Discussion
This is the first study evaluating the transcriptomic profiles of the growing, plateau and atretic folliculogenesis phases in large mammals. It is clear that the value of this transcriptomic data relied heavily on the ability to separate the follicles in their respective phase using flow cytometry. Although it has been shown that flow cytometry is a valid method for evaluating follicular atresia[
14] and distinguishing healthy follicle from atretic ones[
16] it has not been used to distinguish an intermediate growth phase for further molecular analysis. Unlike previous methods[
14,
16] which only took the proportion of apoptotic cells under consideration, in the current study all cell cycle phases (G1, G0/G1, S, G2) were utilized to distinguish the 3 important physiological statuses: growing, plateau and atretic. The BGA analysis indicated that cytometry was indeed able to categorize the follicle samples in three distinct groups based on transcriptomic expression of all probes on the microarray. The intermediate position of the plateau phase is supporting the assumption that the process is progressive but the different group are not on the same axe/line indicating a non-linear process. Such feature is suggesting a shift in gene expression with some being shut off and other activated.
As mentioned before, the growth phase in the bovine follicles can be distinguished in three segments; the FSH-independent growth until a follicular diameter of 2-3 mm is reached, the FSH dependant growth from 3-9 mm and the LH dependant phase from 9 mm to ovulation. Since the GC samples in this study originate from the 6-9 mm range, they can either represent follicles that are still growing, in a plateau phase (reduced FSH support) or in the atresia process (insufficient FSH) excluding dominant follicles which are larger than 8.5 mm. The first contrast analysed, P vs. G, corresponds to the reduction of cell division that is associated with decreasing FSH support[
24]. In this comparison, follicles with a high proportion of mitotic granulosa cells were compared to follicles with roughly equal proportion of mitotic granulosa and apoptotic granulosa, corresponding to a turning point in follicular growth. The shift from growth to atresia is not only gradual it remains reversible for up to one day (if the dominant is ablated) indicating there is a transition period[
25]. The upstream analysis further supports the inhibition of growth by the down-regulation of
MYC,
FOS and
E2F1-2-3 in response to
P-53 and
CDKN1A. This later gene has been recently associated with estrogen inactive dominant follicles[
26]. Using IPA, an analysis of FSH-responsive genes was performed in order to confirm that the growing follicles selected by cytometry had an expression pattern corresponding to gonadotrophin-stimulated growth. Transcripts for genes related to steroidogenesis: Cytochrome P450, subfamily XIX (
CYP19A1, aromatase) was significantly lower in P relative to G and in A relative to P, and Cytochrome P450, family 11, subfamily A, polypeptide 1 (
CYP11A1), was significantly lower in P relative to G. Both showed a significant decreasing linear trend from G to A. These two genes were found to be up-regulated in cultured granulosa cells when stimulated with FSH[
27], demonstrating that the change in expression profile observed between growing and plateau follicles, at least for
CYP11A1 and
CYP19A1, could be associated with FSH deprivation. Mani et al., 2010[
28] found
CYP11A1,
CYP19A1, but also hydroxy-delta-5-steroid dehydrogenase, 3 beta- and steroid delta-isomerase 1 (
HSD3B1) to be up-regulated in cultured bovine GC receiving IGF1, another factor driving GC proliferation and differentiation, however the significantly lower
HSD3B1 mRNA levels in P relative to G according to microarray, were not consistent with qRT-PCR results. This suggests that the role of
HSD3B1 in granulosa may only come later in folliculogenesis, namely in the formation of the corpus luteum and the production of progesterone[
29]. In addition,
HSD3B1 gene expression up-regulation could indeed be induced by FSH alone in rat granulosa[
30], but not in bovine[
28,
31]. The aromatase cytochrome P450 enzyme, encoded by the
CYP19A1 gene is the final rate-limiting step in the synthesis of estrogens from androgens in granulosa cells. In gonadotrophin-responsive follicle, FSH enhances expression of aromatase which consequently enhances estradiol production from granulosa cells which is associated to the expression of survival genes, inhibition of apoptosis genes and promotion of follicular growth[
32]. Using in situ hybridization in a recruited cohort ≥4 mm, it was observed that
CYP19A1 mRNA was significantly lower in atretic follicles compared to healthy follicles[
33].
