Rho-regulated Myosin phosphatase establishes the level of protrusive activity required for cell movements during zebrafish gastrulation

Gastrulation is a tightly regulated series of cell rearrangements that establishes the vertebrate body plan(Keller, 2002). A major driving force in vertebrate gastrulation is convergent extension (CE), a mechanism by which cells move towards the future dorsal side of the embryo and then intercalate between neighboring cells, resulting in an overall dorsoventral narrowing (convergence) and anteroposterior lengthening(extension) of the embryo (Hammerschmidt and Wedlich, 2008; Keller,2002; Solnica-Krezel,2006). Zebrafish lateral mesodermal cells undergo a series of characteristic cell behavior changes during this process. Early in gastrulation, lateral mesodermal cells are highly motile and extend numerous short-lived and randomly oriented lamellipodia and bleb-like protrusions(Concha and Adams, 1998; Sepich et al., 2005; Solnica-Krezel, 2006). As gastrulation proceeds, the cells become more highly polarized and stabilize their protrusions with a dorsal orientation(Concha and Adams, 1998). When the cells approach the dorsal midline they reduce their migratory and protrusive behaviors (Concha and Adams,1998), pack tightly together and undergo intercalation(Yin et al., 2008). Interestingly, several studies have shown that convergence and extension are genetically separable in zebrafish, indicating that some aspects of these processes are unique (Bakkers et al.,2004; Formstone and Mason,2005; Glickman et al.,2003; Myers et al.,2002; Weiser et al.,2007; Yin et al.,2008).

One of the major regulators of CE movements is the non-canonical(β-catenin-independent) Wnt pathway, the vertebrate version of the Drosophila planar cell polarity (PCP) pathway(Mlodzik, 2002). Interestingly, both gain- and loss-of-function of regulators of PCP pathway members results in severe gastrulation defects(Carreira-Barbosa et al., 2003; Carreira-Barbosa et al., 2009; Heisenberg et al., 2000; Jessen et al., 2002; Wallingford et al., 2000),demonstrating that proper cell movements during gastrulation require precise and tightly balanced signaling. The non-canonical Wnts activate several downstream PCP effectors, including the small GTPase Rho, which mediates many of the effects of the PCP pathway during gastrulation(Habas et al., 2003; Marlow et al., 2002) by affecting cell polarity, cell shape changes and movement(Etienne-Manneville and Hall,2002; Hall, 1998). Although most work on small G-proteins during gastrulation has focused on the role of non-canonical Wnts in controlling their activity(Habas et al., 2003), several other regulators of Rho are required for proper gastrulation, including the scaffolding protein Gravin (Akap12), which is essential for mesodermal cells to convert from amoeboid cell movement to intercalative cell behaviors at the end of gastrulation in order for extension to occur(Weiser et al., 2007; Weiser et al., 2008). RhoA activates the Formins, which control actin polymerization, and Rho-dependent kinase (Rock), which phosphorylates the regulatory light chain of non-muscle myosin (Mlc) (Etienne-Manneville and Hall,2002). Both Formin and Rock signaling are required for normal CE movements (Marlow et al.,2002; Sato et al.,2006; Zhu et al.,2006).

In this work, we investigated the role of the Myosin phosphatase complex in CE movements, exploiting the advantages of zebrafish to study its role in cell behavior in vivo. The Myosin phosphatase complex is composed of a catalytic subunit of Protein phosphatase 1 (Ppp1cb) and a targeting subunit called Mypt1(Ppp1r12a) (Ito et al., 2004; Matsumura and Hartshorne,2008). Several sites on Mypt1 are regulated by phosphorylation(Ito et al., 2004; Matsumura and Hartshorne,2008; Muranyi et al.,2002; Wooldridge et al.,2004), including Rock-mediated inhibitory phosphorylation(Feng et al., 1999; Kimura et al., 1996). In addition, a family of inhibitory proteins including Cpi-17 (Ppp1r14a) and Phi-1 (Ppp1r14b) (Eto et al.,1999; Eto et al.,2000) interact with Mypt1 and block Mlc dephosphorylation(Eto et al., 2004; Eto et al., 2000).

