A tensile ring drives tissue flows to shape the gastrulating amniote embryo

See allHide authors and affiliations

Science  24 Jan 2020:
Vol. 367, Issue 6476, pp. 453-458
DOI: 10.1126/science.aaw1965

Shaping the early amniote embryo

Gastrulation is an essential step in development in which the internal tissues of the body are set apart. In birds and mammals, a similar cascade of molecular events is known to specify embryonic territories, but how they are physically remodeled has remained elusive. Working with avian embryos, Saadaoui et al. identified a cable that encircles the embryo as the engine of gastrulation and described the collective cell movements as similar to the motion of a fluid. One side of this contractile ring pulls more strongly than the other, entraining the large-scale tissue movements that shape the early body plan. The embryo margin, previously known to function in molecular regulation, thus emerges as a dual mechanical and molecular organizer of development.

Science, this issue p. 453


Tissue morphogenesis is driven by local cellular deformations that are powered by contractile actomyosin networks. How localized forces are transmitted across tissues to shape them at a mesoscopic scale is still unclear. Analyzing gastrulation in entire avian embryos, we show that it is driven by the graded contraction of a large-scale supracellular actomyosin ring at the margin between the embryonic and extraembryonic territories. The propagation of these forces is enabled by a fluid-like response of the epithelial embryonic disk, which depends on cell division. A simple model of fluid motion entrained by a tensile ring quantitatively captures the vortex-like “polonaise” movements that accompany the formation of the primitive streak. The geometry of the early embryo thus arises from the transmission of active forces generated along its boundary.

During amniote gastrulation, endodermal and mesodermal derivatives internalize through the primitive streak, a transient structure at the midline of the early embryo. In avians, the primitive streak forms from an initially crescent-shaped region at the margin between the embryo proper (EP) and extra-embryonic tissue (EE) (Fig. 1A), which converges toward and extends along the midline. Although myosin-II–driven oriented cell intercalation is known to underlie convergent extension of the prospective primitive streak (1, 2), how the concomitant vortex-like tissue flows arise (3, 4) and how they relate to the formation of the primitive streak has remained elusive.

Fig. 1

Quantitative description of gastrulation movements. (A) Early epiblast. (B to D) Trajectories [(B), t = 4 to 6 hours] and deformation of an initially square grid [(C) and (D)] from the PIV analysis of a memGFP embryo movie. Colors in (C) and (D) show area changes between the initial [(C), t = 0] and final [(D), t = 10 hours] configurations. (E to H) Automated fate mapping (green, EP; magenta, primitive streak); dots show winding motion (arrows) along the margin, quantified in (G) by the time evolution of angular positions [dotted lines in (E) and (F); 0° is posterior]. (H) shows the area of tissue regions versus time (n = 6 embryos; bold lines are averages). (I to L) Decomposition of the tissue-velocity field into divergent and rotational components [(I) and (J)] and contributions to motion along the margin [(K) and (L); colors are as in (I)]. Percentages quantify shaded areas (averages over n = 6 embryos and the indicated time intervals). t0, time of motion onset. Scale bars, 1 mm.

To analyze gastrulation movements, transgenic quail embryos expressing a membrane-bound green fluorescent protein (memGFP) (5) were cultured ex vivo (6) and imaged in their entirety for 12 hours. The resulting movies were processed using particle image velocimetry (PIV) to reconstruct cell trajectories and tissue deformation maps (Fig. 1, B to D, and movie S1). Embryonic territories, originally characterized using anatomical or molecular criteria, could be recognized in these maps (compare Fig. 1, A and C). We designed automated fate-mapping methods that identify and track these territories on the sole basis of tissue movement (Fig. 1, E and F; fig. S1; and supplementary text). To validate this approach, we checked that the inferred location of the embryo margin aligned with the boundary of expression of the ectodermal/EP marker Sox3 in embryos that were fixed after live imaging (fig. S1, K and L). The angular motion of points along the margin, which wind around the EP as they converge to the posterior, captured the progress of gastrulation (Fig. 1, E to G). On the basis of these landmarks, we registered movies of six embryos in space and time to construct an average embryo (movie S2, fig. S1, and supplementary text), which was then used as a reference. The development of cultured embryos was virtually indistinguishable from an embryo imaged directly in the egg (fig. S2 and movie S3).

