Cellular Retinaldehyde-Binding Protein Interacts with ERM-Binding Phosphoprotein 50 in Retinal Pigment Epithelium

Abstract

purpose. To characterize mechanisms of apical localization of visual cycle components in retinal pigment epithelium (RPE) by the identification of cellular retinaldehyde-binding protein (CRALBP) interaction partners.

methods. An overlay assay was used to detect interactions of CRALBP with components of RPE microsomes. Interacting proteins were identified with two-dimensional (2D)-PAGE and liquid chromatography tandem mass spectrometry (LC MS/MS). Protein interactions were characterized by affinity chromatography, peptide competition, and expression of protein domains. Protein colocalization in mouse retina was examined using double-label immunocytochemistry and confocal microscopy.

results. CRALBP bound to a 54-kDa protein in RPE microsomes, which was identified as ERM (ezrin, radixin, moesin)-binding phosphoprotein 50 (EBP50), a PDZ domain protein, also known as sodium/hydrogen exchanger regulatory factory type 1 (NHERF-1). EBP50 and ezrin in solubilized microsomes bound to CRALBP-agarose but not to a control agarose column. CRALBP bound to both recombinant PDZ domains of EBP50 but not to the C-terminal ezrin-binding domain. In outer retina, EBP50 and ezrin were localized to RPE and Müller apical processes. CRALBP was distributed throughout both RPE and Müller cells, including their apical processes.

conclusions. ERM proteins are multivalent linkers that connect plasma membrane proteins with the cortical actin cytoskeleton. EBP50 interacts with ERM family members through a C-terminal domain and binds targets such as CRALBP through its PDZ domains, thus contributing to an apical localization of target proteins. Our results provide a structural basis for apical localization of a retinoid-processing complex in RPE cells and offer insight into the cell biology of retinoid processing and trafficking in RPE. (247 words)

The chromophore of rhodopsin and of cone visual pigments, 11-cis-retinal, is regenerated after its photoisomerization by a complex process called the visual cycle. ^{1 2} Photoisomerization converts 11-cis-retinal to all-trans-retinal, which is reduced to all-trans-retinol by reduced nicotinamide adenine dinucleotide phosphate (NADPH) and photoreceptor retinol dehydrogenase within the rod photoreceptor outer segment. all-trans-Retinol leaves the rod photoreceptor cells and diffuses to the adjacent retinal pigment epithelium (RPE), where it is sequentially esterified by lecithin-retinol acyltransferase (LRAT), converted to 11-cis-retinol by an isomerohydrolase, and oxidized to 11-cis-retinal by NAD(P) and one or more short-chain dehydrogenase-reductases (RDH4/5, RDH11). 11-cis-Retinal then diffuses back into the rod photoreceptor cell, where it regenerates rhodopsin and completes the visual cycle (for reviews of the visual cycle see Refs. ^{3 4 5 6 7} ). A simplified version of this cycle is shown in Figure 1 . Considerable progress has been made in understanding the biochemical basis and enzymology of the visual cycle; however, many aspects of the process remain poorly understood. In particular, the interaction of essential nonmembrane components with membrane-associated enzymes is totally uncharacterized. Nothing is known about mechanisms of retinoid trafficking, and no mechanisms have been identified for the spatial–compartmental organization of cycle components in RPE. Although the cone visual cycle is even less well understood, recent studies of 11-cis-retinal formation in cone-dominated retinas suggest that Müller cells, the major glial cells of the retina, are involved in the regeneration of cone visual pigments. ⁸  

The retina contains several retinoid-binding proteins that have been suggested to play various roles in the visual cycle. Cellular retinaldehyde-binding protein (CRALBP) in particular appears to be a functional component of the cycle. CRALBP has a high affinity for 11-cis-retinoids, ⁶ whose only known function is associated with vision. In addition, the protein is found in RPE and Müller cells, ⁹ both of which project apical process into the matrix surrounding photoreceptor cell inner and outer segments. ^{10 11} Regeneration of rod visual pigment is impaired in mice lacking a functional gene for CRALBP (Rlbp1), and resensitization of cone visual pigment–mediated responses is delayed. ¹² In humans, mutations in RLBP1 cause autosomal recessive retinitis pigmentosa, stationary night blindness, or other related conditions, ¹³ one characteristic of which is delayed rod and cone dark adaptation. ¹⁴ Based on in vivo and in vitro studies from several laboratories, CRALBP is likely to be the major acceptor of 11-cis-retinol from an isomerohydrolase in the visual cycle and to facilitate oxidation of 11-cis-retinol to 11-cis-retinal by 11-cis-retinol dehydrogenase (Fig. 1) . ^{12 15 16 17 18} Molecular mechanisms involving its interaction with other visual cycle components are uncharacterized, and the mechanism of release of 11-cis-retinal from its high-affinity ligand-binding pocket in CRALBP ¹⁹ to the RPE plasma membrane or other proteins is not known. 

