Peropsin modulates transit of vitamin A from retina to retinal pigment epithelium

Peropsin is a non-visual opsin in both vertebrate and invertebrate species. In mammals, peropsin is present in the apical microvilli of retinal pigment epithelial (RPE) cells. These structures interdigitate with the outer segments of rod and cone photoreceptor cells. RPE cells play critical roles in the maintenance of photoreceptors, including the recycling of visual chromophore for the opsin visual pigments. Here, we sought to identify the function of peropsin in the mouse eye. To this end, we generated mice with a null mutation in the peropsin gene (Rrh). These mice exhibited normal retinal histology, normal morphology of outer segments and RPE cells, and no evidence of photoreceptor degeneration. Biochemically, Rrh−/− mice had ∼2-fold higher vitamin A (all-trans-retinol (all-trans-ROL)) in the neural retina following a photobleach and 5-fold lower retinyl esters in the RPE. This phenotype was similar to those reported in mice that lack interphotoreceptor retinoid-binding protein (IRBP) or cellular retinol-binding protein, suggesting that peropsin plays a role in the movement of all-trans-ROL from photoreceptors to the RPE. We compared the phenotypes in mice lacking both peropsin and IRBP with those of mice lacking peropsin or IRBP alone and found that the retinoid phenotype was similarly severe in each of these knock-out mice. We conclude that peropsin controls all-trans-ROL movement from the retina to the RPE or may regulate all-trans-ROL storage within the RPE. We propose that peropsin affects light-dependent regulation of all-trans-ROL uptake from photoreceptors into RPE cells through an as yet undefined mechanism.

The opsins are members of the G protein-coupled receptor superfamily. Visual perception begins with the absorption of a photon by an opsin pigment in a rod or cone photoreceptor cell. The light-absorbing chromophore in most visual opsins is 11-cis-retinaldehyde (11-cis-RAL), 3 which is covalently coupled to the protein through a Schiff-base linkage. Absorption of a photon isomerizes the ligand to all-trans-RAL, activating the pigment. After a brief signaling period, the activated metarhodopsin II/III decays by releasing all-trans-RAL into the outer segment (OS) disk membrane. Restoration of light sensitivity to the resultant apo-opsin occurs after it recombines with another 11-cis-RAL. The conversion of all-trans-RAL back to 11-cis-RAL is carried out by multistep enzyme pathways in the retinal pigment epithelium (RPE) (1) and Müller cells of the retina (2)(3)(4). Non-enzymatic photoisomerization of retinyl lipids also contributes to 11-cis-RAL synthesis and visual pigment regeneration (5). Rhodopsin and the cone-opsin visual pigments are considered bleaching opsins because the retinaldehyde ligand dissociates after photoisomerization.
Besides the visual pigments, the mammalian retina also contains several non-visual opsins. These pigments are more closely related to the rhabdomeric photoreceptor opsins of invertebrates than to the vertebrate visual opsins. The best characterized non-visual opsin is melanopsin (Opn4), which is present in a subset of retinal ganglion neurons and mediates circadian photoentrainment (6,7) and part of the pupillary light response (8). Another non-visual opsin is "retinal G proteincoupled receptor" (RGR) opsin, which is expressed in the internal membranes of RPE and Müller glial cells (9). RGR opsin may function as a photoisomerase that regenerates visual chromophore based on the phenotype of mice that lack RGR opsin (10) and its similarity to the known photoisomerase squid retinochrome (11). RGR opsin was also shown to affect light-dependent mobilization of all-trans-retinyl esters (all-trans-REs) in mouse RPE cells (12). Neuropsin (OPN5) is a UV-sensitive bistable pigment also expressed in the vertebrate retina (13,14) that may play an additional role in photoentrainment (15,16).
Currently, very little is known about this protein despite it having been discovered 20 years ago (17). Mammalian peropsin is 28% identical to rhodopsin and 31% identical to melanopsin. Chromophore regeneration studies of peropsin homologues suggest that it binds all-trans-RAL in the dark and converts it to 11-cis-RAL upon exposure to visible light (18,19) akin to RGR opsin (20). Similar to the other non-visual opsins, peropsin is likely a bistable opsin whereby the retinaldehyde chromophore remains coupled to the protein after photoisomerization (21). Peropsin is located exclusively in the apical microvilli of RPE cells (17). These cellular processes interdigitate closely with the OS of rods and cones. The RPE performs several functions critical to the maintenance of photoreceptor viability and light sensitivity (22). Besides hosting the visual cycle for chromophore regeneration, RPE cells provide nutrition to photoreceptors, maintain fluid and electrolyte balance in the retina, and diurnally phagocytose the shed tips of distal rod and cone OS (23,24). A recent report also suggests peropsin expressed in skin confers photosensitivity to short-wavelength light (25).
In the current study, we set out to define the function of a mammalian peropsin. To this end, we generated mice with a null mutation in the gene for peropsin and studied the resulting ocular phenotype. In particular, we examined retinal morphol-ogy in Rrh Ϫ/Ϫ mice by light and electron microscopy. We also determined the content of visual retinoids in eye tissues from wild-type and Rrh Ϫ/Ϫ mice under different light conditions. Finally, we measured protein levels and catalytic activities of visual cycle proteins. The results of these studies indicate that peropsin (i) modulates movement of vitamin A from the retina to the RPE and/or (ii) modulates storage of vitamin A within the RPE.

