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 Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 Activins have diverse roles in multiple physiologi- are subfertile, ␤B/␤A double mutant females are cal processes including reproduction. Mutations infertile. Strikingly, the activin ␤A and ␤B/␤A-defi- and loss of heterozygosity at the human activin cient ovaries contain increased numbers of func- receptor ACVR1B and ACVR2 loci are observed in tional corpora lutea but do not develop ovarian pituitary, pancreatic, and colorectal cancers. Func- tumors. Microarray analysis of isolated granulosa tional studies support intraovarian roles for ac- cells identifies significant changes in expression tivins, although clarifying the in vivo roles has re- for a number of genes with known reproductive mained elusive due to the perinatal death of activin roles, including Kitl, Taf4b, and Ghr, as well as loss ␤A knockout mice. To study the roles of activins in of expression of the proto-oncogene, Myc. Thus, in ovarian growth, differentiation, and cancer, a tis- contrast to the known tumor suppressor role of sue-specific knockout system was designed to ab- activins in some tissues, our data indicate that activin late ovarian production of activins. Mice lacking ␤A and ␤B function redundantly in a growth stimu- ovarian activin ␤A were intercrossed to Inhbb ho- latory pathway in the mammalian ovary (Molecular mozygous null mice to produce double activin Endocrinology 21: 2458–2471, 2007) knockouts. Whereas ovarian ␤A knockout females

CTIVINS, HOMODIMERS OR heterodimers of ␤Aor growth of both human breast cancer cells (7) and prostate A␤B-subunits, were identified for their roles in posi- cancer cells (8). Mutations in ACVR1B have been observed tively regulating pituitary FSH synthesis and secretion in human pituitary tumors (9) and pancreatic cancers (10). (1–3). The ␤-subunits also dimerize with the inhibin Likewise, ACVR2 mutations and loss of heterozygosity ␣-subunit to produce inhibins [inhibin A (␣:␤A) and in- have been seen in pancreatic (11), colorectal (12), and pros- hibin B (␣:␤B)] that negatively regulate FSH. The ␤A- and tate cancers (13). These studies suggest that the activin ␤B-subunits, encoded by the Inhba and Inhbb genes, signaling pathway functions as a tumor suppressor path- respectively, are produced in multiple tissues during em- way in some tissues to block cell growth and stimulate bryonic and postnatal development and have diverse differentiation. physiological effects. Mice lacking activin ␤A (homozy- Inhba and Inhbb transcripts are prominently expressed gous Inhba null) die at birth secondary to craniofacial by granulosa cells of large preantral and antral follicles (14). defects, whereas homozygous Inhbb null mice are viable In vitro studies have identified multiple intraovarian roles for but exhibit eyelid closure and nursing defects (4–6). activins in granulosa cells including proliferation, potentia- Several studies indicate that activin signaling compo- tion of FSH action, and modulation of steroidogenesis (re- nents, including the receptors ACVR1B (also known as viewed in Refs. 15 and 16). To study the in vivo roles of activins, we produced an ovarian Inhba conditional knock- ALK4) and ACVR2 (also known as ACTR2B), are critical for out (␤A cKO). We also generated the ␤A cKO in the Inhbb growth inhibition. For example, activins inhibit in vitro null background mice to produce mice lacking all ovarian First Published Online July 3, 2007 activins (double mutant mice; herein designated as ␤B/␤A * S.A.P. and C.J.J. contributed equally to this work. dKO). In contrast to extragonadal roles of these proteins as Abbreviations: BMP, Bone morphogenetic protein; cKO, tumor suppressors, activins play redundant roles to block conditional knockout; CL(s), corpus luteum/corpora lutea; Ctgf, connective tissue growth factor; CV, coefficient of vari- terminal differentiation of granulosa cells, leading to dos- ation; dKO, double mutant mice; Gapd, glyceraldehyde age-dependent fertility defects. 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, preg- RESULTS nant mare serum gonadotropin; PRL, prolactin; qPCR, quantitative PCR. Generation of Ovarian Activin ␤A Knockout Mice Molecular Endocrinology is published monthly by The Endocrine Society (http://www.endo-society.org), the tm3Zuk foremost professional society serving the endocrine A floxed Inhba allele was generated (Inhba ; community. herein called ␤Aflox) for subsequent tissue-specific de-

2458 Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2459

letion of the Inhba gene in vivo (Fig. 1A). Exon 2 was ␤BϪ/Ϫ; ␤Aflox/Ϫ, and ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ fe- floxed because it encodes the entire mature domain of males. Neither ␤Aϩ/Ϫ nor ␤Bϩ/Ϫ heterozygous mice the protein and a conditional exon 2 deletion would had fertility defects (5, 17). Double mutant females mimic the original null allele (Inhbatm1Zuk) (17). ␤Aflox/ϩ with only one remaining ␤A-subunit allele (␤BϪ/Ϫ; mice were intercrossed to produce ␤Aflox/flox homozy- ␤Aϩ/Ϫ or ␤BϪ/Ϫ;␤Aflox/Ϫ) had severe fertility defects gous mice, which were viable and fertile and obtained (Table 1), indicating that activin dosage plays an im- at the expected Mendelian frequency. To verify the portant role in ovarian function. Of the 16 mice that presence of loxP sites and their ability to recombine in were studied over 6 months, three of five ␤BϪ/Ϫ; vivo, we crossed the ␤Aflox mice to EIIa-cre transgenic ␤Aflox/Ϫ, and three of 11 ␤BϪ/Ϫ;␤Aϩ/Ϫ female mice mice, which express cre recombinase in multiple tis- were infertile. In the mutant lines containing a single ␤ sues, including germ cells (18). Southern blot analysis -subunit allele, the average number of litters per Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 using a 5Ј probe demonstrated the presence of the month was significantly reduced (89–91%), and the various Inhba alleles, including the recombined de- number of pups per litter was 87–90% lower than the leted (␤A⌬) (data not shown). Because ␤Aflox/Ϫ;EIIa-cre controls (Table 1). Finally, female mice lacking all in- newborn mice demonstrate the neonatal lethality seen traovarian activin ␤A and ␤B alleles (␤BϪ/Ϫ;␤Aflox/Ϫ; in the Inhba homozygous null mice (data not shown), Amhr2cre/ϩ) were infertile (Table 1). Thus, although recombination likely produced a null allele. activin A appears to be functionally dominant [be- Anti-Mu¨ llerian hormone receptor-cre (Amhr2cre/ϩ) cause activin A deficiency results in subfertility vs. knock-in mice were used to generate recombination of normal fertility for activin ␤B deficiency (5)], dosage of ␤Aflox in ovarian granulosa cells (19–21). ␤Aϩ/Ϫ; the activin ␤A and ␤B alleles is important for normal Amhr2cre/ϩ mice were mated to ␤Aflox/flox mice and all fertility. four genotypes were recovered at the expected Men- To assess whether ovulation defects were present in delian frequency. Recombination efficiency of the the activin-deficient females, 3-wk-old females were ␤Aflox locus was determined by Southern blot analysis superovulated, mated and oocytes collected from ovi- of granulosa cells derived from cre-negative ␤Aflox/Ϫ ducts. Four groups of mice were analyzed: wild-type, (control) mice and ␤Aflox/Ϫ;Amhr2cre/ϩ (experimental) ␤Aflox/Ϫ, ␤Aflox/Ϫ;Amhr2cre/ϩ, ␤BϪ/Ϫ;␤Aflox/Ϫ, and mice (Fig. 1B). Recombination of the floxed allele in ␤BϪ/Ϫ;␤Aflox/;Amhr2cre/ϩ. There were no significant granulosa cells from female littermates varied from differences in the number of oocytes released after 60% to 99% (Fig. 1B). Variation in recombination and ovulation, except in mice deficient for all ␤-subunits expression differences are likely the result of mo- (Table 2). Two of four female mice deficient for all sacism in recombination, as has been seen in other activin alleles produced no oocytes. conditional knockout mice using Amhr2-cre mice (21, 22). Loss of the ␤A transcript was confirmed by North- Enhanced Corpus Luteum (CL) Formation in ern blot analysis (Fig. 1C), and no compensatory up- Activin-Deficient Ovaries regulation of the activin ␤B-subunit was found in ␤Aflox/Ϫ;Amhr2cre/ϩ ovaries (Fig. 1D). Because the To determine the causes of subfertility/infertility in the ␤-subunit is also required for formation of inhibin, we activin-deficient mice, ovaries were analyzed at 3 and measured serum inhibin levels in the ␤Aflox/Ϫ; 8 months of age. At 3 months of age, there were no Amhr2cre/ϩmice. The average serum inhibin A level in gross differences in appearance or size of the ovaries the ␤A-deficient mice is significantly reduced from the between ␤Aflox/Ϫ;Amhr2cre/ϩ and control mice control values, and six of nine females had undetect- (␤Aflox/Ϫ) (data not shown). The only visible histological able levels. Loss of dimeric inhibin A was not due to difference at 3 months of age was the presence of reductions in inhibin ␣-subunit mRNA expression be- ovarian follicles containing more than one oocyte cause ovaries and granulosa cells of mutant mice [50% occurrence in the experimental mice vs. none of demonstrated increased Inha mRNA levels over con- the controls (␤Aflox/Ϫ) (Fig. 2, A and B)]. However, at 8 trols (Fig. 1E, inset, lane 4). months of age, ovaries from the ␤Aflox/Ϫ;Amhr2cre/ϩ mice contained a greater abundance of CLs (Fig. 2D) Redundant Roles of Activin ␤A and ␤Bin as compared with control ␤Aflox/Ϫ mice (Fig. 2C). Fe- Female Fertility male mice with only one remaining activin ␤ allele (␤BϪ/Ϫ;␤Aϩ/Ϫ) showed an earlier increase in the num- The fertility of control (␤Aflox/Ϫ) mice was comparable ber of CLs (Fig. 2E) (i.e. the phenotype appears at 3 with wild-type mice (data not shown). However, ␤Aflox/Ϫ; months of age instead of at 8 months of age for the Amhr2cre/ϩ female mice demonstrated a statistically single ␤A conditional knockout). By 8 months of age, significant reduction (43%) in litters per month and the ␤BϪ/Ϫ;␤Aϩ/Ϫ ovaries contained a readily identifi- pups per litter (33%) compared with controls (Table 1 able abundance of CLs (Fig. 2F). and data not shown). Because subfertility could result A more extreme version of the phenotype was vis- from redundancy between the activin A and B iso- ible in mice deficient for all activins (i.e. in ␤BϪ/Ϫ; forms (23), we generated ␤A conditional knockouts in ␤Aflox/Ϫ;Amhr2cre/ϩ females) (Fig. 3). Abundant CLs the activin ␤B null (␤BϪ/Ϫ) background and addition- were visible histologically at 6 wk (Fig. 3C) and 3 ally analyzed the following genotypes: ␤BϪ/Ϫ;␤Aϩ/Ϫ, months of age (Fig. 3D). Also at 8–11 wk of age, there 2460 Mol Endocrinol, October 2007, 21(10):2458–2471 Pangas et al.•In Vivo Roles of Activins in the Ovary Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021

