Retinal Stimulates ATP Hydrolysis by Purified and Reconstituted ABCR, the Photoreceptor-specific ATP-binding Cassette Transporter Responsible for Stargardt Disease*

Many substrates for P-glycoprotein, an ABC transporter that mediates multidrug resistance in mammalian cells, have been shown to stimulate its ATPase activity in vitro. In the present study, we used this property as a criterion to search for natural and artificial substrates and/or allosteric regulators of ABCR, the rod photoreceptor-specific ABC transporter responsible for Stargardt disease, an early onset macular degeneration. ABCR was immunoaffinity purified to apparent homogeneity from bovine rod outer segments and reconstituted into liposomes. All-trans-retinal, a candidate ligand, stimulates the ATPase activity of ABCR 3–4-fold, with a half-maximal effect at 10–15 μm. 11-cis- and 13-cis-retinal show similar activity. All-trans-retinal stimulates the ATPase activity of ABCR with Michaelis-Menten behavior indicative of simple noncooperative binding that is associated with a rate-limiting enzyme-substrate intermediate in the pathway of ATP hydrolysis. Among 37 structurally diverse non-retinoid compounds, including nine previously characterized substrates or sensitizers of P-glycoprotein, only four show significant ATPase stimulation when tested at 20 μm. The dose-response curves of these four compounds are indicative of multiple binding sites and/or modes of interaction with ABCR. Two of these compounds, amiodarone and digitonin, can act synergistically with all-trans-retinal, implying that they interact with a site or sites on ABCR different from the one with which all-trans-retinal interacts. Unlike retinal, amiodarone appears to interact with both free and ATP-bound ABCR. Together with clinical observations on Stargardt disease and the localization of ABCR to rod outer segment disc membranes, these data suggest that retinoids, and most likely retinal, are the natural substrates for transport by ABCR in rod outer segments. These observations have significant implications for understanding the visual cycle and the pathogenesis of Stargardt disease and for the identification of compounds that could modify the natural history of Stargardt disease or other retinopathies associated with impaired ABCR function.

Stargardt disease (Mendelian Inheritance in Man number 248200; also called fundus flavimaculatus) is an autosomal recessive disorder that affects approximately one person in 10,000 (1). Affected individuals experience central vision loss and a progressive bilateral atrophy of the macular region of the retina and retinal pigment epithelium (RPE). 1 A distinctive feature of Stargardt disease is the accumulation of fluorescent, shortwave absorbing material in the RPE, which appears in postmortem samples as lipofuscin-like deposits (2)(3)(4). These clinical and histopathologic features resemble those seen in the more common age-related macular degeneration, suggesting that investigation of the pathophysiology of Stargardt disease may reveal mechanisms relevant to age-related macular degeneration (5)(6)(7)(8).
The gene responsible for Stargardt disease codes for a member of the ABC transporter family, ABCR (9 -11). The spectrum of mutations identified thus far in Stargardt disease patients suggests that these individuals carry at least one partially functional allele. By contrast, some cases of autosomal recessive retinitis pigmentosa, a disease characterized by a progressive degeneration of the rod-rich peripheral retina, are associated with homozygosity of frameshift or splice site mutations in the ABCR gene which are likely to result in complete loss of ABCR function (12,13). An association between a subset of subjects with age-related macular degeneration and variant alleles of the ABCR gene has been reported (14), but interpretation of the data has been controversial (15).
Expression of the ABCR gene is confined exclusively to retinal rod photoreceptors, and the ABCR protein is localized to the rim and incisures of rod outer segment discs, where it is present at a molar ratio of approximately 1:120 with rhodopsin (16 -19). This highly specific localization suggests that ABCR may transport a molecule or ion that plays a specialized role in photoreceptor homeostasis. Although small quantities of ABCR may also be present in the rod outer segment plasma membrane, the localization of the majority of ABCR to the internal disc membranes suggests that the relevant transport event is between the lumenal and cytosolic faces of the disc membrane. Because outer segment turnover occurs via RPE phagocytosis, a derangement in this transport mechanism (with the resulting accumulation of one or more compounds within the outer segment) could plausibly explain the progressive accumulation of material within the RPE in Stargardt disease.
From the preceding discussion it is clear that identifying the molecule or ion transported by ABCR is a prerequisite to understanding the normal role of ABCR and the pathogenesis of Stargardt disease. To date, substrates have been identified for only a few of the over 30 known mammalian ABC transporters. In these few cases, critical information regarding the identity of the substrate(s) has often come from genetic studies. For example, the defect in chloride resorption in humans with cystic fibrosis provided the first indication that CFTR mediates chloride transport (20), and mutations in the human canalicular organic anion transporter, SPGP, have been found in a subset of patients with progressive familial intrahepatic cholestasis, an inherited disorder of hepatobiliary excretion (21). For Pglycoprotein and its close relatives, overexpression of the corresponding genes or cloned cDNAs in a variety of mammalian cells was found to confer resistance to chemotherapeutic agents, and targeted disruption of these genes in the mouse shows that MDR1A is responsible for extrusion of hydrophobic compounds across the blood-brain barrier (22).
For a number of transporters, including CFTR, SPGP, and P-glycoprotein, these observations have been extended by experiments in vitro. Purified and reconstituted CFTR shows chloride channel activity (23), purified and reconstituted Pglycoprotein shows substrate-dependent ATP hydrolysis (24 -27) and ATP-dependent vectorial transport of substrates (28,29), and partially purified recombinant SPGP mediates taurocholate transport and taurocholate-dependent ATP hydrolysis (30). In the in vitro experiments with P-glycoprotein, ATP hydrolysis is stimulated to different extents by compounds that are extruded by this protein in living cells. A second class of compounds, called chemosensitizers, increases the sensitivity of cells to chemotherapeutic agents and also stimulates ATPase activity in vitro, possibly by uncoupling ATP hydrolysis and substrate extrusion. This work suggests that it may be possible to identify substrates and allosteric regulators of other ABC transporters by searching for compounds that stimulate the ATPase activity of the purified and reconstituted transporters.
