The pupal notum is a monolayer columnar epithelium that will become the epidermis of the adult fly. At the time of our analysis, 12-18h after puparium formation (APF), it sits atop a basement membrane that is still developing (Mehaffey et al., 2024), and it is composed of diploid cells that undergo regular mitotic cycles (Guirao et al., 2015) (Figure 1—figure supplement 1A-C). To analyze cell behaviors around wounds, we live-imaged after laser ablation. Epithelial cell borders were labeled by the adherens junction protein p120ctnRFP (Ogura et al., 2018) and nuclei were labeled with histone His2GFP. Two hours after wounding, we observed what appeared to be prominent syncytial (multi-nucleated) cells around the wound (Figure 1A); experiments described below demonstrated that these were indeed syncytia. Some appeared to contain over a dozen nuclei within epithelial borders. For both syncytial and mononuclear cells, it was difficult to assign nuclei precisely to cell borders because notum epithelial cells are not oriented at right angles with respect to the surface, and in 2-D projections, a nucleus was frequently observed outside the cell’s apical border (Figure 1—figure supplement 1D).

Accordingly, using apical area and nuclear density, we estimated the number of nuclei within the 3 largest syncytia in different wounds. The number of nuclei in these syncytia increased over time: at 1 h after wounding the three largest syncytia contained an average 3-13 nuclei, and 2 h after wounding they nearly doubled to 6-20 nuclei (Figure 1B). Interestingly, larger wounds generated syncytia with larger apical areas and more nuclei, proportional to wound size, suggesting syncytium formation is a dynamic and scalable response to injury (Figure 1 C-E).

Live imaging of cells after wounding revealed the gradual loss of p120ctn along the border between two epithelial cells followed by rounding of the resulting syncytium (Figure 1F,I, and Video 2), such that the initial two-cell morphology gradually became a one-cell morphology. We termed the visible loss of p120ctn “border breakdown”. We determined that as p120ctn was lost, the adherens junction protein E-cadherin was also lost (Figure 1—figure supplement 1E-H) indicating adherens junctions disassembly along the border between the cells. Border breakdowns were found spatially clustered in the first three to four rows of cells, 30-50 μm from the wound center, and they occurred primarily within the first hour after wounding, some within 10 minutes after wounding (Figure 1G,H). These events could be explained by wound-induced cell fusion or by wound-induced epithelial-to-mesenchymal transition. To ask whether border breakdowns represented cell fusions, we analyzed cytoplasmic mixing. Individual GFP-labeled cells (clones) were generated at random locations using the flip-out Gal4 technique (Pignoni and Zipursky, 1997). Before wounding, the level of cytoplasmic GFP fluorescence was stable yet exhibited cell-to-cell variability, allowing some differentiation of cells by intensity (Figure 1Iⁱ). Minutes after a nearby laser ablation, GFP was observed to diffuse from labeled cells into neighboring unlabeled cells. GFP mixing between two cells was followed by the eventual loss of their shared p120ctn-labeled cell border, confirming that border breakdown is indeed cell fusion, but that cytoplasmic mixing occurs more than 10 minutes before the border breakdown is first observed to start (Figure 1I, Video 2). GFP mixing always preceded border breakdowns (n=11). Cytoplasmic GFP mixing resulted in a consistent level of fluorescence across the two cells, indicating that GFP mixing is a reliable indicator of plasma membrane fusion. The order of events demonstrates that plasma membrane fusion occurs many minutes before the visible disassembly of the apically located adherens junctions. Thus, epithelial fusion is a rapid local response to wounding, and border breakdown is a lagging indicator of cell fusion.

