FGFR2IIIb-MAPK Activity Is Required for Epithelial Cell Fate Decision in the Lower Müllerian Duct

Abstract

Cell fate of lower Müllerian duct epithelium (MDE), to become uterine or vaginal epithelium, is determined by the absence or presence of ΔNp63 expression, respectively. Previously, we showed that SMAD4 and runt-related transcription factor 1 (RUNX1) were independently required for MDE to express ΔNp63. Here, we report that vaginal mesenchyme directs vaginal epithelial cell fate in MDE through paracrine activation of fibroblast growth factor (FGF) receptor-MAPK pathway. In the developing reproductive tract, FGF7 and FGF10 were enriched in vaginal mesenchyme, whereas FGF receptor 2IIIb was expressed in epithelia of both the uterus and vagina. When Fgfr2 was inactivated, vaginal MDE underwent uterine cell fate, and this differentiation defect was corrected by activation of MEK-ERK pathway. In vitro, FGF10 in combination with bone morphogenetic protein 4 and activin A (ActA) was sufficient to induce ΔNp63 in MDE, and ActA was essential for induction of RUNX1 through SMAD-independent pathways. Accordingly, inhibition of type 1 receptors for activin in neonatal mice induced uterine differentiation in vaginal epithelium by down-regulating RUNX1, whereas conditional deletion of Smad2 and Smad3 had no effect on vaginal epithelial differentiation. In conclusion, vaginal epithelial cell fate in MDE is induced by FGF7/10-MAPK, bone morphogenetic protein 4-SMAD, and ActA-RUNX1 pathway activities, and the disruption in any one of these pathways results in conversion from vaginal to uterine epithelial cell fate.

In mammals, the epithelium of the entire female reproductive tract (FRT), namely oviduct, uterus, cervix, and vagina, develops from Müllerian duct epithelium (MDE) (1, 2). Classic tissue recombination studies have established that during development, organ-specific mesenchyme induces differentiation of MDE into epithelia with unique morphology and functions (3, 4). However, the actual mesenchymal factors that determine epithelial cell fate of MDE remained elusive for decades. Previously, we demonstrated that 2 TGFβ superfamily members, bone morphogenetic protein (BMP)4 and activin A (ActA), induced vaginal epithelial differentiation in MDE by activating the expression of ΔNp63 (5), an N-terminal truncated form of transformation related protein 63 (p63), encoded by Trp63. In the lower FRT, expression of ΔNp63 determines the cell fate of MDE to become vaginal or uterine epithelium (5–8). In developing vagina, mesenchymal BMP4 and ActA activate the originally silent ΔNp63 locus through the binding of SMAD transcription factors on the 5′ sequence adjacent to the transcription start site (TSS) of ΔNp63. This SMAD-dependent activation of the ΔNp63 locus requires runt-related transcription factor 1 (RUNX1), a cotranscription factor of SMADs (5). Thus, conditional deletion of Runx1 in MDE inhibited the activation of ΔNp63 and the subsequent differentiation of vaginal epithelium, leading to vaginal adenosis (5). Moreover, SMAD4 was essential for the activation but not the maintenance of the transcriptional activity at the ΔNp63 locus (5). Furthermore, we demonstrated that the transcriptional activity of the ΔNp63 locus requires ΔNp63 protein itself. Accordingly, we concluded that in MDE the transcription of the ΔNp63 locus is activated by the BMP4/ActA-SMAD/RUNX1 pathway and that the activity of the ΔNp63 locus is self-maintained by ΔNp63 protein independently of mesenchymal factors, BMP4/ActA (5).

BMP4, ActA, and RUNX1 were originally identified as candidate factors involved in vaginal epithelial differentiation by microarray analysis on developing uterus and vagina from mice in which MDE was null or heterozygous for Trp63 (5). The microarray analysis identified several additional candidate genes, fibroblast growth factor (FGF)7 and FGF10 being 2 of those candidates. In many organs, the FGF7/10-FGFR2IIIb axis mediates the signal from mesenchymal to epithelial tissues (9–12). However, both Fgfr2 null (13) and FGFR2IIIb null (14) mice are embryonic and newborn lethal, respectively. Therefore, the function of FGF7/10-FGFR2IIIb signaling in developing FRT is unknown. Because the significance of FGF10-FGFR2 signaling in the maintenance of p63-positive basal cells has been demonstrated in the trachea (15), we explored the function of FGF7/10-FGFR2IIIb in the regulation of ΔNp63 in MDE.

The most common downstream pathway employed by FGFs is the MAPK pathway. Binding of FGF promotes dimerization of FGF receptors (FGFRs) leading to a conformational shift that activates the intracellular kinase domain, resulting in phosphorylation of tyrosine residues in the intracellular domain that function as docking sites for adaptor proteins, including FGFR substrate 2 (FRS2) (16). The phosphorylation of FRS2 by FGFRs allows the recruitment of growth factor receptor-bound 2 (GRB2) adaptor protein and son of sevenless exchange factor, which in turn activates RAS and the downstream RAF-MAPK pathway (16). The RAF is transduced to the activation of MAP2K1/2 or MEK1/2, which activates MAPK3/2 or ERK1/2 by phosphorylation (17). However, the binding of FGFs to FGFRs also leads to activation of alternative pathways such as the phosphoinositide 3-kinase pathway, which occurs when a complex of GRB2/GRB2-associated binding protein 1 is recruited by FRS2 (18). Thus, we examined whether the cell-autonomous MEK-ERK activity in MDE can replace the mesenchymal signals mediated by FGF7/10-FGFR2 using a conditional transgenic mouse line, in which Cre recombinase induces expression of constitutively active mutant of MEK1 (MEK1*) (19).

Our current study provides evidence that FGF7/10-FGFR2IIIb-MAPK signaling in combination with ActA-RUNX1 and BMP4-SMADs pathways is essential and sufficient for the expression of ΔNp63 and the subsequent differentiation of vaginal epithelium in MDE.

