The transcriptional regulator CBX2 and ovarian function: A whole genome and whole transcriptome approach

The chromobox homolog 2 (CBX2) was found to be important for human testis development, but its role in the human ovary remains elusive. We conducted a genome-wide analysis based on DNA adenine methyltransferase identification (DamID) and RNA sequencing strategies to investigate CBX2 in the human granulosa cells. Functional analysis revealed that CBX2 was upstream of genes contributing to ovarian function like folliculogenesis and steroidogenesis (i.e. ESR1, NRG1, AKR1C1, PTGER2, BMP15, BMP2, FSHR and NTRK1/2). We identified CBX2 regulated genes associated with polycystic ovary syndrome (PCOS) such as TGFβ, MAP3K15 and DKK1, as well as genes implicated in premature ovarian failure (POF) (i.e. POF1B, BMP15 and HOXA13) and the pituitary deficiency (i.e. LHX4 and KISS1). Our study provided an excellent opportunity to identify genes surrounding CBX2 in the ovary and might contribute to the understanding of ovarian physiopathology causing infertility in women.


Identification of CBX2.1 and CBX2.2 genomic direct and indirect targets in KGN cells. Within
72 hours post-transfection of CBX2, pre-granulosa cells or KGN did not exhibit any morphological changes similar to the small and astrocytic shape of the male NT2-D1 cell line morphology 20 (data not shown). This suggests that there is no link between CBX2 gene expression and KGN morphological changes. We applied the DamID method that couples whole genome-wide protein-DNA interaction to next-generation sequencing to gain deeper insights into the function of CBX2 isoforms in granulosa cells (GC). We identified 524 and 835 enriched binding sequences of CBX2.1 and CBX2.2, respectively. We expanded CBX2 transcriptional landscape, by the used of RNA sequencing that identifies, contrary to DamID, also genes that are not necessarily physically bound by CBX2 and can be considered indirect targets. Thus, we found 692 and 668 differently expressed genes, respectively. A larger number of 1167 and 810 genes were significantly up or downregulated by CBX2.1 and CBX2.2 silencing, respectively.
To independently validate the DamID and RNA-seq results, we selected a subset of genes regulated by CBX2.1 and CBX2.2 (as shown in Tables 2 and 3) to evaluate their expression using quantitative real-time PCR (RT-qPCR). Genes selection was based on their potential links to sex development, their role in human and animal sexual diseases and their specific expression in tissues involved in sex development (gonads, sex organs, hypothalamus and pituitary).  (Fig. 4a). The genes were reported to be implicated in the female sex development and were found to be controlled by the ovarian specific genes FOXL2 and WNT4 [21][22][23][24][25][26] recently shown to be downregulated by CBX2.1 isoform 66 . The expression levels of NRG1, BMP2, PTGER2, and FZD7 but not ESR1 were significantly increased after CBX2.1 knocking down (1.33-, 1.29-, 1.27-, and 4-fold, respectively) (Fig. 4a). CBX2.1 activated POF1B, DKK1 and SOX9 gene expressions (1.51-, 1.62-and 1.84-fold, respectively) (Fig. 4b). Of particular interest, SOX9 an essential male-specific gene was demonstrated to be a positive downstream target of CBX2. 1 8,27 in In blue are the GO-terms involved in morphogenesis process. In red are all GO-terms related to Molecular Function. The orange colour represents the cluster of genes coding for signalling pathways. Some regulatory processes were over-represented by the yellow colour. The genes presented in the mustard colour were overrepresented in response to a stimulus. All data is filtered according to p < 0.05. (b) Cytoscape representation of GO-enrichment analysis of CBX2.2 targets. The green colour represents the GO-terms which are involved in morphogenesis and differentiation process. In the blue cluster, we found GO-terms involved in the developmental process. The red colour indicates genes responsible for Molecular Function. The yellow colour represents the regulatory processes. The orange colour is the cluster, which contains genes coding for signalling processes. The genes present in the mustard colour cluster were over-represented in response to a stimulus. All data is filtered according to p < 0.05.
It is important to indicate that RT-qPCR did not show that CBX2.1 and CBX2.2 influenced the expression of each other. This agrees with results published by Völkel et al., showing that long CBX2.1 isoform interacts with the polycomb repressive complex-1 (PRC1) components. In total contrast, none of the PRC1 components was identified with the CBX2.2 short isoform 13 . According to the same authors, CBX2.2 forms homopolymers in a PRC1-independent way. Unlike CBX2.1, CBX2.2 lacks the Pc domain, essential for the interaction with the PRC1 partners 13 .

