Mutation of NEKL-4/NEK10 and TTLL genes suppress neuronal ciliary degeneration caused by loss of CCPP-1 deglutamylase function

Ciliary microtubules are subject to post-translational modifications that act as a “Tubulin Code” to regulate motor traffic, binding proteins and stability. In humans, loss of CCP1, a cytosolic carboxypeptidase and tubulin deglutamylating enzyme, causes infantile-onset neurodegeneration. In C. elegans, mutations in ccpp-1, the homolog of CCP1, result in progressive degeneration of neuronal cilia and loss of neuronal function. To identify genes that regulate microtubule glutamylation and ciliary integrity, we performed a forward genetic screen for suppressors of ciliary degeneration in ccpp-1 mutants. We isolated the ttll-5(my38) suppressor, a mutation in a tubulin tyrosine ligase-like glutamylase gene. We show that mutation in the ttll-4, ttll-5, or ttll-11 gene suppressed the hyperglutamylation-induced loss of ciliary dye filling and kinesin-2 mislocalization in ccpp-1 cilia. We also identified the nekl-4(my31) suppressor, an allele affecting the NIMA (Never in Mitosis A)-related kinase NEKL-4/NEK10. In humans, NEK10 mutation causes bronchiectasis, an airway and mucociliary transport disorder caused by defective motile cilia. C. elegans NEKL-4 localizes to the ciliary base but does not localize to cilia, suggesting an indirect role in ciliary processes. This work defines a pathway in which glutamylation, a component of the Tubulin Code, is written by TTLL-4, TTLL-5, and TTLL-11; is erased by CCPP-1; is read by ciliary kinesins; and its downstream effects are modulated by NEKL-4 activity. Identification of regulators of microtubule glutamylation in diverse cellular contexts is important to the development of effective therapies for disorders characterized by changes in microtubule glutamylation. By identifying C. elegans genes important for neuronal and ciliary stability, our work may inform research into the roles of the tubulin code in human ciliopathies and neurodegenerative diseases.


Introduction
Cilia and flagella are antenna-like organelles that are conserved from algae to humans and play important roles in sensation, intercellular communication, development, and homeostasis [1]. The core of both cilia and flagella is the axoneme, a conserved microtubule structure with a ring of nine doublets that may surround inner microtubule singlets in motile cilia [1]. The Tubulin Code hypothesis suggests that post-translational modifications, such as glutamylation, acetylation, glycylation, detyrosination, and others, can increase the diversity of microtubules beyond incorporation of different tubulin isotypes for structural and functional specialization [2].
Glutamylation, or addition of side-chains of the amino acid glutamate, of tubulin tails is emerging as an important factor in organizing the microtubule cytoskeleton [3,4]. Microtubules in neurons and cilia can be reversibly glutamylated. Although the function of glutamylation and other post-translational modifications is not completely understood, enzymes that regulate post-translational glutamylation of microtubules are known. Cytosolic carboxypeptidases (CCPs) such as CCP1 deglutamylate microtubules [5,6], opposing the activity of tubulin tyrosine ligase-like (TTLL) enzymes, which add glutamates to tubulin tails [7][8][9]. Glutamylation side-chains can be as short as a single glutamate, or as long as 20 glutamates [10]. Particular CCP enzymes may reduce the length of glutamate side-chains, and some can remove the branch point glutamate to eliminate glutamylation [5]. Similarly, some TTLL family members initiate glutamate side-chains, while others elongate [9]. TTLLs may also have preferences for α or β tubulin substrates [9]. Therefore, microtubule glutamylation is not uniform, and both side-chain length and tubulin substrate may exert different influences over motors, microtubule-associated proteins (MAPs), or other regulators of microtubule stability.
Glutamylation is essential for neuronal survival and axoneme stability, and a growing list of reports link hyper-or hypo-glutamylation to disease. Dysregulated microtubule glutamylation causes infantile-onset neurodegeneration and the ciliopathy Joubert syndrome in humans, and progressive cerebellar purkinje cell degeneration (pcd) in mice [3,5,[11][12][13][14]. Defects in axonal transport, which may be regulated by the Tubulin Code, contribute to neurodegenerative diseases including Alzheimer's disease and Parkinson's disease [4,15]. Although the importance of regulated glutamylation on ciliary and neuronal microtubules are clearly established, a comprehensive understanding of microtubule glutamylation is necessary to understand and interpret its context-specific effects in different cilia and neuron types.
