RPE65 Is an Iron(II)-dependent Isomerohydrolase in the Retinoid Visual Cycle*

The isomerization of all-trans-retinyl ester to 11-cis-retinol in the retinal pigment epithelium (RPE) is a critical step in the visual cycle and is essential for normal vision. Recently, we have established that protein RPE65 is the isomerohydrolase catalyzing this reaction. The present study investigated if metal ions are required for the isomerohydrolase activity of RPE65. The conversion of all-trans-[3H]retinol to 11-cis-[3H]retinol was used as the measure for isomerohydrolase activity. Metal chelators 2,2′-bipyridine and 1,10-phenanthroline both showed dose-dependent inhibitions of the isomerohydrolase activity in bovine RPE microsomes, with IC50 values of 0.5 and 0.2 mm, respectively. In the same reaction systems, however, lecithin-retinol acyltransferase (LRAT) activity was not affected by these metal chelators. The isomerohydrolase activity inhibited by the metal chelators was restored by FeSO4 but not by CuSO4, ZnCl2, or MgCl2. Moreover, addition of Fe(III) citrate or FeCl3 did not restore the activity, indicating that Fe2+ is the metal ion essential for the isomerohydrolase activity. To confirm this result in recombinant RPE65, we expressed RPE65 in a 293A cell line stably expressing LRAT. In vitro activity assay showed that both metal chelators inhibited isomerohydrolase activity of recombinant RPE65. The addition of FeSO4 restored the enzymatic activity of the recombinant RPE65. Further, two specific iron-staining methods showed that purified RPE65 contains endogenous iron. Inductively coupled plasma mass spectrometry measurements showed that bovine RPE65 binds iron ion with a stoichiometry of 0.8 ± 0.1. These results indicate that RPE65 is an iron-dependent isomerohydrolase in the visual cycle

Photon absorption by the visual pigments results in isomerization of 11-cis-retinal, the chromophore, to all-trans-retinal and, subsequently, triggers a phototransduction cascade (1). Obviously, an efficient regeneration of 11-cis-retinal is necessary for the continuous phototransduction process and normal vision. 11-cis-Retinal is regenerated through a series of sequential reactions termed the visual cycle (2). The most critical step of the visual cycle is the isomerization of all-trans-retinol to 11-cis-retinol in the retinal pigment epithelium (RPE). 3 It has been shown that all-trans-retinol must be first acylated by lecithin-retinol acyltransferase (LRAT), and the resultant retinyl ester serves as a substrate for the enzyme termed isomerohydrolase (3,4). The isomerohydrolase is postulated to concomitantly hydrolyze and isomerize alltrans-retinyl ester to generate 11-cis-retinol (2). The isomerohydrolase activity was known for almost 20 years, but the enzyme has only been identified recently (5,6).
RPE65 is a microsomal protein predominantly expressed in the RPE (7,8). Mutations in RPE65 are associated with some forms of retinal dystrophies such as retinitis pigmentosa and Leber's congenital amaurosis (9 -12). Evidence from the RPE65 gene knock-out mouse has shown that RPE65 is essential for the isomerization of all-trans-retinyl ester to generate 11-cis-retinol (13). Delivery of the RPE65 gene has been shown to restore normal vision in dogs with blindness resulting from RPE65 mutations (14) and in RPE65 knock-out mice (15,16). RPE65 is homologous to ␤-carotene 15,15Ј-monooxygenases, which catalyze the initial step in biosynthesis of vitamin A and related carotenoid enzymes (17)(18)(19). Although the overall identity of RPE65 and carotenoid oxygenases is not high (37% for the mouse protein (17)) there are 4 histidine residues that are absolutely conserved in this protein superfamily. Recently, it has been shown that the 4 conserved histidines are essential for the catalytic activity of ␤-carotene monooxygenases (20 -22). Mutations of these histidines impaired iron binding to ␤-carotene monooxygenase, and thus, it was concluded that these histidines coordinate iron required for its catalytic activity (20).
Recently, we and others have shown that recombinant RPE65 protein has the isomerohydrolase activity in vitro, indicating that RPE65 is the isomerohydrolase in the visual cycle (5,6). In the present study, we show that iron(II) is required for the isomerohydrolase activity of native RPE65 from RPE microsomes as well as RPE65 expressed in 293A cells.
We have further demonstrated that purified RPE65 protein contains endogenous iron.

