| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1From The Netherlands Ophthalmic Research Institute, Amsterdam, The Netherlands; the 3University Medical Centre Nijmegen, Nijmegen, The Netherlands; and the 4Centre for Molecular and Biomolecular Informatics, Nijmegen, The Netherlands.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
METHODS. To dissect the protein scaffold connected to CRB1, the yeast two-hybrid approach was used to screen for interacting proteins. Glutathione S-transferase (GST) pull-down analysis and immunoprecipitation were used to verify protein–protein interactions. The subcellular localization of the proteins was visualized by immunohistochemistry and confocal microscopy on human retinas and immunoelectron microscopy on mouse retinas.
RESULTS. A novel member of the scaffold connected to CRB1, called membrane palmitoylated protein (MPP) subfamily member 4 (MPP4), a membrane-associated guanylate kinase (MAGUK) protein, was identified. MPP4 was found to exist in a complex with CRB1 through direct interaction with the MPP subfamily member MPP5 (PALS1). 3D homology modeling provided evidence for a mechanism that regulates the recruitment of both homo- and heterodimers of MPP4 and -5 proteins to the complex. Localization studies in the retina showed that CRB1, MPP5, and MPP4 colocalize at the outer limiting membrane (OLM).
CONCLUSIONS. These data imply that MPP4 and -5 have a role in photoreceptor polarity and, by association with CRB1, pinpoint the cognate genes as functional candidate genes for inherited retinopathies.
Several members of the membrane associated guanylate kinase (MAGUK) protein family are involved in cell polarity through their role in large multiprotein complexes at tight junctions.3 4 This protein family is characterized by a specific set of protein-binding domains, consisting of one or more PDZ (postsynaptic density 95/discs large/zonula occludens 1) domains, an SH3 (Src homology 3) domain, and a region with homology to a guanylate kinase (GuK) domain.5 6 Some members, such as PALS1 (assigned the name MPP5 by the HUGO Gene Nomenclature Committee; gene.ucl.ac.uk/nomenclature; hosted by the University College London, London, UK), ZO1, and PSD-95, have been shown to be localized at sites of cell–cell contact (e.g., synapses and epithelial tight junctions).7 8 9 10 It is currently assumed that MAGUK proteins serve as scaffolds by recruitment of other MAGUKs, eventually linking them to the cell cytoskeleton or to the carboxyl terminus of transmembrane proteins.3 11 12
The MAGUK protein Stardust is the Drosophila homologue of MPP5, one of the seven mammalian membrane palmitoylated protein (MPP) subfamily members. Stardust mutants exhibit severe disruption in apicobasal polarity in embryonic epithelia.13 In addition, mutants of the zebra fish homologue Nagie oko display severe defects in the organization of the retinal cell layers,14 and loss of Stardust gives rise to an eye phenotype in Drosophila characterized by a shortened stalk membrane and altered rhabdomere morphogenesis resembling the Crumbs mutant phenotype.15 16 17
Crumbs is an apically localized transmembrane protein involved in organizing the apical plasma membrane subdomains.18 19 20 21 Stardust has been found to colocalize with Crumbs and interact physically with the C-terminal ERLI motif of Crumbs through its PDZ domain.22 This interaction was recently also identified for their mammalian homologues MPP5 and CRB110 as well as for MPP5 and CRB3.23 Stardust and Crumbs are both necessary to ensure stability, localization, and function in controlling the apicobasal polarity of epithelial cells.18 22 The Crumbs-Stardust protein complex also recruits the Drosophila protein associated with tight junctions (DPATJ; formerly known as Discs lost). This cytoplasmic multi-PDZ domain protein interacts indirectly, via Stardust, with the cytoplasmic tail of Crumbs. These three proteins colocalize in Drosophila photoreceptors during and after eye development.17 The mammalian CRB1-MPP5-PATJ complex localizes to tight junctions where it may control cell polarity.10 In the mouse retina, the CRB1-MPP5-PATJ proteins colocalize at the apical region adjacent to adherens junctions of photoreceptors.24 The Drosophila Crumbs protein and the human homologue CRB1 are 35% similar in amino acid sequence and contain the same conserved protein motifs.25 Mutations have been identified in the CRB1 gene in individuals with Leber congenital amaurosis (LCA); retinitis pigmentosa (RP) type 12, with preservation of para-arteriolar retinal pigment epithelium (PPRPE); RP with Coats-like exudative vasculopathy; and early-onset RP without PPRPE.26 27 28 29 CRB1 has been found to maintain adherens junctions between photoreceptor cells and Müller glia cells,30 thus preventing delamination of the photoreceptor layer and death of retinal neurons.24 31
