As previously reported, cells of the lateral epidermis elongate toward the AS during DC, especially at the leading edge (Figure 1B and Supplementary Video S1) (Kiehart et al., 2000; Jacinto et al., 2002b). The kinetics of cell elongation has not been thoroughly quantified. Based on time-lapse recordings of embryos expressing E-cadherin-GFP (E-CadGFP), we measured the displacement and elongation of epidermal cells (three rows of cells including leading-edge cells and dorsal-most epidermal cells) (Figure 1C) over about 90 min prior and during DC (Figure 1C and Supplementary Video S2). The measurements were aligned to the onset of DC (t = 0), as defined by the noticeable displacement of the leading edge dorsally. The leading edge moved dorsally slightly with approximately 2 μm in 30 min before the onset of DC and more than 25 μm in 60 min during DC (Figure 1D). We obtained an approximately linear displacement over time (Figure 1D). We calculated the rate of displacements and obtained an increasing rate, especially in the first 30 min of DC (Figure 1E). The length of epidermal cells increased almost linearly (Figure 1F). On average, the cell length extended by 1.3-fold in 60 min during DC (Figure 1F). Plotting the elongation speed revealed pulsatile kinetics (Figures 1F,G).

Previous studies have shown that the contraction of AS cells is required for DC (Solon et al., 2009; Pasakarnis et al., 2016). Here, we asked whether the increased contractility of a single cell would suffice for increased pulling by the AS. We also aimed to estimate the reach of such a pulling force, given the viscoelastic nature of the tissue. We previously described an optochemical approach that controls epithelial cell contractility via Ca²⁺-mediated light activation of non-muscle myosin II (CaLM) on the scale of seconds and at single-cell resolution during tissue morphogenesis (Kong et al., 2019).

We applied CaLM and targeted AS cells at elliptical stage embryos expressing E-CadGFP and inspected the movement of the leading edge (Figure 2 and Supplementary Figure S1). Applying CaLM to a single AS cell induced the rapid cell contraction to half of the cross-section area but no obvious displacement of the leading edge (Supplementary Figure S1). In contrast, when we simultaneously applied CaLM in three neighboring AS cells (Figures 2B,C, Supplementary Video S3), we measured a displacement of about 5 μm dorsally in 5 min, significantly higher than in control (Figures 2B,D). Control displacement was measured from the same embryos but away from the contracting AS cells.

We employed a Vinculin biosensor to assay the force distribution next to the contracting cells (Kong et al., 2019) and obtained an estimation for the reach of the pulling force by the contracting cell. This biosensor is based on the force-dependent conformational change of alpha-Catenin (alpha-Cat), a component of adherence junctions and binding partner of E-Cad (Choi et al., 2012; Rangarajan and Izard, 2012; Yao et al., 2014; Duong et al., 2021). In the closed state, the central mechanosensitive modulatory (M) domain of alpha-Cat is inaccessible to the D1 domain of Vinculin. In contrast, the stretching of alpha-Cat induces a conformational change, which exposes the M domain and allows binding of the D1 domain. alpha-Cat bridges the cadherin complex with the actin cytoskeleton and can thus sense the mechanical load between them. We previously reported that the Vinculin biosensor protein was recruited to cell junctions of target cells in response to CaLM-induced contractility (Kong et al., 2019). Following CaLM in individual AS cells, we measured changes of VincD1-GFP fluorescence normalized by E-CadmTom (Figure 3). Our measurements revealed an increasing ratiomeric signal at the junction next to but not part of the target cell. Next-next and further distant junctions showed less or no difference in the normalized reporter signal. Ninety seconds after uncaging, we detected a 30% increase at junctions of the target cell, 20% at first junctions and 10% at second neighbor junctions (Figure 3C). In more distant junctions, the reporter signal remained relatively stable and was not different from the reporter signal as distant junctions. In summary, our experiments showed that the optochemically controlled contractility generates a pulling force on the surrounding cells. Consistent with previous reports, we showed that acute pulling forces from the region of the AS next to the epidermis could lead to a displacement of the leading edge.

