Loss of Cofilin 1 Disturbs Actin Dynamics, Adhesion between Enveloping and Deep Cell Layers and Cell Movements during Gastrulation in Zebrafish

During gastrulation, cohesive migration drives associated cell layers to the completion of epiboly in zebrafish. The association of different layers relies on E-cadherin based cellular junctions, whose stability can be affected by actin turnover. Here, we examined the effect of malfunctioning actin turnover on the epibolic movement by knocking down an actin depolymerizing factor, cofilin 1, using antisense morpholino oligos (MO). Knockdown of cfl1 interfered with epibolic movement of deep cell layer (DEL) but not in the enveloping layer (EVL) and the defect could be specifically rescued by overexpression of cfl1. It appeared that the uncoordinated movements of DEL and EVL were regulated by the differential expression of cfl1 in the DEL, but not EVL as shown by in situ hybridization. The dissociation of DEL and EVL was further evident by the loss of adhesion between layers by using transmission electronic and confocal microscopy analyses. cfl1 morphants also exhibited abnormal convergent extension, cellular migration and actin filaments, but not involution of hypoblast. The cfl1 MO-induced cell migration defect was found to be cell-autonomous in cell transplantation assays. These results suggest that proper actin turnover mediated by Cfl1 is essential for adhesion between DEL and EVL and cell movements during gastrulation in zebrafish.


Introduction
In the late gastrula stage, different types of cell migrations, including epiboly, involution, convergence, and extension, work together to form the embryonic germ layers and proper body structures [1,2]. At this stage, a zebrafish embryo has three distinct cellular layers: the enveloping layer (EVL), deep cell layer (DEL) and yolk cell, which can be subdivided into the yolk cytoplasmic layer (YCL) and yolk syncytial layer (YSL) adjacent to the EVL. The EVL, a single-celled layer covering the mass of the DEL, has both tight junctions and filopodial activities with the YSL at its vegetal end as it undergoes epiboly [3,4]. In concert with the EVL, the DEL also migrates toward the vegetal pole; and after the shield stage, part of the DEL, hypoblast, involutes at the margin into the space between the non-involuting DEL, epiblast, and yolk cell for a reverse anterior migration. The DEL then initiates convergence and extension movements before midgastrulation [5]. Both epiblast and hypoblast on the lateral side converge toward the dorsal side accompanied by their epibolic and involuting migration, respectively. Eventually, the dorsal-most portion of the DEL reaches the animal and vegetal poles.
The progression of the EVL margin towards the vegetal pole is at least in part dependent on the YSL by coupling to the EVL via tight junctions. Constriction of actin and myosin may be required to pull down the EVL [4]. Unlike the EVL, the molecular regulation of vegetal movement of DEL is less understood. A close association has been observed between DEL and EVL in an Ecadherin-dependent manner [6]. Thus the vegetal movement of DEL is presumably towing by EVL. This is supported by the arrest or delay of epiboly in DEL, but not EVL in zebrafish embryos lacking E-cadherin [6,7,8,9]. These findings reveal that Ecadherin is required for the adhesion between DEL and EVL and this adhesion is essential for a proper DEL epibolic movement.
E-cadherin is the major transmembrane protein component of epithelial apical junctional complex (AJC), composed of tight junction (TJ) and adherens junction (AJ). By interacting with bcatenin and a-catenin, E-cadherin is coupled to actin filaments to consolidate the AJC [10]. The turnover of actin filament has been shown to affect the disassembly of AJC. We thus hypothesized that the actin filament turnover may also be involved in the Ecadherin-dependent association between DEL and EVL during gastrulation.
Highly active actin turnover occurs at the leading edge of motile cells. In a steady state, actin monomers polymerize at the barbed ends of actin filaments, and depolymerize from the pointed ends. Filament severing and nucleation both increase the numbers of barbed and pointed ends, thus promoting actin turnover and further protrusive activities [11]. Actin-depolymerizing factors (ADFs)/cofilins, one kind of actin-binding protein, can sever, stabilize, or nucleate actin filaments in a dosage-dependent manner [12]. They also were reported to be the primary factor enhancing actin turnover [13,14,15], and amplifying and specifying the direction of motile cells during chemotaxis [16]. These small (15,21 kDa) ADFs/cofilins have conserved ADFhomology domains which are present in the closely related twifilins [17] and in some parts of several familiar proteins (e.g., drebrins) [18]. When a cofilin protein binds to an ATP-binding cleft on an actin filament helix, it changes the twist of the helix [19], and then bends the filament, thus accelerating actin filament severing [20]. But the ability to sever actin decreases as the concentrations of ADFs/cofilins increase [21,22,23]. That is, when the concentrations of ADFs/cofilins increase, actin filaments tend to be fully decorated by ADFs/cofilins resulting in a rather stabilized condition. And if the local ratio of ADFs/cofilins to actins increases, the actin tends toward nucleation.
ADFs/cofilins activities are directly activated by Slingshot phosphatase (SSH) [24] and inhibited by LIM kinase (LIMK) by dephosphorylation and phosphorylation on the N terminal serine 3 residue, respectively [25,26]. Both SSH and LIMK are known downstream effectors of small GTPases [26,27]. We have previously demonstrated the necessity of Rho signaling in mediating gastulation in zebrafish [28,29]. In addition, the SSH was also been reported to play an essential role in Xenopus gastrulation [27]. Thus we reasoned that ADFs/cofilins may also be critical regulators in gastrulation by modulating actin turnover. However, in a large insertion mutagenesis screen, a zebrafish mutant harboring an insertion in the first intron of cofilin 1 (cfl1) (previously known as cofilin 2 like) was identified and showed no obvious phenotype at one day after fertilization [30]. The lack of early gastrulation defect in zebrafish cfl1 mutants is contradictory to previous studies described. It might be that the residual maternal cfl1 messages presented in embryos were enough for the embryos to survive through early development. To clarify this issue, we have cloned the zebrafish cfl1 and showed that cfl1 is differentially expressed in DEL, but not in EVL in zebrafish embryos. We further demonstrated that the loss of cfl1 induced by its respective morpholino results in the interference of epiboly, convergent extension, cell protrusions and directed cell migration during gastrulation.

