ABSTRACT
In the developing Drosophila eye, the centrioles of the differentiating retinal cells are not surrounded by the microtubule-nucleating γ-tubulin, suggesting that they are unable to organize functional microtubule-organizing centers. Consistent with this idea, Cnn and Spd-2, which are involved in γ-tubulin recruitment, and the scaffold protein Plp, which plays a role in the organization of the pericentriolar material, are lost in the third-instar larval stage. However, the centrioles maintain their structural integrity, and both the parent centrioles accumulate Asl and Ana1. Although the loading of Asl points to the acquisition of the motherhood condition, the daughter centrioles fail to recruit Plk4 and do not duplicate. However, it is surprising that the mother centrioles that accumulate Plk4 also never duplicate. This suggests that the loading of Plk4 is not sufficient, in this system, to allow centriole duplication. By halfway through pupal life, the centriole number decreases and structural defects, ranging from being incomplete or lacking B-tubules, are detected. Asl, Ana1 and Sas-4 are still present, suggesting that the centriole integrity does not depend on these proteins.
INTRODUCTION
The centrosome is a structured protein complex that recruits microtubule-nucleating proteins and tubulin (Woodruff et al., 2017), acting as the main microtubule-organizing center (MTOC) of the animal cells (Sanchez and Feldman, 2017). Through its ability to nucleate the cytoplasmic microtubule network, the centrosome plays essential roles in various cellular activities, including cytoplasmic transport, cell movement, chromosome segregation and primary cilia formation (Bettencourt-Dias et al., 2011; Arquint et al., 2014; Conduit et al., 2015). The centrosome is also involved in some aspects of the cell division because it organizes the mitotic spindle and dictates its orientation throughout the cell cycle (Roubinet and Cabernard, 2014; Meraldi, 2016). Therefore, the accurate control of centrosome dynamics is instrumental to avoiding a plethora of cellular defects including ciliopathies, microcephalies, aneuploidy and cancer (Zyss and Gergely, 2009; Crasta et al., 2012; Vitre and Cleveland, 2012; Chavali et al., 2014; Godinho and Pellman, 2014).
Centrosome elimination is a common feature during gametogenesis of most organisms (Manandhar et al., 2005; Mikeladze-Dvali et al., 2012; Pimenta-Marques et al., 2016). Female germ cells, indeed, lose or inactivate their centrosomes to avoid multipolar spindles at fertilization. Thus, the assembly of the zygotic centrosome is driven by the sperm-provided centrioles (Schatten, 1994). Centrosome elimination has been also reported in endoreduplicating intestinal cells of Caenorhabditis elegans (Lu and Roy, 2014) and follicle cells of Drosophila melanogaster (Mahowald et al., 1979). The alteration of the centrosome integrity has been also observed in human post-mitotic cells (Zebrowski et al., 2015) and in Drosophila somatic cyst cells (Riparbelli et al., 2009).
Since the organization and integrity of the centrosome depends on a pair of centrioles at its heart (Sluder and Rieder, 1985), understanding centriole dynamics is crucial to deciphering centrosome behavior. We have now a quite detailed knowledge of centriole composition and architecture, and the process of their duplication (Lattao et al., 2017), but there is a poor understanding of how the centrioles are eliminated or inactivated in differentiated cells.
It has been recently reported that the starfish oocytes lose mother centrioles before the daughters (Borrego-Pinto et al., 2016). In this system, the mother centriole retains its microtubule-organizing activity and moves through a dynein-dependent process towards the plasma membrane to be eliminated with the extrusion of the polar body. Thus, the elimination of the parent centrioles is associated with their different activity. The basal bodies of Caenorhabditis sensory neurons also degenerate early in neuronal differentiation, after the formation of the ciliary structures (Serwas et al., 2017).
In an attempt to study the mechanisms underlying centriole elimination in post-mitotic cells, we examined centriole behavior during the developments of the Drosophila eye. The adult Drosophila eye is formed from ∼750 ommatidial units that are derived by a complex differentiation process that requires dramatic cell transformations (Wolff and Ready, 1993; Carthew, 2007). The eye-antennal imaginal disc of the third-instar Drosophila larvae is crossed by the morphogenetic furrow that represents the boundary between the anterior region in which undifferentiated epithelial cells proliferate and the posterior region that holds the differentiating rhabdomeric cells (Cagan, 2009; Treisman, 2013). The epithelial cells ahead of the morphogenetic furrow proliferate randomly, but arrest in G1 as they enter the morphogenetic furrow (Thomas et al., 1994; Treisman and Heberlein, 1998). The cells posterior to the morphogenetic furrow are organized into regularly spaced groups, the ommatidial preclusters, within which the differentiation of the photoreceptors occurs (Tomlinson and Ready, 1987; Wolff and Ready, 1991). The remaining uncommitted cells re-enter the cell cycle and undergo a single round of cell division within the second mitotic wave, before they arrest terminally (Wolff and Ready, 1991; de Nooij and Hariharan, 1995; Firth and Baker, 2005; Escudero and Freeman, 2007). As the distance from the furrow increases, the ommatidial units became complete with the recruitment of the last three rhabdomeric cells and the differentiation of pigment- and lens-secreting cone cells (Ready et al., 1976; Wolff and Ready, 1991). Since the post-mitotic cell differentiation occurs progressively, successive developmental stages are present in the anterior to the posterior region of the imaginal disc. Thus, the neuroepithelial cells of the developing Drosophila eye represent a suitable model to study the changes in centriole dynamics and organization that accompany terminal cell fate specification.
