ABSTRACT
Centrosomes nucleate microtubules and are tightly coupled to the bipolar spindle to ensure genome integrity, cell division orientation and centrosome segregation. While the mechanisms of centrosome-dependent microtubule nucleation and bipolar spindle assembly have been the focus of numerous works, less is known about the mechanisms ensuring the centrosome–spindle coupling. The conserved NuMA protein (Mud in Drosophila) is best known for its role in spindle orientation. Here, we analyzed the role of Mud and two of its interactors, Asp and Dynein, in the regulation of centrosome numbers in Drosophila epithelial cells. We found that Dynein and Mud mainly initiate centrosome–spindle coupling prior to nuclear envelope breakdown (NEB) by promoting correct centrosome positioning or separation, while Asp acts largely independently of Dynein and Mud to maintain centrosome–spindle coupling. Failure in the centrosome–spindle coupling leads to mis-segregation of the two centrosomes into one daughter cell, resulting in cells with supernumerary centrosomes during subsequent divisions. Altogether, we propose that Dynein, Mud and Asp operate sequentially during the cell cycle to ensure efficient centrosome–spindle coupling in mitosis, thereby preventing centrosome mis-segregation to maintain centrosome number.
INTRODUCTION
During mitosis, cells ensure accurate segregation of their genetic material by assembling a bipolar spindle that is tightly coupled with its two duplicated centrosomes. This coupling is ultimately essential for faithful genome segregation as well as for the correct segregation of one centrosome in each daughter cell (Gönczy, 2015; Gordon et al., 2012). Centrosome–spindle coupling might simply arise from the well-known capability of centrosomes to nucleate microtubules (MTs). However, spindle assembly also relies on the nucleation of acentrosomal MTs occurring after nuclear envelope breakdown (NEB). Furthermore, the loss of centrosome components, such as the conserved Asp (ASPM in mammals) protein, induces defects in the maintenance of centrosome–spindle coupling (Gonzalez et al., 1990; Meunier and Vernos, 2016; Prosser and Pelletier, 2017; Schoborg et al., 2015). Our initial aim was to further understand the function of the conserved spindle orientation factor Drosophila Mud (NuMA or NuMA1 in mammals) in controlling centrosome number in epithelial cells. In doing so, we uncovered how both centrosome positioning in interphase and the correct separation of the centrosomes prior to NEB contribute to the initial centrosome–spindle coupling that is required to prevent supernumerary centrosome formation in the subsequent mitosis.
Although the function of Mud in spindle positioning or orientation has been the focus of numerous works (Kotak and Gönczy, 2013; Morin and Bellaïche, 2011), its role in spindle assembly and in the regulation of centrosome number is far less characterized. In vertebrates, NuMA binds MTs and exhibits a poleward flow during mitosis. It also localizes at the spindle pole, where it is essential for spindle pole focusing, centrosome–spindle coupling and centrosome separation (Du et al., 2002; Lydersen and Pettijohn, 1980; Maekawa et al., 1991; Merdes et al., 1996, 2000; Silk et al., 2009; Tousson et al., 1991; Yang et al., 1992). In Drosophila, the loss of Mud function can give rise to neuroblast stem cells with supernumerary centrosomes (Izumi et al., 2006). However, the mechanism by which Drosophila Mud or mammalian NuMA controls centrosome number remains elusive.
In agreement with a role in regulating centrosome number, Mud localizes at the spindle poles (Bowman et al., 2006; Izumi et al., 2006; Siller et al., 2006; Wang et al., 2011). The function of Mud in controlling centrosome number might be dependent on either Asp or Dynein, which have been reported to function with Mud in different contexts. The C. elegans homolog of NuMA (LIN-5) binds to a complex of ASPM-1 (Asp in Drosophila and ASPM in mammals) and Calmodulin (CMD-1), and ASPM-1–CMD-1 localizes LIN-5 at the spindle pole during mitosis (van der Voet et al., 2009). Loss of Asp function in Drosophila neuroblasts can induce centrosome detachment from the spindle, which can in turn result in centrosome mis-segregation during division (Gonzalez et al., 1990; Schoborg et al., 2015). Furthermore, it is important that the respective roles of Asp and Mud in centrosome and spindle dynamics are dissected because it has been proposed that Drosophila Asp is the functional homolog of mammalian NuMA during spindle formation despite the absence of clear sequence homology between mammalian NuMA and Drosophila Asp (Goshima et al., 2005; Ito and Goshima, 2015; Manning and Compton, 2008; Morales-Mulia and Scholey, 2005; Schoborg et al., 2015; Wakefield et al., 2001; Zhang et al., 2009). The NuMA–Dynein interaction has been mainly studied in the context of mitotic spindle orientation, whereby the trimeric Gαi–LGN–NuMA (LGN is also known as GPSM2 in mammals and Pins in flies) complex promotes Dynein cortical recruitment and activity to orient the spindle (Kotak and Gönczy, 2013; Morin and Bellaïche, 2011). However, the role of the spindle pole localization of NuMA, Mud and LIN-5, and the role of Dynein in maintaining correct centrosome numbers remain less explored.
To decipher the function of Mud in centrosome number regulation, we focused on symmetrically dividing epithelial cells. We provide evidence that Asp-dependent localization of Mud at the spindle pole is required for both spindle pole focusing and the maintenance of centrosome–spindle coupling during metaphase. We also identified an Asp-independent function for Mud in centrosome separation during prophase. In Drosophila epithelial cells, we found that this Mud activity requires Dynein function and it is critical for the establishment of centrosome–spindle coupling upon NEB. Using long-term live-imaging, we directly established that the loss of Dynein, Mud or Asp function can cause centrosome mis-segregation resulting in supernumerary centrosome formation during the subsequent division.
RESULTS
mud epithelial cells exhibit spindle defects, and displaced and ectopic centrosomes
To investigate how Mud controls centrosome number, we analyzed its function during mitosis in the Drosophila pupal notum epithelium tissue (Bosveld et al., 2012). For our analyses, we viewed cells in flies hemizygous or homozygous for the mudF01205 or mud4 null alleles (hereafter referred to as mud cells) (Guan et al., 2000; Ségalen et al., 2010). As previously found on fixed neuroblasts (Izumi et al., 2006), live imaging of the centrosomal marker Spd-2:RFP in combination with the MT marker Tubulin:GFP (Tub:GFP) revealed the presence of abnormal centrosome numbers in mitotic epithelial cells deprived of Mud function (Fig. 1A,B). In all imaged cells exhibiting abnormal centrosome numbers (n=28 cells; two cells with one centrosome, 16 cells with three centrosomes, eight cells with four centrosomes, two cells with five centrosomes), we observed bipolar cell divisions with supernumerary centrosomes (cells with three or more centrosomes) either clustering (defined by their close proximity at the spindle pole) or wandering in the cytoplasm during anaphase (Fig. 1C,D; Movie 1). Furthermore, mud spindles exhibited unfocused spindle poles (Fig. 1A,B), a phenotype thus far unreported. This shows that, like NuMA (Merdes et al., 1996, 2000), Mud is required for MT focusing at the spindle pole. Interestingly, the most prevalent phenotype was the presence of cells with two centrosomes where one of the centrosomes was not correctly attached to the spindle (displaced centrosome), rather than aberrant centrosome numbers (Fig. 1A,B).
