The Multipotency of Luteinizing Granulosa Cells Collected fromMature Ovarian Follicles

KATARZYNA KOSSOWSKA-TOMASZCZUK,a,b CHRISTIAN DE GEYTER,a,b MARIA DE GEYTER,a IVAN MARTIN,b

WOLFGANG HOLZGREVE,a ARNAUD SCHERBERICH,b HONG ZHANGb

aWoman’s Hospital and bDepartment of Research, University of Basel, Basel, Switzerland

Key Words. Follicle-stimulating hormone receptor • Ovarian follicle • Granulosa • Mesenchymal lineage • POU5F1

ABSTRACT

Graafian ovarian follicles consist of follicular fluid, onesingle mature oocyte, and several hundred thousands ofgranulosa cells (GCs). Until now, luteinizing GCs have beenconsidered to be terminally differentiated, destined to un-dergo death after ovulation. Present concepts of luteal func-tion, endocrine regulation of early pregnancy, and the re-cruitment of new ovarian follicles are all based on thecyclical renewal of the entire population of GCs. We nowdemonstrate that luteinizing GCs isolated from the ovarianfollicles of infertile patients and sorted with flow cytometrybased upon the presence of their specific marker, the follicle-stimulating hormone receptor (FSHR), can be maintained inculture over prolonged periods of time in the presence ofthe leukemia-inhibiting factor (LIF). Under those condi-tions the markers of GC function such as FSHR andaromatase gradually disappeared. POU5F1 (POU do-main, class 5, homeobox 1), a typical stem cell marker,

was expressed throughout the culture, but germ line cellmarkers such as nanog, vasa, and stellar were not. Mes-enchymal lineage markers such as CD29, CD44, CD90,CD105, CD117, and CD166, but not CD73, were ex-pressed by substantial subpopulations of GCs. The mul-tipotency of a subset of GCs was established by in vitrodifferentiation into other cell types, otherwise not presentwithin ovarian follicles, such as neurons, chondrocytes,and osteoblasts. Follicle-derived stem cells were also ableto survive when transplanted into the backs of immunoin-competent mice, in vivo generating tissues of mesenchy-mal origin. The unexpected findings of multipotency ofcells with prolonged lifespans originating from ovarianfollicles are likely to have a significant impact on evolvingtheories in ovarian pathophysiology, particularly withreference to ovarian endometriosis and ovarian cancer.STEM CELLS 2009;27:210 –219

Disclosure of potential conflicts of interest is found at the end of this article.

INTRODUCTION

Having been laid down in fetal ovaries, the primordial folliclesremain quiescent for decades. Once their development starts,they give rise to primary follicles, which are characterized byslow growth of the enclosed oocyte and by low granulosa cell(GC) proliferation rates [1]. After their transformation intoprimary follicles, both growth of the oocyte and proliferation ofthe granulosa gain momentum, culminating in the rapid growthof the antral follicle, which finally results in the development ofthe mature Graafian follicle destined for ovulation. Severalhundred thousands of GCs exert a multitude of specializedfunctions encompassing the function of the follicle, such asproducing large amounts of estradiol, adapting its follicle-stim-ulating hormone (FSH) and luteinizing hormone receptivity tothe endocrine milieu, nursing the oocyte, and communicatingboth with the enclosed oocyte and the surrounding thecal cells.The signaling leading to ovulation results in luteinization of thetissue. Luteinized GCs are considered to be terminally differen-

tiated, being replaced in the midluteal phase of the menstrualcycle by small, luteinized cells originating from the surroundingtheca [2].

Both the rapid proliferation of the GCs within the growingfollicle and the exertion of such a large variety of specializedfunctions can only be thought of by accepting the notion that thepopulation of GCs in a healthy follicle is not uniform but ratherconsists of subpopulations of differentiated and less differenti-ated cells, the latter being more capable of mitosis. A similarsituation is encountered in other rapidly proliferating tissueswith specialized functions, such as the bone marrow. The bonemarrow contains a variety of specialized hematopoietic cells,such as myelocytes, reticulocytes, or megakaryocytes but alsoundifferentiated cells, which possess many of the characteristicscontributed to stem cells. Within the particular anatomical en-vironment of the bone marrow, stem cells are first capable ofproliferation and then of differentiation, thereby steadily replac-ing aging and apoptotic hematopoietic cells [3]. Other examplesare found throughout the body, such as the intestines, the brain,the placenta, and the testis [4]. However, in contrast with many

Author contributions: K.K.-T., A.S., and H.Z.: conception and design, data collection and interpretation, manuscript writing and finalapproval; C.D.G.: conception and design, financial and administrative support, provision of study material and patients, data collection, dataanalysis, manuscript writing and final approval; M.D.G.: conception and design, provision of study material, data analysis, manuscriptwriting; I.M.: conception and design, data analysis and interpretation, manuscript writing and final approval; W.H.: manuscript writing andfinal approval. A.S. and H.Z. contributed equally to this work.

Correspondence: Prof. Christian De Geyter, University Hospital of Basel, Spitalstrasse 21, CH-4031 Basel, Switzerland. Telephone: 41 612659315; Fax: 41 61 265 9194; e-mail: cdegeyter@uhbs.ch Received March 25, 2008; accepted for publication October 13, 2008; firstpublished online in STEM CELLS EXPRESS October 30, 2008. ©AlphaMed Press 1066-5099/2008/$30.00/0 doi: 10.1634/stemcells.2008-0233

TISSUE-SPECIFIC STEM CELLS

STEM CELLS 2009;27:210–219 www.StemCells.com

other tissue types, cells with stem cell properties have not yetbeen described within ovarian follicles.

The present article is the first to describe how luteinizingGCs can be cultured over prolonged time periods under condi-tions similar to those known to support the survival of adultstem cells. We characterized a subpopulation of these cells,either shortly after their collection or after prolonged periods ofculture in vitro, and demonstrated that these cells under appro-priate conditions both in vitro and in vivo can be differentiatedinto other cell types, normally not encountered in the ovary. Inthis way, for the first time we were able to establish the multi-potency of a subpopulation of cells collected from the antrum ofhuman ovarian follicles.

