Control of convergent yolk syncytial layer nuclear movement in zebrafish

Nuclear movements in eukaryotic cells are important for a wide variety of processes such as mitosis, meiosis, cell migration and cell polarization(Morris, 2000). Although in most cases nuclei are transported along microtubules by microtubule motor proteins (Morris, 2000; Morris, 2003; Reinsch and Gonczy, 1998), the actin cytoskeleton has also been implicated(Reinsch and Gonczy, 1998; Starr and Han, 2003). In addition to these intrinsic transport mechanisms, signals from surrounding tissues and cells have also been involved in nuclear movements. Notably,studies in Drosophila have provided evidence that the migration of the oocyte nucleus, which is required to establish the dorsal-ventral axis of the embryo, crucially depends on localized signals from the surrounding follicular epithelium (Gonzalez-Reyes et al., 1995; Roth et al.,1995).

Extensive nuclear movements have been observed within syncytial environments in metazoan development(D'Amico and Cooper, 2001; Englander and Rubin, 1987; Foe and Alberts, 1983; Trinkaus, 1993; Xiang and Fischer, 2004). One of the best-studied examples is the Drosophila syncytial preblastoderm, where nuclei initially move along the anterior-posterior axis of the embryo by actin-dependent cytoplasmic streaming and are subsequently transported towards the cortex along microtubules by associated motor proteins(Baker et al., 1993; Robinson et al., 1999; von Dassow and Schubiger,1994). Although these studies have provided detailed insights into the molecular and cellular mechanisms that underlie syncytial nuclear migrations, much less is known about how such migrations are coordinated with the movements of surrounding cells and tissues in the developing organism.

To obtain insights into the coordination of syncytial nuclear movements with surrounding cells and tissues, we have chosen to study the yolk syncytial layer (YSL) in the zebrafish gastrula. The YSL and its nuclei play crucial roles in embryo patterning and morphogenesis. Nodal/TGFβ signals emanating from the YSL are thought to be involved in specifying mesoderm and endoderm cell fates in marginal blastomeres along the circumference of the embryo (Chen and Kimelman,2000; Mizuno et al.,1999; Mizuno et al.,1996; Ober and Schulte-Merker,1999; Rodaway et al.,1999). Furthermore, interfering with the actin and microtubule cytoskeleton of the YSL leads to defective epiboly movements of both the YSL and the cellularized blastoderm (Cheng et al., 2004; Solnica-Krezel and Driever, 1994; Zalik et al.,1999), suggesting that proper YSL morphogenesis is required for blastoderm epiboly. Finally, the YSL has been proposed to induce fibronectin expression in the overlying embryonic tissue, which is important for heart progenitor cell migration during gastrulation and somitogenesis stages(Sakaguchi et al., 2006).

The YSL is formed by marginal blastomeres that collapse and deposit their nuclei into the cortical cytoplasm of the yolk cell at blastula stage(Kimmel and Law, 1985; Trinkaus, 1993). These yolk syncytial nuclei (YSN) undergo three to five rounds of divisions before ceasing mitosis at sphere stage. At this stage, the YSL contains several hundreds of YSN that can be subdivided into two main groups according to their position within the YSL (D'Amico and Cooper, 2001). External YSN (eYSN) form a marginal band of YSN located in front of the enveloping cell layer (EVL). Some of the eYSN move towards the vegetal pole (epiboly movements), whereas others recede from the marginal zone and undergo convergence and extension (CE) movements. By contrast, internal YSN (iYSN) are located below the blastoderm and EVL and primarily undergo CE movements. Strikingly, CE movements of iYSN appear very similar to the CE movements of the overlying blastoderm (mesoderm, endoderm and ectoderm progenitors) (D'Amico and Cooper, 2001). It has therefore been hypothesized that iYSN movements are driven by chemotactic signals emanating from overlying mesendoderm cells in the hypoblast (Cooper and Virta, 2007). Alternatively, iYSN may direct mesendoderm movements, or iYSN and mesendoderm may move independently from each other(D'Amico and Cooper, 2001). However, direct experimental evidence for any of these hypotheses is still missing.