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 3Јuntranslated region of the activin ␤A gene. Four alleles shown are: WT, ␤AϪ, ␤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 5Ј 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/Ϫ) mouse. In lanes 2–4, ␤Aflox/Ϫ;Amhr2cre/ϩ 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/Ϫ) mice show expression of the activin ␤A-subunit. Lanes 1 and 4, ␤Aflox/Ϫ;Amhr2cre/ϩ 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/Ϫ (n ϭ 3) and experimental ␤Aflox/Ϫ;Amhr2cre/ϩ(n ϭ 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/Ϫ (n ϭ 8) and experimental ␤Aflox/Ϫ;Amhr2cre/ϩ (n ϭ 9) females show a statistically significant decrease by ELISA for inhibin A (P Ͻ 0.05). Inhibin A levels in mice lacking all ␤-subunits in the ovary (␤Bϩ/Ϫ;␤Aflox/Ϫ;Amhr2cre/ϩ, third column) (n ϭ 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 ␤Aϩ/Ϫ;␤BϪ/Ϫ; lane 4, ␤Aflox/Ϫ;Amhr2cre/ϩ; Gapd was used as control. BHI, BamHI; ERV, EcoRV; PI, PstI. Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2461

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/Ϫ 1 2 3 10 50 7.78 Ϯ 0.48a 0.89 Ϯ 0.04a ␤Aflox/Ϫ; creϩ 0 2 2 10 36 5.19 Ϯ 0.65b 0.51 Ϯ 0.05b ␤BϪ/Ϫ; ␤Aϩ/Ϫ 1 0 1 10 5 0.80 Ϯ 0.49c 0.08 Ϯ 0.08c ␤BϪ/Ϫ; ␤Aflox/Ϫ 1 0 1 5 2 1.00 Ϯ 0.63b,c 0.10 Ϯ 0.07c ␤BϪ/Ϫ; ␤Aflox/Ϫ;creϩ 00 090 0c 0c

Values are average Ϯ SEM. Statistical significance by one-way ANOVA followed by Tukey-Kramer HSD test (P Ͻ 0.05). Values with different superscripts (a–c) indicate statistically different means. n, Number of animals. Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 were statistically significant increases in the number of ing the growth of CLs and production of progesterone antral follicles (P Ͻ 0.05) (Fig. 4A) and in the number of is LH. The LH receptor (Lhcgr) is expressed highly in CLs (P Ͻ 0.01) (Fig. 4B) only in ␤BϪ/Ϫ; ␤Aflox/Ϫ; mural granulosa cells of preovulatory follicles and Amhr2cre/ϩ females, vs. all tested genotypes. No healthy CLs (14, 24, 25). By in situ hybridization for changes were found in the numbers of developing Lhcgr, several healthy CLs in mice lacking one to four primordial, primary, or preantral follicles for any geno- ␤-subunits were detected (Fig. 6), and expression of type at 6–11 wk of age (data not shown). At 8 months LH receptor was higher in those mice with increased of age, ovaries of ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ mice had numbers of CLs. To examine whether CLs in experi- a 6-fold increase in mass [31.5 Ϯ 2.9 mg (n ϭ 8) vs. mental mice could undergo functional regression, we 5.2 Ϯ 2.9 mg (n ϭ 8) in the ␤Aflox/Ϫ cKO mice] (Fig. 3A). examined expression of 20␣-hydroxysteroid dehydro- Also present were follicles with luteinizing granulosa genase (20␣-HSD), which catalyzes the conversion of cells (Fig. 3H), and multiple cysts that were either progesterone into 20␣-dihydroxyprogesterone, a bio- fluid-filled or hemorrhagic (Fig. 3, F and I). Cells lining logically inactive form. Immunostaining for 20␣-HSD in cysts are immunopositive for the granulosa cell the experimental and control mice identified functional marker, inhibin ␣ (Fig. 3, J and K), and negative for regressing corpora even in mice with one or no re- epithelial markers (Fig. 3L and data not shown), and maining ␤-alleles wild type (Fig. 6, F and I). These thus likely derive from ovarian follicles. results suggest that part of the phenotype of the ex- The serum levels of estradiol, prolactin (PRL), pro- perimental mice may be that the rate of CL formation gesterone, and the gonadotropins, FSH and LH, were in activin-deficient mice might be greater than the rate measured to examine ovarian function at 3 months of of regression. age (Fig. 5). There were no significant differences in basal serum levels of LH (Fig. 5B), estradiol (Fig. 5D), Gene Expression Differences in Granulosa Cells or PRL (data not shown) in random cycling female from Activin-Deficient Mice mice for any genotype. However, a statistically significant increase (P Ͻ 0.01) was identified for FSH in mice Several genes have previously been shown to be reg- lacking all activin subunits (Fig. 5A). In addition, there ulated by activin in granulosa cell culture experiments. was a trend of increasing serum progesterone corre- We initially used quantitative PCR (qPCR) to examine sponding with the decreasing number of ␤-subunit granulosa cell gene expression in wild-type mice, alleles, which reached statistical significance (P Ͻ ␤Aflox/Ϫ;␤BϪ/Ϫ mice, and ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ 0.001) at 3 months of age for ␤BϪ/Ϫ; ␤Aflox/Ϫ; mice using a candidate gene approach. FSH receptor Amhr2cre/ϩ females (Fig. 5C). (Fshr) and cyclin D2 (Ccnd2) are reported to be up- In mouse ovaries, CLs can be classified into two regulated by activin (26–29). Sequential loss of the stages, healthy and regressing; regressing CLs can be activin ␤-subunits reduced Fshr mRNA expression further subclassified into functional and structural re- (Fig. 7A), but this trend was not statistically significant gression stages. One of the major hormones support- 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 2 2 4 8 39.1 Ϯ 5.5 ␤Aflox/Ϫ 1 2 3 5 31.6 Ϯ 5.8 ␤Aflox/Ϫ;creϩ 0 2 2 12 36.3 Ϯ 4.4 ␤BϪ/Ϫ; ␤Aϩ/Ϫ 1 0 1 12 35.3 Ϯ 6.8 ␤BϪ/Ϫ; ␤Aflox/Ϫ;creϩ 0 0 0 4 15.8 Ϯ 10.5a

