Role of FGF10/FGFR2b signaling during mammary gland development in the mouse embryo

Mammary gland formation in the mouse initiates around E10 on the surface ectoderm of both lateral flanks of the embryo (Turner and Gomez, 1933). By E11-E12 five mammary placodes are detected as thickenings of the ectoderm. The placodes develop into bud-like structures that are located at precise points along the antero-posterior axis of the embryo. The position of the mammary buds, three thoracic and two inguinal, are reproducible between embryos suggesting a tight genetic control of their induction. Previous reports suggest that mammary placodes form by the migration of ectodermal or epidermal cells, rather than by cell proliferation, along a putative line running in an anterior to posterior direction just dorsal to the limb buds (Balinsky, 1950; Propper, 1978).

Little is known about the genes that regulate the initial phases of mammary placode development. The transcription factor Lef1, an effector of WNT/β-catenin signaling, is the earliest known marker of mammary placode formation. Its inactivation in vivo leads to embryos with only a single pair of inguinal placodes (van Genderen et al., 1994). The development of mammary buds, in both female and male mouse embryos, is dependent upon signaling by parathyroid hormone related-protein (PTHrP) through its receptor (PTHR1), which are expressed in the epithelium and condensed mammary mesenchyme respectively. In the absence of PTHrP, or its receptor, mammary buds fail to elongate and branch into the primitive fat pad of female embryos, or undergo the expected androgen-mediated apoptosis in males (Wysolmerski et al., 1998; Dunbar et al., 1999).

The fibroblast growth factor (FGF) family comprises at least 22 members, many of which have been implicated in multiple aspects of vertebrate development [for review see Ornitz and Itoh (Ornitz and Itoh, 2001)]. In particular, FGF10 has been associated with instructive mesenchymal-epithelial interactions, such as those that occur during branching morphogenesis. For example, in the developing lung, Fgf10 is expressed in the distal mesenchyme at sites where prospective epithelial buds will appear. Moreover, its dynamic pattern of expression and its ability to induce epithelial expansion and budding in organ cultures have led to the hypothesis that FGF10 governs the directional outgrowth of lung buds during branching morphogenesis (Bellusci et al., 1997). Furthermore, FGF10 was shown to be a potent chemoattractant for the distal lung epithelium (Park et al., 1998; Weaver et al., 2000). Consistent with these observations, mice deficient for Fgf10 show multiple organ defects including lung agenesis (Min et al., 1998; Sekine et al., 1999; Ohuchi et al., 2000).

The mammalian Fgf receptor family comprises four genes (Fgfr1 to Fgfr4), which encode at least seven prototype receptors. Fgfr1, 2 and 3 encode two receptor isoforms (termed IIIb or IIIc) that are generated by alternative splicing, and each bind a specific repertoire of FGF ligands (Ornitz et al., 1996). FGFR2-IIIb (FGFR2b) is found mainly in epithelia and binds four known ligands (FGF1, FGF3, FGF7 and FGF10) which are primarily expressed in mesenchymal cells. While mice null for the Fgfr2 gene die early during embryogenesis, those that are null for the Fgfr2b isoform, but retain Fgfr2c, survive to birth (Arman et al., 1998; Xu et al., 1998; De Moerlooze et al., 2000; Revest et al., 2001). Mice deficient for Fgfr2b show agenesis and dysgenesis of multiple organs indicating that signaling through this receptor is critical for mesenchymal-epithelial interactions during early organogenesis.

Mammary gland development has been studied extensively in the post-natal animal, but less is known about the embryonic stages. We have investigated the initial phases of mammary placode development, and demonstrate using molecular markers and scanning electron microscopy that the placodes form asynchronously. Placode 3 is the first to appear, followed by placode 4, then placodes 1 and 5 and finally placode 2. The role of FGF10/FGFR2b signaling in the epithelial/mesenchymal interactions that characterize embryonic mammary gland development is demonstrated through the analysis of the mammary gland phenotypes of Fgf10^–/– and Fgfr2b^–/– embryos.

