A phylogenetic profiling approach identifies novel ciliogenesis genes in Drosophila and C. elegans

Identification of a ciliary cluster using a refined phylogenetic profiling approach

The present study emerged from a long-standing interest in the laboratory in the molecular mechanisms underlying centriole assembly and function in ciliogenesis, using the nematode Caenorhabditis elegans and the fruit fly Drosophila melanogaster as experimental models (Dammermann et al, 2004, 2009; Serwas et al, 2017; Dobbelaere et al, 2020). Given the highly conserved nature of centrioles and cilia, we hypothesized that the early stages of ciliogenesis, including centriole to basal body conversion and initiation of axoneme extension, must also be conserved across eukaryotes. To identify candidate genes potentially involved in these processes we decided to apply a refined phylogenetic profiling approach. The original landmark studies had demonstrated the utility of comparative genomics approaches to identify ciliary genes (Avidor-Reiss et al, 2004; Li et al, 2004). However, we noted that certain proteins such as the core centriolar components CENPJ/SAS-4, STIL/SAS-5/Ana2, and SASS6 tend to be missed in orthology inferences due to their high degree of sequence divergence making them drop below significance thresholds in rapidly evolving species such as C. elegans (Carvalho-Santos et al, 2010; Hodges et al, 2010). In setting out to generate our phylogenetic profiles, we therefore chose a deliberately low BLASTp sequence similarity cutoff of 0.1, with bidirectional best match as a simple but robust (Kristensen et al, 2011) method to infer orthology. We further decided to use the human proteome as the starting point for our analysis, not only because it is the best annotated but also because vertebrates display a lower rate of sequence divergence compared with other eukaryotes (Douzery et al, 2004; Peterson et al, 2004; Raible et al, 2005). Phylogenetic profiling approaches have grown increasingly complex, with recent studies aimed at identifying ciliary genes incorporating information on the ciliary status of species, their evolutionary relationship, and/or training sets of verified ciliary genes, as well as iterative agglomerative clustering to identify orthogroups (Li et al, 2014; Dey et al, 2015; Nevers et al, 2017). We here opted instead for a simple hierarchical clustering approach, incorporating no information beyond the presence/absence pattern computed for each of the 21.732 human protein-coding genes across the genomes of 147 diverse eukaryotes representing the different branches of the eukaryotic tree, as well as 12 prokaryotic species (9 archaea, 3 bacteria) as outgroups (see Fig 1D and Dataset EV1A). In selecting genomes, care was taken to avoid overrepresenting certain branches of the tree and to exclude incompletely sequenced/annotated genomes that would create false negatives in the phylogenetic analysis. Visual inspection of the clustering output revealed a set of 386 genes whose inheritance pattern was clearly distinct from that of other highly conserved genes (see Fig 1E and Dataset EV1B). Further analysis showed this set of genes to comprise key centriolar and ciliary genes, and their presence and absence pattern across eukaryotes to reflect the presence and absence of cilia in those species (see below). Such a single ciliary cluster had not been observed in previous analyses (Li et al, 2014; Dey et al, 2015; Nevers et al, 2017). We therefore sought to confirm the robustness of our analysis by reducing the number of species representing each branch of the tree by 25, 50, and 75%. A clear ciliary cluster could still be identified in all cases, although the percentage of cluster genes remaining dropped below 80% with the most aggressive perturbation (277/386 genes when using 42 genomes, see Dataset EV1). This is in line with published reports of diminishing returns in phylogenetic profiling approaches > 100 genomes (Škunca & Dessimoz, 2015). We conclude that our phylogenetic profiling approach robustly identifies a set of genes putatively linked to cilia.

