FATP1 Inhibits 11-cis Retinol Formation via Interaction with the Visual Cycle Retinoid Isomerase RPE65 and Lecithin:Retinol Acyltransferase*

The isomerization of all-trans retinol (vitamin A) to 11-cis retinol in the retinal pigment epithelium (RPE) is a key step in the visual process for the regeneration of the visual pigment chromophore, 11-cis retinal. LRAT and RPE65 are recognized as the minimal isomerase catalytic components. However, regulators of this rate-limiting step are not fully identified and could account for the phenotypic variability associated with inherited retinal degeneration (RD) caused by mutations in the RPE65 gene. To identify new RPE65 partners, we screened a porcine RPE mRNA library using a yeast two-hybrid assay with full-length human RPE65. One identified clone (here named FATP1c), containing the cytosolic C-terminal sequence from the fatty acid transport protein 1 (FATP1 or SLC27A1, solute carrier family 27 member 1), was demonstrated to interact dose-dependently with the native RPE65 and with LRAT. Furthermore, these interacting proteins colocalize in the RPE. Cellular reconstitution of human interacting proteins shows that FATP1 markedly inhibits 11-cis retinol production by acting on the production of all-trans retinyl esters and the isomerase activity of RPE65. The identification of this new visual cycle inhibitory component in RPE may contribute to further understanding of retinal pathogenesis.

The isomerization of all-trans retinol (vitamin A) to 11-cis retinol in the retinal pigment epithelium (RPE) is a key step in the visual process for the regeneration of the visual pigment chromophore, 11-cis retinal. LRAT and RPE65 are recognized as the minimal isomerase catalytic components. However, regulators of this rate-limiting step are not fully identified and could account for the phenotypic variability associated with inherited retinal degeneration (RD) caused by mutations in the RPE65 gene. To identify new RPE65 partners, we screened a porcine RPE mRNA library using a yeast two-hybrid assay with full-length human RPE65. One identified clone (here named FATP1c), containing the cytosolic C-terminal sequence from the fatty acid transport protein 1 (FATP1 or SLC27A1, solute carrier family 27 member 1), was demonstrated to interact dose-dependently with the native RPE65 and with LRAT. Furthermore, these interacting proteins colocalize in the RPE. Cellular reconstitution of human interacting proteins shows that FATP1 markedly inhibits 11-cis retinol production by acting on the production of all-trans retinyl esters and the isomerase activity of RPE65. The identification of this new visual cycle inhibitory component in RPE may contribute to further understanding of retinal pathogenesis.
In vertebrates, vision begins in photoreceptors (rod and cone) with the absorption of light by the visual pigments, rhodopsin and cone opsins, which consist of two components: opsin (apoprotein) and 11-cis retinaldehyde (11cRAL, 2 chro-mophore). Light causes photoisomerization of 11cRAL to alltrans retinal (atRAL) that dissociates from opsin. The atRAL is reduced to all-trans-retinol (atROL, vitamin A), which is in turn converted to 11-cis retinol (11cROL) and oxidized to 11cRAL in the neighboring retinal pigment epithelium (RPE). The whole process involves both retinoid transport proteins and enzymes and is termed the visual (retinoid) cycle (1)(2)(3).
In the RPE, atROL is first esterified in the smooth endoplasmic reticulum membrane by a lecithin:retinol acyltransferase, LRAT (4,5), to fatty acids to form all-trans-retinyl-esters (atRE). The latter are recognized by the RPE-specific protein RPE65 (MIM 180069) that catalyzes their cleavage and isomerization to the 11cROL (6,7). 11cROL is then oxidized to 11cRAL by the 11cROL dehydrogenase (11cRDH), a member of the short chain alcohol dehydrogenases (8). Cellular retinaldehyde-binding protein (CRALBP) is an abundant carrier of both 11cROL and 11cRAL that facilitates the 11cROL formation and its oxidation to 11cRAL (9,10).
Isomerization is a rate-determining step in the visual cycle. In mice, the level of RPE65 expression is strain-dependent and determine the rate-limited rhodopsin regeneration (11,12). Recently, in vitro assays have shown that multiple disease-associated mutations in human RPE65 shown to decrease protein concentration, directly affect the isomerase activity (13,14). This rate-determining step may be regulated. For example, phosphate-containing compounds, such as ATP and GTP, stimulate the isomerase but have no influence on LRAT activity (15). In contrast, 11cROL is a specific inhibitor of isomerase activity (16).
