Misregulation of the Wnt/β-catenin signaling pathway is a hallmark of several forms of cancer. Components of the Wnt/β-catenin pathway are expressed in ovarian granulosa cells; nevertheless, its potential involvement in granulosa cell tumorigenesis has not been examined. To this end, human (n = 6) and equine (n = 18) granulosa cell tumors (GCT) were analyzed for β-catenin expression by immunohistochemistry. Unlike granulosa cells of normal ovaries, most (15 of 24) GCT samples showed nuclear localization of β-catenin, suggesting that activation of the Wnt/β-catenin pathway plays a role in the etiology of GCT. To confirm this hypothesis, Catnbflox(ex3)/+; Amhr2cre/+ mice that express a dominant stable β-catenin mutant in their granulosa cells were generated. These mice developed follicle-like structures containing disorganized, pleiomorphic granulosa by 6 weeks of age. Even in older mice, these follicle-like lesions grew no larger than the size of antral follicles and contained very few proliferating cells. Similar to corpora lutea, the lesions were highly vascularized, although they did not express the luteinization marker Cyp11a1. Catnbflox(ex3)/+; Amhr2cre/+ females were also found to be severely subfertile, and fewer corpora lutea were found to form in response to exogenous gonadotropin compared with control mice. In older mice, the ovarian lesions often evolved into GCT, indicating that they represent a pretumoral intermediate stage. The GCT in Catnbflox(ex3)/+; Amhr2cre/+ mice featured many histopathologic similarities to the human disease, and prevalence of tumor development attained 57% at 7.5 months of age. Together, these studies show a causal link between misregulated Wnt/β-catenin signaling and GCT development and provide a novel model system for the study of GCT biology.
- Cancer etiology
- Granulosa cell tumor
- Animal model of human disease
- β-catenin, Conditional gene targeting
Ovarian cancer is currently the fifth leading cause of cancer deaths in North American women ( 1). Unlike many other forms of cancer, age-adjusted incidence and death rates from ovarian cancer have improved very little over the past ≥30 years ( 1). Most research in this field has focused narrowly on neoplasms that arise from the ovarian surface epithelium, as these comprise the majority of ovarian cancers ( 2). Comparatively little attention has been paid to the sex cord/stromal tumors, although they are thought to represent up to 10% of all ovarian neoplasms ( 2). The most prevalent tumor in the latter group is the granulosa cell tumor (GCT), which constitutes ∼5% of all ovarian cancers ( 3). Contrary to what is seen in humans, tumors of the ovarian surface epithelium are relatively uncommon in most domestic species. Indeed, in species such as the cow, ewe, and mare, GCT is by far the most frequently observed neoplastic disease of the ovary ( 4).
Although much is now known of the molecular pathogenesis of ovarian epithelial tumors, the etiology of GCT remains unclear ( 5). The proliferation of granulosa cells is controlled in large part by pituitary gonadotropin signaling via the G protein–coupled follicle-stimulating hormone (FSH) receptor. Several groups have therefore sought to identify activating mutations in either the FSH receptor or various G protein subunits in GCT but to no avail ( 6– 9). One group identified a mutation in the KRAS gene in 1 of 10 GCT samples ( 10), although a more recent report ( 11) failed to detect a similar mutation in a separate panel (n = 0 of 9), suggesting that KRAS activation is a rare event in GCT. Aside from this finding, the activation of any other oncogenes has yet to be associated with GCT development ( 5). Similarly, very few instances of inactivation of tumor suppressor genes in GCT have been reported thus far. The tumor suppressor genes TP53 and WT1, which are commonly mutated in a broad range of neoplasia, are thought not to play a role in the pathogenesis of GCT ( 12– 14). One recent report found a lack of expression of the cyclin-dependent kinase inhibitors INK4A and/or INK4B in a significant proportion of GCT ( 15), suggesting a means by which GCT cells may escape cell cycle control. However, the importance of this potential pathogenic mechanism to GCT development remains to be determined. Mice null for Inha (which encodes the inhibin α subunit) develop mixed gonadal stromal tumors in a gonadotropin-dependent manner, and it has therefore been postulated that Inha functions as a tumor suppressor gene ( 16, 17). Whereas the Inha knockout model has been of great use in the study of both ovarian physiology and tumor biology, its relevance to the study of GCT is unclear, particularly in view of the fact that the vast majority of GCT synthesize large amounts of inhibin ( 5, 18). Two additional transgenic mouse models have been reported to develop GCTs, one of which chronically hypersecretes luteinizing hormone (LH; ref. 19), whereas the other expresses the SV40 T-antigen driven by the Inha promoter ( 20). Again, whereas much information has been obtained from these animals, the development of a clinically and biologically relevant mouse model for GCT awaits the identification of relevant genetic lesions and pathogenic mechanisms in spontaneously occurring GCT.