Border breakdowns, representing cell fusion events, occurred mostly within 1 h after wounding, so it was unclear how syncytia grew in size and nuclear number between 1-2 h after wounding. However, an unexpected cell behavior was frequently observed during this time: the apical area of diploid epithelial cells shrank until they disappeared from the epithelial sheet, a phenomenon we termed “apical shrinking” (Figure 2A). Although the appearance of apical shrinking suggested cell extrusion, we were surprised at its high frequency around wounds (Figure 2E), as wounds have already lost many cells and would be expected to prioritize cell survival to speed barrier repair. To understand the fate of apical-shrinking cells, we tracked their nuclei and determined that they did not extrude; rather, all apical-shrinking cell nuclei moved laterally within the plane of the epithelium to join neighboring syncytia (n=7, Figure 2B), suggesting that apical shrinking is an indicator of cell fusion. To better understand this behavior, we analyzed individual GFP-labeled cells so that we could observe the fate of the cytoplasm.

Like with border breakdowns, GFP mixing in the X-Y plane of the epithelium was observed to precede the initiation of apical shrinking, although the interval between GFP mixing and apical shrinking was longer, one or more hours (Figure 2C, Video 3). In X-Z projections, GFP labeled cytoplasm associated with apical shrinking shifted laterally to join with a neighboring cell as shown for two apical-shrinking cells in Figure 2D. For the cell labeled with the yellow arrow, before apical shrinking at 60 min post wound, its GFP-labeled cytoplasm extended from the plane of the adherens junctions (Z_AJ) basally over 8 µm; when apical shrinking was underway at 120 min post wound, its cytoplasm extended less than 5 µm down from the apical plane; and when shrinking was almost complete at 150 min post wound, it extended only 1-2 µm down from the apical plane, appearing as a whisp of cytoplasm connecting the shrinking apical surface with the neighboring syncytia. A similar loss of basal depth over time is observed in the cell labeled with the white arrow. In both these cells, as the cell lost its basal volume, cytoplasm moved laterally to join the neighboring syncytia. Thus, analysis of both nuclei and cytoplasm demonstrated that apical shrinking represented cell fusion events. The spatial distribution of apical shrinking around wounds was similar to border breakdowns, but fusions associated with apical shrinking occurred later and were more numerous (Figure 2E,F). Thus, cell fusion is evident as either border breakdown or apical shrinking, which both occur after GFP sharing.

To determine what percentage of cells around a wound will fuse, we analyzed over 100 single epithelial cells randomly-labeled with cytoplasmic GFP within the radius of observed fusion (80 µm), tracking them for 6.5 h to assess their fate (Figure 3A,B): a full quarter of the cells fused (25%), sharing GFP before borders breakdown or apical shrinking; 67% persisted as diploid cells, most with stable GFP, but infrequently (n=3) with GFP mixing and no subsequent cell fusion; the remaining 7% could not be tracked (Figure 3—figure supplement 1A). All GFP-labeled cells with apical shrinking had previously shared cytoplasm, indicating that all apical shrinking represents fusion. The observation of rare diploid cells with previous cytoplasmic mixing confirms that plasma membrane fusion occurs before the later steps of border breakdown/apical shrinking, as these cells represent fusion initiation events (fusion pore) but were unable to productively stabilize and expand the site of fusion and so returned to the diploid state. Fusing cells were strongly skewed toward the center of the wound: within 70 µm, about half the cells (47%) underwent fusion (Figure 3A, left), and fusion continued for over 300 min after wounding (Figure 3—figure supplement 1B,C). Apical shrinking accounts for more than half of all fusing cells (Figure 3C), and the spatial distribution of fusing cells that shrank vs. lost borders was similar (compare Figures 1G and 2E). However, apical shrinking began later than border breakdowns (Figure 2F, Figure 3—figure supplement 1C), continued for several hours after wounding (Figure 2F, Figure 3—figure supplement 1C), and took longer to complete (Figure 3—figure supplement 1D). Thus, fusion associated with apical shrinking accounted for continuing syncytial growth after border breakdowns subsided. Further, cell fusion is a persistent behavior over the course of wound closure.