Mypt1 is an important regulator of cell migration, adhesion and retraction. Knockdown of Mypt1 in mammalian cells causes excessive cell contractility(Eto et al., 2005). Mypt1(Mbs) loss in Drosophila results in disorganized epithelial movements, cell shape defects and failure of dorsal closure(Mizuno et al., 2002; Tan et al., 2003), whereas loss of Mypt1 (MEL-11) function in C. elegans disrupts cell movements during gastrulation (Piekny et al.,2003; Wissmann et al.,1999). In contrast to these systems, the role of Mypt1 in early vertebrate development is unclear. The Mypt1 mouse knockout is embryonic lethal, but the phenotype remains uncharacterized(Okamoto et al., 2005). Zebrafish mypt1 mutants develop fairly normally although they fail to form the liver owing to incorrect placement of Bmp-expressing cells(Huang et al., 2008). We reasoned that the lack of earlier defects in zebrafish might be due to maternal mypt1 mRNA, and we show here that Mypt1 is essential for CE movements. Mypt1 acts cell-autonomously to control protrusive activity, which is essential for intercalation and extension, whereas it acts non-autonomously to regulate convergence movements. Importantly, inhibitory phosphorylation mediated by Rock downstream of non-canonical Wnt signaling tightly regulates Mypt1 activity, establishing Mypt1 as a key downstream mediator of Wnt/Rho signaling. With the similarity between amoeboid cell movements in gastrulation and in metastatic cells, our work suggests that Mypt1 might function as a tumor suppressor regulating Rho-dependent amoeboid cell behavior in metastasis.

Wild-type WIK/AB zebrafish (Danio rerio) embryos were obtained through natural spawning and were maintained, injected and staged according to established procedures (Westerfield,1993).

Human MYPT1, zebrafish cpi-17 and zebrafish phi-1 clones were obtained from Open Biosystems, and constitutively active mlck (mylk) was a kind gift of Dr Erez Raz(Blaser et al., 2006). Each clone was inserted into pCS2 for the production of mRNA (details of plasmid construction are available upon request). Embryo injection, mRNA synthesis and in situ hybridization were performed as described(Weiser et al., 2007).

HEK 293T cells, L-cells and Wnt5a/L-cells (obtained from the ATCC) were maintained in DMEM containing 10% FBS and cultured in 5% CO₂. Cells were transfected as described (Weiser et al., 2008). Conditioned media were obtained by culturing L-cells or Wnt5a/L-cells in DMEM containing 10% FBS without selective antibiotics for 5 days. The medium was then removed and placed on the HEK 293T cells for 4 hours. The following antibodies were used: anti-Mypt1 (Cell Signaling,1:1000), anti-phospho-T696 Mypt1 (Upstate, 1:500) and mouse anti-Actin(Chemicon, 1:1000).

Embryos were injected and imaged at bud stage for analysis of convergence and/or extension as described (Weiser et al., 2007).

Time-lapse recordings were performed with a Zeiss Axiovert 200M microscope using AxioVision 4 software, DIC optics and a 40× objective at 28°C as described (Weiser et al.,2007). Cell tracking experiments were performed using the ImageJ manual-tracking program to record tracks of cell movement and calculate dorsal migration. Net dorsal migration was calculated by determining the distance between the location of each cell during the first and last frame of a 15-minute time-lapse recording. Total movement was calculated as the sum of the distances between 15 1-minute time points. Persistence is defined as the net migration divided by the total migration. The number of bleb-like protrusions per cell was scored by observing individual mesodermal cells (at least 25 in total from three separate embryos) two to three cell layers away from the notochord through a 15-minute time-lapse movie, and reported as blebs per cell per hour.

Embryos were treated with the Myosin inhibitor blebbistatin (50 μM),with Rho kinase inhibitor III (200 μM) or with DMSO as a control as described (Weiser et al.,2007).

Myc-tagged mypt1 mRNA (100 pg) was injected into one-cell-stage embryos. At the beginning of gastrulation, 300 embryos of each condition were collected, dechorionated, deyolked and lysed. The myc-tagged Mypt1 was immunoprecipitated using anti-myc antibody (Covance). Fifty embryos equivalents were loaded on an acrylamide gel and the level of phosphorylation determined by western blot.

One-cell-stage embryos to be used as donors were injected with 1%fluorescein dextran or with 1 ng mypt1 MO and 1% rhodamine dextran. Fifteen to twenty-five cells of each control and morphant cells were removed from sphere-stage embryos and injected into shield-stage host embryos near the margin, ∼90 degrees from the shield.

Statistical calculations were performed using Microsoft Excel. Error bars represent s.d. and P-values were determined using Student's t-test, comparing each sample with the uninjected control at the same stage.