We show here that the centrosomes of the Drosophila retinal cells lose their activity early during the larval stage, whereas the centrioles start to disappear later by halfway through the pupal life. Centriole elimination begins with the gradual disassembly of the microtubule wall, followed by the disappearance of the cartwheel and the loss of the ninefold symmetry. Moreover, the centrioles of post-mitotic cells fail to duplicate even if they accumulate Asl and Plk4.
RESULTS
Neuroepithelial cells lose functional centrosomes
The eye-antennal disc of the Drosophila third-instar larva is crossed by a distinct groove, the morphogenetic furrow, that establishes the anterior and posterior regions of the disc (Fig. 1A). The apical cytoplasm of the retinal cells in the posterior region showed focal accumulations of tubulin continuous with longitudinal bundles of microtubules (Fig. 1B). These microtubules radiated from electron-dense material lining the plasma membrane of short microvillus-like projections (Fig. 1C). Distinct centrioles were found in the apical cell cytoplasm, but they did not contact the microtubules nor the peripheral electron-dense material (Fig. 1D). These observations point to non-conventional microtubule-organizing centers like those described in the cone cells of the Drosophila ommatidia (Mogensen et al., 1993) and suggest that centrioles of the epithelial cells within the posterior region of the imaginal discs were unable to properly recruit centrosomal material. To verify whether these centrioles lose their ability to recruit the main centrosomal proteins during the differentiation process of the ommatidia, we first analyzed the localization of γ-tubulin, the master protein for microtubule nucleation. γ-tubulin was found in the anterior region of the third-instar larval imaginal disc as small spots associated with the centrioles of the interphase cells and as large aggregates at the poles of the mitotic cells (Fig. 1E). Behind the morphogenetic furrow, only the cells that underwent a new cell division within the second mitotic wave and a few scattered mitotic cells in the more posterior region of the disc displayed distinct accumulations of γ-tubulin at their spindle poles (Fig. 1E). Although γ-tubulin did not accumulate at the interphase centrosomes of the differentiating rhabdomeric cells, a weak labeling was found at their apical surface (Fig. 1F). This staining did not overlap the centrioles but was presumably associated with the nucleation sites for the longitudinal microtubule bundles seen at the plasma membrane.
Loss of centrosome function in the larval imaginal disc eye. (A) Whole-mount of the eye-antennal disc of the third-instar larva: the morphogenetic furrow (mf) represents the boundary between the anterior (a) and the posterior (p) regions of the disc. Anterior is to the left. (B) Behind the morphogenetic furrow (mf), longitudinal bundles of microtubules (arrowheads) cross the cytoplasm of the retinal cells and end in the apical region where strong accumulations of tubulin are found (arrows). (C) The cell membrane bulges out to form short microvillus-like projections containing electron-dense material in which the cytoplasmic microtubules end (arrowheads). (D) Detail of the apical region of a retinal cell showing two separated centrioles that lack PCM and are not in contact with cytoplasmic microtubules. (E–E″) γ-tubulin is associated with the centrioles of the proliferating cells in the anterior region of the eye disc (a) and inside the morphogenetic furrow (mf); behind the morphogenetic furrow, a distinct accumulation of γ-tubulin is found at the spindle poles of the cells within the second mitotic wave (arrowheads), whereas γ-tubulin does not associate with the centriole clusters of the differentiating rhabdomeric cells (asterisks). (F–F″) Weak γ-tubulin spots (arrowheads) that do not overlap with the Plp signal (arrows) are present in the apical region of the rhabdomeric cells. Scale bars: 100 µm (A); 2.5 µm (B); 250 nm (C,D); 15 µm (E,E′), 5 µm (inset E′); 8 µm (F–F″), 5 µm (inset F″).
Loss of centrosome function in the larval imaginal disc eye. (A) Whole-mount of the eye-antennal disc of the third-instar larva: the morphogenetic furrow (mf) represents the boundary between the anterior (a) and the posterior (p) regions of the disc. Anterior is to the left. (B) Behind the morphogenetic furrow (mf), longitudinal bundles of microtubules (arrowheads) cross the cytoplasm of the retinal cells and end in the apical region where strong accumulations of tubulin are found (arrows). (C) The cell membrane bulges out to form short microvillus-like projections containing electron-dense material in which the cytoplasmic microtubules end (arrowheads). (D) Detail of the apical region of a retinal cell showing two separated centrioles that lack PCM and are not in contact with cytoplasmic microtubules. (E–E″) γ-tubulin is associated with the centrioles of the proliferating cells in the anterior region of the eye disc (a) and inside the morphogenetic furrow (mf); behind the morphogenetic furrow, a distinct accumulation of γ-tubulin is found at the spindle poles of the cells within the second mitotic wave (arrowheads), whereas γ-tubulin does not associate with the centriole clusters of the differentiating rhabdomeric cells (asterisks). (F–F″) Weak γ-tubulin spots (arrowheads) that do not overlap with the Plp signal (arrows) are present in the apical region of the rhabdomeric cells. Scale bars: 100 µm (A); 2.5 µm (B); 250 nm (C,D); 15 µm (E,E′), 5 µm (inset E′); 8 µm (F–F″), 5 µm (inset F″).
