G_q/11-Dependent Changes in the Murine Ovarian Transcriptome at the End of Gestation

Abstract

Parturition in rodents is highly dependent on the engagement of the luteal prostaglandin F2 alpha receptor, which, through activation of the G_q/11 family of G proteins, increases the expression of Akr1c18, leading to an increase in progesterone catabolism. To further understand the involvement of G_q/11 on luteolysis and parturition, we used microarray analysis to compare the ovarian transcriptome of mice with a granulosa/luteal cell-specific deletion of Galpha_q/11 with their control littermates on Day 18 of pregnancy, when mice from both genotypes are pregnant, and on Day 22, when mice with a granulosa/luteal cell-specific deletion of Galpha_q/11 are still pregnant but their control littermates are 1–2 days postpartum. Ovarian genes up-regulated at the end of gestation in a Galpha_q/11 -dependent fashion include genes involved in focal adhesion and extracellular matrix interactions. Genes down-regulated at the end of gestation in a Galpha_q/11-dependent manner include Serpina6 (which encodes corticosteroid-binding globulin); Enpp2 (which encodes autotaxin, the enzyme responsible for the synthesis of lysophosphatidic acid); genes involved in protein processing and export; reproductive genes, such as Lhcgr; the three genes needed to convert progesterone to estradiol (Cyp17a1, Hsd17b7, and Cyp19a1); and Inha. Activation of ovarian G_q/11 by engagement of the prostaglandin F2 alpha receptor on Day 18 of pregnancy recapitulated the regulation of many but not all of these genes. Thus, although the ovarian transcriptome at the end of gestation is highly dependent on the activation of G_q/11, not all of these changes are dependent on the actions of prostaglandin F2 alpha.

INTRODUCTION

It is now clear that at least two families of G proteins, G_s and G_q/11, are essential to maintain the functions of granulosa and luteal cells. G_s activation by the luteinizing hormone (LH) receptor in granulosa cells is required for initiating the events leading to the resumption of meiosis and the expansion of cumulus cells, but it is also required for the induction of the progesterone receptor (Pgr) and subsequent rupture of the ovarian follicles [1, 2]. The activation of G_q/11 by the LH receptor in granulosa cells is not required for the resumption of meiosis and the expansion of cumulus cells [3]. Together with the LH receptor-activated G_s, however, the LH receptor-activated G_q/11 is required for optimal induction of the granulosa cell Pgr, follicular rupture, and an optimal ovulatory response [1, 3]. Thus, genetic deletion of Gα_q/11 from granulosa cells compromises the LH receptor-induced expression of the granulosa cell Pgr and ovulation [3].

Studies with mice harboring a granulosa/luteal cell-specific deletion of Gα_q/11 have demonstrated that the differentiation of granulosa cells into luteal cells, as well as the establishment and maintenance of pregnancy, does not require G_q/11 [3, 4]. Parturition, on the other hand, is dependent on the prostaglandin F2α (PGF2α)-provoked activation of luteal G_q/11 [4]. A conditional deletion of Gα_q/11 from granulosa/luteal cells disrupts the actions of PGF2α on luteal cells that normally occur toward the end of gestation [4]. The ovaries of pregnant mice with a conditional deletion of Gα_q/11 from granulosa/luteal cells display a defect in the PGF2α-provoked induction of Ark1c18, the gene encoding for 20α-hydroxysteroid dehydrogenase, the enzyme that catabolizes progesterone to 20α-hydroxyprogesterone [4]. As such, the mice exhibit a delay or failure of parturition because they do not experience the reduction in progesterone levels that precipitate parturition [4].

The crucial role of luteal cell Gα_q/11 in the expression of luteal Ark1c18 at the end of gestation prompted us to examine the hypothesis that luteal cell G_q/11 is also involved in the expression of other ovarian genes that may participate in luteolysis and parturition. This is an important question, because in addition to activating G_q/11, the PGF2α receptor also activates G_12/13 [5], and because there are other hormones, such as oxytocin [6, 7] or lysophosphatidic acid [8, 9], that can activate luteal G_q/11. To test this hypothesis we compared the ovarian transcriptome of mice with a granulosa/luteal cell-specific deletion of Gα_q/11 with the transcriptome of their control littermates on Days 18 and 22 of pregnancy. The results presented show that changes in the ovarian transcriptome that occur at the end of gestation (Day 22 of pregnancy, when mice lacking Gα_q/11 still have not given birth but their control littermates have) are highly dependent on Gα_q/11. Not all of the Gα_q/11-dependent genes are dependent on the actions of PGF2α, however.