Protein interactions with RPE65 may also alter the isomerase activity. For example, RPE65 have been shown to interact with 11cRDH and the retinal G protein-coupled receptor RGR (8,17). RGR belongs to the opsin family and participates in the regeneration of 11cRAL in the RPE (18). Using knock-out mice, RGR was shown to enhance the isomerase activity of RPE65 independently of light (19). Later, RGR was demonstrated to inhibit LRAT activities and to mediate light-dependent translocation of atRE for synthesis of visual chromophore (20). These data established that RGR plays a regulatory role in the visual cycle.
Looking for potential protein partners of the isomerization process using a two-hybrid screening, we found that RPE65 interacts with the fatty acid transport protein 1, FATP1, also named SLC27A1, solute carrier family 27 (fatty acid transporter), member 1. Furthermore, we show that FATP1 reduces 11cROL production in cellular models, affecting level of atRE generated by LRAT activity and inhibiting isomerase activity without affecting level of RPE65 protein. These data raise the possibility that FATP1 alleles may modify RD phenotypes caused by defects in the visual cycle.
Two-hybrid Library and Bait Construction-The two-hybrid library was prepared using CDS III random-primer to prime poly(A) ϩ RNA isolated from porcine RPE following the MATCHMAKER library construction and screening kit instructions. To use human RPE65 protein and fragments (see supplemental materials for construction) as baits, cDNA was ligated in-frame with GAL4 DNA binding domain into pGBKT7 DNA-BD cloning vector to transform the yeast reporter strain, AH109 (Saccharomyces cerevisiae). Human LRAT and tLRAT, which were also used as baits were cloned from retina and constructed according to Ref. (21), respectively.
Yeast Two-hybrid Analysis-The interaction between RPE65 and library-encoded proteins was phenotypically detectable on nutritionally deficient agar plates containing SD/Dropout (DO) medium, a combination of a Minimal SD Base combined and a DO supplement lacking leucine (LEU2 reporter gene), tryptophan (TRP1), adenine (ADE2), and histidine (HIS3) and with the secreted ␣-galactosidase activity (MEL1). Expression vector pGADT7-Rec containing RPE65-interacting proteins were iso-lated and the entire cDNA inserts were sequenced. In silico analysis was performed with in-frame sequences to identify genes. To eliminate false positives, relevant clones were tested again by co-transformation of AH109 yeast with either pGBKT7-RPE65 or pGBKT7-LamC or empty pGBKT7 vectors.
RNA Extraction and RT-PCR Expression Analysis-Porcine tissues were purchased from INRA Rennes (UMR SENAH, Saint-Gilles, France). Porcine retina and RPE were prepared as described below. Total RNAs were collected with RNAxel kit and mRNAs were then purified with Oligotex kit following manufacturer's instructions. 500 ng of each mRNA pool were reverse-transcribed in a 20-l reaction mixture containing 250 ng of random primer and 200 units of Superscript II reverse transcriptase at 42°C for 60 min. One microliter of the cDNA was then amplified in a 20-l PCR using genespecific primers and 2 units of Taq polymerase for 25-30 cycles. The 503-bp RPE65 product was amplified using the primers forward 5Ј-CTGCAGTGACCGATTCAAGC-CATC-3Ј and reverse 5Ј-CACTGCACAGAATTGCAGTG-GCAG-3Ј; the 500-bp FATP1 product was amplified with the primers forward 5Ј-ATGCTGGACCTTCGCACAGCT-GGA-3Ј and reverse 5ЈAATGCGGTAGTACCTGCTGTG-CAC-3Ј; the 300-bp GAPDH product was amplified with the primers forward 5Ј-CCCTGCAAATGAGCCCCAGC-CTT-3Ј and reverse 5Ј-TTGGTCGTATTGGGCGCCTG-GTCA-3Ј. Buffer or genomic DNA contaminations were assessed in all assays by PCR without cDNA or reverse transcriptase. PCR products were analyzed in 2% ethidium bromide-agarose, then purified with a Wizard SV gel kit and sequenced.
