Drosophila PLP forms centriolar-clouds that promote centriole stability, cohesion and MT nucleation

Pericentrin is a conserved centrosomal protein whose dysfunction has been linked to several human diseases. The precise function of Pericentrin, however, is controversial. Here, we examine Drosophila Pericentrin-like- protein (PLP) function in vivo, in tissues that form both centrosomes and cilia. PLP mutant centrioles exhibit four major defects: (1) They are too short and have subtle structural defects; (2) They separate prematurely, and so overduplicate; (3) They organise fewer MTs during interphase; (4) They fail to establish and/or maintain a proper connection to the plasma membrane— although, surprisingly, mutant centrioles can still form an axoneme and recruit transition zone (TZ) proteins. We show that PLP helps to form “ pericentriolar clouds” of electron-dense material that emanate from the central cartwheel spokes and spread outward to surround the mother centriole. The partial loss of these structures may explain the complex centriole, centrosome and cilium defects we observe in PLP mutant cells.


Introduction
Centrioles are complex MT-based structures that duplicate precisely once per cell cycle when a "daughter" centriole is assembled on the side of a "mother" centriole during S-phase (Arquint and Nigg, 2016). In proliferating tissues, centrioles form centrosomes when they recruit pericentriolar material (PCM) around themselves (Conduit et al., 2015). The PCM contains several hundred proteins (Alves-Cruzeiro et al., 2013), including many involved in nucleating and organising MTs, so centrosomes function as the major MT organising centres (MTOCs) in many eukaryotic cells. In interphase, the centrioles organise relatively small amounts of PCM, but the PCM expands dramatically as cells enter mitosis in a process termed centrosome maturation (Palazzo et al., 2000;Conduit et al., 2015). In non-proliferating tissues, the centrioles often migrate to the cell cortex, where the mother centriole organises the assembly of a cilium. The cilium can be motile-so moving the cell, or generating liquid flow around the cell-or immotile, with mechano-and/or chemo-sensory functions. Defects in centriole, centrosome and cilium function have been linked to a wide range of human pathologies, including cancer, microcephaly, dwarfism and obesity (Nigg and Raff, 2009;Bettencourt-Dias et al., 2011).
Pericentrin is one of the best studied centrosomal proteins, and it has been implicated in several aspects of centriole, centrosome and cilium function (Delaval and Doxsey, 2010). In vertebrates, Pericentrin interacts with several PCM proteins, and it appears to be an important driver of mitotic centrosome maturation (Zimmerman et al., 2004;Purohit et al., 1999;Haren et al., 2009;Lee and Rhee, 2011;Lawo et al., 2012). In flies, however, the Pericentrin-likeprotein (PLP) has a more minor role in mitotic centrosome assembly (Martinez-Campos et al., 2004;Richens et al., 2015;Lerit et al., 2015), especially when compared to proteins like Cnn and Spd-2 (Cdk5Rap2/Cep215 and Cep192 in vertebrates, respectively) that form a scaffold upon which the mitotic PCM is assembled Feng et al., 2017). Instead, studies in fly cultured cells suggest that PLP helps to organise the interphase PCM (Mennella et al., 2012). PLP molecules are polarised around the mother centriole, with their C-termini linked to the centriole, and their N-termini extending outwards away from the centriole (Fu and Glover, 2012;Mennella et al., 2012). The MT nucleator γ-tubulin, and the mitotic scaffold proteins Cnn and Spd-2, are also recruited around the mother centriole during interphase, and this recruitment is abolished when PLP is depleted, although it has not been documented how this loss of interphase PCM influences the ability of these centrioles to organise MTs (Mennella et al., 2012). Interestingly, Pericentrin adopts a similarly polarised and extended conformation around the mother centriole in human cells (Lawo et al., 2012;Sonnen et al., 2012). Pericentrin also has several other functions in vertebrate cells: It is required for proper cilia function (Jurczyk et al., 2004); it is cleaved by Separase towards the end of mitosis to promote centriole disengagement (Matsuo et al., 2012;Lee and Rhee, 2012;Kim et al., 2015); it influences the DNA Damage Response (DDR) Antonczak et al., 2016).
Pericentrin defects have been linked to several human diseases (Delaval and Doxsey, 2010). Most importantly, mutations in human Pericentrin cause microcephalic osteodysplastic primordial dwarfism (MOPD) or Seckel syndrome, diseases associated with severe growth retardation during both foetal and post-foetal development (Rauch et al., 2008;Griffith et al., 2008;Bober and Jackson, 2017). Moreover, Pericentrin dysfunction has been linked to several other diseases such as mental disorders (Anitha et al., 2009), and diabetes (Huang-Doran et al., 2011). In none of these cases, however, is it understood how Pericentrin defects contribute to these complex pathologies.
Clearly it is important to understand Pericentrin function within the context of a developing organism, but such an analysis is complicated in vertebrates because centrosome and cilia defects lead to pleiotropic organismal defects.
In vertebrates, the loss of centrosomes activates a p53-dependent pathway leading to cell death or senescence (Lambrus and Holland, 2017), while the loss of cilia leads to severe developmental defects (Mitchison and Valente, 2017). Drosophila is an attractive model system for this type of study, as flies can proceed quite normally through most of development in the absence of centrioles, centrosomes and cilia (Basto et al., 2006).
Here we examine Drosophila PLP function in the Sensory Organ Precursor (SOP) lineage of the pupal notum (Hartenstein and Posakony, 1989), and in testes -tissues in which the centrioles initially form centrosomes and then form cilia (Lattao et al., 2017). Using a combination of live cell imaging, Electron Microscopy (EM) and Electron Tomography (ET) we show that PLP helps to stabilise centrioles and to nucleate centriolar interphase MTs, and it promotes centriole cohesion, and centriole docking at the plasma membrane. We also show that mother centrioles organise electron-dense "pericentriolar clouds" that extend outwards from the cartwheel spokes and surround the centriolar MT triplets. These clouds are greatly diminished in PLP mutant centrioles, and we propose that the partial loss of these structures may explain the pleiotropic centriole, centrosome and cilium defects observed in PLP mutant cells.

