Dilatory is a Drosophila protein related to AZI1 (CEP131) that is located at the ciliary base and required for cilium formation

Cilia are highly structured eukaryotic organelles that perform a variety of motile and sensory functions in development and physiology (Satir and Christensen, 2007; Silverman and Leroux, 2009). Studies of model organisms, such as Chlamydomonas reinhardtii and Caenorhabditis elegans, have greatly advanced our knowledge of ciliogenesis, particularly of intraflagellar transport (IFT) – the specialised trafficking process by which ciliary proteins and vesicles move through the cilium during cilium assembly at the surface of the cell (Dawe et al., 2006; Marshall, 2008; Rosenbaum and Witman, 2002; Silverman and Leroux, 2009; Stephan et al., 2007). Such studies have established that anterograde transport of ciliary cargoes involves kinesin-2 and the IFT-B subcomplex of proteins, whereas retrograde transport requires cytoplasmic dynein-2 and IFT-A proteins (Rosenbaum and Witman, 2002). IFT is required not only to construct the cilium but also for establishing and maintaining the correct localisation of receptor proteins and ion channels.

In Drosophila, cilia are confined to just two cell types – sensory neurons and spermatozoa – which facilitates in vivo analysis of ciliogenesis (Laurençon et al., 2007; Lee et al., 2008). Type I sensory neurons have sensory processes that terminate in a modified cilium, which is the site of sensory reception and transduction (Field and Matheson, 1998; Keil, 1997). The cilium of chordotonal (Ch) mechanosensory neurons has a typical 9+0 axonemal pattern of microtubule doublets. A swelling known as the ciliary dilation divides the Ch cilium into proximal and distal segments, which each have distinct compositions of ion channels (Gong et al., 2004; Kim et al., 2003; Lee et al., 2010). Moreover, in the proximal cilium, each microtubule doublet is furnished with dynein arms, which are linked to mechanically induced motility (Kavlie et al., 2010). The IFT machinery is well conserved in Drosophila (Avidor-Reiss et al., 2004), and several components have been characterised by virtue of the sensory deficits that result from their mutation (Han et al., 2003; Lee et al., 2008). Such analyses have yielded insights that were not apparent from protist studies, such as the involvement of an IFT-A protein in maintaining the functionally distinct segments of the Ch cilium (Lee et al., 2008).

Apart from Type I sensory neurons, the only other cells in flies that bear cilia are the sperm – the flagellum being a modified cilium. Drosophila spermatids arise from spermatocytes that undergo meiosis and differentiation (including flagellum assembly) within a syncytial cyst before the spermatids become bounded (‘individualised’) by separate membranes (Fuller, 1993). This radically different method of ciliogenesis does not involve IFT (Han et al., 2003; Lee et al., 2008; Sarpal et al., 2003). Thus Drosophila is a useful model for dissecting different modes of ciliogenesis.

The basal body, a membrane-tethered derivative of the centriole, is the site of assembly and remodelling of the cilium (Dawe et al., 2006). As well as anchoring the cilium, the basal body is thought to provide a selective gateway regulating the entry of ciliary proteins and vesicles by IFT, and basal body defects therefore lead to truncated cilia (Dawe et al., 2006; Marshall, 2008; Rosenbaum and Witman, 2002; Silverman and Leroux, 2009; Stephan et al., 2007). Little is yet known about how basal body structure and molecular composition relate to this function, although some recent progress has been made in protists (Craige et al., 2010; Graser et al., 2007; Lechtreck et al., 1999). Moreover, the basal body is a highly complex entity, and proteomic and comparative genomic analyses have identified large numbers of centrosome- or basal-body-associated proteins (Andersen et al., 2003; Gherman et al., 2006; Keller et al., 2005; Kilburn et al., 2007; Laurençon et al., 2007; Li et al., 2004), but little is known about how they contribute to the structure and function of this region of the cilium. The part of the cilium just distal to the basal body also appears highly specialised, with axoneme-membrane linkers and membrane elaborations. This region has been termed the transition zone, and, like the basal body, it has been proposed to play a role in IFT regulation (Craige et al., 2010; Omran, 2010; Seeley and Nachury, 2010).

