Intraovarian Activins Are Required for Female Fertility

0888-8809/07/$15.00/0 Printed in U.S.A.

Molecular Endocrinology 21(10):2458–2471 Copyright © 2007 by The Endocrine Society doi: 10.1210/me.2007-0146

Intraovarian Activins Are Required for Female Fertility Stephanie A. Pangas,* Carolina J. Jorgez,* Mai Tran, Julio Agno, Xiaohui Li, Chester W. Brown, T. Rajendra Kumar, and Martin M. Matzuk Departments of Pathology (S.A.P., M.T., J.A., X.L., M.M.M.), Molecular and Human Genetics (C.W.B.), Molecular and Cellular Biology (M.M.M.), and Program in Developmental Biology (C.J.J., M.M.M.), Baylor College of Medicine, Houston, Texas 77030; and Department of Molecular and Integrative Physiology (T.R.K.), The University of Kansas Medical Center, Kansas 66160 Activins have diverse roles in multiple physiological processes including reproduction. Mutations and loss of heterozygosity at the human activin receptor ACVR1B and ACVR2 loci are observed in pituitary, pancreatic, and colorectal cancers. Functional studies support intraovarian roles for activins, although clarifying the in vivo roles has remained elusive due to the perinatal death of activin ␤A knockout mice. To study the roles of activins in ovarian growth, differentiation, and cancer, a tissue-specific knockout system was designed to ablate ovarian production of activins. Mice lacking ovarian activin ␤A were intercrossed to Inhbb homozygous null mice to produce double activin knockouts. Whereas ovarian ␤A knockout females

are subfertile, ␤B/␤A double mutant females are infertile. Strikingly, the activin ␤A and ␤B/␤A-deficient ovaries contain increased numbers of functional corpora lutea but do not develop ovarian tumors. Microarray analysis of isolated granulosa cells identifies significant changes in expression for a number of genes with known reproductive roles, including Kitl, Taf4b, and Ghr, as well as loss of expression of the proto-oncogene, Myc. Thus, in contrast to the known tumor suppressor role of activins in some tissues, our data indicate that activin ␤A and ␤B function redundantly in a growth stimulatory pathway in the mammalian ovary (Molecular Endocrinology 21: 2458–2471, 2007)

CTIVINS, HOMODIMERS OR heterodimers of ␤A or ␤B-subunits, were identified for their roles in positively regulating pituitary FSH synthesis and secretion (1–3). The ␤-subunits also dimerize with the inhibin ␣-subunit to produce inhibins [inhibin A (␣:␤A) and inhibin B (␣:␤B)] that negatively regulate FSH. The ␤A- and ␤B-subunits, encoded by the Inhba and Inhbb genes, respectively, are produced in multiple tissues during embryonic and postnatal development and have diverse physiological effects. Mice lacking activin ␤A (homozygous Inhba null) die at birth secondary to craniofacial defects, whereas homozygous Inhbb null mice are viable but exhibit eyelid closure and nursing defects (4–6). Several studies indicate that activin signaling components, including the receptors ACVR1B (also known as ALK4) and ACVR2 (also known as ACTR2B), are critical for growth inhibition. For example, activins inhibit in vitro

growth of both human breast cancer cells (7) and prostate cancer cells (8). Mutations in ACVR1B have been observed in human pituitary tumors (9) and pancreatic cancers (10). Likewise, ACVR2 mutations and loss of heterozygosity have been seen in pancreatic (11), colorectal (12), and prostate cancers (13). These studies suggest that the activin signaling pathway functions as a tumor suppressor pathway in some tissues to block cell growth and stimulate differentiation. Inhba and Inhbb transcripts are prominently expressed by granulosa cells of large preantral and antral follicles (14). In vitro studies have identified multiple intraovarian roles for activins in granulosa cells including proliferation, potentiation of FSH action, and modulation of steroidogenesis (reviewed in Refs. 15 and 16). To study the in vivo roles of activins, we produced an ovarian Inhba conditional knockout (␤A cKO). We also generated the ␤A cKO in the Inhbb null background mice to produce mice lacking all ovarian activins (double mutant mice; herein designated as ␤B/␤A dKO). In contrast to extragonadal roles of these proteins as tumor suppressors, activins play redundant roles to block terminal differentiation of granulosa cells, leading to dosage-dependent fertility defects.

A

First Published Online July 3, 2007 * S.A.P. and C.J.J. contributed equally to this work. Abbreviations: BMP, Bone morphogenetic protein; cKO, conditional knockout; CL(s), corpus luteum/corpora lutea; Ctgf, connective tissue growth factor; CV, coefficient of variation; dKO, double mutant mice; Gapd, glyceraldehyde 3-phosphate dehydrogenase; Ghr, GH receptor; hCG, human chorionic gonadotropin; 20␣-HSD, 20␣-hydroxysteroid dehydrogenase; HSD, honestly significant difference; Lhcgr, LH receptor; Mmp2, matrix metalloproteinase 2; PMSG, pregnant mare serum gonadotropin; PRL, prolactin; qPCR, quantitative PCR. Molecular Endocrinology is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the endocrine community.

RESULTS Generation of Ovarian Activin ␤A Knockout Mice A floxed Inhba allele was generated (Inhbatm3Zuk; herein called ␤Aflox) for subsequent tissue-specific de2458

Pangas et al. • In Vivo Roles of Activins in the Ovary

letion of the Inhba gene in vivo (Fig. 1A). Exon 2 was floxed because it encodes the entire mature domain of the protein and a conditional exon 2 deletion would mimic the original null allele (Inhbatm1Zuk) (17). ␤Aflox/⫹ mice were intercrossed to produce ␤Aflox/flox homozygous mice, which were viable and fertile and obtained at the expected Mendelian frequency. To verify the presence of loxP sites and their ability to recombine in vivo, we crossed the ␤Aflox mice to EIIa-cre transgenic mice, which express cre recombinase in multiple tissues, including germ cells (18). Southern blot analysis using a 5⬘ probe demonstrated the presence of the various Inhba alleles, including the recombined deleted (␤A⌬) (data not shown). Because ␤Aflox/⫺;EIIa-cre newborn mice demonstrate the neonatal lethality seen in the Inhba homozygous null mice (data not shown), recombination likely produced a null allele. Anti-Mu¨llerian hormone receptor-cre (Amhr2cre/⫹) knock-in mice were used to generate recombination of ␤Aflox in ovarian granulosa cells (19–21). ␤A⫹/⫺; Amhr2cre/⫹ mice were mated to ␤Aflox/flox mice and all four genotypes were recovered at the expected Mendelian frequency. Recombination efficiency of the ␤Aflox locus was determined by Southern blot analysis of granulosa cells derived from cre-negative ␤Aflox/⫺ (control) mice and ␤Aflox/⫺;Amhr2cre/⫹ (experimental) mice (Fig. 1B). Recombination of the floxed allele in granulosa cells from female littermates varied from 60% to 99% (Fig. 1B). Variation in recombination and expression differences are likely the result of mosacism in recombination, as has been seen in other conditional knockout mice using Amhr2-cre mice (21, 22). Loss of the ␤A transcript was confirmed by Northern blot analysis (Fig. 1C), and no compensatory upregulation of the activin ␤B-subunit was found in ␤Aflox/⫺;Amhr2cre/⫹ ovaries (Fig. 1D). Because the ␤-subunit is also required for formation of inhibin, we measured serum inhibin levels in the ␤Aflox/⫺; Amhr2cre/⫹mice. The average serum inhibin A level in the ␤A-deficient mice is significantly reduced from the control values, and six of nine females had undetectable levels. Loss of dimeric inhibin A was not due to reductions in inhibin ␣-subunit mRNA expression because ovaries and granulosa cells of mutant mice demonstrated increased Inha mRNA levels over controls (Fig. 1E, inset, lane 4). Redundant Roles of Activin ␤A and ␤B in Female Fertility The fertility of control (␤Aflox/⫺) mice was comparable with wild-type mice (data not shown). However, ␤Aflox/⫺; Amhr2cre/⫹ female mice demonstrated a statistically significant reduction (43%) in litters per month and pups per litter (33%) compared with controls (Table 1 and data not shown). Because subfertility could result from redundancy between the activin A and B isoforms (23), we generated ␤A conditional knockouts in the activin ␤B null (␤B⫺/⫺) background and additionally analyzed the following genotypes: ␤B⫺/⫺;␤A⫹/⫺,