γ-Tubulin recruitment at the centrosome mainly depends on the pericentriolar protein centrosomin (Cnn) (Megraw et al., 1999). Thus, we asked whether the accumulation of this protein may be reduced during the differentiation of the rhabdomeric cells. We find a weak Cnn staining on centrioles of the interphase epithelial cells located both in the anterior and posterior regions of the imaginal disc and within the morphogenetic furrow (Fig. 2A). By contrast, a strong Cnn accumulation was seen at the poles of the dividing cells in the anterior proliferating region and at the poles of the spindles within the second mitotic wave (Fig. 2A). The centrioles of the interommatidial cells were weakly stained, whereas the majority of the centrioles of the rhabdomeric cells did not accumulate Cnn (Fig. 2A). During the early pupal stages only a few spots of low intensity were observed (data not shown).
PCM dynamics in the third-instar larval eye. Cnn (A,A′) accumulation is less evident than Spd-2 (B,B′). The centrioles of the rhabdomeric cells (arrows) lack both Cnn and Spd-2 (arrowheads). Scale bars: 15 µm (main images); 5 µm (insets).
The incorporation of Cnn into the pericentriolar material (PCM) is facilitated in somatic Drosophila cells by Spd-2 (Conduit et al., 2014a). Thus, we expected a temporal and spatial colocalization of these proteins during the early phases of the eye development. The anti-Spd-2 antibody, indeed, mainly recognized the centrioles of the interommatidial cells, whereas most of the centrioles associated with the apical region of the rhabdomeric cells were devoid of the Spd-2 protein (Fig. 2B).
Centrioles disappeared during eye development
Since the behavior of the centrosome depends on the centrioles at its heart we analyzed their dynamics during eye development. We traced centrioles through the localization of the conserved centriole-specific core protein Sas-4, which provides a link between the cartwheel and the microtubule wall (Hsu et al., 2008; Tang et al., 2011) and may represent a bona fide marker of centrioles in the developing eye.
Distinct centriole pairs were found behind the morphogenetic furrow on the narrow apical regions of the cells that formed the ommatidial preclusters and within the undifferentiated interommatidial cells (Fig. 3A). Each rhabdomeric cell also displayed two apical centrioles. Since the ommatidia consist of eight rhabdomeric cells, we found evenly spaced groups of eight couples of centrioles within the posterior region of the disc (Fig. 3A).
Timeline of centriole reduction during Drosophila eye development. (A–A′′′′) Distinct ommatidial preclusters (p) are visible just behind the morphogenetic furrow (mf) and complete ommatidial units (asterisks) form in the posterior region of the larval eye disc; the rhabdomeric cells have distinct centriole pairs (arrowheads) that are clustered in evenly spaced groups (arrows). (B–B″) The daughter centrioles, identified by Cnb staining, display a lower Sas-4 accumulation (arrowheads). (C–E″) Top panels are surface views of the Drosophila retina at different developmental times; middle and lower panels are details of single ommatidia and surrounding interommatidial cells with their centrioles defined by Sas-4 localization. Centrioles of the photoreceptor cells, when present, are out of focus and are not visible at the apical surface. (C–C″) Before the activation of the apoptotic machinery, the ommatidial and interommatidial cells display distinct centriole pairs. (D–D″) As the cell death reduces the interommatidial cells, the centriole number within the retina cells also decreases and only one Sas-4 spot for each cell is usually found. (E–E″) At later developmental stages only two or three Sas-4 spots are observed. Arrows and arrowheads point to centrioles of ommatidial and interommatidial cells, respectively. b, mechano-sensory bristles. (F) Schematic illustration of the apical profiles of single ommatidia at different developmental times showing two primary pigment cells (yellow) and four cone cells (gray). For simplicity, we did not include the interommatidial cells. The eight photoreceptor cells lie below the cone cells and are not visible. The red spots represent parent centrioles. (G) Quantification analysis of the centriole numbers within the ommatidia at different times of development as defined by Sas-4 localization. n, total number of the ommatidia scored for each developmental stage. Scale bars: 1 µm (A); 0.25 µm (A′–A′′′′); 1 µm (B,B′); 4 µm (B″); 1 µm (C–E); 0.25 µm (C′–E″).
Timeline of centriole reduction during Drosophila eye development. (A–A′′′′) Distinct ommatidial preclusters (p) are visible just behind the morphogenetic furrow (mf) and complete ommatidial units (asterisks) form in the posterior region of the larval eye disc; the rhabdomeric cells have distinct centriole pairs (arrowheads) that are clustered in evenly spaced groups (arrows). (B–B″) The daughter centrioles, identified by Cnb staining, display a lower Sas-4 accumulation (arrowheads). (C–E″) Top panels are surface views of the Drosophila retina at different developmental times; middle and lower panels are details of single ommatidia and surrounding interommatidial cells with their centrioles defined by Sas-4 localization. Centrioles of the photoreceptor cells, when present, are out of focus and are not visible at the apical surface. (C–C″) Before the activation of the apoptotic machinery, the ommatidial and interommatidial cells display distinct centriole pairs. (D–D″) As the cell death reduces the interommatidial cells, the centriole number within the retina cells also decreases and only one Sas-4 spot for each cell is usually found. (E–E″) At later developmental stages only two or three Sas-4 spots are observed. Arrows and arrowheads point to centrioles of ommatidial and interommatidial cells, respectively. b, mechano-sensory bristles. (F) Schematic illustration of the apical profiles of single ommatidia at different developmental times showing two primary pigment cells (yellow) and four cone cells (gray). For simplicity, we did not include the interommatidial cells. The eight photoreceptor cells lie below the cone cells and are not visible. The red spots represent parent centrioles. (G) Quantification analysis of the centriole numbers within the ommatidia at different times of development as defined by Sas-4 localization. n, total number of the ommatidia scored for each developmental stage. Scale bars: 1 µm (A); 0.25 µm (A′–A′′′′); 1 µm (B,B′); 4 µm (B″); 1 µm (C–E); 0.25 µm (C′–E″).