Materials and Methods

Mouse models

All animal procedures were approved by the Animal Care and Use Committee in Ohio State University and Northwestern University. The mouse strains carrying the following alleles were used: Fgfr2^flox (Fgfr2^tm1Dor/J) (20), ROSA^mT-mE (Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,−EGFP)Luo/J}) (21), ROSA^MEK1* (Gt (ROSA)26Sor^{tm8(Map2k1*,EGFP)Rsky}/J) (19), Smad2^flox (Smad2 ^tm1.1Epb/J) (22), Smad3^flox (23), Smad4^flox (Smad4^tm2.1Cxd/J) (24), Wnt7a-Cre (25), and NSG (NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/Szj). C57BL/6J mice were purchased from The Jackson Laboratory, and CD-1 mice were purchased from Charles River Laboratories. Lines originally in the mixed genetic background were crossed to C57BL/6J at least 3 times before experiments. Wnt7a-Cre strain was maintained by breeding with C57BL/6J mice. Conditional knockout (cKO) lines were crossed with the ROSA^mT-mE line to verify the efficiency and specificity of Cre expression. The day of birth was count as postnatal day (PD)1.

Isolation of uterine and vaginal epithelium and mesenchyme

The procedure has been described previously (26). Briefly, pieces of vaginae and uteri from PD2 C57BL/6J mice were incubated with 1% trypsin (1:250; Life Technologies) for 45 minutes on ice, and a subsequent incubation with PBS containing 5mM EDTA for 15 minutes on ice. The residual trypsin was neutralized with DMEM (Life Technologies) containing 10% fetal bovine serum, then mesenchyme and epithelium were mechanically separated using a fine surgical knife and forceps in cold DMEM supplemented with 10-μg/mL Deoxyribonuclease I (DNase I) (Sigma-Aldrich). The separated epithelial and mesenchymal tissues were immediately frozen in liquid nitrogen and stored at −80°C until they were used for quantitative real-time PCR (qRT-PCR).

Quantitative real-time PCR

Total RNA was extracted from tissue samples using TRIzol (Life Technologies). For epithelium and mesenchyme isolated from C57BL/6J uteri and vaginae, cDNA was synthesized using qScript cDNA Synthesis kit (Quanta BioSciences, Inc), and qRT-PCR was performed on a QuantStudio 12K Flex Real-Time PCR System using TaqMan Universal Master Mix II (Life Technologies). For the vaginal tissue of Fgfr2 cKO and conditional heterozygous (cHet) mice, qRT-PCR was performed on a StepOnePlus Real-Time PCR System (Life Technologies) using KAPA SYBR FAST qPCR kits (Kapa Biosystems, Inc). qRT-PCR assays were performed at least twice with triplicates. The relative expression values of target transcripts were calculated by normalizing the threshold cycle (CT) value to that of Actb or Gapdh. Statistical analysis was performed by one-way ANOVA. For the quantitation of FGFs and FGFRs transcripts, TaqMan Probes were purchased from Life Technologies: Fgf1, Mm00438906_m1; Fgf7, Mm00433291_m1; Fgf9, Mm00442795_m1; Fgf10, Mm00433275_m1; Fgf13, Mm00438910_m1; Fgfr21, Mm00840165_g1; Fgfr22, Mm00445749_m1; Fgfr1, Mm00438930_m1; Fgfr2IIIb, Mm01275521_m1; Fgfr2IIIc, Mm01269938_m1; Fgfr3, Mm00433294_m1; Fgfr4, Mm01341852_m1; and Actb, Mm00607939_s1. For the detection of transcripts for mesenchymal factors, the following primer sets referred by PrimerBank (http://pga.mgh.harvard.edu/primerbank/) were used: Fgf7 forward, TGGGCACTATATCTCTAGCTTGC and Fgf7 reverse, GGGTGCGACAGAACAGTCT; Fgf10 forward, TTTGGTGTCTTCGTTCCCTGT and Fgf10 reverse, TAGCTCCGCACATGCCTTC; Inhba forward TCCGAAGGATGGACCTAACTC and Inhba reverse, GCTTTCTGATCGCGTTGAGAAG; Bmp4 forward, TTCCTGGTAACCGAATGCTGA and Bmp4 reverse, CCTGAATCTCGGCGACTTTTT; and Gapdh (internal control) forward, TCCATGACAACTTTGGCATTG and Gapdh reverse, CAGTCTTCTGGGTGGCAGTGA.

Immunofluorescence (IF)

IF assays were performed as previously described (5, 27). Briefly, tissues collected for immunostaining were fixed with Modified Davidson's fixative solution (Electron Microscopy Sciences), processed into paraffin, and sectioned at 5 μm. The sections were heated in 10mM sodium citrate buffer (pH 6.0) containing 0.05% Tween 20 for 35 minutes in an Electric Pressure Cooker. The following primary antibodies were used at the indicated dilution: anti-p63 (clone 4A4) (sc-8421, 1:200), anticytokeratin (CK) 14 (K14) (LL001) (sc-53253, 1:50), and antipan CK (sc-81714, 1:100) from Santa Cruz Biotechnology, Inc; anti-ΔNp63 (PC373, 1:2000) from Millipore; antiprogesterone (P4) receptor (PR) (A0098, 1:200) from Agilent Technologies; anti-RUNX1 (2593–1, 1:400) from Epitomics; antiphospho (p)-Ser¹⁰ histone H3 (pH3) (CST3377, 1:2000), anti-pERK (CST4370, 1:30), anti-MEK (CST9122, 1:100), anti-pSMAD1/5/9 (CST9511, 1:100), and anti-pSMAD3 (CST9520, 1:50) from Cell Signaling Technology; anti-SMAD2 (700048, 1:100) from Life Technologies; and anti-Forkhead Box A2 (FOXA2) (ab40874, 1:2000) from Abcam. Alexa Fluor 594 antimouse IgG and Alexa Fluor 488 antirabbit IgG from Jackson ImmunoResearch were used for the secondary antibodies, and bisbenzimide H 33258 (Hoechst 33258, 1:10 000; Sigma-Aldrich) was used for nuclear staining. Micrographs were captured using a BZ-9000 microscope (Keyence).

