Germline Disruption of Retinal Pigment Epithelium‐Expressed Zebrafish rlbp1b ^−/− Results in Selective Dim Light Visual Behavior Deficits and Provides a Screening Platform for Evaluating the Pathogenicity of Human RLBP1 Variants

ABSTRACT

Cellular retinaldehyde binding protein (CRALBP) plays a crucial role in the visual cycle by chaperoning 11‐cis‐retinoids. Mutations in its encoding gene RLBP1 lead to inherited retinal diseases with the common feature of poor night vision. Zebrafish possess two RLBP1 paralogs, rlbp1a and rlbp1b, with distinct retinal expression profiles, providing a bespoke opportunity to dissect cell‐specific functions of CRALBP. Here, we first resolved conflicting reports on paralog expression by interrogating zebrafish single‐cell RNA‐sequencing datasets, which revealed predominant rlbp1a expression in Müller glia and rlbp1b expression in the RPE. Using CRISPR‐generated zebrafish knockouts, we demonstrated that loss of RPE‐expressed rlbp1b selectively impaired optokinetic responses (OKR) with a ~50% reduction in saccade frequency relative to wildtype. This impaired OKR response is only seen when under dim light conditions with no defect seen in standard or bright light rearing conditions. This recapitulates the night blindness presentation in patients with RLBP1 mutations. Retinoid profiling of rlbp1b knockout larvae showed significant decreases in 11‐cis‐retinal (62% reduced) and all‐trans‐retinal (69% reduced) levels. To explore mechanistic changes following rlbp1b loss, unbiased proteomic profiling was carried out on rlbp1b knockout adult zebrafish eyes. This confirmed the knockout of Cralbpb and revealed significant disruption of proteins involved in vitamin A metabolism, lipid metabolism/storage and ferroptosis. To explore the utility of zebrafish for in vivo pathogenicity assessment of RLBP1, we established a complementation assay using transgenic zebrafish. Although expression of wildtype EGFP‐tagged human RLBP1 did not rescue the visual deficit, expression of zebrafish Cralbp to the RPE restored dim light vision, whereas zebrafish Cralbp harboring the human pathogenic p.R151Q mutation failed to do so. Together, these findings underscore the predominant role of RPE‐expressed CRALBP in sustaining visual function under low‐light conditions and establish a zebrafish platform for functional evaluation of RLBP1 variants.

RLBP1 is a gene that is mutated in several different inherited retinal diseases (IRDs). Here we investigate the consequences of loss of RPE expressed zebrafish Cralbpb using CRISPR knockout models. These consequences include impaired dim light OKR behaviour, altered retinoid profiles, and a disrupted proteomic profile. This work also leverages genetic manipulation technologies in zebrafish for their use in predicting the pathogenicity of novel RLBP1 variants.

Abbreviations

11cRal

11‐cis‐retinal

11cRol

11‐cis‐retinol

AAV

adeno‐associated virus

atRAL

all‐trans‐retinal

CRALBP

cellular retinaldehyde‐binding protein

Cralbpa

cellular retinaldehyde‐binding protein a (zebrafish)

Cralbpb

cellular retinaldehyde‐binding protein b (zebrafish)

CRISPR

clustered regularly interspaced short palindromic repeats

ERG

electroretinogram

IRD

inherited retinal disease

kDa

kilodalton

KO

knockout

OKR

optokinetic response

RDH5

retinol dehydrogenase 5

RLBP1

retinaldehyde‐binding protein 1 gene (Human)

Rlbp1

retinaldehyde‐binding protein 1 gene (Mouse)

rlbp1a

retinaldehyde‐binding protein 1 a gene (zebrafish)

rlbp1b

retinaldehyde‐binding protein 1 b gene (zebrafish)

RPE

retinal pigment epithelium

RPE65

retinal pigment epithelium‐specific protein 65 kDa

scRNA seq

single‐cell RNA sequencing

VMR

visual motor response

VUS

variant of uncertain significance

WT

wildtype

1. Introduction

Cellular retinaldehyde binding protein (CRALBP) is a 36 kilodalton (kDa) protein encoded by the retinaldehyde‐binding protein 1 (RLBP1) gene [1]. CRALBP supports production of 11‐cis‐retinaldehyde (11cRal) during the visual cycle, through its binding of 11cRal and 11‐cis‐retinol (11cRol) [2, 3]. This process is responsible for recycling the visual chromophore enabling phototransduction in photoreceptors and vision [2, 3]. 11cRal is the light‐sensitive aldehyde form of vitamin A that binds to opsins in the eye to enable vision, while 11‐cis‐retinol is its alcohol precursor [4].