So far, only one basal-body-specific protein has been studied in Drosophila (Baker et al., 2004). This protein, Uncoordinated (UNC), is required for ciliogenesis but is not conserved, so its general significance is unclear. Here, we report the characterisation of a new gene that encodes an evolutionarily conserved basal-body-associated protein. This protein, Dilatory (DILA), is a coiled-coil protein that is homologous to vertebrate AZI1 (also known as CEP131). Tagged DILA localises to the base of sensory neuron cilia. Mutations in dila result in flies with ciliary defects in both sensory neurons and sperm. Our analysis of this phenotype suggests that DILA participates in the regulation of transport at the base of the cilium in collaboration with UNC.

In a microarray screen for genes expressed during the development of Ch neurons, we identified CG1625, which we named dilatory (dila) (Cachero et al., 2011) (Fig. 1A). dila has two predicted transcripts, CG1625-RA (3381 bp) and CG1625-RB (3480 bp) (Fig. 1A). RT-PCR shows that both transcripts exist in fly embryos (L.M. and A.P.J., unpublished work). The RA cDNA sequence was identical to that predicted by FlyBase annotation, except that 21 bp of the predicted first intron were found to be part of the second exon (as is predicted already for RB). The RB cDNA seemed to differ from that defined by FlyBase annotation by including the entire predicted first intron of that transcript. CG1625-RA encodes a predicted protein of 1134 amino acids. The predicted protein has a C-terminal region that is highly conserved among Drosophila species and contains five coiled-coil domains (residues 765–1131). The protein shares 18.6% identity (30.4% similarity) with the mouse coiled-coil protein, AZI1 (Aoto et al., 1996) [also known as CEP131 (Andersen et al., 2003)]. Within the coiled-coil domains, this rises to 27.5% identity (42.2% similarity), suggesting that DILA is the AZI1 orthologue.

During embryogenesis, dila mRNA was expressed in developing and differentiating ciliated sensory neurons of both the Ch and external sensory (ES) lineages (Fig. 1B–D). Expression began in the Ch precursor cells before their division and persisted until at least the end of embryogenesis in the differentiated Ch neurons. Expression in ES neurons was generally more transient. To determine whether dila is associated with ciliogenesis in general, we examined expression in the adult testis. Here, dila mRNA was expressed in the cell bodies of elongating spermatids (Fig. 1E). In situ hybridisation did not rule out the possibility that dila was also expressed at a low level in other cells, but reporter gene analysis using the entire intergenic region upstream of dila supported the conclusion that dila is expressed exclusively in the Ch and ES lineages (Cachero et al., 2011). Moreover, by microarray analysis, dila was the highest ranked differentially expressed gene in mid-staged embryonic Ch cells (Cachero et al., 2011). Thus, although we cannot completely exclude low-level expression in other cells, the evidence suggests that dila is expressed exclusively in ciliated cell types.

To generate dila mutations, we mobilised a nearby P-element (P{GT1}CG30001^BG02674) to create imprecise excisions extending in the direction of the dila gene. Three 3′ deletions were generated, of 0.9, 1.4 and 2.0 kb, respectively, in which all or most of the DNA encoding the C-terminal coiled-coil domain was deleted (Fig. 1A). Homozygous embryos showed reduction in mRNA levels that correlated with deletion size, suggesting an increasing instability of the truncated transcripts (supplementary material Fig. S1). Most of the following analysis is based on the largest deletion, dila⁸¹, which is likely to be a strong hypomorph or a null.