Mol Endocrinol, October 2007, 21(10):2458–2471 2459

␤B⫺/⫺; ␤Aflox/⫺, and ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ females. Neither ␤A⫹/⫺ nor ␤B⫹/⫺ heterozygous mice had fertility defects (5, 17). Double mutant females with only one remaining ␤A-subunit allele (␤B⫺/⫺; ␤A⫹/⫺ or ␤B⫺/⫺;␤Aflox/⫺) had severe fertility defects (Table 1), indicating that activin dosage plays an important role in ovarian function. Of the 16 mice that were studied over 6 months, three of five ␤B⫺/⫺; ␤Aflox/⫺, and three of 11 ␤B⫺/⫺;␤A⫹/⫺ female mice were infertile. In the mutant lines containing a single ␤-subunit allele, the average number of litters per month was significantly reduced (89–91%), and the number of pups per litter was 87–90% lower than the controls (Table 1). Finally, female mice lacking all intraovarian activin ␤A and ␤B alleles (␤B⫺/⫺;␤Aflox/⫺; Amhr2cre/⫹) were infertile (Table 1). Thus, although activin A appears to be functionally dominant [because activin A deficiency results in subfertility vs. normal fertility for activin ␤B deficiency (5)], dosage of the activin ␤A and ␤B alleles is important for normal fertility. To assess whether ovulation defects were present in the activin-deficient females, 3-wk-old females were superovulated, mated and oocytes collected from oviducts. Four groups of mice were analyzed: wild-type, ␤Aflox/⫺, ␤Aflox/⫺;Amhr2cre/⫹, ␤B⫺/⫺;␤Aflox/⫺, and ␤B⫺/⫺;␤Aflox/;Amhr2cre/⫹. There were no significant differences in the number of oocytes released after ovulation, except in mice deficient for all ␤-subunits (Table 2). Two of four female mice deficient for all activin alleles produced no oocytes. Enhanced Corpus Luteum (CL) Formation in Activin-Deficient Ovaries To determine the causes of subfertility/infertility in the activin-deficient mice, ovaries were analyzed at 3 and 8 months of age. At 3 months of age, there were no gross differences in appearance or size of the ovaries between ␤Aflox/⫺;Amhr2cre/⫹ and control mice (␤Aflox/⫺) (data not shown). The only visible histological difference at 3 months of age was the presence of ovarian follicles containing more than one oocyte [50% occurrence in the experimental mice vs. none of the controls (␤Aflox/⫺) (Fig. 2, A and B)]. However, at 8 months of age, ovaries from the ␤Aflox/⫺;Amhr2cre/⫹ mice contained a greater abundance of CLs (Fig. 2D) as compared with control ␤Aflox/⫺ mice (Fig. 2C). Female mice with only one remaining activin ␤ allele (␤B⫺/⫺;␤A⫹/⫺) showed an earlier increase in the number of CLs (Fig. 2E) (i.e. the phenotype appears at 3 months of age instead of at 8 months of age for the single ␤A conditional knockout). By 8 months of age, the ␤B⫺/⫺;␤A⫹/⫺ ovaries contained a readily identifiable abundance of CLs (Fig. 2F). A more extreme version of the phenotype was visible in mice deficient for all activins (i.e. in ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ females) (Fig. 3). Abundant CLs were visible histologically at 6 wk (Fig. 3C) and 3 months of age (Fig. 3D). Also at 8–11 wk of age, there

2460 Mol Endocrinol, October 2007, 21(10):2458–2471


Fig. 1. Creation of the Activin ␤A Conditional Gene and Efficiency of in Vivo Recombination A, The activin ␤A conditional-targeting vector to delete exon 2 is shown. The targeting vector was generated by inserting a loxP sites into intron 1, and a PgkNeo cassette flanked by loxP sites after the 3⬘untranslated region of the activin ␤A gene. Four alleles shown are: WT, ␤A⫺, ␤Aflox, and ␤A⌬ (deleted). B, Efficiency of Amhr2-cre in recombining ␤Aflox allele in granulosa cells. Southern blot analysis of granulosa cell DNA derived from four offspring using the 5⬘ probe, which detects a 4.2-kb ␤A null band, a 6.1-kb ␤Aflox band, and a 13.1-kb ␤A⌬ (deleted) band. In lane 1, genomic DNA from a control (␤Aflox/⫺) mouse. In lanes 2–4, ␤Aflox/⫺;Amhr2cre/⫹ mice show recombination. Mice in lane 2 and 3 are littermates. C, Loss of activin ␤A mRNA expression in the ovary. Northern blot analysis of whole ovary RNA from 3-month old mice. Lanes 2, 3, and 5, control (␤Aflox/⫺) mice show expression of the activin ␤A-subunit. Lanes 1 and 4, ␤Aflox/⫺;Amhr2cre/⫹ mice do not show the activin ␤A-transcript. Gapd was used as a control for RNA loading. D, qPCR for activin ␤B-subunit expression in adult control ␤Aflox/⫺ (n ⫽ 3) and experimental ␤Aflox/⫺;Amhr2cre/⫹(n ⫽ 3) ovaries. Although the mean relative expression in cre-positive ovaries decreased, the difference is not statistically significant (ns, not significant). E, Serum inhibin A levels in adult control ␤Aflox/⫺ (n ⫽ 8) and experimental ␤Aflox/⫺;Amhr2cre/⫹ (n ⫽ 9) females show a statistically significant decrease by ELISA for inhibin A (P ⬍ 0.05). Inhibin A levels in mice lacking all ␤-subunits in the ovary (␤B⫹/⫺;␤Aflox/⫺;Amhr2cre/⫹, third column) (n ⫽ 4) are also significantly different from the control. Different letters above the columns represent significantly different means by one-way ANOVA and Tukey-Kramer HSD post hoc test. The dashed line represents the limit of detection of the ELISA. Although dimeric serum inhibin declines, the inhibin ␣-subunit mRNA expression is intact in ovaries of 3-month-old activin-deficient mice (inset, panel D) by Northern blot analysis. Lane 1, wild type (WT); lanes 2–3 ␤A⫹/⫺;␤B⫺/⫺; lane 4, ␤Aflox/⫺;Amhr2cre/⫹; Gapd was used as control. BHI, BamHI; ERV, EcoRV; PI, PstI.



Table 1. Six-Month Breeding Data for Activin-Deficient Female Mice Genotype

␤〈 Alleles

␤〉 Alleles

Total Alleles

n

Total Litters

Average Litter Size

Litters/Month

␤Aflox/⫺ ␤Aflox/⫺; cre⫹ ␤B⫺/⫺; ␤A⫹/⫺ ␤B⫺/⫺; ␤Aflox/⫺ ␤B⫺/⫺; ␤Aflox/⫺;cre⫹

1 0 1 1 0

2 2 0 0 0

3 2 1 1 0

10 10 10 5 9

50 36 5 2 0

7.78 ⫾ 0.48a 5.19 ⫾ 0.65b 0.80 ⫾ 0.49c 1.00 ⫾ 0.63b,c 0c

0.89 ⫾ 0.04a 0.51 ⫾ 0.05b 0.08 ⫾ 0.08c 0.10 ⫾ 0.07c 0c

Values are average ⫾ SEM. Statistical significance by one-way ANOVA followed by Tukey-Kramer HSD test (P ⬍ 0.05). Values with different superscripts (a–c) indicate statistically different means. n, Number of animals.

were statistically significant increases in the number of antral follicles (P ⬍ 0.05) (Fig. 4A) and in the number of CLs (P ⬍ 0.01) (Fig. 4B) only in ␤B⫺/⫺; ␤Aflox/⫺; Amhr2cre/⫹ females, vs. all tested genotypes. No changes were found in the numbers of developing primordial, primary, or preantral follicles for any genotype at 6–11 wk of age (data not shown). At 8 months of age, ovaries of ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ mice had a 6-fold increase in mass [31.5 ⫾ 2.9 mg (n ⫽ 8) vs. 5.2 ⫾ 2.9 mg (n ⫽ 8) in the ␤Aflox/⫺ cKO mice] (Fig. 3A). Also present were follicles with luteinizing granulosa cells (Fig. 3H), and multiple cysts that were either fluid-filled or hemorrhagic (Fig. 3, F and I). Cells lining cysts are immunopositive for the granulosa cell marker, inhibin ␣ (Fig. 3, J and K), and negative for epithelial markers (Fig. 3L and data not shown), and thus likely derive from ovarian follicles. The serum levels of estradiol, prolactin (PRL), progesterone, and the gonadotropins, FSH and LH, were measured to examine ovarian function at 3 months of age (Fig. 5). There were no significant differences in basal serum levels of LH (Fig. 5B), estradiol (Fig. 5D), or PRL (data not shown) in random cycling female mice for any genotype. However, a statistically significant increase (P ⬍ 0.01) was identified for FSH in mice lacking all activin subunits (Fig. 5A). In addition, there was a trend of increasing serum progesterone corresponding with the decreasing number of ␤-subunit alleles, which reached statistical significance (P ⬍ 0.001) at 3 months of age for ␤B⫺/⫺; ␤Aflox/⫺; Amhr2cre/⫹ females (Fig. 5C). In mouse ovaries, CLs can be classified into two stages, healthy and regressing; regressing CLs can be further subclassified into functional and structural regression stages. One of the major hormones support-

ing the growth of CLs and production of progesterone is LH. The LH receptor (Lhcgr) is expressed highly in mural granulosa cells of preovulatory follicles and healthy CLs (14, 24, 25). By in situ hybridization for Lhcgr, several healthy CLs in mice lacking one to four ␤-subunits were detected (Fig. 6), and expression of LH receptor was higher in those mice with increased numbers of CLs. To examine whether CLs in experimental mice could undergo functional regression, we examined expression of 20␣-hydroxysteroid dehydrogenase (20␣-HSD), which catalyzes the conversion of progesterone into 20␣-dihydroxyprogesterone, a biologically inactive form. Immunostaining for 20␣-HSD in the experimental and control mice identified functional regressing corpora even in mice with one or no remaining ␤-alleles wild type (Fig. 6, F and I). These results suggest that part of the phenotype of the experimental mice may be that the rate of CL formation in activin-deficient mice might be greater than the rate of regression. Gene Expression Differences in Granulosa Cells from Activin-Deficient Mice Several genes have previously been shown to be regulated by activin in granulosa cell culture experiments. We initially used quantitative PCR (qPCR) to examine granulosa cell gene expression in wild-type mice, ␤Aflox/⫺;␤B⫺/⫺ mice, and ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ mice using a candidate gene approach. FSH receptor (Fshr) and cyclin D2 (Ccnd2) are reported to be upregulated by activin (26–29). Sequential loss of the activin ␤-subunits reduced Fshr mRNA expression (Fig. 7A), but this trend was not statistically significant when comparisons were made between all three ge-

