Origin and establishment of the germline in Drosophila melanogaster

Abstract

The continuity of a species depends on germ cells. Germ cells are different from all the other cell types of the body (somatic cells) as they are solely destined to develop into gametes (sperm or egg) to create the next generation. In this review, we will touch on 4 areas of embryonic germ cell development in Drosophila melanogaster: the assembly and function of germplasm, which houses the determinants for germ cell specification and fate and the mitochondria of the next generation; the process of pole cell formation, which will give rise to primordial germ cells (PGCs); the specification of pole cells toward the PGC fate; and finally, the migration of PGCs to the somatic gonadal precursors, where they, together with somatic gonadal precursors, form the embryonic testis and ovary.

Introduction

Germ cells fulfill a unique role; instead of contributing to the cells of the body, they enable reproduction. As oocytes and eggs, germ cells pass on genetic instructions to the next generation. In addition, the egg's abundant cytoplasm provides the cytoskeletal structures, organelles including mitochondria (discussed in “Mitochondria are enriched in PGCs”), and reprogramming factors that are critical for embryonic development to the fertilized egg cell. In many species, primordial germ cells, the precursors of egg and sperm, are set aside early during development. The germline–soma dichotomy prevents germ cells from taking on somatic fates and losing their potential for totipotency and restricts somatic cells from reprogramming or forming teratomas. How primordial germ cells are specified differs among species (Extavour and Akam 2003). Germ cell fate can be induced by cell-to-cell interactions and signaling, as it occurs in mice and humans. Alternatively, germ cells can be specified by inheriting maternal factors, which are usually stored within biomolecular condensates (germplasm or germ granules) that are deposited in the egg, as is seen in Drosophila melanogaster (fly), Caenorhabditis elegans (worm), Xenopuslaevis (frog), and Danio rerio (zebrafish) (Voronina et al. 2011). For a discussion of the different mechanisms used in evolution to establish the germ cell fate, and how germ granules and their varying biophysical properties are a hallmark of germ cell fate irrespective of the mode of specification, please refer to recent reviews (Lesch and Page 2012; Strome and Updike 2015; Pamula and Lehmann 2024).

Research in Drosophila melanogaster has led the way for the study of germplasm and germ cell specification. In this review, we will summarize what we have learned about this process (see “Germplasm: history and introduction,” “Oskar mRNA transport and regulation,” “Composition and function of germ granules,” and “Mitochondria are enriched in PGCs”). Starting with a historical perspective, we describe how germplasm was identified by its unique dense morphology in the egg and how the determining function of germplasm was demonstrated by embryo manipulation. The developmental and molecular revolution of the late 20th century led to the identification of genes with germ cell-specific functions and revealed their molecular nature. By studying how germplasm is concentrated in a specific egg region, we describe how the Drosophila field broadly influenced the concepts and underlying cellular mechanisms of RNA localization and spatially restricted translation (see “Oskar mRNA transport and regulation”).

In contrast to the transcription factors and cell signaling pathways that instruct embryonic body patterning and somatic cell fates, similar germ cell-specific transcriptional regulators that instruct germ cell fate have not been identified in any organism. Instead, many germline genes that are essential for germ cell identity have posttranscriptional functions. Indeed, the prominent role of RNA regulatory functions in germ cells is a conserved and defining feature. Proteins such as the ATP-dependent helicase Vasa, the translational repressor Nanos, and the Argonaute protein Piwi were initially identified in Drosophila based on their function in germ cell development. Each of these genes was later found to be a global marker for germ cell fate in organisms ranging from worms to humans, irrespective of the initial mode of germ cell specification (inductive vs inherited) (see “Composition and function of germ granules”). Despite this, without an instructive, fate-specifying transcription factor, the criteria for when during development a cell can be defined as a “germ cell” or a primordial (precursor) germ cell (PGC) are debated (Nicholls and Page 2021). In Drosophila, this debate is particularly salient, as the origin of the germline can be clearly defined as the cells that originate at the posterior pole (pole cells) due to the action of the germplasm and a specialized cellularization process (described in detail in “Formation of the pole cells, the precursors of the PGCs”). However, while pole cells can give rise to “PGCs,” not all pole cells are fated to become PGCs. Under normal circumstances, pole cells that do not make the transition to PGCs die rather than taking on a different, somatic fate. Therefore, we will refer to these first cells as “pole cells” and reserve the term “PGCs” for cells that initiate the PGC program and have begun their migration toward the somatic gonad (described in “The primordial germ cell transcriptional program” and “PGC migration”).

The repression of transcription, specifically of genes expressed in somatic tissues, has broadly been recognized as another defining feature of early germ cell fate (Cinalli et al. 2008). In Drosophila, transcription is actively repressed in pole cells when they first form and is activated significantly later than in the somatic cells. In “The primordial germ cell transcriptional program,” we describe how the transcriptional germ cell program is regulated and transitions from a maternal to a zygotic program of gene expression. A final conserved feature of embryonic germ cells is that they migrate from one location, often at the fringe of the embryo where they were formed or specified, toward the somatic gonad. In Drosophila, this migration process has been separated into various steps based on direct observation of the changing cell and tissue contacts, and on the action of specific gene functions (described in “PGC migration”). Like PGCs in other organisms, robust germ cell-specific transcription is observed only once germ cells interact with the somatic cells of the gonad, at which point PGCs begin to express sex-specific genes that are critical for differentiation into egg and sperm.

We are largely restricting our description of germ cell development to the embryo, except for summarizing the processes of germplasm assembly during oogenesis. The early stages of oogenesis that lead to oocyte determination and the establishment of oocyte polarity and the mechanisms protecting germline fate have been described in other chapters of this book and in recent reviews and will not be covered in depth here (Gleason et al. 2018; Drummond-Barbosa 2019; Hirakata and Siomi 2019; St Johnston 2023).

Together the detailed analysis of the earliest stages of embryonic germ cell development in Drosophila summarized in this review illustrates the deep understanding of the developmental mechanisms that define this unique cell fate and connects the unique morphologies of early germ cells with their privileged function as bearer of genetic information for the next generation.

Germplasm: history

August Weismann first postulated that an inheritance substance was handed down from parent germ cells to the germ cells of the new individual (Weismann 1893). Weismann termed such material Keimplasma (germplasm). While the “germplasm” that Weismann referred to was meant as a property more akin to genetic information that is passed on from generation to generation [the genetic material (DNA) carried by chromosomes], the presence of cytoplasmic regions in the egg marking the future germ cells in some species invoked the hypothesis that specific cytoplasmic substances or structures are responsible for giving rise to germ cells (Eddy 1975).

Hegner was the first to observe a cytoplasmic substance that he hypothesized may be responsible for germ cell specification. He observed cytoplasmic granules at the posterior of insect oocytes, connected these to the PGCs of the next generation, and proposed that these granules could be germ cell determinants (Hegner 1911, 1914). Since then, similar granular structures have been detected in specific areas of embryo-plasm and in PGCs of various animals, including Drosophila (Sonnenblick 1959; Mahowald 1962; Eddy 1975). Together, these studies correlated a specialized cytoplasmic area containing these granular structures to the formation of PGCs, giving rise to the modern concept of germplasm.

Direct experimental proof of germplasm function in germ cell formation came from experiments carried out by Geigy (1931). He irradiated the posterior pole of D. melanogaster eggs, where germplasm was observed, with ultraviolet (UV) light and obtained adult flies that lacked germ cells. Okada et al. (1974) later rescued this germ cell formation defect by transplanting nonirradiated posterior cytoplasm into the UV-irradiated embryos. The final piece of evidence came when Mahowald and Illmensee injected posterior pole cytoplasm into ectopic sites of early Drosophila embryos (anterior pole and midventral sites). They demonstrated the potential of this cytoplasm for determining germ cell fate by transplanting the cells formed at the ectopic sites into genetically distinct host embryos. These ectopically formed germ cells could produce progeny when transplanted into host embryos, confirming that pole plasm is sufficient to induce germ cell fate (Illmensee and Mahowald 1974; Mahowald et al. 1976).

Electron microscope studies by Mahowald revealed the fine structure of polar granules in the germplasm of Drosophila embryos (also called germ granules later in this review), which were believed to be the functional substance that instructs the formation of pole cells. Germ granules are membraneless, electron-dense structures with 200–500 nm diameter (Mahowald 1962, 1968). These images suggested that germ granules were protein–RNA assemblies with polysomes on the surfaces, leading Mahowald to propose that mRNAs stored in germ granules are translated into proteins required for germ cell specification (Mahowald 1968).

A mechanistic understanding of how germ granules form and instruct germ cell specification required the identification of germ granule components and/or genes required for germ granule formation. As biochemical isolation of germ granules proved challenging (Waring et al. 1978), genetic screens were the main approach for discovering the genes required for germ granule formation and functions. Germ granules form during oogenesis and are required for the fertility of the next generation. Therefore, mutants that failed to form germ granules were expected to have grandchildless phenotypes, i.e. the progeny of mutant mothers would lack germ cells and thus fail to produce offspring. Initial screens for grandchildless mutants, however, had very limited success (Mahowald 2001). Major progress was made when screens for maternal-effect lethal mutants uncovered a group of mutants (termed posterior group mutants) that caused abdominal segmentation defects in embryos (Boswell and Mahowald 1985; Lehmann and Nusslein-Volhard 1986, 1991; Schupbach and Wieschaus 1989). It was discovered that for most of these mutants, embryos also lack germplasm and germ cells, suggesting a role of posterior group genes for germplasm assembly. Among these genes, oskar (osk) was shown to have a central and instructive role. Localization of osk mRNA to the posterior of developing oocytes precedes and is required for the localization of other posterior group RNAs and/or proteins (Ephrussi et al. 1991; Kim-Ha et al. 1991). Furthermore, ectopic expression of osk at the anterior pole is sufficient to assemble functional germplasm. This ectopic germplasm is sufficient to form functional pole cells at the anterior of the embryo, as well as a second abdomen in a mirror image of the “normal”, posterior abdomen (Ephrussi and Lehmann 1992). In the following section, we will review the molecular mechanism of germplasm assembly, focusing on osk mRNA regulation, Oskar protein functions, and the interplay between Oskar and other germ granule components. We will also review the mechanism of RNA localization to the germ granules and discuss how translation, specifically regulated by germ granules, underlies PGC specification. (see Table 1 for genes involved in the assembly of germ plasm and germ granule-localized transcripts).

Oskar mRNA regulation

Osk mRNA transport and localization

Given the instructive role of Oskar in germplasm assembly, the localization of osk mRNA to the posterior pole serves as the first crucial step in germplasm biogenesis during oogenesis (Ephrussi et al. 1991; Kim-Ha et al. 1991; Ephrussi and Lehmann 1992). Osk mRNA is first transcribed in the nurse cell nuclei of egg chambers before localizing to the posterior of oocytes (for a description of oogenesis, please refer to Fig. 1a). In situ hybridization showed that in early chambers, osk mRNA fills the oocytes, while in later stages, osk mRNA is concentrated at the oocyte's posterior pole (Ephrussi et al. 1991; Kim-Ha et al. 1991) (Fig. 1a, for staging oogenesis, please refer to Jia et al. 2016). Different osk alleles and posterior group mutants accumulate osk mRNA either within nurse cells, enriched at the anterior, or dispersed throughout the oocyte (Ephrussi et al. 1991; Kim-Ha et al. 1991), suggesting a stepwise process of osk RNA localization: synthesis in nurse cells, transport from nurse cells to the oocyte, transport within the oocyte from anterior to posterior, and sustained localization at the posterior pole (Fig. 1a–c). Here we summarize the molecular mechanism leading to this stepwise localization of osk RNA.

