Coordinated cell-shape changes control epithelial movement in zebrafish and Drosophila

Contraction of an epithelial sheet to close an opening is a common process in animal development and tissue repair, and requires extensive modification of cell shape at the advancing epithelial margin(Martin and Parkhurst, 2004). Many of the cellular and molecular mechanisms controlling this event have been identified by analysis of Drosophila dorsal closure (reviewed by Harden, 2002; Jacinto et al., 2002b). Dorsal closure is one of the last morphogenetic movements during Drosophilaembryogenesis and occurs as the opposing fronts of the lateral epidermis converge dorsally, displacing the underlying amnioserosa(Martinez-Arias, 1993). A central aspect of this process is the uniform elongation and constriction of marginal epidermal cells. This transforms the margin from a loose, scalloped edge into a taut row of cells and probably ensures even epithelial advance(Jacinto et al., 2002a; Young et al., 1993). The observed accumulation of actin and myosin 2 in these cells suggests that the transformation of the margin is achieved by a contractile actomyosin `cable'. However, to what extent this represents a general mechanism controlling epithelial movement is unclear.

Similar to the Drosophila epidermis, the outer epithelium of the zebrafish embryo, the enveloping layer (EVL), is a simple epithelium that spreads over the spherical yolk cell together with the underlying deep cells,to ultimately seal the embryo on the vegetal side(Kane and Adams, 2002; Kimmel et al., 1995) (for illustration, see Fig. 1L). During EVL epiboly, the length of the advancing margin must be reduced as the tissue closes around the vegetal yolk blastopore. Analysis of the related teleost Fundulus heteroclitus has shown that this involves the constriction of marginal EVL cells (Keller and Trinkaus, 1987), similar to Drosophila dorsal closure. However, the mechanistic basis for EVL cell shape change remains to be shown.

Insights come from studies of both Fundulus and zebrafish, and suggest a potential role of the underlying yolk syncytial layer (YSL) during EVL epiboly. The YSL constitutes the surface of the yolk cell directly beneath the blastoderm and undergoes epiboly simultaneously with the EVL. In Fundulus, the leading edges of marginal EVL cells were found to be stably attached to the YSL while undergoing constriction. Furthermore, a dense network of microfilaments was detected in the YSL along the EVL margin in Fundulus (Betchaku and Trinkaus,1978; Keller and Trinkaus,1987). A similar accumulation of actin within the YSL has recently also been reported in zebrafish, and was implicated in EVL epiboly(Cheng et al., 2004). This suggests that the YSL may control changes in marginal EVL cell shape via an actin-dependent mechanism. However, YSL-specific analysis supporting a role for actin during EVL epiboly has so far been missing. Furthermore, the molecular mechanisms controlling actin function in the YSL are still unknown.

In this study, we show that localized recruitment of actin and myosin 2 within the YSL correlates with EVL cell-shape change. Furthermore, we show that this process is dependent on Msn1, a zebrafish ortholog of the Drosophila Ste20-like kinase Misshapen. Similarly, we show that Drosophila Misshapen is required for actin/myosin 2-based cell constriction at the epidermal margin during dorsal closure. These findings point to a conserved mechanism of actin/myosin 2 recruitment and cell constriction during epithelial morphogenesis in zebrafish and Drosophila.

All embryos were obtained from zebrafish AB, TL and Wik wild-type lines,grown at 31°C, and manipulated in E3 zebrafish embryo medium or Danieau buffer.

Wild-type embryos were from the Oregon R strain. UAS lines were expressed using the Gal4 system (Brand and Perrimon,1993). The Gal4 lines e22c-Gal4, LE-Gal4, en-Gal4(Brand and Perrimon, 1993), and c381-Gal4 (AS-Gal4) and flies carrying the GFP balancer TM3,gal4-twi, UAS-2xEGFP (TTG) were kindly provided by the Bloomington stock centre. msn¹⁷² was kindly provided by J. Treisman. Homozygous msn¹⁷² embryos expressing GFP-Actin in the engrailed domains were generated by crossing UAS-GFP-Actin/Cyo; msn¹⁷²/TTGflies to en-Gal4; msn¹⁷²/TTG flies. A strain carrying a UAS-DN-msn construct was a gift from Yong Rao(Houalla et al., 2005). Homozygous UAS-DN-msn flies were crossed with e22c-Gal4,LE-Gal4 or AS-Gal4 flies to express DN-msn in the whole epidermis, marginal cells of the epidermis, or the amnioserosa,respectively.

misshapen-type genes were identified in the zebrafish genome using BLAST analysis. cDNA sequences were obtained from PubMed. Multiple alignments of predicted protein sequences were done using ClustalX(Chenna et al., 2003). Phylogenetic analysis was done using protdist and fitch from the Phylip package [Felsenstein, J. 1989. PHYLIP (Version 3.2). Cladistics 5: 164-166]. The phylogenetic tree was displayed using NJplot(Perriere and Gouy, 1996).

