The germ plasm is anchored at the cleavage furrows through interaction with tight junctions in the early zebrafish embryo

Germ plasm consists of a maternally inherited ribonucleo-protein (RNP) condensate, which controls the formation of the germline in many animals (Strome and Updike, 2015; Aguero et al., 2017). Germ plasm thereby acts as a classical cytoplasmic determinant during embryonic development with the following activities: (1) in the zygote, germ plasm is uniformly distributed and, after the cleavage, period leads to the formation of a subpopulation of embryonic cells containing germ plasm; (2) these cells will be programmed to differentiate into primordial germ cells (PGCs), while the other cells without germ plasm adopt a somatic fate, e.g. neuron, muscle, etc. Proper segregation of germ plasm allows its accumulation in presumptive PGCs, whereas it is subsequently degraded in prospective somatic cells.

The germ plasm specification of PGCs seem to be largely conserved during evolution, because many components like Vasa, Nanos and Piwi are present throughout most animal genomes (Ewen-Campen et al., 2010; Juliano et al., 2010). By contrast, it is currently unknown whether the molecular mechanisms controlling localization of germ plasm are also conserved during evolution.

The positioning of germ plasm during embryogenesis is best understood in invertebrates, because of their powerful molecular-genetic tools. In C. elegans, the entry of sperm determines embryonic polarity (Otto and Goldstein, 1992; Strome and Wood, 1983), which eventually leads to asymmetric localization of germ plasm and germline specification (Seydoux, 2018; Strome and Updike, 2015). In the fly Drosophila, local translation of the germ plasm organizer Oskar (Osk) recruits germ plasm components to the cellular cortex of the posterior pole (Ephrussi and Lehmann, 1992; Kim-Ha et al., 1993; Trcek and Lehmann, 2019). Among vertebrates using germ plasm for germline specification, some key discoveries of its localization were made in the frog Xenopus laevis (Houston, 2013; Aguero et al., 2017). In Xenopus laevis it was shown that during oogenesis germ plasm first accumulates at the prominent Balbiani body (BB), also called mitochondrial cloud (Heasman et al., 1984). Germ plasm then becomes anchored at the vegetal pole and after fertilization is passively inherited during the cleavage period of the most vegetal blastomeres (Ressom and Dixon, 1988; Aguero et al., 2017). At the blastula stage, germ plasm-positive cells internalize into the embryo and then start their migratory journey until they reach the gonads. However, the molecular structure tethering germ plasm to the vegetal pole during the cleavage period of Xenopus embryogenesis is not known.

In zebrafish egg, germ plasm also initially accumulates at the BB and subsequently localizes to the vegetal pole, as in Xenopus (Dosch, 2015; Moravec and Pelegri, 2020; Raz, 2003). However, in contrast to Xenopus, after fertilization, germ plasm streams together with cytoplasm during ‘ooplasmic segregation’ into the forming blastodisc at the animal pole of the zebrafish embryo (Welch and Pelegri, 2014). Subsequently, germ plasm localizes to the cleavage furrows at the four-cell stage, forming four aggregates in close proximity to the apical ends of the furrows (Olsen et al., 1997; Raz, 2003; Yoon et al., 1997). Indeed, maternal mutants affecting the first embryonic cleavages also interfere with germ plasm recruitment (Nair et al., 2013; Yabe et al., 2007). The first described cytoskeletal structure tethering germ plasm in zebrafish was described as a furrow-associated microtubule array (FMA) (Jesuthasan, 1998; Pelegri et al., 1999). However, the FMA starts to disassemble after the third cleavage, leaving the molecular identity of the cellular structure anchoring germ plasm after the eight-cell stage unresolved. A role for cytoskeletal structure for germ plasm transport is further supported by the observation that inhibitors for Rho and Rock activity, which are enriched at the cleavage furrows at the four-cell stage, affect cytoskeletal structures and result in additional mislocalized germ plasm aggregates (Miranda-Rodríguez et al., 2017).

Molecular and genetic screens identified the proteins that are specifically localized to these four germ plasm spots, e.g. zebrafish Piwi (Ziwi) (Houwing et al., 2007), phosphorylated non muscle myosin II (p-NMII) (Nair et al., 2013) and Bucky ball (Buc) (Bontems et al., 2009; Campbell et al., 2015; Riemer et al., 2015; Roovers et al., 2018). Buc appears to exert a central role during germline specification, because it acts as a germ plasm organizer by recruiting other germ plasm components and thereby triggers germline specification (Bontems et al., 2009; Heim et al., 2014; Krishnakumar et al., 2018; Marlow and Mullins, 2008). Buc interacts through Kinesin Kif5Ba with microtubules, which is essential for Buc transport towards the cleavage furrows (Campbell et al., 2015). However, it is not clear which cellular structure anchors Buc after its transport to the four germ plasm spots in the early embryo.

