Core Binding Factor-β Knockdown Alters Ovarian Gene Expression and Function in the Mouse

Core binding factor (CBF) is a heterodimeric transcription factor complex composed of a DNA-binding subunit, one of three runt-related transcription factor (RUNX) factors, and a non-DNA binding subunit, CBFβ. CBFβ is critical for DNA binding and stability of the CBF transcription factor complex. In the ovary, the LH surge increases the expression of Runx1 and Runx2 in periovulatory follicles, implicating a role for CBFs in the periovulatory process. The present study investigated the functional significance of CBFs (RUNX1/CBFβ and RUNX2/CBFβ) in the ovary by examining the ovarian phenotype of granulosa cell-specific CBFβ knockdown mice; CBFβ f/f * Cyp19 cre. The mutant female mice exhibited significant reductions in fertility, with smaller litter sizes, decreased progesterone during gestation, and fewer cumulus oocyte complexes collected after an induced superovulation. RNA sequencing and transcriptome assembly revealed altered expression of more than 200 mRNA transcripts in the granulosa cells of Cbfb knockdown mice after human chorionic gonadotropin stimulation in vitro. Among the affected transcripts are known regulators of ovulation and luteinization including Sfrp4, Sgk1, Lhcgr, Prlr, Wnt4, and Edn2 as well as many genes not yet characterized in the ovary. Cbfβ knockdown mice also exhibited decreased expression of key genes within the corpora lutea and morphological changes in the ovarian structure, including the presence of large antral follicles well into the luteal phase. Overall, these data suggest a role for CBFs as significant regulators of gene expression, ovulatory processes, and luteal development in the ovary.

In response to the LH surge, a preovulatory follicle undergoes dramatic yet precisely regulated changes in the expression pattern of a myriad of genes to bring about ovulation and luteinization (reviewed in reference 1–3). Transcriptional factors induced by the LH surge in preovulatory granulosa cells have been thought to play a crucial role in ovulation and/or luteal formation and function by directly controlling the transcription of a diverse array of genes in periovulatory follicles. Indeed, recent studies using transgenic mouse models have convincingly demonstrated that the induction of specific transcription factors by the LH surge in periovulatory follicles is essential for successful ovulation and/or luteinization. Such transcription factors include progesterone receptor (4, 5), CCAAT/enhancer-binding proteins (C/EBP)-α and -β (6), and Nr5a2 (7). In addition to these transcription factors, recent studies by our laboratory and others (8–10) shed light on a small family of nuclear transcription factors, the core binding factors (CBF), as transcriptional regulators involved in the periovulatory process.

CBF is a heterodimeric transcription factor complex composed of two subunits: CBFα and CBFβ. There are three genes that encode the α-subunit in mammals: Runx1, Runx2, and Runx3. These genes contain a highly conserved domain in their N-terminal region referred to as the runt homology domain, which is responsible for DNA binding (11, 12). Runt-related transcription factor (RUNX) proteins bind to a consensus recognition sequence (5′-PyGPyGGTPy-3′) (13) in the promoter or enhancer region of target genes and function either as an activator or repressor, depending on the promoter context of particular target genes (reviewed in reference 14). The runt homology domain is also responsible for heterodimerization with the β-subunit, which is encoded by only one mammalian gene, Cbfb (15–17). Although CBFβ does not interact with DNA by itself, it greatly enhances the binding affinity of RUNX proteins to DNA (17). In addition, dimerization with CBFβ protects RUNX proteins from ubiquitin-mediated degradation (18). Thus, CBFβ is essential for the functional activity of RUNX proteins (19–21).

Evidence accumulating from knockout mouse models has demonstrated that CBFs play a crucial role in the development and proliferation of many tissue types, primarily those of mesodermal origin, with each of the RUNX proteins playing a unique, nonredundant role. For instance, mice lacking the Runx1 gene die at the fetal stage due to hemorrhaging in the central nervous system and lack of fetal liver hematopoiesis (22, 23). Runx2 null mice die of respiratory failure shortly after birth and show no bone formation (24, 25). A null mutation of Cbfb also resulted in embryonic lethality due to failure in hematopoiesis and bone formation (19, 26, 27). In the ovary, we have documented that the expression of Runx1 and Runx2, but not Runx3, is dramatically increased by the LH surge in granulosa cells and cumulus cells of preovulatory follicles in naturally cycling rats and by human chorionic gonadotropin (hCG) stimulation in pregnant mare serum gonadotropin (PMSG)-stimulated immature rats (8, 9). Similarly, Shimada et al (10) has shown that hCG increases the levels of mRNA for Runx1 and Runx2 in granulosa cells isolated from mouse ovaries. Using a rat granulosa cell culture model, we have demonstrated that the knockdown of Runx1 or Runx2 expression by small interfering RNA affects the expression of several key ovulatory and/or luteal genes, suggesting that these transcription factors are involved in the ovulatory process and/or luteinization in rat ovaries (8, 9). However, the physiological significance of RUNX transcription factors in ovulation and/or luteinization in vivo remains yet to be determined.

