sacrifice. BrdU antibody staining of lung sections from the resulting embryos was carried out as described (Naski et al., 1998). Sections were counterstained with Methyl Green. Proliferation ratios were determined as BrdU-labeled nuclei over total nuclei in fields viewed through a 20× objective. At E10.5, the 20× field generally includes the entire lung bud. Results were determined to be significant at P<0.01.

Tracheas of E18.5 embryos (or of lungs that had been dissected out of E18.5 embryos) were injected with a 50:1 mixture of Mercox resin and catalyst (Ladd Research Industries, #21245). After allowing the plastic to solidify at room temperature for 1 to 2 hours, lung tissue was removed by immersion in 20% KOH for several days. For s.e.m., plastic casts were dehydrated in graded alcohols, critical-point dried in liquid carbon dioxide, mounted on aluminum stubs and coated with gold (deMello et al., 1997).

Paraffin sections were rehydrated, trypsin treated (for CCSP and vWF), microwaved (for SP-C and vWF) in antigen unmasking solution (Vector), incubated in methanol/hydrogen peroxide to block endogenous peroxidases and blocked with diluted serum. The primary antibodies used were: an anti-human antiserum against pro-surfactant C (SP-C) (rabbit; a gift from J. Whitsett;1:150 dilution) (Zhou et al., 1996), an anti-mouse CCSP antibody (rabbit; a gift from F. DeMayo;1:2500 dilution) (Ray et al., 1996) and an anti-human vWF antibody (rabbit; from DAKO; 1: 1000 dilution) (Schor et al., 1998). The secondary antibodies used were: a biotinylated goat anti-rabbit IgG (for SP-C and CCSP) and a Cy3-conjugated anti-rabbit IgG (for vWF). For CCSP, the Vectastain ABC-alkaline phosphatase reagent and AP substrate were used (Vector), followed by Eosin counterstaining. For SP-C, the Vectastain ABC-peroxidase reagent and DAB substrate were used (Vector).

In situ hybridization was performed as described (Sasaki and Hogan, 1993), with modifications. Control and Fgf9^−/− tissues were processed together to ensure identical conditions. Lung tissue was dissected in cold diethyl pyrocarbonate (DEPC) phosphate-buffered saline (PBS), fixed overnight in 4% paraformaldehyde (PFA)/DEPC PBS, dehydrated through graded methanols in DEPC PBT (PBS+0.1% Tween-20) (25%, 50%, 75%, 100%, 100%) and stored at −20°C. For use, tissue was rehydrated and washed with DEPC PBT, digested with proteinase K (10 μg/ml)/DEPC PBT (for 20-25 minutes), rinsed in PBT, fixed in 4% PFA/0.2% glutaraldehyde/DEPC PBT, and rinsed with DEPC PBT. Three-step prehybridization included incubation in (1) 1:1 hybridization solution (50% formamide, 1.3× SSC, 5 mM EDTA, 50 μg/ml yeast tRNA, 0.2% Tween-20, 0.5% Chaps, 100 μg/ml heparin) to PBT mix; (2) hybridization solution; and (3) hybridization solution at 65°C for 1 hour. Tissues were incubated in hybridization solution (∼1 μg/ml) for 3 days at 65°C with digoxigenin (DIG)-labeled probes for Fgf10 (622 bp, provided by B. Hogan) or sonic hedgehog (Shh; 642 bp, provided by A. McMahon). Tissue was then washed with 2× SSC w/0.1% Chaps (three times for 20 minutes at 65°C), 0.2× SSC w/0.1% Chaps (three times for 20 minutes at 65°C) and KTBT (50 mM Tris-HCl, 150 mM NaCl, 10 mM KCl, 1% Triton X-100) twice for 5 minutes. Tissue was preblocked with KTBT/20% sheep serum, incubated with anti-DIG Fab alkaline phosphatase conjugate (Boehringer Mannheim) in KTBT/20% sheep serum overnight at 4°C, washed with KTBT (four times for 1 hour), and left overnight in KTBT at 4°C. For the alkaline phosphatase color reaction, tissue was washed with NTMT (40 mM Tris, 100 mM NaCl, 40 mM MgCl, 0.2% Tween-20) twice for 15 minutes, then incubated in the dark with NTMT/BCIP (4.5 μl/ml)/NBT(3 μl/ml). The reaction was stopped by washing in KTBT and post fixing in 4% PFA/PBS for 10 minutes. Stained tissues were dehydrated through graded methanols in PBT and stored at 4°C. For Fgf10 in situ, n=3 at E12.5 and n=3 at E13.5-E14.5; for Shh in situ, n=2.

