Wnt5a is required for proper epithelial-mesenchymal interactions in the uterus

The mammalian female reproductive tract (FRT) is derived from the Müllerian ducts giving rise to the oviducts, uterine horns, cervix and the anterior vagina (Cunha,1975). The murine FRT is immature at birth and consists of two tubes of simple columnar epithelia surrounded by a mesenchymal sheath fused at the level of the cervix. Distinct cytodifferentiation occurs during postnatal development and differentiation is complete 2 weeks after birth. Uterine horns develop postnatally to form an external myometrium surrounding the mesenchymal(stromal) compartment which contains glands. By contrast, the vagina and cervix do not develop glands and the luminal epithelium undergoes a transition from simple columnar to squamous (stratified) morphology. Experiments in which neonatal epithelium from any part of the FRT is recombined with presumptive uterine or vaginal mesenchyme reveals that the epithelium is developmentally plastic and adopts either a uterine (simple columnar) or vaginal(squamous/stratified) epithelial fate dependent upon the origin of the mesenchyme (Cunha, 1976; Kurita et al., 2001). Recent results using grafts between estrogen receptor α (Esr1) mutant and wild-type FRTs demonstrates that Esr1 function is required in the mesenchyme but not in the epithelium to mediate estrogen-mediated responses in the epithelium (Cooke et al.,1997; Kurita et al.,2000). Taken together, these studies demonstrate that FRT mesenchyme delivers developmental and estrogenic signals to the epithelium.

Wnt genes encode secreted glycoproteins that regulate cell and tissue growth and differentiation (Polakis,2000) and activate multiple signaling pathways through the frizzled receptors and the cytoplasmic signaling protein, dishevelled(Pandur et al., 2002). We identified three members of the Wnt gene family (Wnt4, Wnt5a, Wnt7a)that are expressed at high levels in the adult FRT throughout development(Miller et al., 1998b). At birth, Wnt4 expression is restricted to the uterine mesenchyme. By contrast, Wnt5a is distributed throughout the mesenchyme of the uterus, cervix and vagina, whereas Wnt7a is in the epithelium. During postnatal differentiation, Wnt5a and Wnt7a become restricted to the uterine horns, whereas Wnt4 expression is activated in the stratified epithelium of the cervix and the vagina(Miller et al., 1998b). We noted that the levels and the sites of Wnt expression fluctuate during estrous, suggesting a continued role in the adult(Miller et al., 1998b).

Mice corresponding to all three Wnt genes expressed in the FRT have been generated. Wnt4 mutants fail to form Müllerian ducts and die at birth due to numerous defects, thus an analysis of how Wnt4 contributes to later FRT development is unknown (Vainio et al., 1999). Wnt7a mutants are viable and exhibit malformations in the FRT including shortened and uncoiled oviducts,hypoplastic uterine horns and a vaginal septum(Miller and Sassoon, 1998; Parr and McMahon, 1998). The uterus is most affected with a marked reduction in the stromal compartment,accompanied by a lack of uterine glands and a disorganized myometrium. In addition, Wnt7a mutant uterine epithelium fails to maintain a normal columnar phenotype and becomes stratified upon puberty(Miller and Sassoon, 1998). It was subsequently demonstrated that fetal diethylstilbestrol (DES) exposure transiently represses Wnt7a expression in the Müllerian ducts and is sufficient to recapitulate the Wnt7a mutant FRT phenotypes providing a molecular basis for environmental endocrine disruption(Miller et al., 1998a).

In this study, we examined the role of Wnt5a, which is expressed in the FRT mesenchyme, and thus is a good candidate as a potential mediator of mesenchymal-epithelial interactions(Miller et al., 1998b; Pavlova et al., 1994). Wnt5a mutants die at birth due to a failure to complete anteroposterior body axis development(Yamaguchi et al., 1999). In order to circumvent the neonatal lethality, we grafted neonatal mutant FRT tissue into adult hosts to assess postnatal potential and phenotypes. We find that although the oviduct, uterine and cervical compartments of the FRT develop in the absence of Wnt5a, the mutant uterus fails to form glands that are essential for adult function. In addition, we demonstrate that Wnt5a is required for the complete repertoire of estrogen-mediated cellular and molecular responses. Furthermore, Wnt5a participates in a regulatory loop with Wnt7a and is required for correct regulation of Hoxa10 and Hoxa11, which control anteroposterior patterning of the FRT (Benson et al.,1996; Hsieh-Li et al.,1995). These data shed light upon the mechanism by which uterine stroma mediates both developmental and estrogen-mediated changes in uterine epithelium and reveal that Wnt5a is required in these processes.