There are challenges in demonstrating the functional significance of RUNX transcription factors in the ovary in vivo: 1) null mutation of the Runx1 or Runx2 gene resulted in embryonic or neonatal lethality, respectively (22–25), 2) both Runx1 and Runx2 are expressed in granulosa and cumulus cells of periovulatory follicles (8, 9), and 3) RUNX1 and RUNX2 proteins have been shown to display functional redundancy in rat granulosa cells (28). To alleviate these problems, we generated granulosa cell-specific knockout of the Cbfb gene in mice using a loxP/cre recombinase system. Because CBFβ is a binding partner for both RUNX1 and RUNX2, the deletion of Cbfb is expected to greatly diminish the activity of both CBF complexes (RUNX1/CBFβ and RUNX2/CBFβ) in the mouse periovulatory ovary.

Using this mutant mouse model, we tested our hypothesis that CBFβ plays an important role in the transcriptional regulation of a specific array of periovulatory genes involved in ovulation as well as luteal formation and function in the mouse. Because the expression of Runx1, Runx2, and Cbfb has not been well characterized in the mouse ovary, we first examined the expression profiles of these genes during the periovulatory period using gonadotropin-primed immature mice. Second, to assess the functional significance of CBFβ, the ovarian phenotype of granulosa cell-specific Cbfb knockout mice was evaluated by examining possible defects in fertility, ovulation, and luteinization. Lastly, we determined whether the deletion of Cbfb affects the transcription of ovulatory and/or luteal genes.

Materials and Methods

Animals

All animals were treated in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Animal protocols were approved by the University of Kentucky Animal Care and Use Committees. Immature C57BL/6 mice were obtained from Harlan, Inc. Cbfb flox (20) and Cyp19 Cre (6) mutant mice were used to generate granulosa cell-specific deletion of the Cbfb gene. All mice were maintained on a 12-hour light, 12-hour dark cycle with water and food ad libitum at the University of Kentucky Division of Laboratory Animal Resources. For the gonadotropin-induced preovulatory model, mice (25–27 d old) were injected with PMSG (5 IU, ip), followed 48 hours later with hCG (5 IU, ip). To determine the stage of the estrous cycle, vaginal fluid and cell samples were obtained from adult female mice (3 mo old) daily for 45 days and examined microscopically.

Genotyping

To genotype mice, ear punches were collected and genomic DNA isolated using the Accustart II mouse genotyping kit (Quanta BioSciences, Inc) according to the manufacturer's instructions. PCR was conducted with primers for Cbfb or Cyp19 Cre listed in Table 1, and amplified PCR products were run on 2% agarose gel for examination.

Collection and culture of mouse granulosa cells

Immature female mice (∼25 d old) were injected with 5 IU PMSG to stimulate follicle development. Forty-eight hours later, animals were killed and ovaries collected. Granulosa cells were isolated from the ovaries via follicular puncture as described previously (29) and plated in OptiMEM (Life Technologies) supplemented with gentamycin and 1× insulin/transferrin/selenium and then treated with hCG (1 IU/mL) and cultured for 12 or 24 hours.

Quantification of mRNA expression

Total RNA was isolated from whole mouse ovaries using a Trizol reagent (Invitrogen) and from cultured mouse granulosa cells using an RNeasy minikit (QIAGEN, Inc). Levels of mRNA for genes of interest were measured by quantitative RT-PCR according to the method described previously (29). Relative amount of transcripts was calculated via the 2^-δδCT method (30), normalizing to the mouse gene Rpl19. Oligonucleotide primers for all genes analyzed were designed using the PRIMER3 Program (San Diego Supercomputer Center) and primer sequences are listed in Table 1.

