CTNNB1 in Mesenchyme Regulates Epithelial Cell Differentiation during Müllerian Duct and Postnatal Uterine Development

Abstract

Müllerian duct differentiation and development into the female reproductive tract is essential for fertility, but mechanisms regulating these processes are poorly understood. WNT signaling is critical for proper development of the female reproductive tract as evident by the phenotypes of Wnt4, Wnt5a, Wnt7a, and β-catenin (Ctnnb1) mutant mice. Here we extend these findings by determining the effects of constitutive CTNNB1 activation within the mesenchyme of the developing Müllerian duct and its differentiated derivatives. This was accomplished by crossing Amhr2-Cre knock-in mice with Ctnnb1 exon (ex) 3^f/f mice. Amhr2-Cre^Δ/+; Ctnnb1 ex3^f/+ females did not form an oviduct, had smaller uteri, endometrial gland defects, and were infertile. At the cellular level, stabilization of CTNNB1 in the mesenchyme caused alterations within the epithelium, including less proliferation, delayed uterine gland formation, and induction of an epithelial-mesenchymal transition (EMT) event. This EMT event is observed before birth and is complete within 5 days after birth. Misexpression of estrogen receptor α in the epithelia correlated with the EMT before birth, but not after. These studies indicate that regulated CTNNB1 in mesenchyme is important for epithelial cell differentiation during female reproductive tract development.

The Müllerian duct is the precursor of the female reproductive tract, including the oviduct, uterus, cervix, and anterior vagina. Initially, it forms as a mesoepithelial cord or tube surrounded by mesenchyme that begins to differentiate before birth (1); however, development of the female reproductive tract continues until puberty. This extended period represents a critical time in female reproductive tract development, which, if interrupted or altered, negatively affects fertility.

The mature uterus consists of 3 tissue compartments, the endometrium, myometrium, and serosa. The endometrium is composed of a supporting stroma and both luminal and glandular epithelial cells. The myometrium consists of inner circular and outer longitudinal smooth muscle layers. The serosa is the single outer cell layer that surrounds the outer longitudinal smooth muscle layer. Development of the uterus into these compartments begins soon after birth with the differentiation of the mesenchyme into the stroma and smooth muscle layers. Within the endometrium, uterine gland formation begins postnatally in an ovary- and steroid-independent manner (2, 3). During pregnancy, uterine glands secrete and transport substances that are essential for periimplantation embryo survival, including leukemia inhibitory factor (LIF) (4). Disruption of gland development results in infertility (5–10).

WNT signaling is known to be important for female reproductive tract development. Wnts are expressed by both the Müllerian duct epithelium (Wnt5a, Wnt7a, Wnt11) and mesenchyme (Wnt4, Wnt5a). Wnt4 is required for the initial formation of the Müllerian duct in both sexes (11) and postnally is involved in uterine gland formation, stromal cell proliferation, and decidualization (12). Wnt5a is required for posterior formation of the Müllerian duct and is essential for epithelial-mesenchymal paracrine interactions (13). Wnt7a regulates patterning and morphogenesis of the female reproductive tract. Wnt7a-null mice exhibit a posterior shift in the identity of Müllerian duct-derived tissues. Wnt7a is also required for endometrial gland formation and fertility (5, 14). Wnt11 does not appear to be essential for uterine development (15).

β-Catenin (CTNNB1) is a downstream effector of the canonical WNT-signaling pathway and has roles in female reproductive tract development, function, and cancer formation (9, 16, 17). Conditional deletion of Ctnnb1 within the Müllerian duct mesenchyme using Amhr2-Cre caused a lack of postnatal oviductal coiling without altering Wnt5a and Wnt7a expression (16). Despite changes in the myometrial compartment, endometrial gland development occurred normally, but in reduced numbers (18). Conditional deletion of Ctnnb1 in the postnatal uterus using Pgr-Cre resulted in squamous cell metaplasia, a decrease in uterine glands, and infertility (9). In addition to loss-of-function Ctnnb1 studies, conditional stabilization of CTNNB1 in the female reproductive tract have also been performed. Conditional stabilization of CTNNB1 in the Müllerian duct mesenchyme using Amhr2-Cre induced myometrial hyperplasia and formation of tumors similar to human leiomyomas and endometrial stromal sarcomas (17). In addition, stabilization of CTNNB1 in the postnatal uterus using Pgr-Cre induced endometrial hyperplasia of the uterine glands (9). Furthermore, stabilization of CTNNB1 in female reproductive tract tissues caused subfertility or infertility (9, 17, 19).

CTNNB1 is phosphorylated by glycogen synthase kinase-3β and targeted for degradation by the proteasome. Activation of Frizzled receptors by WNT ligand results in the inhibition of glycogen synthase kinase-3β, allowing nonphosphorylated CTNNB1 to translocate to the nucleus and regulate transcription. Exon 3 of Ctnnb1 contains the coding region for all of the phosphorylation sites; therefore, removal of this exon prevents CTNNB1 from being targeted for degradation, resulting in constitutive stabilization (20). Anti-Müllerian hormone receptor 2 (Amhr2)-Cre knock-in mice exhibit Cre activity within the Müllerian duct mesenchyme at embryonic day (E) 12.5 and postnatal ovarian granulosa cells (21). Studies using Rosa26 lacZ Cre reporter mice and Amhr2-Cre indicate that this Cre is sufficient to conditionally modify gene expression within the endometrial stroma and both layers of the myometrium (18).

