Ectoderm tensile stress controls mesoderm internalization during embryo gastrulation

Abstract

Gastrulation is the dynamic reorganization of embryonic cells into a multi-stratified epithelial system from which inner and outer organs originate and a mature animal emerges. Tissue folding is one of the key morphogenetic processes driving gastrulation. While the biochemical signals are well deciphered, the biomechanical bases of tissue folding are still poorly understood. Here we study mesoderm invagination during Drosophila gastrulation, a process that results from the folding of the ventral embryonic tissue. The prospective mesoderm forms a furrow and eventually internalizes. How mesoderm internalization is mechanically controlled is still unclear. By combining spatio-temporal stress mapping, two-photon optogenetic-based force patterning, and in toto embryo imaging, we show that the ectoderm tensile stress is developmentally fine-tuned on both lateral sides of the embryo to control the internalization of the mesoderm. Thus, the ectoderm-mesoderm mechanical interplay is key to ensuring folding in embryo gastrulation.

The biomechanical basis for tissue folding during morphogenesis remains incompletely understood. Here they show that precisely balanced mechanical tension in tissues surrounding the future mesoderm is essential for its internalization during fruit fly gastrulation.

Introduction

The folding of an epithelial tissue is a fundamental and vital process during embryo development^1–3, showcasing the remarkable ability of cells to cooperate to drive dramatic tissue shape transformations. Cells packed together into tissues maintain cell–cell contacts while coordinately remodeling their cytoskeletons, generating finely tuned patterns of forces that sculpt the epithelium. Like an origami master, developmental signaling guides these collective cellular rearrangements through the interplay of long-range cues and short-range cell–cell interactions⁴. While much of the biochemical machinery underlying tissue folding is well characterized, the mechanical basis remains poorly understood. To address this, we take advantage of the Drosophila model system, which provides powerful genetic tools for visualization and manipulation. Here we study mesoderm invagination during early embryonic gastrulation, a morphogenetic process that begins with the formation of a ventral furrow along the anterior–posterior (AP) axis and culminates in the internalization of the prospective mesoderm^5,6. While the process of furrow formation has been thoroughly studied and experimentally deciphered both biochemically (reviewed in ref. ^7,8) and mechanically^9–12, how the mesoderm tissue eventually internalizes is still poorly understood. The coordination of the activity of mesoderm cells, whether they are located more centrally close to the ventral midline or more peripherally¹³, is necessary for mesoderm internalization^14,15. In addition, several studies support the idea that the ectoderm-mesoderm coordination plays a critical role in mesoderm internalization: both physical ectoderm immobilization and biochemical ectoderm manipulation result in mesoderm internalization failure^16–18. Here, we shed light on the mechanics of mesoderm invagination by experimentally mapping the tensile stress of the different parts of the embryo as a function of time. Using infrared (IR) femtosecond (fs) laser dissection and two-photon optogenetics to modulate the actomyosin cortex, coupled with multi-view light sheet microscopy for in toto embryo imaging, here we demonstrate how the ectoderm-mesoderm mechanical interplay controls mesoderm internalization, opening new avenues to unveil the physical principles governing tissue folding in embryo gastrulation.

Results

The apical side of the ventral tissue is the only domain displaying a major tensile stress variation during mesoderm invagination

During the first phase of gastrulation in Drosophila, the ventral tissue, composed of prospective mesoderm cells, invaginates. To investigate and map the mechanical forces responsible for mesoderm invagination, we quantified the distribution in space and the variation in time of the actomyosin network in different parts of the embryo and measured the associated tensile stresses. Since the Drosophila early embryo is opaque, to measure filamentous actin (F-actin) and Myosin II (MyoII) levels, we developed a quantitative normalization technique to correct for light scattering at different positions in depth and at different points in time. To that end, we injected fluorescent dextran (Supplementary Fig. 1a, j) as a ground-truth signal to normalize measured MyoII and F-actin local average intensity (Fig. 1a, a’, b, b’ and Supplementary Fig. 1b, c, c’, d, d’, Supplementary Movie 1 first and second video and Methods). From early to late cellularization (i.e., during cell formation), MyoII is mostly enriched basally and increases over time until mid-cellularisation in ventral, lateral and dorsal tissues (Fig. 1a, b, left panel). From mid-cellularization onward, basal MyoII is significantly reduced, showing a 2.8 ± 0.9-fold decrease, eventually reaching the same levels as those measured apically during cellularization, when MyoII does not form apical clusters (Fig. 1b, left panel). During the same time period, F-actin is, on the contrary, apically denser. Eventually, F-actin gradually decreases by a factor of 1.8 ± 1.3 in all tissues until the onset of gastrulation (Fig. 1b, right panel). This decrease is consistent with the depletion of F-actin-rich microvilli at the cell apical side during cellularization¹⁹. Starting from the blastula-to-gastrula transition (t = 0), apical MyoII and F-actin increase sharply by 2.2 ± 0.8-fold specifically at the ventral-apical side of the embryo where the blastoderm tissue flattens and eventually bends (Fig. 1b). While total medial-apical F-actin intensity per cell gradually decreases over time in the mesoderm²⁰, the local sharp increase of F-actin is consistent with cell apical constriction driving overall F-actin concentration. During ventral tissue flattening and bending, MyoII and F-actin located on the lateral cortex of mesoderm cells significantly increase by a factor 1.5 ± 0.3 (Fig. 1b’). On the basal side, MyoII and F-actin intensity persistently decrease in all tissues, reaching lower and higher values, respectively, in the ventral tissue compared with the other tissues. This shows that the distribution of MyoII and F-actin is dynamic in space and time across different regions of the embryo and that mesoderm invagination is characterized by a sharp increase in both MyoII and F-actin on the ventral part of the embryo along the apical and lateral cell sides (see summary representation in Supplementary Fig. 1e).