Our individual cell tracking indicated that syncytia were the result of epithelial cell fusions. However, phagocytic hemocytes are known to migrate to the sites of pupal wounds (Antunes et al., 2013). To determine if hemocytes fused with epithelial syncytia around wounds, we labeled hemocytes with cytoplasmic GFP (hml>GFP) and observed their behavior in live imaging. In six wounds, we observed GFP-labeled hemocytes at the basal side of the wound margin; however, we did not observe any events where GFP was transferred from hemocytes to the epithelium, ruling out the fusion of hemocytes with epithelia in pupal wounds (Supplemental Video 4) .

To evaluate the role of cell fusion and syncytia in repairing wounds, we wanted to analyze a mutant that blocked syncytia formation. It has been reported that several individual autophagy (Atg) genes are required for syncytia formation around small laser ablation wounds in the polyploid Drosophila larval epithelium (Kakanj et al., 2022), so we asked if autophagy promotes fusion around laser wounds in the diploid pupal epithelium. Using a dsRNA line known to disrupt syncytia in larvae, we knocked down Atg1 on one side of the wound expressing pnr-Gal4, as illustrated in Figure 4A, reserving the other side of the wound as an internal control, modifying an internally controlled wounding system we have previously reported (O’Connor et al., 2021b). Visualizing Ecad-labeled cell borders across the wound, we found that indeed Atg1 is required for syncytia formation in pupal wounds, like it is in larval wounds (Figure 4B).

Specifically, we observed a reduction in cell fusion, both reductions of border loss and apical shrinking (Figure 4C). We next compared wound closure rates on the two sides of the wound (Figure 4D,E). Because leading edge migration is slightly delayed on the pnr-Gal4 side of the wound even in control wounds (pnr>+), we compared the difference in wound radii on the two sides, comparing control wounds to pnr>Atg1^RNAi wounds. By 5h after wounding, the pnr>+ side lagged behind the other side by only about 10 µm, yet the pnr>Atg1^RNAi side lagged behind the other side by an average of about 25 µm, a reproducible, significant, and large reduction in wound closure when syncytia formation was reduced (Figure 4E). These results suggest that syncytia speed wound closure. It is difficult, however, to disentangle the roles of autophagy and syncytia in this experiment, as the Atg1 knockdown cells may move more slowly because they are defective in the essential cellular function of autophagy and/or lacking syncytia. To address this question, we found a strong correlation between the size of the cell/syncytium and the distance it had moved from the start, measured at the time when the wound was half-closed. This correlation was true for cells/syncytia in control wounds as well as for those when Atg1 was knocked down (Figure 4F,G respectively). These data indicate that syncytia formed through cell fusion speed wound closure.

To understand this phenomenon better, we further investigated control wounds. Because about half the cells fused to form syncytia around wounds, we were able to compare the behavior of syncytial to non-syncytial cells within the same wound, providing a well-controlled environment for assessing how syncytia contribute to wound closure. In live imaging, we observed that syncytia frequently moved faster than mononucleated cells toward the wound, overtaking unfused cells as the wound closed. Figure 5A tracks four GFP-labeled cells near a wound, with cell 1 furthest from and cell 4 closest to the wound (Figure 5Aⁱ). Middle cells 2 and 3 fuse with five others into a large syncytium, which advances toward the wound faster than both non-fusing cells, and by 400 min after wounding the syncytium extends well beyond the unfused cells (Figure 5Aⁱⁱⁱ). This behavior was evident even without GFP labeling: Figure 5B-C show a group of cells at the leading edge of the wound (the wound is recognized by the lack of p120ctn). At 90 min after wounding, this region of the leading edge comprises three mononuclear cells flanked on either side by syncytia (outlined in yellow); the middle of the three mononuclear cells is outlined in orange (5B) and white (5C). By 150 min, the syncytia have pushed out all three mononuclear cells from the leading edge, and at 360 min two of these mononuclear cells are visible about 20 µm back from the wound edge. Indeed, we observed that several hours after wounding, the entire leading edge was occupied by syncytia. To investigate how syncytia came to occupy this position at the front lines of wound healing, we expressed MyoIIGFP/Zip-GFP, which along with actin forms the contractile purse string and makes the leading edge visible, along with p120ctnRFP to label cell borders. We analyzed the persistence of all mononuclear and syncytial cells at the leading edge over the course of closure for three wounds, starting when the leading edge was first visible about 30 min after wounding (Figure 5D-G). At the start, 75% of the perimeter was occupied by syncytia with the rest occupied by 10-13 mononuclear cells (Figure 5D). As the wound closed, the syncytia became larger and displaced all the mononuclear cells, with the last mononuclear cell removed from the leading edge 20-160 min before closure (Figure 5E-G). No mononuclear cell persisted at the leading edge through wound closure, indicating that syncytia outcompete unfused cells to close the wound.