To inhibit maternal Myosin phosphatase we designed a translation-blocking morpholino to mypt1 (mypt1 MO) and also tested ectopic expression of the specific Mypt1 inhibitors Cpi-17 and Phi-1(Eto et al., 2005; Eto et al., 2000; Totsukawa et al., 2000). mypt1 MO (1 ng) reduced endogenous Mypt1 protein levels(Fig. 1I). After 48 hours, mypt1 morphant embryos had a dramatically shorter body axis, wavy notochord, broad somites and serious defects in heart morphology(Fig. 1B) as compared with control embryos (Fig. 1A). This phenotype differs considerably from that of embryos lacking only zygotic mypt1, which show no defect in body axis elongation(Huang et al., 2008). Similar morphogenetic defects were seen in embryos injected with cpi-17 mRNA(Fig. 1C) or phi-1mRNA (not shown). Injection of constitutively active myosin light chain kinase mRNA (ca-mlck), which we predicted would cause the same biochemical effect as a loss of Myosin phosphatase, produced the same phenotype (Fig. 1D). Injection of cpi-17 or ca-mlck resulted in complete cyclopia(Fig. 1C,G), whereas injection of phi-1 or the mypt1 MO resulted in partial cyclopia and small eyes (Fig. 1B and data not shown). A truncated body axis is a common phenotype in embryos with defects in CE movements (Jessen et al.,2002; Solnica-Krezel et al.,1996; Topczewski et al.,2001), whereas cyclopia is often seen in embryos with defects in prechordal plate migration (Heisenberg et al., 2000; Marlow et al.,1998). Interestingly, injection of 100 pg of human MYPT1mRNA has no effect on zebrafish development(Fig. 1E), as reported previously (Huang et al.,2008). However, injection of 100 pg of human MYPT1 mRNA rescued the phenotype of mypt1 morphant embryos(Fig. 1F,J); human MYPT1 is 75%identical and 85% similar to zebrafish Mypt1 and the mRNA matches the MO at only 9 of 25 positions. In addition, co-injection of mypt1 MO with either ca-mlck or cpi-17, each at a dose that does not cause a CE defect when injected alone, resulted in a clear synergistic CE phenotype(65 of 71 embryos for cpi-17 plus mypt1 MO; 52 of 61 embryos for ca-mlck plus mypt1 MO). To confirm that the short body axis phenotype observed at 48 hours was caused by a gastrulation defect,embryos were scored for elongation defects at the end of gastrulation and again at 48 hours. The penetrance of the phenotype was the same at both time points (see Fig. S1 in the supplementary material). The results of all these experiments demonstrate that decreasing Myosin phosphatase activity and increasing Myosin contractility results in specific defects in body axis formation.

To determine the underlying morphogenetic defects that cause the morphant phenotype, in situ hybridization was performed on mypt1 morphants, cpi-17, ca-mlck and control embryos. Morphants(Fig. 2B,D,F; 44 of 52 embryos)and cpi-17 (Fig. 2G;38 of 49 embryos) and ca-mlck(Fig. 2H; 32 of 51 embryos)embryos displayed short, broad notochords and broadened presomitic domains as compared with the wild type (WT) (Fig. 2A,C,E). Anterior migration of the prechordal plate was also defective, as indicated by the mislocalization of hgg1(ctsl1b) staining (39 of 52 embryos). This phenotype indicates that CE is defective during gastrulation in both axial and paraxial tissues. Interestingly, other morphogenetic movements appeared normal, with no apparent defects in epiboly in morphant or cpi-17 embryos (data not shown). This pattern of mislocalized prechordal plate and shortened broadened notochord and presomitic mesoderm is a hallmark of defects in convergence and extension during zebrafish gastrulation(Heisenberg et al., 2000; Marlow et al., 2002; Topczewski et al., 2001). Given the similarity in phenotype between the morphant and mRNA-injected embryos, the mypt1 MO was used for the rest of this work.

