Reinterpretation of evidence advanced for neo- oogenesis in mammals, in terms of a finite oocyte reserve

A primary observation made in mice by proponents of neo- oogenesis has been the incorporation of the thymidine analogue, 5-bromo-2-deoxyuridine (BrdU), by germ cells located in the ovarian surface epithelium (OSE), as detected by immunocytochemistry using anti-BrdU monoclonal antibody: this was interpreted as evidence for mitotic germ cells [7, 10], with the OSE functioning as a classical, germinal epithelium [7‐9, 11]. Johnson et al. [7] discounted the alternative possibilities that BrdU-incorporation arose from either mitochondrial (mt) DNA replication or DNA repair in oocytes, on the basis that "the degree of BrdU incorporation observed in cells due to either of these processes is several log orders less than that seen during replication of the nuclear genome during mitosis." This assumption is invalid because the immunocytochemical technique used is both likely and sensitive enough to detect (a) mtDNA synthesis and (b) DNA repair in meiotically arrested oocytes, as discussed below.

Following the deduced existence of mitotic germ cells in the OSE (above), Bukovsky et al. [8] and Virant-Klun et al. [9] endeavoured to culture OSE derivatives, and subsequently reported the production of "oocyte-like" cells in vitro. Two major limitations are common to both studies.

(a) The criteria used to denote an "oocyte-like" phenotype [8, 9] are morphological, namely: cells with large and rounded morphology in which a large or no nucleus is visible, and which may be surrounded by a structure resembling a zona pellucida (ZP). However, the photomicrographs presented may instead depict those general features of cells undergoing apoptosis, necrosis or - especially - oncosis [46], namely: cell swelling, plasma membrane breakdown, and swollen or lysed nuclei. Structures described as "developing zona pellucida" [8, 9] may reflect cellular swelling, membrane rupture and lysis, and spillage of cytoplasm [46]; the "germinal vesicle" [8, 9], nuclear swelling [46]; and "germinal vesicle breakdown" [8, 9], karyolysis [46]. These considerations underline the importance of validating putative oocytes by immunocytochemical and molecular techniques, rather than by morphological criteria. The attempt by Bukovsky et al. [8] to detect ZP-antigenicity in these cells by immunofluorescence is marred throughout by a high background of staining of the cytoskeleton, which is probably an artefact of desiccation arising from the unconventional step of air-drying cells overnight, prior to fixation. Desiccation and cell death occur extremely rapidly under these conditions [47, 48], with interim activation of survival and death pathways [49]. Regarding the deduced ZP-antigenicity of OSE-derived "germ-like" cells as detected using PS1 antibody [8], it should be noted that Skinner and Dunbar [50] considered their antibody to be non-specific for ZP proteins as it recognises a carbohydrate moiety present on the apical surface of the OSE.

(b) It is immediately apparent that the culture systems of Bukovsky et al. [8] and Virant-Klun et al. [9] are relatively very simple, without addition of the growth factors, cytokines or feeder-cell support that usually are essential to the growth of pluripotent germline cells or ES cells. In fact, the growth of embryonic or germline stem cells under these conditions would be unprecedented. What cells, therefore, could constitute the proliferating populations in these studies?

As cultures were obtained by the conventional technique of scraping of the OSE, the heterogeneity of cells should be considered: an estimated 98% of cells obtained in this way are ovarian epithelial cells [51], and contaminants include extraovarian mesothelial cells, endothelial cells, ovarian somatic and mesenchymal cells, and immune cells [52]. Moreover, cultured OSE demonstrates an epitheliomesenchymal phenotype with contractile functions, and the capacity to differentiate into stroma, granulosa cells or Müllerian epithelia, reflecting its role in vivo as a dynamic tissue involved in post-ovulatory tissue repair and remodelling [52]. Granulosa cells express Oct4 and are multipotent, differentiating into neurons, chondrocytes and osteoblasts [53]. Therefore, in the absence of data from clonal cell analysis, and of unambiguous validation by stem cell-specific markers (see below), the claims of Bukovsky et al. [8] and Virant-Klun et al. [9] for spontaneous in vitro differentiation of germline stem cells into cells of mixed phenotype should be regarded with caution.

