REVIEWS

Specification and epigenetic programming of the human germ line

Walfred W. C. Tang1,2, Toshihiro Kobayashi1,2, Naoko Irie1,2, Sabine Dietmann1,3 and M. Azim Surani1,2 Abstract | Primordial germ cells (PGCs), the precursors of sperm and eggs, are established in perigastrulation-stage embryos in mammals. Signals from extra-embryonic tissues induce a unique gene regulatory network in germline-competent cells for PGC specification. This network also initiates comprehensive epigenome resetting, including global DNA demethylation and chromatin reorganization. Mouse germline development has been studied extensively, but the extent to which such knowledge applies to humans was unclear. Here, we review the latest advances in human PGC specification and epigenetic reprogramming. The overall developmental dynamics of human and mouse germline cells appear to be similar, but there are crucial mechanistic differences in PGC specification, reflecting divergence in the regulation of pluripotency and early development.

Totipotent The germ line provides an enduring link between all which entails global DNA demethylation9–11, genomic The ability of a cell to give rise generations of an organism. In mammals, primordial imprint erasure12,13, X‑chromosome reactivation14–17 and to all cell types (both embryonic germ cells (PGCs) — the founder cells of the germ line reorganization of chromatin modifications18,19. At E13.5, and extra-embryonic) of an organism. — are specified during early embryonic development. XY and XX germ cells undergo mitotic arrest and mei­ Thereafter, PGCs migrate to the developing gonads, otic entry, respectively. This marks the end of the PGC Pluripotent epiblast cells where they undergo meiosis and differentiate into gam­ stage of germline development20. Cells derived from the inner cell etes. Fertilization of the oocyte by the sperm results in The precise timing of human PGC (hPGC) speci­ mass of a blastocyst that give totipotent rise to all lineages of the a zygote that gives rise to all cell lineages of fication is less clear than that of mPGCs, as direct embryo proper. an organism, including the germ line itself. Thus, speci­ studies on early post-implantation embryos are imprac­ fication of PGCs can be regarded as a crucial first step tical. The human germ line is probably established at for the acquisition of totipotency and the continuation around the time of gastrulation, at the second and of the mammalian life cycle (FIG. 1). third week of development21,22. Differences in PGC Investigations on mammalian germ cell develop­ specification mechanisms might be expected, as the ment have been primarily conducted using mouse regulation of pluripotency and early post-­implantation models. Mouse PGC (mPGC) specification is initiated development in mice and humans differ (see BOX 1 for 1Wellcome Trust/Cancer in the early post-implantation embryo, which develops details). Specified hPGCs are first observed during the Research UK Gurdon (BOX 1) 23,24 Institute, Tennis Court Road, as an egg cylinder . Bone morphogenetic protein early fourth week (approximately E24) in the extra-­ 1 University of Cambridge, (BMP) and WNT signals from extra-embryonic tis­ embryonic yolk sac wall near the allantois, which is an Cambridge CB2 1QN, UK. sues induce the expression of PR domain zinc-finger equivalent location to that of mPGCs at E8. Thereafter, 2Department of Physiology, protein 1 (PRDM1; also known as BLIMP1), the key hPGCs migrate and colonize the incipient genital ridge Development and regulator of PGC fate, in a few pluripotent epiblast cells by the early sixth week (approxi­mately E37)24. Gonadal Neuroscience, Downing 2 Street, University of at approximately embryonic day 6.25 (E6.25) . This is hPGCs (also known as gonocytes (XY) or oogonia (XX)) Cambridge, Cambridge followed by the upregulation of two other key specifi­ remain proliferative until around the tenth week, when CB2 3EG, UK. cation factors, PRDM14 (REF. 3) and transcription fac­ they asynchronously enter either mitotic quiescence (in 3Wellcome Trust/Medical tor AP2γ (encoded by Tfap2c)4. Thereafter, PRDM1, male embryos) or meiotic prophase (in female embryos). Research Council Stem Cell Institute, Tennis Court Road, PRDM14 and AP2γ form a core regulatory network Until recently, our understanding of human germline 5–7 University of Cambridge, that induces germ cell fate . At E7.25, a founder popu­ development was mainly based on extrapolation from Cambridge CB2 1QR, UK. lation of approximately 40 mPGCs are formed at the mouse research, supplemented with sporadic immuno­ 8 Correspondence to M.A.S. base of the allantois , which subsequently migrate into histochemistry studies in humans. However, advances [email protected] the hindgut and colonize the genital ridge by E11.5. To in mouse germline biology, particularly over the past doi:10.1038/nrg.2016.88 erase somatic epigenetic memories, migratory mPGCs 15 years, have informed more recent breakthroughs on Published online 30 Aug 2016 undergo genome-wide epigenetic reprogramming, the mechanism of hPGC development. Here, we review

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 1 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

Zygote

Adulthood Gametes Blastocyst Week 1 (E6) Pre-implantation epiblast Pre-im n plan atio develop tat iliz me ion rt nt Fe B i la is fo m d s r n e m in a n a a s Week 7–10 e t r i Week 2 (E14) s g i d o o o i i t n s e e c Post-implantation

M m

a epiblast

g

M

i g n r o a i Genital r t t e i a p o c ridge r n ifi o a c g n e ra d p Mesoderm m ep s m ige GC ing netic hP hPGC precursors? Week 3 (~E17)

Week 5–6 Gastrulation hPGC

Week 4 Figure 1 | Cycle of human germline development. After fertilization, the zygote develops intoNature the blastocyst, Reviews which | Genetics contains pluripotent pre-implantation epiblast cells. These cells give rise to all lineages in the embryo proper, including the germ line. As the blastocyst implants into the uterine wall, it develops a bilaminar embryonic disc and later undergoes gastrulation to form the ectoderm, mesoderm and endoderm germ layers (see also BOX 1). Human primordial germ cells (hPGCs) are probably specified around the time of gastrulation (approximately embryonic day 17(E17)), although the exact timing of hPGC specification in the embryo remains unknown. At week 4, the hPGCs localize near the yolk sac wall close to the allantois, and later migrate through the hindgut to the developing genital ridges. Migratory hPGCs undergo genome-wide epigenetic reprogramming, including global DNA demethylation, to erase imprints and other somatic epigenetic memories. During fetal development and adulthood, gonadal germ cells undergo meiosis and gametogenesis to differentiate into sperm and eggs. Concurrently, the genome is remethylated and acquires appropriate epigenetic signatures for the generation of a totipotent zygote upon fertilization.

Allantois A membranous sac that the signalling principle and the gene regulatory net­ maternally inherited determinants (known as germ develops from the mesoderm (in mice) or the hindgut work for hPGC and mPGC specification. Emerging evi­ plasm) are exclusively allocated to prospective germ cells 26 endoderm (in humans) during dence shows that there are vital mechanistic differences following fertilization (that is, preformation) . However, early embryonic development. between hPGC and mPGC specification that were not in mammals and many other animals, germ cells are The allantois contributes to the anticipated from studies in mice. We also review the latest induced in competent cells by extrinsic signals during formation of the umbilical cord advances in research on germline epigenetic reprogram­ embryogenesis (that is, epigenesis)1,27,28. and placenta. ming in the two species, with a particular focus on DNA In mice, PGC specification occurs in response to Genomic imprint demethy­lation and chromatin reorganization. These complex signalling crosstalk within the egg cylinder An epigenetic phenomenon events have implications for subsequent germline devel­ just before gastrulation1,29,30 (FIG. 2a). BMP signals pro­ that results in monoallelic gene opment and the totipotent state at fertilization, as well as duced in the extra-embryonic ectoderm (ExE) and expression in a parent-of‑­ origin-dependent manner. the transmission of genetic and epigenetic information the proximal visceral endoderm (VE) are essential for to subsequent generations. mPGC induction in the posterior–proximal epiblast Gastrulation at E6.25. Ablation of Bmp2, Bmp4 or Bmp8b31–33 or of The developmental process Mammalian germline induction genes encoding downstream signalling components in which the three germ Induction of germ cell fate by extrinsic signals. There are (for example, type 1 Activin A receptor (Acvr1; also layers (that is, the ectoderm, (REF. 34) mesoderm and definitive two basic modes for germline formation in metazoans, known as Alk2 ), and Smad1, Smad4 and Smad5 endoderm) of the embryo ‘preformation’ and ‘epigenesis’ (REF. 25). In Caenorhabditis (REFS 35–37)) in mice results in loss or a reduction in the are formed. elegans, Drosophila melanogaster and Xenopus laevis, number of mPGCs. In a comprehensive signalling study,

2 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

Ohinata et al.2 found that both BMP2 and BMP4 can form mPGCs regardless of BMP signalling1,30. WNT3 dose-dependently induce PRDM1‑positive mPGCs in is expressed in the posterior VE and epiblast at around isolated epiblasts cultured ex vivo, with BMP4 being the E5.5–E6.5 and is known to induce primitive streak genes more potent inducer2. In addition, the authors reported that are essential for gastrulation38. Although these meso­ that Wnt3 and Ctnnb1 mutants (Ctnnb1 encodes the dermal genes are initially induced in mPGC precursors, downstream signalling transducer β‑catenin) fail to they become repressed in specified E7.5 mPGCs to

Box 1 | Divergence of early embryonic development in mice and humans After fertilization, human and mouse zygotes both undergo a series of axes156. At E5.5, the distal VE (DVE) thickens and forms a specialized cleavage divisions (see the figure) and cell fate decisions, leading to the signalling region, and migrates proximal–anteriorly and becomes the formation of a blastocyst (reviewed in REFS 156,157). The blastocyst is anterior VE (AVE). Reciprocal signalling interactions between the epiblast, composed of a pluripotent inner cell mass (ICM) and an outer the ExE and the VE lead to gastrulation at the posterior epiblast at E6.5, trophectoderm (see the figure). The ICM subsequently segregates into the where the primitive streak (PS) emerges and gives rise to mesoderm and pre-implantation epiblast and hypoblast (also known as the primitive definitive endoderm (reviewed in REF. 29). endoderm). The epiblast primarily develops into the embryo proper, Humans, and possibly all non-rodent mammals, exhibit a planar structure whereas the trophectoderm and the hypoblast give rise to extra-embryonic during the perigastrulation period161. Following implantation at E7–E8, the tissues, including the placenta and the yolk sac. Pre-implantation epiblast moves away from the trophoblast, creating the amniotic cavity. The development is morphologically similar in humans and mice, although post-implantation epiblast and the underlying hypoblast flatten into a round there are notable differences in developmental timing, gene expression, bilaminar embryonic disc, which is sandwiched between the fluid-filled signalling requirement for pluripotency and lineage segregation58,158. For amniotic cavity and the yolk sac (formerly the blastocoel). Cells originating instance, although human and mouse pre-implantation epiblasts share the from the epiblast and the hypoblast form the linings of the amniotic cavity expression of some naive pluripotency factors (for example, transcription (amnion epithelium) and the yolk sac (yolk sac epithelium), respectively. factor CP2-like 1 (TFCP2L1), Krüppel-like factor 4 (KLF4) and T-box 3 (TBX3)), At around E16, gastrulation commences and the PS is formed at the some notable factors, such as oestrogen-related receptor beta (ESRRB) and posterior end of the embryonic disc and extends halfway across the epiblast KLF2, are present in mice but not in humans58,159. Furthermore, components towards the anterior region. The proliferating epiblast cells along the streak of the transforming growth factor-β (TGFβ) signalling pathway (such as undergo epithelial–mesenchymal transition and migrate to the space NODAL, growth differentiation factor 3 (GDF3) and TGF beta receptor 1 between the epiblast and primitive endoderm, giving rise to the mesoderm (TGFBR1)) are highly expressed in the human epiblast58. Inhibition of this layer. The ingressing epiblast cells also replace the hypoblast cells to become pathway in blastocysts downregulates the core pluripotency factor Nanog the definitive endoderm. This process results in the transformation of the in the human epiblast but has no detectable effect in mice, indicating that bilaminar embryonic disc into a trilaminar disc, which contains the three the regulation of pluripotency differs between humans and mice. germ layers for lineage specification. Human and mouse embryos become structurally distinct following In mice, primordial germ cell (PGC) specification occurs at the posterior implantation (see the figure). As the mouse blastocyst implants at epiblast before gastrulation2. This is dependent on the pluripotent state49 approximately embryonic day 4.5 (E4.5), the pluripotent epiblast cells and intricate signalling interaction between embryonic and extra-embryonic become polarized and transform into a cup-shaped epithelium40,160, whereas tissues in the egg cylinder1,30 (FIG. 2a). Less is known about PGC origin and the trophectoderm‑derived extra-embryonic ectoderm (ExE) forms an perigastrulation patterning in planar embryos of humans and other inverted cup on top. Concurrently, the hypoblast develops as the visceral non-rodent mammals. The divergence in pluripotency regulation and endoderm (VE) and envelops the epiblast and ExE. These changes result in embryonic structure between rodents and other mammals may culminate the formation of an egg cylinder with proximal–distal and anterior–posterior in mechanistic differences for the establishment of the germ line.