Values are average Ϯ SEM. Statistical significance by one-way ANOVA followed by Tukey-Kramer HSD test (P Ͻ 0.05). a By ANOVA, none of the groups are statistically significant for mean oocytes. However, 2 of 4 ␤BϪ/Ϫ;␤Aflox/Ϫ;creϩ females yielded no ovulated oocytes, and a t test between wild type and ␤BϪ/Ϫ;␤Aflox/Ϫ;creϩ is statistically significant (P ϭ 0.43). 2462 Mol Endocrinol, October 2007, 21(10):2458–2471 Pangas et al.•In Vivo Roles of Activins in the Ovary

mice and mice that were deficient for all ␤-subunits (␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ), 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)-stimu-

lated females, differentially expressed genes represent Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 genes expression changes induced in growing antral follicles by gonadotropin stimulation with or without endogenous activin. One hundred and eleven nonre- dundant genes showed statistically significant (P Ͻ 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 ϭ 0.005), as would be predicted by loss of activin alleles, the JAK-STAT pathway (P ϭ 0.023), and the cytokine- cytokine receptor interaction pathway (P ϭ 0.016). Fig. 2. Histological Analysis of the Ovaries of Activin ␤A Conditional Mice and ␤Aϩ/Ϫ;␤BϪ/Ϫ Mutant Mice Genes differentially expressed in these pathways are A, Ovary from a 3-month-old ␤Aflox/Ϫ;Amhr2cre/ϩ female highlighted in Supplemental Tables 1 and 2. with several CLs, follicles at different stages, and one follicle Expression of several differentially expressed genes containing two oocytes (arrow). B, Higher magnification of with known ovarian roles were verified by qPCR on polyovular follicle from panel A. C, Ovary from an 8-month- granulosa cells from individual mice collected inde- flox/Ϫ old ␤A control mouse with follicles at different stages pendently from the samples used in the microarray ␤ flox/Ϫ and several CLs. D, Ovary from an 8-month-old A ; analysis (Fig. 7, D–H, and Supplemental Table 1). In Amhr2cre/ϩ mouse with a large increase in the number of CLs. general, the fold changes found for qPCR verification E, Ovary from 3-month-old ␤BϪ/Ϫ;␤Aϩ/Ϫ mouse demonstrat- ing increased numbers of CLs and decreased numbers of in these experiments were larger than fold changes visible developing follicles. F, Ovary from an 8-month-old determined from microarray data. Likely, this results ␤BϪ/Ϫ;␤Aϩ/Ϫ female mouse with a large number of CLs and from the use in the microarray analysis of pooled RNA almost no remaining developing follicles. Scale bar, 400 ␮m. from samples that have variable recombination rates AnF, Antral follicle; Oo, oocyte. (Fig. 1B), leading to more conservative estimates of gene expression changes. We found that connective tissue growth factor (Ctgf) was significantly down-reg- notypes (P ϭ 0.08). These data suggest that in vivo, ulated in the microarray (4.7-fold) and by qPCR (1 ␤ ␤ activin likely augments FSH-induced expression of -allele, 16-fold; 0 -alleles, 26-fold) in mutant mice Fshr, similar to its effects in vitro (28, 29), but is not (Fig. 7D). Importantly, Ctgf has been shown to be ␤ necessary for basal FSH receptor expression. Unex- regulated by several members of the TGF family, pectedly, reduction of activin alleles had no effect on including activin (31). Other significantly down-regu- Ccnd2 expression (Fig. 7B). We also examined ex- lated genes in mutant mice include GH receptor (Ghr) pression of Lhcgr and found that its expression was (1 ␤-allele and 0 ␤-alleles, 14-fold) (Fig. 7E), myelocy- significantly up-regulated (P Ͻ 0.05) in granulosa cells tomatosis oncogene (Myc)(1␤-allele, 13-fold; 0 ␤-al- of mutant mice (Fig. 7C). These gene expression re- leles, 15-fold) (Fig. 7F), and the TATA-binding factor- sults are similar to the previously reported gene ex- associated factor (Taf4b)(1␤-allele, 3.9-fold; 0 pression pattern of Smad4 granulosa cell-specific ␤-alleles, 4.4-fold) (Fig. 7G). It is unknown whether knockout mice, which also show an increase in Lhcgr, these genes are activin target genes. To determine but little effect on Ccnd2 (21). Misexpression of Lhcgr whether Ghr, Myc, or Taf4b are regulated by activin, in activin-deficient granulosa cells is consistent with we treated wild-type granulosa cells with recombinant an inappropriate differentiation of granulosa cells (30). human activin A and measured changes in mRNA To further identify changes in gene expression in expression by qPCR (Fig. 8). Levels of Ghr and Myc granulosa cells, we performed microarray analysis on were unchanged after 5 h, indicating that short-term wild-type and activin-deficient granulosa cells. Gran- activin treatment does not alter their transcription (Fig. ulosa cells were collected from immature wild-type 8, A and B). However, a statistically significant (P Ͻ Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2463 Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021

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/Ϫ;Amhr2cre/ϩ (␤A cKO) mouse (left) showing normal appearance. In contrast, ovaries from ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ 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 ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩmouse with follicles of all stages, but no CL development. C, Ovary from 6-wk-old from ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩmouse with several preovulatory follicles (arrow- heads) and many CLs (one CL indicated). D, Ovary from 3-month-old ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ 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 ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ dKO mouse with a large number of CLs, and few antral follicles. F, Ovary from an 8-month-old ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ 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 ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ 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) granulosa cells as assayed by microarray (2-fold) after 5 h. (Supplemental Table 2) and qPCR (1 ␤-allele, 3-fold; 0 Fewer genes were significantly up-regulated than ␤-alleles, 5-fold) (Fig. 7H). Kit ligand is a positive reg- down-regulated in the microarray analysis, but several ulator of follicle growth (32) and has important growth have important ovarian roles, including kit ligand (Kitl). stimulatory effects on granulosa cells (33). The expres- Of the up-regulated genes, Kitl increased in mutant sion of kit ligand has been shown to be increased by 2464 Mol Endocrinol, October 2007, 21(10):2458–2471 Pangas et al.•In Vivo Roles of Activins in the Ovary Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021

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/Ϫ; group (2) is Fig. 4. Antral Follicle and CL Counts in Activin Mutant Mice ␤Aflox/Ϫ;Amhr2cre/ϩ; group (1) is ␤BϪ/Ϫ;␤Aϩ/Ϫ; group (0) is at 6–11 Wk of Age ␤BϪ/Ϫ; ␤Aflox/Ϫ;Amhr2cre/ϩ. A, FSH values significantly in- Ovaries were serially sectioned and the total number of crease (P Ͻ 0.01) with decreasing numbers of ␤ alleles. antral follicles and CLs were counted for the following geno- Sample sizes for FSH values are (3) n ϭ13; (2) n ϭ 11; (1) n ϭ types: ␤Bϩ/Ϫ;␤Aϩ/Ϫ;Amhr2cre/ϩ (2 ␤ alleles), ␤Bϩ/Ϫ;␤Aflox/Ϫ; 13; (0) n ϭ 6. B, LH values do not change across genotypes. Amhr2cre/ϩ (1 ␤ allele); ␤BϪ/Ϫ;␤Aflox/Ϫ (1 ␤ allele); and ␤BϪ/Ϫ; Sample sizes for LH values are (3) n ϭ 12; (2) n ϭ 12; (1) n ϭ ␤Aflox/Ϫ;Amhr2cre/ϩ (0 ␤ alleles). A, There are statistically sig- 10; (0) n ϭ 3. C, Progesterone levels are statistically signifi- nificant increases in the number of antral follicles (fourth bar) cant (P Ͻ 0.001) for mice with no activin alleles (0) (hatched (P Ͻ 0.05) and B, CL (fourth bar)(P Ͻ 0.01), in mice that are bar) at 3 months, compared with the other mutant mice. deficient in all ovarian activins. Each group is represented by Sample sizes for progesterone are: (3) n ϭ 7; (2) n ϭ 10; (1) three mice, and statistics were preformed by one-way n ϭ 11; (0) n ϭ 5. D, Estradiol values do not change between ANOVA followed by Dunnett’s post hoc test using the 2 ␤ genotypes. Sample sizes for estradiol are: (3) n ϭ 11; (2) n ϭ allele group as the control group. 11; (1) n ϭ 11; (0) n ϭ 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 bone morphogenetic protein (BMP) 15 (33) and inhib- not a,b). ited by GDF9 (34). To determine whether activin has any regulatory role on Kitl expression, Kitl transcript levels were measured in activin-treated wild-type depending on the type of activin allele. A possible expla- granulosa cells by qPCR. Recombinant activin A sig- nation for the more severe effects of activin ␤A defi- nificantly suppressed levels of Kitl in wild-type granu- ciency, is that activin ␤A has a greater bioactivity than losa cells by approximately 7-fold (P Ͻ 0.05) (Fig. 8D). activin ␤B, a hypothesis supported by our previous work These data are consistent with the increase in Kitl (23). In addition, it has been shown in a cell-free assay expression detected in activin-deficient granulosa system that activin A has a higher affinity for both types cells, and thus Kitl is a candidate gene for negative of activin type 2 receptors compared with activin B (ϳ8- regulation by activin in granulosa cells. fold 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 DISCUSSION with multiple oocytes (polyovular follicles; also called MOFs). The natural occurrence of polyovular follicles in Activins have multiple roles during folliculogenesis, in- mice have been reported since 1920 (36), and some cluding regulation of granulosa cell growth and differen- mouse strains are more susceptible to polyovular follicle tiation. Our ovarian activin knockout mice demonstrate formation than others (e.g. C58/J vs. C57L/J) (37). decreasing fertility with sequential deletion of the ␤-sub- Polyovular follicles are found in ovaries of some trans- units, culminating in sterility when all intraovarian activins genic mice, including inhibin-␣ overexpressing mice (38), are removed. As our laboratory previously observed (23), germ cell nuclear factor conditional knockout mice (39), activin dosage plays an important role in ovarian function Bmp15 knockout mice (40) and FSH receptor haploin- Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2465 Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021