Subrenal capsule grafting of Fgfr2 cKO and cHet uteri

The procedure for subrenal grafting of mouse uteri has been described previously (6). Briefly, uteri from PD1 Fgfr2 cKO and cHet mice were cut into pieces (2–3 pieces per uterine horn) and transplanted onto kidneys of 8-week female NSG mice. Two weeks after grafting, host mice were ovariectomized. After 10–14 days of resting time, ovariectomized hosts were randomly assigned into 3 groups and were given an ip injection of corn oil (Sigma-Aldrich), 100-ng 17β-estradiol (E2) (Sigma-Aldrich), or 100-ng E2 + 250-μg P4 (Sigma-Aldrich) (28). Twenty-four hours after hormonal treatment, uterine grafts were harvested for histological analyses. Five host mice were assigned each for E2 and E2+P4 groups, and 3 host mice were assigned for oil group. The experiment was repeated 5 times (2 experiments did not include oil group). For the analysis of epithelial proliferation rate (pH3 labeling index), the images of IF stained sections for pH3 were captured, and the pH3 positive and negative epithelial cells were blindly counted (≥100 cells per sample) in more than or equal to 5 samples for each using the ImageJ software (NIH). The data were subjected to Student's t test for comparison between Fgfr2 cKO and cHet uterus and one-way ANOVA for among different treatments.

Uterine organ culture

Uterine hanging drop organ culture was performed as previously described with minor modifications (5). Briefly, uteri were dissected from PD1 mice, cleaned by removing connective tissues, and cut into 3 pieces per uterine horn in DMEM/F12 (11039; Life Technologies). The uterine pieces were then placed in autoclaved PCR tube caps (AXYGEN) with basal medium (DMEM/F12 with insulin-transferrin-selenium and antibiotic antimycotic; Life Technologies) with/without 20-ng/mL human recombinant BMP4, ActA, and/or FGF10 (Life Technologies), inverted, and incubated. In all experiments, 10nM ICI 182780 (ICI) (Sigma-Aldrich) was added to the medium unless it was specified otherwise. Uterine pieces were cultured up to 3 days with daily medium change, fixed with Modified Davidson's fixative, and processed for histological analysis. There was no difference in the results between CD-1 and C57BL/6J strains. Therefore, we used CD-1 mice as the standard donor of uterine pieces for their larger litter size, unless the experiment required uterine pieces from transgenic or cKO mice.

Activin inhibitor treatment of neonatal mice

SB-505124 (Cayman Chemical), a kinase inhibitor of type I receptor for TGFβ and activin (Activin Receptor-like Kinase (ALK), ALK5, and ALK7), was dissolved in dimethyl sulfoxide (Sigma-Aldrich) at 12 mg/mL. The inhibitor solution or dimethyl sulfoxide was diluted with corn oil to 10% vol/vol immediately before injection. For this experiment, the CD-1 strain was used for their large litter size and higher success rate of nursing pups after treatment (no rejection observed). Newborn CD-1 mice were given 2 daily sc injections of 12-mg/kg SB-505124 in 10-μL/g body weight corn oil from PD1 to PD5. The first injection was performed within 12 hours after the birth, and the vaginal tissues were collected within 6 hours after the last injection at PD5 for histological analysis. Litters (n = 5) were randomly selected from 4 different mothers, and pups in the same treatment group were assigned to the same foster mother.

Results

Expression patterns of FGF signaling molecules in neonatal FRTs

To identify molecules that control epithelial cell fate decision in developing vagina, we have conducted microarray analysis on vagina and uterus from PD2 Trp63 cHet and cKO mice. In the analysis, FGF7 and FGF10 were enriched in developing vagina rather than uterus, irrespective of Trp63 expression (GSE44697) (5). Accordingly, we performed qRT-PCR analyses on epithelium and mesenchyme isolated from PD2 uterus and vagina to confirm the expression patterns of FGFs and FGFRs that were detected in the microarray analysis (Figure 1A). The qRT-PCR assay confirmed that the FGF7 and FGF10 were highly expressed in the vaginal mesenchyme and undetectable in the uterus (Figure 1A), whereas FGFR2IIIb, the receptor for FGF7/10, was expressed in the epithelium of both neonatal uterus and vagina. The expression patterns warranted an investigation on the function of FGF7/10 in vaginal cell fate decision. In contrast to FGF7/10-FGFR2IIIb, FGFR2IIIc was specifically expressed in uterine and vaginal mesenchyme, whereas FGF9, a specific ligand of FGFR2IIIc, was enriched in MDE. Thus, FGF signaling pathways appear to regulate the development of the lower FRT through reciprocal tissue interactions between epithelium and mesenchyme: FGF7/10-FGFR2IIIb signaling mediates signals from mesenchymal to epithelium, and FGF9-FGFRIIIc mediates signals from epithelium to mesenchyme. In addition, FGFR3 and its ligands, FGF1 and FGF9, were enriched in vaginal epithelium, suggesting the presence of autocrine signaling within vaginal epithelium. Fgf13, Fgf21, and Fgf22 were detected at minimum levels in microarray, and the qRT-PCR confirmed that these FGFs were almost undetectable in PD2 uterus and vagina (data not shown).

Figure 1.

Expression patterns of FGF signaling molecules in neonatal mouse FRT. A, Relative mRNA levels of FGF family ligands and receptors in the epithelium (E) and mesenchyme (M) of uterus (Ut) and vagina (Vg) from PD2 C57BL/6J mice. Bars indicate average mean ± SD (n = 3, each contains ≥5 pooled samples). Statistical analysis was performed by one-way ANOVA (a vs b, P < .01; b vs c, P < .05). B, Expression of RUNX1, pSMAD1/5/9, and pERK1/2 in the lower part of MD (dashed line) at E15.5. WD, Wolffian duct. RUNX1 (green) was costained with pan CK (red). pSMAD1/5/9 (green) and pERK1/2 (green) were costained with p63 (red). MDE was undifferentiated at E15.5, because it was mostly negative for CK (53) and p63, whereas WD epithelium was positive for CK, and UGS epithelium was uniformly stained for both CK and p63 (54). MDE showed uniform nuclear staining of RUNX1. pSMAD1/5/9 and pERK1/2 activities in MDE were higher at the junction to UGS, demonstrating a gradient towards the cranial portion. Scale bar, 100 μm.

To assess whether FGFR-MAPK signaling was active in the MDE undergoing vaginal epithelial differentiation, we examined the phosphorylation/activation of ERK1/2 in embryonic FRTs. The expression of ΔNp63 in MDE is initiated at the junction to the urogenital sinus (UGS) around embryonic day (E)16 and progresses anteriorly towards the cervix (6, 29). At E15.5, MDE was still negative for ΔNp63 (Figure 1B). However, the MDE at the junction to UGS was strongly positive for RUNX1 and pSMAD1/5/9 (Figure 1B). Interestingly, pERK1/2 positive cells were also concentrated in the caudal portion of MDE, particularly at the junction to UGS (Figure 1B), suggesting that the FGF-MAPK pathway is involved in the induction of ΔNp63.