Generation of mice that lack peropsin
As a first step toward understanding the function of peropsin, we generated Rrh Ϫ/Ϫ mice by homologous recombination in embryonic stem cells. The knock-out construct is shown in Fig. 1A, and confirmation of homologous recombination by Southern blotting is shown in Fig. 1B. Genotyping of offspring by PCR analysis is shown in Fig. 1C. To confirm abrogation of peropsin expression, we performed immunoblot analysis on RPE homogenates from wild-type and Rrh Ϫ/Ϫ homozygous mice. As expected, the band corresponding to the 37-kDa peropsin protein is absent in lanes containing Rrh Ϫ/Ϫ homogenates ( Fig. 2A). Immunohistochemistry of eyecup sections showed an immunofluorescent signal with antiserum to perop-

Peropsin modulates retinoid transit
sin in the RPE of wild-type sections that was absent in sections from Rrh Ϫ/Ϫ mice (Fig. 2B). Besides the RPE, no other cells exhibited peropsin immunofluorescence.

Retinas of Rrh ؊/؊ mice have normal morphology
Photoreceptors are critically dependent on RPE cells for viability. To test whether the absence of peropsin causes photoreceptor degeneration, we examined Rrh Ϫ/Ϫ retina sections by light microscopy. Rrh Ϫ/Ϫ retinas were indistinguishable from wild type (Fig. 3, A and B). Quantification of retinal thickness across the entire eye revealed no loss of photoreceptor nuclei in the outer nuclear layer (Fig. 3E). Electron microscopy of outer retina sections from Rrh Ϫ/Ϫ mice revealed qualitatively normal RPE (Fig. 4, A and B) with normal basal infoldings (Fig. 4, AЈ and BЈ) and apical microvilli (Fig. 4, AЉ and BЉ). Similarly, the photoreceptor OS appeared normal with regular and well aligned disks (Fig. 4, AЉ and BЉ). No debris was present in the subretinal space between the OS and RPE. Loss of peropsin therefore did not affect distal retinal histology or the ultrastructure of photoreceptor OS and RPE cells.

