Epigenetic Regulation during Primordial Germ Cell Development and Differentiation

Germline development varies significantly across metazoans. However, mammalian primordial germ cell (PGC) development has key conserved landmarks, including a critical period of epigenetic reprogramming that precedes sex-specific differentiation and gametogenesis. Epigenetic alterations in the germline are of unique importance due to their potential to impact the next generation. Therefore, regulation of, and by, the non-coding genome is of utmost importance during these epigenomic events. Here, we detail the key chromatin changes that occur during mammalian PGC development and how these interact with the expression of non-coding RNAs alongside broader epitranscriptomic changes. We identify gaps in our current knowledge, in particular regarding epigenetic regulation in the human germline, and we highlight important areas of future research.


Introduction
Across metazoans there are a diverse array of mechanisms of primordial germ cell (PGC) specification which have important implications for the evolution of the vertebrate body plan [Extavour and Akam, 2003;Evans et al., 2014]. In mammals, despite notable embryological differences, many features of PGC development are conserved [Kobayashi and Surani, 2018;Dion and Leitch, 2020;Hancock et al., 2021]. PGCs are induced by the action of BMP signals on either the pluripotent epiblast [Ohinata et al., 2009;Kobayashi et al., 2017] or, perhaps, its recent descendants, such as the amnion in cynomolgus macaques [Sasaki et al., 2016] (Fig. 1a). This leads to the upregulation of PGC specification genes (including BLIMP1 and TFAP2C), although species-specific differences have been noted Tang et al., 2015;Kojima et al., 2017;Kobayashi et al., 2017;Sybirna et al., 2020]. This is accompanied by maintenance of OCT4 and re-expression of many pluripotency-associated transcription factors [Leitch and Smith, 2013]. Thereafter, PGCs migrate to the genital ridge, where upon colonisation a process of genome-wide epigenetic reprogramming is completed prior to sex-specific differentiation [Hajkova, 2011;Leitch et al., 2013a]. To our knowledge, no mammalian N.B.R. and K.M. contributed equally to this work.
This article is licensed under the Creative Commons Attribution 4.0 International License (CC BY) (http://www.karger.com/Services/ OpenAccessLicense). Usage, derivative works and distribution are permitted provided that proper credit is given to the author and the original publisher. DOI: 10.1159/000520412 species studied so far deviates from this basic developmental blueprint. Like many specification events, induction of a small number of key transcription factors appears to drive segregation of the germline [Nakaki et al., 2013;Magnusdottir and Surani, 2014]. Perhaps unsurprisingly, this is accompanied by epigenomic changes such that upon arrival at the genital ridges, PGCs are already epigenetically distinguishable from their new somatic cell neighbours, which have a distinctive developmental origin [Seki et al., 2007;Hajkova, 2011]. Here, we will briefly summarise these early epigenetic changes that provide the template for genome-wide epigenetic repro-gramming, which is subsequently discussed in detail. Next, we consider how non-coding RNAs map onto these key early germline events and summarise their known germline functions, before finally suggesting important unanswered questions that might constitute future areas of research into the function of the non-coding genome in early mammalian germ cells. Filling in these gaps in our knowledge is a key priority because, in contrast to any other cell lineage, epigenetic alterations in the germline have the potential to impact the development and health of the next generation. A ct iv in A , G S K 3 β i, R O C K i 2 d a ys a b Fig. 1. Mammalian models of primordial germ cell specification. a Simplified depictions of cylindrical mouse (left) and bilaminar human (right) embryos with mouse primordial germ cells (mPGCs) specified at embryonic day (E) 6.25 at the posterior epiblast, and human PGCs (hPGCs) thought to be specified around week 2-3 putatively at the early amnion and/or posterior epiblast. b In vitro models of PGC-like cell (PGCLC) differentiation for mouse (top) and human (bottom). Mouse PGCLCs (mPGCLCs) are derived from 2-inhibitor (2i: PD0325901, CHIR99021)/Leukemia Inhibitory Factor (LIF), embryonic stem cells (mESCs), then induced to epiblast-like cells (mEpiLCs) upon 40-48 h culture with Activin A and fibroblast growth factor 2 (FGF2). mPGCLCs are then isolated from embryoid body (EB) formation upon 4-6 days of treatment with cytokines bone morphogenetic factor 4 (BMP4) (and optionally with BMP8A/B), stem cell factor (SCF), LIF, and epidermal growth factor (EGF). Human PGCLCs (hPGCLCs) are derived from conventional human ESCs (hESC) or induced pluripotent stem cells (iPSCs) in Essential-8 (E8) or conventional hESC media grown in 1 of 3 protocols. (1) 4-inihibitor hESC (4i: CHIR99021, PD0325901, SB203580, SP600125), FGF2, LIF, and transforming growth factor beta 2 (TGFβ2) grown on feeder layers; (2) pre-mesendoderm (pre-ME) cells grown in Activin A, glycogen synthase kinase 3 beta inhibitor (GSK3βi: CHIR99021), and Rho-associated protein kinase inhibitor (ROCKi) for 12 h; or (3) incipient mesoderm-like cells (iMeLC) grown in Activin A, GSK3βi, ROCKi for 2 days. hPGCLCs are isolated from day 4 EBs upon treatment of either of these cell states with ROCKi and cytokines BMP2 or 4, SCF, LIF, and EGF. Initial epigenetic changes triggered by mouse PGC specification (starting around E6.25) have largely been quantified by immunofluorescence staining (Fig. 2). This revealed that the repressive histone modification H3K-9me2 becomes globally depleted from around E8.0 [Seki et al., 2005;Hajkova et al., 2008]. Conversely, H3K9me3 levels remain high at pericentric heterochromatin, with punctae observed in immunostainings, whereas repressive H3K27me3 levels become progressively enriched globally [Seki et al., 2005;Chuva de SousaLopes et al., 2008;Hajkova et al., 2008;Lesch et al., 2013;Sachs et al., 2013]. Repressive modifications of histones H2A and H4 (H2A/H4R3me2) also increase, due to the nuclear translocation of Prmt5 and its association with Blimp1 [Ancelin et al., 2006]. Notably, there is a decline in DNA methylation starting from around E8.5 [Seki et al., 2005;Guibert et al., 2012;Seisenberger et al., 2012] which is associated with downregulation of de novo methyltransferases Dnmt3a and 3b, as well as the Dnmt1 co-factor Uhrf1 [Seki et al., 2005;Kurimoto et al., 2008;Ohno et al., 2013]. However, clearly some DNA methyltransferase activity is maintained as many loci retain high levels of DNA methylation during this period, and conditional deletion of Dnmt1 in pre-gonadal PGCs results in precocious depletion of DNA methylation, in particular at imprinting control regions, meiotic genes, and endogenous retrovirus-intracisternal A particle (ERV-IAP) retrotransposons [Hargan-Calvopina et al., 2016]. How DNA methylation levels are regulated at different genomic regions remains poorly characterised (discussed below). The small number of newly specified PGCs has been a limiting factor for more detailed epigenomic analyses. Instead, the mPGC-like cell (mPGCLC) system ( Fig. 1b) has been used to generate sufficient cell numbers to perform high resolution chromatin-immunoprecipitation followed by next-generation sequencing (ChIP-seq). As mPGCLCs are induced from epiblast-like cells (mEpiLCs), they show a widespread reorganisation of the repressive marks H3K27me3 and H3K9me2 , in keeping with the in vivo data. These germline-competent mEpiLCs were noted to harbour relatively low levels of H3K27me3 across developmental regulators compared with mESCs which cannot directly give rise to PG-CLCs. This perhaps allows mEpiLCs to make rapid transcriptional responses upon receipt of appropriate signalling cues. It was additionally noted that in early mPGCLCs, H3K27me3 marks are recruited across promoters of developmental regulators of somatic lineages -perhaps acting to reinforce their silencing in the germline. Curiously, later (Day 6) mPGCLCs display concomitant upregulation of the active H3K4me3 mark across a number of these genes, marking these promoters as bivalent yet still repressed regions -the significance of which remains unknown. Meanwhile, unlike mEpiLCs, epiblast-derived stem cells (mEpiSCs) show reduced competence for germline fate [Hayashi and Surani, 2009;Hayashi et al., 2011]. These mEpiSCs lose H3K27me3 but gain more stable repressive DNA methylation marks especially over germline-related promoters when compared to germlinecompetent in vivo E6.25 epiblast cells [Hayashi and Surani, 2009;Zylicz et al., 2015], which may contribute to the reduced germline competence of mEpiSCs. Intriguingly, a recent report has described the derivation of self-renewing formative stem cells that exhibit similar properties to mEpiLCs, including the ability to directly give rise to mPGCLCs [Kinoshita et al., 2021]. Expansion culture of in vivo-sourced mPGCs may provide an alternative in vitro system in the future but remains challenging despite some recent advances [Leitch et al., 2013b;Borkowska and Leitch, 2021].
Chromatin dynamics have not been studied in newlyspecified, pre-migratory human PGCs in vivo due to the inherent challenges in accessing early post-implantation human embryos for research (Fig. 3). Human (h)PGCLCs are thus the only available tool that might provide an approximation of nascent hPGCs (Fig. 1b). However, with no in vivo comparisons of newly-specified PGCs, it remains unclear exactly which stage of development hPGCLCs represent [Dion and Leitch, 2020]. Of note, recent analysis of an extremely rare week 3 (CS 7) gastrulating human embryo identified the putative presence of hPGCs [Tyser et al., 2020], and there has been (infrequent) detection of hPGC markers in late week 2 ex vivo cultured human embryos [Chen et al., 2019]. If further rare embryos do become available, or if ex vivo embryo culture beyond day 14  reveals nascent hPGCs, this might provide further relevant comparisons to more accurately stage hPGCLCs. Nevertheless, hPGCLCs will likely continue to be the mainstay of investigation into the early human germ line. hPGCLCs have not been reported to show large-scale global changes in DNA methylation following induction Murase et al., 2020], and there are differing reports as to whether DNA demethylation proceeds following further in vitro culture [Gell et al., 2020;Murase et al., 2020]. This may be in part due to differential regulation of the DNA methyltransferase enzymes and UHRF1 in the different induction and culture systems. Immunostainings have shown decreased H3K9me2 and slightly higher H3K27me3 staining [Sasaki et al., 2015], in keeping with data from the mouse and also with immunostainings from the earliest hPGCs obtained thus far [Gkountela et al., 2013;Tang et al., 2015]. Additionally, ATAC-seq of hPGCLCs identified regions of accessible chromatin that significantly overlap with naive-pluripotency related TFAP2C-bound enhancers. Indeed, the deletion of the TFAP2C-bound POU5F1/OCT4 distal enhancer perturbed hPGCLC induction  However, no rigorous genome-wide analysis of histone modifications in hPGCLCs has yet been reported but would be of interest to the field, especially if compared with fetal hPGC samples.
Thus, following specification, and during migration to the genital ridges, PGCs establish a distinctive epigenome characterized by low levels of DNA methylation and H3K-9me2, as well as elevated H3K27me3 and H2A/H4R3me2s. However, this appears to act as a prelude to a more complex and complete epigenetic reprogramming event upon arrival at the genital ridge, associated with, and/or leading to, further profound changes in epigenetic regulation.

