Interphotoreceptor retinoid–binding protein removes all-trans-retinol and retinal from rod outer segments, preventing lipofuscin precursor formation

Interphotoreceptor retinoid–binding protein (IRBP) is a specialized lipophilic carrier that binds the all-trans and 11-cis isomers of retinal and retinol, and this facilitates their transport between photoreceptors and cells in the retina. One of these retinoids, all-trans-retinal, is released in the rod outer segment by photoactivated rhodopsin after light excitation. Following its release, all-trans-retinal is reduced by the retinol dehydrogenase RDH8 to all-trans-retinol in an NADPH-dependent reaction. However, all-trans-retinal can also react with outer segment components, sometimes forming lipofuscin precursors, which after conversion to lipofuscin accumulate in the lysosomes of the retinal pigment epithelium and display cytotoxic effects. Here, we have imaged the fluorescence of all-trans-retinol, all-trans-retinal, and lipofuscin precursors in real time in single isolated mouse rod photoreceptors. We found that IRBP removes all-trans-retinol from individual rod photoreceptors in a concentration-dependent manner. The rate constant for retinol removal increased linearly with IRBP concentration with a slope of 0.012 min−1 μm−1. IRBP also removed all-trans-retinal, but with much less efficacy, indicating that the reduction of retinal to retinol promotes faster clearance of the photoisomerized rhodopsin chromophore. The presence of physiological IRBP concentrations in the extracellular medium resulted in lower levels of all-trans-retinal and retinol in rod outer segments following light exposure. It also prevented light-induced lipofuscin precursor formation, but it did not remove precursors that were already present. These findings reveal an important and previously unappreciated role of IRBP in protecting the photoreceptor cells against the cytotoxic effects of accumulated all-trans-retinal.

Detection of light by the rod photoreceptor cells in the vertebrate retina begins in the outer segment, an organelle that is packed with the visual pigment rhodopsin, the primary light detector. The light-detecting chromophore of rhodopsin is 11-cis-retinal, which is isomerized by light to all-trans; the photoisomerization activates rhodopsin and begins the reactions of visual transduction (1,2). Photoactivated rhodopsin eventually dissociates to all-trans-retinal and the protein component opsin. All-trans-retinal is a reactive aldehyde with a range of deleterious effects and is removed through reduction to alltrans-retinol. 3 The reduction of all-trans-retinal to all-transretinol occurs primarily in photoreceptor outer segments in a reaction that is catalyzed by retinol dehydrogenase RDH8 (3)(4)(5) and requires metabolic input in the form of NADPH (6,7). The reduction of all-trans-retinal that leaks into photoreceptor inner segments is catalyzed by the broad specificity retinol dehydrogenase RDH12 (3,8). All-trans-retinal that escapes reduction to all-trans-retinol forms lipofuscin precursors (LFPs), 4 such as bis-retinoids, the condensation products of two retinal molecules (9,10). Bis-retinoid adducts are a major component of lipofuscin (11), a fluorescent pigment mixture that accumulates in the lysosomes of the retinal pigment epithelium (RPE), and it has been associated with the development of degenerative diseases of the retina, such as age-related macular degeneration and Stargardt disease.
The all-trans-retinol generated from the reduction of alltrans-retinal is transported out of the rod outer segments and into the adjacent cells of the RPE, where it is recycled to form 11-cis-retinal (12)(13)(14). 11-cis-Retinal is transported from the RPE to the rod outer segments where it combines with opsin to reform rhodopsin. Retinoids have low aqueous solubility (15), and their transport through aqueous space is supported by specialized carriers (16). Such a carrier, the interphotoreceptor retinoid-binding protein (IRBP), is the major soluble protein present in the interphotoreceptor matrix (IPM), the material between the photoreceptor outer segments and the RPE (17,18). IRBP binds the 11-cis and all-trans isomers of both retinal and retinol (19), and it promotes the transfer of all-trans-retinol from photoreceptors to RPE (20) and of 11-cis-retinal in the opposite direction (21). IRBP's role in this trafficking may not be in its transport role per se (22,23) rather than an ability to protect these retinoids from photodecomposition (24 -26). Defects in IRBP cause retinal degenerations in individuals with mutations in the RBP3 gene (27) and Rbp3-null transgenic mice (28). A protective role for IRBP may also be associated with facilitating the clearance of all-trans-retinal generated by light in rod outer segments, which would be promoted by the removal of all-trans-retinol and perhaps of all-trans-retinal itself.
The levels of all-trans-retinol, all-trans-retinal, and lipofuscin precursors can be measured in single rod photoreceptors by imaging their fluorescence (29). The removal of all-trans-retinol by IRBP has been previously characterized with single cell fluorescence imaging in the large photoreceptor cells from amphibian retinas (30). In this study, we have used fluorescence imaging to characterize the effects of IRBP on all-trans-retinol, all-trans-retinal, and lipofuscin components in the much smaller and fragile mouse rod photoreceptors. We found that IRBP can remove both all-trans-retinol and all-trans-retinal in a concentration-dependent manner while preventing the formation of lipofuscin precursors that accumulate with aging and disease in the mammalian retina.