Loss of Mud affects bipolar spindle organization and centrosome numbers. (A) wt and mud mutant epithelial cells expressing Tub:GFP (green) and Spd-2:RFP (red). Yellow arrows, centrosomes coupled to the spindle pole; white arrow, centrosome displaced from the spindle pole; blue arrow, centrosome of a cell harboring only one centrosome; red arrows, ectopic centrosomes; white arrowheads, unfocused spindle poles. (B) Quantification of centrosome numbers and of the number of displaced (one normal, one displaced) centrosomes (top) or unfocused spindle poles (bottom) in wt and mud epithelial cells during mitosis. The distribution of aberrant centrosomes and spindle pole focusing in mudF01205 and mud4 are significantly different from wt (χ2 test; P<0.0004 and P<0.003 for mudF01205, and P<0.0002 and P<0.02 for mud4, respectively). (C) Time-lapse images (t=0 is set to anaphase onset) of two mud epithelial cells expressing Tub:GFP (green) and Spd-2:RFP (red), one cell with three centrosomes (top) and one cell with five centrosomes (bottom); the ectopic centrosomes (red and white arrows at t=−7 min) will either be located close to the spindle pole in anaphase (clustered centrosome, red arrows) or wandering (white arrows) in anaphase (t=0 min) (see Movie 1). Yellow arrows: centrosomes associated with the spindle pole. (D) Quantification of supernumerary centrosome behavior (wandering or clustering) in mud epithelial cells with three, four or five centrosomes during mitosis. (E,F) mud epithelial cells expressing Jup:GFP (green, left; white, middle and right) and Sas-4:RFP (red, left) showing astral MTs (arrowheads) emanating from a displaced (E) or supernumerary centrosome (F). Boxed regions (middle panels) are magnified in the right panels. (G) Time-lapse images showing MT nucleation from a displaced centrosome in mud after MT laser ablation (arrowheads, position of the laser ablation line at 3.780 s after ablation). The mean±s.d. duration to anaphase onset after successfully severing the centrosome–spindle connection was 6.41±3.8 min (n=17 cells). Centrosomes, Spd-2:RFP (red); MTs, Tub:GFP (green); septate junctions, Nrg:GFP (green). (H) Images of wt (left) and mud (right) epithelial cells expressing Tub:GFP (green), YFP:Asl (green) and H2B:RFP (red) during metaphase. The apical side is indicated. (I) Plot of the apical-basal angle of the spindle, αAB, in wt (blue) and mud epithelial cells with normal (red) or aberrant (green) centrosome numbers. αAB values in mud cells with normal and abnormal centrosomes are similar (P>0.1), but different from that in wt cells (P<10−4) as evaluated using the Kolmogorov–Smirnov (KS) test. Scale bars: 1 μm. n, cell numbers. Images are acquired from live cells.
Loss of Mud affects bipolar spindle organization and centrosome numbers. (A) wt and mud mutant epithelial cells expressing Tub:GFP (green) and Spd-2:RFP (red). Yellow arrows, centrosomes coupled to the spindle pole; white arrow, centrosome displaced from the spindle pole; blue arrow, centrosome of a cell harboring only one centrosome; red arrows, ectopic centrosomes; white arrowheads, unfocused spindle poles. (B) Quantification of centrosome numbers and of the number of displaced (one normal, one displaced) centrosomes (top) or unfocused spindle poles (bottom) in wt and mud epithelial cells during mitosis. The distribution of aberrant centrosomes and spindle pole focusing in mudF01205 and mud4 are significantly different from wt (χ2 test; P<0.0004 and P<0.003 for mudF01205, and P<0.0002 and P<0.02 for mud4, respectively). (C) Time-lapse images (t=0 is set to anaphase onset) of two mud epithelial cells expressing Tub:GFP (green) and Spd-2:RFP (red), one cell with three centrosomes (top) and one cell with five centrosomes (bottom); the ectopic centrosomes (red and white arrows at t=−7 min) will either be located close to the spindle pole in anaphase (clustered centrosome, red arrows) or wandering (white arrows) in anaphase (t=0 min) (see Movie 1). Yellow arrows: centrosomes associated with the spindle pole. (D) Quantification of supernumerary centrosome behavior (wandering or clustering) in mud epithelial cells with three, four or five centrosomes during mitosis. (E,F) mud epithelial cells expressing Jup:GFP (green, left; white, middle and right) and Sas-4:RFP (red, left) showing astral MTs (arrowheads) emanating from a displaced (E) or supernumerary centrosome (F). Boxed regions (middle panels) are magnified in the right panels. (G) Time-lapse images showing MT nucleation from a displaced centrosome in mud after MT laser ablation (arrowheads, position of the laser ablation line at 3.780 s after ablation). The mean±s.d. duration to anaphase onset after successfully severing the centrosome–spindle connection was 6.41±3.8 min (n=17 cells). Centrosomes, Spd-2:RFP (red); MTs, Tub:GFP (green); septate junctions, Nrg:GFP (green). (H) Images of wt (left) and mud (right) epithelial cells expressing Tub:GFP (green), YFP:Asl (green) and H2B:RFP (red) during metaphase. The apical side is indicated. (I) Plot of the apical-basal angle of the spindle, αAB, in wt (blue) and mud epithelial cells with normal (red) or aberrant (green) centrosome numbers. αAB values in mud cells with normal and abnormal centrosomes are similar (P>0.1), but different from that in wt cells (P<10−4) as evaluated using the Kolmogorov–Smirnov (KS) test. Scale bars: 1 μm. n, cell numbers. Images are acquired from live cells.
It has been reported that upon overexpression of the centriole duplication kinase Sak (also known as Plk4), ectopic centrosomes that do not contribute to spindle formation and fail to cluster are inactive and lose their ability to nucleate MTs (Basto et al., 2008; Sabino et al., 2015). In mud cells, both the supernumerary centrosomes and the centrosomes not attached to the spindle have astral MTs, indicating that these centrosomes can nucleate MTs. Furthermore, these ectopic and displaced centrosomes are only connected with the mitotic spindle through a small number of MTs, as revealed by using the Sas-4:RFP centrosome and Jupiter:GFP (Jup:GFP) MT markers (Fig. 1E,F). Upon laser-mediated severing of the MTs connecting the displaced or supernumerary centrosomes with the mitotic spindle during metaphase, we consistently observed a reestablishment of the MTs connecting the centrosome with the spindle (all seven cells analyzed, Fig. 1G; Movie 2). Taken together, these data suggest that in the absence of Mud function, both ectopic centrosomes and centrosomes not attached to the mitotic spindle are active as they nucleate MTs.
Since Mud regulates apical-basal (AB) spindle orientation in epithelial cells (Bergstralh et al., 2013; Bosveld et al., 2016b; David et al., 2005; Herszterg et al., 2013; Nakajima et al., 2013), and the displaced or supernumerary centrosomes still nucleate some MTs, we analyzed whether spindle orientation defects were caused by their presence. We found that mud cells with ectopic or displaced centrosomes showed a spindle mis-orientation phenotype identical to that in mud cells with two centrosomes attached to the spindle (Fig. 1H,I). The mud spindle orientation defect is therefore unlikely to be caused by centrosome number or attachment defects. In conclusion, mud cells exhibit spindle defects, and displaced or supernumerary centrosomes in epithelial tissue.