Wnt5a mutant mice were obtained from A. McMahon(Yamaguchi et al., 1999) and maintained in a C57BL6/SV129 mixed background. Neonates were obtained after delivery or C-section on day 19 of gestation. Wnt7a mutant mice were obtained from B. Parr and A. McMahon (Parr and McMahon, 1998) and maintained in a SV129 background. Lef1 mutant mice were kindly obtained from R. Grosschedl(van Genderen et al., 1994)and were maintained in a C57BL6/SV129 mixed background. Nude mice in a C57BL6 background were purchased intact or ovariectomized from Taconic, Germantown,NY. All procedures for handling of mice, housing and maintenance were performed according to approved institutional guidelines. All surgical procedures were prior approved by the institution according to NIH guidelines.

Wild-type and mutant neonate FRT fragments were grafted under the renal capsule of each kidney of the same adult female nude host to ensure identical hormonal conditions. Grafting procedures were performed as previously described (Cunha, 1976). Adult female hosts were ovariectomized 3-4 weeks prior to grafting where indicated. Separation of the neonatal epithelium from the mesenchyme for recombination between wild-type and Wnt5a mutant tissues was performed as previously described (Bigsby et al.,1986). Diethylstilbestrol (DES) administration was delivered i.p. suspended in saline between day 18 to 20 post graft implantation following a previously described protocol (Miller et al., 1998a). Host FRT and neonate grafts were harvested 24 hours after the last injection on day 21, i.e. a developmental stage equivalent to 3-week-old FRT, and were fixed o/n in 4% PAF 4°C and processed for paraffin histology.

Wnt cDNAs encoding HA tagged Wnt4 and Wnt5a were inserted in QCX backbone vectors derived from MLV retrovirus and produced as previously described (Julius et al.,2000; Shimizu et al.,1997). Retroviral supernatants were concentrated by ultracentrifugation, 2 hours at 100 g in a Beckman SW28 rotor. Titer was estimated to 1×10⁶ infection unit by lacZstaining of NIH3T3 cells infected with a parallel QC-lacZ prep. Western blot and in situ immunofluorescence using anti-HAtag high affinity, rat monoclonal antibody (3F10) from Roche Diagnostics (Mannheim) were performed on infected NIH3T3 to confirm the synthesis of the Wnt factors by the retrovirus. Neonate uterine fragments were infected overnight in retroviral supernatant resuspended in DMEM/20% FCS/4 μg/ml hexadimethrine bromide (H9268,Sigma).

In situ hybridizations for the Wnt, Hoxa and Msx1 were performed as previously described (Miller and Sassoon, 1998). Esr1 and Pgr probes were kindly provided by G. Cunha. RNA probes were labeled with ³⁵S-UTP. Black and white dark field images were converted to using Adobe Photoshop to allow superimposing upon phase contrast images.

Wnt5a heterozygote mice were crossed to generate a total of 242 neonates. We obtained 40 Wnt5a^-/- pups, which falls below the number predicted by Mendelian genetics (∼60 pups or 16.53% versus 25%), indicating loss of mutant fetuses in utero. Of 40 mutant pups, 17 were males and 19 females revealing no gender bias in survival during gestation. We could not determine the sex of 4 pups because of extreme reduction in posterior development. The anterior Müllerian-derived structures(oviducts and uterine horns) could easily be identified, whereas posterior derived structures (cervix and vagina) were absent(Fig. 1A,B). The uterine horns are either fused at the midline (Fig. 2) or terminate as a blind pouch(Fig. 1B). The Wnt5amutant uterine horns have an undulated lumen and show a 60 to 90% reduction in length when compared with wild types (see also left panels in Fig. 2). The oviducts are less affected and we observe correct narrowing of the anterior uterine horn at the level of the uterotubal junction with the oviduct. The anterior border of expression for Hoxa10 defines the site of the uterotubal junction(Benson et al., 1996). Using whole-mount in situ hybridization, we confirmed the position of the uterotubal junction in the Wnt5a mutant and in the wild type(Fig. 1). Expression of Wnt7a and Msx1 is detected in the epithelium of the Wnt5a mutant FRT (Fig. 1E-H) whereas the expression of Wnt4 is only slightly reduced when compared with wild-type uterus and is not affected in the ureters(Fig. 1I,J). The expression of the Wnt5a mutant transcript is not affected in the absence of Wnt5a in the FRT (Fig. 1K,L) as previously seen for the Wnt5a mutant limb bud(Yamaguchi et al., 1999). We conclude that loss of Wnt5a affects posterior growth of the Müllerian duct.