MATERIALS AND METHODS

Mice

A colony of Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice were generated by crossing Gα_q^f/f;Gα₁₁^−/− mice and Cyp19Cre transgenic mice as previously described [3]. The colony of Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice was maintained by crossing Gα_q^f/f;Gα₁₁^−/−;Cre⁺ males with Gα_q^f/f;Gα₁₁^−/−;Cre⁻ females [4]. The resulting Gα_q^f/f;Gα₁₁^−/−;Cre⁺ and Gα_q^f/f;Gα₁₁^−/−;Cre⁻ females were used as experimental and control animals, respectively. Genotyping was performed using tail genomic DNA and PCR amplification as described previously [10–14].

Adult females (ages 6–10 wk) were synchronized on estrus by injecting them with progesterone and cloprostenol [15], housed overnight with males of proven fertility, and checked the following morning for vaginal plugs. Day 1 of pregnancy (1 dpc) was counted as the morning when a vaginal plug was found, and the pregnant animals were killed in the morning on Days 18 or 22 of pregnancy as indicated. In some experiments pregnant mice (17 dpc) were injected subcutaneously with vehicle only or 1 μg of cloprostenol, a PGF2α agonist [16], and killed 24 h later for the preparation of ovarian RNA. All animal procedures were approved by the Institutional Animal Care and Use Committee for the University of Iowa.

cAMP Assays

Luteal cells were prepared from the ovaries of mice on 18 or 22 dpc exactly as described previously for pseudopregnant ovaries [4]. The freshly isolated luteal cells were immediately suspended in 500 μl of Dulbecco modified Eagle medium/F12 supplemented with 10 mM Hepes, bovine serum albumin (1 mg/ml), and 1 mM isobutyl methylxanthine and incubated without or with human chorionic gonadotropin (hCG; 100 ng/ml), for 30 min at 37°C. Total cAMP (cells + medium) was then extracted and quantitated as described earlier [17–19], except that we used a commercially available enzyme immunoassay kit from Cayman Chemical instead of a radioimmunoassay. All samples were assayed in a single assay and the intraassay coefficient of variation was 10%.

Microarray Analysis

Transcription profiling was performed in triplicate (n = 3) for each genotype on 18 dpc and 22 dpc, respectively. The profiling was done using ovarian RNA purified from the ovaries and the mouse WG-6 V2.0 Illumina platform. All steps of the transcription profiling (except for the isolation of the RNA) and analyses were performed by the Iowa Institute of Human Genetics. Gene enrichment analysis was carried out using publicly available software (http://bioinfo.vanderbilt.edu/webgestalt).

RNA Isolation and Real-Time PCR

Ovarian RNA was prepared as described earlier [4]. Target and control gene (Gapdh) expressions were measured by quantitative PCR (qPCR) using 5–50 ng of purified RNA, the iTaq Universal SYBR Green One Step Kit (Bio-Rad Laboratories), and a CFX90 real-time PCR detection system (Bio-Rad Laboratories). Each experimental group included a minimum of five replicate animals (see Figure legends), and each sample was run in duplicate on the qPCR reactions. Data were quantitated by the comparative C_q method as described previously [20], using Gapdh as control or normalizer gene, and are always expressed relative to a control group as indicated in each of the figures. There was no statistical difference among the average quantification cycles (C_q) for the control gene (Gapdh) for any group or condition.

The primers used were as follows: Lhcgr: forward, CCTTGTGGGTGTCAGCAGTTA; reverse, TTGTGACAGAGTGGA TTCCACAT; Cyp19a1: forward, GACGGGCCCTGGTCTTGT; reverse, CCGGTCCAAATGCTGCTT; Cyp11a1: forward, CCCGGAGCGGTTCCTT; reverse, CCAATGGGCCTCTGATAATACTG; Cyp17a1: forward, TGGAGGCCACTATCCGAGAA; reverse, CACATGTGTGTCCTTCGGGA; Inha: forward, TGTGGGTAAAGTGGGGGAGA; reverse, GTCTTTGGTGGTGTCTGCGA; Hsd17b7: forward, AGGCGTCGTGATGACCAATA; reverse, AAAAAGCGAAGGAGCCACATT; Enpp2: forward, GGGCTGCACCTGTGATGATA; reverse, GATGTCTCTCTTCTGTAGATCCTT; Serpina6: forward, TTCCAAGGCAGATGAACCTGTA; reverse, GTCTTTGGTGGTGTCTGCGA; Wfs1: forward, TCGTCAGCAGTGAATCCAAGAA; reverse, AATGATGCCCTTGGCGTACT; Jun: forward, CTTGTGCCCCAAGAACGTG; reverse, GTGACACTGGGAAGCGTGTT; Tnc: forward, CAATAACCACAGTCAGGGCGT; reverse, CCCTGGAATTTATGCCCGCT; Ssr4: forward, TGGCTCTTTATGCCGACGTT; reverse, AGGCTCCAGGACACCTGATA; and Gapdh: forward, GACGGCCGCATCTTCTTG; reverse, ACCGACCTTCACCATTTTGTCT.