Mice heterozygous for the targeted allele (Fgf9^+/−) were phenotypically normal and were bred to produce homozygous null mice (Fgf9^−/−). Among 138 heterozygous intercross offspring genotyped at 1 to 3 weeks postpartum, the genotype distribution was: 36%, wild type; 64%, heterozygous; and 0%, null. Newborn pups from heterozygous intercrosses were periodically found dead. All nine of the dead pups yielding usable DNA were Fgf9^−/−. Among an additional 375 embryos harvested at E10.5-E18.5, the genotype distribution was: 24%, wild type; 51%, heterozygous; and 25%, null. These data indicate that Fgf9^−/− mice survive embryonic development but die in the early postnatal period.

Definitive pathology in Fgf9^−/− mice has been observed in the lung and reproductive system. Fgf9^−/− mice exhibit dramatic lung hypoplasia. Male-to-female sex reversal in Fgf9^−/− mice is described elsewhere (Colvin et al., 2001). Total mean bodyweight of E18.5 Fgf9^−/− mice was 84% of that of controls (Table 1). The etiology of embryonic growth retardation is unclear. Variable dilation of cardiac atria and ventricles was observed in E14.5-E18.5 Fgf9^−/− embryos, suggesting possible cardiovascular insufficiency. Fgf9 is expressed in E10.5-E12.5 heart in the region of the developing interventricular septum (Colvin et al., 1999), suggesting a possible role for Fgf9 in cardiac development. However, similar variable cardiac chamber dilation in Fgf10^−/− embryos (J. S. C., unpublished), which lack lungs (Min et al., 1998; Sekine et al., 1999), suggests that cardiac dilation in Fgf9^−/− mice may be secondary to lung hypoplasia.

Fgf9^−/− mice die shortly after birth, apparently owing to hypoxia. Among eight E18.5 Fgf9^−/− pups observed after Caesarian section, all pups moved, seven out of eight breathed and several became somewhat pink. By 30 minutes postpartum, all seven breathing pups exhibited agonal breathing and cyanosis. One naturally delivered Fgf9^−/− pup remained pink and active until sacrificed about 20 minutes postpartum. Thus, E18.5 Fgf9^−/− pups can breathe, and (initially) some ventilate enough to become pink, indicating oxygenation of systemic tissues; however, they become severely cyanotic shortly thereafter. About 40% of Fgf9^−/− embryos exhibit cleft palate (A. C. W., unpublished). Cyanosis and neonatal death are associated with cleft palate in Tgfb3^−/− and Msx1^−/− (Satokata and Maas, 1994) mice, suggesting that cleft palate may contribute to the death of some Fgf9^−/− mice. However, Tgfb3^−/− mice die at 11-20 hours after birth (Kaartinen et al., 1995), a less precipitous decline than observed in Fgf9^−/− mice. These findings suggest that hypoxia secondary to lung hypoplasia causes the neonatal death of Fgf9^−/− mice.

The lung hypoplasia phenotype is 100% penetrant with little variation in severity. All Fgf9^−/− E12.5-E18.5 lungs examined were hypoplastic (n=48). Mean wet weight of E18.5 Fgf9^−/− lungs was 29% that of controls, whereas mean Fgf9^−/− total body weight was 84% that of controls (Table 1). Hence, generalized reduction in embryonic growth cannot account for the decreased lung volume in Fgf9^−/− mice. The number and orientation of lobes is preserved in Fgf9^−/− lungs (Fig. 1A,B), indicating that very early epithelial branching occurs normally. Primary bronchi in E12.5-13.5 Fgf9^−/− lungs were as long as those of controls (data not shown), indicating that longitudinal growth of airways is at least initially preserved. However, in contrast to those of littermate controls, E14.5 Fgf9^−/− lungs failed to fill the thoracic cavity (Fig. 1C,D). The diaphragm in Fgf9^−/− mice was intact (Fig. 1D), and no abdominal organs were observed in the thoracic cavity of Fgf9^−/− mice. These findings suggest that a primary defect in lung growth, rather than compression of the developing lung, causes Fgf9^−/− lung hypoplasia.