Results
Sequence, syntenic, and structural analyses of zebrafish cofilin 1 In zebrafish, there are three homologous genes of ADF/cofilins: cofilin 1 (cfl1, non-muscle form, AY398324), cofilin 1-like (cfl1l, AY398323) and cofilin 2 (cfl2, muscle form, NM_205700). We were interested in cofilins, which are expressed during early embryogenesis and have the potential to regulate gastrulation cell migration. According to the available expression patterns in ZFIN, only cfl1 and cfl1l fit into this category, as cfl2 is mainly expressed in somites and later in myotomes and lenses. By a preliminary screening using an antisense morpholino oligo (MO), we found that knockdown of cfl1l had a minimum effect on gastrulation (our unpublished data). By contrast, the cfl1 MOs showed severe interference with the progression of gastrulation (data will be described later). Thus, we had isolated the zebrafish cfl1 gene by PCR cloning from zebrafish embryonic complementary DNAs. To compare its sequence identities and similarities of zebrafish Cfl1 to cofilins of other animal species, we aligned its amino acid sequence of with that of the non-muscle form of cofilins in mammals (including bovine, human, mouse, and rat), chick, western clawed frog, fish (including fugu and medaka), and the ADFs in the fruit fly (TSR) and nematode (UNC-60) (Fig. 1A). Zebrafish Cfl1 was highly similar to those cofilins examined (82%,90%) except for the fruit fly (50%) and nematode (56%) (Fig. 1B). It was most similar to Cfl1 of medaka with 90% similarity. As shown in the phylogenetic tree (Fig. 1C), zebrafish Cfl1 was grouped together with the chordates, while the invertebrate ADFs (fly and nematode) were placed in an outgroup cluster.
To determine the chromosomal position of cfl1, we conducted a syntenic analysis of three cofilin genes (Fig. 1D). The cfl1 in zebrafish, human, and mouse are located on chromosomes 14, 11, and 9, respectively. The same genes are found in the vicinity, which are efemp2 (EGF-containing fibulin-like extracellular matrix protein 2) and fibp (fibroblast growth factor (acidic) intracellular binding protein), but zebrafish cfl1 is arranged in a reversed orientation compared with those of human and mouse. These analyses suggest that zebrafish cfl1 is evolutionarily related to mammalian cofilin 1 genes.
The primary amino acid sequence of Cfl1 contains 165 amino acids with a molecular weight of ,18 kDa. The characteristic secondary structures are depicted in Fig. 1E. Cfl1 has six a-helices and seven b-sheets with a phosphorylation site at Ser 3 and a nuclear localization sequence (NLS, at residues 21,38). A threedimensional ribbon structure of Cfl1 created by SWISS-MODEL (http://www.swissmodel.expasy.org) is shown in Fig. 1F. Taken together, the zebrafish cfl1 we isolated has high sequence and functional domain similarities compared to its orthologs from other species.

