Selective expression of Kras^G12D in granulosa cells of the mouse ovary causes defects in follicle development and ovulation

Activation of small G-proteins within the RAS superfamily impact multiple downstream signaling cascades, including RAF1/MEK/ERK1/2 and PI3K/AKT/FOXO, in many tissues in a cell- and context-specific manner(Campbell et al., 2007; Cespedes et al., 2006; Gupta et al., 2007; Rocks et al., 2006). In response to growth-regulatory molecules, transient activation of RAS can stimulate controlled proliferation as well as differentiation of cells. Uncontrolled activation of RAS is often associated with oncogenic transformation or senescence of cells(Jackson et al., 2001; Lin et al., 1998; Serrano et al., 1997; Shaw et al., 2007). Specifically, RAS family members become oncogenic by single-point mutations,mainly at codons 12 or 13 (Bourne et al.,1990), leading to constitutive signaling and cell transformation with changes in morphology, increased proliferation and/or inhibition of apoptosis. Tissue-specific activation of oncogenic Kras^G12D causes mammary gland, lung and endometrioid ovarian carcinoma in mouse (Dinulescu et al., 2005; Jackson et al.,2001; Sarkisian et al.,2007). Mutations of KRAS or BRAF in non-invasive and invasive carcinomas of the ovary [involving the ovarian surface epithelium(OSE)] have been reported (Gemignani et al., 2003; Mayr et al.,2006).

In this study, we sought to determine the impact of RAS activation in granulosa cells in vivo. The constitutively active Kras^G12D mutation(Johnson et al., 2001) was selectively expressed in mouse granulosa cells using a Cre-mediated DNA recombination approach. The inappropriate, premature expression of Kras^G12D in granulosa cells blocked granulosa cell differentiation at an early stage, leading to the formation of abnormal follicle-like structures containing non-mitotic, non-apoptotic,non-differentiated and non-tumorigenic cells. Moreover, those follicles that reached the antral stage exhibited impaired responses to hormones, leading to ovulation failure. Thus, transient but not sustained activation of RAS in granulosa cells is crucial for normal follicle growth and successful completion of the ovulation process.

LSL-Kras^G12D;Amhr2-Cre mice were derived from previously described Amhr2-Cre and LSL-Kras^G12Dparental strains (Jamin et al.,2002; Tuveson et al.,2004). Although Amhr2 is highly expressed in granulosa cells of growing follicles it is also known to be expressed in other reproductive tissues, including ovarian surface epithelial cells and the uterus (Arango et al., 2008)(our unpublished observations). Therefore, we sought to obtain a Cre-expressing mouse model that would be more highly specific for granulosa cells. Cyp19-Cre transgenic mice were generated by oocyte microinjection of a DNA fragment in which the 304 bp Cyp19 promoter(GenBank S85356, bp -278 to +26) was ligated to iCre cDNA. To study ovarian responses to exogenous gonadotropins, 21-day-old immature females were analyzed to avoid the complexity of ovarian functions associated with estrous cycles and endogenous surges of gonadotropins. Specifically, immature mice were injected intraperitoneally (ip) with 4 IU eCG (equine chorionic gonadotropin; Calbiochem) followed 48 hours later with 5 IU hCG (human chorionic gonadotropin; American Pharmaceutical Partners, Schaumburg, IL). Ovulated COCs were collected from oviducts 16 hours after hCG injection. Animals were treated in accordance with the NIH Guide for the Care and Use of Laboratory Animals.

Undifferentiated granulosa cells were released from mouse antral follicles by puncturing with a 26.5-gauge needle. Cells were cultured at a density of 1×10⁶ cells/ml in defined medium (DMEM:F12 containing penicillin and streptomycin) in 12-well culture dishes. Cells were cultured overnight to allow attachment to the culture dish and were infected with adenoviral vectors expressing Cre (Ad5-CMV-Cre, generated by the Vector Development Laboratory, Baylor College of Medicine) or GFP as the infection control. The infected cells were treated with or without FSH [NIH-FSH-16,National Hormone and Peptide Program (Al Parlow), Torrance, CA; 100 ng/ml] for time intervals designated in the figure legends.

Mice were injected ip with 50 mg/kg of BrdU in PBS, and were killed 2 hours later. Ovaries were isolated and fixed with 4% paraformaldehyde (PFA)overnight. Incorporated BrdU was detected by immunohistochemistry using BrdU antibody according to manufacturer's instructions (Sigma, St Louis, MO). TUNEL assays were performed on PFA-fixed paraffin-embedded sections using the ApopTag Plus Peroxidase In Situ Apoptosis Detection Kit (Serologicals Corporation, Norcross, GA) according to manufacturer's instructions.