Asp regulates Mud spindle pole localization
To better define the role of Mud during spindle formation and centrosome–spindle coupling, we analyzed in detail the distribution of a functional GFP-tagged Mud (GFP:Mud) during mitosis (Bosveld et al., 2016b). Previously, we reported that GFP:Mud and endogenous Mud were detected at the cortex, tricellular junctions, the centrosome and weakly on the mitotic spindle. Interestingly, in metaphase, GFP:Mud was present around the centrosome where it localized in a ring-like structure closely opposed to the spindle MTs (the spindle pole) (Fig. 2A,B; Fig. S1A,B). This localization likely depends on a flow of GFP:Mud present on the mitotic spindle towards the centrosome (Fig. 2C,D; Fig. S1F, Movie 3). Indeed, when we photobleached GFP:Mud at one spindle pole (n=15), the signal recovered rapidly and GFP:Mud particles flowed on the mitotic spindle towards the centrosome (Fig. 2E; Movie 4). Although less frequent, we also found that GFP:Mud particles can move from the cortex towards the centrosome (Fig. S1E). In order to understand how Mud is localized to the spindle pole, we therefore aimed to analyze the mechanisms that regulate the flow of Mud protein.
The spindle pole localization of Mud is Asp dependent. (A,B) Localizations of GFP:Mud (white in A,B; green in A″,B″), Spd-2:RFP (white in A′; red in A″) and Tub:RFP (white in B′; red in B″) in epithelial cells during metaphase. Arrowheads, GFP:Mud at the spindle pole. The dashed box is magnified in the inset in A″. (C) GFP:Mud (white in C′; green in C″) particle on the mitotic spindle labeled with Tub:RFP (white in C; red in C″). Boxed regions are magnified in the bottom panels. Arrowheads, MT-associated GFP:Mud particles. (D) Displacement of a ChFP:Mud particle (red, arrowhead) on the spindle (Jup:GFP, green) and towards the spindle pole. (E) Kymograph of GFP:Mud signal upon photobleaching at one spindle pole (arrow). Arrowheads indicate the flow of some GFP:Mud particles along the spindle. The dashed line delineates the cell cortex. (F) GFP:Mud (white in F; green in F′) is absent from the spindle pole (arrowheads, Spd-2:RFP, red in F′) in asp mutant epithelial cells. (G,H) Asp:GFP (white in G,H; green in G′,H′) localization at the spindle pole (Spd-2:RFP, red in G′,H′) in wt (G) and mud (H) epithelial cells. (I) Kymograph of ChFP:Mud (white in I′; red in I″) and Asp:GFP (white in I; green in I″). Arrowheads indicate the flows of some ChFP:Mud and Asp:GFP particles. The dashed line indicates the cell cortex. (J) ChFP:Mud (white in J; red in J″) and Asp:GFP (white in J′; green in J″) colocalize at the spindle pole. Boxed regions are magnified in the right panels. (K) Kymograph of GFP:Mud (green) in an asp epithelial cell. Arrowheads, GFP:Mud particles that do not move. Centrosomes (Spd-2:RFP, red). The dashed line indicates the cell cortex. (L) Kymograph of Asp:GFP (green) in a mud epithelial cell. Centrosomes (Spd-2:RFP, red). Scale bars: 1 μm (white horizontal bars), 20 s (yellow vertical bars). Images are acquired from live cells.
The spindle pole localization of Mud is Asp dependent. (A,B) Localizations of GFP:Mud (white in A,B; green in A″,B″), Spd-2:RFP (white in A′; red in A″) and Tub:RFP (white in B′; red in B″) in epithelial cells during metaphase. Arrowheads, GFP:Mud at the spindle pole. The dashed box is magnified in the inset in A″. (C) GFP:Mud (white in C′; green in C″) particle on the mitotic spindle labeled with Tub:RFP (white in C; red in C″). Boxed regions are magnified in the bottom panels. Arrowheads, MT-associated GFP:Mud particles. (D) Displacement of a ChFP:Mud particle (red, arrowhead) on the spindle (Jup:GFP, green) and towards the spindle pole. (E) Kymograph of GFP:Mud signal upon photobleaching at one spindle pole (arrow). Arrowheads indicate the flow of some GFP:Mud particles along the spindle. The dashed line delineates the cell cortex. (F) GFP:Mud (white in F; green in F′) is absent from the spindle pole (arrowheads, Spd-2:RFP, red in F′) in asp mutant epithelial cells. (G,H) Asp:GFP (white in G,H; green in G′,H′) localization at the spindle pole (Spd-2:RFP, red in G′,H′) in wt (G) and mud (H) epithelial cells. (I) Kymograph of ChFP:Mud (white in I′; red in I″) and Asp:GFP (white in I; green in I″). Arrowheads indicate the flows of some ChFP:Mud and Asp:GFP particles. The dashed line indicates the cell cortex. (J) ChFP:Mud (white in J; red in J″) and Asp:GFP (white in J′; green in J″) colocalize at the spindle pole. Boxed regions are magnified in the right panels. (K) Kymograph of GFP:Mud (green) in an asp epithelial cell. Arrowheads, GFP:Mud particles that do not move. Centrosomes (Spd-2:RFP, red). The dashed line indicates the cell cortex. (L) Kymograph of Asp:GFP (green) in a mud epithelial cell. Centrosomes (Spd-2:RFP, red). Scale bars: 1 μm (white horizontal bars), 20 s (yellow vertical bars). Images are acquired from live cells.
Findings in C. elegans have demonstrated that LIN-5 localization at the spindle pole depends on ASPM-1, while cortical LIN-5 localization is independent of ASPM-1 (van der Voet et al., 2009). Accordingly, we found that during Drosophila mitosis, the localization of Mud at the spindle pole was lost in homozygous aspE3-null mutants (hereafter referred to as asp cells) (Gonzalez et al., 1990) (Fig. 2F; Fig. S1D). GFP-tagged Asp (Asp:GFP) colocalized with Cherry-tagged Mud (ChFP:Mud) at the spindle pole (Fig. 2G,J; Fig. S1C). Furthermore, as recently reported (Ito and Goshima, 2015; Schoborg et al., 2015), we confirmed that Asp:GFP exhibits a flow from the spindle to the centrosome (Fig. 2I; Fig. S1G,H, Movies 4 and 5). The Asp:GFP flow appears to be concomitant with the flow of ChFP:Mud (Fig. 2I). When we analyzed GFP:Mud flow in asp cells, the flow of GFP:Mud particles was impaired and GFP:Mud did not accumulate at the spindle pole (Fig. 2K; Movie 3). By contrast, Asp:GFP spindle pole localization was independent of Mud function and Asp:GFP flow appeared to be unaffected in mud cells (Fig. 2H,L; Movie 5). These data show that an Asp-dependent flow of Mud from the spindle towards the centrosome contributes to its accumulation at the spindle pole during metaphase.
Loss of centrosome–spindle coupling can drive formation of supernumerary centrosomes
In C. elegans, ASPM-1 is dispensable for spindle orientation during asymmetric cell division (van der Voet et al., 2009). Having established that Asp is necessary for the localization of Mud at the spindle pole, we tested whether the role of Mud in spindle orientation depends on its spindle pole localization during the symmetrical Drosophila epithelial cell divisions. To this end, we analyzed AB spindle orientation in asp cells. Unlike in mud cells, spindles were not mis-oriented relative to the AB axis in asp cells (Fig. 3A). Accordingly, the cortical pool of Mud appeared to be intact in asp cells (Fig. 2F; Fig. S1D). These data indicate that Mud localization at the spindle poles is not required for AB spindle orientation. In agreement with our observation that impaired spindle orientation does not correlate with defects in centrosome–spindle coupling (Fig. 1H,I), we conclude that the function of Mud in AB spindle orientation can be uncoupled from its localization at the spindle pole.