Centrioles separate prematurely in PLP mutant SOPs
To gain a better understanding of the role of PLP in centrosome organisation during mitosis we imaged the first cell division of either WT ( Figure 1A) or strongly-hypomorphic/null PLP mutant ( Figure 1B) SOPs. PLP mutant SOPs fell into two categories: (1) 10/17 SOPs (~59%) entered mitosis with two centrosomes but, in 5/10 of these cells, centriole separation occurred prematurely compared to WT (arrows, Figure 1A, t=15 and Figure 1B, t=9; Figure 1F); (2) 7/17 SOPs (41%) entered mitosis with centrioles that had already prematurely separated (arrow, Figure 1C, t=-36; Figure 1F). These cells formed multipolar spindles as they entered mitosis (yellow arrow, Figure   1C, t=6), but became bipolar before anaphase onset ( Figure 1C, t=12). Many of the centrosomes that had separated prematurely appeared to duplicate (red arrows, Figure 1C, t=12), so daughter cells often had too many centrosomes (red arrow, Figure 1C, t=36; Figure 1G). Surprisingly, PLP mutant SOPs were not delayed in mitosis ( Figure 1D,E), and there were no detectable defects in spindle alignment relative to the anterior-posterior body axis ( Figure 1H). Thus, cell division is relatively unperturbed in PLP mutant SOPs, but centrioles often separate prematurely, leading to spindle multipolarity and centriole amplification.

