|Home | About | Journals | Submit | Contact Us | Français|
WNT, PI3K or RAS signaling pathways control specific stages of ovarian follicular development. To analyze the functional interactions of these pathways in granulosa cells during follicular development in vivo, we generated specific mutant mouse models. Stable activation of the WNT signaling effector beta-catenin (CTNNB1) in granulosa cells results in the formation of premalignant lesions that develop into granulosa cell tumors (GCTs) spontaneously later in life or following targeted deletion of the tumor suppressor gene Pten. Conversely, expression of oncogenic KRASG12D dramatically arrests proliferation, differentiation and apoptosis in granulosa cells, and consequently, small abnormal follicle-like structures devoid of oocytes accumulate in the ovary. Because of the potent anti-proliferative effects of KRASG12D in granulosa cells, we sought to determine if KRASG12D would block precancerous lesion and tumor formation in follicles of the CTNNB1 mutant mice. Unexpectedly, transgenic Ctnnb1;Kras mutant mice exhibited increased GC proliferation, decreased apoptosis and impaired differentiation and developed early-onset GCTs leading to premature death in a manner similar to the Ctnnb1;Pten mutant mice. Microarray and RT-PCR analyses revealed that gene regulatory processes induced by CTNNB1 were mostly enhanced by either KRAS activation or Pten loss in remarkably similar patterns and degree. The concomitant activation of CTNNB1 and KRAS in Sertoli cells also caused testicular granulosa cell tumors that showed gene expression patterns that partially overlapped those observed in GCTs of the ovary. Although the mutations analyzed herein have not yet been linked to adult GCTs in humans, 1) other components of these pathways may be altered or mutated, 2) these mutations may relate to juvenile GCTs or 3) they may occur in tumors of other tissues where CTNNB1 is mutated. Importantly, our results provide strong evidence that CTNNB1 is the driver in these contexts and that KRASG12D and Pten loss promote the program set in motion by the CTNNB1.
Granulosa cells are the most proliferative cell type in the ovary (Rao MC, Midgley AR et al., 1978; Robison GAR, Butcher RW et al., 1971). Because of this it is surprising that granulosa cell tumors (GCTs; classified among sex-cord/stromal neoplasms) are relatively rare in mammals. This may be determined by apoptosis which is a potent mechanism for eliminating granulosa cells in follicles that fail to differentiate or by luteinization which terminates granulosa cell proliferation (Tilly JL, Tilly KI et al., 1995) and leads to the non-dividing luteal phenotype (Rieske P Pongubala JMR, 2001). Despite their rarity, GCTs remain the most common ovarian cancer subtype in most domestic species. For reasons that are not completely understood, women are particularly prone to developing cancers arising from the ovarian surface epithelium, and for this reason GCTs only represent around 5% of all human ovarian cancer (Jubb I, Kennedy KVF et al., 1993; Schumer ST Cannistra SA, 2003). GCTs can also occur in the testis (Dilworth JP, 1991). Despite the impact of GCTs on domestic species and humans, the molecular mechanisms underlying the etiology of this disease are not yet entirely clear, only a limited number of mouse models have been generated (Edson MA, Nalam RL et al., 2010; Matzuk MM, Finegold et al., 1992; Pangas SA, Li X et al., 2008) and their relevance to domestic and human GCTs remains to be clearly determined.
Recent studies indicate that stage-specific activation of the WNT/FZD/CTNNB1/TCF cascade plays critical roles in controlling normal follicular development and granulosa cell tumor formation (Boyer A, Goff AK et al., 2009). If components of canonical WNT pathway are inappropriately activated or disrupted, granulosa cell fate decisions and follicular growth are dramatically altered (Boerboom D, Paquet M et al., 2005; Boerboom D, White LD et al., 2006; Boyer A, Goff AK et al., 2009; Boyer A, Lapointe E et al., 2010; Lague MN, Paquet M et al., 2008; Vainio S, Heikkila M et al., 1999). Wnt4 null mice provided the first evidence that the WNT/beta-catenin (CTNNB1) canonical pathway exerts potent effects on embryonic gonadal development by suppressing male gonad formation and facilitating ovarian development (Vainio S, Heikkila M et al., 1999). Conversely, expression of a stable, activate, mutant form of CTNNB1 selectively in granulosa cells leads to the formation of precancerous lesions that eventually become granulosa cell tumors (Boerboom D, Paquet M et al., 2005; Boerboom D, White LD et al., 2006). In follicles where granulosa cells escape early transformation but express stable activated CTNNB1, FSH mediated induction of Cyp19a1 and granulosa cell proliferation are enhanced whereas LH mediated granulosa cell differentiation is completely blocked (Fan HY, O'Connor A et al., 2010). Moreover, FSH and other signaling factors canactivate the WNT signaling cascade leading to transcriptional activation of CTNNB1 (Fan HY, O'Connor A et al., 2010; Parakh TN, Hernadez JA et al., 2006). The WNT/CTNNB1 pathway intersects with multiple signaling cascades and transcription factors to dramatically alter cell function during development and in cancer (Boyer A, Goff AK et al., 2009; Jin T, George Fantus I et al., 2008; Wodarz A Nusse R, 1998). For example, FOXL2 acts coordinately with WNT4 to regulate formation of the embryonic ovary and impacts granulosa cell proliferation and differentiation in the postnatal ovary (Boyer A, Lapointe E et al., 2010; Jamieson S, Butzow R et al., 2010; Ottolenghi C, Pelosi E et al., 2007; Uda M, Ottolenghi C et al., 2004; Wang HX, Li TY et al., 2010). Moreover, a specific mutation (C134W) in the human Foxl2 gene has recently been shown to be expressed in nearly all adult GCTs indicating that it plays a critical role in this disease in women (Jamieson S, Butzow R et al., 2010; Kalfa N, Fellous M et al., 2008; Kobel M, Gilks CB et al., 2009; Shah SP, Kobel M et al., 2009).