To characterize the biochemical properties of ABCR with the goal of identifying its substrate, we report here a procedure for solubilizing and functionally reconstituting the purified protein into lipid membranes. Several compounds, including amiodarone, digitonin, and various geometric isomers of retinal, produce a significant increase in ATPase activity of this preparation. Interestingly, amiodarone and digitonin greatly enhance the stimulatory activity of all-trans-retinal but not of each other, suggesting that they act by a similar mechanism which differs from the mechanism by which all-trans-retinal acts. These data, together with the clinical and histological data summarized above, suggest a model in which ABCR is involved in the recycling of retinoids that attends photoactivation of rhodopsin (the visual cycle).
Preparation of ROS-ROS were prepared from dark adapted and frozen bovine retinas under dim red light as described (31). Aliquots were stored at Ϫ80°C in the dark.
Conjugation of Rim 3F4 Antibody to Sepharose Beads-Monoclonal antibody Rim 3F4, which recognizes a linear epitope near the C terminus of bovine ABCR (16), was purified from ascites fluid by caprylic acid precipitation (32) and conjugated to CNBr-activated Sepharose 4 Fast Flow beads (Amersham Pharmacia Biotech) at a ratio of 1 mg of antibody per ml of resin. The resin was discarded after each use to improve reproducibility in the purification of ABCR.
Purification of ABCR from Bovine ROS-All purification procedures prior to the elution step were performed under red dim light at 4°C to minimize aggregation of photobleached rhodopsin. Purified ROS from 4 -8 bovine retinas were diluted into greater than 10 volumes 1ϫ phosphate-buffered saline, pelleted, and solubilized in 1.4 ml of ROS/ CHAPS buffer (10% glycerol, 0.75% CHAPS, 50 mM HEPES, pH 7.0, 0.5 mg/ml crude brain PE, and 0.5 mg/ml egg PC, 140 mM NaCl, 3 mM MgCl 2 , 5 mM ␤-mercaptoethanol). The sample was rotated at 4°C in the dark for 1 h to completely solubilize the ROS membranes. Insoluble material was removed by microcentrifugation at 16,000 rpm for 5 min, and the supernatant was added to 200 l of Rim 3F4-Sepharose and rotated in the dark at 4°C for 3 h. The resin was recovered by microcentrifugation at 3000 rpm and washed three times for 1 min and then two times for 20 min (with gentle rotation) using 1 ml of ice-cold ROS/CHAPS buffer. Three successive elutions were performed at 4°C using 80 l each of 0.2 mg/ml of a peptide (H 2 N-NETYDLPLHPRTAG-COOH) containing the Rim 3F4 epitope (16) in ROS/CHAPS buffer for a total of 45 min. An irrelevant peptide, JN50 (H 2 N-MLRNNL-GNSSDC-CONH 2 ), was used as a control for the specificity of elution.
Reconstitution of ABCR into Proteoliposomes-240 l of immunoaffinity purified ABCR in ROS/CHAPS buffer was added to a premixed solution containing 30 l of 15% octyl glucoside and 90 l 50 mg/ml sonicated lipid in 25 mM HEPES, 140 mM NaCl, 10% glycerol. The sample was mixed, incubated on ice for 30 min, and then rapidly diluted into 2 ml of reconstitution buffer (25 mM HEPES, pH 7.0, 140 mM NaCl, 1 mM EDTA, 10% glycerol) at room temperature and incubated at room temperature for an additional 2 min. To remove detergent, the sample was passed (at room temperature) over a 2-ml Extracti-Gel (Pierce) column that had been thoroughly prewashed with the reconstitution buffer. The flow-through was stored on ice. Light microscopic examination of the reconstituted preparation stained with 1,1Ј-dioctadecyl-3,3,3Ј,3Ј-tetraethylindocarbocyanine shows a heterogeneous distribution of vesicle sizes. Unless noted otherwise all reconstitutions reported here were performed with swine brain polar lipids.
ATPase Assays-ATPase assays were performed using ABCR that had been purified from ROS and reconstituted into liposomes on the same day. ATPase assays were performed in 300 l of reconstitution buffer supplemented with MgCl 2 to a final concentration of 3 mM. Appropriate dilutions and/or mixtures of test compounds were prepared in ethanol and were added to 0.1% of the volume of the ATPase reaction; control reactions received the same volume of ethanol alone. The reactions were started by the addition of 30 l of 10ϫ ATP mix containing 50 mM ␤-mercaptoethanol, 500 M ATP with 50 Ci/ml of 3000 Ci/mmol of [␥-32 P]ATP, and incubated at 37°C for 2 h. For experiments in which the ATP concentration varied, the [ 32 P]ATP was held constant at 5 Ci/ml. At the starting and final time points, triplicate samples of 50 l were added to 200 l of 10% activated charcoal in 10 mM HCl, vortexed vigorously for 2 min, and microcentrifuged to pellet the charcoal. 50 l of the resulting supernatant was counted. For experiments to differentiate the effects of retinal isomers, the ATPase assay was manipulated under dim red light and incubated in the dark. All other ATPase assays are performed in room light. Data points in Lineweaver-Burk plots were fitted to a straight line by a least squares criterion; no attempt was made to to fit the data to more complex curves.
Handling and Storage of Retinoids-All procedures were performed under dim red light or in the dark. Retinal, retinol, and retinoic acid were dissolved in ethanol, and for each experiment their integrity was monitored by determining an absorption spectrum. Retinoids were handled in glass vials to minimize absorption to plastic. For storage, aliquots were dried under argon and maintained in the dark at Ϫ80°C. Oxidized all-trans-retinal was prepared by exposing a sample of alltrans-retinal in ethanol to air at 37°C overnight in a sealed glass vial.
