BIOLOGY OF REPRODUCTION 76, 1045–1053 (2007)Published online before print 28 February 2007.DOI 10.1095/biolreprod.106.059360

Role of Central Nervous System and Ovarian Insulin Receptor Substrate 2 Signaling inFemale Reproductive Function in the Mouse1

Irina Neganova,3 Hind Al-Qassab,3 Helen Heffron,3 Colin Selman,3 Agharul I. Choudhury,3

Steven J. Lingard,3 Ivan Diakonov,3 Michael Patterson,4 Mohammad Ghatei,4 Stephen R. Bloom,4

Stephen Franks,5 Ilpo Huhtaniemi,5 Kate Hardy,5 and Dominic J. Withers2,3

Centre for Diabetes and Endocrinology,3 Rayne Institute, University College London, London WC1E 6JJ, United KingdomDepartment of Metabolic Medicine,4 and Institute of Reproductive and Developmental Biology,5 Imperial CollegeLondon, London W12 0NN, United Kingdom

ABSTRACT

Insulin receptor signaling regulates female reproductivefunction acting in the central nervous system and ovary. Femalemice that globally lack insulin receptor substrate (IRS) 2, whichis a key mediator of insulin receptor action, are infertile withdefects in hypothalamic and ovarian functions. To unravel thetissue-specific roles of IRS2, we examined reproductive functionin female mice that lack Irs2 only in the neurons. Surprisingly,these animals had minimal defects in pituitary and ovarianhormone levels, ovarian anatomy and function, and breedingperformance, which indicates that the central nervous systemIRS2 is not an obligatory signaling component for the regulationof reproductive function. Therefore, we undertook a detailedanalysis of ovarian function in a novel Irs2 global null mouseline. Comparative morphometric analysis showed reducedfollicle size, increased numbers of atretic follicles, as well asimpaired oocyte growth and antral cavity development in Irs2null ovaries. Granulosa cell proliferation was also defective inthe Irs2 null ovaries. Furthermore, the insulin- and eCG-stimulated phosphoinositide-3-OH kinase signaling events,which included phosphorylation of Akt/protein kinase B andglycogen synthase kinase 3-beta, were impaired, whereasmitogen-activated protein kinase signaling was preserved inIrs2 null ovaries. These abnormalities were associated withreduced expression of cyclin D2 and increased CDKN1B levels,which indicates dysregulation of key components of the cellcycle apparatus implicated in ovarian function. Our data suggestthat ovarian rather than central nervous system IRS2 signaling isimportant in the regulation of female reproductive function.

insulin, insulin-like growth factor receptor, kinases, oocytedevelopment, ovary

INTRODUCTION

Insulin receptor (INSR) signaling pathways regulate periph-eral energy homeostasis by acting in skeletal muscle, adiposetissue, and the liver to control carbohydrate, lipid, and proteinmetabolism. It is clear that these signals also act in other

tissues, for example, in the regulation of pancreatic b-cell andhypothalamic functions [1]. Insulin receptor signaling has alsobeen implicated in the regulation of female reproductivefunction through its actions in both the central nervous system(CNS) and ovaries. Female mice that lack insulin receptors intheir neurons show impaired fertility with fewer antral folliclesand corpora lutea due to reduced release of LH [2]. Insulin alsostimulates GnRH secretion in vivo, and intracerebroventricularadministration of insulin restores reproductive behavior indiabetic rats [3, 4]. In vitro studies using immortalized GnRHneuronal cell lines have suggested that insulin regulates GnRHexpression through the mitogen-activated protein kinasepathway [5]. Therefore, CNS insulin signaling plays a role inregulating female reproductive function in rodents. Insulinsignaling has also been shown to have complex roles in ovarianfunction, including the regulation of ovarian steroidogenesis,follicular development, and granulosa cell proliferation [6–8].The strong association of insulin resistance with ovariandysfunction in polycystic ovarian syndrome also suggests arole for insulin signaling in ovarian function [9].

Insulin receptor substrate (IRS) proteins mediate the effectsof the insulin receptor on cellular and whole body physiology,including reproduction [10]. Mice that lack Irs1 displayprofound growth retardation and insulin resistance but havemildly defective reproductive function [11, 12]. Irs3 and Irs4null mice have minimal metabolic, endocrine, and growthphenotypes [10]. In contrast, mice that lack Irs2 developdiabetes due to a combination of insulin resistance andpancreatic b cell dysfunction [11]. Female Irs2 null mice arealso infertile due to reduced pituitary LH levels andgonadotroph cell numbers, and show reduced gonadotropin-stimulated ovulation and markedly reduced numbers of ovarianfollicles and corpora lutea [12]. While CNS IRS2 signalingplays complex roles in energy homeostasis [1], the contributionof CNS IRS2 to female reproductive function has not beenestablished. Ovarian cellular defects and downstream signalingabnormalities in Irs2 null mice have not been characterized.Therefore, we used conditional gene targeting in mice to deleteIrs2 in the CNS and examined their reproductive functions. Inaddition, we generated a novel Irs2 null mouse line andassessed its ovarian morphology, signaling, and cell cycleevents.

MATERIALS AND METHODS

Ethics

All of the in vivo studies were performed in accordance to the UnitedKingdom Home Office Animal Procedures Act (1986) and University CollegeLondon Animal Ethical Committee Guidelines.