The previous studies of CTNNB1 stabilization focused on the consequences in mature animals. In this study we examined the role of CTNNB1 during perinatal Müllerian duct differentiation and postnatal onset of uterine gland formation. CTNNB1 was stabilized within the Müllerian duct mesenchyme using Amhr2-Cre. Mice with the genotype of Amhr2-Cre^Δ/+; Ctnnb1 exon (ex)3^f/+ will be referred to as mutants and Ctnnb1 ex3^f/+as controls. Stabilization of CTNNB1 within the mesenchyme primarily alters epithelial cell gene expression, resulting in profound effects on oviduct and uterine development and differentiation. Furthermore, there was an epithelial to mesenchymal transition (EMT) in the epithelial compartment of the female fetus reproductive tract. However, after birth, epithelial cells begin to express typical epithelial markers in mutant mice, emphasizing the difference in the role of canonical WNT signaling between prenatal and postnatal development. Overall, precise regulation of canonical WNT signaling between the mesenchyme and epithelia is critical for Müllerian duct differentiation and fertility.

Materials and Methods

Mice

Amhr2-Cre^Δ/+ mice were maintained on a C57BL/6J; 129/SvEv mixed genetic background (21). These mice have an internal ribosome entry site-Cre cassette inserted into exon 5 of the endogenous Amhr2 locus, and a neomycin resistance selection cassette for gene targeting has been removed (Δ) by Flp recombinase (21). Ctnnb1 ex3^f/f mice were obtained from the Jackson Laboratory and maintained on a C57BL/6J; 129/SvEv mixed genetic background. Male Amhr2-Cre^Δ/+ mice were mated to female Ctnnb1 ex3^f/f mice to produce offspring with the genotype of Amhr2-Cre^Δ/+; Ctnnb1 ex3^f/+ (mutant) or Ctnnb1 ex3^f/+(control).

All animals were maintained in accordance with the Institutional Animal Care and Use Committee of the M.D. Anderson Cancer Center and the NIH Guide for the Care and Use of Laboratory Animals.

Tissue collection

Reproductive tracts were dissected from female mice at E18.5, postnatal day (P) 5, 3 weeks, 8 weeks, and 20 weeks of age. These time points were chosen for the following reasons: 1) E18.5 is immediately before birth and the final differentiation of the uterus and oviducts; 2) P5 is after birth, but prior to onset of uterine gland formation; 3) at 3 weeks the histology of the female reproductive tract is similar to an adult, but puberty has not been reached; 4) 8 weeks of age is after puberty has occurred and during peak fertility; and 5) 20 weeks is after some aging, but still within the fertile period. Noon on the day of the vaginal plug was considered E0.5. At all time points, the uterine horn length was measured from the uterotubal junction to the uterine bifurcation. The reproductive tracts from the E18.5 samples were collected with the ovary intact for fixation in 4% paraformaldehyde in PBS. For RNA, the ovary and anterior portion of the tract were removed and the posterior uterine region was snap frozen in liquid nitrogen and stored at −80°C. For the postnatal time points, the ovary was removed and a portion of the center of the uterine horn was fixed for histology. Another portion of the uterine horn was collected for RNA extraction. Regions from the vagina, oviduct, and uterotubal junction were also fixed for histology.

Histology

Sections of paraffin-embedded tissues (4 μm) were stained with hematoxylin and eosin. Van Giesen staining was performed on uterine sections from 20-week-old mice to identify collagen-rich regions (22). Briefly, slides were deparaffinized and rehydrated to PBS, stained with Weigert's iron hematoxylin for 15 minutes, and rinsed in water for 10 minutes. Slides were then placed in Van Giesen stain for 5 minutes, dehydrated through a graded series of alcohols to Histo-Clear (National Diagnostics) and coverslips were applied by standard methods.

Morphometry

To determine uterine gland density, nonserial sections (5 slides, 6 sections per slide) of the uterus from at least 4 mutant and 4 control mice at 3, 8, and 20 weeks were stained with hematoxylin and eosin and endometrial gland cross sections were counted. Uterine, endometrial, and myometrial area and myometrial width were determined using Image J v1.41 software (http://rsbweb.nih.gov/ij/index.html). Myometrial width measurements were made from multiple measurements of each uterine section. Uterine area was calculated with Image J software via outline of the outer longitudinal smooth muscle layer, the inner longitudinal smooth muscle layer, the endometrium, and the uterine lumen. Uterine area was calculated as a combination of endometrial and myometrial area measurements.

Myometrial layer cell volume was determined by staining the nuclei of uterine cross sections with 4′,6-diamidino-2-phenylindole. Z-stack images were taken every 2.5 μm for 10 μm. Regions from these stacks (n = 10) were selected from the inner circular and outer longitudinal myometrial layers based upon cellular morphology from 3 to 5 mice per genotype. These regions were analyzed by Imaris software (Bitplane Scientific Solutions) to determine region volume and number of nuclei per region to establish the average volume per cell within the myometrial layers.

