Roles of p63 in the diethylstilbestrol-induced cervicovaginal adenosis

Diethylstilbestrol (DES) is a synthetic estrogen that was prescribed to prevent miscarriage in pregnant women. It has been estimated that two to four million individuals were exposed to DES during pregnancy from 1946 to 1971(Giusti et al., 1995). Women exposed to DES in utero (DES daughters) exhibit genital tract abnormalities,including cervicovaginal adenosis, that are characterized as the development of columnar epithelium in the cervix and/or vagina(Robboy et al., 1981). DES daughters are at risk of developing cervicovaginal clear-cell adenocarcinoma(Herbst et al., 1971), and cervicovaginal adenosis is thought to be the precursor of adenocarcinoma(Robboy et al., 1981). Perinatal exposure of mice to DES generates a spectrum of reproductive tract lesions similar to those observed in humans(Forsberg, 1976; McLachlan et al., 1980; Plapinger and Bern, 1979). Using this animal model, many genes have been identified as a potential cause of DES-induced abnormalities in the female reproductive tract. For example,perinatal DES exposure disrupts expression of Wnt7a(Miller et al., 1998a), Hoxa10 and Hoxa11 (Ma et al., 1998) in the upper Müllerian duct. These genes play important roles in development and/or function of the uterus. However, the mechanism of DES-induced cervicovaginal adenosis is not understood. Estrogen receptor α (ERα) is essential for development of cervicovaginal adenosis induced by neonatal DES-exposure(Couse et al., 2001); however,the target of DES and ERα in the cervicovaginal adenosis is still unclear.

Columnar and squamous epithelia are dramatically different. The major functions of columnar epithelium are absorption and secretion, while stratified squamous epithelia form barriers. In addition, cytoskeletal and cell-adhesion molecules are different in columnar versus squamous epithelia. For example, cytokeratins 5 and 14 are expressed in squamous epithelial cells,and are essential to maintain the integrity of squamous epithelium(Chan et al., 1994; Ehrlich et al., 1995; Hutton et al., 1998; Rugg et al., 1994). It is not understood how squamous and columnar epithelia differentiate from their embryonic precursors. The female reproductive tract is an excellent model with which to study the program of epithelial differentiation because it is lined with two distinct types of epithelia that differentiate from a common precursor. The Müllerian vagina, cervix, uterus and oviduct develop from the embryonic Müllerian duct, which is composed of a uniform layer of pseudo-stratified columnar epithelial cells. In the mouse, the Müllerian duct epithelium undergoes organ-specific morphogenetic changes during postnatal development induced by uterine and vaginal mesenchyme(Cunha, 1976; Kurita et al., 2001a). In the uterus, the epithelium gives rise to columnar luminal and glandular epithelia. In the Müllerian vagina and cervix, the columnar epithelium transforms into a stratified squamous epithelium. In adulthood, columnar uterine and squamous cervicovaginal epithelia meet at the squamocolumnar junction (SCJ) in the cervix. In the mouse, epithelial cells of the Müllerian duct are fully capable of being induced by heterotypic mesenchyme to undergo uterine or vaginal differentiation prior to 7 days postnatal, after which this developmental plasticity is gradually lost. By adulthood, most uterine and cervicovaginal epithelial cells do not change their phenotype in response to induction by heterotypic mesenchyme (Cunha,1976; Kurita et al.,2001a).

Historically, uterine and vaginal epithelial phenotypes have been judged by histology. However, 17β-estradiol (E₂) and progesterone(P₄) modify epithelial morphology in the uterus and vagina, and thus the effects of ovarian steroids must be always considered. For example,uterine epithelium can stratify as a result of hyper-proliferation in response to E₂. In this case, the stratification is reversible, and does not involve expression of squamous-epithelial markers(Kurita et al., 2001a). The uterus of progesterone receptor (PR) knockout mice shows a stratified epithelial phenotype due to hyperplasia caused by unopposed estrogen action(Lydon et al., 1995), which is due to loss of PR in the stromal cells(Kurita et al., 1998). Thus,epithelial stratification (histology) per se is not the most reliable marker distinguishing uterine versus cervicovaginal epithelia. Unequivocal identification of cervicovaginal epithelial differentiation can be achieved by examination of squamous markers such as K14, which are not modified by steroid hormones (Kurita et al.,2001a). Likewise, uterine epithelial differentiation is best assessed by estrogen-independent expression of PR, which is a unique feature of rodent uterine epithelium (Kurita et al., 2000). In this study, we used multiple markers to assess differentiation of uterine and cervicovaginal epithelia.

