A critical step in the closure of the mouse midbrain is the transformation of the neural plate from convex to concave (Figure 1A–C; Nikolopoulou et al., 2017; Vijayraghavan and Davidson, 2017; Juriloff and Harris, 2018). Prior to closure, the cranial neural plate has an open, rams-horn shape (Figure 1C). The neuroepithelial sheet is convex on either side of the midline, with the outer edges of the neural plate tucked under the lateral regions. This curvature reverses during neural fold elevation, when both sides of the neural plate rise up and straighten to produce a concave, V-shaped structure (Figure 1C). The borders of the neural plate subsequently bend inward, appose, and fuse at the dorsal midline to produce a closed tube. To investigate the cell behaviors that drive these structural changes, we used confocal imaging and semi-automated image segmentation (Mashburn et al., 2012; Farrell et al., 2017) to analyze cell behavior at single-cell resolution. The apical profiles of midbrain neuroepithelial cells were relatively homogeneous in area prior to elevation (0 somites, E7.75) (Figure 1D and E). However, a strong pattern emerged during elevation (6 somites, E8.5). Lateral cells on either side of the midline displayed a more than 50% decrease in apical area between 0 and 9 somites (Figure 1F–H, Supplementary file 1). By contrast, the average apical surface area of midline cells did not change significantly during elevation (Figure 2A–C). Additionally, lateral cells became progressively mediolaterally oriented during the same period, whereas midline cell orientation was unchanged (Figure 2—figure supplement 1). These results indicate that lateral cells, but not midline cells, undergo apical remodeling during cranial neural fold elevation.

The finding that midline cells in the midbrain do not remodel during neural fold elevation differs from neural tube closure mechanisms in the spinal cord, in which wedge-shaped midline cells drive tissue bending (McShane et al., 2015; Schoenwolf and Franks, 1984; Smith et al., 1994; Smith and Schoenwolf, 1988), and raises the possibility that lateral cells may be key drivers of elevation. To determine if the apical remodeling of lateral cells is due to apicobasal elongation, we analyzed cell height in the cranial neural plate at different stages of elevation. Cell height in the lateral and midline regions did not change significantly during early elevation (0–7 somites) (Figure 3C and D), even though the average apical area of lateral cells decreased by more than 30% during this period (Figure 1G). By contrast, lateral and midline cells elongated more than 60% along the apical-basal axis after the 7-somite stage, such that lateral cells were consistently taller than midline cells throughout elevation (Figure 3C–E). Thus, cell remodeling in the elevating midbrain occurs in two phases, with an early phase involving apical remodeling in the absence of changes in cell height, and a later phase involving apicobasal elongation (Figure 3J). These results indicate that apicobasal elongation in the neuroepithelium occurs at late stages of elevation, but cannot explain the dramatic structural changes that occur during early elevation.

We next investigated whether apical constriction contributes to early structural changes in the midbrain neuroepithelium. Consistent with this possibility, the conversion of the midbrain neural plate from convex to concave is accompanied by a decrease in the apical span of the tissue without a significant change in the basal span (Figure 3A and B, Figure 3—figure supplement 1A and B). However, apical constriction has not been directly observed in the pseudostratified mammalian neural plate, where the crowded packing of cells has been proposed to hinder this process (Nikolopoulou et al., 2017). To determine if lateral cells undergo apical constriction, we visualized cell morphology in the midbrain neuroepithelium of embryos expressing membrane-GFP in a mosaic pattern, using the inefficient EIIA-Cre recombinase to label individual cells (Figure 3F; Lakso et al., 1996; Muzumdar et al., 2007). Using this approach, we identified five classes of lateral cells (Figure 3G). More than half of lateral cells (51 ± 6%) had a highly constricted apical neck, a hallmark of apical constriction. An additional 13 ± 2% of cells displayed properties consistent with apical constriction, but with a shorter neck domain, suggesting that apical area changes can occur even in the absence of a basal shift in cell volume. The remaining one-third of lateral cells were apically expanded, spindle-shaped, or columnar. Because few midline cells were labeled by this method, we used manual segmentation to investigate cell shape at the midline using antibodies to β-catenin (Figure 3H). In contrast to lateral cells, midline cells tended to be columnar (45 ± 5%) or apically expanded (30%), with relatively few midline cells showing apically constricted morphologies (17 ± 3%) (Figure 3I, Figure 3—figure supplement 2A and B). These results directly demonstrate the presence of apical constriction in the lateral neural plate and reveal a striking regionalization of cell-shape changes along the mediolateral axis.