Immunofluorescence

The following primary antibodies were used for immunofluorescence: CTNNB1 (1:250; Thermo Scientific, Labvision Corp., RB-9035), vimentin (VIM) (1:100; Sigma-Aldrich; V2258), pan-cytokeratin (1:100; Developmental Studies Hybridoma Bank;, TROMA-I), keratin14 (1:400; Covance; PRB-155P), smooth muscle α-actin (ACTA 1:600, Sigma-Aldrich A2547), E-cadherin (1:200; BD Biosciences; 610181), ERα (1:50; Dako; M7047), phospho-histone H3 (1:250 H3S10ph; Millipore; clone 3H10), and Twist1 (1:200; Abcam; ab50887). Immunofluorescence was performed using paraffin-embedded uterine tissue sections (5 μm). Briefly, slides were deparaffinized and rehydrated to PBS. Antigen unmasking was performed with boiling 0.05 M Tris buffer (pH 9.0) for 20 minutes. Nonspecific binding was blocked with 10% normal goat serum in PBS and 0.05% Tween 20 for 30 minutes at room temperature. Slides were then incubated with the primary antibodies in 10% goat serum in PBS and 0.05% Tween-20 overnight at 4°C, washed and incubated with Alexafluor-conjugated secondary antibodies (1:600; Molecular Probes) for 45 minutes. Slides were counterstained with Vectashield mounting media with 4′,6-diamidino-2-phenylindole (Vector Laboratories, Inc), and coverslipped.

In situ hybridization

Localization of Wnt5a and Wnt7a mRNA in uterine tissue sections (10 μm) was conducted by in situ hybridization analysis. Briefly, deparaffinized, rehydrated, and deproteinated uterine tissue sections were hybridized with digoxygenin-labeled antisense or sense cRNA probes generated from linearized cDNAs using in vitro transcription. Slides were hybridized, washed, blocked, and incubated with an antidigoxygenin antibody (Roche). Slides were then washed, stained with nitro-blue tetrazolium chloride/5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt. Slides were then coverslipped with an aqueous mounting media (VectaMount AQ, Vector Laboratories, Inc).

Uterine dye injection

To determine the functionality of the uterotubal junction, 0.25% bromophenol blue was injected into the posterior uterine lumen of 8-week-old mutant and wild-type mice with a 20-gauge needle, as described previously (7).

Quantitative real-time PCR

Quantitative real-time PCR analysis was performed on total RNA extracted from the uterine region at E18.5 and the uterus at 8 weeks from mutant and wild-type mice using Trizol reagent (Life Technologies). Tissues from mice were pooled together (n = 6 uteri per sample). The RNA (3 μg) was reverse transcribed using the SuperScript II first-strand synthesis system for RT-PCR (Invitrogen) with oligo-dT primers. The synthesized cDNA was mixed with SYBR Green PCR master mix (Applied Biosystems) and gene-specific primers. The primers for Wnt4, Wnt5a, and Wnt7a primers were used previously (23). The remaining primers were as follows (5′ to 3′): Goosecoid (Gsc) forward, GAA GCC CTG GAG AAC CTC TTC; and reverse, CCG AGG AGG ATC GCT TCT G; Slug (Snai2) forward, GCG CTC CTT CCT GGT CAA G; and reverse, GCT CCC GAG GTG AGG ATC TC; and Twist1 forward, ACG CTG CCC TCG GAC AA; and reverse, CAG ACG GAG AAG GCG TAG CT. Quantitative real-time PCR analysis was performed on an ABI PRISM 7900HT (Applied Biosystems) according to the manufacturer's instructions. The relative expression level for each amplicon was calculated by normalizing each cDNA to Gadph, and the ΔΔCt method was used to compare wild type to mutant. All experiments were performed in quadruplicate.

Tissue transplantation under the kidney capsule

B6129F1 females were ovariectomized at 6 weeks of age to serve as host mice and allowed at least 10 days to recover. Host mice were ovariectomized because uterine glands develop in the absence of circulating hormones (2, 3). These surgeries were performed following protocols described previously (13). Control and mutant uteri collected at E18.5 were recognized by the presence or absence of the anterior region of the tract and grafted under the kidney capsule of a single host female to ensure identical conditions. All grafts were harvested 21 days later at a stage equivalent to a 3-week-old uterus. Tissues were fixed in 4% paraformaldehyde in PBS at 4°C and processed for paraffin histology.

Results

Infertility and female reproductive tract abnormalities caused by stabilization of CTNNB1 in Müllerian duct mesenchyme

Mutant female mice were infertile, despite the presence of vaginal plugs following mating with wild-type males (data not shown). After birth, a rudimentary oviduct is present, which appears as a blind pouch (Figure 1, B, D, G, and H). Histological analysis of the oviductal remnant at 8 weeks of age reveals that it is similar to the isthmus region of control mice (Supplemental Figure 1, C and D, published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org). No histologic differences were found between the vagina of mutant and control mice (data not shown). At E18.5, there was no difference in length of the uterine region between the mutants and controls (Figure 1A, B, and I; P = .49). However, in prepubescent and adult mutant females, uterine length was decreased compared with that of controls (Figure 1I) at 3 weeks (P < .001), 8 weeks (Figure 1, C and D; P < .006), and 20 weeks of age (P < .002).

Figure 1.