The cell types cultured by Virant-Klun et al. [9] from OSE scrapings from postmenopausal women, termed "putative stem cells", "oocyte-like", or "embryonic", may be re-identified from information in the literature. "Putative stem cells" were identified morphologically as round cells, 2-4 μm in diameter, located below or above the OSE [9]. However, the possibility arises that these are small immune cells, e.g. lymphocytes or plasma cells, which are seen located above and below the OSE in ovarian sections [54]. After enrichment by differential centrifugation, these "putative stem cells" proliferated in culture [9]. Plasma cells, also, can be cultured easily in simple media [55], but the presence of this cell type as a culture contaminant was not considered [9]. Virant-Klun et al. [9] stated that the proliferating "putative stem cells" generated adherent oocyte-like cells, 20-95 μm in diameter, with ZP-like, germinal vesicle-like and polar body-like structures that were ascribed to an oocyte nature. However as stated above, these structures could arise from oncosis in any of the cell types being cultured, causing cell swelling, karyolysis and cytoplasmic leakage. In their cultures, Virant-Klun and colleagues [9] also describe the formation of "embryoid body-like" and "blastocyst-like" structures, interpreted as products of parthenogenetic activation of oocyte-like cells. However, they are far less convincing in appearance than the (parthenogenetic) embryos demonstrated by Hübner et al. [56] to arise from ES cell differentiation into oocytes. Could there be an alternative explanation for the structures produced by Virant-Klun et al. [9]? The aggregates of cells termed "embryoid-body like" could arise from any cell type, rather than being diagnostic of embryoid bodies proper with their complex internal differentiation. And the vesicles formed by these aggregates with continued culture could arise from a contaminating epithelial cell type, such as OSE [52], which has the capacity to polarise and form impermeable junctions. The propensity to form vesicles in culture is a common property of epithelial cells from epithelial linings [57]; and the increased tendency of OSE to line clefts and inclusion cysts in the ovary, with increasing age, may be relevant here [52]. Further clues to the identity of the cells can be gleaned from patterns of transcription: "putative stem cells" expressed OCT4, SOX-2, NANOG and C-KIT, and "blastocyst-like" structures expressed OCT4, SOX-2 and NANOG, from which an embryonic nature of the putative stem cells was inferred by Virant-Klun et al [9]. However, a recent study by Song et al. [58] first showed that the trio of stem cell regulatory genes, Oct4, Sox-2 and Nanog, constitute markers for epithelial stem cells, whose function is vital to regeneration and tissue homeostasis: they are expressed during the regeneration of rat tracheal epithelium in vitro, specifically by epithelial stem cells in the G₀ phase. Expression of Oct4 is associated also with a variety of types of epithelial stem cells, but not their differentiated derivatives [59]. Moreover, human epithelial ovarian cancer cell lines and the multilayered structures, or spheroids, they form in suspension culture are known to highly express stem cell-specific genes, including OCT4, NANOG and NESTIN [60, 61]. It is therefore inferred that the OSE-derivative cultures of Virant-Klun et al. [9] comprise epithelial stem cells, which are responsible normally for maintaining the integrity of the OSE - a property that may be especially important in ovaries of post-menopausal women [54], used here. This inherent regenerative potential may be manifest in culture. Another feature is consistent with the presence of OSE in these cultures - the expression of C-KIT [51]. In fact, both C-KIT and KIT LIGAND are expressed by human, normal OSE [62].

The importance of critically evaluating claims for the validation of cell lines as female (or ovarian) germline stem cells is further illustrated by the recent study of Pacchiarotti et al. [11]. These authors reported the isolation and characterisation of germline stem-cell lines from ovaries of neonatal mice of the TgOG2 strain. (These mice carry an Oct4-GFP transgene where GFP expression is controlled by an Oct4 promoter sequence. They are considered in more detail in section 2.(iv).) Their main conclusions are as follows:

(a) Germline stem cells were identified at the ovarian surface, on the basis of their small size (10-15 μm) and expression of Oct4-GFP, Mvh, c-kit and SSEA-1. These cells were purported to transition into germ cells of intermediate size (20-30 μm), and subsequently into growing oocytes.

