Centriole planar polarity assessment in Drosophila wings

In some specialized epithelial cells, polarized settling of a centriole at the core of the ciliary basal body is essential for coordinated cilia beating (Brooks and Wallingford, 2014; Ohata and Alvarez-Buylla, 2016; Spassky and Meunier, 2017). Excellent examples are multiciliated cells lining the oviduct, brain ventricles, and the airway tract, in which coordinated arrangement between neighbouring basal bodies within the same cell is key to creating directional movement of ovules, cerebrospinal fluid flow and mucus, respectively (Brooks and Wallingford, 2014; Spassky and Meunier, 2017). Thus, deeper insights into the molecular mechanisms that control centriole positioning at the core of the ciliary basal body would have important implications for understanding the physiology and aetiology of pathological states coming from centriole mis-positioning (Bettencourt-Dias et al., 2011; Reiter and Leroux, 2017; Spassky and Meunier, 2017).

The Frizzled (Fz) planar cell polarity pathway (PCP) is a conserved pathway governed by two protein complexes – the Frizzled/Dishevelled/Flamingo/Diego (Fz/Dsh/Fmi/Dgo) complex and the Vang/Prickle/Flamingo (Vang/Pk/Fmi) complex (Adler, 2012; Carvajal-Gonzalez and Mlodzik, 2014; Devenport, 2014; Goodrich and Strutt, 2011; Peng and Axelrod, 2012; Singh and Mlodzik, 2012). The main role of the Fz-PCP pathway during development is to coordinate cells within a tissue in different cellular processes. Recently, this pathway has been shown to be essential for proper translational polarity, which refers to the off-centre movement of centrioles, and/or rotational polarity, which refers to the position of a basal body relative to neighbouring basal bodies. Examples in multiciliated cells occur in the larval skin of Xenopus, mouse ependymal cells and mouse oviduct epithelial cells (Mitchell et al., 2009; Park et al., 2006, 2008; Guirao et al., 2010; Shi et al., 2014, 2016). Similarly, the positioning of cilia and their basal bodies is also regulated by the Fz-PCP pathway in the outer hair cells in the mouse cochlea (Ezan and Montcouquiol, 2013; Jones and Chen, 2008), the mouse node (Song et al., 2010) and the floor plate of zebrafish (Borovina et al., 2010). We have found previously that this centriole polarization function of the Fz-PCP pathway was evolutionarily conserved by showing that centriole polarity was also controlled by this pathway in Drosophila wings (Carvajal-Gonzalez et al., 2016a,b).

In vertebrates, the Fz-PCP pathway and basal body polarization link is mediated, at least in part, by the downstream CPLANE (ciliogenesis and planar polarity effector) complex, formed in part by Inturned (Intu), Fuzzy (Fuz) and WD repeat containing planar cell polarity effector (Wdpcp) (Adler and Wallingford, 2017; Park et al., 2006). This CPLANE complex was first discovered in Drosophila, with its components being termed PCP effectors (PPEs), including In (Inturned) (Park et al., 1996), Fy (Fuzzy) (Collier and Gubb, 1997) and Frtz (Fritz, Wdpcp in vertebrates) (Collier et al., 2005), inter alia. The CPLANE complex, together with Dvl (Dishevelled; Dsh in Drosophila) and RhoA (Rho1 in Drosophila), is involved in early stages of centriole positioning during ciliogenesis, strikingly affecting protein vesicular transport and the docking of centrioles to the apical membrane in vertebrates, which is related to actin polymerization defects (Gray et al., 2009; Mitchell et al., 2009; Park et al., 2006, 2008). Although PPEs deeply affect actin polymerization, their role in centriole polarization remains largely unexplored in Drosophila. In fact, the role of actin in centriole polarization downstream of the Fz-PCP pathway is unknown. In this sense, we have discovered previously that centriole planar polarization was very mildly affected in a multiple wing hairs mutant, the most downstream target of the PPE complex (Carvajal-Gonzalez et al., 2016b). Based on this initial result, we decided to study in more depth the connection between actin, known PPEs, and centriole polarization in Drosophila wings.