EXPERIMENTAL PROCEDURES
Cell Culture-The QBI-293A cell line for the adenovirus packaging was purchased from Qbiogene (Irvine, CA). The QBI-293A cells were cultured in Dulbecco's modified eagle medium (Invitrogen) containing 5% fetal bovine serum (Invitrogen) supplemented with 100 units/ml penicillin G, 100 g/ml streptomycin, and 250 ng/ml amphotericin. Adenoviral Vector Construction-Briefly, to construct an adenovirus expressing RPE65 (Ad-RPE65), a full-length human RPE65 cDNA was cloned into the pShuttle vector (Qbiogene, Canada). The purified plasmid DNA was digested with PmeI then co-transformed with pAdEasy-1 vector (Qbiogene) into BJ5183 competent cells. The recombinant Ad-RPE65 viral DNA was digested with PacI, purified, and transfected into QBI-293A cells to generate virus particles. The resulting virus (Ad-RPE65) was propagated in QBI-293A cells and purified by cesium chloride gradient centrifugation. Adenovirus infectious titer was determined by plaque formation assay as recommended by manufacturer and the virus titer expressed as plaque-forming units/ml.
Co-expression of LRAT and RPE65-QBI-293A cells were cultured until 70 -75% confluency (1.2-1.3 ϫ 10 7 cells/15-cm dish). A human LRAT cDNA in the pcDNA6 (Invitrogen) plasmid was linearized and transfected into the cells. Colonies stably expressing LRAT were selected using a medium containing 10 g/ml blasticidin (Invitrogen) as described previously (21). The selected colonies were subcloned to generate a 293A cell line stably expressing LRAT (293A-LRAT). The 293A-LRAT cells were infected with Ad-RPE65 and cultured for another 24 h. The cells were harvested with a cell scraper (Sarstedt, Newton, NC) and rinsed twice with phosphate-buffered saline, then stored at Ϫ80°C until used.
Isomerohydrolase Activity Assay-Bovine RPE microsomes were prepared from fresh cow eyes as described previously (3). All-trans- [11, H]retinol in ethanol (1 mCi/ml, 52 Ci/mmol, PerkinElmer Life Sciences) was dried under argon and re-suspended in the same volume of dimethyl formamide (DMF). For each reaction, 2 l of the non-diluted all-trans- [11, H]retinol in DMF and 10 g of bovine RPE microsomes or 250 g of 293A-LRAT cell homogenate expressing both LRAT and RPE65 were added into 200 l of a reaction buffer (10 mM 1,3bis[tris(hydroxymethyl)methylamino]propane, pH 8.0, 100 mM NaCl) containing 0.5% BSA and 25 M cellular retinaldehyde-binding protein (CRALBP) (3,23). After 2-h incubation in the dark at 37°C, retinoids generated were extracted by the addition of 300 l of cold methanol and 300 l of hexane. The upper organic phase was collected and analyzed by normal phase HPLC as described previously (3). The identity of each retinoid was based on comparison to retention times of known retinoid standards.
Metal Chelation-The 1,10-phenanthroline and 2,2Ј-bipyridine solutions were prepared as stocks of 20 -1000 mM in DMF and added into reaction together with 10 g of bovine microsomal proteins or 250 g of total homogenate of 293A-LRAT cells expressing both RPE65 and LRAT. The total volume of phenanthroline and 2,2Ј-bipyridine solutions added to the reaction did not exceed 2 l to avoid the possible effect of DMF on isomerohydrolase activity. The total volume of the reaction was 200 l, and the reaction conditions were the same as described above. For the control, the same volume of DMF was added without the chelators. Metal ions were added into the reaction 1 h after the addition of the metal chelators to regenerate the isomerohydrolase activity.
Purification of Bovine RPE65-Bovine RPE microsomes were prepared as described (3). The microsomal protein was solubilized in 1% CHAPS, 20 mM Tris-Cl, pH 8.0 and centrifuged at 100,000 ϫ g for 30 min. The detergent-soluble fractions were passed through a DEAE-Sepharose anion exchange column (1 ϫ 20 cm) equilibrated with a loading buffer. A NaCl gradient from 0 to 0.5 M in the same loading buffer was used for the elution of proteins from the column. Collected gradient fractions were analyzed by SDS-PAGE and Western blotting (3). The fractions that had the highest RPE65 protein as shown in West-ern blotting were combined and concentrated using a Centricon 10,000-kDa cut-off ultrafiltration device (Millipore, Billerica, MA).