Loss of centrosome–spindle coupling drives supernumerary centrosome formation. (A) Plot of the αAB in wt and asp epithelial cells. Angles in asp cells and wt cells are not different (P>0.1, KS test). (B) Quantification of centrosome numbers, and of the fraction of cells with one normal centrosome and one displaced centrosome that showed normal segregation or mis-segregation during mitosis in the indicated genotypes. The distributions are significantly (χ2 test) different in mudRNAi (P<0.02), aspRNAi (P<0.001) and glDN (P<0.001) compared to wt, or when mudRNAi is compared to aspRNAi (P<0.001) or glDN (P<0.001). (C) Quantifications of the centrosome inheritance in wt epithelial cells after the centrosome–spindle connection was severed by laser ablation and in mudRNAi, aspRNAi or glDN cells that exhibited one normally attached centrosome and one displaced centrosome. Mis-segregation is not significantly (Fisher Exact test) different in mudRNAi (P>0.7), aspRNAi (P>0.5) and glDN (P>0.1) cells compared to that seen upon ablation in wt cells. (D–F) Time-lapse images (t=0 is set to anaphase onset) showing the centrosome mis-segregation in mudRNAi (D), aspRNAi (E) and glDN cells (F). Yellow arrows, centrosomes associated with the spindle pole; white arrows, mis-segregated centrosomes; arrowheads, unfocused spindle poles. (G) Time-lapse images of a mitotic cell in a wt tissue expressing Tub:GFP (green) and Spd-2:RFP (red) where the centrosome–spindle connection is severed by laser ablation (arrowheads, position of the laser ablation 0.5 min after laser ablation). Yellow arrow, normal centrosome; white arrow, mis-segregated centrosome. (H) mud epithelial cell labeled with Tub:GFP (green) and Spd-2:RFP (red) during two subsequent mitoses. During the first division, one cell inherits two centrosomes (white and yellow arrows). In the next division, the cell enters mitosis with two ectopic centrosomes (red arrows) and undergoes bipolar cell division. Yellow arrows, centrosomes associated with the spindle pole. Scale bars: 1 μm. n, cell numbers. Images are acquired from live cells.
Loss of centrosome–spindle coupling drives supernumerary centrosome formation. (A) Plot of the αAB in wt and asp epithelial cells. Angles in asp cells and wt cells are not different (P>0.1, KS test). (B) Quantification of centrosome numbers, and of the fraction of cells with one normal centrosome and one displaced centrosome that showed normal segregation or mis-segregation during mitosis in the indicated genotypes. The distributions are significantly (χ2 test) different in mudRNAi (P<0.02), aspRNAi (P<0.001) and glDN (P<0.001) compared to wt, or when mudRNAi is compared to aspRNAi (P<0.001) or glDN (P<0.001). (C) Quantifications of the centrosome inheritance in wt epithelial cells after the centrosome–spindle connection was severed by laser ablation and in mudRNAi, aspRNAi or glDN cells that exhibited one normally attached centrosome and one displaced centrosome. Mis-segregation is not significantly (Fisher Exact test) different in mudRNAi (P>0.7), aspRNAi (P>0.5) and glDN (P>0.1) cells compared to that seen upon ablation in wt cells. (D–F) Time-lapse images (t=0 is set to anaphase onset) showing the centrosome mis-segregation in mudRNAi (D), aspRNAi (E) and glDN cells (F). Yellow arrows, centrosomes associated with the spindle pole; white arrows, mis-segregated centrosomes; arrowheads, unfocused spindle poles. (G) Time-lapse images of a mitotic cell in a wt tissue expressing Tub:GFP (green) and Spd-2:RFP (red) where the centrosome–spindle connection is severed by laser ablation (arrowheads, position of the laser ablation 0.5 min after laser ablation). Yellow arrow, normal centrosome; white arrow, mis-segregated centrosome. (H) mud epithelial cell labeled with Tub:GFP (green) and Spd-2:RFP (red) during two subsequent mitoses. During the first division, one cell inherits two centrosomes (white and yellow arrows). In the next division, the cell enters mitosis with two ectopic centrosomes (red arrows) and undergoes bipolar cell division. Yellow arrows, centrosomes associated with the spindle pole. Scale bars: 1 μm. n, cell numbers. Images are acquired from live cells.
Next, we analyzed whether the increased number of centrosomes in mud cells was dependent on Asp function. To this end, we first characterized the centrosome number in asp cells during mitosis. We found that close to 50% of asp cells exhibited supernumerary centrosomes (Fig. S2A,B). Although the asp mutant phenotype is much stronger compared to that of mud, we initially hypothesized that Asp and Mud could share a similar function at the spindle pole, and that loss of this common function results in the generation of cells exhibiting supernumerary centrosomes. To test this hypothesis, we analyzed how the loss of function of Asp or Mud triggers the formation of supernumerary centrosomes.
Cells with supernumerary centrosomes can arise due to cytokinesis failure, centrosome over-duplication, cell fusion, de novo centriole formation or fragmentation of the pericentriolar mass/centrioles (Gönczy, 2015; Maiato and Logarinho, 2014; Nigg, 2002). In a multicellular mutant context, it is difficult to assess each of these mechanisms given that defects in centrosome numbers can arise during previous divisions and cause adverse secondary effects during subsequent divisions. To test these previously characterized mechanisms, we therefore utilized the GAL4/GAL80ts system in combination with RNA interference (RNAi) constructs under UAS control to knockdown Asp and Mud protein levels in a temperature-controlled manner (McGuire et al., 2003). When hairpin RNA expression is induced during early pupal stages, an efficient knockdown is achieved during the first division wave in the dorsal thorax epithelium (Bosveld et al., 2012). This therefore allows us to analyze the loss of Mud and Asp activity in cells that have not experienced impaired Mud and Asp function during previous divisions (hereafter referred to as aspRNAi and mudRNAi cells) (Fig. S3A,B). In both aspRNAi and mudRNAi cells, time-lapse imaging revealed that there was no cell fusion, defects in cytokinesis or splitting of centrosomes. Therefore, none of the previously well-characterized mechanisms seem to explain the appearance of supernumerary centrosomes in asp and mud cells.
