Ciona embryonic tail bending is driven by asymmetrical notochord contractility and coordinated by epithelial proliferation

Tissue bending and folding, including the formation of the vertebrate neural tube, the germ layer (during gastrulation) and the gut, are essential evolutionarily conserved morphogenetic processes in embryogenesis. However, coupled bending-folding processes also occur on a large scale beyond the tissue level, such as embryonic folding. In vertebrates, embryonic folding changes a disc-like embryo into a curved, cylindrical-shaped one (Tuchmann-Duplessis et al., 1971). Embryonic folding includes longitudinal folding, which produces both head and tail folds, and transverse folding, which produces right and left lateral folds. Failure in any of the bending-folding processes leads to severe birth defects (e.g. craniorachischisis) (Copp and Greene, 2013).

Various mechanisms underlie tissue bending and/or folding in both invertebrates and vertebrates. For example, actomyosin-mediated apical constriction [narrowing at the apical surface of the cell (Lv et al., 2019)] is required for gastrulation in Drosophila melanogaster, Caenorhabditis elegans and Ciona intestinalis (Lee and Goldstein, 2003; Martin et al., 2009; Sherrard et al., 2010) and vertebrate neural tube formation (Haigo et al., 2003; Nishimura et al., 2012). Similarly, basal constriction (narrowing at the basal surface of the cell) is required for the formation of the zebrafish optic cup (Nicolás-Pérez et al., 2016) and the midbrain-hindbrain boundary (Gutzman et al., 2008). Differential growth, due to increased cell proliferation or volume in one layer of cells relative to an adjacent layer, is another important mechanism responsible for tissue bending and folding. For example, hinged folding in the avian gut endoderm is driven by localized differential epidermis proliferation (Miller et al., 1999), whereas chick embryonic heart bending is due to differential hypertrophic growth (Shi et al., 2014).

In addition to these mechanisms, cell apoptosis (Monier et al., 2015), cell mitotic rounding (Kondo and Hayashi, 2013) and differential positioning of the adherens junction (Wang et al., 2012) also contribute to tissue bending and/or folding. Increasing evidence shows that most of the tissue-bending and tissue-folding processes are driven by the interaction of multiple mechanisms. For example, both basal actomyosin constriction and active epithelial migration are required for retinal neuroepithelium bending during zebrafish optic cup morphogenesis (Sidhaye and Norden, 2017), and both cell adhesion and differences in tissue growth are required for the folding during early tooth germ formation (Marin-Riera et al., 2018).

Although tissue bending has been extensively studied at the tissue level, the mechanisms that sculpt the tissue-bending and tissue-folding morphology at the embryo scale are unclear because of anatomical complexity. We addressed this question by taking advantage of the relatively transparent, structurally simple chordate Ciona robusta (also called C. intestinalis type A) embryo, which is an ideal system for determining cellular and molecular mechanisms underlying morphogenesis at a whole-embryo scale. The notochord is responsible for both tail elongation and tail morphology because disruption of notochord intercalation or elongation leads to defects in the notochord and hence tail morphogenesis (Jose-Edwards et al., 2013; Oda-Ishii et al., 2010; Segade et al., 2016; Sehring et al., 2014; Shi et al., 2009). Cooperation between the notochord and muscle is required for tail elongation (Di Gregorio et al., 2002) and a wave-like mitotic pattern of the epidermis is crucial for tail morphogenesis during neurulation (Ogura and Sasakura, 2016). Therefore, tail morphogenesis is likely achieved via highly coordinated behavior between the notochord, muscle and the epidermis. In addition to elongation, as in other species, the embryonic tail in Ciona bends ventrally and surrounds the trunk within the chorion. However, the underlying genetic mechanisms have not been studied because embryonic tail bending is thought to be a passive process that is achieved by the physical barrier of the chorion confining embryonic tail bending during elongation.

In this study, we experimentally investigated the mechanisms underlying embryonic tail bending in chordates. To further identify the mechanical role of each tissue, we also developed a physical model with experimentally measured parameters to simulate embryonic tail bending. Together with experiments and the biomechanical model, the findings reveal how the different tissues coordinate with each other to drive the large-scale morphogenesis at the whole-embryo scale, and provide insights into the complex embryonic bending process in chordates.

