Cellular crowding influences extrusion and proliferation to facilitate epithelial tissue repair.

Epithelial wound healing requires a complex orchestration of cellular rearrangements and movements to restore tissue architecture and function after injury. While it is well known that mechanical forces can affect tissue morphogenesis and patterning, how the biophysical cues generated after injury influence cellular behaviors during tissue repair is not well understood. Using time-lapse confocal imaging of epithelial tissues in living zebrafish larvae, we provide evidence that localized increases in cellular crowding during wound closure promote the extrusion of nonapoptotic cells via mechanically regulated stretch-activated ion channels (SACs). Directed cell migration toward the injury site promoted rapid changes in cell number and generated shifts in tension at cellular interfaces over long spatial distances. Perturbation of SAC activity resulted in failed extrusion and increased proliferation in crowded areas of the tissue. Together, we conclude that localized cell number plays a key role in dictating cellular behaviors that facilitate wound closure and tissue repair.

INTRODUCTION

Epithelial tissues provide a protective barrier for the body and organs throughout development and adult life. As the first line of defense, these multicellular structures are continually remodeled to sustain tissue form and function. Precise control and coordination of cell division and death are required to retain constant cell numbers during homeostasis and repair the tissue after injury. As individual epithelial cells are physically coupled together, the addition or deletion of cells causes neighboring cells to be stretched and squeezed and to move and change shape (Mason and Martin, 2011; Fernandez-Sanchez ; Navis and Nelson, 2016). This cellular reorganization results in a change in mechanical forces that are exerted on and between cells and transduced via specific biochemical signals that influence their behavior. Thus, deciphering how mechanical forces coordinate multicellular behaviors is essential to our understanding of epithelial tissue maintenance and repair after injury.

Mechanical forces generated by living cells are crucial for the control of embryological development, morphogenesis, and tissue patterning (Mammoto and Ingber, 2010; Davidson, 2011; Heisenberg and Bellaiche, 2013). Cells sense and react to forces imposed by neighbors based on changes in membrane curvature and shearing of the underlying actin cortex (Dreher ). Cells can also generate forces by dynamic reorganization of F-actin structures and the associated myosin motors that drive cellular shape changes during tissue morphogenesis (Pasakarnis ). The correct timing, location, and magnitude of mechanical forces subjected on and between cells are critical for many morphogenetic processes (Lemke and Schnorrer, 2017; Vining and Mooney, 2017), yet the inability to visualize these dynamic events in living epithelial tissues has been an impediment to understanding how biophysical forces can direct epithelial cell function.

Here we use high-resolution time-lapse microscopy of epithelial tissues in living zebrafish larvae to track individual cellular behaviors and population dynamics after injury. Our analysis revealed that localized changes in cell numbers result in heterogeneous tension distribution across the epithelium that influences the elimination of damaged or excess cells, or production of new cells. Localized increases in cellular crowding promoted the extrusion of nonapoptotic cells from the wound edge via mechanically regulated stretch-­activated ion channels (SACs), key mediators of mechanosensitivity (Coste ; Sachs, 2010; Olsen ). Interestingly, perturbation of SAC activity resulted in failed cellular extrusion and increased proliferation in crowded areas of the tissue. Together, our data establish a temporal series of mechanically induced cellular and molecular events that facilitate wound closure and tissue repair.

RESULTS AND DISCUSSION

To visualize the cytoskeleton in living epithelial cells during epithelial tissue maintenance and repair after injury in vivo and in real time, we imaged transgenic Et(Gal4-VP16) zebrafish larvae with F-actin fluorescently labeled in surface epithelial cells under homeostatic conditions and immediately after amputation (Figure 1, A and B, and Supplemental Figure S1A). Under homeostatic conditions, epithelial tissues in nonamputated 4-d postfertilization (dpf) larvae eliminate damaged or excess cells by extrusion (Figure 1A and Supplemental Video S1), as identified by the formation of an actomyosin ring in neighboring cells that contracts and ejects the cell from the tissue (Rosenblatt ). We found the number of extruding cells significantly increased after injury from 0.33 ± 0.19 cells under homeostatic conditions to 6.60 ± 0.91 cells at 15 min postamputation (mpa) and continued at similar levels for the next 45 min (Figure 1, B and C, and Supplemental Video S2). Further analysis with cell-type-specific markers revealed that extruded cells originate from both the apical periderm layer and the basal layer of p63-positive progenitor/stem cells (Supplemental Figure S1, B and C).