Retinal degeneration in Irbp-deficient mice
Irbp-deficient (Rbp3 Ϫ/Ϫ ) mice exhibit early-onset retinal degeneration that is evident in photomicrographs of animals as young as 1 month of age (28), and the morphometric measurements of outer nuclear layer thickness show progress to severe disease by 1 year of age ( Fig. 1). While affirming the critical role of IRBP in retinal function, the massive photoreceptor cell death occurring in the knock-out-mouse model limits its usefulness for studies of the visual cycle mechanism. Thus, we focused our efforts on single-cell fluorescence imaging of the effects of IRBP on the levels of all-trans-retinal, all-trans-retinol, and lipofuscin precursors present in isolated rod photoreceptor cells from wild-type and other visual cycle mutant strains.

Removal of all-trans-retinol by IRBP
In dark-adapted mouse rod photoreceptors, there is an increase in outer segment fluorescence excited by UV light (excitation, 360 nm; emission, Ͼ420 nm) following rhodopsin bleaching (31). In the case of metabolically intact rods, this fluorescence is due overwhelmingly to the all-trans-retinol formed from the reduction of the all-trans-retinal released by photoactivated rhodopsin (3,6). In single isolated rods, and in the absence of extracellular lipophilic carriers, fluorescence reaches a peak ϳ30 min after bleaching and then declines as retinol slowly leaves the outer segment. Fig. 2A shows an experiment with an isolated wild-type mouse rod photoreceptor that was bleached in the absence of lipophilic carriers.
After bleaching, there was a large increase in outer segment fluorescence due to the formation of all-trans-retinol; at 30 min after bleaching IRBP was added to the extracellular medium at a final concentration of 2 M, and the outer segment fluorescence declined rapidly. This was a single experiment at one IRBP concentration. The experiment was repeated with different cells and with different concentrations of IRBP. The results from these experiments, showing the decline of outer segment fluorescence as a function of IRBP concentration, are plotted in Fig. 2B. To facilitate comparisons across cells and IRBP concentrations, the outer segment fluorescence was normalized to its value immediately after the addition of IRBP, and the data were then averaged. It appears that all of the all-trans-retinol generated by bleaching can be removed by IRBP, because at higher IRBP concentrations (10 and 20 M), the outer segment fluorescence had declined to its pre-bleach value by the end of the experiment. For each IRBP concentration, single exponentials, e Ϫk⅐t , decaying to 0 with unitary amplitude at t ϭ 0 min, were fitted through the fluorescence data. The fits provided the first-order rate constants for all-trans-retinol removal at each concentration of IRBP.
The rate constant of all-trans-retinol removal increased with the IRBP concentration (p ϭ 0.0001; linear regression). The

All-trans-retinal and retinol clearance by IRBP
dependence of the rate constant on IRBP concentration was linear, with a slope of 0.012 Ϯ 0.001 min Ϫ1 M Ϫ1 (Fig. 3).

