Admp regulates tail bending by controlling ventral epidermal cell polarity via phosphorylated myosin localization in Ciona

Although chordates display diverse shapes and sizes in the adult stage, they have similar shapes during the organogenesis period, called the phylotypic stage (Klaus, 1983). During the phylotypic stage, chordates pass through neurulation and subsequently reach the tailbud stage. At the chordate tailbud stage, the embryo tail elongates along the anterior-posterior (AP) axis, and most tailbud embryos become curved with their tail bending ventrally (Richardson et al., 1997).

The ascidian tunicate Ciona intestinalis type A (Ciona robusta) embryo also shows a curved body shape with its tail bending ventrally (ventroflexion) during the early- to mid-tailbud stages (stage 19 to stage 22, as defined by Hotta et al., 2007), after which bending relaxes again, and eventually the tail bends dorsally (dorsiflexion). This dynamic body shape change occurs even if the egg envelope is removed, suggesting that Ciona tail bending can occur in the absence of external spatial confinement (Hotta et al., 2007; Nakamura et al., 2012; Matsumura et al., 2020; Lu et al., 2020).

During tail extension, notochord cells change their shape by circumferential contraction between stages 21 and 24 (Lu et al., 2019; Mizotani et al., 2018; Sehring et al., 2014). This circumferential contraction, when applied on a compression-resisting system, such as the notochord, is converted into a pushing force along the AP axis of the notochord, thereby elongating the notochord (Lu et al., 2019; Miyamoto and Crowther, 1985).

Although it has been shown that an AP pushing force is exerted by each notochord cell (Sehring et al., 2014; Zhou et al., 2015) and that tail elongation is achieved by the notochord actively producing AP elongating forces (Miyamoto and Crowther, 1985; Sehring et al., 2014; Spemann, 1987; Ubisch, 1939), the mechanism by which the tail bends in the tailbud-stage embryo is only incompletely understood. Recently, Lu et al. (2020) have shown that tail bending in Ciona during the early tailbud stages (stages 18 to 20) is caused by the actomyosin cytoskeleton displaying contraction forces that are different at the ventral side compared with the dorsal side of the notochord. However, the upstream regulators involved in Ciona tail bending and the morphogenetic mechanisms driving tail bending after stage 20 remain unclear.

In this study, we used a combination of genetic, cell biological and biophysical/three-dimensional (3D) imaging experiments to show that Admp regulates cell polarity by determining the localization of phosphorylated myosin light chain (pMLC) at the apex of ventral midline epidermal cells. This ventral epidermal myosin accumulation leads to ventral tail bending by resisting notochord-driven AP tail elongation specifically at the ventral side during mid-tailbud stages.

Previous studies about Admp function did not focus on tail bending but the phenotype was apparent in their images. The knockdown of Admp has been shown to cause reduced ventral tail bending in mid-tailbud-stage Ciona embryos (Imai et al., 2006; Imai et al., 2012; Pasini et al., 2006). As these studies did not focus on the tail-bending morphant phenotype, we decided to mechanistically dissect how Admp functions in ventral tail bending.

To confirm that Admp is indeed required for Ciona ventral tail bending, we first performed microinjection of Admp morpholinos (MOs) and observed the morphant phenotype by recording time-lapse movies (Fig. 1A; Movie 1). We found that ventral tail bending (ventroflexion; Fig. 1A, red arrow) did not occur in Admp morphant embryos at the mid-tailbud stages; in contrast, dorsal tail bending (dorsiflexion; Fig. 1A, yellow arrows) was unaffected in morphant embryos. Comparison of the bending angle of Admp morphant with wild-type (WT) embryos from stage 18 to stage 22, when ventroflexion occurs in WT (Fig. 1B), showed that the degree of ventroflexion was significantly reduced in morphant embryos (Fig. 1C; N=11/11). Likewise, embryos treated with dorsomorphin, an Admp/BMP signaling inhibitor, also displayed a significantly reduced ventral tail bending angle (n=5-12, Fig. 1D; Fig. S1A,B). Taken together, these experiments indicate that Admp/BMP signaling regulates the ventroflexion of ascidian tailbud embryos.

Admp is a BMP ligand, which, in Ciona, has been reported to induce the differentiation of ventral peripheral neurons (Imai et al., 2012; Waki et al., 2015), with the homeobox gene Msxb functioning as a downstream effector of Admp signaling in this process (Imai et al., 2012). We thus investigated whether Msxb might also function as a downstream effector of Admp signaling in Ciona ventroflexion. However, ventroflexion appeared normal in Msxb morphant embryos (Fig. S1C), suggesting that Msxb, does not function as a downstream effector of Admp signaling in ventroflexion, unlike in neuronal differentiation (Roure and Darras, 2016; Waki et al., 2015).

