Rpe65 Is a Retinyl Ester Binding Protein That Presents Insoluble Substrate to the Isomerase in Retinal Pigment Epithelial Cells*

Photon capture by a rhodopsin pigment molecule induces 11-cis to all-trans isomerization of its retinaldehyde chromophore. To restore light sensitivity, the all-trans-retinaldehyde must be chemically re-isomerized by an enzyme pathway called the visual cycle. Rpe65, an abundant protein in retinal pigment epithelial (RPE) cells and a homolog of β-carotene dioxygenase, appears to play a role in this pathway. Rpe65-/- knockout mice massively accumulate all-trans-retinyl esters but lack 11-cis-retinoids and rhodopsin visual pigment in their retinas. Mutations in the human RPE65 gene cause a severe recessive blinding disease called Leber's congenital amaurosis. The function of Rpe65, however, is unknown. Here we show that Rpe65 specifically binds all-trans-retinyl palmitate but not 11-cis-retinyl palmitate by a spectral-shift assay, by co-elution during gel filtration, and by co-immunoprecipitation. Using a novel fluorescent resonance energy transfer (FRET) binding assay in liposomes, we demonstrate that Rpe65 extracts all-trans-retinyl esters from phospholipid membranes. Assays of isomerase activity reveal that Rpe65 strongly stimulates the enzymatic conversion of all-trans-retinyl palmitate to 11-cis-retinol in microsomes from bovine RPE cells. Moreover, we show that addition of Rpe65 to membranes from rpe65-/- mice, which possess no detectable isomerase activity, restores isomerase activity to wild-type levels. Rpe65 by itself, however, has no intrinsic isomerase activity. These observations suggest that Rpe65 presents retinyl esters as substrate to the isomerase for synthesis of visual chromophore. This proposed function explains the phenotype in mice and humans lacking Rpe65.

Light perception in vertebrates is mediated by a group of G protein-coupled receptors called the opsins. Most opsin pigments contain 11-cis-retinaldehyde (11cRAL) 1 as the light-ab-sorbing chromophore. Absorption of a photon induces 11-cis to all-trans isomerization of the chromophore, resulting in the activated species, metarhodopsin II. After a brief period, metarhodopsin II decays to yield apo-rhodopsin and free alltrans-retinaldehyde (atRAL). Before light sensitivity of the pigment can be restored, the atRAL must be chemically re-isomerized to 11cRAL by a metabolic pathway called the visual cycle. Most steps in this pathway take place within cells of the retinal pigment epithelium (RPE) adjacent to the photoreceptors. The key step in this pathway is all-trans to 11-cis re-isomerization of the retinoid, which is catalyzed by an enzyme activity called isomerohydrolase (IMH). IMH has been shown to use fatty acyl esters of retinol as a substrate (1,2), harnessing the energy of ester hydrolysis [⌬G ϭ Ϫ5 kcal/mol (3)] for the endothermic conversion of all-trans-retinol (atROL) to 11-cis-retinol (11cROL) (ϩ4.1 kcal/mol, Ref. 4). IMH has never been purified or cloned.
Leber's congenital amaurosis (LCA) is a severe and relatively common autosomal recessive disease that results in blindness at birth. LCA is frequently caused by mutations in the RPE65 gene (5,6). Rpe65, the product of this gene, is an abundant protein of unknown function in cells of the RPE (7). Rpe65 has high affinity for phospholipid membranes but contains no membrane-spanning domains (8,9). Mice with a knockout mutation in rpe65 massively accumulate all-trans-retinyl esters (atRE) in their RPE and have no detectable 11-cis-retinoids (10). Photoreceptors in rpe65 Ϫ/Ϫ mice are morphologically normal but contain only apo-rhodopsin instead of rhodopsin pigment. These observations led to speculation that Rpe65 may be IMH.
In the current study, we sought to define the function of Rpe65 and its relationship to IMH. We show by several approaches that Rpe65 specifically binds all-trans-retinyl palmitate (atRP), the most abundant retinyl ester in RPE cells (11). Further, we present data suggesting that atRP bound to Rpe65 is the substrate for IMH.

EXPERIMENTAL PROCEDURES
Expression of Bovine Rpe65 in Baculovirus-infected Sf9 Cells-We cloned the complete coding sequence of bovine Rpe65 into pFastBac bacmid (Invitrogen) and used it to transfect Spodoptera frugiperda (Sf9) cells following the manufacturer's procedure. Following infection, adherent cells were lysed in phosphate-buffered saline, pH 7.0, 1% CHAPS, and protease inhibitor mixture (Roche Diagnostics) and centrifuged at 100,000 ϫ g for 30 min. The detergent-soluble fractions were analyzed by gel filtration (Zorbax GF-250) on an Agilent 1100 series liquid chromatograph (mobile phase ϭ 20 mM Na 2 HPO 4, 130 mM NaCl, flow ϭ 1 ml/min, 280 nm detection). Proteins eluting at M r 158 -43 kDa were collected and concentrated using Centricon 30,000 molecular weight cut-off (MWCO) ultrafiltration devices (Millipore). The retained proteins were separated on an Ä KTA FPLC (Amersham Biosciences) using a Mono Q anion exchange column equilibrated with 25 mM Tris-OAc, pH 7.4, 1 mM EDTA, 0.1 mM dithiothreitol, 10 mM NaOAc, flow ϭ 1 ml/min. A gradient to 0.5 M NaOAc over 60 min was initiated 10 min after sample injection. Collected gradient fractions were analyzed by SDS-PAGE and Western blotting to confirm the presence of Rpe65. For reconstitution studies, partially purified Rpe65 or control proteins from non-expressing Sf9 cells were exchanged into reconstitution buffer (10 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM EDTA) using 30,000 MWCO ultrafiltration devices. In some experiments the detergent-soluble fractions were used without further purification.