The nematode C. elegans offers advantages to unraveling the Tubulin Code. Most basic biological processes, including axon development, neuronal transport, and ciliogenesis are conserved between C. elegans and humans [16]. In the worm, cilia are located on the distal most ends of sensory dendrites [17][18][19], are not essential for viability, and are diverse in terms of the composition and arrangement of microtubules, tubulin post-translational modifications, and motor proteins [2,18,20]. In male-specific ciliated sensory neurons, the CCPP-1 deglutamylase and TTLL-11 glutamylase control ciliary ultrastructure and particular ciliary kinesin motors [21,22]. In amphid and phasmid sensory neurons, loss of CCPP-1 leads to degeneration of ciliary microtubules indicating cell-specific differences in the tubulin code [21,22].
In C. elegans amphid and phasmid neurons, ciliary integrity is easily tested using a dye-filling assay [18]. ccpp-1 mutants are dye-filling defective (Dyf) in adults but not early larval stages, reflecting ultrastructural defects and progressive ciliary microtubule degeneration [21]. Using a candidate gene approach, we found that deletion mutations in ttll-4, ttll-5, or ttll-11 glutamylase genes suppressed the ccpp-1 Dyf phenotype. To identify new molecules and pathways that function in ciliary homeostasis, we performed a genetic screen for mutations that suppress the ccpp-1 Dyf phenotype. We identified mutations in ttll-5 and in the never-in-mitosis kinase gene nekl-4. In humans, TTLL5 mutation causes recessive retinal dystrophy [23,24]. C. elegans NEKL-4 is homologous to human NEK10, which is a kinase specific to ciliated cells and contributes to ciliogenesis in cultured human kidney cells [25]. In humans, NEK10 mutation causes bronchiectasis, a disorder of mucociliary transport in the airway due to defective motile cilia [26]. Here we show that NEKL-4 is expressed in all ciliated neurons but does not localize to cilia, suggesting that NEKL-4 indirectly influences regulation of ciliary stability. This work represents a step toward elucidating the molecular pathways by which glutamylation, as a component of the Tubulin Code, regulates cilia; this, in turn, allows us to better understand ciliopathies and neuronal survival in humans.
We also tested candidate genes that are not predicted to act directly in opposition to CCPP-1 deglutamylase activity for a role in the suppression of ciliary degeneration due to loss of CCPP-1 function. Because spas-1/Spastin and mei-1/Katanin are implicated in the severing of glutamylated microtubules [32,33], we tested whether mutations in these genes altered ccpp-1 hyperglutamylation-induced ciliary degeneration. Other candidates included the microtubuledepolymerizing kinesin-13 KLP-7 and the MAP kinase PMK-3, which interact with the deglutamylases CCPP-1 and CCPP-6 to regulate axon regrowth after laser-severing of C. elegans PLM touch receptor neurons [34]. Mutations in these and several other candidate genes failed to suppress the ccpp-1 Dyf phenotype (S1 Fig), suggesting that other unidentified molecules modulate ciliary degeneration due to loss of CCPP-1 loss of function.
A forward genetic screen for suppressors of the ccpp-1 dye-filling defect recovered alleles affecting the glutamylase TTLL-5 and the NIMA-related kinase NEKL-4 To identify genes that act with ccpp-1 to regulate ciliary stability, we conducted a genetic screen for suppressors of the ccpp-1(ok1821) Dyf phenotype. We mutagenized >1000 P 0 ccpp-1 mutants, passaged them to homozygose recessive mutations, performed dye-filling assays on F 2 adults, and screened for non-Dyf animals (Fig 3A). Mutant lines were maintained and retested for suppression of the Dyf phenotype. From >100,000 mutagenized haploid genomes, we isolated 11 viable lines that displayed suppression of the ccpp-1 Dyf phenotype when retested (S1 Table).
The my38 allele recovered in our screen suppressed the Dyf phenotype of ccpp-1 in the amphid neurons in the head (Fig 2A; Fig 3B). We identified my38 as a missense mutation in the ttll-5 glutamylase gene. my38 encodes an L338F substitution, located in the TTL domain (Fig 2B), that is predicted to affect ATP binding [A. Roll-Mecak, personal communication]. Therefore, our screen was able to identify a new allele of one of our candidate suppressor genes. This is important both as a validation of our screening method as well as a way to study the effects of different protein alterations on a phenotype.