The RAS pathway is also critical for normal ovarian function (Fan HY Richards JS, 2010). LH induction of the EGF-like factors amphiregulin (AREG), epiregulin (EREG) and betacellulin (BTC) and their activation of the EGF receptor/RAS/ERK1/2 pathway in granulosa cells of preovulatory follicles regulate ovulation, oocyte maturation and the terminal differentiation of granulosa cells to non-dividing luteal cells (Ashkenazi H, Cao X et al., 2005; Conti M, Hsieh M et al., 2005; Fan HY, Liu Z et al., 2009b; Hsieh M Conti M, 2005; Hsieh M, Theologis A et al., 2006; Park J-Y, Su Y-Q et al., 2004). When the KRAS/ERK1/2 pathway is activated selectively in granulosa cells at earlier stages of follicular development by expressing a mutant KRASG12D, the KrasG12D mutant granulosa cells cease dividing and fail to differentiate (Fan HY, Shimada M et al., 2008). Because apoptosis is also blocked, abnormal follicle-like structures persist and accumulate in the ovary leading to premature ovarian failure (Fan HY, Shimada M et al., 2008). Thus, expression of KrasG12D in granulosa cells does not exhibit oncogenic activity, at least in this context. Rather, it causes granulosa cells to exit the cell cycle and this is associated with elevated levels of the tumor suppressor, PTEN (Fan HY, Liu Z et al., 2009a; Fan HY, Shimada M et al., 2008).
Components of the PTEN/PI3K/AKT/FOXO pathway are expressed in granulosa cells and are activated not only by IGF1 but also by FSH, LH and the EGF-like factors (Fan HY, Liu Z et al., 2009b; Gonzalez-Robayna IJ, Falender AE et al., 2000; Richards JS, Sharma SC et al., 2002). Disruption of Pten in granulosa cells reduces apoptosis, increases the number of ovulating follicles and the persistence of luteal cell structures (Fan HY, Liu Z et al., 2008). However, the disruption of Pten alone rarely leads to GCTs (Lague MN, Paquet M et al., 2008), perhaps because the levels of PTEN are relatively low in normal granulosa cells (Fan HY, Liu Z et al., 2008). Conversely, over-expression of a stable active form of FOXO1, a down-stream target of AKT, severely impairs granulosa cell responses to FSH and LH in culture (Liu Z, Rudd MD et al., 2009).
In the ovary, the proliferative effects of a stable mutant form of Ctnnb1 in granulosa cells are markedly enhanced if the tumor suppressor Pten is deleted coordinately in granulosa cells of the Ctnnb1 mutant strain leading to rapid tumor growth (Lague MN, Paquet M et al., 2008). Furthermore, when mutations mutations in Ctnnb1 and Pten are directed to Sertolie cells, GCT formation in occurs in the testis (Boyer A, Paquet M et al., 2009), indicating that Sertoli cells are also susceptible to the combined tumorigenic effects of these two oncogenes.
The potent anti-proliferative effects of mutant KRASG12D in granulosa cells of small follicles appear to be irreversible because proliferation in the KRASG12D mutant granulosa cells is not markedly reversed by the loss of the tumor suppressor Pten in this context (Fan HY, Liu Z et al., 2009a; Fan HY Richards JS, 2010). These results documented that granulosa cells, unlike ovarian surface epithelial cells (Fan HY, Liu Z et al., 2009a), are highly resistant to these specific oncogenic factors. Therefore, we reasoned that KRASG12D might block the proliferative and transformation-dependent effects of CTNNB1 in GCTs of the ovary and testis. Therefore, we sought to determine if tumor formation and growth initiated by mutant CTNNB1 could be altered (suppressed) in granulosa cells by co-expressing KRASG12D.
Additionally, we sought to determine if there were cell- and stage-specific effects in response to expressing mutant CTNNB, KRASG12D or the loss of PTEN. Thus, we have compared gonad morphology, gene expression profiles and functions in mice expressing mutant Ctnnb1 and Kras or mutant Ctnnb1 and loss of Pten driven by mice expressing either Cyp19Cre or Amhr2Cre. Although these mouse models combine mutations that have not yet been identified in human GCTs, the data obtained provide novel evidence on how these pathways may interact in other tissues where CTNNB1 mutations are common.
Previous studies showed that targeted expression of dominant stable CTNNB1 in granulosa cells using the Ctnnb1tm1Mmt (simplified herein to Ctnnb1) and Amhr2tm3(cre)Bhr (A-cre) alleles led to the formation of abnormal follicles characteristic of precancerous lesions (Boerboom D, Paquet M et al., 2005; Boerboom D, White LD et al., 2006). Ultimately, the (Ctnnb1tm1Mmt/+ ;Amhr2tm3(cre)Bhr/+)(Ctnnb1;A-cre) mice developed GCTs after 6 months of age. Because Amhr2Cre is expressed early during embryonic ovarian development, it was hypothesized that a clonal population of cells in the developing gonad was vulnerable to effects of oncogenic CTNNB1. To determine if expressing dominant stable CTNNB1 at later stages of follicular development could also impair granulosa cell functions leading to precancerous lesions and GCTs, we used the Cyp19Cre (Tg(CYP19A1-cre)1jri) (C-Cre) mice in which recombinase activity is first detected approximately at postnatal day 15–20 and is increased in preovulatory follicles (unpublished). The Ctnnb1tm1Mmt/+; Tg(CYP19A1-cre)1jri (Ctnnb1;C-Cre) mice lived for 6–8 months (Figure 1A) but also eventually succumbed to GCTs. The ovaries of these mice exhibited visible precancerous lesions between 4–6 weeks of age (Figure 1C) similar to those observed previously in the Ctnnb1;A-Cre mice (Boerboom D, Paquet M et al., 2005; Boerboom D, White LD et al., 2006). All wild-type (WT) littermates remained viable and exhibited normal ovarian morphology at 4, 6 and 8–12 weeks of age (Figures 1A, C).