Removal of all-trans-retinal by IRBP
Broken-off rod outer segments (bROS), separated from the metabolic machinery of the inner segment, have no access to NADPH. Thus, bROS cannot reduce all-trans-retinal to -retinol (3,6). Following rhodopsin bleaching in a wild-type mouse rod outer segment, there is a much smaller increase in outer segment fluorescence due to the released all-trans-retinal, which has a lower fluorescence quantum yield compared with retinol (3) (Fig. 4A). Addition of 10 M IRBP at 30 min after bleaching resulted in the decline of outer segment fluorescence, reflecting the removal of all-trans-retinal by IRBP. This was a single experiment at one IRBP concentration. Fig. 4B shows the collected results from experiments with different concentrations of IRBP. As in Fig. 2B, to facilitate comparisons across cells and IRBP concentrations, the outer segment fluorescence was normalized to its value immediately after the addition of IRBP, and the data were then averaged. The continued increase in fluorescence in the absence of IRBP likely reflects the accumulation of all-trans-retinal as it continues to be released by photoactivated rhodopsin (32), but it does not leave the outer segment. Contrary to the situation with all-trans-retinol, fluorescence did not decline to zero after addition of IRBP, suggesting that, at least within the time scale of the experiment, the carrier cannot completely remove the all-trans-retinal present in the outer segment. The removal of all-trans-retinal could not be readily described as a first-order process, and so it was not possible to assign a removal rate constant for each IRBP concentration. Comparing the levels of fluorescence at 30 min after the addition of IRBP (corresponding to 60 min after bleaching), we did find that addition of 2 M IRBP resulted in a significant decline in outer segment fluorescence compared with no addition of IRBP (p ϭ 0.004, t test) and that 10 M IRBP was more effective than 5 M IRBP (p ϭ 0.04, t test). No significant difference, however, was detected between the effectiveness of 2 and 5 M IRBP (p ϭ 0.32, t test). Because of the fragility of mouse bROS and the low fluorescence signal of all-trans-retinal, it was not possible to characterize retinal removal by IRBP to the same extent as for all-trans-retinol.
The outer segments of metabolically intact rods from Rdh8 Ϫ/Ϫ mice are deficient in Rdh8, the enzyme that catalyzes the reduction of all-trans-retinal to -retinol (3,4). Thus, although they have access to the metabolic machinery of the inner segment, and to NADPH, they also cannot reduce alltrans-retinal to -retinol. Following rhodopsin bleaching in Rdh8 Ϫ/Ϫ rods, outer segment fluorescence increases, reflecting the accumulation of released all-trans-retinal. Addition of 5 M IRBP at 30 min after bleaching resulted in decline of outer segment fluorescence (Fig. 5), indicating removal of all-trans-retinal by IRBP. As with wild-type bROS, the removal of all-trans- Error bars represent standard deviations. The errors for the removal rate constants were obtained from the curve fits.

All-trans-retinal and retinol clearance by IRBP
retinal in intact Rdh8 Ϫ/Ϫ rods could not be readily described as a first-order process. Comparing the levels of outer segment fluorescence before and after the addition of IRBP, we found that 30 min of IRBP presence resulted in a significant decline ( Fig. 5B; p ϭ 0.01, paired t test). We have not attempted to further characterize the interaction between IRBP and Rdh8 Ϫ/Ϫ rods.
The results shown in Figs. 4 and 5 indicate that IRBP removes all-trans-retinal from outer segments. Upon comparison with the results in Fig. 2, it appears that IRBP is much less efficacious at removing all-trans-retinal than retinol. We have also exam-ined the ability of IRBP to remove retinol and retinal from purified bovine ROS membranes (Fig. 6). Purified ROS membranes lack an NADPH source, and so any all-trans-retinal released after bleaching will not be reduced (spectra labeled 4,5,6 in Fig.  6). With addition of NADPH, however, all-trans-retinal is converted to all-trans-retinol (spectra labeled 1,2,3 in Fig. 6). 2 M IRBP removed a fraction of 0.06 Ϯ 0.01 (mean Ϯ S.D.) of retinol, whereas 5 M IRBP removed a fraction of 0.14 Ϯ 0.03. IRBP was ineffective at removing all-trans-retinal, and for 2 M IRBP, the difference was Ϫ0.01 Ϯ 0.02 and for 5 M IRBP was Ϫ0.03 Ϯ 0.03.
We further examined this difference in efficacy in single bROS using exogenously added retinol and retinal. We loaded exogenous all-trans-retinal and retinol to bROS from Rpe65 Ϫ/Ϫ mice, which contain negligible amounts of endogenous retinoids (33,34); then, we measured their removal in the presence and absence of IRBP (5 M concentration; Fig. 7). For these experiments, first-order kinetics provided a reasonable description of the removal of all-trans-retinal, presumably because of the high retinoid levels loaded (see "Discussion"). In the absence of IRBP, all-trans-retinal fluorescence declined very slowly, giving a removal rate constant of 0.007 Ϯ 0.003 min Ϫ1 (Fig. 7A). By contrast, the rate constant for the removal of all-trans-retinol in the absence of IRBP was ϳ7.4 times larger, 0.052 Ϯ 0.005 min Ϫ1 (Fig. 7B). The presence of 5 M IRBP resulted in faster removal, with rate constants of 0.05 Ϯ 0.01 min Ϫ1 for retinal (Fig. 7A) and 0.390 Ϯ 0.002 min Ϫ1 for retinol (Fig. 7B); the rate constant for retinol removal was ϳ7.8 times larger than that for retinal.

