Divergent evolutionary strategies preempt tissue collision in fly gastrulation

Metazoan development proceeds through a series of morphogenetic events that sculpt body plans and organ structures. In the early embryonic stages, morphogenetic processes involving growth and deformation occur concurrently. Forces generated in one tissue can thus increase mechanical stress in the neighboring tissue, potentially disrupting spatial patterning, morphological robustness, and consequently decreasing organismal fitness. How organisms evolved mechanisms to reduce or release inter-tissue stresses remains poorly understood. Here we combined phylogenetic survey across a whole insect order (Diptera), quantitative live imaging, and functional mechanical perturbation to investigate the evolution of mechanical stress management during epithelial expansions in the gastrulating fly embryos. We find that two distinct cellular mechanisms exist in Diptera to prevent the accumulation of compressive stress that can arise when the expanding head and trunk tissues collide. In Cyclorrhapha, a monophyletic dipteran subgroup including the fruit fly Drosophila melanogaster, the head-trunk boundary undergoes active out-of-plane deformation to form a transient epithelial fold, called the cephalic furrow (CF), which acts as a mechanical sink to preempt head-trunk collision. Genetic or optogenetic elimination of the CF leads to tissue buckling, yielding deleterious effects of axial distortion that likely results from unmitigated release of compressive stress. Non-cyclorrhaphan flies, by contrast, lack CF formation and instead display widespread out-of-plane division in the head, which shortens the duration of its expansion and reduces surface area increase. Reorienting head mitosis in Drosophila from in-plane to out-of-plane partially suppresses the need for epithelial out-of-plane deformation, suggesting that out-of-plane division can act as an alternative mechanical sink to prevent tissue collision. Our data suggest that programs of mechanical stress management can emerge abruptly under selective pressure of inter-tissue mechanical conflict in early embryonic development.

. We then performed live imaging and found that at gastrulation onset 127 they lack planar polarized Myosin accumulation in the cells predicted to initiate the CF 9 and consequently 128 do not form the CF (Fig. 2a, b; Fig. 2 Movie S1). Instead, the head-trunk interface undergoes a late stage 5 which the buckles are formed without localized cell shape change, all suggest that these are not 139 genetically patterned active deformations, but are consistent with passive buckling. Similar head-trunk 140 buckles could also be observed in the classic eve and btd mutants (Fig. 2e, f, i; Fig. 2 Movie S2), 141 confirming previous reports 9,10 and the independent study by Vellutini et al. 18 , thus validating the use of 142 eve or btd mutants, or global RNAi knockdown (see below), in generating the head-trunk buckles.

143
To rule out the possibility that the head-trunk buckles arise from local alteration of genetic patterning, we 144 mechanically blocked CF formation using an optogenetic system, the opto-DNRho1 system 9,21 , that 145 inhibits actomyosin contractility. To block CF formation, we illuminated the entire CF region on one side of

155
In search of the potential source of compressive stress, we made two observations. First, we observed a 156 correlation between mitosis and buckling in embryos in which the CF is optogenetically inhibited, as the   Fig. 1 Movie S1 top panel). PIV analysis also reveals that the CF 175 breaks the continuity of the flow field by separating it into two persistent, regionally coherent flows in the 176 flanking tissues -a posterior-ward flow in the head and a ventral-ward flow in the trunk that diverges 177 along the A-P axis (Fig. 2j, control at MD1 telophase onset). This suggests that without a sink, the head 178 and the trunk tissue flows would have collided at the head-trunk interface. In contrast, the eve1 KO embryo 179 displays a single, uninterrupted flow field at the onset of gastrulation, as predicted in a previous 180 computational model when the CF is absent 15 (Fig. 2j, eve1 KO at gastrulation onset). This supports the 181 premise that the CF functions as a sink. As the embryo begins to buckle, a local convergent flow emerges 182 at the head-trunk boundary, similar to the one we observed at the CF in the wild-type, suggesting that the 183 buckle also behaves as a sink (Fig. 2j, eve1 KO at MD1 telophase onset). The head-trunk buckle breaks 184 the continuity of the flow field, similar to the CF, further supporting the idea that the head-trunk boundary 185 interfaces two colliding tissue flows. We observed similar flow fields in embryos that lack Btd expression, 186 confirming these analyses ( Fig. 2     reported in the independent study by Vellutini et al. 18 . These data thus support the hypothesis that head 207 expansion contributes to buckling, presumably by driving the posterior-ward flow of the head that collides 208 with the trunk.

7
In the trunk, tissue flow stems from the combined effect of VF formation and GBE 15 the former of which is 210 known to drive the ventral-ward flow 15 , while the latter likely accounts for the A-P divergent flow, in light of 211 our PIV analysis that shows the anterior-ward flow of the trunk ectoderm (Fig. 2j) Fig. 2 Movies S4, S5). These data suggest that trunk 220 extension along the A-P axis contributes to buckling, likely due to a GBE-driven, A-P divergent flow that 221 collides with the head ectoderm.

222
Together, data presented above support our hypothesis that genetically programmed tissue expansion in 223 the head and trunk results in tissue collision, producing head-trunk buckles when the CF does not form.

224
The expanding head and trunk indeed 'fuel' the CF, as both stg and khft mutants form shallower and less 225 persistent CF than the wild-type ( Fig. 2 Supplement 2c, d). Thus, although the CF is initiated by the local 226 increase of actomyosin contractility 9 , its subsequent, extensive invagination requires the expansion of the 227 neighboring tissues. Given the pliability of invagination depth and persistence, we conclude that the CF 228 has the capacity to 'absorb' the tissue surfaces of the expanding neighbors, acting as a bona fide, 229 genetically patterned mechanical sink that guides the tissue flow, thereby preemptively preventing tissue 230 collision and buckling at the head-trunk interface. This interpretation is further corroborated by the in silico 231 simulation performed in the independent study by Vellutini et al. 18 .