To investigate the origin of the lung hypoplasia phenotype, we examined Fgf9^−/− lungs starting at E10.5, shortly after initial formation of the bronchial buds. E10.5 and most E11.5 Fgf9^−/− lungs were grossly and histologically indistinguishable from controls (Fig. 1I,J). By E12.5, lung and mesenchymal tissue hypoplasia were apparent on gross examination of Fgf9^−/− lungs (Fig. 1E,F). Histology (Fig. 1K,L) and quantitative analysis of tissue areas (Table 2) revealed a reduced ratio of mesenchyme to epithelial tissue in E13.5 Fgf9^−/− lungs relative to controls. All E12.5-E14.5 Fgf9^−/− lungs grossly or histologically examined showed lung hypoplasia and reduced lung mesenchyme (n=19). By E12.5, some peripheral epithelial buds in Fgf9^−/− lungs appeared to be dilating, a process that normally coincides with the thinning of mesenchyme and distal epithelial differentiation later in lung development (Fig. 1G,H; Bellusci et al., 1997a; Ten Have-Opbroek, 1981). This observation suggests that the reduction in the amount of mesenchyme in Fgf9^−/− lungs may allow premature dilation of epithelial buds.

Epithelial and mesenchymal proliferation in lung tissue was assessed by BrdU immunohistochemistry after a 1 hour exposure to BrdU in utero (Table 3). Mesenchymal proliferation was significantly reduced in Fgf9^−/− lungs relative to controls at E10.5-E11.5. No difference in epithelial proliferation between Fgf9^−/− lungs and controls was observed at E10.5-E11.5. At E13.5, no difference in mesenchymal or epithelial proliferation between Fgf9^−/− lungs and controls was detected. This suggests that FGF9 drives mesenchymal proliferation primarily early in lung development, when Fgf9 is expressed in both the epithelium and the pleura. Although equal proportions of mesenchymal cells are proliferating in control and Fgf9^−/− lungs at E13.5, there are fewer total mesenchymal cells in E13.5 Fgf9^−/− lungs than in controls. Hence, the normal decline in mesenchymal thickness as lung development progresses would be expected to occur prematurely in Fgf9^−/− lungs.

During early embryonic development, extensive branching of epithelial buds (about 16 generations in humans) generates conducting airways (Fig. 2J). Conducting airways are lined with columnar epithelium and function after birth, in part, to transport air in and out of the lung. Later iterations of epithelial branching (generations 17-23 in humans) generate respiratory airways, which function after birth to exchange oxygen and carbon dioxide between air and blood (Hasleton, 1996a). Walls of respiratory airways contain small alveolar sacs, which are further subdivided into alveoli by postnatal septation. Fgf9^−/− lung hypoplasia could be the result of (1) decreased branching of airways and/or (2) failure of development of distal lung respiratory structures, including the alveolar ducts and sacs.

Histological and morphological examination was used to assess airway branching in Fgf9^−/− lungs. Preservation of number and orientation of lobes in Fgf9^−/− lungs suggests that the earliest stages of proximal branching are preserved (Fig. 1A,B). By E12.5, however, epithelial branching was less extensive in Fgf9^−/− lungs than in controls (Fig. 1E,F). The observation that lung hypoplasia and reduced epithelial complexity in Fgf9^−/− lungs is well established by E12.5, prior to the onset of distal airspace formation, suggests that reduced proximal airway branching contributes significantly to Fgf9^−/− lung pathology. At E13.5, secondary branching of the accessory lobe had occurred in control lungs but was absent from Fgf9^−/− lungs (data not shown). Although peripheral epithelial buds in E12.5-E13.5 control lungs were readily observed in the process of budding into two epithelial branches (arrow in Fig. 1G), peripheral buds in Fgf9^−/− lungs generally appeared dilated and unbranched (Fig. 1H). Sections of Fgf9^−/− lungs at E13.5 (Fig. 1L) and E14.5 (data not shown) showed fewer small diameter airways than controls (arrow in Fig. 1K), also suggesting reduced airway branching. In sections of E18.5 lungs, fewer generations of conducting airways were seen in Fgf9^−/− lungs than in controls (see Fig. 4A,B below). These results indicate impaired airway branching in Fgf9^−/− lungs.

To assess further whether decreased airway branching and/or distal airspace formation contributed to lung hypoplasia in Fgf9^−/− mice, we studied airway branching using plastic epithelial casts of E18.5 lungs. All Fgf9^−/− lung casts generated (n=15) were smaller than controls, but exhibited evidence of both proximal branching and distal airspace formation (Fig. 2A-C). Proximal branching can be observed in underinjected casts, generated by injecting less plastic than is required to expand the lungs fully. Underinjected control casts exhibited an extensive network of proximal branches (Fig. 2B). The scaffolding of proximal branches in control casts was larger than the total size of more fully injected Fgf9^−/− casts (Fig. 2C). The main bronchus of the accessory lobe in control casts is hidden by an extensive network of airway branching (Fig. 2D). In contrast, the main bronchus of the accessory lobe is readily visible in Fgf9^−/− lung casts. Even in overinjected Fgf9^−/− casts, much of the main bronchus of the accessory lobes lacked secondary branching (arrow in Fig. 2E). These findings further demonstrate impaired airway branching in Fgf9^−/− lungs.

