Maintenance and activation of the limited supply of primordial follicles in the ovary are important determinants of reproductive lifespan. Currently, the molecular programme that maintains the primordial phenotype and the early events associated with follicle activation are not well defined. Here, we have systematically analysed these events using microscopy and detailed image analysis. Using the immature mouse ovary as a model, we demonstrate that the onset of granulosa cell (GC) proliferation results in increased packing density on the oocyte surface and consequent GC cuboidalization. These events precede oocyte growth and nuclear translocation of FOXO3a, a transcription factor important in follicle activation. Immunolabelling of the TGFβ signalling mediators and transcription factors SMAD2/3 revealed a striking expression pattern specific to GCs of small follicles. SMAD2/3 were expressed in the nuclei of primordial GCs but were mostly excluded in early growing follicles. In activated follicles, GC nuclei lacking SMAD2/3 generally expressed Ki67. These findings suggest that the first phenotypic changes during follicle activation are observed in GCs, and that TGFβ signalling is fundamental for regulating GC arrest and the onset of proliferation.
Female mammals are endowed with a finite complement of oocytes, which become enveloped by a single layer of flattened somatic granulosa cells (GCs) to form primordial follicles before or soon after birth (Edson et al., 2009). From the time of follicle formation, a steady trickle of follicles starts growing – a significant event in follicle development that defines reproductive lifespan. Throughout reproductive life, a diminishing population of primordial follicles are maintained in developmental arrest, and the factors maintaining this quiescence are unknown. Spatial analysis of primordial and growing follicles in sections of mouse ovary suggests that primordial follicles produce a local inhibitor that maintains them, and their close neighbours, in an arrested state (Da Silva-Buttkus et al., 2009).
Activation of follicle growth involves changes in GC shape coupled with the onset of GC proliferation and oocyte growth (Adhikari and Liu, 2009; Da Silva-Buttkus et al., 2008; McGee and Hsueh, 2000; McLaughlin and McIver, 2009; Picton, 2001). The precise sequence of these events in the mouse and the identity of the activation factor remain unclear (Da Silva-Buttkus et al., 2008). These questions are important, as pinpointing which cell type (GCs or oocyte) and which cellular processes are the targets of this factor will help ascertain its identity and function.
Currently, the picture regarding the regulation of activation of follicle growth is incomplete. The use of mice carrying mutations in, or lacking, specific genes has identified key transcription factors and signalling pathways that play a role. In GCs, FOXL2 is reported to regulate early morphological changes (Schmidt et al., 2004; Uda et al., 2004) and key components of mTOR signalling have recently been associated with early proliferative events (Zhang et al., 2014). Transcription factors such as Sohlh1/2 (Choi et al., 2008), Nobox (Rajkovic et al., 2004), Lhx8 (Pangas et al., 2006), Foxo3a (Castrillon et al., 2003; Hosaka et al., 2004) and Ybx2 (Yang et al., 2005), as well as internal regulators of mTOR (Adhikari et al., 2009; Adhikari et al., 2010) and PI3K/Akt (Liu et al., 2006) signalling are also known to control early oocyte growth. Culture experiments using neonatal rat ovaries have identified possible candidate growth factors that stimulate activation of follicle growth in vitro (Skinner, 2005), and have confirmed the importance of kit ligand (KL) signalling in this process (Hutt et al., 2006; Nilsson and Skinner, 2004; Packer et al., 1994; Parrott and Skinner, 1999). One growth factor family that has attracted much attention is the TGFβ superfamily, with its downstream signalling molecules, the SMAD proteins (Knight and Glister, 2006; Pangas, 2012a,b). SMAD2 and SMAD3 (hereafter SMAD2/3) signalling is of particular interest due to their predominant expression in GCs of primordial and early growing follicles (Billiar et al., 2004; Drummond et al., 2002; Fenwick et al., 2013; Sharum et al., 2017; Xu et al., 2002). TGFβ signalling is an important regulator of cell proliferation (Massagué, 2012); however, the precise relationship between this signalling pathway and the early phenotypic changes in GCs of small follicles have not yet been examined.
