Coordination of cellular differentiation, polarity, mitosis and meiosis – new findings from early vertebrate oogenesis

Abstract

A mechanistic dissection of early oocyte differentiation in vertebrates is key to advancing our knowledge of germline development, reproductive biology, the regulation of meiosis, and all of their associated disorders. Recent advances in the field include breakthroughs in the identification of germline stem cells in Medaka, in the cellular architecture of the germline cyst in mice, in a mechanistic dissection of chromosomal pairing and bouquet formation in meiosis in mice, in tracing oocyte symmetry breaking to the chromosomal bouquet of meiosis in zebrafish, and in the biology of the Balbiani body, a universal oocyte granule. Many of the major events in early oogenesis are universally conserved, and some are co-opted for species-specific needs. The chromosomal events of meiosis are of tremendous consequence to gamete formation and have been extensively studied. New light is now being shed on other aspects of early oocyte differentiation, which were traditionally considered outside the scope of meiosis, and their coordination with meiotic events. The emerging theme is of meiosis as a common groundwork for coordinating multifaceted processes of oocyte differentiation. In an accompanying manuscript we describe methods that allowed for investigations in the zebrafish ovary to contribute to these breakthroughs. Here, we review these advances mostly from the zebrafish and mouse. We discuss oogenesis concepts across established model organisms, and construct an inclusive paradigm for early oocyte differentiation in vertebrates.

1. Introduction

Germ cells undergo fascinating processes during their development. Unlike many other tissues where cells perform a collective tissue function, germ cells differentiate in the gonad to perform an individual function. In the case of the oocyte, this single cell contains the building blocks that initiate the early events of embryonic development following fertilization. The oocyte undergoes a dramatic differentiation process (Fig. 1) that begins when germline stem cells divide, giving rise to oogonia, a mitotic precursor cell of the differentiating meiotic oocyte. Oogonia divide incompletely to generate early oocytes that are inter-connected to sister oocytes via cytoplasmic bridges in a germline cyst. Meiosis initiation transforms the oogonia in the cyst into differentiating oocytes that then separate and become surrounded by somatic granulosa cells in the follicle. In the early differentiating oocyte dramatic nuclear events underlie meiosis, while the cytoplasm in most species becomes polarized. Notably, these intracellular processes occur simultaneously with the changes in cellular organization, all while the oocyte significantly grows in volume.

Figure 1. Early oogenesis in vertebrates.

Major events in early oogenesis from germline stem cells through the formation of the primary follicle are depicted. Germline stem cells give rise to the mitotic oogonial cells. Upon the induction of meiosis, oocytes undergo the stages of prophase I: leptotene, zygotene, pachytene and diplotene. Oocytes arrest at diplotene stage (dyctate) until meiosis resumes during oocyte maturation later in oogenesis. The top row shows the cellular organization of oogonial cells, oocytes and somatic follicle cells at each stage. For the oogonia stage, a germline cyst is on the right and a presumptive cell division pattern in the cyst with cytoplasmic bridges depicted is on the left. This division pattern is based on the Drosophila model for germline cyst construction, and remains to be determined in vertebrates. On the left, five major themes in early oogenesis are depicted: mitosis, meiosis, cellular organization, apoptosis, and the Balbiani body mRNP granule. The blue bars to the right of each major process indicate their timing in the oogenesis pipeline. Within the bars, primary events of each process are depicted at each stage. Indigo bars indicate processes that occur similarly in zebrafish and mouse oocytes. Species-specific information is indicated by a dark aqua-blue bar for zebrafish, and a light aqua-blue bar for mouse. Known regulators are indicated in red (CDK2, RingoA and Tex14 were analyzed in mouse oocytes, Bucky ball in zebrafish oocytes. Retinoic acid is a known regulator in mouse and strongly implicated in zebrafish (Rodriguez-Mari et al., 2013)). Telomeres (red) indicate the bouquet configuration at zygotene, and Bb components (green) depict Bb formation in zygotene through diplotene.. The proposed scheme provides a conserved unified model for vertebrate oogenesis.

The Drosophila model has provided a major paradigm of early oogenesis in the field. Because of technical challenges in investigating early oogenesis in vertebrates, several fundamental questions have been long standing. However, in recent years many of these challenges have been overcome and an unprecedented view of early vertebrate oogenesis has immerged. We will discuss recent discoveries on the formation and structure of the germline cyst in the mouse and the fish Medaka, the dissection of chromosomal movement mechanisms during chromosome pairing in meiosis from the mouse, and mechanisms of oocyte symmetry breaking from zebrafish that are linked to meiotic chromosome pairing. We also discuss recent breakthroughs in zebrafish, Xenopus, and the mouse in understanding the biology of the Balbiani body, a universal oocyte feature, comparing common themes in its formation and potential species-specific functions. We further discuss the coordination of these multiple aspects of early oocyte differentiation.

We focus on early oogenesis from germline cyst formation through diplotene stages of the early follicle. Emphasis is on zebrafish oogenesis while comparing common themes with other species, and we mostly use the zebrafish oogenesis staging nomenclature. In an accompanying paper, we provide the methods established for investigating the zebrafish ovary that enabled some of the recent zebrafish contributions discussed here.

2. Organization of oogonia and early meiotic oocytes in cysts and nests

2.1 Germline cyst formation and structure

The building blocks of the germline cyst are oogonia, which are mitotically active cells that are born from germline stem cells (GSCs) and give rise to oocytes (Fig. 1). GSCs have recently been identified in the Medaka ovary through lineage tracing studies. It was found that the GSCs localize to what has been called a ‘germinal cradle’ (Nakamura et al., 2010). These GSCs specifically express the germline marker nanos2 (Nakamura et al., 2010). Several lines of evidence also indicate that nanos2 marks the GSCs in the zebrafish. These cells reside in a lateral, anterior-posterior band at the ovary periphery termed the germinal zone (Beer & Draper, 2013). The GSCs in Medaka also reside at the ovary surface but are dispersed along thread-like cords throughout the dorsal surface (Nakamura et al., 2010). Loss of the nanos2-expressing cell population in juvenile zebrafish is associated with loss of oocytes and sterility in the adult (Beer & Draper, 2013; Dranow et al., 2013), consistent with nanos2 marking the GSCs. Future studies are needed to characterize the GSC niche that regulates the production of oogonia from these cells.

Oogonia divide several times in mouse (Lei & Spradling, 2013), frogs (Kloc et al., 2004), Medaka (Nakamura et al., 2010) and zebrafish (Leu & Draper, 2010) (Fig. 1- mitosis), like in Drosophila (Matova & Cooley, 2001; Xie, 2013). During these divisions cytokinesis is incomplete and in the absence of cellular abscission, oogonial cells remain connected via cytoplasmic bridges (CBs) and form a germline cyst (Greenbaum et al., 2007; Kloc et al., 2004; Marlow & Mullins, 2008a) (Fig. 1- cellular organization). The germline cyst is engulfed by somatic follicle cells (Elkouby et al., 2016; Leu & Draper, 2010; Nakamura et al., 2010; Pepling, 2012; Selman et al., 1993), and the organization of oogonia and oocytes within germline cysts is highly conserved (Pepling et al., 1999).

The current prevailing model for the oogonial cell division pattern that generates the cyst is the one known from Drosophila. In this model, the cells in the cyst, called cystoblasts, divide 4 times synchronously, giving rise to a 16-cell cyst (Greenbaum et al., 2007; Xie, 2013). In Drosophila, such a pattern is evident by the specialized fusome structure that persists between daughter cells and traces their division planes (Greenbaum et al., 2007; Xie, 2013). A fusome does not form in vertebrates and the pattern and number of divisions of each cell within the cyst has not been addressed. But if the cells develop synchronously, then 2ⁿ cells are expected, where n is the number of cell divisions. Cysts in Xenopus contain up to 16 cells (Kloc et al., 2004), whereas in the Medaka fish and in the mouse, cysts have been identified that are up to 30 or 32 cells, suggesting that an additional round of cell division can occur in the cyst (Lei & Spradling, 2013). Detection of midbodies, a structure of the CB, in clonal meiotic mouse cysts at E14.5 shows that most cells have one or two midbodies and few (~14%) have three or more, but a consistent division pattern or cyst morphology could not be deduced (Lei & Spradling, 2016). Interestingly, the partial breakdown of cysts and their non-clonal aggregation into nests in the mouse demonstrate a cyst-nest aggregation mechanism that is not solely dependent on cell division (Lei & Spradling, 2013), but it is not known if such a mechanism exists in other species.