In adult and pupal Drosophila wings, such PCP phenotypes as hair number defects [multiple hair cell (mhc) phenotypes] or hair misorientation defects are linked to actin polymerization dysfunction that can present at the apical membrane of each epithelial cell (Carvajal-Gonzalez and Mlodzik, 2014; Wong and Adler, 1993). The mhc phenotypes are related to PPEs, including actin polymerization regulators downstream of the core Fz-PCP pathway (Fig. 1A), whereas hair misorientation is controlled by the core members of the Fz-PCP or Fat-PCP pathways (Carvajal-Gonzalez and Mlodzik, 2014; Wong and Adler, 1993). To study the involvement of PPE in centriole polarization, we first tested available RNA interference (RNAi) lines for each of the commonly known PPEs and detected those that generated robust mhc phenotypes in adult wings (Fig. 1B-E; Fig. S1). In our experimental conditions with Sas4-GFP combined with dpp>Gal4, RNAi lines for mwh, frtz and Drok (Rok – FlyBase) produced robust mhc phenotypes in adult and pupal wings (Fig. 1F-Hʺ; Figs S1-S3). Once PCP phenotypes had been confirmed, we calculated centriole polarity by using two existing methods – the average basal body position (ABP) and the quartile (Q) method – and compared centriole polarity in PPE loss-of-function (LOF) versus wild-type (WT) conditions (Fig. 1I,J). Briefly, the ABP method is based on the normalized position of a centriole along the proximal-distal (P-D) axis of the cell, and the Q method quantifies the number of centrioles located within four angular regions (Q1-Q4) of a population of cells, with Q1 being the quartile that represents polarized centrioles (for more details, see Materials and Methods). Unexpectedly, cells with at least two hairs showed a WT polarity of centrioles in frtz-RNAi, Drok-RNAi and mwh-RNAi conditions using the ABP method (Fig. 1I). By contrast, the Q method showed a mild, but statistically significant, reduction in Q1 centriole positioning in frtz-RNAi conditions compared with WT, but no differences compared with WT in the Drok-RNAi and mwh-RNAi experiments (Fig. 1I,J; Table S1 for the statistical analyses). We also observed that frtz-RNAi produced a delay in hair formation with respect to the WT area of the same wing, which also correlates with a delay in centriole polarization (Fig. 1K,L; Fig. S2). This delay could also explain our former mwh mutant data (Carvajal-Gonzalez et al., 2016b), in which we found a mild effect on centriole polarity. Finally, we found that Drok knockdown also generated an increased cell size that did not affect centriole polarization (Fig. S3D).

Altogether, this data set indicates that the number of actin-rich hairs induced by PPE LOF might delay centriole polarity or cell size, but does not affect final centriole position. Our unexpected data on PPE seem to contradict the vertebrate data (Gray et al., 2009; Mitchell et al., 2009; Park et al., 2006, 2008). In vertebrates, the functional link between CPLANE proteins and Rho GTPases and ciliogenesis is very well-established and involves a decreased apical actin assembly, which is important for the orientation of basal bodies and cilia (Park et al., 2006, 2008). In flies, most of the PPE mutants (including mwh, in, fy, frtz, Drok, Rac1, rho) do not show any decreased or increased actin polymerization, and indeed it is believed that PPE mutants fail to restrict the actin polymerization site in the apical membrane. In any case, although this actin defect is present in PPE mutants, we found no effect on centriole polarity.

To determine whether or not actin polymerization is required for centriole polarization, we decided to block actin polymerization acutely by using drug treatment in Drosophila pupal wings in vitro (Fig. 2A). In our culturing conditions, 30-40% of pupal wings were able to fully develop up to the point of producing actin-rich hairs in each epithelial cell and distal wing fold (Fig. 2B-D). Again, using ABP and Q methods, pupal wings aged for 25 h at 29°C plus 10 h cultured in vitro showed partial centriole polarization compared with fully polarized centrioles [pupae aged up to 28.5 hours after puparium formation (APF)] or non-polarized centrioles (pupae aged up to 25 APF) (Fig. 2F,G). Under those in vitro culturing conditions causing a partial centriole polarization (25 APF plus 10 h culture in vitro), when actin polymerization was blocked by cytochalasin D treatment, we found that centriole polarization was completely inhibited (Fig. 2F,G; Table S2 for statistical analyses). This result highlighted that actin polymerization is required for centriole polarization in Drosophila pupal wings, as has been reported for vertebrates previously.

Of the PPEs described, only Cdc42 LOF blocks actin-rich hair formation (Eaton et al., 1995, 1996), a phenotype that we were not able to obtain using available CDC42 stocks, but which is comparable to our cytochalasin D treatment. Furthermore, cytochalasin D interferes with basal body migration and ciliary development in epithelial cells by blocking actin filament formation (Boisvieux-Ulrich et al., 1990), and decreases centriole polarization in planarians (Almuedo-Castillo et al., 2011), and now we found that actin polymerization is required for centriole polarization. Therefore, although PPEs are not required for proper centriole polarization in Drosophila, actin polymerization is required, mimicking the actin and centriole linkage phenotype previously found in vertebrates.