In this study, we show that the movements of mesendoderm progenitors and iYSN are coordinated during zebrafish gastrulation. Our findings suggest that mesendoderm progenitors direct iYSN movement by modulating cortical flow within the YSL responsible for nuclear transport. Furthermore, we show that the ability of mesendoderm progenitors to guide iYSN movements requires E-cadherin expression, suggesting that adhesive contact between mesendoderm progenitors and YSL is involved.

Fish maintenance and embryo collection were carried out as described(Westerfield, 2000). Details can be provided on request.

To label nuclei in live embryos, 1 mg/ml Histone H1 conjugated to Alexa Fluor 488 (H13188, Invitrogen) was injected in one marginal blastoderm cell at 16-cell stage (1.5 hours post-fertilization; hpf) to visualize mesendoderm progenitors, and in the YSL between high and sphere stage (3.3-4.0 hpf) to mark YSN.

mRNA was synthesized and injected as described(Montero et al., 2005; Westerfield, 2000). Details can be provided on request.

To knock down E-cadherin function, a previously described morpholino oligonucleotide designed against the 5′ end of e-cadherin cDNA(Babb and Marrs, 2004; Montero et al., 2005) (Gene Tools) was injected at the one-cell stage (0.25-0.5 ng).

Fluoresbrite White-Green Microspheres (Polysciences Europe, Germany) with a diameter of 0.5 μm were washed twice in water by brief centrifugation and removal of the supernatant, then incubated for 1 hour in 0.1 μg/μl bovine serum albumin to block unspecific binding, briefly centrifuged and re-suspended in water. To insert microspheres into the YSL, a drop of 0.1 nl was injected at the margin of the YSL at 30% epiboly.

MZoep mutant embryos were used as host embryos, and wild-type embryos were used as donor embryos. Between sphere and dome stage (4.0-4.3 hpf), dechorionated embryos were transferred into an agarose chamber. About 20 cells from a donor embryo were transferred into the blastoderm margin of a host MZoep embryo.

For cytochalasin treatment, dechorionated 30-50% epiboly stage (5 hpf)embryos were incubated in 1-2 μg/ml cytochalasin B (Sigma-Aldrich, Germany)in 1× Danieau's buffer for 1-2 hours at 31°C. After treatment,embryos were mounted in low melting point (LMP) agarose containing cytochalasin. For phalloidin experiments, we injected into the YSL 100 pg Lifeact-GFP mRNA at the 1000-cell stage (3 hpf), 500 pg Histone-Alexa 488 at sphere stage (4 hpf) and 20 pg Amino-Phalloidin, Hydrochloride (Cat. No. ALX-350-266-M001, Alexis Biochemicals) at 50% epiboly.

Embryos were sectioned and immunostained as described(Montero et al., 2005). Details can be provided on request.

For analysis of the YSL ultrastructure, transmission electron microscopy was performed as described (Montero et al., 2005). Dechorionated embryos were fixed in 2% gluteraldehyde and 0.5% paraformaldeheyde in 0.1 M phosphate buffer overnight at 4°C.

Two-photon imaging was performed as described(Ulrich et al., 2003). Details can be provided on request.

To analyze YSN and microsphere movements in two dimensions (2D), we used previously described Motion Tracking software(Helenius et al., 2006; Rink et al., 2005). Net dorsal speed was calculated by dividing the net displacement in the x-axis by the total time. We used the emerging notochord and neural tube as anatomical landmarks to determine the embryonic midline.

To analyze nuclear and cell movements in three dimensions (3D), we implemented an automated tracking algorithm, allowing the estimation of nuclear trajectories without user interaction(Stühmer, 2007).

In this section we refer to a cell, nucleus or bead generically as an`element' for convenience, as the same algorithm is used to quantify movement of all three. The velocity of element i at time point t_k can be estimated from the displacement between two consecutive timepoints as

where is the position of the center of element i at time t. To quantify the correlation between the movement of two elements, we calculated the instant similarity measure (Costa Lda et al.,2005). The similarity between the motion of element i and a neighboring element j at time t can be defined as the normalized internal product of the corresponding velocity vectors, yielding the cosine of the smallest angle between these vectors:

Elements moving with parallel velocity give the maximum similarity of 1.0,whereas elements moving in different directions give smaller values. A value of -1.0 indicates movement in opposite directions.