A prominent feature of the germline cyst is the synchronous development of sister oocytes within it. The intercellular CB connections are thought to facilitate this synchrony through shared cytoplasmic regulators (Pepling et al., 1999). Interestingly, in the re-aggregated nests of clonally unrelated cysts in the mouse, each partial cyst develops synchronously, but not in conjunction with other partial cysts in a nest (Lei & Spradling, 2013). This clonal-specific synchronization supports a CB-mediated synchronization mechanism. CBs assemble on midbodies and require the Tex14 protein (Greenbaum et al., 2009; Greenbaum et al., 2007). Tex14 is required for the construction of the male spermatocyte cyst in the mouse since tex14^−/− spermatocytes lack CBs, and mutant males are sterile (Greenbaum et al., 2009; Greenbaum et al., 2007; Greenbaum et al., 2006). Tex14 positive midbodies were also detected in cysts at E14.5 through E17.5 in female mice (Greenbaum et al., 2009; Lei & Spradling, 2016). However, tex14^−/− oocytes lack CBs, but complete oogenesis normally (Greenbaum et al., 2009). While early meiotic stages were not directly addressed, tex14^−/− females produced developing follicles and were fertile (Greenbaum et al., 2009). This surprisingly demonstrates that CBs, the essence of the cyst structure, are dispensable for oogenesis in the mouse.

A direct examination of the roles of the germline cyst in early oogenesis in other species would be insightful. For example, the cyst structure was linked to oocyte polarization in zebrafish (Elkouby et al., 2016) (discussed below). This testable function together with technical advances in the study of oogenesis in zebrafish, such as ovarian culture and live time-lapse imaging analysis of germline cysts [see (Elkouby & Mullins, 2017) in this issue], provide an excellent experimental setting to directly examine potential cyst functions.

In Drosophila the cyst always contains 16 cells following the last mitosis, but only one of these cells differentiates as an oocyte and the remaining 15 become nurse cells. In Drosophila the nurse cells produce the vast majority of factors for the oocyte, delivering them through the persisting ring canals (CBs), while the oocyte is largely transcriptionally quiescent. The cell that will become the oocyte is always one of the two cells derived from the first oogonial division and thus has four ring canals (Fig 1) (Huynh & Johnston, 2004). At these cyst stages in Drosophila, the specialized fusome structure forms in each cell following a cell division, extending through the ring canals between sister cells. The first oogonial cell forming the cyst inherits fusome from its GSC precursor and so contains more fusome than other cyst cells, which is thought to be important to specifying it as the future oocyte (Huynh & Johnston, 2004). Determination of the oocyte versus nurse cell fate is postulated to result from a biased enrichment of inherited components to the fusome and further requires polarized microtubules that direct factors and organelles to the prospective oocyte (Huynh & Johnston, 2004). The later polarity of the oocyte also correlates with the clustered positioning of the ring canals and the fusome of the pro-oocyte, marking the anterior of the prospective oocyte, suggesting that these polarized structures in some manner preset oocyte polarity (Huynh & Johnston, 2004).

In vertebrates the oocyte is transcriptionally active and the Drosophila oocyte-nurse cell relationship does not exist, although there is an intriguing recent report in the mouse that cytoplasmic content from one oocyte may be transferred to another, followed by apoptosis of the former (Lei & Spradling, 2016). It was proposed that similar to Drosophila, mouse oocytes with 3 to 4 CBs (and not two) are selected for further differentiation based on the increase in the percent of such oocytes from ~14% at E14.5 to ~20% at E17.5, during which oogonial cell division has largely ceased (Lei & Spradling, 2016). However, as discussed above, tex14^−/− mouse oocytes lack CBs, yet undergo normal folliculogenesis and tex14^−/− females are fertile (Greenbaum et al., 2009; Greenbaum et al., 2007). This suggests that transfer of cytoplasmic contents between oocytes is not essential to oocyte development or occurs during the very transient period of CB formation following oogonial cell division.

The transfer of cytoplasmic contents between cyst oocytes is proposed to lead to the growth of one oocyte and subsequent death of the other. Indeed, oocytes in mammals face a cell fate decision between apoptosis and follicle formation, and in parallel with primordial follicle formation, the majority of oocytes are eliminated by apoptosis (Tingen et al., 2009). Oocyte apoptosis involves Bcl family proteins (Kim & Tilly, 2004), and likely depends on a balance between Bcl-x, a cell survival factor, and Bax, a cell death factor (Kim & Tilly, 2004; Rucker et al., 2000). Bax is required for oocyte apoptosis during a cell fate decision between follicle growth and atresia (loss of oocytes) at a later folliculogenesis stage in adult mammals (Tingen et al., 2009). Adult bax^−/− female mice showed more follicles and less atresia throughout adult life, leading to a longer functional longevity of the ovary (Perez et al., 1999). Surprisingly, at an earlier P4 stage bax^−/− ovaries have a normal number of primordial and primary follicles (Perez et al., 1999), suggesting that oocytes apoptose prior to P4 in a Bax-independent manner or normalize their number in some manner. Evidence for Bcl-x/Bax pathway function at earlier oogenesis stages comes from a hypomorphic mutation in bcl-x (bcl-x fl^neo/fl^neo) that leads to increased oocyte apoptosis in ovaries at E14.5-E15.5 as detected by histological analysis, as well as a dramatic loss of follicles in P1 ovaries (Rucker et al., 2000).

The investigations of apoptotic mutants in mouse oogenesis largely focused on later folliculogenesis (bcl-X/bax) and meiotic checkpoints (p53, p63) (Bolcun-Filas et al., 2014; Morita et al., 2001). To directly test if apoptosis is required for follicle formation through a cytoplasmic transfer mechanism, it would be very interesting to examine these mutants for early follicle formation with high spatio-temporal resolution in the fetal and newborn ovary.

In Xenopus and Medaka, only a few cells in the mitotic cyst or neighboring it undergo apoptosis (Kloc et al., 2004; Nakamura et al., 2010), in contrast to the mouse where ~20% of the initial cells within cysts ultimately develop into a primordial follicle with the remaining undergoing massive apoptosis (Bolcun-Filas et al., 2014; Burgoyne et al., 2009; Lake & Hawley, 2012; Lei & Spradling, 2013; MacQueen & Hochwagen, 2011). In zebrafish this remains to be fully addressed, although we did not observe massive apoptosis suggesting that zebrafish is more similar to Medaka and Xenopus. Our recent findings that cyst division planes may regulate the oocyte polarization axis (Elkouby et al., 2016) (see section 5.1), further emphasize the need for analysis of cyst formation. Harnessing live time-lapse imaging strategies to monitor oogonial division in the future will determine the number and pattern of cell division that construct the cyst, as well as the number of cells that undergo apoptosis versus differentiation into oocytes. This will provide key information about cyst dynamics and regulation, and its potential roles in oogenesis.

2.2 The transition from a cyst to a follicle

Oogonial cells give rise to early meiotic stage I oocytes that continue to develop in a germline cyst or nest (Fig. 1). A germline nest is a group of oocytes collectively surrounded by follicle cells where it is unknown if they are connected by CBs. Similar to oogonia, groups of leptotene and zygotene meiotic oocytes cluster tightly together and are collectively engulfed by somatic follicle cells in a cyst or nest (Elkouby et al., 2016). Ultimately, the CBs breakdown between the oocytes through completion of cytokinesis (or cellular abscission) and individual oocytes are surrounded by layers of follicle cells, a process called folliculogenesis (Fig. 1, cellular organization). In Medaka and mice, zygotene oocytes are detected in a germline cyst, while individual oocyte follicles are evident at early diplotene (Lei & Spradling, 2016; Nakamura et al., 2010). Whether the intervening pachytene stage oocytes are still found in the cyst or undergo separation has not yet been addressed in these species. In zebrafish our results indicate that zygotene oocytes are still found in cysts and that pachytene oocytes are already separated with a few follicle cells surrounding them (Elkouby et al., 2016). Thus it appears common that folliculogenesis begins after zygotene and by diplotene stages (Fig. 1- cellular organization).

Mutant analysis in the mouse uncovered factors that are required for the transition from cyst to follicle in the developing ovary. In P3-P7 nobox mutant ovaries (an oocyte specific gene), primary follicles were lacking and cyst-like oocyte clusters remained, suggesting that Nobox is required in the oocyte for this transition (Rajkovic et al., 2004). Foxl2 is required for granulosa cell differentiation and foxl2^−/− ovaries failed to form primary follicles, while cyst-like clusters of degenerating oocytes were observed (Uda et al., 2004), demonstrating an expected role for granulosa cells in the cyst to follicle transition. In addition, Notch pathway components are expressed in both oocytes and granulosa cells in the mouse ovary, and pharmacological inhibition of Notch signaling in cultured ovaries lead to an increase in the number of cyst-like clusters and a decrease in primordial follicles ex-vivo (Trombly et al., 2009).

