1 Reconstitution of the oocyte transcriptional network with transcription factors 2 3 Nobuhiko Hamazaki*1, Hirohisa Kyogoku2, Hiromitsu Araki3, Fumihito Miura3, Chisako 4 Horikawa1, Norio Hamada1,4, So Shimamoto1, Orie Hikabe1, Kinichi Nakashima1, 5 Tomoya Kitajima2, Takashi Ito3, Harry G Leitch5,6, Katsuhiko Hayashi*1 6 7 1 Department of Developmental Stem Cell Biology, Graduate School of Medical 8 Sciences, Kyushu University, Maidashi 3-1-1, Higashi-ku, Fukuoka 812-8582, Japan. 9 2 Laboratory for Chromosome Segregation, RIKEN Center for Biosystems Dynamics 10 Research, Kobe, Japan. 11 3 Department of Biochemistry, Graduate School of Medical Sciences, Kyushu University, 12 Fukuoka, Japan 13 4 Department of Obstetrics and Gynecology, Graduate School of Medical Sciences, 14 Kyushu University, Fukuoka, Japan 15 5 MRC London Institute of Medical Sciences (LMS), London, UK 16 6 Institute of Clinical Sciences (ICS), Faculty of Medicine, Imperial College London, 17 London, UK 18 19 Abstract 20 During female germline development oocytes become a highly specialized cell type, and 21 form a maternal cytoplasmic store of crucial factors during oocyte growth. Oocyte 22 growth is triggered at the primordial-primary follicle transition accompanied with dynamic 23 changes in gene expression1, yet the gene regulatory network underpinning oocyte 24 growth remains elusive. Here we identified a set of transcription factors sufficient to 25 trigger oocyte growth. By dissection of the change in gene expression and functional 26 screening using an in vitro oocyte development system, we identified 8 transcription 27 factors, each of which was essential for the primordial-primary follicle transition. 28 Surprisingly, enforced expression of these transcription factors swiftly converted 29 pluripotent stem cells to oocyte-like cells that were competent for fertilization and 30 subsequent cleavage. These transcription factor-induced oocyte-like cells were formed 31 without PGC specification, epigenetic reprogramming or meiosis, establishing that 32 oocyte growth and lineage-specific de novo DNA methylation is separable from the 33 preceding epigenetic reprogramming in PGCs. This study identifies a core set of 34 transcription factors for orchestrating oocyte growth, and also provides an alternative 35 source of ooplasm, which is a unique material for reproductive biology and medicine. 36 37 Main 38 In mouse germline, primordial germ cells (PGCs) go through a sequential 39 differentiation process beginning with PGC specification, followed by migration to the 40 gonad, epigenetic reprogramming (including genome-wide DNA demethylation)2, fate 41 determination3 and sex determination4. Following sex determination, female PGCs enter 42 meiosis thereby becoming oocytes. Oocytes are arrested in the diplotene stage of 43 meiotic prophase I and most are maintained in primordial follicles. Cytoplasmic 44 expansion is triggered upon activation of primordial follicles. Once oocyte growth is 45 triggered, maternal RNAs and proteins are stored in the cytoplasm5,6. Meiosis resumes 46 in full-grown oocytes, establishing metaphase II (MII) oocytes, and is not completed until 47 after fertilization. 48 Oocyte differentiation therefore entails two key processes: oocyte growth and 49 meiosis. Although concurrent, these two features are separable, as evidenced by a 50 seminal study in which Stra8-knockout mice were shown to develop oocyte-like cells 51 that do not enter meiosis. A number of genes essential for early oocyte growth have 52 been identified, including Figla7, Sohlh18, Sohlh29, Lhx810, Nobox11, Taf4b12,13, Yy114 53 and Tbpl215. Transcriptome analysis using oocytes lacking these genes8,10,15-19 and 54 identification of direct target sequences/genes7,8,20-22 has revealed downstream gene 55 cascades involved in early oocyte growth. In addition, a previous microarray analysis 56 uncovered highly dynamic gene expression changes between postnatal day 2 (P2) 57 oocytes in primordial follicles and P6 oocytes in the primary follicles (Primordial to 58 primary follicle transition: PPT)1. Genes enriched at PPT were involved in protein 59 synthesis and transcription1, suggesting a role in oocyte growth. 