To confirm that my31 was a mutation in the nekl-4 gene, we performed transgenic rescue experiments by germline injection of the wild-type genomic copy of nekl-4 (injected @1.5ng/ μL) into ccpp-1;nekl-4(my31) mutants, and restored the ccpp-1 Dyf phenotype (Fig 3C). We also characterized the deletion allele tm4910, which removes portions of exon 6 and exon 7 and replaces part of the kinase domain, including the putative ATP-binding lysine residue, with a glutamine residue. nekl-4(tm4910) single mutants are non-Dyf. nekl-4(my31) and nekl-4 (tm4910) exhibited similar suppression of ccpp-1 Dyf defects in amphid and phasmid neurons. Because both alleles were phenotypically indistinguishable and are likely loss-of-function alleles, we continued with the nekl-4(tm4910) deletion allele for ease of genotyping.

NEKL-4 is expressed in ciliated neurons but does not localize to cilia
We examined NEKL-4 expression pattern and subcellular localization using an extrachromosomal nekl-4p::nekl-4::gfp reporter transgene and endogenously tagged nekl-4::mneongreen and nekl-4::mscarlet CRISPR reporters (Fig 4A). Extrachromosomal NEKL-4::GFP, driven by an endogenous promoter of 1609 base pairs before the start codon, localized in the axons, cell bodies, and dendrites of the ciliated neurons in the head of the worm, as well as the phasmid and PQR neurons in the tail, and was excluded from nuclei (Fig 4B-4C). NEKL-4::GFP localized to the distal dendrite and the base of sensory cilia, but was not observed in the cilium. In the dendrites of some neurons, NEKL-4::GFP accumulated in patches ranging from 2-5 μm in length. Overexpression of a nekl-4p::nekl-4::gfp transgene (injected at a concentration of 13 ng/ μL) resulted in a Dyf phenotype (Fig 3A), suggesting that appropriate expression level or activity of NEKL-4 is important to ciliary stability.
With the CRISPR-generated nekl-4::mneongreen reporter, we were able to resolve fine structures in the cell bodies and dendrites. In the amphid cell bodies, NEKL-4::mNG appeared as thin filaments. A similar pattern was seen in the outer labial (OL), inner labial (IL), and cephalic (CEP) cell bodies, though the filaments appeared to be thicker and brighter. The CRISPR-generated NEKL-4::mNG reporter also localized in discrete, discontinuous patches in the dendrites. This pattern continued to the distal dendrite, terminating before entering the cilium of any neurons. We observed NEKL-4::mNG movement in the distal dendrite (Movie S1) and localization to the cell bodies and dendrites of the ray neurons in the male tail (S3 Fig). The intricate localization pattern and movement of NEKL-4::mNG suggests association with organelles or cytoskeletal components.
Because a nekl-4::gfp reporter was previously reported [41] to localize in phasmid cilia, we examined colocalization of NEKL-4::mScarlet with CHE-13/IFT57, an IFT-B component [42], and NPHP-1, a transition zone protein, to confirm NEKL-4 exclusion from the cilium. NEKL-4::mSC did not show significant overlap with NPHP-1::CFP or CHE-13::YFP, and was mainly restricted to the distal dendrite and periciliary membrane compartment (PCMC) (Fig 4F and  4G). We hypothesize that the discrepancy between our results and those of Yi et al [41] regarding NEKL-4 localization could be due to the higher injection concentration of their construct (50ng/μL vs. 13ng/μL for our extrachromosomal reporter), and the use of the dyf-1 promoter as opposed to the endogenous nekl-4 promoter that we utilized for both our extrachromosomal reporter and CRISPR markers. From our data, we conclude that NEKL-4 functions in ciliated neurons, although the protein itself does not localize to the cilium (Fig 4H).

ttll-4(tm3310) and ttll-5(tm3360) partially suppress ccpp-1 glutamylation defects. ttll-11 activity is essential for glutamylation but not ciliogenesis
To analyze the effect of our suppressor mutations on microtubule glutamylation, we conducted indirect immunofluorescence experiments. We used the monoclonal antibody GT335, which recognizes the branch point of glutamate side-chains [43]. In wild type, GT335 labels the cilia of amphid, labial (OL, IL), and cephalic (CEP) neurons (Fig 5A; S4 Fig). Labial and cephalic cilia are seen at the anterior-most tip of the animal, surrounding the buccal cavity. GT335 staining of labial and cephalic cilia appeared as singular axonemes [44]. In amphid channel cilia, GT335 specifically stains the doublet region middle segments that appear as two triangular bundles located posterior to the labial and cephalic cilia [44]. In wild type animals, GT335 strongly stained labial and cephalic cilia and amphid ciliary middle segments (Fig 5B).