Because we found that PPEs do not affect centriole polarity but actin polymerization is still required in this process, we next investigated whether or not centriole distribution was affected in the distal part of the cell under PPE mutant conditions. To this end, we first developed a better quantitative system with which to measure centriole distribution. In the past and in this study, we and others had measured centriole polarization by using either the distance and position relative to the centre of the cell or the angle of the vector connecting the centre of the cell and the centriole, as in the ABP and Q methods (Carvajal-Gonzalez et al., 2016b; Minegishi et al., 2017; Taniguchi et al., 2011). Although useful, these approaches do not take into consideration the fact that centriole polarization is at least two-dimensional, so there is extra information that could lead to the discovery of otherwise hidden differences. To respond to this question and improve our quantitative analysis, we developed a system to measure and compare the centriole population of a given experimental condition in two dimensions, which we termed the ‘representative polarized centriole distribution’ (RPCD) (Fig. S4).

The RPCD is based on the generation of a centriole positioning library in well-polarized epithelial cells (Fig. S4A-D). The accumulation of single centriole positions in a normalized epithelial cell model produces a density map representing the distribution and enrichment of centrioles (Fig. S4D). This density map can then be segmented using different isolines along which the probability of finding a centriole is the same (Fig. S4E,F). By using this approach, we obtained a model distribution for centrioles that can be used for comparisons between two populations of centrioles (Fig. S5; see Materials and Methods for more details). Briefly, for a given number of cells, we retrieved from our library a model distribution of well-polarized centrioles. This expected distribution was then confronted with the observed centriole coverage value in the experimental condition, and the ratio R_0.05 (observed/expected) will be greater than unity when centrioles are well polarized and less than unity when centrioles are not significantly polarized (P<0.05) (Fig. S5).

As a proof of principle for our RPCD method, we compared centrioles from WT wings before hair formation (25 APF), with pre-hairs or with well-formed hairs (28.5 APF). As shown in Fig. 3A, with the appearance of the actin-rich hairs in WT wings, centrioles polarized towards the distal part, going from non-planar polarized and centred in the apical membrane to planar polarized towards the distal part of the cell (Fig. 3A,B). Furthermore, when we evaluated centriole polarization using RPCD in Fz overexpression (Fz-OE) or Vang knockdown conditions, we confirmed that centriole polarization was disturbed with R_0.05 values being less than unity (Fig. 3C,D; Fig. S7A), obtaining results similar to those previously obtained with the ABP and the Q methods (Fig. S6; Table S3).

Using our RPCD method under PPE conditions, we found that the centriole distribution in the distal part of the cells is not different from that of WT cells with a R_0.05 greater than unity in frtz-RNAi, Drok-RNAi and mwh-RNAi conditions (Fig. 3E,F; Fig. S7). Taking our RPCD, ABP and Q methods data together, we conclude that neither centriole polarity nor centriole population distribution in the distal part of the cell are affected under PPE conditions.

Unexpectedly, when we applied the RPCD method to our in vitro experimental data in wings treated with cytochalasin D, we observed noticeable differences in the density map of centrioles in the cytochalasin D-treated wings compared with all the other conditions (Fig. 3G,H). We found that the distribution of centrioles in cytochalasin D-treated wings generates a density map that is more spread out and homogeneous in the apical planes of cells than the maps observed in Fz-OE, vang-RNAi or ‘no-hair’ conditions (Fig. 3; Fig. S7A-D). To confirm this new phenotype, we used another centriole distribution model for comparison, but in this case for a non-polarized centriole distribution using wings developed up to 25 APF, which we termed a representative non-polarized centriole distribution (RNCD) (Fig. S4H-J). We compared different non-polarized conditions (Fz-OE, vang-RNAi and cytochalasin D treatment) with the RNCD, and found that centriole distributions in wings treated with cytochalasin D were significantly different from the rest of the non-polarized conditions (Fig. 3I). Furthermore, quantification of the centriole distribution after short treatments with cytochalasin D before hairs are formed or after hair formation showed the same result: centrioles became dispersed at the apical planes of the cells (Fig. S7E,F). Altogether, our data indicate that actin is required to keep centrioles in a restricted area before and after centriole polarization takes place.

