Ectodysplasin target gene Fgf20 regulates mammary bud growth and ductal invasion and branching during puberty

Mammary gland development begins with the appearance of epithelial placodes that invaginate, sprout, and branch to form small arborized trees by birth. The second phase of ductal growth and branching is driven by the highly invasive structures called terminal end buds (TEBs) that form at ductal tips at the onset of puberty. Ectodysplasin (Eda), a tumor necrosis factor-like ligand, is essential for the development of skin appendages including the breast. In mice, Eda regulates mammary placode formation and branching morphogenesis, but the underlying molecular mechanisms are poorly understood. Fibroblast growth factor (Fgf) receptors have a recognized role in mammary ductal development and stem cell maintenance, but the ligands involved are ill-defined. Here we report that Fgf20 is expressed in embryonic mammary glands and is regulated by the Eda pathway. Fgf20 deficiency does not impede mammary gland induction, but compromises mammary bud growth, as well as TEB formation, ductal outgrowth and branching during puberty. We further show that loss of Fgf20 delays formation of Eda-induced supernumerary mammary buds and normalizes the embryonic and postnatal hyperbranching phenotype of Eda overexpressing mice. These findings identify a hitherto unknown function for Fgf20 in mammary budding and branching morphogenesis.


Results
Fgf20 is expressed in the embryonic mammary buds. We have previously shown by microarray profiling that a short treatment with recombinant Eda protein upregulates the expression of Fgf17 and Fgf20 in the mammary buds of E13.5 Eda−/− embryos ex vivo 33 . Quantitative RT-PCR was used to validate these findings, as well as the expression of Fgf4 and Fgf9, two Fgf family members reported to be expressed in the mammary buds 40 but not upregulated by Eda in the microarray. In line with the microarray results, after 4 hours Eda-treatment, of these only Fgf17 and Fgf20 were upregulated 5.8-fold (p = 0.042) and 3.8-fold (p = 0.019), respectively (Fig. 1a). However, analysis of the absolute mRNA quantity indicated that Fgf17 is expressed at a very low level, and thus the role of Fgf17 in mammary gland development was not analyzed further.
In order to analyze expression of Fgf20 in embryonic mammary glands, we took advantage of the Fgf20-LacZ knock-in allele 41 and performed X-gal staining on Fgf20 LacZ/+ embryos between E10.5 and E18.5. Expression of Fgf20-LacZ was detected earliest at ~E11.25 in the mammary bud 1 (data not shown), and at E11.5 in the buds 1 and 3 (Fig. 1b). At E13.5, Fgf20-LacZ expression was detected in all mammary buds (Fig. 1b) and accordingly, in situ hybridization with an Fgf20 specific probe showed positive signal in wildtype embryos at the same stage (Fig. 1d). The Fgf20-LacZ expression was still relatively strong in the mammary buds at E15.5 ( Fig. 1e) but was substantially downregulated at E16.5 (Fig. 1f,g). At E18.5, no expression of Fgf20-LacZ could be detected in the mammary glands by X-gal staining (Fig. 1h) or immunohistochemical staining with anti-β-galactosidase antibody, although expression in hair follicles was readily observed (Fig. S1a), as reported previously 39 . At postnatal stages, expression of Fgf20-LacZ was assessed by X-gal staining and anti-β-galactosidase antibody in mammary glands of 3-, 5-and 7-week-old Fgf20 LacZ/+ and Fgf20 LacZ/LacZ mice, and by qRT-PCR in samples from 3 different regions (proximal to nipple, middle, and distal to nipple) of 5-week old glands. No expression was detected in the postnatal mammary gland by any of the methods used ( Supplementary Fig. S1).