After fertilization, Ciona embryos developed within a layer of membrane called the chorion (Fig. 1A, white arrowhead). The early embryo tail bent and subsequently surrounded the trunk inside the chorion (Fig. 1A; Movie 1), indicating that the tail elongates by passively bending itself down because of confinement within the chorion. Unexpectedly and excitingly, when the chorion was removed from the embryos using a chemical approach (Corbo et al., 1997), the dechorionated embryos elongated normally and still bent ventrally toward the trunk (Fig. 1A; Movie 2), although the tail-bending angle (θ) was slightly lower compared with wild-type embryos. To quantitatively evaluate the embryonic tail bending in the embryos, we plotted θ against the tail length (ζ) during tail bending (Fig. 1B). The rose diagram plots show that tail morphogenesis patterns were similar in the embryos with or without the chorion; however, θ in dechorionated embryos was lower compared with embryos with the chorion at the same ζ (Fig. 1C). These findings showed that embryonic tail bending in Ciona is a chorion-independent process and controlled intrinsically by genetics.

To determine whether embryonic tail bending is autonomous, we isolated the trunk and tail parts of Ciona tailbud embryos at stage 19 using needle ablation, cultured the separated parts in filter-fresh seawater for 2 h and examined isolated tail fragment development (Fig. 1D, top). Surprisingly, the cutting tail (CT) bent ventrally as the tail elongated (Fig. 1D,E, n=13/14; Movie 3), and the maximum curvature of the bent tail was frequently located at the anterior part of the tail (Fig. 1D, white arrowhead). We further cut the isolated tail into two segments: anterior cutting tail (A-CT) and posterior cutting tail (P-CT). After 2 h post cutting, the A-CT bent ventrally (n=7/10) (Fig. 1D,E), whereas the P-CT did not bend but elongated straight (n=21/24) (Fig. 1D,E). Together, the findings indicate that embryonic tail bending in Ciona is an autonomous event driven by tail tissues and that the driving forces originate in the anterior part of the tail.

The Ciona embryo tail comprises three main tissues: epidermis, muscle and notochord (Fig. 2A). The muscle is located symmetrically on the left and right of the tail, rather than in the dorsoventral (DV) axis (Fig. 2A, middle), so we ignored its mechanical contribution and mainly investigated the roles that the notochord and epidermis play in embryonic tail bending. Geometric analysis of notochord morphology showed that the anteroposterior (AP) length of the notochord's ventral edge was shorter compared with its dorsal edge (Fig. 2B; Table S1) at tailbud stages 17-22, indicating that the notochord presents as a bending cylindrical shape during embryonic tail bending (Fig. 2B, inset). In addition, individual cell geometry data showed ventral narrowing of notochord cells after intercalation initiation (Fig. S1A). To determine the mechanical mechanisms underlying the ventrally bent shape, we examined the actomyosin distribution in the notochord. Longitudinal sections showed that F-actin (labeled by Phalloidin) is asymmetrically enriched at the ventral side of the notochord at tailbud stages 17-21 (Fig. 2C, n≥6; Movie 4). To avoid bias from the observed angles, we confirmed the ventral enrichment of F-actin from cross-sectional (Fig. S1B) and dorsal (Fig. S1C) views. Ventral enrichment of F-actin was more prominent at the early tailbud stage (stage 17), and the enrichment gradually became symmetrical later (tailbud stage 22) (Fig. 2C,E). Quantitative data showed that the F-actin intensity at tailbud stages 17 and 18 on the ventral side of the notochord increased by 2.5-fold compared with the dorsal side (Fig. 2E).

Nonmuscle myosin II is essential for contractility of actomyosin networks. Its motor function is activated by reversible phosphorylation of the myosin regulatory light chain (MRLC) at serine 19 and threonine 18 (Vicente-Manzanares et al., 2009). Considering results of F-actin locations (Fig. 2C), we examined active myosin II using a specific anti-pS19 MRLC antibody and found that it asymmetrically accumulated at the ventral side of the notochord during embryonic tail bending (Fig. 2D, yellow arrows, n≥6). Quantitative data further revealed that the difference between dorsal and ventral active myosin II was 1.3 times, relatively weaker compared with F-actin (Fig. 2F). Together, the findings indicate that the active actomyosin apparatus is asymmetrically located at the ventral side of the notochord, indicating that asymmetrical actomyosin contractile forces cause notochord cell shape changes.