The study of activation of follicle growth is problematic. Many knockout mouse models are embryonic lethal, and conditional knockouts created to study follicle development use promoters for genes whose expression is amplified after follicles start to grow, such as AMHR2, ZP3 or GDF9. Furthermore, primordial and transitional follicles are difficult to isolate because of their small size (up to 20 µm) and fragility, and they fail to thrive in vitro and provide scant amounts of mRNA and protein for analysis. Conventional GC culture approaches are not useful as they involve the use of GCs aspirated from antral follicles, which differ from flattened cells in primordial follicles in terms of phenotype, hormone and growth factor production and responsiveness, and gene expression patterns (Herrera et al., 2005; Liu et al., 2001; Mora et al., 2012). We therefore chose to investigate follicle activation in situ, in the freshly dissected intact ovary, which maintains the three-dimensional environment that is essential for the health of these very small follicles, as well as allowing precise identification of morphological changes associated with activation of follicle growth.
In this study, our key objectives were to identify which follicle cell type initially undergoes morphological changes as follicles activate and determine how these changes relate to indicators of TGFβ signalling. We used immunohistochemistry and image analysis to carry out systematic measurements of follicle and oocyte growth, GC proliferation and shape change. We demonstrate that GC proliferation is associated with a subtle but clear change in GC shape which precedes the formation of cuboidal cells, suggesting that cuboidalization is an event resulting from increased packing density of GCs on the oocyte surface. Furthermore, we show that these GC-based events occur before the oocyte starts growing. We have gone on to localize and quantify nuclear and cytoplasmic SMAD2/3 in the GCs of primordial and activating follicles and show that loss of nuclear SMAD2/3 is associated with the onset of GC proliferation. This has allowed us to propose that the GCs are a target for the activating signal and that upstream inhibitory factors of the TGFβ-SMAD2/3 pathway are likely candidates for regulating activation of follicle growth.
A new morphological stage for follicles initiating growth
Primordial follicles are surrounded by a few flattened granulosa cells (Fig. 1A,E,I), whereas primary follicles are enveloped in a single layer of cuboidal cells (Fig. 1D,H,L). In our earlier studies we used the term ‘transitional follicle’ to describe follicles that are intermediate between primordial and primary stages and which are surrounded by a mixture of flattened and cuboidal cells (Fig. 1C,G,K) (Da Silva-Buttkus et al., 2008; Mora et al., 2012; Stubbs et al., 2007). Previously, careful measurement of transitional follicles showed that they had more GCs and larger oocytes than primordial follicles, and it was difficult to show with precision which event occurred first, oocyte growth or GC proliferation (Da Silva-Buttkus et al., 2008). Further experience of follicle morphology and granulosa cell adhesion allowed us to identify a new intermediate stage between primordial and transitional, where two adjacent granulosa cells had formed wedge shapes with an elongating contact between them (Mora et al., 2012). Their appearance was suggestive of the first division of a flat cell (Fig. 1M) into two adjacent daughter cells (Fig. 1N); we descriptively named these ‘zip’ follicles, as they showed granulosa cell contacts that were ‘zipping up’ and lengthening (Fig. 1B,F,J). The key feature of a zip follicle is that it exhibits the first substantial region of adhesion between two adjacent granulosa cells (Fig. 1N) which extends perpendicular to the oocyte surface, i.e. it is the first phenotypic sign of the onset of change in shape of GCs leading to cuboidalization (Fig. 1O). Retrospective analysis of images of early pre-antral follicles acquired during our previous study (Mora et al., 2012) showed the changes in GC shape (Fig. 1P-R) and expression of the adhesion molecules N-cadherin and nectin 2, as well as the adhesion-related protein β-catenin in the extending interface between the two adjacent GCs (Fig. 1S-U). These follicles became our focus for further study.
GCs proliferate and change shape before the oocyte grows
The boundary of follicles is not always clear in sections of ovary stained with the nuclear stain DAPI. Therefore, double immunofluorescence for basal lamina-specific laminin and oocyte-specific DDX4 with a DAPI counterstain was carried out to clearly define the basal lamina surrounding the follicle, the oocyte and oocyte nucleus, respectively (Fig. 2A), allowing accurate measurement of their area. Only follicles where the oocyte had a pure blue nucleus with a clear outline were measured, to ensure that the follicle was at its largest cross section (LCS) (Fig. 2B). Follicles with an indistinct basal lamina or which were lying in the ovarian surface epithelium (OSE; outside the OSE basal lamina) were excluded (Fig. 2A). Using ImageJ, measurements of the area of the follicle, oocyte and oocyte nucleus were made, and the number of GC nuclei visible in the image was counted (Fig. 2C). Mean GC height, area and width were calculated from these values.
