Ovary and fimbrial stem cells: biology, niche and cancer origins

The mammalian ovary is covered by a single-layered epithelium that undergoes rupture and remodelling following each ovulation. Although resident stem cells are presumed to be crucial for this cyclic regeneration, their identity and mode of action have been elusive. Surrogate stemness assays and in vivo fate-mapping studies using recently discovered stem cell markers have identified stem cell pools in the ovary and fimbria that ensure epithelial homeostasis. Recent findings provide insights into intrinsic mechanisms and local extrinsic cues that govern the function of ovarian and fimbrial stem cells. These discoveries have advanced our understanding of stem cell biology in the ovary and fimbria, and lay the foundations for evaluating the contribution of resident stem cells to the initiation and progression of human epithelial ovarian cancer.

in defining stem cell identity and function in vivo 22 . Indeed, local niche influences can rapidly convert committed progenitors or even fully differentiated epithelial cells into functional stem cells to facilitate tissue repair followin g injury [22][23][24][25][26] .
Understanding the biology of stem cells of the OSE and of fimbrial epithelia has been hampered by the lack of molecular markers for prospective identification and characterization of stem cell populations. This has been compounded by the inherent difficulties of isolating sufficient epithelia for detailed analyses and by the lack of ex vivo methods for long-term culture of OSE and fimbrial epithelia (for recent reviews, see REFS 23,27,28). Surrogate stemness assays, including long-term DNA label retention and side-population enrichment, have been employed to identify candidate subsets of epithelial cells that display stem-like activities 29-33 . More recently, novel marker genes that were functionally validated by fate-mapping studies were used to document the existence of stem cells that establish the epithelial cell lineages of the ovary and fimbria during development 34 , as well as contributing to lifelong homeostasis and post-ovulatory repair of the ovary 34,35 .
An understanding of the normal biology of stem cells in the ovary and fimbria is crucial for deciphering the mechanisms underlying ovary dysfunction and tumorigenesis in humans. Long-lived stem cells are prime candidates for the gradual accumulation of the mutation cohort that is required to drive initiation of cancer in many adult epithelia 24 , and the same may be true for epithelial ovarian cancer (EOC). The cellular origins

Niche
The supportive local environment in which stem cells reside. The stem cell niche supplies all the extrinsic signalling cues that dictate stem cell maintenance and proliferative activity.

DNA label retention
Introduction of a nucleotide analogue (for example, 5-bromo-2ʹ-deoxyuridine (BrdU) or 5-iodo-2ʹdeoxyuridine (IdU)) or labelling by genetic means (for example, using histone 2B-GFP) for a short period, followed by a prolonged period in the absence of the labelling reagent. After several rounds of cell division, fast-cycling cells rapidly dilute the label, whereas quiescent cells retain most of the original label. These cells are termed label-retaining cells.

Side-population enrichment
A flow-cytometry assay that discriminates for and enriches cell populations that efflux fluorescent dyes (for example, Hoechst 33342) at a higher pace, owing to the expression of ATP-binding cassette transporter proteins within the cell membrane.

Fate-mapping
Permanent labelling of a cell type with a reporter marker (LacZ or a fluorescent protein) that is inherited by its progeny upon cellular division, thus facilitating analyses of its cell-fate decisions and behaviour. Also known as lineage tracing.
of EOC are controversial [36][37][38][39][40][41] . Traditionally, EOC was believed to arise from the ovary itself 36,42 . Accumulating recent evidence, however, has led to a fundamental paradigm shift, according to which EOC may be derived from extra-ovarian sites, notably the fimbria 38,43,44 . These new insights are altering our views of the aetiolog y of EOC and have important implications for future research and preventive therapeutic approaches.
In this Review, we summarize the current knowledge on epithelial stem cell biology in the ovary and fimbria, and the nature of the local stem cell niche components in vivo, and speculate on the cell(s) of origin of EOC. Finally, we discuss current challenges and research directions that may accelerate progress in this relatively immature field.

Structure of the OSE and the fimbrial epithelium
The adult mammalian ovary and fallopian tube (including the fimbria) are lined by a single-layered epithelium that is derived from a common embryonic origin in the pluripotent coelomic epithelium 45 (FIG. 1).
Divergent differentiation during development results in the ovaries and fallopian tubes having distinct histology and phenotypes. In adulthood, the cycling OSE is a simple epithelium that remains incompletely committed and transitions between a squamous and a cuboidal cell shape, depending on positional and oestrus cues (FIG. 1a). By contrast, adult fimbrial epithelia are highly differentiated, assuming specialized columnar shapes comprising ciliated and secretory cell types that are intercalated with rare, basally located cells 31 (FIG. 1b). OSE and fimbrial epithelia express common epithelial (for example, keratins) and mesenchymal (for example, vimentin) markers 1 ; however, their divergence is associated with the selective expression of several epithelial differentiation-specific proteins 31,37,46-49 (FIG. 1).
Despite being generally considered to be anatomically separate, the ovary and fimbria are in fact contiguous at a narrow isthmus. This ovary-fimbria connection is evident in humans (in which it is termed the ovarian fimbriae 37,40 ), non-human primates 46 and mice 34,35 (FIG. 1). In mice, the ovary-fimbria connection is located at the Box 1 | Cyclic epithelial remodelling in the ovary and fimbria Cyclic ovarian surface epithelium (OSE) rupture and repair can be broadly divided into three phases: an ovulatory phase; a rupture phase; and a repair phase. In the ovulatory phase (see the figure, part a), with impending ovulation, the ovulatory follicle becomes hyperaemic as it protrudes above the ovary surface 3,34 . Sequential actions of the pituitary gonadotropins luteinizing hormone (LH) and follicle-stimulating hormone (FSH) trigger the onset of ovulation. The OSE cells and the underlying basal lamina (pink dashed line), together with the tunica albuginea and cell layers of the ovulatory follicle (which include theca cells, granulosa cells and their basement membranes) undergo proteolysis and degradation 156 . Tissue degradation is spatially restricted at the follicular apex. Cultured OSE cells have the autonomous ability to release matrix-digesting proteases, lysosomal enzymes and plasminogen activators and may contribute to follicular disintegration in vivo [156][157][158][159] . In the rupture phase (see the figure, part b), follicular wall disintegration and desquamation of OSE cells at the follicular apex create a wound stigma in the epithelial barrier, through which the cumulus-oocyte complex is released into the periovarian space 160 . In the repair phase (see the figure, part c), wound stigma closure post-ovulation is achieved by cellular proliferation and migration. Increased OSE proliferation has been observed tandem to large follicles 4,161 , the stigma wound periphery 34,161 and post-ovulatory corpora lutea 4 , and at the hilum 35 . Intriguingly, proliferative repair in response to ovulation has not been observed in the primate OSE 3,162,163 . This has led to the notion that the primate OSE is dispensable for post-ovulatory repair 162 . Nevertheless, mechanical OSE ablation studies have shown that the primate OSE is capable of regenerative proliferation 163 . OSE cells may also contribute to the deposition of a new basal lamina and stromal matrix post-ovulation, as exemplified by their intrinsic capacity for de novo synthesis of laminin, collagens (types I, III and IV) and extracellular matrix components in culture 159,164 .
The fimbria also undergoes cyclic remodelling. In humans, there is evidence to suggest that fimbrial epithelium proliferates 165 and becomes pseudo-stratified at the follicular phase (pre-ovulation) before reverting to a single layer at the luteal phase (post-ovulation) 166 . hilum region and is abridged by the ovary ligament, which connects the ovary to the uterus and contains blood vessels and nerves entering the ovary 34 . This ovary ligament is also lined by a single-layered epithelium, which establishes an OSE-ligament-fimbria epithelial continuum (FIG. 1c). Whether a ligament continuum also exists in other species remains unclear.