Fig. 6. Markers of Healthy and Regressing CLs in Activin-Deficient Mice Ovaries from mature female mice of different genotypes. A–C, ␤Aflox/Ϫ mice; D–F, ␤BϪ/Ϫ;␤Aϩ/Ϫ mice; G–I, ␤BϪ/Ϫ; ␤Aflox/Ϫ; Amhr2cre/ϩ 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 mice (51). If growth induction is activin’s primary role at lines displaying this phenotype, it is likely that polyovular the preantral follicle stage, it would be predicted that loss follicle formation occurs through multiple mechanisms; of activin in granulosa cells could result in arrested gran- however, estrogen is one known mediator of polyovular ulosa cells growth, early differentiation of granulosa cells, follicle formation (42), likely caused by improper germ cell or retarded follicle development. Indeed, follicles dem- cyst breakdown (43–46). An additional study has shown onstrating luteinization of granulosa cells were apparent that neonatal estrogen exposure results in an increased in the ovaries of activin-deficient mice. However, our incidence of polyovular follicles in conjunction with sup- data did not demonstrate any changes in numbers of pression of the activin ␤-subunit (47). These data are primordial, primary, or secondary follicles in adult female consistent with polyovular formation in the ␤A condi- mice that lacked all ␤-subunits. Because activin is se- tional knockout, because Amhr2-driven expression of creted and its actions are paracrine in nature, variations cre recombinase is detectable in somatic cells of the in timing of recombination between loxP sites in granu- developing embryonic gonad (19, 20), and could result in losa cells from different follicles may variably change embryonic loss of the ␤A-subunit in these cells. Our data local activin concentrations, which may be insufficient to further suggest a direct role of activin in normal follicle cause complete intraovarian penetrance of the pheno- formation. type. Conditional mutations in activin receptors would Mice lacking both activin ␤A and ␤B are sterile, with a thus contribute valuable information regarding activin complex ovarian phenotype, perhaps in part because function in granulosa cells. activin has stage-specific functions during follicle devel- Deletion of the activin ␤-subunit genes in granulosa opment that also vary with age. In general, in vitro cul- cells affects both activin and inhibin production. The tures demonstrate that activin induces small follicle gonad is the predominant source of circulating inhibin growth and promotes granulosa cell proliferation alone in rodents (52), and our data demonstrate that deletion and in combination with FSH (48–50). However, activins of the ␤-subunit genes in granulosa cells decreases promote preantral follicle growth in isolated follicles de- serum inhibin to undetectable levels. Although previ- rived from immature mice but prevent FSH-stimulated ous in vitro studies show that recombinant activin growth in isolated cultures of preantral follicles from adult increases inhibin ␣-subunit expression in isolated 2466 Mol Endocrinol, October 2007, 21(10):2458–2471 Pangas et al.•In Vivo Roles of Activins in the Ovary Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021

Fig. 8. Gene Expression Changes in Activin-Treated Wild- Type 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 Ϯ SEM. Statistical significance by Stu- dent’s t tests (*, P Ͻ 0.05).

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 Fig. 7. qPCR Analysis of Activin-Deficient Granulosa Cells the activin-deficient females, although not likely all. Fe- Granulosa cell RNA was collected from three wild-type (wt) mice, 3 ␤Aflox/Ϫ;␤BϪ/Ϫ mice (1 ␤ alleles) and 3 ␤BϪ/Ϫ; ␤Aflox/Ϫ; male mice overexpressing FSH demonstrate hemor- Amhr2cre/ϩ (0 b alleles) mutant mice. qPCR results are ex- rhagic and cystic ovaries (55) but have a cystic pheno- pressed as relative quantity (RQ). A, Fshr trends toward lower type much more severe than the phenotype that expression in the mutants, but is not statistically significant presents in activin-deficient females. Interestingly, even by one-way ANOVA. B, Ccnd2 expression is unchanged be- though the antral follicle pool is expanded, superovula- tween genotypes. C, Lhcgr is significantly up-regulated in the tion of immature female mice does not result in an in- activin-deficient granulosa cells (0) vs. the wild-type sample. creased ovulation rate. In fact, the ovulation rate in Ctgf (D), Ghr (E), Myc (F), and Taf4b (G) are significantly PMSG/human chorionic gonadotropin (hCG) stimulated down-regulated in mutant granulosa cells (1, 0) vs. the wt activin-deficient females decreases approximately 60%. cells. H, Kitl is significantly increased in activin-deficient (0) Gene expression analysis of granulosa cells from the granulosa cells compared with wild type. Statistical significance by one-way ANOVA followed by Tukey-Kramer HSD preovulatory period indicates abnormal cell differentia- post hoc tests. Different letters above each column (a and b) tion. Granulosa cells collected from PMSG-stimulated indicate statistical significance at P Ͻ 0.05. 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). Inter- granulosa cells (53), this does not appear to be a estingly, the kit ligand-2 isoform is normally down-regu- necessary physiological role for activin in vivo because lated during early antral follicle formation (56), and ovar- there is an increase in Inha gene expression in activin- ian injection of antibody to the kit ligand receptor deficient ovaries. Because IGF-I is required for activin- disrupts antral follicle granulosa cell proliferation (57). By induced Inha expression (54), compensation by IGF-I in vitro experiments, we were able to show that activin is Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2467

capable of suppressing Kitl expression. During the antral The activin-deficient mouse model has higher levels of follicle period, activin may be partly responsible for the serum progesterone at 3 months of age. Activin signaling down-regulation of kit ligand, resulting in the transition requires the common SMAD transcription factor, SMAD4 from an FSH-driven growth phase, to a differentiation and in granulosa cell-specific knockouts of Smad4, proges- phase required for the preovulatory period. A role for terone levels are also increased (21). However, this is due to activin in regulating the terminal differentiation of ovarian premature luteinization of granulosa cells in preovulatory follicles has been proposed (58, 59). follicles. Unlike Smad4 cKO ovaries, activin-deficient pre- We found that several other genes are down-regulated ovulatory granulosa cells do not luteinize when immature in preovulatory activin-deficient granulosa cells, includ- mice are given supraphysiological injections of PMSG. And ing Ctgf, Ghr, Myc, and Taf4b. To date, only a few direct in contrast to the activin-deficient mouse model, luteinized