1Supported by an MRC Co-operative Component Grant and a Well-come Trust Functional Genomics Award.2Correspondence: Dominic J. Withers, Centre for Diabetes andEndocrinology, Rayne Institute, 5 University Street, University CollegeLondon, London WC1E 6JJ, United Kingdom. FAX: 4420 767 96583;e-mail: d.withers@ucl.ac.uk

Received: 7 December 2006.First decision: 23 December 2006.Accepted: 19 February 2007.� 2007 by the Society for the Study of Reproduction, Inc.ISSN: 0006-3363. http://www.biolreprod.org

1045

Experimental Animals

Mice with a floxed allele of Irs2 (Irs2tm1With) and mice that lack Irs2 in theCNS, Irs2tm1With /Irs2tm1With Tg (Nes-cre)1Kln/0 (hereinafter designated asNesCreIrs2KO mice) have been described previously [1]. Mice with a germ-line deletion of Irs2 were also generated using embryonic stem cells that hadundergone complete deletion of the Irs2 locus following Cre recombinasetreatment during the same targeting strategy [1]. The mice were back-crossedfive times onto a C57Bl/6 background prior to analysis. Genotyping wasperformed as previously described [1]. The mice were maintained on a12L:12D cycle with free access to water and standard mouse chow (4% fat,RM1; Special Diet Services, Witham, Essex, UK) and housed in specific-pathogen-free barrier facilities.

In Vivo Analyses of Reproductive Function

Estrus cycle characteristics were determined by analysis of vaginal smearsfor cell type and morphology with Giemsa staining each day for three completecycles. Fertility was assessed by monitoring daily for 3 mo the numbers ofpregnancies, litters, and pups in continuous mating studies with confirmed malebreeders and 7–9-week-old females.

Hormone Stimulation, Sample Collection, and Hormoneand Peptide Measurements

For the superovulation studies, 8–10-wk-old mice received 10 IU eCG(Folligon; Intervet) by i.p. injection, followed 46 h later by 8 IU hCG (Intervet).Oocytes were flushed from the oviduct 14–16 h after the second injection. Forbromodeoxyuridine (BrdU) incorporation studies, the mice received an i.p.injection of 100 mg/kg BrdU (Roche) 2 h before tissue harvesting. Bloodsamples were collected by tail vein bleeding or by cardiac puncture ofterminally anesthetized mice at specific stages of the estrus cycle, and the serumwas either stored at �808C prior to analysis or allowed to coagulate at 48Covernight, centrifuged, and frozen at �208C. Pituitaries were harvested,homogenized in 1 M urea in PBS, and stored at�808C prior to analysis. Serumestrogen and progesterone were measured using immunofluorometric assays(Delfia; Perkin Elmer, Turku, Finland). RIAs for pituitary LH, FSH, andprolactin were performed as previously described [13, 14].

Histomorphometric Analyses and Immunohistochemistry

Ovaries were fixed in 10% buffered formalin, embedded in paraffin, and cutinto 5-lm sections. For histomorphometric analysis, sections were stained withhematoxylin-eosin. Follicles, which were classified according to diameter andnumber of granulosa cell layers, were counted in every tenth section, startingfrom the fifth section, through the entire ovary. Only those follicles with aclearly visible oocyte nucleus were counted, to avoid double counting. Atreticfollicles were defined according to the following morphological criteria: follicleand oocyte shapes, localization of the oocyte in the follicle, presence of adegenerated oocyte, altered zona pellucida, the presence of granulosa cellsinside the zona pellucida, the presence of more than three pyknotic nuclei in thegranulosa cells, and macrophage invasion. Follicle growth, follicle and oocytesize (diameter), and total follicular and antral cavity area were measured usingthe LUCIA software (Nikon UK, Kingston-upon-Thames, UK). For immuno-histochemistry, sections were deparaffinized and antigen retrieval wasperformed by microwave boiling of the sections in 10 mM citrate buffer (pH6.0) twice for 10 min each. Sections were incubated in 20% (v/v) normal goatserum in PBS with 4% (w/v) BSA for 30 min at room temperature, and thenwith the indicated primary antibodies overnight at 48C. Detection of boundantibodies was performed using the avidin-biotin complex method (VectastainABC Elite kit, Vector Laboratories). Sections were also counterstained withhematoxylin, dehydrated in a graded ethanol series, cleared in xylene, andmounted. The Zymed proliferation cell nuclear antigen (PCNA) and BrdUdetection kits were used for PCNA and BrdU staining, respectively.

Western Blotting

Mice (8–10 wk of age) were either superovulated as described above ortreated with insulin as described previously [15]. In brief, mice were anesthetizedand injected i.v. with saline or 5 U of soluble insulin, and the ovaries wereremoved and snap frozen 10 min later. Tissue preparation and lysis, Westernblotting, and quantification were performed as previously described [15]. Theblots were stripped and probed for actin when loading controls were required.

Antibodies

Affinity-purified rabbit anti-IRS1 and anti-IRS2 antibodies were fromUpstate Biotechnology (Dundee, UK). The anti-AKT, anti-pAKTSer473, anti-

glycogen synthase kinase (GSK)3B, anti-pGSK3A/BSer9/21, anti-mitogen-activated protein kinase (MAPK) and anti-pMAPKThr202/Tyr204 antibodies werefrom Cell Signaling Technology (Beverly, MA). Rabbit polyclonal antibodiesto cyclin D2 and CDKN1B (p27KIP1) and mouse anti-actin were from SantaCruz Biotechnology (Insight Biotechnology, Wembley, UK).

Real-Time PCR

Real-time PCR was performed as previously described using FAM/TAMRA primers (Applied Biosystems, Foster City, CA) [15]. The followingprimer sets were used: Irs1 Mm00439720_s1, Irs2 (MIRS 396412), and Hprt1Mm01545399_m1. The relative amounts of mRNA were calculated from aninternal standard curve following normalization to the Hprt1 mRNA levels.

Statistical Analysis

All statistical analyses were performed using the GraphPad Prism4 software(GraphPad Software, San Diego, CA), and paired and unpaired t-tests and two-way ANOVA with Bonferroni post-hoc tests were performed as appropriate. P, 0.05 was regarded as statistically significant.