In situ localization of Runx1, Runx2, Cbfb and Lhcgr mRNA

Ovaries were collected from PMSG/hCG-primed immature mice at defined times after hCG injection or from cycling mice (2–3 mo old) on the day of estrus. Frozen ovaries were sectioned at 8 μm and mounted on Probe On Plus slides (Fisher Scientific). In situ hybridization analysis was carried out as described previously (31). Briefly, partial cDNA fragments were amplified using primers for mouse Runx1, Runx2, and Cbfb using total RNA samples isolated from ovaries 12 hours after hCG. The amplified PCR fragment was cloned into pCRII-TOPO vector. Sequences of the cloned DNA were verified commercially (Eurofins Genomics). Plasmids containing partial cDNA for mouse Runx1, Runx2, Cbfb, and Lhcgr (supplied by Dr Joanne Richards, Baylor College of Medicine, Houston, Texas) were linearized using the appropriate restriction enzymes. Sense and antisense riboprobes were synthesized using the corresponding linearized plasmids and labeled with [α-³⁵S]uridine 5-triphosphate (10 mCi/mL) or fluorescein-12-uridine 5-triphosphate. The ovarian sections hybridized with fluorescein-labeled probes were incubated with an antifluorescein antibody (Roche Applied Sciences), and signals were amplified using a TSA-plus fluorescein kit (Roche Applied Sciences). The sections were counterstained with propidium iodide for 20 minutes. Sections hybridized with the ³⁵S-labeled riboprobes were dipped in Kodak NTB2 emulsion (Eastman Kodak), exposed at 4°C for 4 weeks, developed with Kodak D19, and counterstained with a hematoxylin solution. Specific signals were visualized with an Eclipse E800 Nikon microscope under fluorescent optics or bright- and dark-field optics.

Immunohistochemical analyses and Oil-Red O staining

Ovaries collected from PMSG/hCG-primed immature mice at 3 days after hCG or from cycling mice (2–3 mo old) on the day of estrus were embedded in optimal cutting temperature compound (Sakura Finetek USA, Inc) and stored at −70°C. Frozen ovaries were sectioned at 10 μm and mounted on Probe On Plus slides (Fisher Scientific). Sections were fixed in cold acetone for 10 minutes, pretreated with normal donkey serum, and then treated with primary antibodies for secreted frizzled-related protein 4 (SFRP4; 1:250, HPA009712; Sigma-Aldrich), RUNX1 (1:100, sc-8563; Santa Cruz Biotechnology), RUNX2 (1:200, sc-10758; Santa Cruz Biotechnology), or cytochrome P450 (1:250, ABS235; Millipore) for 16 hours at 4°C. Sections were rinsed with PBS, incubated with Alexa Fluor donkey antirabbit IgG (A-21206) or Alexa Fluor donkey antigoat IgG (A-11055) as appropriate (Life Technologies), counterstained with propidium iodide, and mounted with a mounting medium (fluorogel with 1,4-diazabicyclo[2.2.2]octane; EMS Supply). Digital images were captured using an Eclipse E800 Nikon microscope, with exposure time kept constant for sections incubated with the same primary antibody.

For Oil-Red O staining, sections were mounted on Probe On Plus slides (Fisher Scientific), fixed in 4% paraformaldehyde for 5 minutes, incubated in 60% isopropyl alcohol for 5 minutes, stained with Oil-Red O (Abcam) for 10 minutes, washed in 60% isopropyl alcohol, counterstained with hematoxylin, and mounted with a mounting medium.

Western blot analysis

Whole-cell lysate was extracted from the granulosa cells of Cbfb f/+, Cbfb f/f, and Cbfb f/f * Cyp19 Cre mouse ovaries collected at 12 hours after hCG using a nuclear extraction kit (Active Motif). Cell lysates were denatured by boiling for 5 minutes, separated using SDS-PAGE, and transferred onto a nitrocellulose membrane. Membranes were incubated with a primary antibody against CBFβ (Abcam) overnight at 4°C. β-Actin (Cell Signaling Technology Inc) was used as a loading control. Blots were incubated with the respective secondary horseradish peroxidase-conjugated antibody (Santa Cruz Biotechnology) for 1 hour. Peroxidase activity was visualized using the SuperSignal West Pico chemiluminescent substrate (Pierce Chemical Co).

