Oocyte regulation of metabolic cooperativity between mouse cumulus cells and oocytes: BMP15 and GDF9 control cholesterol biosynthesis in cumulus cells

Bi-directional communication between oocytes and companion somatic cells is essential for the development and function of ovarian follicles and promotes the production of mature oocytes competent to undergo fertilization,preimplantation development and development to term. Although granulosa cells provide essential nutrients and stimuli for oocyte growth and development,oocytes are not merely passive recipients of such support, but rather active regulators of follicular development. Oocytes affect the development and function of all stages of follicles beginning with the formation of primordial follicles (Soyal et al.,2000). Oocytes promote the primary to secondary follicle transition (Dong et al., 1996; Elvin et al., 1999b; Galloway et al., 2000; Juengel et al., 2002; Latham et al., 2004),granulosa cell proliferation and differentiation before the luteinizing hormone (LH) surge (Gilchrist et al.,2003; Gilchrist et al.,2000; Gilchrist et al.,2001; Otsuka et al.,2005; Otsuka et al.,2000; Vanderhyden et al.,1992; Vitt et al.,2000), the preantral to antral follicle transition(Diaz et al., 2007a; Diaz et al., 2007b; Orisaka et al., 2006) and cumulus expansion and ovulation after the LH surge(Buccione et al., 1990; Diaz et al., 2006; Dragovic et al., 2005; Dragovic et al., 2007; Joyce et al., 2001; Su et al., 2004; Vanderhyden et al., 1990).

Recently emerging evidence points to the existence of an oocyte-granulosa cell regulatory loop by which complementary signaling and metabolic pathways drive the development and function of both the oocytes and follicular somatic compartments. For example, Slc38a3, which encodes a sodium-coupled neutral amino acid transporter, and Aldoa, Eno1, Ldha, Pfkp, Pkm2 and Tpi1, encoding enzymes in the glycolytic pathway, are highly expressed in cumulus cells compared with mural granulosa cells, and their expression in cumulus cells is promoted by oocyte-derived paracrine factors(Eppig et al., 2005; Sugiura et al., 2005). Moreover, the uptake of L-alanine and L-histidine, two preferred substrates of SLC38A3 (Gu et al., 2000), and the activity of glycolysis in cumulus cells, are promoted by factors secreted by fully grown oocytes at the germinal vesicle stage(Eppig et al., 2005; Sugiura et al., 2005). Since oocytes themselves are unable to take up L-alanine and poorly metabolize glucose for energy production, they obtain these amino acids and products of glycolysis, which are essential for their development and function, from cumulus cells (Biggers et al.,1967; Colonna and Mangia,1983; Donahue and Stern,1968; Eppig et al.,2005; Haghighat and Van Winkle, 1990; Leese and Barton, 1984; Leese and Barton, 1985). Thus, oocytes benefit their own development by enhancing metabolic cooperativity between granulosa cells and oocytes (for a review, see Sugiura and Eppig,2005).