Implantation

Embryonic day 0 1 2.5 3 3.5 4.5 5.5 6.5

ExE Mouse

Blastocyst VE AVE Pre-implantation Zygote Morula ICM epiblast DVE Post-implantation epiblast PS

Trophectoderm Hypoblast Mesoderm Amnion epithelium Hypoblast Human Yolk sac epithelium

Embryonic day 0 1 3 4 5 6 14 16

Fertilization Implantation Gastrulation

Nature Reviews | Genetics NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 3 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

a Mouse (E6.5) P BMP4 and BMP2 BMP8B A P ExE D ExE • BMP4 • BMP8B AVE PS

mPGC precursors Epiblast AVE

• CER1 BMP2 WNT3 AVE (antagonistic signals) • DKK1 WNT3 • LEFTY1 PS

Epiblast PVE

b Human (E17)

Amnion epithelium • BMP4 or peripheral epiblast • BMP8B A P

Epiblast PS Amniotic cavity hPGC precursors?

Hypoblast (VE) Mesoderm Yolk sac epithelium Embryonic disc

Yolk sac • CER1 Anterior • DKK1 hypoblast BMP2 WNT3 • LEFTY1

Figure 2 | Signalling for mammalian germline induction. a | At the time of mouse primordialNature germ cell Reviews (mPGC) | Genetics specification and gastrulation, the embryo forms an egg cylinder with anterior–posterior (A–P) and proximal–distal (P–D) axes (left panel). Bone morphogenetic protein 4 (BMP4) from the extra-embryonic ectoderm (ExE), together with BMP2 and WNT3 from the posterior visceral endoderm (PVE), induces a few PR domain zinc-finger protein 1 (PRDM1)‑positive mPGC precursors at the most proximal–posterior epiblast, the site with highest levels of BMP signals1,30. The anterior VE (AVE) expresses antagonists of BMPs (for example, cerberus 1 (CER1) and left–right determination factor 1 (LEFTY1)) and WNT (for example, dickkopf 1 (DKK1)), which prevents mPGC and mesoderm induction at the anterior epiblast1. Smad2‑mutant and forkhead box H1 (Foxh1)-mutant embryos, which lack a functional AVE, exhibit ectopic induction of PRDM1‑positive cells in the anterior epiblast1. The expression of BMP8B in the ExE may have a role in restricting the AVE domain from extending to the posterior epiblast1. The expression patterns of key signals for mPGC specification are illustrated in the right panel. b | A hypothetical signalling model for PGC induction in humans and other non-rodent mammals. The mouse embryonic day 6.5 (E6.5) embryo is transformed into a planar structure (left panel)40,41, which corresponds to the trilaminar human embryo at approximately E17 (right panel). In this model, the unfolded cup-shaped mouse epiblast and VE correspond to the upper and lower layers of a planar embryonic disc, respectively, with the nascent mesoderm sitting between the two layers at the posterior end (left panel). The ExE corresponds to the structural equivalent of the amnion epithelium or the peripheral epiblast in humans, and the extra-embryonic VE corresponds to the Primitive streak yolk sac epithelium. As a result, the proximal–distal axis in the mouse egg cylinder becomes a peripheral–central axis in A structure at the posterior end humans and non-rodent mammals. The projection of key PGC specification signals predicts that BMP4 and BMP8B would of the embryo where epiblast be expressed as a ring-like domain that surrounds the epiblast, whereas the posterior hypoblast and epiblast would be cells ingress to form the mesoderm and the definitive the site of BMP2 and WNT3 signalling. Inhibitory signals from the anterior hypoblast would restrict PGC induction to the endoderm. Formation of the posterior epiblast and nascent mesoderm, where BMP dosage is the highest. It is important to note that post-implantation primitive streak is the first development is fundamentally different in humans and mice, and this model serves to translate the signalling patterns in visible sign of gastrulation. mice for comparison with those in non-rodent mammals. PS, primitive streak.

4 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

Box 2 | Pluripotent states and initial attempts of in vitro PGC induction and the mesoderm at the posterior epiblast is still unclear. It is possible that the high expression levels of BMP at Pluripotency, a transient state during early embryonic development, can be the most proximal–­posterior epiblast favours germ cell maintained in vitro as self-renewing pluripotent stem cells (for reviews, see induction over meso­derm formation (FIG. 2a). Although REFS 54,55,162). In mice, naive pluripotent cells in the inner cell mass (ICM) of the role of Nodal signalling has not been investigated, it is embryonic day 3.5 (E3.5)–E4.5 blastocysts give rise to mouse embryonic stem cells (mES cells), which are traditionally maintained in in vitro culture with fetal calf serum probably crucial for mPGC specification, as Bmp4 and 39 and leukaemia inhibitory factor (LIF)163,164. Cultured mES cells can contribute to all Wnt3 are absent in Nodal-null embryos . somatic lineages and the germ line when introduced back into blastocysts. Because The signalling mechanism for PGC specification and serum contains various undefined signalling molecules, serum–LIF-cultured mES cells gastrulation in non-rodent mammals, which develop are in metastable states and cycle in and out of naive pluripotency162. Conversely, as a planar embryonic disc instead of an egg cylinder as cultured mouse epiblast stem cells (mEpiSCs) are traditionally derived with the addition in rodents, is less clear. To extrapolate the patterning in of Activin A and fibroblast growth factor 2 (FGF2) from E5.5–E6.5 mouse post- mice to other mammals, some embryologists have pro­ 165,166 implantation epiblasts that are poised for lineage differentiation . Although posed a flattened model that unfolds the egg cylinder mEpiSCs express the core pluripotency factors Pou5f1, Nanog and SRY-box 2 (Sox2), into a planar embryonic disc40,41 (FIG. 2b). The predicted they upregulate the expression of somatic markers (for example, orthodenticle signalling pattern in this model is approximately in homeobox 2 (Otx2), eomesodermin (Eomes), forkhead box 2 (Foxa2) and T) and lack the expression of naive pluripotency genes (for example, PR-domain containing 14 line with the patterns recently observed in rabbit peri­ 42,43 (Prdm14), T-box 3 (Tbx3) and zinc-finger protein 42 (Zfp42; also known as Rex1)). gastru­lation planar embryos . In rabbits, both BMP2 They can contribute to chimeric embryos when grafted into the post-implantation and BMP4 are expressed as a ring-like domain in the epiblast62,167 but exhibit limited contribution when injected into pre-implantation peripheral epiblast and trophoblast42, whereas BMP4 blastocysts165,166. Thus, mEpiSCs are considered to be in a ‘primed’ pluripotency state. and WNT3 antagonists cerberus 1 (CER1) and dick­ Putative germ cells are generated at a very low frequency by spontaneous kopf 1 (DKK1), respectively, are expressed in the anterior differentiation of serum–LIF-cultured mES cells in vitro168,169. The fact that in vivo mouse hypoblast43. These signals probably restrict rabbit PGC primordial germ cells (mPGCs) are specified from E5.75 to E6.25 post-implantation induction and gastrulation to the posterior epiblast, epiblasts led to the speculation that mEpiSCs derived from the post-implantation where PRDM1-positive putative PGCs and the primitive epiblast might be germline competent. However, mEpiSCs also demonstrate a very streak are observed42. Currently, there is little informa­ limited capacity for PGC-like cell (PGCLC) specification (up to 1.5%)170. Recent studies show that mEpiSCs acquire properties more similar to those of anterior primitive streak tion on WNT3 and BMP8A expression in non-rodent cells than to those of pre-gastrulating epiblasts62, which might explain the low germline mammals. Notably, a similar BMP signalling pattern is competence in these cells. To achieve robust mouse PGCLC induction, an intermediate also observed in pig embryos around the time of PGC entity in between naive mES cells and primed mEpiSCs is required. Such a induction44. Overall, despite differences in development, germline-competent epiblast-like state has been established recently in vitro49 (FIG. 3a). comparative biology suggests that BMP signalling is very Human ES cells (hES cells) can also be derived from blastocyst culture171. Human and likely to be a conserved signal for PGC specification in mouse ES cells have similar embryonic origins, although they exhibit different mammals, including humans45–47 (see below). morphology, gene expression patterns and signalling requirements for the 172 maintenance of pluripotency . For example, mES cells are dependent on bone Germline competence and in vitro PGC induction. morphogenetic protein (BMP) and LIF signalling pathways, whereas conventional hES In mouse embryos, only a few cells in the proximal–­ cells require FGF and transforming growth factor-β (TGFβ) signalling173. Conversely, hES cells share similar properties with mEpiSCs and are considered to be in a primed posterior epiblast acquire the PGC fate, whereas the pluripotency state172. However, there are some notable differences between hES cells rest of the epiblast cells differentiate into the ecto­ and mEpiSCs. For example, hES cells express naive pluripotency genes PRDM14 and derm, endoderm and mesoderm. However, ex vivo cul­ ZFP42, which are absent in mEpiSCs, and both factors are essential for maintenance of ture models suggest that most of the pre-gastrulation hES cell pluripotency107,174. Thus, conventional hES cells probably represent a distinct E5.5–E6.25 epiblast cells are germline competent1,48. pluripotent state that has characteristics of both naive mES cells and primed mEpiSCs. Initial efforts to reconstitute PGC specification in vitro Previous studies have shown that conventional hES cells can differentiate into human resulted in limited success, mainly owing to the lack of PGCLCs at a low frequency (up to 5%) by spontaneous differentiation, which can be germline-competent pluripotent states (BOX 2). Taking 46,53,175 promoted by BMP signalling . The low germ cell induction capacity suggests that in vivo development into account, a two-step model hES cells do not represent a germ-cell-competent state. Recent evidence indicates that was developed to reconstitute mPGC specification cells with early mesoderm characteristics exhibit high competence for human germ cell 49 (FIG. 3a) fate45,47,61 (FIG. 3b). in vitro . Accordingly, the pre-implantation epiblast-like naive mouse embryonic stem cells (mES cells)50,51 are first differentiated into post-implantation maintain germ cell identity. However, one of the WNT3 mouse epiblast-like cells (mEpiLCs)49. Upon exposure target genes, meso­dermal factor T (also known as to BMP4 and supporting cytokines (FIG. 3a), mEpiLCs brachyury), is unexpec­tedly required for mPGC speci­ efficiently give rise to mouse PGC-like cells (mPGCLCs) Nodal signalling fication30. Although PRDM1‑positive mPGC precursors that resemble in vivo mPGCs. Remarkably, these cells A signal transduction pathway are formed in T-knockout mouse embryos, they fail to undergo gametogenesis and yield functional sperm and that is essential for the formation of the mesoderm sustain PRDM1 expression and do not upregulate the eggs after transplantation into the gonads of neonatal 49,52 and the endoderm, and for mPGC specifier PRDM14. In the absence of Wnt3, over­ and adult mice . This in vitro mPGCLC specifica­ axis determination in expression of T alone without cytokines is sufficient to tion system has since been used to address mechanistic vertebrates. Nodal signalling induce PRDM1 and PRDM14 expression by activating aspects of mPGC development (discussed below). is activated by transforming their putative enhancers. Thus, some mesodermal fac­ Until recently, a robust in vitro PGC induction model growth factor-β (TGFβ) family factors Activin and Nodal, tors induced by WNT3–β‑catenin signalling during gas­ was not available for humans. Human ES cells (hES cells) and is transduced by trulation indeed promote germ cell fate. Precisely how grown under standard culture conditions (BOX 2) are SMAD2 and SMAD3. BMPs and WNT3 cooperate to induce both germ cells believed to be in a ‘primed’ pluripotent state and exhibit

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 5 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

45,53 Trophoblast low germline competence . However, several recent specification with high efficiency (up to 50%) after The outermost layer of studies have proclaimed the establishment of naive hES 4–5 days (FIG. 3b), as opposed to less than 5% efficiency extra-embryonic tissues cell culture conditions (for a review, see REFS 54,55). with conventional hES cells45. These hPGCLCs exhibit that attaches the embryo These studies also led to a distinct hES cell type that a transcription profile that is globally similar to that of to the uterine wall and forms the placenta. self-renews under a four-inhibitors (4i) condition, com­ in vivo hPGCs. Notably, hPGCLCs and hPGCs share prising transforming growth factor β1 (TGFβ1), fibro­ a unique gene expression profile, which includes PGC Pluripotent states blast growth factor 2 (FGF2) and leukaemia inhibitory genes (for example, PRDM1, TFAP2C and NANOS3), Pluripotency refers to the factor (LIF), together with inhibitors of extracellular pluripotency genes (for example, NANOG, POU5F1 ability of a cell to differentiate signal-­regulated kinase 1 (ERK1; also known as MAPK3) (also known as OCT4), transcription factor CP2‑like 1 into any cell of the three germ layers in the embryo proper. and ERK2 (also known as MAPK1), glycogen synthase (TFCP2L1) and Krüppel-like factor 4 (KLF4)), lineage The pre-implantation epiblast kinase 3β (GSK3β), JUN amino-terminal kinase (JNK) specifiers (for example, SRY-box 17 (SOX17), GATA- represents a naive pluripotent and mitogen-activated kinase (MAPK) p38 (REF. 56). binding protein 4 (GATA4) and TEA domain transcrip­ state, whereas the post- Remarkably, hES cells cultured in 4i medium directly tion factor 4 (TEAD4)) and cell surface markers (for implantation epiblast (poised for lineage differentiation) (without any pretreatment) respond to BMP2 or BMP4 example, CD38 and alkaline phosphatase, liver/bone/ 45 represents a ‘primed’ treatment, and undergo human PGCLC (hPGCLC) kidney (ALPL)) . These hPGCLCs probably represent pluripotent state.