Discussion
In this work, we showed the gene expression landscape of CBX2 isoforms in the ovary based on data-driven from profiling genes and transcriptome data. Unbiased GO analysis data obtained from the whole genome protein/ DNA interaction and RNA-seq methods revealed a greatly expanded "atlas" of CBX2 new targets implicated in several developmental and functional pathways in KGN. Of particular interest, we demonstrated that CBX2.1 targets are over-represented for GO-term associated with urogenital system development, thereby supporting the involvement of CBX2.1 in human sex development as has been reported by Biason-Lauber and co-workers 8,27 . Several genes with diverse functions related to folliculogenesis, steroidogenesis and ovarian disease like PCOS, POF were found to be regulated directly and indirectly by CBX2 isoforms. Multiple categories of CBX2.1 and CBX2.2 related genes are implicated in generic development, morphogenesis and differentiation events. Our findings are in harmony with recent data revealing the implication of murine Cbx2 gene in the central nervous system development in mice 29 . We showed significant enrichment of genes involved in immune responses, supporting the results of Katoh-Fukui in Cbx2/M33 knock out mice with immunological deficits due to spleen development abnormalities 14 . Yet, no immunological deficit is found to be influenced by CBX2 mutations in human patients 8 . This could be explained by the difference between human and mouse phenotypes and the presence of alternative pathways of CBX2 gene in humans. Among CBX2.1 and CBX2.2-regulated genes were factors involved in the regulation of insulin-like growth factor (IGF) receptors, which were reported to be required for sex determination in mice 30 . In the double knock out insulin-Igf1 receptor null embryos, a delay in ovarian differentiation has been observed, suggesting that in mouse gonads lacking insulin/Igf signalling remain undifferentiated with no clear pathway decision of either testicular or ovarian pathways for several days 31 . CBX2.1 related genes were found to be highly interconnected in ovarian developmental processes and supported an active contribution of CBX2.1 in ovarian function and maintenance. Genes regulated by CBX2.1 like CYP19A1, KISS1, and ESR2 were found to work together to determine and maintain the ovary phenotype [32][33][34][35] . Unlike CBX2.1, the CBX2.2 network appeared to be much more limited, most likely due to the novelty of CBX2.2's functions and the lack of animal model studies. Taken together, the network-based transcriptome and profiling data offered a solid starting point for the elucidation of detailed connections of CBX2 isoforms genes in the human ovarian pathway.
A new single-cell RNA-seq analysis using human fetal gonad cells and their neighbouring somatic cells from 15 embryos between 4 and 19 weeks post-fertilization (GEO accession GSE86146) 36 , demonstrated that the CBX2 transcript is highly expressed in both sexes. The CBX2 expression level was remarkably higher in the female than in the male embryo 37 . Additionally, human follicular transcriptome data (GSE107746) obtained from various developmental stages of oocytes (primordial, primary, secondary, antral and preovulatory) and the corresponding GC published by Zhang et al., showed that CBX2 expression was consistently high at all stages of follicular

ESR1
Maintaining the female phenotype of the endocrine somatic cells of the ovary by inhibiting male cell development 128 .
ERKO mice develop testis-like features. Null ERα mutations in human females exhibit profound estrogen resistance and have features analogous to those in the knock out mouse 67 .

NRG1
Induced by luteinizing hormone (LH)/hCG to activate the MAKP3/1 pathway to promote GC differentiation and controls ovulation and luteinization related events 74 .
Not reported

BMP2
BMP2 with FOXL2 ensure expression follistatin in the developing ovary. It amplifies FSH-induced estradiol production in sheep granulosa cell 23 .
In mice, BMP2 null mutation is embryonic lethal and foetuses contain a low number of primordial germ cells leading to POF 129 .
Mice deficient in Ptger2 have ovulatory defects that are related to an abnormality in cumulus expansion 131 .