SDS-PAGE, Western Blots, and Determination of ABCR Concentra-tion-These were performed as described (18). Western blots were performed with affinity purified rabbit anti-human ABCR antibodies and were standardized using bovine ROS.

RESULTS
Purification and Reconstitution of ABCR-The purification strategy employed in this work is essentially that of Illing et al. (16) modified by the use of a mixture of CHAPS, PC, and PE (ROS/CHAPS buffer) for the solubilization and chromatography steps. Mixtures of CHAPS/PE or CHAPS/PC have been used previously to solubilize and purify several photoreceptor membrane proteins in native form, including cone visual pigments (33), the cGMP-gated channel (34), and the Na/Ca-K exchanger (35). In preliminary experiments, we observed that the ATPase activity from the ABCR preparation (see below) was higher if solubilization and purification were performed with CHAPS than with either octyl glucoside or dodecyl maltoside. In brief, the ABCR purification involves (a) purifying bovine ROS on a sucrose step gradient, (b) solubilizing the ROS in a mixture of CHAPS, PC, and PE (ROS/CHAPS buffer), (c) selectively binding ABCR to monoclonal antibody Rim 3F4conjugated Sepharose, and (d) selectively eluting ABCR with a synthetic peptide corresponding to the Rim 3F4 epitope. All but the last of these steps are carried out under dim red light to minimize denaturation and nonspecific trapping of rhodopsin. Fig. 1A shows the results of the purification procedure. The specificity of the immunoaffinity purification step was assessed by halving a sample of Rim 3F4-Sepharose to which ABCR had been bound and carrying out parallel elutions with either the Rim 3F4 peptide or an irrelevant peptide, JN50. The material eluted with the JN50 peptide will be referred to throughout this paper as "mock-eluted." Coomassie staining and Western blotting show that the ABCR polypeptide, with an apparent molecular mass of approximately 250 kDa, is eluted with the Rim 3F4 peptide but is barely detectable following elution with the JN50 peptide (Fig.  1A). Although we cannot rule out the presence of minor ROSderived contaminants in the final immunoaffinity purified preparation, they are likely to be considerably less abundant than ABCR as judged by the apparent homogeneity of the ABCR preparation. The extensive washing of the columnbound material should remove all low molecular weight compounds initially present in the ROS membrane sample.
Reconstitution of ABCR into brain lipid membranes was accomplished by gently mixing immunoaffinity purified ABCR in ROS/CHAPS buffer with an excess of sonicated swine brain polar lipids (referred to hereafter as "brain lipid"). Efficient removal of CHAPS with a high yield of ABCR ATPase activity was most effectively achieved by absorption to Extracti-Gel resin (Pierce). This method of detergent removal is rapid and efficient and results in minimal losses of ABCR, most likely because the proteolipid complexes are effectively excluded by the low molecular weight pore size of the Extracti-Gel resin. In developing this protocol we have avoided ultracentrifugation of the reconstituted ABCR proteolipid. This circumvents problems associated with inefficient recovery of membranes in the presence of 10% glycerol, used here as a protein stabilizer, and with dispersing a compact proteolipid pellet. Several other methods for detergent removal and proteolipid formation were found to be less effective for reconstitution of ABCR as follows: rapid dilution (36) followed by ultracentrifugation resulted in inefficient removal of detergent and poor recovery of proteoliposomes; gel filtration followed by ultracentrifugation (26,37) resulted in loss of ABCR ATPase activity; and detergent removal by binding to Bio-Beads (Bio-Rad) followed by ultracentrifugation (38) efficiently removed detergent but was accompanied by significant absorption of ABCR protein to the beads.
Purified ABCR solubilized in ROS/CHAPS buffer or reconstituted in brain lipid possesses considerable ATPase activity ( Fig. 1B and see Fig. 7 below). We ascribe this activity to ABCR because it is far lower in the mock-eluted sample. In the experiments described below in which various compounds were used to stimulate ATPase activity, we observed variation in ATPase stimulation relative to the basal ATPase activity of up to 1.5-fold for identical experimental protocols carried out on different ABCR preparations. This variation may reflect differences between ABCR preparations in the level of minor contaminating ATPases, which would presumably contribute to the basal activity but not the stimulated activity, and/or differences in the absolute amount of lipid present in the sample which therefore changes the ratio of lipid to the added compounds, most of which are hydrophobic. To avoid normalizing and averaging data from different ABCR preparations, all of the data presented in the figures show the results of single representative experiments.
Stimulation of ABCR ATPase Activity by Low Molecular Weight Compounds-As one approach to identifying substrates or allosteric regulators of ABCR, we surveyed a group of structurally diverse compounds for their ability to stimulate or inhibit the ATPase activity of purified and reconstituted ABCR. Our earlier localization of ABCR to ROS disc membranes suggested the possibility that it might be involved in the movement of retinal or retinol (9,16,18), and therefore the collection includes a variety of retinoids. The structures of these compounds are shown in Fig. 2, and they fall into four groups as follows: retinoids (compounds 1-5), known P-glycoprotein substrates or sensitizers (compounds 6 -14), detergents and digitonin-related (compounds [15][16][17][18][19][20][21][22], and compounds selected for their structural diversity (compounds . The compounds were initially tested at 20 M (Fig. 3), a concentration at which many P-glycoprotein substrates and sensitizers enhance ATP hydrolysis by P-glycoprotein (24 -27). Most of the compounds had little or no effect on ATPase activity, and none significantly inhibited ATPase activity. However, seven compounds, three retinal isomers (all-trans-retinal, 11-cis-retinal, and 13-cis-retinal), amiodarone, digitonin, dehydroabietylamine, and 2-tert-butylanthroquinone, reproducibly activated the ATPase activity 2-5-fold (the fold activation depending on the compound), a degree of activation similar to that reported for purified and reconstituted P-glycoprotein in the presence of the most potent P-glycoprotein substrates and sensitizers (26,27,39). Most of the experiments involving retinal were carried out with the all-trans-isomer. Except where noted below, these experiments were carried out under room lights and therefore some degree of photoisomerization would have occurred during the 2-h incubation.
