Mechanical coordination is sufficient to promote tissue replacement during metamorphosis in Drosophila

Abstract

During development, cells coordinate to organize in coherent structures. Although it is now well established that physical forces are essential for implementing this coordination, the instructive roles of mechanical inputs are not clear. Here, we show that the replacement of the larval epithelia by the adult one in Drosophila demands the coordinated exchange of mechanical signals between two cell types, the histoblasts (adult precursors) organized in nests and the surrounding larval epidermal cells (LECs). An increasing stress gradient develops from the center of the nests toward the LECs as a result of the forces generated by histoblasts as they proliferate and by the LECs as they delaminate (push/pull coordination). This asymmetric radial coordination of expansive and contractile activities contributes to epithelial replacement. Our analyses support a model in which cell–cell mechanical communication is sufficient for the rearrangements that implement epithelial morphogenesis.

Substitution of larval abdominal epithelium with adult tissue during fruit fly development is regulated by interplay of physical forces induced by adult tissue proliferation and delamination of larval cells.

Introduction

While tissues are mostly static in adult individuals, during development they undergo complex rearrangements creating an endless array of shapes by means of a few basic cellular processes. Cells grow, die, and constantly and dynamically change positions generating diversity in a precise coordinated way.

Whereas in some morphogenetic processes cells move individually over large distances (e.g., germ cell migration; Kunwar et al, 2006), during epithelial morphogenesis, when cell–cell junctions restrict free movements, cells act coordinately in groups (e.g., during neural tube folding; Nikolopoulou et al, 2017) In these cases, the emergence of tissue organization cannot be just explained by the genetic makeup of individual cells, but from their interactions. Some morphogenetic events proceed without growth and are the result of coordinated geometrical changes and of cell rearrangements (e.g., Drosophila gastrulation; Sawyer et al, 2010). Others, however, are dependent on the local control of cell growth, either in size or in number. This is the case of the folding of the wing disk (Bryant & Levinson, 1985; O'Brochta & Bryant, 1985; Gong et al, 2004; Baena‐Lopez et al, 2005; Lecuit & Lenne, 2007; Sopko et al, 2009) or the expansion of the abdominal histoblasts (Ninov et al, 2007) in Drosophila.

In this report, we explored the instructive capabilities of mechanics on tissue organization during the replacement of the abdominal epithelium in flies. This process relies on maintaining a balance between the expansion of histoblast founders and the delamination of larval epidermal cells (LECs), so that epithelium integrity is ensured. Histoblasts are grouped in four nests by hemisegment that spread and fuse in a stereotyped way to cover the whole surface of the abdomen. Histoblast proliferation and growth and the apical constriction of LECs appear to be non‐autonomous processes that depend on a mutual exchange of signals (Ninov et al, 2007). Yet, how these two events are coordinated is not well understood.

Here, we reveal that the nests’ growth is dynamic and anisotropic and that it relies on contributions from cell divisions and size changes. These cellular actions are spatially coordinated in a characteristic radial pattern. This radial layout marks the distribution of mechanical parameters, such as relative pressure, power, or stress. A tensional stress gradient is created and sustained from the center of the nests toward the LECs. Remarkably, the mechanical constraints directing the generation of the stress gradient originate on the efforts generated by histoblasts, as they proliferate and by LECs, as they delaminate and die (push/pull mechanical coordination). In the absence of histoblast proliferation, LEC delamination does not occur and no stress gradient is created. Otherwise, blocking LEC delamination does not affect histoblast proliferation rate. Histoblasts remain as a power source, and the stress gradient is still established, although shallow and positionally‐shifted.

In summary, an asymmetric mechanical coordination of the different expansive and contractile activities of histoblasts and LECs in a radial configuration is essential for the progression of epithelial replacement in the abdomen of Drosophila, during metamorphosis. Our analyses support a model in which cell–cell mechanical interactions are important for the control of the rearrangements that underlay epithelial morphogenesis (Manning et al, 2010).

Results

Ventral nests’ kinematics

The growth of the histoblast nests during metamorphosis follows a precise spatiotemporal pattern. By 15 h after puparium formation (APF), the ventral and spiracle histoblast nests initiate their expansion in a coordinated manner. They fuse by 18–22 h and zip up with other segments by 28–32 h (Fig 1A and Movie EV1). The ventral nest area increases with irregular oscillations as the LEC tissue territory decreases (Fig 1B and C, and Appendix Fig S1). The full expansion of the ventral nest takes approximately 750 min and proceeds in three stages: an early phase 1 wherein the average rate of growth of the nest increases from 0.0018 to 0.0038 min⁻¹ (0–250 min); an intermediate phase 2 when the growth rate is sustained (250–570 min); and a late phase 3 in which growth slows down to 0.0025 min⁻¹ (570–750 min). Nest growth is anisotropic and oriented. At stage 1, the ventral nests have a typical elongation ratio (ER) of 2.8 and are tilted anteriorly (dorso‐ventral orientation of 1.7 radians), while at the end of expansion, their ER is 1.6, and their orientation has drifted to 1.57 radians (Fig 1D and E).

Figure 1. Ventral nests’ growth.

1. Ventral nest expansion at 0 min (onset); 245 min, when ventral and spiracle nests fuse; 515 min, when intersegment nests fuse; and 750 min, when most LECs are gone. We employed for these analyses the following stock: w; If/CyO; ZCL2207::hh‐DsRed/MKRS. The green signal corresponds to ZCL2207 (septate junction GFP marker) outlining all apical cell membranes and the red signal to nuclei from the posterior compartment expressing RFP under the control of the hedgehog promoter. Arrows point to the spiracular and ventral nests on the top panel and to the junction between these two nests in the second panel. Images of confocal sections have been rotated to maintain consistent alignment of the dorso‐ventral axis (dorsal is up). The outlines of the confocal images are marked with a yellow dotted line. Scale bar = 40 μm.
2. Comparison of manual versus PIV surface area kinetics [evolution of nests (red—manual, black—PIV) and LECs (cyan—manual, blue—PIV)] (averaged from six movies—see Appendix Fig S1).
3. Nest rate of growth from a representative single nest. Note the similarity in behavior (red—manual, black—PIV) in three sequential phases: phase 1 (Ph 1), accelerated growth; phase 2 (Ph 2), sustained growth; and phase 3 (Ph 3), decelerated growth.
4. Progression of the elongation ratio (red—manual, black—PIV) of nests (averaged from six movies). Note the initial fast decay on the elongated shape of the nests.
5. Nest orientation evolution (averaged from six movies) (red—manual, black—PIV). Anterior is π (3.1415 rads), dorso‐ventral is π/2 (1.57 rads), and posterior is 0.