In silico characterization of the ciliary phylogenetic cluster

We next set out to examine the 386 genes in the cluster by performing a comprehensive literature and database analysis, including any information on their putative orthologs in nonvertebrate experimental models (Dataset EV2A). We classified each gene based on its level of functional characterization overall and any previously reported link to cilia on a scale of 0–3 (0—no information, 1—some indication of centriole/cilium-related function, 2—confirmed function, and 3—clearly established molecular mechanism). We further set out to provide a short description of its proposed function, as well as any potential or confirmed ciliopathy association. At time of writing, 234 of the 386 genes in the cluster had a reported function in cilium assembly or motility (i.e., with a score of 2 or above), with the other 152 genes as yet functionally uncharacterized or not linked to cilia (Fig 2A). How comprehensive is this list for known ciliary complexes and structures? As shown in Fig 2B, more than 50% and in many cases nearer 100% of components reported to be associated with basal bodies (including centriolar and transition zone components), cilium assembly (IFT and BBS components), and motility (inner and outer dynein arm components and dynein assembly factors, nexins, N-DRC, radial spoke and central apparatus components and the only recently identified microtubule inner proteins [MIPs]) can be found in the cluster, indicating essentially complete coverage of core components. Notably missing are components linked to centrosomes (PCM and centriolar satellites), subdistal appendages of centrioles and ciliary membrane trafficking, in line with the proposed recent evolutionary origin of centrosomes and the intracellular pathway of ciliogenesis discussed above (see also Dataset EV2B). Signaling genes (components of GPCR and Hh signaling pathways, TRP channel proteins) are likewise absent, reflecting the more recent association of those pathways with cilia in opisthokonts (Sigg et al, 2017). A comparison of the present cluster with the original comparative genomics studies (Avidor-Reiss et al, 2004; Li et al, 2004) and more recent refined phylogenetic profiling approaches (Li et al, 2014; Dey et al, 2015; Nevers et al, 2017) revealed considerable overlap, but also components unique to one or other type of study (Fig 2C, Dataset EV2C). The enrichment of known ciliary components in the present study suggests our approach is similarly if not more effective at identifying bona fide components compared with other recent phylogenetic profiling approaches, although a direct comparison is complicated by the inclusion of known components as “seeds” in the latter analyses.

Of the 386 genes in the ciliary cluster, all but one can be found in two or more of the six major eukaryotic groups other than unikonts shown in Fig 1D (counting cryptophytes and haptophytes and disicristates and excavates each as one group). All genes therefore likely originate in the ancestral eukaryote and may have formed part of the original complement to assemble motile cilia. As discussed by David Mitchell in his 2016 review (Mitchell, 2016), diversification into modern organisms has involved not only acquisition of new features but also loss of existing ones, including part or all of the machinery to form cilia. Within the species represented in our analysis, cilia can be inferred to have been lost 13 times independently of each other over the course of evolution (4× in opisthokonts, 1× in amoebozoa, 3× in archaeplastidae, 2× in alveolates, 3× in stramenopiles, see Dataset EV1A), which helps explain the clear definition of the ciliary cluster in Fig 1E. Clustering species based on the major ciliary subcomplexes within the cluster reveals seven other patterns (Fig 2D and Dataset EV2C). The machinery for ciliary motility but not cilia themselves has been entirely lost twice (once in nematodes, once in the arthropod Ixodes scapularis) and once in part (in the crustacean Daphnia pulex). These and other losses (e.g., of BBSome components in ciliated fungi) give rise to the grouping of genes functioning together in ciliary subcomplexes seen in Fig 1E and can be used to predict the function of known genes based purely on their pattern of conservation with a high degree of accuracy (Fig 2E and Dataset EV2D). For the 152 novel candidate genes within the cluster, ~ 2/3 can be assigned as putative ciliary assembly or motility genes, while for the remaining 1/3 the phylogenetic footprint is insufficient to make a confident prediction (Fig 2F and Dataset EV2D). But is there any experimental evidence for a ciliary function for these genes? Indeed, there is. As summarized in Fig 2G, candidate genes have been identified in large-scale proteomic, expression, and localization screens performed in different experimental models, albeit unsurprisingly (given that those selfsame screens were used to isolate and characterize ciliary genes) at a lower rate than known genes within the cluster (see also Dataset EV2E).