GST Pull-down Assay-The FATP1c nucleotide sequence isolated from the two-hydrid contains the native TGA stop codon and untranslated sequence. The full-length was subcloned into pGEX-4T1 vector using EcoRI and XhoI restriction sites. To produce glutathione S-transferase (GST) and GST-FATP1c fusion proteins, Escherichia coli BL21 cells were transformed with pGEX-4T1 plasmids and growth at 30°C for 3-4 h in 2xYT medium with 100 g/ml ampicillin and 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside. Bacteria were sonicated 6 times for 15 s on ice in 1.5 ml BBIP buffer (phosphate-buffered saline with 5% glycerol, 5 mM MgCl 2 , 0.1% Triton X-100, and protease inhibitor mixture). The extracts were incubated with 1% Triton X-100 for 1 h at 4°C and centrifuged twice at 12,000 ϫ g for 15 min. Aliquots of supernatants were incubated with 50 l of glutathione-Sepharose beads for 1 h at 4°C and washed twice with 200 l of ice-cold BBIP with 150 mM NaCl to isolate GST proteins. RPE homogenate (1.3 mg/ml protein) was solubilized in BBIP supplemented with 0.5% Triton X-100 for 1 h at 4°C and centrifuged at 20,000 ϫ g for 20 min. 50 g of RPE protein were incubated with GST alone or GST-FATP1c bound to glutathione-Sepharose beads overnight at 4°C. Beads were washed five times with 200 l of ice-cold BBIP-150 mM NaCl-0.1% Triton X-100 and resuspended in Laemmli sample buffer. An equal amount of bound proteins was separated in each lane by SDS-PAGE and analyzed by Western blot using anti-RPE65 (1:5,000) and anti-GST (1:100,000), then detected with horseradish peroxidase-conjugated IgG (1:10,000) and enhanced chemiluminescence reagents.

FATP1 Inhibits 11-cis Retinol Formation
FATP1 Antibody and Western Blotting-A rabbit polyclonal FATP1 antibody was produced against GST-FATP1c purified on glutathione beads and dialyzed against phosphate-buffered saline. Antiserum was affinity-purified onto GST-FATP1c-glutathione-Sepharose column and used for Western blot analysis. Samples were solubilized in Laemmli sample buffer and ran in 10% SDS-PAGE. The proteins were transferred to polyvinylidene difluoride membranes, incubated overnight at 4°C with specific primary antibodies (1:1,000), subjected to alkaline phosphatase-conjugated IgG (1:5,000) and revealed with BCIP/ NBT-purple liquid substrate.
For the immunoprecipitation assays, adherent cells were grown 3 days post-infection, and lysed into 50 mM Tris, pH 8.0 with protease inhibitor mixture and different detergent (IGEPAL CA-630, radioimmune precipitation assay buffer) and salt (0 -150 mM NaCl) compositions to optimize the lysing conditions. Immunoprecipitation was performed according to instructions of the immunoprecipitation starter pack using antibodies against FATP1c and RPE65. The immune complexes were precipitated with protein G-Sepharose, washed several times and separated by SDS-PAGE, followed by Western blotting.
Porcine RPE Cell Isolation and Subcellular Fractionation-Fresh porcine eyes obtained from local slaughterhouse were opened on ice and the neuroretina removed from eyecups. The RPE was gently brushed and centrifuged for 15 min at 1170 ϫ g. The pellet was resuspended in 150 mM NaCl, 10 mM Tris-HCl, pH 7.4, and homogenized in a Teflon-glass potter. Unbroken cells, nuclei, mitochondria, and heavy membranes were pelleted at 30,000 ϫ g for 20 min. The supernatant was ultracentrifuged at 105,000 ϫ g for 1 h to separate the cytosolic fraction from the pelleted microsomal fraction. The protein concentration was determined with a BCA protein assay kit. Similar subcellular fractionation was performed onto 293 and Sf9 cells.
Immunofluorescence Analysis of Cells and Retina-Human RPE65 cDNA was fused to N-terminal enhanced green fluorescent protein (EGFP) gene in pEGFP-C1 vector. EGFP-RPE65 was stably expressed into 293 cells (293-R) using Lipofectamine 2000 transfection method (Invitrogen) followed by Geneticin selection. In the meantime, 293 cells stably expressing ECFP-LRAT and EGFP-RPE65 (293-LR) were generated with the same method. GFP-positive cells were sorted by flow cytometry. A pCMV-FLAG-FATP1 (see supplemental materials for construction) was transiently transfected in 293-R cells using Lipofectamine 2000, according to the manufacturer's procedures. Cells were grown on FluoroDish (FD35-100, World Precision Instruments, UK) for an additional 24 -48 h after transfection in Dulbecco's modified Eagle's medium with 10% heat-inactivated fetal bovine serum plus penicillin-streptomycin and fixed with 4% paraformaldehyde.
Immunocytochemistry was then performed according to Ref. 22 using anti-FLAG (1:1000). Confocal imaging was per-formed with a Zeiss LSM 5 LIVE DUO system. The image acquisition was obtained with the Zeiss LSM AIM software. For immunohistochemistry, C57BL/6J eyes were rapidly removed and fixed in 4% PFA, 24 h at 4°C. Eyes cups were embedded in 5% low-melting agar and cut into 50-m sagittal sections. Sections were permeabilized with 0.1% SDS, saturated 20 min with 10% fetal calf serum, and incubated overnight with the primary antibodies (rabbit anti-FATP1, 1/250; mouse anti-RPE65, 1/500). The secondary antibodies were diluted in phosphatebuffered saline (Alexa568-conjugated anti-rabbit, Alexa488conjugated anti-mouse) and incubated 1 h. Sections were then mounted in Fluorsave TM Reagent (Calbiochem).
