|Home | About | Journals | Submit | Contact Us | Français|
Tubulin glutamylation is a post-translational modification (PTM) occurring predominantly on ciliary axonemal tubulin and has been suggested to be important for ciliary function 1,2. However, its relationship to disorders of the primary cilium, termed ‘ciliopathies’, has not been explored. Here, in Joubert syndrome (JBTS) 3, we identify the JBTS15 locus and the responsible gene as CEP41, encoding a centrosomal protein of 41 KDa 4. We show that CEP41 is localized to the basal body/primary cilium, and regulates the ciliary entry of TTLL6, an evolutionarily conserved polyglutamylase enzyme 5. Depletion of CEP41 causes ciliopathy-related phenotypes in zebrafish and mouse, and induces cilia axonemal glutamylation defects. Our data identify loss of CEP41 as a cause of JBTS ciliopathy and highlight involvement of tubulin PTM in pathogenesis of the ciliopathy spectrum.
Joubert syndrome (OMIM 213300) is characterized by cerebellar hypoplasia, and neurological features including ataxia, psychomotor delay and oculomotor apraxia with a pathognomonic “molar tooth sign” on brain imaging. JBTS is frequently accompanied by various multiorgan signs and symptoms including retinal dystrophy, nephronophthisis, liver fibrosis and polydactyly, conditions associated with disorders of the ciliopathy spectrum of diseases that include Meckel-Gruber Syndrome (MKS), Bardet-Biedl Syndrome (BBS) and Nephronophthisis (NPHP). Though several causative genes have been found for these disorders, they account for less than 50% of cases 6,7. We recruited a consanguineous two-branch Egyptian family (MTI-429) with five affected members (Fig. 1a–b, Table 1). We excluded linkage to previously identified JBTS loci using a panel of highly informative markers. Analysis of the family using whole genome Illumina 5K SNP Linkage chip Ver. IV scan identified a 5 Mbp region of linkage on chromosome 7q31.33–32.3, with a peak multipoint LOD score of 3.71, thus defining the JBTS15 locus. Haplotype analysis suggested a candidate interval between rs766240 and rs4728251 delineating the peak of highest significance (Supplementary Fig. 1a–b).
In order to further narrow the interval, we re-analyzed MTI-429 with the denser Affymetrix 250K NspI SNP array by applying a linkage-free IBD model 8. The combination of the 5K SNP and 250K SNP linkage analyses generated a 2.8 Mbp IBD interval between rs17165226 and rs2971773 containing 26 genes (Fig. 1c). Direct sequence analyses of candidate genes within the interval led to the identification of a homozygous c.33+2T>G base change from reference sequence NM_018718, which was predicted to abolish the consensus splice donor site from exon 1 of the CEP41 gene (Fig. 1d, Supplementary Fig. 2a). To confirm a mutation-specific splicing defect, we evaluated CEP41 transcripts from MTI-429 primary patient fibroblasts (MTI-429-IV-1 and -IV-6). The RT-PCR result showed an absence of mature CEP41 mRNA products in both affected patient cells, likely attributed to nonsense mediated decay (Fig. 1e).
We next screened an additional 832 ciliopathy patients: 720 of JBTS and 112 of MKS (many of whom were excluded for mutations in known ciliopathy genes) by directly sequencing CEP41 and found two additional consanguineous families with homozygous mutations: c.97+3_5delGAG in an Egyptian JBTS family (MTI-1491) and c.423-2A>C in a Portuguese JBTS family (COR-98) (Fig. 1b–d, Supplementary Fig. 2a). These mutations were predicted to abolish the consensus splice donor site from exon 2 and the splice acceptor site from exon 7, respectively. Moreover, we confirmed that the mutation in MTI-1491 led to skipping of exon 2, thereby generating a premature stop in exon 3 (Supplementary Fig. 2b). Interestingly, in addition to the JBTS patients, the MTI-1491 family included one individual that was consistent with a phenotype of BBS and lacked the pathognomonic “molar tooth sign”, and this patient was heterozygous for the c.97+3_5delGAG mutation (Supplementary Fig. 2b–d), suggesting CEP41 may modify other ciliopathy conditions. From our cohort screen, we further identified heterozygous CEP41 mutations (c.83C>A, c.107T>C, c.265C>G, c.536G>A, c.1078C>T), which altered highly conserved amino acid residues among vertebrates or led to a premature stop codon from five different families (Fig. 1d, Supplementary Fig. 2e, Supplementary Table 1). Each of these CEP41-mutated patients was additionally sequenced at the known JBTS genes, and in four there was an additional heterozygous potentially deleterious variant. It was notable that all homozygous mutations in CEP41 were splice site mutations and identified only in JBTS patients while heterozygous variants were present in several ciliopathies including BBS and MKS. Our findings suggest that constitutive disruptions of CEP41 result in JBTS, but that it may also serve as a modifier in the broader class of ciliopathies.