Radial topographical asymmetries

To evaluate the dynamics and topology of expanding and contracting areas, we employed divergence fields (Fig EV1). From the onset of expansion, divergence was positive (expansion) within the nest. Yet, it was scattered and in most cases associated with cell divisions. Likewise, within the LECs, divergence was preferentially negative (contraction) and linked to individual extrusions (Fig 2A). Negative divergence was also found at the nest–LEC and intersegment borders. The difference in divergence between the histoblasts and the LECs increased over time (Fig 2B), and a constriction belt negative divergence developed surrounding the nest, 20 μm away from the nest–LEC border. Visual inspections and divergence's radial kymographs (Fig 2C and Movie EV2) indicated that as the nests grow, expansions and contractions were distributed in a characteristic radial pattern. Divergence was preferentially positive in the nest center, weak at the periphery, and negative in the LECs.

Figure EV1. Divergence and vorticity of velocity fields.

1. Diagram representing tissue/cell increase (growth—brown) and decrease (shrinkage—blue) in area.
2. Corresponding velocity fields and divergence fields for an expansive movement and a contractive one. In expansive movements, the PIV arrows point outwards (diverge) and divergence is positive (red). In contraction movements, the PIV arrows point inwards (converge) and divergence is negative (blue).
3. Rotating velocity field showing positive (counterclockwise) or negative vorticity (clockwise).

Figure 2. Divergence field.

1. Snapshots of the divergence of the velocity field from a representative movie. Note the areas of positive divergence in the nests, areas of positive and negative divergence in the LECs, and an increase in magnitude on the differences between them over time. Scale bar = 40 μm.
2. Frequency distribution of the divergence magnitude in the nests and LECs in the different growth phases. Nests are always shown in red and LECs in blue.
3. Color‐coded radial kymographs of divergence for a representative movie. Low divergences are in blue, and high divergences are in red (see scales). Note the overall positive divergence in the nests and negative in the LECs.

Other geometrical contributions besides divergence, such as vorticity, did not appear to constitute a relevant hallmark of the nest' growth dynamics (Appendix Fig S2).

Individual cell contribution to tissue kinematics

To evaluate the contribution of cell proliferation and size to the growth of the histoblast nests, the numbers and areas of individual histoblasts and LECs were averaged from three segmented time‐lapse movies (Fig 3A and B; see Movie EV3). Cell intercalations (T1 transitions) and the delamination of LECs were also monitored.

Figure 3. Histoblast and LEC number and area kinetics.

1. Cell numbers in nests (blue—left scale) and LEC territories (red—right scale) (profiles are averaged from three segmented movies).
2. Mean cell area evolution of histoblasts (blue—left scale) and LECs (red—right scale). Note the progressive increase in mean histoblast area (profiles are averaged from three segmented movies).
3. Net change rate in cell number variation for nests (red line) versus nests’ growth (black line). Note the temporal correlation in phases, but the reduced magnitude in cell numbers (profiles are collected from single nests and normalized; control black lines are retrieved from Fig 1).
4. Net change rate in cell number variation for LECs (blue line) versus tissue growth (black line). Note the full pattern correlation and the oscillations in LEC net rate numbers (profiles are collected from single nests and normalized; control black lines are retrieved from Fig 1).
5. Evolution of the net change rate of histoblast MCA (red line) versus net change rate of nest growth (black line). Note the delay in reaching the sustained growth phase and the reduced magnitude of the MCA rate (profiles are collected from single nests and normalized; control black lines are retrieved from Fig 1).
6. Evolution of the net change rate of LEC MCA (blue line) versus net change rate of LEC tissue growth (black line). Note the oscillatory MCA rate around 0 and the tendency toward negative values (MCA decrease) (profiles are collected from single nests and normalized; control black lines are retrieved from Fig 1).

Histoblast number increased with time but always at a smaller rate than the nests’ growth (Fig 3C). Histoblast size [mean cell areas (MCAs)] also increased, but it was uncoupled from the nests’ growth: e.g., nests' accelerated growth, in phase 2, preceded histoblast MCA speeding up by 200 min (Fig 3E). LEC number, conversely, decreased, while the average size of individual LECs remained constant (Fig 3D and F). Thus, contrary to histoblasts where both number and size are relevant for growth, a change in LEC number was their solely contribution to tissue elimination.

Histoblast number, size, and anisotropy were also spatially controlled. The rates of growth of the histoblasts at the center or at the periphery of the nest were initially comparable, but the peripheral ones became larger as the tissue expanded (Fig 4A and B). Their numbers, however, increased irrespectively of their position (Fig EV2). On the other hand, the central histoblasts were more rounded than the peripherals, at all times, although this difference diminished progressively (Fig 4C and D). Both histoblasts and LECs, which were initially elongated along the nest edge (90°) (averaged ER of 2.5 and 3.5, respectively), rounded up as expansion progressed (Appendix Fig S3).

Figure 4. Histoblast size and shape.

1. Histoblast mean cell area (MCA) increases over time for all (black—the whole nest), central (blue), and peripheral (red) histoblasts.
2. Snapshots at specific time points of segmented movies depicting histoblasts with color‐coded cell areas. Color scale on the right. Scale bar = 40 μm. Note the progressive increase in the histoblasts size, specially the peripheral ones.
3. Histoblast elongation ratio decreases over time for all (black—the whole nest), central (blue), and peripheral (red) histoblasts.
4. Color‐coded snapshots of cell elongation at different time points. Color scale on the right. Scale bar = 40 μm.