Metabolic processes are often organized into complexes of several proteins for efficiency of enzymatic processing and to route intermediates into dedicated metabolic pathways. Work in our laboratories, focused on identifying interaction partners of CRALBP, has provided evidence for a retinoid-processing complex in RPE. ²⁰ Here, we demonstrate with two in vitro assays that CRALBP interacts with ERM (ezrin, radixin, moesin)-binding phosphoprotein50 (EBP50), a well-characterized PDZ domain protein. ²¹ EBP50 was independently discovered as a regulatory factor for sodium-hydrogen exchanger type 3 (NHE3) and is also known as sodium-hydrogen exchanger regulatory factor type 1 (NHERF-1). ²² We will refer to the protein as EBP50 in this communication. The PDZ domain protein is responsible for localization of proteins to the apical plasma membrane of polarized epithelia, ^{23 24} but has not been previously associated with the visual process. Furthermore, we confirm that EBP50, CRALBP, and ezrin, an interaction partner of EBP50, are present in apical processes of RPE cells and in apical microvilli of Müller cells, the major glial cell of the retina. Our findings provide a structural basis for localization of a retinoid-processing complex to the apical membranes of RPE and Müller cells and offer mechanistic insight into the cell biology of retinoid processing in RPE. 

Materials and Methods

Animals

All procedures with live animals were approved by the University of Washington Animal Care Committee and were in accord with the recommendations of the American Veterinary Medical Association Panel on Euthanasia and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Antibodies

The following antibodies were used: Anti-CRALBP (mAb B2), mouse monoclonal antibody to human recombinant CRALBP (rCRALBP); ²⁵ anti-CRALBP (UW55), rabbit polyclonal antibody to human rCRALBP; anti-interphotoreceptor retinoid-binding protein (IRBP) (mAb F7), mouse monoclonal antibody to bovine IRBP (Huang J, and Saari J, unpublished data, 1995); anti-EBP50, rabbit polyclonal antibody to human EBP50 C terminus ²¹ ; anti-ezrin, rabbit polyclonal antibody ²⁶ ; anti-RDH5 (UW41), rabbit polyclonal antibody made with a peptide epitope from RDH5 (SKYGVEAFSDSLRREL) (Huang J, Palczewski K, Saari J, unpublished data, 1995). 

Preparation of RPE Microsomes

Bovine eyes were purchased from Schenk Packing Company (Stanwood, WA) and dissected to yield RPE cell suspensions from which microsomes were prepared as previously described. ^{27 28}  

Purification of RPE Microsomes

For identification of EBP50, microsomes were washed with 0.6 M NaCl, 20 mM MOPS (3-(N-morpholino)propanesulfonic acid [pH 7]), 0.1 mM dithiothreitol (DTT), and protease inhibitors (HALT, 1:100 dilution of stock; Pierce Biotechnology, Rockford, IL) by homogenizing the pellet in a glass–glass homogenizer and centrifuging at 50,000g for 10 minutes. The resultant pellet was dissolved in 2 mM dodecylmaltoside (DM), 0.5% polyoxyethylene¹⁰ dodecylether X-080 (Genapol; Clariant Functional Chemicals, Mount Holly, NC), 0.1% Tween 80, 20 mM MOPS (pH 7), and centrifuged at 100,000g for 10 minutes. The supernatant was applied to a Dspin (Maxi/H) column (Vivascience AG, Hannover, Germany) and centrifuged at 2000g. The column was then washed at least 3 times with 0.5% polyoxyethylene¹⁰ dodecylether X-080 (Genapol), 0.1% Tween 80 in MOPS-DTT buffer. A purified microsomal membrane fraction was eluted from the column with two applications of the initial buffer containing 300 mM NaCl. For some procedures, the purified fraction was desalted on spin mini columns (catalog no. 89849; Pierce Biotechnology) or dialyzed. 

Preparation of Bovine CRALBP, Recombinant his-Tagged Human CRALBP, and IRBP

Bovine CRALBP complexed with 11-cis-retinal was purified from bovine retinas as described. ²⁸ rCRALBP was prepared and purified as described. ²⁹ The apo-protein was either used as such or reconstituted with 11-cis-retinal, by using methods described previously. ^{28 29} IRBP was purified from bovine retinas as described. ^{28 30} All procedures involving 11-cis-retinal/CRALBP were performed in dim red light. 

Preparation of Recombinant EBP50 and EBP50 Domains

Plasmid constructs were transformed into Escherichia coli Bl21, and protein expression was amplified in liquid cultures (Luria-Bertani [LB] broth plus ampicillin, 100 μg/mL) followed by induction with isopropylthio-β-d-galactoside. Cells were harvested by centrifugation at 10,000g for 10 minutes. Bacterial lysates were prepared by suspending the cells in 10 mM phosphate buffer (pH 7.2), containing 100 mM NaCl and protease inhibitors (HALT; Pierce Biotechnology), followed by sonication. Lysates were centrifuged at 60,000g for 10 minutes. Domains of EBP50 were expressed as glutathione S-transferase (GST) fusions and purified with GST columns (GST-Trap; Amersham Biosciences, Piscataway, NJ), according to the manufacturer’s instructions. Proteins were eluted with 10 mM glutathione and 50 mM Tris (pH 8) and equilibrated with an appropriate buffer by dialysis or gel filtration. 