RESULTS

Reproductive Phenotype of NesCreIrs2KO Mice

NesCreIrs2KO mice lack Irs2 expression in the CNS andhypothalamus [1], while their expression of pituitary andovarian Irs2 mRNA is equivalent to that in control animals,i.e., the relative mRNA levels expressed as percentages of thewild-type (WT) control are: Irs2 in NesCreIrs2KO pituitaries,98.65 6 12.5%; Irs2 in NesCreIrs2KO ovaries, 107.4 615.2%; n ¼ 12, P . 0.05. There were no changes in Irs1expression in the ovary, hypothalamus, and pituitary ofNesCreIrs2KO mice (data not shown). Therefore NesCreIr-s2KO mice represent a good model for determining thecontribution of CNS IRS2 pathways to female reproductivefunction. The NesCreIrs2KO mice exhibited an extended estruscycle due to a mildly prolonged diestrus phase (Table 1).Diestrus pituitary prolactin levels were reduced in NesCreIr-s2KO mice, whereas the serum estradiol and progesterone andpituitary FSH and LH levels were similar to those of WT mice(Table 1). No differences in the timing of ovulation and thenumbers of oocytes recovered were seen when NesCreIrs2KOmice and WT mice were superovulated (total oocytesrecovered per animal: WT, 23.6 6 2.3 vs. NesCreIrs2KO21.1 6 2.1, n ¼ 5, P . 0.05). Morphological studies showedthat follicles at all developmental stages and corpora lutea werepresent in these animals and in WT mice, which is consistentwith their normal responses to superovulation (Fig. 1, A and Band data not shown). In continuous mating studies, the averageduration between litters and the size and number of litters overthe study period were similar in NesCreIrs2KO and WT mice(Table 1). NesCreIrs2KO mice were able to suckle and reartheir progeny and there were no differences in the postpartumsurvival rates of these mice compared to WT females (data notshown). Therefore, somewhat surprisingly, deletion of Irs2 inthe CNS has a minimal effect on reproductive function in mice.

A previous analysis of ovarian function in Irs2 null micewas limited to the demonstration of reduced numbers ofovarian follicles and corpora lutea, defective ovarian responsesto superovulation, and reduced circulating gonadal steroidlevels [12]. There are currently limited options for ovarian-specific gene deletion throughout folliculogenesis. Therefore,we examined in detail the ovarian morphology and ovarian cellproliferation and we analyzed the signaling events and cellcycle components downstream of IRS2 in our novel Irs2 nullmouse on a C57Bl/6 background.

After superovulation, reduced numbers of oocytes wererecovered from 8–10-week-old Irs2 null mice (total oocytes

1046 NEGANOVA ET AL.

recovered per animal: WT, 23.6 6 2.3 vs. Irs2 null 8.3 6 0.8,n¼ 5, P , 0.01). Irs2 null ovaries showed increased numbersof atretic follicles, reduced numbers of large antral follicles,and an absence of preovulatory follicles (Fig. 1, C and D).Additional characteristic features of Irs2 null ovaries were thepresence of oocytes trapped in corpora lutea (Fig. 1E) and thepresence of occasional follicles with double oocytes (Fig. 1F).We explored further the mechanisms underlying these defects.

IRS2 and IRS1 Protein Expression by Ovarian Cells

Strong specific IRS2 staining was seen in oocytes andgranulosa cell cytoplasm of the early primordial follicles in theWT mice but not in the Irs2 null animals (Fig. 2, A–D). Less-extensive IRS2 staining was detected in the thecal and stromalcells at all stages of follicular development in the WT mice(Fig. 2, A–C). IRS1 was expressed in a range of ovarian celltypes in the WT mice but not in the Irs1 null mice (Fig. 2, E–H). Staining was more intense in the granulosa cells than inthecal cells and increased with follicular growth. In the Irs2null mice, there was an apparent reduction in the expression ofIRS1 protein in the ovary (Fig. 2, I–K). This was confirmed bythe demonstration of reduced Irs1 mRNA expression in Irs2null ovaries (relative Irs1 mRNA levels expressed aspercentage of the WT control: 73 6 2%, n ¼ 3, P , 0.001).These findings suggest that IRS2 plays complex roles inovarian cellular function. Therefore, we characterized in detailthe follicular morphologies of the WT and Irs2 null mice.

Analysis of Follicular Development in Irs2 Null Ovaries

Morphometric analysis confirmed a marked reduction in thenumber of follicles of all sizes in Irs2 null ovaries and that 75%of the large antral follicles were atretic in Irs2 null mice (Fig. 3,A and B). When there were fewer than three folliculargranulosa cell layers, no differences were observed between thedevelopment of these follicles in Irs2 null and WT mice (Fig.3C). However, the follicles at later stages were reduced in size(Fig. 3C), which suggests that Irs2 null ovaries are defectivefor granulosa cell proliferation within growing follicles.Complex interactions occur between oocytes and otherfollicular components, with oocytes promoting follicular

growth and directing granulosa cell differentiation [16–19].In Irs2 null mice, the early oocytes grew as quickly as those inthe WT mice but growth was impaired at later stages (Fig. 3D).Antral cavity development and expansion were also markedlyattenuated in Irs2 null ovaries (Fig. 3E). After superovulationin the Irs2 null mice there was a minimal increase in antralcavity size (data not shown). These morphological parametersdemonstrate that in Irs2 null follicles, three events that areimportant for normal follicular development from the growingpool to the preovulatory stage are markedly disturbed.

Granulosa Cell Proliferation in Irs2 Null Follicles

We used PCNA expression and BrdU incorporation toassess further cell proliferation. In the WT ovaries, nuclearPCNA staining was seen in cuboidal granulosa cells in theprimary follicles and in most of the granulosa cells of folliclesat later stages (Fig. 4, A–D). Thecal cell staining was seen asthe follicles developed granulosa cell layers. Oocytes werePCNA positive in the primary follicles onwards (Fig. 4A). Inantral follicles, PCNA was expressed in all the granulosa celllayers, with the highest levels seen in those around the oocyte(Fig. 4C). Superovulation induced preovulatory follicles with

TABLE 1. Reproductive characteristics of NesCreIrs2KO mice.

