CC2D2A Is Mutated in Joubert Syndrome and Interacts with the Ciliopathy-Associated Basal Body Protein CEP290

Subjects

Subjects were recruited worldwide and collected under the approval of the Human Subjects Divisions at the University of Washington, the University of Michigan, and Johns Hopkins University. Characteristic brain-imaging findings, combined with developmental delay and ataxia, were the minimal criteria for JSRD, whereas subjects with BBS were ascertained as described by Beales et al.²⁴ Cerebellar vermis hypoplasia was a sufficient imaging criterion in subjects not evaluated by MRI. Known genes and loci for JSRD (AHI1, CEP290, NPHP1, TMEM67, chromosomal loci 9q34, and 11p11.2–q12.1) were excluded in the consanguineous families by either haplotype analysis for homozygosity via microsatellite markers or by direct sequencing. Nonconsanguineous families with mutations in NPHP1 and AHI1 were also excluded from the analysis.

SNP and Microsatellite Marker Genotyping

Single nucleotide polymorphisms (SNPs) were genotyped with the GeneChip Human Mapping 50K Array Xba 240 (Affymetrix Inc., Santa Clara, CA) under standard conditions. The arrays were scanned with a GeneChip Scanner 3000 and calls were assigned with GeneChip Operating Software (GCOS) and Genotyping Analysis Software. Overlapping homozygous intervals were identified with Excel (Microsoft Corporation, Redmond, WA) and shared haplotypes were identified with Stata Corp. (College Station, TX; code available on request). Microsatellite markers on chromosome 4p15 were genotyped with ABI primers and an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Austin, TX).

DNA Sequencing

Sequencing was performed according to the manufacturer's instructions with Big Dye terminator version 3.1 and an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems). Forward and reverse strands of all 38 CC2D2A exons were sequenced, as well as ∼30 bp of flanking intronic sequence. Primers and conditions used are depicted in Table S1 available online.

Sequence Analysis

Alignments were performed with Clustal W2 and Boxshade 3.21. Homology calculations were performed with Vector NTI10 (Invitrogen Corporation, Carlsbad, CA). Phylogenetic trees were constructed with BLAST and TreeFam.

RT-PCR for Human Transcripts

Fetal eye and liver mRNA was isolated with the RNeasy Mini Kit (QIAgen, Valencia, CA) and all other RNAs were purchased: adult retina from Clontech (Mountain View, CA), adult brain, kidney, liver, heart, lung, and skeletal muscle from Ambion (Austin, TX), and fetal brain and kidney from Virogen (Watertown, MA). RT-PCR analysis of CC2D2A, CC2D2B, RPGRIP1L, and HPRT (control) was performed on total RNA from adult brain, retina, heart, kidney, liver, and skeletal muscle, as well as 12–16 week human fetal brain, eye, kidney, and liver. The primers are depicted in Table S1. Amplification was carried out in PCR buffer containing 2.5 mM MgCl₂ (32 cycles of denaturation for 30 s at 94°C, annealing for 30 s at 55°C [50°C for RPGRIP1L], and extension for 1 min at 72°C, with an initial denaturation step for 4 min at 94°C). The number of cycles was titrated so that the amount of product was not saturated in any tissue.

Immunocytochemistry

hTERT-RPE1 cells (kindly provided by Uwe Wolfrum) were cultured as described previously¹⁹ and stained with anti-CEP290 (rabbit polyclonal, kindly provided by Erich Nigg) 1:100 or GT-335 (mouse monoclonal, kindly provided by Carsten Janke) 1:1000. hTERT-RPE1 cells were seeded on glass slides, grown overnight, and then transfected for 24 hr with eCFP-CC2D2A with Effectene (QIAgen), according to manufacturer's instructions. Primary cilium formation in hTERT-RPE1 cells was induced by serum starvation (0.2% serum) for another 24 hr. Prior to fixation, ciliated hTERT-RPE1 cells were washed briefly in PBS, fixed in ice-cold methanol for 10 min, and blocked with 2% BSA in PBS for 20 min. Slides were incubated with the primary antibodies for 1 hr at room temperature and subsequently washed with PBS. Subsequently, slides were incubated with secondary antibodies that were described previously. Washing with PBS was repeated. Cells were then embedded in Vectashield with DAPI (Vector Laboratories, Burlingame, CA) and analyzed with a Zeiss Axio Imager Z1 fluorescence microscope equipped with a 63× objective lens (Carl Zeiss, Sliedrecht, The Netherlands). Optical sections were generated through structured illumination by inserting an ApoTome slider into the illumination path, followed by processing with AxioVision (Carl Zeiss) and Photoshop CS2 software (Adobe Systems, San Jose, CA).

