Sphingosine-1-phosphate receptors regulate individual cell behaviours underlying the directed migration of prechordal plate progenitor cells during zebrafish gastrulation

Progenitor cells destined to form ectodermal, mesodermal and endodermal fates undergo several different cell movements that shape the embryonic body axis during vertebrate gastrulation. Different populations of mesodermal cells have unique cell behaviours underlying directional and co-ordinated cell movements. For example, prechordal plate progenitor cells undergo directed anterior migration as a small group of coherent cells, whereas in the presumptive notochord and presomitic mesoderm cells undergo co-ordinated rearrangement, called convergence and extension (C&E), as a large sheet of cells (reviewed by Heisenberg and Tada,2002; Solnica-Krezel,2005).

In Xenopus, once the mesoderm has involuted, the most anterior dorsal mesodermal cells migrate directionally toward the animal pole across the blastocoel roof of the ectoderm using fibronectin as a substrate(Nagel and Winklebauer, 1999). In zebrafish, the first internalised axial hypoblast cells that become fated to the prechordal plate move as a cohesive sheet of cells toward the animal pole of the embryo using the epiblast as a substrate(Montero et al., 2005; Ulrich et al., 2003). The cohesive property of prechordal plate progenitor cells is thought to provide a mechanism for effective directed migration(Ulrich et al., 2005), and is reminiscent of the collective migration seen in border cells in Drosophila and in lateral line progenitor cells in zebrafish(reviewed by Lecaudey and Gilmour,2006; Rorth,2007). However, the mechanism underlying this process remains poorly understood, in particular is collective migration really important or can cells migrate as individuals?

The identification of directional cues for the anterior migration of prechordal plate progenitor cells has fascinated developmental biologists for several decades. The best characterised is platelet-derived growth factor(Pdgf) in Xenopus, as Pdgf signalling is required for the orientation of cells toward the animal pole and for their directionality(Nagel et al., 2004). By contrast, Pdgf and its key intracellular transducer PI3K are necessary for cell polarisation and motility, but not for directionality, in zebrafish(Montero et al., 2003). Despite the migration defects in embryos with compromised Pdgf pathway activities, prechordal plate progenitor cells are largely still capable of undergoing directed migration in Xenopus and zebrafish, suggesting the existence of another cue(s) in this process.

What are the genetic pathways involved in the regulation of coherence of cells during anterior migration of the presumptive prechordal plate?E-cadherin and Rab5-mediated endocytosis regulate the cell coherence of prechordal plate progenitors, and these are mediated by Wnt11(Ulrich et al., 2005), a member of the non-canonical Wnt/planar cell polarity (PCP) pathway (reviewed by Tada et al., 2002; Seifert and Mlodzik, 2007). Indeed, silberblick (slb)/wnt11 mutant embryos exhibit reduced anterior migration of prechordal plate progenitor cells(Ulrich et al., 2005). The expression of E-cadherin (cdh1) in the presumptive prechordal plate is regulated by snail genes that are required for anterior migration of the prechordal plate in zebrafish(Blanco et al., 2007). Furthermore, Liv1 (Slc39a6 - Zebrafish Information Network), a zinc transporter, controls nuclear localisation of Snail, and is required for anterior migration of the prospective prechordal plate during zebrafish gastrulation (Yamashita et al.,2004). Moreover, the secreted Wnt antagonist Dkk1 modulates anterior migration of mesodermal cells by interacting with Knypek/Glypican4, a cofactor for Wnt11, in zebrafish (Caneparo et al., 2007).

The bioactive lipid, sphingosine-1-phosphate (S1P), is a signalling molecule that acts through binding to a family of seven-pass transmembrane,G-protein-coupled receptors (S1PRs) that have been implicated in the regulation of cytoskeletal rearrangements, cell motility and cell adhesion in a variety of cell types (reviewed by Spiegel and Milstien, 2003). In addition to the role for S1P as a ligand, it acts as a second messenger within the cell, and its intracellular levels are regulated by the balance between its production by Sphingosine kinases and its degradation by S1P lyases (reviewed by Alvarez et al.,2007). However, little is known about the functions of S1P and its receptors in regulating directed cell migration in the embryo.