DNA Methylation
Mouse Since the first reports of DNA methylation changes in mouse PGCs, it has been clear that DNA methylation drops across almost all DNA elements during migration and following colonisation of the genital ridge [Monk et al., 1987;Kafri et al., 1992;Hajkova et al., 2002]. These findings have been replicated in more recent studies utilising genome-wide approaches, demonstrating that during epigenetic reprogramming in the gonad, DNA methylation reaches its nadir, achieving levels lower than at any other stage of development [Seisenberger et al., 2012;Hill et al., 2018]. The exact timing of DNA demethylation differs according to genetic background (which does make direct comparison between different studies challenging), however, in most cases this commences around E11.0 and is complete by E13.5. Critically, it is during this process that genomic imprints are erased [Hajkova et al., 2002;Seisenberger et al., 2012]. The removal of DNA methylation from imprinted differentially methylated regions (DMRs) is a unique event during mammalian development and allows subsequent re-establishment of sex-specific genomic imprints. In males imprint re-establishment appears coupled to global increases in DNA methylation, a process that begins soon after epigenetic reprogramming [Lees-Murdock and Walsh, 2008], with substantial restoration of global DNA methylation clearly evident by E16.5 [Li et al., 2004;Seisenberger et al., 2012]. The histone modification H3K-36me2, deposited by NSD-1 over broad euchromatic regions occupying 59% of the genome, has recently been shown to play a critical role in this de novo DNA methylation process in mouse spermatogonia by E16.5, particularly over non-transposable element (TE) regions [Shirane et al., 2020]. In contrast, low levels of DNA methylation are maintained in mouse oogoonia until oocyte growth commences postnatally [Smallwood et al., 2011]. Intriguingly, a recent in vitro study inducing oocyte-like cells from pluripotent stem cells by transcription factor overexpression has demonstrated that at least some of the de novo DNA methylation that occurs during oocyte growth can occur independently of the preceding process of epigenetic reprogramming [Hamazaki et al., 2021]. Despite the loss of DNA methylation from promoter and regulatory elements, there is not a generalised transcriptional activation. However, loss of DNA methylation has been associated with activation of selected germlinespecific genes [Maatouk et al., 2006;Hackett et al., 2012b;Velasco et al., 2010], and more recently, a larger group of germline reprogramming responsive (GRR) genes have been described. These GRR genes become upregulated as a result of promoter DNA demethylation, and robust expression also requires simultaneous depletion of polycomb marks [Hill et al., 2018]. How these findings relate to the global hypertranscription reported in mPGCs [Percharde et al., 2017] has not been specifically addressed and is an important topic for future study (see later). In particular, it will be interesting to determine when hyper-transcription commences (it was noted as early as E9.5) and how this relates to the epigenomic landscape at this timepoint.
Most TEs are also subject to DNA demethylation [Seisenberger et al., 2012;Ohno et al., 2013], however, this does not result in widespread upregulation of expression, with the exception of specific evolutionary young elements such as certain loci of ERV-IAP retrotransposons (IAPA_MM, IAPEZI), long interspersed nuclear element-1 (L1) (L1mdTf_II, L1MdA_II/III), long terminal repeat (LTR) elements (MMERVK10C), and early transposon (ETn) elements [Ohno et al., 2013;Hill et al., 2018]. This may suggest both an alternative mechanism for maintaining transcriptional silencing of those elements that are DNA demethylated but not expressed, as well as a method of post-transcriptional silencing of those that are upregulated. In contrast, the majority of L1A and ERV-IAP retrotransposons retain substantial levels of DNA methylation [Hajkova et al., 2002;Seisenberger et al., 2012], which may be required for their transcriptional repression. Which locus-specific mechanisms maintain the relatively higher levels of DNA methylation on these loci during reprogramming remains largely unknown, although there is some evidence that repressive histone modifications play a role [Liu et al., 2014]. Notably, when mESCs are transitioned from serum to 2-inhibitor(2i) + leukemia inhibitory factor (LIF) + vitamin C conditions there is DNA demethylation and loss of H3K9me2 at certain TE loci. This is accompanied by an initial burst of TE de-repression which is subsequently quelled by gain of H3K27me3 [Walter et al., 2016]. Intriguingly, this switch in repressive histone modifications is reminiscent of the global changes that occur in pregonadal PGCs [Seki et al., 2005].
However, this classic view of relatively limited TE expression in PGCs is challenged by the above-mentioned observation of hypertranscription in mPGCs between E9.5 and E15.5, including higher expression of TEs in mPGCs than in somatic cells [Percharde et al., 2017]. This elevated expression of TEs is evident at E9.5 when DNA methylation remains high and is observed in TEs that both retain and lose DNA methylation at later timepoints. Therefore, as for single copy genes, future work that combines cell number-normalised TE expression analysis with epigenomic analysis would be of particular interest. Other important factors that impact whether (and which) TEs show elevated expression upon demethylation include the additional activatory mechanisms such as transcription factors that are recruited to their loci [Kunarso et al., 2010;Pontis et al., 2019] with older TEs carrying a high mutational burden that may impact their competence for transcription [Rollins et al., 2006;Castro-Diaz et al., 2014. Additionally, subsequent higher TE expression may not be detected in steady-state RNA-seq datasets due to the activity of posttranscriptional mechanisms in controlling transcript levels. Nevertheless, the postulated presence of TE transcripts, even at these early PGC stages, may potentially suggest the involvement of post-transcriptional mechanisms in preventing deleterious retrotransposition events (see later discussion).
The mechanism of DNA demethylation has sparked intense debate. The rapid reduction in DNA methylation following colonisation of the gonad that occurs without a change in doubling time [Tam and Snow, 1981], and possibly even within a single cell cycle [Hajkova et al., 2010], has suggested an active process. However, a mechanism of active DNA demethylation that could explain this global loss of DNA methylation has not yet been described in mammals. With the discovery of Tet enzymes the possibility of replication-coupled, Tet-assisted passive demethylation by hydroxylation of 5mC seemed both an attractive proposition and a highly plausible mechanism [Dawlaty et al., 2011;Hackett et al., 2013;Yamaguchi et al., 2013]. However, direct measurements of 5hmC levels in mPGCs reveal levels that are an order of magnitude lower than would be required for this to be the major mechanism [Hill et al., 2018]. Furthermore, deletion of Tet1 in mPGCs drops 5hmC by a further order of magnitude, while DNA demethylation proceeds unencumbered. In fact, the role of Tet1-mediated hydroxylation in mPGCs appears to be to protect newly demethylated DNA from inappropriate re-acquisition of DNA methylation, a mechanism which mirrors the role of Tet3 in the zygote [Amouroux et al., 2016;Hill et al., 2018]. This mechanistic distinction is consistent with the observation that imprint abnormalities occur only sporadically and not universally in the offspring of Tet1/2 double knock-out embryos [Dawlaty et al., 2013]. These data could be adequately explained by a failure to protect the newly DNA demethylated genome from ectopic DNA methylation rather than due to a primary problem with imprint erasure. Finally, Tet1 has been shown to play an independent role as a transcriptional activator, both of GRR genes and evolutionary young retrotransposons [Hill et al., 2018]. So, while Tet-mediated hydroxylation is important in maintaining a demethylated state and may play a role in demethylating specific loci, this does not appear to be the mechanism of global DNA demethylation during epigenetic reprogramming. Indeed, the mecha-nism remains enigmatic, although there is evidence that the base excision repair (BER) pathway is active in mP-GCs and required for DNA demethylation [Hajkova et al., 2010]. An alternative explanation is that replicationcoupled DNA demethylation is sufficient to explain genome-wide DNA demethylation, essentially representing a continuous passive process secondary to the downregulation of Uhrf1 and repression of Dnmt3a/b in pre-gonadal mPGCs [Ohno et al., 2013;Kagiwada et al., 2013]. These and other potential mechanisms have been reviewed in detail elsewhere [Ooi and Bestor, 2008;Hackett et al., 2012a;Seisenberger et al., 2013;Hill et al., 2014;Saitou, 2021], and undoubtedly will remain an area of fervent research activity.