RNA sequencing and transcriptome assembly

Total RNA was extracted from cultured granulosa cells using an RNeasy minikit (QIAGEN), treated with deoxyribonuclease I and run on a 1% agarose gel to verify its integrity. Library construction and RNA sequencing were conducted by the DNA Services division of the Roy J. Carver Biotechnology Center at the University of Illinois (Chicago, Illinois). Briefly, strand-specific RNA sequencing libraries for 12 individual samples were prepared using a TruSeq stranded RNA sample preparation kit (Illumina). These samples represent three independent samples for each of the four conditions examined. Libraries were pooled in equimolar concentration, and the pool was quantitated by quantitative PCR and sequenced on two lanes on a HiSeq2500 using a HiSeq SBS sequencing kit version 4. The sequencing produced more than 466 million single reads of excellent quality. Fastq files were generated and demultiplexed with the bcl2fastq version .8.4 Conversion Software (Illumina). Adaptor sequences were trimmed from raw RNA sequencing reads using a FastX tool kit (http://hannonlab.cshl.edu/). RNA sequencing data were analyzed using Tuxedo pipeline. Briefly, each set of RNA sequencing reads was mapped on the NCBI annotation of mouse genome reference sequence (GRCm38.p4_genomic.fna) using Tophat (version 2.0.10) and Bowtie (version 1.0.0). Cufflinks suite (version 2.1.1) was used to assemble complete transcripts and quantify their expression by examining each Tophat mapping result with the mouse reference whole transcriptome (GRCm38.p4_genomic.gff). Cuffmerge was performed to merge all assembled transcriptomes into a master transcriptome, and then Cuffdiff was used to find differentially expressed transcripts between RNA sequencing samples. The level for each transcript was expressed as fragments per kilobase of transcript per million mapped reads (FPKM). A P value was adjusted using a q value (30). We defined the threshold for significant differential expression as a value of q < 0.05.

Immunoassay of progesterone

Concentrations of progesterone in serum samples collected from mutant mice were assayed using an Immulite kit on an Immulite 1000 machine (Siemens Healthcare Diagnostics). Assay sensitivity was 0.2 ng/mL, and the intraassay and interassay coefficients of variation were 6.3% and 9.1%, respectively.

Statistical analysis

All data are presented as mean ± SEM. Data were analyzed using the SPSS program (one way ANOVA or t test as appropriate) to determine significance. If an ANOVA revealed significant effects of treatments, the means were compared by a Duncan's post hoc test. Values were considered significantly different if p < 0.05.

Results

Expression of Runx1, Runx2, and Cbfb in the mouse ovary during the periovulatory period

The expression pattern and localization of components of the CBF complex before or at various time points after hCG injection were examined in PMSG/hCG-primed periovulatory mouse ovaries. Consistent with expression profiles observed for Runx1 and Runx2 in rat periovulatory ovaries (8, 9), the levels of mRNA for Runx1 were highest at 6–12 hours after the hCG injection and declined by 24 hours (Figure 1A). The intense signal for Runx1 mRNA was localized to the granulosa cells of periovulatory follicles at 6 hours after the hCG injection, but the signal was also detected, albeit faintly, in newly forming corpora lutea (CL) at 24 hours after hCG (Figure 1B). Levels of mRNA for Runx2 were also increased after hCG, reaching the highest levels at 12 hours and remaining elevated at 24 hours (Figure 1C). Comparable with the real-time PCR data, a strong signal for Runx2 mRNA was localized to the granulosa cells of periovulatory follicles and newly forming CL (Figure 1D). Cbfb mRNA was ubiquitously expressed and levels were constant throughout all time points examined during the periovulatory period (Figure 1, E and F).

Creation and verification of granulosa cell-specific knockdown of Cbfb

To generate ovarian granulosa cell-specific deletion of the Cbfb gene, Cbfb f/f (20) mice were bred with Cyp19 Cre mice (6). In Cbfb f/f * Cyp19 Cre (hereafter called f/f * cre) mice, exon 5 of the Cbfb gene is flanked by LoxP sites, resulting in a loss of functional CBFβ in Cre recombinase-expressing granulosa cells by deleting portions of the protein necessary for binding RUNX proteins (17).