To determine whether embryonic tail bending requires actomyosin contractility, we inhibited actomyosin activity during early (tailbud stages 16 and 17) and later (tailbud stage 21) embryonic tail bending stages using chemical inhibitors Y-27632 and blebbistatin, respectively. We observed embryonic tail-bending defects in all treatment groups, with more severe tail-bending abnormality at early tailbud stages (tailbud stages 16 and 17) (Fig. S2). To determine whether contractility is the driving force behind notochord bending, we performed genetic manipulation experiments to disrupt the actomyosin apparatus, specifically in the notochord. Previous studies in Ciona and Drosophila have reported that introducing an unphosphorylatable form of MRLC (T18A-S19A) decreases myosin activity and alters tissue morphogenesis (Denker et al., 2015; Kasza et al., 2014). Therefore, we overexpressed the nonphosphorylatable MRLC mutant (T18A-S19A) in the notochord. The T18A-S19A-expressing notochord showed less tail-bending morphology (n=47/72; Fig. 3A, bottom) compared with the control group (Fig. 3A, top). Quantitative data showed that θ in the T18A-S19A-expressing embryos was 39.9°±19° (mean±s.d., n=47; Fig. 3B), significantly lower compared with the control group (θ=75.4°±17°, n=20; Fig. 3B).

Cofilin, a highly conserved actin-binding protein, plays an important role in many cellular processes (Lappalainen and Drubin, 1997), such as the furrow formation in cytokinesis (Abe et al., 1996; Nakano and Mabuchi, 2006) and cell elongation in Ciona notochord morphogenesis (Sehring et al., 2014). In mammals, cofilin phosphorylation at serine 3 prevents cofilin-actin binding and stabilizes F-actin. Overexpression of the nonphosphorylatable cofilin mutant cofilin S3A inhibits RhoA-stimulated actin polymerization (Arber et al., 1998; Sotiropoulos et al., 1999). Ciona has only one cofilin gene, in which serine 5 is equivalent to the serine 3 in mammalian cofilin (Sehring et al., 2014). To disrupt F-actin dynamics, we made Ciona cofilin S5A mimic the nonphosphorylated cofilin acting as a dominant-negative cofilin mutant and electroporated it into Ciona embryos. The results showed that the difference between ventral and dorsal F-actin enrichment was greatly reduced in the cofilin S5A-expressing notochord (n=16/24; Fig. 3C, bottom) compared with the control group (n=15/19; Fig. 3C, top). Quantitative data showed that θ in the cofilin S5A group was 52.03°±17.9° (n=43; Fig. 3D), which was lower compared with the control group (θ=80.20°±19.77°, n=21; Fig. 3D). Thus, disruption of both myosin II and F-actin activities and localization data showed that asymmetrical actomyosin contractility along the ventral side of the notochord is required for embryonic tail bending.

Next, we measured the tail epithelial cell geometry and counted the number of cells at the dorsal and ventral sides of the notochord, respectively. Although the cell geometry was different (Fig. S3A-E), the more prominent phenotype was the number of cells (Table S1). BrdU staining of Ciona embryos showed that the dorsal region, including the dorsal midline of the tail epidermis and neural tube, contained more dividing cells compared with the ventral region, including the ventral midline of the tail epidermis and endodermal strand (Fig. 4A-C; Movie 5). The number of divided dorsal epithelial cells was significantly higher compared with the ventral epithelial cells (Fig. 4D).

To determine the role of differential epidermis proliferation activity in embryonic tail bending, we blocked cell proliferation by incubating embryos at tailbud stage 16 using the DNA replication inhibitor aphidicolin (Ikegami et al., 1978). BrdU staining confirmed the efficiency of cell cycle inhibition by aphidicolin (Fig. S3H). Embryos developed normally, at least without detectable abnormality during the early tailbud stages (tailbud stages 17-19) in the aphidicolin-treated group (Fig. S3F,G). However, embryos showed abnormal morphology with a shorter and ill-bent tail at tailbud stage 22 (Fig. 4E-G; Fig. S3F), indicating that cell proliferation plays an essential role in embryonic tail bending at later tailbud stages.

