JNK Signalling Controls Remodelling of the Segment Boundary through Cell Reprogramming during Drosophila Morphogenesis

Reprogramming of a specific group of Drosophila epidermal cells allows the mixing of normally segregated populations and the release of mechanical tension that arises during morphogenesis.


Introduction
Patterning of tissue progenitors through specific gene expression precedes tissue morphogenesis. Once cells are committed to a particular lineage, they generally keep to it throughout development. Nonetheless, plasticity of segmental lineages is commonly observed during the stages of boundary sharpening, like for example during Drosophila segmentation [1,2,3,4] and rhombomere formation in the vertebrate hindbrain [5,6,7,8,9,10]. In contrast, during later development, reprogramming of patterned cells is mostly associated with pathological conditions (e.g. regeneration) [11] or experimental procedures (e.g. cloning, grafting, or overexpression of selector genes) [12]. Rare cases of fate switching have nonetheless been reported during somitogenesis and hindbrain segmentation in the chick embryo [5,13,14] and during Caenorhabditis elegans embryogenesis [15]. Still, whether patterning can be re-adjusted during late tissue morphogenesis remains elusive.
Dorsal closure in Drosophila embryos is a powerful model of epithelial morphogenesis and wound-healing [16,17,18]. It proceeds through cell stretching and a zipping mechanism that lead to the convergence and suture of the lateral leading edges (LE) at the dorsal midline (see Video S1). This cell movement is believed to be collective and uniform. By looking at dorsal closure in live Drosophila embryos, we reveal a highly stereotyped pattern of cell reprogramming and intercalation, resulting in the remodelling of segment boundaries during late epithelial morphogenesis.

Cell Mixing and Intercalation at the Segment Boundaries during Dorsal Closure
Tracking of the dorsal ectoderm cells using confocal live imaging revealed several unexpected cell rearrangements taking place within the leading edge ( Figure 1A-D and Video S2). First, we observed that in abdominal segments, one cell from each anterior compartment mixes with the posterior compartment by the end of dorsal closure. We designate these versatile cells the mixer cells (MCs; yellow in Figure 1B-D). These cells have been noticed recently and have been qualified as an aberration in patterning [19]. Second, we show that two cells from the ventral ectoderm intercalate into the leading edge, posterior to each MC ( Figure 1C,D). The two intercalating cells, one from the anterior compartment (anterior intercalating, AI; green in Figure Figure 1D and Video S3). This striking pattern of remodelling is spatially and temporally regulated along the leading edge, with a degree of fluctuation, from embryo-toembryo, in the timing and number of intercalating cells ( Figure 1E).
To investigate the mixing mechanism, we analysed the origin and identity of the MCs during dorsal closure. Originally, the MCs occupy the dorsal-anterior corner of each anterior compartment ( Figure 1C,D). They are clearly identifiable as part of a single row of cells, known as the groove cells, which form a morphological furrow that marks each segment border, perpendicularly to the leading edge [20,21]. Like other groove cells, the MCs express higher levels of the actin anti-capping protein Enabled (Ena) (Figure 2A; Figure S1A). The anterior nature of the MCs was confirmed by looking at endogenous Patched (Ptc) expression, which is indeed present throughout the process of cell mixing ( Figure 2A; Figure S1B). Thus, both its initial position as well as the expression of Ena, Ptc, and of compartment specific drivers (ptc-gal4 positive and en-lacZ negative; see Figure S1C) show that the MC is the dorsal-most anterior groove cell.

De Novo Expression of Engrailed in the MC Induces Its Shifting to the Posterior Compartment
The MC behaviour challenges the compartment boundary rule stating that cells from different compartments cannot mix due to different cell affinities that sort them out [2,3,4,10,22,23]. One possible explanation for the violation of this law is that the MCs may be re-programmed to acquire posterior identity. Strikingly, the analysis of endogenous protein levels revealed that the MCs start expressing the selector protein Engrailed (En) [24] prior to their shifting towards the posterior compartment (Figure 2A,B; Figure S1 data). The profile of En accumulation in the MCs is distinct from bona fide posterior En-expressing cells present in the neighbouring posterior compartment ( Figure 2B; Figure S2), suggesting that En expression in the MCs is controlled by a different mechanism. Double staining for endogenous En and Ptc shows that the MCs express both markers (Figure 2A; Figure S1B), Ptc first then both, which supports the idea that the MCs were originally anterior cells that subsequently acquired posterior identity. Consistent with previous work showing that ectopic expression of En in anterior cells is sufficient to determine posterior-type cells [25,26], these results suggest that the MCs undergo anterior-to-posterior reprogramming through de novo expression of the En posterior determinant, thus favouring their mixing into the posterior compartment.
To demonstrate a direct role of En in MC formation, we inhibited its function in the anterior compartment by inducing en RNAi using the ptc-gal4 driver. These embryos showed a decrease of En expression in the MC ( Figure 3A). In addition, they exhibited a significant number of segments (40%) with aberrant cell mixing, i.e. with partial or no mixing at all ( Figure 3B). These results indicate that de novo expression of En in the MCs is essential for their reprogramming and mixing behaviour.

