The human Usher syndrome (USH) is a complex ciliopathy with at least 12 chromosomal loci assigned to three clinical subtypes, USH1-3. The heterogeneous USH proteins are organized into protein networks. Here, we identified Magi2 (membrane-associated guanylate kinase inverted-2) as a new component of the USH protein interactome, binding to the multifunctional scaffold protein SANS (USH1G). We showed that the SANS-Magi2 complex assembly is regulated by the phosphorylation of an internal PDZ-binding motif in the sterile alpha motif domain of SANS by the protein kinase CK2. We affirmed Magi2's role in receptor-mediated, clathrin-dependent endocytosis and showed that phosphorylated SANS tightly regulates Magi2-mediated endocytosis. Specific depletions by RNAi revealed that SANS and Magi2-mediated endocytosis regulates aspects of ciliogenesis. Furthermore, we demonstrated the localization of the SANS-Magi2 complex in the periciliary membrane complex facing the ciliary pocket of retinal photoreceptor cells in situ. Our data suggest that endocytotic processes may not only contribute to photoreceptor cell homeostasis but also counterbalance the periciliary membrane delivery accompanying the exocytosis processes for the cargo vesicle delivery. In USH1G patients, mutations in SANS eliminate Magi2 binding and thereby deregulate endocytosis, lead to defective ciliary transport modules and ultimately disrupt photoreceptor cell function inducing retinal degeneration.
The human Usher syndrome (USH) is an autosomal recessive disorder and the most frequent cause of combined deaf-blindness (1–4). Depending on the clinical characteristics, age of onset, severity and progression of symptoms, USH is divided into three subtypes: USH 1–3 (5). USH is genetically heterogeneous with at least 12 chromosomal loci described so far (3,6,7). The proteins encoded by the 10 known USH causing genes belong to very different protein families. Nevertheless, previous molecular analyses elucidated the integration of all USH1 and USH2 proteins in functional protein networks organized by the scaffold proteins harmonin, whirlin and SANS (2,3).
Recent data established a protein network organized by the USH proteins whirlin (USH2D) and SANS (USH1G, scaffold protein containing ankyrin repeats and SAM domain) in the periciliary region of vertebrate photoreceptor cells (8–12). This periciliary membrane complex (PMC) is localized at the collar-like extension of the apical inner segment and the connecting cilium, which bridges the biosynthetic active inner segment with the light sensitive outer segment, a modified primary cilium (13). It has been suggested that the PMC plays an important role connecting the modules of the molecular transport in these two photoreceptor compartments, namely the microtubule-based transport through the inner segment with the intraflagellar transport across the connecting cilium into the outer segment (8,13). Furthermore, this membrane compartment is homologous to the membrane of the ciliary pocket at the base of prototypic primary cilia, a microdomain, which seems to be a major site for endocytosis and exocytosis, regulating ciliary morphogenesis and homeostasis (14,15).
Previous data, based on protein interaction studies in heterologous systems and immunolocalization in the retina, suggest the hypothesis that SANS may participate in microtubule-based transport from the Golgi apparatus through the inner segment to the periciliary compartment and facilitates the reloading of ciliary cargo on the transporter module in the PMC of photoreceptor cells for the ciliary delivery (8,11,16). The molecular domain structure of SANS is very efficient for its scaffolding function (17). The N-terminus contains three ankyrin repeats followed by the central domain (CENT), a sterile alpha motif (SAM) and a C-terminal PDZ-binding motif (PBM) class-I (Fig. 1A). All three parts of the SANS molecule, the ankyrin repeats, the CENT and the SAM-PBM are known to mediate protein–protein interactions and their roles have been previously studied (8,17–19). In the present study, we aimed to enlighten the precise cellular function of SANS by identifying novel functional module partners of SANS. For this, we adopted yeast-2-hybrid (Y2H) screens of a retinal cDNA library. We identified Magi2 (membrane-associated guanylate kinase inverted-2) as a novel potential binding partner of the SANS' C-terminus. Magi2, also known as S-SCAM (synaptic scaffolding molecule), belongs to the MAGUK protein family of PDZ domain-containing scaffold proteins. Previous studies have indicated that Magi2 plays an essential role in synaptic development and maintenance (20,21). More specifically, Magi2 serves as a scaffold for a variety of proteins assembling synaptic protein complexes (22–24) and there is growing evidence that Magi2 participates in endocytosis as well (20,23,25).
Here we demonstrated that direct binding to the USH1G scaffold protein SANS links Magi2 to the USH protein interactome. We showed that the assembly of the SANS-Magi2 complex is procured by a novel non-canonical internal PBM within the SAM domain of SANS. We revealed that this interaction is triggered by CK2-mediated phosphorylation of this PBM. We provided evidence that Magi2 mediates endocytosis in cultured cells and gathered data which suggest a possible role at the ciliary pocket of photoreceptor primary cilia. Furthermore, we demonstrated that these endocytotic processes could be negatively regulated by the binding of phosphorylated SANS to Magi2. Finally, our data suggest that Magi2-mediated endocytosis and its regulation by SANS are essential for the ciliogenesis and maintenance of primary cilia. Since several mutations in the USH1G gene affect the Magi2-binding site in the SAM domain of SANS, we hypothesize that the deregulation of the physiological modulation of the SANS-Magi2 complex may underlay the pathophysiological processes leading to neurosensory degenerations in human USH patients.
The USH1G protein SANS directly interacts with the MAGUK protein Magi2 via its SAM domain
To gain further insights into the composition of USH protein networks and thereby into their functions, we searched for novel proteins directly interacting with the USH1G scaffold protein SANS. For this, we performed a Y2H screen on a retinal cDNA library using the C-terminus of SANS, containing a SAM and a class I PBM, as bait (SAM-PBM) (Fig. 1A). We identified clones encoding three PDZ domain-containing scaffold proteins as putative interaction partners, namely the USH2D protein whirlin (8), the PDZ-domain-containing RING finger protein 4 and the MAGUK protein Magi2 (membrane-associated guanylate kinase inverted-2). Here we focus on the analysis of Magi2.