Eda levels influence the expression of Fgf20 in vivo. The observation that Eda induced the expression of Fgf20 in the embryonic mammary buds ex vivo prompted us to study the influence of Eda on Fgf20 expression levels in vivo by analyzing the Fgf20-LacZ expression in Eda null and Eda-overexpressing (K14-Eda) embryos. In Eda −/− embryos there was a slight delay in the onset of Fgf20-LacZ expression at E11.5 followed by somewhat decreased signal at E12.5 compared to control or K14-Eda embryos (Fig. 2a,b). At E13.5-E14.5 expression in K14-Eda embryos appeared more intense (Fig. 2c,d), and at E15.5, Fgf20-LacZ expression levels correlated with the Eda status (Fig. 2e). Together, these data show that loss-and gain-of Eda influence Fgf20-LacZ expression, Figure 1. Fgf20 is induced by Eda and is expressed in embryonic mammary glands. (a) qRT-PCR analysis of Fgf4 (n = 4), Fgf 9 (n = 4), Fgf17 (n = 6) and Fgf20 (n = 7) expression in E13.5 Eda −/− mammary buds after 4 h treatment with Eda protein ex vivo. Values represent mean ± SD. (b,c) X-gal-stained whole mounts of Fgf20 LacZ/+ embryos at E11.5 (b) and E13.5 (c) showing positive staining in the developing mammary buds (numbered). (d) In situ hybridization of a WT embryo with an Fgf20 specific probe at E13.5. (e,f) X-Gal stained whole mount of E15.5 whole embryo (e) and dissected skin of E16.5 embryo (f) showing staining in the developing mammary buds (numbered) and hair follicles. (g,h) Representative figures of histological sections of X-Gal whole mount-stained mammary glands of Fgf20 LacZ/+ embryos at E16.5 (g) and E18.5 (h). *p < 0.05. At least two litters of Fgf20 LacZ/+ embryos per stage were analyzed. *p < 0.05. mb, mammary bud. although modestly, yet clearly cues other than Eda have a more prominent impact on Fgf20 expression during embryogenesis. The Wnt pathway is the most likely positive regulator: the murine Fgf20 promoter is known to be highly responsive to β-catenin/Lef1 in promoter-reporter assays 39 .
Absence of Fgf20 compromises mammary bud formation. To elucidate the role of Fgf20 in mammary gland development, we first analyzed the expression of placode markers Wnt10b and PTHrP by RNA in situ hybridization in the mammary buds of Fgf20 LacZ/+ and Fgf20 LacZ/LacZ mice (Figs 3 and 4). At 46-48 somite stage (E11.5-E11.75) Wnt10b expression in the two genotypes was indistinguishable indicating that Fgf20 deficiency does not impede induction of mammary gland development (Fig. 3a). At E12.5, however, Wnt10b expression domain appeared smaller in Fgf20 LacZ/LacZ embryos, the difference being most pronounced in bud 3 (Fig. 3b), which is the first bud to form 11 . Quantification of the Wnt10b expression domain confirmed a significant difference between the two genotypes (p = 0.0007) (Fig. 3b'). At E13.5, the same was observed with the PTHrP probe, or when Fgf20-LacZ expression was assessed by X-gal staining (Fig. 4). For a more detailed morphological analysis, EpCAM-stained mammary buds 3 were visualized by whole mount confocal microscopy in 3D (Fig. 3c,d). Quantification revealed that Fgf20 LacZ/LacZ buds were substantially smaller than control buds at E13.5 (p = 1.098E-13) and E15.5 (p = 2.234E-6). In attempt to gain insights into the molecular mechanisms underlying the Fgf20 LacZ/ LacZ bud phenotype, we analyzed expression of Edar, Lef1, and Dkk4 at E12.5, and Lef1 protein at E13.5. No gross difference in Edar, or Lef1 expression was detected, but Dkk4 expression was somewhat reduced in Fgf20 LacZ/LacZ embryos ( Supplementary Fig. S2), as previously shown in hair placodes (Huh et al. 39 ).
Scientific RepoRts | 7: 5049 | DOI:10.1038/s41598-017-04637-1 nipples, and at least in the neck region, a ductal tree was readily observed in compound mutants ( Supplementary  Fig. S3). In conclusion, in the absence of Fgf20, all mammary buds formed, yet a clear reduction in bud size and a slight delay in appearance of supernumerary mammary buds in K14-Eda embryos was evident.

