Mutation of key residues of RPE65 abolishes its enzymatic role as isomerohydrolase in the visual cycle

Abstract

RPE65 is essential for isomerization of vitamin A to the visual chromophore. Mutations in RPE65 cause early-onset blindness, and Rpe65-deficient mice lack 11-cis-retinal but overaccumulate alltrans-retinyl esters in the retinal pigment epithelium (RPE). RPE65 is proposed to be a substrate chaperone but may have an enzymatic role because it is closely related to carotenoid oxygenases. We hypothesize that, by analogy with other carotenoid oxygenases, the predicted iron-coordinating residues of RPE65 are essential for retinoid isomerization. To clarify RPE65's role in isomerization, we reconstituted a robust minimal visual cycle in 293-F cells. Only cells transfected with RPE65 constructs produced 11-cis-retinoids, but coexpression with lecithin:retinol acyltransferase was needed for high-level production. Accumulation was significant, amounting to >2 nmol of 11-cis-retinol per culture. Transfection with constructs harboring mutations in residues of RPE65 homologous to those required for interlinked enzymatic activity and iron coordination in related enzymes abolish this isomerization. Iron chelation also abolished isomerization activity. Mutating cysteines implicated in palmitoylation of RPE65 had generally little effect on isomerization activity. Mutations associated with Leber congenital amaurosis/early-onset blindness cause partial to total loss of isomerization activity in direct relation to their clinical effects. These findings establish a catalytic role, in conjunction with lecithin:retinol acyltransferase, for RPE65 in synthesis of 11-cis-retinol, and its identity as the isomerohydrolase.

Regeneration of 11-cis-retinal, the chromophore of all visual pigments (opsins), occurs by a process in the retinal pigment epithelium (RPE) termed the visual cycle that involves isomerization of all-trans-retinyl esters to 11-cis-retinol. In outline, lecithin:retinol acyltransferase (LRAT) (1, 2) esterifies incoming all-trans-retinol to all-trans-retinyl esters, the substrate for the putative isomerohydrolase (IMH) (3, 4). IMH is postulated to perform a concerted hydrolysis and isomerization yielding 11-cis-retinol, which is trapped by cellular retinaldehyde-binding protein (CRALBP) and then oxidized to 11-cis-retinal by 11-cis-retinol dehydrogenase/RDH5 (11cisRDH). An alternate mechanism proposes generation of 11-cis-retinol via a retinyl carbocation intermediate (5). Mutations of RDH5, RPE65, LRAT, and CRALBP in the human (6-12) and cognate disruptions in the mouse (13-16) result in mild to severe blindness due to derangement of RPE retinoid metabolism. CRALBP and RDH5 disruptions do not block the visual cycle but reduce its efficiency (13, 16). Although mutation or loss of LRAT and RPE65 each block the visual cycle, they do so differently. With LRAT loss, no vitamin A accumulates in the RPE, obviating retinoid metabolism (15). With RPE65 loss, all-trans-retinyl ester accumulates to very high levels, but 11-cis-retinoids are not formed (14).

However, the details of the key isomerization step are still unclear. This is partly because the putative IMH (3), or alternate isomerase (5), has yet to be identified at the molecular level, and partly due to uncertainties in understanding the mechanism of isomerization (5, 17, 18). The precise role of RPE65 in isomerization has been particularly perplexing, although its absolute necessity is demonstrated by the Rpe65-deficient mouse (14). Mutations in RPE65 (≈60 to date) give rise to a spectrum of retinal dystrophies ranging in severity and age of onset from Leber congenital amaurosis (LCA) or autosomal recessive child-hood-onset severe retinal dystrophy (arCSRD) to autosomal recessive retinitis pigmentosa (arRP) (6-9, 19, 20). It has been proposed that the role of RPE65 is to present all-trans-retinyl ester to the IMH (18, 21). Alternatively, because it belongs to a family of carotenoid oxygenases, RPE65 may play a direct catalytic role in the visual cycle. This family, including bacterial (22), plant (23), invertebrate (24), and vertebrate (25, 26) carotenoid oxygenases, shares four absolutely conserved histidines that coordinate the iron required for activity of these enzymes (22, 27). Because RPE65 retains these histidines, it might also bind catalytic iron to play a role in isomerization.