The apical positioning of centrioles is perturbed in PLP mutant SOPs
When viewed along the apical-basal axis, the spindles in WT SOPs were well aligned with the cortex during metaphase-early-anaphase (Figure 2A, t=6-12), but the posterior centriole pair initially moved basally during late-anaphase-telophase (t=15-18mins), before moving back to the apical cortex as the centrioles separated (t=21-27)-as described previously (Jauffred et al., 2013). At the end of mitosis all 4 centrioles were tightly clustered at the cortex, close to the spindle mid-body-remnant (t=30-33). In PLP mutant SOPs the spindles aligned with the cortex during metaphase-early anaphase (Figure 2B,, and the posterior centriole pair moved basally during lateanaphase-telophase (t=15-18min), but the movement of the centrioles back to the apical cortex was often delayed and erratic ( Figure 2B, t=21-33; Figure   2C,D). PLP helps to organise electron-dense "prericentriolar clouds" that surround the mother centriole.
These live-cell studies suggest that PLP mutant SOP centrioles exhibit two independent defects: (1) Mother and daughter centrioles separate prematurely either prior to mitosis (potentially allowing centrioles to overduplicate), or during mitosis; (2) The centrioles do not migrate normally to the apical cortex at the end of mitosis. To better understand the nature of these defects we analysed centriole ultrastructure by electron tomography (ET) in the pupal-notum (at 24hr after pupae formation [APF], when all mitotic divisions in the SOP lineage are completed) and also in 3 rd instar larval wingdiscs-a tissue where centriole ultrastructure is better preserved than in the pupal-notum (unpublished observations). We analysed 8 WT centrioles (all single centrioles) and 18 PLP mutant centrioles (5 single, 4 pairs, and 9 aggregated centrioles) from the pupal notum, and 7 WT (6 pairs, 1 single) and 5 PLP mutant (3 pairs, 2 single) centrioles from wing discs.
A striking feature of the tomograms of WT wing disc centrioles was the presence of "clouds" of electron dense material that surround the mother centriole; these pericentriolar clouds originated close to the outer spokes of the central cartwheel and extended outwards through the gaps between the MT blades ( Figure 3A). Similar structures were present in WT notum centrioles, but structural details were hard to discern ( Figure 3B). In PLP mutant centrioles, similar electron-dense clouds emanated from the cartwheel spokes and spread past the MT blades, but the outer regions that normally surround the mother centriole were greatly diminished. ( Figure 3C-E).
Importantly, previous super-resolution light microscopy studies identified diffuse clouds of PLP-fibrils organised around the mother centriole in an approximately nine-fold symmetric manner (Mennella et al., 2012). Taken together, these data suggest that PLP is a major component of the outer region of these pericentriolar-clouds.
In WT centriole pairs, the pericentriolar centriole clouds directly contacted the proximal end of the daughter centrioles (n=6/6 centriole pairs examined) (arrows, Figure 3A). In PLP mutant centriole pairs (n=3/3), the daughter was still in contact with the mother, but the outer region of the fibrils was greatly reduced, so the daughter centriole was positioned much closer to the mother (arrows, Figure 3C). This raises the possibility that these pericentriolar clouds normally help promote mother/daughter centriole engagement, potentially explaining why PLP mutant centriole pairs tend to separate prematurely.

PLP helps to organise centriolar MTs during interphase
Previous studies have shown that PLP promotes the organisation of the interphase PCM in cultured Drosophila cells (Mennella et al., 2012).
Interestingly, MTs were often associated with the pericentriolar clouds (arrowheads, Figure 3A,C). To test if PLP was involved in organising interphase centriole MTs we counted the number of MTs within 100nm of the centriole MT wall (pooling data from both the pupal-notum and wing-disc).
Significantly more MTs were associated with the centrioles in WT compared to PLP mutant tissues ( Figure 3F). Thus, PLP helps organise interphase centriole MTs these cells, but PLP mutant centrioles can still associate with some MTs.

Centrioles are too short and their organisation is subtly perturbed in the absence of PLP
Our EM analysis revealed two further aspects of the PLP mutant centriole phenotype. First, in two of the five singlet or mother centrioles we observed in PLP mutant wing discs several of the centriolar MT blades were missing an outer B-MT (arrows, Figure 3H)-something we have never observed in WT centrioles. This suggests that the PLP may help to assemble and/or stabilise the centriolar B-MTs. Second, centrioles were significantly shorter in the PLP mutant tissues ( Figure 3G). Thus, centriole structure is slightly perturbed in PLP mutants.

The inability of PLP mutant centrioles to interact with the cortex is correlated with their inability to nucleate interphase MTs
We measured the distance of the centrioles from the cell cortex in our ET images of the pupal notum and found that, in agreement with our live-cell studies, WT centrioles were closely associated with the cortex, while PLP mutant centrioles were more widely dispersed ( Figure 4A-C). We counted the number of MTs associated with each of these centrioles and found that PLP mutant centrioles that associated with more MTs tended to be located closer to the cortex ( Figure 4D). Thus, an inability to efficiently organise interphase MTs could contribute to the failure of mutant centrioles to become properly positioned at the cell cortex.