For all three dila deletions, homozygous individuals reached the adult stage, and these mutant flies showed uncoordinated behaviour. Mutant flies had a dramatically reduced climbing ability in bang tests compared with wild-type controls (Fig. 2A). In addition, flies homozygous for the largest deletion, dila⁸¹, were incapable of standing or righting themselves (supplementary material Movie 1); they died very soon after eclosion. Coordinated locomotion depends on sensory feedback (Caldwell et al., 2003). Given the expression pattern of dila, these phenotypes therefore suggest a defect in sensory neuron structure or function. The uncoordination of dila mutant flies was more severe than that observed for atonal mutants, which lack Ch organs (Fig. 2A). This suggests that all types of ciliated neuron (both Ch and ES) are defective in dila mutants, as is observed for ciliogenesis genes such as nompB and rempA (Han et al., 2003; Lee et al., 2008).

The P-element used to generate dila deletions was inserted at the 5′ end of an adjacent gene, CG30001, the function of which is unknown (Fig. 1A). To establish whether CG1625 was responsible for the observed phenotypes rather than CG30001, we carried out a rescue experiment. A fragment containing the CG1625-RA open reading frame (ORF) was inserted into a UAS vector in frame with a multimer FLAG tag, and transformant lines generated. dila⁸¹ mutant flies containing UAS-dila-FLAG driven by either elavGal4 or scaGal4 showed a strong rescue of the locomotion phenotype (supplementary material Movie 1).

In dila⁸¹ mutants, Ch neurons are grossly normal in most respects (L.M. and A.P.J., unpublished) but show defects in their ciliary sensory processes. To study the ciliary morphology of Ch neurons, we examined the ultrastructure of Johnston's organ (JO) – a large array of Ch organs in the adult antenna. Each scolopale of JO houses the ciliary sensory processes of two or three Ch neurons (Fig. 2B). Cilia of dila⁸¹ Ch neurons showed a range of ultrastructural defects. Some cilia were shortened, thereby appearing missing in crosssection (Fig. 2C,D). Other cilia were present but usually had a disorganised axonemal structure, often with missing microtubule doublets (Fig. 2G,H), or were occasionally grossly abnormal in structure or orientation (Fig. 2O). In addition, the doublets mostly lacked their associated dynein arms (Fig. 2E,F), which are thought to be required for the ciliary beat that occurs in response to mechanical stimulation and acts as a signal amplification mechanism (Göpfert and Robert, 2003; Kavlie et al., 2010). The disorganised axoneme patterns suggest that other axoneme-associated components are also missing, given that loss of dynein arms alone does not alter axoneme organisation (Kavlie et al., 2010). In contrast to the ciliary compartment, basal sections showed no obvious defects in the ciliary rootlet or basal bodies and their associated fibrous structures (Fig. 2I–M). Beyond the basal bodies is a region in which the axoneme forms an electron-dense ring in transverse sections (Fig. 2J,L). In studies of Ch neurons in other insects, this region (termed the ‘ciliary base’) was found to have delicate ‘champagne-glass-shaped extensions’ connecting the axoneme to the membrane (Young, 1973). We suggest that at least part of this region corresponds to the transition zone of other cilia (Omran, 2010; Seeley and Nachury, 2010), and here we refer to it as the putative transition zone. This region showed no obvious defects in dila mutant cilia, although the fibrous structures were difficult to detect in our hands (Fig. 2J,K), In summary, in dila⁸¹ mutants, ciliary defects appear confined to the cilium beyond these basal structures.

In dila⁸¹ embryos and larvae, Ch neuronal cilia were also missing or shortened as judged by anti-horse radish peroxidase (HRP) antibody staining. The anti-HRP antibody recognizes sugar residues on multiple glycoproteins that are transported into the Ch neuron cilium and secreted into the scolopale lumen to form two bands (Jan et al., 1985) (Fig. 3A,B,F). In addition to highlighting the shortened cilia, HRP immunoreactivity was strikingly accumulated at the base of the cilia in embryos (Fig. 3C) and larvae (Fig. 3D,E). Another scolopale lumen protein, Eyeshut (EYS) (Husain et al., 2006), showed a similarly abnormal basal accumulation (Fig. 3D,E). The deposit of EYS at the base of the cilia was also seen in the IFT-associated dynein mutant, beethoven (btv) (Lee et al., 2008). Together, this and the defects observed by electron microscopy suggest a disruption of transport into the cilium.