To analyze the specific effects of oncogenic KrasG12D versus loss of Pten in the Ctnnb1;C-Cre mutant mice, the Krastm4Tyj/+;Ctnnb1tm1Mmt/+; Tg(CYP19A1-cre)1jri (Ctnnb1;Kras;C-Cre) and Ptentm1Hwu/tm1Hwu;Ctnnb1tm1Mmt/+; Tg(CYP19A1-cre)1jri (Ctnnb1;Pten;C-Cre) mice were generated and checked twice weekly for evidence of tumor growth (distended abdomen) and viability. Because granulosa cells present in the abnormal follicles of the Kras;C-Cre mice are non-proliferative (Fan HY, Shimada M et al., 2008), we predicted that KrasG12D would suppress the proliferative effects of mutant Ctnnb1. Unexpectedly, precancerous lesions appeared in the Ctnnb1;Kras;C-Cre mice between 4–5 weeks of age (Figure 1C) and bilateral tumors were observed by 3 months of age (Fig. 1B). All Ctnnb1;Pten;C-Cre mice died within 2–3 months of age due to tumor burden (Figure 1A; not shown). Histological analyses showed evidence of precancerous lesions by 3 weeks of age and large GCTs by 6–8 weeks of age (Figure 1C). These results are similar to the Ctnnb1;Pten; A-Cre mice that we reported earlier (Lague MN, Paquet M et al., 2008). Although the tumors of the Ctnnb1;Kras;C-Cre mice grew more slowly than those of the Ctnnb1;Pten;C-Cre mice, they grew faster than those in the Ctnnb1;C-Cre mice indicating that the presence of KrasG12D enhances tumor growth mediated by CTNNB1 but does so at a slower rate compared to the loss of Pten.
We next sought to determine the physiological consequences of either expressing KRASG12D or disrupting Pten in the Ctnnb1;C-Cre mutant mice. As shown, serum levels of FSH and LH were elevated in the Ctnnb1;Kras;C-Cre and Ctnnb1;Pten;C-Cre mutant mice at 6 weeks of age compared to WT controls, whereas serum estradiol and progesterone levels were reduced in the mutants (Figure 2A). Despite elevated gonadotropins, granulosa cell differentiation was blocked as indicated by the loss of granulosa cell specific marker genes (Fshr, Lhcgr, Cyp19a1, Cyp11a1 Inha, Amh) in ovaries of the Ctnnb1;Kras;C-Cre and Ctnnb1;Pten;C-Cre mice at 6 weeks (Figure 2B). Granulosa cells of immature mice prior to major tumor formation also failed to respond to exogenous eCG as indicated by the lack of induction of specific granulosa cell marker genes in immature Ctnnb1;Kras;C-Cre and Ctnnb1;Pten;C-Cre mice treated with eCG to induce follicular development (Figure 2C).
To determine if proliferation was altered in the Ctnnb1;Kras;C-Cre and Ctnnb1;Pten;C-Cre granulosa cells compared to WT controls, BrdU uptake and levels of phospho-histone-H3 (pHH3) were examined. To determine if apoptosis was altered TUNEL assays detecting DNA fragmentation and immunostaining for FOXO1, an apoptosis related factor in granulosa cells (Liu Z, Rudd MD et al., 2009) that is absent in the abnormal follicle present in the Kras;C-Cre mice (Fan HY, Shimada M et al., 2008) were performed. FOXO1 and DNA fragmentation were absent/reduced in the GCTs of the Ctnnb1;Kras;C-Cre and Ctnnb1;Pten;C-Cre mutant mice but were detected in growing follicles of WT and Ctnnb1;C-Cre ovaries (Figure 3A). Immunolabeling of BrdU and pHH3 was highest in the ovaries (GCTs) of the Ctnnb1;Pten;C-Cre mice at 5–6 weeks of age, compared to that observed in the Ctnnb1;Kras;C-Cre mutant mice and WT mice Figure 3A and B). Thus, apoptosis is reduced and proliferation is increased in the GCTs of the double mutant mice.
To determine the expression levels of CTNNB1, PTEN and components of the KRAS pathway in the mutant mouse ovaries, ovaries were collected and cell lysates were prepared for Western blot analyses. As shown, levels of CTNNB1 were low in ovaries of immature WT mice compared to ovaries in the Ctnnb1;C-Cre, Ctnnb1;Kras;C-Cre and Ctnnb1;Pten;C-Cre mice (Figure 4). The results also confirmed the expression of the mutant CTNNB1 in the relevant ovaries and that levels of PTEN protein were reduced selectively in the Ctnnb1;Pten;C-Cre ovaries. Levels of phosphoERK1/2 and phospho-P90RSK, down stream targets of KRAS were elevated in the Ctnnb1;Kras;C-Cre and Ctnnb1;C-Cre samples but were low in the Ctnnb1;Pten;C-Cre granulosa cells. Of particular note, the levels of phospho-AKT and phospho-GSK3β were elevated in all samples expressing mutant CTNNB1 (Figure 4B).
