## Introduction

During early embryogenesis, cells undergo dramatic changes in shape, size and polarity, which display a large cell-to-cell (spatial) variation. Recent studies have revealed that, in early embryogenesis characterised by rapid and synchronous cleavage divisions, the developing process is highly regulated by cellular mechanics through the cytoskeleton1,2,3,4,5,6,7,8,9,10. Furthermore, a close correlation between cell division and force during the subsequent early developmental process has been observed in the zebrafish epiboly11,12,13, epithelium of Xenopus laevis embryo14, Drosophila wing disc15,16 and the dorsal midline of the ascidian embryo17,18. However, little is known about how individual cells temporarily change their mechanical properties in early embryogenesis after the embryo has divided into two hemispheres such as animal and vegetal poles.

To address this issue, we investigated the spatiotemporal change in single-cell mechanical properties of the ascidian embryo that exhibits mosaic development19 using atomic force microscopy (AFM)20 (Fig. 1a) providing a non-destructive measurement to simultaneously determine the topography and stiffness of embryonic cells (‘Methods’). Because the cell lineages of the ascidian embryo have been determined and the embryonic cells are invariant in space during the early developmental stages21, the analysed individual cells in embryo samples were identified by comparing AFM topographic and optical microscopic images of samples (‘Methods’ and Fig. 1b). Thus, we obtained information about how individual cells in animal and vegetal hemispheres behave as a mechanical system by mapping the upper surface region of different ascidian embryos.

Here, we show how the stiffness of embryonic cells evolves during early embryogenesis from the fertilised egg to the stage before entering gastrulation (Fig. 1c). The fertilised egg undergoes symmetrical cleavage from the 1-cell to 16-cell stage. Then, the embryo begins to deform anisotropically where the subsequent cell divisions are no longer synchronised in animal and vegetal hemispheres21. In the animal hemisphere, the cell division occurs synchronously and a spherical-like shape is still maintained. On the other hand, in the vegetal hemisphere, the shape of the embryo becomes rather flat, and the cells divide asynchronously so that the apical constriction of endodermal cells begins to occur at around the 64-cell stage and initiates gastrulation at the late 112-cell stage22. Immunofluorescence revealed that the apical shrinkage in the vegetal hemisphere was relaxed and the apical and circum–apical accumulation of myosin was reduced as the embryo was treated with Rho kinase (ROCK) inhibitor Y-2763222. This observation indicated that a Rho pathway controlled the accumulation of myosin in the apical region of the vegetal hemisphere before gastrulation4,18,22.

We observed that the stiffness of embryonic cells increased at the onset of cell division and then decreased after the cell divided. Furthermore, we observed in the vegetal hemisphere that cell stiffening occurred even after finishing cell division or without cell division, displaying a characteristic cell-to-cell mechanical variation in the interphase. The Y-27632 treatment experiments indicated that the cell stiffening occurring in the interphase was strongly associated with myosin, which is regulated through the ROCK pathway, while the cell stiffening at the timing of cell division was associated with other pathways.

## Results

### Mechanical properties of embryonic cells measured by AFM

We first checked how well cell mechanical properties in an embryo were estimated by AFM force measurement with a tandemly arranged double-colloidal probe cantilever (Fig. 1a, ‘Methods’). The approaching force–distance curves measured in the upper regions of embryonic cell surfaces were well fitted to the conventional Hertzian contact model23 (Fig. 1d, ‘Methods’). The estimated apparent Young’s modulus E changed within several hundred nm in the indentation depth, but was almost constant in the further indentation (Supplementary Fig. 1). Thus, we used approaching force–distance curves obtained from the indentation of 1–2 μm to estimate the height H and E of an embryo, the latter of which was calibrated using the modified Hertzian contact model24 considering the tilt angle of the sample surface (‘Methods’). The AFM images acquired at 3 min/frame showed that the cell shapes in H and E images were identical, and the cell locations were in good agreement with the 3D virtual embryo model determined from confocal microscopic images (Fig. 1b). These results indicated that the E-value measured by AFM mainly reflected the mechanical properties of cells that directly contacted the AFM probe. We noticed that the magnitude of E-values was largely varied among native embryo samples even under the same condition, which indicated embryo-to-embryo mechanical variation in a chordate sample. The variation was also likely sensitive to the temperature of seawater from which the native chordate samples were collected and the culture day. However, because the origin of the magnitude variation of E has not been elucidated, we estimated the average E of embryonic cells and the significant difference of E between different states or conditions in this study.