In situ staining of zebrafish embryos were performed as previously described (Barth and Wilson,1995). In situ probes were synthesized from cDNA using a DIG RNA labeling kit (Roche). The msn1 in situ probe corresponded to bases 1230-2341 of the msn1 cDNA sequence (PubMed Accession number AAH55134.1).

The following morpholino oligonucleotides (MOs)(Heasman, 2002) were used: msn1MO-splice (5′-ACACACAACTTACccttaaagtgga-3′); msn1MO-splice5bp (5′-ACACAGAAGTTACCGTTAAACTGCA-3′), msn1MO-ORF (5′-CTCGCCATATTAAAGACGAAAAAAC-3′, msn1MO-UTR (5′-GAATAGACCTCTCAGTCAGTCGTCC-3′), msn2-ORF (5′-GCGGGAGAGTCGTTCGCCATCTTTC-3′), msn3-ORF (5′-TGCGTTTTCTGACATTTTCACAACG-3′), e-cadMO (5′-ATCCCACAGTTGTTACACAAGCCAT-3′). Embryos were injected at the one-cell stage as previously described(Barth and Wilson, 1995). YSL injection was performed at the sphere stage. For rescue of msn1MO phenotypes,200 pg of full-length msn1 mRNA (PubMed Accession number AAH55134.1)and 2 ng of msn1MO-splice were injected into the YSL. Effective YSL-injection was monitored by co-injecting Alexa 488 Histone H1(Molecular Probes), which labels YSL nuclei(D'Amico and Cooper, 2001).

Total RNA was purified from zebrafish embryos using Trizol reagent(Invitrogen). mRNA was then purified using PolyAtract reagents (Promega). cDNAs were generated using the SMART RACE cDNA Amplification Kit (Clontech). The msn1 splicing pattern was analyzed by PCR amplification using the primers P1 (5′-TGTACCAGAAGCAAGAGCGAGC-3′) and P2(5′-CATCCATGCACTTGATAGCAGCC-3′).

Embryos were collected on yeasted apple juice agar plates for 24 h and aged for 48 h at 25°C. They were then dechorionated in bleach, rinsed in water,mounted in a 1:1 mixture of acetic acid and Hoyer's medium, and incubated for 12 h at 65°C.

Zebrafish embryos were fixed overnight in 2% or 4% paraformadehyde at 4°C, then washed in 0.1% Triton in PBS (PBT) and dechorionated. They were then incubated for 1 h in 0.5% Triton in PBS, followed by a 5-h incubation in block solution (10% normal goat serum, 1% DMSO, 0.1% Triton in PBS). Embryos were then incubated overnight at 4°C in block solution containing Phalloidin and/or primary antibodies. They were then washed in PBT, and incubated for 5 h at room temperature with secondary antibodies. Antibody and Phalloidin staining of Drosophila embryos were performed as previously described (Kaltschmidt et al.,2002). Alexa 488 Phalloidin and Alexa 594 Phalloidin (Molecular Probes) were used at 1:200 dilutions. The following primary antibodies and dilutions were used: rabbit anti-E-cadherin at 1:750(Babb et al., 2001), rabbit anti-α-catenin and mouse anti-β-catenin at 1:500 (Sigma), mouse anti-ZO-1 (a gift from S. Tsukita) at 1:25, rabbit anti-Claudin-1 at 1:400(Zymed), rabbit anti-phospho-myosin light chain 2 Ser19 (Cell Signalling) at 1:100, rabbit anti-Msn at 1:50 (Houalla et al., 2005), rabbit anti-myosin 2 at 1:500(Kiehart and Feghali, 1986),and mouse anti-Armadillo at 1:50 (Developmental Studies Hybridoma Bank, ref. N2 7A1). Zebrafish embryos were mounted on agarose-coated dishes in PBT medium, Drosophila embryos were mounted in Vectashield (Vectorlabs). Images were acquired on a Zeiss LSM and a Leica TCS-SP2 confocal microscope.

Time-lapse, multiple focal plane (4D) microscopy of zebrafish embryos expressing GAP-43-GFP and cytosolic GFP was performed as previously described(Montero et al., 2005). Live imaging of Drosophila embryos was performed as previously described(Jacinto et al., 2000).