Importantly, the most prevalent phenotype in both aspRNAi and mudRNAi was the presence of displaced centrosomes that were not coupled to the spindle poles (Fig. 3B). As centrosomes were frequently displaced from the spindle (20.5% in aspRNAi and 9.9% in mudRNAi), we analyzed whether such displaced centrosomes could induce centrosome mis-segregation resulting in supernumerary centrosome formation upon the subsequent mitosis. We therefore live-imaged aspRNAi or mudRNAi cells with one correctly attached centrosome and one displaced centrosome, and then analyzed the centrosome inheritance during mitosis. We found that in cells with one unattached centrosome, both centrosomes were frequently segregated into the same cell, one cell gaining an extra centrosome and leaving the other one deprived of centrosomes (31.3% in aspRNAi and 34.3% in mudRNAi, Fig. 3B–E). In aspRNAi or mudRNAi cells, we never observed mitotic cells without centrosomes, suggesting that these cells are either eliminated or do not re-enter mitosis. Similar observations were made in asp and mud mutant epithelial tissues (Fig. S2C–E). Importantly, centrosome mis-segregation could be induced in wild-type (wt) cells by severing the centrosome–spindle connection in metaphase using laser ablation (Fig. 3C,G; Movie 6). More precisely, the severing of the centrosome–spindle connection by laser ablation in 17 cells resulted in seven divisions with centrosome mis-segregation, nine divisions with correctly inherited centrosomes and one division during which both centrosomes appeared to detach from the spindle but were correctly inherited. This demonstrates that loss of centrosome–spindle coupling is sufficient to induce centrosome mis-segregation. Finally, to test whether mis-segregated centrosomes could induce ectopic centrosome formation during subsequent cell divisions, we followed mud cells in which centrosome mis-segregation occurred (n=6 cells). The six cells with an extra centrosome were not eliminated from the tissue, but continued to proliferate and entered the subsequent mitosis with four centrosomes forming bipolar spindles (Fig. 3H; Movie 7). Largely based on fixed tissue analyses, it has been speculated that centriole mis-segregation during asymmetric cell division might induce supernumerary centrosome formation upon the loss of Centrosomin function (Lucas and Raff, 2007). Taken together, our results directly show that lack of centrosome–spindle coupling leads to centrosome mis-segregation resulting in supernumerary centrosome formation during the subsequent mitosis.
Asp and Mud mediate centrosome–spindle coupling via different mechanisms
Centrosome–spindle coupling initiates after the two centrosomes have separated and coincides with NEB; coupling is then maintained until telophase (reviewed in Tanenbaum and Medema, 2010). Accordingly, by monitoring localization of Tub:GFP and Spd-2:RFP, we observed that upon entry into mitosis, the apically located centrosomes relocalize around the nucleus (Fig. S2F; Movie 8). At the nucleus, astral MTs start to form around the centrosomes and the two centrosomes separate. Finally, the two diametrically opposed centrosomes associate with the acentrosomal MTs that form upon NEB, as visualized by the appearance of the chromatin-generated MTs. Since centrosome–spindle coupling plays a key role in maintaining the correct number of centrosomes in asp and mud cells, we aimed to elucidate whether Asp and Mud function similarly by better characterizing their localization in early mitosis.
Time-lapse imaging of Asp:GFP and GFP:Mud revealed that the two proteins localize at the spindle pole prior to the formation of the chromatin-generated nuclear MTs and NEB (Fig. 4A,B; Movies 9 and 10). Asp:GFP is present at the spindle pole when the astral MT network is formed, while GFP:Mud accumulates there during centrosome separation in prophase. Importantly, in prophase, Asp:GFP levels at the poles remain constant but increase after NEB, while GFP:Mud levels continue to increase during metaphase (Fig. 4D). Furthermore, in contrast to the localization of Asp:GFP, which is diffuse within the nucleus in G2 and prophase cells, GFP:Mud is present in small punctate structures at the nuclear envelope (NE), where it partially colocalizes with the nucleopore complex protein Nup107. This NE localization of Mud is independent of Asp (Fig. 4E–G; Fig. S3C–E). The different localization and dynamics of Mud and Asp might suggest that they have distinct functions in centrosome–spindle coupling. We therefore analyzed in more detail the centrosome dynamics during the early stages of mitosis in cells deprived of Asp and Mud function.
Dynein, Asp and Mud mediate centrosome–spindle coupling through distinct mechanisms. (A–C) Asp:GFP (A), GFP:Mud (B) and Gl:GFP (C) localization during mitosis, shown in white in the top panels and green in the bottom panels. Tub:RFP is in red in the bottom panels. t=0 is set to anaphase onset. (D) Plot (mean±s.e.m.) of the Asp:GFP (red), GFP:Mud (blue) and Gl:GFP (green) signal intensities at the spindle pole normalized to their respective intensities at 8 min prior to NEB. (E) Asp:GFP nuclear localization (white, top; green, bottom) during G2 and prophase. DNA, H2B:RFP (red, bottom). (F–H) GFP:Mud localization (white, top; green, bottom) during G2 and prophase in wt (F), asp (G) and glDN (H) cells. DNA, H2B:RFP (red, bottom). Arrowheads, GFP:Mud punctate structures at the NE. (I) Quantification of the different types of centrosome–spindle coupling defects (displacement, delayed attachment or failed attachment as indicated in the diagram) in epithelial cells of indicated genotypes that showed abnormal centrosome–spindle coupling during mitosis (see Movie 11). (J) Images of a wt and mud epithelial cell at NEB expressing Tub:GFP (green) and Spd-2:RFP (red). θ is used to quantify centrosome separation and denotes the angle of the centrosomes relative to the centre of the nucleus at NEB. (K–L) Plots of the θ angles between the two centrosomes at NEB in indicated genotypes. θ=180°, maximal separation; θ<180°, incomplete separation. Defects in centrosome separation in mudRNAi cells correlate with centrosome–spindle coupling defects (delayed or failed coupling) whereas normal centrosome separation correlates with normal centrosome–spindle coupling (L, P<10−4, KS test). Scale bars: 1 μm. n, centrosome numbers (D), cell numbers (I,K,L). Images are acquired from live cells.
Dynein, Asp and Mud mediate centrosome–spindle coupling through distinct mechanisms. (A–C) Asp:GFP (A), GFP:Mud (B) and Gl:GFP (C) localization during mitosis, shown in white in the top panels and green in the bottom panels. Tub:RFP is in red in the bottom panels. t=0 is set to anaphase onset. (D) Plot (mean±s.e.m.) of the Asp:GFP (red), GFP:Mud (blue) and Gl:GFP (green) signal intensities at the spindle pole normalized to their respective intensities at 8 min prior to NEB. (E) Asp:GFP nuclear localization (white, top; green, bottom) during G2 and prophase. DNA, H2B:RFP (red, bottom). (F–H) GFP:Mud localization (white, top; green, bottom) during G2 and prophase in wt (F), asp (G) and glDN (H) cells. DNA, H2B:RFP (red, bottom). Arrowheads, GFP:Mud punctate structures at the NE. (I) Quantification of the different types of centrosome–spindle coupling defects (displacement, delayed attachment or failed attachment as indicated in the diagram) in epithelial cells of indicated genotypes that showed abnormal centrosome–spindle coupling during mitosis (see Movie 11). (J) Images of a wt and mud epithelial cell at NEB expressing Tub:GFP (green) and Spd-2:RFP (red). θ is used to quantify centrosome separation and denotes the angle of the centrosomes relative to the centre of the nucleus at NEB. (K–L) Plots of the θ angles between the two centrosomes at NEB in indicated genotypes. θ=180°, maximal separation; θ<180°, incomplete separation. Defects in centrosome separation in mudRNAi cells correlate with centrosome–spindle coupling defects (delayed or failed coupling) whereas normal centrosome separation correlates with normal centrosome–spindle coupling (L, P<10−4, KS test). Scale bars: 1 μm. n, centrosome numbers (D), cell numbers (I,K,L). Images are acquired from live cells.
Analyzing images taken at earlier time points for cells that later showed displaced centrosomes revealed that aster formation occurred normally in mudRNAi and aspRNAi cells. Interestingly, this also revealed that the defects in centrosome–spindle coupling observed upon reduction of Asp or Mud functions have different origins. In mud and mudRNAi cells, centrosome–spindle coupling was frequently not established or delayed at NEB, while in most asp and aspRNAi cells centrosome–spindle coupling initiated normally at NEB, but instead the centrosome was later displaced away from the spindle pole during metaphase (Fig. 4I; Movie 11). In our analyses, we found that the displaced centrosomes remained connected to the spindle through a few MTs (in asp, mud, aspRNAi and mudRNAi conditions). Importantly, aspRNAi expression in mud cells only mildly increased centrosome attachment failure and centrosome displacement (Fig. 4I). Combined, these findings are consistent with the notion that the establishment of centrosome–spindle coupling upon NEB is mainly regulated by Mud activity.