Hormone Assays

Serum estradiol and Inhibin A concentrations were measured using ELISA by the Ligand Assay and Analysis Core of the University of Virginia. The measurable ranges for estradiol and inhibin A are 3–300 pg/ml and 10–950 pg/ml, respectively. The intraassay and interassay coefficients of variation for estradiol are 6.1% and 8.9%, respectively, and the corresponding values for Inhibin A are 1.5% and 4.1%.

RESULTS

Microarray Analysis

Ovarian RNA was prepared from Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ females on 18 dpc (three mice of each genotype) and 22 dpc (three mice of each genotype). On 18 dpc, mice from both genotypes are pregnant. Because the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice give birth on 20–21 dpc, they are 1–2 days postpartum on 22 dpc. The Gα_q^f/f;Gα₁₁^−/−;Cre⁺ are still pregnant on 22 dpc because of the previously described delay or failure of parturition [4]. The 12 RNA samples were subjected to expression profiling using the mouse WG-6 V2.0 Illumina platform. Array data were imported using the Bead array software package (http://www.bioconductor.org/packages/release/bioc/html/beadarray.html) and were quantile normalized. Differential expression was determined using the Limma software package (http://bioconductor.org/packages/release/bioc/html/limma.html).

A multidimensional scale plot of the microarray data (Fig. 1) clearly shows that based on the similarity of gene expression, the 12 mice used for the microarray analysis can be segregated into 4 groups according to genotype and length of gestation. Ovarian gene expression was rather similar between the two genotypes on 18 dpc (when mice from both genotypes are pregnant) but was very different on 22 dpc, when the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice have given birth but the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice have not. Also, the three mice of each length of gestation and genotype clustered very close to each other, but there was more variability in the group of Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 22 dpc than in any of the other three groups.

FIG. 1.

Multidimensional plot of ovarian gene expression in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 18 and 22 dpc. The blue and red circles enclosed in the rectangles highlight the overall similarity of gene expression in the individual ovaries of the four groups of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice analyzed on 18 and 22 dpc as indicated. On 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

Because we used 12 mice and there are a large number of probes represented in the mouse WG-6 V2.0 Illumina chips (∼45 000), the results of the differential expression can be analyzed using rather stringent criteria for statistical significance (P < 0.001). A comparison of gene expression between the two genotypes on 18 dpc revealed no differentially expressed probes, either when using this criterion or when a more relaxed (P < 0.005) level of significance was used. A comparison of gene expression in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 22 and 18 dpc using P < 0.001 generated a list of 288 differentially expressed probes. This group of probes was then ranked to include only probes that were differentially expressed by at least 100-fold (2 log-fold), to generate the heat map presented in Figure 2. As expected from the above discussion, gene expression was generally similar in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁺ and Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 18 dpc. This map also clearly shows two clusters of probes (shown in red, about 40 probes) that were down-regulated in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ on 22 dpc, and one cluster of probes (shown in green, about 20 probes) that was up-regulated in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 22 dpc. The down- or up-regulation of these clusters of probes was not obvious in the ovaries of the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 22 dpc, however.

FIG. 2.

Heat map of ovarian gene expression in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 18 and 22 dpc. The probes included in this heat map were first selected by comparing gene expression in the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 22 and 18 dpc. On 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum. Only those probes that were differentially expressed with a P < 0.001 were chosen. This list was sorted again based on the extent of the differential expression, and only those probes that were differentially expressed by at least two orders of magnitude (100-fold) are shown. An increase in expression is shown by the different shades of green, and a decrease in expression is shown by the different shades of red. The clusters shown contain duplicate probes and probes for unknown genes. The identity of the up- and down-regulated genes (after removal of duplicate probes and probes for unknown genes) is shown in Tables 1 and 2.