Dechorionated zebrafish embryos were incubated in 50 μm(±)-blebbistatin or (+)-blebbistatin (Calbiochem) after the 50% epiboly stage, and were compared to control embryos in Danieau buffer; 4 ng of e-cadMO was injected into one-cell-stage embryos, which were then incubated in blebbistatin after 50% epiboly.

Embryos were fixed in 4% paraformaldehyde, 2% glutaraldehyde in 0.1 M phosphate buffer overnight, then postfixed in 1% osmium tetroxide (Science Services) for 1 h, dehydrated through a graded series of ethanol, and infiltrated in Embed-812 resin (Science Services) overnight. Samples were cured for 48 h at 65°C. Ultrathin sections (70 nm thick) were cut on an Ultracut microtome (Leica Microsystems). Samples were viewed in a Morgagni EM(FEI).

Images of Phalloidin-stained zebrafish embryos were analyzed using ImageJ software(http://rsb.info.nih.gov/ij/). The length of marginal EVL cells was defined as the distance between the middle of the leading edge of the cell and the animal-most region of the cell. Cell width was defined as the distance between the centers of the lateral cell boundaries of a cell. Length/width ratio of a cell was correlated with the local Phalloidin density in the adjacent YSL. Phalloidin density was defined as the average pixel intensity within a rectangle immediately adjacent to the leading edge of the cell. Data presented were based on one experiment. Repeats of the experiment yielded similar results. Phalloidin signal intensity profiles were generated by plotting pixel intensity along a line from the EVL margin towards the vegetal pole. Several profiles were generated for each embryo. Average profiles of multiple embryos are shown.

Drosophila Misshapen: NP_524679.2

Zebrafish Msn1: AAH55134.1

Zebrafish Msn2: fj33a09 (EST)

Zebrafish Msn3: AL921244

Human TRAF2 and NCK interacting kinase (Traf2_NCK): Q9UKE5

Human MAP4 kinase 4: O95819

Human MAP4 kinase 6: NP_733763.1

Mouse Traf2_NCK: XP_130797.3

Mouse MAP4 kinase 4: P97820

Mouse MAP4 kinase 6: NP_795712.1

C. elegans Mig-15: NP_509747.1

Previous analysis indicated a potential role of the actin cytoskeleton within the YSL for EVL epiboly (Cheng et al., 2004). We therefore addressed how actin distribution in the YSL relates to the morphology and arrangement of EVL cells during epiboly (for illustration of tissue organization during epiboly, see Fig. 1L). To visualize actin distribution in addition to outlining the EVL cells, we stained embryos with fluorescently labeled Phalloidin at different epiboly stages. Throughout this study, epiboly stages reflect the position of the margin of the deep cell layer.

At 30% epiboly, the EVL margin had advanced towards the vegetal pole to the same extent as the underlying deep cells. Marginal EVL cells were oval shaped and loosely aligned. A diffuse band of actin was present at the YSL cortex in the vicinity of the EVL margin (Fig. 1A,D). At the onset of mesendoderm ingression, actin within the YSL became sharply concentrated along the EVL margin. Simultaneously, the marginal EVL cells were aligned along the equator, and typically displayed a rectangular shape (Fig. 1B,E). At 75% epiboly, the EVL margin was positioned clearly ahead of the deep cells. A dense ring of actin had formed within the YSL along the EVL margin, and the majority of EVL cells were clearly elongated in the movement direction(Fig. 1C,F).

We noticed that at 75% epiboly, both actin concentration in the YSL and marginal EVL cell shape had become increasingly variable. Interestingly, we found a direct correlation between the length/width ratio of individual marginal EVL cells and the amount of actin in the adjacent YSL(Fig. 1M, see Materials and methods). This suggests that local recruitment of actin in the YSL is linked to the elongation and narrowing of EVL cells that occurs as the epithelium advances towards the vegetal pole.

Given the evidence from Drosophila that actin/myosin 2-based contraction occurs at the leading edge of the epidermis during dorsal closure,we looked for myosin 2 in the zebrafish YSL by staining embryos with an antibody against phosphorylated (activated) myosin light chain 2. Weak staining was detected at the boundaries between EVL cells, and at late stages of epiboly, the signal became enriched within the YSL along the EVL margin,similar to the Phalloidin signal (Fig. 1G,H). Thus, both actin and myosin 2 reside within the YSL in the direct vicinity of marginal EVL cells as these cells undergo dynamic changes in shape.