SOX9
Stimulates the differentiation of Sertoli cells 132 .

POF1B
Regulates ovarian function 133 Assumed to be a causative candidate of POF 134
In humans, it is a PCOS risk candidate 136 .

FZD7
WNT signalling regulation 137 Not reported Potential role in lipid metabolism 138 Not reported

NTRK2
Involved in the development and the maturation of the central and peripheral nervous systems 140 .
In the knock out mice reduced the number of secondary follicles and a decrease in granulosa cell proliferation 141 .

AKR1C1
Implicated in the inactivation and formation of male and female sex hormones 142 .
In Akr1c1 deficient mice high progesterone levels and display a delay in parturition of several days 143 .

FZD5
Induce the beta-catenin pathway 144 Not reported

SOX4
Heart function 145 Reported in human ovarian cancer 146

RSPO3
Regulation of Wnt/beta-catenin signalling. It has a possible role in folliculogenesis and development of germ cells of fish 147 . Not reported www.nature.com/scientificreports www.nature.com/scientificreports/ development compared with GC 38 . In mice, a study revealed a spectrum of meiotic abnormalities in Cbx2 deficient oocytes at the diplotene stage with abnormal synapsis and non-homologous chromosome interactions in Cbx2 (XX −/− ) mutant oocytes 15 . This phenotype observed in fetal oocytes lacking Cbx2 might suggest that Cbx2 is expressed in follicles and have a functional role for chromatin remodelling required for the establishment and repair of homologous chromosome pairing.
The gene Cytochrome P450 family 19 subfamily A member 1 (CYP19A1), estrogen receptor 2 (ESR2), (Kisspeptin) KISS1 and FSHR were found to be the upregulated by CBX2.1. The CYP19A1 or aromatase A gene is responsible for the aromatization of androgens into estrogens in many tissues in female and male 39 . A previous animal study showed that Nr5a1, also called steroidogenic factor-1 (Sf-1), depletion leads to reduced Cyp19a1 expression and low estradiol levels resulting in testis differentiation in the XX gonad 40 . The gene ESR2  www.nature.com/scientificreports www.nature.com/scientificreports/ was associated with follicular growth 41 . Mutations of ESR2 were found to be responsible for 46,XY and 46,XX DSD, both with gonadal dysgenesis 42 . The genes KISS1 and FSHR were related to ovarian diseases like POF and PCOS 43,44 . Recent studies have shown that kisspeptin-1 and its receptor are expressed in the mammalian ovary and are critical for initiating puberty and regulating ovulation in sexually mature females via the central control of the hypothalamic-pituitary-gonadal axis 45 . Data gathered recently suggested a putative role of kisspeptin signalling in follicular development, oocyte maturation, steroidogenesis, and ovulation 46 . Additionally, loss-of-function of KISS1 is associated with hypogonadotropic hypogonadism leading to reproductive function failure and female infertility 47,48 . As for FSHR, it plays a major role in the development of follicles and steroidogenesis in the ovary 49 and a loss-of-function of FSHR causes ovarian dysgenesis 50 .
We identified upregulated factors such as SRY-box 9 (SOX9) and doublesex and mab-3 related transcription factor 1 (DMRT1) 8,27,51,52 as CBX2.1 downstream masculinizing factors. Heterozygous loss-of-function mutations in human SOX9 cause sex development disorder in XY males 53,54 while gain-of-function mutations, such as gene duplication, can lead to XX female DSD 55 . According to Ledig and co-workers, a partial deletion of DMRT1 causes 46,XY ovotesticular sexual disorder 56 . The ovotestis formation is caused by the disturbed action of DMRT1 in germ cells as well as in Sertoli cells, causing female reprogramming of the testis 56 .