Table 2. Superovulation of 21-Day-Old Wild-Type and Activin-Deficient Mice Genotype

␤〈 Alleles

␤〉 Alleles

Total Alleles

Total No. of Animals

Average Oocytes

Wild type ␤Aflox/⫺ ␤Aflox/⫺;cre⫹ ␤B⫺/⫺; ␤A⫹/⫺ ␤B⫺/⫺; ␤Aflox/⫺;cre⫹

2 1 0 1 0

2 2 2 0 0

4 3 2 1 0

8 5 12 12 4

39.1 ⫾ 5.5 31.6 ⫾ 5.8 36.3 ⫾ 4.4 35.3 ⫾ 6.8 15.8 ⫾ 10.5a

Values are average ⫾ SEM. Statistical significance by one-way ANOVA followed by Tukey-Kramer HSD test (P ⬍ 0.05). a By ANOVA, none of the groups are statistically significant for mean oocytes. However, 2 of 4 ␤B⫺/⫺;␤Aflox/⫺;cre⫹ females yielded no ovulated oocytes, and a t test between wild type and ␤B⫺/⫺;␤Aflox/⫺;cre⫹ is statistically significant (P ⫽ 0.43).


Fig. 2. Histological Analysis of the Ovaries of Activin ␤A Conditional Mice and ␤A⫹/⫺;␤B⫺/⫺ Mutant Mice A, Ovary from a 3-month-old ␤Aflox/⫺;Amhr2cre/⫹ female with several CLs, follicles at different stages, and one follicle containing two oocytes (arrow). B, Higher magnification of polyovular follicle from panel A. C, Ovary from an 8-monthold ␤Aflox/⫺ control mouse with follicles at different stages and several CLs. D, Ovary from an 8-month-old ␤Aflox/⫺; Amhr2cre/⫹ mouse with a large increase in the number of CLs. E, Ovary from 3-month-old ␤B⫺/⫺;␤A⫹/⫺ mouse demonstrating increased numbers of CLs and decreased numbers of visible developing follicles. F, Ovary from an 8-month-old ␤B⫺/⫺;␤A⫹/⫺ female mouse with a large number of CLs and almost no remaining developing follicles. Scale bar, 400 ␮m. AnF, Antral follicle; Oo, oocyte.

notypes (P ⫽ 0.08). These data suggest that in vivo, activin likely augments FSH-induced expression of Fshr, similar to its effects in vitro (28, 29), but is not necessary for basal FSH receptor expression. Unexpectedly, reduction of activin alleles had no effect on Ccnd2 expression (Fig. 7B). We also examined expression of Lhcgr and found that its expression was significantly up-regulated (P ⬍ 0.05) in granulosa cells of mutant mice (Fig. 7C). These gene expression results are similar to the previously reported gene expression pattern of Smad4 granulosa cell-specific knockout mice, which also show an increase in Lhcgr, but little effect on Ccnd2 (21). Misexpression of Lhcgr in activin-deficient granulosa cells is consistent with an inappropriate differentiation of granulosa cells (30). To further identify changes in gene expression in granulosa cells, we performed microarray analysis on wild-type and activin-deficient granulosa cells. Granulosa cells were collected from immature wild-type


mice and mice that were deficient for all ␤-subunits (␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹), and analyzed by using the Affymetrix (Santa Clara, CA) Mouse Genome 430 2.0 Arrays, which represents over 39,000 mouse transcripts. Differentially expressed genes are presented in Supplemental Tables 1 and 2, which are published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals. org. Because granulosa cells were collected from pregnant mare serum gonadotropin (PMSG)-stimulated females, differentially expressed genes represent genes expression changes induced in growing antral follicles by gonadotropin stimulation with or without endogenous activin. One hundred and eleven nonredundant genes showed statistically significant (P ⬍ 0.05) differences in the final analysis, whereas 14 genes were expected to change by chance alone. Eighty-one genes of these genes were down-regulated (Supplemental Table 1), whereas 30 genes were up-regulated (Supplemental Table 2). Pathway analysis demonstrated that three signaling pathways were significantly disrupted: the TGF␤ pathway (P ⫽ 0.005), as would be predicted by loss of activin alleles, the JAK-STAT pathway (P ⫽ 0.023), and the cytokinecytokine receptor interaction pathway (P ⫽ 0.016). Genes differentially expressed in these pathways are highlighted in Supplemental Tables 1 and 2. Expression of several differentially expressed genes with known ovarian roles were verified by qPCR on granulosa cells from individual mice collected independently from the samples used in the microarray analysis (Fig. 7, D–H, and Supplemental Table 1). In general, the fold changes found for qPCR verification in these experiments were larger than fold changes determined from microarray data. Likely, this results from the use in the microarray analysis of pooled RNA from samples that have variable recombination rates (Fig. 1B), leading to more conservative estimates of gene expression changes. We found that connective tissue growth factor (Ctgf) was significantly down-regulated in the microarray (4.7-fold) and by qPCR (1 ␤-allele, 16-fold; 0 ␤-alleles, 26-fold) in mutant mice (Fig. 7D). Importantly, Ctgf has been shown to be regulated by several members of the TGF␤ family, including activin (31). Other significantly down-regulated genes in mutant mice include GH receptor (Ghr) (1 ␤-allele and 0 ␤-alleles, 14-fold) (Fig. 7E), myelocytomatosis oncogene (Myc) (1 ␤-allele, 13-fold; 0 ␤-alleles, 15-fold) (Fig. 7F), and the TATA-binding factorassociated factor (Taf4b) (1 ␤-allele, 3.9-fold; 0 ␤-alleles, 4.4-fold) (Fig. 7G). It is unknown whether these genes are activin target genes. To determine whether Ghr, Myc, or Taf4b are regulated by activin, we treated wild-type granulosa cells with recombinant human activin A and measured changes in mRNA expression by qPCR (Fig. 8). Levels of Ghr and Myc were unchanged after 5 h, indicating that short-term activin treatment does not alter their transcription (Fig. 8, A and B). However, a statistically significant (P ⬍



Fig. 3. Histological Analysis of the Ovaries of Activin Double Knockout Mice Ovaries from mice lacking activin ␤A and ␤B have a dramatic increase in overall size and CL number. A, Gross morphological comparison of 8-month-old ovaries from an ␤Aflox/⫺;Amhr2cre/⫹ (␤A cKO) mouse (left) showing normal appearance. In contrast, ovaries from ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ mouse (right) are approximately six times larger and show hemorrhagic cysts (arrows) and increased numbers of CLs (white areas). B, Ovary from 21-d-old ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹mouse with follicles of all stages, but no CL development. C, Ovary from 6-wk-old from ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹mouse with several preovulatory follicles (arrowheads) and many CLs (one CL indicated). D, Ovary from 3-month-old ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ dKO mouse with an elevated number of CLs in two different magenta tones; the darker staining CLs are younger and the lighter staining ones are older. E, Ovary from 8-month-old ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ dKO mouse with a large number of CLs, and few antral follicles. F, Ovary from an 8-month-old ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ dKO mouse with a large number of CLs, few antral follicles, and a large number of cysts. G, Higher magnification of ovary from panel D showing two CLs, one of them with pale, highly steroidogenic cells (asterisk). H, Higher magnification of ovary from panel E showing a degenerating oocyte (Oo) surrounded by highly steroidogenic granulosa cells. I, higher magnification of ovary from panel F showing multiple cysts. J, Immunohistochemistry for inhibin ␣ in an ovary from an 8-month old ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ dKO. Cells lining the cyst are inhibin ␣ positive, whereas CLs are negative. Cyst in panel J is shown at a higher magnification in panel K. L, Cells lining the cyst are negative for CK19, an epithelial marker. Inset, Positive control for CK19 immunostaining in the ovarian surface epithelium (ose). Scale bar, 400 ␮m in all panels, except in panel J, where scale bar is 200 ␮m.