Immunohistochemistry was performed on 4% PFA-fixed paraffin-embedded 5-μm sections using the VectaStain Elite Avidin-Biotin Complex Kit as directed by the manufacturer (Vector Labs, Burlingame, CA). Sections were probed with primary antibodies against FOXO1 or PCNA (Cell Signaling, CA) and visualized using a 3,3V-Diaminobenzidine Peroxidase Substrate Kit (Vector Labs). For immunofluorescence, ovaries were PFA fixed, embedded in OCT compound (Sakura Finetek USA, Torrance, CA) and stored at -80°C before sectioning. Sections were probed with anti-KRAS (Santa Cruz Biotechnology,Santa Cruz, CA), anti-phospho-AKT, anti-phospho-ERK1/2, anti-cleaved caspase 3, or anti-phospho-histone H3 (Cell Signaling Technology) antibodies and visualized with Alexa Fluor 594-conjugated goat anti-rabbit IgG (Molecular Probes, Eugene, OR). Digital images were captured using a Zeiss Axiphot microscope with 5-40× objectives. For all the experiments, exposure time was kept the same for control and Kras mutant samples.

Plasmids for Nr5a2 and Cyp11a1 probes were as described previously (Boerboom et al.,2005; Falender et al.,2003). A cDNA fragment of Mkp3 was amplified by RT-PCR from mouse ovary total cDNA and subcloned into the pCR-TOPO4 vector(Invitrogen, Carlsbad, CA). In situ hybridization was performed as previously reported (Falender et al.,2003; Hsieh et al.,2005). Tissue histology and the radioactive probe were visualized under light- and dark-field illumination, respectively.

Reverse transcription (RT)-PCR was performed using the SuperScript One-Step RT-PCR System with the Platinum Taq Kit (Invitrogen) and 100 ng samples of ovarian total RNA that had been isolated using the RNeasy Mini Kit (Qiagen,Germantown, MD). Approximately 0.625 μCi of [α-³²P]dCTP(3000 Ci/mmol; MP Biomedicals, Irvine, CA) were added to each reaction to generate radioactive signals. Primer sequences and amplification conditions used are available upon request. Samples were separated by electrophoresis on 5% PAGE gels, dried and exposed to Biomax XAR film (Eastman Kodak, Rochester,NY) to generate the presented images.

Quantitative (q) RT-PCR was performed using the Rotor-Gene 3000 thermocycler (Corbett Research, Sydney, Australia). Relative levels of gene expression were normalized to β-actin.

The RAS-binding domain (RBD) of the mouse PI3K p110α subunit (PIK3CA;aa 220-311) (Rodriguez-Viciana et al.,1996) and of mouse RAF1 (aa 55-131)(Campbell-Valois and Michnick,2007) were PCR amplified from a mouse ovary cDNA pool and subcloned into pGEX 4T1 vector. Recombinant GST-PI3K RBD and GST-RAF1 RBD were expressed in the Rosetta-pLysS E. coli strain (Novagen) and affinity purified using glutathione-agarose beads (Sigma). Ovaries were homogenized in lysis buffer (20 mM NaF, 10 mM MgCl₂, 100 mM NaCl, 10% glycerol,0.5% Triton X-100, 20 mM HEPES, pH7.5). The lysates were incubated with agarose slurries linked with mouse RAF1 RBD or PI3K RBD to bind RAS-GTP. The agarose beads were washed and resuspended in Laemmli sample buffer prior to western blot analysis.

Western blots were performed utilizing 30 μg of lysate protein, 10%SDS-PAGE gels and transfer to Immobilin membrane (Millipore). Membranes were incubated with the following antibodies at 1:1000 dilutions:anti-phospho-FOXO1, anti-phospho-ERK1/2, anti-phospho-AKT, anti-AKT (all from Cell Signaling Technology) and anti-RAS (Upstate Biotechnology).

Mkp3 siRNA (sc-39001) was purchased from Santa Cruz Biotechnology. Scrambled siRNA duplex (Ambion) was used as control. Transfection of siRNA (50 nM) into cultured granulosa cells was accomplished using the HVJ Envelope Vector Kit (Ishihara Sangyo, Tokyo, Japan) as previously reported(Shimada et al., 2007). The culture medium was replaced 5 hours after transfection and the cells were treated with 250 ng/ml amphiregulin (R&D Systems) for up to 4 hours.