However, the comprehensive mechanism by which oocytes leave the cyst to become individual follicles is not understood. For example, it is unclear whether single oocytes separate from the cyst one by one, or the entire cyst breaks down and releases all oocytes at once. Recent observations provide some interesting insight. In Medaka single early diplotene oocytes that are already individually surrounded by follicle cells were detected adjacent to a cyst and found to be motile (Nakamura et al., 2010). Gaining motility could help oocytes to break free from the cyst, either collectively or one by one. In mice, apoptosis is thought to contribute to cyst breakdown. Apoptosis and cyst breakdown events overlap in the ovary, but cyst breakdown was reported to precede the majority of germ cell loss, suggesting that apoptosis is not a major factor in cyst breakdown (Lei & Spradling, 2013, 2016). Since only a few apoptotic cyst cells were detected in Medaka (Nakamura et al., 2010) that could contribute to oocyte separation, additional regulation likely is acting.

In zebrafish, we detected oocytes that appeared to be transitioning from late zygotene to early pachytene that had long cytoplasmic bridges containing microtubules, characteristic of abscission stages (Menon et al., 2014; Wickstrom et al., 2010), that were always at the periphery of a cyst (Elkouby et al., 2016). These oocytes may be completing cytokinesis and undergoing abscission as they separate from the cyst (Elkouby et al., 2016). Intriguingly, abscission completion requires the action of follicle cells in the Drosophila testis (Lenhart & DiNardo, 2015). Involvement of follicle cells in abscission would suggest a mechanism for separation from the cyst that is coupled with initiation of folliculogenesis, but this remains to be investigated. It is possible that completion of cytokinesis, oocyte motility and apoptosis are not mutually exclusive as separation mechanisms. These could be operating to varying degrees between species and during different stages of separation.

3. The zygotene chromosomal bouquet

A hallmark of meiosis I is recombination between homologous chromosomes, which increases genetic diversity of gametes. Homologous recombination depends on chromosomes first finding and pairing with their homolog via protein-bridging complexes called synaptonemal complexes. These events take place in the first stages of prophase I, the leptotene and zygotene stages, while homologous recombination is executed at the subsequent pachytene stage (Bolcun-Filas et al., 2014; Burgoyne et al., 2009; Lake & Hawley, 2012; MacQueen & Hochwagen, 2011) (Fig. 1- meiosis). Chromosomal pairing and synapsis universally involve a transient polarized nuclear arrangement, called the zygotene chromosomal bouquet configuration (Scherthan, 2001). In the bouquet, all telomeres are clustered at one pole of the nuclear envelope (NE), while the free looping ends of their chromosomes face the other side (Fig. 2A). In the process of forming the bouquet, telomeres are tethered to and rotate around the NE, driving chromosomal movements in the nucleus. This shuffling of chromosomes facilitates their homology searches for correct pairing. Telomere clustering in the bouquet ceases their rotation around the NE and is thought to stabilize the interactions between homologous chromosomes and reinforce their pairing and synapsis. The bouquet was discovered over a century ago, but how telomeres are attached to the NE, what drives their rotation, and how these processes are regulated have been unknown. The regulation and mechanisms of bouquet formation have only recently been identified.

Figure 2. Mechanisms of telomere NE attachment and clustering during bouquet formation.

(A) Telomere clustering and bouquet formation. Telomere tethering to perinuclear microtubules that emanate from the centrosome enable their rotation around the NE. Telomere movements shuffle entire chromosomes in the nucleus and allow for homology searches. Eventually, these movements lead to clustering of telomeres apposing the centrosome, and the cessation of movements then stabilizes correct homologous chromosome pairing. (B) A model for telomere attachment to the NE. (1) At the onset of meiosis, telomeres are found randomly within the nucleus. The Trf1 protein is bound as a dimer to telomeres as part of the Shelterin complex, a hexameric protein complex that binds to telomeric repeat DNA. Terb1 binds to Terb2, which can bind to Majin. Majin is anchored to the inner nuclear membrane (INM). At leptotene-zygotene stages, RingoA activates CDK2, which phosphorylates SUN1 in the nucleoplasm. This phosphorylation is required for telomere attachment to SUN1 by an unknown mechanism. It is not known if the phosphorylation is transient or persists during the following steps. (2) In response to a yet unknown signal, and perhaps stochastically, Terb1 binds one Trf1 monomer displacing the other. This recruits telomeres to the Terb1/Terb2/Majin complex. (3) Terb1 then binds telomeric DNA and a nearby SUN1 protein on the INM. SUN1 is bound to a KASH protein in the outer nuclear membrane (ONM), which in turn is associated with microtubules in the cytoplasm. This mechanism connects the intranuclear telomeres with the cytoplasmic perinuclear microtubules, allowing for telomere movements around the NE (in A).

3.1 Mechanisms of bouquet formation

The rapid chromosomal movements during bouquet formation at leptotene and zygotene stages are facilitated by a unique connection between chromosome telomeres, a SUN domain protein in the inner NE, a KASH domain protein in the outer NE, and microtubules in the cytoplasm (Fig 2A). Telomere association with the SUN1 protein on the inner NE is required for bouquet formation and synapsis. In sun1^−/− mutant mouse spermatocytes and oocytes and C. elegans oocytes, telomeres or pairing centers (in C. elegans) do not attach to the NE, chromosomes do not rotate around in the nucleus and synapsis is abrogated (Ding et al., 2007; Penkner et al., 2009; Sato et al., 2009; Shibuya et al., 2014a).

Telomere attachment to SUN1 is regulated by phosphorylation of the SUN1 N-terminal domain in both mouse (Ortega et al., 2003; Viera et al., 2015; Viera et al., 2009) and in C. elegans (Penkner et al., 2009). SUN1 phosphorylation is induced in mice by the Cyclin-dependent kinase, Cdk2. cdk2^−/− mice are viable, but sterile, and in mutant spermatocytes and oocytes, telomeres do not localize to SUN1 on the NE and synapsis is abrogated (Ortega et al., 2003; Viera et al., 2015; Viera et al., 2009). Importantly, Cdk2 purified from testes was shown to phosphorylate the N-terminus of SUN1 in vitro (Viera et al., 2015). Furthermore, the atypical non-cyclin Cdk activator protein RingoA, regulates these Cdk2 meiotic functions (Mikolcevic et al., 2016). ringoA mutant mice are also viable but sterile and phenocopy cdk2^−/− in a lack of SUN1 telomere attachment to the NE (Mikolcevic et al., 2016). Importantly, RingoA protein co-localizes with Cdk2 to telomeres during leptotene to zygotene stages and unloads from telomeres at pachytene, consistent with bouquet-specific telomere regulation (Mikolcevic et al., 2016). RingoA is required in vivo for Cdk2 to localize to telomeres. Furthermore, RingoA and Cdk2 physically interact in co-immunoprecipitation experiments (Mikolcevic et al., 2016). Consistent with RingoA acting as a non-canonical cyclin, it was shown that Cdk2 phosphorylation of the SUN1 N-terminus depends on RingoA in in vitro kinase assays (Mikolcevic et al., 2016). Overall, RingoA functions as an activator of Cdk2, which then phosphorylates SUN1, inducing telomeres to attach to SUN1 and the NE (Fig. 2B).

How telomeres associate with SUN1 on the inner NE in vertebrates was unknown until recently. In the fission yeast (S. pombe), two inner NE proteins, Bqt3 and Bqt4, bind telomeric proteins and mediate their binding to the SUN/KASH complex (Chikashige et al., 2006), and in the budding yeast (S. cerevisiae), the NDJ1 protein functions similarly (Conrad et al., 1997; Trelles-Sticken et al., 2000). However, neither homologs or analogs of Bqt3,4, or Ndj1 have been identified in vertebrates. Recently, a series of breakthroughs from the Watanabe lab have dissected this mechanism in mice. The Watanabe group discovered three meiosis-specific proteins, Terb1, Terb2 and Majin that mediate an analogous mechanism in vertebrates via interaction with the ubiquitous telomeric protein Trf1 (Fig. 2B) (Shibuya et al., 2015; Shibuya et al., 2014a). Mutants of all three proteins are viable but sterile, exhibiting abnormal telomere distribution, loss of chromosomal movements and the bouquet organization during zygotene, as well as aberrant chromosomal synapsis (Shibuya et al., 2015; Shibuya et al., 2014a).