Isomerase Assay in Living Cells-This assay was done as described previously (23). Briefly, the 293T-LC cells transfected with expression vectors for RPE65 and FATP1 were incubated with 5 M atROL in the medium for indicated times and lysed in 0.5 ml of 10 mM HEPES 0.2% SDS. During 293 cell transfection, pRK5 (empty expression vector) was added for single protein expression to normalize plasmid amount. Proteins (10 g) were analyzed by immunoblot using RPE65 antiserum. Esters were saponified with 0.5 mM KOH at 55°C for 15 min. Retinoids were extracted with hexane and analyzed by high performance liquid chromatography (HPLC). Identified peaks were confirmed by spectral analysis and coelution with authentic retinoid standards.
In Vitro Isomerase Assay-The Sf9 cells were harvested at 2 days post-infection with baculoviruses expressing FATP1, RPE65, LRAT, and GFP (irrelevant protein), homogenized in 40 mM HEPES buffer (pH 7.5). The isomerase assay mixtures contained 20 mM HEPES, 100 mM NaCl, 6% bovine serum albumin, 10 M all-trans retinyl palmitate (atRP), 6.0 mM sodium cholate, and 1 mg of protein. After incubation for 30 min, 1 h, 1.5 h, and 2 h in the dark at 37°C, the reactions were quenched by adding 0.2% SDS and 2 volumes of methanol. Retinoids were analyzed as described above.
LRAT Assay in Living Cells-293T cells in 12-well plates were transfected with pRK5 and LRAT or FATP1-and LRAT-expressing plasmids. At 36 h post-transfection, cell medium were replaced with fresh medium containing 5 M atROL and incubated for 15 min, 30 min, and 1 h in the dark at 37°C. 293-LR cells in 6-well plates were transfected with pRK5 or pCMV-FLAG-FATP1. At 36 h post-transfection, cell medium were replaced with fresh medium containing 10 M atROL and incubated for 0.5, 1, 3, 6, and 18 h in the dark at 37°C. After washing with phosphate-buffered saline, the cells were pelleted by lowspeed centrifugation and lysed before retinoid analysis.
In Vitro LRAT Assay-The Sf9 cells were harvested at 2-3 days post-infection with baculoviruses expressing FATP1 and/or LRAT, homogenized in 10 mM bis Tris propane-HCl buffer (pH 7.5), 100 mM NaCl. The LRAT assay mixtures con-tained 10 -20 M atROL, 0.5% bovine serum albumin, and 200 g of protein. After incubation for different times (as indicated) in the dark at 37°C, the reaction was stopped, and retinoids were analyzed as described above.

RESULTS
Yeast Two-hybrid Screen and Identification of FATP1 as an RPE65-interacting Protein-From pig RPE we constructed yeast two-hybrid GAL4 AD/library, which was used to transform AH109 yeast expressing a GAL4 DNA-BD/human RPE65 fusion protein. GAL4 DNA-BD/RPE65 alone failed to activate transcription of the HIS3 and LacZ reporter genes, indicating that RPE65 does not function as a transcriptional activator and can be used as bait (data not shown). A total of 1.2 ϫ 10 6 individual clones were screened by growth selection of yeast colonies. Among these, 112 positives clones were identified. Additional controls and sequence analysis allowed us to further identify 12 (11%) true positive clones. One of these clones contained a cDNA encoding part (the C-terminal 307 amino acids) of porcine FATP1, known to play a role in fatty acid metabolism. The 307-amino acid fragment (herein referred to as FATP1c) is highly conserved among mammalian species and notably shows 95% identity with its human ortholog (supplemental Fig. S1A). Importantly, based on the proposed membrane topology model of FATP1 by Lewis et al. (24), the main part of FATP1c projects into the cytosol while its N-terminal part is embedded in the lipid bilayer membrane (supplemental Fig. S1B).
We evaluated the strength of these interactions by assessing yeast densities from colonies expressing the Adeϩ/Hisϩ/ Leuϩ/Trpϩ phenotype grown to mid-log phase (Fig. 2B). The truncated RPE65 fragments and tLRAT showed a significantly stronger interaction than full-length RPE65. Interestingly, yeast transformed with RPE65-N147 grew significantly faster (up to 50% increase) than those expressing other fragments and reached at least 80% of the standard value, suggesting that the N terminus of RPE65 was sufficient to achieve the highest interaction. In any case, yeast growth differences were not consistent with expression variability among the polypeptides (Fig. 2C).
