Disturbed retinoid metabolism upon loss of rlbp1a impairs cone function and leads to subretinal lipid deposits and photoreceptor degeneration in the zebrafish retina

The RLBP1 gene encodes the 36 kDa cellular retinaldehyde-binding protein, CRALBP, a soluble retinoid carrier, in the visual cycle of the eyes. Mutations in RLBP1 are associated with recessively inherited clinical phenotypes, including Bothnia dystrophy, retinitis pigmentosa, retinitis punctata albescens, fundus albipunctatus, and Newfoundland rod–cone dystrophy. However, the etiology of these retinal disorders is not well understood. Here, we generated homologous zebrafish models to bridge this knowledge gap. Duplication of the rlbp1 gene in zebrafish and cell-specific expression of the paralogs rlbp1a in the retinal pigment epithelium and rlbp1b in Müller glial cells allowed us to create intrinsically cell type-specific knockout fish lines. Using rlbp1a and rlbp1b single and double mutants, we investigated the pathological effects on visual function. Our analyses revealed that rlbp1a was essential for cone photoreceptor function and chromophore metabolism in the fish eyes. rlbp1a-mutant fish displayed reduced chromophore levels and attenuated cone photoreceptor responses to light stimuli. They accumulated 11-cis and all-trans-retinyl esters which displayed as enlarged lipid droplets in the RPE reminiscent of the subretinal yellow-white lesions in patients with RLBP1 mutations. During aging, these fish developed retinal thinning and cone and rod photoreceptor dystrophy. In contrast, rlbp1b mutants did not display impaired vision. The double mutant essentially replicated the phenotype of the rlbp1a single mutant. Together, our study showed that the rlbp1a zebrafish mutant recapitulated many features of human blinding diseases caused by RLBP1 mutations and provided novel insights into the pathways for chromophore regeneration of cone photoreceptors.

Mutations in RLBP1 have been associated with a group of autosomal recessive retinal diseases comprising autosomal recessive retinitis pigmentosa (Maw et al., 1997), Bothnia dystrophy (Burstedt et al., 1999), Newfoundland rod-cone dystrophy (Eichers et al., 2002), retinitis punctata albescens (RPA) (Morimura et al., 1999), and fundus albipunctatus (Katsanis et al., 2001). These retinal disorders differ in certain aspects of the disease such as time of onset, severity, and progression. However, they all cause prolonged dark adaptation and night blindness in early childhood. Upon progression, the involvement of the macula and RPE is possible, resulting in loss of central vision and legal blindness at the age of 20-40 (Burstedt and Golovleva, 2010). Another interesting common feature in patients with RLBP1-associated disease is a characteristic fundus appearance with homogeneously distributed, punctate white-yellow deposits. It was found that impaired CRALBP function can be caused by pathogenic mutations that either tighten or weaken the interaction of CRALBP with 11cisRAL . The crystal structure of wild-type and mutant CRALBP provided insights into the accompanying molecular changes within the protein (He et al., 2009).
The involvement of cone photoreceptors at later stages of the disease makes it imperative to elucidate the role of CRALBP in visual pigment regeneration. While visual pigment regeneration in rods is very well established (e.g., reviewed in Kiser et al., 2014), the processes involved in 11cisRAL recycling in cones are less well understood. Besides the canonical visual cycle, a Müller glia cell-dependent intraretinal visual cycle has been proposed (Fleisch et al., 2008;Collery et al., 2008;Wang et al., 2009;Xue et al., 2015;Sato and Kefalov, 2016) as well as recycling pathways via light-dependent mechanisms both in the RPE and MGCs (Kolesnikov et al., 2021;Morshedian et al., 2019;Kaylor et al., 2017;Babino et al., 2015). Together, these pathways might meet the demand for chromophore supply of cones during bright daylight illumination and prevent competition between cone and rod photoreceptor for the unique chromophore.
In contrast to mammals, the zebrafish eyes harbor duplicates of Cralbp and express rlbp1a in RPE cells and rlbp1b in MGCs in a nonoverlapping pattern (Fleisch et al., 2008;Collery et al., 2008). In the current study, we exploited the fish to target Cralbp in an intrinsically cell-specific manner and investigated the contributions of the RPE and the MGC to cone visual pigment regeneration. The cone-dominant retina of the zebrafish and the practically exclusively cone-driven retinal responses in larval stages (Branchek, 1984;Bilotta et al., 2001) facilitated the investigation of cone-specific defects upon knockout of either rlbp1 paralog in larvae up to 5 days postfertilization (dpf). Moreover, our model has the advantage over previous morpholino-mediated knockdown studies in that we could assess the progression of the disease beyond 5 dpf. Importantly, our fish model recapitulated the key characteristics of the human disease, including progressive retinal thinning and the appearance of subretinal lipid deposits, resembling the yellow-white dots in patient fundi.