Data information: In (B) and (D), the top panel represents the central histoblasts, and the bottom panel, the peripheral. We employed an arbitrary annotation and consider peripheral cells present in the two outermost rows of cells all throughout the perimeter of the nest at the earliest recorded time point.

Figure EV2. Ventral nest expansion masks.

1. Snapshots displaying the areas of central (blue) and peripheral (red) domains over time.
2. Tissue area evolution for the total nest (black), central clone (green), and peripheral clone (red).
3. Cell number evolution of the total nest (black), central clone (green), and peripheral clone (red).

Data information: Scale bar = 40 μm. Note the sustained equivalent number of cells in central and peripheral clones.

The number of T1 transitions increased with time. Initially, most intercalations oriented along the D/V axis. Later, this directionality was lost. The histoblasts at the center displayed more T1 transitions than the peripheral ones, and intercalated more frequently in the D/V orientation (Appendix Fig S4).

Larval epidermal cells extrusions were mostly anisotropic and followed a precise spatiotemporal pattern. LECs preferentially extruded (> 50%) at the nest border. Those in anterior lateral positions delaminated at phase 1, while those located posteriorly extruded later. This pattern showed a spatial association with that followed by the nests’ shape changes (Appendix Fig S5).

In brief, the distinct growth phases of the nests appear to be dynamically modulated. Oscillatory cell divisions and an increase in cell apical surfaces are accountable for nests’ growth. Tissue anisotropy dynamics, on the other hand, is mediated by changes in cell main axis directionality, orientation of cell intercalations, and, to a lesser extent, LEC delamination.

Mechanics of cell replacements

Previous reports (Ninov et al, 2007) have uncovered the active role of the actin cytoskeleton on the invasive behavior of the histoblasts. Yet, the active and passive mechanical components implementing the anisotropic growth of histoblast nests (Figs 1 and 4) and the delamination of LECs are unknown. From the 2D divergence fields (Fig 5A and B), we modeled the apex expansion or contraction of the tissue by fitting experimental data and generating analytic hydrodynamic regression (HR) flow solutions, employing the linear Stokes equation (Hernandez‐Vega et al, 2017). These models accurately described the cell growth, contraction and expansion components, and kinetics of tissue replacement. The numerical values of the modeled divergence fields let to infer: (i) pressure fields (the relative distribution over time of cortical pressure at the epithelium apex) (Fig 5C); (ii) power density fields (the pattern of the relative rates of mechanical energy) (Fig 5D); and (iii) the distribution of the stress associated with expansions or contractions (see below).

Figure 5. Histoblast mechanical landscape.

1. Snapshots of the velocity field generated by PIV from a representative movie. Measured velocity vectors in each point of the grid are represented in red.
2. Snapshots of the computed divergence field (expansion—red, contraction—blue) at different time points.
3. HR‐modeled pressure field (positive pressure—red, negative pressure—blue).
4. HR‐modeled power density (positive power/active contribution—red, negative power/resisting tissue—blue).
5. Relative surface stress over time (positive tension/high contraction—red, negative tension/high expansion—blue).

Data information: Images were taken from the same animal as in Fig 1A. Scale bar = 40 μm.

The pressure fields progressed dynamically. Within the nest, strong positive pressure developed in central areas, while in the LECs, negative values spread throughout the tissue (Fig 5C and Movie EV4). Meanwhile, power was positive at the center of the nest and negative at the periphery. In the LECs, positive power was linked to cell extrusions and always associated with surrounding negative (tissue resistance) spots (Fig 5D and Movie EV5). Last, stress within the nest was increasingly negative, especially in the central areas, remaining neutral at the periphery. In the LECs, stress was incrementally positive. In a way, stress patterns inversely correlated with those of pressure (Fig 5E and Movie EV6).

Strains, relative pressures, power densities, and tensional stresses display distinct patterns dynamically generated during nest expansion. These patterns along the observed differences in geometries, morphologies, and dynamics between central and peripheral histoblasts suggest, altogether, that the expansion of the nests and the delamination of the LECs are mechanically coupled in a radial configuration. To study this possibility, we built radial kymographs for inferred mechanical parameters (Fig EV3).

Figure EV3. Construction of radial kymographs.

1. Velocity, divergence, or mechanical parameter maps were employed to generate distance masks in reference to the nest contour (white) at different time points. Scale bar = 40 μm.
2. The different masks for each distance band (power) at five different distances to the nest border are represented on the left in sequential diagrams from top to bottom (at 400 min as a representative time). On the right, cumulative representations of the mean values for each distance band (mean over the pixels that are at the same distance to the edge) stretching (red arrows) from left to right (in blocks of 5 μm) and spanning from −40 to +40 μm (x‐axis) with respect to the nest edge (0) at the corresponding time (400 min).
3. Full cumulative kymograph representing mean value variations along the nests and LECs at different distances (x‐axis) to the nest–LEC edge (black bar) during the temporal span (y‐axis) of the expansion process (750 min). The colored LUT codes for the relative power intensities (positive and negative) for (B) and (C).
4. Representation of the average (dark blue) and SEM (light blue mask) values of relative power (y‐axis) at different distance (x‐axis) in reference to the nest–LEC edge at one particular time (700 min).

Averaged power density kymographs showed a bimodal distribution with two maxima, at the central area of the nests and far in the LEC domain. Negative values were found in between, at both sides of the border between histoblasts and LECs. This power distribution indicated that the tissue at, and adjacent to, the nest–LEC border passively responded to the opposite inputs of both epithelia (Fig 6A): pushing efforts from the center of the nest and pulling exerted by the LECs. These power sources increased over time (Fig 6B).

Figure 6. Density power and surface stress radial kymographs.

1. Density power radial kymograph of a representative time lapse. Each horizontal line displays for each time point (left scale) the mean of the power value at every distance (bottom scale) to the nest–LEC border (vertical black line). Each phase is represented in a colored background with three different densities of purple. Value scale is on the right.
2. Surface stress radial kymograph of a representative time lapse as in (A). Phases are represented in different densities of brown.
3. Graph representing the mean (line) and standard deviation (shadow) of the power density radial kymographs (left scale) at different distances (bottom scale) by phases (in coded densities of purple) for all samples.
4. Graph representing the mean and standard deviation of the surface stress radial kymographs as in (C). Values color‐coded in different densities of brown.