Loss of Cfl1 interferes with epiboly movement
To further investigate the role of Cfl1 during embryogenesis in zebrafish, we separately injected embryos with two non-overlapping antisense MOs, tMO 1 and tMO 2 targeting the start codon and neighboring 59-untranslated region (UTR) sequence of cfl1 mRNA at 222 to +3 and 250 to 226, respectively, to block translation of cfl1. At 10 hours post fertilization (hpf), untreated embryos had completed epiboly and had entirely enclosed the yolk sphere ( Fig. 2A). By contrast, knockdown of cfl1 by tMO 1 delayed or even arrested epiboly progression ( Fig. 2B-C) in a dosedependent manner and similar inhibition of epiboly was observed using the non-overlapping tMO 2 (Table 1). Embryos injected with 10 ng tMO 1 , showed severe retardation in epiboly progression, and some of them could not complete gastrulation and ceased at about the 50% epiboly stage at 10 hpf (Fig. 2D). Embryos injected with lower tMO 1 dosages had milder epiboly defects, most of them achieved 80%,90% epiboly (Fig. 2C), and some of them only exhibited a malformed tail-bud (Fig. 2B). To further confirm that the severely retarded embryos are developmentally arrested, we monitored those 10 ng tMO 1 -injected embryos, which did not reach 100% epiboly at 10 hpf, until 12 hpf when the un-injected control embryos developed to 6-somite stage, and found that 235 out of 247 of those embryos (95.665.4%) still arrested at 60,80% epiboly in three trials.
To examine whether the zygotic Cfl1 expression can be inhibited by cfl1 MOs, we inserted a stretch of cfl1 nucleotides ,200-bp, which contains their respective MO target site, into a pCS2+ XLT vector with a green fluorescent protein (GFP) gene and co-injected these constructs with their respective MO. Embryos (81.0613.8%) injected with the pCS2+ XLT with tMO 1 binding site (tMO 1 plasmid) expressed GFP fluorescence (Fig. 3A). By contrast, none of embryos co-injected with tMO 1 showed GFP fluorescence (Fig. 3B). Similarly, embryos (79.4611.9%) injected with the pCS2+ XLT with tMO 2 binding site (tMO 2 plasmid) expressed GFP fluorescence, but not in those embryos co-injected with tMO 2 (Fig. 3C). Both MOs appeared to be equally effectively, we thus used tMO 1 for the rest of experiments unless otherwise stated.
To confirm that the defects we observed in cfl1 morphants were specifically due to the loss of Cfl1 activity, we examined whether overexpression of cfl1 could rescue the cfl1 morphant defects. cfl1 was successfully overexpressed as revealed by the presence of GFP fluorescence in zebrafish embryos injected with mRNAs prepared from a PCS2+ XLT vector containing the full cfl1 coding sequence (data not shown). More importantly, the inhibition of epiboly could be significantly (P,0.05) reduced in embryos co-injected with 100 pg cfl1-egfp mRNAs and 7.5 ng tMO 1 (Table 2). Furthermore, depending on the sequence specificity of MOs, MOs may have caused secondary effects like p53-dependent apoptosis [31]. To examine whether the effect of the cfl1 MO was mediated by p53, we co-injected the cfl1 tMO 1 (7.5 ng) with a p53 MO (7.5 ng) into zebrafish embryos and found that both epiboly progression and mortality of co-injected embryos did not significantly (P.0.05) differ from those embryos injected with only the cfl1 tMO 1 ( Table 2).

cfl1-tMO 1 causes unsynchronized migration of EVL and DEL and over-polymerization of filamentous actin
To examine the effects of cfl1 knockdown on the actin cytoskeleton structure, the filamentous actin (F-actin) was stained by rhodamine phalloidin, and to better differentiate individual cells nuclei were visualized by a nuclear dye DAPI (49,6diamidino-2-phenylindole). The double-stained embryos were examined under confocal microscopy ( Fig. 4). At the lower magnification, the margins of EVL and DEL did not appear to migrate together during epiboly in cfl1 morphants (Fig. 4D) compared to the synchronized movements in control embryos (Fig. 4A). At a higher magnification, it clearly showed that the EVL was 2 or 3 rows of cells ahead of the DEL (Fig. 4E, F). This uncoordinated migration between the EVL and DEL was even more obvious in embryos injected with a higher dosage (10 ng) of cfl1-tMO 1 (data not shown). At the higher magnification, the YSL actin cable (arrow) and cell boundary actin were notably intense and condensed in cfl1 morphants ( Fig. 4E) compared to control embryos (Fig. 4B).

