Mammalian ovaries undergo considerable remodeling during the lifetime of the organism, leading to the supposition that somatic stem cells account for or contribute to this cyclic regeneration. While much of ovarian stem cell research has been focused on germ cells, recent interest in normal somatic stem cells has been driven by their possible links to ovarian cancer stem cells. While evidence for stem cell biology with regards to granulosa cells is scant, recent work has isolated potential somatic stem cells for the theca and ovarian surface epithelium. Additionally, evidence for potential cancer initiating cells for ovarian epithelial carcinomas continues to mount.
Reproductive organs undergo considerable remodeling during the lifetime of the mammalian organism, leading to the supposition that somatic stem cells account for or contribute to this cyclic regeneration. While much of ovarian stem cell research has been focused on germ cells, recent interest in normal somatic stem cells has been driven by their possible links to ovarian cancer stem cells. Given that the ovarian surface epithelium is the postulated source for 90% of human ovarian cancers (Gondos, 1975; Herbst, 1994; Auersperg et al., 1998), understanding presumptive stem/progenitor function in this less well understood component of the ovary may reveal mechanisms of tumor progression with resultant important clinical implications. The current status of our understanding of stem cell biology in the somatic components of the ovary is schematized in Figure 1.
2. Ovarian development
An understanding of the salient features of normal ovarian development is necessary before delving into stem cell functions of its somatic compartments and begins with the formation, migration, and organogenesis of the germ cells (see Loffler and Koopman, 2002 and Oktem and Oktay, 2008 for comprehensive reviews).
2.1. Germ cells
Primordial germ cells (PGCs) form at E6.5 in the proximal epiblast where as few as six Blimp-1 expressing cells are detected (Ohinata et al., 2005). Blimp-1 appears to initiate lineage specificity by repressing Hox and other somatic genes while extra-embryonic ectoderm BMP 2, 4, and 8b signaling (Molyneaux and Wylie, 2004; Lawson et al., 1999; Ying et al., 2000) expands this population which expresses tissue-specific alkaline phosphatase and Stella (Saitou et al., 2002) at embryonic day 7.2 (E7.2) in the mouse (Ginsburg et al., 1990) and as early as week 3 in the endoderm of the yolk sac wall of the human (Hacker and Moore, 1992). PGC migration to the mesonephros occurs between E8-E12 in the mouse (Loffler and Koopman, 2002) and week 8 of gestation in the human (Hacker and Moore, 1992). Mouse primordial germ cells undergo proliferation and imprint erasure as they traverse from the proximal primitive streak at the base of the allantois to the hindgut endoderm (E8.5–9.5) and then to the genital ridges. Progression of the PGCs to the developing urogenital ridges appears to involve largely uncharacterized chemotactic signals (De Felici et al., 2005; Molyneaux et al., 2003; Godin et al., 1990), integrins (Anderson et al., 1999), and C-kit signaling (Buehr et al., 1993). As they migrate, PGCs proliferate 170-fold by mitotic division (Tam and Snow, 1981) and lose imprinting imposed by DNA methylation (E10.5–12.5; Hajkova et al., 2002; Lee et al., 2002; Reik and Walter, 2001) and complex chromatin modification by methylation or acetylation of histones on lysine and arginine residues (Hajkova et al., 2008; Seki et al., 2007; Seki et al., 2005; Kimmins and Sassone-Corsi, 2005). This erasure is presumably necessary to reset epigenetic memory in germ cells and its complex regulation remains a continuing challenge for somatic nuclear transfer. The transmembrane protein Fragilis induces the germ cell gene Stella which represses the developmental homeobox genes, while Oct4, Sox2, and Nanog, preserve the pleuripotency of the PGCs (Saitou et al., 2002). The PGCs lose their migratory phenotype at E12.5 (Ginsburg et al., 1990) and at E13.5 female germ cells undergo meiosis in an anterior to posterior wave which denotes ovarian differentiation to become oogonia, a process mediated by Stra8 which is stimulated by retinoic acid. In contrast to the embryonic ovary, the embryonic testes expresses CYP26b1 which degrades retinoic acid so Stra8 is not expressed (Bowles et al., 2006; Koubova et al., 2006).
Normal migration and colonization of the PGCs are necessary for further ovarian development, as ovarian dysgenesis with degeneration of ovarian somatic cells occurs in germ cell deficient mice (Merchant-Larios and Centeno, 1981; Behringer et al., 1990; Hashimoto et al., 1990). Death of germ cells occurs when E12.5 ovaries are placed ectopically beneath the renal capsule, an event followed by formation of testicular tubules, again indicating the important role of the germ cells in ovarian development (Taketo et al., 1993). Furthermore, there is evidence to suggest developmental germ cell tumors may result from incomplete migration of the PGCs to the developing ovary (Göbel et al., 2000; Schneider et al., 2001).