Absence of Fgf20 delays ductal growth in puberty. Macroscopic analysis of pubertal and adult
Fgf20 LacZ/LacZ females revealed the presence of the normal number of nipples. To examine the impact of Fgf20 deficiency on postnatal mammary morphogenesis, 4 th mammary glands of 5-week-old WT and Fgf20 LacZ/LacZ were analyzed ( Fig. 5a-c). The number of the ductal ends was reduced by 35% (p = 0.018) and TEBs by 51% (p = 0.008) in Fgf20 LacZ/LacZ mice compared to WT controls (Fig. 5d,e). Also, the extent of ductal outgrowth (i.e. penetration to the fat pad) was significantly compromised (p = 0.037) (Fig. 5f). These data clearly show that absence of Fgf20 greatly retards ductal outgrowth during puberty. The ductal characteristics were, however, quite variable among the Fgf20 LacZ/LacZ mice: often the ductal tree was very rudimentary and barely contained any TEBs while in some mice the ductal tree was only modestly affected ( Fig. 5a-c). Quantification of the maximum width of the five largest TEBs/ductal tips in each specimen confirmed a significant difference between Fgf20 LacZ/LacZ and WT mice (p = 0.029) (Fig. 5g). Ki-67 expression analysis in TEBs evidenced a decrease in the number of proliferating cells in Fgf20 mutants (p = 0.0038) (Fig. 5h,i).

No evidence for a systemic pubertal defect in Fgf20
LacZ/LacZ females. We detected Fgf20 expression only in the embryonic mammary glands (see above), yet Fgf20 LacZ/LacZ mammary glands displayed a remarkable postnatal phenotype (Figs 5 and 6). To assess whether the pubertal phenotype could be caused by a systemic defect due to the germline deletion of the Fgf20 gene, we analyzed various parameters in the mutant animals. We found no difference in the onset of puberty, nor in the weight of the animals at the onset of, or during puberty (at 3, 5, or 7 weeks of age), or the weight of ovaries and uteri ( Supplementary Fig. S4). Yet, 18% of 7-week-old Fgf20 LacZ/LacZ females (n = 22) had completely closed vaginas, whereas a similar defect was not observed in WT mice (n = 9). These mice were not used for mammary gland analyses. The estrus cycles analyzed from vaginal smear cytology of WT and Fgf20 LacZ/LacZ females were normal, and serum estradiol levels of the 7-week-old Fgf20 LacZ/LacZ females in diestrus were similar to those of WT littermates ( Supplementary Fig. S4). Finally, we performed mammary fat pad transplantations in which 1 mm 3 pieces of adult Fgf20 +/+ mammary glands were transplanted into the cleared fat pad of 3-to-4-week old WT or Fgf20 LacZ/LacZ females and allowed to grow for 5 weeks before analysis. WT epithelium grew equally well in the fat pad of both recipients ( Supplementary Fig. S4). Collectively, these data indicate that there is no gross systemic defect in Fgf20 LacZ/LacZ females, which could explain the pubertal mammary phenotype.
Absence of Fgf20 normalizes the hyperbranching phenotype of K14-Eda mice. Our data showing that Fgf20 expression levels are modulated by Eda (Fig. 1) and loss of Fgf20 delays ductal growth at puberty (Fig. 5) prompted us to study the effects of Fg20 deficiency on ductal branching at other developmental stages, as well as the crosstalk with the Eda pathway. At E18, the number of ductal ends in the mammary glands of Fgf20 LacZ/LacZ embryos was similar to that of wildtype mice (p = 0.638) (Fig. 6a,a'). However, mammary glands of 3-week-old Fgf20 LacZ/LacZ mice contained somewhat lower number of ductal tips than those of WT controls (p = 0.0321) (Fig. 6b,b'). At 7 weeks of age, the decrease in the ductal outgrowth and number of ductal ends in Fgf20 LacZ/LacZ mice was prominent (p = 0.0039 and p = 0.0051, respectively) (Fig. 6c,c' ,e), even more pronounced than at 5 weeks of age (Fig. 5d). However, at 12 weeks of age, the number of ductal ends was similar in both genotypes (p = 0.363) (Fig. 6d,d').