Radioactive and whole-mount in situ hybridization protocols were based on previously described methods (Winnier et al., 1995). The following mouse cDNAs were used as templates for the synthesis of digoxigenin or ³⁵S-labeled riboprobes: a 360 bp Lef1 probe provided by Dr Grosschedl, a Fgfr2-TK and IIIb probes previously described (De Moerlooze et al., 2000), a 584 bp Fgf10 cDNA (Bellusci et al., 1997), a 1.5 kbp Bmp4 probe (Winnier et al., 1995) and a 622 bp Fgf7 cDNA (kindly provided by Dr Mason).

E11.5 and E12.5 mouse embryos (C57BL/6) were extracted quickly from the uteri, washed 6 times in filtered PBS and fixed in a solution of sodium cacodylate 0.1 M pH 7.6/glutaraldehyde 2% at room temperature for 1 hour and then overnight at 4°C. They were washed three times in 0.2 M sodium cacodylate for 1 hour at room temperature and transferred in a solution of sodium cacodylate 0.1 M pH7.6/OsO₄ 0.1% for 1 hour at room temperature. After a 5 minute wash in distilled water, the embryos were dehydrated in graded ethanols (70% to 100%) and then in amyl acetate (30% to 100%). They were critical-point dried in liquid carbon dioxyde, mounted on aluminum stubs and coated with gold.

Fgf10^–/– and Fgfr2b^–/– embryos were generated as previously described (Sekine et al., 1999; De Moerlooze et al., 2000) and were on the C57BL/6 background. C57BL/6 or wild-type littermates mice were used as control embryos at different stages of development. The number of Fgf10^–/– embryos used in this study at the different stages were as follows: E11.5 (n=5), E12.5 (n=2), E13.5 (n=2), E14.5 (n=3 females), E18.5 (n=9 females). The number of Fgfr2b^–/– embryos used in this study were as follows: E11.5 (n=3), E12.5 (n=11), E13 (n=6), E14.5 (n=1 female), E15.5 (n=2 females), E16.5 (n=2 females).

Embryos were removed at E10.5 and E11.5. At E11.5, the heads were surgically removed and the remaining body of the embryos cut into halves along the dorsal-ventral axis. Embryos were placed on Nucleopore filters, which were then laid on the surface of 500 μl F12: DMEM medium containing 50 Units/mg penicillin and streptomycin, 1% glutamine and 10% heat-inactivated fetal calf serum in NUNCLON dishes [technique adapted from Lebeche et al. (Lebeche et al., 1999)]. We investigated the local effects of FGF10 on these cultures by implanting heparin beads (Sigma) impregnated with human recombinant FGF10 (Research and Development) (100 μg/ml) in the flanks of the embryos in the area of mammary placode formation. The embryos were usually incubated for 24 hours at 37°C under CO₂ and then fixed for 2 hours in 4% PFA and processed for whole-mount in situ hybridization. BSA-impregnated beads were used as controls.

The mammary gland 4 from Fgf10^–/– (n=3) and wild-type fetuses (n=3) at E18.5 were freshly dissected and transplanted into cleared mammary fat pads of syngenic mice (Medina, 1996). In these experiments, 21-24 days old females were used as transplant recipients. The endogenous mammary epithelium was surgically removed from the fourth inguinal glands to provide a cleared mammary fat pad. Mutant and wild-type mammary glands were transplanted separately into contralateral glands of each recipient (n=3) to ensure an identical host environment. After 4 weeks, the mice were sacrificed and the fourth inguinal mammary glands dissected. The epithelium was stained using Carmin Red as previously described (Faraldo et al., 1998).

Apoptotic cells were detected by the incorporation of terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling (TUNEL) using the ApopTagRPlus In Situ Apoptosis Detection Kit (Oncor, USA) as recommended by the manufacturer.