In this study, we sought to search for a novel genetic pathway that controls the collective migration of prechordal plate progenitor cells at the onset of zebrafish gastrulation. In our morpholino (MO)-based screen, we have identified miles apart (mil) as a genetic suppressor of defective migration of prechordal plate cells in slb/wnt11 mutant embryos. mil was initially isolated as a heart mutant with a cardia bifida phenotype, and it encodes a S1P receptor, Edg5, also called S1pr2(Kupperman et al., 2000). We analysed the cell behaviour of the presumptive prechordal plate based on DIC time-lapse movies of the living zebrafish gastrula, and found that, in slb embryos injected with mil-MO, the cells migrated with increased motility but decreased directionality, without restoring the coherence of cell migration in slb embryos. Furthermore, we showed that Mil controls cell motility through the Pdgf/PI3K pathway but modulates individual cell behaviours underlying cell coherence separately from this pathway. These results highlight the unexplored role of the motility and coherence of individual cells, regulated by the Mil/S1P signal, in the directed migration of prechordal plate progenitors.

The maintenance of fish and the collection of embryos were performed essentially according to Westerfield(Westerfield, 2000). The allele we used was homozygous slb^tx226, which had been selected for slb carriers that consistently showed the defective movement of the prechordal plate, and slb mutant embryos were obtained from in-crossing of the homozygotes. The embryos were kept at 28.5°C for time-lapse analysis or at 31°C for in situ hybridisation.

The sequences of morpholinos used in this study were: mil-MO,5′-CCGCAAACAGAGCCGAACTAGTCAT-3′ (ATG-MO)(Matsui et al., 2007); and edg1-MO, 5′-TTAGGTCATCCATGGTTTGCACTGG-3′ (ATG-MO). mil-MO and edg1-MO were, respectively, injected at concentrations of 4.3 ng/nl and 6.2 ng/nl. mil-MO injection phenocopied the cardia bifida condition in the mil mutant (87.5%, n=56). To validate the specificity of the edg1-MO, RNA encoding GFP tagged with amino acid sequences that include the sequence corresponding to the edg1-MO, or irrelevant sequences (ILK) as a negative control, was co-injected with edg1-MO, and GFPexpression was examined at dome stages. GFP expression from the construct containing the sequence corresponding to the edg1-MO was suppressed (GFP-positive: 0%, n=19), whereas GFPexpression was retained in the negative control (GFP-positive: 100%, n=36). Together with the ability of the edg1-MO to cancel the edg5-morphant phenotype when co-injected (see Fig. 6B,E,H to compare with Fig. 1B,F,I), these data validate the specificity of the edg1-MO.

For overexpression studies, the following constructs were used:pCS2-dn-PI3K, pCS2-p110CAAX and pCS2-PH-GFP [all described previously by Montero et al. (Montero et al.,2003)]. As for the mil construct, a PCR-amplified full-length fragment from a gastrula library was cloned into pCS2+, and RNA for injection was made after linearisation with NotI. All of the RNAs were synthesised in vitro essentially as described(Smith, 1993).

Antisense RNA probes were synthesised with a digoxigenin RNA-labelling kit(Roche), using plasmids containing cDNA for ntl(Schulte-Merker et al., 1994), hgg1 (Thisse et al.,1994), dlx3 (Akimenko et al., 1994), hlx1(Fjose et al., 1994) and sprouty4 [(Furthauer et al.,2004) originally published as sprouty2]. Whole-mount in situ hybridisation was carried out as described previously(Barth and Wilson, 1995).

For time-lapse imaging, embryos were manually dechorionated at the shield stage and mounted in 0.8% low-melting-point agarose (Sigma) in embryo medium. For DIC observation, images were taken at 1 frame per minute for 1 hour, with approximately 18 z-levels spaced 3 μm apart, using a 40×water immersion lens on an Axioplan 2 (Zeiss) compound microscope and a Hamamatsu Orca ER digital camera. Two-photon microscopy was performed as described by Montero et al. (Montero et al., 2003).