Surprisingly the Sas-4 signal had a different intensity within the centrioles of the same pair (Fig. 3B). To decipher whether this signal was a property of only one parent centriole, we counterstained the eye imaginal discs with an antibody against centrobin (Cnb), which specifically recognizes the daughter centrioles (Gottardo et al., 2016). The Cnb stain overlapped with the centrioles that had lower Sas-4 intensity (Fig. 3B) that can be, therefore, taken as daughters. The lower Sas-4 accumulation at the daughter centrioles was also observed when the parent moved away (Fig. 3B).
The eye imaginal disc dramatically changed shape during the transition to the pupal stage becoming a thin disc-like epithelium in which the ommatidia are separated by a matrix of unpatterned interommatidial cells arranged in double or triple rows. At ∼25 h after puparium formation (APF), the interommatidial cells sort into single rows disposed in a precise hexagonal pattern (Fig. 3C). At this developmental stage, the centrioles of the rhabdomeric cells were barely detectable because the apical surface of these cells had retracted below the cone cells. By contrast, the centrioles of the interommatidial, cone and primary pigment cells could be easily identified because they were present at approximately the same focal plane at the surface of the ommatidial units. Each cell had one distinct centriole pair at this stage of development. Therefore, apical views of the retina show groups of 12 centrioles, eight from cone cells and four from primary pigment cells, surrounded by the centriole pairs of several interommatidial cells (Fig. 3C). At ∼45 h APF the interommatidial cells flattened and distinct mechano-sensory bristles were visible at the anterior vertex of each ommatidium (Fig. 3D). The majority of the interommatidial, cone and pigment cells showed single Sas-4 spots (Fig. 3D). By 65 h APF, only two or three spots of Sas-4 were found in the apical region of the ommatidial units (Fig. 3E). A quantification of the centrioles as defined by Sas-4 staining confirmed the progressive reduction of their number from 25 h to 65 h APF (Fig. 3F,G).
Centrioles of the pupal eye lost the scaffold protein Plp, but maintained the core proteins Asl and Ana1
It has recently been suggested that centriole elimination in Drosophila somatic cells and female germ cell line could be a consequence of the loss of different components of the PCM (Pimenta-Marques et al., 2016). This prompted us to verify whether the centriole disappearance during later stages of eye development was also associated with the loss of distinct centriolar constituents that play the main roles in centriole and centrosome biogenesis. Since we found that γ-tubulin was the first PCM component to be lost, and then Cnn and Spd-2, we asked whether the dynamics of the main core centriole proteins involved in PCM recruitment and organization might be also affected during eye development.
A distinct role in PCM organization is played by the coiled-coil Pericentrin-like protein (Plp), which is radially arranged around the centriole wall and organizes a distinct scaffold before the PCM is recruited (Mennella et al., 2012). Moreover, the direct interaction between Plp and Cnn is required for normal centrosome organization and activity during interphase and mitosis of the Drosophila syncytial embryo (Lerit et al., 2015; Richens et al., 2015). We then sought to analyze the distribution of this peripheral centriolar component during the eye development. Plp was associated with all the centrioles of the rhabdomeric cells in the larval eye imaginal disc (Figs 1E and 4A), but found that they start to disappear early in the pupal stage. By 25 h APF, the anti-Plp antibody recognized seven or eight small spots within the apical surface of each ommatidium out of 12 recognized at this stage by the Sas-4 antibody (Fig. 4A). At 45 h APF there were usually six Sas-4 spots in the apical region of each ommatidium but only two or three of them maintained a detectable Plp signal (Fig. 4A).
Dynamics of the centriole core proteins during eye development. Distributions of Plp (A–A″), Ana1 (B) and Asl (C) at different stages of eye development. All the centrioles accumulate Ana1 and Asl, but some of them lack Plp during the pupal stages. Scale bars: 0.25 µm.
The assembly of the PCM around the centrioles requires the products of the genes asterless (asl) and anastral spindle 1 (ana1) (Lattao et al., 2017). The recruitment of Spd-2 seems, indeed, to be initially supported by Asl (Conduit et al., 2014a), although additional observations have suggested that Drosophila centrioles lacking Asl may efficiently recruit PCM (Galletta et al., 2014). Asl, in turn, is recruited and maintained at the centriole by Ana1 (Fu et al., 2016; Saurya et al., 2016). These proteins extend from the inner centriole to the outermost part of it and are usually recruited at the daughter centrioles when they acquire motherhood (Fu and Glover, 2012). Both Ana1 (Fig. 4B) and Asl (Fig. 4C) were found within the centrioles of the rhabdomeric cells during the larval stage. When the apical surface of the photoreceptor cells turned by 90° at the beginning of pupal life, the staining for Asl and Ana1 was barely detectable. By contrast, the centrioles of the cone and primary pigment cells displayed strong Ana1 and Asl signals (Fig. 4B,C). Thus, the apical surface of each ommatidium at 25 h APF displayed a cluster of 12 centrioles surrounded by the centriole pairs of the interommatidial cells (Fig. 4B,C). By 45 h APF, we find six or seven spots of Ana1 or Asl within each ommatidial unit (Fig. 4B,C). At 65 h APF the number of spots reduced to two or three (Fig. 4B,C).