MATERIALS AND METHODS

Collection of Luteinizing GCsLuteinizing GCs were collected by transvaginal ultrasound-guidedaspiration from infertile patients treated with controlled ovarianhyperstimulation for assisted reproduction. Patients were treatedwith various exogenous gonadotropins including human meno-pausal gonadotropins (Menopur, Ferring Pharmaceuticals, Switzer-land, http://www.ferringusa.com; Merional, IBSA Institut Bio-chimique SA, Switzerland, http://www.ibsa.com) and recombinantFSH (Gonal-f, Merck Serono S.A., Geneva, Switzerland, http://www.merckserono.net; Puregon, Organon, Pfaffikon, Switzerland,http://www.organon.com) followed by 10,000 IU of human chori-onic gonadotrophin (Pregnyl; Organon). After removal of the cu-mulus oophorus-oocyte-complexes, the freshly collected follicularaspirates were centrifuged for 5 minutes at 111g. GCs were sepa-rated from other cells by density gradient centrifugation on 5 ml ofFicoll PLUS (Amersham Biosciences, Uppsala, Sweden, http://www.amersham.com) for 20 minutes at391g. GCs were clearlyvisible in the interphase layer, isolated by pipetting, washed twice in10 ml of Dulbecco’s modified Eagle’s medium (DMEM) and cen-trifuged again at 111g for 5 minutes for final collection of the cells[5]. The purified cells were placed in freezing medium (fetal calfserum [FCS] with 10% (v/v) dimethyl sulfoxide and stored at�80°C until flow cytometry with fluorescence activated cell sorting[FACS]). To reduce interpatient variability, each experiment wasperformed with mixed populations of GCs collected from at leasteight patients after informed consent was obtained from each pa-tient. This study was approved by our local ethics committee.

Cell CultureGCs were cultured in DMEM containing a high concentration ofglucose (4500 mg/l; Gibco, Basel, Switzerland, http://www.invitrogen.com), supplemented with 15% (v/v) fetal calf serum(Gibco), penicillin/streptomycin (50 �g/ml), L-glutamine (3 mmol/l), �-mercaptoethanol (10 mM stock solution in DMEM), recombi-nant FSH (100 ng/ml or 3 � 10�4 IU/ml, Gonal-f; Serono), and1,000 IU/ml leukemia-inhibiting factor (LIF) (Chemicon Interna-tional, Temecula, CA, http://www.chemicon.com) [6]. Becausecells were highly sensitive to trypsin, a cell scraper was used forpassages. Identical culture conditions were used for the incubationof bone marrow stromal cells to check for potential contaminationwith fibroblasts.

Identification of the Luteinizing GCs Using FACSand SortingGCs were identified by the presence of FSH receptor (FSHR) andsubsequently sorted using FACS. GC identification and sortingwere performed by a dual labeling technique, in which GCs wereidentified as CD3-negative cells, distinguishing them from CD3-positive leukocytes (anti-CD3-APC monoclonal mouse antibodies;Becton Dickinson, Franklin Lakes, NJ, http://www.bd.com) [7].GCs were kept frozen at �80°C and were thawed on the day ofperforming FACS. The first polyclonal goat antibody, raised againsta peptide mapping near the N terminus of the FSHR of human

origin (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com), was added for 30 minutes and kept on ice in the dark.The second donkey anti-goat IgG antibody labeled with fluoresceinisothiocyanate (FITC) (Santa Cruz Biotechnology Inc.) was used forincubation for 30 minutes on ice in the dark. Isotype controls wereused. Isolated populations of FSHR-positive cells, considered to bepure GCs, were used for prolonged culture. As GCs were culturedeither immediately after their aspiration from ovarian follicles orafter thawing, their viability was tested using propidium iodideexclusion and calcein tests (Live/Dead Kit, Invitrogen, Carlsbad,CA, http://www.invitrogen.com).

FACS AnalysisCell suspensions were incubated for 30 minutes at 4°C with fluo-rochrome-conjugated antibodies against the indicated protein or anisotype control. All antibodies were purchased from Becton Dick-inson except the antibodies against CD105 (Serotec Ltd., Oxford,U.K., http://www.serotec.com) and FSHR (Santa Cruz Biotechnol-ogy Inc.). Cells were washed, resuspended in phosphate-bufferedsaline, and analyzed with a FACSCalibur (Becton Dickinson).

Reverse Transcriptase-Polymerase Chain ReactionTotal RNA was extracted from GCs using a RNeasy Total RNA kitfrom Qiagen (Hilden, Germany, http://www1.qiagen.com). Thequantity of RNA was measured by optical density at absorbance of260 nm (ND-1000 Spectrophotometer, NanoDrop Technologies,Wilmington, DE, http://www.nanodrop.com). Total RNA (1 �g)was reverse-transcribed into single-strand cDNA using the cDNAsynthesis kit (Boehringer Mannheim, Mannheim, Germany, http://www.boehringer.com). Primers were synthesized by Microsynth(Balgach, Switzerland): FSHR forward 5�TGGGCTGGATTTTT-GCTTTTG and reverse 5�CCTTGGATGGGTGTTGTGGAC (an-nealing temperature 55°C, DNA product size 529 base pairs [BP]);aromatase forward 5�CAAGTGGCTGAGGCAT and reverse5�GAGAATAGTCGGTGAA (55°C, 429 bp); POU5F1 (POU do-main, class 5, homeobox one, OCT4) [8]; stellar [9]; vasa [9]; nanog[9]; LIF receptor (LIFR) [10]; nestin [11]; neurofilament [11]; and�-3-tubulin [12]. cDNA amplification primers for POU5F1, FSHR,and LIFR were designed to span introns to eliminate genomic DNAcontamination. The �-actin polymerase chain reaction (PCR) prod-uct was used as an internal control (RapidScan). The single-strandcDNA was subjected to 35 cycles of PCR amplification using oneof the primer sets. The amplified products were separated on 1% or2% agarose gels. The reverse transcriptase-PCR products wereanalyzed by DNA sequencing (ABI system, PE Applied Biosys-tems, Foster City, CA, http://www.appliedbiosystems.com); mRNAfrom bone marrow was used as a positive control for POU5F1 [13].

ImmunohistochemistryFor the morphological examination, freshly collected or frozen/thawed GCs were fixed in 1% paraformaldehyde overnight at 4°C,stained with hematoxylin/eosin (H&E), and observed microscopi-cally at various magnifications (Dialux 20; Leitz, Wetzlar, Ger-many, http://www.leica-microsystems.com).