The positions of five randomly chosen cells were plotted every minute,using OpenLab5.0 software (Improvision). The `speed' of a cell was measured asμm/minute using gross length of migration. The `persistence' of a cell was calculated by the quotient of net migration per gross migration over every 14 minutes. Assuming that cells change their relative positions more dynamically in less coherent conditions, the `coherence<sup>-1</sup>' of migration of a group of cells was defined as the rate of change in cell-cell distance per length of cell migration. Here, the change of distance between two cells was determined, then normalised for each cell by dividing by its net migration(see Fig. S1 in the supplementary material). This calculation was applied to all 10 combinations among five cells, every 9 minutes. For each experimental condition, three to five embryos from independent experiments were analysed. To test the significance between two mean values, Welch's t-test based on an unequal variance was applied and described as`mean±s.e.m.'.

Approximately 25 cells at the leading edge from three to five embryos were randomly chosen from time-lapse images, and, of those, cells with a distinctive cellular protrusion were used for analysis with ImageJ software. The `length' and `angle' of the largest protrusion were determined by assuming a line from the centre of the nucleus to the tip of the protrusion: the length being its length in μm, and the angle being measured from the direction of general migration (i.e. the upper-vertical axis). Here, the mean and standard deviation of angles are calculated in cosine, as measured anticlockwise from the upper-vertical axis. White dots indicate the central position of the nucleus, and green lines show the length and angle of a protrusion. Red `fans'represent the average length and angle (in degrees) of the protrusions,including their standard deviations (respectively depicted by its height and open angle). To test the significance in the length between two mean values,Welch's t-test based on an unequal variance was applied and described as `mean±s.e.m.'.

Cell culture experiments were carried out according to Montero et al.(Montero et al., 2003) with slight modification. In brief, 50 embryos injected with 100 pg cyclops RNA together with the RNA(s) indicated in each experiment were manually dechorionated at the dome stage, and dissociated in L-15 medium containing 1×trypsin-EDTA (Biowhittaker). After stopping the enzymatic reaction by the addition of chick serum, cells were harvested by centrifugation at 150 g for 2 minutes and re-suspended in 2.5 ml of fresh L-15 medium, containing 1 mg/ml insulin, 0.3 mg/ml L-glutamine,100 U/ml penicillin and 100 μg/ml streptomycin. Cells were then cultured for 1 hour at 25°C on a plastic Petri dish coated with 50 μg/ml fibronectin (Sigma), prior to treatment with 50 ng/ml Pdgf-AA (Sigma).

The grey value along a line of 3-pixel weight was measured using ImageJ software. The x-axis of the graph represents the distance, while the y-axis indicates the relative signal intensity (as normalised by total intensity), along the line.

In silberblick (slb)/wnt11 embryos, prechordal plate progenitor cells move anteriorly, but slower than those in wild-type(WT) embryos because of the reduced cohesion of cells; cell cohesion is thought to be required for prechordal plate progenitor cells to undergo collective migration (Montero et al.,2005; Ulrich et al.,2005). In order to identify genetic cascades that regulate anterior migration of the prospective prechordal plate, we have undertaken a candidate approach in which morpholino antisense oligonucleotides (MOs) are injected into slb homozygous embryos as a sensitised background to detect either enhancement or suppression of the slb phenotype. During the course of this approach, we identified miles apart (mil)as a suppressor of the slb phenotype. mil was originally isolated as a mutant exhibiting a bifurcated heart, and encodes a sphingosine-1-phosphate (S1P) receptor, Edg5(Kupperman et al., 2000). When the position of the anterior prechordal plate was visualised with respect to the anterior edge of the neural plate at the end of gastrulation, the presumptive prechordal plate of the slb embryo was located more posteriorly than that of the WT embryo(Fig. 1A,C,E,G)(Heisenberg et al., 2000). Strikingly, the deficit in anterior movement of the slb prechordal plate was largely rescued by injection of mil-MO(Fig. 1C,D,G,H; 65.4%, n=52). When WT embryos were injected with mil-MO, the shape of the prospective prechordal plate was slightly flatter and wider, and its positioning was slightly more anterior than in WT embryos(Fig. 1A,B,E,F). In addition,the anteroposterior length of the posterior prechordal plate was increased in embryos with compromised Mil functions in WT and slb embryos(Fig. 1I-L; compare with Fig. 1A-D and Fig. S2 in the supplementary material). In contrast to the prechordal plate, slbembryos exhibit a shorter and thicker notochord than do WT embryos, implying defective C&E movements in the presumptive notochord(Heisenberg et al., 2000), but this defect was not rescued by injection of mil-MO(Fig. 1O,P).