0.05), increase was seen for Taf4b (1.8-fold) (Fig. 8B) after 5 h. Fewer genes were significantly up-regulated than down-regulated in the microarray analysis, but several have important ovarian roles, including kit ligand (Kitl). Of the up-regulated genes, Kitl increased in mutant

granulosa cells as assayed by microarray (2-fold) (Supplemental Table 2) and qPCR (1 ␤-allele, 3-fold; 0 ␤-alleles, 5-fold) (Fig. 7H). Kit ligand is a positive regulator of follicle growth (32) and has important growth stimulatory effects on granulosa cells (33). The expression of kit ligand has been shown to be increased by

2464

Mol Endocrinol, October 2007, 21(10):2458–2471

Fig. 4. Antral Follicle and CL Counts in Activin Mutant Mice at 6–11 Wk of Age Ovaries were serially sectioned and the total number of antral follicles and CLs were counted for the following genotypes: ␤B⫹/⫺;␤A⫹/⫺;Amhr2cre/⫹ (2 ␤ alleles), ␤B⫹/⫺;␤Aflox/⫺; Amhr2cre/⫹ (1 ␤ allele); ␤B⫺/⫺;␤Aflox/⫺ (1 ␤ allele); and ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹ (0 ␤ alleles). A, There are statistically significant increases in the number of antral follicles (fourth bar) (P ⬍ 0.05) and B, CL (fourth bar) (P ⬍ 0.01), in mice that are deficient in all ovarian activins. Each group is represented by three mice, and statistics were preformed by one-way ANOVA followed by Dunnett’s post hoc test using the 2 ␤ allele group as the control group.

bone morphogenetic protein (BMP) 15 (33) and inhibited by GDF9 (34). To determine whether activin has any regulatory role on Kitl expression, Kitl transcript levels were measured in activin-treated wild-type granulosa cells by qPCR. Recombinant activin A significantly suppressed levels of Kitl in wild-type granulosa cells by approximately 7-fold (P ⬍ 0.05) (Fig. 8D). These data are consistent with the increase in Kitl expression detected in activin-deficient granulosa cells, and thus Kitl is a candidate gene for negative regulation by activin in granulosa cells.

DISCUSSION Activins have multiple roles during folliculogenesis, including regulation of granulosa cell growth and differentiation. Our ovarian activin knockout mice demonstrate decreasing fertility with sequential deletion of the ␤-subunits, culminating in sterility when all intraovarian activins are removed. As our laboratory previously observed (23), activin dosage plays an important role in ovarian function


Fig. 5. Serum Hormone Levels from Random Cycling 3-Month-Old Activin Mutant Mice Bars represent number of ␤ alleles (3, 2, 1, 0) present. Genotypes are as follows: group (3) is ␤Aflox/⫺; group (2) is ␤Aflox/⫺;Amhr2cre/⫹; group (1) is ␤B⫺/⫺;␤A⫹/⫺; group (0) is ␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹. A, FSH values significantly increase (P ⬍ 0.01) with decreasing numbers of ␤ alleles. Sample sizes for FSH values are (3) n ⫽13; (2) n ⫽ 11; (1) n ⫽ 13; (0) n ⫽ 6. B, LH values do not change across genotypes. Sample sizes for LH values are (3) n ⫽ 12; (2) n ⫽ 12; (1) n ⫽ 10; (0) n ⫽ 3. C, Progesterone levels are statistically significant (P ⬍ 0.001) for mice with no activin alleles (0) (hatched bar) at 3 months, compared with the other mutant mice. Sample sizes for progesterone are: (3) n ⫽ 7; (2) n ⫽ 10; (1) n ⫽ 11; (0) n ⫽ 5. D, Estradiol values do not change between genotypes. Sample sizes for estradiol are: (3) n ⫽ 11; (2) n ⫽ 11; (1) n ⫽ 11; (0) n ⫽ 4. Mean and SEM are shown. Statistical analysis by one-way ANOVA, followed by Tukey-Kramer HSD post hoc tests. Statistical difference is shown by different letters above the bars (i.e. a is statistically different from b, but not a,b).

depending on the type of activin allele. A possible explanation for the more severe effects of activin ␤A deficiency, is that activin ␤A has a greater bioactivity than activin ␤B, a hypothesis supported by our previous work (23). In addition, it has been shown in a cell-free assay system that activin A has a higher affinity for both types of activin type 2 receptors compared with activin B (⬃8fold difference) (35). A subfertility defect was uncovered when we generated a conditional knockout for the activin ␤A-subunit. In addition, activin ␤A-deficient females developed follicles with multiple oocytes (polyovular follicles; also called MOFs). The natural occurrence of polyovular follicles in mice have been reported since 1920 (36), and some mouse strains are more susceptible to polyovular follicle formation than others (e.g. C58/J vs. C57L/J) (37). Polyovular follicles are found in ovaries of some transgenic mice, including inhibin-␣ overexpressing mice (38), germ cell nuclear factor conditional knockout mice (39), Bmp15 knockout mice (40) and FSH receptor haploin-



Fig. 6. Markers of Healthy and Regressing CLs in Activin-Deficient Mice Ovaries from mature female mice of different genotypes. A–C, ␤Aflox/⫺ mice; D–F, ␤B⫺/⫺;␤A⫹/⫺ mice; G–I, ␤B⫺/⫺; ␤Aflox/⫺; Amhr2cre/⫹ mice. A, B, D, E, G, and H, in situ hybridization of LH receptor showing high number of healthy CLs in all genotypes. A, D, and G, bright-field; B, E, and H, dark-field; C, F, and I, immunohistochemistry for 20␣-HSD showing high number of functional regressive CLs in mice that lack only one activin ␤A-subunit. Box in panel D is shown at a higher magnification as panel E. Box in panel G is shown at a higher magnification as panel H.

sufficient mice (41). Given the frequency of transgenic lines displaying this phenotype, it is likely that polyovular follicle formation occurs through multiple mechanisms; however, estrogen is one known mediator of polyovular follicle formation (42), likely caused by improper germ cell cyst breakdown (43–46). An additional study has shown that neonatal estrogen exposure results in an increased incidence of polyovular follicles in conjunction with suppression of the activin ␤-subunit (47). These data are consistent with polyovular formation in the ␤A conditional knockout, because Amhr2-driven expression of cre recombinase is detectable in somatic cells of the developing embryonic gonad (19, 20), and could result in embryonic loss of the ␤A-subunit in these cells. Our data further suggest a direct role of activin in normal follicle formation. Mice lacking both activin ␤A and ␤B are sterile, with a complex ovarian phenotype, perhaps in part because activin has stage-specific functions during follicle development that also vary with age. In general, in vitro cultures demonstrate that activin induces small follicle growth and promotes granulosa cell proliferation alone and in combination with FSH (48–50). However, activins promote preantral follicle growth in isolated follicles derived from immature mice but prevent FSH-stimulated growth in isolated cultures of preantral follicles from adult

mice (51). If growth induction is activin’s primary role at the preantral follicle stage, it would be predicted that loss of activin in granulosa cells could result in arrested granulosa cells growth, early differentiation of granulosa cells, or retarded follicle development. Indeed, follicles demonstrating luteinization of granulosa cells were apparent in the ovaries of activin-deficient mice. However, our data did not demonstrate any changes in numbers of primordial, primary, or secondary follicles in adult female mice that lacked all ␤-subunits. Because activin is secreted and its actions are paracrine in nature, variations in timing of recombination between loxP sites in granulosa cells from different follicles may variably change local activin concentrations, which may be insufficient to cause complete intraovarian penetrance of the phenotype. Conditional mutations in activin receptors would thus contribute valuable information regarding activin function in granulosa cells. Deletion of the activin ␤-subunit genes in granulosa cells affects both activin and inhibin production. The gonad is the predominant source of circulating inhibin in rodents (52), and our data demonstrate that deletion of the ␤-subunit genes in granulosa cells decreases serum inhibin to undetectable levels. Although previous in vitro studies show that recombinant activin increases inhibin ␣-subunit expression in isolated



Fig. 8. Gene Expression Changes in Activin-Treated WildType Granulosa Cells Wild-type granulosa cells were untreated (control) or treated with 50 ng/ml recombinant human activin A for 5 h, and RNA levels measured by qPCR. Data from four independent experiments performed on four separate days are shown. Ghr expression (A) or Myc expression (B) is unchanged in activin-treated compared with control-treated granulosa cells. Taf4b significantly increases (C), and Kitl expression significantly decreases (D), upon activin treatment. Data are mean ⫾ SEM. Statistical significance by Student’s t tests (*, P ⬍ 0.05).

