1. Cell Biology
  2. Developmental Biology
Download icon

Cytoskeletal tension and Bazooka tune interface geometry to ensure fusion fidelity and sheet integrity during dorsal closure

  1. Piyal Taru Das Gupta
  2. Maithreyi Narasimha  Is a corresponding author
  1. Tata Institute of Fundamental Research, India
Research Article
  • Cited 1
  • Views 782
  • Annotations
Cite this article as: eLife 2019;8:e41091 doi: 10.7554/eLife.41091

Abstract

Epithelial fusion establishes continuity between the separated flanks of epithelial sheets. Despite its importance in creating resilient barriers, the mechanisms that ensure stable continuity and preserve morphological and molecular symmetry upon fusion remain unclear. Using the segmented embryonic epidermis whose flanks fuse during Drosophila dorsal closure, we demonstrate that epidermal flanks modulate cell numbers and geometry of their fusing fronts to achieve fusion fidelity. While fusing flanks become more matched for both parameters before fusion, differences persisting at fusion are corrected by modulating fusing front width within each segment to ensure alignment of segment boundaries. We show that fusing cell interfaces are remodelled from en-face contacts at fusion to an interlocking arrangement after fusion, and demonstrate that changes in interface length and geometry are dependent on the spatiotemporal regulation of cytoskeletal tension and Bazooka/Par3. Our work uncovers genetically constrained and mechanically triggered adaptive mechanisms contributing to fusion fidelity and epithelial continuity.

https://doi.org/10.7554/eLife.41091.001

Introduction

Epithelial fusion results in the formation of continuous epithelial sheets from the fusion of two epithelial flanks that are spatially separated by an intervening tissue or a gap, and accomplishes the closure of the neural tube, palate and epidermis in many organisms (Jacinto et al., 2001; Kiehart et al., 2017). In these contexts, fusion must not only ensure stable epithelial continuity but must also maintain morphological (size matching) and molecular (matching gene expression patterns) symmetry between the two fusing flanks. Defects in epithelial fusion morphogenesis can have disastrous consequences on embryonic viability resulting not only from the loss of tissue integrity but potentially also from the failure to maintain symmetry in gene expression patterns. An understanding of the molecular, cellular and physical principles that govern fusion fidelity and impart integrity to the newly formed seam is therefore essential from both fundamental and clinical perspectives.

Embryonic segmentation is a striking example of a molecular or genetic pre-pattern that also has a morphological correlate, and is a hallmark of both vertebrate and invertebrate body plans. Although the gene regulatory networks that govern segmentation were first identified in Drosophila over 40 years ago, the mechanisms that ensure segment continuity and the maintenance of symmetric segmentation gene expression patterns during fusion morphogenesis are remarkably poorly understood. Drosophila dorsal closure is a morphogenetic movement that relies on epithelial fusion and accomplishes the closure of a curved epithelial sheet, the embryonic epidermis, through the meeting of their separated edges in the midline. The embryonic epidermis is morphologically and molecularly segmented along the anterior-posterior axis of the embryo. This segmentation is evident in the morphologically identifiable segmentation grooves and in the columnar expression patterns of segmentation genes during and after fusion. Epithelial fusion during dorsal closure must therefore not only ensure strong and stable epithelial continuity, but also the faithful matching of morphology and segmentation gene expression patterns.

Earlier studies on epithelial fusion driven morphogenetic movements including Drosophila dorsal closure (DC), ventral closure in C. elegans and neural tube closure in the chick have identified the cellular origins of forces that drive fusion. These studies have demonstrated roles for a) patterned and heterogeneous apical constriction in driving contraction of the intervening tissue, b) cell elongation and intercalation in driving the movement of the flanks and c) the supracellular actin cable and actin based lamellipodial and filopodial protrusions assembled in the leading edge cells in enabling proximity and recognition between fusing partners during fusion (Eltsov et al., 2015; Haigo et al., 2003; Heller et al., 2014; Jacinto et al., 2000; Kiehart, 2015; Kiehart et al., 2000; Meghana et al., 2011; Millard and Martin, 2008; Narasimha and Brown, 2004; Nishimura et al., 2012; Peralta et al., 2008; Saravanan et al., 2013; Sokolow et al., 2012; Solon et al., 2009; Toyama et al., 2008). Genetic and biophysical studies on Drosophila dorsal closure have revealed that the contraction of the amnioserosa to which the epidermal flanks are attached provides the major force that brings the flanks into close proximity (Harden et al., 2002; Narasimha and Brown, 2004; Pasakarnis et al., 2016; Scuderi and Letsou, 2005). Additionally, two forces originate in the epidermal flanks: a retarding force in the cells of the lateral epidermis and a driving force in the leading edge (or Dorsal Most Epidermal/DME) cells. The latter has been attributed to the apical supracellular actomyosin cable assembled in the DME cells of the two epithelial flanks at its fusing interfaces (Kiehart, 1999; Kiehart et al., 2000). Recent studies have argued that the actin cable is dispensable for driving closure, but demonstrate an effect on dorsal closure dynamics in its absence and suggest a role for the actin cable in facilitating scar less closure(Ducuing and Vincent, 2016; Pasakarnis et al., 2016). An actin cable is also assembled in wound, ventral and eyelid closure, but its requirement for the latter has been ruled out (Heller et al., 2014; Raich et al., 1999; Rodriguez-Diaz et al., 2008; Williams-Masson et al., 1997). Dynamic, short, actin based filopodia and lamellipodia that emanate from the fusing cell interfaces during dorsal closure are thought to contribute to forces that enable further proximity between the fusing flanks and to cell recognition and adhesion priming between fusing partners (Eltsov et al., 2015; Jacinto et al., 2000; Millard and Martin, 2008).

Surprisingly little is known about the mechanisms that ensure symmetry between the two fusing flanks and establish seamless epithelial continuity during Drosophila dorsal closure (Kiehart et al., 2017). Uncovering these mechanisms is of outstanding importance given the requirement of both stable epithelial continuity and geometric and molecular symmetry for the maintenance of integrity and the subsequent patterning of the structures engaged in fusion. Dorsal closure accomplishes the covering of the dorsal surface of the embryo by the cuticle producing epidermis and the alignment and registry of the embryonic segments Lb-A8. Pioneering studies that visualised labelled Drosophila embryonic epidermal segment compartments fusing during dorsal closure in real time, demonstrated their faithful pairing and alignment, and hinted at the requirement for cell pair matching between the fusing epidermal flanks (Jacinto et al., 2000; Millard and Martin, 2008). Studies based on electron microscopy images revealed filopodial interdigitations between fusing cell partners during Drosophila dorsal closure and suggested that such filopodial interdigitations must enable the fusing cell pairs to recognize each other and establish contact (Jacinto et al., 2000; Eltsov et al., 2015). An alternative possibility is that fusion fidelity is achieved through the spatiotemporal regulation of fusion, ensuring that only one pair of cells is proximate enough to fuse at any given time. Such a model would necessitate the spatiotemporal regulation of distance between the two flanks and of adhesion between the two fusing partners, one pair at a time. Whether this relies on the regulation of adhesion or contractility also remains unclear. Additionally, the nature and regulation of junctional changes that might impart mechanical integrity to and enable seamless continuity of the fused epithelial sheet remain unknown.

Using live confocal microscopy, quantitative morphodynamics and genetic perturbations, we qualitatively and quantitatively examine the progress of fusion of the embryonic epidermis during Drosophila dorsal closure. We investigate the basis of fusion fidelity and epithelial continuity and uncover their dependence on changes in geometry and cell number in the fusing flanks. We further find that cytoskeletal tension and the polarity regulator Bazooka/Par 3 modulate fusing interface geometry and fusion front cell numbers to impart fidelity and enable seamless, stable epithelial continuity. Our findings reveal that fusion is not genetically pre-set to be accurate at the length scale of fusing cell pairs. Instead, we find that active, adaptive mechanisms that are spatially constrained to embryonic segment compartments and are mechanically triggered, contribute to fusion fidelity. Our work also uncovers for the first time, junction remodelling events that accompany epithelial fusion during dorsal closure.

Results

Chronology of segment fusion during dorsal closure

In a dorsal-up view of the embryo (used throughout in this study) with anterior to the left, the upper and lower arcs are the right and left arcs of the embryo. Fusion begins at the anterior and posterior canthi, proceeds inwards and results in the precise registry of the borders of embryonic segments Lb to A8 (Figure 1A). To follow the fusion of individual segment compartments, we labelled the posterior compartment (stripe) of each segment with UAS Actin GFP or UAS GFP driven by engrailed Gal4 (enGal4) in embryos that also carried a single copy of the ubi::ECadherin GFP transgene. This enabled the visualization of cell outlines in both compartments (the anterior compartment is referred to as the interstripe; Figure 1A, Figure 1—figure supplement 1O–P and Figure 1—video 1). At approximately 3 hr prior to the fusion (−190 mins) of the last segment (A3), the leading edge of the epidermis is just visible in a dorsal up view. About two hours prior to fusion (−130 mins), the segments Lb and A7 fuse. The remaining segments fuse after the onset of zippering (−100 mins) in quicker succession (Figure 1B, Figure 1—figure supplement 1N; n = 6 embryos).

Figure 1 with 5 supplements see all
Chronology of segment fusion and pair-wise analysis of fidelity descriptors.

(A) Top: Low magnification image of a Drosophila embryo at the onset of dorsal closure (DC) showing the segments (Lb-A8, posterior compartments/stripes are labelled with GFP) that fuse during dorsal closure (HG- hindgut and AS-amnioserosa). Bottom: Time-lapse images showing the progression of DC in one such embryo (white arrows show the fusion of the Lb and A7 stripes, yellow arrows mark the anterior and posterior canthi). (B) Prospective and retrospective time lines depicting the chronology of fusion during dorsal closure (times are mean ± sd, n = 5 embryos). (C) Epidermal landmarks and the descriptors used to assess fidelity during dorsal closure. DME - Dorsal Most Epidermal cells (also called Leading Edge cells; see Materials and methods). (D) DME cell number dynamics of the upper (red) and lower (blue) arc over the course of dorsal closure in a representative embryo in a retrospective time scale (0-completion of closure, red arrows denote time of equalization between the two arcs). (E) Normalized difference in DME cell number between the fusing arcs over the course of dorsal closure (median ± range, n = 16 embryos). (F) Frequency distribution of DME cell number differences between contralateral stripe and interstripe pairs at t-100 and tF (n = 21 pairs of stripes/interstripes from seven embryos). (G) Normalized DME cell number difference in stripes and interstripes at t-100 and tF (median ± range, n = 21 stripe/interstripe pairs from seven embryos). (H) Variance in DME cell number differences between contralateral and ipsilateral partner stripes and interstripes at t-100 and tF (median ± range, n = 21 pairs of contralateral and 28 pairs of ipsilateral compartments from seven embryos). (I) Length dynamics of the upper (red) and lower (blue) arcs plotted over the course of DC in a representative embryo in a retrospective time scale (0-completion of closure, red arrows denote time of equalization). (J) Normalized difference in arc length between the fusing arcs over the course of dorsal closure (median ± range, n = 16 embryos). (K) Normalized width difference between contralateral partner segments and compartments at t-100 and tF (median ± range, n = 21 pairs from seven embryos). Scale bar- 20 µm. (* - p<0.01, ** - p<0.001, *** - p<0.0001).

https://doi.org/10.7554/eLife.41091.002

Measuring symmetry, fidelity and plasticity in fusion

Earlier work examined the final outcome of fusion of labelled segment compartments during dorsal closure and suggested that segment alignment (inferred from the precise registry of compartment boundaries) must involve cell matching in the two flanks (Jacinto et al., 2000; Millard and Martin, 2008). These studies did not address how symmetric the two flanks were, or whether, when and how is this symmetry achieved. To address these questions, we performed real-time, 3D confocal microscopy of dorsal closure in the embryos described above, and measured the lengths of the upper and lower arcs, the widths of each segment or segment compartment at the leading edge and the number of DME cells in each arc, each fusing segment or segment compartment during the course of dorsal closure. A population level analysis of cell numbers and widths of the compartments of all segments revealed that (i) the posterior compartment has fewer DME cells and is narrower than the anterior compartment and (ii) the posterior compartments of the last fusing segments (T3, A1-A4) exhibit the greatest plasticity, inferred from their heightened ability to modulate cell number and width (Figure 1—figure supplement 1A–M).

In order to obtain an estimate of fusion fidelity, and specifically to assess how symmetric or well matched the fusing right and left epidermal flanks or segments are, we performed a pairwise analysis of lengths and cell numbers between contralateral fusing partners (right and left flanks of the same segment) and measured the differences in these parameters between the fusing partners at multiple time-points during the course of fusion. For the analysis of fusion of segments or compartments, we defined two time points: one at the time of fusion of that segment (tF) and another at an early time point (tE). We also assessed whether fusing (contralateral) partners were more ‘matched’ than ipsilateral (non-fusing neighbours on the same side) compartments (Figure 1C; See Materials and methods).

The evolution of high fidelity during dorsal closure

We first performed a pair-wise analysis of changes in DME cell number and arc length in the two flanks over the course of dorsal closure (Figure 1D and I and Figure 1—figure supplement 2A–B). Contrary to the expectations of a model in which fidelity is pre-set by invariant and equal cell numbers in the two fusing flanks, we found significant differences in both cell number and arc length between the fusing flanks at early time points. For both parameters, significant reductions in the disparity between the two fusing flanks were observed later. Reductions in the disparity in absolute and normalized arc length preceded the reduction in disparity in absolute and normalized cell number (100 and 50 minutes prior to closure for arc length and DME cell number respectively; Figure 1E and J; n = 16 embryos). These results demonstrate that the fusing flanks become more matched or symmetric with respect to DME cell numbers and arc lengths over time (Figure 1—figure supplement 1Q) and hint at the existence of distinct cellular mechanisms that accomplish changes in geometry and cell number to achieve symmetry and fusion fidelity. Surprisingly, mechanisms that contribute to equalization begin to operate well before fusion, suggesting they must operate at a distance.

Differential contributions of segment compartments to equalization in cell number and geometry of the fusing fronts

To identify the cellular mechanisms that contribute to equalization in arc length and cell number between the two fusing flanks and to examine whether arc length and cell number equalization occurred uniformly along the anterior-posterior extent of the flank, we determined widths of and cell numbers in each pair of segment compartments that align and fuse during dorsal closure. For this analysis, we focused on the central segments that showed the maximum plasticity. Indeed, a higher proportion of central stripes showed a reduction in DME cell number disparity between fusing partners than the peripheral stripes or the central interstripes (Figure 1—figure supplement 1E–J; n = 6). On average, the stripes had only half as many DME cells as the interstripes (Figure 1—figure supplement 1A–D; n = 48 pairs).

Within the central stripes, DME cell number differences between contralateral fusing pairs ranged from 2 to 0 at t-100 but reduced to 1 and 0 at tF. Overall, the normalized disparity in cell number between contralateral stripes reduced by two-fold at fusion (Figure 1G). Significantly, while no stripe pairs with a difference in cell number of two were found at fusion, close to half of the stripe pairs examined showed a difference in cell number of one even at fusion (46.2%; n = 21 pairs; Figure 1F and Figure 1—figure supplement 1F–G). This demonstrates that a cell number difference of one between the fusing fronts of the posterior compartment at fusion is surprisingly common. In contrast, no significant reduction in cell number differences was observed between contralateral interstripe pairs during the course of dorsal closure, and differences of one (in approximately 30%) or two cells (in 13%) persisted even at fusion (Figure 1F and Figure 1—figure supplement 1I–J). Indeed, the normalized difference in DME cell number of the stripe pair at fusion was similar to that exhibited by the interstripe at both time points (Figure 1G). This suggests that stripe pairs match cell numbers more effectively than the interstripe pairs at fusion.

In contrast, interstripes more effectively matched the widths of their fusing fronts (a 3 to 4-fold decrease in disparity at fusion) than the stripes, and contributed substantially to the equalisation of segment width (Figure 1K). Together, these results reveal that the anterior and posterior compartments of the central embryonic segments contribute different mechanisms to the temporal evolution of fidelity in fusion, and suggest that their ability to modulate the length and geometry of their fusing fronts may be genetically constrained.

Differences in cell number are sensed and corrected within each segment compartment

The distinct contributions of the anterior and posterior compartments to fusion fidelity suggest that their differential gene expression patterns might contribute to it. We wanted to determine whether each segment compartment behaved autonomously to achieve symmetry in cell numbers. To address this, we examined whether ipsilateral compartments (adjacent stripes/interstripes on the same flank that do not fuse with each other) also exhibit a tendency to reduce disparity in cell numbers between them as we have demonstrated in contralateral partners (fusing partners on opposite flanks). For this, we compared the variance in DME cell number between ipsilateral and contralateral partners. Ipsilateral stripe pairs exhibited a significant increase in the variance in DME cell numbers at fusion compared to contralateral stripe pairs, but no significant differences in the variances in DME cell numbers were observed at early time points or between ipsilateral interstripes at fusion (Figure 1H; n = 21 contralateral pairs and 28 ipsilateral pairs). The observation that contralateral stripe pairs are more ‘symmetric’ with respect to DME cell numbers than ipsilateral stripes suggests that the segment compartment is the smallest unit within which differences in cell number are sensed and corrected. Together, our findings suggest that distinct genetically hardwired mechanisms operating within the posterior and anterior compartments must enable cellular rearrangements that facilitate changes in segment DME cell number.

Cell addition to the leading edge does not reduce cell number disparity but may contribute to width equalization

To determine how differences in cell number between the fusing segment compartments are reduced during the course of dorsal closure, we visualised the evolution of changes in cell number by live confocal microscopy. We observed that all DME cell number changes in the stripe resulted from an increase in cell number in one or both flanks (Figure 1—figure supplement 3A–D). Two sources of cell addition were found: ‘mixer cells’ (MC) - DME cells that are plastic and cross from the anterior compartment and become integrated into the leading edge of the posterior compartment (and turn on engrailed as described [Gettings et al., 2010]) and ‘posterior intercalating cells’ (PIC; Millard and Martin, 2008) – cells that move from a row ventral to the DME cells in the posterior compartment and become incorporated in the leading edge (Figure 1—video 2). Cell addition occurred at different times in different stripes but reduced cell number disparity in only a third of the cases. Cell addition also occurred when there was no initial disparity, and in some cases cell addition increased the disparity (Figure 1—figure supplement 3A–D,A’–D’, E). In every instance, cell number addition was followed by a transient or sustained equalization of compartment width (Figure 1—figure supplement 3A”–D” and F, n = 19 pairs). This suggests that cell addition may be primarily deployed to equalize the width of the posterior compartment.

Collectively, our results so far reveal that cell number matching occurs with less than perfect fidelity within segment compartments and point to the importance of matching fusing front geometry in the face of differences in cell number to ensure segment alignment. Our findings also suggest that mechanisms other than cell addition must contribute to width matching.