Here, we show that the germ plasm nucleators Buc and its Xenopus homolog Velo1 use conserved mechanisms for their anchorage, whereas Drosophila Osk localizes by a distinct mode. We mapped the localization motif in the Buc protein and used the isolated peptide to purify its interactors from zebrafish embryos. Among numerous proteins, we identified subunits of the NMII complex, which is a known cytoskeletal component of adherens junctions, tight junctions and midbodies (Liu et al., 2012; Vicente-Manzanares et al., 2009). In addition, ZO2 and the adherence receptor of tight junctions claudin-d (Cldn-d) were identified to be associated with the Buc localization domain. Furthermore, we discovered that TJ protein ZO1 colocalizes with the four germ plasm aggregates at the cleavage furrows at the eight-cell stage. Electron microscopy (EM) of zebrafish embryos uncovered TJ-like structures at the cleavage furrows that are in proximity to germ plasm at the eight-cell stage. Moreover, overexpressing the tight junction receptor Cldn-d led to the formation of ectopic germ plasm aggregates in zebrafish embryos. Taken together, our results identify TJs as the cellular structures that recruit germ plasm at the onset of zebrafish embryogenesis.

Consistent with its function as a germ plasm organizer, Buc localizes to the germ plasm throughout early embryogenesis (Bontems et al., 2009; Heim et al., 2014; Riemer et al., 2015). To address whether this localization mechanism is conserved between zebrafish and Xenopus, we injected mRNA encoding GFP-fusions of these germ plasm organizers into one-cell zebrafish embryos (Fig. 1A). At 2.5-3 h post fertilization (hpf), we analysed whether the GFP-fusion proteins colocalized with the germ plasm using an antibody against the endogenous Buc protein, which is tightly associated with the germ plasm (Riemer et al., 2015), and an antibody detecting β-catenin to label the membrane of the cleavage furrows. Western blot of in vitro translated proteins confirmed the specificity of the Buc antibody (Fig. S1). Injections of mRNA encoding Buc-GFP colocalized with zebrafish germ plasm, recapitulating the positioning of the germ plasm (Fig. 1B,C, Fig. S2A) (Bontems et al., 2009). Similarly, Velo1-GFP colocalized with the germ plasm (Fig. 1B,D, Fig. S2B), suggesting that zebrafish Buc and Xenopus Velo1 are targeted by a similar molecular machinery for germ plasm recruitment.

To test whether the localization mechanism also detects the invertebrate germ plasm organizer short Osk, mRNA of sOsk fused to GFP was also injected. In contrast to Buc and Velo1, sOsk-GFP did not overlap with the germ plasm in injected zebrafish embryos (Fig. 1B, Fig. S2C), but instead localized to the nuclei, as previously shown in insect cells and in Drosophila embryos (Jeske et al., 2017; Kistler et al., 2018), similar to control injections of a GFP that resulted in a ubiquitous subcellular localization, including the nucleus (Fig. 1B,E, Fig. S2C,D). These results suggest that the recruitment machineries of germ plasm in zebrafish and Xenopus are conserved, but not interchangeable between zebrafish and Drosophila. To test whether this non-overlapping recruitment mechanism is also true for Buc in Drosophila, we tested whether ectopic localization of Buc to the anterior pole in Drosophila embryos is sufficient to recruit endogenous germ plasm and the subsequent formation of ectopic PGC, as shown for Osk (Ephrussi and Lehmann, 1992). We fused the buc ORF to GFP and a bicoid-3′-UTR to direct its translation to the anterior pole of Drosophila embryos (Fig. S3A). As a control, we used sOsk ORF fused to the bicoid-3′-UTR (sOsk) (Tanaka and Nakamura, 2008). Immunolabelling of stage 4-5 fly embryos showed that sOsk-GFP can be detected at the anterior pole of the embryos and was sufficient to induce the formation of ectopic PGCs (Fig. S3B,D) by the ectopic localization of germ plasm, including Vasa protein at the anterior pole (Fig. S4). By contrast, although Buc was expressed anteriorly, it was neither anchored to the embryo cortex nor did it ectopically aggregate germ plasm or form ectopic PGCs (Fig. S3E, Fig. S4). These results show that Buc is not recognized by the localization machinery in Drosophila that anchors Osk to the cortex and is not sufficient to aggregate germ plasm ectopically, suggesting that zebrafish and flies use different mechanisms for germ plasm recruitment.

To identify the protein domain of Buc that is responsible for its recruitment to the four germ plasm aggregates, we generated systematic deletions of Buc fused to GFP (schematically shown in Fig. 2A), injected their mRNAs into zebrafish one-cell stage zygotes and scored the number of embryos with GFP foci at 3 hpf as depicted in Fig. 1A.

An N-terminal fragment [amino acids (aa) 1-361, Fig. 2A] that corresponds to the previously identified buc^p43 mutant allele localized correctly and with the same penetrance as full-length Buc (Fig. 2A-C, Fig. S5A). Next, we split this fragment into two halves (aa1-158 and 159-361) and analysed their localization. Buc1-158 localized, whereas Buc159-361 showed ubiquitous fluorescence, similar to control embryos injected with GFP mRNA (Fig. 2A,B, Fig. S5B,C). We then split Buc1-158 into two fragments and in addition removed the first ten amino acids (Buc11-88), which show a low conservation in teleost evolution (Škugor et al., 2016). Buc11-88 was sufficient to recapitulate germ plasm localization, whereas Buc89-158 showed no specific localization (Fig. 2A,B,E, Fig. S5D). Further splitting of Buc11-88 disrupted the localization activity of both resulting fragments (Fig. 2A,B, Fig. S5E,F), suggesting that aa11-88 contains the residues sufficient to target the protein to the germ plasm spots. To confirm that Buc does not contain other motifs involved in localization, we generated a deletion of the isolated motif aa11-88 (BucΔ11-88) in full-length Buc. This protein did not localize (Fig. 2A,B,F). We therefore concluded that aa11-88 is sufficient and necessary for the localization of Buc and named the protein region BucLoc. However, this experiment does not allow us to distinguish whether BucLoc is autonomously anchored at the cleavage furrow or recruited by endogenous Buc in the germ plasm granules. The identified BucLoc domain is part of the so called BUVE domain that shows the highest homology to Xenopus Velo1 (Boke et al., 2016; Bontems et al., 2009; Krishnakumar et al., 2018).