Fig. 1. Actomyosin spatio-temporal distribution and stress map before and during mesoderm invagination.

a Drosophila embryo cross-section time-lapse from mid cellularization to mesoderm internalization. Top row MyoII labelled with two copies of Sqh::mCherry without endogenous MyoII. Bottom row F-actin labelled with two copies of Moe::ABD::mKate2. Scale bar 50 µm. b MyoII (left) and F-actin (right) intensity variation as a function of position in the embryo (ventral, dorsal and lateral), cell position (apical and basal) and time as shown in (a). b’ as in b for the cell lateral side. n = 25 cells, 5 embryos, and shaded region = mean ± SEM. Time-lapse showing images before and after laser ablation on the cell apical c, basal d, and lateral e side in embryos labelled with two copies of Sqh::mCherry,Utrophin::GFP. Scale bars 20 µm. c’, c” Maximum recoil speed of the actomyosin network after IR fs ablation at the cell apical side. c’ shows recoil speed along the DV axis for a cut performed along the AP axis. c” shows recoil speed along the AP axis for a cut performed along the DV axis. n = 5 embryos for each tissue and stage of ablation for AP ablations, similarly for DV ablations. d’ and d” as in c’ and c” for the cell basal side. n = 5 embryos for each tissue and stage of ablation for AP ablations, similarly for DV ablations. e’ Maximum recoil speed of the actomyosin network after IR fs ablation at the cell lateral side. n = 5 embryos for each tissue and stage of ablation. Red indicates the embryo ventral side, green lateral, and blue dorsal. For ablations, error bar = mean ± SD. Statistical test: One-way ANOVA with post-hoc multiple comparisons, p values are included in the source data, provided as a Source data file.

A previous study used a force inference method based on cell shape changes in 2D along the embryo cross-section and MyoII intensity distribution (named video force microscopy) to predict forces in different parts of the embryo²¹. To directly probe with spatio-temporal precision the tensile stresses generated by the actomyosin network, we took advantage of the dissecting power of IR fs laser²² and imaged the actomyosin recoil after ablation on the apical (using an improved methodology compared to previous work¹⁶—Supplementary Fig. 1f and Supplementary Movie 1 third video and Methods), basal (Supplementary Fig. 1g, Movie 1 fourth video and Methods) and lateral side (Supplementary Fig. 1h and Supplementary Movie 1 fifth video and Methods) of different cell populations (Supplementary Fig. 1i). Further, we analysed the spatio-temporal behaviour of the tissue around the cut, extracting both a temporal and a spatial relaxation rate. We found that neither quantity varied substantially across conditions, except for specific cases (refer to Supplementary Fig. 2a–f’ and Methods). Based on our theoretical framework, this validates the hypothesis that the maximal initial recoil velocity is a proxy for mechanical stress in the direction orthogonal to the cut (Methods). This results in a spatio-temporal stress map of the blastoderm epithelium during mesoderm invagination. Remarkably, our data indicate that the ventral tissue is the only part of the embryo that shows a stark variation in tensile force during the blastula-to-gastrula transition (Fig. 1c–e’). While the tensile force persistently increases from late cellularization to mesoderm bending along a direction parallel to the AP axis (Fig. 1c”), along the dorsal-ventral (DV) direction the force reaches a peak during ventral flattening and then reduces in magnitude (Fig. 1c’), in line with the fact that mesoderm cell strain rate is greater along DV than AP^12,23. While tensile stress mildly but significantly increases also on the lateral side of mesoderm cells during mesoderm bending (Fig. 1e’), all other variations of MyoII and F-actin intensity in other parts of the embryo do not reflect any major change in mechanical tension. Overall, the spatio-temporal stress map shows that the ventral tissue is the main domain displaying a major increase in recoil-based cortical tensile stress during mesoderm invagination.

The brd gene controls the ectoderm tensile stress on the ventrolateral left and right sides of the embryo

The prospective mesoderm that undergoes invagination encompasses the ventral portion of the embryo and is composed of about 1000 cells of the 6000 that build the blastoderm epithelium^5,6. While our tensile stress analysis indicates that the ventral tissue is the active force generator during mesoderm invagination, the remaining tissue may produce passive forces as a reaction to the change in the boundary conditions (in agreement with Newton’s third law). To investigate how the mechanics of mesoderm-neighbouring tissues participate in mesoderm invagination, we thoroughly addressed the bearded family of genes that is expressed ubiquitously in the blastoderm epithelium but is silenced in the mesoderm by Snail^24,25. Bearded proteins suppress the activity of the E3 ubiquitin ligase Neuralized that is expressed in a ventral-to-dorsal gradient, and that is necessary for mesoderm invagination^{17,24,26–29}. In a bearded knockout embryo (brd −/−), the activity of Neuralized extends beyond the ventral region, and the mesoderm forms a furrow that eventually unfolds, resulting in invagination failure¹⁷. In brd −/− embryos, the initial extension of the ectoderm during mesoderm invagination is not affected (Supplementary Fig. 3g). To further dissect the brd −/− mutant, we implemented multi-view light sheet microscopy to image the embryo in toto over time (see Methods). Our data confirm that mesoderm invagination is aborted in a brd −/− background, while it is rescued in a brd +/− background (Fig. 2a, a’, b, b’ red curve and g and Supplementary Movie 2 first video, in comparison with d, d’, e and e’ red curve and h and Supplementary Movie 2 sec video) in line with the fact that the complete bearded loss of function effect is achieved in a homozygous mutant. Our multi-dimensional image analyses show that, during the ventral flattening and bending phases, the rate of apical constriction of mesoderm cells in a brd −/− embryo is overall the same as in control conditions (until t = 9 min – Fig. 2b, b’ in comparison with e and e’). During the next 10 min (from t = 10 to t = 20 min), the invagination is stalled, the mesoderm cells start to dilate (undergoing cell rounding and eventual mitosis) and the ventral tissue unfolds (Fig. 2b, b’ black curve compared to e and e’ black curve). In addition, in a brd −/− background both left and right prospective ectoderm tissues displace symmetrically towards the ventral midline as in control conditions but at a rate approximately threefold slower (R = 0.32 ± 0.09) (Fig. 2c, c’ in comparison with f and f’). Overall, our analyses of morphology and tissue dynamics support the idea that, during mesoderm invagination, brd controls the rate of displacement of the left and right lateral ectoderms towards the ventral midline and does not directly control the apical constriction rate of mesoderm cells, in line with the fact that brd is silenced ventrally.