The area of zip follicles was significantly larger than that of primordial follicles (Fig. 2D). The increased follicle area was due to an increased GC area (Fig. 2E), number of GCs (Fig. 2F) and GC height (Fig. 2G), but not to a larger oocyte (Fig. 2H). Of critical importance, the oocytes (and the oocyte nuclei) in zip follicles were the same size as in primordial follicles (Fig. 2H,I). The combination of increased GC number with the absence of oocyte growth resulted in a higher packing density of GCs on the oocyte surface (Fig. 2J). If we go on to compare oocyte area to the number of GCs seen in the LCS, we see that there is no significant oocyte growth until there are six or more GCs in the LCS (Fig. 2K). Interestingly, in the earliest stages of follicle growth, the oocyte nucleus grows as the oocyte grows, reaching a maximal size as the oocyte exceeds an area of ∼500 µm2 (a diameter of ∼25 µm, Fig. 2L).
These data indicate that the GCs start to proliferate and change shape before the onset of oocyte growth, suggesting that, in the mouse, the GCs are the first cells to respond, at least in terms of morphology, to an activation signal. Previously, we had considered that the response to such a signal could be a change in cell shape followed by GC proliferation (Da Silva-Buttkus et al., 2008; Mora et al., 2012). However, the current data led us to consider the alternative scenario that, in response to the unknown signal, flat GCs start to occasionally proliferate and become more packed on the oocyte surface, resulting in an increase in GC height and cuboidalization, as we observed (Fig. 2G).
Proliferation increases in ‘zip’ follicles
The increased number of GCs in zip follicles prompted us to quantify granulosa cell proliferation at the primordial, zip, transitional and primary stages. Ki67 is generally accepted to be an accurate marker of cell proliferation, particularly in cells undergoing rapid proliferation such as cell lines and malignant cells. However, Ki67 labelling in primordial follicles is rare in both mouse and human (Da Silva-Buttkus et al., 2008; Stubbs et al., 2007) because of the infrequency of cell division, so we chose to use the proliferation marker PCNA, which has been used previously in studies on GC proliferation (Gougeon and Busso, 2000; Lundy et al., 1999; Oktay et al., 1995; Wandji et al., 1997; Wandji et al., 1996). PCNA has a longer half-life of ∼20 h (Bravo and Macdonald-Bravo, 1987), compared with ∼1 h for Ki67 (Scholzen and Gerdes, 2000), so label persists longer after S phase or after cells have re-entered G0. It is therefore possible to detect GCs that are in the process of, or have recently undergone, cell division in cells with a low proliferation rate.
The proportion of zip follicles with one or more PCNA-positive GCs (Fig. 3A-C) was significantly higher than that in primordial follicles (Fig. 3D). Similarly, the proportion of PCNA-positive GCs per follicle was significantly higher in zip follicles than in primordial follicles (Fig. 3E). We confirmed our findings with a second marker of proliferation, minichromosome maintenance protein 2 (MCM2; Fig. 3F-J), which we have previously used to examine GC proliferation in human GCs (Stubbs et al., 2007). These observations support the concept that the zip phenotype is associated with increased GC proliferation and accompanying change in cell shape, and the hypothesis that GCs are the first cell type to respond phenotypically to an unknown activation signal.
FOXO3 exclusion from the oocyte nucleus occurs after the onset of GC proliferation
We have shown that the onset of GC proliferation occurs before that of oocyte growth. We recognise that the oocyte is a metabolically active cell and that morphological changes are likely to occur following internal molecular events. Therefore, we wanted further evidence to support the hypothesis that the GCs, rather than the oocyte, are the first to undergo phenotypic changes. FOXO3 is a transcription factor that is expressed in oocyte nuclei and undergoes nuclear exclusion as follicles activate growth (Fig. 4A). This is a well-documented event that is linked to PI3K signalling (John et al., 2008; Liu et al., 2007a), but the precise stage at which this occurs has not been previously quantified.
The intensity of FOXO3A staining was quantified in greyscale images (Fig. 4B,C). There was no difference in oocyte nucleus FOXO3 intensity between primordial and zip follicles. Transitional and primary oocyte nuclei had significantly less staining (Fig. 4D). Ooplasmic FOXO3 intensity remained constant in follicles from the primordial to the primary stage (i.e. as they activated growth; Fig. 4E). In a further analysis aimed at controlling for variability in staining intensity between follicles, sections and ovaries, the difference between nuclear and cytoplasmic FOXO3 intensity within each oocyte was calculated and compared. The difference between nuclear and cytoplasmic intensity was similar in primordial and zip follicles, was significantly less marked in transitional follicles, while in primary follicles the nucleus was lighter than the cytoplasm (Fig. 4F).