Mutations in RLBP1 were first linked to autosomal recessive retinitis pigmentosa (RP) [5]. Since then RLBP1 variants are causally linked to several other inherited retinal diseases (IRDs) including Retinitis Punctata Albescens [6], Fundus albipunctatus [7], Bothnia dystrophy ([8]), and Newfoundland rod‐cone dystrophy [9]. These RLBP1 linked forms of impaired vision share common characteristics including night blindness, delayed dark adaptation, poor night vision, and white lesions on the retina [10]. CRALBP itself does not have catalytic activity but acts to enhance the kinetics of visual cycle reactions through its binding to 11cRal and 11cRol [11]. CRALBP is present in both the retinal pigment epithelium (RPE) and Müller glia and as such is regarded to support both the classical visual cycle and the intraretinal visual cycle [12]. Mouse Rlbp1 knockout models show defects in retinoid levels consistent with the function of Rlbp1 in supporting the visual cycle [3, 13, 14]. Using in vitro bovine RPE membranes, the presence of CRALBP was reported to enhance the rate of production of 11cRol by the RPE isomerohydrolase [15]. This is thought to occur due to sequestration of 11cRol bound to CRALBP, which reduces its feedback inhibition on the retinal isomerohydrolase [15]. The retinal isomerohydrolase referred to here was later identified as RPE65 which utilizes all‐trans‐retinyl esters as a substrate for hydrolysis of the ester and isomerisation of the retinoid [16, 17, 18]. Using the same in vitro RPE system, CRALBP promotes oxidation of 11cRol to form 11cRal rather than esterification and storage [19], partially due to CRALBP acting as a substrate carrier for retinol dehydrogenase 5 (RDH5) [20]. The role of CRALBP on the kinetics of the intraretinal visual cycle has also been examined. The presence of CRALBP enhances threefold the photoisomerization of atRAL to 11cRAL, by the retinal G protein‐coupled receptor (RGR) in cultured cells [21]. In vivo data supports these functions as CRALBP knockout models show less 11cRal and an accumulation of retinyl esters [3, 14]. CRALBP may also support the visual cycle by protection of 11‐cis retinoids from bleaching and facilitating intracellular transport. 11cRAL bound to bovine CRALBP was resistant to bleaching even under intense light conditions [22]. CRALBP has been shown to interact with ERM‐binding phosphoprotein50/sodium‐hydrogen exchanger regulatory factor1 (EBP50/NHERF1) which is localized to RPE apical processes [23] and enhances 11cRol release from RPE cells [24].

Here, we report visual behavior phenotypes in CRISPR knockouts of zebrafish rlbp1a and rlbp1b. Zebrafish are a versatile model of human vision primarily due to their cone‐dominant retina, photopic vision, and their amenability to genetic manipulation [25]. Zebrafish are particularly suited to studying CRALBP function as two paralogs of RLBP1 are present in the zebrafish genome, with each paralog showing specific expression in the RPE or Müller glia [14, 26, 27]. This allows investigations into the cell‐specific roles of CRALBP in the classical and intraretinal visual cycle. However, some previous studies suggest that rlbp1a and rlbp1b show enriched expression in the RPE and Müller glia, respectively; whereas others report the opposite pattern [14, 26, 27]. Here, we address this problem using scRNAseq databases to resolve the expression of rlbp1a in Müller glia and rlbp1b in RPE, respectively.