In this study we identified a second MPP subfamily member, MPP4, existing in a complex with CRB1 through direct interaction with MPP5. The presence of MPP4 and -5 in this CRB1 protein complex implicates these proteins in photoreceptor polarity and putatively in inherited retinal dystrophies.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
The following constructs were made by PCR using the a cloning system (Gateway; Invitrogen, Groningen, The Netherlands), according to the manufacturer’s procedures, using full-length constructs as a template: the intracellular domains of bovine CRB1 (bCRB1intra) and human CRB1 (hCRB1intra); the full-length human MPP5 (MPP5FL); the SH3 and HOOK domains of human MPP5 (MPP5SH3+HK); the C terminus of human MPP5 from the HOOK domain (MPP5HK-end); the PDZ domain of human MPP5 (MPP5PDZ); the coiled-coil domain of human MPP5 (MPP5CC); the GuK domain of human MPP5 (MPP5GuK); full-length human MPP4 (MPP4FL); the PDZ domain of human MPP4 (MPP4PDZ); the C terminus of MPP4 containing the E domain and GuK domain (MPP4Edom-end); the C terminus of MPP4 (MPP4365-637); the SH3 and HOOK domains of human MPP4 (MPP4SH3+HK); and the C terminus of MPP4 from the SH3 domain (MPP4SH3-end). Gene-specific primers that were used to make these constructs are listed in Table 1 . The attB1 and attB2 linkers were attached to the 5' end of, respectively, the sense and antisense primers.
|
Yeast Two-Hybrid
A GAL4-based yeast two-hybrid system (Hybrizap; Stratagene), with yeast strain PJ69-4
, was used to identify proteins that interact with CRB1intra and MPP5. The pBD-hMPP5CC, pBD-hMPP5GuK, and pBD-hMPP5SH3+HK constructs were used as baits on an oligo-dT primed human retina cDNA library, representing 2.1 x 106 primary cDNA clones. In total, 8.2 x 105, 1 x 106, and 6.8 x 105 clones were plated, respectively. The human and bovine pBD-CRB1intra constructs were used to screen respectively a pretransformed oligo-dT primed human and a randomly primed bovine retina cDNA library by yeast cell-to-cell mating, resulting in screening 1.0 x 107 and 1.4 x 107 clones.33 In subsequent yeast two-hybrid experiments, different domains of MPP5 and -4 were tested for interaction by cotransformation into the YRG-2 yeast strain. Interactions were quantified in a liquid ortho-nitrophenyl-ß-D-galactopyranoside (ONPG) assay for ß-galactosidase activity.34
Antibodies
Bacterially expressed full-length MPP5 protein was used for immunization of chicken. The yolk was processed with an IgY Purification Kit (Eggcellent Chicken; Pierce Biotechnology, Rockford, IL) according to the manufacturer’s protocol, and IgY antibodies were consequently affinity purified on a protein-coupled NHS-activated HP column (Hi-Trap; Amersham Biosciences, Roosendahl, The Netherlands).
Production of AK2, AK5, AK7, AK4, and AK8 antibodies have been described.24 Anti-c-myc monoclonal mouse antibodies (clone 9E10) were purchased from Roche; anti-rat MUPP1 (clone 43) and anti-ß-catenin (clone 14) mouse monoclonal antibodies from BD Biosciences (Leiden, The Netherlands); anti-6x His antibody from Santa Cruz Biotechnology (Heerhugowaard, The Netherlands); anti-FLAG monoclonal mouse antibody (clone M2), monoclonal anti-chicken IgG (clone CG-106), and rat monoclonal anti-uvomorulin (clone Decma-1) from Sigma-Aldrich (Amsterdam, The Netherlands). Secondary antibodies conjugated to Alexa 488, Cy3, and Cy5 were obtained from Molecular Probes (Leiden, The Netherlands) and Jackson ImmunoResearch Laboratories (West Grove, PA). Secondary antibodies conjugated to horseradish peroxidase, were purchased from Sigma-Aldrich and Zymed (Uden, The Netherlands).