60 However, there has been limited further advance towards a comprehensive 61 description and functional dissection of the gene regulatory network specifically 62 orchestrating oocyte growth. Here, we identify a set of transcription factors that 63 comprise the underlying gene regulatory network and validate these findings with 64 functional screening. Furthermore, we successfully reconstitute the network in 65 pluripotent stem cells, thereby generating oocyte-like cells that are competent for 66 fertilization and subsequent cleavage divisions. 67 68 Characterization of PPT in oogenesis 69 To identify the gene regulatory network, we utilized our recently established culture 70 system that recapitulated female germline differentiation using embryonic stem cells 71 (ESCs)23. In this culture system, ESCs are differentiated into PGC-like cells (PGCLCs) 72 and then undergo oogenesis to give rise to MII oocytes in the presence of supporting 73 gonadal somatic cells. Using female germline cells in this culture system and their in 74 vivo counterparts, we mapped the trajectory of the female germline cycle by RNA-seq 75 analysis (Fig. 1a and Extended Data Fig.1a). Principal component analysis (PCA) 76 revealed that a highly dynamic gene expression change occurred in oocytes between in 77 vitro differentiation culture day 11 (IVD.D11) and day 13 (IVD.D13), which corresponded 78 to the period between postnatal day 1 (P1) and P8 in vivo. As previous microarray 79 analysis identified PPT between P2 and P6 oocytes1, we considered the transition 80 between IVD11 and IVD13 to represent PPT in vitro. Oocytes also exhibited a rapid 81 increase in their cytoplasmic and nuclear volumes between IVD11 and IVD13, as 82 observed in cytoplasmic expansion at PPT in vivo24, along with a decrease in their 83 nucleus/cytoplasmic (N/C) ratios (Fig. 1b and c). Notably, the magnitude of gene 84 expression changes in PPT was as large as the transition between MII oocytes and 85 early stage preimplantation embryos, a period known as oocyte-to-embryo transition 86 (OET), during which zygotic genome activation (ZGA) occurs25 (Fig. 1a). 87 During PPT, the upregulated genes included those encoding oocyte-specific 88 structural proteins as previously reported1. However, we also identified genes involved 89 in preimplantation development and DNA methylation of the maternal genome (Fig.1d). 90 In addition, upregulation of a subset of long terminal repeat (LTR) retrotransposon 91 families23 and the decline of the ratio of mean X chromosome transcripts in relation to 92 mean transcripts from the autosomes (X:A ratio)1 were observed at PPT (Extended 93 Data Fig.1b and c). As previous studies have described that growing oocytes use TATA 94 promoters, while ZGA is governed by a reciprocal switch from TATA promoter usage 95 back to GC promoter usage26,27, we hypothesised that the gene expression changes 96 observed during PPT might also be coupled with a change in promoter usage. We found 97 that GC nucleotides were enriched in the promoter region of genes highly transcribed 98 until IVD.D11, whereas AT nucleotides were enriched in that of genes highly transcribed 99 from IVD.D13 (Extended Data Fig. 1d). The AT nucleotides were specifically enriched at 100 20-30 bp upstream from the transcriptional start sites, and included TATA-box motifs 101 (Extended Data Fig. 1e-g). These results demonstrate that promoter usage was shifted 102 from GC promoter to TATA promoter usage upon PPT. 103 104 Functional screening of PPT genes 105 To identify key transctiption factors responsible for PPT, we applied weighted gene 106 co-expression network analysis (WGCNA) to our RNA-seq data and identified 34 107 modules (Extended Data Fig. 2a). Among these modules, we focused on five modules 108 (MEcyan, MEdarkred, MEmagenta, MEwhite and MEpurple) whose genes were 109 specifically