Other genes like empty spiracles homeobox 2 also known as EMX2, Dickkopf-related protein 1 (DKK1) and neurotrophic receptor tyrosine kinase 1 (NTRK1) are gonadal dual-functional factors upregulated by CBX2.1 in ovarian GC. It has been reported that EMX2 is indispensable for the formation of the embryonic structures Müllerian and Wolffian ducts in the female and male embryo 57,58 . A nonsense mutation of EMX2 resulted in uterus didelphysis in Chinese women with incomplete Müllerian fusion 59 . In agreement with our data, a murine study showed that Emx2 is downregulated in Cbx2-deficient gonads 14 . In women, genetic variation in DKK1 may result in hyperandrogenism and metabolic dysfunction of PCOS 60 . Other data suggested that mice Dkk1 plays a backup or fail-safe role in preventing Wnt signalling which is in harmony with the possible antineoplasic role of CBX2 61 . We also found that CBX2.1 stimulated NTRK1 expression, which has been reported to be involved in the assembly of primordial follicles 62 to facilitate the progression of follicular development within the ovary 62 . Together, our data indicate that CBX2.1 might be required within ovarian cells for follicular fate regulation which agrees with previous findings showing Cbx2 mutant female mice with small ovaries and significant germ cell loss 11 .
We found angiopoietin-like-4 (ANGPTL4), a factor yet novel to the scene of sex development, that we found it upregulated by CBX2.1 in GC. Importantly, murine Angptl4 has been found to play a role in lipid metabolism, which can provide cellular energy and mobilize substrate for progesterone synthesis in breeding females 63 . Together, we suggested a putative correlation between CBX2.1 and ANGPTL4 to maintain hormone metabolism in the ovary. ANGPTL4 has been also reported to be an apoptosis survival factor capable of preventing metastasis by inhibiting vascular growth and protecting from tumour cell invasion 64 . The explanation of the role of ANGPTL4 in ovarian physiopathology, if any, seems to be more challenging.
Some of the most important factors negatively regulated by CBX2.1 are estrogen receptor 1 (ESR1), prostaglandin E receptor 2 (PTGER2) and bone morphogenetic protein 2 (BMP2). The genes seem to be interconnected with FOXL2 65 . CBX2.1 was demonstrated to downregulate the female determining factor FOXL2 in testis and ovary gonads 27,66 . ESR1 was reported to cooperate with FOXL2 to restrain SOX9 in the ovary 65 . In women, ESR1 deficiency was associated with clinical features of estrogen resistance, including primary amenorrhea, the absence of breast development, a small uterus and enlarged multicystic ovaries 67,68 . In humans, PTGER2 and BMP2 gene are prerequisite for ovulation and are activated by female factor FOXL2 21,69 . The repressive effect of CBX2.1 on these ovarian factors together with the stimulation of male-typical factors, such as SOX9 and DMRT1, indicate a sort of dual-function of CBX2.1 in the development of the human gonads. It seems not to be an isolated example. Other genes like the WNT4, ESR2 and SF-1 have a necessary role for ovarian and testicular development function [70][71][72] as demonstrated by the fact that genetic variants in these genes cause gonadal dysgenesis in 46,XY individuals 70,73,74 , with ovarian failure in women and ovotesticular DSD 42,70,74 . Some authors suggested that sex determination is sensitive to gonad genes dosage at multiple steps in the gonads pathway 75 , which might be the case of the CBX2 gene. Recent preliminary reports of CBX2.1 genetic variations in 46, XX individuals with gonadal abnormalities lend further weight to an essential role of CBX2 in human ovarian and testicular development 76 .
There is very little information available about the role of the second isoform CBX2.2 in any process in women. Recently, Sproll et al. showed the existence of two CBX2.2 genetic variants that fail to regulate the expression of genes essential for sexual development, leading to a severe 46,XY DSD defects 9 .
Among the factors upregulated by CBX2.2, were the bone morphogenetic protein 15 (BMP15) and the transforming growth factor-beta 2 (TGF-β2) genes. Previous studies showed that BMP15 affected the production of estradiol and progesterone [77][78][79] . Studies in animal have shown that the activation of the primordial follicles is mediated by bmp15 80 . In humans, the defected BMP15 in patients was found to cause ovarian failure 81 . Mounting evidence supported the implication of TGF-β2 in female reproduction and development 82 . TGF-β superfamily members may play different roles in the development of follicles across the species 80 . Tgf-β2 knock out mice study showed multiple developmental defects, including cardiopulmonary, skeletal, ocular, and urogenital system defects 83 . In women, dysregulation of this transforming growth factor circuitry was associated with PCOS 84,85 and fertility problems 86 . These findings may point to potential roles of CBX2.2 in regulating transforming growth factors networks reportedly found crucial during follicular development 38,87,88 .