Identity of stem cells in the ovary and fimbria
The adult ovary is composed of germ cells (oocytes or oogonia) and somatic cells (OSE, granulosa and theca cells) that are supported on a stromal mesh (tunica albuginea and interstitial cells). Although the existence of stem cell pools that establish and maintain germline and ovarian somatic cell lineages has been described,  Figure 1 | The adult human and mouse ovary and associated tissues. The diagrams show the organization and anatomy of the adult human and mouse ovary; the fallopian tube (also known as the oviduct), including finger-like fimbria; the ovary ligament; and the uterus. A thin bursa membrane encapsulates the mouse ovary but is absent in humans. The placement of the ovary ligament differs between humans and mice. In humans, the ovary and the fimbria are anatomically contiguous at the ovarian fimbriae, whereas the ovary ligament abridges the ovary-fimbria connection in mice. A single-layered epithelium lines the ovary, fimbria and ovary ligament. a | The ovary is entirely encapsulated by the ovarian surface epithelium (OSE), which is a single-layered simple epithelium. Compared with other types of epithelial cell, the cells of the OSE are uncommitted and express both epithelial and mesenchymal markers. Cells of the OSE adopt cuboidal or squamous cell shapes, depending on oestrus cues and on their proximity to a follicle or to a corpus luteum, respectively. b | The adult fimbrial epithelium comprises highly differentiated columnar ciliated and secretory epithelial cells, and basally located cells that contact the basement membrane. c | In mice, an OSE-ligament-fimbria epithelial continuum exists at the ovary hilum; whether this continuum is also present in other mammalian species is unknown. The table lists epithelial differentiation-specific markers that are expressed in the OSE 37,46 , in fimbria 31,37,46-49 or in both 31,37 .

Cellular quiescence
A reversible, non-dividing cell state. Some stem cells are quiescent under steady state, a property that is believed to sustain lifelong tissue maintenance and preservation of the stem cell compartment.
their identity and precise locations are still the subject of much debate.

Germline stem cells.
A long-standing dogma posits that female mammals are born with a fixed ovarian follicle reserve. Several studies 50-55 have challenged this dogma by finding that the OSE monolayer in adult mice and humans contains putative germline stem cells. These cells expressed the oocyte-specific mouse VASA homologue MVH (also known as DDX4), could self-propagate in vitro and generated new primordial follicles following their xenograft onto donor ovarian tissue in vivo. Independent studies have failed to reproduce these seminal findings 56-59 , and the concept of postnatal oogenesis has yet to gain universal acceptance. However, the study of potentially persisting oogenesis in adult female mammals is an exciting area of reproductive research that holds tremendous promise for the treatment of infertilit y pathologies.