To explore whether dila is involved in ciliogenesis generally, we examined sperm production (Han et al., 2003). We tested the fertility of flies in which peripheral nervous system (PNS) defects had been substantially rescued by UAS-dila-FLAG expression (see above). Male PNS-rescued flies showed a very low level of fertility, suggesting a primary defect in spermatogenesis (Fig. 4A). By contrast, PNS-rescued females show a normal level of fertility (L.M. and A.P.J., unpublished).

dila⁸¹ mutant testes are of normal size and shape but few, if any, mature motile sperm were evident in the 2-day-old adult testis compared with wild type (data not shown). Phase-contrast microscopy of dila⁸¹ testis squashes (n=11) showed no defects in meiotic cysts and a very low level of defects in onion-stage spermatocytes (seven examples had one to four abnormal cells in a single 64-cell cyst; data not shown). This suggests that defects occur after mitosis and meiosis and during sperm differentiation (spermiogenesis). During the elongating stage (when flagella are formed), sperm nuclei showed substantial disorganisation in the dila⁸¹ testis. In wild-type syncytial cysts, nuclei tightly associate with each other to form a well-organised cluster (Fig. 4B,C). This organisation was strongly disrupted in the dila⁸¹ testis, with some nuclei misorientated within the nuclear clusters and others dispersed along the flagellar bundles (Fig. 4D,E). In addition, substantial disorganisation of the flagellar bundles was observed (Fig. 4H).

Several basal body markers showed impaired localisation in dila mutant sperm. GFP–pericentrin/AKAP450 centrosomal targeting domain (PACT) is a marker of the rod-shaped centriole (Martinez-Campos et al., 2004), whereas γ-tubulin labels the collar-shaped centriolar adjunct surrounding the centriole (Blachon et al., 2008; Wilson et al., 1997). In wild-type sperm these locate to the apical tip of the sperm nuclei (Fig. 4C,F). In dila mutant sperm, both markers were often abnormal in location (Fig. 4D,G), with γ-tubulin often missing from the scattered nuclei (Fig. 4E). Similar defects were observed for the coiled-coil protein YURI, which is proposed to be required for correct attachment of the basal body to the nuclear membrane during differentiation (Texada et al., 2008).

Electron microscopy confirmed that sperm were often chaotically organised within their cysts in dila mutants (Fig. 4I). There were also individualisation defects, with many sperm tails sharing syncytial cytoplasm or lacking surrounding cytoplasmic membrane (Fig. 4I,J). Possibly as a consequence, many flagella were found separated from their mitochondrial derivatives (Fig. 4J). Mixed axoneme polarities were observed, consistent with flagella lying in both orientations within the same elongating spermatid cyst (Fig. 4J). However, despite these defects, flagellar axoneme structure appeared intact – an observation that is in striking contrast to Ch neuronal cilia (Fig. 4J,K).

To indicate the site of dila function in sensory neurons, we investigated the subcellular localisation of DILA protein. When expressed in larval Ch neurons using a scaGal4 driver, FLAG-tagged DILA protein (DILA–FLAG) localised to a pair of dots at the base of the cilia (Fig. 5A). To explore this, we examined the location of DILA–FLAG relative to Sas-4 (Basto et al., 2006). This protein is a centriole marker in Drosophila embryonic cells, and Sas-4 mutant flies have ciliogenesis defects that lead to the uncoordinated behaviour. We established that, after neuronal differentiation, Sas-4 was localised at the distal and proximal basal bodies (Fig. 5B). Co-labelling showed that DILA–FLAG localisation overlapped with Sas-4, but was often strongest between and beyond the Sas4-positive foci (Fig. 5B and data not shown). The most consistent and prominent localisation was in the cilium just beyond the distal basal body. This corresponds to the ciliary base region described above as being putatively equivalent to the transition zone (Fig. 5B,D). Occasionally, variable protein accumulation was observed in the cilium itself (Fig. 5A) but this appeared to be dependent on the expression strength of the DILA–FLAG protein, suggesting that it is an artifact of protein overexpression. To minimise this possibility, we examined DILA–FLAG insertion lines that had a low level of expression (i.e. leaky expression) in the absence of a Gal4 driver gene. One such line showed a barely detectable level of protein in Ch neurons. In embryos, this expression was basal-body-associated (data not shown), whereas in larvae it was specifically confined to the region distal to Sas-4, suggesting localisation at the putative transition zone (Fig. 5C).