We next examined the effect of the ttll gene suppressor mutations on GT335 staining in the ccpp-1 mutant background. We analyzed both the presence of GT335 staining and the appearance of staining to examine subtle differences in glutamylation patterns. We observed a variation in the GT335 staining pattern of some mutants: the amphids were stained but the pattern was abnormal. We classified this staining pattern as an "abnormal amphid" phenotype ( Fig  5B). We observed abnormal phenotypes in ciliary glutamylation in ttll mutants and suppressor strains. In some mutants, including ccpp-1, labial and cephalic cilia stained, while amphid cilia did not, consistent with cell-specific regulation of the Tubulin Code [2,22,34,45] (Fig 5B). In ccpp-1 mutant amphid channel cilia, GT335 staining was often undetectable, and occasionally detectable but abnormal (Fig 5B). We observed some restoration of GT335 staining in ampid channel cilia in 15% of the ccpp-1;ttll-4 double mutant animals (Fig 5C). In ccpp-1;ttll-5 double mutants, we observed a higher incidence of abnormal amphid middle segments compared to the ccpp-1 single mutant (Fig 5C). In the ccpp-1;ttll-11(tm4059) double mutants, GT335 staining was undetectable (Fig 5C). ccpp-1;ttll-11b(gk482) mutants, in contrast, had some restoration of amphid and labial and cephalic GT335 staining (Fig 5C).
We also used a polyE antibody, which detects side-chains of 3 or more glutamates [5]. PolyE staining of wild-type males strongly labelled amphid middle segments and labial cilia, suggesting that microtubules in these cilia are decorated with side chains of at least 3 glutamates. Mutations in ttll-4 and ttll-5 did not detectably reduce polyE staining. The ttll-11 (tm4059) mutation, however, abolished polyE staining (S5 Fig). Our results on the effect of the suppressor mutations on polyE staining in ccpp-1 mutants closely matched our observations with GT335. In ccpp-1 mutants, amphid channel cilia polyE staining was rarely observed (in 10% of animals) and more often appeared abnormal (S5 Fig). In ccpp-1;ttll-5(tm3360) and ccpp-1;ttll-11b(gk482) double mutant animals, we observed increases in the incidence of polyE stained amphid cilia (S5 Fig). We note that the absence of GT335 and polyE staining indicates an absence of glutamylation, but not necessarily an absence of MTs. Taking into account our antibody staining results, Dyf phenotype suppression, and previously published transmission electron microscopy data [21], we propose that our data indicates that loss of microtubule glutamylase activity suppresses microtubule degeneration due to loss of CCPP-1 function, and that microtubule glutamylation is not essential for dye filling or ciliogenesis. Overall, our data suggests that TTLL-11 is essential for adding the initiating "branch point" glutamate and that TTLL-4 and TTLL-5 are not essential for initiating glutamate side-chains. TTLL-4 and TTLL-5 might act primarily as elongases, but loss of their function was not sufficient to reduce the chain length to less than 3 glutamates.

Discussion
We used a forward genetic screen and candidate approach to identify genes that act with the ccpp-1 deglutamylase to control ciliary stability. We identified tubulin code writers and effectors of ciliary stability as modulators of degenerative phenotypes associated with CCPP-1 loss. Our data provides insight into how glutamylation is precisely controlled to maintain ciliary structure and function and to prevent neurodegeneration. Ciliary degeneration in ccpp-1 mutants can be suppressed by altering microtubule glutamylation through the loss of function of tubulin glutamylating enzymes TTLL-4, TTLL-5, or TTLL-11 (refer to model in Fig 7). Our findings are in agreement with studies in Chlamydomonas, where reducing microtubule glutamylation has a stabilizing effect on axonemal defective mutants ( [51,52]; for a summary of phenotypes, refer to S3 Table).