The Wnts comprise a large family of secreted glycoprotein-signaling molecules that, via Frizzled (Fzd) receptors, act to exert numerous roles in embryonic development ( 21, 22). A key effector of the canonical Wnt signal transduction pathway is β-catenin, a multifunctional protein that also plays essential roles in mediating cell-cell adhesion by interacting with-type II cadherins near the cell surface ( 23). In Wnt signaling however, a distinct pool of β-catenin protein localizes to the cytoplasm, where it resides in a large multiprotein complex that includes adenomatous polyposis coli (APC) and Axin (reviewed in refs. 21, 22, 24). In the resting state, components of this complex ensure the phosphorylation and subsequent ubiquitination and degradation of β-catenin, thereby limiting the free intracytoplasmic pool of β-catenin. In the presence of Wnt signal, β-catenin dissociates from this complex and translocates to the nucleus, where it acts to modulate the transcriptional activity of a wide range of target genes. Given the effect of Wnt signaling on cell proliferation and cell fate decisions during embryogenesis, it is not surprising that misregulated Wnt signaling is a hallmark of many forms of cancer ( 21, 24). Indeed, the APC gene has been found to be mutated in 80% of colorectal cancers, and mutations in the genes encoding β-catenin, APC, Axin-1, and Axin-2 have been detected in tumors of the ovarian surface epithelium ( 24– 27). Whereas Wnt signaling plays a key role in the embryonic development of the ovary ( 28– 30), several recent reports have also described the expression of Wnt2 and Wnt4 and the Wnt signaling pathway components Sfrp4, Fzd1, and Fzd4 in adult rodent ovarian granulosa cells ( 31– 33). Interestingly, the pattern of expression and hormonal regulation of these Wnt effectors suggests a likely role for Wnt signaling in follicular development in adult mice. Another implication of these findings is that misregulation of Wnt signaling in granulosa cells could contribute to GCT development.
The objective of this study was to investigate the possible involvement of the Wnt/β-catenin signaling pathway in GCT etiology. We evaluated archived human and equine GCT samples for β-catenin expression by immunohistochemical analysis and found that a large proportion featured β-catenin expression localized to the nucleus, indicative of activation of the Wnt/β-catenin pathway. To confirm and further study the role of β-catenin in the pathogenesis of GCT, transgenic Catnbflox(ex3)/+; Amhr2Cre/+ mice were generated that express a dominant stable mutant of β-catenin in granulosa cells. These were found to develop multiple ovarian lesions resembling disorganized follicles and cystic structures, which later gave rise to GCT with a high penetrance. This study therefore represents an important advance in our understanding of the etiology of GCT and also provides a valuable model system for the further study of GCT biology.
Materials and Methods
Animals and hormone treatments. Catnbflox(ex3)/+; Amhr2Cre/+ mice were derived from previously described Amhr2Cre and Catnbflox(ex3) parental strains, and genotyping was done by PCR as described ( 34, 35). Ovaries for histopathologic analysis were fixed in buffered 10% formalin for 24 hours before paraffin embedding, sectioning, and H&E staining. To study ovarian response to exogenous gonadotropins, 23- to 24-day-old Catnbflox(ex3)/+; Amhr2Cre/+ and Catnbflox(ex3)/+ mice were injected with either equine chorionic gonadotropin (Gestyl, Professional Compounding Center of America, Houston, TX; 5 IU, i.p., injections at t = 0 hour and t = 48 hours, animals sacrificed at t = 96 hours) or saline vehicle alone (n = 3-7/genotype and treatment group). All procedures were approved by the Institutional Animal Care and Use Committee and were conform to the USPHS Policy on Humane Care and Use of Laboratory Animals.