Functional analysis of cluster genes reveals signature ciliary phenotypes in Drosophila and C. elegans

Based on our in silico analysis, we were confident that within this cluster we had a set of genes highly enriched in key players in cilium assembly and motility. To test whether candidate genes indeed have a cilium-related function, we performed a comprehensive characterization by tissue-specific RNAi in Drosophila and mutant analysis in C. elegans. Drosophila is an attractive experimental model for such questions in that it possesses multiple types of cilia, including motile sperm flagella that have been reported to form in an IFT-independent manner in the cytoplasm, as well as sensory cilia that form in the canonical, IFT-dependent manner in olfactory and mechanosensory neurons (Han et al, 2003; Jana et al, 2016). Candidate genes can therefore be easily classified into genes required for general and compartmentalized cilium biogenesis and cilium motility. Using the GAL4-UAS system to deplete each candidate gene in the male germline (scoring male fertility), sensory neurons (assessing coordination in the negative geotaxis assay), and the whole animal (testing for nonspecific, pleiotropic phenotypes), we compared the resultant phenotypes to those for known ciliary genes within the cluster (Fig 3A–C). As can be seen in Fig 3B, centriolar and ciliary transition zone components presented moderately strong phenotypes in both tissues, consistent with their role in both canonical and compartmentalized cilium biogenesis (Basiri et al, 2014; Vieillard et al, 2016; Jana et al, 2018). In contrast, IFT components presented strong phenotypes almost exclusively in neurons, confirming and generalizing their reported role specifically in compartmentalized cilium biogenesis (Han et al, 2003). This was also true for the IFT-associated BBSome, not previously functionally characterized in the fly, although phenotypes here were generally weak. In contrast, depletion of axonemal dyneins with the exception of certain outer arm dyneins strongly affected male fertility. Gravitaxis was also frequently perturbed, reflecting the role of dynein activity in mechanosensory neurons (Zur Lage et al, 2019). In total, 88% of known ciliary genes within the cluster presented significant phenotypes in neurons, sperm or both (Fig 3C and D, see also Dataset EV3A). Remarkably, this was true also for 54 of 69 novel genes (78%). Depleting candidate genes in the whole animal tended to reveal no additional phenotypes, ruling out broader cellular functions. Given the limitations of the assays employed in this primary screen, we conclude that most if not all novel candidate genes are indeed cilium-linked.

To confirm and extend our findings, we turned to C. elegans as a complementary experimental model. Unlike Drosophila, C. elegans possesses exclusively nonmotile cilia, present in postmitotic sensory neurons, where they mediate the perception of chemosensory and mechanosensory stimuli. Compromised ciliary function leads to defects in complex behaviors such as chemotaxis, foraging, and male mating. However, critically, cilia are dispensable for viability and fertility, allowing mutants to be propagated in their homozygous state (Inglis et al, 2007). To investigate the potential role of candidate genes in cilium assembly, we took advantage of the dye-fill assay as a simple and robust method to assess cilium structural integrity (Hedgecock et al, 1985). This assay is based on incubating worms with dilute solutions of the lipophilic dye DiI. In wild type, this dye is taken up by a subset of neurons (12 amphid neurons in the head, four phasmid neurons in the tail) through their exposed ciliary endings and accumulates in their cell bodies (Fig 3E). Defects in cilium assembly result in impaired dye-filling, which can be easily scored under the fluorescence microscope. For known ciliary genes within the cluster, the dye-fill assay revealed significant defects for 60 of 68 genes (88%), aided by the low animal-to-animal variability in C. elegans, with IFT mutants presenting stronger phenotypes than those affecting the transition zone (centriolar genes cannot be assayed given their essential role in embryonic development; Fig 3F and G, see also Dataset EV3A). Remarkably, 26 of 37 candidate genes (70%) likewise presented a dye-fill phenotype. Phenotypes were generally weak, reflecting the fact that stronger, dye-fill null mutants have been previously identified in exhaustive large-scale screens using this assay (Perkins et al, 1986). With ecdysozoa and particularly nematodes having experienced widespread gene loss (Guijarro-Clarke et al, 2020), including within the ciliary cluster, not all 386 genes could be assayed in Drosophila or C. elegans. Our results are, however, consistent with our hypothesis that essentially all cluster genes function in some aspect of cilium assembly or motility.