Why do syncytia move faster? To address this question, we considered the geometry of fusion because different fusion orientations could provide different benefits for wound repair. We observed that fusions sometimes occur between two adjacent cells equidistant from the wound, as diagramed in the top panel of Figure 6A and exemplified in Figure 1Iⁱⁱⁱ-I^iv (cells 1,2). This breakdown of a radial border produces a syncytium elongated along the wound edge.

Alternatively, fusion may occur between adjacent cells at different distances from the wound, as diagramed in the lower panel of Figure 6A and exemplified in Figure 1F (cells 3,4). This breakdown of a tangential border produces a spoke-like syncytium pointing into the wound. To analyze the frequency and timing of these two different axes of fusion, we calculated the orientations of all lost borders in four wounds. (The angle of the lost border can be determined for border breakdowns; it is difficult to determine for apical shrinking because the other fusion partner is hard to identify). In four analyzed wounds, we identified 235 border breakdowns: 39 radial borders and 196 tangential borders (Figure 6B). Thus, there were about five-fold more tangential borders lost to fusion than radial. Nonetheless, the two types had similar temporal and spatial distributions around the wounds (Figure 6—figure supplement 1).

As a wound closes, fewer cells can occupy the ever-smaller leading edge, necessitating a rearrangement of cell adhesions to allow cell intercalation, as depicted in the top of Figure 6C. Indeed, when cells do not fuse after wounding, as in Drosophila wing discs, the impact of intercalations on tissue fluidity is known to be a rate-limiting step of wound repair (Tetley et al., 2019). As shown in the bottom of Figure 6C, the requirement for cell intercalation would be reduced specifically by radial border fusion.

To consider the contribution of radial border fusions to wound closure, we took three approaches. First, we analyzed existing data from the three wounds shown in Figure 5D-G, and asked how many leading-edge cells fused radially vs. intercalated: 17-41% of cells removed from the leading edge were removed through fusion, reducing the burden of intercalation substantially (Figure 6D). Interestingly, the larger the percentage of cells that fused rather than intercalated, the faster the wound closed (Figure 6D).

Second, we analyzed tissue fluidity as measured by the cell shape index, ! = #⁄√%, where P is cell perimeter and A is cell area. This dimensionless index characterizes the cell jamming transition in vertex models of epithelia: for π < 3.81, a modeled isotropic epithelium is in a jammed state with energy barriers that limit intercalations, giving rise to a solid-like tissue; but for π > 3.81, the energy barriers vanish, yielding an unjammed fluid-like tissue (Bi et al., 2016; Park et al., 2015). In real epithelia, the solid-fluid transition may take place at different values of π (Wang et al., 2020), but higher π still implies greater tissue fluidity. We segmented images of the pupal notum before and after wounding and calculated the distributions of cell shape index for both syncytia and non-syncytial cells (Figure 7A-B). The mean cell shape index increased significantly after wounding with syncytia having a higher shape index than nearby post-wound non-syncytial cells (p = 0.001) and both having higher shape indices than pre-wound cells (p < 10^-6). The post-wound epithelium is thus substantially more fluid-like than it was before wounding