We validated the interaction between SANS and Magi2 in independent complementary assays on different levels, in vitro, in cell culture and in situ. First, we affirmed the binding of SANS' SAM-PBM to the identified PDZ domain of bovine Magi2 in Y2H 1:1 assays (Fig. 1B). Second, we affirmed the interaction between both proteins in vitro by glutathione-S-transferase (GST)-pull down assays using the last PDZ domain of murine Magi2, here referred to as PDZ6, which pulled down recombinant SANS full-length with and without the PBM (Fig. 1C). To elucidate, which SANS' domains participate in the interaction, we carried out GST-pull downs adopting GST-tagged Magi2-PDZ6 domain and different 3×FLAG-tagged domains of SANS (Fig. 1D). We found that the PDZ6 domain pulls down only the SANS-C-terminus with and without the PBM, but neither the N-terminus nor the central domain of SANS. In contrast, we did not recover any SANS domain in GST-pull down assays with GST alone or other PDZ domains of Magi2, namely GST-Magi2-PDZ4 and -PDZ5 (Fig. 1E). Quantification of the recovered SANS in the GST-pull downs interestingly showed that the recovery was 2.5-fold higher in the case of the SANS full-length lacking the C-terminal PBM compared with the PBM-containing constructs (Fig. 1C and D). Furthermore, in experiments adopting constructs of the SANS' C-terminus without the PBM the recovery was even 4.7-fold higher. Our GST-pull down results conclusively confirmed that SANS directly binds to Magi2 in vitro and that this interaction is mediated by the SAM domain in the C-terminus of SANS. In addition, they provide evidence that the C-terminal PBM of SANS is not necessary for the binding of SANS to the PDZ6 domain of Magi2 and may in fact actually reduce the affinity of SANS to Magi2.
Next, we adopted the bioinformatics tool POW (PDZ domain–peptide interaction prediction website (http://webservice.baderlab.org/domains/POW/) to determine the binding motif in SANS for the PDZ6 domain of Magi2. POW analysis predicted that the PDZ6 domain of Magi2 interacts with an internal five amino acid short SDLDL motif (amino acid 422–426) in the SAM domain with 2.3-fold higher predictor confidence as with the C-terminal PBM of SANS. The alignment of amino acid sequences of SANS' SAM domains from diverse vertebrate species revealed that the SDLDL motif is conserved in all Eutheria species (Table 1). In the marsupial Monodelphis domestica, in birds and the lower vertebrate animal models, zebrafish and frog, non-polar amino acids are replaced by polar amino acids, which even increase the predictor confidence for Magi2-PDZ6 binding.
|Species||NCBI no.||Alignment of SAM domain sequences|
|Species||NCBI no.||Alignment of SAM domain sequences|
To analyze this interaction in cells, we performed co-precipitation assays with recombinant 3×FLAG-SANS and mRFP-Magi2-PDZ6 or full-length GFP-SANS and Magi2-HA, respectively, and precipitated the mRFP- or GFP-tagged polypeptides applying the Trap® bead system (Fig. 1F and G). In these experiments, we showed that SANS co-precipitated with the PDZ6 domain of Magi2 and that vice versa Magi2 co-precipitated with SANS. Next, we performed membrane targeting assays, transfecting IMCD3 cells with MyrPalm-eCFP-tagged SANS and/or Magi2-HA (Fig. 1H). In singly transfected cells, the N-terminal membrane-anchoring MyrPalm-tag attached eCFP-SANS to the plasma membrane (Fig. 1H,a), while Magi2-HA was found in the cytoplasm and nucleus (Fig. 1H,b). However, in co-transfected cells, Magi2 co-localized with SANS at the plasma membrane (Fig. 1H,c), indicating that Magi2 is recruited to the membrane by binding to SANS. This binary interaction between Magi2 and SANS was affirmed by reciprocal experiments utilizing MyrPalm-eCFP-Magi2 and GST-SANS (Supplementary Material, Fig. S1). In this case, Magi2 recruited the cytoplasmic SANS to the membrane in co-transfected cells, which is not the case in the control with MyrPalm-eCFP alone and GST-SANS.
Next, we checked to what extend disease-causing mutations affect SANS-Magi2 complex assembly. So far, mutations in MAGI2 have not been linked to any disease. In silico analyses on the known pathogenic USH disease causing mutations in the USH1G gene SANS by database screening (https://grenada.lumc.nl/LOVD2/Usher_montpellier/) (26) revealed a variety of pathogenic mutations in SANS. With one exception the identified mutations lead to truncations of the SANS molecule, which lack the SAM domain (Supplementary Material, Fig. S2) (17,27–33). Although these truncations also affect other domains, most disease causing mutations in SANS have effects on the predicted Magi2-binding motif and thereby eliminate the SANS-Magi2 complex formation in human USH1G patients.
The interaction of SANS and Magi2 is regulated by phosphorylation
Several previous studies indicated that the binding of molecules to Magi2 is regulated by phosphorylation (34,35). This prompted us to examine, whether the interaction of SANS and Magi2 also depends on phosphorylation. For this, we combined the previously introduced robust membrane targeting assay in IMCD3 cells with the application of d-ribofuranosyl-benzimidazole (DRB), a potent inhibitor of carboxyl-terminal domain kinases (36). In contrast to DMSO-treated controls (Fig. 2A), DRB inhibited the recruitment of Magi2-HA to the membrane-bound MyrPalm-eCFP-SANS (Fig. 2B), suggesting that phosphorylation regulates the assembly of the SANS-Magi2 complex.