The BUVE domain was recently shown to be responsible for Velo1 aggregation at the BB during Xenopus oogenesis (Boke et al., 2016). The localization of Velo1 to the BB is driven by aggregation of a prion-like domain with two regions enriched in aromatic amino acids within the BUVE motif (Boke et al., 2016). A sequence alignment of Buc with Velo1 showed the conservation of these aromatic amino acids regions of the PLDs in Buc between aa24-30 and 64-71 (Fig. 3A, marked in red), suggesting that these regions might also be required for the formation of the four germ plasm aggregates at the eight-cell stage in zebrafish.

To investigate the importance of these two potential PLD domains in Buc, the colocalization of deletion variants with the germ plasm was analysed. Therefore, mRNA of deletion variants of the BucLoc domain (shown schematically in Fig. 3D) fused to mCherry were injected into one-cell embryos, and their colocalization to germ plasm aggregates marked by Buc-GFP was examined at 3 hpf. As a positive control, we used the entire BucLoc domain (aa11-88) that shows colocalization with the endogenous germ plasm (Fig. 3B,C, quantification in E). To narrow down the localization motif further, the N-terminal 20 amino acids were removed, deleting the N-terminal domain with aromatic amino acids (PLD1; aa21-30). Indeed, Buc 31-88 showed a slight reduction in germ plasm localization (Fig. 3D,E, Fig. S6). However, when we deleted additional ten C-terminal amino acids (Buc31-78), localization was restored to nearly wild-type frequency (Fig. 3D,E, Fig. S7B). By contrast, deleting four additional N-terminal amino acids (Buc35-78) almost completely abrogated localization (Fig. 3D,E). These results suggest that the first PLD does not seem to be necessary for localization.

To examine the role of the second domain with aromatic amino acids (Buc64-71), we generated internal deletions in Buc31-78. When we removed the second PLD (Δ64-71), no fluorescence could be detected in the embryos (Fig. S7C), suggesting that aa64-71 might affect protein stability or translation. Therefore, we analyzed two variants with five amino acid deletions within the second domain. Strikingly, removing parts of the second domain (BucΔ62-66 or BucΔ67-71) caused no clear reduction in germ plasm localization (Fig. 3D,E, Fig. S7D,E). In contrast, when we kept the second domain intact, but removed sequences C-terminal to the second domain (Buc31-71), the localization efficiency dropped to 15% (Fig. 3D,E, Fig. S7F). These results suggest that the two motifs enriched in aromatic amino acids are not required for positioning Buc to the four germ plasm aggregates in the zebrafish embryos.

As Buc forms clusters with the germ plasm in the proximity of the cleavage furrows, we aimed to identify the cellular structure that is essential for its anchorage. As our results show that BucLoc domain is sufficient for the recruitment of Buc to the germ plasm foci, we used this protein motif as a bait to identify its cellular binding partners directly by co-immunoprecipitation followed by mass spectrometry analysis. Embryos were injected at the one-cell stage with mRNA encoding BucLoc-GFP, lysed at the stage of the formation of germ plasm foci and immunoprecipitated using GFP-tag (Fig. 4A). Embryos injected with mRNA encoding GFP were used as a negative control, and transgenic embryos for full length Buc-GFP were used to control for mRNA overexpression.

In this analysis, we found 1817 protein candidates that potentially interact with full-length Buc and BucLoc but not with GFP. From those, 213 proteins were strongly enriched for BucLoc interaction (Fig. 4B, see Table S1 for the full list of candidates of the mass spectrometry) and therefore represent candidates for the subcellular network required for germ plasm localization. Among the candidates that were strongly enriched was myosin light chain (Fig. 4C), which is a subunit of the non-muscle myosin II (NMII) protein complex. Interestingly, phosphorylated NMII (p-NMII) colocalizes with germ plasm RNAs at the two- and four-cell stage in zebrafish embryos (Nair et al., 2013). To investigate whether p-NMII also colocalizes with Buc and could therefore play a role in germ plasm localization, we performed immunohistochemistry for Buc and p-NMII. Indeed, we found that Buc colocalizes with p-NMII in early stage IB oocytes (Fig. 4D) and during zebrafish embryogenesis (256 cell stage, Fig. 4E,F). This might suggest a role of NMII in germ plasm localization; however, further experiments might provide a better understanding of this.

As the NMII associates with various cellular structures (Liu et al., 2012; Nair et al., 2013; Vicente-Manzanares et al., 2009), we therefore screened the list of potential BucLoc interactors for a defined subcellular localization. We detected ZO2 and ZF-A89 as highly enriched in the pull-down assay; the latter is a homolog of Cldn-d (UniProtKB/Swiss-Prot record; description: Claudin-like protein ZF-A89) (Table S1). This suggest that we might have purified components of tight junctions. Claudins are adherence receptors essential for the formation of TJs, suggesting that Buc and germ plasm aggregate at TJs within the cleavage furrows.