Secondary analysis identifies candidate genes involved in the early stages of ciliogenesis

Although our primary screen in Drosophila allowed the identification and general classification of candidate genes, the precise nature of the process disrupted in each case remained to be determined. Thus, defects in male fertility could arise from a lack of sperm flagella or lack of sperm motility. Similarly, neuronal defects could arise from a lack of cilia or lack of mechanosensory function. To classify novel ciliary genes according to their function and identify candidates for further study, we employed a set of secondary assays for both sperm and neuronal cilia, comparing their depletion phenotypes to those for representatives of the various functional classes. For any gene scoring significantly in the coordination assays as well as those chosen for further characterization based on their spermatogenesis phenotype, the chordotonal organs responsible for proprioception located in the animals' legs (Fig 4A and B) were dissected and examined by DIC microscopy, as well as by immunofluorescence microscopy using antibodies to markers for the apical ciliary membrane (NompC) and basal bodies (Sas-4; Dobbelaere et al, 2020). Chordotonal organs are made up of multiple scolopidia, each of which contains a pair of ciliated nerve endings ensheathed by a glial cell and attached with their ciliary tips to the cuticle via a cap cell (Kernan, 2007). We found defects in cilium biogenesis to manifest themselves in different ways, with depletion of centriole biogenesis and IFT components resulting in the total absence or reduced numbers of cilia, while depletion of other components yielded more subtle defects including shorter (as assessed by the position of the ciliary dilation) or morphologically abnormal cilia (Fig 4B). Depletion of candidate genes primarily resulted in the latter type of defect, with eight of 36 genes examined displaying shortened or abnormal cilia (selected examples shown in Fig 4B and C, quantitation in Dataset EV3B). Supernumerary cilia, formed by conversion of daughter centrioles into mothers, a phenotype associated with depletion of Centrobin (Gottardo et al, 2015), were not observed for any candidate gene.

For any gene scoring significantly in the male fertility assay as well as those chosen for further characterization based on their neuronal phenotype, the process of spermatogenesis was examined in two ways. First, testes were dissected, fixed, and stained to visualize DNA and the actin cytoskeleton. This allowed us to monitor the different stages of sperm differentiation using nuclear morphology, as well as the formation of actin cones during individualization (Fig 4D). Second, we dissected the seminal vesicle where mature sperm are stored and filmed the movement of sperm tails using dark field microscopy (Fig 4E). Examining known ciliary genes, we found that centriolar and transition zone components gave the strongest phenotypes, arresting sperm morphogenesis at early stages of differentiation, whereas motility genes only affected later stages of differentiation (Fig 4F and G). Interestingly, sperm found in the seminal vesicle were generally found to be at least partly motile, with defects in ciliary motility primarily manifesting themselves in an empty seminal vesicle as sperm fail to move from the testes to this storage organ. Depletion of candidate genes resulted in a wide spectrum of phenotypes, with some genes resulting in defects as early as the early spermatid stage, similar to Sas4 and CEP290, while others progressed as far as the individualization stage. Overall, 15 of 18 genes examined yielded significant phenotypes in this assay (Fig 4F and G, quantitation in Dataset EV3B).

In total, 19 of 37 genes (51%) examined based on their primary screen phenotype presented clear defects in either neurons or sperm. In our further characterization, we focused on a handful of those genes which presented particularly strong phenotypes consistent with a function in early stages of ciliogenesis and/or displayed an evolutionary conservation pattern consistent with a central role in cilium biogenesis.

