The Ciliopathy Protein CC2D2A Associates with NINL and Functions in RAB8-MICAL3-Regulated Vesicle Trafficking

Abstract

Ciliopathies are a group of human disorders caused by dysfunction of primary cilia, ubiquitous microtubule-based organelles involved in transduction of extra-cellular signals to the cell. This function requires the concentration of receptors and channels in the ciliary membrane, which is achieved by complex trafficking mechanisms, in part controlled by the small GTPase RAB8, and by sorting at the transition zone located at the entrance of the ciliary compartment. Mutations in the transition zone gene CC2D2A cause the related Joubert and Meckel syndromes, two typical ciliopathies characterized by central nervous system malformations, and result in loss of ciliary localization of multiple proteins in various models. The precise mechanisms by which CC2D2A and other transition zone proteins control protein entrance into the cilium and how they are linked to vesicular trafficking of incoming cargo remain largely unknown. In this work, we identify the centrosomal protein NINL as a physical interaction partner of CC2D2A. NINL partially co-localizes with CC2D2A at the base of cilia and ninl knockdown in zebrafish leads to photoreceptor outer segment loss, mislocalization of opsins and vesicle accumulation, similar to cc2d2a-/- phenotypes. Moreover, partial ninl knockdown in cc2d2a-/- embryos enhances the retinal phenotype of the mutants, indicating a genetic interaction in vivo, for which an illustration is found in patients from a Joubert Syndrome cohort. Similar to zebrafish cc2d2a mutants, ninl morphants display altered Rab8a localization. Further exploration of the NINL-associated interactome identifies MICAL3, a protein known to interact with Rab8 and to play an important role in vesicle docking and fusion. Together, these data support a model where CC2D2A associates with NINL to provide a docking point for cilia-directed cargo vesicles, suggesting a mechanism by which transition zone proteins can control the protein content of the ciliary compartment.

Fig 1. CC2D2A associates with NINL.

(a) Yeast two-hybrid interaction assays were performed with different fragments of CC2D2A fused to the GAL4 DNA binding domain (BD) and full length NINL isoform A and B, fused to the GAL4 activation domain (AD). Activation of the reporter genes, which indicates a physical interaction, was dependent on coiled-coil (CC) domains 1 and 2 of CC2D2A and either NINL isoform A or B. (b) The top panel of the immunoblot (IB) shows that FLAG-tagged CC2D2A, but not the FLAG-tagged LRRK2 that was included as a negative control, was co-precipitated with HA-tagged NINL isoform B using a rat monoclonal antibody directed against the HA-epitope. Protein input is shown in the lower panel; anti-HA precipitates are shown in the middle panel. (c) In a reciprocal experiment, HA-tagged NINL^isoB was co-precipitated with FLAG-tagged CC2D2A, but not with FLAG-tagged LRRK2. Protein input is shown in the lower panel; anti-FLAG precipitates are shown in the middle panel.

Fig 2. CC2D2A and NINL co-localize at the ciliary base in hTERT-RPE1 cells and in zebrafish retina.

(a, a’ and inset) When expressed alone, eCFP-tagged CC2D2A (green signal) localizes to the ciliary base (basal body, accessory centriole). The cilium is marked by anti-polyglutamylated tubulin (red signal, a’ and inset). eCFP-tagged CC2D2A (green signal; b) also (partly) localizes to the ciliary transition zone, which was visualized using anti-RPGRIP1L as a marker (red signal; b). (c, c’ and inset) mRFP-tagged NINL isoform B was localized at the ciliary base (cilium in green, c’ and inset). (d and inset) mRFP-tagged NINL isoform B (red signal) localizes adjacent to the ciliary transition zone (anti-RPGRIP1L; green signal). (e-e” and inset) Co-expression of mRFP-tagged NINL isoform B (red signal) and eCFP-tagged CC2D2A showed co-localization of both proteins around the ciliary base (yellow signal). (f) In wild-type larval zebrafish retina (4 dpf), Cc2d2a marked by anti-Cc2d2a antibodies (red signal) is localized apically to the photoreceptor basal body (marked by anti-centrin antibodies, green signal). (g) Ninl, stained with anti-Ninl antibodies, (red signal) is localized at the zebrafish photoreceptor ciliary base, partially overlapping with and apical to the green centrin signal. (h) Cc2d2a localization is unaffected by ninl knockdown and (i) Ninl localization is normal in cc2d2a ^-/- larvae. (j) Schematic representation of the localization of Ninl and Cc2d2a in zebrafish photoreceptor cells. (f-i) are immunostainings on cryosections from 4 dpf larvae. Nuclei were stained with DAPI (blue signal) in all panels. Scale bars are 10 μm in a-e, and 4 μm in f-i.

Fig 3. Zebrafish ninl knockdown causes loss of axonemes and outer segments, opsin mislocalization and vesicle/vacuole accumulation.