CYP19A1 is the only gene in this study showing an expression pattern significantly different between each phase, which confirms it as an accurate marker, closely correlated to follicular health.
Some well-established granulosa cell proliferation markers also showed the expected pattern of expression across the three follicular stages, providing additional support to the precise categorization by cytometry. In mice ovaries, Cyclin D2 (
CCND2) mRNA is expressed in proliferating granulosa, as determined by proliferating cell nuclear antigen (
PCNA) co-localization[
34], a marker for GC proliferation[
35]. The mRNA and protein levels for both genes are increased in response to FSH[
34,
36]. This is in accordance with the current study, as
CCND2 transcript levels were found to be significantly lower in A follicles relative to G and P follicles, while
PCNA was more highly expressed in G follicles relative to both P and A follicles. In addition, both genes showed significant decreasing trends in their expression levels when comparing each successive phase. These result suggest that although both these genes are indeed good indicators of GC proliferation, higher
PCNA expression is characteristic of actively growing follicles, while
CCND2 expression will only go down significantly in GC once the follicle has entered advanced atresia.
Focusing on the cell cycle category, it is clear as illustrated by the 21 annotations for which there was a significant association with genes in the G vs. P dataset that it is a focal point of change in GC as the follicle enters the plateau phase. There is a clear trend towards the down-regulation of later phases of the cell cycle such as S-, G2- and M-phase, as well as cytokinesis, while the resting G1 phase is showing an increasing trend. Looking at the contrasting cell death category of functions, there is a considerably smaller overlap with the genes in the P vs. G dataset, as only 5 categories had a significant association. This indicates that the plateau phase of follicular growth, at the GC level, takes place through a negative modulation of the cell cycle machinery, more so than by the immediate activation of apoptotic or cell death pathways. This is represented in the functional analysis only by a slight positive trend of apoptosis and the broad cell death annotation not yet trending positively. The down regulation of cell survival annotations is also interesting to note as FSH mostly supports growth and proliferation by promoting downstream survival signals such as insulin like growth factors (IGF), depending on the growth stage[
37]. Without those signals GC inevitably undergo apoptosis.
Among the genes contributing to the prediction that the proliferation of cells function was decreased in P vs. G, was
CKS1 and 2CDC28 protein kinase regulatory subunit 1B (
CKS1B; also known as
CKS1). A study investigating the role of
CKS1B using siRNA, showed that depleting
CKS1B in mouse embryonic fibroblasts (MEF) cells resulted in the cessation of cell proliferation[
38].
CKS1B mRNA was significantly lower in the GC of P and A follicles relative to G follicles and there was a significant decreasing trend from G to A. Its depletion could therefore be an early indicator of GC proliferation slowing in P follicles and it remains low in A follicles where GC proliferation has stopped entirely.
As discussed previously, while the growing phase of 6-9 mm follicles is characterized by ample FSH support and their plateau phase results from a reduced supply of FSH, the atretic phase, however, is the beginning of the irreversible follicle demise which is characterised by the up regulation of specific genes. One of the genes being upregulated is
ID3. ID3 is more expressed in A relative to P and relative to G, and also showed a significant increasing trend from G to A. In mural granulosa cultured
in vitro with FSH and/or COC,
ID3, expression was decreased by FSH but increased by COC[
39].
ID3 acts as a dominant negative of basic helix loop helix (bHLH) transcription factor, therefore FSH and COC may regulate granulosa cell function by tuning activity of bHLH factors through
ID3. This indicates that follicles categorized as atretic likely came from an environment low in FSH, allowing for higher expression of
ID3, compared to follicles with high or higher FSH levels such as G and P respectively, where
ID3 expression was inhibited.