To further analyze the contribution of differential epidermis proliferation dorsally and ventrally, we used the mosaic expression pattern of electroporated transgenes in Ciona embryos (Zeller et al., 2006). Ectopic Ci-cdc45 overexpression inhibited epithelial cell proliferation (Fig. S3I), although the underlying mechanism was unclear. Ci-cdc45 was ectopically overexpressed under the control of the klf 1/2/4 promotor, exclusively expressed in the dorsal and ventral midline of the epidermis of Ciona embryos (Pasini et al., 2006). We found that the epithelial cell proliferation inhibited by Ci-cdc45 overexpression leads to shorter tail length compared with the control embryo at tailbud stage 23 (Fig. 4H,I), mimicking the effect of aphidicolin treatment (Fig. 4E,F). Tail bending was normal at the early tailbud stage (tailbud stage 19) (Fig. 4H), but we found two distinct phenotypes because of two different expression patterns of the transgene at a later tailbud stage. Tail bending was normal when Ci-cdc45 was expressed symmetrically in the dorsal and ventral midline of the epidermis [Fig. 4H,J, pattern I (blue)]. In contrast, when Ci-cdc45 was dominantly expressed in the dorsal midline of the epidermis, the embryonic tail bent dorsally and presented a more diverse morphology [Fig. 4H,J, pattern II (pink)].

Together, our findings show that both asymmetrical actomyosin constriction of the notochord and differential epidermis proliferation are required for embryonic tail bending in Ciona.

The above results show that both asymmetrical actomyosin constriction of the notochord and differential cell division in the epidermis are required for the embryonic tail bending in C. robusta, but we are yet to determine the main driving force in this event. From a mechanics perspective, both of them can drive tail bending, as long as the induced force is large enough. We measured the Young's moduli, the parameter that describes material stiffness, of the notochord, muscle and epidermis by nanoindentation using an atomic force microscope (AFM) (Fig. 5A,B). Measurements showed that the Young's moduli of the notochord, muscle and epidermis decrease in turn (Fig. 5C). It is more likely that the main driving force of embryonic tail bending originates from the bending of the notochord, the stiffest tissue, than the asymmetrical division of the relatively soft epidermal cells. To verify our hypothesis, a finite element model was built using Abaqus (Hibbitt et al., 2011), as shown in Fig. 5D.

In the finite element simulations, we considered two extreme situations: the contraction of the notochord serves as the exclusive driving force, and the differential cell division serves as the exclusive driving force. The contraction of the notochord and the cell division are treated by including a growth field. The enlargement of the tissues arising from cell elongation and division can be simulated by introducing a positive growth ratio. The contraction of the tissues, by contrast, can be achieved by introducing a negative growth ratio. To simulate the bending of the tail, a simple and typical method is to include a gradient growth field (Fig. S4A). For the notochord-driving scenario, the elongation and contraction of the notochord are considered together and included by a gradient growth field, whereas the epidermal cells are treated as a passive material, undergoing elastic deformation. For the differential cell division-driving scenario, the differential cell division is included by a gradient growth field, whereas the notochord is treated as a passive material. In the simulation, we can adjust the magnitude of the growth field to fit the bending angle and the elongation of the tail observed in the experiments. However, the stress or strain level in the tail does not influence the relative contributions of the epidermis and notochord. The results show that, if the contraction of the notochord serves as the main driving force for tail bending, the epidermis should be stretched along the AP axis (Fig. 5E, top). On the contrary, if the differential cell division drives the tail bending, the epidermal cells at the ventral side should be squeezed (Fig. 5E, bottom). To test which scenario is correct, we measured the epidermal cell geometry and found that the dorsal and ventral epidermal cells flattened gradually in the DV direction as the tail bends (Table S1). Owing to the Poisson's effect, a material tends to expand in the directions perpendicular to the direction under compression. Conversely, if the material is stretched, it usually tends to shrink in the perpendicular directions. Therefore, the flattening of the epidermis on both dorsal and ventral sides suggests that the epidermal cells are stretched along the AP axis. It is worth noting that elongation of the notochord is neglected in the exclusive epidermis-driving scenario. The elongation of the notochord will stretch the epidermal cells and counteract the compression of the epidermal cells to some extent. Thus, we cannot totally exclude the possibility that cell division plays a role in tail bending. In this vein, the differential cell division may drive the tail bending, only up to a point. To quantify the contributions of the notochord and epidermis, we further propose a theoretical physical model (see Supplementary Materials and Methods). Based on geometric parameters of the same embryo for different stages (Table S1), we calculated the contributions of the epidermis differential cell division and the notochord contraction. The results show that the bending moment induced by the epidermis is dispensable compared with the moment required to bend the tail (Fig. S4). Therefore, the contribution of differential cell division in the epidermis to the tail bending is relatively limited at the initial stage.