In Ciona, Admp is expressed in the endoderm and lateral epidermis (Imai et al., 2012). In vertebrates, Admp is first expressed dorsally within the embryo, and then Admp protein physically moves to the opposite side to specify the ventral fate, but it is difficult to predict the site of Admp activity from its gene expression pattern. Moreover, Admp promotes bmp4 expression and controls the positioning of bmp4 expression during regeneration of left-right asymmetric fragments in planarians (Gaviño and Reddien, 2011).

In Ciona, Admp expression appears normal in Bmp2/4 morphants, but Bmp2/4 expression is suppressed in Admp morphants (Imai et al., 2012). Furthermore, the BMP target Smad is phosphorylated by Admp signaling, followed by translocation of phosphorylated Smad into the nucleus and activation of target genes (Blitz and Cho, 2009; De Robertis, 2009; Imai et al., 2012). In line with this, Smad phosphorylation and activation in ventral epidermal cells is reduced in Ciona Admp morphants at the late gastrula stage (Fig. 1E; Waki et al., 2015).

To determine when and where within the tailbud stages Ciona embryo Admp/BMP signaling is activated, we performed antibody staining of phosphorylated pSmad1/5/8 (Fig. 1E). Consistent with a previous study (Waki et al., 2015), pSmad staining was observed in ventral midline epidermal cells after the late gastrula stage (Fig. 1E), whereas no specific signal was detected in other regions, including the notochord, from the gastrula to the initial tailbud period. This indicates that Admp/BMP signaling is specifically activated in ventral midline epidermal cells.

Asymmetric activation of actomyosin contractility in notochord cells has recently been proposed to be responsible for ventroflexion between stages 18 and 20 (Lu et al., 2020). To test whether Admp functions in ventroflexion by affecting asymmetric actomyosin contraction within notochord cells, we analyzed actin localization in Admp signaling-defective embryos. We found that in both dorsomorphin-treated and Admp morphant embryos, asymmetric actin localization in notochord cells remained unchanged (Fig. S2). This indicates that Admp/BMP signaling affects ventroflexion independently of the proposed function of asymmetric actomyosin contraction in notochord cells.

Next, we investigated the dynamics of dorsal and ventral epidermal cell rearrangements during ventral tail bending from stage 18 to stage 24 (Fig. S3). Cell-cell intercalation of ventral epidermal cells started at stage 19 and was completed by stage 24. The tail epidermis of the ascidian embryo completes elongation along the AP axis by arranging epidermal cells in a row along this axis through cell-cell intercalation (Fig. S3; Hotta et al., 2007). Interestingly, in the period between stages 19 and 22, which reflects the early phase of epidermal cell-cell intercalation when ventroflexion occurs, intercalation was not associated by AP elongation of the ventral tail, whereas during later stages of epidermal cell-cell intercalation (stage 22 to stage 24), intercalation was accompanied by ventral tail elongation (Fig. 2A,B). We thus hypothesized that epidermal cell dynamics during the early intercalation period contribute to ventroflexion. To test this hypothesis, we compared epidermal cell dynamics between WT and ventroflexion-deficient Admp morphant embryos.

Between stages 20 and 22, the ventral epidermis in WT embryos showed a preferential accumulation of junctional F-actin in the medio-lateral (ML) direction (ML accumulation) (Fig. 2C). Antibody staining of pMLC also showed such ML accumulation, especially from stage 19 to stage 22 (Fig. S4). In contrast, no such ML accumulation was observed in Admp morphant embryos between stages 20 and 22 (Fig. 2C). In addition, although the AP/ML aspect ratio of ventral epidermal cells decreased in WT embryos from stage 18 to stage 22, no such decrease was found in Admp morphants (Fig. 2D). This suggests that Admp is required for proper asymmetric junctional actin accumulation and ML elongation of ventral epidermal cells during early intercalation. At stage 24, the tail epidermis in WT embryos became organized into eight distinct single-cell rows as a result of cell-cell intercalations (Fig. 2E,F) (Hotta et al., 2007; Pasini et al., 2006). Moreover, these eight rows, consisting of three rows of dorsal, two rows of lateral, and three rows of ventral epidermal cells, were closely aligned (Fig. 2F, WT). In contrast, the ventral epidermal cells in Admp morphant embryos were disorganized into one or two rows, making it difficult to clearly distinguish between midline and medio-lateral cells (Fig. 2F, Admp MO; orange/red color). Dorsomorphin-treated embryos showed a similarly disordered ventral midline intercalation phenotype (Fig. S5), further supporting the notion that Admp regulates ordered ventral epidermal cell-cell intercalation.