Granulosa and theca stem cells.
No stem cell activity has been detected within adult granulosa or theca cell lineages. However, early work has described a small proportion of adult granulosa cells that display stemlike characteristics in culture, including the ability to divide and form colonies without substrate anchorage 60 . More recently, in vivo fate-mapping studies have revealed two distinct somatic subpopulations within embryonic mouse ovaries that establish the granulosa lineages of the adult medullary and cortical follicles 61,62 . Putative theca precursors have also been identified in the ovaries of newborn mice; these cells differentiated into mature steroidogenic cells in vitro and rapidly invaded the theca layers of follicles within the host ovary post-transplantation 63 .
The first evidence of putative stem cells on the ovary surface came in 2008, when 5-bromo-2ʹ-deoxyuridine (BrdU) or 5-iodo-2ʹ-deoxyuridine (IdU) and doxycyclin e-inducible histone 2B-GFP pulse-chase surrogate stemness assays were used to identify a subset of stem-like cells in the adult mouse OSE 29 (FIG. 2). These OSE cells displayed cellular quiescence, DNA label retention and enhanced in vitro colony formation 29 . The label-retaining cells also displayed Hoechst 33342 dyeeffluxing cytoprotective capacity, a trait that is believed to be specific to stem cells, owing to their intrinsic ability to have efflux pumps on their membrane. The location of these label-retaining OSE cells, juxtaposed at ovulatory follicles, suggests that they may participate in post-ovulatory wound closure. Subsequently, a subset of verapamil-sensitive OSE cells expressing lymphocyte antigen 6A2-6E1 (LY6A; also known as SCA1), which is a common haematopoietic stem cell marker, with Hoechst-effluxing capability was identified in adult mouse ovaries 32 (FIG. 2). In culture, Ly6a + cells had enhanced sphere-forming abilities and were responsive to transforming growth factor-β1 (TGFβ1) and l eukaemia-inhibitory factor (LIF), two growth factors that are abundant in follicular fluid 32 . A role for these label-retaining and Ly6a + OSE cells in epithelial regeneration in vivo, however, has not been formally established.
More recently, two studies 34,35 used in vivo fatemappin g methodologies to provide direct evidence for the existence and locations of self-renewing epithelial stem cells in the adult mouse ovary. Fluorescence-activated cell sorting was used to isolate a subpopulation of OSE cells located at the ovary hilum that displayed high ALDH activity 35 (FIG. 2). ALDH activity, which confers drug resistance and cell protection, has been widely used to identify potential stem cells in various tissues 66 . In culture, Aldh1 + hilum OSE cells generated large epithelial spheres at higher frequency and exhibited enhanced colony formation compared to their Aldh1counterparts. Expression analyses showed that Aldh1 + cells expressed Leu-rich repeat-containing G protein-coupled receptor 5 (LGR5), a seven-transmembrane receptor that is a facultative component of the WNT receptor complex and is expressed by stem cells in various epithelial tissues 8,9,16,[67][68][69] . Lineage tracing with an Lgr5-Cre-knock-in mouse model expressing enhanced GFP (Lgr5-egfp-ires-CreER T2 ) led to the conclusion that hilum cells expressing ALDH1 and LGR5 are the major OSE stem cell reservoir that supplies epithelia l cells to the entire ovary surface in vivo (BOX 2).
Subsequent experiments using the same Lgr5-Creknock-in allele 34 challenged the existence of an exclusive Aldh1 + Lgr5 + stem cell compartment in the ovary hilum. By combining Lgr5 reporter gene profiling and endogenous single-molecule mRNA fluorescence in situ hybridization (FISH) expression analyses, these studies reported multiple clusters of Lgr5 + cells located at interfollicular cleft regions throughout the ovary surface, in addition to the population at the hilum (FIG. 2). Longterm lineage marking analyses indicated that both hilum and extrahilum Lgr5 + cell populations were capable of generating phenotypically distinct epithelial lineages over the 16-month reproductive lifetime of the mouse, thus formally establishing the entire Lgr5 + cell pool as OSE-resident stem cells .
Whether the ovary epithelium contains additional stem cell subsets with distinct lineage features remains an open question. Definitive answers to such questions would require the discovery of specific genes that mark putative stem cell pools for isolation and downstream characterization, using clonal marking techniques and/ or in vivo conditional gene ablation strategies to establish stem cell identity.
Interestingly, human OSE cells express LGR5 (REF. 34), suggesting that putative LGR5 + stem cells are present in the ovaries of other mammalian species. Future experimentation relies on a validated antibody against human LGR5 that can be used to purify and characterize LGR5 + stem cell activity using functional assays, such as in vivo regeneration following transplantation and in vitro sphere-forming assays.
Organoids Three-dimensional cell culture structures that recapitulate the multipotent cellular differentiation and functional complexity of the native tissue of origin.

Fimbrial stem cells.
Fate-mapping studies that formally document somatic stem cells in the adult mouse fimbria are currently lacking.
LGR5 is not expressed in the adult mouse fimbria (although human fimbria highly expresses LGR5) 34 . Furthermore, fate-mapping of Lgr5 + epithelial cells at the ovary-ligament-fimbria junction showed that these cells did not contribute to in vivo homeostasis of the fimbria 34 , suggesting the existence of as-yet undiscovered stem cell populations in this tissue.
Several studies have described stem-like epithelial cells in the fallopian tube in humans 31 and in the oviduct in mice 30,33 , which were concentrated at the fimbriated end (FIG. 2). In mice, two phenotypically distinct populations of stem-like fimbrial epithelial cells, which were distinguished by positive 33 and negative 30 expression of mast-stem cell growth factor receptor KIT, were identified on the basis of long-term label-retention assays. In culture, Kit − label-retaining cells, which did not express oestrogen or progesterone receptors (ERa − PR − ) formed self-renewing organoids that contained mature epithelia that are normally present in the native distal (ERa + PR − ) and proximal (ERa + PR + ) oviduct 30 . Remarkably, the organoids also expressed progestogen-associated endometrial protein (PAEP; also known as glycodelin), which is a marker of mature endometrium 30 . This suggests that the fimbrial label-retaining cells had considerable plasticity ex vivo. In humans, fimbrial stem-like cells were identified as tubulin β4 (TUBB4) − , paired box 8 (PAX8) − , CD44 + basally-located cells 31 that were capable of generating in vitro spheres that contained differentiated Only the Lgr5 + populations located at the hilum and extrahilum regions have been formally established to contribute to in vivo OSE homeostasis and post-ovulatory regeneration 34 . The inset in the Lgr5 + box is a schematic representation of clonal fate-mapping using the ROSA4 colour lineage reporter 34 , which demonstrated the involvement of several Lgr5 + stem cells in post-ovulatory wound repair. Each coloured (red, RFP; yellow, YFP; cyan, CFP, membranous) clone represents the progeny of a single Lgr5 + OSE stem cell. BrdU, 5-bromo-2ʹ-deoxyuridine; CK6B, cytokeratin 6B; EPCAM, epithelial cell adhesion molecule; ERa, oestrogen receptor α; H2B, histone 2B; IdU, 5-iodo-2ʹ-deoxyuridine; ITGA6, integrin α6; K5, keratin 5; LEF1, lymphoid enhancer-binding factor 1; LIF, leukaemia inhibitory factor; Paep, progestogen-associated endometrial protein; PR, progesterone receptor; TGFβ1, transforming growth factor-β1.