The presence of FSHR and POU5F1 in GCs was demonstratedwith immunohistochemical analysis using antibodies against humanFSHR (Santa Cruz Biotechnology Inc.) and POU5F1 (Abcam,Cambridge, UK, http://www.abcam.com), respectively, followingstandard protocols. The secondary antibody against FSHR consistedof FITC-labeled donkey anti-goat antibodies (Santa Cruz Biotech-nology Inc.). The secondary antibodies against POU5F1 consistedof biotin-conjugated rabbit anti-goat antibodies (DAKO, Glostrup,Denmark A/S, http://www.dako.com) or donkey anti-rabbitTexRed. Stainings for immunohistochemical analysis were per-formed by incubation with the ABC-alkaline phosphatase complexkit (DAKO), counterstained with H&E, and mounted. Sections ofmouse ovaries were used as positive controls for the detection ofPOU5F1.

Differentiation In VitroThe multilineage differentiation capacity of the sorted luteinizingGCs was evaluated by their differentiation into cell types normally

211Kossowska-Tomaszczuk, De Geyter, De Geyter et al.

www.StemCells.com

not encountered within the antrum of ovarian follicles, such asneuronal cells, osteoblasts, and chondrocytes. Freshly collected andsorted GCs were cultured in neuroinductive medium for 10 days.For differentiation of GCs after long-term culture in vitro, GCs werefirst incubated in medium supplemented with LIF for 2 weeks andthen for 3 weeks in one of the three specific differentiation mediadescribed below. After 5 weeks, the pellets were harvested forhistological examination and gene expression analysis.

Differentiation toward the neurogenic lineage was induced byDMEM supplemented with 10% FCS and 30 �mol/l transretinoicacid (Sigma-Aldrich, http://www.sigmaaldrich.com) [14]. Differen-tiation toward the chondrogenic lineage was induced in DMEMculture medium supplemented with 10% FCS, ITS-1 (insulin, trans-ferrin, selenium; Sigma-Aldrich), 0.1 mM ascorbic acid 2-phos-phate, 10 ng/ml transforming growth factor-�1, and 10�7 M dexa-methasone [15]. Osteogenic differentiation was induced in DMEMculture medium supplemented with 10% FCS, 0.1 mM ascorbic acid2-phosphate, 10�2 M �-glycerophosphate, and 10�8 M dexameth-asone [15]. Neurodifferentiation was performed in monolayers,chondrodifferentiation was performed in three-dimensional (3D)cell cultures [15], and osteodifferentiation was performed in both.For the 3D cell culture approximately 3.5 � 105 cells were culturedin pellets in conical microtubes (Sarstedt, Numbrecht, Germany,http://www.sarstedt.com/php/main.php) on an orbital shaker. Themedia were changed three times weekly. For histological examina-tion pellets were fixed in 4% formalin overnight at 4°C, paraffin-embedded, and sectioned (7 �m thickness). The sections collectedfrom the osteogenic culture medium and the respective controlswere stained with H&E or incubated for 10 minutes with alizarinred, washed extensively with water, and observed microscopically(Dialux 20; Leitz). An alternative procedure consisted of stainingwith an antibody against bone sialoprotein (anti-BSP, Immundiag-nostik AG, Bensheim, Germany, http://www.immundiagnostik.com). Decalcification of osteodifferentiated pellets was performedwith Osteodec (Bio-Optica, Milan, Italy. http://www.bio-optica,it).After chondroinduction sections were stained with Safranin O orAlcian Blue.

Quantitative Real-Time PCRPrimers for real-time PCR were synthesized by Microsynth(COLL1, COLL2, osteocalcin [OC] [15], and osteopontin [OP]forward 5�CTCAGGCCAGTTGCAGCC and reverse 5�CAAAAG-CAAATCACTGCAATTCTC) or by Roche (BSP [15] and SOX9forward 5�CCCGCACTTGCACAACG and reverse 5�TCCAC-GAAGGGCCGCT). Power SYBR Green PCR Master Mix(AB Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) for real-time PCR and TaqMan glyceral-dehyde-3-phosphate dehydrogenase (GAPDH) Control Reagent (PEApplied Biosystems) were used as internal controls. cDNA wassubjected to 40 cycles of amplification using the ABI PRISM 7000Sequence Detector System (AB Applied Biosystems). Expression ofthe different genes was presented as the percentage of expression ofGAPDH, a housekeeping gene, by using the formula: 1/2�Ct, where�Ct � gene � GAPDH, �Ctq � control gene � GAPDH, ��Ct ��Ct � �Ctq, and the �� CT Method, fold change, was presentedusing the formula: 2–��Ct.

Differentiation In VivoThe multilineage differentiation capacity of GCs was assessed byimplantation into the backs of immunoincompetent mice. For thatpurpose, GCs were cultured in vitro for 3 weeks in 3D cell culturesand transplanted in nude mice (CD-1 nu/nu, 1-month old; CharlesRiver Laboratories, Wilmington, MA, http://www.criver.com) inaccordance with institutional guidelines. Four to 8 weeks afterimplantation, the mice were sacrificed. The constructs were har-vested and fixed overnight in 1% paraformaldehyde, paraffin-em-bedded, and sectioned. Sections were then stained by H&E andobserved microscopically. Immunohistochemical analysis for BSPwas performed with a BSP-biotin conjugated antibody (Cedar Lane,Hornby, ON, Canada, http://www.cedarlanelabs.com) followed byincubation with ABC-alkaline phosphatase complex (DAKO),counterstaining with H&E, and mounting. For immunohistofluores-cence of BSP, polyclonal rabbit-anti-human BSP antibodies from

Alexis Biochemicals (Lausen, Switzerland, http://www.axxora.com) with secondary goat anti-rabbit PE antibodies from BectonDickinson were used. To distinguish human from murine cells,immunohistofluorescence was performed with anti-human mono-clonal HLA-ABC-biotin conjugated antibody with avidin-FITC sec-ondary antibodies from Becton Dickinson. Some sections were alsostained with Safranin O to assess the formation of cartilage.

RESULTS

The cellular content of follicular fluid aspirated during oocytecollection for assisted reproduction consisted of a mixture ofluteinizing GCs, both single and in clumps, erythrocytes, andlarge epithelial cells, probably also arising from the vaginalepithelium. Most of the erythrocytes were excluded during theFicoll density gradient purification. With FACS a subpopulationof FSHR-bearing cells from the follicular aspirates was consis-tently identified and separated from any contaminating cells. Asillustrated in supporting information Fig. 1, the relative numberof cells expressing FSHR among the entire population of cells inthe unsorted follicular aspirates ranged between 7% and 50%.This broad range corresponds to the individual characteristics ofinfertile women, from whom the GCs were collected, and to thetechnical variabilities of transvaginal, ultrasound-guided aspirationof ovarian follicles. Pooling of GCs from different patients wasused to overcome this variability. After sorting and adhesion to theculture dish, all cells expressed the FSHR, as demonstrated both bycytofluorimetry and immunocytochemistry (Fig. 1A–1C). Afterthawing and sorting, approximately 40% of GCs survived after 1day in culture (supporting information Fig. 2). The sorted cellswere characterized as luteinizing GCs through their expression ofboth FSHR and aromatase (Fig. 2A). These purified luteinizingGCs were then used for prolonged culture.