To further confirm the anchorage of Buc at TJs, we used an antibody that specifically detects the zona occludens protein ZO1, which is shown to directly interact with Cldn proteins to mark the TJ for colocalization analysis. As controls, we used an antibody against E-Cadherin to label the adherens junction and Kif23 to label the midbody. Fascinatingly, Buc perfectly colocalized with the TJ marker ZO1 (Fig. 5D, Fig. S8C), whereas no overlap could be detected with E-Cadherin (Fig. 5B, Fig. S8A) or Kif23 (Fig. 5C, Fig. S8B). These data show that Buc localizes to the ZO1-positive foci at the cleavage furrows, suggesting colocalization of the germ plasm aggregates with TJs. To investigate the localization of Cldn-d, we generated an antibody. Co-labelling of eight-cell embryos with Buc and Cldn-d antibodies showed partial overlapping signals of both proteins (Fig. S9).

By contrast, at germ plasm-free cleavage furrows, we did not detect Buc and Cldn-d colocalization at the membrane. These data support a role of Cldn-d in tethering Buc at the germ plasm cleavage furrows of early zebrafish embryos. However, the absence of Cldn-d protein at those cleavage furrows without germ plasm could also be explained as differences in the temporal development of these furrows.

Taken together, our data show that Buc is associated with Cldn-d and ZO1-positive foci at the cleavage furrows supporting the idea that the TJ protein complex might be functionally involved in the association of Buc and the germ plasm. One limitation of our work is that we cannot show a triple staining of Buc together with Cldn-d and ZO1 due to technical reasons. We therefore suggest further studying the localization of these proteins with regard to other TJs and germ plasm markers.

To verify that the ZO1- and Buc-positive structures at the distal cleavage furrows of the eight-cell embryos are TJs, we used electron microscopy to search for characteristic TJ structures at the cleavage furrows of eight-cell stage embryos. Indeed, electron microscopy showed electron-dense membrane sections resembling TJ-like structures at the cleavage furrows where germ plasm is localized (Fig. 5E). In contrast, we did not find these structures at those cleavage furrows, where germ plasm is not accumulated (Fig. 5F). This finding supports the results of staining with the antiserum against ZO1, which also showed four spots in eight-cell stage embryos. The results of the electron microscopy show in addition, for the first time, that early zebrafish embryos already have TJ-like structures at the eight-cell stage. However, we believe the data we provided here are preliminary and limited; we think further electron microscopy is needed to show detailed structure of those embryonic TJs with respect to germ plasm.

Claudins are one family of receptors, which physically connect the TJs in the epithelial and endothelial tissues of vertebrates. Claudins are transmembrane proteins that bind to the PDZ domains of scaffolding zonula occludens (ZO) proteins through their cytoplasmic C-terminal YV (tyrosine-valine) motifs (Furuse et al., 2014; McCarthy et al., 2000). More than 50 claudins with restricted tissue expression patterns have been identified in teleost fishes (Kolosov et al., 2013).

The zebrafish genome encodes only five ZO proteins with numerous functions during early zebrafish embryogenesis (Kiener et al., 2007; Schwayer et al., 2019). Fascinatingly, only two claudins, Cldn-d and -e are maternally expressed according to the Zfin database (https://zfin.org/) (Fig. S10). Moreover, the role of the maternally expressed Xenopus Xcla, which is an ortholog of Cldn-d, was previously characterized to function in a different process from germ plasm localization, including a dominant-negative version of the receptor (Brizuela et al., 2001). Based on the colocalization of Buc with TJs and the interaction of ZO1 with Cldn-d, we addressed the hypothesis that Cldn-d could act as a membrane anchor for germ plasm.

To analyse a potential function of Cldn-d in germ plasm tethering, we injected cldn-d mRNA into one-cell zebrafish embryos transgenic for Buc-GFP to detect germ plasm localization in vivo. Compared with uninjected control embryos, the injection of Cldn-d mRNA led to a significantly higher number of Buc-GFP-positive spots at 2-3 hpf (Fig. 6A,B,E). To control the specificity of mRNA overexpression, we injected the same concentration of cldn-a mRNA, but did not detect a change of germ plasm spots, similar to uninjected controls (Fig. 6A,C,E).

We believe that our results of control and injected embryos showed no significant difference (Fig. 6G), suggesting that the expression of Cldn-d either caused a fragmentation of the existing four germ plasm spots into more but smaller aggregates or, alternatively, that Cldn-d caused the formation of additional germ plasm aggregates. Independent of the mechanism, these results indicate that Cldn-d might be involved in germ plasm tethering in the early zebrafish embryo. However, further investigation is needed to check the specificity of the function of Cldn-d. It might be worth checking whether Cldn-d overexpression induces ectopic TJ formation versus solely ectopic germ plasm spots, and whether the ectopic spots lead to extra germ cell formation.

The C-terminal amino acids tyrosine and valine are crucial for the interaction of claudins with ZO proteins (Itoh et al., 2014). We therefore generated a Cldn-d mutant lacking this interaction motif (C-terminal YV, named Cldn-dΔYV), which was previously shown to act as a dominant-negative form of claudin-d (Brizuela et al., 2001). Notably, cldn-dΔYV-injected embryos showed a significantly reduced number of germ plasm spots in comparison with uninjected embryos (Fig. 6A,D,E). However, cldn-dΔYV-injected embryos displayed severe developmental defects, in which the cells did not attach to each other, suggesting that the dominant-negative receptor might also disrupt TJs formed during later embryogenesis (Fig. 6D,D′).