We report the crucial role of p63 as an identity switch in differentiation of Müllerian duct epithelium. P63 (KET, p51A, p51B, p40 or p73L) is a homologue of the p53 tumor suppressor gene(Yang et al., 1998). p63^–/– mice have skin defects and lack organs arising from epidermis such as mammary and salivary glands(Mills et al., 1999; Yang et al., 1999). As development of uterus, cervix and vagina occurs mostly during postnatal stages, the phenotype of p63^–/– mice in the mature female reproductive tract is unknown because of newborn lethality. Through rescue of p63^–/– cervicovaginal rudiments by grafting, we have shown that cervicovaginal epithelium of p63^–/– mice expresses the full spectrum of uterine epithelial markers. In this study, we describe the ontogeny of p63 in the mouse female reproductive tract and demonstrate a key role for p63 in DES-induced cervicovaginal adenosis

Mice were maintained in accordance with the NIH Guide for Care and Use of Laboratory Animals, and the UCSF LARC committee approved all procedures described here. BALB/c, CD-1, C57B6 and adult female nude mice were purchased from Charles River (Wilmington, MA). p63^–/–embryos were produced by heterozygous intercrosses of p63^Brdm2 mice(Mills et al., 1999). Pregnant females were sacrificed at 16-18 days postcoitum, and embryos were harvested. Female reproductive organs from 18 p63^–/– and 32 p63⁺ (hetero p63^+/– and wild-type p63^+/+) embryos were used. p63^–/– mice were identified visually as limbless. The genotypes of embryos were confirmed by PCR. Uterine and vaginal rudiments were divided into four parts [uterus, lower cervix and upper Müllerian vagina (cervicovaginal), lower Müllerian vagina and sinus vagina] and were grafted under the kidney capsule of athymic mice. Hosts were ovariectomized 4 weeks after grafting. Two weeks after the ovariectomy, hosts were injected with 125 ng E₂ in 100 μl corn oil(tocopherol-stripped, ICN Biomedicals, Aurora, OH) or 100 μl corn oil alone daily for 3 days before termination. Silastic capsules filled with 25 mg DES(Wang et al., 2001) were implanted into some hosts bearing p63^–/– and p63⁺ uterine grafts at the time of grafting. Two weeks later, the grafts were harvested for analysis.

ERα^–/– mice on a C57BL6J/129Svj mixed genetic background were produced and genotyped as described previously(Lubahn et al., 1993).

For the UtE/VgM recombination experiment with ERα^–/– mice, hosts were ovariectomized at the time of grafting and 25 mg DES Silastic-capsules were implanted into half of the hosts. The capsules were removed 2 weeks after the grafting, and tissues were harvested 2 weeks after removal of the DES-capsule.

The detailed protocol for tissue recombination has been described previously (Kurita et al.,2001a). Results were based upon analysis of 8-11 tissue recombinants/group from at least three independent experiments.

The methods for epithelial separation and primary culture have been described previously (Kurita et al.,2000). Briefly, epithelial cells were embedded in a 1:1 mixture of growth factor reduced Matrigel (BD Bioscience, Franklin Lakes, NJ) and rat tail collagen, and cultured in a 1:1 mixture of DME and Ham's F-12 media(Gibco, Gaithersburg, NY) with transferrin (5 μg/ml) (Sigma) and insulin(10 μg/ml) (Gibco).

Methods for IHC have been described(Kurita et al., 1998). Mouse monoclonal antibodies were used at the following concentrations:anti-ERα 1D5 (1:50), anti-cytokeratin 10 (1:25, DAKO, Carpenteria, CA),anti-p63 4A4 (1:100, Santa Cruz Biotechnology, Santa Cruz, CA), anti-p63 Ab2(1:100, Lab Vision, Fremont, CA), anti-Ki67 (1:100, Novacastra Laboratories,Burlingame, CA), anti-K8 LE41 (1:5) and anti-K14 LE001 (1:5, gift from Dr E. B. Lane, University of Dundee, Dundee, UK). Rabbit and goat polyclonal antibodies were used at the following concentrations: anti-P-cadherin goat(1:100) and anti-p130 rabbit (1:200) (Santa Cruz Biotechnology), anti-PR rabbit (1:100, DAKO) and anti-involucrin rabbit (1:2000, Covance, Princeton,NJ). Anti-K19 rabbit monoclonal antibody (1:1) was obtained from Dr Robert Pytela, UCSF, San Francisco, CA. Positive signals were visualized as brown precipitates utilizing 3,3′-diaminobenzidine tetra-hydrochloride(Sigma).