Interestingly, dila mRNA expression was present in undifferentiated precursors and their daughter cells (Fig. 1B), suggesting that the protein might also be translated before cell cycle exit and differentiation. When expressed in precursor cells, the DILA–FLAG fusion protein colocalised at centrosomes with the pericentriolar material (PCM) marker CP190 (Fig. 6) (Butcher et al., 2004). The completeness of this colocalisation suggests that DILA–FLAG is associated with the PCM rather than with the centriole itself, which is usually observed as a defined dot surrounded by the PCM. Despite this centrosomal localisation, neuronal cell division is normal in dila⁸¹ embryos, as judged by the pattern of neurons in the late embryonic PNS (L.M. and A.P.J., unpublished). Moreover, the distribution of centrosome markers [CP190, Drosophila pericentrin-like protein (dPLP), TACC and γ-tubulin] is not altered in dila⁸¹ sensory precursor cells (Gergely et al., 2000; Martinez-Campos et al., 2004) (L.M. and A.P.J., unpublished). Thus dila is not required to organise centrosomes prior to differentiation.

UNC was reported as a new Drosophila protein that might play a role in the transformation of the centriole to the basal body (Baker et al., 2004). Like DILA, UNC is a coiled-coil protein that is required for ciliogenesis (Baker et al., 2004) and is present at sensory precursor centrosomes prior to neuronal differentiation (Cachero et al., 2011) before becoming localised to the distal and proximal basal bodies of late embryonic Ch neurons (Baker et al., 2004). However, when UNC–GFP localisation was examined in mature larval Ch neurons, we found that UNC–GFP moves distally from the basal bodies, suggesting that it also localised at the putative transition zone (Fig. 7A). These similarities between UNC and DILA led us to ask whether UNC and DILA require the function of each other for correct localisation. When expressed in dila⁸¹ mutant larvae, UNC–GFP was delocalised and, in fact, not visible in Ch cilia (n=15 larvae) (Fig. 7B,C), suggesting that either UNC–GFP protein is unstable when not localised or it is too widely dispersed to be detectable. Similarly, when DILA–FLAG was expressed in unc²⁴ mutant larvae, it was strongly delocalised: either it diffused along the distal tip of cilium or, otherwise, formed aggregates in the dendrites (Fig. 7D). Thus, DILA and UNC require each other for stable localisation at the base of the cilium. By contrast, mutation of dila or unc did not detectably alter the localisation of the core centriole and basal-body protein, Sas-4 (Fig. 7E,F). This suggests that DILA and UNC perform a function distinct from that of the core proteins of the basal body that are inherited from the centrosome – a conclusion consistent with the absence of overt basal body structural defects in dila mutant Ch neurons.

The hypothesis that DILA and UNC function together in the transition zone was explored in genetic interaction experiments using dila²⁴⁴, a moderate allele with a mildly uncoordinated phenotype (supplementary material Movie 1). When one copy of the unc gene was mutated in dila²⁴⁴ homozygote flies, their uncoordinated phenotype became much more severe (Fig. 7G). This suggests that these two proteins function together in a common pathway or process at the base of the cilium, and they might potentially interact directly.