Tensional stress, likewise, was initially low, all throughout, but quickly progressed toward negative and positive values at the center of the nests and at the LEC territory, respectively (Fig 6C). In this way, a gradually cumulative tensional stress gradient, sustained over time, was developed (Fig 6D).

Mechanical coordination of histoblast proliferation and LEC extrusions

The coordination between histoblasts and LECs demands a tight coupling of cell behaviors. Yet, the contributions of histoblast proliferation and LEC extrusions to the topology and magnitude of power and stress fields remains to be clarified.

We first inhibited cell divisions by overexpressing Dacapo, a CDK inhibitor, with a histoblast‐specific driver (Esg‐Gal4). We found that blocking histoblast proliferation prevented nest expansion. The rate of nests’ growth was null and even became negative at times (slight nest shrinkage). LECs responded by suppressing, non‐autonomously, their delamination (Fig EV4). Nests' morphometric parameters were, nonetheless, preserved: The loss of nest anisotropy, observed in the wild type, was not as pronounced but occurred, and the nests' main orientation was conserved. Qualitative analysis of individual cells indicated that, overall, the growth of histoblasts was reduced. The histoblasts at the nest center became smaller, while the peripherals just moderately increased their size. The area of individual LECs remained essentially the same (Fig EV5).

Figure EV4. Tissue growth after inhibition of histoblast division or LEC death.

1. Phenotype of esg‐Gal4/UAS‐Dacapo and 32B‐Gal4/UAS‐P35‐overexpressing animals. Snapshots at three representative time points of WT pupae (upper panels—images were taken from the same animal as in Figs 1A and 5), non‐dividing nests (medial panels), and non‐dying LECs (bottom panels). Scale bar = 40 μm. Note how the nests in either condition fail to expand and to recapitulate WT growth.
2. Area size variation in nests and LECs (solid and dotted lines, respectively) in WT (black) or after overexpressing Dacapo in histoblasts (green) or P35 in LECs (magenta).
3. Net rate of growth in pupae of the same conditions as in (B).

Figure EV5. Cellular phenotypes after inhibition of histoblast division or LEC death.

1. WT histoblasts (red—peripheral, blue—central) and LECs (green).
2. Histoblasts from esg‐Gal4/UAS‐Dacapo‐expressing pupae.
3. Histoblasts from 32B‐Gal4/UAS‐P35‐expressing pupae.

Data information: Scale bar = 5 μm.

Tissue mechanics appeared to be largely disturbed. While initially the radial distribution of power was not affected, it became essentially flat with time (Fig 7A and B). Power was diminished both in nests and in LECs, and only sporadic LEC delamination adjacent to the nests led to occasional positive power spots. In terms of stress, no gradient was created (Fig 7C and D). Negative stress in the nests was negligible, and positive stress in the LECs was extremely reduced. In summary, an initial trigger (nest area expansion as a result of histoblast proliferation) appears to mechanically instruct LECs to delaminate and die, thus implementing tissue replacement.

Figure 7. Density power and surface stress upon interference in histoblast growth or LEC delamination.

1. Density power radial kymograph of a representative time‐lapse movie of the nest expansion process after blocking histoblast proliferation by overexpressing Dacapo. Each horizontal line displays at each time point (left scale) the mean of the power value at every distance (bottom scale) to the nest–LEC border (vertical black line). Each phase is represented in a colored background with three different densities of purple. Power value color scale is on the right.
2. Comparison of the means and standard deviations of the density power radial kymographs in phase 3 for Dacapo‐overexpressing nests (green) and wild type (purple).
3. Surface stress radial kymograph of a representative time‐lapse movie as in (A). Phases are represented in different densities of brown. Stress values color scale is on the right.
4. Comparison of the means and standard deviations of the surface stress radial kymographs in phase 3 for Dacapo‐overexpressing nests (green) and wild type (brown).
5. Density power radial kymograph of a representative time‐lapse movie after blocking LEC delamination by overexpressing P35. Power value color scale is on the right.
6. Comparison of the means and standard deviations of the density power radial kymographs in phase 3 for P35‐overexpressing LECs (green) and wild type (purple).
7. Surface stress radial kymograph of a representative time‐lapse movie as in (E). Phases are represented in different densities of brown. Power value color scale is on the right.
8. Comparison of the means and standard deviations of the surface stress radial kymograph in phase 3 for P35‐overexpressing LECs (green) and wild type (brown).

Secondly, overexpressing P35 with a LEC‐specific driver (32B‐Gal4) delayed and prevented LEC delamination and death and, non‐autonomously, blocked nest expansion. We found a reduction of the rate at which the LECs shrink, and subsequently, histoblasts became unable to expand, despite their proliferation rate being unaffected (Fig EV4). The shape of the nests was preserved, and as in the wild type, elongation was progressively lost and orientation shifted posteriorly. However, the suppression of LEC delamination abolished the differences in size and shape between central and peripheral histoblasts. Hence, the increment in size of peripheral histoblasts, non‐autonomously, depended on a LEC mechanical input (pulling). Conversely, the LEC MCAs and their elongation ratio decreased, probably as a consequence of compression by trying‐to‐expand histoblasts near the LEC–nest border (Fig EV5). Cell death was also frequently observed between the histoblasts, potentially due to hyper‐compression (see Discussion).

