Fig 1.
Experimental observation of gastruloids and shape quantification using LOCO-EFA.
A: Representative images of fixed gastruloids after 48 h, 72 h and 96 h. Cell nuclei were stained with DAPI. Shown is the mid-plane of a z-stack. Scale bar: 100 μm (48 h and 72 h), 200 μm (96 h). B: Image of a gastruloid with a binary mask created from that image. Numerical values of L2/L1 and L3/L1 for this gastruloid are given. Scale bar: 100 μm. C: LOCO-EFA analysis of measured gastruloids at 72 h (n = 132, two biological replicates) and 96 h (n = 413, 8 biological replicates). Each data point is a gastruloid. The scatter plot shows the scaled LOCO-EFA coefficients L2/L1 and L3/L1. The 3 insets show images of the gastruloids highlighted in the LOCO-EFA scatter plot. Cell nuclei were stained with DAPI. Shown is the mid-plane of a z-stack. For each gastruloid, the scaled LOCO-EFA coefficients L2/L1 and L3/L1 are given. Scale bar: 100 μm.
Fig 2.
Differential adhesion cannot explain the shape distribution of gastruloids.
A: Wholemount immunostaining of Brachyury (magenta) in 72 h (left) and 96 h (right) gastruloids. Shown is a maximum z-projection. Cell nuclei were stained with DAPI (grey). Scale bars: 200 μm. B: Example shapes resulting from a simulation of differential adhesion for 120, 30,000, 60,000, 90,000 and 120,000 Monte Carlo steps. Different cell types / adhesion strengths are indicated by color. C,D: Quantification of shapes resulting from simulations of differential adhesion. Shown are LOCO-EFA coefficients averaged over a population of shapes. The shaded areas (in the experimental data) and vertical bars (in the simulation data) indicate standard deviations. C: The surface tension γ(c, M) was varied. The slope was kept constant at a value corresponding to a tension γ(τ, τ+ 1) between two adjacent cell types of 2.5. D: The slope of the adhesion gradient was varied. The offset was kept constant at 25. L2/L1 increases with increasing slope but L3/L1 remains approximately constant. E: Scatter plot of the scaled LOCO-EFA coefficients L3/L1 versus L2/L1 for the experimental data and simulations of differential adhesion with offset 25 and slope 2.5.
Fig 3.
Simulations including crawling-driven CE reproduce average LOCO-EFA coefficients.
A: Schematic of the filopodial-tension model with self-organized preferred direction. Green and blue arrows indicate the cells’ polarization directions. The highlighted cell s can extend filopodia within the shaded blue cone around its polarization direction. Such filopodia are currently attached to cells s1, s2 and s3, indicated by the link between the cell centers. The new polarization direction of cell s is the weighted average of its old polarization direction and the average polarization direction of cells s1, s2 and s3. The rightmost panel shows the new polarization direction of cell s as a blue arrow. The old polarization direction is indicated by a pink arrow. B: Example shapes resulting from a simulation with one cell type for 100, 25,000, 50,000, 75,000 and 100,000 Monte Carlo steps. C,D: Quantification of shapes resulting from simulations of crawling-driven CE. Shown are LOCO-EFA coefficients averaged over a population of shapes. The shaded areas (in the experimental data) and vertical bars (in the simulation data) indicate standard deviations. C: The interaction strength with the medium was varied. The pulling force was kept constant at 15. D: The pulling force was varied. The surface tension with the medium was kept constant at 5. E: Scatter plot of the scaled LOCO-EFA coefficients L3/L1 versus L2/L1 for the experimental data and simulations of crawling-driven CE with pulling force λF = 15 and surface tension γ(c, M) = 5. These parameters gave the smallest 2D Kolmogorov-Smirnov test statistic comparing experiment and simulation (D = 0.40, p = 9.3 ⋅ 10−9). F: Example of experimentally observed shape (bottom) with highly similar simulated shape (top) and their corresponding scaled LOCO-EFA coefficients L3/L1 and L2/L1. Bottom: a 96 h fixed gastruloid. Shown is the mid-plane of a z-stack. Cell nuclei were stained with DAPI. Scale bar: 200 μm.
Table 1.
Default parameters used in the simulations.
Table 2.
Example of edge list arrays.
Fig 4.
Simulations of crawling-driven CE with two cell types create more extreme shapes.
