Cadherin Fat2 directs cellular mechanics to promote epithelial rotation

Left and right symmetry breaking is involved in many developmental processes that are important to form bodies and organs. One of them is the epithelial rotation of developing organs. However, how epithelial cells move, how they break symmetry to define common direction of their collective movement and what function rotational epithelial motions have in morphogenesis remain elusive. Here, we identified a dynamic actomyosin network with preferred retrograde contractility at the basal side of the rotating follicle epithelium in Drosophila oogenesis. We provide evidence that unidirectional epithelial rotation is a result of actomyosin asymmetry cue transmission onto a tissue plane synchronized by the atypical cadherin Fat2, a key planar cell polarity regulator in Drosophila oogenesis. We found that Fat2 directs actomyosin contractility to move the epithelial tissue in order to provide directed elongation of follicle cells. In contrast, loss of Fat2 results in anisotropic non-muscle Myosin II pulses that are disorganized in plane and deform cell shape, tissue and Drosophila eggs. Our data indicate that directed elongation of follicle cells is critical for proper Drosophila egg morphogenesis. Together, we demonstrate the importance of atypical cadherins in the control of cell mechanics, left/right symmetry breaking and its propagation onto the tissue scale to facilitate proper organ morphogenesis. This process may be evolutionarily conserved in rotating animal organs.

Functional organ morphogenesis [1][2][3] has been linked to left and right (LR) turns and rotations of epithelial sheets 4 5 6 7 8 9 10 relative to the organ or body anterior-posterior (AP) axis. A primary determinant of this LR chirality has been associated with the cytoskeleton in different species 11 12 13 14 15 . In rotating Drosophila organs such as hindgut 10 and male genitalia 6  Epithelial rotation is initially slow during early oogenesis (stages 1-5: average speed ~ 0.2 µm/min) 19 accelerates in mid oogenesis (stages 6-8: average speed ~ 0.5-0.6 µm/min) 8,9,19 24 and stops at stage9, 8 . It has been shown that microtubules (MTs) predict the direction of epithelial rotation in early and mid-oogenesis and their global alignment is regulated by the atypical cadherin Fat2 25 24 . Fat2 is a key PCP regulator of the actin cytoskeleton 25 , basement membrane components 26 20 and its function is required for epithelial rotation and elongation of Drosophila egg chambers 25 9 . The Fat2 asymmetric planar polarized pattern on the basal lagging membrane side of a follicle cell depends on MTs during fast epithelial rotation 9 . There is no evidence that MTs represent the active force-generating mechanism that drives epithelial rotation, which has been recently shown to involve the actin-rich protrusions 19 23 . However, non-muscle myosin II (Myo-II) that generally provides contractility and force generation to actin cytoskeleton is missing on actin-rich protrusions 19 . Therefore, motivated by the observation that pharmacological depletion of non-muscle myosin II (Myo-II) leads to no epithelial rotation 9 , we hypothesized that the basal actin filaments that contain Myo-II are better candidates to fulfill the force generating function. To test this hypothesis, we investigated the function of Myo-II, its connection to the PCP pathway in Drosophila epithelial rotation and the role of their interplay in egg chamber morphogenesis.