expressed around PPT. From these modules, we extracted 27 transcription- 110 related genes that fulfilled the GO term “regulation of transcription, DNA-templated 111 (GO:0006355)” (Fig. 1e), and then investigated the functional requirement of these 112 genes for PPT by loss-of-function analysis using ESCs harboring the Blimp1-mVenus 113 (BV) and Stella-ECFP (SC) reporter genes23 (Extended Data Fig. 2b and c). Of the total 114 26 gene-knockout (KO) ESC lines, 11 differentiated normally beyond IVD13, 8 were 115 arrested between IVD11 and IVD13, and 7 were arrested at an earlier stage than 116 IVD.D3 (Extended Data Fig. 2c-e). Transcriptome analysis of these KO oocytes, except 117 for those arrested before IVD3, revealed that the 8 KO-oocytes (Figla, Sohlh1, Lhx8, 118 Nobox, Stat3, Tbpl2, Dynll1 or Sub1-KO oocytes) were arrested before or around PPT 119 (Fig. 1f). These 8 KO-oocytes did not show the representative features of PPT, such as 120 a decrease in the X/A ratio, propensity for TATA promoter usage and upregulation of 121 specific retrotransposons, with the single exception that Dynll1-KO oocytes showed a 122 decreased X/A ratio (Extended Data Fig. 3a-c). Analysis of reciprocal gene expression 123 in each line of KO-oocytes revealed that the 8 genes had a mutual effect on the 124 expression, and an imputed transcriptional network illustrated that Lhx8, Sohlh1, Nobox, 125 and Tbpl2 formed a core network to which Stat3, Dynll1, Sub1 and Figla were tightly 126 connected (Extended Data Fig. 3d and e). 127 128 PPT induction by transcription factors 129 Next we tested whether these 8 genes (hereinafter PPT8) are sufficient to drive 130 PPT and induce competence for oocyte growth. We overexpressed PPT8 in 131 BVSC“NCh”-ESCs, which consisted of BVSC ESCs with the mCherry gene inserted into 132 the Npm2 locus (Extended Data Fig. 4a). Consistent with evidence that Npm2 is 133 specifically expressed at later oogenesis stages in vivo28 and from IVD.D13 in vitro (Fig. 134 1d), Npm2-mCherry expression became visible in the nucleus of BVSCNCh-ESC- 135 derived oocytes from 15 days of IVD culture (Extended Data Fig. 4b and c). 136 Overexpression of PPT8 was controlled by the Shield1-degradation system, in which 137 PPT8 proteins become stable upon the addition of Shield1 (Fig. 2a). In suspension 138 culture with Shield1, BVSCNCh-ESCs harboring PPT8 (PPT8-BVSCNCh-ESCs) 139 immediately started to express Stella-CFP, but not Blimp1-mVenus or Npm2-mCherry, 140 and were deemed to cease proliferating, as the aggregations decreased in size (Fig. 2b 141 and Extended Data Fig. 5a). These phenotypes were transgene-dependent, because no 142 Stella-CFP expression was observed in the parental BVSCNCh-ESCs. Weak Stella- 143 CFP expression was detected in the aggregations of PPT8-BVSCNCh-ESCs without 144 Shield1, possibly due to leaky expression of PPT8, but these aggregations did not 145 decrease in size (Extended Data Fig. 5a and b). A similar phenotype was also observed 146 in PPT8-BVSCNCh-ESCs cultured in self-renewal conditions: Stella-CFP became 147 visible on day 1 of culture with Shield1, and the cells were no longer proliferative 148 (Extended Data Fig. 5c). The PPT8-BVSCNCh-ESCs in aggregations with Shield1 149 increased in both cellular and nuclear size (Extended Data Fig. 5d) and some of the 150 Stella-CFP-positive cells expressed DDX4/MVH, a conserved germ cell-specific marker 151 (Extended Data Fig. 5e). At 25 days of culture, Npm2-mCherry became visible in Stella- 152 CFP-positive cells (Fig. 2c). However, these oocyte-like cells showed irregular 153 morphology, which in some cases included an unusual cavity in the cytoplasm. 