Next, we wanted to compare both intensity of nuclear FOXO3 in oocytes and onset of GC proliferation with oocyte growth. GC proliferation occurred before nuclear FOXO3A exclusion in the oocyte growth trajectory. A significant increase in GC number was observed in oocytes that were 20 and 25 µm in diameter, whereas significant FOXO3A exclusion was seen when oocytes were between 30 and 35 µm in diameter (Fig. S1A), when follicles are reaching the primary stage and starting to multilayer (Fig. S1B). This further supports our argument that GC behaviour changes first.
SMAD2/3 labelling decreases as follicles develop
As the TGFβ superfamily is the major group of growth factors involved in follicle development (Knight and Glister, 2006; Pangas, 2012b), and SMAD2/3 is strongly expressed in primordial and early growing follicles (Billiar et al., 2004; Drummond et al., 2002; Fenwick et al., 2013; Sharum et al., 2017; Xu et al., 2002) we went on to quantify and localize SMAD2/3 during activation of follicle growth. SMAD 2/3 was expressed in the GCs, but not in the oocyte (Fig. 5A-D). Upon inspection, it became clear that some nuclei were SMAD2/3 positive, whereas others were negative. SMADs are known to shuttle between the cytoplasm and nucleus under both basal and stimulated conditions; however, increased nuclear dwell time is a feature of active SMAD signalling (Hill, 2009). This prompted us to examine the expression of nuclear SMAD in more detail in individual follicles. To quantify nuclear and cytoplasmic SMAD, we applied a set threshold to the green (SMAD) channel. Using a sampling circle of fixed pixel diameter, we measured the proportion of pixels exceeding the threshold in nuclei and in a neighbouring area of cytoplasm (Fig. 5E). This approach clearly showed that there were nuclei that had negligible levels of SMAD (Fig. 5A′ to D′,E, white arrows). The majority of nuclei in primordial follicles were SMAD2/3 positive, but the proportion of nuclei lacking SMAD2/3 (or with negligible levels), increased as follicle development progressed (P≤0.0001, Chi-square test for trend; Fig. 5F). The key observation was that there was a significant increase in the proportion of SMAD2/3-negative nuclei (with <1% positive pixels) in zip follicles compared with primordial follicles (P=0.0021, Fisher's Exact test; Fig. 5G).
We went on to examine how follicle stage affected the distribution of nuclei with negligible (Fig. 5H, arrow), intermediate and high (Fig. 5H, arrowhead) intensity of nuclear SMAD2/3. Primordial and zip follicles had nuclei with either negligible or very high levels of SMAD2/3, as shown by the peaks at either end of the distribution, as well as intermediate levels. As development progressed, an increasing percentage of follicles had negligible SMAD2/3 (Fig. 5H, arrows), with a decreasing percentage having very high levels (Fig. 5H, arrowheads). Finally, we examined whether there was a relationship between nuclear intensity of SMAD and the shape of the GC and showed that flat GCs had higher levels of SMAD2/3 than cuboidal GCs (Fig. 5I). Interestingly, total SMAD levels in the GC layer, including both nuclear and cytoplasmic expression, declined with follicle stage, especially from the primary stage onwards (Fig. S2).
In summary, primordial follicles have the highest proportion of SMAD2/3-positive nuclei, suggesting that signalling through the SMAD2/3 pathway may play a role in the maintenance of the primordial pool. A similar pattern of SMAD2/3 expression in early stage follicles from adult mouse ovaries was also seen, confirming that the findings were not exclusively related to the immature ovary (Fig. S3).
Nuclear localization of SMAD2/3 disappears with GC proliferation
The variability of nuclear SMAD2/3 expression in primordial, zip and transitional follicles, and the presence of increasing numbers of nuclei totally lacking SMAD2/3 as development progressed, led us to ask whether there was an association between nuclear SMAD expression and GC proliferation. For this, we used a cell cycle marker with a shorter half-life, Ki67, to get an accurate snapshot in time of proliferation and nuclear SMAD2/3 occupancy. Sections from three ovaries were examined, encompassing 169 follicles with a total of 828 nuclei (Table S1). The key finding was that expression of SMAD2/3 and Ki67 was mutually exclusive in the majority of nuclei, with Ki67 expression occurring predominantly in nuclei lacking SMAD2/3 (Fig. 6A-E, white arrows). Of the GCs lacking nuclear SMAD2/3, 47% were positive for Ki67. Conversely, 82% of Ki67-positive GCs were negative for SMAD2/3. Only a small proportion (5%) of GC nuclei contained both SMAD2/3 and Ki67 (Fig. 6F). Each ovary demonstrated a similar expression pattern (Fig. S4). As follicles progressed through development, the percentage of SMAD2/3-positive nuclei declined, whereas the percentage of Ki67-positive nuclei increased (Fig. 6G).