Early Lgr5 + stem cells in somatic lineage specification.
Expression of LGR5 was first detected in surface and subsurface somatic cells in the embryonic day E12.5-E13.5 ovary anlagen, coincident with female sex determination 34, 61 . In vivo lineage tracing during this embryonic period showed that early Lgr5 + ovarian cells possessed lineage-differentiation capabilities, contributing to both the future OSE 34 and granulosa cell lineages 61 . Granulosa contribution by early Lgr5 + cells was restricted to the pre-granulosa lineage of adult cortical follicles, which is responsible for folliculogenesis throughout life 61 . A distinct forkhead box L2 (Foxl2) + Lgr5 − pool established the adult medullary follicles 62 . In the developing Müllerian duct, expression of LGR5 was restricted to the anteriormost region (that is, the primordia of the oviduct, including the fimbria). These early Lgr5 + cells contributed to the epithelia of the adult oviduct, including the fimbria, as well as the ovary-ligament-fimbria junction at the ovary hilum 34 . Intriguingly, examination of Lgr5-knockout female gonads has shown that OSE and granulosa cell lineage differentiation proceeds normally in the absence of LGR5 (REF. 61). Instead, Lgr5-mutant ovaries had variable degrees of defects in the differentiation of germ cells, which do not normally express LGR5 (REF. 61). It is plausible that the closely related LGR4, which is also expressed in ovarian somatic cells 61,70 , may provide functional redundancy with LGR5 during ovary organogenesis.

OSE and fimbrial stem cell regulation
The behaviour of epithelial stem cells is governed by a combination of intrinsic programmes and extrinsic stimuli derived from the local niche. Perturbation of these regulatory cues is instrumental in driving formation of epithelial cancer in many tissues, including the intestine 8 , stomach 9 and skin 71 .
Currently, little is known about the extrinsic regulation of stem cell fate (for Lgr5 + cells and other potentially undiscovered cells) in the ovary and fimbria. However, a large body of work has characterized Lgr5 + stem celldriven epithelial systems in other tissues. Insights gained from these studies may facilitate the identification of stem cell niche components and regulatory signals in the ovary and fimbria and uncover similarities between different epithelia.

The WNT-enriched Lgr5 + stem cell transcriptome.
Comparative microarray analyses have shown that Lgr5 + and Lgr5 − cells have distinct transcriptional profiles 34 . Consistent with the role of WNT signalling in the maintenance and renewal of adult epithelial stem cells 20,68,[72][73][74] , Lgr5 + OSE stem cells are enriched in expression of several WNT signalling components, including the WNT ligand WNT4, as well as WNT targe t genes axis inhibition 2 (Axin2) and the gene encoding tumour necrosis factor receptor family member Troy (Tnfrsf19) 34 (FIG. 3a). Activation of WNT signalling relies on the binding of secreted WNT ligands to their Frizzled receptors and to low-density lipoprotein receptor-related protein 5 (LRP5)-LRP6 co-receptor complexes on the cell surface. This leads to stabilization and entry of β-catenin into the nucleus to induce transcription of target genes through interactions with the T-cell factor (TCF) and lymphoid enhancer factor (LEF) transcription factors. Thus, the ability to produce their own WNT4 ligands indicates that Lgr5 + OSE stem cells have an intrinsic ability to regulate their own fate through an autocrine WNT signalling loop, as has been recently demonstrated for the interfollicular epidermis 20 .
A finely balanced WNT signal strength is crucial to achieving optimal stem cell function. For instance, overactive WNT signalling in mice leads to adenoma and tumour formation in the gastrointestinal tract and Box 2 | Ovulatory re-epithelialization by Lgr5 + ovarian surface epithelium (OSE) stem cells in mice In mice, follicular rupture can occur anywhere on the ovary surface (whether this is also true in mono-ovulating species such as humans or primates remains unexplored). An aldehyde dehydrogenase 1 (Aldh1) + Leu-rich repeat-containing G protein-coupled receptor 5 (Lgr5) + stem cell pool restricted to the hilum would require massive cell proliferation and extensive migration to repair distant ovulatory stigma wounds. This would be no easy feat, given that surface barrier continuity is typically restored within days following ovulation in the mouse 2,4 . Increased proliferation in OSE cells around the ovary hilum, which is suggestive of participation in post-ovulatory wound repair, has been reported 2,167 , although this was not confirmed by other studies 168,169 . There was also no significant difference in the proliferative activity of Lgr5 + stem cells at the hilum compared to those at other regions of the ovary 34 .
Epithelial restoration occurring in a fast and spatially localized manner intuitively seems to favour a model of post-ovulatory repair by local, wound-adjacent stem cells. The widespread distribution of Lgr5 + stem cells throughout the mouse ovary surface is consistent with a model of rapid and effective re-epithelialization by these cells at the wound perimeter. Supporting this notion is the observation that Lgr5 + OSE stem cells at wound stigmas were highly proliferative 34 . During regular tissue homeostasis (that is, in the absence of ovulation), Lgr5 + stem cells were relatively quiescent and had a slow cellular turnover. This is consistent with previous reports of negligible levels of proliferation within non-ovulating regions in various species 2,3,159,163,168 . In this context, widespread proliferation of the OSE does not seem to be required during homeostatic expansion of the ovary surface associated with folliculogenesis. In response to local ovulatory damage, however, Lgr5 + stem cells at OSE sites flanking ovulatory follicles were activated to contribute progeny to repair the subsequent stigma wound 34 . Clonal fate-mapping using the ROSA4 colour lineage reporter has documented the recruitment of multiple Lgr5 + OSE stem cells during the re-epithelialization process (FIG. 2). Comprehensive clonal lineage-tracing studies carried out in conjunction with real-time cell tracking should help to establish the population kinetics and migration patterns of Lgr5 + stem cells onto the rupture wound during homeostatic repair.
the stomach 8,9 , and de novo hair follicle morpho genesis and formation of hair follicle tumours 71 , whereas inhibition of WNT causes depletion of intestinal stem cell and crypt compartments 74 . Although it remains unclear how internal WNT signals within Lgr5 + OSE stem cells are regulated, Lgr5 (which is a WNT target gene) is likely to bind to its secreted R-spondin ligands (RSPO1 to RSPO4) to augment the OSE WNT signals that are initiated by WNT ligands 75 . At the same time, Troy, which is a negative regulator of WNT signalling within Lgr5 + stem cells in the intestine and the gastric corpus 76,77 , might fine-tune the signal strength of WNT and RSPO signalling to restrict the OSE stem cell zone.
Robust WNT-RSPO signalling has been shown to be important during embryonic development of the ovary anlagen. Females lacking Wnt4 or Rspo1 were born with oocyte-depleted masculinized ovaries, indicative of an aberrant male phenotypic pathway during ovarian somatic cell differentiation [78][79][80] . The physiological role of Troy could not be ascertained from loss-of-function studies, as mice deficient for Troy were viable and fertil e, displaying no abnormalities in their ovaries 81,82 .
Whether a similar WNT-intrinsic programme regulates adult fimbrial stem cell function remains largely unexplored. Active canonical WNT signallin g has been reported in the adult mouse oviduct 83 . Candidate regulatory niche signals emitted from the mature epithelial progeny (epithelial niche) are depicted, as well as the ovulatory follicle with its follicular fluid, and surface stroma (non-epithelial niches). In the inset, presumptive stem-niche units comprising OSE cell clusters and the underlying follicular structure (follicle or corpus luteum) on the ovary surface are indicated by orange arcs; potential stem cell niches at the ovary hilum are indicated by magenta arcs. b | The small intestinal stem cell niche is defined by geometrical arrangement of stem cells (dark pink) and their differentiated Paneth cell progeny surrounded by mesenchyme at the base of the crypts. Paneth cells and mesenchyme function as intestinal stem cell niches. c | The hair follicle stem cell niche is composed of stem cells (dark pink) residing at the outer bulge and hair germ, supported by niche signal-producing secondary hair germ, dermal papilla and stem cell progeny in the inner bulge. ALDH1, aldehyde dehydrogenase 1; AXIN2, axis inhibition 2; BMP, bone morphogenetic protein; DLL, delta-like protein; EGF, epidermal growth factor; ER, oestrogen receptor; FGF, fibroblast growth factor; FSH, follicle-stimulating hormone; FZD, Frizzled; HGF, hepatocyte growth factor; KGF, keratinocyte growth factor; KL, KIT ligand; LGR, Leu-rich repeat-containing G protein-coupled receptor; LH, luteinizing hormone; PR, progesterone receptor; SFRP, secreted Frizzled-related protein; TGFβ, transforming growth factor-β.