Growth differentiation factor 9 (GDF9) and bone morphogenetic protein 15(BMP15) are two well-characterized oocyte-derived growth factors that play crucial roles in follicle growth and ovulation in all mammalian species studied, including rodents (Dong et al.,1996; Elvin et al.,1999b; Yan et al.,2001), domestic ruminants(Bodin et al., 2007; Galloway et al., 2000; Juengel et al., 2002) and humans (Chand et al., 2006; Di Pasquale et al., 2006; Dixit et al., 2006; Palmer et al., 2006). GDF9 and/or BMP15 are probably major players of the `oocyte-granulosa cell regulatory loop', and participate in many of the aforementioned functions of oocytes (for reviews, see Eppig,2001; Erickson and Shimasaki,2001; Matzuk et al.,2002; McNatty et al.,2004). Genetic targeting or spontaneous mutations of either Gdf9 or Bmp15 in mammals affect fertility in females (for reviews, see Juengel and McNatty,2005; Pangas and Matzuk,2004). Particularly in mice, deletion of Gdf9 by homologous recombination(Gdf9^tm1Zuk/Gdf9^tm1Zuk, hereafter Gdf9^-/-) causes arrest of folliculogenesis at the primary stage and female infertility since the cuboidal granulosa cells fail to proliferate (Dong et al.,1996; Elvin et al.,1999b). Deletion of Bmp15(Bmp15^tm1Zuk/Bmp15^tm1Zuk, hereafter Bmp15^-/-) results in reduced female fertility with the primary defects in ovulation and fertilization(Yan et al., 2001). A more dramatic reduction of fertility was observed in double mutant Bmp15^-/-Gdf9^+/- (hereafter DM) than in Bmp15^-/- females. The cumuli oophori ovulated in DM females are fragile and unstable (Yan et al., 2001) indicating that GDF9 and BMP15 are essential for the normal development of cumulus-oocyte complexes (COCs). Although in-vitro studies using recombinant GDF9 and BMP15 demonstrate that both growth factors,either alone or in combination, play significant role(s) at all stages of follicular development (Elvin et al.,1999a; Elvin et al.,2000; Hayashi et al.,1999; Hussein et al.,2005; McNatty et al.,2005a; McNatty et al.,2005b; Otsuka et al.,2001a; Otsuka and Shimasaki,2002; Otsuka et al.,2001b; Otsuka et al.,2000; Vitt et al.,2000), controversy persists owing to differences in recombinant protein preparations (for a review, see Pangas and Matzuk, 2005). It has been suggested that the role of BMP15 in mouse follicular development is restricted to the period after the LH surge(Gueripel et al., 2006; Li et al., 2006; Yoshino et al., 2006). These studies are contradicted by evidence that cumuli oophori of DM mice are abnormal even before the LH surge because they are unable to undergo normal expansion in vitro even when co-cultured with normal wild-type oocytes(Su et al., 2004). However,the extent of the role of BMP15 in the differentiation and function of cumulus cells before the LH surge is unknown.

The first objective of the present study was to determine the effects of BMP15 and GDF9 on cumulus cells before the LH surge by analyzing the transcriptomes of cumulus cells from wild-type (WT), Bmp15^-/- and DM mice using microarrays and bioinformatics methods. We report that cumulus cell metabolic pathways, particularly glycolysis and cholesterol biosynthesis, are highly affected by Bmp15and Gdf9 mutation. To follow up on these findings, we conducted a detailed analysis of cholesterol biosynthesis in oocytes and cumulus cells and the ability of oocytes to promote the cholesterol biosynthetic pathway in cumulus cells.

Adult (4- to 5-month-old) female Bmp15^-/- and DM mice on the B6/129 genetic background and similarly aged WT B6129F1 mice produced in the research colonies of the authors were used for the microarray and the subsequent real-time PCR validation experiments. Other experiments were conducted with normal 22- to 24-day-old female B6SJLF1 mice. All animal protocols were approved by the Administrative Panel on Laboratory Animal Care at The Jackson Laboratory, and all experiments were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals.

Female WT, Bmp15^-/- and DM mice were primed with 7.5 IU equine chorionic gonadotropin (eCG, EMD Biosciences, Calbiochem, La Jolla, CA)for 48 hours to stimulate follicular development. Cumulus-cell-oocyte complexes (COCs) were released by puncturing large antral follicles with a pair of 26-gauge needles. Released COCs were collected and washed three times by passing through three dishes of medium. Cumulus cells were then stripped off oocytes by passing COCs several times through a glass pipette with an inner diameter slightly narrower than the oocyte. After removing all of the denuded oocytes from the dish, cumulus cells were transferred into a 1.5 ml centrifuge tube, and collected by gentle centrifugation. The resulting pellets were resuspended in 350 μl RLT buffer (Qiagen, Valencia, CA) after removing the supernatant, and were snap frozen in liquid nitrogen and temporarily stored at -80°C until RNA isolation. Three sets of WT, Bmp15^-/- and DM cumulus cell samples were collected and employed in this microarray study. For each sample, about 75-100 COCs,obtained from 3-4 mice, were used for cumulus cell collection. Four additional sets of cumulus cell samples were collected and used for subsequent real-time RT-PCR analysis. Medium used for cumulus cell isolation was MEM-α(Invitrogen Corporation, Grand Island, NY) supplemented with 3 mg/ml crystallized lyophilized bovine serum albumin (Sigma, St Louis, MO), 75 mg/l penicillin G (Sigma) and 50 mg/l streptomycin sulfate (Sigma). Milrinone(Sigma), a selective inhibitor of oocyte-specific phosphodiesterase (PDE3),was added into the medium at a concentration of 5 μM to prevent the fully grown GV-stage oocytes from undergoing maturation during the process of COC and cumulus cell isolation and culture.