The transport of osk mRNA depends on microtubules (Theurkauf et al. 1993; Pokrywka and Stephenson 1994, 1995). Distinct types of motors transport the mRNA from the nurse cells into the oocyte and from the oocyte anterior to the oocyte posterior. Nurse cell-to-oocyte transport of osk mRNA is dependent on Dynein, Egalitarian (Egl), and Bicaudal D (BicD) (Clark et al. 2007; Bullock and Ish-Horowicz 2001). Egl binds RNA cargo, the Dynein Light chain, and BicD, which bridges RNA cargo/Egl with the dynein motor (Navarro et al. 2004; Dienstbier et al. 2009; McClintock et al. 2018; Goldman et al. 2019; Neiswender et al. 2021). Mapping the osk 3′UTR identified a region (nucleotide 499-759, called region 2) that is both necessary and sufficient to drive RNA from nurse cells to oocytes (Jambor et al. 2014). A 67-nucleotide stem–loop structure termed oocyte entry sequence (OES) within region 2 is essential for dynein-mediated transport of osk RNA (Jambor et al. 2014; Mohr et al. 2021). OES structurally resembles the transport/localization signal (TLS) of the fs(1)K10 mRNA, which directs the oocyte localization of fs(1)K10 (Serano and Cohen 1995). OES is potentially bound by Egl, which connects osk RNA to BicD and dynein for minus-end-directed transport from the nurse cells into the oocyte (Mohr et al. 2021). This mode of transport dominates the early stage of oogenesis (stages 1−7), resulting in a pronounced accumulation of osk mRNA in the oocyte and specifically at the anterior margin of oocytes by stage 8 (Ephrussi et al. 1991; Kim-Ha et al. 1991; Theurkauf et al. 1992; Jambor et al. 2014).

Moving osk mRNA to the posterior pole of the oocyte depends on kinesin-mediated, plus-end-directed transport, with kinesin I playing an important role (Clark et al. 1994; Brendza et al. 2000; Zimyanin et al. 2008). During mid-oogenesis (stages 8–10), the microtubule network in oocytes is reorganized and becomes enriched at the cortex, creating a slight posterior bias of the microtubule plus-end (Theurkauf et al. 1992; Clark et al. 1994; Zimyanin et al. 2008). This bias allows a net posteriorly directed transport of osk mRNA by kinesin. Curiously, this transport requires kinesin heavy chain (Khc), the force-generating component of the motor, but not kinesin light chain (Klc), which bridges the Khc with cargos, suggesting there may be a novel cargo loading mechanism for osk RNA transport (Palacios and St Johnston 2002). A genetic screen identified Tropomyosin I (Tm1) as being required for posterior localization of osk RNA (Erdelyi et al. 1995). It was later found that this Tm1 allele caused a defect in the plus-end-directed transport of osk mRNA, specifically due to a loss of Kinesin-1 activity (Zimyanin et al. 2008; Gaspar et al. 2017). Further genetic, biochemical, and structural dissection revealed that Khc directly binds RNA, and that the C/I isoforms of Tm1, an atypical RNA-binding tropomyosin, bind the RNA and Khc, forming a stabilized RNA-Khc-Tm1 tripartite complex, thus allowing kinesin-mediated transport (Veeranan-Karmegam et al. 2016; Gaspar et al. 2017; Dimitrova-Paternoga et al. 2021; Heber et al. 2024). Notably, while the Dynein-Egl-BicD complex is used to transport various nurse cell-synthesized transcripts to oocytes in addition to osk (Clark et al. 2007), the Khc-Tm1C/I complex appears to be specific for osk transport within oocytes (Veeranan-Karmegam et al. 2016).

Initial attempts to map the cis-elements that direct RNA localization using lacZ-osk hybrid transgenes suggested that the 3′UTR was necessary and sufficient for posterior localization in oocytes (Kim-Ha et al. 1993). Subsequently, it was found that the exon junction complex (EJC) components Barentsz, Y14, and Mago Nashi are required for osk mRNA localization, suggesting that RNA splicing is linked to localization (Newmark and Boswell 1994; Hachet and Ephrussi 2001; Mohr et al. 2001; van Eeden et al. 2001; Palacios et al. 2004). Indeed, removing the first intron of osk mRNA completely abolishes posterior localization in oocytes (Hachet and Ephrussi 2004). Further mapping identified 28 nucleotides flanking the splice site as the “spliced oskar localization element” (SOLE). SOLE, in association with the deposited EJC, directs osk RNA localization to the oocyte posterior (Ghosh et al. 2012). One possibility is that the SOLE RNA motif and the EJC are specifically recognized by the Khc-Tm1C/I transport machinery, once osk mRNA exits the nurse cell nuclei. However, how this specific interaction is established remains an open question. Furthermore, the identification of osk RNA null mutations (see “A noncoding function of osk RNA”) revealed that the previously observed localization of an intronless lacZ-osk3′UTR mRNA was accomplished by “hitchhiking” of the osk 3′UTR on the endogenous, spliced osk mRNA (Hachet and Ephrussi 2004; Jenny et al. 2006). This hitchhiking is mediated by palindromic sequences within the OES/SL2B element in the osk 3′UTR. These sequences allow osk to self-associate (“kiss”) in trans and promote higher-order RNA–protein condensates (Jambor et al. 2011, 2014; Bose et al. 2024).

The distinct transport mechanisms in nurse cells vs oocytes suggest a requirement for motor switching after entry into oocytes. However, both dynein and kinesin are loaded onto the osk RNPs shortly after their export from nurse cell nuclei, and both remain loaded within oocytes (Sanghavi et al. 2013; Gaspar et al. 2017), suggesting that kinesin-dependent transport is inhibited in the nurse cells and dynein-dependent transport needs to be inactivated in the oocytes. A recent biochemical and structural study revealed that Tm1 inhibits kinesin in nurse cells to allow dynein-mediated transport (Heber et al. 2024). How the inhibition by Tm1 is relieved in oocytes remains unknown; however, one candidate is the double-stranded RNA-binding protein Staufen (Stau), a component of osk RNPs in oocytes, that is required for osk posterior localization. Loss of Stau causes osk RNA to accumulate at the anterior of oocytes (Ephrussi et al. 1991; St Johnston et al. 1991; Micklem et al. 2000). Stau binds several stem–loop structures, termed Staufen recognized structures (SRSs), in the osk 3′UTR. Mutating SRSs strongly disrupted osk mRNA posterior localization just as in stau mutants (Laver et al. 2013; Mohr et al. 2021). Recent studies show that Stau competes with Egl for osk binding, and that oocyte-enriched Stau displaces Egl from osk mRNA, thereby inactivating osk association with dynein and allowing the switch to kinesin-mediated transport in oocytes (Mohr et al. 2021; Gaspar et al. 2023).

Posteriorly directed transport of osk RNA lasts until the end of stage 10 of oogenesis, when the microtubule network in oocytes reorganizes again to initiate fast kinesin-driven cytoplasmic streaming that efficiently mixes the cytoplasmic contents in the oocytes (Theurkauf et al. 1992; Serbus et al. 2005). Simultaneously, nurse cells dump their cellular contents into oocytes while undergoing programmed cell death. Nurse cell dumping and cytoplasmic streaming also contribute to osk localization and late-stage accumulation (Glotzer et al. 1997; Snee and Macdonald 2004; Sinsimer et al. 2011). However, separation-of-function kinesin alleles that are specifically defective in cytoplasmic streaming, but not plus-end-directed transport, showed largely normal osk RNA localization in late-stage oocytes, suggesting that cytoplasmic streaming is not strictly necessary for osk RNA localization (Serbus et al. 2005; Lu et al. 2018). Once osk RNPs reach the oocyte posterior, their maintenance requires Myosin V and the actin network (Krauss et al. 2009). The low microtubule density at the posterior allows Myosin V to out compete kinesin and anchor osk mRNA at the posterior cortex (Lu et al. 2020). Oskar protein (long isoform) is also required for maintaining the anchorage of osk RNA (Vanzo and Ephrussi 2002), by promoting the F-actin network organization through locally regulating the endocytosis pathway (Jankovics et al. 2001; Dollar et al. 2002; Vanzo et al. 2007; Tanaka and Nakamura 2008) (see “Osk mRNA generates 2 functionally distinct proteins: long osk and short osk”).

Translational control of osk RNA

While osk RNA is present at high abundance throughout the early stages of oogenesis, Oskar protein becomes detectable only after the RNA reaches the posterior pole (Kim-Ha et al. 1995), suggesting a translational repression mechanism before localization and/or translational activation after localization. The 80kD Bruno protein binds specifically to 3 distinct regions within the osk 3′UTR [Bruno response elements (BREs)]. Mutations in these binding sites abolish the translational repression of osk, cause precocious production of Oskar protein throughout the oocyte, and cause anterior patterning defects in embryos (Kim-Ha et al. 1995). Subsequently, it was found that lack of the protein Cup also caused precocious activation of osk translation before reaching the posterior (Wilhelm et al. 2003; Nakamura et al. 2004). Cup binding to eIF4E blocks the eIF4G-eIF4E interaction and thereby impedes cap-dependent translational initiation (Wilhelm et al. 2003; Nakamura et al. 2004; Kinkelin et al. 2012). It has been proposed that Bruno directly recruits Cup to osk RNA. Loss of the RNA-binding protein Bic-C also causes premature accumulation of Oskar protein (Saffman et al. 1998). Like its C. elegans homolog GLD-3, Bic-C likely represses translation by recruiting the CCR4-NOT deadenylation complex, which shortens the poly-A tail and impedes cap-dependent initiation (Chicoine et al. 2007). Targeting cap-dependent initiation, however, cannot fully explain the translational repression of osk, because other RNA-binding proteins, including ME31B, Hrp48, and PTB, are also required for osk translational repression but have not been implicated in suppressing cap-dependent translation initiation (Nakamura et al. 2001; Yano et al. 2004; Besse et al. 2009). These proteins, potentially together with Bruno, Cup, and Bic-C, may assemble stable silencing osk RNP particles, which repress translation by more globally limiting the access of the translation machinery (Nakamura et al. 2001; Chekulaeva et al. 2006; Besse et al. 2009; Bose et al. 2022).

Despite the identification of numerous proteins implicated in the translational repression of osk RNA during transport, a unifying and comprehensive mechanism for translational activation or derepression of osk RNA at the posterior of oocytes is still lacking (Wilson et al. 1996; Micklem et al. 2000; Castagnetti and Ephrussi 2003; Vazquez-Pianzola et al. 2011; Ryu and Macdonald 2015) (Fig. 1d). The repressor Bruno has been suggested to be inactivated by a yet unknown mechanism when reaching the posterior (Kim et al. 2015). In favor of an activation mechanism, deletion of one of the RNA-binding domains of Staufen abolished Oskar protein accumulation at the posterior without affecting osk RNA localization (Micklem et al. 2000). Furthermore, the RNA-binding protein Makorin 1 (Mkrn1) binds to the osk 3′UTR at a site partially overlapping with the BRE-C element in osk RNA and competes with Bruno for osk binding (Reveal et al. 2010; Dold et al. 2020). Loss of Mkrn1 abolishes Oskar protein accumulation at the posterior of oocytes without affecting osk mRNA localization, suggesting that Mkrn1 plays a role in activating osk translation by displacing Bruno from the 3′UTR (Dold et al. 2020). An A-rich region bound by poly-A-binding protein (PABP) near the Mkrn1 binding site stabilizes Mkrn1 binding to the osk 3′UTR, consistent with previous findings that PABP and the cytoplasmic PABP Orb are required for osk translation (Chang et al. 1999; Castagnetti and Ephrussi 2003; Vazquez-Pianzola et al. 2011; Dold et al. 2020). It remains unknown, however, how the translational derepression functions of Mkrn1 and Stau are specifically activated at the oocyte posterior.

One unique feature of osk RNPs at the posterior of oocytes, compared to the localizing osk RNP transport particles, is that they form submicron-sized granules (termed “founder granules”), each containing dozens to hundreds of osk mRNA molecules (Little et al. 2015; Bose et al. 2022; Eichler et al. 2023). These founder granules contain RNA-binding proteins like Staufen, Bruno, Hrp48, and PTB, a set of proteins also present in osk RNP transport particles. Specifically, condensation properties of these proteins at the posterior may affect their distribution and functionality, potentially allowing translation to occur. Indeed, it was found that founder granules have solid-like material properties. Artificially fusing the intrinsically disordered region (IDR) of the human FUS (FUsed in Sarcoma) gene to the osk RNP promoted a more liquid-like state and impeded the translation of osk mRNA molecules (Bose et al. 2022). Superresolution and single-molecule level characterization of these RNA-binding proteins and osk RNA in the founder granules, especially with alleles that are defective in osk translation (Micklem et al. 2000; Reveal et al. 2010; Dold et al. 2020), may provide more mechanistic insight into the translational regulation of osk. In “Translational regulation by germ granules,” we also discuss the translational regulation of nanos mRNA by germ granules in early embryos. Translational regulation by RNP granules or condensates has been an active field of research and has been reviewed comprehensively elsewhere (Sankaranarayanan and Weil 2020; Parker et al. 2022; Zhang et al. 2023).