The assessment of tissue dynamics by PIV, after blocking the delamination of the LECs, showed an overall reduction on the average divergence in the LEC domain; while in the nests, a progressive increase, comparable to the wild type, was sustained. From these data, we inferred a severe impairment in power density and stress patterns. All throughout the expansion, the power was low and only slightly recovered with the late removal of some LECs by macrophages (Fig 7E and F). Conversely, the relative stress in the nest was unaffected, while it was strongly reduced for LECs. Radial kymographs pointed, both for power and for tension, to strong differences with the wild‐type condition. The histoblasts at the periphery of the nests generated positive power and were not pulled by the adjacent negative LECs. Regarding stress, the relative radial differences were small within the nests, with the LECs in proximity to the nest displaying negative values. This low stress suggests that LEC delamination is a major determinant for the generation of stress in normal conditions (Fig 7G and H). In brief, upon blockage of LEC delamination, as the histoblasts keep proliferating, the tissue undergoes general compression.

Tissue stress tensional pattern

To validate the stress gradient inferred by HR, we applied targeted laser microsurgery. We performed 20‐μm‐long laser cuts and measured their retraction velocities by PIV. The laser cuts were done parallel to the nest–LEC border at 24‐h APF, both in the nest tissue itself, at central and peripheral positions (at 40 and 10 μm of the edge, respectively), and in the LECs (Fig 8A and D). Laser cuts resulted in tissue retractions with a velocity proportional to the stress level (Fig 8B and Movie EV7). The data were essentially equivalent for cuts performed along the A/P or the D/V axis, despite the nest anisotropic growth. Cuts at the periphery of the nest showed, on average, faster retraction velocity (higher stress) than central ones (P < 0.001), and much less than the LECs (P < 0.001; Fig 8C). These results strongly support a central–peripheral LEC‐graded tension profile and, indirectly, the bimodal force profile detected by HR.

Figure 8. Surface stress patterns.

1. Laser microsurgery was performed at the different sides of the ventral nests during phase 2 (300 min) in parallel cuts [in the LEC tissue at 40 μm of the nest–LEC border (cyan) and within the nests at 10 μm (red) and 40 μm (green) of the edge]. Myo‐GFP transgenic pupae were used to visualize the cortex displacements. This example illustrates at the anterior edge the laser cut positions. Scale bar = 15 μm.
2. Cortex recoil velocity (pixels per second) at different time points after laser cutting (+10 s) for a single experiment. Note the exponentially decaying speed.
3. Inferred stress differences (brown) and mean recoil velocities (blue) as measured by PIV and in the LECs (cyan—n = 18), 10 μm (red—n = 16), and 40 μm within the nest (green—n = 20). Error bars represent standard deviations.
4. Snapshots of a laser cut in the nest area. PIV arrows are shown in red. Scale bar = 5 μm.
5. Comparison of the mean laser recoil velocities in phase 2 for Dacapo‐overexpressing nests (green) with wild type (blue) at different distances (−40—n = 16, −10—n = 22, +40—n = 14) as in (C). Error bars represent standard deviations.
6. Comparison of the mean laser recoil velocities in phase 2 for P35‐overexpressing nests (green) with wild type (blue) at different distances (−40—n = 17, −10—n = 20, +40—n = 15) as in (C). Error bars represent standard deviations.

To experimentally validate the inferred mechanical relationships between histoblasts and LECs, laser microsurgery was employed after inhibition of divisions or extrusions. Applying the same approach as above, we found that the stress gradient, as inferred by the HR analysis, was abolished in the absence of histoblast proliferation (Fig 8E, compare to Fig 7F). Further, when LECs delamination and death were prevented, high‐tension levels were only found amongst them. In this condition, the stress gradient, as expected, became shallow (Fig 8F, compare to Fig 7H).

In summary, our findings suggest that a radial stress gradient builds up as a result of an initial instructive mechanical push from the expanding histoblasts to the LECs. LECs respond by delaminating and positively feed back by pulling from the nest. Interfering with histoblast proliferation blocks the whole process, and impairing LEC delamination disturbs shape changes and the expansion of the nest.

Discussion

Collective cell organization and migration are key actions in many physiological and developmental contexts (Friedl & Gilmour, 2009). Considering the role of physical forces in these processes (Banerjee & Marchetti, 2019), how mechanics may instruct cells activities during morphogenesis remains an open question. We have uncovered that for the substitution of the larval epidermis by the adult one during metamorphosis, mechanical communication is sufficient.

During the abdomen epidermal replacement in Drosophila, growth is mostly the result of stereotypically and anisotropically distributed tissue expansions and contractions. At the cellular level, nest's expansion is the consequence of the combined action of proliferation and apical surface growth. Changes in the cell main axis directionality, orientation of cell intercalations, cell growth anisotropy, and LEC delamination contribute to modulate the nests’ shape. We have found that the active and passive elements underlying nest expansion and LEC delamination are radially distributed. The mechanical activity of the histoblasts at the center of the nests and of distant LECs gets coordinated to generate an outbound stress gradient. This mechanical link is compromised when the proliferation of histoblasts is impaired. In this condition, the elimination of LECs is also non‐autonomously blocked, power becomes essentially null everywhere, and no stress gradient is created. Likewise, preventing the death of the LECs does not affect the proliferation of histoblasts, power weakens just among LECs, and the stress gradient, although shallow and shifted, is sustained.

To note, as we focused in ventral nests, which share anterior and posterior lineages, we did not evaluate the changes in the growth pattern that could be associated with the establishment and refinement of compartment borders, already analyzed on dorsal histoblasts (Umetsu et al, 2014).

Cell proliferation and shape changes direct histoblast nest anisotropic growth

The nests’ growth proceeds through different phases (acceleration, sustainment, and slowness). These phases greatly correlate with the rates of histoblast’ divisions. We then expect, considering that histoblast proliferation is dependent on EGFR signaling (Ninov et al, 2009), that the activity of this pathway will be spatial and temporally regulated. Yet, we do not know the source of the signal that would activate the pathway, either Vein or Spitz (Shilo, 2005). It does not originate in the LECs, as this will lead to more divisions in peripheral than in central histoblasts, when it is not the case (Appendix Fig S4), and it will probably be autocrine. We do not understand either how the pathway could be slowed down; by changes in the expression of agonists, antagonists, or the receptor itself, or by chemical (Hariharan & Bilder, 2006; Kango‐Singh & Singh, 2009; Fernandez & Kenney, 2010; Kim et al, 2011; Gumbiner & Kim, 2014; Sharif & Wellstein, 2015) or mechanical inhibition of growth. Further questions arise when focusing in the dynamics and anisotropy of the nests’ growth rate. Why intersegmental fusions always precede the contralateral fusions at the dorsal and ventral midlines? This may well respond to strategic evolutionary needs, ensuring proper bounding between segments, before the initiation of collective displacements.