Lineage specifiers a Mouse Competence for mPGC fate Transcription factors that direct competent cells to differentiate 2i/LIF into a specific cell lineage. Activin A and FGF2 BMP4, BMP8B, SCF, LIF and EGF mPGCLC

2 days 6 days Naive mES cell mEpiLC

Epiblast PS

E4.5 E5.75 E6.5 E8.5 E9.5

Early post-implantation Implantation development Gastrulation Somitogenesis

b Human E6 E14 E16 E23 (week 4) E29 (week 5) PS Epiblast Mesoderm

FGF2, FGF2 BMP2 or BMP4, TGFβ1, SCF, LIF, EGF LIF and 4i and ROCKi

4–8 days Conventional 4i hES cell hPGCLC hES cell Activin A BMP4, SCF, and LIF, EGF FGF2 GSK3βi and ROCKi 2 days 4–8 days Feeder-free hiMeLC hPGCLC hiPS cell BMP4, SCF, LIF, EGF and ROCKi

Competence for hPGC fate

Nature Reviews | Genetics

6 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

pre-migratory hPGCs in which global epigenetic repro­ pre-gastrulating mEpiLCs and anterior primitive streak- gramming, including global DNA demethylation­ 45 and like mouse epiblast stem cells (mEpiSCs)47,62 (BOX 2). imprint erasure57, has just been initiated. These independent studies suggest that hPGCs may Several recent studies have shown that hES cells cul­ originate at the onset of gastrulation from meso­dermal tured in 4i medium exhibit a transcriptome that is more precursors, and not from the pre-gastrulation epiblast, similar to that of conventional primed hES cells than as in mice (FIG. 3). In fact, the specification of PGCs to that of naive pluripotent pre-implantation epiblast from the mesoderm is a well-­conserved mechanism in cells in the blastocyst58–60. The key distinction between many organisms, including a basal insect (the cricket)28 hES cells cultured in 4i conditions compared with con­ and a tetrapod ancestor (the axolotl)27. Even in mice, in ventional conditions is the upregulation of early meso­ which mPGCs are induced before mesoderm formation, dermal genes (for example, T, mix1 homeobox-like 1 mesodermal factors (such as T) play an essential part (MIXL1), runt-related transcription factor 1 (RUNX1) in regulating­ the expression of key mPGC specifiers30. and platelet-derived growth factor receptor alpha (PDGFRA)) in the cells grown in 4i conditions, which Acquisition of germline competence. How do cells might signify their competence for germ cell fate45. Two acquire competence for germline fate? Recent mouse subsequent studies have also demonstrated that cells studies reveal dynamic epigenetic changes during the bearing early mesodermal characters have an enhanced transition from naive mES cells to germline-­competent capability for hPGCLC specification47,61. In one of the mEpiLCs63–66. In particular, in mEpiLCs, regulatory reports, human induced pluripotent stem cells (hiPS elements of post-implantation epiblast genes accu­ cells) cultured in a pre-formulated commercial medium mulate the active enhancer marks histone H3 lysine 4 were induced into human ‘insipient mesoderm-­like cells’ monomethylation (H3K4me1) and H3K27 acetylation (hiMeLCs), which are germline competent and can give (H3K27ac), whereas naive pluripotency-associated rise to PRDM1- and AP2γ‑positive hPGCLCs with up to enhancers lose their active states63. Promoters of devel­ 40% efficiency when exposed to BMP2 or BMP4 (REF. 47) opmental genes and germ cell specifiers (for example, (FIG. 3b). Similarly to hES cells cultured in 4i medium, PRDM1 and PRDM14) become bivalently marked hiMeLCs express early meso­dermal genes. Compared by H3K4me3 and H3K27me3, and are considered to with mouse pluripotent stem cells, hiMeLCs are consid­ be poised for activation upon mPGC specification64. ered to be at a developmental state in between that of However, the mechanistic link between germline compe­ tence and the rewired enhancer and promoter epigenetic landscape in mEpiLCs requires further investigations. ◀ Figure 3 | Reconstitution of mouse and human PGC specification in vitro. a | The Unlike mEpiLCs and hiMeLCs, hES cells cultured in 4i embryonic day 3.5 (E3.5)–E4.5 pre-implantation epiblast in mice can give rise to naive medium can self-renew while retaining germline com­ self-renewing embryonic stem (ES) cells under basal culture conditions consisting of petence seemingly indefinitely. This state is also revers­ leukaemia inhibitory factor (LIF), glycogen synthase kinase 3β inhibitor (GSK3βi) and ible, as it can be gained or lost by switching between 4i extracellular signal-regulated kinase 1 (ERK1) and ERK2 inhibitor (referred to as 2i/LIF)50. and standard culture conditions45. The properties of Priming of naive mouse ES cells (mES cells) with the addition of Activin A and fibroblast growth factor 2 (FGF2) for 2 days yields epiblast-like cells (EpiLCs), which are these cells make them tractable models to elucidate the transcriptionally similar to the E5.75 pre-gastrulating epiblast49. Like its in vivo counterpart, mechanistic basis of germline competence in humans. mouse EpiLCs (mEpiLCs) are germline competent and can robustly give rise to mouse primordial germ cell-like cells (mPGCLCs) in response to high dosage of bone Regulatory network for mPGC specification morphogenetic protein 4 (BMP4) with BMP8B, LIF, stem cell factor (SCF) and epidermal A tripartite transcription factor network. The signals growth factor (EGF). Transcriptomic and epigenetic profiles of day 6 mPGCLCs are globally that induce mPGC specification also induce gastrula­ similar to those of in vivo E9.5 migratory mPGCs. XY mPGCLCs can differentiate into sperm tion and mesoderm formation. This accounts for the W/Wv after transplantation into the seminiferous tubules of KIT neonatal mice, which are expression of early mesoderm genes (such as T, eomeso­ 49 sterile and lack endogenous germ cells . mPGCLC-derived spermatozoa can yield healthy dermin (Eomes), homeobox A1 (Hoxa1) and Hoxb1) fertile offspring following intracytoplasmic injection of sperm into oocytes. Similarly, XX in PRDM1‑positive mPGC precursors67,68. T initially PGCLCs can mature into functional oocytes following aggregation with female gonadal 30 somatic cells and transplantation into the ovarian bursa of adult mice52. b | Despite their facili­tates mPGC specification , although somatic origin from the blastocyst, conventional human ES cells (hES cells) cultured in genes (in general) need to be repressed in nascent FGF2‑containing medium are considered to be in a primed pluripotent state similar to that mPGCs to maintain germ cell identity. The establish­ in post-implantation epiblasts and exhibit low germline competence (see also BOX 2). ment of mouse germ cell fate is governed primarily by Conventional hES cells can reversibly adapt to a recently reported ‘4i’ condition, which PRDM1, PRDM14 and AP2γ, which are upregulated in consists of transforming growth factor β1 (TGFβ1), FGF2 and LIF, together with inhibitors of mPGC precursors at E6.25–E6.75 (REFS 2–4) (FIG. 4a). The ERK1, ERK2, GSK3β, JUN amino-terminal kinase (JNK) and mitogen-activated protein kinase knockout of Prdm1, Prdm14 or Tfap2c in mice impairs p38 (REF. 56). Self-renewing 4i hES cells express early mesoderm genes and are competent mPGC specification2–4,69. The resultant embryos exhibit for human PGCLC (hPGCLC) formation in the presence of BMP2 or BMP4 and other reduced numbers of alkaline phosphatase‑positive cytokines (up to 50% efficiency)45. Alternatively, human induced pluripotent stem cells (hiPS mutant mPGCs, which are lost between E8.5 and E12.5. cells) cultured under feeder-free conditions with a pre-formulated commercial medium can give rise to hPGCLCs with approximately 20% efficiency in response to BMP4 (REF. 47). A common pheno­type observed among these cells is 2,4,67,70 When cultured under Activin A and GSK3βi for 2 days, these hiPS cells can differentiate into the derepression of mesodermal genes , suggesting human incipient mesoderm-like cells (hiMeLCs), which exhibit enhanced competence for that the three transcription factors together are essen­ hPGCLC formation (up to 40% efficiency). hPGCLCs generated from 4i hES cells and tial for suppressing the somatic programme. PRDM1 hiMeLCs share a similar transcriptome and probably represent ­pre-migratory-stage and PRDM14 are also responsible for the upregulation hPGCs47,57. PS, primitive streak; ROCKi, Rho-associated protein kinase 1 inhibitor. of germ cell and pluripotency genes, and the initiation

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 7 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

a Mouse In line with other transcription factors, PRDM1, WNT3–β-catenin BMP–SMAD PRDM14 and AP2γ are independently involved in diverse developmental processes. For instance, PRDM1 is a well-known transcriptional repressor that is involved Germ cell genes in lymphocyte differentiation71, whereas PRDM14 main­ Mesoderm genes 70,72,73 NANOG tains pluripotency in ES cells . AP2γ is expressed PRDM1 AP2γ POU5F1 T SOX2 in various tissues and plays an essential part in troph­ ectoderm specification74,75. Genome-wide maps of Pluripotency genes DNA binding show that PRDM1, PRDM14 and AP2γ PRDM14 Epigenetic cooper­ate in an intricate manner to upregulate germ reprogramming Cell proliferation cell and pluripotency genes, repress somatic fate, initi­ 5,6 Migration ate migration and reset the epigenome . In particular, PRDM1 predominantly binds to promoters5,64, whereas 72 b Human PRDM14 is enriched in distal regulatory elements . AP2γ binds to targets of both PRDM1 and PRDM14 WNT3–β-catenin BMP–SMAD to activate or repress target genes5,6. As none of these factors demonstrates histone-modifying activities, they Germ cell genes probably modu­late transcriptional activities by recruit­ PRDM14? ing epigenetic modifiers. PRDM14 was shown to inter­ NANOG POU5F1 act with SUZ12, a component of Polycomb repressive Somatic genes SOX17 complex 2 (PRC2), to confer repressive H3K27me3 in AP2 ? Pluripotency genes Mesoderm genes γ mES cells73. However, two recent interactome studies Epigenetic show that, instead of interacting with SUZ12, PRDM14 Endoderm genes PRDM1 reprogramming forms a complex with ETO family protein CBFA2T2 Cell proliferation Migration (also known as MTGR1), which has been shown to recruit histone deacetylases (HDACs) for gene silen­ 76,77 Figure 4 | Gene regulatory network models for mouse and human PGC specification. cing . Deletion of Cbfa2t2 (REFS 76,77), or interrup­ a | In mice, bone morphogenetic protein (BMP)–SMAD and WNT3–Natureβ‑catenin Reviews signalling | Genetics tion of the interaction between CBFA2T2 and PRDM14 induces expression of PR domain zinc-finger protein 1 (PRDM1) and PRDM14, as well as (REF. 76), mimics the effect of Prdm14 knockout on mesoderm genes in primordial germ cell (PGC) precursors30. The precise mechanism for the pluri­potency maintenance and mPGC specification. initial upregulation of PRDM1 at embryonic day 6.25 (E6.25) remains unclear, but it is known Similarly, PRDM1 has been reported to interact with his­ that the mesodermal factor T is required for sustaining PRDM1 expression and triggering tone modifiers, including HDAC2, lysine-­specific 30 PRDM14 expression . Maintenance of PRDM1 and of PRDM14 expression are mutually ­histone demethy­lase 1A (KDM1A), EHMT1 and protein 3,70 interdependent . PRDM1 and PRDM14 induce the expression of transcription factor arginine­ N-methyltransferase 5 (PRMT5)78. AP2γ and, together, the three transcription factors form a core specification network (pale orange box)5,7 to upregulate the expression of germ cell genes and pluripotency genes (such as Nanog, Pou5f1 and SRY-box 2 (Sox2)), repress mesoderm genes, temporarily inhibit Pluripotency factors in mPGC development. cell proliferation18, and initiate epigenetic reprogramming and migration. b | In humans, Re‑expression of pluripotency genes is a hallmark of SOX17 is critical for PGC specification and is upstream of PRDM1 expression45. PRDM1 mPGC development. Apart from POU5F1, which is represses SOX17‑induced endoderm genes, BMP- and WNT-induced mesoderm genes and expressed throughout mPGC induction79,80, other pluri­ other somatic genes. Together, SOX17 and PRDM1 establish the human germ cell potency factors, such as SOX2 (REFS 67,81) and develop­ programme, whereas the role of AP2γ and PRDM14 in human PGC (hPGC) specification mental pluripotency-associated protein 3 (DPPA3)68,82, remains to be clarified. Notably, SOX2 is absent in hPGCs. BMP signalling176 and PRDM1 are transiently repressed in epiblast cells and mPGC pre­ (REF. 177) may contribute to the rapid downregulation of SOX2 during hPGC specification. cursors and re‑expressed in specified mPGCs at approxi­ It is unclear whether BMP–SMAD signalling activates SOX17 directly or indirectly through mately E7.25–E7.75. Consistent with the presence of the activation of other transcription factors. The role of WNT3–β‑catenin signalling and T in hPGC specification remains unknown but might be less important. Arrows and blunt-ended pluripotency factors, mPGCs can readily­ dedifferentiate arrows depict positive and negative regulation, respectively. Dashed arrows indicate into pluripotent embryonic germ cells (EG cells) under 83–85 postulated regulations. appropriate culture conditions . Owing to their requirement for pre-implantation development, the involvement of pluripotency factors of epigenetic reprogramming. The latter is evident from in mPGC specification has been unclear. Notably, con­ the repression of the histone-lysine N-methyltransferase ditional knockout or knockdown of Pou5f1 (REF. 86), EHMT1, which confers H3K9me2 erasure, and other Nanog 87,88, Sox2 (REF. 89) or Sall4 (REF. 90) after specifi­ factors that confer DNA methylation, including DNA cation results in the loss of mPGCs through defects in methyltransferase 3a (Dnmt3a), Dnmt3b and ubiquitin-­ apoptosis or proliferation, highlighting the importance like PHD and RING finger domain-containing 1 (Uhrf1) of pluripotency factors in mPGC maintenance. Making (discussed below)2,3,67. Importantly, overexpression of use of the mPGCLC induction model, we recently found any two of the three transcription factors, or of Prdm14 that overexpression of Nanog alone in mEpiLCs, but alone, is sufficient to induce mPGCLCs from competent not of POU5F1 or SOX2, can induce germ cell fate66. mEpiLCs5,7. Thus, PRDM1, PRDM14 and AP2γ form a Unexpectedly, Nanog was found to induce mPGCLC tripartite transcription factor network that is necessary by direct activation of Prdm1 and Prdm14 enhancers and sufficient for PGC induction in mice. independently of BMP4. Conversely, loss of Nanog was