Third, we adapted the computational vertex model from Tetley et al. (2019) to include probabilistic cell-cell fusions to directly investigate the role of cell fusion in tissue fluidity and wound closure. We generated four initial cell configurations with similarly sized wounds and used these to run four matched pairs of simulations – one that proceeded to close without any cell fusions and the other that allowed fusions to occur with spatiotemporal probabilities that matched our experimental observations (Figure 6B, Figure 6—figure supplement 1A). On average, the probabilistic fusion-allowed simulations had 68 ± 14 (mean ± SD) fusions and closed the wound in about 1/3 less time than those without fusions (Figure 7C). Selected frames from one of the matched simulation pairs are shown in Figure 7D-E and full videos of this pair are available as Video 5. Importantly, the modeled syncytia had no special properties: same contractility as other cells, same target shape index, and same target area as the sum of the fused pair. Just the presence of these larger cells reduced the need for intercalations during wound closure and thus sped up the process.

In contrast to radial border fusions, tangential border fusions might provide a way for cellular resources that would be trapped in distal cells to move toward the wound to contribute to closure. To test this hypothesis, we generated small flip-out clones expressing actin-GFP and wounded such that unlabeled cells intervened between the labeled cell and the leading edge (Figure 8). Like cytoplasmic GFP in Figures 1I and 2C, we observed actin-GFP to equilibrate between cells soon after fusing. For syncytia that did not have access to the leading edge (n =2), actin-GFP levels remained uniform (Figure 8A,B). In contrast, syncytia with access to the leading edge (n = 10) first uniformly distributed actin (Figure 8Cⁱⁱ), but once the leading edge was contacted, they redistributed actin-GFP to it (Figure 8C,D). Kymographs of actin-GFP confirm that regardless of location, actin equilibrates between fusing cells within 5 minutes (Figure 8Bⁱⁱⁱ, D^iv); however, nearly all actin that originated in the distal cell is redistributed to the leading edge in a syncytium positioned there (Figure 8D^iv). In one striking instance, actin-GFP appeared to travel through three cells to arrive at the leading edge from its initial location three rows back (Figure 8E-G, Video 6). Importantly, although actin is labeled from only one of the fusing cells, it likely represents the location of actin from all fusing cells, explaining why syncytia are better able to occupy the leading edge. We envision that other resources would also be concentrated in subcellular regions as needed by syncytia. This ability of syncytia to concentrate pooled actin to the leading edge may allow syncytia to outcompete their smaller mononuclear neighbors and close wounds more rapidly.

Drosophila lines used in this study are in Table S1. All Drosophila lines were maintained on standard cornmeal-molasses media supplemented with dry yeast. All flies, except those used in heat-shock flip clonal analysis, were raised at room temperature. For clonal analysis experiments flies were raised at 18 degrees Celsius until the 3^rd instar stage when they were heat shocked in a circulating water bath at 37°C for 3 minutes. They were then allowed to develop at room temperature to 15-18hr after puparium formation (APF) before wounding experiments (described below) were conducted. To knock down Atg1, Atg1^RNAi was crossed to the stock for internally controlled wounding analysis, expressing Ecad-GFP and TubP-Gal80^ts ubiquitously and pnr-Gal4, UAS-nls-mCherry in the central region of the notum, similar to our previous system (O’Connor et al., 2021b). Crosses were moved to 29° for 6 days to allow expression of the RNAi construct and dissected between 12-15 hr APF.

Following imaging, pupae were kept mounted as described above and allowed to continue to develop and eclose for 3-4 days. Pupae that continued developing until they were able to crawl out of the partially dissected case were classified as ‘survived’ and their data acquired form these samples were used for analysis. If a pupae did not survive to eclosion, the associated datasets were not used in the study.