Isolation of FATP1-RPE65/LRAT Protein Complexes-To confirm these interactions, a GST-FATP1c fusion protein purified with glutathione-Sepharose beads was incubated with detergent-solubilized porcine RPE proteins (Fig. 2A). The native porcine RPE65 was retained in a dose-dependent man-ner by GST-FATP1c but not by the GST protein regardless of concentration. However, we could not attribute any relative affinity for RPE65 because the GST fusion protein showed variable binding efficiencies that remained low when compared with input RPE65 signal. We then evaluated recombinant protein interaction by co-immunoprecipitation (Co-IP) experiments carried out with Sf9 cells co-expressing human FATP1, RPE65, LRAT, and CRALBP (Fig. 2B). Using increasing amounts of anti-FATP1, FATP1 co-precipitated with both RPE65 and LRAT but not with CRALBP. The reverse Co-IP with anti-RPE65 demonstrated that both FATP1 and CRALBP specifically interact with RPE65 but that LRAT did not. We concluded that FATP1 does biologically interact with RPE65 and LRAT, and that these interactions are independent of each other.

FATP1 Expression and Co-localization with RPE65 in the
Retinal Pigment Epithelium-We first examined the distribution of FATP1 in pig tissues by semi-quantitative RT-PCR using tissue-specific poly(A)ϩ RNA. FATP1 transcripts were detected in various tissues (Fig. 3A). The highest expression level was found in the RPE (strong and specific RPE65 expression served as a control). High levels were also found in heart, brain, and skeletal muscle whereas lung, kidney, liver, and skin showed lower levels. No expression was detected in intestine and thymus. FATP1 transcripts were also detectable in both neural retina and RPE (data not shown).
The protein expression analysis by immunoblotting with the FATP1c antibody showed an immunoreactivity at 63 kDa in tissues where FATP1 is known to be well expressed, such as white adipose tissue, skeletal muscle, and heart. The immunoreactivity was lower in liver and absent in intestine. An immunoreactive band was also detected in RPE/retina (Fig. 3B).
We further investigated the FATP1 subcellular distribution in porcine RPE cells. Separation of membranes into pellet (P; nucleus, mitochondria, lysosomes, cytoskeleton), microsomes (M; ER and plasma membrane) and cytosol (C) by differential centrifugations and subsequent Western blotting demonstrated the presence of a 63-kDa FATP1 band that, like RPE65 and LRAT, was localized in microsomes (Fig. 3C). Weak FATP1 and RPE65 bands were also detected in the P fraction (30,000 ϫ g pellet). In contrast, CRALBP was found in the C and P fractions.
Lastly, Fatp1 expression was studied in the mouse retina using an antiserum against residues 628 -640 of the murine

FATP1 Inhibits 11-cis Retinol Formation
Fatp1 sequence. This antibody recognized a 63-kDa band in both RPE and neuroretina by Western blotting (Fig. 3D). As shown in Fig. 3E, Fatp1 was observed mainly in the RPE monolayer, and at a lower level in the ganglion cell layer, but nowhere else in the neuroretina. In the RPE, Fatp1 co-localized with RPE65 except in the apical microvilli where only Fatp1 immunoreactivity was detected.
FATP1 Inhibits RPE65-mediated 11-cis Retinol Production-We evaluated the impact of FATP1 on the RPE65 and LRAT activities. 293T cells stably expressing LRAT and CRALBP (293T-LC) were transiently transfected with plasmids highly expressing RPE65 or FATP1 alone or RPE65 plus FATP1. Following addition of 5 M atROL to the culture medium, the synthesis of 11cROL was measured over the time (Fig. 4A). In the presence of RPE65, 11cROL was detectable as early as 30 min after addition of atROL and increased within the next 2 h. In contrast, the cells which only expressed FATP1 did not show any isomerase activity. When RPE65 and FATP1 were co-expressed, the formation of 11cROL was lower at each time point. Importantly, the expression levels of RPE65 in all conditions were similar (Fig. 4B), indicating that the decreased isomerase activity in the cells transfected with RPE65 plus FATP1 was not due to altered expression levels of RPE65. Because FATP1 interacts dosedependently with RPE65 in protein-protein interaction assays, we looked for a dose-dependent effect on RPE65 isomerase activity. Fig. 4C clearly demonstrated that the level of inhibition depends on the FATP1 expression in the 293T-LC. Finally, to prove the co-localization of FATP1 with RPE65 in 293 cells, we stably expressed an EGFP-tagged human RPE65 in 293 cells and transiently transfected a FLAG-tagged human FATP1. We detected green fluorescent with a punctuated distribution in the cytoplasm that superimposed with an ERtracker (data not shown). In some cells, we revealed FATP1 expression with an anti-FLAG antibody (red fluorescence) and in a few of them, the confocal analysis demonstrated a perfect co-localization of FATP1and RPE65-tagged proteins (white merge, Fig. 4D).