CABLES1, CCDC6, and TEX9 as novel basal body components related to Bld10/CEP135

Among the candidate genes in the screen that caught our attention were CABLES1, CCDC6 and TEX9, which presented significant phenotypes in neurons as well as sperm in our primary and secondary analyses (information on all genes further characterized in this study is collated in Dataset EV3C). TEX9 has previously been localized to centriolar satellites and the ciliary base in vertebrate cells (Gupta et al, 2015; Nevers et al, 2017), but remains functionally uncharacterized. CCDC6 has an extensive literature as a recurrent fusion partner for the oncogenic tyrosine kinases ROS1 and RET (Cerrato et al, 2018), yet itself remains entirely uncharacterized, although interestingly it was reported among the interactors in pulldowns of the IFT-B component Cluap1/IFT38 (Beyer et al, 2018). Finally, CABLES1 has been reported to interact with CDK5 and c-ABL and linked to control of cell proliferation and/or differentiation (Zukerberg et al, 2000; Wang et al, 2010), with no apparent connection to centrosomes or cilia. Given the severity of the neuronal and spermatogenesis phenotypes, ultrastructural analysis was performed on RNAi-depleted animals (and in the case of CABLES1 also deletion mutants) for all three candidate genes. In the case of neuronal cilia, we observed a general disorganization and shortening of ciliary axonemes, with missing and displaced doublet microtubules toward the ciliary tip (Fig 5A). More striking was the phenotype in sperm, where the central pair of microtubules was frequently missing, while the outer microtubule doublets were unaffected (Fig 5B). Such a phenotype has hitherto been reported in Drosophila only for a single gene, the centriolar cartwheel component Bld10/CEP135 (Carvalho-Santos et al, 2012). A re-examination of CEP135 by RNAi and mutant analysis confirmed these earlier observations but also revealed defects in mechanosensation and ciliary ultrastructure in chordotonal neurons, similar to CABLES1 and CCDC6 (Fig 5A). We should note that the latter finding stands in contrast to initial reports on this gene in the fly which reported no apparent defects in coordination (Mottier-Pavie & Megraw, 2009), which we attribute to the use of a stronger loss of function mutant than in the original work.

This phenotypic similarity to CEP135 prompted us to examine the localization of our novel components in the fly, with GFP transgenic fusions expressed under the general Ubq promoter. Remarkably, GFP fusions for both TEX9 and CCDC6, like CEP135 (Tian et al, 2021), localized to the center of the centriole barrel marked by Ana1, but extending further along the length of the extended spermatocyte basal body (Fig 5C). The third protein, CABLES1, also localized to spermatocyte basal bodies, albeit much more weakly, precluding detailed analysis. No localization was observed to centrosome-organizing centrioles elsewhere in the fly. Instead, all three proteins were recruited as centrioles convert into basal bodies during spermatogenesis (Fig 5D). In the course of our studies, we came across one other known ciliary cluster gene, CEP104, which shared some of the phenotypic features of the above-named components. Ultrastructural analysis of both RNAi-depleted and mutant animals revealed the same signature ciliary defects in sperm and neurons (Fig 5A and B). CEP104 in vertebrates and Chlamydomonas has been shown to be a microtubule plus-end tracking protein that translocates during ciliogenesis from the distal end of centrioles to the ciliary tip where it contributes to stability particularly of the central pair of microtubules (Jiang et al, 2012; Satish Tammana et al, 2013; Louka et al, 2018; preprint: Legal et al, 2023). Cep104 in Drosophila shares this localization, forming a central focus at the distal end of sperm basal bodies that co-localizes with the most distal populations of CCDC6 and TEX9 (Fig 5C).

We then have four (with CABLES1 potentially five) components that share a localization to the core of basal bodies, with CEP135 at the proximal end and CEP104 at the distal end, and function in ciliogenesis, particularly in formation of the central pair and organization of ciliary tips. We therefore propose that these components form part of a common ciliogenesis pathway with CEP135 at its base. Consistent with such a pathway, loss of CEP135 is independent of CABLES1, CCDC6, and TEX9 (Fig 5F), but strongly affects the localization of all other components without affecting centriole assembly itself (Fig 5E).

