|Home | About | Journals | Submit | Contact Us | Français|
Axis elongation in Drosophila occurs through polarized cell rearrangements driven by actomyosin contractility. Myosin II promotes neighbor exchange through the contraction of single cell boundaries, while the contraction of myosin II structures spanning multiple pairs of cells leads to rosette formation. Here we show that actomyosin cables form at a higher frequency than expected by chance during elongation, indicating that cable assembly is an active process. Multicellular cables are sites of increased mechanical tension as measured by laser ablation. Fluorescence recovery after photobleaching experiments show that myosin II is stabilized at the cortex in regions of increased tension. Myosin II is recruited in response to an ectopic force and relieving tension leads to a rapid loss of myosin, indicating that tension is necessary and sufficient for cortical myosin localization. These results demonstrate that myosin II dynamics are regulated by tension in a positive feedback loop that leads to actomyosin cable formation and efficient tissue elongation.
Morphogenesis involves a combination of biochemical signaling pathways and the translation of these signals into the forces that move cells. Mechanical force is increasingly appreciated as an input that can regulate cell behavior (Gorfinkiel et al., 2009; Solon et al., 2009). Forces acting between cells and tissues regulate gene expression, cell division, and tumor cell progression (Orr et al., 2006; Wozniak and Chen, 2009). A primary source of force generation is actin-based contractility mediated by the myosin II motor protein. Localized actomyosin contractility is required for cell division and migration (Robinson and Spudich, 2004) and patterned myosin II activation at the tissue level can lead to structural transformations such as grooves, tubes, and placodes in multicellular epithelia (Dawes-Hoang et al., 2005; Escudero et al., 2007; Nishimura et al., 2007).
In the Drosophila embryo, polarized actomyosin contractility in the plane of the tissue drives the cell rearrangements that elongate the body axis (Irvine and Wieschaus, 1994; Zallen and Wieschaus, 2004; Butler et al., 2009). Actomyosin contractility promotes local neighbor exchange through the contraction of single cell boundaries (Bertet et al., 2004), and the contraction of actomyosin structures spanning multiple pairs of cells leads to the formation of rosettes (Blankenship et al., 2006). Actomyosin organization into multicellular cable-like structures has been implicated in epithelial advance (Franke et al. 2005), compartment boundary formation (Major and Irvine, 2006), and wound healing (Wood et al., 2002), and can occur in response to a variety of signaling pathways (Escudero et al., 2007; Nishimura et al., 2007). However, the cell biological mechanisms that regulate actomyosin cable formation are not well understood.
Here we show that actomyosin cables in intercalating cells of the Drosophila embryo form at a higher frequency than expected by chance and multicellular contractile structures sustain increased tension as measured by laser ablation. Cell boundaries under increased tension have higher levels of myosin and a lower rate of myosin dissociation from the cortex. An ectopic force is sufficient to recruit myosin and, conversely, relieving tension leads to loss of myosin from the cortex. These results indicate that myosin II localization is stabilized by tension in a mechanism that leads to higher-order actomyosin cable formation and efficient tissue elongation.
Myosin II directs polarized cell rearrangements in the Drosophila embryo through the contraction of specific boundaries between cells (Figure 1A–D) (Bertet et al., 2004; Zallen and Wieschaus, 2004). Time-lapse imaging showed that adjacent myosin-positive edges often contract simultaneously (Figure 1D; Supplemental Movie 1). Contraction of these structures leads to multicellular rosettes that form predominantly during the period of rapid intercalation in stage 8 (Figure 1E, blue line) (Blankenship et al., 2006). This may explain why rosettes were overlooked in studies at stage 7, when the germband has only reached one-third of its final length (Bertet et al., 2004; Rauzi et al., 2008).