Immunohistochemical and immunofluorescence analyses. Immunohistochemistry was done on formalin-fixed or paraformaldehyde-fixed, paraffin-embedded 7-μm sections using VectaStain Elite avidin-biotin complex method kits (Vector Labs, Burlingame, CA) as directed by the manufacturer. Sections were probed with primary antibodies against β-catenin (PharMingen, San Diego, CA), proliferating cell nuclear antigen (PCNA, Novocastra Laboratories, Newcastle upon Tyne, United Kingdom), or cyclin D2 (Lab Vision Corp., Fremont, CA), and staining was done using the 3,3′-diaminobenzidine peroxidase substrate kit (Vector Labs) as directed. Human and equine GCT samples consisted of archived, paraffin-embedded tumor fragments obtained during surgical resections at the University of Texas M.D. Anderson Cancer Center (Houston, TX) or the Centre Hospitalier Universitaire Vétérinaire de l'Université de Montréal (St. Hyacinthe, Québec, Canada), respectively. Normal human ovary samples (n = 4) were obtained from the same establishment as the GCT and were removed from premenopausal women undergoing surgery for nonovarian gynecologic conditions. Preovulatory equine follicular samples (n = 3) were obtained as previously described ( 36). For immunofluorescence experiments, ovaries were embedded in OCT compound (Sakura Finetek USA, Inc., Torrance, CA) and stored at −70°C before the preparation of 5-μm sections, which were fixed for 1 hour in PBS-buffered 4% paraformaldehyde at 4°C. Sections were then sequentially probed with primary anti-type IV collagen (Biogenesis, Inc., Brentwood, NH) and secondary Alexa Fluor 594–conjugated goat anti-rabbit IgG antibodies (Molecular Probes, Eugene, OR) as previously described ( 37). Slides were mounted using VectaShield with 4′,6-diamidino-2-phenylindole (Vector Labs).
Semiquantitative reverse transcription-PCR. Reverse transcription-PCR (RT-PCR) was done using the SuperScript One-Step RT-PCR system with Platinum Taq kit (Invitrogen, Carlsbad, CA), and 100 ng samples of ovarian total RNA that had been isolated using the RNeasy mini kit (Qiagen Sciences, Germantown, MD). Reactions were formulated as directed by the manufacturer, except 0.625 μCi of [α-32P]dCTP (specific activity = 3,000 Ci/mmol, MP Biomedicals, Irvine, CA) were added to each reaction to generate quantifiable radioactive signal and to increase assay sensitivity. Oligonucleotides used for Amhr2 were 5′-CATGGCTCCAGAGCTCTTGGACA-3′ and 5′-TGGGTCTGCGTCCCAGCAATCT-3′, oligonucleotides used for Catnb1 (BCAT-GF2, AS5) and ribosomal protein L19 (Rpl19) were as previously described (refs. 34, 38, 39; Sigma-Genosys, The Woodlands, TX). Cycling conditions were 50°C for 30 minutes and 94°C for 2 minutes followed by 18 cycles (Rpl19), 22 cycles (Amhr2), or 24 cycles (Catnb1) of 94°C for 15 seconds, 55°C for 30 seconds, and 72°C for 1 minute. A final extension step of 72°C for 7 minutes was also done. Preliminary experiments were done for Amhr2 and Rpl19 to ensure that the cycle numbers selected fell within the linear range of PCR amplification (data not shown). For Catnb1, the cycle number selected was the lowest capable of detecting the recombined Catnb1 mRNA species in ovaries from 3-day-old Catnbflox(ex3)/+; Amhr2Cre/+ mice. Samples were separated by electrophoresis on 2% TAE-agarose gels, dried, and exposed to Biomax XAR film (Eastman Kodak Co., Rochester, NY) for 30 to 240 minutes at −70°C to generate the presented images. The relative radioactive signal strengths from the RT-PCR products were subsequently quantified using a Storm 860 PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA).
In situ hybridization. In situ hybridization was done on paraformaldehyde-fixed, paraffin-embedded sections of ovaries from a 12-week-old Catnbflox(ex3)/+; Amhr2Cre/+ mouse. Probe preparation, slide preparation, hybridization, washing, and exposure/developing steps were done as previously described ( 40, 41), using a Cyp11a1 cDNA fragment as a template for probe synthesis (a generous gift from Dr. Keith Parker, University of Texas Southwestern, Dallas, TX). A 16-hour exposure time to NTB-2 emulsion (Eastman Kodak) was required to generate the image shown. A sense Cyp11a1 riboprobe failed to generate any signal when hybridized under the same conditions (data not shown).