Loss of the NIMA related kinase NEKL-4 opposes the ccpp-1 effect on ciliary stability and may act independently of microtubule glutamylation state. Loss of NEKL-4 function suppresses the ccpp-1 Dyf phenotype while animals overexpressing NEKL-4::GFP are Dyf. Our findings are consistent with previous reports on the roles of NIMA kinases in regulating axonemal stability [52][53][54]. nekl-4(tm4910) does not strongly affect GT335 staining, arguing against a role for NEKL-4 in the regulation of a glutamylase or deglutamylase enzyme, although we cannot completely rule out this possibility. NEKL-4 is also unlikely to be a reader of ciliary glutamylation because it is largely absent from cilia. Future studies should prioritize ultrastructural analysis by performing transmission electron microscopy to characterize the effect of suppressor mutations on ciliary microtubule loss in ccpp-1 mutants. How does NEKL-4 regulate ciliary stability despite not localizing to cilia? Several processes including IFT loading/unloading, flagellar length sensing, and ciliary entry of proteins depend on the phosphorylation state of kinesin-II and other IFT components at the base of the cilium [55][56][57][58]. Potential substrates of the NEKL-4 human homolog NEK10 include kinesins, IFT components, and other ciliary proteins that have homologs in C. elegans ( [26,59]; S4 Table).
NEKs may also regulate endocytosis. C. elegans NEKL-2 and NEKL-3 control clathrinmediated endocytosis and trafficking in non-ciliated cells [69][70][71]. NEKL-4 localizes to the periciliary membrane compartment, an active site of endocytosis, a process necessary for regulating ciliary membrane volume, ciliary protein levels, and IFT [72]. Consistent with this idea, a reduction in endocytosis can rescue the Dyf defect of a dominant active G protein α subunit (gpa-3QL) mutant [73]. What could be the mechanism by which loss of CCPP-1 results in ciliary instability? Our data argue against a role for microtubule severing by katanin (MEI-1) and spastin (SPAS-1) in modulating the loss of ciliary stability in ccpp-1 mutants. Our results also did not support a role for other known effectors of microtubule glutamylation-the MAP kinase PMK-3 and the microtubule-depolymerizing kinesin-13 (KLP-7), in the degradation of ciliary microtubules in the absence of CCPP-1 function. Previous studies in Chlamydomonas revealed a link between tubulin turnover and both tubulin polyglutamylation [74] and NEK kinase activity [52]. Changes in IFT due to loss of CCPP-1 function could affect tubulin transport [75], or the binding of structural MAPs [76,77] could also mediate ccpp-1 loss induced loss of ciliary stability. Identification of readers of the tubulin code in amphid and phasmid neurons could identify molecules and mechanisms that modulate ciliary degeneration in response to changes in microtubule glutamylation.
NIMA-related kinases play major roles in microtubule structure regulation and processes such as mitosis, ciliogenesis, and post-translational modifications [54,[78][79][80]. Consequently, mutations of NEKs may result in cancer [81][82][83] and ciliopathies, such as polycystic kidney disease [26,80,84]. Many NEKs regulate ciliary disassembly through promotion of microtubule instability, such as Tetrahymena Nrk2, Nrk17, and Nrk30 [54], Chlamydomonas FA2, CNK2, and CNK11 [52,85,86], and human NEK7 [53]. Our data on the effects of NEKL-4 on ciliary stability offers insight into mucociliary disease associated with mutations in human NEK10. Our data also expands upon an evolutionarily-conserved link between molecules involved in microtubule glutamylation and NEK kinases [52,87]. Our studies provide a model for the exploration of this poorly-understood link between NEK kinases and microtubule glutamylation, and identify NEKs as a modulator of the degenerative effects associated with dysregulated levels of microtubule glutamylation. The data shown in this work will inform future studies on neurodegenerative disorders and ciliopathies associated with mutations in microtubule glutamylation enzymes and NEK kinases.