### Cells stiffen at the timing of initial symmetrical cell division

To elucidate the spatiotemporal dynamics of E in the ascidian embryo during early cleavage, we first measured E of the fertilised egg during the initial symmetric development from the one-cell to the eight-cell stage (Fig. 2a–d). The fertilised egg is known to form cortical actin filaments before the cleavage division25,26. The AFM measurement showed that, during the cytokinesis of the first cell division from the one-cell to the two-cell stage (m1 in Fig. 2b, c), the cell began to increase E, which was more pronounced in the equatorial region when a furrow appeared, and then the blastomeres rapidly reduced E after completing cell division (Fig. 2b, #4 and #5). Furthermore, the same increasing behaviour in E was observed in subsequent cell divisions from the two-cell to four-cell stage (m2 in Fig. 2b, c) and from the four-cell to the eight-cell stage (m3 in Fig. 2b, c). The significant change in E (Fig. 2d) was reminiscent of the mechanical behaviours observed in single mitotic cells adhered on a substrate in vitro27,28 where the cortical actin filaments are formed during mitosis28,29.

We also observed that chemical treatment with an actin filament polymerisation inhibitor, Latruncurin A (LatA), led to a significant reduction of E that attained a constant value of <50 Pa, which was lower than that in native and unfertilised embryos, and no longer caused cell division (Fig. 2e–g). This result implies an intimate relationship between the change in E and the formation of actin filaments.

The average E was significantly increased to ~300 Pa at the timing of the cell division (Fig. 2d). The value was the same order of magnitude as the instantaneous elastic modulus of a Xenopus laevis embryo obtained by stress relaxation measurement, which was about tenfold lower than that in vitro single mitotic cells, which is typically around a few kPa27. During interphase, the average E was minimised to ~100 Pa (Fig. 2c, d), which was comparable with that of the unfertilised egg24, indicating that the E-value was the lowest stiffness of single cells in the native embryo.

### Animal hemisphere: epidermal cells stiffen at the timing of cell division

In the later stages from the 32-cell to the 112-cell stage, the significant increase in E at the timing of cell division appeared subsequently in the animal hemisphere (m1 and m2 in Fig. 3a–d). The temporal change in E was almost synchronised among epidermal cells at the timing of cell division (Fig. 3b). The time evolution of cell stiffness observed in the animal hemisphere (Fig. 3b–d) is summarised schematically in Fig. 3e.

Approaching cell division, epidermal cells began to increase E in the cell–cell boundary region (#1–4 in Supplementary Fig. 2a) and then stimulated furrow formation with a distinct increase of E, while the intracellular stiffness was increased gradually (#5 and #6 in Supplementary Fig. 2a). After completing the cell division, E decreased rapidly in the intracellular and cell–cell boundary regions (#7–9 in Supplementary Fig. 2a).

Single cells have a round shape during mitosis, which is called mitotic rounding28,30. During mitotic rounding, actin filaments accumulate in cortical regions and facilitate cell stiffening31. AFM images showed rounding of embryonic cells during mitosis as E increased (Supplementary Fig. 2b). Furthermore, the cell stiffening was significantly reduced by treatment with LatA in the animal hemisphere (Supplementary Fig. 3a–c) where the cell stiffening remained unchanged by treatment with DMSO (Supplementary Fig. 3d–f). Furthermore, using a live imaging technique (‘Methods’), we observed clear recruitment and release of actin filaments in cortical regions of the animal hemisphere at the timing of cell division (Fig. 4). The results indicated that the change in E observed in epidermis cells at the timing of cell division was associated with the formation and release of cortical actin filaments, the underlying mechanism of which is the same as that in single mitotic cells adhered on the substrate in vitro28,29.