The basal body is a complex structure with functions in membrane attachment, axoneme templating and ciliary transport docking (Marshall, 2008). Several proteins are known to be recruited to the basal body during its conversion from the centrosome, and are required for basal-body-specific functions (Bettencourt-Dias and Carvalho-Santos, 2008). Our data suggest that DILA belongs to this group of proteins: DILA–FLAG is localised to the basal body and putative transition zone, and dila mutants show defects in cilia. Mutations of general centrosomal proteins can also be expected to cause ciliary defects, in these cases due to absent or defective centrosomes or basal bodies (Basto et al., 2006; Blachon et al., 2008; Martinez-Campos et al., 2004). The evidence suggests that dila is not generally required for centrosome duplication or maturation: expression outside of sensory neurons and sperm is very low or absent, dila mutants do not show defects in PCM assembly or male meiosis, and basal body structure appears normal. Nevertheless, dila mRNA is expressed before ciliogenesis in dividing Ch neural precursors and DILA–FLAG can localise to centrosomes in these cells. It is possible that acquisition of DILA is an early step in converting the centrosome into basal body.

What is the role of DILA at the base of the cilium? In Ch neurons, several aspects of the dila mutant phenotype suggest an insufficiency in transport of components into the growing cilium. First, sensory cilia are truncated with disarrayed axonemes, which implies a shortage of building blocks, such as tubulin and associated proteins, being delivered into the cilium and is a common feature of IFT gene mutations. Second, HRP-reactive and EYS proteins remain at the ciliary base rather than being transported and secreted into the scolopale lumen. Third, axonemal microtubules mostly lack their dynein arms and other components required for their ordered arrangement. Studies in Chlamydomonas have shown that dynein arm precursors are preassembled and docked at the basal body for transport into the cilium through IFT (Hou et al., 2007). The phenotypes observed for dila all suggest impaired IFT in Ch neuron ciliogenesis. It is striking that dila mutant sperm have normally constructed axonemes, in contrast to sensory neurons. This difference correlates with the fact that IFT is not involved in sperm ciliogenesis (Han et al., 2003), thereby further implicating IFT insufficiency as the cause of the axoneme defects of sensory neurons. However, given the restricted localisation of DILA and its function in both neurons and sperm, it is unlikely that DILA forms part of the core IFT complexes. We propose that DILA ensures correct IFT function by facilitating the docking of IFT particles and vesicles (Marshall, 2008). This is emerging as a crucial regulatory function of the ciliary base. Notably, this function has been ascribed to the fibrous structures attached to both the basal body and the transition zone – both of which would be consistent with the DILA–FLAG localisation (Nigg and Raff, 2009; Silverman and Leroux, 2009). Consistent with this role, IFT components and cargoes have been found associated with these structures, presumably parked transiently (Deane et al., 2001; Stephan et al., 2007).

A few proteins are known to affect IFT by contributing directly to the structural integrity of the fibrous structures, including Chlamydomonas CEP290 (Craige et al., 2010), at the transition zone, and CEP164 (Graser et al., 2007) and Spermatozopsis P210 (Lechtreck et al., 1999) at the basal body. In dila mutant cilia, the structure of the basal body appears grossly normal, suggesting that dila plays a distinct role from these proteins, perhaps including a more direct role in docking rather than in structural integrity. The docking platform has been proposed not only to regulate selective entry into the cilium but also to be site of activation of anterograde kinesin motors, deactivation of retrograde dynein motors (Rosenbaum and Witman, 2002) and processing of tubulin (Stephan et al., 2007). We suggest that DILA participates directly in these processes, or that it might act as a scaffold to recruit the proteins that carry out these functions.

Genetic interaction and localisation evidence suggest that DILA and UNC proteins work together at the base of the cilium, perhaps defining an entity that is distinct from core structure of the basal body. Moreover, both are expressed before basal body conversion (Baker et al., 2004; Cachero et al., 2011). Interestingly, DILA and UNC are both coiled-coil proteins, which raises the possibility that they interact directly. Indeed co-dependence of localisation, such as we observed for these proteins, is a common feature of members of a protein complex, and in the context of ciliogenesis has been observed for IFT-A components and for TRPV channel subunits (Gong et al., 2004; Lee et al., 2008). dila and unc also show some phenotypic similarities with yuri, which also encodes a coiled-coil protein required for Ch neuron and sperm differentiation (Texada et al., 2008). The location of YURI protein in sensory neurons is not yet known.