In order to circumvent the perinatal lethality of the Wnt5amutant, we grafted newborn (postnatal day 0/embryonic day 19-20) FRT fragments under the renal capsule of cycling nude females (8 week old). Wnt5a^-/-, Wnt5a^+/– and Wnt5a^+/+ FRT tissues were grafted and grown for 3 weeks in the same host to ensure identical host conditions. FRTs were cut into size-matched fragments along the anteroposterior axis prior to grafting(Fig. 2). FRTS harvested from a total of four Wnt5a^+/+, four Wnt5a^+/– and five Wnt5a^-/- pups,and were analyzed using grafting procedures with multiple grafts per individual. The ovaries, oviduct and uterine regions, which normally express Wnt5a, developed correctly for all genotypes(Fig. 2). The epithelial cells display the normal characteristics of square-shaped ciliated cells for the oviduct and tall columnar cells for the uterus in the Wnt5a mutant compared with wild-type grafts. The smooth muscle cell layers formed normally in the absence of Wnt5a, although they appear thinner when compared with control grafts. This is in contrast to the Wnt7a mutant, which shows a hyperplastic and hypertrophied smooth muscle compartment(Miller and Sassoon, 1998). Smooth muscle myosin heavy chain in situ hybridization revealed correct differentiation of smooth muscle layers in the Wnt5a mutant (data not shown). We observed grafts that developed stratified epithelium in two out of five independent Wnt5a mutant grafts which were derived from the most posterior portion of the neonate Wnt5a mutant Müllerian ducts(Fig. 3A-C). The morphological columnar-to-squamous junction of the epithelium was accompanied at the molecular level by the correct boundary of Msx1 and Wnt7aexpression and activation of Wnt4 in the stratified epithelium(Fig. 3D-F; data not shown). This result reveals the potential to form a cervix in the Wnt5amutant despite the lack of a morphologically defined cervix at birth.

We observed that few glands develop in wild-type uterine grafts grown in adult cycling females. Uterine glands normally appear by 1-2 weeks after birth in situ. In the grafts, a small number of glands appear after 5 weeks of growth under the renal capsule showing an abnormal delay when compared with gland formation in the uterus in situ (data not shown). We reasoned that this delay in glandulargenesis may be caused by precocious exposure of neonatal uterine grafted tissues to adult levels of sex hormones present in the cycling female hosts as perinatal exposure of the FRT to sex hormones is linked to deficient glandulargenesis (Branham et al.,1985a; Branham et al.,1985b; Gray et al.,2001). As normal (postnatal) glandulargenesis occurs in the immature uterus in the absence of high levels of circulating steroid hormones,we placed grafts into female hosts that were ovariectomized 2 weeks prior to the grafting procedures. Under these conditions, numerous and normal appearing glands developed in all the grafts (n=18; Fig. 4A). In addition, the tissue organization of the grafts and the luminal folds were indistinguishable from uteri in situ at an equivalent stage of post-natal development (3 weeks). Thus, this procedure for obtaining normal morphogenesis and maturation of wild-type grafted uterine tissues allows for assessment of the outcome of mutant FRT development, particularly in the case of perinatal lethality. Using ovariectomized hosts, 10 out of 12 grafts derived from eight independent Wnt5a mutant mice did not develop glands whereas the remaining two grafts developed very few glands (Fig. 4B; data not shown). The overall morphology of the Wnt5amutant grafts is otherwise normal, although the smooth muscle layers are moderately thinner compared with the wild-type grafts.