Gel Overlay Assays

Washed RPE microsomes were dissolved in SDS sample application buffer, subjected to SDS-PAGE, and transferred to an Immobilon-P membrane (Millipore Corp., Bedford, MA) using a specialized device (Mighty Small SE250; Amersham Biosciences). After the membrane was blocked with milk proteins (3% wt/vol) for 1 hour at room temperature, the blot was incubated for 2 hours at 5°C in the dark with purified bovine 11-cis-retinal/CRALBP (5–10 μg/mL) in 10 mM Tris-HCl (pH 8), 150 mM NaCl, and 0.05% Tween-20. The membrane was washed and incubated with anti-CRALBP (either UW55 or mAb B2), washed, and then incubated with alkaline phosphatase-conjugated goat anti-rabbit IgG or rabbit anti-mouse IgG. Incubation with a chromogenic substrate for alkaline phosphatase (NBT, BCIP; Promega, Madison, WI) completed the assay. 

Peptide Competition Experiments

The synthetic peptides AEVENTAF, AEVENTAL, EVENAAG, and EVENAAD were obtained from GenMed Synthesis, Inc. (San Francisco, CA). Solutions of the peptides in 50 mM MOPS (pH 7) were diluted into 10 mM Tris (pH 8) 150 mM NaCl, and 0.05% Tween 20, to a final concentration of 150 μM. Peptides were incubated with the PVDF membrane in the CRALBP overlay assay for 5 minutes before addition of CRALBP (see description of overlay assay). 

Two-Dimensional Gel Electrophoresis of RPE Microsomes

Two-dimensional (2D) gel electrophoresis was performed with a commercial system (Protean; Bio-Rad, Hercules, CA), according to the manufacturer’s instructions. A sample of purified RPE microsomes (100–200 μg protein) was dialyzed against 10 mM MOPS (pH 7) and 2 mM DM and mixed with 130 μL of isoelectric focusing (IEF) sample buffer (8 M urea, 2% CHAPS, 50 mM DTT, 0.2% ampholytes [BioLyte 4/7; Bio-Rad], and a trace of bromophenol blue) and left for 10 minutes at room temperature. The sample was passively absorbed to a gel strip (pH 4–7; DryStrip; Amersham Biosciences) overnight at room temperature, and IEF was performed for 8 to 12 hours according to the manufacturer’s instructions. After IEF, proteins were resolved by SDS-PAGE on 10% gels (Ready Gels, catalog no. 161-1137; Bio-Rad). 2D gels were electroblotted to polyvinylidene difluoride (PVDF) and stained (Gelcode Blue Stain; Pierce Biotechnology) or analyzed with the CRALBP overlay assay. 

Protein Identification by Mass Spectrometry

Identification of EBP50 by liquid chromatography electrospray tandem mass spectrometry LC MS/MS used methods described in detail elsewhere. ^{31 32} Briefly, stained gel spots or bands were excised, the stain washed away, the proteins digested in gel with trypsin, and the peptides extracted for mass spectrometric analysis. LC MS/MS was performed with a commercial system (CapLC; Micromass, Beverly, MA) and a quadrupole time-of-flight mass spectrometer (QTOF2; Micromass). Protein identifications from MS/MS data were performed with the system software (ProteinLynx Global Server, MassLynx, ver. 3.5; Micromass), and the Swiss-Prot and NCBI protein sequence databases (http://expasy.hcuge.ch/ provided in the public domain by the Swiss Institute of Bioinformatics, Geneva, Switzerland; and www.ncbi.nlm.nih.gov/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD). 

Preparation of CRALBP Agarose

Agarose beads (1 mL; Aminolink; Pierce Biotechnology) were washed with 30 mL of 100 mM sodium phosphate (pH 7.2) 150 mM NaCl (coupling buffer) and gently agitated with 2 mL of 11-cis-retinal/CRALBP (2.6 mg/mL). Sodium cyanoborohydride was added to give a final concentration of 50 mM. After 1 hour at room temperature, the incubation was continued with gentle mixing at 5°C for 15 hours and then centrifuged (500g, 2 minutes). The beads were washed with 4 mL of coupling buffer. Residual reactive groups were blocked by incubation with 2 mL of 1 M Tris-HCl (pH 7.4) and 50 mM Na-cyanoborohydride, for 60 minutes at room temperature with gentle agitation. Finally, the beads were washed three times with 5 mL 1 M NaCl and three times with 5 mL immunoprecipitation buffer (8 mM sodium phosphate, 2 mM potassium phosphate [pH 7], 140 mM NaCl, 10 mM KCl). IRBP-agarose was prepared according to the same protocol. 