Bovine Rpe65 Antibody-The synthetic peptide from bovine Rpe65, NFITKINPETLETIK, was coupled to keyhole limpet hemocyanin and injected (ϳ500 g) subcutaneously as an emulsion with complete Freund' s adjuvant (Invitrogen) into New Zealand White rabbits. Intramuscular boosts were done at 3-week intervals using incomplete Freund' s adjuvant. Total IgG was purified from whole serum on a protein A column. The IgG fraction (15 mg), was immobilized on 1 ml of Affi-Gel 10 matrix according to the manufacturer's instructions.
SDS-PAGE and Western Blot Analysis-Protein samples were resolved by SDS-PAGE on a 4 -12% gradient gel and transferred to polyvinylidene difluoride membranes (Millipore) in a semidry transblot apparatus (BioRad) according to the manufacturer's instructions. After blocking in 5% nonfat milk, blots were reacted with the anti-Rpe65 peptide Ab at 1:2000 dilution. Bands were detected using enhanced chemiluminescence according to the manufacturer's instructions (Kirkegaard & Perry laboratories).
Retinoid Preparation and Synthesis-Retinoids were obtained from Sigma (atRP, atROL, atRAL, and ␤-carotene) or were synthesized from 11cRAL (11cRP and 11cROL) according to published procedures (12). All retinoids were purified by HPLC to Ͼ98% purity (ratio of sample peak area to total chromatogram peak area) and quantified by UV-vis spectroscopy using published molar extinction coefficients (13). Purified retinoid stocks were stored in ethanol at Ϫ80°C prior to use.
Spectrophotometric Analysis of Retinoid Binding-Divided quartz cuvettes (containing two compartments separated by a quartz window) were used in mixing studies to detect retinoid binding. An absorbance baseline (400 -250 nm) was established in which each cell contained a protein sample (0.5-1.0 mg/ml of detergent-solubilized Rpe65 in 10 mM HEPES, pH 7.2, 100 mM NaCl, 0.1% CHAPS) in one compartment and the test retinoid (atRP, atROL, ␤-carotene, 11cRP, 11cROL, atRAL, or 11cRAL) at 40 M in the same buffer in the adjacent compartment. The sample cell was mixed and absorbance measurements were taken at the indicated times using the un-mixed cell as a reference (temp ϭ 30°C). Control studies were performed in an identical fashion using either heat-denatured protein or lipid-free bovine serum albumin (1 mg/ml).
Gel Filtration and Immunoprecipitation-[1-14 C]palmitate-atRP (80 M, 2 Ci) was added (in DMF at 0.25% v/v) to detergent-solubilized proteins from either non-transfected Sf9 cells or Sf9 cells expressing Rpe65. Samples were rocked overnight at 4°C and centrifuged at 45,000 ϫ g, 30 min to remove particulate debris. Sample aliquots from the binding mixtures were analyzed by gel filtration chromatography (described above) with on-line photodiode array and radiometric ([ 14 C]) detection (525TR FSA, Packard Instruments) and in immunoprecipitation studies. Briefly, 50 l of Rpe65 IgG-Affi-Gel 10 beads were incubated with 30 l of sample (ϳ50 g of protein) in phosphate-buffered saline/0.3% CHAPS at 4°C overnight. Similar incubations were performed on control samples containing retinoid alone (C1) and protein alone (C2). Following incubation, the mixtures were centrifuged at 15,000 ϫ g, 30 s and the supernatant (unbound) fractions were collected. The beads were washed three times with 500 l of phosphatebuffered saline/0.3% CHAPS and resuspended in 50 l of of 2ϫ Laemmli buffer (4% SDS). Aliquots of each fraction were taken for scintillation counting and for Western blot analysis. Radoactivity in C1 was used to correct for nonspecific retinoid binding, while C2 was used to confirm immunolabeling of Rpe65.
Preparation of BODIPY-PC and Pyranine Proteoliposomes-Proteoliposomes were prepared according to published methods (14) with minor modifications. Chloroform stocks of dioleoyl phosphatidylcholine (DOPC) and dilinoleoyl phosphatidylcholine (DLPC) (Avanti Polar Lipids) were mixed (85:15 mol/mol, respectively) and dried under a stream of argon. Identical samples containing 0.5 mol% of the phosphatidylcholine analog, ␤-BODIPY 500/510 C12-HPC, (Molecular Probes) were treated in a similar fashion. The dried phospholipid (PL) residues were resuspended in hexane and placed on dry ice. Methanol was added to 5% v/v, and the samples were placed under high vacuum for 12-16 h. Partially purified Rpe65 or control proteins (30,000 MWCO retentates from non-transfected Sf9 cells) were added to resuspend the PL residues. Samples that did not contain BODIPY-PC were supplemented with pyranine (Molecular Probes) at 2 mM in reconstitution buffer. Samples were agitated for 2 h at room temperature. Proteoliposomes were prepared from the resulting multilamellar vesicles (MLV) by the extrusion method (15,16). The MLV suspensions were extruded through 5-m polycarbonate membranes (Nucleopore Corp) using a Liposofast lipid extruder (Avestin). Vesicles were fractionated and washed by repeated centrifugation (233,000 ϫ g, 45 min) and resuspension in reconstitution buffer. In some experiments, vesicles were fractionated by gel filtration as described above. The washed proteoliposomes (final phospholipid concentration ϭ 2.5-3.6 mM, final protein concentration ϭ 0.8 -1.5 mg/ml) were used immediately or stored at 4°C for 48 h before use.