This observation raises the possibility that Mil activity might regulate anterior migration of the presumptive prechordal plate but not of the prospective notochord, where cells undergo C&E. If this were the case,anterior movement of the lateral mesoderm would be unaffected in milmorphants. To test this, we examined the expression of sprouty4 to see to what extent lateral mesodermal cells have migrated at 60% epiboly(Caneparo et al., 2007). Despite the fact that axial mesodermal cells migrated faster in the mil morphant, anterior movement of lateral mesodermal cells was unchanged when compared to the WT embryo (90.3%, n=31; Fig. 1M,N). These results suggest that Mil activity is required only for prechordal plate progenitor cells to undergo directed anterior migration, and not for other populations of mesodermal cells.

In order to explore the cellular mechanisms by which the abrogation of Mil function restores the reduced migration of prechordal plate progenitor cells in slb embryos, we tracked the cells based on DIC time-lapse movies at the onset of gastrulation, and analysed the cell behaviours that underlie collective migration. We extracted three parameters: first, speed, by the total length of their paths over the time; second, persistence, by the gained distance over the total length of their paths; and third, coherence, by changes in relative distances between pairs of cells over time for all the combinations of five cells (a total of 10 combinations per time point).

Cell tracking analysis revealed that the reduced migration of slbhypoblast cells is closely associated with a lower persistency in directed migration and a reduced coherence of cell migration(Fig. 2A-F,J; see Movies 1 and 2 in the supplementary material) (Ulrich et al., 2005; Witzel et al.,2006). Surprisingly, when Mil function is compromised, slb hypoblast cells migrated faster than did WT cells, but more randomly and with more space between the cells, thus maintaining a lower cell cohesion than WT cells (Fig. 2G-I,J; see Movie 3 in the supplementary material). Occasionally,we observed that an isolated single cell popped out but then travelled back into the group of cells (Fig. 2H; observed in four movies out of five). These results indicate that slb cells with compromised Mil function have acquired more motility, but that their directionality and coherence remain as low as slb cells. Consistent with this, leading edge cells of the slb embryo with compromised Mil function formed longer lamellipodia than did both slb cells and WT cells, but were unable to stabilise their processes toward the direction of movement when compared with WT cells(Fig. 3A-C).

We further sought to identify differences in behaviours of slbcells with compromised Mil function when compared with WT cells, and investigated involution of the axial hypoblast in a sagittal view. Whereas both WT and slb hypoblast cells internalised as three layers of cells, as shown previously (Montero et al., 2005; Ulrich et al.,2005) (Fig. 4A and Movie 8 in the supplementary material), slb cells with compromised Mil function underwent internalisation as one or two layers of cells(Fig. 4B and Movie 9 in the supplementary material). Similarly, abrogation of Mil function in the WT embryo leads to the scattered cell behaviour phenotype during internalisation,as seen in the slb embryo with compromised Mil function (see Movie 10 in the supplementary material). Consistent with the observation from DIC movies (Fig. 2G,H), there was more space in between the hypoblast cells in slb embryos injected with mil-MO than in WT embryos, confirming that prechordal plate progenitor cells migrate individually in slb embryos when Mil function is compromised.