Retinoid dynamics in Rrh ؊/؊ mice
A possible function of peropsin is light-dependent regulation of the visual cycle in RPE cells through interaction with G q or another G protein. Alternatively, because peropsin binds all-trans-RAL in the dark and converts it to 11-cis-RAL upon light exposure (18), peropsin may function as a photoisomerase for synthesis of chromophore to regenerate visual pigments.
To test these possible functions of peropsin, we measured levels of visual retinoids in neural retinas and RPE from 1-month-old wild-type (129/Sv) and Rrh Ϫ/Ϫ mice exposed to different light conditions. Retinoids were examined in mice after overnight dark adaptation, immediately following light exposure, and at different times during recovery in darkness as described previously (26). Wild-type and Rrh Ϫ/Ϫ mice exposed to a 1,000-lux light for 5 min exhibited 2-fold reductions in levels of 11-cis-RAL (Fig. 5A), indicating ϳ50% photobleach of rhodopsin in mice of both genotypes (p values Ͻ0.001). Levels of 11-cis-RAL were similar in wild-type and Rrh Ϫ/Ϫ retinas under all tested conditions (p values Ͼ0.122) (Fig. 5A). These results suggest that peropsin plays no significant role in the synthesis of visual chromophore.
Two notable differences emerged between wild-type and Rrh Ϫ/Ϫ mice in the retinoid dynamics study. First, levels of all-trans-ROL were 2-fold higher in Rrh Ϫ/Ϫ retinas at 5 min after the photobleach (p ϭ 0.003) with reversion toward normal during recovery (Fig. 5B). Second, all-trans-REs were 5-fold lower in dark-adapted Rrh Ϫ/Ϫ RPE (p Ͻ 0.001) (Fig. 5C). Retinyl esters normally increase in the RPE after a photobleach due to uptake and esterification of all-trans-ROL released by the photoreceptors. Retinyl esters increased in Rrh Ϫ/Ϫ RPE after light exposure but more slowly than in wild-type RPE (Fig. 5C). After 45 min of postbleach recovery, all-trans-RE levels were similar in mice of the two genotypes. Besides these changes, levels of visual retinoids were similar between wild-type and Rrh Ϫ/Ϫ retinas and RPE under the various light conditions.
Retinyl esters normally build up in mouse RPE during the first few months of postnatal life (27). To test whether this occurs in RPE without peropsin, we measured all-trans-RE in eyes from dark-adapted wild-type and Rrh Ϫ/Ϫ mice at 1 through 4 months of age. Although all-trans-RE increased 4-fold in wild-type mice, we observed only a slight increase in all-trans-RE in Rrh Ϫ/Ϫ RPE during the same period (p Ͻ 0.001) (Fig. 5D). The failure of Rrh Ϫ/Ϫ mice to build their stores of all-trans-REs with age indicates that loss of peropsin has a longterm impact on vitamin A metabolism in the RPE.

Peropsin modulates retinoid transit Studies on mice lacking both peropsin and IRBP
The phenotypic pattern of elevated all-trans-ROL in the retina and reduced all-trans-RE in the RPE (Fig. 5, B and C) was previously observed in Irbp Ϫ/Ϫ mice that lack IRBP (26,28). IRBP binds all-trans-ROL and 11-cis-RAL during transit through the extracellular space between photoreceptor OS and the apical RPE. Transit of all-trans-ROL from the retina to the RPE is delayed but not blocked in mice lacking IRBP or peropsin, implying the existence of redundant transit pathways. If IRBP and peropsin were components of separate pathways, we would predict a more severe retinoid phenotype in mice lacking both IRBP and peropsin. To test this possibility, we generated mice with null mutations in both the Irbp and Rrh genes. We repeated the retinoid dynamics experiment on Irbp Ϫ/Ϫ and Irbp Ϫ/Ϫ ;Rrh Ϫ/Ϫ mice. Similar to Rrh Ϫ/Ϫ mice, levels of alltrans-ROL were elevated in Irbp Ϫ/Ϫ (p ϭ 0.019) and Irbp Ϫ/Ϫ ; Rrh Ϫ/Ϫ retinas (p Ͻ 0.001) relative to wild-type retinas following exposure to light (Fig. 5B). Furthermore, in a dark-adapted state, the levels of all-trans-RE in Irbp Ϫ/Ϫ and Irbp Ϫ/Ϫ ;Rrh Ϫ/Ϫ RPE were comparable with that in Rrh Ϫ/Ϫ (Fig. 5C). Therefore, loss of both peropsin and IRBP caused similar slowing of all-trans-ROL transit as loss of peropsin or IRBP alone. These observations suggest that peropsin and IRBP may act at similar points in the retinoid transport pathway.