8 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

found to impair the efficiency of mPGCLC induction by of some hPGC genes, and the derepression of develop­ BMP4. As Nanog is expressed in the posterior epiblast mental genes45,47,57. The expression of key transcription where mPGCs are specified, BMP signalling and Nanog factors, such as SOX17, AP2γ and POU5F1, is mildly might act cooperatively to induce mouse germ cell fate affected. Interestingly, the loss of PRDM1 in humans in vivo. causes the upregulation of endoderm genes (for example, GATA4, GATA6 and FOXA2)45, indicating that PRDM1 A distinct hPGC specification network plays a part in repressing endo­dermal genes that could Crucial role for endodermal factor SOX17. otherwise be induced by SOX17 (REF. 101). Overall, Immunohistochemistry and recent RNA sequen­cing PRDM1 is important for the initi­ation of the germ cell (RNA-seq) studies have shown that hPGCs retain a core programme and the repression of somatic genes that are set of genes that are essential for mouse germ cell devel­ presumably induced by upstream BMP signalling and opment21,22,57,91. This includes specification genes (for SOX17 (FIG. 4b). PRDM1 may also have a role in initiating example, PRDM1 and TFAP2C), germline-­specific genes global DNA demethylation by repressing the expression (such as NANOS3, dead end homologue 1 (DND1), of DNMT3B47,57. deleted in azoospermia-like (DAZL) and DEAD box polypeptide 4 (DDX4)) and pluripotency factors (for Potential partnership of SOX17 and POU5F1. How example, NANOG, POU5F1 and DPPA3). Nonetheless, can SOX17, a transcription factor with diverse roles careful comparison of mouse and human PGC expres­ in the development of somatic lineages, be involved in sion profiles has revealed some key differences. For the establishment of human germ cell fate? SOX pro­ example, hPGCs lack the core pluripotency gene SOX2 teins modulate transcriptional activities with a partner (REF. 92) but express inner cell mass (ICM)-associated transcription factor that binds to an adjacent DNA naive pluripotency factors TFCP2L1 and KLF4 (REF. 57). sequence98. It is likely that SOX17 regulates transcrip­ Furthermore, hPGCs express several lineage specifiers tion at distinct sets of loci by interacting with different that are absent in mPGCs, including trophectoderm partners in different cellular contexts. Pluripotency regu­lator TEAD4 (REF. 57) and endoderm regulator factor POU5F1 is one of the known interaction part­ SOX17 (REF. 93). ners of SOX17 (REF. 102). In both human and mouse Among these differences, the absence of SOX2 ES cells, it is well established that POU5F1 acts in and the presence of the endoderm specifier SOX17 conjunction with SOX2 to maintain pluripotency95. are particularly noteworthy, as SOX2 is required for Interestingly, POU5F1 ‘switches’ partners from SOX2 mPGC proliferation89, whereas SOX17 is dispensable to SOX17 during primi­tive endoderm differentiation for mPGC specification94. Although both SOX17 and from mES cells103. As a result, SOX17–POU5F1 binds to SOX2 belong to the SRY-related HMG-box transcrip­ the enhancers of endoderm genes and apparently acti­ tion factor family, it is unlikely that SOX17 is simply vates their expression, inducing primitive endoderm a replacement for SOX2 in hPGCs, as these proteins fate. Notably, one of the targets of SOX17–POU5F1 in belong to different subfamilies. SOX2, together with mice is Prdm1 (REF. 103). As POU5F1 is highly expressed SOX1 and SOX3, belongs to the SOXB1 subfamily, throughout hPGCLC induction from competent hES which has crucial roles in pluripotency maintenance cells45, it is possible that a similar switching of partners and neuroectoderm differentiation95. SOX17, together might occur for the establishment of human germ cell with SOX7 and SOX18, belongs to the SOXF subfamily, fate. A possible scenario is that both germ cell genes and which is essential for endoderm differentiation96, fetal endodermal genes are poised for SOX17–POU5F1 acti­ haematopoiesis97 and cardiovascular development98. vation in human germline-competent cells. One of the Moreover, SOX17 cannot replace SOX2 in pluripotency potential downstream targets, PRDM1, may then act in maintenance and somatic cell reprogramming99. In fact, tandem to activate germ cell genes, and repress endo­ overexpression of SOX17 in mES cells induces exit from dermal genes and other somatic genes. If so, the rapid pluripotency and initiates endoderm differenti­ation100. downregulation of SOX2 upon hPGCLC specification By contrast, the closely related SOXB1 subfamily mem­ and its continual absence in hPGCs might be necessary bers, SOX1 and SOX3, can replace the function of to allow the SOX17–POU5F1 interaction and avoid SOX2 as a regulator of pluripotency and neuron dif­ neuro­ectoderm differentiation. The correct dosage of ferentiation95,99. It is important to note that all SOXB1 SOX17 and POU5F1 may also be important for hPGC members (that is, SOX1, SOX2 and SOX3) are absent in fate, as both transcription factors exhibit dose-depend­ hPGCs57,92, suggesting that SOXB1‑mediated functions ent action in directing differentiation100,104. Intriguingly, are not required in hPGCs. although PRDM1 represses endodermal factors in Surprisingly, SOX17 is among the first transcription hPGCs, it is also expressed in the primitive endoderm factors to be upregulated during the specification of where the endodermal genes are expressed103, indicating 45 Inner cell mass hPGCLCs from competent hES cells . The knockout that the repressive role of PRDM1 is context depend­ (ICM). A compact mass of cells of SOX17 in competent hES cells abolishes hPGCLC ent. Another SOX family protein, SOX15, is highly located at the embryonic pole specification, and this can be rescued by SOX17 over­ expressed in both hPGCs and mPGCs47,91. SOX15 is the of the blastocyst. The ICM expression. PRDM1 expression is compromised in the sole member of the SOXG subfamily and is co‑­expressed gives rise to the epiblast and (REF. 45) 105 the hypoblast, which form the absence of SOX17 , indicating that it acts down­ with SOX2 in mES cells . SOX15 physic­ally interacts 2,69 embryo proper and the yolk stream of SOX17. As in mice , the loss of PRDM1 dur­ with POU5F1 (REF. 105) and shares some common targets sac, respectively. ing hPGCLC specification results in the down­regulation with SOX2 in mES cells106. However, the loss of Sox15

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 9 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

in mES cells does not affect the maintenance of pluri­ reprogramming in the early germ line erases parental potency. In addition to some SOX family proteins, the epigenetic memories and facilitates gametogenesis. trophectoderm specifiers­ AP2γ and TEAD4, and the Here, we focus on epigenetic resetting in the mouse and pluripotency factors Nanog, TFCP2L1 and KLF4, all human early germ line, and discuss its potential effect show early expression in hPGCLCs45,91. Their potential on germ cell maintenance, differentiation and epigenetic functions and combinatorial effects in ­hPGCLC speci­ inheritance. fication merit further investigations. Genome-wide DNA demethylation. In mice, PGCs are The paradox of PRDM14 in hPGC development. specified from the post-implantation epiblast, which is As PRDM14 is indispensable and sufficient for PGC hypermethylated and primed for lineage differentiation. specification in mice3,7,70, its role in hPGCs is of par­ To reset the epigenome, mPGCs undergo genome-wide ticular interest. PRDM14 is strongly expressed in DNA demethylation11, X‑chromosome reactivation14–17 germline-competent hES cells but is rapidly downregu­ and chromatin modification reorganization19 as they lated during hPGCLC induction by BMP2 or BMP4 migrate and colonize the genital ridge from E7.5 to (REF. 45). Thereafter, PRDM14 is modestly expressed in E13.5 (FIG. 5). During this time, global CpG methyla­ specified hPGCLCs. As PRDM14 has an integral role in tion levels drop from approximately 70% in the epi­ the core pluripotency circuit in hES cells107,108, its rapid blast to exceptionally low levels of approximately 4% suppression is probably necessary for the exit from pluri­ in E13.5 mPGCs9,111. As a result, almost all genomic potency. Indeed, the overexpression of PRDM14 in hES features, including imprint control regions, become cells upon random differentiation prevents the upregu­ hypomethy­lated. Methylation is then re‑established lation of genes of the three germ layers (for example, in a sex-specific manner after E13.5 in males and after T, MIXL1, GATA4 and paired box 6 (PAX6))108. Partial birth in females109. knockdown experiments suggest that PRDM14 may not The current view is that global DNA demethylation be required for hPGCLC specification61, although addi­ in the germ line is primarily achieved through a passive tional evidence is required to verify this result. In mice, mechanism. Shortly after mPGC specification, PRDM1 PRDM14 also facilitates global DNA demethylation by and PRDM14 repress the expression of DNMT3A and repressing the expression of de novo methyltransferases DNMT3B, as well as UHRF1 (REFS 5,67), a recruitment Dnmt3a and Dnmt3b70, but its role in hPGC reprogram­ factor of DNMT1 that is essential for the maintenance of ming is unclear. Gonadal hPGCs undergoing global DNA methylation112. As a result, both maintenance and DNA demethylation exhibit low and heterogeneous de novo methylation activities are apparently repressed, expression of PRDM14 (REF. 91), with predominant local­ enabling replication-coupled DNA demethylation when ization in the cytoplasm45. Moreover, even in the pres­ mPGCs proliferate9,113–115. ence of nuclear PRDM14, conventional hES cells exhibit Recent evidence suggests that the enzymatic con­ very strong expression of DNMT3A and DNMT3B, and version of 5‑methylcytosine (5mC) to 5‑hydroxy­ the hES cell genome remains highly methylated57. Thus, methylcyotsine (5hmC) also plays a part in cumulative evidence suggests that PRDM14 may have a demethylation, especially for imprints, which are pro­ less prominent role in hPGC development. Intriguingly, tected from demethylation until the mPGCs settle in the when transfected into mES cells, human PRDM14 can genital ridge at around E9.5–E10.5 (REF. 10). Ten-eleven substitute for mouse PRDM14 to rescue defects in translocation enzymes (TET1, TET2 and TET3) oxi­ Prdm14‑knockout mES cells76. This suggests that the dize 5 mC to 5hmC and other downstream derivatives, two orthologues are functionally conserved and that which can subsequently be removed enzymatically and the behaviour of human PRDM14 may hinge on precise replaced by unmethy­lated cytosine, or can be diluted interacting partners and cellular contexts. The paradox passively during replication116. Concurrent with TET1 of PRDM14 in human germline development remains and TET2 upregulation at E9.5–E11.5, a transient global to be fully addressed. increase in 5hmC coupled with a reduction of 5mC is Overall, the crucial role of SOX17 in hPGC speci­ observed in mPGCs10,117. Nonetheless, knockout ­studies fication and the apparent diminished role of PRDM14 in mice suggest that a large extent of germline DNA suggest a clear divergence in the mechanisms for PGC demethylation can still occur independently of TET1 specification and the initiation of epigenetic reprogram­ and TET2 (REFS 118–120). However, TET1 and TET2 ming between humans and mice. It will be of interest are required for efficient erasure of imprints118,119 and to establish whether SOX17‑mediated germline induc­ demethylation of meiotic gene promoters for germ cell tion is a general mechanism that is conserved among differentiation120. non-rodent mammals, which develop as a planar Recently, three independent studies showed that the embryonic disc. overall DNA demethylation dynamics in hPGCs are similar to those in the mouse germ line57,91,121 (FIG. 5a). Resetting the germline epigenome At week 5, migratory hPGCs in the hindgut already In mammals, global epigenetic reprogramming occurs exhibit lower levels of 5mC than neighbouring soma57. during pre-implantation development and in the early As the hPGCs settle in the genital ridge, DNA methy­ germ line. Reprogramming during pre-implantation lation levels­ drop to a minimum of approximately development resets the zygotic epigenome for naive 4.5% by week 8 (REFS 57,91). Interestingly, germ cells pluripotency (reviewed in REFS 109,110), whereas the of both sexes (isolated by the hPGC surface marker