To confirm the purity of the purified RPE65, the purified protein was resolved on 10% polyacrylamide gels containing 0.5% SDS as described previously (24). The SDS-PAGE gels were stained with 0.25% Coomassie Brilliant Blue R-250 in 40% methanol and 7% acetic acid, and destained using 7% acetic acid and 20% ethanol.
Staining of Iron-containing Proteins in Native Polyacrylamide Gels-Native PAGE was performed as described by Laemmli (24) except that SDS was omitted from the gels and buffers. Pre-cast native polyacrylamide gradient gels (4 -20%) were purchased from Jule Inc. (Milford, CT). Two different methods were used for specific staining of iron bound to proteins loaded in the gel. The first method was based on the ability of iron to catalyze the oxidation of diaminobenzoate by H 2 O 2 (25). After electrophoresis, the gel was immersed in 50 mM sodium acetate at pH 5.0. Hydrogen peroxide was added to 40 mM, and diaminobenzoic acid dihydrochloride from freshly prepared 0.8 M stock solution was added to 80 mM. The gel was gently agitated until the colored bands appeared (usually about 1 h). The gel was rinsed in water and then placed in 7% acetic acid. The gels were analyzed by the imager Chemi Genius 2 (Syngene, Frederick, MD) within several hours after staining.
The second iron staining method is based on the reaction of potassium ferricyanide with iron(II) ions (26) as shown in Reaction 1 as follows.
Following electrophoresis, the gel was stained in the dark by freshly prepared 100 mM potassium ferricyanide in 50 mM Tris-HCl and 100 mM NaCl at pH 7.5 for 10 min. A solution consisting of 10% trichloroacetic acid and 10% methanol was prepared immediately prior to being used as the color development solution. The color development was carried out for 2 h. The development solution was changed at least two times to reduce the background. The gel was recorded and analyzed by the imager as described above.
Measurements of Stoichiometry of Iron in RPE65 Using ICP-MS-The iron concentration measurements were performed by Elemental Analysis, Inc. (Lexington, KY). High resolution inductively coupled plasma mass spectrometry (ICP-MS) was used to measure iron concentrations in RPE65 purified from bovine RPE. The RPE65 protein aliquot (50 l) was weighed into a precleaned, preweighed trace metal-free polypropylene centrifuge tube. Concentrated nitric acid (HNO 3 , Fisher Brand Optima Grade, 250 l) was added, and the tube was allowed to sit overnight. Hydrogen peroxide solution (30% H 2 O 2 , Fluka Trace-SelectUltra, 80 l) was added, and the tube was capped. The sample was placed into an ultrasonic bath at 60°C for 2 h. The clear digestate was then made up to a volume of ϳ5 ml with water (ultrapure, 18.2 M⍀-cm deionized water). The tube was reweighed, and then a solution of internal standard (scandium) was added and weighed. Instrument calibration standards were prepared from a commercial High-Purity multielement stock solution. The scandium internal standard was added to the calibration standards at the same level as in the samples (20 ppb). Five standard points (0, 2, 5, 10, and 20 ppb) were used to generate a linear calibration curve. The correlation coefficient (R) of this curve was 0.99996. The instrument limit of detection was calculated as three times the standard deviation of ten measurements of iron in a blank solution. The limit of detection in sample is the product of the instrument limit of detection and the sample dilution factor.

Inhibition of the Isomerohydrolase Activity in Bovine RPE by Metal
Chelators-To determine if endogenous metal ions are necessary for the isomerohydrolase activity, we used two common metal chelating agents, 2,2Ј-bipyridine and 1,10-phenanthroline, in the isomerohydrolase activity assay. These chelating agents are lipophilic and can efficiently penetrate the cell membranes. As the isomerohydrolase activity is associated with the RPE microsomal membranes, we expected that these agents would be efficient in extracting metal ions from membrane-associated enzymes. Although all-trans-retinyl ester is known to be the substrate of the isomerohydrolase, the poor solubility of retinyl ester in the reaction system limits its use in this assay. Therefore, alltrans-[ 3 H]retinol was used to generate retinyl esters by LRAT in the microsomes of the RPE; these esters were then isomerized/hydrolyzed to 11-cis-[ 3 H]retinol. Incubation of the bovine RPE microsomes with all-trans-[ 3 H]retinol resulted in the formation of retinyl esters and 11-cis-[ 3 H]retinol as shown by the HPLC elution profile (Fig. 1A). As expected, the addition of 5 mM 1,10-phenanthroline or 2,2Ј-bipyridine to the RPE microsomes resulted in an almost complete inhibition of the 11-cis-[ 3 H]retinol generation (Fig. 1, B and C), whereas the production of retinyl ester was not affected significantly, suggesting that LRAT activity is not affected by these metal chelators. The inhibition of isomerohydrolase activity appeared to be chelator concentration-dependent (Fig. 1, D and E). The apparent IC 50 is 0.2 mM for phenanthroline and 0.5 mM for bipyridine. These results indicate that some endogenous metal ion is essential for the isomerohydrolase activity.