Despite the fact that slb cells with compromised Mil activity retain a lower coherence of cell migration to the direction of motion, they can gain the net distance as efficiently as WT cells due to their acquisition of greater motility (Fig. 5). In this extreme circumstance, cells appear be capable of migrating individually rather than in a cluster of coherent cells.

It has been shown that the two structurally related S1P receptors, Edg1(also called S1pr1) and Edg5, are positive and negative regulators,respectively, of cell migration in lymphocytes and endothethial cells (e.g. Yamaguchi et al., 2003). To test this possibility in directed migration of the prospective prechordal plate, we first injected WT embryos with edg1-MO. Opposite to the mil-morphant phenotype, directed migration of the anterior and posterior prechordal plate was perturbed in edg1 morphants(Fig. 1B,F,J and Fig. 6A,D,G). Consistent with their opposing activities in cultured cells, the reduced migration of edg1-morphant prechordal plate cells was rescued by co-injection of mil-MO (Fig. 6B,E,H). Next, we tested whether edg1 is an enhancer of the slbphenotype as opposed to mil being identified as a suppressor. When edg1 function was compromised, slb embryos exhibited a more posteriorly displaced prechordal plate but the notochord was unaffected(Fig. 1C,G,K, Fig. 6C,F,I, data not shown). These results suggest that Edg1 and Mil act in a mutually antagonistic manner to regulate directed migration of the presumptive prechordal plate, and that Edg receptors control directed migration of prechordal plate progenitors but not of other mesodermal cells.

Pdgf signalling through the intracellular mediator PI3K is one genetic pathway that has been shown to regulate the directed migration of mesendoermal cells in zebrafish and Xenopus(Montero et al., 2003; Nagel et al., 2004; Symes and Mercola, 1996). Therefore, we investigated a possible interaction of the Mil/S1P signal with the Pdgf/PI3K pathway in directed migration of prechordal plate progenitor cells. We hypothesised that Mil controls directed migration by negatively regulating the Pdgf/PI3K pathway, as inhibition of the Pdgf/PI3K pathway leads to a reduction in cell migration (Montero et al., 2003), and this phenotype is the opposite of the mil morphant. To test this hypothesis, we first investigated process formation in response to Pdgf in dissociated mesendodermal cells from zebrafish blastulae. Although mesendodermal cells formed lamellipodia-like long processes when treated with Pdgf, Pdgf treatment of cells overexpressing mil RNA resulted in a reduction in the frequency and length of processes (Fig. 7A,B). Next, to test whether Edg5 inhibits the intracellular effecter PI3K, we monitored the membrane localisation of PH-GFP induced by Pdgf, and of membrane-RFP as a reference. Pdgf-induced membrane localisation of PH-GFP was inhibited by the presence of mil RNA (Fig. 7C-F). These results indicate that the Mil/S1P signal might control cell polarisation through the Pdgf/PI3K pathway.

The possible interaction of Mil with the Pdgf signal, as well as with Wnt11, inspired us to identify which parameters (coherence, directionality and motility) are regulated by Mil through or in parallel with these pathways in the directed migration of prechordal plate progenitors. Because Mil can modulate cell adhesion underlying coherence of cell migration in a fibronectin-dependent manner (Matsui et al., 2007), we hypothesised that the reason why slb cells(when Mil function is compromised) can move faster than WT cells even though they retain a lower coherence of cell migration might be because Mil can modulate cell cohesion independently of Slb/Wnt11 activity. To test this, we injected WT embryos with mil-MO and analysed the cell behaviour of prechordal plate progenitors. Cells with compromised Mil function migrated persistently but less coherently than did WT cells (see Fig. S1 and Movie 4 in the supplementary material; see also Fig. 8C). If Mil modulates cell cohesion through PI3K, prechordal plate progenitor cells with increased PI3K activity would lead to phenotypes similar to that of mil-morphant cells. To test this possibility, we expressed RNA encoding a constitutively active form of PI3K (p100CAAX) in the WT embryo to examine the cell behaviour of migrating prechordal plate progenitors. In contrast to abrogation of Mil function, the expression of p100CAAXRNA did not cause cells to change their cohesive properties in WT embryos(Montero et al., 2003) (see Fig. S4 and Movie 5 in the supplementary material; see also Fig. 8C). Conversely, a reduction in the motility of WT cells expressing dominant-negative (dn)-PI3K was not due to a lower coherence of cell migration (see Fig. S4 and Movie 6 in the supplementary material; see also Fig. 8C). These results suggest that Mil regulates the coherence of cell migration independently of Wnt11 activity and PI3K.