CFAP97D1 as part of a family of CFAP97 domain-containing proteins required for axoneme assembly/stability

CG14551/CFAP97D1 drew our attention due to its strong phenotype specifically in neurons. CG14551 is the only Drosophila ortholog of vertebrate CFAP97D1 and 2, part of a family of proteins that also includes CFAP97/KIAA1430 (Hemingway in Drosophila; Fig 6A). CFAP97D1 in mice has been found to be required for sperm flagellar axoneme integrity and male fertility (Oura et al, 2020), while its paralog CFAP97D2 remains uncharacterized. Similarly, Hemingway was found to be required for sperm flagellum assembly and ciliary motility in the fly, including in auditory sensory neurons, but with no apparent defects in coordination (Soulavie et al, 2014; vertebrate CFAP97 is presently uncharacterized). It was then surprising to see a phenotype for CFAP97D1 exclusively in neurons. Further characterization including by mutant analysis confirmed this specificity. Ultrastructural analysis of chordotonal cilia revealed broken axonemes and missing axonemal microtubule doublets, particularly toward the ciliary tip (Fig 6B), a phenotype reminiscent of what has been observed for sperm flagella with vertebrate CFAP97D1 and Drosophila Hemingway (Soulavie et al, 2014; Oura et al, 2020). CFAP97D1 was not localized in that study, while staining for Hemingway revealed no apparent localization to mature sperm. In contrast, Drosophila CFAP97D1 showed a clear ciliary localization, particularly in the area of the ciliary dilation toward the tip of chordotonal cilia (Fig 6C). Such a localization is consistent with the proposed function of CFAP97 domain-containing proteins in cilium elongation/stability. Our data suggest that this function is not specific to motile cilia/flagella. Consistent with this, the Human Protein Atlas (proteinatlas.org, Uhlén et al, 2015) notes that while CFAP97D1 is testis specific, CFAP97D2 is more broadly expressed in ciliated tissues, including those bearing exclusively nonmotile cilia such as the pancreas, while CFAP97 is essentially ubiquitous.

Unique role for CFAP299 in ciliary membrane deposition and ciliogenesis

CG8138/CFAP299 along with CG14013, CG14017, and CG3528 is one of four Drosophila orthologs of vertebrate C4orf22/CFAP299 (Fig 6A), a largely uncharacterized protein reported to be testis-enriched and tentatively linked to spermatogenesis (Li et al, 2019a), but according to the Human Protein Atlas also expressed in other ciliated tissues including the lung and oviduct. CG8138 initially presented with a strong phenotype in male fertility in the primary and secondary screen. Ultrastructural analysis of both RNAi-depleted and mutant animals revealed what to our knowledge is a unique and unprecedented phenotype: a seemingly intact flagellar axoneme but a total failure of membrane deposition to complete sperm individualization (Fig 6D). As described above, Drosophila spermatogenesis is highly unusual in that flagella form in an IFT-independent manner in the cytoplasm and are then extruded in a process reminiscent of what is believed to occur in the malarial parasite Plasmodium (Sinden et al, 1976; Han et al, 2003). The proteins responsible for axoneme elongation in the cytoplasm, whether unique to that context or shared with canonical compartmentalized ciliogenesis, are not known. Tantalizingly, CFAP299 is also conserved in Plasmodium, yet its wider phylogenetic distribution, including in humans, suggests a function also in canonical ciliogenesis. Indeed, CG8138/CFAP299 mutant animals also presented severe defects in chordotonal neurons, with highly aberrant axonemes lacking ninefold symmetry (Fig 6B). CG8138 shares extensive homology with its paralogs CG14013, CG14017, and CG3528, also putatively expressed in neurons and sperm, raising the possibility of functional redundancy. Consistent with this, weaker phenotypes were also observed for CG14013 and CG3528 (see Dataset EV3A). How CFAP299 and its orthologs contribute to ciliogenesis is as yet unclear. No specific localization was observed for N- and C-terminal GFP fusions in either sperm or neurons. However, its presence in the Chlamydomonas flagellar proteome (Pazour et al, 2005) suggests a role directly in cilium extension.