We also found that treatment with ROCK inhibitor Y-27632 did not completely suppress the increase in E at the timing of cell division (Fig. 3f–h). At the same time, the chemical did not arrest the cell cycle, but the cell shape was different from that in the native embryo (m2 in Fig. 3f), which showed abnormal development as reported previously32,33.

### Cell cycle-arrested embryonic cells display a periodic mechanical change

The change in E that occurred at the timing of cell division was not necessarily synchronised with the cell division. This was found in an embryo sample treated with nocodazole (NOC) that inhibits the formation of microtubules and facilitates cell cycle arrest. In a cell cycle-arrested fertilised egg, we observed a slight but significant increase and decrease in E with a time interval of 30–40 min (Fig. 5a–c), which was still the same order as that in the native sample (Fig. 2b, c). A similar relationship between actin filaments and microtubules has been observed in cells in vitro, so that single cells treated with NOC still accumulated cortical actomyosin and increased rounding force34, and the actin filaments exhibited repeated polarisation35. Furthermore, we observed such a change in E of a NOC-treated embryo with inhibited cell division at later stages, such as the 32-cell stage (Fig. 5d–f). These results suggested that the remodelling of cortical actin filaments in embryonic cells, which contributed to the cell stiffness measured by AFM, was highly conserved during developing processes and less associated with the dynamics of microtubules that facilitate cytokinesis in cell division.

### Vegetal hemisphere: spatiotemporal heterogeneity of single-cell stiffening

We observed complex behaviour of E in the vegetal hemisphere (Fig. 6a–h and Supplementary Fig. 4) rather than in the animal hemisphere, showing that the mechanical properties of embryonic cells were independently regulated in different germ layers. First, at cell stages from the 32-cell to the 44-cell stage (#1–6 in Fig. 6b), we tracked a set of adjacent cells, such as A6.1 and B.6.1 cells, which divide into A7.1 and A7.2 cells, and B7.1 and B7.2 cells, respectively. The observation showed that both A6.1 and B6.1 cells began to increase E while approaching cell division (m1 in Fig. 6b–d), which was similar to the mechanical behaviours observed in the animal hemisphere (Fig. 3b–d). Importantly, after completing the cell division, the evolution of E was significantly different between A7.1 and B7.1 cells: in the A7.1 cell, the increased E was rapidly reduced, while in the B7.1 cell, the increased E was maintained until the later cell stages towards gastrulation (Fig. 6c, e).

Next, we tracked a set of adjacent cells, such as A7.3 (#5–12 in Supplementary Fig. 4), at the later cell stages from the 64-cell to the 76-cell stage in addition to A7.1, A7.2 and B7.1 cells (#7–12 in Fig. 6b). In the 64-cell stage, from which cell fate is determined (Fig. 6a)36,37, A7.3 cells adjacent to A7.2 cells began to increase E approaching cell division into A8.5 and A8.6 (pre-notochord) cells (Supplementary Fig. 4b–d), which was similar behaviour to B7.1 cells (Fig. 6f). At the same time of the stiffening of A7.3 cells, both A7.1 and A7.2 cells began to increase E (Supplementary Fig. 4b, c). Interestingly, the stiffening of A7.1 and A7.2 cells occurred in the subcellular region and not with accompanying cell division (#2 and #3 arrows in Fig. 6g), and the rate of cell stiffening was different between them (Fig. 6c, f and Supplementary Fig. 4c). Apparently, E was enhanced sequentially in the direction from posterior (B7.1 cells) to anterior (A7.1. and A7.2 cells) in the vegetal hemisphere (Fig. 6b, c and Supplementary Fig. 4).