Immunocytochemistry

The methods we used for tissue fixation, immunocytochemistry, and laser scanning confocal microscopy have been described. ^{12 25}  

Results

CRALBP Interaction with a 54-kDa Component in RPE Microsomes

We used an SDS-PAGE overlay assay to search for components in RPE microsomes that interacted with CRALBP. Bovine RPE microsomes, in the absence of an overlay with CRALBP (Fig. 2A , lane 1), contain CRALBP, which migrates at its normal position in SDS-PAGE (∼36 kDa). Proteolysis fragments are also usually present (Fig. 2 , bracket, lanes 1 and 2). After overlaying with CRALBP, we also detected CRALBP at an apparent mass of ∼54 kDa, indicating that a component in RPE microsomes bound the protein (Fig. 2A , lane 2, arrowhead). Both native and recombinant CRALBP bound the 54-kDa component in their apo- and holo(11-cis-retinal) forms. IRBP, another retinoid-binding protein, did not bind to the 54-kDa component after overlay with IRBP and probing with anti-IRBP (results not shown). This suggests that interaction of the 54-kDa protein with CRALBP is relatively specific. 

CRALBP, a water-soluble protein, was consistently detected in washed preparations of bovine RPE microsomes. The explanation for this is not clear; however, interactions with EBP50 and/or other visual cycle components are likely to be involved. ²⁰  

2D Gel Electrophoresis of RPE Microsomes

Attempts to identify the 54-kDa interaction partner by sequence analysis of the excised band from a 1D gel were not successful because of the presence of comigrating proteins such as tubulin and cytokeratins. However, we were able to separate the 54-kDa component from other proteins by 2D gel electrophoresis. Purified RPE microsomes were subjected to 2D gel electrophoresis, and the resultant blot was stained with Coomassie blue. A limited number of components were stained, including a prominent set of approximately five isoelectric variants with apparent molecular masses of ∼54 kDa (Fig. 3A , arrows). A duplicate gel was transferred to PVDF film and analyzed with the CRALBP overlay assay. Components corresponding to the five isoelectric variants plus others of slightly lower mass were evident as CRALBP interaction components (Fig. 3B) . Five Coomassie blue–stained components (arrows) were individually excised from the gel, digested in situ with trypsin, and identified by LC MS/MS sequence analysis. Sequences exactly matching those of a mouse homologue of rabbit NHERF-1, also known as EBP50, were identified in each of the excised spots. Overall, we identified six peptides comprising 60 residues or 18% of the total 358 amino acids. The identification suggests that the multiple 2D gel spots detected with the overlay assay represent charge variants of EBP50. Because EBP50 is a phosphoprotein, ²¹ it is possible the electrophoretic variants result from the various amounts of phosphate linked to the protein. 

The mass of EBP50 calculated from its amino acid sequence is 38.6 kDa, whereas the protein migrates in SDS-PAGE with an apparent mass closer to 50 kDa in other studies ²¹ or 54 kDa in this study. The reason for this anomalous migration has not been established, but is perhaps related to abnormal SDS binding due to the PDZ domains and/or the presence of phosphate. 

Binding of CRALBP

To confirm the identification of EBP50 as a CRALBP-interaction partner, we examined the ability of human recombinant (rEBP50) to bind CRALBP. rEBP50 was subjected to the CRALBP overlay assay and the results (Fig. 2B , lane 3) demonstrate clear interaction with CRALBP. Anti-EBP50 labeled a component in bovine RPE microsomes with a migration identical with that of the component labeled in the CRALBP-overlay assay (Fig. 2B , lanes 1 and 2). The apparent mass of bovine EBP50 (54 kDa) was larger than that of the human recombinant protein (50 kDa), as was noted earlier. ³³  

Binding of EBP50 to CRALBP-Agarose

The overlay assay requires that the interaction partner of CRALBP retain some of its structure in the presence of SDS or that its structure renatures after electrophoresis and transfer to PVDF. Although unlikely, interaction of CRALBP and EBP50 in the overlay assay may result artificially from this denaturation–renaturation process. To examine this possibility more thoroughly, we determined whether RPE microsomal EBP50, dissolved in a nonionic detergent, would interact with CRALBP attached to a solid support. Solubilized RPE microsomes were applied to CRALBP-agarose, which was then extensively washed with 0.3 M NaCl. The remaining bound proteins were eluted with 2 M NaCl, analyzed by SDS-PAGE, transferred to PVDF membranes, and probed with the CRALBP overlay assay or with anti-EBP50. The same procedure was performed with IRBP-agarose as a control for the experiment. The results, shown in Figure 4 , indicate that EBP50 was detectable in the eluate from CRALBP-agarose by both the CRALBP-overlay assay and anti-EBP50 antibody (Fig. 4A , lanes 2 and 4). The labeled components of lower molecular weight in Figure 4A , lane 2, have not been identified. Thus, interaction of CRALBP and EBP50 is not dependent on the procedures used in the overlay assay. The same assays failed to detect EBP50 in the eluate from IRBP-agarose (Fig. 4B , lane 3), indicating that the affinity of EBP50 for CRALBP is not related to the derivatized-agarose used in construction of the affinity absorbent. 