Consistent with our previous results 34 , the number of ductal ends was significantly higher in K14-Eda mice compared to WT controls at E18 (p = 0.00009) and 3 weeks of age (p = 0.0019) (Fig. 6a,a' ,b,b'). The hyperbranching phenotype was apparent also at 7 (p = 0.034) and 12 weeks of age (p = 0.0004) (Fig. 6c,c' ,d,d'). Surprisingly, even though Fgf20 null mammary glands did not display a growth phenotype at E18, the K14-Eda phenotype was greatly attenuated in Fgf20 LacZ/LacZ background (p = 0.0005) (Fig. 6a'). Also at later stages, loss of Fgf20 normalized the K14-Eda phenotype, although at 7 weeks of age, the difference did not reach statistical significance (p 3wk = 0.0046; p 7wk = 0.1521, p 12wk = 0.0011). These data identify Fgf20 as a critical mediator of Eda in mammary ductal growth and branching.
At late puberty, the terminal end buds of Fgf20 LacZ/LacZ mice are larger and more proliferative.
Since the growth delay of the Fgf20 mutants was most pronounced at 7 weeks of age, we analyzed the ducts and TEBs of Fgf20 LacZ/LacZ and WT glands in more detail at this stage. The architecture of the ducts appeared normal based on all criteria used: histology, hormone receptor expression, the distribution of basal (K14) and luminal (K8) keratins, and the expression of basal cell marker α-SMA ( Supplementary Fig. S5). Accordingly, FACS analysis did not show significant differences in the percentage of luminal (CD29 lo CD24+) or basal (CD29 hi CD24+) cells between WT and the Fgf20 LacZ/LacZ mice at 7 week of age, nor at 3 weeks when the growth phenotype was first evident (Supplementary Fig. S5).
Analysis of TEBs, however, revealed that the epithelium appeared more cellular in Fgf20 LacZ/LacZ mice compared to WT mice (Fig. 7a). TEB area, measured from the carmine alum whole mount images, was larger in Fgf20 LacZ/LacZ mice at the same age (Fig. 7b). Quantification of Ki-67 and cleaved caspase-3 positive cells in TEBs revealed that the proportion of the proliferating cells was significantly higher in Fgf20 LacZ/LacZ mice compared to WT controls (Fig. 7c,c'), but there was no difference in the proportion of apoptotic cells (Fig. 7d,d'). ERα and PR expression was indistinguishable between WT and Fgf20 LacZ/LacZ TEBs (Fig. 7e,f). TEBs consist of a mass of luminal K8+ body cells surrounded by α-SMA+/p63+ cap cell layer. The expression patterns of body and cap cell markers were unchanged in 7-week old Fgf20 LacZ/LacZ mice (Fig. 7g-i) indicating intact TEB architecture and cell identities.

Discussion
In the current study, we have unveiled a role for Fgf20 in two stages of embryonic mammary gland development: budding and branching morphogenesis. Even though Fgf20 was dispensable for mammary placode induction, the buds were smaller in size. The molecular mechanism underlying the bud growth defect remain elusive. Furthermore, loss of Fgf20 delayed, but did not prevent, the formation of supernumerary mammary buds in K14-Eda embryos. Perinatally, Fgf20 null mammary glands did not differ from the WT controls, yet the K14-Eda hyperbranching phenotype was greatly attenuated in Fgf20 null background. The most plausible explanation for these seemingly contradictory findings is redundancy of Fgf20 with other Fgf ligands, the most prominent candidate being Fgf9, a member of the same Fgf subfamily. Fgf9 is expressed in embryonic mammary glands 40 , shares similar biochemical properties with Fgf20 including receptor specificity 1, 2 , and redundancy between these two Fgfs has already been demonstrated in developing teeth 38 , kidney 42 , and cochlea 43 . Other Fgfs reported to be expressed in mammary bud epithelium are Fgf4, Fgf8, and Fgf17 40 , which may further compensate for loss of Fgf20.
Scientific RepoRts | 7: 5049 | DOI:10.1038/s41598-017-04637-1 cochlea, epithelially expressed Fgf20 positively regulates epithelial progenitor proliferation via the mesenchyme, whereas intraepithelial Fgf20 signaling is essential for sensory cell differentiation 41,43 . In hair follicles, Fgf20 is dispensable for placode formation, but is necessary for condensation of the underlying mesenchymal cells, which in turn is required for further follicular downgrowth 39 . The target genes regulated by Fgf20 have remained elusive in all organs studied so far.