Human
The availability of fetal samples from human pregnancy terminations has allowed for the elucidation of the transcriptional and epigenetic changes which occur during the colonisation of the gonad Guo et al., 2015;Tang et al., 2015;Li et al., 2017aLi et al., , 2020. In these studies, human germ cells were isolated via cKIT or AP/cKIT surface marker cell sorting from embryos between 5 and 24 weeks, a period that spans epigenetic reprogramming and subsequent differentiation of hPGCs. Additionally, single-cell analysis of more mature fetal germ cells (hFGCs) has revealed sex-specific sub-populations emerging from week 9.
In keeping with the mouse, hPGCs arriving at the gonad already have lower methylation levels than the surrounding tissue or that of the inner cell mass (ICM). Once in the gonads, hPGCs continue DNA demethylation, with levels decreasing to a trough of approximately 5% by week 7, as measured by bulk whole genome bisulfite sequencing [Guo et al., 2015;Tang et al., 2015]. Using single cell bisulfite sequencing for hPGCs, Li et al. [2020] both corroborated earlier findings and also revealed the DNA methylation states of later hFGCs. Most intriguingly, all subpopulations of male and female hFGCs analysed showed persistent low levels of DNA methylation (4-10%). Overall, this indicates that the very low DNA methylation established in gonadal hPGCs are maintained in hFGCs for months. While this mirrors the low levels of DNA methylation maintained in mouse oogoonia, it contrasts markedly with the rapid restoration of DNA methylation observed in male gonocytes (also known as prospermatogonia [McCarrey, 2013]) in mice. When exactly DNA remethylation occurs during sex-specific differentiation of hFGCs therefore remains an open and interesting area for future study. DOI: 10.1159/000520412 Gene bodies, enhancers, promoters, and intergenic regions all show very low methylation levels, as do CpG islands Guo et al., 2015;Tang et al., 2015;Li et al., 2020]. Promoter methylation falls to 10% in germ cells compared to 80% in the soma, while only 12% of demethylated promoter-controlled genes show upregulation in hPGCs. These upregulated genes are mainly associated with germline processes, such as DAZL and piRNA pathway associated genes DDX4 and PI-WIL1/2. Meiotic gene expression is found upregulated only in female hFGC subpopulations from week 11 onwards [Li et al., 2017a;Chitiashvili et al., 2020], but these meiotic genes are not transcribed in male hFGCs at matched time points, despite similarly low levels of DNA methylation, suggesting roles for activatory processes perhaps linked to dynamic signalling as germ cells mature within the gonadal niche.
DNA methylation at most imprinted DMRs in hPGCs has already fallen to below somatic and earlier ICM levels during migration, but like the mouse, imprint erasure appears to be completed only after gonadal colonisation [Guo et al., 2015;Tang et al., 2015;Li et al., 2020]. The maternally methylated DMRs of PEG10 and IGF2R appear to show slower dynamics Li et al., 2020] but are completely erased in mature sperm.
Notably, several genomic regions were found to retain methylation at a higher level than the genome average and are termed "escapees" Dietmann et al., 2020]. Interestingly, these correspond to gene ontology terms relating to brain-expressed loci, with potential links to neurological and metabolic disorders in humans, albeit with the caveat that neuronal genes are statistically much larger in terms of genome size [Sahakyan and Balasubramanian, 2016;Lopes et al., 2021], and possibly more likely to be enriched in genomic analyses. A similar trend was seen in recently profiled pig PGCs [Zhu et al., 2021]. Thus, it has been proposed that DNA methylation states at these regions might be heritable, with implications for transgenerational inheritance of epialleles Dietmann et al., 2020] (discussed below).
Although TEs are highly variable between species, the overall rules of DNA methylation at TEs appear to follow the pattern observed in mouse in both human Guo et al., 2015;Tang et al., 2015;Li et al., 2020] and pig PGCs [Zhu et al., 2021]. While DNA demethylation follows the genome-wide pattern at most TEs, DNA methylation is retained at specific TE families -most notably evolutionary young TEs, for instance L1HS and AluY in humans, PRE-1 in pigs, and IAPEz in mouse [Zhu et al., 2021]. Single-cell ATAC-seq probing chromatin accessibility meanwhile showed the primateexclusive SINE-VNTR-Alus (SVAs), followed by the HERVKs, HERV1s, Alu elements and SINEs had enrichment for accessibility in hPGCs and hFGCs compared to the soma [Li et al., 2020]. However, despite DNA demethylation and increased accessibility, TE expression remains low compared to their juxtaposed neighbouring somatic cells, as measured by standard scRNA or bulk RNA-seq techniques. The only exception noted so far is that levels of SVA transcripts were noted to creep upwards from week 5.5 to week 9, while HERVK and HERVH transcripts are also detectable Li et al., 2020;Pontis et al., 2021].
Less information is available regarding the mechanism of DNA demethylation in humans. UHRF1, and to a lesser extent DNMT3B, are repressed in migratory cells, however DNMT1 still remains expressed in all germ cells stages at a similar or higher levels than the soma [Guo et al., 2015;Tang et al., 2015;Li et al., 2017a]. The downregulation of the former 2 factors in migratory PGCs is supported by cross-species studies in pig: DNMT3B and UHRF1 were both shown to be expressed at lower levels than the epiblast and soma [Zhu et al., 2021]. Of the TET enzymes, TET1 shows the highest expression compared to the soma in gonadal germ cells [Guo et al., 2015;Tang et al., 2015], with stainings suggesting 5hmC is present in migratory PGCs, before falling rapidly at the gonad at week 7 . Intriguingly, the BER pathway was found to be upregulated in both human and pig PGCs [Guo et al., 2015;Zhu et al., 2021].

