|Home | About | Journals | Submit | Contact Us | Français|
Follicle-stimulating hormone (FSH) regulation of aromatase gene expression in vitro requires the transcriptional coactivator beta-catenin. To ascertain the physiological significance of beta-catenin in granulosa cells during folliculogenesis, mice homozygous for floxed alleles of beta-catenin were intercrossed with Amhr2cre mice. Conditional deletion of beta-catenin in 8-wk-old females occurred in derivatives of the Müllerian duct, granulosa cells and, surprisingly, in brain, pituitary, heart, liver, and tail. Female mice deficient for beta-catenin were infertile, despite reaching puberty and ovulating at the expected age, indications of apparently normal ovarian function. In contrast, their oviducts were grossly distended, with fewer but healthy oocytes. In addition, their uteri lacked implantation sites. Together, these two phenotypes could explain the complete loss of fertility. Nevertheless, although the ovary appeared normal, with serum estradiol concentrations in the normal range, there was marked animal-to-animal variation of mRNAs encoding beta-catenin and aromatase. Similarly, inhibin-alpha and luteinizing hormone receptor mRNAs varied considerably in whole ovaries, whereas pituitary Fshb mRNA was significantly reduced. Collectively, these features suggested cyclization recombination (CRE)-mediated recombination of beta-catenin may be unstable in proliferating granulosa cells, and therefore may mask the suspected steroidogenic requirement for beta-catenin. We tested this possibility by transducing primary cultures of granulosa cells from mice homozygous for floxed alleles of beta-catenin with a CRE-expressing adenovirus. Reduction of beta-catenin significantly compromised FSH stimulation of aromatase mRNA and subsequent production of estradiol. Collectively, these data suggest that FSH regulation of steroidogenesis requires beta-catenin, a role that remains hidden when tested through Amhr2cre-mediated recombination in vivo.
Estrogen contributes to folliculogenesis, development of the female reproductive tract, female sexual behavior, lipid metabolism, bone remodeling, and development of several cancers . Ovarian granulosa cells are the major source of circulating estrogen. Biosynthesis of estrogen is regulated by the conversion of thecal-derived C19 androgens to C18 estrogens by the enzyme aromatase (officially designated CYP19A1). The expression of aromatase in granulosa cells is induced upon binding of follicle-stimulating hormone (FSH) to its guanine nucleotide-binding protein-coupled receptor (GPCR) and subsequent activation of the cAMP/protein kinase A (PKA) pathway [2, 3]. Transcriptional activation of the rat aromatase gene requires the PKA substrate CREB, the orphan nuclear receptor SF1 (officially designated NR5A1), and GATA4 [4, 5]. Production of the androgen substrate by theca interna cells is regulated by luteinizing hormone (LH) via G protein-coupled LH receptors [6, 7]. Thus, the extent of estrogen production by a follicle depends on both the regulated expression of the aromatase gene by FSH and the availability of LH-regulated androgen substrate .
Recently, we reported that β-catenin (officially designated CTNNB1), a transcriptional coactivator, is required for FSH-regulated expression of aromatase in a human granulosa tumor cell line (KGN) and in primary cultures of rat granulosa cells . Beta-catenin is the intracellular mediator of the canonical WNT pathway that signals though Frizzled family receptors and low-density lipoprotein receptor-related protein (LRP) coreceptors [10–14]. WNT proteins participate in embryogenesis, cell polarity, specification of cell fate, and tumorigenesis [15, 16]. Growing evidence indicates that the WNT family of signaling molecules also plays critical roles in regulating ovarian function. WNT4 is required for early ovarian development; mice null for Wnt4 have sex-reversed ovaries that express genes associated with testis development and a reduced number of oocytes at birth . In contrast, mice with granulosa cell-specific alleles encoding constitutively active β-catenin are subfertile and develop ovarian lesions that progress to granulosa cell tumors .
Hormonal regulation of the Wnt family of genes also has been detected in rodent ovaries. Wnt4 expression is elevated in response to the LH receptor agonist human chorionic gonadotropin (hCG), and Wnt4 is highly expressed in corpora lutea . In addition, expressions of Wnt4 and secreted frizzled-related protein 4 are low in genetically modified mice (Tg(Cga-LHB/CGB)94Jhn) that hypersecrete a chimeric form of LH . Collectively, these studies suggest that normal ovarian function requires regulated contributions from WNTs and β-catenin.
Beta-catenin acts in the canonical WNT signaling pathway as a downstream coactivator that restores transcription to genes normally repressed by complexes that contain Groucho-related proteins, such as TLE1, histone deacetylases, and other corepressors [12–14]. WNTs transduce their signals by binding to Frizzled receptors in cooperation with LRP coreceptors to activate signaling cascades that inhibit the phosphorylation and degradation of β-catenin. The resulting accumulation of β-catenin promotes its association with a variety of transcription factors, including both the T-cell factor/lymphoid enhancer-binding protein (TCF/LEF) and nuclear receptor family of transcriptional activators [21–24].