Images were collected using a Nikon Ti2 Eclipse with X-light V2 spinning disc (Nikon, Tokyo, Japan) with a 40X 1.3 NA oil-immersion objective or 60X 1.4 NA oil-immersion objective.

Unless otherwise noted samples were imaged pre-wounding, immediately after wounding, every 2 min for 30 min, and then every 10 min for 6 h. Images were pre-processed in NIS-Elements using combinations of background subtraction, rolling ball correction, local contrast, and Denoise a.i. Assembly of figure panels was done using Affinity Designer and frames were centered on the entity in focus, compensating for frame shift due to wounding.

A single pulse of a 3rd harmonic (355 nm) Q-switched Nd:YAG laser (5 ns pulse-width, Continuum Minilite II, Santa Clara, CA) was used for laser ablation. Laser pulse energies were kept to 1.9 μJ +/- 0.1 μJ, increased from our previous report (O’Connor et al., 2021b) to keep the wound size similar between old and new ablation rig.

Individual border breakdown events were manually observed using FIJI by identifying syncytia late in the video and back-tracking to determine which borders broke down to form each syncytium. These borders were traced back to the first frame after wounding to develop the map on Fig 1G. For each border breakdown event, distance from the center of the wound and time after wounding were recorded in Microsoft Excel. These events were categorized as tangential or radial based on the orientation of the border-to-be-lost relative to a vector pointing outward from the center of the wound. Specifically, the line tool in FIJI was used to measure both the angle of the border with respect to the horizon, 8, and the angle of the line from the center of the wound to the center of the border with respect to the horizon, α. If |cos(, − .)| ≥ cos 45°, then the border was classified as radial; otherwise, the border was classified as tangential.

Shrinking cells were identified in live microscopy videos by beginning at the end of the video and playing backwards in FIJI. Backwards, a shrinking cell appears to bloom from the epithelial layer, characterized by the expansion of a bright puncta of p120ctnRFP. Each cell that underwent this behavior was marked on a single frame of the video, then the distance from the center of the wound when shrinking started was denoted as well as the time that the shrinking started and completed. All cells that shrank were then manually traced back to the first frame after wounding to develop the map of shrinking cells (Fig 2E).

Nuclei and cell borders do not align in Z-projections of the pupal notum as the cells are non-prismatic. To estimate the number of nuclei per syncytium, the pre-wounding density of nuclei per unit area was determined for the circular region where syncytia form after wounding. Next, the apical area of the three largest syncytia/cells around a wound was measured at 0h, 1 h and 2 h post wounding. The area of each syncytia was multiplied by the nuclear density to yield the approximate number of nuclei per syncytia. The number of nuclei in each of the three syncytia was averaged to give a value for each of three samples in Figure 1B.

The three largest syncytia were determined by eye in FIJI for six samples, three ablated at 1.9 µJ and three ablated at 3 µJ ablation. The apical area of syncytia was measured 3 h post wounding using the p120ctnRFP signal. Initial wound size was calculated by measuring Myosin II marked leading edge when it became apparent 30-160 min post wounding.

Unfused cells at the leading edge were identified using FIJI by a lack of border breakdowns. Each unfused cell was manually observed over the duration of wound closure and the time at which it departed from the leading edge was noted. A count of unfused cells at the leading edge was created in Excel and formal graphs were generated using Prism 9. To measure percent of the leading edge comprised of unfused vs. syncytial cells, each unfused cell’s leading edge contact was measured in FIJI using the polygon line tool. The total circumference of the wound was measured using the polygon line tool and unfused cell measurements were subtracted from the total to infer the syncytial occupancy at the leading edge. Formal histograms were generated using Prism 9.