PLP mutant sensory cells lack cilia due to a failure to establish and/or maintain the position of the centrioles at the cell cortex.
A lack of PLP/Pericentrin leads to defects in cilia function in flies and vertebrates (Martinez-Campos et al., 2004;Jurczyk et al., 2004). We therefore used Serial Block Face-Scanning Electron Microscopy (SBF-SEM) to reconstruct 3 notum bristle sensory organs from 72hr APF WT and PLP mutants (a time when organ assembly is complete). These organs had a complex cellular organisation, but this was not detectably perturbed in PLP mutants ( Figure 5A,B). However, while centrioles, a transition zone (TZ) and a ciliary axoneme were all detectable in the 3 WT organs, none of these structures were detectable in the 3 PLP mutant organs ( Figure S1). This suggests that cilia fail to form in PLP mutants because the centrioles are not correctly positioned within the sensory neurons.
To better understand the reason for this centriole mis-positioning, we analysed centriole behaviour in the developing pupal-notum, starting at 20:50h APF when cell divisions in the SOP lineage are complete, but before the sensory cilia are starting to form. We labelled centrioles with Cep104-GFP, a protein that localizes to both centrioles and cilia (Satish Tammana et al., 2013;Jiang et al., 2012) (unpublished observations). In WT organs a prominent array of MTs starts to form at 20:50hr APF that is associated with the forming bristle cell; several centrioles are detectable at this stage, but the most brightly labelled centriole (that will form the basal body of the cilium) is slightly displaced below the MT array (arrow, Figure 5C, 20:50hr APF). By 22:30hr APF, this centriole has established a sub-apical position close to the bristle cell MT array, and this position is maintained as ciliogenesis initiates (arrow Figure 5C, 22:30-25:00hr APF; Figure 5E). In PLP mutant sensory organs this MT array is formed, and a prominently labelled centriole is initially detectable close to it (arrow, Figure 5D 20:50hr APF), but this centriole fails to establish a stable sub-apical position relative to the MT array, and it usually became further displaced as development proceeded (arrow, Figure 5D, 22:30-25:00hr APF; Figure 5E). At 30hr APF a single brightly labelled centriole pair remained positioned close to the MT array in WT sensory organs, but in PLP mutant organs this centriole pair was invariably displaced away from the MT array by several microns, and often it had prematurely separated ( Figure 5E, F). Thus, a centriole pair destined to form the basal body is specified in PLP mutant sensory neurons, but it fails to establish and/or maintain its proper position within the neuron, and so ultimately cannot organise a cilium.

PLP mutant spermatocyte centrioles are mis-oriented and fail to dock properly at the PM, but they form an axoneme and recruit TZ proteins
Apart from ciliated sensory neurons, the only other cell type that forms cilia/flagella in Drosophila are the cells of the sperm lineage (Lattao et al., 2017). The centrioles in primary spermatocytes are longer than in other fly tissues (~1µm compared to ~100-150nm) and they grow short cilia from both mother and daughter centrioles (Gonzalez et al., 1998;Lattao et al., 2017).
After meiosis, these centrioles will form the basal bodies of the sperm flagellum. WT spermatocytes contain 2 centriole pairs, but there are usually too many centrioles in PLP mutant spermatocytes and these are often short and fragmented ( Figure 6A-C), as reported previously (Martinez-Campos et al., 2004;Galletta et al., 2014).
An ET analysis revealed that WT centrioles (n=8) always organised a cilium, and the distal end of the centriole was often connected to the plasma membrane (PM), close to the position where the centriole MT triplets became axoneme MT doublets (inset, Figure 6D). In PLP mutant spermatocytes we found only a single centriole (1/13) that organised a cilium, and both the centriole and cilium were shorter than normal ( Figure 6E). We also observed a single centriole (1/13) making a side-on, rather than end-on, connection to the plasma membrane ( Figure 6F). In most cases, however, PLP mutant centrioles were found close to the PM, but were not detectably connected to the PM ( Figure 6G,H). These centrioles were often clustered, and they usually formed an axoneme-like structure, as the MTs transitioned from triplets to doublets at the presumed distal ends (arrows, Figure 6G,H). Thus, PLP mutant centrioles can form short axoneme-like structures, but they very rarely form a cilium.
We next tested whether these axoneme-like structures can recruit TZ proteins. Surprisingly, all of the TZ proteins we examined were strongly recruited to the presumed distal ends of the PLP mutant centrioles in a manner very similar to that observed in WT centrioles that formed cilia ( Figure   7A,B) (Vieillard et al., 2016;Pratt et al., 2016). Thus, PLP mutant centrioles appear to efficiently assemble axoneme-and TZ-like structures even though they are often not directly associated with the PM.