Total RNA was extracted from cumulus cells using the RNeasy Micro Kit(Qiagen) according to the manufacturer's instructions. The RNA quality and yield of each sample were determined using the Bioanalyzer 2100 and RNA 6000 Pico LabChip assay (Agilent Technologies, Palo Alto, CA) in combination with Quant-iT RiboGreen Reagent according to supplied protocols (Invitrogen). Total RNA (10 ng) isolated from each sample was used in the two-round cDNA synthesis and subsequent in vitro-transcription according to the Two-Cycle Eukaryotic Target Labeling Assay [Affymetrix Expression Analysis Technical Manual:Section 2: Eukaryotic Sample and Array Processing(http://www.affymetrix.com/support/technical/manual/expression_manual.affx)]. Equal amounts (15 μg) of fragmented and biotin-labeled cRNA from each sample were then hybridized to Affymetrix GeneChip Mouse Genome 430 2.0 Arrays for 16 hours at 45°C. Post-hybridization staining and washing were performed according to manufacturer's protocols using the Fluidics Station 450 instrument (Affymetrix).

After post-hybridization staining and washing, the arrays were scanned with a GeneChip 3000 laser confocal slide scanner (Affymetrix) and the images were quantified using Gene Chip Operating Software version 1.2 (GCOS, Affymetrix). Probe level data were imported into the R software environment and expression values were summarized using the RMA (Robust MultiChip Average) function(Irizarry et al., 2003) in the R/affy package (Gautier et al.,2004). Using the R/maanova package (Wu, 2003), an analysis of variance (ANOVA) model was applied to the data, and Fs test statistics were constructed along with their permutation P-values(Cui and Churchill, 2003; Cui et al., 2005). False discovery rate (FDR) (Storey and Tibshirani, 2003) was then assessed using the R/qvalue package to estimate q-values from calculated Fs test statistics. Three pairwise comparison analyses: DM vs WT, Bmp15^-/- vs WT, and DM vs Bmp15^-/-, were generated, and the significantly changed transcripts were identified using the criteria of Fs P<0.01. Results were annotated using information provided by Affymetrix (12/20/2005 release). Full data sets are available at http://www.ncbi.nlm.nih.gov/geo/(Acc. no. GSE7225).

Gene identifiers and their corresponding Fs p-values, and fold changes were uploaded into IPA 3.1 (Ingenuity Pathway Analysis, Ingenuity System, http://www.ingenuity.com)and GenMAPP 2.0 (Gene Map Annotator and Pathway Profiler, http://www.genmapp.org)/MAPPFinder 2.0 to identify the pathways and functions associated with significantly changed transcripts.

Real-time RT-PCR analyses were carried out using total RNA isolated from target cells (cumulus cells or oocytes). RNA isolation was accomplished using the RNeasy Micro Kit (Qiagen). In vitro transcription was carried out using QuantiTect Reverse Transcription Kit (Qiagen) at 42°C for 15 minutes. Real-time PCR was then conducted to quantify the steady-state mRNA levels of the tested genes using QuantiTect SYBR Green PCR Kits (Qiagen) on the ABI 7500 Real-time PCR System (Applied Biosystems, Foster City, CA). The threshold cycle (Ct) was used for determining the relative expression level of each gene by normalizing to the Ct of Rpl19 mRNA. The method of 2^-ddCt was used to calculate relative fold change of each gene as described previously (Su et al.,2007). To ensure only target gene sequence-specific, non-genomic products were amplified by real-time PCR, careful design and validation of each primer pair, as well as cautious manipulation of RNA were undertaken as described exactly in previous studies (Su et al., 2007). Primers used for real-time PCR are shown in Table 1.

COCs were isolated from 22-day-old eCG-primed B6SJLF1 mice. Microsurgical oocytectomy (OOX) was carried out as described previously(Buccione et al., 1990). Fully-grown oocytes were isolated from the same age eCG-primed B6SJLF1 mice as described previously (Su et al.,2007). COCs, OOX cumulus cells (without oocytes) and OOX cumulus cells + oocytes (two fully grown oocytes/μl medium) were cultured in a drop of medium covered with mineral oil at a density of one COC or OOX cumulus cell/μl medium in a four-well plate (Nuclon, Denmark). Medium used for culture was the same as that used for mutant cumulus cell isolation. Cells were cultured at 37°C in a modular incubation chamber (Billups Rothenberg,Del Mar, CA) infused with 5% O₂, 5% CO₂ and 90%N₂ for 20 hours, and then were collected in RTL buffer for RNA isolation.