Fluorescence Resonance Energy Transfer (FRET) Studies and Fluorescence Microscopy-The binding/internalization of atRP in Rpe65 and control proteoliposomes was determined by FRET analysis where the emission of atRP (400 -500 nm) was used to excite the BODIPY-PC or pyranine fluorescent dyes. atRP was added (in DMF, 0.1% v/v) to the washed proteoliposomes at 2 mol% relative to total phospholipids, and the samples were rocked at 33°C. Aliquots of the binding mixtures were removed at the indicated times and analyzed on an Agilent liquid chromatograph equipped with an on-line fluorescence detector (flow rate ϭ 0.5 ml/min; temp ϭ 25°C). The instrument executed two injections with a 1-min interval between injections. During the first injection, excitation was set to 325 Ϯ 10 nm; emission was set to either 520 Ϯ 10 nm for BODIPY-PC proteoliposomes or 510 Ϯ 10 nm for pyranine proteoliposomes. For the second injection, excitation was set to 460 Ϯ 10 nm, emission acquisition parameters were unchanged. Spectral data for all runs were acquired in the region of 400 -600 nm to confirm FRET spectra. The association of atRP with liposomes containing either Sf9-Rpe65 or Sf9 protein was examined by phase contrast fluorescence microscopy on a Zeiss Axiovert 135 inverted microscope using a Plan-Neofluor 40ϫ/0.75 objective and appropriate filters (DAPI/ Hoechst/AMCA, 360 Ϯ 40 nm exciter, 460 Ϯ 25 nm emitter) (Chroma Technology).
Preparation of Proteoliposomes-Microsomal proteins were prepared from homogenates of bovine RPE as previously described (17). Microsomal membranes (5-10 mg/ml) were treated with 365-nm UV light (5 min on ice) to destroy endogenous retinoids, and were homogenized in solubilization buffer (25 mM Tris OAc, pH 8.0, 1 mM dithiothreitol, 1 mM EDTA, 1 M leupeptin, 0.5% Genapol X-100). The sample was rocked for 1 h at 4°C followed by centrifugation at 175,000 ϫ g, 35 min. Solubilization buffer in the supernatant fraction was exchanged for reconstitution buffer as described above (final protein concentration ϭ 2-4 mg/ml, final detergent concentration ϭ 0.25%). The protein sample was added to an equal volume of PL suspension containing dioleoyl phosphatidylcholine/dioleoyl phosphatidylethanolamine/phosphatidyl inositol/phosphatidic acid (25:60:5:10, mol/mol, respectively in reconstitution buffer, 10 mM total lipid) to obtain a final protein concentration of 1-2 mg/ml and a lipid concentration of 5 mM. Apo-CRALBP (18) was added to a final concentration of 10 M, and the sample was agitated at 37°C for 1 h. The sample was treated with SM-2 Bio-Beads to remove the detergent and generate proteoliposomes (19,20). Vesicles were fractionated and washed as described above. To prepare Sf9-atRP and Rpe65-atRP proteoliposomes, mixtures containing varied amounts of solubilized protein from either non-expressing Sf9 cells (1-3 mg/ml) and atRP (20 -60 M) or Rpe65-expressing Sf9 cells (3-5 mg/ml) and atRP (20 -60 M) were prepared and mixed at 4°C for 1 h. Sf9-atRP proteoliposomes were generated by adding the mixed samples to an equal volume of the PL suspension followed by detergent removal and vesicle fractionation as described above for bovine RPE proteoliposomes. Rpe65-atRP proteoliposomes were prepared similarly except that the binding mixtures were first applied to a gel filtration column to obtain the atRP-bound Rpe65 complex (see fraction 18, Fig. 2A). The recovered fraction was added to an equal volume of the PL suspension and vesicles were generated as described. Liposomes containing only atRP were prepared by mixing various concentrations of atRP (10 -60 nmol in chloroform) with a similar PL mixture (5 mM) before solvent evaporation and resuspension in reconstitution buffer. To facilitate sample comparison in enzyme assays, vesicles containing similar amounts of protein (0.2-0.4 mg/ml) and/or atRP (3-5 M) were used.
Proteoliposome Fusion and Enzyme Assays-The ability of Sf9-atRP-, Rpe65-atRP-, or atRP-liposomes to serve as substrate vehicles for enzymes in proteoliposomes containing solubilized bovine RPE microsomal protein was determined by inducing fusion of the vesicles with polyethylene glycol (21,22). A 200-l aliquot of the respective substrate liposomes was added to 200 l of RPE proteoliposomes, and the samples were preincubated for 5 min at 37°C. Prewarmed polyethylene glycol was added to 10% w/v, and incubation was resumed for 1.5 h at 37°C. Samples were quenched with 200 l of chilled methanol and 100 l of water. Retinoids were extracted into hexane (2 ϫ 500 l) and analyzed by HPLC as previously described (17).