(a-b) Paraffin sections stained with Hematoxylin/Eosin of control (a) and ninl knockdown larvae (b) demonstrating shortened outer segments and grossly preserved retinal lamination in the morphants. (c-d’) Bodipy-stained cryosections highlight the shortened (brackets c’-d’) and dysmorphic outer segments of ninl knockdown larvae (d and d’) compared to the long cone- or rod-shaped outer segments of controls (c and c’). (e-f) Axonemes and connecting cilia marked with anti-acetylated alpha-tubulin and anti-Ift88 antibodies are severely shortened and reduced in numbers in ninl knockdown larvae (arrowhead in f). (g-h’) Immunofluorescence with anti-opsin antibody 4D2 demonstrates mislocalization of opsins within the cell body in ninl knockdown larvae (arrow in h’) compared to controls (g) where opsins are restricted to the outer segment. (i) Quantification of the intracellular opsin accumulation in ninl morphant photoreceptors compared to control: each single datapoint in the scatter graph displays the averaged mean grey value from one larva. The mean value and the Standard Error of the Mean (SEM) are displayed as bars. The difference is statistically significant (*** = p<0.0001, Student’s t-test). (j-l’) Transmission electron microscopy of control (j) and ninl knockdown larvae (k-l’) demonstrates absent or shortened and dysmorphic outer segments (OS) and accumulation of large vacuoles (v, arrow in l’) and smaller vesicular structures (bracket in k” and white arrowheads in l’) in morphants. Black arrowheads point to the connecting cilium in k and k”. k’ and k” are the boxed areas in k and l’ is the boxed area in l. (m) Quantification of the % of photoreceptors displaying these phenotypes. Absolute numbers of photoreceptors are also indicated. Error bars indicate 95% Confidence Intervals. The differences between morphant (red bars) and controls (blue bars) are statistically significant (*** = p<0.0001, Fisher’s exact test). Larvae in all panels are 4 dpf old. Scale bars are 30 μm in a-b, 15 μm in c-d and g-h, 3 μm in c’-d’ and g’-h’, 4 μm in e-f, 0.5 μm in j-k and l and 150nm in k’-k” and l’. OS outer segment, CC connecting cilium, m mitochondria, n nucleus, v vacuole.

Fig 4. Genetic interaction between ninl and cc2d2a.

(a-d) Partial ninl knockdown enhances the cystic kidney phenotype of cc2d2a mutants. (a-c) Glomerulus and proximal pronephric tubules highlighted in the transgenic line Tg(wt1b-EGFP). (a) Injection of a low dose of ninl atgMO (0.75 ng/nl) causes no cysts in wild-type larvae. (b) cc2d2a-/- larvae display small dilatations of the proximal tubules (arrow) in ~40% of cases. (c) Injection of this low dose of ninl atgMO in the cc2d2a-/- background leads to large dilatations of the proximal tubules and glomerular space (arrow) in 89% of mutants. g glomerulus, p pancreas. (d) Quantification of the glomerular + proximal tubular area displayed as a scatter plot, demonstrating a significant increase in proximal pronephric area in cc2d2a-/- larvae injected with low-dose ninl atgMO. The bars represent the mean and standard error of the mean (SEM) for each treatment group and each datapoint is an individual fish. (e-g’) Immunohistochemistry with anti-opsin antibody (4D2, green) on retinal cryosections of 4dpf cc2d2a-/- uninjected larvae (f-f”) and cc2d2a-/- larvae injected with subphenotypic doses of ninl MO (g’g”’), that cause no mislocalization in wild-type fish (e-e’), demonstrates that partial ninl knockdown increases the mislocalization of opsins (e’-g’). (h) Quantification of the mean intracellular fluorescence displayed as a scatter plot shows significant increase in intracellular fluorescence in cc2d2a-/- larvae injected with low dose of ninl atgMO. The bars represent the mean and standard error of the mean (SEM) for each treatment group and each datapoint represents the mean intracellular fluorescence from 10 photoreceptors in one individual fish. Cell membrane and outer segments are stained with bodipy (red in e-g). Nuclei are counterstained with DAPI. Scale bars are 100 μm in (a-c) and 4 μm in (e-g’). (i) Pedigree of a consanguineous family with one affected boy (UW48-3) and 4 unaffected siblings. UW48-3 carried a homozygous missense CC2D2A mutation as well as a frameshift mutation in NINL leading to premature truncation. (j) Pedigree of a family where the affected individual (UW36-3) carries the same homozygous CC2D2A mutation as in (i) but no additional rare deleterious variants. (k) Pedigree of a family where the affected individual (UW07-3) carries compound heterozygous C5ORF42 frameshift mutations and a nonsense mutation in NINL. (l) Pedigree of a family where the affected individual (UW57-3) carries compound heterozygous TMEM67 mutations and a missense NINL mutation. The phenotype of the affected individuals is detailed in italic on each pedigree under the corresponding mutations. MTS Molar Tooth Sign, DD Developmental Delay, ESRF End-Stage Renal Failure.

Fig 5. Ninl is required for correct Rab8A localization.