In situ hybridization was performed using ovarian sections derived from eCG-primed (44-46 hours) B6SJLF1 mice as described previously(Eppig et al., 2002). ³³P-labeled cRNA probes were prepared using target gene-specific PCR products amplified from cDNA of B6SJLF1 ovaries. The length and region of probes were: Mvk, 859 bp, NM_023556, 604-1462; Fdps, 1018 bp, NM_134469, 16-1033; Sqle, 809 bp, NM_009270, 1364-2172; Cyp51, 1205 bp, NM_020010, 730-1934; Sc4mol, 806 bp,NM_025436, 428-1233.

Levels of cholesterol in cumulus cells and/or oocytes were compared by assessing the incorporation of [1-¹⁴C]acetate into cholesterol using a protocol adapted from previous reports(Friberg et al., 2007; Rung et al., 2006; Rung et al., 2005). Briefly,for comparing cholesterol synthesis in WT, Bmp15^-/-, and DM COCs, 150 COCs of each genotype were cultured in medium supplemented with 10 μCi [1-¹⁴C]acetic acid, sodium salt (Amersham Biosciences,Buckinghamshire, UK) for 5 hours. For testing the effects of OOX on cholesterol synthesis in WT cumulus cells, 150 COCs, OOX cumulus cells or OOX cumulus cells + oocytes were initially cultured in a drop of radioisotope-free medium covered by mineral oil at a density of 1 COC or OOX cumulus cell/μl medium in a four-well plate for 15 hours, and then transferred to fresh medium(375 μl/well) where cumulus cells of the intact COC group were stripped off and oocytes discarded. Finally, 10 μCi (50 μl) [¹⁴C]acetate was added and cells were cultured for additional 5 hours. At the end of culture, cells and media were collected, and cholesterol in the cells and media was extracted and subjected to thin layer chromatography (TLC). For comparing the levels of cholesterol synthesized in WT cumulus-enclosed oocytes and denuded oocytes, 400 cumulus-cell-enclosed and denuded oocytes were incubated with 10 μCi [¹⁴C]acetate in 425 μl medium for 5 hours. They were washed four times in 2.5 ml fresh medium. After washing, the cumulus-cell-enclosed oocytes were denuded, and resultant oocytes were collected for cholesterol extraction. Equal numbers of oocytes were also collected from the denuded oocyte group incubated without cumulus cells, and subjected to cholesterol extraction and TLC separation.

To compare levels of cholesterol synthesized in WT and DM cumulus-enclosed oocytes, and to test effects of co-culturing with fully-grown WT oocytes on cholesterol synthesis in DM cumulus cells and oocytes, 100 WT and DM COCs or 100 DM COCs + WT oocytes (four oocytes/μl medium) were initially cultured in a drop of medium covered by mineral oil at a density of one COC/μl of medium in a four-well plate for 15 hours. Then 2.5 μCi[¹⁴C]acetate was added and cultured for an additional 5 hours. At the end of culture, complexes were washed four times in 2.5 ml fresh medium,and cumulus-cell-enclosed oocytes were denuded, and resultant oocytes and cumulus cells were collected. For each TLC run, oocytes collected from four independent experiments, each containing 100 oocytes, were pooled to produce a total of 400 denuded oocytes, and cholesterol was extracted. This experiment was then replicated four times.

Unlabeled cholesterol (10 μg) was added to each sample at the beginning of cholesterol extraction to serve as a carrier and external control. The TLC plates were silica gel 60, 20×20 cm (Merck, Darmstadt, Germany). Compounds in extracted samples were separated on TLC plates using a mobile phase solvent mixed of petroleum ether:diethyl ether:acetic acid (60:40:1,v/v). Bands separated on plates were visualized using iodine (Sigma, Grand Island, NY) staining. Unlabeled cholesterol (10 μg) was loaded directly on the plate in a separate lane to identify the location of radioisotope-labeled cholesterol bands. Dried plates were placed onto Fuji Phosphor Imaging Plates(Fuji Medical Systems USA, Stamford, CT), and exposed for at least 2 days,scanned using a Fuji Phosphor Imager (Fuji Medical Systems USA), and the intensity of each corresponding cholesterol band quantified using Fuji Phosphor Imaging system software (Fuji Medical Systems USA).