We have developed a non-RPE cell culture system coexpressing visual cycle proteins that converts exogenous all-trans-retinol to 11-cis-retinoids. Using this system, we tested the effect of mutating RPE65's conserved histidines and other residues, and the effect of chelation on the production of 11-cis-retinoids. We suggest that the minimal IMH complex consists of LRAT and RPE65, and requires no obligate additional components to produce 11-cis-retinol.

Methods

Generation of Expression Vectors and Site-Directed Mutagenesis. Canine RPE65 and bovine CRALBP cDNAs were subcloned into the bicistronic expression vector pVitro2 (InvivoGen, San Diego). Bovine LRAT and bovine RDH5 cDNAs were subcloned into the bicistronic expression vector pVitro3 (InvivoGen). Non-human homologs were used to discriminate from any endogenous human transcripts in 293-F FreeStyle cells. Site-directed mutagenesis of the RPE65 ORF was done by using the QuikChange XL site-directed mutagenesis kit (Stratagene). All constructs and mutants were sequenced to verify the orientation and accuracy of the ORFs and/or the changes introduced. Plasmids were prepared by using the QiaPrep8, Midi, Maxi, or Mega (Qiagen) scale kits as required.

Cell Culture. Human 293-F FreeStyle (Invitrogen) suspension cells were grown in serum-free FreeStyle 293 expression medium (Invitrogen) and transfected according to the supplier's protocols. Briefly, a typical transfection experiment used 3 × 10⁷ cells in 28 ml of FreeStyle medium mixed with 2 ml of OptiMem-I reduced serum medium containing 40 μl of 293fectin transfection reagent (Invitrogen) and 30 μg of each expression plasmid under study. Cells were grown with shaking at 125 rpm on an orbital shaker platform in a 37°C incubator with a humidified atmosphere of 8% CO₂ in air. All-trans retinol was added 24 h posttransfection, at 1:1,000 dilution to preclude effects of vehicle on the cells. The final concentration (2.5 μM) used is close to the calculated physiological concentration of 3.8 μM after a 2% rhodopsin bleach (28). Cells were harvested at 6 h after substrate addition. Longer durations (12, 24, and 48 h) did not improve conversion rates, presumably because this was a transient expression system.

Retinoid Extraction, Saponification, and Retinoid Analysis. All procedures were performed under red or yellow safelights, using tubes covered with aluminum foil. Pellets from a 19-ml volume of cells were resuspended in 3 ml of ice-cold methanol and vortexed for 1 min. Hexane (3 ml) was added, and the mixture was vortexed for 1 min. The phases were separated by centrifugation at 3,000 × g for 3 min, and the upper phase was reserved. The lower phase was extracted with a second 3-ml hexane, vortexed for 1 min, and centrifuged, and the upper phase was taken off. The pooled hexane extracts were dried with argon. The dried samples were saponified to convert retinyl esters to isomeric retinols (29). Saponified samples were dissolved in hexane (200 μl for retinoids from 19 ml of cells). Aliquots (20 μl) were analyzed on an isocratic HPLC system (Agilent Technologies) employing 3 × 150-mm Lichrospher Si-60 5-μm normal phase columns (Merck; packed by Alltech Associates) with a diode array detector set at 318, 325, and 328 nm, eluting with a hexane (85.4%):ethyl acetate (11.2%):dioxane (2.0%):octanol (1.4%) mobile phase (30) at 0.8 ml/min. This gave a baseline separation of 11-cis, 13-cis, 9-cis, and all-trans-retinols, in that order, as verified by authentic standards and absorption spectra (29). Chromatographic data were collected and analyzed by using chemstation software (Agilent Technologies). Synthetic 11-cis-retinal (Rosalie Crouch, Medical University of South Carolina, Charleston) was reduced to 11-cis-retinol with sodium borohydride. Standard curves for all-trans- and 11-cis-retinols were prepared by chromatographing 0-150 and 0-2,500 ng of pure standards for each isomer to give area (mAU × s) vs. mass plots, and were used to calculate yields of 11-cis-retinol.

