Comparing germband extension in wild-type and twist mutant embryos

To investigate whether mesoderm invagination has an impact on GBE, we carried out a quantitative analysis of GBE in wild-type and twist mutants, which lack mesoderm invagination. We chose to analyse twist mutants rather than snail or twist snail mutants, because although ventral furrow formation fails in all these mutants, some contractility remains in mesodermal cells in twist mutants alone, which decreases the width of the mesoderm and makes the ventrolateral field of cells more comparable with wild type. In snail or twist snail mutants, no contractility remains and the uninvaginated mesoderm takes significant space at the surface of the embryo [31].

We acquired movies of the ventral side of embryos from before the onset of mesoderm invagination until the middle of GBE, for both wild-type and twist mutants (Fig 1F–1H’). Four wild-type and 6 twist movies were analysed throughout this study. We labelled cell membranes with Gap43mCherry, and Myosin II with GFP-tagged knock-in of MRLC to visualise all Myosin II molecules (called sqhGFP^KI here; [39]). Strains were constructed to ensure that these labels were expressed at the same level in wild-type and twist embryos (Methods). We also controlled the temperature during imaging (20.5 +/− 1°C) to limit variability. Cell contours were segmented from the Gap43mCherry channel and tracked over time (Methods and S1 and S2 Movies) to calculate metrics describing key cell and interface behaviours (Fig 1F) [40]. Myosin II density and planar polarity were quantified from the Myosin II channel as before (Fig 1F) [26].

Our field of view captured the convergence and extension of the ectodermal tissue on the ventrolateral surface of the embryo but also included mesodermal cells and mesectodermal cells. The inclusion of the ventral midline enabled us to precisely measure orientations with respect to this landmark (Methods). To restrict our analysis to the extending ectoderm, we carefully excluded mesodermal and mesectodermal cells based on (i) whether they had invaginated (wild type only); (ii) their proximity to the midline; and (iii) the timing of their cell divisions [41,42] (Fig 1G and 1H’ and S1, S3 and S4 Movies).

Next, we checked that the ectodermal cells analysed in wild type and twist are comparable in terms of DV and AP genetic patterning. We examined the expression patterns of key DV patterning genes: single-minded (sim) is expressed in the mesectoderm, ventral nervous system defective (vnd) is expressed in the ectoderm abutting sim, and intermediate neuroblast defective (ind) is expressed next, abutting vnd [43]. While there is some de-repression of both sim and ind in the non-invaginated mesoderm and in the mesectoderm in twist mutants, the expression of all 3 genes is similar in the ectoderm (S2 Fig). We also checked the expression patterns of the Leucine-rich repeat (LRR) cell surface receptors Tolls 2, 6, 8, and Tartan, which are required for the polarised distribution of Myosin II during GBE downstream of the AP patterning genes [22,44–47]. We confirmed that these genes are expressed with the same patterns in twist mutants as in wild type (S2 Fig). Therefore, the regions of the ectodermal germband we analyse here in wild type and twist mutants are indistinguishable in term of patterning.

To compare averaged metrics between wild type and twist mutants, we synchronised the movies in time, using the start of tissue extension along the AP axis (measured as AP tissue strain rate) in each movie as time zero as before (Methods) (S3A and S3B Fig) [3,4,26,47]. To check that timelines of wild-type and twist mutant movies were comparable once synchronised, we checked the timings of 2 developmental events: (i) when Myosin II becomes detectable apically in the ectoderm; and (ii) when the first cell divisions occur (Methods). We found that appearance of apical ectodermal Myosin II (around 10 minutes before GBE) and of the first cell divisions in the lateral ectoderm (around 32 minutes after GBE) are grouped in time across movies (S3C Fig). We note that mesectoderm cell divisions are delayed by nearly 10 minutes in twist mutants compared to wild type, which might be a consequence of misregulation of mesodermal and mesectodermal patterning genes. We conclude that our movie synchronisation gives comparable developmental time windows for the 2 genotypes.

We next determined the number of cells tracked at each time-point in wild-type and twist movies (S3D–S3F Fig). At early time points, we track a total of approximately 800 cells (from 4 movies) for wild type and a total of 2,000 cells (from 6 movies) for twist. By the end of the period of analysis, over 2,500 cells are included for both wild type and twist.

In summary, the above characterisation demonstrates that we have acquired comparable live imaging datasets of thousands of tracked cells of the ectodermal germband for wild-type and twist embryos covering a developmental window of 15 minutes prior to 30 minutes after the onset of GBE. We confirmed that DV and AP patterning in the extending ectoderm is unchanged in twist mutant compared to wild type, allowing us to explore the mechanical impact of mesoderm invagination on GBE independently of patterning.

The invaginating mesoderm pulls on the extending ectoderm

As the mesoderm invaginates, cells of the ectodermal germband undergo translation ventrally in wild type, indicating that these 2 epithelial tissues are mechanically connected. This movement towards the midline is much reduced in twist embryos as expected from the lack of invagination (Fig 2A and 2A’). This mechanical linkage suggests that the invaginating mesoderm pulls on the adjacent ectodermal germband, generating a DV-oriented tensile stress. Gradients of cell elongation and apical area increase have been found as a signature of epithelial tissues being subject to tension from extrinsic forces generated by neighbouring morphogenetic movements [3,4,30,35]. To ask whether the ectoderm is being pulled by the invaginating mesoderm, we analysed these metrics for all ectodermal cells in the field of view for 4 wild-type and 6 twist embryos (Methods and Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Ectodermal cells during GBE are stretched and rotate in response to ventral pull from mesoderm invagination.