In the present study, CBX2.2 upregulated dual-functional factors in gonads like the early expressing gonadal factors, Homeobox protein Hox-A13 (HOXA13) and SRY-box 4 (SOX4). In humans, HOXA13 mutations were found to affect uterine development 89,90 and produced hand-foot-genital syndrome in females 91 with a decrease in androgen expression in males 90 . These data imply that CBX2.2 might play a role in the normal expression of HOXA13 in the early developing ovary, mirroring the situation in the human testis 92  www.nature.com/scientificreports www.nature.com/scientificreports/ both ovaries and testes 93 . A recent animal study revealed that Sox4 was among Foxl2 positively regulated genes in the mice ovary 94,95 and showed to repress transcription of Sox9 in fetal gonads, raising the possibility that SOX4 may function as a new feminizing C SOX factor in the regulation of the ovarian determination.
The testis expressed-14 gene (TEX14) is a masculinizing factor downregulated by CBX2.2. It has been reported to be required for intercellular bridges in vertebrate germ cells 96 . In females, these embryonic intercellular bridges have been proposed to have a role in the development of the primordial germ cells 96 and Tex14 −/− ovaries have fewer oocytes relative to control ovaries in mice 97 . The human fetal ovary expresses TEX14 after the 12 th week of gestation, suggesting that the growth of oogonia may be induced by cellular precursor transport from neighbouring oogonia via the TEX14 channels 98 .
CBX2.2 negatively regulated dual-functional factors as mitogen-activated protein kinase kinase kinase 15 (MAP3K15) and Aldo-keto reductase family 1 member C1 (AKR1C1). The two genes appear to serve as markers involved in steroidogenesis 99,100 . Reduced levels of activated mitogen-activated protein kinase (MAPK) contribute to excessive ovarian androgen production in women with PCOS 101 . Besides, the change of expression patterns of MAPKs in rat ovaries was significantly higher during the secondary and antral follicle stages than those in the primordial follicles, primary follicles and corpora lutea indicating their possible involvement during follicular growth and development 102,103 . We assume that CBX2.2 could be implicated in the optimal control of the MAPKs signalling pathway in GC during differentiation and proliferation processes. We studied one of the Aldo-keto reductase steroidogenic enzymes, which is the AKR1C1. Recently collected data showed that it is expressed in adrenal tissue and may be involved in the fine regulation of androgen and androgen receptors (AR) availability in adipose tissue in men and women 104,105 . Aldo-keto reductases (AKR1C1-C4), 5α-reductases and retinol dehydrogenase were found to be expressed in the ovary, indicating that the human ovary might produce dihydrotestosterone via the alternative steroid backdoor pathway 106 . This pathway seems to be considerably enhanced in the polycystic ovary syndrome 104 . Little is known about the direct involvement of AR actions in the female. Nonetheless, previous results based on global AR knock out female mice 105,107 demonstrated that they are subfertile, have defective folliculogenesis and ultimately develop POF. In the human ovary, androgen precursors are crucial for estrogen synthesis and hyperandrogenism in pathologies such as the polycystic ovary syndrome 108 . Taken together, our study provides a piece of indirect evidence about the role of CBX2.2 in the regulation of androgen receptor in the ovary that could be through the control of AKR1C1 expression. However, further studies are paramount to show how these new targets fit into the expanding CBX2.2-regulated network and how CBX2.2 activation and suppression can impact our understanding of ovarian functionality in humans.
We concentrated our study on the developmental side of CBX2 since variants of CBX2 in human leads to developmental defects like gonadal dysgenesis in women and men 8 , Although yet no defect in CBX2 is known in later stages we suggested that CBX2 and some of its targets might be involved in adult ovarian dysfunction such