Requirement of Ferrous Ion for Retinol Isomerohydrolase Activity-To determine which particular metal ion is required for the isomerohydrolase activity, we added back various metal ions individually into the isomerohydrolase assay mixture after inhibition of the enzymatic activity by 5 mM 1,10-phenanthroline.
After bovine RPE microsomes were preincubated with all-trans-[ 3 H]retinol and 5 mM of phenanthroline for 1 h, various metal ions were added separately into the reaction system and incubated for another 1 h. The isomerohydrolase activity was evaluated by the production of 11-cis-[ 3 H]retinol. Of all the metal ions tested, only 6 mM FeSO 4 (Fig.  2B) and 6 mM ferrous ammonium sulfate (data not shown) partially restored the isomerohydrolase activity. In contrast, the addition of the same concentrations of CuSO 4 , ZnCl 2 , or MgCl 2 did not result in the formation of detectable 11-cis-[ 3 H]retinol (Fig. 2, C-E), demonstrating that ferrous ion is specifically required for the isomerohydrolase activity. Moreover, the addition of 6 mM FeCl 3 (data not shown) or ferric citrate (Fig. 2F) did not restore the isomerohydrolase activity, indicating that Fe 2ϩ rather than Fe 3ϩ is the active form of metal ion in the isomerohydrolase.
Further, we examined if the restoration of the isomerohydrolase activity is dependent on the concentration of the Fe 2ϩ added into the reaction. The results showed that generated 11-cis-retinol increased with FeSO 4 concentrations within a range of 0 -6 mM (Fig. 3). At concentrations higher than 6 mM, further increases of FeSO 4 decreased the isomerohydrolase activity (Fig. 3).

Isomerohydrolase Activity of Recombinant RPE65 Is Inhibited by Metal Chelators and Restored by Exogenous Fe(II)-Recently, we have
shown that recombinant RPE65 expressed in the 293 cell line has robust isomerohydrolase activity (5). Here we have examined if the metal chelators inhibit the enzymatic activity of recombinant RPE65. The in vitro isomerohydrolase assay showed that recombinant RPE65, when co-expressed with LRAT, generated significant amount of 11-cis-retinol from all-trans-retinol in the absence of the chelators (Fig. 4, A and B). The addition of 1 mM of bipyridine significantly decreases the 11-cisretinol generated by the recombinant RPE65 (Fig. 4C). Similar inhibition of the enzymatic activity was observed for another chelator 1,10phenanthroline (data not shown). Consistent with the observation in bovine RPE microsomes, the 11-cis-retinol production was restored when 6 mM FeSO 4 was added to the reaction 1 h after the metal chelat-  FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 ing (Fig. 4D), demonstrating that iron(II) is essential for the isomerohydrolase activity of recombinant RPE65.

Iron in RPE65
Detection of Iron in Purified RPE65 Protein-As RPE65 was found to be the isomerohydrolase of the visual cycle (5, 6) and it is homologous to the known iron-containing monooxygenases (17)(18)(19), it was of interest to determine if RPE65 is an iron-containing protein. RPE65 was purified from the RPE microsomes as described under "Experimental Procedures." SDS-PAGE demonstrated that the purified bovine RPE65 is electrophoretically homogeneous (Fig. 5A). The identity of the purified RPE65 was confirmed by Western blot analysis as described previously (5).
We employed a sensitive assay for iron detection based on the ability of iron to catalyze the oxidation of diaminobenzoate by hydrogen peroxide (25). The diaminobenzoate is oxidized by H 2 O 2 to a brown insoluble pigment that can visualize as little as 5 ng of protein-bound iron (25). Purified bovine RPE65 (50 g, molecular mass of 61 kDa (27)) was loaded on a native polyacrylamide gel, side by side with a positive control, human transferrin (Sigma, 100 g, molecular mass of 78 kDa), a well known iron-binding protein (28), and with bovine serum albumin (BSA, 100 g) as negative control. As shown by the specific iron staining, both the purified RPE65 and transferrin showed intensive iron staining. It is noteworthy that the RPE65 band in the native gel appears   to be smear, possibly due to partial precipitation of the protein and different secondary structures of the protein in the absence of detergents. As expected, BSA did not show any detectable iron staining, demonstrating the specificity of the staining. The dialysis of bovine RPE65 against 20 mM bipyridine in the activity measurement buffer results in the absence of iron staining.