(A, A’) Cell trajectories (coloured lines) of germband cells over 5–10 minutes of GBE overlaid on Gap43Cherry maximum intensity projection at 10 minutes of GBE. Anterior to the left, ventral centrally. (A) wild type: germband cells move significantly ventrally and begin to move posteriorly. (A’) twist: germband cells move only slightly ventrally and begin to move posteriorly. (B, B’) Analysis of cell shape strain rate, projected along DV axis (proportion/minute) in ectodermal cells in wild type (B) and twist (B’) at 2.5 minutes into GBE. Cell shape strain rate of each cell is indicated by coloured square at the centre of each cell (cell outlines, grey). (C) DV-projected cell shape strain rate in proportion/minute (abbreviated as pp/min throughout the manuscript) of ectodermal cells, summarised for wild type (magenta) and twist (green) against time of GBE. The average time-period of mesoderm invagination, from first sustained apical constriction until all mesoderm cells are internalised, is indicated with a yellow bar here and in (E). Timing of mesoderm invagination was assessed by eye for all wild-type movies (S4A Fig). (D and D’) DV projected cell shape strain rate (pp/min) of germband cells, summarised for wild type (D) and twist (D’) against time of GBE and distance from mesectoderm (μm) (see Methods). (E) Cell area strain rate (pp/min) of germband cells, summarised for wild type (magenta) and twist (green) against time of GBE. (F and F’) Cell area strain rate (pp/min) of ectodermal cells, summarised for wild type (F) and twist (F’) against time of GBE and distance from mesectoderm (μm) (see Methods). (G) Average DV cell shape strain rate from −7.5 to 7.5 minutes of GBE summarised for wild type (magenta) and twist (green) against distance from mesectoderm (μm). Thinner lines show standard error. (H) Average orientation of principal axis (degrees from AP axis) of ectodermal cells against time of GBE for wild type (magenta) and twist (green). Note that if cells are randomly oriented, then the average cell angle would be 45 degrees. (I) Our “axial shape elongation” measure is a combined measure of cell elongation and cell orientation, giving a value of 1 for cells strongly elongated in DV, minus 1 for cells strongly elongated in AP and 0 for isotropic cells or elongated cells at 45 degrees to embryonic axes (see Methods). (J) Axial shape elongation measure for ectodermal cells against time of GBE for wild type (magenta) and twist (green). (K) Average orientation of cell interfaces from the AP axis of ectodermal cells for wild type (magenta) and twist (green) plotted against time of GBE. (L) Summary of average ectodermal cell behaviour in response to tensile force from invaginating mesoderm (note that angle of rotation shown is arbitrary). Throughout the manuscript: Dark grey shading indicates periods of significant difference (p < 0.01; see Methods), and ribbons show standard error between genotypes; in explanatory diagrams of the measures being plotted on graphs (for example, Fig 2C), the + directional arrow indicates where values will be positive, and the – direction arrow indicates where values will be negative. Data associated with this figure can be found in S1 Data. AP, antero-posterior; DV, dorso-ventral; GBE, germband extension.

https://doi.org/10.1371/journal.pbio.3002611.g002

In wild-type embryos, the rate of change in apical cell lengths along DV (DV cell shape strain rate, Fig 2B and 2C) and in apical areas (Fig 2E) both increase significantly from 10 minutes before the start of GBE, showing that cells are starting to stretch. The rate of DV stretching increases rapidly, peaking at about 3 minutes into GBE, before rapidly decreasing until 7 to 9 minutes into GBE (Fig 2C and 2E). After this time, the cells start shrinking, reducing their DV length and cell area until the end of the period analysed, 30 minutes into GBE. To compare the timing of these cell behaviours in the ectoderm to the behaviour of the mesoderm, we documented the key events of mesoderm invagination in our 4 wild-type movies (S4A Fig). We find that the period of cell stretching corresponds to when the mesoderm starts contracting and then invaginates in the movies (about -13 to +7 minutes, yellow bar in Fig 2C and 2E). This is consistent with previous reports of cell stretching in the ectoderm in response to mesoderm invagination [3,4,27,30]. We next plotted cell shape changes and apical area changes on spatiotemporal plots to examine how they evolve in time and space. In wild type, the rate of change for DV cell shape and apical area have the same spatiotemporal pattern (red signal in Fig 2D and 2F), confirming that the cells are stretching in DV and indicating that the cells are under DV-oriented apical tension. Also, there is a gradient in the observed cell stretching from ventral to lateral, suggesting that the source of tension is positioned ventrally, as expected if the mesoderm is pulling (Fig 2G). Consistent with this notion, in twist embryos, cell stretching in DV is virtually abolished (Fig 2C–2F’) and there is no gradient of DV cell elongation (Fig 2G). We conclude that the invaginating mesoderm generates a tensile force that is transmitted across the germband, causing ectodermal cells to stretch along DV, with a ventral to dorsal gradient. In wild type, following this period of intense cell stretching in DV, cells shrink again along this axis (Fig 2C), eventually regaining an isotropic shape [3].

This tissue-level tensile force could also cause ectodermal cells to reorient, so we quantified cell orientation (Fig 2H–2K). First, we plotted cell orientation over time and find for both wild type and twist that cells are slightly DV-oriented already 15 minutes before the start of GBE (average angle greater than 45 degrees, Fig 2H). This trend increases rapidly in wild type until about 5 minutes into GBE, reaching a maximum average angle of 75 degrees, showing that cells become very DV-oriented. Cells in twist mutants also become briefly more aligned with the DV axis, but to a much lesser extent than wild type, peaking at about 60 degrees. A combined measure of cell elongation and cell orientation, which we call “axial cell elongation bias,” confirms that in wild type, cells rotate to become more DV-oriented as they stretch in DV (Fig 2I and 2J and Methods). The strongest effect is seen from 5 minutes prior to the start of GBE to 15 minutes after (peaking around 7 minutes) and is mostly absent in twist. This demonstrates that mesoderm invagination causes cells not only to stretch but also to realign perpendicular to the ventral midline. We also plotted average orientation of cell junctions from the AP axis (Fig 2K). At −15 minutes before GBE starts, the average orientation is approximately 45 degrees for both wild type and twist, showing that there are no bias in directionality of interfaces just before mesoderm invagination. In wild type, the average orientation increases towards DV from −10 minutes as cells stretch in DV, peaking at 8 minutes at approximately 55 degrees, consistent with DV-oriented interfaces becoming more aligned with the DV axis as cells stretch. In contrast for twist, the average orientation changes little over time (Fig 2K), consistent with the finding that there is little DV stretch in cells in twist mutants.

Extrinsic tissue-level forces can drive cells to move past each other, leading to passive cell intercalations [5,6]. To ask whether the DV-oriented pull of mesoderm invagination causes cell slippage in the ectoderm, we quantified the cell intercalation strain rate as previously [40] (Methods). We find that the intercalation strain rates along the DV axis is negligible or negative during the period of DV cell stretch, with no difference between wild type and twist mutants (S4B Fig). This demonstrates that the cell stretching caused by mesoderm invagination in the ectoderm is not accompanied by passive cell rearrangement.

In summary, our results show that in wild type, the contraction force from the invaginating mesoderm causes the ectodermal cells of the germband to stretch in DV and change orientation, without causing cells to rearrange along the axis of tension (Fig 2L). We next asked whether this mechanical force exerted by the invaginating mesoderm onto the ectoderm impacts on axis extension.