Histone Modifications
In addition to DNA demethylation, the chromatin landscape also changes during reprogramming of mP-GCs, suggesting the presence of global epigenome reprogramming that includes global alterations in histone modifications, histone exchange, and changes to nuclear architecture.

Mouse
Alteration to histone modifications in mPGCs have generally been considered to occur in two main phases, with distinct changes noted in pre-gonadal and gonadal PGCs (Fig. 2). In early mPGCs, H3K27me3, H3K4me2, and H3K9ac levels rise from E7.5 to E11.5. H3K9me2 declines during this window, although H3K9me3 stays relatively constant over constitutive heterochromatin [Seki et al., 2005[Seki et al., , 2007Hajkova et al., 2008]. At E11.5, repressive marks (H3K9me3 and H4/H2AR3me2) and permissive Sex Dev 2021;15:411-431 DOI: 10.1159/000520412 mark (H3K9ac) diminish , while different studies observe either loss Mansour et al., 2012] or maintenance of H3K27me3 [Kagiwada et al., 2013] during this time window. It has also been observed that the histone variant H2A.Z is removed from demethylating mPGCs, with H1 chaperone NAP1 entering the nucleus at the point where 5mC drops in the gonad, while H3.3 chaperone HIRA is also expressed . Replacement of canonical histones with their histone variant counterparts has been suggested as one mechanism which could underly the rapid disappearance of these marks , although dynamic changes in linker H1 and the histone chaperone NAP1 were not observed in a subsequent study [Kagiwada et al., 2013]. More recently, dynamic regulation of histone H1 subtypes has been observed in vivo using knock-in mouse lines in which endogenous H1 subtypes are tagged [Izzo et al., 2017]. Notably, H1.4 and H1.10 appear to aggregate at the nuclear membrane around E11.25 before becoming undetectable at E11.5, before re-appearing again at the nuclear envelope at E11.75 and spreading throughout the nucleus, emphasizing a degree of dynamic control that could easily be missed by using alternative methods [Izzo et al., 2017].
Following reprogramming, chromatin re-compacts with nuclei shrinking Kagiwada et al., 2013], chromocenters returning, and H3K9me3 and H3K27me3 being re-established . Meanwhile, H3K9me2 slightly increases from E13.5 [Abe et al., 2011] but remains low compared to the soma [Deguchi et al., 2013], while H3K9ac and H4/H2AR3me2 do not appear to return after reprogramming . For histone H1, differences between the sexes begin to show post-reprogramming, with female mPGCs showing higher H1.1 levels than the soma and male mP-GCs showing lower H1.4 than the surrounding soma [Izzo et al., 2017].
The function of a small number of histone modifying enzymes/complexes has been directly assessed in PGCs. Deletion of polycomb repressive complex 2 (PRC2) component embryonic ectoderm development (Eed) and the accompanying H3K27me3 loss in mPGCs led to defects in both male and female mPGCs, with the female germline being more sensitive to premature upregulation of sex-specific differentiation genes at E11.5 and E13.5 [Lowe et al., 2019], emphasising sexually-divergent dependencies on H3K27me3 and perhaps on additional histone modifications yet to be studied. The related histone mark H2AK119ub, which is deposited by polycomb repressive complex 1 (PRC1), is also present at E12.5 [Yo-kobayashi et al., 2013]. Co-regulation by PRC1 and PRC2 complexes is well described [Laugesen et al., 2019]; however, knockout of PRC1 component Rnf2 in mPGCs does not affect H3K27me3 deposition. Instead, it leads to a downregulation of Oct4 in female mPGCs at E12.5, while double knockout of Rnf2 and Ring1, another PRC1 component, leads to a reduction of mPGCs in both sexes. Loss of H2AK119ub affects female mPGCs more than male, with the major defects occurring in the regulation of meiosis. Female cells begin the process of meiosis at E13.5 in wildtype PGCs, but upregulation of Stra8, Rec8, and Syp3 occurs at E11.5 in Rnf knock-out cells when retinoic acid (RA) is produced in the gonad [Yokobayashi et al., 2013]. Male germ cells enter mitotic arrest at E13.5 [McLaren, 1984;Vergouwen et al., 1991] with meiotic entry prevented due to the degradation of RA by Cyp26b1 [Bowles et al., 2006]. However, knockout of PRC1 components is insufficient to trigger precocious meiotic entry. Notably, when Rnf2 knock-out PGCs of either sex are treated with RA in vitro, an early and increased expression of Stra8 is observed, suggesting that PRC1 plays a role in repressing the meiotic programme in both sexes [Yokobayashi et al., 2013]. Deletion of the X-linked H3K27me3 demethylase Utx appears to reduce the germline transmission of male mESCs in chimaeric animals, and when chimaeric embryos are assessed, Utx null mPGCs exhibited reprogramming abnormalities in the gonad with higher levels of H3K27me3 than controls. However, such abnormalities are compatible with fertility in males, while the knockout is embryonic lethal in females [Mansour et al., 2012;Welstead et al., 2012]. Meanwhile, the aforementioned arginine methyltransferase Prmt5 forms a complex with Blimp1 in the germline at E8.0, allowing translocation into the nucleus and leading to methylation of H2A/ H4R3 in germ cells [Ancelin et al., 2006]. Deletion of Prmt5 in mPGCs leads to a loss of H2A/H4R3me2 and apoptosis, with complete loss of germ cells by E15.5-E16.5 [Kim et al., 2014;Li et al., 2015]. Knock-out mPGCs show activation of the DNA damage response [Kim et al., 2014] and splicing defects , but also higher expression of TE elements, including IAP-LTRs, IAP-GAG, and L1 elements, indicating a role for Prmt5 in TE silencing in early mPGCs [Kim et al., 2014]. Later, in male gonadal mPGCs, Prmt5 translocates back to the cytoplasm where it methylates PIWI family proteins Kim et al., 2014]. This post-translational modification enables interaction with Tudor-domain proteins, essential for subsequent PIWI-interacting RNA (piRNA) biogenesis and TE silencing by DNA methylation [Aravin et al., 2008] (discussed below). Thus, Prmt5 DOI: 10.1159/000520412 is implicated in a range of epigenetic mechanisms to regulate TE expression, including histone modifications, non-coding RNAs, and DNA methylation.