Because the image segmentation algorithm that is used for automated tracking in 3D depends on the image resolution, estimated positions can adopt only discrete values. The resulting quantization error is especially large in the z-direction. To reduce this effect of quantization noise, the positions are filtered along the trajectory with a mean filter kernel that averages over the four previous and future time points.

To analyze the movement of iYSN in the presence of mesendoderm progenitors in our transplantation experiments, we calculated the normalized distance difference Δd_ij(t) between each nearest-neighbor iYSN/mesendoderm progenitor pair (i,j) at consecutive time points.

where A value of zero indicates that the distance stays constant over the time interval. Positive or negative values indicate that the iYSN and progenitor are moving towards or away from each other, respectively.

The flow fields were computed with the FlowJ plugin for ImageJ (NIH, USA)(Abramoff et al., 2000), using the Lucas-Kanade method (Lucas and Kanade,1981). Flow vectors where drawn using a custom-built vector field visualization plugin for ImageJ that exports the computed vectors into SVG(Scalable Vector Graphics) format.

By investigating the constitutive properties of the cortex and the length and time scales involved in progenitor migration, we develop a suitable physical description of YSN dynamics. The cytoskeleton is viscoelastic,displaying elastic behavior on short time scales and viscous behavior on long time scales. The time scale for elastic relaxation is system dependent, though typical values are of the order of seconds to minutes(Pullarkat et al., 2007), e.g.≈40 seconds for chick fibroblasts(Thoumine and Ott, 1997). The time scale for cortical flow induced by a transplanted patch of progenitors is given by a/U_p≈50 minutes, where a≈50 μm is the radius of a progenitor patch and U_p≈1 μm/minute is the speed with which the progenitor patch moves. Since the time scale for cortical flow is much longer than that for elastic relaxation, the cortex may be described as an incompressible viscous fluid for the present purpose. In general, the flow profile is a function of the Reynolds number (the ratio of inertial to viscous forces) and the system geometry (Landau and Lifshitz, 1987). We overestimate the Reynolds number as Re=ρU_pa/η≈10^-5, where ρis the density, which we take to be that of water, and η is the viscosity,which we underestimate to be 10× that of water, approximately the measured viscosity of cytoplasm in tissue culture cells(Luby-Phelps, 2000) (the viscosity of the cortex is expected to be much higher). To specify the geometry, we model the cortex as an infinite 2D fluid in the plane immediately below the progenitor patch. We specify that the fluid in contact with the patch moves with a velocity U, whereas the rest of fluid is free to move, satisfying the boundary conditions that the velocity is U at the edge of the patch and zero infinitely far away. This is equivalent to flow past a circle, for which approximate solutions for the flow profile are known for low Re (Lamb,1932; Van Dyke,1975). Importantly, in the low Re regime, the velocity decays from the progenitor patch very slowly as log r, r being the distance from the patch. Finally, we verify that the YSN move along with the cortical flow (advectively), as opposed to diffusively, by estimating the Péclet number, Pe=Ua/D, the ratio of diffusive to advective time scales, where D is the diffusivity. To estimate D, we use the Stokes-Einstein-Sutherland relation with an effective viscosity of ≈1 Pa-s, interpreted from measurements of microspheres in fibroblasts (Luby-Phelps,2000), giving D≈10^-4 μm²/s. This is an underestimation of D, as the YSN are larger than the 0.16μm beads described previously(Luby-Phelps, 2000), resulting in greater effective viscosity. Nevertheless, we obtain Pe≈50,which means that advection dominates diffusion and the YSN move with the progenitor-induced cortical flow.