A: Example shapes resulting from a simulation of same type pulling for 100, 25,000, 50,000, 75,000 and 100,000 Monte Carlo steps. B,C: Quantification of shapes resulting from simulations of CE driven by crawling and differential adhesion with two cell types. Shown are LOCO-EFA coefficients averaged over a population of shapes. The shaded areas (in the experimental data) and vertical bars (in the simulation data) indicate standard deviations. B: The surface tension between the two cell types was varied and the pulling force was kept constant at 20. C: The pulling force was varied. The surface tension between the two cell types was kept constant at -4. D: Scatter plot of the scaled LOCO-EFA coefficients L3/L1 versus L2/L1 for the experimental data and simulations of crawling-driven CE with pulling force λF = 20 and surface tension γ(1, 2) = −4. These parameters gave the smallest 2D Kolmogorov-Smirnov test statistic D = 0.15 and the largest p-value p = 0.15. E: Example of experimentally observed shape (bottom) with highly similar simulated shape (top) and their corresponding scaled LOCO-EFA coefficients L3/L1 and L2/L1. Bottom: a 96 h fixed gastruloid. Shown is the mid-plane of a z-stack. Cell nuclei were stained with DAPI. Scale bar: 200 μm.
Fig 5.
Time lapse imaging of gastruloids reveals hallmarks of CE.
A: Maximum z-projection of a gastruloid after 91 h (left) and 96 h (right) of differentiation. Cell trajectories are overlayed. The outline of the gastruloid is indicated by a yellow dashed line. Approximately 1 in 16 of the cells expressed mCherry-GPI, which fluorescently labeled the membrane. 14 of these cells were traced over the duration of the experiment, which had a total of 31 time points (10 min intervals). Circles with a white core show the position of the cells at 91 h. Diamond shapes show the position of the cells at 96 h. Dividing cells are indicated by a circle with a dark core. For each dividing cell, both daughter cells were traced. The inset at 91 h shows a zoom-in on cell number 9 that was traced. Scale bars: 200 μm, and 10 μm for the inset. B: Deformation of shapes defined by the position of selected cells. Scale bars: 200 μm. C: Stills of a simulation of a mosaic gastruloid at 20,000 Monte Carlo steps (left) and 100,000 Monte Carlo steps (right), showing 20 cells in grey and 180 cells in white. Trajectories of 12 grey cells shown at 17 time points are overlayed. Trajectories were corrected for the gastruloid’s rotation and drift. Trajectories start at 20,000 Monte Carlo steps (circles) and end at 100,000 Monte Carlo steps (diamond shapes). D: Scheme showing how cell positions were projected on the long axis of a measured or simulated gastruloid. Positions of cells were rotated to have the new x coordinate correspond to the position on the long axis and the new y coordinate to the position on the short axis. The movement along the long axis vl or short axis vs was measured for each cell and normalized to the total length of the gastruloid L. E: Distance moved versus end position along the long axis for simulated and measured cell trajectories. Large, pink circles represent cells of the in vitro gastruloid. Small, black circles represent cells of the simulated gastruloid. F: Net movement (final position—initial position) for each cell plotted as a vector on top of each cell center at the final Monte Carlo step. The net movement was corrected for rotation and drift as described in Materials and Methods. For visualization, the size of the vector was divided by a factor of 20 to avoid overlapping arrows. Arrows are colored according to their direction using a cyclic color map.
Fig 6.
Inhibition of the ROCK pathway prevents elongation but not cell type segregation.
A: Fixed gastruloids at 96 h treated with (left) or without (right) 10 μM of the ROCK inhibitor Y-27632 during the final 24 h of differentiation. Shown is a maximum z-projection. Cell nuclei were stained with DAPI. Scale bars: 200 μm. B: Wholemount immunostaining of Brachyury/T and Sox2 in fixed gastruloids at 96 h treated with (top) or without (bottom) 10 μM Y-27632. Cell nuclei were stained with DAPI. Scale bars: 200 μm. C: Wholemount immunostaining of Brachyury/T and Cer1 in fixed gastruloids at 96 h treated with (top) or without (bottom) 10 μM Y-27632. Cell nuclei were stained with DAPI. Scale bars: 200 μm. D: LOCO-EFA analysis of measured gastruloids at 96 h treated with (n = 110) or without (n = 106) 10 μM Y-27632. The results of two biological replicates were combined. Each data point is a gastruloid. The scatter plot shows the scaled LOCO-EFA coefficients L2/L1 and L3/L1.