FGFR2 is essential for vaginal epithelial cell fate decision

In order to investigate the requirement of FGF7/10-FGFR2IIIb signaling in the differentiation of vaginal epithelium, we generated MDE-specific Fgfr2 cKO (Fgfr2 cKO, Fgfr2^flox/flox; Wnt7a-Cre⁺) mice (20, 25). cHet (Fgfr2 cHet, Fgfr2^flox/wt; Wnt7a-Cre⁺) littermates were used as controls. Because Wnt7a-Cre is also expressed in embryonic ectoderm, in which FGFR2 plays a critical role in limb development (14), the limbs of Fgfr2 cKO mice were truncated or absent (Supplemental Figure 1). Most Fgfr2 cKO mice died within the first 5 days of postnatal development. Therefore, the analysis on vaginal phenotypes was performed by PD4. The vaginal epithelium of Fgfr2 cKO mice was completely negative for ΔNp63 (Figure 2A). The phenotype was identical to that of Trp63 null vaginal epithelium: negative for K14, a differentiation marker for squamous epithelium of vagina (30), but positive for PR, which is indicative of uterine differentiation (Figure 2A) (26). These observations establish the essential role of FGFR2 in the induction of vaginal epithelial cell fate in MDE.

Figure 2.

FGFR2 is essential for the commitment of MDE to vaginal epithelial cell fate. A, IF analysis of epithelial marker expression in the vaginae (Vgs) of PD4 Fgfr2 cHet and cKO mice (n ≥ 4). Vaginal epithelium of Fgfr2 cKO mice lacked the expression of vaginal differentiation markers ΔNp63 (green) and K14 (red), however, expressed uterine differentiation marker PR (green). The phenotype was identical to that of ΔNp63 null mice (5). Scale bar, 100 μm. B, Conditional deletion of Fgfr2 did not alter RUNX1 expression in vaginal epithelium. RUNX1 (green) was comparably expressed in the cervical/vaginal epithelium in Fgfr2 cHet and cKO at PD2. At PD4, RUNX1 expression decreased as the epithelial cells differentiate in Fgfr2 cHet Vg, whereas it was maintained strong in Fgfr2 cKO Vg. Scale bar, 100 μm. C, pSMAD1/5/9 IF analysis in PD2 Vg of Fgfr2 cHet and cKO mice. Nuclear pSMAD1/5/9 (green) was detected in the epithelium of both Fgfr2 cHet and cKO mice. Scale bar, 20 μm. D, Relative mRNA expression analysis for vaginal mesenchymal factors in the PD1 Vgs of Fgfr2 cHet (n = 7) and cKO (n = 5) mice, and the PD1 uterus (Ut) (n = 4) from wild-type (WT) mice. Bars indicate average mean ± SE. Statistical analysis was performed by one-way ANOVA (a vs b, P < .05; a vs c, P < .01; b vs c, P < .01; b vs d, P < .01). Fgf7 was significantly higher in the vaginal tissue of Fgfr2 cKO than cHet mice.

RUNX1 was equally expressed in vaginal/cervical epithelia of Fgfr2 cKO and cHet mice (Figure 2B). Thus, FGFR2 is required for the expression of ΔNp63 independently of RUNX1 expression, and RUNX1 is not downstream of FGFR2. We had previously determined the expression of RUNX1 diminishes when MDE differentiates into vaginal epithelium (5). Accordingly, RUNX1 expression was down-regulated in the vaginal epithelium of Fgfr2 cHet mice but not in Fgfr2 cKO mice at PD4 (Figure 2B).

We next tested whether the loss of FGFR2IIIb in MDE inhibits expression of ΔNp63 through inhibition of BMP signaling. The activation/phosphorylation of BMP-regulated SMADs (pSMAD1/5/9) was equally detected in the nucleus of MDE in both PD2 Fgfr2 cKO and cHet vaginae (Figure 2C), suggesting that the inactivation of FGFR2 signaling in MDE does not alter BMP signaling activity.

This was further confirmed by comparing the mRNA expression levels of BMP4 and ActA as well as FGF7 and FGF10 in the vaginal tissues from PD1 Fgfr2 cKO and cHet mice (Figure 2D). There was no evidence that the loss of epithelial FGFR2IIIb interferes with the differentiation of vaginal epithelium through down-regulation of BMP4 or ActA. The levels of Inhba, encoding the subunit of ActA, were comparable between Fgfr2 cKO and cHet vaginae. The levels of Fgf10 and Bmp4 appeared up-regulated in Fgfr2 cKO than cHet vaginae; however, the difference did not reach statistical significance. Among 4 genes tested, only the level of Fgf7 showed a significant difference between Fgfr2 cKO and Fgfr2 cHet vaginae, and the level was up-regulated in Fgfr2 cKO vaginae. This implied increase of Fgf7 in the mesenchyme may indicate a presence of feedback regulation on FGF signals in the reciprocal communication between vaginal epithelial and mesenchymal tissues.

Uterine phenotype was not altered by loss of FGFR2

Although vaginal epithelial differentiation was inhibited, loss of FGFR2 in MDE did not appear to affect neonatal uterine epithelial differentiation. However, Filant et al reported that conditional deletion of Fgfr2 by Pgr-Cre (Pgr^Cre/+) induced aberrant p63 expression in neonatal and adult uterine epithelium (31). In addition, Li et al demonstrated that stromal cells control epithelial proliferation through production of FGFs in the mouse uterus (32). Accordingly, we addressed the function of FGFR2 in mature uterus by growing the uteri from neonatal Fgfr2 cKO and cHet mice as subrenal grafts (28). Three weeks after grafting, uterine pieces from PD1 Fgfr2 cKO and cHet mice developed normal uterine histology with uterine glands (Figure 3A, FOXA2), and vaginal epithelial markers, ΔNp63 and K14, were never detected in epithelial-specific Fgfr2 cKO uteri (n = 12). Because epithelial differentiation was normal in Fgfr2 cKO uteri, we next examined the cellular response to steroid hormones. IF assay indicated that the expression of PR was properly regulated by E2 and P4 in both Fgfr2 cKO and cHet uteri (Figure 3B): PR was predominantly expressed in the epithelium in the ovariectomized hosts, E2 up- and down-regulated PR in stromal and epithelial cells, respectively (26), and the down-regulation of PR in the epithelium was inhibited by P4 (33). The proliferation of uterine epithelial and stromal cells is also regulated by the balance between E2 and P4 (28, 34, 35), and the proliferation response of Fgfr2 cKO and cHet uteri was normal as assessed by pH3 (Figure 3, C and D): E2 increased the pH3-positive cells in the epithelium, and cotreatment with P4 blocked the induction of pH3 in the epithelium but induced pH3 in the stroma. In summary, the loss of FGFR2 does not affect the response of uterine epithelium to E2 and P4 in PR regulation and proliferation.