Characteristic WTb NesCreIrs2KOb,c

Estrous cycle (n ¼ 9)a

Proestrus (days) 1.2 6 0.1 1.5 6 0.2Estrus (days) 1.3 6 0.2 1.3 6 0.2Metestrus (days) 1.2 6 0.1 1.4 6 0.2Diestrus (days) 1.2 6 0.1 1.9 6 0.2*Total duration (days) 4.9 6 0.1 6.1 6 0.2*

Hormone measurement (n ¼ 7–13)a

Estradiol (nmol/L) 0.016 6 0.002 0.017 6 0.04Progesterone (nmol/L) 4.198 6 0.7 3.43 6 1.0LH (ng/ml) 300.6 6 37.3 218.5 6 34.3Prolactin (ng/ml) 4048 6 840.7 979.5 6 126.9**FSH (ng/ml) 46.07 6 2.8 36.63 6 4.2

Breeding performance (n ¼ 6)a

Litters 23 19Pups 172 131Pups per litter 7.5 6 0.7 7.2 6 0.7Litter per female 3.8 6 0.2 3.2 6 0.3Durations between litters (days) 26.1 6 1.9 28.7 6 2.7

a n ¼Number of mice per genotype.b All results are presented as means 6 SEM.c Values are significantly different; *P , 0.05, **P , 0.01.

FIG. 1. Ovarian anatomies of WT, NesCreIrs2KO, and Irs2 null mice. A–C) Low magnification photomicrographs of typical WT (A), NesCreIrs2KO(B), and Irs2 null (C) ovarian morphologies. In C, the arrows indicateatretic follicles. D–F) High magnification of photomicrographs of Irs2 nullovaries demonstrating atretic follicles (D), oocytes trapped in corporalutea (E; arrow), and a follicle that contains a double oocyte (F). Bar¼ 0.5mm (A-C) or 100 lm (D–F). The brown staining is peroxidase stainingfrom the immunohistochemistry shown in detail in subsequent Figures.

IRS2 REGULATES OVARIAN FUNCTION 1047

strong cumulus PCNA staining (Fig. 4D). In contrast, Irs2 nullovaries displayed a marked reduction in granulosa cell PCNAstaining in all follicular classes (Fig. 4, E–H). Aftersuperovulation, there was little PCNA staining in the Irs2 nullovaries, which suggests that the vast majority of granulosa cellsin Irs2 null ovaries do not proceed to a rapid phase ofproliferation (Fig. 4H).

To assess S phase progression, we analyzed BrdUincorporation in the ovaries. In small follicles with two orfewer granulosa cell layers, there were no differences in thenumbers of BrdU positive cells in the WT and Irs2 nullovaries. In the WT mice, increased follicle size was associatedwith a marked increase in BrdU labeling (Fig. 4I). In the largeantral follicles, there was a general reduction in BrdU stainingbut the cumulus cells showed a high frequency of positivestaining (Fig. 4J). BrdU labeling was reduced in Irs2 nullfollicles of all sizes (Fig. 4, K and L). Superovulationincreased BrdU labeling in Irs2 null follicles but to asignificantly lesser extent to that seen in WT ovaries(percentage of BrdU-positive cells in secondary follicles(WT vs. Irs2 null): without stimulation, 13.8 6 2.3 vs. 3.6 60.9, n¼ 6, P , 0.05: after superovulation, 19.3 6 3.4 vs. 1.76 1.2, n¼ 6, P , 0.05; and percentage of BrdU-positive cellsin the large antral follicles: without stimulation, 26.9 6 2.1vs. 9.6 6 0.3, n ¼ 6, P , 0.05; after superovulation, 33.1 63.3 vs. 17.8 6 2.8, n ¼ 6, P , 0.05).

Defective PI3 Kinase Signaling Pathways and AbnormalExpression of Cell Cycle Regulators Cyclin D2 andCDKN1B in Irs2 Null Ovaries

IRS2 recruits the PI3 kinase/AKT and MAP kinasepathways, which together mediate the effects of insulin andIGF1 [10]. These downstream signaling events were notexamined in the Irs2 null mice described previously [12].Insulin induced a 5-fold increase in Akt phosphorylation in WTovaries, which was indicative of PI3 kinase pathway activation,although this was largely abrogated in Irs2 null ovaries (Fig. 5,A and B). Treatment with hCG stimulated AKT phosphory-lation in the ovaries of mice of either genotype but both thebasal and peak-stimulated AKT phosphorylation were attenu-ated in Irs2 null ovaries, despite similar levels of AKTexpression (Fig. 5, C and D). AKT phosphorylates andinactivates glycogen synthase kinase-3 b (GSK3B) [20, 21],which in turn phosphorylates cyclin D, thereby targeting it fordegradation [22]. Insulin increased GSK3B phosphorylation inthe WT ovaries but not in the Irs2 null ovaries (Fig. 5, E andF). In contrast, insulin-stimulated MAP kinase activation wasequivalent in the ovaries of both genotypes and hCG-stimulated MAP kinase activation was also largely preserved(Fig. 5, G and H and data not shown).

We also examined the expression levels of cyclin D2, whichis a critical component of the cell cycle apparatus that regulatesG1/S phase progression, and the cell cycle inhibitor CDKN1B.

FIG. 2. Localization of IRS2 and IRS1 inovarian follicles. A–C) IRS2 immunostainingof WT ovaries in primordial and primaryfollicles (A), secondary follicles (B), andgranulosa cells (C) of early antral follicles.D) IRS2 staining is absent from the Irs2 nullovaries. E–G) IRS1 immunostaining of WTovaries in primordial and primary follicles(E), secondary follicle and thecal cells (F),and in a large preantral follicle (G). H) Lackof IRS1 staining in an Irs1 null ovary. I–K)IRS1 staining of Irs2 null ovary demonstratesdecreased expression of IRS1 in the primary(I), secondary (J), and large preantralfollicles (K). L) There is no staining when theIRS1 antibody is omitted in Irs2 null ovarystaining. Bar ¼ 100 lm.