To elucidate which interaction partner needs to be phosphorylated for the SANS–Magi2 interaction, we identified putative phosphorylation sites present in the interacting domains of SANS-SAM (S422) and Magi2-PDZ6 (S1152) by BLAST analyses (http://www.phosphosite.org). Subsequently, we mutated these sites to phospho-mimicking and dephospho-polypeptides. We found that in cells co-transfected with the phospho-mimicking SANS construct (MyrPalm-eCFP-SANS S422E) and HA-tagged full-length Magi2 (Magi2-HA), Magi2 was recruited to the cell membrane (Fig. 3A). In contrast, in cells co-transfected with the dephospho-construct of SANS (MyrPalm-eCFP-SANS S422A) and Magi2-HA, Magi2 was no longer found at the cell membrane (Fig. 3B). In cells co-transfected with wild-type full-length SANS and the dephospho-construct of Magi2 (Magi2-HA S1152A), we observed that Magi2 was still recruited to SANS at the cell membrane (Fig. 3C). In conclusion, present results indicate that the phosphorylation of SANS is essential for the assembly of the SANS-Magi2 complex.
Next, we addressed whether the interaction of the endogenous proteins depends on phosphorylation. Preliminary quantitative real-time PCRs (qPCRs) revealed that SANS and Magi2 were expressed in murine IMCD3 cells and knock down experiments validated the specificity of the antibodies for both proteins (Supplementary Material, Figs S3 and S4). Indirect immunofluorescence demonstrated that in untreated and DMSO-treated cells both proteins were expressed in the pericentriolar region of the centrosome, visualized by antibodies against the pericentriolar marker protein PCM1 (Supplementary Material, Fig. S5A). In DRB-treated cells, SANS was no longer concentrated in the pericentriolar matrix, where Magi2 and PCM1 remained, but was found in the nucleus instead (Supplementary Material, Fig. S5B). The disappearance of SANS from the pericentriolar region was more evident comparing the intensity profiles of SANS and Magi2 in control and DRB-treated cells in the region of interest (ROI, defined by PCM1-staining) as exemplarily shown (Supplementary Material, Fig. S5C and D).
The SAM domain of SANS is phosphorylated by CK2
DRB is widely used as a potent inhibitor for CK2 (casein kinase 2) (36) and BLAST analysis (http://www.phosphosite.org) predicted the serine S422 in the SAM domain of SANS as a putative phosphorylation site for the protein kinase CK2. To experimentally validate whether SANS-SAM is a substrate for CK2, we performed phosphorylation in vitro assays by P32 incorporation (Fig. 4). Quantification of the radioactivity incorporation revealed that the phosphorylation of GST-SAM-PBM was significantly increased ∼4-fold in comparison with GST alone (Fig. 4B), indicating that CK2 phosphorylates S422 in the SAM domain of SANS.
Magi2-mediated endocytosis is negatively regulated by SANS binding
Previous studies have indicated a function of Magi2 in endocytotic processes in neurons (23,24,37,38). To analyze the role of the Magi2-SANS complex in endocytosis, we performed transferrin uptake assays in IMCD3 cells. Fluorescence microscopy analysis of the time course of Alexa 647-tagged transferrin (Tf647) uptake showed that fluorescent Tf647-vesicles appeared first in the cytoplasm before they concentrated in the perinuclear region of the cell. Immunofluorescent counterstaining of Magi2 revealed the close association of Magi2 with Tf647 vesicles at each time point, which was most prominent after 30 min (Fig. 5A, lower panel). To confirm that Tf647 was taken up by endocytotic processes, we treated cells with dynasore, which blocks clathrin-mediated endocytosis by inhibiting the dynamin GTPase (39). In dynasore-treated cells, Tf647 was no longer found in the cytoplasm, but remained at the cell membrane, indicating that the monitored transferrin uptake is based on clathrin-mediated endocytosis (Fig. 5B).
Next, we examined whether the formation of the Magi2-SANS complex is related to endocytosis. For this, we measured the transferrin uptake after knock down of Magi2 and SANS, respectively, using specific short hairpin RNA (shRNA), which we previously validated in IMCD3 cells (Supplementary Material, Figs S3 and S4). Knock down of Magi2 significantly reduced the Tf647 uptake by IMCD3 cells to 57% (Fig. 6A), while the overexpression of GFP-Magi2 significantly increases the Tf647 uptake by 43% compared with GFP-transfected cells (Supplementary Material, Fig. S6). shRNA-mediated knock down of SANS significantly increased the transferrin uptake by IMCD3 cells by 62% compared with mock-transfected cells and Tf647 uptake in cells transfected with the shRNA-empty vector showed a slight increase by 32% (Fig. 6B). Furthermore, we analyzed whether the observed endocytosis is dependent on the phosphorylation. For this, we treated IMCD3 cells with the kinase inhibitor DRB, which induces dephosphorylation as shown above and measured the uptake of Tf647. Surprisingly, we observed a 0.63-fold increase of Tf647 fluorescence in DRB-treated cells (Fig. 6C). Taken together, these data displayed that Magi2 mediates endocytosis and that SANS' phosphorylation negatively affects endocytosis, suggesting a negative regulation of Magi2-mediated endocytosis by the phosphorylation-dependent SANS-Magi2 binding.