Tissue transglutaminase (
TGM2; transglutaminase 2 (C polypeptide, protein-glutamine-gamma-glutamyltransferase)) mRNA, which is significantly up regulated in A relative to P was shown to be expressed at different levels for varying degrees of atresia in mice follicles, characterized by hematoxylin-eosin and TUNEL, suggesting it is a good indicator of the degree of follicular atresia[
40].
TGM2 is expressed in all organs and is the most ubiquitous of all the transglutaminases[
41]. It has also been well documented in other tissues that
TGM2 mRNA coincides with apoptosis
in vivo, such as the liver[
42].
TGM2 is believed to sensitize cells to apoptosis by hyperpolarizing mitochondria, an event which precedes the loss of transmembrane potential, a decrease in
GSH levels and consequently an increase in the production of reactive oxygen species (ROS)[
43]. The indication that
TGM2 transcript levels increase with the degree of atresia makes it an interesting marker for the A vs. P contrast, as both follicle type will contain apoptotic granulosa, albeit at different stages, which can explain why classical granulosa apoptosis markers such as Fas (TNF receptor superfamily, member 6;
FAS)[
44] and BCL2-associated X protein (
BAX)[
45,
46] reviewed in[
47] are not significantly different in their expression between the two conditions. However, B-cell CLL/lymphoma 2 (
BCL2), a marker for apoptosis resistance[
46,
48], was higher in atretic follicle GC than in both P and G follicles which is unexpected[
48] but not the first time such results are obtained. Valdez et al., 2005[
49] who worked with dominant follicles collected at different days of the first follicular wave found that although the number of non-viable cells increased from days 4, to days 6 and 8, there was only a significant increase in the ratio
BCL2:BAX from day 4 to 6, with no significant difference between day 4 and 8. It can therefore be suggested that the translocation of
BAX from the cytoplasm to the mitochondria and its interaction with
BCL2[
50] takes precedence over expression of those factors in the activation of apoptosis in bovine GC. More studies looking at protein localization will be needed to elucidate this concept. It is important to note that previous studies did not necessarily evaluate the presence of those markers in contexts matching the current study. For example,[
44] found
FAS mRNA levels to be higher in GC of the two largest, atretic subordinates than in the healthy dominant follicle. In small- and medium-sized follicle, although in an
in vitro culture, it was found that neither
BAX nor
BCL2 were modulated by FSH or Insulin-like growth factor 1 (
IGF1)[
28]. Since FSH is thought to modulate the dynamics of the different phases in the current study, this corresponds to the steady
BAX levels between the three growth phases, and indicates that the increase expression of
BCL2 in A is likely driven by some other factor.
Tumor necrosis factor receptor superfamily, member 21 (
TNFRSF21, also known as death receptor 6
: DR6), much like other members of the TNFR family, has been shown to induce apoptosis when over expressed. In hen ovaries it was found that both transcript and protein levels were higher in cells of atretic follicles relative to healthy follicles[
51]. In the current study, transcript levels were significantly up-regulated in A follicles compared to P, presenting
TNFRS21 as a marker of late follicular atresia in GC. In addition, initial experiments following
TNFRSF21 discovery showed that ectopic expression of this death receptor induced apoptosis in HeLa S3 cervical carcinoma cells[
52]. As the
Functions analysis of IPA shows, the image is not precise when comparing two tissues which are both atretic, but differing in the stage of atresia they have reached. Being among the qRT-PCR-validated genes shown to be significantly higher in A follicle GC relative to P follicle GC,
TNFRSF21 can now suggested as an indicator of advanced follicular atresia.
Desmoglein-2 (
DSG2)[
53] and chaperonin containing TCP1, subunit 2 (beta) (
CCT2)[
54] and disabled-2 (
DAB2)[
55] were significantly up-regulated in A and P relative to G. Although these genes have not been studied in granulosa specifically, it was shown in other cell types that they promote apoptosis since siRNA’s directed against their transcripts lead to a decrease in cell death.
STK17A has been shown to induce apoptosis in multiple cell lines[
56,
57]. It seems to act in early atresia as his expression levels were higher in both P and A relative to G. Overexpression of FOS-like antigen 1 (
FOSL1 or fra-1) in cultured cell lines has been shown to increased apoptosis[
58].