Having analyzed the AS tissue, we asked whether increased contractility in the lateral epidermis would counteract the pulling by the AS cells. As a preliminary test for a contribution of the epidermal cells to the force balance between AS and epidermis, we increased cell contractility by the expression of constitutively active RhoA (Rho^V14) (Lee et al., 2000) in four- to five-cell-wide epithelial strips with an engrailed-GAL4 driver (Supplementary Figure S2). In the case of a passive contribution and cell elongation as a mere consequence of DC, we would expect no delay of DC. Alternatively, an active contribution would be visible by a delay in DC. Rho^V14-expressing cells but not in control embryos (Supplementary Figures S2A and B) showed an isotropic cell shape compared to the engrailed negative cells in those embryos (Greig and Bulgakova, 2020). The inhibition of cell elongation indicates that contraction was induced. By the end of embryonic development, we observed the dorsal hole in the cuticles of Rho^V14-expressing embryos, which may indicate potential DC defects.

To better test the hypothesis that lateral epidermal cells actively contribute to DC, we turned to an acute interference instead of the chronic expression of activated RhoA. We applied CaLM to lateral epidermal cells in three rows of cells including the leading-edge cells and avoided the AS-epidermal interface. As a readout, we measured the movement of the leading edge (Figure 4A). We first induced contractions in a region 6–8 cells wide (region of interesting, white dashed boxes in Figure 4B), avoiding the actin cable (Figure 4B). Strikingly, we observed that the leading edge of the target region failed to move dorsally, whereas the distant regions moved as normal (Figures 4B,C). We measured the leading-edge displacements nearby and CaLM application region in three individual embryos after CaLM application. The leading-edge displacements far away from the CaLM application region in the same embryo were measured as controls. The leading-edge movements were blocked, and in some cases, it even moved toward the ventral, and on the contrary, the leading edge in the same embryos far away from the CaLM application region remained to move dorsally (Figure 4C). Scanning the complete region of interest with the UV laser, we bleached, unfortunately, the GFP signal labeling the cell outlines. To follow the shapes of epidermal cells, we applied CaLM with higher-magnification objectives but a smaller field of view. In this way, we spot-illuminated the target cells one by one (Figure 4D). As a consequence of the induced contractility, cell elongation was explicitly inhibited in the target cells (Figure 4D, Supplementary Video S4). The leading edge close to the target cells moved even slightly ventrally for 2.5 min and was blocked there for a couple of minutes and finally started to move dorsally again (white arrowheads in Figures 4D,E). In contrast with control (leading edge in the same embryos far away from the CaLM application region), the leading-edge movements were significantly blocked and even to the opposite direction for 2.5 min due to CaLM-induced contractions in the lateral epidermal cells (Figure 4E). Our data show that optochemically induced contractions in the lateral epidermal cells temporarily and locally antagonized DC. Together with the genetic Rho^V14 expression data (Supplementary Figure S2), our results suggest that DC requires elongation in epidermal cells.

We employed microsurgical tissue cuts to assay the mechanical tension in the lateral epidermis. Ablation studies suggest that the lateral epidermis is under tension during DC (Kiehart et al., 2000). However, it remains unclear whether the transition in tissue tension occurs in the lateral epidermis. Tension is determined by the ratio of velocity and phenomenological friction. As friction is difficult to measure, we relied on the initial recoil velocity after the cut as a direct measure for tension. We first performed a tissue scale cut parallel to the leading edge in the lateral epidermis (Figure 5A). With particle image velocimetry (PIV), we calculated the initial recoil velocities in both directions toward and away from the AS (Figures 5B,C, Supplementary Video S5). Our measurements revealed larger initial recoil velocities at the DC stage than at the elliptical stage, which was consistent with cell elongation during DC. Our data revealed the developmental upregulation of tissue tension in the lateral epidermis.

We compared the initial recoil velocities after ablation in different directions, toward the AS and lateral epidermis, in each stage. The initial recoil velocities toward the AS are greater than toward the lateral epidermis during the DC stage (Figure 5C). Our data agreed with the previous studies that AS contractions generate pulling forces to the lateral epidermis during DC (Figure 1A) (Solon et al., 2009; Saias et al., 2015). Interestingly, the initial recoil velocities away from the AS were significantly greater than toward the AS during the elliptical stage (Figure 5C). It did not agree with the previous biophysical model in which the tissue tension in AS and lateral epidermis is balanced with each other initially, and a temporal increase in AS surface tension could drive closure (Saias et al., 2015). During the elliptical stage, the pulling force of AS is balanced by the later epidermis since the interface does not move. The force balance changes between the lateral epidermis and AS at the transition from the elliptical stage to DC. The movement during DC indicates that changes in the cell biology of AS lead to an imbalance with the AS taking over or the lateral epidermis giving in (Figure 1A). Our data suggest that the tissue tension in AS and lateral epidermis is not balanced with each other at the elliptical stage.