Highly dynamic Myo-II behaviour at the basal side of the follicle epithelium
In order to understand Myo-II function in epithelial rotation, we first investigated the behaviour of the Myo-II regulatory light chain (MRLC, called Spaghetti Squash, sqh in Drosophila) fused to GFP (MRLC::GFP) in a null sqh AX3 mutant 27 at the basal cortex of the follicle epithelium using ex vivo live imaging. We analyzed Myo-II behaviour in three different situations: slow (stage 4), fast (stage 7) and no epithelial rotation (stage 7 of fat2 58D/103C mutants in a null sqh AX3 background since fat2 mutant egg chambers display no epithelial rotation 9,19 ) (Extended Data Fig. 1b). High-speed confocal live imaging of basal cortex of the follicle epithelium during slow, fast and no rotation (Figure 1b and Movies 1,2,3). We distinguished individual MRLC::GFP dot-like signals with an average size of 363 nm ± 0.05 nm (n = 136) and an average speed of 2.12 µm/min ± 0.8 (n = 101) for slow, 2.44 µm/min ± 0.96 (n = 105) for fast and 1.99 µm/min ± 0.62 (n = 100) for no epithelial rotation. This speed was consistent with the speed of anterograde flow of actomyosin during zebrafish gastrulation 28 . We also observed large intense MLRC::GFP dots (1.01um ± 0.14 um, n= 50, Figure 1b and Movie 2) close to the lagging end of migrating follicle cells, which were lost in the fat2 mutant follicle epithelium ( Figure 1b and Movie 3), suggesting an unknown function in epithelial rotation. Taken together, we discovered highly dynamic behaviour of Myo-II at the basal cortex of the Drosophila follicle epithelium.

Global actomyosin retrograde movement is regulated by atypical cadherin Fat2 in the follicle epithelium
To find out whether the small (~360nm) MRLC::GFP dots moved in a specific direction with respect to the egg chamber axis, we quantified directions of MRLC::GFP movement expressed as angles ranging from 0° to 360° where 0° represented the anterior and 180° the posterior of egg chambers (Extended Data Fig. 1c and Online Methods). In contrast to the situation with no epithelial rotation, which showed no clear preference in direction of MRLC::GFP movement (Figure 1b), we observed that during slow rotation, MRLC::GFP showed weak preferred movement perpendicularly to the AP axis of egg chambers ( Figure 1b) that was strongly reinforced during fast rotation (Figure 1b and Extended Data Fig. 1d). Similarly, labeling actin filaments with a LifeAct 29 molecule fused to GFP (LifeAct::GFP) showed strong preference in LifeAct::GFP movement perpendicularly to the AP axis of egg chamber during fast rotation (Extended Data Fig. 1d). This preference of small MRLC::GFP dots to move perpendicularly to the AP axis of egg chambers as well as the loss of their preferred direction in fat2 mutant egg chambers observed in live imaging was corroborated by the analysis of MRLC::GFP signal in fixed wild type and fat2 mutant egg chambers during early and mid-oogenesis (Extended Data Fig. 2).
Having established the global trend of Myo-II movement, we next asked whether individual MRLC::GFP dots moved randomly along their preferred direction (perpendicular to the AP axis of egg chambers) during slow and fast epithelial rotation.
To this end, we calculated how frequently MRLC::GFP dots moved within defined angle range of four (90 degrees) quadrants: Anterior (315°≤45°), Up (45°≤135°), Posterior (135°≤225°) and Down (225°≤315°). The data revealed preferred movement within Up and Down quadrants and an asymmetry in movement of MRLC::GFP dots within these quadrants in individual egg chambers (Figure 1b). This asymmetry was initially small during slow rotation and became prominent during fast rotation. In contrast, rather weak to no asymmetry has been detected in fat2 mutant egg chambers ( Figure 1b).
Next we asked in what way this MRLC::GFP movement asymmetry relates to the direction of epithelial rotation. In order to define the average percentage of MRLC::GFP dots moving with or against epithelial rotation, we unified the direction of epithelial rotations in the direction Up for all analyzed egg chambers and detected that on average 59% and 77% MRLC::GFP dots moved against epithelial rotation during slow and fast rotation, respectively ( Figure 1c and Extended Data Fig. 3a). This was not true for the fat2 mutant egg chambers (Figure 1c and Extended Data Fig. 3a), where no preferred global direction was identified in individual egg chambers. We also observed that actin molecules preferably moved (78%) against fast epithelial rotation, based on LifeAct-GFP, and this movement was comparable to the MRLC::GFP movement during fast rotation (Extended Data Fig. 3b and Figure 1c). Thus, we discovered that MRLC::GFP dots preferred to move against epithelial rotation during slow and fast rotation.
Contractile actomyosin rings, Myo-II pulses and Myo-II asymmetric localization on the cell membrane represent features that have been previously linked to tissue movements during epithelial morphogenesis 28 30 31 6 . However, we observed neither of these, indicating that known mechanisms of actomyosin mediated collective epithelial movement do not play a role in this system.
Taken together, our results revealed novel temporally regulated and spatially coordinated Fat2-depenent actomyosin movement at the basal side of the follicle epithelium in the direction opposite to epithelial rotation (retrograde movement). Altogether, our data provide evidence that Fat2 propagates the initial, intracellular actomyosin-based asymmetry onto the tissue level to facilitate epithelial rotation of Drosophila egg chambers.