Stabilization of CTNNB1 Prevents Oviduct Formation, Decreases Uterine Horn Length, and Compromises Uterotubal Junction Function. Gross morphology of reproductive tracts from control (A, C, E, and F) and mutant (B, D, G, and H) females at E18.5 (A, B, E, and G) and 8 weeks (C, D, F, and H). Uterine horn lengths were measured at E18.5, 3 weeks, 8 weeks, and 20 weeks (I). Blue dye was injected into the uterine lumen at 8 weeks in control (J) and mutant (K) mice. The dye entered the oviduct of mutant, but not control mice. Scale bars represent 1 mm (E–H, J, and K) and 2 mm (A–D). Asterisks indicate the absence of the oviduct, white arrowheads indicate where uterine horn length was measured, and black arrows indicate the uterotubal junction.

The uterotubal junction, connecting the oviduct to the uterus, is present in the mutant females; however, this junction does not function properly. In control mice, blue dye injected into the uterine lumen was unable to pass through the uterotubal junction into the oviduct and remained in the uterus. In mutant mice, blue dye passed through the uterotubal junction and into the oviduct remnant. The blue dye did not exit the oviduct but remained within the rudimentary oviduct lumen (Figure 1, J and K), which indicates that it is a blind pouch with no opening. Histologic analysis of the uterotubal junction of mutant mice revealed no obvious defects (Supplemental Figure 1, A and B) other than an increased density of uterine stroma underlying the epithelium in the tip of the uterine horn. The hypoplastic oviducts and dysfunctional uterotubal junction are likely contributing factors to the infertility of the mutant mice.

Stabilization of CTNNB1 results in endometrial and myometrial hypoplasia

Measurements of the inner circular and outer longitudinal myometrial layers revealed that the width of the inner circular layer is increased (P < .001) in the mutant mice compared with that of controls (Figure 2, A–E). The uterine area (P < .05) and endometrial area (P < .005) of 8-week-old mutant mice were reduced relative to those of controls (Figure 2, F–H). However, no differences in total myometrial area (P = .23), inner circular myometrial area (P = .22), and outer longitudinal myometrial area (P = .28) were found between the mutant and control mice (Figure 2H and data not shown). Uterine area is a combination of endometrial and myometrial area, indicating that the decrease in uterine area is specifically due to a reduction in endometrial area. Uterine area is reduced in the mutants, whereas myometrial area is unchanged, which resulted in an increase in the width of the inner circular layer of the mutants. Initial observations suggested that there was a decreased density of cells within the inner circular myometrial layer in the mutant mice, which may indicate hypertrophy (Figure 2, A, B, F, and G). However, measurements of cell volume within the inner circular and outer longitudinal myometrial layers revealed no differences at 8 weeks of age between mutants and controls (Supplemental Figure 2). Despite the fact that Amhr2-Cre should stabilize CTNNB1 within the myometrium, there are very few effects on the development and differentiation of the myometrium in mutant mice. This is consistent with immunostaining, showing that little or no increase of CTNNB1 was detected within the myometrium of mutant mice at 8 weeks of age (data not shown).

Figure 2.

Uterine Area Is Reduced and Inner Circular Myometrial Width Is Increased When CTNNB1 Is Stabilized. Immunofluorescence analysis of smooth muscle ACTA in control (A and C) and mutant (B and D) uteri at 8 weeks of age. Measurements of myometrial layer width at 8 weeks (E). Histologic analysis of uterine area (F and G) and morphometric area measurements within regions of the uterus (H; *, P < .001). ACTA-positive cells within the endometrium are red blood cells that exhibit autofluorescence. Arrows indicate width of the myometrial layers. Scale bars represent 200 μm (A, B, F, and G) and 100 μm (C and D). Endo, endometrium; Myo, myometrium; OL, outer longitudinal myometrium; IC, inner circular myometrium.

The decrease in uterine horn length and endometrial area indicates that there could be differences in cell proliferation in mutant mice compared with controls. Therefore, cell proliferation was examined in uteri of E18.5, P5, and 8-week-old mice by immunofluorescence for phospho-histone H3 and quantification of cells for each uterine cell compartment (Figure 3). At E18.5, fewer proliferating cells were detected in both the epithelia (P < .0001) and mesenchyme (P < .01) of mutant mice relative to controls (Figure 3, A, B, and E). By P5, less proliferation was detected in the epithelia (P < .0001) and stroma (P < .0001), but not the myometrial layers (P = .44) of mutant mice (Figure 3, C, D, and F). At 8 weeks, fewer proliferating cells were detected but only in the outer longitudinal myometrium (P < .001, data not shown). Terminal deoxynucleotide transferase-mediated dUTP nick end labeling staining was also performed at E18.5 and P5, and no differences in numbers of apoptotic cells were detected (data not shown).

Figure 3.

Stabilization of CTNNB1 Decreases Cell Proliferation Perinatally. Immunofluorescence labeling of phospho-histone H3 (H3S10ph; A and B) at E18.5 and P5. H3S10ph-positive cells were quantified within uterine cellular compartments at E18.5 (E) and P5 (F). Asterisks indicate P < .001. Scale bar represents 25 μm (A–D). Epi, epithelium; Mes, mesenchyme; LE, luminal epithelium; St, stroma; Myo, myometrium.