Previous reports have shown that Lef1 is expressed initially in the epithelium of the mammary placode at E11-E12 (van Genderen et al., 1994; Foley et al., 2001). A detailed temporal analysis using whole-mount in situ hybridization, shows that Lef1 is expressed in the emerging placodes in a dynamic fashion between E11.5 and E11.75 (Fig. 1). Early E11.5 embryos show no Lef1 expression between the fore- and hindlimbs (Fig. 1A), but at a slightly later stage, Lef1 expression appears as a short line (Fig. 1B) which progressively changes through a comet-shape (Fig. 1C) to a disc (Fig. 1D). To gain more insight into the surface features of the mammary placodes, the embryos were visualized by scanning electron microscopy (SEM). At E11.5, placode 3 can be seen as a knob-like structure elevated above the surrounding epidermis (Fig. 1E,F).

Using the shape dynamics of Lef1 expression in the forming placode, together with SEM, the timing of mammary placode formation was determined. Placode 3 is the first to form as shown in Fig. 1, this is followed by placode 4, then placodes 1 and 5 and finally placode 2 (Fig. 2A-C). Analysis by SEM of the E12.5 embryo shows that at this stage of development, five mammary buds are localized on the flank of the embryo that showed distinctive features. At E12.5, the first mammary bud to appear (number 3) was far less elevated above the epidermis than the newly formed mammary bud 2 (Fig. 1E,F and Fig. 2D-F). It appears that when the mammary buds first form they are elevated above the forming epidermis before subsiding.

The phenotypic similarities between Fgf10 and Fgfr2b null mice indicate that FGF10 is the major ligand for FGFR2b. Hybridization of E11.5 and E15.5 embryo sections with a probe that recognizes Fgfr2b and Fgfr2c revealed a high level of Fgfr2 in the epithelium of the mammary bud (Fig. 3A,B). Confirmation that Fgfr2b was the isoform present was shown with IIIb- and IIIc-specific probes (data not shown). Using whole-mount in situ hybridization, a transient expression of Fgf10 was seen between E10.5 and E11 as a fine segmented line extending between the fore- and hindlimb (Fig. 3C,D). This domain of expression corresponds to the territory of the mammary line that is seen as a ridge in other mammals. Vibratome sections at E10.5 showed that this expression corresponds to the most ventral epithelial part of the dermamyotome (Fig. 3E,F). At E11.5, when mammary bud 3 has formed, no significant signal was detected (data not shown). This was confirmed by radioactive in situ hybridization at E11.5 (Fig. 3G). However, by E15.5 a Fgf10 signal was observed in the mammary fat pad precursor localized around the mammary bud (Fig. 3H). At E12.5, sections through mammary placodes were analyzed with radioactive antisense probes to the known ligands of FGFR2b, namely Fgf1, Fgf3, Fgf7 and Fgf10. Expression of Fgf1, Fgf3 or Fgf10 was not detected at this developmental stage. However, Fgf7 was detected in the mesenchyme surrounding the mammary bud at E12.5, but by E15.5 its expression had decreased and extended into the adjacent fat pad precursor (Fig. 3I,J).