Next, we measured the tension on the cellular scale. We cut individual junctions of epidermal cells at the leading edge prior to and during DC (Figures 5D,E). Surprisingly, the initial recoil velocities significantly decreased by almost 50% from the elliptical stage to DC, in contrast, to expect given the increased tension on the tissue scale and elongated cells (Figures 5E–G). Our data show that the cellular material properties vary in the lateral epidermis during morphogenesis. The opposite outcome between the tissue and individual junction levels may be due to the experimental procedure. The laser is focused on the cell border for junction cuts, including the cortex with cortical actomyosin (Figure 5D). In contrast, the complete cell, including the cytoplasm and the cytoskeleton, is ablated for tissue cut (Figure 5A).

We searched for any developmental changes in the cytoskeleton between the elliptical and DC stages. Actomyosin was our first candidate as it is the main regulator for elliptical cell contraction and shape (Koenderink and Paluch, 2018). We examined the levels of active non-muscle myosin II by the fluorescence intensity of an mCherry-tagged spaghetti squash (sqh), the regulatory light chain of the non-muscle myosin II (MRLC) in Drosophila, during elliptical and DC stages (Figure 6). We detected increased MRLC-mCherry intensities in AS cells in the DC stage compared to those in the elliptical stage. We also detected a corresponding increase of MRLC-mCherry in the lateral epidermal cells (Figure 6). The averaged, MRLC-mCherry intensities were slightly higher in the AS, and the relative upregulation of ≈1.4-fold in AS and 1.2-fold in the epidermis was significantly different (Figure 6C). Using a phospho-specific myosin antibody (P-Myo) (Losick and Duhaime, 2021), we confirmed the developmental upregulation of active myosin both in AS and in the lateral epidermal cells (Supplementary Figure S3). P-Myo staining was comparable in AS and the lateral epidermal cells at the elliptical stage (Supplementary Figure S3B). Notably, P-Myo staining in the lateral epidermal cells was more intense than in AS cells from DC stage embryos (Supplementary Figure S3B). The average mean intensities of P-Myo staining were upregulated 1.2-fold in AS and 1.7-fold in the lateral epidermis from the elliptical to the DC stage (Supplementary Figure S3C). Consistent with our tissue ablation experiments, our data exclude that the epidermal cell shape change is a simple relaxing process. Furthermore, the opposite results from junctional ablation suggest the transition of other factors from the elliptical stage to AS.

The microtubule cytoskeleton is essential for regulating the cell shape (Dogterom and Koenderink, 2019) and for the fusion of the left and right sheets of the lateral epidermis in the late DC (Takács et al., 2017). We detected a specific upregulation of microtubules in the lateral epidermis by time-lapse imaging and by staining of fixed embryos (Figures 7A,B, Supplementary Video S6). We quantified the intensities of alpha-Tubulin-GFP during the course of DC in both epidermis and AS, revealing an increasing ratio between the epidermis and AS signals over the course of more than an hour (Figure 7C). We measured a corresponding increased staining in the epidermis in fixed embryos (Figures 7D,E). Whereas alpha-Tubulin staining in the AS was similar in the elliptical to the DC stage (factor of 1.1), we detected a significant 1.6-fold increase in the epidermis (Figures 7B,D,E).

Next, we employed a high-resolution imaging with an AiryScan detector to obtain an insight into the organization of microtubules (Figures 7F,G). Whereas microtubules were apparently oriented in the lateral direction during the elliptical stage in both epidermis and AS, the pattern striking changed in AS. During DC, the microtubules appeared isotropic as no preferred orientation was obvious. In contrast, the lateral orientation remained unchanged in the epidermal cells (Figures 7F,G, Supplementary Videos S7 and S8). With these high-resolution images, we quantified staining intensity at the tissue interphase by line profiles. We detected a sharp change of intensity at the tissue interphase during DC but only a slight difference during the elliptical stage (Figure 7H), consistent with the previous quantification.