Lower retinol and retinal levels after bleaching in the presence of IRBP
In the retina of the living animal, rod outer segments are continuously bathed in the IRBP of the IPM. We therefore examined the effect of the continuous presence of physiological

All-trans-retinal and retinol clearance by IRBP
concentrations of IRBP on the levels of retinal and retinol attained in rod outer segments after bleaching. Fig. 8A shows the effect of 2 and 5 M IRBP on the levels of outer segment fluorescence (excitation, 360 nm; emission, Ͼ420 nm) attained after bleaching in metabolically intact wild-type rods. Fluorescence levels were significantly reduced in the presence of 2 M IRBP (p Ͻ 0.00001 at 60 min after bleaching), and 5 M IRBP reduced them significantly more than 2 M (p ϭ 0.0036 at 60 min after bleaching). The outer segment fluorescence signal excited by 360 nm is dominated by retinol but does include a contribution from retinal. The retinol and retinal signals can be separated by exciting outer segment fluorescence with 340 and 380 nm (10), and Fig. 8, B and C, shows the effect of IRBP on the individual retinoid levels. Retinol levels were significantly reduced by 2 M IRBP, and 5 M IRBP reduced them significantly more than 2 M (p ϭ 0.042 and p ϭ 0.036 respectively, at 60 min after bleaching; Fig. 8B). Retinal levels were also reduced by IRBP, but the effect was statistically significant only for 5 M IRBP (p ϭ 0.031 at 60 min after bleaching; Fig. 8C).

IRBP prevents formation of lipofuscin precursors
Addition of exogenous, or release of endogenous, all-transretinal in rod outer segments that cannot reduce it to retinol due to lack of access to sufficient amounts of NADPH has been shown to result in the formation of LFP fluorophores (35), likely to be bis-retinoids (36,37). We first examined whether LFPs form after bleaching in the outer segments of metabolically intact rods that lack Rdh8. In the absence of the enzyme, the all-trans-retinal released by photoactivated rhodopsin is not reduced to retinol even though there is no lack of NADPH. Fig.  9A shows that in an intact Rdh8 Ϫ/Ϫ mouse rod, there is an increase in the outer segment LFP fluorescence (excitation, 490 nm; emission, Ͼ515 nm) following rhodopsin bleaching. Results from Rdh8 Ϫ/Ϫ rods (OE, n ϭ 12; Fig. 9B) showed a significant increase in LFP fluorescence after bleaching (p ϭ 0.01; linear regression with slope 0.027 Ϯ 0.007). By contrast, in wildtype mouse rods (E, n ϭ 11; Fig. 9B) there was no significant change in LFP fluorescence levels after rhodopsin bleaching (p ϭ 0.58; linear regression with slope Ϫ0.009 Ϯ 0.015).
Because IRBP can remove all-trans-retinal, and all-trans-retinal can form LFPs, we then examined whether IRBP can prevent LFP formation after rhodopsin bleaching. Formation of LFPs after rhodopsin bleaching can be detected only in outer segments that cannot reduce all-trans-retinal, i.e. outer segments that lack either NADPH or Rdh8. We therefore measured LFP formation in bROS and in intact Rdh8 Ϫ/Ϫ rods. Apart from wild-type rods, we also examined Abca4 Ϫ/Ϫ rods, which accumulate high LFP levels due to the deletion of a transporter protein needed for efficient removal of retinal from outer segment disk membranes (36,38). For an experiment, a dark-adapted rod was selected under infrared light, and the outer segment LFP fluorescence was measured. The rod was then bleached and kept in the dark, and LFP fluorescence was measured 60 min later (Fig. 10, A and B). The experiment was carried out with rods in the presence or absence of 5 M IRBP.
For wild-type bROS, the presence of IRBP resulted in significantly less formation of LFP compared with control (p ϭ 0.0016; Fig. 10C). The presence of IRBP also resulted in less LFP formation in Abca4 Ϫ/Ϫ bROS (0.0048; Fig. 10D). Interestingly, IRBP had no detectable effect on the high LFP levels present in intact Abca4 Ϫ/Ϫ rods (p ϭ 0.467; Fig. 10E). Finally, the presence of IRBP did result in lower LFP levels in intact Rdh8 Ϫ/Ϫ rods, but the difference was not statistically significant (p ϭ 0.05; Fig. 10F).
Our findings are summarized in the schematic in Fig. 11, which illustrates the functional coupling of IRBP and RDH8