Epithelial morphogenetic processes are regulated by cell–cell and cell–matrix contacts and by cortex contractility (Ladoux & Mege, 2017). In most cases of epithelial tissue expansion, the cells at the leading edge are those directing the process, e.g., dorsal closure in Drosophila (Martin‐Blanco & Knust, 2001). In others, the movement is collective and traction forces apply all throughout the tissue (Trepat et al, 2009; Serra‐Picamal et al, 2012). In the pupal abdomen, the growth of the nests is strongly influenced by an increase on the average surface area of outer histoblasts. The histoblasts contiguous to the LECs change their shape as expansion proceeds, become larger, show a characteristic flattening, and exhibit low levels of E‐cadherin (Ninov et al, 2007). They appear to be very plastic, with tunable and dynamic adhesions and planar invasive capabilities. Yet, although a dynamic actin cytoskeleton is essential for histoblast expansion (Ninov et al, 2007), an actomyosin contractile ring at the leading edge, which in other processes helps to stretch the epithelium (Martin‐Blanco & Knust, 2001), is not observed. Indeed, it is not advantageous here, as it will prevent the 2D spreading of the nests. Although leader cells may provide guidance for migration, they do not appear to play a major role in force generation.

Inference of epithelial mechanical parameters

During morphogenesis, the material properties of tissues are dynamically modulated. Cells may flow like a liquid, as they change positions, or become stiff once the tissue achieves structural stability (Lecuit et al, 2011). At the mesoscopic level, cells behave as active gels with non‐conventional properties. When the length scale of cellular structures largely exceeds that of the cytoskeleton components, both elastic (Kopf & Pismen, 2013; Banerjee et al, 2015) and fluid models (Arciero et al, 2011; Lee & Wolgemuth, 2011; Recho et al, 2016; Blanch‐Mercader & Casademunt, 2017; Blanch‐Mercader et al, 2017) could be employed to describe morphogenetic dynamics. Yet, no continuum model is available, that could permit to understand the overall rheology of epithelial tissues across timescales.

HR has proven to be a powerful method to extract biomechanical parameters from 2D + T movies (Hernandez‐Vega et al, 2017). Like other methods, such as video force microscopy (VFM; Brodland et al, 2010) or Bayesian analyses (Ishihara & Sugimura, 2012; Ishihara et al, 2013, 2017), HR fails to provide absolute measurements. Yet, relative spatial and temporal profiles of mechanical power and cortical stress are easily and properly inferred. We adapted HR to the analysis of the expansion of the histoblast nests by modeling the nests and LECs as a dynamic planar cortex immersed in an incompressible viscous Newtonian fluid (cytoplasm). The expansion or constriction of the cortex reaches elastic equilibrium dissipating its elastic energy by generating flows. From these flows, the out‐of‐equilibrium stresses could be extracted by low Reynolds number hydrodynamics. In other words, relative pressures, power densities, and tensional stresses could be inferred from the cortical elastic stresses estimated from surface flows.

The analyses by HR of wild‐type and mutant conditions let to infer temporal and topological differences in power density and stress throughout the epithelial replacement process. In terms of stress, these differences were fully validated by laser microsurgery. Their magnitude (predicted relative stress versus measured stress), however, did not quantitatively correlate. HR probably overestimates the stress differences as it takes into account tissue expansion or contraction only. Underlying asymmetries influencing structural tension (the mechanical properties of the cells themselves, or the stress exerted by the underlying muscles, or the 3D deformation of the tissue adapting to the abdominal cylindrical shape) were not considered. Clarifying these problems will demand further studies and technical improvements.

Mechanical coupling of growth and death

It has been claimed that the mechanochemical coupling of cell adhesion and contractility is at the basis of tissue morphogenesis (Howard et al, 2011). During their expansion, histoblasts do not confront a free space but a well‐defined larval epithelium. Along this process, LECs do not behave as passive element and they actively delaminate as the epithelial nests grow (Ninov et al, 2007, 2010; Nakajima et al, 2011; Bischoff, 2012; Arata et al, 2017; Teng et al, 2017). Delamination, to avoid any type of transient or permanent gap in the epithelium, is coordinated with the expansion of the nests, and our data indicate that the initial pressure imposed by the proliferation of histoblasts is mechanically transmitted to the LECs. As a result, delamination and apoptosis are triggered in distant LECs, which pull outward from the leading edge of the nests. Histoblasts at the boundary expand along the main stress axis as a result of the pushing and pulling forces generated, respectively, at the center of the nest, and at the distant LECs.

While the expansion of the nests is prevented upon blocking LEC death, histoblast proliferation is not affected (Ninov et al, 2007). In this scenario, the overcrowded histoblasts are unable to enlarge and die in high numbers. Overcrowding result in mechanical cell competition by compression. Local compressive stress is expected to escalate, and cell elimination ensures the recovery of cell density (Shraiman, 2005; Brás‐Pereira & Moreno, 2018). Tissue crowding, by convergence or corralling, has been proposed to drive cell elimination in multiple tissues; e.g., the Drosophila pupal thorax (Marinari et al, 2012; Levayer et al, 2016). Yet, while local increase in cell density, or a caspase‐dependent process, would promote live cell extrusion or delamination, it is debatable, and in the case of the histoblasts remains to be determined. Regarding mechanical coordination, both, the roles of cell competition (fitness fingerprints; Levayer et al, 2016) or the JNK pathway (Eisenhoffer et al, 2012; Marinari et al, 2012) might be instrumental.