To determine the efficiency of Cbfb knockout, granulosa cells were obtained from the ovaries of f/+, f/f, and f/f * cre mice at 12 hours after hCG administration and used for RT-PCR with primers spanning exon 5 of the Cbfb gene (Figure 2A). In granulosa cells isolated from ovaries of wild-type mice (f/+ or f/f), two different sizes of PCR fragments corresponding to different mRNA transcript variants were amplified. This revealed the expression of variants 1/2, and 4 (V1, V2, and V4) in the mouse ovary, with V4 as the predominant form in periovulatory granulosa cells. In f/f * cre mice, these intact transcript variants and the truncated transcript (69 bp fragment) resulting from the removal of exon 5 were detected. We estimated approximately 60% knockout efficiency in periovulatory granulosa cells of f/f * cre mice, based on the ratio of intact vs mutant transcripts (Figure 2B). Similarly, Western blot analyses revealed that the levels of CBFβ protein were significantly reduced, but not completely eliminated, in the granulosa cells of periovulatory ovaries obtained from f/f * cre mice. Of note, the magnitude of Cbfb deletion varied among mutant mice (Figure 2C).

To determine whether the reduction of CBFβ protein affects the hCG-induced up-regulation and cellular localization of its heterodimeric partners, the expression of Runx1 and Runx2 was assessed in ovaries of wild-type or f/f * cre mice collected 10 hours after hCG injection. As expected, both RUNX1 and RUNX2 proteins were localized to the granulosa cells of ovulatory follicles in wild-type mice. Similarly, immunopositive staining for RUNX1 and RUNX2 proteins was localized to the granulosa cells of ovulatory follicles in f/f * cre mice (Figure 2D), indicating that knockdown of Cbfb did not visibly alter the localization patterns of RUNX1 and RUNX2 in preovulatory follicles before ovulation. However, RUNX2 staining was markedly reduced in the CL of both PMSG/hCG stimulated immature and mature f/f * cre mice compared with those from wild-type mice (Supplemental Figure 1), indicating that CBFβ knockdown did not prevent hCG-induced induction of Runx1 and Runx2 expression but dampened the sustained expression of Runx2 in CL.

Characterizing the ovarian phenotype of f/f * cre mice

The fecundity of the knockdown mice was tested by mating wild-type (f/+ or f/f) and mutant (f/f * cre) females with fertile males for 6 months. Mutant mice exhibited decreased fertility. Notably, 3 of 15 mutant females were infertile throughout the breeding period. The remaining fertile mutant females (12 mice) produced smaller litters on average and significantly fewer pups overall than wild-type females (Figure 3, A and B). Furthermore, fewer cumulus oocyte complexes (COCs) were collected from f/f * cre mice when superovulation was induced (Figure 3C). Levels of serum progesterone did not differ in PMSG-primed f/f * cre and wild-type females at 12 hours and days 1 and 3 after the hCG injection (Figure 3D), but f/f * cre females showed lower serum progesterone levels than wild-type females throughout pregnancy (Figure 3E and Supplemental Figure 2).

Cbfb knockdown alters the levels of mRNA for Runx1 and Runx2 in granulosa cells in vitro

To further assess the knockout efficiency of Cbfb expression and its impact on Runx1 and Runx2 expression, granulosa cells were isolated from the ovaries of gonadotropin-primed f/f * cre (n = 6) and wild-type (n = 4) mice. The granulosa cells were treated with or without hCG (1 IU/mL) and cultured for 12 or 24 hours. This culture model was used to assess the impact of Cbfb deletion in granulosa cells while minimizing contamination and influence of other types of cells. Similar to the data shown in Figure 2B, the levels of Cbfb mRNA were 50%–60% reduced in f/f * cre mice compared with the wild-type mice. Also, hCG had no effect on the levels of Cbfb mRNA in granulosa cell cultures (Figure 4A). Consistent with the in vivo data, Runx1 expression was increased by hCG treatment at 12 hours and returned to control levels by 24 hours in the granulosa cells of wild-type mice. Meanwhile, the levels of Runx1 mRNA were elevated significantly above the wild-type in granulosa cells from f/f * cre mice without hCG stimulation and remained elevated at 24 hours (Figure 4B). Runx2 mRNA levels were also increased by hCG at 12 hours and continued to increase at 24 hours in wild-type granulosa cells. However, in f/f * cre mice, the levels of Runx2 mRNA were reduced (Figure 4C). These data suggest that Cbfb knockdown affected transcription of Runx1 and Runx2 in granulosa cells.