Uterine gland development is delayed when CTNNB1 is stabilized

Histologic analysis of the E18.5 uterus revealed a more cuboidal than columnar epithelium and increased cell density in the mesenchyme of mutant mice (Figure 4, A and B). Large masses of cells (Supplemental Figure 3B) were found in the deep stromal region of the endometria of mutant mice compared with controls at all postnatal time points. In control mice at 3 weeks of age, uterine glands have formed within the endometrium (Figure 4C). Interestingly, in 3-week-old mutant mice no uterine glands were present (Figure 4D). However, by 8 weeks, uterine glands were present in the endometrium of the mutants (Figure 4, E and F). At 20 weeks, uterine glands increased in both mutants and controls (Figure 4, G and H). Quantification of uterine gland cross sections revealed that there were fewer glands at all postnatal time points in the mutant mice relative to controls (P < .001; Figure 4I). These results suggest that the postnatal period of gland formation, which occurs independently of ovarian hormones, can be delayed by alterations in canonical WNT signaling.

Figure 4.

Uterine Gland Formation Is Delayed until after Puberty in Mice with CTNNB1 Stabilized in the Mesenchyme. Histologic comparison of uteri from control (A, C, E, and G) and mutant (B, D, F, and H) mice collected at E18.5 (A and B), 3 weeks (C and D), 8 weeks (E and F), and 20 weeks (G and H). Uterine glands per uterine cross section were counted at all postnatal time points (I). Epi, epithelium; Mes, mesenchyme; LE, luminal epithelium; GE, glandular epithelium; St, stroma; Myo, myometrium. Asterisks indicate P < .001. Scale bars represent 50 μm (A–H).

The Amhr2-Cre allele is active in the Müllerian duct mesenchyme at E12.5 but also in granulosa cells of the postnatal ovary (21, 24). Uterine gland development occurs in an ovary-independent manner prior to puberty (2, 3). However, the ovary does secrete factors that may influence uterine development (25). Mericskay et al. (13) determined that neonatal uteri do not develop endometrial glands when transplanted underneath the kidney capsule if the host animal is not ovariectomized (13). Therefore, uteri from control and mutant mice were placed under the kidney capsule of 6-week-old ovariectomized wild-type mice to determine whether stabilization of CTNNB1 within the ovarian granulosa cells directly affected uterine development (Figure 5, A and B). Uterine glands were detected in the endometrium of control mice, but not mutants (Figure 5, C–H). This suggests that the absence of uterine glands in the mutants 3 weeks after transplantation is due to the stabilization of CTNNB1 in the Müllerian duct mesenchyme-derived tissues, and not the ovarian granulosa cells.

Figure 5.

The Absence of Uterine Gland Development in Mutant Mice Is Not due to Ovarian Defects. Uteri from control (A, C, E, and G) and mutant (B, D, F, and H) mice were collected at E18.5 and transferred into the kidney capsule of ovariectomized mice for 3 weeks. Uteri were evaluated by immunofluorescence against pan-cytokeratin (C and D) and hematoxylin and eosin staining (E–H) for presence of endometrial glands. No glands were detected in mutant uteri. Arrowheads indicate uterine glands. Scale bars represent 50 μm (G and H) and 100 μm (C–F). LE, luminal epithelium; GE, glandular epithelium; Myo, myometrium.

CTNNB1 is increased in both the epithelium and mesenchyme of mutant uteri

At E18.5, CTNNB1 is expressed in the epithelium but not mesenchyme of control uteri, whereas expression is found in both the epithelium and mesenchyme in mutant uteri (Figure 6, A and B). At P5, CTNNB1 is expressed weakly in the luminal epithelium but not stroma and myometria of controls, but abundantly in the luminal epithelium, stroma, and myometria in mutant uteri (Figure 6, C and). At 8 weeks of age, CTNNB1 is expressed by both the luminal epithelium and glandular epithelium of controls and mutants (Figure 6, E and F). The abundance of CTNNB1 within the epithelia in E18.5 and P5 mutants is intriguing, because it suggests that paracrine signaling from the stroma to the epithelium is affecting CTNNB1 expression via canonical WNT signaling. CTNNB1 is expressed in the luminal epithelium, glandular epithelium, and stroma of controls during the period of early gland development (P11, Figure 6G).

Figure 6.

Stabilization of CTNNB1 in the Mesenchyme Increases CTNNB1 in Both the Epithelium and Mesenchyme and an EMT Occurs at E18.5. Immunofluorescence localization of CTNNB1 (A, B, G, and H) in the uteri of control (A, C, E, and G) and mutant (B, D, F, and H) mice at E18.5 (A, B, and H), P5 (C and D), 8 weeks (E and F), and P11 (G). Scale bars represent 50 μm. Epi, epithelium; Mes, mesenchyme; LE, luminal epithelium; St, stroma; GE, glandular epithelium.