Using whole-mount in situ hybridization with a Lef1 probe, the appearance of mammary placodes was monitored in Fgfr2b^–/– embryos taken between E11.5 and E14.5. At E11.5 a characteristic placode 3 was clearly detected in wild-type or heterozygous embryos from the same litter, but not in the homozygous mutants (data not shown). The absence of placode 3 was also confirmed by histological analysis of serial sections. At E12.5, five pairs of mammary buds were clearly observed in the control embryos using the Lef1 probe as a marker (Fig. 4A). By contrast, only a single bud was located in the inguinal region of the homozygous mutant (Fig. 4C). Mammary bud identity was confirmed at the molecular level using a Bmp4 probe (Bmp4 is expressed in the condensed mesenchyme) (Phippard et al., 1996) and by sectioning and histological analysis (data not shown). Although the absence of limbs in the Fgfr2b^–/– embryo makes it difficult to discriminate between the two inguinal buds, its relative position suggested it was number 4. We will therefore refer to this bud in mutant mice as bud 4. At E13, Lef1 expression was no longer detected in the inguinal region indicating the disappearance of bud 4 (compare Fig. 4G and 4E). A TUNEL assay was used to detect apoptotic cells, and this showed that the epithelium of the mutant mammary bud at E12.5 undergoes extensive apoptosis (Fig. 4I,J). The absence of bud 4 at E13 and E14.5 was also confirmed by histological analysis of serial sections (data not shown). In addition, an examination of embryonic skin from wild-type (n=2) and Fgfr2b null (n=2) female embryos at E16.5 failed to detect mammary bud development in the mutants, while five pairs were clearly seen in the wild-type skins (data not shown). In conclusion, placode 4 is the only placode to form in the Fgfr2b^–/– embryos. A bud arising from this placode forms transiently at E11.5 and is then lost within a day through apoptosis. Hence, signaling through FGFR2b by one or more of its ligands is necessary to maintain mammary bud 4 after E12.5 and to induce the other mammary placodes.

FGFR2b is the main receptor for FGF10 as evidenced by the remarkable similarity of phenotypes exhibited by embryos where these genes have been inactivated (De Moerlooze et al., 2000; Ohuchi et al., 2000). To examine the effect of Fgf10 abrogation on mammary placode induction and maintenance, we used whole-mount in situ hybridization with Lef1 and Bmp4 probes on Fgf10^–/– embryos. At E11.5 and E12.5 a single inguinal bud was detected in a similar position to that seen in the Fgfr2b^–/– embryos, and we therefore refer to it as bud 4 (Fig. 5 and data not shown for Bmp4). Sectioning of the E12.5 mutant embryos confirmed the absence of bud 1, 2, 3 and 5 but the presence of bud 4 (Fig. 5H-J). At E13.5, the identification of mammary buds by whole-mount in situ hybridization using a Lef1 probe was more difficult, as this gene is also expressed in the developing hair follicles (Fig. 6A,C). However, close examination allowed discrimination of the mammary buds as large and round structures from the hair follicles. The presence of a typical mammary bud at this stage was also confirmed by sectioning (data not shown). Therefore we can conclude that one pair of inguinal mammary placodes is induced in the Fgf10^–/– mutants, but unlike the Fgfr2b null mice, the corresponding bud is maintained at E13.5.

Fgf10 expression in the mammary fat pad precursor at E15.5 suggested that it might direct the growth of the mammary epithelium toward the fat pad. As bud 4 is maintained in these mice we tested this idea. Skins containing mammary gland 4 from E18.5 Fgf10^–/– female embryos (n=9) were dissected and examined for the extent of ductal branching after staining. Fig. 6G,H shows that in the Fgf10^–/– mammary gland, the epithelial bud penetrated the mammary fat pad (n=9). In 5 cases it was not branched and in 4 it only formed two short branches at the extremity of the main duct. This result is in contrast to wild-type mammary glands that showed secondary and tertiary branches (Fig. 6E,F). During the dissection we noticed that the white adipose tissue forming the fat pad was very thin in the Fgf10^–/– embryos compared to the wild type (data not shown), suggesting that the absence of branching in the mutant epithelium may be due to an abnormal recipient fat pad. The competence of Fgf10-deficient epithelium to undergo ductal branching was tested by transplantation of mammary bud 4 into a normal fat pad. The dissected glands (epithelium + mesenchyme) were grafted into cleared mammary fat pads of wildtype syngenic C57Bl/6 females. Four weeks post-grafting the wild-type epithelium (n=2) was fully competent to invade and branch into a wild-type stroma (Fig. 6I,J), as was the epithelium (n=1) from Fgf10^–/– (Fig. 6K,L). Epithelium from both wildtype and mutant mice grew radially from the implant and showed prominent terminal end buds.