All-trans-retinal and retinol clearance by IRBP
that results in efficient removal of all-trans-retinal and alltrans-retinol from photoreceptor outer segments and in decreased formation of LFPs. Also shown are lesser pathways of all-trans-retinal removal, involving direct interaction with IRBP, as well as leakage into photoreceptor inner segments where it is reduced by RDH12 (3,8). In addition to all-transretinal, the broad substrate specificity of RDH12 enables the reduction of 11-cis-retinal, as well as short chain aldehydes resulting from lipid peroxidation. Loss-of-function mutations in the gene encoding RDH12 are associated with Leber congenital amaurosis (39,40).

Discussion
Although the potential of IRBP to facilitate transport of retinoids between photoreceptors and RPE has been clear (20), the specific role of IRBP in visual cycle function has not been fully established. Consistent with the notion that IRBP might not be essential for the transport of retinoids between photoreceptors and RPE (22), this study shows that all-trans-retinol can leave mouse rod outer segments even in the absence of IRBP (Fig. 2), in agreement with previous results (31,32). Early experiments also reported that the recovery kinetics of 11-cis-retinal levels following exposure to light are not significantly affected in mice that lack IRBP (23,41), suggesting that IRBP may have only a marginal role in visual cycle activity. However, more recent studies demonstrated that mice that lack IRBP have slower rhodopsin regeneration kinetics and accumulate all-trans-retinal and retinol in the retina following light exposure (42). In addition, IRBP was shown to play a critical role in cone pigment regeneration (26,42). One possibility that may account for the negative conclusions of the early experiments could be the use of mice harboring a variant of the Rpe65 protein that slows down the visual cycle (42).
This study shows that IRBP facilitates the removal of the all-trans-retinol generated after bleaching from mouse rods (Fig. 2). This is in agreement with previous findings in amphibian photoreceptors (30,43,44). One striking difference is that the rate constants for removal from mouse rods are much higher than those for frog rods (30); for example, the rate constant for removal by 1 M IRBP is 0.020 Ϯ 0.001 min Ϫ1 for mouse compared with 0.007 Ϯ 0.001 min Ϫ1 for frog, whereas for 20 M IRBP it is 0.29 Ϯ 0.05 min Ϫ1 compared with 0.061 Ϯ 0.007 min Ϫ1 . One possible explanation is that with their diameter being about 5-6 times smaller than that of frog, mouse outer segments have a much larger surface-to-volume ratio, which would allow for faster removal of retinol. Another contributing factor may be that the IRBP used for both the mouse and frog experiments is of bovine origin, and it might be more effective for mammalian compared with amphibian rods. Another apparent difference is that in mouse rods the rate constant for retinol removal increases linearly with the IRBP concentration (Fig. 3), whereas in frog rods it appears to level off at . F, metabolically intact rod photoreceptors from Rdh8 Ϫ/Ϫ mice without added IRBP (f, n ϭ 11) and in the presence of 5 M IRBP (Ⅺ, n ϭ 9). Asterisks denote statistical significance. Error bars represent standard deviations. Figure 11. Schematic of the functional coupling of IRBP and RDH8 that facilitates efficient removal of all-trans-retinol (AtROL) from photoreceptor outer segments following rhodopsin bleaching, which is necessary to prevent formation of LFPs. Also shown are lesser pathways of alltrans-retinal (AtRAL) removal, involving direct interaction with IRBP, as well as leakage into the photoreceptor inner segment and reduction by RDH12, a broad specificity reductase for multiple short-chain aldehydes. CRALBP, cellular retinaldehyde-binding protein; 11cRAL, 11-cis-retinal; 11cROL, 11-cis-retinol; scALDs, short-chain aldehydes; scAOLs, short-chain alcohols; RPE, retinal pigment epithelium; OS, outer segment; CC, connecting cilium; IS, inner segment; OLM, outer limiting membrane.