Manually removing duplicate probes and probes for unknown genes further reduced the up-regulated cluster to the 11 genes shown in Table 1. This table shows that the magnitude of the increased expression of the 11 genes in the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice was much lower or absent compared with the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice (compare the log of fold increase in the Cre⁻/Cre⁻ and Cre⁺/Cre⁺ columns) and, with the exception of Akr1c18 and Chrna1, the up-regulation of the other genes in the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice was no longer statistically significant (adjusted P value <0.001 in the Cre⁻/Cre⁻ column and >0.001 in the Cre⁺/Cre⁺ column). The same type of analysis reduced the down-regulated clusters to the 28 genes shown in Table 2. Again, the decreased expression of these genes was much lower or absent (compare the log of fold decrease in the Cre⁻/Cre⁻ and Cre⁺/Cre⁺ columns) and no longer statistically significant in the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice (adjusted P value <0.001 in the Cre⁻/Cre⁻ column and >0.001 in the Cre⁺/Cre⁺ column).

TABLE 1.

Ovarian genes up-regulated at the end of gestation in a Gα_q/11-dependent fashion by at least two orders of magnitude and P < 0.001.

^aOn 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

^bThese genes were chosen for further analysis as discussed in the text.

TABLE 2.

Ovarian genes down-regulated at the end of gestation in a Gα_q/11-dependent fashion by at least two orders of magnitude and P < 0.001.

^aOn 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

^bThese genes were chosen for further analysis as discussed in the text.

We next used pathway enrichment software to categorize the 138 probes that were individually up-regulated as well as the 150 probes that were individually down-regulated (P < 0.001) into Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Using the default criteria for this analysis (at least two genes per group and P < 0.01 for the group) we identified 34 up-regulated genes that can be grouped into 8 KEGG pathways (Table 3) and 49 down-regulated genes that can be grouped in 7 KEGG pathways (Table 4).

TABLE 3.

Pathway enrichment analysis of ovarian genes that are up-regulated at the end of gestation in a Gα_q/11-dependent fashion.

^aGene symbols identified as being enriched for a given pathway are shown underneath the KEGG pathway name.

^bThese genes were chosen for further analysis as discussed in the text.

TABLE 4.

Pathway enrichment analysis of ovarian genes that are down-regulated at the end of gestation in a Gα_q/11-dependent fashion.

^aGene symbols identified as being enriched for a given pathway are shown underneath the KEGG pathway name.

^bThese genes were chosen for further analysis as discussed in the text.

The entire microarray data set is included herein (MS Excel file) as Supplemental Table S1 (available online at www.biolreprod.org).

Validation of the Microarray Analysis by qPCR

Akr1c18 was the gene with the highest magnitude of Gα_q/11-dependent up-regulation (Table 1), and the validation of this result by qPCR has already been accomplished elsewhere [4]. We chose to further analyze two up-regulated genes, Tnc and Jun, based on a combination of the magnitude of up-regulation (Table 1) and their grouping into the “focal adhesion” KEGG pathway (Table 3). We also chose to further analyze nine down-regulated genes based on the magnitude of down-regulation (Serpina6 and Enpp2; Table 2), their known involvement in reproductive functions (Cyp19a1, Cyp17a1, Hsd17b7, Lhcgr, and Inha; Table 2), their grouping into a KEGG pathway (Wfs1 and Ssr4 in the “protein processing in endoplasmic reticulum” KEGG pathway in Table 4), or a combination of all three criteria—magnitude of down-regulation, their known involvement in reproductive functions, and their grouping into the “steroid hormone biosynthesis” KEGG pathway (Cyp19a1, Cyp17a1, and Hsd17b7; Tables 2 and 4).

Quantitative PCR validation of the microarray data for the up-regulated (Fig. 3) and down-regulated (Figs. 4 and 5) genes on 18 and 22 dpc clearly recapitulated the microarray data. Their expression increased (Fig. 3) or decreased (Figs. 4 and 5) at the end of gestation in the Gα_q^f/f;Gα₁₁^−/−;Cre⁻, mice but the magnitude of this change was blunted or absent in the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice.

FIG. 3.

Quantitative PCR validation of the changes in expression of ovarian Tnc and Jun at the end of gestation in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. For each gene, data are expressed relative to the gene's expression in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 18 dpc (blue bar on the left panel of each graph). Each bar is the mean ± SEM of five different animals analyzed on the indicated days. Statistically significant differences (P < 0.05) were calculated using ANOVA and are shown by the brackets and asterisks. On 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

FIG. 4.

Quantitative PCR validation of the changes in expression of ovarian Serpina6, Enpp2, Wfs1, and Ssr4 at the end of gestation in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. For each gene, data are expressed relative to the gene's expression in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 18 dpc (blue bar on the left panel of each graph). Each bar is the mean ± SEM of 5–10 different animals analyzed on the indicated days. Statistically significant differences (P < 0.05) were calculated using ANOVA and are shown by the brackets and asterisks. Some bars are not visible because of the scale used. On 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

FIG. 5.