This new analysis using RNCD and RPCD methods highlighted that centriole distribution is affected by actin polymerization in a different manner than under Fz-OE or vang-RNAi experimental conditions. Under core PCP mutant conditions, centrioles do not polarize (compare core PCP results with RPCD) but remain restricted to the central area in the apical-most planes of the epithelial cell (compare core PCP with RNCP) (Fig. S4K). In contrast, in the cytochalasin D treatment, the centrioles did not polarize (compare the cytochalasin D results with RPCD) and became dispersed in the apical planes (compare the cytochalasin D results with RNCD) (Fig. S4K).

Next, to analyse whether actin polymerization is sufficient to polarize centrioles, we went back to our hair misorientation phenotypes. Under Fz-PCP mutant conditions, actin polymerization still takes place in the apical membrane, but the actin-rich hair is generated at the wrong place within that membrane, which correlates with centriole distribution defects (Fig. S6). Under those circumstances, we wondered whether, under hair misorientation conditions, centrioles were still able to follow the actin-rich hair or were completely uncoupled (Fig. 4A,B). To answer this question, we measured the co-occurrence of the two structures, actin-rich hairs and centrioles, at the same angle (see Materials and Methods for details; Fig. 4C-G). We analysed the percentage of co-occurrence in cells with hair misorientation phenotypes using Fz-GOF (gain of function) and Vang-LOF conditions, but only in those cells in which centrioles were able to migrate towards the cell periphery. In all the conditions tested, cells with hair misorientation had a lower ability to locate centrioles at the base of the actin-rich hairs (Fig. 4H). These cells had 20% co-occurrence in Fz-GOF and 26% in Vang-LOF, both lower than the 46% estimated in WT cells (Fig. 4H). We concluded that, under PCP LOF and GOF conditions, the centriole loses its ability to follow the actin-rich hair positioning in the apical membrane of Drosophila wing epithelial cells. Together with the cytochalasin D treatment, our data revealed that actin polymerization is necessary but not sufficient to polarize centrioles in Drosophila wings. Our results suggest that the connection between the actin-based hair and the centriole depends on the Fz-PCP pathway but is independent of known PPEs, so that new downstream players need to be found.

In summary, by using complementary quantitative methods, we found three distinct centriole polarization phenotypes. Under PPE mutant conditions, centriole polarization is delayed, but the final distribution and polarity are completely unaffected (Fig. 4I; Fig. S4K). However, PCP core mutants do not polarize centrioles as they remained mostly centred in the cell (Fig. 4I; Fig. S4K). In contrast, actin polymerization is required both to polarize centrioles and to keep them in a restricted area of the cell (Fig. 4I; Fig. S4K) before and after hair formation. In conclusion, we hypothesize that in Drosophila pupal wings there are at least two mechanisms controlling centriole positioning – one a general mechanism governed by actin polymerization and independent of PCP, the main function of which is to maintain centriole positioning, and the other a mechanism controlled by the core PCP pathway in charge of moving the centriole to the proper position. Further experiments will be required to dissect the mechanism behind this off-centred movement downstream of the core PCP pathway.

Fly lines were cultured and crossed on semi-defined medium and maintained at the indicated temperatures (25°C or 29°C). For RNAi studies, the following lines were used: Fz RNAi (VDRC stock 105493/KK), Vang RNAi (VDRC stock 7376/GD and 100819/KK), mwh (VDRC stock 41514/GD and 45265/GD) Drok (VDRC stock 3793/GD and 104675/KK), fritz (VDRC stock 103611 and 10088/GD), ds (VDRC stock 36219/GD and 4313/GD), fuzzy RNAi (VDRC stock 108550/KK), Cdc42 RNAi (VDRC stock 100794/KK), Rac1 (VDRC stock 49247/GD), inturned (VDRC stock 27252/GD and 103407/KK) and rho RNAi (VDRC stock 107502/KK and 51952/GD). For overexpression and RNAi studies, the GAL4/UAS system was used to direct the expression of UAS constructs to distinct wing areas, linked to the localized expression of developmental genes such as decapentaplegic (dpp), expressed in a central stripe, and nubbin (nub), expressed in the whole wing. For cytochalasin D experiments, UAS-GFP-CG2919 (Asl) flies were crossed with UAS-dcr²nub-GAL4 to direct the expression of asterless coupled to green fluorescent protein (GFP) to the whole wing.

For analysis of wing trichomes, adult wings were removed, incubated in wash buffer [PBS and 0.1% Triton X-100 (PBS-t)] and mounted on a slide in 80% glycerol in PBS. Adult wings were imaged at room temperature using an Olympus BX51 direct microscope. Images were acquired with an Olympus DP72 camera and CellD software (Olympus).

