Ephrin-Eph signalling drives the asymmetric division of notochord/neural precursors in Ciona embryos

Embryogenesis is a process during which a variety of cell types are produced in a spatially and temporally coordinated manner. Certain cell divisions during embryogenesis, termed asymmetric cell divisions, result in the production of two daughter cells specified to adopt distinct developmental fates, thus contributing to the generation of cell diversity(Horvitz and Herskowitz, 1992; Knoblich, 2001; Roegiers and Jan, 2004; Betschinger and Knoblich,2004). Although many such divisions have been described, the underlying molecular mechanisms are currently understood in only a few cases analysed in Drosophila melanogaster and Caenorhabditis elegans. For the fly sensory organ precursors (SOPs) to generate two distinct siblings, an asymmetrically distributed protein called Numb plays a crucial role by attenuating Notch signals in one of the daughter cells and thus creating differential transduction of the signalling pathway between the daughter cells (Rhyu et al.,1994; Guo et al.,1996). A different mechanism controls the asymmetric cell division of the Drosophila larval neuroblasts, in which the transcription factor Prospero is asymmetrically inherited between the daughter cells and acts as a binary switch between the self-renewal and differentiation fates adopted by the daughters (Hirata et al.,1995; Knoblich et al.,1995; Choksi et al.,2006). In both these cases, segregation of these factors in mother cells appears to take place intrinsically(Gho and Schweisguth, 1998; Siegrist and Doe, 2006). In another example, in C. elegans, during the cell division of an embryonic cell named EMS, extrinsic cues, provided by a Wnt ligand, govern the binary choice between endoderm and mesoderm fates of the resultant daughter cells (Thorpe et al., 1997; Rocheleau et al., 1997). This extrinsic and directional Wnt signal, acting on the mother cell, results in a differential nuclear accumulation of β-catenin between the daughter cells, and thus to their differential fate specification(Goldstein et al., 2006; Nakamura et al., 2005; Lin et al., 1998). These examples reveal that a variety of molecular strategies can be used to generate two distinct cell fates following an asymmetric cell division.

One model system that could be particularly amenable to the study of asymmetric cell division is the invertebrate chordate embryos of ascidians. Ascidian embryogenesis is characterised by a fixed cell-cleavage pattern and a small number of cells, which undergo fate restriction during the early cleavage stages. These characteristics have led to a complete documentation of cell lineages up to the onset of gastrulation(Nishida, 1987). The cell lineages have revealed that a number of cell divisions are coupled with the differential segregation of cell fates between the sibling cells. We are studying one such example, the generation of notochord and neural precursors,from two pairs of mother cells named A6.2 and A6.4, located in an anterior-marginal position of the 32-cell-stage embryo(Fig. 1). Following cleavage of these cells, the daughter cells that reside on the animal-pole side become neural precursors and give rise to a part of the larval central nervous system, whereas daughter cells arising on the vegetal-pole side become precursors of the larval notochord (Fig. 1). It has been shown that the fibroblast growth factor (FGF)pathway - involving the small GTPase Ras, MAPK/ERK kinase (MEK) and extracellular-signal-regulated kinase (ERK) - plays a crucial role during this differential fate specification (Nakatani and Nishida, 1997; Kim and Nishida, 2001). When FGF-ERK1/2 signals are inhibited, each mother cell generates two neural precursors, instead of one notochord and one neural precursor (Minokawa et al.,2001; Yasuo and Hudson,2007). Conversely, if isolated mother cells are treated with exogenous FGF, both daughters adopt a notochord fate(Minokawa et al., 2001; Nakatani et al., 1996). Consistent with these results, activation of ERK1/2 is detected in notochord/neural mother cells but remains active only in the notochord precursors following cell division (Yasuo and Hudson, 2007). Thus, differential ERK activation between these daughter cells, which is the first known sign of an asymmetry between them, is the determinative event driving their binary cell fate choice. It has been proposed that the differential activation of ERK1/2, and thus differential fate specification of the two daughter cells, might be explained by a directional FGF signal coming from the vegetal region (neighbouring endoderm precursors) (Minokawa et al.,2001). However, transcripts of FGF9/16/20, encoding the FGF ligand responsible for ERK1/2 activation and early notochord induction,are detected widely in vegetal cells, including in the notochord/neural mother cells themselves, suggesting that it is unlikely to provide an asymmetric positional cue (Imai et al.,2002; Bertrand et al.,2003).

