Data processing pipeline for superpixel motion-based analysis of VE tissue morphodynamics.

(A) Diagram of E5.5 mouse embryo; Epi = epiblast, ExE = extra-embryonic ectoderm, EPC = ectoplacental cone, emVE = embryonic visceral endoderm, exVE = extra-embryonic visceral endoderm, DVE = distal visceral endoderm. (A’) Lateral view of example E5.5 Hex-GFP:mem-tdTomato embryo at two time-points, imaged using a ZEISS Z.1 light-sheet at a 10-minute interval. MIP = max intensity projection. Scale bar = 50 μm. (B) Polar geodesic projection of the apical VE surface of embryo in A’ . Re-projections enable visualisation of entire VE and the application of sophisticated motion analysis tools to analyse time-lapse data. Dotted line shows position of emVE-exVE boundary – the limit of DVE cell migration. Blue cross = distal tip. (C) Overview of VE motion-tracking pipeline for quantitative analysis of time-lapse data-set. (D’-D’’) Automatic staging of DVE migration from superpixel-based motion-tracking. (D) VE superpixel motion tracking on VE polar projection; red arrows show motion-flow vectors. DVE-core motion highlighted in green. Dotted line shows mean consensus angle vector of DVE which is used to align embryos along their A-P axis. The persistence of motion direction ddirect of DVE was separated from the surrounding VE. (D’’) The motion persistence of DVE showed three phases; 1) pre-migration/initiation, 2) migration to boundary, 3) post-boundary, as shown in the max intensity projections. (D’’’) Example embryo showing last timepoint of each phase. (E) A flow-like pattern of motion across the VE was observed by superpixel motion tracking. Colours denote mean vector angle. (F) To track sub-regions of the VE, a deformable grid was seeded at the start of each phase. By tracking the motion within each sub-region the grid becomes deformed and multiple parameters are analysed. (F’) To quantify parameters and to combine data from across embryos, 2D tracking data is converted back to 3D coordinate-space for analysis which can be summarised in a multi-embryo polar plot. Note re-projections have no scale bar as they are non-linear.

Morphological asymmetry in the ectoplacental cone predicts the position of the A-P axis, but not its polarity.

(A) Diagram showing EPC in relation to the anterior-posterior (A-P) axis. It is unclear whether the tilt (morphological asymmetry) of the EPC aligns with the A-P axis even prior to DVE migration. (B) Example E5.5 embryos imaged by a light-sheet microscope, with examples of EPC-tilted to the anterior-left and posterior right of the A-P axis, showing that DVE cells can migrate away or towards the same side as the EPC. (C) Radial plot showing direction of EPC in relation to the A-P axis (black line = median line). Not all embryos showed a clear tilt in their EPC (6/20). In embryos with a morphologically tilted EPC there was an even split between anterior left or posterior right with an average offset of 51°. This differed significantly from a random distribution (χ2 test for deviation from expected uniform probabilities; using 90º sectors, p=<0.05) suggesting that the EPC predicts the axis, but not its polarity (i.e., direction of DVE migration).

Multi-embryo tissue morphodynamic anaylsis reveals distinct phases in DVE migration.

(A) Multi-embryo heat-maps of VE tissue motion behaviour showing average behaviour of 32 sub-regions of the DVE of anteriorwards velocity, total speed (in any direction), curl strain fraction per hour, and surface area change. Arrows show median motion vector of superpixels in each region. Arrow length is proportional to the speed value. Central ring is the distal most VE. Black ring denotes the emVE-exVE boundary. Data is converted to 3D coordinate-space for quantification. In motion tracking note the anterior direction of the DVE migration (arrows), and the bi-rotational flow of the surrounding emVE during migration (curl strain fraction). This flow pattern continues into the post-boundary phase where the nascent DVE cells continue to move anteriorly from the distal tip. In surface area change, note that the anterior emVE undergoes a decrease in surface area (orange and red colour) during the migration phase. This decrease in area does not happen during the post-boundary phase. Note that in the ‘pre-migration/initiation’ phase, DVE cell show higher instantaneous anteriorward speed than surrounding regions, but movement is not directionally persistent at this stage. (B-B’’) The distal VE regions shows the highest anteriorwards velocity and total speed at migration stages (migration to boundary and post-boundary) (one-way ANOVA, p=<0.001, followed by Tukey’s HSD Test for specific comparisons). Interestingly, even in the pre-migration/initiation phase, when the DVE shows low directional persistence (see Figure 1D’), there is still significantly higher motion in the DVE cell population compared to surrounding cells and the exVE (one-way ANOVA, p=<0.001, followed by Tukey’s HSD Test for specific comparisons).