To determine the p63-positive epithelial index, images of p63-IHC were captured with a DC330 camera (Dage-MTI, Michigan City, IN), interfaced with a computer. Lengths of basement membrane were measured on the images with Scion Image 1.62a (Scion, Frederick, MD), and the length of basement membrane associating with p63-positive cells against the length of total basement membrane was calculated. Approximately 2000-14,000 μm of basement membrane was analyzed for each tissue recombinant.

Neonatal CD-1, BALB/c and C57B6 mice and neonatal mice from ERα^+/– breeding cages were treated with DES by the following protocol previously described(McLachlan et al., 1980). Briefly, newborn mice were injected subcutaneously with 2 μg DES in 20μl corn oil or 20 μl corn oil from day 1. The neonatal mice received DES or oil every 24 hours for 5 or 7 days. mice were sacrificed at P1, P2, P3, P4,P5, P7, P10, P14, P21, P35 and P60. mice to be harvested at P60 were ovariectomized at P35 and treated with 125 ng E₂ in 100 μl corn oil or 100 μl corn oil alone everyday for 3 days before termination.

P63 was undetectable in the Müllerian duct on embryonic day 16 (E16)(not shown). At E18, the sinus vagina (SVG) consisted of a solid epithelial cord that was uniformly positive for p63 and K14(Fig. 1A-C). By contrast, p63 was only detected in a small subset of columnar epithelial cells in the Müllerian vagina (MVG) (Fig. 1A,C,D) and the cervix (Fig. 1A). Other vaginal epithelial differentiation markers including K14 (Fig. 1B) were not detectable in MVG at this stage (Kurita et al., 2001a). In the MVG, p63-positive cells were consistently detected at the junction between SVG and MVG(Fig. 1C).

During postnatal development, the number of p63-positive cells gradually increased in cervicovaginal epithelium (CVE) as the p63-positive epithelial layer extended cranially from the MVG into cervix. On postnatal day 1 (P1),p63 was strongly detected in a substantial portion of Müllerian-vaginal and cervical epithelial cells (Fig. 1E), while K14 (Fig. 1F) and other markers of CVE (K5, K19 and p-cadherin) were still undetectable (data not shown). By P2, some p63-positive epithelial cells became squamous (with a ∼90° change in nuclear polarity) and formed a basal layer in the cervix and MVG (Fig. 1G). Some of these p63-positive basal cells were weakly positive for K14 (Fig. 1H). However,many p63-positive epithelial cells in the cervix and MVG were columnar and remained in the luminal position (Fig. 1G, black arrows). These observations suggest that in Müllerian duct epithelium, expression of p63 proceeds squamous differentiation. As p63 was already detected in cervicovaginal epithelium at E18, there was at least a three-day gap between expression of p63 and K14. By P5, a p63-positive basal epithelial layer was fully formed and continuous throughout MVG and cervix (not illustrated). By P14, the SCJ was formed, and differences in expression of p63 (Fig. 1I), K14 (Fig. 1J)and PR (Fig. 1K) were clearly visible in CVE versus uterine epithelium (UtE).

Although E₂ and/or P₄ regulate histodifferentiation and functional cytodifferentiation of UtE and CVE in adulthood, E₂and/or P₄ did not modify the adult expression pattern of p63 in the female reproductive tract. Therefore, p63 was always expressed in CVE but not in UtE (Fig. 1L-N). E₂-treatment stimulated proliferation and caused stratification of UtE (Fig. 1N, blue arrows), but the uterine epithelial cells remained negative for p63 and K14 (not shown),which were always expressed in adult CVE.