To further confirm that RPE65 is an iron-containing protein, we also used a second method for iron staining based on the well known reaction of potassium ferricyanide with protein-bound iron atoms to form royal blue complexes (26). Transferrin and BSA were again used as positive and negative controls, respectively. The result showed that RPE65 protein as well as transferrin was stained, whereas BSA was negative in staining (Fig. 5C). These iron-specific staining results demonstrated that RPE65 is bound by endogenous iron.
The stoichiometry of iron binding in purified bovine RPE65 was determined using ICP-MS. We also used human transferrin and bovine serum albumin as positive and negative controls, respectively ( Table 1). The protein concentrations for all samples were 4.4 M. The iron concentration in bovine RPE65 was measured in duplicate. We obtained an iron stoichiometry of 0.8 Ϯ 0.1 for bovine RPE65 ( Table 1). The iron stoichiometry for human transferrin was calculated to be 1.9, which agrees well with documented data showing that each transferrin molecule contains 2 iron atoms (28). As expected, the bovine serum albumin sample iron content was close to the level of detection.

DISCUSSION
Conversion of all-trans-retinyl ester to 11-cis-retinol catalyzed by the isomerohydrolase is a critical step in the regeneration of 11-cis-retinal in the visual cycle and thus, essential for normal vision. Although the isomerohydrolase activity has been known for almost 20 years, the enzyme that catalyzes this reaction has never been purified. Recently, we and others have established that RPE65 is the isomerohydrolase in the visual cycle (5,6). In the present study, we provide the first direct evidence indicating that RPE65 is an iron-binding protein, and endogenous iron(II) is an essential co-factor for the isomerohydrolase. This finding provides new insights into the mechanism for the isomerohydrolase activity.
The present study showed that deprivation of metal ions from the RPE microsomes by common metal chelators 1,10-phenathroline and 2,2Ј-bipyridine inhibits the isomerohydrolase activity, suggesting that some endogenous metal ion bound to the isomerohydrolase is required for its activity. Previous studies showed that the isomerohydrolase activity is associated with the microsomal membranes of the RPE (3, 29, 30), which is consistent with the finding of its hydrophobic substrate, alltrans-retinyl ester (3). Therefore, we used hydrophobic metal chelators, 1,10-phenanthroline and 2,2Ј-bipyridine, to facilitate their penetration into the membrane. As a result of the metal chelating, the isomerohydrolase activity was abolished by these chelators. In the same reaction mixture, however, the LRAT activity is not inhibited by these chelators, indicating that LRAT does not require endogenous metal ions for its enzymatic activity. Furthermore, the inhibition of isomerohydrolase by these two metal chelators resulted in increased levels of retinyl ester, further supporting the previous finding that retinyl ester is the substrate of the isomerohydrolase (3). It is noteworthy that the IC 50 values for both metal chelators used in the study are relatively high, consistent with the high concentrations of metal chelators required to inhibit ␤-carotene monooxygenase, a homologue of RPE65 (31,32). One possible explanation for the high IC 50 is that iron ions may be tightly bound to the isomerohydrolase and ␤-carotene monooxygenases. Another possible explanation is that the isomerohydrolase activity assay used bovine RPE microsomes, because purified RPE65 loses its enzymatic activity possibly due to the detergent used in the purification process. The microsomes may contain other metal binding proteins and have high levels of endogenous metal ions.
To identify the metal ion required by the isomerohydrolase, we have tested several divalent ions that are likely cofactors of enzymes, including Fe 2ϩ , Mg 2ϩ , Zn 2ϩ , and Cu 2ϩ . After the near complete inhibition of the isomerohydrolase activity by 1,10-phenanthroline, the metal ions were added separately into the isomerohydrolase assay system. Among these metal ions tested, only Fe 2ϩ restored isomerohydrolase activity in a concentration-dependent manner, whereas Mg 2ϩ , Zn 2ϩ , and Cu 2ϩ did not. Interestingly, our results showed that the iron oxidation state is important, because Fe 2ϩ but not Fe 3ϩ ion restored the isomerohydrolase activity, suggesting the isomerohydrolase specifically requires Fe 2ϩ for its enzymatic activity.