ccpp-1 EMS mutagenesis screen
To identify the pathway by which ccpp-1 mutations cause the Dyf phenotype [21], we performed a forward genetic screen for suppressors of the Dyf defect of ccpp-1 mutants. Ethyl methanesulfonate (EMS; Sigma Aldrich Cat. M0880) was used to mutagenize ccpp-1(ok1821) mutants {either Strain PT2168-ccpp-1(ok1821);myIs1[pkd-2::gfp + unc-122p::gfp] pkd-2 (sy606);him-5(e1490) or OE4160-ccpp-1(ok1821)}. L4 larvae were suspended in 1 ml of M9 buffer [89] with a final concentration of 47 mM EMS for four hours, before washing three times in M9 and depositing worm suspension onto an NGM plate seeded with OP50 E. coli to recover. We then transferred 10 or 20 of these mutagenized P 0 hermaphrodites to new NGM plates and allowed them to lay eggs on 100mm NGM plates seeded with OP50 for one day or two days before removing them. When the F 1 hermaphrodites reached adulthood, we synchronized the age of F 2 progeny by hypochlorite treatment of gravid adult F 1 animals [89] and replating embryos on NGM plates with OP50. We then incubated plates for four days at 20˚C or five days at 15˚C (to allow the age-synchronized F 2 progeny to reach adulthood) before washing them off plates using M9 to be dye-filled using the method described below. ccpp-1 mutants without suppressor mutations typically are completely Dye-filling defective (Dyf) by adulthood. Therefore, to isolate suppressor mutants, we picked F 2 adult animals that were non-Dyf or incompletely Dyf. We isolated up to five suppressor mutant isolates corresponding to a single P 0 plate to allow for some lethality, but in most cases kept only a single suppressor from each P 0 plate, except as noted in S1 Table. In total, we examined more than 100,000 haploid genomes. We isolated 93 non-Dyf or partially Dyf F 2 animals, of which 79 either died, were sterile, or were Dyf (not suppressed) in a subsequent secondary dye-filling assays to confirm the phenotype. 14 isolates were viable and produced non-Dyf or partial Dyf progeny in secondary screens, of which we consider 11 to be independent isolates. S1 Table is based on the qualitative analysis where amphid and phasmid neurons were not scored separately.

Whole-genome sequencing
To facilitate identification of suppressor mutations, whole-genome sequencing (WGS) was performed using the strategy described in Doitsidou et al [90]. Mutants were crossed with PT3037, a strain in which ccpp-1 had been introgressed into the Hawaiian wild isolate genomic background by backcrossing ten times into CB4856. From crosses of PT3068 (my31) X PT3037, at least 20 suppressed F 2 s were isolated; these F 2 s were pooled for WGS to identify genomic intervals that lacked PT3037-derived single-nucleotide polymorphisms (SNPs); novel SNPs within those intervals potentially represent the suppressor mutation.
Adapter-ligated libraries were prepared from genomic DNA using the manufacturer's protocol (Illumina, San Diego, CA) from PT3037 and the pooled suppressed F 2 s. A HiSeq 2000 instrument (Illumina, San Diego, CA) was used to sequence the genomic libraries. To identify the my31 mutation, we used a pipeline described in Wang et al [91]. Briefly, the pipeline was as follows: BFAST for alignment [92], SAMtools for variant calling [93], and ANNOVAR for annotation [94]. We compared sequence data with C. elegans genome version WS220 to identify mutations. To identify mapping intervals that might contain the my31 suppressor, we plotted Hawaiian SNP density against chromosome position. Non-Hawaiian SNPs that were also identified in PT3037 were removed as background. Candidate mutations for my31 were identified as novel homozygous non-synomymous coding variants within the mapping interval.
To facilitate identification of suppressor mutations, mutants were crossed with PT3037 (a strain in which ccpp-1 had been introgressed into the Hawaiian wild isolate genomic background by backcrossing ten times into CB4856). From crosses of PT3068 (my31) X PT3037, at least 20 suppressed F 2 s were isolated; these F 2 s were pooled for whole genome sequencing using the strategy described in Doitsidou et al [90] to identify genomic intervals containing N2 single nucleotide polymorphisms (SNPs) as potentially containing the suppressor mutation.
Adapter-ligated libraries were prepared from genomic DNA using the manufacturer's protocol (Illumina, San Diego, CA) from PT3037 and the pooled suppressed F 2 s. A HiSeq 2000 instrument (Illumina, San Diego, CA) was used to sequence the genomic libraries. To identify the my31 mutation, we used a pipeline described in Huang et al [95]. Briefly, the pipeline was as follows: BFAST for alignment [92], SAMtools for variant calling [93], and ANNOVAR for annotation [94]. We compared sequence data with C. elegans genome version WS220 to identify mutations. To identify N2 intervals that might contain the my31 suppressor, we plotted Hawaiian SNP density against chromosome position. Non-Hawaiian SNPs that were also identified in PT3037 were removed as background.