Quantitative PCR validation of the changes in expression of the genes known to be involved in reproductive functions at the end of gestation in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. For each gene, data are expressed relative to the gene's expression in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice on 18 dpc (blue bar on the left panel of each graph). Each bar is the mean ± SEM of 8–10 different animals analyzed on the indicated days. Statistically significant differences (P < 0.05) were calculated using ANOVA and are shown by the brackets and asterisks. On 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

Effect of a PGF2α Receptor Agonist on the Expression of Luteal Gα_q/11-Dependent Genes

The engagement of the PGF2α receptor by uterine-derived PGF2α is recognized as the main determinant of luteolysis and parturition in rodents [4, 21–29], and the increased expression of one of the genes found in the microarray (Akr1c18; Table 1) has already been shown to be dependent on the activation of G_q/11 by the PGF2α receptor [4, 26, 27]. To determine whether PGF2α also influenced the expression of the 11 Gα_q/11-dependent genes shown in Figures 3–5, we injected mice of both genotypes with cloprostenol (a PGF2α agonist) on 17 dpc and analyzed ovarian gene expression 24 h later. Figure 6 shows that the expression of only one (Jun) of the two chosen up-regulated genes was increased when pregnant Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice were injected with cloprostenol, and that this response was absent in the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. The expression of ovarian Tnc was not affected by cloprostenol in either genotype.

FIG. 6.

Effect of cloprostenol on the expression of Tnc and Jun in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 17 dpc. Mice were injected with vehicle or 1 μg of cloprostenol on 17 dpc and killed 24 h later. For each gene, data are expressed relative to the gene's expression in the ovaries of mice of the same genotype injected with vehicle only. Each bar is the mean ± SEM of five different animals. Statistically significant differences (P < 0.05) were calculated only by comparing expression in each genotype (t-test) and are shown by asterisks.

The effect of cloprostenol on the down-regulated genes was also mixed (Figs. 7 and 8). This PGF2α agonist decreased the expression of ovarian Serpina6, Ssr4, Enpp2, Lhcgr, Hsd17b7, and Cyp19a1 in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice but not in Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. Cloprostenol had no effect on the expression of ovarian Wfts1, Cyp17a1, or Inha in either mouse genotype.

FIG. 7.

Effect of cloprostenol on the expression of Serpina6, Enpp2, Wfs1, and Ssr4 in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 17 dpc. Mice were injected with vehicle or 1 μg of cloprostenol on 17 dpc and killed 24 h later. For each gene, data are expressed relative to the gene's expression in the ovaries of mice of the same genotype injected with vehicle only. Each bar is the mean ± SEM of 5–10 different animals. Statistically significant differences (P < 0.05) were calculated only by comparing expression in each genotype (t-test) and are shown by asterisks.

FIG. 8.

Effect of cloprostenol on the expression of the genes known to be involved in reproductive functions in the ovaries of Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 17 dpc. Mice were injected with vehicle or 1 μg of cloprostenol on 17 dpc and killed 24 h later. For each gene, data are expressed relative to the gene's expression in the ovaries of mice of the same genotype injected with vehicle only. Each bar is the mean ± SEM of 5–10 different animals. Statistically significant differences (P < 0.05) were calculated only by comparing expression in each genotype (t-test) and are shown by asterisks.

Physiological Correlates of the Gα_q/11-Dependent Changes in Reproductive Gene Expression That Occur at the End of Gestation

The decrease in expression of steroidogenic enzyme genes and the Lhcgr that occurred at the end of gestation (Table 2 and Fig. 5) implies that the decline in progesterone and estradiol levels that occur during this time period [27, 30] is coordinately regulated in a Gα_q/11-dependent fashion. The dependency of the decline in progesterone on Gα_q/11 has already been documented [4]. Figure 9A shows that estradiol levels in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 18 dpc were similar and that they declined on 22 dpc in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice but remained elevated in Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. Figure 9B similarly shows that the responsiveness of luteal cells to LH/CG was similar in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice on 18 dpc, and that it declined on 22 dpc in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice but remained elevated in Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. Lastly, the data in Figure 9C show that circulating levels of Inhibin A, another ovarian product, follow the same pattern predicted by the changes in the ovarian expression of Inha documented in Table 2 and Figure 5.

FIG. 9.