A noncoding function of osk RNA

Osk RNA also fulfills a noncoding function, which was only revealed when RNA and protein null alleles were identified (Jenny et al. 2006). Osk RNA null females produce no eggs because oogenesis is arrested at stage 7. This noncoding function involves aspects of oogenesis that are not affected in Oskar protein null alleles. Based on experiments that mutated sequence elements in the osk 3′UTR RNA and altered the dosage of potential trans-acting factors, it has been proposed that osk RNA acts as a sponge for RNA-binding proteins. One identified factor is Bruno, where, in the absence of osk RNA, excessive free Bruno leads to a dominant oogenesis phenotype (Kanke et al. 2015). Other yet unknown osk RNA-binding factors are predicted to affect oocyte nucleus condensation and positioning of the microtubule organizing center (MTOC) during the early stages of oogenesis (Kenny et al. 2021). These noncoding effects are mediated by the osk 3′UTR (Jenny et al. 2006). They can be functionally uncoupled from the role of Oskar protein in germplasm assembly, as this function can occur ectopically at the anterior pole independent of the osk 3′UTR while under the control of the bicoid 3′UTR (Ephrussi and Lehmann 1992). RNAs with dual roles as protein-coding and noncoding (CNC) have been observed in several animal and plant species and are emerging as regulators and sensors in development and differentiation (Sampath and Ephrussi 2016). Indeed, a decoy function similar to the osk RNA-Bruno interaction has been assigned to the noncoding RNA NORAD in controlling the active pool of Pumilio protein, thereby protecting genome stability in mammalian cells (Lee et al. 2016).

Osk mRNA generates two functionally distinct proteins: Long Oskar and Short Oskar

At the posterior of oocytes, osk mRNA is translated into two protein isoforms, by using two different in-frame AUG codons (Markussen et al. 1995; Rongo et al. 1995). These two start codons produce a long isoform (Long Oskar, 606 amino acids) that has an additional N-terminal extension compared to the short isoform (Short Oskar, 467 amino acids). The two isoforms have different subcellular localizations and serve distinct functions (Fig. 2a and b). Short Oskar directly assembles into germ granules and is necessary and sufficient for germplasm assembly, embryo patterning, and germ cell formation (Markussen et al. 1995). The N-terminal extension in Long Oskar inhibits Short Oskar function by a yet unknown mechanism. Long Oskar is not a component of germ granules based on immunostaining but ensures the robustness of germplasm assembly. In Long Oskar mutants, osk RNA localization is not maintained in the late oocytes and embryos; the germ granules appear aggregated and dislodged from the posterior cortex, resulting in a significantly reduced number of pole cells in embryos and semisterility in the offspring (Vanzo and Ephrussi 2002). Mechanistic studies showed that Long Oskar is localized to the endocytic membranes, stimulates the endocytosis pathway at the posterior cortex, and induces long F-actin filaments protruding from the posterior cortex, which potentially anchor assembled germ granules and founder granules (Vanzo et al. 2007). Furthermore, Rabenosyn-5 (Rbsn-5), a Rab5 effector protein required for the early endocytic pathway, is crucial for germ plasm assembly. Genetic analysis placed Rbsn-5 downstream of Long Oskar, suggesting that endocytosis promotes actin filament protrusions (Tanaka and Nakamura 2008). A mechanistic link between the Long Oskar-induced endocytosis and actin remodeling was identified through a genetic screen: an endosomal protein Mon2, in response to the endocytic activation by Long Oskar, interacts with the actin filament remodeling complex (Spir/Capu/Rho1) to instruct the formation of long F-actin projections (Tanaka et al. 2011). Forming F-actin projections also requires the interaction between Long Oskar and Yolkless, the receptor of yolk protein on the endocytic vesicles (Tanaka et al. 2021). Long Oskar also plays a unique role in localizing mitochondria to the posterior pole of embryos, discussed in detail in “Mitochondria are enriched in PGCs.”

Composition and function of germ granules

Germ granule components and assembly

The assembly of germ granules occurs at the posterior of stage 8–14 oocytes, beginning with the synthesis of short Oskar protein. Short Oskar recruits binding partner protein Vasa, which in turn recruits Tudor. Tudor recruits Aubergine (Aub) and other proteins that contain symmetric dimethylated Arginine (sDMA) modifications. Such a hierarchical process was first suggested by genetic analysis and then elucidated at the biochemical and molecular level (see below for details). Meanwhile, mRNAs localize to germ granules through diffusion and trapping, mediated by protein–RNA interactions and RNA self-recruitment (see “RNA localization to germ granules”). A select list of germ granule proteins and RNAs are summarized in Table 1.

While the N-terminal extension of Long Oskar is mostly unstructured, Short Oskar has 2 clearly structured domains: an N-terminal LOTUS (Limkain, Oskar, and Tudor containing proteins 5 and 7) domain (residue 139-240) and a C-terminal SGNH-like domain (or OSK domain, residue 400-606, homologous to the SGNH hydrolases), which are connected by a largely disordered linker sequence (Jeske et al. 2015; Yang et al. 2015; Kistler et al. 2018) (Fig. 2c). These structural features allow Short Oskar to mediate multivalent intermolecular interactions, which underly the function of Short Oskar as an inducer or nucleator of germ granules (Kistler et al. 2018). The Oskar LOTUS domains dimerize constitutively in vitro (Jeske et al. 2015, 2017; Yang et al. 2015). The IDR can mediate weak and multivalent interactions, which drive the phase separation of IDR-containing proteins and the formation of biomolecular condensates (Choi et al. 2020; Kato et al. 2022). Indeed, Short Oskar expressed in tissue culture cells autonomously assembles into granular structures with a gel-like property similar to germ granules (Jeske et al. 2015; Kistler et al. 2018). Removing the LOTUS domain or linker sequence dampened granule formation in vitro (Kistler et al. 2018). The SGNH-like domain is a novel RNA-binding domain with little sequence specificity in vitro (Jeske et al. 2015; Yang et al. 2015), potentially allowing Short Oskar to recruit various mRNAs into germ granules. Almost all osk missense mutations isolated from genetic screens map to the SGNH-like domain, suggesting an essential role for RNA-binding in the assembly of germ granules (Lehmann and Nusslein-Volhard 1986; Kim-Ha et al. 1991). RNA also mediates multivalent intermolecular interactions with RNA-binding proteins, which may drive the assembly and possibly specificity of these RNP condensates (Van Treeck and Parker 2018). Interestingly, a nuclear localization signal (NLS) was found inside the linker sequence of Oskar (Kistler et al. 2018), allowing Short Oskar to form characteristic hollow-sphere nuclear bodies within pole cells described earlier by EM (Mahowald 1968). The function of the nuclear bodies, which contain Oskar and Vasa protein, remains unclear. Deleting the NLS eliminated nuclear bodies and moderately reduced pole cell number but did not affect germline development or fertility (Kistler et al. 2018).

Vasa was identified as one of the posterior group mutants through chemical mutagenesis and the use of a monoclonal antibody associated with polar granules (Schupbach and Wieschaus 1986; Hay et al. 1988b; Lehmann and Nusslein-Volhard 1991). The vasa gene was identified from an ovarian cDNA library using this antibody (Hay et al. 1988b) and, independently, through chromosomal walking using existing vasa alleles (Lasko and Ashburner 1988). Both studies identified Vasa as a germplasm component and a conserved DEAD-box RNA helicase with similarity to eIF4A. Later, Vasa was identified as a direct interactor of Short Oskar through the LOTUS domain (Lasko and Ashburner 1990; Breitwieser et al. 1996). Several oskar alleles that fail to form germplasm also fail to bind Vasa in vitro (Breitwieser et al. 1996). In vasa mutant ovaries, Short Oskar protein, but not Long Oskar, fails to accumulate (Rongo et al. 1995), suggesting that Vasa specifically stabilizes Short Oskar and enables its accumulation during germplasm assembly. Several regions of Short Oskar, most notably the N-terminal LOTUS domain, can bind Vasa (Breitwieser et al. 1996; Jeske et al. 2015, 2017; Kistler et al. 2018). A crystal structure of the Oskar LOTUS domain in association with the Vasa C-terminal domain has been solved. A mutation in Vasa that disrupts this interaction abolishes germplasm assembly (Jeske et al. 2017). The LOTUS domain stimulates the helicase activity of Vasa in vitro, but its functional significance in vivo is unclear (Jeske et al. 2017). Vasa mutants that lose the DEAD-box helicase activity also fail to localize posteriorly and do not maintain germplasm (Liang et al. 1994; Dehghani and Lasko 2015), suggesting that Vasa serves not just as a binding partner of Oskar, but rather recruits RNA using its helicase activity during germplasm assembly. In addition to germplasm and germ cell formation, and independent of its association with Oskar, Vasa has multiple roles during oogenesis. This includes oocyte polarity regulation activated by double-stranded DNA checkpoint control and transposable element silencing mediated by Vasa's role in nuage assembly (Tomancak et al. 1998; Ghabrial and Schupbach 1999; Abdu et al. 2002; Xiol et al. 2014; Dehghani and Lasko 2017) (see below).

Tudor is another germ granule component required for germ granule assembly (Boswell and Mahowald 1985; Bardsley et al. 1993). Embryos of tud null mutant females lack germ granules, and certain hypomorphic alleles make granules with reduced size or changed morphologies, suggesting a structural role of Tudor in germ granules (Boswell and Mahowald 1985; Thomson and Lasko 2004; Arkov et al. 2006). Tudor is a large protein of 2,515 amino acids with 11 copies of an “extended” Tudor domain, forming an aromatic cage that binds sDMA (Liu et al. 2010). Numerous germ granule proteins have been shown to contain sDMA modifications, catalyzed by the germplasm-localized methyltransferase Capsuleen (Csul) in concert with its associated protein Valois (Vls) (Cavey et al. 2005; Gonsalvez et al. 2006; Anne et al. 2007). Tudor can thus serve as a scaffold for recruiting DMA-modified proteins into germ granules. A well-characterized example is the Piwi-family protein Aub, which is symmetrically dimethylated at its N-terminus by Csul and bound specifically by domain 11 of Tudor (Kirino et al. 2009; Nishida et al. 2009; Liu et al. 2010). Tudor is required for Aub localization to germplasm. Mutating the sDMA sites of Aub abolishes Tudor binding in vitro and localization to germplasm in embryos (Nishida et al. 2009; Kirino et al. 2010; Liu et al. 2010; Webster et al. 2015; Huang et al. 2021). Piwi and Ago3, two additional Piwi-family proteins, might also be localized to germplasm through an sDMA-Tudor interaction (Kirino et al. 2009). Vasa also contains sDMAs that are deposited by Csul (Kirino et al. 2010), but Vasa localization to the germplasm does not depend on Tudor (Thomson and Lasko 2004). Conversely, Vasa is required for Tudor localization to the nuage in nurse cells (Anne 2010). Therefore, Tudor is believed to be recruited to germplasm through its interaction with Vasa. Alternatively, Tudor could also be recruited through direct interaction with Vls and Csul (Anne and Mechler 2005).

Aub, a PIWI-family protein responsible for piRNA biogenesis, is enriched in germ granules (Harris and Macdonald 2001). Loss of Aub leads to the activation of transposable elements, which causes DNA double-strand breaks and DNA damage checkpoint activation (Chen et al. 2007). Checkpoint activation disrupts the microtubule reorganization within the stage 8 oocyte, disabling the posterior localization of osk mRNA and accumulation of Oskar protein, thereby indirectly impacting germ granule assembly. Loss of the checkpoint kinase Chk2 can partially rescue Oskar accumulation (Klattenhoff et al. 2007; Dufourt et al. 2017). Aub may also play a more direct role in germplasm assembly, as it has been implicated in recruiting select maternally deposited mRNAs to germ granules (see “RNA localization to germ granules”). It has been proposed that germplasm-localized Aub, as well as Piwi and Ago3, allows maternally generated piRNA to be transferred to the germline of offspring and immunizes the germline against transposon activation (Brennecke et al. 2008).

Vasa, Tudor, and Aub are also components of the nuage, germline-specific perinuclear granules that are broadly associated with the silencing of transposable elements throughout the animal kingdom (Eddy 1975; Lim and Kai 2007; Lim et al. 2013). Like germ granules, the nuage is thought to be assembled by phase separation (Brangwynne et al. 2009). This results in the condensation and high concentration of some of the same components found in germ granules (Pamula and Lehmann 2024). However, the mode by which components associate and function differs between nuage and germ granules: nuage assembles along the nuclei of developing male and female germ cells and requires Vasa but not Oskar protein. Furthermore, nuage is associated with the silencing of transposable elements by piRNAs, small germline-specific RNAs (Aravin et al. 2007). Finally, while piRNAs are also localized to the germplasm to provide maternal immunity against transposable elements to the next generation (Brennecke et al. 2008), mRNAs associated with germplasm (discussed in “RNA localization to germ granules”) are not found in the nuage. For detailed reviews on piRNA function in silencing transposable elements and the role of the nuage, please refer to the following articles (Aravin et al. 2007; Lim et al. 2009; Pek et al. 2012; Hirakata and Siomi 2016).