2.2. Ovarian somatic tissue and ovarian surface epithelium
Figure 2 gives a schematic overview of ovarian somatic development. The development of the somatic gonad begins on E10 in the mouse (4 weeks gestation in the human) as a thickening of the coelomic epithelium on the ventro-medial side of the mesoderm (Swain and Lovell-Badge, 1999). The indifferent gonad is essentially a mass of blastema (the primordial mesenchymal cell mass) surrounded by the coelomic-epithelium derived surface epithelium. This mesenchymal cell mass contains elements which are destined to become the supportive (granulosa), steroidogenic (theca and granulosa), or structural (stroma) cells of the ovary. After the primordial germ cells arrive in the developing gonad (∼E12–13.5), an arrangement of loose cords called the ovigerous cords begins to form around clusters of germ cells (Odor and Blandau, 1969; Konishi et al., 1986). Developing somatic cells (presumed pre-granulosa cells) then separate the germ cell clusters into individual oocytes surrounded by a monolayer of granulosa cells, forming the primordial follicle (Pepling and Spralding, 1998; Merchant-Larios and Chimal-Monroy, 1989). While the embryonic origins of the granulosa cells are still a matter of debate, there is evidence suggesting that the ovarian surface epithelium is at least a partial source of granulosa cells (Sawyer et al., 2002).
2.3. Genetic regulation of ovarian development
Compared to the genetic regulation of testicular development, the genes responsible for ovarian development are relatively unknown and it was once thought the ovaries developed passively as a result of the absence of testicular determining genes. Only a handful of genes required for the formation of the ovary have been identified and studied, almost exclusively in the mouse. Of these, Wnt-4 is the gene most clearly associated with ovarian development, with homozygous mutant males exhibiting normal testicular development while their female counterparts are virilized with absence of the Müllerian duct and morphologically masculinzed gonads with subsequent degeneration of meiotic-stage oocytes (Vainio et al., 1999). Follistatin is a downstream component of Wnt-4 signaling which has also been associated with regulating normal ovarian organogenesis (Yao et al., 2004). Recent data has shown that Wnt-4 activation is regulated by Respondin-1 mediation of the canonical β-catenin pathway and that β-catenin stabilization in the XY gonad is sufficient to cause male-to-female sex reversal (Chassot et al., 2008; Maatouk et al., 2008). Additionally, a member of the forkhead transcription factors, Foxl2, has been identified as a gene that appears to repress the male genetic program and insure normal granulosa cell development around growing oocytes, allowing for normal ovarian development (Ottolenghi et al., 2005; Schmidt et al., 2004; Uda et al., 2004). While work continues on the genetic regulation of ovarian development, it is interesting to note that prenatal ovarian development occurs independent of steroid hormone action (Couse et al., 1999).
3. Granulosa and theca cells
The granulosa and theca cells of the ovary serve to support the germ cells within the developing follicle. Initially indistinguishable from the ovarian stroma, theca cells surround the developing follicle, form the two layers known as the theca externa and interna, and produce the androgens which are ultimately converted to estradiol by the granulosa cells. While converting theca-produced androgens into estradiol via aromatase, the granulosa cells form the multilayered cumulus oophorus and later the Graafian or pre-ovulatory follicle which surrounds the germ cells. After ovulation, both the granulosa and theca cells contribute to the corpus luteum which is responsible for producing the estrogen and progesterone necessary to support a developing pregnancy. While the complex biology of these cells suggests an obvious somatic stem cell-mediated process, substantive evidence to this end is lacking for the granulosa cells but is just recently being elucidated for the theca cells.
3.1. Granulosa cells
The origin of the pre-granulosa cells (the flattened somatic cells of the primordial follicle surrounding the oocyte) is not known but evidence supports three proposed sources: the developing ovarian blastema (Pinkerton et al., 1961), the mesonephric cells of the rete ovarii (Byskov and Lintern-Moore, 1973), and the developing ovarian surface epithelium (Gondos, 1975; Sawyer et al., 2002). Until recently, most would have agreed that the cells that give rise to the granulosa cells (and their derived structures) during folliculogenesis have already segregated with each primordial follicle, and separate each germ cell from the surrounding ovarian stroma by a basal lamina where they lie dormant prior to follicular recruitment. While Bukovsky (1995; 2004; 2008) suggests that the tunica albuginea immediately underlying the ovarian surface epithelium gives rise to both post-natal oocytes and their associated pre-granulosa cells (Bukovsky et al., 2004), the existence of germline stem cells and the possibility of post-natal oogenesis and their further development remains to be determined definitively (Johnson et al., 2005; Eggan et al., 2006, see Tilly et al., 2008 for a comprehensive review).