Statistical methods. One-way ANOVA was used to test for differences between groups, with P < 0.05 considered statistically significant. For the data presented in Fig. 2C, the 3-day time point was omitted from ANOVA analysis, as only one value was available. As one-way ANOVA indicated a statistically significant effect of age on relative Amhr2 mRNA levels, Tukey's test was used to individually compare all pairs. All tests were done using Prism software version 4.0a (GraphPad Software, Inc., San Diego, CA).
Misregulated Wnt/β-catenin signaling in ovarian GCT. To obtain evidence of misregulated Wnt signaling in GCT, β-catenin expression and subcellular localization were evaluated by immunohistochemical analysis using archived human GCT and normal adult ovarian samples. As GCT is the most frequently observed ovarian tumor in mares, equine ovaries and GCT were also evaluated. All human and equine GCT samples and all normal ovarian samples showed some degree of β-catenin staining within the follicular granulosa cells ( Fig. 1A-D ). In every case, staining localized to the plasma membrane, cytoplasm, and perinuclear region. Nuclear staining indicative of activation of the Wnt/β-catenin signaling pathway was observed in subsets of human (n = 1 of 6, Fig. 1C) and equine (n = 14 of 18, Fig. 1D) GCT but was not observed in normal ovarian samples ( Fig. 1A and B).
Catnbflox(ex3)/+; Amhr2Cre/+ mice develop multiple follicle-like ovarian lesions. To verify the importance of misregulated Wnt/β-catenin signaling in granulosa cell tumorigenesis, a strategy to constitutively activate the Wnt/β-catenin pathway in granulosa cells was devised using genetically engineered mice ( Fig. 2A ). The previously described Catnbflox(ex3) strain of mice features loxP sites flanking the third exon of the β-catenin gene ( 34). Upon Cre-mediated excision, the recombined allele encodes a β-catenin protein that, while still functional, lacks a series of phosphorylation sites that are required for its degradation. To target recombination to granulosa cells, Catnbflox(ex3)/flox(ex3) mice were mated to Amhr2Cre/+ mice in which the Cre transgene was knocked-in to the anti-Mullerian hormone type II receptor locus ( 42). Genetic recombination in the resulting Catnbflox(ex3)/+; Amhr2Cre/+ mice was assessed by RT-PCR on ovarian RNA at different times after birth. Results showed the presence of the shorter β-catenin mRNA species derived from the recombined allele (Δex3) as early as 3 days of age, the abundance of which subsequently increased over time ( Fig. 2B). To relate this finding to the normal postnatal pattern of ovarian Amhr2 expression, Amhr2 mRNA levels were evaluated by RT-PCR in ovaries from wild-type mice at the ages described in Fig. 2B. Results showed that Amhr2 is highly expressed in the mouse ovary at 3 days of age and that mRNA levels remain fairly constant throughout the first 7.5 months of life ( Fig. 2C). Together with previous reports of prenatal ovarian transgene expression in Amhr2cre mice ( 35, 42), these data indicate that Cre-mediated recombination in Catnbflox(ex3)/+; Amhr2Cre/+ mice follows the expression pattern of Amhr2, and that recombination likely occurs prenatally/perinatally to some extent.
Histopathologic analyses revealed several anomalies in the ovaries of Catnbflox(ex3)/+; Amhr2Cre/+ mice. Unlike control mice that lack the Cre transgene ( Fig. 3A, a-d ), the ovaries of Catnbflox(ex3)/+; Amhr2Cre/+ mice featured multiple abnormal follicle-like structures that were observable as early as the fifth or sixth week of age ( Fig. 3A, e-h and Fig. 3B, a-c). The majority of these structures consisted of solid nests of disorganized, pleiomorphic granulosa cells ( Fig. 3A, f-h, arrows and Fig. 3B, a). Some of these featured a small, antrum-like space filled with a pale eosinophilic substance, but oocytes were never observed within these lesions ( Fig. 3B, b). Less frequently, cystic structures lined with multiple layers of granulosa cells were found. These were often blood filled and typically contained cords of granulosa cells that subdivided their lumens into multiple compartments ( Fig. 3A, g-h, arrowheads and Fig. 3B, c). Whereas the solid and cavitary lesions were never found to grow beyond the size of antral follicles, the cystic lesions occasionally became larger than the ovary itself, particularly in mice beyond 6 months of age (data not shown). Detailed necropsies failed to detect any extraovarian anomalies in Catnbflox(ex3)/+; Amhr2Cre/+ mice.