JNK Signalling Controls En Expression and Cell Mixing
The differentiation of the dorsal leading edge, to which MCs belong, is under the control of the conserved JNK pathway. Embryos lacking the activity of the JNKK/hemipterous (hep) gene do not express the LE reporter line puckered-lacZ (puc-lacZ), fail to undergo dorsal closure and die later in development [27]. Interestingly, JNKK mutant embryos are completely lacking cell intercalation and MC shifting ( Figure 4A). The expression of Ptc and Ena is normal in these embryos, showing that the identity of the groove cells is not affected in JNKK mutants ( Figure 4A; Figure  S3A). In contrast, expression of En could not be detected in the MCs ( Figure 4A top; Figure S3A), which indicates that JNK signalling is essential for de novo En expression. To distinguish compartment specific activities, a dominant negative form of Drosophila JNK/Basket (Bsk DN ) was expressed either in the anterior or in the posterior compartment using the ptc-gal4 or en-gal4 driver, respectively. The extinction of JNK activity was assessed by the loss of puc-lacZ expression ( Figure 4A). Embryos expressing Bsk DN in the posterior compartment (en.bsk DN ) showed no phenotype ( Figure 4A). In contrast, expression of Bsk DN in the anterior compartment (ptc. bsk DN ) led to the complete absence of MC intercalation, as is observed in JNKK mutant embryos ( Figure 4A,C; Figure S4 and Video S4). The same result was obtained when blocking JNK signalling through overexpression of the JNK phosphatase Puckered (Puc) ( Figure 4C; Figure S3B). Absence of JNK activity in ptc.bsk DN or ptc.puc embryos (but not with the en-gal4 driver) also led to the abolition of En expression in the MCs ( Figure 4A; Figure S3B). Interestingly, although most (87%) ptc.bsk DN embryos were able to complete dorsal closure, 92% of them showed a high degree of segment mismatching at the dorsal midline (53% of A1-A6 segments showed defects; Figure 4D). This suggests that MC formation and intercalation play a role in segment adjustment at the time of suture, consistent with a previous hypothesis [19]. In contrast, matching was normal in en.bsk DN embryos. Together, these results indicate that JNK signalling is essential in the anterior compartment, most likely in the MCs, to promote anterior-to-posterior reprogramming through de novo expression of En, compartment mixing, and segment adjustment.
In order to address the effect of excess JNK activity in the process, we ectopically expressed either a wild type (Hep) or an activated form of DJNKK (Hep act ) in the anterior compartment using the ptc-gal4 driver. These gain-of-function conditions induced a dramatic increase in the number of intercalated cells and the formation of ectopic MCs at the segment boundaries ( Figure

Author Summary
Multicellular organisms are assembled from different cell types, each following a particular fate depending on their history and location. During development, cells are organized into compartments, which are essential for the correct formation of organs. Within the compartments, cells follow two general rules: (i) cells that have acquired a given fate cannot change their differentiation state and (ii) cells from one compartment stay together and never mix with cells from other compartments. In this work, we identified a group of unique cells in Drosophila melanogaster embryos called mixer cells which move from one compartment of the epidermis to another, breaking the compartment boundary rule. Our data show that this unique behaviour depends on the nuclear reprogramming of the mixer cells, which change their fate and acquire the identity of the destination compartment de novo. We show that the shift in identity and compartment mixing are due to the expression of a single gene (Engrailed), under the control of JNK, a signalling pathway that is conserved across species. Interestingly, this process of reprogramming and mixing provides a mechanism of tension relaxation to the tissue during morphogenesis that allows dorsal closure of the Drosophila embryo (an event that resembles wound healing). This work reveals a novel model of cell plasticity that is amenable to genetic study, with potential application in the field of regenerative medicine.
express Ptc, Ena, and En like normal MCs. These results show that more lateral groove cells are competent for reprogramming, but they are restricted by the field of JNK activity in the leading edge.