10 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

KIT (also known as SCFR)) stay hypomethylated Notably, naive pluripotency genes (for example, at week 19 (REF. 91), around 9 weeks after the start of TFCP2L1 and KLF4) are expressed in the human ICM, meiotic entry in females and mitotic quiescence in naive hES cells and hPGCs, all of which exhibit a glo­ males. Immunohistochemistry studies show that this bally hypomethylated genome57,60,134,135. It is tempting KIT-positive population represents rare germ cells that to speculate that naive pluripotency factors may have retain hPGC characteristics, whereas the majority of roles in facilitating genome-wide DNA demethyla­ germ cells differentiate into KIT-negative pre-spermato­ tion and/or safeguarding the globally hypomethylated gonia or oogonia122–124. Human oogonia probably remain genome. However, the comprehensive extent of DNA hypomethylated before birth, as in mice, but it remains demethy­lation seen in hPGCs (down to 4% average unclear whether the more advanced male germ cells at CpG ­methy­lation in contrast to 30–40% in the ICM and week 19 have begun remethylation125. naive hES cells) suggests that additional factors are in As in mPGCs, DNA demethylation in hPGCs is place to facilitate robust DNA demethylation. associated with the repression of UHRF1, DNMT3A and DNMT3B. Furthermore, hPGCs exhibit transiently ‘Escapees’ from global DNA demethylation. Despite high levels of 5hmC, which are coupled with TET1 and global DNA demethylation, some genomic loci — referred TET2 upregulation57. Notably, most imprints are already to as ‘escapees’ — remain methylated in hPGCs and demethylated when hPGCs arrive at the genital ridge57, mPGCs9–11,57. In both species, the vast majority of escap­ indicating earlier imprint erasure dynamics in humans ees are associated with retrotransposable elements9–11,57,91. compared with mice. In particular, evolutionarily young and potentially haz­ ardous retrotransposons, such as IAP in mice136 and Chromatin reorganization safeguards the germ line. SVA (SINE–VNTR–Alu) in humans57, remain relatively DNA methylation is an important epigenetic silencer highly methylated, which may contribute to their repres­ that modulates gene expression and maintains genome sion. Escapees are also found in pericentromeric satel­ stability in mammalian cells126. Loss of DNA methy­ lite repeats57 and in sub­telomeric regions11. Retention lation in somatic lineages causes derepression of of DNA methylation in these regions may maintain ­retrotransposons, proliferation defects and cell death. chromosome stability, and ensure proper chromosome Notably, hypomethylated hPGCs and mPGCs remain alignment and segregation during mitosis. Notably, proliferative without signs of global gene or retrotrans­ a minority of escapees are single-copy sequences9–11. poson activation57. Similarly to PGCs, Dnmt1, Dnmt3a In humans, ‘repeat-poor’ escapees are widely distributed and Dnmt3b triple-knockout mES cells can self-renew in in the genome, including at promoters, gene bodies and the absence DNA methylation127, although they perish enhancers57. Some escapees are associ­ated with genes that upon differentiation. Recent studies in mES cells suggest are expressed in the brain and participate in neuronal that global DNA demethylation triggers reorganization differentiation, whereas others are found in ubiquitously of repressive chromatin modifications to repress retro­ expressed genes, such as the circadian regu­lator casein transposons and maintain genome stability128,129. In fact, kinase 1 delta (CSNK1D)57. The functions of these repeat- DNA demethylation in mPGCs is also accompanied by poor escapees, particularly those found in ­regulatory global chromatin modification reorganization18,19. Apart regions, remain to be determined. from the depletion of H3K9me2, repressive H3K27me3 and H2A/H4R3me2s (symmetrical dimethylation of Potential mechanisms for escapee methylation. The arginine 3 on H2A and H4) are both enriched during existence of escapees suggests that DNA methylation the course of mPGC development, whereas H3K9me3 pathways are not completely repressed in PGCs. In is retained predominantly at pericentric heterochroma­ both hPGCs and mPGCs, DNMT1 remains strongly Retrotransposons tin19,130 (FIG. 5a). In addition to modulating gene expres­ expressed in the nucleus, whereas DNMT3A, DNMT3B DNA elements that can amplify sion, these marks have been implicated in the repression and UHRF1 are not detectable57,113,121. DNMT1 is gener­ themselves in the genome. 131 During the process of of retrotransposons in mPGCs . For instance, loss of ally regarded as a maintenance DNA methyltransferase, retrotransposition, H3K9 N-methyltransferase SETDB1 expression causes although recent studies suggest that DNMT1 exhibits retrotransposon DNA is first depletion of H3K9me3 and H3K27me3 in endogenous de novo methylation activity on unmethylated DNA both transcribed into RNA, then retroviruses (ERVs)132, which is associated with the in vivo and in vitro137. Given these findings, it is possible reverse transcribed into DNA, derepres­sion of many ERVs (for example, intracister­ that DNMT1 acts independently of UHRF1 to confer followed by insertion into a new genomic site. nal A particle (IAP)) and reduced numbers of male and/or to maintain methylation at escapee loci. Escapees germ cells. Likewise, loss of arginine methyltransferase in humans are enriched for H3K9me3 and KRAB- Krüpple-associated box PRMT5 in mPGCs results in male and female sterility associated protein 1 (KAP1; also known as TIF1β)- zinc-finger protein that is associ­ated with the depletion of H2A/H4R3me2s binding sites57, suggesting that DNMT may be targeted (KRAB-ZFP). The largest Krüppel-associated individual family of and the upregulation of long interspersed element 1 to the escapees through the KAP1– 133 transcriptional repressors in (LINE1) and IAP elements . Global reorganization box zinc-finger protein (KRAB-ZFP) co‑­repressor com­ mammals. KRAB-ZFPs contain of repressive histone modifications is also observed in plex. The zinc-finger domain of KRAB-ZFP binds to DNA-binding C2H2 zinc-fingers hPGCs, albeit with slightly different dynamics (FIG. 5a). specific DNA sequences, whereas the KRAB domain and a KRAB domain that Thus, it is likely that one of the purposes of germline interacts with KAP1, which recruits HDACs, histone interacts with the KRAB-associated protein 1 chromatin reorganization is to safeguard genome methyltransferase SETDB1 and/or DNMT1 for hetero­ 138 (KAP1) co-repressor complex integrity while PGCs undergo DNA demethylation to chromatin formation . In humans, there are approxi­ for epigenetic silencing. ­exceptionally low levels (FIG. 5b,c). mately 400 KRAB-ZFPs, presumably with different

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 11 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

DNA-binding specificities and hence different targets. attention in recent years144–146. As DNA methylation is For instance, ZFP91 and ZFP93 have been shown to tar­ stably heritable through cell division, it is a candidate get and repress human SVA and L1PA retro­transposons, for the transmission of epigenetic information through respectively139, whereas ZFP57 maintains methylation the germ line. Indeed, a classical example of epigenetic at imprinted loci140,141. Many KRAB-ZFPs are highly inheritance involving DNA methylation is the trans­ expressed in hPGCs and may be involved in recruiting mission of genomic imprints147,148. Parent-of‑origin- DNMT1 to confer DNA methylation at retro­transposon specific DNA methylation is established at imprint and single-copy escapee loci57. Parallel mechanisms, such control regions concurrently with global remethylation as PIWI–PIWI-interacting RNA (piRNA) pathways, during gametogenesis and is transmitted to the zygote may also be involved in the methy­lation of retrotrans­ following fertilization. Imprints are maintained during posons57,91,142. Notably, PRDM9 has been shown to be global demethylation in the pre-implantation embryo essential for meiotic recombination143, suggesting that and persist in somatic cells, but they are erased dur­ KRAB-ZFPs may also have other crucial functions dur­ ing germline reprogramming and re‑established in ing germ cell development. each generation. Genomic imprinting is therefore an intergenerational epigenetic inheritance phenomenon. Epigenetic inheritance through DNA methylation. Indeed, the comprehensive nature of germline methy­ The inheritance of non-genetic information through lation and, to a lesser extent, DNA demethylation the mammalian germ line has received considerable during pre-implantation development would reduce

a

E6.5 E7.5 E8.5 E9.5 E10.5 E11.5 E12.5 E13.5 Mouse Global level of epigenetic modification 5mC Mouse High Low H3K9me2 Human High Low H3K9me3 (pericentric heterochromatin) Safeguard hypomethylated genome? H3K27me3

H2A/H4R3me2s

Human Week 3 Week 4 Week 5 Week 6 Week 8 Week 10

b Function of DNA demethylation c Genome stability Imprint erasure Retrotransposons (evolutionarily old)

LINE-L2

Retrotransposons (evolutionarily young) X reactivation Xa Xa

SVA

Xi Xa Pericentromeric repeats

Activation of meiotic and genome defence genes

DAZL Unmethylated CpG Methylated CpG

Nature Reviews | Genetics 12 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