To better characterize the origin of failed centrosome–spindle coupling established in the absence of Mud function, we analyzed centrosome separation in more detail (Fig. 4J–L; Fig. S2G,H). In mudRNAi cells, the centrosomes failed to fully separate and centrosomes were frequently not located on opposing sites of the nucleus when NEB took place. Moreover, we found that mudRNAi cells in which centrosome–spindle coupling was not established or delayed, also displayed increased centrosome separation defects upon NEB (Fig. 4K–L).
The role of Mud in centrosome separation is mainly independent of Asp. In fact, centrosome separation at the onset of NEB was delayed in mud (P<0.002) and mudRNAi (P<0.003) cells, but not in asp (P>0.6) and aspRNAi (P>0.7) cells, and the expression of aspRNAi in mud cells only slightly enhanced the Mud loss-of-function defects observed during centrosome separation (P<0.03) (Fig. 4K; Fig. S2G,H). Since Mud is not localized at the spindle pole in cells lacking Asp function (Fig. 2F; Fig. S1D), these observations also suggest that this pre-NEB role of Mud in centrosome separation is unlikely to depend on Mud spindle pole distribution and might depend on the Asp-independent Mud distribution at the cortex and/or around the NE. Finally, the minor function of Mud in the maintenance of centrosome–spindle coupling in metaphase is largely independent of its earlier pre-NEB role in centrosome separation, since in mudRNAi cells the occurrence of centrosome displacements in metaphase was not preceded by pre-NEB defects in centrosome separation (Fig. S2I,J).
While in cultured mammalian cells and in the C. elegans embryo, NuMA and LIN-5, respectively, participate in centrosome separation (De Simone et al., 2016; Silk et al., 2009), our data now establish a link between defects in centrosome separation and the establishment of centrosome–spindle coupling. Furthermore, our data indicate that an Asp-independent and Mud-dependent centrosome separation function is required to establish the centrosome–spindle coupling upon NEB, and that Asp, with a minor contribution from Mud, is then subsequently required to maintain centrosome–spindle coupling.
Dynein-dependent centrosome positioning ensures centrosome–spindle coupling
To further analyze the role of Mud in centrosome separation, we focused on the Dynein minus-end-directed motor protein. Dynein is localized at the nuclear envelope during early prophase (Gönczy et al., 1999; Splinter et al., 2010) and binds NuMA (Merdes et al., 1996; Nguyen-Ngoc et al., 2007). Furthermore, Dynein is essential for centrosome attachment to the nuclear envelope, and the positioning and separation of the centrosomes during mitosis, as well as centrosome–spindle coupling (De Simone et al., 2016; Gönczy et al., 1999; Quintyne et al., 1999; Raaijmakers et al., 2012; Robinson et al., 1999). To determine whether Mud-dependent centrosome separation required Dynein activity, we analyzed the localization and function of the Dynactin subunit p150 (Glued; also known as DCTN1-p150). GFP-tagged Glued (Gl:GFP) localized at the cortex, cytoplasm, NE, spindle poles and asters before NEB, but was quickly lost from the poles during metaphase, suggesting that the Dynein complex could be required for centrosome–spindle coupling during the early stages of mitosis (Fig. 4C,D; Fig. 5A; Movie 12). We did not observe any changes in Gl:GFP localization in aspRNAi and mud cells. In contrast with the critical roles of Calmodulin, and the Klp10A and Klp61F kinesins, in regulating the Asp flow (Ito and Goshima, 2015; Schoborg et al., 2015), Asp:GFP and GFP:Mud spindle pole localizations were normal in cells overexpressing a dominant-negative version of Glued under GAL4/GAL80ts control (glDN) (Fig. 4H; Fig. S4A–C). Likewise, Gl:GFP localization appeared to be unaffected in asp and mud cells (Fig. S4D,E).
Dynein regulates epithelial centrosome positioning. (A) Gl:GFP (white, top; green, bottom) NE localization during G2 and prophase. DNA, H2B:RFP (red, bottom). Arrowhead, Gl:GFP at the NE. (B) Side views of wt, glDN and mud cells expressing Jup:GFP (green) and Sas-4:RFP (red), and asp cells expressing Tub:RFP (red) and Sas-4:GFP (green) showing the centrosome localization (yellow arrows) relative to the apical-basal axis. (C) Top (top panels) and side (bottom panels) views of a mitotic glDN cell expressing Jup:GFP (green) and Sas-4:RFP (red). Arrows, centrosomes. t=0 is set to anaphase onset. (D) Diagrams illustrating the three different sequential processes (centrosome positioning at the nucleus, centrosome separation, and maintenance of centrosome–spindle coupling) regulated by Gl, Mud and Asp during mitosis; the size of the symbol reflects their relative contribution. Loss of Gl or Mud mainly results in failed or delayed establishment of centrosome–spindle coupling, while loss of Asp mainly results in failure to maintain centrosome–spindle coupling. Scale bars: 1 μm. Images are acquired from live cells.
Dynein regulates epithelial centrosome positioning. (A) Gl:GFP (white, top; green, bottom) NE localization during G2 and prophase. DNA, H2B:RFP (red, bottom). Arrowhead, Gl:GFP at the NE. (B) Side views of wt, glDN and mud cells expressing Jup:GFP (green) and Sas-4:RFP (red), and asp cells expressing Tub:RFP (red) and Sas-4:GFP (green) showing the centrosome localization (yellow arrows) relative to the apical-basal axis. (C) Top (top panels) and side (bottom panels) views of a mitotic glDN cell expressing Jup:GFP (green) and Sas-4:RFP (red). Arrows, centrosomes. t=0 is set to anaphase onset. (D) Diagrams illustrating the three different sequential processes (centrosome positioning at the nucleus, centrosome separation, and maintenance of centrosome–spindle coupling) regulated by Gl, Mud and Asp during mitosis; the size of the symbol reflects their relative contribution. Loss of Gl or Mud mainly results in failed or delayed establishment of centrosome–spindle coupling, while loss of Asp mainly results in failure to maintain centrosome–spindle coupling. Scale bars: 1 μm. Images are acquired from live cells.
Consistent with previous reports for other tissues that express loss-of-function alleles for Dynein heavy chain or other components of the Dynein complex (Heald et al., 1996; Morales-Mulia and Scholey, 2005; Robinson et al., 1999), specific overexpression of glDN at the pupal stage caused displaced centrosomes in mitotic cells and such cells exhibited unfocused spindle poles (Fig. 3B,F). In agreement with these previous studies (Morales-Mulia and Scholey, 2005; Robinson et al., 1999), we also found that centrosomes from glDN cells failed to attach to the spindle or showed a delayed attachment, and glDN cells showed strong defects in centrosome separation (P<10−4) (Fig. 4I,K). Importantly, as we found for cells deprived of Asp or Mud activity, displaced centrosomes were frequently mis-segregated into one daughter cell, and glDN cells exhibited supernumerary centrosomes (Fig. 3C,F; Fig. S2B). As indicated above in the context of the Mud or Asp loss of function, we never observed mitotic cells without centrosomes upon loss of Dynein activity, suggesting that these cells are either eliminated or do not reenter mitosis.