Although our previous study indicated that β-catenin upon interaction with SF1 is required for FSH-regulated expression of aromatase, these studies relied largely on RNAi-mediated reduction of β-catenin as well as overexpression of SF1 and a constitutively active form of β-catenin in cell lines and primary cultures of granulosa cells . To further address the physiological significance of β-catenin in the ovary, we used the Cre/loxP system of recombination to permit tissue-specific deletion of β-catenin alleles . Such an approach was mandated because global deletion of Ctnnb1 genes results in early embryonic death due to a defect in the ectoderm cell layer at gastrulation . Here, we report that Amhr2cre-mediated recombination of β-catenin occurs in an unexpected range of tissues, causing infertility overtly through impaired development and patterning of the oviduct and uterus without apparent alteration of ovarian function. However, in vitro studies with primary cultures of granulosa cells bearing floxed alleles of β-catenin unmask a requirement for the coactivator with respect to FSH regulation of steroidogenesis, a role that remains hidden when tested through Amhr2cre-mediated recombination in vivo.
All procedures were approved by the Institutional Animal Care and Use Committee at Washington State University. Amhr2-Cre (Amhr2tm3(cre)Bhr) transgenic mice were provided by Dr. Richard Behringer (University of Texas M.D. Anderson Cancer Center, Houston, Texas) and have been described previously . In a first cross, Amhr2cre mice were mated with mice homozygous for loxP sites flanking exons 2 and 6 of the Ctnnb1 gene (Ctnnb1tm2Kem or Ctnnb1flox/flox; The Jackson Laboratory, Bar Harbor, ME). The offspring inheriting both Amhr2cre and a floxed deleted allele then were mated with homozygous Ctnnb1flox/flox mice to obtain control mice lacking the Cre transgene and mice conditionally deleted for β-catenin in both alleles in tissues expressing Amhr2cre.
For identification of Ctnnb1 alleles and the Amhr2cre transgene, DNA from 2-mm tail biopsies of 21-day-old mice was collected. Tail samples were lysed in a buffer containing 1 M NaOH and 0.5 M EDTA at 75°C for 3 h; 1 μl of genomic DNA was used for PCR. DNA extraction and genotyping of additional tissues were done as described for tails, with the addition of a proteinase K incubation followed by phenol chloroform extraction. DNA isolation from granulosa cells collected into Trizol was performed as directed by the manufacturer's protocol. The Cre gene was detected using the following primers to generate a 200-bp product: forward, 5′-GCA CTG ATT TCG ACC AGG TT-3′; and reverse, 5′-GCT AAC CAG CGT TTT CGT TC-3′. Ctnnb1 floxed, wild-type, and floxed deleted alleles were detected using a combination of RM41, RM42, and RM43 primer sets, whereas generation of a 631-bp product identified the floxed deleted allele using primer sets RM68 and RM69, as described previously .
Superovulation was carried out in mutant and control female mice. Mice received a single intraperitoneal injection of 5 IU of eCG per mouse (Sigma-Aldrich, St. Louis, MO) for 48 h, followed by 5 IU of hCG per mouse (Sigma-Aldrich) for an additional 22 h.
Oocytes at the germinal vesicle stage were liberated from antral follicles using 26-gauge needles, placed in 10-μl drops of Weymouth medium (Gibco BRL, Gaithersburg, MD) supplemented with 10% fetal bovine serum and 0.23 mM sodium pyruvate overlaid with Squibb mineral oil, and incubated at 37°C, 5% CO2. To obtain oocytes arrested at metaphase II, oocytes were maintained in culture for 16 h, and those exhibiting polar body extrusion were fixed for analysis. Oocytes then were embedded in a fibrin clot and attached to a microscope slide before being subjected to immunofluorescence staining as described previously [28, 29].
The 8-wk-old female Ctnnb1 wild-type (n = 4) and mutant (n = 3) mice were subjected to a continuous mating study. Female mice were housed with a proven male stud for 6 mo. Mice were monitored daily for identification of a vaginal plug, and the subsequent number of pups and litters was recorded. For implantation studies, female mice were placed with male mice and checked for the presence of a vaginal plug the following morning. Uteri were collected 6 days after the identification of a plug to visualize implantation sites.
Ovaries were fixed overnight at 4°C in 4% PBS-buffered paraformaldehyde. Paraffin-embedded sections were sectioned at 6 μm and stained with hematoxylin and eosin.