FIGURE 1:

Increased cell density at the wound site promotes cell extrusion. (A, B) Maximum intensity projection still images from time-lapse confocal movies of (A) nonamputated and (B) amputated Et(Gal4-VP16)) 4 dpf zebrafish larvae acquired every 15 min. Arrowheads indicate extrusions. White box indicates zoom-in region. Scale bar, 25 µm (see Supplemental Video S1). (C) Quantification of cell extrusions after amputation in untreated (n = 72) and apoptosis inhibitor (AI)-treated (n = 56) larvae over time, which were significantly increased compared with nonamputated controls both untreated (n = 28) and AI-treated (n = 26). Results are expressed as mean ± SD. (D) Quantification of cell density from maximum intensity projections of confocal images of fixed amputated larvae (black, n = 132) and nonamputated controls (gray, n = 62) over time. Error bars represent mean value ± SD. P values were calculated using two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons tests (*p < 0.05, ***p < 0.001, ****p < 0.0001). (E, F) Density maps of homeostatic control (E) and amputated (F) DAPI-stained larvae at 30 mpa. Color-coded spheres denote distance to the nearest neighboring cell. (G, H) Maximum intensity projection still images from time-lapse confocal movies of amputated Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) larvae. Arrowheads indicate cells during the process of extrusion. Scale bars, 20 µm. (I) Quantification of distance to nearest neighboring cell during the process of extrusion. Mean values are plotted from n = 12 extruding cells and n = 12 nonextruding cells analyzed from three independent data sets. P value was determined using a two-tailed, unpaired Student’s t test with Welch’s correction (****p < 0.0001).

Movie S1

Epithelial extrusion during homeostasis. Maximum intensity confocal projection images of 4 dpf (day post-fertilization) larvae acquired every 15 minutes for 1.5 hours of Et(Gal4-VP16) during homeostasis. Scale bar, 50 μm.

Movie S2

Visualizing early cellular events after injury in a living epithelial tissue. Maximum intensity confocal projection images acquired every 15 minutes for 1.5 hours of Et(Gal4-VP16) after injury. Scale bar, 50 μm.

The extrusion pathway can operate in two distinct manners: one that is stimulated by cellular crowding and the other that is induced by caspase-activation to remove apoptotic/damaged cells (Eisenhoffer ). To determine whether the cells being eliminated by extrusion were undergoing apoptosis, we examined cleaved caspase-3 activity after amputation. We observed increased levels of activated caspase-3 in larvae treated with UV-C to induce DNA damage, whereas activated caspase-3 was rarely detected in extruding cells (0.27%, 2/742 cells) near the wound edge (Supplemental Figure S1, D–H). Moreover, the number of extrusion events after amputation was also unchanged in amputated larvae treated with a pharmacological inhibitor of apoptosis (Apoptosis Inhibitor II, NS3694) (Gauron ) when compared with untreated larvae (Figure 1C). We conclude from this data that nonapoptotic cells are cleared from the wound site by extrusion after amputation.

Cellular crowding plays a key role in promoting extrusion of nonapoptotic cells (Eisenhoffer ; Eisenhoffer and Rosenblatt, 2013; Marinari ), and as a result, we quantified the distance between all nuclei under homeostatic conditions and at different times after amputation. Interestingly, amputation led to a unique and localized crowded area near the injury site (<100 μm) with a significantly increased number of cells per area (34.1%) within 10 mpa (Figure 1, D–F, and Supplemental Video S3). The cell density increase occurs 5 min before the observed extrusion events (Figure 1, C and D), suggesting a temporal link between crowding and nonapoptotic cell extrusion. To directly test whether localized cellular crowding promotes extrusion, we quantified the mean distance between individual cells that would go on to extrude and their nearest neighbors in Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) larvae with all nuclei and cell membranes fluorescently labeled and found the likelihood of a cell being extruded dramatically increased as the cell density increased by more than a factor of 1.4 (Figure 1, G–I). These results are consistent with the 1.4–1.6× critical threshold range, previously reported using in vitro crowding studies (Eisenhoffer and Rosenblatt, 2013) and mathematical modeling (Shraiman, 2005), and support the idea of a critical crowding concentration that activates extrusion of nonapoptotic cells in living tissue.