(b) Cell populations containing the putative stem cells were isolated from disaggregated suspensions of whole ovaries by fluorescence-activated cell sorting for Oct4-GFP expression, and propagated using a feeder-based culture system. It was deduced that the derived lines consisted of ovarian germline stem cells from their expression of germ-cell and stem-cell markers (namely, Gcna1, c-kit, Oct4, Nanog and GFR-α1).

However, many of these assumed marker specificities are incorrect and the above conclusions are therefore unwarranted, as discussed in detail below. Rather, it is proposed that the cultures consisted of monolayers of OSE, together with a proportion of early oocytes and/or oogonia. That is, a complex co-culture system is envisaged containing both somatic and germ-cell types. It is notable that the culture medium used by Pacchiarotti et al. [11] was optimised for spermatogonial stem cells (SSC) [63], as was that employed by Zou et al. [10] for FGSC. These media are considered further in section 2.(v), as potentially being mitogenic for growth-arrested oogonia.

(a) Rather than providing direct evidence for germline stem cells, the localisation of small cells (≤15 μm) expressing Oct4, Mvh and SSEA-1, and subtending the OSE, is compatible with residual oogonia [64‐66]. In fact, the authors acknowledged the likely existence of oogonia in these neonatal ovaries.

(b) These putative germline stem cell lines show a striking resemblance in morphology and growth characteristics (with a low mitotic rate) to previously established mouse and human OSE cell lines [67‐69], growing in monolayers as epithelial colonies with cobblestone appearance, with a tendency towards multilayering at the centre. (Compare, for example, the cellular morphology in Figure three 'N' of Pacchiarotti et al. [11] with that of mouse OSE in Figure two 'A' of Roby et al. [67] and in Figure four 'B' of Szotek et al. [69].) Like established lines of mouse OSE cells at low passage [67], these putative stem cells lacked tumorigenicity in mouse xenograft systems. Furthermore, markers reportedly expressed by these cultures are not germline specific: GFR-α1 is expressed by OSE [70]; and co-expression of c-kit, Oct4 and Nanog was discussed in section 2.(ii), in the context of the OSE as a regenerative epithelium.

(c) Concerning the structures described as "embryoid bodies", patterns of gene expression were entirely consistent with OSE, as a mesoderm-derived, multipotent epithelium with stromal characteristics. For example, nestin [60] and Gata-4 [69] are markers for OSE stem cells. FoxA2 is known to be expressed in uterine glands [71], and expression in this culture system may therefore be indicative of OSE cells undergoing Müllerian-type differentiation towards endometrioid cells [72]. In short, the structures described resemble those spheroids that are formed by both normal OSE [68, 73] and ovarian cancer-derived cell lines [60].

Detection of Gcna-1 in these cell lines requires further comment, as this antigen is considered specific to the nuclei of germ cells in the neonatal and foetal gonad, from zygotene through pachytene stages of meiotic prophase. It is relevant that Alton and Taketo [74] observed immunocytochemical staining for Gcna1 in a large number of cells either in, or protruding from, the OSE in foetal mouse ovaries at 18.5 d.p.c., which was attributed to oocytes in the process of exfoliation. However, that those cells did not express Mvh [74] is incompatible with their identification as oocytes. It is therefore suggested that Gcna-1 may be expressed by OSE, especially during the neonatal period or in culture. Another germ cell-specific gene, VASA, is expressed by ovarian epithelial cancers, which arise from transformation of the OSE [75]. Now that candidate stem cells for OSE have been identified by Szotek et al. [69], it will be of interest to determine if genes involved in germ-cell specification also are involved in normal epithelial regeneration, or differentiation. As well as increasing understanding of the etiology of ovarian epithelial cancers, this information will help clarify the origin of cell lines claimed to represent ovarian germline stem cells [8, 9, 11] on the basis of expression of germ-cell markers.