Comparing Myosin II planar polarisation in wild-type and twist mutants

Myosin II–driven polarised cell intercalation is the main cell behaviour causing the convergence and extension of the Drosophila germband [18,22]. During this process of cell neighbour exchange, DV-oriented junctions between AP cell neighbours shrink, so that these cells lose contact with each other, and then a new cell junction grows in the perpendicular orientation (Fig 3A). Cell junction shrinkage is driven by Myosin II, which is enriched in shrinking junctions under the control of the AP patterning genes [19,20]. In addition to genetic regulation, there is some evidence in the Drosophila germband that Myosin II can be recruited through a mechanosensitive feedback mechanism [24,48]. Since mesoderm invagination generates a DV-oriented tensile force within the ectoderm, mechanosensitive recruitment or stabilisation of Myosin II might contribute to the enrichment of Myosin II on DV-oriented cell junctions. Alternatively, cortical Myosin II might be diluted along DV-oriented cell junctions as they are stretched by the invaginating mesoderm. We therefore asked how Myosin II concentration on DV-oriented junctions is affected by mesoderm invagination.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Myosin II bipolar enrichment in wild-type and twist mutants.

(A) Diagram showing Myosin II–driven cell intercalation. Myosin II (orange) shows planar polarised localisation, being enriched in a bipolar manner on the DV-oriented junctions (between cells neighbouring each other along the AP axis). Interface shrinkage (brown arrows) leads to a transient 4-way vertex, followed by growth of a new AP-oriented interface. (B and C) Normalised Myosin II signal at the level of adherens junctions extracted from confocal image stacks for quantification in wild type (B) and twist (C) at 10 minutes of GBE. See Methods and S7 Fig for details of extraction and normalisation of signal. (B’ and C’) Quantified interface Myosin II signal of ectodermal cells corresponding to images in (B) and (C) for wild type (B’) and twist (C’). Dark blue corresponds to lowest Myosin II density through to red for highest Myosin II density. Note there are no interfaces at this time point with highest Myosin II density (red), so green-yellow shows highest Myosin II in these frames. (D) Average density of Myosin II on ectodermal cell interfaces, summarised for wild type (magenta) and twist (green) against time of GBE. (E) Unprojected bipolarity of Myosin II of ectodermal cell interfaces, summarised for wild type (magenta) and twist (green) against time of GBE (f.i. denotes fluorescent intensities throughout the manuscript). (F) Rate of change of unprojected bipolarity of Myosin II of ectodermal cell interfaces, summarised for wild type (magenta) and twist (green) against time of GBE. (G) Proportional rate of interface length change, for DV-oriented (oriented more than 45 degrees from the AP axis) interfaces only, summarised for wild type (magenta) and twist (green) against time of GBE. (H) Rate of Myosin II density change, for DV-oriented (oriented more than 45 degrees from the AP axis) interfaces only, summarised for wild type (magenta) and twist (green) against time of GBE. Data associated with this figure can be found in S2 Data. AP, antero-posterior; DV, dorso-ventral; GBE, germband extension.

https://doi.org/10.1371/journal.pbio.3002611.g003

First, we quantified fluorescence as a readout of Myosin II density at the level of the adherens junctions in the ectoderm, for our wild-type and twist movies (Fig 3B and 3C’) (Methods). We plotted average fluorescence intensities for these cell–cell interfaces over time and observed that in both wild type and twist, Myosin II density starts low before GBE (Fig 3D), consistent with our previous measures [26]. Myosin II levels then increase, gradually at first and then more rapidly until approximately 15 minutes after the start of GBE, after which they continue to increase more slowly. Wild-type and twist curves are virtually identical demonstrating that accumulation of Myosin II at apical cell–cell junctions is unaffected in twist mutants (see also S3C Fig).

Myosin II is expected to be enriched on membranes oriented approximately parallel to the DV embryonic axis [19,20], and this is apparent in our movies of both wild type and twist (see S3 and S4 Movies). We quantified this Myosin II planar polarisation by measuring Myosin II bipolar distribution around the membrane of each cell as before [26]. Because there are differences in cell orientation between wild type and twist (see Fig 2H–2K), we plotted the average bipolarity of germband cells, independently of the orientation of the bipolarity. In both wild-type and twist embryos, this unprojected Myosin II bipolarity measure starts to increase around the start of GBE, increases quickly in the first 15 minutes of GBE and then stays high for the remaining 15 minutes, with no statistically significant difference between the 2 genotypes (Fig 3E). Spatiotemporal plots for these data show that these temporal patterns are similar for all cells in the field of view and that there is no obvious ventral to dorsal gradient in wild type (S4D and S4D’ Fig). Although there is no significant difference in unprojected bipolarities at any time between the 2 genotypes, the rate of increase appeared marginally faster between 2 and 12 minutes of GBE in the wild type compared to twist mutants (Fig 3E). Therefore, we calculated the rate of change of Myosin II bipolarity, and we found that in twist embryos, it is slightly reduced compared to wild type, showing some statistically significant differences around 5 minutes of GBE (Fig 3F).

Because these differences are seen while cells are stretching in DV (see Fig 2C–2F’), we asked whether cell interface stretching is associated with an increased rate of Myosin II recruitment. We plotted the proportional rate of stretching of DV-oriented interfaces and confirm that significant stretching occurs in wild type, but not in twist, between approximately −10 and 7 minutes of GBE (Fig 3G). We then plotted the rate of Myosin II density change for DV-oriented interfaces and observed that the rate of Myosin II recruitment increases rapidly during this time-period not only in wild type but also in twist mutants whose DV-oriented interfaces undergo very little stretching (Fig 3H). Apart from some very brief bursts of significance, there is no difference between wild type and twist for this measure (Fig 3H). Altogether, this indicates that if there is any increase in Myosin II localisation early in GBE in response to the physical pull by the mesoderm on the ectoderm, it is at the limit of what we can detect with our measures. Interestingly, our results show, however, that Myosin II density is maintained in DV-oriented junctions that are stretching under tension from mesoderm invagination, compared to unstretched DV-oriented junctions in twist embryos, indicating that a homeostatic mechanism maintains Myosin II cortical density during junctional stretching.

Comparing the rates of junctional shortening and neighbour exchange in wild-type and twist mutants

In addition to intrinsic forces within the germband driving its convergence and extension, ectoderm extension is facilitated by an extrinsic pull (along AP) from the posterior endoderm [3,4,9]. Both the convergence and the extension in the germband could be mechanically counteracted by the pull along DV from the invaginating mesoderm. If so, in the absence of mesoderm invagination, we might expect junctional shortening to speed up due to the lack of a DV tension opposing junctional shrinkage. Thus, boundary conditions imposed by the endoderm and mesoderm invaginations could affect the behaviour of ectodermal cells during GBE. Junctional shortening rates could also be influenced by mesoderm invagination in the opposite way (i.e., junctional shortening being faster in wild type than in twist mutants) through mechanosensitive Myosin II recruitment, as considered above.