Cell Culture
Human embryonic kidney (293HEK) and Madin-Darby canine kidney type II (MDCKII) cells were grown in DMEM (Invitrogen) containing 1% penicillin/streptomycin and 10% fetal bovine serum. Stably transfected MDCKII/CRB1 clones were generated by transduction of MDCKII cells with pBABE-CMV-Puro-CRB1 retroviruses, and subsequently selected with 2 µg/mL puromycin.
GST Pull-Down, Coimmunoprecipitation, and Western Blot Analysis
Arabinose inducible BL21-AI cells were transformed with GST-hCRB1intra/pDest15, or His-MBP-hMPP5PDZ/pDest566 and IPTG-inducible BL21-DE3 cells with GST-hMPP5SH3+HK/pDest15 or His-MBP-hMPP4HK-end/pDest566. BL21-DE3 cell lysates were prepared according to a 1.5% sarkosyl protocol with DNase added before centrifugation.35 BL21-AI cell lysates were prepared (B-PER; Pierce Biotechnology, Etten-Leur, The Netherlands), with a protease inhibitor cocktail (Roche, Almere, The Netherlands) plus 1 µg/mL pepstatin A and 5 mM dithiothreitol (DTT). For GST pull downs, equal amounts of blocked (1.5 mg/mL BSA) glutathione Sepharose beads (4B; Amersham Pharmacia, Uppsala, Sweden) with glutathione S-transferase (GST), beads with GST fusion proteins, or beads alone were incubated with 0.5 mL of lysates containing His-MBP-fusion proteins for 2 hours at 4°C. After several washes with lysis buffer and TBS containing 1% Triton X-100 and 2 mM DTT, the beads were boiled, and proteins were resolved on SDS-polyacrylamide gels.
For coimmunoprecipitation experiments, 293HEK cells were transfected with pBabe-CMV-Puro/Hygro-CRB1/MPP5/MPP4, with a commercial reagent (Fugene 6; Roche) or calcium phosphate. After 48 hours, cells were lysed in 50 mM HEPES (pH 7.4), 150 mM sodium chloride, 10% glycerol, 0.5% Triton X-100, 1.5 mM magnesium chloride, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride (PMSF), a protease inhibitor cocktail (Roche), and 10 µg/mL aprotinin (Sigma-Aldrich). Either protein LA-agarose (Sigma-Aldrich) was used to bind the primary antibodies after incubation of precleared supernatants with 10 to 15 µg antibody for 4 to 16 hours at 4°C, or antibodies were precoupled to protein G beads (Dynabeads; Dynal Biotech ASA, Oslo, Norway) before incubation of supernatants for 2 hours at 4°C. For immunoprecipitation with anti-MPP5 antibody SN47, mouse monoclonal anti-chicken IgG was precoupled to the protein G beads (15 µg/reaction), followed by a second round of coupling of chicken anti-MPP5 antibody SN47 (10 µg/reaction) and incubation with cell lysates for 2 hours at 4°C. The beads were washed three times in 10% glycerol/PBS or lysis buffer, respectively and boiled in sample buffer with ß-mercaptoethanol, and the immunocomplexes were resolved by SDS-PAGE. For Western blot analyses, proteins were electrophoretically transferred onto nitrocellulose membranes, which were then blocked, incubated with primary and secondary antibodies (conjugated to horseradish peroxidase) in 0.3% to 5% milk powder/TBS, and washed in TBS. The bands were visualized with a chemiluminescence reagent (ECL; Amersham Biosciences).
N-Glycosylation Experiments
After they reached 70% confluence, stable clones of MDCKII cells expressing hCRB1 were cultured for 1 to 5 days in DMEM supplemented with 1% penicillin/streptomycin, 10% fetal bovine serum, and 5 µg/mL tunicamycin dissolved in dimethylsulfoxide (DMSO) or only DMSO for the control cells.
Expression Profiles
Total RNA was isolated from different human tissues and from an ARPE-19 cell-line (Dunn KC, et al. IOVS 1995;36:ARVO Abstract 766). For the semiquantitative RT-PCR, 3.1 µg RNA was reverse transcribed using random hexanucleotides.36 A touchdown PCR was performed for 28 and 33 cycles on 62 ng cDNA for CRB1, MPP5, MPP4, and the housekeeping gene GAPDH, which served as a standard. The following primer pairs were used: 5'-ACCAATGTATTCAACAGGGACC-3'(sense) and 5'-TCGTTTCCGTTGTAGTGTCTCC-3' (antisense) for CRB1, 5'-GTATGGAACTAGCATAGATTCTG-3' (sense) and 5'-CAAGATCGGAATTCACAATTGCC-3' (antisense) for MPP5, 5'-CACCTGTATGGCACTAGTGTGG-3' (sense) and 5'-CATACACCTCATATTCGATGGC-3' (antisense) for MPP4, and 5'-ACCACAGTCCATGCCATCAC-3' (sense) and 5'-TCCACCACCCTGTTGCTGTA-3' (antisense) for GAPDH.