Next, we analyzed in more detail when and how glDN expression causes defects in centrosome–spindle coupling, and whether this relates to centrosome–spindle coupling establishment or maintenance. The phenotype of glDN cells is much stronger than that in mud cells, and we frequently observed that in glDN cells the two centrosomes were not attached to the spindle and were localized basally in these cells (Fig. 5B,C). This latter observation prompted us to evaluate the interphase centrosome localization in glDN cells. In wt epithelial cells upon mitosis exit, the centrosomes resume an apical localization in interphase and remain there until the next mitosis (Jauffred et al., 2013). Consistent with these findings, we observed that in wt, mud and asp cells the interphase centrosomes were apically localized (Fig. 5B). However, centrosomes were dispersed throughout the cytoplasm in glDN cells during interphase (Fig. 5B). Accordingly, we found that the two centrosomes in glDN cells often failed to be positioned at the nuclear periphery upon entry into mitosis (17 out of 60 cells) (Fig. 5C). The mechanisms controlling centrosome positioning in epithelial cells during interphase are poorly characterized (Agircan et al., 2014; Tang and Marshall, 2012). Our results suggest that Dynein-dependent centrosome positioning at the nuclear periphery upon entry into mitosis is essential to ensure centrosome–spindle coupling.
DISCUSSION
The coupling of the centrosomes to the bipolar spindle is instrumental to genome integrity, cell division orientation and centrosome segregation. We provide evidence that at least three mechanisms operate sequentially to control centrosome–spindle coupling in symmetrically dividing epithelial cells to ensure faithful centrosome inheritance (Fig. 5D): (1) a Dynein-dependent mechanism ensuring NE positioning of centrosomes upon entry into mitosis, (2) a centrosome separation mechanism that is mainly dependent on Mud and Dynein, and which promotes the establishment of centrosome–spindle coupling, and (3) a mechanism that is mainly dependent on Asp that maintains this coupling during metaphase if it has been correctly established prior to NEB. It has been shown previously that, during mitosis, loss of centrosome–spindle coupling can result in centrosomes mis-segregating into the wrong daughter cell, and it has been proposed that such mis-segregated centrosomes could drive supernumerary centrosome formation (Schoborg et al., 2015; Chavali et al., 2016; Lucas and Raff, 2007). By using live-imaging, we have directly shown that centrosome mis-segregation due to impaired centrosome–spindle coupling upon loss of Mud function can drive supernumerary centrosome formation during the subsequent division in dividing epithelial cells. Future work will establish how the three sequentially acting mechanisms can be molecularly orchestrated in time so that faithful centrosome segregation and the maintenance of centrosome number during the subsequent mitosis can be ensured.
We have also observed that asp, mud or glDN cells do not fully abrogate centrosome–spindle coupling. This points to the existence of additional parallel or upstream actors that are involved in the establishment and maintenance of centrosome–spindle coupling and that are likely to act in conjunction with Asp, Mud and Dynein. In particular, the kinesin-5 motor protein Klp61F regulates Asp flux in Drosophila (Ito and Goshima, 2015), while the mammalian homolog Eg5 (also known as KIF11) is needed for centrosome separation (Slangy et al., 1995; Gaglio et al., 1996). Therefore, it would be interesting to analyze whether this motor protein is essential for centrosome–spindle coupling and the maintenance of centrosome number. Likewise, the human CEP215–HSET (also known as CDK5RAP2–KIFC1) complex was recently shown to be essential for centrosome–spindle coupling and centrosome number regulation (Chavali et al., 2016). The respective Drosophila homologs, Centrosomin (Cnn) and kinesin-14 (Non-claret disjunctional; Ncd) are required for centrosome–spindle coupling and centrosome number regulation (Endow et al., 1994; Lucas and Raff, 2007; Zhang and Megraw, 2007). Importantly, supernumerary centrosome formation in the absence of Cnn or CEP215–HSET function was proposed to be caused by centrosome mis-segregation (Chavali et al., 2016; Lucas and Raff, 2007). Based on our findings on the sequential activities of Dynein, Mud and Asp in the initiation and the maintenance of centrosome–spindle coupling, it will be important to analyze the interplay of Mud, Dynein or Asp with Klp61F, Cnn or Ncd, in order to further decipher the mechanisms ensuring correct centrosome–spindle coupling and centrosome segregation to prevent the formation of cells with supernumerary centrosomes.
MATERIALS AND METHODS
Fly stocks and genetics
Drosophila melanogaster stocks used were: UAS-mudRNAds, TRiP.JF02911 (Ni et al., 2011), UAS-aspRNAds, TRiP.GL00108 (Ni et al., 2011), hs-flp, act5C-FRT-yellow-FRT-GAL4, act5C-GAL4, tub-GAL80ts, UAS-GFP.DCTN1-p150 (Gl:GFP), mud4, Nrg:GFP, Jup:GFP, mRFP:Nup107, and lexAop-UAS-TagBFP:CAAX obtained from the Bloomington Stock Center. Additional stocks were: mudF01205 (Ségalen et al., 2010), aspE3 (Gonzalez et al., 1990), aspt25 (Schoborg et al., 2015), E-Cad:GFP (Huang et al., 2009), GFP:Mud (Bosveld et al., 2016b), ChFP:Mud (Bosveld et al., 2016b), ubi-GFP:Asp (Rujano et al., 2013), ubi-Tub:GFP (Grieder et al., 2000), ubi-Tub:RFP (Basto et al., 2008), ubi-Spd-2:RFP (Conduit et al., 2014), ubi-Sas-4:RFP (Lucas and Raff, 2007), ubi-Sas-4:GFP (Peel et al., 2007), ubi-H2B:RFP (Bosveld et al., 2012), UAS-D82glued (glDN) (Allen et al., 1999), ubi-YFP:Asl (Varmark et al., 2007), tub-FRT-GAL80-FRT-GAL4 UAS-mRFP (a gift from Enrique Martin-Blanco, Molecular Biology Institute of Barcelona, Spain).
Temperature gating using GAL4/GAL80ts (McGuire et al., 2003) was performed by transferring late third-instar larvae raised at 18°C to 29°C. After 24–30 h [10–12 h after pupa formation (APF), as identified by the onset of head eversion], pupae were mounted and imaged at 29°C (Bosveld et al., 2016a). All experiments with dsRNA constructs and glDN were performed using GAL4/GAL80ts gating. Because loss of Gl function is lethal (data not shown), to analyze whether continued proliferation can result in supernumerary centrosome formation, centrosome numbers in glDN cells were analyzed in pupae at 20–24 h APF during the second division. All other experiments were performed during the first round of cell divisions in pupae at 12–18 h APF (Guirao et al., 2015). Gl:GFP expression was analyzed in somatic clones induced in second-instar larvae by using the flip-out system (20 min heat-shock at 37°C) and analyzed 2 days later (Basler and Struhl, 1994). The efficacy of the RNAi lines was assessed by expressing mudRNAi or aspRNAi in flip-out clones in tissue expressing either GFP:Mud or Asp:GFP, respectively. Knockdown was analyzed in clones constitutively expressing GAL4 (20 min heat-shock at 37°C in second-instar larvae and analyzed 3–4 days later) and in clones expressing GAL4 under GAL80ts control (Bosveld et al., 2016a). Briefly, embryos or larvae were raised at 18°C for 7 days. Upon a 15 min heat-shock at 37°C, larvae were returned to 18°C. After 4–5 additional days at 18°C, late third-instar larvae were transferred to 29°C to induce RNAi expression. After 24–30 h at 29°C, pupae at 12 h APF were mounted and imaged at 29°C.