In Fundulus, EVL cells at the margin appear to be connected to the YSL via tight junctions (Betchaku and Trinkaus, 1978). To characterize this linkage in zebrafish, we first analyzed zebrafish embryos at 75% epiboly using transmission electron microscopy. We detected structures at the EVL-to-YSL interface that resemble tight junctions but not adherens junctions(Fig. 1K). Consistent with this, antibodies against zebrafish ZO-1(Fig. 1I) and human Claudin-1(data not shown), both known components of vertebrate tight junctions(D'Atri and Citi, 2002),predominantly labeled the leading edges of marginal EVL cells, and the immediately adjacent YSL. In addition, both markers labeled the boundaries between EVL cells. In contrast, antibodies against E-cadherin, α-catenin and β-catenin, components of adherens junctions(Nagafuchi, 2001), strongly labeled the borders between EVL cells, but only very weakly labeled the EVL-to-YSL interface (Fig. 1J,and data not shown), indicating that the linkage between these tissues does not primarily involve adherens junctions.

Together, these results suggest that actin and phospho-myosin light chain 2 recruitment within the YSL to the EVL margin and the physical linkage between both tissues via tight junctions are involved in controlling cell constrictions at the EVL margin.

The Drosophila MAP4 kinase Misshapen and its orthologs in other organisms have been implicated in the control of actin-based cell movement,including Drosophila dorsal closure(Ruan et al., 2002; Ruan et al., 1999; Su et al., 1998; Su et al., 2000; Xue et al., 2001). We therefore determined the specific requirement for Misshapen-type molecules in actin-based cell constriction both during EVL epiboly in zebrafish and Drosophila dorsal closure.

To address a potential role of Misshapen-like proteins during zebrafish epiboly, we first identified orthologues of Drosophila misshapen in the zebrafish genome. We identified three genes (msn1-3) encoding MAP4 kinase proteins with high similarity to Drosophila Misshapen. Phylogenetic analysis suggests that Msn1-3 each clearly group with one of the three Misshapen-related proteins in mammals(Fig. 2A).

To assess potential functions of these genes during zebrafish epiboly, we performed gene `knock-down' experiments by injecting specific morpholino-oligonucleotides (MOs) targeting the coding region of each gene into one-cell stage embryos. msn1 MOs caused clear defects during gastrulation (described below). Weaker gastrulation defects were observed in the case of an msn2 MO, while no defects were observed in the case of an msn3 MO (data not shown). We co-injected MOs against msn1and msn2, but no obvious enhancement of the msn1 morphant phenotype was observed (data not shown). This suggests that both genes function non-redundantly. For the remainder of this study, we focus specifically on msn1 function.

Comparison of the predicted protein sequences of DrosophilaMisshapen and Msn1 revealed 77% identity of the N-terminal kinase domain, and 73% identity of the C-terminal CNH (citron homology) domain(Fig. 2B).

To characterize the RNA expression pattern of msn1 during embryogenesis, we performed whole-mount in situ hybridization using a probe against the non-conserved central domain of the message. msn1 was detected at early cell division stages(Fig. 2C), and is therefore maternally provided. At shield and bud stages, we detected weak, ubiquitous expression throughout the embryo (Fig. 2D,E).

To address the role of msn1 during gastrulation, we injected 1-8 ng of three independent msn1-specific MOs at the one-cell stage, and observed similar embryonic defects in each case. The splicing morpholino msn1MO-splice was used for the phenotypic analysis presented in this study. A quantification of the effects of the other MOs is included in our description of YSL-specific MO injections below.

msn1MO-splice targets the exon 1/intron 1 boundary of the pre-mRNA. Sequence analysis of the msn1 mRNA of embryos injected with this MO revealed changes in the splicing pattern, that eliminate the exon sequence encoding amino acids 1-19 of the predicted protein (see Fig. S1A in the supplementary material). Usage of a potential alternative start codon(encoding amino acid 56) would truncate the N-terminal Ser/Thr kinase domain(amino acids 25-289). This suggests that msn1MO-splice probably interferes significantly with msn1 function, but may not lead to complete protein loss.

Morphant embryos appeared morphologically normal and displayed no obvious defects in patterning at shield stage (see Fig. S1B-E in the supplementary material). However, during subsequent stages, we observed epiboly defects in 65% of the embryos. In addition, defects in the convergence and extension of the main body axis were frequently observed (see Fig. S1F-L in the supplementary material). This suggests that msn1 is required for the process of epiboly as well as for convergence and extension movements.