To identify the mechanisms that regulate the distinct behaviors of lateral and midline cells, we sought to identify mutants that disrupt this pattern. In a genetic screen for mouse mutants with embryonic defects (García-García et al., 2005), we identified two mutants with severe defects in cranial neural closure (Figure 4A–C). The mutations in these strains mapped to premature stop codons in Ift122 and Ttc21b (Ift139), which encode components of the conserved intraflagellar transport A (IFT-A) complex (Figure 4—figure supplement 1A and B). The IFT-A complex directs the trafficking of structural and signaling proteins in cilia, microtubule-based cellular organelles that modulate Shh signaling (Wong and Reiter, 2008; Bangs and Anderson, 2017). Consistent with these functions, mutant embryos from both strains exhibited a reduction in the number of cilia (Figure 4—figure supplement 1C–E). Mutations in IFT-A complex components have been shown to cause exencephaly, but how these proteins influence cranial neural closure is not known (Tran et al., 2008; Cortellino et al., 2009; Murdoch and Copp, 2010; Qin et al., 2011; Bangs and Anderson, 2017). In contrast to littermate controls, which completed neural fold elevation, apposition, and fusion in 24 hr, Ift122 and Ttc21b mutants failed to generate V-shaped neural folds in the midbrain, forebrain, and anterior hindbrain regions of 7-somite embryos (Figure 4D–F, Figure 4—figure supplements 2 and 3). These defects did not recover and the cranial neural folds of mutant embryos remained unelevated at all stages analyzed, leading to highly penetrant exencephaly at E10.5 (Figure 4A–C). Thus, the cranial closure defects in Ift122 and Ttc21b mutants arise from an early failure in cranial neural fold elevation.

To determine the cellular basis of these exencephaly defects, we analyzed cell shape in Ift122 and Ttc21b mutants. Mutant embryos displayed a striking expansion of the apical cell surface in the lateral midbrain (Figure 4G–I). Lateral cells in Ift122 and Ttc21b mutants displayed a 55% and 93% increase in average apical cell area, respectively, compared with littermate controls (Figure 4J–M), as well as altered cell orientation (Figure 4—figure supplement 4). These defects did not result from reduced cell proliferation, as mutant embryos had a normal frequency and distribution of mitotic cells along the mediolateral axis, and normal cell density in the underlying mesenchyme (Figure 4—figure supplements 5 and 6). These results indicate that Ift122 and Ttc21b are required for cell-shape changes in the lateral midbrain neuroepithelium.

To determine if the global pattern of cell remodeling is affected in IFT-A mutants, we examined cell-shape changes throughout the entire mediolateral axis of the midbrain in Ift122 and Ttc21b mutants. In wild-type littermate controls, apically expanded cells were present at the midline and at the outer margins of the tissue. These domains were separated by a broad domain of apically constricted cells spanning 30–40 cell diameters along the mediolateral axis and more than 60 cells along the anterior-posterior axis, encompassing a region of more than 2000 lateral cells on either side of the midline (Figure 5A and F–H). The difference between midline and lateral populations was eliminated in Ift122 and Ttc21b mutants (Figure 5B–E). In mutant embryos, lateral cells were apically expanded and midline cells were apically constricted compared with controls, whereas cell shape at the outer margins of the neural plate was independent of IFT-A activity (Figure 5F–H). Moreover, the difference in height between wild-type midline and lateral cells was abolished in Ift122 and Ttc21b mutants (Figure 5I and J, Figure 5—figure supplement 1A and B). These cell remodeling defects were associated with a failure to fully convert the cranial region from convex to concave in Ift122 and Ttc21b mutants (Figure 5—figure supplement 1C–F). These results demonstrate that Ift122 and Ttc21b are required for patterned apical remodeling in the midbrain neuroepithelium. In their absence, midline and lateral cells adopt a uniform cell morphology.