Protein domain identification, homology analysis and alignments
Sequences of ttll -4, ttll-5, ttll-11, and nekl-4 were obtained from WormBase [29] and protein sequences were obtained from WormBase or UniProt [96]. Structural predictions and protein domain information were generated with InterPro [97] and ScanProsite [98]. The PEST proteolytic degradation site in NEKL-4 was identified using EMBOSS epestfind [36]. NEKL-4 homology to NEK10 was predicted using protein BLAST [99] and human NEK10 and C. elegans NEKL-4 kinase domains were aligned using T-COFFEE Expresso [30]. All scans and queries described above were performed using default settings. Gene diagrams were created with the WormWeb Exon-Intron graphic maker [100], and protein diagrams were created with DOG 2.0 software [101,102]. The NEK10/NEKL-4 kinase domain alignment figure was generated with BoxShade [103] from the T-COFFEE sequence alignment.

Dye-filling assays
Three days prior to performing dye-filling assays, healthy, non-crowded plates containing many gravid hermaphrodite animals were washed and bleached to synchronize the worms. Day 1 and Day 2 adults were washed from the plate with 1mL M9 buffer and transferred to 1.5mL tubes. A stock solution of 2.5mg/mL DiI (Thermo Fisher Cat. D282) was added to a 1:1000 dilution and worms were incubated with gentle shaking for 30 min at room temperature. Prior to scoring, worms were briefly spun at 2000RPM and plated on new seeded NGM plates for approximately 1hr to excrete excess dye. The presence and intensity of dye in the amphid and phasmid neurons was visually scored using a dissecting microscope. Worms scored as non-Dyf were similar in brightness to wild type, and Dyf worms had no staining. Partial Dyf worms had dye present, but distinguishably less bright than wild type. Kruskal-Wallis one-way ANOVA analysis (where non-Dyf = 2, partial Dyf = 1. and Dyf = 0) and posthoc Dunn's multiple comparison test were performed in Prism (Graphpad Software).

Antibody staining
We performed immunofluorescence microscopy using the protocol described in O'Hagan et al [22]. We synchronized animals by bleaching and fixed as Day 1 adults. Fixation was accomplished by washing animals from 3 NGM plates using M9 buffer, then washing animals in a 15 ml conical tube 3 more times with M9 over one hour. Worms were chilled on ice before washing in ice-cold Ruvkun buffer (80 mM KCl, 20 mM NaCl, 10 mM EGTA, 5 mM spermidine-HCl, 15 mM Pipes, pH 7.4 and 25% methanol) plus 2% formaldehyde in 1.5ml centrifuge tubes. The tubes were immersed in liquid nitrogen and melted under tap water to crack the worms' cuticles. Worms were then washed twice with Tris-Triton buffer (100 mM Tris-HCl, pH 7.4, 1% Triton X-100 and 1 mM EDTA), suspended in Tris-Triton buffer+1% β-mercaptoethanol, and incubated overnight at 37˚C. The next day, worms were washed with 1X BO 3 buffer (50 mM H 3 BO 3 , 25 mM NaOH) + 0.01% Triton, and suspended in 1X BO 3 + 0.01% Triton buffer + 10 mM DTT for 15 minutes with gentle agitation at room temperature. Worms were then washed with 1X BO 3 buffer (50 mM H 3 BO 3 , 25 mM NaOH) + 0.01% Triton, and suspended in 1X BO 3 + 0.01% Triton buffer + 0.3% H 2 O 2 for 15 minutes with gentle agitation at room temperature. After washing once with 1X BO 3 + 0.01% Triton buffer, worms were washed for 15 minutes in Antibody buffer B (1X PBS, 0.1% BSA, 0.5% Triton X-100, 0.05% sodium azide, 1mM EDTA) with gentle agitation at room temperature. Fixed worms were stored in Antibody buffer A (1X PBS, 1% BSA, 0.5% Triton X-100, 0.05% sodium azide, 1mM EDTA) at 4˚C for up to one month before antibody staining. Animals were stained overnight at room temperature with a 1:450 dilution (in Antibody Buffer A) of GT335 (Adipogen Cat. AG-20B-0020-C100), a monoclonal antibody which binds the branch point of both monoglutamylated and polyglutamylated substrates [43], or polyE (Adipogen Cat. AG-25B-0030-C050), a polyclonal antibody which only binds polyglutamylated substrates (3 or more glutamate residues) [5]. Stained worms were washed with several changes of Antibody B Buffer with gentle agitation at room temperature over several hours. After rinsing with Antibody Buffer A, Alexa-fluor 568-conjugated donkey anti-mouse secondary antibody (Invitrogen Cat. A10037) was added at a final dilution of 1:2500 and incubated for 2 hours at room temperature with gentle agitation. Worms were then washed with several changes of Antibody Buffer B over several hours before mounting on 2% agarose pads for imaging.