Fig. 7. qPCR Analysis of Activin-Deficient Granulosa Cells Granulosa cell RNA was collected from three wild-type (wt) mice, 3 ␤Aflox/⫺;␤B⫺/⫺ mice (1 ␤ alleles) and 3 ␤B⫺/⫺; ␤Aflox/⫺; Amhr2cre/⫹ (0 b alleles) mutant mice. qPCR results are expressed as relative quantity (RQ). A, Fshr trends toward lower expression in the mutants, but is not statistically significant by one-way ANOVA. B, Ccnd2 expression is unchanged between genotypes. C, Lhcgr is significantly up-regulated in the activin-deficient granulosa cells (0) vs. the wild-type sample. Ctgf (D), Ghr (E), Myc (F), and Taf4b (G) are significantly down-regulated in mutant granulosa cells (1, 0) vs. the wt cells. H, Kitl is significantly increased in activin-deficient (0) granulosa cells compared with wild type. Statistical significance by one-way ANOVA followed by Tukey-Kramer HSD post hoc tests. Different letters above each column (a and b) indicate statistical significance at P ⬍ 0.05.

granulosa cells (53), this does not appear to be a necessary physiological role for activin in vivo because there is an increase in Inha gene expression in activindeficient ovaries. Because IGF-I is required for activininduced Inha expression (54), compensation by IGF-I

on Inha expression through an alternative pathway may occur in activin-deficient females. Loss of dimeric serum inhibin likely favors FSH release due to loss of negative feedback by inhibin. The increased FSH may contribute to a number of defects, including a larger growing antral follicle pool present in the activin-deficient females, although not likely all. Female mice overexpressing FSH demonstrate hemorrhagic and cystic ovaries (55) but have a cystic phenotype much more severe than the phenotype that presents in activin-deficient females. Interestingly, even though the antral follicle pool is expanded, superovulation of immature female mice does not result in an increased ovulation rate. In fact, the ovulation rate in PMSG/human chorionic gonadotropin (hCG) stimulated activin-deficient females decreases approximately 60%. Gene expression analysis of granulosa cells from the preovulatory period indicates abnormal cell differentiation. Granulosa cells collected from PMSG-stimulated activin-deficient immature mice express more Kitl as measured by microarray and qPCR experiments, and kit ligand is a known mitogen for granulosa cells (33). Interestingly, the kit ligand-2 isoform is normally down-regulated during early antral follicle formation (56), and ovarian injection of antibody to the kit ligand receptor disrupts antral follicle granulosa cell proliferation (57). By in vitro experiments, we were able to show that activin is


capable of suppressing Kitl expression. During the antral follicle period, activin may be partly responsible for the down-regulation of kit ligand, resulting in the transition from an FSH-driven growth phase, to a differentiation phase required for the preovulatory period. A role for activin in regulating the terminal differentiation of ovarian follicles has been proposed (58, 59). We found that several other genes are down-regulated in preovulatory activin-deficient granulosa cells, including Ctgf, Ghr, Myc, and Taf4b. To date, only a few direct downstream targets of activin are known in granulosa cells. Ctgf is regulated by activin and other TGF␤ family members (31), but the consequence of loss-of-function of Ctgf in granulosa cells is unknown, and the contribution to the activin phenotype is unclear. In activin-deficient granulosa cells from immature mice, Myc and Taf4b are expressed less than in wild-type granulosa cells, and activin A up-regulates Taf4b in granulosa cells in culture. These experiments thus identify Taf4b as potential activin-response gene, although the activin pathway may not be the primary regulatory pathway because the expression changes are small, although significant. Taf4b homozygous null female mice are infertile, with defects in the inhibin/activin system and loss of expression of Inha, Inhba, Inhbb, and Ccnd2 (60). Thus, activin may be part of a positive regulatory loop involving TAF4B. However, Inha is expressed in activin-deficient ovaries as is Ccnd2. Differences in expression levels of Taf4b may explain the overall differences when comparing Taf4b knockout ovaries to activin-deficient ovaries (i.e. reduced levels of Taf4b in the activin-deficient mouse, compared with compete absence of Taf4b expression in Taf4b knockouts). The most striking phenotype of ␤B/␤A-deficient females is the progressive accumulation of CLs. Our data indicate a defect during structural regression because 20-␤HSD immunostaining is detectable in CLs of activin-deficient mice. Whereas no definitive data exist for activin in luteolysis, a number of studies suggest that activin plays an active role in regression of CLs. First, activin receptors localize to CLs in many species (61–63). And second, the activin antagonist follistatin is highly expressed in newly formed rodent CLs but undetectable during luteolysis (64), resulting in a permissive environment for activin signaling. Whereas the mechanism(s) by which activin could control structural regression of CLs is unknown, a recent study using a novel in vitro coculture system as a model for human luteolysis indicates that activin A from steroidogenic cells may directly induce matrix metalloproteinase 2 (Mmp2) expression through paracrine signaling to fibroblasts (62). MMPs are critical for tissue remodeling, and increased expression and activity of MMP2 is associated with luteolysis in women (65), and MMP2 levels also increase during the late luteal phase of monkeys (66). Whereas the hypothesis that activin-induced MMP2 expression is responsible for structural luteolysis remains to be tested in intact CLs, the activindeficient dKO mouse model should provide a unique system to test the roles of various hormones and growth factors to examine CL physiology, and in particular luteal regression.


The activin-deficient mouse model has higher levels of serum progesterone at 3 months of age. Activin signaling requires the common SMAD transcription factor, SMAD4 and in granulosa cell-specific knockouts of Smad4, progesterone levels are also increased (21). However, this is due to premature luteinization of granulosa cells in preovulatory follicles. Unlike Smad4 cKO ovaries, activin-deficient preovulatory granulosa cells do not luteinize when immature mice are given supraphysiological injections of PMSG. And in contrast to the activin-deficient mouse model, luteinized follicles/structures do not accumulate in the Smad4 cKO ovary, even though progesterone levels remain increased. Therefore, both models have higher levels of progesterone, although it appears to be through different mechanisms. Studies of human and macaque luteal cells find that activin inhibits progesterone synthesis (67, 68). Therefore, it is possible that the increase in progesterone in the activin-deficient mice results from a lack of activin activity in CLs, which cannot be compensated for by another pathway. BMPs and GDF9 are known to inhibit FSH-induced progesterone production and have been hypothesized to be luteinization inhibitors in granulosa cells of antral follicles (69, 70). Our data from the activin-deficient mouse model support this hypothesis and suggest that the BMP and GDF9 pathways, and not the activin pathway, may be the primary luteinization inhibitors in preovulatory follicles because the BMP and GDF9 pathways are disrupted in Smad4 cKO, but not in the activin-deficient conditional knockout. An alternative hypothesis is that during the preovulatory period, activin, GDF9 and/or BMP pathways may be redundant with respect to luteinization inhibition, with the BMP and/or GDF9 pathways compensating for loss of activin in the activin-deficient mouse model and preventing premature luteinization. The activin-deficient dKO ovary mouse model also highlights the widespread, cell-specific effects of activins in various cell types. In other tissues, activin signaling components act as tumor suppressors; for example, loss of activin receptors and SMAD signaling proteins, such as during the development of some pancreatic cancers, results in tumorigenesis. No tumor development was seen in the activin-deficient mouse models. Thus, in different cellular contexts, which may include expression of different transcription factors that interact with the downstream SMAD signaling proteins, the activins and their signaling components can either function as tumor suppressors to inhibit growth/stimulate differentiation, or as oncogenes to stimulate growth/inhibit differentiation. Thus, both loss-offunction and gain-of-function mutations in activin signaling pathway components may be observed to positively or negatively regulate cancer initiation and/or progression.

MATERIALS AND METHODS ES Cell Technology and Southern Blot Analysis A 21.6-kb of isogenic DNA sequence encompassing the twoexon mouse activin ␤A gene sequence was isolated from a 129S6/SvEv library and used to generate a conditional tar-



geting vector (␤Aflox). The targeting vector contains activin ␤A intron 1 sequence, a loxP site, activin ␤A exon 2 propeptide sequence, activin ␤A 3⬘untranslated region, polyadenylation sequence, 3⬘ sequences downstream of the activin ␤A gene, and a Pgk-Neo expression cassette flanked by loxP sites inserted into the 3⬘ downstream sequence. The floxed activin ␤A construct was electroporated into the ES cells (␤A5-F12) with a null mutation at the activin ␤A locus (␤Am1) (17) and clones isolated by positive-negative selection. The 23/35 ES cell clones (66%) targeted correctly at the activin ␤A locus. Two ES cell clones, ␤Aflox-B6 and ␤Aflox-C3 were injected into blastocysts as described (71) to produce chimeric mice. Male chimeras were fertile and transmitted the ␤Aflox allele to F1 progeny. Chimeras were mated to C57Bl6/J females to produce 129S6/SvEv/C57BL6/J hybrid mice. The phenotype of mice generated from the two different lines was identical. Southern blot analysis was used for genotyping. Southern blot analysis of granulosa cells was performed with mural granulosa cells isolated from large antral follicles of 3-wk-old control and experimental mice treated with PMSG for 48 h before granulosa cell collection as described in (72). Genotyping analysis of the Amhr2-cre allele and EIIa-cre and R26R transgene was determined using PCR. The sequences of the primers for the Amhr2-cre genotyping were 5⬘-CGCATTGTCTGAGTAGGTGT and 5⬘-GAAACGCAGCTCGGCCAGC. The sequences of the primers for EIIa-cre genotyping were 5⬘-CCGGGCTGCCACGACCAA and 5⬘-GGCGCGGCAACACCATTTTT.

ginia Ligand Core Facility (Specialized Cooperative Centers Program in Reproduction Research NICHD/NIH U54 HD28934). Assay information is available at (http://www.healthsystem.virginia.edu/internet/crr/ligand.cfm). The mouse FSH RIA has a sensitivity of 2.0 ng/ml, and an average intraassay coefficient of variation (CV) of 10.1 and interassay CV of 13.3%. The mouse LH Sandwich IRMA has a sensitivity of 0.07 ng/ml, and an average intraassay CV of 4.7% and interassay CV of 13.5%. The progesterone RIA has a sensitivity of 0.10 ng/ml and an average intraassay CV of 4.5% and interassay CV of 6.9%. The estradiol RIA has a sensitivity of 10 pg/ml, and an intraassay CV of 5.33%, and an interassay CV of 12.1%. The inhibin A ELISA has a sensitivity of assay 1 pg/ml and interassay variation of 6–7%. The percentage of cross reactivity of the inhibin A assay with other ligands was the following: inhibin B 0.012, activin A 0.002, activin B 0.001, and pro-␣ C 0.01. Values that fell below the assay threshold were given the threshold value. Serum samples were diluted in PBS to fall within the detectable range, when necessary. National Hormone and Peptide Program (NHPP) made PRL measurements by using a highly sensitive double-antibody method with reagents provided by Dr. A. F. Parlow (NHPP, Torrance, CA). Mouse PRL (AFP10777D) was used for iodination, rabbit antimouse PRL (AFP131078) was used as antiserum, and mouse PRL (AFP6476C) was used as reference preparation.