A hallmark of apical constriction is the requirement for apically localized actomyosin contractility (Martin and Goldstein, 2014). To determine if this is the mechanism by which Ift122 and Ttc21b promote apical remodeling in lateral cells, we analyzed the localization of F-actin and the phosphorylated (active) form of myosin II in Ift122 and Ttc21b mutants. Wild-type cranial neuroepithelial cells display a strong accumulation of F-actin and phosphomyosin at the apical cell cortex, which is often assembled into mediolaterally oriented actomyosin cables in the chick and mouse neural plate (Nishimura et al., 2012; McGreevy et al., 2015). In line with these observations, we observed frequent supracellular cables in the elevating cranial neural plate. Cables were present at a range of orientations, with a strong mediolateral bias (Figure 6A,B,E and F). By contrast, fewer actomyosin cables were present in Ift122 and Ttc21b mutants, and the cables that did form were not consistently oriented with respect to the mediolateral axis (Figure 6A–F). The ratio of phosphomyosin to F-actin at adherens junctions was also decreased in Ttc21b mutants (Figure 6G and H), consistent with the stronger apical constriction defects in this mutant. These results demonstrate that Ift122 and Ttc21b are required for apical actomyosin organization in the midbrain neuroepithelium.

Cilia are signaling organelles that are critical for Shh signaling and cell fate. Our finding that cilia proteins are also required for cell remodeling suggests that fate and morphology may be directly linked. In the spinal cord, the loss of IFT-A complex function typically results in ligand-independent activation of Shh signaling and an expansion of Shh-dependent ventral cell fates (Tran et al., 2008; Cortellino et al., 2009; Qin et al., 2011; Bangs and Anderson, 2017), although strong disruption of IFT-A function can result in a loss of Shh-dependent cell fates (Liem et al., 2012). To test whether Ift122 and Ttc21b pattern cell fate during cranial neural fold elevation, we analyzed the expression of Nkx6.1, a target of Shh signaling. In wild-type embryos, Nkx6.1 levels were highest at the midline during midbrain neural fold elevation and decreased with increasing distance from the midline (Figure 7A and B), consistent with results at later stages (Qiu et al., 1998; Tran et al., 2008; Qin et al., 2011; Tang et al., 2013). In addition, Nkx6.1 was expressed at lower levels in the anterior hindbrain, revealing differential regulation along the anterior-posterior axis (Figure 7A). In Ift122 and Ttc21b mutants, the mediolateral pattern of Nkx6.1 expression was abolished, and Nkx6.1 was expressed at equivalent, intermediate levels in midline and lateral cells (Figure 7A–C). In addition, the mediolateral extent of the Nkx6.1 domain was expanded, reaching all the way to the neural plate borders in Ift122 and Ttc21b mutants. By contrast, the anterior-posterior pattern of Nkx6.1 expression was unaffected, indicating that this axis of Shh regulation is independent of IFT-A activity. These results raise the possibility that deregulated Shh signaling could underlie the cranial closure defects in Ift122 and Ttc21b mutants.

To investigate the role of Shh signaling in midbrain cell remodeling, we examined cell morphology in mutants lacking the Shh effector Gli2, which is required to generate ventral Shh-dependent cell types (Mo et al., 1997; Matise et al., 1998; Bai and Joyner, 2001; Wong and Reiter, 2008; Bangs and Anderson, 2017). Because Shh signaling is normally highest at the midline, we asked if the unique architecture of midline cells requires Gli2 function. Consistent with the effects of Gli2 at later stages, Gli2 mutants failed to establish ventral cell fates in the elevating midbrain, including FoxA2 expression in the floor plate, indicating that Gli2 is required for midline cell identity (Figure 8—figure supplement 1). Midline cells in Gli2 mutants had a significant decrease in apical area compared with wild type (Figure 8D–F). In addition, midline cells in Gli2 mutants were significantly taller than in wild-type controls (Figure 8G and H). By contrast, lateral cell morphology was unaffected in Gli2 mutants (Figure 8A–C,G and H), and no defects in cell orientation were observed in either region (Figure 8—figure supplement 2). These data demonstrate that Gli2 activity is necessary for the short, apically expanded architecture of midline cells but not for apical constriction in lateral cells. Despite these severe midline defects, Gli2 mutants complete closure normally in both cranial and spinal regions (Mo et al., 1997; Matise et al., 1998; Bai and Joyner, 2001), indicating that the specialized architecture of midline cells is dispensable for cranial neural tube closure.