We have previously shown that Eda regulates expression of Fgf20 in embryonic hair follicles and teeth where Fgf20 functions as one of the major downstream effectors of the Eda pathway 38,39 . Here, we identify Fgf20 as a mediator of Eda in the developing mammary glands: absence of Fgf20 delayed formation of supernumerary buds and normalized the hyperbranching mammary phenotype of K14-Eda mice, an effect maintained until adulthood. However, our data implicate the existence of other downstream targets of Eda besides Fgf20, since at E18 and at the onset of puberty, the ductal trees of Eda-null mice are more severely affected than those of Fgf20 LacZ/LacZ mice 34 . Our earlier studies have identified several other Eda-induced factors that can enhance branching morphogenesis such as PTHrP, Egfr ligands, and Wnt pathway agonists 34 . Hence, the Eda-null and K14-Eda branching phenotypes are likely the combinatorial result of changes in the expression level of multiple Eda target genes.
Our data show that Fgf20 has a considerable impact on postnatal mammary morphogenesis since its absence led to defective TEB formation and delayed ductal invasion during puberty. However, the ductal growth defect was transient: the ductal trees caught up to the WT glands between 7 and 12 weeks of age. We propose that this also explains the counterintuitive finding of increased cell proliferation in Fgf20 LacZ/LacZ TEBs at 7 weeks of age. In WT glands, the percentage of proliferative cells in the TEBs decreases between 3 weeks of age and late puberty (7 weeks) 26 , whereas Fgf20 LacZ/LacZ mammary glands begin their growth burst at 7 weeks of age.
The embryonic phenotype and the subtle reduction in the number of branches in 3 weeks old Fgf20 LacZ/LacZ mice implicates that the defect underlying the pubertal ductal phenotype may arise before puberty. We were unable to detect Fgf20 expression during puberty, not even by qRT-PCR, a finding in line with a recent study assessing Fgf20 expression in mammary glands of 3, 5, and 10 week old mice 27 . Thus, it is plausible that Fgf20 deficiency during embryogenesis leads to qualitative changes in the mammary stem/progenitor cells that fully manifest only during puberty. Fittingly, a recent study implicated epithelial Fgfr1/2 signaling in proper mammary stem cell function during development 26 . However, we cannot exclude the possibility that Fgf20 is expressed during puberty in a rare cell population that escaped our analysis. To answer the question whether Fgf20 has a role in pubertal development independent of its embryonic function must await for the generation of a conditional Fgf20 mouse.
The mammary phenotype of Fgf20 LacZ/LacZ mice resembles the phenotypes generated by K14-Cre-mediated deletion of Fgfr1 26 and MMTV-Cre-mediated (mosaic) deletion of Fgfr2 24 , which both display compromised TEB formation, reduced number of branch points, and pubertal ductal outgrowth defect that normalizes in the adulthood. A complete failure in TEB maintenance is observed in mice inducibly overexpressing a transgene encoding a soluble form of Fgfr2b 25 . Interestingly, upon cessation of transgene expression 6 weeks after its induction, TEBs reform and branching is resumed. These data are suggestive of Fgfr signaling being essential for the functionality rather than survival of mammary stem/progenitor cells driving TEB formation and ductal invasion.
Pubertal ductal morphogenesis is a complex hormone regulated process, which involves cellular functions such as proliferation, apoptosis, migration, ECM degradation, and a tight interplay between epithelial and mesenchymal compartments 6,15 . A great number of genetically manipulated mice, and experiments using slowly-releasing protein pellets in vivo, are known to cause a pubertal mammary phenotype 23,44 . These studies show that loss of tissue integrity in TEBs readily leads to ductal outgrowth defects. However, this is unlikely the case in Fgf20 LacZ/LacZ mice, as the expression pattern of body and cap cell markers was unaltered. Another important class of pubertal phenotypes is caused by loss-or gain-of-function of matrix remodeling enzymes such as matrix metalloproteinases (MMPs), which regulate ductal invasion and branching via their ability to sculpt the ECM 45,46 , and Fgfr1/2 stimulation has been shown to induce the expression of Mmp3 and Mmp9 in several breast cancer and immortalized mammary epithelial cell lines [47][48][49][50][51] .