While granulosa cell growth and differentiation during folliculogenesis is a complex and interesting interplay of paracrine and endocrine factors, definitive evidence that the development of the granulosa-derived structures of the follicle is a stem cell-mediated process has yet to be produced. It could be argued that given the complex changes that occur during folliculogenesis, the pre-granulosa cells residing in the primordial follicle should be labeled fate-determined progenitor cells capable of forming the various structures of the developing follicle. However, there has been a lack of evidence so far to support the presence of asymmetric division, pluri- or multi- potency, or indefinite self-renewal in granulosa cells that characterizes stem cell biology.
3.2. Theca cells
While the presence of probable theca cell precursors in the ovarian stroma and interstitium was proposed by Hirshfield in 1991, identification and isolation of these cells has proven elusive. Recent work by Honda and his colleagues (Honda et al., 2007) led to isolation of ‘putative thecal stem cells’ after enzymatic and mechanical dissociation of newborn mice ovaries and growth of the resulting cell suspension in serum-free germline stem cell (GS) media. Non-adherent anchorage independent spheres exhibited the morphology of ovarian interstitial somatic cells and expressed gene profiles suggestive of theca cells, not germ or granulosa cells. By supplementing the media with serum, luteinizing hormone, insulin-like growth factor-1, stem cell factor, and granulosa cell-conditioned media in a stepwise manner, they were able to induce subcultures of these cells to differentiate into lipid producing, androgen secreting cells which morphologically resembled theca cells. Furthermore, transplantation of these cells isolated from whole-body green fluorescence protein-expressing transgenic mice into the ovaries of wild-type recipients showed scattered interstitial GFP with aggregation of GFP cells immediately adjacent to developing follicles and subsequent GFP expression in both theca interna and externa during folliculogenesis (see Figure 3). While there are still questions to be answered such as the exact location of these thecal precursors in vivo, the cell surface marker profile of these cells, and the niche in which these cells reside, the ability to isolate and characterize these cells represents a significant step towards understanding follicular development.
4. Granulosa cell tumors and thecomas
Adult granulosa cell tumors (GCT), the most common ovarian stromal tumor, account for approximately 2–5% of all ovarian cancers. Juvenile GCTs are 20 to 50 times more rare (Schumer and Cannistra, 2003; Colombo et al., 2007). Unlike the epithelial ovarian cancers, sex-cord stromal ovarian tumors (including granulosa cell tumors and thecomas) do not seem to have a demonstrable hereditary component. In addition, reports of oncogene involvement are inconclusive (Semczuk et al., 2004; Shen et al., 1996; Enomoto et al., 1991), though there is data to suggest dysregulation of the canonical Wnt/β-catenin pathway may play a role in the development of granulosa cell tumors (Boerboom et al., 2005). Additionally, an imbalance in chromosomes 4, 9, and 12 have been reported repeatedly in thecomas, suggesting that genes in these regions may contribute to the development of these tumors (Streblow et al., 2007; Liang et al., 2001; Shashi et al., 1994). Evidence for stem cells in these tumors is, however, circumspect at best. Based solely upon morphological and histological factors, reports have implicated putative somatic stem cell involvement in certain ovarian stromal tumor subtypes, such as sertoli-leydig cell tumors in women; however, evidence that would definitively confirm these observations is currently lacking.
5. Ovarian surface epithelium and related malignancies
5.1. Ovarian surface epithelium
The simple squamous-to-cuboid single-layered epithelial cell structure of the normal human ovarian surface epithelium (OSE) belies its complex biology. Several studies have shown that rather than being a passive structure during ovulation, the OSE plays an active role in both follicular rupture and subsequent ovarian remodeling. The fact that OSE can transition back and forth between epithelial and mesenchymal phenotypes has been well-established (Kruk and Auersperg, 1992; Auersperg et al., 1999; Salamanca et al., 2004; reviewed in Ahmed et al., 2007) and this epithelial-mesenchymal transition is believed to be part of the normal process of post-ovulatory ovarian remodeling. Additionally, the OSE has been shown to contribute to repairing the ovarian stroma after ovulation by producing and remodeling components of the extracellular matrix (Kruk and Auersperg, 1992; Kruk and Auersperg, 1994; Auersperg et al., 2001; Salamanca et al., 2004).