The above results suggest a function of msn1 in multiple gastrulation movements. However, since these movements are probably interdependent, the specific role of msn1 during EVL/YSL epiboly remained unclear. We therefore directly addressed the role of msn1 in the YSL by injecting morpholinos into this tissue prior to the onset of epiboly. Embryos injected with 8 ng msn1MO-splice displayed no obvious abnormalities until 50% epiboly (approximately 75 min after injection, Fig. 3A,H), after which half of the embryos showed a delay in epiboly of the deep cell layer and the EVL,while convergence and extension occurred at the normal rate(Fig. 3B,I,O). The deep cell epiboly delay probably occurred as a secondary consequence of impaired EVL epiboly, as the EVL margin limits how far deep cells can progress towards the vegetal pole.

The observed phenotypes are most likely to be due to a reduction of msn1 function in the YSL, since co-injection of wild-type msn1 mRNA into the YSL suppressed the epiboly defect to a large extent, and the injection of a 5-base-mispair MO (msn1MO-splice5bp)caused a significantly less-penetrant phenotype(Fig. 3O). Furthermore, YSL injection of two additional MOs caused similar phenotypes: 1 ng of msn1MO-ORF (targeting the open reading frame) resulted in a clear delay of epiboly in 79% of the embryos (n=70), while injection of 8 ng of msn1MO-UTR (targeting the 5′UTR) caused epiboly delay in 24% of the embryos (n=74).

To specifically identify the role of msn1 in the YSL for EVL epiboly, we analyzed morphant embryos with respect to three main aspects: (1)EVL cell shape, (2) localization of actin and myosin 2 in the YSL, and (3) the linkage between EVL and YSL. We analyzed changes in EVL cell shape between 50%and 75% epiboly both in live embryos, using multi-photon confocal microscopy,and in Phalloidin-stained embryos. At 50% epiboly, marginal cells of control embryos were typically oval shaped and displayed protrusive activity at their leading edges. Shortly after, protrusive activity ceased, and during subsequent stages the average length/width ratio of marginal cells increased,reflecting the elongation and narrowing of these cells. In addition, cell shapes at the margin became increasingly variable, as some cells underwent leading edge constriction while adjacent cells became wider. Frequently, the collapse of the leading edge of cells resulted in the elimination of these cell from the margin (Fig. 3C-G,P; see Movie 1 in the supplementary material). Morphant embryos showed normal cell morphology at 50% epiboly, but failed to undergo the later changes observed in the controls. The rate of margin progression towards the vegetal pole was slowed, the average the length/width ratio remained essentially constant, and little variability in cell shape was observed at 75% epiboly (Fig. 3J-N,P; see Movie 2 in the supplementary material).

To address whether these cell-shape defects correlated with impaired recruitment of actin and myosin 2 in the YSL, we co-stained embryos with Phalloidin and the phospho-myosin 2 antibody. At 50% epiboly, diffuse Phalloidin signal was observed in the YSL of control and morphant embryos(Fig. 4A,B). In control embryos, the Phalloidin and anti-phospho myosin 2 signals became markedly enriched within the YSL along the EVL margin as the embryos reached 75%epiboly (Fig. 4C,E,G). In morphant embryos, this enrichment of both markers was clearly weaker(Fig. 4D,F,H).

To address whether reducing msn1 function in the YSL interferes with the formation of tight junctions between EVL and YSL, we stained morphant embryos with the anti-ZO-1 antibody. No clear abnormalities in the staining pattern were observed (data not shown), arguing against a defect in tight junction formation in these embryos.

The above results indicate that msn1 is required in the YSL for the recruitment of actin and myosin 2 and for cell-shape changes of marginal EVL cells.

The above results show that actin and myosin 2 recruitment correlates with the progression of EVL/YSL epiboly. To directly address the role of actin/myosin 2 contraction in the YSL, we tested myosin 2 function during EVL/YSL epiboly using the specific myosin 2 inhibitor blebbistatin(Straight et al., 2003).

We found that embryos exposed to (±)-blebbistatin at the onset of gastrulation epibolized at a slower rate than control embryos, typically completing the process with significant delay, or arresting prior to its completion. In contrast, other aspects of gastrulation were only very mildly affected, resulting in the formation of a largely normal body axis. Embryos treated with an inactive form of the drug, (+)-blebbistatin, showed no significant developmental defects (Fig. 5A-C). This suggests that epiboly is directly impaired by blebbistatin.