Owing to the difficulty of experimental measurement of constriction force in the notochord in vivo, we took the value of the constriction force in actin bundles of cell groups for reference: f_c is in the order of 10 nN (Effler et al., 2006; Poirier et al., 2012; Soiné et al., 2015; Zhang and Robinson, 2005). Though the reference value may deviate from the actual value of Ciona, this does not affect the underlying mechanical mechanisms. The rotational angle α was calculated to be ∼24.2° when f_c=20 nN, which is equal to the tail bending angle θ (Fig. S5). This result suggests that actomyosin constriction at the ventral side of the notochord is sufficient to drive the embryonic tail bending in C. robusta.

In Ciona embryogenesis, the embryonic tail bends and elongates smoothly and continually. In this study, we presented data to show that Ciona embryonic tail bending is an autonomous large-scale morphogenic process mediated by tail tissues and controlled by genetics. During the early tailbud stage, the notochord generates asymmetrical contractile forces to drive embryonic tail bending, whereas differential epidermis proliferation dorsally and ventrally plays a permissive role and allows embryonic tail bending (Fig. 6). These results show a robust tail-bending mechanism at the embryonic scale and deepen our understanding of embryonic tail bending commonly found in invertebrates and vertebrates.

The notochord, typically found in chordate embryos, is located in the midline of the embryonic tail. It acts as a signaling center and provides structural support (Stemple, 2005). A mechanical challenge for the Ciona embryo to overcome during embryonic tail bending is mechanical resistance produced by the stiff notochord. Our analysis of the actomyosin network in Ciona embryos during early tail bending showed asymmetrical accumulation of actomyosin at the ventral side of the notochord, which generates mechanical forces to drive tail bending by contractility that is orthogonal to the bending axis. Polarized actomyosin contractility, such as apical or basal constriction, plays an essential mechanical role in tissue bending and/or folding (Gutzman et al., 2008; Haigo et al., 2003; Lee and Goldstein, 2003; Martin et al., 2009; Nicolás-Pérez et al., 2016; Nishimura et al., 2012; Sherrard et al., 2010). The notochord, and hence embryonic tail bending, show defects as actomyosin organization and activities decrease because of myosin II and cofilin mutations. Our physical model predicted that the polarized actomyosin contractility along the ventral side of the notochord is sufficient to drive embryonic tail bending. Therefore, the notochord is the key driver of tail bending in Ciona embryogenesis.

Polarized actomyosin localization has been found in the Ciona notochord and plays an indispensable role in notochord morphogenesis. For example, the tip polarization of actin is important for notochord convergence/extension (CE) (Munro and Odell, 2002), whereas anterior polarization of myosin and the contractile ring are important for notochord elongation (Newman-Smith et al., 2015; Sehring et al., 2014). The actomyosin in the notochord displays clear DV polarity patterning during the early tailbud stage, which is required for embryonic tail bending. How the actomyosin is directed to accumulate along the ventral side of the notochord at the early tailbud stage is unclear. Some proteins, such as those found in the extracellular matrix, also show polarity during notochord morphogenesis (Oda-Ishii et al., 2010; Veeman et al., 2008). Interestingly, during notochord CE, the notochord preferentially accumulates basement membrane marker laminin dorsally and the apical cell polarity molecule atypical protein kinase C (aPKC) ventrally (Oda-Ishii et al., 2010), which might provide a polarizing cue for polarized actomyosin enrichment.