Recently, Tilly et al. [15] cited their findings from busulphan (BU) treatment of female mice as key evidence for neo- oogenesis, based on their understanding that this chemotherapeutic, alkylating agent targets replicative - and not postmeiotic - germ cells in females, as well as males. By their reasoning, inhibition of de novo oocyte formation by BU treatment leads to exhaustion of the oocyte reserve by normal processes during oestrus cycling: "Young adult female mice treated with busulfan exhibit a gradual loss of the entire primordial follicle reserve over a 3-wk period without a corresponding cytotoxic effect on primordial follicles [7]. Such an outcome would be expected if busulfan were, as past studies contend [76], selectively eliminating replicative germ cells that support primordial oocyte formation. The net result would be the normal rate of follicle loss via atresia no longer partially offset by de novo follicle formation, leading to accelerated depletion of the follicle reserve without the need for a corresponding increase in the rate of oocyte death." However the major premise here, that BU targets only replicative (and, by definition, premeiotic) germ cells in both females and males without causing atresia in postmeiotic cells (oocytes and spermatids), is seen to be incorrect from what is discussed below. Furthermore, it is deduced that the data of Johnson et al. [7] provide direct evidence against neo-oogenesis, and against precursors to oocytes being supplied from bone marrow precursors. To this end, it is necessary to consider the known effects of BU on female and male, murine reproductive function.

Johnson et al. [12] modified their concept of neo- oogenesis to specify that oocyte progenitors are supplied to the ovary by the bone marrow via the circulatory system. This came from experiments on wild type (wt) and Atm-deficient (Atm^-/-) mice in which sterile, depleted ovaries were reportedly repopulated with oocytes derived from EGFP-labelled progenitors, following peripheral blood cell transplantation (PBCT). Subsequently there have been other reports of successful engraftment of donor somatic cells as oocytes following CT and BMT [13], with the provisos that: only a low percentage of designated immature oocytes are donor-derived (around 0.1% of total oocytes in recipients) when bone marrow or peripheral-blood cells are transplanted; designated follicles are never observed beyond preantral stages (i.e. maturing antral or Graafian follicles); and donor cell-derived mouse offspring have never been produced. (Meanwhile, other attempts to reproduce these findings have proved entirely unsuccessful [14, 23].) The general consensus is that any de novo follicles do not undergo ovulation, although they may support the depleted ovary [13]. What, therefore is the functional relevance of this proposed, renewing population of early-stage oocytes? Arguments leading to alternative identities for those cells designated as de novo, immature oocytes [12, 13] are given below.

Another challenge to the concept of a fixed ovarian pool at birth was made by Zou et al. [10], who claimed to have isolated female germline stem cell (FGSC) lines from both neonatal and adult mice ovaries (the adult mice being of unspecified age), having first identified putative FGSC in the OSE of neonatal and adult mice by BrdU-incorporation (see section (i), above). Remarkably, FGSC lines were shown to be capable of reassembly into follicles on reintroduction into a sterile ovary, and produced viable offspring that transmitted a transgene through the germline. The authors take their considerable achievements as validating the existence of a germline stem cell population in the ovary, but do not consider the possibility that their lines arise from quiescent oogonia present in the postnatal ovary, which are induced to proliferate in culture under conditions devised originally to be highly mitogenic for SSC (Figure 1). Arguments leading to this conclusion are presented below. A starting premise is the existence of oogonia in the postnatal mouse ovary, as documented previously by Pepling and Spradling [33], and Greenbaum et al. [109]: about 10% of germ cells persist within small germline cysts containing 2-4 cells at 26.5 d.p.c., or day 7 postnatal [33].

It has been amply demonstrated in mouse models that during homeostasis there predominates self-tolerance to ovarian antigens, sustained by a mechanism dependent specifically on regulatory T cells (Treg). (Treg, either thymus derived or produced by activation of naïve T cells, function to suppress activation of T, B and NK cells.) In the steady-state ovary, atretic follicles contain degenerating ooyctes that are the targets of autoreactive CD4⁺ T cells [126]. Concurrently, ovarian antigens continuously stimulate Treg in the regional lymph nodes to maintain self-tolerance [127, 128]. Autoimmune ovarian disease (AOD) results from loss of functional Treg, e.g. by thymectomy or iatrogenic effects, so that self-tolerance is converted into an active T cell response [128‐130]. In their study of AOD in the thymectomised mouse model, Wheeler et al. [128] further demonstrated that ovarian antigen-specific Treg are capacitated by autoantigen exposure within regional, draining lymph nodes; and that AOD development can be abrogated in the thymectomised recipients by transfer of Treg from the lymph nodes of normal female donors. This Treg-based mechanism may be directly relevant to the studies under consideration here [12, 13].