SANS and Magi2 knock down and inhibition of endocytosis affect ciliogenesis of primary cilia
Previous studies demonstrated that endocytosic processes are engaged in ciliogenesis (40,41). Immunofluorescence labeling of SANS and Magi2 in starved IMDC3 cells revealed the localization both proteins in the periciliary region at the base of primary cilia (Supplementary Material, Fig. S7A and B). Furthermore, in transferrin uptake-assays fluorescent Tf647 vesicles were found in the periciliary region at the base of primary cilia of IMCD3 cells (Supplementary Material, Fig. S7C) highlighting the periciliary compartment as a region of prominent endocytosis. This prompted us to test whether the complex partners contribute to ciliogenesis and/or maintenance of primary cilia. We knocked down either SANS or Magi2 via the specific shRNA and induced ciliogenesis in IMCD3 cells by serum starvation. Knock down of SANS resulted in heterogeneous ciliary phenotypes in transfected cells, namely no cilia were found in ∼77% of the cells, while cilia of the few ciliated cells were either shorter (∼12%; <1.9 µm) or longer (∼4%; >2 µm) than control primary cilia or exhibited multiple cilia (∼6.8%) (Fig. 7). The variable ciliary phenotype may reflect SANS participation in several cellular functional modules related to cilia.
Knock down of Magi2 abolished ciliogenesis in 82% of transfected IMCD3 cells compared with control (Fig. 8A and C). In 18% of Magi2 shRNA-transfected cells cilia were present, but reduced in length (<1.5 µm) compared with control (Fig. 8B). To examine the link between endocytosis and ciliogenesis, we treated IMCD3 cells with the dynamin-inhibitor dynasore during serum starvation. Inhibition of clathrin-dependent endocytosis eliminated primary cilia formation in 70% of dynasore-treated cells compared with control cells (Fig. 8B and D). Analysis of the ciliary shape revealed a reduction in length of the few emerging primary cilia (<1.7 µm) compared with control (Fig. 8Bc,d). Taken together, these results indicated that SANS affected ciliogenesis and that Magi2 and clathrin-dependent endocytosis were essential for the ciliogenesis of primary cilia in IMDC3 cells.
Subcellular localization of Magi2-USH protein complexes in mouse retina
Next, we were interested in the role of the Magi2-SANS protein complexes in the mammalian retina. For this, we examined where Magi2 and SANS are co-expressed in situ, which is a necessary prerequisite for the interaction of proteins within a functional module in vivo. We performed immunofluorescence labeling of Magi2 in longitudinal cryosections through the murine retina. Magi2 was present throughout the retina, except for the nuclear layers and the photoreceptor outer segments (Fig. 9B). The most prominent staining was observed in the outer plexiform layer, at the outer limiting membrane and the inner segment of photoreceptor cells. Immunofluorescence double-labeling with antibodies against Magi2 and centrin-3, a ciliary marker which stains the connecting cilium, the basal body and the adjacent daughter centriole (36) revealed the localization of Magi2 in the periciliary region of the apical inner segment (Fig. 9C). Next, we illuminated the putative co-localization of Magi2 and SANS by double-labeling of both interacting proteins in retinal cryosections. Immunofluorescence microscopy analysis revealed a substantial co-staining of Magi2 and SANS in the inner segment and in the ciliary region of photoreceptor cells (Fig. 9D). To elucidate the localization of SANS and Magi2 in the ciliary region of photoreceptor cells, we triple labeled both proteins and centrin-2 in retinal cryosections. Fluorescent microscopy revealed the considerable co-localization of SANS and Magi2 at the base of the photoreceptor connecting cilium (Fig. 9E). Next, we were interested to determine whether SANS and Magi2 molecules form complexes in the retina in situ. For this, we carried out in situ proximity ligation assays (PLAs) on mouse retinal sections applying antibodies against SANS and Magi2. Counterstaining with centrin-2 and 4′,6-diamidin-2′-phenylindoldihydrochlorid (DAPI) enabled us to determine the localization and subsequent quantification of the PLA signals in retinal layers (Supplementary Material, Fig. S8). Our analysis revealed the concentration of SANS-Magi2 PLA signals in the inner segment and the ciliary region of retinal photoreceptor cells. Higher magnification images of the ciliary region showed the localization of the SANS-Magi2 PLA signal in the periciliary region of photoreceptor cells (Fig. 9F and G). Taken together, immunocytochemistry demonstrated co-localization of SANS and Magi2 at the base of the photoreceptor cilium and PLA data indicate that both proteins can also interact there.
Magi2 is part of an endocytosis module in the periciliary region of photoreceptor cells
Subsequently, we investigated the precise localization of Magi2 in the ciliary and periciliary region of photoreceptor cells. Adopting immunoelectron microscopy, we have previously shown that SANS is localized in the apical inner segment in the periciliary region, at the centriole-basal body complex, and in lower amounts in the connecting cilium of mouse and human rod photoreceptor cells (8,19). Electron microscopic analysis demonstrated connatural subcellular localization of Magi2 (Fig. 10A and B), further supporting our light microscopy results. In addition, we detected Magi2 at dents of the plasma membrane of the ciliary pocket coating membrane vesicles in the cytoplasm of the apical inner segment (Fig. 10B,b). These data suggest that Magi2 associates with membrane vesicles budding from the ciliary pocket into the periciliary region of photoreceptor cells.
To explore further putative components of the endocytotic system in the periciliary compartment of photoreceptor cells, we stained cryosections through the mouse retina for the transferrin receptor (TfR), which is characteristically associated with receptor-mediated endocytosis. Immunofluorescence microscopy showed that antibodies against the transferrin receptor labeled the apical inner segment of retinal photoreceptor cells (Fig. 10C). The high resolution of immunoelectron microscopy revealed that antibodies against the extracellular parts of the receptor decorated predominantly the extracellular face of the plasma membrane in the ciliary pocket of photoreceptor cells (Fig. 10D and E). Additional less intense labeling of the transferrin receptor was present along other parts of the membrane of the apical inner segment. Taken together, present data suggest that transferrin receptor-mediated endocytosis takes place at the ciliary pocket of photoreceptor cells.