The CEP41 gene has been poorly characterized except for its expression analysis in human organs including brain, testis and kidney 9. CEP41 encodes for Centrosomal Protein 41 KDa, predicted to contain two coiled-coils and a rhodanese-like domain (RHOD), which is structurally related to the catalytic subunit of the Cdc25 class of phosphatases 10. However, we found that it lacks phosphatase activity in the in vitro para-Nitrophenylphosphate (pNPP) phosphatase assay (not shown). The RHOD domain therefore may be an enzymatically inactive version like some described RHOD domains in other proteins 10 functioning in protein interactions.
We next examined CEP41 gene expression at the mRNA level and CEP41 protein subcellular localization. In zebrafish, cep41 was expressed in the various ciliary organs including Kupffer’s vesicle (KV), ear and heart as well as brain and kidney, regions predominantly affected in JBTS (Fig. 2a, Supplementary Fig. 3). In several ciliated cell lines such as mouse inner medullary collecting duct (IMCD3) cells and human retinal pigment epithelial (hTERT-RPE1) cells, endogenous CEP41 was predominantly noted at the centrioles and cilia (Fig. 2b). The cilia-associated expression/localization of CEP41 prompted us to assess a possible role in cilia-related function. Accordingly, we performed knockdown experiments using translation blocking morpholino anti-sense oligonucleotides (MOs) in zebrafish. In cep41 MOs-injected embryos (morphants), we observed peripheral heart edema and tail defects along with ciliopathy-related phenotypes including hydrocephalus, abnormal ear otolith formation and smaller eyes 11–14 (Supplementary Fig. 4a–b). We further found that injection of cep41 MOs induced a decrease in production of the protein with a dose-dependent phenotypic severity in the embryos (Supplementary Fig. 4b–c).
Cilia of KV, a structure that corresponds to the mammalian embryonic node, are essential to mediate lateral asymmetry 15. Accordingly, the defect of left/right (L/R) asymmetry in the heart, a well-established ciliopathy phenotype in mammals, is also shown in zebrafish 16–18. To examine whether cep41 depletion results in the phenotype, we injected cep41 MOs into Tg(myl7:egfp) zebrafish, a myocardium-specific transgenic reporter line, and found heart asymmetry defects such as inversion or failure to develop asymmetry of the ventricle and atrium (Fig. 2c). We also generated a Cep41 knockout mouse line using a genetrap strategy (Supplementary Fig. 5a–c) and characterized its phenotype at E10–13. The homozygous Cep41Gt/Gt embryos showed a range of phenotypes: malformed hindbrain, exencephaly, brain hemorrhage, dilated pericardial sac and lethality as well as unexpected normal development in some homozygous mutants (Fig. 2d, Supplementary Fig. 5d–e, Supplementary Table 2). Although exencephaly, dilated heart and embryonic lethality suggest possible ciliary roles of Cep41 in mouse 16,19,20, the phenotypic variability including normal development suggests the presence of extragenic phenotypic modifiers. We next investigated a genetic rescue using human CEP41 in zebrafish cep41 morphants, and found partial rescue of the cilia-associated morphant phenotype (Fig. 2e). The data suggest a potential evolutionarily conserved role of CEP41 in ciliary function.