Immunohistochemistry
Eight human postmortem retinas, with enucleation times of 8 to 24 hours, were obtained from the cornea bank in Amsterdam and treated in accordance with the guidelines of the Declaration of Helsinki for the use of human tissue in research.
Frozen human retina sections, 10 µm thick, were treated essentially as described previously24 using PBS buffer and 1% BSA. Sections were imaged on a confocal laser-scanning microscope (model 501; Carl Zeiss Meditec, Jena, Germany).
Immunoelectron Microscopy
Immunoelectron microscopy on mouse retina sections was performed as described previously.37 Ultrathin sections were examined and photographed (model 201 electron microscope; Phillips, Eindhoven, The Netherlands).
Molecular Modeling of MPP4 and -5
The amino acid sequences of MPP4 (Swiss Prot accession number Q96JB8; http://www.expasy.org; provided in the public domain by Swiss Institute of Bioinformatics, Geneva, Switzerland) and MPP5 (accession number Q8N3R9) were submitted to the 3D-PSSM fold recognition server38 to search the protein data base (PDB) for homology modeling templates. The best hit in both cases (E-value 0.07) was PDB file 1KJW, the SH3-GuK module of Postsynaptic Density Protein 95 (PSD-95), solved at 1.8-Å resolution.39 With 
40% sequence identity in the aligned regions, the modeling template PSD-95 can be expected to be very similar to the target structures,40 except for structurally divergent loop regions. Consequently, homology models for MPP4 and -5 were built with What If,41 using a backbone-dependent rotamer library42 (see Fig. 3B , left side). Flexible HOOK residues (blue in Fig. 3B ) were deleted, and the independent SH3 and GuK domains of MPP4 and -5 were arranged in all four possible permutations. Finally, the side-chain rotamers at the domain interface were optimized with YASARA (Yasara Biosciences, http://www.yasara.org/index.html) by minimizing the NOVA force field energy.43 The parameters of the NOVA energy function have been optimized based on known high-resolution x-ray structures, so that the function has stable minima as close as possible to these structures. The relative domain-binding energies of the four models were then calculated as described previously.44 Coordinate files of the models are available from the authors on request.
|
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Identification of a Novel Interactor with MPP5
Although the MAGUK proteins in general and the MPP family of proteins in particular contain different conserved putative protein–protein interaction domains that allow the buildup of a scaffold, only a few of these domains in MPP5 have known ligands (Fig. 1) . We used the conserved epitopes of human MPP5, for which no partners have yet been identified, as baits in yeast two-hybrid screens of a human oligo-dT-primed retina cDNA library. We did not identify interactors for the coiled-coil domain (amino acid [aa] 1-122) nor for the GuK domain (aa 470-675; data not shown). However, a bait construct containing the SH3-HOOK region of human MPP5 was found to interact with MPP4. Four different clones containing the C terminus of MPP4 were identified, with the GuK domain flanked by strands E and F (Fig. 1) , starting at aa 319 (#228), aa 365 (#73), aa 368 (#221) and aa 390 (#78) (Fig. 2A) . The binding affinities of these different clones for MPP5 were measured semiquantitatively in a liquid ortho-nitrophenyl-ß-D-galactopyranoside (ONPG) assay, revealing that the peptide stretch of MPP4 containing the regions E, GuK, and F is essential for the interaction. However, when more amino acids were present at the N terminus, the binding affinity increased.