In this study, we revealed by complementary interaction assays and in situ localization analyses that the USH1G protein SANS directly interacts with the MAGUK protein Magi2. Our data showed that SANS binds via its SAM domain to the PDZ6 domain of Magi2 displaying a novel non-canonical targeting mode for PDZ domain interactions. PDZ domains mediate protein–protein interactions via targeting PBMs at the C-terminus (42), but less frequently also recognize internal peptide sequences of ligand proteins (43). We have previously described a canonical PDZ/target interaction between both N-terminal PDZ domains of the USH2D protein whirlin and the SANS' C-terminal PBM (8). More recently, an unexpected binding mode for PDZ and SAM domains has been described for the USH1C protein harmonin and SANS (7). Yan and coworkers demonstrated a highly stable complex of harmonin and SANS by synergistic interaction of PDZ1 domain of harmonin with the C-terminal PBM-I SANS and the N-terminus of harmonin with the SAM domain of SANS. Here, we provide evidence that an internal PBM consisting of five highly conserved amino acids in the central part of the SAM domain of SANS directly access the PDZ6 domain of Magi2. In contrast to the assembly of the SANS/harmonin complex, the C-terminal PBM of SANS does not contribute to the SANS–Magi2 interaction, but may even reduce the affinity of SANS to Magi2. Although SAM domains are among the most common protein domains in eukaryotic genomes (44), to our knowledge the present binding module is the first between a SAM and a PDZ domain. The depicted interaction mode between SANS and Magi2 via the SAM domain provides SANS an expanded variance in its scaffold function, enabling SANS to bind with its C-terminal PBM-I to a second PDZ domain-containing protein like whirlin in the PMC (8).
Although it is well accepted that interactions mediated by PDZ domains are not static, there are only a few molecular examples known, how the plasticity of PDZ domains in ligand binding is regulated (43,45). Here, we demonstrated that the phosphorylation of the ligand SANS regulates its binding to a PDZ domain of Magi2. Our data conclusively show that the assembly of the SANS-Magi2 complex is strictly dependent on the phosphorylation of serine S422 in the SAM domain of SANS by the protein kinase CK2. S422 is the first of the five amino acids predicted as the internal PDZ-binding site for the PDZ6 domain of Magi2. From this, it is apparent that phosphorylation introduces the conformational change that facilitates the binding of SANS to Magi2. The differential regulation of molecular targeting to the scaffold molecule SANS by reversible phosphorylation provides the plasticity to participate in different functional cellular modules, which we and others have previously described for SANS in the protein interactome related to USH (8,19,46). To our knowledge, it is the first regulation mechanism characterized in the USH protein interactome so far, and in the future it will be interesting to address whether the phosphorylation by CK2 or other kinases regulates only the assembly of the SANS-Magi2 complex or also controls the interaction between other interactome partners, which have been previously identified (8,11,18,19). It is certainly possible that the scaffold function of SANS is modulated by phosphorylation in general.
There is emerging evidence that Magi2 plays important roles in the development and function of synapses, participating in receptor endocytosis and postendocytotic trafficking (20,21,23). Our findings coherently affirm that Magi2 participates in receptor-mediated endocytosis. The localization of Magi2 at synapses of the inner and outer plexiform layer suggests that Magi2 also supports endocytosis in retinal neurons. However, our present data revealed Magi2 as a component of the PMC in the apical photoreceptor inner segment, facing the ciliary pocket. In diverse cell lines, the ciliary pocket of primary cilia has been recently demonstrated as a hot spot for frequent endocytosis (14,15,47,48). Here we showed that Magi2 is essential for ciliogenesis and/or maintenance of primary cilia in IMDC3 cells. Furthermore, we demonstrated that the inhibition of endocytosis also drastically reduced the formation of primary cilia indicating that endocytosis is required for cilia morphogenesis. Our data support previous findings on endocytotic processes at the ciliary base for controlling ciliary and periciliary membrane homeostasis (49). Interestingly, endocytosis inhibition phenocopies the ciliary phenotype induced by Magi2 depletion, suggesting that Magi2-mediated endocytosis is essential for ciliogenesis and/or maintenance of primary cilia.
The association of Magi2 with putative endocytotic vesicles in the periciliary compartment suggests that Magi2 also mediates endocytosis in photoreceptor cells. This hypothesis is further supported by the clustering of transferrin receptors, a hallmark for receptor-mediated endocytosis (50), at the ciliary pocket membrane of photoreceptor cells. In addition, the abundance of transferrin receptors suggests that their ligand transferrin is taken up in the ciliary pocket of photoreceptor cells. Transferrin is mainly synthesized in the retinal pigment epithelial cells and assists in shuttling iron to photoreceptor cells (51). Since iron is an important cofactor of various enzymes but can also generate harmful free radicals (52), the control of its endocytosis is essential for the cellular homeostasis (53). In the retina, unbalanced iron homeostasis leads to photoreceptor pathology (54), and previous studies have monitored death of photoreceptor cells caused by either a lack or an excess of iron (55,56).
In primary cilia, it has been previously hypothesized that membrane cargoes cross the periciliary diffusion barrier by endocytosis to enable ciliary import (41,57). Our present findings support this hypothesis indicating that Magi2 expression and the maintenance of the endocytosis machinery are essential for ciliogenesis and/or retention of primary cilia. It is worth to speculate that Magi2-related endocytosis participates in the tight control of ciliary membrane cargo delivery and membrane retrieval via endocytosis at the base of primary cilia in general and at the PMC of photoreceptor cells in particular.
There is growing evidence that the scaffold protein SANS integrates into distinct protein complexes which associate with functional modules related to the microtubule cytoskeleton, namely microtubule-based transport and centriolar/centrosomal function (8,9,19,58). The present SANS knock down resulted in diverse ciliary phenotypes: the deprivation, but also the facilitation of cilia growth as well as multiple ciliation. These findings suggest that SANS participates in the regulation of different aspects of ciliogenesis and/or maintenance. As known from other centrosomal components the formation of multiple cilia may be based on a role of SANS in the regulation of centriole duplication (59). Failure of centriole duplication control leads to an increase of centrioles and basal bodies nucleating primary cilia growth as found in SANS-depleted cells. The directly opposed phenotype in ciliary length may reflect on the one hand a ciliary transport defect in cells with either no or short cilia and on the other hand an amplified ciliary growth in cells with elongated cilia, respectively.