Fig. 2. The brd gene controls the mechanical tension of the ventrolateral ectoderm.

a Time-lapse cylindrical projection of a brd −/− embryo labelled with two copies of Sqh::mCherry (MyoII). V ventral, LL lateral left, LR lateral right and D* dorsal shown twice. n = 4 embryos. Scale bar 100 µm. a’ Embryo scale kymograph. Dashed lines indicate constriction and eventual dilation of the ventral tissue. V, LL, LR, and D* as in (a). b Plot of apical constriction and mesoderm internalization as a function of time (related to a). b’ Plot of rate of apical constriction and rate of mesoderm internalization as a function of time (related to a). c Plot of lateral left (LL) and lateral right (LR) tissue displacement towards ventral midline as a function of time (related to a). c’ Plot of the displacement rate of lateral left (LL) and lateral right (LR) tissues (related to a). d and d’ Control (brd +/−) embryo shown as in a and a’, respectively. n = 4 embryos. e and e’ as in b and b’ for control (brd +/−) embryos. f and f’ as in c and c’ for control (brd +/−) embryos. Shaded region = mean ± SD. g Cross-section time-lapse of a brd −/− embryo. Scale bar 50 µm. h as in g for a control (brd +/−) embryo. i Angular plot showing the distribution of the barycentre of the ventrolateral MyoII contractile zones. n = 6 embryos. j Representative time-lapse showing actomyosin recoil (yellow arrows) after laser ablation (red line) on the ventrolateral side of a brd −/− (top) and brd +/− (bottom) embryo labelled with two copies each of Sqh::mCherry (MyoII) and Utrophin::GFP (F-actin). Scale bar 20 µm. k Plot showing the distribution of maximum recoil velocities of the actomyosin network after laser ablation on the ventrolateral side of the embryo in a brd −/−, brd +/−, and wild-type background. n = 5 embryos. Error bar = mean ± SD. Statistical test: One way ANOVA, p value = 0.0018, with post-hoc multiple comparison, p values in k: 0.0112, 0.0486, >0.9999. Source data are provided as a Source data file.

Next, we further explored possible differences in cytoskeletal distribution that could account for alterations in the ectoderm tissue behaviour. During mesoderm invagination in a brd −/−, MyoII is upregulated in the ventrolateral zone of the embryo at the apical side of cells (Supplementary Fig. 3a, a’, b, b’)¹⁷. Our data show that MyoII is upregulated bilaterally on both the right and left lateral ectoderm in a zone located 60° from the ventral midline measured along a cross-section (Fig. 2i and Supplementary Fig. 3c). Therefore, we can hypothesize that the altered distribution of MyoII could result in a change in the mechanical properties of the two tissues located on the ventrolateral sides of the embryo. Ventrolateral cells have been shown to form periodic and contractile actomyosin pulses in the apical-medial region favouring cell intercalation^30–32. These pulses increase both in frequency and amplitude in brd mutants¹⁷. Cell contractility can result in active fluctuations favouring epithelial fluidization or upregulation of the overall tensile stress at the tissue scale^33,34. To directly test tensile stress, we used IR fs ablation as described previously. Remarkably, the ventrolateral tension is 3.2 ± 1.8-fold higher in brd −/− compared with control conditions (i.e., brd +/−), where tension shows values similar to those in a wild-type background (Fig. 2j and Supplementary Figs. 2g, h and 3d, e). Overall, our data show that the brd pathway controls the mechanical properties of the two ventrolateral regions of the embryo during mesoderm invagination (Supplementary Fig. 3f). This leads us to hypothesize that a change in tissue mechanics of the epithelia neighbouring the mesoderm could interfere with the internalization of the ventral tissue.

Bilateral tensile stress upregulation on the ventrolateral left and right ectoderm results in mesoderm internalization failure

The analyses on brd −/− embryos suggest that an upregulation of the ectoderm tensile stress could abort mesoderm invagination. This idea is also supported by the fact that ectopic upregulation of Rho in the two lateral ectoderms leads to mesoderm invagination failure¹⁷. In order to further test this hypothesis, we devised a new experimental approach to test different tensile stress patterns by modulating tissue mechanical tension dynamically in an arbitrary fashion. To that end, we implemented two-photon Cry2-RhoGEF2 optogenetics coupled with multiview light-sheet microscopy to upregulate the amount of Rho guanine nucleotide exchange factor RhoGEF2 with spatio-temporal specificity while simultaneously performing in toto embryo imaging³⁵. This results in increased actomyosin contractility in the activated regions³⁶. Our tensile stress analysis shows that the tension of the activated tissue can increase by up to 4.5 ± 2.8-fold compared with control conditions (Fig. 3i, j and Supplementary Fig. 2g, h), reaching magnitudes that are similar to those measured in a brd −/− background (see Figs. 3j and 2k). This shows that two-photon optogenetics is an effective tool to faithfully reproduce complex tensile stress patterns. To mechanically phenocopy the brd −/− embryo, shortly before gastrulation, we activated two 10- to 12-cell-wide regions along the DV direction located on the left and right ventrolateral apical sides of the embryo 60° from the ventral midline (as in brd −/−, Supplementary Fig. 3c) (see Fig. 3a and refer to Methods). The width of the activated regions was chosen ad hoc to avoid activation of a 20-cell-wide DV region on the ventral side (i.e., the mesoderm). In activated embryos, the initial extension of the ectoderm, during mesoderm invagination, is not affected (Supplementary Fig. 3g). In 60° bilateral tension upregulation embryos, the ventral tissue forms a shallow furrow that fails to internalize (Fig. 3a’, b, b’ g and Supplementary Movie 3 sec video compared with control shown in Supplementary Fig. 4a, a’, b, b’, g and Supplementary Movie 3 first video). In addition, the two ectoderms displace slower than in control conditions (Fig. 3c, c’ and Supplementary Fig. 4c, c’) at a speed comparable to that measured in a brd −/− background (Fig. 2c, c’). To test if bilateral tension upregulation is necessary to abort mesoderm invagination, we activated only one 60° side and monitored tissue behaviour and dynamics. Under this condition, tissue dynamics are the same as in control embryos, and the mesoderm properly invaginates (Supplementary Fig. 4d–f’, h and Supplementary Movie 3 third video), showing that bilateral tension upregulation is necessary to block the ingression of the ventral tissue. A reduced size of the activation region can affect ventral furrow propagation dynamics but can no longer abort internalization (Supplementary Movie 3 fifth video). We then wondered if the angular position of the tension upregulation zone plays a role in the invagination process. To test this, we shifted the activation region from 60° to 90° from the ventral midline (i.e., the most lateral zone of the embryo) on both left and right sides. 90° bilateral tension upregulation does not affect mesoderm invagination that proceeds with dynamics similar to control conditions (Fig. 3d–f’, h, Supplementary Fig. 4i and Supplementary Movie 3 fourth video). Remarkably, 90° bilateral tension upregulation does not affect the rate of mesoderm cell apical constriction that is instead partially reduced in the 60° bilateral tension upregulation, highlighting the spatial range of mechanical coupling (Figs. 3e, e’, b, b’). Overall, our data and analyses demonstrate that the upregulation of the ectoderm tensile stress can result in mesoderm invagination failure if tension is specifically increased bilaterally on both ventrolateral sides of the embryo.