CRALBP-Agarose Interaction with Ezrin and RDH5

We used antibodies to another visual cycle enzyme and to ezrin to determine whether other putative components of the cycle bound to CRALBP. Solubilized RPE microsomes applied to CRALBP-agarose and pass-through and eluted fractions were analyzed by immunoblot. Staining of the blot with anti-RDH5 or anti-ezrin indicates that RDH5 or a similar, cross-reacting dehydrogenase binds to CRALBP-agarose (Fig. 5A , lane 2) but not to IRBP-agarose (Fig. 5A , lane 3) and that ezrin binds to CRALBP-agarose (Fig. 5B , lane 2). Binding of RDH5 to CRALBP is likely to be direct, because the retinoid-binding protein is known to influence the kinetics of the dehydrogenase reaction ^{6 18} and because RDH5 has been identified in anti-CRALBP immunoprecipitates. ²⁰ Ezrin is likely to bind indirectly to CRALBP through its interaction with EBP50. ²¹ Ezrin (93 kDa) did not bind CRALBP directly in the overlay assay (Fig. 2A) . 

Binding of CRALBP to PDZ-1 and -2 Domains of EBP50

EBP50 comprises two nonidentical PDZ domains and a C-terminal domain that binds ezrin or another ERM family member. To determine the sites of binding of CRALBP to EBP50, we expressed GST fused to full-length EBP50, PDZ1+PDZ2, PDZ1, PDZ2, and the EBP50 C terminus. These constructs have been described previously. ³⁴ All constructs containing either PDZ1 or PDZ2 bound CRALBP when tested with the overlay assay (Fig. 6A , lanes 1–4). In contrast, the C-terminal domain, which does not have a PDZ domain, did not bind CRALBP (Fig. 6A , lane 5). 

Peptide Competition Studies

PDZ domains generally bind their targets through their C-terminal four amino acids, which form H-bonds with a β-strand in the PDZ domain. If CRALBP binds to EBP50 through its PDZ domains, peptides corresponding to the C-terminal sequence of CRALBP should compete with the binding of the protein in the overlay assay. We tested this hypothesis by performing the CRALBP overlay assay after preincubation with a 2600-fold molar excess of C-terminal CRALBP peptides or control peptides. The results, shown in Figure 6B , indicate that the peptides AEVENTAL (mouse CRALBP C-terminal sequence is in italic) and AEVENTAF (bovine and human CRALBP C-terminal sequence is in italic) abolished binding of CRALBP in the overlay blot (Fig. 6B , lanes 2 and 3). Peptides EVENAAG and EVENAAD, which do not conform to a PDZ-binding motif, did not inhibit the interaction of CRALBP with EBP50 (Fig. 6B , lanes 4 and 5, respectively). These findings provide further evidence that CRALBP interacts with the PDZ domains of EBP50 through its C-terminal amino acids. 

CRALBP, EBP50, and Ezrin in RPE and Müller Cells

We examined frozen sections of retina from dark-adapted mice with antibodies to CRALBP, EBP50, and ezrin using laser scanning confocal microscopy. Low magnification showed CRALBP to be distributed throughout the retina from the inner limiting membrane to the RPE, as noted earlier (Figs. 7A 7G) . ⁹ This distribution is due to its presence in Müller cells, which span nearly the complete thickness of the retina, and in RPE cells. At higher magnification, CRALBP was widely distributed within RPE from the basal infoldings to the tips of the apical processes (Figs. 7D 7J) . RPE nuclei were seen as dark objects surrounded by fluorescence. At low magnification, EBP50 was evident in inner retina in a radial orientation resembling a Müller cell distribution (Fig. 7B) . Identification of these inner retinal cells requires more detailed studies. At higher magnification, diffuse labeling of EBP50 within RPE cells was largely confined to the apical processes and terminated abruptly at their bases (Fig. 7E) . A similar distribution was observed in Müller apical processes (Figs. 7B 7E) . Distribution of EBP50 clearly overlapped that of CRALBP in RPE apical processes, as evident in the merged images (Figs. 7C 7F) . EBP50 was also present along the basal RPE membrane and in a punctate distribution pattern over RPE and Müller apical processes and weakly over Müller cells as they traversed the outer nuclear layer (Fig. 7E , described later). Ezrin appeared to fill RPE apical processes from their tips to their bases and to be largely absent in RPE soma (Figs. 7H 7K) . A weaker signal was detected over Müller cell apical processes, which was more evident at higher microscope sensitivity settings. Ezrin distribution overlapped that of CRALBP in RPE apical processes, as indicated by the merged images (Figs. 7I 7L) . Both ezrin, ^{11 35 36} and EBP50 ³⁶ had been noted in RPE microvilli, and ezrin was reported in Müller microvilli. ^{35 37}  

The punctate signal (speckles) in Figure 7E was a consistent feature of labeling with anti-EBP50. Centrifugation, filtration, dilution of the antibody into nonionic detergents, or changing the source of the secondary antibody did not affect the pattern or intensity. The speckles were not a surface phenomenon, because they are present in 3-μm optical sections obtained from the middle of a 50-μm physical section (Figs. 7B 7E) . Finally, the distribution of speckles generally paralleled the diffuse labeling of EBP50 over RPE and Müller apical processes and was diminished in regions of retina with lesser amounts of EBP50 (e.g., basal RPE, area between the tips of RPE and Müller cell apical processes and choroid). Thus, the punctate labeling appears to be a feature of the distribution of EBP50. 