We suspected that defective ventroflexion in Admp morphant embryos involves changes in ventral epidermal cell morphologies (Fig. 2F). To further investigate in detail the morphological change that occurs in the ventral epidermal cells during this period, we monitored changes in single ventral epidermal cell morphology by 3D imaging. This revealed that ventral epidermal cells acquire a distinctive boat-like morphology (boat cell, Fig. S6), characterized by a larger area on the basal surface (Fig. 3A,B, yellow areas) compared with the apical surface (Fig. 3A,B, red areas), and ridges at both ends oriented along the ML direction. Almost all anterior ventral epidermal cells showed this shape (Movie 2), consistent with previous reports that tail bending only occurs in the anterior tail of Ciona (Lu et al., 2020). The shape of boat cells is characterized by a triangular cross-section where the apical surface is entirely constricted (triangular section of boat cell, TSBC), and a square cross-section where some apical surface is left (square section of boat cell, SSBC) (Fig. 3A,B). In Admp morphant embryos at stage 22, the number of TSBCs in ventral epidermal cells was strongly reduced (Fig. 3D,E; Admp MO, n=12; WT, n=7; P=0.05×10⁻⁵, two-tailed, unpaired Student's t-test), whereas the number of square sections of cells that were not boat cells was increased, indicative of a reduced number of boat cells in all ventral epidermis sections of morphant embryos (Movie 2).

To understand how this distinctive boat-cell morphology (Fig. 3A,B) arises, we performed both F-actin/phalloidin staining and antibody staining for pMLC. This showed an accumulation of both F-actin and pMLC at the apical side of TSBCs (Fig. 4A,B, WT arrowheads). Interestingly, the localization of pMLC to the apical side was significantly decreased in Admp morphant embryos (Fig. 4C; Admp MO, n=7; WT, n=9; P=0.01, two-tailed, unpaired Student's t-test), suggesting that Admp triggers the formation of TSBCs, and thus boat-cell shape, by localizing pMLC to the apical side of ventral midline cells.

To test whether Admp/BMP signaling can ectopically affect the localization of pMLC and thereby generate TSBCs (Fig. 4D), we performed ectopic Bmp expression experiments. In WT embryos, apical pMLC accumulation and TSBCs were not observed in epidermal cells, except in ventral epidermal cells, where pSmad signal was also detected (Fig. 4D, frontal section of WT). In contrast, in embryos ectopically expressing BMP, pSmad signal was detected in all epidermal cells, accompanied by apical pMLC accumulation and TSBC formation not only in ventral tail epidermal cells but also within the remainder of the tail epidermis (Fig. 4D).

This suggests that Admp/BMP signaling is sufficient to induce the localization of pMLC to the apical side of epidermal cells (Fig. 4E), leading to the formation of boat cells. This cell-shape change, again, might resist tail elongation at the ventral tail region, leading to ventroflexion.

To investigate whether the ventral epidermis indeed locally resists tail elongation, eventually leading to ventroflexion, we cut either ventral or dorsal epidermal cells at their apex along the AP axis using an ultraviolet (UV) laser cutter (Fig. 5, yellow lines). When dorsal midline epidermal cells were cut, no effect was observed. However, cutting ventral midline epidermal cells led to a strong relaxation of ventroflexion, indicative of stress release along the AP axis at ventral midline epidermal cells (Fig. 5A,B; Movie 3). These findings suggest that apically accumulated pMLC in boat cells generate AP stress in the ventral midline epidermis, resisting tail elongation and thus enabling ventroflexion.

We further investigated whether the relaxation of ventroflexion by UV laser cutting depends on whether the cuts were oriented along the AP axis or ML axis in ventral midline cells. This showed that the relaxation of ventroflexion was slower in AP cuts compared with ML cuts (Movie 4), consistent with the notion that ML cuts more efficiently interfere with AP stress in the ventral epidermis midline than AP cuts.