Therefore, we examined whether mesoderm invagination impacts on the rate of junctional shrinkage of the extending ectoderm throughout early and mid-GBE (0 to 30 minutes). We find that while there is a consistent trend over time for the proportional rate of junctional shrinkage to be higher in wild-type interfaces compared to twist, this difference is statistically not significantly different (Fig 4A). Because wild-type cell interfaces have been stretched by mesoderm invagination, they will generally be longer in length when they initiate junctional shrinkage for intercalation. Therefore, we also compared the speed of junctional shrinkage in microns per minute. We find that wild-type cell junctions initially shrink significantly more quickly than twist in absolute terms (S4E Fig). In summary, although wild-type interfaces start longer and shrink more quickly in absolute terms, in proportional terms, they shrink at approximately the same rate (no significant difference) but with a tendency for wild-type junctions to shrink slightly quicker.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Junctional shortening speed and rate of neighbour exchange in wild-type and twist mutants.

(A) Proportional rate of interface shrinkage plotted against time to T1 swap during GBE (data for 0–30 minutes of GBE) summarised for wild type (magenta) and twist (green). (B) Average density of Myosin II on shrinking interfaces, plotted against time to an intercalation event (swap) during GBE (data for 0–30 minutes of GBE) summarised for wild type (magenta) and twist (green). (C) Example of detection of neighbour exchange events in the germband in wild type at 20 minutes of GBE. Loss of neighbours is shown as blue lines between cell centroids, while gain of neighbours is shown in red. Cell centroids are shown as black dots. Underlaid is an image of Gap43Cherry signal at the level of adherens junctions extracted for tracking. (D) Rate of neighbour exchange. Number of interfaces involved in a T1 swap expressed as a proportion of the total number of DV-oriented interfaces for all tracked ectodermal germband cells, summarised for wild type (magenta) and twist (green) over time of GBE. Data associated with this figure can be found in S3 Data. DV, dorso-ventral; GBE, germband extension.

https://doi.org/10.1371/journal.pbio.3002611.g004

Since Myosin II is proposed to be the main source of mechanical stress driving junction shrinkage [19,20], we next quantified the density of Myosin II specifically on shrinking interfaces (Fig 4B). We find that wild-type Myosin II densities tend to be higher than those of twist mutants, but with only a very brief period of statistically significant difference between the 2 genotypes, at 5 minutes prior to neighbour exchange. This is consistent with the very mild effect on Myosin II in response to mesoderm invagination described in the section above. The slight decrease in Myosin density on shrinking interfaces in the twist mutant compared to wild type could explain the slight decrease in their proportional contraction rate.

The proportional speed in junctional shortening should influence the rate at which cells exchange neighbours, and from above our prediction would be that there should be no difference between wild type and twist. We tracked discrete cell neighbour exchanges (T1 swaps) over time (Fig 4C) (see Methods). We find that the number of T1 swaps increase from the start of GBE until about 20 minutes and that wild-type and twist embryos show a comparable proportion of T1 swaps throughout the course of GBE, with no discernible delay in twist compared to wild type (Fig 4D).

In summary, we conclude that there is no significant effect of mesoderm invagination on the proportional rate of junction shrinkage. However, wild-type junctions start longer and shrink more quickly in absolute terms and therefore are not delayed in reaching the point of neighbour exchange compared to twist mutants. Consistent with these observations, we find no difference in the proportion of cells undergoing intercalation events between wild-type and twist mutants. Therefore, there is no significant positive or negative effect of mesoderm invagination on the process of junctional shortening leading to cell intercalation events. However, we do see a slightly (not significantly) higher proportional rate of junction shrinkage in wild type compared to twist, accompanied by a slightly increased Myosin II density on shrinking junctions. Therefore, junctional shrinkage appears to be mildly affected by a mechanosensitive response of Myosin II, and this may mask the effects of boundary conditions on this process.

Comparing junctional elongation rate and AP cell elongation in wild-type and twist mutants

Following shrinkage of Myosin II−enriched cell junctions, cells exchange neighbours and cells previously separated from each other along the DV axis come into contact, forming a new junction (approximately AP-oriented) to drive extension of the tissue along the AP axis (Fig 3A) [9]. Cell junction elongation has been less studied than cell junction shrinkage in GBE. There might be active mechanisms involved, but biophysical modelling suggests that new junction growth could also occur passively, following the resolution of unstable 4-way junctions [49]. Independently of the mechanism by which junctions elongate, active or passive, stress is expected to influence the dynamics of junctional elongation. For example, as for junctional shortening rates, the balance of boundary conditions on the extending ectoderm could impact junctional elongation rates. Indeed, while the posterior pull by endoderm invagination would promote junction extension, the counteracting orthogonal pull by mesoderm invagination would oppose junction extension. Therefore, an increase in junctional elongation rates might be expected in twist mutants, since new AP-oriented junctions would grow in the absence of an orthogonal tension (along DV) that could slow junction growth. The change in boundary conditions in twist mutants might also increase the AP cell shape strain rates since in absence of an orthogonal pull from mesoderm invagination, cells might elongate more due to the endoderm invagination pulling along AP.

We first quantified the proportional rate of elongation of new junctions in wild type and twist (Fig 5A). We restricted our analysis to 0 to 20 minutes of GBE, when the endoderm pull along AP appears to have its strongest effect, based on the rate of AP cell elongation (measured as AP cell shape strain rate; [3,4] and see below). We found that new junctions grow significantly quicker in twist compared to wild type just after the T1 swap (Fig 5A). Then, junction elongation rates gradually become similar between the 2 genotypes. This suggests that in wild type, the DV tension in the germband that is caused by mesoderm invagination slows the initial rate of elongation of new junctions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Junctional elongation speed and AP cell elongation in wild type and twist mutants.

(A) Proportional rate of interface growth in ectodermal germband cells over time from the time of T1 swap summarised for wild type (magenta) and twist (green) (data for 0–20 minutes of GBE). (B) AP cell shape strain rate plotted against time of GBE summarising data from the ectoderm of the full imaged region of 4 wild-type and 6 twist movies. (C, C’) Spatiotemporal plots of AP cell shape strain rate plotted against AP position and time of GBE summarising wild-type (C) and twist (C’) movies. Data associated with this figure can be found in S4 Data. AP, antero-posterior; GBE, germband extension.

https://doi.org/10.1371/journal.pbio.3002611.g005

We next measured AP cell shape strain rate. Consistent with our previous studies, cells elongate along AP from the onset of GBE until about 20 minutes in wild type (Fig 5B), with a characteristic spatiotemporal gradient of elongation highest towards the posterior of the embryo (Fig 5C) [3,4]. In twist mutants, AP cell shape strain rates are significantly elevated compared to wild type (Fig 5B), with a more pronounced posterior spatiotemporal gradient (Fig 5C’). As expected, in both genotypes, AP cell elongation is accompanied by an increase in cell area, showing that the cell apices are under tension (S5A and S5A’ Fig). In wild type, cells are stretched first along DV by mesoderm invagination and then along AP by endoderm invagination, causing a substantial apical cell area increase across the ectoderm (S5A Fig). In twist mutants, the loss of the pull from mesoderm invagination causes a strong diminution in this cell area increase (S5A’ Fig). There might be a limit up to how much the cells can increase their apical area, which, in turn, might limit how cells will stretch along AP in wild type. In twist, the stretching along DV does not take place because of the loss of mesoderm invagination, and this might allow the cells to stretch more along AP in response to endoderm invagination, as we observe (Fig 5B and 5C’).