In this study, we first present evidence that the notochord/neural mother cell acquires, during its cell cycle, the autonomous capacity to divide asymmetrically and give rise to one notochord and one neural precursor. The acquisition of this functional polarity by the mother cell depends on cellular interactions between the mother cell and neighbouring epidermal precursors,positioned on the animal side. The molecular nature of this signal is a GPI-membrane-bound form of ephrin ligand, signalling via the Eph receptor(ephrin-Eph signalling). Disruption of contacts with the signalling cells or inhibition of the ephrin-Eph signals results in the symmetric division of the mother cell and the generation of two notochord precursors. Finally, we demonstrate that the ephrin-mediated signal acts to attenuate ERK activation in the neural precursor, and is thus responsible for creating the differential ERK activation between sibling cells.

Adult Ciona intestinalis (Ciona) were purchased from the Roscoff Marine Biological Station (Roscoff, France). Embryo culture,microinjection of unfertilised eggs, UO126-treatment, and isolation and recombination of blastomeres have been described previously(Hudson et al., 2003; Hudson and Lemaire, 2001). All data presented in this study were collected from at least two independent experiments.

In situ hybridisation of Ciona embryos was carried out as described previously (Hudson and Yasuo,2006). Dig-probes were synthesised from the following cDNA clones: Ci-Bra (Corbo et al.,1997); Ci-ETR (Hudson et al., 2003); Ci-ephrin-Ad (ciad008n17 from the Kyoto Gene Collection Plates) (Satou et al.,2002). For immunohistochemistry, embryos were fixed in 4%paraformaldehyde/0.2% glutaraldehyde in artificial sea water for 30 minutes at room temperature, washed with PBS, 0.1% Triton X-100 (PBST), blocked in PBST,5% goat serum, 1% Roche Blocking Reagent, and incubated overnight in anti-dpERK1/2 (Sigma, 1:1000) at 4°C. After washing in PBST, embryos were incubated overnight in anti-mouse-HRP (Molecular Probes, 1:100) at 4°C and then washed prior to signal detection, which was performed using TSA Plus Fluorescence Systems (PerkinElmer) with cyanine 5 as fluochrome. Western blot analyses were carried out following standard protocols with anti-dpERK1/2(Cell Signaling Technology) and anti-rabbit-HRP (Jackson ImmunoResearch) used at a dilution of 1:350 and 1:3000, respectively. A total of 20 embryos dissolved in lysis buffer were used for each analysis.

Ephrin-Ad-MO (5′-GGTAGTAGGTAAATTGAGTTGCCAT-3′) was purchased from GeneTools LLC and injected at a concentration of 0.5 mM. A dominant-negative form of Eph3 was constructed using a cDNA clone, cieg009e01,from the Kyoto Gene Collection Plates. cDNA fragments corresponding to the entire extracellular domain, transmembrane domain and a part of intracellular domain were PCR-amplified. For ephrin-Ad RNA, cDNA fragments corresponding to the entire ORF was PCR-amplified from ciad008n17. Amplified cDNA fragments were subcloned in pRN3 (Lemaire et al.,1995) and used to synthesise RNAs. dnEph3 and ephrin-Ad RNAs were injected at a concentration of 0.125 and 0.25 mg/ml, respectively.

Because the expression pattern of FGF9/16/20 does not explain differential ERK1/2 activation between notochord and neural precursors (see above), we speculated that an inhibitory mechanism may exist to attenuate ERK1/2 activation in the neural precursors. To test for inhibitory influences from neighbouring cells, we isolated A4.1 blastomeres (green in Fig. 1) - the progenitors of notochord and neural precursors - from eight-cell-stage embryos, cultured them until control embryos reached the 110-cell stage and analysed them for the number of notochord and neural precursors(Fig. 2). The presence of notochord and neural precursors was monitored by expression of Ci-Bra(Corbo et al., 1997) and Ci-ETR (Hudson et al.,2003), respectively. The mean number of cells expressing Ci-Bra in the A4.1-derived partial embryos was significantly higher than that predicted from the cell lineage (6.4 versus 4; Fig. 1 and Fig. 2B,K). By contrast, Ci-ETR-positive cells rarely formed (average of 0.8 cells), despite the fact that the cell lineage predicts four neural precursors(Fig. 1 and Fig. 2G,K). Thus, when A4.1 progenitor cells are isolated, additional notochord precursors form at the expense of neural precursors, indicating that interactions with other cells are required for correct neural fate specification.