Morphological, Histological, and Immunohistochemical Analysis

Nineteen to 21-d-old and 6-month-old experimental and control female mice were injected with PMSG (ip, 5 IU/mouse), and given hCG (ip, 5 IU/mouse) 48 h later. Mice were bred to C57/129 hybrid known fertile males. The following morning oocytes were recovered from the ampulla of the oviduct in M2 medium and counted.

All experimental animals were maintained in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory animals. Embedding and staining procedures were performed by standard protocols at the Baylor College of Medicine Pathology Core Services laboratory. 20␣-HSD staining (1:300 dilution) was performed as described (21). Rabbit polyclonal anti-inhibin ␣ antibody was used at a 1:500 dilution and was a gift from W. Vale (The Salk Institute, La Jolla, CA). TROMA-III (anti-cytokeratin 19) rat monoclonal antibody was used at 1:10 dilution and was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences. Immunohistochemistry was performed on at least four samples in duplicate. For follicle histomorphometric analysis, a minimum of three ovaries was counted for each genotype. Follicle classification was based on Pederson and Peters (73). For assessment of primordial, primary, and preantral follicles, follicle counts were carried out as described (74, 75). Briefly, ovaries of each genotype were serially sectioned at 5 ␮m, and every tenth section was kept. Follicles were counted from five of the largest sections, and normalized to the total area of the section. Counts and area were collected using the AxioVision 4.0 software (Carl Zeiss, Jena, Germany). For assessment of antral follicles and CLs, the total number of antral follicles and CLs in each ovary were counted. To avoid double counting antral follicles, only follicles with a visible oocyte nucleus were counted. For CL counts, images were taken of each section (every 50 ␮m), ordered sequentially, and each CL followed through the sections, comparing CLs with those on previous sections to avoid double counting. Serum Analysis Mice were anesthetized by isoflurane inhalation (Abbott Laboratories, North Chicago, IL), blood recovered by closed cardiac puncture, and serum separated by centrifugation in Microtainer tubes (Becton Dickinson, Franklin Lakes, NJ) and stored at ⫺20 C. FSH, LH, estradiol, progesterone, and inhibin A measurements were made by the University of Vir-

Superovulation and Collection of Oocytes

Granulosa Cell Preparation For granulosa cell isolation, female mice 19–21 d of age were injected ip with 5 IU pregnant mare serum gonadotropins (Calbiochem, La Jolla, CA) for 44–46 h to stimulate follicle growth and granulosa cells collected by puncturing antral follicles with fine-gauge needles (72). RNA was extracted from mutant and control granulosa cells for microarray experiments and qPCR verification immediately upon harvest. For treatment of wild-type cells, 19–21 d CD1 female were used for granulosa cell collections and cells plated at a density of 5.5 ⫻ 105 cells/ml as described (76). Cells were incubated overnight in DMEM-F12 medium supplemented with 2% fetal calf serum, 1⫻ insulin/transferring/selenite (Invitrogen Life Technologies, Carlsbad, CA), 10 U/ml penicillin and streptomycin and 250 ng/ml recombinant follistatin-288 to block endogenous activin signaling. After 18 h, cells were washed twice in PBS and incubated with medium only with 50 ng/ml recombinant human activin A (R&D Biosystems, Minneapolis, MN) in DMEM-F12 medium supplemented with 0.5% heat inactivated fetal bovine serum, 1⫻ ITS, and 10 U/ml penicillin and streptomycin. Cells were treated for 5 h, then harvested for RNA. A minimum of four independent replicates was performed. RNA Isolation and Northern Blot Analysis For Northern blot analysis, total ovarian RNA was isolated from individual mice using RNA STAT-60 reagent (Leedo Medical Laboratories, Houston, TX). Twelve micrograms of each RNA sample were used for electrophoresis and transferred to nylon membranes as described previously (14) and probed with radioactive cDNA probe of 426 bp against activin ␤A (1049–1474 of the clone X69619) generated using [32P]dATP and the Strip-EZ kit (Ambion, Inc., Austin, TX).



Phosphorimaging plates were scanned and analyzed using Image Quant software (Molecular Dynamics, Inc., Sunnyvale, CA). A background level for each blot was determined and subtracted. Blots were stripped and reprobed for glyceraldehyde 3-phosphate dehydrogenase (Gapd) as a loading control.

(Microsoft). Statistical differences were tested using one-way ANOVA for multiple comparisons, followed by Tukey-Kramer honestly significant difference (HSD) or Dunnett’s post hoc tests as indicated in the text. Two-tailed Student’s t test was used for single comparisons at ␣ ⫽ 0.05. Statistics were performed on no less than three independent experiments.

Microarray Analysis and qPCR

Acknowledgments

RNA was isolated using the QIAGEN (Valencia, CA) RNeasy kit. RNA integrity, concentration, and quality were checked by the Baylor College of Medicine Microarray Core Facility. Labeling, hybridization, washing, scanning, and initial microarray analysis were performed by the Baylor College of Medicine Microarray Core Facility using standard Affymetrix protocols and the Affymetrix Mouse Genome 430 2.0 Array. Independent chips were analyzed from three independent wild-type and two independent activin-deficient (␤B⫺/⫺; ␤Aflox/⫺;Amhr2cre/⫹) granulosa cell samples. Each sample represented granulosa cells collected from a minimum of 2 immature mice stimulated for 44–46 h PMSG. Initial analysis was performed using the Microarray Analysis Suite 5.0 (Affymetrix) software to determine signal intensity and detection (absent, present, and marginal), and then data files were imported into GeneSpring GX for expression analysis. Data were filtered based on 1) MAS 5.0 call of “present” in a minimum of two samples; 2) a raw signal value of greater than 300 in a minimum of one of two conditions; 3) normalized expression change of 2-fold or greater. This resulted in a list of 280 genes, which were then analyzed for statistical significance at ␣ ⫽ 0.05 by one-way ANOVA; this should have resulted in 14 genes changes by chance alone. The final differentially expressed gene list contained 111 genes: 81 upand 30 down-regulated. Genes lists of up-regulated and down-regulated were imported into DAVID Bioinformatic software (http://david.abcc.ncifcrf.gov/home.jsp) for pathway analysis (77). The complete microarray data set will be available at NCBI Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo) with the accession no. GSE7150.

We thank Michal Klysik, Sankar Sridaran, Samuel Ogbonna, and Mujtaba Ali for genotyping, Dr. Kathleen Burns for ES cell injections, Dr. Heiner Westphal (National Institutes of Health) for the EIIa-cre transgenic mice, Dr. Richard Behringer (M. D. Anderson Cancer Center) for the Amhr2-cre mice, Dr. Teresa Woodruff (Northwestern University) for the recombinant follistatin-288, and Dr. Geula Gibori (University of Illinois, Chicago, IL) for the 20␣-dihydroxyprogesterone antibody. We thank Dr. Alfred Parlow at National Hormone and Peptide Program (Torrance, CA) for PRL measurements.

Received March 19, 2007. Accepted June 28, 2007. Address all correspondence and requests for reprints to: Martin M. Matzuk, M.D., Ph.D., The Stuart A. Wallace Chair and Professor, Department of Pathology, Baylor College of Medicine, One Baylor Plaza, Houston, Texas 77030. E-mail: [email protected]. This research was supported by National Institutes of Health (NIH) Grant HD32067 (to M.M.M.), National Institute of Child Health and Human Development/NIH U54 HD28934 (hormone analysis), and a Ruth L. Kirshstein National Research Service Award (5F32HD46335) and Burroughs Wellcome Career Award in the Biomedical Sciences (to S.A.P.). Data Deposition: Affymetrix microarray data files (five files) have been deposited to the NIH Gene Expression Omnibus under the series accession no. GSE7150 Disclosure Statement: The authors have nothing to disclose.