In Vitro Iron Chelator Studies. Cell pellets from 30-ml cultures of 293-F cells transfected with pVitro3/LRAT+RDH5 and pVitro2/CRALBP+RPE65 were homogenized in 2 ml of isomerase activity buffer {modified from ref. 4; 10 mM 1,3-bis[Tris(hydroxymethyl)-methylamino]propane (pH 8.0)/100 mM NaCl containing 1 mM CTP and one Complete protease inhibitor tablet per 50 ml (Roche Applied Science)} in a glass homogenizer. Homogenates were centrifuged at 3,000 × g for 20 min, and supernatants were reserved for assay. One-microliter serial dilutions of 800 mM 2,2-dipyridyl (2,2-DP) in ethanol, or ethanol alone, were added to 400-μl aliquots of 293-F extract for final concentrations of 0, 250, 500, 1,000, and 2,000 μM 2,2-DP. All-trans retinol was added (final concentration, 12.5 μM), and the reactions were incubated in the dark at 37°C for 3.5 h with shaking at 750 rpm. Retinyl esters were extracted and saponified, and retinols were analyzed as above.

Immunoblot Analysis. Cell pellets (≈2 × 10⁶ cells) from 1-ml culture aliquots were lysed in 100 μl of CytoBuster detergent (Novagen), incubated on ice for 10 min, and centrifuged at 13,000 × g for 10 min, and the supernatant was harvested for SDS/PAGE analysis. Denatured samples were separated on 12% BisTris NuPage (Invitrogen) gels and electrotransferred to nitrocellulose membranes. Blots were probed with antibodies by standard procedures and developed in color substrate. The primary antibodies used were as follows: rabbit anti-bovine RPE65 antibody (1:4,000); multiple antigenic peptide LRAT antibody (1:4,000); affinity-purified double peptide rabbit anti-LRAT antibody (1:2,000; gift of Dean Bok, University of California, Los Angeles); rabbit anti-RDH5 multiple antigenic peptide antibody (1:2,000); and rabbit anti-CRALBP antibody (1:20,000; gift of John Saari, University of Washington, Seattle). The secondary antibody used was alkaline phosphatase-conjugated goat anti-rabbit IgG (1:10,000; Novagen). Densitometry of bands on immunoblots was performed by using scion image software (release Alpha 4.0.3.2, Scion, Frederick, MD).

Confocal Microscopy. 293-F cells were centrifuged at 500 rpm for 3 min. The pelleted cells were fixed in 1 ml of 2% paraformaldehyde for 15 min at 24°C and centrifuged at 26 × g for 3 min, and the fixative aspirated off then washed three times with 1 ml of PBS, centrifuging at 26 × g for 3 min after each wash. The fixed cells were blocked in 3% goat serum in ICC buffer (1× PBS, pH 7.3/0.5% BSA/0.1% Triton X-100/0.1% NaN₃) and incubated overnight in primary antibodies: rabbit anti-bovine RPE65 antibody (1:400), rabbit anti-LRAT multiple antigenic peptide antibody (1:200), and mouse monoclonal anti-CRALBP antibody (1:1,000; gift of John Saari) diluted in ICC. Cells were washed and incubated in ICC buffer containing Alexa 488-conjugated goat anti-rabbit IgG (Molecular Probes/Invitrogen) and Alexa 568-conjugated goat anti-mouse IgG (Molecular Probes/Invitrogen). Cells were mounted in Gel Mount (Biomeda) and imaged on a Leica SP2 laser scanning confocal microscope. Images of immunolabeled and negative control samples were obtained in identical scanning conditions. Files were imported into photoshop 7.0 (Adobe Systems) and converted to PSD format.