The expression of Ki67 in nuclei lacking SMAD2/3 led us to conclude that this signalling pathway is associated with the maintenance of GCs in cell cycle arrest, and furthermore, that the onset of GC proliferation is associated with the exclusion of SMAD2/3 from the nucleus.
Here, we have shown that as follicles activate, GC proliferation precedes phenotypic changes in the oocyte, including growth of the oocyte and its nucleus, and FOXO3 export from the oocyte nucleus. Furthermore, we have shown that the canonical TGFβ signalling mediators and transcription factors SMAD2/3 are expressed in the nuclei (>80%) of primordial GCs, suggesting that this pathway is involved in inhibition of GC proliferation and maintenance of the primordial pool. We have previously proposed that primordial follicles themselves produce a local inhibitor that prevents activation of follicle growth, and that follicles only start to grow in regions where levels of this putative, and as yet unidentified, inhibitor are reduced (Da Silva-Buttkus et al., 2009). The data presented here indicate that the putative inhibitor is likely to be a member of the TGFβ superfamily that signals via the SMAD2/3 pathway. Moreover, as follicles begin to grow, the loss of nuclear SMAD2/3 in GCs indicates that perturbation of this signalling pathway is a key molecular event associated with further proliferation and differentiation.
Key spatial information is lost when whole ovaries are homogenized for standard RNA or protein techniques. Therefore, identifying the genes associated with the primordial phenotype and the molecular changes that occur as follicles activate requires detailed image analyses. Based on the morphometric analyses in this study, as follicles activate, GCs initially proliferate and subsequently change shape. This is consistent with reports in other species that have also suggested that changes in GCs precede oocyte enlargement during follicle activation (Braw-Tal, 2002; Hirshfield, 1991; Lintern-Moore and Moore, 1979; Stubbs et al., 2007). Importantly, the observed change in shape is likely to occur as a consequence of an increased number of GCs occupying the surface of the oocyte. GCs of single-layered follicles are tethered basally to the basal lamina via hemidesmosomes (Da Silva-Buttkus et al., 2008) and apically to the oocyte via strong heterotypic N-cadherin–E-cadherin and Nectin-2–Nectin-3 adherens junctions (Mora et al., 2012). The latter attachments are strong and cannot be disrupted even in Ca2+-free conditions. Flat cells have little intercellular contact with each other, but a large area of contact with the oocyte, which has a finite surface area. As the GCs divide parallel to the oocyte surface (see Fig. 3G, arrow, for an example), their contact area with the oocyte decreases. Following division, the daughter cells become abutted to each other, forming two wedge-shaped cells with an increased area of lateral contact between them, which start to express Nectin 2 and N-cadherin, characteristic of adherens junctions (Mora et al., 2012), (Fig. 1). Adherens junctions play a role in both intercellular adhesion and intracellular signalling (Andl and Rustgi, 2005; Erez et al., 2005). The increased intercellular contact and expression of adherens junctions as GCs divide, become more packed and change shape, may lead to increased intracellular signalling from these junctions, further stimulating GC proliferation. This could explain our previous observation that cuboidal cells proliferate more than flat cells (Da Silva-Buttkus et al., 2008).