Movie S3

Cellular crowding promotes extrusion at the wound site after injury. Maximum intensity confocal projection of time-lapse imaging of Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) after amputation. Images were acquired every 5 minutes and 51 seconds for 1.5 hours. Scale bar, 50 μm.

Rapid increases in cell density can result in an increase in mechanical forces that are transduced via the activity of SACs, key mechano-regulators of crowding-induced nonapoptotic extrusion (Eisenhoffer ; Kim ). We compared wound closure and crowding-induced extrusion after injury following perturbation of SACs. After 60 min of observation, a total of 28 extrusion events (n = 6 larvae) had occurred in untreated amputated larvae, whereas treatment with gadolinium trivalent cations (Gd3+) (Yang and Sachs, 1989) or the spider venom peptide GsMTx4 (Bae ) to perturb SACs resulted in a failure of cells to complete extrusion and detach from the tissue over time (Figure 2, A–G). Failure to complete the extrusion process resulted in the accumulation of undetached cells at the amputation site after 30 min and a significant increase in wound area when compared with untreated injured larvae, from 901.2 ± 121.5 μm2 to 2558 ± 260.6 μm2 and 1750 + 144.6 μm2, respectively (Figure 2, B’, C’, and H). Collectively, our results suggest that rapid cellular crowding after injury induces nonapoptotic cell extrusion that is regulated by mechanosensitive SACs and contributes to proper wound repair.

FIGURE 2:

Blocking SACs causes failed extrusion. (A–C) Maximum intensity projection confocal images of the wound site at 30 mpa in fixed untreated (A), Gd3+-treated (B), or GsMTx-4-treated (C) amputated Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) larvae. Scale bar, 10 µm. Rotated images of the wound site (A’, B’, C’). Scale bar, 50 µm. Arrowheads denote gaps and areas of the wound that have not fully closed. (D–F) SEM images of the wound site at 30 mpa in untreated (D), Gd3+-treated (E), or GsMTx-4-treated (F) amputated larvae. Scale bar, 10 µm. (G) Quantification of the total number of successful extrusion events across six embryos in untreated (black), Gd3+-treated (light gray), and GsMTx-4-treated (dark gray) amputated Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) larvae over time. (H) Quantification of the area of the wound site in untreated (n = 51), Gd3+-treated (n = 52), and GsMTx-4-treated (n = 56) amputated larvae. P values were calculated using Kruskal-Wallis multiple comparisons tests (***p < 0.001, ****p < 0.0001).

The generation and transmission of mechanical forces can influence cell behaviors associated with wound repair (Zulueta-Coarasa and Fernandez-Gonzalez, 2017), yet how these events are coordinated in space and time is not well understood. To better define shifts in cell- and tissue-level forces that initiate changes in cellular behavior after injury, we tracked both groups and individual cells within the first 30 min after amputation. We found that cells within the tail fin exhibited an increase in average speed when compared with similar regions in homeostatic controls (Figure 1, G and H, and Supplemental Figure S2, A–F). Importantly, this analysis revealed that speed varied depending on the distance from the wound site. Based on our cell-tracking experiments, we defined three equal regions or zones near the wound site, termed Zone 1 that was closest (<66 μm) to the wound, Zone 2 that was between ∼ 66 and 133 μm away from the wound site, and Zone 3 that was between ∼133 and 200 μm from the wound site. These analyses revealed that cells further from the injury site move with increased speed compared with those closest to the wound edge, 33% increase in speed in Zone 1 (n = 36) and 98% increase in speed in Zone 3 (n = 36), respectively (Supplemental Figure S2, A–F). Interestingly, both Gd3+ and GsMTx-4 treatments led to increased movement of cells in Zone 3 (Supplemental Figure S2, A–F), indicating that SAC inhibition affects the ability to sense crowding and extrude. These data suggest that collective displacement of cells from several hundred microns away increases crowding at the wound site, and importantly, predicts areas of the tissue where cells undergo the most movement.