Previous studies in zebrafish have suggested that iYSN undergo convergence movements similar to mesendoderm progenitors overlying the YSL during gastrulation (D'Amico and Cooper,2001). To quantitatively determine the extent to which iYSN and mesendoderm convergence movements are coordinated, we recorded timelapse movies of paraxial and lateral regions of gastrulating zebrafish embryos in which we fluorescently labeled iYSN and nuclei of overlying mesendoderm progenitors (Fig. 1A-C). To analyze and compare the trajectories of iYSN and mesendoderm cells, we developed an image analysis software that allows automated tracking of nuclei in three dimensions (3D) over time (Fig. 1C-F; Movie 1) (Stühmer,2007). We quantified movement coordination between iYSN and mesendoderm progenitors by determining the similarity between the movements of mesendoderm cells and iYSN at short (0-40 μm), medium (40-80 μm) and long (80-120 μm) distances. The similarity value corresponds to the cosine of the smallest angle between two velocity vectors(Costa Lda et al., 2005). Thus, cells and iYSN moving in the same direction will have high similarity values (close to 1), whereas cells and iYSN moving in opposite directions will have low similarity values (close to -1). We found that for short distances 52% of values were higher than 0.5, indicating high movement similarity(Fig. 1G). The similarity values were only slightly reduced for medium and long distances(Fig. 1H,I), suggesting that movement coordination of iYSN and mesendoderm slowly decays with increasing distance. We also determined movement similarity among iYSN and among mesendoderm progenitors, and found a high degree of movement similarity in both cases (see Fig. S1 in the supplementary material). These observations indicate that cellular and nuclear movements are highly coordinated not only between the mesendoderm and YSL, but also within each layer. In addition, we observed that mesendoderm progenitors move faster than the underlying iYSN[v(mesendoderm)=1.46 μm/minute; v(iYSN)= 1.06μm/minute; P<0.05, n=6 embryos]. Taken together, our observations extend previous findings(D'Amico and Cooper, 2001) and show that iYSN and mesendoderm progenitors exhibit highly coordinated convergence movements over long length scales during gastrulation.

To obtain insight into the coordination between iYSN and mesendoderm progenitor movements, we investigated how iYSN move within the YSL. It has previously been shown (Solnica-Krezel and Driever, 1994) that the YSL contains a dense network of microtubules where iYSN are embedded. To test whether microtubules might be involved in iYSN movements, we recorded timelapse movies of the microtubule network by expressing a GFP fusion of the microtubule-associated protein Tau in the YSL (Geldmacher-Voss et al.,2003; Kaltschmidt et al.,2000) (see Movie 2 in the supplementary material). The movement of labeled microtubules in relation to iYSN was analyzed by determining the microtubule flow pattern (for details, see Materials and methods) and comparing it with the movement of iYSN therein. We found that during later stages of gastrulation (8-10 hpf), the YSL microtubule network converges towards the dorsal side, similar to the iYSN(Fig. 2A,D), suggesting that iYSN move together with the microtubule network rather than along microtubules. This view was supported by our observation that the movements of the plus ends of microtubules, visualized by expressing a GFP fusion of the microtubule plus end-binding protein 3 (EB3)(Nakagawa et al., 2000; Stepanova et al., 2003) in the YSL, did not exhibit any preferential orientation within the YSL and that iYSN did not show any obvious plus end-directed movements(Fig. 2B,E; see Movie 3 in the supplementary material). To determine whether convergence movements are restricted to iYSN and microtubules, we recorded timelapse movies of cortical actin movements, visualized by expressing Lifeact-GFP(Riedl et al., 2008) (see Movie 4 in the supplementary material). Similar to the iYSN and the microtubule network, cortical actin exhibited pronounced convergence movements(Fig. 2C,F), indicating that the actin and microtubule cytoskeleton undergo convergence movements together.

These observations suggest that iYSN converge passively together with the YSL cytoskeleton, rather than independently within the YSL. To test whether there is indeed cortical flow within the YSL that could account for iYSN convergence, we injected fluorescently labeled polystyrene microspheres(diameter 0.5 μm) into the YSL and recorded their movements during gastrulation. To analyze these timelapses, we automatically tracked the movements of iYSN and injected microspheres in two dimensions (2D)(Fig. 2G; see Movie 5 in the supplementary material). We observed that the injected beads moved at a similar speed and converged to the dorsal side as their neighboring iYSN(Fig. 2H). This supports the hypothesis that iYSN converge together with the actin and microtubule cytoskeleton by cortical flow towards the dorsal side of the embryo.