Corpus luteum
The vascularized by-product of a recently ruptured follicle. Over time, the corpus luteum regresses into the centre of the ovary to become stromal or interstitial tissue.

Epithelial-mesenchymal transition
(EMT). A biological process by which epithelial cells dissolve their tight junctions with one another and convert into a free, migratory form.
Moreover, human pathological studies have suggested a correlation between aberrant WNT signalling activation and altered fallopian tubal homeostasis 84 . Together, these findings support a role for WNT in fimbrial stem cell specification. As in the ovary, WNT-RSPO signalling promotes embryonic Müllerian duct development. Mouse Wnt4-mutants lacked a Müllerian duct at birth, in addition to having ovary defects 78 , whereas inactivation of Wnt9b 85 or Wnt7a 86 led to absence of the duct or partial elongation of the Müllerian duct.
Putative niche components. Niche signals can be supplied by epithelial (that is, the differentiated stem cell progeny) and/or non-epithelial sources 87 . As niche signals typically function over short distances 88 , resident stem cells are often located close to their niche cells for optimal niche-to-stem-cell signal transmission. The base of the intestinal crypts of Lieberkühn, for instance, comprises discrete niche units containing multipotent Lgr5 + stem cells that are intercalated with their antimicrobial peptide-secreting Paneth cell progeny and surrounded by mesenchyme (for a recent review, see REF. 24). Paneth cells and mesenchyme represent crucial intestinal niche sources, secreting essential stemness signals to balance stem cell maintenance and differentiation ex vivo and in vivo 75,[89][90][91][92][93] (FIG. 3b). Similarly, each hair follicle represents a discrete stem cell niche entity. Multipotent Cd34 + K15 + Lgr5 + stem cells 16,18,94 reside within the lower bulge and the hair germ and rely on a range of stem cell activation and quiescence signals from the inner bulge, hair germ and dermal papilla to achieve homeostatic hair follic le regeneration (for recent reviews, see REFS  Using close proximity as a criterion to define stem cell niches, candidate niche components of the ovary are the mature epithelial progeny of stem cells (epithelial niche), as well as the underlying follicular structure (the follicle or ovulatory by-product, known as the corpus luteum, and its follicular fluid) and the surface stroma at the OSE interface (non-epithelial niches) (FIG. 3a). The fimbria probably shares some of the OSE niche components, given its proximity to the ovary surface during ovulation. Although this is speculative, it is thought that the ovary surface may contain several interspersed niche units comprising clusters of OSE cells and their underlying follicular structures (FIG. 3a). Within each unit, OSE stem cells respond to the unique set of paracrine cues derived from their epithelial and non-epithelial niche components, to achieve the appropriate self-renewal and differentiation. A potential Aldh1 + Lgr5 + stem cell niche at the mouse ovary hilum has also recently been identified 35 (FIG. 3a).
Exactly which WNT or WNTs are the endogenous WNT-modulating niche signals for OSE stem cell activity is currently unknown. Loss-of-function mouse models have failed to provide direct evidence of this: Wnt2-null mice have placental defects, and survivin g females do not show an ovary defect 102 , probably because of the compensatory function of other WNTs. Furthermore, conditional ablation of WNT4 in adult granulosa cells (using an Amhr2-Cre driver) efficiently blocked folliculogenesis, but a detailed analysis of the OSE was lacking 97 .