Because cranial neural tube closure occurs normally in the absence of proper midline morphology, we hypothesized that a failure of apical constriction in lateral cells could be responsible for the cranial closure defects in Ift122 and Ttc21b mutants. In addition, the expanded Nkx6.1 expression in Ift122 and Ttc21b mutants raises the possibility that increased Shh signaling in lateral cells could underlie the defects in apical constriction. To test these hypotheses, we investigated whether spatially restricted Shh signaling is required for apical constriction and cranial neural closure. We ectopically activated the Shh signaling response throughout the midbrain by expressing a constitutively active variant of the Shh receptor Smoothened (SmoM2) (Jeong et al., 2004) using Wnt1-Cre2 (Lewis et al., 2013). SmoM2-expressing embryos have expanded Nkx6.1 expression throughout the midbrain, consistent with uniform activation of the Shh response (Figure 9C and D). SmoM2 expression did not affect cell proliferation or cell orientation (Figure 9—figure supplements 1 and 2). However, SmoM2 expression resulted in a 50% increase in apical area in lateral cells (Figure 9G–I), similar to the defects in Ift122 mutants, but less severe than the defects in Ttc21b mutants. The effects of SmoM2 were localized, as cells that did not express Wnt1-Cre2 apically constricted normally, suggesting that activated SmoM2 acts cell autonomously to regulate cell shape (Figure 9E and F). SmoM2-expressing embryos did not display morphological defects at the midline, perhaps because Smoothened activation did not further enhance the already high Shh response in this region (Figure 9G,J and K). Notably, SmoM2-expressing embryos exhibited 100% penetrant exencephaly (12/12 SmoM2-expressing embryos compared with 0/13 littermate controls) (Figure 9A and B). These results demonstrate that disruption of apical constriction in lateral cells alone—in the absence of structural changes at the midline—is sufficient to prevent cranial neural closure. Thus, patterned Shh signaling in the midbrain neuroepithelium is required for spatially regulated apical remodeling events that drive cranial neural closure, and dysregulation of Shh activity leads to altered cell remodeling and exencephaly (Figure 9L and M).

The Ift122^TR2 and Ttc21b^TF2 alleles were identified in an ongoing forward genetic screen (García-García et al., 2005). The Ift122^TR2 mutation was mapped to a single C to A mutation at position 115,899,529 on chromosome 6, resulting in a premature stop codon that is predicted to truncate the protein at amino acid 575 (out of 1138). The Ttc21b^TF2 allele was mapped to a single C to A mutation at position 66,242,780 on chromosome 2, resulting in a premature stop codon that is predicted to truncate the protein at amino acid position 187 (out of 1,315). Both alleles created a new restriction site. The presence of the Ift122^TR2 allele was genotyped by PCR amplification with the primers TR2F 5’ CTGGTTGTAATCTGACTCGTTGA 3’ and TR2R 5’ ACTCCCAAGCAAGCGAACT 3’ followed by restriction digest with HpyCH4III (New England Biolabs). The presence of the Ttc21b^TF2 allele was genotyped by PCR amplification with the primers TF2F 5’ AGAATGATGTGCAACCTTGTTGA 3’ and TF2R 5’ TTATCTGGCTCACGGTCTCC 3’ followed by restriction digest with NmuCI (ThermoFisher Scientific). The following previously described mouse strains were used in this study: Wnt1-Cre2 [Tg(Wnt1-cre)2Sor] (Lewis et al., 2013), EIIA-Cre [Tg(EIIa-Cre)C5379Lmgd/J] (Lakso et al., 1996), mT/mG [Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J] (Muzumdar et al., 2007), SmoM2 [Gt(ROSA)26Sortm1(Smo/YFP)Amc/J] (Jeong et al., 2004), and Gli2 [Gli2^tm2.1Alj/J] (Bai and Joyner, 2001). All lines were maintained on an FVB/N background except SmoM2, which was maintained on a C57BL/6J background, and Gli2, which was maintained on an SWR/J background. Timed pregnant mice were euthanized at E7.5-E12.5. Noon on the day of the vaginal plug was considered E0.5 and embryos were staged by counting the number of somites. Analysis of wild-type embryos in Figures 1–3 and associated supplements was performed on FVB/N embryos. Control embryos were wild-type and heterozygous littermate controls of Ift122, Ttc21b, and Gli2 mutants (designated WT in the corresponding figures), or embryos bearing Wnt1-Cre2 or SmoM2 alone (designated control in Figure 9 and Figure 9—figure supplements 1 and 2). The presence or absence of exencephaly was analyzed in E10.5-E12.5 embryos. Mutant and transgenic embryos were processed in parallel with littermate controls.