To exclude the possibility that the reduced number of germ plasm foci in cldn-dΔYV-injected embryos is a secondary result caused by a defect in cell attachment, we targeted its expression to two blastomeres in a 16-cell embryo. Injection of cldn-dΔYV mRNA into 16-cell stage embryos still reduces the number of germ plasm spots. At this stage, the junctions are matured and the germ plasm-containing cells can easily be distinguished from somatic cells, as they hold the central position in the marginal row of four blastomeres (‘middle blastomeres’). We injected cldn-dΔYV into two middle blastomeres surrounding one germ plasm spot (Fig. 7A) using uninjected and wild-type cldn-d-injected embryos as controls. The number of Buc spots was counted immediately after injection (16-cell stage) and then followed up at regular time intervals (see Table S2). In this assay, embryos developed normally and did not show developmental defects (Fig. 7B,C). Interestingly, we still observed a significant reduction in the number of germ plasm spots in cldn-dΔYV-injected embryos (Fig. 7B-D) compared with uninjected and cldn-d-injected controls (Fig. 7D). More than 35% of cldn-dΔYV-injected embryos lost a germ plasm spot, whereas only 6% of the embryos injected with cldn-d showed germ plasm spot reduction (Fig. 7D, Table S1). These results support our model that TJs might be involved in the localization of germ plasm at the cleavage furrows and that the Cldn-d receptor might be component of TJs in the early zebrafish embryo (Fig. 8).

Our data show that the machinery of germ plasm recruitment and anchorage is different between vertebrates and invertebrates. We identified that the N-terminal BucLoc domain (aa11-88) is necessary and sufficient for the localization of Buc to the four germ plasm aggregates at the cleavage furrow at the eight-cell stage embryo. Our colocalization and protein-protein interaction data from the immunoprecipitations suggest that Buc, together with other germ plasm components, are linked to de novo-forming TJs at cleavage furrows.

Our results showed: (1) that myosin light chain co-immunoprecipitated with Buc and that p-NMII colocalizes with Buc protein, suggesting that germ plasm might become anchored to one of the cellular structures through NMII (Fig. 4); (2) co-immunoprecipitations of Buc co-purified ZO and claudin (Fig. 4); (3) that germ plasm colocalizes with the TJ protein ZO1 and with claudin-d (Fig. 5, Figs S8-S10); (4) the presence of TJ-like structures by electron microscopy at the cleavage furrows in the 8-cell zebrafish embryo (Fig. 5); (5) and that cldn-d injection caused the formation of a higher number of germ plasm spots, whereas Cldn-d with a mutated interaction motif for ZO proteins (C-terminal YV motif) functions as a potential dominant negative, resulting in fewer germ plasm spots (Figs 6, 7). Taken together, these results might support the model that newly forming TJs at the cleavage furrows represent the anchorage hub for the germ plasm in zebrafish.

Invention of multicellularity requires cell adhesion and a more advanced form of reproduction. With our finding, it will be possible to address whether germ plasm localization at TJs was already occurring at the origin of Metazoa or whether it is a derived mechanism acquired during vertebrate evolution. The isolated BucLoc motif does not show homology with known protein domains, making it impossible to trace its biochemical function. However, it suggests a conserved localization mechanism among vertebrates.

Both zebrafish Buc and Xenopus xVelo are positioned at the cleavage furrow, whereas Drosophila sOsk is not targeted by the vertebrate localization system. Despite the functional equivalence of Buc and Osk shown previously (Krishnakumar et al., 2018), we provide evidence here that the cellular mechanisms of germ plasm anchoring differ between vertebrates and invertebrates. The germ plasm aggregation activity of these germ plasm nucleators appears to be conserved, whereas the mechanisms of their anchoring appear to be different. The observed different mechanisms of germ plasm anchoring are therefore probably consistent with the different morphology of early embryos. Whereas in Drosophila the attachment of germ plasm to the apical surface within the syncytial early embryo guarantees the transfer of germ plasm into the budding PGC, we think that in zebrafish the anchoring of germ plasm to the TJs of cleavage furrows is a prerequisite for the asymmetric distribution of germ plasm during subsequent cell divisions; this allows the inheritance of the four germ plasm aggregates into four PGCs. However, further experiments disrupting TJs in early embryos might be needed to prove this model.