To induce the expression of KRAS^G12D in granulosa cells, the previously described LSL-Kras^G12D mice were crossed with either the Amhr2-Cre knock-in mice(Amhr2^cre/+)(Jamin et al., 2002) or with transgenic mice in which Cre expression is driven by the Cyp19promoter (Cyp19-Cre). The ovarian expression pattern of the Amhr2-Cre allele has been described previously(Jorgez et al., 2004; Pangas et al., 2006; Boerboom et al., 2006). The generation of the Cyp19-Cre mouse strain is described in Materials and methods. To monitor the Cre activity in the ovaries, the Cyp19-Cre mice were crossed to the ROSA26 reporter mouse strain that expresses β-galactosidase (β-gal) only in Cre-expressing cells (Soriano, 1999). Ovaries were stained for β-gal activity using X-Gal substrate as previously reported (Jorgez et al.,2004). No β-gal activity was detected in ovaries of ROSA26 mice lacking Cre (Fig. 1A). In the ROSA26;Cyp19-Cre mice, β-gal was detected at low levels in granulosa cells of small follicles at postnatal day 10 (Fig. 1D) and at increased levels in granulosa cells of all antral follicles(Fig. 1B,E). Injections of ROSA26;Cyp19-Cre mice with equine chorionic gonadotropin (eCG), a known inducer of endogenous Cyp19 expression, stimulated follicle growth and increased Cre activity (Fig. 1C,F). Cre activity was not detected in theca cells or oocytes throughout postnatal development (Fig. 1D-F). These results indicate that the Cyp19-Cre mouse strain exhibits specific expression of Cre in the granulosa cells, with minimal leakage in other cell types.

Examination of Kras^G12D mRNA in immature LSL-Kras^G12D;Amhr2-Cre and LSL-Kras^G12D;Cyp19-Cre mice demonstrated that the Kras^G12D allele was efficiently recombined and expressed at levels comparable to the endogenous Kras gene(Fig. 1G). Since endogenous KRAS protein is highly expressed in granulosa cells of growing follicles,expression of the mutant allele is being induced in the same cell type as the endogenous gene (Fig. 1H).

For fertility tests, LSL-Kras^G12D;Amhr2-Cre and LSL-Kras^G12D;Cyp19-Cre females were bred to wild-type males continuously for 6 months. The average number of ∼7-8 pups per litter for the control mice (LSL-Kras^G12D) was not different from that of our C57BL/6J mouse colony. However, the LSL-Kras^G12D;Amhr2-Cre and Kras^G12D;Cyp19-Cre females (n=6, respectively)were subfertile over the 6-month period, with most pups being born in the first 2 months (Fig. 2A).

To determine the cause of reduced fertility in the Kras^G12D mutant mice, we tested their ability to ovulate by injecting immature mice with 4 IU of eCG and 46 hours later 5 IU of human chorionic gonadotropin (hCG). Whereas the control littermates ovulated many COCs at 16 hours after hCG injection, most LSL-Kras^G12D;Amhr2-Cre mice did not ovulate at all, and only a few COCs were observed in the oviducts of LSL-Kras^G12D;Cyp19-Cre mice(Fig. 2B). The histological data from the LSL-Kras^G12D;Amhr2-Cre mice are presented because the block of ovulation was more complete in this strain. However, the overall histological patterns in the two mutant strains were similar. Specifically, ovulation failure in the mutant mice was associated with defects of COC expansion and with the germinal vesicle breakdown of oocytes(Fig. 2D,F), whereas expanded COCs and meiotic oocytes with condensed chromosomes were present in the preovulatory follicles of control mice, at 8 hours post-hCG(Fig. 2C,E). In control mice,most large antral follicles ovulated by 16 hours after hCG(Fig. 2G) and had well developed corpora lutea (CLs) at 48 hours post-hCG(Fig. 2I). By contrast, antral follicles containing unovulated COCs remained in the mutant mouse ovaries at 16 hours post-hCG (Fig. 2H) and unovulated oocytes were trapped at the center of the CL at 48 hours post-hCG(Fig. 2J). More than 40 sections of Kras^G12D mutant ovaries (n=4) were examined, and trapped oocytes were present in 80-90% of the newly formed CLs. Because comparative analyses of the two mutant mouse strains revealed similar phenotypes and because expression of the Cyp19-Cre transgene is more specific for granulosa cells than is Amhr2-Cre, data from the Kras^G12D;Cyp19-Cre mice (C-Cre) mice are presented.