Embryos were dissected in ice-cold PBS and fixed overnight at 4°C in 4–8% paraformaldehyde (PFA, Electron Microscopy Sciences) or Dent’s fixative (4:1 methanol:DMSO). Embryos fixed in Dent’s fixative were rehydrated in successive 30 min washes of 75:25, 50:50, and 25:75 methanol:PBS at room temperature (RT). Rehydrated embryos were then washed 3 × 30 min in PBS + 0.1% TritonX100 (PBTriton) at RT. Embryos were then incubated in blocking solution (PBS + 3% BSA, 0.1% TritonX100) for 1 hr at room temperature. Embryos were then incubated in staining solution (PBS + 1.5% BSA, 0.1% TritonX100) containing primary antibodies overnight at 4°C. Embryos were then washed 3 × 30 min in PBTriton and incubated in staining solution containing Alexa Fluor conjugated secondary antibodies (1:500, ThermoFisher) for 1 hr at room temperature. Embryos were subsequently washed 3 × 30 min in PBTriton at RT and stored in PBTriton at 4°C until imaging. Antibodies used for embryos fixed in Dent's fixative were: rat anti-ZO-1 (DSHB R26.4C, 1:100), rabbit anti-Arl13b (Caspary et al., 2007) (1:1000), rabbit anti-phosphohistone H3 (Upstate 06–570, 1:1000), and mouse anti-β-catenin (BD Biosciences 610153, 1:300). Antibodies used for embryos fixed in 4% PFA were rabbit anti-laminin (Sigma L9393, 1:1000), rabbit anti-N-cadherin (Cell Signaling Technology D4R1H, 1:500), chicken anti-GFP (abcam ab13970, 1:1000), mouse anti-β-catenin (BD Biosciences 610153, 1:300), mouse anti-Nkx6.1 (DSHB F55A10, 1:50), and rabbit anti-FoxA2 (abcam ab108422, 1:1000). Embryos fixed in 8% PFA were stained with rabbit anti-diphosphomyosin regulatory light chain antibody (Cell Signaling Technology 3674, 1:100). Alexa 546-conjugated phalloidin (Molecular Probes), and DAPI (ThermoFisher) were used as counterstains.

Embryos were dissected and fixed in 4% PFA for 2 hr at room temperature and then washed 5 × 30 min in PBTriton. Embryos were then transferred into 15% sucrose for 30 min and subsequently into 30% sucrose overnight at 4°C. Embryos were then placed anterior down in a cryoblock in OCT (Tissue-Tek) and frozen on dry ice. Embryos were stored at −80°C until sectioning. Embryos were sectioned on a cryostat (Leica) from anterior to posterior in 14 μm sections, with sections adsorbed onto Superfrost slides (Fisher). Cryosections from the midbrain/hindbrain region were then washed 3 × 15 min in PBTriton at RT, blocked for 30 min in blocking solution (see above), stained for 30 min with primary antibodies as above, washed 3 × 15 min in PBTriton, incubated with secondary antibodies, and washed 3 × 15 min in PBTriton. Stained sections were then mounted under a coverglass in fluorescence mounting media (Dako).

For whole-mount confocal analysis, stained embryos were mounted dorsal side down in PBTriton in Attofluor cell chambers (ThermoFisher A7816), using a small fragment of broken coverglass with small dabs of vacuum grease (Dow Corning) to mount the embryo on a #1.5 coverglass (Dow Corning). Embryos were then imaged by inverted confocal microscopy on either a Zeiss LSM700 equipped with a Plan-NeoFluar 40x/1.3 oil immersion objective, or a Leica SP8 equipped with a HC PL Apo 40x/1.3 oil immersion objective. Images were captured by tile-based acquisition of contiguous z-stacks of 50–150 μm depth with 0.9–1.2 μm optical slices and 0.3–0.5 μm z-steps. Tiled images were computationally stitched together with 10% overlap per tile using Zen (Zeiss) or LAS-X (Leica) software, resulting in visible seams in some images. Maximum-intensity projections of the entire z depth were created for analysis in the same software. For confocal imaging of cryosections, slides were imaged on an inverted Zeiss LSM700 equipped with a Plan-Apochromat 20x/0.8 air objective. Z-stacks of 10–14 μm depth were imaged with 1.8–2.0 μm optical slices and 1.0–1.2 μm z-steps. For bright-field imaging, embryos were imaged in PBTriton on a Zeiss Stemi 508 stereomicroscope equipped with a Canon EOS DSLR camera and EOS Utility software (Canon).