Characterization of Catnbflox(ex3)/+; Amhr2Cre/+ ovarian lesions. To relate the formation of the observed ovarian lesions to the Cre-mediated gene recombination, ovaries from 12-week-old Catnbflox(ex3)/+ and Catnbflox(ex3)/+; Amhr2Cre/+ mice were analyzed by immunohistochemistry for the expression of β-catenin ( Fig. 4A-C ). Predictably, all cells within the lesions expressed high amounts of β-catenin, which localized mainly to the nucleus in many cells ( Fig. 4B and C). Interestingly, normal-looking antral follicles in both Catnbflox(ex3)/+ and Catnbflox(ex3)/+; Amhr2Cre/+ ovaries expressed β-catenin, particularly in the cytoplasm of the granulosa cells located nearest to the antrum ( Fig. 4A and B, arrowheads). As no inappropriate expression of β-catenin was observed in the normal-looking follicles or corpora lutea of Catnbflox(ex3)/+; Amhr2Cre/+ ovaries when compared with Catnbflox(ex3)/+ controls, it can be deduced that many follicles developed and ovulated without any granulosa cells undergoing genetic recombination. This suggests (a) that the recombination process was relatively inefficient; (b) that much of the recombination occurred early in the lives of the Catnbflox(ex3)/+; Amhr2Cre/+ mice; and (c) that the age-dependent increase in the levels of the truncated β-catenin transcript observed in Fig. 2B probably resulted mostly from the proliferation of the few granulosa cells in which recombination had occurred, rather than from ongoing recombination events.
As the follicle-like ovarian lesions did not seem to grow significantly after ∼6 weeks of age, the proliferative status of the cells in the lesions was assessed by PCNA immunohistochemistry in 12-week-old Catnbflox(ex3)/+ and Catnbflox(ex3)/+; Amhr2Cre/+ mice. Results showed a limited number of cells within the lesions to be proliferating, consistent with their slow growth ( Fig. 4H and I, arrows). Interestingly, the pattern of PCNA staining within the lesions was not unlike that observed in normal corpora lutea, in which a small number of PCNA-positive cells (presumed to be vascular endothelial cells) were observed interspersed among the nonproliferating luteal cells ( Fig. 4H, inset). This raised the possibility that the lesions could be vascularized, corpus luteum–like structures, and that their constitutive granulosa cells may be luteinized. To test this, indirect immunofluorescence microscopy was done on frozen sections of ovaries from Catnbflox(ex3)/+; Amhr2Cre/+ mice to detect the vascular basement membrane marker type IV collagen. Results showed the lesions to be densely populated with collagen IV–positive basement membranes in a manner similar to normal, highly vascularized corpora lutea and contrasting with normal follicles in which the granulosa cell layer is avascular ( Fig. 5A ). Unlike functional corpora lutea however, in situ hybridization analysis did not detect mRNA encoding the luteinization marker Cyp11a1 within the lesions ( Fig. 5B). Furthermore, immunohistochemical analyses showed some cells within the lesions to be positive for cyclin D2 ( Fig. 4E and F, arrows), a marker of proliferating granulosa cells that is not expressed in the PCNA-positive cells of the corpus luteum ( Fig. 4E, inset). Together, these results suggest that the follicle-like lesions present in the ovaries of Catnbflox(ex3)/+; Amhr2Cre/+ mice are vascularized in a manner similar to corpora lutea but contain slowly proliferating, nonluteinized granulosa cells.