In contrast to in DV, we do not have a precise registration of the position of the cells along the AP axis in our movies, since the field of view is positioned around the middle of the embryo, with no landmarks available for registration (see Methods). To address a possible impact of position variability along AP, we examined individual spatiotemporal plots for AP cell shape strain rates for each wild-type and twist movie (S5B and S5C Fig). We find that the variation of the anterior edge of the field of view means that in some plots but not others, some AP cell elongation signal contributed by the cephalic furrow is included (the cephalic furrow separates the head from the trunk of the embryo and forms just outside the anterior edge of the field of view). Also, the variation of the posterior edge of the field of view means that more or less of the posterior gradient is in view (this will depend on the centering of the field of view and variation in embryo shapes). To control for these variations, we selected a 100 μm-long central region on the plots, systematically excluding the signal at the anterior due to cephalic furrow formation (boxed regions in S5B and S5C Fig) (Methods). We find that AP cell elongation is also higher in this central region in twist mutants compared to wild type, with several bursts of significance (S5D Fig), consistent with our analysis with the entire field of view (Fig 5B).

In summary, we conclude that a change in the DV boundary conditions in twist mutants causes both an increase in the rate of elongation of new junctions following T1 transitions and an increase in AP cell elongation contributing to GBE. This is in contrast with our finding that the change in boundary conditions does not affect junction shrinkage.

Comparing the orientation of cell interfaces during cell intercalation in wild-type and twist mutants

Next, we asked whether the observed changes in the alignment of cells and cell junctions with the embryonic axes, in twist compared to wild type, could impact tissue extension, since this could lead to polarised cell intercalation being less well aligned with the embryonic axes, which might lessen how effective convergence and extension is.

First, we examined Myosin II bipolarity alignment with the embryonic axes in twist mutants compared to wild type. We projected our bipolarity cell measure along the anterior-posterior axis to assess the extent to which Myosin II is enriched at the anterior and posterior of each cell (Figs 6A, S6A and S6A’). We find that the rapid increase of AP-projected Myosin II polarity is delayed by approximately 5 minutes and significantly reduced in twist compared to wild type, until approximately 10 minutes into GBE. This difference is more pronounced than for the unprojected data (compare Fig 6A with Fig 3E), showing that the Myosin II bipolarity of cells is less well aligned with the embryonic axes early in GBE, in embryos where mesoderm invagination is defective. We conclude that the Myosin II–enriched junctions in twist mutants are less well aligned perpendicular to the AP axis than in wild type.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Orientations of growing and shrinking junctions and their impact on AP intercalation strain rates in wild-type and twist mutants.

(A) Myosin II bipolarity, projected along AP, of all tracked ectodermal cells, summarised for wild type (magenta) and twist (green) plotted against time of GBE. (B) Orientation (degrees from AP axis) of shrinking interfaces at 6.5 minutes prior to T1 swap of all tracked ectodermal germband cells, summarised for wild type and twist over time of GBE. (C) Orientation (degrees from AP axis) of growing interfaces in ectodermal germband cells, summarised over the first 30 minutes of GBE for wild type and twist plotted against time from neighbour exchange event. (D) Orientation (degrees from AP axis) between cell centroids of pairs of “new neighbours” in ectodermal germband cells, summarised over the first 30 minutes of GBE for wild type and twist, plotted against time from neighbour exchange event. (E) Rate of productive neighbour exchange. Rate of net gains along the AP axis, using a continuous angular measure for productivity of T1 swaps to axis extension (see Methods), expressed as a proportion of the total number of DV-oriented interfaces for all tracked ectodermal cells, summarised for wild type (magenta) and twist (green) over time of GBE. (F) AP-projected intercalation strain rate of all tracked ectodermal cells, summarised for wild type and twist plotted against time of GBE. Data associated with this figure can be found in S5 Data. AP, antero-posterior; DV, dorso-ventral; GBE, germband extension.

https://doi.org/10.1371/journal.pbio.3002611.g006

While the number of intercalation events is not reduced in twist compared to wild type (Fig 4D), the changes in average cell junction orientation (Fig 2K) and AP-projected bipolarity of Myosin II (Fig 6A) suggest that shrinking interfaces involved in T1 transitions may be poorly aligned with the DV axis. So, we measured the orientation of shrinking interfaces shortly before neighbour exchange and throughout GBE (Fig 6B) (Methods). We find no statistically significant difference between the orientations of shrinking interfaces in wild type and twist. In the first 5 minutes of GBE, there is a trend for shrinking interfaces to be less well aligned with the DV-axis in twist mutants, but this difference does not exceed 15 degrees (Fig 6B). These results suggest that the orientation of actively shrinking interfaces is not predominantly controlled by the global tissue movements and tension induced by mesoderm invagination. Because the maximum deviation in shrinking junctions misalignment we observe is 15 degrees at the start of GBE, then the maximum expected difference in strain rate projected along the DV axis from this misalignment would be: 1 –cos(15) = 3.4%, predicting a very minor effect. Consistent with this, we found no significant reduction in DV intercalation rate contributing to convergence, in twist embryos compared to wild-type embryos (see S4B Fig, the negative values show the convergence rate in DV).

Previously, in other mutants affecting the orientations of shrinking interfaces, we had seen that the angle between junction shrinkage and growth was maintained (approximately 90 degrees), so growing junctions were also misoriented [50]. So, we also assessed the orientation of new junction growth. We did not detect a statistically significant difference in orientation of growing interfaces in twist mutants compared to wild type, although the trend is for twist to be less well aligned with the AP axis compared to wild type (Fig 6C). When plotting the angles between the centroids of newly neighbouring cells, we do see that these centroid–centroid angles are significantly less well aligned with the DV axis in twist mutants than in wild type for the first few minutes after a neighbour exchange (Fig 6D). Therefore, although T1 neighbour exchanges occur at normal rates in twist mutants (Fig 4D), the resulting cell packing is mildly perturbed compared to wild type soon after the neighbour exchange, but this effect is short-lived. We also quantified the rate of “productive” neighbour exchanges contributing to tissue extension (that is net T1 gains along the AP axis, taking into account the varying angles of T1 events with respect to the embryonic axes, see Methods) and observe that it is very similar in twist compared to wild type (Fig 6E). This shows that minor observed differences in angles of T1 events between wild type and twist do not impact on the ability of T1 events to extend the ectodermal germband.