Ovaries from LSL-Kras^G12D;Amhr2-Cre and LSL-Kras^G12D;Cyp19-Cre mice were consistently larger and increased progressively in size, as compared with control littermates(Fig. 3A). Histological sections of the Kras mutant ovaries revealed multiple abnormal small`follicle-like' structures compared with controls(Fig. 3, C and D compared with B). These small follicle-like structures lacked an antrum and consisted of nests of disorganized, pleiomorphic granulosa cells. Many of these structures contained an oocyte of abnormal appearance that was displaced to the periphery of the `follicle' rather than being central(Fig. 3D). Cells within these abnormal follicle-like structures failed to express the granulosa cell marker genes Nr5a1 (see Fig. S1A-D in the supplementary material), Cyp11a1 (see Fig. S1E-H in the supplementary material) and Foxo1 (data not shown), indicating that normal granulosa cell differentiation had been blocked at an early stage of follicle growth.

To characterize the Kras mutant ovaries at the cellular level, we examined the proliferative rate of the developing follicles. In antral follicles, Kras^G12D-expressing granulosa cells demonstrated slightly increased levels of proliferation, based on BrdU incorporation (Fig. 3E,F,I) and immunofluorescence for the mitosis marker phospho-histone H3 (pHH3)(Fig. 3G,H,J). By contrast,only a limited number of cells within the abnormal follicle-like structures were positive for these proliferation markers. Immunohistochemical staining for proliferating cell nuclear antigen (PCNA) further proved that these abnormal structures are negative for proliferation markers(Fig. 3K,L).

Since the cells in the aberrant ovarian lesions were non-mitotic, the progressive enlargement of Kras mutant ovaries and the increased number of abnormal follicle-like structures might be caused by repression of apoptosis, a common feature in the mammalian ovary that serves to eliminate atretic follicles (Wang et al.,2006). Both the TUNEL assay and immunostaining for cleaved caspase 3 (CC3) were analyzed in the Kras mutant ovaries. Immature control and Kras mutant mice were primed with eCG and hCG to stimulate increased follicle growth. At 2 hours after hCG, DNA fragmentation(Fig. 3M) and caspase 3 cleavage (Fig. 3O) were detected in multiple pre- and early-antral follicles in control ovaries. By contrast, these apoptosis markers were markedly reduced in the Krasmutant ovaries, where the abnormal follicle-like structures were completely devoid of fragmented DNA and cleaved caspase 3(Fig. 3N,P). These results show that apoptosis was repressed by the Kras^G12D mutation in granulosa cells.

Because the Kras^G12D knock-in mice failed to ovulate,the expression of genes crucial for granulosa/cumulus cell differentiation and ovulation was analyzed in wild-type and mutant ovaries. As shown in Fig. 4A, Fshr mRNA was readily detected in ovaries of immature control mice and increased∼2.5-fold in response to eCG and was associated with the growth of preovulatory follicles. Other genes highly induced by eCG were Lhcgr,a marker of differentiated granulosa cells in preovulatory follicles, Areg, which encodes an EGF-like factor, and Cyp11a1, which encodes the steroidogenic enzyme leading to progesterone biosynthesis. Whereas expression of Fshr and Lhcgr was selectively reduced by the ovulatory stimulus of hCG, genes associated with ovulation (Areg,Ptgs2 and Tnfaip6) and luteinization (Cyp11a1) were upregulated markedly by hCG (Fig. 4A). By contrast, the induced expression of these genes was reduced/altered in ovaries of the Kras^G12D mutant mice. Notably, levels of Fshr mRNA were reduced in the ovaries of immature(untreated) Kras mutant mice indicating that constitutively active KRAS impairs the expression of this gene at an early stage of granulosa cell differentiation.

To determine whether the decreased levels of Fshr and Lhcgr mRNAs were the direct effect of mutant Kras^G12D expression, we isolated undifferentiated granulosa cells from immature LSL-Kras^G12D mice and cultured them in serum-free medium followed by infection with an adenoviral vector expressing Cre recombinase driven by the CMV promoter (Ad-CMV-Cre). In control cells, addition of FSH to the medium upregulated Fshr and induced the expression of Lhcgr mRNAs. However, the FshrmRNA level decreased in the granulosa cells expressing Kras^G12D, and the inductive effect of FSH on Lhcgr mRNA was totally abolished(Fig. 4B). This experiment confirmed our observations in vivo (Fig. 4A) and provided direct evidence that KRAS^G12D reduces Fshr mRNA levels and blocks FSH-mediated induction of luteinizing hormone (LH) receptors in granulosa cells.