In summary, we detected a developmental upregulation in microtubules specific for the epidermis. Large-volume electron tomography previously revealed that microtubules orient the protrusions of epidermal cells (Eltsov et al., 2015). The direct interaction of microtubules with F-actin bundles facilitates an initial contact during zippering between epidermal cells from the left and right tissue sheets. Expanding this previous finding, our data suggest a role of microtubules already during the early DC.

The following transgenes were used: E-Cadherin-GFP (Huang et al., 2009), E-Cadherin-mTomato (Huang et al., 2009), MRLC-mCherry (Martin et al., 2009), and UAS-Vinculin D1-GFP (Kong et al., 2019). The following transgenes are from the Bloomington Drosophila Stock Centre (Whitworth, 2019), and a Flybase Symbol (Thurmond et al., 2019) for each transgene is provided here. engrailed-GAL4 (Bloomington number 30564, Flybase Symbol, Dmel\P{en2.4-GAL4}e16E), AS-Gal4 (Bloomington number 3734, Flybase Symbol, Dmel\P{GawB}c381), alpha-Tubulin-GFP (Bloomington number 7253, Flybase Symbol, Dmel\P{GAL4:VP16-nos.UTR}CG6325^MVD1), and UAS- Rho^V14 (Bloomington number 7330, Flybase Symbol, Dmel\P{UAS-Rho1.V14}5.1) were used.

First instar larvae were dechorionated with bleach, washed with water, mounted in a 1:1 (v/v) mixture of Hoyer’s medium (Hoyer, 1882) and lactic acid, and incubated at 65°C overnight. Photographs were taken with the darkfield illumination.

Embryos were fixed, stained, and mounted as previously described (Zhang et al., 2014). Embryos were fixed by 37% formaldehyde for alpha-Tubulin staining. Antibodies against the following antigens were used: engrailed (mouse anti-Engrailed, 4.7 μg/mL, Hybridoma center 4D9), DCAD2 (rat anti-E-Cadherin, 7 μg/mL, Hybridoma center) (Oda et al., 1994), P-Myo (Rabbit anti-Phospho-myosin, 1:200, Cell Signaling), and alpha-Tubulin (mouse anti-alpha-Tubulin, 8 μg/L, Sigma B512). Secondary antibodies were labeled with Alexa dyes (Invitrogen, 5 μg/mL). GFP was stained by a nanobody labeled with ATTO 488 dye (GFP-Booster, chromotek). DNA was stained by DAPI (Sigma). Specimens were mounted in Aqua polymount (Polysciences).

Fluorescent images from fixed embryos were acquired on a laser scanning confocal microscopy (Carl Zeiss, ZEISS LSM 980 with Airyscan 2). Images in Figure 6 and Supplementary Figure S3 were acquired with a ×63 oil objective (Carl Zeiss, ×63/oil, NA1.4) with 0.5 μm intervals for Z-stacks. Images in Figure 7B and S2 were acquired with a ×40 water objective (Carl Zeiss, ×40/W, NA1.2) with 0.5 μm breaks for Z-stacks. For super-resolution imaging (Figures 7F,G and Supplementary Videos S7, S8), images were acquired with a ×63 oil objective (Carl Zeiss, ×63/oil, NA1.4, Airyscan model) with 0.17 μm interval for Z-stacks and processed by a ZEISS Airyscan Joint Deconvolution (Carl Zeiss). Image stacks were processed with “max intensity” in ImageJ/Fiji (Schindelin et al., 2012). 3D reconstruction was carried out by Imaris9 (Supplementary Videos S7, S8).

Embryos were prepared as described previously (Kong et al., 2019). Briefly, the staged embryos were collected and dechorionated with 50% hypochlorite bleach for 90 s, aligned on an agar block, and attached on the coverslips by homemade glue covered with halocarbon oil. Cross-sectional images were recorded from the dorsal side for E-CadGFP (488 nm excitation) on a spinning disc microscope with a 40x or ×100 oil objective (Carl Zeiss, ×40/oil, NA1.2, 100x/oil, NA1.4) with an emCCD camera (Carl Zeiss, AxioCam MRm). The apical plane of the embryo was acquired with axial sections of each 0.5 µm and a frame rate of 0.2/min or 1/min. Images in Figure 6A were acquired on a laser scanning confocal microscopy (Carl Zeiss, ZEISS LSM 980 with Airyscan 2) with a ×63 oil objective (Carl Zeiss, ×63/oil, NA1.4), and the apical plane of the embryo was acquired with axial sections of each 0.5 µm. Images in Figure 7A were acquired on a laser scanning confocal microscopy (Carl Zeiss, ZEISS LSM 980 with Airyscan 2) with a ×63 oil objective (Carl Zeiss, ×63/oil, NA1.4), and the apical plane of the embryo was acquired with axial sections of each 1 µm and a frame rate of 0.2/min.