Yeast Two-Hybrid

Fragments of CC2D2A were tested for binding with proteins associated with ciliary disorders or cilia function (Table S2). The N-terminal region until the C2 domain (1–998 aa), the C2 domain with the remaining C-terminal part (992–1561 aa), and the C2 domain alone (992–1177 aa) were cloned in pAD as well as pBD Gateway-adapted vectors. Yeast two-hybrid screening with ciliary proteins was performed in both directions. The BD Yeastmaker kit (Clontech) was used for transformation of yeast strains PJ69-4alpha and PJ69-4A with pBD and pAD vectors, respectively, according to the manufacturer's protocol. Yeast strains were mated and cotransformed yeast colonies were selected on plates lacking leucine (-L) and tryptophan (-W). Activation of the reporter genes was detected on selection plates lacking histidine (-LWH) and adenine (-LWHA). In addition, X-β-gal colorimetric filter lift assays (LacZ reporter gene) were performed to confirm the interaction as described.³

GST Pull-Down

We cloned the part of CEP290 (coiled-coil domain 4-6 encoded by amino acid residues 703–1030) that interacts with CC2D2A by yeast two-hybrid experiments with pDEST15, and we transformed this plasmid in BL21-DE3 cells. IPTG (Sigma-Aldrich, St. Louis, MO) was used to induce the expression of N-terminally GST-tagged CEP290 or GST alone. Purification of GST-CEP290^cc4-6 (∼78 kDa) or GST alone (∼26 kDa) was performed as described.³ Bacterial lysates were added to 50 μl of glutathione Sepharose 4B coupled beads (Amersham Biosciences, Piscataway, NJ). Beads were washed twice with 1× STE (10 mM Tris-HCl [pH 8.0]; 1 mM EDTA; 150 mM NaCl) and incubated for 2 hr. Beads were again washed twice and incubated for 2 hr with lysates of COS-1 cells expressing a Flag-tagged fragment of CC2D2A (amino acids 1–998 encoding the N-terminal part until the C2 domain), followed by five washes with Triton-X buffer (50 mM Tris-HCl [pH 7.5]; 150 mM NaCl; 0.5% Triton X-100). Finally, we added Laemmli Sample Buffer (Bio-Rad Laboratories, Hercules, CA) to the beads and performed western blotting with mouse anti-Flag (1:1000, clone M2; Sigma-Aldrich), secondary IRDye800 goat anti-mouse antibody (1:20,000; Rockland) and the ODYSSEY system (Li-Cor, Lincoln, NE), as described.³

RT-PCR Methods for Zebrafish Transcripts

All zebrafish work was performed under an approved protocol at the Fred Hutchinson Cancer Research Center. At 48 hpf, 40 snl/snl zebrafish were identified by the sinusoidal tail phenotype. Likewise, 40 snl/+ and +/+ fish from the same cross were pooled. Whole zebrafish embryos were homogenized by mortar and pestle and passed through QIAshredder columns (QIAgen). Total RNA was isolated from the homogenate with the RNeasy Mini Kit (QIAgen). cDNA was reverse transcribed from total RNA with random primers and reagents supplied in the RETROscript Kit (Ambion). Primers for sentinel and the odc1 (ornithine decarboxylase 1) housekeeping control gene are depicted in Table S1. Amplification was carried out for 34 cycles.

Genotyping for ZF

dCAP primers at the site of the sentinel mutation were designed to create an EcoRV restriction enzyme site in the PCR product from wild-type fish. Zebrafish genomic DNA was isolated with standard techniques. We used an annealing temperature of 64°C for 35 amplification cycles. Digested fragments were analyzed on 2% agarose gels.

Antibody Staining

Zebrafish embryos were fixed in Dent's fix (80% methanol/20% DMSO) at 4°C overnight. After gradual rehydration, they were washed several times in PBST (1× PBS with 0.5% Tween20) and blocked in PBS-DBT (1% DMSO/1% BSA/0.5% Tween20) with 10% normal sheep serum (NSS) (Sigma-Aldrich) at room temperature for 2 hr. Primary antibody incubation in PBS-DBT 10% NSS (1:2000 monoclonal anti-acetylated tubulin 6-11B-1 [Sigma-Aldrich]) was at 4°C overnight. Embryos were washed in PBST and blocked in PBS-DBT 10% NSS at RT for 1 hr and then incubated in 1:1000 goat anti-mouse Alexa Fluor 488 (Invitrogen) in PBS-DBT 10% NSS at 4°C overnight. After rinsing in PBS, the embryos were washed with methanol and equilibrated in clearing solution (1/3 benzoyl-alcohol and 2/3 benzoyl-benzoate) and examined with a Zeiss LSM510 confocal microscope.