We set out to directly assay the ability of lateral cells to both converge dorsally and extend in the anterior-posterior direction using photo-uncaging(Sepich and Solnica-Krezel,2005). Groups of cells were uncaged at shield stage, ∼90 degrees from dorsal (Fig. 3A,D). Four hours later (bud stage), the embryos were analyzed to determine the extent of cell migration. Lateral cells displayed severe defects in both convergence and extension (Fig. 3E,F) compared with controls(Fig. 3B,C). The cells were unable to migrate dorsally to the extent seen in controls, and they also failed to undergo full extension (Fig. 3G,H). Interestingly, morphant cells displayed a highly disorganized distribution pattern (Fig. 3F) as compared with the tightly packed and organized cells in the controls (Fig. 3C). This disorganization has been observed in embryos that lack the proper intercalative cell behaviors of presomitic mesoderm late in gastrulation(Weiser et al., 2007). Anterior migration of the axial mesoderm was also affected(Fig. 3I), consistent with the observed head defects.

We next used cell transplantation to determine whether the CE defects in mypt1 morphants are cell-autonomous or non-autonomous. WT and morphant cells, separately labeled, were transplanted together into WT embryos at shield stage (Fig. 3J). The embryos were then allowed to complete gastrulation and were assayed for convergence and extension. Transplanted control cells underwent both convergence and extension and formed a narrow extended belt of intercalated cells on the dorsal side of the embryo (green in Fig. 3K). Surprisingly, mypt1 morphant cells were able to migrate dorsally to the same extent as WT cells; however, these cells failed to undergo extension properly (red in Fig. 3K; 13 of 15 transplants). Morphant cells remained primarily in a clump with other morphant cells and failed to undergo intercalation (Fig. 3L). Taken together with the whole-embryo results, these observations demonstrate that Mypt1 has a cell-autonomous role in extension,and a non-autonomous role in convergence. Moreover, our studies further support the emerging view that convergence and extension are mechanistically separable (Bakkers et al.,2004; Glickman et al.,2003; Weiser et al.,2007).

We next focused on the underlying cell behaviors of mypt1morphants to determine which aspect of gastrulation is defective. We analyzed the polarity of mesodermal cells by calculating the length-width ratios of mesodermal cells at bud stage in control embryos compared with embryos injected with the mypt1 MO (Fig. 4A). Control mesodermal cells are highly polarized and elongate at bud stage in both presomitic and axial mesoderm(Fig. 4B,C). Both lateral and axial mesodermal cells were significantly less polarized in morphant embryos(Fig. 4B,C). This clearly indicates that mypt1 morphant cells are less polarized than control cells.

To determine the effect of Mypt1 loss-of-function on cell motility, we analyzed presomitic mesodermal cells just prior to bud stage using time-lapse and DIC optics. Migration speeds and paths for individual cells were calculated and recorded (from 25 cells in total from two embryos of each condition; Fig. 4D,E). Morphant cells moved more slowly than control cells, with a ∼50% reduction in net dorsal migration. Control cells at this stage migrated in a highly polarized manner, whereas morphant cells not only moved more slowly (total migration reduced), but they also changed direction more often [persistence (the ratio of total to net migration) increased from 1.36 to 1.95], further reducing convergence movements (Fig. 4E). These observations indicate that Mypt1 is required for the proper dorsal migration of mesodermal cells, without which convergence is impaired.

Since mypt1 morphants are defective in extension as well as in dorsal migration of paraxial mesoderm (Fig. 3), we hypothesized that mypt1 is required for mediolateral intercalation. To test this, we utilized time-lapse analysis of mesodermal cell shape throughout gastrulation in control and morphant embryos. Control cells during early gastrulation are highly active and extend many bleb and lamellipodial protrusions and migrate in an amoeboid fashion(Fig. 5A and see Movie 1 in the supplementary material). By the end of gastrulation, control cells were densely packed, exhibited few protrusions and were mesenchymal in cell behavior (Fig. 5B and see Movie 2 in the supplementary material). The few protrusions that were observed at this stage were primarily either bleb-like protrusions along the notochordal-paraxial boundary or lamellipodial protrusions that extended between cells. By contrast, mypt1 morphant mesodermal cells extended increased numbers of bleb-like protrusions at both early and late gastrula stages and exhibited amoeboid behaviors throughout gastrulation(Fig. 5C,D and see Movies 3 and 4 in the supplementary material). Control cells averaged less than 20 blebs per hour at early stages and less than four at late gastrula stages, whereas morphant embryos displayed over 50 blebs per hour throughout gastrulation(Fig. 5F). This increased blebbing correlates with defects in dorsal migration, as shown in the dose-response curve (see Fig. S2 in the supplementary material). In contrast to lamellipodia, which display a strong dorsal bias in orientation(Solnica-Krezel, 2006), blebs in both WT and morphant cells appeared to be randomly oriented (data not shown). Interestingly, co-injection of mypt1 mRNA not only rescued axis elongation of mypt1 morphant embryos(Fig. 1), but also rescued blebbing (Fig. 5E,F).