EMX2
Homeodomain transcription factor EMX2 is critical for the central nervous system and urogenital development.
Emx2 mutant mice died soon after birth because of the absence of kidneys indicating an essential role in the morphogenesis of the urogenital system 58 .

ESR2
Nuclear receptor transcription factors have a crucial role in reproductive function. Subfertility and reduced litter sizes and granulosa cell defect 32 .
KISS1 knock out female mice demonstrated an abnormal reproductive system with abnormal reproductive system phenotype. A clinical case of hypogonadotropic was associated with a loss-of-function of KISS1 48 .

BMP2
BMP2 with FOXL2 ensure expression follistatin in the developing ovary. It amplifies FSH-induced estradiol production in sheep GC.
In mice, BMP2 null mutation is embryonic lethal and foetuses contain a low number of primordial germ cells leading to POF 129

LHX4
Acts as a transcriptional regulator that is involved in the control of differentiation and development of the pituitary gland.

POF1B
Regulates ovarian function Assumed to be a causative candidate of POF 153

FSHR
Follicle stimulating hormone receptor Mutations in the FSHR cause primary ovarian failure in females and impaired spermatogenesis in males 50 .

BMP15
Stimulation of ovarian granulosa cell growth and proliferation and downregulates FSH receptor expression.
In mice: subfertile, in human: ovarian dysgenesis 154 .   www.nature.com/scientificreports www.nature.com/scientificreports/ as PCOS and POF. To lend more weight to our hypotheses, we compared our selected CBX2.1 and CBX2.2 targets with the existing RNA-seq datasets of female embryonic and mature gonad cells 37 .We found that CBX2 is greatly and specifically expressed during all stages of female fetal gonadal cells (FFGC) including mitotic, retinoid acid (RA) responsive, meiotic, oogenesis, endothelial, early granulosa, mural granulosa and late granulosa phases 38 . Substantially, some of the major CBX2.1 and CBX2.2 downstream targets like SOX4, ANGPTL4, AKR1C1, BMP2, EMX2, DMRT1, CYP19A1, FZD5 and TEX14 showed also high and specific expression patterns in the same fetal stages. In our study, the ANGPTL4, DMRT1, EMX2, CYP19A1 genes were found significantly activated by CBX2.1. Using the same available public dataset, ANGPTL4 and DMRT1 were abundantly expressed during the FFGC and particularly in the mitotic, RA-responsive and meiotic stages. New findings indicated that in mice the lack of Dmrt1 in the fetal ovary resulted in the formation of many fewer primordial follicles in the juvenile ovary 109 . EMX2 and CYP19A1 showed high expression in the early, mural and late granulosa stages. To the best Figure 5. (a) Relative expression levels (2 −ΔΔCt ) of the RNA-seq of CBX2.1 downstream genes. ANGPTL4, NTRK1, CYP19A1, DMRT1, EMX2, ESR2 and KISS1 were upregulated after CBX2.1 overexpression (Over Exp). CBX2.1 silencing assay (siCBX2.1) reduced significantly the genes: ANGPTL4, DMRT1, ESR2 and KISS1. NTRK1, CYP19A1 and EMX2 gene did not show any expression in response to siCBX2.1. All graphs are the average of three independent experiments, error bars represent SD from the mean (SEM), and values are expressed as relative to control =1; ***P < 0.001; **P < 0.01; *P < 0.05. non-significant differences are not indicated. (b) Effect of CBX2.2 on DamID downstream targets: AKRC1, TGFα, AMIGO2 and RSPO3. Gene expression levels showed a substantial downregulation after CBX2.2 overexpression (Ov. Exp). The silencing of CBX2.2 (si CBX2.2) significantly stimulated the expression of AMIGO2, RSPO3 and AKR1C1 genes compared to scrambled siRNA. TGBβ2, NTRK2, FZD5 and SOX4 genes were significantly upregulated by CBX2.2. In the siCBX2.2 samples, NTRK2, FZD5 and TGFβ2 expression levels were found to be negatively regulated. TGFα and SOX4 expressions showed no effect relative to the scrambled sample. All graphs are the average of three independent experiments, error bars represent SD from the mean (SEM), and values are expressed as relative to control =1; ***P < 0.001; **P < 0.01; *P < 0.05. non-significant differences are not indicated. www.nature.com/scientificreports www.nature.com/scientificreports/ of our knowledge, Emx2 is one of the genes which was found to be necessary for the survival of the female and male gonads in mice. Nonetheless, the gene was not reported to play a crucial role at various stages during oocyte development 110 . We showed that CBX2.1 downregulated BMP2 ovarian marker. According to the transcriptome data of Zhang and co-workers 38 , this gene was found highly expressed during the meiotic, oogenesis, early granulosa, mural granulosa and late granulosa stages. This is not surprising given the well-proven expression of BMP2 in GC and germ cells 111 . Consistent with this, BMP2 was reported to be important for the follicular development 111 and a predictor marker of embryo quality 112 . We demonstrated that genes like SOX4 and FZD5 are positively regulated by CBX2.2 and seem to have consistent high expressions during all stages of the FFGC. Noteworthy, a recent mice study indicated that Sox4 plays an important role in mouse gonad development by promoting gonad germ cell differentiation 93 . The AKR1C1 and TEX14 are CBX2.2 downregulated genes and exhibited high expression profiles during all ovarian fetal cells. In the first place, the results indicated that these factors are most likely node genes that establish a robust regulatory network governed by CBX2 gene in the fetal ovary, which may be involved in the control of these genes pathways during fetal germ cells and their neighbouring GC development 38 . Overall, this data could offer a solid reference dataset for the implication of the CBX2 and the candidate genes in the regulation of follicular development and could be valuable factors that are important in several stages of ovarian life, spanning from development, maintenance, reproductive potential and function.