Generation of intrinsically cell-specific KO lines using CRISPR/Cas9
The nonoverlapping expression pattern of rlbp1a and rlbp1b in the zebrafish retina allowed us to target each Cralbp separately and to create intrinsically cell-specific knockout (KO) lines. The rlbp1a−/− line harbors a 19 bp deletion in exon four causing a frameshift and an early stop codon. The rlbp1b−/− line displays an in-frame 18 bp deletion in exon 4 (see Methods Table 2 for sequences). Both KOs were confirmed in retinal sections using immunohistochemistry with a CRALBP antiserum ( Figure 1). In wild types, CRALBP immunofluorescence is present in the RPE cells and the MGCs. In rlbp1a−/−, the staining is only present in MGCs, while in rlbp1b−/−, staining is restricted to the RPE. Double-KO animals were generated by crossing the two single-mutant lines. The loss of Cralbp proteins in the retina of double-KO was confirmed by immunohistochemistry. Both single-KO lines had reduced levels of mRNA of the targeted gene, indicative of nonsense-mediated decay (Figure 1B and C). Notably, the expression of the respective paralogs was not increased. Therefore, transcriptional adaptation, often observed in zebrafish knockout models (Rossi et al., 2015), through upregulation of the paralogs, was not observed for Cralbps.

Chromophore regeneration through 11-cis-retinyl esters is dependent on Cralbpa
To assess the effect of rlbp1a knockout on ocular retinoids, we determined their concentrations and composition in adult eyecups and 5 dpf whole larval extracts of mutants and controls. The term 11cisRAL always refers to protein-bound (e.g., by opsin or Cralbp) and free 11cisRAL in our experiments. Representative high-performance liquid chromatography (HPLC) traces are shown in  expression was similar to wild type [t(4) = 0.849, p = 0.444], but rlbp1b mRNA levels were twofold reduced [t(4) = 2.84, p = 0.094]. Statistics: t-test comparing KO vs. wild type for each target with Benjamini-Hochberg correction for multiple testing. Significance: **p < 0.01. Abbreviations: doubleknockout (dbKO), retinal pigment epithelium (RPE), outer segment (OS), outer nuclear layer (ONL), outer plexiform layer (OPL), and inner nuclear layer (INL). See Figure 1-source data 1 for expression data.
The online version of this article includes the following source data for figure 1: Source data 1. Source data for Figure 1.
Chromophore levels in heterozygotes were comparable to wild types ( Figure 2A, top panel). In rlbp1a−/− adults, 11cisRAL concentration was about threefold reduced when compared to wild-type siblings ( Figure 2A, top panel). In DA rlbp1a−/− larvae, 11cisRAL levels were halved relative to wild type ( Figure 2B, top panel). In line with previous reports (Babino et al., 2015), wild-type larvae maintained high levels of 11cisRAL even upon prolonged exposure to BL ( Figure 2B, top panel). After one hour of redark adaptation, wild-type larvae regenerated the chromophore to levels of dark-adapted animals. In rlbp1a−/− larvae, ambient illumination caused a steep decline of 11cisRAL when compared to controls. After 1 hr of BL bleach, 11cisRAL levels were further reduced. Upon redark adaption, rlbp1a−/− larvae regenerated 11cisRAL to only 57 % of the level observed in dark-adapted larvae.
We also determined the concentration of 11cisREs which are Rpe65 dependently produced in the fish and serve as a reservoir for chromophore precursors under photopic conditions (Babino et al., 2015). Eyes of adult rlbp1a−/− fish displayed significantly higher 11cisRE concentrations than their wild type and heterozygous siblings (Figure 2A, middle panel). We observed a similar trend in larval samples. However, the effect was not as pronounced as in the adult eye ( Figure 2B, middle panel). Illumination with BL led to a decrease in 11cisRE in both wild type and rlbp1a−/− larvae ( Figure 2B, middle panel). Based on this observation and the reduced chromophore concentration in rlbp1a−/− larvae we assume that the esters were hydrolyzed upon light exposure but were not utilized for 11cisRAL production. Instead, we observed an increased concentration of atRE in rlbp1a−/− adult fish ( Together, our biochemical analyses showed that rlbp1a−/− adult and larval fish display a significant reduction of 11cisRAL concentration under all light conditions and an accumulation of 11-cis and atRE.