One example of the specificity of ABCR can be seen in comparing the stimulatory effects of digitonin-like compounds ( The specificity of ATPase stimulation was assessed by comparing preparations that had been identically reconstituted following elution from the Rim 3F4 immunoaffinity column with either the Rim 3F4 peptide or the irrelevant peptide JN50 (Fig. 4, A-E). In each case, ATPase stimulation is observed only in the sample that was eluted with the Rim 3F4 peptide. The concentration dependence of ATPase stimulation by each of the stimulatory compounds was determined and compared with that of verapamil, a P-glycoprotein sensitizer and a potent activator of P-glycoprotein ATPase in vitro (26). Verapamil had no effect on ATPase activity at any concentration up to 200 M (Fig. 4F). Each of the stimulatory compounds has a characteristic concentration dependence of activation as follows: all-  (Fig. 4, B, D and E); and digitonin activates the ATPase with a sharp optimum at approximately 20 M, indicative of a high affinity stimulatory interaction and a lower affinity inhibitory interaction (Fig. 4C). The dose-response curves suggest a mechanistic distinction between all-transretinal, which appears to act at a single site, and amiodarone, dehydroabietylamine, and 2-tert-butylanthroquinone which appear to act through a cooperative mechanism that involves two or more sites.
As an additional method for assessing the origin of the ATPase activity in the reconstituted ABCR preparation, we examined the sensitivity of basal and all-trans-retinal-stimulated ATPase activity to 1 mM sodium azide, 1 mM ouabain, and a range of sodium vanadate concentrations. Sodium azide and ouabain at 1 mM had no effect, and sodium vanadate showed a very modest concentration-dependent inhibition (Fig. 5). These experiments rule out contamination by two of the most abundant membrane ATPases, the mitochondrial F 1 F 0 -ATPase which is sensitive to azide, and the Na/K-ATPase which is sensitive to ouabain. Moreover, all-trans-retinal does not appear to be a general activator of membrane ATPases as it had no effect on the ATPase activity of crude mouse brain or liver membranes at concentrations of 100, 50, 25, 12, or 6 M. Based on (a) the high purity of the ABCR protein in this preparation, (b) the low ATPase activity and the lack of ATPase stimulation by all-trans-retinal in mock-eluted controls, (c) the insensitivity of the basal and all-trans-retinal-stimulated ATPase activity to azide or ouabain, and (d) the lack of stimulation of other membrane ATPases by all-trans-retinal, we believe that both the basal and all-trans-retinal-stimulated ATPase activities observed here are derived from ABCR and not from contaminating ATPases. In the text below we will assume that this is the case.
Factors Affecting the All-trans-retinal-dependent Stimulation of ATP Hydrolysis-The ATPase activity of purified and reconstituted ABCR in the presence or absence of all-transretinal exhibits a broad pH optimum, and the reaction continues linearly for at least 180 min at 37°C (Fig. 6, A and B). The basal and all-trans-retinal-stimulated ATPase activities of ABCR are sensitive to lipid environment, a property shared with P-glycoprotein (40). Crude sheep brain PE, brain polar lipid, and Escherichia coli polar lipid support all-trans-retinalstimulated ATPase, but egg PC and soybean PC do not ( Fig. 6C and data not shown). In contrast to the reconstituted sample, purified ABCR solubilized in ROS/CHAPS buffer shows no stimulation by all-trans-retinal (data not shown), suggesting that the precise arrangement of protein-lipid contacts is a critical factor in the ability of retinal to activate the ATPase activity. The lack of all-trans-retinal stimulation may be related to the high basal V max observed for ABCR in a mixed CHAPS/lipid environment relative to that in lipid. A similar difference has been observed between detergent solubilized and reconstituted P-glycoprotein in ATPase stimulation by both substrates and sensitizers (26).
An analysis of the ATPase activity of ABCR in the presence of varying ATP concentration shows simple Michaelis-Menten behavior for ATP regardless of whether ABCR is solubilized in ROS/CHAPS buffer (Fig. 7, A and C) or reconstituted in brain lipid (Fig. 7, B and D). In ROS/CHAPS buffer, the K m for ATP is 278 M, and the V max is 27 nmol of ATP/min/mg ABCR, assuming that all of the ABCR is functional. For comparison, highly purified P-glycoprotein in detergent has a K m for ATP of 940 M and a V max of 321 nmol of ATP/min/mg (26). Reconstituted ABCR in the absence or presence of 80 M all-transretinal shows a K m for ATP of 33 or 725 M and a V max of 1.3 or 29 nmol of ATP/min/mg of ABCR, respectively. These data suggest that one effect of a lipid environment is to constrain the ATP hydrolytic activity of ABCR and that this constraint is lost when the protein is solubilized in a CHAPS/lipid mixture. In these calculations we assume that all of the ABCR present in the reaction (and quantitated by Western blotting) is functionally reconstituted and has equivalent access to ATP. As these assumptions are unlikely to apply, the calculated V max represents a lower limit of the true V max , a consideration that makes any quantitative comparison between solubilized and reconstituted preparations problematic. In the Lineweaver-Burk plot shown in Fig. 7D the parallel shift of the curve of 1/V versus 1/[ATP] upon addition of all-trans-retinal indicates that alltrans-retinal acts as an "uncompetitive activator" analogous in effect but opposite in sign to the action of a classical uncompetitive enzyme inhibitor (41,42). The simplest interpretation of these kinetic data is that all-trans-retinal binds to and alters the ABCR-ATP intermediate at the rate-limiting step in the ATP hydrolytic pathway.