The CT combination of cyclophosphamide (CY) and BU causes catastrophic damage to oocytes and ovarian failure [12, 13], and is likely to increase apoptosis, which functions normally to promote oocyte clearance and tissue remodelling. According to current thinking, efficient apoptosis provides a safeguard against autoimmunity. But a high burden of apoptosis is strongly implicated in the development of the autoimmune state, by cellular spillage or increased exposure of the immune system to autoantigens [131]. By this reasoning, CT would serve as a trigger for autoimmunity in the ovary, whereby the load of apoptotic cells may exceed the clearance capacity of macrophages and/or dendritic cells.

Cyclophosphamide (CY), the chemotherapeutic alkylating agent used in combination with BU as an ovotoxic agent [12, 13], constitutes possibly an additional trigger for autoimmunity. CY has established immune-enhancing effects, involving both the stimulatory and suppressive arms of adaptive immunity: CY treatment augments effector T-cell stimulation, while selectively depleting Treg numbers and function [132]. These actions of CY are achieved by alteration of subsets of dendritic cells in lymphoid tissue, which normally maintain peripheral tolerance via Treg activation [133].

In a highly relevant study that used the non-obese diabetic (NOD) mouse model, Brode et al. [132] demonstrated the potential of Treg to abrogate organ-specific autoimmunity, and deduced the existence of a temporal window between disease induction and development of an autoreactive T cell response during which suppression would be effective. A single injection of CY induced onset of the autoimmune syndrome, type 1 diabetes (T1D), which was synchronous with selective reduction in Treg in lymph nodes through apoptosis, and with reduced suppressive capacity of Treg in vitro. Furthermore, the ensuing autoreactive T cell response could be suppressed, and development of T1D abrogated, by transfer of antigen-specific Treg from a non-diabetic, syngeneic donor to CY-treated, NOD recipients - provided that Treg were received between 1 and 8 d after CY treatment. Thereafter, the developing autoreactive T cell response predominated. A mechanism was proposed for the action of CY, whereby an imbalance created between CD4⁺CD25^- T cells and Treg leads to priming of autoreactive T cells and development of autoimmunity. (Normally, interaction of Treg with dendritic cells within lymph nodes suppresses the priming of naïve autoreactive T cells.)

A mechanism is therefore proposed from (A) and (B) for the studies considered [12, 13], by which CT treatment induced ovarian failure and apoptosis, and caused depletion of ovarian antigen-specific Treg, thereby promoting activation of effector T cells and inducing autoimmunity to ovarian antigens. A proportion of oocytes may have survived, but a switch from self-tolerance to intolerance caused these to be eventually cleared. It is in keeping with this hypothesis that autoimmune premature ovarian failure in humans is characterised by inflammatory infiltration into developing follicles and the production of anti-ovarian autoantibodies, while primordial follicles are spared [134‐136]. Also, immunosuppressive, corticosteroid therapy may lead to resumption of menses in women with autoimmune oophoritis with secondary amenorrhea [136].

In mouse models, reconstitution of the immuno-haematopoietic system by BMT or transfer of HSC attenuates autoimmunity and may achieve disease remission (reviewed by Kaminitz et al. [137]). The mechanism of such modulation may involve resetting immune homeostasis [138‐140] or reversal of spontaneous autoimmunity, for example by clonal deletion, anergy, or suppression [137, 141]. In the proposed scenario of the ovary with developing autoimmunity to ovarian antigens following CT (A, B), immuno-modulation would be restored by the haematopoietic chimaerism induced by syngeneic BMT, with resumption of self-tolerance. And it is suggested that the specific tolerogenic mechanism would involve restoration of suppressive Treg function by BMT, either from donor stem cells developing in the recipient's thymus, or from bone marrow acting as a natural reservoir for homeostatic trafficking of functional, activated Treg [142]. The transient immunosuppressive and lymphopenic effect of CY [143], given in combination with BU as ovotoxic treatment, also would provide a niche for homeostatic expansion of Treg following BMT [130]. Consequently, remaining primordial follicles would grow and reach ovulation, rather than be cleared as in control, CT-treated mice not receiving BMT.