A final open question in this context is the morphogenetic meaning of cell polarization. Polarization has been linked to the orientation of lamellipodial/filopodial protrusions and actin stress fibers, and, on a more mesoscopic level, with the alignment of cells with respect to the organism axes (Devenport, 2014). Cell polarization dictates the direction of growth and might be coordinately affected by forces from neighboring cells. During expansion, we have found that cell elongation affects the orientation of anisotropic growth (Mangione & Martin‐Blanco, 2018). While we did not analyze/discriminate stress polarity in a radial configuration, we found that the final global polarization of the abdominal epidermis responds to the axial alignment of histoblast orientation. This constitutes an emerging property of the tissue, dictated by the activity of the Ft/Ds/Fj pathway, which define a patterned cell adhesiveness landscape (Mangione & Martin‐Blanco, 2018). Whether this is linked to mechanical efforts, remains to be explored.

In summary, we have uncovered general principles, at the cell, tissue, and biomechanical level, directing the development of the adult abdominal epidermis within the Drosophila pupae. In this scenario, the plasticity and adaptability of histoblasts and LECs ensures a robust process, where events are self‐compensated by biomechanical means.

Materials and Methods

Fly stocks

• w; If/CyO; ZCL2207::hh‐DsRed/MKRS

• y,w; UAS‐P35; ZCL2207::hh‐DsRed/MKRS

• w; tubGal80 ^ts ; P{GawB}32B

• w; Esg‐Gal4; ZCL2207::hh‐DsRed/MKRS

• w; tubGal80 ^ts ; UAS‐Dacapo

• y,w,cv,sqh ^AX3 ; P{sqh‐GFP.RLC}

Target expression

The Esg‐Gal4 was used as a histoblast‐specific driver (w; Esg‐Gal4; ZCL2207::Hh‐DsRed/MKRS) and crossed with w; tubGal80 ^ts ; UAS‐Dacapo to block histoblast proliferation. To interfere with LEC delamination at the replacement time (after 15 APF), w; tubGal80 ^ts ; P{GawB}32B flies were crossed with yw; UAS‐P35; ZCL2207::hh‐DsRed/MKRS and maintained at 18°C until pupal stages. Pupae were shifted to 29°C at 6–8 h of APF, kept for 6 h at this temperature, and then imaged at 25°C O/N.

Live imaging and processing

Pupae carrying the ZCL2207 septate junction marker (Mangione & Martin‐Blanco, 2018) were removed from the pupal case and placed on a drop of Voltalef 10S halocarbon oil (VWR International) in a glass‐bottom culture plate (MatTek) with a wet piece of paper to avoid specimen desiccation. Pupae were oriented to monitor the 3^rd or 4^th ventral nests at mounting. This required lifting the ventrally located legs.

Confocal live imaging of late pupa was performed with either a Leica TCS SP5 or Zeiss LSM 700 or LSM 780 confocal microscopes at 1,024 × 1,024 pixel resolution (pixel size 0.312–0.326 μm) with a 40× oil immersion objective (HCX PL APO 40×/1.25–0.75 OIL CS), with a 400–600 Hz acquisition frequency. Laser intensity was kept to a minimum to avoid photobleaching and minimize phototoxicity. The animals survived the dissection and data acquisition, and developed into adult stages. Z‐stacks were acquired every 1–1.2 μm for 20–22 μm in the red and green channels. Time lapses were captured every 5 min from 15 to 25–30 h APF.

ImageJ/Fiji (NIH Image) software was used for image processing and to create time‐lapse movies. The movies were cleaned out, cropping in each Z‐stack the undesired out‐of‐the‐plane signal interfering with the analysis (macrophages, trachea, or cell debris). For segmentation, Z‐planes were projected with maximum intensity; contrast was enhanced with 3% of saturated pixels and normalized; noise outliers (ratio = 1, threshold < 5) were removed, and the image was despeckled. Images were saved as single TIFF files for loading into the Packing Analyzer software suite (Aigouy et al, 2010).

For PIV measurements, despeckling was applied to the cleaned Z‐planes that were projected with maximum intensity. An FFT band‐pass filter (parameters: up to 8, down to 5) was applied to the projection, and the images were saved as single TIFFs.

Image segmentation

To track cells in 2D, we used the Packing Analyzer software (Aigouy et al, 2010). This software is based on watershed algorithms and allows the automated detection of fluorescently labeled cell borders, retrieving vertex and cell position information. Several steps of manual user corrections were applied.

For analysis, only cells remaining in the field of view at all time points were considered. The Packing Analyzer software yielded a csv table with vertex information for each cell. This vertex information was compiled and organized by homemade software written in Mathematica. These scripts generated Excel files containing, for every time point and for each cell, its identity, center coordinates, vertex number and coordinates, area, perimeter, x component of velocity, y component of velocity, velocity magnitude, neighbors’ number, and neighbors’ identities. A second and third file indicating the orientation of intercalation events, at each time, and the identities of each dividing cell, for every time point, were also generated. The data were then exported to homemade written algorithms in MATLAB that reconstruct individual matrices for each parameter per time point. These matrices display the corresponding parameter values within the area occupied by each cell. On these matrices, binary TIFF masks from the Packing Analyzer were employed to extract nest, LECs, central domain and peripheral domain parameters, and their means, for each segmented time‐lapse data. The binary masks were also used to calculate the orientation of cell divisions, according to the positions of daughter cells. To facilitate the interpretation of the data, graphs were averaged (coursed grain T − 1, T, T + 1 means). Photoshop CS5 was used to mount the Figures.

PIV and hydrodynamic regression analysis

MATLAB software (MatPIV—http://folk.uio.no/jks/matpiv/index2.html) was used for the 2D analysis (Supatto et al, 2005), on windows of 128 pixels (3/4 overlaps), resulting in a 39 × 39 pixel matrix, with x and y velocity components. To fit the divergence from the PIV measurements for all time points (5‐min intervals), and to infer mechanical parameters, we employed hydrodynamic regression (HR), assuming that the process was governed by the linear Stokes equation and that the tissue was incompressible (Hernandez‐Vega et al, 2017).

