Tension regulates myosin dynamics during Drosophila embryonic wound repair

Coordinated cell movements drive embryonic development and tissue repair, and are implicated in cancer metastasis (Gaggioli et al., 2007; Hidalgo-Carcedo et al., 2011). Cell movements are often coordinated by supracellular cytoskeletal networks formed by actin and the molecular motor non-muscle myosin II (Blankenship et al., 2006; Fernandez-Gonzalez et al., 2009; Lienkamp et al., 2012; Heller et al., 2014; Rozbicki et al., 2015).

Embryonic wound closure is a conserved process driven by the collective migration of cells to rapidly seal a lesion. During embryonic wound repair, actin and myosin are polarized in the cells adjacent to the wound, accumulating at the interface with the wounded cells and forming a supracellular cable around the wound (Martin and Lewis, 1992; Brock et al., 1996; Kiehart et al., 2000; Davidson et al., 2002; Kofron et al., 2002; Wood et al., 2002; Clark et al., 2009). The actomyosin cable contracts (Fernandez-Gonzalez and Zallen, 2013), serving as a purse string that coordinates the migratory behaviours of the cells adjacent to the wound (Brock et al., 1996; Wood et al., 2002; Abreu-Blanco et al., 2012). Embryonic wound closure can last minutes or hours, with total myosin levels at the purse string remaining relatively constant (Fernandez-Gonzalez and Zallen, 2013). The dwell times of myosin motors on actin filaments in vitro are of just a few seconds (Norstrom et al., 2010; Nagy et al., 2013; Hundt et al., 2016), suggesting that myosin motors may constantly turn over during wound closure. However, the mechanisms that regulate myosin dynamics in supracellular networks are not well understood.

Physical forces can regulate myosin dynamics both in vitro and in vivo. In vitro, non-muscle myosin II motors under mechanical load are stabilized on actin filaments (Kovacs et al., 2007). In vivo, mechanical tension controls the cortical localization of myosin in mitotic Dictyostelium cells (Effler et al., 2006), and in the ectoderm (Fernandez-Gonzalez et al., 2009) and mesoderm (Pouille et al., 2009) of early Drosophila embryos. It is currently unknown whether mechanical forces affect myosin dynamics during embryonic wound repair.

We investigated myosin dynamics during the repair of laser-induced wounds in the epidermis of Drosophila embryos. Wound repair occurred in a fast phase followed by a slow phase (Fig. S1) (Abreu-Blanco et al., 2012). We conducted our measurements within the fast phase (unless otherwise indicated), when wound area was 40–60% of its maximum value.

We compared myosin dynamics in supracellular networks displaying different degrees of contractility, the purse string and actomyosin cables that segment the epidermis (Simone and DiNardo, 2010) (Movie 1). Myosin fluorescence correlates with tension (Fernandez-Gonzalez et al., 2009). In embryos expressing myosin regulatory light chain (MRLC, a subunit of myosin II motors; known as Sqh in flies) tagged with the green fluorescent protein (GFP), fluorescence at the purse string was 1.9-fold greater than in epidermal cables (P<0.005, Fig. 1C,D,F). Similar results were found for GFP-tagged myosin heavy chain (MHC, another myosin subunit; known as Zip in flies; Fig. S2A,B,D). We used ultraviolet laser ablation to quantify relative tension levels (Hutson et al., 2003; Zulueta-Coarasa and Fernandez-Gonzalez, 2015) (Fig. 1A,B). The retraction velocity after severing the purse string was 40% greater than that of epidermal cables (P<0.02, Fig. 1E, inset). Thus, assuming constant viscoelastic properties, our results indicate that tension is greater in wound-associated purse strings than in epidermal cables.

To investigate whether myosin turnover at the purse string is affected by contractile force, we used fluorescence recovery after photobleaching (FRAP) of MRLC–GFP (Fig. 1C,D). At the purse string, MRLC turned over with a characteristic half-time, t_1/2, of 37.5±3.0 s (mean±s.e.m.; Fig. 1E,G,H). In epidermal cables, MRLC displayed a t_1/2 of 12.2±2.6 s, which was significantly shorter than at the purse string (P<0.001, Fig. 1G). We found that MHC:GFP was also stabilized at the wound margin (Fig. S2C,E,F). Taken together, our data indicate that myosin motors are stabilized at the wound edge during embryonic wound repair.

To understand how myosin turns over at the wound edge, we developed a mathematical model to extract the rates of myosin assembly (k_on) and disassembly (k_off) in the photobleached regions (Pines et al., 2012). Using the model, we estimated the assembly rate constant at the purse string to be 0.012±0.003 s⁻¹, significantly lower than the 0.050±0.015 s⁻¹ in epidermal cables (P<0.01, Fig. 1I). The disassembly rate constant was also significantly lower at the purse string (0.011±0.001 s⁻¹) than in epidermal cables (0.047±0.009 s⁻¹, P<0.001, Fig. 1J). Taken together, our data indicate that the turnover of myosin is reduced at the wound edge, suggesting a role for tension in regulating myosin dynamics during wound closure.