Levels of visual-cycle proteins in retinas and RPE from wildtype and Rrh ؊/؊ mice
After uptake into the RPE cell, all-trans-ROL binds to CRBP1 (29). All-trans-ROL bound to CRBP1 is the preferred substrate for lecithin-retinol acyltransferase (LRAT), which esterifies retinol with a fatty acid from phosphatidylcholine (30,31). Rpe65, the retinoid isomerase in RPE cells, uses all-trans-REs, such as all-trans-retinyl palmitate, as substrate to synthesize 11-cis-ROL (32)(33)(34). Changes in the levels of CRBP1, LRAT, or Rpe65 could affect all-trans-RE levels in Rrh Ϫ/Ϫ mice. Accordingly, we measured levels of these proteins in RPE homogenates from dark-or light-adapted wild-type and Rrh Ϫ/Ϫ mice by immunoblotting. We observed no significant difference in the   (Fig. 6). These data exclude the possibility that loss of peropsin affects the levels of other RPE proteins involved in the synthesis or depletion (or mobilization) of retinyl esters. These data also exclude the possibility that peropsin modulates expression of these proteins in a light-dependent fashion.

Vitamin A uptake and processing by RPE eyecup explants
To determine whether peropsin affects the rate of all-trans-ROL esterification in RPE cells, we measured de novo synthesis of all-trans-RE from exogenous all-trans-[ 3 H]ROL by dark-and light-adapted wild-type and Rrh Ϫ/Ϫ mouse eyecups. Although levels of endogenous retinyl esters differed markedly between wild-type and Rrh Ϫ/Ϫ eyecups (p Ͻ 0.001), we observed no differences in newly synthesized all-trans-[ 3 H]RE between dark-or light-adapted wild-type or Rrh Ϫ/Ϫ eyecups (p ϭ 0.900) (Fig. 7). These data suggest that the lower retinyl esters in Rrh Ϫ/Ϫ versus wild-type RPE are not caused by reduced retinyl ester synthesis.

Vitamin A processing in RPE homogenates
To measure retinyl ester synthesis and hydrolysis rates directly, we prepared protein homogenates from dark-and light-adapted eyecups from Rrh Ϫ/Ϫ or 129/Sv mice. For retinyl ester synthesis, we added all-trans-ROL substrate to the homogenates and measured the amount of retinyl ester formed by HPLC (Fig. 8A). We observed no difference in the rates of ester formation in eyecup homogenates from mice of either genotype (p ϭ 0.295). We also observed no difference in the rates of ester synthesis by homogenates from dark-or light-adapted mice (p values Ͼ0.055). These results are in agreement with those from the explant experiments, providing further evidence that retinyl ester synthase activity in RPE cells is not affected by loss of peropsin.
A third possible explanation for the low retinyl esters in the RPE of dark-adapted Rrh Ϫ/Ϫ mice is increased retinyl ester hydrolase activity. The protein responsible for simple hydrolysis of all-trans-REs in the RPE has not yet been identified. To assay for this activity, we prepared RPE-containing eyecup homogenates from dark-or light-adapted wild-type and

Peropsin modulates retinoid transit
Rrh Ϫ/Ϫ mice. These homogenates were incubated with alltrans-retinyl acetate (all-trans-RAc), a relatively soluble form of retinyl ester. We removed aliquots at different times during the incubations and quantitated remaining all-trans-RAc by HPLC. Approximately 90% of the all-trans-RAc substrate was consumed during the 30-min incubation (Fig. 8B). The rates of all-trans-RAc depletion were similar in homogenates from the four mouse experimental conditions (Fig. 8B), although the data did reveal a statistically significant difference (p values Ͻ0.050): retinyl ester hydrolysis activity in Rrh Ϫ/Ϫ homogenates was actually slightly slower that the 129/Sv homogenate control. Therefore, these differences do not explain our in vivo observations and may not be biologically relevant. Furthermore, we observed no differences in the rates of retinyl ester hydrolysis between homogenates from dark-adapted versus light-exposed mice. These results suggest that peropsin plays no role in regulating retinyl ester hydrolysis.