Fig. 3. Ectoderm mechanical tension upregulation aborts mesoderm invagination.

a Time-lapse cylindrical projection of a 60° left- and right-activated embryo resulting in bilateral tension upregulation on the ventrolateral sides of the embryo labelled with one copy of Sqh::mCherry (MyoII). V ventral, LL lateral left, LR lateral right, and D* dorsal shown twice. Dashed red rectangle indicates region of activation. n = 3 embryos. Scale bar 100 µm. a’ Embryo scale kymograph related to a. Dashed white lines indicate constriction and eventual dilation of the ventral tissue. Dashed red lines indicate the region of activation. V, LL, LR, and D* as in a. b Plot of apical constriction and mesoderm internalization as a function of time (related to a). b’ Plot of rate of apical constriction and rate of mesoderm internalization as a function of time (related to a). c Plot of lateral left (LL) and lateral right (LR) tissue displacement towards ventral midline as a function of time (related to a). c’ Plot of the displacement rate of lateral left (LL) and lateral right (LR) tissues (related to a). d and d’ 90° left-right activated embryo shown as in a and a’, respectively. n = 3 embryos. e and e’ as in b and b’ for 90° left-right activated embryos. f and f’ as in c and c’ for 90° left-right activated embryos. Shaded region = mean ± SD. g Cross-section time-lapse of a 60° left-right activated embryo. Scale bar 50 µm. h as in g for a 90° left-right activated embryo. i Representative time-lapse showing actomyosin recoil (yellow arrows) after laser ablation (red line) on the ventrolateral side of an activated (top) and control (bottom) embryo labelled with one copy of Sqh::mCherry (MyoII). n = 8 embryos. Scale bar 20 µm. j Plot showing the distribution of maximum recoil velocities of the actomyosin network after laser ablation on the ventrolateral side of the embryo in activated and control conditions. N = 8 embryos. Error bar = mean ± SD. Statistical test: two-tailed Mann–Whitney U test, p value: 0.0002. Source data are provided as a Source Data file.

Tensile stress in the ventrolateral ectoderm is developmentally fine-tuned to promote mesoderm internalization

Bilateral upregulation of tensile stress on the ventrolateral ectoderm regions aborts mesoderm invagination. How would mesoderm invagination be affected if the ventrolateral tension were downregulated instead, or, in the most extreme case, if tissues were physically disconnected from each other? To answer this question, we implemented two-photon Cry2-OCRL optogenetics to depolymerize F-actin and, therefore, to lower tissue tensile stress dynamically with spatio-temporal specificity³⁷. Our IR fs laser-based dissection trial indeed shows that, after Cry2-OCRL activation, tissue tension is reduced to approximately half of control levels (R = 0.482 ± 0.302, Fig. 4i, j). We next downregulated tension with different spatial patterns along the DV axis. Given the fact that tension upregulation on the two ventrolateral regions results in mesoderm internalization failure, we predicted that tension downregulation would, on the contrary, preserve or even favour internalization. Surprisingly, 60° bilateral tension downregulation also results in a mesoderm furrow that fails to internalize (Fig. 4a, a, g and Supplementary Movie 4 sec video compared with control in Supplementary Fig. 5a, a’ and g and Supplementary Movie 4 first video). Nevertheless, under these conditions, tissue dynamics differ from previous experiments based on 60° bilateral tension upregulation or brd mutations. Indeed, in 60° bilateral Cry2-OCRL activation trials, apical constriction of mesoderm cells proceeds at a faster pace compared with control conditions and the ectoderm displacement is overall maintained (Fig. 4b, b’, c, c’ and Supplementary Fig. 5b, b’, c, c’). 60° unilateral tension downregulation preserves mesoderm internalization (Supplementary Fig. 5d–f’ and h and Supplementary Movie 4 third video). This shows that the tensile stress of the two ventrolateral regions is fine-tuned to promote mesoderm internalization: both ventrolateral tension upregulation and downregulation result in mesoderm internalization failure. While 90° bilateral tension downregulation does not interfere with mesoderm invagination, showing tissue dynamics similar to control conditions (Fig. 4d–f’, h and Supplementary Movie 4 fourth video and Supplementary Fig. 5a–c’, g), we then wondered if further tension downregulation up to complete ablation of the ectoderm tissue at bilateral 90° could impair mesoderm internalization. To achieve such an extreme phenotype, we performed sustained activation that eventually leads to failed cellularization, resulting in an aperture in the epithelium (Supplementary Fig. 5i). This makes Cry2-OCRL activation an ideal tool to partially disconnect tissues (Supplementary Fig. 5j, k). Remarkably, 90° bilateral tissue ablation and 180° tissue ablation still do not impair mesoderm internalization (Supplementary Fig. 5j and Supplementary Movie 4 fifth and sixth videos). Overall, our data show that the two ventrolateral parts of the ectoderm participate in mesoderm internalization while the remaining ectoderm tissue is dispensable.