qPCR qPCR was performed as described (21). cDNA was prepared from 200 ng of total RNA using SuperScript III First Strand Synthesis Kit (Invitrogen) in a 50-␮l reaction. Real-time qPCR was performed as previously described (21) using Applied Biosystems (ABI; Foster City, CA) Prism 7500 Sequence Detection System, Taqman Master Mix and Gene Expression Assays. Taqman (ABI) Assays used for qPCR: Ccnd2, Mm00438071; Ctgf, Mm00439093; Fshr, Mm00442819; Gapd, 4352339E; Ghr, Mm00439093; Inhbb, Mm03023992; Lhcgr, Mm00442931; Myc, Mm00487803. Primers to Kitl, Taf4b and Gapd were designed using the Primer Expression 2.0 Software (ABI) and amplified using Sybr Green Master Mix (ABI) with Gapd as a control. Primer sequences available upon request. The Kitl primer set recognizes both splice variants, Kl-1 and Kl-2. Dissociation curves were generated for Sybr Green qPCRs to verify amplification of a single PCR product. All qPCR data were analyzed by the ⌬⌬ cycle threshold method using the ABI 7500 System Software (version 1.2.3) as described by ABI and normalized to the endogenous reference (Gapd). One control sample was randomly as the calibrator sample. The mean and SEM were calculated, and the relative amount of target gene expression for each sample compared with the control was plotted in Excel (Microsoft, Redmond, WA). Statistical Analysis Statistical analysis was carried out using the JMP version 5.1 (SAS Institute Inc., Cary, NC) statistical package or in Excel

REFERENCES 1. Chang H, Brown CW, Matzuk MM 2002 Genetic analysis of the mammalian TGF-␤ superfamily. Endocr Rev 23: 787–823 2. Vale W, Rivier J, Vaughan J, McClintock R, Corrigan A, Woo W, Karr D, Spiess J 1986 Purification and characterization of an FSH releasing protein from porcine ovarian follicular fluid. Nature 321:776–779 3. Vale W, Bilezikjian LM, Rivier C 1994 Reproductive and other roles of inhibins. In: Knobil E, Neill J, eds. The physiology of reproduction. 2nd ed. New York: Raven Press; 1861–1878 4. Matzuk MM 1995 Functional analysis of mammalian members of the transforming growth factor-␤ superfamily. Trends Endocrinol Metab 6:120–127 5. Vassalli A, Matzuk MM, Gardner H, Lee K, Jaenisch R 1994 Activin/inhibin ␤B subunit gene disruption leads to defects in eyelid development and female reproduction. Genes Dev 8:414–427 6. Schrewe H, Gendron-Maguire M, Harbison ML, Gridley T 1994 Mice homozygous for a null mutation of activin ␤B are viable and fertile. Mech Dev 47:43–51 7. Cocolakis E, Lemay S, Ali S, Lebrun JJ 2001 The p38 MAPK pathway is required for cell growth inhibition of human breast cancer cells in response to activin. J Biol Chem 276:18430–18436 8. Zhang Z, Zhao Y, Batres Y, Lin MF, Ying SY 1997 Regulation of growth and prostatic marker expression by activin A


9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

in an androgen-sensitive prostate cancer cell line LNCAP. Biochem Biophys Res Commun 234:362–365 Alexander JM, Bikkal HA, Zervas NT, Laws Jr ER, Klibanski A 1996 Tumor-specific expression and alternate splicing of messenger ribonucleic acid encoding activin/transforming growth factor-␤ receptors in human pituitary adenomas. J Clin Endocrinol Metab 81:783–790 Su GH, Bansal R, Murphy KM, Montgomery E, Yeo CJ, Hruban RH, Kern SE 2001 ACVR1B (ALK4, activin receptor type 1B) gene mutations in pancreatic carcinoma. Proc Natl Acad Sci USA 98:3254–3257 Hempen PM, Zhang L, Bansal RK, Iacobuzio-Donahue CA, Murphy KM, Maitra A, Vogelstein B, Whitehead RH, Markowitz SD, Willson JK, Yeo CJ, Hruban RH, Kern SE 2003 Evidence of selection for clones having genetic inactivation of the activin A type II receptor (ACVR2) gene in gastrointestinal cancers. Cancer Res 63:994–999 Mori Y, Yin J, Rashid A, Leggett BA, Young J, Simms L, Kuehl PM, Langenberg P, Meltzer SJ, Stine OC 2001 Instabilotyping: comprehensive identification of frameshift mutations caused by coding region microsatellite instability. Cancer Res 61:6046–6049 Rossi MR, Ionov Y, Bakin AV, Cowell JK 2005 Truncating mutations in the ACVR2 gene attenuates activin signaling in prostate cancer cells. Cancer Genet Cytogenet 163: 123–129 Elvin JA, Yan C, Wang P, Nishimori K, Matzuk MM 1999 Molecular characterization of the follicle defects in the growth differentiation factor-9-deficient ovary. Mol Endocrinol 13:1018–1034 Findlay JK, Drummond AE, Dyson ML, Baillie AJ, Robertson DM, Ethier JF 2002 Recruitment and development of the follicle; the roles of the transforming growth factor-␤ superfamily. Mol Cell Endocrinol 191:35–43 Muttukrishna S, Tannetta D, Groome N, Sargent I 2004 Activin and follistatin in female reproduction. Mol Cell Endocrinol 225:45–56 Matzuk MM, Kumar TR, Vassalli A, Bickenbach JR, Roop DR, Jaenisch R, Bradley A 1995 Functional analysis of activins in mammalian development. Nature 374:354–356 Lakso M, Pichel JG, Gorman JR, Sauer B, Okamoto Y, Lee E, Alt FW, Westphal H 1996 Efficient in vivo manipulation of mouse genomic sequences at the zygote stage. Proc Natl Acad Sci USA 93:5860–5865 Jamin SP, Arango NA, Mishina Y, Hanks MC, Behringer RR 2002 Requirement of Bmpr1a for Mullerian duct regression during male sexual development. Nat Genet 7:7 Jorgez CJ, Klysik M, Jamin SP, Behringer RR, Matzuk MM 2004 Granulosa cell-specific inactivation of follistatin causes female fertility defects. Mol Endocrinol 18:953–967 Pangas SA, Li X, Robertson EJ, Matzuk MM 2006 Premature luteinization and cumulus cell defects in ovarian-specific Smad4 knockout mice. Mol Endocrinol 20:1406–1422 Boerboom D, Paquet M, Hsieh M, Liu J, Jamin SP, Behringer RR, Sirois J, Taketo MM, Richards JS 2005 Misregulated Wnt/␤-catenin signaling leads to ovarian granulosa cell tumor development. Cancer Res 65:9206–9215 Brown CW, Houston-Hawkins DE, Woodruff TK, Matzuk MM 2000 Insertion of Inhbb into the Inhba locus rescues the Inhba-null phenotype and reveals new activin functions. Nat Genet 25:453–457 Peng XR, Hsueh AJ, LaPolt PS, Bjersing L, Ny T 1991 Localization of luteinizing hormone receptor messenger ribonucleic acid expression in ovarian cell types during follicle development and ovulation. Endocrinology 129: 3200–3207 Segaloff DL, Wang HY, Richards JS 1990 Hormonal regulation of luteinizing hormone/chorionic gonadotropin receptor mRNA in rat ovarian cells during follicular development and luteinization. Mol Endocrinol 4:1856–1865 Nakamura M, Nakamura K, Igarashi S, Tano M, Miyamoto K, Ibuki Y, Minegishi T 1995 Interaction between activin A


27.

28.

29.

30.

31.

32.

33.

34.

35.

36. 37.

38.

39.

40.

41.

42.

43.

44.

45.

and cAMP in the induction of FSH receptor in cultured rat granulosa cells. J Endocrinol 147:103–110 Minegishi T, Tano M, Nakamura K, Nakamura M, Igarashi S, Ito I, Shinozaki H, Karino S, Ibuki Y, Miyamoto K 1996 Regulation of follicle-stimulating hormone receptor. Horm Res 46:37–44 Ogawa T, Yogo K, Ishida N, Takeya T 2003 Synergistic effects of activin and FSH on hyperphosphorylation of Rb and G1/S transition in rat primary granulosa cells. Mol Cell Endocrinol 210:31–38 El-Hefnawy T, Zeleznik AJ 2001 Synergism between FSH and activin in the regulation of proliferating cell nuclear antigen (PCNA) and cyclin D2 expression in rat granulosa cells. Endocrinology 142:4357–4362 Eppig JJ, Wigglesworth K, Pendola F, Hirao Y 1997 Murine oocytes suppress expression of luteinizing hormone receptor messenger ribonucleic acid by granulosa cells. Biol Reprod 56:976–984 Harlow CR, Davidson L, Burns KH, Yan C, Matzuk MM, Hillier SG 2002 FSH and TGF-␤ superfamily members regulate granulosa cell connective tissue growth factor gene expression in vitro and in vivo. Endocrinology 143: 3316–3325 Thomas FH, Vanderhyden BC 2006 Oocyte-granulosa cell interactions during mouse follicular development: regulation of kit ligand expression and its role in oocyte growth. Reprod Biol Endocrinol 4:19 Otsuka F, Shimasaki S 2002 A negative feedback system between oocyte bone morphogenetic protein 15 and granulosa cell kit ligand: its role in regulating granulosa cell mitosis. Proc Natl Acad Sci USA 99:8060–8065 Joyce IM, Clark AT, Pendola FL, Eppig JJ 2000 Comparison of recombinant growth differentiation factor-9 and oocyte regulation of KIT ligand messenger ribonucleic acid expression in mouse ovarian follicles. Biol Reprod 63:1669–1675 del Re E, Sidis Y, Fabrizio DA, Lin HY, Schneyer A 2004 Reconstitution and analysis of soluble inhibin and activin receptor complexes in a cell-free system. J Biol Chem 279:53126–53135 Engle E 1927 Polyovular follicles and polynuclear ova in the mouse. Anat Rec 35:341–343 Jagiello G, Ducayen M 1973 Meiosis of ova from polyovular (C58-J) and polycystic (C57 L-J) strains of mice. Fertil Steril 24:10–14 McMullen ML, Cho BN, Yates CJ, Mayo KE 2001 Gonadal pathologies in transgenic mice expressing the rat inhibin ␣-subunit. Endocrinology 142:5005–5014 Lan ZJ, Xu X, Cooney AJ 2003 Generation of a germ cell nuclear factor conditional allele in mice. Genesis 37: 172–179 Yan C, Wang P, DeMayo J, DeMayo F, Elvin J, Carino C, Prasad S, Skinner S, Dunbar B, Dube J, Celeste A, Matzuk M 2001 Synergistic roles of bone morphogenetic protein 15 and growth differentiation factor 9 in ovarian function. Mol Endocrinol 15:854–866 Danilovich N, Roy I, Sairam MR 2001 Ovarian pathology and high incidence of sex cord tumors in follitropin receptor knockout (FORKO) mice. Endocrinology 142:3673–3684 Chen Y, Jefferson WN, Newbold RR, Padilla-Banks E, Pepling ME 2007 Estradiol, progesterone and genistein inhibit oocyte nest breakdown and primordial follicle assembly in the neonatal mouse ovary in vitro and in vivo. Endocrinology 148:3580–3590 Iguchi T, Fukazawa Y, Uesugi Y, Takasugi N 1990 Polyovular follicles in mouse ovaries exposed neonatally to diethylstilbestrol in vivo and in vitro. Biol Reprod 43:478–484 Iguchi T, Takasugi N 1986 Polyovular follicles in the ovary of immature mice exposed prenatally to diethylstilbestrol. Anat Embryol (Berl) 175:53–55 Iguchi T, Takasugi N, Bern HA, Mills KT 1986 Frequent occurrence of polyovular follicles in ovaries of mice exposed neonatally to diethylstilbestrol. Teratology 34:29–35