To evaluate the effect of LIF on the prolonged survival ofGCs in vitro, sorted luteinizing GCs were split into two groupsand cultured separately. One group was cultured in DMEMsupplemented with LIF, whereas another group was culturedwithout LIF. The expression of LIFR in sorted luteinizing GCswas first confirmed (supporting information Fig. 6). The lutein-izing GCs cultured without LIF consistently died within 2 weeks(Fig. 1D), whereas those cultured in medium supplemented withLIF remained viable for up to 4 months and could be passaged(Fig. 1E, 1F). After 7 days with LIF, GCs retained their mor-phology, constructed intercellular connections, and becamestrongly attached to the culture dish (supporting informationFig. 3). In contrast, GCs cultured in the absence of LIF becamefolded after 7 days (green arrow) and lost intercellular connec-tions (red arrow). Moreover, expression of FSHR remainedpresent for at least 7 days in cells cultured with LIF, whereas itwas mostly lost after 7 days in the absence of LIF (supportinginformation Fig. 3). The expression of FSHR, aromatase, andPOU5F1 was then examined on sorted luteinizing GCs culturedin the presence of LIF at various time intervals (after 7, 21, 28,and 56 days) and compared with that on freshly collected sortedand unsorted luteinizing GCs (Fig. 2A). After approximately 7days, the luteinizing GCs progressively lost their ability toexpress FSHR and after 8 weeks also that of aromatase.POU5F1 was expressed in the freshly collected luteinizing GCsand remained expressed in the luteinizing GCs throughout theirculture in medium supplemented with LIF (Fig. 2A). The ex-pression of POU5F1 was confirmed by immunocytochemicalanalysis in the nucleus of some GCs attached to unmarked GCs(Fig. 2B). The same immunostaining for POU5F1 in both oo-cytes and GCs inside antral follicles of mouse ovaries was usedas a positive control (data not shown). Double stainings forFSHR and POU5F1 were performed in GCs after 1, 3, 5, 7, and

212 Stem Cell Characteristics among Ovarian Granulosa Cells

9 days in culture to examine whether the FSHR-expressing GCswere coexpressing POU5F1 and to exclude any contaminationof extrafollicular stem cells. Throughout culture in the presenceof LIF, 1%–3% of GCs expressing FSHR also expressedPOU5F1 (supporting information Fig. 5). In addition, the po-tential overgrowth of the GC culture by contaminating fibro-blasts or other cells was excluded by the observation that nocells survived in the absence of LIF. Furthermore, in another setof experiments bone marrow stromal cells were cultured in thesame medium either supplemented or not supplemented withLIF. Under those conditions the bone marrow stromal cellsremained viable over prolonged time periods in both mediaand expressed neither FSHR nor POU5F1 (supporting infor-mation Fig. 5B).

As visualized with light microscopy, cultured luteinizingGCs exhibited two distinct morphologies: epithelial (between5% and 35% of all cells) or fibroblastic (Fig. 2C, 2D). Theepithelial-like cells disappeared after about 3 weeks in cul-ture, whereas the remaining cells retained their fibroblasticmorphology.

As the transcription factor POU5F1 remained expressedthroughout the prolonged culture, we examined other markers ofpluripotency, characteristic for germ cells, such as nanog, stel-lar, and vasa (supporting information Fig. 5A, 5C). All specificmarkers of germ cells, however, were negative. All PCR prod-ucts yielded the expected fragment sizes. There was no contam-ination of genomic DNA in any of the samples tested, and allnegative controls (RT�/�) processed without reverse transcrip-tase yielded no amplification product.

As GCs originate from the mesoderm, we subsequentlyexamined the mesenchymal cells characteristic of the freshlycollected GCs using various markers of mesenchymal stem cells

(MSCs). Cells were positive for markers CD29, CD44, CD90CD105, CD117, and CD166 but not for CD73 (Fig. 2E). CD117was positive in only 4.5% � 3% of GCs (mean � SD, 10different donors). The typical marker for hematopoietic cells,CD45, was present only in freshly isolated GCs probably due toa contamination with blood cells, although CD34 was not ex-pressed (data not shown).

The multipotency of a subpopulation of cells in the follicularaspirates was assessed. Neuronal markers for neurodifferentia-tion of freshly isolated GCs were thus examined. Two neuronalmarkers, nestin and �-3-tubulin, were weakly expressed infreshly collected GCs, whereas another, neurofilament, was notfound to be expressed in freshly collected GCs (Fig. 3). Subse-quently, the multilineage differentiation capacity of GCs wasassessed after prolonged culture. The cells were differentiated invitro to neuronal, osteoblastic, and chondrogenic lineages, re-spectively, under conditions known to direct the differentiationof MSCs. A clonal analysis of GCs was attempted but failedprobably because of the deleterious effect of trypsin and theincreased general sensitivity of passaged GCs. The capacityof luteinizing GCs to undergo neurogenic differentiation afterprolonged culture in medium supplemented with LIF wasexamined as well. After 5 days GCs cultured as monolayersin medium containing retinoic acid developed neuron-likestructures. After 8 days of culture, approximately 15% of allcells displayed the distinct morphology suggestive of neurons(Fig. 3A, 3B). Various neuronal markers, such as nestin,neurofilament, and �-3-tubulin, were found to be expressedin GCs cultured in retinoic acid-enriched medium but not inthe control medium supplemented with LIF (Fig. 3C). Allexperiments were performed in triplicate. Brain tissue was

Figure 1. Purification and long-term culture of luteinizing granulosa cells (GCs). FSHR in GCs (A, B, C). (A): Fluorescence-activated cell sorting(FACS)/sorting results of freshly collected GCs. (B): Pure population of GCs after FACS/resorting of already sorted cells. R2, FSHR-positive cells;R3, CD3-positive cells. (C): Immunocytochemical analysis of GCs for FSHR after FSHR sorting. GCs cultured in medium (D) after 5 days withoutLIF, (E) after 5 days in the presence of LIF, and (F) after 1-month in the presence of LIF. Abbreviations: FITC, fluorescein isothiocyanate; FSHR,follicle-stimulating hormone receptor.