To determine specific genes regulated by CTNNB1 in granulosa cells in vivo, microarray analyses were done using RNA samples prepared from ovaries of 4 week old control, Ctnnb1;C-Cre, Ctnnb1;Kras;C-Cre, and Ctnnb1;Pten;C-Cre mutant mice. In silico analyses determined that many genes were up-regulated (619) or down-regulated (203) by expression of dominant stable CTNNB1 alone (Figure 5). Of the 619 up-regulated genes 130 (21%) were specific to CTNNB1 alone, whereas 440 (71%) were up-regulated not only in the Ctnnb1;Pten;C-Cre mutant mice but also in the Ctnnb1;Kras;C-Cre mutant strain. These included CTNNB1 target genes many of which are positive (Wnt4, Wnt2, Wnt6, Wnt5b, Wnt16, Tcf7, Tcf12) as well as negative (Wif1, Notum, Nkd1 and Axin2) regulators of the canonical WNT pathway (Table 1, Figure 6). Of note, two negative regulators of the WNT/CTNNB1 pathway, Ndk2 and Dkk2 were increased in the Ctnnb1;Kras;C-Cre, and Ctnnb1;Pten;C-Cre mutant cells but not in the Ctnnb1;C-Cre mutant cells whereas up-regulation of Ndk1 and Dkk4 were selectively up-regulated in the Ctnnb1;C-Cre mutant cells. Expression of several genes was verified by real-time RT-PCR and in situ hybridization (Figure 6) (Supplemental Figures 1&2). Of the 203 down-regulated genes 22 (11%) were specific for CTNNB1 alone whereas 177 (87%) were also down-regulated in both double mutant strains. Within the latter cluster, many of the genes were either granulosa cell specific (Fshr, Nr5a2, Esr2, Prkar2b, Fst) or oocyte specific (Zp3, Oosp1, Nlrp4, Gdf9, Gpr1) but also included one tumor suppressor (Rassf4) that binds RAS and is associated with RAS-dependent apoptosis (Table 1).
Equally impressive was the high number of genes that were up-regulated in the CTNNB1 expressing cells by either KRASG12D (2136) or the loss of Pten (1977) (Figure 5). Of these, a large number (1362) were up-regulated in both the Ctnnb1;Pten;C-Cre (64%) and Ctnnb1;Kras;C-Cre (69%) mutant ovaries. These results were unexpected given the markedly different effects of expressing KrasG12D (Fan HY, Shimada M et al., 2008) or disrupting Pten alone (Fan HY, Liu Z et al., 2008) or in combination in granulosa cells (Table 1) (Fan HY, Shimada M et al., 2008). A notable number of up-regulated genes were related to RAS signaling including the Ret and Met proto-oncogenes, Etv5 a potential regulator of RET (Tyagi G, Carnes K et al., 2009), Epidermal growth factor receptor pathway substrate 8, Eps8 (Xu M, Shorts-Cary L et al., 2009), Rrad, Rassf8, several guanine nucleotides exchange factors (Arhgef2, 3, 5, 12 and Rapgef4, also known as Epac) and Rab family members (Rab6, 20, 24) involved in intracellular trafficking. Components of the PI3K pathway (Pik3c3, Pik3r3) were increased ~3-fold and protein kinase C, beta (Prkcb) was elevated ~30-fold (Table 1). The parallel up-regulation of components of the RET/MET/RAS and PI3K pathways by KRASG12D or loss of Pten underscores the overlapping impact of these mutations and indicates that these two pathways are tightly linked in these cancer cells.
An impressive number of genes were also down-regulated and showed extensive overlap between in the Ctnnb1;Kras;C-Cre (83%) and Ctnnb1;Pten;C-Cre (82%) mutant ovaries. Most genes down-regulated in this cluster reflected the loss of granulosa cell markers, including several components of the IGF1 pathway (Igf1, Igf1r, Irs2, Socs2, Foxo1) and the inhibin/activin/Bmpr1b pathway. Because the mutant ovaries lack corpora lutea and are essentially devoid of oocytes, markers of these cell types were also markedly reduced (Table 1). However, three of the most highly down-regulated genes were osteoglycin (Ogn) that encodes a blocker of metastasis, leucocyte cell derived chemotaxin (Lect1) that encodes a potent blocker of angiogenesis and Uchl1 that encodes a ubiquitin carboxy-terminal hydrolase that is associated with cancer progression in a cell context specific manner (Fellenberg J, Lehner B et al., 2010; Hussain S, Foreman O et al., 2010). Other genes of particular interest that were increased in the tumors include Apcdd1, Enpp2, Etv1, Fst, Cnr1, Sox7 and Tbx3 (Table 1, Figure 6).
As male Ctnnb1;Pten;A-Cre male mice develop granulosa cell tumors of the testis (GCTT)(Boyer A, Paquet M et al., 2009), we next sought to determine if activated KRAS and dominant-stable CTNNB1 could combine to cause GCTT in the Ctnnb1;Kras;A-Cre model. Male Ctnnb1;A-Cre mice developed a seminiferous tubule degeneration phenotype by 5 weeks of age, resulting in progressive loss of spermatogenesis, testicular atrophy with reduced testis size and sterility (Figure 7A, B, D; Table 2), which was consistent with previous reports (Boyer A, Hermo L et al., 2008; Tanwar PS, Zhang L et al., 2010). Whereas male Kras;A-Cre mice were phenotypically normal and had grossly unaltered fertility and spermatogenesis (Fig 7A–C, Table 2 and data not shown), Ctnnb1;Kras;A-Cre mice had a degenerative phenotype similar to that observed in the Ctnnb1;A-Cre model, but with a somewhat earlier onset (~ 4wks). Furthermore, GCTT that were histologically indistinguishable from those observed in the Ctnnb1;Pten;A-Cre model were found in most Ctnnb1;Kras;A-Cre testes after 4–5 months of age (Fig 7A, E, F). GCTTs in Ctnnb1;Kras;A-Cre mice grew more slowly than those of Ctnnb1;Pten;A-Cre mice, resulting in testis/tumor weights at 4–6 months of age that were similar to those of Ctnnb1;Pten;A-Cre mice at 5 weeks of age (Fig 7A&B and Table 2). The tumors did not spread or metastasize, and did not compromise viability or animal well-being up to 8 months of age (not shown). The diagnosis of GCTT in the Ctnnb1;Kras;A-Cre testes was supported by the detection of ectopic expression of FOXL2 (Fig 7G) and Wnt4 (Fig 8), both markers of early granulosa cell differentiation (Ottolenghi C, Pelosi E et al., 2007). FOXL2 expression was in fact comparable to levels expressed in ovarian tissue, as had been previously observed in the Ctnnb1;Pten;A-Cre model.