46. Iguchi T, Todoroki R, Takasugi N, Petrow V 1988 The effects of an aromatase inhibitor and a 5␣-reductase inhibitor upon the occurrence of polyovular follicles, persistent anovulation, and permanent vaginal stratification in mice treated neonatally with testosterone. Biol Reprod 39: 689–697 47. Kipp JL, Kilen SM, Bristol-Gould S, Woodruff TK, Mayo KE 2007 Neonatal exposure to estrogens suppresses activin expression and signaling in the mouse ovary. Endocrinology 148:1968–1976 48. Li R, Phillips DM, Mather JP 1995 Activin promotes ovarian follicle development in vitro. Endocrinology 136:849–856 49. Miro F, Hillier S 1996 Modulation of granulosa cell deoxyribonucleic acid synthesis and differentiation by activin. Endocrinology 137:464–468 50. Zhao J, Taverne MA, van der Weijden GC, Bevers MM, van den Hurk R 2001 Effect of activin A on in vitro development of rat preantral follicles and localization of activin A and activin receptor II. Biol Reprod 65:967–977 51. Yokota H, Yamada K, Liu X, Kobayashi J, Abe Y, Mizunuma H, Ibuki Y 1997 Paradoxical action of activin A on folliculogenesis in immature and adult mice. Endocrinology 138: 4572–4576 52. Woodruff T, Besecke L, Groome N, Draper L, Schwartz N, Weiss J 1996 Inhibin A and inhibin B are inversely correlated to follicle-stimulating hormone, yet are discordant during the follicular phase of the rat estrous cycle, and inhibin A is expressed in a sexually dimorphic manner. Endocrinology 137:5463–5467 53. LaPolt PS, Soto D, Su JG, Campen CA, Vaughan J, Vale W, Hsueh AJ 1989 Activin stimulation of inhibin secretion and messenger RNA levels in cultured granulosa cells. Mol Endocrinol 3:1666–1673 54. Kubo T, Shimasaki S, Kim H, Li D, Erickson GF 1998 Activin-induced inhibin ␣-subunit production by rat granulosa cells requires endogenous insulin-like growth factor-I. Biol Reprod 58:712–718 55. Kumar TR, Palapattu G, Wang P, Woodruff TK, Boime I, Byrne MC, Matzuk MM 1999 Transgenic models to study gonadotropin function: the role of follicle-stimulating hormone in gonadal growth and tumorigenesis. Mol Endocrinol 13:851–865 56. Thomas FH, Ethier JF, Shimasaki S, Vanderhyden BC 2005 Follicle-stimulating hormone regulates oocyte growth by modulation of expression of oocyte and granulosa cell factors. Endocrinology 146:941–949 57. Yoshida H, Takakura N, Kataoka H, Kunisada T, Okamura H, Nishikawa S-I 1997 Stepwise requirement of c-kit tyrosine kinase in mouse ovarian follicle development. Dev Biol 184:122–137 58. Findlay JK 1993 An update on the roles of inhibin, activin, and follistatin as local regulators of folliculogenesis. Biol Reprod 48:15–23 59. Hillier SG, Miro F 1993 Inhibin, activin, and follistatin. Potential roles in ovarian physiology. Ann NY Acad Sci 687: 29–38 60. Freiman RN, Albright SR, Zheng S, Sha WC, Hammer RE, Tjian R 2001 Requirement of tissue-selective TBP-associated factor TAFII105 in ovarian development. Science 293: 2084–2087 61. Cameron V, Nishimura E, Mathews L, Lewis K, Sawchenko P, Vale W 1994 Hybridization histochemical localization of

activin receptor subtypes in rat brain, pituitary, ovary, and testis. Endocrinology 134:799–808 Myers M, Gay E, McNeilly AS, Fraser HM, Duncan WC 2007 In vitro evidence suggests activin-A may promote tissue remodeling associated with human luteolysis. Endocrinology 148:3730–3739 van den Hurk R, Van de Pavert SA 2001 Localization of an activin/activin receptor system in the porcine ovary. Mol Reprod Dev 60:463–471 Erickson GF, Shimasaki S 2003 The spatiotemporal expression pattern of the bone morphogenetic protein family in rat ovary cell types during the estrous cycle. Reprod Biol Endocrinol 1:9–28 Duncan WC, McNeilly AS, Illingworth PJ 1998 The effect of luteal “rescue” on the expression and localization of matrix metalloproteinases and their tissue inhibitors in the human corpus luteum. J Clin Endocrinol Metab 83:2470–2478 Young KA, Hennebold JD, Stouffer RL 2002 Dynamic expression of mRNAs and proteins for matrix metalloproteinases and their tissue inhibitors in the primate corpus luteum during the menstrual cycle. Mol Hum Reprod 8:833–840 Di Simone N, Lanzone A, Petraglia F, Ronsisvalle E, Caruso A, Mancuso S 1994 Effect of activin-A on progesterone synthesis in human luteal cells. Fertil Steril 62:1157–1161 Brannian JD, Woodruff TK, Mather JP, Stouffer RL 1992 Activin-A inhibits progesterone production by macaque luteal cells in culture. J Clin Endocrinol Metab 75:756–761 Shimasaki S, Zachow RJ, Li D, Kim H, Iemura S-I, Ueno N, Sampath K, Chang RJ, Erickson GF 1999 A functional bone morphogenetic protein system in the ovary. Proc Natl Acad Sci USA 96:7282–7287 Vitt UA, Hayashi M, Klein C, Hsueh AJ 2000 Growth differentiation factor-9 stimulates proliferation but suppresses the follicle-stimulating hormone-induced differentiation of cultured granulosa cells from small antral and preovulatory rat follicles. Biol Reprod 62:370–377 Matzuk M, Finegold M, Su J, Hsueh A, Bradley A 1992 ␣-Inhibin is a tumor-suppressor gene with gonadal specificity in mice. Nature 360:313–319 Elvin JA, Clark AT, Wang P, Wolfman NM, Matzuk MM 1999 Paracrine actions of growth differentiation factor-9 in the mammalian ovary. Mol Endocrinol 13:1035–1048 Pedersen T, Peters H 1968 Proposal for a classification of oocytes and follicles in the mouse ovary. J Reprod Fertil 17:555–557 Hu YC, Wang PH, Yeh S, Wang RS, Xie C, Xu Q, Zhou X, Chao HT, Tsai MY, Chang C 2004 Subfertility and defective folliculogenesis in female mice lacking androgen receptor. Proc Natl Acad Sci USA 101:11209–11214 Cheng G, Weihua Z, Makinen S, Makela S, Saji S, Warner M, Gustafsson JA, Hovatta O 2002 A role for the androgen receptor in follicular atresia of estrogen receptor ␤ knockout mouse ovary. Biol Reprod 66:77–84 Pangas SA, Jorgez CJ, Matzuk MM 2004 Growth differentiation factor 9 regulates expression of the bone morphogenetic protein antagonist, gremlin. J Biol Chem 279: 32281–32286 Dennis Jr G, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA 2003 DAVID: database for annotation, visualization, and integrated discovery. Genome Biol 4:P3

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

Molecular Endocrinology is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the endocrine community.