The first report of a global Cralbp knockout mouse showed defects in retinoid levels and delayed dark adaptation by electroretinography (ERG) but could not differentiate the specific roles of RPE versus Müller‐expressed Cralbp [3]. Morpholino‐based knockdown of Müller glia‐ or RPE‐ expressed paralogs in zebrafish suggested that both are important to support vision [26, 27]. Initial cell‐specific investigations of Cralbp in mice suggested a key role for Cralbp in Müller cells. AAV‐mediated Cralbp expression in Müller cells rescues transretinal ERG responses in Rlbp1 knockout mice, but no rescue was seen when Cralbp expression was returned to RPE cells only [28]. However, this finding has limitations as transretinal ERGs were performed without the RPE in situ [28]. A subsequent study sought to elucidate Cralbp function in Müller glia, by examining ERGs in Rlbp1 heterozygous versus wildtype mice, wherein the RPE/RPE65‐based visual cycle was reportedly blocked pharmacologically [29]. In this model, ERG‐based dark adaptation of M‐cones is impaired in Rlbp1 heterozygotes, indirectly supporting a requirement of Cralbp function in Müller glia for features of cone electrophysiology [29]. However, these interpretations have not been validated by an independent methodology. This importance of CRALBP in Müller cells is challenged by more recent papers that show no or limited effects of CRALBP knockout specifically in Müller cells [14, 30]. For example in zebrafish, RPE‐ but not Müller glia‐ expressed Cralbp knockouts show impaired ERG responses and impaired retinoid metabolism [14]. Recently, defects in M‐cone response, M‐cone dark adaptation, and retinoid metabolism were reported specifically in RPE‐Cralbp knockout mice but not in Müller glia‐Cralbp knockout mice supporting a predominant role for RPE‐Cralbp in supplying visual chromophore to cones [30]. However, to date, no studies have used mammalian germline knockouts to investigate the requirement of Müller glia‐ or RPE‐expressed CRALBP for visual behavior.

Decades of research on RLBP1 recently culminated in a successful phase 1/2 safety and efficacy report for RLBP1 gene therapy [31]. This provides hope for a future treatment for RLBP1‐linked diseases. However, gene therapy approaches rely on accurate identification of pathogenic mutations. In databases of RLBP1 clinical variants, 57% are classed as variants of uncertain significance (VUS) with 9% as pathogenic, 12% as conflicting, and 22% as benign (CONVART.org). Classifications are based on frequency of variant occurrence, predictive modeling, and experimental analysis of function. There is currently no systematic in vivo functional assay to classify the pathogenicity of CRALBP variants. Here, we address this by exploring the utility of humanized zebrafish models for RLBP1 pathogenicity prediction.

2. Methods

2.1. Zebrafish Husbandry and Ethics Statement

Zebrafish larvae up to 131 h post fertilization (hpf) were raised at 27°C under a 14 h/10 h light–dark cycle in 10 cm Petri dishes of E2 Media (0.137 M NaCl, 5.4 mM KCl, 5.5 mM Na₂HPO₄, 0.44 mM KH₂PO₄, 1.3 mM CaCl₂, 1.0 mM MgSO₄, and 4.2 mM NaHCO₃, conductivity ~1400–1600 μS, pH ~ 7). Adult zebrafish were housed in a recirculating Aquatics Habitats system and fed Gemma micro (Skretting) twice daily (once daily on weekends) and brine shrimp (Artemia sp.) once daily. All experiments with zebrafish were performed according to ethical exemptions and approvals granted by the UCD Animal Research Ethics Committee (AREC‐20‐16‐KENNEDY) and authorizations from the Health Products Regulatory Authority (Project authorization AE18982/P186).