TEX14
A limitation of our study is that we did not use primary cultured human GC, but a KGN tumour cell line. Primary GCs can be isolated from follicular fluid during oocyte retrieval procedures. This retrieval takes place after ovulation when GCs undergo a process of luteinization which involves structural and genomic changes that lead to the terminal differentiation of follicular cells with increased progesterone production 113 . Luteinized GCs stop their proliferation, making the long-term cultivation of GCs primary cells extremely challenging 114 . Furthermore, the differentiation of GC into luteinized cells has effects on intracellular signalling and cell cycle www.nature.com/scientificreports www.nature.com/scientificreports/ regulation, rendering these cells not suitable to represent developing GCs. The use of the animal models as an alternative remains problematic 37 as many genes are expressed differently and preferentially in human and mice. Also, Homo sapiens seem to be the only mammalian species to have two CBX2 isoforms. We therefore resorted to the use of cell lines. The KGN cell line is very well characterized 66,115-119 and represents the most suitable alternative in vitro model and the closest to human ovarian cells to ascertain the role of CBX2.1 and CBX2.2 in human ovary cells.
In all, the combination of different hypothesis-generating NGS-approaches allowed us to shed light on the transcriptionally CBX2-dependent landscape in the ovarian pathway. Certainly, new knowledge in CBX2 network using novel advanced technologies for detecting and sequencing exomes are essential to prove the exact role of www.nature.com/scientificreports www.nature.com/scientificreports/ the new putative CBX2 regulated genes in the developing ovary network. It would be an important investigation, not only to advance our understanding regarding gonadal development but also to expand our ability to diagnose, counsel and properly accompany patients affected by ovarian dysfunctions like infertility and cancer. Methods cell culture. The human ovarian granulosa-like tumour commercially available cell line or KGN 116 has been provided by RIKEN BRC. It has been established from a tumour specimen enucleated from a 63-year-old woman who was diagnosed with a local recurrence of granulosa cell carcinoma after menopause. A portion of the granulosa tumour tissue obtained was used as the source of the cell culture 116 . We maintained KGN cells in Dulbecco's essential medium/Ham's F12 medium supplemented with 10% fetal calf serum, 5% Penicillin/Streptomycin at 37 °C in a 5% CO 2 as described in Nishi et al. protocol 116 . Cells were transfected by 2 µg of plasmids encoding for CBX2.1 (SC303599, OriGene Rockville, Maryland, USA), C-Myc-CBX2.2 (RC216313 OriGene Rockville, Maryland, USA) and pCMV6-empty plasmid was used as a control vector (PS100010 OriGene Rockville, Maryland, USA). We transfected cells with Metafectene (Biontex Laboratories, Munich, Germany) with ratio 1:4 (transfection reagent: DNA). siRNA duplexes (purchased from Microsynth) were introduced into cells using Lipofectamine RNAiMAX (Invitrogen) in two consecutive rounds at a final concentration of 40 nM. Experiments were typically performed 48 hours after the first transfection. siRNA duplex for CBX2 isoforms silencing was designed to alter specifically CBX2.1 and CBX2.2 so that siCBX2.1-oligos are not targeting CBX2.2 and vice-versa (si-oligo sequences are available upon request).