Temporal and spatial expression profile analyses of cfl1
The unsynchronized movements of EVL and DEL in cfl1 morphants led us to suspect that cfl may be expressed differentially and result in different responses upon knockdown of cfl1. The RT-PCR analysis showed that cfl1 was expressed throughout development and was present in all adult tissues examined (data not shown). To gain the temporal and spatial profiles of cfl1 during embryogenesis, we thus performed whole-mount in situ hybridization (WISH) in different stages of embryos (Fig. 5). The WISH analysis revealed that the expression domains of cfl1 during early development were ubiquitous until the sphere stage ( Fig. 5A-D). The cfl1's expression was notably reduced at 30% epiboly ( Fig. 5E) and the shield stages (Fig. 5F), and it was absent from the future ventral side of the embryo in the bud stage (Fig. 5G). cfl1 was strongly expressed throughout the embryo body during early segmentation period (Fig. 5H); it was later restricted to the central nervous system, lateral line, and pronephric duct at 26,31 hpf ( Fig. 5I and J); and expression domains were evident in the brain, pharyngeal arches, lateral line, and pronephric duct in the larval stages (3,5 days post-fertilization, Fig. 5K-N). We were in particular interest in the expression of cfl1 at different cell layers of a gastrulating embryo, thus we performed cryo-section for those 70% epiboly stage embryos underwent WISH against cfl1. Interestingly, we found that the cfl1 expression domain was restricted to blastomere boundaries of the DEL, but not those of the EVL (Fig. 5O).  Loss of Cfl1 retards vegetal migration in the DEL but not the EVL It is known that a close contact association exists between EVL and DEL [6]. We therefore examined whether the unsynchronized movements of EVL and DEL is due to the loss of their association by transmission electron microscopy (TEM). Clear disruptions of EVL-DEL attachments were observed in cfl1 morphants as evidenced by spaces in part of the contact zones between the DEL and EVL (Fig. 6B) compared to that in control embryos (Fig. 6A).
To further compare dynamic interactions between the DEL and EVL, we performed live imaging analysis of a gastrulating embryo from 7.5 to 8.5 hpf under confocal microscopy. We focused on the dorsolateral sites of embryos injected with membrane-bound GFP-GAP43 mRNAs and the StdMO or cfl1 tMO 1 . While both the DEL and EVL showed highly coordinated migration toward posterior sites in control embryos (Fig. 7, left column; Supplementary Movie S1), cells of the DEL in cfl1-morphants did not form tight attachments and showed a clear gap with EVL cells (Fig. 7, right column; Supplementary Movie S2). In addition, a very dynamic change in cell shape was observed in DEL cells of control embryos, but not in those of cfl1 morphants (Supplementary Movie S2).