Finally, we measured the intercalation strain rate in AP to ask whether the minor differences that we observe in junction growth speed and orientation cause any difference in the extension of the tissue by cell intercalation (Fig 6F). This is a continuous measure of how much cell intercalation is contributing to tissue extension in AP [3,40] and therefore will integrate effects on the numbers of T1 events, speed and orientation of junctional shrinkage, and junctional growth (see S6B Fig for the definition of intercalation strain rate). We find no convincing difference between wild-type and twist embryos, though there is a trend for twist intercalation strain rate to be slightly higher, with several short bursts of significant differences between wild-type and twist. This tendency for twist intercalation strain rate to be mildly higher is consistent with our findings that the junctional growth rate is increased after T1 events (see Fig 5A).

In summary, although there are measurable differences in orientation of shrinking and growing interfaces, the effect on tissue extension is negligible. Considering all our results together, we conclude that the polarised cell intercalations causing GBE are not significantly augmented by mesoderm invagination and that they are robust to the extrinsic force that it exerts.

Comparing tissue extension rate in wild type and twist mutants

Finally, we sought to compare how the different effects analysed above impact on total tissue extension during GBE. We used our measure of tissue strain rate along AP to address this [3,40]. We find that AP tissue strain rates are higher in twist mutants than in wild type, showing that germband extension is proceeding faster in twist mutants than in wild type (S6C Fig). This elevated rate of AP extension is expected from our observations that the AP cell elongation is higher in twist than in wild type, while there is little difference in the AP intercalation strain rate. This is because in our mathematical description, tissue strain rate is the sum of cell shape strain rate and intercalation strain rate (S6B Fig) [3,40]. Since AP cell shape strain rate varies along AP (S5B and S5C Fig), so will AP tissue strain rate. So as for the AP cell shape strain rate, we analysed AP tissue strain rate for a central 100 μm-wide region as before, to control for slight variability in the AP positioning of the field of view between movies. We confirm that in this central region, AP tissue strain rates are still significantly higher in twist mutants compared to wild type (S6D Fig), confirming results for the full field of view (S6C Fig).

Mechanical impact of mesoderm on the ectoderm

Previous studies looking at the mechanical impact of mesodermal cell contraction have focused on cell deformation within the mesoderm domain [35–38]. These studies have shown that while the ventral-most region of the mesoderm undergoes apical constriction, a region of about 3 cells on either side (the most lateral of those being the mesectoderm cells) do not constrict but instead stretch along DV. Beyond the mesectoderm cells, the lateral ectoderm cells were thought to be “inert” and mechanically buffered by the stretching of compliant tissue at its ventral (and also dorsal) edge [13,35,36]. In this paper, we show that ectodermal cells do respond mechanically to mesoderm invagination. We detect a ventral to dorsal gradient of cell stretching in the lateral ectoderm, which is lost in twist mutants (Fig 2G). This suggests that forces generated by the contractile activity of the apical actomyosin in the mesoderm do transmit to the ectoderm via the apical adherens junctions. When mesoderm invagination begins, the level of apical Myosin II is still very low in the lateral ectodermal cells (Fig 3D), which might explain why ectodermal cells stretch significantly under tension from the invaginating mesoderm, relieving the mechanical stress. We also show that ectodermal cell stretching is not accompanied by cell slippage.

Mechanisms regulating Myosin II density at the cortex and consequences for polarised cell intercalation

The stretching of ectodermal cell–cell junctions could trigger a mechanosensitive response leading to an increase in Myosin II enrichment, or, alternatively, a dilution of Myosin II (Fig 7). Although we do not find evidence of a significant increase in Myosin II density (see below), we find a clear absence of dilution. This suggests that a homeostatic mechanism maintains Myosin II density at the cortex upon rapid stretching. It might be that the AP patterning system controlling Myosin II enrichment at DV-oriented interfaces senses the density of Myosin II at the cortex and maintains it at a specific density. Alternatively, the homeostatic mechanism might be a more general mechanosensitive response to stretch. One consequence of maintaining Myosin II density in stretching DV-oriented junctions is that these junctions undergo shrinkage to initiate neighbour exchange events in wild-type embryos, at the same rate as unstretched junctions in twist mutants. In contrast, in the case of the lateral mesoderm, the already low levels of apical medial Myosin II become further diluted and are easily overcome by the pulling forces of the central mesoderm. This causes an extreme apical stretching of cells, rather than homeostatic maintenance of apical actomyosin density [35]. This highlights the fact that responses of cells to apical tension is cell-type specific.

We did find that Myosin II recruitment rate was temporarily marginally higher in wild type than in twist, providing further evidence of a mechanotransductive response. It is possible that the homeostatic mechanism discussed above is triggered by mechanical stretch and overshoots slightly, leading to a marginally increased rate of Myosin II recruitment. However, this increase is at the limit of detection and did not have a significant effect on the proportional shrinkage of cell junctions during neighbour exchange.

Mechanosensitive feedback in Myosin II recruitment has been reported at cable-like actomyosin enrichments spanning several junctions connected to each other later in GBE and at parasegmental boundaries, due to linked junctions pulling on each other [24,48]. In both cases, when the “cables” were cut by laser dissection to decrease tension in the linked junctions, a modest decrease in Myosin II recruitment was measured (less than 10%). Considering that mesoderm invagination pulls on the ectoderm when apical actomyosin is very low and just starting to increase (Fig 3D), it is perhaps unsurprising that any mechanosensitive increase in Myosin II would be challenging to detect.

Comparison with other studies

Our results are in contrast to some other studies that have reported a stronger effect of mesoderm invagination on Myosin II localisation and GBE, but there are inconsistencies in results between studies, perhaps due to the challenge of acquiring comparable datasets between wild type and mutants, so the precise effect of mesoderm invagination on GBE has not been clear [3,4,27–30]. Indeed, the first study noting a possible reduction of cell intercalation in mutant embryos lacking mesoderm invagination was from our own laboratory [3]. Since this study, we have improved cell tracking techniques (for example, manual correction of mistracked cells and more precise exclusion of the mesoderm/mesectoderm cells) and started imaging earlier which has improved movie synchronisation and allowed us to collect more cells for analysis around the time of mesoderm invagination (see Methods). Additionally, there are now more fluorescently tagged proteins available, and we have taken care here to choose markers that are expressed at the same level between wild-type and twist embryos. For these reasons, along with the consistency between our different analyses, we are confident in the results that we present here.