RNA localization to germ granules

High-throughput in situ hybridization projects have identified more than 100 mRNAs localized to the germplasm of Drosophila embryos (Lecuyer et al. 2007; Tomancak et al. 2007). A few of them have been validated in vivo and shown to associate with germ granules (Little et al. 2015; Trcek et al. 2015, 2020). Nanos mRNA is the best-studied case. The localization of nanos and its local translation at the posterior pole are essential for the anterior–posterior patterning of the embryo. Together with the sequence-specific RNA-binding protein Pumilio, Nanos inhibits translation of the gap gene hunchback (hb) in the posterior of the embryo, thereby allowing the expression of posterior segmentation genes such as knirps (Hulskamp et al. 1989; Irish et al. 1989; Struhl et al. 1992). Loss of Nanos (or Pumilio) causes the abdominal phenotype of posterior group mutants (Wang and Lehmann 1991; Gavis and Lehmann 1992; Kobayashi et al. 1996). In pole cells, Nanos controls PGC specification, but not germ cell formation (discussed in detail below). This is a conserved role for Nanos throughout the animal kingdom (Kobayashi et al. 1996). The 3′UTR of nanos is necessary and sufficient to direct its posterior localization (Gavis and Lehmann 1992). Further deletion analysis of the nanos 3′UTR failed to identify a single consensus sequence, or a “zip code”, that is necessary or sufficient for localization. Rather, different regions within nanos 3′UTR act additively and redundantly to direct localization (Gavis et al. 1996). Other germplasm-localized mRNAs [e.g. germ cell-less (gcl), polar granule component (pgc), and Cyclin B (cycB)] have also been shown to be directed by their respective 3′UTRs (Dalby and Glover 1992; Jongens et al. 1992; Nakamura et al. 1996; Rangan et al. 2009). However, like nanos, a specific nucleotide-based signal sequence that can dictate their localization has not been identified in any of these mRNAs, so far.

Short Oskar is considered the core seed component of germ granules that recruits not only germ granule-associated proteins but also germplasm-localized mRNAs through its C-terminal SGNH-like RNA-binding domain (Jeske et al. 2015; Yang et al. 2015). Indeed, several germplasm-localized mRNAs were enriched in a Short Oskar pull-down from embryo lysate (Jeske et al. 2015). However, the specificity or selectivity of this domain for germplasm mRNAs has not been clearly established. Aub was also identified in a sensitized genetic screen to affect nanos localization (Becalska et al. 2011). Aub CLIP (cross-linking and immunoprecipitation) and iCLIP (individual-nucleotide resolution CLIP) experiments showed that Aub bound select maternal mRNAs through partial piRNA target pairing (Rouget et al. 2010; Barckmann et al. 2015; Vourekas et al. 2016). Further, posteriorly and germ cell-localized mRNAs were enriched among Aub targets (e.g. nanos, gcl, and pgc), suggesting Aub/piRNA binding can recruit select maternal mRNAs into germ granules (Barckmann et al. 2015; Vourekas et al. 2016). The enrichment of posteriorly and germ cell-localized mRNAs was largely lost in tud mutants, suggesting that Aub/piRNA binds to localized mRNA in germ granules (Vourekas et al. 2016). Posteriorly localized mRNAs are significantly longer than nonposterior localized mRNAs and so contain a higher number of piRNA target sites, which may contribute to their enrichment. Posterior mRNAs are also more abundant than other mRNAs (Vourekas et al. 2016). Thus, Aub/piRNA may function as a sticky trap in germ granules for generally longer and more abundant mRNAs. Consistent with this model, nanos localization is defective in aub mutant embryos. This defect is not due to the epistatic effect of checkpoint activation on Oskar synthesis in aub−/− oocytes, as Oskar accumulation can be rescued by Chk2 mutation or ectopic expression, while nanos localization remains defective (Dufourt et al. 2017). A recent genetic screen identified additional trans-acting factors that may contribute to RNA localization in germ granules (Curnutte et al. 2023). While it remains unclear whether these proteins act directly on RNA localization or more indirectly on germ plasm assembly, these results highlight the multitude of factors acting additively and redundantly to ensure proper granule assembly and mRNA localization.

Nanos mRNA, as well as most of the other germplasm-localized maternal mRNAs, is synthesized in nurse cells during oogenesis and, rather than being transported specifically as described for osk RNA, predominantly enters the oocyte during nurse cell dumping. The posterior localization of nanos coincides with cytoplasmic streaming in oocytes and continues until the last stage of oogenesis (stage 14) (Wang and Lehmann 1991; Forrest and Gavis 2003). By fluorescently labeling nanos mRNA in vivo using the MCP-MS2 system, the process of nanos localization in oocytes has been followed through live imaging (Forrest and Gavis 2003). Nanos mRNA reaches the posterior through cytoplasmic streaming and diffusion and becomes entrapped and anchored by germ granules (Forrest and Gavis 2003). Single-molecule RNA FISH (smFISH) and superresolution microscopy showed that unlocalized (or localizing) nanos mRNA (as well as gcl, pgc, and cycB) exists as RNP complexes containing single RNA molecules (Fig. 1e). These assemble into multicopy homotypic RNA clusters once they are localized to germ granules (Little et al. 2015; Trcek et al. 2015). The homotypic clusters of different mRNAs appear spatially separated, occupying distinct territories within a germ granule, while the protein components (Oskar, Vasa, Tudor, and Aub) appear more uniform within granules (Trcek et al. 2015, 2020; Niepielko et al. 2018). The cluster size, namely the RNA copy number of a cluster, is proportional to the abundance of the corresponding mRNA within an embryo and is a predictor of cluster position within the granule (Trcek et al. 2015, 2020). These observations suggest that the accumulation of mRNA in germ granules is mainly driven by self-association after an initial stochastic seeding event (Niepielko et al. 2018). The homotypic clustering in germ granules is similar to that of other RNP phase separation or condensation examples, such as in P bodies and stress granules observed in cells or in vitro, and may rely on similar underlying physical properties (Jain and Vale 2017; Langdon et al. 2018; Van Treeck and Parker 2018; Tauber et al. 2020a). Specifically, mRNA folding and multivalent interactions between mRNA molecules seem to play a role in the homotypic association of nanos mRNA (Tian et al. 2024). The functional significance of homotypic clustering in germ granules is still unclear since, for example, the relative position of a cluster within the granule does not seem to correlate with its translational activity.

It should be noted that posteriorly localized founder granules, the RNP granules containing osk RNA and Staufen protein, are distinct and separate from germ granules (Little et al. 2015; Trcek et al. 2015). While germ granules and localized mRNAs are actively incorporated into pole cells through microtubule-mediated transport (Lerit and Gavis 2011), osk RNA is degraded during pole cell formation by Decapping Protein 1 (DCP1) and Aub (Eichler et al. 2020). Forcing osk RNA into germ granules using the nanos 3′UTR is toxic to pole cells (Eichler et al. 2020), suggesting a functional purpose for this separation. Furthermore, superresolution imaging suggests that multiple osk mRNAs localize to individual founder granules but do not form homotypic clusters (Eichler et al. 2023). The formation of founder granules depends on RNA-binding proteins Bruno and Hrp48, which bind to osk mRNA and mediate phase separation through their prion-like domains (Bose et al. 2022, 2024). Thus, while RNP association governs osk localization, homotypic clustering may depend on the specific physical environment provided by germplasm proteins, allowing mRNAs to recognize each other.

As mentioned in “Germ granule components and assembly” and “RNA localization to germ granules,” phase separation plays an important role in condensate formation and RNA regulation, including germ granules. For more detailed and comprehensive information on phase separation and its cellular and developmental functions, we recommend these reviews (Sankaranarayanan and Weil 2020; Tauber et al. 2020b; Roden and Gladfelter 2021; So et al. 2021; Kato et al. 2022; Putnam et al. 2023; Pamula and Lehmann 2024).

Translational regulation by germ granules

Germ granules have been speculated to regulate translation since EM images showed the presence of polysomes on their surface (Mahowald 1962). The first functional proof of their role in translational control came from a study on nanos expression (Gavis and Lehmann 1994). Nanos protein is exclusively synthesized at the posterior pole of embryos to give rise to the Nanos gradient specifying abdominal segmentation (Lehmann and Nusslein-Volhard 1991). High concentrations of Nanos at the posterior pole are needed for PGC specification (Wang and Lehmann 1991; Kobayashi et al. 1996; Gavis et al. 2008). However, the amount of nanos mRNA that is localized to the posterior pole only accounts for a small fraction (∼4%) of the total nanos mRNA in embryos (Bergsten and Gavis 1999; Trcek et al. 2015). The unlocalized nanos mRNA is translationally repressed, while only the localized fraction is translationally activated (Fig. 2d). The posterior group mutants that fail to assemble germ granules in embryos can neither localize nor translationally activate nanos mRNA, suggesting that localization and translational activation of nanos mRNA are tightly coupled and might rely on a shared set of molecular machinery in germ granules (Gavis and Lehmann 1994).

Translational repression of the unlocalized nanos mRNA is mediated by the nanos 3′UTR. Screening fragments of nanos 3′UTR in vivo identified a conserved 100-nucleotide RNA segment that forms a stem–loop structure (Dahanukar and Wharton 1996; Gavis et al. 1996; Smibert et al. 1996). Using this sequence, Smibert et al. (1996, 1999) pulled down an RNA-binding protein Smaug from embryo lysates. Smaug is responsible for nanos translational repression by binding to the stem–loop structure [named Smaug response element (SRE)]. Dahanukar et al. (1999) identified the same protein using a yeast three-hybrid approach. The mechanism by which Smaug represses nanos translation has been studied through a series of biochemical studies and validation in vivo. Smaug binds to the SRE using its SAM domain and recruits the deadenylation complex CCR4-NOT to degrade the poly-A tail of the respective target mRNA (Aviv et al. 2003; Semotok et al. 2005; Zaessinger et al. 2006; Pekovic et al. 2023). Smaug also recruits Cup, which binds eIF4E and blocks eIF4G recruitment, impeding translational initiation (Nelson et al. 2004). Smaug forms a stable repression complex with target mRNAs and other RNA-binding proteins to ensure robust repression (Jeske et al. 2011; Gotze et al. 2017). In addition to translational repression, Smaug binding also leads to mRNA degradation in the soma, as part of the maternal-to-zygotic transition (Smibert et al. 1996; Semotok et al. 2005, 2008; Tadros et al. 2007; Chen et al. 2014).

How germ granule-localized nanos mRNA is protected from Smaug-mediated translational repression has been elusive. The challenge of uncoupling nanos localization from nanos translation derepression has prevented a molecular understanding of how nanos translation is achieved in germ granules. Mutations in posterior group genes abolish germ granule assembly and RNA localization, precluding the opportunity to study the translational control of localized mRNA separately. Oskar has been proposed to antagonize Smaug's function through direct interaction with Smaug. A yeast two-hybrid experiment first suggested that Oskar binds to the RNA-binding domain of Smaug, potentially blocking the Smaug RNA interaction (Dahanukar et al. 1999). Furthermore, overexpressing Oskar in embryos reduces nanos RNA in pull-down experiments with Smaug, suggesting Oskar indeed interferes with Smaug-nanos binding (Zaessinger et al. 2006). However, pull-down experiments using purified recombinant Oskar and the Smaug RNA-binding domain did not show any interaction (Kubikova et al. 2023). Instead, this study identified an interaction between the Oskar linker sequence with the N-terminal domain of Smaug (Kubikova et al. 2023), which may underlie the recruitment of Smaug into germ granules (Chen et al. 2024b; Siddiqui et al. 2024). Mutating the overrepresented glutamine and asparagine residues in the linker region of Oskar into glycine prevented the enrichment of Smaug in germ granules, potentially by breaking the interaction between the linker region and Smaug (Chen et al. 2024b). Interestingly, nanos translation was reduced significantly in these mutant germ granules, suggesting that the Oskar-Smaug interaction protects nanos from translational repression (Chen et al. 2024b).