In order to explore possible roles of CEP41, we examined primary cultured fibroblasts of patients (MTI-429-IV-1 and -IV-6). We first tested whether CEP41 mutant cells were devoid of CEP41 protein. Consistent with RT-PCR results (Fig. 1e), neither patient fibroblast line tested produced detectable CEP41, suggesting that these mutant cells are nearly null for CEP41 (Fig. 3a–b). In order to examine the effect of CEP41 loss on cilia assembly, we induced ciliogenesis using serum starvation-mediated cell cycle arrest in confluent cells and visualized cilia by co-immunostaining using anti-ARL13B (cilia marker) 21 and GT335 (centrioles/cilia marker) 22 antibodies. In control fibroblasts, cilia were evident in 70 % of the total stained cells by 48 hr and 72 hr and nearly all co-stained with both markers (Fig. 3c, Supplementary Fig. 6). However, in the CEP41 mutant fibroblasts, cilia were stained positively with ARL13B but not with GT335 (Fig. 3c). We quantified this effect and found that, whereas the percent of ARL13B-positive ciliated cells in the mutant fibroblasts was approximately equal to control fibroblasts, the percentage of GT335-positive ciliated cells was dramatically reduced in mutant fibroblasts (Supplementary Fig. 6). The GT335 antibody was originally generated to recognize the glutamylated forms (both mono- and polyglutamylation) of tubulin 23. Therefore, the data suggest a potential role of CEP41 in regulating tubulin glutamyl posttranslational modifications (PTMs).
Microtubules are the major structural scaffolds of the ciliary axoneme, composing the 9+0 or 9+2 arrangement, and undergo several PTMs including acetylation, detyrosination, glycylation and glutamylation (Supplementary Fig. 7). We therefore tested the effect of CEP41 deficiency on these tubulin PTMs and found no significant defects other than glutamylation in CEP41 mutant cells (Supplementary Fig. 8). In addition, the result of immunostaining using PolyE antibody (specific for polyglutamylated tubulins) in mutant fibroblasts suggested that CEP41 might function in regulating both tubulin mono- and polyglutamylation (Supplementary Fig. 9). Concurrently, we observed an effect of cep41 deficiency on tubulin glutamylation as well as mildly reduced glycylation of the placode cilia in zebrafish (Fig. 3d, Supplementary Fig. 10). Forced expression of exogenous CEP41 in the mutant fibroblasts remarkably increased the percentage of cells displaying glutamylated cilia (Fig. 3e), suggesting that glutamylation of the cilium is dependent on the expression of CEP41. Furthermore, we found that cell lines from patients with mutations in other JBTS genes including TMEM216 and INPP5E displayed no such glutamylation defect (not shown), suggesting the phenotype of tubulin glutamylation phenotype is not a non-specific consequence of ciliopathy mutation. Together these data suggest that CEP41 is required for ciliary glutamylation, but not necessary for initial cilia assembly.
Recent studies have shown that defective tubulin PTM is associated with altered ciliary axonemal structure 24–26. Accordingly, we investigated whether the CEP41-dysfunctional cilia exhibiting glutamylation defects gave rise to abnormal axonemal structure. Transmission electron microscopy (TEM) analysis demonstrated that the depletion of cep41 resulted in apparent structural defects in zebrafish renal cilia: specifically A tubules of the outer doublet microtubules of the axoneme were collapsed and/or duplicated (Fig. 3f, Supplementary Fig. 11). Previous studies have suggested that ciliary structural disruption affects ciliary motility 27,28, thus we examined the effect on the cilia of KV and kidney in zebrafish cep41 morphants and found disabled motility of both cilia (Supplementary Fig. 12, Supplementary Mov. 1–6). Our data suggest that CEP41 is involved in ciliary structural formation and motility by playing an essential role in tubulin glutamylation at the cilium.
We next investigated how CEP41 modulates microtubule PTM at the ciliary axoneme. We noted that only microtubules of ciliary axonemes failed glutamylation, whereas those of centrioles were properly modified in CEP41 mutant patient cells. (Fig. 3c). Furthermore, the lack of a Tubulin Tyrosine Ligase (TTL) domain in CEP41, requisite for enzyme activity 5, implied that it is unlikely to serve as a glutamylase. The main enzymes mediating tubulin glutamylation are members of the conserved TTL-like (TTLL) family 5. Among several identified TTLL factors, we found consistently strong localization of TTLL6 at the basal body (the organizing structure at the base of cilium, derived from a mother centriole) and the cilium (Supplementary Fig. 13). In addition, previous studies have suggested that TTLL6 may be involved in ciliary function in several organisms including zebrafish 5,24,29. We therefore examined the effects of ttll6 deficiency using a ttll6 translational blocking MO 30 in zebrafish and observed ciliopathy-related morphological phenotypes, similar to although less severe than what we observed following cep41 knockdown (Fig. 4a). Additionally, ttll6 morphants showed A-tubule axonemal defects, similar to those of cep41 morphants (Fig. 4a), besides diverse axonemal structural defects, which were consistent with a previous report 30 (not shown). These data prompted us to test a possible functional relationship between CEP41 and TTLL6 by pair-wise co-immunoprecipitation (coIP), and found that the two proteins were part of a complex (Fig. 4b). Given evidence suggesting a candidate regulator associated with the transport of a polyglutamylase between the basal body and the cilium 29, we investigated a possible role for CEP41 in TTLL6 localization. Following efficient knockdown of Cep41 in IMCD cells using siRNA (Supplementary Fig. 14), we found localization of TTLL6 restricted mainly to the basal bodies, suggesting a block of entry of TTLL6 into the cilium (Fig. 4c). These data suggest that CEP41 functions in tubulin glutamylation by mediating transport of TTLL6 between the basal body and cilium.