Recent advances in stem cell technologies have made it possible to derive oocyte-somatic cell complexes and recapitulate follicle development in vitro (Hikabe et al., 2016). Despite this advance, it has not yet been possible to produce true primordial follicles in vitro, partly due to a lack of molecular details associated with the pre-granulosa cell phenotype. In this report, we have used high-resolution confocal imaging to analyse the sub-cellular expression of SMAD2/3. The general localization of SMAD2/3 is consistent with previous studies indicating that SMAD2/3 is predominantly expressed in GCs of small (single and multi-layered) pre-antral follicles (Fenwick et al., 2013; Sharum et al., 2017; Xu et al., 2002). Although the antibody used in this study does not discriminate phosphorylated from non-phosphorylated forms, the expression of these transcription factors in the nuclei of GCs of primordial-transitional follicles implies that this signalling pathway is actively regulating genes in these cells. TGFβ signalling is known to regulate a variety of cellular responses depending on the context, and in epithelial cells is often associated with maintaining a cytostatic phenotype (Massagué, 2012; Zhang et al., 2017). Since GCs from single-layered follicles are typically characterized as having low rates of proliferation, with some ‘epithelial-like’ qualities, including expression of cytokeratin and attachment to a basal lamina (Da Silva-Buttkus et al., 2008; Mora et al., 2012), we propose a role for TGFβ signalling in this context.
Previous spatial analysis has proposed the existence of a locally produced inhibitory factor (Da Silva-Buttkus et al., 2009), which we now believe to be a TGFβ family member. Candidate ligands known to activate SMAD2/3 include TGFβ-1, -2, -3, activins, nodal and several members of the growth differentiation factor (GDF) sub-family (Wakefield and Hill, 2013). Binding of the ligand to type I and II TGFβ receptors results in phosphorylation of SMAD2/3, which promotes their translocation to the nucleus to form transcriptional complexes to regulate target genes (Massagué, 2012). In epithelial cells, these complexes can regulate the expression of genes involved in the cell cycle, including Myc and the CDK inhibitors Cdkn1a, Cdkn1b and Cdkn2b (which encode p21CIP1, p27KIP1 and p15INK4B, respectively) (Chen et al., 2002; Frederick et al., 2004; Hannon and Beach, 1994; Lecanda et al., 2009; Seoane et al., 2001). Interestingly, p27KIP1 is expressed in primordial GCs and deletion of p27KIP1 in mice causes premature activation of follicles (Rajareddy et al., 2007). Although our findings have highlighted an association between TGFβ-SMAD2/3 signalling and the phenotype of primordial GCs, further studies will be required to elucidate the molecular link between this pathway and regulation of the cell cycle.
We have also shown that, reciprocally, a significant proportion of GCs lacking nuclear SMAD2/3 are undergoing proliferation, as indicated by the proliferation marker Ki67. Modulation of TGFβ signalling can occur through ligand availability, expression of receptors and co-receptors, internal antagonism of SMADs and SMAD-independent pathways, as well as interaction or availability of transcriptional co-factors (Massagué, 2012; Wakefield and Hill, 2013). Since growing preantral follicles are known to express a range of extracellular antagonists, including follistatin, HTRA1, TWSG1, CTGF and GREM2 (Fenwick et al., 2011; Harlow et al., 2002; Shimasaki et al., 1989), it is plausible these proteins could act on neighbouring follicles to inhibit TGFβ signalling in this context. Indeed, direct inhibition of the Type I TGFβ receptors in vitro leads to increased follicle activation in neonatal mouse ovaries (Wang et al., 2014). We have also demonstrated that modulation of serine-threonine kinase receptor associated protein (STRAP), can affect early follicle development (Sharum et al., 2017), presumably through known interactions that influence downstream SMAD2/3 signalling (Datta and Moses, 2000). Thus, there are multiple mechanisms that can potentiate the observed change in SMAD2/3 signalling in early growing follicles.
On the other hand, exclusion of SMAD2/3 from GC nuclei of early growing follicles is slightly paradoxical, since this coincides with increased expression of GDF9 in oocytes, a TGFβ ligand that acts on GCs to promote proliferation (Elvin et al., 1999; Hayashi et al., 1999; Vitt et al., 2000). TGFβ signalling is known to be context dependent (Massagué, 2012); for example, in endothelial cells, exposure to low levels of TGFβ supports a SMAD2/3-dependent cytostatic phenotype, whereas elevated levels initiate a change to a SMAD1/5/8-mediated proliferative phenotype (Goumans et al., 2002). Whether a similar mechanism exists in quiescent versus proliferating GCs is not known; however, it is interesting to note that the reduction in nuclear SMAD2/3 in growing follicles coincides with an increase in SMAD1/5/8 expression in GCs (Fenwick et al., 2013; Sharum et al., 2017). Moreover, in cultured human biopsies, activin A, which signals via SMAD2/3, was found to inhibit or promote activation of primordial follicles depending on the concentration of the ligand (Ding et al., 2010). Increased exposure to certain TGFβ ligands may therefore initiate a change in the way GCs respond to TGFβ signalling. Importantly, TGFβ ligand-receptor complexes can also drive SMAD-independent pathways, including PI3K/AKT, mTOR and MAPK (Lamouille et al., 2012; Lamouille and Derynck, 2007; Lee et al., 2007). Activation of PI3K/AKT and mTOR pathways both promote early follicle growth in mice (Castrillon et al., 2003; Liu et al., 2007a,b; Zhang et al., 2014). The findings presented here suggest stage-dependent responses to TGFβ signalling in GCs.