Comparison of the Stimulatory Activities of Different Retinoids-To explore the stereochemical specificity of ABCR for all-trans-retinal, we compared the stimulatory activity of various retinoids (Fig. 8). The three geometric isomers of retinal and a photoisomeric mixture of retinal isomers (containing all-trans-, 7-cis-, 9-cis-, 11-cis-, 13-cis-, and various di-cis-retinal isomers) are all equally active in ATPase stimulation ( Fig.  8A and data not shown). All-trans-retinol, all-trans-retinoic acid, and ␤-ionone have stimulatory activity, but in each case it is lower than that seen with all-trans-retinal (Fig. 8, C and D). Like amiodarone, ␤-ionone shows little activity below 20 M and a progressive increase in activity at least up to 100 M. Partially oxidized all-trans-retinal has a dose-response curve that is indistinguishable from that of pure all-trans-retinal (Fig. 8B), suggesting that ABCR does not specifically recognize oxidized retinoids preferentially over all-trans-retinal, and therefore the activity of all-trans-retinal in this assay is not due to small amounts of contaminating oxidation products. We note that ABCR is not efficiently stimulated by the two hydrophobic nonretinoid aldehydes tested, ␤-ionone and hexanal (compounds 37 and 40 in Fig. 2), which show little or no activity at 20 M.
Synergistic Activation of ABCR by All-trans-retinal, Amiodarone, and Digitonin-The different shapes of the dose-response curves for all-trans-retinal, amiodarone, and digitonin suggest that these compounds may interact with ABCR in different ways. To examine this possibility we asked whether mixtures of these compounds activate ABCR in an additive or nonadditive manner. Fig. 9 shows an experiment in which these and various other compounds, each at a fixed concentration, were assayed singly or in combination for their effects on ABCR ATPase. In this experiment, verapamil and progesterone were included as negative controls. ␤-Ionone was included because it produced significant ATPase activation when present at concentrations equal to or greater than 50 M (Fig. 8C).
As described above and shown in Figs The linear time course of ATP hydrolysis over a 3-h incubation implies that the standard 2-h reaction time used to quantitate ATPase activity is within the linear range. C, comparison of retinal stimulated ATPase following ABCR reconstitution with brain polar lipid, brain PE, or egg PC.

FIG. 7. K m for ATP and V max for ATP hydrolysis determined for purified ABCR in ROS/CHAPS buffer or following reconstitution in the presence or absence of all-trans-retinal.
A and C, the ATP concentration dependence of ATP hydrolysis by purified ABCR in ROS/CHAPS buffer shows simple Michaelis-Menten behavior. B and D, the ATP concentration dependence of ATP hydrolysis by purified and reconstituted ABCR in the absence or presence of 80 M all-transretinal. C and D show the Lineweaver-Burk plots derived from the data shown in A and B, respectively. Based only on the intercepts of the fitted lines, the K m for ATP and the V max of ATP hydrolysis are as follows: 278 M and 27 nmol/min/mg (ROS/CHAPS buffer), 33 M and 1.3 nmol/min/mg (reconstituted, no addition), and 725 M and 29 nmol/ min/mg (reconstituted, plus 80 M all-trans-retinal). All-trans-retinal addition therefore increases both the K m for ATP and the V max of ATP hydrolysis more than 20-fold. In these calculations of reaction velocity, we assume that all of the ABCR present in the reaction is functional and has access to ATP. As this assumption is unlikely to apply, any quantitative comparison between solubilized and reconstituted preparations is problematic. The calculated V max values therefore represent a lower limit of the true V max . At the 1-3 mM ATP concentration present in the outer segment the reaction velocity will be close to V max . To quantitate the concentration dependence of the synergistic interaction between all-trans-retinal and amiodarone, digitonin, or ␤-ionone, experiments were conducted in which one compound was added at a fixed concentration, and a second compound was added at a variety of concentrations (Fig. 10). In Fig. 10, the degree of synergy can be assessed by comparing the experimental curve in the presence of both compounds with the curve calculated for simple additivity. In one set of experiments the dose-response curve for ATPase activation by all-transretinal was determined in the presence or absence of 40 M amiodarone (Fig. 10A), 20 M digitonin (Fig. 10B), or 50 M ␤-ionone (Fig. 10C). In this experiment, a somewhat greater synergy was observed with all-trans-retinal and ␤-ionone compared with the experiment shown in Fig. 9. In a second set of experiments, the dose-response curve for ATPase activation by amiodarone was determined in the presence or absence of 100 M all-trans-retinal (Fig. 10D) or 20 M digitonin (Fig. 10E). At this concentration, all-trans-retinal shows appreciable synergy with 40 M amiodarone (Fig. 10A). This experiment shows that the synergy between all-trans-retinal and amiodarone is minimal below 20 M amiodarone and then rises steeply at amiodarone concentrations greater than 30 M (Fig. 10D). In Fig.  10E, the combination of amiodarone and digitonin shows greater than additive activation of ATPase below 40 M amiodarone, but this effect is absent above 40 -50 M amiodarone. The greater than additive behavior of digitonin and amiodarone at low amiodarone concentration may simply reflect the nonlinear concentration dependence of ATPase stimulation by amiodarone alone, as can be seen from the shape of the amiodarone dose-response curve in Fig. 10E (e.g. the degree of stimulation by 50 M amiodarone is greater than the sum of that produced individually by 20 and 30 M amiodarone). If the stimulatory mechanism(s) of digitonin and amiodarone are similar, as suggested by their ability to synergize at higher concentrations with all-trans-retinal but not with each other, then this could account for the greater than additive behavior of these two compounds at low amiodarone concentration.