Daughter centrioles failed to recruit Plk4
To understand how long the centrioles maintain their duplication properties, we looked at the localization of Plk4, a protein kinase at the head of the centriole duplication process (Bettencourt-Dias et al., 2005; Habedanck et al., 2005). Plk4 has a distinct cell cycle-dependent localization on the centrioles of the anterior region of the larval imaginal disc where the undifferentiated cells proliferated randomly. At the beginning of interphase, Plk4 was associated with only one centriole of each pair (Fig. 5A). This centriole expressed more Sas-4 and was identified as the mother on the basis of the lack of Cnb (Fig. 3B). As interphase progressed the daughter centrioles gradually accumulated Plk4 and reached the same fluorescence intensity as the mothers. Both the parent centrioles soon displayed a small daughter lacking Plk4. The tight centriole pairs then migrated to the opposite poles of the cells during prometaphase/metaphase (Fig. 5A) and disengaged at anaphase (Fig. 5A). At the end of telophase, the parent centrioles were widely separated and each sister cell inherited a pair of centrioles with only the mother expressing Plk4 (Fig. 5A). Electron microscopy (EM) analysis of the anterior region of the larval eye imaginal disc confirmed the presence of interphase cells with duplicated centrioles (Fig. 5B). The short procentrioles formed by the nine A-tubules and some growing B-tubules were orthogonal to the proximal end of the mother centrioles that were in turn formed by the nine doublet microtubules (Fig. 5B).
Centriole duplication failure in post-mitotic retinal cells. (A) Cell cycle-dependent Plk4 localization in proliferating neuroepithelial cells relative to the centriole marker Sas-4: note that Plk4 docking is restricted to the centriole expressing more Sas-4; I, interphase, M, metaphase, A/T, anaphase/telophase. (B) EM analysis of proliferating late interphase cells showing two pairs of engaged centrioles; the daughter centriole is formed by A-tubules and some growing B-tubules and it is orthogonal to the mother that is built by a set of nine complete doublets. At the end of the larval stage (C) and during the early pupal life (D–D″) Plk4 is restricted to the centrioles expressing more Sas-4. (E–E″) The apical region of the cells behind the morphogenetic furrow shows pairs of disoriented centrioles: one of them is shorter with an incomplete wall and looks like a procentriole. (F,F′) Differently sized centrioles are found at the poles of late telophase spindles. Scale bars: 0.8 µm (A); 500 nm (B,E); 1 µm (B′,B″,E′,E″,F′); 1 µm (C); 0.25 µm (D–D″); 2 µm (F).
Centriole duplication failure in post-mitotic retinal cells. (A) Cell cycle-dependent Plk4 localization in proliferating neuroepithelial cells relative to the centriole marker Sas-4: note that Plk4 docking is restricted to the centriole expressing more Sas-4; I, interphase, M, metaphase, A/T, anaphase/telophase. (B) EM analysis of proliferating late interphase cells showing two pairs of engaged centrioles; the daughter centriole is formed by A-tubules and some growing B-tubules and it is orthogonal to the mother that is built by a set of nine complete doublets. At the end of the larval stage (C) and during the early pupal life (D–D″) Plk4 is restricted to the centrioles expressing more Sas-4. (E–E″) The apical region of the cells behind the morphogenetic furrow shows pairs of disoriented centrioles: one of them is shorter with an incomplete wall and looks like a procentriole. (F,F′) Differently sized centrioles are found at the poles of late telophase spindles. Scale bars: 0.8 µm (A); 500 nm (B,E); 1 µm (B′,B″,E′,E″,F′); 1 µm (C); 0.25 µm (D–D″); 2 µm (F).
No sign of centriole duplication was seen behind the morphogenetic furrow, except within the second mitotic wave area. The retinal cells of the posterior region of the imaginal disc displayed a centrioles pair in which only the mother showed a distinct Plk4 accumulation (Fig. 5C). However, the mother centrioles did not support procentriole assembly despite the fact that they showed accumulation of Plk4. At 25 h APF each retinal cell expressed two Sas-4 spots, but only one Plk4 spot was present (Fig. 5D). As the eye development proceeded the number of centrioles expressing Plk4 further reduced. The Plk4 labeling was barely detectable by 45 APF when each cone and pigment cell displayed a single Sas-4 spot (data not shown).
EM analysis revealed that distinct procentrioles closely associated with the mothers only in the region just posterior the morphogenetic furrow where the cells start to divide to enter the second mitotic wave (data not shown). Behind this tight proliferative area each cell had only two centrioles and these had lost their orthogonal configuration and became un-orientated (Fig. 5E). One centriole of the pair was shorter and looked like a procentriole (Fig. 5E). This centriole was usually composed of a distinct cartwheel and single A-tubules; only occasionally were one or two B-tubules observed. Therefore, in differentiating rhabdomeric cells, the procentrioles did not complete their doublet content and failed to elongate properly. Despite examining serial sections from several cells in the posterior region of the larval imaginal disc (n=47), we never observed parent centrioles in association with a newly formed daughter. The different length of the parent centrioles during interphase is unusual, since sister cells inherit, at the end of mitosis, two disengaged and equally sized centrioles. Serial sections of late telophase cells revealed that there was a pair of different sized centrioles at the cell poles (Fig. 5F) pointing to the failure of centriole elongation during the previous interphase.