Proliferating GCs of early growing follicles express a range of growth factors capable of regulating oocyte growth via the PI3K pathway (Adhikari and Liu, 2009; Thomas and Vanderhyden, 2006). Activation of the PI3K pathway leads to phosphorylation and nuclear exclusion of FOXO3 in these cells, which is a key molecular indicator of growth (Castrillon et al., 2003; Liu et al., 2007a). Based on our findings, it is likely that GCs proliferate and change shape prior to oocyte growth. The earliest molecular events that control these morphological changes are likely to be regulated by TGFβ signals in GCs as proposed (Fig. 7). This raises a number of questions in terms of identifying the specific TGFβ factor(s) that act on primordial GCs, the downstream targets of SMAD2/3, and the molecular events that precipitate SMAD2/3 nuclear exclusion. Clarifying these details would potentially provide a new framework for improving our understanding of how the ovarian reserve is maintained and how some follicles are activated to grow.
MATERIALS AND METHODS
Ovary collection, culture and immunohistochemistry
C57BL/6 female mice were housed in accordance with the Animals (Scientific Procedures) Act of 1986 and associated Codes of Practice. Ovaries were dissected from mice at 4 and 12 days post-partum (d4 pp; d12 pp), at which stages they contain a preponderance of primordial and early growing follicles, respectively (Fenwick et al., 2011; Kerr et al., 2006; Sharum et al., 2017). In the d12 mouse ovary, somatic cells from growing follicles are proposed to originate from an initial developmental wave, which eventually contributes to fertility during the first three months after puberty, whereas somatic cells from primordial follicles contribute to fertility after this (Mork et al., 2012; Zheng et al., 2014). Given these distinct developmental origins, some analyses were also carried out on adult ovaries for comparison (Fig. S3). Ovaries were fixed in 10% neutral buffered formalin (NBF; 48 h) for later paraffin embedding and serial sectioning. Sections (5 µm) of ovary were dewaxed and rehydrated. Antigen retrieval was performed by boiling slides in citrate buffer (10 mmol l−1 citric acid; pH 6.0; 4×5 min) or Tris buffer (100 mmol l−1, pH 10.0) with 5% (w/v) urea and non-specific binding blocked using 10% (v/v) goat serum (Invitrogen, ThermoFisher, UK) with 4% (w/v) BSA (Sigma-Aldrich, Poole, UK). Primary antibodies were applied overnight at 4°C. Primary antibodies used were rabbit anti-MVH (3 µg ml−1; ab13840, Abcam, Cambridge, UK); rabbit anti-laminin (2.7 µg ml−1; ab11575, Abcam); mouse monoclonal anti-SMAD 2/3 (0.25 µg ml−1; 133098, Santa Cruz Biotechnology); rabbit polyclonal anti-Ki67 (1:200; ab15580, Abcam); rabbit monoclonal anti-MCM2 (3619 Cell Signaling). After washes in PBS, sections were incubated in an appropriate Alexa Fluor™ fluorescently labelled secondary antibody (1:200; Invitrogen) for 40 min. All sections were mounted in Prolong Gold medium containing 4,6-diamino-2-phenylindole (DAPI, Invitrogen). Labelling of rabbit monoclonal anti-FOXO3a (1:200; 2497; Cell Signalling Technology) was visualized with a peroxidase-conjugated avidin biotin complex (ABC kit; Vector Labs, Peterborough, UK) and 3,3′-diaminobenzidine tetrahydrochloride (DAB brown kit; Invitrogen). Granulosa cell proliferation was assessed using a PCNA staining kit, with a DAB detection system (93-1143; Zymed; Thermo Fisher, UK) according to the manufacturer's instructions.
Transmission electron microscopy
Ovaries were fixed for 24 h in 3% v/v glutaraldehyde in 0.1 mol l−1 cacodylate buffer (pH 7.2), post-fixed in 1% osmium tetroxide in cacodylate buffer for 1 h and embedded in Epon (TAAB Laboratories Equipment Ltd). Ultrathin sections were stained in a saturated solution of uranyl acetate in 50% ethanol, followed by Reynold's lead citrate, and examined using a Tecnai Spirit (FEI) electron microscope.