Fig. 4. Ectoderm mechanical tension downregulation aborts mesoderm invagination.

a Time-lapse cylindrical projection of a 60° left and right activated embryo resulting in bilateral tension downregulation on the ventrolateral sides of the embryo labelled with Gap43::mCherry (membrane). V ventral, LL lateral left, LR lateral right, and D* dorsal shown twice. Dashed red rectangle indicates region of activation. n = 3 embryos. Scale bar 100 µm. a’ Embryo scale kymograph related to (a). Dashed white lines indicate constriction and eventual dilation of the ventral tissue. Dashed red lines indicate the region of activation. V, LL, LR, and D* as in a. b Plot of apical constriction and mesoderm internalization as a function of time (related to a). b’ Plot of rate of apical constriction and rate of mesoderm internalization as a function of time (related to a). c Plot of lateral left (LL) and lateral right (LR) tissue displacement towards ventral midline as a function of time (related to a). c’ Plot of the displacement rate of lateral left (LL) and lateral right (LR) tissues (related to a). d, d’ 90° left-right activated embryo shown as in a and a’, respectively. n = 3 embryos. e, e’ as in b, b’ for 90° left-right activated embryos. f, f’ as in c, c’ for 90° left-right activated embryos. Shaded region = mean ± SD. g Cross-section time-lapse of a 60° left-right activated embryo. Scale bar 50 µm. h as in g for a 90° left-right activated embryo. i Representative time-lapse showing actomyosin recoil (yellow arrows) after laser ablation (red line) on the ventrolateral side of an activated (top) and control (bottom) embryo labelled with Gap43::mCherry (membrane), mCherry::Cry2::dOCRL, and one copy of Moe::ABD::mKate2 (F-actin). Upon activation, the cytoplasmic mCherry::Cry2::dOCRL localizes to the membrane, causing a decrease in cytoplasmic mCherry::Cry2::dOCRL intensity. n = 5 embryos. Scale bar 20 µm. j Plot showing the distribution of maximum recoil velocities of the actomyosin network after laser ablation on the ventrolateral side of the embryo in activated and control conditions. n = 5 embryos. Error bar = mean ± SD. Statistical test: two-tailed Mann–Whitney U test, p value for j: 0.0317. Source data are provided as a Source data file.

Discussion

Epithelial folding is a vital tissue transformation during embryo development. Mesoderm invagination in the developing Drosophila embryo is one of the most iconic and well-studied model systems for tissue folding. The mesoderm tissue, located on the ventral side of the blastoderm epithelial monolayer, initially flattens, then bends to form a furrow and finally internalizes. While much is known about the biochemical signals and the mechanical forces driving mesoderm furrowing, how the mesoderm finally internalizes is still not well understood. While previous work has provided insights into the cytoskeletal modulation and ensuing cell shape changes within the mesoderm^13,38, here, we provide a thorough mechanical analysis of this process at the embryo scale. We unveil an ectoderm-mesoderm mechanical interplay that is key for mesoderm internalization, opening new avenues to decipher the physics underlying such a dramatic tissue shape transformation.

Our quantitative tensile stress map supports the idea that the process of mesoderm invagination is fully powered by the ventral tissue generating active tensile forces at the apical^9,12 and eventually lateral³⁹ side of cells. This is reflected by the fact that among dorsal, lateral and ventral tissues, only the mesoderm epithelium undergoes a stark tensile stress upregulation at the onset of gastrulation. Previous work has advanced the possibility that the downregulation of MyoII at the basal side of mesoderm cells is required for ventral tissue invagination⁴⁰. This idea was inferred by the evidence that basal activation of the Cry2-RhoGEF2 protein complex shows increased levels of cell basal MyoII accompanied by mesoderm internalization failure. Our quantitative actomyosin intensity analysis combined with the tensile stress map corroborates the notion that changes in cortical MyoII intensity per se do not systematically reflect tensile stress changes (in agreement with⁴¹). For instance, the mesoderm basal side shows overall constant (or at most a minor increase of) tensile stress despite the sharp MyoII downregulation during the different phases of the blastula-to-gastrula transition. This supports the idea that downregulation of the mesoderm basal MyoII is unnecessary to ensure basal tension downregulation for mesoderm invagination and that basal activation of Cry2-RhoGEF2 ectopically upregulates and possibly remodels the actomyosin cytoskeleton in a way that is no longer relatable to control conditions. MyoII activity can also modulate cortical stiffness⁴². This raises the possibility that basal cortical stiffness, rather than tensile forces, is downregulated in mesoderm cells to enable invagination. In a recent study, Brillouin microscopy was applied to estimate the longitudinal elastic modulus of the mesoderm tissue during invagination⁴³. This work does not report a Brillouin shift at the basal side of mesoderm cells. In addition, we show that the viscoelastic temporal and spatial relaxation rates, which characterize the mechanics of the basal cortex, remain unchanged in the course of the downregulation of the mesoderm basal MyoII (Supplementary Fig. 2e, f). Overall, this evidence argues against stiffness modulation at the basal side of ventral cells during mesoderm invagination.

Previous work has highlighted the interdependency between ectoderm and mesoderm movements¹⁶ and shown how changes in ectoderm Rho signalling can drastically impact mesoderm internalization¹⁷. This study provides deeper insights into the ectoderm-mesoderm mechanical interplay. Our dynamic tensile stress patterning coupled with in toto embryo imaging demonstrates that bilateral tensile stress upregulation on the ventrolateral left and right ectoderm results in mesoderm invagination failure. Therefore, our data support a model in which the mesoderm tissue is the active force generator while the ectoderm, physically connected to the mesoderm, generates passive resistive forces. If the resisting tensile stress is upregulated on both sides of the mesoderm boundaries, the ventral tissue is constrained in a two-way tug-of-war that results in internalization failure. In line with this model, a logical inference is that ectoderm tensile stress downregulation would in principle preserve, if not accelerate, mesoderm internalization. Our data contradict this reasoning, demonstrating that bilateral tensile stress downregulation on the ventrolateral left and right ectoderm also results in mesoderm invagination failure. This counterintuitive evidence shows that the ectoderm resistive forces also play a key role in mesoderm invagination. Remarkably, under these conditions, the apical constriction rate of mesoderm cells doubles. This highlights a model based on an interplay between mesoderm active forces and ectoderm passive forces to finely control mesoderm strain rate. How might the strain rate of the ventral tissue affect mesoderm internalization? The dynamics of mesoderm cells during ventral tissue internalization are yet to be thoroughly characterized. Overall, pairs of AP-constricting mesoderm cell rows, symmetrically located on the left and right sides of the mesoderm midline, sequentially meet in the centre before spreading on each other (see Supplementary Movie 4 sixth video), resembling the limbs of two tapes pulled through the slider of a zipper. This results in an inward-directed movement. Our analysis shows that bilateral tension downregulation on the ventrolateral left and right sides of the embryo results in the doubling of mesoderm cell apical constriction rate. Hypothetically, excessively rapid constriction could collapse the mesoderm tissue towards the ventral midline, preventing the sequential encounter of symmetrically located cell rows and aborting mesoderm internalization.