FSH and LH transiently activate ERK1/2 and PI3K pathways in granulosa cells(Cottom et al., 2003; Gonzalez-Robayna et al.,2000). Recently, FSH has been shown to activate RAS, indicating that granulosa cells have factors that mediate G-protein receptor coupling to RAS (Wayne et al., 2007). Therefore, we analyzed components of the RAF1/MEK1/ERK1/2 and PI3K/AKT cascades in both control and Kras^G12D mutant ovaries. First, we measured RAS activity by RAS-GTP pull-down assay. Whereas levels of total RAS did not change throughout the ovulation period, levels of active,GTP-bound RAS were undetectable in ovaries at postnatal day 23, increased slightly in response to eCG (48 hours) and then increased markedly (but transiently) 2 hours after hCG (Fig. 5A). When RAS-GTP was measured by GST-RAF1 and GST-p110αpull-down assays in Kras mutant ovaries, high levels of RAS-GTP were present compared with control mice (Fig. 5B). These results indicate that Kras^G12Dmutant protein interacts with both PI3K and RAF1 in ovaries.

The phosphorylation status of selected RAS downstream kinases was also analyzed. Levels of phospho-MEK1/2 (MAP2K1/2 - Mouse Genome Informatics) and phospho-ERK1/2 were negligible in ovaries of immature mice prior to and 48 hours after eCG treatment (Fig. 5C; NT and eCG, respectively), increased dramatically 2 hours post-hCG and declined by 4 hours (see Fig. S2A in the supplementary material). Levels of phospho-AKT were low in ovaries of immature mice but increased markedly after eCG treatment. Phospho-AKT was further increased 2 hours post-hCG stimulation, was maintained at 4 hours(Fig. 5C and see Fig. S2A in the supplementary material) and returned to a basal level 8 hours post-hCG, a pattern similar to that of PDK1 (see Fig. S2A in the supplementary material). In addition, the total amount of the known AKT target, FOXO1, was reduced in Kras mutant ovaries compared with controls. The total amounts of ERK1/2 and AKT, as well as of actin (loading control), were not altered by gonadotropin treatment (Fig. 5C).

In comparison to their phosphorylation patterns in wild-type mice, elevated levels of phospho-MEK1/2 and phospho-AKT were observed in ovaries of Kras mutant mice even without hormonal stimulation(Fig. 5C, NT), and were only marginally increased in response to hCG, indicating that KRAS^G12Dexerted stimulatory effects on these pathways(Fig. 5C). By contrast, the levels of phospho-ERK1/2 were undetectable in the same Kras mutant ovaries and increased only marginally after hCG treatment. These data suggested that potent inhibitory factors selectively reduced ERK1/2 phosphorylation.

To determine the cell-specific pattern of phospho-ERK1/2 and phospho-AKT in ovaries, immunofluorescent staining was performed using phospho-specific antibodies. In wild-type ovaries, phospho-ERK1/2 was only detected in the large antral follicles 2 hours after hCG(Fig. 5E and see Fig. S2B in the supplementary material). By contrast, levels of phospho-ERK1/2 were markedly reduced in Kras mutant ovaries treated in the same manner(Fig. 5F). In control ovaries,phospho-AKT was first detected after eCG stimulation in the mural layer of granulosa cells in the large antral follicles, whereas the signal was weak in the cumulus cells (Fig. 5H and see Fig. S1C in the supplementary material). However, within 2 hours of hCG treatment, phospho-AKT was present in all granulosa/cumulus cells(Fig. 5I and see Fig. S1D in the supplementary material). By comparison, the phospho-AKT signal was already high in granulosa cells of the 23-day-old Kras mutant mice before eCG treatment. Phospho-AKT was also detected in some of the abnormal follicle-like structures (Fig. 5J, arrows). The progressive pattern of AKT phosphorylation was not seen in Krasmutant follicles. Rather, all granulosa/cumulus cells were positive for phospho-AKT after eCG treatment alone (Fig. 5K) or eCG/hCG treatment (2 hours)(Fig. 5L).