We observed variation between and within individual GT335 staining experiments with respect to pattern and intensity. Environmental factors may cause changes in levels of ciliary glutamylation as measured by GT335 intensity [104]. It is therefore possible that small variations in experimental techniques or day-to-day lab conditions could have impacted our results. However, the GT335 staining phenotype of each mutant was generally consistent across trials.

Confocal imaging
Day 1 adult hermaphrodites were anaesthetized with 10 mM levamisole and mounted on 4% agarose pads for imaging at room temperature. Confocal imaging was performed with a Zeiss LSM 880 inverted microscope with an Airyscan superresolution module using a LSM T-PMT detector and ZenBlack software (Carl Zeiss Microscopy, Oberkochen, Germany). Laser intensity was adjusted to avoid saturated pixels. Images were acquired using a 63x/1.4 Oil Plan-Apochromat objective in Airyscan Fast mode and deconvolved using Airyscan processing. Image files were imported into Fiji/ImageJ [105] with the BioFormats Importer plugin for linear adjustment of contrast and creation of maximum intensity projections, and Adobe Photoshop CS5 was used to trace the outlines of the worm if needed. Images were placed in Adobe Illustrator CS5 for figure assembly.

KAP-1::GFP localization analysis
Nematodes were mounted as above. Widefield images were acquired on a Zeiss Axio Observer with Colibri 7 LEDs and ZenBlue software (Carl Zeiss Microscopy, Oberkochen, Germany) using a Photometrics Prime 95B sCMOS camera (Teledyne Photometrics, Tucson, AZ). A 100x/1.4 Oil Plan-Apochromat objective was used for imaging the phasmid cilia of Day 1 hermaphrodite animals. Images were imported into Fiji/ImageJ [105] and the length of KAP-1:: GFP labeling was measured for each cilium using maximum intensity projections. Length was measured starting at the transition zone, which is distinct and bright, and ending when fluorescence was no longer visible. Kruskal-Wallis one-way ANOVA analysis and posthoc Dunn's multiple comparison test were performed in Prism (Graphpad Software).

CRISPR constructs
For creation of the nekl-4::mneongreen and nekl-4::mscarlet-I endogenous CRISPR tags, we followed the Mello Lab protocol [107] and used pdsDNA [108] that included a short flexible linker sequence between the 3' end of nekl-4 and the start of mneongreen or mscarlet-I. The guide sequence was designed using CRISPOR [109,110] and silent mutation sites for PAM modification and diagnosis were located using WatCut silent mutation scanning [111]. Reagents used are as follows. dg357 was a gift from Dominique A. Glauser and was used in accordance with the C. elegans group license with AlleleBiotech. pSEM90 was a gift from Thomas Boulin.
Supporting information S1 Fig. Candidate gene mutations that did not suppress the ccpp-1 dye-filling defect. For experiments involving the mei-1(or1178) temperature-sensitive allele, eggs from plates kept at the permissive temperature (15˚C; [112]) were picked to fresh plates and shifted to 25˚C after hatching for 4 days before scoring. Localization of NEKL-4::mNeonGreen in adult males. Tail is flat against the coverslip in a. to visualize ray dendrites. Tail in b. is folded but the filamentous pattern in ray cell bodies is visible. c-d. Localization of NEKL-4::mNeonGreen in L4 molt males. Both images are different sections from the same z-stack. Arrows indicate ray cell bodies, arrowheads indicate ray dendrites (numbered). Scale = 10μm.    [26]. Original list contains genes with phosphopeptides depleted more than twofold upon NEK10 deletion in human bronchial epithelial cell (HBEC) culture. This list contains genes with known C. elegans homologs from the following relevant classes in the NEK10 list: axonemal dyneins and assembly factors, kinesins, intraflagellar transport, ciliary length control. (XLSX) S5