2.2. Zebrafish rlbp1a and rlbp1b Knockout Line Generation and Genotyping

To generate the rlbp1a ^−/− zebrafish, two CRISPR guides (5′‐AGGAGATGAGGTGGCCAAAA‐3′, 5′‐GAGCACACGAGCTCATGAAAGGT‐3′) were designed to produce a 101 base pair (bp) deletion in exon 3. These CRISPR guides were injected into in‐house stocks of WT (Tü) zebrafish, and a germline transmitting founder fish was identified. Primers used for genotyping rlbp1a are as follows: rlbp1a F—5′‐TAAAGGCTAAAGATGAGCTGAA‐3′ and rlbp1a R—5′‐AAGCAAAGAATTTCCCCCGT‐3′. Sanger sequencing of genomic DNA amplicons from the rlbp1a ^−/− zebrafish line confirmed the presence of the correct size deletion. The zebrafish rlbp1b ^−/− line used was previously described [32]. Genotyping primers for rlbp1b are as follows: rlbp1b F—5′‐CCACAGAGTGGAACATTTCGA‐3′, rlbp1b R—5′‐ACCTTTCAGCATTCACAATCCT‐3′, and rlbp1b poison—5′‐GATGATGCAACACTATTGCCCA‐3′. Zebrafish rlbp1a ^−/−;rlbp1b ^−/− double knockout lines were generated by incrossing the rlbp1a ^−/− and rlbp1b ^−/− CRISPR knockout lines.

Genomic DNA extraction from whole larvae or adult caudal fin biopsies was carried out as follows. Tissue was added to 50 μL of lysis buffer (10 mM Tris–HCl, 50 mM KCl, 0.3% NP40, 0.3% Tween, pH 8.3, 0.02% 20 mg/mL) proteinase K (Invitrogen 4333793) in 200 μL PCR tubes. The tissue was then incubated with periodic vortexing for 30 min to overnight. The proteinase K in the mix was then heat‐inactivated at 90°C for 10 min. Tubes were spun at 10 000 RPM for 10 min to pellet insoluble material. 5 μL of the resulting supernatant was used as a PCR template.

PCR reactions were carried out using the OneTaq quick‐load 2× Master Mix with standard buffer (New England Biolabs M0486). Reactions were set up to contain 12.5 μL OneTaq enzyme master mix, 1 μL forward primer, 1 μL reverse primer, 1 μL poison primer (only for rlbp1b), and nuclease free water to 25 μL. PCR cycling conditions used for rlbp1a were 95°C 3 min, 35 cycles of 95°C 30 s, 58.9°C 25 s, 72°C 25 s, and a final elongation of 72°C for 5 min. The WT amplicon measures 346 bp, and the deletion amplicon measures 245 bp. PCR conditions for rlbp1b were 95°C 3 min, 34 cycles of 95°C 30 s, 61.1°C 30 s, and 72°C 1 min, and a final elongation of 72°C for 5 min. PCR products were run on 2% agarose gels stained with SYBR safe DNA gel stain (Invitrogen S33102). Brightfield images of larval zebrafish were taken using an Olympus SZX10 microscope and a color camera (HDKE09422053) with live 120–125 hpf larvae mounted in 9% methylcellulose.

2.3. Alignments

Homologous sequences to the human CRALBP protein were identified using a BLAST search with human CRALBP protein as the template sequence. Identified sequences of CRALBP homologs in model organisms were then input into the TCoffee alignment tool (tcoffee.crg.eu) using the espresso setting. The alignment file was opened using JalView (jalview.org) and colored according to percentage identity. Domains were identified in human CRALBP using the NCBI conserved domain search and mapped manually to the alignment. Pathogenic mutations identified in RLBP1 were taken from [6] and manually mapped to the alignment.

2.4. Single Cell RNAseq Analysis

Single cell RNA sequencing data was obtained from the [33] dataset using the viz.stjude.cloud web tool (viz.stjude.cloud). This web tool was used to generate heatmaps for gene expression by cell type for rlbp1a and rlbp1b. The Daniocell (daniocell.nichd.nih.gov) web tool was also used to generate expression patterns for rlbp1a and rlbp1b.