DVE cells collectively migrate in ratchet-like manner.

(A) Overview of cell movement analysis in cell tracked Lifeact-GFP time-lapse data-set, imaged at a 5-minute interval. Each cell vector angle is assessed then weighted by its velocity. (A’) Example embryo showing DVE cell tracks (cyan) on 2D polar geodesic projection. Red dotted line denotes emVE-exVE boundary. (A’’) Heat-map of weighted angle histogram of DVE cell velocity where each cell’s contribution is weighted by its speed. Cell vectors are in 22.5° bins for the +/-1 hour from onset of collective migration. Note the increase in coordinated anteriorward motion at the onset of the migration phase (white dashed line). (B) Graph showing Rayleigh distribution statistic of DVE cell tracks at pre-migration/initiation and migration phases, which shows the increase in coordination in cell vectors at the onset of migration (0 = random distribution, 1 = uniform distribution). (C) 2D geodesic polar projection of an example E5.5 Hex-GFP:mem-tdTomato embryo imaged every 10 minutes for 11 hours. Digitally extracting the A-P midline (coloured vertical line in C) throughout the time-lapse enables a kymograph to be plotted. (C’) Kymograph of the Hex-GFP channel along A-P mid-line of the embryo in C, showing the DVE cell population migrating collectively at the onset of the migration phase (vertical dotted line) to reach emVE-exVE boundary (horizontal dotted line). (C’’) Kymograph of superpixel motion tracking of A-P mid-line of embryo in C. Plotting anteriorwards velocity shows that the DVE movement is intermittent, undergoing start-stop behaviour (arrows). (D-D’) Graphs showing anteriorwards velocity and cumulative velocity (i.e., anteriorwards displacement/ min) of Hex-GFP population of embryo in C. In the pre-migration/initiation phase, where there is a relatively low level of directionally coordinated motion amongst DVE cells (A’’,B), there is little anteriorwards velocity. In the migration phase when directional coordination increased (A’’,B), anteriorwards velocity increases as the DVE migrated collectively (see position of Hex-GFP in D’’. t-5h - t0h). (E) Histogram of anteriorwards velocity at all timepoints from 9 Hex-GFP embryos during migration phase. In >90% of time-points Hex-GFP progresses anteriorly, with minimal posteriorwards motion. (E’) Plot showing mean time between start-stop events for each Hex-GFP embryo (n = 9); mean ratchet period was 1.13 ± 0.29 hours (mean ± stdev).

Vertex model shows that DVE start-stop motion is accompanied by T1 transitions in surrounding cells.

(A) Example of vertex model; DVE cells (grey) were given parameters corresponding to a solid-like cell cluster (a small target shape index) with finite self-propulsion force (F). emVE cells (blue) were modelled as passive (no self-propulsive force, F = 0) and were less rigid. Surrounding exVE cells (red) do not evolve in the simulation. (B-B’’) Parameter space of vertex model formed by varying DVE self-propulsive force (F) and rigidity of emVE cells (p0, target shape index), coloured by: (B) maximum DVE velocity, (B’) coefficient of variation of DVE velocity, and (B’’) cumulative T1 transition events (Methods). We note that only a specific region of parameter space gives rise to intermittent DVE movements. (C) Kymograph through the midline of the simulation in A, and line plots showing the position of the DVE centre-of-mass and velocity – confirming that start-stop behaviour occurs in the simulation. (D) Density-map of T1 events in example in silico simulation (top panel) and in vivo Lifeact-GFP embryo (bottom panel), showing that the majority of T1 events occur ahead of the migrating DVE (high density - black, low density - white). (E) Temporal analysis of DVE velocity and T1 events in an example in silico simulation where F = 0.01 and p0 = 3.6, showing peaks of DVE velocity (blue line) closely aligned with peaks of T1 events (orange line). Also see Movie S10 for an example vertex model simulation.

The anterior epiblast shows a distalwards motion during DVE migration.