The next important question is how SANS contributes to the endocytosis module in the ciliary pocket of the primary cilia of photoreceptor cells. Our previous work gave the rise to the hypothesis that SANS might be part of a protein complex in the intracellular transport module for cargo transport through the inner segment and of vesicle cargo targeting to the PMC of photoreceptor cells (8,19). Here, we demonstrated that SANS and Magi2 are co-expressed and can assemble into a protein complex in the periciliary region of photoreceptor cells. These data suggest that the interaction between SANS and Magi2 may participate in targeting cargo vesicles to periciliary membranes, which is followed by subsequent vesicle docking and fusion. However, this vesicle membrane delivery resembles an exocytosis process, but not endocytosis mediated by Magi2 as discussed above. Interestingly, we showed that the Magi2-mediated endocytosis is negatively regulated by binding of phosphorylated SANS. One possible model for the SANS-Magi2 function in the periciliary region of photoreceptor cells is that, upon approaching of the SANS transport complex at the periciliary compartment, SANS is phosphorylated by the periciliary resident protein kinase CK2 (36,60) (Fig. 11B). Thereby, the affinity of phospho-SANS for Magi2 increases and the phospho-SANS-Magi2 complex assembles, which in turn shuts down endocytosis. This may facilitate exocytosic processes placing the cargo vesicular into the periciliary membrane for further translocation into the outer segment (Fig. 11A).
In conclusion, our study uncovers a novel mechanism, by which SANS controls Magi2-mediated endocytosis which may be conjoin to vesicular cargo delivery machinery in ciliated photoreceptor cells and cilia in general. Most disease causing mutations in SANS also eliminate the SANS-Magi2 complex formation. These defects may unbalance the regulation of these processes, ultimately leading to ciliary pathogenesis observed in human USH1G patients.
MATERIAL AND METHODS
Antibodies and fluorescent dyes
Following antibodies were used: anti-TfR (sc-65877) and anti-Magi2 (sc-25664); anti-tubulin, anti-FLAG and anti-Magi2 (SAB1404762) from Sigma–Aldrich (Hamburg, Germany); anti-HA antibodies from Roche Diagnostics (Mannheim, Germany), anti-RFP antibodies from ChromoTek (Planegg-Martinsried, Germany), anti-actin antibodies from Millipore GmbH (Schwalbach, Germany) and anti-GST from GE Healthcare (Munich, Germany). Novel SANS-CENT domain antibodies ( amino acid 339–384) were generated in guinea pig and validated by specific antigen-recognition, shRNA knock down and qPCR (Supplementary Material, Fig. S3). Different centrin antibodies (61) were used as ciliary markers. Anti-GFP antibodies were a gift from W. Clay Smith (Gainesville, FL, USA) (62). Alexa647®-conjugated transferrin and secondary antibodies conjugated to Alexa488®, Alexa568® or Alexa647® were purchased from Molecular Probes® (Life Technologies, Darmstadt, Germany) or from Rockland Inc. (Gilbertsville, PA, USA). Biotinylated secondary antibodies (Vector Laboratories, Burlingame, CA, USA) were used for pre-embedding electron microscopy. Nuclear DNA was stained with DAPI (1 µg/ml) (Sigma–Aldrich).
Y2H screen of retinal cDNA libraries was performed as previously described (63). Briefly, SAM-PBM domain of human SANS (NCBI:NM_173477.3; amino acid 385–461) fused to the DNA-binding domain of GAL4 transcription factor was used as bait to screen a bovine oligo-dT primed retinal cDNA library, fused to DNA-activation domain of GAL4. Interactions were identified and analyzed by assessment of the HIS1, ADE3, LacZ and MEL1 reporter genes. To confirm interactions, bait and prey were co-transformed in the yeast strain PJ694α and analyzed as described above.
A clone of murine Magi2 (NCBI: NM_001170745) was used lacking the encoding sequence of the first PDZ domain of the canonical isoform Magi2 long (NCBI: NM_001170746). Nevertheless, the PDZ domains are consecutively numbered from PDZ1 to PDZ6 and sequence information in text and figures alluded to the canonical isoform (UniProtKB:Q9WVQ1). The PDZ4 (amino acid 777–859), PDZ5 (amino acid 919–1009) and PDZ6 (amino acid 1139–1221) domains of murine Magi2 were subcloned in the pDEST15 vector (Gateway® pDEST™15, Invitrogen™), expressed in Escherichiacoli BL21AI and bound to glutathione–sepharose beads (GE Healthcare) as described in (61). 3×FLAG-tagged recombinant polypeptides were generated in HEK293 cells transfected with the appropriate expression plasmids. 24 h post-transfection, cells were washed with PBS (phosphate-buffered saline; 140 mm NaCl, 2.7 mm KCl, 10 mm Na2HPO4 × 2H2O, 1.8 mm KH2PO4, pH 7.4) and lysed in lysis buffer (50 mm Tris–HCl, pH 7.5, 150 mm NaCl, 0.5% Triton X-100) containing protease inhibitor cocktail (PI-mix; Roche Diagnostics). Cell lysates were incubated for 2 h with GST fusion proteins bound to glutathione–sepharose beads at 4°C under constant shaking. Samples were washed with lysis buffer and precipitated protein complexes were eluted with SDS-sample buffer and subjected to SDS–PAGE and western blot employing the Odyssey InfraRed imaging system (LI-COR Biosciences, Lincoln, NE, USA) for detection.