Discussion
Peropsin is a non-visual opsin of ancient origin in both vertebrates and invertebrates. In the vertebrate eye, it is expressed in the apical microvilli of RPE cells facing the OS of photoreceptors across the interphotoreceptor matrix (IPM). Unlike the visual opsins in rods and cones, invertebrate peropsin homologues bind all-trans-RAL in the dark and isomerize it to 11-cis-RAL upon exposure to light (18,19). This "reverse" photoisomerase activity of peropsin is similar to that of RGR opsin (10) and squid retinochrome (35), which are thought to synthesize 11-cis-RAL chromophore for regeneration of visual opsins. Peropsin is unlikely to serve this function, however, because wild-type and Rrh Ϫ/Ϫ retinas contained similar levels of 11-cis-RAL following a photobleach (Fig. 5A). Based on sequence homology, peropsin is likely a bistable pigment whereby the chromophore remains covalently coupled to the opsin after photoisomerization (36). Peropsin includes the conserved (E/D)R(Y/W/F) and NPXXYX 5,6 F motifs, which are critical for the interaction of a G protein-coupled receptor with its G protein (37,38). These observations suggest that peropsin is a light-signaling molecule in the RPE. A recent report suggests vertebrate peropsin may also act as a light sensor in skin keratinocytes (25). The normal retinal anatomy observed in Rrh Ϫ/Ϫ mice (Figs. 3 and 4) indicates that peropsin is not required for photoreceptor viability or the maintenance of RPE ultrastructure. Rods and cones diurnally shed their distal OS, which are phagocytosed by cells of the RPE (23). Animals with impaired OS phagocytosis, such as the Mertk Ϫ/Ϫ mouse (39), exhibit whorl-like debris in the subretinal space between the OS and RPE cells and rapid photoreceptor degeneration (39). Neither of these morphologic features were present in Rrh Ϫ/Ϫ mice, suggesting no essential or non-redundant role for peropsin in OS phagocytosis.
Because chromophore consumption by photoreceptors is coupled to light intensity, RPE cells may use photometric information from peropsin to regulate visual retinoid metabolism. To explore this possibility, we determined visual retinoids in retinas and RPE from normal and Rrh Ϫ/Ϫ mice overnight darkadapted and at different times after a bright light. These studies revealed two persistent changes with loss of peropsin: (i) levels of all-trans-ROL were 2-fold higher in Rrh Ϫ/Ϫ versus wild-type retinas after the photobleach (Fig. 5B), and (ii) retinyl esters were 5-fold lower in dark-adapted Rrh Ϫ/Ϫ versus wild-type RPE (Fig. 5C). Following light exposure, photoreceptors release alltrans-ROL into the IPM where it binds IRBP (40) and transits the short distance to the RPE. On the apical surface of the RPE, which does not contain the Stra6 receptor for retinol-binding protein (41), the mechanism of all-trans-ROL uptake from the IPM is not clear. It may involve a docking or receptor-mediated interaction with another protein on the villous membrane surface. The greater affinity of all-trans-ROL to CRBP1 inside the cell drives uptake of all-trans-ROL (1), which is rapidly fattyacylated by LRAT to form a retinyl ester. The high activity and low K m of LRAT toward its all-trans-ROL substrate (42) means that all-trans-REs are an indicator of all-trans-ROL, which is usually present at low concentrations in RPE cells. Mice lacking IRBP (26) or CRBP1 (43) exhibit elevated all-trans-ROL in the retina and reduced all-trans-REs in the RPE due to impaired transit of all-trans-ROL from photoreceptor OS to RPE. The similar retinoid phenotypes in Rrh Ϫ/Ϫ , Irbp Ϫ/Ϫ , and Crbp1 Ϫ/Ϫ mice supports the hypothesis that peropsin plays a role in the movement of all-trans-ROL from photoreceptors to the RPE.