Multicellular contractile structures presumably represent intracellular myosin filaments connected across cell boundaries through adherens junctions. These structures are described here as actomyosin cables. Actomyosin cables appear to represent sites of increased tension, as cell boundaries in cables were aligned perpendicular to the anterior-posterior (AP) axis (Figure 1D). We quantified the extent of alignment in living embryos using a computer algorithm to identify cell boundaries in confocal images (Experimental Procedures). Alignment was measured as the fraction of AP edges (cell-cell interfaces oriented at 75–105° relative to the AP axis) that were directly connected to at least one other AP edge. Wild-type embryos displayed a sharp increase in alignment at the onset of intercalation that plateaued at ~40% of all AP edges (Figure 1E). Embryos zygotically mutant for eve or maternally mutant for bcd, nos and tsl are defective for elongation and showed significantly reduced alignment (Figure 1F), while alignment occurred normally in twist snail mutants that fail to generate mesoderm but are able to elongate (Figure 1G; 45±3% of edges were aligned in twist snail at stage 8, n=3 embryos, compared to 40±2% in wild type). These results indicate that the timing of alignment correlates with intercalary behavior and does not require external forces from the ventral furrow.
The spatial distribution of myosin II was analyzed in time-lapse movies of living embryos that express fluorescently-tagged myosin regulatory light chain from its endogenous promoter (Myo:mCherry, Martin et al., 2008). Myo:mCherry localized predominantly to interfaces between anterior and posterior cells (Figure 2A,B). One-third of all AP edges were myosin-positive (39±2%, n=2609 AP edges in 5 embryos), and nearly two-thirds of the myosin-positive edges were associated with multicellular cables (62±2%) (Figure 2C–E). Analysis of single edge behaviors over time revealed that alignment often precedes asymmetric myosin redistribution (Supplemental Figure 1A,B).
The organization of myosin into multicellular cables could arise randomly from independent and unrelated single-cell events. Alternatively, cables could form through a mechanism that actively promotes the accumulation of myosin in adjacent cells. To distinguish between these possibilities, we used statistical methods to calculate the probability that myosin-positive cell edges would be connected by chance. Monte Carlo methods (Fernandez-Gonzalez et al., 2005) were used to simulate random distributions of myosin in the endogenous cellular sheet using the same number of edges that were myosin-positive in vivo. Simulations were repeated 20,000 times for each image (25 images in 5 embryos, 5 time points/embryo). To recapitulate myosin planar polarity, myosin was assigned to edges oriented at 60–120° relative to the AP axis, a restriction that includes a majority of myosin-positive edges in vivo (Supplemental Figure 2A–C).
In contrast to the association of myosin-positive edges with cables in random simulations (44±0.02%), the distribution of myosin II was significantly more clustered in vivo (Figure 2C–E) (62±2%, P=2.0×10−8). Similar results were obtained at a range of threshold values for myosin II (Supplemental Figure 2D–F). These results demonstrate that myosin II is preferentially associated with multicellular contractile structures in intercalating cells.
The nonrandom organization of myosin II into multicellular structures suggests that actomyosin cable formation is actively regulated. To ask if actomyosin cables display distinct mechanical properties, we performed laser ablation of individual boundaries between cells. A 365 nm UV laser was focused on an ~1 μm diameter junctional region labeled with E-cadherin:GFP. This is predicted to sever the plasma membrane and cortical cytoskeleton on both sides of the cell-cell interface (Farhadifar et al., 2007).
Laser ablation induced a local relaxation of cortical tension and an increase in the distance between the vertices attached to the ablated edge (Figure 3; Supplemental Movies 2–4). AP edges displayed a greater response to ablation than DV edges (Figure 3A–C,G; Supplemental Movies 2–4), consistent with previous studies (Rauzi et al., 2008) and with the planar polarized distribution of myosin II (Zallen and Wieschaus, 2004). The response to ablation was abolished by injecting the Rho-kinase inhibitor Y-27632 (Figure 3D,E,H), a drug that blocks myosin cortical localization (Supplemental Figure 1C,D) and arrests germband extension (Bertet et al., 2004). These results indicate that myosin contractility, and not membrane integrity per se, is required for net axial forces between cells in the germband. Conversely, increasing myosin activity by injecting Calyculin A, a serine/threonine phosphatase inhibitor that acts on myosin phosphatase (Ishihara et al., 1989) and leads to increased myosin at the cortex (Supplemental Figure 1E), converted the mechanical properties of DV edges into those of AP edges (Figure 3F,H). These results indicate that differences in myosin II activity are responsible for the differences in tension at AP and DV cell boundaries.