Human
In humans, only immunofluorescence data are currently available [Gkountela et al., 2013Guo et al., 2015;Tang et al., 2015;Eguizabal et al., 2016]. These show that H3K9me3 and H3K9me2 are depleted in hPGCs in comparison to neighbouring somatic cells, although distinct and bright punctae of H3K9me3 remain within the nucleus. H3K27me3, although initially enriched at week 4, is thereafter downregulated in hPGCs in comparison to somatic neighbours. Active marks such as H3K4me1/3 increase during this time period to levels comparable to the soma, while H3K9ac shows a marked increase even above the levels observed in neighbouring cells [Eguizabal et al., 2016]. The higher numbers of hPGCs make epigenomic studies using ChIP-seq or CUT&RUN [Brind'Amour et al., 2015;Skene and Henikoff, 2017] feasible despite the challenges in obtaining samples, and locus specific information awaits such studies. However, the asynchrony of hPGC development will be a challenge in interpreting such data, unless the distinct germ subpopulations are cleanly isolated using unique cell-surface markers [Li et al., 2020;Mishra et al., 2021]. Indeed, detailed immunostaining of histone modifications in different hFGC subpopulations (stained with the appropriate stage-specific hFGC markers), has not yet been undertaken.
More globally, in both mouse and human PGCs, nuclear size appears to increase during this window, suggesting a widespread dismantling of chromatin structure while demethylation occurs Gkountela et al., 2015;Guo et al., 2015;Tang et al., 2015]. This may be necessary to allow the final stages of DNA demethylation or alternative functions such as recruiting the DNA repair pathway for rapid active DNA demethylation.