TGFβ and bone morphogenetic protein (BMP) niche signals.
Secreted TGFβ and BMP is another candidate paracrine niche signal. The biological effects of TGFβ and BMP signalling are mediated by ligandreceptor interactions at the cell surface, leading to nuclear translocation of SMAD transcription factors to regulate stemness-promoting genes. TGFβ signalling within the niche has been implicated in epithelial stem cell quiescence and differentiation, as well as induction of epithelia l-mesenchymal transition (EMT) 103 . EMT has also been reported to increase expression of stemness markers in human mammary epithelial cells 104 . Interestingly, EMT involvement in post-ovulatory wound repair is supported by evidence that artificially generated scratch wounds were healed by migratory OSE cells in the presence of the EMT inducer pro-epiderma l growth factor (EGF) 105 .
BMP4 or TGFβ1 has been shown to induce EMT of OSE cells in vitro 106,107 . Adult OSE cells, including Lgr5 + OSE stem cells, express all known TGFβ and BMP receptors (BMPR1A, BMPR1B, BMPR2, TGFβRI and TGFβRII) and downstream signalling effectors (SMAD1, SMAD5 and SMAD8, and the co-SMAD SMAD4), indicating that the ovary epithelium is a target of TGFβ and BMP signalling 34, [108][109][110] (FIG. 3a). Furthermore, disruption of SMAD3 leads to OSE hyperproliferation 111 . Candidate paracrine BMP signals emitted from the follicles include BMP2 and BMP7, the expression of which in granulosa cells and theca cells, respectively, peaked at ovulation, before rapidly diminishing within lutein cells post-ovulation 112 (FIG. 3a). Also of potential importance is the expression of all three TGFβ isoforms (TGFβ1, TGFβ2 and TGFβ3) at the surface stroma at the OSE interface 108,109 (FIG. 3a). These secreted molecules might gain direct access to the adjacent OSE stem cells following breakdown of OSE basemen t membranes during ovulation.

Cancer cell of origin
The cell that sustains the first cancer-promoting mutation or mutations that initiate tumour development.

Follicular fluid milieu.
The follicular fluid is a rich source of growth factors, steroid hormones and gonadotropins, which regulate growth and differentiation (FIG. 3a). Expulsion of follicular fluid at the stigma wound that is created during ovulation potentially brings the regulatory milieu within signal-receiving range of OSE stem cells at wound margins. A centripetal morphogenetic gradient of follicular fluid milieu arising from the wound epicentre probably establishes the boundaries to facilitate spatially targeted re-epithelialization at the follicular apex, whereas wound-distant regions remain unperturbed and maintain stem cell quiescence. Expulsed follicular fluid may also bathe the fimbrial fringes at the ovulatory rupture site (FIG. 3a).
During follicular maturation, the composition of the follicular fluid milieu, and hence the nature and strength of potential niche signals, dramatically changes. Among the growth factors that are present in high concentrations in ovulatory follicular fluid, EGF and TGFα 105,[113][114][115] , hepatocyte growth factor 116,117 , keratinocyte growth factor 116,118 and KIT ligand 116 exerted the most potent mitogenic effects on OSE cells in vitro. Oestrogen, which is an OSE mitogen, also reaches peak levels in the follicular fluid of ovulatory follicles. A dramatic shift from oestrogen to progesterone production occurs post-ovulation in the corpus luteum 119 . As progesterone has been shown to exert a growth-inhibitory effect on cultured OSE cells 119 , this steroid hormone may be an important endogenous niche signal to prevent OSE overproliferation. Large amounts of the gonadotropins luteinizing hormone (LH) and follicle-stimulating hormone (FSH) are deposited into the ovulatory follicular fluid by the blood circulation; these long-range signals are known to stimulate OSE proliferation in vitro and in vivo 4,120,121 . Importantly, the cell-surface receptors for all of these putative growthstimulating and growth-inhibitory regulatory factors and hormones are present in the OSE and in the fimbrial epithelium 3,117,[122][123][124][125] . In sheep, expression of FSH and LH receptors was highest in OSE overlying large follicles 123 , suggesting that gonadotropins probably exert the strongest growth-stimulating effects in receptor-enriched OSE cells on top of ovulatory follicles.

Dissection of niche regulatory mechanisms in vivo and in vitro.
Definitive proof of the existence of the OSE and fimbrial stem cell niche, and its characterization, awaits genetic studies to decipher the mechanisms that control the OSE and Lgr5 + stem cells. In vivo ablation strategies targeting the removal of candidate niche cells will also provide direct evidence of a role for the niche in maintaining Lgr5 + stem cell activities. However, the lack of Cre lines to precisely modulate these signalling pathways in the ovary remains a major technical barrier to such studies.
Recently, advances have been made in growing ovary fragments, whole ovaries or oviductal tissues with intact basal lamina and follicular structures in an alginate hydrogel matrix [126][127][128] or as multicellular spheroids 129,130 . Although the use of these models has facilitated analyses of the function of stem cells within their native microenvironments, the main limitation of current 3D culture systems is their inability to sustain long-term ovary growth and follicle maturation, which precludes the study of ovulatory re-epithelialization. An alternative approach is the development of near-physiological Matrigel-based 3D ex vivo culture systems that are capable of sustaining the long-term growth of functional OSE and fimbrial epithelia, for studying the regulation of stem cell behaviour in real time. Such culture systems have achieved considerable success for the intestine 75,89 , the stomach 9 , the colon 131,132 and the mammary gland 133 . 3D-based culture studies are particularly useful for real-time tracking of Lgr5egfp + OSE and fimbrial stem cells, combined with clonal fate-mapping strategies in a supportive in vitro environment. Aspects of Lgr5 + stem cell behaviour -mode and rate of cell division, self-renewal and migration properties -during regular tissue homeostasis versus in response to ovulatory wound repair could easily be studied, as could the effects of specific regulatory growth factors and hormonal influences on Lgr5 + stem cell characteristics and function. It is therefore of paramount importance to optimize culture conditions for the long-term growth of near-physiological OSE and fimbrial epithelium in vitro.