Results

A Minimal in Vitro Visual Cycle. We designed a robust non-RPE cell culture system reproducing the visual cycle, quantitatively accumulating 11-cis-retinoids at concentrations close to physiological levels by transient transfection of 293-F cells. Conversion was robust, allowing the use of unlabeled retinol. Not surprisingly, the bulk of retinoids found in 293-F cells (control or transfected) were retinyl esters (data not shown). All analyses used saponified samples. Incubation of all-trans-retinol in FreeStyle medium (no cells) produces an isomeric mixture of 13-cis, 9-, 13-di-cis, 9-cis, and all-trans-retinols, to a level of 20% 13-cis-retinol at 24 and 48 h. No 11-cis-retinol was detected, however (data not shown). 293-F cells cotransfected with pVitro3/LRAT+RDH5 and pVitro2/CRALBP+RPE65 grown for 24 h and then cultured for a further 6 h with 2.5 μM all-trans-retinol accumulated up to 20% of retinol as 11-cis-retinol, whereas cells transfected with pVitro3/LRAT+RDH5 alone accrued no 11-cis-retinol (Fig. 1A). A 30-ml culture could accumulate up to 800 ng of 11-cis-retinol (Fig. 1B), surpassing observations made with RPE primary cultures (see refs. 28 and 31). Absorption spectra (Fig. 1 A Inset) confirmed the identity of 11-cis and other retinols (29). Using pVitro2 constructs containing either RPE65 or CRALBP, we dissected the necessity of the various components for production of 11-cis-retinoids (Fig. 1B). Transfection of no DNA, LRAT+RDH5 or CRALBP vectors together or alone did not yield 11-cis-retinoids above background levels in the absence of RPE65. Although 11-cis-retinoids were generated by transfection of pVitro2/RPE65 or pVitro2/RPE65+CRALBP alone, 11-cis-retinol output was increased 10-fold when cells were cotransfected with pVitro3/LRAT+RDH5, showing the necessity for LRAT to provide substrate for RPE65 in the physiological isomerization mechanism. Paradoxically, there is higher production of 11-cis-retinol in the presence of LRAT but absence of CRALBP. This effect is correlated with higher expression of RPE65 in these cells (data not shown).

Fig. 1.

Specific transfection of RPE65 and LRAT induces 11-cis-retinol synthesis. (A) 11-cis-retinol synthesis by 293-F cells transfected with RPE65. Synthesis of 11-cis-retinol is seen only in 293-F cells transfected with pVitro2/RPE65+CRALBP and pVitro3/LRAT+RDH5 (red trace), and not when pVitro3/LRAT+RDH5 alone is transfected (blue trace). Traces are representative of multiple analyses. Peak identification: 1, 11-cis-retinol; 2, 13-cis-retinol; 3, all-trans-retinol. (Inset) Absorption spectra of peaks 1, 2, and 3 confirm the identity of the isomers (29). (B) Isomerization of all-trans- to 11-cis-retinol occurs optimally with coexpression of RPE65 and LRAT. Cultures are transfected with six different combinations of pVitro2/RPE65+CRALBP, pVitro2/RPE65, pVitro2/CRALBP, and/or pVitro3/LRAT+RDH5, or no DNA, and incubated with 2.5 μM all-trans-retinol, and retinols are analyzed after 6 h. Data represent means and SDs of three independent sets of experiments.

Western blot analysis of total lysates of 293-F cells transiently transfected with the pVitro2 and pVitro3 vectors expressing CRALBP+RPE65 and LRAT+RDH5, respectively, showed immunoreactive bands at 22, 34, 35, and 61 kDa for LRAT, RDH5, CRALBP, and RPE65, respectively (Fig. 2A) and none in the absence of specific transfection. Immunofluorescence confocal microscopy was used to image RPE65 and CRALBP. As expected, CRALBP coexpresses with RPE65 in transfected cells (Fig. 2B).

Fig. 2.

RPE65, CRALBP, LRAT, and RDH5 are expressed only in transfected 293-F cells. (A) Immunoblot analysis of whole-cell lysates probed with antisera to, from top, RPE65, CRALBP, RDH5, and LRAT. Lane 1, untransfected 293-F cells; lane 2, 293-F cells transfected with an irrelevant plasmid; lane 3, 293-F cells transfected with pVitro2/RPE65+CRALBP and pVitro3/LRAT+RDH5. (B) Single-channel and merged confocal images of 293-F cells transfected with pVitro2/RPE65+CRALBP and labeled to demonstrate coexpression of RPE65 and CRALBP. Immunolabeling: (a) rabbit antibody to RPE565 (green); (b) mouse monoclonal antibody to CRALBP (red); (c) merged triple-labeled image; (d) merged image of untransfected 293 cells labeled with RPE65 and CRALBP antibodies. DAPI labels cell nuclei (blue) in all images.