Our finding of CEP41 mutations in JBTS patients provides the first evidence directly linking defective tubulin glutamylation at the cilium to a cause of ciliopathies. Consistent with previous studies implicating tubulin PTMs in cilia function 23,24,31,32, our data suggest CEP41-mediated ciliary glutamylation is essential for axonemal formation. Moreover, we found that CEP41 is required for transport of TTLL6 to modulate tubulin glutamylation, although it remains to be determined how these molecules co-function. Most likely, these proteins enter through the ciliary diffusion barrier at the transition zone 33,34 and are then transported along ciliary microtubules by intraflagellar transport (IFT) motor proteins. Thus examining whether CEP41 and TTLL6 form a complex with other factors at these locations will follow as a future study. Because tubulin glycylation is partially affected in cep41 morphants and the phenotype of cep41 morphants is more severe than ttll6 morphants, it is likely that CEP41 modulates other TTLL protein family members 5,30. Evidence for this possibility has been described in recent studies suggesting involvement of both tubulin- glycylation and glutamylation, each regulated by different TTLL-family members, in maintaining ciliary structure and motility 30,35–37.
MTI-429 family and additional families were recruited worldwide based upon the presence of at least one individual with a neuroradiographically proven ‘molar tooth sign’ associated with any JBTS or related disorder (JSRD) phenotype. Whenever possible, patients underwent a full diagnostic protocol as previously reported 38 and a standardized clinical questionnaire was obtained to assess extent of multi-organ involvement. We used standard methods to isolate genomic DNA from peripheral blood of the affected and unaffected family members after obtaining informed consent from all participating. Human subject research was approved by the Ethics Boards of Leeds (East), CASA Sollievo della Sofferenza Hospital/CSS-Mendel Institute, Hôpital Necker-Enfants Malades, Human subjects divisions at the University of Washington, University of Michigan Institutional Review Board and Human Research Protection Program, University of California, San Diego.
A 5K whole genome linkage SNP-scan was performed with family MTI-429 using the Illumina Linkage IVb mapping panel 39, and analyzed with easyLinkage-Plus software 40, which runs Allegro version 1.2c in a PC Windows interface to calculate multipoint LOD scores. Parameters were set to autosomal recessive with full penetrance, and disease allele frequency of 0.001. Fine mapping on the pedigree was performed with the Affymetrix 250K Nsp1 SNP array, and data were searched for common shared homozygous intervals from all affected family members using a custom script implemented in Mathematica (B. Merriman, unpublished). This script identifies all homozygous intervals longer than 2 Mbp for which there are no more than 1 % heterozygous calls (to permit potential genotyping errors).
Mutational screening of CEP41 was performed by direct sequencing of the 11 coding exons and the adjacent intronic junctions in patients. PCR products obtained were treated with Exonuclease I (EXO) (Fermentas) and shrimp alkaline phosphatase (SAP) (Promega), and both strands were sequenced using a BigDye terminator cycle sequencing kit with an ABI3100 automated sequencer (Applied Biosystems). The primers and optimized PCR conditions used are depicted (Supplementary Table 3). Segregation of the identified mutations was investigated in all available family members. All identified mutations in CEP41 were not encountered in 96 ethnically matched controls (188 chromosomes) upon direct sequencing.
Genetic location is according to the March 2006 Human Genome Browser build hg18. The ciliary proteome was searched using web-based tools 41,42. Protein sequence conservation was determined using ClustalW multiple amino acid sequence alignment (see URL section).
Full-length human CEP41 was cloned into the TOPO blunt vector, and then shuttled into EGFP- and HA- containing vectors. Human and zebrafish cep41 open reading frame were amplified by RT-PCR and cloned into the pCS2+ vector in order to synthesize RNA for injection into zebrafish embryos. Mouse Cep41 open reading frame was amplified and cloned into the GST-, EGFP-, and FLAG-conjugated vectors for biochemical assays.