|
|
Furthermore, to test for a physical interaction between MPP4 and -5 in the presence or absence of CRB1, we used 293HEK cells overexpressing MPP4, and/or MPP5, and/or CRB1 in immunoprecipitation experiments (Figs. 2C 2D) . Anti-FLAG antibody coimmunoprecipitated MPP5 from cells overproducing 3xFLAG-MPP4 and MPP5, as well as from cells overproducing 3xFLAG-MPP4, MPP5, and CRB1 (Fig. 2C) . The 80-kDa recombinant and the endogenously expressed 70-kDa MPP5 coimmunoprecipitated with MPP4 (Fig. 2C) . In a reciprocal experiment, anti-MPP5 antibody SN47 coimmunoprecipitated MPP4 from cells overproducing 3xFLAG-MPP4 and MPP5, as well as from cells overproducing 3xFLAG-MPP4, MPP5, and CRB1 (Fig. 2D) . Coimmunoprecipitation of MPP4 or CRB1 with immunoprecipitated endogenous 70- and 80-kDa MPP5 was below detection levels (Fig. 2D , and data not shown). Recombinant MPP5 was efficiently immunoprecipitated by SN47 from cell lysates overproducing MPP5, whereas endogenous MPP5 was not (data not shown). By Western blot analysis we could not detect endogenous expression of CRB1 (data not shown). These results show that MPP4 interacts with MPP5 in the absence as well as presence of CRB1. To test for the presence of a protein complex containing MPP4 and CRB1, we used 293HEK cells overexpressing 3xFLAG-MPP4 and/or myc-CRB1. Anti-MPP4 antibody (AK4) immunoprecipitated MPP4 from 3xFLAG-MPP4-overproducing cell lines (Fig. 2E) , and coimmunoprecipitated CRB1 from cells overproducing 3xFLAG-MPP4 and myc-CRB1 (Fig. 2F) , though at a much lower level.
In a reciprocal experiment, anti-myc antibody immunoprecipitated CRB1 from myc-CRB1 overproducing cell lines (data not shown), and coimmunoprecipitated MPP4 from cells overproducing 3xFLAG-MPP4 and myc-CRB1 at similar low levels (Fig. 2G) . This result, as well as the anti-Flag immunoprecipitation of MPP4, indicates the presence of MPP4 and CRB1 in the same complex, though probably through endogenously expressed MPP5. The calculated size of CRB1 was 154 kDa, whereas CRB1 polyclonal antibodies recognized a protein >220 kDa on Western blot from 293HEK/CRB1 and MDCKII/CRB1 cells. The PROSITE program predicted 23 putative N-glycosylation sites (Swiss Institute of Bioinformatics). Therefore, we tested to see whether CRB1 is glycosylated in MDCKII/CRB1 cells. Incubation with tunicamycin shifted the molecular weight toward the expected 154 kDa (Fig. 2H) , indicating extensive N-glycosylation of CRB1. The endogenous Crb1 from mouse retina24 appears to be more than 220 kDa, whereas the calculated molecular weight is 153 kDa. This suggests that N-glycosylation also occurs in vivo in mouse retina.
Computer-Based Molecular Dynamics and In Vivo Interaction of Homo- and Heterodimers of MPP4 and -5
The interacting domains of MPP4 and -5 were analyzed by using a panel of deletion variants of both proteins in the yeast two-hybrid system (Fig. 3A , above dotted line). A fragment of MPP5 containing the SH3-HOOK region, flanked by strands A and D, specifically interacts with a fragment of MPP4 containing the GuK domain, flanked by strands E and F. When either of these strands was absent, the interaction was fully disrupted (data not shown). Furthermore, no interaction was found between the SH3-HOOK region of MPP5 and the full-size SH3-GuK module of MPP4, or between the full-length MPP5 and -4 proteins (data not shown). These results are analogous to the intra- and intermolecular interactions reported for the MAGUK protein PSD-95. In PSD-95, the interaction between the SH3 and GuK domains can either occur within a single peptide chain or between separate peptide chains. It has been proposed before that the interactions observed in PSD-95 are a conserved feature among MAGUK proteins.45 A similar mechanism could thus be expected for MPP4/MPP5 based on our experimental results and the high percentage of sequence identity to PSD-95, especially at the domain interface. To analyze this possibility, we built 3D homology models of the SH3-GuK domains of MPP4 and -5 (Fig. 3B , left), based on the crystal structures of PSD-95.39 45 All four permutations of domain interactions were analyzed: MPP4SH3-MPP4GuK, MPP4SH3-MPP5GuK, MPP5SH3-MPP4GuK, and MPP5SH3-MPP5GuK. In all cases, the domain interfaces were stabilized by salt bridges and hydrophobic interactions, most prominently by a conserved tyrosine (Y413 in MPP4 and Y466 in MPP5) that forms the core of the SH3 domain, but belongs to the E-strand just before the GuK domain (Fig. 3B , below dotted line). The domain binding energies of the four models are listed in Table 2 .
|
To validate the model further, we analyzed the binding affinities of the SH3 and GuK domains in MPP4 and -5 (Fig. 3A , right). We were able to confirm in the sensitive liquid ONPG assay that the MPP4 domains interact more strongly than the ones in MPP5.