Catnbflox(ex3)/+; Amhr2Cre/+ mice are subfertile. To assess the fertility of female Catnbflox(ex3)/+; Amhr2Cre/+ mice, five ∼6-week-old females were placed in the presence of adult male mice for a period of 6 months. Three Catnbflox(ex3)/+; Amhr2Cre/+ females became pregnant, gave birth normally, and were able to raise pups until weaning. However, the Catnbflox(ex3)/+; Amhr2Cre/+ mice produced significantly (P < 0.05) fewer and smaller litters than observed in Catnbflox(ex3)/+ control animals, resulting in the birth of over 15-fold fewer offspring overall (Supplementary Table). Interestingly, the Catnbflox(ex3)/+; Amhr2Cre/+ females also failed to produce any offspring until the end of the third month of the mating experiment, suggesting either a delay in the onset of puberty or that the mechanism underlying the subfertility phenotype can be overcome with age.
The severity of the observed subfertility phenotype in Catnbflox(ex3)/+; Amhr2Cre/+ mice was somewhat unexpected, as histologic analyses revealed the presence of normal antral follicles and corpora lutea in the ovaries of these mice (e.g., Fig. 3A, f-h; Fig. 4B; and Fig. 5B), suggestive of grossly normal functioning of the ovarian cycle. To evaluate the ovarian response to gonadotropin stimulation more directly, immature Catnbflox(ex3)/+ and Catnbflox(ex3)/+; Amhr2Cre/+ females were given two doses of equine chorionic gonadotropin, which has both FSH and weak LH activity in the mouse. In Catnbflox(ex3)/+ control mice, equine chorionic gonadotropin treatment induced a wave of follicular growth and maturation followed by ovulation and formation of numerous corpora lutea after 96 hours (Supplementary Fig. A and B). In contrast, far fewer corpora lutea were consistently observed in the ovaries of Catnbflox(ex3)/+; Amhr2Cre/+ following equine chorionic gonadotropin treatment, despite the presence of numerous follicles (Supplementary Fig. C and D). These data show that the subfertility phenotype observed in female Catnbflox(ex3)/+; Amhr2Cre/+ mice results, at least in part, from an impaired follicular response to gonadotropins.
Older Catnbflox(ex3)/+; Amhr2Cre/+ mice develop granulosa cell tumors. Acute abdominal distension was observed in a subset of Catnbflox(ex3)/+; Amhr2Cre/+ mice that were part of an experiment designed to evaluate the long-term evolution of the ovarian lesions. This was found to be caused by the formation of unilateral ( Fig. 6A ) or bilateral ovarian tumors, which consisted mostly of blood-filled cysts and were often very large in size (up to 8 cm3). Splenomegaly was also noted in animals bearing large tumors, and this was associated with extensive extramedullary hematopoiesis ( Fig. 6B), presumably secondary to the sequestration of large volumes of blood in the cystic ovarian tumors. Foci of extramedullary hematopoiesis were also observed in the liver and within the tumors themselves in some cases (data not shown).
Microscopic examination revealed the tumors to be GCT, composed of sheets of granulosa cells and granulosa cell–lined cysts of variable sizes ( Fig. 6C). Distinct histologic patterns were observed, the most common of which was a solid sheet of disorganized granulosa cells with randomly distributed, fenestrated antrum-like spaces filled with an eosinophilic material, suggestive of aborted follicles ( Fig. 6C and D, arrowheads). Another pattern consisted of discrete islands and cords of granulosa cells, often featuring centrally located areas of necrosis ( Fig. 6F, arrows). Surprisingly, well-organized areas of ossification and mineralization were commonly found in a subset of tumors that were extensively examined ( Fig. 6H; n = 10 of 11), and this could be readily detected in living mice by radiography ( Fig. 6G, arrows). Immunohistochemistry predictably revealed that the granulosa cell component of these tumors expresses β-catenin although at variable levels and with a generally lesser degree of nuclear localization than seen in the follicle-like pretumoral lesions ( Fig. 6E).
No instance of extraovarian spread of the GCT was detected in Catnbflox(ex3)/+; Amhr2Cre/+ mice (n = 0 of 13), although animals bearing large tumors were sacrificed prematurely for humane reasons, precluding long-term studies. The overall prevalence of GCT formation in Catnbflox(ex3)/+; Amhr2Cre/+ females seemed to increase over time, as no instances were detected before 19 weeks of age (n = 0 of 28) but increased rapidly to 44% at 6 months (n = 4 of 9) and 57% at 7.5 months (n = 8 of 14). Two additional animals that were monitored over a longer term both developed GCT after 12 months, suggesting that most or all Catnbflox(ex3)/+; Amhr2Cre/+ females eventually develop ovarian tumors. Despite the size and rate of growth of the tumors, we did not observe increased mortality in Catnbflox(ex3)/+; Amhr2Cre/+ females bearing GCT.