Cbfb knockdown alters the levels of gene transcripts in granulosa cells

To further identify genes whose expression is differentially regulated by knockdown of Cbfb expression, granulosa cells from wild-type (n = 3) and f/f * cre (n = 3) mice were cultured with hCG for 12 and 24 hours and total RNA isolated from these cells was used for RNA sequencing. Transcriptomes were assembled and differentially expressed genes were analyzed using Tophat and Cufflinks programs. The Cuffdiff analysis revealed 2476 and 3391 transcripts as differentially expressed (q < 0.05) in cultured granulosa cells of wild-type and f/f * cre mice obtained at 12 and 24 hours, respectively. It is important to note that included in these data are several genes previously identified as targets of RUNX1 or RUNX2. Among them are Cyp11a1 (9), Abcb1a (8), and Ptgs2 (32). A selection of those genes highly down- or up-regulated is depicted in Figures 5 and 6, respectively. All genes meeting the further criteria of fold change of 2 or greater and FPKM value of 10 or greater for at least one time point analyzed are listed in Supplemental Tables 1–3. These criteria were chosen to identify genes with both dramatic changes in expression and high levels of expression. Among the differentially expressed transcripts, 82 and 185 met these criteria at 12 and 24 hours, respectively.

A selection of genes with decreased expression in f/f * cre granulosa cells are depicted in Figure 5. Selected genes were divided into two groups based on their expression profile in the wild type: one group showing the highest expression at 24 hours after hCG treatment (Figure 5, A–D) and the other showing transient expression, peaking at 12 hours and decreasing by 24 hours after hCG (Figure 5E). Notable among the genes showing a decreased expression in granulosa cells of f/f * cre mice are Lhcgr, Sfrp4, Prlr, Wnt4, and Edn2. These genes are known to play roles in ovulation or luteal formation and function (2, 33–37). Additionally, these data include several key steroidogenic genes such as Cyp11a1, Hsd3b1, and StAR (Figure 5D). Many of the genes identified, however, have not been characterized in terms of their role in the ovary.

In addition to the genes down-regulated in f/f * cre mice, many genes were also found to be up-regulated in granulosa cells by the knockdown of Cbfb expression (Figure 6). Of these up-regulated genes, one group of genes demonstrated their peak expression in the wild type at 12 hours after hCG stimulation (Figure 6, A and B), whereas another group exhibited constant or increased levels of expression over 24 hours (Figure 6, C–F). Notable among the up-regulated genes are members of the epidermal growth factor family (eg, Areg, Btc, and Ereg) and members of the inhibin protein complex (Inha and Inhba).

Among the genes identified as differentially regulated using RNA sequencing, we further selected 15 genes to determine whether these genes are regulated by hCG and to verify RNA sequencing results via real-time PCR. These genes are marked in red in Figures 5 and 6. Levels of mRNA for genes selected were significantly altered in f/f * cre samples, verifying RNA sequencing results (Supplemental Figure 3). Treatment with hCG increased the levels of mRNA for most these genes, except for Mlh1 and Hsd3b1.

Cbfb knockdown alters luteal gene expression

Because several well-known luteal genes were down-regulated in cultured granulosa cells of f/f * cre mice, we further determined whether Cbfb knockdown affects expression of these genes in the corpus luteum in vivo. Expression was examined in ovaries of wild-type and f/f * cre mice obtained at 3 days after hCG injection. The levels of mRNA for Cbfb, Lhcgr, and Sfrp4 were lower in f/f * cre mice compared with those in wild-type mice (Figure 7A). Several other genes examined failed to reach the threshold of statistical significance between wild-type and f/f * cre mice, likely due to the high variation among samples. This variation may be due to differences in peak expression time of these genes in the corpus lutea or the expression of these genes in other cell types in addition to luteal cells in the whole ovary.

To compare luteal expression of Lhcgr mRNA and SFRP4 protein and ovarian morphology between mutant and wild-type mice, ovaries from 3 days after hCG stimulation were examined. Compared with the ovaries of the wild type, f/f * cre mice showed reduced levels of Lhcgr mRNA in CL, with the extent of reduction varying among individuals (Figure 7B). This variation likely reflects the differing levels of Cbfb knockdown observed previously in Figure 2C. Expression of SFRP4, a well-known luteal marker (38), was compared between wild-type and f/f * cre mice by immunohistochemistry (Figure 7C). As expected, SFRP4 staining was localized to corpus lutea in ovaries of both wild-type and f/f * cre mice. However, SFRP4 staining was strong and consistent throughout the CL in wild-type mice, whereas in f/f * cre mice, many cells at the center of CL failed to stain for SFRP4, potentially indicating failure of a subpopulation of granulosa cells to reach their terminally differentiated state as luteal cells. Noteworthy is the presence of large antral follicles in ovaries of f/f * cre mice at 3 days after hCG administration, implying that at least some of the stimulated follicles failed to ovulate (Figure 7C).