We further tested whether Mil-mediated cell motility is dependent upon the Pdgf/PI3K pathway. As the phenotype caused by dn-PI3K shows reduced migration and is opposite apparently to that of the mil morphant, we performed epistasis analysis of the interaction between the two signals by examining details of the cell behaviours based on DIC time-lapse movies. Prechordal progenitor cells expressing dn-PI3K migrated much slower than did WT cells,while their directionality was largely unaffected(Montero et al., 2003)(Fig. 8A,C; see also Movie 6 in the supplementary material). The reduced motility of cells expressing dn-PI3K was not reversed by knocking down Mil function, which suggests that Mil regulates cell motility through PI3K (see Fig. S4 and Movie 7 in the supplementary material; see also Fig. 8C). However, these cells migrated much less persistently than did cells with reduced PI3K activity only. The lower persistency of cell movement,when both PI3K and Mil functions are compromised, correlated with the randomised orientation of lamellipodia of leading edge cells with respect to their direction of motion (Fig. 8C). These results prompt us to propose that Mil controls the directed migration of prechordal plate progenitor cells in both a Pdgf/PI3K-dependent and -independent manner.

In this study, we demonstrate that prechordal plate progenitor cells are able to migrate efficiently as individual cells in the gastrula when Mil function is compromised, rather than as a cluster of coherent cells, and that this cell behaviour is more prominent in the slb/wnt11 embryo. This brings into question the previous view that prechordal plate progenitors undergo collective migration as a group of cohesive cells that presumably creates a force to drive directed migration(Montero et al., 2005; Ulrich et al., 2005).

Cases for which collective migration in the embryo have been well documented are lateral line primodia in zebrafish and border cells in Drosophila. In contrast to the mesenchymal nature of prechordal plate progenitor cells in zebrafish, lateral line primodia cells undergo collective migration as a sheet of epithelia, such that only a few front-line cells lead their migration while the rest of the cells follow the leader cells(Haas and Gilmour, 2006). By contrast, border cells in Drosophila undergo collective migration as mesenchymal cells by changing their position within the cluster in the absence of leader cells (Prasad and Montell,2007). There is no evidence that leader cells exist within prechordal plate progenitor cells, but we observed that cells located at the leading edge retain their position at least for the duration of the time-lapse movies (e.g. Movies 1 and 8 in the supplementary material). This suggests that prechordal plate progenitor cells are capable of migrating as individual cells, and that this might be similar to the traditional case of individual cell migration during chemotaxis in Dictyostelium (reviewed by Dormann and Weijer, 2006; Franca-Koh et al., 2006).

Prechordal plate progenitor cells can migrate as isolated single cells in extreme situations. Interestingly, single cells do not travel alone far away from the group of prechordal plate progenitors and in all cases they come back to the group. This raises the possibility that the isolated cell might lose its identity as a presumptive prechordal plate progenitor; for example, goosecoid expression could have been lost. In MZoep embryos, a single WT cell can internalise but is incapable of migrating anteriorly(Carmany-Rampey and Schier,2001), presumably its anterior mesendodermal cell identity has been lost. Alternatively, the isolated cell might lose motility through the loss of a community effect that normally maintains cells as a group. However,it is difficult to distinguish between these possibilities, as it is technically challenging to visualise cell identity in the living embryo in such a short period. Also, it will be interesting to test whether or not cells with higher motility and lower directionality can migrate individually when they are isolated from surrounding cells with lower motility and lower coherence.

