Planar polarization of cilia in the zebrafish floor-plate involves Par3-mediated posterior localization of highly motile basal bodies

Cilia are conserved microtubule-based organelles nucleated from a modified centriole called the basal body (BB). Motile cilia generate oriented fluid-propelling forces, whereas primary cilia receive and transduce mechanical and/or chemical signals (Goetz and Anderson, 2010; Wallingford, 2010). To achieve their functions, cilia often display an asymmetric localization within the apical cell surface, a form of planar cell polarity (PCP) called ‘translational polarity’, which involves BB off-centering.

In many ciliated vertebrate tissues, such as the mouse cochlea and ependyma, the laterality organ of mouse and zebrafish, and the zebrafish floor-plate (FP), cilium polarity requires PCP proteins, such as Van Gogh like 2 (Vangl2), Frizzled (Fz3/6), Cadherin EGF LAG seven-pass G-type receptors (Celsr1-3) and Dishevelled 1, 2 and 3 (Dvl1-3). These proteins localize asymmetrically and are required for proper cilia/BB positioning (Montcouquiol et al., 2003; Borovina et al., 2010; Mirzadeh et al., 2010; Song et al., 2010; Boutin et al., 2012). However, the mechanisms of BB positioning downstream of PCP proteins remain poorly understood. Translational polarity requires non-muscle myosin II in murine ependymal multiciliated cells (Hirota et al., 2010), Rac1 in monociliated cells of the mouse node and cochlea (Grimsley-Myers et al., 2009; Hashimoto et al., 2010) and G protein signaling in mouse cochlear hair cells (Ezan et al., 2013; Tarchini et al., 2013). Some ciliary proteins are themselves also involved in cilia planar polarization (Ross et al., 2005; Jones et al., 2008; Mirzadeh et al., 2010; Mahuzier et al., 2012; Ohata et al., 2015). However, the link between these different actors and how they impact BB movements is unclear.

Understanding the mechanisms of cilium polarization would greatly benefit from a dynamic analysis of BB movements. So far, because of technical limitations, live imaging of cilium polarization has been performed only in cochlear explants in which confined Brownian motion of BB was observed (Lepelletier et al., 2013) and in the mouse node (Hashimoto et al., 2010) and ependyma (Hirota et al., 2010), with limited temporal resolution. In this study, we used the zebrafish FP to investigate the dynamics of the polarization process in live embryos. The FP is a simple monociliated epithelium, the posterior-positioned motile cilia of which propel the embryonic cerebrospinal fluid (Kramer-Zucker et al., 2005; Borovina et al., 2010; Thouvenin et al., 2020).

We show that planar polarization of BBs and their associated cilia is progressive during somitogenesis and is accompanied by a change in the behavior of the BBs, which are highly motile at early stages and spend an increasing amount of time in contact with the posterior membrane as development proceeds. We found that BBs always contacted membranes at Par3-enriched apical junctions. Par3 became enriched at the posterior apical side of FP cells before BB polarization. Par3 depletion and overexpression of either wild-type (wt) or dominant-negative forms of Par3 disrupted FP polarization. Furthermore, Par3 distribution along apical junctions was disrupted in a vangl2 mutant. Thus, we propose that a major role of the PCP pathway in the FP is to drive Par3 asymmetric localization, which, in turn, mediates BB posterior positioning.

Posterior positioning of the BB in the zebrafish FP is visible as soon as 18 h post-fertilization (hpf) (Mahuzier et al., 2012) and is maintained at least until 72 hpf (Mathewson et al., 2019). From 24 hpf onward, it is instrumental in propelling the cerebrospinal fluid (Borovina et al., 2010; Fame et al., 2016). During late gastrulation stages, ectodermal cell centrioles are already slightly posterior (Sepich et al., 2011).

To define the time-course of FP cell polarization, we assessed the BB position along the AP axis on fixed embryos from the six somite (6s) to the 26s stage (Fig. 1B,C). We focused all our studies on a single stripe of cells, the medial FP. For each cell, we defined a BB polarization index (p.i.; Fig. 1A). BBs were posteriorly biased at 6s and the polarization state did not significantly change until 10s. After 10s, there was a progressive increase in polarization, mostly the result of an increase in the percentage of cells with a BB in contact with the posterior membrane, with a disappearance of anterior BBs and a reduction of median BBs. The polarization state of the FP did not further increase between 18s and 26s (Fig. 1C). We did not detect any gradient of polarization index along the AP axis (Fig. S1A), and single nonpolarized cells were often intermingled among their polarized neighbors (Fig. S1B), arguing against a polarization wave.

This static view of BB polarization could reflect very different individual BB behaviors: e.g. a slow, regular BB movement towards the posterior or a unique fast movement at a specific time-point. In order to understand the dynamics of the process, we turned to live imaging and followed BB movements within the apical surface of individual FP cells at different developmental stages (4-21s), using Centrin-GFP (Pouthas et al., 2008) to label BBs and membrane Cherry (mCherry) (Megason, 2009) to label membranes. BBs were surprisingly motile within the apical surface (Fig. 2A-D; Movies 1-4), moving both anteriorly and posteriorly (Fig. 6A, wt; Fig. S1D, first column), with a clear BB movement orientation bias along the AP axis (70% of BB movements) (Fig. 6A,B, wt). At early stages, BBs contacted the anterior and posterior membranes several times per hour (Fig. 6A,B, wt).