Kinetic Analysis of ATPase Stimulation by All-trans-retinal and Amiodarone-To explore the mechanism by which amio-  darone, retinal, and the combination of the two compounds stimulate ATP hydrolysis by ABCR, we measured the effect of amiodarone on the apparent affinity for all-trans-retinal and the effect of all-trans-retinal and amiodarone individually and in combination on the K m for ATP and the V max for ATP hydrolysis (Figs. 10F and 11). A Lineweaver-Burk plot of the relationship between the rate of ATP hydrolysis at 50 M ATP and the concentration of all-trans-retinal (Fig. 10F) shows by its fit to a straight line that all-trans-retinal acts without cooperativity at a single class of binding site. The 1.7-fold increase in V max produced by the addition of 40 M amiodarone is associated with a 2-fold increase in the apparent affinity (a decrease in K app from 9.6 to 4.6 M) of all-trans-retinal for this site or sites. (We refer to this value as an apparent affinity or K app rather than a true K m because the binding and release of all-trans-retinal may be complex and the reaction was not performed at saturating ATP concentration.) As revealed by a Lineweaver-Burk analysis, amiodarone and all-trans-retinal alter the K m of ATP and the V max for ATP hydrolysis via distinct mechanisms (Fig. 11). The principal effect of amiodarone (at 40 M) is to increase V max with only a modest change in K m for ATP, as measured from the intercept of the best fitting straight line and the horizontal axis, which corresponds to Ϫ1/K m . Amiodarone addition in the absence of all-trans-retinal lowers the K m for ATP by approximately 2-fold ( Fig. 11B), whereas amiodarone addition in the presence of 100 M all-trans-retinal raises the K m for ATP approximately 2-fold (Fig. 11C). Importantly, these changes are accompanied by a 2-fold decrease in the ratio K m /V max (the slope of the best fitting straight line), an effect that is analogous, but opposite in sign, to that of classical noncompetitive enzyme inhibitors. We will therefore refer to amiodarone as a "noncompetitive activator". The simplest model to account for the kinetic behavior of amiodarone is one in which ABCR is assumed to exist in an equilibrium mixture of enzymatically active and inactive species. By analogy with the diminution in the number of active enzymes resulting from the action of a noncompetitive inhibitor, we propose that amiodarone binding shifts the equilibrium between active and inactive enzymes in favor of the active species.
In contrast to amiodarone, the effect of all-trans-retinal, either alone or in combination with amiodarone, is to increase the V max for ATP hydrolysis and the K m for ATP by the same factor, thereby leaving the ratio K m /V max (the slope of the best fitting straight line) unchanged (Figs. 7D and 11D). This effect on kinetic parameters is analogous but opposite in sign to that displayed by classical uncompetitive enzyme inhibitors, and we will therefore refer to all-trans-retinal as an uncompetitive activator. The simplest model that accounts for this kinetic behavior is one in which all-trans-retinal interacts specifically with an intermediate in the ATPase reaction pathway. By analogy with the specific trapping of enzyme-substrate intermediates produced by the action of a classical uncompetitive enzyme inhibitor, we propose that the binding of all-transretinal creates an accelerated pathway for ATP hydrolysis. The data can also be formalized along the lines of a double-displacement ("ping pong") kinetic scheme in which the progress of one substrate, ATP, through the reaction pathway is accelerated by the binding of the second component, all-trans-retinal, at an intermediate point in the pathway (41, 42).
The two different modes of interaction between ABCR and amiodarone or all-trans-retinal predicts that their combined effects will by multiplicative rather than additive, and therefore accounts in a simple way for their ability to act simultaneously and synergistically on ABCR. The simple Michaelis-Menton behavior and uncompetitive mode of ATPase activation exhibited by all-trans-retinal strongly suggest that this compound is a transport substrate. By contrast, the cooperative behavior and the noncompetitive mode of ATPase activation exhibited by amiodarone implicate this compound as an allosteric effector. These observations suggest that the division of ATPase activators into noncompetitive and uncompetitive classes may prove to be generally useful in the study of other ABC transporters.

Purification and Reconstitution of ABCR-The experiments
reported here define a purification procedure and a set of lipid, detergent, and buffer conditions that permit the isolation of nearly homogenous ABCR and its reconstitution into lipid membranes in functional form. The purified and reconstituted preparation has ATPase activity that appears to derive from ABCR rather than from a contaminating ATPase. The stimulated ATPase activity of ABCR is highly sensitive to its local membrane environment, as has been observed for P-glycoprotein (40). PE supports functional reconstitution as do lipid mixtures containing PE but PC does not. When ABCR is solubilized in a mixture of PE, PC, and CHAPS (ROS/CHAPS buffer) it retains ATPase activity, but this ATPase is not stimulated by all-trans-retinal. This effect of CHAPS is unlikely to reflect denaturation of ABCR, since sensitivity to all-transretinal appears upon transfer of ABCR from this CHAPS/lipid mixture to brain lipid alone. These observations suggest instead that the mixed CHAPS/lipid micelle permits cycles of ATP hydrolysis that are uncoupled from conformational events at the site to which all-trans-retinal binds.
Retinal Stimulation of ABCR ATPase-The finding that alltrans-retinal (and other geometric isomers of retinal) stimulates the ATPase activity of ABCR can be explained by any of three mechanisms as follows: (a) all-trans-retinal exerts a general effect on the lipid environment that alters the conformation of ABCR; (b) all-trans-retinal binds to ABCR as an allosteric effector of the ATPase cycle but is not itself a substrate for transport; and (c) all-trans-retinal is a substrate for transport, and its presence accelerates ATP hydrolysis by a conformational coupling between the transport domain and one or both ATP binding domains.
The first mechanism is unlikely because (a) it does not account for the specificity of ATPase stimulation by all-transretinal relative to the other compounds tested, (b) it seems implausible that retinal should exert an effect on bulk membrane properties in a reaction in which lipid is present at 1-1.5 mg/ml and retinal is present at 10 M (equivalent to 3.3 g/ml), and (c) a quantitative analysis of the concentration dependence of all-trans-retinal stimulation indicates that it acts with simple Michaelis-Menten behavior.