RNA was isolated from tissues using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Total RNA (150 ng) was reverse transcribed by oligo(dT) primers and Superscript II Reverse Transcriptase (Invitrogen). For specific primers, the sequences are as follows: cyclophilin B (Ppib), forward, 5′-CAA AGA CAC CAA TGG CTC ACA G-3′ and reverse, 5′-CCA CAT CCA TGC CCT CTA GAA C-3′; β-catenin (Ctnnb1), forward, 5′-CGC AAG AGC AAG TAG CTG ATA TTG-3′ and reverse, 5′-CGG ACC CTC TGA GCC CTA GT-3′; FSH beta (Fshb), forward, 5′-CAA TAC CCA GAA AGT ATG TAC CTT CAA G-3′ and reverse, 5′-GGC ACA GCC AGG CAA TCT T-3′; LH receptor (Lhr), forward, 5′-CCT TGT GGG TGT CAG CAG TTA C-3′ and reverse, 5′-TTG TGA CAG AGT GGA TTC CAC AT-3′; inhibin-α (Inha), forward, 5′-GCC CAG GAG GCT GAG GAA-3′ and reverse, 5′-CCT GGT GGC TGC GTA TGT G-3′; LH beta (Lhb), forward, 5′-CAC CTT CAC CAC CAG CAT CT-3′ and reverse, 5′-GCA CAC GAG GCA AAG CA-3′; Aromatase (Cyp19a1), forward, 5′-CCA TCA TGG TCC CGG AAA C-3′ and reverse, 5′-GGC CCA TGA TCA GCA GAA GT-3′. All primer sets were optimized for the appropriate primer concentration using a concentration gradient (75, 150, and 300 nM final concentration in 25-μl reaction). The primer efficiency was determined to be 95%–110% for all primer sets.
All samples were assayed in duplicate for each gene, and cyclophilin B (Ppib) was used to normalize for cDNA variability between samples. Real-time PCR was performed on 15 ng of cDNA with SYBR green master mix (Applied Biosystems) using a 7000 ABI prism sequence detection system (Applied Biosystems). Standard thermocycler conditions were: 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 1 min. Relative fold change in target mRNAs was quantified using the ΔΔCT method .
Female Ctnnb1flox/flox mice were maintained in the Eastlick Vivarium on the campus of Washington State University. Mice 22–24 days of age were injected subcutaneously with 0.75 mg of 17β-estradiol in propylene glycol daily for 3 days. Ovaries were harvested, trimmed of fat and bursa, and incubated for 30 min at 37°C, 5% CO2, in 6 mM EGTA in Dulbecco modified Eagle medium (DMEM)/F-12 (Invitrogen) media. Ovaries then were incubated in 0.5 M sucrose in DMEM/F-12 media for 30 min. Granulosa cells were collected from ovaries by puncturing the follicles with a 30-gauge needle. Granulosa cells were seeded onto a 35-mm dish (four ovaries per dish) in DMEM/F-12 media containing 10% fetal bovine serum and 100 U/ml penicillin per 100 ng/ml streptomycin (Invitrogen, Grand Island, NY) for 4 h at 37°C, 5% CO2, prior to adenoviral transduction. Medium and unattached cells were removed, and the granulosa cells were exposed to recombinant adenoviruses in serum-free medium supplemented with penicillin/streptomycin for 13 h. At 24 h after transduction, cells were incubated in the presence or absence of 100 ng/ml purified human FSH (National Hormone and Peptide Program, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD) and 10−7 M testosterone propionate in serum-free conditions for 24 h.
Protein from primary granulosa cells was collected from the Trizol fraction according to the manufacturer's protocol. Protein concentrations were estimated using Coomassie Plus Protein Assays (Pierce, Rockford, IL). Samples were separated on 10% SDS-PAGE Tris·HCl gels, and resolved proteins were transferred to polyvinylidene fluoride (Bio-Rad Laboratories, Hercules, CA) membranes. Western blot analysis for β-catenin was performed using anti-β-catenin (no. 610154; BD Transduction Laboratories) at a final concentration of 1:5000. Following incubation with horseradish peroxidase-conjugated secondary antibody (1:20000 final concentration; Pierce), antigen-antibody complexes were detected by chemiluminescence using Immobilon Western detection reagent (Millipore, Billerica, MA). Membranes were reprobed with anti-Akt (no. 9272; Cell Signaling Technology) at a final concentration of 1:5000 to determine sample loading.
Serum samples were analyzed by the Center for Reproductive Biology Assay Core at Washington State University. Serum samples and assay standards were extracted as described previously . Briefly, duplicate 100-μl samples were extracted with 3 ml of methyl-tert-butyl ether for 1 min on a multi-tube vortex. The solvent and aqueous phases were allowed to separate for 5 min, and the aqueous phase was frozen over liquid nitrogen. The solvent fraction was decanted into 12 × 75 mm borosilicate glass assay tubes and dried overnight under a fume hood. Any remaining solvent was removed at 37°C under air. Once dried, the samples and standards were reconstituted in 100 μl of FTA hemaglutination buffer (Becton Dickinson and Co., Sparks, MD) made with ultra-PURE water (Invitrogen Corp.). Reconstituted samples were immediately assayed using a double-antibody radioimmunoassay (DSL-4400; Diagnostic System Labs, Webster, TX) with the following modifications. Antibody was added to all tubes except total counts and nonspecific binding (NSB), and tubes were incubated for 3 h at 4°C before the addition of the 125I-labeled estradiol, followed by overnight incubation at 4°C. Bound and free estradiol was separated following the manufacturer's protocol. The detection limit of the assay was 1.25 pg/ml, and the intraassay and interassay coefficients of variation were 5.3% and 4.4%, respectively.