To address the timing of these cellular interactions, we performed similar isolation experiments on A6.2 and A6.4 mother cells at the start or at the end of their cell cycle. According to the cell lineage, each of these mother cells should generate two notochord and two neural precursors at the 110-cell stage(Fig. 1). When isolated shortly before their division (at the late 32-cell stage), these cells showed equivalent fate attribution, expressing Ci-Bra and Ci-ETR in an average of 2.0 and 1.8 cells, respectively(Fig. 2E,J,L and Fig. 1). By contrast, when they were isolated early in the cell cycle (at the 24-cell stage), the resultant partial embryos had an average of 3.2 cells positive for Ci-Bra and 1.3 cells positive for Ci-ETR(Fig. 2D,I,L). These results show that the cellular interactions required for the suppression of notochord fate and promotion of neural fate are still occurring at the 24-cell stage but that they have terminated before the A6.2 and A6.4 mother cells begin cytokinesis. Thus, the mother cells acquire the autonomous capacity to divide asymmetrically during their cell cycle. These results indicate, rather unexpectedly, that the notochord/neural fate decision is not dependent on differential signalling to the daughter cells but on the prior polarisation of the mother cell.

We hypothesised that the inhibitory signals implicated in suppressing notochord fate in the neural lineage may arise from the anterior-animal lineage derived from a4.2 (blue in Fig. 1), because these cells are in contact with the A6.2 and A6.4 mother cells on the side on which the neural precursors will form. Using the recently reported bioinformatics tool, 3D Virtual Embryo version 1.0(Tassy et al., 2006), we identified contacts between animal cells and notochord-neural mother cells at the 24-cell stage. A6.2 contacts a5.3 with a contact area of 23.8% relative to the entire cell surface of A6.2, whereas the A6.4 blastomere contacts a5.3(10.9%), a5.4 (16.4%) and b5.3 (15.5%) with a total area of contact of 42.8%. a5.3 and a5.4 cells originate from the a4.2 anterior-animal cell of the eight-cell-stage embryo. We demonstrated that animal cells are indeed the source of the notochord-inhibitory (neural inducing) signals by cell co-isolation and recombination experiments. In A4.1 cells co-isolated with a4.2 cells, the average number of Ci-Bra-positive cells at the equivalent of the 110-cell stage was reduced to 3.2 cells(Fig. 2C,K) compared with 6.4 cells in A4.1-derived partial embryos. Concomitantly, expression of Ci-ETR was recovered with an average of 3.4 positive cells compared with 0.8 cells in A4.1-derived partial embryos(Fig. 2H,K). Similar results were obtained when A4.1 and a4.2 cells were isolated separately and then recombined (Fig. 2K) or when A4.1 was recombined with b4.2 (mean of 3.1 cells positive for Ci-ETRexpression, n=16), but not when A4.1 was recombined with the vegetal cells A4.1 or B4.1 (mean of 0.0 and 0.0 cells positive for Ci-ETRexpression; n=15 and n=10, respectively). These results indicate that cell contact between mother cells and animal cells is required to promote neural fate and suppress notochord fate in the neural precursors.

We next sought to identify the molecular nature of the animal-derived signals. Candidate molecules are predicted to be expressed in the animal cells at around the 8- to 32-cell stages and to be secreted or membrane-bound ligands capable of modulating the FGF signalling pathway (see Introduction). The ephrin-Eph signalling molecules provide ideal candidates. Activation of Eph receptor tyrosine kinases (RTKs) has been shown to inhibit the Ras-ERK1/2 pathway in a range of cell lines (Elowe et al., 2001; Miao et al.,2001). In particular, it has been shown that activation of EphA leads to a rapid attenuation of ERK1/2 activity induced by other RTKs,including those for platelet-derived growth factor (PDGF), epidermal growth factor (EGF) and vascular endothelial growth factor (VEGF)(Miao et al., 2001). Eph receptors are activated by ephrin ligands. As ephrins are membrane-bound proteins, ephrin-Eph signals are mediated in a cell-contact-dependent manner. We first analysed the expression profiles, during early embryogenesis, of all five ephrin genes annotated in the Ciona genome(Satou et al., 2003). Among the profiles obtained, the only appropriate spatial and temporal expression was observed for Ci-ephrin-Ad, which encodes a GPI-anchored form of ephrin ligand. Ci-ephrin-Ad starts to be expressed in all animal cells at the 16-cell stage and its expression is maintained through the 24-and 32-cell stages (Fig. 3A-D).