Cell deformations along the AP axis were more important at early stages (4-10s) (Fig. 2A,B) than at later stages of development (14-21s) (Fig. 2C,D), probably because of convergence-extension movements, but many long BB movements did not correlate with cell deformation (see, for example, the two long movements around 55 and 75 min in Fig. 2A). This suggests that BBs are actively moving within FP cell apical surfaces. One possible explanation for the presence of unpolarized cells is that they had undergone mitosis immediately before. However, mitoses were rare in FP cells at early stages (6/79 cells, 9 embryos at 4-8s) and absent at later stages of development (118 cells from 15 embryos at 13-21s). Thus, the impact of cell shape changes and mitosis on FP polarization is likely very small.

Therefore, during early polarization stages, BBs actively move along the AP axis of FP cells and make multiple contacts with the anterior and posterior membranes.

In order to characterize BB behavioral changes during development, we determined the percentage of time that BBs spent in contact with the posterior membrane (Fig. 2E). During early stages, BBs spent, on average, 44% of their time in contact with the posterior membrane, versus more than 70% during later stages (13-21s). This was largely due to an increase in the number of cells in which the BB did not detach from the posterior membrane (Fig. 2C). We refer to this situation as ‘posteriorly docked BB’. During early stages (4-8s), we did not observe any cell with posteriorly docked BB (41 cells, five embryos), whereas they made up 34% of the FP cell population during the 13-17s stages (13/38 cells, six embryos) and almost half (46%) the FP population during later stages (17-21s, 27/59 cells, seven embryos). There was also a decrease in the frequency of BB direction changes, as well as an increase in the mean duration of BB/posterior membrane contact events and mean polarization index, suggesting that, as development proceeds, BB movements become confined posteriorly (Fig. S1D, first line). Posteriorly docked BBs made a significant contribution to these behavioral changes. In order to determine whether changes in the behavior of nonposteriorly docked BBs contributed to the increase in FP polarization during somitogenesis, we quantified the same parameters, but taking into account only these motile BBs (Fig. S1D, second line): the same trend in BB behavior change was observed.

To further characterize the behavior of nonposteriorly docked BBs, we quantified the frequency of contact events between the BB and either the anterior or posterior membrane (Fig. 2F,G, respectively). Posterior contacts were more frequent compared with anterior ones even at 4-8s (compare Fig. 2F versus Fig. 2G), confirming that FP cells already had a posterior polarization bias by these early stages. Contacts with the anterior membrane were frequently observed during early stages (50% of BBs made at least one anterior contact per hour, see, for example, t=70′ in Fig. 2B), but almost never at later stages (3/57 cells). Contact frequency with the posterior membrane was also significantly higher during earlier stages (1.3 contact/h) than during later stages (0.8 contact/h, Fig. 2G). This reduction in the number of contacts could be the result of an increase in their duration (Fig. S1D, plot second column, second line) and of a reduction in BB speed. Indeed, we found that BBs moved faster during earlier stages (Fig. S1C, 0.2 µm/min at 4-8s versus 0.1 µm/min at 13-21s). Thus, the observed changes in FP polarization are explained both by an increase in the posteriorly docked BB population and by behavioral changes in other BBs.

Live imaging revealed the presence of membrane digitations extending between the BB and anterior or posterior apical membranes (‘ant/post membranes’ hereafter) (Fig. 3; Movies 5, 6). Digitations always formed at the same apicobasal level as the BB (i.e. ∼1 µm under the apical surface). During early stages, in the Δt2 min and Δt5 min movies, we detected such digitations in 44% of FP cells (26/59 cells, nine embryos), most of which were linking the posterior membrane and the BB (83%, 45/54 digitations, Fig. 3A, upper row, white arrows; Movie 5), although digitations from the anterior membrane were also seen, with a similar lifetime (Fig. 3A, second row; Fig. 3B,C; Movie 6). Posterior digitations (i.e. originating from the posterior membrane) were followed by a posterior-directed BB movement in 67% of cases (26/39), whereas anterior digitations were followed by a BB anterior movement in only 22% of cases (2/9) (Fig. 3D). Membrane digitations were rarely seen during later stages (after 14s, 9/40 cells, ten embryos).

In order to better characterize these digitations, we imaged the floor-plate of wt embryos every 10 s (Δt10 s movies, Fig. S2A-C). In these movies, digitations were seen in 80% of FP cells (25/31 cells from 17 embryos). Almost all of them extended from ant/post membranes (92%, 88/95, Fig. S2A) and pointed towards the BB (95%, 90/95 digitations, 25 cells from 14 embryos), whereas 45% touched the BB. As with our Δt2 min and Δt5 min movies, there was a poor correlation between digitation position (anterior or posterior membrane) and BB movements (Fig. S2C), suggesting that these digitations are a consequence, rather than a cause, of the forces exerted on BBs.