In the analyses, the time (t) equals 0 was set at: (1) the anaphase onset defined by the initial cell elongation and/or centrosome movement towards the cortex, (2) the NEB identified by the first appearance of acentrosomal chromatin-generated nuclear MTs, or (3) the beginning of the time-lapse movies (analyses of flow, ablation experiments and long-term imaging of mud tissues).
Imaging, photobleaching and laser ablations
Pupae were dissected and fixed as previously described (Ségalen et al., 2010). The primary antibody used was rabbit anti-Mud antibody (1:1000; Yu et al., 2006) and the fluorescent secondary antibody used was Alexa-Fluor-488-conjugated goat-anti-rabbit-IgG antibody (1:500, Molecular Probes). Images were collected with a confocal microscope and a 63× NA 1.4 oil DICII PL APO objective (LSM880NLO, Carl Zeiss).
Pupae were mounted for live-imaging as previously described (Bosveld et al., 2016a). Pupae were imaged at 25°C or 29°C (GAL4/GAL80ts gating) with either an inverted confocal spinning disk microscope from Nikon or Zeiss using 60× NA1.4 oil DIC N2 PL APO VC, 63× NA1.4 oil DICII PL APO or 100× NA1.4 OIL DIC N2 PL APO VC objectives, and a CoolSNAP HQ2 (Photometrics), an EMCCD Evolve (Photometrics) or a CMOS (Hamamatsu) camera. To improve the signal-to-noise ratio, movies and images for display were denoised with Fiji software (Schindelin et al., 2012).
Photobleaching of GFP:Mud and Asp:GFP was performed in a disk of 1.3 μm diameter drawn around the centrosome with a 491 nm laser at 100% power (12 iterations following five initial image acquisitions). Kymographs of GFP:Mud and Asp:GFP flows were generated by re-slicing the images (using a line width of 2 μm) through the mitotic spindle using the two centrosomes as reference points.
MT severing was performed in cells expressing Tub:GFP, Spd-2:RFP and Nrg:GFP. Images were acquired with a two-photon laser-scanning microscope (LSM880 NLO, Carl Zeiss) equipped with a 63× NA 1.4 oil DICII PL APO objective (digital zoom 3×) in single-photon bidirectional scan mode lasting δt=376 ms. MTs for ectopic or wandering centrosomes were severed following two intial acquisitons by using the Ti:Sapphire laser (Mai Tai DeepSee, Spectra Physics) at 890 nm with <100 fs pulses with a 80 MHz repetition rate set at 80% power. The centrosome–spindle connection was severed in metaphase cells expressing Tub:GFP and Spd-2:RFP typically by six consecutive ablations near the spindle pole. A confocal z-stack (14 z-sections at 0.5 μm spacing) was acquired every 30 s with a 63× NA 1.4 oil DICII PL APO objective (digital zoom 3×) in single-photon bidirectional scan mode.
To analyze supernumerary centrosome formation in mud cells, tissues expressing Tub:GFP and Spd-2:RFP were imaged every 3 min (26 z-sections at 0.9 μm spacing) by using an inverted confocal spinning disk microscope (Nikon) with a 60× NA1.4 oil DIC N2 PL APO VC objective. Once dividing cells in which centrosome mis-segregation occurred were identified, only the Tub:GFP signal was acquired for 4 h, to prevent photobleaching of Spd-2:RFP, after which both the Spd-2:RFP and Tub:GFP signal were acquired. The second division occurs 6–10 h after the first division (Guirao et al., 2015)
Image quantifications
GFP:Mud, Asp:GFP and Gl:GFP intensities at the centrosome were measured with Fiji software (Schindelin et al., 2012) every 1 min during mitosis. Measurements were performed on average projected and bleach-corrected image stacks in a 1.3 μm diameter disk centered on the centrosome identified through the Tub:RFP signal. The measured signals were corrected by subtracting the cytoplasmic signals and normalized to their intensities at 8 min prior to NEB.
AB spindle orientation (αAB) was determined by measuring the orientation of the centrosomes (marked by Spd-2:RFP or Sas-4:GFP) relative to the plane of the epithelial tissue (labeled by Tub:GFP or Tub:RFP) with a custom ImageJ plugin (available from the corresponding author on request). Centrosome numbers and αAB were measured in mitotic cells during metaphase. Centrosome separation was analyzed in cells where no obvious defects in αAB were observed, by measuring the angle formed by the centrosomes and the nucleus center at NEB (hereafter referred to as θ centrosomes–spindle) (Fig. 4J). In 17 out of 60 glDN cells, the two centrosomes were not localized at the nuclear periphery upon NEB and failed to attach to chromatin-derived spindle MTs. They were omitted from the centrosome separation analyses. Spindle defects, centrosome separation, centrosome mis-segregation and centrosome–spindle coupling, and GFP:Mud and Asp:GFP flows were analyzed in cells harboring two centrosomes. Centrosome–spindle coupling defects and centrosome mis-segregation were only analyzed in cells with two centrosomes in which one centrosome was not correctly attached to the spindle during mitosis.
Colocalization (Pearson's correlation coefficient) between GFP:Mud and mRFP:Nup107 at the NE was measured in G2 interphase cells by using Fiji software on single-plane confocal images by drawing a 0.39 μm-wide line delineating the NE.
Statistics
Statistical significance of αAB and θ centrosomes–spindle were assessed by using the Kolmogorov–Smirnov test (KS test). The Fisher exact test was used to evaluate differences in spindle pole unfocusing and in the frequency of centrosome mis-segregation in mitotic cells with one normally attached centrosome and one displaced centrosome, while the χ2 test was used for the comparison of the fraction of mitotic cells with abnormal centrosome numbers and displaced centrosomes, as well as the analyses of spindle pole unfocusing.
Acknowledgements
We thank Renata Basto (UMR 144, Institut Curie, France), Jordan W. Raff (Sir William Dunn School of Pathology, University of Oxford, UK), the Bloomington Stock Center and the Transgenic RNAi Project at Harvard Medical School for reagents; the Developmental Biology Curie imaging facility PICT-IBiSA@DBB for help with microscopy; A. Baffet, I. Cristo, D. Pinheiro and M. A. Rujano for helpful discussions and comments on the manuscript.
Footnotes
Author contributions
Conceptualization: F.B., Y.B.; Methodology: F.B., Y.B.; Validation: F.B., Y.B.; Formal analysis: F.B., Y.B.; Investigation: F.B., A.A.; Data curation: F.B.; Writing - original draft: F.B.; Writing - review & editing: F.B., A.A., Y.B.; Visualization: Y.B.; Supervision: F.B., Y.B.; Project administration: Y.B.; Funding acquisition: Y.B.
Funding
This work was supported by the European Research Council (ERC Advanced, TiMoprh, 340784), the Fondation ARC pour la Recherche sur le Cancer (SL220130607097), the Agence Nationale de la Recherche (ANR Labex DEEP; 11-LBX-0044, ANR-10-IDEX-0001-02), the Centre National de la Recherche Scientifique, the Institut National de la Santé et de la Recherche Médicale, and Institut CURIE and PSL Research University funding or grants.
References
Competing interests
The authors declare no competing or financial interests.