Morpholino Injections

Antisense morpholino oligonucleotides blocking the function of CEP290⁵ were obtained from GeneTools (Corvalis, OR). Morpholinos were diluted to a working concentration of 5 ng/nl in 0.2 M KCl, and 4 nl were pressure-injected into one- or two-cell stage zygotes.

Homozygosity Mapping

To identify candidate loci for JSRD, we performed homozygosity mapping in consanguineous families with JSRD as described by Lander and Botstein.²⁵ The JSRD diagnosis was established by the presence of the MTS on brain MRI (Figures 1A–1E) and/or vermis hypoplasia on CT scan with supportive clinical findings: developmental delay/intellectual disability, ataxia, and/or abnormal eye movements. We genotyped affected and unaffected siblings in 28 consanguineous families with JSRD from multiple ethnic groups, with a cutoff of 80 consecutive homozygous SNPs to identify candidate regions of homozygosity in affected offspring of parents separated by six to nine meioses. Regions of homozygosity shared by affected and unaffected siblings were excluded. To further narrow these candidate regions, we also identified shared haplotypes of homozygous SNPs within and across families (Figures 1F–1H).

In Levanten Arab family UW50, we identified a shared haplotype of 252 consecutive homozygous SNPs spanning 7.6 Mb on chromosome 4p15 in two cousins, consistent with inheritance from a common ancestor (Figure 1F). These cousins shared no other regions of homozygosity greater than 50 consecutive homozygous SNPs, and the shared haplotype was heterozygous in available family members predicted to be obligate carriers. In Saudi Arabian families UW36 and UW48, the affected individuals shared a distinct haplotype of 64 consecutive homozygous SNPs at 4p15, providing additional evidence for a JSRD gene at this locus and narrowing the interval to 2.3 Mb (Figure 1G). The affected individuals did not share any other homozygous haplotypes greater than 38 consecutive SNPs. A fourth family (UW41) also had a large homozygous interval at chromosome 4p15 (Figure 1H). Segregation of the shared haplotype in each family was consistent with autosomal-recessive inheritance.

Mutation Detection

The 2.3 Mb shared haplotype in families UW36 and UW48 encompasses 14 genes (Figure 2A). We sequenced the coding regions and exon-intron boundaries of 13 of these genes in all four families, detecting two homozygous missense mutations (c.3364C → T p.P1122S; c.4582C → T p.R1528C) and one homozygous nonsense mutation (c.2848C → T p.R950X) in CC2D2A (Table 1 and Figure 3). As expected from the shared haplotype, we found that the two affected individuals in families UW36 and UW48 carry the same mutation (p.P1122S). P1122 is conserved across all species with related genes encoding this C2 domain, including all animals and protists, but not plants or fungi. R1528 is conserved in all vertebrates except pufferfish. Both missense mutations are predicted to be deleterious by SIFT and PolyPhen.^26,27 Sequence changes in the other 12 genes within the interval were either known SNPs or noncoding (not shown).

We then sequenced all CC2D2A exons in families that could not be excluded by segregation analysis by using four microsatellite markers at 4p15 (Figure 4). We identified compound heterozygous mutations in two families and a single frameshift mutation in a third nonconsanguineous family (Table 1 and Figure 3), yielding seven different CC2D2A mutations in 6 out of 70 JSRD families ascertained by the MTS ± other findings. A second cohort of 40 consanguineous JSRD families with renal disease was similarly evaluated, yielding one homozygous mutation (p.L1551P) in family F871, for a total prevalence of 7/110 (6%). The L1551P change is also predicted to be possibly damaging by Polyphen and tolerated by SIFT; however, this leucine is evolutionarily conserved as far as opossum, and proline is not present in any species. All mutations segregated as expected for autosomal-recessive inheritance, and the missense mutations were absent from >210 control chromosomes.