Using in vivo mutant analysis, protein domain deletion analysis in isolated cultured spermatocytes ex-vivo, imaging experiments and extensive biochemistry, the following model was constructed (Shibuya et al., 2015; Shibuya et al., 2014a). Majin is a carboxy-terminal tail-anchored class protein spanning the inner NE through its single transmembrane domain, with its N-terminus located in the nucleoplasm. Majin directly binds Terb2, which can bind Terb1. Prior to leptotene, Majin brings Terb1 and Terb2 to the NE (Fig 1A). During leptotene-zygotene stages, Terb1 recruits telomeres to the Majin/Terb2/Terb1 complex by binding both the telomeric protein Trf1, which is part of the Shelterin complex, and telomere DNA via a MYB DNA binding domain (Fig 1A–2). Majin can further bind telomeric DNA via a basic domain near its transmembrane domain, reinforcing the complex. Once this complex forms, Terb1 can bind a nearby SUN1 protein. Thus both Majin and the SUN/KASH complex reinforce telomere attachment to the NE (Fig. 2B).

The bouquet chromosomal movements via SUN/KASH telomere tethering to perinuclear microtubule is demonstrated from fission yeast to mammals, as well as in plants (Chikashige et al., 2006; Cowan et al., 2002; Ding et al., 2007; Morimoto et al., 2012; Murphy et al., 2014; Sato et al., 2009). Interestingly, some species have developed a variation on this zygotene bouquet theme. In Drosophila, chromosome pairing occurs during the mitotic divisions that precede meiosis and utilizes centromeres instead of telomeres (Christophorou et al., 2015; Christophorou et al., 2013; Lake & Hawley, 2012). In C. elegans, pairing occurs during meiosis, but telomeres are again replaced by other specialized chromosomal regions, called pairing centers (Sato et al., 2009). However, even in these non-conventional forms of pairing, chromosomal movements still depend on microtubules via the same chromosome-SUN/KASH-microtubule connection (Christophorou et al., 2015; Hiraoka & Dernburg, 2009; Penkner et al., 2009; Sato et al., 2009). An exception to this mechanism is exhibited in S. cerevisiae, where actin and not microtubules are required for bouquet formation and synapsis (Trelles-Sticken et al., 2005).

The engagement of telomeres with the SUN/KASH complex facilitates the bouquet movements, as it links telomeres to cytoplasmic perinuclear microtubules (Fig. 2). SUN/KASH complexes can interact with a variety of cytoskeletal components in many ubiquitous, as well as cell-specific contexts (Starr & Fridolfsson, 2010). The connection of SUN/KASH complexes to microtubules is mediated by the dynein motor protein that binds KASH proteins and microtubules (Starr & Fridolfsson, 2010). Indeed, in both C. elegans and S. pombe bouquet telomere-SUN/KASH complexes are bound to microtubules via dynein (Chikashige et al., 2006; Sato et al., 2009). In mouse spermatocytes the p150 subunit of Dynactin, a dynein regulatory complex, co-localizes with telomeres in WT but not sun1 or terb-1 mutants, and this interaction persisted beyond the bouquet stages (Shibuya et al., 2014a).

3.2 The bouquet in a broader context of early germ cell differentiation

While these events are mostly characterized in mouse spermatocytes, they are likely conserved in oocytes and across species. Nonetheless, gender and/or species-specific modifications to these mechanisms are plausible. The synapsis checkpoint pathways, which directly cross-talk with homolog pairing processes, differ between spermatocytes and oocytes where they induce immediate arrest and apoptosis in the former, but are more tolerated in the latter (Burgoyne et al., 2009). The complexes and events discussed above might be further subjected to sex specific regulation.

Bouquet mechanisms have also been co-opted for other aspects of meiocyte differentiation. In the fission yeast, bouquet formation and the normal localization of telomeric proteins during the bouquet regulate the integrity of the spindle pole body and spindle formation in the first meiotic division (Tomita & Cooper, 2007). In zebrafish oocytes, where animal-vegetal polarity is key to embryonic development and has been traced to the onset of meiosis, bouquet microtubules are required for the initial symmetry-breaking of the oocyte along the bouquet nuclear axis (Elkouby et al., 2016). The theme arising is of a bouquet configuration of broader cellular significance, functioning in chromosomal pairing, but also in other key developmental aspects of meiocytes, at the nexus of their early differentiation.

Many of the nuclear events of the bouquet configuration are now well understood. Much less is known about the functions of microtubules and the centrosome as the cytoplasmic counterpart of the bouquet. In C. elegans, the dynein motor protein localizes to the microtubule-KASH protein connection sites, where it repels the clustering of such sites (Sato et al., 2009). This was proposed to act as a licensing mechanism, providing counter forces that only correct homologous pairing with sufficient affinity could overcome (Sato et al., 2009). What regulates the centrosome and the perinuclear organization of microtubules during vertebrate bouquet formation is unknown. In C. elegans, presumptive microtubules were detected by TEM, but a global cellular organization was not deduced at bouquet stages (Sato et al., 2009). In mouse isolated spermatocytes, perinuclear microtubule cables were clearly observed at the zygotene stage (Shibuya et al., 2014b). In zebrafish oocytes perinuclear microtubules were detected in vivo specifically at zygotene, but not in oogonia or pachytene (Elkouby et al., 2016), consistent with a role for perinuclear microtubule cables acting specifically in bouquet formation. It will be important to determine in future experiments what initiates the perinuclear organization of microtubules and how the centrosome is regulated during this process, as well as how these events are coordinated with the formation of the Terb1/Terb2/Majin/Sun1 telomere attachment complex on the nuclear side of the NE.

4. Primary cilia in the zebrafish meiotic cyst?

An interesting feature of the zygotene oocyte that was recently detected in the zebrafish ovary is a novel acetylated microtubule cable emanating from the centrosome (Elkouby et al., 2016). These acetylated cables were absent from the oogonial cyst, but detected as a few short cables in the leptotene cyst, then fully elaborated in the zygotene cyst, and finally absent at pachytene stage. The structure and function of these cables is still unclear, but we postulate three potential scenarios (Fig. 3): 1. midbody acetylated microtubules that extend through the CB (cytoplasmic bridge); 2. cables in the cytoplasm that extend around the nucleus; 3. primary cilia-like structures. In Figure 3 we depict pairwise CB connections between oocytes in the cyst. Such pairwise CBs were detected in the late cyst (Elkouby et al., 2016). CBs in earlier cysts were detected in zebrafish by transmission electron microscopy (TEM), where up to three CBs were detected in the same cyst (Marlow & Mullins, 2008a). The number of CBs per oocyte are expected to vary between one to at least three for some or all oocytes in the early cyst. For simplicity, pairwise connections are shown.

Figure 3. Models for acetylated tubulin cables in the germline cyst.

1. Cables extending through the CB and midbody of sister oocytes. 2. Cables curve around and remain within the cytoplasm. The acetylated cables may comprise a new type of cilium-like structure that is fully cytoplasmic or partially cytoplasmic (not shown). 3a. Cables are a primary cilium. 3b. Cables are a long subtype of primary cilia. For simplicity, pairwise CB connections between oocytes in the cyst are depicted. The number of CBs per oocyte may vary from one to four or more depending on the number of mitotic divisions and when each CB undergoes abscission.

The cables could represent the acetylated microtubules of the CB midbody (Danilchik et al., 1998; Perdiz et al., 2011; Piperno et al., 1987; Schatten et al., 1988) (Fig. 3-1). During cytokinesis, CB microtubules become acetylated, and then are deacetylated for abscission and cytokinesis to complete (Menon et al., 2014; Wickstrom et al., 2010). This is consistent with the detection of acetylated cables at zygotene and non-acetylated microtubules in late zygotene-early pachytene CBs, presumably during abscission (Elkouby et al., 2016). Since the CB detected in late zygotene to early pachytene oocytes connects oocytes adjacent to their centrosomes, in this model both ends of the cable would be localized to or near its centrosomes. However, the centrosome localizes to only one end of the cable (Elkouby et al., 2016). In addition, the minus ends of microtubules attach to the centrosome, with the plus ends extending away (Akhmanova & Hoogenraad, 2015; de Forges et al., 2012), so it is unlikely that the plus end of the acetylated cable connects to another centrosome. The ratio of cables per oocytes in a cyst is still unclear and this model would indicate one cable per pair of oocytes.

The acetylated cable extending from the oocyte centrosome highly resembles the structure of a primary cilium (Fig. 3-3). A primary cilium is comprised of an axoneme, a structure composed of stereotypic organized cables of acetylated microtubules that emanate from the centrosome and extend as a cellular protrusion into the extracellular space (Reiter et al., 2012; Satir & Christensen, 2007). While most primary cilia are based from a centrosome that is localized at the cell periphery close to the plasma membrane, examples of cilia with the centrosome more internally based exist (Malicki & Avidor-Reiss, 2014). During mammalian spermatogenesis, a specialized primary cilium extends from a centrosome that is localized at the nuclear periphery rather than at the cell periphery (Fawcett et al., 1970), similar to the acetylated cables we observe. This type of primary cilium is termed a cytoplasmic cilium (Malicki & Avidor-Reiss, 2014).