Loss of Cfl1 perturbs convergent extension but not involution during gastrulation
To further investigate how Cfl1 affects cell movements during gastrulation, we conducted cell tracing assays using caged Q-  rhodamine dextran [32]. Caged Q-rhodamine was injected with or without the cfl1 tMO 1 into one-cell stage embryos and later uncaged by brief exposure to ultraviolet light to mark a selected group of cells. To monitor dorsal convergence movements, we marked cells in the lateral blastoderm margin at 90u from the dorsal embryonic shield, and fluorescent cells were traced during gastrulation. The marked cells in the control StdMO-injected embryos moved dorso-anteriorly, and had extended along the anterior-posterior (AP) axis at 10.5 hpf (Fig. 8A, upper panels), as previously reported [33,34]. By contrast, the anterior migration and convergence of lateral mesendodermal cells toward the dorsal side were notably impaired in cfl1 morphants (Fig. 8A, lower panels). Measuring the degree of convergence (h = sin 21 (r/R)) from the origin of the marked cells in control and treated embryos at different times after recording and plotting the results against the time (hpf) showed that the fitted slope of cfl1 tMO 1 -injected embryos (y = 10.7x -69.7) clearly deviated from that of StdMOinjected ones (y = 15.6x -96.2) (Fig. 8C). To examine the effect of cfl1 tMO 1 on AP extension movements, cells of dorsal embryonic shields were marked and monitored as previously described in StdMO-and cfl1 tMO 1 -injected embryos (Fig. 8B). In control StdMO-injected embryos, marked cells were found in the dorsal axial hypoblast along the entire AP axis at 10.5 hpf (Fig. 8B, upper panels) which is consistent with a previous report [34]. By contrast, the anterior movement of marked cells was perturbed, exhibiting a shortened axial mesoendoderm in cfl1 tMO 1 -injected embryos (Fig. 8B, lower panels). Measuring the angle (Q) between two arrows of the anterior and posterior ends of the marked axial mesoendoderm connecting to the center of the embryos (Fig. 8B) in control and treated embryos at different times after recording and plotting the results against times of measurement demonstrated that the fitted slope (33.2x -215.8) of cfl1 tMO 1 -injected  embryos was also lower than that of StdMO-injected ones (y = 41.7x -261.9) (Fig. 8D). Furthermore, we also conducted WISH to reveal the anterior edge of the neural plate and prechordal plate by probing against cathepsin L 1 b (cstl1b) and distalless homeobox gene 3b (dlx3b), respectively. The body axis was severely shortened and broadened in tMO 1 -treated embryos (Fig. 8E) compared to that of control embryos (Fig. 8F).
To examine whether the internalization of epiblast (i.e. involution) was affected by the reduction of cfl1 translation, we have treated embryos with 5.0 or 7.5 ng of tMO 1 or stdMO, fix and subjected them to WISH against cstl1b. The appearance of cstl1b is an indication of formation of prechordal plate and thus the occurrence of involution. In two independent trials, involution occurred in all embryos treated with both dosages of tMO 1 tested   Fig. 10A-D, and the average number of protrusions formed in each treatment is presented in Fig. 10E.

Cfl1 mediates pseudopod formation of involuting cells in a cell-autonomous manner
To further evaluate the cell autonomy of Cfl1 functions, cells from embryos injected with MOs and rhodamine-dextran, were transplanted into untreated host (control) embryos and cfl1 morphants respectively. Only those transplanted cells undergoing epiboly were selected for further analysis. The protrusive activities of these transplanted cells were examined by confocal microscopy, and representative snapshots were taken to show the polygonal shape and pseudopod formation by StdMO-injected cells in control hosts (arrow heads in Fig. 11A; Supplementary Movie S7) and in cfl1 morphants ( Fig. 11C; Supplementary Movie S8), while cfl1 tMO 1 -injected cells remained rounded with blebbing-like structures in control hosts (arrow in Fig. 11B; Supplementary Movie S9) and in cfl1 morphants ( Fig. 11D; Supplementary Movie S10). Furthermore, to reveal the migration velocity and direction of movement, we took time-lapse recordings under an epifluorescence microscope to trace cells from both StdMO-and cfl1 MOinjected embryos transplanted into untreated or cfl1 morphants. After analysis, we found that both the exact velocity (Vcl: curvilinear velocity; curvilinear distance/time) and directed velocity (Vsl: straight line velocity; straight line distance/time) were significantly reduced in cfl1 tMO 1 -injected cells compared to StdMO-injected cells when transplanted into an untreated host embryo (Fig. 11E). This indicates that cofilin functions cell autonomously in transplanted cell migration. In the same logic, both Vcl and Vsl of cfl1 tMO 1 -injected cells significantly reduced when transplanted into cfl1 morphants. However, the Vcl and Vsl of StdMO-injected cells is also significantly reduced when transplanted into cfl1 morphants, although was not as severe as cfl1 tMO 1 -injected cells in the same background (Vsl, Fig. 11E).