Because GCTs develop in the ovaries and testes of the Ctnnb1;Pten;A-Cre mutant mouse strain (Boerboom D, Paquet M et al., 2005; Boerboom D, White LD et al., 2006; Boyer A, Goff AK et al., 2009; Boyer A, Paquet M et al., 2009; Lague MN, Paquet M et al., 2008), we used these mice to determine if the tumors in the ovary and testis express similar or distinct genes and if expression of KrasG12D would impact the growth of GCTs in the Amhr2-Cre expressing mouse strains. Ovaries and testes were collected from control, Ctnnb1;A-Cre, Kras;A-Cre, Pten;A-Cre, Ctnnb1;Pten;A-Cre and Ctnnb1;Kras;A-Cre mutant mice at distinct ages when tumor sizes were similar (based on published and preliminary studies). Ovarian and testis weights were determined (Table 2) and the tissues used to prepare RNA. Real time RT-PCR was used to determine the expression of selected genes that were prominently regulated on the microarrays and in tumors of the C-Cre mutant strains (Figures 5 and and6).6). As shown, many CTNNB1 target genes were slightly or significantly up-regulated in ovaries of the Ctnnb1;A-Cre mice and were dramatically increased further when either the Pten gene was depleted or KrasG12D was expressed in the presence of mutant CTNNB1 (Figure 8). These results confirm those obtained in the C-Cre strains. Two genes (Nr5a2, Nr0b1) were reduced in ovaries of the double mutant mice and three (Foxo3a, Tcf12 and Wnt4) showed no change. When the same genes were analyzed in the testis samples, several CTNNB1 target genes (Axin2, Enpp2, Etv1, Nfat5, Ndk1, Tcf12 and Wnt4) were increased in the Ctnnb1;A-Cre alone mutant testes. Several of these genes were also increased in the Ctnnb1;Pten;A-Cre mice whereas expression of Axin2, Ndk1 and Peg3 increased most dramatically in the Ctnnb1; Kras;A-Cre mutant mice. The increased expression of Foxo3 and Wnt4 and the lack of expression of Cd83 and Wnt16 in testes of the Ctnnb1;A-Cre mice were notably different from the expression patterns of these genes observed in the ovaries of mice of the same genotype.
These studies document for the first time that oncogenic KRASG12D as well as loss of Pten promote the altered genetic program set in motion by dominant stable CTNNB1 that drives granulosa proliferation and cell fate decisions (Boerboom D, White LD et al., 2006; Boyer A, Paquet M et al., 2009). Most impressive and unexpected were the phenotypic data showing that oncogenic KRASG12D and loss Pten exert remarkably similar effects in magnitude and direction, leading to early granulosa cell tumor formation and growth in ovaries of both the Ctnnb1;C-Cre and Ctnnb1;A-Cre mouse strains. These results were totally unexpected based on the completely opposite effects of expressing KRASG12D alone (Fan HY, Shimada M et al., 2008) compared to disrupting Pten alone (Fan HY, Liu Z et al., 2008) in granulosa cells. Whereas premature expression of KRASG12D in granulosa cells of small follicles causes granulosa cell cycle arrest and elevated levels of PTEN in these cells (Fan HY, Shimada M et al., 2008), loss of Pten enhances granulosa cell proliferation, leading infrequently to GCTs (Fan HY, Liu Z et al., 2008; Lague MN, Paquet M et al., 2008). Therefore, we predicted that expression of KRASG12D would block or reduce granulosa cell proliferation and the appearance of precancerous lesions observed in the in the Ctnnb1;A-Cre and Ctnnb1;C-Cre mutant mice. However, contrary to this hypothesis, the Ctnnb1;Kras;A-Cre and Ctnnb1;Kras;C-Cre mutant mice develop precancerous lesions similar to those observed in the Ctnnb1;A-Cre and Ctnnb1;C-Cre mouse strains. Furthermore, GCTs develop in these mice with 100% penetrance. Although the growth rate of the GCTs in the Ctnnb1;Kras;A-Cre and Ctnnb1;Kras;C-Cre mutant mice is slower than in the Ctnnb1; Pten;C-Cre and Ctnnb1;Pten;A-Cre mice, the phenotypic outcomes are ultimately similar. Collectively, these results indicate that the potent mechanisms by which KRASG12D alone acts to block or terminate granulosa cell proliferation in small antral follicles is reversed and/or superceded by events initiated and driven by dominant stable CTNNB1. Moreover, because ovaries of the Ctnnb1;Pten;A-Cre mice expressed reduced levels of KRAS and ERK1/2 phosphorylation compared to the Ctnnb1;C-Cre and Ctnnb1;Kras;C-Cre mice, it appears that the presence of KRAS and ERK1/2 reduces the rate of tumor growth but cannot completely block proliferation as observed when KRASG12D is expressed alone in the Kras;C-Cre mice (Fan HY, Shimada M et al., 2008).