Numerical data on flows can be obtained from modeled velocity fields fitted by regression. To do so, we employed analytical solutions of the dynamics of an incompressible viscous Newtonian fluid with negligible inertia. This Newtonian description of dynamics at low Reynolds number can be generalized to viscoelastic materials, such as the cortex, which exhibit an initial linear response of their strain rate to stress when these are small enough. The dynamics is governed by the volume continuity equation (1), ensuring volume conservation, and the linear Stokes equation (2):

$$\text{div}\mathbf{V}=0$$ (1)

$$\mathrm{\nu }\mathbf{\Delta }\mathbf{V}=\mathbf{grad}\mathrm{P}$$ (2)

with ν the dynamic viscosity, V the velocity field, Δ the Laplace operator, grad the gradient, and P the pressure.

The Stokes equation provides a unique solution for the pressure field for a given velocity field. Otherwise, we can exploit the linearity and locality of the Stokes equation to express the fluid velocity as a sum of elementary known solutions. This approach simplifies considerably the regression to obtain a theoretical approximation to the measured velocity field. From the velocity, one can determine, the corresponding stress σ _ij and, assuming that the viscosity is constant, the variation of the local value of the tension at the cortex, τ, in both principal directions. Finally, we defined and mapped a mechanical surface power density Π, the cortical elastic energy produced per unit of time and surface. At any given moment, the surface power density Π, supplied by the cortex at each point of its surface, is opposite to the rate of change in internal cortical elastic energy density, and can be derived from the measured stresses and strains, in local Cartesian coordinates, in the plane tangent to the cortex surface. Π quantifies the cortical elastic energy released per unit of time and surface (see theoretical considerations in Hernandez‐Vega et al, 2017).

To perform the regression analysis, and the modeling of the continuous elastic surface, we first defined Nε elementary solutions of equations (1) and (2). To model a planar hydrodynamic cortex, we distributed multiple elementary solutions of the linear Stokes equation at different points of a planar 2D grid. Each elementary solution consists in a dipole built as two Stokeslets pointing in opposite directions a distance δ away from each other, and distributed on a geometrical surface, modeling the cortex. A Stokeslet provides the 3D velocity field, and corresponding pressure, created by a point force in a viscous fluid at low Reynolds numbers. Due to the linearity of the Stokes equation, the velocity and pressure fields around the cortex can be expressed as a linear combination of the N_ε elementary solutions, V _e and p_e. Each solution e (1 ≤ e ≤ N_ε) is centered at a given position on a grid at the cortex and oriented perpendicularly to it. A multiplicative coefficient β_e was applied to each elementary solution. The independent variables of the regression were the coordinates of the grid points, X_e, Y_e, and Z_e, and the dependent variables, the coordinates of the grid velocity vectors, ${\mathrm{V}}_{\mathbf{e}}^{\mathbf{m}}$. We performed the regression iteratively with Monte Carlo simulations, considering the mean square errors (MSE) between measured and simulated velocity fields. It should be noted that such an approach, when restricted over the 2D cortex, isolates the leading order term of the general multipolar expansion of the velocity field on the cortex surface. Thus, it is not explicitly dependent on the details of the 3D velocity field in the bulk fluid. Recasting the flow fields in this way can, in some cases, present the drawback that simultaneous interpretation of the elastic moduli and the stress environment is hindered. We here consider only the relative stresses, pressures, and power, mitigating the need to consider separate elastic moduli altogether.

It should be further noted that HR does not determine the mechanical properties of single cells, but rather it allows for a principled coarse‐graining of the mechanical parameters at the mesoscopic level. It is capable of estimating the mechanical activity linked to tissue expansion and contraction, identifying the domains where dynamic stress (τ) and mechanical power density (Π) concentrate.

We assumed that the movements of the epithelium recapitulated a fluid flow, which is a solution of the Stokes hydrodynamic equation, that relates measured movements and deformations to mechanical stress and energy. Thus, from the 2D divergence in the plane (we did not consider vorticity), positive or negative, we modeled apex expansion or contraction.

The power density identified the positions (areas) within the epithelium resisting (negative power), or actively contributing (positive), to the replacement process. Positive power can be generated by active extension or constriction events at cell apices, while negative power denotes the passive reaction to deformation of surrounding tissues. On the other hand, positive tensional stress maps to areas of high contraction, while negative stress maps to areas of high expansion (relative compression).

Radial kymographs calculation

A nest binary mask was created for each time point in segmented movies and inverted (value 0 in nest, 1 outside nest). These masks were used to calculate the internal and external distance maps (distance map plug‐in in ImageJ). The distance map calculates for each pixel the distance to the more proximal pixel with value 1. The internal and external distance maps were added to create the complete distance map. The distance mask reflects the distance of each pixel to the nest–LEC border. From the mask, areas of 6 pixel width (approximate one cell diameter) at increasing distances were used to calculate the mean mechanical parameters for each time point. They were subsequently represented in radial kymographs with MATLAB software.

Laser microsurgery

Laser cuts were performed with a Zeiss Axiovert 200M microscope in a custom platform with an Olympus MMI Microdissector and a 63× objective. Cuts were 20 μm long in all cases and at 10 or 40 μm (histoblasts), or at 10–15 μm (LECs) of the histoblast–LEC border. Acquisition was continuous (interval 0.5 s).

Movies were rotated to display cuts in a horizontal position and were cropped to 160 × 160 pixels with the cut in the center. PIV with a window size of 64 and overlap 3/4 was applied to 10 precut and 60 post‐cut time points. The window size and overlap led to a 7 × 7 grid. For each time point, the retraction velocity was considered to be the mean of the magnitude of the y component of the PIV velocity vector excluding the central row (cut row).

Author contributions

CP‐R and EM‐B performed the measurements. CP‐R, P‐AP, JB, and EM‐B analyzed and quantified the data. P‐AP developed the HR algorithms. JB designed the tracking scripts. CP‐R, P‐AP, JB, and EM‐B interpreted the data. CP‐R, P‐AP, JB, and EM‐B contributed to the conception and design of the study. EM‐B wrote the manuscript. All authors provided critical revisions and approved the final manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Movie EV5

Movie EV6

Movie EV7

Review Process File

The EMBO Journal (2020) 39: e103594