Iron Chelation Blocks Retinoid Isomerization. To determine the sensitivity of the isomerization activity of expressed RPE65 to iron withdrawal, we incubated extracts of homogenized 293-F cells cotransfected with pVitro2/RPE65+CRALBP and pVitro3/LRAT with 0-2,000 μM of a membrane-permeant heavy metal chelator 2,2-DP. We used this in vitro approach because 2,2-DP at concentrations above 50 μM has a toxic effect on 293-F cells (data not shown). At concentrations of 1,000 μM and 2,000 μM, 2,2-DP reduces isomerization activity in this preparation to 35% or less of activity in the presence of vehicle only (Fig. 3). These effects were not due to inhibition of LRAT because 2,2-DP, at the concentrations used, did not affect total retinyl ester accumulation (data not shown).

Fig. 3.

Isomerization activity is inhibited by the iron chelator 2,2-DP. Extracts of 293-F cell cultures transfected with pVitro2/RPE65+CRALBP and pVitro3/LRAT+RDH5 vectors were treated with 2.5 μM all-trans-retinol for 3.5 h in the presence of 0-2,000 μM 2,2-DP. Retinoids were extracted, saponified, and analyzed by HPLC. Values for 11-cis-retinol synthesis in the presence of chelator are normalized to activity in the presence of ethanol vehicle alone, taken as 100% activity. Results are means and SDs of three independent experiments.

Mutational Analysis of RPE65. Residues implicated in RPE65 function and mutants associated with human disease were generated and tested. The consolidated data are presented in Table 1. These activities are a combined phenotypic effect of each mutant on enzymatic activity and stability, as if each were expressed as a homozygous allele in vivo.

Table 1. Isomerization activity of RPE65 mutants.

Mutant	Conserved family/conserved RPE65*	Activity,† %
Wild type		100
Iron-binding
H180A	Yes/yes	<2
H241A	Yes/yes	<2
H313A	Yes/yes	<2
H527A	Yes/yes	<2
E417A	Yes/yes	<2
Cysteines
C231S	No/no	55
C329S	No/yes	26
C330T	No/no	83
CC329/330SS	No/yes/no	<2
Pathogenic mutations
G40S	Yes/yes	<2
R44Q	Yes/yes	<2
H68Y or A	No/yes	<2
A132T	No/yes	50
H182Y or A	No/yes	10
Y368H	No/yes‡	<3
E417Q	Yes/yes	<2
G528V	Yes/yes	<2
K294T	Yes§/yes	16
C330Y	No/no	<2
A434V	No/no	55
Other criteria
Y368F	No/yes‡	56

^*Original wild-type residue.

^†Values are average of at least three determinations and are expressed as percentage of wild type.

^‡Conserved as aromatic: Y or F.

^§Conserved as basic: K or R.

Mutation of Conserved Iron-Binding Residues Abolishes Isomerization Activity. As shown in Fig. 4A, mutation to alanine of the conserved histidine residues (H180, H241, H313, and H527), paralogous to the iron-binding histidines of BCMO1, severely reduced or abolished 11-cis-retinol synthesis in cells transfected with these mutants compared with wild-type-transfected cells. E417A was also equally inactive; its paralog in BCMO1 (E405) is also involved in iron binding (27). In addition, E417Q, associated with retinal dystrophy (9), lacked activity (Table 1). Although in three mutants there were somewhat lower levels of RPE65, there is sufficient expression such that the loss of isomerization activity cannot simply be attributed to loss of RPE65 protein in transfected cells (Fig. 4B). In particular, in the case of the H180 and E417 mutants, the level of expressed protein was comparable to wild type, but these mutants had essentially no isomerization activity.

Fig. 4.