All-trans-retinal and retinol clearance by IRBP
higher concentrations of IRBP (30). This leveling off in frog rods became evident at higher concentrations of IRBP, 50 -100 M; it was consistent with retinol removal involving the binding of IRBP to a plasma membrane receptor with an affinity of ϳ40 M (30). In mouse rods, because of the high rates of retinol removal, it was not possible to accurately determine the removal rate constant for IRBP concentrations greater than 20 M. The linear dependence of the removal rate constant on IRBP concentrations up to 20 M is not inconsistent with the participation of a plasma membrane receptor that has an affinity for IRBP of ϳ40 M. Our experiments, however, do not address the presence of an IRBP partner on the outer segment plasma membrane that may be playing a role in retinol removal (45). The removal of retinol is likely to be a complex multistep process, so its description with first-order rate constants is an approximation.
IRBP also facilitates the removal of all-trans-retinal from rod outer segments, and the removal is enhanced by increased concentrations of IRBP (Fig. 4). In contrast to retinol, not all of the all-trans-retinal generated by light in either bROS (Fig. 4) or in intact Rdh8 Ϫ/Ϫ rod outer segments (Fig. 5) can be removed by IRBP, at least within the time scale of the experiments. Several considerations affect a straightforward interpretation of the results of Figs. 3 and 4. One, the fragility of the bROS limits the time scale of the experiments; two, the weak fluorescence of all-trans-retinal limits the signal-to-noise ratio of the measurements; three, it is possible that endogenously generated alltrans-retinal has specialized/non-specialized interactions with outer segment components that hinder its removal. The experiments of Fig. 7 addressed these concerns using Rpe65 Ϫ/Ϫ mouse bROS, which contain negligible amounts of endogenous retinoids (33,34), and loading them with high concentrations of exogenous all-trans-retinal and retinol. This also dramatically improved the signal-to-noise ratio of the all-trans fluorescence measurements and allowed quantification of the kinetics of retinal removal. Retinal left the bROS ϳ7.4 times more slowly than retinol in the absence of lipophilic carriers, and ϳ7.8 times more slowly than retinol in the presence of 5 M IRBP. IRBP has comparable binding strengths for alltrans-retinal (0.195 M) and all-trans-retinol (0.1-0.2 M) (19), so it is unlikely that they are the explanation for the more efficacious removal of retinol. This is also consistent with the much slower removal of all-trans-retinal found in the absence of IRBP. Thus, the slower removal of all-transretinal appears to be due to its interaction with rod outer segment components, with an obvious possibility being reversible binding to phosphatidylethanolamine, which is present in high concentrations (46). Michael addition reactions with protein amino acid side chains, such as lysine's, is another possibility. The effect of such interactions would be even more pronounced at the lower concentrations of alltrans-retinal generated via release from photoactivated rhodopsin (Figs. 3 and 4). In contrast, all-trans-retinol, does not appear to interact strongly with outer segment components (47), allowing for its faster removal. This interpretation is further supported by the biochemical experiments of Fig. 6, showing that IRBP can remove the all-trans-retinol generated after the photoactivation of rhodopsin, but it is ineffec-tive at removing all-trans-retinal. The difference in removal rates indicates that the reduction of all-trans-retinal to alltrans-retinol facilitates the faster clearance of the photoisomerized rhodopsin chromophore and its recycling. The reduction of all-trans-retinal to -retinol is dramatically impaired in mouse rod photoreceptors that lack both Rdh8 and Rdh12 (3,48). Determination of retinoid levels in the eyes of these mice (48) shows delayed clearance of all-transretinal, consistent with a role for the reduction reaction in the facilitation of the removal.