All experiments were repeated at least three times independently, and data are presented as mean ± s.e.m. Student's t-test was conducted to evaluate differences when there were only two groups. For experiments with more than two groups of treatments, one-way ANOVA followed by Tukey's HSD test was used to evaluate differences between groups using JMP software (SAS Institute, Cary, NC). P<0.05 was considered significantly different.

Three pairwise comparisons, DM vs WT, Bmp15^-/- vs WT and DM vs Bmp15^-/-, identified the most highly affected transcripts in mutant cumulus cells. As shown in Fig. 1, compared with WT, there were 7640 and 5332 unique transcripts whose levels of expression were significantly changed in DM and Bmp15^-/- cumulus cells,respectively. Interestingly, when compared with WT, there were 4147 (2958 +1189, 54.3%) and 1839 (1522 + 317, 34.5%) transcripts whose expression was changed only in DM or Bmp15^-/- cumulus cells, respectively(Fig. 1). Since these transcripts were not commonly changed in the two groups, they are regulated in cumulus cell either by the full complement of GDF9 and BMP15, or only by BMP15. There were 744 transcripts commonly altered in all three pairwise comparisons (Fig. 1) and these were considered the transcripts most highly affected by mutations of Bmp15 and Gdf9 and were used for bioinformatic pathways and functions analyses.

Validation of data was carried out on two groups of selected transcripts using quantitative real-time RT-PCR (Fig. 2). The first group of transcripts was representative of those whose steady-state levels appeared highly changed in DM(Fig. 2A,B) and Bmp15^-/- (Fig. 2F,G) cumulus cells by microarray analysis. The second group was representative of those in DM (Fig. 2C,D,E) and Bmp15^-/-(Fig. 2H,I,J) cumulus cells involved in specific metabolic pathways described in the following section(see Fig. 3). In all cases,quantitative differences between groups were similar in both microarray and RT-PCR data, thus validating use of microarray data for further bioinformatic analyses and testing of physiological mechanisms.

To identify biological themes underlying effects of the mutations on transcript levels in cumulus cells, IPA and GenMAPP/MAPPFinder bioinformatic packages were used to carry out pathway and function analyses on the 744 transcripts whose levels of expression were commonly affected in the mutant groups as shown in Fig. 1. As shown in Fig. 3A, seven canonical pathways (of 113 in the IPA pathway library) were significantly affected. Surprisingly, all pathways identified were metabolic and the majority of changed transcripts involved in these pathways were downregulated in mutant cumulus cells. When IPA analysis was conducted using only downregulated transcripts, the same pathways were found to be significantly affected as when all the changed transcripts were used. No pathways were significantly affected when only upregulated transcripts were used in the IPA analysis. Of the seven identified pathways, glycolysis/gluconeogenesis and sterol biosynthesis were the two pathways most affected. As shown in Fig. 3C (and see Fig. S1 in the supplementary material), most of the transcripts encoding enzymes for sterol biosynthesis and glycolysis/gluconeogenesis, respectively, were downregulated in Bmp15^-/- and DM cumulus cells.

IPA also identified 25 categories of molecular and cellular functions that were associated with the 744 transcripts. The 10 most affected are shown in Fig. 3B. Lipid metabolism and small molecule biochemistry were the most highly affected functions and cholesterol biosynthesis (sterol biosynthesis) was the major subcategory of these two functions (see Tables S1 and S2 in the supplementary material). Canonical pathways and molecular and cellular functions identified by IPA were essentially the same as those identified by GenMAPP/MAPPFinder analyses as downregulated in the mutant cumulus cells (see Tables S3 and S4 in the supplementary material).