Cbfb knockdown alters gene expression and lipid accumulation in CL of cycling animals

Next, we determined the impact of Cbfb knockdown in CL of cycling mice by examining the expression of Lhcgr and SFRP4. Ovaries of wild-type and f/f * cre mice (2–3 mo old) were collected on the morning (10:00 am) of estrus, a time when ovulation has already occurred and newly forming CL are easily distinguished from CL generated from previous cycles in the ovary. As shown in Figure 8A, the expression of Lhgcr mRNA was localized to newly forming CL (Figure 8A, arrows) but not in CL from previous cycles (Figure 8A, arrowheads). Moreover, the signal for Lhgcr mRNA in newly forming CL was reduced in ovaries of f/f * cre mice (Figure 8A), similar to our findings in PMSG/hCG-primed mice in Figure 7B. In contrast to the Lhgcr mRNA expression profile, positive staining for SFRP4 protein was detected in all CL, with the highest intensity in CL generated from previous cycles (Figure 8B, arrowheads). There was marked reduction of SFRP4 staining intensity in CL of f/f * cre mice compared with wild-type mice (Figure 8B and Supplemental Figure 4).

In addition to the difference in specific luteal gene expression, many CL in f/f * cre mouse ovaries stained heavily with Oil-Red O (used for staining of neutral fat and lipid droplets), whereas Oil-Red O staining was less visible in the CL of wild-type mice (Figure 8, C and D). Similar heavy Oil-Red O staining was observed in the CL of mature (5 mo old) and pregnant f/f * cre mice (Supplemental Figure 5). This excessive lipid accumulation in CL was most noticeable in those f/f * cre mice that were infertile for 6 months of breeding (Supplemental Figure 5C).

Discussion

CBFβ is a common non-DNA binding partner for all RUNX proteins and is critical for DNA binding activity and stability of the CBF transcription factor complex (17, 18). Previous studies have documented increases in Runx1 and Runx2 expression after the LH surge in periovulatory follicles of rats (8, 9). To determine the physiological impact of RUNX1 and RUNX2 in the ovary, we used a granulosa cell-specific Cbfb knockout mouse model in which the activity of both CBF complexes (RUNX1/CBFβ and RUNX2/CBFβ) is impeded. The present study demonstrates for the first time that CBFβ plays a crucial role in normal ovarian function; granulosa cell-specific Cbfb mutant mice were subfertile, showed evidence of compromised ovulation and luteinization, and displayed altered expression of many known ovulatory and luteal genes.

Knockdown of CBFβ led to several phenotypic changes in female mice. Most notably, 20% of f/f * cre mice were infertile over a 6-month breeding period. Mice maintaining some level of fertility displayed a subfertile phenotype characterized by decreased litter size and cumulative number of pups, fewer COCs collected after superovulation, and reduced progesterone levels during gestation. The infertility of a considerable percentage of the mice assayed may be explained by the variation among individuals in the degree of Cbfb knockdown. Mice exhibiting complete infertility likely fell below the threshold at which the transcriptional activity of CBF complexes is no longer sufficient to sustain normal function. Additionally, daily vaginal cytology of cycling mice revealed that f/f * cre females have significantly longer estrous periods and a shorter duration of diestrus than wild-type mice (Supplemental Figure 6), indicating some degree of disturbance in the estrous cycle caused by Cbfb knockdown.

Our data also provide evidence of auto- or cross-regulation for the expression of RUNX family members. Whereas the levels of Runx1 mRNA were elevated, the hCG-induced increase in Runx2 mRNA was down-regulated in the granulosa cells of f/f * cre mice, indicating that CBFs exert opposing effects on their own gene expression in periovulatory granulosa cells. In support of this finding, a previous study using rat granulosa cell cultures showed that Runx2 suppresses the transcription of Runx1 by directly binding to its promoter region (32). Taken together, these data indicate that RUNX2/CBFβ facilitates the rapid down-regulation of Runx1 expression after ovulation by acting as a repressor while enhancing its own gene expression in periovulatory granulosa cells and/or luteal cells.