We also provided evidence of an epidermis-dependent mechanism regulating tail bending at later tailbud stages. BrdU staining showed that dorsal midline cells in the tail epidermis divide faster compared with the ventral side during embryonic tail bending. Spatial differences in cell proliferation rates can cause folding and/or bending shapes during development (Mao et al., 2013; Miller et al., 1999; Tozluoǧlu et al., 2019). However, our results revealed that this bias in cell proliferation is required for embryonic tail bending at late tailbud stage rather than early tailbud stage. It appears that the tail epidermis plays a dominant role in producing the tail-bending structure at late tailbud stage. Consistent with this, tail-bending deficiencies were observed at late tailbud stage of miniature (dwarf) tailbud embryos, which are composed of relatively smaller epidermal cells compared with wild-type embryos, indicating that overall decrease in the number of epidermal cells changes the mechanical state of the tailbud embryos (Matsumura et al., 2020). In this study, we believe that the faster cell proliferation in the dorsal midline epidermis can release the accumulated mechanical stress generated by asymmetrical notochord contractility and/or promotes polarized tissue shrinkage of tail epidermis. Both mechanisms are able to facilitate embryonic tail bending ventrally, which is confirmed by the fact that cell proliferation inversion at the dorsal midline of the tail epidermis causes embryonic tail bending dorsally at later tailbud states. In addition, tail epidermal cells tend to divide parallel to the AP axis during the early tailbud stages (Negishi et al., 2016; Ogura et al., 2011), which supports the idea that this oriented cell proliferation might also participate in the relief of local tissue resistance along the AP axis, facilitating embryonic tail bending. Mechanical stretching promotes epithelial cell proliferation (Gudipaty et al., 2017). Therefore, it would be interesting to analyze whether mechanochemical feedback exists between mechanical stretching and differential epidermis proliferation (speed and/or orientation) in Ciona embryonic tail bending.

The notochord and tail epidermis are both essential for Ciona embryonic tail bending. However, tail bending likely originates from the overall effect of the high coordination between different tail tissues. Indeed, the polarized contractility of the notochord plays a major role in shaping the bending tail at early tailbud stages, whereas bias of differential epidermal proliferation ensures the robustness of tail bending at later tailbud stages. However, our data did not rule out the possibility that other tissues and their interactions also contribute to embryonic tail bending. For example, embryos treated with SU5402, the fibroblast growth factor signaling inhibitor, showed morphologically defective tail bending and elongation, although the notochord preserved its bending property (Ikuta et al., 2010), indicating that the role of the notochord in driving embryonic tail bending depends on the synergistic effect of other tissues. In addition, a posterior-to-anterior zippering event occurs during the same period as early embryonic tail bending in the neural tube (Hashimoto et al., 2015), which asymmetrically localizes on the dorsal side of the embryos. Whether this cellular process affects embryonic tail bending is unclear.

According to von Baer's laws (von Baer, 1928), vertebrate embryos converge on a common physical structure and hence show a similar morphology during early embryogenesis, called phylotypic stages (Hall, 1997). For example, at the beginning of neurulation, the embryo in chordates is commonly C-shaped (Gilbert and Barresi, 2016). Embryonic tail bending helps embryos elongate continuously within the chorion without mechanical damage. In tunicates, the closest living relatives of vertebrates (Delsuc et al., 2006), the embryos also show a bending morphology at a similar tailbud stage. This interesting phenomenon gives rise to the question, why and how do embryos assume a bending shape from the viewpoint of evolution and development? Our study of the mechanism of ascidian embryonic bending is a first step toward our understanding of embryo bending in an evolutionary perspective.

Embryonic tail bending within the chorion is a chorion-independent process under genetic control. The head and trunk are not required for tail bending, suggesting that it is a self-organized event. Furthermore, we showed the existence of ventrally localized actomyosin in the notochord and differential epidermis proliferation between the dorsal and ventral midline during embryonic tail bending. Through a combination of genetic perturbation and chemical drug manipulation, we revealed that both asymmetrical notochord contractility and differential epidermis proliferation are important for embryonic tail bending. Asymmetrical notochord contractility is sufficient to drive embryonic tail bending, whereas differential epidermis proliferation is a passive response to mechanical forces. Our findings provide insights into how different tissues coherently coordinate to achieve a large-scale morphogenetic event at the embryonic level. The mechanisms underlying embryonic tail bending in structure-simple chordate embryos will help us understand their more complex counterparts in vertebrate systems.

Adult C. robusta individuals were collected from the Yellow Sea in Weihai, China and kept in a recirculating aquarium at 18°C until analysis.

All procedures involving live animals were performed in filter-fresh seawater. Developing embryos were staged according to the criteria provided by Hotta et al. (2007).

Fertilized eggs were dechorionated, as previously described (Corbo et al., 1997). The fertilized eggs (10 min post fertilization) were immersed in seawater containing 1% sodium thioglycolate, 0.05% protease E and 32 µl of 10 N NaOH. The chorion was removed within 5 min at room temperature (RT) by gently pipetting. Dechorionated eggs were washed four times with filter-fresh seawater and transferred into 1% agarose-coated plastic Petri dishes for culturing or immediately used for plasmid electroporation. Embryos were cultured at 16°C.