107 individually labeled GFP cells were analyzed across 5 wounds over 6.5 h. The position of the ablation was optimized to place as many individually labeled cells within 40-80 µm from the center of the wound as possible. After wounding it was possible to identify a mixing event by the decrease in intensity from the source cell with a corresponding increase in intensity of a previously unlabeled neighbor. Intensity differences made it possible to distinguish instances where two source cells were adjacent to each other but only one had a mixing event. However, large patches of source cells were not evaluated because inter-patch mixing was not distinguishable. To evaluate if border breakdowns were preceded by mixing, the 11 individually labeled cells that had border breakdowns were tracked back to the start of the video and confirmed to have a mixing event. There was never an instance where a labeled source cell had a border breakdown without a prior mixing event occurring.

A pigmented scar forms at the site of laser ablation making identifying the exact moment a wound is closed difficult. Since the scar is approximately the same size in each sample, the time point at which the Myosin II signal disappeared below the scar was used as a proxy for closure.

A cell border was counted as a location of wound-induced fusion if it met three criteria: (1) adherens junctions along the cell border were lost; (2) the lost adherens junctions were not restored; and (3) the loss of adherens junctions was accompanied by a morphological change of the neighboring cells, for example, the moving apart of tri-cellular junctions previously connected by the lost cell border. Since border-breakdown fusions are observed most frequently within 1 hr after wounding, we counted all such fusions occurring within 1 hr 40 min of wounding.

When comparing the fusion frequency in pnr versus control domains, we used a landscape-oriented rectangle centered on the wound with its height defined by the inner radius of the zone of nuclear membrane damage visualized by the release of nls-mCherry from the nucleus, which is approximately the same as the zone of lysis (O’Connor et al., 2021a).

Replacing the annular region with an extended rectangle excludes possible heterotypic fusions along the pnr-control border just above and below the wound. To calculate the frequency of fusion, in addition to the raw number of fusion borders, the total number of cell borders was also estimated by area-based scaling of the border count within a randomly selected 60-x-60-pixel (16.8-x-16.8-µm) region in the pre-wound image.

Since fusions can also appear as apical shrinking rather than border breakdown, we also quantified the number of apical-shrinking cell fusions in pnr versus control domains. Apical-shrinking cell fusions were defined as diploid cells that decreased in surface area over time and fully disappeared within a 5-hr movie. As with fusion analysis, cells near the pnr-control border were excluded. Apical shrinking cells were counted manually and their spatial locations marked using FIJI’s “Multi-point” tool.

The distance between the leading edge and the wound center over time was used to calculate the speed of wound closure. The distance was calculated using the coordinates of the wound center and fixed points on the leading edge. To identify fixed positions on the leading edge that are unbiased and consistent over time, four reference lines, all crossing the center of the wound, were drawn using the red channel with only the nuc-mCherry signal: line 1 along the edge of the pnr domain labeled by nuc-mCherry, line 2 perpendicular to line 1, and line 3 and 4 are 30° from line 2. These references lines were then overlaid on the green channel with the E-cadherin signal.

To determine the speed of wound closure, the intersection of E-cadherin marked leading edge and reference lines 2-4 were used as the position of the leading edge. That is, at each time point, a total of 6 positions of the leading edge were recorded, among which 3 will be in the pnr domain and the other 3 in the ctrl domain. The average distance (d) between the leading edge and the center of the wound were calculated at each time point for both the pnr domain (d_pnr) and the control domain (d_ctrl). The difference between control and pnr distance (Δd = d_pnr – d_ctrl) was calculated over time for both the control pupae with no genetic manipulation in the pnr domain (pnr > +) and experimental animals with Atg1 knocked down in the pnr domain (pnr > Atg1^RNAi). To determine whether Δd is statistically different between control and experimental pupae at each time point, two-way ANOVA (fit using full model) was used, comparing each cell mean with the other cell mean in that row.