MAPK15 and ELMOD as highly conserved regulators of ciliogenesis

The final two genes, MAPK15 and ELMOD, have to different degrees been previously characterized, but were nevertheless chosen for further analysis based on the type of protein they encode, their highly conserved nature and the severity of the phenotype associated with their depletion/mutation. MAPK15 (also known as ERK7 or 8) is an atypical mitogen-activated protein kinase, atypical in the sense that it does not function as part of a classical three-tiered MAPKKK–MAPKK–MAP kinase cascade (Coulombe & Meloche, 2007). Originally linked to cellular homeostasis and the maintenance of genomic integrity (Deniz et al, 2022), MAPK15 was more recently found to also regulate ciliogenesis by promoting the apical migration of basal bodies in Xenopus, reportedly by phosphorylating the actin regulator CapZIP (Miyatake et al, 2015). In C. elegans, putative mapk-15 loss of function mutations were found to perturb neuronal morphogenesis (McLachlan et al, 2018) but also ciliary architecture as well as the localization of many ciliary proteins (Kazatskaya et al, 2017; Piasecki et al, 2017). Ciliogenesis was also reported to be affected in human RPE1 cells (Kazatskaya et al, 2017). With the kinase localizing to the ciliary base, MAPK-15 was proposed to act in a gating capacity to regulate ciliary trafficking and thereby ciliogenesis (Kazatskaya et al, 2017).

What caught our attention is that MAPK-15 is one of only 15 genes universally conserved in ciliated species and lost in nonciliated ones, a remarkably short list that also includes the core centriolar structural components SASS6 and CENPJ/SAS-4, suggesting a central role in cilium biogenesis (see Dataset EV2D). Consistent with this, MAPK15 scored strongly in the primary and secondary screens in the fly as well as in the worm. Given that available mutants in C. elegans are only partial gene deletions, a complete knockout was generated by CRISPR/Cas9-mediated gene editing. This mutant presented an even stronger phenotype (with an average of 4.3 and 0.3 dye-filled neurons in amphids and phasmids, respectively, compared to 6.9 and 0.4 for the original mutant), with cilium length and dendrite extension (indicative of transition zone dysfunction in the worm; Schouteden et al, 2015) both strongly reduced (Fig 7A and B). Ultrastructural analysis revealed that mapk-15 null mutants display severe defects in transition zone organization and axoneme elongation, with additional distortions of the periciliary membrane previously noted for loss of the microtubule organizing center at the ciliary base (Garbrecht et al, 2021; Fig 7C–E). This combination of phenotypes in the worm is remarkable in that IFT mutants do not affect transition zone organization, while loss of transition structures barely affects cilium elongation (Schouteden et al, 2015; Fig 7A and B). Indeed, no other previously characterized C. elegans gene shares this dual phenotype. The Drosophila phenotype is likewise remarkable: severe defects in both sperm and neuronal axoneme elongation and structural integrity (Fig EV2A and B). Defects in flagellar morphogenesis in particular are telling, given that this process is IFT-independent in the fly. MAPK15 thus clearly functions in IFT-independent step in axoneme elongation, consistent with its conservation in Plasmodium, which entirely lacks IFT components. MAPK15 has previously been reported to localize to the (acentriolar) ciliary base in C. elegans and basal bodies in vertebrates, as well as to the periciliary membrane compartment in C. elegans (Kazatskaya et al, 2017). We could confirm those observations for the worm (Figs 7F and EV1A) and extend them to Drosophila, where MAPK15 co-localized with the core centriolar component Sas4 to basal bodies in both chordotonal neurons and sperm (Fig EV2C and D). What is worth noting, however, is that MAPK15 does not localize to centrioles elsewhere in the worm or fly. Instead, MAPK15 is recruited to basal bodies at a time comparable to the expression of the very first ciliary components (1.5-fold stage in C. elegans embryogenesis; Serwas et al, 2017; Fig EV1B), meiosis I in Drosophila spermatogenesis (Basiri et al, 2014; Fig EV2C). In summary, then, what MAPK15 represents is a universally conserved regulator that is recruited at the onset of ciliogenesis to initiate the assembly of otherwise largely independent ciliary structures (the ciliary transition zone and axoneme).