Other germplasm-localized mRNAs, including gcl and pgc, are also specifically translated in the germplasm dependent on their respective 3′UTRs (Jongens et al. 1992; Hanyu-Nakamura et al. 2008; Rangan et al. 2009). Like nanos, the majority of gcl and pgc mRNA is distributed throughout the soma and translationally repressed, and this repression is overcome in the germplasm (Rangan et al. 2009; Trcek et al. 2020). However, gcl and pgc do not have SRE sequences. The elements that control the translational regulation of gcl and pgc in the embryo have not been mapped, and their putative repressors have not been identified. Not all the germplasm-localized mRNAs undergo translational derepression. CycB mRNA is localized to germ granules, but its translation is specifically repressed by Nanos and Pumilio (Asaoka-Taguchi et al. 1999; Kadyrova et al. 2007; Weidmann et al. 2016). In nanos or pumilio mutants, CycB is translated and causes excessive pole cell divisions (Asaoka-Taguchi et al. 1999; Kadyrova et al. 2007). Curiously, pgc mRNA contains a Pumilio binding site and is subject to repression by Nanos and Pumilio in the germarium (Flora et al. 2018), but is actively translated in germplasm where Nanos and Pumilio are concentrated (Hanyu-Nakamura et al. 2008). Thus, it remains unclear what determines whether an mRNA is protected from translational repression or not in the germplasm. One potential explanation could be the presence of a cofactor for a specific mRNA repressor in the germplasm. Furthermore, translation in germplasm appears temporally regulated. While nanos is translated as soon as the egg is activated, Pgc protein appears in pole cell nuclei in stage 4–5 embryos and disappears abruptly at stage 6, suggesting a translational repression mechanism is specifically active beginning at stage 6 in the germ cells (Hanyu-Nakamura et al. 2008; Rangan et al. 2009) (for embryogenesis staging, refer to Campos-Ortega and Hartenstein 1997). Germ granules go through morphological and compositional changes throughout development, which may direct temporal changes in translational regulation in the germplasm (Hakes and Gavis 2023). However, it is also possible that the observed differences in translational timing are due to the levels of protein production by the respective transcripts.

Localized translation in the germplasm suggests a direct role of germ granules in translational activation. This is in contrast to most of the well-studied RNP granules, which store translationally silent mRNA (Anderson and Kedersha 2008; Buchan and Parker 2009; Formicola et al. 2019; Aoki et al. 2021). Whether germ granules are the exact compartment for localized translation was unknown until recently. If so, how is translation achieved within a presumably condensed and seemingly unfavorable granule environment? Nanos has served as a paradigm to explore the mechanisms of this regulation by combining nanos mRNA visualization with SunTag, a technique that allows single-molecule imaging of translation in vivo (Wang et al. 2016; Wu et al. 2016; Yan et al. 2016). SunTag-nanos recapitulates the translational regulation of native nanos in embryos and directly demonstrates the association of translating polysomes with germ granules. However, polysomes are preferentially distributed on the surface of germ granules, with the nanos 3′UTR embedded inside the granules (Chen et al. 2024b) (Fig. 2d). A higher-order structure of germ granules organized into an inner core and outer shell region, visible through superresolution microscopy, may impact translational efficiency (Ramat et al. 2024). Only about 30%–50% of germ granule-localized mRNAs are translated (Chen et al. 2024b). Untranslated mRNAs tend to have the 5′UTR, the coding sequence, and the 3′UTR of the transcript embedded inside the germ granules, suggesting that circularization of mRNAs may favor translational repression. In support of this idea, the 5′ sequence of nanos appears to shift from inside germ granules in oocytes, to the germ granule surface as oocytes become activated for embryogenesis, suggesting that there is a developmental signal acting on germ granules to shift from translational dormancy to activity during the oocyte-to-embryo transition (Chen et al. 2024b; Ramat et al. 2024).

Function of mRNAs localized in germ granules

Some of the germplasm-localized mRNAs have been studied more extensively because they have specific functions important for germ cell formation, development, and specification (Fig. 3 and Table 1). Nanos encodes an RNA-binding protein and translational repressor that is thought to repress the translation of mRNAs that can interfere with germ cell specification, a conserved function it shares throughout animal germline development (Asaoka et al. 1998; Asaoka-Taguchi et al. 1999; Deshpande et al. 1999; Sato et al. 2007; Lee et al. 2017; Asaoka et al. 2019). Germ cell-less (gcl) encodes an ubiquitin ligase adaptor protein required for the formation of the pole cells, the PGC precursors, of the Drosophila embryo (Jongens et al. 1992; Pae et al. 2017). Polar granule component (pgc) encodes a short peptide that inhibits mRNA Polymerase II-dependent transcriptional elongation, thereby blocking new transcription in pole cells (Hanyu-Nakamura et al. 2008). Finally, Cyclin B is translationally repressed in pole cells by the activity of Nanos and Pumilio (Asaoka-Taguchi et al. 1999; Kadyrova et al. 2007). CycB promotes cell division of PGCs once they reach the embryonic gonad and the translational block is released (Asaoka-Taguchi et al. 1999).

Together, germplasm-enriched mRNAs and small RNAs establish a maternal program that dominates early germ cell development in the fly. Localized translation and protection from degradation allow these RNAs to contribute to the formation of the pole cells (gcl) that will give rise to PGCs, the specification of PGCs (nanos), the transcriptional program of PGCs (pgc), PGC cell cycle regulation (cycB), and protection from transposable elements (piRNAs). In the next sections, we will describe the events that assure selection and transgenerational inheritance of mitochondria, the specialized cellular processes that lead to the formation of pole cells that will establish the PGC pool, the distinct transcriptional program that is established in pole cells so they attain their fate as PGCs, and the process of pole cell migration from their site of origin to the embryonic gonad.

Mitochondria are enriched in PGCs

Mitochondria, the major regulators of energy and metabolism, are transmitted solely from the mother through the oocyte to the next generation. Mitochondria have their own genome (mtDNA). mtDNA replicates rapidly during oogenesis to provide the embryo with the mitochondria needed for development. From the vast oocyte pool, the pole cells inherit fewer than 0.1% of mitochondrial genomes (mtDNA) from the egg (Hurd et al. 2016).

The mitochondrial genome lacks the repair and recombination mechanisms safeguarding the nuclear genome. Thus, without some form of selection, mutations would accumulate in the mitochondrial genome of germ cells through generations and would inevitably be fatal for the continuity of the species. Research in Drosophila revealed several selection mechanisms acting specifically during germline development that depend on developmentally regulated mitophagy and selective replication of mtDNA (Hill et al. 2014; Ma et al. 2014; Lieber et al. 2019; Palozzi et al. 2022). Because multiple mtDNA molecules can coexist within a single mitochondrion, a mutation in one mtDNA molecule can be functionally complemented by another. Thus, selection is achieved by developmentally controlled fission of mitochondria, such that mitochondrial genomes are reduced to approximately one copy per mitochondrion, which can then be targeted for functional selection (Lieber et al. 2019; Chen et al. 2020). This selection process occurs early in oogenesis at the level of individual mitochondria. Consequently, the egg and resulting embryo carry a lower mutational burden than the mother. Together, these mechanisms ensure that functionally active mitochondria are passed on to the next generation.

Direct, live observation and high-resolution microscopy methods are beginning to reveal how mitochondria are faithfully transmitted to the oocyte and eventually to the next generation for transgenerational transmission. Whether these processes are linked to the active selection of functional mitochondria is less clear. The process of mitochondrial enrichment in the oocyte is tightly linked to establishing the polarity of the egg and the anteroposterior axis of the embryo during oogenesis (recently reviewed by St Johnston 2023). Briefly, during early oogenesis, asymmetric division of the germline stem cell results in a new stem cell and a differentiating daughter cell that will undergo four synchronous divisions, leading to a 16-cell cyst that is connected by intercellular bridges [reviewed in flybook chapter (Yamashita 2018) and briefly in Fig. 1a]. Only one cell within the cyst will become the oocyte, and the other cells will serve as nurse cells that feed the oocyte either through active, regulated transport from the nurse cells to the oocyte during early oogenesis (as described in “Oskar mRNA transport and localization” for osk RNA) or through dumping of nurse cell contents into the oocyte during later stages of oogenesis [reviewed in flybook chapter (Berg et al. 2024)]. The fusome, a structure composed of ER, actin, and actin-associated proteins, connects the 16 cells of the early egg chamber and orients the transport of cellular components into the developing oocyte (de Cuevas and Spradling 1998; Grieder et al. 2000). The fusome supports translocation of mitochondria into the oocyte and is enriched for actively replicating mitochondria (Cox and Spradling 2003, 2006; Hill et al. 2014). Subsequently, a noncentriolar MTOC, also referred to as “Balbiani body,” forms in the future oocyte and controls dynein-mediated, microtubule minus-end-directed transport from the nurse cells into the oocyte. The Balbiani body has also been implicated in the selection of mitochondria that will be incorporated into the pole cells (PGCs) (Cox and Spradling 2003, 2006). However, disruption of mitochondrial association with the Balbiani body only minimally affected mitochondrial selection in pole cells (Chen et al. 2020). Indeed, oocyte-directed transport during early oogenesis only indirectly contributes to germplasm formation as described in “Oskar mRNA transport and localization” (Hurd et al. 2016; see below). Instead, when mitochondria were imaged with a mitochondria matrix-targeted fluorescent protein (mito-EYFP), an enrichment of mitochondria at the posterior pole was observed only later, during mid-oogenesis, when the nurse cells empty their contents into the oocyte, and the entire oocyte cytoplasm is moved during cytoplasmic streaming (Hurd et al. 2016). Early studies had suggested that mitochondrial large and small ribosomal (mtlr and mtls) RNA but not mitochondria were associated with germ granules and had linked this enrichment with a germ plasm-specific translational role of mitochondrial ribosomes for germplasm function (Kobayashi et al. 1993; Amikura et al. 2001). However, mtlrRNA perfectly colocalizes with mito-EYFP but not with nanos mRNA, a germ granule marker, demonstrating that entire mitochondria rather than isolated mitochondrial rRNAs are concentrated at the posterior pole (Hurd et al. 2016). This concentration of mitochondria is consistent with increased respiratory activity at the posterior pole (Akiyama and Okada 1992).

Mitochondria and germplasm RNAs such as nanos are initially captured at the posterior pole of the oocyte during cytoplasmic streaming (described in “RNA localization to germ granules”) and incorporate into PGCs as they form in the embryo (Fig. 4b). Hurd et al. (2016) found depletion of Long Oskar specifically blocked mitochondria localization while depletion of Short Oskar or other germplasm components had no effect. Furthermore, mislocalization of either full-length Long Oskar or the unique N-terminal extension of Long Oskar alone enriches mitochondria at the anterior pole, while Short Oskar does not (Fig. 4a). These observations demonstrated that Long Oskar is necessary and sufficient to trap mitochondria. Mutations in the gene Tropomyosin II (TMII, also called Tm1 in flies) reduce mitochondrial localization placing TMII function downstream of Oskar for mitochondrial capture (Hurd et al. 2016). Without clear separation of function, it remains open whether enhanced endocytosis and F-actin projections downstream of TMII, discussed in “Osk mRNA generates two functionally distinct proteins: Long Oskar and Short Oskar” are also required for mitochondrial localization.

Long Oskar function leads to an increase of mtDNA from ∼700 mtDNA copies in wild-type pole cells compared to about ∼200 mtDNA copies in long osk mutant pole cells (Hurd et al. 2016). A reduction in ATP production and other mitochondrial functions in long osk mutants could have direct consequences for mitochondrial quality and survival. Nevertheless, pole cells with reduced mtDNA number, which form in embryos laid by long oskar mutant mothers, can give rise to fertile offspring. Thus, reduction of mtDNA copy number is not detrimental for reproduction. Strikingly, the number of PGCs is reduced in long oskar mutants (Vanzo and Ephrussi 2002). This may suggest that a reduced number of mtDNA copies impairs PGC formation. However, as described above, Long Oskar is also needed to anchor osk RNA and enhance the production of Short Oskar protein via localized endocytosis. Thus, reduced levels of Short Oskar may explain the decrease in pole cell number. Additional experiments, such as challenging mitochondrial function or multigenerational transmission under low copy number conditions, are needed to assess whether and to what extent mtDNA copy number variation in pole cells affects reproductive fitness and whether the enrichment of mitochondria in PGCs is linked to any mitochondrial quality control mechanism.