Given that centrosomes are the main microtubule-organizing centers of animal cells and their positioning in many systems depends on microtubules, we investigated microtubule dynamics within the apical surface of FP cells, using GFP-tagged EB3, a microtubule plus end-binding protein. Live imaging of moving EB3-positive comet-like structures (EB3 comets) revealed highly dynamic microtubules originating from the centrosome/BB and directed to apical junctions (Fig. S2D; Movie 7). The time interval between an EB3 comet coming from the BB touching a spot at the ant/post membrane and the occurrence of either a digitation at this spot or a BB movement towards it was very short: 10 s in 50% of cases and less than 1 min in 95% of cases (Fig. S2E). These observations suggest that the mechanical forces responsible for back-and-forth BB movements are mediated by microtubules.

Overall, our dynamic analysis reveals a highly motile behavior of BBs in FP cells during early somite stages. As somitogenesis proceeds, BB motility decreases. BBs progressively stop shuttling from anterior to posterior cell junctions and their contacts with the posterior membrane last longer. We also uncover membrane digitations forming at precise spots of ant/post apical membranes and show that these spots are linked to BBs by dynamic microtubules. These results suggest the existence of a molecular complex at precise spots of apical ant/post membranes, which is able to exert mechanical forces on the BB via microtubules, and the distribution of which becomes biased to the posterior side of each FP cell during development.

Par3 is a conserved polarity protein that positions at apical junctions of neuroepithelial cells during zebrafish neurulation and is important for lumen formation (Hong et al., 2010; Buckley et al., 2013). Moreover, Par3 has been shown to modulate centrosome positioning in Drosophila (Inaba et al., 2015; Jiang et al., 2015), making it a good candidate for BB attraction in the zebrafish FP. In order to test this hypothesis, we first assessed Par3 localization by immunostaining (Fig. 4A). At the 14s stage, Par3 localized at apical junctions of FP cells (Fig. 4A). Par3 local enrichments or ‘patches’ were detected on ant/post membranes and in close contact with posteriorly docked BBs (Fig. 4A, white arrows). This distribution was confirmed using a phosphorylated-Par3 antibody (BazP1085) (Krahn et al., 2009) (Fig. S3A). Par3 patches were also present in FP cells in which the BB was not yet in contact with the posterior membrane (Fig. 4A; Fig. S3A, right panels), showing that this enrichment precedes stable BB/posterior membrane contact.

In order to test whether Par3 is asymmetrically enriched in FP cells, we mosaically expressed Par3-RFP (Alexandre et al., 2010) and Centrin-GFP. Quantification of Par3 expression showed that, among fully polarized (p.i.=1) individual Par3-RFP-expressing FP cells, almost all had a Par3-RFP post/ant ratio greater than 1 (Fig. 4B). To determine whether Par3 posterior enrichment preceded BB/posterior docking, we imaged BB movements and quantified the Par3-RFP posterior/anterior ratio at each time-point; Par3-RFP was enriched posteriorly before BB/posterior docking (12/14 cells, 12 embryos) (Fig. 4C; Movie 8). In contrast, BBs of FP cells with weak or no posterior Par3 enrichment remained unpolarized [either making no contact (2/5 cells, five embryos) or unstable contacts (3/5 cells, five embryos) with the posterior membrane] (Fig. 4D; Movie 9).

Thus, Par3 forms patches at FP apical ant/post membranes and BBs dock posteriorly at the level of Par3 patches. In addition, Par3 is enriched posteriorly before BB/posterior membrane contact. Together, these results strongly suggest that Par3 is a key player in BB posterior positioning.

During the second half of somitogenesis, Par3 formed a continuous belt at apical junctions of FP cells, although it was locally enriched, forming patches that associated with BBs, as described above. In contrast, at the 4-8s stages, Par3 formed discrete patches at FP apical ant/post membranes, but not at lateral membranes (i.e. apical membranes of medial FP cells in contact with lateral FP cells). These patches were aligned with the AP axis of the embryo (Fig. 4E, white arrows). Strikingly, BB/membrane contacts occurred exclusively at these patches (58 cells from 18 embryos) (Fig. 4F; Movies 10, 11).

In 40% of these cells (23/58), Par3 patches ‘stretched’ towards the BB (Fig. 4F, yellow arrows) and covered a membrane digitation (Fig. 4F, t=0′ and t=64′; Movie 11). Of these digitations, 92% occurred at a Par3 patch (36/39 digitations from 23 cells and 14 embryos). The presence of membrane digitations and their overlap with Par3 patches point to the existence of mechanical forces between BBs and membranes at Par3 patches and suggests that Par3 could be required for local force generation. This role of Par3 could be more general, because, after cytokinesis in dividing FP cells, the centrosomes also always rapidly (within 10 min) moved back towards Par3 patches (9/9 cells from nine embryos, Fig. S3B; Movie 12).