Anchoring the germ plasm to the TJs is different from anchoring to the posterior cell cortex in Drosophila oocytes and embryos long before cellularization take place. In Drosophila embryos, the posterior localization of germ plasm is essential for its recruitment into the budding PGCs at the posterior pole. However, in zebrafish embryos, the role of germ plasm anchorage is different. Germ plasm aggregation is detectable from the four-cell stage onwards, and depends on the furrow-associated microtubule-array (FMA) (Jesuthasan, 1998; Pelegri et al., 1999) and on the activity of Rho1 and ROCK (Miranda-Rodríguez et al., 2017). However, as the FMA disassembles after the third cleavage, germ plasm needs to be anchored at cleavage furrows. Thus, anchoring to the TJ could represent a functional model that can explain the asymmetric distribution of germ plasm during subsequent cell divisions, with only one of the daughter cells inheriting the TJ and germ plasm at a time. Such an asymmetric distribution of germplasm would explain that starting from the four germ plasm sites at the four-cell stage, only four PGCs arise. It is unclear how germ plasm becomes asymmetrically distributed between the daughter cells and limits the number of aggregates, although several mechanisms are possible: (1) removal of the germ granules from the cytoplasmic pool of germ granules during subsequent cleavage divisions; (ii) stabilization of the aggregate and prevention of its fragmentation; and/or (3) facilitation of compaction of the aggregate by phase separation. From the 32-cell stage onwards, germ plasm is associated with a cup-like structure near one spindle pole, which controls asymmetric segregation until the sphere stage (Braat et al., 1999; Knaut et al., 2000; Yoon et al., 1997). Only after the mid-blastula transition (MBT) if germ plasm localized in perinuclear clusters, allowing symmetric distribution during PGC divisions (Strasser et al., 2008) that form four PGC clusters (Kane and Kimmel, 1993; Dosch, 2015; Knaut et al., 2000; Wolke et al., 2002).

We believe that our results show a specific function of Cldn-d for germ plasm anchorage, as injection of Cldn-d caused a significant increase in the number of germ plasm spots, whereas Cldn-a had no effect. The specific role of Cldn-d is supported by the fact that co-immunoprecipitation experiments revealed a specific interaction between Buc and Cldn-d but not with Cldn-a. This differential biological activity suggests a particular capability of Cldn-d to function as a specific anchorage receptor for germ plasm aggregates. Fascinatingly, injection of cldn-d had a similar activity on forming extra germ plasm spots compared with the injection of Buc (Bontems et al., 2009). It is unclear yet whether the observed higher number of germ plasm spots represent additional germ plasm spots or a fragmentation of the germ plasm spots already present. Our results suggest that the maternal load of Cldn-d would be limiting and only sufficient to form four spots. Indeed, the loss of spots after injection of dominant-negative Cldn-d seems to support this hypothesis. A specific function of Cldn-d in the anchorage of germ plasm aggregates could explain the release of the germ plasm from TJs by the exchange or dilution of the maternal Cldn-d with other claudins. Alternatively, the release of the germ plasm could be achieved by the post-translational modifications of Buc or components of the TJs.

The recruitment of Buc and Xenopus Velo1 to the four germ plasm aggregates in zebrafish suggest also a potential role for Velo1 in germ plasm anchorage in Xenopus. A combination of molecular markers, including ZO1, a biotin-based permeability assay and electron microscopy showed that TJs are formed from maternally derived proteins as early as the two-cell stage of Xenopus at the boundary between the pre-existing apical membrane and the newly formed basolateral membrane (Fesenko et al., 2000; Merzdorf et al., 1998). Moreover, live imaging studies with Dria-EGFP transgenic Xenopus showed that germ plasm aggregates and enters the embryo along cleavage furrows, which are found to be enriched with membranes until embryonic stage 10 after MBT, when germ plasm lose their contact with membranes and become cytoplasmic, presumably associated with perinuclear structures (Taguchi et al., 2012). Based on these similarities it will be valuable to investigate whether the TJ-mediated germ plasm anchoring is also used in Xenopus PGCs.

Sequence analysis of Buc did not reveal any characterized domain within the protein (Bontems et al., 2009; Krishnakumar et al., 2018). Sequence comparison with 15 related Buc proteins revealed a conserved 100 amino acid N-terminus, which was named BUVE motif (Buc-Velo) (Bontems et al., 2009). The BUVE domain was shown to be essential for the formation of the amyloid-like aggregates in the BB in Xenopus oocytes (Boke et al., 2016). The BUVE domain contains potential prion like domains (PLDs) (Alberti et al., 2009), which were shown to be essential for the aggregation process, based on the fact that the replacement of critical residues with charged amino acids inhibited the aggregation. These results might suggest that the BUVE domain of Velo1 and Buc is required for amyloid-like germ plasm aggregation in the BB. However, Velo1 variants in which the potential PLDs were replaced with unrelated PLDs were inactive, whereas the replacement with the related sequences from zebrafish Buc were active, revealing a sequence specificity (Boke et al., 2016). Surprisingly intrinsic disorder prediction of Buc showed that the N terminus (aa1-150) is the largest ordered sequence in Buc (Krishnakumar et al., 2018). Nevertheless, we tested for the role of the predicted PLDs within the N-terminus when we identified the region between aa11 and aa88 to be essential and sufficient for the localization of Buc to the four germ plasm spots at the cleavage furrows (Fig. 2). Even though current results cannot rule out the possibility of the exogenous Buc11-88 peptide being incorporated into endogenous Buc protein aggregates, detailed mapping revealed that neither of the two potential PLDs within this sequence is essential for its localization, but short amino acid segments positioned C-terminal from them are essential (Fig. 3). But, this result does not rule out a functional role for the PLDs, as the identified additional regions may be required for proper positioning of the PLDs. However, co-immunoprecipitation and mass spectrometry analyses showed that the domain between aa11 and aa88 interacts with about 213 peptides, including myosin light chain and Cldn-d (Fig. 4), suggesting a role for the BucLoc domain as protein-protein interaction module. However, 213 peptides are an unexpected high number of interactions and includes probably a number of indirect interactions. Future experiments will identify the direct interaction partners of Buc and help to unravel the precise mechanism of Buc anchorage.