In conclusion, our results identify a hitherto unknown function for Fgf20 in both embryonic and postnatal mammary gland morphogenesis. Our data suggest that compromised Fgf20 signaling during embryogenesis results in qualitative changes in TEBs that are thought to harbor the majority of stem cells driving branching morphogenesis during puberty 52,53 . To our knowledge, in addition to Fgf10 11 , Fgf20 is the only Fgf family member with a proven in vivo function in mammary gland development. Furthermore, we discovered Fgf20 as an important mediator of Eda in mammary gland budding and branching morphogenesis. Future studies should shed light on the molecular mechanisms downstream of Fgf20 in mammary gland morphogenesis.

Materials and Methods
Mice. The generation and genotyping of Fgf20 LacZ/LacZ , K14-Eda, and Eda −/− (Tabby; Jackson Laboratories, stock no 000314) mouse strains have been described previously 41,54 . Fgf20 LacZ/LacZ and K14-Eda mice were maintained in the C57Bl/6 background (K14-Eda > 10 generations and Fgf20 LacZ/LacZ > 5 generations) and Eda −/− mice in the B6CBA background. Embryonic ages were defined based on the appearance of vaginal plug and external criteria including limb morphogenesis 55 . The sex of embryos older than E14 was defined by PCR with Sry-specific primers or anatomy, and only female mice were used for analysis unless otherwise stated. All mouse studies were approved and carried out in accordance with the guidelines of the Finnish national animal experimentation board.
In situ hybridization. For whole mount in situ hybridization with digoxigenin-labeled RNA probes, E11.5-E13.75 embryos were fixed in 4% PFA overnight at 4 °C and dehydrated with rising methanol series. In situ hybridization was performed with inSituPro robot (Intavis AG) as previously published 29,38 or manually using a similar protocol. The digoxigenin-labeled RNA probes for Wnt10b, Edar, Dkk4, Lef1, PTHrP and Fgf20 have been described earlier 34 ; Fgf20 probe corresponded to the open reading frame. BM Purple AP substrate Precipitating Solution (Boehringer Mannheim) was used for detection of digoxigenin-labeled RNA probes. Radioactive in situ hybridization was performed on paraffin sections using 35 S-UTP labeled (Amersham) probe specific to Fgf20 as described 38 . Immunohistochemical stainings. For immunohistochemical and hematoxylin-eosin stainings, the 4 th mammary glands of WT and Fgf20 LacZ/LacZ mice were dissected, spread on microscope slides, and fixed with 4% PFA overnight at 4 °C. Alternatively, 13.5 trunks were dissected. The samples were dehydrated, embedded in paraffin, and 5 µm sections were cut. Slides were deparaffinized by standard methods. In immunohistochemical stainings antigen retrieval was performed by heating the slides in microwave oven in TE buffer, pH 9.0 (keratin-8 (K8), keratin-14 (K14), progesterone receptor (PR) and estrogen receptor α (ERα) stainings), or in 10 mM sodium citrate buffer pH 6.0 (β-Galactosidase, cleaved Caspase-3, Ki-67, α-smooth muscle actin (α-SMA), Lef1, and p63 stainings). Primary and secondary antibodies used are listed in Supplementary information. Samples were imaged with a Zeiss Axio Imager M2 microscope equipped with an AxioCam HRc camera (Zeiss) and processed in Photoshop.