When entering the 76-cell stage, the A7.2 cell was softer than the surrounding cells such as A7.1, A8.5-6 and A8.13–14 cells (#4 in Fig. 6g, f), although these cells attained stiffness higher than ~100 Pa in the interphase of the 44-cell stage (Fig. 6f). In addition, as shown in Fig. 6f, we found that A7.6 cells (trunk lateral cells) retained an almost unchanged E (#11–12 in Fig. 6b), which were significantly softer than pre-notochord cells such as A8.5-6 and A8.13–14 cells (#12 in Fig. 6b) and pre-endoderm cells such as A7.5 (#11 and #12 in Fig. 6b and #4 in Figs. 6g) and B7.1 (#5 and #6 in Fig. 6b) cells. It was also observed that trunk lateral cells were softer than pre-mesenchyme such as B8.5 cells (#11 and #12 in Fig. 6b and #4 in Fig. 6g). Furthermore, A8.7-8 cells (pre-neuron cells) (#11 and #12 in Fig. 6b and #4 in Fig. 6g) were observed to retain an unchanged E that was similar to trunk lateral cells. These results indicated that the change in E was different among cells with different cell fates before cell differentiation. The time evolution of cell stiffness observed in the vegetal hemisphere (Fig. 6b–g) is summarised schematically in Fig. 6h.

We found that E maintained at a higher value in B7.1 cell was significantly suppressed by treatment with Y-27632 (Fig. 6i–k), which was different from the behaviours observed in the change of E at the timing of cell division (Fig. 3f–h). In addition, the Y-27632 treatment suppressed stiffening of A7.1 and A7.2 cells occurring without cell division at the 76-cell stage (Fig. 6i–k). These results indicated that the cell stiffening occurring without cell division in the interphase was strongly associated with myosin, which is regulated through the ROCK pathway, while the cell stiffening at the timing of cell division was associated with other pathways.

## Discussion

During mitosis in fertilised eggs, actin filaments accumulate in the cortical region and contraction ring25,26. Our AFM analyses showed that the fertilised egg exhibited a remarkable increase in E, which was more pronounced in the equatorial region when a furrow appeared (#1–5 in Fig. 2b) similarly to that during the division of cells adhered on a substrate in vitro27. Similar cell stiffening at the timings of cell division was observed in subsequent stages from the two-cell to eight-cell stage (#6–12 in Fig. 2b) and the stages until the 112-cell stage before entering gastrulation in both animal (Fig. 3b, c) and vegetal (Fig. 4b, c) hemispheres. Furthermore, we observed that depolymerisation of actin filaments led to a significant reduction in E (Fig. 2e–g and Supplementary Fig. 3a–c) and that clear recruitment and release of actin filaments in cortical regions of the animal hemisphere at the timing of cell division (Fig. 4). Based on the results, we concluded that AFM force mapping measurement can mechanically monitor the remodelling of cortical actin filaments in embryonic cells during early embryogenesis.

The total stress operating in the developing embryo includes ‘active’ stress defined as contractile traction force or tension exerted in cells and between adjacent cells and ‘passive’ stress as the intracellular stiffness38. Previous studies have revealed that hydrodynamic pressure and tension due to the actomyosin cortex lead to the mitotic cell rounding of single cells28,39 and a confined cell system34. In stages that embryonic cells were tightly confined as adjacent cells, we observed that the mitotic cell rounding appeared as E increased (Supplementary Fig. 2). Therefore, it is reasonable to consider that the indentation force measured by AFM involves tensile force and hydrodynamic pressure in addition to force resulting from the elastic modulus of cells. Assuming that T is measured predominantly by AFM at a shallow indentation depth with an indentation force of less than 200 pN, we found that T was correlated positively with E (Supplementary Fig. 5). This correlation between E and T does not contradict our consideration that the observed change in E of mitotic cells involves the change in the tensile forces and hydrodynamic pressure reported in the previous studies28,34,39,40. Therefore, E estimated in this study is regarded as a measure of total stress operating in the developing embryo.