Modulation of fusion front width is achieved by changes in fusing interface lengths within each segment compartment

Earlier work had demonstrated that the segment boundaries align at fusion (Jacinto et al., 2000; Millard and Martin, 2008). To determine the mechanisms by which the fusing fronts of the flanks become matched in length, we examined fusion in sqh::Utrophin GFP embryos in which the stripes were marked with nuclear RFP. This revealed that the dorsal-ventral aligned (D/V) interfaces of the DME cells at the posterior compartment (stripe) boundaries became aligned, with the anterior boundary more pronouncedly linear than the posterior boundary, through dynamic alterations in the length of the fusing anterior -posteriorly (A/P) oriented cellular interfaces despite differences in cell number in the fusing compartments (Figure 2—figure supplement 1B and Figure 6—video 3B). This suggests that anisotropic cell shape changes that principally modulate fusing interface lengths must contribute to the correction of cell number and width disparities that persist in segment compartments at fusion.

Spatiotemporally regulated remodelling of cell junctions converts staggered en-face cellular contacts at the fusing front to an interlocking pattern after fusion

Earlier studies have attributed the accurate registry of segment compartments following fusion to filopodia-based mechanisms that ensure cell partner recognition and cell-to-cell adhesion (Jacinto et al., 2000; Millard and Martin, 2008). To identify the nature of the contacts that formed between fusing segments to ensure segment alignment and epithelial continuity, we examined the formation of the fusion front at high temporal and spatial resolution during and after fusion in embryos carrying ubi::ECadherin GFP. We also examined the nature of contacts in fixed enGal4 >UAS GFP embryos post-closure that were labelled with ECadherin and GFP. Our analysis revealed that despite a difference of one or two DME cells in the fusing segment compartments, the compartment boundaries aligned to form '+' shaped junctions. Such junctions formed between the straight, en-face contacts along the anterior-posterior axis in the dorsal midline (A/P oriented) and the aligned D/V interfaces of the cells at the compartment boundaries and ensured that the fusing fronts were now equal in width. Invariably, the shorter of the two fronts elongated till the boundaries were in alignment. In contrast, cells within segment compartments often contacted two cells on the opposite flank forming 'T'shaped junctions. Such junctions were formed between straight, en-face contacts formed along the anterior-posterior axis in the dorsal midline (A/P oriented) and the D/V interfaces of the cells within the compartment (Figures 2A and 5A,D, and Figure 2—video 1). We therefore wanted to examine whether these junctional configurations persist even after fusion.

Figure 2 with 5 supplements see all
Dynamic changes in the geometry and molecular composition of the fusing interface.

(A, B) High resolution time lapse images of embryos expressing ECadh GFP (A; n = 8 embryos) and Utrophin GFP (B; n = 9 embryos) showing the progressive remodelling of en-face contacts at fusion to an interlocking pattern after fusion at the canthus (red arrowheads). The cluster plots alongside A and B show the normalized intensity of ECadh GFP and Utrophin GFP respectively at the fusing interfaces before, during and after fusion (circles indicate individual data points; black lines indicate median ± range; n = 30 sets of 10 interfaces (from five embryos) for ECadh GFP and 37 sets of 5 interfaces each (from six embryos) for Utrophin GFP; (*- p<0.03, **- p<0.001, ***- p<0.0001)). (C) High resolution time lapse images of embryos expressing sqh GFP (C; n = 3 embryos) showing the intensity receding at the midline some interfaces away from the canthus as the interfaces are interlocked (red arrowheads). (D) High resolution time lapse images of embryos expressing Bazooka GFP (Baz GFP; n = 11 embryos) showing the recruitment of Bazooka upon fusion and its progressive enrichment with interlocking at the fusing front (red arrowheads). (E) Top: Embryo expressing Baz GFP (heat map) and enGal4-UAS NLS RFP (white nuclei) showing differential expression of Bazooka along the D/V interfaces (white arrows) and non-leading edge A/P interfaces (yellow arrows) in stripes (black brackets) compared to interstripes. Bottom: Line intensity profiles of the same image at the fusing interfaces and at the D/V interfaces of DME cells. (F) Fixed preparations showing localization of Vinculin (Vinc) in the DME cells during (top) and after (bottom) fusion in wildtype embryos (n = 8 embryos; red arrowheads show Vinculin enrichment at unfused and en-face contacts). (G) Schematic representation of the dynamic changes in the localization of molecules at the fusing interfaces. (H) Pictorial summary of the dynamics of the molecular composition and its correlation with dynamic modulations of interface geometry. Scale bar- 20 µm.

https://doi.org/10.7554/eLife.41091.008

Both kinds of contacts ('+' and 'T') resolved upon fusion to form an interlocking pattern, with the A/P interfaces becoming angled with respect to the dorsal midline to form tricellular junctions with the D/V interfaces. This pattern evolved from the en-face contacts through the gradual sliding and relaxation of straight, horizontal, A/P oriented interfaces, rendering them less taut, and their subsequent repositioning along the adjacent D/V interface which was shifted away from the dorsal midline (Figure 2A). Junction remodelling culminated in the formation of nearly evenly spaced, 'Y'shaped tricellular junctions between two adjacent A/P interfaces and the D/V interface and obliterated the midline seam approximately 15 minutes after fusion. This remodelling was associated with a small decrease in ECadherin GFP intensity in the fusing interfaces as they interlocked (Figure 2A and Figure 2—video 1). These findings reveal that fusion does not always culminate in the accurate matching of cell pairs and suggest that the final conformation of junctions (interlocked, tricellular) confers stability and mechanical strength to the newly formed seam, rendering it less susceptible to tears. Thus, the geometric and molecular remodelling of the newly formed ECadherin-based contacts that includes changes in interface length, position and possibly also tension accomplishes strong and stable epithelial continuity upon fusion (Figure 2G,H). To identify the mechanisms that contribute to the remodelling of contacts between the two fusing flanks, we examined the spatiotemporal organization of the actomyosin cytoskeleton and of regulators of adhesion.

Spatiotemporal regulation of actomyosin organisation: clearance of actin from fused interfaces

The transition in interface geometry that we described above is reminiscent of the changes that accompany cell intercalation during germ band extension in Drosophila, in which the conversion from Type I to Type II junctions involves the resolution of one or more four-point vertices by junction growth in the perpendicular direction. This remodelling restores the formation of the more stable three-point vertices (Bertet et al., 2004; Blankenship et al., 2006; Simões et al., 2010). Regulation of junction length during growth involves both passive relaxation and active cytoskeletal changes (Collinet et al., 2015; Hara et al., 2016). We therefore examined cytoskeletal organization and dynamics at the fusing interfaces during fusion. Changes in the levels of actin (visualised using Utrophin GFP) and myosin (visualised using sqh GFP) accompanied interface remodelling. Specifically, while both were enriched in unfused interfaces and en-face contacts, their levels reduced significantly in angled, bent and interlocked interfaces (Figure 2B,C and Figure 2—video 2, 3). Live imaging with Utrophin GFP revealed that actin is ‘cleared’ from the interlocking interfaces (Figures 2B and 6C, Figure 2—video 2 and Figure 6—video 3A). All three proteins (actin, myosin and ECadherin) were specifically regulated in the fusing interfaces (intensities in Figure 2A–B are normalized to that of D/V oriented interfaces in the same cells). The D/V oriented interfaces of DME cells showed no change in the levels of actin or ECadherin (Figure 2—figure supplement 1E–F). These results suggest that the spatiotemporally regulated, planar polarized distribution dynamics of the actomyosin cytoskeleton and possibly also of ECadherin may contribute to the resolution of the fusing front to a seamless, interlocked geometry.

The polarity regulator Par3/Bazooka is known to influence both actomyosin organization and cadherin dependent adhesion and guide junction remodelling. Indeed, the conversion from Type I to Type II junctions during cell intercalation is associated with the appearance of Bazooka. The mutual exclusion of myosin and Bazooka enables junction shortening in the Type I interfaces that do not contain Bazooka. Conversely, Bazooka-enriched interfaces lengthen (Bertet et al., 2004; Blankenship et al., 2006; Simões et al., 2010). We therefore wished to examine whether Bazooka may also be spatiotemporally regulated at the fusing interfaces during the fusion of the epidermis.

Spatiotemporal regulation of Bazooka recruitment to the fusing interfaces

To determine whether Bazooka may also enable junction remodelling in the context of ‘reepithelialisation’ or the formation of new contacts during epithelial fusion, we examined the distribution of Bazooka during and after fusion. Live imaging of genomic Bazooka GFP (Baz GFP) embryos revealed striking spatiotemporal changes in Bazooka distribution. Bazooka was absent from pre-fusion DME interfaces that were rich in actin and myosin but became recruited to these interfaces as the en-face contacts began to form. Its intensity at these interfaces increased as the contacts began to bend and reposition (11.2 ± 0.3 mins after initial contact; n = 71 interfaces), and peaked when the interfaces became configured to create an interlocking cell arrangement (Figure 2D and Figure 2—figure supplement 1A; n = 11 embryos and Figure 2—video 4). Bazooka colocalised with ECadherin containing cell junctions at the fusing interfaces (Figure 5—figure supplement 1G). These results reveal the complementary patterns of distribution of actomyosin and Bazooka at the fusing interfaces. While the intensities of Bazooka GFP in the fusing interfaces of both the stripe and the interstripe were comparable, Bazooka distribution in other A/P or D/V oriented interfaces was lower in the stripe than in the interstripe (Figure 2E). This differential distribution within segment compartments may regulate compartment specific differences in the cellular behaviours that contribute to fusion fidelity.

Spatiotemporal regulation of cytoskeletal tension: Vinculin clearance from fused interfaces

We wished to determine whether the junctions that Bazooka is newly recruited to might also be more relaxed. For this, we examined the distribution of Vinculin, an adherens junction and focal adhesion protein whose recruitment to the cytoplasmic face of the adherens junction is dependent on force or tension dependent changes in the conformation of the junctional protein α−Catenin (Hara et al., 2016; Seddiki et al., 2018; Yonemura et al., 2010). We examined Vinculin distribution using an antibody that recognizes Vinculin and in embryos carrying a Vinculin GFP transgene. In sharp contrast to Bazooka, Vinculin in the DME cells was enriched in unfused interfaces and in en-face contacts where its highest levels were found in dot like structures. After fusion, the dots were no longer visible at the apical membrane, and lower levels of Vinculin were detectable along the length of the interface (Figure 2F). Live imaging with Vinculin GFP revealed that while the dots observed pre-fusion persist for some time after fusion, they appear to have repositioned basally. This was particularly evident in orthogonal sections from such embryos and in fixed embryos carrying Vinculin GFP (Figure 2—figure supplement 1C–D). This reciprocity suggests that Bazooka enrichment at the newly formed cell junctions is associated with reduced interfacial tension. These findings establish the reciprocal dynamics of actomyosin/interfacial tension and Bazooka at fusing DME cell interfaces (Figure 2G–H). Vinculin GFP also showed a striped expression pattern in the epidermis (Figure 2—figure supplement 1C–D), suggesting that tension within segment compartments might be differentially regulated.

To determine whether the spatiotemporal regulation of interfacial tension and Bazooka contribute to fusion fidelity and seamless epithelial continuity, we used genetic perturbations that modulate cytoskeletal tension or Bazooka levels and distribution within the posterior compartment and examined their effects on fusion fidelity, epithelial continuity and the mechanisms that contribute to it.

Cytoskeletal tension entrains the temporal evolution of fidelity and modulates cell behaviour

The observations described above demonstrate that the supracellular actomyosin cable that assembles along the fusing DME cell interfaces/arcs and provides a force for closure prior to fusion, disappears after fusion. This suggests that there is a progressive relaxation of the fusing interfaces after fusion. We therefore examined the role of tension generated by the cytoskeleton on fusion fidelity and epithelial continuity. For this, we measured the fidelity descriptors described earlier (DME cell numbers and widths of the fusing segment compartment fronts as well as the alignment of segment boundaries) in embryos expressing a contraction defective form of myosin (ZipDN) previously shown to function in a dominant negative manner, in the stripes (Dawes-Hoang et al., 2005; Saravanan et al., 2013) (Figure 3A–B, Figure 3—figure supplement 1A and Figure 3—video 1). We perturbed the stripes rather than just the DME cells since the cell rearrangements we had documented include cells that are not initially at the leading edge. In such embryos, both the fusion of stripe pairs and dorsal closure was delayed (Figure 3—figure supplement 1B–D). Additionally, the arc lengths and DME cell numbers were more variable than controls during early closure (Figure 3—figure supplement 1E–F; n = 16 and 10 for control and UAS ZipDN respectively). Despite this, both parameters reduced in mutant embryos with the progression of closure, albeit with different kinetics compared to control embryos (Figure 3—figure supplement 1E–F). Within individual embryos,neither parameter exhibited the temporal evolution that was observed between contralateral partners in control embryos (Figure 3C–D). Instead, the mutant embryos appeared invariant over time compared to control embryos, and the maximum reduction in disparity was also lower. The stripe and interstripe widths did not show any change between early (t-100) and late (tF) time points (Figure 3E–F; n = 40 stripes/interstripes). Additionally, unlike controls, the interstripes in these embryos also did not show a significant reduction in width disparity (Figure 3G; n = 20 pairs). Thus, cytoskeletal tension generated by actomyosin interactions entrains the temporal progression to high fusion fidelity. ZipDN expressing stripes were also significantly less well aligned at and after fusion compared to controls (Figure 4C; n = 10 pairs each for control and Zip DN). In a fraction of such embryos and in a smaller percentage of embryos expressing a phosphorylation defective myosin regulatory light chain (SqhAA, Jordan and Karess, 1997), fusions with ipsilateral or non-partner contralateral stripes were observed (Figure 4D–E; n = 10 pairs). These results strongly suggest that cytoskeletal tension exerts both a restraining and a permissive force that guides the temporal evolution of fidelity during epithelial fusion.

Figure 3 with 2 supplements see all
Tissue tension entrains the temporal evolution of fidelity in epithelial fusion.

(A, B) Time-lapse images of dorsal closure in control embryos (A), and in embryos expressing UAS ZipDN driven by enGal4 (B). Red arrowheads denote scalloped leading edge. (C, D) Normalized arc length (C) and DME cell number (D) differences in control and in UAS ZipDN embryos (median ± range, n = 16 for control, n = 10 for UAS ZipDN). (E–F) Absolute widths of central stripes (E) and interstripes (F) of control and UAS ZipDN embryos at t-100 mins and at tF (median ± range, n = 40 stripes/interstripes from 10 embryos). (G) Normalized difference in width of contralateral partners at t-100 and tF (median ± range, n = 20 pairs from segments A2-A3 from 10 embryos). (# - p<0.05, *- p<0.01, **- p<0.001, ***p<0.0001). Scale bar- 40 µm.

https://doi.org/10.7554/eLife.41091.014
Tissue tension modulates the alignment and matching of contralateral partner stripes at the midline.

(A, B) Time series images showing the positions of the anterior boundaries (red lines) of a pair of contralateral partner stripes from segment A3 in control embryos (A) and in embryos expressing UAS ZipDN (B). Scale bar-10 µm. (C) Temporal evolution of alignment of contralateral partner stripes in control and UAS ZipDN (mean ± sd, n = 10 pairs from 10 embryos, *- p<0.01, ** -p < 0.001, *** -p < 0.0001). (D) UAS ZipDN and UAS SqhAA expressing embryos showing non-partner fusion (yellow arrowheads) and cis-fusion (red arrowheads). Scale bar- 20 µm. (E) Frequency distribution of fusion defects in control embryos, and in embryos overexpressing UAS ZipDN or UAS SqhAA.

https://doi.org/10.7554/eLife.41091.017

Cytoskeletal tension modulates changes in interface geometry during contact remodelling

We then examined whether cytoskeletal tension also influences junction remodelling after fusion. Expression of either ZipDN or SqhAA in stripes resulted in the poor resolution of junctions to an interlocking pattern after fusion (in 95% of stripes in UAS ZipDN, n = 80 stripes and 96.5% stripes in UAS SqhAA, n = 102 stripes; Figure 5A–C). Surprisingly both perturbations also affected junction remodelling in the interstripe (81.3% of interstripes in UAS ZipDN, and 88.5% interstripes in UAS SqhAA). The two perturbations however produced qualitatively different effects on epithelial organization within the stripe. Whereas the expression of ZipDN produced pronounced epithelial disorganization at the dorsal midline, the expression of SqhAA had a more uniform effect on all interfaces resulting in an unobliterated midline seam, suggesting a block in junction repositioning (Figure 5A–C). Consistent with this, the frequency distribution of angled A/P interfaces in DME cells from stripes expressing SqhAA was significantly lower and that of straight A/P interfaces significantly higher than in control stripes (Figure 5—figure supplement 1E). The angles subtended by the interfaces were also lower than that of controls (Figure 5—figure supplement 1F). A similar analysis could not be done on UAS ZipDN embryos on account of the greater epithelial disorganization at the midline. These results suggest that spatiotemporally regulated changes in cytoskeletal tension in the fusing DME interfaces drive timely interface remodelling. These experiments do not however enable us to separately examine the influence of tension generated by the actomyosin cytoskleton at the fusing interfaces, compartment boundaries (Monier et al., 2010), or elsewhere within the stripe.

Figure 5 with 1 supplement see all
Cytoskeletal tension and Bazooka influence interface remodelling, cell shape and epithelial organisation during fusion.

(A–G) Interface geometries of fusing DME cells and epithelial organisation in embryos expressing UAS GFP (control; A, D n = 8, 14 embryos), UAS ZipDN (B, n = 14 embryos), UAS SqhAA (C, n = 15 embryos), UAS Baz GFP (E, n = 9 embryos), UAS Baz∆ (F, n = 10 embryos) and UAS Baz RNAi (G, n = 14 embryos) after fusion stained for ECadh (magenta) and GFP (marking the engrailed compartment, green). A’- G’ are single channel images showing ECadh localization. Dashed yellow lines mark the boundaries of expressing stripes (bracketed). The dashed red lines in A-G mark the midpoint along the A/P axis (as shown in cartoons inset in A, D). The red boxes in A’-G’ are magnified in i, i’ and i’’. Scale bar- 20 µm. (H–K) Frequency distribution of interface geometries (H–J) and distribution of interface angles (K) of stripe DME cells at the leading edge in embryos expressing UAS GFP (control, n = 135 interfaces, 11 embryos), UAS Baz GFP (n = 123 interfaces, nine embryos), UAS Baz∆ (n = 54 interfaces, seven embryos) and UAS Baz RNAi (n = 126 interfaces, 11 embryos) driven by enGal4. (** -p < 0.01, ***-p < 0.0002).

https://doi.org/10.7554/eLife.41091.018

Modulation of cytoskeletal tension in the DME cells influences the straightness of the fusion seam

To determine whether tension in the whole stripe or just in the DME cells governed its effects on fusion fidelity, we also expressed ZipDN, SqhAA and SqhEE (a phosphomimetic version of the myosin regulatory light chain) exclusively in the DME cells using a leading edge cell specific Gal4 (LEGal4) that is expressed patchily. All three perturbations influenced the straightness of the midline seam formed upon fusion as well as its obliteration after fusion. Specifically, expression of both Zip DN and SqhAA rendered the seam (the blue line drawn by joining the interfaces that contribute to it) less straight (bent or wavy) with respect the anterior posterior axis or dorsal midline (the red line in Figure 5—figure supplement 1A,A’–C,C’). In contrast, the seam in embryos expressing SqhEE was straight rather than jagged (Figure 5—figure supplement 1D,D’). All three perturbations produced patches of defective interlocking (Figure 5—figure supplement 1A”, A”’, D’, D”’), suggesting that the spatiotemporally regulated balance of forces between the A/P and D/V oriented interfaces is critical for orderly interlocking and DME cell spacing both along the D/V and A/P axes.