We had previously observed that Wnt7a mutant females do not develop uterine glands (Miller and Sassoon, 1998), suggesting a potential genetic interaction between Wnt7a and Wnt5a. However, we had analyzed the Wnt7auterine phenotype from samples obtained directly from postnatal mutant females. To compare directly our results under the same experimental conditions, we tested Wnt7a mutant uterine grafts in ovariectomized hosts. The Wnt7a mutant uterine grafts (n=3) fail to develop glands in ovariectomized hosts and recapitulate the myometrial and epithelial phenotypes that have been reported in our previous studies(Miller and Sassoon, 1998)(Fig. 4C).

Lef1 is a transcription factor that interacts with β-catenin and mediates the canonical Wnt pathway(McKendry et al., 1997). Lef1 is expressed in the Müllerian duct mesenchyme(Oosterwegel et al., 1993);however, Lef1 mutant mice die several days after birth prior to cytodifferentiation of the FRT and uterine gland formation. As Lef1is required for glandulargenesis in the mammary gland and mediates numerous epithelial-mesenchymal interactions during development(van Genderen et al., 1994),we evaluated the potential role of Lef1 in uterine postnatal development using grafting procedures. We find that Lef1 is not required for uterine development. Moreover, glandulargenesis proceeds normally in Lef1 mutant grafts (n=2). In addition we generated grafts(n=2) from double Wnt7a/Lef1 mutants that are indistinguishable from Wnt7a mutant grafts(Fig. 4D,E). Taken together,these results show that Wnt5a and Wnt7a are required for gland formation in the uterus and participate in a signaling pathway that does not require Lef1.

Uterine gland formation initiates on postnatal days 7-9 (P7-9). and they continue to grow and increase in number until puberty. By P15, Wnt7ais expressed exclusively in luminal epithelium but not in glandular epithelium(Fig. 4G). Wnt7a is also expressed in the deep folds of the luminal epithelium that start to form between P5 and P7 (Brody and Cunha,1989). In wild-type grafts, we observe a sharp boundary of Wnt7a expression at the transition between luminal and glandular epithelium (Fig. 4J). Wnt4 expression is restricted to the stroma between the luminal epithelium and adjacent uterine glands(Fig. 4I). Wnt5aexpression is abundant throughout the stroma that extends from the subepithelial region up to the inner smooth muscle layer. This domain of expression includes stroma surrounding the folding luminal epithelium and more distal glands (Fig. 4H). In addition, low but detectable levels of Wnt5a expression are observed in luminal and glandular epithelium. These patterns of expression suggest that Wnt7a and Wnt5a act in juxtaposed compartments to control gland formation. The sharp boundary of Wnt7a expression at the site of glandular invagination suggests a mechanism whereby Wnt7a is repressed locally to allow luminal epithelium to participate in gland formation. If this model is correct, then Wnt5a may provide a permissive environment for proper regulation of Wnt7a. To determine if Wnt5a expression is required in the stroma to mediate gland formation, we performed recombinant graft experiments using wild-type and Wnt5a mutant (–/–) uterine fragments(Fig. 5A). When wild-type stroma is recombined with wild-type epithelium or Wnt5a mutant epithelium, recombinant grafts form glands(Fig. 5B,D). When Wnt5a mutant stroma is recombined with Wnt5a mutant or wild-type epithelium, no glands developed, except for one mutant mesenchyme/wild-type epithelium graft that developed a single gland(Fig. 5C,E; data not shown). These results demonstrate that Wnt5a is required in the stroma for glandulargenesis.