The dynamic change of E in embryonic cells during early development from the fertilised egg to the 112-cell stage before entering gastrulation can be classified into two types. The first is the ROCK-independent cell stiffening that E of cells increases at the onset of cell division and then decreases after finishing the cell division (Figs. 2b–d,  3b–d, and  6b–d). This was commonly observed in both animal and vegetal hemispheres as well as in symmetrical developing stages. The second type is the ROCK-associated cell stiffening, which was observed only in the vegetal hemisphere (Figs. 6b–h and  7): The increased E of cells did not decrease even after finishing the cell division and was consequently maintained in the interphase as shown in the division of B6.1 (Fig. 6b, c) and A7.3 cells (Supplementary Fig. 4b, c), while E of cells increased without cell division and was consequently maintained in the interphase as shown in A7.1 and A7.2 cells (Fig. 6b, c).

In this context, previous studies have shown that Y-27632 treatment strongly inhibits apical and circum–apical myosin, but not basolateral myosin where myosin that facilitates the apical construction22 began to accumulate in apical regions from around the 64-cell stage and then in the basolateral region (Fig. 7). These results suggest that the change of E in the second type observed in the vegetal hemisphere reflects the apical formation of actomyosin, which is highly regulated in subcellular regions in the direction from posterior (B7.1 cell) to anterior (A7.1. and A7.2 cells) (Fig. 6b, c, g, h and Supplementary Fig. 4).

We noticed that no apparent cell-to-cell variation in myosin density in apical regions has been observed in previous studies by immunofluorescence22, implying that the apical actomyosin network is remodelled without changing the myosin density and facilitates a mechanically stabilised structure. Thus, we consider that AFM provides a complementary approach to quantitative investigations of dynamic structural changes in embryonic cells during the developmental process.

Previous studies have reported that the number distribution of cell mechanical properties exhibits a large variation in single cells41,42 and the cell stiffness and tension are very broadly varied over a long intercellular distance in a cell monolayer43,44,45,46. In contrast, the embryonic cells in the vegetal hemisphere exhibited broad spatiotemporal heterogeneity occurring at the single-cell level. From the 64-cell to 76-cell stage, at the timing that A7.2 cells (endodermal cells) began to increase E, the adjacent cells such as A8.5-6 and A8.13–14 cells (pre-notochord cells) also increased E (Fig. 6a–h). It is known that, at the 64-cell stage, notochord precursor cells begin to express a specific gene, so-called brachyury, which is a key transcription factor for notochord differentiation in ascidians47. Our results implied that the mechanical properties of embryonic cells may be associated with gene expression, as reported in single cells48.

### Conclusions

Our AFM measurements of the ascidian embryo showed single-cell stiffening and cell rounding at the timing of cell division, which have been commonly observed in single mitotic cells in vitro and monolayer cultures on a substrate. Furthermore, at stages from the 32-cell to the 112-cell stage before entering gastrulation, we found characteristic cell stiffening and softening in the vegetal hemisphere, which were different from those in the animal hemisphere. Treatment with Y-27632 showed that the single-cell stiffening without cell division observed only in the vegetal hemisphere was associated with myosin that is regulated through the ROCK pathway, while the single-cell stiffening at the timing of cell division in both hemispheres was mediated via other pathways. Taken together with the previous immunofluorescence observations indicating that apical and circum–apical myosin were specifically inhibited by Y-27632, but no apparent spatial variation in myosin density was observed, we concluded that the spatiotemporal change in E of the vegetal hemisphere reflected the formation and remodelling of the apical actomyosin network. The direct AFM measurements of cell stiffness in the ascidian embryo, a model of early chordate development, can facilitate a better understanding of how embryonic single cells in animals are spatiotemporally regulated via mechanical cues.