We next used particle image velocimetry (PIV; see Materials and Methods) to experimentally measure the velocity and strain rate fields between time-lapse confocal images after injury and after inhibiting the cells’ ability to sense mechanotranductive cues during crowding using Gd3+. Globally, the vector fields show a net flow toward the wound site, indicative of directed cell migration toward the amputated area (Supplemental Figure S3, A–D). Velocity flow maps validated our cell speed analysis and showed a gradual increase in tissue flow velocity from the wound edge (Zone 1) to several cell rows back (Zones 2 and 3) up to 10 mpa (Figure 3, A–I, Supplemental Figure S3, A–D, and Supplemental Video S4). Importantly, starting at 10 mpa, and upon increase in cell density and crowding (Figure 1D), flow velocity was similar in all three zones and constantly decreased to reach a plateau of <0.01 µm/s (Figure 3, E and G–I), indicating a temporal link between crowding and decrease in cell migration. Our data also revealed spatially localized strain fields that were elevated in Zones 2 and 3, indicative of extension due to migration (Figure 3, A, B, and E). In contrast, we observed compressive strain fields near the wound in Zone 1, presumptively due to crowding. Interestingly, the spatial trends in both strain rate and velocity were also present in Gd3+-treated fish, yet persisted for a longer period of time in all three zones (Figure 3, C–F). This dynamic analysis also showed cell speed was ∼ 0.1 µm/s higher and occurred for longer in Zones 2 and 3 after Gd3+ treatment (Figure 3, G and H), indicating that upon inhibition of SACs, cells lack the ability to mechanically sense their neighbors and maintain a high migration speed despite a continuous increase in cell density. Together, these data suggest that continuous migration and a lack of extrusion to remove the excess cells contribute to observed increased cell crowding and wound width associated with failed mechanosensation.

FIGURE 3:

Blocking SACs causes spatial and temporal changes in strain patterns and cell velocities after injury. (A–D) Strain rate maps of untreated (A, B) and Gd3+-treated (C, D) tails 0-1 and 3–4 mpa. Color code indicates the magnitude of strain rate in s-1. Colored squares highlight Zone 1 (black), Zone 2 (blue/gray), and Zone 3 (light gray). (E, F) Normalized strain rate of untreated (E) and GD3+-treated (F) amputated tails as a function of time. Curves represent strain rate fluctuations in the defined spatial zones. Results are expressed as mean ± SEM obtained as an average of n = 9 and n = 10 larvae, respectively, over three independent experiments. (G–I) Average velocity of cells located in defined spatial zones in untreated (circle, n = 9) and GD3+-treated (triangle, n = 10) amputated larvae from 0 to 30 mpa.

Movie S4

Strain rate after amputation. Strain rate maps of untreated (left) and Gd3+ treated (right) tail fins over 30min post-amputation; time-frame = 1min. Color code gives the magnitude of simple strain rate in s−1. Blue represents negative strain rate (contraction), red represents positive strain rate (extension) and green represents no strain.

Our data suggest that cell extrusion is related to an increase in cell density at the edge of the wound caused by collective movement of cells, or collective cell migration (Friedl and Gilmour, 2009), toward the injury site. Cell movement involves tight regulation of F-actin-binding and nonmuscle myosin II activity to control focal adhesion formation and protrusion necessary for migration (Watanabe ; Chen ). Consequently, we next examined F-actin filaments and nonmuscle myosin II in epithelial cells within regions of the injured tissue with predicted differential velocity and strain. Within 5 mpa, phosphorylated-myosin light chain II was localized to several rows of cells within <100 μm of the amputation site (Zone 1) and was significantly enriched on the edges of cells facing the wound site (Figure 4A and Supplemental Figure S3, E and F). This expression correlated with the appearance of filopodia protrusions from epithelial cells also located within <100 μm of the amputation site (Figure 4C and Supplemental Figure S3G). No change in myosin localization or filopodia extension was detected in areas of the tissue several cell rows back from the wound site (i.e., Zones 2 and 3). These data suggest a collective movement of epithelia cells toward the wound site where leader cells at the wound edge drive migration of follower cells (Friedl and Gilmour, 2009). By 30 mpa, we observed cytoplasmic myosin localization largely restricted to epithelial cells at the wound site undergoing cytoskeletal reorganization during extrusion (Figure 4A). Interestingly, the polarized localization of phosphorylated-myosin was expanded by several cell lengths and persisted for a greater duration in Gd3+-treated cells that could no longer perceive crowding-induced mechanotransductive cues (Figure 4B). We also observed increased numbers of filopodia and lamellipodia protrusions from epithelial cells in Gd3+-treated larvae (Figure 4D and Supplemental Figure S3G). These data suggest that myosin-mediated cytoskeletal reorganization (in Zone 1) drives collective movement of cells to facilitate wound closure and contributes to the observed rapid increases in cell number and crowding-induced extrusion at the injury site. Further, these data support the idea that collective cell migration during wound closure is comprised of leader cells at the front that drive migration and follower cells that are dragged along (Trepat ; Haeger ).