The inability of the Wnt5a mutant FRT to form glands may reflect a requirement for Wnt5a directly at sites of gland formation or an earlier requirement during fetal development to promote the survival of a unique population of cells in the FRT that direct this process. To test these possibilities, uterine fragments were infected with retroviral vectors that are expressed only in the stroma, based on histological markers such as human placental alkaline phosphatase or β-galactosidase (data not shown). Ectopic Wnt5a expression in Wnt5a mutant neonatal uterine grafts rescues gland formation in discrete regions of the uterine grafts in three out of three independent grafts whereas no glands are formed in uterine grafts derived from the same tissues infected with an empty vector(n=2) (Fig. 5F,G). Interestingly, Wnt4 overexpression is unable to rescue gland formation in the Wnt5a mutants (n=2)(Fig. 5H) demonstrating that Wnt4, which is also expressed in the uterine mesenchyme is unable to substitute for Wnt5a and that specific roles likely exist for each ligand in this system.

We evaluated the morphological and cellular responses of the Wnt5amutant uterine graft to the potent estrogen, DES (diethylstilbestrol), which normally elicits pronounced cellular and molecular changes in uterine tissue. Grafts were allowed to grow in ovariectomized hosts for 18-20 days followed by three daily injections of DES or saline followed 24 hours by harvesting the grafts. Wild-type (n=3) and Wnt5a^+/– grafts(n=6) responded to DES exposure with the typical changes associated with the estrogenic response, i.e. hypertrophy and hyperplasia of the luminal and glandular epithelial cells and the distension of the stromal compartment which is associated with the changes in vascular permeability that occur upon estrogenic compounds exposure (Korach and McLachlan, 1995) (Fig. 6,compare B with A). All Wnt5a mutant grafts (n=4)exposed to DES responded by a normal increase in cellularity and thickness of the epithelial compartment (although no glandular response is measured)(Fig. 6D,H). The average luminal epithelium height was assayed in three independent grafts of each genotype. The epithelium height changed from 19.4±3.8 μm in the absence of DES to 38.4±0.2 μm after DES exposure in the Wnt5a mutant graft (compare Fig. 6G,H), showing no significant difference with the wild-type grafts, 20.1±1.5 μm in saline conditions to 38.3±5.1 μm after DES exposure (Fig. 6E,F). By contrast, the global response of the Wnt5a mutant grafts was abnormal in appearance. The uterine walls of the mutant grafts did not enlarge in response to DES, but instead underwent an unusual dilation. In addition, we note that two out of the four Wnt5a mutant DES exposed grafts did not display stromal edema (Fig. 6D;data not shown).

We analyzed the expression of Wnt7a in saline and DES exposed Wnt5a mutant grafts and compared the pattern with wild-type grafts grown in the same hosts (Fig. 6I-L). We observe that Wnt7a is expressed throughout the luminal epithelium of 3-week-old Wnt5a^+/– control and Wnt5a^-/- mutant grafts grown in saline injected ovariectomized host (Fig. 6I,K). Wnt5a^+/– and wild-type grafts show the expected downregulation of Wnt7a following exposure to DES(Fig. 6J; data not shown). By contrast, exposure to DES is unable to repress Wnt7a in the Wnt5a mutant graft (Fig. 6L). In addition Hoxa10 and Hoxa11, which are also repressed by DES exposure in utero(Block et al., 2000; Miller et al., 1998a) are not repressed by DES exposure in the Wnt5a mutant grafts(Fig. 6M-T, compare P with N and T with R).

To determine if prolonged exposure to high levels of a synthetic estrogen adequately reproduces the endogenous regulation of Wnt7a and the Hoxa genes by Wnt5a, we analyzed uterine grafts grown in intact cycling hosts that were sacrificed at different period of the estrous cycle. Expression patterns of Wnt7a, Hoxa10 and Hoxa11 are normal in Wnt5 mutant grafts harvested at diestrus from the host, when levels of estrogen are low and levels of progesterone are high(Fig. 7A). During proestrus,when the levels of estrogen are high and the levels of progesterone are low,we find that Wnt7a levels remain high in Wnt5a mutants(Fig. 7B). Similarly, Hoxa10 and Hoxa11 remain high during proestrus in the Wnt5a mutant grafts. These results show that estrogen mediated repression of Wnt7a in the epithelium and repression of Hoxa10 and Hoxa11 in the stroma is dependent upon Wnt5a expression.