Transgenerational the likelihood of transgenerational­ epigenetic inheritance whereas small RNAs in the sperm may contribute to epigenetic inheritance (TEI) through DNA methylation146. Moreover, a large intergenerational inheritance of diet-induced metabolic (TEI). Transmission of 151,152 epigenetic information through portion of the genome becomes fully remethylated disorders . Taken together, recent evidence indicates the germ line that affects during gametogenesis, representing another barrier that DNA methylation is less likely to be the primary phenotypic traits in more than to TEI. However, recent studies have identified some mechanism for environment-induced epigenetic inheri­ one generation without single-­copy loci and retrotransposon loci that can tance, but this cannot be entirely excluded at present. changes in DNA sequence. escape both waves of demethylation9,11,57, and some of Moreover, it remains possible that some retrotransposon these loci remain partially methylated in gametes57. escapees, such as IAP in mice and SVA in humans, may It is possible that methylation at these escapee loci give rise to metastable epialleles, which are vulnerable to might be susceptible to environmental factors, and environmental influence. This is exemplified by agouti that such epigenetic information could be potentially viable yellow (Avy) mice and axin fused (AxinFu) mice, transmitted to subsequent generations with phenotypic in which an IAP insertion in the vicinity of A or AxinFu consequences. causes ectopic gene expression and TEI phenotypes144. The inheritance of environment-induced metabolic and behavioural traits has been reported in mammals, Conclusions and perspectives with most phenotypes persisting for one to two gener­ Humans and mice diverged approximately 60 million ations (reviewed by REFS 144,145). For instance, male years ago, and since then mice have acquired a unique mice exposed to in utero undernutrition yield off­ egg cylinder structure for perigastrulation devel­ spring that exhibit obesity and glucose intolerance149; opment153, whereas humans and other non-­rodent however, although there are locus-specific DNA mammals develop as a planar embryonic disc. The methylation changes in the sperm of exposed males, germ line of both species arises during this period differential methylation is not detected in the offspring. of developmental divergence. Recent breakthroughs More recently, it was shown that the paternal diet does in in vitro hPGCLC specification and direct studies not have consistent effects on the sperm methylome150, on in vivo hPGCs have provided extensive informa­ tion on human germline development. Although BMP signalling appears to be a conserved pathway for ◀ Figure 5 | Epigenetic reprogramming in mouse and human PGCs. a | Genome-wide human and mouse germ cell induction, the competent DNA demethylation occurs in mouse primordial germ cells (mPGCs) during the onset of state and gene regulatory network for PGC specification migration and early settlement at the genital ridge (at approximately embryonic day 7.5 have diverged between the two species. Nevertheless, (E7.5)– E13.5). Concurrently, mPGCs exhibit global chromatin modification reorganization. the extent and the dynamics of epigenetic reprogram­ In particular, histone H3 lysine 9 dimethylation (H3K9me2) — a repressive chromatin ming in hPGCs and mPGCs appear to be similar, albeit modification associated with silenced genes — becomes globally depleted. As there is not identical. Importantly, the mechanism of initiation 178 intimate crosstalk between H3K9me2 and DNA methylation pathways , loss of H3K9me2 of the epi­genetic programme may differ owing to the in mPGCs may facilitate or be a consequence of global DNA demethylation. Conversely, involvement of SOX17 as a key germline specifier, H3K9me3 levels remain high at pericentric heterochromatin, whereas repressive and to an as yet uncertain contribution of PRDM14. H3K27me3 levels become progressively enriched globally. Symmetrical dimethylation of arginine 3 on H2A and H4 (H2A/H4R3me2s) is transiently increased from E8.5 to E10.5, Notably, both germ lines feature some single-copy and along with nuclear localization of the protein arginine N-methyltransferase 5 (PRMT5)130,133. retrotransposon demethylation escapees, which may be The overall global epigenetic reprogramming dynamics in the human germ line are similar important for genome stability and could have potential to those in mPGCs, but with subtle differences. Notably, the loss of 5‑methylcytosine (5mC) for epigenetic­ inheritance. in association with the loss of H3K9me2 is also observed in human PGCs (hPGCs). However, In the light of these discoveries, new questions H3K27me3 is only transiently enriched in hPGCs during migration and becomes depleted emerge that merit further investigation. What is the after PGCs arrive at the genital ridge57,179. H2A/H4R3me2s levels stay constant throughout molecular basis for germline-competent states? How do hPGC development (W.W.C.T., unpublished observations). The broken areas in the profiles transcription factors with diverse functions in different indicate postulated dynamics in hPGCs. b | Germline DNA demethylation erases parental lineages act combinatorially to specify and maintain epigenetic memories and drives germ cell differentiation. Allele-specific methylation at germ cell fate? What triggers epigenetic reprogramming imprinting control regions (ICRs) are erased in PGCs and re‑established in a sex-specific manner later in development. In XX PGCs, loss of DNA methylation and repressive in hPGCs? How are escapees targeted, and do they have H3K27me3 at the inactivated X chromosome (Xi) lead to X reactivation. This ensures that functional significance in the germ line? As functional each oocyte contains an active X chromosome (Xa) for early embryonic development after studies in human are not possible, further advances in fertilization. In somatic cells, some germ-cell-specific meiotic genes and genome defence in vitro hPGC development are required to address some genes (for example, DEAD box polypeptide 4 (DDX4), deleted in azoospermia-like (DAZL), of these questions. Under current conditions, hPGCLCs synaptonemal complex protein 3 (SYCP3), PIWI-like genes and Krüppel-associated box do not progress beyond the pre-migratory stage and zinc-finger protein (KRAB-ZFP) genes) are silenced by promoter DNA methylation. therefore do not undergo comprehensive epigenetic Demethylation of their CpG island promoters in PGCs facilitates germ cell differentiation reprogramming or upregulate the expression of meiotic and maintains genome stability. c | Despite global loss of DNA methylation, both genes45,47. Development of in vitro conditions that enable evolutionarily old (for example, long interspersed element subgroup 2 (LINE‑L2)) robust differentiation of hPGCLCs towards later gonadal and evolutionarily young (for example, the mouse intracisternal A particle (IAP) and the human endogenous retrovirus group K (HERVK)) retrotransposons remain repressed in stages and beyond will be essential. Indeed, a recent PGCs. Repressive chromatin modifications are probably redistributed and/or retained at report shows that haploid mouse spermatid-like cells repetitive sequences to safeguard genome stability. Notably, some evolutionarily young can be derived from mES cells using an entirely in vitro retrotransposons (for example, the LTR of the IAP in mice and SINE–VNTR–Alu (SVA) in stepwise protocol154. These haploid cells apparently humans) and pericentromeric satellite repeats remain partially methylated, which may yield fertile offspring after injection into oocytes, albeit contribute to their repression. at a low frequency (approximately 2%). Observations in

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 13 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

this study have yet to be reproduced, but this report has the mechanism for neuroectoderm specification has also raised the prospects for in vitro reconstitution of meiosis diverged between humans and mice155. In view of these and gametogenesis in humans. differences, studies on non-rodent mammals, such as Mice have been and will remain a valuable model for rabbits42, pigs44 and non-human primates, in combina­ mammalian development. However, studies on human tion with in vitro human models, will provide insights germline development suggest that not all observations into the mechanisms of PGC specification and other in mice can be faithfully extrapolated to humans. Indeed, early cell fate decision in humans.

1. Ohinata, Y. et al. A signaling principle for the specification and early development of germ cells in 43. Idkowiak, J., Weisheit, G., Plitzner, J. & Viebahn, C. specification of the germ cell lineage in mice. Cell 137, mice. Dev. Biol. 278, 440–458 (2005). Hypoblast controls mesoderm generation and axial 571–584 (2009). 20. McLaren, A. Primordial germ cells in the mouse. patterning in the gastrulating rabbit embryo. A comprehensive signalling study that shows that Dev. Biol. 262, 1–15 (2003). Dev. Genes Evol. 214, 591–605 (2004). BMP–phosphorylated SMAD (pSMAD) and WNT3 21. De Felici, M. in Oogenesis (eds Coticchio, G., 44. Valdez Magana, G., Rodriguez, A., Zhang, H., Webb, R. signalling pathways are indispensable for mPGC Albertini, D. F. & De Santis, L.) 19–37 (Springer, & Alberio, R. Paracrine effects of embryo-derived specification from the post-implantation epiblast 2012). FGF4 and BMP4 during pig trophoblast elongation. during a restricted time window. 22. Leitch, H. G., Tang, W. W. & Surani, M. A. Primordial Dev. Biol. 387, 15–27 (2014). 2. Ohinata, Y. et al. Blimp1 is a critical determinant of germ-cell development and epigenetic reprogramming 45. Irie, N. et al. SOX17 Is a critical specifier of human the germ cell lineage in mice. Nature 436, 207–213 in mammals. Curr. Top. Dev. Biol. 104, 149–187 primordial germ cell fate. Cell 160, 253–268 (2015). (2005). (2013). This paper shows robust in vitro induction of 3. Yamaji, M. et al. Critical function of Prdm14 for the 23. Fuss, A. Über extraregionare Geschlechtszellen bei hPGCLCs from germline-competent hES cells and establishment of the germ cell lineage in mice. einem menschlichen Embryo von 4 Wochen. Anat. Am. illustrates that SOX17 is essential for hPGC Nat. Genet. 40, 1016–1022 (2008). 39, 407–409 (in German) (1911). specification. This paper shows that Prdm14 is crucial for the 24. Witschi, E. Migration of the germ cells of human 46. Kee, K., Gonsalves, J. M., Clark, A. T. & Pera, R. A. upregulation of pluripotency genes and the embryos from the yolk sac to the primitive gonadal Bone morphogenetic proteins induce germ cell initiation of epigenetic reprogramming during folds. Contrib. Embryol. 32, 67–80 (1948). differentiation from human embryonic stem cells. mPGC specification. 25. Extavour, C. G. & Akam, M. Mechanisms of germ cell Stem Cells Dev. 15, 831–837 (2006). 4. Weber, S. et al. Critical function of AP‑2γ/TCFAP2C specification across the metazoans: epigenesis and 47. Sasaki, K. et al. Robust in vitro induction of human in mouse embryonic germ cell maintenance. preformation. Development 130, 5869–5884 (2003). germ cell fate from pluripotent stem cells. Cell Stem Biol. Reprod. 82, 214–223 (2010). 26. Weismann, A., Parker, W. N. & Rönnfeldt, H. The Germ- Cell 17, 178–194 (2015). This article shows that Tfap2c, together with plasm: A Theory of Heredity (Scribner, 1893). 48. Tam, P. P. & Zhou, S. X. The allocation of epiblast cells Prdm1, represses mesodermal gene expression 27. Chatfield, J. et al. Stochastic specification of to ectodermal and germ-line lineages is influenced by during mPGC specification. primordial germ cells from mesoderm precursors in the position of the cells in the gastrulating mouse 5. Magnusdottir, E. et al. A tripartite transcription factor axolotl embryos. Development 141, 2429–2440 embryo. Dev. Biol. 178, 124–132 (1996). network regulates primordial germ cell specification in (2014). 49. Hayashi, K., Ohta, H., Kurimoto, K., Aramaki, S. & mice. Nat. Cell Biol. 15, 905–915 (2013). 28. Ewen-Campen, B., Donoughe, S., Clarke, D. N. & Saitou, M. Reconstitution of the mouse germ cell 6. Magnusdottir, E. & Surani, M. A. How to make a Extavour, C. G. Germ cell specification requires zygotic specification pathway in culture by pluripotent stem primordial germ cell. Development 141, 245–252 mechanisms rather than germ plasm in a basally cells. Cell 146, 519–532 (2011). (2014). branching insect. Curr. Biol. 23, 835–842 (2013). This paper shows robust in vitro induction of 7. Nakaki, F. et al. Induction of mouse germ-cell fate by 29. Tam, P. P. & Loebel, D. A. Gene function in mouse mPGCLCs from ground-state mES cells through transcription factors in vitro. Nature 501, 222–226 embryogenesis: get set for gastrulation. Nat. Rev. a post-implantation epiblast-like state. (2013). Genet. 8, 368–381 (2007). 50. Ying, Q. L. et al. The ground state of embryonic stem This study, along with reference 5, shows that 30. Aramaki, S. et al. A mesodermal factor, T, specifies cell self-renewal. Nature 453, 519–523 (2008). overexpression of Prdm1, Tfap2c and/or Prdm14 mouse germ cell fate by directly activating germline 51. Boroviak, T., Loos, R., Bertone, P., Smith, A. & is sufficient for the induction of mPGC fate. determinants. Dev. Cell 27, 516–529 (2013). Nichols, J. The ability of inner-cell-mass cells to self- 8. Lawson, K. A. & Hage, W. J. Clonal analysis of the 31. Ying, Y. & Zhao, G. Q. Cooperation of endoderm- renew as embryonic stem cells is acquired following origin of primordial germ cells in the mouse. derived BMP2 and extraembryonic ectoderm-derived epiblast specification. Nat. Cell Biol. 16, 516–528 Ciba Found. Symp. 182, 68–84; discussion 84–91 BMP4 in primordial germ cell generation in the (2014). (1994). mouse. Dev. Biol. 232, 484–492 (2001). 52. Hayashi, K. et al. Offspring from oocytes derived from 9. Seisenberger, S. et al. The dynamics of genome-wide 32. Lawson, K. A. et al. Bmp4 is required for the in vitro primordial germ cell-like cells in mice. Science DNA methylation reprogramming in mouse primordial generation of primordial germ cells in the mouse 338, 971–975 (2012). germ cells. Mol. Cell 48, 849–862 (2012). embryo. Genes Dev. 13, 424–436 (1999). 53. Kee, K., Angeles, V. T., Flores, M., Nguyen, H. N. & 10. Hackett, J. A. et al. Germline DNA demethylation 33. Ying, Y., Liu, X. M., Marble, A., Lawson, K. A. & Reijo Pera, R. A. Human DAZL, DAZ and BOULE genes dynamics and imprint erasure through Zhao, G. Q. Requirement of Bmp8b for the generation modulate primordial germ-cell and haploid gamete 5‑hydroxymethylcytosine. Science 339, 448–452 of primordial germ cells in the mouse. Mol. Endocrinol. formation. Nature 462, 222–225 (2009). (2013). 14, 1053–1063 (2000). 54. Wu, J. & Izpisua Belmonte, J. C. Dynamic pluripotent 11. Guibert, S., Forne, T. & Weber, M. Global profiling of 34. de Sousa Lopes, S. M. et al. BMP signaling mediated stem cell states and their applications. Cell Stem Cell DNA methylation erasure in mouse primordial germ by ALK2 in the visceral endoderm is necessary for the 17, 509–525 (2015). cells. Genome Res. 22, 633–641 (2012). generation of primordial germ cells in the mouse 55. Weinberger, L., Ayyash, M., Novershtern, N. & References 9–11 describe the detailed embryo. Genes Dev. 18, 1838–1849 (2004). Hanna, J. H. Dynamic stem cell states: naive to primed genome-wide DNA demethylation dynamics in 35. Tremblay, K. D., Dunn, N. R. & Robertson, E. J. Mouse pluripotency in rodents and humans. Nat. Rev. Mol. mPGCs. embryos lacking Smad1 signals display defects in Cell Biol. 17, 155–169 (2016). 12. Hajkova, P. et al. Epigenetic reprogramming in mouse extra-embryonic tissues and germ cell formation. 56. Gafni, O. et al. Derivation of novel human ground state primordial germ cells. Mech. Dev. 117, 15–23 (2002). Development 128, 3609–3621 (2001). naive pluripotent stem cells. Nature 504, 282–286 13. Lee, J. et al. Erasing genomic imprinting memory in 36. Chu, G. C., Dunn, N. R., Anderson, D. C., Oxburgh, L. (2013). mouse clone embryos produced from day 11.5 & Robertson, E. J. Differential requirements for 57. Tang, W. W. et al. A unique gene regulatory network primordial germ cells. Development 129, 1807–1817 Smad4 in TGFβ-dependent patterning of the early resets the human germline epigenome for (2002). mouse embryo. Development 131, 3501–3512 development. Cell 161, 1453–1467 (2015). 14. Sugimoto, M. & Abe, K. X chromosome reactivation (2004). 58. Blakeley, P. et al. Defining the three cell lineages of the initiates in nascent primordial germ cells in mice. 37. Chang, H. & Matzuk, M. M. Smad5 is required for human blastocyst by single-cell RNA-seq. Development PLoS Genet. 3, e116 (2007). mouse primordial germ cell development. Mech. Dev. 142, 3151–3165 (2015). 15. Chuva de Sousa Lopes, S. M. et al. X chromosome 104, 61–67 (2001). 59. Pastor, W. A. et al. Naive human pluripotent cells activity in mouse XX primordial germ cells. 38. Liu, P. et al. Requirement for Wnt3 in vertebrate axis feature a methylation landscape devoid of blastocyst or PLoS Genet. 4, e30 (2008). formation. Nat. Genet. 22, 361–365 (1999). germline memory. Cell Stem Cell 18, 323–329 (2016). 16. Tam, P. P., Zhou, S. X. & Tan, S. S. X‑chromosome 39. Brennan, J. et al. Nodal signalling in the epiblast 60. Takashima, Y. et al. Resetting transcription factor activity of the mouse primordial germ cells revealed patterns the early mouse embryo. Nature 411, control circuitry toward ground-state pluripotency in by the expression of an X‑linked lacZ transgene. 965–969 (2001). human. Cell 158, 1254–1269 (2014). Development 120, 2925–2932 (1994). 40. Beddington, R. S. & Robertson, E. J. Axis 61. Sugawa, F. et al. Human primordial germ cell 17. Monk, M. & McLaren, A. X‑chromosome activity in development and early asymmetry in mammals. commitment in vitro associates with a unique PRDM14 foetal germ cells of the mouse. J. Embryol. Exp. Cell 96, 195–209 (1999). expression profile. EMBO J. 34, 1009–1024 (2015). Morphol. 63, 75–84 (1981). 41. Behringer, R. R., Wakamiya, M., Tsang, T. E. & 62. Kojima, Y. et al. The transcriptional and functional 18. Seki, Y. et al. Cellular dynamics associated with the Tam, P. P. A flattened mouse embryo: leveling the properties of mouse epiblast stem cells resemble the genome-wide epigenetic reprogramming in migrating playing field. Genesis 28, 23–30 (2000). anterior primitive streak. Cell Stem Cell 14, 107–120 primordial germ cells in mice. Development 134, 42. Hopf, C., Viebahn, C. & Puschel, B. BMP signals and (2014). 2627–2638 (2007). the transcriptional repressor BLIMP1 during germline 63. Buecker, C. et al. Reorganization of enhancer patterns 19. Seki, Y. et al. Extensive and orderly reprogramming of segregation in the mammalian embryo. Dev. Genes in transition from naive to primed pluripotency. genome-wide chromatin modifications associated with Evol. 221, 209–223 (2011). Cell Stem Cell 14, 838–853 (2014).