Increasing evidence suggest that germ granules in many different organisms are formed by phase separation. Germ plasm consists of spherical units of protein RNA aggregates that show a highly dynamic exchange with the surrounding cytoplasm (recently reviewed by Dodson and Kennedy, 2020; So et al., 2021). Indeed, the BucLoc motif has previously been shown to play a crucial role in aggregating the BB in the Xenopus oocyte, which is probably the largest biomolecular condensate in the animal kingdom (Boke et al., 2016). However, our results show that the prion-like domains in the BucLoc motif, which control BB assembly, are not required for germ plasm anchoring in the embryo. Moreover, germ plasm aggregates can be detected at the four-cell stage prior to their anchorage to the de novo developing TJ, suggesting that germ plasm aggregates are probably formed by phase separation independent of their subsequent anchorage.

Interestingly, ZO proteins also induce the assembly of liquid-like condensates (reviewed by Canever et al., 2020; Citi, 2020). The condensation of ZO proteins in cell culture and zebrafish embryos induces the assembly of TJs, revealing unexpected activity in the cytoplasm to control the formation of TJs (Beutel et al., 2019; Schwayer et al., 2019). Our finding that Buc and ZO1 colocalize raises the issue of whether Buc indeed autonomously induces the formation of condensates or whether this activity is mediated by ZO1. However, we have previously shown Buc condensates in HEK293 cells, which do not form TJs, supporting the autonomous phase separation activity of Buc (Krishnakumar et al., 2018).

In conclusion, we have found that vertebrates and invertebrates use different germ plasm anchorage mechanisms, with evolutionary conservation between vertebrates. We have also discovered that TJs anchor germ plasm during early zebrafish embryogenesis and that microinjection of cldn-d induced extra germ plasm spots. Therefore, we think that germ plasm in zebrafish is probably anchored to the TJs via Cldn-d receptor protein (Fig. 8).

Zebrafish (Danio rerio) used in this study were AB*TLF (wild type) and Buc-GFP transgenic zebrafish lines (Riemer et al., 2015). Fish were raised and maintained according to the guidelines from Westerfield (2000) and according to regulations from the Georg-August University Goettingen.

Previously synthesized capped RNA was diluted with 0.1 M KCl and 0.05% Phenol Red (Sigma Aldrich). 2 nl of RNA was injected into one-cell stage embryos using a PV820 (World Precision Instruments). Injected embryos were incubated in E3 medium at 28°C until they reached the developmental stage of phenotype evaluation.

To study whether non-functional Cldn-d has an influence on matured TJs, we conducted cldn-dΔYV injections in 16-cell embryos. In this assay, we injected the RNA directly into two cells next to a germ plasm-localizing tight junction. As a control we used uninjected and cldn-d RNA-injected embryos. The number of Buc spots was counted immediately after injection and then followed up at regular time periods. The spots were counted from z-stacks at ∼2 hpf using an ImageJ plug-in for each z-section if the signal-to-background ratio was beyond a specific threshold but with size cut offs to exclude pixel artefacts. A detailed description of the injection procedure at the 16-cell stage has been previously published (Bontems et al., 2009; Krishnakumar et al., 2018).

Flies were kept and crossed at room temperature or 25°C. To collect embryos, the flies were kept in cages with apple juice agar plates at 25°C. Experiments were approved by the Lower Saxony State Office for Consumer Protection and Food Safety (AZ14/1681). The pUASp bcd3′UTR plasmid expressing sOsk (Tanaka and Nakamura, 2008) was used to replace the sosk ORF with Buc ORF-GFP. A germline-specific mat-Gal4VP16 driver was used to express UASp-based transgenes in oogenesis. Antibody staining and fluorescent in situ hybridization were performed as described previously (Pflanz et al., 2015). The antibodies used were anti-PY20 (1/500, Biomol; p-Tyrosine, bml-sa240), chicken anti-GFP (1/1000, Synaptic Systems, 132 006), anti-Osk (kindly provided by Anne Ephrussi, European Molecular Biology Laboratory, Heidelberg, Germany) and anti-Vasa (1/5000; Pflanz et al., 2015). Anti-mouse, anti-rabbit and anti-chicken antibodies coupled to Alexa 488, 568 or 647 were used as secondary antibodies (Invitrogen, A-11001, A-11004, A-21235, A-11008, A-11011, A-21244, A-11039, A78950, A78952; 1/1000). Embryos were embedded in DPX to provide clearing and to protect from bleaching.

Co-immunoprecipitation (co-IP) was performed to identify the Buc protein interactome. BucLoc interactors were isolated directly from the embryos at the stage where Buc is localized to the putative primordial germ cells to determine the protein network involved in Buc localization. RNA encoding BucLoc-eGFP was introduced into embryos at the one-cell stage, lysed at a higher stage and immunoprecipitated using the GFP tag. As a negative control, embryos injected with RNA encoding eGFP were employed. Embryos from the transgenic Buc-GFP line were used as a positive control for overexpression artefacts. Mass spectrometry was used to analyse the co-immunoprecipitated proteins of the three samples. Each sample was prepared from 500 deyolked high stage embryos after homogenization on ice in lysis buffer [10 mM Tris (pH 7.5), 150 mM NaCl, 0.5 mM EDTA, 0.5% NP-40 and 1× complete protease inhibitor cocktail (Roche)]. The supernatant was subsequently used for the co-IP using a GFP-binding protein coupled to magnetic beads (GFP-Trap_M; ChromoTek) following the manufacturer's instructions. After pulling down, the magnetic beads and their bound proteins were either incubated with 2× SDS loading buffer for 5 min at 96°C and analysed via SDS-PAGE and western blotting or sent for mass spectrometry (Core Facility of Proteome Analysis), as described previously (Krishnakumar et al., 2018).