213Kossowska-Tomaszczuk, De Geyter, De Geyter et al.

www.StemCells.com

used as a positive control, and the expression of all markerswas confirmed by sequencing (data not shown).

The osteoblastic differentiation potential of cultured lutein-izing GCs was examined by alizarin red and BSP staining andgene expression of various osteoblastic markers, such as BSP,OC, and OP. Because GCs cultured as monolayers in osteoin-ductive medium exhibited typical changes in their cellular mor-phology (Fig. 5A) but were too sensitive and became detachedfrom the culture plate, a 3D culture system was introduced.Under those conditions and in the presence of an osteoinductivemedium, previously luteinizing GCs were stained positivelywith alizarin red and marked with anti-BSP antibodies, whereasthe same cells cultured in medium supplemented with LIF orsections of mouse ovaries remained negative (Fig. 4). Thematrix of osteo-differentiated cell pellets was demonstrated to

be mineralized, as documented by rapid dissolution of crystal-lized structures by treatment with an acidic decalcificationbuffer, Osteodec (data not shown). Real-time PCR showed thatexpression of BSP was increased 14-fold, expression of OP wasincreased 66-fold, and expression of OC was increased 3-fold inosteo-differentiated tissue pellets, compared with expression ofcontrol cells cultured with medium supplemented with LIF (Fig.5B). Compared with bone marrow-derived stem cells (BMSCs)cultured as a monolayer in a similar osteoinductive mediumduring the same time [16], BSP was found to be expressed 8times more in BMSCs than in osteo-differentiated GC pellets,but both OP and OC were expressed more in the GC pellets (5-and 1.5-fold, respectively).

The chondrogenic differentiation potential of luteinizingGCs cultured over prolonged periods of time was demonstrated

Figure 2. Characterization of luteinizing granulosa cells (GCs). Characterization of luteinizing GCs, sorted with fluorescence-activated cell sortingand cultured in medium supplemented with LIF. (A): Reverse transcriptase-polymerase chain reaction analysis shows progressive loss of FSHR/aromatase expression during prolonged culture, but not of POU5F1 (OCT4). (B): POU5F1 staining of GCs. Three GCs stained for POU5F1 arevisible. Two GCs are fibroblast-type cells (asterisk and black arrow). One GC is an epidermal-type cell and did not attach to the culture dish (greenarrow). One GC presents negative nucleus staining (asterisk) and one presents positive nucleus staining (black arrow). In both cells cytoplasm isnegative for POU5F1 staining (asterisk and black arrow). Epithelial-like (C) and fibroblast-like (C) morphology of GCs after 10 days of culture. (E):Immunophenotyping results for GC-derived multipotent cells. The following markers of mesenchymal stem cells were detected: CD29, CD44, CD105,CD117, and CD166; however, CD73 was negative. The red line indicates the respective markers, the gray shaded area represents the isotype control.Abbreviations: bp, base pairs; fGC, freshly collected granulosa cells; FSHR, follicle-stimulating hormone receptor; M, marker; RT�/�, negativecontrol; sGC, sorted granulosa cells, cultured GCs after 7, 21, 48, and 56 days, respectively.

Figure 3. Neurogenic differentiation of GCs. Neurogenic differentiation of GCs after prolonged culture in medium supplemented with leukemia-inhibiting factor. (A, B): Neuron-like morphology after neurogenic induction of GCs. (C): Reverse transcriptase-polymerase chain reaction results ofGC neural induction showing expression of the neuronal markers, nestin, �-3-tubulin, and neuro-3-filament. Abbreviations: GC, granulosa cell; N,nestin; NF, neuro-3-filament; sGC, sorted GCs; T, �-3-tubulin.

214 Stem Cell Characteristics among Ovarian Granulosa Cells

by the presence of glycosaminoglycan (GAG) in GC pelletscultured in 3D cultures in chondroinductive medium using Sa-franin O staining (Fig. 6A, 6B). After chondrogenic differenti-ation, the tissue sections were weakly positive for Safranin Ostaining, whereas GCs cultured in medium supplemented withLIF were negative. Sections of mouse ovary were also negativefor GAG staining (data not shown). With real-time PCR theexpression of various genes specific for chondrogenic differen-tiation was upregulated in GCs cultured in chondrogenic differ-entiation medium, compared with GCs cultured with controlmedium with LIF: collagen-1 (COLL1) 4.5-fold, collagen-2(COLL2) 6.5-fold, and SOX9 12.5-fold (Fig. 6C). These valueswere also compared with those for expanded primary chondro-cytes cultured as a monolayer in the same chondrogenic mediumduring the same period [15]. Expression of COLL2 and SOX9was higher in chondrocytes (2.5- and 1.9-fold, respectively), butthat of COLL1 was higher in chondro-induced GCs (1.6-fold).

The capacity of GCs cultured in medium supplemented withLIF to survive and differentiate in vivo into other, distinct tissue

types was examined through subcutaneous transplantation intothe back of immunoincompetent, nude mice. The implants wereharvested either 4 or 8 weeks after transplantation. After 8weeks, the implanted cell pellets appeared to be more integratedwithin the murine tissue than after 4 weeks (Fig. 7A, 7B) andexpression of BSP was detected (Fig. 7C, 7D, 7F). Those cellswere always surrounded by murine cells also showing expres-sion of BSP (Fig. 7E, 7F). The distinct origin of both cell typeswas tested by HLA-ABC staining, which is specific for humantissue. Eight weeks after transplantation some GAG depositionwas also detected, as demonstrated through the Safranin Ostaining (Fig. 7G, 7H, 7I).

DISCUSSION

Up to now, luteinizing GCs were considered to be terminallydifferentiated, unavoidably becoming apoptotic a few days after

Figure 4. Osteogenic differentiation of granulosa cells (GCs). Osteogenic differentiation of GCs after prolonged culture in medium supplementedwith leukemia-inhibiting factor. Stainings for alizarin red (A–C) and BSP (D–F) for three-dimensional-cultured luteinizing GCs and sections of mouseovaries. Abbreviation: BSP, bone sialoprotein.

Figure 5. Osteogenic differentiation ofgranulosa cells (GCs). (A): GCs cultured inmonolayers in control medium exhibiting fi-broblast-like cells and in osteoinductive me-dium after 5 days, showing epithelial-likemorphology of cells. (B): Real-time poly-merase chain reaction results for BSP, OP,and OC expression in control GCs, in GCsafter osteodifferentiation, and in bone mar-row stem cells. Abbreviations: BSP, bone sia-loprotein; OC, osteocalcin; OP, osteopontin.