To test whether Par3 is required for posterior BB positioning in the FP, we analyzed FP polarization in the MZPard3ab mutant devoid of maternal and zygotic Pard3ab protein (Blasky et al., 2014). MZpard3ab^−/− embryos displayed convergence-extension and cell intercalation defects with variable severity. In severely affected embryos (19/55, 34%), these defects precluded the analysis of FP polarity. In mildly affected embryos, in which the medial FP could be unambiguously identified, the BB polarization index was significantly reduced to a moderate extent (Fig. 5A; Fig. S4A). A reduction in BB polarization was also observed in Mpard3ab^+/− siblings (heterozygotes devoid of maternal Pard3ab; Fig. 5A; Fig. S4A), suggesting that maternal stores of Par3 are important for BB polarization, as confirmed by the strong reduction in Par3 protein level (Fig. S4C).

In order to obtain further insight into the mechanisms by which Par3 impacts BB positioning, we expressed a dominant-negative form of Par3, Par3Δ6, which localizes less efficiently to apical junctions and more to microtubules (von Trotha et al., 2006) and induces a neural ventricle defect similar to that induced by the Pard6g mutant (Munson et al., 2008; Buckley et al., 2013). Injection of Par3Δ6-GFP mRNA at the one-cell stage led to a significant reduction in BB polarization, often accompanied by FP cell intercalation defects along the midline (Fig. 5B; Fig. S4B). Par3Δ6-GFP was then injected into one cell at the 16- to 32-cell stage to obtain mosaic expression. In isolated FP cells expressing high Par3Δ6-GFP levels, the BB always colocalized with the most intense GFP patch, either along apical junctions or in the middle of the apical surface (41 cells, 11 embryos) (Fig. 5C). This strongly suggests that Par3Δ6-GFP foci, similar to wt Par3 foci, are able to attract the BB. However, unlike wt Par3, it does not localize efficiently to the posterior apical junctions (17/41) and, thus, its overexpression perturbs BB polarization.

To confirm that Par3 mislocalization along apical junctions is sufficient to perturb BB polarization, we overexpressed the wt form of Par3 (150 pg mRNA per embryo instead of 50 pg in previous injections). As expected, Par3-RFP still localized to FP apical junctions and did not affect apicobasal polarity, but did significantly disrupt BB posterior positioning (Fig. 5D). Furthermore, Par3-RFP-negative cells adjacent to Par3-RFP overexpressing cells polarized normally, showing that the Par3 overexpression effect is cell autonomous (391 cells, 20 embryos, P=0.19, Wilcoxon test).

These results show that Par3 is necessary for BB posterior positioning in the FP and that it acts by attracting or capturing the BB at posterior apical junctions.

Vangl2, a PCP protein, is involved in zebrafish FP PCP (Borovina et al., 2010), but the mechanisms linking Vangl2 to BB posterior positioning are unknown. Thus, we analyzed the dynamics of FP polarization in the vangl2^m209 mutant (Solnica-Krezel et al., 1996). At 18s, the BB of vangl2^m209/m209 FP cells was mispositioned at the center of the apical surface, which could suggest either a loss of BB motility or misoriented BB movements (Fig. 7A). Live imaging of vangl2^m209/m209 FP revealed several BB behavior changes in vangl2^m209/m209; surprisingly, during early stages in vangl2^m209/m209 FP cells, BBs were still motile and even faster than wild type (Fig. 6D). Moreover, BB movements were still biased along the AP axis (60% of movements; Fig. 6B), but BBs made more lateral movements and 10% fewer AP movements compared with wild type (Fig. 6A,B; Movie 13). The length of BB movements, which was shorter in the lateral than in the AP direction in controls, was equivalent in all directions in vangl2^m209/m209 mutants (Fig. 6A).

Despite the preserved AP bias in BB movements, vangl2 mutants showed a striking loss of AP bias in BB/membrane contacts. The overall proportion of BB movements resulting in BB/membrane contacts was the same as in wild type (16% of movements), but the positions of these contacts were very different: in wild type, most contacts occurred with the posterior membrane (75%) and almost none with the lateral membranes (3%), whereas, in vangl2^m209/m209, BB contacts occurred equally with anterior, posterior or lateral membranes (∼33% each) (Fig. 6D, middle barplot). In addition, vangl2^m209/m209 BBs spent less time in contact with membranes (Fig. 6D, right plot).

At later stages, almost half of the BBs remained in contact with the posterior membrane in wild type (Fig. 2E) (posteriorly docked BBs). In vangl2^m209/m209 embryos, the vast majority of BBs did not stably dock at any membrane. Instead, BBs remained at the center of the apical surface. The few BB movements of wt cells as well as the many BB movements seen in vangl2^m209/m209 cells were much smaller than during early stages and had no preferential orientation (Fig. 6A). BB/membrane contacts were half as frequent than during early stages, both in wild type and vangl2^m209/m209, but we could still detect a significant difference in their position between wild type and vangl2^m209/m209, with more lateral and anterior contacts in vangl2^m209/m209 (Fig. 6E, middle barplot).

In conclusion, in vangl2 mutants, BBs still show fast, oriented movements along the AP axis, but make aberrant membrane contacts with lateral membranes during early stages, and fail to attach to the posterior membrane during later stages.