We sought to identify the cause of low retinyl esters in the RPE of Rrh Ϫ/Ϫ mice. Three proteins directly affect all-trans-RE levels in the RPE: CRBP1, LRAT, and Rpe65. These proteins were present at similar abundances under the four conditions of light exposure and genotype (Fig. 6), ruling out the possibility that light exposure or loss of peropsin affects CRBP1, LRAT, or Rpe65 abundance. The rates of all-trans-RE synthesis by RPEcontaining eyecups were also similar under these four conditions (Fig. 7). Enzyme assays on RPE homogenates from wildtype and Rrh Ϫ/Ϫ mice in light-or dark-adapted states also showed no difference in LRAT catalytic activity (Fig. 8A). Finally, we found that the rates of retinyl acetate hydrolysis in RPE homogenates from dark-or light-adapted wild-type and Rrh Ϫ/Ϫ mice were similar (Fig. 8B). Together, these findings suggest that the lower retinyl esters in Rrh Ϫ/Ϫ RPE are probably not caused by changes in retinyl ester synthase or hydrolase activities of the visual cycle but may reflect regulation by peropsin of all-trans-ROL uptake into the RPE.
Several possible mechanisms that explain peropsin modulation of all-trans-ROL uptake into RPE cells are consistent with the results presented here. (a) Peropsin may serve as the receptor or docking site for holo-IRBP on the RPE apical microvillus membrane. This would place IRBP in position to transfer its all-trans-ROL cargo to CRBP1 inside the cell. The affinity of IRBP binding to peropsin may be regulated by ambient light. (b) Peropsin may modulate the affinity of the IRBP receptor on the apical RPE surface for its IRBP ligand in a light-dependent fashion. The IRBP receptor has not been identified. (c) Peropsin may modulate by light the affinity of CRBP1 inside the cell to the IRBP receptor, to phospholipid headgroups on the inner villous membrane, or to cytoskeletal proteins associated with the inner villous membrane. Any of these actions would inhibit transfer of alltrans-ROL from IRBP to CRBP1 and hence uptake of all-trans-ROL into the RPE cell. (d) Finally, peropsin may modulate the affinity of CRBP1 for its all-trans-ROL ligand or the affinity of holo-CRBP1 to LRAT for esterification of all-trans-ROL.
How does regulating uptake of all-trans-ROL into RPE cells benefit the visual system? Given the low K m of LRAT for alltrans-ROL (42) and the low turnover rate of Rpe65 (32,44), the RPE visual cycle supplies chromophore to rods in dim light with low fluxes of all-trans-ROL. In contrast, the retinoid isomerase in Müller cells, DES1, has a much higher turnover rate than does Rpe65 (3,4). Furthermore, Müller cells release the chromophore precursor, 11-cis-ROL, into the IPM (3,45). Because only cones can utilize 11-cis-ROL to regenerate visual pigments (44,46), the Müller cell visual cycle is optimized to provide cones with chromophore precursor under bright light conditions. In dim light, peropsin may accelerate uptake of all-trans-ROL into RPE cells, diverting all-trans-ROL from the Müller cell pathway. In the absence of peropsin, all-trans-ROL uptake by RPE cells is reduced, explaining the lower all-trans-RE levels in dark-adapted Rrh Ϫ/Ϫ mice (Fig. 5C). In bright light, peropsin may limit uptake of all-trans-ROL into the RPE, diverting it to Müller cells. Thus, peropsin may serve as a diverter valve, send-