Stabilization of CTNNB1 induces EMT in the Müllerian duct before birth

E-cadherin (CDH1) is typically expressed by epithelia and VIM by mesenchyme. To determine whether epithelial and mesenchymal differentiation was altered in the mutants, immunofluorescence was performed for CDH1 (Figure 7, A, B, E, and F) and VIM (Figure 7, C, D, G, and H) at E18.5 and 8 weeks. At E18.5, CDH1 was expressed in the epithelia and VIM was expressed in the mesenchyme in control mice (Figure 7, A and C). However, in mutant mice, VIM was expressed in both the epithelia and mesenchyme (Figure 7D) and CDH1 was undetectable (Figure 7B). The E18.5 VIM-expressing epithelium is columnar and does not express markers of stratified epithelia. This was determined by the absence of KRT14 expression by the uterine luminal epithelium in both mutant and control mice, but its presence in the vaginal epithelia (data not shown). The expression of VIM by the epithelium was not found in adult mice (Figure 7H).

Figure 7.

EMT Occurs at E18.5. Immunofluorescence localization of CDH1 (A, B, E, and F) and VIM (C, D, G, and H) in the uterus of control (A, C, E, and G) and mutant (B, D, F, and H) mice at E18.5 (A–D) and 8 weeks (E–H). Asterisk indicates loss of CDH1 in the epithelium and arrowhead indicates VIM expression within the epithelium. Scale bars represent 25 μm. Epi, epithelium; Mes, mesenchyme; LE, luminal epithelium; St, stroma; GE, glandular epithelium.

Several WNTs are known to be expressed by the Müllerian duct and uterus, including Wnt4 (mesenchyme), Wnt5a (epithelia and mesenchyme), and Wnt7a (epithelia). Transcript levels for these 3 Wnts assessed by quantitative RT-PCR were increased in the Müllerian duct of mutant mice at E18.5 compared with controls (Figure 8, A–C).

Figure 8.

Stabilization of CTNNB1 Alters Abundance and Expression Patterns of Wnts and TWIST1. Real-time RT-PCR (qPCR) of Wnt4 (A), Wnt5a (B), Wnt7a (C), Goosecoid (Gsc) (D), Slug (Snai2) (E), and Twist1 (F) in uteri from control and mutant uteri at E18.5. In situ localization of Wnt5a (G and H) and Wnt7a (I and J) mRNA and immunohistochemical analysis of TWIST1 (K and L) expression patterns at E18.5. Blue dash lines surround the epithelium (H and J), and white dash lines indicate TWIST1 restriction to the mesometrial region in mutant mice (L). Scale bars represent 100 μm (G–J) and 50 μm (K and L). Asterisks indicate P < .0001. Epi, epithelium; Mes, mesenchyme; M, mesometrial region; A-M, antimesometrial region.

Loss of epithelial expression of CDH1 is a key feature of an EMT event and CTNNB1 acts directly on several EMT genes, including Gsc, Snai2, and Twist1, which in turn act upon Foxc2 (26). Thus, expression levels of several EMT genes were investigated by quantitative RT-PCR. Stabilization of CTNNB1 has no effect on Foxc2 mRNA levels at E18.5 (data not shown). However, Gsc and Snai2 mRNA are decreased in mutant mice and Twist1 mRNA is increased relative to controls (Figure 8, D–G). TWIST1 is expressed throughout the entire mesenchyme of control mice before birth (Figure 8K). However, TWIST1 expression is restricted to the mesenchyme on the mesometrial side of the uterus in mutant mice at E18.5 (Figure 8L). Alterations in Snai2 and Twist1 gene expression may be sufficient to induce an EMT event before birth in mutant mice.

At birth, newborn pups are removed from the maternal endocrine environment, which is rich in estrogens and progesterone. We hypothesize that this change in endocrine environment might be responsible for the discontinuation of the EMT event after birth, as shown by epithelial expression of CDH1 (Figure 7, E and F) and stromal expression of VIM (Figure 7, G and H). Thus, immunofluorescence for estrogen receptor α (ESR1) was performed at E18.5 and P5. ESR1 is expressed by the mesenchyme (E18.5) and stroma (P5) in both control and mutant mice but was also expressed by the epithelia in mutant mice at E18.5 (Figure 9, A–D).

Figure 9.

ESR1 Is Expressed by the Epithelia before Birth in Mice with Stabilized CTNNB1. Immunohistochemical localization of ESR1 at E18.5 and P5 in control and mutant mice. Asterisk indicates expression of ESR1 in the epithelia. Epi, epithelium; Mes, mesenchyme; LE, luminal epithelium; St, stroma; myo, myometrium. Scale bar represents 100 μm.

Formation of endometrial stromal sarcomas occurs as early as 3 weeks of age in mutant mice

Stabilization of CTNNB1 is known to cause tumor formation in many different tissues, including the uterus (9, 17). Large masses of cells were observed within the endometrium of mutant mice at 3, 8, and 20 weeks of age, but not E18.5 or P5 (Supplemental Figure 3B). Further examination of these masses suggests that they are endometrial stromal sarcomas. They are collagen rich, smooth muscle α actin (ACTA) negative, and ESR1 positive, which are characteristics of endometrial stromal sarcomas (Supplemental Figure 3).

Discussion

Stabilization of mesenchymal CTNNB1 prevents normal female reproductive tract growth

To understand the role of canonical WNT signaling during Müllerian duct differentiation and uterine development, female mice were generated in which CTNNB1 was stabilized in the Müllerian duct mesenchyme at E12.5 using Amhr2-Cre. These mice suffer severe morphologic defects in reproductive tract development, including a failure of oviduct formation, defects in the function of the uterotubal junction, reduced uterine growth, and delay in uterine gland formation. Perhaps not surprisingly, these mutant females were infertile. A previous report found that mice with the same genotype were subfertile (19), whereas ours were infertile. It is possible that differences in genetic background or environment contribute to this variation in fertility.