Scoring follicle stage
The stage of follicle development was assessed on the basis of granulosa cell shape. Primordial follicles were enveloped by a single layer of flattened GCs; zip follicles had two wedged-shaped GCs abutting each other, with or without adjacent flat GCs; transitional follicles had a mixture of flattened and cuboidal GCs surrounding the oocyte; primary follicles were surrounded by a complete layer of cuboidal GCs and primary+ follicles were developing a second layer of one or more GCs adjacent to the oocyte.
Image acquisition and analysis
For all studies, digital images were taken encompassing two or three entire ovary sections from three ovaries. For DAB-stained sections, images were acquired with a DS-Fi1 digital camera (Nikon, Kingston-upon-Thames, UK) attached to an E600 microscope (Nikon), using the NIS-Elements AR image analysis program (version 3.10; Nikon). For immunofluorescence, RGB images were acquired using a Leica inverted SP5 confocal laser-scanning microscope (Leica Microsystems, Wetzlar, Germany) with a ×40 oil immersion objective. A single image of the entire section was compiled from individual images using DoubleTake (Version 2.2.8; http://echoone.com/doubletake/). This allowed the numbering of individual follicles for analysis, ensuring that follicles were not measured twice or omitted. Original high-power DAB-stained or RGB images were analysed using ImageJ (http://imagej.nih.gov/ij), with reference to the annotated, compiled whole-ovary section image. Images were calibrated to pixels/µm. Using ‘Measure and Label’ in ImageJ, each measurement (whether follicle, oocyte or individual GC) was numbered on the image, allowing data to be matched to images.
SMAD2/3 and Ki67 expression was quantified by analysing confocal RGB images that had been separated into individual 8-bit greyscale images. In the relevant channel for the specific antibody, a threshold was set by adjusting the channel threshold until the degree of labelling visually matched that seen in the original image. This threshold was then maintained for analysis of the whole section. The area of positive pixels that exceeded this threshold was measured using ImageJ and presented as a proportion of the area of the structure in question, whether a nucleus or a sampling circle. For simplicity, we refer to this measure as ‘intensity’. Light micrographs of FOXO3A immunohistochemistry (where DAB was used to visualize immunoreactivity and haematoxylin to visualize nuclei) were analysed using ImageJ. Images were viewed as 8-bit greyscale images and calibrated using intensity values from lightest and darkest nuclei. This calibration was maintained throughout the analysis. Images of haematoxylin staining alone were also analysed.
Data handling and analysis
ImageJ measurements and cropped images of individual follicles with ImageJ annotations were imported into custom-built databases (Filemaker Pro). This also allowed automatic calculation of follicle and oocyte diameter (d) (2×√area/π), oocyte circumference (πd), and mean GC height [(Follicle diameter−oocyte diameter)/2]. Scoring of follicle stage (Fig. 1), GC shape and number of GC nuclei could be carried out within the database from the imported images. This simplified scoring and reduced operator error during data transcription. Statistical comparisons were carried out using Prism 6 for Mac OSX (www.graphpad.com). Data were generally non-normally distributed, so non-parametric analyses were used as stated in figure legends (Kruskal-Wallis test with a Dunn's post test).
The authors would also like to acknowledge Norah Spears (University of Edinburgh) and Sofia Granados Aparici (University of Sheffield) for useful discussion.
Conceptualization: K.H., S.F., M.A.F.; Methodology: K.H., J.M.M., C.D., M.A.F.; Validation: K.H.; Formal analysis: K.H., C.D.; Investigation: K.H., J.M.M., C.D., R.C., M.A.F.; Resources: C.D.; Data curation: K.H.; Writing - original draft: K.H., M.A.F.; Writing - review & editing: K.H., S.F., M.A.F.; Visualization: K.H.; Supervision: K.H., S.F., M.A.F.; Project administration: K.H.; Funding acquisition: K.H., S.F., M.A.F.
This work was supported by grants received from the Biotechnology and Biological Sciences Research Council (BBSRC) UK awarded to K.H., S.F. and M.A.F. (BB/H00002X/1), the Medical Research Council (MRC) UK awarded to S.F. and K.H. (G0802782) and the Genesis Research Trust (UK).
The authors declare no competing or financial interests.