Effect of mutation of iron-coordinating residues on RPE65 activity and expression. (A) Mutation of iron-coordinating residues abolishes isomerase activity. Synthesis of 11-cis-retinol is normalized to RPE65 immunoreactivity quantified by densitometry. Relative activities of mutants are compared with wild type, expressed as the mean and SD of three determinations. (B) Effect of histidine and glutamate mutations on RPE65 expression. Equivalent volumes of whole-cell lysates of transfections with pVitro2/RPE65+CRALBP and pVitro3/LRAT+RDH5 constructs analyzed by immunoblot using RPE65 (Upper) and CRALBP (Lower) antibodies. The expression level of all mutants except H180A was lower than wild type. Lane 1, untransfected control; lane 2, wild-type RPE65; lane 3, H180A; lane 4, H241A; lane 5, H313A; lane 6, H527A; lane 7, E417A. The lower band in Upper is a nonspecific reactant present in both untransfected and transfected 293-F cells.

Effect of “Palmitoylation Switch” Cysteine Mutations on Isomerization Activity. Isomerization activity of three cysteine residues (C231, C329, and C330) implicated in the “palmitoylation switch” (32) hypothesis of RPE65 action were analyzed (Fig. 5A). Of these, only C329 is absolutely conserved in RPE65. Mutation of C231 to serine (as in zebrafish RPE65) resulted in a modest reduction in activity that was not statistically significant. Likewise, the isomerization activity of the single mutants C329S and C330T (as in chicken RPE65) was not statistically different from wild type. However, activity in the double mutant CC329/330SS was abolished, as was that of C330Y, a pathogenic human mutation (33) (Fig. 5A). We evaluated the expression of these mutants by immunoblot (Fig. 5B). Expression levels of C231S and C330T (naturally occurring variants) were comparable to wild type, whereas the inactive mutants C330Y and CC329/330SS were lower, suggesting that they are less stable. Although the relative activity of C329S was similar to wild type, it was also expressed at a lower level, again suggesting compromised stability due to mutation of a conserved residue.

Fig. 5.

Effect of mutation of “palmitoylation switch” cysteines. (A) Individual conservative changes do not affect 11-cis-retinol synthesis by RPE65, but nonconservative (C330Y) or double (CC329/330SS) mutations abolish activity. Differences between wild type, C231S, C329S, and C330T are not significant. Synthesis of 11-cis-retinol is normalized to RPE65 immunoreactivity quantified by densitometry. Relative activities of mutants are compared with wild type, expressed as the mean and SD of three independent determinations. (B) Effect of cysteine mutations on RPE65 expression. Equivalent volumes of whole-cell lysates of cells transfected with pVitro2/RPE65+CRALBP and pVitro3/LRAT+RDH5 constructs analyzed by immunoblot with primary antibodies to RPE65 (Upper) and CRALBP (Lower). Lane 1, wild-type RPE65; lane 2, C231S; lane 3, C329S; lane 4, C330T; lane 5, C330Y; lane 6, CC329/330SS. The lower band in Upper is a nonspecific reactant present in both untransfected and transfected 293-F cells (see Fig. 4).

Pathogenic Mutations Reduce Isomerization in Direct Relation to Clinical Effect. G40S, R44Q, and G528V are mutants of residues absolutely conserved in the superfamily that are associated with human disease. All had <2% of wild-type activity (Table 1). Besides the four conserved histidines, there are three histidines (H59, H68, and H182) conserved only among RPE65 and its vertebrate paralogs. Mutations of two of these (H68Y and H182Y) are associated with human disease. H182Y (and H182A) had ≈10%, whereas H68Y (and H68A) had <2% of wild-type RPE65 activity (Table 1). Besides these, other mutants (some pathogenic, some of uncertain pathogenicity) involved residues conserved in RPE65 but not generally, including K294T (6, 20), A434V (6, 20), and Y368H (19). K294T and A434V were reported (6, 20) as being of uncertain pathogenicity. We found that K294T had 16%, whereas A434V had 55% wild-type activity. Y368H had only <3% activity as wild type, but Y368F had 56%; not surprising, because phenylalanine is the paralogous residue in BCMO1 and BCMO2.