Statistical measurements for the in vivo experiments are summarized using a five-number data summary (box-and-whisker plot) to display the data spread, degree of skew, and the presence of outliers. The box contains the middle 50% of the data and is divided into quartiles. The upper edge indicates the 75th percentile of the data set (first quartile), and the lower edge represents indicates the 25th percentile (third quartile). The interquartile range is the difference between the third quartile and the first quartile and represents values less influenced by extremes. The line in the box represents the median; a median line that is not equidistant indicates skewed data. The largest and smallest values for each data set are represented by the ends of the vertical lines. Any observation greater than 1.5 interquartile range is considered an outlier. All experiments were analyzed using the nonlinear Mann-Whitney U-test. Significance was set at P < 0.05.
Amhr2 is expressed in the Müllerian duct and gonads as early as Embryonic Day 13.5 [32, 33]. In females, the Müllerian duct will differentiate into the oviducts, uterus, cervix, and upper vagina . Original characterization of the Amhr2cre mouse (Amhr2tm3(cre)Bhr) reported cyclization recombination (CRE) activity as early as Embryonic Day 11.5 in male and female urogenital ridges, and at Embryonic Day 12.5, activity was associated only with male and female gonads and Müllerian ducts . During postnatal development, CRE activity was detected in secondary and small antral follicles but absent in primordial or primary follicles . Since this characterization, the Amhr2cre mouse has been used extensively as a means to target deletion in granulosa cells [18, 35–37].
To achieve conditional recombination of β-catenin, mice homozygous for a floxed allele of β-catenin (Ctnnb1tm2Kem , denoted henceforth as Ctnnb1flox/flox) were bred to mice heterozygous for a “knock-in” allele containing a Cre-Neo cassette inserted into the fifth exon of Amhr2 (Amhr2tm3(Cre)Bhr , denoted henceforth as Amhr2cre/+). CRE-mediated excision of the regions between exons 2 and 6 of the Ctnnb1 gene converted the floxed allele into a recombined allele no longer capable of encoding a functional β-catenin protein .
Originally, tail DNA was used to follow the genotype of each animal and identify Ctnnb1flox/flox; Amhr2cre/+ mice. All litters yielded mice with the expected Mendelian ratio of recombined β-catenin alleles—wild-type:heterozygous:homozygous pups (1:2:1)—ruling out any embryonic lethal events.
Reports indicate that tissue-specific expression of Amhr2 is confined to members of the embryonic Müllerian duct lineage [32, 33, 38] and granulosa cells in secondary and small antral follicles , and that recombination in these tissues may vary [32, 39]. To gauge both specificity and extent of CRE-mediated recombination, PCR was performed on DNA from granulosa cells, oviduct, and uterus using a battery of primers that identify wild-type Ctnnb1 (221 bp), floxed Ctnnb1 (324 bp), recombined Ctnnb1 (631 bp; denoted floxdel), and Cre (200 bp; Fig. 1A and data not shown). To verify that the recombination of Ctnnb1 was restricted to members of the Müllerian duct lineage and granulosa cells, we profiled additional tissues, including the brain, pituitary, heart, liver, and tail (Fig. 1B).
Mice were designated as wild type if they lacked Cre and were either homozygous or heterozygous for the floxed Ctnnb1 allele (Fig. 1A, left). Mice were designated as mutant if they harbored a hemizygous allele for Cre and were heterozygous for the floxed and deleted (Floxdel) Ctnnb1 alleles (Fig. 1A). The latter genotype was expected because CRE-mediated recombination within a given cell rarely results in deletion of both alleles . As expected, PCR analysis of granulosa cells, uterus, and oviduct obtained from mutant mice clearly indicated occurrence of CRE-mediated recombination (Fig. 1). In contrast, we were surprised to find various levels of recombined Ctnnb1 in the brain, pituitary, heart, and liver from the mutant mice because these organs arise independently from the Müllerian duct lineage. This apparent leaky expression of Cre and subsequent recombination of Ctnnb1 in a variety of organs mandated physiological characterization from organs that comprise the reproductive tract and pituitary gonadal axis.
Ovaries from mature (data not shown) and immature (29–31 days old) Ctnnb1 mutant mice had normally appearing gross and histological phenotypes (Fig. 2A). In addition, superovulation of immature mice resulted in normal folliculogenesis and ovulation, as indicated by similar numbers of corpora lutea counted in serial sections from the ovaries of mutant and wild-type mice (data not shown). Furthermore, oocytes isolated from the antral follicles of immature (28–31 days old) wild-type (n = 226 oocytes) and mutant (n = 63 oocytes) mice underwent spontaneous meiosis as evidenced by germinal vesicle breakdown in culture. Similarly, the oocytes from both groups successfully extruded a polar body and exhibited normal MII metaphase configuration (Fig. 2B).
Similarly to the immature mice, non-hormonally stimulated adult animals also exhibited no difference in the number of corpora lutea when comparing Ctnnb1 wild-type and mutant ovaries (data not shown). Collectively, these data emphasize that the ovaries of Ctnnb1 mice appear normal.