Stem cells as putative EOC cell(s) of origin
EOC comprises four phenotypically distinct subtypes, including serous, endometrioid, mucinous and clear cell carcinomas. Depending on their clinicopathology and molecular profile, these EOC subtypes can be classified as either type I or type II tumours. Slow-growing type I tumours, which encompass low-grade serous, low-grade endometrioid, clear cell and mucinous carcinomas 41 , are often preceded by a well-recognized OSE precursor intermediary in benign or borderline tumours 42 . They are defined by several somatic mutations, including mutations in the genes encoding β-catenin, PTEN, PIK3CA, AT-rich interactive domain-containing protein 1A (ARID1A), GTPase KRAS, Ser/Thr protein kinase BRAF and Ser/ Thr protein phosphatase PPP2R1A 41 . By contrast, type II tumours, comprising high-grade serous, high-grade endometrioid, undifferentiated and malignant-mixed carcinomas 41 , are aggressive and are invariably detected at advanced stages when the disease has spread beyond the ovaries. High-grade serous carcinomas (HGSCs), which are the most prevalent type II tumours, have a unique genetic fingerprint involving mutations in the tumour suppressor gene p53, as well as alterations in components of the breast cancer type 1 susceptibility protein (BRCA1)-BRCA2, RB1, RAS-PI3K and NOTCH signalling pathways and the forkhead box protein M1 (FOXM1) transcription factor network 134 .
The identity of the transformed EOC precursor(s) remains controversial. The central question is whether a single precursor cell generates the entire EOC spectrum, or whether each EOC subtype is derived from a distinct cell and tissue source. Although the long-held paradigm posits that EOC is the result of the transformation of an OSE cell, recent compelling evidence has suggested that a mutant fimbrial cell may be the cancer-initiating culprit for some EOCs, notably HGSCs. This OSE versus fimbrial cancer cell of origin controversy (BOX 3) has been extensively reviewed elsewhere 36-41,43,135 . Genetically engineered mouse models for the in vivo transformation of OSE 35, [136][137][138][139][140][141][142][143][144]145 or fimbria 48,146-148 support a paradigm in which both tissues are potential sources of ovarian carcinogenesis.
Epithelial stem cells have been shown to be cells of origin in many human epithelial cancers. Their longevity allows for the accumulation of the range of genetic mutations that are needed for perturbation of growth controls leading to metaplasia and malignant transformation. The plasticity of transformed stem cells may also account for the phenotypic heterogeneity present in EOC.
Several epithelial cancers have been shown to originate from normal stem cells. For example, constitutive activation of the WNT pathway through conditional deletion of the adenomatous polyposis coli tumour suppressor in Lgr5 + stem cells (but not in their t ransitamplifying and differentiated progeny), or expressio n of the oncogenic β-catenin in Polycomb complex protein BMI1-and prominin 1-expressing stem cells, resulted in epithelial cell transformation and rapid formation of adenomas in the mouse small intestine [149][150][151] and stomach 9 .
Recently, meta-analysis of human ovarian cancer expression microarrays has revealed an EOC tumour cluster, designated stem-A (proliferative) and stem-B tumours, that expressed markers typical of epithelial stem cells, notably LGR5 and CD133 (also known as prominin 1) 152 . Thus, EOC may have stem-like components. Of note, stem-A tumours were linked with poor clinical prognosis and with high expression of p roliferationrelated genes 152 and several WNT-planar cell polarity (PCP) pathway genes, including the WNT signalling receptor gene FZD7, the knockdown of which decreased cell proliferation and migration 153 .
From this perspective, the recent identification of LGR5 + cells in healthy OSE and fimbrial epithelium in humans has led to speculation regarding their potential role as the EOC cell of origin 34 . A stem-driven cancerinitiating scenario may be proposed, by which an initiating mutation that is sustained by a stem cell (LGR5 + and/or others) leads to its neoplastic transformation and differentiation along several Müllerian tumour lineages (FIG. 4).
Also of potential importance in EOC genesis is the identification of an enriched Aldh1 + and Lgr5 + stem cell population at the ovary-fimbrial border in the mouse ovary hilum 35 . Compared to their differentiated progeny, OSE cells of the ovary hilum show enhanced proliferation in vitro and increased transformation potential in vivo following inactivation of tumour suppressor genes transformation related protein 53 (Trp53) and Rb1, which are often lost in human HGSCs 35 .
In humans, the ovary-fimbrial border (termed the ovarian fimbriae) represents an epithelial transitional zone with overlapping OSE and fimbrial marker expression 36,37,40 and a robust stem cell programme 65 . It is plausible that regulation of stem cell fate may be less defined in this region, making the cells more susceptible to carcinogenesis. Furthermore, we speculate that combined niche signals emitted by the OSE and fimbrial niches enable resident stem cells to differentiate into OSE and Müllerian tumour lineages. Epithelial transitional zones have been implicated as major sources of many epithelial cancers. For example, a small population (~40 cells) of cuboidal epithelial cells at the squamocolumnar junction of the uterine cervix is believed to be the target precursor for cervical carcinogenesis 154 , whereas Barrett's oesophagus, which is a precursor of oesophageal adenocarcinoma, is thought to originate from the embryonic-like squamous epithelium at the oesophagus-stomach border 155 .
The availability of the Lgr5-Cre mouse model should make it an ideal in vivo platform for evaluating the behaviour of ovary-resident Lgr5 + OSE and ovary-fimbrial transitional epithelia following conditional targeting of key EOC-associated oncogenes and/or tumour suppressor genes. Such a model will directly assess whether transformation of an Lgr5 + stem cell is the initial transformation event that leads to epithelial precancerous lesions and ultimately to EOC.