CC2D2A Mutations in Other Ciliopathies

We have shown recently that MKS genes contribute both primary loci and second site modifying alleles to the pathogenicity of BBS.²⁰ We therefore hypothesized that genes contributing to JSRD may also be involved in the manifestation of the BBS phenotype. To explore this possibility, we sequenced CC2D2A in 96 BBS patients preselected for having mutations in zero or one known causative ciliopathy locus. We detected one novel sequence variant, encoding a heterozygous K478E amino acid change in a European BBS family. Importantly, this variant is highly conserved as far as stickleback and predicted to be nontolerated by SIFT. Consistent with a potential modifying, but not causal, role in ciliary disease, we detected this variant in one of 192 ethnically matched control chromosomes; further investigations will be required to ascertain the potential effect of this allele on CC2D2A function. Overall, the paucity of novel variants identified in BBS patients suggests that CC2D2A is not a major contributor to BBS; however, screening of a larger or more ethnically diverse cohort will be necessary to test this possibility exhaustively.

In parallel, we also evaluated the potential role of CC2D2A in isolated nephronophthisis. By using homozygosity mapping, we excluded the 4p15 locus in 86 of 90 consanguineous families via a recessive model. In the four families that could not be excluded, we did not identify any CC2D2A mutations. These results indicate that CC2D2A mutations are not a major cause of isolated nephronophthisis.

Gene/Protein Structure

The 38 exon CC2D2A gene encodes a predicted 1620 amino acid product with coiled-coil and C2 domains (Figures 2B and 2C) described by Noor et al. and Tallila et al.^22,23 Tallila et al.²³ also reported an additional predicted exon between exons 29 and 30 that is not in the reference sequence (NM_001080522), whose presence we confirmed in multiple tissues via RT-PCR (not shown). The nonsense and frameshift mutations in JSRD patients precede the C2 domain and are likely to trigger nonsense-mediated decay, whereas the missense mutations occur in the C2 domain and C-terminal region, supporting a functional role for these parts of the protein (Figure 2C).

A human paralog (CC2D2B) is present on chromosome 10q23.33. Ensembl predicts a 322 amino acid CC2D2B product that corresponds to NP_001001732 and shares homology with the C-terminal region of CC2D2A; however, ESTs mapping to chromosome 10q23.33, comparison of the human genomic sequence to other mammals, and gene-prediction programs (reviewed in Brent²⁸) reveal a 1341 amino acid predicted protein with 33% identity/46% similarity to CC2D2A (Figure S1). The C2 domains of CC2D2A and CC2D2B are 42% identical and 55% similar, whereas the C-terminal regions are 45% identical and 60% similar. CC2D2A orthologs are found in all vertebrates and in sea urchin, jellyfish, and insects, whereas CC2D2B orthologs are found only in mammals, implicating CC2D2B in processes restricted to this class (Figure 2D). Interestingly, the C-terminal region of both proteins is distantly related to the CEP76 protein, a component of the centrosome (Figure S2).²⁹

Human Phenotype

We observed a spectrum of phenotypes in our cohort (Table 1). Subject UW49-II:1 has chorioretinal coloboma, renal, and liver disease (COACH phenotype). Liver biopsy displayed “chronic inflammation and fibrosis of the portal triads, intact sinuses, and central veins” and required living-related donor liver transplantation at 10 years of age. Her creatinine has been mildly elevated and her “renal echogenicity was in the upper limit of normal” on renal ultrasound, but she has not had a renal biopsy. Subject UW48-IV:7 may also have the COACH phenotype, but details of an abnormal renal ultrasound and mild hepatosplenomegaly are not available. Subjects UW48-IV:7 and UW41-IV:1 also have evidence of retinal dystrophy. Subjects UW50-VI:1 and UW50-VI:3 have encephaloceles similar to patients with MKS, whereas subjects UW50-VI:1 and UW49-II:1 have agenesis of the corpus callosum seen in a subset of individuals with JSRD. Affected individuals in UW36 and UW46 did not exhibit retinal, renal, or liver disease; however, UW36 was only 1 year old at last follow-up, and we found only a single CC2D2A mutation (p.V1097FfsX1) in UW46. No affected individuals had polydactyly. Although olfactory dysfunction has been observed in BBS,³⁰ it was not noted clinically in subjects with CC2D2A mutations, none of whom were formally tested for anosmia. After identifying CC2D2A mutations in subjects with JSRD, we reviewed the brain MRI images from two of the affected individuals with homozygous CC2D2A mutations described by Noor et al.,²² and we identified the characteristic imaging features of Joubert syndrome in both subjects (one shown in Figures 4E and 4E′). No overt findings of liver or kidney disease were identified in this family.