To assess the impact of conditional deletion of Ctnnb1 on fertility, four wild-type and three Ctnnb1 mutant mice (8 wk of age) were placed in the presence of proven male studs for a period of 6 mo (Table 1). All wild-type mice became pregnant, gave birth, and raised pups to weaning. In contrast, the mice with a recombined Ctnnb1 gene failed to produce a single litter in the same 6-mo period, despite plugs being observed after mating. This result suggests a discrepancy between apparently normal ovaries and reproductive viability.
Subsequent examination of the reproductive tract revealed a striking phenotype in the oviducts of immature (29–31 days old) Ctnnb1 mutant mice, namely, partial oviductal formation that lacked the coiling typical of wild-type mice (Fig. 3A). Although we did not examine oviduct differentiation on a cellular level, others have reported the absence of the mucosal folding in the epithelial layer of the utero-oviductal junction in mice that carry both a conditional recombined allele and a nonfunctional allele of Ctnnb1 .
Despite partial oviductal formation, the oviductal ampulla of immature Ctnnb1 mutant mice responded to superovulatory doses of eCG and hCG to the same extent as immature control littermates, as noted by swelling of the ampulla and accompanying striations (n = 3 for both groups; data not shown). Nevertheless, although oocytes were found in the ampulla of control and Ctnnb1 mutants, fewer oocytes were recovered from the mutants (average of 63 vs. 15, respectively). Additionally, there was no evidence of entrapped oocytes in corpora lutea in the serial sections of ovaries from hormonally stimulated mice. The decreased oocyte number in the ampulla is thus likely the result of the oviductal deformity because normal folliculogenesis and ovulation occur in mutant mice.
In addition to oviductal deformity, Ctnnb1 mutant mice presented with thinner uteri that appeared to have a large amount of associated fat along with diminished tone (Fig. 3B). Uteri from control littermates (n = 24) tended to weigh more than uteri from mice carrying the recombined Ctnnb1 allele (n = 11; P = 0.19; Fig. 3C). This difference is most likely a result of the subtle but obvious morphological differences that include changes in the uterine mesenchyme. Such an outcome is consistent with previous reports indicating that conditional recombination and deletion of Ctnnb1 in the uterus result in deficiency of myometrial smooth muscle and replacement by adipose .
Because Ctnnb1 mutant mice are capable of ovulating oocytes that appeared within the ampulla, we examined the ability of the embryos to implant in the uterus. Wild-type mice had noticeable swellings along the uterus indicating sites of implantation 6 days after an observed plug (Fig. 3D). In contrast, sites of implantation were absent in Ctnnb1 mutant mice 6 days after breeding. One possible mechanism is the noted deformity in the oviduct, but it is also possible that there are contributions from the uterus. The inability to support implantation is consistent with the reports that WNT via β-catenin-directed vascularization is essential for placental development [40–42], owing to the maternal as well as fetal contributions to the placenta.
The observation that CRE-mediated recombination of β-catenin failed to alter ovarian function was unexpected in light of our recent report  that targeted reduction of β-catenin in primary cultures of rat granulosa cells or in an established human granulosa cell line (KGN) reduced FSH-regulated expression of aromatase. Therefore, to address this apparent discrepancy, we examined the levels of the mRNAs encoding β-catenin and aromatase in granulosa cells from 8-wk-old Ctnnb1 mutant and wild-type mice as well as serum levels of estradiol to determine the impact of conditional CRE-mediated recombination of Ctnnb1.
Quantitative real-time PCR analysis of mRNA isolated from granulosa cells from adult randomly cycling mice displayed significant animal-to-animal variation in the levels of β-catenin mRNA, as indicated by box-and-whisker plots (Fig. 4A). Although not statistically significant, levels of β-catenin mRNA in the mutant mice displayed a reduced trend when taking out the contributions of a single outlier identified through the box-and-whisker plots (P = 0.1; Fig. 4A). Similarly, expression of granulosa cell aromatase mRNA had a skewed distribution in Ctnnb1 mutant mice that trended toward lower levels when compared with their wild-type littermates (Fig. 4B). Despite these trends, however, estradiol concentrations were not statistically different between the two groups in randomly cycling mice, although the range of variation was far greater in the Ctnnb1 mutant mice (Fig. 4C). To exclude the possibility that these results were due solely to different stages of the estrous cycle, wild-type (n = 4) and mutant (n = 4) immature mice were treated with eCG to synchronize the stage of the estrous cycle, and again no difference (P = 0.7) was demonstrated in estradiol concentrations (99 ± 19 vs. 89 ± 15 pg/ml, respectively).
To further explore the disconnection between levels of serum estradiol and the apparent changes in Ctnnb1 and Cyp19a1 mRNA, we measured levels of the pituitary mRNAs encoding Lhb and Fshb (Fig. 5A) because both hormones contribute to estrogen production either during folliculogenesis or after ovulation. We also measured levels of the ovarian mRNAs encoding Inha and Lhr (officially Lhcgr) because both contribute to the steroidogenic response (Fig. 5B).