The second mechanism, if correct, would imply a novel strategy for modulating the activity of a rod outer segment protein, i.e. sensing the concentration of free retinal or a related retinoid. Under this assumption, ABCR activity might be modulated by the level of all-trans-retinal released by photoactivated rhodopsin or of free 11-cis-retinal imported from the RPE for rhodopsin regeneration. This mechanism would be in contrast to the regulatory mechanisms identified thus far among outer segment proteins that involve phosphorylation or binding to calcium, GTP, or cGMP. Arguing against a simple allosteric mechanism is the effect of all-trans-retinal on the K m for ATP and the V max for ATP hydrolysis which indicates that all-transretinal acts via an uncompetitive interaction with ABCR.
The third possible mechanism, that various geometric isomers of retinal and/or other retinoids are transported by ABCR in a reaction that is coupled to ATP hydrolysis, can only be definitively distinguished from an allosteric mechanism by directly demonstrating ATP-dependent vectorial transport of these compounds in a reconstituted membrane system. At present, the strongest evidence in favor of this hypothesis is the observation that all-trans-retinal acts without cooperativity at a single class of binding site(s) on ABCR and exhibits uncompetitive stimulation of ATPase activity. This kinetic behavior implies that all-trans-retinal binds to an intermediate in the ATPase reaction pathway and that this binding accelerates a rate-limiting step in ATP hydrolysis and/or release of the hydrolysis products. This behavior is precisely that predicted for a transported substrate, and it was not observed with any of the 37 nonretinoid compounds tested in this study. The localization of ABCR to the rod disc membrane also favors retinal (and/or retinol) as the natural ligand since it predicts that the endogenous ligand should be present within ROS, a property fulfilled by all-trans-retinal, 11-cis-retinal, and all-trans-retinol. Given the relatively simple chemical composition of ROS (45) and the absence, at present, of any data implicating other transport processes involving the photoreceptor disc membrane, there are few competing candidate substrates for ABCRmediated transport. Finally, we also favor this hypothesis because, as described below, various pieces of circumstantial evidence suggest that retinal may require active transport across or extraction from the disc membrane and that Stargardt disease is associated with an accumulation of retinal or its derivatives in ROS.
Retinoid Metabolism in the Rod Outer Segment-Photoactivation of rhodopsin results in the hydrolysis and release of all-trans-retinal which is reduced within the outer segment to all-trans-retinol by all-trans-retinol dehydrogenase (named for the reverse reaction). All-trans-retinol is transported to the RPE where it is esterified, isomerized to the 11-cis-configuration, released from the ester linkage, and returned to the outer segment (46). Many of the steps in this process are still poorly understood. In particular, most of the enzymes involved have yet to be purified, and it is not known which steps involve active transport or passive diffusion. The flux of retinal through this pathway can be extremely high under physiological conditions: viewing the blue sky on a cloudless day produces 20,000 photoisomerizations per rod/s in the human retina (47).
Analysis of mice that were exposed to physiological light levels indicates that reduction of all-trans-retinal is the ratelimiting step in the visual cycle (48); following a light flash or under constant illumination, all-trans-retinal accumulates in the retina without a concomitant accumulation of other retinoids. In vitro experiments suggest that this accumulation may increase photoreceptor noise because the combination of free all-trans-retinal and opsin produces a partially active species (49 -52). By contrast, electrophysiological analyses of single photoreceptor cells have failed to reveal a desensitizing effect of exogenously applied all-trans-retinal (reviewed in 53), suggesting either that the in vitro effect is of little physiological relevance, that exogenous all-trans-retinal cannot readily gain access to opsin, or that there are mechanisms in the intact cell that minimize the desensitizing effect of all-trans-retinal. If ABCR facilitates the reduction of all-trans-retinal by all-transretinol dehydrogenase, then one might predict that ABCR dysfunction would lead to an increase in photoreceptor noise. In this regard, it is interesting that individuals with Stargardt disease show an elevated rod threshold and a delayed recovery of rod sensitivity following light exposure (54).
When all-trans-retinal is released from photoactivated rhodopsin, it presumably enters the hydrophobic environment of the disc membrane. In the disc membrane, free all-trans-retinal can react to form a Schiff base with the amine present in PE (55,56), which constitutes 40% of ROS lipid (45). If we assume that Schiff base adducts between all-trans-retinal and PE can form on either the luminal or the cytosolic face of the disc membrane then it seems plausible to suppose that those adducts on the cytosolic face will be available for reduction by all-trans-retinol dehydrogenase, whereas those on the lumenal face will be sequestered from the enzyme active site. This line of reasoning assumes that the active site of all-trans-retinol dehydrogenase has access to substrates associated with only one side of the disc membrane, consistent with the known asymmetric disposition of membrane-associated proteins. In the absence of an active mechanism for transmembrane flipping of the PE-all-trans-retinal Schiff base adduct, we would predict that this adduct would accumulate on the luminal face of the disc membrane. Its abundance there would presumably reflect the rate of spontaneous hydrolysis of the PE-all-transretinal Schiff base and subsequent equilibration of all-transretinal to the cytosolic leaflet of the disc membrane.