Tissue recombinants were constructed with newborn uterine (UtM) or Müllerian-vaginal mesenchyme (VgM) plus UtE or Müllerian-vaginal epithelium (VgE) derived from either P1 neonatal or adult BALB/c mice. When VgE from neonatal mice was recombined with UtM (VgE+UtM), the epithelial tissue developed a columnar uterine phenotype as previously reported(Cunha, 1976; Kurita et al., 2001a)(Fig. 2A). Even though a substantial portion of Müllerian-vaginal epithelial cells expressed p63 at the time of tissue recombination (Fig. 1E), the entire epithelium became negative for p63 in VgE+UtM tissue recombinants after 1 month of growth(Fig. 2A). Thus, UtM turned-off p63 expression in VgE. When neonatal-UtE was recombined with VgM (UtE+VgM), a squamous vaginal epithelium developed that was positive for p63(Fig. 2B). Homotypic recombinants (UtE+UtM and VgE+VgM) behaved as expected(Fig. 2C,D). Thus, based upon p63 ontogeny and tissue recombination studies, p63 expression is induced in Müllerian duct epithelial cells by VgM during embryonic-neonatal development.

In contrast to newborn UtE and VgE, p63 expression was not changed when adult UtE and VgE were grown in association with heterotypic mesenchyme. When adult-VgE was recombined with UtM, the adult-VgE maintained its squamous phenotype, and p63 was still highly expressed in the basal layer after 1 month of in vivo growth (Fig. 2E). Likewise, adult-UtE recombined with VgM retained the original p63-negative columnar uterine phenotype (Fig. 2F). Therefore, most epithelial cells of the uterus and vagina lose their developmental plasticity by adulthood as reported previously(Cunha, 1976; Kurita et al., 2001a).

The rudiments of uterus, cervix and Müllerian vagina were rescued from p63^–/–, wild-type (p63^+/+)and heterozygous (p63^+/–) embryos (E16-18) by grafting. Uterine and cervicovaginal phenotypes of p63^+/+and p63^+/– mice were indistinguishable by all criteria examined. Thus, both p63^+/+ and p63^+/– progeny were referred as p63⁺. In uterine grafts of p63^–/– mice, the epithelium appeared completely normal in morphology and expression of differentiation markers(Table 1). In cervicovaginal and Müllerian-vaginal grafts from p63^–/–mice, the epithelium failed to differentiate into squamous and instead formed a uterine-like columnar epithelium (Fig. 3). Moreover, in the p63^–/–cervicovaginal grafts deep epithelial invaginations or glands were always observed(Fig. 3A,B, black arrows). The glandular epithelium in the p63^–/–cervicovaginal grafts became hyperplastic and multilayered in response to E₂-treatment(Fig. 3B). However, markers for squamous differentiation (Fig. 3C-F) or keratinization (Fig. 3G-J) were never detected in p63^–/– CVE(Table I). Although negative for cervicovaginal differentiation markers, p63^–/– CVE expressed markers common to both uterine and cervicovaginal epithelia such as ERα (Esr1 – Mouse Genome Informatics) and p130 (Fig. 3K-N, Table 1). Most importantly, p63^–/– CVE expressed a high level of PR in the absence of E₂ [Fig. 3, compare p63⁺ (O) and p63^–/– (P)], which is a feature unique to rodent UtE. Unlike normal UtE, which downregulates PR in response to E₂, the epithelium of p63^–/–cervicovaginal grafts remained strongly positive for PR when treated with E₂ (Fig. 3R). However, this pattern of PR expression is actually identical to that of adult-UtE + VgM tissue recombinants (Table 1). It is known that PR in UtE is downregulated by an action of uterine stroma induced by E₂(Kurita et al., 2000), and the stromal activity to downregulate PR in UtE is specific for uterine stroma. Therefore, when adult-UtE is recombined with VgM, the adult-UtE retained the original uterine phenotype and expressed high levels of PR in the absence of E₂, but the UtE also strongly expressed PR in the presence of E₂ (Kurita et al.,2001a). In conclusion, the profile of differentiation markers for p63^–/– CVE was identical to that of normal UtE for the all criteria examined (Table 1). Hence, in the absence of p63, Müllerian CVE differentiated into UtE.