The primary cilium is known to have mechanical functions (Satir & Christensen, 2007). One context for such a mechanical role of the cable emerges from its specific formation at zygotene. During the formation of the zygotene chromosomal bouquet, dramatic forces are in action. Centrosome based microtubules extend toward the nucleus, where they bind SUN/KASH proteins spanning the NE that bind the telomeres. The telomeres thus linked to cytoplasmic microtubules rotate on the NE and shuffle chromosomes inside the nucleus.

A centrosome based primary cilium or acetylated microtubule cable may generate or regulate the bouquet forces in several ways. It could serve as an anchor point for the movements, provide counter forces, regulate movement direction, stabilize telomere clustering, and once clustering is achieved, it could halt microtubule movements. A mechanical primary cilium could also stabilize other cellular structures or simply keep the entire oocyte from rotating, protecting it from physical damage. It is possible that the cables do not extend as cellular protrusions into the extracellular space, but remain in the cytoplasm (Fig. 3-2). Such a cytoplasmic organization could also account for these mechanical functions. In addition, the length of the cables has yet to be defined, and some cables appear long and curved. They could represent a long type of primary cilium (Fig. 3-3b) that could also account for these proposed roles.

A primary cilium has signaling roles as well (Satir & Christensen, 2007). The synchronous formation of acetylated cables in the cyst may collectively respond to a signal and in turn synchronize oocyte development within the cyst. Further investigation will unveil whether the acetylated cables constitute primary cilia in early oocytes, as well as the function of this intriguing cytoskeletal feature in the cyst.

5. The Balbiani body – a universal feature of differentiating oocytes

The Balbiani body (Bb) is a large RNA-protein granule that is universally conserved in oocytes of insects (Cox & Spradling, 2003; Jaglarz et al., 2003), arthropods (Jedrzejowska & Kubrakiewicz, 2007; VonWittich, 1845), fish (Marlow & Mullins, 2008b), frogs (Dumont, 1978), birds (Carlson et al., 1996; Rodler & Sinowatz, 2013; Ukeshima & Fujimoto, 1991), rodents (Pepling et al., 2007; Weakley, 1967), primates (Barton & Hertig, 1972), and humans (Albamonte et al., 2013; Hertig, 1968). In all cases ultrastructural analysis shows that the Bb is located at the nuclear periphery and is composed of aggregated mitochondria and electron dense material, likely corresponding to RNA. The Bb has been most studied in Xenopus and zebrafish, where Bb localization of specific mRNAs, proteins, and mitochondria are directly demonstrated and its role in oocyte polarization is readily evident (reviewed in (Escobar-Aguirre et al., 2017)). While Bb formation has been poorly understood, new studies have begun to unravel the mechanisms of oocyte symmetry breaking that lead to its formation (Elkouby et al., 2016) as we discuss below.

Interestingly, the mouse Bb contains the Trailer Hitch protein, a member of an mRNA regulating mRNP complex (Pepling et al., 2007), possibly indicating a conserved role in mRNA regulation. However, the mouse Bb is enriched with endoplasmic reticulum (ER) and the Golgi apparatus (Pepling et al., 2007), and Drosophila Trailer Hitch is also an ER resident protein (Wilhelm et al., 2005). Trailer Hitch could therefore represent the enriched ER in the mouse Bb, rather than specifically regulating Bb mRNAs. There is no evidence currently that the Bb polarizes the early mammalian oocyte. However, a potential role for the Bb has been recently proposed during the transition from cyst to primordial follicle in the mouse (Lei & Spradling, 2016) as discussed below.

5.1 The zebrafish Bb and AV polarization

In non-mammalian vertebrates, the animal-vegetal (AV) axis of the egg lays the foundation for the embryonic body plan. This foundation is manifested by the segregation of specific mRNAs and proteins to either pole of the egg that shape its ensuing form following fertilization. The origins of these polarized components reside in the establishment of the oocyte AV axis in early oogenesis.

Two pathways deliver mRNA-protein (mRNPs) granules to the oocyte vegetal pole during oogenesis: the Balbiani body pathway, which initially determines the vegetal pole, and the “late pathway”, which localizes mRNAs to the defined vegetal pole (Escobar-Aguirre et al., 2017). The Balbiani body (Bb) is an aggregate of mRNPs and mitochondria that initially forms adjacent to the nucleus, enlarges and resides in the cytoplasm, and then localizes to the oocyte cortex, where it dissociates (Bontems et al., 2009b; Elkouby et al., 2016; Gupta et al., 2010; Heim et al., 2014; Marlow & Mullins, 2008a; Riemer et al., 2015). Prior to the arrival of the Bb to the plasma membrane, the oocyte cortex is still radially symmetric. Only when the Bb associates with the cortex, disassembles, and docks its contents to the membrane, is this cortex region defined as the vegetal pole. Thus, it is the Bb that first establishes the definitive AV axis of the oocyte.

Importantly, the Bb delivers factors to the vegetal pole that are key for later embryogenesis. The AV axis functions in embryonic patterning and germ cell specification. Since this is the topic of a recent review (Escobar-Aguirre et al., 2017), we only briefly discuss it here. The oocyte and egg vegetal pole harbors factors that regulate dorsal-ventral axis formation via activation of a Wnt signaling pathway (Langdon & Mullins, 2011). In zebrafish and Xenopus, Wnt ligand transcripts are trafficked from the vegetal pole to the future dorsal side of the embryo by microtubules and the microtubule motor linker protein Syntabulin (Colozza & De Robertis, 2014; Lu et al., 2011; Nojima et al., 2010; Tran et al., 2012). Other factors localized to the oocyte vegetal pole function in establishing the germline. Germ cells in zebrafish and Xenopus are specified by the inheritance of germ cell fate determinants, called germ plasm (Extavour & Akam, 2003; Kobayashi et al., 1994). Importantly, mRNA of the dorsal determinants wnt8 and syntabulin, as well as components of the germ plasm, first localize to the vegetal pole in early oocytes via the Balbiani body (Kosaka et al., 2007; Lu et al., 2011; Nojima et al., 2010).

Recent studies have begun to reveal the molecular mechanisms that govern Bb formation, its cortical localization, and finally its disassembly and the tethering of its mRNP cargo to the oocyte cortex. Forward maternal-effect genetic screens in zebrafish recovered two genes that regulate some of these processes, bucky ball (buc) and microtubule actin cross-linking factor 1 (macf1; previously called magellan) (Dosch et al., 2004; Gupta et al., 2010; Marlow & Mullins, 2008a). Buc is the only known protein required for Bb formation. Indeed, in buc^−/− mutant oocytes, the Bb does not form and vegetally-fated mRNAs disperse in the cytoplasm, while animally-destined mRNAs expand radially (Bontems et al., 2009b; Heim et al., 2014; Marlow & Mullins, 2008a; Riemer et al., 2015). The Buc protein localizes to the Bb (Elkouby et al., 2016; Heim et al., 2014; Nijjar & Woodland, 2013; Riemer et al., 2015) and can bind to the Rbpms2 (Hermes) mRNA binding protein (Heim et al., 2014; Nijjar & Woodland, 2013). Buc protein structure and its mechanism of Bb formation are discussed below (section 5.3). Briefly, Buc and its Xenopus ortholog XVelo are predicted intrinsically disordered proteins that have a prion-like domain in their N-terminus (Boke et al., 2016; Bontems et al., 2009b). XVelo forms a stable amyloid -sheet lattice postulated to nucleate Bb mRNP granules, a function likely conserved by zebrafish Buc (Boke et al., 2016).

Macf1 is required for the Bb to localize to the oocyte cortex and disassemble. In macf1^−/− mutants the Bb forms, but fails to dock at the oocyte plasma membrane and disassemble, thus trapping the vegetally-destined mRNAs within the cytoplasm, preventing their final localization (Gupta et al., 2010). During its cortical localization, as well as its disassembly while docking mRNAs to the cortex, the Bb may interact with various cytoskeletal elements. Microtubules form an elaborate network in Xenopus oocytes (Gard, 1991, 1999), while actin forms a cortical network in zebrafish and Xenopus, and is also detected intranuclearly in both (Feric & Brangwynne, 2013; Gard, 1999; Gupta et al., 2010; Marlow & Mullins, 2008a), and cytokeratins were detected in the Xenopus Bb (Gard et al., 1997). The Macf1 protein is a member of the spectraplakin family of cytoskeletal linker proteins that connect cytoskeletal elements in various physiological functions (Liem, 2016). Macf1 may thus link cytoplasmic microtubules or cytokeratins in the Bb to cortical actin in docking the Bb to the oocyte cortex and in disassembling it (Escobar-Aguirre et al., 2017; Gupta et al., 2010).