1048 NEGANOVA ET AL.

The basal expression of cyclin D2 was reduced and that ofCDKN1B was elevated in Irs2 null ovaries compared to WTovaries, which implies an imbalance between these cell cycleregulators (Fig. 5, I–L). Superovulation induction with eCGfollowed by hCG injection caused a significant reduction incyclin D2 expression 24 h after hCG treatment (Fig. 5, I and J).This effect was significantly impaired in Irs2 null ovaries. Thesame treatment induced a significant increase in CDKN1Bexpression in the WT ovaries but no significant alteration inIrs2 null ovaries, which also demonstrated a higher basal levelof CDKN1B (Fig. 5, K and L).

Primary follicles from mice of both genotypes showed nodifferences in cyclin D2 immunostaining, with a strong patternof staining seen in most granulosa cells (Fig. 6, A and F). In thesubsequent stages of follicular growth, both the numbers ofgranulosa cells that expressed cyclin D2 and the intensity ofstaining were reduced in Irs2 null compared to WT ovaries(Fig. 6, B–E and G–J). In WT ovaries, 46 h after theadministration of eCG alone, most of the stained cells weregranulosa cells surrounding the oocyte (Fig. 6C). In largepreovulatory follicles (with or without prior superovulationinduction) most of the stained granulosa cell were cumuluscells, whereas the cells in the mural granulosa were mostlyunstained (Fig. 6, D and E). These results are consistent withthe patterns of PCNA staining and BrdU incorporationdescribed above. In contrast, in the Irs2 null ovaries in whichlarge antral follicles were found, few cells expressed cyclin D2(Fig. 6, I and J), which suggests that impaired proliferation ofcumulus cells contributes to defective ovulation in these mice.

High levels of CDKN1B expression were found in thecorpora lutea of the WT mice, indicating that these cells wereno longer dividing (Fig. 6K). In contrast, most of the granulosacells from growing small and antral follicles did not expressCDKN1B (Fig. 6, K–O). In the WT animals, eCG stimulationdid not induce the expression of CDKN1B in granulosa andcumulus cells (Fig. 6M). However, in Irs2 null ovaries, somepositive staining was observed from the stage of small antralfollicles (Fig. 6Q). Moreover, 46 h after eCG stimulation,CDKN1B staining was observed in cumulus cells and in muraland thecal cells in the large antral follicles (Fig. 6R). Aftersuperovulation of the WT mice, no CDKN1B staining was seenin the cumulus cells and the majority of the mural granulosacells (Fig. 6, N and O). This is consistent with the high level ofcyclin D2 expression seen 8.5 h after hCG treatment in thecumulus cells in the preovulatory follicles of the WT mice (Fig.6, D and E). In contrast, in Irs2 null ovaries, the granulosa cellsshowed positive staining for CDKN1B 46 h after eCG injection(Fig. 6R). After administration of hCG, positive staining forCDKN1B was seen in the cumulus cells, suggesting thatproliferation was arrested in these cells (Fig. 6, S and T).

DISCUSSION

Insulin signaling pathways in the CNS have been implicatedin the regulation of female reproductive function [2, 10].Insulin receptor substrate proteins 1 and 2 are the majordownstream effector molecules recruited by the activatedinsulin receptor [10]. Female Irs2 null mice are infertile, with

FIG. 3. Comparative analysis of the ovarian morphologies of WT and Irs2null mice. A) Distribution of follicles according to diameter (lm) and total

3

number of atretic follicles in WT and Irs2 null ovaries. B) Proportion ofnormal and atretic follicles of different diameters (lm) in WT and Irs2 nullfollicles. C) Effect of Irs2 deletion on ovarian follicle growth. Each barrepresents the mean 6 SEM (n¼ 6 ovaries per genotype); *, P , 0.05. D)Effect of Irs2 deletion on oocyte growth. E) Effect of Irs2 deletion on antralcavity development during follicular growth. Values represent means 6 SEMand represent data from 194 WT and 173 Irs2 null follicles. *, P , 0.05.

IRS2 REGULATES OVARIAN FUNCTION 1049

defects in hypothalamic and ovarian functions, which suggestthat IRS2 mediates the effects of the insulin receptor onreproductive function in a number of tissues [12]. Furthermore,since Irs2 null mice display defects in leptin action, IRS2 mayintegrate the actions of insulin and leptin on energyhomeostasis and fertility [12]. Surprisingly, NesCreIrs2KOmice, which lack Irs2 in all their neurons, displayed minimalreproductive abnormalities with almost normal hormone levels,ovarian functions, and fertility rates. There are several possibleexplanations for these findings. IRS1 may compensate in theCNS of NesCreIrs2KO mice, although we detected noalterations in the expression of Irs1 in the hypothalami ofthese animals. The timing of deletion of Irs2 is different in Irs2null and NesCreIrs2KO mice, as the nestin transgene isexpressed from E9 onwards, whereas Irs2 expression neveroccurs in the global null animal. Therefore subtle develop-mental differences may explain the phenotype. Taken together,these results and the demonstration of marked hyperphagia andobesity in NesCreIrs2KO mice suggest that CNS Irs2expression is required for the effects of insulin on energybalance but not on reproductive function. Furthermore, wehave demonstrated that leptin action in NesCreIrs2KO mice isnormal, which suggests that IRS2 is not an obligatory point ofconvergence for the actions of insulin and leptin [1].