To determine if the p63^–/– cervicovaginal mesenchyme (CVM) was able to induce cervicovaginal epithelial differentiation in Müllerian duct epithelial cells, E17.5 UtE from p63^–/– and p63⁺ embryos were recombined with p63^–/– and p63⁺ VgM. When p63⁺ UtE was recombined with p63^–/– VgM(Fig. 3S), p63 expression and cervicovaginal differentiation was induced by the p63^–/– VgM in the p63⁺ UtE. By contrast, p63^–/– UtE could not be induced to become CVE by p63⁺ VgM(Fig. 3T). The epithelial marker expression profile in the p63^–/– UtE +p63⁺ VgM tissue recombinants was identical to p63^–/– cervicovaginal grafts (not shown). Thus, the uterine phenotype of p63^–/– CVE was due to the absence of p63 in the epithelial cells even though p63^–/– CVM has ability to induce normal cervicovaginal epithelial differentiation.

As CVE differentiated into UtE in the absence of p63, it appeared that the uterine phenotype might be the default differentiation endpoint for embryonic Müllerian duct epithelium, suggesting that induction by UtM may not be required for differentiation of UtE. Whether embryonic UtE requires instructive induction from UtM to establish the proper uterine marker expression pattern was tested in vitro and in vivo.

At E18, UtE did not express PR (Fig. 4A) or ERα (Fig. 4B). When undifferentiated UtE from E18 embryos (E18-UtE) was cultured in an extracellular matrix for 7 days, only a small subset of the cultured epithelial cells expressed PR(Fig. 4E), which was expressed in the entire UtE in situ by P5 (Fig. 4C). In addition, whereas ERα was undetectable in UtE in situ at P5 (Fig. 4D), ERαwas expressed in a substantial proportion of cultured E18-UtE cultured to the age equivalent of P5 (Fig. 4F). The comparison of PR and ERα expression in vivo versus in vitro demonstrates that the gene expression pattern of developing UtE is strongly affected by presence or absence of UtM (compare Fig. 4C,D with E,F). By contrast, when adult (2-month-old) UtE were cultured under the same conditions as the E18-UtE, the adult-UtE retained its original expression pattern of PR and ERα (Fig. 5G,H). Thus, the absence of PR expression in the cultured E18-UtE was not due to the culture conditions.

The role of UtM in the differentiation of UtE was also examined in vivo by tissue recombination experiments. E18-UtE was recombined with bladder mesenchyme (BLM) or UtM, and the UtE + UtM and UtE + BLM tissue recombinants were grafted onto contralateral kidneys of the same hosts. The grafts were grown for a month, and then marker expression was analyzed. The UtE recombined with UtM expressed both PR and ERα as expected(Fig. 4I,J). By contrast, UtE recombined with BLM in the same host failed to express PR(Fig. 4K). Thus, UtM appears to play an essential role in inducing proper uterine epithelial differentiation in Müllerian duct epithelium. As grafts of p63^–/– CVE expressed the full and correct set of uterine differentiation markers, mesenchymal cells within the Müllerian-vaginal, cervical and uterine domains appear to have uterine inductive activity.

Cervicovaginal adenosis consists of focal areas of columnar surface epithelium or simple columnar submucosal glands in the cervix or vagina, which are normally lined with squamous epithelium(Robboy et al., 1986). In the cervix and Müllerian vagina of neonatally DES-treated mice cervicovaginal adenosis appeared as deep-inclusions or glands with openings into the lumen of cervicovaginal canal (Fig. 5A-D), or simple columnar surface epithelium within the vaginal or cervical canal (Fig. 5E-G). In DES-induced cervicovaginal adenosis, squamous differentiation markers were not expressed (Table 1). Instead,the columnar adenosis lesions expressed high levels of PR in the absence of E₂ (Fig. 5C,G) and other markers indicative of UtE. Indeed, cervicovaginal adenosis expressed the full spectrum of uterine markers (Table I).