Bazooka modulates interface remodelling during fusion

As mentioned earlier, the polarity regulator Par3/Bazooka has been shown to influence both actomyosin organization and cadherin dependent adhesion, and to guide the conversion from Type I to Type II junctions during cell intercalation (Bertet et al., 2004; Blankenship et al., 2006; Simões et al., 2010). To examine whether Bazooka might also influence interface remodelling during the formation of new contacts during fusion (reepithelialisation), we expressed GFP tagged full length Bazooka, Bazooka RNAi and a GFP tagged C terminally truncated (Δ969–1464/BazΔ) Bazooka transgene in the stripes. The latter was previously demonstrated to exhibit dominant negative effects in the context of neuronal polarity and shown to prevent its membrane recruitment by interfering with its interactions with phosphoinositides (Krahn et al., 2010; Shi et al., 2003). To examine the effects of Bazooka RNAi in the stripes, we used UAS GFP (in fixed preparations) or UAS NLS RFP (for live imaging) to label the stripe, and sqh::Utrophin GFP to enable the visualization of changes in cell and interface geometry and cytoskeletal organization in real time. We chose Utrophin GFP (rather than ECadherin GFP or sqh GFP) since Bazooka has been shown to regulate both ECadherin and myosin albeit in different contexts. It was therefore possible that ECadherin GFP or sqh GFP may modify the phenotype resulting from Bazooka knockdown.

We first analysed the effect of Bazooka perturbations on interface remodelling in fixed preparations. The expression of Baz RNAi or BazΔ in stripes resulted in a significant increase of fused interfaces that were curved/squiggly compared to controls, and a significant decrease in angled interfaces. In contrast Bazooka overexpression did not significantly alter the distribution of interface types (Figure 5D–J). Further, the average angle of the fusing interfaces was significantly reduced upon expressing Baz RNAi (32.3 ± 2˚) and BazΔ (25.1 ± 2.7˚) compared to the control (39.7 ± 2˚) and Baz overexpression (39.1 ± 1.7˚) (Figure 5K). We also examined interface morphologies in zygotic baz4 mutant embryos. Dorsal closure fails in the majority of such mutants precluding a detailed analysis. However, in the few embryos that do progress through dorsal closure, the seam is characterized by straight interfaces (Figure 5—figure supplement 1J). These results suggest that Bazooka function modulates interface morphology to facilitate interlocking.

We then examined fusing interface morphologies in the three perturbations in real time. For the full length and truncated Bazooka transgenes, we used their GFP tags to visualize interface morphologies at the fusing DME interfaces. As mentioned above, we used sqh::Utrophin GFP to visualize interface morphologies in Baz RNAi. Live imaging of fusion in stripes overexpressing full length Bazooka recapitulated the recruitment dynamics seen in genomic Baz GFP embryos. Overexpressed Bazooka was recruited to enface contacts of DME cells and subsequently became enriched (15–20 minutes after the initiation of fusion) in interlocking interfaces (Figure 6A and Figure 6—video 1A-B). This suggests that Bazooka overexpression does not influence its subcellular distribution dynamics. In contrast both the subcellular distribution and recruitment kinetics of BazΔ differed from the full length Bazooka. Α vesicular, cytosolic pool of BazΔ was present in addition to the membrane pool, and BazΔ recruitment to en-face DME interfaces was both delayed and less tight (11.8 ± 1.2 mins and 20.4 ± 2.2 mins for UAS Baz GFP and UAS BazΔ respectively; Figure 6B, Figure 6—figure supplement 1F and Figure 6—video 2). While en-face contacts quickly interlocked in genomic Bazooka or full-length Bazooka overexpressing stripes (15.9 ± 1.2, 16.4 ± 1.0 mins respectively for genomic Baz GFP, UAS Baz GFP; n = 41 and 47 interfaces respectively), expression of BazΔ prolonged the time taken to interlock (24.6 ± 1.5 mins, n = 26 interfaces; Figure 6—figure supplement 1E). Whereas interlocked interface morphologies were evident 15 minutes after fusion in stripes from control embryos expressing sqh::Utrophin GFP, such morphologies could not be easily discerned in the Baz RNAi expressing stripes even 45 minutes after fusion. Additionally, the fusing fronts of Baz RNAi expressing stripes remained enriched with actin (Figure 6C–D and Figure 6—videos 3A, 4). These results reveal that spatiotemporally regulated changes in Bazooka distribution modulate interface remodelling.

Figure 6 with 6 supplements see all
Bazooka localizes at newly formed junctions and influences interface remodelling and fusion fidelity.

(A–D) Time-lapse images of embryos expressing UAS Baz GFP (A, n = 6 embryos), UAS Baz∆ (B, n = 6 embryos), UAS NLS RFP alone (C, n = 6 embryos) or UAS Baz RNAi and UAS NLS RFP (D, n = 7 embryos) in the stripes. Embryos in C, D also express Utrophin GFP. Grey broken lines in A, B mark the dorsal opening. Magenta arrowheads in A, B show Bazooka localization at the newly formed contacts between epithelial cells from the opposing flanks; magenta arrows (in B) indicate stripe overhangs after fusion and yellow arrows (in B) indicate improperly positioned mixer cells along the A/P axis. Green and yellow arrowheads in C, D indicate respectively, the dynamics of interface remodelling and of Utrophin GFP, magnified in the insets provided alongside (region magnified is marked by the dashed yellow squares). (E, F) Absolute (E) and normalized differences (F) in DME cell number between contralateral partner stripes at fusion (tF) of UAS Actin GFP (control), UAS Baz GFP, UAS Baz∆, Utrophin GFP/enGal4-UAS NLS RFP (UTRN GFP, control) and of UAS Baz RNAi (with Utrophin GFP/enGal4-UAS NLS RFP) expressing embryos (n = 30 pairs of stripes from 10 embryos for control, 15 pairs of stripes from five embryos for UAS Baz GFP, 12 pairs of stripes from four embryos for UAS Baz∆, 15 pairs of stripes from five embryos for UTRN GFP and 15 pairs of stripes from six embryos for UAS Baz RNAi). (G) Constriction index of stripes at fusion in control (enGal4 >UAS Actin GFP, n = 26 stripes from four embryos), enGal4 >UAS Baz GFP (n = 26 stripes from four embryos), enGal4 >UAS Baz∆ (n = 24 stripes from four embryos), UTRN GFP (n = 30 stripes from five embryos) and enGal4 >UAS Baz RNAi (n = 30 stripes from five embryos). (median ±range; *- p<0.01, **- p<0.001, ***- p<0.0001). Scale bar- 20 µm.

https://doi.org/10.7554/eLife.41091.020

Bazooka influences cell dynamics and epithelial organization in fusing stripes

Bazooka perturbations also influenced epithelial organization within the stripe. In Baz RNAi and BazΔ expressing stripes, cells in the stripe appeared to transgress the dorsal midline leading to an obliterated seam. In contrast, Bazooka overexpression within the stripe resulted in greater ‘order’, evident in the arrangement of cell rows (Figure 5E and Figure 6—video 1). These results suggest that Bazooka levels/function influence cell dynamics within the stripe. Cell shape (aspect ratio of the DME cells measured as ratio of D/V by A/P lengths) was significantly altered (p=0.018) upon Bazooka overexpression (mean aspect ratio 4.3 ± 1.3, n = 59 cells from six embryos) compared to control cells (mean aspect ratio 4.9 ± 1.3, n = 58 cells from seven embryos) but no significant differences were observed in BazΔ or Baz RNAi expressing cells (BazΔ: 4.9 ± 1.5, n = 58 cells from six embryos and Baz RNAi: 4.6 ± 1.5 n = 42 cells from six embryos). However, both perturbations marginally increased the number of cells with higher aspect ratio (≥6).

Bazooka entrains actomyosin contractility during epithelial fusion

Two mechanisms have previously been suggested to explain the effect of Bazooka on cell behaviour and junction remodelling: i) the regulation of cell adhesion, and ii) the regulation of actomyosin contractility. We first examined the effects of Bazooka perturbations on ECadherin levels and distribution (Figure 5D–G). Overexpression of Baz or BazΔ in stripes resulted in a reduction in ECadherin intensity in the entire stripe, giving it a distinct paler appearance, compared to either control or the unperturbed interstripes. In contrast, stripes expressing Baz RNAi did not show an obvious reduction in ECadherin intensity. While both control and Bazooka overexpressing DME cells maintained a marginal anisotropic enrichment of ECadherin at the fused interfaces (compared to the D/V interfaces), ECadherin distribution in BazΔ and Baz RNAi expressing DME cells was more isotropic. Overall, the effect of Bazooka perturbations on ECadherin levels was modest at best.

We then examined the distribution of actin in Baz RNAi expressing stripes using sqh::Utrophin GFP. In contrast to control embryos, in which Utrophin GFP appeared to be significantly reduced at fused DME cell interfaces approximately 10 minutes after fusion (10.9 ± 0.9 mins, n = 5 embryos), the fusing fronts in Baz RNAi expressing cells were enriched with Utrophin GFP even 20 minutes (n = 5 embryos) after fusion (Figure 6C–D and Figure 6—videos 3A, 4). This observation uncovers a role for Bazooka in actin remodelling or clearance during fusion and suggests that Bazooka might influence interface remodelling through the regulation of actin dynamics. The presence of GFP tags in the Bazooka overexpression transgenes did not allow the examination of Utrophin GFP at their DME cell interfaces.

Bazooka dependent cell shape changes and rearrangements regulate fusion fidelity

Bazooka perturbations also influenced the modulation of fusing front width and cell number during the evolution to high fidelity. Notably, there was an increase in cell number disparity between partner stripes at fusion, with an increase in the proportion of stripes exhibiting a disparity of two cells with all three perturbations (23.5% and 41.6%, for UAS Baz and UAS BazΔ; and 0% for control; 0% for UTRN GFP and 16.6% for Baz RNAi; Figures 6E–F and 7D). Bazooka perturbations also altered the constriction index of the fusing front (width/cell number) with Bazooka overexpression producing a significant increase and BazΔ and Baz RNAi modestly reducing it (Figure 6G). Despite these changes, stripes overexpressing UAS Baz aligned by transient but pronounced changes in stripe width (Figure 6A and Figure 6—video 1). In contrast, stripes expressing BazΔ or Baz RNAi were kinked, exhibited overhangs at fusion and often contained more than one engrailed negative cell at the leading edge (Figures 6B and 7C,E and Figure 6—videos 2, 4). In a small fraction of Baz RNAi expressing embryos, obvious mismatches of segments and epithelial tears were observed (Figure 7B,E and Figure 5—figure supplement 1H–I). These results strongly suggest that spatiotemporally regulated changes in Bazooka levels, distribution and/or function regulate the adaptive changes in cell number, cell rearrangements and cell shape to contribute to fusion fidelity, epithelial continuity and epithelial order.

Bazooka downregulation reduces fusion fidelity and epithelial order.

(A–C): Post fusion images of the dorsal midline from movies of embryos expressing UAS Baz RNAi driven by enGal4 (embryos also express Utrophin GFP and UAS NLS RFP) showing abnormalities in fusion fidelity including aberrant fusion (A, yellow dashed lines indicate the stripes), overhangs between contralateral partners at fusion (B, dashed yellow line marks the anterior boundary of the stripe) and stripe width disparities (C, black brackets indicate wider stripes and red bracket indicates constricted stripes). Scale bar- 20 µm. (D) Prevalence of defects in fusion, fusion fidelity and epithelial organization observed in the Bazooka perturbations tested. (E) Graphic summary of the cellular, subcellular, molecular and physical changes that enable interface remodelling and ensure fusion fidelity and epithelial continuity.

https://doi.org/10.7554/eLife.41091.027

Tension-dependent Bazooka recruitment promotes intercalatory dynamics and interface remodelling in induced discontinuities

The work described above uncovers a role for the timely recruitment of Bazooka at the fusing interfaces for the molecular and cellular changes that confer fidelity to epithelial fusion. Earlier work on Drosophila germband extension uncovered the mutual exclusion of Bazooka and myosin from A/P and D/V interfaces of cells respectively and suggested mutual negative regulation (Blankenship et al., 2006; Simões et al., 2010). To determine whether cytoskeletal tension might influence the recruitment of Bazooka to DME interfaces, we used laser ablation to create a linear discontinuity within the stripe along its A/P axis during (pre-fusion) and after (post-fusion) the completion of dorsal closure in embryos overexpressing Bazooka in the stripe. This creates two flanks on the same side of the embryo. In pre-fusion stripes, such an ablation did not impede the fusion of the contralateral partners at the dorsal midline. The dorsal part of the cut stripe surged ahead after ablation, and remodelled its fusing interfaces with orderly Bazooka recruitment and the appearance of interlocked interfaces. The cut ends of the stripe were depleted of Bazooka for about 20 minutes after ablation when punctae of Bazooka became detectable. Notably, Bazooka recruitment was detected as the distance between the cut ends reduced, and preceded actual contact between the fusing interfaces (Figure 6—figure supplement 1A–B and Figure 6—video 5). These results suggest that tension release promotes Bazooka recruitment. Similar results were observed when stripes were cut post-fusion except that both the recruitment of Bazooka and the reestablishment of continuity were more rapid presumably due to the lack of interference from fusion as part of dorsal closure (Figure 6—figure supplement 1C–D and Figure 6—video 5). Remarkably in both cases, directed, intercalatory movements of cells within the cut stripes contributed to the establishment of a linear seam. The onset of muscle contraction however precluded the assessment of the precise nature of interface remodelling. These results suggest that Bazooka recruitment, which is influenced by tension, may be a generic regulator of epithelial continuity.

Discussion

Much work has informed our understanding about the nature and origin of forces that drive cell sheet movements during fusion morphogenesis. Biophysical studies in the genetically and cell biologically tractable model, dorsal closure, have uncovered the remarkable resilience of the process to perturbations in the participating tissues and suggested that multiple forces contribute to its progression (Hutson et al., 2003; Kiehart, 1999; Kiehart et al., 2000; Martin and Parkhurst, 2004; Martin and Wood, 2002). While the cellular and molecular origins of the forces have been identified, how fidelity in fusion is achieved and how stable epithelial continuity is established is poorly understood (Kiehart et al., 2017). These are questions of fundamental importance. They address not only how contacts formed during re-epithelialisation become strong and stable but also how the physical sizes of (geometric/morphological symmetry and scaling) and the genetic pre-patterns in the fusing flanks are matched and maintained in the face of large-scale cell movements. In addition to establishing epithelial continuity and integrity, fusion morphogenesis also ensures the positions of cells of specific fates. A failure in fusion can thus have consequences not only on tissue integrity and morphology but also on tissue function. The work we describe here uncovers the cellular and molecular mechanisms that ensure fusion fidelity and impart mechanical integrity to the newly formed seam.

Fidelity in epithelial fusion: the logic and the mechanisms

Our work uncovers for the first time, the principles that govern fidelity in epithelial fusion during Drosophila dorsal closure. Our findings demonstrate that fidelity does not rely on pre-set, invariant and equal parameters (DME cell numbers and fusing front geometry) in the two fusing flanks. Rather, spatially constrained adaptive mechanisms that primarily modulate the length and geometry of the fusing front result in the temporal evolution to high fidelity (Figure 1—figure supplement 1Q). Genetically hardwired mechanisms (segmentation genes) spatially constrain adaptive mechanisms (cell shape changes and cellular rearrangements) to distinct embryonic segments and segment compartments but do not predetermine the exact numbers of cells in each fusing segment or arc. The adaptive mechanisms are modulated by changes in cytoskeletal tension and are mediated by anisotropic cell shape changes that alter fusing interface lengths, and possibly also by cell rearrangements. These mechanisms ensure that fused segments are matched in width despite a cell number disparity of one or two. Our findings thus argue that cell number matching during dorsal closure (Jacinto et al., 2000; Millard and Martin, 2008) operates at less than perfect fidelity.

Establishing stable epithelial continuity by interface remodelling after fusion

Our results also uncover for the first time, changes in interface geometry that accompany fusion. These changes, characterized by the rapid remodelling of horizontal en-face interface configurations formed at fusion to angled interfaces that reposition along the D/V interfaces, result in the formation of tricellular vertices and cell interlocking across the fusing front. This conversion obliterates the straight seam evident at fusion, and we speculate, achieves a more stable junctional configuration that imparts mechanical integrity to the epithelial sheet (Figure 7F). This remodelling presumably also obviates the need to precisely match cell numbers at fusion.

System variables, spatial constraints and the temporal evolution to high fidelity

Our quantitative spatiotemporal analysis of disparities in cell number and geometry in the fusing fronts revealed that neither parameter was pre-set to be identical in the two fusing fronts at the outset. Rather, significant reductions in disparity were observed over the course of closure, first in arc length and then in cell number, suggesting that the system evolves to become better matched. Our findings also uncover error correction mechanisms that primarily modulate of the width of the fusing front to ensure segment alignment even in the face of disparities in cell number that persist at fusion.

The observation that fusing front geometry and cell numbers are differentially regulated along the anterior-posterior axis of the embryo and within each embryonic segment suggests genetic control by homeotic and segment polarity genes. Indeed, homeotic genes of the Abd A/B complex have been shown to regulate cell mixing between anterior and posterior compartments in segments T3-A6, the same segments that contribute to the cell number changes we observe during the evolution of fidelity (Gettings et al., 2010; Roumengous et al., 2017). Thus, genetically hardwired mechanisms must spatiotemporally constrain cellular mechanisms that contribute to fidelity. Whether and how segment polarity genes confer differences in cell behaviour that explain the differential contributions of anterior and posterior compartments of embryonic segments to fusion fidelity remain to be determined.

Our results also reveal that cell rearrangements occur well before fusion, even in the absence of cell number disparity, and coincide temporally with width equalization. Conversely, disparities in cell number at fusion are corrected by rapid modulation of fusion front length. These results argue that both cell rearrangements and cell shape changes enable the equalization of geometry albeit at different times and with different efficiencies/speeds. The early cell number and width modulation we observe suggests that sensing mechanisms must operate at a distance within each segment compartment. The width adjustments that occur after fusion operate up close and align compartment boundaries. These findings suggest that distinct mechanisms separated in time and distinguished by different length scales of operation must contribute to sensing cell number and width within each compartment and enable D/V interfaces at the compartment edges to sense and find each other.