FIGURE 4:

Changes in phospho-myosin and F-actin subcellular localization during collective cell movement after injury. (A, B) Maximum intensity projection images of fixed untreated (A) and GD3+-treated (B) larvae stained for F-actin and p-myosin at homeostasis and 0, 5, and 30 mpa. Scale bar, 50 μm. (C, D) Maximum intensity projection still images from time-lapse confocal movies of untreated (C) and GD3+-treated (D) Et(Gal4-VP16)) larvae at homeostasis and at 10 mpa. Zones 1 and 2 are composed of leader and follower cells, respectively. Arrowheads denote observable filopodial extensions. Scale bar, 5 µm.

An increase in tensile forces across intercellular adhesions can result from this “tug of war” between leader and follower cells during collective cell migration (Mayor and Etienne-Manneville, 2016). To examine the pulling forces exerted on follower cells and the preceding cells, we used CellFIT, the Cellular Force Inference Toolkit, which assumes that cell shapes are determined by interfacial tensions at cell edges (Brodland ). We found that edge tension was significantly increased in cells located more anterior (>100 μm away) when compared with those closer to the amputation site (Supplemental Figure S4, A–C). We then used UV laser ablation to quantify relative tension levels on cells located greater than or less than 100 μm away from the wound site. The retraction velocity after severing the cell–cell junction between follower cells (Zone 3, or 133–200 μm from the wound site) was 67.6% greater than that of leader cells (located in Zone 1, 0–66 μm from the wound site) (Supplemental Figure S4, D–H). The observation of elevated junctional tension associated with cells being extended while bringing up the rear of a collective movement of cells is in line with our elevated strain fields and velocity measurements (Figure 3, A, B, and E, and Supplemental Figure S3, A and B). These results support a model where collective migration during wound closure promotes SAC-regulated crowding-induced extrusion events at the injury site, and pulling forces exerted by leader cell migration fosters long-range communication via mechanotransduction in follower cells several diameters away.

Tightly controlled proliferation of progenitor cells upon or after amputation is a hallmark of epimorphic regeneration (Kawakami ; Rojas-Munoz ). To determine how mechanosensation of cellular crowding could impact the proliferative response required for successful regeneration, we analyzed the number of cells actively synthesizing DNA, as measured by BrdU incorporation, at different locations within the tissue 3 h after injury. Amputated larvae showed an increase in the number of proliferating cells compared with homeostatic controls that were located >100 μm back from the wound edge, areas of predicted elevated tension due to stretching from the preceding cells (Figure 5, A–E). This finding is in agreement with the notion that cell stretching induces proliferation (Gudipaty ). Conversely, inhibiting the cells’ capacity to sense mechanical forces by altering SAC activity by Gd3+ treatment led to increased proliferation in areas of increased cell density, closer to the wound site (Figure 5, D–F). These data suggest that the ability to detect and respond to crowding-induced mechanical forces at the wound site is required for both the elimination of nonapoptotic cells by extrusion and the spatial and temporal control of proliferation to replace the cell types lost to injury.

FIGURE 5:

Proliferation in crowded areas after inhibition of SACs during regeneration. (A–D) Maximum intensity projection images of BrdU incorporation (green dividing cells) during homeostasis (A, B) and at 3 h after amputation (C, D) in fixed Gd3+-treated (B, D) and untreated (A, C) DAPI-stained (blue nuclei) larvae. Scale bar, 50 µm. (E) Quantification of the average number of BrdU+ cells within Zones 1 (black), 2 (blue), and 3 (gray) during homeostasis and at 3 h after amputation in untreated (n = 37 and n = 52) and GD3+-treated (n = 33 and n = 53) larvae. P values were calculated using a two-way ANOVA with Tukey’s multiple comparisons test (*** p < 0.001, **** p < 0.0001). (F) Cell density (cells/µm2) near the proliferating cells in untreated (n = 97) and Gd3+-treated (n = 173) amputated larvae. P value was determined using two-tailed, unpaired Student’s t test with Welch’s correction (****p < 0.0001).