Epithelial lung casts also allowed us to assess distal airspace development. Outer surfaces of control (Fig. 2A,F) and Fgf9^−/− (Fig. 2C,G) lung casts showed comparable density of alveolar sacs, indicating that alveolar sac formation can occur in the absence of Fgf9. Scanning electron microscopy of lung casts showed similar structure of distal alveolar sac clusters in control (Fig. 2H) and Fgf9^−/− (Fig. 2I) casts. This indicates that development of several generations of distal respiratory airways can occur without Fgf9. Normal density of air sacs on the lung surface and normal structure of distal airway clusters, together indicate preservation of significant distal airway development in Fgf9^−/− lung. In contrast, Pod1^−/− mice (Tcf21 – Mouse Genome Informatics), which show moderately reduced proximal airway branching, exhibit nearly complete absence of distal airspace formation (Quaggin et al., 1999). Fgf9^−/− lungs clearly have fewer alveolar sacs overall than controls (Fig. 2A,C). This could result from either impaired branching of proximal airways, or from generalized impairment of both proximal branching and distal airspace formation.

To carry out gas exchange after birth, the distal lung must undergo extensive morphological changes late in fetal development. These changes include thinning of the epithelium, differentiation of pneumocytes, development of capillaries and close juxtaposition of capillaries to flattened pneumocytes that line the walls of alveolar sacs. Gas exchanged between blood in the capillaries and air in the alveolar sacs must diffuse across the three layers of the alveolar sac wall (i.e. the squamous Type I pneumocyte, the intervening basement membrane and the capillary endothelial cell). Pneumocytes in alveolar sac walls must be thin and closely associated with thin capillary endothelial cells to allow efficient diffusion.

To assess distal differentiation, we analyzed E18.5 Fgf9^−/− lungs using plastic-embedded sections and transmission electron microscopy. Similar distal architecture was observed in Fgf9^−/− and control lungs (Fig. 3A,B). Abundant capillaries and close apposition of capillaries to alveolar spaces were seen in both Fgf9^−/− and control lungs (Fig. 3A,B), as was normal air:blood diffusion barrier morphology (Fig. 3E-H). Transmission electron microscopy revealed lamellar bodies within Type II pneumocytes in control and Fgf9^−/− lungs (Fig. 3I,J). Type II cells in Fgf9^−/− lungs also exhibited characteristic surface microvilli (arrowhead in Fig. 3J). These results indicate that distal lung differentiation is preserved in Fgf9^−/− mice. Fgf9 is expressed in the developing visceral pleura (Colvin et al., 1999). However, pleural structure, including surface microvilli on pleural mesothelial cells, tight junctions between pleural cells, and the underlying basement membrane, appears to be preserved in Fgf9^−/− lungs (Fig. 3K,L).

Finally, sections of proximal airways at E18.5 indicate normal cellular differentiation and epithelial organization in Fgf9^−/− lungs (Fig. 3C,D). Proximal epithelium includes both ciliated cells and Clara cells. Clara cells are nonciliated secretory cells found mainly in terminal and respiratory bronchioles (Hasleton, 1996a). Both control and Fgf9^−/− lung sections exhibited histologically normal cilated cells (arrowheads in Fig. 3C,D) and Clara cells (arrows in Fig. 3C,D).

Sections of E18.5 control (Fig. 4A) and Fgf9^−/− lungs (Fig. 4B) both show large proximal epithelial branches, but control lung sections show more smaller diameter conducting airways (arrow in Fig. 4A) than do Fgf9^−/− sections. Control and Fgf9^−/− lungs exhibit similar distal epithelial structures, including alveolar ducts and sacs, which are particularly evident in the less expanded upper regions of the lung (leftmost regions of Fig. 4A,B). To evaluate epithelial cell differentiation in E18.5 lung, we analyzed expression of proximal and distal lung epithelial cell markers using immunohistochemistry. Clara cell secretory protein (CCSP) is produced by Clara cells in the proximal lung epithelium. Staining with CCSP antibody revealed CCSP expression in Clara cells in E18.5 control (Fig. 4C) and Fgf9^−/− (Fig. 4D) lungs. Surfactant C (SP-C) is expressed in Type II pneumocytes in distal lung epithelium (Wert et al., 1993). Sections from E18.5 control (Fig. 4E) and Fgf9^−/− (Fig. 4F) lungs showed normal expression of SP-C in Type II pneumocyte cytoplasm. Extensive expression of CCSP and SP-C indicates that normal differentiation of Clara cells and Type II pneumocytes, respectively, can occur in the absence of Fgf9.