The decrease of the 11-cis production could result from an action of FATP1 on LRAT or RPE65 or both. To demonstrate a direct inhibition of RPE65, homogenates prepared from baculovirus-infected Sf9 cells that expressed RPE65, FATP1, or both were incubated with atRP, the isomerase substrate (Fig. 5A). The kinetics of 11cROL production in Sf9 was similar to that of 293T-LC: FATP1 markedly inhibited the synthesis of 11cROL from atRP in the presence of RPE65, the expression of which remained unchanged (Fig. 5B). In addition, we checked that RPE65 and FATP1 co-localized in the same membrane fraction from co-infection of Sf9 cells (Fig. 5C). Therefore, we concluded that FATP1 directly inhibits the isomerase activity of RPE65.
FATP1 Decreases atRE Production-To measure LRAT activity, atRE production was followed in 293T-LC cells during 1 h of atROL incubation. Under this condition, FATP1 expression did not affect atRE production (Fig. 6A). The next experiment was designed in 293 cells stably expressing LRAT and RPE65 (293-LR) incubated with 4 nmol atROL and atROL and atRP were monitored for 18 h to allow the reaction to reach a plateau. The results of the experiment, shown in Fig. 6B, demonstrate that atROL is rapidly esterified within the first hour but reached at steady-state phase after 6 h at a value close to the accumulation of atRP, a characteristic of a reaction at equilibrium. In the presence of FATP1, the cellular uptake of atROL was increased but the amount of atRP produced was decreased after 3 h of incubation. Thus, it is apparent that FATP1 does not affect the initial activity of LRAT but enhances the transition to the steady-state condition in which the amount of atRP does not appear to change.
To confirm this finding, Sf9 cells were also used to express LRAT, FATP1, or both. In the presence of LRAT, the addition of atROL resulted in the production of retinyl esters. The amount of esters formed after 1 h of incubation decreases slowly with decreasing the level of LRAT present in this experiment (Fig. 7A). The kinetics of accumulation of retinyl esters showed that at high level of LRAT, the atROL was very quickly and fully converted into retinyl esters (Fig. 7B). In contrast, with a much lower level of LRAT (L/5), the synthesis of retinyl ester reached a plateau at 30 min of reaction, and all atROL was not esterified. As shown in Fig. 7C, co-expression of FATP1 under similar conditions did not alter the rate of retinyl ester synthesis at a high level of LRAT but significantly decreased this rate (16.5 Ϯ 2.1%) after 30 min of reaction when the level of LRAT is reduced by four. Co-expression of FATP1 itself did not affect the expression of LRAT (Fig. 7D). Finally, we asked whether the substrates of FATP1, the palmitate, CoA, and ATP could modify the observed effect on the synthesis of retinyl esters. In the series of experiments, the dampening effect of FATP1 was very marked (60.6 Ϯ 5.8% reduction). In these conditions, the substrates did not alter the activity of LRAT (Fig. 7E). As for RPE65, we verified that LRAT was localized with FATP1 in the same membranes (Fig. 7F). In conclusion, these data demonstrated that FATP1 inhibits the 11cROL formation by interacting with both RPE65 and LRAT proteins and altering their functions.

DISCUSSION
RPE65, responsible for severe early-onset congenital blindness (25,26), is the key-enzyme for regeneration of 11-cis conformation of vitamin A. The molecular mechanisms regulating this rate-limiting step are not yet fully known and could explain phenotypic variability in RD caused by analogous gene mutations (27). Our work has consisted of identifying, through yeast two-hybrid screening, new RPE65 protein partners, studying their modality of interaction with RPE65 and defining their regulatory role in the visual cycle. We demonstrated that FATP1 is a regulatory partner of both LRAT and RPE65 in the microsomal membrane of RPE cells.