In conclusion, our high-resolution in vivo imaging-based approach has yielded a detailed map of cellular events that generate and result from shifts in mechanical forces within a living epithelial tissue after injury. Our analyses revealed that wound-induced collective cell movements lead to an increase in cell numbers at the injury site and removal of excess cells by extrusion. These results support the idea that wound-induced collective migration leads to spatially graded, and rapidly shifting, cytoskeleton-­dependent changes in tension (Trepat ; Gayrard ). The collective cell migration and crowding-induced epithelial cell extrusions promoting wound closure and repair in the larval zebrafish fin after injury are similar to that seen after fusion of the mammalian secondary palate during development (Kim ) and after fin amputation in adult zebrafish (Chen ). The analyses presented here identify a critical threshold of crowding that promotes extrusion in living epithelia that is consistent with previous reports for crowding-induced extrusion and delamination (Shraiman, 2005; Marinari ). This is supported by the observation that perturbation of SACs causes defective nonapoptotic extrusion and cell accumulation (Figure 2, A–F) (Eisenhoffer ; Kim ), as well as aberrant proliferation in crowded areas (Figure 5, D’–F). Our data support the conclusion that the inability of cells to sense changes in mechanical forces may lead to the breakdown of both contact inhibition of locomotion (Abercrombie and Heaysman, 1954; Stramer and Mayor, 2016) and growth (Stoker and Rubin, 1967; Abercrombie, 1979; McClatchey and Yap, 2012), hallmarks of oncogenic cells. Together, our results highlight an essential role for mechanical forces in coordinating multicellular behaviors during epithelial tissue maintenance and regeneration after injury.

MATERIALS AND METHODS

Further information and requests for resources and reagents should be directed to and will be fulfilled by corresponding author George T. Eisenhoffer (gteisenhoffer@mdanderson.org).

Zebrafish and wound healing assay

Zebrafish were maintained under standard laboratory conditions with 14 h light and 10 h dark cycles. Embryos were kept in E3 embryo medium at 28.5°C and staged as described in Kimmel . For the wound closure and repair assay, 4 dpf larvae were anesthetized with 0.04% tricaine and amputated ∼100 μm (98.4 μm average, n = 285) posterior of the notochord using a scalpel. Assays were performed in the wild-type AB* background. The zebrafish used in this study were handled in accordance with the guidelines of the University of Texas MD Anderson Cancer Center Institutional Animal Care and Use Committee.

Transgenic zebrafish lines

The GAL4 enhancer trap line Et(Gal4-VP16) was used to drive expression of Tg(UAS-E1b:nsfB-mCherry) or Tg(UAS-E1b:Lifeact-EGFP). The Et(Gal4-VP16) line was also used in combination with Tg(p63:EGFP) (Eisenhoffer ). The transgenic line Tg(-8.0cldnB:lyn-EGFP) was previously described in Haas and Gilmour (2006).

The Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) transgenic line was generated by combining p5E (–3.5Ubi) (Mosimann ), pME-H2A-EGFP (no stop), and p3E-2A-mCherry-CAAX-poly A into the pDestTol2pA3 destination vector. Injections were performed with 25 pg of the purified plasmid along with 50 pg of Tol2 transposase mRNA into wild-type Tü strain developing zebrafish embryos at the one-cell stage. Potential carriers were identified by fluorescence expression and raised to adulthood.

Pharmacological treatments and UV exposure during the wound healing assay

For apoptosis inhibitor assays, embryos were treated with 10 μM apoptosis inhibitor NS3694 (Santa Cruz Biotech) for 18 h prior to amputation and during recovery in dimethyl sulfoxide (DMSO) diluted in E3. For gadolinium assays, larvae were treated with 100 μM gadolinium(III) chloride hexahydrate (Sigma Aldrich) in E3 for 3 h prior to amputation and during recovery. For GsMTx-4 assays, larvae were treated with 3 μM GsMTx-4 (Abcam) in E3 for 3 h prior to amputation and during recovery. For UV assays, larvae were placed in a Spectrolinker XL-1000 UV crosslinker (Spectronics Corporation) at 4 h before amputation and exposed to a UV-C dose of ∼420 μJ/cm2 over 2 min.