The CCSP staining pattern indicates a lack of small diameter conducting airways in Fgf9^−/− lung. The diameter of the most distal CCSP positive airways (arrowheads in Fig. 4C,D) is larger in Fgf9^−/− sections than in controls. Because CCSP is expressed in proximal airways, and airway diameter decreases as branching progresses, this indicates that branching of proximal airways is reduced in Fgf9^−/− lung. As CCSP is expressed primarily in the distalmost conducting airways (i.e. the terminal and respiratory bronchioles), this observation suggests that branching of the smaller conducting airways is regulated by Fgf9.

To assess vascular differentiation in Fgf9^−/− lungs, we used immunohistochemistry to analyze vonWillebrand Factor (vWF) expression in E18.5 lung. vWF antibody selectively stains large to medium sized arteries and veins (Kuzu et al., 1992). Expression of vWF in large blood vessels was comparable in control and Fgf9^−/− lungs (Fig. 4G,H). Plastic sections and transmission electron microscopy both showed normal capillary distribution and morphology in Fgf9^−/− lung (Fig. 3B,F). Finally, Hematoxylin and Eosin stained sections of Fgf9^−/− lungs exhibited normal histology and distribution of blood vessels (data not shown). These findings suggest that development of lung vasculature can occur in the absence of Fgf9. Hence, loss of blood supply secondary to reduced mesenchyme is unlikely to account for Fgf9^−/− lung hypoplasia.

Fgf10 and Shh are both essential for lung embryogenesis: Fgf10^−/− mice lack lungs (Min et al., 1998; Sekine et al., 1999) and Shh^−/− mice have dysmorphic lungs (Pepicelli et al., 1998). To determine whether aberrant expression of Shh and/or Fgf10 contributes to the Fgf9^−/− lung phenotype, we used whole-mount in situ hybridization to assess expression of Shh and Fgf10 in embryonic Fgf9^−/− lungs. Shh is expressed in the epithelium of developing mouse lung from E11.5 through E18.5 (Bellusci et al., 1997a). E14.5 control (Fig. 5A) and Fgf9^−/− (Fig. 5B) lungs showed comparable epithelial Shh expression.

During early lung embryogenesis, Fgf10 is detected in foci of high expression in lung mesenchyme distal or adjacent to budding airway epithelium. Fgf10 expression is observed in mesenchyme at the distal tip of the prospective lobes starting at about E10.0. By E12.5, Fgf10 is also observed in mesenchyme immediately adjacent to lung buds (Bellusci et al., 1997b). At E12.5, control (Fig. 5C,E) and Fgf9^−/− (Fig. 5D,F) lungs exhibited comparable expression of Fgf10 in mesenchyme distal to budding airways.

At E13.5 (data not shown) and E14.5, Fgf10 expression in Fgf9^−/− (Fig. 5H,J) lungs was reduced relative to controls (Fig. 5G,I). In control E13.5-E14.5 lungs, mesenchymal foci of Fgf10 expression were observed between most distal lung buds (arrowheads in Fig. 5I). Interbud mesenchyme in E13.5-E14.5 Fgf9^−/− lungs often lacked Fgf10 expression (arrowheads in Fig. 5J). Lung organ culture experiments have indicated that Fgf10 expression in lateral interbud mesenchyme induces branching and outgrowth of new distal buds (Bellusci et al., 1997b). Hence, the lack of Fgf10 in interbud mesenchyme in Fgf9^−/− lungs may impair future branching of lung buds. In Fgf9^−/− lungs, some buds that lacked adjacent Fgf10 expression (arrow in Fig. 5H) were dilated with little surrounding mesenchyme. Similar premature bud dilation and mesenchymal hypoplasia were observed in E12.5 Fgf9^−/− lungs (Fig. 1H). This suggests that reduced Fgf10 expression in E13.5-E14.5 Fgf9^−/− lungs may be secondary to a lack of mesenchyme. Dilation of airway buds in E12.5-E14.5 Fgf9^−/− lungs and reduced Fgf10 expression in E13.5-E14.5 Fgf9^−/− lungs are both consistent with premature termination of airway branching in Fgf9^−/− lungs.