Physiological correlates of the changes in ovarian reproductive gene expression at the end of gestation in Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. For A and B, each bar shows the mean ± SEM of five to eight different animals analyzed on the indicated days. C) The effects of hCG on cAMP accumulation in freshly isolated luteal cells (see Materials and Methods for details). These data are presented as fold over basal level for ease of comparison. The basal levels of cAMP among the four groups shown varied between 30 and 50 pmol/mg protein. Each bar shows the mean ± SEM of three to four different animals analyzed on the indicated days. Statistically significant differences (P < 0.05) were calculated using ANOVA and are shown by the brackets and asterisks. On 18 dpc Gα_q^f/f;Gα₁₁^−/−;Cre⁻ and Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are pregnant. On 22 dpc the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice are still pregnant but the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice are 1–2 days postpartum.

DISCUSSION

The murine parturition pathway is initiated by the increased expression of the Ark1c18 gene in luteal cells. This gene encodes for 20α-hydroxysteroid dehydrogenase an enzyme that inactivates progesterone by converting it to 20α-hydroxyprogesterone. In pregnant rodents the expression of luteal Akr1c18 is initially silenced by the actions of prolactin and placental lactogens [29, 31–38]. As pregnancy progresses, expression of the luteal PGF2α receptor increases [21–24], and it is engaged by uterine-derived PGF2α [25], which, through the activation of G_q/11 [4], increases the expression of the Ark1c18 gene and the conversion of progesterone to 20α-hydroxyprogesterone [26–29]. The resulting drop in progesterone ultimately allows uterine contractions and parturition to begin (reviewed in Davis and Rueda [25], Ratajczak and Muglia [39], Stocco et al. [40], and Henkes et al. [41]). We cannot assume, however, that all actions of PGF2α are mediated by G_q/11, because the PGF2α receptor also activates G_12/13 [5]. In addition, luteal G_q/11 can also be activated by other hormones, such as oxytocin [6, 7] or lysophosphatidic acid [8, 9], and these could be involved in luteolysis and parturition as well.

The results of the microarray analysis reported here show that Ark1c18 is the ovarian gene that is most robustly up-regulated at the end of gestation in a Gα_q/11-dependent fashion. Because we recently showed that the increased expression of ovarian Akr1c18 provoked by activation of the PGF2α receptor or observed at the end of gestation is dependent on Gα_q/11 [4], the finding of Akr1c18 as the ovarian gene that is most robustly up-regulated in our microarray is a reassuring finding that serves as a positive control. The data presented here and in a previous publication [4] also show that the up-regulation of this gene is greatly reduced in magnitude but not abolished in the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. Nevertheless, this residual up-regulation detected in Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice does not appear to be physiologically significant because these mice experience a delay or failure of parturition [4]. When grouped by the magnitude of the difference in expression, the other up-regulated genes code for proteins involved in a variety of cellular functions, including components of the extracellular matrix (Tnc), control of the cell cycle (Cdkn2b), solute transport (Slc46a3), and nicotinic receptor signaling (Chrna1). When grouped by their presumed involvement in cellular functions, they segregate into some categories, such as focal adhesion, extracellular matrix, endocytosis, and cytokine signaling, that may be involved in the regression of the corpus luteum and other ovarian remodeling that occurs at the end of gestation and the reentry into the estrous cycle. Although the “metabolic pathway” KEGG grouping has the largest number of genes and the highest level of statistical significance we chose not to further consider genes belonging to this KEGG grouping because of its broadness. We instead validated the microarray results of the up-regulated genes with Tnc and Jun, two up-regulated genes placed in the focal adhesion KEGG pathway. Tnc codes for tenascin C (Tn-C), a nonstructural extracellular matrix protein that is highly up-regulated during active tissue remodeling [42]. Tn-C is expressed in several ovarian compartments, including degenerating luteal cells [43–45]. Jun was also placed in this KEGG pathway, but because it is a component of the AP-1 transcription factor [46], it can modulate the expression of many genes and participate in many pathways. Jun is present in the porcine corpus luteum, and under some conditions it can be induced by PGF2α [47].