Fixation and immunofluorescence of zebrafish larvae

Zebrafish larvae were fixed overnight at 4°C with 4% paraformaldehyde in phosphate-buffered saline (PBS), 0.05% Triton X-100 (PBSTx 0.05%), washed 15 min in PBSTx 0.5%, and blocked 2 h at room temperature in block buffer (PBS, 1% DMSO, 2 mg/ml bovine serum albumin, 0.5% Triton X-100, 10% heat-inactivated goat serum). Larvae were then incubated overnight at 4°C in primary antibodies (below) in block buffer. Samples were washed for 2 h in PBSTX 0.5%, incubated in block solution for 2 h at room temperature, and then incubated overnight at 4°C in secondary antibodies (below) in block solution. Specimens were then washed 1 h in PBSTx 0.5%, incubated in 4′,6-diamidino-2-phenylindole (DAPI) (ThermoFisher; 1:1000) in PBSTx 0.5% for 30 min, and rinsed with PBS, and the tail portion of the larvae was mounted on sealed glass slides containing 80% glycerol.

BrdU incorporation and detection

For proliferation assays, larvae were soaked in 10 mM BrdU (Sigma Aldrich), E3 + 5% DMSO for 1 h, followed by 30 min recovery and fixation. Larvae were then washed 15 min in PBSTx 0.5%, 20 min in ddH2O, 45 min in 2 N HCl in ddH2O + 0.5% Triton-X, and washed 15 min in PBSTx 0.5%. Detection was performed according to the above protocol.

Primary antibodies

Primary antibodies used were rat anti-BrdU (Abcam; 1:200), rabbit anti-cleaved caspase-3 (BD Pharmingen; 1:700), and rabbit anti-phospho-Myl9 (Thr18, Ser19) (Invitrogen; 1:200).

Secondary antibodies

Secondary antibodies used were goat anti-rabbit Alexa Fluor 488 (Invitrogen; 1:200), donkey anti-rabbit Alexa Fluor 568 (Invitrogen; 1:200), goat anti-rat Alexa Fluor 488 (Invitrogen; 1:200), and Phalloidin Alexa Fluor 647 (Invitrogen; 1:50).

Image acquisition and time-lapse confocal microscopy

Confocal images were acquired as z-stacks using 20× objective (filopodial extension images were acquired using 40× water-immersion objective) on a Zeiss LSM 800 Confocal Microscope and Zen 2 software. Time-lapse images were acquired immediately after amputation by anesthetizing day 4 embryos with 0.04% tricaine in E3 and mounting in 1% low-melt agarose in a 10-mm MatTek culture dish. Images presented as maximum intensity projections made on Zen 2.

Image quantification

All quantifications were conducted using maximum intensity projections. Extruding cells were defined by presence of a multicellular actin ring and defined nuclei above the actin ring. Apoptotic cells were quantified by manually counting activated caspase-3 positive cells posterior to notochord. Cell density was calculated by manually counting DAPI-stained nuclei located posterior to the notochord and defining the area using the “Contour” tool in Zen 2. Proliferating cells in amputated embryos were quantified by manually counting BrdU-positive cells within three regions of width equal to two-thirds the wound-to-notochord distance (∼66.6 μm). Density near BrdU-positive cells was measured by defining a square of 1143 μm2 centered on the BrdU-positive cell nuclei and manually counting DAPI-stained nuclei within the defined region. Localization of anti-phospho-Myl9 was measured in cells located two cell lengths behind the wound edge by using the Contour tool in Zen 2 to trace around cell membranes and measure fluorescence intensity for the anterior half and posterior half of the cell, where polarization = anterior intensity ÷ posterior intensity. Sinuosity coefficient was determined by measuring the curvilinear distance of cell junctions in mosaic Et(Gal4-VP16);Tg(UAS-E1b:Lifeact-EGFP) embryos and dividing by the Euclidean distance between the corresponding tricellular junctions.