DamiD. In principle, the DamID technique is based on in vivo expression of a chromatin protein of interest fused to DNA adenine methyltransferase (Dam) 120 . CBX2-Dam identification was achieved as previously described by Eid et al. (2015). Briefly, human CBX2.1 and CBX2.2 cDNA (OriGene, Rockville, Maryland, USA) were amplified and cloned then recombined into the destination vector to generate Dam-CBX2 construct. DamID was performed using a lentiviral transduction protocol 120 . Genomic DNA was isolated and used as templates to amplify methylated genomic fragments. DNA libraries were then prepared using the TruSeq DNA LT Sample Prep Kit (Illumina) and the libraries were sequenced on a HiSeq. 2000 sequencer (Illumina).
Illumina RNA sequencing. RNA sequencing provides far higher coverage and greater resolution of the dynamic nature of the transcriptome 121 . Total RNA was isolated and physical integrity was examined on Agilent Bioanalyser 2100 and Agilent 2200 TapeStation. Three replicates for each sample were performed for transcriptome analysis. RNA library preparation for sequencing was achieved according to the protocols of the functional genomics centre of Zurich (Switzerland) 122 . Typical cut-offs for candidate selection are Log2Ratio>1 (2x Fold) or Log2Ratio <−1 (2x Fold in the other direction) and an FDR (false discovery rate) of 0,05. The resulting p-values were FDR corrected using the Benjamini-Hochberg method. Genes with FDR values equal or smaller than ≤0,05 were considered differentially expressed.
Go enrichment analysis. GO-terms with p-values ≤ 0.05 and more than five target genes associated with the corresponding GO-term were defined as significant. CBX2.1 and CBX2.2 target genes were clustered depending on GO-terms and visualized using a spring-embed layout with Cytoscape v3. 3.0 123 . We used ToppCluster 124 to explore the functional significance of the binding patterns of CBX2.1 and CBX2.2. GO-enrichment permits to analyse functional features of gene sets, clustering them by their involvement in pathways related to Molecular Function, Biological Process and/or Cellular Component. GO-terms were considered as significantly enriched when value equal to or smaller than ≤0,05. Downstream genes were clustered according to their GO-terms. PathwayStudio 11 (Elsevier) is an exhaustive resource of easily searchable data from biology articles describing interactions between molecules, cell processes and diseases 125 . The platform allowed us to analyse the connection of CBX2 related genes in the human ovary genetic network.
Quantitative Rt-pcR. In this study, we performed the RT-PCR as a technical validation approach. Gene expressions were evaluated using KAPA SYBR FAST qPCR Kit (KAPA BIOSYSTEMS). All samples were run in triplicates and the normalized relative expression values (2 −ΔΔCt ) 126 of multiple independent experiments were plotted against the control vector set at 1. Statistical analyses were conducted using GraphPad Prism version 6.07 (Software, La Jolla, California, USA) and data sets were analysed for statistically significant differences using unpaired Student's t-test 127 with confidence intervals set at 95%. Our gene-specific primer sequences and RT-qPCR conditions for CBX2 isoforms and selected targets amplification are available upon request.