Effects of Rpe65 on IMH Activities in Mouse RPE-Wild-type (C57BL/6) and strain-matched rpe65 Ϫ/Ϫ knockout mice (10) were reared in 12-hr light/dark cycles at 20 -40 lux for 5 weeks. Mice (n ϭ 12 per group) were anesthetized by intraperitoneal injection of ketamine (200 mg/kg) and xylazine (10 mg/kg) and sacrificed by cervical dislocation. Eyes were removed and the anterior segments, vitreous and retina were discarded. The remaining RPE/eyecups were homogenized in BTP buffer (12 eyecups/ml, 10 mM BTP, pH 7.5, 1 mM dithiothreitol, 1 M leupeptin, 2 mM CaCl 2 ) using a Duall glass-glass homogenizer, and centrifuged at 175,000 ϫ g for 15 min. The resulting pellets were homogenized in BTP buffer containing 0.1% CHAPS (w/v), and the centrifugation step was repeated. The supernatant fraction, which contained ϳ95% of the endogenous retinoids, was decanted, and the sedimented protein was homogenized in BTP buffer containing 1% CHAPS followed by brief sonication (3 ϫ 5 s bursts, 20 watt output) and mixing (45 min at 4°C). The samples were centrifuged at 175,000 ϫ g for 30 min. The soluble fractions were removed and subjected to gel filtration to remove excess detergent, as described above (final CHAPS concentration ϭ 0.2%). The recovered protein was concentrated and assayed for IMH activity plus or minus added Rpe65 from expressing Sf9 cells. Briefly, 400 g of each protein sample (100 l) were added to a reaction mixture containing 25 M apo-CRALBP in BTP buffer (100 l) and either 100 g of Rpe65 (50 l) or 50 l of reconstitution buffer. 1 l of a 400-M solution of atRP in DMF was added to the reaction mixtures (final atRP concentration ϭ 1.6 M), the reaction mixtures were sonicated as described above, and incubated at 37°C for 1 h with vigorous shaking. Control reactions to measure potential catalytic activity of Rpe65 were performed as above except that the RPE proteins were substituted with heat-denatured protein. [ 3 H]atROL or [ 3 H]atRP were used as substrate. All reactions were quenched with 100 l of methanol and the retinoids extracted into 600 l of hexane. Solvent was removed under a stream of nitrogen and the sample residues were resuspended in 10% dioxane/hexane. Retinoids were analyzed by HPLC as described on an Agilent Zorbax RX-SIL column (4.6 ϫ 250 mm) and a mobile phase of 10% dioxane/hexane, flow ϭ 1 ml/min.
Determination of Vesicle Size-Average diameters of the lipid vesicles used in FRET analyses and in fusion assays were determined by quasi-elastic light scattering (QELS). QELS measurements were performed on a Nicomp 380 Submicron Particle Sizer equipped with a helium-neon laser (532 nm) and a computer-controlled auto-correlator (Particle Sizing Systems). Size determinations were made after 15-20 min of analysis time. These data were compared with data obtained from polystyrene microspheres of known size (Duke Scientific Corporation).

Interaction between Rpe65 and atRP Induces a Change in the UV Absorption Spectrum-A specific interaction between an
apoprotein and its light-absorbing ligand invariably induces a change in the absorption spectrum. An indication of this change is the presence of one or more nodal isosbestic points in overlaid spectra from before and after interaction of the ligand with the protein (23). To test for possible binding of atRE to Rpe65, we added samples of atRP and Rpe65 separately to the two chambers of a divided cuvette. In this and all subsequent experiments, we used baculovirus-infected Sf9 cells as a source of Rpe65. These cells contained no detectable retinoids, hence the starting Rpe65 protein was in the apo or unbound state. UV absorption spectra were obtained at different times after sample mixing in the cuvette. Incubation of atRP with Rpe65 caused a reduction in UV absorption at 355 nm and an increase at 280 nm, resulting in an isosbestic point at 295 nm (Fig. 1A). The absorption maximum ( max ) of atRP dispersed in binding buffer was 355 nm. Thus, incubation with Rpe65 reduced the UV absorption of free atRP. We repeated this mixing experiment, substituting the geometric isomer, 11-cis-retinyl palmitate (11cRP) for atRP. Here, no change in UV absorption was observed during incubation with Rpe65 (Fig. 1B), suggesting that the interaction with atRP is specific for the all-trans isomer. We used the same spectral assay to test for possible interactions between Rpe65 and atRAL or 11cRAL. Similar to 11cRP, we observed no changes in the UV-absorption spectra of the retinaldehydes upon incubation with Rpe65 (not shown). We also tested for an interaction between Rpe65 and ␤-carotene. Again, incubation with Rpe65 resulted in no change in the absorption spectrum (not shown). However, incubation of Rpe65 with atROL induced a hyperchromatic shift in near-UV absorption with appearance of an isosbestic point at 285 nm (Fig. 1C). As a positive control, we performed similar spectral analysis on 11cROL and cellular retinaldehyde-binding protein (apo-CRALBP), which specifically binds 11cROL and 11cRAL (24). As expected, we observed an isosbestic point in the overlaid spectra (Fig. 1D). Together these results suggest that Rpe65 interacts with atRP and atROL, but does not interact with 11cRP, atRAL, 11cRAL, or ␤-carotene.
Co-elution of Rpe65 and atRP by Gel Filtration and Immunoprecipitation-If Rpe65 binds atRE, atRP should co-elute with Rpe65 during gel filtration chromatography. We incubated a CHAPS detergent extract of Rpe65-expressing Sf9 cells with an excess of [ 14 C]atRP. As a control, we incubated a CHAPS extract of non-expressing Sf9 cells with [ 14 C]atRP. We separated both samples by gel filtration chromatography and monitored atRP elution by measuring 14 C-disintegrations per min (dpm) with an online flow-scintillation analyzer. We also collected 1-min fractions for analysis by SDS-PAGE. The [ 14 C]atRP eluted in two peaks (Fig. 2A). The first peak was in the column void volume, similar to the profile observed when [ 14 C]atRP in CHAPS was chromatographed alone (not shown) or with a CHAPS extract of non-expressing Sf9 cells (Fig. 2A). The second peak of 14 C-dpm was in fractions 17-19 of the Rpe65-expressing sample and contained the preponderance of eluted Rpe65 by protein gel electrophoresis (Fig. 2B). No second peak of 14 C-dpm was observed in the sample derived from non-expressing Sf9 cells. We obtained UV absorption spectra at the two peaks of 14 C-dpm in the Rpe65 experiment. The spectrum from the first peak showed a max of 325 nm ( Fig. 2A,  inset), characteristic of atRP in CHAPS. The spectrum from the second peak showed a max of 280 nm, characteristic of proteins, with less intense absorption at 325 nm ( Fig. 2A, inset). These data suggest that atRP binds to Rpe65.