(a-a’) Expression of a rhodopsin-promoter driven cherry-tagged Rab8a in wild-type photoreceptors is mostly concentrated in one or several puncta (arrows a-a’) whereas it is diffuse in the majority of ninl morphant photoreceptors (b-b’). (c) Proportion of Rab8a-cherry expressing photoreceptors with punctate expression versus diffuse expression (bars represent 95% confidence interval; ** P<0.001, Fisher’s exact test). (d-d’) Endogenous Rab8a localization as seen by immunohistochemistry using an anti-Rab8a antibody (green) displays similar puncta (arrowheads) in wild-type photoreceptors, while the number of puncta is decreased in ninl morphant photoreceptors (e-e’). (f) Quantification of the number of Rab8a puncta displayed in the form of a scatter plot indicating that significantly fewer endogenous Rab8 puncta per μm² are present in ninl morphants compared to uninjected controls (*P = 0.01, unpaired Student’s t-test; bars represent standard error of the mean). Scoring was performed blinded as to injection status for (c) and (f). Outer segments are counterstained with bodipy in (d-e). Nuclei are counterstained with DAPI. All images are cryosections of 4 dpf larvae. Scale bars are 4 μm in all panels.

Fig 6. NINL interactome screen identifies MICAL3.

(a) Strep-SILAC and TAP (tandem affinity purification) experiments show that NINL interacts specifically with MICAL3 (Yellow). The solid line between NINL and MICAL3 symbolizes a direct interaction, whereas the dashed lines indicate interactions determined by IP. (b) Co-immunoprecipitation of eGFP-MICAL3 with FLAG-NINL^isoB, but not with FLAG-STRAD. The immunoblot (IB) in the top panel shows that eGFP-tagged MICAL3 co-immunoprecipitated with FLAG -tagged NINL (lane 2), whereas FLAG-tagged STRAD used as a negative control (lane 3) did not. The anti-GFP immunoprecipitates are shown in the middle panel; protein input is shown in the bottom panel. Reciprocal IP experiments using anti-FLAG antibodies confirmed the co-immunoprecipitation of eGFP-tagged MICAL3 with FLAG-tagged NINL^isoB (lane 2) and not with STRAD (lane 3) shown in the top panel. The anti-FLAG immunoprecipitations are shown in the middle panel; protein input is shown in the bottom panel. A co-immunoprecipitation experiment using untagged eGFP as a negative control (right panel) showed that eGFP-tagged MICAL3 immunoprecipitates with FLAG-tagged NINL^isoB but not with untagged eGFP.

Fig 7. NINL and CC2D2A co-localize with MICAL3 and are required for correct MICAL3 localization.

(a) Schematic of a photoreceptor for orientation. (b-d’) Co-localization of endogenous MICAL3 (green signal; b) and polyglutamylated tubulin (red signal; c) in rat retina (P20) by co-immunostaining radial cryo-sections. The yellow signal in the merged image (d’) indicates co-localization at the base of the photoreceptor connecting cilium. (b’-d’) are high magnification images of the boxed areas in (b-d). (e-f”) Centrosomal co-localization of NINL^isoB, CC2D2A and MICAL3 in hTERT-RPE1 cells. mRFP-NINL^isoB (red signal, e) localizes to the basal body of the cilia marked with polyglutamylated tubulin (cyanid signal; e) and overlaps with GFP-tagged MICAL3 (green signal, e’) at the ciliary base when co-expressed (yellow signal, e”). Co-expression of eCFP-CC2D2A (green signal, f) and mRFP-MICAL3 (red signal, f’) resulted in partial overlap at the base of the cilia (yellow signal, f”). (g) Endogenous MICAL3 (green signal) detected by immunostaining clusters at the ciliary base (white arrows; cilium marked with anti-acetylated tubulin in red) of hTERT-RPE1 cells treated with non-targeting siRNA. (h, i) Knockdown of NINL (h) or CC2D2A (i) expression by siRNA results in dispersed distribution of MICAL3 throughout the cell body (brackets) with retention of some MICAL3 puncta at the ciliary base (arrows). qPCR analysis of NINL (j) and CC2D2A (k) siRNA treated hTERT-RPE1 cells. Cells were transfected with 10nM siRNA and all qPCR data were normalized against GUSB levels. Bar and error bars refer to mean and standard deviation, respectively (n = 3, on two biologicial replicates). *: P<0.05; **: P<0.01 versus non targeting siRNA (NT) (student’s t-test). Nuclei are counter stained with DAPI in all panels (blue signal). Scale bars: are 5 μm in d, 1 μm in d’ and 10 μm in e-i.

Fig 8. Proposed model for CC2D2A and NINL function in trafficking, docking and fusion of rhodopsin-carrier vesicles.

1) CC2D2A binds NINL and thus provides a docking point at the base of the connecting cilium for incoming vesicles. 2) NINL binds MICAL3 which in turn binds RAB8 that is coating the rhodopsin-carrier vesicles. Since NINL also associates with the cytoplasmic dynein1 motor complex, it provides a link between the carrier vesicles and the motor generating the movement along the microtubules. 3). MICAL3 subsequently interacts with ELKS and its redox activity promotes remodeling of the docking complex resulting in fusion of the vesicle at the periciliary region.