That KRAS^G12D interacts with RAF1 and activates MEK1/2 in granulosa cells but failed to increase ERK1/2 phosphorylation suggested that potent negative-feedback mechanisms were operative in these cells. To test this, granulosa cells isolated from 21-day-old LSL-Kras^G12D mouse ovaries were infected with Ad-CMV-Cre. Expression of Kras^G12D mRNA was detected by RT-PCR after adenoviral infection (Fig. 6A). Phospho-ERK1/2 increased gradually from 8 to 24 hours post-infection, but decreased to a basal level at 48 hours(Fig. 6B). Phospho-AKT also increased post-infection, but remained elevated at 48 hours, a time when the phospho-ERK1/2 level decreased, indicating that both the ERK1/2 and PI3K pathways in granulosa cells were transiently activated by KRAS^G12Dbut that negative-feedback mechanisms were induced to selectively block the ERK1/2 pathway in response to constitutively active KRAS^G12D. Additional LSL-Kras^G12D granulosa cells were infected with Ad-CMV-Cre or Ad-CMV-GFP (as control) for 48 hours and then stimulated with FSH, forskolin (Fo) or AREG for 20 minutes. As shown in Fig. 6C, each agonist induced ERK1/2 phosphorylation in control cells. However, in the cells expressing KRAS^G12D, the responses to agonists were markedly reduced indicating that KRAS^G12D activates a negative-feedback mechanism that represses ERK1/2 phosphorylation.

To elucidate specific changes in ovarian gene expression associated with the Kras^G12D mutation, microarray analyses were undertaken using RNA prepared from ovaries of LSL-Kras^G12D;Amhr2-Creversus LSL-Kras^G12D mice at 26 days of age. The microarray data showed that the Mkp3 (Dusp6) gene was upregulated in the Kras mutant ovaries, and this was confirmed by RT-PCR(Fig. 7A). This gene encodes MAPK phosphatase 3 (MKP3), which is an ERK1/2-specific protein phosphatase(Camps et al., 2000; Keyse, 2000; Li et al., 2007; Urness et al., 2007; Woods and Johnson, 2006). Expression of Mkp3 was induced in granulosa cells both in vivo (2-4 hours) and in vitro (1-2 hours) by hCG and AREG stimulation, respectively(Fig. 7B,J). In situ hybridization showed that Mkp3 mRNA is highly expressed in pre-ovulatory follicles 4 hours after hCG treatment(Fig. 7C,D), but is undetectable in 23-day-old ovaries (Fig. 7E,F). Ovaries of LSL-Kras^G12D;Amhr2-Cre mice(23 days old) exhibited elevated expression of Mkp3 mRNA in growing follicles, as compared with wild type (Fig. 7G,H, arrows).

To provide further evidence that MKP3 is functionally involved in the negative regulation of ERK1/2 activity, Mkp3 mRNA was depleted in cultured granulosa cells by RNAi. Mkp3 siRNA (50 nM) efficiently decreased Mkp3 mRNA in unstimulated cells or those exposed to AREG,the most potent stimulator of ERK1/2 in granulosa cells(Fig. 7I). AREG induced rapid but transient phosphorylation of ERK1/2 in control granulosa cells. However,in cells treated with Mkp3 siRNA, the dephosphorylation of ERK1/2 was significantly delayed (Fig. 7J,K). Lastly, ERK1/2 activity is required for the induction of Mkp3, because the MEK1/2 inhibitor PD98059 blocked the AREG-induced Mkp3 expression in cultured granulosa cells(Fig. 7M,N).

Activation of the RAS small G-protein family is crucial for FSH and EGF-like factor-induced signaling events in cultured granulosa cells via stimulation of downstream kinases such as ERK1/2 and AKT(Wayne et al., 2007). Based on in vitro studies, the ERK1/2 pathway is presumed to be essential for COC expansion and meiotic resumption of oocytes(Diaz et al., 2006; Fan et al., 2003; Shimada et al., 2006; Su et al., 2002). FSH-mediated stimulation of the PI3K pathway is also presumed to impact follicular development (Alam et al., 2004; Alliston et al., 2000; Park et al., 2005; Richards et al., 2002; Zeleznik et al., 2003). By analyzing the effects of expressing a constitutively active form of KRAS(KRAS^G12D) selectively in granulosa cells of two mouse models, we report the first detailed investigation of the consequences of mutant KRAS activation during mammalian follicle development and ovulation. In both models, we observed two distinct follicular phenotypes, suggesting that the impact of Kras^G12D is dependent on the stage of granulosa cell differentiation. If recombination occurred at early stages of follicle development, many abnormal follicle-like structures were observed, whereas if recombination occurred later, events associated with ovulation were impaired(Fig. 8). These changes in ovarian function caused the Kras^G12D mutant mice to be subfertile and to exhibit premature ovarian failure. Because Amhr2-Cre is expressed earlier in follicular development than Cyp19-Cre (our unpublished observations), the Kras^G12D;Amhr2-Cre mice exhibited more-severe ovulation detects. The expression of the Cyp19-Cre transgene is highly specific for granulosa cells. Although Cyp19-Cre is expressed in follicles at slightly later stages of growth than Amhr2-Cre, the Kras^G12D;Cyp19-Cre and Kras^G12D;Amhr2-Cre phenotypes are very similar, verifying that mutant KRAS^G12D impacts granulosa cell function in a stage-specific manner.