Despite similar serum levels of estradiol, Fshb mRNA expression was significantly downregulated (P < 0.05) in the pituitaries of Ctnnb1 conditional mutants (Fig. 5A). The accompanying apparent rise in ovarian Inha (Fig. 5B) suggests that Fshb could be responding to negative signals from Inha. The expression of Lhb mRNA in the pituitary and Lhr in the whole ovary were skewed to indicate a higher expression in Ctnnb1 mutants than wild-type mice (Fig. 5).
The opposing trends in Fshb and Lhr mRNA could be a compensatory response required to maintain serum levels of estradiol in the normal range in mice with a conditionally deleted allele of Ctnnb1. Although it is tempting to link the reduction of Fshb mRNA directly with the detection of recombined alleles of β-catenin in the pituitary, the genotyping was carried out on the whole pituitary. Thus, it remains unclear whether recombination of β-catenin alleles was cell specific and confined to gonadotropes or was more widespread, involving one or several of the other four cell types in the pituitary (lactotropes, somatotropes, thyrotropes, and corticotropes). Alternatively, changes in Fshb and Lhr may be unrelated and reflect incidental noise due to tissue and animal-to-animal variation as well as the widespread and unexpected recombination of Ctnnb1 in a range of tissues. Instead, the apparently normal ovarian morphology and normal levels of serum estradiol may be an indication that CRE-mediated recombination of β-catenin in proliferating granulosa cells may be unstable, and therefore may mask the suspected steroidogenic requirement for β-catenin.
To circumvent many of the caveats noted above, we elected to tease out the effects of Ctnnb1 recombination by employing short-term primary cultures of granulosa cells. This allowed for assessment of the effect of β-catenin deletion in granulosa cells that at most only undergo a single round of division. This approach also avoided confounding effects from recombination that occur in tissues derived from within and outside the Müllerian duct lineage.
Following isolation of granulosa cells from mice homozygous for floxed Ctnnb1 alleles (Ctnnb1flox/flox), these cells were transduced for 24 h with an adenovirus encoding CRE recombinase or an adenovirus encoding enhanced green fluorescent protein (EGFP) . At harvest, DNA was collected from the cultures and genotyped as described previously for mouse tissues to confirm recombination of Ctnnb1 in the cells exposed to the CRE recombinase adenovirus (Fig. 6A). Further confirmation that the adeno-CRE recombinase effectively recombined the LoxP sites was demonstrated by RT-PCR and Western blot analysis that showed notable reduction of β-catenin mRNA and protein (Fig. 6, B and C).
To assess the functional consequence of reduced levels of β-catenin after CRE-mediated recombination, granulosa cells were transduced with either EGFP control adenovirus or CRE recombinase adenovirus and then incubated for 24 h with either vehicle or FSH. Quantitative real-time PCR analysis indicated that the 24-h treatment with FSH markedly induced aromatase mRNA compared with EGFP controls (Fig. 7, upper). This upregulation of aromatase mRNA correlated positively with elevated levels of estradiol measured in the media (Fig. 7, lower). In contrast, in granulosa cells exposed to the adeno-CRE recombinase, the response of aromatase mRNA to FSH was significantly (P < 0.01) muted and corresponded with lower levels of estradiol (P < 0.05) in the media. Together, these experiments firmly establish that FSH regulation of aromatase gene expression requires β-catenin.
The Amhr2cre transgene has been used extensively to drive granulosa cell-specific recombination of a gene of interest [35–37]. However, the data reported here indicate that CRE-mediated recombination of β-catenin directed by the Amhr2 promoter in mice may be too unstable to yield an ovarian phenotype, whereas the oviduct and uterus are markedly affected. Our data also indicate that Amhr2cre caused recombination of Ctnnb1 unexpectedly in a range of other organs, including brain, pituitary, heart, liver, and tail. These numerous sites of β-catenin recombination probably contribute directly and indirectly to the spectrum of phenotypic changes observed, including infertility that is most likely due to abnormal oviductal and uterine development. In addition, this wide-ranging pattern of CRE-mediated recombination resulted in changes in the expression of a number of genes associated with the hypothalamic-pituitary-gonadal (HPG) axis, including Fshb and, to a lesser extent, the marked animal-to-animal variation of Lhb in the pituitary, as well as Ctnnb1 and Cyp19a1 in granulosa cells, and Inha and Lhr in the ovary. Although ovarian histology appears unaffected in mice with recombined alleles of Ctnnb1, and serum estradiol resides within a normal range, the importance of β-catenin to ovarian steroidogenesis may be masked by compensatory responses that promote selective proliferation of granulosa cells in growing follicles that escape CRE-mediated recombination. Indeed, our experiments with primary cultures of granulosa cells support this view by demonstrating clearly that FSH regulation of aromatase gene expression and subsequent production of estradiol require β-catenin.