Figure 3.

Uterine phenotypes of Fgfr2 cKO mice. A, Fgfr2 cHet and cKO uterine grafts developed uterine glands expressing FOXA2 (green). Epithelium was highlighted by pan CK (red). Scale bar, 100 μm. B, Regulation of PR (green) in Fgfr2 cHet and cKO uterine grafts treated with oil (vehicle), E2, or E2 plus P4 (E2P4). PR was identically regulated in Fgfr2 cHet and cKO uterine grafts by E2 and P4. Dot lines indicate the border between the epithelium (Ep) and stroma (St). Scale bar, 100 μm. C and D, Regulation of cell proliferation in the Fgfr2 cHet and cKO uterine grafts. C, Representative staining for pH3 (green) with CK (red). Scale bar,100 μm. D, pH3 labeling index (pH3 LI, %) was calculated as the numbers of pH3-positive cells per total cell numbers (≥100 cells per sample) in the epithelium. Cell proliferation response of Fgfr2 cHet and cKO was comparable for each treatment, and the difference was not significant (ns) (P > .05) by Student's t test. pH3 LI was significantly higher in E2-treated group for both Fgfr2 cHet and cKO uterine epithelium compared with oil and E2P4-treated groups (one-way ANOVA). Bars indicate average mean ± SD (n ≥ 5).

MEK-ERK activation rescues the phenotype of Fgfr2 cKO vagina

In many developing organs, the RAF-MAPK pathway plays critical roles in the transduction of FGF-FGFR signaling. Hence, the activity of ERK was assessed by IF assay of pERK1/2 in the cervical/vaginal epithelium in PD2 Fgfr2 cKO and cHet mice (Figure 4A). The loss of FGFR2 did not completely abolish the ERK activities, and the signal for pERK1/2 was sporadically detected in the epithelium of Fgfr2 cKO cervix and vagina (Figure 4A). Nevertheless, the signal for pERK1/2 in PD2 vaginal epithelium was significantly reduced in Fgfr2 cKO compared with Fgfr2 cHet mice (Figure 4A). Because the FGFR2 signal can be transduced through different pathways, we tested whether activation of MEK1-ERK1/2 could replace FGFR2 using a Cre-inducible transgenic mouse line for a MEK1* (19). When MEK1* was induced in the MDE by Wnt7a-Cre, Fgfr2 null vaginal epithelium differentiated normally with expression of ΔNp63 and K14 (Figure 4B), and PR expression that was observed in Fgfr2 cKO mice (Figure 2A) was absent at PD5 (n = 6). In contrast, uterine epithelium of MEK1*-positive Fgfr2 cKO and cHet mice was normal at PD5 (data not shown). Absence of ΔNp63 expression in uterine epithelium particularly confirmed that MEK1* expression itself was insufficient to transform MDE into vaginal epithelium.

Figure 4.

MEK1 activity rescues the loss of FGFR2 in vaginal epithelial differentiation. A, IF analysis of pERK1/2 in the vagina of Fgfr2 cHet and cKO mice at PD2. pERK1/2 (green) was costained with K14 (red). The loss of FGFR2 reduced ERK activity in the cervical/vaginal epithelium. However, the signal for pERK1/2 was still detectable in the epithelium of Fgfr2 cKO cervix and vagina. B, Vaginal phenotype of Fgfr2 cHet and cKO mice with MEK1*. MEK1* expression restored the vaginal epithelial differentiation in Fgfr2 cKO mice with the expression of p63 (red, left), RUNX1 (green, left), and K14 (red, right) at PD5. Accordingly, PR (green, right) was absent in the epithelium. C, IF analysis of MEK and pERK1/2 in the vaginae of PD1 nontransgenic/wild-type (WT) and MEK1* transgenic mice. The expression of MEK1* in the MDE up-regulated pERK1/2 (green) in the cervical/vaginal epithelium, yet did not alter p63 (red) expression pattern significantly. Scale bar, 100 μm.

To further confirm that MEK1* expression itself does not alter ΔNp63 expression pattern in MDE, development of vagina and uterus was assessed from PD1 to PD6 in MEK1* transgenic and control mice in normal C57BL/6J background. The expression pattern of ΔNp63 in vaginal epithelium was comparable between MEK1* transgenic and nontransgenic control mice at PD1 even though the pERK1/2 was up-regulated in MEK1*-positive MDE (Figure 4C). Likewise the expression of MEK1* did not induce expression of RUNX1, ΔNp63, or K14 in uterine epithelium (Supplemental Figure 2A). These observations establish that FGFR2-MAPK signaling is essential but not sufficient to induce vaginal epithelial cell fate in MDE. Although the expression of MEK1* did not affect the differentiation cell fate of uterine epithelium, MEK1* transgenic mice developed endometrial hyperplasia by the age of 3 months with 100% penetrance (n = 4) (Supplemental Figure 2B), entailing the critical role of MEK-ERK activity being downstream of FGFR2 activating mutations in endometrial carcinogenesis (36).

To demonstrate the specificity of MEK-ERK activity being downstream of FGFR2, we also tested whether vaginal phenotypes of Smad4 cKO (Smad4^flox/flox; Wnt7a-Cre⁺) mice (5) could be rescued by the expression of MEK1*. Although the number of sporadic basal cells appeared increased, the expression of MEK1* did not restore ΔNp63 in the vaginal epithelium of Smad4 cKO mice (Supplemental Figure 3). Thus, FGF7/10-FGFR2IIIb-ERK1/2 signaling is indispensable for the induction of ΔNp63 in MDE in addition to RUNX1 and SMAD4.