14 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

64. Kurimoto, K. et al. Quantitative dynamics of chromatin primordial germ cells. Development 136, 4011–4020 specific reprogramming in mice. Genome Res. 23, remodeling during germ cell specification from mouse (2009). 616–627 (2013). embryonic stem cells. Cell Stem Cell 16, 517–532 88. Chambers, I. et al. Nanog safeguards pluripotency 112. Sharif, J. et al. The SRA protein Np95 mediates (2015). and mediates germline development. Nature 450, epigenetic inheritance by recruiting Dnmt1 to 65. Zylicz, J. J. et al. Chromatin dynamics and the role of 1230–1234 (2007). methylated DNA. Nature 450, 908–912 (2007). G9a in gene regulation and enhancer silencing during 89. Campolo, F. et al. Essential role of Sox2 for the 113. Kagiwada, S., Kurimoto, K., Hirota, T., Yamaji, M. early mouse development. eLife 4, e09571 (2015). establishment and maintenance of the germ cell line. & Saitou, M. Replication-coupled passive DNA 66. Murakami, K. et al. NANOG alone induces germ cells Stem Cells 31, 1408–1421 (2013). demethylation for the erasure of genome imprints in primed epiblast in vitro by activation of enhancers. 90. Yamaguchi, Y. L. et al. Sall4 is essential for mouse in mice. EMBO J. 32, 340–353 (2013). Nature 529, 403–407 (2016). primordial germ cell specification by suppressing 114. Ohno, R. et al. A replication-dependent passive 67. Kurimoto, K. et al. Complex genome-wide somatic cell program genes. Stem Cells 33, 289–300 mechanism modulates DNA demethylation in mouse transcription dynamics orchestrated by Blimp1 for (2015). primordial germ cells. Development 140, 2892–2903 the specification of the germ cell lineage in mice. 91. Guo, F. et al. The transcriptome and DNA methylome (2013). Genes Dev. 22, 1617–1635 (2008). landscapes of human primordial germ cells. Cell 161, 115. Arand, J. et al. Selective impairment of methylation 68. Saitou, M., Barton, S. C. & Surani, M. A. A molecular 1437–1452 (2015). maintenance is the major cause of DNA methylation programme for the specification of germ cell fate in This article and reference 57 reveal the reprogramming in the early embryo. Epigenetics mice. Nature 418, 293–300 (2002). transcriptional network and epigenetic Chromatin 8, 1 (2015). 69. Vincent, S. D. et al. The zinc finger transcriptional reprogramming dynamics in hPGCs isolated from 116. Hackett, J. A., Zylicz, J. J. & Surani, M. A. Parallel repressor Blimp1/Prdm1 is dispensable for early axis human embryos, using high-quality RNA-seq and mechanisms of epigenetic reprogramming in the formation but is required for specification of bisulfite sequencing (BS‑seq) analyses. germline. Trends Genet. 28, 164–174 (2012). primordial germ cells in the mouse. Development 132, 92. Perrett, R. M. et al. The early human germ cell lineage 117. Yamaguchi, S. et al. Dynamics of 5‑methylcytosine 1315–1325 (2005). does not express SOX2 during in vivo development or and 5‑hydroxymethylcytosine during germ cell This paper and reference 2 show that PRDM1 is upon in vitro culture. Biol. Reprod. 78, 852–858 reprogramming. Cell Res. 23, 329–339 (2013). the earliest lineage-restricted marker of mPGCs (2008). 118. Yamaguchi, S., Shen, L., Liu, Y., Sendler, D. & Zhang, Y. and that it is indispensable for mPGC specification. 93. de Jong, J. et al. Differential expression of SOX17 and Role of Tet1 in erasure of genomic imprinting. Nature 70. Grabole, N. et al. Prdm14 promotes germline fate and SOX2 in germ cells and stem cells has biological and 504, 460–464 (2013). naive pluripotency by repressing FGF signalling and clinical implications. J. Pathol. 215, 21–30 (2008). 119. Dawlaty, M. M. et al. Combined deficiency of Tet1 DNA methylation. EMBO Rep. 14, 629–637 (2013). 94. Hara, K. et al. Evidence for crucial role of hindgut and Tet2 causes epigenetic abnormalities but is 71. Martins, G. & Calame, K. Regulation and functions of expansion in directing proper migration of primordial compatible with postnatal development. Dev. Cell 24, Blimp‑1 in T and B lymphocytes. Annu. Rev. Immunol. germ cells in mouse early embryogenesis. Dev. Biol. 310–323 (2013). 26, 133–169 (2008). 330, 427–439 (2009). 120. Yamaguchi, S. et al. Tet1 controls meiosis by 72. Ma, Z., Swigut, T., Valouev, A., Rada-Iglesias, A. & 95. Sarkar, A. & Hochedlinger, K. The Sox family of regulating meiotic gene expression. Nature 492, Wysocka, J. Sequence-specific regulator Prdm14 transcription factors: versatile regulators of stem and 443–447 (2012). safeguards mouse ESCs from entering extraembryonic progenitor cell fate. Cell Stem Cell 12, 15–30 (2013). 121. Gkountela, S. et al. DNA demethylation dynamics in endoderm fates. Nat. Struct. Mol. Biol. 18, 120–127 96. Kanai-Azuma, M. et al. Depletion of definitive gut the human prenatal germline. Cell 161, 1425–1436 (2011). endoderm in Sox17‑null mutant mice. Development (2015). 73. Yamaji, M. et al. PRDM14 ensures naive pluripotency 129, 2367–2379 (2002). 122. Gaskell, T. L., Esnal, A., Robinson, L. L., Anderson, R. A. through dual regulation of signaling and epigenetic 97. Kim, I., Saunders, T. L. & Morrison, S. J. Sox17 & Saunders, P. T. Immunohistochemical profiling of pathways in mouse embryonic stem cells. Cell Stem Cell dependence distinguishes the transcriptional germ cells within the human fetal testis: identification of 12, 368–382 (2013). regulation of fetal from adult hematopoietic stem cells. three subpopulations. Biol. Reprod. 71, 2012–2021 74. Auman, H. J. et al. Transcription factor AP‑2γ is Cell 130, 470–483 (2007). (2004). essential in the extra-embryonic lineages for early 98. Kamachi, Y. & Kondoh, H. Sox proteins: regulators of 123. Kerr, C. L., Hill, C. M., Blumenthal, P. D. & postimplantation development. Development 129, cell fate specification and differentiation. Development Gearhart, J. D. Expression of pluripotent stem cell 2733–2747 (2002). 140, 4129–4144 (2013). markers in the human fetal ovary. Hum. Reprod. 23, 75. Werling, U. & Schorle, H. Transcription factor gene 99. Nakagawa, M. et al. Generation of induced pluripotent 589–599 (2008). AP‑2γ essential for early murine development. stem cells without Myc from mouse and human 124. Kerr, C. L., Hill, C. M., Blumenthal, P. D. & Mol. Cell. Biol. 22, 3149–3156 (2002). fibroblasts. Nat. Biotechnol. 26, 101–106 (2008). Gearhart, J. D. Expression of pluripotent stem cell 76. Nady, N. et al. ETO family protein Mtgr1 mediates 100. Niakan, K. K. et al. Sox17 promotes differentiation markers in the human fetal testis. Stem Cells 26, Prdm14 functions in stem cell maintenance and in mouse embryonic stem cells by directly regulating 412–421 (2008). primordial germ cell formation. eLife 4, e10150 extraembryonic gene expression and indirectly 125. Wermann, H. et al. Global DNA methylation in fetal (2015). antagonizing self-renewal. Genes Dev. 24, 312–326 human germ cells and germ cell tumours: association 77. Tu, S. et al. Co‑repressor CBFA2T2 regulates (2010). with differentiation and cisplatin resistance. J. Pathol. pluripotency and germline development. Nature 534, 101. Seguin, C. A., Draper, J. S., Nagy, A. & Rossant, J. 221, 433–442 (2010). 387–390 (2016). Establishment of endoderm progenitors by SOX 126. Smith, Z. D. & Meissner, A. DNA methylation: roles References 76 and 77 show that co‑repressor transcription factor expression in human embryonic in mammalian development. Nat. Rev. Genet. 14, CBFA2T2 interacts with PRDM14 and thus has a stem cells. Cell Stem Cell 3, 182–195 (2008). 204–220 (2013). role in the regulation of pluripotency and mPGC 102. Stefanovic, S. et al. Interplay of Oct4 with Sox2 and 127. Tsumura, A. et al. Maintenance of self-renewal ability specification. Sox17: a molecular switch from stem cell pluripotency of mouse embryonic stem cells in the absence of DNA 78. Fog, C. K., Galli, G. G. & Lund, A. H. PRDM proteins: to specifying a cardiac fate. J. Cell Biol. 186, 665–673 methyltransferases Dnmt1, Dnmt3a and Dnmt3b. important players in differentiation and disease. (2009). Genes Cells 11, 805–814 (2006). Bioessays 34, 50–60 (2012). 103. Aksoy, I. et al. Oct4 switches partnering from Sox2 to 128. Brinkman, A. B. et al. Sequential ChIP-bisulfite 79. Scholer, H. R., Dressler, G. R., Balling, R., Sox17 to reinterpret the enhancer code and specify sequencing enables direct genome-scale investigation Rohdewohld, H. & Gruss, P. Oct‑4: a germline-specific endoderm. EMBO J. 32, 938–953 (2013). of chromatin and DNA methylation cross-talk. transcription factor mapping to the mouse t‑complex. 104. Niwa, H., Miyazaki, J. & Smith, A. G. Quantitative Genome Res. 22, 1128–1138 (2012). EMBO J. 9, 2185–2195 (1990). expression of Oct‑3/4 defines differentiation, 129. Walter, M., Teissandier, A., Perez-Palacios, R. & 80. Yeom, Y. I. et al. Germline regulatory element of Oct‑4 dedifferentiation or self-renewal of ES cells. Nat. Genet. Bourc’his, D. An epigenetic switch ensures transposon specific for the totipotent cycle of embryonal cells. 24, 372–376 (2000). repression upon dynamic loss of DNA methylation in Development 122, 881–894 (1996). 105. Maruyama, M., Ichisaka, T., Nakagawa, M. & embryonic stem cells. eLife 5, e11418 (2016). 81. Yabuta, Y., Kurimoto, K., Ohinata, Y., Seki, Y. & Yamanaka, S. Differential roles for Sox15 and Sox2 in This paper shows that repressive chromatin Saitou, M. Gene expression dynamics during germline transcriptional control in mouse embryonic stem cells. modifications repress retrotransposons and specification in mice identified by quantitative single- J. Biol. Chem. 280, 24371–24379 (2005). safeguard genome stability during chemical-induced cell gene expression profiling. Biol. Reprod. 75, 106. Masui, S. et al. Pluripotency governed by Sox2 via global DNA demethylation in mES cells. 705–716 (2006). regulation of Oct3/4 expression in mouse embryonic 130. Ancelin, K. et al. Blimp1 associates with Prmt5 and 82. Sato, M. et al. Identification of PGC7, a new gene stem cells. Nat. Cell Biol. 9, 625–635 (2007). directs histone arginine methylation in mouse germ expressed specifically in preimplantation embryos 107. Chia, N. Y. et al. A genome-wide RNAi screen reveals cells. Nat. Cell Biol. 8, 623–630 (2006). and germ cells. Mech. Dev. 113, 91–94 (2002). determinants of human embryonic stem cell identity. 131. Ng, J. H. et al. In vivo epigenomic profiling of germ 83. Leitch, H. G. et al. Embryonic germ cells from mice Nature 468, 316–320 (2010). cells reveals germ cell molecular signatures. Dev. Cell and rats exhibit properties consistent with a generic 108. Tsuneyoshi, N. et al. PRDM14 suppresses expression 24, 324–333 (2013). pluripotent ground state. Development 137, of differentiation marker genes in human embryonic 132. Liu, S. et al. Setdb1 is required for germline 2279–2287 (2010). stem cells. Biochem. Biophys. Res. Commun. 367, development and silencing of H3K9me3‑marked 84. Matsui, Y., Zsebo, K. & Hogan, B. L. Derivation of 899–905 (2008). endogenous retroviruses in primordial germ cells. pluripotential embryonic stem cells from murine 109. Messerschmidt, D. M., Knowles, B. B. & Solter, D. Genes Dev. 28, 2041–2055 (2014). primordial germ cells in culture. Cell 70, 841–847 DNA methylation dynamics during epigenetic 133. Kim, S. et al. PRMT5 protects genomic integrity (1992). reprogramming in the germline and preimplantation during global DNA demethylation in primordial germ 85. Durcova-Hills, G., Tang, F., Doody, G., Tooze, R. & embryos. Genes Dev. 28, 812–828 (2014). cells and preimplantation embryos. Mol. Cell 56, Surani, M. A. Reprogramming primordial germ cells 110. Saitou, M., Kagiwada, S. & Kurimoto, K. Epigenetic 564–579 (2014). into pluripotent stem cells. PLoS ONE 3, e3531 (2008). reprogramming in mouse pre-implantation 134. Guo, H. et al. The DNA methylation landscape of 86. Kehler, J. et al. Oct4 is required for primordial germ development and primordial germ cells. Development human early embryos. Nature 511, 606–610 (2014). cell survival. EMBO Rep. 5, 1078–1083 (2004). 139, 15–31 (2012). 135. Smith, Z. D. et al. DNA methylation dynamics of the 87. Yamaguchi, S. et al. Conditional knockdown of Nanog 111. Kobayashi, H. et al. High-resolution DNA methylome human preimplantation embryo. Nature 511, induces apoptotic cell death in mouse migrating analysis of primordial germ cells identifies gender- 611–615 (2014).