To confirm that atRP binds specifically to Rpe65, we prepared an immunoaffinity matrix containing immobilized IgG from antisera against Rpe65. We incubated CHAPS detergent extracts of Rpe65-expressing and non-expressing Sf9 cells with an excess of 14 C-labeled atRP. Both samples were incubated with the immunoaffinity matrix. We analyzed the unbound, wash, and eluted fractions for 14 C-dpm. We also did immunoblot analysis of the unbound and eluted fractions with the Rpe65 antibody. Most 14 C-dpm were in the unbound fraction due to the excess of added [ 14 C]atRP (Fig. 2C). This fraction also showed considerable Rpe65 immunoreactivity due to incomplete binding to the immunoaffinity matrix. Importantly, significant 14 C-dpm (ϳ15%) were present in the Rpe65-containing eluted fraction. 14 C-dpm were scarcely detectable in the equivalent fraction of the non-expressing control sample (Fig.  2C). We confirmed chemical integrity of the eluted [ 14 C]atRP by HPLC analysis following gel filtration and immunoprecipitation. Together these results identify Rpe65 as the protein binding partner for atRP.
Since atROL showed spectral evidence for an interaction with Rpe65 (Fig. 1C), we repeated the gel filtration experiment, substituting [ 3 H]atROL for [ 14 C]atRP. Here, we observed no co-elution of 3 H-dpm with Rpe65 (not shown). This suggests that the interaction of atROL with Rpe65 may be too weak to survive chromatographic separation.
Rpe65 Solubilizes atRP and Facilitates Its Removal from Membranes-Retinyl esters are virtually insoluble in aqueous solutions and thus do not exchange between membrane compartments (25,26). A potential function for Rpe65 may be to solubilize atRE and to facilitate their removal from membranes. To test this possibility, we employed a novel fluorescent resonance energy transfer (FRET) binding assay in liposomes, using the fluorescent probes, ␤-BODIPY-500/510 C 12 -HPC (BODIPY-PC) (27) and pyranine (28). Table I shows the maximal excitation and emission wavelengths for these fluorophores and for atRP in liposomes. The excitation spectra of BODIPY-PC and pyranine overlap with the emission spectrum of atRP. On the other hand, the excitation spectrum of atRP overlaps little with the excitation spectrum of BODIPY-PC or pyranine. To establish conditions of FRET we prepared liposomes containing: (i) BODIPY-PC alone; (ii) BODIPY-PCϩ atRP; (iii) pyranine alone; or (iv) pyranineϩ atRP. We excited each type of liposome at 325 nm, the excitation maxi-mum for atRP, and measured the emission spectrum. Relatively weak emissions were detected from liposomes containing BODIPY-PC or pyranine alone (Fig. 3, A and B), due to limited direct fluorescence of these probes with 325-nm excitation. The

FIG. 2. Co-elution of atRP with Rpe65 during gel filtration chromatography and immunoprecipitation.
A, gel filtration of detergent extracts from expressing and non-expressing Sf9 cells following incubation with excess [ 14 C]atRP. Eluates were analyzed for 14 C-dpm using an online flow-scintillation detector. Normalized radiograms from Rpe65-expressing (red) and non-expressing (blue) extracts are shown. The inset shows normalized UV absorption spectra at the two peaks of [ 14 C]atRP dpm in the Rpe65 eluate. Note the higher 280-nm and lower 325-nm absorption in the second (bound) peak. B, protein gel electrophoresis of selected gel filtration fractions from panel A. Following SDS-PAGE, the gel was stained with Sypro-red (per the manufacturer's procedure). C, immunoprecipitation analysis of detergent extracts from Rpe65-expressing and non-expressing Sf9 cells following incubation with [ 14 C]atRP. The affinity matrix contained immobilized IgG against Rpe65. Histograms show the percent of total 14 C-dpm in the unbound, wash, and eluted fractions. Insets above show excerpts of immunoblots with the Rpe65 antibody on the unbound and eluted fractions from Rpe65-expressing and non-expressing Sf9 extracts. fluorescent emissions were 4-fold higher from liposomes containing atRPϩ BODIPY-PC or atRPϩ pyranine (Fig. 3, A and  B). The increased fluorescent emissions in the presence of atRP were due to FRET.
Next, we used the FRET assay to measure Rpe65-dependent uptake of atRP into the lipid bilayer and aqueous interior compartments of liposomes. We prepared four types of liposomes containing partially purified Rpe65 from expressing, or equivalent chromatographic fractions from non-expressing Sf9 cells, and BODIPY-PC or pyranine. Similar quantities of total protein were used in the Rpe65-expressing and non-expressing liposomes. We added free atRP in dimethylformamide (DMF) to media containing each type of liposome, incubated for different times, and measured fluorescent emission at 520 or 510 nm with 325-nm excitation. After a 30-min incubation, the atRPdependent FRET signal at 520 nm was ϳ6-fold higher in BODIPY-PC liposomes containing Rpe65 compared with Sf9controls (Fig. 3C). A similar experiment was performed on pyranine-containing liposomes. Here, the 510-nm FRET signal was also about 6-fold higher in liposomes containing Rpe65 compared with Sf9-controls (Fig. 3D). These results show significantly higher uptake of atRP into the Rpe65-containing compared with Sf9-control liposomes.
The BODIPY fluorophore is coupled to a fatty-acyl chain in phosphatidylcholine, and is thus confined to the internal lipid bilayer. Pyranine is a hydrophilic trisodium salt, confined to the aqueous interior of liposomes. Accordingly, the BODIPY-PC signal with 325-nm excitation is due to FRET from atRP in the bilayer of liposomes while the pyranine signal is from atRP in the aqueous lumina. The onset of FRET was faster in the bilayer than in the aqueous interior of Rpe65containing liposomes (Fig. 3, C and D). The Rpe65-dependent membrane FRET showed early saturation while the luminal FRET showed sigmoidal kinetics with an initially slow rate of increase. These kinetic data make sense considering that free atRP in the medium must pass through the liposome membrane to reach the aqueous interior.