Linked edges associated with multicellular cables displayed a significantly stronger response to ablation than isolated contractile edges (Figure 3A,B; Supplemental Movies 2,3). These differences were apparent in terms of the distance retracted by the vertices of the ablated edge (Figure 3G, P=7.0×10−4) and the peak retraction velocities (Figure 3H, P=0.002). The retraction distance after ablation correlated with the length and amount of myosin in the cable rather than the properties of the ablated edge (Figure 3I, Supplemental Figure 3A–F). These results indicate that mechanical properties are influenced by higher-order cellular organization.
Initial retraction velocities after ablation are considered to be proportional to the endogenous forces at cell boundaries (Hutson et al., 2003; Peralta et al., 2007). By estimating the initial velocity as the peak velocity after ablation, the tension at linked AP edges was greater than the tension at isolated AP and DV edges in a ratio of Tlinked AP:Tisolated AP:TDV=3.1:1.7:1 (Figure 3H). In an independent method, we modeled cell boundaries as mechanical equivalent circuits consisting of a spring and dashpot in parallel (Figure 3J, Supplemental Figure 3G–L). This approach can be used to estimate the relaxation time (τ), a measure of local viscoelastic properties, and the asymptotic distance retracted by the vertices attached to the cut edge (D), which is proportional to the tension on the edge when the local viscoelastic properties are homogeneous. Values for τ are constant throughout the tissue (Figure 3J, Supplemental Figure 3K,L). Therefore, on the spatiotemporal scale of vertex recoil, the local viscoelastic properties in the Drosophila germband are homogeneous. Values for D indicate that the tension at linked AP edges is greater than the tension at isolated AP and DV edges in a ratio of Tlinked AP:Tisolated AP:TDV=2.8:1.7:1 (Figure 3J), consistent with the peak retraction velocity measurements. These results demonstrate that cortical tension is significantly higher in actomyosin cables than in isolated contractile edges.
These findings contrast with a previous report using an infrared laser (Rauzi et al., 2008). Low-power infrared ablation induces a lesion that is smaller (<0.2 μm) than the lesion induced by UV ablation in this study (1 μm), and generates retraction velocities that are several times slower (0.11 μm/s vs. 0.5 μm/s reported here), suggesting that the infrared ablation experiments only partially disrupt the actomyosin network.
Cell intercalation occurs in an epithelial sheet in which forces within cells can trigger mechanotransduction pathways in neighboring cells. If mechanical tension acts as a signal to promote myosin II localization, then myosin levels should be higher at cell boundaries in cables. Consistent with this prediction, we found that linked AP edges had higher levels of myosin II regulatory light chain than isolated AP edges, regardless of stage (Figure 4A) or edge length (Figure 4B). These results demonstrate that cortical myosin levels are increased in multicellular cables.
The accumulation of myosin in cables could occur through an increase in the rate of myosin association or, alternatively, a decrease in the rate of myosin dissociation. To distinguish between these possibilities, we photobleached a 1.0 × 1.4 μm region at the cortex of cells expressing the myosin regulatory light chain fused to GFP (Myo:GFP, Supplemental Movie 5). Myo:GFP levels immediately after bleaching were similar in all cases (Figure 4C), indicating that photobleaching was complete. Myo:GFP recovered to more than half of pre-bleach fluorescence levels within 30 s (Figure 4D). The rate of recovery was similar for isolated and linked AP edges (Supplemental Figure 4E), indicating that tension does not affect the rate of myosin recruitment to the cortex. By contrast, the mobile fraction was significantly lower for linked AP edges in cables than for isolated AP edges (Figure 4D, P=0.03). The mobile fraction decreased dramatically from >75% at edges with low levels of myosin to <40% at edges with high levels of myosin (Figure 4E, P=2.0×10−4). These results suggest that myosin II dissociation from the cortex is inhibited in cortical domains with increased contractile activity.