The concentration of IRBP in the IPM varies significantly across species, at least from 2 to 13 M. In rabbit and rat IPM, it was reported to be 2 and 3 M, respectively (49). The results of Fig. 8 show that concentrations of IRBP in the physiological range, 2 and 5 M, greatly reduce the levels of both all-trans-retinol and all-trans-retinal attained in rod outer segments after bleaching. Consistent with this effect, mice that lack IRBP have higher levels of all-trans-retinal and all-trans-retinol in the retina after light exposure compared with wild-type controls (42).
Accumulation of all-trans-retinal due to metabolic limitations has been shown to result in excess formation of LFPs in rod photoreceptor outer segments (10). All-trans-retinal also accumulates in Rdh8 Ϫ/Ϫ rod outer segments, which lack the enzyme needed for the reduction of retinal to retinol (3,48,50). Fig. 9 shows that in Rdh8 Ϫ/Ϫ rod outer segments, there is also increased LFP formation following light exposure, consistent with the increased accumulation of A2E in Rdh8 Ϫ/Ϫ RPE (50,51). Thus, accumulation of all-trans-retinal, regardless of cause, results in increased LFP formation.
IRBP can partially prevent the excess formation of LFPs that originate from all-trans-retinal accumulation in wildtype and Abca4 Ϫ/Ϫ metabolically compromised rod outer segments (Fig. 10). Because IRBP has no effect on the levels of already formed LFPs (Fig. 10C), its effect is likely to be due to its ability to remove all-trans-retinal. We have not carried out any experiments to directly test a protective role for IRBP, but such a role is consistent with its protective effect against all-trans-retinal cytotoxicity (52), as well as with the increased accumulation of bis-retinoids in the RPE of IRBPdeficient mice (42).
In intact rods with a functioning Rdh8 enzyme, even in the absence of IRBP, it has not been possible to resolve any LFP formation from all-trans-retinal released by photoactivated rhodopsin (Fig. 10C) (10,35). In the intact retina, rods will be in the presence of IRBP, which strongly promotes the clearance of all-trans-retinal (Fig. 8). This makes it unlikely that LFPs originate from free-dwelling all-trans-retinal, and it is consistent with the suggestion that they form from all-trans-retinal trapped in the form of a Schiff base with phosphatidylethanolamine (10). Another possibility would be that a small amount of chromophore remains trapped in aggregated opsin within which the Schiff base is retained.
In summary, we have examined the removal of all-trans-retinol and all-trans-retinal by IRBP from mouse rod outer segments. IRBP is much more efficacious in removing all-transretinol than all-trans-retinal, indicating an important role of the reduction reaction in the clearance of chromophore. The All-trans-retinal and retinol clearance by IRBP extracellular presence of IRBP strongly suppresses the levels of both all-trans-retinol and all-trans-retinal following bleaching and limits the formation of LFPs from accumulated all-transretinal. The ability of IRBP to facilitate the clearance of alltrans-retinal from photoreceptor outer segments is strong evidence of an important role for IRBP in protecting against the cytotoxic effects of retinoids in the outer retina.