Irs2 was expressed in a number of ovarian cell types, withhigh levels in the follicular granulosa cells, which suggests thatdefective insulin and IGF1 actions in these cells may accountlargely for the infertility phenotype of Irs2 null mice. Analysesof follicular development and growth demonstrated a series ofinterrelated abnormalities in these animals. While earlyfollicular development and oocyte development were notdifferent between the WT and Irs2 null mice, once the follicleshad reached the stage in which growth became gonadotropin-dependent there was a marked decrease in their numbers. Therewas a significant reduction in the number and size of antralfollicles due to reduced antral cavity development andgranulosa cell proliferation and an increase in follicular atresia.Normally, in primordial follicles, the oocyte is surrounded by asingle layer of nondividing cells that are arrested in G0; thesecells can be induced by appropriate stimuli to undergoproliferation. Our studies demonstrate that in primordial andsmall follicles, cell proliferation is largely preserved in Irs2 nullovaries. However, the rapid phase of follicular growth thatoccurs when granulosa cells acquire enhanced FSH and LHresponsiveness and produce estrogen was abrogated in Irs2 nullovaries. Terminal differentiation (luteinization), which occurswhen the LH surge causes granulosa cells in preovulatoryfollicles to exit the cell cycle, was also abnormal in the Irs2 nullovaries. Taken together, these findings demonstrate disruptionof the co-ordinated growth pattern within follicles and loss of

3

FIG. 4. Expression of PCNA and BrdU incorporation in WT and Irs2 nullfollicles. A–D) WT ovarian PCNA expression in primordial (thick arrow)and primary follicles (indicated by a short arrow in thecal cell and by along arrow in granulosa cells) (A), small transitional follicle granulosa andthecal cells (B), granulosa and thecal cells in a large antral follicle (C), andin cumulus cells (D) surrounding an oocyte 8.5 h after superovulationinduction. E–H) PCNA expression in Irs2 null ovaries in a primary follicle(E), small transitional follicle (F), granulosa cells of a large follicle (G), andlow-level expression in granulosa cells (H) surrounding an oocyte 8.5 hafter hCG treatment. I, J) WT ovarian BrdU incorporation in a transitionalfollicle (I) and in cumulus cells of a preovulatory follicle aftersuperovulation (J). K, L) Irs2 null ovarian BrdU incorporation in atransitional follicle (K) and after superovulation induction (L). Sectionsshown are representative of ovaries from six animals of each genotype anddemonstrate typical patterns of PCNA and BrdU staining. Bar ¼ 100 lm.

1050 NEGANOVA ET AL.

the bidirectional communication between oocytes and gran-ulosa cells during follicle development.

The defects seen in our novel Irs2 null mouse model were notas severe as those described in the original knockout mouse.Differences in genetic background are likely to account for thisfinding, with the current model using mice on a C57Bl/6background, while the earlier studies were performed in a mixed129Sv/C57Bl/6 strain. Consistent with this observation are thefindings that the metabolic phenotype of Irs2 null mice isstrongly influenced by genetic background, and that the fecundityof C57Bl/6 mice is greater than that of 129Sv mice [23].

IRS2 lies downstream of both INSR and IGF1R and is able torecruit the MAP kinase and PI3 kinase signaling cascades,which have been implicated in ovarian function [6–8, 10].However, we found that MAP kinase expression and phos-phorylation in response to insulin or hCG were largelypreserved in Irs2 null ovaries. Other signaling components,such as IRS1 and SHC, may be involved in the recruitment andactivation of the MAP kinase in Irs2 null ovaries, althoughnormal MAP kinase activity in response to these hormones isinsufficient to preserve follicular function in these mice. Incontrast, we found significant signaling defects in the PI3 kinasecascade in Irs2 null ovaries, with both insulin and hCGstimulation of AKT being impaired. Previous studies haveimplicated PI3 kinase/AKT signaling in the maintenance of thepreovulatory follicle granulosa layers and have demonstrated

that FSH induces a biphasic increase in AKT phosphorylation inrat granulosa cells [24]. We have shown that the metaboliceffects of FSH in the mouse cumulus-oocyte complexes aremediated by the PI3 kinase pathway [25]. However, insulinappears to stimulate a larger increase in AKT serine phosphor-ylation than LH/hCG in the rat ovary in vivo [26]; our data frommouse ovaries are consistent with this observation.

A number of substrates of activated AKT that regulate cellsurvival and proliferation have been identified, includingcomponents of the apoptotic machinery, such as caspase 9and BAD, which may be involved in IGF1-mediated protectionof granulosa cell apoptosis [27, 28]. Members of the FOXOforkhead transcription factor subfamily (e.g., FOXO1a andFOXO3a) are also important substrates of AKT. Mice that lackFoxo3a display follicular activation and depletion of oocyteswith secondary infertility, which suggests an important role forthis protein in ovarian function [29]. AKT also phosphorylatesGSK3B, which may be involved in the regulation of ovariangene expression by FSH [21, 30]. Therefore, the major roles ofFOXO and GSK3B as downstream effectors of AKT signalinginvolve the co-ordination of cell proliferation and differentia-tion in the ovary, and defects in these events in Irs2 null ovariesare likely to account for many of the observed abnormalities.The AKT/forkhead signaling cassette has been demonstrated toregulate directly cyclin D2 function in rat granulosa cells, andGSK3B has also been shown to regulate D-type cyclin function

FIG. 5. Phosphorylation of AKT, GSK3B, and MAP kinase and expression of cyclin D2 and CDKN1B in WT and Irs2 ovaries. A) Western blot of phospho-AKT and total AKT in insulin-stimulated ovarian lysates from WT and Irs2 null mice. B) Densitometric quantification of A. C) Western blot of phospho-AKTand total AKT in hCG-stimulated ovarian lysates from WT and Irs2 null mice. D) Densitometric quantification of C. E) Western blot of phospho-GSK3B andtotal GSK3B in insulin-stimulated ovarian lysates from WT and Irs2 null mice. F) Densitometric quantification of E. G) Western blot of phospho-MAPkinase and total MAP kinase in insulin-stimulated ovarian lysates from WT and Irs2 null mice. H) Densitometric quantification of G. I and K) Western blotsfor cyclin D2 (I) and CDKN1B (K) in ovarian lysates from basal and hCG-stimulated WT and Irs2 null mice. Actin loading controls are also presented. J andL) Densitometric quantification of I and K, respectively. Representative blots are shown and quantification represents the mean 6 SEM of results obtainedin three independent experiments. *, P , 0.05.