The observation that iYSN and mesendoderm progenitors exhibit highly coordinated convergence movements during gastrulation suggests interaction between the two. To test whether mesendoderm progenitors are required for iYSN convergence movements, we investigated iYSN movements in MZoep mutant embryos, which lack most of their mesendoderm progenitors(Gritsman et al., 1999). By analyzing iYSN movement of MZoep embryos in 2D at late gastrulation stages (8-11 hpf), we found that mutant iYSN from lateral and ventral regions of the gastrula exhibited strongly reduced convergence movements compared with wild-type embryos (Fig. 3A,B,D-F; Movies 6-10; see Table S1 in the supplementary material)(D'Amico and Cooper, 2001). To quantify the degree of convergence of iYSN movements, we calculated their net speed along the dorsoventral axis. We found that, in MZoep, the net dorsal speed is significantly lower than in wild type(Fig. 3C; see Table S1 in the supplementary material). By contrast, longitudinal YSN movements in dorsal regions appeared largely unchanged in mutant embryos from early to mid-gastrulation stages (6-8 hpf; data not shown). In late gastrulation stages(8-11 hpf), longitudinal movements of mutant iYSN in dorsal regions occurred predominantly in an anterior direction, whereas dorsal iYSN in wild-type embryos undergo longitudinal movements in both anterior and posterior directions (Fig. 3A,B)(D'Amico and Cooper,2001).

Notably, injection of microspheres into the YSL of mutant embryos revealed that similar to iYSN movements, the convergent cortical flow was strongly reduced in MZoep mutants, suggesting that mesendoderm influences iYSN convergence movements by modulating cortical flow (see Fig. S2 and Movie 11 in the supplementary material). To exclude the possibility that Oep protein itself is needed within the YSL for iYSN movements independently of the presence or absence of mesendoderm progenitors, we injected oep-flag mRNA (Zhang et al., 1998)into the YSL of MZoep mutant embryos. We observed no rescue of the MZoep iYSN mutant phenotype in the YSL-injected embryos (see Fig. S3,Table S1 and Movie 12 in the supplementary material).

Endoderm progenitors have previously been suggested to move in close contact to the YSL plasma membrane (Warga and Nusslein-Volhard, 1999). To test whether endoderm progenitors are specifically required for iYSN convergence, we analyzed casanova(cas) mutant embryos, which lack endoderm but not mesoderm progenitors (Alexander et al.,1999; Dickmeis et al.,2001; Kikuchi et al.,2001). In contrast to the situation in MZoep mutant embryos, iYSN convergence movements appear largely unaffected in casmutants (see Fig. S4 and Movie 13 in the supplementary material), indicating that endoderm progenitors are not specifically required for iYSN convergence. Taken together, these data suggest a crucial function of mesoderm progenitors in iYSN convergence movements.

Although these findings demonstrate a requirement for mesendoderm progenitors in iYSN convergence, they do not indicate whether mesendoderm progenitors are also sufficient to direct iYSN convergence movements. To address a possible instructive function of mesendoderm for iYSN movement, we transplanted a small number of blastodermal cells into the lateral side of MZoep mutant embryos and monitored the movements of iYSN relative to the position of the transplanted cells. The majority of these cells underwent ingression and expressed mesendodermal markers, as previously reported(Carmany-Rampey and Schier,2001) (data not shown), indicating that they adopt a mesendodermal fate. From mid-gastrulation stages onwards, the transplanted cells converged toward the dorsal side (Fig. 4;see Movies 14 and 16 in the supplementary material). Strikingly, iYSN in the transplanted side of MZoep mutant embryos converged together with the transplanted cells, whereas iYSN on the opposite, non-transplanted side continued to show only very reduced convergence movements(Fig. 4A-E; see Movies 15 and 16 in the supplementary material). This was confirmed when we determined the net dorsal speed of iYSN movement in the presence and absence of transplanted cells (Fig. 4C; see Table S1 in the supplementary material). To analyze the convergence movements of iYSN relative to the transplanted cells in more detail, we imaged the transplanted region at higher magnifications and tracked their movements in 3D(Fig. 4F,G; see Movie 14 in the supplementary material). We found that, as in wild-type embryos, iYSN and transplanted cells move with high similarity values (see Materials and methods) at short as well as at long distances(Fig. 4H,I). These results suggest that convergence movements of mesendoderm progenitors are able to induce convergence movements of underlying iYSN.