Formation of the pole cells, the precursors of the PGCs

Pole cells are the first cells to form from the syncytial nuclei of the early Drosophila embryo. Only nuclei that migrate to the posterior pole and its germplasm form pole cells. In the earliest descriptions of pole cell formation in 1923, Huettner described how nuclei “bulge[d] out posteriorly” before being “constricted off” (Huettner 1923). About 30–40 pole cells are made from the nuclei that reach the posterior pole, though there is significant natural variability in this number (Sonnenblick 1950). Named for their origin at the posterior pole of the embryo, pole cells develop into the PGCs of Drosophila, while all other nuclei will give rise to yolk nuclei or develop into somatic cells (see Table 2 for genes involved in germ cell formation).

Pole cells form earlier and independent of somatic cellularization

As described above, the maternally deposited germplasm contains everything the embryo requires to make PGCs. The presence of germplasm has a decisive role for the mode of cellularization that will ensue at the posterior pole as opposed to the rest of the early embryo. Upon fertilization, the embryonic nuclei divide synchronously without cell divisions, thereby forming a syncytium. Between nuclear cycles 8 and 9, the syncytial nuclei migrate to the embryo cortex. Irrespective of their final fate, cortical nuclei and their associated centrosomes reorganize the cytoskeletal machinery (Raff and Glover 1989). This results in a membrane bud that surrounds each nucleus with an actin-rich cap and a myosin-rich base (Warn et al. 1985; Foe et al. 2000). These buds have the potential to constrict, but Rho1 (the Drosophila RhoA homolog) activity is restrained in the soma by the Arf-GEF Steppke (Step), preventing early cellularization (Lee and Harris 2013; Lee et al. 2015). While nuclei in the future somatic region of the embryo continue their synchronous divisions at the cortex and will not form cells until the 14th nuclear cycle, those nuclei that reach the posterior pole are encapsulated by cell membranes and exit the cell cycle in G2 (Su et al. 1998) (Fig. 5a). Actin-rich buds at the posterior pole develop into “pole” cells, as early as nuclear cycle 9. This initial budding process does not require the synthesis of new zygotic transcripts. Indeed, nuclei were found dispensable for pole cell cellularization—as long as centrosomes migrated to the posterior pole, upon decoupling centrosomes from their nuclei (Raff and Glover 1989; Cinalli and Lehmann 2013; Lerit et al. 2017). In contrast, formation of the cells that will give rise to all the somatic tissues and organs of the embryo, larva, and adult relies on zygotic genome transcription. Specialized early zygotic transcripts orchestrate highly specialized steps to grow cell membranes between interphase nuclei (Foe and Alberts 1983; Tam and Harris 2024).

Direct, in vivo imaging revealed that two orthogonal constrictions give rise to two pole cells from one pole bud (Cinalli and Lehmann 2013). The first furrow, called the bud furrow, forms a contractile ring at the base of the bud, pinching it off from the rest of the embryo. This furrow is unusual as it constricts in a mitotic spindle-independent manner. The second, orthogonal furrow resembles a traditional anaphase furrow constriction as it divides one pole bud into two pole cells via microtubule-dependent cytokinesis (Fig. 5b). Both constrictions depend on Rho1 activation and the recruitment of Anillin, the formin Diaphanous, and Septins to both contractile rings (Castrillon and Wasserman 1994; Afshar et al. 2000; Field et al. 2005; Cinalli and Lehmann 2013). Consistent with a role for Rho activation, mutations in two RhoGEFs, Pebble and RhoGEF2, affect pole cell formation (Lehner 1992; Padash Barmchi et al. 2005).

Germ cell-less a direct regulator of germ cell formation

Germ Cell-Less (gcl), a germplasm-localized mRNA, encodes a protein (Gcl) that is important for pole cell formation but does not play a role in germplasm assembly. In embryos derived from gcl null mothers, pole cells fail to form but the rest of the embryo develops normally, resulting in a grandchildless phenotype (Jongens et al. 1992, 1994; Robertson et al. 1999). Gcl controls the constriction of the bud furrow, while the anaphase constriction proceeds normally. Conversely, the bud furrow is unaffected when the anaphase constriction is inhibited by injection of colcemid to depolymerize microtubules.

As a protein located at the nuclear envelope, Gcl was thought to play a role in transcriptional regulation. It has been argued that pole buds in gcl null embryos fail to form because they gain expression of somatic genes normally repressed in pole buds (Leatherman et al. 2002; Colonnetta et al. 2021). However, blocking transcription with the injection of α-amanitin did not rescue the gcl phenotype, suggesting that ectopic transcriptional activation is not responsible for the failure in cellularization (Cinalli and Lehmann 2013). Furthermore, misexpression of Gcl at the anterior pole, while altering the nuclear division cycles of somatic nuclei by a yet unknown mechanism, did not interfere with transcription (Cinalli and Lehmann 2013). There is also evidence suggesting that Gcl promotes centrosome segregation in pole buds, and that in gcl null embryos, germplasm is inadequately distributed, resulting in cellularization failure (Lerit et al. 2017). However, experiments described below suggest that these phenotypes are likely a consequence of the failure of pole cell formation rather than a primary effect due to loss of Gcl function.

Gcl encodes a C3-ubiquitin ligase adapter responsible for targeting the receptor tyrosine kinase Torso for degradation specifically at the posterior pole (Pae et al. 2017) (Fig. 5c). Torso overexpression has been shown to antagonize pole cell formation and its activity is normally repressed at the posterior pole (Martinho et al. 2004; de Las Heras et al. 2009). While Gcl is localized to the nuclear envelope during interphase, it transitions to the cell membrane during anaphase where it degrades Torso (Pae et al. 2017). The gcl null phenotype is rescued by disrupting the Torso pathway, suggesting Torso repression at the posterior pole is essential for pole cell formation. When the degron motif of Torso is mutated and replaced with alanines, preventing it from being targeted for degradation by Gcl, embryos that inherit one maternal copy of Torso-Deg exhibit a dominant PGC formation defect, while Torso's RTK function remains intact (Pae et al. 2017). This demonstrates that Torso is the only target for Gcl-mediated degradation. Steppke RNAi results in early cellularization in the soma as well as the germline, leading to ectopic anterior pole cells, as well as ectopic posterior pole cells that do not receive germplasm. In addition to rescuing the loss of PGCs in gcl^−/− embryos, Steppke RNAi was also able to rescue PGC formation in rho^1720/+ embryos, which have a defect in PGC formation (Lee et al. 2015). This suggests that Steppke suppresses early cellularization everywhere in the embryo by inhibiting Rho1 actomyosin activity through a GCL-independent manner.

Interestingly, RNAi against MEK and MAPK, the effectors of Torso responsible for activating its downstream transcriptional signaling cascade, did not rescue the gcl null phenotype, suggesting that Torso employs a transcription-independent pathway to disrupt pole cell formation (Pae et al. 2017). Gcl mislocalization to the anterior pole alone is insufficient to promote pole cell formation. Since pole cells can form in the absence of Gcl and Torso pathway activities, the existence of a second pathway that promotes cellularization independent of Gcl has been proposed. Indeed, the rescue of the gcl phenotype by knocking down Steppke, an ARF-GEF that acts as a repressor of Rho1 activity, suggests a more specific role for Rho1 in pole cell formation (Lee et al. 2015).

Thus, pole cell formation seems to rely on two parallel pathways, whose activities are restricted to the germplasm region (Fig. 5c). One pathway depends on the localization of gcl RNA to the germplasm. The local translation of Gcl at the posterior pole suppresses somatic signaling via the Torso pathway, allowing pole cell formation to occur. The other pathway depends on Rho activation and may be activated by centrosomes migrating into the germplasm region. How this second pathway acquires germplasm specificity is still unknown.

The PGC transcriptional program

While overexpression of Gcl at the posterior can lead to the formation of additional pole cells, these extra cells do not develop into functional PGCs. This is in contrast to overexpression of Oskar, the germplasm master regulator, which results in additional, functional PGCs (Ephrussi and Lehmann 1992; Smith et al. 1992). Thus, additional, germplasm-dependent factors besides precocious cellularization must support PGC fate. How does germplasm direct PGC specification? What factors are necessary and sufficient to activate the transcriptional program that will eventually lead to the specification of oocyte and sperm? While these questions remain to be fully answered, invaluable insights have been gained into one fundamental process that distinguishes germ cells from their somatic counterparts in the embryo: transcriptional repression.

In Drosophila, newly formed pole cells are transcriptionally quiescent, a process that may allow germ cells to ignore signals from nearby somatic cells in the dynamic embryo (Fig. 6). The first hint of this quiescence came from early work in the 1970s, when it was observed that pole cells were not labeled with the mRNA precursor [H3]-uridine, and hybridization experiments failed to detect poly(A) containing RNAs in the nuclei of pole cells (Lamb and Laird 1976; Zalokar 1976). Even the presence of a potent transcriptional activator, GAL4-VP16, was not sufficient to induce transcription in the newly formed germ cells of the embryo (Van Doren et al. 1998b). Using antibodies specific to Serine2 (Ser2) phosphorylation in the repeated C-terminal domain (CTD) of RNA polymerase II (RNAPII), Seydoux and Dunn (1997) discovered that both C. elegans and Drosophila early germ cells shared a common transcriptional phenotype—the absence of a phosphorylated Ser2 CTD. In eukaryotic cells, phosphorylation of Ser2 CTD is mediated by a highly conserved complex called the positive transcription elongation factor b (P-TEFb) (Aoi and Shilatifard 2023). This complex, formed by CDK9 and Cyclin T, has been shown to stimulate transcriptional elongation by phosphorylating the CTD of RNAPII. The absence of phosphorylated Ser2 CTD in Drosophila pole cells suggested that P-TEFb, and thus transcriptional elongation, may be blocked. This, along with the observation that ribosomal RNA production was active in the pole cells, suggested that the transcriptional repression may be due to a global repression of RNAPII in the newly formed germline (Seydoux and Dunn 1997).

After the early stages of broad transcriptional repression, pole cells undergo a transition from maternally provided to zygotically produced transcripts. This transition is temporally delayed compared to the soma (Siddiqui et al. 2012). Robust transcription in PGCs is only observed when germ cells reach the somatic gonad (stages 14 and 15 of embryogenesis). At this stage, sex-specific transcription in germ cells is controlled by the sex of the somatic cells encapsulating the germ cells and germ cell-autonomous factors, such as X chromosome dosage and the master regulator of sex determination, Sex lethal (Sxl) (Schüpbach 1982; Casper and Van Doren 2009; Li et al. 2021; Grmai et al. 2022) (see Table 3 for genes involved in germ cell transcription).

Inhibition of somatic mRNA transcription—a conserved mechanism for regulation of gene expression in PGCs

The search for a global inhibitor of transcriptional elongation hit a breakthrough in 2004 when Martinho et al., discovered that the gene polar granule component (pgc) regulated transcriptional repression in the Drosophila germline (Martinho et al. 2004). pgc mutants expressed somatic genes such as slam at high levels in pole cells, suggesting that pgc represses somatic transcriptional targets. Importantly, pole cells mutant for pgc contained phosphorylated Ser2 CTD significantly earlier than wild-type germ cells, connecting pgc to the global repression of Ser2 phosphorylation described by Seydoux and Dunn (1997).

Pgc was originally identified as a germplasm-localized noncoding RNA that is essential for germline development (Nakamura et al. 1996). Later it was shown that the pgc gene encodes a small open reading frame for a 71 amino acid polypeptide (Hanyu-Nakamura et al. 2008). Pgc protein is present in the germ cells of wild-type Drosophila embryos and is sufficient to repress CTD Ser2 phosphorylation in somatic cells. Nakamura and colleagues demonstrated that Pgc protein directly interacts with P-TEFb and prevents its recruitment to chromatin, thereby preventing RNAPII elongation. Yet, Pgc protein alone failed to inhibit the kinase activity of P-TEFb in vitro, suggesting that Pgc does not directly inhibit the enzymatic activity of P-TEFb, but rather acts by competing for the recruitment of P-TEFb to sites of RNAPII initiation (Hanyu-Nakamura et al. 2008). This discovery demonstrated that Pgc protein acts to repress global RNAPII transcriptional elongation by blocking P-TEFb from phosphorylating Ser2 on the CTD, and that this activity is critical to establishing the PGC program. Loss of Pgc protein allows for somatic gene expression in germ cells but also indirectly affects germ cell gene expression, likely through the ectopic expression of somatic regulators that degrade maternal RNAs such as miRNAs (Hanyu-Nakamura et al. 2019).