Given its ability to differentiate between two cell types and its role in the cyclical disruption and repair that occurs with ovulation, OSE biology seems an intuitive candidate to study in order to understand stem cell mediated processes. We studied the normal OSE as a somatic stem cell source given that repeated ovulation is thought to predispose the OSE, the postulated origin of 90% of ovarian carcinomas, to malignant transformation (Mahdavi et al., 2006). Szotek and colleagues have recently identified a putative somatic stem/progenitor cell in the ovarian surface epithelium (Szotek et al., 2008). A transgenic mouse model of doxycycline inducible green fluorescence protein tagged histones (Tet-on-H2B-GFP) was used; after an initial prolonged pulse, GFP expression or fluorescence can be followed or chased (Tumbar et al., 2004; Brennand et al., 2007). After a chase period of several months, we identified slowly-cycling or quiescent cells in the OSE by retention of label (see Figure 4A), while mitotically active cells which should, by diluting their GFP label with each division, be unlabelled. Using quiescence and label retention as evidence for asymmetric division of the OSE, we then characterized the OSE LRCs by expression of the epithelial markers (cytokeratin-8 and E-cadherin) and the mesenchymal marker, Vimentin, and enrichment in the cytoprotective ABC transporter-associated Hoescht dye-excluding side population (SP), which has been associated with stem/progenitor cells in a variety of tissues and malignancies (Goodell et al., 1996; Jonker et al., 2005; Szotek et al., 2006; Ono et al., 2007; Rossi et al., 2008). When examined for mitotic activity before and after ovulation by iodo-deoxy-uracil (IdU) incorporation, these GFP LRCs were induced to proliferate after ovulation, indicating responsiveness to the estrous cycle (see Figure 4B). Finally, these LRCs showed increased growth potential compared to their non-GFP counterparts in in vitro colony formation assays (see Figure 4C). Furthermore, we observed that these LRCs were consistently found adjacent to CD31+ vascular endothelial cells (Movie; Szotek, unpublished observations). These assays and characteristics collectively point to these cells as the putative somatic stem/progenitor cell of the OSE (Szotek et al., 2008).
Another property that has been attributed to the OSE is the ability post-natally to contribute new follicles to the pool of primordial follicles. Studies published by Bukovsky and colleagues present data in which they contend the tunica albuginea of the post-natal ovary immediately underlying the OSE is capable of producing new primordial follicles (1995; 2004; 2008). Recent studies have reported the isolation of cells from OSE of postmenopausal women and women with premature ovarian failure which express developmental embryonic markers including Oct-4, VASA, Nanog, c-Kit and Sox-2 (Virant-Klun et al., 2008a). Furthermore, the authors claim that these cells are capable of developing embryoid body-, oocyte- and blast-like structures in culture (Virant-Klun et al., 2008b). Though there has been some evidence to suggest the existence of germ line stem cells and post-natal oogenesis, such cells do not appear to ovulate, are not found in the Fallopian tube, and do not contribute to pregnancies (Johnson et al., 2005; Eggan et al., 2006, reviewed in Tilly et al., 2008).
5.2. Epithelial ovarian carcinoma
It is estimated that over 21,000 women will be diagnosed with ovarian carcinoma in 2008 and of these, approximately 15,000 will succumb to their disease (NCI SEER database: see http://seer.cancer.gov/statfacts/html/ovary.html). Survival is directly related to stage, with 5-year survival rates of 92.7% for those diagnosed with localized disease, to 30.6% for those with distant disease at diagnosis (NCI SEER). Unfortunately, two-thirds of patients already have evidence of distant disease at diagnosis (NCI SEER). While the majority of patients respond to initial therapy, most recur with chemoresistant disease. Hence re-evaluation of the current treatment paradigms of ovarian cancers is needed.
Cancer stem cells (CSC), or tumor-initiating cells, are postulated to be specialized cells within tumors which are responsible for propagating cancer growth (Al-Hajj and Clarke, 2004). CSCs are thought to have the ability to give rise to daughter non-tumorigenic cancer cells while retaining their ability to self-renew and form tumors (Al-Hajj and Clarke, 2004). Small populations of clonogenic cells capable of tumorigenesis, self-renewal, differentiation, and chemoresistance in vitro and in vivo have been identified as CSCs in a variety of solid tumors (Al-Hajj et al., 2003; Collins et al., 2005; Dalerba et al., 2007; Li et al., 2007). There are several characteristics of epithelial ovarian carcinomas (EOCs) that indicate that it may be a stem cell-driven disease. Firstly, though the OSE is the postulated source of EOCs (Gondos, 1975; Herbst, 1994; Auersperg et al., 1998), EOCs can generate differentiated subtypes that recapitulate the histology of other normal gynecologic tissues (Landen et al., 2008). Secondly, the high rate of chemoresistant recurrence after initial treatment success suggests that there are cells within the cancer population which are 1) capable of repopulating then entire tumor burden from a small number of cells and 2) exhibit cytoprotective mechanisms thought to exist on somatic stem cells.