Lost in Translation? Non-Coding RNAs and the Epitranscriptome in Early Germline Development
In addition to covalent modifications of chromatin, roles for small (<200 nt) and long (>200 nt) non-coding RNAs as well as RNA modifications have also been explored in the context of mammalian germline development (Fig. 2, 3).
Germline small RNAs include ∼22 nt microRNAs (miRNAs), 25-32 nt PIWI-interacting RNAs (piRNAs), as well as ∼22 nt endogenous-silencing RNAs (endo-siR-NAs). Each have distinct biogenesis pathways, with the small RNAs acting as guide RNAs for their associated Argonaute proteins, leading primarily to post-transcriptional gene silencing via Watson-Crick base pairing. miRNAs miRNAs may play a role in triggering mPGC specification, with the inhibition of let-7 miRNA biogenesis by Lin28a appearing to be required for PGC development [West et al., 2009]. In an in vivo chimaera assay, the depletion of Lin28a led to a reduction of the Stella-positive mPGCs in the gonad, which was rescued upon the introduction of Blimp1 with an ectopic 3′-UTR carrying a mutated let-7 miRNA target site [West et al., 2009]. This suggests a model in which downregulation of let-7 via its repressor Lin28a is required to allow appropriate expression of key PGC specification gene Blimp1. However, let-7 miRNAs are also known suppressors of the pluripotent state, whilst encouraging the somatic programme, whereas Lin28 antagonises this, maintaining pluripotency. So, whether this regulatory circuit is required for specification versus maintenance of the germline fate is less clear. Meanwhile, the role of LIN28A in hPGCLC specification remains as of yet unstudied.
Thereafter, in mouse gonadal germ cells, the miR-17/92 family, miR-181 family, and miR-290 cluster, among a handful of others, were observed to be highly expressed, with putative roles in proliferation of bipotential early mPGCs Bhin et al., 2015;Fernandez-Perez et al., 2018], while depletion of somatic let-7 miRNAs continued to be observed. This putative function was ascertained via the use of a conditional germline-specific knockout (AP-Cre) of the miRNA biogenesis factor Dicer1 in PGCs Maatouk et al., 2008], which identified defective proliferation of mPGCs as well as arrested spermatogenesis. In addition, while Dicer is important for the biogenesis of miR-NA, it is also required for the biogenesis of other small RNA types such as endo-siRNA, with additional roles in post-transcriptional gene silencing [Czech et al., 2008;Kim et al., 2009], thus confounding the interpretation of the specific roles of miRNAs in mPGCs.
A chief observation by the above studies was the high expression of all members of the miR-290 cluster in mP-GCs. This cluster of 7 miRNA hairpins, largely sharing the same seed sequence, is associated with developmentally regulated expression in pluripotent cells, with its expression robustly seen in the ICM of mouse blastocysts, as well as in naive mESCs [Parchem et al., 2014]. Intrigu-ingly, miR-290 expression drops upon the specification of the epiblast as well as in mEpiSCs, both sharing primedpluripotent states. Meanwhile, miR-290 cluster expression picked up again exclusively in the germline in the embryo, with additional trophoblast expression also continuously seen [Parchem et al., 2014;Paikari et al., 2017]. This pattern of behaviour contrasted with that of a related family of miRNAs, the miR-302 cluster, which is upregulated in E5.5-6.5 epiblasts, and remains present in late epiblasts, but is subsequently absent in mPGCs upon their specification. Taken together, this has led to the model of alternating expression levels of these clusters in the progression from ICM/naive-mESC, to epiblast, and then germline cell fate.
Crucially, the generation of a miR-290 knock-out mouse demonstrated the key role of this cluster, with pervasive (yet incompletely penetrant) embryonic lethality. In embryos that survived, a vast reduction in mPGC numbers was seen at both E11.5 and E13.5. Females were rendered infertile, while males remained fertile due to the continued viability of the very few mPGCs that survived [Medeiros et al., 2011]. A further study also revealed catastrophic placenta defects that could also explain part of the embryonic lethality seen, as the miR-290 cluster is indeed observed to be expressed in the trophoblast [Paikari et al., 2017]. This experiment together with Dicer knockouts described above therefore bolsters the observation that miRNAs are important for the regulation of gene networks in mPGCs.
In the human, no studies have yet comprehensively profiled the miRNAs of isolated hPGCs nor of their in vitro counterpart, the BMP-induced hPGCLCs. However, germ cell cancers, and the seminoma cell line TCAM-2, have been extensively profiled for miRNAs, with the equivalent of the miR-290 cluster in the human, the miR-371 cluster (composed of 3 hairpins, producing 6 mature miRNAs), spotted to be highly abundant [Murray et al., 2011;Novotny et al., 2012]. In addition, naive-hESCs have also been profiled for miRNAs, with a similar observation noted [Faridani et al., 2016;Dodsworth et al., 2020]. Therefore, it may be reasonable to expect that the miR-371 cluster is also expressed in the human germline, considering the similar behaviour of this cluster in both mouse and human naive ESCs, as well as its presence in seminoma whose transcriptome overlaps somewhat with that of hPGCs. A concerted approach towards profiling the miRNAs in human germline development and identifying their mRNA targets will thus certainly be useful to the field. Upon sex specification later in gonadal germ cell development, miRNA profiles in mouse fetal germ cells (mFGCs) diverge. In particular, miR-17/92 and the let-7 families decreased in female mFGCs as they entered meiosis, while their levels increased in male mFGCs as they entered mitotic arrest at E13.5 Bhin et al., 2015]. miR-29b was meanwhile found to be exclusively expressed in female mFGCs from E13.5 with putative targets Dnmt3a and Dnmt3b [Takada et al., 2009], potentially contributing to low DNA methylation levels in contrast to their remethylated male counterparts at the same time points. Full profiles of miRNAs in both male and female mFGCs beyond E15.5, as well as hFGCs, however, remain unknown. piRNAs piRNAs are restricted to metazoa, with signatures of 2′-O-methyl-modified 3′-ends and a 5′-U bias, while deploying the PIWI-clade of Argonautes (PIWI proteins), distinguishing them from miRNAs [Weick and Miska, 2014;Czech et al., 2018;Özata et al., 2020]. In mammals, their expression is restricted to the germline, with a range of lengths seen depending on the PIWI protein they are found associated with.
Unlike miRNAs, piRNAs do not play a role in PGC specification but instead appear later during sex-specific differentiation. In the mouse, 3 PIWI proteins are expressed in the male only (Piwil1/Miwi, Piwil2/Mili, and Piwil4/Miwi2), with their individual knockouts leading to spermatogenic defects and complete male infertility [Deng and Lin, 2002;Kuramochi-Miyagawa et al., 2004;Carmell et al., 2007]. Piwil2/Mili expression begins in both sexes at E12.5, followed by male-exclusive Piwil4/ Miwi2 at a window of E13.5 to postnatal day (P)3, then male-exclusive Piwil1/Miwi from P14 at the beginning of the pachytene-stage of meiosis [Aravin et al., 2008]. The coincidence of Piwil2/Mili and Piwil4/Miwi2 in male mP-GCs yields 26-29 nt long piRNAs (pre-pachytene piR-NAs). Longer 30 nt pachytene piRNAs (not discussed here) are produced in the male adult upon the co-expression of Piwil2/Mili and Piwil1/Miwi from P14. In particular, Piwil2/Mili and other PIWI pathway members have been shown to be strongly activated upon DNA demethylation [Hackett et al., 2012b], suggesting a direct role of DNA methylation regulation on their expression.
The knockout of either Piwil2/Mili and Piwil4/Miwi2 leads to the arrest of spermatogenesis at early prophase of meiosis I (pre-pachytene). Molecularly, these knockouts correspond to the upregulation of TEs, in particular, evolutionary young L1s (as the case for Piwil4/Miwi2 and DOI: 10.1159/000520412 cofactor Mael) [Carmell et al., 2007;Aravin et al., 2009], with DNA methylation defects over TEs noted. Murine pre-pachytene piRNAs derive from long single-stranded precursors, subsequently processed by numerous members of an extensive PIWI pathway into multiple smaller transcripts [Özata et al., 2020]. Homotypic Piwil2-Piwil2 and, to a smaller extent, heterotypic Piwil2-Piwil4 interactions in the perinuclear nuage amplify primary-piR-NAs to generate secondary/antisense-piRNAs by a pingpong amplification loop [DeFazio et al., 2011].
The majority of pre-pachytene piRNAs are known to map to TEs, whereas more non-TE mapping is seen for pachytene piRNAs [Aravin et al., 2008]. Pull downs for Piwil2/Mili (26 nt preference) and Piwil4/Miwi2 (29 nt preference) show this difference, with Piwil2/Mili (also seen in the adult stage) already binding piRNA that map to fewer TEs than does Piwil4/Miwi2 at E16.5. The rules governing piRNA-mediated TE targeting are similar to the seed-based mechanism employed by miRNAs [Goh et al., 2015;Shen et al., 2018], albeit less thoroughly studied.
In addition to the endonucleolytic cleavage of these TE transcripts, Piwil4/Miwi2 is known to translocate into the nucleus from E16.5 [Aravin et al., 2008]. This has subsequently been shown to result in transcriptional silencing of especially the evolutionary young TEs, with downstream members of the DNA methylation machinery Dn-mt3l, the murine-specific Dnmt3c, as well as the Dnmt3l-Miwi2/Piwil4 co-interactor Spocd1, shown to recapitulate the DNA methylation TE defects of a Miwi2 knockout [Aravin et al., 2008;Barau et al., 2016;Vasiliauskaitė et al., 2018;Zoch et al., 2020;Dura et al., 2021]. Additionally, a Tex15 knockout was also shown to recapitulate this DNA methylation defect, although its involvement in the DNA methylation pathway remains unknown and may instead function via its DUF3715 domain, found also in Tasor of the HUSH complex, linking Piwil4/Miwi2 nuclear silencing to Setdb1-linked H3K9me3 repression [Ninova and Tóth, 2020;Schöpp et al., 2020;Yang et al., 2020], with H3K9me3 establishment previously shown to be implicated downstream Piwil4/Miwi2 and linked to DNA methylation [Liu et al., 2014;Pezic et al., 2014].
In humans, 4 PIWI paralogues are present, with PIWIL3 noted to be oogenic-specific in mammals but absent in the Muridae family of rodents. Indeed, robust populations of piRNA can be detected in the mature mammalian oocytes of human, cow, macaque, and golden hamster [Roovers et al., 2015;Williams et al., 2015;Hasuwa et al., 2021;Loubalova et al., 2021;Zhang et al., 2021].
In the developing human germline, piRNAs have been detected mainly by oxidising whole fetal gonadal samples to enrich for 2′-O-methyl-modified 3′-end small RNAs. In summary, piRNAs could not be detected when first trimester samples were studied. Among second trimester samples, Gainetdinov et al. [2017] were able to investigate the very low levels of oxidized male week 20 piRNA dataset from Williams et al. [2015] to conclude that these pre-pachytene piRNA target the evolutionary youngest TEs, as do later pachytene piRNAs, but that later pachytene piRNAs also map to older TEs. Meanwhile, Reznik et al. [2019] claimed to identify piRNAs with a ping-pong signature, although they could only study the 0.53% of Repbase-mapped small RNA reads in their non-oxidized heterogenous testis small RNA samples which are otherwise populated with somatic small RNA types. For female samples, 28/27 nt-peak piR-NAs were detected in Williams et al. [2015] and Roovers et al. [2015] oxidized late second trimester samples, with a large divergence in TE-mapping rates noted (2% versus 30% respectively) and with the detection of ping-pong signatures only in Williams et al. [2015]. In all papers, no attempt was made to isolate pure germ cell populations. A more concerted approach to rigorously profile the emergence of piRNAs and their sequences in the context of the human germline, and with attention to the specific germsubpopulations, is therefore warranted.
Additionally, microscopy-based descriptions of PIWI paralogue distribution and expression dynamics in the developing human germline have been noted. Cytoplasmic PIWIL1 was detected only in female mitotic-hPGCs, with single granules seen later in hFGCs in females, but not males [Fernandes et al., 2018]. Cytoplasmic PIWIL2 was detected in both male and female hPGCs and hFGCs [Fernandes et al., 2018], while PIWIL4 localised to the perinuclear/inter-mitochondrial cement in both sexes [Fernandes et al., 2018;Reznik et al., 2019], with instances of nuclear PIWIL4 localization [Reznik et al., 2019;Guo et al., 2021] in male hFGCs also reported. PIWIL3 was not detected in the developing fetus [Fernandes et al., 2018], which has only been observed in adult oocytes instead [Roovers et al., 2015;Fernandes et al., 2018;Yang et al., 2019].