Effects of Cralbpb on chromophore regeneration
Expressed in MGCs, rlbp1b was implicated as a critical component of the proposed intraretinal visual cycle (Xue et al., 2015;Sato et al., 2017). Therefore, we investigated putative effects brought up by the rlbp1b knockout on ocular retinoid concentrations and composition. We detected comparable amounts of 11cisRAL in rlbp1b−/−, rlbp1b+/−, and wild-type siblings ( Figure 2C, top panel). The eyes of the rlbp1b−/− and rlbp1b±fish displayed lower 11cisRE and atRE levels than the wild-type control siblings ( Figure 2C, middle and bottom panel).
We next challenged larval eyes with BL illumination. Subsequently, we extracted retinoids and subjected them to HPLC analysis. We observed that genetic disruption of rlbp1b did not alter 11cisRAL concentrations in DA, bleached, and bleached and redark-adapted rlbp1b−/− larvae when compared to controls ( Figure 2D, top panel). Also, retinyl ester levels at this stage were not significantly altered ( Figure 2D, middle and bottom panel).
In adult double mutant fish, we found a reduction in 11cisRAL levels ( Figure 2E, top panel). Additionally, these fish displayed high levels of RE ( Figure 2E, middle and bottom panel).
In conclusion, rlbp1b deficiency did not affect ocular retinoid concentration in such a dramatic fashion as observed in rlbp1a-deficient eyes. Additionally, the double mutant displayed decreased 11cisRAL and elevated RE concentrations when compared to wild-type controls. This pattern of alterations in ocular retinoid content was comparable to that found in rlbp1a−/− fish.

Cone-specific retinal function in larvae is dependent on Cralbpa
To investigate the contribution of the two Rlbp1 genes on cone photoreceptor responses, we measured DA white-light electroretinogram (ERG) in 5 dpf larvae. At this developmental stage, ERG responses are exclusively cone driven in zebrafish (Branchek, 1984;Bilotta et al., 2001). As a readout for retinal function, we used the b-wave amplitude generated by ON-bipolar cells. Maximum amplitude was assessed after stimulation with increasing light intensities within a range of five log units. Representative traces can be found in Figure 3-figure supplement 1 .
We found a significant effect of genotype on b-wave amplitude in the rlbp1a-mutant line. In rlbp1a−/− larvae, amplitudes were significantly reduced when compared to heterozygous and  wild-type siblings ( Figure 3A). Knockout of rlbp1b did not affect ERG responses under the applied conditions ( Figure 3B).
To explore the functional consequences of Cralbpa deficiency during prolonged photopic illumination, we measured light-adapted ERG responses with constant background illumination (LA) ( Figure 4A, left panel). In addition, we determined cone responses directly after a 1 hr bleach with BL ( Figure 4A, middle panel). These conditions paralleled our retinoid analysis protocol. To investigate recovery of rlbp1a−/− larvae after such an intense bleach, we redark-adapted larvae for 1 hr after the bleach (reDA) ( Figure 4A, right panel). The prolonged illumination reduced the sensitivity in mutant and wild-type larvae when compared to the DA state. For comparison, the dotted line represents the mean DA b-wave amplitude of wild types at 0.01 % light intensity.
The responses of rlbp1a−/− were diminished under all light conditions when compared to wild types. The sensitivity was shifted approximately one log unit and the maximum amplitude remained below wild-type levels in all illumination conditions examined. Notably, the responses of cone photoreceptors were not completely absent in the mutant, even after exposure to bright bleaching light. Also, we observed substantial recovery of retinal function following redark adaptation, albeit with reduced sensitivity.
This prompted us to investigate the possibility that rlbp1b compensates for the lack of rlbp1a. Therefore, we repeated the experiments with double-KO larvae. Again, we found a one log unit shift in sensitivity and reduced maximum amplitudes in DA double-KO responses ( Figure 3C). Also under photopic conditions ( Figure 4B), the phenotype was similar to the one seen in rlbp1a−/−. However, immediately after a strong bleach, but also following redark adaptation, the sensitivity shifted more than one log unit in double-KO compared to the respective wild-type controls. Interestingly, responses were severely attenuated but still recordable at 100 % light intensity in the bleach condition.
Together, our analyses demonstrated that cone function depends on RPE-expressed Cralbpa. We observed no pronounced changes of cone responses in DA Cralbpb-deficient larval eyes and only a minor additive effect in the double mutant under the applied conditions.