In the embryonic lung, mesenchymally expressed Fgf10 (Fig. 6A, top panel) and Fgf7 have been previously shown to regulate epithelial development (Bellusci et al., 1997b). Here we show that an epithelial signal, FGF9, regulates lung mesenchymal development (Fig. 6A, top panel). While airway branching is reduced in Fgf9^−/− lungs, it is completely absent in Fgf10^−/− lungs (Min et al., 1998; Sekine et al., 1999). This suggests that FGF10 signaling initiates individual branch formation, whereas FGF9 signaling titrates the overall amount of branching in developing lung. FGF10, which is restricted to foci of high expression in mesenchyme, stimulates localized proliferation of adjacent epithelial cells (Bellusci et al., 1997b), and acts as a chemotactic factor for lung epithelial cells in vitro (Park et al., 1998). New lung buds apparently grow toward localized foci of high Fgf10 expression (Fig. 6B, top). In contrast, Fgf9 is expressed throughout lung epithelium and pleura at E10.5, and stimulates mesenchymal proliferation (Fig. 6B, top).

The reduced amount of mesenchyme in Fgf9^−/− lungs suggests that Fgf9 regulates airway branching by controlling mesenchymal expansion. Regulation of epithelial branching by lung mesenchyme is well documented (Hilfer, 1996; Hogan and Yingling, 1998). In fact, lung graft experiments have shown that the quantity of mesenchyme surrounding lung buds is proportional to the extent of branching observed (Masters, 1976). Also, as developing lung leaves the pseudoglandular phase (at about E16), the amount of mesenchyme decreases; concurrently, airway lumens dilate and the epithelial lining thins as differentiation of the epithelium progresses (Fig. 6B, bottom).

There are several mechanisms by which hypoplasia of lung mesenchyme could reduce airway branching. First, loss of mesenchyme could reduce Fgf10 expression, thus removing the signal that instructs individual epithelial buds to branch. Reduced Fgf10 expression in E13.5-E14.5 Fgf9^−/− lungs is consistent with this hypothesis. Second, loss of mesenchyme in Fgf9^−/− lungs could remove a signal that prevents premature epithelial differentiation during the pseudoglandular period. Investigation of this possibility will require detailed temporal studies of epithelial cell morphology and gene expression in Fgf9^−/− versus control lungs. Finally, thick lung mesenchyme in the early embryo may mechanically inhibit dilation of distal epithelial buds. Thinning of mesenchyme in Fgf9^−/− lungs would thus allow premature epithelial bud dilation by decreasing a physical constraint on epithelial expansion. Indeed, some epithelial buds in Fgf9^−/− lungs appear to be dilating at E12.5-E13.5, whereas control lung buds are still branching. Shh mutant mice support the hypothesis that the amount of mesenchyme around budding airways affects airway dilation: lungs in which Shh is overexpressed using the surfactant C promoter show increased mesenchyme and reduced airway dilation (Bellusci et al., 1997a).

Dynamic changes in foci of high Fgf10 expression in the mesenchyme appear to stimulate epithelial branching early in lung development (Bellusci et al., 1997b; Hogan and Yingling, 1998). Why airway branching stops is less well understood. Fgf10 may contribute to the termination of branching by stimulating epithelial expression of the gene for bone morphogenetic protein 4 (Bmp4), which inhibits endoderm proliferation (Fig. 6B, middle) and is essential for distal lung differentiation (Fig. 6B, bottom) (Quaggin et al., 1999; Weaver et al., 2000; Weaver et al., 1999). Our results suggest a model in which Fgf9 regulates the number of cycles of airway branching by regulating the amount of lung mesenchyme and the amount of mesenchymal Fgf10 expression. Early in normal lung development, when Fgf9 is expressed in both the epithelium and in the pleura, FGF9 signaling to mesenchyme is maximized, and proximal airway branching is prolific (Fig. 6B, top). After E12.5, when Fgf9 is expressed only in the pleura, the amount of lung mesenchyme gradually decreases and airway branching subsequently slows (Fig. 6B, lower). Fgf9^−/− lungs appear to ‘run out’ of mesenchyme prematurely, causing reduced airway branching. In this model, distal lung development occurs when mesenchyme surrounding developing airways decreases, Fgf10 expression is decreased, and epithelial Bmp4 expression is high (Fig. 6B, bottom). Normal lung lobe formation indicates that early airway branching proceeds normally in Fgf9^−/− lungs, presumably occurring before the onset of critical mesenchymal depletion.