Uterine horn length is decreased postnatally in mutant mice. The presence of fewer proliferating cells in both epithelia and mesenchyme prenatally likely contributes to the reduced uterine size. After birth, fewer proliferating cells were found in the epithelia and stroma, but not in the myometrium. As a consequence, no differences in myometrial area were observed. Overall it appears that less cell proliferation within the epithelia and mesenchyme/stroma results in uterine growth reduction of both horn length and endometrial area, whereas myometrial growth and development are largely unaffected by stabilization of CTNNB1. We observed that higher levels of Wnt5a and Wnt7a are accompanied by less cell proliferation within the epithelium and mesenchyme/stroma, contributing to the decrease in uterine horn length and/or the absence of the oviduct. Deletion studies found that Wnt7a suppresses cell death and controls patterning and growth of the oviduct, uterus, and vagina (5, 30). Additionally, Wnt5a promotes cell proliferation and inhibits apoptosis in cell culture (31, 32). This suggests that the observed increases in Wnt expression levels in mutant mice were likely unrelated to the observed changes in cell proliferation.

The loss of the oviduct in the mutants is consistent with the idea that Ctnnb1 mediates anti-Müllerian hormone (AMH)-induced Müllerian duct regression in males (27). We hypothesized that stabilization of CTNNB1 in the mesenchyme of females would cause Müllerian duct regression. However, the loss of only the oviduct and not the uterus was surprising. Müllerian duct development appears normal in the mutant mice prior to E17.5 (data not shown). Therefore, the loss of the oviduct does not coincide with the normal pattern of AMH-induced Müllerian duct regression found in male fetuses. This suggests that the loss of the oviduct in the mutants is independent of the AMH pathway. The oviduct phenotype of our mutants is distinctly different from other oviduct mutants that result in patterning differences, such as Hoxa10- (28) and Wnt7a-null mice (5) or when a dominant-active allele of smoothened (SmoM2) is conditionally overexpressed in the Müllerian duct-derived mesenchyme (29).

Uterine gland development is delayed by the stabilization of CTNNB1

Endometrial glands are present in all mammalian uteri and transport and/or produce secretions required for embryo survival and implantation. A number of mouse models exist that exhibit defects in uterine gland development, including deletion of Wnt5a (13), Wnt7a (5), conditional deletion of Wnt4 in the postnatal endometrium (12), Lif (4), Hoxa11 (10), conditional deletion of Dicer in the Müllerian duct mesenchyme (7, 8), and conditional deletion of Ctnnb1 in the Müllerian duct mesenchyme and the postnatal endometrium (9, 18, 33), as well as a uterine gland knockout sheep model (34). All of these models are infertile, with the exception of Wnt5a, which is neonatal lethal. The analysis of postnatal uterine development in the Wnt5a-null mice was performed via kidney capsule grafting (13). In the current study, gland development was not ablated, but rather delayed and reduced in the mutants. Grafting uteri under the kidney capsule of recipients shows a similar absence of uterine glands. This suggests that either the stabilization of CTNNB1 within the granulosa cells of the ovary is not responsible for the absence of uterine glands or that stabilization of CTNNB1 in the granulosa cells prior to E18.5 is sufficient to influence uterine development. However, the prepubertal period of uterine gland formation occurs independently of the ovary (2, 3), so it is unlikely that alterations in ovarian gene expression during midgestation would have lasting effects on the initiation of uterine gland development 1 week after birth. Furthermore, the delay in uterine gland formation is comparable to what is observed when mice are treated postnatally with estradiol or progesterone. Differences in uterine CTNNB1 and ESR1 expression were also detected in the hormone-treated mice (35). Increased abundance of CTNNB1 occurs in mouse models of delayed uterine gland formation. In this case, increased CTNNB1 in the mesenchyme exhibits the same phenotype, suggesting that CTNNB1 is sufficient for delaying uterine gland formation. ESR1 is not essential for normal uterine gland development, but an increase in ESR1 is associated with delayed uterine gland development. However, ESR1 expression is reduced after birth in the mutants, indicating that it is not directly regulated by CTNNB1. Increased levels of Wnt5a and Wnt7a in the mutants, combined with their known role in uterine gland formation and an abundance of epithelial CTNNB1, strongly suggest that WNT signaling is responsible for the delay in uterine gland formation.

Previous studies have determined that uterine gland formation, or adenogenesis, occurs within a defined postnatal period in a number of species. If this window is bypassed, as in the mouse models mentioned previously or uterine gland knockout sheep (34), gland development does not occur and fertility is compromised. In rodents, adenogenesis begins within 1 week of birth (36) and occurs independently of steroid hormone production by the ovary prior to puberty (3) and after parturition, but not after puberty (35). The formation of uterine glands after puberty suggests that the period before puberty is sensitive to effects of mesenchymal CTNNB1. The frequent regeneration of uterine glands in women with every menstrual cycle provides numerous opportunities for deviations in gland development that could impact fertility. Similarities with mouse uterine gland development and regeneration suggest that understanding the cellular and molecular events regulating adenogenesis in mice may shed light upon this process in humans.