Future directions and outlook
In contrast to many adult epithelia, our understanding of stem cell biology in the ovary and fimbria has remained frustratingly poor. Until recently, we could only

Box 3 | Ovary versus fimbria as epithelial ovarian cancer (EOC) source
Traditionally, the ovarian surface epithelium (OSE), or its cortical inclusion cyst, is regarded as the sole EOC source (FIG. 4a). The incessant ovulation hypothesis posits that the cyclic rupture and repair trauma that is endured by OSE increases cell proliferation and consequently accumulation of deleterious somatic mutations 170 . Cortical inclusion cysts form in the superficial ovarian cortex as a result, and exposure of the entrapped cyst-lined OSE cells to the activated stromal milieu 42 and OSE-secreted milieu within the cystic lumen 171 causes their transformation (FIG. 4a). Unlike many epithelial cancers, in which carcinogenesis is accompanied by a block in differentiation, the ovary epithelium undergoes metaplasia to acquire highly complex histology that resembles Müllerian duct-derived fallopian tube (serous EOC), endometrium (mucinous EOC), endocervix (endometrioid EOC) or vagina (clear cell EOC) 1 (FIG. 4a). Consistent with the capability of OSE to give rise to the wide EOC histological spectrum, ectopic expression of the Müllerian-expressed homeobox genes Hoxa9, Hoxa10 and Hoxa11 induced immortalized OSE cells to differentiate along Müllerian lineages in vitro 172 , and subsequent intraperitoneal inoculation of these transformed cells generated tumours resembling serous, endometrioid and mucinous EOCs, respectively 172 . Excessive gonadotropin stimulation and repeated exposure to the inflammatory factors within follicular fluid were subsequently identified as factors contributing to EOC development 135 .
Failure to identify convincing EOC precursor lesions in the OSE has led to speculation that these carcinomas either arise de novo from epithelial inclusion cysts without an intermediary lesion 42 or are derived from an extra-ovarian source or sources.
An alternative hypothesis posits that many EOCs, notably high-grade serous carcinomas (HGSCs), are derived from the fimbria. Pre-neoplastic p53 signaturecontaining lesions termed serous tubal intraepithelial carcinomas (STICs) have been identified in the fimbria of carriers of breast cancer 1 (BRCA1) and BRCA2 mutations predisposed to ovarian cancer and in patients with sporadic pelvic-ovarian HGSC, whose ovary histology was otherwise normal 44, [173][174][175][176][177] . Secretory fimbrial cells have an intrinsic delayed response to DNA damage compared to their ciliated counterparts 178 . This trait might make secretory fimbrial cells vulnerable to the accumulation of deleterious mutations in their genomes following repeated exposure to the pro-inflammatory follicular fluid milieu. The current model of fimbrial involvement in HGSC genesis (FIG. 4b) proposes ectopic implantation of abnormal STIC fimbrial cells into the ovary stroma as cortical inclusion cysts through the ovulatory rupture stigma site 39 ; analogous to the entrapped-OSE model, exposure of the implanted fimbrial epithelia to an unfavourable ovary microenvironment leads to HGSC. As a subset of HGSC cases showed no fimbrial STIC involvement 173,179 , the entrapped-OSE model might still apply.
speculate on the existence of somatic stem cells involved in post-ovulatory wound repair and remodelling in these tissues. With the discovery of LGR5-marked stem cells in the ovary, it is now possible to investigate how these stem cells execute wound repair and maintain ovary homeostasis. Future efforts should focus on characterizing the OSE stem cell niche components and signals, as well as identifying stem cells that are responsible for lifelong homeostasis of the adult fimbrial epithelium.
A more comprehensive understanding of normal stem cell biology in the ovary and fimbria is also expected to shed light on the genesis of EOC, which is one of the most fatal, yet least understood, reproductive malignancies. Of paramount importance will be investigating the role of LGR5 + cells in EOC initiation. The discovery of a transformation-susceptible Aldh1 + Lgr5 + stem cell niche at the mouse ovary hilum is particularly promising. Whether a similar cancer stem cell niche exists in humans is unknown, and this warrants future investigation. Armed with these new stem cell (and potentially cancer stem cell) markers, and recently developed mouse ovarian cancer models, we are now well placed to make important advances in our understanding of ovary and fimbrial epithelial stem cell biology and cancer biology in the next decade The development of near-physiological 3D human ovary and fimbrial culture systems and gene-editing technologies will also provide essential tools for studying ovary and fimbrial stem cell biology and for disease modelling. Given the immense heterogeneity and complexity of EOC, elucidating how stem cells execute tissue repair during normal homeostasis, and identifying OSE and fimbrial niche components and signals, are crucial steps towards understanding how tumours develop in these tissues.  During incessant ovulation, a normal or mutated OSE stem cell (dark pink) at ovulatory wound margins becomes entrapped within the ovary cortex to form a cortical inclusion cyst. The entrapped stem cell is exposed to aberrant paracrine signals from the activated stromal niche (orange area) and/or autocrine signals within the cystic lumen (grey area). Over time, the entrapped stem cell undergoes Müllerian metaplasia, followed by malignant transformation. Depending on the nature of the niche signals and acquisition of critical mutations, the transformed OSE stem cell differentiates along different Müllerian lineages to generate the different type I EOC cell types. Type II tumours, conversely, arise de novo from cortical inclusion cysts (or from fimbria, see part b). b | A putative fimbrial stem cell as high-grade serous carcinoma (HGSC) cell of origin is shown. A stem cell (dark pink) in the fimbria or ovarian fimbriae (marked with *) develops a p53 signature, followed by development of serous tubal intraepithelial carcinoma (STIC). During ovulation, the STIC cell (or cells) dislodges and becomes entrapped within the ovary cortex to form a cortical inclusion cyst. Exposure to the activated stromal niche (orange area) and/or autocrine signals within the cystic lumen (grey area) induces HGSC carcinogenesis and metastasis. No Müllerian metaplasia is involved.