Discussion
We have previously demonstrated that Rho mediates gastrulation cell movements in zebrafish via Rho-associated kinase [29] and Diaphanous [28], by regulating the contractile activity of actin filaments. Herein, we further delineated an essential role of a Rho downstream effector, cofilin, in zebrafish gastrulation by demonstrating the inhibitions of epiboly, convergence, and extension, but not involution by knocking down cfl1. The effects of cfl1 knockdown were shown to be cell-autonomous and presumably via affecting actin turnover as evident by the over-polymerization of F-actin seen in cfl1 morphants. The cfl1 MOs also inhibited coordinated movement of EVL and DEL due to the disruption of their attachments and cellular contacts between two layers.
The uncoordinated movements between the EVL and DEL observed in cfl1 morphants (Fig. 4) has been reported in Ecadherin-deficient embryos [6,7,8,9]. Tight adhesion between the DEL and EVL is known to be exerted by E-cadherin mediated AJC that presumably allows the towing of DEL toward vegetal pole during gastrulation [6]. To examine whether the uncoordinated movement between the EVL and DEL was due to the disrupted EVL-DEL association, we showed the physical disrup-  The cryo-section was taken from the box region as depicted in the embryo carton shown on the right. b, brain; ba, bronchial arches; ll, lateral line system; pd, pronephric duct; pa, pharyngeal arches; EVL, enveloping layer; DEL, deep cell layer. Scale bars: 400 mm for the 4-and 5-dpf embryos, 200 mm for the others and 50 mm for the cryo-section photograph. doi:10.1371/journal.pone.0015331.g005  . Live cell imaging reveals loose cell-cell interactions between the DEL and EVL in cfl1 morphants. Embryos injected with designated MOs and membrane-bound green fluorescent protein (GFP) mRNA, were dechorionated, immobilized, and examined by confocal microscopy for 10-min recordings at 10-s intervals per frame. The snapshots of representative embryos from the StdMO-(left column) or cfl1MOinjected embryos (right column) are shown. The recording times in minutes are denoted in the lower left corner of the left column. While StdMOtreated embryos showed continuous tight attachments between the DEL and EVL, DEL cells in cfl1 morphants did not form a tight connection with the EVL, which was further evidenced by the more-rounded cell shapes. Arrowheads indicate spaces between the EVL and DEL. A single-cell-thick EVL is to the right; yolk cells are toward the left; the animal pore is on the top; and the vegetal pore is at the bottom. doi:10.1371/journal.pone.0015331.g007 tion of EVL/DEL attachment in cfl1 morphants by TEM analysis (Fig.6) and time-lapse cellular recording ( Fig. 7 and supplementary Movie S2). Because actin turnover is critical for maintaining AJC [35], thus Cofilin 1 may be required for the E-cadherin-mediated adhesion of EVL/DEL. Furthermore, cryo-sections of cfl1-WISHlabeled embryos showed restricted cfl1 mRNA signals in the DEL but not the EVL at the 70% epiboly stage (Fig. 5). This differential expression of cfl1 suggests that Cfl1 might only function in the DEL. Indeed, we clearly demonstrated that loss of Cfl1 only hampered movement of the DEL but not the EVL (Fig. 4,7). Although the EVL was shown to have filopodial activities toward the YSL [3], the lack of cfl1 expression suggests that its filopodial activities may be regulated by factors other than Cfl1. In addition, tight junctions which exist between the EVL and YSL might passively drive epibolic migration of the EVL epiboly [4]. With intact DEL-EVL association, the DEL can be towed by the EVL during epiboly. However, this interaction between DEL and EVL was apparently lacking in cfl1 morphants, which might have resulted in the uncoordinated migration.
The clear separation of margins of the DEL and EVL in cfl1 morphants (Fig. 4,6B,7) suggests that cell-cell interactions between the DEL and EVL were affected. This pathological phenomenon is similar to that seen in E-cadherin (cadherin 1, cdh1)-deficient embryos [6,7,8,36]. DEL-EVL but not DEL-DEL or EVL-EVL cell attachments were disrupted in both cfl1 and cdh1 knockdown embryos. E-cadherin interacts with actin cytoskeleton by a complex with b-catenin which binds to the actin-binding protein a-catenin [37,38]. These serial bindings may stabilize the membrane-bound cdh1 and form tight cell adhesions in these adherent junctions. It might be possible that knockdown of cfl1 reduces actin turnover, disrupts the proper structures of actin filament networks in the DEL near the contact regions with the  EVL, and results in unstable membrane-bound cdh1 and abnormal detachments from the EVL. Unlike the EVL, the force driving the DEL to undergo epiboly is still unclear. However, active cell protrusions between the DEL and EVL at the margins of the epiboly were observed in wild-type zebrafish embryos [39]. These active cellular protrusions may assist in maintaining the tight interaction between the DEL and EVL. Thus, EVL could lead the DEL towards the vegetal pole to complete the epiboly. It is clear that the differential expression of cfl1 in the DEL but not in EVL (Fig. 5O) allowed us to disrupt this interaction by simply knocking down cfl1 activity in DEL and further demonstrated the importance of proper interaction of both layers in epiboly progression.
ADF/Cofilin has recently been suggested to play a functional node in cell biology [40]. Recently, it has been found that instead of its conventional role in actin depolymerization, cofilin can either promoting actin filament assembly or disassembly depending on their cellular concentrations and their ration to actin and other actin binding proteins [41]. It can also compete for actin at actin branches with ARP2/3 [42], which is known to be critical actin branches during embryogenesis [43,44,45,46,47]. Cofilin also contains functions unrelated to actin. It has been suggested to chaperon actin, which is lacking nuclear location signal, into nucleus [40], Upon cell stress, it can be translocated to mitochondrial [48] to induce the release of cytochrome c and subsequent apoptosis [49,50]. The phospho-cofilin can directly activate phospholipase D (PLD) [51], which can mediates chemotaxis [52] and membrane lipid metabolism. It certainly deserves further investigations to determine whether cofilin exerts its effects on gastulation cell movements via any of its above-described novel functions in addition to the control of actin dynamics.
The lack of early phenotype in cfl1 insertion mutants in zebrafish [30] is puzzling and contradictory to the necessity of cofilin as demonstrated in Xenopus embryos [27]. In this study, we demonstrated that Cfl1 is also essential for proper gastrulation cell movements in a cell-autonomous manner by mediating actin dynamics. Its differential expression in the EVL and DEL, and different migration patterns of the EVL and DEL in cfl1 morphants further suggest that differential actin regulation is necessary to maintain adhesion of embryonic layers for cohesive epibolic movement in zebrafish. The lack of gastrulation defect observed in cfl1 mutants is presumably due to the presence of maternal cfl1 that helps the mutant embryos pass the gastrulation process.