Pgc is not the only factor that acts to repress transcription in early germ cells (Fig. 6). Numerous other proteins are thought to play a role in transcriptional repression after Pgc protein has been degraded. Concomitant with Pgc-mediated transcriptional quiescence, there is a dramatic change in the chromatin architecture of pole cells (Schaner et al. 2003; Li et al. 2014; Schulz and Harrison 2019). When pole cells first form, their chromatin is already distinct from that of somatic cells. At their formation, pole cell nuclei lack the activating chromatin mark H3K4me due to the activity of the H3K4 demethylase Su(var)3-3 (Schaner et al. 2003; Rudolph et al. 2007). Loss of Su(var)3-3 results in premature accumulation of H3K4me in pole cells, suggesting that active removal of H3K4me is occurring during the same developmental period as pgc-mediated transcriptional repression. Interestingly, pgc mutant pole cells acquire H3K4me marks, suggesting a dual role of both Pgc and Su(var)3-3 in maintaining germline chromatin organization from an early stage (Martinho et al. 2004). Yet Su(var)3-3 is not the only chromatin remodeler that is essential for maintaining early transcriptional quiescence in the germline. Pole cell chromatin organization is also mediated in part by a member of the SWI/SNF chromatin remodeler complex called osa (Schaner et al. 2003; Martinho et al. 2004). Like pgc, depletion of osa results in promiscuous transcription (Martinho et al. 2004). Importantly, each of these chromatin remodelers likely plays an important role in repressing robust transcription even after Pgc protein disappears from pole cells, as only low levels of transcripts are detected in the germline even after Pgc protein is lost (Van Doren et al. 1998b; Martinho et al. 2004; Hanyu-Nakamura et al. 2008). Thus, the chromatin landscape must be tightly controlled for proper germ cell transcriptional programming, even in the absence of active RNAPII transcription. Despite this, little is known about how these chromatin changes directly affect germline transcriptional activation and specification.

Inhibition of translation by Nanos—a conserved mechanism for regulation of gene expression in PGCs

While Pgc inhibits RNAPII transcription at the earliest stages of germline development, once zygotic transcription is activated in PGCs, additional mechanisms are necessary to protect PGCs from activating somatic transcriptional programs. The evolutionary conserved RNA-binding protein Nanos is essential for this protection (Gilboa and Lehmann 2004; Wang and Lin 2004; Lehmann 2012). Nanos binds to target RNAs and promotes their degradation through deadenylation (Wreden et al. 1997; Chagnovich and Lehmann 2001; Cinalli et al. 2008). Nanos mutant embryos fail to give rise to functional germ cells, and loss of Nanos in the adult germline results in the rapid depletion of GSCs (Lehmann and Nusslein-Volhard 1991; Arrizabalaga and Lehmann 1999; Gilboa and Lehmann 2004; Wang and Lin 2004). While it is clear that Nanos is required for the maintenance and viability of germ cells, how Nanos actively promotes germ cell fate is unknown. Studies ranging from flies to frogs show that nanos mutant PGCs ectopically express somatic genes (Deshpande et al. 1999; Köprunner et al. 2001; Hayashi et al. 2004; Lai et al. 2012; Lee et al. 2017; Asaoka et al. 2019). Since nanos is only known as a translational repressor, the precociously transcribed genes are unlikely to be direct targets of Nanos. Instead, they are likely activated by regulators of somatic fate that are inappropriately translated in the absence of Nanos (Asaoka et al. 1998; Hayashi et al. 2004). Importantly, PGCs from nanos mutant embryos are unable to maintain their chromatin state, suggesting a role for Nanos in establishing proper germ cell chromatin organization (Schaner et al. 2003). How Nanos influences the PGC transcriptional landscape by repressing somatic programs is yet to be discovered.

A transcriptional program for PGC fate

So how might the germ cell program be specified? No “master regulator” transcription factor for germ cell fate has been identified in PGCs (Cinalli et al. 2008). Recent studies have turned to transcriptomics in hopes of identifying a potential germline-specific factor. In one such study, Li et al. (2021) used single-cell RNA sequencing (scRNA-seq) of Drosophila pole cells to identify genes that exhibit germline-specific expression. Very few germline-restricted candidate transcription factors were identified, and those that were did not strongly correlate with the activation of zygotically transcribed genes in the germline. The authors hypothesized that germline-specific transcription may not require germline-restricted transcription factors. Instead, Li et al. found several transcription factors that were enriched in the female germline compared to that of the male, yet none that were enriched in the male compared to the female, suggesting that the male germline program may be the default program early in development (Li et al. 2021). Nonetheless, a few putative germline-specific transcription factors were exciting candidates for further study. One germline-restricted factor of note was the putative DNA-binding protein HP6. HP6 was identified in scRNA-seq of PGCs from the embryo, the germline of the larval ovary, and the germline stem cells of the adult ovary (Rust et al. 2020; Slaidina et al. 2020, 2021; Li et al. 2021). However, despite the germline-restricted expression of HP6, subsequent studies found that it had minimal impact on germline development (Grill et al. 2023). While HP6 is not a key piece in the germline transcription puzzle, future studies that focus on additional candidate factors will be essential to our understanding of germline transcriptional activation.

If indeed germline-specific expression does not require activation from germline-restricted transcription factors, then known factors with roles in somatic specification may also play essential roles in PGC specification. One such candidate factor is the conserved zinc-finger transcription factor Ovo (OVO in mammals). Ovo is required for germline development and is a key regulator of sex determination in the female germline (Oliver et al. 1987, 1990). The ovo gene locus produces two separate germline-enriched isoforms, Ovo-A and Ovo-B, as well as a somatic-specific isoform Svb (Shavenbaby), which is essential for epidermal differentiation (Payre et al. 1999; Andrews et al. 2000; Salles et al. 2002). These isoforms share a common C-terminal zinc-finger DNA-binding domain (Andrews et al. 2000; Salles et al. 2002). While Ovo-A and Ovo-B both function in oogenesis, maternally deposited Ovo-B is the predominant isoform in pole cells (Andrews et al. 2000; Bielinska et al. 2005; Hayashi et al. 2017; Benner et al. 2024). Interestingly, Ovo-B is necessary for expression of nanos, vas, and piwi transcripts in the germline, as well as suppression of somatic transcripts in pole cells (Yatsu et al. 2008; Hayashi et al. 2017). However, suppression of Ovo-B function did not affect embryonic pole cell development, migration, or gonad coalescence, during which pole cells are activating their transcriptional program, but did lead to germ cell depletion during larval stages. Interestingly, maternally provided Ovo-A seems sufficient for germ cell survival in the ovary, while male germ cells rely on maternal and zygotically synthesized Ovo-B (Hayashi et al. 2017). Therefore, it is likely that additional, yet unknown, factors play a role in activating the earliest stages of the PGCs’ transcriptional program, while Ovo-B function may be necessary for sustaining germ cell-specific transcription patterns during postembryonic stages of germline development (Hayashi et al. 2017).

While the existence of PGC-specific transcription factors is still unclear, decades of work point to a clear role for transcriptional repression in the specification of PGCs. From global RNAPII inhibition to chromatin-based mechanisms that remodel the PGC genome, transcriptional repression is an essential step in establishing germ cell fate. One hypothesis is that the repression of somatic programs is sufficient to “allow” PGCs to activate their programming. This would support the notion that the “ground state” of cell identity is the primordial germline, and no instructive activation of germline-specific programs is necessary to promote germ cell specification. Yet inhibition alone cannot fully account for the conserved, robust transcriptional program that is activated in PGCs. Instead, germline-specific functions critical to early PGC development may require specific germline gene activation programs that have not yet been elucidated. These germline-specific functions include the regulation of PGC migration during midembryogenesis and the initiation of PGC differentiation during somatic gonad coalescence. Notably, these germline-specific functions are conserved across evolution, suggesting that a transcriptional activation program may also be conserved across species regardless of whether germ cells are formed through germplasm or inductive signaling (Extavour and Akam 2003). One can envision two ways by which PGC fate can be activated and maintained. The first is driven by “instructive cues” that trigger a gene activation program, such as a germline-specific transcription factor. The second would be a repressive mechanism, or “repressive mode,” whereby a germ cell-specific program is the mere outcome of the repression of somatic programs. Importantly, these mechanisms are not necessarily mutually exclusive. It is likely that a repressor program that actively and specifically represses the somatic programs in the germline intersects with an activation program that may not function exclusively in the germline. Together both of these programs are likely necessary to achieve and maintain germ cell fate. Future research focused on how the coordinated action of both repressive and activating programs instructs the PGC transcriptional landscape may be the key to truly understanding PGC fate.

PGC migration

Germ granules distinguish PGCs from all other embryonic cells. Sheila Counce and Anthony Mahowald capitalized on these distinctive structures to map out PGC locations over embryonic development using light and electron microscopy, respectively (Counce 1963; Mahowald 1968). This was not trivial in D. melanogaster; however, in some species, such as Drosophila willistoni or Drosophila virilis, polar granules were large and conspicuous enough to clearly identify PGCs, even when ensconced by other cells in the embryo interior. These early studies were remarkably accurate in tracing the developmental PGC migration path through the midgut and into the mesoderm. Subsequent studies using the first germ granule-specific antibody (later found to be targeting Vasa) confirmed these results and allowed the unambiguous tracking of PGCs from formation to colonization of the gonads (Hay et al. 1988a, 1988b). Vasa remains an excellent marker for PGCs in Drosophila and many other animals. In the following sections, we summarize distinctive steps in PGC migration and note essential genes in PGCs and somatic cells (Fig. 7). Many of these genes were identified using large-scale EMS and misexpression screens (Moore et al. 1998a; Starz-Gaiano et al. 2001; Kunwar et al. 2003). See Table 4 for genes involved in germ cell migration and a recent review for a comparative description of germ cell migration (Barton et al. 2024a).

Entry into the embryo

As described above, pole cells, the PGC precursors, are born as a cluster of cells at the embryo posterior before somatic cellularization occurs in the blastoderm (Starz-Gaiano and Lehmann 2001). PGCs sit on top of the posterior endoderm primordium, also called the posterior midgut. When gastrulation commences, the posterior midgut invaginates, enclosing PGCs and carrying them into the embryo interior. The posterior midgut invagination travels anteriorly along the dorsal embryo surface while also folding further into the embryo, coincident with germ-band extension. At stage 9 of embryogenesis, PGCs reside within the midgut pocket at approximately 50% embryo length (anterior to posterior) and 50% embryo depth (dorsal to ventral).

Early electron microscopy studies from Eric Wieschaus's laboratory noted that the morphological changes associated with posterior midgut invagination were delayed in the cells beneath PGCs, suggesting that PGC-posterior midgut adhesion resisted these changes (Sweeton et al. 1991). Indeed, in osk mutant embryos that lack PGCs, these same midgut cells altered their morphology in concert with neighboring cells. The cell–cell adhesion molecule mediating PGC-posterior midgut attachment was later found to be E-Cadherin, encoded by the shotgun locus (Kunwar et al. 2008; DeGennaro et al. 2011). A maternally provided peroxiredoxin, Jafrac1, maintains maternal E-cadherin levels, ensuring PGCs hold on to each other and the midgut tightly enough to be successfully swept into the embryo (DeGennaro et al. 2011). Other genes that affect PGC entry include the terminal class genes involved in the Torso pathway. Mutant embryos develop a characteristic “pole hole” phenotype, where blastoderm cells beneath PGCs do not form properly (Perrimon et al. 1985; Ambrosio et al. 1989; Warrior 1994) (as this gene class dramatically affects embryonic development, we did not include them in Table 2). Furthermore, loss of Tre1, a G protein coupled receptor (GPCR), Hedgehog, and Jak-Stat pathway components affect PGC ingression (Li et al. 2003; Kim et al. 2021). Tre1 and Hedgehog signaling have more defined roles in subsequent PGC migration steps and will be described in more detail below.