Less speculative, experimental evidence for the existence of ovarian cancer stem cells was first reported with the identification and isolation a single tumorigenic clone by anchorage independent spheroid formation from the ascites of a patient with advanced disease (Bapat et al., 2005). The authors then went on to characterize this clone and provide immunohistologic evidence that suggested differentiation along epithelial, granulosa, and germ cell lineages. These clones were shown to form tumors and metastasize in nude mice and retained tumorigenic ability with sequential transplant (Bapat et al., 2005). The authors concluded that their findings suggest that stem/progenitor cell-driven biology may contribute to the aggressive behavior of EOCs.
In a step further, by using the murine transgenic epithelial ovarian cancer cell line MOVCAR-7, produced when the large T antigen is driven by the Müllerian Inhibiting Substance type II receptor promoter (Connolly et al., 2003), Szotek and colleagues identified a putative cancer stem cell within this cell line by using the dye efflux marker SP (Szotek et al., 2006). This subset of cells were found to form palpable tumors when injected into the dorsal fat pad of nude mice, faster and at fewer inoculated numbers than the non-SP cells (see Figure 5A). The SP cells also were cell-cycle arrested and exhibited resistance to conventional chemotherapeutic agents whereas the non-SP cells were and did not (see Figure 5B). We also observed that Müllerian Inhibiting Substance, the protein responsible for the regression of the Müllerian duct during development, was able to inhibit the growth of these SP cells in vitro (see Figure 5C). Additionally, verapamil-sensitive SPs were identified in human ovarian cancer cell lines and in the ascites of a small number of patient, suggesting that the SP could be used to identify the CSC of ovarian cancers in patients though these human cells were not functionally characterized in this study.
In the human, a recent study has identified a subpopulation of putative CSCs from primary human ovarian tumors (Zhang et al., 2008). In this study, ovarian serous adenocarcinomas were disaggregated and grown in conditions selecting for anchorage independent spheroid formation (see Figure 6A). After several passages, purified sphere-forming cells were isolated and found to express various stem cell markers (stem cell factor, Notch-1, Nanog, ABCG2, and Oct-4), demonstrate chemoresistance to ovarian cancer therapeutics, and form palpable tumors in athymic nude mice with inoculation of as few as 100 purified cells (see Figure 6B) compared to no growth with injection of 1 × 106 non-spheroid cells. Further characterization of these cells identified an enrichment for the hyaluronate receptor CD44 and CD117 (c-kit) in the spheroid cells (see Figure 6C). The authors found that, similar to spheroid cells, CD44(+) CD117(+) cells isolated from primary human tumors were able to serially propagate the original tumor at injections of only 100 cells whereas up to 1×105CD44(-) CD117(-) cells formed no tumors. These findings lead the authors to assert that EOCs are derived from a CD44(+) CD117(+) CSC population.
While the initial studies identifying and isolating CSCs in ovarian cancer have yielded valuable insight into the biology of this devastating disease, the therapeutic implications of these findings have yet to be realized. Further studies characterizing the CSC population in a wider population of patients with correlations made to stage at diagnosis, response to treatment, and, ultimately, survival are necessary to take the next step to design therapies targeting these specialized cells. In order to achieve sustained response and convert ovarian cancer to a manageable disease, it is apparent that treatment will need to be patient specific, cancer specific and stem cell specific.
The authors would like to thank Dr. Jose Teixeira for his critical reading of this text, Dr. Paul Szotek for the use of unpublished observations, and Ms. Caroline Coletti for her editorial expertise.
HLC was supported by NIH/MGH T32 in Cancer Biology # 2T32CA071345-11. DTM is supported by NIH/NCI Grant 5R01CA017393-30, the McBride Family Fund, the Commons Development Group, and the Ovarian Cancer Research Fund (New York). PKD is supported by Harvard Stem Cell Institute Grant DP-0010-07-00, Brigham and Women's SPORE Grant 5P50CA105009-03, and NIH/NCI Grant 5R01CA017393-33.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.