Noting that the EP maintains an approximately constant area, whereas the EE tissue steadily expands (Fig. 1H), we sought to distinguish area changes from other contributions to tissue movement. A decomposition into divergent (area changes) and rotational (incompressible) components indicated that gastrulation movements can be understood as the sum of three simpler flows: (i) a radial, outward movement of the expanding EE tissue; (ii) an area-preserving flow with two vortices within the EP; and (iii) at later stages, inward movement driven by areal contraction along the streak (Fig. 1, I and J, and movie S4). Although large-scale flows in the epiblast have been proposed to passively ensue from the deformation of the mesendoderm (1, 2, 7), we found that rotational movement persists after the mesendodermal crescent has converged onto the midline (Fig. 1, I and J) and that areal contraction makes a limited contribution to continued movement toward the streak (Fig. 1, K and L), suggesting that other forces must be at play.

For a viscous fluid that is described by the Stokes equations, these forces could be derived from the Laplacian of the velocity field (see the supplementary text). When applied to tissue flows in the epiblast, this suggested a pattern of tangential forces along the embryo margin extending well into its anterior half (Fig. 2A and movie S4). When flow is driven by active internal stresses, here by cell contractility, forces inferred in this way should be understood as apparent external forces arising from the spatial variations of the active stress. The force pattern of Fig. 2A thus pointed to a tensile margin around the EP, with a tissue-scale tension that decays from posterior to anterior: Tissue along the sides of the EP is drawn toward the posterior, where tension is higher; tissue in the posterior is thrust forward because the margin is curved (Fig. 2B). To test this hypothesis, we formulated a fluid-mechanical model that is based on the Stokes equations, with source terms for nonuniform area changes and for active tensile stresses along the margin (Fig. 2, C and D, and materials and methods). Area changes are taken from the experiment, whereas tensions along the margin, which moves with the tissue, are fit to the observed motion at each time step (Fig. 2C). Although strongly constrained—aside from the tensions, the initial position of the margin and its width are the only free parameters—the model recapitulates the full course of tissue movements over 8 hours (movie S5), with >90% accuracy for the average reference embryo (Fig. 2, F and G, and materials and methods). Considering the deformations that would result from each source term taken separately (Fig. 2E), active tensions largely account for the shaping of the embryo, whereas area changes are mostly responsible for EE expansion. As further abstractions, a “synthetic embryo” in which the source terms are replaced by simple mathematical functions of space and time (fig. S3 and movie S6; see also supplementary text, table S1, and fig. S4) is sufficient to quantitatively capture the movement of the tissue, and its essential features can be recovered analytically in the limit of a thin margin (fig. S5 and supplementary text).

Fig. 2

Quantitative fluid-mechanical model for gastrulation. (A) Apparent forces (negative of the Laplacian of the velocity field; averages over n = 6 embryos; magenta, presumptive primitive streak). (B) Diagram illustrating how apparent forces (black) arise from graded tensions along the margin (magenta). (C to E) Quantitative model for gastrulation movements. (C) Tension/viscosity profiles (millimeters per hour; averages over 1-hour intervals) from a fit to the reference average embryo (n = 6 embryos). (D and E) Tissue flows in the model as the resultant of area changes (taken from experiment) plus an incompressible flow driven by tension along the margin [magenta line in (D)]. (D) Velocity fields (colors as in Fig. 1I). (E) Deformation maps from each source term taken separately and together. (F and G) Velocity field (F) and deformation map (G) for the average embryo (right panels show deviation between model and experiment). Scale bars, 1 mm.