To address the potential role of Ci-ephrin-Ad in directing the notochord-neural binary cell fate choice, we first overexpressed it by injecting in vitro synthesised RNA into Ciona eggs. Injected eggs were subsequently fertilised and cultured to the 64- and 110-cell stages for the analysis of Ci-Bra and Ci-ETR expression, respectively. Consistent with the idea that ephrin-Eph signals inhibit notochord fates,expression of Ci-Bra was blocked following Ci-ephrin-Ad injection(Fig. 4A-C). Furthermore,notochord precursors now expressed the neural marker Ci-ETR,suggesting that both of the daughter cells of A6.2 and A6.4 cells had adopted a neural fate (Fig. 4E,F,G). This effect of Ci-ephrin-Ad overexpression phenocopied that of the inhibition of FGF-MEK signals (Minokawa et al.,2001; Yasuo and Hudson,2007), suggesting that activation of ephrin-Eph signals can inhibit the FGF-Ras-MEK-ERK pathway in ascidian embryonic cells. This was directly addressed by using an antibody recognising the activated,diphosphorylated form of ERK1/2, for which there is a single representative in the Ciona genome (Satou et al.,2003). Strikingly, activation of ERK1/2 in Ci-ephrin-Ad-injected embryos was blocked as efficiently as in sibling embryos treated with UO126, a pharmacological inhibitor for MEK (Favata et al., 1998) (Fig. 5A). Furthermore, immunohistochemistry on whole embryos showed that ERK1/2 activation was inhibited throughout the embryo following Ci-ephrin-Ad injection (Fig. 5C).

Ephrin-Eph signals are known to act bi-directionally; receptor signalling is referred to as forward signalling and ephrin-ligand signalling as reverse signalling (Cowan and Henkemeyer,2002; Murai and Pasquale,2003). We thus asked whether the effect of Ci-ephrin-Ad on ERK1/2 activity was mediated via forward or reverse signalling. Transcripts of two of the five Ciona Eph receptor genes, namely Ci-Eph2 and Ci-Eph3, showed a broad distribution during the early cleavage stages. We generated a dominant-negative form of Ci-Eph3 (dnEph3) by removing most of its intracellular domain. Co-injection of RNAs for dnEph3 and Ci-ephrin-Ad completely suppressed the ephrin-mediated inhibition of ERK1/2 activation (Fig. 5A),demonstrating that Ci-ephrin-Ad attenuates ERK1/2 activation via Eph receptors.

To address whether ephrin-Ad-Eph signalling is required in the embryo to suppress notochord fate and promote neural fate in neural precursors, we inhibited it by antisense morpholino (MO)-mediated knockdown of Ci-ephrin-Ad,and by the dominant-negative form of Eph3. In embryos injected with either ephrin-Ad-MO or dnEph3 RNA, expression of Ci-ETR was repressed,whereas Ci-Bra was expressed ectopically in the neural precursors(Fig. 4D,H). Thus, in the absence of ephrin-Ad-Eph signalling, the neural precursors transform into a notochord fate. Concomitantly, an ectopic activation of ERK1/2 was observed immunohistochemically in the neural lineages of dnEph3-injected embryos(Fig. 5D). These results show that ephrin-Ad signalling via Eph receptors is required to inhibit ERK, and thus to repress notochord fate and to induce neural fate in the neural precursors. Thus, ephrin-Ad is a strong candidate to be the animal-derived signal identified from our embryological studies.

The studies described above identified ephrin-Ad as a major factor controlling the binary fate choice of the notochord and neural precursors in the context of whole embryos, but did not reveal whether it does so by acting on the mother cell or on the neural-fated daughter cell. To address this issue, we isolated notochord/neural mother cells from uninjected or ephrin-Ad-MO-injected embryos at the late 32-cell stage, and monitored Ci-Bra and Ci-ETR expression at the equivalent of the early gastrula stage (Fig. 6A). Similar to the result shown in Fig. 2L, mother cells isolated from control late 32-cell-stage embryos developed into four-cell partial embryos expressing Ci-Bra and Ci-ETR in average of 1.6 and 1.9 cells, respectively(Fig. 6B). By contrast, the resultant partial embryos developed from mother cells isolated from ephrin-Ad-MO-injected embryos produced more notochord cells (2.9 cells) at the expense of neural cells (0.5 cells) (Fig. 6B). These results show that ephrin-Ad is indeed acting on the notochord/neural mother cells and is required for their autonomous capacity to generate two distinct daughter cells, with ephrin-Ad promoting the neural daughter cell fate.