Cytoskeletal tension as a trigger for adaptive mechanisms

Our work suggests that tension generated by the cytoskeleton may serve as a sensing mechanism that sets in motion adaptive mechanisms that contribute to fusion fidelity. Indeed, tension entrained the temporal evolution of all the fidelity parameters examined. Perturbations that lowered tension also affected segment matching in a small fraction of embryos examined, as had been previously observed in heterozygous zip (non muscle myosin heavy chain) mutants that also expressed ZipDN in the engrailed compartment (Franke et al., 2005). How and where tension is sensed and how it triggers these adaptive mechanisms are areas of future investigation.

Bazooka: a key regulator of interface geometry

Our work uncovers the importance of modulating interface geometry for ensuring fusion fidelity and epithelial continuity. Interface length was modulated by anisotropic cell shape changes and ensured fusion fidelity even in the face of large cell number disparities induced by Bazooka overexpression. These results suggest a role for Bazooka in modulating junction length. Consistent with this, Bazooka colocalised with ECadherin at fusing interfaces (Figure 5—figure supplement 1G) and high intensity Bazooka GFP patches shifted dynamically along the length of the interface during junction remodelling, sparing the vertices or tricellular junctions (Figures 2D–E and 6A, Figure 2—figure supplement 1A, Figure 6—figure supplement 1A,C, Figure 2—video 4, Figure 6—video 1 and Figure 6—video 5). These findings suggest that the dynamic repositioning of Bazooka patches modulates interface length and geometry.

Interface remodelling accomplished the repositioning of A/P oriented en-face contacts and enabled cell interlocking. Our work demonstrates that this remodelling depends on the reciprocal spatiotemporal regulation of cytoskeletal tension and Bazooka. Bazooka has been previously shown to contribute to anisotropic cell shape change in the epidermis and to the resolution of unstable junctional intermediates during cell intercalation (Bertet et al., 2004; Blankenship et al., 2006; Bulgakova et al., 2013; Simões et al., 2010). The former has been shown to rely on the regulation of ECadherin turnover. The latter may rely on its ability to exclude myosin from resolving interfaces. Our results identify one potential mechanism that may contribute to Bazooka’s effects on interface remodelling: the regulation of interfacial tension. Our results revealed that Bazooka and actomyosin contractility/cytoskeletal tension (actin, myosin and vinculin) were reciprocally regulated in space and time at the fusing/fused interfaces (Figure 2H). Notably, Utrophin GFP appeared to accumulate at fusing interfaces in cells lacking Bazooka (Figure 6D). These results tempt the speculation that Bazooka may mediate its effects on junction remodelling through the regulation of actomyosin contractility and through it, interfacial tension (Figure 7E). Additionally, Bazooka heterogeneities within a fusing interface may enable the modulation of tension along each interface. Thus, Bazooka may regulate the modulation of interface tension locally and its relative depletion at the vertices may help maintain high tension there. Indeed, Bazooka has been shown to interact either genetically or physically with myosin, the Rho1 GTPase effector ROCK and the Drosophila β-Catenin homolog Armadillo, and the effects we see on actin may be mediated by any of these interactions (Simões et al., 2010; St Johnston and Sanson, 2011). Our results also revealed that Bazooka downregulation has only a modest effect on ECadherin anisotropy. We have however not examined it effects on ECadherin turnover at the fusing DME cell interfaces, which is technically challenging on account of the fact that the interfaces are dynamic. What molecular interactions drive the effects of Bazooka we observe is a matter currently under investigation.

Bazooka also influenced epithelial organization, producing noticeable differences in cell packing within the stripe. While Bazooka over expression produced neatly ordered cell rows, both BazΔ and Baz RNAi increased cell dynamics. These results suggest that the relative distribution of Bazooka at cellular interfaces, which is both genetically hardwired (A/P vs D/V boundary differences) and emergent (fusing interface recruitment and dynamics within the interfaces), controls cell shape and cell dynamics through its effects on interface morphology. Whereas overexpressed Bazooka maintained these differential distributions, the recruitment of BazΔ specifically to the fusing DME interfaces was delayed, enabling us to examine the effect of dysfunctional Bazooka recruitment. The efficacy of BazΔ in disrupting fusion fidelity, epithelial organization and interface remodelling as well as its phenotypic similarities with Baz RNAi allow us to conclude that the spatiotemporal regulation of Bazooka entrains the temporal evolution of changes in interface length and geometry that contribute to fusion fidelity, seamless epidermal continuity and epithelial order.

Adaptive mechanisms in the regulation of fusion fidelity

Initially identified as a regulator of apicobasal polarity, Bazooka’s roles have expanded to include adhesion junction remodelling (during cell intercalation), anisotropic cell shape change and boundary positioning (during the formation of segment boundaries) (St Johnston and Sanson, 2011). Bazooka and myosin also exhibit reciprocally planar polarized patterns within the same cells and can influence each other negatively (Blankenship et al., 2006; Simões et al., 2010). Our ablation experiments suggest that release of interfacial tension is sufficient to recruit overexpressed Bazooka to the wound edge interfaces and to drive healing by intercalatory movements and cell interlocking rather than by wound edge contraction. Combined with the effects that Bazooka knockdown produces on actin, our findings tempt the suggestion that a biomechanical feedback loop operating at the scale of single interfaces facilitates junctional changes that accompany fusion. In this feedback loop, Bazooka negatively influences actomyosin contractility in the DME cells, which in turn promotes Bazooka recruitment (Figure 7F). Potentially, such a mechanism can enable tension-dependent regulation of interface geometry, analogous to what has been suggested for the regulation of pulsed constrictions (Munjal et al., 2015). Together, out results identify Bazooka as a tension sensitive regulator of interface geometry. What molecular interactions promote Bazooka localization and dynamics at fusing interfaces in a tension sensitive manner and enable the modulation of interface geometry remain questions for future investigation.

The spatiotemporal regulation of tension also governs the directional temporal progression of fusion during neural tube closure in the ascidian C. intestinalis. The sequential fusion of individual cell pairs provides the force dynamics (relaxation) necessary for the fusion of the subsequent cell pairs (Hashimoto et al., 2015). Thus, junction relaxation may be a conserved feature in the progression of epithelial fusion. Junction relaxation and elongation have recently been demonstrated to be active processes that are regulated by directed cytoskeletal tension in neighbouring cells (Bardet et al., 2013; Collinet et al., 2015; Hara et al., 2016). Whether junction remodelling during fusion requires tension in neighbouring cells and whether Bazooka is a generic regulator of tension will be interesting to determine.

Conclusions, Significance and Implications

Our work sheds light on the mechanisms that ensure epithelial continuity and fusion fidelity during epithelial fusion. Our results highlight the role of adaptive changes in cell behaviour - anisotropic cell shape changes and cellular rearrangements- in the refinement of spatial patterns established by gene regulatory networks to ensure morphological and molecular symmetry across the newly formed seam. It uncovers changes in interface geometry that contribute to the formation of a strong and stable seam. Our results also identify cytoskeletal tension and the polarity protein Bazooka as key regulators whose mutual dependence and exclusion might enable a self-organised biomechanical feedback mechanism that operates within an interface to entrain the dynamics of cells and cellular interfaces and ensure fusion fidelity and stable epithelial continuity. How general the cellular and subcellular mechanisms that we have identified are in other contexts of reepithelialisation will be interesting to determine.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or referenceIdentifiersAdditional
information
Genetic reagent
D. melanogaster: ubi::ECadherin GFPDrosophila stockOda and Tsukita, 2001.
Kind gift of Tadashi Uemura, KyotoUniversity, Japan
Flybase_FBtp0014096
D. melanogaster: Bazooka GFPDrosophila stockBuszczak et al., 2007 Bloomington DrosophilaStock Centre (BDSC)BDSC_51572 FlybaseFBst0051572
D. melanogaster:
y w sqhAX3; sqh GFP
Drosophila stockBloomington Drosophila Stock Centre (BDSC)BDSC_57144 Flybase_FBst0057144
D. melanogaster: sqh::Utrophin GFPDrosophila stockRauzi et al., 2010.Made by Thomas Lecuit,IBDM, Marseille, France, obtained from Richa Rikhy,IISER Pune, IndiaFlybase_FBal0270167
D. melanogaster:
UAS ZipperDN YFP
Drosophila stockDawes-Hoang et al., 2005.Kind gift of Andrea Brand, University of Cambridge, UKFlybase_FBtp0021364
D. melanogaster:
UAS SqhAA
Drosophila stockJordan and Karess, 1997. Bloomington Drosophila Stock Centre (BDSC)BDSC_64114 Flybase_FBst0064114
D. melanogaster:
UAS Bazooka GFP
Drosophila stockBenton and St Johnston, 2003.Kind gift of Daniel StJohnston, University of Cambridge University, UKFlybase_FBtp0017662
D. melanogaster: UAS Bazooka GFPΔ969–1464Drosophila stockKrahn et al., 2010. Kind gift of Richa Rikhy,IISER Pune, IndiaFlybase_FBtp005
850
D. melanogaster: UAS Actin5C GFPDrosophila stockBloomington Drosophila Stock Centre (BDSC)BDSC_9257 Flybase_FBst0009257
D. melanogaster: UAS Bazooka RNAiDrosophila stockWeng and Wieschaus, 2017.Bloomington Drosophila Stock Centre (BDSC)BDSC_35002 Flybase_FBst0035002
D. melanogaster: engrailedGal4Drosophila stockBloomington Drosophila Stock Centre (BDSC)BDSC_30564 Flybase_FBst0030564
D. melanogaster: w1118Drosophila stockBloomington DrosophilaStock Centre (BDSC)BDSC_5905 Flybase_FBst0005905
D. melanogaster: enGal4-ubi::ECadherin GFPDrosophila stockRecombinant, this studyN/A
D. melanogaster: enGal4-UAS GFPDrosophila stockRecombinant, this studyN/A
D. melanogaster: enGal4-UAS NLS RFPDrosophila stockBloomington DrosophilaStock Centre (BDSC)BDSC_30557 Flybase_FBst0030557
D. melanogaster: Vinculin GFPDrosophila stockVienna Drosophila Resource Centre (VDRC)v318227 Flybase_FBst0491774
D. melanogaster: sqh::Utrophin GFP; UAS Baz RNAi Drosophila stockThis study
D. melanogaster: baz4Drosophila stockBloomington Drosophila Stock Centre (BDSC)BDSC_23229 Flybase_FBst0023229
D. melanogaster:
UAS SqhEE
Drosophila stockBloomington DrosophilaStock Centre (BDSC)BDSC_64411 Flybase_FBst0064411
D. melanogaster:
LEGal4
Drosophila stockBloomington Drosophila Stock Centre (BDSC)BDSC_58801 Flybase_FBst0058801
Antibodies
Anti GFP A11122PrimaryInvitrogenAB_221569
DCAD2 (Anti ECadherin)PrimaryDSHBAB_528120
Anti Vinculin N19 (SC7649)PrimarySanta Cruz BiosciencesAB_2288413

Drosophila stocks

Request a detailed protocol

ubi::ECadherin GFP [ECadh GFP; Oda and Tsukita, 2001], Bazooka GFP [Baz GFP, Bl 51572; Buszczak et al., 2007], y w sqhAX3; sqh GFP (sqh GFP, Bl 57144) (all from Bloomington Drosophila Stock Centre, Indiana University, USA) and sqh::Utrophin GFP [Utrophin GFP/UTRN GFP; (Rauzi et al., 2010); a kind gift from Richa Rikhy, IISER Pune] were used to visualize junction and cytoskeletal dynamics during fusion. UAS ZipperDN YFP [UAS ZipDN; kind gift from Andrea Brand, University of Cambridge; (Dawes-Hoang et al., 2005), UAS SqhAA (Bl 64114) and UAS SqhEE (Bl 64411) [(Jordan and Karess, 1997), both from Bloomington Drosophila Stock Centre], UAS Bazooka GFP [UAS Baz GFP/UAS Baz; kind gift from Daniel St Johnston, University of Cambridge; (Benton and St Johnston, 2003)], UAS Bazooka RNAi [UAS Baz RNAi, Bl 35002; (Weng and Wieschaus, 2017) (from Bloomington Drosophila Stock Centre) and UAS Bazooka GFP Δ969–1464 [UAS BazΔ; kind gift from Richa Rikhy, IISER Pune; (Krahn et al., 2010)] were driven with engrailed Gal4 (enGal4, Bl 30564) or enGal4-UAS NLS RFP (Bl 30557)(both from Bloomington Drosophila Stock Centre) to perturb cytoskeletal tension or Bazooka levels in the posterior compartment of embryonic segments (stripes) along with UAS GFP or UAS Actin5C GFP (UAS Actin GFP, Bl 9257, Bloomington Drosophila Stock Centre). baz4 (Bl 23229, Bloomington Drosophila Stock Centre) was also used to examine the effect of loss of zygotic Bazooka in embryos. Stable enGal4-ubi::ECadherin GFP and enGal4-UAS GFP recombinants were built to visualize the epidermal cells and the stripes. LEGal4 (Bl 58801, from Bloomington Drosophila Stock Centre) was used to drive SqhAA, SqhEE or ZipDN in the DME cells. The stock sqh::Utrophin GFP/CyO; UAS Baz RNAi/TM3SerGFP was built to visualize actin organization upon downregulating Bazooka. Vinculin GFP (VDRC v318227) was used to determine the distribution of Vinculin.

Genotypes examined

Request a detailed protocol
  • Figure 1, Figure 1—figure supplements 1, 2 and 3: w; enGal4-ubi::ECadherin GFP/+; UAS Actin5C GFP/+ to estimate epidermal arc lengths, stripe width and DME cell number.

  • Figure 2A w; ubi::ECadherin GFP was used to visualize ECadherin.

  • Figure 2B w; sqh::Utrophin GFP/+ was used to visualize actin localization in stripes and interstripes.

  • Figure 2C y w sqhAX3; sqh GFP was used to visualize myosin.

  • Figure 2D and Figure 2—figure supplement 1A: y1 w Baz GFP was used to examine native Bazooka localization.

  • Figure 2E y1 w Baz GFP/+; enGal4-UAS NLS RFP/+ was used to visualize Bazooka localization in stripes and interstripe.

  • Figure 2F w1118 was used to examine Vinculin localization.

  • Figures 3 and 4A–C and Figure 3—figure supplement 1: w; enGal4-ubi::ECadherin GFP/+; UAS ZipDN/+ was used to perturb tissue tension. w; enGal4-ubi::ECadherin GFP/+; UAS Actin5C GFP/+ was used as a control.

  • Figures 4D–E and 5A–C: w; enGal4/+; UAS ZipDN/+ and w; enGal4-UAS GFP/+; UAS SqhAA/+ were used to perturb tissue tension. w; enGal4-UAS GFP/+ was used as a control.

  • Figure 5D–G w; enGal4/UAS Baz GFP and w; enGal4/+; UAS Baz GFPΔ969–1464/+ and w; enGal4-UAS GFP/UAS Baz RNAi were used to perturb Bazooka levels/function. w; enGal4-UAS GFP/+ was used as a control.

  • Figure 6A and Figure 6—figure supplement 1A: w; enGal4/UAS Baz GFP was used to visualize Bazooka localization and examine the effect of overexpression in real time with or without ablation.

  • Figure 6B w; enGal4/+; UAS Baz GFPΔ969–1464 /+ was used to perturb Bazooka function and examine its effect in real time.

  • Figure 6C w; sqh::Utrophin GFP/enGal4-UAS NLS RFP was used to visualize stripe cell shape and interface morphologies and actin organization at the leading edge in real time in an unperturbed condition.

  • Figures 6D and 7A–D: w; sqh::Utrophin GFP/enGal4-UAS NLS RFP; UAS Baz RNAi/+ was used to down-regulate Bazooka function in stripes and examine its effect upon interface remodelling and actin organization in real time.

  • Figure 2—figure supplement 1B w; sqh::Utrophin GFP/enGal4-UAS NLS GFP used to visualize actin dynamics in stripes and interstripes.

  • Figure 2—figure supplement 1C,D Vinculin GFP was used to determine the distribution of Vinculin during fusion.

  • Figure 5—figure supplement 1A–D w; LEGal4/+; UAS ZipDN/+, w; LEGal4/+; UAS SqhAA/+ and w; LEGal4/+; UAS SqhEE/+ were used to perturb cytoskeletal tension in the DME cells. w; LEGal4/UAS GFP was used as a control.

  • Figure 5—figure supplement 1E–F w; enGal4-UAS GFP/+; UAS SqhAA/+ was used to assess the distribution of interface morphologies and angles. w; enGal4-UAS GFP/+ was used a control.

  • Figure 5—figure supplement 1G w; enGal4/UAS Baz GFP was used to determine Bazooka localization in the DME cells.

  • Figure 5—figure supplement 1H–I w; enGal4-UAS GFP/UAS Baz RNAi was used to knock down Bazooka in stripes.

  • Figure 5—figure supplement 1J y1 w baz4 was used to assess the effect of zygotic Bazooka loss of function on junction geometries.

  • Figure 6—figure supplement 1E–F w; enGal4/UAS Baz GFP and w; enGal4/+; UAS Baz GFPΔ969–1464/+ was used to assess the time taken to interlock and the time of appearance of Bazooka at fused interfaces in the Bazooka perturbations. y1 w Baz GFP was used as a control in Figure 6—figure supplement 1E.

Staging of embryos

Request a detailed protocol

Embryos were harvested from a 12–16 hr collection. For live imaging, embryos in which the spiracles were completely visible on the dorsal surface and the distance between the contralateral stripes of segments A8 and head segment Lb was approximately 30 μm apart were chosen. The scalloped leading edge of the epidermis was just visible in a dorsal-up view. Also, the anterior end of the hindgut was positioned beneath the amnioserosa at approximately the posterior 1/3rd of the distance between the two canthi. Embryos in which all embryonic segments had fused, the hindgut was completely uncoiled, and the yolk showed three lobes were used as post-closure stage embryos in fixed preparations (Figure 5 and Figure 5—figure supplement 1).

Immunohistochemistry

Request a detailed protocol

Embryos were fixed essentially as described (Narasimha and Brown, 2006) and mounted using Vectashield mounting medium (Vector Laboratories, Inc, CA). The following primary antibodies were used: anti-ECadh antibody (DCAD2, 1:10; Drosophila Studies Hybridoma Bank), rabbit anti-GFP antibody (A11122, 1:1000; Invitrogen) and anti-Vinculin antibody (SC 7649, 1:50; Santa Cruz Biosciences). The following secondary antibodies were used: Alexa Fluor anti-rat 633 (A21094), anti-rabbit 488 (A11034) and anti-goat 568 (A11057) (1:200, Invitrogen). The stained embryos were imaged using 60X oil (NA 1.4) immersion lens on an Olympus FV 1200 confocal microscope. Approximately 20–25 optical slices, 0.3 μm apart were used for maximum intensity projections on ImageJ. Figures were prepared using Adobe Photoshop and Adobe Illustrator (Adobe Systems, USA).