Peropsin modulates retinoid transit
ing all-trans-ROL to the RPE or Müller cell visual cycle depending on light conditions.

Animal use and care statement
This study was carried out in accordance with recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and the Association for Research in Vision and Ophthalmology statement for the use of animals in ophthalmic and vision research. Euthanasia was performed by cervical dislocation on deeply anesthetized (10 mg/kg xylazine and 100 mg/kg ketamine) mice to minimize pain and distress.

Targeted disruption of the Rrh gene and generation of Rrh ؊/؊ mice
A mouse Rrh genomic fragment was used to construct a targeting vector (Fig. 1A). We replaced a 4.3-kb fragment starting at the initiator methionine in exon 1 and encompassing the 5Ј 60% of intron 1 with a LacZ-coding region followed by a neomycin resistance cassette (PGK-Neo). This insert deletes the first 36 codons of the Rrh open reading frame (present within exon 1), removes the exon 1-intron 1 splice donor site, and introduces two transcription termination sites into the transcription unit: one at the 3Ј-end of the LacZ reporter and a second one at the 3Ј-end of the PGK-Neo cassette. Within that part of the Rrh open reading frame that remains in the targeted allele (starting in exon 2), the next in-frame ATG codon (i.e. potential initiator methionine codon) is at codon 81 between the second and third predicted transmembrane segments. If transcription proceeded through the remaining Rrh exons and the resulting mRNA were to be translated, the translation product would include only five of the seven predicted transmembrane segments. Thus, the knock-out allele is very likely to completely eliminate Rrh function. The 5Ј homology arm extends from the initiator methionine codon in exon 1 to an EcoRI site 6.7 kb 5Ј of that location; the 3Ј homology arm extends 2.6 kb from a HindIII site in intron 1, 60% of the distance from exon 1, to the 3Ј-end of exon 2. The thymidine kinase gene was inserted in the vector adjacent to the 3Ј homology arm to counterselect against non-homologous recombinants with ganciclovir treatment. The targeting vector was linearized at the 5Ј edge of the 5Ј homology arm and electroporated into ES cells. After selection, the G418-and ganciclovirresistant clones were screened by Southern blotting and introduced into blastocysts. The progeny of the resulting chimeric mice were screened by Southern blotting (Fig. 1B). Genomic DNA was digested with KpnI and probed with a fragment from Rrh exon 2 and a fragment from LacZ. Tail DNA was analyzed by PCR to detect the presence of the Neo cassette (5Ј-GGCCA-CAGTCGATGAATCCAGAA-3Ј and 5Ј-TGATGCCGCCGT-GTTCCGGCTGT-3Ј) or the Rrh wild-type allele (5Ј-TGAAC-TTACACACCAGATCTGCAG-3Ј and 5Ј-CTGAGTGA-GGCTTCGGATTTTAAC-3Ј). Homozygous mutants were crossed with 129/Sv mice for at least eight generations and confirmed to be on a background containing the normal Rpe65 Leu/Leu-450 and wild-type Rd8 alleles. It should be noted that for all studies comparisons were made using 129/Sv and Rrh Ϫ/Ϫ of a semimixed background.

Antibodies and immunoblotting
A custom immunogen affinity-purified antibody was made in rabbits by immunizing with a peptide corresponding to the carboxyl-terminal 41 residues of mouse peropsin (CHKKFRK-AMLAMFKAQPHLAVPEPSTLPMDMPQSSLAPVRI) conjugated to keyhole limpet hemocyanin (YenZym, San Francisco, CA). Anti-Rpe65 antiserum was prepared as described previously (47). Anti-LRAT antiserum was a gift from Palczewski and co-workers (48). Protein extracts from dissected eyecups were homogenized in phosphate-buffered saline (PBS) ϩ 0.05% n-dodecyl ␤-D-maltoside containing protease inhibitor, incubated with Benzonase nuclease (Invitrogen) overnight at 4°C, and clarified by microcentrifugation. SDS-polyacrylamide gels were run with 20 g of protein/lane, transferred to PVDF membrane (Millipore), and blocked with Odyssey Blocking Buffer (LI-COR Biosciences) for 1 h. Membranes were incubated with primary antibody (1:1,000) overnight at 4°C followed by three washes in PBS ϩ 0.1% Tween 20 (PBST). Secondary antibodies (donkey anti-mouse IRDye 680 or donkey anti-rabbit IRDye 800, LI-COR Biosciences) were applied for 1 h at 1:10,000 dilutions followed by three washes in PBST and imaged using a LI-COR Biosciences Odyssey CLx infrared imaging system.

Immunohistochemistry
After euthanizing the mice, a corneal window was cut into whole enucleated eyes and fixed for 30 min at room temperature or overnight at 4°C in 4% paraformaldehyde diluted in PBS. Eyecups were marked for orientation, and the anterior section and lens were removed. Tissues were cryopreserved first in 10% and then 30% sucrose solutions in PBS before embedding in OCT (TissueTek) and cut into 10-m sections. Sections were prepared as described previously (4). Primary (YenZym) and secondary (Alexa Fluor 488 donkey anti-rabbit) antibodies were used at 1:100 dilutions. Images were collected using an Olympus FluoView FV1000 confocal laser-scanning microscope equipped with 40ϫ and 60ϫ oil immersion objectives. Images were minimally processed using Fiji software (49).