While the mechanisms driving mesoderm internalization are still unclear, the scientific community agrees overall on the idea that mesoderm invagination is the result of two sequential processes: (1) ventral furrow formation and (2) internalization, driven by distinct mechanisms (as supported by ref. ⁴⁴) operating at the cellular and embryo scales. Ventral furrow formation relies on contractions of the medial-apical actomyosin cortex of mesoderm cells, which generate a supracellular contractile network. Contractile forces are then transmitted through this network to the anterior and posterior poles of the embryo along the AP axis, forming a tensile string that indents the ventral epithelium^9,12,45. This first phase of mesoderm invagination is independent of the ectoderm-mesoderm interplay. Mesoderm internalization is characterized by a still poorly studied cell-zippering mechanism and is intimately dependent on the DV force transmission resulting from the ectoderm-mesoderm mechanical interaction on the left and right ventrolateral sides of the embryo. This interaction could be facilitated by the upregulation of cell-cell adhesion sites in the ventral-most region of the ectoderm on both left and right sides of the embryo¹⁸.

Which cellular mechanisms could drive the internalization of the ventral tissue? Mesoderm cell lateral tension³⁹ and de novo medial-apical junction formation⁴⁶, in coordination with medial-apical actomyosin remodelling, are possible mechanisms that may drive tissue involution and that need to be addressed in the future.

Shortly after the mesoderm tissue forms a furrow and begins internalization, MyoII is apically upregulated in the ectoderm, leading to a tensile stress buildup with minute-scale temporal precision. This study shows that such an increase in mechanical tension is developmentally fine-tuned, and that a change in tension magnitude or timing can be lethal, causing mesoderm internalization failure and aborting gastrulation. While much is known about the spatial gene patterning and signaling pathways that regulate MyoII and cortical forces, how the magnitude and timing of tensile stresses are fine-tuned across different regions of the embryo remains an open question. Cell mechanotransduction^18,47–49 and spatio-temporal control of transcriptional and translational dynamics⁵⁰ are promising avenues of research that merit deeper investigation.

Method

Experimental models and study participants

Fly stocks and genetics

All fly stocks and crosses were maintained at room temperature (22 °C). Drosophila stocks used in this study for live imaging were: Myosin-mCherry without endogenous MyoII contribution (Sqh^AX3, BDSC #57144) (Sqh^AX3; Sqh::mCherry), F-actin-mKate2 (Moe::ABD::mKate2, BDSC #76270), Snail-GFP (Snail::sGFP, VDRC #318449), Myosin-mCherry with F-actin-GFP (Sqh::mCherry, Utrophin::GFP), Myosin-GFP (Sqh::GFP, a gift from Xiabo Wang), used for basal depth guidance in acquiring Dextran intensities. Bearded control embryos with MyoII-mCherry and F-actin-GFP (Sqh::mCherry, Utrophin::GFP; brd^-/TM6), Bearded mutant embryo with MyoII-mCherry and F-actin-GFP (Sqh::mCherry, Utrophin::GFP; brd^-/ brd^-). For opto-recruitment of RhoGEF2 with MyoII-mCherry, embryos were collected from Sqh::mCherry/UASp Cry2::RhoGEF2; _posk Gal4/UASp CIBN::pmGFP mothers. For opto-recruitment of OCRL-Pip2 with membrane-mCherry, embryos were collected from _pubi: Gap43::mCherry/+; UASp CIBN::pmGFP/ tub^mat Gal4; UASp mCherry::Cry2::dOCRL/ tub^mat Gal4 mothers. For opto-recruitment of OCRL-Pip2 with membrane-mCherry and F-actin-mKate2 to perform ablations on F-actin upon opto-activation, embryos were collected from _pubi:Gap43::mCherry/+; UASp CIBN::pmGFP /tub^mat Gal4; Sqh::Moe::ABD::mKate2/ UASp mCherry::Cry2::dOCRL mothers. Optogenetic lines were a gift from S. De Renzis.

For obtaining complete loss of function of all Brd family genes, embryos homozygous for both Df(3)Brd-C1 and Df(3)E(spl)δ−6 deletions were considered (referred in this work as brd −/−), as described previously⁵¹. Brd deficient lines were a gift from F. Schweisguth.

Method details

Confocal time-lapse imaging

Embryos were mounted in water on glass bottom plates (MatTek) after dechorionation with bleach. For embryos expressing optogenetic constructs, the flies and crosses were kept in a dark environment, and embryo preparation was performed using commercially available red light to prevent unwanted activation. Confocal imaging was carried out on either Zeiss LSM 780 or 980 Airyscan (Carl Zeiss) using a 40 × 1.1 NA water-immersion objective for deep tissues and a 40 × 1.2 NA objective for surface imaging with 488 nm argon or 561 nm DPSS laser lines. Imaging data were acquired using Zeiss Zen black software and processed in ImageJ. For cross-section time-lapse imaging, embryos were mounted with the posterior pole facing the objective, with a z-depth of 150 μm set for continuous imaging at a frame rate of 1 min.

In toto 4D imaging

Individual embryos at the appropriate stage were collected under halocarbon oil (Sigma), dechorionated in bleach, and washed thoroughly before mounting. Embryos were mounted in a glass capillary filled with 1% low-melting agar (VWR Life Science), with their anterior-posterior axis aligned parallel to the capillary. A small section of the solidified agar containing the embedded embryo was pushed outward from the capillary for imaging on a Bruker Luxendo Multi View Light-Sheet Microscope (MuVi-SPIM) equipped with an Olympus 20 × 1.0 NA detection objective, using 488 nm and 594 nm light sheets. For in-toto imaging, z-stacks were defined with a 1 μm step size; the stage was rotated 90 degrees between each Z-stack acquisition to capture opposing orthogonal views. Consequently, for two-color imaging at each time point, four 3D stacks of the embryo were recorded. A frame rate of 1 min and 20 s was used for brd +/− and brd −/− background embryos (related to Fig. 2). A frame rate of 33 s was used for Cry2-RhoGEF2 and Cry2-OCRL opto-activated embryos (related to Figs. 3 and 4 and Supplementary Figs. 4 and 5).