IRS2 REGULATES OVARIAN FUNCTION 1051

FIG. 6. Cyclin D2 and CDKN1B immunolocalization in ovaries from WT and Irs2 null mice. A–E) Immunostaining for cyclin D2 in WT ovaries.Immunostaining for cyclin D2 is present in the primary and secondary follicles (A), in granulosa cells of transitional follicles (B), in the cumulus and muralgranulosa cells (C) 46 h after eCG treatment, and 8.5 h after superovulation induction in cumulus cells and mural granulosa cells (D and E). F–J)Immunostaining for cyclin D2 in Irs2 null ovaries. Immunostaining for cyclin D2 is detected in the primary and secondary follicles (F), in transitionalfollicles (G), 46 h after eCG in a small antral follicle (H), and after superovulation induction (I and J). K–O) Immunostaining for CDKN1B in WT ovaries.CDKN1B immunostaining is absent in the secondary follicle and present in the corpus luteum (CL) (K), in small growing transitional follicles (L), in largeantral follicles 46 h after eCG (M), and after superovulation induction (N and O). P–T) Immunostaining for CDKN1B of Irs2 null ovaries. Immunostainingfor CDKN1B is detected in the secondary and small transitional follicles (P), in a small antral follicle (Q), 46 h after eCG treatment (R), and 8.5 h aftersuperovulation induction (S and T). Bar ¼ 100 lm.

1052 NEGANOVA ET AL.

[31]. AKT/forkhead signaling also regulates CDKN1B func-tion through both transcriptional and phosphorylation events.Deletion of cyclin D2 in mice results in impaired granulosa cellproliferation, impaired follicular growth, trapped oocytes in thecorpora lutea, and ovulatory failure [32, 33]. Mice that lackCDKN1B have normal follicular growth but have impairedluteinization of granulosa cells [34]. In Irs2 null ovaries, wehave demonstrated defects in the regulation of both cyclin D2and CDKN1B due to impaired upstream AKT signaling, whichmay explain the functional abnormalities of ovarian cellularfunction in these mice. Reduced AKT activity may lead to theearlier expression of CDKN1B seen in the granulosa andcumulus cells of Irs2 null follicles. Disordered co-ordination offollicular and oocyte functions may also account for thefrequent observation of trapped oocytes in Irs2 null follicles.

In summary, we have shown that CNS IRS2 signalingpathways are not required for female reproductive function. Incontrast, signaling events downstream of IRS2 in the ovaryplay critical roles in follicular development and ovulation byregulating key components of the cell cycle apparatus involvedin the co-ordination of cell proliferation and differentiation.

ACKNOWLEDGMENTS

We thank P. Smith and A. Parlow of the National Hormone and PituitaryProgram and Pituitary Hormones and Antisera Centre for reagents.

REFERENCES

1. Choudhury AI, Heffron H, Smith MA, Al-Qassab H, Xu AW, Selman C,Simmgen M, Clements M, Claret M, Maccoll G, Bedford DC, HisadomeK, et al. The role of insulin receptor substrate 2 in hypothalamic and betacell function. J Clin Invest 2005; 115:940–950.

2. Bruning JC, Gautam D, Burks DJ, Gillette J, Schubert M, Orban PC, KleinR, Krone W, Muller-Wieland D, Kahn CR. Role of brain insulin receptor incontrol of body weight and reproduction. Science 2000; 289:2122–2125.

3. Burcelin R, Thorens B, Glauser M, Gaillard RC, Pralong FP.Gonadotropin-releasing hormone secretion from hypothalamic neurons:stimulation by insulin and potentiation by leptin. Endocrinology 2003;144:4484–4491.

4. Kovacs P, Morales JC, Karkanias GB. Central insulin administrationmaintains reproductive behavior in diabetic female rats. Neuroendocri-nology 2003; 78:90–95.

5. Salvi R, Castillo E, Voirol MJ, Glauser M, Rey JP, Gaillard RC,Vollenweider P, Pralong FP. Gonadotropin-releasing hormone-expressingneurons immortalized conditionally are activated by insulin: implication ofthe mitogen-activated protein kinase pathway. Endocrinology 2006; 147:816–826.

6. Poretsky L, Cataldo NA, Rosenwaks Z, Giudice LC. The insulin-relatedovarian regulatory system in health and disease. Endocr Rev 1999; 20:535–582.

7. Willis D, Mason H, Gilling-Smith C, Franks S. Modulation by insulin offollicle-stimulating hormone and luteinizing hormone actions in humangranulosa cells of normal and polycystic ovaries. J Clin Endocrinol Metab1996; 81:302–309.

8. Adashi EY, Resnick CE, Payne DW, Rosenfeld RG, Matsumoto T, HunterMK, Gargosky SE, Zhou J, Bondy CA. The mouse intraovarian insulin-like growth factor I system: departures from the rat paradigm.Endocrinology 1997; 138:3881–3890.

9. Franks S, Gilling-Smith C, Watson H, Willis D. Insulin action in the normaland polycystic ovary. Endocrinol Metab Clin North Am 1999; 28:361–378.

10. Withers DJ. Insulin receptor substrate proteins and neuroendocrinefunction. Biochem Soc Trans 2001; 29:525–529.

11. Withers DJ, Gutierrez JS, Towery H, Burks DJ, Ren JM, Previs S, ZhangY, Bernal D, Pons S, Shulman GI, Bonner-Weir S, White MF. Disruptionof IRS-2 causes type 2 diabetes in mice. Nature 1998; 391:900–904.

12. Burks DJ, de Mora JF, Schubert M, Withers DJ, Myers MG, Towery HH,Altamuro SL, Flint CL, White MF. IRS-2 pathways integrate femalereproduction and energy homeostasis. Nature 2000; 407:377–382.