It is now clear that cell death related genes operate through many different pathways. The regular apoptosis in granulosa cells must occur with minimal inflammation and other processes seem to be activated; such as autophagy. Oxidized low density lipoprotein (lectin-like) receptor 1 (
OLR1) mRNA expression levels were measured in the current study in order to evaluate the potential action of autophagy in the later stages of atresia, as was previously observed in quail granulosa cells[
59]. Up-regulation of
OLR1 expression was associated to non-apoptotic, autophagic cell death, as determined by vacuoles and actin remodelling[
60].
OLR1 expression, although evaluated as significantly higher in A than P follicle GC by microarray data, was not observed to be significantly different between phases by qRT-PCR. However, qRT-PCR results did confirm the significant up-regulation of xin actin-binding repeat containing 1 (
XIRP1), a gene involved in actin cytoskeleton re-organization[
61] and the most up-regulated gene in A compared to P in microarray data, may also indicate that autophagy is more prominent in late atretic follicle than in early atretic follicles. This could correspond to a greater proportion of the granulosa cell population having entered apoptosis and consequently a greater quantity of cellular debris needing to be digested.
Transcript levels for BCL2, ID3, TGM2, TNFRSF21 and XIRP1 were all significantly higher in A follicles GC relative to P and G, indicating that they are indicators of late atresia. CCT2 and STK17A were significantly higher in both P and A relative to G, and could therefore be used to differentiate between growing and early-atretic follicles. The same can be proposed for CKS1B, CYP11A1 and CYP19A1, for which the levels seem to decrease significantly when the follicle passes from growing phase to the plateau phase and remain at lower levels throughout atresia, except for CYP19A1 whose transcription decreases even more. Finally, DSG2 and GADD45A were significantly higher in A follicles compared to G follicles, with P follicles showing an intermediate levels, which resulted in a significant increasing trend from G to A.
The difference in functional activity between the P vs. G contrast and the A vs. P contrast, as predicted by IPA software, is striking. While 21 functional annotations were predicted to be decreased in P relative to G and only were predicted to 4 increased, no functions were predicted to be decreased in A relative to P but no less than 54 were predicted to be increased. This is a strong suggestion that the plateau or static phase really corresponds to a quiescent or resting stage, intermediate between the active, FSH-supported growth phase and the atretic process.
The functions predicted to be increased, in a statistically significant way, in the A follicle GC relative to P follicle GC revolve around tumorigenesis, cell cycle progression and survival of cells as well as apoptosis and cell death, development of various cell types, differentiation of various cell type, homeostasis, proliferation and multiple cell movement related functions. For example, the functional annotation with the highest z-score is cell movement of neutrophils. Neutrophils are polymorphonuclear leukocytes, or white blood cells, and have been shown to be recruited to the thecal compartment of pre-ovulatory follicles just before ovulation[
62], in corpus lutea (CL) and increase during CL regression, suggesting a role in both ovulation and luteolysis[
63]. These cells are also present in higher numbers among thecal cells in atretic follicles relative to healthy follicles[
64]. The results of the current study complement these results by specifying that infiltration and movement of leukocytes is a hallmark specific to the advanced atretic follicles. It also suggests that granulosa are likely the source of signals attracting these white blood cells to the surrounding thecal compartment and that atresia may indeed be a differentiation process similar to ovulation and later luteolysis.
Based on microarray results,
GADD45A was modulated and linked to 9 of the predicted activated transcription factors in the P vs. G contrast. However, qRT-PCR validation showed that relative expression was only significantly higher in A follicles GC, relative to G, with P at an intermediate level, resulting in a significant increasing trend from G to A. It is therefore possible that these transcription factors drive the expression of
GADD45A mRNA as early as the plateau phase, but this only result in a significant difference in transcript levels when the atretic stage is reached. This qualifies
GADD45A as another indicator of late-atresia.