Restoration of DV cell length after being stretched by mesoderm invagination

Following mesoderm-induced DV cell and cell junction stretching, cells contract to regain an approximately isotropic shape by the end of GBE [3]. When comparing junctional shrinkage leading to neighbour exchange, we found that the proportional rate of length change was not significantly different, but in absolute terms (μm/minute), wild-type junctions shrink faster than those in twist mutants. This is because wild-type DV-oriented junctions have been stretched, and these cell junctions are both restoring their length after stretching and undergoing shrinkage for intercalation events simultaneously. The transient marginal increase in the rate of Myosin II recruitment in wild type compared to twist (Fig 3) may be due to a mechanotransductive mechanism that acts to restores junctional length after stretching. Other mechanisms might also contribute to the restoration of cell shape after stretching, including the material properties of the cells, the activity of medial myosin networks at the apices of cells, and three-dimensional constraints to their shape (the volume of cells is thought not to change in the early embryo [51]). Nevertheless, the possibility remains that mechanosensitive recruitment or stabilisation of Myosin II on the cortex of stretched junctions is important for cell junctions to resist overstretching and subsequently restore their length. In the context of GBE, however, this does not significantly help to drive junctional shortening for cell intercalation as we have shown by comparing proportional junctional shrinkage rates.

Boundary conditions during germband extension and impact on junction shortening and elongating rates

Loss of mesoderm invagination would be expected to change the boundary conditions during GBE. The extending germband is pulled along AP as a consequence of posterior endoderm invagination, while mesoderm invagination pulls along DV. These 2 pulls can be considered as boundary conditions acting in perpendicular directions on the extending germband. Even after mesoderm invagination is complete, the DV tension may remain as the mesectodermal cells meet and adhere to each other at the midline, which could hold tension in the ectoderm. This DV-oriented force could counteract both the AP pulling force and forces generated within the germband, promoting convergence and extension, such as those generated by Myosin II contraction at shrinking interfaces. If so, the rate of junction shortening and elongation would increase in the absence of mesoderm invagination. We find a clear increase in the initial rate of junction growth, and in cell elongation along AP, in twist mutants (Fig 5). This suggests that mesoderm invagination does act as a counteracting force. Interesting, following the initial dissipation of stress from resolving the unstable 4-way vertex, the rate of junctional elongation slows and become comparable in wild-type and twist mutants, suggesting that it is the initial phase of junctional growth that is affected by the global force balance of boundary conditions. We note that despite the increase in initial junctional elongation rate seen in twist mutants, we see little impact of this on the overall rate of cell intercalation.

We do not, however, detect a similar increase in the rate of junctional shortening in twist mutants (in fact, it tends towards a reduction, Fig 4). Based on the above, it is plausible that an increase in rate of junctional shortening due to mechanical effects might be cancelled out by a decrease in rate of junctional shortening linked to a decrease in Myosin II mechanosensitive recruitment in absence of cell stretching. In this model, the opposing force of mesoderm invagination to ectoderm extension would be balanced by the increased contractility of the stretched interfaces.

Junction orientation and tissue extension

We addressed whether mesoderm invagination impact ectoderm extension by helping to align the junctions of rearranging cells along the main embryonic axes. We do find a significant effect of mesoderm invagination on cell and junction orientation up to approximately 15 minutes of GBE. However, the effect on the shortening cell junctions themselves is marginal, even within this first 15 minutes of GBE, suggesting that their orientation is relatively insensitive to tissue-level forces. This could be explained by the rapidly increasing Myosin II planar polarisation of the ectoderm during this period. We and others have shown that Myosin II–enriched interfaces form supracellular cable-like structures that straighten throughout the course of GBE [24,26,47]. Thus, this tissue-autonomous straightening might explain why the orientation of shortening junctions is mostly unaffected by mesoderm invagination. Consistent with this, [24] report that these cable-like structures appear normal in twist snail double mutants later in GBE.

We also found only marginal differences in the orientation of growing junctions between wild-type and twist mutants. This is expected since junctions grow in an orientation perpendicular to shortening junctions, and this angle is tightly constrained [50]. Also, invagination of the posterior endoderm (which still occurs in twist mutants) will act to pull growing junctions to be parallel with the AP axis [9]. Overall, we find that the minor differences we observe in the orientation of shrinking and growing interfaces do not have an impact on the productivity of cell intercalation events for tissue extension.

Fly maintenance and strains

We used a CRISPR knock-in line in the spaghetti-squash locus (sqh-eGFP^29B, referred to in this work as sqhGFP^KI) to tag endogenous Myosin II regulatory light chain (MRLC) with GFP [39]. The line is homozygous viable and so all the molecules of MRLC are labelled with GFP. We used Gap43mCherry (on the X) to mark cell membranes [52]. A recombinant chromosome between sqhGFP^29B and Gap43Cherry was made using standard Drosophila genetics. We used the null mutation twi¹ (Bloomington Stock Center). twist homozygous embryos were collected from sqhGFP^29B, Gap43Cherry; twi¹/Cyo stock and wild type from sqhGFP^29B, Gap43Cherry; Gla/Cyo stock. Standard fly husbandry was used to maintain stocks.

In situ hybridisation chain reaction and antibody staining

In situ hybridisation chain reaction (HCR) and immunostaining methods are as described previously [47]. Stage 5 to 8 embryos were collected from twi¹/CyO stocks. Morphology of the ventral furrow and ectopic expression of ind or sim within the mesoderm was used to identify mutants. The HCR probe sets were designed by Molecular Instruments. Primary antibodies used were mouse anti-phospho-Tyrosine (pTyr) (Cell Signalling #941) (1:1,000) and rabbit anti-phospho-Histone3 (pHis) (Cell Signalling #9701; 1:200). Secondary antibodies conjugated to fluorescent dyes were obtained from Jackson ImmunoResearch Laboratories, Invitrogen, and Life Technologies. Streptavidin with Alexa Fluor 405 conjugate was from Thermo Fisher Scientific.

Embryos were imaged on an inverted SP8 Confocal Microscope (Leica Systems), with a 40X 1.3NA oil-immersion objective. Either a PMT or an HyD detector was used alongside a 405/488/546/647nm laser line. Image stacks of 1 μm Z separations were captured using the Leica Application Suite X Software.

Movie acquisition

Embryos were mounted ventral/ventrolaterally and a central region of the embryo was imaged on a spinning disc confocal microscope as previously [26]. The anterior edge of the field of view was typically just posterior to the cephalic furrow separating the head from the trunk of the embryo (Fig 1F). Z-stacks, with 1 μm step size, were collected sequentially for each channel. Embryos were imaged every 30 seconds from approximately mid-cellularisation (stage 5) until the onset of cell divisions within the field of view (approximately stage 8). We imaged at a controlled temperature (20.5°C +/− 1°C) to minimise variability, since Drosophila developmental processes are temperature dependent [53,54].