We have identified mil/edg5 as a suppressor of the slb/wnt11 phenotype in this study. This raises the intriguing possibility that Mil might act in the non-canonical Wnt/PCP pathway. However,several lines of evidence do not support this possibility. First, although abrogation of Mil function with MOs restores the anterior migration of prechordal plate progenitors in slb embryos, it fails to rescue defects in the slb presumptive notochord. Second, while injection of dkk1-MO facilitates anterior migration of the prechordal plate in WT embryos, as in the mil morphants, it cannot rescue the defective anterior migration of the slb prechordal plate(Caneparo et al., 2007),suggesting that Dkk1-regulated migration is dependent upon Slb/Wnt11 function in this context. Indeed, it has been shown that Dkk1 acts in the non-canonical Wnt pathway by physically and functionally interacting with Knypek/Glypican4,a known co-factor of Wnt11 (Caneparo et al., 2007; Topczewski et al.,2001). Third, the cardia bifida phenotype caused by mil-MO injection is still observed in the slb homozygous background (data not shown). These results support the notion that the Mil signal acts independently of the non-canonical Wnt pathway.

What molecule(s) mediates downstream of the Mil/S1P signal? In mammalian cultured cells, it has been to shown that Gα_12/13 is activated by EDG5 (Gonda et al.,1999). Importantly, abrogation of α_12/13 function leads to impaired mediolateral elongation of mesendodermal cells undergoing convergence of the lateral mesoderm and extension of the notochord during zebrafish gastrulation (Lin et al.,2005). However, it implies that α_12/13 is unlikely to mediate EDG5/S1P signalling directly, as embryos with altered EDG5/EDG1 activities exhibit little defect in mesodermal cells undergoing convergence and extension. It remains to be investigated which heteromeric G protein mediates signalling downstream of Mil/S1P in the directional migration of prechordal plate progenitors. In addition, how different G protein-coupled receptors (GPCRs) regulate the movements of distinct cell populations during gasrulation needs to be addressed, as another GPCR Agtrl1b predominantly mediates the movements of the lateral mesoderm of the zebrafish gastrula(Scott et al., 2007; Zeng et al., 2007).

The best candidate for a directional cue is Pdgf, but it appears that Pdgf regulates anterior migration of the lateral mesoderm as well as of the anterior axial mesoderm in Xenopus and zebrafish(Nagel et al., 2004) (data not shown), whereas the Mil/S1P signal is more specific to prechordal plate progenitor cells in zebrafish. Do Edg receptors regulate directional migration cell-autonomously within the prospective prechordal plate? Considering the fact that the expression of both mil and edg1 is ubiquitous during gastrulation (Kupperman et al.,2000) (data not shown), the localisation of the bioactive lipid S1P might be restricted to either the presumptive prechordal plate or the overlying neurectoderm. Although it will be challenging to visualise the short life of the bioactive lipid in the embryo, the temporal and spatial localisation of its producing and degrading enzymes, such as Sphingosine kinase and S1P lyase, respectively, might allow us to speculate where a potential gradient of the bioactive lipid is, as in the case for the gradient of retinoic acid in the zebrafish hindbrain(White et al., 2007). Alternatively, S1P receptors might be required cell-non-autonomously in the overlying neurectoderm, as it appears that there is correlation of tissue movements between the presumptive prechordal plate and the hypothalamus (see Fig. S3 in the supplementary material). To clarify the apparent interdependency of the two tissue movements, we will need to manipulate the movement of one tissue to see possible alteration in the other tissue. This will require the use of a transgenic approach rather than a transplantation one, as transplanted cells normally disperse in the host and thereby lose coherence of cell movement, which might mediate the interaction of two distinct populations.