215Kossowska-Tomaszczuk, De Geyter, De Geyter et al.

www.StemCells.com

ovulation. Indeed, prolonged culture in vitro of GCs collectedfrom preovulatory follicles has not been possible beyond 10

days. Instead, various researchers have attempted to use GCsfrom granulosa tumors [5, 17] or immortalized GCs [18] to

Figure 6. Chondrogenic differentiation ofGCs. Chondrogenic differentiation of GCsafter prolonged culture in medium supple-mented with LIF. Safranin O staining incontrol GCs (A) and in GCs after chondroin-duction (B). (C): Real-time polymerasechain reaction results for expression ofCOLL1, COLL2, and SOX9 in control GCsin regular culture medium, in GCs afterchondroinduction, and in chondrocytes. Ab-breviation: GC, granulosa cell.

Figure 7. In vivo differentiation of granulosa cells (GCs). HLA-ABC (green) staining of pellets containing GCs (A) 8 weeks after implantation at 20�magnification and (B) 4 weeks after implantation at 10� magnification. Staining for BSP using immunohistochemistry (C) 8 weeks after transplantation (red) at 20�magnification, (D) BSP using immunohistofluorescence (red) at 20� magnification, and (E) for HLA-ABC (green) at 20� magnification. (F): Double staining forBSP and HLA-ABC at 20� magnification. (G): Staining for Safranin O, 8 weeks after transplantation (red) at 10� magnification. (H): Staining for HLA-ABC at10� magnification. (I): Double staining for Safranin O and HLA-ABC 10� magnification. Abbreviation: BSP, bone sialoprotein.

216 Stem Cell Characteristics among Ovarian Granulosa Cells

construct GC lines suitable for research purposes. We were ableto culture luteinizing GCs, collected from infertile womentreated with controlled ovarian hyperstimulation for assistedreproduction over prolonged time periods. The variable fertilitystatus of single patients certainly affects GC function. There-fore, in all experiments the samples of several patients werepooled to reduce this potential confounding factor. The crucialdifference between our present approach and earlier trials [5]was the use of LIF, a cytokine commonly used in culture mediasupporting the development and growth of stem cells. LIFpromoted the long-term survival of luteinizing GCs, whereas inthe absence of LIF these cells invariably became apoptotic. LIFis a glycoprotein with a remarkable range of biological actionsin different tissues, such as long-term maintenance of mouse,but not human, embryonic stem cells [19]. In a number oftissues, such as the brain [20], the gut [21], and bone marrow[22], LIF has been shown to be important for stem cell self-renewal.

LIF has been detected in both fetal and adult human ovaries[10] and may be involved in the transition of primordial toprimary follicles [23]. LIF is present in the follicular fluid, andits secretion can be enhanced by human chorionic gonadotropin[24, 25]. LIF receptor activity has been detected in oocytes andpreimplantation human embryos [26], suggesting a role of gen-ital tract LIF in the process of follicular development andimplantation [27].

Although LIF permitted the prolonged survival of luteiniz-ing GCs, they progressively lost their major characteristics, suchas the FSHR and aromatase. The overgrowth of the luteinizingGCs by a subpopulation of other cells such as fibroblasts wasexcluded by the extraction of a pure population of GCs withFACS based on the FSHR. Apart perhaps from the oocyte [28],GCs are the only cell type in the female body possessing theFSHR. With both flow cytometry and immunocytochemicalanalysis we demonstrated that all experiments were performedwith a highly homogeneous and almost pure population of GCs.The sorted GCs continued to possess all of typical characteris-tics such as aromatase and FSHR over a period of at least 10days. As follicles mature, the amount of mRNA for FSHR isknown to decrease, whereas that of aromatase increases [29].The presence of FSHR in sorted GCs was confirmed by immu-nocytochemical analysis.

The two different morphologies of GCs, epithelial and fi-broblastic, found to be present in the medium during initialculture, correspond to different intrafollicular locations, fromwhich the GCs were removed during transvaginal ultrasound-guided follicular aspiration. There is evidence that GCs origi-nating from close to the basal membrane are columnar, whereasthose originating from the middle layer are rounded and thoseoriginating from the central part of the follicle, close to theoocyte, are flattened [30]. Some authors argue that the elongatedGCs may have lost aromatase activity, cytoplasmic changescompatible with luteinization [31]. When cultured in monolay-ers, GCs invariably become luteinized and convert their epithe-lial morphology into a fibroblastic one, explaining why the lattermorphology became dominant during prolonged culture.

The progressive loss of all characteristics of GCs duringprolonged culture and the continued expression of POU5F1 inluteinizing GCs gave rise to the hypothesis that some follicularcells may exhibit stem cell properties. POU5F1, also known asOCT4, is a transcription factor and one of the two isoformsproduced by the OCT4 gene, which is considered to be a mainregulator of differentiation and self-renewal [32–36]. The ex-pression of POU5F1 has not been demonstrated previously ingranulosa. Using immunostaining we confirmed the presence ofPOU5F1 in the nucleus of human GCs and in mouse ovaries.

To confirm the multipotency of GCs, we first examinedtypical mesenchymal markers of multipotent MSCs. With theexception of CD73, subpopulations of GCs expressed thosemarkers, thereby adding to the notion that during prolongedculture in the presence of LIF these cells possess many but notall attributes of the MSC lineage. We then provided evidence ofthe multipotency of a GC subpopulation by demonstrating theirdifferentiation potential when cultured with specific neuroin-ductive, chondroinductive, and osteoinductive culture media.

Chondroinduction was demonstrated by real-time PCR anddetection of GAG, a method commonly used for the detection ofcartilage matrix [15]. Safranin O staining was used becauseAlcian Blue staining, commonly applied for GAG staining, alsostained Call-Exner bodies in ovarian tissue, as described previ-ously [37]. Safranin O staining was weak, probably becausesome GCs died early during chondrogenic differentiation. Ex-pression patterns of cartilage-related genes during chondrogenicdifferentiation in GCs and expanded primary chondrocytes con-trol were similar. The COLL1 and COLL2 genes are expressedduring cartilage development. COLL1 is expressed by cells firstentering differentiation, whereas COLL2 is expressed in differ-entiated cells [38]. SOX9 is a key regulator of chondrogenesis[39]. The difference in expression levels of the three genessuggest that GCs under the conditions described undergo earlychondrogenic differentiation.