Transepithelial migration

Clustered PGCs disperse from each other and transmigrate as single cells through the posterior midgut from stages 9 through 10 of embryogenesis. This process requires directed PGC migration away from the cluster and morphological changes in the surrounding endoderm driven by an epithelial-to-mesenchymal transition (EMT). At this step, PGC migration is oriented autonomously by Tre1 (Kunwar et al. 2003, 2008; Kamps et al. 2010; LeBlanc and Lehmann 2017; Lin et al. 2020, 2022), an orphan GPCR. Tre1 functions as a prototypical guidance receptor in PGCs (Lin et al. 2020; Kim et al. 2021); it orients migration but is not necessary for migration itself. Other cellular processes regulated by Tre1 include asymmetric neuroblast cell division, immune cell transepithelial migration, and astrocyte morphogenesis (Yoshiura et al. 2012; Thuma et al. 2018; Chen et al. 2024a), suggesting a common link to cell polarity and cytoskeletal remodeling. In PGC migration, Tre1 signals through canonical heterotrimeric G proteins, including Gbeta13f and Ggamma1 (Kunwar et al. 2008). The G-alpha (Ga) counterpart in this complex is purported to be Gao in neuroblast division, but this awaits further validation in other biological contexts, including PGC migration (Yoshiura et al. 2012). Further downstream, Tre1 signaling regulates RhoGEF2, a Rho1-specific guanine exchange factor (GEF) involved in various aspects of PGC migration (Lin et al. 2022). In wild-type PGC clusters, RhoGEF2 is enriched in the cluster center at the rear of each PGC, enforcing front–back polarity. This localization is lost in Tre1 mutant PGCs, impairing cluster separation (Lin et al. 2022). Smoothened, the receptor for Hedgehog, and PI(4,5)P2 have also been implicated in mediating Tre1 signaling through localization of F-actin and PIP5 kinase in germ cells (Kim et al. 2021).

The redundant lipid phosphate phosphatases, Wunen and Wunen 2 (called Wunens hereafter), are also required for efficient PGC cluster dispersal (Renault et al. 2010). Wunens are integral membrane proteins containing an extracellular catalytic domain that dephosphorylates and inactivates phospholipids, including the well-known signaling molecules Spingosine-1-phosphate (S1P) and Lysophosphatidic acid (LPA) (Burnett and Howard 2003). Given the shared migration phenotype between Tre1 and Wunen loss, a tantalizing hypothesis is that PGC-expressed Wunens create a self-generated phospholipid signaling gradient recognized by Tre1. However, the ligand for Tre1 remains unknown, and Wunens regulate additional aspects of PGC migration beyond transepithelial migration, including PGC death (Starz-Gaiano et al. 2001; Hanyu-Nakamura et al. 2004; Renault et al. 2004).

Directionally migrating PGCs require space to move through the endoderm to fully separate from each other. This space is generated by a developmental endoderm EMT, requiring the endoderm transcription factors Huckebein (Hkb) and Serpent (Srp) (Jaglarz and Howard 1994; Warrior 1994; Seifert and Lehmann 2012). The nascent adherens junctions linking endoderm cells break down during EMT, allowing directionally migrating PGCs to squeeze through (Jaglarz and Howard 1995). Hkb and srp mutant endoderm maintain junctions, whereas ectopic junction breakdown is sufficient for PGC-midgut exit (Seifert and Lehmann 2012). FGF signaling affects the integrity of the midgut pocket and maintains proper adhesion levels in the midgut for PGC egress (Parés and Ricardo 2016).

Once PGCs have exited the midgut, they move toward the dorsal surface of the posterior midgut in a Wunen-dependent manner (Zhang et al. 1997). In addition to PGC expression, Wunens are expressed in somatic embryonic regions that PGCs avoid such as the ventral surface of the midgut and the nervous system, and thus establish permissive zones for these subsequent steps in PGC migration (Zhang et al. 1996, 1997; Starz-Gaiano et al. 2001; Renault et al. 2004). These Wunen-expressing somatic regions are responsible for guiding PGCs initially toward the mesoderm and splitting a single PGC cluster into 2 populations, each migrating toward the bilateral gonad. If the hypothesized connection between Wunens and Tre1 holds, Tre1 may continue functioning beyond transepithelial migration. However, PGCs that pass through the blastoderm prior to gastrulation migrate to the somatic gonad in the absence of Tre1 (Kunwar et al. 2003), favoring an involvement for Tre1 limited to the early stages of directed transepithelial migration.

Directed migration to mesoderm

From stages 10 to 11 of embryogenesis, PGCs transit from the posterior endoderm to the mesoderm, splitting into 2 groups moving in an anterior direction toward the dorsal–lateral regions of the embryo. Many PGCs use a fast-moving anteriorly directed intermediate tissue, the caudal visceral mesoderm (CVM), as a stepping stone to reach the lateral mesoderm, while later migrating PGCs can directly move from endoderm to mesoderm when these tissues are more closely apposed (Broihier et al. 1998; Stepanik et al. 2016). Within the mesoderm, PGCs are attracted to somatic gonadal precursors (SGPs). SGPs are specified in the dorsolateral mesoderm via the mutual action of the transcription factors Srp, which promotes fatbody development, and the homeotic gene, Abdominal A, which inhibits Srp within the fifth to seventh abdominal segments (corresponding to parasegments 10–12) (Boyle et al. 1997).

This complicated cellular choreography requires several components: (1) successful specification and maintenance of CVM and SGPs, (2) CVM migration, (3) secreted signals from the SGPs, and (4) intracellular signaling components in PGCs to decode these guidance cues. The genes necessary for SGP and CVM specification and CVM migration are listed in Table 2. Disrupting these genes often causes embryonic patterning defects, indirectly affecting PGC migration.

The first gene that specifically perturbed PGC migration to the mesoderm without affecting SGP development was hmgcr (Van Doren et al. 1998a), which encodes for 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR). HMGCR has a well-known role in producing mevalonate for cholesterol biosynthesis, but mevalonate is also an intermediate product for many other signaling molecules, including isoprenoids (Santos and Lehmann 2004). Flies are auxotrophic for cholesterol since they lack the enzymatic steps to synthesize it; thus, cholesterol modifications cannot be directly linked to HMGCR function (Deshpande et al. 2001; Kim et al. 2021). Ectopically expressing HMGCR is strikingly sufficient to attract PGCs into foreign environments, suggesting it functions as a rate-limiting enzyme in producing a potent PGC chemoattractant (Van Doren et al. 1998a).

Subsequent studies implicated intermediate enzymes involved in isoprenoid synthesis in the HMGCR-attractant pathway (Santos and Lehmann 2004) and defined a potential mechanism of attractant export through an ABC transporter (Ricardo and Lehmann 2009). More than 20 years after the discovery of HMGCR function in PGC migration (Van Doren et al. 1998a), at least one of the guiding isoprenoids was recently identified. Juvenile hormone (JH), an isoprenoid with well-described roles in metamorphosis and female adult reproduction (Yamanaka 2024), was found to attract migrating PGC in vivo and in vitro (Barton et al. 2024b). This embryonic role for JH as a paracrine signaling factor was surprising as JH was viewed as acting systemically on molting behavior and reproduction (De Velasco et al. 2004). The JH synthesizing enzyme, Jhmat (Juvenile hormone acid methyl transferase), is expressed in SGPs, and JH reporters are activated near Jhmat expressing cells. JH receptors Met/GCE activity restricts Jhmat expression in the soma, suggesting a negative feedback mechanism limiting JH hormone production in the embryo (Barton et al. 2024b). At least in vitro, PGC migration toward an artificial JH gradient does not require PGC transcription, questioning the role of the classic JH receptors Met/GCE in mediating this aspect of the migratory response. Other factors shown to influence PGC attraction to the mesoderm include Hedgehog and Wnt-mediated lipid signaling (Deshpande et al. 2001, 2023; McElwain et al. 2011). In general, cell type-specific manipulation and direct observation of signal responses are needed to distinguish roles in SGP specification and function from more direct effects on PGC migration. How PGC guidance is regulated within somatic tissues and how they are interpreted by PGCs will be an exciting area of future discovery.

Recent work has clarified how PGCs move and identified new intracellular signaling components required to direct PGC migration to the mesoderm. Unlike the canonical cell migration model where cells extend actin-filled protrusions, PGCs maintain a constant shape and use cell-scale retrograde cortical actin flow for motility (Lin et al. 2022). This cell motility mode has been referred to as fast amoeboid migration (Liu et al. 2015; Logue et al. 2015; Ruprecht et al. 2015) and is also observed in cancer cells (Tozluoğlu et al. 2013). Surprisingly, directed PGC migration requires the master energy sensor AMPK, which phosphorylates RhoGEF2, releasing it from inhibitory microtubule interactions. Free RhoGEF2 establishes regions of high actomyosin contractility to guide cortical flow and migration (Lin et al. 2022). Future studies will need to clarify how these molecules are connected to guidance receptor signaling. Other intracellular signaling components implicated in PGC guidance and F-actin polarization include WASP, dPIP5K, and dWIP (Kim et al. 2021).

Gonad coalescence

After PGCs reach the SGPs, which are set aside from the mesoderm in 3 clusters within abdominal segments 5–7, both cell populations move anteriorly during germ-band retraction and coalesce into a compact gonad from stages 12–14 of embryogenesis (Boyle and DiNardo 1995; Moore et al. 1998a, 1998b; Riechmann et al. 1998). Gonad compaction is driven by somatic cells and occurs even when PGCs are not present (Brookman et al. 1992). Known genes required for gonad coalescence include ena (Sano et al. 2012), foi, and shotgun (Van Doren et al. 2003). Somatic cells tightly ensheath the germ cells and their interaction and cross-signaling with PGCs is critical for the development of the gonads into ovaries or testis (Gilboa and Lehmann 2006).

Regulated germ cell death

While pole cells are destined to become PGCs and have not been reported to contribute to any other cell type, about a third of them will die, likely due to their inability to fully activate the PGC program. Lost PGCs die during their migration (Starz-Gaiano and Lehmann 2001; Coffman et al. 2002; Renault et al. 2004). PGC death is attributed largely to a Wunen-mediated alternative death pathway that is partially offset by the tumor suppressor P53 (Starz-Gaiano et al. 2001; Starz-Gaiano and Lehmann 2001; Renault et al. 2004; Yamada et al. 2008; Slaidina and Lehmann 2017). Wunen acts protectively in germ cells, such that PGCs that form in the center of the germplasm or have artificially increased levels of Wunen have a higher probability to reach the gonad, while those at the periphery are more likely to die (Slaidina and Lehmann 2017). This PGC-intrinsic cell death is independent of classic apoptosis and is activated by DNaseII release from lysosomes (Tarayrah-Ibraheim et al. 2021). DNaseII released from lysosomes translocates to the PGC nuclei to induce double-strand breaks, activating the ATR DNA damage response pathway which ultimately results in PGC death. Wunens may regulate PGC survival through lipid metabolism, presumably through the uptake of the dephosphorylated lipid product (Renault et al. 2004; Tarayrah-Ibraheim et al. 2021). One hallmark of Drosophila PGCs is their lack of lipid droplets (Allis et al. 1979). Thus, PGCs may be uniquely dependent on environmental lipid uptake for their survival.

Summary

Formation, specification, and migration of PGCs are essential prerequisite developmental steps to generate and protect the next generation. Genetic analysis and cell biological studies have shaped a picture of the intricate and changing interactions between germ cells and the soma. Initially, at the specification stage, it seems essential that germ cells are different from somatic cells and somatic differentiation is prevented at multiple levels—pole cells form hours before somatic cells, and somatic transcription and signaling through soma-specifying signaling pathways are suppressed. During migration, germ cells depend on somatic cells as they are carried along by changing associations with somatic cell populations until they reach the somatic gonad. The cells of the somatic gonad and the germ cells codevelop, where the soma is critical in shielding the germ cells from premature differentiation and fostering their differentiation into egg and sperm. As this review highlights, while our understanding of some aspects of early germ cell biology is quite advanced, many gaps in our understanding remain and will require new methods to observe this dynamic process live and manipulate PGCs and their environment precisely in space and time.

Acknowledgments

We apologize to authors whose work we could not cite or discuss in detail. We thank members of the Lehmann Lab past and present for discussion and Melissa Pamula for assistance in generating figures. Drosophila research depends on the community-shared use of Drosophila stocks and their maintenance and distribution by the Bloomington Drosophila Stock Center (NIH P40OD018537) and on the genetic and genomic knowledge curated in FlyBase (Öztürk-Çolak et al. 2024).

Funding

Work in the Lehmann lab on germline development has been funded by the National Institute of Child Health and Human Development with grant R01/R37 HD41900 and presently by R01HD110546 to RL, and by investigator support to RL from the Howard Hughes Medical Institute. SG is supported by the American Cancer Society Postdoctoral Fellowship PF-21-116-01-RMC. BL is supported by the National Institute of General Medical Sciences of the National Institutes of Health under award number 1R35GM157022.

Data availability

The data underlying this article are found in the article.

Literature cited

Author notes