Altered expression of transcripts in mutant cumulus cells may be the result of chronic deficiencies in BMP15 and/or GDF9 throughout follicular development and may not reflect the acute regulatory response of cumulus cells to oocyte-derived factors. To address this possibility, we tested the effects of WT oocytes on expression of transcripts by WT cumulus cells cultured for only 20 hours. Transcripts chosen for analysis were those whose levels of expression were affected in Bmp15^-/- and DM cumulus cells and validated in experiments shown in Fig. 2, with emphasis given to transcripts encoding enzymes of the cholesterol biosynthesis pathway. As shown in Fig. 4, for all selected transcripts, OOX resulted in a pattern of mRNA expression in WT cumulus cells that was similar to that in mutant cumulus cells. Co-culture of OOX cumulus cells with WT oocytes prevented these changes. Specifically, OOX caused a dramatic increase in the expression of the transcripts whose levels were upregulated in Bmp15^-/- and DM cumulus cells, and this change was prevented by co-culture of OOX cumulus cells with WT oocytes(Fig. 4A). OOX also dramatically reduced expression of transcripts whose levels of expression were downregulated in Bmp15^-/- and DM cumulus cells, and this reduction did not occur when oocytes were present(Fig. 4B). Most interestingly,of the 16 transcripts selected from the 17 transcripts encoding enzymes for cholesterol biosynthesis, 15 were found to be expressed at significantly lower levels in OOX cumulus cells than in cumulus cells of intact COCs(Fig. 4C). Co-culture of OOX cumulus cells with WT oocytes sustained elevated steady-state expression of these transcripts. Similar changes were observed for transcripts encoding enzymes involved in other metabolic pathways, i.e. glycolysis, purine metabolism, pyrimidine metabolism, pentose phosphate, fructose and mannose metabolism and inositol metabolism (Fig. 4E). In contrast to the downregulation of transcripts encoding enzymes required for cholesterol biosynthesis in OOX cumulus cells, a dramatic upregulation of Cyp27a1 mRNA encoding cholesterol 27 hydroxylase,which functions in cholesterol metabolism, was observed in OOX cumulus cells,and this upregulation was prevented by co-culture with oocytes(Fig. 4D).

We next determined whether the reduced expression of transcripts encoding enzymes in the cholesterol biosynthesis pathway reflects changes in cholesterol synthesis in cumulus cells. As shown in Fig. 5A,B, compared with WT COCs, levels of de-novo-synthesized cholesterol in Bmp15^-/- and DM COCs were dramatically reduced, to about 55% and 25% of WT level, respectively. OOX resulted in more than a 90%reduction of de-novo synthesized cholesterol in WT OOX cumulus cells(Fig. 5C,D). Synthesis of cholesterol in OOX cumulus cells was elevated when they were co-cultured with WT oocytes. However, this increase was only to 50% of the control level(Fig. 5C,D).

Steady-state levels of transcripts encoding enzymes in the cholesterol biosynthesis pathway were compared in cumulus cells and oocytes obtained from WT mice. As shown in Fig. 6A,transcripts, Mvk, Pmvk, Fdps, Sqle, Cyp51, Sc4mol and Ebp,were expressed at much higher levels in cumulus cells than in oocytes,relative to levels of Rpl19 mRNA in the respective cell types. Because possible differences in levels of Rpl19 mRNA in oocytes and cumulus cells would bias this comparison, we compared expression by unbiased in situ hybridization. Robust levels of Mvk, Fdps, Sqle, Cyp51, and Sc4mol transcripts (Fig. 6B), and Pmvk and Ebp transcripts (not shown),were detected in cumulus cells, as well as the periantral granulosa cells, but not in oocytes (Fig. 6B). The results described above suggest that oocytes are deficient in sterol biosynthesis and require products of this pathway to be supplied by cumulus cells. To test this possibility, levels of cholesterol synthesized in cumulus-cell-enclosed oocytes and denuded oocytes were compared. After stripping cumulus cells from cumulus-cell-enclosed oocytes, >fivefold more radiolabeled cholesterol was found in cumulus-cell-enclosed than denuded oocytes (Fig. 7A). Differences between labeled cholesterol in denuded oocytes and cumulus-cell-enclosed oocytes could result from differences in availability of labeled acetate. Note, however, that the unidentified radiolabeled bands near the origin are about the same intensity in both groups(Fig. 7A), suggesting similar availability of labeled acetate substrate for their synthesis. These results,therefore, suggest that cumulus cells provide oocytes with newly synthesized cholesterol. Support for this conclusion is the observation that lower levels of [¹⁴C]cholesterol were detected in DM cumulus-cell-enclosed oocytes compared with WT oocytes, and levels of [¹⁴C]cholesterol in DM oocytes were partially promoted by co-culture of DM cumulus-cell-enclosed oocytes with WT oocytes (Fig. 7B). These changes in DM oocytes are coincident with changes in[¹⁴C]cholesterol levels in DM cumulus cells(Fig. 7C). Cumulus cells were collected from the same COCs used for measuring levels of[¹⁴C]cholesterol in oocytes(Fig. 7B).