Embryos were prepared as described previously (Kong et al., 2019). Briefly, embryos were collected and dechorionated with 50% hypochlorite bleach for 90 s, dried in a desiccation chamber for ∼10 min, covered with halocarbon oil, and injected dorsally into the vitelline space in the dark at room temperature (∼22°C). After injection, the embryos were incubated at room temperature in the dark for about 10 min before uncaging. Nitrophenol-EGTA, AM (NP-EGTA ester, Invitrogen) was prepared in a 1x injection solution (180 mM NaCl, 10 mM HEPES [pH 7.2], 5 mM KCl, 1 mM MgCl₂); 1 mM NP-EGTA, AM was injected for Ca²⁺ uncaging in AS cells (Figures 2, 3, Supplementary Figure S2, and Supplementary Video S3), and 2 mM NP-EGTA, AM was injected for Ca²⁺ uncaging in epidermal cells (Figure 4 and Supplementary Video S4).

A 375 nm diode laser (DPSL-375, up to 200 mW, Rapp OptoElectronic) was employed for Ca²⁺ uncaging. The laser was mounted on the epiport of a spinning disc microscope (Zeiss). The laser was applied on the “REO-SysCon-Zen” platform (Rapp OptoElectronic), while the images were recorded on a spinning disc microscope with a ×100 oil objective (Carl Zeiss, ×100/oil, NA1.4) with a CCD camera. An emCCD camera (Photometrics, Evolve 512) was used in Supplementary Figure S3.

For cell dynamics, cross-sectional images were recorded in the GFP channel (488 nm excitation) with a frame rate of 0.2/s (one frame every 5 s), immediately following uncaging. Flies carried the UASt-VincD1GFP; E-cad-mTomato transgenes were crossed with the AS-Gal4 driver flies. Embryos from this cross expressing VincD1GFP in AS cells were employed for uncaging. Cross-sectional images were recorded from the dorsal side in the GFP channel (488 nm excitation) with a CCD camera at a frame rate of 1/s, immediately following uncaging. The recording was switched to the emCCD camera with a 488 nm excitation for VincD1GFP and a 561 nm excitation for E-Cad-mTomato. The GFP and mTomato channels were recorded simultaneously at a frame rate of 0.1/s. The apical plane of the embryo was acquired with four axial sections of each 0.5 µm.

Stage 13 or 14 embryos expressing E-Cad-GFP were prepared as live imaging. Cross-section images were recorded in the GFP channel with a frame rate of 1/s on a spinning disc microscope (Zeiss, 100x/oil, NA1.4) with a CCD camera. A 355 nm pulsed laser (DPSL355/14, 355 nm, 70 µJ/pulse, Rapp OptoElectronic) was employed for ablation and manipulated on the “REO-SysCon-Zen” platform (Rapp OptoElectronic). The laser was mounted on the epiport of a spinning disc microscope.

Tissue laser ablation was applied with a 30 μm line along the axis in the lateral epidermis with 5% of laser power during the recording mode. Particle image velocimetry (PIV) of E-CadGFP images after tissue laser ablation (Figures 5A–C and Supplementary Video S5) analysis was performed using square interrogation windows of side 64 pixels by the “PIVlab” (Thielicke and Sonntag, 2021) in MATLAB.

Junction ablation was performed with 5% of laser power and a 200 ms (around 40 pulses) exposure time during the recording mode (100x oil, NA 1.4) (Figures 5D–G). The displacement of both ends (L(t)) of ablated junctions was measured manually in Fiji/ImageJ. The displacement (Figure 5F) was fitted as a Kelvin–Voigt fiber model (Fernandez-Gonzalez et al., 2009) to the Eq. 1 on Prism8 (Liang et al., 2016).

ε ( t ) = L ( t ) − L ( 0 ) = F 0 E . ( 1 − e − [ ( E μ ) *t ] ) (1) where

F₀ is the tensile force present at the junction before ablation, E is the elasticity of the junction, and μ is the viscosity coefficient related to the viscous drag of the cell cytoplasm.