Transgenic mouse models are considered to be among the most powerful and promising tools presently available for the study of the biology of various forms of cancer and for the development of treatments ( 43, 44). Despite some recent advances ( 45, 46), the development of mouse models of ovarian cancer has lagged behind many other neoplastic diseases. In the case of tumors that arise from the ovarian surface epithelium, this can be attributed in part to our poor understanding of the biology of this particular cell type, leading to difficulties in targeting mutations specifically to epithelial cells ( 43). For GCT, the problem has been further compounded by a dearth of information regarding the genetic lesions involved in GCT development ( 5). Although a few strains of transgenic mice have been shown to develop GCT ( 16, 17, 19, 20), none of these were engineered to mimic a molecular pathogenic process known to be relevant to the etiology of GCT, and as such, their value as animal models for GCT remains unclear. In this report, we show for the first time that a significant subset of GCT feature inappropriate activation of the Wnt/β-catenin signaling pathway, presumably secondary to genetic lesions in one or more of the pathway component genes. To show the importance of misregulated Wnt/β-catenin signaling to the etiology of GCT, we employed the recently developed Amhr2Cre mouse strain to target an activating mutation of the β-catenin gene specifically to granulosa cells, which resulted in the development of GCT with a high penetrance. This study therefore provides significant novel insight into the molecular pathogenesis of GCT, and we propose that the Catnbflox(ex3)/+; Amhr2Cre/+ mouse strain represents the most scientifically relevant model for the study of GCT that has been derived to date.
As stated above, our finding of inappropriate nuclear localization of the β-catenin protein in spontaneously occurring human and equine GCT presumably results from genetic mutation(s) resulting in the stabilization of β-catenin. In other cancers, genetic lesions have been identified in the CTNNB1 (β-catenin) gene that result in hypophosphorylation of the β-catenin protein, but mutations also frequently occur in the APC and AXIN genes, whose protein products form the scaffold that must bind β-catenin to ensure its degradation ( 24). More recently, loss of expression of the Wnt/β-catenin signaling antagonist SFRP1 by an epigenetic mechanism has also been proposed to be involved in the development of colorectal and ovarian cancer ( 47, 48). Future studies will be required to identify genetic or epigenetic defects in the aforementioned genes in spontaneously occurring GCT. In addition to determining the prevalence of genetic lesions in the Wnt/β-catenin pathway in GCT, these studies will serve to better define the molecular events leading to GCT development and possibly correlate specific genetic lesions with clinical variables such as staging and survival.
Another interesting finding reported in the present work is the development of follicle-like lesions in post-pubertal Catnbflox(ex3)/+; Amhr2Cre/+ mice. As the GCT that develop at a later age in these mice apparently arise from these lesions, it would seem reasonable to propose that they could represent a pretumoral state. By analogy, these lesions could be compared with the colonic polyps observed in humans suffering from familial adenomatous polyposis, a disease which results from mutations in the APC gene and that evolves into colonic cancer with a nearly 100% incidence by midlife ( 2). Interestingly, this phenotype was mimicked to some extent in Catnbflox(ex3)/+; Krt1-19Cre/+ and Catnbflox(ex3)/+; Tg-FabplCre mice, which feature the same activating mutation of Catnb used in our GCT model but targeted to the intestinal epithelium rather than granulosa cells ( 34). These mice developed hundreds to thousands of adenomatous intestinal polyps but failed to develop intestinal carcinomas, possibly due to premature death due to anemia and cachexia. In this regard, the relative inefficiency of the Cre-mediated recombination process noted in the Catnbflox(ex3)/+; Amhr2Cre/+ model was fortuitous, as it resulted in a mild phenotype with a limited number of pretumoral lesions that minimally affected the health of the animals, allowing time for progression to the tumoral state to occur. Our finding that the development of GCT in Catnbflox(ex3)/+; Amhr2Cre/+ mice proceeds through a pretumoral intermediate step is a novel concept and has to our knowledge not been reported or proposed to occur in the pathogenesis of GCT in any species. Whereas it is thus quite possible that a pretumoral intermediate is a feature of the human disease, proof of this will require the identification of such lesions in human ovaries, which could be difficult given the low prevalence of GCT.