Recombinant GST-SAM-PBM was bound to beads as described above and subjected to kinase assays. Briefly, samples were divided in equal aliquots and incubated with 2 µCi [γ-32P]-ATP (PerkinElmer, Rodgau, Germany) in 500 ml kinase buffer (20 mm Tris–HCl, pH 7.5, 50 mm KCl, 10 mm MgCl2, 1 mm EGTA, PI-mix) including 62.5 units CK2 (Calbiochem, EMD Millipore) or kinase buffer alone as control for 1 h at 30°C. Reactions were stopped by washing with kinase buffer. Samples were subjected to SDS–PAGE (input and size control by Coomassie staining) and analyzed by measuring the incorporated and residual radioactivity using a 2200CA Tri-Carb® Liquid Scintillation Analyzer (Canberra Packard, Frankfurt, Germany) or autoradiography. Imaging was done by exposition of X-ray films.
GFP- or RFP-fused polypeptides were immobilized at Trap® agarose beads (ChromoTek) and used for co-precipitation assays according to the manufacturer's protocol. Briefly, cell lysates from co-transfected HEK293 cells (mRFP-tagged PDZ6 domain of Magi2 or mRFP alone together with 3×FLAG-SANS, or GFP-SANS and Magi2-HA) were suspended in lysis buffer (10 mm Tris–Cl, pH 7.5, 150 mm NaCl, 0.5 mm EDTA, 0.5% NP-40), spun and the supernatant was diluted to 500 µl in dilution buffer (10 mm Tris–Cl, pH 7.5, 150 mm NaCl, 0.5 mm EDTA). Fifty microliters were separated as input (total cell lysate) and samples were added to equilibrated beads for 2 h at 4°C under constant shaking. After washing, precipitated protein complexes were eluted with SDS-sample buffer and subjected to SDS–PAGE and western blots.
We cultured HEK293 (human embryonic kidney cells) in Dulbecco's modified Eagle's medium (DMEM) for high expression levels and IMCD3 (mouse inner medullary collecting duct cells) cells in DMEM-F12 for intrinsic protein analyses containing 10% heat-inactivated fetal calf serum (FCS). Cells were transfected with plasmids using Lipofectamine® LTX and Plus Reagent (Invitrogen™, Karlsruhe, Germany) according to manufacturer's instructions. To induce ciliogenesis, IMCD3 cells were starved 24 h after seeding or transfection in low serum medium containing 0.5% FCS for additional 48 h.
Membrane targeting assay
Human SANS (amino acid 2–461) and murine Magi2 (amino acid 2–1112) were subcloned in the MyrPalm-eCFP vector (plasmid 14867, Addgene, Cambridge, MA, USA) (64) containing an N-terminal membrane-anchoring peptide and eCFP. MyrPalm-eCFP-SANS was single- or co-transfected with murine Magi2 (amino acid 2–1112) C-terminally tagged with HA (Magi2-HA). 24 h post-transfection, cells were subjected to immunocytochemistry. For kinase inhibition assays, transfected cells were incubated in medium containing 150 mm DRB (BIOMOL, Hamburg, Germany) or DMSO (dimethyl sulfoxide) as control.
Expression plasmids containing SANS or Magi2 were used for mutagenesis to generate dephospho-mimicking constructs. Primers were designed on Agilent homepage (http://www.genomics.agilent.com) and mutagenesis PCR was performed according to the manufacturer's instructions (Agilent Technologies, Waldbronn, Germany) (for used primers, see Supplementary Material, Table S1).
Cells were fixed with methanol containing 0.05% EGTA, air-dried and subsequently washed with PBS and incubated with 0.01% Tween 20 for 10 min. shRNA-transfected cells or cells in transferrin uptake assays were fixed with 3% PFA for 15 min at 4°C, washed with PBS and incubated with NH4Cl and 0.01% Tween 20 for 10 min each. Subsequently, cells were incubated in blocking solution (0.5% cold-water fish gelatin, 0.1% ovalbumin in PBS) for at least 30 min before primary antibodies were incubated overnight at 4°C. After washing, samples were incubated with secondary antibodies and DAPI for 1.5 h at room temperature. After washing, cover slips were mounted in Mowiol (Roth, Karlsruhe, Germany). Samples were analyzed with a Leica DM6000B microscope (Leica, Bensheim, Germany) and images were processed with Adobe Photoshop CS for reducing gray scales, merge images, changing colors, changing image size and higher magnifications (Adobe Systems, San Jose, CA, USA).
shRNA against Magi2 and SANS
shRNAs against SANS were purchased from OriGene (Rockville, MD, USA). Twelve hours after transfection, IMCD3 cells were treated with puromycin (Roth) for selection of transfected cells. For Magi2 knock down, double-stranded 21-nucleotide oligos were annealed overnight and cloned with the restriction sites for HpaI and HindIII into the shRNA vector (pAAV2.1-sc-shRNA-CMV-eGFP). Magi2 gene-specific (NCBI Gene ID: 50791) inserts named Oligo 1–3 were 21-nucleotide sequence separated by AAGTTCTCT non-complementary spacers from the reverse complement of the same 21-nucleotide sequence (for used oligos, see Supplementary Material, Table S1). shRNA were transfected into IMCD3 cells with 50% confluence using LTX. Knock down efficiency was validated 72 h posttransfection by anti-SANS or anti-Magi2 western blots and by qPCR. shRNA03/Oligo3 showed strongest knock down efficiency compared with empty vectors and normalized to actin or to tubulin, respectively. Seventy-two hours of posttransfection cells were prepared for uptake assays and/or immunocytochemistry.