Extraction of initial recoil values (Figure 5G) as fitting parameters for the Eq. 2 was introduced.

As fitting parameters for the above equation was introduced i n i t i a l   r e c o i l = d ε ( 0 ) d t = F 0 μ (2)

The image stacks were processed with “max intensity” in ImageJ/Fiji for the leading-edge movement and cell elongation analyses (Figures 1, 2, 4 and Supplementary Figure S1). The leading edge was measured manually in ImageJ/Fiji every frame. Including leading-edge cells, three rows of the lateral epidermal cells from dorsal to ventral in each embryo were analyzed. Five cells from each row were selected randomly for measurements. The cell length was measured along the dorsal–ventral direction manually in ImageJ/Fiji every frame. All the lengths were normalized by L ₀ (cell length at 0 min, the onset of DC).

To analyze cell dynamics after CaLM (Figures 2, 3 and Supplementary Figure S2), the projected cross-sectional images were filtered with a Gaussian filter and then segmented and tracked with a “Tissue Analyzer” (Aigouy et al., 2016) in ImageJ/Fiji. Cell area measurements were carried out with ImageJ/Fiji. The value of cell area from A₀ (the first frame after UV illumination) was set to 1, and all the cell areas were normalized by A₀.

To analyze the VinculinD1-GFP to E-Cadherin-mTomato ratio (Figure 3), the image stacks were processed with “max intensity” under ImageJ/Fiji. The ratio was generated by plugin “Ratio plus” in ImageJ/Fiji, followed by the manual measurements of the intensity. The mean integrated density was measured, as shown in Figure 3, and normalized as follows: I = I t I 0 (3) where the I ₀ is the measurement from the first frame.

To analyze the intensities of MRLC-mCherry (for non-muscle myosin II in Figure 6), the image stacks were processed with “max intensity” in ImageJ/Fiji. Images from five embryos at each stage were performed. The mean gray value for each measurement was performed in a 12.5 × 12.5 µm box from the AS or the lateral epidermis in ImageJ/Fiji. Five independent measurements were performed in each stage embryo and each tissue. The regions for measurements were selected randomly, while the AS and lateral epidermis interface were avoided.

To analyze the intensity of alpha-Tubulin-GFP (Figures 7A,C), the image stacks were processed with “max intensity” in ImageJ/Fiji. The mean gray value for each measurement was performed in a 12.5 × 12.5 µm box from the AS or the lateral epidermis. The regions for measurements were selected randomly and consistent in each frame, while the AS and lateral epidermis interface were avoided.

Mixtures of embryo were stained in the same tubes, imaged with the same microscopic settings, and images were processed and quantified in parallel (Figures 7 and Supplementary Figure S3). For alpha-Tubulin staining in Figures 7B,D,E, the mean gray value for each measurement was measured in a 15 µm × 15 µm box in three embryos, whose position was selected randomly while avoiding the tissue interphase. Three independent measurements were conducted for each stage and tissue. In the case of the phosphorylated regulatory light chain of the myosin II (P-Myo) (Supplementary Figure S3), we measured the mean gray value in a 10 μm × 10 µm box in five embryos for each stage embryo and tissue, again avoiding the tissue interphase.

The ratio of fluorescence intensities in Figures 6C, 7E, and Supplementary Figure S3C was calculated by dividing the mean from the elliptical stage with each independent measurement from the DC stage. Numbers were normalized to the mean intensity from the elliptical stage separately for both tissues.

For line profiles across the tissue interphase (Figure 7H), a line with a width of 10 pixels was place over the tissue interphase. Twelve line profiles from three embryos and aligned to the interphase were averaged for each of the two stages.

Statistical analysis was performed with Prism8. Two-way ANOVA and Sidak’s multiple comparison test estimate the p values in Figures 2D, 4E, 5F. Unpaired two-sided t-test estimates the p values in Figures 3, 5G. Two-way ANOVA estimates the p values in Figures 6, 7, and Supplementary Figure S3. In Figure 5C, two-way ANOVA estimates the p values between the particles toward the same directions (the AS or the lateral epidermis) from different stages; The Mann–Whitney test estimates the p values between particles toward the AS and the lateral epidermis from the same stages.