PLP helps to organise interphase centriole MTs in spermatocytes
Our studies in the wing disc and notum indicate that PLP helps to organise interphase centriolar MTs, and this defect may contribute to the failure to position mutant centrioles at the apical surface. We therefore counted centriolar MT number in WT and PLP mutant spermatocytes by EM. We typically observed many MTs around WT centrioles (n=8), and these usually had their minus ends capped close to the centriole surface ( Figure 8A-C). In PLP mutants (n=13) the number of MTs associated with the centrioles was greatly reduced, and we only observed a single MT that had its minus end capped at the centriole surface ( Figure 8A; arrow, Figure 8D).
The endogenous PLP protein is normally concentrated proximally at spermatocyte centrioles (Galletta et al., 2014) (Figure 8E), and centriolar MTs were preferentially associated with the proximal half of these centrioles ( Figure 8C and 8F). A GFP-PLP fusion protein expressed from the Ubqpromoter localises along the length of the spermatocyte centrioles ( Figure   8G), and this led to an increase in the number of MTs associated with the centrioles ( Figure 8G-H), and more of these MTs were associated with the centriole distal end ( Figure 8F, H). These data strongly suggest that PLP is involved in organising interphase centriolar MTs in spermatocytes.

The PCM proteins Spd-2 and Cnn help to recruit and/or maintain PLP at interphase centrioles, but antagonise the ability of PLP to organise MTs
In cultured cells PLP recruits PCM proteins such as Cnn and γ-tubulin to interphase centrioles (Mennella et al., 2012). In vivo, however, Cnn normally cooperates with Spd-2 to form a scaffold that recruits the mitotic PCM, and the role of these proteins in interphase PCM assembly is unclear (Conduit et al., 2015). We wondered, therefore, whether PLP functions to promote MT nucleation at interphase centrioles by recruiting Spd-2 and Cnn. Surprisingly, the centriolar levels of the PCM components Asl, Cnn, Spd-2, γ-tubulin and Polo-GFP were not detectably perturbed in PLP mutants ( Figure 9A; Figure   S2), indicating that PLP is not required to recruit and/or maintain these proteins at interphase spermatocyte centrioles.
Although PLP does not seem to recruit Spd-2 or Cnn to interphase spermatocyte centrioles, we reasoned that PLP might still promote interphase centriolar MTs by influencing their behaviour in some way. We therefore quantified PLP, g-Tub and MT nucleation in cnn;Spd-2 double mutant spermatocytes ( Figure 9B,C). Surprisingly, the centriolar levels of PLP, but not γ-tubulin, were significantly reduced in the absence of Spd-2 and Cnn ( Figure 9C-E), indicating that Spd-2 and Cnn play some part in recruiting and/or maintaining PLP at interphase centrioles. Even more surprisingly, the number of interphase centriolar MTs dramatically increased in the absence of Spd-2 and Cnn ( Figure 9F,G). Thus, at least in spermatocytes, Spd-2 and Cnn do not promote MT organisation at interphase centrosomes, but rather they appear to suppress it.