To determine whether physical forces affect myosin dynamics around the wound, we used laser ablation to sever the purse string, thereby partially releasing the tension that it sustained. We cut the purse string 10 min after wounding in embryos expressing MRLC–GFP, and we monitored fluorescence immediately after cutting in regions adjacent to the severed area (Fig. 2A), in regions that were far away (at least 30 μm, Fig. 2B) from the severed area, and in sham-irradiated controls (Fig. 2C). We measured a significant loss of 19.8±6.3% (mean±s.e.m.) of MRLC–GFP fluorescence in regions of the purse string adjacent to the severed area (15 out of 18 regions, P<0.02, Fig. 2D,D′). In contrast, MRLC–GFP fluorescence did not change in segments away from the ablation site or in controls (Fig. 2D,D′). These data indicate that tension is, at least in part, necessary to maintain myosin at the purse string around embryonic wounds.

To determine how loss of tension affects myosin dynamics at the purse string, we used FRAP to measure myosin turnover at ∼1 min after cutting the purse string in segments adjacent to or far from ablation sites, and in sham-irradiated controls (Fig. 2E,G; Movie 2). Myosin turnover was significantly faster close to the site of ablation than in controls (t_1/2 of 19.5±2.7 s vs 30.2±2.5 s, respectively, P<0.01, Fig. 2H). Our results indicate that tension is necessary to stabilize myosin and maintain it at the wound edge.

To investigate whether tension is sufficient to stabilize myosin, we sought scenarios in which tension at the purse string increased. Using laser ablation, we found that the retraction velocity after severing the cable was 31.7% greater in the slow phase of wound closure (25 min after wounding, <10% of the maximum wound area) than in the fast phase (5 min after wounding, 40–60% of the maximum wound area, P<0.01, Fig. 3A,B,E, inset). Thus, tension at the purse string increased as the wound closed.

To determine the effect of increased tension on myosin dynamics, we compared MRLC–GFP turnover during the fast and slow phases of wound repair (Fig. 3C,D). Each purse string was photobleached during both the fast and slow phases. The t_1/2 of MRLC–GFP at the wound margin during the slow phase was 50.0±3.4 s, significantly longer than the 34.5±5.5 s in the fast phase (mean±s.e.m.; P<0.02, Fig. 3E–G). Using our mathematical model, we found that the k_on did not display significant changes (Fig. 3H), but the rate of myosin disassembly from the purse string, k_off, showed a significant decrease of 2-fold in the slow phase compared to the fast phase (P<0.02, Fig. 3I). Sequential photobleaching of epidermal cables did not cause any differences in the turnover of MRLC–GFP (Fig. S3). Taken together, our data show that increased tension is sufficient to stabilize myosin at the purse string, possibly through a decreased rate of myosin disassembly.

Phosphorylation of the MRLC at a serine and a threonine residue is necessary for myosin motor activity (Sellers et al., 1985; Jordan and Karess, 1997; Komatsu et al., 2000), minifilament formation (Scholey et al., 1980), and, to a lesser extent, efficient actin binding (Sellers et al., 1982). We investigated the role of phosphorylation on the stabilization of myosin at the wound border and tissue repair. We used embryos expressing GFP-tagged phosphomimetic (Sqh-AE, threonine-20 is mutated into an alanine and serine-21 is mutated into a glutamic acid), unphosphorylatable (Sqh-AA) or wild-type (Sqh-TS) forms of the MRLC (Vasquez et al., 2014) (Fig. 4A–E). Recent data indicate that Sqh-AE may not fully mimic the phosphorylated state of myosin, but rather displays partially defective motor activity (Vasquez et al, 2016). Endogenous myosin activity was depleted by injection of the Rho kinase (Rok) inhibitor Y-27632 (Fernandez-Gonzalez et al., 2009) in embryos expressing the phosphorylation mutant MRLC variants. Y-27632 treatment at 8–10 min after wounding led to a 46.0±3.2% loss of wild-type myosin from the wound edge within 1 min, a greater percentage than the 25.8±4.3% loss of Sqh-AA, and the 8.5±4.7% loss of Sqh-AE (mean±s.e.m.; P<0.01; Fig. S4, Movie 3). Our data suggest that cortical localization of phosphorylation mutant myosin only partially requires Rok activity.

Stabilization at the wound edge was disrupted for Sqh-AA and Sqh-AE (Fig. 4F). Upon endogenous myosin deactivation, both Sqh-AA (t_1/2 of 15.7±3.1 s) and Sqh-AE (21.2±3.2 s) turned over faster than wild-type myosin (36.7±3.7 s; P<0.01 and P<0.05, respectively; Fig. 4F′). Sqh-AE was stabilized (51.8±4.2 s) in the presence of endogenous myosin (P<0.01). In contrast, the turnover of Sqh-AA, which cannot form minifilaments with other motors (Scholey et al., 1980), was not affected by the presence of endogenous myosin (18.1±3.2 s). Our data suggest that motor activity and the ability to form minifilaments are crucial for myosin II motors to be stabilized in the purse string during embryonic wound repair.