Next, we asked how boat cells change their shape during the ventral tailbud period (Fig. 6A). To this end, we analyzed the subcellular distribution of pMLC as a proxy for actomyosin contraction in ventral midline cells (ventral view in Fig. S4). This revealed that pMLC first emerged at the ML junction of ventral midline epidermal cells at stage 21 (Fig. 6A, stage 21 arrowheads) and localized to the apical side of TSBCs at stage 22 (Fig. 6A, stage 22 arrowheads). At stage 23, the trapezoid shape became apparent at the midline plane (Fig. 6, stage 23; Fig. S4) and, finally, the apical accumulation of pMLC disappeared at stage 24 (Fig. 6, stage 24; Fig. S4). These observations suggest that TSBC formation is driven by ML contraction. The ML-directed localization of pMLC and the cell shape change to a boat-like morphology correspond with ventroflexion, and pMLC disappearance and the cell shape change away from the boat-like morphology correspond with relaxation.

Originally, the Admp/BMP pathway was identified to be central in establishing, maintaining and regenerating the dorsoventral axis among bilaterian animals (Gaviño and Reddien, 2011). In the ascidian, both gain- and loss-of-function experiments demonstrated that Admp expressed in the B-line medial vegetal cells acts as an endogenous inducer of the ventral epidermis midline (Pasini et al., 2006). Admp is required for sensory neuron differentiation of the ventral epidermis via expression of the Tbx2/3 and Msxb genes (Waki et al., 2015). Our finding that tail bending was not regulated by Msxb (Fig. S1C) suggests that Admp functions in ascidian tail bending through different effector pathways than those implicated in ventral epidermal cell-fate specification. Admp controls ventral but not dorsal tail bending by determining early ventral epidermal cell intercalation (Fig. 2) and the shape of ventral epidermal midline cells (Fig. 3) through the localization of pMLC (Fig. 4). Importantly, this does not exclude the possibility that genes other than Admp act on tissues other than the ventral epidermis to control ventroflexion.

We divided the intercalation period into two phases: during early intercalation, the ventral epidermis does not elongate in an AP direction; during late intercalation, in contrast, the ventral epidermis elongates. Interestingly, ventral tail bending (ventroflexion) occurs during early intercalation, with ventral epidermal cells displaying a flat shape elongated along the ML axis (Fig. 2C,D). We assume that this ML cell elongation is caused by the accumulation of actomyosin in ML-oriented protrusion-like extensions of ventral epidermal cells (Fig. 6B) that extend cells along the ML axis (Fig. S7). No such actomyosin localization and ML elongation was found in Admp morphant embryos, suggesting that Admp is required for ventral epidermal cell polarization and protrusion formation.

When ventral epidermal cell intercalation is completed (approximately by stage 24), ventral epidermal cells drastically change their polarity into the AP direction in WT embryos. This does not occur to the same extent in Admp morphants (Fig. 2D), suggesting that Admp is also required for this later shift in cell polarity.

Ventral tail epidermal cells in WT but not Admp morphant embryos arrange into three ordered rows at stage 24. This suggests that Admp may be required for the cell autonomous ML intercalation of ventral epidermal cells by controlling ML cell polarization and protrusion formation. Notably, in Admp morphants, the ventral epidermis was disordered but kept a three-cell width, suggesting that some intercalation of ventral epidermal cells might still occur in the absence of Admp.

How does the intercalation of ventral midline cells contribute to ventroflexion? The ventral epidermis undergoing early cell intercalation does not elongate along the AP axis from stage 20 to stage 22, different from the already fully intercalated dorsal epidermis (Fig. 2B). Assuming that the notochord functions as the main force-generating structure driving tail elongation (Dong et al., 2011; Hara et al., 2013; Lu et al., 2019), the lack of ventral epidermis elongation between stages 20 and 22 might locally resist the global notochord-mediated tail elongation, thereby causing the tail to bend ventrally. The lack of ventral epidermis elongation along the AP direction during early intercalation is likely due to the Admp-dependent polarization of ventral epidermal cells along the ML direction. This cell polarization perpendicular to the AP direction can limit epidermal AP elongation during intercalation, which again, by resisting global notochord-driven tail elongation, leads to ventroflexion.

Ventral epidermal cells take a distinct boat-like shape, which likely contributes to ventroflexion (Fig. 4; Fig. S7). The preferential accumulation of pMLC in ventral epidermal cells along ML junctions (Fig. S4Db′,Eb′) is found at the apical side of cell boundaries of TSBCs and/or SSBCs and might correspond to protrusion-like extensions formed between interdigitating boat cells (Fig. 6B). The lack of such polarized distribution of pMLC suggests that Admp might be required for both planar and apicobasal polarization of these cells.