endo-siRNAs
Unlike piRNAs, endo-siRNAs are more ubiquitously expressed [Kim et al., 2009]. They derive from doublestranded RNA duplexes which can occur as a consequence of convergent sense-and antisense-transcription across the genome, which can typically occur across TE sequences when they run antisense to gene bodies or on self-complementary single-stranded RNA sequences. Duplexes are recognised as a substrate for DICER1 cleavage, thus merging with the miRNA biogenesis pathway. These endo-siRNA are then loaded into AGO1-4, silencing targets based on complementary seed-based pairing [Xia et al., 2013;Piatek and Werner, 2014;Svoboda, 2014;Taborska et al., 2019].
While a clear role for the endo-siRNA pathway is seen in adult murine oocytes [Murchison et al., 2007;Nagaraja et al., 2008], with the involvement of an oocyte-specific Dicer isoform [Flemr et al., 2013], much lower levels of endo-siRNAs were noticed to be upregulated in early mPGC development coinciding with the onset of DNA demethylation [Berrens et al., 2017]. The same observation was noted in parallel, with a Dnmt1 conditional knockout in mESCs artificially causing passive demethylation, whereupon endo-siRNAs were upregulated to silence TEs before the establishment of repressive histone marks. However, any functional roles for these endo-siRNAs in mPGC development were not assessed. A hint of the role of endo-siRNA in PGC development can however be delineated with the observation that a more severe impact on spermatogenesis resulted as a consequence of conditional Dicer1 knockout, compared to Dgcr8 knockout (Vasa-Cre), with Dicer1 implicated in both miRNA and endo-siRNA biogenesis, whereas Dgcr8 is only implicated in miRNA biogenesis [Zimmermann et al., 2014].
lncRNAs Long non-coding RNAs (lncRNAs) are distinguished from small non-coding RNAs by having sizes greater than 200 nt, while similarly being non-protein coding [Hombach and Kretz, 2016]. A small proportion have been reported to play roles in development by different mechanisms of action, such as in transcriptional, posttranscriptional, and epigenetic regulation of chromatin in either cis or trans [Liu and Lim, 2018].

Xist/Xact
In female mPGCs, the inactive X chromosome is reactivated alongside DNA demethylation with a decrease in Xist clouds between E9.5 and E11.5, followed by a slight window of X dosage excess until E14.5 [Sugimoto and Abe, 2007;Chuva de SousaLopes et al., 2008;Sangrithi et al., 2017], an event recapitulated in the demethylated ICM [Mak et al., 2004], thus ensuring that each oocyte contains an active X chromosome for early embryonic development after fertilisation. This dosage excess in mP-GCs normalises back down to an X:autosome (X:A) ratio of 1 in developing oogonia from E15.5 [Li et al., 2017b;Sangrithi et al., 2017].
In the human, earlier work suggested that the X chromosome was already activated in hPGCs due to the noted lack of H3K27me3 foci over the X chromosome in week 7 hPGCs compared to neighbouring somatic cells . A recent study by Chitiashvili et al. [2020] proposed a more subtle and unique human germline behaviour in reactivation. The X was claimed to be reactivated but slightly dampened, such that X:A ratios were <1. The bi-allelic presence of the primate-exclusive X chromosome active lncRNA XACT was also noted, doubling as a unique marker of hPGCs. X reactivation was also noted in later hFGC subpopulations, followed by an XIST-independent inactive state in oogenic hFGCs. However, a different study relying on earlier bulk RNA datasets [Sangrithi et al., 2017] observed the opposite behaviour with X chromosome dosage excess observed during initial reactivation, followed by a return to an X:A ratio of 1 in hPGCs. Further, more detailed studies will therefore be needed to resolve these contrasting observations.

Others
The roles of other lncRNAs have not been thoroughly explored in PGC development. The only study to focus on lncRNAs in mPGC development identified a differential increase in 2,500 and decrease in 1,000 lncRNAs in the progression from E12.5 to E15.5 in male mPGCs [Bao et al., 2013]. No follow-up functional validation of candidate lncRNAs have yet been performed. Meanwhile in hPGCs, lncRNAs derived from the TE HERVH were found to be abundantly expressed Pontis et al., 2021]. While HERVH lncRNAs in hESCs are known to associate with OCT4 to regulate pluripotency [Lu et al., 2014], HERVH lncRNAs expressed in hPGCs were distinct from that of hESCs, with any functions remaining unknown.
The rich bulk and single-cell transcriptomic datasets of mPGCs and hPGCs nevertheless present an opportunity for the systematic identification and exploration of lncRNA expression dynamics in both male and female PGC development. The equivalent in vitro model systems, combined with methods for CRISPRi lncRNA screens  further provide an opportunity for the functional validation of any candidate lncRNA involved in PGC specification and development.

RNA Modifications m6A
Post-transcriptional modifications of transcribed RNA, or the epitranscriptome, can also contribute to the regulatory landscape of developing cells. The N 6 -methyl-DOI: 10.1159/000520412 adenosine (m6A) modification is the most abundant covalent modification of mRNA [Lasman et al., 2020] and has been characterised in the developing mouse germline. m6A reader Ythdc2 knockouts or knockdowns lead to both male and female infertility due to a defect in switching from a mitotic to a meiotic programme: female mPGCs fail to properly execute meiosis, while male germ cells fail to progress through meiosis, leading to apoptosis [Bailey et al., 2017;Hsu et al., 2017;Wojtas et al., 2017;Zeng et al., 2020]. In parallel, m6A writers Mettl3/14 conditional knockouts (Vasa-Cre) led to the inhibition of meiosis in male mPGCs [Lin et al., 2017;Xu et al., 2017], with observations in female mPGCs lacking.

Polyuridylation
Terminal uridylyl transferases (TUTases) are known to polymerise untemplated U nucleotides at the 3′-end of RNA, or occasionally a single U addition, known to impact the regulation of the transcript, as well as especially the let-7 family of miRNAs as mentioned earlier [Thornton et al., 2012;Lim et al., 2014]. In relation to TEs, knockouts of Tut4/7 were recently shown to cause the surge of L1 retrotransposition events in mammalian cells [Strzyz, 2018;Warkocki et al., 2018]. Indeed L1s were found to be 3′-uridylated in human cells and mouse testes [Warkocki et al., 2018]. Their roles in mammalian germline development (in particular PGCs) are hitherto unknown, although a role was identified for the TUTases in the clearance of transcripts in later development in both mouse spermatogenesis and oogenesis [Morgan et al., 2017[Morgan et al., , 2019.
In both these cases, the roles of post-transcriptional covalent modification or tailing of RNA in shaping the regulatory landscapes in developing hPGCs remain uncharacterised and present an opportunity for further work.

Model Organisms for Mechanistic Studies
For mechanistic studies, the mouse will remain an essential tool due to the ease of embryo and gene manipulation. Indeed, the lack of available tools and/or limited access to embryos makes it difficult to conceive that new molecular mechanisms will be uncovered in other mammalian model organisms or in human fetal samples. While the PGCLC system has revolutionised the field, any mechanistic insights will inevitably require in vivo confirmation. Recent work on PGCs in pigs [Kobayashi et al., 2017;Zhu et al., 2021] and rats [Leitch et al., 2010;Northrup et al., 2011;Encinas et al., 2012;Kobayashi et al., 2020], as well as the observation of female PGC piR-NAs in a non-murine rodent, the golden hamster Ishino et al., 2021;Loubalova et al., 2021;Zhang et al., 2021], do show promise and will hopefully make an increasing contribution to complement mouse studies in the years to come. While recent studies in nonhuman primates have been illuminating, their use does carry an increased ethical burden, particularly for studies into fundamental mechanisms that could be first undertaken in other species. Without doubt, mechanistic studies in model organisms will continue to be crucial and allow confirmatory studies in human using available fetal tissue and in vitro systems.