Light and electron microscopy
Mice under deep anesthesia with isoflurane were fixed by intracardiac perfusion with 2% formaldehyde and 2.5% glutaraldehyde in 0.1 m sodium phosphate buffer, pH 7.2. Light cautery was applied at the superior pole of the cornea to mark the orientation before enucleation of the eyeball. The eyecup was trimmed into temporal and nasal hemispheres. All tissues were immersed for 1 h in 1% osmium tetroxide dissolved in 0.1 M sodium phosphate buffer (pH 7.4) followed by dehydration in a graded series of alcohols. The temporal hemispheres were embedded in an Epon/Araldite mixture. Light microscope sections were cut at 1 m, stained with 1% toluidine blue and 1% sodium borate, and then photographed with a Zeiss Axiophot microscope fitted with a Planapo 20ϫ lens and a CoolSNAP digital camera. The nasal hemispheres were cut in quadrants and embedded in Araldite 502 (Ted Pella, Redding, CA) for electron microscopy. Ultrathin sections for electron microscopy were cut on a Leica Ultracut ultramicrotome. The sections were collected on 200-mesh copper grids and stained with uranium and lead salts before viewing on a Zeiss 910 electron microscope. Electron Peropsin modulates retinoid transit microscopic images were acquired from the superior nasal quadrant sections using a KeenView TM digital camera.

Retinoid dynamics assay
Overnight dark-adapted mice were anesthetized using a mixture of xylazine and ketamine at 10 mg/kg and 100 mg/kg, respectively, and pupils were dilated using eye drops containing 2.5% phenylephrine HCl solution. Mice were bleached with 1,000-lux white light for 5 min in a Ganzfield dome and returned to dark for 0, 5, 15, 45, and 90 min before sacrifice, enucleation, and dissection of RPE eyecup and retina. Equal numbers of male and female mice were analyzed at 4 -6 weeks of age. Measurements represent analysis from a single eye from four mice for each time point. Tissues were homogenized using a glass-glass homogenizer in 500 l of PBS ϩ 200 mM hydroxylamine ϩ protease inhibitor and analyzed for retinoids by HPLC as described previously (12).

All-trans-[ 3 H]retinol uptake in mouse eyecups
Eyecups were prepared from 129/Sv and Rrh Ϫ/Ϫ overnight dark-adapted mice. Eyes were enucleated, and the anterior section and lens were removed. The retina was separated, and the eyecup containing RPE was washed in PBS at room temperature. The eyecups were then submerged in a solution containing 5 M holo-IRBP with 1 eq of all-trans-[ 3 H]retinol (ϳ10,000 dpm/pmol) in DMEM for 30 min at 37°C followed by three washes in PBS. Eyecups were homogenized using a glass-glass homogenizer in a total volume of 500 l of PBS and extracted twice in cold hexane/methanol (1.5 ml each). The non-polar phase containing the retinoid was dried, resuspended in 100 l of hexane, and analyzed by HPLC equipped with a Packard 500TR flow scintillation counter calibrated for tritium.

LRAT assay
RPE eyecups were collected from overnight dark-adapted mice or mice additionally exposed to 1,000-lux UV-filtered white light for 2 h. All subsequent procedures were performed under dim red light. Homogenates were prepared in PBS ϩ protease inhibitor, and cells were disrupted using a glass-glass homogenizer. Samples contained 10 g of protein homogenate in 10 mM Tris (pH 8.0), 2 mM CaCl 2 , 2 mM MgCl 2 , 1 mM DTT, and 1% BSA. After 2 min of preincubation in a 37°C water bath, the reaction was initiated with the addition of 2 M all-transretinol delivered in DMSO in the dark. The reaction was quenched after 4 min with cold methanol and extracted with hexane for retinoid analysis.

Retinyl ester hydrolase assay
RPE homogenates were prepared as above. Samples contained 10 g of protein homogenate in 50 mM Tris (pH 8.0) with 1% BSA. After 2 min of preincubation in a 37°C water bath, the reaction was initiated with the addition of 10 M all-trans-RAc delivered in ethanol in the dark. Aliquots were removed at 0, 1, 5, 15, and 30 min from the bulk reaction mixture and quenched with cold methanol, and the retinoids were extracted in hexane.

Statistical analysis
Data were analyzed in the Statistical Package for Social Scientists, version 24 (SPSS; IBM), using separate but parallel analyses of variance (linear mixed models) with the relevant measure (retinal thickness, 11-cis-RAL, all-trans-ROL, all-trans-RE, all-trans-RAc, or pixel density) as the dependent variable and genotype, time, and light status as independent variables as necessary. Post hoc tests were performed only when the initial omnibus tests showed significance. Statistical significance was defined as p Ͻ 0.050.