In DES-induced uterine SQM of the mouse, the metaplastic foci(Fig. 5H-J, red arrows)expressed p63, K14 (Fig. 5H,I),K19, P-cadherin and involucrin (Table 1). In the metaplastic squamous epithelium, PR was expressed in the presence of E₂ (Table 1), but was undetectable in the absence of E₂(Fig. 4J, red arrows). Taken together, all of these markers of uterine SQM are identical to that of CVE. In both cervicovaginal adenosis and uterine squamous lesions, the epithelium expressed both ERα (Fig. 5D) and p130 (Table 1). Thus, based upon the morphology and the expression of all markers tested (Table 1),cervicovaginal adenosis is not persistent undifferentiated Müllerian duct epithelium, but the development of UtE in the cervix and vagina. Likewise,uterine SQM appears to be the development of CVE in the uterus. Although uterine SQM was always associated with expression of p63, whether p63 was essential for the development of uterine SQM was not clear. To answer the question, uterus from E17.5 p63^–/– and p63⁺embryos were grafted into untreated or DES-treated nude mice. Expression of squamous differentiation markers was examined 2 weeks after the grafting. In the absence of DES, no epithelial cells expressed squamous markers in the p63^–/– and p63⁺ uterine grafts. With DES-treatment, p63- (Fig. 5K) and K14- (Fig. 5L)-positive basal epithelial cells were detected in four out of 12 p63⁺ uterine grafts, but these markers were not detected in 16 p63^–/– uterine grafts(Fig. 5M). In the presence of DES, p63^–/– uterine grafts showed stratification of epithelium (Fig. 5M, black arrow), but the stratified epithelial cells were negative for K14 and involucrin (not shown). Therefore, p63 is not just a marker but is essential for development of uterine SQM.

The ontogeny of p63 expression was examined during treatment of neonatal mice with DES. In 5-day-old oil-treated mice (control), a continuous layer of p63-positive basal epithelial cells extended from the Müllerian vagina into the cervical canal (Fig. 6A). However, when neonatal wild-type mice were treated with DES from P1 through P5, induction of p63 was greatly inhibited in the upper part of Müllerian vagina and the cervix. At P5, this was manifest as large gaps in the p63-positive basal epithelial layer in the fornix and common cervical canal (Fig. 6B, green arrows), which are the primary site for development of cervicovaginal adenosis(Forsberg and Kalland, 1981). Two days after the last DES-injection (P7)(Fig. 6D), most of the gaps in the p63-positive layer in the CVE were filled-in with p63-positive cells. However, small patches of p63-negative epithelial cells in the fornix and common cervical canal (Fig. 6D,F, green arrows) persisted into adulthood as adenosis(Fig. 6G).

When DES-treatment was extended from birth to P7, p63 expression in CVE did not recover (compare Fig. 6D,E), and many large patches of p63-negative columnar cells were observed in CVE (Fig. 6E, green arrows). This result suggests that a continuous high level of DES is required to fully block induction of p63 expression in CVE. Presumably, discontinuation of DES-treatment at P5 allows mesenchymal and epithelial tissues to interact and induce p63 in Müllerian duct epithelium.

ERα^+/+, ERα^+/–and ERα^–/– mice were treated with DES from P1 to P5. Confirming the results described above, DES disrupted p63 expression in CVE in ERα^+/+ and ERα^+/– mice. By contrast, the p63-positive basal layer of cervicovaginal epithelial cells developed normally in DES-treated ERα^–/– mice(Fig. 6C), which was identical to that of the oil-treated ERα^+/+ control(Fig. 6A). Therefore, DES action on p63 in developing CVE requires signaling through ERα.

In the cervix and Müllerian vagina, ERα is highly expressed in both epithelial and mesenchymal cells by P3(Kurita et al., 2001a). Therefore, DES action may be mediated by ERα in either epithelial and/or mesenchymal cells. To test whether DES disrupts induction of p63 via ERαin epithelial or mesenchymal cells, the four types of tissue recombinants were constructed with UtE and VgM from ERα^+/+ and ERα^–/– mice(ERα^+/+UtE+ERα^+/+ VgM, ERα^+/+UtE+ERα^–/– VgM, ERα^–/–UtE+ERα^+/+ VgM and ERα^–/–UtE+ERα^–/– VgM) and grafted into ovariectomized female nude mice. In untreated hosts, a p63-positive squamous basal epithelial layer formed in all four types of UtE+VgM tissue recombinants(Fig. 7A, parts a and b, Fig. 7B) indicative of normal cervicovaginal epithelial differentiation. The basal cells were also strongly positive for K14 (not shown). When hosts were treated with DES, squamous basal cells were not detected in the UtE+VgM tissue recombinants prepared with ERα^+/+ UtE (ERα^+/+UtE+ERα^+/+ VgM and ERα^+/+UtE+ERα^–/– VgM) as assessed by histology and K14 expression (not shown), indicating that DES had inhibited cervicovaginal differentiation (Fig. 7A,parts c and e). By contrast, when ERα^–/– UtE were used to construct the tissue recombinants (ERα^–/–UtE+ERα^+/+ VgM and ERα^–/–UtE+ERα^–/– VgM tissue recombinants), a p63-positive squamous basal epithelial layer was induced even when the hosts were treated with DES (Fig. 7A, parts d and f, red arrows). These results clearly demonstrate that DES acts via ERα in the epithelial cells to inhibit induction of p63 in CVE. DES action via mesenchymal ERα does not inhibit p63 expression and squamous differentiation of CVE.