Live imaging

Request a detailed protocol

The selected embryos were placed on 22 mm X 40 mm cover glass (Corning Technologies) in a thin film of Halocarbon oil 700 (Sigma Life Sciences) and imaged using either a Zeiss 710 Meta confocal microscope (Carl Zeiss, Germany) and a Plan-Neufluar 63X oil immersion objective (NA 1.4) or an Olympus FV1000 (60X oil immersion objective, NA 1.4). Optical sections 1–1.2 μm apart that captured the entire amnioserosa, leading edge and DME cells were acquired at a temporal resolution of approximately one to four 3D frames/minute for studies on fusing front length and dynamics. Maximum intensity projections (MIP) at each time point were made using the Image5D plugin of ImageJ and assembled as a time series using ImageJ. For the visualization of molecular dynamics, a smaller field of view encompassing 3–4 central fusing stripes were imaged with a 2–2.5X zoom and projections of 7–8 slices (step size 0.3–1 μm) were used to visualize the geometry of the fusing front. Figures were prepared using Adobe Photoshop and Adobe Illustrator (Adobe Systems, USA).

Quantitative morphodynamic analysis

The parameters and formulae used to quantify the fidelity of fusion are shown in Figure 1C. The measurements were made as described below.

Difference in arc and stripe length and cell number

Request a detailed protocol

The outlines of the arcs formed by the leading edge or fusing front (from the anterior canthus to the posterior canthus) of each epidermal flank were traced out manually using ImageJ from 190 minutes prior to fusion to the end of dorsal closure. The average of six measurements for every frame was used. The normalized arc length difference at a given time t was calculated as follows.

(1) Norm.ΔArclength=(LuL1)/<L>

Lu and Ll are the lengths of the upper arc and the lower arc. <L> is the average of the two arc lengths at each time point and is determined by the formula:

(2) <L>=(Lu+L1)/2

Similarly, the width at the leading edge or fusing front of each stripe or interstripe was traced out manually as the distance between the compartment boundaries identified by GFP, YFP or RFP. Measurements were made at every time frame starting from 100 minutes prior to fusion (t-100) for the central stripes and from the beginning of dorsal closure (early time point/tE) for the peripheral stripes to the end of fusion of the fusing pair (tF). The two starting points were necessary since the peripheral stripes remained invisible in a dorsal-up view 100 minutes prior to their fusion. As a result, the time intervals over which the change in width is measured is different for the central and peripheral stripes (Figure 1—figure supplement 1K and L). For all other analyses, only the central stripes (A2-A4 for Figure 1K and A2-A4 for Figure 3E–G) were analysed and the differences in width in the 100 minute interval were calculated. The normalized width difference at a given time was calculated as follows.

(3) Norm.ΔWidth=(WuW1)/<W>

Wu and Wl are the widths of the stripe or interstripe from the upper flank and the lower flank respectively. <W > is the average width of the contralateral partners at each time point and is determined by the formula:

(4) <W>=(Wu+W1)/2

To determine the difference in total DME cell number between the two fusing flanks, the total number of DME cells in each epidermal flank was counted at five-minute intervals from 100 minutes prior to fusion to the end of dorsal closure. Whole cells that were located between the two canthi were counted and the average of six counts was used. DME cell number counts at earlier time points could not be obtained since the DME cells were not distinctly visible in a dorsal-up view. The normalized difference in DME cell number at a given time was calculated as follows.

(5) Norm.ΔDME=(NuN1)/<N>

Nu and Nl are the total number of DME cells between the anterior and posterior canthi in the upper and the lower flanks respectively. <N > is the average number of DME cells in the two arcs at each time point and is calculated as follows:

(6) <N>=(Nu+N1)/2

DME cell numbers at the leading edge or fusing front of the stripe (the labelled GFP/RFP positive compartment) or interstripe (the unlabelled compartment between two GFP/RFP positive compartments) were counted at five-minute intervals starting from 100 minutes prior to fusion (t-100) for the central stripes and from the early time point tE (beginning of dorsal closure) for the peripheral stripes (sometimes excluding the Lb/A7 segment on account of non-visibility) to the end of fusion (tF) of the fusing pair (Figure 1—figure supplement 1A–B and E–J). For all other analyses, only the central stripes (A2-A4 in Figures 1F–H and 6E–F) were analysed and the differences in cell number in the 100 min interval were calculated. For stripe DME cell numbers, all cells within the labelled compartment including any GFP negative cells contained within it were counted. The normalized difference in DME cell number of a pair of stripes/interstripes (from central segments A2, A3 and A4) at any given time t was calculated as follows.

(7) Norm.ΔDME=(nun1)/<n>

nu and nl are the DME cell numbers in the stripes or interstripes from the upper flank and lower flank. <n > is the average number of DME cells in the fusing pair of stripes or interstripes at time t and is calculated as follows.

(8) n= (nu + nl)/2

The image contrast was artificially increased post-acquisition for clear visualization of interstripe DME cells during counting (Figure 1—figure supplement 1O–P).

Variance of ΔDME of stripe/interstripes

Request a detailed protocol

For each embryo, three pairs of contralateral partner and four pairs of ipsilateral partner stripes and interstripes belonging to segments A2, A3 and A4 were used. The variance of the difference in DME cell numbers (ΔDME) between the contralateral partners in an embryo was calculated as follows.

(9) σ2 = (X-μ)2N

σ2 is the variance in ΔDME of partner pairs at a given time t. X is the ΔDME of each partner pair at time t. μ is the mean ΔDME and N is the sample size.

The average variance of ΔDME of partner pairs from multiple embryos was calculated as follows.

(10) Var.ΔDME=σ2/v

σ2 is the variance in ΔDME in each embryo (equation 9) and ν is the total number of embryos.

Alignment of stripes

Request a detailed protocol

A straight line joining the anterior ends of the posterior compartment of segment A3 between the upper and lower flank was drawn and the angle α subtended by it to the horizontal line (A/P axis) was measured using ImageJ. The degree of alignment (θ) was measured as follows.

(11) θ=I90αI

Constriction index of stripes

Request a detailed protocol

The width of the fusing fronts (W) as well as the number of DME cells (N) in one of the two flanks of the fusing stripe pair from the central segments (A2-A4) was measured at fusion. The constriction index for each stripe was calculated by the formula:

(12) Constrictionindex=W/N

Intensity measurements

Request a detailed protocol

To measure ECadherin or actin intensity at the fusing interfaces, a line was drawn along the fusing A/P interfaces of ten (for ECadherin intensity) or five (for actin intensity) DME cells in the pre-fusion stage and mean intensity was measured using ImageJ. Their intensities at the midline seam were similarly determined at fusion (en-face contacts) and after fusion (interlocked interfaces). The mean fluorescence intensity (I) at each stage was calculated and normalized to the intensity of ECadherin or actin at the D/V oriented interfaces (IDV) of the same cells.

Norm.intensity=I/IDV

Estimation of types of cell rearrangements

Request a detailed protocol

Cell rearrangments and cell number changes at the leading edges of the stripes of segments T2 to A6 were visualised and determined by live imaging of enGal4-ubi::ECadherin GFP/+; UAS Actin5C GFP/+ embryos. Based on the ΔDME between partner stripes before and after the observed cell rearrangement, they were divided into the following types: (i) Type 1: ΔDME = 0 at all times, no addition in either partner; (ii) Type 2: ΔDME remains unchanged, addition occurs in both partner stripes; and (iii) Type 3: ΔDME reduces, addition occurs in the stripe with the lower cell number. To assess the change in stripe width upon cell addition, widths of contralateral partner stripes were measured at three instances: a) approximately 10 min prior to (pre-CR), b) immediately after (at CR) and c) 10 min after (post-CR) cell addition is complete in both stripe partners. The normalized disparity in width was then calculated as given in equation 3 above.

Assessing interface remodelling defects

Request a detailed protocol

To assess the defects in cell interlocking in tension defective fixed embryos (Figure 5A–C), the following stripes or interstripes were considered: (i) stripes (or interstripes) that were at least 10 cells away from the canthi in a closure stage embryo and (ii) all the stripes (or interstripes) except the central most stripe (or interstripe) in a post closure embryo. Stripes or interstripes from segments Lb and A7 were not considered. A stripe (or interstripe) was categorized as interlocking defective only when at least 70% of the cells in that compartment (analysed on one side of the midline) lacked interlocked (Y-shaped/tricellular) interfaces.

In fixed embryos expressing UAS Baz GFP, UAS BazΔ or UAS Baz RNAi in the stripes (Figure 5D–G), the geometries of the fusing interfaces were classified into three categories: (i) ‘angular’ if the interfaces subtended an angle between 20o and 90o to the A/P axis; (ii) ‘straight’ if the interfaces subtended an angle between 0o and 20o to the A/P axis and (iii) 'curved' if the interfaces exhibited concave/convex curvatures or squiggles. Interfaces connecting two D/V interfaces and of at least 1 μm in length within stripes that belonged to segments T2-A6 were considered for analysis.

Measurement of interface angles

Request a detailed protocol

A straight line was drawn along the interfaces (described above) between two vertices. The acute angle that this line subtended to the horizontal was measured using ImageJ.

Laser ablation

Request a detailed protocol

Stripes of segment A2/A3 of embryos expressing Baz GFP in the stripe (enGal4/UAS Baz GFP) were chosen for laser ablation. Laser pulses (at 800 nm wavelength) were generated by a titanium-sapphire pulsed femtosecond laser (MaiTai DeepSee, Spectra Physics) coupled to a Zeiss 710 Meta microscope and delivered on to the sample using a 63X1.4 NA oil immersion objective essentially as described previously (Meghana et al., 2011). The laser power at the sample plane measured using a 10X 0.3 NA dry objective was approximately 720 mW. For ablations, 18% laser power was used with 18 iterations (pixel dwell time/iteration = 8.15µs) over a rectangular ROI positioned a row ventral to the DME cells. The ROI spanned the width of the stripe (approx. 10–15 µm) and was 1.05 µm in height. After ablation, the entire stripe was imaged for approximately 35 to 60 min at a temporal resolution of four 3D frames/minute. To assess the distance between the cut edges, a straight line was drawn between the furthest points on the two edges post ablation (approximately between the centers of the two cut edges) and the length of this line was measured at every time point using ImageJ.

Statistical analysis

Request a detailed protocol

All statistical analyses were done using GraphPad Prism six software. Cluster plots show median ± interquartile range (the mean is indicated by the red dot in Figure 3—figure supplement 1E–F). The Mann-Whitney Unpaired t-test was used to assess the significance of difference between central tendencies. The f-test was performed to assess the significance in variance between the sets of data in Figure 1G. For Figures 1, 2A–B, 3C–G, 4C and 6E–G, Figure 1—figure supplements 1 and 3E–F, Figure 3—figure supplement 1B–F and Figure 6—figure supplement 1E–F, data from multiple live imaging sessions were pooled since it was not possible to analyze a statistically significant number of embryos from a single imaging session. Bar graphs show mean ± sd (Figure 4C, Figure 1—figure supplement 1N, Figure 3—figure supplement 1C) or mean ± sem (Figure 5H–J, Figure 1—figure supplements 1A–B, K–L and 3F and Figure 5—figure supplement 1E). No statistical method was employed to predetermine the sample size. No data were excluded from the analysis.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
    Confocal Microscopy of Drosophila embryos
    1. M Narasimha
    2. NH Brown
    (2006)
    In: J. E Celis, editors. Cell Biology: A Laboratory Handbook (Third Edition). London: Elsevier Academic Press. pp. 77–86.
    https://doi.org/10.1016/b978-012164730-8/50136-2
  35. 35
  36. 36
    Real-time imaging of cell-cell adherens junctions reveals that Drosophila mesoderm invagination begins with two phases of apical constriction of cells
    1. H Oda
    2. S Tsukita
    (2001)
    Journal of Cell Science 114:493–501.
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
    An actin-mediated two-step mechanism is required for ventral enclosure of the C. elegans hypodermis
    1. EM Williams-Masson
    2. AN Malik
    3. J Hardin
    (1997)
    Development 124:2889–2901.
  54. 54

Decision letter

  1. Elisabeth Knust
    Reviewing Editor; Max Planck Institute of Cell Biology and Genetics, Germany
  2. Anna Akhmanova
    Senior Editor; Utrecht University, Netherlands

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Cytoskeletal tension and Bazooka/Par3 modulate interface geometry to ensure fusion fidelity and seamless epithelial continuity during Drosophila dorsal closure" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

All reviewers agree on the fact that there are interesting results presented in this manuscript, but that they are "buried", and the manuscript is difficult to read. Therefore, we must decline to consider the paper further in its present form. If the authors are able to address the points made by the referees below and write the manuscript in a way that allows an easy understanding, we would be willing to consider a revised version. This would be considered as a new submission with no guarantee of acceptance.

Reviewer #1:

Dorsal closure in the Drosophila embryo has been used extensively as a model to unravel the genetic, molecular and biophysical principles of epithelial morphogenesis. During this process, the two lateral epidermal sheets move dorsalwards and eventually fuse at the dorsal midline. Up to now, very little is known about the fusion process, in particular with respect to the precision by which cells of both sides recognise each other in order to form a uniform epithelial sheet.

In this manuscript, the authors set out to describe this process in great detail, focusing on the number and shape of the dorsalmost epidermal cells (DME) and the length covered by cells in stripes and interstripes, i. e. in the posterior and anterior compartments of the segments, respectively. Interestingly, they found that precision emerges over time, in particular with respect to the arc length, but that a certain degree of "imprecision" with respect to the cell number in contralateral stripes/interstripes is tolerated. Addition of cells from neighbouring compartments or from more ventral areas reduced the discrepancy in cell number in only some cases. They move on to show that the geometry of the junctions after fusion changes, from an "en-face" geometry to an "interlocked" pattern. This change is associated with a decrease of actin, Sqh and vinculin expression, being high in unfused cells, lower in en-face fused cells and even lower in "interlocked" cells. They further show that the proper change in junctional geometry depends on tension. As shown in other processes, e.g. germ band extension, expression of Bazooka (Baz)/Par3 shows an opposite behaviour, in that it is highest in "interlocked" cells. Upon overexpressing a truncated version of Baz, which is unable to localise at the membrane, this process is affected and goes along with reduced ECadherin expression, precocious interlocking, increased cell number disparity and changes in stripe width. Overall, most of the data presented here, which were obtained by an extremely careful analysis, add novel very important aspects for understanding epithelial fusion, showing that precision is not pre-set, but emerges.

The experiments are carefully described, and the data are very well presented. However, I have a problem to follow the conclusion made on the role of Baz in this process. From the experiments presented in Figure 5E-5G they suggested that "Bazooka enrichment is associated with reduced interfacial tension". To corroborate this assumption, they first showed that impaired tension (achieved by expression ZipDN or SqhAA) affects junction remodelling. Do the authors have any idea why dominant negative Zip and Sqh have different effects on the process?

They go on to express either GFP tagged full-length Baz, a GFP tagged C-terminally truncated Baz (Baz[Δ969-1464]) or Baz-RNAi. Expression of either full-length Baz or Baz[Δ969-1464] resulted in similar phenotypes, including weaker ECadherin staining, whereas expression of Baz-RNAi had only mild effects, but showing increased E-Cad staining. Baz-GFP was recruited with a similar dynamics as endogenously-tagged Baz. In contrast, Baz[Δ969-1464] recruitment was delayed, which was associated with an extended time required to reach the interlocked stage. Finally, they induced a wound in the epidermis on one side of the embryo and observed the fusion event in the wound. From all this they conclude that "… changes in interface length and geometry are dependent on regulation of cytoskeletal tension and Baz/Par3" (Abstract). While the data are convincing with respect to the importance of cytoskeletal tension in this process, the role of Baz is unclear. The fact that overexpression of Baz versions, in particular the truncated one, impairs this process, does not imply that Baz plays a role in this process, given that expression of RNAi has a different outcome. Rather, expressing this form results in a dominant phenotype, but does not provide convincing evidence on the role of Baz in this process.

Reviewer #2:

This manuscript by Das Gupta and Narasimha employs an elegant combination of genetics, live imaging, and morphometric analyses to study re-epithelialization in Drosophila dorsal closure. The authors first perform a detailed descriptive analysis of segment matching in terms of cell number and cell shape and uncover a role for myosin regulation (or, in their terms, tension) in this process. The authors then explore the resolution of junctional geometry following fusion. The results suggest that reciprocal regulation of Bazooka/Par-3 and vinculin localization downstream of cytoskeletal tension are required to effectively restore epithelial integrity.

While the study includes an impressive volume of detailed, quantitative analyses, the manuscript suffers from poor organization and lack of structure, which makes it challenging for the reader to tease apart the real significance of the authors' contributions. The finding that cell numbers between opposing sides of the embryo are not initially precisely regulated is interesting, as are the authors' data on reorganization of junctions following fusion. However, these two components of the story have not been well melded together, and it is difficult for the reader to see the forest for the trees. Despite these qualms, the videos are stunning, and the experiments are, for the most part, well controlled and carefully presented in the figures and captions-too much so at times, in terms of the amount of detail presented.

Major Concerns:

1) In the Introduction and Discussion section, the authors could do a better job explaining how the aspects of dorsal closure they are studying might have implications beyond this particular system. Concepts key to the authors' work, such as fidelity and precision, as they apply to this particular system, should be more clearly defined in the Introduction.

2) The authors perturb myosin as a way to make inferences about the role of tissue tension in this process. However, their strategy is somewhat problematic, as it relies on constitutive expression of a DN-zipper construct in engrailed stripes. While it is likely that tension does play a role in this process, it is difficult to interpret whether the observed changes in precision parameters occur because of a direct role for actomyosin-based tension in cellular rearrangements (which I believe that the authors want to imply) or that other aspects such as cell proliferation might be impaired. To address this problem, the authors should treat embryos with ROCK inhibitor Y-27632 during dorsal closure, which will allow more temporal control over manipulation of contractility.

3) As written, the manuscript is largely descriptive and difficult to digest. The novel points of the paper, although beautiful, are subtle and might be lost on readers unfamiliar with the extensive literature on dorsal closure. In its current form, this study may be better suited for a more specialist audience than the readers of eLife. This outcome would be a shame because the problem of re-epithelialization is of broad interest, and the authors' observations regarding interface remodelling after fusion are interesting, but they are regrettably buried in a sea of difficult-to-follow quantifications of unclear broader biological significance.

4) To begin to unclutter the story, Figure 1 and Figure 2, a tour de force of measurements and quantifications, could be simplified, with many of the detailed comparisons between stripes and interstripes removed or moved to supplemental. Similarly, Figure 7 and Figure 2—figure supplement 1 could be presented with fewer time points, just those illustrating the authors points. It is hard for the reader to digest so many panels.