Because granulosa cells are highly proliferative and KRAS^G12Dcan induce tumorigenic transformation of several cell types(Campbell et al., 2007; Sarkisian et al., 2007; Shaw et al., 2007; Tuveson et al., 2004), we anticipated that expression of KRAS^G12D in granulosa cells might lead to enhanced proliferation and oncogenic transformation of these cells. Oncogenic transformation was not observed and alterations in proliferation were critically dependent on when recombination and expression of KRAS^G12D were initiated. In the abnormal follicle-like structures,no evidence for proliferation was observed. However, in the large antral follicles, proliferation was increased, indicating that the effects of mutant KRAS were dependent on the stage of granulosa cell differentiation. Furthermore, expression of KRAS^G12D led to impaired apoptosis of granulosa cells in the abnormal follicle-like structures, whose growth appeared to be self-limiting. Expression of KRAS^G12D also profoundly altered the fate and differentiation of granulosa cells. Specifically, expression of mutant KRAS^G12D in granulosa cells of small growing follicles completely disrupted normal follicular development and granulosa cell differentiation, as known markers of granulosa cell function(Fshr and Nr5a2)(Richards, 1994) were not detected. This altered cell fate led to a novel and unexpected ovarian phenotype, with follicle-like structures that were devoid of mitotic,apoptotic and differentiated cells (Fig. 8). The behavior of the granulosa cells in these abnormal follicle-like structures appears to be similar to the premature senescence observed in primary cells in culture expressing mutant forms of HRAS(Lin et al., 1998).

Follicles in which granulosa cells escaped the recombination events at an early stage of development continued to grow to the antral stage. However,follicles with granulosa cells expressing KRAS^G12D at this stage also exhibited impaired function. Specifically, most antral follicles failed to ovulate even if exposed to exogenous hormones. Ovulation failure was associated with impairments in expansion of cumulus cells, in meiotic maturation of the oocytes and in expression of ovulation-related genes. This phenotype is similar to that of the mutant mouse model with an EGFR signaling defect (Hsieh et al., 2007). The altered response of KRAS^G12D-expressing granulosa cells to LH/hCG appears to be related to low levels of Fshr and the inability of FSH to induce expression of Lhcgr mRNA, and therefore to the loss of the crucial LH-ERK1/2 signaling pathways. This conclusion is supported by the reduced expression of specific genes known to be essential for COC expansion and ovulation (Richards,2005), including Ptgs2, Has2 and Tnfaip6.

The mechanisms by which KRAS^G12D alters granulosa cell functions and fate appear to be mediated by both the ERK1/2 pathway and the PI3K pathway. Specifically, our results show for the first time that treatment of mice in vivo with exogenous hormones, eCG and hCG, leads to transient activation of RAS and that this is associated with the transient phosphorylation of ERK1/2 and AKT in granulosa cells. By contrast, granulosa cells expressing KRAS^G12D exhibit elevated levels of RAS-GTP, as would be expected. In these cells, KRAS^G12D interacts with RAF1 and initially leads to increased phosphorylation of ERK1/2. These results support our recent study showing that RAS-GTP in rat granulosa cells interacts with RAF1 directly (Wayne et al.,2007), as well as the studies of others who have shown that KRAS^G12D selectively activates RAF1 and/or PI3K in a cell- and context-specific manner (Campbell et al.,2007; Cespedes et al.,2006; Gupta et al.,2007; Tuveson et al.,2004). However, KRAS^G12D-mediated ERK1/2 phosphorylation is transient and becomes markedly reduced in the mutant granulosa cells in vivo and in culture. This transient activation of ERK1/2 is mediated, at least in part, by the upregulation of MKP3, a specific ERK1/2 phosphatase (Camps et al.,2000; Keyse, 2000; Li et al., 2007; Urness et al., 2007; Woods and Johnson, 2006). Mkp3 mRNA was rapidly induced in granulosa cells of wild-type mice in response to hCG and was elevated in Kras^G12D mutant ovaries (Fig. 7) and cultured granulosa cells expressing KRAS^G12D (data not shown). Because induction of Mkp3 mRNA by AREG in granulosa cells was blocked by the MEK1/2 inhibitor PD98059, these results reinforce the notion that ERK1/2 induces expression of this negative-regulatory factor in granulosa cells. Conversely, reducing Mkp3 expression by a siRNA approach prolonged the presence of phospho-ERK1/2 in response to AREG by up to 60 minutes. Collectively, these results provide evidence that MKP3 is regulated in murine ovarian granulosa cells and controls the duration of ERK1/2 phosphorylation.