SIGTR ES cell line (AW0157), derived from 129P2/OlaHsd mice carrying a gene trap insertion 43 in the intron 1 of Cep41, was obtained from the European Conditional Mouse Mutagenesis Program (EUCOMM) and injected into C57BL/6 recipient blastocysts at the Mouse Biology Program, UC Davis. High-percentage chimeras (≥ 70 %) were bred to C57BL/6 for germline transmission. Gene-trapped mice were isolated by PCR for β-galactosidase-neomycin fusion gene and genotyped by Southern blot analysis hybridizing a ~10 kb (wild-type allele) and a ~8 kb (gene-trapped allele) product (Supplementary Fig. 5).
AB wild type zebrafish strains were used for in situ hybridization carried out by standard protocols using a DIG-labeled sense and anti-sense RNA probe for cep41. To knockdown zebrafish cep41, a translational blocking morpholino antisense oligonucleotide (MO, Gene Tools Inc) was used: 5′-CATCTTCCAGCAGCAGAGCTTCGGC-3′, diluted to appropriate concentrations in deionized sterile water, and injected into one-two cell stage embryos, obtained from natural spawning of zebrafish lines. To rescue the phenotypes in MO-injected embryos (morphants), RNA transcribed in vitro with the SP6 mMessage mMachine kit (Ambion) was co-injected. For characterization of ciliary defects in zebrafish embryos, the morphological phenotype of either cep41 or ttll6 morphants were observed until 5 days post-fertilization (dpf) and quantified under bright-field microscopy based upon previously established criteria 44. For western blot analysis at 1–2 dpf zebrafish control embryos and cep41 morphants, about 50 embryos with each genotype were deyolked and the embryo lysates were extracted with RIPA buffer. For immunostaining with GT335, Ac-Tub and PolyG Abs in whole-mount zebrafish embryos, 3 dpf control embryos and cep41 morphants were fixed in Dent’s fixative (80 % MeOH: 20 % DMSO) at 4 °C overnight and incubated with GT335 Ab (1:400), Ac-Tub (1:400) and PolyG (1:300) as primary and with anti-goat-mouse 594 (1:1000) as secondary in diluted blocking solution (10 % normal goat serum: 0.5 % Tween 20 in PBS).
IMCD3, hTERT-RPE1, human fibroblast, and human embryonic kidney (HEK293) cells were grown in appropriate DMEM or MEM media supplemented with 10–20 % fetal bovine serum (FBS) at 37 °C in 5 % CO2. Healthy human female and male control fibroblast cells were obtained from ATCC, and patient fibroblasts from skin biopsies were propagated in culture (≤ 5 passage number). Human fibroblast cells were transfected using the Basic Nucleofector Kit for Primary Mammalian Fibroblasts (Lonza). Other cells were transfected at 60–80 % confluency with plasmids or siRNAs using Lipofectamine 2000 (Invitrogen). The transfected cells were incubated for 24–72 hr with FBS or without FBS, dependent on the experimental purpose.
Images of immunofluorescent stained cells were obtained on a Deltavision RT Deconvolution microscope (Olympus IX70), under the same parameters for each experiment. Taken images were edited and analyzed using Adobe Photoshop CS. For electron microscopy, a standard protocol 45 was used except for one modification that tannic acid was included in the fixative to enhance the final contrast of the images. Formvar-coated slot grids (Electron microscopy Sciences) were used for sections (60–70 nm) to maximize visibility of the tissue and cross section performed to observe cilia axonemal structure.
Zebrafish embryos (12 hpf for cilia in the KV and 2.5 dpf for the renal cilia) were transferred with embryo media to glass bottom culture dishes (MatTek), and 2.5 dpf embryos were anesthetized in tricaine solution (~0.016 mg/ml). Images were acquired for 30 sec – 2 min using a Perkin Elmer UltraView Vox Spinning Disk Confocal with EMCCD Hamamatsu 14 bit 1K × 1K camera, and edited with Volocity imaging software (Perkin Elmer).