## Methods

### Ascidian samples

Adult Ciona intestinalis type A (Ciona robusta) was collected in the vicinity of the Maizuru Fisheries Research Station (Kyoto University) and Misaki Marine Biological Station (The University of Tokyo) through the National BioResource Project (NBRP). Eggs and sperm were obtained by dissection of the ascidian gonoducts. Unfertilised eggs were dechorionated by natural seawater (Nihon Aquarium) containing 1.0% sodium thioglycolate (Wako Pure Chemical Industries Ltd.) and 0.05% actinase E (Rikaken Pharmaceutical Co. Ltd.). After washing the dechorionated eggs in seawater several times, they were artificially inseminated and reared on 0.1% gelatin (Wako Pure Chemical Industries Ltd.)-coated dishes in seawater at 18 °C for 3 hpf (hour post-fertilised). For AFM measurements, the embryos were weakly attached to a cell culture dish (Iwaki) in seawater to prevent unexpected translational movement during the experiments. During the preparation, the embryos occasionally failed to undergo normal development during the early developmental stage. Therefore, we showed experimental data of embryos developed until the gastrula stage.

### AFM measurements

A customised atomic force microscope46 was used to map the relative height and apparent Young’s modulus E of ascidian embryo samples. The atomic force microscope was mounted on an upright optical microscope (Eclipse FN1, Nikon) with a liquid-immersion objective lens (plan fluor, ×10, Nikon) for the optical lever system, as reported previously46. We used a rectangular cantilever (Biolever-mini, BL-AC40TS-C2, Olympus) with a nominal spring constant of <0.1 N/m. Before the experiments, the spring constant of the cantilever was determined in liquid environments using a thermal fluctuation procedure built in a commercial atomic force microscope (MFP-3D, Asylum Research). To achieve a well-defined contact geometry and prevent the contact between the cantilever beam and the sample surface, two silica beads with a radius R of ca. 5 μm (Funakoshi) were arranged tandemly from the AFM tip with epoxy glue (Nichiban).

The AFM force mapping measurements were conducted in a part of the upper region of the embryo at 18 °C with a piezoscanner (E-761, Physik Instruments) controlled with LabVIEW software (National Instruments), where the embryo samples were weakly adhered on a dish so that abnormal cell division was highly prevented. However, the weak adhesion occasionally caused small unexpected fluctuations. Thus, the mapped positions by AFM were changed manually to trace the approximately same region of the embryo in the developing process according to the sample fluctuation. The scan range was about 72 × 72 μm with a spacing of 3 μm. The acquisition time for a single mapping image was ~3 min with a slow scan speed, which reduced the unexpected movement of embryo samples. The maximum loading force was set at 0.6 nN. The cells mapped by AFM were identified using the Four-Dimensional Ascidian Body Atlas database21.

### Estimate of the apparent Young’s modulus and tension

The E of embryonic cells was estimated from approaching force–distance curves using the modified Hertz contact model24. According to the conventional Hertz contact model23, as a spherical probe with a radius of R indents perpendicular to the flat surface of a cell with E with an indentation depth of δ, the loading force F is given by

$$F = \frac{{4R^{1/2}E}}{{3\left( {1 - \nu _{\mathrm{p}}^2} \right)}}\delta ^{3/2},$$
(1)

where νp is the Poisson’s ratio of the cell, which was assumed to be 0.5. Because the surface of the embryo samples was not flat but tilted, F changed depending on the tilt angle θ of the sample surface that can be estimated from an AFM topographic image. In this case, E for soft materials with <~10 kPa was calibrated with a modified Hertz contact model 24:

$${\it{F}} = \frac{{4R^{1/2}\left( {E\cos ^{5/2}\theta } \right)}}{{3\left( {1 - \nu _{\mathrm{p}}^2} \right)}}\delta ^{3/2}.$$
(2)

Because the modified Hertz contact model was valid for θ <40°24, we discussed the phenomena observed in mapping regions that satisfied θ <40°. Images of E were scaled with ν = log10E (Pa).