downstream targets of activin are known in granulosa follicles/structures do not accumulate in the Smad4 cKO Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 cells. Ctgf is regulated by activin and other TGF␤ family ovary, even though progesterone levels remain increased. members (31), but the consequence of loss-of-function Therefore, both models have higher levels of progesterone, of Ctgf in granulosa cells is unknown, and the contribu- although it appears to be through different mechanisms. tion to the activin phenotype is unclear. In activin-defi- Studies of human and macaque luteal cells find that activin cient granulosa cells from immature mice, Myc and inhibits progesterone synthesis (67, 68). Therefore, it is pos- Taf4b are expressed less than in wild-type granulosa sible that the increase in progesterone in the activin-defi- cells, and activin A up-regulates Taf4b in granulosa cells cient mice results from a lack of activin activity in CLs, in culture. These experiments thus identify Taf4b as po- which cannot be compensated for by another pathway. tential activin-response gene, although the activin path- BMPs and GDF9 are known to inhibit FSH-induced pro- way may not be the primary regulatory pathway because gesterone production and have been hypothesized to be the expression changes are small, although significant. luteinization inhibitors in granulosa cells of antral follicles Taf4b homozygous null female mice are infertile, with (69, 70). Our data from the activin-deficient mouse model defects in the inhibin/activin system and loss of expres- support this hypothesis and suggest that the BMP and sion of Inha, Inhba, Inhbb, and Ccnd2 (60). Thus, activin GDF9 pathways, and not the activin pathway, may be the may be part of a positive regulatory loop involving primary luteinization inhibitors in preovulatory follicles be- TAF4B. However, Inha is expressed in activin-deficient cause the BMP and GDF9 pathways are disrupted in ovaries as is Ccnd2. Differences in expression levels of Smad4 cKO, but not in the activin-deficient conditional Taf4b may explain the overall differences when compar- knockout. An alternative hypothesis is that during the pre- ing Taf4b knockout ovaries to activin-deficient ovaries ovulatory period, activin, GDF9 and/or BMP pathways may (i.e. reduced levels of Taf4b in the activin-deficient be redundant with respect to luteinization inhibition, with mouse, compared with compete absence of Taf4b ex- the BMP and/or GDF9 pathways compensating for loss of pression in Taf4b knockouts). activin in the activin-deficient mouse model and preventing The most striking phenotype of ␤B/␤A-deficient females premature luteinization. is the progressive accumulation of CLs. Our data indicate a The activin-deficient dKO ovary mouse model also high- defect during structural regression because 20-␤HSD im- lights the widespread, cell-specific effects of activins in munostaining is detectable in CLs of activin-deficient mice. various cell types. In other tissues, activin signaling com- Whereas no definitive data exist for activin in luteolysis, a ponents act as tumor suppressors; for example, loss of number of studies suggest that activin plays an active role activin receptors and SMAD signaling proteins, such as in regression of CLs. First, activin receptors localize to CLs during the development of some pancreatic cancers, re- in many species (61–63). And second, the activin antago- sults in tumorigenesis. No tumor development was seen in nist follistatin is highly expressed in newly formed rodent the activin-deficient mouse models. Thus, in different cel- CLs but undetectable during luteolysis (64), resulting in a lular contexts, which may include expression of different permissive environment for activin signaling. Whereas the transcription factors that interact with the downstream mechanism(s) by which activin could control structural re- SMAD signaling proteins, the activins and their signaling gression of CLs is unknown, a recent study using a novel in components can either function as tumor suppressors to vitro coculture system as a model for human luteolysis inhibit growth/stimulate differentiation, or as oncogenes to indicates that activin A from steroidogenic cells may directly stimulate growth/inhibit differentiation. Thus, both loss-of- induce matrix metalloproteinase 2 (Mmp2) expression function and gain-of-function mutations in activin signaling through paracrine signaling to fibroblasts (62). MMPs are pathway components may be observed to positively or critical for tissue remodeling, and increased expression and negatively regulate cancer initiation and/or progression. 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 ac- MATERIALS AND METHODS tivin-induced MMP2 expression is responsible for structural luteolysis remains to be tested in intact CLs, the activin- ES Cell Technology and Southern Blot Analysis deficient dKO mouse model should provide a unique system to test the roles of various hormones and growth fac- A 21.6-kb of isogenic DNA sequence encompassing the two- tors to examine CL physiology, and in particular luteal exon mouse activin ␤A gene sequence was isolated from a regression. 129S6/SvEv library and used to generate a conditional tar- 2468 Mol Endocrinol, October 2007, 21(10):2458–2471 Pangas et al.•In Vivo Roles of Activins in the Ovary

geting vector (␤Aflox). The targeting vector contains activin ␤A ginia Ligand Core Facility (Specialized Cooperative Centers intron 1 sequence, a loxP site, activin ␤A exon 2 propeptide Program in Reproduction Research NICHD/NIH U54 sequence, activin ␤A3Јuntranslated region, polyadenylation HD28934). Assay information is available at (http://ww- sequence, 3Ј sequences downstream of the activin ␤A gene, w.healthsystem.virginia.edu/internet/crr/ligand.cfm). The and a Pgk-Neo expression cassette flanked by loxP sites mouse FSH RIA has a sensitivity of 2.0 ng/ml, and an average inserted into the 3Ј downstream sequence. The floxed activin intraassay coefficient of variation (CV) of 10.1 and interassay ␤A construct was electroporated into the ES cells (␤A5-F12) CV of 13.3%. The mouse LH Sandwich IRMA has a sensitivity with a null mutation at the activin ␤A locus (␤Am1) (17) and of 0.07 ng/ml, and an average intraassay CV of 4.7% and clones isolated by positive-negative selection. The 23/35 ES interassay CV of 13.5%. The progesterone RIA has a sensi- cell clones (66%) targeted correctly at the activin ␤A locus. tivity of 0.10 ng/ml and an average intraassay CV of 4.5% and Two ES cell clones, ␤Aflox-B6 and ␤Aflox-C3 were injected into interassay CV of 6.9%. The estradiol RIA has a sensitivity of blastocysts as described (71) to produce chimeric mice. Male 10 pg/ml, and an intraassay CV of 5.33%, and an interassay chimeras were fertile and transmitted the ␤Aflox allele to F1 CV of 12.1%. The inhibin A ELISA has a sensitivity of assay progeny. Chimeras were mated to C57Bl6/J females to pro- 1 pg/ml and interassay variation of 6–7%. The percentage of Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 duce 129S6/SvEv/C57BL6/J hybrid mice. The phenotype of cross reactivity of the inhibin A assay with other ligands was mice generated from the two different lines was identical. the following: inhibin B 0.012, activin A 0.002, activin B 0.001, Southern blot analysis was used for genotyping. Southern and pro-␣ C 0.01. Values that fell below the assay threshold blot analysis of granulosa cells was performed with mural were given the threshold value. Serum samples were diluted granulosa cells isolated from large antral follicles of 3-wk-old in PBS to fall within the detectable range, when necessary. control and experimental mice treated with PMSG for 48 h National Hormone and Peptide Program (NHPP) made PRL before granulosa cell collection as described in (72). Geno- measurements by using a highly sensitive double-antibody typing analysis of the Amhr2-cre allele and EIIa-cre and R26R method with reagents provided by Dr. A. F. Parlow (NHPP, transgene was determined using PCR. The sequences of the Torrance, CA). Mouse PRL (AFP10777D) was used for iodi- primers for the Amhr2-cre genotyping were 5Ј-CGCATT- nation, rabbit antimouse PRL (AFP131078) was used as an- GTCTGAGTAGGTGT and 5Ј-GAAACGCAGCTCGGCCAGC. tiserum, and mouse PRL (AFP6476C) was used as reference The sequences of the primers for EIIa-cre genotyping were preparation. 5Ј-CCGGGCTGCCACGACCAA and 5Ј-GGCGCGGCAACA- CCATTTTT. Superovulation and Collection of Oocytes

Morphological, Histological, and Nineteen to 21-d-old and 6-month-old experimental and con- Immunohistochemical Analysis trol female mice were injected with PMSG (ip, 5 IU/mouse), and given hCG (ip, 5 IU/mouse) 48 h later. Mice were bred to All experimental animals were maintained in accordance with C57/129 hybrid known fertile males. The following morning the National Institutes of Health (NIH) Guide for the Care and oocytes were recovered from the ampulla of the oviduct in Use of Laboratory animals. Embedding and staining proce- M2 medium and counted. dures were performed by standard protocols at the Baylor College of Medicine Pathology Core Services laboratory. 20␣-HSD staining (1:300 dilution) was performed as de- Granulosa Cell Preparation scribed (21). Rabbit polyclonal anti-inhibin ␣ antibody was used at a 1:500 dilution and was a gift from W. Vale (The Salk For granulosa cell isolation, female mice 19–21 d of age were Institute, La Jolla, CA). TROMA-III (anti-cytokeratin 19) rat injected ip with 5 IU pregnant mare serum gonadotropins monoclonal antibody was used at 1:10 dilution and was (Calbiochem, La Jolla, CA) for 44–46 h to stimulate follicle obtained from the Developmental Studies Hybridoma Bank growth and granulosa cells collected by puncturing antral developed under the auspices of the NICHD and maintained follicles with fine-gauge needles (72). RNA was extracted by the University of Iowa, Department of Biological Sciences. from mutant and control granulosa cells for microarray ex- Immunohistochemistry was performed on at least four sam- periments and qPCR verification immediately upon harvest. ples in duplicate. For follicle histomorphometric analysis, a For treatment of wild-type cells, 19–21 d CD1 female were minimum of three ovaries was counted for each genotype. used for granulosa cell collections and cells plated at a den- Follicle classification was based on Pederson and Peters (73). sity of 5.5 ϫ 105 cells/ml as described (76). Cells were incu- For assessment of primordial, primary, and preantral follicles, bated overnight in DMEM-F12 medium supplemented with follicle counts were carried out as described (74, 75). Briefly, 2% fetal calf serum, 1ϫ insulin/transferring/selenite (Invitro- ovaries of each genotype were serially sectioned at 5 ␮m, gen Life Technologies, Carlsbad, CA), 10 U/ml penicillin and and every tenth section was kept. Follicles were counted streptomycin and 250 ng/ml recombinant follistatin-288 to from five of the largest sections, and normalized to the total block endogenous activin signaling. After 18 h, cells were area of the section. Counts and area were collected using the washed twice in PBS and incubated with medium only with AxioVision 4.0 software (Carl Zeiss, Jena, Germany). For as- 50 ng/ml recombinant human activin A (R&D Biosystems, sessment of antral follicles and CLs, the total number of antral Minneapolis, MN) in DMEM-F12 medium supplemented with follicles and CLs in each ovary were counted. To avoid dou- 0.5% heat inactivated fetal bovine serum, 1ϫ ITS, and 10 ble counting antral follicles, only follicles with a visible oocyte U/ml penicillin and streptomycin. Cells were treated for 5 h, nucleus were counted. For CL counts, images were taken of then harvested for RNA. A minimum of four independent each section (every 50 ␮m), ordered sequentially, and each replicates was performed. CL followed through the sections, comparing CLs with those on previous sections to avoid double counting. RNA Isolation and Northern Blot Analysis