13. Wynick D, Small CJ, Bacon A, Holmes FE, Norman M, Ormandy CJ,Kilic E, Kerr NC, Ghatei M, Talamantes F, Bloom SR, Pachnis V. Galaninregulates prolactin release and lactotroph proliferation. Proc Natl Acad SciU S A 1998; 95:12671–12676.

14. Lopez J, Talamantes F. Secretion kinetics of prolactin and growthhormone by mouse anterior pituitaries in long term organ cultures. Life Sci1983; 32:2103–2106.

15. Simmgen M, Knauf C, Lopez M, Choudhury AI, Charalambous M,Cantley J, Bedford DC, Claret M, Iglesias MA, Heffron H, Cani PD,Vidal-Puig A, et al. Liver-specific deletion of insulin receptor substrate 2does not impair hepatic glucose and lipid metabolism in mice.Diabetologia 2006; 49:552–561.

16. Hsueh AJ, Billig H, Tsafriri A. Ovarian follicle atresia: a hormonallycontrolled apoptotic process. Endocr Rev 1994; 15:707–724.

17. Richards JS, Russell DL, Robker RL, Dajee M, Alliston TN. Molecularmechanisms of ovulation and luteinization. Mol Cell Endocrinol 1998;145:47–54.

18. Richards JS, Russell DL, Ochsner S, Hsieh M, Doyle KH, Falender AE,Lo YK, Sharma SC. Novel signaling pathways that control ovarianfollicular development, ovulation, and luteinization. Recent Prog HormRes 2002; 57:195–220.

19. Robker RL, Russell DL, Yoshioka S, Sharma SC, Lydon JP, O’MalleyBW, Espey LL, Richards JS. Ovulation: a multi-gene, multi-step process.Steroids 2000; 65:559–570.

20. Cohen P, Frame S. The renaissance of GSK3. Nat Rev Mol Cell Biol2001; 2:769–776.

21. Frame S, Cohen P. GSK3 takes centre stage more than 20 years after itsdiscovery. Biochem J 2001; 359:1–16.

22. Diehl JA, Cheng M, Roussel MF, Sherr CJ. Glycogen synthase kinase-3beta regulates cyclin D1 proteolysis and subcellular localization. GenesDev 1998; 12:3499–3511.

23. Kubota N, Tobe K, Terauchi Y, Eto K, Yamauchi T, Suzuki R, TsubamotoY, Komeda K, Nakano R, Miki H, Satoh S, Sekihara H, et al. Disruptionof insulin receptor substrate 2 causes type 2 diabetes because of liverinsulin resistance and lack of compensatory beta-cell hyperplasia. Diabetes2000; 49:1880–1889.

24. Gonzalez-Robayna IJ, Falender AE, Ochsner S, Firestone GL, RichardsJS. Follicle-Stimulating hormone (FSH) stimulates phosphorylation andactivation of protein kinase B (PKB/Akt) and serum and glucocorticoid-lnduced kinase (Sgk): evidence for A kinase-independent signaling byFSH in granulosa cells. Mol Endocrinol 2000; 14:1283–1300.

25. Roberts R, Iatropoulou A, Ciantar D, Stark J, Becker DL, Franks S, HardyK. Follicle-stimulating hormone affects metaphase I chromosomealignment and increases aneuploidy in mouse oocytes matured in vitro.Biol Reprod 2005; 72:107–118.

26. Carvalho CR, Carvalheira JB, Lima MH, Zimmerman SF, Caperuto LC,Amanso A, Gasparetti AL, Meneghetti V, Zimmerman LF, Velloso LA,Saad MJ. Novel signal transduction pathway for luteinizing hormone andits interaction with insulin: activation of Janus kinase/signal transducer andactivator of transcription and phosphoinositol 3-kinase/Akt pathways.Endocrinology 2003; 144:638–647.

27. Datta SR, Brunet A, Greenberg ME. Cellular survival: a play in three Akts.Genes Dev 1999; 13:2905–2927.

28. Hu CL, Cowan RG, Harman RM, Quirk SM. Cell cycle progression andactivation of Akt kinase are required for insulin-like growth factor I-mediated suppression of apoptosis in granulosa cells. Mol Endocrinol2004; 18:326–338.

29. Castrillon DH, Miao L, Kollipara R, Horner JW, DePinho RA.Suppression of ovarian follicle activation in mice by the transcriptionfactor Foxo3a. Science 2003; 301:215–218.

30. Cunningham MA, Zhu Q, Unterman TG, Hammond JM. Follicle-stimulating hormone promotes nuclear exclusion of the forkheadtranscription factor FoxO1a via phosphatidylinositol 3-kinase in porcinegranulosa cells. Endocrinology 2003; 144:5585–5594.

31. Park Y, Maizels ET, Feiger ZJ, Alam H, Peters CA, Woodruff TK,Unterman TG, Lee EJ, Jameson JL, Hunzicker-Dunn M. Induction ofcyclin D2 in rat granulosa cells requires FSH-dependent relief fromFOXO1 repression coupled with positive signals from Smad. J Biol Chem2005; 280:9135–9148.

32. Moons DS, Jirawatnotai S, Tsutsui T, Franks R, Parlow AF, Hales DB,Gibori G, Fazleabas AT, Kiyokawa H. Intact follicular maturation anddefective luteal function in mice deficient for cyclin- dependent kinase-4.Endocrinology 2002; 143:647–654.

33. Moons DS, Jirawatnotai S, Parlow AF, Gibori G, Kineman RD, KiyokawaH. Pituitary hypoplasia and lactotroph dysfunction in mice deficient forcyclin-dependent kinase-4. Endocrinology 2002; 143:3001–3008.

34. Tong W, Kiyokawa H, Soos TJ, Park MS, Soares VC, Manova K, PollardJW, Koff A. The absence of p27Kip1, an inhibitor of G1 cyclin-dependentkinases, uncouples differentiation and growth arrest during the granulosa-.luteal transition. Cell Growth Differ 1998; 9:787–794.

IRS2 REGULATES OVARIAN FUNCTION 1053