One simple mechanism underlying abnormal estrogen responses in Wnt5a mutant FRTs is that estrogen receptor expression is altered. However, we find that Esr1 is correctly expressed in the Wnt5a mutant uterine grafts grown in ovariectomized hosts and expression is activated in epithelium and the smooth muscle in both controls and Wnt5a mutant grafts upon DES exposure(Fig. 8A-F). Expression of Esr1 is also normal in mutant grafts grown in an intact cycling and harvested at different stages of the estrous (data not shown).

The progesterone receptor (Pgr) gene is a major target of estrogen signaling. In mice, Pgr is repressed by estrogen in the uterine epithelium in adult females (Kurita et al., 2000). We found that Pgr is repressed by DES in the ovariectomized adult host consistent with previously reported observations(Fig. 8K,L). In contrast to adult host uteri, Pgr is not repressed by DES in the epithelium of control grafts, rather it is repressed in the stroma and activated in the epithelium and smooth muscle layers (Fig. 8N). The differences in regulation of Pgr between the 3-week-old wild-type grafts and the adult host may reflect the previously noted differences in the estrogenic response between prepubertal females and sexually mature females (Korach and McLachlan, 1995); however, we note that Pgr gene expression and regulation in the Wnt5a mutant grafts is identical to control grafts (Fig. 8O,P). These results show that key aspects of the genomic response to estrogenic signals is preserved in the Wnt5a mutant. We then analyzed Wnt5a expression following DES exposure. Wnt5a transcripts are restricted primarily to the stroma in control grafts grown in mock-injected ovariectomized hosts (Fig. 8G), whereas Wnt5a mutant transcripts are present in both the stroma and epithelium in mutant grafts(Fig. 8I). DES exposure increases Wnt5a levels in the epithelium and the smooth muscle in both control and Wnt5a mutant grafts(Fig. 8H,J) revealing that Wnt5a signaling is not required for the regulation of its own gene product by estrogen. Therefore, in contrast to Wnt7a, which is primarily repressed by DES exposure, Wnt5a undergoes a spatial change in expression similar to the situation observed with Esr1. This result and the fact that Esr1 itself is correctly regulated in the Wnt5a mutant (Fig. 8A-F) suggest that Wnt5a regulation by estrogenic stimuli is genetically downstream of Esr1.

To determine if genes expressed in the epithelium other than Wnt7aare misregulated in the Wnt5a mutant, we analyzed the regulation of Msx1, a homeobox gene whose expression is maintained specifically in the luminal and glandular uterine epithelium of the adult(Pavlova et al., 1994). We found that DES represses Msx1 in the Wnt5a mutant grafts as in control grafts (Fig. 8Q-T)indicating that Msx1 regulation by estrogen is Wnt5aindependent in marked contrast to what is seen with other patterning genes examined in this study. Thus, Msx1 represents a potential developmental and hormone-sensitive pathway that is not subject to control by Wnt genes.