Reverse transcription and qPCR
Total RNA was isolated from cells by RNA isolation kit from Macherey-Nagel (Düren, Germany). Reverse transcription was performed with 1 µg of total RNA with the SuperScript III First Strand synthesis kit (Invitrogen) following manufacturer's instructions with a mixture of random hexamers and oligo-dT primers. qPCR was performed on CFX96 real-time system (Bio-Rad, Munich, Germany) using the SYBRGreen iTAQ according to manufacturer’s instructions. In a total volume of 20 μl, diluted cDNA (1:5) and 4 nm of each primer/reaction were used. Cycling conditions were 95°C for 30 s, 45 cycles at 95°C for 5 s, 60°C for 30 s followed by plate read, and melt curve analysis for 65–95°C using primers, specific for murine Magi2/SANS (Supplementary material, Table S1.) Quantification was done with three independent experiments.
Transferrin uptake assay
At 75% confluence IMCD3 cells were incubated in serum-free media for 2 h at 37°C to remove endogenous iron. Alexa-647 labeled transferrin (Tf647®) was diluted in serum-free media (10 µg/ml) and incubated for 30 min at 4°C. After washing with pre-warmed, serum-containing medium, cells were incubated for additional 5–30 min at 37°C in serum-containing medium. Washing with PBS containing 0.5% acidic acid to remove surface-bound transferrin was followed by fixation with 3% PFA. Quantification of non-processed images was performed with CellProfiler 2.0 (r11710) cell image analysis software (Broad Institute, Cambridge, MA, USA). All experiments were done three to five times; at least five images from each sample were used for quantification. In overexpression experiments, cells were transfected 24 h before uptake assay with GFP or GFP-Magi2 (amino acid 2–1112) using LTX®. In dephosphorylation experiments, cells were grown over night in DRB-medium and DRB was added to all media used during uptake assay. To inhibit endocytosis, IMCD3 cells were treated with 80 µm dynasore (Dynasore hydrate D7693, Sigma–Aldrich) solved in DMSO.
Eyes of mice were cryofixed in melting isopentane and cryosectioned as described elsewhere (65). Cryosections were placed on poly-l-lysine-precoated coverslips, incubated with 0.01% Tween 20 PBS, washed several times, covered with blocking solution and incubated for minimum 30 min followed by overnight incubation at 4°C with primary antibodies. Washed cryosections were incubated with secondary antibodies in blocking solution containing DAPI (Sigma) for 1.5 h at room temperature. After washing, sections were mounted in Mowiol (Roth). Specimen were analyzed on a Leica DM6000B microscope, images were processed with Leica imaging software and Adobe Photoshop CS. Images in Figures 8, 9, 10F and G, 11 C, Supplementary Material, Figures S4A,B and S6 were 3D deconvoluted with Leica imaging software (three iteration steps).
Proximity ligation assay
In situ PLA was developed to support the interaction of two co-localized proteins by visualizing protein–protein interactions with a single-molecule resolution (66). Following PLA probes were purchased from Olink Bioscience (Uppsala, Sweden): Duolink PLA probe anti-rabbit PLUS, anti-mouse PLUS, anti-guinea pig MINUS and Duolink in situ Detection Reagent Red. PLAs were performed according to manufacturer’s protocol adapted to our immunohistochemistry protocol, applied on unfixed cryosections of murine retina. Briefly, cryosections were incubated overnight at 4°C with primary antibodies, subsequently counterstained with anti-centrin-2 for 2 h before fixation with 2% paraformaldehyde in PBS. Next, PLA probes were added to the sections for 2 h at room temperature. Ligation was performed for 30 min at 37°C and amplified for 100 min. For quantification of PLA signals, we applied the open source software ImageJ (http://rsb.info.nih.gov/ij/) to separate the different layers of photoreceptor cells, namely RPE via DIC image, nuclear layers via DAPI-staining and the ciliary region via centrin staining. In these regions, the program counted red fluorescent particles automatically. We used five different images of one representative experiment for quantification of absolute number of PLA signals per area in each region. To test the significance, the number of signals in the experiment were alluded to the controls.
For immunoelectron microscopy, we applied the pre-embedding labeling protocol previously described in (67). Ultrathin sections were analyzed in a transmission electron microscope (Tecnai 12 BioTwin; FEI, Eindhoven, The Netherlands). Images were obtained with a charge-coupled device camera (SIS Megaview3; Surface Imaging Systems), acquired by analySIS (Soft Imaging System) and processed with Adobe Photoshop CS.
Animals and tissue dissection
All experiments described herein are conforming to the statement by the Association for Research in Vision and Ophthalmology as to care and use of animals in research. C57BL/6J mice were maintained under a 12 h light–dark cycle, with food and water ad libitum. After sacrifice of the animals in CO2 and decapitation, subsequently appropriate tissues were dissected.
All values represent means ± SD. To probe for significance of observed differences, the Student's t-test was performed (unpaired, two tailed, assuming equal variance) with individual data-points from minimum three independent experiments. P-value of >0.05 was considered to be significant.
This work was supported by the BMBF ‘HOPE2’ (01GM1108D to U.W.), DFG (GRK 1044 to U.W.), Forschung contra Blindheit-Initative Usher Syndrom (to H.K. and U.W.), ProRetina Deutschland eV (to U.W. and K.B.), the FAUN-Stiftung, Nurnberg (to U.W.), European Community FP7/2009/241955 (SYSCILIA) (to H.K., R.R. and U.W.) and FP7/2009/242013 (TREATRUSH) (to U.W.), the LSBS and The Foundation Fighting Blindness C-CMM-0811-0547-RAD03 (to H.K.).
The authors thank Ulrike Maas, Elisabeth Sehn and Gabi Stern-Schneider for excellent technical assistance; Dr W. Clay Smith for providing the anti-GFP antibodies; Dr Stylianos Michalakis for providing the shRNA AAV-vector; Anna Valeria Etz for testing the usability of the software CellProfiler and Dr Kerstin Nagel-Wolfrum for abundant discussion.
Conflict of Interest statement. None declared.