Equally impressive, and supporting the phenotypic outcome of the different Ctnnb1 mutant strains, was the high degree of similarity among the gene expression profiles of the precancerous lesion- and tumor- bearing ovaries in the Ctnnb1;Kras;C-Cre and Ctnnb1;Kras;A-Cre and Ctnnb1;Pten;C-Cre and Ctnnb1;Pten;A-Cre mutant female mice. Many of the genes up-regulated by mutant CTNNB1 alone in the precancerous ovaries were expressed at even higher levels in ovaries of the Ctnnb1;Kras;C-Cre and Ctnnb1;Kras;A-Cre and Ctnnb1;Pten;C-Cre and Ctnnb1;Pten;A-Cre mutant mice. Notable among these were negative regulators of the WNT/CTNNB1 pathway observed in our previous studies, including Wif1, Axin2 and Ndk1 (Boerboom D, White LD et al., 2006; Fan HY, O'Connor A et al., 2010). Additional CTNNB1 target genes, identified in the more extensive microarray analyses presented in this study, include the positive regulators, such as Wnt16, as well as the negative regulators such as Apcdd1, Enpp2, Etv1, Peg3, Nkd2, Notum and Tbx3 of the canonical WNT pathway.
APCDD1 (adenomatosis polyposis coli down-regulated 1) is a membrane-bound glycoprotein interacts with WNT3 and LRP5 to inhibit WNT signaling (Shimomura Y, Agalliu D et al., 2010; Takahashi M, Fujita M et al., 2002). In colorectal cancer cells, up-regulation of the Apcdd1 gene promotes cell proliferation (Takahashi M, Fujita M et al., 2002). Based on the dramatic induction of Apcdd1 expression in GCTs of the ovary and testis, it may be a key driver of CTNNB1 in granulosa cell tumor formation and growth. Enpp2 encodes ectonucleotide pyrophosphatase/phosphodiesterase family member 2 (autotaxin) that is important for lipid signaling, proliferation and cell migration (Tania M, Khan MdA et al., 2010). The Ets variant transcription factor Etv1 impacts gastrointestinal tumor formation where it is regulated by a KITL/MAPK1/3 pathway (Chi P, Chen Y et al., 2010). Etv5 is expressed in granulosa and Sertoli cells and regulates Ret expression and signaling in the testis (Eo J, Han K et al., 2008; Tyagi G, Carnes K et al., 2009).
Because Etv5 is a CTNNB1 target gene that increases expression of Ret, the increase in Etv5 likely leads to enhanced activation of this pathway and tumor growth as in other tissues (Tyagi G, Carnes K et al., 2009). Cnr1 and Tbx3 are CTNNB1 target genes known to impact proliferation and pluripotency and thus may also contribute to the GCT formation (Trazzi S, Steger M et al., 2010; Wagner RT, Xu X et al., 2010). Peg3 is a maternally imprinted gene that has the potential to inhibit apoptosis (Broad KD, Curley JP et al., 2009). CD83 and NFAT5 are normally related to immune-cell functions (Breloer M Fleischer B, 2008; Drews-Elger K, Ortells MC et al., 2009). Their roles in GCTs remain unknown.
The data generated using the two different mouse strains expressing Cre recombinase document unequivocally that CTNNB1 can alter granulosa cell fate decisions and differentiation at early (Amhr2Cre) as well as at later (Cyp19Cre) stages of follicle development by blocking the expression of many granulosa cell specific marker genes, such as Cyp19, Nr5a2, Lhcgr and Fshr (Figures 5, ,6,6, ,8)8) (Boerboom D, White LD et al., 2006; Fan HY, O'Connor A et al., 2010). However, the mutant granulosa cells continue to express markers characteristic of granulosa cells at early stages of gonadal development and follicle formation, such as Wnt4 and Foxl2. These results suggest that the CTNNB1 expressing granulosa cells are locked into a specific, primordial stage of differentiation. That these CTNNB1 expressing granulosa cells are highly susceptible to transformation by either KrasG12D or loss of Pten, indicates that each potently and rapidly enhances the effects of CTNNB1 in these cells. These results provide strong evidence that if CTNNB1 or other genetic lesions that directly up-regulate the canonical WNT pathway are mutated or deleted, CTNNB1 is likely to be the strong driver irrespective of what other mutations occur. Moreover, many of the genes that were expressed at elevated levels in the GCTs of the ovary and/or testis are known to be targets of CTNNB1 in other cancer cell types and impact cancer cell growth and invasiveness (Chi P, Chen Y et al., 2010; Harper K, Arsenault D et al., 2010; Heinrich MC Corless CL, 2010; Jane-Valbuena J, Widlund HR et al., 2010; Jiang X, Yu Y et al., 2010; Shimomura Y, Agalliu D et al., 2010; Takahashi M, Fujita M et al., 2002; Tania M, Khan MdA et al., 2010; Wagner RT, Xu X et al., 2010).
The molecular mechanisms by which expression of KrasG12D and loss of Pten exert such similar effects on the genetic program initiated by stable CTNNB1 remain to be determined. However, increased phosphorylation of CTNNB1 by activated AKT and the loss of FOXO1 may both contribute to enhanced transcriptional activity of CTNNB1 (Boyer A, Goff AK et al., 2009). Furthermore, AKT phosphorylation was increased in both mutant models and expression of specific components of the PI3K and RAS signaling pathways were increased to a similar extent in tumor of the Ctnnb1;Kras and Ctnnb1;Pten mutant mice indicating that these two pathways converge in the presence of active CTNNB1. This convergence may involve the up-regulation of the proto-oncogenes Ret and Met that occurred in the double mutant mice and may be mediated by phosphorylation and increased expression of Etv5 and/or Etv1.