Returning to the study of Lee et al. [13], several observations can now be reinterpreted:

(i) Beneficial effects on fertility are attenuated when BMT is postponed from 1 week to 2 months following CT. This can be explained by the priming of an autoreactive T cell response after 1 week, in accord with Brode et al. [132] (see (B), above). BMT at 2 months is ineffective, and there is progressive destruction of the surviving oocyte reserve by immune clearance.

(ii) Postponement of mating after CT and BMT by two months versus 1 week results in decreased fertility. This can be explained by exhaustion of the surviving oocyte reserve in the 2-month interim period by oestrus cycling (occurring every 3-5 days), compared with the suspension of oestrus cycling during consecutive pregnancies.

This line of reasoning also may account for the observations of Johnson et al. [12] with Atm^-/- mice. Although ovaries of mutant females are described as devoid of oocytes and developing follicles [12], evidence exists for the persistence of residual germ cells: rare and abnormal oocytes were recorded in Atm^-/- mice that were 20 days old [144], and between 17 and 29 days old [145]. This accords with the observation of Johnson et al. [12] of low-level, germ cell-specific gene expression (Oct4, Mvh, Dazl and Stella) in ovaries from adult Atm^-/- mice. Crucially, Johnson et al. [12] noted that, following ovotoxic CT and BMT, ovaries from Atm^-/- mice contained a small number (maximum, 25) of follicles at 2 and 11.5 months after BMT, while non-transplanted mice did not. It is suggested here that ovarian failure caused by Atm-deficiency also may induce autoimmunity to ovarian antigens, resulting in clearance of those rare, residual oocytes. Thus, BMT to Atm^-/- mice would restore tolerance, allowing those surviving oocytes to develop to antral stages [12].

This hypothesis of CT-induced autoimmunity and its suppression by immune-cell transfer gives a prediction for the system of Eggan et al. [14], where female mice were subjected to CT (with BU and CY) and 1 d later were connected parabiotically to untreated partners, to provide a shared circulatory system. In this case, the priming of an ovarian autoreactive T cell response in the CT-treated mouse would be suppressed by functional, ovarian antigen-specific Treg transfusing from the untreated partner, thereby imposing a state of dominant self-tolerance in both parabionts (Figure 2D). That Eggan et al. [14] found no evidence for restoration of fertility in the CT-treated mice by parabiosis, nor in CT-treated (non-parabiotic) mice by BMT, most likely reflects their use of superovulation as the measure of ovarian function. (The technique was used primarily to harvest large numbers of oocytes, to ascertain any contribution of blood-derived precursors to neo- oogenesis.) However, there is inherent variance in the superovulatory response in mice (which is apparent in the data presented), and a dependence of superovulation on the instantaneous number of hormonally responsive, antral follicles. Consequently, the superovulated yield would not reflect the size of the primordial and growing follicle pool - especially when these are close to exhaustion following CT. Therefore, superovulation would not provide an accurate indicator of long-term fertility. The measures taken by Johnson et al. [12] and Lee et al. [13] to assess fertility are more effective, analysing total numbers of non-atretic immature follicles per ovary, and recording live-birth pregnancies over time, respectively.

To paraphrase Oktay and Oktem [146], further investigation is needed into the mechanism of rescue of fertility in CT treated females by BMT. Establishing the validity or otherwise of the neo- oogenesis concept is crucial to understanding germinal function, and to preserving fertility. The proposed involvement of the immune system may, on the other hand, offer possibilities for preserving ovarian function in women undergoing CT, and for treatment of AOD and primary ovarian insufficiency, e.g. by Treg-based immunotherapy [130].