Next, we analyzed EVL cell shape and actin distribution in blebbistatin-treated embryos during late stages of epiboly. We found that the EVL margin of treated embryos had progressed significantly less towards the vegetal pole compared to untreated embryos. In addition, we noted unusually wide marginal EVL cells, and found that the level of actin accumulation in the YSL along the margin was markedly reduced(Fig. 5D,E). These findings suggest that blebbistatin inhibition of myosin 2 partially impairs actin accumulation in the YSL and slows EVL progression towards the vegetal pole.

To address whether the effect of blebbistatin on EVL epiboly may be caused indirectly through an effect of the drug on deep cell epiboly, we tested whether EVL epiboly is generally dependent on the deep cells. Consistent with previous findings, we observed that the EVL advanced at a normal rate when we severley impaired deep cell epiboly through morpholino-based inactivation of half baked/E-cadherin in the embryo(Fig. 5F,G)(Kane et al., 1996; Kane et al., 2005). This suggests that EVL epiboly is independent of the deep cells and is directly inhibited by blebbistatin. In addition, we performed blebbistatin treament of half baked/E-cadherin morphants, which enabled us to better visualize and quantify the overall effect of the drug on the EVL. Morphant embryos with or without drug treatment showed a similar delay of deep cell epiboly 9 h after fertilization. However, in the case of no drug treatment, the EVL margin had almost reached the vegetal pole and was positioned 186±9 μm(mean±s.e.m., n=10) ahead of the deep cell margin(Fig. 5F,G), while in drug-treated embryos, the EVL margin had progressed only 71±6 μm(mean±s.e.m., n=11) ahead of the deep cells.

Given the apparent role of cell constriction at the margin both during EVL epiboly in zebrafish and dorsal closure in Drosophila, we investigated whether Misshapen may play a similar role during Drosophila dorsal closure. Consistent with such a role, we found that Misshapen is localized to the cortex of cells of the epidermis and cells of the underlying amnioserosa throughout dorsal closure(Fig. 6A,B).

To characterize the role of Misshapen during dorsal closure, we first determined the range of embryonic defects in embryos homozygous for msn¹⁷², a likely null allele of misshapen(Treisman et al., 1997). Based on the preparation of cuticles, 2% of mutant embryos displayed an anterior dorsal hole; 38% had closed but showed misalignment of segments; while 12%showed embryonic lethality without obvious defects during dorsal closure(n=48, Fig. 6C,D and data not shown). To determine whether these phenotypes may result from defects of cell constriction at the margin, we investigated the organization of the actin cytoskeleton in msn and control embryos. In control embryos,actin was detected as a continuous `cable' in marginal cells of the epidermis from stage 13 onward. Marginal cells were elongated in the movement direction and were aligned into an apparently taut row(Fig. 6E). In msn¹⁷² embryos, we detected intermittent breaks in the actin distribution at the margin, typically affecting groups of 2-4 cells. On average, these breaks in the actin cable were found at a frequency of two per epithelial margin at stage 13-14 (n=10). Affected cells were wider and less elongated than neighboring cells with normal actin localization(Fig. 6F).

To determine how these defects affect the closure process, we performed live imaging of embryos expressing GFP-Actin in the engrailed domain(see Materials and methods). In control embryos, we consistently observed perfect segment alignment along the seam where the opposing epithelial fronts have met (Fig. 6G,I; see Movie 3 in the supplementary material). In msn¹⁷² embryos, we observed a similar range of dorsal closure phenotypes as was detected by our cuticle preparations. In particular, we found that 30% of the embryos displayed segment misalignments during the final phase of dorsal closure(Fig. 6H,J; see Movie 4 in the supplementary material; and data not shown).

The segment misalignment defect of msn¹⁷² embryos was typically accompanied by unusual irregularities in the width of segments during the closure process. Some segments appeared abnormally wide, while neighboring segments became excessively constricted. This probably resulted in the failure of proper matching of opposing segments during the zippering phase of the closure process. The formation of filopodia and lamellipodia at the margin, which had been previously implicated in the control of segment matching (Jacinto et al.,2000), appeared normal in these embryos.

We therefore conclude that Drosophila Misshapen is required during dorsal closure for actin-based elongation and constriction of cells at the advancing epidermal margin.

To determine whether Drosophila Misshapen function is specifically required in the epidermis of the embryo, we expressed a dominant-negative version of the gene [DN-msn, previously described by Houalla et al.(Houalla et al., 2005)] in the whole epidermis or in marginal cells alone. In both cases, analysis of the actin cytoskeleton revealed defects similar to those of msn¹⁷² mutants. We consistently observed breaks in the localization of actin at the margin of the epidermis(Fig. 7A,D,G). These occurred at a similar frequency to that in msn mutant embryos.