Fat2-dependent epithelial rotation directs elongation of follicle cells
Because loss of Fat2 leads to static and round egg chambers 9 , we next wished to understand what role Fat2-dependent epithelial rotation plays in egg chamber morphogenesis. Epithelial rotation has been clearly linked to proper PCP of basement membrane and actin filaments, which is required for egg chamber elongation along the AP axis 18,19 . In addition, integrin-based adhesions to ECM can modulate speed of epithelial rotation and impact shape/stretching of follicle cells 18 . Thus, we wondered whether epithelial rotation could define the shape of follicle cells and so measured their roundness (Online Methods). Indeed, we found that follicle cells are on average significantly more elongated during fast rotation than follicle cells in static fat2 mutant egg chambers and less elongated during slow rotation ( Figure 3a). We also pharmacologically perturbed epithelial rotation by using actin-depleting drug

Fat2 suppresses anisotropic and premature Myo-II pulses
We next asked what role directed elongation plays in egg chamber morphogenesis.
Interestingly, when we analyzed individual fat2 mutant follicle cells, besides weak Myo-II asymmetries and rounder cell phenotype, we also observed spatially unequal (anisotropic) MRLC::GFP pulses (Figure 4a  In summary, we propose that Fat2 synchronizes the initial default LR Myo-II asymmetry in one direction in the plane (the global Fat2 function), reorients and reinforces actin cytoskeleton within follicle cells (the local Fat2 function), resulting in transmission of actomyosin-based LR asymmetry onto tissue scale. This facilitates epithelial rotation, directed elongation of follicle cells and cell-shape-dependent local and global alignment of MTs. As MTs have been reported to be relevant for turn-over of the Fat2 (zig-zag) asymmetric pattern during fast epithelial rotation 9 , we further hypothesize that Fat2 uses epithelial rotation to align MTs for establishment of its planar polarized asymmetric pattern in a positive-feedback amplification mechanism 9 . This mechanism thus provides a robust Fat2-MTs-dependent platform that guarantees stable global and local alignment of actomyosin and MTs cytoskeleton during fast epithelial rotation (stage 6-8), which upon its establishment and maintenance 9 likely becomes independent of epithelial rotation as observed in 19 . We envision that this is further critically required for globally aligned, physiological Myo-II pulses at the basal side of follicle cells in order to properly elongate follicle cells and egg chambers along their AP axis in later Drosophila oogenesis ( Figure 5).

Discussion
One of the best understood processes that involves actomyosin flow in LR symmetry breaking is single and four cell C. elegans embryo 15 , where active torque generates force that leads to LR symmetry breaking and embryo asymmetry.

Scientific Computing Facility at MPI-CBG (Robert Haase) for developing an in-house Fiji
Plugin. We are grateful to Stephan Grill for comments on the manuscript.

AUTHOR INFORMATION
Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Correspondence and requests for materials should be addressed to viktorin@mpi.cbg.de and tomancak@mpi-cbg.de.

Movie 3
Time-lapse movie of MRLC::GFP (green) signals moving at the basal cortex of the static fat2 58D/103C mutant mid-oogenesis egg chamber (stage 7). Membrane marker stains cell outlines (red). Note that MRLC::GFP is locally (within a cell) polarized (directed subcellular movement shown in Figure 2) but its global polarity is lost (Figure 1). Large MRLC::GFP dots were also lost. Cell outlines display deformations linked to MRLC::GFP increased intensity (Figure 4). Frame interval = 6s. Scale bar = 5 µm. Anterior is on the left.
Anterior is on the left.