After deprivation of endogenous metal ions, the addition of Fe 2ϩ enhanced isomerohydrolase activity in a Fe 2ϩ concentration-dependent manner within the range of 0 -6 mM. When Fe 2ϩ concentration is higher than 6 mM, the isomerohydrolase activity decreases with the increasing concentrations of Fe 2ϩ in the reaction mixture. This decrease of the isomerohydrolase activity may be ascribed to the free radical generation caused by oxidation of ferrous ion to ferric ion (Fenton reaction) (33) and/or concomitant protein modification by these radicals, in the presence of high and possibly toxic concentrations of iron.
Recently, we demonstrated that recombinant RPE65 expressed in four different cell lines has the isomerohydrolase activity (5). In this study we showed that deprivation of iron also inhibits the isomerohydrolase activity of recombinant RPE65 expressed in 293A-LRAT cells. Moreover, the isomerohydrolase activity of recombinant RPE65 can be restored by the addition of FeSO 4 into the reaction. These results further confirm that the isomerohydrolase activity of RPE65 is dependent   FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 on iron(II). It should be noted that in Fig. 4A the recombinant RPE65 appears to migrate faster in SDS-PAGE than the native enzyme from bovine RPE. We have shown that the molecular mass of native bovine RPE65 has molecular mass higher than theoretical value by ϳ1000 Da probably due to post-translational modifications (27). It is not known if the recombinant RPE65 has similar modifications. The enzymatic activity of RPE65 requires the presence of LRAT (5) and CRALBP (34), suggesting isomerohydrolase may function as an enzymatic complex. To further determine if the iron required for the activity is bound in RPE65 itself or in its functional partners, we have detected iron in purified RPE65 protein. The iron in RPE65 was demonstrated qualitatively by specific iron-staining methods in a native polyacrylamide gel. The same batch of the purified RPE65 protein showed apparent homogeneity on SDS-PAGE, indicating that the smear band of RPE65 in the native gel is not a result of impurity of RPE65. The quantitative ICP-MS measurements showed that stoichiometry of iron in purified bovine RPE65 equals to 0.8 Ϯ 0.1. Because a fraction of RPE65 could lose bound iron during purification process, the ICP-MS results suggest that the real stoichiometry for iron bound in bovine RPE65 is 1.

Iron in RPE65
RPE65 shares significant sequence homology with ␤-carotene monooxygenase, an enzyme belonging to the non-heme iron-containing oxygenases (18). The recombinant human ␤-carotene monooxygenase was shown to be sensitive to the metal chelating agents such as o-phenanthroline and 2,2Ј-bipyridine (32). This inhibition was similar to that of the native enzyme and corroborated an earlier assumption that iron is the metal ligand in the active site of the enzyme (31). The crystal structures of a number of oxygen-activating mononuclear nonheme iron-containing enzymes have displayed a common structural motif in which two histidines and one carboxylate occupy a face of the iron coordination sphere, termed the 2-His-1-carboxylate facial triad (35). Recently, the x-ray structure of the carotenoid oxygenase, the first member of the ␤-carotene monooxygenase family, has been solved (36). It was shown that the endogenous Fe 2ϩ is bound to the enzyme, being coordinated by four conserved histidines (36). Recently, it was also reported that the substitution of any one of iron-coordinated histidines in ␤-carotene monooxygenase caused a complete loss of the enzyme activity (20). The crystal structure of RPE65 has not yet been determined. However, recent studies demonstrated that mutations at the conserved histidine and glutamate residues in RPE65 abolish its isomerohydrolase activity, suggesting that these conserved histidine and aspartate/glutamate residues may be located in the iron-binding site in RPE65 (21,22). It is likely that Fe 2ϩ bound to RPE65 has a similar coordination to its conserved histidines.
The interesting feature of the substrate in the complex of carotenoid oxygenase is the trans-cis isomerization at the methyl substituted double bond of isoprenoid tail upon the binding to the enzyme (36). Based on this observation, we hypothesize that similar mechanism for retinoid isomerization may be used by RPE65 to convert all-trans-retinyl ester to 11-cis-retinol. Although the exact function of the iron bound to RPE65 remains to be investigated, our data demonstrate that iron(II) likely serves as a cofactor for the isomerohydrolase. It is also possible that the endogenous iron is important for the structure and stability of RPE65 protein.
In summary, our data provide the first biochemical evidence showing that RPE65 is an iron-containing protein and iron bound to RPE65 is required for its isomerohydrolase activity.