Discussion

Many cell-signaling and metabolic pathway components are organized into multimeric protein complexes that provide precise temporal and spatial localization of the pathway, maximize efficiency, and control cross-talk or diffusion (leakage) of pathway intermediates. ^{38 39} Interaction domains, sometimes present on scaffold proteins, often mediate formation of these protein complexes through their affinity for sequence and/or structural motifs of other proteins. PDZ domain scaffolds, named after the first three PDZ proteins discovered (PSD95/Discs large/ZO-1), participate in the organization of numerous complexes, including those at postsynaptic densities of neuronal dendrites, at tight junctions, within the apical processes of polarized epithelial cells, and in Drosophila photoreceptor cells. ^{38 39}  

Given the complexity of the visual cycle and the multiple functions that can arise from all-trans-retinol, one of the cycle intermediates, we sought and obtained evidence that some of the components of the cycle in RPE are organized into a functional complex. ²⁰ In this study, we found that CRALBP interacts with a well-known scaffold protein, EBP50, which had not been previously associated with the visual cycle. We identified the EBP50–CRALBP interaction with RPE microsomes by using an overlay assay and confirmed the interaction by showing that recombinant human EBP50 bound CRALBP and that a CRALBP-affinity column bound EBP50, ezrin, and RDH5 in RPE microsomes. Analysis of recombinant truncations of EBP50 and of peptide competition experiments indicated that CRALBP interacts with both PDZ domains of EBP50. Furthermore, we showed that interaction of these components is likely in vivo because they occur within the apical processes of RPE cells ^{11 35 36} and of Müller cells, which project into the same extracellular compartment in the retina, bordered by photoreceptor and RPE cells. The presence of CRALBP, EBP50, and ezrin in Müller apical processes is intriguing in light of the considerable indirect evidence supporting a role for these glial cells in cone visual pigment regeneration. ^{6 8}  

EBP50 has two related PDZ domains that interact with a number of soluble and membrane-bound proteins in various tissues ^{21 22 23 24} and a C-terminal domain that binds proteins of the ERM family, such as ezrin. ²¹ ERM proteins are multivalent linkers that connect plasma membrane proteins with the cortical actin cytoskeleton. An actin-binding domain near the C terminus establishes liaison with the F-actin, and N-terminal FERM (4.1 ERM) domains mediate interaction with plasma membrane proteins, either directly or indirectly, by binding a scaffold protein such as EBP50 or the related exchanger 3 kinase A regulatory protein (E3KARP). EBP50 thus contributes to the development of epithelial polarity by tethering plasma membrane proteins to the actin bundle within apical microvilli or at the apical side of an epithelial cell. Considerable evidence has accumulated in vitro that suggests that EBP50 plays multiple functional roles in the regulation of the transmembrane receptors and transporters, of recycling endosomes, and of other proteins localized to apical plasma membranes of polarized epithelia. ^{21 22 23 24} The initial characterization of NHERF-1 knockout mice revealed a renal phosphate wasting phenotype probably resulting from inappropriate accumulation of sodium-phosphate cotransporter type IIa (Npt2) at internal sites of the glomerular epithelial cell instead of its normal localization in the apical plasma membrane. ²⁴ Thus, EBP50 has the potential to cross-link CRALBP to the actin cytoskeleton in RPE apical processes. 

Structural Basis for the Interaction of CRALBP and EBP50

PDZ domains interact with their targets by a β-sheet augmentation mechanism in which the C-terminal four amino acids of the target orient antiparallel to the β2 strand of the PDZ domain and thereby extend or augment the β-sheet. ^{38 39} Based on comparisons of all known targets of EBP50, the consensus C-terminal sequence for PDZ1 is X(S/T)XL, where X is any amino acid. ³⁴ The C-terminal sequence of mouse CRALBP (NTAL) conforms to this consensus ⁴⁰ ; however, the C terminus of bovine CRALBP used in these studies is NTAF. ⁴⁰ Specificity for C-terminal leucine has been suggested to arise from insertion of its side chain into a tight hydrophobic pocket of PDZ1. ⁴¹ However, the investigators drew their inference from a crystal structure of a PDZ1 dimer in which the C-terminal leucine of one rPDZ1 molecule inserted into the PDZ1 domain of another at an angle that precluded β-sheet augmentation. Thus, it is possible that the hydrophobic cavity accommodates another hydrophobic residue such as phenylalanine, if the C-terminal peptide is oriented to allow formation of a classic β-sheet. No structural information is available regarding the PDZ2 domain of EBP50. 