Developmental Timing and the Problem of Asynchrony
It has become clear, especially from single cell studies, that while mPGCs differentiate into either mitotically (male) or meiotically (female) arrested gonocytes from E13.5, around a week after they were first specified, hP-GCs persist much longer, with asynchronous differentiation into later hFGCs observed [Li et al., 2017a;Chitiashvili et al., 2020;Hwang et al., 2020;Guo et al., 2021]. In particular, founder hPGC subpopulations persist up till the latest examined timepoint of week 26, while male mitotic-arrest hFGCs start to emerge from week 10, and 3 sequential subpopulations of female hFGCs (RA-responsive, meiotic, and oogenic) emerge from week 11, 14, and 18, respectively, whilst also all persisting up until week 26 [Li et al., 2017a] (Fig. 3). This vastly prolonged differentiation process and coexisting germ cell subpopulations in different stages of maturity differs from that of the mouse and will have implications on both their differing biology as well as the bulk study of human germ cells beyond week 10, with single-cell studies or unique cell-surface markers required for the isolation of pure subpopulations [Li et al., 2020;Mishra et al., 2021].

ncRNAs in the Germline -A Long Way to Go
Genetic perturbation of mPGCs as well as small RNA profiling of isolated, fluorescently-marked mPGCs have allowed the characterisation of some of the roles of miR-NAs, piRNAs, and endo-siRNAs in germline development. However, far less is known about the role of ln-cRNAs. Indeed, we hope this review serves to highlights that our overall understanding of the role of the non-coding genome during early germline development remains limited even in the mouse. In particular, the interactions between ncRNAs and underlying chromatin requires fur-Sex Dev 2021;15:411-431 DOI: 10.1159/000520412 ther investigation. In human studies, even more work is needed. While some progress has been made in characterising piRNAs in second trimester human gonads, little else is known. Looking forwards, the combination of surface-marker cell sorting of pure hPGC populations, and perhaps of hFGC subpopulations, together with small RNA profiling, should help improve the characterisation of small RNAs in human germline development. Where appropriate, such studies could be complemented by knockout of core small RNA components during PGCLC induction and differentiation in vitro. To this end, a promising protocol recapitulating the demethylated and remethylation states and expression of PIWI pathway components in in vitro derived mouse gonocytes has recently been published [Ishikura et al., 2021].

Regulating TEs in the DNA Hypomethylated State
In male mPGCs the window of the demethylated state is short (E11.5-E15.5), with piRNAs emergent from E13.5, whereas the demethylated state persists postnatally in female germ cells. In male and female hPGCs, this demethylated state extends for a much longer period (week 5 to week 26, the latest characterised time point). A locus-specific characterisation of repressive histone marks during PGC development using low-input techniques such as ULI-ChIP or CUT&RUN/CUT&Tag may help shed light on any additional layers of transcriptional repression of TEs. Specific care could also be made to explore TE transcript levels at the individual locus level, rather than the bulk family analytic methods predominantly used [Lanciano and Cristofari, 2020;Stow et al., 2021]. Which additional epigenetic mechanisms help to suppress the activity of TEs during this demethylated time window remain unknown. Given the recent report of hypertranscription in mPGCs, this seems particularly pertinent [Percharde et al., 2017]. It will be interesting to assess if hypertranscription is observed in PGCs of other species, including humans. Intriguingly, in the human testis, high transcript abundance has been observed and proposed to play a role in reducing germline mutation rates across these expressed genes, in a process referred to as widespread "transcriptional scanning" [Xia et al., 2020]. However, hypertranscription is not a reported feature in PGCs of invertebrate model organisms [Oulhen and Wessel, 2017;Lebedeva et al., 2018]. Either way, if confirmed it would seem essential that high levels of TE expression must be regulated somehow to prevent retrotransposition, particularly in view of the tight control of genomic stability in PGCs [Hill and Crossan, 2019]. Perhaps recently described mechanisms of TE suppression found in other contexts may also be operating in PGCs? For instance, an emergent class of small RNAs, precisely cut fragments of tRNAs (tRFs), have been detected in mature sperm, mESCs, and TS cells, with roles in TE regulation described [Sharma et al., 2016;Schorn et al., 2017;Schorn and Martienssen, 2018;Boskovic et al., 2020]. Their presence and roles in PGCs, meanwhile, remain undescribed. Additionally, roles for the m6A modification across TEs in regulating ERV and L1s have recently been described in mESCs [Chelmicki et al., 2021;Hwang et al., 2021;Xiong et al., 2021]. However, a potential role in TE regulation in PGCs has been overlooked in m6A studies thus far. The recent development of techniques to profile m6A RNA modifications directly over long reads may overcome this dearth of knowledge [Pratanwanich et al., 2020]. Indeed, some of these post-transcriptional mechanisms may not alter transcript abundance at all but instead impact the translation of TE proteins required for their retrotransposition. This can be measured by ribosome profiling or by simply detecting TE protein products.

Transgenerational Epigenetic Inheritance
A great deal of interest has been sparked by the socalled "escapee" genes and loci that apparently resist DNA demethylation in PGCs [Hajkova et al., 2002;Hackett et al., 2013;Tang et al., 2015]. Whether such loci truly escape reprogramming or are in fact targeted for active maintenance remains an open question. Such loci might have implications for transgenerational epigenetic inheritance (TEI). However, although TEI is a well described mechanism in other model organisms [Castel and Martienssen, 2013;Heard and Martienssen, 2014;Miska and Ferguson-Smith, 2016;Perez and Lehner, 2019;Miska and Rechavi, 2021] it remains a largely unproven notion in mammals [Heard and Martienssen, 2014;Otterdijk and Michels, 2016;Miska and Ferguson-Smith, 2016;Blake and Watson, 2016]. Such loci would have to further evade zygotic epigenetic reprogramming and the passive DNA demethylation that occurs during pre-implantation development to have an effect on the next generation and/or continue their passage through the germline. Indeed, recent studies have demonstrated that newly introduced epialleles are completely erased during passage through the naive cell state [Carlini et al., 2021] and that metastable IAP epialleles are highly resistant to environmental perturbations in the first place [Bertozzi et al., 2021]. The studies that have rigorously addressed this question have not shown DNA methylation marks that can be maintained following even a single passage through the entire germline cycle [Radford et al., 2014;Kazachenka et al., 2018]. DOI: 10.1159/000520412 Conclusion Despite significant advances in our understanding of epigenetic reprogramming and the role of the non-coding genome in mammalian germline development, this review has highlighted that many basic processes and mechanisms remain poorly understood. New technologies are emerging as a key enabling factor, overcoming the challenges of small cell numbers, limited access to samples, and cellular heterogeneity that have long challenged germline studies. Improved single-cell transcriptomic protocols and analysis pipelines, single-cell multiomics, low-input methods for assessing histone modifications and small non-coding RNAs, and epitranscriptome co-sequencing of long reads may help to provide mechanistic insight in the future. However, these technologies must be applied with rigour, to the right experimental system, and interpreted with a caution that reflects the complexity of epigenetic regulation in the germline. Indeed, that so many questions remain unanswered likely reflects the intricate interactions of overlapping regulatory systems, or alternatively, that we are still missing key parts of the puzzle. As such, while there is a recent ferment in favour of studying human biology directly in human cells, organoids, or embryo-models [Posfai et al., 2021], it seems just as likely that major advances in our understanding of human biology will continue to flow from research on model organisms. This may also be aided by CRISPR/Cas9 technology, which makes multiple genetic manipulations in mammals feasible, allowing more complex genetic experiments in the future. Regardless of model organism or approach, there remain a great many mysteries to uncover in germline epigenetics and the non-coding genome, which will keep the field busy in the decades to come.