Furthermore, we observed abnormalities in myosin 2 localization. While control embryos showed continuous myosin 2 localization along the margin,frequent disruptions of the pattern were observed in embryos expressing DN-msn (Fig. 7B,E,H). These gaps in actin or myosin 2 localization correlated with reduced elongation and leading edge constriction of the affected cells(Fig. 7C,F,I). This suggests that Misshapen is required in marginal cells of the epidermis to mediate actin/myosin 2-based cel-shape changes during dorsal closure.

During dorsal closure, the amnioserosa (AS) cells change shape and constrict apically. This process is known to be driven by an apical contractile apparatus that is regulated by Drac1 and Crumbs(Harden et al., 2002). To investigate if Misshapen also plays a role in the AS, we interfered with Msn function by ectopically expressing DN-msn specifically in the AS cells. We found that less than 10% of the embryos (n=120) displayed cuticle phenotypes, namely with head problems that vary from an anterior hole to dorsal-anterior holes (data not shown). This phenotype suggests that Misshapen does not play a major role in amnioserosa contraction.

In this study, we provide new insight into the control of coordinated cell-shape changes at the advancing margin of an epithelial sheet and suggest that conserved mechanisms mediate this process both in zebrafish and Drosophila (see Fig. 8for an overview). In zebrafish, we found that the accumulation of actin and myosin 2 in the YSL is dependent on the Ste20-like kinase Msn1 and is required for cell-shape changes of marginal EVL cells during epiboly. Similarly, we found that Drosophila Misshapen, the ortholog of Msn1, is required for actin/myosin 2-based constriction of marginal cells of the epidermis during dorsal closure.

Previous studies have suggested that the YSL plays an important role during EVL epiboly in teleosts. Analysis of epiboly movements of Fundulus heteroclitus has shown that marginal EVL cells are physically linked to the YSL as they undergo changes in shape(Betchaku and Trinkaus, 1978; Keller and Trinkaus, 1987). Furthermore, both in Fundulus and zebrafish, the accumulation of actin within the YSL at the EVL/YSL interface has been described(Betchaku and Trinkaus, 1978; Cheng et al., 2004). Our study builds on these findings and specifically focuses on the molecular mechanism of actin recruitment in the YSL and the role of this process for cell-shape changes of the EVL.

We show that after the 50% epiboly stage, the advancement of the EVL margin towards the vegetal pole becomes in part dependent on the YSL. As in Fundulus, we find that the YSL in zebrafish is probably coupled to the EVL via tight junctions. Furthermore, our findings suggest that actin/myosin 2-based contraction of the YSL mediates cell-shape changes at the EVL margin. First, we demonstrated that cell elongation and leading edge constriction at the margin correlates with the accumulation of actin within the adjacent YSL. Secondly, we showed that YSL-specific inactivation of Msn1 impairs both actin/myosin 2 accumulation in the YSL and cell-shape changes of the EVL. Our data suggest that EVL epiboly is driven by YSL contraction, and implicate substrate contraction as an important mechanism to coordinate cell-shape changes of an advancing epithelial sheet.

Based on our results, we propose a specific model of how actin/myosin 2-based YSL contraction promotes EVL epiboly. Actin and myosin 2 within the YSL become recruited to the junctional complex between the EVL and YSL via a Mns1-dependent mechanism. This results in local constriction of the leading edges of marginal EVL cells. As a consequence, the overall length of the EVL margin shortens and progresses towards the vegetal pole of the spherical yolk cell (for illustration, see Fig. 8).

A central question is how the constriction of cells is coordinated at the EVL margin. The presence of a cable-like zone of actin along the closing margin of an epithelial sheet has previously invoked a purse-string mechanism(Kiehart, 1999; Williams-Masson et al., 1997). However, such a model is probably insufficient to account for the dynamic cell-shape changes of marginal EVL cells. Our findings suggest that YSL contraction may be increased at local `hot spots' in the vicinity of the EVL margin, causing some marginal cells to become highly constricted, while adjacent cells widen. This results in high variability of EVL cell shape observed during late stages of epiboly. Whether this is regulated autonomously within the YSL or involves signals from the EVL is currently unknown. Alternatively, contractility may be evenly distributed within the YSL, and cell constriction may occur particularly in those areas of the EVL margin that pose the least resistance. These models should be tested in future studies.

Our analysis of EVL epiboly reveals several similarities to dorsal closure in Drosophila. In both cases, the advancing front of an epithelial sheet becomes physically linked to the underlying substrate via junctional complexes (this study; Narasimha and Brown, 2004). Furthermore, both processes depend on actin/myosin 2-based constriction of marginal cells that is mediated by a Misshapen molecule.