Osteoblastic differentiation was demonstrated by stainingwith alizarin-red, a dye assessing the presence of calcium inmineralized matrices. BSP staining was also performed andconfirmed the osteoblastic differentiation of GCs. BSP and OPare prominent components of bone extracellular matrix. Theyare expressed by differentiated osteoblastic cells and serve asindicators of osteoblastic differentiation of BMSCs [15]. After 3weeks of osteoblastic differentiation, the expression of BSPmessenger RNA exhibited a 14-fold increase and expression ofOP messenger RNA exhibited a 66-fold increase compared withthose of GCs cultured in LIF medium. OC messenger RNAshowed a limited threefold increase in differentiated GCs. Thislow expression level is explained by OC starting to be expressedlater during the osteoblastic differentiation process, betweendays 16 and 30 of culture, resulting in a maximal althoughlimited expression at day 21.

The differentiation potential of GCs into the neuronal lin-eage was less pronounced among freshly isolated GCs than afterprolonged culture. Neuroinduction of freshly collected GCsinduced the expression of only two neuronal markers, nestin and�-3-tubulin, but not neurofilament. Obviously, the size of thesubpopulation of GCs with multipotent stem cell characteristicsis smaller shortly after follicular aspiration than after prolongedculture.

Because some MSC markers were not uniformly present inthe sorted GCs and although we cannot entirely exclude theovergrowth of the sorted GCs by contaminating mesenchymalstem cells admixed from other tissues and expanding duringprolonged culture, we hypothesize the presence of several sub-populations of GCs in preovulatory follicles, each expressingdifferent MSC markers according to their degree of differenti-ation. In many organs, adult tissues typically contain variouscell populations, including multipotent stem, progenitor cells,and terminally differentiated cells [40].

For the first time, prolonged culture of luteinizing GCs inmedium supplemented with LIF allows the selection of lessdifferentiated GCs, which exhibited a certain degree of plastic-ity, as they could be differentiated in vitro into three distinctlineages: neuronal, chondrocytic, and osteoblastic, all normallynot found in healthy ovarian follicles. Both the survival of GCsafter prolonged culture in the presence of LIF and their ability todifferentiate into cells of the mesodermal lineage were also

217Kossowska-Tomaszczuk, De Geyter, De Geyter et al.

www.StemCells.com

confirmed in vivo. The transplanted and differentiated humancells were surrounded by specific mouse cells differentiated intothe same direction, all within mesenchymal lineage.

Recently, much controversy has arisen concerning the pos-sible presence of stem cells in the germinal layer of the mouseovary and their potential for replacing germ cells postnatally[41]. The ovarian surface epithelium of adult human femalesand bone marrow MSCs were reported to be a source of germcells [42]. However, the possibility that the luteinizing GCscultured in the presence of LIF were germ cells was ruled out bythe lack of expression of nanog, vasa, and stellar.

The high concentration of LIF in the follicular fluid ofmature follicles and the presence of LIFR in GCs both suggestthat our findings are physiologically relevant. Multipotent stemcells in ovarian follicles may be involved in the early origin ofsome forms of ovarian cancer as well as in the origin of ovarianendometriosis, which is considered to arise from undifferenti-ated, metaplastic cells in the ovary. This hypothesis is supportedby the abundant secretion of LIF by endometrial cells [27].Evidence for the presence of stem cells in the ovarian folliclewas provided recently by the identification of stem cells fromthe thecal layer in the neonatal mouse ovary [43]. A model of

how GCs arise from a population of stem cells has been dis-cussed previously [44]. Here, we demonstrate the presence ofmultipotent follicular cells, characterized as GCs, which survivein the presence of LIF.

ACKNOWLEDGMENTS

We gratefully thank Dr. Alicia Rovo and Dr. Andrea Barberofrom the University Hospital of Basel, Switzerland, for provid-ing bone marrow tissue and chondrocytes and Dr. Jacek Kowal-ski from the Saint Adalbert Specialist Hospital in Gdansk,Poland, for his contribution in interpreting various tissue sec-tions. This work was funded by Grant 320000-113517/1 fromthe Swiss National Foundation.

DISCLOSURE OF POTENTIAL CONFLICTS

OF INTEREST

The authors indicate no potential conflicts of interest.

REFERENCES

1 Gougeon A. Regulation of ovarian follicular development in primates:facts and hypotheses. Endocr Rev 1996;17:121–155.

2 Niswender GD, Juengel JL, Silva PJ et al. Mechanisms controlling thefunction and the life span of the corpus luteum. Physiol Rev 2000;80:1–24.

3 Stem Cell Information. The National Institutes of Health recourse ofstem cell research. Available at http://stemcells.nih.gov/. Accessed Jan-uary 10, 2006.

4 Young HE, Duplaa C, Katz R et al. Adult-derived stem cells and theirpotential for use in tissue repair and molecular medicine. J Cell Mol Med2005;9:753–769.

5 Zhang H, Vollmer M, De Geyter M et al. Characterization of an immor-talized human granulosa cell line (COV434). Mol Hum Reprod 2000;6:146–153.

6 Gough NM, Gearing DP, King JA et al. Molecular cloning andexpression of the human homologue of the murine gene encodingmyeloid leukemia-inhibitory factor. Proc Natl Acad Sci U S A 1988;85:2623–2627.

7 De Neubourg D, Robins A, Fishel S et al. Quantification of insulin-likegrowth factor I receptors on granulosa cells with flow cytometry afterfollicular stimulation. Hum Reprod 1998;13:161–164.

8 Abdel-Rahman B, Fiddler M, Rappolee D et al. Expression of transcrip-tion regulating genes in human preimplantation embryos. Hum Reprod1995;10:2787–2792.

9 Ezeh UI, Turek PJ, Reijo RA et al. Human embryonic stem cell genesOCT4, NANOG, STELLAR, and GDF3 are expressed in both seminomaand breast carcinoma. Cancer 2005;104:2255–2265.

10 Abir R, Fisch B, Jin S et al. Immunocytochemical detection and RT-PCRexpression of leukaemia inhibitory factor and its receptor in human fetaland adult ovaries. Mol Hum Reprod 2004;10:313–319.

11 Scintu F, Reali C, Pillai R et al. Differentiation of human bone marrowstem cells into cell with a neural phenotype: diverse effects of twospecific treatment. BMC Neurosci 2006;7:14–27.

12 Dozier JH, Hiser L, Davis JA et al. B class II tubulin predominates innormal and tumor breast tissues. Breast Cancer Res 2003;5:R157–R169.

13 Pochampally RR, Smith JR, Ylostalo J et al. Serum deprivation of humanmarrow stromal cells (hMSCs) selects for a subpopulation of earlyprogenitor cells with enhanced expression of Oct-4 and other embryonicgenes. Blood 2004;103:1647–1652.