Pupae were collected from vials at white stage and cultured at 29°C for different time points (25 h or 28.5 h). Pupae were dissected in PBS-t and fixed with 4% paraformaldehyde for 1 h at room temperature. Then, pupae were washed in PBS-t three times for 5 min each wash and blocked in PBS-t with 2% bovine serum albumin (BSA) for 45 min. Samples were incubated with primary antibody overnight at room temperature in PBS-t-BSA. After overnight incubation, samples were washed five times in PBS-t and incubated for 90 min in fluorescent phalloidin to stain F-actin and fluorescent secondary antibodies, both diluted in PBS-t-BSA. To continue the staining, five washes in PBS-t were performed and pupal wings were then detached from the pupal cage. Finally, samples were mounted on slides with medium for fluorescence (Vectashield). To stain the cellular membrane anti-Fmi (also known as Starry night, Stan; from DSHB clone 74, dilution 1:20) was used. Secondary antibody conjugated with the fluorophore Alexa 405 was used at 1:200 (A-31553, Invitrogen) and Alexa 594-phalloidin was used at 1:200 (Invitrogen).

Pupal wings were oriented with the distal part always pointing to the right side of the image and images were acquired using an Olympus FV 1000 confocal microscope. After acquisition, images were processed using ImageJ (Fiji) to select the wing sections and generate the cell borders mask (Tissue Analyzer plugin), and Adobe Photoshop CC 2015 to create the colour-coded mask based on the number of hairs.

For cytochalasin D treatment, UAS-GFP-Asl and UAS-dcr2;nub-GAL4, which directs the expression of GFP-asterless to the whole wing, were used. White pupae from the crossing were cultured for 25 h at 29°C and removed from the pupal case in aseptic conditions. Dissection was carried in modified M3 medium (MM3) supplemented with 20-hydroxyecdysone at 100 ng/µl (Sigma). Then, each pupa was carried separately to 200 µl tubes and cultured at 29°C for 10 h in 100 µl of supplemented MM3 medium. For inhibition of actin polymerization, cytochalasin D was used at 5 ng/µl and treatment was carried out in supplemented MM3 medium in the same conditions. After incubation, pupae were washed in PBS-t for 5 min and the staining was performed as described above.

Each centriole was assigned to a specific quartile (Q) depending on their relative position to the anterior-posterior axis of the cell (Taniguchi et al., 2011). Specifically, we drew two virtual lines that crossed the centroid of the cell with a π/4 and −π/4 orientation. These lines divided the space of the cell in four different regions. If the centriole position was located between −π⁄4 and π⁄4, then it was assigned to the Q1 region. If the centriole was located between π⁄4 and 3π/4, it was assigned to Q2. In the case of being located between −π⁄4 and −3π/4, then it was assigned to Q3, and, finally, the rest of centrioles were assigned to Q4. A higher proportion of Q1 values represents polarized centrioles.

where i represents the index of the set of n matrices, E is an all-ones matrix with the same dimension of C_i, and tr indicates the trace of the matrix. Note that the divider of this equation represents the total sum of the elements of C_i. This accumulated density represents an intuitive way to visualize centriole position and polarity in a set of cells, as shown in Fig. 3.

where q_obs represented the q value observed for the external sample at the isoline k. Values of OER₀₅ higher or lower than 1 indicated polarized or non-polarized conditions, respectively.

The previous approach was modified to generate a non-polarized centriole library from cells without hair formation. In order to do this, we selected 5129 cells from different fields, wings and experiments from 25 APF wings. Then, all the previously described steps were performed (isoline generation and calculation of q_exp05 for different cell library sizes and isolines) in order to compare centriole samples against a non-polarized model.

Data were analysed using one-way ANOVA followed by the Bonferroni multiple comparisons post-test (GraphPad Prism) for co-occurrence, Q and ABP methods. In the case of RPCD and RNCD, the data were compared with the model using the OER₀₅, which measures the statistical significance (values below 1 represent a P-value below 0.05) but also using t-test to compare different experimental groups.

We are most grateful to Jordan Raff for many reagents and all members of the Carvajal lab for helpful suggestions, comments and discussions during the development and execution of the project. We thank the Bloomington and VDRC Stock Centers for fly strains, and the DSHB for antibodies. Confocal microscopy was performed at the UEX microscopy core facility (Servicio de Técnica Aplicadas a Biociencias, STAB). We thank Dr Francisco Centeno and Dr Sonia Mulero for helpful discussions and comments on the manuscript.