Similar to EVL epiboly in zebrafish, Drosophila dorsal closure is probably not mediated by a simple purse-string mechanism. In embryos with impaired misshapen function, local disruptions of the actin/myosin 2 cable did not prevent the overall closure process. This is consistent with prior studies showing that ablation of marginal epidermal cells does not prevent completion of dorsal closure(Kiehart et al., 2000). We observed that cells lacking actin and myosin were abnormally wide, while the adjacent cells contracted excessively. This led to segment mismatches at the end of dorsal closure. Therefore, it is likely that actin/myosin 2 cable mediates coordinated cell constrictions at the margin to ensure even closure of the embryo.

In addition, our study revealed important differences between epithelial-sheet sealing in fish and flies. In zebrafish, cell-shape changes at the EVL margin are probably mediated by the underlying substrate, while in Drosophila, marginal cells appear to actively constrict(Jacinto et al., 2002a). We speculate that in zebrafish, actin/myosin 2-based tissue contraction along the circumference of the embryo may be most efficiently achieved in the continuous cytoplasm of the YSL. Similar observations have been made in the case of the C. elegans embryo, in which most cells of the outer epithelium fuse prior to the formation of contractile rings of microfilaments(Priess and Hirsh, 1986).

Another difference between EVL epiboly and dorsal closure lies in the cellular mechanism that shortens the epithelial margin during the closure process. In both cases, this involves the elimination of marginal cells. However, in Drosophila, cells leave the margin through the fusion of the opposing fronts of the epidermis on the anterior and posterior ends of the dorsal hole (Wood et al.,2002). In contrast, we find that in zebrafish, similar to Fundulus, cells along the whole circumference of the embryo are continuously removed from the margin and accommodated into rows further behind(Keller and Trinkaus, 1987). This process involves the complete elimination of the leading edges of EVL cells, probably as a result of localized contraction of the YSL.

A mechanism similar to EVL epiboly has been proposed for closure of epidermal wounds in Drosophila. Reduction of an epidermal wound margin involves the exclusion of marginal cells through actin-based leading-edge constriction (Wood et al.,2002). Furthermore, wound healing as well as dorsal closure of the Drosophila embryo require the function of an actin/myosin 2-based cell constriction. Moreover, wound closure in Xenopus oocytes depends on the accumulation of actin filaments and myosin 2 at the wound margin(Mandato and Bement, 2001). This suggests that further analysis of the molecular mechanism of actin/myosin 2-mediated YSL contraction will improve our understanding of the mechanisms of epithelial morphogenesis as well as wound healing.

Molecular insight into the control of actin-based YSL contraction comes from our analysis of the role of the Ste20-like kinase Msn1, which is required for actin and myosin 2 recruitment within the YSL. One candidate molecule likely to mediate Msn1 function is myosin light chain 2, as we observed decreased levels of the phosphorylated (activated) form of the protein within the YSL of msn1 morphant embryos. Whether Msn1 functions through known activators of myosin light chain 2, such as Rho kinase, or whether it affects myosin 2 function more indirectly remains to be established. Other candidates for Msn1 effectors are suggested from studies of Misshapen in Drosophila. During dorsal closure, Misshapen functions upstream of the JNK pathway (Su et al.,1998; Su et al.,2000). Additionally, Misshapen has been described to interact with the actin-binding protein Bifocal during photoreceptor growth cone migration(Ruan et al., 2002; Su et al., 2000). Future studies are needed to determine which downstream targets of Misshapen-like molecules are conserved between Drosophila and zebrafish.

We thank B. Habermann for help with identifying the msn1-3 genes and for performing the phylogenetic analysis. Thanks to Michaela Wilsch-Bräuninger for transmission electron microscopy work and to the laboratory of S. Eaton for help with fly work. We thank the Bloomington Stock Center, Y. Rao and J. Treisman for fly stocks, Y. Rao for the anti-Msn antibody, and D. Kiehart for the anti-myosin 2 antibody. We are grateful to Franziska Friedrich for assistance with the artwork. We thank the staff of the fish and imaging facilities at the MPI-CBG for excellent assistance. We thank C. Dahmann, A. Oates, and L. Rohde for critical reading of the manuscript. This work was supported by grants from the Emmy-Noether-Program of the DFG,the MPG and the Humboldt foundation. M.K. was supported by fellowships from the DFG and the Marie Curie Fellowship Association.