We checked that wild-type embryos survived to larval hatching post acquisition. twi homozygous embryos were identified by failure of ventral furrow formation during movie acquisition and then left to develop until the end of embryogenesis to confirm their phenotype by cuticle preparation (twi is an embryonic lethal gene so homozygous embryos die before larval hatching).

Movie tracking

To track at comparable apico-basal positions in all cells regardless of embryo curvature, acquired z-stacks were converted into stacks of curved quasi-two-dimensional representations, following the curved surface of the embryo at each time point. Next, automated segmentation, with manual intervention to increase tracking accuracy, was performed as previously [26]. We segmented cell contours in the Gap43Cherry channel at the level of adherens junctions (see example S1 and S2 Movies). During GBE, we used the Myosin II signal concentrated at the level of adherens junctions as a guide to ensure we were segmenting at the correct plane. Note that prior to GBE, apical Myosin II is very weak and apical adherens junctions are slightly more basal, based on the ruffled appearance of cell membranes more apically and imaging data of Cadherin-GFP in other datasets from our laboratory. Therefore, in earlier time frames, we used images from up to 4 μm deeper into the embryo to accurately segment cell contours.

Any inaccurately tracked cells were removed prior to analysis based on cell velocity relative to neighbours, cell area, cell area change, and number of frames over which the cell could be tracked.

The ventral midline (middle of the presumptive mesoderm in twist) was followed over time and used to define the orientation of the AP axis, and the DV axis was set perpendicular to this. “Distance from mesectoderm” is defined as the distance from the first row of centroids of ectodermal cells at 30 minutes of GBE.

Note on number of cells in movies

In wild-type movies, initially, the mesoderm takes up a large part of the field of view leaving ectoderm only visible on one side of the embryos at the start of the movies (see example S3 Movie). As the mesoderm invaginates, the ectoderm on the other side of the embryo comes into view, and the number of ectodermal cells visible continues to increase as mesoderm invagination completes and ectodermal cells converge towards the midline during GBE. Note that the ectoderm that is visible often moves slightly dorsalwards and out of the field of view prior to mesoderm invagination, accounting for the reduction in cell numbers between −15 and 0 minutes of GBE (S3D and S3E Fig).

In twist movies, cell numbers vary less throughout the course of the movie because the mesoderm primordia is narrower in twist than in wild-type so takes up less space initially but does not invaginate successfully so still takes up part of the field of view throughout GBE (see example S4 Movie and S3F Fig). Precise cell numbers vary slightly between movies due to differences in exact DV location of mounting of the embryo and slight variations in the number of cells successfully tracked.

Movie synchronisation

Movies of the same genotype were synchronised together using a threshold of 0.01 pp/minute AP-projected tissue strain rate as previously [3]. This nonzero threshold was chosen due to some minor fluctuations in this strain rate around the start of GBE. Average AP-projected tissue strain rate curves were then calculated for each genotype, and used to synchronise movies so that average AP-projected tissue strain rate was 0 pp/minute at time = 0 minutes GBE (S3A and S3B Fig).

For the purposes of checking synchronisation, the appearance of apical myosin in the ectoderm was assessed by eye as signal discernible above background in raw movies (S3C Fig). Cytokinesis timing was assessed through by-eye observation of Myosin II−enriched cytokinesis rings.

Exclusion of mesoderm and mesectodermal cells from analysis

Cell types were initially broadly defined using rules based on their proximity to the midline at a given time point and then were refined by manually selecting cells of incorrect cell-types based on their patterns of cell division within the time-period of analysis and redefining them where necessary. Note that in wild type, the width of the mesoderm is approximately 18 cells, while in twist mutants, the width of the mesoderm is reduced to approximately 8 to 10 cells, and this was taken into account when defining initial midline proximity rules [41] (see examples S3 and S4 Movies).

For cell division timing, we used comparison to fixed embryos with mesectodermal (sim HCR) and dividing cells (anti-phospho-H3) marked to ensure accurate cell type identification throughout the analysed time-period (S1A and S1B Fig). The pattern of cell division was of particular use for the twist movies because the boundary between ectoderm and mesectodermal cells is not as straight as in wild type (S1B–S1B” Fig).

Selection of a 100 μm-wide region to control for variations in AP registration

In contrast to DV, our movies are not registered precisely along the AP axis because of lack of landmarks. To control for variations in AP registration, we compared both the imaged ectoderm for the full field of view and also for a smaller 100 μm “central” region along AP (S5B and S5C Fig). The anterior boundary of the central window was chosen to exclude cells that underwent higher AP cell shape change compared to elsewhere in the field of view, due to the proximity to the cephalic furrow (which typically formed just outside the anterior edge of the field of view) (white dotted line in S5B and S5C Fig). The posterior boundary was positioned 100 μm further along AP, thus defining a central region. This central region was used for quantitative comparison of AP cell shape (Figs 5B and S5D) and AP tissue strain rates (S6C and S6D Fig) between wild type and twist.

Movie analysis

Normalisation of Myosin channel

Background signal was subtracted by setting pixels of intensities up to fifth percentile set to zero for each time point. The fifth percentile was chosen as this removed background signal without removing any structured (i.e., nonuniform) Myosin signal, as assessed by eye. Intensities varied slightly between experiments due to differences in laser intensity, and, therefore, histograms of pixel intensities were stretched: The 98.5 percentile of the data at 30 minutes GBE was set to a greyscale value of 200 with lower value pixels scaled by the same factor and leaving the top 55 values for a tail of high value pixels. We choose to stretch histograms using a reference time point, rather than on a per time point basis, to preserve the observed increase in Myosin intensity over time. Examples images and histograms of movie frames before and after normalisation can be seen in S7A–S7D’ Fig.

Following normalisation, we quantified “nonjunctional” Myosin II from the same apicobasal position in the cell (+/− 2 μm from adherens junctions) as we quantify junctional Myosin II (see above).

Note that “nonjunctional” Myosin II is defined as the remaining pixels in the cell after the definition of junctional and vertex pools of Myosin II (see details in Myosin quantification section above) and will include both cytoplasmic signal and some medial web associated Myosin II. We compared the “nonjunctional” Myosin II intensities between wild type and twist (see S7E Fig) and find that they are very similar, confirming that that we are not inappropriately brightening one set of movies compared to the other.

Statistics and graphical representations

Ribbon plots were generated in R with 30-second bins and profiles smoothed over 3 bins for presentation only (significance tests were calculated on unsmoothed data). We used a mixed effects model to test for differences between genotypes with a p-value of 0.01 as previously [50]. This estimates the p-value associated with a fixed effect of differences between genotypes, allowing for random effects contributed by differences between embryos within a given genotype. Significant differences are highlighted in grey on plots. Ribbons show standard error between genotypes. Spatiotemporal plots were generated in custom software otracks [40]. Other plots (neighbour exchanges, synchronisation plots) were generated in Prism with 1-minute bins.