Transplanted morphant cells, which converge but fail to extend(Fig. 3), were also less polarized than transplanted and host WT cells and they extended numerous bleb-like protrusions (Fig. 3M). These results demonstrate that these highly blebbing amoeboid cells are unable to undergo extension, but are fully capable of dorsal convergence. This property is very reminiscent of gravin morphant cells, which, uniquely among the genes involved in CE, are highly blebbing and fail to extend (Weiser et al.,2007). However, whereas gravin morphant cells exhibit normal blebbing activity at the start of gastrulation and fail to shut down blebbing after migration to the dorsal side, mypt1 morphant cells show increased blebbing activity from the start of gastrulation(Fig. 5C). We therefore conclude that while the failure to shut down blebbing prevents intercalation and extension in a cell-autonomous manner, increased blebbing does not prevent convergence as long as the surrounding cells have normal Mypt1 activity.

We sought to determine whether the mypt1 MO phenotype was mediated by effects on acto-myosin contractility and Rho signaling. Rho, acting through Rock, phosphorylates and activates non-muscle myosin(Etienne-Manneville and Hall,2002). We used two potent pharmacological inhibitors of Rho-Myosin signaling, blebbistatin, which inhibits non-muscle myosin, and Rho kinase inhibitor III, a cell-permeable Rock inhibitor, both of which are effective in zebrafish embryos (Berndt et al.,2008; Weiser et al.,2007). Treatment with blebbistatin completely abrogated blebbing in mypt1 morphants (Fig. 6A,D), providing direct evidence that Mypt1 is upstream of Myosin II. Interestingly, blebbing was reduced but not completely eliminated in mypt1 morphants when treated with the Rock inhibitor(Fig. 6B,D and see Movie 5 in the supplementary material). This contrasts with gravin morphants, in which both blebbistatin and the Rock inhibitor eliminate ectopic blebbing(Weiser et al., 2007). Other Mlc kinases are present during zebrafish gastrulation, including Mlck and Dap(Blaser et al., 2006). We suggest that in the absence of opposing Myosin phosphatase activity, these other kinases can elicit sufficient phosphorylation of Mlc to cause some membrane blebbing, even if Rock is inhibited.

Mypt1 is not a constitutively active phosphatase but instead is regulated by many inputs, including binding proteins and phosphorylation at various sites. Since the Rho-Rock pathway plays an essential role in zebrafish gastrulation (Habas et al.,2001; Marlow et al.,2002), we asked whether Rock not only activates Mlc to regulate cell movements, but also targets Mypt1 in embryos. Cell culture experiments have shown that inhibitory phosphorylation of Mypt1 at threonine 696 is induced by Rock, either directly (Feng et al., 1999; Kimura et al.,1996) or by activating Dap kinase (also called Zip kinase), which also phosphorylates this site (Hagerty et al., 2007). Phosphorylation at this site reduces the affinity of Mypt1 for Myosin and thus inhibits the function of the Myosin phosphatase complex (Kimura et al.,1996).

We hypothesized that if Rock regulates Mypt1 then loss of this inhibitory phosphorylation site would render Mypt1 hyperactive during gastrulation. To test this, we mutated T696 to either alanine or aspartic acid to eliminate phosphorylation and create a phosphomimetic, respectively. Injection of 100 pg of mypt1 mRNA (Fig. 7B), or as much as 400 pg (data not shown), had no effect on early development. By contrast, injection of 100 pg of mypt1 T696A mRNA caused a severe CE defect (Fig. 7A,B), as well as a partially cyclopic head. Examination of cells expressing T696A mypt1 revealed inhibition of cell movement and protrusive activity (Fig. 7C). T696A cells displayed a rounded cell morphology, with reduced polarity and migratory ability, resembling cells with reduced acto-myosin contractility. The phosphomimetic mutant T696D, like WT mypt1, failed to elicit a CE defect (Fig. 7B). Importantly,whereas 100 pg of WT mypt1 mRNA was required to rescue the CE defect in mypt1 morphants, just 20 pg of the T696A mRNA produced the same extent of rescue (Fig. 7B). By contrast, mypt1 T696D completely failed to rescue the phenotype(Fig. 7B). These results indicate that mutation of threonine 696 to an alanine renders Mypt1 constitutively active in vivo, as has been observed in vitro(Feng et al., 1999), whereas mutation to aspartic acid renders Mypt1 constitutively inactive. Since WT Mypt1 is much less active than Mypt1 T696A, these results demonstrate that Mypt1 activity is regulated in vivo through residue 696.