154 Because follicular somatic cells are crucial for oocyte growth, we co-cultured PPT8- 155 BVSCNCh-ESCs and E12.5 female gonadal somatic cells with Shield1, which mimicked 156 our previously described reconstituted ovaries (rOvaries)23 (Fig. 2d). Strikingly, PPT8- 157 BVSCNCh-ESC-derived oocyte-like cells grew uniformly and became positive for both 158 Stella-CFP and Npm2-mCherry with formation of follicle structures with the expression 159 of GDF9, an oocyte-secreted factor important for folliculogenesis29, and multiple layers 160 of granulosa cells (Fig. 2e-g). Notably, Npm2-mCherry was detectable from 8 days of 161 culture, which was earlier than that in PPT8-BVSCNCh-ESCs without somatic cells (Fig. 162 2c), and even earlier than that in rOvaries using PGCLCs (Extended Data Fig. 4b). 163 PPT8-BVSCNCh-ESCs in rOvaries exhibited a rapid increase in size (Extended Data 164 Fig. 5f). Without Shield1, parental BVSCNCh-ESCs or PPT8-BVSCNCh-ESCs 165 proliferated extensively in rOvaries and never formed follicle-like structures (Extended 166 Data Fig. 5g). 167 The transcriptomes of PPT8-BVSCNCh-ESCs with Shield1 alone (PPT8- 168 BVSCNCh-ESCs w/o Soma) and PPT8-BVSCNCh-ESCs cultured with Shield1 and 169 gonadal somatic cells (PPT8-BVSCNCh-ESCs+Soma) were shifted to that of P1 170 oocytes in vivo on day 5 of culture (Fig. 2h), suggesting that the initial transformation 171 process was independent of the somatic cells. On the other hand, the further shift of the 172 transcriptomes of PPT8-BVSCNCh-ESCs w/o Soma after 5 days of culture was slower 173 than those of PPT8-BVSCNCh-ESCs+Soma, suggesting that the somatic cells 174 accelerated the PPT process. Both PPT8-BVSCNCh-ESCs+Soma at 14 days of culture 175 and PPT8-BVSCNCh-ESCs w/o Soma at 25 days of culture exhibited transcriptional 176 features beyond PPT, such as the expression of maternal factor genes and specific 177 retrotransposons, a decrease in the X/A ratio and propensity for TATA promoter usage 178 (Extended Data Fig. 5h-k). These results demonstrate that the expression of PPT8 was 179 sufficient to induce PPT and the competence for oocyte growth directly in ESCs. These 180 induced oocyte-like cells, hereinafter termed directly induced oocyte-like cells (DIOLs), 181 were capable of growing to form secondary follicle structures when combined with 182 gonadal somatic cells. 183 To identify a minimum set of factors sufficient for DIOL induction, we generated 66 184 ESC lines, containing different combinations of the PPT8 transgenes (Extended Data 185 Fig. 6a-d). We found that DIOLs were induced from ESC lines harboring Nobox, Figla, 186 Tbpl2 and Lhx8 (NFTL) transgenes and no DIOLs were induced from ESC lines that 187 lacked any of these genes. However, some ESC lines harboring NFTL showed few or 188 no DIOL inductions (Extended Data Fig. 6d and e), possibly due to an insufficient level 189 and/or an inappropriate balance of the expression levels of these genes, whereas all 190 PPT8-ESC lines showed a robust induction of DIOLs. Therefore, although NFTL is the 191 minimum necessary set, we used PPT8-ESCs in subsequent experiments. 192 To assess whether PPT8 are suffcient to induce DIOLs from somatic cells, mouse 193 embryonic fibroblast (MEFs) were obtained from PPT8-BVSCNCh-ESCs (Extended 194 Data Fig. 7a). However, no DIOLs were induced from the MEFs (Extended Data Fig. 7b). 195 This suggests that a pluripotent state might be required for PPT to induce oocyte 196 features. In keeping with this notion, DIOLs were consistently induced from iPSCs 197 reprogrammed from adult tail fibroblasts (Extended Data Fig. 7c and d). These results 198 indicate that production of DIOLs from somatic cells is feasible through an iPSC 199 intermediate, emphasising the reciprocal link between the germline cycle and the 200 pluripotent state30. 201 202 Oocyte growth separable from PGC fate 203 The trajectory of the transcriptomes in the first 5 days of culture suggested that 204 DIOLs were induced from ESCs without passing through early germ cell differentiation 205 processes prior to oocyte growth. The transcriptome analysis revealed that genes 206 essential for PGC specification, such as Blimp1/Prdm1, Prdm14, and Tfap2c, were not 207 expressed during DIOL induction (Extended Data Fig. 8a). Moreover, DIOLs were 208 successfully induced from Blimp1/Prdm1-KO PPT8-BVSCNCh-ESCs, which fail to 209 specify PGCs31 (Extended Data Fig. 8b and c). These results demonstrate that DIOL 210 induction occurred directly from ESCs without transition through a PGC intermediate. 