ActA, BMP4, and FGF10 are essential and sufficient to induce ΔNp63

Previously, we demonstrated that BMP4 alone was sufficient to induce ΔNp63 in the epithelium of newborn uterus, and ActA synergistically enhanced the BMP4 action in an organ culture system (5). However, this result was inconsistent with our current in vivo study that indicates the essential role of FGFR2 signaling in ΔNp63 expression. We suspected a presence of factors in the organ culture system that activated MAPK in the uterus. Detailed examination of organ cultured uteri detected PR expression in mesenchyme but not epithelium (Figure 5A), indicating estrogen action. Furthermore, RUNX1, which can be induced in uterine epithelium by estrogens (5), was also expressed in the epithelium of organ-cultured uteri (n = 12) (Figure 5A). Accordingly, we refined the culture conditions to minimize estrogenic activity coming from the plastics used in hanging drop culture. In addition, the culture medium was supplemented with ICI, an estrogen receptor (ER) antagonist, at 10nM. In organ culture with ICI, PD1 uteri (n = 12) retained the original uterine expression patterns of PR and RUNX1 (Figure 5A). Moreover, ICI down-regulated pERK1/2 in the epithelium of cultured uteri (Figure 5A), suggesting that extraneous estrogenic activity in the medium had replaced FGF7/10 by activating the ERK MAPK pathway in the previous organ culture study. Because ERα is expressed only in the mesenchyme in neonatal uteri (37, 38), we hypothesize that estrogenic activity in the medium induced growth factors in the uterine mesenchyme, and the growth factors induced RUNX1 and ERK MAPK activity in the epithelium.

Figure 5.

Uterine organ culture study. ActA, BMP4, and FGF10 are essential and sufficient to induce ΔNp63 in cultured newborn uterus. A, Estrogenic activity in the uterine organ culture system induced RUNX1 and pERK1/2. Presence of estrogen activity was indicated by the expression pattern of PR; PR was expressed in mesenchymal (Ms) but not epithelial (Ep) cells. Expression of RUNX1 in uterine epithelium also indicates the estrogen activity (5). The expression pattern of PR and RUNX1 became that of normal neonatal uteri when 10nM ICI was added to the media. ICI also repressed the activity of ERK in uterine epithelium. B, Three-day incubation with 20-ng/mL ActA, but not BMP4 or FGF10, up-regulated RUNX1 (green) in the neonatal uterine epithelium in the presence of 10nM ICI. p63 (red) expression was induced only when all 3 factors were added to the medium. C, ActA induced RUNX1 in uterine epithelium in a dose-dependent manner. Neonatal uterus was incubated with different concentrations of ActA (μg/mL) overnight. D, ActA and BMP4 were sufficient to induce p63 (red) in MEK1* transgenic uterine epithelium in 3-day culture. Note, although MEK1* transgenic uteri never expressed ΔNp63 in vivo, a sporadic expression of p63 (arrow) was occasionally detected in the epithelium after 3 days of culture. E, The loss of FGFR2 decreases the sensitivity of uterine epithelium to ActA in RUNX1 expression. Uterine pieces from PD1 Fgfr2 cKO and cHet mice were cultured with different concentrations (0, 0.02, or 0.05 μg/mL) of ActA overnight, and the expression of RUNX1 (green) was analyzed. Although ActA at 0.02 μg/mL effectively induced RUNX1 in Fgfr2 cHet uteri, Fgfr2 cKO uteri required 0.05-μg/mL ActA. However, the difference disappeared after 3 days in vitro as Fgfr2 cKO uteri also expressed RUNX1 with 0.02-μg/mL ActA (data not shown). Scale bar, 100 μm.

In refined culture conditions, RUNX1 was induced by ActA in a dose-dependent manner (Table 1 and Figure 5, B and C). However, ΔNp63 was induced only when all 3 factors, FGF10, BMP4, and ActA, were supplemented in the medium at the final concentration of 20 ng/mL each (Table 1 and Figure 5B). In the same culture conditions, MEK1* could replace the supplementation of FGF10 in the medium, thus BMP4 and ActA were sufficient to induce ΔNp63 in MEK1* transgenic uterine epithelium (n = 5) (Figure 5D and Table 2). In contrast, FGF10, BMP4 plus ActA did not induce ΔNp63 in Fgfr2 cKO uteri (n = 3) (data not shown). These observations indicate that 3 signaling pathways, FGF7/10-FGFR2IIIb-ERK, ActA-RUNX1, and BMP4-SMADs, are required for the differentiation of MDE into vaginal epithelium. Notably, although these 3 pathways are independently required for activation of ΔNp63, Fgfr2 cKO uteri required a higher concentration of ActA to express RUNX1 compared with Fgfr2 cHet uteri (Figure 5E), indicating the presence of crosstalk between these pathways.

Table 1.

Combination Effect of Activin A, BMP4 and FGF10 in Uterine Organ Culture (20 ng/ml each)

Combinations
Factors
Activin A	−	+	−	−	+	+	−	+
BMP4	−	−	+	−	+	−	+	+
FGF10	−	−	−	+	−	+	+	+
Uterine pieces
N	8	7	6	7	6	8	7	7
ΔNp63	8 (−)	7 (−)	6 (−)	7 (−)	6 (−)	8 (−)	7 (−)	3 (±) 3 (+)1 (++)
RUNX1	8 (±)	1 (++) 6 (+++)	5 (±) 1 (+)	6 (±) 1 (+)	2 (++) 4 (+++)	2 (++) 6 (+++)	6 (±) 1 (+)	2 (++) 5 (+++)

−, negative; ±, Single positive cells were detected at least in one histology section; +, <25% of epithelium was consistently positive; ++, ≥25% and <75% of epithelium was positive; +++, ≥75% of epithelium was positive.

Table 2.

MEK1* Transgene Replaces FGF10 in Induction of ΔNp63 in Uterine Organ Culture

Combinations
Factors
Activin A	−	+
BMP4	−	+
FGF10	−	−
Uterine pieces
N	3	5
ΔNp63	3 (±)	5 (++)
RUNX1	3 (±)	5 (+++)

Activin A and BMP4 were added at 20 ng/ml each.