The occurrence of dila phenotypes in both Ch neurons and sperm suggests a common function for DILA in these processes. One function for the basal body is to act as a template for axoneme formation. However, the fact that axoneme construction appears to proceed normally in dila mutant sperm flagella appears to rule out a role for dila in axoneme templating. An important function of the basal body in sperm differentiation appears to be maintaining the association between the nucleus and the axoneme during elongation, and this seems to underlie the defects in unc and yuri mutants, possibly owing to a failure in centrosome to basal body conversion or in correctly orientating basal body positioning during individualisation (Baker et al., 2004; Texada et al., 2008). Although this phenotype remains to be explored fully, dila phenotypes, such as the disorganised sperm orientations within a cyst, are consistent with such a function. Ciliogenesis in sperm differentiation is highly unusual, and sperm differentiation itself is highly specialised. It might be that DILA and other proteins have been recruited to a different role required for a highly derived basal body function during spermiogenesis. Alternatively, similar to the neuronal phenotype, dila might be required to target or transport components required for sperm individualisation.

UNC and YURI have no clear orthologues in vertebrates, although it has been suggested that UNC is functionally equivalent to human OFD1, a centrosomal and basal body coiled-coil protein mutated in the ciliopathy, oro-facial-digital syndrome 1 (Baker et al., 2004; Romio et al., 2004). In contrast to UNC and YURI, DILA is predicted to be orthologous to human AZI1. AZI1 was first discovered in the sperm of mouse (Aoto et al., 1996) and is implicated to be present at the human centrosome (as CEP131) in a proteomic analysis (Andersen et al., 2003). In a recent study, morpholino knockdown of CEP131 in zebrafish resulted in developmental phenotypes consistent with defective cilia (Wilkinson et al., 2009). Conservation of protein structure, localisation and mutant phenotype imply that DILA, and its AZI1 homologues, is likely to be a fundamental component of the ciliary base. Therefore, our study suggests that AZI1 is a candidate gene for human ciliopathies and that it might interact with OFD1 during ciliogenesis.

Flies are maintained on standard medium at 25°C. We used OreR or w¹¹¹⁸ as wild-type controls. The P-element insertion line BG02674 and unc²⁴ were obtained from the Bloomington Stock Center (Indiana University, Bloomington, IN). unc-gfp flies were described previously (Baker et al., 2004). The drivers used were scaGal4 and elavGal4 (Bloomington). PACT–GFP flies have been described previously (Martinez-Campos et al., 2004).

Imprecise excision mutagenesis of dila was performed using P element line P{GT1}CG30001^BG02674 within CG30001, a gene downstream of dila. The P element was mobilised by crossing to a Δ2–3 transposase line. Male progeny from females containing both the P element and the transposase were chosen as candidate deletion alleles, balanced over CyO, Kr-GFP, then made homozygous and the presence of deletions determined by PCR using primers dila del A and dila del B (supplementary material Table S1). Approximately 300 excision lines were analyzed and three deletions were detected, named dila⁵⁷, dila²⁴⁴ and dila⁸¹.

Primers used to clone the full length CG1625-RA cDNA are listed in supplementary material Table S1 (CG1625 cDNA 5′ and 3′). RNA was extracted from wild-type adult flies using the RNeasy Plus Mini kit (Qiagen). First-strand synthesis was performed using the ImProm-II reverse transcription system (Promega). The PCR product was cloned into pSC-A. To construct the pUAS-FLAG-dila fusion, three repeats of FLAG (DYKDDDDK) were introduced into the N- or C-terminus of the CG1625-RA cDNA in pSC-A using site-directed mutagenesis (QuickChange IIXL, Stratagene). After sequencing, the modified FLAG–DILA fusion fragments were cloned into pUAST digested with EcoRI and XhoI. The resulting construct was transformed into Drosophila embryos by microinjection. Both N- and C-terminal fusions rescue the dila⁸¹ behavioural phenotype. Experiments shown in Fig. 5 employ the C-terminal fusion.