The crucial role of non-canonical Wnts in controlling RhoA signaling during gastrulation led us to hypothesize that non-canonical Wnts, working through Rho and Rock, might promote inhibitory phosphorylation of Mypt1 at position 696. To test this, we first turned to an in vitro system. WNT5A- and control-conditioned media were added to HEK 293T cells transfected with myc-tagged Mypt1, and the level of Mypt1 phosphorylation was detected using an anti-phospho-T696 antibody. T696 exhibits basal levels of phosphorylation in HEK 293 cells; however, upon treatment with Rock inhibitor, the level of phosphorylation was considerably reduced, indicating that much of this phosphorylation is Rock dependent (Fig. 7E). Treatment with human WNT5A-conditioned medium resulted in a greater than 2-fold increase in Mypt1 phosphorylation. Expression of constitutively active RhoA, or incubation with the phosphatase inhibitor okadaic acid, increased Mypt1 phosphorylation to a similar extent(Fig. 7E). Importantly,inhibition of Rock blocked the increase in Mypt1 phosphorylation caused by WNT5A, and reduced phosphorylation to below control levels.

We next determined whether Mypt1 phosphorylation is regulated in zebrafish. We injected 100 pg of myc-tagged mypt1 mRNA into embryos, a level that causes no phenotypic effects on embryo development, and determined the level of phosphorylation by western blot. In zebrafish, basal Mypt1 phosphorylation is Rock-dependent and was stimulated by expressing constitutively active RhoA, or the non-canonical Wnts Wnt11 or Wnt5b(Pipetail) (Fig. 7F), providing clear in vivo evidence that Mypt1 phosphorylation is regulated by non-canonical Wnt signaling in zebrafish embryos. In addition, co-injection of mypt1 T696A with wnt11 mRNA rescued the CE phenotype in wnt11 embryos (Fig. 7G). Together, these experiments provide strong evidence that stimulation of the non-canonical Wnt pathway can promote Rock-dependent inhibitory phosphorylation of Mypt1, both in mammalian cells and in zebrafish embryos.

Since our loss-of-function data demonstrated that Mypt1 acts to limit the extent of blebbing, and because we found that Rho causes phosphorylation of an inhibitory site in Mypt1, we hypothesized that hyperactivation of the RhoA pathway would phenocopy the ectopic blebbing seen in mypt1 morphants. To test this, we injected 10 pg of constitutively active rhoA mRNA(ca-rhoA) and assayed protrusive activity early in gastrulation. ca-rhoA-expressing cells displayed a high level of protrusive activity and reduced movement (Fig. 6C and see Movie 6 in the supplementary material). These cells extended a large number of membrane blebs and the tissue organization was severely perturbed. Mesodermal cells overexpressing Wnt5b also displayed rounded cell morphology and increased cell blebbing(Fig. 7D,H). These experiments strongly support the hypothesis that Rho-dependent inhibition of Mypt1 is essential for regulating protrusive activity during gastrulation.

Mypt1 is a highly conserved regulator of acto-myosin contractility(Matsumura and Hartshorne,2008). In this work, we demonstrate that Myosin phosphatase is a tightly controlled regulator of essential cellular behaviors during vertebrate gastrulation. Two key functions of Mypt1 are required for gastrulation. Mypt1 cell-autonomously controls cell shape and protrusive activity. Cells lacking Mypt1 fail to convert from amoeboid to mesenchymal behavior and do not undergo intercalative movements required for extension. These cells are, however, able to migrate dorsally without delay, indicating that the amoeboid-to-mesenchymal transition is not required cell-autonomously for convergence. If the entire tissue is lacking Mypt1, however, the cells remain amoeboid and fail to undergo net dorsal migration, resulting in severe convergence defects. Finally, we demonstrate that Mypt1 is tightly regulated by inhibitory phosphorylation downstream of the non-canonical Wnt pathway.