As the cell-cell intercalation of the boat cell progresses, the shape of these cells changes from a triangular to a trapezoidal shape (Fig. 6A). This shape change occurs during the late intercalation period (Fig. 2B) and leads to ventral epidermis elongation along the AP axis (Fig. 6B, green dotted line). Thus, ventral epidermal cell elongation along the ML axis during early intercalation locally resists notochord-mediated tail elongation, thereby triggering ventroflexion (Fig. 6B, stage 20 to stage 22). During later intercalation, in contrast, the ventral epidermal midline cells enlarge their apical area and elongate along the AP axis, thereby relaxing the local resistance against tail elongation (Fig. 6B, yellow dotted line, stage 23).

How does Admp/BMP signaling regulate both the cell-cell intercalation and the apicobasal polarity of the ventral epidermal cells? Our findings suggest that Admp is required for the preferential localization of pMLC not only at ML junctions between intercalating cells (Fig. S4), but also at the apical side by pSmad signaling (Fig. 4D). Recent studies show that SMAD3-driven cell intercalation underlies secondary neural tube formation in the mouse embryo (Gonzalez-Gobartt et al., 2021). Moreover, the BMP-Rho-ROCK1 pathway is thought to target myosin light chain to control actin remodeling in fibroblasts (Konstantinidis et al., 2011). Finally, BMP regulates cell adhesion during vertebrate neural tube closure and gastrulation (von der Hardt et al., 2007; Smith et al., 2021). Yet, how BMP/Smad signaling regulates the localization of pMLC in ventral epidermal cells is still unclear.

Our findings demonstrate that Admp is required for ventroflexion of the ascidian tail during tailbud stages (stages 18 to 22). We propose that Admp phosphorylates Smad in the ventral epidermis. pSmad, in turn, allows early cell intercalation within the ventral epidermis by controlling the localization of the pMLC, leading to ventral epidermal cells taking a boat-like shape. This cell-shape change limits ventral epidermal elongation along the AP axis, thereby locally resisting global notochord-driven tail elongation causing the tail to bend down (Fig. 6).

The notochord has recently been proposed to display asymmetric contraction forces before stage 20 by the asymmetrical localization of actomyosin in notochord cells (Lu et al., 2020). However, Admp morphant embryos displaying straight tails still have a ventral bias in notochord actomyosin localization (Fig. S2). This suggests that Admp is not required for asymmetrical notochord actomyosin localization, and that this asymmetric localization is not sufficient to cause ventroflexion. One possibility is that the ventral accumulation of actomyosin in the notochord might be involved in earlier morphogenetic events, such as notochordal cell intercalation, giving rise to a transient ventral groove (Munro and Odell, 2002).

Our study provides insights into the molecular and mechanical mechanisms underlying conserved shape changes of chordate embryos, such as tail bending. Tail bending in tailbud-stage embryos is a still understudied morphogenetic process, even though many genes, including Admp, with a crucial function in tail bending have been identified in zebrafish (Esterberg et al., 2008; Willot et al., 2002) and frog (Dosch and Niehrs, 2000; Kumano et al., 2006). In invertebrate non-chordate animals, such as sea urchins and hemichordates, Admp is expressed within the embryonic ectoderm (Chang et al., 2016; Lowe et al., 2006). It would thus be interesting to investigate whether the regulation of pMLC subcellular localization by Admp is conserved in primitive chordate embryogenesis and causes a change in body shape in these animals.

Ciona robusta (Ciona intestinalis type A) adults were obtained from Maizuru Fisheries Research Station (Kyoto University, Japan), Onagawa Field Center (Tohoku University, Sendai, Japan) and Misaki Marine Biological Station (University of Tokyo, Japan) through the National Bio-Resource Project (Japan) and Roscof Marine Station (Roscof, France). Eggs were collected by dissection of the gonoducts. To distinguish WT and morphant embryos, embryos were stained with NileBlue B (Wako, Japan), which was gifted by Prof. Hiroki Nishida (Osaka University, Japan). After artificial insemination, fertilized eggs were incubated at 20°C until fixation or observation. Developmental stages followed Hotta's stages (Hotta et al., 2007, 2020). To inhibit phosphorylation of myosin, Y27632 (10 µM, Nacalai Tesque, Japan) was applied to embryos at 7 h post fertilization (late neurula, stage 16). Dorsomorphin (10 µM, Sigma-Aldrich) was applied to embryos after fertilization.