Centrioles of post-mitotic cells lose their structural integrity
The mother centrioles within the anterior and the posterior regions of the larval imaginal disc were built by nine doublets of microtubules and a central cartwheel (Fig. 6A). This organization persisted through the early pupal life. By 45 h APF all the centrioles scored (n=27) still maintained a central cartwheel, but often they displayed distinct defects of the microtubule wall. These defects were more evident in cross sections and ranged from incomplete to lacking B-tubules (Fig. 6B,C). We excluded the possibility that these centrioles were remnant procentrioles that failed to grow. The assembly of the B-tubule occurs, indeed, during centriole elongation from arc-like projections starting from the external side of the A-tubule, whereas the disassembly of the B-tubule occurred at the opposite side (Fig. 6B) or at the middle region of the same tubule (Fig. 6C). The majority of the centrioles scored during later stages of eye development (75%, n=34) lost their cartwheel and the centriole wall was often built by eight singlet-doublet tubules (Fig. 6D). Thus, the diameter of these centrioles (148±2.3 nm; n=9) was reduced compared to the diameter of the centrioles that maintained the ninefold symmetry (165±1.9 nm: n=21).
Centriole disassembly during late pupal stages. Cross sections of mother centrioles from retinal cells during larval (A) and pupal stages (B–D). The mother centrioles during the larval stage and early pupal life always display complete sets of nine doublets with a distinct cartwheel (A). The disassembly of the B-tubule occurs at the side opposite to the region of its assembly (B, arrows) or at the middle of the tubule (C, arrowheads). (D) Abnormal centriole symmetry in older retinal cells. Scale bars: 250 nm (A–D).
Centriole disassembly during late pupal stages. Cross sections of mother centrioles from retinal cells during larval (A) and pupal stages (B–D). The mother centrioles during the larval stage and early pupal life always display complete sets of nine doublets with a distinct cartwheel (A). The disassembly of the B-tubule occurs at the side opposite to the region of its assembly (B, arrows) or at the middle of the tubule (C, arrowheads). (D) Abnormal centriole symmetry in older retinal cells. Scale bars: 250 nm (A–D).
DISCUSSION
Recent findings suggest that the centrioles may be modified during development, yet still maintain their function. In the Caenorhabditis embryo, indeed, the centrioles remodel to nucleate the axoneme of sensory neurons (Nechipurenko et al., 2017), and during Drosophila spermiogenesis the sperm basal body is modified both in protein composition and in ultrastructure (Khire et al., 2016). However, the centrioles do not persist at the base of the mature sensory cilia in Caenorhabditis suggesting that they are dispensable for cilia maturation and maintenance (Serwas et al., 2017). Similarly, the centrioles of the differentiating Drosophila ommatidia gradually lose their structural organization and then disappear. It is unclear why it would be advantageous to remove the centrioles from the developing Drosophila eye. Perhaps, centriole elimination could prevent the assembly of too many centrosomes and unnecessary mitotic spindles that may impair cell differentiation by imposing abnormal divisions. It has been, indeed, proposed that centrosome inactivation in differentiated cells may function as a barrier restricting cell cycle re-entry (Wong et al., 2015).
PCM is lost before pupation
γ-Tubulin and Cnn were the first PCM components that disappeared from the centrioles during the development of the Drosophila eye (Fig. 7). This agrees with similar observations showing the loss of γ-tubulin within the posterior region of the larval eye imaginal disc (Fernandes et al., 2014). These findings might be explained by the spatial localization of Cnn and γ-tubulin within the PCM. 3D-structured illumination microscopy has indeed revealed that the PCM components are organized in two discrete layers around the centriole of Drosophila somatic cells (Fu and Glover, 2012; Mennella et al., 2012): an inner layer formed by the fibrous proteins Asl, pericentrin and Plp that span outwards from the centriole wall, and a peripheral matrix composed of Spd-2, Cnn and γ-tubulin (Mennella et al., 2014). Thus, the more-external proteins could be the first to be removed in differentiating retinal cells. Since Cnn organizes a scaffold to ensure the proper PCM architecture during interphase and mitosis (Conduit et al., 2014b; Lerit et al., 2015; Feng et al., 2017), it is possible that the reduction in the amount of Cnn at the centrioles leads to the instability or defective recruitment of γ-tubulin.
Schematic representation of centriole and PCM dynamics during ommatidia development in the Drosophila eye. Parent centrioles are fully evident until early pupal stages (red, mothers; green, daughters), but become barely distinguishable by 45 h APF when the centriole number reduces (blue, unspecified parent centrioles). The centriole-associated proteins assayed at different developmental stages are listed in the table: +, present; −, absent; and +/−, partially present.
Schematic representation of centriole and PCM dynamics during ommatidia development in the Drosophila eye. Parent centrioles are fully evident until early pupal stages (red, mothers; green, daughters), but become barely distinguishable by 45 h APF when the centriole number reduces (blue, unspecified parent centrioles). The centriole-associated proteins assayed at different developmental stages are listed in the table: +, present; −, absent; and +/−, partially present.