During gastrulation, paraxial mesodermal cells undergo a characteristic series of changes beginning with cells that produce blebs, lamellipodia and filopodia and move in an amoeboid manner, followed by more dorsal oriented,mesenchymal type movements as the cells approach the dorsal midline(Concha and Adams, 1998; Sepich et al., 2005; Weiser et al., 2007). At the midline, cells shut down protrusive activity, pack tightly together and intercalate to extend the axis (Fig. 8A). In our previous work we found that paraxial mesodermal cells lacking Gravin showed normal protrusive activities at the start of gastrulation and converged dorsally, but the cells specifically failed to shut down blebbing when they reached the dorsal side and were unable to undergo normal intercalative behavior, resulting in extension but not convergence defects (Weiser et al., 2007). mypt1 morphant cells transplanted into WT hosts act similarly to gravin morphant cells, converging normally but failing to intercalate and extend. Unlike gravin morphants, however, the mypt1morphant cells have hyper-blebbing activity throughout gastrulation,demonstrating that this type of protrusive activity is fully consistent with dorsal migration of the paraxial mesodermal cells, but, as in the case of gravin morphants, it is inconsistent with packing, intercalation and extension (Fig. 8A). Interestingly, metastatic tumor cells are able to move at very similar speeds when using either lamellipodial or blebbing-type protrusive activities(Sanz-Moreno et al., 2008; Wolf and Friedl, 2006; Wolf et al., 2003), and zebrafish germ cells move rapidly using only blebbing activity(Blaser et al., 2006). Thus,hyper-blebbing activity is consistent with rapid movement, but as we suggest below, there are other reasons why mesodermal cells do not only adopt this type of protrusive activity.

In contrast to the cell-autonomous role of Mypt1 in extension, Mypt1 has a surprising, non-cell-autonomous essential role in convergence. When all the cells in the embryo are depleted of Mypt1 (or are overexpressing either of two Myosin phosphatase inhibitors or a constitutively active Myosin light chain)the embryos have severe convergence defects. During convergence, the paraxial mesoderm cells migrate up an adhesive gradient that is under the control of Bmp signaling, with the lowest adhesion ventrally and the highest adhesion dorsally (Hammerschmidt and Wedlich,2008; Myers et al.,2002; Solnica-Krezel,2006; von der Hardt et al.,2007). Since the transplanted mypt1 morphant cells are able to migrate dorsally in WT hosts, they are able to follow the adhesive gradient. However, when all the cells lack Mypt1, the mesodermal cells are much less dorsally directed in their migration, which explains the overall reduction in convergence (Fig. 8A).

Interestingly, acto-myosin contractility is required for the maintenance of tension in gastrula-stage cells in zebrafish and in Xenopus, and in zebrafish for the separation of germ layers and for mechano-transduction in other tissues (Clark et al.,2007; Krieg et al.,2008; Skoglund et al.,2008). Our results suggest that in zebrafish, in which mesodermal cells migrate dorsally as individual cells, they make productive adhesive interactions (von der Hardt et al.,2007) and pull against each other to gain traction. Thus, we suggest that if cells have too much acto-myosin contractility, owing to a reduction in Mypt1 function, they are unable to productively move towards the dorsal side, explaining the observed non-cell-autonomous defects in mypt1 morphants. Similarly, too little acto-myosin contractility,caused by overexpression of a constitutively active mypt1, also results in severe CE defects. We propose that the mesodermal cells very carefully regulate their level of acto-myosin contractility in order to establish optimal conditions for convergence(Fig. 8B). An essential role for Mypt1, therefore, is to modulate the level of Myosin phosphorylation in order to establish this balance.

One of the conundrums in the study of CE is that for many factors, both inhibition and overexpression cause the same general phenotype. The same is true in our studies, in which the mypt1 MO and overexpression of the constitutively active mypt1 produce embryos with severe CE defects. However, examination of the mesodermal cell behaviors reveals very different effects. Whereas the morphant cells have hyper-blebbing activity throughout gastrulation, the cells overexpressing the constitutively active mypt1 have much reduced protrusive activity. A similar effect was seen with the hyaluronan synthetase Has2, which acts via Rac1 activation(Bakkers et al., 2004). Both gain- and loss-of-function of Has2 caused CE defects. However, gain of Has2 function produced supernumerary lamellipodia, whereas loss of Has2 caused an absence of lamellipodia. Taken together, these results suggest that several features of CE are finely balanced such that tipping either way causes different cell behaviors, with the net effect being a defect in convergence and extension.