The similarity in mammary placode phenotype between Fgf10^–/– and Lef1^–/– embryos raised the interesting possibility that FGF10 induces Lef1 expression. To test this idea, FGF10-coated beads were grafted onto the flank of E10.5 and E11.5 embryos. After 30 hours of culture the FGF10-coated beads failed to induce Lef1 expression in the surrounding ectoderm of E10.5 or E11.5 embryos (Fig. 7A,B). Importantly, endogenous Lef1 expression corresponding to the normal mammary buds was detected, indicating that endogenous placode formation occurred normally (Fig. 7B). Note that the positions of the beads were dorsal, ventral and coincident with the putative mammary line. The positive control used in this experiment was the induction of Sprouty2 in the lung endoderm by FGF10-coated beads (Mailleux et al., 2001).

In mouse and rabbit embryos, proliferation index studies on epidermis and mammary buds have suggested that placode formation primarily involves cell migration, since cells within the buds showed less proliferation than the surrounding epidermis (Balinsky, 1950). This hypothesis received support from SEM studies with E13.5 rabbit embryos that showed individual cells along a raised mammary line having loose intercellular contacts and exhibiting pseudopodia, a hallmark of cell motility (Propper, 1978).

Until recently, no molecular marker was available to monitor the emergence of mammary epithelial cells during the initial phases of placode development. The transcription factor Lef1 has been shown to participate in Wnt signaling by complexing with β-catenin to form a transcriptional complex that modulates the expression of WNT-responsive genes (Behrens et al., 1996; Eastman and Grosschedl, 1999). Lef1 is expressed at the very early stages of placode formation (van Genderen et al., 1994). We show here that Lef1 has a very dynamic expression pattern that appears to mark the cells that aggregate to form the mammary placode. For each placode, Lef1 expression goes from a line, to a comet shape and finally to a characteristic disc. These observations would suggest that epithelial cells are recruited locally along a mammary line and migrate to a precise location to form the mammary placode.

The timing of placode formation has not been addressed to date, but might be expected to occur sequentially, from anterior (placode 1) to posterior (placode 5) in line with other aspects of mouse development. However, using Lef1 expression to monitor placode formation, they were found to emerge between E11.5-E11.75 in the order 3, 4, 1 and 5 and then 2. This order of placode appearance is supported by SEM observations that show that they initially form an epidermal mound that subsides and become undetectable by E14.5 (Fig. 1, Fig. 2 and S. B. and A. de Maximy, unpublished data). One explanation for asynchrony is that placode formation is autonomous. This is consistent with placode 4 formation in Fgf10^–/– and the Fgfr2b^–/– embryos, although we cannot exclude the idea that the single inguinal placode is actually a fusion of placodes 4 and 5 which fail to separate. The autonomous nature of placode formation is also supported by our recent work on the Extratoes mice, which have a deletion in the Gli3 gene (Hui and Joyner, 1993). The Gli3 null embryos exhibit a lack of induction of placode 3 and 5, while the other placodes are induced normally (S. B. and A. de Maximy, unpublished data).

In contrast to the rabbit, a distinguishable mammary line seen as an elevation on the surface ectoderm has not been observed in the mouse (Bellusci and de Maximy, unpublished data; Balinsky, 1950; Propper, 1978). However, the observation of a line of transient Fgf10 expression in the dermomyotome between E10.5 and E11, prior to placode formation, may be indicative of such a mammary line. The chemoattractant properties attributed to FGF10 in the migration of lung epithelium during branching morphogenesis, suggests a similar role in directing the migration of the epithelial cells along such a hypothetical mammary line. Alternatively, FGF10 may act to specify ectodermal cells destined to form mammary placodes.