Questions remain as to the molecular and cellular mechanisms by which mesendoderm progenitors interact with the YSL to direct iYSN convergence. To determine whether physical contact between YSL and mesendoderm progenitors is required, we visualized the contact zone between mesendoderm progenitors and iYSL using transmission electron microscopy. We found that the plasma membranes of the iYSL and overlying mesendoderm progenitors were often directly adjacent to each other and showed electron-dense structures resembling adherens junctions (Fig. 5A).

E-cadherin (cdh1 gene in zebrafish) has previously been shown to be highly expressed in the YSL and mesendoderm progenitors, and to represent a key component regulating cell-cell interaction both within and between the forming germ layers during gastrulation(Fig. 5B)(Babb and Marrs, 2004; Kane et al., 2005; Krieg et al., 2008; McFarland et al., 2005; Montero et al., 2005; Shimizu et al., 2005). To determine whether E-cadherin is also required for the coordination of iYSN and mesendoderm progenitor convergence movements, we interfered with E-cadherin expression by injecting a previously characterized morpholino oligonucleotide(MO) targeted against e-cadherin and monitored iYSN and mesendoderm convergence movements in 3D over time. We analyzed the movements of iYSN and mesendoderm at mid-gastrulation stages (7-8 hpf)(Fig. 5C; see Movie 17 in the supplementary material) and found that the similarity values for movement coordination between iYSN and mesendoderm were reduced in morphant embryos compared with wild-type embryos. In the morphants, only 35% of the values were higher than 0.5, in contrast to the 50% observed in the wild type (compare Fig. 5E with Fig. 1J). In addition, we observed that mesendoderm progenitors moved slightly slower in e-cadherin morphant embryos than in wild-type embryos (see Table S2 in the supplementary material). To verify that the effects in the morphant embryos are specifically due to the loss of E-cadherin function, we also analyzed previously described e-cadherin loss-of-function mutants(weg) (Kane et al.,1996; Kane et al.,2005) (Fig. 5D; see Movie 18 in the supplementary material). In weg mutant embryos, the similarity values for movement coordination between iYSN and mesendoderm were also diminished (35% of the values higher than 0.5)(Fig. 5F), congruous with our observations in e-cadherin morphant embryos. Taken together, these data suggest that E-cadherin expression is needed to coordinate iYSN with mesendoderm progenitor cell movements.

E-cadherin is dynamically linked to the actin cortex of cells(Adams and Nelson, 1998; Gumbiner, 2005). This,together with our previous observation that iYSN move together with the actin and microtubule cytoskeleton by cortical flow, suggests that E-cadherin coordinates mesendoderm progenitor and iYSN movements by interacting with the YSL actin cytoskeleton. To determine whether partial disruption of the actin cytoskeleton can phenocopy loss of E-cadherin function in this process, we incubated embryos with the actin depolymerizing drug cytochalasin B during gastrulation, which has previously been shown to effectively interfere with the actin cytoskeleton in the zebrafish gastrula(Zalik et al., 1999) (see Movies 19 and 4 in the supplementary material). When we analyzed iYSN and mesendoderm movements in cytochalasin B-treated embryos(Fig. 6A,C; see Movie 20 in the supplementary material), we found reduced similarity values for movement coordination between iYSN and mesendoderm when compared with wild-type embryos(only 34% of the values were higher than 0.5)(Fig. 6E), similar to the situation in e-cadherin mutants and morphants. To test whether polymerized actin is required locally within the YSL, we injected the actin-polymerizing drug phalloidin (Cooper,1987; Wehland et al.,1977) directly into the YSL. We found that phalloidin effectively interfered with actin distribution within the YSL (compare Fig. 6B with Fig. 2C; see Movies 21 and 4 in the supplementary material) and that the similarity values for movement coordination between mesendoderm and iYSN were reduced in phalloidin-injected embryos compared with wild-type embryos (only 41% of the values were higher than 0.5) (Fig. 6D,F; see Movie 22 in the supplementary material). This suggests that polymerized actin within the YSL plays a crucial function in this process. Taken together, these results suggest that mesendoderm control of iYSN convergence movements requires E-cadherin expression and that the YSL actin cytoskeleton plays an important function therein.