Reviewer #3:

This manuscript focuses on the important problem of fusion fidelity and seamless epithelial continuity in the dorsal closure model for epithelial cell sheet migration and fusion. The authors find that differences in the length of the leading edge (here referred to as "arc length") of the fusing cell sheets is resolved long before fusion and that small differences in arc cell number are tolerated at fusion. The authors conclude that precision evolves over time and because arc length matches before cell number, suggest that two processes lead to these emergent properties. Such differences in arc cell number are accommodated by cell shape changes that are modulated by cell tension. When tension is perturbed by the expression of dominant negative myosin, matching is also perturbed. The authors also have documented changes in the structure of the formed seam with time, a process they call remodeling, and show that an initially flat interface between the zipped cell sheets becomes interlocked. Interestingly, the process of interlocking is accompanied by a loss of cadherin, actin and myosin at the interface between cells that came from contralateral sides of the embryo. The decrease in myosin occurs concomitantly with an increase in Bazooka/Par3, whereas patterns of vinculin accumulation is more similar to myosin. The authors conclude that such patterns of myosin and Bazooka accumulation are consistent with reduced tension in the formed seam. As described below, I do not understand how the observations that accompany attempts to reduce tension in stripes influence junctional remodeling, at least in part because I find Figure 6 very hard to understand and I find comparing the heat maps in Figure 7 hard to compare to the presentation of data in the rest of the Figures.

Overall, this is an interesting topic and a revised manuscript would make a compelling publication in eLife. My sense is that modifications to the presentation, and few if any new experiments, will make it ready for publication.

Specific comments:

The only way the authors could do these experiments were through live imaging and the data set they present is rich and valuable. Nevertheless, such live imaging requires the expression of fluorescently tagged cadherin, to label cell junctions and GFP-actin to delineate stripes vs. interstripes. Can the authors be sure that the expression of these transgenes doesn't affect the processes they are studying? Perhaps analysis of fixed and stained, control and experimental embryos could confirm that at least the overall pattern of mismatching, segment and intersegment widths and cell numbers is preserved? The analysis I envisage is to replicate the measurements on several fixed embryos stained with antibodies to cadherin and engrailed, to confirm that the patterns are not distorted in control vs. experimental animals.

Critical reading of the Ducuing and Vincent, (2016) and Pasakarnis et al., (2016) papers suggest that the authors of both studies have over interpreted their data to rule out a contribution of actin-myosin rich cables to wild type closure. Numerous other studies show that the cables are not strictly required for closure, but nevertheless normally contribute to the process. Similarly, the cables present but not necessary for wound healing. This does not mean that they do not normally facilitate the process. The authors should amend their Introduction appropriately. Indeed, the authors own observations that show expressing dominant negative myosin constructs in epithelial stripes slows closure is consistent with (but does not prove) a role for the purse strings in closure.

It is not clear to me why that authors normalize the difference in upper and lower arc lengths using Norm. ΔArc Length = (Lu – Ll)/(Lu + Ll). Shouldn't the ΔArc Length be normalized with the length of one arc, not two? That is, it would seem that (Lu – Ll)/((Lu + Ll)/2) is a more appropriate measure of the normalized arc length difference. A similar issue arises from the calculation of Norm. ΔWidth and with DME cell number. I think equations 4 and 5 are more appropriate normalization schemes. It seems that the authors' normalization scheme in equations 1, 2 and 3 underestimates all differences by two fold.

Shouldn't the authors be evaluating standard deviation rather than standard error of the mean? I'm not an expert on statistics, but see Barde and Barde, 2012, Perspectives in Clinical Research 3: 113.

Subsection “Population analysis of precision descriptors” states: "(control in Figures S3E and S3F)". I don't understand how the data in the two cited panels, which pertain to embryos expressing a dominant negative form of non muscle myosin constitute a "control". The authors should explain.

Subsection “Population analysis of precision descriptors”: there is an "are" before "remain" that should be deleted.

Subsection “Cell addition to the leading edge contributes to cell number and width equalization”. The authors conclude "This suggests that cell addition may be deployed to equalize number or width of the posterior compartment." I don't see how, taken at face value, that sentence is consistent with the observations reported in the rest of the paragraph. At the very least, it is misleading since cell addition occurs even when no disparity exists (previous to addition). The wording here should be amended to better reflect the observations reported in the remainder of the paragraph and accurately reflect a consistent interpretation of the data.

Subsection “Subsection “Cytoskeletal tension contributes to the evolution of precision and prevents segment misalignment”. On a sentence that starts with "Furthermore…" and focuses on the role of tension in cell matching, the authors describe the effects of the expression of dominant negative myosin in stripes and note that in one quarter of embryos there were fusions with "non-partner contralateral stripes". Such non-partner fusions were described by Franke et al., 2005 in embryos that had compromised myosin function (see their Figure 7). The authors should cite that work.

I am having trouble understanding how the experiments described in subsection "Bazooka and tension-dependent cell interlocking contributes to establishment of a continuous epithelial sheet" fits with their model of how tension influences remodeling of the formed seam. The authors show that cadherin, myosin, actin and vinculin are all decreased at the interface of the epidermal cell sheets from opposite flanks once the flanks are zipped. Moreover, they show that Bazooka is increased concomitantly with the decrease in myosin. They argue that tension is also decreased, and this allows an initially straight seam to become zig-zagged or interlocked. But here, they attempt to reduce tension by expressing either dominant negative myosin heavy chain or non phosphorylatable light chain and this disturbs remodeling. Shouldn't it do the opposite (i.e., like the over expression of Bazooka, described later in subsection "Bazooka and tension-dependent cell interlocking contributes to establishment of a continuous epithelial sheet")?

I find Figure 6 very confusing. First, D – F are labeled "closure" and G-I are labeled post closure, but each panel has a label "zipping" with an arrow. Isn't all zipping done post closure? Similarly, panels A, B and C have cartoons that are sketches of the interfaces seen in the micrographs and they too are labeled with zipping and arrows. But those arrows, in the control panel A are beneath mature interlocked cells, whereas the middle has T or + boundaries that show cells early in fusion. The authors clearly understand what they are trying to show. They need to figure out how to modify the Figure so it is more easily interpretable by the reader. I find the magnified inserts and schematics very helpful in A, B and C (but NOT the zipping arrows, which I don't fully understand). Why not have all of the panels comparable to A, B and C? I don't understand the point of showing zipping in two directions in A, B and C. The authors describe no differences between the anterior and posterior seams formed, so why bother showing both? The layout of panels, D – F is compelling because the dorsal opening is shown so it helps the reader orient the panel properly. Why not do that for A, B and C? Finally, are each pair D and G, E and H and F and I from the same embryo? If not, why not?

The authors wait until the Discussion section to tell us what the significance of a truncated form of Baz is. They add a sentence in the results to inform the reader.

I don't understand why the authors have switched to a heat map presentation in Figure 7. Of course, heat maps are useful for highlighting relative intensities, but the bulk of the rest of the paper depends on inverse fluorescent gray scale to depict cellular morphologies. So that the reader can better understand the effects of the various permutations of baz expression, the authors should present the data as they have for the other Figures. If heat maps are important, put them in a supplemental figure.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for submitting your revised article "Cytoskeletal tension and Par3 tune interface geometry to ensure fusion fidelity and sheet integrity in dorsal closure" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Anna Akhmanova as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

This is a potentially interesting paper that addresses cell matching during the process of dorsal closure. It is from an accomplished lab that has contributed key findings to the field. The authors show that initially, the lateral epidermal cells that surround the dorsal opening are different in number and the perimeter of the dorsal opening measured on the right side of the embryo is different than the left side. The embryo adjusts length earlier than cell number and fusion between advancing cells sheets can occur even if the number of cells is not perfectly matched. The authors go on to show that an initially T or + seam that forms becomes interlocked and jagged. They go on to explore what might be responsible for regulating the process and investigate both tension (by perturbing myosin function and with laser cuts) and by manipulating Baz/Par3. They find effects with both, some of which are new and some of which are not (for example, the finding that DNZip causes segment misalignment). Overall, this is a very careful analysis, and the data add novel and important aspects to our understanding of epithelial fusion, showing that precision is not pre-set, but emerges.

This revised version of the manuscript is a major improvement to the manuscript submitted previously. The manuscript is easier to read, since they have put quite a few data into the Supplementary material. The figures are also significantly de-cluttered and it is now easier to digest the main points. We also appreciate the authors' efforts to address our point regarding better spatiotemporal control of their myosin experiments – we agree there is no easy way to do this without the development of new tools (ie. optogenetics as they point out) – and we concede that the alternative Gal4 driver (LE-Gal4) is a good compromise.

However, although the reviewers agree that the manuscript is easier to read, there are some concerns summarised below:

- If the authors could justify the assertions they summarize in their manuscript's Abstract, the paper would be a very nice contribution. Unfortunately, it is not clear that the assertions made are justified based on the data presented. Furthermore, the writing can still be improved. The reader is still left with a lot of detailed data. In order not to get lost, it would be helpful to add a summary sentence after each paragraph, if possible.

- The new Figure 5 and Figure 6 are still very difficult to follow. Is it possible to reduce the number of panels or perturbations in the main figures and move the remainder to supplementary figures? Some of the difficulty arises from the lengthy labels, e.g., "enGal4>UAS GFP," "enGal4>UAS ZipDN," "enGal4>UAS BazGFP ∆969-1464." Figure 5H is still rather confusing, in particular due to the inclusion of the significance within the columns. The authors may consider a modification of this graph, perhaps by showing three individual columns for each genotype, with the significance on top. Can the authors come up with more compact and understandable ways to label the different scenarios?

- A very interesting finding that is not well, if at all, highlighted is when equalization occurs. If the reviewers understand Figure 1—figure supplement 1Q correctly, don't the data suggest "action at a distance"? If geometry equalization occurs so early and number equalization occurs later, but still there is tens of microns between advancing cell sheets, how does one row of DME cells "know" to adjust its length and/or? Please clarify.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Cytoskeletal tension and Bazooka tune interface geometry to ensure fusion fidelity and sheet integrity in dorsal closure" for further consideration at eLife. Your revised article has been favorably evaluated by Anna Akhmanova (Senior Editor), a Reviewing Editor, and one reviewer.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below in reviewer #2's comments. As this version is the second round of submission, at this point, we would encourage you to hire a professional scientific editor to help clean up the manuscript. Otherwise the hard work that you have put into your experiments and analyses becomes obscured, which would be a great loss to the readership of eLife.

Reviewer #2:

I greatly appreciate the changes the authors have made to restructure the presentation of data, in particular by moving some of the data to supplementary material. However, before acceptance to eLife, this manuscript's writing still needs significant work. Many sentences contain vague modifiers. As one of the reviewers pointed out, each paragraph would benefit from a clear topic sentence that segues from the previous paragraph and introduces the main point of the new paragraph. For clarity, many sentences with multiple clauses could be made into multiple separate sentences.

General comment: the errors in measurements should be reported to the same decimal place as the value reported, not 11.81 {plus minus} 1.2 but rather 11.8 {plus minus} 1.2 (subsection “Bazooka modulates interface remodeling and DME cell dynamics during fusion” and other places).

Regarding the Discussion section, some of the comments are unfounded, such as "thermodynamically stable junctional configuration" (subsection “Fidelity in epithelial fusion: the logic and the mechanisms”), especially without a citation.

In the Discussion section, it is not clear to me what mechanisms are "emergent" and what the authors mean by this word. I suggest scaling back the claim of emergent mechanisms.

Sometimes the authors write "E Cadherin" or "E-cadherin" or "Ecadherin." These inconsistencies are a result of sloppy editing.

Throughout the manuscript, the authors write "this" without identifying a noun that follows that word. Such sentence structure is sloppy, as it is rarely clear which previous ideas the authors intend the word "this" to mean. A simple search for "this" and adding a noun (e.g., "observation," "result," "concept," etc.) would strengthen the clarity of these sentences.

I have outlined some example writing suggestions below, but this careful scrutiny needs to be applied to the entire manuscript, likely with the help of a professional editor.

Abstract:

- I believe the Abstract is still poorly written. For a broad journal like eLife, it should be accessible to a general life-sciences audience.

- Unclear writing: "We investigate this…": This what? Many ideas were presented previous to that sentence.

- "We find that although differences in cell number and length between the two fusing arcs reduce as fusion progresses, small differences in cell number between fusing epidermal segments persist at fusion" – this sentence is extremely confusing, and a general life-sciences audience will not know what the arcs are.

- The word "Rather" negates the previous sentence, which was also negated by the word "Although." This writing is unclear. Is it a positive result or a negative result that "small differences in cell number persist"?

- "Segment compartment boundaries" are not defined. "Front length" of what?

- "Emergent mechanisms" – it's not at all clear in the Abstract what the emergent mechanisms are.

Introduction:

- Even the first sentence is confusing. The comma after "healing" should not be there; the comma's presence forced me to read the sentence a number of times before I could follow what the authors were trying to say.

- The verb should be "has" because the subject of the sentence is the singular "study".

- The comma in” cell partners during Drosophila dorsal closure and eyelid closure, suggested that” is also incorrect and confusing, and the whole sentence should be rewritten for clarity.

- "It" is vague.

- What do the authors mean by "emergent"?

- This comma also is incorrect.

- At what time scale and in what process are the authors referring to "de novo"?

Results section:

- Is the list of measurements in subsection “Defining and measuring symmetry, fidelity and plasticity in fusion” missing the word "and" somewhere?

- Subsection “Defining and measuring symmetry, fidelity and plasticity in fusion”: what are "measures of plasticity in cell number"?

- The sentence in subsection “The emergence of symmetry and fidelity during dorsal closure” is confusing because of the use of "later" and "preceding" in reverse order. Can the authors rewrite this sentence more clearly?

- Subsection “The emergence of symmetry and fidelity during dorsal closure”: To what "mechanisms" are the authors referring?

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: better sentence structure would have the modifier "For this analysis" at the beginning of the sentence instead of the end.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: what is a frequency of stripes? Do the authors mean number of stripes?

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: I don't understand the anthropomorphizing word "tolerated" here. What is tolerating one fewer cell?

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: comma needed after "pairs"; for sentences with two independent clauses separated by a conjunction, there needs to be a comma before the conjunction.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: because of how confusingly written this paragraph is, the two different mechanisms are obscured.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: "them" is an unclear pronoun.

- Subsection “Spatiotemporally regulated remodeling of cell junctions converts staggered en-face cellular contacts at the fusing front to an interlocking pattern after fusion”: what is "their" referring to?

- Subsection “Spatiotemporal regulation of tension and Bazooka at fusing interfaces”: missing a period.

- Subsection “Spatiotemporal regulation of tension and Bazooka at fusing interfaces”: since supplementary figures are possible in eLife, why are these data not shown?

- Subsection “Spatiotemporal regulation of tension and Bazooka at fusing interfaces”: why is this finding surprising?

- Subsection “Cytoskeletal tension entrains the evolution of precision and modulates cell behaviour and interface geometry”: which parameters?

- Subsection “Cytoskeletal tension entrains the evolution of precision and modulates cell behaviour and interface geometry”: should be "were" instead of "was".

- Subsection “Cytoskeletal tension entrains the evolution of precision and modulates cell behaviour and interface geometry”: what do the authors mean by "evolution"?

- Lines Results section, Discussion section: some letters are missing.

- Subsection “Bazooka modulates interface remodeling and DME cell dynamics during fusion”: whereas, not "where as".

https://doi.org/10.7554/eLife.41091.030

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Reviewer #1:

Dorsal closure in the Drosophila embryo has been used extensively as a model to unravel the genetic, molecular and biophysical principles of epithelial morphogenesis. During this process, the two lateral epidermal sheets move dorsalwards and eventually fuse at the dorsal midline. Up to now, very little is known about the fusion process, in particular with respect to the precision by which cells of both sides recognise each other in order to form a uniform epithelial sheet.

In this manuscript, the authors set out to describe this process in great detail, focusing on the number and shape of the dorsalmost epidermal cells (DME) and the length covered by cells in stripes and interstripes, i. e. in the posterior and anterior compartments of the segments, respectively. Interestingly, they found that precision emerges over time, in particular with respect to the arc length, but that a certain degree of "imprecision" with respect to the cell number in contralateral stripes/interstripes is tolerated. Addition of cells from neighbouring compartments or from more ventral areas reduced the discrepancy in cell number in only some cases. They move on to show that the geometry of the junctions after fusion changes, from an "en-face" geometry to an "interlocked" pattern. This change is associated with a decrease of actin, Sqh and vinculin expression, being high in unfused cells, lower in en-face fused cells and even lower in "interlocked" cells. They further show that the proper change in junctional geometry depends on tension. As shown in other processes, e.g. germ band extension, expression of Bazooka (Baz)/Par3 shows an opposite behaviour, in that it is highest in "interlocked" cells. Upon overexpressing a truncated version of Baz, which is unable to localise at the membrane, this process is affected and goes along with reduced ECadherin expression, precocious interlocking, increased cell number disparity and changes in stripe width. Overall, most of the data presented here, which were obtained by an extremely careful analysis, add novel very important aspects for understanding epithelial fusion, showing that precision is not pre-set, but emerges.

The experiments are carefully described, and the data are very well presented.

We thank the reviewer for these positive comments on the manuscript. Indeed, very little is known about whether and how precise cell matching occurs between the fusing flanks. Our work shows for the first time that while segment/compartment boundaries are respected, small differences in cell number between the fusing compartments is well tolerated. This begs an understanding of the nature of contacts formed upon fusion and motivated us to uncover the morphological/geometric and molecular changes that accompany it. Our work demonstrates for the first time that the contacts formed upon fusion remodel to generate an interlocked cell packing geometry at the dorsal midline. We also identified Bazooka and cytoskeletal tension as regulators of cell morphology, cell dynamics and interface remodelling. Both are themselves spatiotemporally regulated during the process, in a reciprocal manner.

However, I have a problem to follow the conclusion made on the role of Baz in this process. From the experiments presented in Figure 5E-5G they suggested that "Bazooka enrichment is associated with reduced interfacial tension". To corroborate this assumption, they first showed that impaired tension (achieved by expression ZipDN or SqhAA) affects junction remodelling. Do the authors have any idea why dominant negative Zip and Sqh have different effects on the process?

This is an interesting question that we also pondered about. ZipDN has stronger effects on epithelial order but milder effects on interlocking. In contrast, SqhAA has stronger effects on interlocking and milder effects on epithelial order. One explanation is that perturbing actinmyosin interactions (ZipDN) and perturbing actomyosin contractility by reducing phosphorylation on the myosin RLC (SqhAA) produce qualitatively distinct effects. An alternate possibility is that as both transgenes are expressed in otherwise wildtype backgrounds, their effects are different on account of the varying stoichiometries of wildtype and mutant myosins in each perturbation. At present, we cannot distinguish between these two possibilities. A strict assessment of the reason behind the difference will require an analysis of their relative stoichiometries and the ability to target the mutants solely to the fusing interfaces at different times during the process. In the new data we provide using the leading edge Gal4 (Figure 5—figure supplement 1), we show that a phosphomimetic myosin RLC mutant (SqhEE, that enhances contractility) also effectively abolishes interlocking and renders the midline seam ramrod straight. In contrast, the other two perturbations resulted in a wavy dorsal midline and produced patches of defective interlocking. This local perturbation (in contrast to the expression in stripes) lends support to the idea that the downregulation of contractility in the leading edge cell fusing interfaces is necessary for their relaxation and remodeling. It also brings out the importance of the spatiotemporal regulation of actomyosin based contractility (see also response to reviewer 3).