(A) Diagram of E5.5 mTmG mouse embryo after Ttr-CRE mediated recombination to label the VE with membrane-GFP and the epiblast and ExE with membrane-tdTomato (magenta). (A’) Example of an E5.5 Ttr-CRE/mTmG embryo imaged by light-sheet microscopy showing a single z-plane (left) and a a max intensity projection (right). White dashed line shows emVE-exVE boundary. (B) Digital 2D geodesic polar projections of the apical VE (left) and basal surface (right) of the epiblast in the embryo in A’. White dashed line shows emVE-exVE and epiblast-ExE boundary, respectively. (C) Superpixel motion tracking of the VE shows the migration of the DVE and the bi-lateral pattern of motion in the VE. (D) Superpixel motion tracking of the epiblast during DVE migration shows a posteriorwards flow in the anterior epiblast (yellow-green tracks). Also, see Movies S11-S12 for examples of superpixel tracking. (E) Multi-embryo heat-maps of epiblast superpixel tracking during the 3 phases of DVE migration show a flow of motion across the epiblast even during pre-migration/initiation phase. (F) Manual tracking of sub-sets of VE (green tracks left panel; black dots denote final position), anterior epiblast (magenta tracks, right panel) and posterior epiblast (cyan tracks, right panel). Black cross denotes distal tip. (F’) 3D cell tracking of anterior epiblast and posterior epiblast from embryo in F, here shown with temporally-coloured tracks. Also, see Movie S13 for an example time-lapse with VE and epiblast cell tracking.

Midline A-P projection analysis shows retrograde motion behaviour.

(A) Overview of analysis method for digital extraction and linear re-projection of the anterior-posterior mid-line of light-sheet imaged embryos. (B) Example linear A-P re-projection of Lifeact-GFP embryo (upper panel) and corresponding superpixel motion-tracking (lower panel) showing the migration of the DVE to the left of the projection (purple tracks), while the epiblast moves to the right (yellow-orange colours). See Movie S14 for an example mid-sagittal re-projection. (C) Multi-embryo heat-maps of A-P linear projections showing the average motion behaviour of the epiblast and VE at 3 migration phases in the mid-sagittal tracking analysis.

Examples of light-sheet time-lapse data and geodesic projections of VE and epiblast tissues in three fluorescent reporter mouse lines.

(A) Example Hex-GFP; membrane td-tomato, (B) Lifeact-GFP, and (C) mTmG crossed with Ttr-Cre line. Each panel shows a max intensity projection, mid-sagittal optical section, and rectangular and polar projections of the apical surface of the VE (upper panels) and basal surface of the epiblast (lower panels). We note that the Lifeact-GFP mouse line consistently showed an almost complete absence of fluorescence signal in the ExE tissue. This does not reflect the level of F-actin in the ExE as shown by immunofluorescent staining. Note projections have no scale bar as they are non-linear.

Superpixel motion tracking of the VE tissue on geodesic projections from three fluorescent mouse lines.

(A) Example Hex-GFP; membrane td-tomato, (B) Lifeact-GFP, and (C) mTmG crossed with Ttr-Cre line. Each panel shows superpixel tracks on a rectangular geodesic projection and polar geodesic projections of the apical surface of the VE. Tracks are coloured by the direction of motion. Note re-projections have no scale bar as they are non-linear.

Sub-regional tracking of the VE tissue using MOSES superpixel motion tracking.

(A) Overview of method for tracking the motion of VE sub-regions using superpixel motion tracking. (B) Example Hex-GFP;mem-tdTomato embryo showing the start, middle and end of migration in a geodesic polar projection. (C) Superpixel motion tracking of embryo in B. As the tissue moves the grid is deformed, enabling analysis of the tissue behaviour within each tissue sub-region and comparison across embryos. (D) Final timepoint of superpixel motion tracking from all (n = 21) embryos in this dataset. All data is converted back to 3D coordinates for analysis. For each stage (pre-migration/initiation, migration, post-boundary) the grid is re-set and re-tracked by motion behaviour. Note re-projections have no scale bar as they are non-linear. HXMT = Hex-GFP;membrane-tdTomato, LA = Lifeact-GFP, TTRmTMG = mTmG;Ttr-Cre embryo.

Overview of analysis of motion superpixel tracks from time-lapse data.

(A) Example Lifeact-GFP embryo reprojected as a geodesic polar projection from 4D light-sheet imaging. (B) A uniform grid is seeded at the start of each migration phase and the motion in each sub-region tracked throughout the time-lapse. (C) Superpixel tracks are constructed for each region. (D) The apical surface coordinates of each track are converted back to 3D coordinate space. (E-G) Three parameters of motion behaviour calculated for each superpixel track: total speed, mean speed and anteriorwards velocity.