The transcriptome of cumulus cells before the LH surge was highly affected by deletion of Bmp15, and this effect was enhanced in Bmp15^-/-Gdf9^+/- (DM) mice. Thus both BMP15 and GDF9 are important regulators of cumulus cell development and function before the LH surge. The most highly affected processes were metabolic, with glycolysis and sterol biosynthesis affected most dramatically. These effects could reflect acute regulation of these transcripts by oocytes,rather than an effect of chronic deprivation of BMP15 and GDF9 throughout follicular development, because removal of the oocyte from WT COCs results in the same alterations in steady-state transcript levels as observed in mutant cumulus cells. Synthesis of cholesterol from acetate was reduced by removal of oocytes from WT COCs and in cumulus cells of mutant mice. Two lines of evidence indicate that mouse oocytes are deficient in their ability to synthesize cholesterol. First, levels of expression of transcripts encoding enzymes of the cholesterol synthesis pathway were very low in oocytes compared with cumulus cells. Second, cumulus-cell-denuded oocytes convert acetate to cholesterol poorly. Thus oocytes probably depend upon cumulus cells to provide them with cholesterol, and oocytes stimulate this activity in cumulus cells via BMP15 and GDF9. Supporting this conclusion, lower levels of radiolabeled cholesterol were detected in cumulus-cell-enclosed DM oocytes, which was coincident with the reduced ability of DM cumulus cells to synthesize cholesterol. Co-culture of the DM complexes with fully grown WT oocytes partially restored levels of radiolabeled cholesterol in both DM cumulus cells and oocytes.

The glycolysis/gluconeogenesis pathway is highly affected in cumulus cells by Bmp15 and Gdf9 mutation. Steady-state levels of most of transcripts encoding enzymes of the glycolytic pathway were decreased in both Bmp15^-/- and DM cumulus cells. That oocytes control glycolysis in cumulus cells was known from previous studies(Sugiura et al., 2005), but the breadth of impact of oocytes on expression of diverse transcripts encoding enzymes needed for this pathway had not been realized prior to this analysis of mutant cumulus cell transcriptomes.

As indicated above, effects of Bmp15^-/- and DM on expression of transcripts encoding enzymes required for glycolysis in cumulus cells was anticipated based on previous studies(Sugiura et al., 2005). By contrast, the effects of these mutations on cholesterol biosynthesis in cumulus cells was entirely unexpected. Almost all (13/17) transcripts encoding enzymes required for cholesterol biosynthesis were downregulated in cumulus cells of both mutants, and this correlated with a reduction of de-novo-synthesized cholesterol from acetate. Therefore, BMP15 and GDF9 control the rate of cholesterol biosynthesis in cumulus cells at least in part by promoting expression of transcripts encoding cholesterol biosynthetic enzymes. Removal of oocytes resulted in downregulation in levels of 15/17 of transcripts in this pathway, as well as reduction in cholesterol synthesis in cumulus cells without oocytes. Co-culture of OOX cumulus cells with WT oocytes completely prevented the decrease in steady-state transcript levels in WT OOX cumulus cells. Although cholesterol synthesis still decreased in OOX cumulus cells co-cultured with oocytes, synthesis did not decrease to levels found in OOX cumulus cells not co-cultured with oocytes. Thus oocytes acutely promote cholesterol biosynthesis in cumulus cells. Although oocyte-derived factors,probably including BMP15 and GDF9, promote the expression of transcripts encoding enzymes essential for cholesterol synthesis, it appears that an additional interaction between oocytes and cumulus cells is necessary for full capacity to synthesize cholesterol. This additional interaction appears to require contact between oocytes and cumulus cells.

Cholesterol 27 hydroxylase, CYP27A1, functions mainly in liver, converting cholesterol to bile acids (Bjorkhem,1992). The steady-state level of Cyp27a1 mRNA was markedly elevated in Bmp15^-/- and DM cumulus cells and also in WT cumulus cells without oocytes. Oocytes, via BMP15 and GDF9, could therefore suppress cholesterol degradation in cumulus cells and provide another avenue, besides promoting cholesterol biosynthesis, for promoting elevation of cholesterol levels in cumulus cells.