Mammary bud area and volume quantification. Wnt10b expression area was quantified manually from
images with the help of Fiji ImageJ software. For whole-mount immunofluorescence staining E13.5 and E15.5 mouse embryos were fixed in 4% PFA at 4 °C overnight. After washing the samples with PBS for 3-4 hours, they were permeabilized with 0.3% TritonX-100 in PBS for 1-2 hours at room temperature, blocked (5% normal donkey serum, 0.5% BSA, and 0.3% TritonX-100 in PBS) for 1 h, and incubated at 4 °C with rat anti-mouse CD326 (EpCAM, BD Pharmingen, 552370, 1:1,000) and 10 µg/ml Hoechst 33258 (Molecular Probes/Invitrogen) in blocking buffer for 2 days. EpCAM was detected with an Alexa Fluor 647 -conjugated secondary antibody (Molecular Probes/Invitrogen). The ventral skin around mammary gland 2 and 3 was dissected and mounted with Vectashield (Vector Laboratories) and visualized using a Zeiss LSM700 laser scanning confocal microscope. For mammary placode and bud volume quantification, the area of mammary primordium was outlined manually based on EpCAM expression and bud morphology. Surface rendering and volume quantification were performed with Imaris 8.3 software (Bitplane).
Mammary cell preparation, cell labelling, and flow cytometry. Single cell suspension of mammary gland was prepared according to the protocol modified from Shackleton et al. 58 . Briefly, the 4 th mammary glands were cut into small pieces after removal of the lymph node. The tissues were digested in a mixture of 5 ml collagenase I buffer (10% FBS, 100 mg/ml streptomycin, 10 U/ml penicillin, 300 U/ml collagenase I (ThermoFisher) and 100 U/ml hyaluronidase (Sigma) in DMEM/F12 (1:1) medium for 1-2 hours at 37 °C with moderate shaking. The cell suspension was washed in PBS and digested further in 0.25% trypsin-EDTA for 5-10 minutes. The red blood cells were removed by incubation in red blood cell lysing buffer (Biolegend) on ice for 5 minutes. The single cell suspension was passed through 40 µm cell strainer (BD Bioscience) before stained with the mixture of antibodies on ice for 30 minutes. After washing in PBS, the dead cells were labeled with Fixable Viability Dye eFluor 780 (eBioscience) for 30 minutes on ice. Flow cytometry was carried out by BD LSR II, and data analysis was done by Flowjo. Monitoring the onset of puberty, estrous phase, and measurement of estradiol. Onset of puberty was assessed by monitoring the vaginal opening (VO) by visual examination of vulva 59 every morning 5 days/week (Mon-Fri) starting at the age of 18 days until the appearance of VO. In case of VO occuring during the weekend, the earliest, latest, and average times of VO were defined and separate comparisons of Fgf20 LacZ/LacZ and WT mice were done using average VO time as well as extreme VO times (eg. VO WTlatest vs. VO Fgf20LacZ/LacZ earliest and vice versa).
Estrus phase was defined by examining the vaginal cytology collected by vaginal lavage with PBS using a small pipet and stained by crystal blue as previously described 60 . For monitoring the regularity of estrus cycles, 7-week and 12-week-old Fgf20 LacZ/LacZ and WT females were examined 5 days/week in the mornings for at least two weeks.
Estradiol levels were measured from serum of 7-week-old mice in diestrus by highly sensitive gas chromatography-tandem mass spectrometry 61 . In case of obtaining zero value from the measurement (n = 3 in both WT and Fgf20 LacZ/LacZ ), value equal to ½ LLOD (lower limit of detection) of estradiol (0.15 pg/ml) was used for the sample 61 . Mice were sacrificed, blood samples were immediately taken by heart puncture and kept overnight at 4 °C. Mammary glands were used for FACS analysis and immunohistochemistry and uterus and ovaries were carefully dissected and weighted. Serum was dissociated the following day by centrifugation in at 3000 rpm at 4 °C. Minimum of 250 µl of serum was required for mass spectrometry analysis.
Mammary fat pad transplantations. For mammary fat pad transplantations, 3-4 week old WT (n = 6) and Fgf20 LacZ/LacZ (n = 5) recipient females were anesthetized and the fat pad of left 4 th mammary gland was cleared until the lymph node as described 62 . ~1 mm 3 pieces of adult (12-13-week-old) WT donor (n = 4) mammary glands were transplanted into cleared fat pads. Five weeks later transplanted mammary glands were collected, stained by Carmine alum, and ductal ends quantified.
Statistical analysis. P-values were calculated with unpaired t-test assuming unequal variances unless otherwise stated.
Data availability. The datasets generated during the current study are available from the corresponding author on reasonable request.