The buc^−/− and macf1^−/− mutants provide compelling evidence for the importance of the Bb for oocyte polarization and embryo development. Oocytes of both mutants develop into symmetrical eggs, with a radially expanded or multiple ectopic animal poles and no vegetal pole (Gupta et al., 2010; Marlow & Mullins, 2008a). Moreover, the dorsal determinants and germ plasm discussed above, all localize to the vegetal pole via the Bb and are mislocalized in both mutants, further demonstrating its developmental roles.

5.2 XVelo/Bucky ball intrinsically disordered protein aggregation in Bb mRNP granule

An intriguing model for Bb aggregation arises from its structure that resembles that of a hydrogel, a structure of various mRNP granules. A hydrogel is a polymerized meshwork that can form in vitro by the polymerization of a substrate accumulating over a specific concentration threshold (Han et al., 2012; Toretsky & Wright, 2014). The content of the hydrogel is therefore less soluble than the remaining cytoplasm (Brangwynne et al., 2009; Han et al., 2012; Toretsky & Wright, 2014). In a physiological context in C. elegans, a similar phase separation of mRNP complexes was demonstrated, where embryonic germ plasm P-granules form condensed liquid droplets in the embryo posterior, while they disassemble anteriorly (Brangwynne et al., 2009; Wang et al., 2014a). The P granules are ~1000-fold more viscous than water, a viscosity similar to glycerol (Brangwynne et al., 2009). The polymerizing substrates of P-granules are the MEG1 and MEG 3 proteins (Wang et al., 2014b). The condensation of mRNPs into a hydrogel-like P-granule and their dissolution is controlled by the phosphorylation state of MEG1, 3 (Brangwynne et al., 2009; Wang et al., 2014a).

One type of hydrogel and hydrogel-like forming substrates are intrinsically disordered proteins (IDPs), like the MEG proteins (Toretsky & Wright, 2014). IDPs are proteins that lack a fixed or ordered three-dimensional structure, as their amino acid sequence does not dictate folding into a consistent tertiary structure. However, under certain conditions IDPs can fold into a structure that in high concentrations often form a lattice that scaffolds and phase-separates mRNA binding proteins (Guo & Shorter, 2015; Nott et al., 2015; Toretsky & Wright, 2014; Zhang et al., 2015). The human homolog of the piRNA protein Vasa, DDX4, contains an intrinsically disordered region (IDR), as is predicted for other Vasa homologs (Nott et al., 2015). Other mRNA binding proteins contain predicted IDRs adjacent to their RRM RNA binding domain (Guo & Shorter, 2015; Zhang et al., 2015). These are thought to promote the phase separation of mRNA granules, where the RRM domain binds RNA, while the IDR comprises the lattice-like meshwork of the phase separating granule (Guo & Shorter, 2015; Nott et al., 2015; Zhang et al., 2015). Interestingly, it has been proposed that the RNA molecules in the complex could regulate the phase transition, by modifying the interactions between RNA binding proteins, in an RNA specific manner, either strengthening interactions and leading to a more separated viscous granule, or weakening them and leading to a more soluble granule (Guo & Shorter, 2015; Nott et al., 2015; Zhang et al., 2015).

Intriguingly, the Bucky ball amino acid sequence indicates that it is an IDP (Bontems et al., 2009a; Toretsky & Wright, 2014), suggesting that it could act similarly to IDPs in other granules, by forming a scaffolding phase-separating lattice that aggregates mRNAs and mRNPs during Bb formation (Escobar-Aguirre et al., 2017). Indeed, such a mechanism was recently suggested for the Buc Xenopus homolog XVelo (Boke et al., 2016). Mature Bbs isolated from Xenopus oocytes exhibit features of an amyloid structure, such as resistance to dissolution at high salt concentration and high temperature, and it stained positively for an amyloid dye in vivo. Furthermore, the N-terminus of XVelo, which is conserved among Buc orthologs (the BUVE in (Bontems et al., 2009b)), contains a prion-like domain that is adjacent to the extended predicted IDR. This domain in XVelo was shown to be necessary and sufficient for Bb localization in ectopic expression studies in vivo, and formed an SDS resistant amyloid-like structure in vitro.

Consistent with forming a stable lattice of an amyloid scaffold, exogenous XVelo-GFP within the Bb showed very slow fluorescence recovery after photobleaching (FRAP). In contrast, mutated forms of the XVelo prion domain recovered much more rapidly, suggesting that the prion domain is required for formation of the stable amyloid lattice. Other RNA binding proteins with prion domains (known as well as predicted) either did not localize to the Bb or showed fast recovery in FRAP experiments when ectopically expressed, suggesting that the ability of XVelo to form an amyloid lattice in the Bb is unique.

XVelo could further aggregate the Bb-localized mRNA Xcat (nanos) in vitro, as well as mitochondria in Xenopus egg extracts. For both of these aggregation processes, a predicted RNA binding domain at the XVelo C-terminus was required. XVelo forms that retained the N-terminal prion-like domain but lacked the C-terminus formed an amyloid aggregate that could not recruit either mRNA or mitochondria. This suggests that while the conserved XVelo N-terminal prion domain is required for the formation of an amyloid-like matrix, its C-terminus is required for its aggregation of putative Bb components (Boke et al., 2016). However, the XVelo amyloid structure not only recruited Xcat in vitro, but also the control mCherry mRNA, arguing that other factors in normal physiological conditions are required to confer the specificity for mRNA aggregation in the Bb, and self-assembly is not sufficient for normal Bb formation.

As expected for protein orthologs, ectopic expression of zebrafish Buc in Xenopus oocytes localized to the Bb where it co-localized with XVelo (Boke et al., 2016). Swapping of the XVelo prion domain with that of the Buc protein resulted in normal XVelo Bb localization in oocytes, and with similarly low FRAP recovery, suggesting that a normal Bb amyloid matrix is formed. In contrast, the swapping of the XVelo prion domain with that of an aggregating form of the FUS protein did not result in Bb localization and exhibited rapid FRAP recovery, consistent with the inability of other prion domain mRNA binding proteins to localize to the Bb and form a stable slow-FRAP lattice. In the context of the homologous function of Buc and XVelo, the observation that XVelo requires the predicted RNA binding C-terminal domain to aggregate putative Bb components is intriguing. While Buc is comprised of 639 amino acids, a mere 37 amino acid truncation of its C-terminus in the buc^p106 mutant abolishes Bb formation (Bontems et al., 2009b; Marlow & Mullins, 2008a) and a transgene expressing this form cannot rescue Bb formation in a buc mutant (Heim et al., 2014). Although this predicted truncated Buc protein still contains the N-terminal prion-like domain, Buc aggregation into a Bb was not observed. Thus the N-terminal prion-like domain may be sufficient to aggregate in vitro, but may be insufficient in vivo to aggregate on its own. Alternatively the Buc^p106 truncated protein may be unstable or insufficiently stable under normal physiological conditions to aggregate into a Bb.

The XVelo self-assembly observations from overexpression, cell-free, and in vitro studies provide important information about the aggregating properties of XVelo and Buc in forming the Bb. In the future, it will be important to investigate these properties and functions of the endogenous XVelo/Buc protein. CRISPR/Cas9 genome editing can now be used to generate alleles that lack specific domains of Buc in zebrafish, enabling a protein structure-function analysis in vivo. Such studies will add to our understanding of the specific functions of Buc domains in Bb aggregation, assembly, and dissociation, which establish oocyte polarity and lay the groundwork for patterning the early embryo.

5.3 Symmetry-breaking of the zebrafish oocyte - oocyte patterning is coupled with meiosis

Previously the earliest polarity detected in the oocyte was the mature Bb at mid diplotene stage (Fig. 1- Balbiani body). The origins of oocyte polarization have only very recently been uncovered. We now know that oocyte polarity initiates during the earliest stages of vertebrate oogenesis, which were previously difficult to access by whole-mount three-dimensional morphological analysis. We established methods to overcome these limitations (Elkouby et al., 2016) and provide a comprehensive description of these methods in the accompanying Methods paper. This enabled us to examine Bb component localization earlier in oogenesis, identify the earlier origins of oocyte polarity by tracing back symmetry-breaking, and to reveal key steps in early Bb formation (Fig. 4A). We first utilized telomere dynamics as a definitive molecular meiotic progression marker to characterize and define the earliest meiotic stages of oogenesis, the oogonia, leptotene, zygotene, pachytene and diplotene stages (Elkouby et al., 2016) (accompanying Methods paper). We then investigated early polarization and Bb formation during this pipeline of oogenesis.