For tail-cut experiments, embryos at tailbud stage 19 were used, unless otherwise specified. All procedures were performed in a room maintained at ∼20°C. A hand-held glass needle was prepared by pulling a glass capillary. Each embryo was carefully cut along the boundary between trunk and tail to obtain isolated tail fragments (CT). The A-CT or P-CT was obtained by further cutting CT fragments.

Plasmid DNA (60 µg or 80 µg in 80 µl) was mixed with 420 µl of 0.77 M mannitol in 4 mm cuvettes. Then 300 µl of dechorionated fertilized eggs was added and electroporated. Electroporated embryos were cultured in Millipore-filtered seawater at 16°C or 18°C.

A C. robusta klf 1/2/4 promoter was PCR-amplified by the forward primer 5′-CTCAAGCTTCGAATTCGGCATATGATACATCATGTGTGTCT-3′ and reverse primer 5′-GGCGACCGGTGGATCCTTCAAACCAATTACCATTCGTCTA-3′. The purified PCR product was inserted into promoter-less enhanced green fluorescence protein 1 (pEGFP-1) to create klf>eGFP using the standard In-Fusion^® cloning protocol (TaKaRa Bio). Cr-cdc45 was amplified using the primer obtained from Sehring et al. (2014). The eGFP::Cr-cdc45 was obtained from cytomegalovirus (CMV)>eGFP::Cr-cdc45 digested with BamH1 and Not1. The eGFP complementary DNA (cDNA) of klf>eGFP was replaced by a cDNA of eGFP::Cr-cdc45 to create klf>eGFP::cdc45. Cr-cofilin S5A was PCR-amplified by the forward primer 5′-ATGGGCGTAAGTGCCGGAATTA-3′ and the reverse primer 5′-TTAGACATATTCATATGCGTTACT-3′. The PCR products were used to create entry clones using the pCR8/GW/TOPO system (Invitrogen). The entry clone was used to generate cofilinS5A-mCherry expression constructs using the destination vector Minos-B3-eBra-bpFOG-B5::R1-ccdB/CmR-R2-mCherry (Dong et al., 2009). The following expression constructs have been described previously: MRLC-mCherry (Dong et al., 2011), MRLC T18A-S19A-mCherry (Denker et al., 2015) and cofilin-mCherry (Sehring et al., 2014).

Time-lapse movies were taken using either a Nikon Eclipse Ni fluorescence or a Nikon A1 confocal microscope. To observe the development of chorionated and dechorionated tailbud embryos, or cut tails (CT, A-CT and P-CT), corresponding embryos or isolated tails were captured under a microscope equipped with a DS-Ri2 camera, the interval between frames was 30 s. The duration depended on embryo development at ∼20°C and was ∼2 h for all CT developments.

To inhibit cell proliferation, embryos were treated with 2 µg/ml aphidicolin or dimethyl sulfoxide (DMSO; control) at the start of the tailbud stage. All embryos were cultured and imaged at ∼20°C. Imaging was performed using the Eclipse Ts2R inverted microscope (Nikon), and images of developing embryos were captured manually every 30 min until 3.5 h post-treatment. To inhibit actomyosin activity, embryos at 8 h post fertilization (hpf), 9 hpf and 11 hpf at 18°C were treated with 100 µM blebbistatin or 200 µM Y-27632 and imaged subsequently in a room maintained at 20°C.

Embryos at tailbud stages 17-22 were collected and fixed with 4% paraformaldehyde (PFA) in seawater for 2 h at RT. The fixed embryos were washed thrice with phosphate-buffered saline (PBS) containing 0.1% Triton X-100 (PBST) for 15 min each time. To detect F-actin, the fixed embryos were stained with 1/200 Alexa Fluor^® 488-conjuated Phalloidin (Invitrogen, 11791-020) in PBS overnight at 4°C. They were washed thrice with PBS, mounted using Vectashield (Vector Laboratories) mounting medium with 4′,6-diamidino-2-phenylindole (DAPI), and stored at 4°C.