Time lapse imaging.
Egg chambers were cultured and life imaging performed as described 9 . An inverted LSM 700 Zeiss confocal microscope was used with 63x/1.45 water immersion lens. Timelapse movies were taken with an interval of 6s for 300s-600s.

Fixation and immunohistochemistry.
Adult fly ovaries were dissected in 1xPBS and fixed with 4% p-formaldehyde for 20minutes. Immunostaining followed standard protocols. We used polyclonal GFP tag antibody conjugated with Alexa Fluor 488 (Molecular Probes) in dilution of 1:100 and rhodamine-phalloidin in dilution of 1:200. Images were acquired on inverted LSM700 Zeiss confocal microscope with 63x/1.45 oil immersion lens.

Drug treatments.
To inhibit polymerization of actin filaments, Latrunculin A (10 µM in 1% DMSO, Enzo Life Sciences) was used for ca. 10mins before direct imaging. To deplete actin protrusions and follicle cell adhesion to the ECM, we used Arp2/3 inhibitor (CK-666, 250 µM, Sigma) for ca. 1h as described 19 .
Image processing, data analysis and statistics.

Measurement of direction of MRLC::GFP (small dot-like signals) movement
To measure the direction of MRLC::GFP movement in individual follicle cells and globally in the epithelial tissue, we measured angle relatively to the AP axis of egg chambers (Extended Data Figure 1c) with 'Angle' tool in Fiji. Before angle measurement, time-lapse movies were corrected for bleaching and cell membranes were registered for their movement to make them static. Angles were then measured on time (60s) projections of MRLC::GFP signals. Altogether, we measured movement of MRLC::GFP relatively to the cell membrane.
To unify fat2 mutant egg chambers with no epithelial rotation in order to compare with the weak and fast epithelial rotation data, the higher value of MRLC::GFP movement identified for Up (45°<135°) and Down (225°<315°) quadrants was artificially assigned to the Down quadrant to mimic as if all egg chambers rotated upwards, i.e. the preferred MRLC::GFP movement was against (retrograde) epithelial rotation (used in Figure 1c and Figure 2b,c,d).

Measurement of Myo-II size and velocity
The size of small and big MRLC::GFP signal was measured as a diameter over 5 and 10 independent egg chambers, respectively. Myo-II velocity was measured on a displacement of Myo-II signals in 30s-60s in original time-lapse movies over 10 independent egg chambers.

Measurement of velocity of epithelial rotation
The velocity was defined as an average velocity over 3 independent measurement of cell membrane movement in the most central part of the confocal plane.

Angular correction
To define a direction of the MRLC::GFP movement within follicle cells in fat2 mutant egg chambers, time projected MRLC::GFP pattern served as a definition of the main MRLC::GFP arrays that were reoriented within the existing smallest angle (≤90°) to achieve the perpendicular orientation to the AP axis (used in Figure 2). We hypothesized that Fat2 reorients Myo-II of follicle cells either by spatial regulation of the intracellular actomyosin dynamics or via remodeling of adherens junctions. Thus, in both cases, we concluded that it will be an energy-demanding process and angularly corrected follicle cell perpendicularly to the AP axis of egg chambers (based on the time projected Myo-II pattern) in the smallest possible angle (e.g. as shown in Figure 2a, 45° clockwise and not 135° anti-clockwise).

Quantification of global Myo-II and actin filament alignment
To measure global alignment of MRLC::GFP and actin filaments in fixed tissues, we used Fiji software 'Directionality' http://imagej.net/Directionality.

Measurement of follicle cell shape and elongation direction
Outlines of follicle cells were used to measure Roundness parameter (Shape The time series were smoothened with a Gaussian filter with window of 10 data points