To rule out that the increased fluorescent signals in liposomes containing Rpe65 versus control proteins were due to different levels of BODIPY-PC or pyranine, we excited both liposome populations at 460 nm, the maximum emission wavelength of atRP, and measured fluorescent emission at 520 or 510 nm. For both probes the emission signals were virtually identical in Rpe65 and Sf9-control liposomes (not shown). Thus, the increased FRET signals in the Rpe65-containing liposomes (Fig. 3, C and D) were not caused by different levels of the probes. It is also unlikely that orientation of the fluorophores contribute to the observed differences between Rpe65 and Sf9-control liposomes. Pyranine, atRP bound to soluble Rpe65, and atRP within the bilayer should experience unrestricted isotropic motion. The motion of BODIPY coupled to phosphatidycholine is partially constrained. However, this constraint is similar in the Rpe65 and control liposomes.
To confirm Rpe65-dependent uptake of atRP, we analyzed the Rpe65-containing and Sf9-control liposomes by fluorescence microscopy after addition of atRP. We excited with light at 360 nm and imaged using a 435-495-nm bandpass filter. Under these conditions, only direct atRP fluorescence is de-tected. The control liposomes showed a ring-like pattern of fluorescence (Fig. 3E), suggesting that the atRP was restricted to the liposome membranes. We also consistently observed small fluorescent droplets in the medium surrounding the control liposomes (Fig. 3E). The pattern of fluorescence was dramatically different in Rpe65-containing liposomes. Here, we observed prominent atRP fluorescence in both the membrane and aqueous interior (Fig. 3F). Interestingly, the small fluorescent droplets were never seen in media surrounding these Rpe65 liposomes. Together, the results presented in Fig. 3 show that Rpe65 not only binds atRE but can also extract them from membranes into the aqueous interior of liposomes. These characteristics suggest a possible function for Rpe65.
Rpe65 Is Required for Isomerization of atRP but Contains No Intrinsic Isomerase Activity-A potential role for Rpe65 as an atRE-binding-protein may be to present this otherwise insoluble substrate to IMH for isomerization. To test this possibility, we prepared RPE membranes from wild-type (C57BL/6) and Emission spectra were obtained with 325-nm excitation from liposomes containing BODIPY-PC alone or BODIPY-PC plus atRP. B, atRP-pyranine FRET. Emission spectra were obtained from liposomes containing pyranine alone or pyranine plus atRP, as in panel A. C, Rpe65-dependent uptake of atRP into liposome membranes revealed by atRP-BODIPY-PC FRET. Liposomes containing protein from Rpe65expressing or non-expressing Sf9 cells plus BODIPY-PC were incubated in media containing free atRP for the indicated times. 520-nm emission was acquired with 325-nm excitation. D, Rpe65-dependent uptake of atRP into the aqueous interior of liposomes revealed by atRP-pyranine FRET. Liposomes containing protein from Rpe65-expressing or nonexpressing Sf9 cells plus pyranine were incubated in media containing free atRP for the indicated times. 510-nm emission was acquired with 325-nm excitation. E, fluorescence light microscopy of liposomes containing protein from non-expressing Sf9 cells incubated in medium containing free atRP. The excitation and detection conditions permitted direct visualization of atRP. Note the ring-like fluorescent labeling of the liposome, suggesting membrane localization of atRP. Also note the fluorescent lipid droplets in the medium due to insolubility of atRP. F, fluorescence microscopy of liposomes containing protein from Rpe65expressing cells incubated in medium containing free atRP, as in panel E. Note the more uniform fluorescent labeling indicating atRP in both the membrane and aqueous interior of this liposome. The apparent fluorescent intensities of the liposome images in panels E and F do not reflect the absolute levels of atRP. rpe65 Ϫ/Ϫ knockout mice (10). After extracting to remove endogenous retinoids, the membranes were solubilized in detergent and assayed for production of 11cROL from exogenous atRP substrate plus or minus expressed Rpe65. Representative chromatograms of retinoids formed during these reactions are shown in Fig. 4, A-D. As in previous experiments, we confirmed identification of the retinoids by spectral analysis (insets in Fig.  4, A-C) and co-migration with standards (Fig. 4E). Fig. 4F shows the quantitation of 11cROL synthesized under the four assay conditions. Thirteen pmoles of 11cROL were produced by the solubilized RPE membranes from wild-type mice. Addition of exogenous Rpe65 to these solubilized membranes increased synthesis of 11cROL to 21 pmol. No 11cROL was produced by solubilized RPE membranes from rpe65 Ϫ/Ϫ mice. Importantly, addition of exogenous Rpe65 to solubilized rpe65 Ϫ/Ϫ membranes fully restored the IMH activity (19 pmol of 11cROL produced). We also measured the effect of Rpe65 on IMH activity in proteoliposomes prepared from detergent-solubilized bovine RPE. First, we induced fusion of these proteoliposomes with a second preparation of liposomes that contained only atRP. Here, 1.3 pmol of 11cROL were produced (Fig. 4G). A similar amount of 11cROL was produced (1.4 pmol) when the atRP liposomes were replaced with liposomes that contained atRP plus non-expressing Sf9 proteins. However, dramatically higher levels of 11cROL were produced (11.3 pmol) when the bovine RPE proteoliposomes were fused with liposomes that contained atRP plus Rpe65 (Fig. 4G). The small amounts of 11cROL produced in the fusion assay with liposomes containing atRP alone or atRP plus non-expressing Sf9 proteins are probably due to endogenous Rpe65 carried-through during preparation of the bovine RPE proteoliposomes. A small amount of 13-cis-retinol (13cROL) was also detected in each assay where 11cROL was produced (Fig. 4, A, B, and D). In a control experiment, we added 11cROL to heat-denatured RPE proteoliposomes. Following incubation, a similar relative amount of 13cROL (plus atROL) was produced, suggesting that the 13cROL in the Fig. 4 experiments is a thermal degradation product of 11cROL.