Given that no structural information was available to delineate interacting domains, we used the fulllength RPE65 open reading frame as bait in the two-hybrid screen. To analyze specific protein-protein interactions, we selected three RPE65 fragments, N147, 140 -318 and 311C. The N-terminal 147 residues showed the highest interaction of the three fragments, suggesting that preferential FATP1 interaction may biologically occur within the N terminus of RPE65. Interestingly, the crystal structure of native RPE65 recently described shows that its N-terminal region could be available for interacting proteins (supplemental Fig. S2 and Ref. 28).
FATP1 was first identified in a functional screen for adipocyte proteins that facilitate fatty acid uptake into 3T3-L1cells (29). Later, FATP1 function was redefined as an acyl-CoA synthetase with broad specificity for both long (palmitate) and very long chain fatty acids (30,31). FATP1 belongs to a large evolutionarily conserved family of integral membrane proteins with a molecular weight of 63 kDa. To date, five members have been described in mice and six in humans (FATP 1-5/6) (32), which show tissue-specific expression patterns. Here, we characterized the expression pattern of porcine FATP1 by the simultaneous analysis of mRNA and protein. The higher expression observed in skeletal muscle, heart, white adipose tissue, brain, and skin is consistent with that previously reported in mouse and human (29,(32)(33)(34). Moreover, we show for the first time that the RPE and retina also contains high levels of FATP1 expression.
Lewis et al. (24) depicted FATP1 as an integral membrane protein composed of one transmembrane domain near the N terminus of the protein; the N terminus faces the extracellular/ luminal space, whereas the C terminus projects into the cytosol. In light of our observations, FATP1 interacts with the visual cycle proteins LRAT and RPE65 via its cytosolic domain, con- sistent with its established membrane topology. Note that the two-hybrid system in yeast identifies only areas of interaction outside of the membrane bilayer. This has also forced to truncate the transmembrane domains of LRAT to study its interaction with FATP1. The FLAG-FATP expressed in 293-LR cells mainly localized in a cytoplasm network with a reticular pattern and superimposes with EGFP-RPE65 fluorescence, suggesting a correct targeting to the ER membrane. Moreover, after subcellular fractionation, FATP1 appears confined to RPE microsomes like RPE65 and LRAT. All together, our present results show that FATP1 functions mainly at the ER membrane. The proposal that FATP1 may not function as a plasma membrane fatty acid transporter was originally made after in vitro expression of functional tagged FATP1 proteins in intracellular membranes thus resulting in its characterization as a long-chain fatty acyl-CoA synthetase (24,30). The acyl-CoA synthetase catalyzes esterification of long-chain fatty acids (LCFAs) with coenzyme A (CoA) and ATP as essential cofactors, the initial step required for all cellular LCFA utilization. This enzyme was indeed identified in the microsomal fraction and there was no evidence for acylation by the plasma membrane (35). Moreover, recent data obtained in cultured human myotubes suggested that FATP1 may function within an intracellular protein complex (36). FATP1 enhanced fatty acid import to this protein complex, trapping fatty acids and facilitating their access to other enzymatic pathways. Similar data were obtained with FATP4, that shares 62.5% aa identity with FATP1, and which has an acyl-CoA-synthetase activity and is in fact localized to the ER membrane (37).
Our data show that both LRAT and RPE65 co-localize in the microsomal fraction of porcine RPE cells but suggest that they do not form protein complexes. Indeed, no direct protein-protein interaction was seen using both two-hybrid and co-immunoprecipitation assays. These results are in agreement with the recent finding that LRAT is not an acyl-transferase for RPE65 and does not modify its membrane association and isomerase activity, but only serves to provide atRE (6). In contrast, FATP1 interacts both with LRAT and RPE65 but does not link them together, suggesting that it may regulate their activities independently.
FATP1 markedly decreased the isomerase activity of RPE65 throughout the kinetics of 11cROL production without affecting its expression levels, suggesting a direct enzymatic regulation. The mechanism of this regulation is not clear at present. It may be the result of subtle changes at the level of substrate binding or by an allosteric effect. The fact that RPE65-mediated isomerization is a limiting step of the visual cycle raises the intriguing question of the biochemical significance of this FATP1-inhibiting effect. This inhibition might be either transient (strong interaction) or constitutive (stable interaction) and thus participate in the apparent low isomerase activity of RPE65 (23). However, FATP1 appeared less abundant than RPE65 in the RPE and consequently its inhibitory effect on isomerase activity might involve a mechanism of negative cooperativity, FATP1 acting as a heterotropic modulator. Even if we do not provide evidence that the protein interaction is sufficient for inhibition, this is strongly suggested by the dose-dependence of both proteinprotein interaction and isomerase inhibition. In addition, we propose that FATP1 acts on the isomerase activity rather than hindering the access of the substrate to RPE65 catalytic site because: 1) FATP1 interacts with RPE65 via extramembrane domains, whereas access of substrate to RPE65 is likely within the hydrophobic core of the ER membrane (17), and 2) isomerase inhibition is important in the homogenate from Sf9 cells under saturating substrate concentrations.