Knockout of RPE-expressed Cralbpa leads to subretinal lipid deposits and age-dependent photoreceptor degeneration
Histological examination of rlbp1a −/− larvae at 5 dpf revealed an overall normal stratification of the different retinal layers ( Figure 5A and B). In the ventral retinal periphery, we found fewer rod outer segments ( Figure 5, black arrows), indicating either slower development or an early loss of rod outer segments (OSs). This observation is further supported by the decrease in rhodopsin mRNA levels at this age Figure 6A. The thickness of the outer nuclear layer (ONL), containing the photoreceptor nuclei, was unchanged ( Figure 5G1). However, already at this developmental stage, round subretinal lesions became detectable that appeared colorless upon staining with Richardson's solution ( Figure 5B2 and H1). This phenotype became more apparent at 34 dpf ( Figure 5C, D and H2) and in 6-month-old fish ( Figure 5E, F and H3). The lesions appeared to increase not only in number but also in size. Occasionally, we detected these structures in wild-type siblings ( Figure 5C2). However, these tended to be smaller and sparser.
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Source data for Figure 2.  Importantly, we also found that already at 34 dpf, the OSs of both rods and cones were severely shortened, dysmorphic, or sometimes completely absent ( Figure 5D). The double-KO ( Figure 7) largely recapitulated the findings in rlbp1a−/− fish. In adults, it appeared that rod outer segments were slightly better preserved in double-KO than in the rlbp1a−/− single mutant (compare Figures 5F,  7F). Analysis of rhodopsin mRNA expression revealed lower expression at 5 dpf and in adults in Statistical analysis was performed on data in (A2, C2) after square root transformation to mitigate the impact of outliers. Pairwise comparisons between wild type (WT) and mutants at each light intensity were computed and corrected for multiple testing using the Benjamini-Hochberg correction. Significance levels of pairwise comparisons depicted as asterisks: **p < 0.01, ***p < 0.001, ****p < 0.0001. See   RPE cells form lipid droplet-like structures called retinosomes (Imanishi et al., 2004a;Imanishi et al., 2004b;Palczewska et al., 2010;Orban et al., 2011). They serve as storage sites for retinyl esters and can be visualized with lipophilic dyes, similar to other lipid droplets. The high concentration of RE in rlbp1a−/− and the location of the lesions between photoreceptor OS and the RPE layer, led us to investigate the possibility that the observed structures were enlarged retinosomes in the RPE.
To test this, we stained cryosections with BODIPY TR methyl ester, a lipophilic dye that stains neutral lipids and endomembranous structures like mitochondria and photoreceptor OS. Indeed, we found an increased amount of very bright spherical structures positive for BODIPY staining in the subretinal space in rlbp1a−/− and double-KO retinas (Figures 5 and 7). These droplets were rarely present in wild types (Figures 5,7,8G) and rlbp1b−/− fish ( Figure 8G).