The observation that mesendoderm progenitors are required and sufficient to direct iYSN convergence movements suggests a function of mesendoderm progenitors in iYSN movements. It has previously been hypothesized that YSN movements might be driven by chemotactic signals emanated by mesendodermal cells in the hypoblast (Cooper and Virta,2007). To test this hypothesis, we quantified the distance difference between each iYSN migrating ahead of the transplanted cells(considering the direction of movement of the cells) and its nearest neighboring transplanted mesendoderm progenitor. If the iYSN have the tendency to move toward the transplanted cells, as a result of a chemotactic signal,their distance should decrease over time (for details, see Materials and methods). We found that the normalized distance difference between consecutive timepoints was equally distributed around 0 (exact mean value=-0.003±0.027) (Fig. 7A), indicating that iYSN and transplanted cells keep a constant distance between each other. These results argue against mesendoderm progenitors directly attracting iYSN, e.g. through the release of a chemotactic signal.

Instead of attracting iYSN by chemotactic signaling, converging mesendoderm progenitors might direct iYSN convergence by creating a dorsal-directed cortical flow within the underlying YSL cytogel. Our finding that E-cadherin-mediated adhesion between mesendoderm and YSL is involved in this process suggests that converging mesendoderm progenitors create this cortical flow by dragging the YSL plasma membrane and associated cortical cytoskeleton. To investigate the plausibility of this hypothesis, we considered the force balances for this system. First, we noted that for mesendoderm progenitors to exert a force on the YSL, they must have a substrate for migration other than the YSL itself. Consistent with this, we have previously shown that anterior axial mesendoderm progenitors can use the overlying non-internalizing ectoderm progenitors as a substrate on which to migrate(Montero et al., 2005). Next,we analyzed the propagation of forces exerted by a transplanted patch of cells on the YSL cortex. Modeling the cortex as a 2D viscous fluid and the patch of cells as a translating solid material, we found (1) that the forces exerted by the patch are long range, inducing cortical flow even far away from the progenitors (see Fig. 7B for the flow profile), and (2) that the YSN are advected with this induced flow(for details, see Materials and methods). Indeed, the convergence movement of the YSN in our transplantation experiments displays this behavior, supporting the hypothesis that transplanted mesendoderm progenitors induce iYSN convergence movements by modulating cortical flow within the YSL.

In zebrafish, iYSN and overlying mesendoderm progenitors have been suggested to undergo similar convergence movements during gastrulation(D'Amico and Cooper, 2001). Here, we show that mesendoderm progenitors control iYSN convergence movements by modulating cortical flow within the YSL responsible for iYSN transport. Furthermore, we provide evidence that the coordination between mesendoderm and iYSN convergence movements depends on E-cadherin expression, suggesting that adhesive contact between mesendoderm and YSL is involved.

Previous studies have demonstrated that YSN undergo both epiboly and convergence and extension movements(D'Amico and Cooper, 2001; Solnica-Krezel and Driever,1994; Trinkaus,1951). Furthermore, interfering with the actin and microtubule cytoskeleton of the YSL has been shown to disrupt epiboly movements of the YSL(Cheng et al., 2004; Hsu et al., 2006; Koppen et al., 2006; Solnica-Krezel and Driever,1994; Zalik et al.,1999), suggesting a crucial function of the cytoskeleton for YSN epiboly. By contrast, the role of the YSL cytoskeleton for iYSN convergence and extension movements had not yet been addressed. Our observation that iYSN,together with the actin and microtubule cytoskeleton converge by cortical flow within the YSL indicates that iYSN are not actively transported along microtubules, as has been demonstrated for other types of nuclear migrations(Morris, 2000; Morris, 2003; Reinsch and Gonczy, 1998; Xiang and Fischer, 2004). Instead, the actin and microtubule cytoskeleton appear to function simply as a medium in which the iYSN are entrained and advected. This notion is further supported by our observation that beads injected into the YSL undergo convergence movements similar to neighboring iYSN. These findings are reminiscent of previous experiments in the rainbow trout, demonstrating that chalk particles implanted into the YSL cytoplasm undergo pronounced convergence movements (Long,1980).