In vitro studies of mouse lung have shown that endoderm, but not the surrounding mesenchyme, migrates towards foci of Fgf10 expression (Weaver et al., 2000). Unknown epithelial-mesenchymal interactions must occur to allow mesenchyme to continue to surround endoderm as it migrates (Weaver et al., 2000). Pleural Fgf9 could mediate this process by stimulating mesenchymal proliferation and/or mesenchyme migration. In vitro studies have shown that exogenous FGF9 can stimulate cell migration into the developing ovary from neighboring mesonephric tissue (Colvin et al., 2001).

Development and differentiation of distal, respiratory airways may involve mechanical and molecular processes distinct from those regulating proximal lung development. Loss of epithelial Bmp4 expression in Pod1^−/− mice (Quaggin et al., 1999) and loss of BMP4 activity in mice expressing a dominant negative BMP receptor (Weaver et al., 1999) both result in failure of distal lung development. Pod1^−/− lungs exhibit moderately reduced proximal airway branching, and a profound defect in terminal branching, showing no alveolar ducts, sacs or alveoli. In contrast, significant formation of distal airspaces and differentiation of Type I and Type II pneumocytes occur in Fgf9^−/− lung, suggesting preserved Bmp4 expression. Interestingly, Fgf10 expression is not altered in Pod1^−/− lung (Quaggin et al., 1999).

Reduced airway branching is observed in several knockout mice. Fgf9^−/− mice are unique in that, other than reduced branching, they exhibit relatively normal lung morphology. Loss of BMP4 activity (Weaver et al., 1999), or loss of Pod1 (Quaggin et al., 1999) or Shh (Pepicelli et al., 1998), each results in reduced branching with dysmorphic lung development. Our findings demonstrate that differentiation of proximal and distal lung epithelium and significant development of distal airways can occur in the absence of Fgf9. The small, but well-formed lungs in Fgf9^−/− mice suggest that titration of Fgf9 expression could effectively regulate lung size in vivo.

Analysis of male-to-female sex reversal in Fgf9^−/− mice indicates that Fgf9 also regulates growth of the embryonic testis, in part by stimulating mesenchymal proliferation (Colvin et al., 2001). As in developing Fgf9^−/− lung, differentiation of all testicular cell types can occur without Fgf9. These results suggest that Fgf9 regulates embryonic organ size in at least two systems by regulating mesenchymal expansion.

Reciprocal signaling between epithelial and mesenchymal FGFs drives embryogenesis of multiple organs (Ohuchi et al., 2000). Epithelial tissues tend to express FGFRb splice forms, while mesenchymal tissues tend express FGFRc splice forms (Orr-Urtreger et al., 1993). FGF7 and FGF10 activate only the ‘b’ splice form of FGFR2 (FGFR2b) (Igarashi et al., 1998; Ornitz et al., 1996), while FGF9 preferentially activates FGFRc splice forms (Ornitz et al., 1996; Santos-Ocampo et al., 1996). Lungs of Fgfr3/Fgfr4 double-null mice are histologically normal at birth (Weinstein et al., 1998), indicating that the lethal Fgf10^−/− and Fgf9^−/− lung phenotypes must involve decreased signaling through FGFR1 and/or FGFR2. These findings suggest that, in the developing lung, FGF10 and FGF7 signal from mesenchyme to epithelium through FGFR2b, while FGF9 signals from epithelium and pleura to mesenchyme, primarily through FGFR1c or FGFR2c (Fig. 6A, top). Similar epithelial-mesenchymal reciprocal signaling between FGF10 in mesenchyme and FGF4, FGF8, FGF9 and FGF17 in the overlying AER regulates embryonic mouse limb bud outgrowth (Lewandoski et al., 2000; Martin, 1998). Like Fgf9 and Fgf10 in lung, generation of null mutants has demonstrated that Fgf8 and Fgf10 are essential for limb embryogenesis (Fig. 6A, bottom; Lewandoski et al., 2000; Min et al., 1998; Sekine et al., 1999).

Overexpressing Shh in airway epithelium results in increased mesenchymal proliferation and increased mesenchymal tissue (Bellusci et al., 1997a), suggesting that FGF9 and hedgehog signaling may jointly regulate mesenchymal development in embryonic lung (Fig. 6B, top). Vascular defects in Shh^−/− mice (Pepicelli et al., 1998), indicate that Shh may regulate development of mesenchymal vasculature. In the developing limb, Shh expression in the zone of polarizing activity (ZPA) is required to maintain Fgf9 expression in the apical ectodermal ridge (AER; Sun et al., 2000; Fig. 6A, bottom). This finding, and normal Shh expression in Fgf9^−/− lung, suggest that Shh may also act upstream of Fgf9 during lung embryogenesis. Curiously, data from Shh^−/− lungs (Pepicelli et al., 1998) and from transgenic lungs that overexpress Shh (Bellusci et al., 1997b) indicate that Shh negatively regulates Fgf10 expression in developing lung (Fig. 6B, middle).