Click here to expand this figure.

Click here to expand this figure.

ELMOD1-3, represented in Drosophila and C. elegans by a single ortholog, ELMOD/ELMD-1, are part of a family of ELMO (Engulfment and cell motility) domain-containing proteins, named after its founding member CED-12/ELMO1, which acts to promote apoptotic cell engulfment by activating CED-10/Rac1 GTPase (Wu et al, 2001; Zhou et al, 2001). In contrast, ELMOD family members are thought to act as ARF/ARL GAPs, with broad cellular effects, including on the actin cytoskeleton (Li et al, 2019b). All three vertebrate ELMOD proteins have recently been localized to basal bodies and linked to ciliogenesis, albeit with seemingly opposing effects: While loss of ELMOD1 and 3 reduced ciliation in mouse embryonic fibroblasts, loss of ELMOD2 actually resulted in an increase, with those cilia displaying abnormal morphology, including apparent splaying of axonemes and ciliary rootlet defects (Turn et al, 2021, 2022; ultrastructural analysis has so far not been performed). Common to all three perturbations is a mislocalization of ciliary proteins, suggesting ciliary trafficking is at least partly responsible for the observed defects (Turn et al, 2021, 2022). In our primary and secondary screen in Drosophila, ELMOD presented with strong phenotypes in both sperm and neurons and rather less of a phenotype in C. elegans. However, the latter result was based on a nonsense mutant of uncertain consequence (elmd-1(gk386113), Trp24Ter; Thompson et al, 2013). Similarly, weak phenotypes were reported recently for another mutant, elmd-1(syb630), a partial gene deletion (preprint: Cevik et al, 2021). We therefore obtained a full gene deletion, which yielded a much stronger phenotype (with an average of 4.9 and 0.5 dye-filled neurons in amphids and phasmids, respectively, compared to 10.4 and 3.8 for the original mutant) and was used for further analysis. Remarkably, in both C. elegans and Drosophila ELMOD presented phenotypes nearly indistinguishable from MAPK15. Thus, in C. elegans both cilium and dendrite lengths were much reduced, with axonemal, transition zone, and periciliary membrane compartment defects matching those of MAPK15 (Fig 7A–E). Likewise, in Drosophila ELMOD depletion led to defects in axoneme elongation and structural integrity in both sperm and chordotonal neurons (Fig EV2A and B). ELMOD and MAPK15 localization was also similar if not quite identical: In Drosophila, both proteins localize to basal bodies (Fig EV2C and D), while in C. elegans both display a dual localization to the ciliary base as well as the periciliary membrane compartment, although here the domain occupied by ELMOD is more extensive (Figs 7F and EV1A). Finally, neither protein localizes to centrioles, with both being recruited to basal bodies early in ciliogenesis with near-identical timing in Drosophila sperm and C. elegans neurons (Figs EV1B and EV2C). These similarities led us to posit that these proteins may function in the same molecular pathway. Consistent with this, elmd-1;mapk-15 double mutants in C. elegans presented a phenotype no more severe than that of either single mutant (Fig 7A and B). Finally, we sought to investigate whether there were any localization interdependencies between these two ciliary components. We found loss of MAPK-15 to impair proper ciliary targeting of ELMD-1, while MAPK-15 localization was not affected by loss of ELMD-1 (indeed, levels were markedly increased; Fig 7G). Our results therefore indicate that ELMD-1 functions downstream of MAPK-15 in a ciliogenesis program that serves to convert centrosome-organizing centrioles into basal bodies capable of supporting axoneme extension and transition zone assembly (Fig 7H).

Reagents and Tools table

Methods and Protocols