The isolation of the Buc localization domain is a useful approach for determining the mechanism of Buc localization. This tool may be used to investigate which proteins BucLoc interacts with in order to properly locate Buc to the germ plasm during zebrafish development. The molecular network involved in Buc localization is identified, which provides information of the mechanism that underpins the process of localization. In total, 3464 protein candidates interact. From those, 1817 candidates were identified that interacted with both Buc-GFP and BucLoc-GFP. We were not interested in every candidate with an interaction with Buc-GFP, as they might interact with any other region outside BucLoc. Therefore, we applied a set of criteria to identify candidates that significantly interacted with BucLoc. First, any peptide below a background threshold of five in BucLoc-GFP was considered not to be significant and was sorted out. Furthermore, only proteins with counts in BucLoc-GFP that were at least twice as high as in the negative control GFP were considered as significant. To further reduce overexpression artefacts, enrichment in the positive control and in the sample had to be within a magnitude of ±4-fold. Applying these selection criteria, the number of potential BucLoc interaction proteins could be restricted to 213 interaction candidates (see Table S4 for the full list of mass spectrometry candidates).

Embryos were fixed and stained as previously described (Riemer et al., 2015) with the antibodies listed in Table S5.

Living embryos of the transgenic Buc-GFP line were imaged to analyse the localization of Buc-GFP. For imaging with the stereo microscope SteREO Lumar.V12 (Zeiss), embryos were manually dechorionated and mounted in 1.5% agarose-coated dishes filled with 1× E3 medium. Images were analysed using the software Axio Vision Rel. 4.8 (Zeiss). For imaging with the LSM780 confocal microscope (Zeiss), embryos were mechanically dechorionated, placed in a fluorodish (World Precision Instruments) with a handmade grid, covered with 1× E3 and imaged from below. Images were analysed using the ZEN 2011 software (Zeiss), as described previously (Riemer et al., 2015). Quantification of confocal images was performed with ImageJ software.

Electron microscopy was performed at the facility for transmission electron microscopy (Max Planck Institute for Biophysical Chemistry, Göttingen). Embryos were fixed by high-pressure freezing method and imaged with a Philips CM120 electron microscope using a TemCam 224A slow scan CCD camera, as described previously (Kanagaraj et al., 2016).

Western blotting was performed to detect the specificity of Buc antibody as described previously (Krishnakumar et al., 2018). Fluorescent signal was detected with Li-Cor Odyssey CLx Infrared Imaging system (Li-Cor) and analysed with the Image Studio Software (Li-Cor). Western blots with in vitro-translated proteins confirmed that the Buc antibody did not cross-react with GFP or other proteins and thus specifically highlights endogenous germ plasm (Fig. S1). Details of the antibodies used are provided in Table S6.

Proteins were synthesized with the TnT SP6 Quick Coupled Transcription/Translation System (Promega).

The templates of all the constructs that were used in this study were amplified from reverse-transcribed cDNA, which was made from total ovarian RNA. Constructs were cloned with either restriction digestion or gateway cloning (Table S3).

Pairwise sequence alignment was used to compare protein sequences, using the Needleman-Wunsch algorithm with the EMBL-EBI alignment software EMBOSS Needle (McWilliam et al., 2013).

Fold amyloid (Fernandez-Escamilla et al., 2004), APPNN (Família et al., 2015), FISH amyloid (Gasior and Kotulska, 2014) and aggrescan (Conchillo-Solé et al., 2007) algorithms were used to predict PLDs in BucLoc.

Overlaps in protein interactions between each co-IP sample were analysed using a Venn diagram generator (http://jura.wi.mit.edu/bioc/tools/venn3way/index.php). The Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/) was used to classify the BucLoc-GFP interaction candidates in collaboration with Dr Thomas Lingner to see whether a large number of interaction candidates are involved in the same biochemical pathway. Approximately 62% of the proteins could be assigned to various molecular pathways, including as signalling (11%), splicing (2%), adherens and tight junctions (2%), and mRNA transport, surveillance and destruction (8%). Nevertheless, the co-IP withBucLoc identified 213 interaction candidates involved in various molecular pathways that might play a role in the localization of Buc during early embryogenesis (Table S1).

All the statistical analyses of the experiments have been carried out using Microsoft Excel and the Prism software (GraphPad Software). Error bars indicate the standard deviation of averages. For each injection experiment, at least three independent replicates were used.

We thank Prof. E. A. Wimmer for providing the facilities to perform this research and G. Kracht for technical assistance. Authors also thank Dr Sabine Klein for the useful input on the manuscript. Data from the PhD theses of Stephan Riemer, Alexander Goloborodko, Nadia Rostam, Roshan Priyarangana Perera and Pritesh Krishnakumar (Georg August University Göttingen, 2014-2020) are presented in this paper.