The results in Fig. 4 establish that IMH activity is dependent upon the presence of Rpe65. To rule out the possibility that Rpe65 is IMH, we incubated liposomes containing expressed Rpe65 plus atRP or atROL and analyzed for production of 11cROL by HPLC. No 11cROL was produced in either assay. We also incubated heat-denatured membranes from wild-type mouse RPE with atRP and Rpe65. Here again, no measurable 11cROL was produced. These data indicate that Rpe65 has no intrinsic isomerase activity. DISCUSSION In this study we demonstrated that Rpe65 binds atRE. We showed spectrally that apo-Rpe65 and atRP combine to yield a FIG. 4. Stimulation of IMH activity in mouse and bovine RPE by Rpe65. A, HPLC chromatogram at 318 nm of retinoids synthesized following incubation of solubilized RPE membranes from wildtype (C57BL/6) mice with atRP but without added Rpe65. Retinoid peaks are labeled. Elution time in min are indicated in panel E. Inset shows the UV spectrum acquired for the 11cROL peak. B, chromatogram of retinoids synthesized following incubation of solubilized RPE membranes from wild-type (C57BL/6) mice with atRP plus 100 g Rpe65. Inset shows the UV spectrum of 11cROL. C, chromatogram of retinoids synthesized following incubation of solubilized RPE membranes from rpe65 Ϫ/Ϫ mice with atRP but without added Rpe65. D, chromatogram of retinoids synthesized following incubation of solubilized RPE membranes from rpe65 Ϫ/Ϫ mice with atRP plus 100 g Rpe65. Inset shows the UV spectrum for 11cROL. E, chromatogram of authentic retinoid standards. F, histogram showing pmol of 11cROL produced from atRP by solubilized mouse RPE membranes under the indicated assay conditions. G, histogram showing pmol of 11cROL produced by bovine RPE proteoliposomes from atRP plus non-expressing Sf9 proteins or expressed Rpe65. new molecular species. We showed that atRP co-elutes with Rpe65 during gel filtration chromatography and immunoprecipitation. Also, we showed that liposomes containing Rpe65 take up free atRP from the medium much more rapidly than do control liposomes, and that a significant fraction of the atRP in Rpe65-containing liposomes is found within the aqueous interior. Significantly, we demonstrated that the conversion of atRP to 11cROL by IMH is dependent on Rpe65, but that Rpe65 alone does not catalyze this isomerization reaction.
Rpe65 is 37% identical to ␤-carotene-15,15Ј-dioxygenase (␤-CDO), which catalyzes the conversion of ␤-carotene to atRAL (29,30). Given this level of identity, the ligand-binding pocket of Rpe65 and the catalytic site of ␤-CDO may be conserved structural features. Fig. 5 shows the molecular structures for atRP, ␤-carotene, and atROL. Despite the similarity of these molecules, we observed no spectral change upon mixing ␤-carotene with Rpe65, suggesting that Rpe65 does not bind ␤-carotene. Also, Rpe65 does not possess ␤-CDO catalytic activity (31). The homology between ␤-CDO and Rpe65 makes sense considering that both proteins interact with all-trans-retinoids.
To explore the functional significance of atRP binding by Rpe65, we analyzed the uptake of atRP into liposomes containing proteins from Rpe65-expressing and non-expressing Sf9 cells. Since the Sf9 proteins were mixed with phospholipids before extrusion, Rpe65 was present on both inner and outer leaflets of the liposomal membrane. We delivered atRP to the liposome-containing media in a small volume of water-miscible solvent (DMF), which resulted in fine dispersal of atRP as a dilute emulsion. We used a fluorescent analog of phosphatidylcholine, BODIPY-PC, and a water-soluble fluorescent probe, pyranine, to report the presence of atRP within the lipid bilayer and aqueous lumina of liposomes, respectively. We determined the Rpe65 dependence of atRP uptake by comparing the strength of fluorescent emission from liposomes containing Rpe65 versus a similar quantity of Sf9 control proteins. The FRET signal in the Rpe65-containing liposomes compared with control liposomes was ϳ6-fold higher with both BODIPY-PC and the pyranine probes. These ratios actually underestimate the effect of Rpe65 on atRP uptake into liposomes, since the BODIPY-PC and pyranine contribute direct fluorescence to the 520-nm or 510-nm FRET emissions (Fig. 3, A and B) in both Rpe65 and Sf9 control liposomes. Thus, Rpe65 strongly stimulated the uptake of atRP into the liposomal membranes and effected its transfer to the aqueous interior.