Animals
Wild-type (Sv/129) mice were from Harlan Laboratories (Indianapolis, IN). Animals were 2-5 months old, kept in cyclic light with a 12-h light cycle (06:00 -18:00). All animal procedures were carried out in accordance with protocols approved by the Institutional Animal Care and Use Committees of the Medical University of South Carolina and of the University of Michigan, and with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association. For experiments, animals were dark-adapted overnight and sacrificed under dim-red light, and the retinas were excised under either dim-red or infrared light in mammalian physiological solution (in mM: 130 NaCl, 5 KCl, 0.5 MgCl 2 , 2 CaCl 2 , 25 hemisodium-HEPES, 5 glucose, pH 7.40). Isolated rod photoreceptor cells were obtained as described (31).

Materials
Native IRBP was extracted from the soluble IPM fraction of bovine retinas and purified by a combination of concanavalin-A affinity, ion-exchange, and S-300 size-exclusion chromatography (53,54). The concentration of the purified IRBP was determined by absorbance spectroscopy and amino acid analysis. Purified IRBP was shipped to Charleston on dry ice; before experiments, it was dialyzed with physiological solution containing 1 mM DTT (to prevent IRBP aggregation) using Amicon Ultra 0.5-ml centrifugal filters with 50-kDa cutoff (EMD Millipore, Billerica, MA). The final concentration of IRBP was estimated with absorption spectroscopy, using an extinction coefficient of 0.116 ϫ 10 6 M Ϫ1 cm Ϫ1 at 280 nm (55). For an experiment, the solution containing IRBP was prepared to twice the desired final concentration; a volume of the IRBP solution was then added to an equal volume of the solution containing the photoreceptors. The final concentration of DTT, 0.5 mM, resulted in a slight increase in outer segment fluorescence, but control experiments showed that it did not affect the kinetics of retinol and retinal generation and removal.
Other control experiments showed that freezing and thawing of IRBP did not affect its efficacy. All other reagents were from Sigma.

Fluorescence imaging
Fluorescence measurements were carried out at 37°C on the stage of an inverted Zeiss Axiovert 100 microscope (Carl Zeiss, Thornwood, NY) with a 40ϫ oil immersion objective lens (NA ϭ 1.3) (31). For measuring the overall rod outer segment retinol and retinal fluorescence signal, fluorescence was excited with a broadband (40-nm bandwidth) 360-nm filter, and the emission was collected through a long-pass 420-nm filter (emission, Ͼ420 nm). The particular excitation was selected to allow for comparison with extensive previously published data from mice (31,32) as well as frogs (30). Previous work has established that in metabolically intact rods from wild-type and Abca4 Ϫ/Ϫ mice, this fluorescence signal is mostly due to retinol; in metabolically intact rods from Rdh8 Ϫ/Ϫ mice and in metabolically compromised bROS, it is mostly due to retinal (3,6). Biochemical measurements of all-trans-retinol in excised retinas are in good agreement with single-cell fluorescence measurements (31,32). For metabolically intact rods with adequate metabolic substrate, the contributions of retinal and retinol to outer segment fluorescence can be separated by exciting their fluorescence with narrow bandpass (10-nm bandwidth) filters centered at 340 and 380 nm (10). The contribution of retinol is referred to as Fex-340(ROL) and that of retinal as Fex-380(RAL). Experiments using excitation with both 340 and 380 nm have fewer time points, as we tried to minimize measuring light exposure to avoid retinoid photobleaching. LFP fluorescence was measured with 490 nm excitation and emission Ͼ515 nm (35).Formeasurementswithendogenousfluorophores,fluorescence images were initially recorded for the dark-adapted cell; the cell was subsequently bleached with Ͼ530 nm of light for 1 min, and fluorescence images were recorded at different times after bleaching. For measurements with exogenously supplied retinal and retinol to Rpe65 Ϫ/Ϫ bROS, bovine serum albumin was used as a lipophilic carrier (at 1% concentration) (3). In these loading experiments, the bROS is exposed to a bath solution containing the retinoid-carrying BSA for 10 min, and not all of the retinoid on the BSA present in the bathing solution is transferred to the bROS. Image acquisition and analysis were carried out using Slidebook (Intelligent Imaging Innovations, Denver, CO).

Experiments with bovine ROS membranes
ROS were prepared from frozen bovine retinas as described previously (32). Purified membranes were suspended in physiological solution at a rhodopsin concentration of 20 M. An aliquot was removed and diluted to 10 M to serve as the unbleached reference (spectrum 0 in Fig. 6). The rest was bleached at room temperature for 5 min, and divided into six samples. All subsequent manipulations were carried out under dim red light. NADPH (final concentration 100 M) was added to three of the samples as well as IRBP (final concentrations 0, 2, and 5 M). The final rhodopsin concentration was 10 M. After incubating at 37°C for 30 min, samples were centrifuged in an Eppendorf refrigerated 5415R centrifuge for 10 min at 16,100 ϫ g. After two washes with physiological solution to remove NADPH, the resulting pellets were solubilized in 1% N,N-dimethyldodecylamine N-oxide (Ammonyx LO) and spun to remove any unsolubilized material, and the supernatants were scanned on a Cary 300 spectrophotometer. The experiment was repeated three times.

Statistical analysis
Statistical significance was tested with t test and linear regression. In the figures, statistically significant differences are indicated with an asterisk.