A large body of work has established the importance of the non-canonical Wnt signaling pathway as an essential regulator of CE movements in vertebrate gastrulation (Veeman et al.,2003). Among the essential targets is Rho, which activates Rock(Habas et al., 2001). One major target of Rock is Myosin light chain, which induces acto-myosin contractility (Clark et al.,2007). Here, we provide the first in vivo evidence that non-canonical Wnt signaling, acting through Rock, also regulates Mypt1 phosphorylation. We show in both cultured cells and zebrafish embryos that two non-canonical Wnts and activated Rho enhance T696 phosphorylation, whereas the Rock inhibitor suppresses phosphorylation at this site. Importantly,conversion of T696 to a phosphomimetic aspartic acid inactivates Mypt1,whereas conversion to an alanine produces a constitutively active Mypt1 that causes severe CE defects when expressed in zebrafish embryos, as well as rescuing the morphant phenotype at a 5-fold lower dose than is effective for WT Mypt1. Therefore, we propose that one of the crucial functions of the non-canonical Wnt pathway during gastrulation is not only to promote phosphorylation of Mlc through Rho/Rock, but also to partially inhibit Mypt1,thus creating the balance in blebbing behavior and contractility needed to maintain migration and intercalation (Fig. 8B). This precise molecular control would allow the Wnts to fine-tune acto-myosin contractility and precisely control the conversion from amoeboid to mesenchymal to intercalation of mesodermal cells as gastrulation continues.

The most striking phenotype of cells lacking Mypt1 is their excessive bleb-like protrusive activity. Blebs differ from other cellular protrusions in that they are driven by acto-myosin contraction and hydrostatic pressure on the membrane, and not by actin polymerization(Charras et al., 2005). Blebbing is an important form of locomotion(Charras and Paluch, 2008)observed not only in zebrafish gastrula-stage mesodermal cells(Weiser et al., 2007),zebrafish germ cells (Blaser et al.,2006) and delaminating zebrafish neural crest cells as they undergo an epithelial-to-mesenchymal transition(Berndt et al., 2008), but also in Fundulus blastula- and gastrula-stage deep cells(Trinkaus, 1973). Perhaps the most studied form of amoeboid motion is in tumor cell invasion(Friedl and Wolf, 2003). Highly metastatic cells can utilize an amoeboid form of movement(Friedl, 2004; Wolf and Friedl, 2006)characterized by hyperactivation of the Rho-Rock pathway leading to increased acto-myosin contractility and blebbing(Sanz-Moreno et al., 2008; Wolf et al., 2003).

In our current and previous studies of cell movement in zebrafish, we observed that mesodermal cells undergo a conversion from amoeboid to mesenchymal movements during gastrulation and that these cellular behaviors contribute differently to gastrulation. Loss of the metastasis inhibitor Gravin in zebrafish embryos results in a failure to shut down blebbing, which led us to propose that Gravin has an important role as a tumor suppressor that inhibits amoeboid movement (Weiser et al.,2007). As we show here, loss of Mypt1 results in mesodermal cells that not only fail to shut down blebbing but that also exhibit a hyper-blebbing phenotype even at the start of gastrulation. This raises the intriguing possibility that Mypt1 could also function as a metastasis inhibitor. Several recent lines of research have indicated that Rock is a key regulator of invasive cell behaviors in vivo(Sanz-Moreno et al., 2008; Wilkinson et al., 2005; Wolf and Friedl, 2006; Wyckoff et al., 2006). Our work indicates that a key target of Rock in promoting amoeboid behavior is Mypt1. This raises the interesting possibility that loss of Mypt1 function, or constitutive inhibition of Mypt1 by Cpi-17 or Phi-1, could provide an alternative route toward amoeboid behavior in tumor cells, resulting in enhanced metastatic behavior. Thus, it would be interesting to carefully examine Mypt1 regulation and function in controlling tumor cell invasion in vivo.

In conclusion, we have identified Mypt1 as a crucial regulator of several cell behaviors required for gastrulation. Mypt1 regulates cell shape and protrusive activity and is required for intercalation cell-autonomously. Mypt1 also has a non-autonomous role in dorsal convergence, most likely by maintaining the cohesive tissue tension required for mesodermal cells to pull against each other. Finally, we demonstrated that Mypt1 is under the strict control of the non-canonical Wnt pathway and RhoA signaling. Thus, Mypt1 is a crucial regulator of morphogenesis during gastrulation, and is a likely candidate to be involved in mediating invasive cell behavior in metastasis.