We exploited the fluorescent properties of atRP to visualize it directly in liposomes by fluorescence microscopy. In the control experiment, lipid droplets of atRP were visible in the media (Fig. 3E), resembling the atRE inclusions in RPE cells of rpe65 Ϫ/Ϫ mice (10) and dogs (32). We never observed these  (46,47) synthesizes atRE from atROL delivered by CRBP (33). These atRE, which are insoluble in water (25,26), distribute in the lipid bilayer. Rpe65 functions to extract insoluble atRE from the membrane and present them to IMH. IMH effects the conversion of atRE to 11cROL (1,3,48). Finally, 11cROL is oxidized to 11cRAL by 11cROL dehydrogenase (11cRDH) in ER membranes (49). 11cRAL is bound to CRALBP in the RPE cytoplasm (24). LRAT, Rpe65, 11cRDH, and CRALBP have been shown to associate with one another by immunoprecipitation (50). For this reason, the proteins are depicted as components of a multimeric complex or visuosome. B, visual cycle in rpe65 Ϫ/Ϫ RPE cells. Synthesis of atRE by LRAT is normal in rpe65 Ϫ/Ϫ mutants (37). IMH is also present at normal levels in these mutants. However, without Rpe65 to solubilize and present atRE, IMH is starved for substrate and hence no measurable 11cROL is produced. Instead, atRE accumulate in the ER membrane and bleb off as lipid droplets. Without 11cROL, 11cRDH is also starved for substrate, and hence no measurable 11cRAL chromophore is produced. A similar process likely occurs in patients with LCA. droplets in the media of Rpe65-containing liposomes (Fig. 3F). Thus, Rpe65 on the outer surfaces of liposomes may interact with these lipid droplets to extract atRP. Rpe65 may also dissociate from the liposome membranes to scavenge atRP from these lipid droplets. It is likely that these atRP-containing droplets acquire a skin of phospholipids from the medium, and hence resemble to Rpe65 the surface of a phospholipid membrane at its aqueous interface. Mobilization of atRP in these lipid droplets explains the higher levels of atRP in membranes from Rpe65-containing liposomes (Fig. 3C). Gene therapy of rpe65 Ϫ/Ϫ dogs with recombinant virus expressing normal Rpe65 caused disappearance of the atRE inclusions from the treated eye (32). Thus, Rpe65 can mobilize atRE from these lipid inclusions in vivo. The presence of atRP in the aqueous interior of liposomes in the current study (Fig. 3, D and F) is further evidence that Rpe65 extracts atRP from membranes.
Multiple proteins have been characterized that bind isomers of retinol and retinaldehyde including CRALBP (24), cellular retinol-binding protein (33), retinol-binding protein (34), and interphotoreceptor retinoid-binding protein (35). However, no binding proteins for retinyl esters have previously been described. The listed binding proteins are thought to stabilize their retinoid ligands and to prevent them from reacting with cellular components. Retinyl esters, however, are relatively inert. What then is the need for an atRE-binding protein such as Rpe65? Retinyl esters represent the principal storage pool of vitamin A (11,36), and comprise the sole substrate for chemical synthesis of visual chromophore in the RPE (1, 2). The insolubility of atRE in water infers the requirement for a binding protein to extract them from the lipid bilayer and deliver them in a solubilized form to IMH.
Solubilized RPE membranes from rpe65 Ϫ/Ϫ mice contained no measurable IMH activity (Fig. 4, C and F). However, we observed virtually complete rescue of this biochemical phenotype with addition of Rpe65 to the assay mixture (Fig. 4, D and  F). Similarly, we observed ϳ9-fold stimulation of IMH activity in bovine RPE proteoliposomes with addition of Rpe65 (Fig.  4G). The lack of stimulation in the non-expressing Sf9-control experiment shows that the stimulation effect seen with Rpe65 is not caused by extraneous Sf9 proteins. A possible interpretation of the data in Fig. 4 is that Rpe65 and IMH are the same protein. However, no 11cROL was produced when Rpe65 was incubated with liposomes containing atRP or atROL alone, or with atRP plus heat-denatured RPE proteins from wild-type mice. These observations rule out the explanation that Rpe65 and IMH are the same protein. It remains a formal possibility that Rpe65 is a subunit of a larger IMH complex. The abundance of Rpe65 in RPE cells argues against this possibility. A more likely explanation for the critical dependence of isomerase activity on the presence of Rpe65 is that this protein is required for substrate access by IMH.
Our model for the function of Rpe65 is shown in Fig. 6A. We suggest that the role of Rpe65 is to extract insoluble atRE from ER membrane and present them to IMH. This hypothesis explains the phenotype in rpe65 Ϫ/Ϫ mice (Fig. 6B). Synthesis of atRE by LRAT is normal in these mutants (37). However, without Rpe65 to solubilize atRE, IMH is starved for substrate and hence no measurable 11cROL is produced. Instead, atRE accumulate at up to 1,000-fold the level in wild-type RPE due to decreased utilization (38). The atRE saturate the ER membrane and bleb off as lipid droplets, which float above the cytosol following centrifugation (38) and are visible in rpe65 Ϫ/Ϫ RPE by electron microscopy (10).
Rpe65 is also present in cone photoreceptors, albeit at much lower levels than in RPE cells (39,40). Cone-dominant chickens and ground squirrels have been shown to contain 11cRE and atRE in retina in addition to RPE (17,41,42). The cellular localization of these retinyl esters within the retina is not known. Recently, a new retinoid pathway that mediates conepigment regeneration under daylight conditions was described in chicken and ground squirrel retinas (17). Indirect evidence presented in that study suggests that these esters predominantly accumulate in Mü ller glial cells. However, it is possible that cones also contain low levels of atRE, which may be solubilized by endogenous Rpe65. The presence of Rpe65 in cones may explain the loss of cone function in rpe65 Ϫ/Ϫ mice (43,44).
Recently, Rpe65 was shown to react covalently with a biotinylated derivative of all-trans-retinyl chloroacetate (45). This reaction was competed by addition of atROL and all-transretinyl acetate. Although competition with fatty acyl esters of atROL such as atRP was not investigated, these data agree with the results presented here.
In summary, we have shown that Rpe65 is a binding protein for atRE. This binding function serves to solubilize atRE, which are otherwise confined to membranes and lipid inclusions. We have also shown that Rpe65 is required for IMH activity but has no intrinsic isomerase activity, suggesting that Rpe65 donates atRE substrate to IMH.