Asymmetric cell divisions are intensively studied as a mechanism used to generate cell diversity during development, and also to control the self-renewal versus differentiation choice during stem-cell divisions. Two basic types of mechanisms can generate two daughter cells that adopt distinct cell fates. The first, generally referred to as asymmetric cell division,relies on the polarisation of the mother cell, which intrinsically gives rise to distinct daughter cells (Knoblich,2001; Roegiers and Jan,2004; Betschinger and Knoblich,2004) (see also Horvitz and Herskowitz, 1992; Morrison and Kimble, 2006). The second type does not involve polarisation of the mother cell but involves the asymmetric placement of equivalent daughter cells relative to extrinsic instructive cues, thus resulting in their adoption of distinct fates (Horvitz and Herskowitz,1992; Morrison and Kimble,2006). We have shown that the cell division of the notochord/neural mother cell in Ciona embryos can be considered as truly `asymmetric' because, in isolation, this cell is able to divide autonomously to generate two daughter cells with distinct fates. This asymmetric cell division is dependent on cellular interactions between the mother cell and adjacent animal cells, and requires ephrin-Eph signalling. We have found that Ci-ephrin-Ad, present in animal cells, signals via Eph receptors to the mother cells, which is required for them to divide into two distinct daughter cells, with ephrin-Ad promoting the neural daughter cell fate. Our study also demonstrates that the ephrin-Eph signal acts via attenuation of ERK activation in the neural precursor. Based on these findings, we propose the following model for the asymmetric cell division of notochord/neural mother cells. While broadly expressed FGF9/16/20 activates the Ras-MEK-ERK pathway in the mother cell, directional ephrin-Ad signals from the animal cells act in parallel to functionally polarise the mother cell. Local activation of the ephrin-Eph signal in the mother cell becomes manifest,following cell division, as the selective attenuation of ERK activation in the neural-fated daughter cell, thus resulting in the differential fate specification of the daughter cells as notochord (ERK on) and neural (ERK off)precursors.

It will be of great interest in future studies to unravel the precise molecular cascade acting downstream of ephrin-Ad-Eph signalling in order to understand how this activity becomes spatially restricted within the mother cell prior to its division, and how this leads to the differential attenuation of ERK activation between the daughter cells following division. RTKs,including Eph, commonly signal through cytoplasmic proteins with SH2 domains,which bind either directly to phosphotyrosine sites on the activated receptor or to phosphorylated docking proteins. A variety of SH2 proteins have been identified as potential Eph receptor-binding partners(Hock et al., 1998; Holland et al., 1997). Among them is the p120-Ras GTPase-activating protein (p120-RasGAP), which is a negative regulator of the Ras small G protein(Hock et al., 1998; Holland et al., 1997) and has been shown to be required for EphB2- and EphA2-mediated attenuation of ERK1/2 in NG108 neuronal cell lines and mouse embryonic fibroblasts, respectively(Elowe et al., 2001; Tong et al., 2003). It is tempting to speculate that activation of ephrin-Ad-Eph signals results in the recruitment of p120-RasGAP to a sub-membrane domain on the animal-pole side of the A6.2 and A6.4 mother cells, and that the subsequent cytokinesis of the mother cells results in the partitioning of the RasGAP-enriched sub-membrane domain only to the future neural daughter cells.

During animal development, ephrin-Eph signals are well-known for their ability to control cell behaviours such as attraction/repulsion,adhesion/de-adhesion and migration in a range of developmental contexts(Poliakov et al., 2004). Our current study adds to the list a new role for ephrin-Eph signals - a role in cell fate specification coupled with an asymmetric cell division. The current study also provides compelling evidence that activation of the ephrin-Eph signalling pathway can result in attenuation of ERK1/2 activation in the context of embryonic development, in addition to the examples in cell lines and during C. elegans oocyte maturation(Elowe et al., 2001; Miao et al., 2001; Miller et al., 2003). RTK-mediated ERK1/2 activation is responsible for the control of a wide variety of developmental processes requiring spatial and temporal precision(Christen and Slack, 1999; Corson et al., 2003; Eblaghie et al., 2003; Gabay et al., 1997; Sawada et al., 2001; Shinya et al., 2001). Thus, it will be important to determine whether ephrin-Eph signalling acts more generally, in additional developmental contexts, as a potent local regulator of the RTK-ERK signalling pathway.

In conclusion, we have revealed a novel mechanism of asymmetric cell division based on ephrin-Eph signalling and selective ERK attenuation,broadening our knowledge of the diverse mechanisms used to control these specialised cell divisions. It will now be important to address whether ephrin-Eph signalling is playing similar roles in additional developmental and cellular contexts.

We thank Nori Satoh and colleagues for the Ciona Gene Collection Plates; Evelyn Houliston and Andrea Pasini for critically reading the manuscript; and special thanks goes to A. Pasini for suggesting that we look at ephrins. Many thanks to the staff of the UMR7009 for all kinds of help and support. This study was funded by the Centre National de la Recherche Scientifique (CNRS), including ATIP and ATIP plus, the Université Paris VI, the French Ministry of Research (GIS `Génomique Marine'), the Association pour la Recherche sur le Cancer (ARC) and the Agence Nationale de la Recherche (ANR).