According to the liquid droplet model49,50, the cell–cortex tension T was estimated by the following equation:

$$F = 2T\left( {\frac{1}{R} + \frac{1}{r}} \right)\left( {2\pi R} \right)\delta$$
(3)

where r is the cell radius, which was assumed to be 15 μm. To estimate the apparent T of embryonic cells, we fitted with Eq. (3) approaching force–distance curves of <200 pN, typically corresponding to an indentation of less than 1.0 μm, which was one order of magnitude smaller than the cell diameter.

### Chemical treatments

To inhibit actin filament polymerisation, the embryos were treated with 0.1–0.2 μM latrunculin-A (LatA, Sigma-Aldrich) during time-lapse AFM mapping measurements. To inhibit Rho kinase (ROCK) and thereby suppress myosin II phosphorylation, the embryos were treated with 100 μM Y-27632 (Sigma-Aldrich) during time-lapse AFM mapping measurements. To inhibit the formation of microtubules and thereby halt cell division, the embryos were treated with 5 μM nocodazole (NOC, Sigma-Aldrich) during time-lapse AFM mapping measurements. To quantify the chemical dependence of E, we analysed the treated cells after incubation for at least 10 min with latA and 20 min with Y-27632.

### Live imaging of cytoskeletons

For cytoskeleton imaging, lifeact-GFP was gifted by Dr. Benoit Godard of IST Austria. Lifeact-GFP is an actin marker for the visualisation of actin filaments (F-actin)51. The mRNA was synthesised by using mMESSAGE mMACHINE Kit (Ambion). According to the protocol provided by the company, reaction mixture (2× NTP/CAP 10 µL, 10× reaction buffer 2 µL, enzyme mix 2 µL, linearised DNA 1 µg and nuclease-free water to 20 µL) was prepared in 1.5-mL tube. The reaction mixture was incubated at 37 °C for 2 h. TURBO DNase was added and incubated at 37 °C for 15 min. Nuclease-free water and LiCl precipitation solution were added and chilled for 30 min at –20 °C. The reaction mixture was centrifuged at 4 °C for 15 min at maximum speed to pellet the RNA. The pellet was washed twice in 70% ethanol and centrifuged at maximum speed for 5 min. After removing 70% ethanol, nuclease-free water was added and stored at –20 °C.

The mRNA-encoding lifeact-GFP was injected into unfertilised eggs. After fertilisation, embryos were incubated for 3 h at 20 °C and mounted on a glass-based dish (IWAKI). The embryos were immersed into seawater with FM4–64 to stain cell membrane. Time-lapse 3D imaging was performed using a confocal laser scanning microscope (fv1000, Olympus) equipped with a ×40 oil-immersion lens. During the observation, the temperature was kept using a temperature controller. Step size of the z axis was set at 2 µm, and 20–30 image confocal stacks were acquired per sample. The time interval was 3 min. These images were reconstructed to 3D image by software (Imaris, bitplane).

### Fluorescence image analysis

Fluorescence images of F-actin (lifeact-GFP) and cell membrane (FM4–64) were loaded and converted to greyscale images. Cell membrane region was manually determined with a threshold using membrane fluorescence images according to the Otsu method. Then, the intensity ratios of F-actin (lifeact-GFP) in the cell membrane region (corresponding to cortical F-actin) and in the cytoplasmic region (corresponding to cytoplasmic F-actin) to the cell membrane (FM4–64) were calculated during the developmental process.

### Statistics and reproducibility

For data analysis, an unpaired two-tailed Student’s t test was used to determine P values. The P values <0.05 were considered to be significant. The number of individual experiments is in the corresponding figure legend.

### Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.