Our results suggest that gastrulation is best understood as a tissue-wide process, arising from a tensile margin and a fluid-like response of the epithelial epiblast. We designed additional experiments to challenge this description. First, to detect a tensile margin, we performed circular ultraviolet laser cuts (8) at different locations in the epiblast (Fig. 3, A to C, and movie S7). Cuts along the margin revealed anisotropic tissue strains. Cuts inside the EP showed significantly lower strains, which can be ascribed to passive tissue deformation (see the supplementary text). As a control, linear cuts at the same locations showed a strong correlation between the final opening width (i.e., tissue strain) and the initial opening velocity (a more direct correlate of tissue stress; supplementary text and fig. S7). These experiments further revealed that tension runs all the way to the anterior margin. Although this could not be inferred from motion alone (motion in the model depends essentially on differences in tension), allowing for tension in the anterior recovers patterns of shear stress that agree with the observed tissue strains (supplementary text and fig. S20B). To connect tissue-scale motion and cellular-scale behaviors, we analyzed embryos that were fixed after live imaging for different time intervals (Fig. 3, D to O, and figs. S8 to S13). Cell segmentation of entire embryos showed a gradual increase in cell areas in the EE tissue, likely contributing to its expansion (Fig. 3, E, I, K, and O). Cell shapes, which initially had an isotropic distribution, became elongated along the margin, consistent with a state of tension (Fig. 3, F, I, L, and O). Quantification of junctional phosphorylated myosin II revealed localized myosin anisotropy, a correlate of active force generation (9), at the margin (Fig. 3, G to I and M to O). The location of this large-scale supracellular ring aligned with the location of the embryo margin determined from the motion of the tissue before fixation, and its width agreed with that inferred from the model (fig. S13, supplementary text, and fig. S6). High-resolution live imaging of transgenic embryos expressing a tdTomato-myosin II reporter revealed the progressive formation of dynamic, tangential actomyosin supracellular cables spanning five to 20 cells at the margin (fig. S14A and movie S8), coincident with the site of apparent forces inferred from tissue motion (fig. S15). In the posterior margin, these cables contracted, driving oriented intercalations (2). In the anterior, supracellular cables were also visible but extended tangentially, concomitant with cell elongation and oriented divisions (fig. S14, B and C, and movie S9)—indicative of stress dissipation (10). Thus, the margin exerts active tension in the posterior and passive tension in the anterior.

Fig. 3

Mechanical, cellular, and molecular characterization of the embryo margin. (A to C) 250-μm circular laser cuts in a single memGFP embryo (A) and representation of all laser cut experiments (B); ellipses show anisotropic strain amplified fourfold for visibility. (C) Tangential versus radial strain anisotropy (bars, mean±SE; **P < 0.01, ****P < 0.0001, paired t test). Red indicates the posterior margin, green the anterior margin, and blue the EP. (D to O) Apparent forces [(D) and (J)] in embryos imaged until the indicated stages (dashed line, margin) and cell areas [(E) and (K)], cell shape anisotropy [(F) and (L)], and junctional phosphorylated myosin anisotropy [tangential (green) versus radial (magenta) in (G) and (M) and averaged in 100 × 100 μm boxes in (H) and (N)] mapped in these embryos and averaged across embryos [(I) and (O); mean ± SE of radial profiles in n = 2 and n = 3 embryos]. Boxed regions are shown in figs. S8 to S11 and individual embryos from (I) and (O) in figs. S12 and S13. Green curve in (N) is the Gaussian fit with standard deviation w, as indicated. Scale bars, 1 mm.

Second, we sought to identify the cellular basis of tissue fluidity. Because cell rearrangements contribute to stress relaxation in epithelial tissues (11), and most cell rearrangements in the early avian embryo are associated with cell division (12), we reasoned that cell division may be required for a fluid-like behavior. Treatment with hydroxyurea (HU) efficiently suppressed cell division (fig. S16A) but induced apoptosis in the long term (fig. S16B). When HU was combined with the apoptosis inhibitor Q-VD-OPh, both cell division and apoptosis were suppressed (fig. S16, A, C, and D) and the topology of the epithelium was greatly stabilized compared with controls (fig. S16, E and F, and movie S10). Whereas Q-VD-OPh alone only slightly delayed the progress of gastrulation (figs. S1J and S17, A to E), embryos incubated in both HU and Q-VD-OPh showed a marked slowdown by 6 to 8 hours of treatment and failed to form a primitive streak (n = 6; Fig. 4, A and B; fig. S1J; fig. S17, D and F; and movie S11). Tissue expansion persisted, but rotational movements were abolished, consistent with a suppression of fluidity (fig. S17G and movie S12). As a control, treatment with aphidicolin and Q-VD-OPh produced quantitatively similar results (fig. S17, H to J). The model suggested that the amplitude of the tension/viscosity ratio dropped over time (Fig. 4C). Laser cuts in embryos treated with HU and Q-VD-OPh revealed that tensions were still present, if not increased, along the margin (fig. S17, K to M, and movie S13), and a supracellular ring was still observed in fixed embryos (Fig. 4D and fig. S18), implying that the slowdown resulted from an increase in viscosity and not from a decrease in tension. Thus, fluidity of the embryonic epithelium is required for primitive streak formation and emerges from cell division.