Serpina6 and Enpp2 are the genes that exhibit the largest Gα_q/11-dependent decrease in expression at the end of gestation. The finding that these two genes are expressed in the ovary and down-regulated at the end of gestation is novel and could have potential implications in reproductive functions. Serpina6 codes for corticosteroid-binding globulin (CBG), a circulating protein that binds cortisol and progesterone [48]. In the rat, the liver is the main source of circulating CBG, and its expression can be stimulated with estrogens [49, 50]. Circulating CBG levels increase during the second and third trimesters in pregnant women and influence progesterone levels in the blood and at the maternal-fetal interface [51, 52]. Our data show that Serpina6 is expressed in the pregnant mouse ovary and that it declines at the end of gestation in a Gα_q/11-dependent fashion. The functions of ovarian Serpina6 and its possible contribution to circulating CBG remain to be determined, however. Enpp2 codes for autotaxin [53–55], an extracellular enzyme that synthesizes lysophosphatidic acid (LPA). Lysophosphatidic acid is an extracellular phospholipid that activates many cellular functions by engaging one or more of the six known LPA receptors [8, 9], but its role in reproduction in general and ovarian physiology in particular are poorly understood [9]. The human ovary was initially reported as a site of high Enpp2 expression [56], but this was not confirmed by the expression data available on public databases [57]. Autotaxin expression was shown in the rat corpus luteum but not in developing or mature follicles [58]. In addition, LPA is present in very high concentrations in human follicular fluid isolated from women undergoing in vitro fertilization, and circulating LPA as well as circulating autotaxin increase during human pregnancy [59, 60]. Lastly, several LPA receptor subtypes are expressed in the ovary [9, 58], and addition of LPA to luteal cells impacts steroid and interleukin biosynthesis [61, 62]. When considered together, these data suggest that ovarian-derived LPA is an autocrine/paracrine regulator of ovarian functions. This latter possibility is currently being investigated.

When grouped by their presumed involvement in cellular functions (Table 4), the genes that are down-regulated at the end of gestation in a Gα_q/11-dependent fashion segregate into several KEGG pathways, including three—“protein processing in endoplasmic reticulum,” “N-glycan biosynthesis,” and “protein export”—that are involved in protein synthesis, processing, and glycosylation. We validated the microarray results of the down-regulated genes with Wfs1 and Ssr4, two down-regulated genes placed in the “protein processing in endoplasmic reticulum” KEGG pathway, the down-regulated pathway with the highest level of statistical significance. Wfs1 codes for an endoplasmic reticulum protein called WFS1 that is involved in calcium homeostasis and the unfolded protein response. In humans, mutations of the WFS1 gene result in a disorder (called Wolfram syndrome) characterized by neurodegeneration and diabetes mellitus [63]. Ssr4 codes for SSR4, one of the four signal sequence receptor proteins that are part of a complex of membrane proteins needed for N-glycosylation [64] and to translocate secreted proteins across the endoplasmic reticulum membrane [65]. There is no available information about the expression or possible functions of these two genes in the ovary, but because of the secretory nature of luteal cells [66], the down-regulation of these ovarian genes could also be involved in the transition that occurs as the mice resume the estrous cycle after delivery.

The group of down-regulated genes in the “steroid hormone biosynthesis” KEGG pathway is also of interest because it contains the three enzymes required to convert progesterone to estradiol (Cyp17a1, Hsd17b7, and Cyp19a1). These three genes, as well as two other ovarian genes known to be involved in reproductive functions (Lhcgr and Inha), are also among the genes with the largest Gα_q/11-dependent decrease in expression at the end of gestation. Thus, we also validated the microarray results with these five additional genes. Two of these (Cyp19a1 and Hsd17b7) were previously shown to decrease at the end of gestation [30, 67], but our finding that they are dependent on Gα_q/11 is new. The changes in expression of these five reproductive genes have clear physiological correlates. The decrease in Lhcgr reduces the responsiveness of luteal cells to hCG, the decrease in Inha results in a reduction in the circulating levels of Inhibin A, and the decrease in the three steroidogenic enzymes is associated with a decrease in circulating estradiol levels. We hypothesize that the Gα_q/11-dependent decrease in expression of Lhcgr, Cyp17a1, Cyp19a1, and Hsd17b7, and the increase in expression of Akr1c18 coordinate the drop in circulating progesterone and estradiol that occurs at the end of gestation [4, 25–29, 39–41]. The decrease in Lhcgr reduces the stimulus for progesterone synthesis, and the increase in Akr1c18 enhances progesterone metabolism. Because the magnitude of progesterone production is several orders of magnitude higher than that of estradiol (ng/ml levels of circulating progesterone and pg/ml levels for circulating estradiol), a decline in progesterone production alone (such as that induced by the changes in Lhcgr and Akr1c18) would have to be very large to translate into a significant drop in estradiol production. Thus, the additional decrease in expression of Cyp17a1, Cyp19a1, and Hsd17b7 is likely to contribute to a reduction in estradiol synthesis even if the substrate (progesterone) was still relatively high.