Discussion

Our results underscore the essential role of RPE65 in isomerization of vitamin A in the RPE, as first shown by the Rpe65-deficient mouse (14). We show that cotransfection of RPE65 and LRAT supports high-level synthesis of 11-cis-retinol. The abolition of 11-cis-retinol synthesis by mutation of key residues of RPE65, homologous to catalytic iron-coordinating residues in bona fide enzyme paralogs (22, 27), strongly argues for a catalytic role for RPE65 in isomerization. Inhibition of isomerization by a chelator of ferrous iron further implicates catalytic iron.

HEK293 cells, immortalized from apparent neuronal lineage cells fortuitously present in human embryonic kidney (34), are reported to express low levels of RPE65, IRBP, and RDH5, and possibly LRAT, under certain conditions (35, 36), and to esterify retinols, but have not been shown to make 11-cis-retinoids (36). Chromophore production was suggested (37), but whether it was 9-cis or 11-cis-retinal was not established. In contrast, we find that untransfected 293-F cells do not express detectable LRAT, RPE65, or RDH5. Although a latent isomerization pathway in 293 cells has been suggested, mostly anecdotally, we find no evidence for such: 293-F cells not transfected with RPE65 do not produce 11-cis-retinoids (see Fig. 1). However, specifically transfected 293-F cells accumulate up to 800 ng of 11-cis-retinol per culture, equal to 2.8 nmol of 11-cis-retinol, equivalent to ≈70% of the chromophore needed for the ≈4 nmol of rhodopsin in the human eye (38). This is a significant achievement for these cells. Using this robust system, we tested the necessity for RPE65 isomerase activity of the core iron-coordinating residues predicted (22) and empirically proven (27) to be required for activity in the carotenoid oxygenase family. Mutation of H180, H241, H313, H527, or E417 [for location relative to Synechocystis apocarotenal oxygenase (ACO) residues (22), see Fig. 6] leads to a loss of isomerization activity, as we predict. Thus, directed expression of RPE65 (and LRAT) leads to major synthesis of 11-cis-retinol by 293-F cells that can be abrogated by mutation of specific residues of the introduced entity (i.e., RPE65). Input from endogenous components is not needed, because the presence or absence of product (11-cis-retinol) correlates exactly with the state of the exogenous transfected components (predicted functional or nonfunctional RPE65). Thus, we conclude that RPE65 and LRAT together constitute the physiological “isomerohydrolase” of the RPE.

Fig. 6.

Comparison of RPE65 with ACO. (A) Selected residues of RPE65 are plotted to equivalent positions on the carbon backbone of the ACO crystal structure (22) by using the visual molecular dynamics program (Version 1.8.3; www.ks.uiuc.edu/Research/vmd) (42). Residues are iron-coordinating (blue), conserved residues associated with disease (pink), conserved/nonconserved cysteines (yellow), and weak pathogenic/nonpathogenic mutations (red). The iron-coordinating residues are clustered in the proposed substrate binding tunnel. Residues more sensitive to mutation tend to occur in the core. (B) Sequence alignment of RPE65 and ACO showing key iron-coordinating, cysteine, and other residues. The original alignment obtained with DeepView/Swiss-PdbViewer (Version 3.7; http://ca.expasy.org/spdbv) was manually adjusted to maximize gap suppression and alignment with predicted structural features (β-sheets and α-helices). Residues are color-coded as above. ^*, identity; ·, similarity.

RPE65 triply palmitoylated at C231, C329, and C330 is proposed to be the active form that transports all-trans-retinyl ester to the IMH (32). Of these, only C329 is conserved in RPE65s. Mutation of C231 to serine, paralogous to T228 in ACO (22) (Fig. 6B) and predicted to be located on the surface of the protein (Fig. 6A), had little effect on activity. Our data suggest that, of these three sites, the most important is C329, paralogous to I320 in ACO (Fig. 6B), located near its substrate binding tunnel (22). C329 (Fig. 6A) may have access to the substrate. We found that the C330T mutation had no effect on RPE65 activity, whereas C329S had 25% of wild-type activity in culture, probably because of decreased stability. However, the double mutant CC329/330SS is inactive, suggesting that one or the other of these adjacent cysteines is required for isomerization, or, alternatively, the double mutation destabilizes the protein. In the inactive C330Y pathogenic mutant (33), we hypothesize that the large tyrosine side group destabilizes the protein or sterically hinders C329. Taken together, these data suggest a reevaluation of the “palmitoylation switch” hypothesis.