Embryos at the desired stage were collected on grape agar plates, dechorionated and fixed for 15 minutes with 3.7% formaldehyde in PBS. Adult testes were dissected in PBS. Samples subjected for RNA in situ hybridisation were fixed in 3.7% formaldehyde in PBS for 20 minutes, and stored in methanol at −20°C overnight before use. For antibody staining, testes were squashed and prepared as previously described (Blachon et al., 2008). For staining of larval chordotonal organs, third-instar larvae were dissected in a drop of PBS. The larval pelt was fixed with 3.7% formaldehyde in PBS for 20 minutes.

dila probes were synthesised by PCR amplification with dila 5′ and dila T7 primer sets (supplementary material Table S1), using cDNA extracted from wild-type embryos as template. The probe was digoxigenin labelled (Roche) and purified with RNeasy Mini Spin Columns kit (Qiagen). In situ hybridisation was carried out on the embryos and testes by standard protocols. For embryos, homozygote dila mutant embryos were distinguished from heterozygote siblings by the absence of Kr-GFP expression from a transgene carried on the balancer (CyO, Kr-GFP).

Primary antibodies used in these studies were as follows: mouse MAb22C10 (1:200), and mouse anti-EYS antibody (1:200) (both from Developmental Biology Hybridoma Bank, Iowa), mouse and rabbit anti-GFP antibody (1:500, Molecular probes, Invitrogen) and mouse anti-FLAG antibody (1:500, Sigma, F1804). The following centrosome antibodies were also used: rabbit Sas-4 (1:250) (Martinez-Campos et al., 2004), rabbit CP190 (1:500) (Kellogg and Albert, 1992) and mouse γ-tubulin (1:500, Sigma, GTU88). Secondary antibodies (1:500) were from Molecular Probes (Invitrogen). Samples were mounted with vectashield (Vector laboratories). Confocal microscopy was carried out using Zeiss PASCAL and Leica TCS SP2 microscopes.

Transmission electron microscopy analysis of testis and Johnston's Organ from the adult antenna were performed as previously described (Lee et al., 2008). Sample embedding, sectioning and electron microscope examination were performed by Electron Microscopy Research Services of Newcastle University Medical School using a Philips CM100 Compustage (FEI) microscope with images collected using an AMT CCD camera (Deben).

For the bang test, 2- to 4-day-old flies were collected under CO₂ after eclosion. Assays were carried out in a 100 ml cylinder. After a few minutes of acclimatisation, the cylinder was manually banged twice on a padded surface. Flies were left for 5 minutes before counting the number of flies above the 50 ml mark. Tests were repeated three times for each strain and flies were rested for 5 minutes between tests. For male fertility tests, individual males (2-day-old) were placed with two virgin OreR females in vials for 7 days at 25°C. The parent flies were then removed and all resulting F1 progeny was counted. Seven control crosses and 14 dila⁸¹ mutant crosses were analysed. To test the genetic interaction between dila and unc, the uncoordinated phenotype of adult flies was scored according to the following: class I flies are able to stand and walk properly; class II flies walk in an uncoordinated way, falling easily and with wings held erect; class III flies are completely unable to walk and become stuck in the food.

We thank members of the Jarman laboratory for help with experiments, especially Sadie Kemp and Petra zur Lage. We are grateful to Giuseppa Pennetta for experimental advice, to Jordan Raff for providing antibodies and to Mar Carmena for expert advice on analysis of meiosis. We thank Tracey Davies of Newcastle University Medical School Electron Microscopy Research Services for excellent technical services. This work was funded by The Wellcome Trust (077266). Deposited in PMC for immediate release.