We then investigated whether these behavioral defects could be due to Par3 defects. Par3 localized at apical junctions in vangl2^m209/m209, but there was a significant difference in the number and prominence of Par3 patches (prominence: height of Par3 fluorescence peak relative to the highest and nearest local fluorescence minimum; Fig. S4D). In wild type, 90% of FP cells had at least a major Par3 patch (Fig. 7B, yellow arrows), with 39% of cells also having smaller secondary patches (Fig. 7C), whereas, in vangl2^m209/m209 embryos, the number of FP cells with at least one Par3 patch was unchanged (∼90% of cells), but the number of cells with more than one patch was increased (54% of cells). In addition, the prominence of Par3 patches was decreased in vangl2^m209/m209 embryos (Fig. 7D). Thus, Par3 forms more numerous, smaller patches in vangl2 mutants, showing a role for Vangl2 in Par3 clustering.

To further analyze a potential link between abnormal BB behavior and Par3 patches mislocalization in vangl2^m209/m209 embryos, we made time-lapse movies of mutant embryos mosaically injected with Par3-RFP (seven embryos, 17 cells, see two examples in Fig. 7F and Movies 13 and 14). In vangl2 mutants, BBs still contacted the membrane only at Par3 patches (Fig. 7F), suggesting that Vangl2 did not directly affect the ability of Par3 patches to attract BBs. However, the distribution of Par3 patches was very different. Early vangl2^m209/m209 embryos displayed more cells with lateral Par3 patches compared with wild type (70% versus 20%; Fig. 7E). In addition, they had more cells with an anterior Par3 patch (82% versus 67%) and less cells with a posterior patch (65% versus 87%) compared with wild type.

These results strongly suggest that, in vangl2^m209/m209 embryos, abnormal BB behavior and polarization failure are due to the fragmentation and mispositioning of Par3 patches along the apical junctions of FP cells.

In this study, we analyzed the dynamics of BB posterior positioning in the embryonic zebrafish FP. We show that, during early somitogenesis, BBs are highly motile and able to contact apical junctions several times per hour. As somitogenesis proceeds, BBs settle down posteriorly at junctions enriched in Par3, and we show that Par3 enrichment is essential for BB posterior localization. In the PCP mutant vangl2, the poorly oriented movements of BBs correlated with Par3 patch fragmentation and lateral spreading (Fig. 8). This led us to propose a model in which Par3 posterior enrichment, downstream of the PCP pathway, increases BB attraction forces from the posterior side, which eventually results in its docking at the posterior Par3 patch (Fig. 8).

Our live-imaging studies revealed that BBs make rapid back and forth movements within the FP cell apical surface. This contrasts with mouse cochlear hair cells, in which live imaging of explants over a longer time scale suggested very slow (10-50 nm/h) and regular BBs movements to the lateral cortex (Lepelletier et al., 2013). Our live imaging also uncovered a clear AP bias in BB movements, indicating that the orientation of the forces underlying these movements is biased along the polarization axis from early developmental stages onwards. It will be interesting to investigate whether this high motility of the BB is conserved in other tissues undergoing cilia translational polarity. Strikingly, live imaging also revealed that, in vangl2^m209/m209 embryos, BBs are still highly motile. Major differences between these BBs and wt BBs are that the former make many contacts with lateral membranes and that their contacts are shorter. This suggests that there are cues at the membrane organizing/driving BB movements, and that these cues are disorganized in PCP mutants.

FP polarization is concomitant with cilia growth and most BBs are already off-center by mid-somitogenesis (Fig. 1), suggesting that a fully grown cilium is not required for the process. This is in line with the observation that the MZoval/ift88 zebrafish mutant, devoid of cilia, displays correct BB planar polarity (Borovina and Ciruna, 2013). The polarization process ends before lumen opening (24 hpf), which also excludes a function of cilia-mediated directional fluid flow as a mechanical instructive cue, as shown in ependymal cells (Guirao et al., 2010). However, several proteins of the cilium base are required for PCP in both the zebrafish FP and the cochlea (Jones et al., 2008; Mahuzier et al., 2012; May-Simera et al., 2015). This may be explained by the ability of these proteins to modulate PCP protein stability and/or subcellular localization, as shown for Rpgrip1l on Dvl1 and Dvl2 (Mahuzier et al., 2012), BBS8 and Ift20 on Vangl2 (May-Simera et al., 2015), and by microtubule-nucleating functions at BB subdistal appendages (Kodani et al., 2013).

The high motility of the BBs was unexpected, given that they anchor a growing cilium to the membrane. A possibility is that the growing cilium is still internal during early stages and only protrudes outside when anchored to the posterior junction belt. Such inside cilia have been described recently in other systems (Hong et al., 2010; Insinna et al., 2019; Matsumoto et al., 2019). Alternatively, fluidity of the apical membrane may allow these ciliary movements before the appearance of the tight junction belt and the apical cortical actin meshwork at the neural rod stage (Ciruna et al., 2006). Finally, ciliary transition zone (TZ) and BB rootlet maturation may not be complete while cilia are still growing. Such an adhesion/anchoring role for TZ proteins has been described in Caenorhabditis elegans (Schouteden et al., 2015).