233
The precision and robustness with which the CF is formed suggest that its spatial patterning and temporal 234 dynamics are under selective pressure. Thus, CF formation must confer fitness 9,14 . To test this, we asked 235 whether head-trunk buckling has a deleterious effect on embryonic development. We eliminated the CF 236 bilaterally using the opto-DNRho1 system such that the phenotypic effects can only be attributed to loss 237 of the CF, but not altered genetic patterning. We then imaged the ventral half of the embryo to monitor 238 embryonic development for ~1.5 hours after the onset of gastrulation. We first confirmed that optogenetic 239 perturbation indeed eliminates CF initiation as gastrulation commences (Fig. 3a, b; Fig. 3 Movie S1). The

240
VF initiates normally, indicating the effect of optogenetic perturbation is restricted to the CF. VF formation 241 continues with closure as the mesectodermal cells meet, resulting in a straight ventral midline. After the 242 ventral midline forms, strikingly, we observed an increased frequency of midline distortion or rotation in 243 embryos in which the CF is blocked, as compared to the sham control that maintains a bilaterally 244 symmetric body plan (Fig. 3a, b, e; Fig. 3 Movie S1). In embryos in which the CF is optogenetically 8 eliminated, the extent to which the ventral midline becomes distorted is variable, with strong midline 246 distortion often associated with bilateral asymmetry of head-trunk buckling. These results suggest that 247 releasing compressive stress via buckling is intrinsically stochastic, whereas programmed, active 248 deformation, such the CF, reduces such stochasticity. We further confirmed that btd mutants (Fig. 3c, d, f;

9
We further characterized the cell division plane in the Chironomus head. We observed out-of-plane 280 divisions in MD1 (data not shown), but since the MD1 cells are located near the anterior pole where the 281 embryo surface is highly curved and thus challenging to analyze, we focused on the more accessible 282 MD2. Of all the MD2 cells for which the division orientation can be determined unambiguously, about 50% 283 divide out-of-plane ( Fig. 4 Supplement 3b). Cells of each division mode occupy a distinct spatial domain, 284 which we termed MD2o and MD2i, respectively, for out-of-plane and in-plane domains (Fig. 4c). MD2o   We hypothesized that in Chironomus out-of-plane divisions attenuate the degree of head expansion, 295 thereby avoiding the need to release compressive stress via out-of-plane tissue deformation. To test this 296 possibility, we quantified the temporal dynamics of MD2 surface expansion. As the cell enters mitosis, its 297 apical area increases due to mitotic rounding. Prior to telophase onset, an in-plane dividing on average 298 cell reaches an apical area ~1.9-fold of the initial area prior to rounding. Following cytokinesis, the two 299 daughter cells reach a combined area of 2-fold, after which they shrink back and occupy a combined 300 surface area identical to that of the mother cell. The total duration of expansion is ~24 min ( Fig. 4e; Fig. 4 301 Supplement 4a). In contrast, the out-of-plane dividing cell expands only to ~1.6-fold by telophase for a 302 duration of only ~9 min. Furthermore, for the daughter cell that remains on the surface following 303 cytokinesis, the occupied surface area decreases rapidly, first to the size of the mother cell and then

309
We found that MD2o expands for ~12 min and to a maximum of ~1.4-fold of the initial area, while MD2i 310 expands for ~25 min and to a maximum of ~1.6-fold (Fig. 4f), indicating that the differences observed 311 between in-and out-of-plane divisions in individual cells are conserved at the tissue level. Taken 312 together, our results reveal that the orientation of cell division constitutes a critical parameter that likely 313 controls the accrued compressive stress, suggesting that non-cyclorrhaphan flies use out-of-plane 314 divisions as an alternative mechanism to avoid tissue collision during ectodermal expansion.

316
During Drosophila gastrulation, spatial-temporally restricted expression of the mitotic spindle anchoring 317 protein Inscuteable (Insc) is required for out-of-plane divisions in MD9 26,27 . We cloned the Chironomus 318 ortholog of insc and found that it is expressed throughout the entire head region (Fig. 4g). Although we 319 have not been able to assay its functional requirement in division orientation (see Discussion), the 320 expression pattern of insc implies that the entire head region of the Chironomus embryo is genetically 321 conducive to out-of-plane division.

322
Taking advantage of the fact that in Drosophila Insc is necessary and can be sufficient to instruct out-of-

336
To test this, we examined the effect of Insc OE and found that although 60% of embryos (Class II) undergo

402
Previous studies provided some hints. For example, one hypothesis proposes that the CF serves as a 403 temporary storage of cells that will subsequently contribute to future head development 35

409
CF formation in Drosophila has been previously shown to represent a morphogenetic process whose 410 spatial precision depends on not only the robustness of genetic patterning, but also mechanical self-411 organization 9 . Our surprising findings of the CF as an evolutionary novelty, and its co-evolution with the

423
indicating that the CF appeared once (arrow) in the phylogeny of Diptera, and then did not disappear.

424
Highlights indicate families for which species were evaluated (light) and studied (dark) in this work. See

952
For injections in Chironomus, embryos were collected, prepared, and injected essentially as described 40 .

953
Embryos were injected before the start of cellularization (approximately four hours after egg deposition), 954 and then kept in a moist chamber until the onset of gastrulation. Throughout all procedures, embryos 955 were kept at 25°C (± 1 C). Owing to their small size, Chironomus embryos (200 mm length) were always 956 injected into the center of the yolk (50% of A-P axis). Embryos were injected with dsRNA typically at 957 concentrations of 300 to 700 ng/ml; mRNA was injected typically at concentrations of 1.5-2.5 μg/μl (Cri-