Expression of CRB1, MPP5, and MPP4
Expression analysis by semiquantitative RT-PCR on a panel of RNAs from several human tissues showed that CRB1 is predominantly expressed in the retina. An increase in the number of PCR cycles identified a lower level of expression in brain, testis, and fetal eye (Fig. 4) . MPP5 is more ubiquitously expressed and is also present in the retina. MPP4 is highly expressed in the retina. An increase in the number of PCR cycles identified a lower level of expression in brain, testis, ARPE cell line and fetal eye.
|
Immunolocalizations of CRB1, MPP4, and MPP5 in Human Retina
Immunohistochemistry and confocal laser scanning microscopy were used to determine the subcellular protein localization of CRB1, MPP4, and MPP5. Anti-CRB1 antibodies AK2 and -5 detected CRB1 at the outer limiting membrane (OLM) of human retina. Using monoclonal antibodies against human ß-catenin as a marker for the adherens junction, localization of CRB1 apical to the adherens junction was detected (Figs. 5A 5B 5C 5D) . Anti-MPP5 antibodies (SN47) detected the protein at the OLM, also apical to the AJ (Figs. 5E 5F 5G 5H) , where it colocalized with CRB1 (Figs. 5I 5J 5K 5L) . MPP4 was detected at the OLM and in the outer plexiform layer (OPL) of human retina with AK4 (and AK8) antibodies. At the OLM, MPP4 was detected apical to the adherens junction. However, the intensity of the signal at the OLM was much lower than the MPP4 staining intensity in the OPL (Figs. 5M 5N 5O 5P 5Q) .
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Mechanism of Interaction and Implications for Regulation
The interaction between MPP5 and -4 involves the C-terminal end of MPP4. Alternative interaction through L27 domain dimerization was excluded. In a yeast two-hybrid experiment, no L27 domain binding was found between MPP4 and -5, whereas the L27 domains of PATJ and MPP5 did show interaction (data not shown).
Detailed analysis revealed that the E and F strands flanking the GuK domain of MPP4 are essential for binding. This is in full agreement with the folding according to the crystal structure of PSD-95.39 45 In PSD-95, the SH3-HOOK region interacts both in cis and trans with the E-GuK-F region in vitro. The intramolecular interaction initially prohibits intermolecular interactions by preventing a mechanism called 3D domain swapping. Domain swapping allows proteins to assemble dimers or higher order oligomers by exchanging complementary substructures.48 49 50 The need of a cofactor or modulator has been postulated to switch the preference from intra- toward intermolecular interactions in vivo, thus enabling specific regulation of MAGUK multimerization at the cytoplasmic membrane.39 45
The 3D homology modeling of MPP5 and -4 pointed to a PSD-95-like interaction mechanism that also matched the results of our yeast two-hybrid binding assays. GST pull-down analysis and immunoprecipitation experiments both confirmed the interactions biochemically and provided in vivo evidence for the proposed regulatory mechanism of heterodimerization. The full-length proteins that were the targets for immunoprecipitation did coprecipitate the full complex of CRB1, MPP4, and MPP5 from HEK293 cells, indicating that in these cells the regulatory factor is present (Figs. 2C 2D) . Good candidates for regulation of the dimerization of MPP5 are members of the 4.1 protein family, as the variable hinge or HOOK region of MPP5 contains a conserved 4.1 binding motif. MPP4 does not contain this particular conserved motif in the same region, but does have a predicted -helical stretch. We propose that regulation of dimerization of these MAGUK proteins is one of the factors that governs a dynamic variation of proteins that are present at this polarity-associated protein scaffold.
SH3 domains of tyrosine kinases are usually involved in protein–protein interactions by binding to proline-rich sequences,51 52 and have been described to couple substrates to enzymes, thereby regulating enzymatic activities.53 However, based on the 3D homology models presented herein and the interactions identified in this study, the SH3 module of MPP4 and -5 seems to be functionally different from the conventional ones. The SH3 domain of MPP5 interacts with the GuK domain of MPP4 as well as with its own GuK domain.