Mesenchymal CTNNB1 regulates epithelial cell differentiation

Epithelial-mesenchymal interactions are essential during embryonic development, most notably in tubule formation. These interactions allow for communication between different cell types in order to direct and coordinate cell behavior (37–41). Elegant tissue recombination studies of the rodent female reproductive tract have demonstrated that the mesenchyme directs and specifies patterns of epithelial cell development, whereas the epithelium is necessary for mesenchymal differentiation and organization (37, 42). Interestingly, the only direct effects of CTNNB1 stabilization on the mesenchyme in our mutant are a decrease in endometrial area, fewer proliferating cells within the mesenchyme/stroma, and the presence of endometrial stromal sarcomas. However, this results in significant alterations in the epithelium specifically before birth, including fewer proliferating cells, abundance of cytoplasmic CTNNB1, misexpression of ESR1, delayed uterine gland formation, and induction of an EMT event. Because CTNNB1 cannot signal in a paracrine manner, these effects on the epithelium must be mediated through an unknown factor(s). It is unclear whether these epithelial changes are due to the action of a factor(s), perhaps up-regulation of a WNT(s), which in turn acts upon the epithelia. This is corroborated by the CTNNB1 expression within the epithelium, in addition to the mesenchyme when CTNNB1 is stabilized in the mesenchyme.

An EMT event is induced before birth by stabilization of mesenchymal CTNNB1

EMT is a cellular program that allows polarized epithelial cells to become mesenchymal, resulting in changes to cell morphology and migration capability. This program is involved in a number of developmental events (26), neural crest migration, as well as formation of the secondary palate, cardiac valve, and mesoderm (43). Stabilization of CTNNB1 in the Müllerian duct mesenchyme induces an EMT event before birth. This EMT event involves a down-regulation of CDH1 coinciding with an up-regulation of VIM in the epithelium. Interestingly, the developing Müllerian duct at E12.5 and E13.5 looks morphologically epithelial yet it expresses VIM but not CDH1. Subsequently, VIM expression is lost and CDH1 expression is acquired that persists (1). Thus, during normal development the Müllerian duct transitions from a mesoepithelial to epithelial character. Perhaps stabilization of CTNNB1 in the mesenchyme delays or extends the initial mesoepithelial character of the Müllerian duct epithelium or initiates a de novo EMT. This would also explain subtle differences observed in epithelial histology of mutants before birth (Figure 4, A and B). The effect of the extended EMT appears to be restricted to impaired Müllerian duct development and delayed uterine gland formation, because cell morphology of the uterus and vagina is normal. Furthermore, there is an increase in Twist1 and a decrease in Snai2 and Gsc expression in the mutant mice. TWIST1 is expressed by mesenchyme in both mutant and control mice, but is restricted to the mesometrial side of the uterus in mutant mice. This may be due to the anti-mesometrial-specific localization of Amhr2-Cre activity. Twist1 is considered a master regulator of EMT, because it can transcriptionally repress Cdh1 (45) and induce an EMT event alone (46). However, TWIST1 is not expressed by the uterine epithelia and must be acting via another EMT gene.

Snai1 binds directly to Esr1 cis-elements to down-regulate expression in Esr1-positive MCF-7 cells (47). It is unclear whether other EMT genes can bind directly to Esr1 cis-elements. Furthermore, Esr1 negatively regulates EMT and Snai2 expression in a model of breast cancer (48). ESR1 can undergo ligand-independent activation via growth factors, in the absence of estrogen, which is likely the case before birth in the mutant mice. Expression of ESR1 in the epithelia before birth is likely due to the EMT, in which the epithelial cells are behaving as mesenchyme. After birth, the release from maternal hormones may down-regulate ESR1 in the epithelia, and subsequently end the EMT or vice versa.

CTNNB1 in female reproductive tract tissues

Stabilization of CTNNB1 using Amhr2-Cre has been previously reported. However, these earlier studies focused primarily on the role of CTNNB1 in granulosa cell tumors or endometrial stromal sarcomas and leiomyomas. The current study was performed to specifically elucidate the effects of CTNNB1 stabilization within the Müllerian duct mesenchyme during perinatal Müllerian duct differentiation and development.

Conditional deletion of Ctnnb1 in the Müllerian duct mesenchyme causes infertility and results in reduced coiling of the oviduct and smaller uteri (16). Interestingly, we found that stabilization of CTNNB1 in the same tissue results in a similar phenotype. The similarity in phenotype between conditional deletion and conditional stabilization of CTNNB1 is not unprecedented. Both overexpression and deletion of Pax6 result in microphthalmia, although this occurs via distinct mechanisms (38). Perhaps the same situation is occurring with Ctnnb1.

Deletion and stabilization of CTNNB1 using progesterone receptor (Pgr)-Cre, which is active postnatally in both the epithelium and mesenchyme, results in squamous cell metaplasia and endometrial gland hyperplasia, respectively (9). These phenotypes were not observed when CTNNB1 is stabilized or deleted in the mesenchyme, indicating that these are likely due to the effect of stabilization of CTNNB1 in the epithelium or the onset of stabilization at midgestation in the current model.