Despite the wide range of conditional CRE-mediated recombination of Ctnnb1, several of the reproductive tract anomalies probably reflect a local site of action. For example, Arango et al.  reported previously that Amhr2-mediated conditional recombination of β-catenin resulted in female mice with uteri grossly deficient in smooth muscle and replaced instead with adipose. The analysis of the weight and gross morphology of the uterus in Ctnnb1 mutant mice produced in this study revealed uteri that were thin, fatty, and lacked the classic balloonlike appearance demonstrated by the wild-type controls. Because uterine weight is one of the traditional bioassays for estrogen production, it is tempting to conclude that alterations in uterine morphology reflect changes in steroid levels. Nevertheless, because estrogen levels were in the normal range, the alterations in uterine morphology more likely reflect a local site of action rather than a change in steroid levels. This possibility is consistent with the observation that Ctnnb1 mutant mice enter puberty, an estrogen-triggered event , at the expected age (data not shown). Thus, changes in uterine morphology probably reflect diminished production of β-catenin in mesenchymal cells, causing a switch from myogenesis to adipogenesis .
Another likely local effect of reduced β-catenin is the failure of oviducts from Ctnnb1 mutant mice to differentiate properly, as evidenced by the lack of a typical coiled appearance observed in wild-type, age-matched mice (Fig. 3A). A similar phenotype has been reported in mice lacking Wnt7a, a member of the WNT family that has been shown to play a role in patterning of the Müllerian duct [45, 46]. Consistent with these studies, Deutscher and Yao  reported that female mice with one β-catenin null allele and one conditionally recombined allele of β-catenin have an even more severe uterine and oviductal phenotype compared with wild-type littermates . Our results emphasize the exquisite sensitivity of the oviduct to subtle changes in β-catenin because a dramatic phenotype occurred with only a single, conditionally recombined allele of Ctnnb1.
In contrast to the uterine and oviduct phenotypes, no gross morphological or histological phenotype was observed in the ovaries of conditionally deleted Ctnnb1 mice by 8 wk of age (Fig. 2). In addition, both wild-type and mutant mice had similar numbers of corpora lutea following eCG/hCG stimulation (data not shown), suggesting that the loss of fertility in the Ctnnb1 mutants (Table 1) cannot be explained by ovarian failure. Despite similar numbers of corpora lutea, fewer oocytes were captured and recovered from the ampulla of Ctnnb1 mutant mice. Yet, oocytes collected from Ctnnb1 mutant mice underwent spontaneous maturation to the same degree as wild-type mice, suggesting that β-catenin deficiency has little or no impact on oocyte quality. Instead, the presence of fewer oocytes in Ctnnb1 mutant mice suggests that partial formation of the oviduct and the lack of coiling are the likely culprits for inefficient capture and release of oocytes from the ampulla. Alone, however, inefficient capture of otherwise healthy oocytes would be expected to lead to subfertility, as opposed to a complete loss of fertility. Because Ctnnb1 mutant mice also display a complete failure of implantation, presumably due to incomplete patterning of the uterus during development [32, 39], it seems most likely that oviductal and uterine defects supply a one-two punch that explains the complete loss of fertility in mice with recombined alleles of Ctnnb1.
Establishing a link between changes in β-catenin, aromatase, and estradiol in Ctnnb1 mutants was problematic. Granulosa cells from mature female mice with recombined alleles of β-catenin displayed marked animal-to-animal variation in the mRNAs encoding both the coactivator and aromatase (Fig. 4, A and B). In contrast, serum estradiol levels were essentially the same (Fig. 4C). Although the use of randomly cycling females could be a factor, we view this as unlikely because use of exogenous gonadotropins to synchronize follicular maturation in immature mice also failed to reveal a difference in estradiol concentrations between wild-type and mutant mice. Another possibility is that leaky expression of CRE in tissues outside of the Müllerian duct lineage, along with unexpected interactions along the HPG axis, also obfuscates analysis of the relationship between β-catenin and FSH regulation of aromatase gene expression in the ovary. For example, recombination of floxed alleles of β-catenin was detected in the brain and pituitary and, to a lesser extent, in the heart, liver, and tail (Fig. 1). In addition, elevated trends in the mRNAs encoding Lhb, Inha, and Lhr, along with the significant decrease in Fshb mRNA (Fig. 5), suggest that the apparently normal levels of estradiol in mice with recombined alleles of β-catenin may reflect a yet unidentified net compensatory response. Similar interactions may explain why Jorgez et al.  also found a lack of association between changes in FSH and estradiol when Amhr2cre was used to target deletion of follistatin; FSH levels changed significantly without affecting serum levels of estradiol, yet female mice displayed a range of fertility defects.