Image processing and cell tracking using Imaris

z-Stack rotations were done using Bitplain Imaris ×64 9.0.2 using the “Blend” display mode. Wound area measurements were acquired using the “Contour Drawing Mode” within the “Surfaces” module on Imaris. Cell migration measurements for Et(Gal4-VP16)zc1044a; Tg(UAS:LifeAct-GFP) larvae were acquired using the “Cell” module on Imaris to detect membranes using the following settings: detection type, membrane; filter type, local contrast; cell tracking algorithm, autoregressive motion; cell smallest diameter, 12.5 μm; membrane detail, 1.25 μm; intensity ≥40 μm; quality ≥0.93. Measurements for Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) larvae were acquired using the “Spots” module on Imaris to detect nuclei using the following settings: estimated diameter, 11 μm; quality, >21; tracking algorithm, autoregressive motion; maximum gap size, 3. Values for instantaneous speed were calculated for each frame within the first 30 mpa, along with values for displacement after 60 mpa, for each cell tracked. Nearest neighbor analysis was conducted by using Imaris to detect cells as above along with the “Measuring Points” module on Imaris to measure the distance between the center of the cell of interest and its neighbors at the frame where extrusion begins. The extrusion threshold was calculated by dividing the average nearest neighbor distance of control cells by the average nearest neighbor distance of extruded cells.

PIV and strain rate measurements

The velocity field for 20× time-lapse images on the Tg(Ubi:H2A-EGFP-2A-mCherry-CAAX) line was obtained using an open source MATLAB code (PIVlab) (Thielicke and Stamhuis, 2014, 2018), with four passes (128 × 128, 64 × 64, 32 × 32, and 16 × 16 pixel-size interrogation window with 50% overlap each). Strain rate was calculated using the formula E  =  (∂  +  ∂)/2, with i, j ∈ (x, y) and velocity field, u, obtained from PIV measurements.

Inference of tension using CellFIT

For tension measurements, image segmentation was performed using an automated ImageJ plug-in Tissue Analyzer. Tension maps were generated on segmented images using CellFIT, the Cellular Force Inference Toolkit (Brodland ). CellFIT formulation was applied to the entire mesh using nearest segment tangent vectors.

Scanning electron microscopy

Samples were fixed in 3% glutaraldehyde + 2% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.3). Samples were washed with 0.1 M cacodylate buffer (pH 7.3), postfixed with 1% cacodylate-buffered osmium tetroxide, and washed with 0.1 M cacodylate buffer and then in distilled water. Samples were treated with Millipore filtered 1% aqueous tannic acid, washed in distilled water, treated with Millipore filtered 1% aqueous uranyl acetate, and then rinsed with distilled water. They were dehydrated with a series of increasing concentrations of ethanol, transferred to increasing concentrations of hexamethyldisilazane, and air-dried overnight. Samples were mounted onto double-stick carbon tabs (Ted Pella, Redding, CA) and mounted onto glass microscope slides. Samples were coated under vacuum using a Balzer MED 010 evaporator (Technotrade International, Manchester, NH) with platinum alloy for a thickness of 25 nm and then flash-carbon-coated under vacuum. Samples were then transferred to a desiccator until examination and imaging in a JSM-5910 scanning electron microscope (JEOL USA, Peabody, MA) at an accelerating voltage of 5 kV.

Laser ablation assays

Images were acquired with a water-immersion Plan Apo VC 60× 1.2 NA Nikon objective using a Yokogawa CSU-X1 spinning disk that is currently integrated to an inverted Nikon Ti-microscope by 3i (Intelligent Imaging Innovations). Samples were excited with a 3i 488-nm solid-state laser line, and emission was collected with a 525/30 nm emission filter. Ablation was performed with a 3i 532-nm solid-state pulsed laser (>60 µJ pulses at 200 Hz) that is integrated to a 3i Vector, which is a diffraction-limited high-speed X,Y scanner that allows interactive examination of living specimens with a rapid single-point seek time of 0.3 ms and fast ROI scanning (1500 lines/s). The system comprises a 3i mSwitcher unit that enabled simultaneous use with Spinning Disk Confocal for this multimodal acquisition.

Photoablations were conducted by aiming a laser of a duration of 6 ms and raster block size 10 at a 0.52-μm spot at the junction of two apical epithelial cells of day 4 Tg(-8.0cldnB:lyn-EGFP) larvae (Haas and Gilmour, 2006) that were anesthetized and mounted as previously described. Time-lapse images acquired continuously as 10-slice z-stacks. Recoil was calculated on Fiji ImageJ by measuring the average displacement of all visibly intact cell junctions immediately encircling the cell ablation site ∼5.5 s after photomanipulation. Images are presented as maximum intensity projections.

Statistical analysis

All statistical analysis were conducted using GraphPad Prism 7.0 software. All results expressed as mean ± SEM, unless otherwise stated.

Click here for additional data file.