Figure 4. Dynamics of oocyte polarization and early Bb formation.

(A) The model depicts the size and cellular morphology of specific meiotic stages in the zebrafish and specifies events in oocyte polarization and Bb formation. Oocyte symmetry is broken during the zygotene bouquet stage, when Bb precursors first localize in the cytoplasm apposing the telomere bouquet. This is executed by the centrosome organizing center (MVC, Blue dashed circle). A nuclear cleft forms at pachytene, which is most pronounced at the onset of diplotene and then recedes until the nucleus resumes a spherical shape in ~50 m in diameter mid-diplotene oocytes. The centrosome localizes roughly to the center of the cleft and dissociates in the early ~25 m diameter diplotene oocyte. The panels below the cartoons show immunostained oocytes representative of each stage. Key steps in Bb formation are shown in green boxes. During cleft stages the aggregates are still amorphous and gradually consolidate into a single spherical granule of the mature Bb, in parallel to cleft closure. Scale bars are 5 m for oogonia-zygotene, and 10 m for pachytene and diplotene. Figure panels modified from (Elkouby et al., 2016). (B) Conservation of early Bb formation at the onset of meiosis. Left, a mouse oocyte at E14.5 shows early aggregation of Bb Golgi (white arrow, GM130, red) around the centrosome (Pericentrin, green), similar to the zebrafish Bb. Scale bar is 5 m. Image from (Lei & Spradling, 2016). Right, histological section of a Thermobia ovary showing a zygotene oocyte (dashed yellow ellipse), stained with Methylene Blue. A Bb mitochondrial aggregate (black arrow) is detected apposing the presumptive telomere cluster of the zygotene bouquet (white arrowheads, where chromosomes “touch” the NE), as also depicted in the zebrafish zygotene oocyte in (A). Scale bar is 5 m. Image modified from (Tworzydlo et al., 2016a).

Since the early Bb forms in all organisms intimately adjacent to the nucleus, we hypothesized that the nuclear polarity of the zygotene bouquet stage (discussed in section 3) may be involved in Bb formation and oocyte polarization (Elkouby et al., 2016). We found that the centrosome localizes specifically in the cytoplasm apposing the telomere cluster during the bouquet stage, and seems to organize microtubules to radiate towards and around the nucleus (Elkouby et al., 2016), consistent with their organization in the bouquet of mouse and zebrafish spermatocytes (Saito et al., 2014; Shibuya et al., 2014b). As discussed in section 3, during bouquet formation microtubules emanate from the centrosome and associate with SUN/KASH-telomere complexes, facilitating chromosomal movements (Chikashige et al., 2006; Ding et al., 2007; Mikolcevic et al., 2016; Morimoto et al., 2012; Page & Hawley, 2003; Saito et al., 2014; Sato et al., 2009; Scherthan, 2001; Shibuya et al., 2015; Shibuya et al., 2014a; Trelles-Sticken et al., 2000; Viera et al., 2015).

We traced the symmetry breaking of the oocyte to the zygotene chromosomal bouquet (Fig. 4A; Fig. 1- Balbiani body) (Elkouby et al., 2016). We found that oocyte symmetry is broken when Bb components, such as Buc, GasZ and mitochondria transition from radial distribution in the pre-meiotic oogonial stage to localization around the centrosome in the cytoplasm apposing the telomere cluster of the bouquet. We treated cultured ovaries with nocodazol to depolymerize microtubules and simultaneously monitored telomere distribution and Bb mitochondrial localization. Strikingly, we found that microtubules are concomitantly required to cluster the telomeres of the bouquet and localize the Bb mitochondrial precursors, demonstrating that these nuclear and cytoplasmic events are mechanistically coordinated. Therefore, the bouquet stage centrosome functions as a global cellular organizer that couples meiotic genetic events with oocyte patterning, which we termed the meiotic-vegetal center or centrosome organizing center (MVC; Fig. 4A).

Following the zygotene bouquet stage, using immunohistochemistry on fixed ovaries, vital dye staining of live ovaries and ultrastructural analysis, we detected a nuclear morphology, where the NE is highly concave and forms what we term a nuclear cleft (Fig. 4A) (Elkouby et al., 2016). The nuclear cleft formed apposing the centrosome, which localized in the cleft cytoplasm thereafter (Elkouby et al., 2016). Strikingly, Bb precursors aggregate around the centrosome within the novel nuclear cleft as early as in pachytene stages (Figs. 3A). Using quantitative imaging and MATLAB programing, we detected a robust enrichment of dazl mRNA, the Buc protein and mitochondria in the nuclear cleft cytoplasm compared to the remaining cytoplasm (Elkouby et al., 2016). Other reports also detected a polarized aggregate of Buc prior to mature Bb stages (Heim et al., 2014; Riemer et al., 2015). The nuclear cleft then gradually rounds out during early to mid diplotene, giving rise to the mature Bb.

The mechanism of nuclear cleft formation is still unknown. However, when the Drosophila oocyte is polarized along its dorsal-ventral axis, a similar nuclear morphology is evident that requires MTOC derived microtubules that push against the nucleus, (Tongtong Zhao et al., 2012). During this process, the nucleus migrates from the posterior to the anterior of the oocyte, where it establishes the dorsal-anterior position of the oocyte (Gonzalez-Reyes et al., 1995; Lei & Warrior, 2000; Roth et al., 1995; Tongtong Zhao et al., 2012). MTOCs posterior to the nucleus nucleate growing microtubules that push the nucleus anteriorly, causing an indentation in the NE very similar to the zebrafish oocyte nuclear cleft (Tongtong Zhao et al., 2012). The organization of microtubules during zebrafish cleft formation is not known, but the centrosome localizes within the cleft, consistent with such a mechanism. Immediately prior to cleft formation, during the zygotene bouquet configuration, microtubules emanate from the centrosome and radiate perinuclearly. It is possible that microtubules rearrange following the zygotene bouquet organization to push the NE and form the cleft at pachytene.

These two cases of nuclear clefts could, however, differ in their regulation. In the zebrafish the changes in nuclear morphology are accompanied by changes in the oocyte nuclear lamina composition. While post-cleft oocytes with spherical nuclei express both Lamin B1 and Lamin A/C, cleft stage oocytes only express Lamin B1 (Elkouby et al., 2016). During cleft stages, when the NE presumably requires more flexibility, Lamin A/C, which adds rigidity to the lamina (Swift et al., 2013; Vargas et al., 2012), is specifically absent. It is important to determine the function of the nuclear cleft in Bb formation. One possibility is that the cleft could help separate Bb components from the remaining cytoplasm as the Buc/XVelo amyloid aggregates form and mRNP phase-separation presumably takes place (as discussed above in section 5.2).

The zygotene bouquet symmetry breaking events lie functionally upstream to the Bucky ball protein. In buc^−/− oocytes both the bouquet and the presumptive Bb mitochondrial aggregate initially form normally (Elkouby et al., 2016). Only later during diplotene stages does the mitochondrial aggregate disperse. These results support the potential function of Buc as an intrinsically disordered prion-like nucleating protein (as discussed above in 5.2), which becomes concentrated in a nuclear cleft following the bouquet stage and aggregates with mitochondria, vegetally destined transcripts, and other proteins to form the large mRNP granule of the Bb.

5.4 The Bb and the chromosomal bouquet are linked– a universal theme in oocytes?

Altogether these findings for the first time connect the Bb and the bouquet, two universal oocyte features, in establishing cellular polarity by the centrosome organizing center (MVC). The universal conservation of both the Bb and the zygotene chromosomal bouquet suggests that such a coupling of their formation may be conserved. Indeed, a recent report shows that the Bb of the insect Thermobia domestica is first aggregated apposing the telomere cluster of the Thermobia oocyte zygotene bouquet (Fig. 4B) (Tworzydlo et al., 2016b). Interestingly, Thermobia ovarioles show a more “vertebrate”-like mode of oogenesis, where all germ cells in the germarium are fated to become oocytes, as opposed to some becoming nurse cells like in Drosophila (Tworzydlo et al., 2016b).