It has been demonstrated that Spd-2 helps the incorporation of Cnn into the centrosome in Drosophila cells (Dix and Raff, 2007; Giansanti et al., 2008; Conduit and Raff, 2010). Moreover, Cnn seems to be required to maintain Spd-2 within the PCM (Conduit et al., 2014a). We, thus, expected a similar behavior of Spd-2 and Cnn in the Drosophila eye imaginal discs. However, the Spd-2 signal was present within the centrioles of the whole imaginal disc, whereas a distinct Cnn labeling was restricted to only the anterior region of the disc, the morphogenetic furrow and the dividing cells of the second mitotic wave. It has been reported that Spd-2 is present at the centrosome of the Drosophila somatic cells with two distinct populations (Fu and Glover, 2012): one inner population close to the centriole wall and another that localized at the PCM during centrosome maturation and presumably interacts with Cnn. Therefore, the most extensive Spd-2 labeling with respect to the Cnn staining may be due to the inner Spd-2 population. However, we were surprised to find that Spd-2 was stably associated with the centrioles of the interommatidial cells in the posterior region of the eye imaginal disc, whereas it is lost or is barely detectable at the centrioles of the rhabdomeric cells. Since all these cells are post-mitotic, the diverse timeline of the PCM loss might reflect a different differentiation degree for the cells scored. Rhabdomeric cells are, indeed, fully differentiated, whereas most of the interommatidial cells have still to be finally committed.
Although γ-tubulin is not associated with the centrioles of the retinal cells in the posterior region of the larval imaginal disc, a distinct population of this protein is found within the apical protrusions of the rhabdomeric cells. Remarkably, Spd-2 and Cnn appear to be redundant regarding ensuring the proper localization of γ-tubulin in these cytoplasmic domains, which are the sites for the nucleation of longitudinal bundles of microtubules and may represent a non-conventional microtubule-organizing center. Accordingly, differentiated animal cell types often display non-centrosomal microtubule-organizing centers that function apically in many epithelia to generate longitudinal microtubule bundles (Bartolini and Gundersen, 2006; Sanchez and Feldman, 2017).
The ‘young configuration’ of the daughter centriole does not prevent Asl accumulation
The recruitment of PCM requires Asl (Conduit et al., 2014a), which accumulates into the daughter centrioles at mitosis (Novak et al., 2014; Fu et al., 2016). Thus, in flies, Asl plays a crucial role in allowing the daughter centriole to mature in a mother able to recruit PCM and to duplicate (Novak et al., 2014; Fu and Glover, 2016). The appropriate loading of Asl is ensured by Ana1, which is recruited later in interphase (Fu et al., 2016). At the end of mitosis, sister cells, therefore, inherit a pair of centrioles both expressing Asl and Ana1. This is also the case for the post-mitotic retinal cells in which daughter centrioles display Ana1 and Asl, suggesting that they have acquired a motherhood condition. However, the parent centrioles of retinal cells are unable to recruit the main PCM proteins and lose the scaffold protein Plp. Moreover, daughter centrioles maintain ‘young’ characteristics and look like procentrioles, suggesting that they fail to undergo full maturation. Daughter centrioles are, indeed, shorter than the mothers and do not complete their microtubule wall. It has been shown that Asl is incorporated into daughter centrioles of the syncytial embryos not as they acquire its complete structure at the interphase exit but later in mitosis, but rather when the parent centrioles have disengaged (Novak et al., 2016). Presumably, the accumulation of Asl to the daughter centrioles of the retinal cells could require the travel through mitosis rather that their full growth. It is, therefore, possible that the acquisition of the motherhood condition needs additional players in these cells.
Sas-4 has an important role in fly cells to promote the recruitment of Asl to daughter centrioles during mitosis (Conduit et al., 2014a; Novak et al., 2016). However, in contrast to the constant level of Asl on both parent centrioles of the retinal cells, we observed a low intensity of the Sas-4 signal at the daughter centrioles, consistent with a reduced antigen amount due to their incomplete structure. This finding suggests that the accumulation of Asl is not directly correlated to the amount of Sas-4, but that a low threshold level of Sas-4 might be enough to recruit high levels of Asl.
Plk4 loading is not sufficient to trigger centriole duplication
It has been suggested that daughter centrioles have to be modified before they can duplicate (Wang et al., 2011) and this modification in flies might be the disengagement and the accumulation of Asl (Novak et al., 2014). Daughter centrioles that lack Asl are unable to duplicate even if they disengage from their mothers at the end of mitosis (Novak et al., 2014). However, the parent centrioles in post-mitotic retinal cells of the Drosophila eye are disengaged and accumulate Asl, but they are unable to duplicate. The need for Asl in centriole duplication is explained by its crucial role in the loading of Plk4, the master protein that marks the site where the new centriole is going to be built (Bettencourt-Dias et al., 2005; Ohta et al., 2014) both in Drosophila (Dzhindzhev et al., 2010) and humans (Sonnen et al., 2012). The inability of the daughter centrioles to duplicate is expected because they lack Plk4, but it is surprising that the mother centrioles never duplicate even if they accumulate Plk4. This suggests that the loading of Plk4 is not sufficient, in this system, to allow centriole duplication. The usual pathway of centriole duplication might be thus inactivated in post-mitotic retinal cells despite the presence of the proteins directly involved in centriole assembly.
It has been recently demonstrated that Plk4 influences the proper radial positioning of Asl on mature somatic centrioles (Galletta et al., 2016). However, we do not find a direct correlation between Asl accumulation at the centrioles and Plk4 recruitment. Plk4 is lost early during eye development, whereas the Asl signal is maintained at the centrioles until later stages. Moreover, although all the centrioles of the retinal cells accumulate Asl, only the mothers displays a distinct Plk4 signal. Since the only significant difference between the parent centrioles is the incomplete structure of the daughters, the asymmetric localization of Plk4 may be explained by additional components that delay the accumulation of Plk4 until the daughter centrioles acquire their complete organization. The low accumulation of Sas-4 on the daughter centrioles might play some roles in this process. Therefore, loading Plk4 to the daughter centriole does not only require the accumulation of Asl during the previous cell cycle, but Plk4 recruitment might also depend on the correct assembly of the nine doublet microtubules that form the centriole wall.