Fig. 4

Manipulation of tissue viscosity and hydrodynamic effects in gastrulation. (A to D) Effect of HU+Q-VD-OPh on gastrulation movements (n = 5 embryos). (A) Trajectories (individual embryo, t = 4 to 6 hours). (B) Time evolution of angular positions along the margin (average embryo). (C) Tension/viscosity profiles from model fit to average embryo. (D) Junctional phosphorylated myosin anisotropy (compare Fig. 3M). (E to H) Predictions from the synthetic model [(E) and (G)] and experimental response to centered (F) and off-centered (H) cuts generating a new tissue border. Scale bars, 1 mm.

Third, we challenged model predictions for hydrodynamic effects in gastrulation. The embryo, which draws the surrounding tissue to the posterior, is akin to a swimmer and should move forward over time. Indeed, embryos exhibited a slow anterior-ward movement, in quantitative agreement with the model (fig. S19). At odds with the view that vortex-like flows are shaped by a confining boundary (7, 13, 14), our model suggests that they are governed by the distribution of active forces, with boundary conditions playing a limited role in the intact epiblast. The progress of gastrulation is predicted to be weakly sensitive to the distance to epiblast border, reaching >80% of its maximum rate when it is just 50% larger in radius than the EP (fig. S5D and supplementary text). Indeed, circular cuts centered on the margin, which removed most of the EE tissue and brought the epiblast border closer to the EP, had almost no effect on tissue flow and streak formation (Fig. 4, E and F, and movie S14). By contrast, in the case of off-centered cuts, which bring the border even closer to one side of the EP, the model predicted, and experiments confirmed, that the interaction between the EP and the border induces a rotation of the axis, leaving only one apparent vortex and resulting in a bent streak (Fig. 4, G and H, and movie S14).

Our study demonstrates the power of fluid-mechanical approaches (9, 15, 16) to capture large-scale morphogenetic movements and identifies a simple mechanical basis for gastrulation. Although tissue-wide flows in the embryonic disk were previously interpreted as being a passive consequence of primitive streak formation (1, 2, 7), we find instead that both are part of a broader process that is driven by tensile forces all along the margin and shapes the embryo as a whole (see the supplementary text and fig. S20 for further discussion of alternative models). Supracellular actomyosin cables, which have been shown to drive local cell rearrangements, stabilize compartment boundaries, and act as purse strings in wound healing and embryonic tissue closure (17, 18), effect large-scale remodeling of the surrounding tissue through nonuniform contraction. Our finding that the embryo margin, which was previously identified as a molecular organizer of early development (1921), is also defined by a specific mechanical state and cellular behaviors hints that mechanical and molecular cues may combine in the establishment of the amniote body plan.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S20

Table S1

References (2238)

Movies S1 to S14

References and Notes

Acknowledgments: We thank V. Hakim, P.-F. Lenne, A. Martinez Arias, and F. Schweisguth for critical reading of the manuscript and P. Caldarelli for the in situ hybridization shown in fig. S1, K and L. Funding: The research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC Grant Agreement no. 337635, Institut Pasteur, CNRS, Cercle FSER, Fondation pour la Recherche Medicale, and the Vallee Foundation. Author contributions: M.S., F.C., and J.G. conceived the study. D.R. and J.R. generated transgenic lines. M.S. performed experiments. F.C. developed quantitative analysis methods and theoretical models. F.C. and J.G. wrote the paper with input from M.S. Competing interests: The authors declare no competing interests. Data and materials availability: All experimental data are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article