Wingless Inhibits Groove and MC Formation
Each MC has a mirror-image counterpart at the LE parasegment (PS) boundary (MC* in Figure 5A) that never develops into a MC. Interestingly, the asymmetry of the MC pattern correlates with Wingless (Wg) activity across the segment [28] and the presence of the groove at the segment boundary ( Figure 5A) [20,21]. In addition, in JNK gain-of-function embryos, extra MCs only appear along the segment boundary ( Figure 4B), suggesting that only groove cells can differentiate into MCs. To test this hypothesis, we made use of specific wg mutant embryos in which an ectopic groove is formed at the PS boundary [20]. In this context, MCs* were transformed into ectopic MCs at the PS boundary ( Figure 5B). Like genuine MCs, transformed MCs* express Ptc, Ena, and most importantly En, which suggests that Wg suppresses the MC pathway at the PS boundary. To test whether Wg itself can repress MC formation, Wg was expressed ectopically in the MCs (ptc.wg) where it is not normally active [28]. This blocks MC reprogramming and cell remodelling ( Figure 5C; Video S6). Consistently, En expression is no longer detected in MCs. These results indicate that Wg has a nonpermissive function at the PS boundary through the blocking of groove cell differentiation, thus restricting the MC pathway to the segment boundary ( Figure 5D). Therefore, only dorsal groove cells are competent for MC formation ( Figure 5D).

Local Tissue Tension Modifies the Dynamics of Cell Intercalation
Dorsal closure is characterised by dramatic cell elongation (3fold in the DV axis) accompanied by the formation of a LE supracellular actin cable and amnioserosa contraction, all of which contribute to tissue tension (see Video S1) [29][30][31][32][33]. To test the effect of tension on the intercalation process, we applied laser ablation to live embryos expressing bCatenin-GFP. The tension in tested segments was assessed in three conditions (control, amnioserosa ablation, and cable ablation) by measuring the ectoderm recoil after a single cell ablation at the leading edge ( Figure 6B). Increase in LE tension was induced by ablation of the pulling amnioserosa, while its release was induced through a double ablation of the actin cable on each side of a test segment ( Figure 6A). We next compared the dynamics of LE insertion in the controls and in embryos mechanically challenged by laser. In control embryos, the PI cell (red in Figure 6C) takes, on average, 14 min to complete insertion in the leading edge. This time increases dramatically when tension is reduced in the cable (cable ablation condition; 60 min, Figure 6C middle panel, 6D), while it is shortened (4 min) in conditions of higher tension generated by amnioserosa ablation (Figure 6C bottom panel, 5D). These data show that the dynamics of intercalation depends on local tissue tension and suggest a role of intercalation in tension modulation. Improper tension release along the leading edge, in the absence of cell intercalation, could therefore explain the reduced ability of segments to match with their counterparts, as observed in JNK mutant conditions ( Figure 4D). To perturb tension genetically, we analyzed the pattern of intercalation in zipper (zip, encoding MyoII) mutants, in which a reduced tissue tension has been reported [34]. Interestingly, these embryos show a reduced level of cell intercalation, supporting our model of a link between tissue tension and the rate of intercalation ( Figure 4C).
Based on these results, we propose that the MC pathway provides an adaptive response to tissue tension by allowing an increase of cell number in the leading edge. Indeed, one major consequence of boundary remodelling is the addition of intercalating cells (AI and PI), which increases the cellular number of the leading edge by approximately 10% (Figure 1E). The adaptive nature of cell intercalation is reflected by the flexibility in the number (from 0 to 3) of intercalating cells ( Figure 1E), which contrasts with the robustness of MC reprogramming assessed by de novo expression of En. In our model, MC formation would weaken the segment boundary (i.e., through a change in cell affinity), making it a preferred site competent for tensiondependent intercalation. Our data and work published by Peralta et al. [35] indicate that the width remains constant on average with only slight oscillations during dorsal closure. Therefore, for constant width the increase in the number of cells implies that each cell is less stretched, thus inducing tension relaxation in individual cells. MC formation and associated local cell intercalation thus provide each segment with a tuneable relaxation compartment, important for tension release during morphogenesis ( Figure 6E).
In this study we unravel the mechanism of a unique case of breaching of the segment boundary during late morphogenesis, i.e. post-patterning and post-boundary sharpening. This process is shown to be highly stereotyped and developmentally regulated  through JNK signalling. Our data indicate that it takes place through a two-step mechanism, involving first MC formation, which is then followed by cell intercalation. Indeed, de novo expression of En in the dorsal groove cell always precedes intercalation ( Figure 2B). Furthermore, we can observe MC formation and mixing without intercalation like in the thoracic segments ( Figure 1E), but intercalation was never observed in the absence of MC formation: for example, when MC reprogramming is blocked in JNK loss-of-function conditions, no intercalation occurs ( Figure 4A). Cell mixing thus takes place through a novel morphogenetic mechanism involving plasticity of the segment boundary and compartment relaxation via patterned intercalation. It would be interesting to see if plasticity of boundaries can be a general mechanism for fine tuning late morphogenesis. Intriguingly, late expression of En in anterior cells has been reported at the anterior-posterior boundary in the wing imaginal disc. But contrary to the MC process, the so-called ''S. Blair cells'' do not mix with the posterior En-expressing cells [36], and their function remains elusive. It would be interesting to reinvestigate their late behaviour using time lapse approaches [37].
Images were taken with a Zeiss LSM 510 Meta confocal microscope using 640 1.3 NA or 663 oil immersion objectives.