Maintaining the proper centriole structure in post-mitotic retinal cells does not depend on the PCM
The external PCM proteins γ-tubulin, Cnn, Spd-2 and Plp are lost during larval or early pupal stages, but the centriole architecture is not affected by their depletion. This apparently contrasts with the finding that the down-regulation of the PCM leads to centriole elimination in the Drosophila female germline (Pimenta-Marques et al., 2016). Thus, centriole deconstruction may experience different pathways in meiotic and post-mitotic cells. The first structural defects of the retinal cell centrioles are observed during later pupal stages when Sas-4, Ana1 and Asl are still detected. Therefore, maintaining the proper centriole organization does not directly depend on these proteins. Centriole deconstruction in post-mitotic retinal cells starts with the progressive disassembly of the B-tubules pointing to the low stability of the external components of the centriole wall. It has been reported that antibodies against polyglutamylated tubulin lead to the progressive destabilization of the microtubule wall in vertebrate cultured cells, suggesting the involvement of this tubulin isoform in maintaining the stability of the centriole microtubules (Bobinnec et al., 1998). We do not know whether polyglutamylated tubulin also influences the dynamics of the centrioles in the Drosophila retinal cells. It is, however, remarkable that the abnormal centrioles we found in post-mitotic retinal cells look like the incomplete centrioles observed in vertebrate cells following the incubation with antibodies against polyglutamylated tubulin.
MATERIALS AND METHODS
Drosophila strains
We used fly stocks containing Ana1–GFP (Blachon et al., 2009) and Asl–GFP (Dzhindzhev et al., 2010) transgenes. Oregon-R stock was also used as control. Flies were raised on a standard Drosophila medium at 24°C.
Antibodies
We used the following antibodies: mouse anti-γ-tubulin-GTU88 (1:100; Sigma-Aldrich); rabbit anti-Cnb (1:200; Sigma-Aldrich); mouse anti-acetylated tubulin (1:100; Sigma-Aldrich); mouse anti-Sas4 (1:200; Gopalakrishnan et al., 2011); rabbit anti-Spd-2 (1:500; Rodrigues-Martins et al., 2007); chicken anti-Plp (1:1500; Rodrigues-Martins et al., 2007); rabbit anti-Cnn (1:400; Vaizel-Ohayon and Schejter, 1999); rabbit anti Plk4 (1:50; Bettencourt-Dias et al., 2005). The secondary antibodies used (1:800) were Alexa Fluor 488 and 555-conjugated anti-mouse-IgG, anti-rabbit-IgG and anti-chicken-IgG, and were obtained from Invitrogen.
Immunofluorescence preparations
Eye imaginal discs from larvae and pupae at different times after puparium formation were dissected in phosphate-buffered saline (PBS) and fixed in cold methanol for 10 min at −20°C. For antigen localization, the samples were washed for 20 min in PBS and incubated for 1 h in PBS containing 0.1% bovine serum albumin (PBS-BSA, from Sigma-Aldrich) to block non-specific staining. The samples were incubated overnight at 4°C with the specific antisera in a humid chamber. After washing in PBS-BSA, the samples were incubated for 1 h at room temperature with the appropriate secondary antibodies. In all cases, DNA was visualized after an incubation of 3–4 min in Hoechst 33258 (1 µg/ml, Sigma-Aldrich). Imaginal discs were mounted in small drops of 90% glycerol in PBS. Images were taken by an Axio Imager Z1 microscope (Carl Zeiss), using an 100× objective, equipped with an AxioCam HR cooled charge-coupled camera (Carl Zeiss). Gray-scale digital images were collected separately and then pseudocolored and merged using Adobe Photoshop 5.5 software (Adobe Systems).
Transmission electron microscopy
The eye imaginal discs were isolated from third-instar larvae and pupae at different stages of development and transferred in 2.5% glutaraldehyde buffered in PBS overnight at 4°C. Samples were subsequently rinsed in PBS and post-fixed in 1% osmium tetroxide in PBS for 2 h at 4°C. The material was washed in PBS, dehydrated in a graded series of ethanol, embedded in a mixture of Epon-araldite resin, and then polymerized at 60°C for 48 h. Thin sections (50–60 nm thick) were obtained with a Reichert Ultracut E ultramicrotome equipped with a diamond knife, mounted upon copper grids, and stained with samarium triacetate and lead citrate. Samples were observed with a Tecnai Spirit Transmission Electron Microscope (FEI) operating at 100 kV and equipped with a Morada CCD camera (Olympus).
Acknowledgements
We would like to thank Jay Gopalakrishnan (Center For Molecular Medicine, University of Cologne, Germany), Monica Bettencourt-Dias (Gulbenkian Institute of Science, Oeiras, Portugal), Ana Rodrigues-Martins (Gulbenkian Institute of Science, Oeiras, Portugal) and Eyal Schejter (Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel) for generously providing the antibodies used in this study. We also thank Tomer Avidor-Reiss (Department of Biological Sciences, University of Toledo, Toledo, USA) and Nikola Dzhindzhev (Department of Genetics, University of Cambridge, Cambridge, UK) for the flies carrying the GFP transgene.
Footnotes
Author contributions
Conceptualization: G.C., M.G.R.; Methodology: M.G.R., M.G.; Formal analysis: M.G.; Investigation: M.G.R., V.P.; Data curation: V.P., M.G.; Writing - original draft: G.C.; Supervision: G.C.
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
References
Competing interests
The authors declare no competing or financial interests.