211 Following specification, PGCs undergo epigenetic reprogramming, during which 212 DNA methylation dramatically decreases to 2.9% in the genome of E16.5 oocytes, and 213 this level of DNA methylation is maintained until the non-growing oocyte stage (~2%)32-34. 214 Remarkably, methylome analysis of DIOLs at 5 days of culture with gonadal somatic 215 cells (DIOL.D5) revealed that around 27% of CpGs in the DIOL.D5 genome were 216 methylated, as observed in the parental PPT8-BVSCNCh-ESC line (Extended Data Fig. 217 9a). The pattern of DNA methylation in DIOL.D5 mirrored that in the parental ESC line, 218 whereas that in non-growing oocytes in vivo showed extensive demethylation in the 219 genome (Fig. 3a and Extended Data Fig. 9b and c). Despite a trend of severe loss of 220 DNA methylation at differentially methylated regions (DMRs) of imprinting loci in female 221 ESCs35, we found that DMRs in the H19 and Rasgrf1 loci remained at ~50% methylation 222 levels in the PPT8-ESCs and in DIOL.D5, whereas these DMRs are demethylated in 223 non-growing oocytes in vivo (Fig. 3b). These results demonstrated that no genome-wide 224 DNA demethylation occurred during DIOL induction. 225 Despite their highly methylated genome, we observed growth of DIOLs alongside 226 the proliferation of granulosa cells under an in vitro growth (IVG) condition23 (Fig. 3c). 227 Transzonal projections (TZPs) were formed between DIOLs and the surrounding 228 granulosa cells (Extended Data Fig. 9d). DIOLs became full-grown (fgDIOLs) with large 229 germinal vesicles (GVs) at 11 days of culture. Importantly, under an in vitro maturation 230 (IVM) condition, fgDIOLs underwent GV breakdown (GVB) and bore a polar body (Fig. 231 3c), demonstrating that DIOLs had a potential to reach a stage morphologically similar 232 to that of MII oocytes, which therefore we designated MII-DIOLs. During oocyte growth 233 in vivo, a progressive gain of de novo DNA methylation is accomplished in the genome 234 of oocytes32-34. Therefore, we evaluated de novo DNA methylation in fgDIOLs. 235 Methylome analysis revealed that the pattern at high-methylation regions in fgDIOLs 236 closely resembled that in full-grown oocytes in vivo (fgOocytes) (Fig. 3d and Extended 237 Data Fig. 9e). On the other hand, in low-methylation regions the pattern was more 238 similar to that of DIOL.D5 than fgOocytes (Fig. 3d). The pattern of de novo DNA 239 methylation at gene bodies, which is characteristic during oocyte growth36, was 240 correlated between fgDIOLs and fgOocytes (R2=0.738); however, the absolute levels of 241 de novo DNA methylation at these regions was lower in fgDIOLs (Fig. 3e). These results 242 suggest that de novo DNA methylation was appropriately added to the genome during 243 maturation of fgDIOLs, but on a background of elevated global DNA methylation carried 244 over from ESCs and DIOL.D5. Analysis of DMRs of the maternally methylated 245 imprinting loci (Igf2r, Impact, Mest and Snrpn) revealed that the appropriate pattern of 246 de novo methylation was observed in the genome of fgDIOLs (Fig. 3f). However, the 247 methylation level in fgDIOLs was lower than that in vivo. This could be due to the 248 heterogeneity of DIOLs, as individual reads showed that either all or none of the CpGs 249 were methylated (Extended Data Fig. 9f). Of note, more than half the reads in all loci 250 showed complete DNA methylation, suggesting that a certain number of oocytes had 251 correctly laid down maternal imprints. Taken together, these results demonstrate that 252 oocyte growth represented by cytoplasmic maturation and de novo DNA methylation in 253 the nucleus was accomplished without PGC specification or genome-wide DNA 254 demethylation. 255 256 Developmental potential of DIOLs 257 Finally, we validated the developmental potential of MII-DIOLs. 258 Immunofluorescence analysis revealed that MII-DIOLs formed a spindle-like stru