The endocrine disruptor, DES, perturbs morphogenesis and epithelial differentiation in the mouse and human female reproductive tract. Accordingly,DES elicits cervicovaginal adenosis, which involves permanent changes in epithelial cell fate. This paper integrates development of DES-induced epithelial abnormalities into the context of normal differentiation of the Müllerian duct. P63 is proposed to be an essential epithelial identity switch, which allows Müllerian duct epithelial cells to undergo cervicovaginal squamous differentiation. Based upon the data presented here,we propose the following models of epithelial differentiation in the Müllerian duct (Fig. 8).

Epithelial cells in the Müllerian duct are initially p63-negative and undifferentiated (Fig. 8A,stage 1). When induced by CVM, p63 is expressed in the Müllerian duct epithelial cells. This step is essential for subsequent squamous differentiation. If p63 is not expressed, the Müllerian duct epithelial cells differentiate into UtE. When p63 is initially expressed in Müllerian duct epithelial cells, these cells remain columnar and undifferentiated for about 3 days (Fig. 8A, stage 2). At this stage, epithelial cell fate can be altered by heterotypic mesenchymal induction. The expression pattern of p63 is subsequently stabilized by mesenchymal induction, and the squamous cervicovaginal and columnar uterine cell fate gradually becomes manifest and irreversible by adulthood (Fig. 8A, stage).

This process seems to be a general principle among many organs. Besides female reproductive tract, we have studied the ontogeny of p63 in other developing mouse organs. For all organs examined, expression of p63 precedes the expression of squamous markers and morphological changes (90° nuclear polarity change) by several days. Squamous differentiation appears to consist of at least three steps: induction of p63, stabilization of p63 and subsequent squamous differentiation. As proposed above, expression of p63 itself does not transform columnar epithelium into squamous immediately, but p63 expression appears to be an essential step for squamous differentiation in general,because epithelial cells in the forestomach and seminal vesicle of p63^–/– mice also failed to undergo squamous differentiation (T.K., unpublished).

In embryo, the Müllerian duct is composed of uniform, undifferentiated columnar epithelial cells (i.e. stage 1 in Fig. 8A). UtM and CVM already have acquired their organ-specific inductive identities. Therefore, squamous epithelial differentiation inducing signals (red arrows) are expressed only in the CVM, whereas common uterine epithelial differentiation inducing signals(blue arrows) are expressed throughout the cervicovaginal and uterine regions of the Müllerian duct. In response to induction by CVM (red arrows),epithelial cells in the cervicovaginal area express p63 (P1-5). Subsequently,the p63-positive epithelial cells differentiate into squamous CVE as instructed by CVM. Because the signals inducing squamous differentiation are expressed only in CVM, epithelium in the uterus never expresses p63 and differentiates into columnar UtE, which requires mesenchymal inductive activity (blue arrows). By P5, the adult expression pattern of p63 is established in CVE and UtE, but the status of p63 and thus the developmental fate of the epithelial cells can still be altered by heterotypic mesenchyme(stage 2 in Fig. 8A). By P7, a substantial percentage of uterine and vaginal epithelial cells are insensitive to heterotypic uterine or cervicovaginal mesenchymal induction(Cunha, 1976), and the uterine and cervicovaginal epithelial phenotypes become irreversibly determined in the first few weeks of postnatal development. Accordingly, in adulthood almost all uterine and cervicovaginal epithelial cells cannot be re-programmed by uterine or vaginal mesenchyme to express an alternative epithelial phenotype (stage 3 in Fig. 8A).

In the p63^–/– mice, mesenchymal activity to induce uterine (blue arrows) and cervicovaginal (red) epithelial differentiation is normal. However, as the Müllerian duct epithelial cells cannot express p63, the entire Müllerian duct epithelium differentiates into UtE both in the uterine and cervicovaginal anlagen.