To determine the distance syncytia covered during wound healing, syncytia were analyzed (1) at the leading edge when wound is half closed and (2) within the bow-tie region between reference line 3 and 4. ROIs were drawn along syncytia of interest using the “Freehand selections” tool of FIJI. In addition to syncytia location when wound is half closed, the starting position of cells that later fuse to form syncytia were also recorded in live-imaging videos, tracking the syncytia back in time. In addition to area, the coordinates of ROIs outlining syncytia at different time points were obtained using a self-written FIJI Macro. Within each ROI, the coordinate closest to the center of the wound was recorded using python, as the location of the syncytia. The distance a certain syncytium moved was calculated by subtracting the distance between the syncytium and the wound center when the wound was half closed from the distance between the syncytium and the wound center immediately after injury.

The change in the number of cells when the leading edge forms at 30 min (N_start) to when the wound is closed (N_end) is equal to the number of intercalations plus the number of radial fusion events. Thus, (intercalations = ΔN – radial fusions). For the same three samples used in Fig 4D,G, we determined ΔN from N_start at 30 min and N_end when the wound had closed. Radial fusions were tallied by manually observing border breakdown events between the leading-edge cells and intercalations were calculated. Each radial fusion was counted as one prevented intercalation event.

Images of pupae labeled with Ecad-GFP were segmented into individual cells with Cellpose (Stringer et al., 2021); cells that were not readily segmented were excluded from subsequent analysis. To avoid perimeter inflation due to pixelation, the boundaries of segmented cells were smoothed by taking the midpoint between every other pair of boundary pixels. Smoothing and calculation of perimeters, areas and cell shape indices (#⁄√%) were performed in Mathematica (Wolfram Research Inc., Champaign, IL).

Simulations of wound healing were run using vertex models implemented in Mathematica (Wolfram Research Inc., Champaign, IL) following the methods of Tetley et al. (2019)and using parameters that match experimental observations from the Drosophila pupal notum as detailed in Han et al. (2024) with the following exceptions: the mean line tension on each edge (Λ) was specified in terms of a target shape parameter (ρ₀) such that Λ=−2ρ0ΓA0\begin{document}$\Lambda=-2 \rho_{0} \Gamma \sqrt{A_{0}}$\end{document}, where the contractility (Γ) and target cell area (A₀) were taken from Han et al. (2024) and ρ₀ = 3.745 was chosen so that the actual cell shape index matched experimental pre-wound measurements; and the line tension variability (α_m) was decreased from 1.0 x 10^-2 to 9.3 x 10^-4 so that the intercalation rate in unwounded cell sheets matched experimental observations. Pre-wound configurations were generated by performing a Voronoi tessellation of a circular area with 320 randomly distributed seed points and then equilibrating the cell configuration without stochastic fluctuations until it reached a steady state. Wounds were introduced by removing all cell borders within a circular patch ∼8 cells across and removing the wound’s contributions to the energy Hamiltonian. To handle wound closure and fusion, the following additions were made to the model: (1) purse-string tension on the leading edge of the wound was incorporated as in Tetley et al. (2019) with a normalized strength of 0.115; (2) traction/crawling forces of cells on the leading edge of the wound were incorporated as a uniform tensile stress acting inwardly on the edges and vertices of the wound area (similar to the uniform stress applied to the outer edge of the patch to represent tension from surrounding cells); this inward stress had a normalized strength of Σ_in⁄KA₀ = 0.13, where K is the area elasticity modulus; (3) the cell-cell interfaces at which fusion would occur were selected stochastically with probabilities that depended on time, distance from the wound, and the radial or tangential orientation of the interface – all chosen to match experimental observations reported here; and (4) the fusion of two cells was handled by eliminating the interface between those cells and setting the target area of the fused cell to the sum of the target areas of its two progenitors. All cells, fused or not, had the same contractility and target shape parameter.

Actin-GFP intensity was analyzed using the kymograph tool in FIJI after drawing a 11-pixel line through the middle of the syncytia. Profile plot values were exported from NIS elements to Excel and graphs were generated using Prism.