(A-C) H2B- EGFP blastocysts stained with phalloidin (red) after 8-9h of control culture (A), or nocodazole (B) or cytochalasin D (C) treatment. Disruption of microtubules with nocodazole arrests cells in metaphase (arrowheads). Disruption of actin with CCD inhibits cytokinesis (arrows) and collapses the blastocoel. (D, G) In a typical imaging session, four cells have been tracked to show the persistence (D) or inhibition (G) of cell movement. (E, H) In GAP-RFP⁺ cells, protrusions (arrowheads) persist (C) or disappear (H) after drug treatment. (F) Velocities of GAP-RFP⁺ cells in control, nocodazole and CCD treated embryos. The mean value is shown for 9 (nocodazole, CCD) or 13 (GAP-RFP) cells, and the standard deviation as error bars. The mean cell velocity is significantly decreased after CCD treatment compared to control (nested ANOVA). Scale bar 20μm.

(A) Schematic representation of the constructs used. WDN is a dominant-negative of Wnt9A, deleted for the C-terminal cysteine-rich domain. CDN is a dominant-negative of Nr2f1 (also known as COUP-TFI), with a point mutation in the DNA binding domain (orange). GDN is a dominant-negative of Gata6 in which the Gata6 DNA binding domain is fused to the engrailed (En) inhibitory transactivator domain. (B-D) Control embryos (8-cell stage) cultured for a day after single cell injection of RFP (red) at the 2-cell-stage. After co-injection with CDN (B) or Gata6 (C) mRNA, immunostaining of Nr2f1 (B) or Gata6 (C) is detectable in the nuclei of RFP-positive cells, as expected. Three- (B1-C1) or two- (B2-C2) channel merged pictures. Scale bar 50μm. β-catenin localisation (D) in uninjected, WDN and Wnt9a injected embryos (8-cell stage). (E) The ratio between nuclear and cytoplasmic β-catenin (n=8 cells) is significantly decreased and increased after WDN and Wnt9A injection respectively (T- test). (F-G) Cells were injected at the blastocyst stage with RFP mRNA only (control) or together with one of the constructs shown in (A). Analysis of the fate of surface (F) and deeper (G) injected ICM cells, as a percentage of daughter cells observed at the end of the track at the surface and in the deeper compartment respectively. (*) The fate of GDN injected cells (Fisher test) and that of gata+Wnt injected cells are highly significantly different from the control. The numbers of daughter cells analysed are: 60 (surface control), 13 (CDN), 33 (GDN), 7 (WDN), 29 (deep control), 33 (gata), 6 (Wnt), 44 (gata+Wnt).

(A) Trajectories of ICM cells in an H2B-EGFP blastocyst. Each lineage is coloured and numbered. Time from the start of imaging is indicated. The position of the blastocoel is highlighted by a dashed white line. Scale bar 50μm. (B) In a typical tracked blastocyst, the cumulative distance travelled by each cell over time is represented, using a colour per cell. (C) Mitosis (red cell) results in movement of neighbouring cells (in green with its trajectory). An example of metaphase and anaphase are shown in C1 and C2 respectively. (D) Injection of membrane-localised GAP-RFP (red) reveals protrusions (arrowheads) of moving ICM cells. (E) The orientation of 163 protrusions from 14 cells was calculated in relation to the trajectory of the cell (x axis). The frequency of observations of the angular sectors is represented by the length of the sector. (F) Comparative trajectories of the centre of gravity of the cell (red) and that of its nucleus (green). On the left, the embryo is shown at the start of imaging. In the middle, the cell (pale red), its nucleus (pale green) and their respective parallel trajectories are represented at the scale of the embryo (dashed black line). On the right, a close-up is shown, with a scale bar of 10μm.

(A) The GAP-RFP⁺ surface cell kept its position throughout the movie. b, blastocoel. (B) The RFP⁺ surface cell reached the deeper layers of the ICM by active cell movement. After division, it gave rise to one deeper daughter cell, which died (apoptotic debris, asterisk), whereas the other daughter cell moved back to the surface and divided. (C-D) The RFP⁺ deeper cell divided to give rise to a daughter cell (c) that remained deeper (C). The other daughter cell (d) moved to the surface (D). Two distinct z planes shown in C and D focus on each sister cell. Scale bar 25μm. (E) The blastocoel was traced in each z-plane of the transmission channel (left panels) ; its contour is shown in green in the green-channel images (right panels). (F) The shortest distance of each nucleus to the cavity is plotted over time. For each lineage, the initial cell is represented in red, the first generation in orange and brown and the second in dark and light blue and dark and light green. There was a general trend towards a decrease of the distance to the cavity, as the cavity expanded and the ICM was more compressed. A dotted black line estimates the limit between the two compartments over time. asterisks : asymmetric divisions leading to the separation of sister cells in two different compartments ; black crosses : apoptoses.

(A-D) The number of cells and score of each model are shown as the sum of 7 independent simulations. The number of cells with a given gene expression state (PE markers- red, epiblast markers- blue, indeterminate- yellow) is represented over time. straight line : sorted cells (with congruent gene expression state and position); dashed line : mispositioned cells. (A) Random model (equal probability). (B) Cell sorting model. If the gene expression state is congruent with the position of a cell, the probability that the cell will move to the alternative position is lower than 0.5. Conversely, if cells are mispositioned, the probabilities of movement and apoptosis are increased. (C) Induction model. If the gene expression state is congruent with the position of a cell it is less likely to change (i.e. stable gene expression). (D) In the absence of selective apoptosis of mispositioned cells the score of the Cell sorting model is significantly improved (D1 compared to B, p<0.001 Student test), but not that of the Cell sorting + Induction model (D2 compared to C2). (E) Theoretical models of PE formation. (a) Position-based model: cells in contact with the blastocoel cavity receive inducing signals, and hence differentiate into PE as a result of position. (b) ‘Salt and Pepper’ model: cells sort according to their gene expression patterns. (c) New model of PE formation, emphasising the importance of both position and gene expression.