The reasons underlying the severe subfertility phenotype of Catnbflox(ex3)/+; Amhr2Cre/+ mice remain unclear. Although our observation that fewer corpora lutea form in response to exogenous gonadotropin in the ovaries of immature Catnbflox(ex3)/+; Amhr2Cre/+ mice despite the presence of normal-looking follicles suggests that they do not respond with normal efficiency to hormonal stimuli, the reason(s) underlying this deficiency are unknown. One possibility is that the pretumoral lesions somehow interfere with follicular development and/or luteinization, perhaps by secreting local-acting factor(s). Another cause could be that the early loss of granulosa cells to pretumoral lesion formation may result in the partial depletion of ovarian follicular reserves. Some extent of unintended Cre-mediated genetic recombination may also occur in a subset of oocytes, resulting in nonviable embryos. In addition, the possibility that extraovarian factors contribute to the subfertility phenotype cannot be excluded. For instance, Amhr2 has been shown to be expressed in the gravid uterus of rats ( 49), raising the possibility of a uterine phenotype in Catnbflox(ex3)/+; Amhr2Cre/+ mice, although no gross histopathologic anomalies were noted. Furthermore, the fact that Catnbflox(ex3)/+; Amhr2Cre/+ mice are able to become pregnant and raise pups to the age of weaning suggests the absence of any major uterine or endocrine disruptions. The resolution of these issues will ultimately require additional endocrine-related studies, possibly including systematic follicle counting, gonadotropin measurements, and embryo transfer experiments.
The GCT that arise in Catnbflox(ex3)/+; Amhr2Cre/+ mice bear many histopathologic similarities to the human disease, including the multiplicity of histologic patterns that were found ( 3). Most striking however were the disorganized, follicle-like structures surrounding eosinophilic fluid-filled spaces, which closely resembled the Call-Exner bodies that are a pathognomonic feature of human GCT ( 3). Coupled with the fact that the genetic modifications in Catnbflox(ex3)/+; Amhr2Cre/+ mice mimic the genetic lesions that are presumed to occur in human GCT, these morphologic similarities provide further validation of Catnbflox(ex3)/+; Amhr2Cre/+ mice as a model of GCT that faithfully replicates the human disease. The finding that the pretumoral lesions are vascularized is also an important finding in this respect, as GCTs are highly vascularized tumors that are prone to spontaneous rupture and hemorrhage, which can occasionally endanger the lives of patients ( 3). As the granulosa cell layer of follicles is normally avascular, this finding suggests that activation of the Wnt/β-catenin pathway either promotes vascular growth, or represses a putative mechanism that prevents blood vessels from infiltrating the granulosa cell layer of normal follicles. In either case, the presence of vessels in even small pretumoral lesions indicates that vascular infiltration is an event that occurs early in the development of GCT in Catnbflox(ex3)/+; Amhr2Cre/+ mice and thus possibly in human GCT as well. One apparent discrepancy between the GCT occurring in Catnbflox(ex3)/+; Amhr2Cre/+ mice and human GCT is the presence of foci of apparent osseous metaplasia that were observed in most murine tumors. Ossifying ovarian neoplasms are considered very rare ( 50) and ossification in human GCT has not been reported to our knowledge. The significance of this finding remains unclear, but it could relate to factors that are unique to rodent ovarian physiology.
In summary, this report describes for the first time the role of misregulated Wnt/β-catenin signaling in the development of GCT, providing important insight into the etiology of this poorly understood form of ovarian cancer. Furthermore, the Catnbflox(ex3)/+; Amhr2Cre/+ mouse model described herein will provide cancer biologists and clinical scientists with a powerful tool to study many aspects of GCT biology and disease management, which will hopefully lead to the development of new diagnostic and therapeutic approaches.
Grant support: NIH grants HD16272 (J.S. Richards), HD07495 (J.S. Richards), and HD30284 (R.R. Behringer); Canadian Institutes of Health Research fellowship (D. Boerboom); and Lalor Foundation fellowship (S.P. Jamin).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received March 28, 2005.
- Revision received May 24, 2005.
- Accepted August 2, 2005.
- ©2005 American Association for Cancer Research.