ActA induces RUNX1 through SMAD-independent signaling

ActA is a member of TGFβ superfamily, and its canonical downstream signaling is through SMAD-dependent pathways. However, RUNX1 expression was normal in the vagina of Smad4 cKO mice (5) (Supplemental Figure 3), suggesting that ActA action is independent of SMADs. Alternatively, SMAD2 and SMAD3 may induce RUNX1 in MDE independently of SMAD4 as previously demonstrated in the induction of MAD1 in epidermis (39). Hence, we generated cKO mice for both Smad2 and Smad3 to examine the requirement of the downstream receptor-regulated R-SMADs of ActA in the differentiation of vaginal epithelium. Conditional deletion of Smad2 and/or Smad3 by Wnt7a-Cre did not cause gross abnormalities in newborn mice. As assessed at PD5, the expression of RUNX1, ΔNp63, and subsequent epithelial stratification in vagina was not affected by the deletion of Smad2 and/or Smad3 (Figure 6A). It was not due to incomplete deletion of Smad2 and/or Smad3, because the ablation of SMAD2 and SMAD3 in the vaginal epithelium was confirmed by IF (Figure 6A).

Figure 6.

SMAD2/3 are dispensable for ActA action in ΔNp63 induction on MDE. A, IF analysis for RUNX1 (green), p63 (red), SMAD2 (green), and pSMAD3 (green) in PD5 vagina of Smad2/3 cHet and/or cKO mice. SMAD2/3 is dispensable for the differentiation of vaginal epithelium. RUNX1 and p63 were normally expressed in the vaginal epithelium of Smad2/3 cKO mice. B, Activin receptor kinase inhibitor (SB-505124) down-regulated RUNX1 (green) in the vaginal fornix. Accordingly, MDE in the fornix failed to express p63 (red) and K14 (red) and demonstrated uterine epithelial differentiation with PR (green) expression. Dashed lines indicate the border between the epithelium (Ep) and mesenchyme (Ms). Scale bar, 100 μm.

In order to test the requirement of ActA in vivo, newborn mice were treated with SB-505124, a selective inhibitor of activin/TGFβ type I receptors (ALK4/ACVR1B, ALK5/TGFBR1, and ALK7/ACVR1C), from birth to PD5 (n = 5). SB-505124-treatment down-regulated RUNX1 (Figure 6B). The neonatally SB-505124-treated mice lacked ΔNp63 expression in a substantial portion of vaginal epithelium in the fornix (Figure 6B). Accordingly, epithelial cells in the vaginal fornix demonstrated uterine-like differentiation, negative for K14 and positive for PR (Figure 6B), similar to vagina of Runx1 cKO mice (5).

In conclusion, vaginal epithelial cell fate in MDE is induced by 3 mesenchymal signals, FGF7/10, ActA, and BMP4, each of which is transduced by MAPK, RUNX1, and SMAD pathways, respectively (Figure 7).

Figure 7.

Models. Epithelial cell fate decision in the lower FRT. Vaginal mesenchymal cells instruct vaginal epithelial cell fate in MDE by secreting BMP4, ActA, and FGF7/10. In MDE, BMP4 signal is transduced by canonical SMAD-dependent pathway, whereas ActA signal activates RUNX1 expression through a SMAD-independent pathway. FGF7/10 signal is transduced by the FGFR2IIIb-MAPK pathway in MDE. Three signaling pathways in concert activate expression of ΔNp63, and afterwards, the transcriptional activity of ΔNp63 locus is maintained by ΔNp63 itself (5). The disruption of one of these pathways interferes with ΔNp63 expression and subsequently cervical/vaginal epithelial differentiation, resulting in cervical and vaginal adenosis.

Discussion

The mechanism of vaginal development, particularly the epithelial origin has been long debated. In the mouse, a cell lineage tracing experiment indisputably established that the vaginal epithelium arises solely from MDE (2). Thus, the distinctive epithelia of uterus and vagina share their cellular origin (40, 41). Although it has long been known that epithelial differentiation in MD-derived organs is under the control of underlying mesenchymal cells (3), experimental evidence for the molecular mechanisms underlying vaginal epithelial differentiation was totally unknown until genetically engineered mice became available. Through a series of studies with genetically engineered mice, our group previously demonstrated that ΔNp63 is the master regulator of vaginal epithelial differentiation in MDE, and the expression of ΔNp63 was induced by mesenchymal paracrine factors, BMP4 and ActA, through SMAD4/RUNX1-dependent mechanisms (5). In this study, we demonstrated that in addition to BMP4 and ActA, the FGF7/10-FGFR2IIIb-MAPK pathway is essential for activation of ΔNp63 locus in MDE. We also determined for the first time that ActA action on MDE is mediated by RUNX1, and SMAD2, SMAD3, and SMAD4 are dispensable. SMAD4 is essential for the activation of ΔNp63 locus in MDE, and SMAD4 binds the up-stream of ΔNp63 TSS in vaginal but not uterine epithelial precursors (5). Because activin-regulated SMADs (SMAD2/3) are dispensable for the expression of RUNX1 and ΔNp63, it is reasonable to assume that BMP4 induces the formation of SMAD transcriptional complexes and their binding to the SMAD-binding element on the 5′ of ΔNp63 TSS (Figure 7). Nevertheless, further studies are required to fully establish whether SMAD-mediated BMP4 action is essential for the induction of vaginal epithelial cell fate in lower MDE.

The mechanism of vaginal epithelial differentiation has drawn the interest of medical researchers due to its relevance to the etiology of vaginal adenosis, which is nearly always observed adjacent to the diethylstilbestrol (DES)-associated vaginal adenocarcinoma (42). DES induces vaginal adenosis through inhibition of ΔNp63 expression via ERα in MDE (6, 8). Our mouse studies have demonstrated that vaginal adenosis occurs when any one of 3 key signaling pathways is disrupted in the developing vagina. Previously, we observed the down-regulation of RUNX1 in the precursor of vaginal epithelium by neonatal DES exposure (5). Thus, we proposed that DES-ERα induces vaginal adenosis by inhibiting ΔNp63 expression through the repression of RUNX1 in MDE. It should be noted that the DES action is mediated solely by epithelial ERα (6). Although epithelial ERα is inhibitory on ΔNp63 expression, uterine organ culture study in our current report indicates that estrogen action via mesenchymal ERα promotes the expression of ΔNp63 in the newborn mouse uterus through activation of RUNX1 and MAPK pathways. Therefore, it is unlikely that DES represses the expression of RUNX1 through the down-regulation of ActA in the vaginal mesenchyme. Our previous observation of pSMAD2/3 expression patterns in the vagina and uterus of DES-treated mice also support this model: pSMAD2/3 was expressed in the epithelium of PD2 vagina, and DES did not down-regulate the expression (5). Because ERα is expressed in the mouse vaginal but not uterine epithelium for the first approximately 5 days of postnatal development, the opposite functions of epithelial vs mesenchymal ERα explain why neonatal DES exposure has opposite effect on ΔNp63 expression in vaginal and uterine epithelia: DES promote expression of ΔNp63 in uterine epithelium (uterine squamous metaplasia) through mesenchymal ERα, whereas epithelial ERα inhibits ΔNp63 expression in vaginal epithelium (vaginal adenosis).