Although Mkp3 is induced in granulosa cells of preovulatory follicles and is initially elevated in these cells in the Kras^G12D mutant ovaries, Mkp3 mRNA was not expressed in the granulosa cells contained within the abnormal follicle-like structures. Thus, the absence of phospho-ERK1/2 in these cells also indicates that other potent mechanisms impact and reduce ERK1/2 signaling in these mutant cells. For example, RAS can mediate the epigenetic silencing of genes via its ability to induce CpG methylation at promoter regions of certain genes(Gazin et al., 2007). Moreover, the mediators of RAS epigenetic silencing include Mapk1(Erk2), Pdpk1 (Pdk1) and Dnmt1(Gazin et al., 2007). Thus, it is tempting to speculate that the cells within the abnormal follicle-like structures have undergone specific epigenetic changes to prevent their proliferation, apoptosis and differentiation.

In contrast to ERK1/2, phosphorylation of AKT in granulosa cells of growing and large antral follicles was enhanced by the presence of KRAS^G12Din vivo and in KRAS^G12D-expressing granulosa cells in culture. Since our GST pull-down assays showed that KRAS^G12D interacts directly with the p110α subunit of PI3K as previously reported(Rodriguez-Viciana et al.,1994; Rodriguez-Viciana et al., 1996), it is likely that KRAS^G12D stimulates the PI3K pathway directly, leading to the prolonged activation of AKT in granulosa cells. KRAS^G12D also impairs the expression of FOXO1 that may be mediated by prolonged activation of AKT. Since FOXO1 has been shown to impair granulosa cell differentiation (Park et al., 2005; Rudd et al.,2007), one might have predicted that the mutant cells would exhibit increased responsiveness to FSH, which is not the case. Rather, the PI3K pathway appears to regulate additional functions in granulosa cells. Because Pdpk1 is a factor implicated in RAS-mediated epigenetic gene silencing (Gazin et al.,2007), it is possible that the PI3K pathway is crucial for dictating the fate of granulosa cells in small follicles.

In summary, transient activation of RAS and the phosphorylation of downstream targets, such as the RAF1/MEK1/ERK1/2 and PI3K/AKT cascades, appear to be crucial for mediating appropriate responses of granulosa cells to the gonadotropic hormones FSH and LH, leading to progressive follicular development and ovulation. Conversely, persistent expression of a constitutively active form of KRAS (KRAS^G12D) impairs ovulation and the expression of ovulation-related genes. Moreover, if expressed at an early stage in follicle development, KRAS^G12D dramatically alters granulosa cell fate by precluding granulosa cell differentiation,proliferation and apoptosis, thus impairing granulosa cell responses to gonadotropins and leading to premature ovarian failure(Fig. 8). This marked divergence in granulosa cell function suggests that the potent epigenetic silencing of the promoters of specific genes might provide the basis of how activation of RAS alone can cause quiescence/senescence, rather than transformation, of these cells. These results also provide novel evidence that granulosa cells in vivo possess mechanisms that make them extremely impervious to tumorous transformation and that instead lead to premature ovarian failure.

We thank Dr Tyler Jacks for the LSL-Kras^G12D mice; Dr Richard Behringer for the Amhr2-Cre mice; Dr Michael Mancini and members of the Microscopy Core (HD-07495) for their time and assistance; and Yuet Lo for her many contributions. We thank Rebecca Robker and Darryl Russell for making the template of the ovary that appears in modified form in Fig. 8. This work is supported by NIH grants HD-16229 and HD-07495 (J.S.R.), and in part by JSPS-18688016(M.S.). H.-Y.F. is supported in part by an NIH Postdoctoral Training Grant(HD-07165).