An important issue raised by our findings is their potential relatedness to the pathogenesis of adult and juvenile forms of GCTs in women. It has recently been shown that the vast majority of adult-form GCTs in women bear a common mutation (C134W) in the FOXL2 gene, suggesting a pivotal role for FOXL2 in one or more phases of disease development (Jamieson S, Butzow R et al., 2010; Shah SP, Kobel M et al., 2009). However, it remains to be determined if and how the disease processes initiated by the manipulation of the WNT/CTNNB1, PI3K/AKT and/or KRAS pathways in our models relates to the mutant FOXL2-driven human disease. Although our data seem to indicate that Foxl2 mRNA levels are not meaningfully altered in the various transgenic models, FOXL2 activity may be modulated by other means, such as post-translational modifications or cofactor binding. The biological activities of mutant FOXL2 are as yet poorly understood, and only a handful of studies have begun to elucidate the mechanisms by which it may contribute to tumorigenesis (Fleming NI, Knower KC et al., 2010; Kim JH, Yoon S et al., 2011; Lee K, Pisarska MD et al., 2005). It is plausible that mutant FOXL2 initiates cellular processes resulting in the hyperactivation of the WNT/CTNNB1, PI3K/AKT and/or KRAS pathways that may enhance tumor development. For example, FOXL2 and WNT4 are critical for granulosa cell specification, both regulate aromatase and both promote follicle development (Boyer A, Lapointe E et al., 2010; Fan HY, O'Connor A et al., 2010; Fleming NI, Knower KC et al., 2010; Garcia-Ortiz JE, Pelosi E et al., 2009; Moumne L, Batista F et al., 2008; Uda M, Ottolenghi C et al., 2004). If this also occurs in adult GCTs, the mutant cells may be more responsive to WNT/CTNNB1 activation. In addition, KGN cells that are derived presumably from an adult GCT express the FOXL2 C134W mutant and exhibit constitutive activation of ERK1/2 (Steinmetz R, Wagoner HA et al., 2004). Chemical or siRNA disruption of ERK1/2 blocked proliferation indicating that activation (not a mutation) of a RAS-ERK1/2 cascade is critical for the proliferative potential of these cells. Because RAS can activate the PI3K pathway, this pathway may also contribute to the FOXL2 mutant phenotype. Hence it is not yet clear if mutant FOXL2 is sufficient or necessary to drive GCT formation. Another possibility is that the molecular mechanisms involved in tumor development in our mouse models are partially or completely unrelated to FOXL2. In this regard, the juvenile form of GCT in humans is characterized by an earlier (often prepubertal) onset, is not associated with FOXL2 mutations (Jamieson S, Butzow R et al., 2010; Shah SP, Kobel M et al., 2009) and the molecular mechanisms controlling this GCT form are ill-defined (Jamieson S, Butzow R et al., 2010). It therefore remains to be determined if the mechanisms involved in juvenile GCT development are more closely related to those that are activated in the mouse GCT models described herein, and therefore that the latter could be thought of as being more analogous to the juvenile rather than the adult human GCT disease. Clearly, many more studies are needed to clarify the drivers and helpers involved at various phases of this complex disease in the ovary.
Based on the gene profiling results in the GCTs of the ovary and the observations that cell morphology and tumor architecture are similar in GCTs of the ovary and testis (Boyer A, Paquet M et al., 2009), we anticipated similar patterns of gene expression in the GCTs of the testis and ovary. However, there are distinct as well as similar gene expression patterns between the GCTs in ovary and testis. Because the gene expression profiles of tumors in the Ctnnb1;Kras;A-Cre male mice are more similar to those of the ovarian GCTs than are the profiles in the Ctnnb1;Pten;A-Cre strain, tumors in the Ctnnb1;Kras;A-Cre male mice may appear to be more differentiated towards the granulosa cell phenotype than the tumors of Ctnnb1;Pten;A-Cre male mice. The higher levels of Wnt4 and Foxl2 mRNA expression in the Ctnnb1;Kras;A-Cre testes are consistent with the idea that these cells have assumed a primordial granulosa cell-like genetic program. Their slower growth rate likely relates to the effects of KRASG12D that impair granulosa cells proliferation in the ovary. Because Wnt4 and Foxl2 but not Wnt16 are induced in the GCTs of the testis, these factors may serve as critical positive regulators, perhaps with some functions redundant of WNT16 in this context.
Explanations for the major differences between the Ctnnb1;Pten;A-Cre and Ctnnb1;Kras;A-Cre gene expression profiles in the GCTs of the testis are not completely obvious and are somewhat surprising given the dramatic similarities in the GCTs in the ovaries of mice with these genotypes. However, the differences may relate to when mutant CTNNB1 is first expressed as well as the time of onset and speed of tumor development. For example, tumors in the 5 week Ctnnb1;Pten;A-Cre model are approximately the same size as the tumors in the 5 month old Ctnnb1;Kras;A-Cre mice. Although these times were selected to compare early growth stages of tumor development across genotypes, selection criteria based on size may overlook critical molecular events being driven by the loss of Pten versus the expression of KRASG12D at specific stages of cell differentiation. Although cell morphology appears similar in both models, cell morphology is not always an indicator of molecular changes. The lack of more dramatic changes in gene expression in the male GCTs may also reflect the presence of more testicular tissue associated with the tumors than in the ovaries. This may “dilute out” some of the genes that are prominently expressed in the tumor cells.