They go on to express either GFP tagged full-length Baz, a GFP tagged C-terminally truncated Baz (Baz[Δ969-1464]) or Baz-RNAi. Expression of either full-length Baz or Baz[Δ969-1464] resulted in similar phenotypes, including weaker ECadherin staining, whereas expression of Baz-RNAi had only mild effects, but showing increased E-Cad staining. Baz-GFP was recruited with a similar dynamics as endogenously-tagged Baz. In contrast, Baz[Δ969-1464] recruitment was delayed, which was associated with an extended time required to reach the interlocked stage. Finally, they induced a wound in the epidermis on one side of the embryo and observed the fusion event in the wound. From all this they conclude that "… changes in interface length and geometry are dependent on regulation of cytoskeletal tension and Baz/Par3" (Abstract). While the data are convincing with respect to the importance of cytoskeletal tension in this process, the role of Baz is unclear. The fact that overexpression of Baz versions, in particular the truncated one, impairs this process, does not imply that Baz plays a role in this process, given that expression of RNAi has a different outcome. Rather, expressing this form results in a dominant phenotype, but does not provide convincing evidence on the role of Baz in this process.

In the earlier version of the manuscript, we had analysed the effects of Bazooka using three perturbations targeted to the stripe: full length Bazooka overexpression, overexpression of a truncated version of Bazooka and Bazooka RNAi. The phenotypes of the former two were assessed both with live imaging (GFP tagged constructs) and in fixed preparations. The latter was only analysed in fixed preps (with UAS GFP/engrailed antibody to mark the engrailed domain). In the present version, we have examined Bazooka RNAi embryos using fixed preparations in greater detail (Figure 5, Figure 5—figure supplement 1) and also in real time using sqh::Utrophin GFP (Figure 6, Figure 7). We chose Utrophin GFP as it is neutral (binds actin) and is unlikely to modify the phenotype of Bazooka RNAi, a fear we had with using ECadherin GFP or sqh GFP. We also analysed epithelial order and junction remodeling in fixed preparations as well as in real time. Our new analysis reveals that the expression of either Baz RNAi or BazΔ in stripes produces a significant increase in the percentage of fused interfaces that were curved/squiggly compared to controls, and a significant decrease in both angled and straight interfaces. In contrast Baz overexpression only modestly reduced the proportion of straight interfaces (Figure 5). In Baz RNAi and BazΔ expressing stripes, cells in the stripe appeared more “fluid” and mobile and transgressed the dorsal midline leading to an obliterated seam, suggesting that Bazooka restrains/ regulates cell dynamics and ensures epithelial order at the midline (Figure 5). In contrast Bazooka overexpression led to the ordered stacking of cell rows and increased order at the dorsal midline and reduced cell shape anisotropy (squarish rather than rectangular cells; Figure 5,Figure 5—figure supplement 1G and Video 8A,B).

We also reexamined the effect of the three perturbations on the levels of E cadherin. Both Baz overexpression transgenes reduced the intensity of Ecadherin throughout the stripe. The most striking effect in Baz RNAi and BazΔ was a reduction in the anisotropy in the distribution of Ecadherin in DME cells. This was not observed upon Baz overexpression. In 15% of Baz RNAi expressing stripes, an enrichment of E Cadherin specifically at the fusing interfaces was observed even after fusion. The mobility of the interfaces we are examining makes it difficult to determine whether Bazooka influences the relative distributions of mobile and immobile ECadherin at these interfaces (Figure 5).

We have now examined the effect of Baz RNAi on actin remodeling during fusion in real time and find that the clearing of actin from the fusing interfaces after fusion is impaired in the absence of Baz (Figure 6, Figure 7, Video 10, Video 11). In Baz RNAi, actin persists for longer at these interfaces. Thus, our view of Bazooka function based on the effects of RNAi is that Bazooka restrains cell movement within a sheet and negatively regulates the actomyosin contractility at the fusing interfaces. The temporally regulated appearance of Bazooka at the fusing interfaces thus ensures junction relaxation and stabilization of adhesion in the midline. Our results do not rule out a role for Bazooka dependent ECadherin regulation. Precisely how Bazooka influences junction and cytoskeletal remodeling is an ongoing project in the lab.

Reviewer #2:

This manuscript by Das Gupta and Narasimha employs an elegant combination of genetics, live imaging, and morphometric analyses to study re-epithelialization in Drosophila dorsal closure. The authors first perform a detailed descriptive analysis of segment matching in terms of cell number and cell shape and uncover a role for myosin regulation (or, in their terms, tension) in this process. The authors then explore the resolution of junctional geometry following fusion. The results suggest that reciprocal regulation of Bazooka/Par-3 and vinculin localization downstream of cytoskeletal tension are required to effectively restore epithelial integrity.

While the study includes an impressive volume of detailed, quantitative analyses, the manuscript suffers from poor organization and lack of structure, which makes it challenging for the reader to tease apart the real significance of the authors' contributions. The finding that cell numbers between opposing sides of the embryo are not initially precisely regulated is interesting, as are the authors' data on reorganization of junctions following fusion. However, these two components of the story have not been well melded together, and it is difficult for the reader to see the forest for the trees. Despite these qualms, the videos are stunning, and the experiments are, for the most part, well controlled and carefully presented in the figures and captions-too much so at times, in terms of the amount of detail presented.

We thank you for your valuable comments. In the revised version of the manuscript, we have both rewritten and reorganized the sections (and therefore figures) to ensure better clarity, readability and better melding of the two parts.

Major concerns:

1) In the Introduction and Discussion section, the authors could do a better job explaining how the aspects of dorsal closure they are studying might have implications beyond this particular system. Concepts key to the authors' work, such as fidelity and precision, as they apply to this particular system, should be more clearly defined in the Introduction.

We have expanded the Introduction to include our definitions of precision/fidelity, highlighted the need to understand the underlying mechanisms both from the perspective of how molecular and morphological symmetry is ensured during large-scale cell movements and also from the perspective of reepithelialisation. We have also posed alternative hypotheses of how this might be achieved.

2) The authors perturb myosin as a way to make inferences about the role of tissue tension in this process. However, their strategy is somewhat problematic, as it relies on constitutive expression of a DN-zipper construct in engrailed stripes. While it is likely that tension does play a role in this process, it is difficult to interpret whether the observed changes in precision parameters occur because of a direct role for actomyosin-based tension in cellular rearrangements (which I believe that the authors want to imply) or that other aspects such as cell proliferation might be impaired. To address this problem, the authors should treat embryos with ROCK inhibitor Y-27632 during dorsal closure, which will allow more temporal control over manipulation of contractility.

We agree that our perturbations targeted to the engrailed compartment do not allow us to confine them temporally and spatially to the period of dorsal closure during which time neither the DME nor the amnioserosa cells divide. We have also confirmed this in our experiments using UAS histone RFP that allows us to see mitotic figures and watch divisions as they happen. The use of drugs is also problematic: they do not allow the spatial control of contractility as they cannot be injected specifically into the relevant cells. The amnioserosa will also be affected and this can influence the phenotypes we see. Given the spatiotemporal dynamics of actomyosin we see, and the fact that different stripes in the same embryo are in different stages of fusion at any given stage, the ideal way of addressing this is question is to use optogenetics to trigger and target the perturbations to single interfaces. This is currently not possible. Another compounding feature is that changes in cytoskeletal tension regulate more than just cell rearrangements: they influence junction remodeling at fusing interfaces and affect dorsal closure dynamics. We therefore reasoned that a good compromise would be to drive the three perturbations that alter tension, ZipDN, SqhAA and SqhEE exclusively in the leading edge cells using the leading edge Gal4 whose expression levels increase and patchiness decreases as dorsal closure progresses. While the patchiness does not allow us to assess fidelity at the level of each compartment, we find effects on straightness of the dorsal midline that suggest reduced fidelity (Figure 5—figure supplement 1). In both ZipDN and SqhAA, the dorsal midline is wavy along the AP axis where as in SqhEE, the midline is ramrod straight. In all three perturbations, and in particular, in SqhEE, poor interlocking is observed in post-closure embryos. These experiments have allowed us to confirm that altering contractility in just the leading edge cells during dorsal closure (at stages when proliferation is complete) influences junction remodeling, fusion fidelity and epithelial order at the dorsal midline.

3) As written, the manuscript is largely descriptive and difficult to digest. The novel points of the paper, although beautiful, are subtle and might be lost on readers unfamiliar with the extensive literature on dorsal closure. In its current form, this study may be better suited for a more specialist audience than the readers of eLife. This outcome would be a shame because the problem of re-epithelialization is of broad interest, and the authors' observations regarding interface remodelling after fusion are interesting, but they are regrettably buried in a sea of difficult-to-follow quantifications of unclear broader biological significance.

We have incorporated these suggestions in rewriting this version of the manuscript. We have now moved quantifications that are not essential to the supplementary materials along the lines suggested in 4. We hope you will agree that the most important results now stand out.

4) To begin to unclutter the story, Figure 1 and Figure 2, a tour de force of measurements and quantifications, could be simplified, with many of the detailed comparisons between stripes and interstripes removed or moved to supplemental. Similarly, Figure 7 and Figure 2—figure supplement 1 could be presented with fewer time points, just those illustrating the authors points. It is hard for the reader to digest so many panels.

Thank you for this suggestion. We have incorporated these changes in the in the revised figures.

Reviewer #3:

This manuscript focuses on the important problem of fusion fidelity and seamless epithelial continuity in the dorsal closure model for epithelial cell sheet migration and fusion. The authors find that differences in the length of the leading edge (here referred to as "arc length") of the fusing cell sheets is resolved long before fusion and that small differences in arc cell number are tolerated at fusion. The authors conclude that precision evolves over time and because arc length matches before cell number, suggest that two processes lead to these emergent properties. Such differences in arc cell number are accommodated by cell shape changes that are modulated by cell tension. When tension is perturbed by the expression of dominant negative myosin, matching is also perturbed. The authors also have documented changes in the structure of the formed seam with time, a process they call remodeling, and show that an initially flat interface between the zipped cell sheets becomes interlocked. Interestingly, the process of interlocking is accompanied by a loss of cadherin, actin and myosin at the interface between cells that came from contralateral sides of the embryo. The decrease in myosin occurs concomitantly with an increase in Bazooka/Par3, whereas patterns of vinculin accumulation is more similar to myosin. The authors conclude that such patterns of myosin and Bazooka accumulation are consistent with reduced tension in the formed seam. As described below, I do not understand how the observations that accompany attempts to reduce tension in stripes influence junctional remodeling, at least in part because I find Figure 6 very hard to understand and I find comparing the heat maps in Figure 7 hard to compare to the presentation of data in the rest of the Figures.

Overall, this is an interesting topic and a revised manuscript would make a compelling publication in eLife. My sense is that modifications to the presentation, and few if any new experiments, will make it ready for publication.

We thank the reviewer for the positive comments.

Specific comments:

The only way the authors could do these experiments were through live imaging and the data set they present is rich and valuable. Nevertheless, such live imaging requires the expression of fluorescently tagged cadherin, to label cell junctions and GFP-actin to delineate stripes vs. interstripes. Can the authors be sure that the expression of these transgenes doesn't affect the processes they are studying? Perhaps analysis of fixed and stained, control and experimental embryos could confirm that at least the overall pattern of mismatching, segment and intersegment widths and cell numbers is preserved? The analysis I envisage is to replicate the measurements on several fixed embryos stained with antibodies to cadherin and engrailed, to confirm that the patterns are not distorted in control vs. experimental animals.

We had already done this. In one set of experiments in which we were examining the effect of Bazooka RNAi (presented in the earlier version of the manuscript), we used the engrailed antibody to mark the engrailed domain in controls and mutants, and Ecadherin antibody to label cell outlines, with no additional transgenes. We did this also to see whether the addition of UAS GFP/RFP to label the compartment may have reduced phenotypic severity. A second set of controls used UAS GFP (neutral) to mark stripes in controls and mutants (controls for UAS Baz RNAi, UAS SqhAA, UAS SqhEE) in which the stripes were labelled with UAS GFP and cell boundaries with Ecadherin antibody. Although a dynamic analysis of the parameters could not done in these two data sets, DME cell numbers and their differences, compartment widths and interface geometry were comparable in both sets. For the real time analysis, we used two control genotypes: i) Ecadherin GFP with en>UAS actin GFP (Zip DN experiments) and ii) sqh::Utrophin GFP (also neutral in that it expresses a GFP that can bind actin) with en>UAS NLS RFP (UAS Baz RNAi experiments). In each case, the mutant and controls both had ECadherin GFP or sqh::UtrGFP. Furthermore, the two control genotypes yielded similar results for DME cell numbers and cell number difference evolution, stripe widths as well as for the timescales for interface remodelling suggesting that the parameters we are measuring are not influenced by the transgenes used for live imaging. Specifically, in this manuscript, we show (in Figures 6F,G) that the values of the differences in DME cell number and constriction index (stripe width/stripe cell number) are in the same range in two different controls used: ECadhGFP with en>UAS actin GFP and sqh::Utrophin GFP with en>UAS NLS RFP.

Critical reading of the Ducuing and Vincent, (2016) and Pasakarnis et al., (2016) papers suggest that the authors of both studies have over interpreted their data to rule out a contribution of actin-myosin rich cables to wild type closure. Numerous other studies show that the cables are not strictly required for closure, but nevertheless normally contribute to the process. Similarly, the cables present but not necessary for wound healing. This does not mean that they do not normally facilitate the process. The authors should amend their Introduction appropriately. Indeed, the authors own observations that show expressing dominant negative myosin constructs in epithelial stripes slows closure is consistent with (but does not prove) a role for the purse strings in closure.

We fully agree with this view and have amended our statement in the Introduction and also included a statement in the Discussion section. Indeed, our new results (Figure 5—figure supplement 1) with leading edge Gal4 that is expressed only in the DME cells also confirms the view that tension generated in the leading edge cell entrains the spatiotemporal dynamics of dorsal closure.

It is not clear to me why that authors normalize the difference in upper and lower arc lengths using Norm. ΔArc Length = (Lu – Ll)/(Lu + Ll). Shouldn't the ΔArc Length be normalize with the length of one arc, not two? That is, it would seem that (Lu – Ll)/((Lu + Ll)/2) is a more appropriate measure of the normalized arc length difference. A similar issue arises from the calculation of Norm. ΔWidth and with DME cell number. I think equations 4 and 5 are more appropriate normalization schemes. It seems that the authors' normalization scheme in equations 1, 2 and 3 underestimates all differences by two fold.

We have now incorporated these suggestions in the new Graphs presented, and in the Materials and methods section.

Shouldn't the authors be evaluating standard deviation rather than standard error of the mean? I'm not an expert on statistics, but see Barde and Barde, 2012, Perspectives in Clinical Research 3: 113.

In the majority of the data presented (except frequency/percentage distributions of data values), the graphs show all the data points, the horizontal bars indicate the median and interquartile range, and the Mann-Whitney test was performed to detect significant differences in the shape and spread of the distribution as well as differences in the medians between the genotypes/conditions being compared. The current Figure 1H is also now represented in this way rather than as a histogram with mean and SEM. We have used standard deviation where we were keen to determine the data spread/scatter/variability. We have replaced the error bars in the current Figure 4C to be standard deviations rather than SEM. In some graphs in which we are expressing differences in relative frequency (eg: Figure 5H and Figure 1—figure supplement 3F), we have used the SEM (which relies on the SD and the sample size) as we were interested in determining the accuracy of the mean.

Subsection “Population analysis of precision descriptors” states: "(control in Figures S3E and S3F)". I don't understand how the data in the two cited panels, which pertain to embryos expressing a dominant negative form of non muscle myosin constitute a "control". The authors should explain.

We should have said control panels in Figure 1—figure supplement 3E and Figure 1—figure supplement 3F. The control we were referring to is not Zip DN but to the controls presented on the left hand side of each graph (ECadherin GFP with en>UAS actin GFP), for comparison.

Subsection “Population analysis of precision descriptors”: there is an "are" before "remain" that should be deleted.

This has been corrected.

Subsection “Cell addition to the leading edge contributes to cell number and width equalization”. The authors conclude "This suggests that cell addition may be deployed to equalize number or width of the posterior compartment." I don't see how, taken at face value, that sentence is consistent with the observations reported in the rest of the paragraph. At the very least, it is misleading since cell addition occurs even when no disparity exists (previous to addition). The wording here should be amended to better reflect the observations reported in the remainder of the paragraph and accurately reflect a consistent interpretation of the data.

We agree. What we see is that in a small percentage of stripes, the cell addition reduces cell number disparity but in the majority, there is a correlation between width equalization and cell addition. We have corrected this to reflect what our observations indicate.

Subsection “Cytoskeletal tension contributes to the evolution of precision and prevents segment misalignment”. On a sentence that starts with "Furthermore…" and focuses on the role of tension in cell matching, the authors describe the effects of the expression of dominant negative myosin in stripes and note that in one quarter of embryos there were fusions with "non-partner contralateral stripes". Such non-partner fusions were described by Franke et al., 2005 in embryos that had compromised myosin function (see their Figure 7). The authors should cite that work.

We thank you for pointing this out to us. We have now cited this work.

I am having trouble understanding how the experiments described in subsection "Bazooka and tension-dependent cell interlocking contributes to establishment of a continuous epithelial sheet" fits with their model of how tension influences remodeling of the formed seam. The authors show that cadherin, myosin, actin and vinculin are all decreased at the interface of the epidermal cell sheets from opposite flanks once the flanks are zipped. Moreover, they show that Bazooka is increased concomitantly with the decrease in myosin. They argue that tension is also decreased, and this allows an initially straight seam to become zig-zagged or interlocked. But here, they attempt to reduce tension by expressing either dominant negative myosin heavy chain or non phosphorylatable light chain and this disturbs remodeling. Shouldn't it do the opposite (i.e., like the over expression of Bazooka, described later in subsection "Bazooka and tension-dependent cell interlocking contributes to establishment of a continuous epithelial sheet")?

We understand your concern. In the experiments presented we perturb tension in the entire DME cell and in the entire engrailed compartment. We however show that tension is anisotropic. Thus, although tension is normally reduced in the fusing interfaces, in our perturbations, it is also abolished in the non-fusing interfaces, as for instance the DV oriented interfaces of the leading edge cells along which it is positioned and potentially also in the apicomedial pool. In both germ band extension and pulsed constriction in the amnioserosa, it has been shown that junction relaxation relies on directed actomyosin contractility that pulls the vertices outward. In our context, the neighbours could be DME cells in the same flank or in the opposite flank, where cytoskeletal tension may need to be high. Alternatively, tension in the compartment boundaries may also have an effect. Our results with LE Gal4 (which is patchy) suggest that perturbing cytoskeletal tension only in the leading edge cells is sufficient to influence both interface remodeling and fusion fidelity. Indeed, in the work we present in Figure 5—figure supplement 1, we show that all affect interlocking in patches, with SqhEE having the strongest effect. We believe that we have now established a nice model system in which can address autonomy, non-autonomy and the impact of tension anisotropies within the DME cells in real time.