PDZ Domains of EBP50 and CRALBP

Proteins with multiple PDZ domains frequently recognize several targets, allowing the assembly of multiprotein complexes. ³⁸ Of the targets identified for EBP50, most bind to PDZ1 and only YAP65 and PLC-β3 have been identified as targets of PDZ2. ³⁴ However, our studies indicate that both individual PDZ domains of EBP50 bind CRALBP when tested with the overlay assay. Given the relative abundance of CRALBP in RPE cells, it is likely that EBP50 tethers a complex of two CRALBPs to the actin cytoskeleton (Fig. 8) . CRALBP, in turn, could interact with other visual cycle components such as RDH5, ²⁰ a fraction of which has been localized to the plasma membrane of RPE cells (Fig. 8) . ⁴² Recent proteomic and kinetic evidence further supports the interaction of CRALBP and RDH5. ^{18 20} However, three caveats are in order. First, the overlay assays are not quantitative, and we do not know whether both PDZ domains can bind CRALBP simultaneously. Second, the presence of other target proteins in RPE with higher affinities for the PDZ domain and/or in large amounts could preclude the binding of CRALBP in vivo. Finally, other proteins that bind to CRALBP ²⁰ could mask sites important for interaction with EBP50. 

Dynamics of Complex Formation

CRALBP appears to be uniformly distributed across the RPE cell from basal infoldings to apical processes (Figs. 7A 7D 7G 7J) . However, both EBP50 and ezrin are found only in RPE apical microvilli, presumably because they are tethered to the cortical actin cytoskeleton. Thus, if CRALBP associates with EBP50 and ezrin in vivo, only a fraction of the total retinoid-binding protein could be involved in complex formation at any given time. Interactions of ezrin with the cytoskeleton and with membrane proteins are regulated by conformational changes that alternatively expose or bury binding domains. Activation of ezrin, induced, for instance, by binding PIP₂ or by phosphorylation of T567, disrupts interactions between N- and C-terminal domains that mask binding sites for actin, EBP50, and other proteins. ^{23 33} The dynamics of a cycle of masking and unmasking have been measured in cells in culture. ⁴³ If similar cycling of ezrin occurs in vivo in RPE cells, complexes of EBP50 and CRALBP would be alternatively confined to the actin cytoskeleton within the apical microvilli or free to diffuse to other parts of the RPE cells. Perhaps such a dynamic arrangement is essential for the unloading and reloading of a retinoid-processing complex in RPE. 

In summary, the interactions of EBP50, ezrin, actin, and CRALBP provide a structural basis for the formation of a retinoid-processing complex localized to the apical membranes of RPE and Müller cells. These interactions offer mechanistic insight into the cell biology of retinoid processing in RPE cells and suggest that cellular processing and intercellular diffusion of the lipophilic retinoids will be as complex as those described for cholesterol. 

 Supported by National Eye Institute Grants EY01730, EY02317, EY14239, EY06603, and GM39066, and by an unrestricted award from Research to Prevent Blindness, Inc. (RPB) to the University of Washington. JCS is a Senior Scientific Investigator of RPB.

 Submitted for publication September 8, 2003; revised Ocotber 9, 2003; accepted October 10, 2003.

 Disclosure: M. Nawrot, None; K. West, None; J. Huang, None; D.E. Possin, None; A. Bretscher, None; J.W. Crabb, None; J.C. Saari, 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.

 

Figure 1.

 

Depiction of some of the known reactions involved in regeneration of rod visual pigments. The tip of a rod photoreceptor and one disc membrane are shown, with the adjacent retinal pigment epithelium (RPE) above. The enzymes of RPE are shown associated with a continuous, internal membrane in the absence of evidence of their localization. CRALBP is depicted as an acceptor of 11-cis-retinol or 11-cis-retinal generated by isomerohydrolase or 11-RDH, respectively. Isom, isomerohydrolase; LRAT, lecithin-retinol acyltransferase; RDH, photoreceptor retinol dehydrogenase; 11-RDH, 11-cis-retinol dehydrogenases (RDH5). Adapted, with permission, from Saari JC. Biochemistry of visual pigment regeneration. The Friedenwald Lecture. Invest Ophthalmol Vis Sci. 2000;41:337–348. © Cadmus Professional Communications.

Figure 2.

 

(A) Overlay assay. Lane 1: RPE microsomal proteins were resolved by SDS-PAGE, transferred to a membrane, and stained with anti-CRALBP. Lane 2: As for lane 1, except that the blot was incubated with CRALBP (overlay) before staining with anti-CRALBP. Lane 3: RPE microsomal proteins stained with Coomassie blue. Arrowhead: CRALBP-binding component subsequently identified as EBP50. (B) CRALBP binds to rEBP50. SDS-PAGE. Lane 1: RPE microsomes were analyzed with the CRALBP overlay assay. Arrowhead: component identified as EBP50. Lane 2: RPE microsomes were analyzed with anti-EBP50. Lane 3: rEBP50 was analyzed with the CRALBP overlay assay. Anti-EBP50 labels a component in bovine RPE microsomes with a migration identical with that labeled by the overlay assay. rEBP50 (human) binds CRALBP strongly and migrates more rapidly than bovine EBP50 in RPE microsomes.

Figure 3.