Another open question is how Par3-mediated BB polarity impacts the function of ciliated tissues. BB posterior positioning could allow the coordinating of cell-intrinsic PCP with tissue polarity. In the mouse node, it has been proposed that membrane bending at the level of apical junctions facilitates oriented cilia tilting and beating (Hashimoto et al., 2010). Such membrane bending has not been documented in zebrafish. However, in MZvangl2^−/− embryos, in which BBs remain at the center of the apical surface, ciliary beating is not properly oriented (Borovina et al., 2010), which suggests that BB posterior position favors posterior tilting of motile cilia through unknown mechanisms.

We proposed the asymmetric maturation of cell junctions as a possible cause for posterior BB positioning. Accordingly, Par3 accumulated in patches at the posterior apical junctions of FP cells before BB posterior docking, and perturbation of Par3 localization affected BB polarization. Interestingly, in Drosophila early gastrula ectoderm, aPKC loss of function leads to Par3 accumulation as discrete patches that recruit centrosomes (Jiang et al., 2015). Centrosome docking at Par3 patches has also been observed in Drosophila germ stem cells and is crucial for oriented division (Inaba et al., 2015). Together with these published data, our results strongly suggest that Par3 may be broadly involved in recruitment of centrosomes/BBs in different systems.

Our live-imaging data also strongly suggest that Par3 is involved in generating mechanical forces on the BB to pull it toward the membrane. First, BB/membrane contacts occur exclusively at Par3 patches. Second, membrane digitations support the existence of mechanical forces between Par3 patches and BBs. Third, the predominance of posterior digitations over anterior digitations (Fig. 3B) suggests that more force is exerted on the BB from the posterior side, where Par3 is enriched. Such membrane digitations have been previously observed during cell division in the C. elegans zygote (Redemann et al., 2010), in the Ciona intestinalis embryo ectoderm (Negishi et al., 2016) and in rare cases at the immunological synapse (Yi et al., 2013). In all cases, the existence of pulling forces between the centrosome and the membrane has been proposed.

We propose that digitations are a consequence of mechanical forces between the BB and Par3 patches rather than a driver of BB movement. First, even in Δt10 sec movies, only half of BB movements were associated with digitations and, second, there was a poor correlation between the location of a digitation and the direction of BB movement. One can wonder why digitation formation occurs for some BB movements and not others. A possibility is that digitation formation depends on both mechanical forces pulling the membrane and cortical stiffness.

Our results suggest that mechanical forces between Par3 and the BB could be exerted by microtubules. Dynamic microtubules link the BB and Par3 patches before BB movements and/or digitation formation. Microtubules are required for digitation formation in several systems (Redemann et al., 2010; Negishi et al., 2016) and, thus, are likely to transmit mechanical forces between BB and Par3 patches that lead to BB movements and/or membrane digitations.

An interesting further question concerns the mechanisms that regulate microtubule dynamics to lead to BB movements. BB movements towards Par3 patches could involve local microtubule depolymerization at the patch, coupled to microtubule anchoring by dynein, as proposed for the immunological synapse (‘end-on-capture-shrinkage’ mechanism) (Yi et al., 2013). Indeed, Par3 can interact with Dynein (Schmoranzer et al., 2009) and with microtubules, directly (Chen et al., 2013) or indirectly (Benton and St Johnston, 2003). Consistent with a role for cortical dynein, a recent study in mouse ependymal cells showed a role for cortical dynein in the off-centering of BB clusters (Takagishi et al., 2020). Moreover, dynein cortical localization depends on Daple, which is a known partner of Par3.

Par3 could also regulate microtubule depolymerization via Rac1, which mediates Par3 function in the mouse cochlea (Landin Malt et al., 2019). In different systems, Par3 regulates the local activity of Rac via the RacGEFs Tiam1 and Trio (Nishimura et al., 2005; Zhang and Macara, 2006; Matsuzawa et al., 2016). Par3 can increase microtubule catastrophe rate by inhibiting Trio in neural crest cells (Moore et al., 2013), and Rac1 can regulate microtubule dynamics via CLIP-170 or stathmin in other systems (Fukata et al., 2002; Wittmann et al., 2004).

Interestingly, microtubules also actively maintain BB polarity at later stages in the FP, but whether microtubules act as mechanical-force generators or as tracks for PCP protein transport and asymmetric localization is unknown: indeed, Vangl2 asymmetric localization depends on microtubules (Mathewson et al., 2019). This role of microtubules in PCP protein localization appears widely conserved (Shimada et al., 2006; Vladar et al., 2012). Furthermore, Par3 apical localization also depends on microtubules in the zebrafish embryo neural tube (Buckley et al., 2013). Thus, it might prove difficult to disentangle the different roles of microtubules in the asymmetric positioning of BBs of FP cells.