Discussion
Our results showed that cones in the zebrafish retina rely on RPE-expressed Cralbpa for chromophore regeneration and retinal function. We found that the knockout of rlbp1a impaired cone-driven ERG responses and led to disturbed OS morphology of both rods and cones and age-dependent thinning of the photoreceptor nuclear layer. HPLC analysis of rlbp1a−/− revealed an accumulation of retinyl esters and a reduction in 11cisRAL that was associated with this visual impairment. To our knowledge, we present the first animal model that recapitulates both the age-dependent retinal thinning and the appearance of subretinal lesions that are commonly found in patients suffering from RLBP1 mutations.
Our data indicate that in early stages, the lack of 11cisRAL (Figure 2A, B and E, top panel) is most likely the primary cause for the reduction in cone-mediated retinal function in the rlbpa1−/− fish. However, at later stages, the disturbed OS morphology and the loss of photoreceptors are likely to contribute to a deterioration of vision in these fish. In addition, it seemed that the integrity of both rods and cones is affected in rlbp1a−/− and double-KO fish. The elevated retinyl ester levels in rlbp1a−/− and double-KO zebrafish (Figure 2A, B and E, middle and lower panel) display as enlarged retinosomes found upon histological examination. ONL thinning appeared, concomitant with the increased number of enlarged retinosomes, only with age. It is interesting to note, that chromophore regeneration was not completely blocked but only attenuated in Cralbpa deficiency. This was true not only for rlbp1a−/−, but also for the double-KO.
These findings are consistent with the prolonged dark adaptation in patients with Bothnia dystrophy , and in mouse models (Saari et al., 2001;Lima de Carvalho et al., 2020). Previously, Stecher et al., 1999 proposed that CRALBP facilitates the hydrolysis of 11cisRE. More recently, it was reported that the 11cisRE pool in the zebrafish RPE is mobilized in a light-dependent and background illumination during recordings. BL: recordings immediately after 1 hr exposure to bleaching light at 20'000 lux. reDA: recordings after bleaching and 1 hr redark adaptation. Dotted lines depict the mean responses of wild-type controls after dark adaptation at 0.01 % light intensity to illustrate the shift in sensitivity under photopic conditions. Data were collected in N ≥ 3 independent experiments and from total n = 15 larvae per genotype and light condition Only intensity levels where wild-type values were above the dark-adapted threshold were considered in the analysis. Pairwise comparisons between WT and mutants at each light intensity were computed and corrected for multiple testing using the Benjamini-Hochberg correction. Significance levels of pairwise comparisons depicted as asterisks: *p < 0.05, ***p < 0.001, ****p < 0.0001. See Figure 4-source data 1 for b-wave amplitudes.
The online version of this article includes the following source data for figure 4: Source data 1. Source data for Figure 4.   (Babino et al., 2015). We now demonstrated that the mobilization of the 11cisRE pool is Cralbpa dependent and plays a critical role for cone visual pigment regeneration. Interestingly, we found that 11cisRE concentrations also declined in the absence of Cralbpa upon light exposure. This result raises the possibility that 11cisRE hydrolysis can occur Cralbpa independently. Thus, Cralbpa may be mainly required to assist the oxidation of 11cisROL to 11cisRAL, as earlier proposed by Saari et al., 1994. Additionally, Cralbpa may act as a chaperone for chromophores in the RPE and prevent the light-dependent isomerization of the released 11-cis-ROL diastereomer to the trans-diastereomer of retinol in the RPE. The latter would be reesterified by LRAT (lecithin retinol acyltransferase), an enzyme that has a high capacity for RE production. This reaction sequence would explain the increase in atRE concentration in illuminated rlbp1a−/− larvae.
As an acceptor of 11-cis-retinoids RPE-expressed Cralbpa is found at a convergence point for potentially all chromophore producing pathways of the RPE. Besides its well-known function in the classical visual cycle downstream of RPE65 and its role in the light-dependent 11cisRE mobilization described in this study CRALBP was also shown to support photic chromophore regeneration by RGR in bovine RPE (Zhang et al., 2019). On the other hand, knockout of MGC-expressed rlbp1b in zebrafish did not affect 11cisRAL content, nor morphology and only marginally affected retinal function in the double-KO in BL conditions. This observation excludes that the MGC-expressed Cralbp carries out the same function in chromophore regeneration as its counterpart in the RPE. Accordingly, it was previously shown that 11cisRE was mostly present in retinosomes of the RPE (Babino et al., 2015). However, we do not close out that Cralbpb may play different roles in the visual cycle. For instance, Kiser et al., 2018 suggested that CRALBP in MGCs could serve as a chromophore storage site by binding 11-cis-retinoids. Also, RGR in MGCs was implicated in light-dependent chromophore regeneration (Morshedian et al., 2019) and Cralbpb may drive this reaction during BL conditions. In this line of thought our measurements of 11cisRAL in the rlbp1b-KO could be masked by RGR-bound retinoids and provide an explanation for the more severe phenotype in double-KO, despite the apparently normal amounts of 11cisRAL in rlbp1b-KO. Further research is needed to explore the role of Cralbpb in the zebrafish retina.
Early work in mice failed to reproduce the progressive loss of photoreceptors as well as the characteristic fundus findings seen in patients (Saari et al., 2001). A more recent study (Lima de Carvalho et al., 2020) reported ONL thinning in Rlbp1−/− mice between 8 and 12 months of age. Such changes may have gone unnoticed in earlier studies on Rlbp1−/− mice due to the younger age of animals at examination (5.5 months). Also in patients, the onset of degenerative processes can vary between individuals. Our zebrafish model recapitulates the early impairment of rods and subsequent cone dysfunction that culminates in photoreceptor degeneration at a relatively early age. Therefore, it resembles more closely an early-onset form of CRALBP-associated diseases, such as RPA or Bothnia dystrophy.
The white-yellow fundus lesions in patients with CRALBP-associated disease are usually found upon imaging the central retina, including the macular region and optic disc (Burstedt and arrows: rod outer segments. Red star: subretinal lipid deposits. Red arrow: missing cone outer segment. (A3, B3) Ventral peripheral retina. Black arrows indicate rod outer segments. These are less abundant, but not completely absent in KO retinas. (G) Quantification of outer nuclear layer (ONL) thickness central retinal sections in 5 dpf (G1), 1-month-old (G2), and adult (G3) retinas. Data are from n ≥ 3 biological replicates per genotype and age. Shown is the mean ± standard deviation (SD) at specified distances from the optic nerve (dorsal and ventral). Effect of genotype on ONL thickness (repeatedmeasures analysis of variance [ANOVA]) at 5 dpf (G1) Golovleva, 2010). The lesions appear to be restricted in size towards the center. Interestingly, the white dots are often reported to be absent in the macula (Demirci et al., 2004;Naz et al., 2011) or at least the fovea (Dessalces et al., 2013). The white dots are not present at an early age (Gränse et al., 2001) and only become prominent in the teens and early adulthood, and slowly disappear with age (Burstedt et al., 2001). The macular sparing suggests that these structures are more likely to be associated with rod photoreceptors. Based on the localization of these structures between the photoreceptor and RPE layer upon OCT imaging and the fact that other visual cycle defects (e.g., mutations in RDH5, RPE65) result in a similar punctate fundus appearance, we suggest that they could represent retinyl ester-filled, enlarged retinosomes, similar to the ones we found in the zebrafish retina. If this and their association with rod photoreceptors, as mentioned above, holds true, it is curious, that despite an accumulation of retinyl esters, Rlbp1−/− mice were not reported to have enlarged retinosomes or similar structures. So, why does the cone-dominant zebrafish produce this phenotype? First, considering cone density, the larval retina nicely mimics Individual samples (open circles) are shown with mean ± standard deviation (SD). Statistics: t-test comparing double-KO vs. wild type. *p < 0.05, **p < 0.01, ***p < .001. See Figure 6-source data 1 for expression data.
The online version of this article includes the following source data for figure 6: Source data 1. Source data for Figure 6. the environment of foveal cones, even if rods are not completely absent. At this age, the accumulation of retinyl esters was not very pronounced yet and the number of lesions was much lower than in older retinas. On the other hand, juvenile and adult retinas of zebrafish are cone rich but contain roughly 40 % of rods that are regularly interspersed throughout the retina (Fadool, 2003) in contrast to the localized distribution in humans. The older zebrafish retina, therefore, resembles more the outer macular region. Complementary pathways for cones are thought to help them compete for available chromophores with the more abundant rod photoreceptors in the human retina. The density of rods peaks just outside of the macular region. This is likely to exacerbate the need for alternative and larger chromophore sources. The dramatic increase of enlarged retinosomes at the age of 1 month seems therefore in line with the hypothesis that the white puncta are associated with the presence of rods. It seems likely, that the different visual cycle pathways in RPE and MGC would be fine-tuned according to the type of cells they are in contact with. In line with this idea, single-cell transcriptomic analysis of the human retina and RPE (Voigt et al., 2019a;Voigt et al., 2019b) revealed differences in peripheral and macular gene expression. Based on these datasets  suggested higher importance for MGC-expressed RLBP1 in the peripheral retina compared to the fovea. The MGC-mediated cycle may become more important in later stages with the increase of rod contribution as it was previously suggested in Fleisch et al., 2008. This could explain the absence of an ERG phenotype in rlbp1b−/− larvae and the only subtle changes in the capacity of double-KO (compared to rlbp1a−/−) to respond to strong bleaching lights and to recover thereafter. It remains to be determined whether the knockout of rlbp1b will have a more pronounced effect on retinal function at later stages when rods are more abundant. But it is worthwhile to mention that Cralbpb influenced the retinyl ester storage in adult rlbp1b−/− and double-KO in the present study.
Differences in density and distribution of photoreceptors and the expression pattern or activity of visual cycle proteins could in part explain the discrepancies between different animal models with mutations in CRALBP. Also, the differences in 11cisRE content between species might add to this. In addition, the light environment to which the photoreceptors are exposed during life (nocturnal vs. diurnal species) may contribute to the phenotype. Together, these species-specific differences may provide an explanation for the absence of subretinal lipid droplets in mouse eyes.
In conclusion, knockout of RPE-expressed rlbp1a in zebrafish recapitulates the functional impairment and degenerative phenotype in RLBP1 disease. Importantly, our model shows an increase in retinyl esters and enlarged retinosomes, providing an explanation for the yellow-white fundus lesions commonly found in RLBP1 patients (see Figure 9 for a graphical overview of our working model). It is therefore an excellent model not only to further dissect the disease mechanism, but also to develop potential cures for RLBP1-associated disease.   . Retinoid metabolism in healthy and Cralbpa-deficient retinas. Illustration of a healthy (leftmost) and Cralbpa-deficient (rightmost) retina with the three neural layers (nuclei in dark blue). Rod and cone photoreceptors are depicted with dark blue or turquoise outer segments, respectively. Both retinal pigment epithelial (RPE) cells (gray-brown) and Müller glia (yellow-brown) contact photoreceptors and are potential sources for chromophore. Between the illustrations, schematics for retinoid metabolism in healthy and Cralbpa-deficient retinas are shown with black arrows indicating the sequence of reactions or transport/diffusion (for simplicity, enzymes or transport proteins other than Cralbp are not shown). In healthy retinas, Cralbpa facilitates the generation of 11cisRAL from both atRE (via the canonical visual cycle) and 11cisRE (through light-dependent hydrolysis, yellow arrow). Therefore, lack of RPE-expressed Cralbpa leads to a reduction of 11cisRAL (gray, dotted arrows) and an accumulation of retinyl ester (bold, black arrow), which manifests as an increased number of enlarged retinosomes in RPE cells (yellow circles in schematic and illustration on the right). Both rod and cone outer segment morphology is affected and with age the outer nuclear layer is thinning and the RPE may become atrophic. The contribution of Müller glial cell-expressed Cralbpb to chromophore regeneration is smaller than expected and the enzymes involved as well as the retinoid species supplied to photoreceptors (11cisRAL/11cisROL?) remain to be further investigated.