Discussion
Here we have used live-cell imaging combined with EM and ET to systematically catalogue the centriole, centrosome and cilium defects in PLP mutant pupal notum sensory organs and spermatocytes. We show that WT mother centrioles are normally surrounded by electron-dense clouds of material that extend outwards from the cartwheel spokes. In the absence of PLP, the outer region of these pericentriolar clouds is greatly reduced; this is in good agreement with previous studies that identified diffuse PLP-fibrils organised around the mother centriole in an approximately nine-fold symmetric manner (Mennella et al., 2012). We show that centrioles exhibit multiple, complex, centriole defects in the absence of PLP, several of which are in agreement with defects described in previous reports, but some of which have not been reported previously. Importantly, we propose that the loss of the outer-region of these pericentriolar clouds could explain most, if not all, of the defects we observe in PLP mutant centrioles.
The structure of centrioles is subtly perturbed in the absence of PLP: centrioles were too short and we occasionally observed centriole MT doublets that were missing an outer-B-MT. To our knowledge, these observations provide the first evidence that PLP is involved in maintaining centriole structure per se, and we suggest that the pericentriolar-clouds, which pass close by the centriolar MTs as they spread outwards from the central cartwheel, may have a role in helping to establish and/or maintain these B-MTs. It is unclear why centrioles are too short in the absence of PLP, but this could be a result of a general destabilisation of centriole structure.
Mother and daughter centrioles tend to separate prematurely in the absence of PLP, and this can lead to centriole overduplication. Thus, PLP, like Pericentrin in vertebrates (Matsuo et al., 2012;Lee and Rhee, 2012;Kim et al., 2015), helps to maintain mother/daughter centriole cohesion. The outer regions of the pericentriolar clouds appear to connect mother and daughter centrioles, potentially explaining how PLP might contribute to cohesion.
Importantly, this mechanism cannot be the only way that centriole cohesion is maintained in flies, as most centrioles still separated at the correct time in our live-cell analysis of SOPs, and we could still observe engaged mother and daughter centrioles in our EM and ET analysis, even when the pericentriolar clouds were greatly reduced. Thus, PLP contributes to mother/daughter centriole cohesion in flies, but is not essential for it. In vertebrates, PLP appears to be cleaved by Separase at the end of mitosis to promote centriole disengagement (Matsuo et al., 2012;Lee and Rhee, 2012); it will be fascinating to see if the outer-regions of the pericentriolar clouds are lost as centrioles disengage at the end of mitosis in fly cells.
It has previously been shown that PLP recruits PCM components to the interphase centriole in fly cultured cells (Mennella et al., 2012), but the role of PLP in organising interphase centriolar MTs has not been directly assessed.
Our EM studies demonstrate that PLP plays an important part in organising these MTs; the number of interphase centriolar MTs decreased in the absence of PLP (although they were not abolished) and increased when PLP was overexpressed. Most surprisingly, however, several PCM proteins appear to be recruited to interphase spermatocyte centrioles normally in the absence of PLP, indicating that interphase centrioles likely use different mechanisms to recruit PCM in cultured cells and in spermatocytes. It is unclear how PLP promotes MT organisation at interphase centrioles, but our data strongly suggests that in spermatocytes it does not do so by recruiting or regulating Spd-2 and Cnn. These two proteins cooperate to form a scaffold that is essential for mitotic centrosome assembly in flies Feng et al., 2017), but they are also present at interphase centrioles (Mennella et al., 2012;Fu and Glover, 2012). Surprisingly, however, the ability of spermatocyte centrioles to organise MTs during interphase was dramatically increased in the absence of Spd-2 and Cnn. A possible explanation for this result is that Spd-2 and Cnn normally interact with PLP at interphase centrioles, but these interactions inhibit the ability of PLP to promote MT nucleation. Taken together, our findings suggest that interphase centrioles can organise MTs independently of key proteins that organise MTs at mitotic centrosomes, and that the organisation of interphase centriolar MTs may be complicated, and may use different mechanisms in different cell types.
In the absence of PLP, centrioles often fail to migrate properly to the cell cortex, and they fail to establish and/or maintain a proper connection to the PM in cells that form cilia. This presumably explains why PLP mutants exhibit severe cilia defects (Martinez-Campos et al., 2004). This is different to the situation in vertebrate cells where cilia lacking Pericentrin are also dysfunctional, but this has been attributed to a failure to recruit IFT and Polycystin2 proteins (Jurczyk et al., 2004). It is unclear why centrioles fail to migrate to, and/or establish a connection with, the PM in PLP mutant flies, but we noticed that mutant centrioles that associated with more MTs tended to be located closer to the apical cortex. Our sample size is small, so we remain cautious in interpreting this data, but it is consistent with the possibility that an inability to properly organise interphase MTs may contribute to the inability of the mutant centrioles to migrate properly to the apical cortex and to establish and/or maintain a connection with the PM. Another possibility is that the centriole clouds can directly or indirectly interact with PM and so help to establish and/or maintain the cortical position of the centrioles.
Finally, our analysis of centriole behaviour in PLP mutant spermatocytes revealed that even when these centrioles fail to connect properly to the PM, they can still assemble an axoneme-like structure. The centriolar MTs transitioned from a triplet to a doublet organisation at their presumed distal ends (as normally occurs when centrioles form an axoneme), and, most remarkably, the presumed distal regions recruited several TZ proteins that were organised in a very similar manner to WT TZs. The ability to recruit and organise TZ proteins is very surprising, as many of these proteins contain membrane-association domains, yet these centrioles are not associated with the PM. Thus, basal bodies appear capable of organising a TZ even when they fail to dock at the PM.