A major interest in RPE65 is its role in hereditary blindness (6-9, 19, 20). In general, mutations of residues predicted as essential for important folding (e.g., conserved glycines) or metal coordination (Fig. 6 A and B) roles resulted in severe to total loss of activity (<10% wild type). Sequence anomalies described as rare nonpathogenic variants or of uncertain pathogenicity, such as A434V and K294T (20) (Fig. 6), tested the power of the system. A434V, predicted to be nonpathogenic, is heterozygous in an index patient but homozygous in an unaffected sibling. Because A434V has 55% of wild-type activity, one homozygous for this mildly hypomorphic allele would be expected, as was the case, to be perceived as unaffected. K294T (6, 20), however, was only 16% of wild-type activity. Residue 294 and its paralogs are conserved as a basic residue (K or R; Fig. 6B). If homozygous or with a pathogenic coallele, we predict significant loss of visual function. In a heterozygote, the only case described, no major loss would be expected, as observed.

In light of our data, and the structure of ACO (22), we propose a model for RPE65 action in concert with LRAT. In the RPE, LRAT is essential for IMH activity, because it supplies the obligate substrate (4), otherwise unavailable, and we find that RPE65 activity is markedly enhanced by coexpression with LRAT. Physiological (39) and biochemical (17) data supports direct transfer of retinyl esters to the IMH without the dilution inherent in retrieval from storage sites (e.g., ref. 32). In the binding tunnel of RPE65, as in ACO, four conserved histidines bind catalytic iron. We propose that all-trans-retinyl palmitate binds to RPE65 and interacts with the active site ligands (iron, dioxygen, and/or water). It is possible that the trans to cis isomerization of substrate seen in ACO (22) also occurs in RPE65. We propose that attack by one or more of these groups catalyzes events leading to hydrolysis of the ester bond, but without double-bond scission. The leaving palmitate could form a thioester bond with the adjacent C329 thiolate, with CRALBP favoring release of 11-cis-retinol. Thus, the fatty acyl group could palmitoylate RPE65 at C329 or C330. The palmitate bound to RPE65 could then be transferred to LRAT, as was shown in an in vitro system (32). Thus, LRAT is necessary both to supply substrate and to efficiently remove the palmitoyl product of the reaction, restoring RPE65 to the active state. This model reconciles several seemingly contradictory observations by postulating that the RPE65-LRAT complex is in fact the elusive isomer(ohydrol)ase. The total absence of 11-cis-retinoids in the Rpe65-deficient mouse (14) is a very stringent phenotype not seen in the rather leaky phenotypes of retinoid-binding protein knockouts, even that of CRALBP (16). Furthermore, there is no IMH activity in the LRAT knockout mouse (15). Also, the abundance of RPE65 [≈50% of total RPE microsomal protein (40)] can be explained if it is a particularly slow enzyme (low k_cat with concomitantly low K_m). In support of this, a “turnover number” for production of 11-cis-retinal per molecule of RPE65 was recently calculated to be ≈10 min^-1 (41). A slow enzyme facilitating the kinetically unfavored all-trans to 11-cis isomerization has a marked advantage over lack of such. A “massively parallel” process (i.e., much RPE65) together with nonlimiting substrate (via highly active LRAT) could compensate for the apparent slowness and provide product in the quantity needed. The clinching data provided by this study (a minimal visual cycle, the effect of mutation of the canonical iron-binding residues of RPE65, and its sensitivity to iron withdrawal) complete the picture. Taken together, these data support an enzymatic role for RPE65 in retinol isomerization and a role for it as the isomer-(ohydrol)ase. Finally, the system described provides a functional assay of RPE65 mutations.