Recent analysis of Bb formation in mice oocytes reveals that Bb components initially localize around the centrosome, followed by their further aggregation, similar to that of Bb components in the zebrafish. The mouse Bb Golgi component associates with the centrosome at E14.5, the onset of meiosis in mouse oocytes (Fig. 4B) (Lei & Spradling, 2016), similar to the initial polarization at zygotene stages of zebrafish oocytes (Fig. 1- Balbiani body). The size of the Golgi and the number of mitochondria increases and aggregates around the centrosome within the germline cyst through E18.5 (Lei & Spradling, 2016), similar to the aggregation in pachytene zebrafish oocytes (Fig. 1- Balbiani body). Finally, the mature mouse Bb is detected in stages of the primordial follicle at P0-P7 (Lei & Spradling, 2016; Pepling et al., 2007), similar to the zebrafish early diplotene stages during early folliculogenesis (Fig. 1- Balbiani body). These studies from mice and Thermobia suggest that the mechanisms revealed in zebrafish oocytes may act in these species as well, highlighting a potentially universal coupling of Bb formation with meiosis.

5.5 Potential regulation of polarity by cyst mitotic division plane

While symmetry-breaking of the zebrafish oocyte was first detected at zygotene stage, the organization of the oocytes in the germline cyst suggests earlier polarization. We found that during apparent abscission of the cytoplasmic bridge connecting oocytes at a late zygotene to early pachytene stage, the oocyte likely separates from the cyst (Elkouby et al., 2016). At this point the centrosome was localized to the cytoplasm adjacent to the CB (cytoplasmic bridge) in both presumptive connected sister oocytes (Elkouby et al., 2016) (Fig. 3, microtubules between oocytes depict the CB). Bb precursors are already aggregated around the centrosome at this stage, revealing an alignment of the polarization axis to a previous oogonial division plane (Elkouby et al., 2016). If the centrosome position is maintained between mid zygotene through early pachytene stages (Shibuya et al., 2014b), it suggests that the centrosome may localize to the cytoplasm adjacent to the CB prior to bouquet formation and thus determine the future axis of polarization.

The organization of oogonia and meiotic oocytes in germline cysts is widely conserved. The localization of the centrosome and Bb components adjacent to the CB in zebrafish is the first evidence for a potential role of mitotic cyst division in controlling meiotic oocyte polarization. Mitochondria were previously detected near the centrosome and CB of oogonia in the Xenopus mitotic cyst (Kloc et al., 2004), but their relationship to later Bb formation and cell polarity was unclear. Polarization thus may occur in two steps: first, the centrosome localizes to the cytoplasm adjacent to the CB of cyst oocytes, positioning the future oocyte AV polarity axis orthogonal to the previous cell division plane. Then polarization is in effect executed during bouquet stages through the MVC (Elkouby et al., 2016). Such a dual and dynamic role of the centrosome in cell division and the immediately following polarization during differentiation was beautifully demonstrated in the developing C. elegans gut (Feldman & Priess, 2012; Yang & Feldman, 2015). In this system the apical-basal polarity of differentiating gut epithelial cells is aligned with their previous cell division plane. In these cells, after the centrosome functions as a microtubule organizing center (MTOC) during mitosis, it re-localizes to the future apical pole together with polarity proteins.

Thus, Bb formation, its upstream positioning and assembly, are interwoven with key events in early oocyte differentiation, from potential regulation by mitotic divisions in the germline cyst, through coupling with chromosomal pairing during the meiotic bouquet, to maturing during folliculogenesis.

5.6 The mouse Bb and the transition from cyst to follicle

The Bb has recently been associated with oocyte survival and primordial follicle formation in the mouse (Lei & Spradling, 2016). In the mouse, oocytes reside in cysts until ~E17.5, and primordial follicles are fully formed by P4 (Lei & Spradling, 2013, 2016). Consistent with zebrafish (Elkouby et al., 2016), Bb formation in the mouse is detected when oocytes still reside in germline cysts and completes in the primordial follicle, thus overlapping with the transition from cyst to follicles (Lei & Spradling, 2016). It has been proposed that the Bb selectively forms in oocytes that will become follicles, as opposed to those that undergo apoptosis at this stage (Lei & Spradling, 2016) (discussed in section 2.1). It was proposed that Bb components, such as Golgi, mitochondria and centrosomes, are transported between oocytes through cytoplasmic bridges in the cyst, contributing to Bb formation and oocyte growth in the recipient oocytes, while the “donor” oocytes are apoptosed (Lei & Spradling, 2016). This has been deduced by detection of increased amounts of these components in some oocytes in the cyst and reduced detection in nearby apoptosing oocytes of the same clone. The authors also detected localization of some components near cytoplasmic bridges, possibly in transit to the neighboring cell (Lei & Spradling, 2016). Live time-lapse imaging analysis of organelles or other components transiting through CBs is needed to support this intriguing model, since other mechanisms for such a differential detection could also be involved.

Intriguingly, CBs between oocytes are absent in tex14^−/− female mice, although they exhibit normal fertility (Greenbaum et al., 2009). While fewer oocytes are detected in P2.5 tex14^−/− ovaries, the number of germ cells in these ovaries during cyst stages is normal and oocytes complete folliculogenesis normally (Greenbaum et al., 2009). If transport of Bb components and other organelles through cytoplasmic bridges of prospective apoptosing oocytes is required for oocyte progression, females mutant for tex14^−/− would be expected to have reduced fertility.

Importantly, based on the mutually exclusive detection of Bb formation and apoptosis in early mouse oocytes, it was proposed that Bb formation in an oocyte favors its continuous differentiation and follicle formation, and prevents apoptosis. Whereas oocytes that do not form a Bb would undergo apoptosis (Lei & Spradling, 2016). The notion of the Bb as a possible marker for fit oocytes could reflect a new important function in mammals, irrespective of its association with apoptosis. However, a mutually exclusive detection of apoptosis and Bb formation could also result reciprocally from Bb degradation by apoptotic processes. It is possible that Bb degradation precedes other apoptotic phenotypes, such as nucleus degeneration, since the mitochondrial outer membrane provides the platform for initiating apoptosis (Czabotar et al., 2014; Gillies & Kuwana, 2014) and mitochondria are highly enriched in the Bb. Further studies are needed to address these issues and distinguish between the causes or effects of Bb formation and apoptosis.

6. Concluding remarks

The advances in our understanding of early oocyte differentiation and ovarian development described here, such as organization of germline stem cells, cyst formation and breakdown, bouquet mechanics, oocyte polarity, and Bb formation provide mechanistic insights into long standing questions in germ cell biology. The Bb arises as a central aspect of early oocyte differentiation in the meiotic cyst and early follicle. The Bb may be positioned by the last mitotic cell division plane, and it is formed in mechanistic coordination with the chromosomal bouquet in a likely conserved fashion. Furthermore, it may be an indicator of oocytes that will perdure to become primordial follicles in the mouse. Many intriguing questions remain to be addressed in each of these stages. The Bb provides a new context for investigating how multiple facets of oocyte differentiation are coordinated with and/or by the meiotic groundwork. Such coordination and mechanistic crosstalk between pathways will ultimately generate a more unifying paradigm for the earliest stages of oogenesis in mammalian and non-mammalian vertebrates.

Dissecting the mechanisms that control early oocyte differentiation is challenging. The recent observations that we discuss here have become possible due to technical advances in experimental methodology and design, such as quantitative microscopy, early gonad culturing, and time-lapse analysis. In addition, searching for novel regulators by ovary or maternal-effect forward genetic screens is limited to identifying genes that do not cause embryonic lethality when mutant or to hypomorphic mutations that allow survival to adulthood but are insufficient for oogenesis. A second challenge is mutant genes that cause defects in oogenesis that lead to oocyte loss due to meiotic checkpoint-induced apoptosis. Advances in genome editing now allow for a powerful reverse genetics approach. Mutations in meiotic and early oogenesis-specific genes can be readily made by CRISPR/Cas9 (Gagnon et al., 2014; Montague et al., 2014). Importantly, to overcome embryonic lethality and apoptotic phenotypes, the generation of conditional alleles is now possible in zebrafish by introducing loxP sites at specific loci by either CRISPR/Cas9 or methods that utilize homologous recombination (Ablain et al., 2015; Hoshijima et al., 2016). This will allow the analysis of gene functions that have been previously inaccessible for investigation in early oogenesis. Thus, the combination of these new methods holds great promise for further breakthroughs in the near future.

Highlights.

• Understanding early oocyte differentiation is key to advancing our knowledge of germ cell biology, reproduction and health.

• We review recent advances in the field, including breakthroughs on germline stem cells, the germline cyst, meiotic chromosomal pairing, symmetry breaking and oocyte polarity, and the Balbiani body, a universal oocyte mRNP granule.

• We compare and discuss findings from zebrafish, Medaka, Xenopus, and mouse.

• We discuss an emerging theme of meiosis as a groundwork for multiple processes in early oocyte differentiation.

• We propose am inclusive paradigm for early oocyte differentiation in vertebrates.