For pS19 MRLC staining, we followed the procedure as described previously by Sherrard et al. (2010) with minor modifications. Embryos at tailbud stage 17-22 were collected and fixed in 100 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; pH 6.9-7.0), 100 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA; pH 7.0), 10 mM MgSO₄, 2% formaldehyde, 0.1% glutaraldehyde and 300 mM dextrose for 1 h. Next, the embryos were washed thrice with PBST and then thrice with PBS. To reduce unreacted aldehydes, the embryos were treated with 0.1% sodium borohydride in PBS for 20 min and then washed thrice with PBS. Fixed embryos were blocked using 0.1% bovine serum albumin in PBS for 1 h at RT and then incubated with ser19 phospho-myosin (1/250; Cell Signaling Technology, 3671) for 48 h at RT. Next, they were washed thrice within 8 h and incubated with Alexa Fluor^® 568-conjugated anti-rabbit immunoglobin G (IgG) secondary antibody (1/250; Invitrogen, A11011) for 48 h at RT. Finally, the embryos were rinsed thrice and then mounted for imaging, as described above.

To detect cell proliferation, we used the standard procedures for BrdU staining, as described previously (Sehring et al., 2014). Embryos were treated with BrdU (Invitrogen, B23151) at 8.3, 9, 10, 11 and 12 hpf for 45 min at 18°C. The BrdU-incorporated embryos were washed twice with seawater for 10 min each time at 18°C and quickly fixed with 4% formaldehyde in seawater for 2 h at RT. To detect BrdU, the embryos were washed thrice with PBST, permeabilized in acetone for 10 min at −20°C and again washed thrice with PBST. Before immunostaining, the embryos were incubated in 4 N HCl for 10 min at RT to denature DNA, washed thrice, blocked with PBST containing 10% goat serum for 1 h at RT, and incubated with mouse anti-BrdU primary antibody (1/150; Invitrogen, B35128) for 16 h at 4°C. Next, the embryos were again washed thrice, incubated with Alexa Fluor^® 568-conjugated anti-mouse IgG (1/300; Invitrogen, A21202) overnight at 4°C, rinsed thrice and finally mounted for imaging, as described above.

To thoroughly detect the F-actin distribution of the notochord, embryos positioned on the longitudinal, axial and coronal planes of the notochord were selected for imaging (Fig. 2; Fig. S2). The embryos were imaged using an A1 confocal microscope (Nikon). To measure relative fluorescence intensities (F-actin or pS19 MRLC) along the dorsal or ventral notochord edges, we obtained mean gray levels averaged over a line drawn along the notochord edges of interest using ImageJ (https://imagej.nih.gov/ij/index.html). For each measurement, we measured the maximum fluorescence intensity projections from the longitudinal view. To analyze the distribution of BrdU-positive cells, we rendered 3D projections of confocal images of BrdU-stained embryos in NIS-Elements AR ver. 4.40.00 (Nikon) and manually counted the number of BrdU-positive cells in the dorsal and ventral midline of the tail epidermis, neural tube and endodermal strand.

To further understand the mechanical mechanism underlying the tail bud bending, a three-dimensional finite element model was built (Fig. 5D). Its geometrical parameters were taken according to our experimental measurements (Table S1) and the Young's modulus of each component was taken from our AFM tests. The simulations were made using the software Abaqus (Dassault Systems). The system was meshed by C3D8R elements and a mesh sensitivity analysis was made to ensure that the element sizes did not interfere with the numerical results.

To quantify the tail length of an embryo, the length of the midline (Fig. 1B, green) of the tail was defined as tail length ζ. To quantify the embryonic tail-bending angle, we assigned the cross-angle of the magenta and blue lines (Fig. 1B) as the bending angle θ. The magenta line is along the boundary that separates tail and trunk territories, whereas the blue line intersects with the green line at the tail length.

To quantitatively analyze the embryo geometry, we labeled embryos using Phalloidin (Alexa Fluor^® 488) at tailbud stages 17-22. Phalloidin staining helps to easily define cell boundaries of C. robusta embryos. We measured several aspects related to embryonic- or tissue-level geometries from the longitudinal view in median sections using ImageJ (https://imagej.nih.gov/ij/index.html) (Figs S1 and S3; Table S1).

All graphs in this study were produced using Matlab, Microsoft Excel or gglot2 in R. The significance of differences was calculated using two-tailed Student's t-tests. P<0.05 was considered statistically significant.

We thank lab members in the B.D. laboratory, Dr Prachi Richa from the University of Göttingen and Dr Xin Liang from Tsinghua University for the discussion and critical comments on the manuscript.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.185868.reviewer-comments.pdf