Our results from vangl2 mutants led us to propose that the role of PCP in BB posterior docking is mediated, at least in part, by Par3 localization (Fig. 8). How PCP proteins act on Par3 localization in the FP remains to be uncovered. In FP cells, Vangl2 localizes anteriorly (Davey et al., 2016); thus, the effect of Vangl2 on Par3 could be mediated by Dvl. Indeed, Vangl2 is required for asymmetric localization of Dvl in planar polarized tissues and Dvl can recruit Par3 to the posterior membrane in Drosophila sensory organ precursors (Banerjee et al., 2017). Dvl could also recruit Par3 via Daple, which colocalizes with Par3 in the mouse cochlea and can bind Dvl and Par3 in yeast two-hybrid assays (Siletti et al., 2017). Recent studies have shown that Par3 is planar polarized in several systems, such as Drosophila ommatidia (Aigouy and Le Bivic, 2016) and sensory organ precursors (Besson et al., 2015), Xenopus embryo ectoderm (Chuykin et al., 2018) and mouse cochlea (Landin Malt et al., 2019), suggesting that, in addition to its classic role in apicobasal polarization, Par3 might also be involved in PCP across species.

Finally, given that asymmetric centriole positioning is now recognized as a conserved PCP readout (Carvajal-Gonzalez et al., 2016a,b), it will be interesting to investigate whether Par3 has a conserved role in centriole/BB positioning in other species in which BB/centriole off-centering has been described and also depends on PCP proteins, such as in Clytia hemisphaerica embryos (Momose et al., 2012) or Drosophila pupal wings (Carvajal-Gonzalez et al., 2016b).

Wt and mutant zebrafish embryos were obtained by natural spawning. We used wt AB or (TL×AB) hybrid strains, vangl2^m209 mutants (Solnica-Krezel et al., 1996) (ZDB-GENO-190204-5) and pard3ab^fh305 mutants (Blasky et al., 2014) (ZDB-FISH-150901-20689). To produce embryos lacking maternal stores of Pard3ab, pard3ab^−/− females were crossed with pard3ab^+/− males, which led to 50% MZPard3ab^−/− and 50% MPard3ab^+/− offspring. To obtain early-stage embryos (4-8s), embryos were collected at 10:00 h and incubated for 9 h in a 33°C incubator. To obtain later stages (14-20s), embryos were collected at 10:00 h and incubated for 2 h in a 28°C incubator before being placed overnight in a 24°C incubator. All our experiments were performed in agreement with the European Directive 210/63/EU on the protection of animals used for scientific purposes, and the French application decree ‘Décret 2013-118’. Research by our laboratory has been approved by our local ethical committee ‘Comité d'éthique Charles Darwin’ [authorization number 2015051912122771 v7 (APAFIS#957)]. The fish facility has been approved by the French ‘Service for animal protection and health’ with approval number A-75-05-25.

mRNAs were synthesized from linearized pCS2 vectors using the mMESSAGE mMACHINE SP6 transcription kit (Ambion). The following amounts of mRNA were injected: 22 pg for Centrin-GFP, 40 pg for mCherry or membrane-GFP (Gap43-GFP), 50 pg for EB3-GFP, 50 pg for Par3-RFP live imaging, 150 pg Par3-RFP for overexpression experiments, 60 pg for Par3Δ6-GFP. mRNAs were injected at the one-cell stage for whole-embryo expression and at the 16- to 32-cell stage in a single blastomere for mosaic expression. Plasmid details are provided in Table S1.

For immunostaining, embryos were fixed in Dent fixative (80% methanol, 20% DMSO) at 25°C for 2 h, blocked in 5% goat serum (Sigma), 1% bovine serum albumin (Sigma) and 0.3% Triton (Sigma) in PBS for 1 h at room temperature and incubated overnight at 4°C with primary antibodies and 2 h at room temperature with secondary antibodies. Apical junctions were labelled with ZO1, BB were labelled with Centrin, cilia were labelled with acetylated Tubulin, total Par3 was determine using a polyclonal serum from Millipore, a phosphorylated subpopulation of Par3 enriched in patches was labelled with Baz P1035 and mbCherry were labelled with serum against DsRed. The yolk was then removed and the embryo mounted dorsal side up in Vectashield medium (Vector Laboratories) on a slide. Imaging was performed using a Leica TCS SP5 AOBS upright confocal microscope with a 63× oil lens. A list of the antibodies used is provided in Table S1.

Embryos were dechorionated manually and mounted in 0.5% low-melting agarose in E3 medium (Westerfield, 2007). Movies were recorded at the temperature of the imaging facility room (22°C) on a Leica TCS SP5 AOBS upright confocal microscope using a 63× (NA 0.9) water immersion lens. The embryos were mounted either dorsal side up (for 4-8s or after 18s, when the FP cell apical surface is large) or on their side (for 13-18s embryos, when the apical surface of FP cells is narrower and when the images are more blurred when taken from a dorsal view, possibly because of the thickness of the overlying neural tube). The anterior side of the embryo was positioned on the left and its AP axis aligned horizontally. A z-stack was acquired every 2 or 5 min (or, in some cases, every 4 min) for most analyses (Δt2 min and Δt5 min movies) and every 10 s for some movies (digitation analysis and microtubule dynamics in Fig. S2, Δt10 s) with a z-step of 0.3 µm. For each time-point, the z-stack extended from the most dorsal side of the notochord to neural cells above the FP for Δt2 min and Δt5 min movies, but was narrower for Δt10 s movies (to allow fast acquisition and reduce photobleaching and photodamage). For embryos mounted on the side, the z-stack extended through the full width of the FP. In every case, the z-stack encompassed the apical surface of FP cells with the moving BBs. For each time-point, we then made a z-projection from a 3 µm-thick substack that encompassed the apical centrioles/BB.