Serum Analysis For Northern blot analysis, total ovarian RNA was isolated from individual mice using RNA STAT-60 reagent (Leedo Mice were anesthetized by isoflurane inhalation (Abbott Lab- Medical Laboratories, Houston, TX). Twelve micrograms of oratories, North Chicago, IL), blood recovered by closed each RNA sample were used for electrophoresis and trans- cardiac puncture, and serum separated by centrifugation in ferred to nylon membranes as described previously (14) and Microtainer tubes (Becton Dickinson, Franklin Lakes, NJ) and probed with radioactive cDNA probe of 426 bp against activin stored at Ϫ20 C. FSH, LH, estradiol, progesterone, and in- ␤A (1049–1474 of the clone X69619) generated using hibin A measurements were made by the University of Vir- [32P]dATP and the Strip-EZ kit (Ambion, Inc., Austin, TX). Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2469

Phosphorimaging plates were scanned and analyzed using (Microsoft). Statistical differences were tested using one-way Image Quant software (Molecular Dynamics, Inc., Sunnyvale, ANOVA for multiple comparisons, followed by Tukey-Kramer CA). A background level for each blot was determined and honestly significant difference (HSD) or Dunnett’s post hoc subtracted. Blots were stripped and reprobed for glyceralde- tests as indicated in the text. Two-tailed Student’s t test was hyde 3-phosphate dehydrogenase (Gapd) as a loading used for single comparisons at ␣ ϭ 0.05. Statistics were control. performed on no less than three independent experiments.

Microarray Analysis and qPCR Acknowledgments

RNA was isolated using the QIAGEN (Valencia, CA) RNeasy We thank Michal Klysik, Sankar Sridaran, Samuel Og- kit. RNA integrity, concentration, and quality were checked bonna, and Mujtaba Ali for genotyping, Dr. Kathleen Burns for by the Baylor College of Medicine Microarray Core Facility. ES cell injections, Dr. Heiner Westphal (National Institutes of

Labeling, hybridization, washing, scanning, and initial mi- Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 Health) for the EIIa-cre transgenic mice, Dr. Richard Beh- croarray analysis were performed by the Baylor College of ringer (M. D. Anderson Cancer Center) for the Amhr2-cre Medicine Microarray Core Facility using standard Affymetrix mice, Dr. Teresa Woodruff (Northwestern University) for the protocols and the Affymetrix Mouse Genome 430 2.0 Array. recombinant follistatin-288, and Dr. Geula Gibori (University Independent chips were analyzed from three independent Ϫ Ϫ of Illinois, Chicago, IL) for the 20␣-dihydroxyprogesterone wild-type and two independent activin-deficient (␤B / ; Ϫ ϩ antibody. We thank Dr. Alfred Parlow at National Hormone ␤Aflox/ ;Amhr2cre/ ) granulosa cell samples. Each sample and Peptide Program (Torrance, CA) for PRL measurements. 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 (Af- fymetrix) software to determine signal intensity and detection Received March 19, 2007. Accepted June 28, 2007. (absent, present, and marginal), and then data files were Address all correspondence and requests for reprints to: imported into GeneSpring GX for expression analysis. Data Martin M. Matzuk, M.D., Ph.D., The Stuart A. Wallace Chair were filtered based on 1) MAS 5.0 call of “present” in a and Professor, Department of Pathology, Baylor College of minimum of two samples; 2) a raw signal value of greater than Medicine, One Baylor Plaza, Houston, Texas 77030. E-mail: 300 in a minimum of one of two conditions; 3) normalized [email protected]. expression change of 2-fold or greater. This resulted in a list This research was supported by National Institutes of of 280 genes, which were then analyzed for statistical signif- Health (NIH) Grant HD32067 (to M.M.M.), National Institute of icance at ␣ ϭ 0.05 by one-way ANOVA; this should have Child Health and Human Development/NIH U54 HD28934 resulted in 14 genes changes by chance alone. The final (hormone analysis), and a Ruth L. Kirshstein National Re- differentially expressed gene list contained 111 genes: 81 up- search Service Award (5F32HD46335) and Burroughs Well- and 30 down-regulated. Genes lists of up-regulated and come Career Award in the Biomedical Sciences (to S.A.P.). down-regulated were imported into DAVID Bioinformatic software (http://david.abcc.ncifcrf.gov/home.jsp) for path- Data Deposition: Affymetrix microarray data files (five files) way analysis (77). The complete microarray data set will be have been deposited to the NIH Gene Expression Omnibus available at NCBI Gene Expression Omnibus (GEO) database under the series accession no. GSE7150 (http://www.ncbi.nlm.nih.gov/geo) with the accession no. Disclosure Statement: The authors have nothing to GSE7150. disclose.

qPCR

qPCR was performed as described (21). cDNA was prepared REFERENCES from 200 ng of total RNA using SuperScript III First Strand ␮ Synthesis Kit (Invitrogen) in a 50- l reaction. Real-time qPCR 1. Chang H, Brown CW, Matzuk MM 2002 Genetic analysis of was performed as previously described (21) using Applied the mammalian TGF-␤ superfamily. Endocr Rev 23: Biosystems (ABI; Foster City, CA) Prism 7500 Sequence De- 787–823 tection System, Taqman Master Mix and Gene Expression 2. Vale W, Rivier J, Vaughan J, McClintock R, Corrigan A, Woo Assays. Taqman (ABI) Assays used for qPCR: Ccnd2, W, Karr D, Spiess J 1986 Purification and characterization Mm00438071; Ctgf, Mm00439093; Fshr, Mm00442819; of an FSH releasing protein from porcine ovarian follicular Gapd, 4352339E; Ghr, Mm00439093; Inhbb, Mm03023992; fluid. Nature 321:776–779 Lhcgr, Mm00442931; Myc, Mm00487803. Primers to Kitl, 3. Vale W, Bilezikjian LM, Rivier C 1994 Reproductive and Taf4b and Gapd were designed using the Primer Expression other roles of inhibins. In: Knobil E, Neill J, eds. The phys- 2.0 Software (ABI) and amplified using Sybr Green Master iology of reproduction. 2nd ed. New York: Raven Press; Mix (ABI) with Gapd as a control. Primer sequences available 1861–1878 upon request. The Kitl primer set recognizes both splice 4. Matzuk MM 1995 Functional analysis of mammalian mem- variants, Kl-1 and Kl-2. Dissociation curves were generated bers of the transforming growth factor-␤ superfamily. for Sybr Green qPCRs to verify amplification of a single PCR ⌬⌬ Trends Endocrinol Metab 6:120–127 product. All qPCR data were analyzed by the cycle 5. Vassalli A, Matzuk MM, Gardner H, Lee K, Jaenisch R 1994 threshold method using the ABI 7500 System Software (ver- Activin/inhibin ␤B subunit gene disruption leads to defects sion 1.2.3) as described by ABI and normalized to the en- in eyelid development and female reproduction. Genes Dev dogenous reference (Gapd). One control sample was ran- 8:414–427 domly as the calibrator sample. The mean and SEM were 6. Schrewe H, Gendron-Maguire M, Harbison ML, Gridley T calculated, and the relative amount of target gene expression 1994 Mice homozygous for a null mutation of activin ␤B are for each sample compared with the control was plotted in viable and fertile. Mech Dev 47:43–51 Excel (Microsoft, Redmond, WA). 7. Cocolakis E, Lemay S, Ali S, Lebrun JJ 2001 The p38 MAPK pathway is required for cell growth inhibition of human Statistical Analysis breast cancer cells in response to activin. J Biol Chem 276:18430–18436 Statistical analysis was carried out using the JMP version 5.1 8. Zhang Z, Zhao Y, Batres Y, Lin MF, Ying SY 1997 Regula- (SAS Institute Inc., Cary, NC) statistical package or in Excel tion of growth and prostatic marker expression by activin A 2470 Mol Endocrinol, October 2007, 21(10):2458–2471 Pangas et al.•In Vivo Roles of Activins in the Ovary