These results indicate that increased tension correlates with decreased myosin II dissociation from the cortex, resulting in the stabilization of cortical myosin II. Inhibition of myosin dissociation could occur through a reduction in lateral diffusion along the membrane or decreased exchange of protein with the cytoplasm. The extent of lateral diffusion can be measured by quantifying the shape of myosin intensity profiles during fluorescence recovery after photobleaching (Figure 4F). If lateral diffusion contributes to the recovery of fluorescence, then the width of the bleached region should change over time (de Beco et al. 2009). We approximated the width of the bleached region as the standard deviation (σ) of a Gaussian curve fit to the observed intensity profile (Supplemental Figure 4A–D). The width of the bleached region decreased significantly during the recovery of fluorescence at isolated AP edges (P=0.0057), but was constant for recovery at linked AP edges (Figure 4F, P=0.65). The stabilization of myosin cortical localization by tension therefore occurs in part through an inhibition of lateral diffusion.
These results demonstrate that myosin dynamics are not uniform in intercalating populations but instead correlate with local differences in mechanical tension. To ask if tension is necessary for myosin cortical localization, we performed ablations in embryos expressing Myo:GFP. Single-edge ablations led to a significant loss of myosin II in the unablated cortical regions of the affected cells (Supplemental Figure 4F). To relieve tension at a greater distance from the ablation site, we ablated a ~50 μm line parallel to the AP axis. Line ablations led to a significant decrease in myosin intensity in the intact edges of ablated cables, while myosin intensity outside the ablated region was unaffected (Figure 4G, n=27 ablations, P=0.0018). These results demonstrate that tension is necessary for cortical myosin II localization in intercalating cells.
Conversely, we asked if tension is sufficient for cortical myosin localization using a micropipette aspiration approach to introduce ectopic forces in intercalating cells. Myo:GFP was rapidly recruited to the apical surface in response to microaspiration (Figure 4H–J) (13/17 experiments). By contrast, two control GFP markers were not affected by microaspiration (Figure 4H; 0/10 Resille:GFP; 0/5 Spider:GFP). The pressure applied in microaspiration experiments was estimated to be 0.15 nN/μm2 – 0.4 nN/μm2, corresponding to applied forces of 1–3 nN. These values are similar to the forces required to recruit myosin in Dictyostelium (8–15 nN, Effler et al., 2006) and lower than the forces required to induce Twist transcription in Drosophila (60±20 nN, Desprat et al., 2008). Together, these results indicate that mechanical tension is necessary and sufficient for myosin cortical localization in intercalating cells.
A major challenge in developmental biology is to understand how cell behavior and cytoskeletal activity are coordinated to produce higher-order tissue organization. Here we provide evidence that myosin II is organized into multicellular contractile structures that form nonrandomly in intercalating cells and sustain increased mechanical tension. Mechanical tension is sufficient to promote cortical myosin localization and, conversely, relieving tension leads to a rapid decrease in cortical myosin. These studies demonstrate that myosin II not only generates tension, but myosin II dynamics can also be regulated by tension, generating a positive feedback loop that allows cells to dynamically respond to changes in their mechanical environment.
External forces have been shown to recruit myosin to the cortex during cell division and apical constriction (Effler et al., 2006; Pouille et al., 2009). In intercalating cells, myosin is distributed in a planar polarized fashion in response to striped patterns of gene expression that concentrate contractile proteins in specific cortical domains (Zallen and Wieschaus, 2004). We propose that the recruitment of myosin by tension amplifies these initial subtle asymmetries, reinforcing contractile activity and increasing the number of cells engaged in contractile behavior. This positive feedback loop could explain the formation of multicellular actomyosin cables that promote rosette formation and efficient elongation in intercalating populations (Blankenship et al., 2006).