Apical area was measured in 100 μm x 100 μm regions in maximum-intensity projections of tiled images, either at the midline or in a pair of regions on either side of the midline, approximately midway between the midline and the lateral extent of the neural plate and midway between the pre-otic sulcus and the cranial flexure. For the analysis of cell area throughout the mediolateral axis, a continuous series of 100 μm (anterior-posterior) x 20 μm (mediolateral) regions from the midline to the lateral edge were analyzed. Cells contained entirely within these regions were segmented using SeedWater Segmenter software (Mashburn et al., 2012). Cell areas were quantified and area maps were generated using the MorphoLibJ plugin (Legland et al., 2016) in the FIJI redistribution of ImageJ (Schindelin et al., 2012; Schneider et al., 2012). Cell orientation was evaluated in the same regions using SEGGA software (Farrell et al., 2017). Cells were assigned a mediolateral orientation if they were oriented at 0–45° with respect to the mediolateral axis or an anterior-posterior orientation if they were oriented at 45–90°. Cell height was measured in cryosections by drawing a perpendicular line in FIJI from the apical to the basal surface between two apparent cell edges using phalloidin and laminin or β-catenin (Figure 3C–E), or in XZ-reconstructions of embryos stained with phalloidin (Figure 8G and H). The ratio of the apical span to the basal span of the tissue was calculated by manually drawing segmented lines in FIJI from one lateral extreme of the neural plate to the other in cryosections.

Three-dimensional lateral cell shapes were analyzed using the 3D Project tool in FIJI in cells labeled by EIIA-Cre-driven mosaic recombination of the mT/mG locus, which frequently resulted in individually labeled cells. Cells throughout the lateral midbrain region were analyzed and were manually assigned to shape categories based on examination of their apical, mid-, and basal cross-sectional areas. Cells were considered apically constricted if their apical surface was smaller than their basal surface. EIIA-Cre produced little to no labeling in the midline, and midline cell 3D analysis was performed by manual segmentation of cells labeled with β-catenin using ITK-SNAP software (Yushkevich et al., 2006). Comparison of apical and basal areas of midline cells was performed in FIJI.

For analysis of cell proliferation, the percentage of phosphohistone H3-positive cells was calculated in contiguous 100 μm x 100 μm regions along the mediolateral axis (Figure 4—figure supplement 5) or in a single 100 μm x 100 μm region midway between the midline and the lateral edge of the tissue (Figure 9—figure supplement 1). Mesenchymal cell density was calculated by counting the number of individual DAPI-labeled nuclei in a 50 μm x 100 μm region in transverse sections.

The number and angle of multicellular F-actin and phosphomyosin cables were analyzed manually using FIJI in a pair of 100 μm x 100 μm lateral regions on either side of the midline in each embryo. A cable was defined as three or more consecutive edges of high-intensity signal with no gap or diminishment along its length. Apical F-actin and phosphomyosin intensity were analyzed in 50 cells in a 50 μm x 50 μm lateral region in embryos stained for phalloidin (F-actin) and phosphorylated myosin II by calculating the mean intensity of a line drawn along the entire apical cortex of each cell. All intensity quantifications were performed on unprocessed maximum intensity projections.

Statistical analyses and graph generation were performed in Prism software (Graphpad) or with the circular plugin (Agostinelli and Lund, 2017) in the R software package (R Development Core Team, 2020). All results are reported as mean ± standard deviation (SD). Summary significance levels are as follows: ***p<0.001, **p<0.01, *p<0.05. Statistical tests were Welch's t-test, which does not assume equal SDs between conditions, the Kolmogorov-Smirnov test for comparing distributions, the standard one-way ANOVA with Tukey’s multiple comparisons, which was used when the variance between replicates was expected to come only from measurement error, the Brown-Forsythe and Welch one-way ANOVA using Dunnett’s T3 multiple comparisons test, which does not assume equal SDs between conditions, the two-way ANOVA with Sidak’s multiple comparisons for comparing multiple conditions, and the Watson nonparametric two-sample test for homogeneity for examining circular distributions. Details of the statistical tests, n values, and p values for each experiment can be found in Supplementary file 1. Figures were assembled using Photoshop and Illustrator (Adobe). For display purposes, some plots did not show cells outside the x-axis range, which were generally <2% of cells, except in Figure 5D (<4% of cells excluded) and Figure 9K (<8% of cells excluded). All cells were included in the statistical analysis. Each embryo was considered a biological replicate. Formal power analyses were not conducted. For mutant analyses, in which mutant embryos were compared with stage-matched littermate controls, an n of three to five embryos per genotype was targeted.