GADD45a has been shown to cause cell cycle arrest at the G2-M transition by binding Cdc2 within the Cdc2/cyclin B1, a complex required to complete the G2-M transition, causing the dissociation of the complex and preventing cell cycle progression[
65]. It has also been shown that transfection of a Gadd45a expression vector has induced apoptosis, reportedly by interacting
MEKK4/MTK1 and activating the
JNK/p38 signalling which induces apoptosis[
66]. Although they extracted RNA from both the thecal cells (TC) and granulosa cells (GC) at once, from bovine follicles,[
67] found significantly higher levels of
GADD45A transcripts in the second largest follicle of the first follicular wave (7.8 ± 0.2 mm) than in the largest one (10.7 ± 0.7 mm). In addition, they found, using in situ hybridization, that
GADD45A mRNA was expressed in both TC and GC in atretic follicles but only in GC in healthy follicles, as classified by progesterone and estradiol concentration in follicular fluid. They suggest that these results indicate that
GADD45A activity is increasingly needed for the progression of apoptotic cell death in follicular atresia and stress the fact that stage-specific gene expression may closely reflect the follicle’s growth or atresia. All of the seven transcription factors predicted to be activated by IPA and reported to drive
GADD45A expression in the literature: breast cancer 1, early onset (
BRCA1), forkhead box O3 (
FOXO3 or
FKHRL1), forkhead box O4 (
FOXO4 or
AFX1), myogenic differentiation 1 (
MYOD1), tumor protein p53 (
TP53) (see also Figures
5‐
6), tumor protein p63 (
TP63) and tumor protein p73 (
TP73) have been shown to be involved with apoptosis[
68‐
74].
Very recently, two large scale genomic analysis were published by Rodgers (12, 13) where small follicles are compared to large ones and healthy follciles from less than 5 mm compared to atretic ones. For the second study, closer to the work reported here but in a different follicle class (smaller than 5 mm compared to us 6-9 mm), the focus is on the extracellular matrix and demonstrate how important the matrix is for the survival and growth of the follicle. Only one gene was validated with PCR, the aromatase (
CYP19A1) and as in our results, decreases significantly in atretic follicles. For their upstream regulators (our Figure
5 and their table eight) several pathways are common between these 2 follicles classes: P53 is activated and MYC is inhibited. These are complementary datasets that will lead to a better understanding of the folliculogenesis process.
IPA also revealed significant overlap between differently-expressed genes (DEG) of both contrasts and the NRF2-mediated oxidative stress response canonical pathway, which leads to the activation of the transcription factor NRF2 (
NFE2L2, nuclear factor (erythroid-derived 2)-like 2). This transcription factor responds to environment insult[
75] including reactive oxygen species (ROS)[
76]. It was observed that ROS generated by the mitochondria play an important role in the release of cytochrome
c and other molecules which lead to the activation of apoptosis[
77]. Indeed, in granulosa, FSH acts to suppress ROS, generated by steroidogenesis and metabolic activities, within the granulosa by stimulating the synthesis of the antioxidant glutathione (
GSH)[
78]. In the same study, it was shown that blocking
GSH synthesis lead to GC apoptosis. Multiple target genes of
NFE2L2 were modulated in both contrast based on microarray data. Among them,
SOD2 expression is upregulated by
NFE2L2 according to the IPA database. Its protein encodes the mitochondrial isoform of the superoxide dismutases. It removes the superoxide anion in the dismutation reaction producing hydrogen peroxide and molecular oxygen.
SOD2 transcript levels were greater in P and A follicles relative to G. This may indicate that withdrawal of FSH leads to an increased ROS production, which in turn leads to the activation of the
NFE2L2 oxidative response pathway and the expression of
SOD2 antioxidant enzyme to prevent runaway ROS accumulation as part of the atretic process. Although more target genes would have needed to be validated by qRT-PCR to confirm the intermediate role of
NFE2L2, this results still indicates a role for increased
SOD2 transcription in follicular atresia. More validation with PCR and protein measurements are now required to dissect each of these specific pathways but strong hypothesis can now be put forward to better explain the follicular dynamics.