Observations in Drosophila suggest that cytoplasmic streaming is involved in nuclear distribution within the syncytial preblastoderm and in mixing and dispersal of cytoplasmic components within the oocyte(Dahlgaard et al., 2007; Serbus et al., 2005; Theurkauf, 1994; von Dassow and Schubiger,1994). The mechanisms underlying cortical flow within the YSL are currently unknown. However, it is conceivable that, analogous to the situation in the Drosophila preblastoderm(von Dassow and Schubiger,1994), acto-myosin contraction combined with local changes in actin polymerization within the YSL controls this streaming. Whether iYSN converge solely by cortical flow or whether other transport mechanisms also contribute to iYSN convergence is presently unclear. Our preliminary observations of iYSN being stretched and pulled apart (not shown) suggest that there are cytoskeletal forces that pull on the nuclear envelope and possibly contribute to iYSN movements.

During gastrulation, highly organized cell and tissue movements lead to the formation of the embryonic body axis(Solnica-Krezel, 2005; Stern, 1992). Consistent with this and previous observations (D'Amico and Cooper, 2001), we found that in gastrulating zebrafish embryos cellular and nuclear movements of mesendoderm and YSL are highly coordinated. Furthermore, our observations suggest that iYSN convergence movements depend on mesendoderm convergence and that mesendoderm directs iYSN convergence movements. Interestingly, YSN have previously been shown to undergo epiboly movements independently of the overlying blastoderm in zebrafish and killifish(Fundulus) (Kane et al.,1996; Trinkaus,1951; Trinkaus,1984). This apparent difference in blastoderm dependence between epiboly and convergence movements of YSN is probably due to different processes controlling these movements. Although YSN epiboly has been proposed to depend on parallel filaments of microtubules polarized towards the vegetal pole (Solnica-Krezel and Driever,1994), we show that iYSN convergence movements are directed by cortical flow through contact with the overlying mesendoderm. Our observation that microspheres injected into the YSL strongly converge to the dorsal side,while showing only very reduced epiboly movements (not shown) points at the interesting possibility that YSN epiboly involves active nuclear migration,whereas iYSN convergence does not.

Previous studies have hypothesized that iYSN movements might be driven by chemotactic signals emanating from mesendodermal cells in the hypoblast(Cooper and Virta, 2007). However, we found no evidence for mesendoderm attracting neighboring iYSN,arguing against the assumption that mesendoderm directs iYSN convergence through chemotactic signaling. By contrast, we provide evidence that iYSN converge by cortical flow within the YSL and that mesendoderm progenitors control this cortical flow through E-cadherin-mediated contact with the YSL. Mesendoderm progenitors most probably establish a physical link to the YSL plasma membrane via E-cadherin, which then enables mesendoderm progenitors undergoing convergence movements to drag the YSL plasma membrane and associated cortical cytoskeleton in the same direction. In such a scenario,E-cadherin-mediated adhesion between mesendoderm and YSL would be required to establish sufficient friction between these layers to allow the effective pulling of the mesendoderm on the YSL. The plausibility of this notion is supported by our hydrodynamic analysis showing that the flow of a viscous fluid induced by a solid patch moving over its surface closely resembles the cortical flow within the YSL induced by mesendoderm progenitors moving over it.

Questions remain about the specific role of coordinated convergence movements of the blastoderm and underlying iYSN during embryogenesis. Clearly,transcription from iYSN is crucial for a variety of different embryonic processes, ranging from mesendoderm cell fate induction to heart progenitor cell migration (Chen and Kimelman,2000; Mizuno et al.,1999; Mizuno et al.,1996; Ober and Schulte-Merker,1999; Rodaway et al.,1999; Sakaguchi et al.,2006). In addition, dorsally restricted expression of the homeobox transcription factor hex within the iYSL has been associated with dorsoventral patterning of the gastrula, suggesting that the spatially regulated transcription of this gene within the YSL is important for its activity (Ho et al.,1999).

Future experiments will have to determine whether iYSN convergence movements are indeed required for blastoderm patterning and/or morphogenesis by specifically interfering with iYSN convergence movements within the YSL.