14 Portmann-Lanz CB, Schoeberlein A, Huber A et al. Placental mesenchy-mal stem cells as potential autologous graft for pre- and perinatalneuroregeneration. Am J Obstet Gynecol 2006;194:664–673.

15 Barbero A, Ploegert S, Heberer M et al. Plasticity of clonal populationsof dedifferentiated adult human articular chondrocytes. Arthritis Rheum2003;48:1315–1325.

16 Frank O, Heim M, Jakob M et al. Real time quantitative RT-PCRanalysis of human bone marrow stromal cells during osteogenic differ-entiation in vitro. J Cell Biochem 2002;85:737–746.

17 Nishi Y, Yanase T, Mu Y et al. Establishment and characterization of a

steroidogenic human granulosa-like tumor cell line, KGN that expressesfunctional follicle-stimulating hormone receptor. Endocrinology 2001;142:437–445.

18 Tajima K, Hosokawa K, Yoshida Y et al. Establishment of FSH-respon-sive cell lines by transfection of pre-ovulatory human granulosa cellswith mutated p53 (p53val135) and Ha-ras genes. Mol Hum Reprod2002;8:48–57.

19 Daheron L, Opitz SL, Zaehres H et al. LIF/STAT3 signaling fails tomaintain self-renewal of human embryonic stem cells. STEM CELLS2004;22:770–778.

20 Bauer S, Patterson PH. Leukemia inhibitory factor promotes neural stemcell self-renewal in the adult brain. J Neurosci 2006;26:12089–12099.

21 Kalabis J, Patterson MJ, Enders GH et al. Stimulation of humancolonic epithelial cells by leukemia inhibitory factor is dependent oncollagen-embedded fibroblasts in organotypic culture. FASEB J2003;17:1115–1117.

22 Jiang Y, Vaessen B, Lenvik T et al. Multipotent progenitor cells can beisolated from postnatal murine bone marrow, muscle, and brain. ExpHematol 2002;30:896–904.

23 Nilsson EE, Kezele P, Skinner MK. Leukemia inhibitory factor (LIF)promotes the primordial to primary follicle transition in rat ovaries. MolCell Endocrinol 2002;25:188:65–73.

24 Arici A, Oral E, Bahtiyar O et al. Leukaemia inhibitory factorexpression in human follicular fluid and ovarian cells. Hum Reprod1997;12:1233–1239.

25 Coskun S, Uzumcu M, Jaroudi K al. Presence of leukemia inhibitoryfactor, interleukin-12 in human follicular fluid during follicular growth.Am J Reprod Immunol 1998;40:13–18.

26 Van Eijk MJ, Mandelbaum J, Salat-Baroux J et al. Expression ofleukaemia inhibitory factor receptor subunits LIFR� and gp130 inhuman oocytes and preimplantation embryos. Mol Hum Reprod 1996;2:355–360.

27 Arici A, Engin O, Attar E et al. Modulation of leukemia inhibitory factorgene expression and protein biosynthesis in human endometrium. J ClinEndocrinol Metab 1995;80:1908–1915.

28 Meduri G, Charnaux N, Driancourt MA et al. Follicle-stimulatinghormone receptors in oocytes? J Clin Endocrinol Metab 2002;87:2266 –2276.

29 Slomczynska M, Duda M, Slezak K. The expression of androgen recep-tor, cytochrome P450, aromatase and FSH receptor mRNA in the porcineovary. Folia Histochem Cytobiol 2001;39:9–13.

30 Rodgers RJ, Irving-Rodgers HF, van Wezel IL et al. Dynamics of themembrana granulosa during expansion of the ovarian follicular antrum.Mol Cell Endocrinol 2001;171:41–48.

31 Gutierrez CG, Glazyrin AN, Robertson GW et al. Ultra-structural char-acteristics of bovine granulosa cells associated with maintenance ofoestradiol production in vitro. Mol Cell Endocrinol 1997;134:51–58.

32 Pan GJ, Chang ZY, Scholer HR et al. Stem cell pluripotency andtranscription factor Oct4. Cell Res 2002;12:321–329.

33 Cauffman G, Liebaers I, Van Steirteghem A et al. POU5F1 isoformsshow different expression patterns in human embryonic stem cells andpreimplantation embryos. STEM CELLS 2006;24:2685–2691.

218 Stem Cell Characteristics among Ovarian Granulosa Cells

34 Liedtke S, Enczmann J, Waclawczyk S et al. Oct4 and its pseudogenesconfuse stem cell research. Cell Stem Cell 2007;1:364–366.

35 Zangrossi S, Marabese M, Broggini M et al. Oct-4 expression in adulthuman differentiated cells challenges its role as a pure stem cell marker.STEM CELLS 2007;25:1675–1680.

36 Kotoula V, Papamichos S, Lambropoulos AF. Revisiting Oct4 ex-pression in peripheral blood mononuclear cells. STEM CELLS 2008;26:290 –291.

37 van Wezel IL, Irving-Rodgers HF, Sado Y et al. Ultrastructure andcomposition of Call-Exner bodies in bovine follicles. Cell Tissue Res1999;296:385–394.

38 Bonaventure J, Kadhom N, Cohen-Solal L et al. Reexpression of carti-lage-specific genes by dedifferentiated human articular chondrocytescultured in alginate beads. Exp Cell Res 1994;212:97–104.

39 Jenkins E, Moss JB, Pace JM et al. The new collagen gene COL27A1

contains SOX9-responsive enhancer elements. Matrix Biol 2005;24:177–184.

40 Hochedlinger K, Jaenisch R. Nuclear reprogramming and pluripotency.Nature 2006;441:1061–1067.

41 Johnson J, Canning J, Kaneko T et al. Germline stem cells andfollicular renewal in the postnatal mammalian ovary. Nature 2004;428:145–150.

42 Bukovsky A, Svetlikova M, Caudle M. Oogenesis in cultures derivedfrom adult human ovaries. Reprod Biol Endocrinol 2005;3:17–30.

43 Honda A, Hirose M, Hara K et al. Isolation, characterization, and in vitroand in vivo differentiation of putative thecal stem cells. Proc Natl AcadSci U S A 2007;104:12389–12394.

44 Rodgers RJ, Lavranos TC, van Wezel IL et al. Development of theovarian follicular epithelium. Mol Cell Endocrinol 1999;151:171–179.

See www.StemCells.com for supporting information available online.

219Kossowska-Tomaszczuk, De Geyter, De Geyter et al.

www.StemCells.com