The decreased expression of Inha and circulating Inhibin A predicts that follicle-stimulating hormone (FSH) levels will rebound at the end of gestation in the Gα_q^f/f;Gα₁₁^−/−;Cre⁻ mice but remain low in the Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice. We attempted to test this hypothesis by measuring FSH in both groups of mice on 22 dpc, but most of the samples collected had undetectable levels of FSH.

The failure of parturition and lack of progesterone withdrawal observed in the Ptgfr null mice shows that activation of the luteal PGF2α receptor is the trigger for parturition in mice (reviewed in Davis and Rueda [25], Ratajczak and Muglia [39], Stocco et al. [40], and Henkes et al. [41]). Because we have recently shown that the phenotype of Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice is very similar to that of the Ptgfr null mice, and because luteal cells derived from Gα_q^f/f;Gα₁₁^−/−;Cre⁺ mice fail to respond to PGF2α with an increase in Akr1c18 expression, we concluded that the effects of the luteal PGF2α receptor on the induction of Akr1c18 are mediated by the activation of G_q/11 [4]. These results, together with the Gα_q/11 dependency of the changes in gene expression that occur at the end of gestation, suggest that they are induced by PGF2α. Indeed, (in addition to Akr1c18) some of the genes identified here have already been shown to be targets of PGF2α. The decreased expression of ovarian Hsd17b7 and Cyp19a1 that occurs at the end of gestation was shown to be blunted in Ptgfr null mice [67], but the data reported here show for the first time that this effect is dependent on the granulosa/luteal Gα_q/11. Also, a previous study has shown that administration of PGF2α to pseudopregnant rats decreases the binding of ¹²⁵I-hCG to the corpus luteum and the ability of hCG to stimulate progesterone synthesis [68]. Our findings that Lhcgr declines at the end of gestation in a Gα_q/11-dependent fashion are also new. When we tested the effects of a PGF2α agonist on the expression of 11 of the Gα_q/11-dependent genes characterized here, we found that only 7 of them (Jun, Serpina6, Ssr4, Enpp2, Lhcgr, Hsd17b7, and Cyp19a1) are regulated by PGF2α, whereas the other 4 (Tnc, Wfs1, Cyp17a1, and Inha) are not. It is possible, however, that some of the PGF2-insensitive mRNAs have a relatively long half-life, and therefore changes in expression may not be detectable within 24 h. In contrast, the microarray and the qPCR experiments done at the end of gestation were done by comparing animals on 18 and 22 dpc (a 96-h difference). Another possibility is that the expression of these genes is dependent on the actions of endocrine or paracrine/autocrine hormones that also activate luteal G_q/11 (such as oxytocin, see Blanks and Thornton [6] and Mitchell et al. [7]) or LPA [8, 9]. When engaged by oxytocin, the oxytocin receptor is known to activate G_q/11 [6, 7], and oxytocin has been shown to induce progesterone withdrawal and parturition in Ptgr null mice [69]. The actions of LPA in the corpus luteum need to be characterized, but some of the LPA receptors are known to activate G_q/11 [8, 9]. The changes in the expression of Gα_q/11-dependent but PGF2α-independent genes may also not be a direct effect of the activation of G_q/11. It may be secondary to the decrease in estradiol levels that occurs as a consequence of G_q/11 activation. Lastly, the dependency of luteal Cyp17a1 expression on Gα_q/11 is also curious because Cyp17a1 is expressed in theca cells, not granulosa cells [70], and the deletion of Gα_q/11 took place initially in granulosa cells [3, 4]. We therefore hypothesize that the regulation of luteal Cyp17a1 expression is due to Gα_q/11-dependent changes in a granulosa/luteal cell-specific product that acts on theca/luteal cells to inhibit Cyp17a1 expression.

In summary, the results presented here have identified 138 and 150 ovarian genes that are up-regulated or down-regulated, respectively, at the end of gestation in a Gα_q/11-dependent fashion. Genes that participate in the structure/function of the extracellular matrix are among those that are up-regulated, whereas some of the down-regulated genes participate in the processing and transport of proteins in the endoplasmic reticulum. Several ovarian genes known to be involved in reproductive functions and two novel ovarian genes that may participate in reproductive functions were also down-regulated in a Gα_q/11-dependent fashion. These changes in ovarian gene expression are likely to be involved in the transition that occurs at the end of gestation and in the reentry into the estrous cycle. PGF2α, acting thought the luteal prostaglandin F2α receptor, a G_q/11-coupled receptor, is a prominent regulator of luteolysis, but not all of the genes that are regulated at the end of gestation in a Gα_q/11-dependent fashion are targets of PGF2α