This study has also demonstrated that CRALBP forms complexes with RPE65. The absence of interaction between FATP1 and CRALBP supports the specificity of the FATP1 interaction with RPE65 and LRAT. CRALBP is an acceptor of 11-cis retinol that promotes retinoid isomerization. A high affinity binding site for 11cROL appeared sufficient to drive isomerization by mass action, 11cROL being selectively removed and oxidized to 11cRAL by 11cRDH (9,10). Therefore, FATP1 and CRALBP probably exert opposite effects by different molecular mechanisms of regulation of RPE65 activity. FATP1 did not affect the initial activity of LRAT, suggesting that protein interaction does not directly affect the acyltransferase. In RPE cells, LRAT catalyzes a CoA-independent esterification of atROL, which results in a rapid burst of retinyl ester synthesis, followed by a plateau where the amount of retinyl ester does not change (38). Similar curves were observed in the present study within 293-LR and Sf9 cells expressing adequate low level of LRAT activity under saturating atROL concentrations. Saari et al. (39) have previously demonstrated that the reversibility of LRAT may contribute to the depletion of atRE. Moreover, the co-expression of FATP1 and LRAT in both cell lines resulted in significant decreases in the level of atRE production when the LRAT reaction has reached equilibrium. Therefore, one explanation could be that FATP1 stimulates the reversal of LRAT. Because FATP1 can mobilize palmitate as a substrate to produce palmitoyl-CoA, we would expect that it participates in a palmitate exchange in the LRAT reaction. However, addition of the substrates for FATP1 did modify neither the LRAT reaction nor the regulation by FATP1, suggesting that the acyl-CoA synthetase activity has no direct effect on the LRAT activities. This can be explained in part since the retinyl esters are synthesized by LRAT in a CoA-independent reaction (38). For the reversal of the LRAT reaction, it is unclear whether palmitate generated in situ by LRAT is equivalent to palmitate added to homogenates with respect to accessibility to the acyl-CoA synthetase. Further studies are needed to determine whether the acyl-CoA synthetase activity of FATP1 may be crucial for the reversal of the LRAT reaction.
The observations reported in this work suggest that FATP1 could operate on both LRAT and RPE65, a feature that may have physiological significance for the mobilization of atROL pools. Light exposure of photoreceptor cells results in the release of atROL, which may be taken up by RPE and esterified by LRAT (1). FATP1 may regulate the synthesis of atRE by acting on the reverse activity of LRAT, which functions as a retinyl ester hydrolase. Therefore, in response to light, FATP1 could limit the production of atRE in RPE while in darkness it might inhibit the production of 11cROL acting on RPE65. As another example of regulation of the visual cycle in RPE, RGR was shown to both inhibits LRAT activities and enhances isomerase activity in darkness (19,20). The explanation given by Radu et al. (20) is that activation of RGR by light stimulate LRAT and REH activities and the processing of atRE between lipid droplets and ER membranes where it can be converted to 11cROL by RPE65 in darkness. Accordingly, FATP1 and RGR could act on the synthesis of visual chromophore in regulating both LRAT and RPE65 in a light-dependent manner. Defects in nearly every step of the visual cycle are responsible for RD (3). Retinitis pigmentosa (RP) and Leber congenital amaurosis (LCA) involve the impaired synthesis of visual chromophore leading to the degeneration of photoreceptor cells and are partly caused by loss-of-function alleles in RPE65 and LRAT (2). Macular dystrophy involves the accumulation of toxic lipofuscin fluorophores such as A2E derived from atRAL in the RPE, which ultimately degenerate and cause photoreceptor cell death (3). As FATP1 inhibits the production of 11cROL, a gain-of-function mutation in FATP1 would potentially be deleterious in RP and LCA by phenotypic convergence. In contrast, it might have an important protective role in Stargardt disease by limiting the accumulation of atRAL and A2E. Further in vivo experiments will be necessary to validate this assumption.
A growing number of reports provide evidence of multifaceted roles of energetic metabolism proteins (40). By its acyl-CoA synthetase activity, FATP1 activates fatty acids as the first step of many kinds of metabolism in different cell types. Our data demonstrate that in RPE cells, the targeting of FATP1 to the ER membrane may indicate a specific requirement for the metabolic retinoid cycle. This study provides new perspectives on the involvement of non-retinoid metabolism-related proteins in the regulation of the visual cycle.