Antibodies
The following primary antibodies were used: Guinea-Pig anti-Asterless
Live imaging of pupae centrioles and MTs was performed in a Nikon Eclipse TE200-E spinning disk confocal system, equipped with an EM-CCD Andor iXon+ camera, controlled by the Andor IQ2 software. Pupae were prepared for imaging as previously described (Jauffred et al., 2013). Testis squashes of whole testis and spermatocyte cysts were performed as previously reported (Roque et al., 2012;Dix and Raff, 2007). Immuno-fluorescence images were acquired in an Olympus Fluoview FV-1000 microscope and software. Each slide imaged had both a control and a mutant testis in opposite sides of the slide to guarantee equal staining. Imaging conditions were maintained between slides.

Image analysis
4D centriole tracking was performed in 8-bit converted images with Fiji (Schindelin et al., 2012) using the Trackmate plugin (Tinevez et al., 2016) using the following parameters: detection with sub-pixel localization using LOG method, threshold of 200 and estimated blob diameter set to 0.9µm. No initial threshold was applied to detections. The simple LAP tracker was used to create the tracks. Both linking max distance and gap-closing max distance were set to 2µm and Gap-closing max frame gap was set to 2. Centriole positions in 4D were extracted from Trackmate and exported into Prism (Graphpad) for analysis and plotting. Angles of division were calculated using the line function of Fiji and measured in relation to the anterior to posterior axis. Circular plots were computed using the circular package in Rstudio.
Immuno-fluorescence images were analysed in Fiji, using a purposed-written macro (available on request) to automatically segment and extract the mean sum of marker fluorescence per centriole. Asterless staining was presented as mean sum per centriole, while all other markers where presented as ratio (marker mean sum)/(Asterless mean sum) per centriole. All data analysis and plotting was performed with Prism (Graphpad).

Electron Microscopy
Samples, processing and modelling of electron microscopy and tomography data of testis centrioles and pupal samples was performed as previously published (Roque et al., 2012;Pratt et al., 2016). Data was automatically acquired in an FEITecnai T12 at 120KV using a Gatan OneView digital camera with the Navigator function of SerialEM software (Kremer et al., 1996;Mastronarde and Held, 2017). For SBFSEM samples were prepared according to (Wilke et al., 2013) with some modifications. Briefly, samples were fixed overnight in 2.5% glutaraldehyde, 4% Paraformaldehyde and 0.1% tannic acid at 4°C (from a freshly prepared 10% stock) in 0.1M PIPES buffer, pH 7.2. Samples were then washes twice for 30min in 0.1M PIPES, followed by a 30 min wash in 50mM glycine in 0.1M PIPES to quench free aldehydes, and another 30 min wash in 0.1M PIPES. Samples were then embedded in 4% low melting point agarose plus 4% porcine gelatin (Melford, cat. L1204).
Small cubes of agarose with one pupae each were then further fixed in 1.5% potassium ferricyanide and 2% osmium tetroxide in 0.1M PIPES for 1 hour at 4°C. Samples were then washed three times for 10 min in water. Next samples were incubated in thiocarbohydrazide for 20min at room temperature, followed by three 10min washes in MQ water. Samples were then incubated in 2% osmium tetroxide in MQ water for 30min at 4C. After three 10 min MQ washes, samples were incubated in 1% uranyl acetate in MQ water overnight at 4°C. Samples were washed three times for 10 min in MQ water followed by en-bloc staining with lead aspartate solution for 30 min min at room temperature. After three 10 min washes in MQ water, dehydration was performed in ice with pre-cooled solutions of 30%, 50%, 70%, 90%, 100%, 100% anhydrous ethanol for 10 min each, followed by ice-cold acetone for 10 min and a further 10 min in acetone at room temperature. Embedding was performed in acetone:Durcupan resin (Sigma cat. 44610) mix of 25% for 3 hours, 50% overnight, 75% for 3 hours and four times 100% Durcupan resin freshly prepared for 8-14 hours each. Samples were embedded in Beem capsules and cured at 60°C for 48 -72 hours. Samples were mounted onto 3View pins using conductive silver epoxy and volumes acquired in a Zeiss Merlin Compact VP FEG-SEM equipped with Gatan 3View microscope.
Sections were acquired using Digital Micrograph 2.0 every 50 nm, at 3.5KeV with 30 m aperture and VP set to 50 Pa. Images were aligned and modelled using the software package IMOD (Kremer et al., 1996).  Table   S1. In (G) a Wilcoxon signed-rank test was used to compare the median of PLP mutant to the WT value of 4.