The lack of placodes 1, 2, 3 and 5 in both Fgf10^–/– and Fgfr2b^–/– mice was based on the absence of expression of the molecular markers Lef1 or Bmp4 as well as by direct histological examination. Apart from placode 4, these findings suggest a model where FGF10 might regulate Lef1 expression that in turn helps to specify the mammary epithelium. This would be consistent with Lef1^–/– embryos also having a similar mammary phenotype, with a single inguinal placode. The possibility that LEF1 regulates Fgf10 seemed unlikely since its expression precedes Lef1 in mammary placode development. However, FGF10-coated beads did not induce Lef1 expression when placed in the epidermis close to or within the proposed mammary line, indicating that FGF10 is either not involved in Lef1 induction, or it is required earlier to help specify the mammary epithelium. It is also possible that the epithelium is only competent to respond to FGF10 for a short period of time, or that other growth factors act in synergy with FGF10 to induce Lef1 expression.

In the Fgfr2b^–/– embryos, bud 4 is formed but undergoes apoptosis after E12.5, while in Fgf10^–/– embryos this bud is maintained. This finding suggests that an additional FGFR2b ligand is involved in the maintenance of the inguinal mammary bud. In situ hybridization analysis for genes encoding known FGFR2b ligands during this stage of development showed that Fgf7 was the only one detected at E12.5 in the surrounding mesenchyme of the mammary bud (Cunha and Holm, 1996), and therefore may act redundantly with Fgf10 to maintain placode integrity.

In female embryos between E12 and E16, the mammary bud shows a low level of proliferation termed the resting phase. At late E16 proliferation increases and the mammary bud elongates to form the mammary sprout. The sprout grows rapidly downward, penetrating the mammary fat pad precursor tissue that underlies the mammary placode [for reviews see Sakakura and Robinson et al. (Sakakura, 1987; Robinson et al., 1999)]. As the ductal epithelium penetrates the fat pad it begins to branch. PTHrP is expressed in the mammary epithelium and appears to signal to PTHR1 expressed in the surrounding mesenchyme. Disruption of the PTHrP gene leads to an absence of epithelial bud elongation and subsequent ductal branching and to the degeneration of the mammary epithelium (Wysolmerski et al., 1998). As Fgf10 is expressed in the presumptive fat pad, it is plausible that FGF10 could be a downstream target of the PTHrP/PTHR1 signaling pathway. However, our results indicate that FGF10 was not critical for the growth of the epithelium into the mammary fat pad to form the mammary sprout, although this result only applies for the mammary bud 4. FGF10 could certainly play a role in the directional growth of the other buds. Interestingly, the epithelial sprout of the mutant mammary gland did not ramify extensively after penetrating the fat pad, but this abnormality was not apparent when the Fgf10-deficient epithelium was transferred into a wild-type stroma. This suggests that the branching defect is due to a defect in the Fgf10^–/– fad pad that is unable to support proper branching. Consistent with this finding is a recent work demonstrating that FGF10 has a role in the development of adipose tissue, where it plays a role in the differentiation of the pre-adipocytes into adipocytes (N. Itoh unpublished data).

In conclusion, we have shown that Lef1 expression in the mammary placode is dynamic, that mammary placode formation is asynchronous and involves two different signaling pathways, a FGF10/FGFR2b-dependent pathway for placodes 1, 2, 3 and 5 and a FGF10/FGFR2b-independent pathway for placode 4. Our results also suggest that one or several members of the FGF family are involved in mammary bud 4 maintenance and that Fgf10 expression is not crucial for penetration of the mammary duct of bud 4 into the fat pad precursor.

We are grateful to Prof. H. Boulekbache and to Alice Anne de Maximy for the scanning electron microscopy, to Dominique Morineau and Daniel Meur for the photographs and to Prof. Daniel Medina for the mammary gland transplants into the cleared fat pad. This work was supported by the Human Frontier Science Program grant RG0051/1999-M (S. B. and N. I.), the Association pour la Recherche sur le Cancer grant 5214 (S. B.) and the Action Concertée Incitative 81/2001- biologie du developpement et physiologie intégrative (SB).