Ethics Statement
All animal handling procedures were approved by the use of laboratory animal committee at National Taiwan University, Taipei, Taiwan (IACUC Approval ID: 97 Animal Use document No. 55).  Morpholino oligonucleotides (MOs) and mRNA preparation Antisense MOs were purchased from Gene Tools (Philomath, OR) or Open Biosystems (Huntsville, AL). The standard MO (StdMO; sequence: 59-CCTCTTACCTCAGTTACAATTT-ATA-39) without sequence homology to any known zebrafish DNA sequences was used as a control treatment. To knock down zebrafish cfl1 expression, two translational-blocking MOs (tMO), tMO 1 (targeting the start codon and neighboring 59-untranslated region (UTR) sequence of cfl1 mRNA, 222 to +3; sequence: 59-CATGGCTGTGTCTCTGTGCTAGTCG-39) and tMO 2 (targeting 59-UTR sequences of cfl1 mRNA, 250 to 226; sequence: 59-TCTGGGTTGGAATTGAATGAGCTGT-39) were used. To counter p53 activity, a p53 MO (sequence: 59-GCGCCA-TTGCTTTGCAAGAATTG-39) was used. To examine the translation blocking efficiency of cfl1 tMO 1 and tMO 2 , two pCS2+ XLT constructs were generated with an insertion of cfl1 gene fragmen from 243 to +189 and 252 to +51, respectively. The cfl1 tMO 1 and tMO 2 were co-injected with respective pCS2+ XLT plasmid and the % of embryos expressing GFP fluorescence was determined. To prepare cfl1 expression vector, the full coding region of cfl1 was inserted into pCS2+ XLT vector, which contains the green fluorescent protein (GFP) sequence following the insertion site. The cfl1-gfp construct was linearized by NotI, and the capped RNA was transcribed according to the manufacturer's instruction using a mMESSAGE mMACHINEH SP6 Kit (Applied Biosystems, Foster City, CA 94404).

Microinjection procedures
Glass capillaries (1.1460.50 mm, O.D. 6 I.D., World Precision Instrument, Sarasota, FL) were pulled using a horizontal puller (P-97, Sutter Instrument, Navato, CA). Embryos were collected after spawning, and then immobilized by an injection trough on a 100mm 1% agar plate. An injection pipette was forced through the chorion, into the yolk cell and reached the area adjacent to the blastomeres where 2.3-4.6 nL of the desired solution was ejected using a Nanoliter injector (World Precision Instrument). Before injection, all solutions were diluted by 1x Danieau's buffer with phenol red (0.25% (w/v), pH 7.4) to the desired concentrations. Microinjection was completed before the embryos had reached the 4-cell stage. After injection, embryos were recovered from the trough and cultured at 28.5uC for further experiments.