Expression

CC2D2A transcript could not be detected via human multiple-tissue northern blots (not shown); however, we were able to detect transcript in all adult and fetal tissues tested via RT-PCR (Figure 2E). Interestingly, expression was substantially higher in fetal brain than in adult brain, which may indicate an increased requirement in this tissue during development. High levels of expression were also detected in retina and kidney, commonly affected in JSRD.

Colocalization of CC2D2A with CEP290 at the Basal Body

To investigate whether CC2D2A and CEP290 function in the same location, we transfected hTert-RPE1 cells with eCFP-tagged full-length CC2D2A. As described previously in COS-7 cells,²² eCFP-CC2D2A is seen in the cytoplasm; however, after inducing cilia formation by serum starvation, eCFP-CC2D2A also localizes to the base of the cilia, as indicated by costaining with the anti-polyglutaminated tubulin antibody GT-335 (Figures 5A–5C). Costaining with CEP290 antibody³ demonstrates that recombinant CC2D2A at the basal body colocalizes with endogenous CEP290 (Figures 5D–5F).

Physical Interaction between CC2D2A and CEP290

Based on the hypothesis that CC2D2A might function in the primary cilium/basal body protein network,³¹ we assessed 23 ciliopathy-associated proteins (including CEP290, RPGRIP1L, RPGRIP1, RPGR [MIM 312610], and BBS1-12 [MIM 209900]) for interaction with CC2D2A via a yeast two-hybrid mating assay (Table S2). These proteins were chosen on the basis of their involvement in a ciliopathy-associated protein network.³¹ Of all combinations tested, only CEP290 was found to interact with CC2D2A (Figure 5G) and the interaction was confirmed by a GST pull-down assay (Figures 5H–5J). The interaction is specific for the N-terminal 998 amino acids of CC2D2A and an internal fragment of CEP290 (amino acids 703–1130) containing coiled-coil domains 4-6.

Modeling Loss of cc2d2a Function in Zebrafish

In a screen for mutations modifying aminoglycoside-induced mechanosensory hair cell damage, Owens et al. identified a nonsense mutation (sentinel [snl]) that lies N-terminal to the C2 domain in the sole zebrafish ortholog of CC2D2A (cc2d2a).³² The Cc2d2a protein is more closely related to CC2D2A (60% identical, 74% similar) than CC2D2B (34% identical, 48% similar; Figure S1), with higher homology in the C2 domain and C-terminal region. We found that cc2d2a transcript is reduced substantially in snl/snl fish versus their wild-type and heterozygous siblings (Figure 6A), presumably because of nonsense-mediated decay. To further explore CC2D2A function and develop a zebrafish model for JSRD, we evaluated snl/snl fish for phenotypes reminiscent of JSRD and ciliary dysfunction including pronephric cysts, laterality defects, brain malformation, as well as defects in cilium number and morphology.³³ Although other obvious defects were not observed, 33% of the snl/snl fish (confirmed by genotyping) exhibited pronephric cysts by 6 days postfertilization (dpf) compared to 0% of their wild-type and heterozygous siblings (Figures 6B and 6E). We used anti-acetylated tubulin antibody³⁴ to visualize motile cilia in the lumen of pronephric tubule and caudal region of the central canal. We were unable to discern any overt differences with respect to cilium number or morphology between snl/snl fish and their wild-type and heterozygous siblings at 48 hpf, although subtle differences could not be excluded.

Functional Interaction with CEP290

To assess the functional significance of the CC2D2A-CEP290 physical interaction, we evaluated the effects of decreased cep290 function in the snl/snl background. Sayer et al. found that knockdown of cep290 function in zebrafish via cep290 morpholinos resulted in pronephric cyst formation,⁵ so we titrated splice blocking (cep290splMO) and translation blocking (cep290atgMO) morpholinos to give a low prevalence of pronephric cysts in wild-type fish. We then injected offspring of a snl/+♂ × snl/+♀ cross and observed a striking exacerbation of the cyst phenotype in injected snl/snl fish compared to their injected siblings (snl/+ and +/+) and uninjected snl/snl fish (Figures 6B–6E). The cysts were larger, more prevalent, and evident 2 days earlier in the injected snl/snl fish, suggesting the cooperative action of CC2D2A and CEP290 in the pronephros. Cilia in injected snl/snl fish also appeared similar in morphology and number compared to wild-type.