For immunofluorescence, cells were fixed in 100 % methanol at −20 °C for 10 min. Primary antibodies used for immunofluorescence are: rabbit anti-CEP41, raised in rabbit to a purified bacterially expressed protein of GST-fused CEP41, (PRF&L, PA), mouse anti-acetylated-tubulin (Sigma), mouse GT335 Ab (gift from C. Janke), rabbit PolyE (gift from M. Gorovsky) and rabbit anti-ARL13B (gift from T. Caspary). We used Alexa 488- or Alexa 594-conjugated secondary antibodies (Molecular Probes) and Hoechst 33342 nuclear dye. All antibodies were diluted in 4 % normal donkey serum in PBS, primary antibodies were incubated for overnight at 4 °C and secondary antibodies were applied for 1 hr at room temperature. For immunoblotting, cells were extracted with RIPA lysis buffer 3 days after transfection and boiled with SDS sample buffer. Same concentrations of lysates were loaded for paired experiments. Primary antibodies used for western blotting were: rabbit anti-GFP (Roche), mouse anti-Flag (Sigma), and mouse anti-alpha-Tubulin (Sigma). Bound antibodies were detected using horseradish peroxidase-conjugated secondary antibodies (Pierce). For co-immunoprecipitation (co-IP), whole cell extracts (WCE) were prepared from confluent HEK293 cells transiently transfected with 14 μg plasmid DNA in 10 cm tissue culture dishes. WCE supernatants were processed for IP experiments by using 2 μg primary antibodies and protein A/G mixed agarose beads (Pierce).
The χ2 staticstic was computed manually with P value assigned for 1 degree of freedom in the characterization of mouse embryonic phenotype (Supplementary Table 2). For other studies, Student’s two-tailed non-paired t-tests were carried out to determine the statistical significance of differences between samples. P < 0.05 was considered statistically significant for all tests.
We thank the Marshfield Clinic Research Foundation, Center for Inherited Disease Research (supported by the US National Institutes of Health and National Heart, Lung, and Blood Institute) for genotyping support. We thank the International JSRD Study Group, E. Bertini, and the French Society of Foetal Pathology for patient referrals, J. Meerloo at the UCSD Microscopy Core (P30NS047101), T. Meerloo, Y. Jones and M. Farquhar and the CMM Electron Microscopy Core Facility at UCSD, S. Wirth and B. Willis for mutant mouse generation, C. Janke at the Institute Curie Research Center for GT335 antibody, TTLLs plasmids and technical advice, I. Drummond and N. Pathak at the Massachusetts General Hospital for ttll6 MO, M. Gorovsky at the Univ. of Rochester for polyE and polyG antibodies, S. Audollent at the Hôpital Necker-Enfants Malades for technical help, B. Sotak, N. Akizu, A. Crawford, V. Cantagrel, and E-J. Choi for stimulating scientific discussion and comments. This work was supported by the US National Institutes of Health R01NS048453 and R01NS052455 (J.G.G.), R01DK068306 (F.H.), R01NS064077 (D.A.D.) the American Heart Association 09POST2250641 (J.E.L.), the Italian Ministry of Health (Ricerca Finalizzata Malattie Rare and Ricerca Corrente 2011), the Telethon Foundation Italy GGP08145, and the Pierfranco and Luisa Mariani Foundation (E.M.V.), Grants-in-Aid from MEXT and JSPS, #23117517 and #23570209 (K.I.), the BDF Newlife, the Medical Research Council (G0700073) and the Sir Jules Thorn Charitable Trust 09/JTA (C.A.J.), l’Agence National pour la Recherche ANR 07-MRAR-Fetalciliopathies (T.A.-B.), Simons Foundation Autism Research Initiative (J.G.G.) and the Howard Hughes Medical Institute (F.H. and J.G.G.).
AUTHOR CONTRIBUTIONSJ.E.L. M.S.Z. and J.G.G. designed the study and experiments with substantial contributions from B.M. and S.F.N. helped fine mapping, J.L.S., S.L.B., J.O., F.B., M.I., A.M.S., T.A.-B., C.V.L., I.A.G., A.C., F.H., C.A.J., D.A.D., and E.M.V. performed genetic screening, and J.E.L., J.L.S., J.S., J.O., F.B., M.I., T.A.-B., I.A.G., D.A.D., C.M.L., and J.H.L. performed mutation analysis. M.S.Z., S.E.M., H.R.R., I.R., I.P.C., E.B. and E.M.V. identified and recruited patients, K.I. and M.S. shared critical reagents, and J.S. helped genotyping of mutant mice. J.E.L. performed microscopy, biochemical assays, zebrafish and mouse experiments. J.E.L. and J.G.G. interpreted the data and wrote the manuscript.