How is mechanical tension translated into myosin II stabilization at the cortex? Evidence from the literature suggests three models of mechanotransduction (Vogel and Sheetz, 2006). First, forces have been shown to influence gene expression in normal cells as well as during tumorigenesis (Orr et al., 2006; Wozniak and Chen, 2009), suggesting that tension could lead to changes in the expression of myosin II regulatory proteins. Tension has been shown to promote β-catenin-dependent expression of Twist (Farge, 2003; Desprat et al., 2008), a transcription factor that regulates apical myosin localization (Dawes-Hoang et al., 2005). However, the rapid recruitment of myosin in response to ectopic forces in intercalating cells suggests that the effect of tension is likely to be independent of transcription. A second possibility is that increased tension at the plasma membrane could alter the trafficking of secreted signaling proteins. Such a mechanism has been proposed to occur during Drosophila mesoderm invagination in which mechanical tension potentiates the activity of the secreted Twist target gene, Fog (Pouille et al., 2009). However, Twist and Fog are not expressed or required in intercalating cells, suggesting that myosin localization during intercalation occurs through a different mechanism. Finally, mechanical tension can alter signaling pathways directly through force-dependent changes in protein interactions (Sawada et al., 2006; del Rio et al., 2009). Myosin itself could act as the mechanosensor in this context, as tension favors the ADP-bound form of myosin II in vitro, stabilizing its association with actin (Cremo and Geeves, 1998; Veigel et al., 2003; Kovács et al., 2007). Mechanical tension alters the activity of several myosin and kinesin motors and may represent a general mechanism regulating motor protein function (Spudich, 2006; Kee and Robinson, 2008).
Multicellular actomyosin cables are characteristic of many developmental processes including epithelial closure (Franke et al., 2005; Solon et al., 2009), tracheal tube invagination (Nishimura et al., 2007) and neural plate bending and elongation (Nishimura and Takeichi, 2008). The role of mechanical tension in regulating myosin dynamics could serve to promote contractile activity and maintain the integrity of contractile cables in the presence of interruptions caused by cell shape changes, cell division, or cell death. In the Drosophila embryo, spatially regulated mechanical forces may also act as a long-range signal to allow cells to maintain planar polarity despite the transient nature of local cell interactions during morphogenesis.
Embryos were dechorionated in 50% bleach for 2 min, washed with water, and mounted in halocarbon oil 27 (Sigma) between a cover slip and an oxygen-permeable membrane (YSI). The anterior ventrolateral region of the germband was imaged with an Ultraview RS5 spinning disk confocal (Perkin Elmer) controlled by Metamorph software (Universal Imaging) using a 40X oil-immersion objective (NA 1.3, Zeiss).
We developed a computer algorithm in Matlab (Mathworks)/DIPImage (TU Delft) for automated identification of cell outlines in confocal images. An independent algorithm was used to track cell behavior in Figure 1E (Ori Weitz and JAZ, unpublished results). Myosin intensity (Iedge) was measured as the average Myo:mCherry intensity within a 3-pixel (1 μm) wide mask corresponding to a single interface. To account for variations in image intensity, myosin intensity (I) for each edge was calculated as:
where IDV was the average intensity of the ten closest DV edges and Ibackground was the average pixel value outside the embryo.
Dechorionated embryos were mounted in halocarbon oil 700 (Sigma) and imaged with an Ultraview RS5 spinning disk confocal (Perkin Elmer). An N2 Micropoint laser (Photonics Instruments) tuned to 365 nm was used to ablate cell interfaces labeled with E-cadherin:GFP or Myo:GFP. Imaging was performed before and after ablation using a 63X oil immersion lens (NA 1.4, Zeiss) that was also used to focus the Micropoint laser. All vertices in the field were identified and tracked using the segmentation algorithm above. Edges responded similarly to ablation whether identified by orientation or myosin intensity.