Whole-mount in situ hybridization (WISH) and cryosection
Embryos injected with MOs were grown in 0.3x Danieau's buffer supplemented with 0.2 mM 1-phenyl-2-thiourea (Sigma) to inhibit pigment formation [53], fixed at the desired developmental stages with 4% fresh paraformaldehyde in phosphate-buffer saline (PBS) overnight. After being manually dechorionated with fine forceps, embryos then transferred and stored into 100% methanol at 220uC. WISH was performed according to Thisse et al. [54,55] using a DIG-labeled, SP6-RNA polymerase-made riboprobes hybridized to cfl1 or cfl1l. For cryosections, stained samples were immersed in PBS containing 30% sucrose overnight, and embedded in Optimal Cutting Temperature (OCT) compound embedding medium (Sakura, Tokyo, Japan) at 220uC, and 10-mm frozen cross-sections were cut with a CM 1900 rapid sectioning cryostat (Leica, Heidelberg, Germany) and attached to poly-L-lysine-coated slides (Electron Microscopy Sciences, Ft. Washington, PA). Stained embryos and sections were observed under a stereomicroscope (Mz75, Leica Microsystems) and a DM5000B DIC system, respectively. Photographs were taken using a digital camera (Coolpix 995, Nikon, Melville, NY).

Transmission electron microscopy (TEM)
Embryos injected with MOs were dechorionated at 8 hpf, fixed with 2% fresh paraformaldehyde and 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH = 7.4) at 8.5 hpf for 2 h at room temperature, washed by the same buffer, and then post-fixed with 1% OsO 4 in the same buffer for 2 h on ice. They were then rinsed with distilled water, stained with 0.5% aqueous uranyl acetate overnight at room temperature, dehydrated with ethanol series then 100% acetone, and embedded in Spurr's low-viscosity embedding mixture (Electron Microscopy Sciences). Ultrathin sections were cut with diamond knives, double-stained with uranyl acetate and lead citrate, and examined with a Hitachi H-7650 transmission electron microscope (Toronto, Ontario, Canada).

Ventral leading edge observations
Embryos injected with designated MOs and membrane-bound GFP-GAP43 messenger RNA (mRNA) were dechorionated and mounted in 1% low-melting-point (LMP) agarose (AMRESCO, Solon, OH) at 6.5 hpf. Embryos then were monitored in regions of the ventral leading edges under confocal microscope (TCS SP5, Leica Microsystems). Time-lapse images of suitable embryos were taken at 10-s intervals for 10 min.

Cell tracing assay
Embryos were injected with 1% (w/v) caged Q-rhodamine dextran (Invitrogen, Carlsbad, CA) and 7.5 ng cfl1 tMO 1 dissolved in injection buffer mixed with 1% Q-rhodamine were injected into embryos at one cell stage. The embryos were grown in dark until 50% epiboly and immobilized at an injection through on a 50 mm 1% agar plate. To uncage Q-rhodamine, a beam of ultraviolet light through a DAPI filter set under an automatic upright microscope system (DM5000B, Leica Microsystems, Wetzlar, Germany) with a 40X water objective, was directed for 1,3 s at the lateral blastoderm margin [33]. Both dark and bright field images were taken using a CoolSNAP fx charge-coupled device (CCD) camera (Roper Scientific, Tucson, AZ). Images were collected and analyzed using Simple PCI Imagine System software (Compix, Sewickley, PA).

Cell protrusion assay
Embryos injected with MOs were dechorionated and mounted in 0.8% LMP agarose at the sphere stage for lateral cell observations, or at the shield stage for prechordal plate cell observations. Differential interference contrast (DIC) optical timelapse images were acquired and analyzed by the Simple PCI Imagine System software for 15 min at 15-s intervals under the DM5000B system with a 40X water immersion objective.

Cell transplantation assay
Donor embryos were injected with dextran, tetramethylrhodamine (10,000 MW, anionic, fixable, Invitrogen) and designated MOs. After the 1k-cell stage, host embryos were dechorionated by 10 mg/mL protease (Sigma, St. Louis, MO) and mounted in 0.8% LMP agarose. Cells from donor embryos were then transplanted into host embryos at approximately the sphere stage using CellTram vario (Eppendorf, Hamburg, Germany). The single focal plane time-lapse recordings of transplanted cells were collected at 30-s intervals for 1 h during late gastrulation (at 7,9 hpf) using the 40x water-immersion objective under the DM5000B system and analyzed by the SimplePCI Imagine System software. The transplanted cells were also observed under confocal microscopy (TCS SP5, Leica Microsystems) and 10-min movies were recorded at 15-s intervals.

Supporting Information
Movie S1 Time-lapse confocal imaging of dorsolateral cell migration in StdMO-injected embryo. Embryos were also injected with membrane-bound green fluorescent protein mRNA to reveal the membrane. Shown here is a 10-min movie at 15-s intervals time-lapse confocal image sequence. The cells were migrating toward the vegetal pole.