Recent searches for transcriptional targets of Pdgf in cultured cells revealed that Pdgf upregulates sgpl1 (sphingosine-1-phosphate lyase 1), which encodes an enzyme that degrades active S1P(Chen et al., 2004). Importantly, embryonic fibroblasts from sgpl1^-/- mice in cultures showed reduced migration in response to Pdgf(Schmahl et al., 2007),indicating that the ability of cells to inactivate active S1P is required for the proper response to Pdgf signalling to mediate cell motility. Moreover,Pdgf stimulates the production of S1P by activating and translocating Sphingosine kinase, an S1P-producing enzyme, to the membrane in lymphocytes(Hobson et al., 2001). This further suggests a possible auto-regulatory loop between Pdgf and S1P signals. Therefore, it will be fascinating to investigate this issue in directed migration in the gastrulating embryo.

Recent studies in medaka revealed that maternal-zygotic fgfr1mutants exhibit defective migration of the axial mesoderm but not of the lateral mesoderm (Shimada et al.,2008), and this phenotype is reminiscent of the edg1-morphant phenotype, raising the possibility that the Mil/S1P signal might also co-operate with the Fgf signal. It will be intriguing to test this possibility in zebrafish, as Fgf is capable of activating PI3K as well as Pdgf. Furthermore, the question of how prechordal plate cells acquire greater motility and/or can sense directional cues more efficiently than the other mesodermal cells remains unsolved.

We demonstrated that overexpression of mil RNA inhibits Pdgf-induced cell polarisation in dissociated mesendodermal cells by interfering with Pdgf-mediated activation of PI3K, and that the facilitated migration of the mil morphant is blocked totally by dn-PI3K. These results indicate that Edg5 can act upstream of PI3K to negatively regulate the Pdgf pathway. However, cell-tracking analysis of mil morphants and of embryos with increased or reduced PI3K activity revealed that coherence of cell migration mediated by Mil is likely to be independent of PI3K activity(Fig. 9). By contrast, the directionality of migrating cells is totally lost in the milmorphants when PI3K is blocked, implying that Pdgf and Edg signalling might co-operate to mediate the directionality of prechordal plate progenitor cells(Fig. 9). This is different from the mode of regulation of directionality in Xenopus, as blocking the Pdgf signal is sufficient to lose the directionality of anterior axial mesodermal cells (Nagel et al.,2004). Taken together, these results suggest that Mil modulates cell motility through the Pdgf/PI3K pathway, but that it modulates coherence independently of the pathway.

Does Mil modulate cell cohesion passively as a consequence of increased cell motility or instructively by unknown mechanisms? We favour the latter possibility for several reasons. First, abrogation of Mil function leads to the reduced coherence of cell migration in WT embryos, whereas WT embryos with increased or decreased PI3K activity do not show any changes in coherence of cell migration. This implies that cell motility and coherence are independently regulated. Second, when E-cadherin-mediated cell cohesion was decreased in slb embryos, slb embryos with compromised Mil function acquired more motility than WT embryos. Moreover, even E-cadherin-morphant cells, when Mil function was compromised, were able to migrate as efficiently as WT cells (data not shown), suggesting the presence of a Mil mediator of cell coherence that functions independently of E-cadherin-dependent cell adhesion. Third, abrogation of Mil function in the WT embryo leads to the scattered cell behaviour phenotype during internalisation, as seen in the slb embryo with compromised Mil function (see Movie 10 in the supplementary material). This supports the notion that Mil modulates cell cohesive properties independently of Slb activity (Fig. 9). Fourth, the mil cardia bifida phenotype might be explained by the modulation of adhesive properties through the interaction with fibronectin(Matsui et al., 2007). Whether or not Mil modulates the cell cohesive property underlying the directed migration of collective cells in the embryo will require further investigation.

We thank Jon Clarke and Steve Wilson for critical comments on the manuscript, members of the UCL fish group for fruitful discussions, and Andrew French for helpful comments on cell coherence analysis. This work was supported by grants from the MRC and Royal Society to M.T., and the Wellcome Trust and BBSRC jointly to M.T. and Steve Wilson.