Pharmacological inhibitors were injected ventrally into the perivitelline space of embryos at stage 7 or stage 8. Rho-kinase inhibitor (Y-27632 dihydrochloride, Tocris Bioscience) was injected at 100 mM. Calyculin A (Sigma) was injected at 1 μM. Injected solutions are diluted ~50-fold in the embryo.
Dechorionated embryos were lined up on an agarose pad and transferred to a coverglass covered with heptane glue. Embryos were dried in a sealed container with silica beads for 6 min and covered with a 1:1 halocarbon oil 27:700 mixture (Sigma). Embryos were imaged with an Ultraview RS5 spinning disk confocal and a 40X oil immersion objective (NA 1.3, Zeiss). A pulled glass micropipette connected to a manual piston pump (CellTram Air, Eppendorf AG) was mounted in a micromanipulator on the microscope stage. The micropipette tip was broken against a coverglass and inserted into the perivitelline space of embryos expressing sqh-sqh:GFP, resille:GFP or spider:GFP. Negative pressure was applied to the micropipette until cell displacement was observed. Insertion of the micropipette occasionally resulted in cell wounding and myosin accumulation prior to the application of negative pressure; these experiments were discarded.
Dechorionated embryos were mounted in halocarbon oil 27 (Sigma) between a cover slip and an oxygen-permeable membrane (YSI). The ventrolateral germband was imaged with an LSM 510 laser scanning confocal (Zeiss) using a 40X oil-immersion objective (NA 1.3, Zeiss) and zoom 4. Embryos expressing Myo:GFP were imaged at an optical slice thickness of 1.7 μm to include a majority of junctional myosin. A 1.0 μm × 1.4 μm cortical region was photobleached. Fluorescence intensity in the bleached region was measured at each time point using custom Matlab routines. Intensities were background corrected by subtracting the post-bleach fluorescence. Only experiments in which cell boundaries remained in focus throughout the movie were analyzed.
Mean values were compared using Student's t test, with Holm's correction when more than two groups were considered (Glantz, 2002). The variances of multiple data sets were compared using the F-test. To compare time curves, the areas under the curves were used as the test statistic. The significance of correlation coefficients was evaluated by transforming the correlation value into a t statistic using the Matlab corrcoef function (Mathworks). Monte Carlo simulations were carried out for each image by randomly assigning myosin to the same number of edges that were identified as myosin-positive in vivo. Edges with an orientation of 60–120° were randomly selected to recapitulate myosin planar polarity. The percentage of myosin-positive edges in cables was calculated for each in vivo image or simulation (20,000 simulations/image). The distribution of myosin-positive edges in vivo was compared to the in silico distribution using the Kolmogorov-Smirnov test.
We thank Ori Weitz for developing the computational methods for automated cell tracking in Figure 1E. We are grateful to Richard Zallen for valuable input throughout this project and Dene Farrell, Morgan Reeds, and Justina Sanny for manually correcting segmented movies. We thank Adam Martin and Eric Wieschaus for fly stocks and Kathryn Anderson, Emily Marcinkevicius, Mimi Shirasu-Hiza, Gillian Siegal, Masako Tamada, Athea Vichas, Ori Weitz, Eric Wieschaus, Frederik Wirtz-Peitz, and Richard Zallen for comments on the manuscript. This work was supported by a Burroughs Wellcome Fund Career Award in the Biomedical Sciences, a March of Dimes Basil O'Connor Starter Scholar Award, a Searle Scholar Award, a W. M. Keck Foundation Distinguished Young Scholar in Medical Research Award, and NIH/NIGMS R01 grant GM079340 to JAZ. JAZ is an Early Career Scientist of the Howard Hughes Medical Institute.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.