All bar-plots, boxplot and violin plots and statistical tests (as indicated in figure legends) were generated with R (version 3.3.2) and Rstudio (version 1.1.463). Throughout the figures, the boxplots show five summary statistics (the median, two hinges and two whiskers), and all ‘outlying’ points individually. The lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles). The upper whisker extends from the hinge to the largest value no further than 1.5 * IQR from the hinge (where IQR is the inter-quartile range, or distance between the first and third quartiles). The lower whisker extends from the hinge to the smallest value at most 1.5 * IQR of the hinge. Data beyond the end of the whiskers are called ‘outlying’ points and are plotted individually. For statistical tests, nonsignificant (ns), P>0.05; *P<0.05; **P<0.01; ***P<0.001; and ****P<0.0001.

In all our images, the AP axis (easily visualized because of the underlying notochord, the cells of which have an elongated shape orthogonal to the AP axis) is horizontal and the anterior side of the embryo is toward the left. To assess the polarization of FP cells, we used Fiji to first make a 3 µm-thick z-projection around the centrioles. We then manually measured the distance ‘a’ by drawing a line between the most posterior centriole and the posterior membrane that was parallel to the AP axis of the embryo (i.e. parallel to the horizontal axis of the image) and the distance ‘b’ by drawing a similar line, at the same level and also parallel to the AP axis, between the anterior and posterior membrane (Fig. 1A, dorsal view). We used a similar method for embryos mounted on the side (Fig. 1A, lateral view). The p.i. was then calculated as 1-(a/b). In rare cases, the ant/post membranes in our z-projection around the centrioles were blurry (Fig. 2B); however, in these cases, going slightly more basally through the z-stack allowed us to see a sharper membrane, which was usually located immediately beneath the boundary of the above blurry region.

To follow the evolution of the p.i., the distances ‘a’ and ‘b’ were measured manually at each time-frame in Fiji. These distances and the p.i. were then plotted using Python Matplotlib (Python 2.7.13) and analyzed with a custom Python script to extract relevant information, such as the frequency of contact with the posterior membrane or percentage of total time spent in contact with the posterior membrane (Fig. 2 and Fig. S1).

For automatic tracking of BB movements (Fig. 5), BBs were tracked using ImageJ TrackMate (Tinevez et al., 2017). The movements were then manually curated to only keep active BB movements and not BB movements resulting from the global shift of cells (especially during early stages, when convergence-extension movements are still important), and to indicate whether a contact with an anterior, posterior or lateral membrane occurred. The results were then processed using a custom Python script to calculate each movement length and angle relative to the horizontal axis (i.e. the AP axis of the embryo) and plotted using Python Matplotlib and R ggplot2. We defined AP movements as those with angles relative to the horizontal axis inferior to 45° [i.e. in the intervals (315-45°) and (135-225°) in Fig. 6A].

For all analyses, lateral membranes were defined as those more parallel to the horizontal axis, and ant/post membranes as those more orthogonal to the horizontal axis. The transition between lateral and ant/post membranes is evidenced by ‘Y’-shaped tricellular junctions (Fig. 4A) or as sharp turns in the membrane (Fig. 7F).

Fluorescence intensity was measured along the AP length of isolated labeled FP cells in Fiji. A custom Python script was then used to extract the first quarter (cell anterior side) and last quarter (cell posterior side) of the fluorescence intensity values to determine the area under each curve (corresponding to fluorescence intensity), calculate the posterior/anterior ratio and plot it along with the p.i. (see BB movements analysis section above).

Fluorescence intensity from immunostained embryos was measured along ant/post membranes of medial FP cells and exported to MatLab (R2018a). For each cell, the ‘findpeaks’ function was used to detect Par3 fluorescence peaks and measure their prominence (Fig. S4D), which was then normalized by the minimal Par3 fluorescence value along the junction.

Adobe Photoshop was used to assemble the figures. Fig. 1A was prepared with Microsoft PowerPoint and Fig. 8 with Adobe Illustrator.

We are grateful to the aquatic animal and cell-imaging facilities of the Institut de Biologie Paris-Seine (FR3631, Sorbonne Université, CNRS, Paris, France) for assistance. We thank Teresa Ferraro for helping with image analysis, Marie Breau for live-imaging advice, Alexis Eschstruth for technical assistance and Sophie Gournet for helping with Fig. 8. We thank Paula Alexandre for the kind gift of pCS2-Par3-RFP, Jon Clarke for pCS2-Par3Δ6-GFP, Andreas Wodarz for the BazP1085 antibody and Maximilien Furthauer for the vangl2^m209 line. We thank Nicolas David, Marie Breau and Pierre-Luc Bardet for critical reading and insightful comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.196386