To detect activation of the Admp/BMP signaling pathway, we followed the same method described previously (Waki et al., 2015). To detect pSmad, we used a rabbit primary antibody (1/50, ab216482, abcam) and an Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody from the Alexa Fluor 488 Tyramide SuperBoost Kit, (b40922, Invitrogen). The signal was visualized with a Tyramide Superboost Kit (Invitrogen) using horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG and Alexa Fluor 488 tyramide.

The method of pMLC antibody staining was as follows. Embryos were fixed in 3.7% formaldehyde in seawater for 30 min and then rinsed with PBS with 0.2% Triton X-100 (PBST) for 3 h. Embryos were incubated in PBST containing 10% goat serum for 3 h at room temperature or overnight at 4°C. Embryos were then incubated in the primary antibody (anti-rabbit Ser19 phosphorylated-1P-myosin, 1/50, 3671S, Cell Signaling Technology) diluted to 1:50 incubated for 3 h at room temperature or overnight at 4°C. The samples were then washed with PBST for 3 h. A poly-HRP secondary antibody (goat anti-rabbit IgG, Alexa Fluor 488, Tyramide SuperBoost Kit) was applied for 3 h and washed in PBST for 3 h. The Alexa Fluor dye tyramide (Alexa Fluor 488 Tyramide SuperBoost Kit) was added to the reaction buffer for 5-8 min to induce a chemical HRP reaction. Embryos were dehydrated through an isopropanol series and finally cleared using a 2:1 mixture of benzyl benzoate and benzyl alcohol.

pMLC accumulation was quantified by measuring the intensity along the ventral tail epidermis using Fiji image analysis software. The signal in the brain region was taken as the positive control because its signal was detected even in Y27632-treated embryos, indicating RhoA kinase (ROCK)-independent expression. The relative intensity of pMLC normalized to the intensity of the brain region in each individual was calculated by ImageJ for comparative analysis among different individuals.

UV laser cutting experiments were performed on tailbud Ciona embryos. An inverted Axio Observer Z1 (Zeiss) microscope equipped with a confocal spinning disk (Andor Revolution Imaging System, Yokogawa CSU-X1), a Q-switched solid-state 355 nm UV-A laser (Powerchip, Teem Photonics), a C-APOCHROMAT 63×/1.2 W Korr UV-VIS-IR water immersion objective (Behrndt et al., 2012) and a home-made cooling stage were used. The membranes of tail epidermal cells of tailbud embryos were labeled with FM-64 (Thermo Fisher Scientific). Each ventral midline epidermal cell was cut along the apicobasal axis (5-10 µm lines each) by applying 25 UV pulses at 0.7 kHz. The embryos were imaged every 0.2 s with an exposure time of 150 ms. Single fluorescent images were used to measure tail relaxation 3 s post ablation, and the percentage of relaxation was calculated as the area of movement of the tail region 3 s after laser cutting.

The MOs (Gene Tools) against Msxb and Admp, which block translation, were designed according to previous studies (Imai et al., 2006; Waki et al., 2015) and are as follows: Admp, 5′-TATCGTGTAGT TTGCTTTCTATATA-3′; Msxb, 5′-ATTCGTTTACTGTCATTTTTAATTT-3′. These MOs were injected at 0.25-0.50 mM into an unfertilized egg and incubated until observation. To visualize the phenotype of Admp MO embryos at the single-cell level, embryos were stained using Alexa Fluor 546 phalloidin (1/50 diluted in PBS, A22283, Thermo Fisher Scientific) and imaged using an Olympus fv1000 microscope.

The DNA constructs used for overexpression of bmp2/4 under the Dlx.b upstream sequence (Ciinte.REG.KH.C7.630497–632996|Dlx.b) were used previously (Imai et al., 2012). These DNA constructs were introduced by electroporation.

Ciona intestinalis adults were provided by Dr Yutaka Satou (Kyoto University) and Dr Manabu Yoshida (the University of Tokyo) with support from the National Bio-Resource Project of AMED, Japan. We thank Dr Hidehiko Hashimoto and Dr Yuji Mizotani for technical information about 1P-myosin antibody staining. We thank Dr Kaoru Imai and Dr Yutaka Satou for valuable discussion about Admp and for the DNA construct of Bmp2/4 under the Dlx.b upstream sequence. We thank Ms Maki Kogure for constructing the FUSION360 of the intercalating epidermal cell.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200215.reviewer-comments.pdf.