The adult FRT expresses multiple members of the Hox and homeogene families,Bmp4, and several members of the Wnt gene family(Ma et al., 1998; Miller et al., 1998b; Pavlova et al., 1994; Taylor et al., 1997; Ying and Zhao, 2000). Mouse mutants generated to the Hoxa10, Hoxa11 and Wnt7agenes reveal a requirement for these genes during postnatal development(Benson et al., 1996; Hsieh-Li et al., 1995; Miller and Sassoon, 1998; Parr and McMahon, 1998). In cases where gene mutation leads to perinatal lethality, postnatal FRT development has not been examined. Using grafting techniques as a means to circumvent the neonatal lethality of the Wnt5a mutant, we find that Wnt5a is required to appropriately establish the development of the posterior region of the FRT. Wnt5a mutant FRTs have short and coiled uterine horns of normal diameter and lack defined cervical/vaginal structures. These findings are in contrast to our previous observations for the Wnt7a mutant FRT, which shows complete posterior development whereas the uterine horns are atrophic (Miller and Sassoon, 1998) (this study). Although the Wnt5a and Wnt7a phenotypes differ, they share specific characteristics described for the FRT of different Hox gene mutants(Fig. 9A). In the Hoxa13 mutant, the caudal region of the Müllerian ducts does not develop (Warot et al., 1997). In fact other aspects of the phenotypes of posterior Hoxa gene mutants and Wnt5a mutant are similar as the growth of the genital tubercle and the limb buds is also severely affected in the double Hoxa13/d13 mutant as in the Wnt5a mutant(Warot et al., 1997). Taken together, the phenotypic similarities of Wnt5a and Hoxa13/d13 mutant FRTs suggest that they may act in a common pathway during development to regulate posterior growth of the Müllerian ducts(Fig. 9A). Whereas a phenotypic similarity between Wnt5a and posterior Hoxa mutant FRTs is found, we note that the atrophic uterine horns of the Wnt7a mutants and the reduction of the glandular and stromal compartments in the adult resemble the phenotypes described for the Hoxa11 and Hoxa10/Hoxa11 transheterozygotes mutant FRTs(Branford et al., 2000)(Fig. 9A). We reported previously that Hoxa10 and Hoxa11 expression is normal at birth in the Wnt7a mutant FRT; however, Hoxa10/11 expression is not maintained in the mature FRT(Miller and Sassoon, 1998). Similarly, the expression of Wnt7a is normal in the Hoxa11mutant neonates and subsequently declines (data not shown). Therefore Wnt7a and Hoxa11 are independently activated but then maintain expression of each other. These genetic analyses suggest that in addition to their homeotic functions along the anteroposterior axis(Benson et al., 1996; Branford et al., 2000; Hsieh-Li et al., 1995), Hoxa10 and Hoxa11 participate in a common morphogenetic pathway with Wnt7a that directs growth along the radial axis of uterine horn and subsequent stromal/epithelial differentiations required to generate the glandular compartment (Fig. 9A).

We found that Wnt5a provides a specific signal derived from stromal cells that permits the luminal epithelium to form uterine glands. Little is known regarding the developmental mechanisms that direct gland formation in the FRT (Gray et al.,2001). Analyses of the Wnt7a and Wnt5a mutants demonstrate the requirement of both genes in glandulargenesis of the uterus(Miller and Sassoon, 1998)(this study). The fact that Wnt7a is expressed in uterine epithelium and that Wnt5a is expressed in uterine stroma is consistent with longstanding observations that cytodifferentiation of the uterus requires epithelial-mesenchymal paracrine interactions. Although Wnt5a is expressed throughout the uterine mesenchyme, we observed that Wnt7ais downregulated specifically in the invaginating epithelium that gives rise to the glands during postnatal development. Based upon these observations, we propose that highly regionalized repression of Wnt7a is required to allow luminal epithelium to change fate, invaginate and form glands and that Wnt5a is required for this downregulation (see model in Fig. 9B). Although global hormonal repression of Wnt7a and local repression of Wnt7aduring pre-pubertal development may well reflect completely different pathways, we find that Wnt5a is required for downregulation of Wnt7a in response to DES.

The complete loss or transient repression of Wnt7a expression during perinatal FRT development leads to global disorganization of the uterine epithelium and a disruption of gland formation later in adult life(Miller et al., 1998a; Miller and Sassoon, 1998). This is in contrast to Wnt5a mutant FRT, which maintains a normal columnar epithelial phenotype but still fails to form glands. These observations suggest that Wnt7a is required to maintain a columnar epithelial phenotype and if downregulation of Wnt7a is blocked, gland formation will not occur as seen in the Wnt5a mutant. Alternatively,if Wnt7a expression is disrupted, epithelial cells may attempt to participate in gland formation giving rise to an abnormal multilayered epithelium that is not permissive for gland formation. Chimeric analyses in mice has shown that uterine glands are monoclonal in origin(Lipschutz et al., 1999),raising the intriguing possibility that the repression of Wnt7a may occur in a single cell that then gives rise to a gland (see model in Fig. 9B). We note that Wnt5a is expressed throughout the mesenchyme, suggesting that an additional factor may cooperate with Wnt5a to restrict glandulargenesis at specific sites of the luminal epithelium. Alternatively,it is possible that Wnt7a repression is a stochastic event that occurs in a unique cell and that Wnt5a is simply required for subsequent growth. Experiments to address these models are in progress.