To establish the role of myosin stabilization during embryonic repair, we measured wound closure dynamics in embryos expressing wild-type or phosphorylation mutant myosin (Fig. 4I–N). In embryos in which endogenous myosin was active, wounds closed independently of the expression of phosphorylation mutant myosin (Fig. 4I–K,N). In contrast, in embryos expressing the phosphorylation mutant constructs and where endogenous myosin activity was inhibited by injection of Y-27632 after the purse string had formed and wound closure had begun, wounds did not close, but rather expanded again (Fig. 4L–N). Taken together, these results suggest that motor activity facilitates myosin stabilization at the wound edge and promotes rapid wound closure.

Our results indicate that myosin is mechanically regulated during the collective cell movements that drive embryonic wound repair. We find that dynamic motor activity is crucial for the stabilization of myosin at the cortex, consistent with recent work showing that myosin phosphorylation regulates contractility in epithelial morphogenesis (Kasza et al., 2014; Vasquez et al., 2014; Vasquez et al., 2016). Several actin regulators are sensitive to load, including formins (Courtemanche et al., 2013; Higashida et al., 2013; Jégou et al., 2013), and the filament-severing protein cofilin (Hayakawa et al., 2011; Tojkander et al., 2015). The mechanical dependence of myosin dynamics could thus be a result of direct effects on the motor, indirect effects on the actin filaments bound by the motor, or a combination of both. Changes to the physical properties at the wound edge may also locally affect the dynamics of myosin and actin during wound repair. Recent advances in genome editing and the development of new methods to probe tissue mechanics in living animals (Campàs et al., 2014; Bambardekar et al., 2015) will be of enormous value to investigate the interplay between cell mechanics and myosin dynamics during collective cell migration in vivo.

Live imaging was conducted using sqh-sqh:GFP (sqh encodes MRLC, and here sqh is expressed under its own promoter) (Royou et al., 2004), sqh-sqh^TS:GFP, sqh-sqh^AE:GFP, and sqh-sqh^AA:GFP (Vasquez et al., 2014), zip-zip:GFP (zip encodes for MHC, here expressed under its own promoter) (Buszczak et al., 2007) and endo-DE-cadherin:mTomato (Huang et al., 2009).

Embryos at stage 14–15 of development were dechorionated in 50% bleach for 2 min and rinsed with water. Embryos were aligned on an apple juice–agar pad, glued ventrolateral-side-down onto a coverslip using heptane glue, and covered with a 1:1 mix of halocarbon oil 27 and halocarbon oil 700 (Sigma-Aldrich). Embryos were imaged using a Revolution XD spinning disk confocal (Andor) with a 60× oil-immersion lens (NA 1.35; Olympus). Images were captured with an iXon Ultra 897 camera (Andor) and Metamorph (Molecular Devices) as the image acquisition software. 16-bit Z-stacks were acquired at 0.2–0.3 μm steps every 4–30 s (7–16 slices per stack). Maximum intensity projections were used for analysis. In Fig. 4A–E, the contrast of the images was adjusted independently for each treatment. In all other figures, the same linear contrast adjustment was applied to all the images in each experiment.

Embryos were dechorionated and aligned as above, dehydrated for 10–15 min and covered in a 1:1 mix of halocarbon oil 27 and halocarbon oil 700. Embryos were injected with water or 100 mM Y-27632 (Tocris Biosciences) in water using a Transferman NK2 micromanipulator (Eppendorf) and a PV820 microinjector (WPI) coupled to our spinning disk confocal.

Wounds were created using a pulsed Micropoint N₂ laser (Andor) tuned to 365 nm. The laser produced 120 μJ pulses with a duration of 2–6 ns. To wound embryos, ten laser pulses were delivered at each of seven spots over a 14 μm line. To reduce or measure tension, ten laser pulses were delivered at a single spot, and samples were imaged immediately before and 1.72 s after spot ablation. In sham-irradiated controls, we completely attenuated the laser using a neutral density filter.

To track repair dynamics, wounds were delineated by using Livewire, a semi-automated optimal path search method for image segmentation integrated into SIESTA, an image analysis platform that we develop (Fernandez-Gonzalez and Zallen, 2011; Leung and Fernandez-Gonzalez, 2015). The wound closure rate was the area difference between the time after injection and the last time point of each movie, divided by the elapsed time. To measure recoil velocity after laser ablation, the positions of the tricellular vertices connected by the ablated cytoskeletal networks were manually tracked in SIESTA, and the velocity of retraction was calculated.

We compared sample means using a non-parametric Mann–Whitney test (Glantz, 2002). For comparisons of more than two groups, we used a Kruskal–Wallis test to reject the null hypothesis, and Dunn's test for pairwise comparisons. Sequential FRAP experiments were contrasted using Wilcoxon signed-rank test. For time series, error bars indicate s.e.m. For box plots, error bars show standard deviation, the box indicates the s.e.m. and magenta lines denote the mean. *P<0.05; **P<0.01; ***P<0.001.

We are grateful to Adam Martin for reagents; to Sérgio Simoes, Tony Harris, Jon Rocheleau, Will Ryu, and Jen Zallen for useful discussions; and to Miranda Hunter, Jessica Yu and Michael Wang for comments on the manuscript. Flybase provided important information for this work.