Loss of functional thoracic cavity volume in utero, for example, because of diaphragmatic hernia, fetal pleural effusions or intrathoracic neoplasms, limits embryonic lung growth (Hasleton, 1996b). Our data indicate that Fgf9^−/− lung hypoplasia represents a primary defect in lung growth. No abdominal contents were ever found in the thoracic cavity of Fgf9^−/− embryos, and Fgf9^−/− mice could breathe and fill the lungs with air (data not shown)., indicating that the diaphragm is intact and functional. Because Fgf9 is expressed in the developing pleura, it is possible that malformation of the pleura could result in leakage of lung fluid into the thoracic cavity, constricting growth of thoracic organs. Selective loss of lung mesenchyme and the observation of cardiac dilation in some Fgf9^−/− mice suggest otherwise. In addition, preservation of tight junctions between pleural cells and of the pleural basement membrane suggests that the barrier function of the pleura is retained in Fgf9^−/− lungs. Normal vascular histology, vWF expression and capillary formation in Fgf9^−/− lungs indicate that lung hypoplasia in Fgf9^−/− mice is not secondary to gross vascular dysgenesis.

Fgf9^−/− lungs fail to fill the thoracic cavity and lack the sharp contours of normal lungs, suggesting that Fgf9 is an important regulator of lung size and shape. Pleural expression of Fgf9 could potentially be altered by physical contact with structures surrounding the developing lung (such as the wall of the thoracic cavity in normal lung or intestine displaced into the thoracic cavity secondary to diaphragmatic hernia), allowing precise fitting of the developing lung into the available space. In support of this model, rat visceral pleural cells in vitro exhibit altered gene expression when subjected to fluid shear stress simulating the rubbing of the lung pleura against the chest wall (Waters et al., 1997).

Our data indicate that Fgf9 stimulates fetal lung cell proliferation and regulates fetal lung size. These results suggest that exogenous FGF9 could potentially enhance human lung growth after surgical lung resection, a commonly used treatment for lung carcinoma. In addition, aberrant Fgf9 upregulation could contribute to cancerous lung growth. Like prostate cancer, most lung carcinoma is epithelial in origin. During normal lung embryogenesis, epithelial Fgf9 apparently stimulates mesenchymal proliferation by activating mesenchymal FGFRc splice forms. However, in vivo analysis of malignant transformation of prostate epithelial cells has demonstrated a switch from exclusive expression of FGFRb in epithelial cells to exclusive expression of FGFRc; upregulation of epithelial FGFs was also observed (Yan et al., 1993). In the lung, a similar upregulation of epithelial Fgf9, and aberrant expression of FGFRc splice forms in the epithelium, could create an autocrine loop driving pathologic epithelial proliferation.

In summary, we have demonstrated that Fgf9 is essential for lung embryogenesis and for postnatal survival in mice, and we have identified a novel role for Fgf9 in regulating lung mesenchyme. We provide evidence that Fgf9 affects lung size by stimulating mesenchymal proliferation and mesenchymal Fgf10 expression. Similar to Fgf function in other developmental systems, FGF9 signaling from the epithelium and reciprocal FGF10 signaling from the mesenchyme appear coordinately to regulate airway branching and lung growth in mouse embryos. As Fgf9^−/− mice die at birth, study of these mice could not address Fgf9 function in postnatal lung. Failure of alveogenesis in Fgfr3/Fgfr4 double-null mice (Weinstein et al., 1998), and in vitro activation of FGFR3 and FGFR4 by FGF9 (Ornitz et al., 1996), suggest a possible role for Fgf9 in alveogenesis, which occurs postnatally in mice.

We are grateful for technical assistance from E. Taylor, M. Scott, A. Johnson and B. Coleman (histology), E. Spinaio, X. Hua, L. Li, H. Walker and M. Wuerffel (animal husbandry), M. Veith (scanning electron microscopy), L. LaRose and M. Levy (transmission electron microscopy) and J. Waggoner (figure preparation). We thank B. Hogan for the Fgf10 in situ probe, A. McMahon for the Shh in situ probe, J. Whitsett for the SP-C antibody and F. DeMayo for CCSP antibody. Our thanks to D. deMello and E. Crouch for helpful discussions, and to R. Pierce, S. Brody, R. Mecham and R. Cagan for review of the manuscript. This work was supported by grants from Monsanto/Searle and the American Heart Association (# 974-0221N).