CRISPR/Cas9 mutagenesis
Generation of stable homozygous KO lines for rlbp1a and rlbp1b was achieved by using the CRISPR/ Cas9 approach. Oligos for single guide RNA (sgRNA) synthesis were designed using the CHOPCHOP web application (https:// chopchop. cbu. uib. no/): 5′-GGAGCTCAGAGGAATAATCA-3′ and 5′-GGCAG-GCCAAGGAGATGAGG-3′ were chosen for rlbp1a and rlbp1b, respectively. Synthesis of sgRNAs was achieved by polymerase chain reaction (PCR) using a high fidelity Phusion polymerase (New England Bio Labs., Ipswich, MA) and in vitro transcription of the product: Target-specific oligos (target_oligo) were synthesized (Microsynth AG, Balgach, Switzerland) containing a T7 promoter site, the above target-specific sequence, and part of the tracerRNA sequence. A second, partially overlapping, oligo (rev_oligo) was synthesized (Microsynth AG, Balgach, Switzerland) to contain the reverse tracerRNA sequence. After PCR and cleanup, the dsDNA was in vitro transcribed (MEGAshortscript T7 Transcription Kit, Ambion, Thermo Fisher Scientific, Waltham, MA) and the resulting sgRNA was purified (Megaclear Kit, Ambion, Thermo Fisher Scientific, Waltham, MA) and precipitated using ethanol. Injection mixes containing 400 ng/µl sgRNA, 1000 ng/µl Cas9-GFP (MJ922; kind gift of C. Mosimann), and 300 mM KCl were incubated for 10 min at 37 °C to allow for ribonucleoprotein complex formation. One-cell stage embryos were injected with 1 nl of injection mix and raised to adulthood. Injected F0 fish were outcrossed with wild types and the offspring were genotyped using primers designed around the target region (see Table 1) to check for successful germline transmission. F1 animals carrying the same mutation (−19 bp in rlbp1a and −18 bp in rlbp1b, see Table 2) were incrossed to obtain homozygous mutants (F2).