In summary, our results show that stable expression of CTNNB1 in granulosa or Sertoli cells alters their genetic program and predisposes them to oncogenic transformation by either expression of mutant KRASG12D or loss of PTEN. Because KRASG12D alone is a potent suppressor of granulosa cell proliferation, these oncogenic effects of KRAS G12D were unexpected. Moreover, the striking similarities in the gene expression patterns in the GCTs of the ovary and testis when either mutant KRASG12D is expressed or Pten is lost in the mutant CTNNB1 cells provide strong evidence that CTNNB1 is the driver in these contexts and that KRASG12D and Pten loss promote the program set in motion by the CTNNB1. Although the mutations analyzed herein have not yet been linked to adult GCTs in humans, they may be related to juvenile GCTs or to tumors in other tissues where CTNNB1 is mutated. In this regard, these results may have clinical relevance and help explain why tumor growth is so rapid and pervasive in tissues where mutations in Ctnnb1 as well as alterations in the RAS or PI3K pathways are common.
Immature C57BL/6 mice were obtained from Harlan, Inc. (Indianapolis, IN). Animals were housed under a 14:10h, light:dark schedule and were treated in accordance with the NIH Guide for the Care and Use of Laboratory Animals. Mice expressing Ctnnb1tm1Mmt/Ctnnb1tm1Mmt (Boerboom D, Paquet M et al., 2005), Krastm4Tyj/ Krastm4Tyj (L-S-L-KrasG12D) (Tuveson DA, Shaw AT et al., 2004),Ptentm1Hwu/ Ptentm1Hwu (Ptenfl/fl) (Li G, Robinson GW et al., 2002),were used to generate mice harboring granulosa cells specific mutations of each gene alone or in combination using the two cell-specific Cre recombinase strains of mice Amhr2tm3(cre)Bhr/+) Amhr2-Cre (Jamin SP, Arango NA et al., 2002) and Tg(CYP19A1-Cre)1jri (Cyp19-Cre) (Fan HY, Shimada M et al., 2008). All the mutant mouse strains are in the C57BL/6 background.
Ovaries were fixed in 4% paraformaldehyde, embedded in Optional Cutting Temperature compound (Sakura Finetek USA Inc.) and stored at −70°C before the preparation of 7 µm sections. Serial sections were stained with hemotoxylin and eosin according to established procedures (Fan HY, O'Connor A et al., 2010). The TUNEL assays were done using the ApopTagPlus apoptosis detection kit (Chemicon International. Temecula, CA) as reported previously (Fan HY, Liu Z et al., 2008). Sections were also probed with primary antibodies to specific proteins (FOXO1, BrdU and phosphohistone H3 from Cell Signaling Technology, Danvers, MA); and (BrdU from Sigma Chemical Co, St. Louis, MO) as indicated in the text and secondary Alexa Fluor 594- or 488-conjugated goat anti-rabbit IgG antibodies (Molecular Probes) as previously described (Fan HY, Liu Z et al., 2009b). Slides were mounted using VectaShield with DAPI (Vector Laboratories). Digital images were captured using a Zeiss Axioplan 2 microscope with 5–63× objectives. For all the experiments, exposure time was kept the same for control and mutant samples. Proliferation was analyzed by BrdU incorporation into cells of mice injected with 50mg/kg BrdU in saline 2 hr prior to sacrifice. Ovaries were embedded as above, sections and immno-stained for BrdU (Sigma Chemical Company, St. Louis, MO).
Cell extracts containing 30µg protein were resolved by SDS-PAGE and transferred to PVDF membranes (Millipore Corp., Bedford, MA) and analyzed as previously (Fan HY, Liu Z et al., 2009b) using primary antibodies to CTNNB1 (from Santa Cruz Biotechnology, Inc.) and PTEN, phospho90RSK1/2/3, phospho-ERK1/2, phosphoAKT, phosphoGSK3β, FOXL2 (from Cell Signaling Technology) and ACTIN (from Cytoskeleton, Inc, Denver, CO) at 1:1000 dilutions or as indicated in the Figure legend.
Total RNA was isolated using the RNeasy Mini kit (Qiagen Sciences, Germantown, MD). RNA quality was assessed and then riboprobes were generated from WT and mutant RNA and hybridized to Mouse 430.2 microarray chips (Affymetrix, Santa Clara, CA) in the Microarray Core Facility of the Baylor College of Medicine as previously described (Hernandez-Gonzalez I, Gonzalez-Robayna IJ et al., 2006). Microarray data were analyzed as previously reported using the Robust Multi-array Averaging function (Irizarry RA, Hobbs B et al., 2003) from the Affy package (v1.5.8) through the BioConductor software (http://www.bioconductor.org/ (Gentleman RC, Carey VJ et al., 2004)). The microarray data have been deposited to GEO; the accession number is GSE27656. http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=blczregkwwekmrw&acc=GSE27656
Reverse transcription was done using the SuperScript One-Step RT-PCR system with Platinum Taq kit (Invitrogen, Carlsbad, CA). The QPCR was performed using the Rotor-Gene 3000 thermocycler (Corbett Research, Sydney, Australia). Relative levels of mRNAs were calculated using Rotor-Gene 6.0 software and normalized to the levels of endogenous beta-actin in the same samples.
The data are represented as means ± SEM. QPCR data is expressed at the ratio of WT (n=1) to mutant. Data were analyzed by using GraphPad Prism Programs (ANOVA or t-test; GraphPad Prism, San Diego, CA) and Dunnett’s post-hoc test after ANOVA to compare all genotypes to the control. Values were considered significantly different if P≤0.05 or P≤0.01.
The authors thank the Microarray Core Facility at Baylor College of Medicine (Lisa K White, director) for performing the microarray analyses. Supported, in part, by NIH-HD-16229 and the Eunice Kennedy Shriver NICHD/NIH cooperative agreement [U54-(HD07945) as part of the Specialized Cooperative Centers Program in Reproduction and Infertility Research (SCCPIR) (JSR), an operating grant from the Canadian Institutes of Health Research (DB) and a Discovery Grant from the National Sciences and Engineering Research Council (DB).
Conflict of Interest: The authors have nothing to declare.