4D data digital reconstruction

The stacks obtained from the Luxendo MuVi SPIM were then fused using the ASTEC fusion pipeline⁵², and finally, a single image with isotropic resolution (0.29 μm pixel resolution) was obtained for each time point. Stacks, once aligned and fused, are then cylindrically projected like a carpet to extract embryo surface dynamics using the MATLAB protocol described in ref. ¹⁶. Bright streaks appearing in the cylindrical projections are artifacts resulting from the fusion and unrolling of the data (similarly as in ref. ¹⁶).

Multiphoton laser manipulations

Ablation of the actomyosin network was conducted by employing a tunable Spectra-Physics Mai Tai DeepSee IR fs laser (Newport Corp., Irvine, CA, USA) integrated into a Zeiss 780 LSM confocal microscope (Carl Zeiss), tuned to 950 nm. The ablation process utilized the bleaching mode in Zeiss Zen black software, over an ROI of 20 μm long thin line instead of a previous method of using a rectangular ROI¹⁶, this creates a clean ablation and helps in direct probing of tensile stress. A laser power of 115 mW at the focal point, with 10 iterations and 2.55 µs pixel dwell time, was utilized. The imaging was performed at a frame rate of 780 ms for apical, lateral, and basal ablations using a Zeiss 40 × 1.1 NA infrared-corrected water-immersion objective. For ventral tissue basal ablations, a pre-Z stack of Snail::GFP was acquired to ensure the presumptive mesoderm tissue specificity. To ensure lateral and dorsal tissue specificity, the tissue was observed post-ablation until gastrulation movements became apparent. One ablation was carried out per embryo at each stage. For example, the MC-stage apical AP-axis tensile study used five embryos, with one ablation on each embryo. For brd −/− and brd +/−, ablations were performed on the ectoderm at a 60° angle to the ventral midline when mesoderm furrowing began. To prevent affecting tensile stress, no embryo received more than one ablation. For optogenetic activation and ablation assays, activation lasted 4 min and 30 s, followed immediately by ablation as previously described.

For ablation experiments, a positional tolerance assay was performed for mounting the embryo at ᾱ=60°, 90°, and 180° from the ventral midline. For 60° and 90°, embryos were mounted at appropriate angles, and a Z-stack of 90 µm thickness was acquired to measure the accuracy of the mounting angle relative to the ventral midline. For 180°, embryos were mounted dorsally, and the slide was rotated 90° so that the posterior end faces the objective. In this case, angle measurements from the mounting point to the ventral midline were obtained from the cross-sectional image (refer to Supplementary Fig. 1i).

To perform In-toto imaging with photomanipulation, a 920 nm IR laser (Alcor, Spark Lasers), operating at an 88 fs pulse duration and a 79.73 MHz repetition rate, was coupled to a Bruker Luxendo MuVi SPIM³⁵. For Cry2-OCRL and Cry2-RhoGEF2 opto-recruitment, the embryo preparation was carried out in red light. A rectangular region of interest (200 × 80 μm) was determined in the Lux software after rotating the embryo at the required angles (eg, 60° and 90°). A shorter region of interest (80 × 50 μm) was utilized to test the dimensional role of ROI in optogenetic experimental assays. A pre-activation 3D acquisition was followed by continuous firing of the laser until an effect was observed. This was followed by 4D in-toto imaging. In the case of Cry2-RhoGEF2 90° bilateral and 60° unilateral, the activation was started slightly earlier and for a prolonged duration to produce a stronger phenotype, thereby validating that contractility in these tissue regions does not affect ventral tissue internalization. For Cry2-OCRL and Cry2-RhoGEF2 weak activation, a lower laser power of ~21% (maximum estimated power after transmittance of 920 nm IR laser through the 1.0 NA objective at 21% is ~115 mW) was utilized. For stronger optogenetic-based ablation of tissue using Cry2-OCRL, a higher laser power of ~24% (maximum estimated power after transmittance of 920 nm IR laser through the 1.0 NA objective at ~24% is ~131 mW) was utilized with an increased number of iterations. During the initial attempts, the ablated ectodermal region caused purse-string constriction. This issue was resolved by ablating a larger region of ectoderm, which prevented purse-string constrictions while retaining the ablated region.

Dextran dye injection

Injection needles were prepared using a needle puller (Narishige) with borosilicate glass capillaries containing filaments (Sutter Instruments). The needles were then cut to achieve the appropriate diameter for injection.

For cross-section injection and imaging, embryos were prepared after dechorionation with bleach, and were then glued laterally to the edge of a cut glass slide. The glass slide was adhered to the coverslip using double-sided tape, creating a space for the unglued posterior end of the embryo to face the objective. After desiccation for approximately ~12 min, the embryos were immersed in halocarbon oil (Sigma) to prevent further drying.

Embryos were then injected mid-laterally prior to cellularization onset during the 13th cell cycle with 1 mg/mL dextran (ANTONIA Red Dextran 20, Merck #18650) diluted in water. After injection, the embryos were allowed a minimum of 15 min for homogenized distribution of the dye (refer Supplementary Fig. 1j) before being imaged on a 780 LSM confocal microscope (Carl Zeiss) at a Z position of 150 μm from the posterior pole of the embryo. To image dextran-RED fluorescence, the same excitation and emission settings used for imaging cross-sections of embryos labeled with MyoII (Sqh::mCherry) and F-actin (MoeABD::mKate2) were adopted to ensure consistent normalization during intensity quantification.

Embryo time alignment

Embryo T0 was determined by the appearance of MyoII apical clusters in ventral tissue, which is the onset of blastula-to-gastrula transition. In the case of embryos without a MyoII marker, T0 is determined 2 min before the onset of apical cell shape change in ventral tissue (a reproducible time interval between cell shape change and MyoII accumulation)

Quantification and statistical analysis

2D cell segmentation and area measurements

2D segmentations of mesoderm cell surface (analysis was performed 6 μm from the apical) were carried out using the Tissue Analyzer⁵³ plugin in ImageJ on cylindrical projected time-lapse images. Individual cell area measurements obtained from the segmented dataset were averaged and plotted against time using GraphPad Prism7. F-actin marker (Utrophin::GFP) or membrane marker (Gap43::mCherry) was used to mark the cell boundaries for segmentation.

Furrow depth measurement

Furrow depth was measured using cross-section movies, which were then rotated to attain ventral tissue at the bottom. The manual tracking plugin in ImageJ was used for this analysis. After setting the XY calibration and frame interval details based on the image properties, a track was created in the plugin. The displacement of the tip of the ventral apical tissue undergoing furrowing in the Y axis was measured relative to the ventral apex.