I find Figure 6 very confusing. First, D – F are labeled "closure" and G-I are labeled post closure, but each panel has a label "zipping" with an arrow. Isn't all zipping done post closure? Similarly, panels A, B and C have cartoons that are sketches of the interfaces seen in the micrographs and they too are labeled with zipping and arrows. But those arrows, in the control panel A are beneath mature interlocked cells, whereas the middle has T or + boundaries that show cells early in fusion. The authors clearly understand what they are trying to show. They need to figure out how to modify the Figure so it is more easily interpretable by the reader. I find the magnified inserts and schematics very helpful in A, B and C (but NOT the zipping arrows, which I don't fully understand). Why not have all of the panels comparable to A, B and C? I don't understand the point of showing zipping in two directions in A, B and C. The authors describe no differences between the anterior and posterior seams formed, so why bother showing both? The layout of panels, D – F is compelling because the dorsal opening is shown so it helps the reader orient the panel properly. Why not do that for A, B and C? Finally, are each pair D and G, E and H and F and I from the same embryo? If not, why not?

We understand the confusion. This was a semantic error: zipping was wrongly used to indicate the direction of fusion (from the canthi inwards). We simply wanted to indicate that the segments at the anterior and posterior ends of the ellipse fuse first while the central segments fuse last and that remodeled interfaces are found outwards from the canthi while remodeling interfaces are found closer to the canthi. We have now removed “zipping”and “remodeling” from the figures and instead indicated the position of the midpoint along the AP axis to orient the reader to region of the dorsal epidermis from which the image is obtained. The data in each pair are not from the same embryo because this data was obtained from fixed preparations of embryos that were either still completing closure (some amnioserosa still visible) or immediately after closure is complete (no amnioserosa visible on the dorsal surface).

The authors wait until the Discussion section to tell us what the significance of a truncated form of Baz is. They add a sentence in the results to inform the reader.

We have done this now.

I don't understand why the authors have switched to a heat map presentation in Figure 7. Of course, heat maps are useful for highlighting relative intensities, but the bulk of the rest of the paper depends on inverse fluorescent gray scale to depict cellular morphologies. So that the reader can better understand the effects of the various permutations of baz expression, the authors should present the data as they have for the other Figures. If heat maps are important, put them in a supplemental figure.

We have switched to the grey scale images and have used some heat maps where spatiotemporal regulation of intensity is what we want to show, in the supplementary figures.

[Editors' note: the author responses to the re-review follow.]

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

This is a potentially interesting paper that addresses cell matching during the process of dorsal closure. It is from an accomplished lab that has contributed key findings to the field. The authors show that initially, the lateral epidermal cells that surround the dorsal opening are different in number and the perimeter of the dorsal opening measured on the right side of the embryo is different than the left side. The embryo adjusts length earlier than cell number and fusion between advancing cells sheets can occur even if the number of cells is not perfectly matched. The authors go on to show that an initially T or + seam that forms becomes interlocked and jagged. They go on to explore what might be responsible for regulating the process and investigate both tension (by perturbing myosin function and with laser cuts) and by manipulating Baz/Par3. They find effects with both, some of which are new and some of which are not (for example, the finding that DNZip causes segment misalignment). Overall, this is a very careful analysis, and the data add novel and important aspects to our understanding of epithelial fusion, showing that precision is not pre-set, but emerges.

This revised version of the manuscript is a major improvement to the manuscript submitted previously. The manuscript is easier to read, since they have put quite a few data into the Supplementary material. The figures are also significantly de-cluttered and it is now easier to digest the main points. We also appreciate the authors' efforts to address our point regarding better spatiotemporal control of their myosin experiments – we agree there is no easy way to do this without the development of new tools (ie. optogenetics as they point out) – and we concede that the alternative Gal4 driver (LE-Gal4) is a good compromise.

We thank the reviewers for their positive comments on our revised manuscript. Our manuscript demonstrates for the first time that precision emerges during the process of fusion and identifies the cellular and molecular mechanisms that contribute to it. Although Zip DN has previously been shown junction remodeling events we describe here for the first time. We also demonstrate for the first time, the requirement of Bazooka for ensuring fusion fidelity and to enable junction remodeling. While the effects of some of the perturbations we have analysed may seem small yet significant, we believe from our examination of zygotic Bazooka mutant embryos that these are the result of incomplete loss of function. Indeed, Bazooka mutant embryos often fail in dorsal closure (new data added in this revision), precluding our ability to perform the kind of detailed analysis done on the other perturbations. The junctional morphologies evident in these mutants convince us that the effects are real and are stark. We also concede that there is variability in interface morphologies at fusion. Our quantitative analysis of remodeling in control embryos and in mutants captures this variability but also uncovers significant shifts in the frequency distribution of junctions of different morphologies. In this revised version of the manuscript, we present data that captures the differences in frequency distribution of interface orientation angles as well as the central tendencies of the three described categories of interfaces for all the Bazooka perturbations (Figures 5H-K) and for myosin perturbation SqhAA (now in Figure 5—figure supplement 1).

However, although the reviewers agree that the manuscript is easier to read, there are some concerns summarised below:

- If the authors could justify the assertions they summarize in their manuscript's Abstract, the paper would be a very nice contribution. Unfortunately, it is not clear that the assertions made are justified based on the data presented. Furthermore, the writing can still be improved. The reader is still left with a lot of detailed data. In order not to get lost, it would be helpful to add a summary sentence after each paragraph, if possible.

We do believe that we have justified all the assertions in the Abstract: (i) that precision is not preset but evolves; (ii) that cell matching happens at less than perfect fidelity; (iii) that fusion happens despite differences in cell number; (iv) that changes in interface length and geometry ensures segment alignment and seamless continuity and (v) that tension generated by the actomyosin cytoskeleton and the polarity protein Bazooka modulate cellular rearrangements and changes in interface morphology that mediate precision and seamless continuity. We address the other concerns below.

- The new Figure 5 and Figure 6 are still very difficult to follow. Is it possible to reduce the number of panels or perturbations in the main figures and move the remainder to supplementary figures? Some of the difficulty arises from the lengthy labels, e.g., "enGal4>UAS GFP," "enGal4>UAS ZipDN," "enGal4>UAS BazGFP ∆969-1464." Figure 5H is still rather confusing, in particular due to the inclusion of the significance within the columns. The authors may consider a modification of this graph, perhaps by showing three individual columns for each genotype, with the significance on top. Can the authors come up with more compact and understandable ways to label the different scenarios?

We have split the erstwhile Figure 5H to include additional panels H-J and have now moved panels Figure 6 H,I to Figure 6—figure supplement 1. We have made the labels less long.

- A very interesting finding that is not well, if at all, highlighted is when equalization occurs. If the reviewers understand Figure 1—figure supplement 1Q correctly, don't the data suggest "action at a distance"? If geometry equalization occurs so early and number equalization occurs later, but still there is tens of microns between advancing cell sheets, how does one row of DME cells "know" to adjust its length and/or? Please clarify.

Yes. Our results in Figure 1—figure supplement 1Q reveal that mechanisms that contribute to equalization happen well before fusion and suggest action at a distance. We are currently investigating whether sensing operates across the amnioserosa or along the circumference of each segment compartment and whether this is mechanically or chemically regulated. Our results also show that although the mechanisms kick in well before fusion, final adjustments in compartment width after fusion ensure segment alignment in compartments in which there is still a stagger or width discrepancy. This suggests that while there are hardwired mechanisms that are deployed to ensure precision, their deployment is based on an as yet unidentified ‘sensing’ mechanism. The mechanisms that contribute to the evolution of precision are in some ways reminiscent of “kinetic proofreading” during translation. We have now discussed this in the revised text.

[Editors' note: the author responses to the re-review follow.]

Reviewer #2:

I greatly appreciate the changes the authors have made to restructure the presentation of data, in particular by moving some of the data to supplementary material. However, before acceptance to eLife, this manuscript's writing still needs significant work. Many sentences contain vague modifiers. As one of the reviewers pointed out, each paragraph would benefit from a clear topic sentence that segues from the previous paragraph and introduces the main point of the new paragraph. For clarity, many sentences with multiple clauses could be made into multiple separate sentences.

General comment: the errors in measurements should be reported to the same decimal place as the value reported, not 11.81 {plus minus} 1.2 but rather 11.8 {plus minus} 1.2 (subsection “Bazooka modulates interface remodeling and DME cell dynamics during fusion” and other places).

We have now corrected this throughout the manuscript and ensured that we use the same decimal place (single) for both the value and its error.

Regarding the Discussion section, some of the comments are unfounded, such as "thermodynamically stable junctional configuration" (subsection “Fidelity in epithelial fusion: the logic and the mechanisms”), especially without a citation.

We have now removed any mention of thermodynamic stability from the manuscript.

In the Discussion section, it is not clear to me what mechanisms are "emergent" and what the authors mean by this word. I suggest scaling back the claim of emergent mechanisms.

We have used the word “adaptive” rather than emergent to indicate mechanisms that are put into action as a response to the sensing of disparities and arise as a consequence of cell dynamics. It was also in this context that we had used the word “emergent”.

Sometimes the authors write "E Cadherin" or "E-cadherin" or "Ecadherin." These inconsistencies are a result of sloppy editing.

We have now corrected this throughout the manuscript and have ensured that we use the same notation throughout.

Throughout the manuscript, the authors write "this" without identifying a noun that follows that word. Such sentence structure is sloppy, as it is rarely clear which previous ideas the authors intend the word "this" to mean. A simple search for "this" and adding a noun (e.g., "observation," "result," "concept," etc.) would strengthen the clarity of these sentences.

We have now corrected this throughout the manuscript and ensured that a noun follows “This”.

I have outlined some example writing suggestions below, but this careful scrutiny needs to be applied to the entire manuscript, likely with the help of a professional editor.

We greatly appreciate your thorough reading of our manuscript and thank you for your comments and suggestions.

Abstract:

- I believe the Abstract is still poorly written. For a broad journal like eLife, it should be accessible to a general life-sciences audience.

We have edited the manuscript in the hope that it will be easily understood by the broad readership of eLife.

- Unclear writing: "We investigate this…": This what? Many ideas were presented previous to that sentence.

We have removed this sentence from the Abstract now.

- "We find that although differences in cell number and length between the two fusing arcs reduce as fusion progresses, small differences in cell number between fusing epidermal segments persist at fusion" – this sentence is extremely confusing, and a general life-sciences audience will not know what the arcs are.

We have introduced flanks and segments in the Abstract now and do not refer to arcs. We define arcs in the Results section.

- The word "Rather" negates the previous sentence, which was also negated by the word "Although." This writing is unclear. Is it a positive result or a negative result that "small differences in cell number persist"?

We wanted to convey that small differences in cell number persist at fusion and that fidelity with respect to cell number matching is less than perfect (ie, the disparity in cell number between the fusing pairs is not always zero). We have now edited the Abstract so that neither word (although or rather) is used.

- "Segment compartment boundaries" are not defined. "Front length" of what?

We have introduced segmentation in the abstract now. We meant length or width of the fusing fronts of the flank or segment.

- "Emergent mechanisms" – it's not at all clear in the Abstract what the emergent mechanisms are.

We have removed “emergent” from the Abstract.

Introduction:

- Even the first sentence is confusing. The comma after "healing" should not be there; the comma's presence forced me to read the sentence a number of times before I could follow what the authors were trying to say.

We have edited this sentence.

- The verb should be "has" because the subject of the sentence is the singular "study".

This has been corrected to “Earlier studies on epithelial fusion events … have identified”.

- The comma in “cell partners during Drosophila dorsal closure and eyelid closure, suggested that” is also incorrect and confusing, and the whole sentence should be rewritten for clarity.

This sentence has been corrected.

- "It" is vague.

This has been corrected. “It” has been replaced by “Dorsal closure”.

- What do the authors mean by "emergent"?

We have used the word “adaptive” rather than emergent to indicate mechanisms that are put into action as a response to the sensing of disparities that are arise as a consequence of cell dynamics. It was also in this context that we had used the word “emergent”.

- This comma also is incorrect.

This sentence has been corrected.

- At what time scale and in what process are the authors referring to "de novo"?

We used de novo to indicate that remodelling occurs during the formation of new junctions at fusion. We wanted to distinguish it from the remodeling of existing junctions as seen in germband extension. We now simply use “formation of new contacts” or “formation of the new seam” where de novo previously appeared in the text.

Results section:

- Is the list of measurements in subsection “Defining and measuring symmetry, fidelity and plasticity in fusion” missing the word "and" somewhere?

It is indeed missing an “and”. We have corrected this.

- Subsection “Defining and measuring symmetry, fidelity and plasticity in fusion”: what are "measures of plasticity in cell number"?

We used “plasticity” to simply mean the ability to modulate cell number or geometry. We have corrected this sentence.

- The sentence in subsection “The emergence of symmetry and fidelity during dorsal closure” is confusing because of the use of "later" and "preceding" in reverse order. Can the authors rewrite this sentence more clearly?

We have changed this sentence now.

- Subsection “The emergence of symmetry and fidelity during dorsal closure”: To what "mechanisms" are the authors referring?

We suggest that distinct cellular mechanisms must enable differences in cell number and differences in the geometry within each fusing compartment.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: better sentence structure would have the modifier "For this analysis" at the beginning of the sentence instead of the end.

We have changed this sentence now.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: what is a frequency of stripes? Do the authors mean number of stripes?

As mentioned in Materials and methods section, stripes refer to the posterior compartment of the embryonic segment. We agree that frequency of stripes is confusing. We meant “proportion of stripes” of a given kind. We have now changed “frequency” to “proportion.”

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: I don't understand the anthropomorphizing word "tolerated" here. What is tolerating one fewer cell?

We have removed “tolerated”. We used it to imply that despite small differences in cell number (of one or two cells) in the fusing segment compartments, the compartment boundaries aligned. The sentence has been reworded.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: comma needed after "pairs"; for sentences with two independent clauses separated by a conjunction, there needs to be a comma before the conjunction.

We have added this now.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: because of how confusingly written this paragraph is, the two different mechanisms are obscured.

We have now split this section into paragraphs and added two new paragraphs with separate section headings to emphasise the two mechanisms we are referring to. Subsection “Bazooka modulates interface remodelling during fusion” and subsection “Bazooka entrains actomyosin contractility during epithelial fusion”.

- Subsection “Differential contributions of segment compartments to equalization in fusing front DME cell number and geometry”: "them" is an unclear pronoun.

We have edited this sentence to read “For this, we determined the variance in DME cell numbers between ipsilateral and contralateral stripe pairs”.

- Subsection “Spatiotemporally regulated remodeling of cell junctions converts staggered en-face cellular contacts at the fusing front to an interlocking pattern after fusion”: what is "their" referring to?

“their” referred to A/P interfaces. We have changed this sentence now.

- Subsection “Spatiotemporal regulation of tension and Bazooka at fusing interfaces”: missing a period.

We have now added this.

- Subsection “Spatiotemporal regulation of tension and Bazooka at fusing interfaces”: since supplementary figures are possible in eLife, why are these data not shown?

We have now included this data in Figure 2—figure supplement 1E-F.

- Subsection “Spatiotemporal regulation of tension and Bazooka at fusing interfaces”: why is this finding surprising?

We have removed the word “surprising”. The finding is unexpected and has not been previously described. As we now speculate in our Discussion section, the segment compartment specific differences in Vinculin and Bazooka levels may underlie the compartment specific contributions to fusion fidelity.

- Subsection “Cytoskeletal tension entrains the evolution of precision and modulates cell behaviour and interface geometry”: which parameters?

We have now detailed the parameters we have examined in these embryos.

- Subsection “Cytoskeletal tension entrains the evolution of precision and modulates cell behaviour and interface geometry”: should be "were" instead of "was".

We have corrected this.

- Subsection “Cytoskeletal tension entrains the evolution of precision and modulates cell behaviour and interface geometry”: what do the authors mean by "evolution"?

We have used evolution to indicate the temporal progression to high fusion fidelity. We wished to convey that fidelity is not pre-set in fusing segment compartments at the outset. We have now prefixed “evolution” with “temporal” to indicate this.

- Results section, Discussion section: some letters are missing.

Perhaps the letter Δ is missing because the fonts are different. I cannot find any letters missing.

- Subsection “Bazooka modulates interface remodeling and DME cell dynamics during fusion”: whereas, not "where as".

We have corrected this.

https://doi.org/10.7554/eLife.41091.031

Article and author information

Author details

  1. Piyal Taru Das Gupta

    Department of Biological Sciences, Tata Institute of Fundamental Research, Mumbai, India
    Contribution
    Resources, Formal analysis, Investigation, Visualization, Methodology, Writing—review and editing, Performed all experiments, Analysed data, Prepared figures, Contributed to manuscript editing and reviewing
    Competing interests
    No competing interests declared
  2. Maithreyi Narasimha

    Department of Biological Sciences, Tata Institute of Fundamental Research, Mumbai, India
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Conceived, designed and supervised the project, Acquired funding, Contributed to the image acquisition and analysis, Wrote the manuscript
    For correspondence
    1. maithreyi@tifr.res.in
    2. mnarasimha.tifr@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8398-8023

Funding

Tata Institute of Fundamental Research (Studentship)

  • Piyal Taru Das Gupta

Department of Atomic Energy, Government of India (Intramural Funds)

  • Maithreyi Narasimha

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Andrea Brand, Thomas Lecuit, Richa Rickhy, Daniel St Johnston, Tadashi Uemura, the Bloomington Drosophila Stock Center and the Developmental Studies Hybridoma Bank for reagents, FIJI/ImageJ (NIH) for image analysis software, Himanshu Sinha for suggestions for statistical analysis, Modhura Ganguly and Somnath Ghosal for providing some immunostained embryos, members of the MN lab and the Department of Biological Sciences for discussions, and Sudeepa Nandi and Anwesha Guru for comments on the manuscript. MN was supported by intramural funds and PTDG by a studentship from the Tata Institute of Fundamental Research/Department of Atomic Energy (DAE) of the Government of India.

Senior Editor

  1. Anna Akhmanova, Utrecht University, Netherlands

Reviewing Editor

  1. Elisabeth Knust, Max Planck Institute of Cell Biology and Genetics, Germany

Publication history

  1. Received: August 14, 2018
  2. Accepted: April 3, 2019
  3. Version of Record published: April 17, 2019 (version 1)

Copyright

© 2019, Das Gupta and Narasimha

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 782
    Page views
  • 104
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    Mario Ruiz et al.
    Research Article Updated
    1. Cell Biology
    2. Immunology and Inflammation
    Sophie Isabel Roper et al.
    Research Article