In general, cells accumulate cholesterol from two sources: de-novo synthesis and uptake of extracellular cholesterol via specific receptors for cholesterol carriers. Since evidence is presented here that oocytes are deficient in their ability to produce cholesterol using an endogenous synthetic pathway, they could, theoretically, acquire cholesterol via uptake from their micro-environment via receptor-mediated selective uptake. However,receptors for either HDL-cholesterol (i.e. SCARB1, scavenger receptor class B,member 1, also known as SR-BI) or LDL-cholesterol (i.e. LDLR) are not expressed by mouse oocytes (Sato et al.,2003; Trigatti et al.,1999) suggesting that mouse oocytes are unable to take up carrier-borne cholesterol. What is the source of oocyte cholesterol needed for oocyte development and subsequent embryogenesis? Although cholesterol synthesis was very low in denuded oocytes, much more radiolabeled cholesterol was found in cumulus-cell-enclosed oocytes suggesting that cholesterol was first synthesized by cumulus cells and then transferred to oocytes. It could be argued that cumulus cells stimulate oocytes to synthesize cholesterol. However, this is unlikely because of the poor expression of transcripts encoding the enzymes required for cholesterol synthesis in oocytes. We therefore conclude that a portion of the cholesterol that is either produced or taken up by cumulus cells is transferred to oocytes and that cumulus cells are the source of cholesterol for mouse oocytes.

Do cumulus cells themselves synthesize all of the cholesterol destined for oocytes? SCARB1 is a receptor of HDL cholesterol(Acton et al., 1996). Expression of Scarb1 mRNA was reported to be restricted to theca cells before the LH surge, and detected in granulosa cells only after the LH surge (Li et al., 1998). However, our analysis of the WT cumulus cell transcriptome shows expression of both Scarb1 and Ldlr mRNA. It is therefore possible that these receptors could take up oocyte-destined cholesterol into cumulus cells. However, little LDL cholesterol is present in follicular fluid(Perret et al., 1985; Simpson et al., 1980). Moreover, deletion of Ldlr does not affect fertility in mice(Ishibashi et al., 1993). Scarb1^-/- female mice are infertile(Trigatti et al., 1999). However, the infertility can be reversed by transplanting Scarb1^-/- ovaries to ovariectomized WT recipients, or by lowering the elevated level of plasma cholesterol in Scarb1^-/- mice with the HDL cholesterol-lowering drug probutol (Miettinen et al.,2001). Thus infertility of Scarb1^-/- females is not caused by the absence of SCARB1 in the ovary, but rather indirectly by extra-ovarian defects resulting from the absence of SCARB1(Miettinen et al., 2001). Therefore, although some cholesterol destined for transfer to oocytes could be taken up initially by cumulus cells, cholesterol synthesized by the cumulus cells may be the main source of oocyte cholesterol.

Cholesterol-enriched lipid rafts are present in membranes of mouse oocytes and preimplantation embryos, and treating zygotes with the cholesterol-depleting drug, methyl-β-cyclodextrin, prevented embryonic development beyond 2- to 4-cell stages in culture(Comiskey and Warner, 2007). This indicates that cholesterol deposition in mouse oocytes and embryos is essential for supporting preimplantation development. Furthermore, earlier studies suggested that the full sterol synthetic pathway, i.e., the ability to convert acetate to cholesterol, is not operative in mouse preimplantation embryos until the blastocyst stage (Pratt,1978; Pratt,1982). Therefore, cholesterol and other sterols stored in oocytes are probably required for preimplantation development. Results presented here indicate that cumulus cells provide this cholesterol to oocytes. Therefore,mouse oocytes are promoting their own developmental competence by stimulating cholesterol synthesis in cumulus cells, some of which is then provided to oocytes. Preimplantation development was significantly delayed in DM mice(Su et al., 2004), and this delay could result, at least in part, from lower levels of cholesterol provided to DM oocytes.

We thank Sonya Kamdar and the Jackson Laboratory Gene Expression Service for help with microarray studies, Robert Wilpan of the Protein Chemistry Service for providing facilities and equipment needed for TLC; Anders Friberg and Håkan Billig, Göteborg University, Sweden, for kindly sharing protocols for analysis of cholesterol synthesis in granulosa cells, Bernard Fried, Lafayette College for helpful advice on TLC techniques, and Tom Gridley and Mary Ann Handel for comments and suggestions on this manuscript. Funding was supplied by the following grants: HD23839 (Y.-Q.S., K.S., J.J.E.), HD21970(K.S., K.W., M.J.O., J.J.E.), HD33438 (S.A.P., M.M.M.) and 5F32HD46335(S.A.P.). The Jackson Laboratory Gene Expression Facility is supported by HL072241/Shared Microarray Facilities grant.