in an androgen-sensitive prostate cancer cell line LNCAP. and cAMP in the induction of FSH receptor in cultured rat Biochem Biophys Res Commun 234:362–365 granulosa cells. J Endocrinol 147:103–110 9. Alexander JM, Bikkal HA, Zervas NT, Laws Jr ER, Klibanski 27. Minegishi T, Tano M, Nakamura K, Nakamura M, Igarashi S, A 1996 Tumor-specific expression and alternate splicing of Ito I, Shinozaki H, Karino S, Ibuki Y, Miyamoto K 1996 messenger ribonucleic acid encoding activin/transforming Regulation of follicle-stimulating hormone receptor. Horm growth factor-␤ receptors in human pituitary adenomas. Res 46:37–44 J Clin Endocrinol Metab 81:783–790 28. Ogawa T, Yogo K, Ishida N, Takeya T 2003 Synergistic 10. Su GH, Bansal R, Murphy KM, Montgomery E, Yeo CJ, effects of activin and FSH on hyperphosphorylation of Rb Hruban RH, Kern SE 2001 ACVR1B (ALK4, activin receptor and G1/S transition in rat primary granulosa cells. Mol Cell type 1B) gene mutations in pancreatic carcinoma. Proc Natl Endocrinol 210:31–38 Acad Sci USA 98:3254–3257 29. El-Hefnawy T, Zeleznik AJ 2001 Synergism between FSH 11. Hempen PM, Zhang L, Bansal RK, Iacobuzio-Donahue CA, and activin in the regulation of proliferating cell nuclear Murphy KM, Maitra A, Vogelstein B, Whitehead RH, antigen (PCNA) and cyclin D2 expression in rat granulosa Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 Markowitz SD, Willson JK, Yeo CJ, Hruban RH, Kern SE cells. Endocrinology 142:4357–4362 2003 Evidence of selection for clones having genetic inac- 30. Eppig JJ, Wigglesworth K, Pendola F, Hirao Y 1997 Murine tivation of the activin A type II receptor (ACVR2) gene in oocytes suppress expression of luteinizing hormone recep- gastrointestinal cancers. Cancer Res 63:994–999 tor messenger ribonucleic acid by granulosa cells. Biol Re- 12. Mori Y, Yin J, Rashid A, Leggett BA, Young J, Simms L, prod 56:976–984 Kuehl PM, Langenberg P, Meltzer SJ, Stine OC 2001 31. Harlow CR, Davidson L, Burns KH, Yan C, Matzuk MM, Instabilotyping: comprehensive identification of frameshift Hillier SG 2002 FSH and TGF-␤ superfamily members reg- mutations caused by coding region microsatellite instability. ulate granulosa cell connective tissue growth factor gene Cancer Res 61:6046–6049 expression in vitro and in vivo. Endocrinology 143: 13. Rossi MR, Ionov Y, Bakin AV, Cowell JK 2005 Truncating 3316–3325 mutations in the ACVR2 gene attenuates activin signaling in 32. Thomas FH, Vanderhyden BC 2006 Oocyte-granulosa cell prostate cancer cells. Cancer Genet Cytogenet 163: interactions during mouse follicular development: regula- 123–129 tion of kit ligand expression and its role in oocyte growth. 14. Elvin JA, Yan C, Wang P, Nishimori K, Matzuk MM 1999 Reprod Biol Endocrinol 4:19 Molecular characterization of the follicle defects in the 33. Otsuka F, Shimasaki S 2002 A negative feedback system growth differentiation factor-9-deficient ovary. Mol Endocri- between oocyte bone morphogenetic protein 15 and gran- nol 13:1018–1034 ulosa cell kit ligand: its role in regulating granulosa cell 15. Findlay JK, Drummond AE, Dyson ML, Baillie AJ, Robertson mitosis. Proc Natl Acad Sci USA 99:8060–8065 DM, Ethier JF 2002 Recruitment and development of the 34. Joyce IM, Clark AT, Pendola FL, Eppig JJ 2000 Comparison follicle; the roles of the transforming growth factor-␤ super- of recombinant growth differentiation factor-9 and oocyte family. Mol Cell Endocrinol 191:35–43 regulation of KIT ligand messenger ribonucleic acid expres- 16. Muttukrishna S, Tannetta D, Groome N, Sargent I 2004 sion in mouse ovarian follicles. Biol Reprod 63:1669–1675 Activin and follistatin in female reproduction. Mol Cell En- 35. del Re E, Sidis Y, Fabrizio DA, Lin HY, Schneyer A 2004 docrinol 225:45–56 Reconstitution and analysis of soluble inhibin and activin 17. Matzuk MM, Kumar TR, Vassalli A, Bickenbach JR, Roop receptor complexes in a cell-free system. J Biol Chem DR, Jaenisch R, Bradley A 1995 Functional analysis of 279:53126–53135 activins in mammalian development. Nature 374:354–356 36. Engle E 1927 Polyovular follicles and polynuclear ova in the 18. Lakso M, Pichel JG, Gorman JR, Sauer B, Okamoto Y, Lee mouse. Anat Rec 35:341–343 E, Alt FW, Westphal H 1996 Efficient in vivo manipulation of 37. Jagiello G, Ducayen M 1973 Meiosis of ova from polyovular mouse genomic sequences at the zygote stage. Proc Natl (C58-J) and polycystic (C57 L-J) strains of mice. Fertil Steril Acad Sci USA 93:5860–5865 24:10–14 19. Jamin SP, Arango NA, Mishina Y, Hanks MC, Behringer RR 38. McMullen ML, Cho BN, Yates CJ, Mayo KE 2001 Gonadal 2002 Requirement of Bmpr1a for Mullerian duct regression pathologies in transgenic mice expressing the rat inhibin during male sexual development. Nat Genet 7:7 ␣-subunit. Endocrinology 142:5005–5014 20. Jorgez CJ, Klysik M, Jamin SP, Behringer RR, Matzuk MM 39. Lan ZJ, Xu X, Cooney AJ 2003 Generation of a germ cell 2004 Granulosa cell-specific inactivation of follistatin nuclear factor conditional allele in mice. Genesis 37: causes female fertility defects. Mol Endocrinol 18:953–967 172–179 21. Pangas SA, Li X, Robertson EJ, Matzuk MM 2006 Prema- 40. Yan C, Wang P, DeMayo J, DeMayo F, Elvin J, Carino C, ture luteinization and cumulus cell defects in ovarian-spe- Prasad S, Skinner S, Dunbar B, Dube J, Celeste A, Matzuk cific Smad4 knockout mice. Mol Endocrinol 20:1406–1422 M 2001 Synergistic roles of bone morphogenetic protein 15 22. Boerboom D, Paquet M, Hsieh M, Liu J, Jamin SP, Be- and growth differentiation factor 9 in ovarian function. Mol hringer RR, Sirois J, Taketo MM, Richards JS 2005 Mis- Endocrinol 15:854–866 regulated Wnt/␤-catenin signaling leads to ovarian granu- 41. Danilovich N, Roy I, Sairam MR 2001 Ovarian pathology losa cell tumor development. Cancer Res 65:9206–9215 and high incidence of sex cord tumors in follitropin receptor 23. Brown CW, Houston-Hawkins DE, Woodruff TK, Matzuk knockout (FORKO) mice. Endocrinology 142:3673–3684 MM 2000 Insertion of Inhbb into the Inhba locus rescues the 42. Chen Y, Jefferson WN, Newbold RR, Padilla-Banks E, Inhba-null phenotype and reveals new activin functions. Nat Pepling ME 2007 Estradiol, progesterone and genistein in- Genet 25:453–457 hibit oocyte nest breakdown and primordial follicle assem- 24. Peng XR, Hsueh AJ, LaPolt PS, Bjersing L, Ny T 1991 bly in the neonatal mouse ovary in vitro and in vivo. Endo- Localization of luteinizing hormone receptor messenger ri- crinology 148:3580–3590 bonucleic acid expression in ovarian cell types during folli- 43. Iguchi T, Fukazawa Y, Uesugi Y, Takasugi N 1990 Polyovu- cle development and ovulation. Endocrinology 129: lar follicles in mouse ovaries exposed neonatally to dieth- 3200–3207 ylstilbestrol in vivo and in vitro. Biol Reprod 43:478–484 25. Segaloff DL, Wang HY, Richards JS 1990 Hormonal regu- 44. Iguchi T, Takasugi N 1986 Polyovular follicles in the ovary of lation of luteinizing hormone/chorionic gonadotropin recep- immature mice exposed prenatally to diethylstilbestrol. tor mRNA in rat ovarian cells during follicular development Anat Embryol (Berl) 175:53–55 and luteinization. Mol Endocrinol 4:1856–1865 45. Iguchi T, Takasugi N, Bern HA, Mills KT 1986 Frequent 26. Nakamura M, Nakamura K, Igarashi S, Tano M, Miyamoto occurrence of polyovular follicles in ovaries of mice ex- K, Ibuki Y, Minegishi T 1995 Interaction between activin A posed neonatally to diethylstilbestrol. Teratology 34:29–35 Pangas et al.•In Vivo Roles of Activins in the Ovary Mol Endocrinol, October 2007, 21(10):2458–2471 2471