Live Imaging, Laser Ablation, and Image Treatment
Embryos were dechorionated in bleach, then staged and placed dorsal side down on a coverslip. Embryos were then coated with halocarbon oil and covered with a hermetic chamber containing a piece of damp paper for hydration. This mounting system ensures normal development of 95% of embryos. Videos last from 2 to 5 h with stacks of 25 images (thickness from 30 to 40 mm) taken every 5 min.
Image and video assembly was done using ImageJ. Stacks are projected using either a maximal intensity or an average projection. Cell intercalations were analysed by tracking manually each cell with ImageJ. Graphs were made using Microsoft Excel. Video S3 was made using Microsoft PowerPoint and Alcoosoft PPT2Video converter.
Ablations were performed using a two-photon pulsed Spectraphysic's Tsunami laser combined with a Zeiss LSM 510 Meta confocal microscope for imaging. The power was calibrated in each experiment using a test embryo and ablations were performed with the Zeiss ''bleach'' macro to control the size and timing of each cut. For MC ablation, the actin cable on the dorsal side was targeted, while for amnioserosa ablation, the laser beam was focused on the apical area to destroy adherens junctions and the cytoskeleton in a region of interest of 10-30 micrometers long, parallel to the AP axis. To determine cable tension, we used the classical definition of tension in a purse string as the magnitude of the pulling force exerted by the string [29]. The application of Newton's second law under the conditions of low Reynolds number (viscous fluid) shows that the initial recoil speed of a cable after the cut is proportional to the contribution of the suppressed force, i.e. tension [32,42].

Protein Level Quantification in MCs
ImageJ was used to quantify En and b-Galactosidase levels on projections of non-saturated stacks of images. For a given segment, the absolute intensity of En in the MCs was normalised to the average absolute intensities of the bona fide En-expressing cells of the leading edge. An average of these relative intensities was calculated for stages of intercalation as shown in Figure 2. For each embryo, only segments A2, A3, and A4 were considered as they are most representative of mixing and intercalation. Relative intensity in the MCs is the ratio of absolute MC intensity/average of absolute intensities in bona fide En cells.

Statistical Analysis
All analyses were performed using the Mann-Whitney nonparametric test, which does not assume any condition on the distribution and is adapted to independent experiments and small sample sizes. p values were computed using the statistics toolbox from the Matlab software. Table S1 Timing and standard deviations of closure and cell intercalation described in Figure 1E. Video S1 Dorsal closure of a wild type embryo. Confocal time-lapse images are taken from an embryo expressing ubiquitous aCatenin-GFP (green) and h-Actin-CFP (red) in the posterior compartment. Genotype: bcatenin-GFP; en-gal4, UAS-h-actin-CFP.