NATURE REVIEWS | GENETICS ADVANCE ONLINE PUBLICATION | 15 ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

REVIEWS

136. Lane, N. et al. Resistance of IAPs to methylation 152. Sharma, U. et al. Biogenesis and function of tRNA Proc. Natl Acad. Sci. USA 100, 11457–11462 reprogramming may provide a mechanism for fragments during sperm maturation and fertilization (2003). epigenetic inheritance in the mouse. Genesis 35, in mammals. Science 351, 391–396 (2016). 169. Hubner, K. et al. Derivation of oocytes from mouse 88–93 (2003). 153. Johnson, A. D. & Alberio, R. Primordial germ cells: embryonic stem cells. Science 300, 1251–1256 137. Jeltsch, A. & Jurkowska, R. Z. New concepts in DNA the first cell lineage or the last cells standing? (2003). methylation. Trends Biochem. Sci. 39, 310–318 Development 142, 2730–2739 (2015). 170. Hayashi, K. & Surani, M. A. Self-renewing epiblast stem (2014). 154. Zhou, Q. et al. Complete meiosis from embryonic stem cells exhibit continual delineation of germ cells with 138. Wolf, G., Greenberg, D. & Macfarlan, T. S. Spotting the cell-derived germ cells in vitro. Cell Stem Cell 18, epigenetic reprogramming in vitro. Development 136, enemy within: targeted silencing of foreign DNA in 330–340 (2016). 3549–3556 (2009). mammalian genomes by the Krüppel-associated box 155. Zhang, X. et al. Pax6 is a human neuroectoderm cell 171. Thomson, J. A. et al. Embryonic stem cell lines derived zinc finger protein family. Mob. DNA 6, 17 (2015). fate determinant. Cell Stem Cell 7, 90–100 (2010). from human blastocysts. Science 282, 1145–1147 139. Jacobs, F. M. et al. An evolutionary arms race between 156. Rossant, J. & Tam, P. P. Blastocyst lineage formation, (1998). KRAB zinc-finger genes ZNF91/93 and SVA/L1 early embryonic asymmetries and axis patterning in 172. Nichols, J. & Smith, A. Naive and primed pluripotent retrotransposons. Nature 516, 242–245 (2014). the mouse. Development 136, 701–713 (2009). states. Cell Stem Cell 4, 487–492 (2009). 140. Li, X. et al. A maternal-zygotic effect gene, Zfp57, 157. De Paepe, C., Krivega, M., Cauffman, G., Geens, M. & 173. Vallier, L., Alexander, M. & Pedersen, R. A. Activin/ maintains both maternal and paternal imprints. Van de Velde, H. Totipotency and lineage segregation Nodal and FGF pathways cooperate to maintain Dev. Cell 15, 547–557 (2008). in the human embryo. Mol. Hum. Reprod. 20, pluripotency of human embryonic stem cells. J. Cell Sci. 141. Mackay, D. J. et al. Hypomethylation of multiple 599–618 (2014). 118, 4495–4509 (2005). imprinted loci in individuals with transient neonatal 158. Niakan, K. K. & Eggan, K. Analysis of human embryos 174. Son, M. Y., Choi, H., Han, Y. M. & Cho, Y. S. Unveiling diabetes is associated with mutations in ZFP57. from zygote to blastocyst reveals distinct gene the critical role of REX1 in the regulation of human stem Nat. Genet. 40, 949–951 (2008). expression patterns relative to the mouse. Dev. Biol. cell pluripotency. Stem Cells 31, 2374–2387 (2013). 142. Williams, Z. et al. Discovery and characterization 375, 54–64 (2013). 175. Clark, A. T. et al. Spontaneous differentiation of germ of piRNAs in the human fetal ovary. Cell Rep. 13, 159. O’Leary, T. et al. Tracking the progression of the cells from human embryonic stem cells in vitro. 854–863 (2015). human inner cell mass during embryonic stem cell Hum. Mol. Genet. 13, 727–739 (2004). 143. Baudat, F. et al. PRDM9 is a major determinant of derivation. Nat. Biotechnol. 30, 278–282 (2012). 176. Rao, J. et al. Stepwise clearance of repressive meiotic recombination hotspots in humans and mice. 160. Bedzhov, I. & Zernicka-Goetz, M. Self-organizing roadblocks drives cardiac induction in human ESCs. Science 327, 836–840 (2010). properties of mouse pluripotent cells initiate Cell Stem Cell 18, 341–353 (2016). 144. Daxinger, L. & Whitelaw, E. Understanding morphogenesis upon implantation. Cell 156, 177. Lin, I. Y. et al. Suppression of the SOX2 neural effector transgenerational epigenetic inheritance via the 1032–1044 (2014). gene by PRDM1 promotes human germ cell fate in gametes in mammals. Nat. Rev. Genet. 13, 153–162 161. Moore, K. L., Persaud, T. V. N. & Torchia, M. G. The embryonic stem cells. Stem Cell Rep. 2, 189–204 (2012). Developing Human: Clinically Oriented Embryology (2014). 145. Bohacek, J. & Mansuy, I. M. Molecular insights into (Elsevier/Saunders, 2013). 178. Du, J., Johnson, L. M., Jacobsen, S. E. & Patel, D. J. transgenerational non-genetic inheritance of acquired 162. Hackett, J. A. & Surani, M. A. Regulatory principles of DNA methylation pathways and their crosstalk with behaviours. Nat. Rev. Genet. 16, 641–652 (2015). pluripotency: from the ground state up. Cell Stem Cell histone methylation. Nat. Rev. Mol. Cell Biol. 16, 146. Heard, E. & Martienssen, R. A. Transgenerational 15, 416–430 (2014). 519–532 (2015). epigenetic inheritance: myths and mechanisms. 163. Evans, M. J. & Kaufman, M. H. Establishment in 179. Gkountela, S. et al. The ontogeny of cKIT+ human Cell 157, 95–109 (2014). culture of pluripotential cells from mouse embryos. primordial germ cells proves to be a resource for human 147. Surani, M. A., Barton, S. C. & Norris, M. L. Nature 292, 154–156 (1981). germ line reprogramming, imprint erasure and in vitro Development of reconstituted mouse eggs suggests 164. Martin, G. R. Isolation of a pluripotent cell line from differentiation. Nat. Cell Biol. 15, 113–122 (2013). imprinting of the genome during gametogenesis. early mouse embryos cultured in medium conditioned Nature 308, 548–550 (1984). by teratocarcinoma stem cells. Proc. Natl Acad. Sci. Acknowledgements 148. McGrath, J. & Solter, D. Completion of mouse USA 78, 7634–7638 (1981). The authors thank J. Hackett for reading of the manuscript embryogenesis requires both the maternal and 165. Brons, I. G. et al. Derivation of pluripotent epiblast and members of the Surani laboratory for helpful discussions. paternal genomes. Cell 37, 179–183 (1984). stem cells from mammalian embryos. Nature 448, Our work was funded by a Wellcome Trust Investigator Award 149. Radford, E. J. et al. In utero undernourishment 191–195 (2007). to M.A.S, and by a Britain Israel Research and Academic perturbs the adult sperm methylome and 166. Tesar, P. J. et al. New cell lines from mouse epiblast Exchange (BIRAX) Initiative and a Croucher Cambridge intergenerational metabolism. Science 345, 1255903 share defining features with human embryonic stem International Scholarship to W.W.C.T. Research at the Gurdon (2014). cells. Nature 448, 196–199 (2007). Institute is funded by a core grant from the Wellcome Trust 150. Shea, J. M. et al. Genetic and epigenetic variation, but 167. Huang, Y., Osorno, R., Tsakiridis, A. & Wilson, V. (092096) and Cancer Research UK (C6946/A14492). The not diet, shape the sperm methylome. Dev. Cell 35, In vivo differentiation potential of epiblast stem cells authors apologize to colleagues whose work could not be 750–758 (2015). revealed by chimeric embryo formation. Cell Rep. 2, cited owing to length limitations. 151. Chen, Q. et al. Sperm tsRNAs contribute to 1571–1578 (2012). intergenerational inheritance of an acquired 168. Toyooka, Y., Tsunekawa, N., Akasu, R. & Noce, T. Competing interests statement metabolic disorder. Science 351, 397–400 (2016). Embryonic stem cells can form germ cells in vitro. The authors declare no competing interests.

16 | ADVANCE ONLINE PUBLICATION www.nature.com/nrg ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2016 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.