In addition to altering the expression of its binding partners, Cbfb knockdown significantly altered the expression of numerous other transcripts. The genes identified as differentially regulated are diverse, including cell membrane receptors (Lhcgr, Prlr, Ptgfr, Ptger2) and nuclear receptors (Esr2), factors involved in lipid metabolism and steroidogenesis (Lipg, Cyp11a1, Cyp11b1, Cyp19a1, Hsd17b7, Hsd3b1, StAR, Abcb1a), and members of the epidermal growth factor family (Areg, Ereg, Btc). Of particular interest are genes known to be involved in ovulation and/or luteinization. For instance, we found that expression of Edn2 is markedly down-regulated in Cbfb knockdown mice. Edn2 expression is selectively and highly induced in the granulosa cells of the periovulatory follicles of mice (39). A more recent study by Cacioppo et al (40) further demonstrated that the loss of Edn2 resulted in reduced ovulation and CL formation in mice. Expression of Lhcgr, which was found to be necessary for ovulation, steroidogenesis, and fertility (41), was also reduced in the CL of both gonadotropin-primed and cycling Cbfb knockdown mice. Transcripts for several key luteal cell-specific genes, such as Prlr, Cyp11a1, Sgk1, Wnt4, and Sfrp4 were also found to be down-regulated in Cbfb knockdown mice (37, 38, 42–44).

Along with reduced expression of specific luteal genes, f/f * cre mice showed atypical ovarian morphological appearances. For instance, large antral follicles were present in the ovary during the luteal phase (Figure 7C and Supplemental Figure 1), suggesting a failure of some follicles to undergo ovulation, perhaps due to reduced expression of key mediators of ovulation. These follicles subsequently fail to develop into CL. In the CL of f/f * cre mice, a portion of cells in the center do not show positive staining for SFRP4, suggesting a failure to fully differentiate to luteal cells. Lastly, in f/f * cre mice a high percentage of CL show excessive lipid accumulation, suggesting abnormal lipid metabolism. Taken together, these data indicate that CBFβ plays a critical role in ovulation and luteal development and/or function by up-regulating the expression of key ovulatory and luteal-specific genes.

In addition to down-regulated genes, a significant number of transcripts had elevated levels in the granulosa cells of Cbfb knockdown mice. Among them are transcripts for epidermal growth factor-like peptides (Areg, Ereg, Btc) and Sult1e1. Notably, the expression of these genes was reported to be transiently induced by hCG in the granulosa cells of periovulatory follicles (45–47). Moreover, these genes are known to play a crucial role in the ovulatory process (30, 48, 49). Considering the fact that Runx2 expression continued to increase after ovulation, these data suggest that RUNX2/CBFβ is involved in the down-regulation and/or suppression of ovulatory-specific genes during the late ovulatory period, thereby facilitating the ovulatory to luteal transition and CL formation.

Another intriguing finding is the considerable overlap of genes identified as targets of CBFs in our data and those previously identified as differentially regulated in the ovaries of Cebpa/b knockout mice (6), indicating possible cooperation between these factors to initiate the characteristic transcriptional program of periovulatory follicles. Indeed, in Cebpa/b knockout mice, the levels of Runx2 expression were markedly reduced (6), whereas in our study Cbfb knockdown resulted in reduced mRNA levels for Cebpb and elevated expression of Cebpa (Supplemental Table 4), providing further evidence for reciprocal regulation between these transcription factors, which have been shown to physically interact to induce transcription (50).

Overall, this study provides the first in vivo evidence demonstrating that CBF transcription factors are important transcriptional regulators crucial for successful ovulation and luteal formation/function in mice. Using a unique in vivo model of granulosa cell-specific Cbfb knockdown, we have verified several previously known RUNX1 or RUNX2 downstream target genes and, more importantly, identified a number of new CBF-regulated genes in granulosa cells of the periovulatory ovary.

Acknowledgments

We thank Drs Patrick Hannon and Yohan Choi and Ms Katherine Rosewell for their critical reading of the manuscript.

This work was supported by National Institutes of Health (NIH) Grants RO1HD061617, RO3HD066012, and PO1HD71875.

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• CBF

  core binding factor

 
• C/EBP

  CCAAT/enhancer-binding protein

 
• CL

  corpora lutea

 
• COC

  cumulus oocyte complex

 
• FPKM

  fragments per kilobase of transcript per million mapped reads

 
• hCG

  human chorionic gonadotropin

 
• PMSG

  pregnant mare serum gonadotropin

 
• RUNX

  runt-related transcription factor

 
• SFRP4

  secreted frizzled-related protein 4.

References