46. Iguchi T, Todoroki R, Takasugi N, Petrow V 1988 The ef- activin receptor subtypes in rat brain, pituitary, ovary, and fects of an aromatase inhibitor and a 5␣-reductase inhibitor testis. Endocrinology 134:799–808 upon the occurrence of polyovular follicles, persistent 62. Myers M, Gay E, McNeilly AS, Fraser HM, Duncan WC 2007 anovulation, and permanent vaginal stratification in mice In vitro evidence suggests activin-A may promote tissue treated neonatally with testosterone. Biol Reprod 39: remodeling associated with human luteolysis. Endocrinol- 689–697 ogy 148:3730–3739 47. Kipp JL, Kilen SM, Bristol-Gould S, Woodruff TK, Mayo KE 63. van den Hurk R, Van de Pavert SA 2001 Localization of an 2007 Neonatal exposure to estrogens suppresses activin activin/activin receptor system in the porcine ovary. Mol expression and signaling in the mouse ovary. Endocrinol- Reprod Dev 60:463–471 ogy 148:1968–1976 64. Erickson GF, Shimasaki S 2003 The spatiotemporal expres- 48. Li R, Phillips DM, Mather JP 1995 Activin promotes ovarian sion pattern of the bone morphogenetic protein family in rat follicle development in vitro. Endocrinology 136:849–856 ovary cell types during the estrous cycle. Reprod Biol En- 49. Miro F, Hillier S 1996 Modulation of granulosa cell deoxyri- docrinol 1:9–28 Downloaded from https://academic.oup.com/mend/article/21/10/2458/2738443 by guest on 30 September 2021 bonucleic acid synthesis and differentiation by activin. En- 65. Duncan WC, McNeilly AS, Illingworth PJ 1998 The effect of docrinology 137:464–468 luteal “rescue” on the expression and localization of matrix 50. Zhao J, Taverne MA, van der Weijden GC, Bevers MM, van metalloproteinases and their tissue inhibitors in the human den Hurk R 2001 Effect of activin A on in vitro development corpus luteum. J Clin Endocrinol Metab 83:2470–2478 of rat preantral follicles and localization of activin A and 66. Young KA, Hennebold JD, Stouffer RL 2002 Dynamic ex- activin receptor II. Biol Reprod 65:967–977 pression of mRNAs and proteins for matrix metalloprotein- 51. Yokota H, Yamada K, Liu X, Kobayashi J, Abe Y, Mizunuma ases and their tissue inhibitors in the primate corpus luteum H, Ibuki Y 1997 Paradoxical action of activin A on folliculo- during the menstrual cycle. Mol Hum Reprod 8:833–840 genesis in immature and adult mice. Endocrinology 138: 67. Di Simone N, Lanzone A, Petraglia F, Ronsisvalle E, Caruso 4572–4576 A, Mancuso S 1994 Effect of activin-A on progesterone 52. Woodruff T, Besecke L, Groome N, Draper L, Schwartz N, synthesis in human luteal cells. Fertil Steril 62:1157–1161 Weiss J 1996 Inhibin A and inhibin B are inversely correlated 68. Brannian JD, Woodruff TK, Mather JP, Stouffer RL 1992 to follicle-stimulating hormone, yet are discordant during Activin-A inhibits progesterone production by macaque lu- the follicular phase of the rat estrous cycle, and inhibin A is teal cells in culture. J Clin Endocrinol Metab 75:756–761 expressed in a sexually dimorphic manner. Endocrinology 69. Shimasaki S, Zachow RJ, Li D, Kim H, Iemura S-I, Ueno N, 137:5463–5467 Sampath K, Chang RJ, Erickson GF 1999 A functional bone 53. LaPolt PS, Soto D, Su JG, Campen CA, Vaughan J, Vale W, morphogenetic protein system in the ovary. Proc Natl Acad Hsueh AJ 1989 Activin stimulation of inhibin secretion and Sci USA 96:7282–7287 messenger RNA levels in cultured granulosa cells. Mol En- 70. Vitt UA, Hayashi M, Klein C, Hsueh AJ 2000 Growth differ- docrinol 3:1666–1673 entiation factor-9 stimulates proliferation but suppresses 54. Kubo T, Shimasaki S, Kim H, Li D, Erickson GF 1998 Ac- tivin-induced inhibin ␣-subunit production by rat granulosa the follicle-stimulating hormone-induced differentiation of cells requires endogenous insulin-like growth factor-I. Biol cultured granulosa cells from small antral and preovulatory Reprod 58:712–718 rat follicles. Biol Reprod 62:370–377 55. Kumar TR, Palapattu G, Wang P, Woodruff TK, Boime I, 71. Matzuk M, Finegold M, Su J, Hsueh A, Bradley A 1992 ␣ Byrne MC, Matzuk MM 1999 Transgenic models to study -Inhibin is a tumor-suppressor gene with gonadal speci- gonadotropin function: the role of follicle-stimulating hor- ficity in mice. Nature 360:313–319 mone in gonadal growth and tumorigenesis. Mol Endocrinol 72. Elvin JA, Clark AT, Wang P, Wolfman NM, Matzuk MM 1999 13:851–865 Paracrine actions of growth differentiation factor-9 in the 56. Thomas FH, Ethier JF, Shimasaki S, Vanderhyden BC 2005 mammalian ovary. Mol Endocrinol 13:1035–1048 Follicle-stimulating hormone regulates oocyte growth by 73. Pedersen T, Peters H 1968 Proposal for a classification of modulation of expression of oocyte and granulosa cell fac- oocytes and follicles in the mouse ovary. J Reprod Fertil tors. Endocrinology 146:941–949 17:555–557 57. Yoshida H, Takakura N, Kataoka H, Kunisada T, Okamura 74. Hu YC, Wang PH, Yeh S, Wang RS, Xie C, Xu Q, Zhou X, H, Nishikawa S-I 1997 Stepwise requirement of c-kit ty- Chao HT, Tsai MY, Chang C 2004 Subfertility and defective rosine kinase in mouse ovarian follicle development. Dev folliculogenesis in female mice lacking androgen receptor. Biol 184:122–137 Proc Natl Acad Sci USA 101:11209–11214 58. Findlay JK 1993 An update on the roles of inhibin, activin, 75. Cheng G, Weihua Z, Makinen S, Makela S, Saji S, Warner and follistatin as local regulators of folliculogenesis. Biol M, Gustafsson JA, Hovatta O 2002 A role for the androgen Reprod 48:15–23 receptor in follicular atresia of estrogen receptor ␤ knockout 59. Hillier SG, Miro F 1993 Inhibin, activin, and follistatin. Po- mouse ovary. Biol Reprod 66:77–84 tential roles in ovarian physiology. Ann NY Acad Sci 687: 76. Pangas SA, Jorgez CJ, Matzuk MM 2004 Growth differen- 29–38 tiation factor 9 regulates expression of the bone morpho- 60. Freiman RN, Albright SR, Zheng S, Sha WC, Hammer RE, genetic protein antagonist, gremlin. J Biol Chem 279: Tjian R 2001 Requirement of tissue-selective TBP-associ- 32281–32286 ated factor TAFII105 in ovarian development. Science 293: 77. Dennis Jr G, Sherman BT, Hosack DA, Yang J, Gao W, 2084–2087 Lane HC, Lempicki RA 2003 DAVID: database for annota- 61. Cameron V, Nishimura E, Mathews L, Lewis K, Sawchenko tion, visualization, and integrated discovery. Genome Biol P, Vale W 1994 Hybridization histochemical localization of 4:P3

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