Immunohistochemistry
Whole larvae or adult eyes were fixed in 4 % formaldehyde for 30 min at room temperature (RT). After washing with 1× phosphate-buffered saline (PBS), the tissue was cryo-protected in 30 % sucrose overnight at 4 °C. Samples were embedded in Tissue Freezing Medium clear (EMS-72593, Electron Microscopy Sciences, Hatfield, PA). Samples were sectioned at 12 µm thickness using a Cryostar NX50 (Thermo Fisher Scientific, Waltham, MA) and collected on SuperFrost Plus Adhesion Slides (Thermo Fisher Scientific, Waltham, MA). Sections were spread across multiple slides and wild-type controls were collected next to heterozygous and/or homozygous samples on the same slide to achieve equal conditions during staining and washing steps for all genotypes. Sections were stored at least once overnight at −80 °C. Before staining, the slides were dried for 30 min at 37 °C. Sections were bleached in 10 % H 2 O 2 solution in PBS for 30 min at 65 °C. Slides were washed with 1× PBS and mounted on a Shandon Sequenza Slide Rack (Thermo Fisher Scientific, Waltham, MA). Slides were incubated with blocking solution (10 % normal goat serum, 1 % BSA in 0.3 % PBS Triton X-100) for 2 hr at RT and with primary antibodies overnight at 4 °C. Samples were washed with 1× PBS and secondary antibodies were applied for 1.5 hr at RT in darkness. After washing with 1× PBS, samples were stained with CellTrace BODIPY TR methyl ester (1:200 in PDT [1 % dimethyl sulfoxide, 0.1 % Triton X-100 in PBS], cat# C34556, Thermo Fisher Scientific, Waltham, MA) for 20 min at RT and  Table 2. Wild-type sequences of rlbp1a and rlbp1b in the target region and the corresponding mutated sequence (underlined: target sequence; dashes: deletion).

Real-time PCR
Both eyes of n = 50 larvae per sample or one adult eye per sample were collected and in total three biological replicates were used for each genotype. RNA was isolated using the ReliaPrep RNA Tissue Miniprep System (Promega, Madison, WI). SuperScript III First-Strand Synthesis SuperMix (Thermo Fisher Scientific, Waltham, MA) with a mix of oligo dT primers and random hexamers was used for reverse transcription of the RNAs. Primers used for real-time PCR are summarized in Table 3. Realtime PCR was performed using SsoAdvanced Universal SYBR Green Supermix (catalog #172-5270, Bio-Rad Laboratories, Hercules, CA) on a Bio-Rad CFX96 C1000 Touch Thermal Cycler (Bio-Rad Laboratories, Hercules, CA) using 1 ng template cDNA. Relative normalized expression was calculated by normalizing to the wild-type samples and using actb1, rpl13a, and g6pd as reference genes.

Semithin sections
Whole larvae and adult eyes were fixed at least overnight in 4 % formaldehyde at 4 °C and embedded in Technovit 7,100 (Kulzer, Wehrheim, Germany). Using a microtome (LeicaRM2145, Leica Microsystems, Nussloch, Germany) samples were sectioned at 3 µm thickness and stained in Richardson's solution (1 % methylene blue, 1 % borax) for 5 s, washed for 10 min in ddH 2 O, and coverslipped with Entellan mounting medium (Merck, Darmstadt, Germany). Images were acquired on an Olympus BX61 microscope (Olympus, Shinjuku, Tokyo, Japan) and adjusted for brightness and contrast using Fiji ImageJ (Schindelin et al., 2012). Quantifications were performed on whole retinal sections using Fiji ImageJ. Filenames were randomized to mask the genotype during quantification. ONL thickness was measured dorsally and ventrally every 35 µm (5 dpf), every 100 µm (1 month), and every 200 µm (adults), starting at the optic nerve. Lipid droplets were counted manually with the Cell Counter plugin.