Displacement and recoil velocity measurements

The initial recoil velocity was measured using the manual tracking plugin in ImageJ. Initial calibrations were done, as mentioned previously. Each click recorded the coordinates of the region to be tracked (here, the cut ends, either the MyoII cortex for basal ablations or the actomyosin networks for apical and lateral ablations). The coordinate data and tracings were then saved for further calculation and plotting.

To calculate the displacement, the equation S = √ ((y₂ - y₁)² + (x₂ - x₁)²) was used, where x and y represent the xy coordinates. The initial maximum recoil velocity was determined from the first derivative (dS/dT) of the displacement over time.

To calculate the distance traveled by the VL and DL under both decoupled and control conditions, the step distances were first calculated using the equation described above, followed by the calculation of the cumulative distance. Subsequently, the difference (VL-DL) in traveled distance within the left (L) and right (R) lateral ectoderms was determined (refer to Supplementary Fig. 5k).

Spatio-temporal velocity field analysis and viscoelastic parameter extraction

Spatial and temporal relaxation rates were measured processing the same image stacks and tracking data as for the initial recoil velocity analysis described above, using a custom Python pipeline (full algorithmic and theoretical details in Supplementary document: Modeling and Code)

For the spatial relaxation rate α, pixel intensities were first normalised per stack to the 0.5th–99.5th percentile range. The cut position was detected automatically from the inter-frame intensity difference image by correlating a template encoding the expected ablation length (20μm) after Gaussian smoothing, yielding cut endpoints. For Bearded-mutant and optogenetic stacks, cut coordinates were read from a red annotation line drawn on the first post-cut frame in Fiji. A region of interest centered on the cut site was extracted, rotated 90° clockwise to align the cut axis horizontally, and smoothed. Dense TV-L1 optical flow (OpenCV DualTVL1OpticalFlow) was then computed between each pair of consecutive frames (five intervals per embryo), independently on each side of the cut to prevent velocity discontinuities from propagating across the ablation site. The x-component of the flow field was averaged over a ±40 px band centred on the cut, excluding a ±24 px exclusion zone around the cut itself. The resulting velocity profile vₓ(x) was fitted to a decaying exponential vₓ(x)=sign(x)·B·exp(−α| x |) by Levenberg–Marquardt least squares, yielding the spatial relaxation rate α. Fits were retained when the optimizer converged, the root-mean-square residual RMSE < 0.2 μm s⁻¹, and the fit amplitude B ≥ RMSE to account only for reliable measurements where recoil was sufficient for spatial signal. The per-embryo mean across retained frames (up to 5 per embryo) was used as the statistical unit for all group comparisons.

For the temporal relaxation rate k, the inter-vertex distance d(t) measured by manual tracking was fitted to the Kelvin–Voigt recoil model d(t) = d₀ + A(1−e⁻ᵏᵗ), where d₀ is the inter-vertex distance at the first post-cut frame, A the plateau amplitude, and k the temporal relaxation rate (s⁻¹). Fits were retained when the coefficient of determination R² ≥ 0.60.

For each condition, per-embryo mean values of α and k were computed across retained frames and used for statistical comparisons.

Embryo cross-section MyoII accumulation angle measurements

The ventrolateral tissue MyoII enrichment angle in brd−/− was measured using the ImageJ angle tool, fixing ventral midline = 0° and dorsal = 180°.

Ectoderm extension length measurements

The lateral ectoderm extension was measured using the ImageJ manual tracking plugin. Two cells, separated by 190 μm along the AP axis parallel to the ventral midline in the ventrolateral region (60° from the ventral midline), were tracked from the onset of ventral cell apical constriction until mesoderm internalization (unfolding in the case of brd −/− and Cry2-RhoGEF2 60° bilaterally activated). The percentage difference in length was then calculated (refer to Supplementary Fig. 3g).

Intensity measurements

MyoII intensity (Fig. 1b, b’), F-actin intensity (Fig. 1b, b’ right panel), and dextran intensity (Supplementary Fig. 1b) were measured on cross-section embryo images using the point-picker plugin in ImageJ, which calculates the local average intensity within a 10 × 10 pixel region of interest (ROI). MyoII and F-actin intensities, measured at the basal side of the cell, were normalized using a standard curve derived from dextran intensity measurements at the same depths. These dextran measurements were performed in ventral, lateral, and dorsal tissues, following the basal cortex using MyoII::GFP localization as a guide until the basal cortex reached a depth of 35 µm. To assess comparability of MyoII intensity after Cry2-RhoGEF2 activation at 60° and 90°, intensity ratios relative to control were measured using background-corrected data after activating a Cry2-RhoGEF2 background embryo at 60° (VL) and 90° (L) from the ventral midline (Supplementary Fig. 4i). Imaging was performed using 44× magnification on the Luxendo MuVi SPIM microscope.

Statistics

A Mann–Whitney U test was performed between groups. A one-way ANOVA was performed in case of multiple groups, followed by post-hoc multiple comparison test. p values are provided in the figures or in the legends. A value p < 0.05 is considered statistically significant and marked by ‘*’, p < 0.01 is marked by ‘**’, p < 0.001 is marked by ‘***’ and p < 0.0001 is marked by '****'. All statistical analyses were conducted in GraphPad PRISM 7.

Reporting summary

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

Author contributions

A.B.T. and M.R. planned the experiments. A.B.T. performed the experiments, data processing, and quantification. A.B.T. and M.R. analyzed the data. A.J. measured the spatio-temporal relaxation rates. A.J. and J.E. developed the physical theory. M.R. and A.B.T. wrote the manuscript with the support of A.J. and J.E.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

A.B.T.’s PhD scholarship was supported by the Region SUD PACA Research Program. This work was supported by the Region SUD PACA Research Program, the ANR PRC program (ANR-22-CE13-0024), project PPIA 4D-OMICS [21-ESRE-0052], and the financial support from Inserm to the Booster Program Mecacell3D.

Data availability

The data supporting the findings of this study and materials are available on request from the corresponding author (M.R.).

Code availability

The custom Python code used for the analyses in this study is publicly available at https://doi.org/10.5281/zenodo.21723698.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77010-4.