Analysis of MOSES superpixel tracking of VE tissue.

(A) Superpixel motion tracking of the emVE and exVE tissues confirms that emVE is significantly more active (higher mean speed and total speed) than the exVE (for both comparisons; Student’s t-test, p=<0.001). (B) Comparison of the anterior emVE and posterior emVE that are immediately adjacent to the DVE during the initiation of migration phase. There is no significant difference in motion (mean speed, total speed or anteriorwards velocity) or tissue area change (area fold change or area rate change) between these two tissue regions (all comparisons; Student’s t-test, p=>0.05). (C) Superpixel tracking of the distal VE during each phase shows that the migration phase has the highest anteriorwards velocity (one-way ANOVA, p=<0.01, followed by Tukey’s HSD Test on initiation of migration vs. migration (p =<0.001), migration vs post-boundary (p=<0.001), initiation of migration vs post-boundary (p=>0.05)). In the post-boundary phase the distal VE region contains the nascent Hex-GFP cells that continue to migrate anteriorly, but do so at a lower speed than the initial DVE.

Motion vector distribution in emVE during DVE migration.

(A) Diagram of polar view of VE. emVE is the central, blue coloured region. (B) For each sub-region of the emVE the median motion vector of superpixel tracks is calculated and plotted as a windrose plot for a total of 21 embryos. While the DVE region (sectors 1-8) show highly aligned anteriorwards vectors, there is a bi-rotational flow around the emVE (sectors 9-16). The length of lines is plotted by average superpixel velocity for each region – note the longer lines in the DVE region (sectors 1-8) compared to those of the surrounding emVE (sectors 9-16). Analysis of the distribution of vector angles (see Table S2 for Rayleigh distribution statistic) shows a significant Rayleigh distribution statistic across each emVE sub-region, showing this pattern of motion is conserved across embryos.

Analysis of ratchet-like DVE movement and T1 transitions in silico and in vivo.

(A) Example simulation at t = 0 and t = 6000 with cells coloured by shape index. (B) Flow maps of cells in simulation in A show a similar rotational flow to the mouse visceral endoderm. (C) Example T1 transition in simulation showing the intercalation of cell 2 between 1 and 3. (D) 6 example vertex simulations varying DVE self propulsive force, F, from 0.01 - 0.02, and emVE target shape index, p0, from more rigid to less rigid; 2.6 - 3.6. Orange line; T1 count. Blue line; DVE motion. (E) DVE motion and T1 transitions in 5 Lifeact-GFP E5.5 embryos. Magenta line; T1 count. Green line; DVE motion. Dashed vertical line shows the onset of directional DVE migration in embryos.

Motion analysis of epiblast tissue dynamics and epiblast cell tracking.

(A) Example motion superpixel grid tracking of epiblast during DVE migration with anterior epiblast region (red) and posterior epiblast (blue) regions highlighted at the start and end of the phase. Inset showing anterior and posterior epiblast regions at the end of DVE migration. Anterior epiblast becomes elongated and shifts posteriorly. (B) Motion vector plots of anterior epiblast regions during DVE pre-migration/initiation and migration phases. (C) Analysis of mean speed and anteriorwards velocity components of superpixel tracking of anterior epiblast during migration phases. (D,E) VE and epiblast tracking of two example embryos. Top row shows superpixel tracking of VE and epiblast. Bottom row shows cell tracking of VE (green tracks), anterior epiblast (magenta tracks) and posterior epiblast (cyan tracks).

Paxillin localisation in the E5.5 embryo.

(A) Whole-mount immunofluorescence of PAXILLIN in an example E5.5 embryo, showing a maximum intensity projection (MIP) and single z-section of the embryonic-half of the embryo. PAXILLIN is enriched along the basal surface of the VE and epiblast. (B) Diagram of Cre-inducible EGFP-PAXILLIN reporter line. Ttr-Cre is expressed specifically in the VE and Sox2-Cre in the epiblast. (C) Example live imaging of E5.5 Ttr-Cre EGFP-PAXILLIN embryo and (D) Sox2-Cre EGFP-PAXILLIN embryo, showing the basal localisation of the protein throughout the VE and epiblast. A total of n = 15 embryos were imaged for both crosses. All scale bars 25 μm.

Phase length

Migration phase vector angles