Samples for retinoid analysis and light treatments
Adult zebrafish eyes were collected in ambient room light and immediately frozen in liquid nitrogen and stored at −80 °C before HPLC analysis. For retinoid measurements in zebrafish larvae, light treatments were performed similar to previously described methods (Babino et al., 2015). Samples were grouped according to light condition: overnight DA, DA and exposure to ambient light (ambient), DA and 30 min bleach (30 min BL), DA and 1 hr bleach (1 hr BL), and DA, 1 hr bleach followed by 1 hr redark adaptation (1 hr reDA). Zebrafish larvae (n = 100 per group) were sequentially placed in darkness overnight and immediately collected for HPLC analysis or subjected to indicated light treatments. For light treatments, larvae were placed into open 6 cm Petri dishes wrapped with aluminum foil and exposed to either ambient room light or a strong bleaching light with 20'000 lux using a KL 1500 HAL halogen cold light source (Schott, Mainz, Germany) for the indicated time (30 min or 1 hr). Data are the results of three independent experiments for each age group and KO line.

Electroretinography
Standard white-light electroretinography (ERG) was performed as described (Makhankov et al., 2004), with adaptations. For DA ERG, 5 dpf larvae were placed in darkness for at least 1 hr before measurements. Larvae were sacrificed before the removal of one eye for measurements. For ERG on single mutants, offspring of heterozygous crosses were used and the trunks were used for genotyping after measurements. Five stimulations of 100 ms duration and with increasing light intensity (100 % corresponding to 24'000 µW/cm 2 ) were applied with an interstimulus interval of 15'000 ms. For light-adapted ERGs, an additional background illumination of 20 µW/cm 2 was applied and larvae were not DA before measurements. For bleaching experiments, larvae were exposed to BL from a KL 1500 HAL halogen cold light source (as described in light treatments for retinoid analysis) for 1 hr before measurements. To reduce the number of larvae as well as resources for genotyping, we did not directly use double-heterozygous incrosses for double-KO ERGs. Instead, we incrossed wild-type siblings from such a cross to yield closely related offspring as control and double-KO fish were incrossed to yield homozygous mutant offspring. Measurements were blinded by allocating wild type and double-KO larvae to wells of a 24-well plate according to a random key. For experiments including different light conditions, larvae were derived from the same clutch for all light conditions, but measurements were derived from individual retinas, due to the invasive nature of our protocol.

HPLC analysis of retinoids
For retinoid extraction, a whole eye from wild type and mutant adult zebrafish or 100 zebrafish mutant larvae were transferred to 2 ml Eppendorf microcentrifuge tubes (Eppendorf, Hamburg, Germany). 200 µl of 2 M hydroxylamine (pH 8) and 200 µl of methanol were added and eyes or larvae were homogenized and incubated for 10 min at RT. Then, 400 µl of acetone was added to the homogenate and the mixture was vortexed. Retinoids were then extracted with 500 µl of hexanes by vortexing vigorously. For phase separation, the mixture was centrifuged at 5000 × g for 1 min at RT. The upper organic phase was collected and transferred into a fresh tube. This was repeated and the collected organic phases containing the retinoids were dried using the SpeedVac system (Vacufuge Plus, Eppendorf, Hamburg, Germany). The dried samples were reconstituted in hexanes and used for HPLC separation. Retinoid analyses were performed using normal phase Zorbax Sil (5 mm, 4.6 3 150 mm) column (Agilent Technologies, Santa Clara, CA, USA). The retinoids were eluted by stepwise gradient starting with 0.5 % ethyl acetate in hexane over 15 min, followed by 20 min of 10 % ethyl acetate in hexane with a continuous flow rate of 1.4 ml/min. Finally, the column was equilibrated for 10 min with 0.5 % ethyl acetate in hexane. For molar quantification of retinoids, the HPLC was scaled with authentic standard compounds purchased by Toronto Chemicals (Toronto, Canada).

Statistical analysis
The number of biological replicates is defined in the figure legends as 'n'. It describes the number of individual larvae/fish used in the study. In Figure 2B and D, it refers to samples consisting of a pool of 100 larvae, each. Statistical analysis was performed as indicated in the figure legends using R base (R Development Core Team, 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www. R-project. org/.) and the rstatix package (Kassambara, 2020. rstatix: Pipe-Friendly Framework for Basic Statistical Tests. R package version 0.6.0. https:// CRAN. R-project. org/ package= rstatix). Data were plotted using ggplot2 (H. Wickham. ggplot2: Elegant Wickham, 2016.) and ggpubr (Kassambara, 2020. ggpubr: 'ggplot2' Based Publication Ready Plots. R package version 0.4.0. https:// CRAN. R-project. org/ package= ggpubr) packages. Final figures were assembled using Adobe InDesign and the illustration in Figure 9 was created in Adobe Illustrator.