Distribution of Protein Expression in the Human Retina
Whereas MPP5 is expressed ubiquitously,7 both CRB1 and MPP4 genes are expressed more selectively in eye and brain.54 MPP4 RNA is also present in liver, spleen, heart,46 47 54 55 and testis, but at much lower levels. Previous experiments showed that Crb1 and Mpp4 RNA is expressed in the ONL and inner photoreceptor segments of the retina.46 54 Crb1 RNA was also detected at low levels in the inner nuclear layer.
The protein complex CRB1-MPP5-MPP4 localizes subapically to the adherens junction at the OLM of the retina. The localization of Mpp4/MPP4, Mpp5/ MPP5, and Crb1/CRB1 appears to be conserved between mice and humans.24 Our results on MPP4 localization partially overlap with the positioning described for MPP4 in mouse retina.47 Differences in genetic background or detection level may explain why Mpp4 was not detected in cones or OLM in previous studies47 and may explain the localization of MPP4 in the connecting cilia of bovine and porcine.56
We observed that Mpp4 is detected at intracellular vesicles and at the plasma membrane. Mpp4 is also located at the presynaptic membrane and proximal vesicles. The localization of MPP4 at more than one functionally different cell structure suggests participation in different protein complexes. Some MAGUK proteins target and anchor glutamate receptors to the synaptic terminals.57 It has been proposed that the complex involving the MAGUK protein CASK acts as a nucleation site for the assembly of proteins involved in synaptic vesicle exocytosis and synaptic junctions.58 59 It is tempting to speculate on possible functions of MPP4 in vesicle targeting or fusion complexes at the photoreceptor synapses and the region apical to the adherens junction.
Implications for Inherited Retinal Degenerations
Altogether, these facts provide strong evidence for the involvement of CRB1, MPP5, and MPP4 in a common pathway that determines the polarity of photoreceptors in the retina. Based on the recruitment of both MPP4 and -5 to the CRB1 protein scaffold, the disruption of retinal lamination observed with loss of mouse Crb124 and the zebra fish MPP5 homologue Nagie oko14 and the high expression of MPP4 in the retina, we propose that MPP5 and MPP4 are functional candidate genes for inherited retinal degenerations.
The MPP4 gene has been screened for mutations in RP with 300 patients, and it was mapped in a locus for autosomal recessive retinitis pigmentosa (RP26 locus) at 2q31-33, but no mutations were identified. Recently, mutations in the neighboring CERKL gene were found to cause RP26.60 However, based on the more severe phenotype that is often observed in patients with mutations in CRB1, MPP4 remains a candidate gene for similar eye disorders. Mutational changes in members of the complex that are closely linked could lead to a similar disrupting effect of the protein complex in the retina. Mutation analysis in selected patient panels could reveal the involvement of either MPP4 or MPP5 in inherited retina disorders.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
Supported in part by Grant QLG3-CT-2002-01266 from the European Commission (EC), Grant 912-02-018 from the Netherlands Organization for Scientific Research (JW, FPMC), and Grant N-CB-0600-0003 from the Foundation Fighting Blindness (FPMC, AIdH).
Submitted for publication December 3, 2004; revised January 25, 2005; accepted January 26, 2005.
Disclosure: A. Kantardzhieva, None; I. Gosens, None; S. Alexeeva, None; I.M. Punte, None; I. Versteeg, None; E. Krieger, None; C.A. Neefjes-Mol, None; A.I. den Hollander, None; S.J.F. Letteboer, None; J. Klooster, None; F.P.M. Cremers, None; R. Roepman, None; J. Wijnholds, None
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked "advertisement" in accordance with 18 U.S.C. 
1734 solely to indicate this fact.
Corresponding author: Jan Wijnholds, The Netherlands Ophthalmic Research Institute, Meibergdreef 47, 1105 BA, Amsterdam, The Netherlands; j.wijnholds{at}ioi.knaw.nl.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
). As a negative control, the pBD-GAL4 domain, fused to CRB1intra, and the pAD-GAL4 domain, fused to MPP4PDZ, was used. (B) Homology modeling of MPP4 and -5. The initial models are shown on the left, covering the SH3 (yellow, green) and GuK domains (red, light blue). By swapping the domains, one obtains a heterodimer, half of which is shown in the middle: the SH3 domain of MPP5 bound to the GuK domain of MPP4. Energy calculations predicted a high binding energy for this interaction 
). (B) In rod photoreceptors, the lateral plasma membranes of the spherules were strongly stained (