 

2D gel electrophoresis of microsomes. (A) A fraction from RPE microsomes purified by the Dspin procedure was analyzed by 2D gel electrophoresis and stained with Coomassie blue. (B) A duplicate 2D gel was transferred to a membrane and analyzed with the CRALBP overlay assay. The leftmost lane shows the sample analyzed by 1D SDS-PAGE only. The five isoelectric variants in (A, arrows), which correspond to the overlay-assay–positive components in (B), were excised for in situ tryptic digestion and LC-MS/MS. The overlay-positive components of slightly lower molecular weight may be proteolytic fragments. The isoelectric variants may result from heterogeneous phosphorylation.

Figure 4.

 

Affinity chromatography. (A) CRALBP-agarose. Solubilized RPE microsomes were applied to the affinity column. After a wash with 0.3 M NaCl, bound proteins were eluted with 2 M NaCl. Fractions loaded and eluted from the columns were analyzed with either a CRALBP overlay assay or with anti-EBP50. Lane 1: applied fraction, overlay assay; lane 2: eluted fraction, overlay assay; lane 3: applied fraction, anti-EBP50; lane 4: eluted fraction, anti-EBP50. Arrowhead: component identified as EBP50; arrow: position of CRALBP. (B) CRALBP- and IRBP-agarose. Solubilized RPE microsomes were applied and eluted as in (A) and analyzed with anti-EBP50. Lane 1: applied fraction; lane 2: CRALBP-agarose, eluted fraction; lane 3: IRBP-agarose, eluted fraction. The results indicate that EBP50 bound to CRALBP-agarose but not IRBP-agarose.

Figure 5.

 

RDH5 and ezrin bind to CRALBP-agarose. Solubilized RPE microsomes were applied and eluted from CRALBP- or IRBP-agarose as in Figure 4 , and the applied and eluted fractions were analyzed with anti-RDH5 or anti-ezrin. (A) Analysis with anti-RDH5. Lane 1: applied fraction; lane 2: eluted fraction, CRALBP-agarose; lane 3: eluted fraction, IRBP-agarose. (B) Analysis with anti-ezrin. Lane 1: applied fraction; lane 2: eluted fraction. Ezrin binding to CRALBP is probably mediated by EBP50.

Figure 6.

 

(A) Expressed domains of EBP50. Domains of EBP50 were expressed as GST-fusion proteins and analyzed with a CRALBP overlay assay. Lane 1: full-length EBP50, residues 1-358; lane 2: PDZ1+PDZ2, residues 1-248; lane 3: PDZ1, residues 1-138; lane 4: PDZ2, residues 138-248; lane 5: C-terminal domain, residues 241-358. Details regarding the constructs are provided in the text. Note that CRALBP bound to both PDZ1 and PDZ2. (B) Peptide competition experiments. Peptides corresponding to the C-terminal sequences of bovine and human/mouse CRALBP and control peptides were tested for their ability to compete with CRALBP in the overlay assay with solubilized RPE microsomes. Lane 1: overlay assay, no peptides. Lane 2: AEVENTAL, mouse CRALBP sequence is italic. Lane 3: AEVENTAF, bovine/human CRALBP sequence is italic. Lane 4: EVENAAD. Lane 5: EVENAAG. Arrowhead: position of EBP50; arrow: position of CRALBP.

Figure 7.

 

Colocalization of CRALBP, EBP50, and ezrin in retina. Frozen sections of dark-adapted mouse retina were examined by laser scanning confocal microscopy after staining with anti-CRALBP and anti-EBP50 (A–F) or with anti-CRALBP and anti-ezrin (G–L). (A–C, G–I) Low-magnification, full-thickness retina. (D–F, J–L) High-magnification, outer retina. The optical sections shown are approximately 3 μm thick and were obtained from the interior of the microtome section. (A, D) Anti-CRALBP, green. (B, E) Anti-EBP50, red. (C, F) Merged images. (G, J) Anti-CRALBP, green. (H, K) Anti-ezrin, red. (I, L) Merged images. RPE-S, retinal pigment epithelial cell soma; RPE-AP, retinal pigment epithelial cell apical processes; M-AP, Müller cell apical processes; OLM, outer limiting membrane; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer.

Figure 8.

 

Hypothesis for a retinoid-processing complex in apical RPE. Left: depiction of an apical RPE process filled with oriented actin fibers cross-linked to each other and to the plasma membrane. Right: an enlarged depiction of one actin filament and the plasma membrane. Ezrin binds actin through a C-terminal domain and binds EBP50 through an N-terminal binding site. CRALBP is shown associated with PDZ1 and PDZ2 of EBP50. RDH5, a known interaction partner of CRALBP, is shown in the plasma membrane, as reported by Mata et al., ⁴² although it is found elsewhere in RPE as well. Other RDHs and visual cycle enzymes may be interaction partners. Ezrin undergoes cycles of rapid activation and inactivation, which may be involved in retinoid processing on CRALBP.

The authors thank Anita Hendrickson, PhD, for valuable discussions and Gregory Garwin for preparation of the illustrations and for critical reading of the manuscript.