Given our previous in vitro studies indicating that FSH regulation of aromatase requires β-catenin , the apparently normal morphology and function of the ovary were surprising, especially because Amhr2cre is expressed readily in secondary and small antral follicles but not primordial or primary follicles . Moreover, Amhr2cre-mediated deletion of follistatin (officially Fst) and Sf1 markedly impaired ovarian function [35, 37], suggesting that either β-catenin has little impact on ovarian function or that CRE-mediated recombination of β-catenin may be unstable in proliferating granulosa cells. For example, if granulosa cells with normal levels of β-catenin have a selective proliferative advantage over granulosa cells that are deficient in β-catenin because of CRE-mediated recombination, then reoccurring cycles of folliculogenesis may lead to steady loss of recombined alleles. In this regard, it is important to note that β-catenin mediates the effect of cadherins on cell-cell interactions through the planar polarity pathway, as well as effects emanating from the canonical WNT signaling pathway . Thus, alterations in β-catenin are likely to affect both pathways. Two recent reports support this possibility. First, conditional expression of a dominant, stable form of β-catenin alters the fate of murine granulosa cells and promotes the formation of granulosa cell tumors . Second, conditional reduction of β-catenin reduces expression of cyclin D2 and retards expansion of neonatal pancreatic beta cells . Because FSH specifically regulates cyclin D2 in granulosa cells , it is reasonable to suspect that reductions in β-catenin may affect proliferation of granulosa cells through alterations of cyclin D2 as well as disrupt cell-cell communication critical for cell survival. In short, conditional deletion of β-catenin appears to be unstable in granulosa cells compared with other derivatives of the Müllerian tract or with conditional deletion of follistatin or Sf1 [35, 37].
Consequently, establishing short-term primary cultures of mouse granulosa cells followed by transduction with CRE-expressing recombinant adenovirus provided a means to remove granulosa cells from the dynamics of a growing follicle, isolate them from other CRE-mediated changes in the HPG axis, and allow for direct assessment of the action of FSH on aromatase gene expression. Because primary cultures of granulosa cells undergo limited cell division, their use also increases the likelihood of stably maintaining recombined alleles of β-catenin.
Although FSH regulation of aromatase gene expression requires components from FSH and Frizzled/LRP receptor signaling pathways, the point of intersection between these two pathways remains unclear. A recent report , along with unpublished data from our laboratory, indicates that activation of the gonadotropin-releasing hormone receptor (GNRHR), another GPCR, stimulates nuclear accumulation of β-catenin in gonadotrope-specific cell lines. Activation of GNRHR also stimulates phosphorylation of glycogen synthase kinase-3β (GSK3B) that lies upstream of β-catenin in the canonical WNT signaling pathway ( and our unpublished data). This result raises the possibility that regulation of GSK3B and subsequent nuclear accumulation of β-catenin are common to hormones, such as FSH, that signal through GPCRs. Nevertheless, additional studies will be needed to address whether FSH directly activates the WNT/Frizzled receptor pathway or instead signals through targets that lie downstream of the WNT/Frizzled receptor.
Although it is well established that FSH regulation of aromatase gene expression is mediated by several transcription factors, including CREB [3, 51], SF1 [2, 52–54], and GATA4 [4, 5], the data from the primary cell culture experiments (Figs. 6 and and7)7) collectively suggest that β-catenin also plays an essential role in mediating the actions of the gonadotropin. The FSH regulation of aromatase gene expression thus appears to use intracellular signals associated with two membrane receptor pathways: cAMP from GPCRs that bind FSH, and β-catenin from the Frizzled/LRP receptor pathway that normally mediates the actions of WNT.
Beta-catenin associates with the type II aromatase promoter, as evidenced by chromatin immunoprecipitation analysis of extracts from KGN cells . Follicle-stimulating hormone regulation of aromatase gene expression also requires the binding of orphan nuclear receptor SF1, which in turn interacts with β-catenin via four acidic residues within the ligand-binding domain of the orphan receptor . Although not examined directly in this study, we suspect that SF1 is the downstream target of β-catenin in mouse granulosa cells that regulates aromatase expression.
In conclusion, conditional deletion of Ctnnb1 occurs across an unexpectedly wide range of tissues in the Amhr2cre mouse, including the pituitary. This widespread pattern of CRE-mediated recombination complicates analysis of reproductive tract defects because contributions from the pituitary and other organs, such as the hypothalamus, cannot be ruled out. Nevertheless, it is clear that deficiencies in β-catenin are associated with overt defects in the oviduct and uterus that render females completely infertile. Although ovarian function in mice deficient for β-catenin appears normal, our in vitro studies with primary cultures of granulosa cells bearing floxed alleles of β-catenin underscore the importance of the coactivator with respect to FSH regulation of steroidogenesis, a role that remains hidden when tested through Amhr2cre-mediated recombination in vivo.
We would like to thank Dr. Richard Behringer for generously providing the Amhr2cre mice. We are grateful to Rong Nie for processing the ovaries for histology and David deAvila in the Center for Reproductive Biology Assay Core at Washington State University for assaying the serum and media samples. The authors would also like to acknowledge Dr. Pat Hunt and several members of her lab who contributed to the oocyte data collection. We also acknowledge help with statistical analysis provided by Dr. Bryan Slinker in the School of Veterinary Medicine at Washington State University. We are indebted to April Binder for graciously agreeing to genotype the pituitary samples and for taking over care of the mouse colonies. The deepest gratitude is expressed to all members of the Nilson and Hunzicker-Dunn laboratories for helpful discussion and evaluation of the data and to Maria Herndon for critical review of the manuscript.
1Supported in part by National Institutes of Health (NIH) grant HD-21921 to M.H.-D. and an NIH minority fellowship to J.A.H.G.