A Model for ABCR in the Visual Cycle-The considerations outlined in the preceding three paragraphs, together with the localization of ABCR to the disc membrane and the retinal-dependent stimulation of ATP hydrolysis by ABCR, suggest that ABCR acts in the visual cycle to flip PE-all-trans-retinal adducts from the lumenal to the cytosolic face of the disc membrane, move free all-trans-retinal from the lipid phase of the In this model all-trans-retinal is released into the disc membrane and is transported or presented to all-trans-retinol dehydrogenase on the cytosolic face of the disc by ABCR. ABCR may also facilitate the export of all-transretinol by extracting it from the disc membrane as postulated for the action of P-glycoprotein and its hydrophobic substrates by the hydrophobic vacuum cleaner model (59). Ultimately, all-trans-retinol is exported across the plasma membrane to the adjacent RPE. A hypothetical path of 11-cis-retinal movement is indicated by arrows in the lower half of A. If 11-cis-retinal partitions into the disc membrane, ABCR may facilitate its removal from the lumen of the disc and/or its binding to opsin to regenerate rhodopsin. Whether retinal is released from or attached to opsin with a specific orientation or location relative to the membrane is currently unknown. B, model for the accumulation of shortwave absorbing material in Stargardt disease. The distal region of a rod outer segment and the adjacent RPE cell is shown. Left, in the normal retina there is minimal accumulation of retinoid derivatives in the disc membrane and therefore only a slow accumulation of these or related derivatives in the RPE. Right, in the Stargardt disease retina, retinoids accumulate in the disc membrane (represented by a heavy outline of the disc membrane), and these or related derivatives accumulate in the RPE in phagolysosomes containing ingested ROS (represented by filled cytoplasmic inclusions). disc membrane to a juxtamembrane location, or possibly reorient all-trans-retinal in the bilayer (Fig. 12A). In the first case, ABCR activity would resemble the PC flippase activity of mouse MDR-2 (57) and human MDR3 (58), whereas in the latter cases, ABCR activity would resemble the drug extrusion activity of P-glycoprotein, especially as envisioned by the "hydrophobic vacuum cleaner model" (59) in which P-glycoprotein is proposed to extract hydrophobic compounds from the lipid phase and deliver them to the extracellular space. In each case, we hypothesize that the effect of ABCR activity would be to deliver more efficiently all-trans-retinal to all-trans-retinol dehydrogenase, leading to lower photoreceptor noise, more rapid recovery following illumination, and decreased accumulation of all-trans-retinal or its adducts within the disc membrane. In these models of ABCR action, the proposed vectorial movement is topologically equivalent to importing a substrate from the extracellular space, a direction of transport opposite to the direction of lipid flippase activity demonstrated for mouse MDR2 (57) and the human MDR1 and MDR3 P-glycoproteins (58) and opposite to the direction of drug extrusion mediated by P-glycoprotein. Substrate import by ABC transporters is commonly observed in bacteria but has thus far not been reported in multicellular organisms (60).
These models may account for the presence of ABCR in rods but not cones, a difference that may be related to differences in the visual cycle between these cell types (61). Since the lumenal face of the cone disc membrane remains contiguous with the plasma membrane, free all-trans-retinal within the cone disc membrane or all-trans-retinal-PE adducts formed on the lumenal face of the cone disc could react with other amines in the extracellular space and/or bind to extracellular proteins such as interphotoreceptor retinol-binding protein.
The stimulation of ATPase activity by 11-cis-retinal and to a lesser extent by all-trans-retinol suggests that ABCR may also play a role in the movement of these compounds (Fig. 12A). The facile reaction of 11-cis-retinal and opsin in vitro has generally been interpreted to mean that this reaction does not require additional cofactors or catalysts in vivo. However, it is possible that in vivo the selective pressure for rapid and complete regeneration, which is thought to be one of the rate-limiting steps in dark adaptation, has led to the evolution of enzymes that accelerate this reaction. The relative abundance of ABCR together with the rapid lateral diffusion of opsin within the disc membrane (62) should permit frequent encounters between these proteins. Whereas any models for the role of ABCR in this reaction are necessarily speculative, if ABCR facilitates the binding of 11-cis-retinal to opsin, an ABCR-dependent acceleration in regeneration rate should be demonstrable in vitro.
Implications for the Pathogenesis of Stargardt Disease and Other Retinopathies-The work presented here supports the idea that Stargardt disease and other retinopathies caused by mutations in ABCR arise from defects in retinoid metabolism (Fig. 12B). This idea was originally proposed as an explanation for the delayed recovery of rod sensitivity observed in Stargardt disease patients (54). The persistence of this delay in rod recovery despite high levels of dietary vitamin A (63) distinguishes it from classical night blindness of dietary origin (64), and more specifically suggests a model in which the defect in Stargardt disease is not related to the availability of 11-cisretinal but rather to a defect in removing the product of photobleaching, all-trans-retinal.
The model outlined above may also be relevant to recent structural analyses of the fluorescent, shortwave absorbing compounds that accumulate with age in the human RPE. These analyses show that the major component is a diretinal-ethanolamine derivative, A2-E (65, 66). A2-E has been hypothe-sized to form from the condensation of all-trans-retinal and PE in the outer segment and then to accumulate over time in the RPE as a residual product of ROS digestion (65). Although it is not known definitively whether A2-E comprises the fluorescent, shortwave absorbing material that accumulates to high levels in the RPE of Stargardt disease patients, one in vivo study suggests that this compound or a similar one is present at high concentration in these patients (67,68).
The severe defect in rod vision characterized by homozygosity for ABCR null mutations is in contrast to the milder impairment of rod vision experienced by many patients with Stargardt disease. If ABCR is important only as a transporter and not as a structural protein, then the clinical data would suggest that a complete absence of ABCR-mediated transport leads to the death of rod photoreceptors, the final common event in retinitis pigmentosa.
Implications for Pharmacologic Intervention in Retinopathies-The in vitro assay described here may lend itself to the identification of compounds that either enhance or diminish the activity of ABCR. Those that enhance ATPase activity might represent lead compounds for drug development with the goal of enhancing ABCR activity in patients with Stargardt disease or in a subset of individuals at risk for age-related macular degeneration. In vitro, such compounds might increase the ATPase activity of ABCR when added alone or in combination with a transported compound. We note, however, that this assay represents only a screening tool; any interpretation of altered ATPase activity will require a detailed understanding of the mechanism of action of the test compound because, in theory, ATPase activity could be increased by either promoting or circumventing the normal transport cycle. Although we cannot make any firm predictions regarding the in vivo effect on ABCR function of compounds like amiodarone and digitonin, the existence of such compounds suggests that environmental or drug effects may be relevant to degenerative retinal diseases. In this context, it may be of interest to study visual function in patients receiving amiodarone, a drug that has been widely used in the treatment of cardiac arrhythmias at tissue concentrations at or above 40 M (69).