Neither Wnt7a nor Wnt5a has been clearly linked to the canonical Wnt signaling pathway that requires members of the lef1/tcf1 family. In the chick limb bud, β-catenin and Lef1 retroviral infections induce morphogenetic outcomes similar to Wnt3a infection and distinct from Wnt7a overexpression(Kengaku et al., 1998). Wnt5a has been implicated in Ca²⁺ signaling and has been demonstrated to antagonize canonical Wnt signaling(Miller et al., 1999; Topol et al., 2003). Our results demonstrate that Lef1 is dispensable for uterine morphogenesis and gland formation, suggesting that canonical Wnt signaling is not required for Wnt signaling in the uterus. We note that Tcf1 is also expressed in the uterus and may rescue the lack of Lef1 in the uterus, although it does not do so in many other structures dependent upon epithelial-mesenchymal interactions previously examined. The early embryonic lethality of the double Tcf1/Lef1 mutants precludes analyses of FRT development in the double mutant (Galceran et al., 1999).

It has been demonstrated that estrogen induction of uterine epithelial proliferation is dependent upon Esr1 expression in the stroma which then signals via an unknown ligand to the epithelium(Cooke et al., 1997; Kurita et al., 2000). We have demonstrated previously that DES represses Wnt7a in the neonate FRT(Miller et al., 1998a) and it was later demonstrated that Wnt7a repression requires the expression of Esr1 in the FRT (Couse et al.,2001). We observe here that downregulation of Wnt7a and Hoxa10 and Hoxa11 genes by estrogens is abolished in absence of Wnt5a (Figs 6, 7). However, Wnt5amutant uterine grafts undergo an abnormal dilation and show an increase in epithelial thickness following DES exposure. Based on these results, we propose that there are Wnt5a dependant and independent responses to estrogenic stimulation (Fig. 9C). Factors such as Msx1 may be part of a Wnt-independent regulatory response to estrogen as shown in this study. Candidate factors that link Wnt5a to estrogenic signaling may include Wnt7a and Hoxa genes that are misregulated in the Wnt5a mutant FRT. Wnt7a and Hoxa genes are developmental factors required for normal morphogenesis of the FRT(Branford et al., 2000; Gendron et al., 1997; Hsieh-Li et al., 1995; Miller and Sassoon, 1998; Parr and McMahon, 1998; Warot et al., 1997), and are expressed throughout adult life (Benson et al., 1996; Lim et al.,1999; Ma et al.,1998; Miller et al.,1998b; Pavlova et al.,1994). Expression of Hoxa10 in the uterus is required for successful embryo implantation through the regulation of PGE2 receptors subtypes EP3 and EP4 (Benson et al.,1996; Lim et al.,1999). The expression of genes involved in Wnt signaling is modified during the implantation period(Kao et al., 2002; Paria et al., 2001; Pavlova et al., 1994). We note that Wnt7a mutant females are sterile although their ovaries are functional following transplantation into wild-type recipients(Parr and McMahon, 1998). Taken together, these data implicate uterine Wnt gene expression as crucial regulators of uterine adult function.

We found that wild-type neonate uterine grafts grown in cycling hosts show highly impaired and delayed gland formation. By contrast, neonate uterine fragments grown in ovariectomized hosts develop normally and form uterine glands. We conclude that precocious exposure to endogenous adult levels of ovarian hormones is sufficient to disrupt crucial perinatal patterning events in the FRT. Indeed, precocious exposure to DES, 17 β-estradiol, progestin or tamoxifen alter FRT morphogenesis and glandulargenesis(Branham et al., 1985a; Branham et al., 1985b; Cunha et al., 1991; Gray et al., 2001). The mechanisms underlying how hormonal teratogens permanently alter FRT development have not been completely elucidated; however, these studies support a model whereby precocious exposure to estrogens exerts a teratogenic effect upon the FRT through a perturbation of patterning gene expression in the FRT and a permanent change in gene regulation in response to hormone challenge.

We thank T. Kurita and G. Cunha (UCSF) for teaching us the renal capsule grafting technique and fruitful discussions. We thank Drs B. Kaiser (Bronx Hospital) and L. Carta (Mount Sinai Hospital) fordiscussions and technical advice. This work was supported by NIH-NIA R01 AG13784 to D.S. and J.K., and a grant awarded by NIEHS through the Superfund Basic Research Program (P42 ES07384) to D.S.