DES binds to ERα in the Müllerian duct epithelial cells and blocks expression of p63, which is a prerequisite for subsequent squamous differentiation. As a result, p63 is induced in only a subpopulation of cervicovaginal epithelial cells, even though the mesenchymal signals to induce squamous epithelial differentiation (red arrow) are not affected by DES. Like most developmental processes, the timing of mesenchymal specification of uterine and cervicovaginal epithelial differentiation is tightly regulated. When DES treatment is stopped at P5, most epithelial cells in cervix and vagina respond to the mesenchymal signals (red arrows) and express p63 by P7. However, at least some Müllerian duct epithelial cells have lost sensitivity to the mesenchymal signals, as epithelial cell fate is already irreversibly fixed in some epithelial cells by this stage. These p63-negative epithelial cells in cervix and vagina have missed the time window to express p63, and thus, remain simple columnar and differentiate into UtE, which persists into adulthood as cervicovaginal adenosis. The tissue recombinant studies using ERα^+/+ and ERα^–/– UtE and VgM demonstrate that DES inhibits expression of p63 in Müllerian duct epithelial cells through epithelial ERα. Because DES elicits abnormal epithelial differentiation via epithelial ERα, epithelial (and not mesenchymal/stromal) regulatory molecules are clearly the immediate targets of DES action in inducing adenosis. However, our model does not exclude a subordinate role of mesenchymal genes such as homeobox genes in the development of adenosis.

The correct and full expression of uterine epithelial markers in the p63^–/– CVE implies the expression of uterine inductivity in the CVM as well as in the UtM. However, the molecular signals comprising uterine inductive activity may not be identical in CVM versus UtM. Many developmental regulatory genes have been detected only in the uterine regions of Müllerian duct, e.g. Wnt and Hox genes(Ma et al., 1998; Miller et al., 1998b; Pavlova et al., 1994). Our data do not exclude involvement of these molecules in normal uterine epithelial development. However, it is unlikely that a single molecule regulates all aspects of uterine epithelial differentiation, and instead a balance of several genes may be important for induction of normal UtE. CVM appears to express a reasonably complete combination of signals for inducing uterine epithelial differentiation based upon analysis of p63^–/– cervix and Müllerian vagina. The squamous differentiation inducing activity in CVM is also likely to be a combination of multiple factors.

Wnt7a has been suggested as an immediate target of DES in development of vaginal adenosis because Wnt7a was downregulated by DES in the Müllerian duct, and the Wnt7a^–/– mouse developed`vaginal adenosis' (Miller et al.,1998a). However, in the paper by Miller et al., the term`adenosis' was used to describe `epithelial inclusions'. The glands in the Wnt7a^–/– mouse appeared to be lined by keratinized squamous epithelium, which is not strictly adenosis. Such glands seem to be absent in younger Wnt7a^–/– mice(more than 4 months old), which is also not in agreement with the phenotype of neonatally DES-treated mice. Therefore, it is not clear if downregulation of Wnt7a plays a role in development of columnar epithelium in cervix/vagina. Wnt7a may be upstream of p63, or the glands observed in Wnt7a^–/– mice may have possibly developed via a different mechanism from DES-induced adenosis.

In DES-daughters, cervical/vaginal clear cell adenocarcinoma is rare, even though cervical/vaginal adenosis is commonly found. Although adenosis is thought to be the substrate from which clear cell adenocarcinoma develops, the mechanism by which adenosis develops into adenocarcinoma is unclear. We have shown that DES actions on CVM do not play a role in formation of adenosis. However, growth, cell death and differentiation of epithelial tissue in the female reproductive tract are regulated by stromal cells during embryogenesis as well as adult period (Buchanan et al.,1998; Cooke et al.,1997; Kurita et al.,2001b; Kurita et al.,1998). Thus, it is likely that the DES-caused changes in the stromal cells also play important roles in the subsequent development of cervicovaginal adenocarcinoma.

We thank Dr Dennis B. Lubahn for the ERα knockout mice, Dr E. Birgit Lane for the anti-K14 LE001 and anti-K8 LE041 mouse monoclonal IgGs. This work was supported by NIH Grant AG 13784, AG 15500, DK52707 and DK 47517.