Although MEK1* transgenic uteri never expressed ΔNp63 in vivo, a sporadic expression of p63 was occasionally detected in the epithelium after 3 days in vitro (Figure 5D), indicating that the organ culture system does not perfectly mimic in vivo and that the modified culture condition with 10nM ICI still contains the activity that promotes the induction of ΔNp63 in MDE. This suggests that more factors are involved in the regulation of cell fate decision in lower MDE to become vaginal or uterine epithelium. In this regard, exposure to estrogen is known to down-regulate the expression of developmental regulators, such as HOX genes, in neonatal mouse uteri (43, 44). Thus, in uterine organ culture a residual estrogenic activity may promote the expression of ΔNp63 by down-regulating uterine factors that counteract against inductive signaling for ΔNp63. Accordingly, BMP4, ActA, and FGF7/10 may be insufficient to induce vaginal epithelial cell fate in uterine epithelium in vivo.

The expression patterns of FGFs and FGFRs in neonatal vagina and uterus proposed by Nakajima et al conflict with our conclusions (45), because they claimed that FGFR2IIIc was expressed in MDE, and MDE in the uterus did not express FGFR2IIIb. Previous tissue recombination studies demonstrated that MDE in both neonatal uterus and vagina was equally responsive to induction of vaginal epithelial cell fate by vaginal mesenchyme (3, 30, 37, 46). Thus, our conclusion that FGFR2IIIb is equally expressed in neonatal uterine and vagina epithelia better explains the function of FGF7/10 as epithelial cell fate inducers. The same group also proposed that FGF7/10-FGFR2-MAPK induced K14 expression in p63-positive vaginal epithelium based on the effect of U0126, a MEK1/2 inhibitor, in vaginal organ culture. However, because U0126 inhibits MEK-ERK activities in both vaginal epithelium and mesenchyme, the effect of U0126 on cultured vaginae cannot be attributed to the inhibition of FGFR2 in vaginal epithelium alone. The potential role of FGFR2 in the expression of K14 in ΔNp63-positive vaginal epithelium is a subject of future studies.

We speculate that FGF9-FGFR2IIIc and FGF7/10-FGFRIIIb axes mediate the feedback communication loop between vaginal epithelium and mesenchyme during development. Because FGF9 was down-regulated in vagina from PD2 to PD5 (45), vaginal mesenchyme may monitor the differentiation status of epithelial cells by the level of FGF9 signal, and the expression levels of ΔNp63-inducers (eg, FGF7/10, BMP4, and ActA) may be under the control of FGF9. The elevated FGF7 in the vaginae of Fgfr2 cKO mice may be due to a high level of FGF9 expression in the epithelium.

Previously, Filant et al reported a sporadic expression of p63 in the uterine epithelium of Fgfr2^flox/flox; Pgr^Cre/+ mice, which express Cre in both epithelial and stromal cells (31). It appears to contradict with our finding that epithelial FGFR2IIIb is essential for the expression of ΔNp63 in neonatal vagina as well as uterus. Because the conditional deletion of Fgfr2 by Pgr^Cre/+ abolishes the FGF9-mediated signal from epithelium to stroma by inactivating FGFR2IIIc in stromal cells, the aberrant epithelial-stromal tissue interactions likely played a role in the uterine expression of p63 in Fgfr2^flox/flox; Pgr^Cre/+ mice. It should be noted that Wnt7a-Cre is expressed in embryonic MDE, whereas Pgr^Cre/+ turns on in uterine epithelium around PD5. Thus, it is possible that the concomitant loss of FGFR2 in both uterine epithelium and stroma induces expression of p63 in uterine epithelium through a totally different mechanism from induction of ΔNp63 in MDE. Alternatively, the ectopic expression of p63 in Fgfr2^flox/flox; Pgr^Cre/+ uteri may be due to an incomplete recombination of the floxed allele by Pgr^Cre/+ in the uterine epithelium, which has been reported in a previous study (47). In this theory, the expression of p63 in the uterus of Fgfr2^flox/flox; Pgr^Cre/+ mice occurred in Fgfr2-positive epithelial cells as a consequence of aberrant epithelial-stromal tissue interactions because of the loss of FGFR2 in the stroma. Studies that address the function of epithelial and stromal FGFR2 separately will clarify the mechanism of p63 expression in the uterus of Fgfr2^flox/flox; Pgr^Cre/+ mice.

A series of tissue recombination studies with Esr1 and Pgr null uterus and vagina established the critical roles of stromal cells in the actions of E2 and P4 on the epithelium in FRT (26, 28, 30, 33, 35, 48–51). FGFs are proposed mediators of estrogen action on the uterine epithelium: estrogen induces the expression of FGFs in uterine stroma, and the FGFs in turn stimulate the proliferation of uterine epithelial cells (32). Our study demonstrated that the loss of FGFR2 in the epithelium does not affect the estrogen action on uterine epithelium. This may be explained by redundant functions of FGFRs in uterine epithelium. FGFR1/2 double null mutation may affect the sensitivity of uterine epithelium to estrogen-induced proliferation.

Studies using genetically engineered mice have deepened our understanding in the mechanism underlying normal and abnormal development of FRTs. Our current study provides insight into the pathogenesis of vaginal adenosis and adenocarcinoma by identifying the signaling pathways that regulate the cell fate decision in MDE. New cases of vaginal adenocarcinomas are still reported in women without history of DES exposure (52). These cases are likely caused by a disruption of BMP4-SMAD, ActA-RUNX1, or FGF7/10-FGFR2IIIb-MAPK pathways by chemical exposure during fetal development. Further studies on the effect of environmental chemicals on these signaling pathways will help identify the causal factors of vaginal adenosis and adenocarcinoma.