2.5. Immunofluorescence Imaging

Zebrafish larvae < 131 hpf were fixed in 4% paraformaldehyde (PFA) (Sigma) in phosphate‐buffered saline (PBS) overnight. Following this, the fixed larvae were incubated in 25% sucrose (Sigma) in PBS for 24 h or until submerged; this was followed by an incubation in 35% sucrose in PBS for 24 h or until they submerged. Following this, the larvae were mounted in cryoblocks containing OCT embedding matrix and frozen. 25 μm cryosections were taken with a Leica CM1860 UV cryostat and mounted on Epredia Superfrost Plus Adhesion microscope slides. Sections were post‐fixed for 15 min with 4% PFA in PBS. Sections were bleached for 30 min to remove RPE pigment in a Coplin jar containing 3% hydrogen peroxide (Sigma) and 180 mM KOH (Sigma) in deionized water. Sections underwent antigen retrieval in 0.1 M citrate buffer pH 6 in a pressure cooker for 10 min. Following this, sections were blocked with normal goat serum (Sigma) in PBSDT (PBS plus 0.1% Tween 1% DMSO) for 1 h. After blocking, slides were incubated overnight with a 1:1000 dilution of a CRALBP antibody (Proteintech 15356‐1‐AP). Following washes, sections were incubated with a 1:500 dilution of anti‐rabbit IgG Alexa Fluor 647 conjugate (Cell signaling technologies 4414) and a 1:100 dilution of Hoechst. Following washes, sections were covered with Aqua‐Poly/Mount (Polysciences) and a coverslip was added. Sections were imaged with a Zeiss LSM 800 Airy confocal microscope using a Plan‐Apochromat 63×/1.4 Oil DIC M27 objective. Images shown contain the optic nerve and are maximal frontal orthogonal projections.

2.6. Visual Behavior Assays: OptoKinetic Response

Optokinetic response (OKR) assays were performed based on a previous publication [34]. All OKR assays were carried out on < 131 hpf zebrafish larvae. Larvae were raised under two different conditions. The first being raised in an incubator at 27°C with a 14:10 h light:dark cycle and a light intensity of 450 lx. The second raising paradigm differs from the first, where at 99 hpf the larvae are transferred to a 60 mm dish and placed in a zebrafish 1 L mating tank lined with aluminium foil inside a wooden box also lined with aluminium foil. They are then exposed to 81 000 lx light from a Schott KL 2500 LED light source at 100% power for 24 h. After the 24‐h incubation, the plate is placed on a table in the fish facility under ambient light conditions where the larvae are taken one at a time to the OKR assay. The OKR assay was performed using a stereo microscope and a custom rotating base that holds a 3D printed drum consisting of black and white stripes spaced at 0.02 cycle per degree (CPD). The < 131 hpf zebrafish larvae are placed in the centre of a 60 mm Petri dish containing a 9% methylcellulose (Sigma viscosity: 25 cP) solution in embryo media and orientated so that its dorsal side is pointing upward. The drum is then rotated for 30 s clockwise, then 30 s anticlockwise, and the number of saccadic eye movements are counted and summed together. All OKR assays were carried out between 12 and 3 pm to minimize variation due to the time of day. Standard OKR assays were carried out using the Schott KL 2500 LED light source set at 22.7% power with the room lights turned on for a light level of 36 000 lx. Dim light OKR assays were carried out with the same light source but set to 0.2% power and the room lights turned off for a light level of 5.6 lx. For rlbp1a and rlbp1b OKRs, parent fish heterozygous for each gene were incrossed to yield a clutch of eggs with a Mendelian ratio of all genotypes. Following the OKR assay, larvae were saved for genotyping. For the rlbp1a ^−/−;rlbp1b ^−/− double knockout, parent fish that were rlbp1a ^+/−;rlbp1b ^−/− were incrossed to give a clutch with a Mendelian ratio of all rlbp1a genotypes in the rlbp1b ^−/− background. Light measurements are the average of triplicate values taken with a Sauter SO 200 K Lux meter. All graphs and statistical analyses were completed using Graphpad Prism version 8.

2.7. Visual Behavior Assays: Visual Motor Response

The visual motor response (VMR) assay was carried out using the rlbp1b line. As for the OKRs, heterozygous parent fish were incrossed to give a clutch of all rlbp1b genotypes in Mendelian ratios. These larvae were then used for the VMR and genotyped afterwards. VMR assays were carried out using the Zebrabox (Viewpoint Technologies) and associated software using the quantization live recording format [34]. Capture settings were integration period 1 s, sensitivity 25, burst threshold 25, and freeze threshold 3. One larva was placed per well in square 96 well plates and loaded into the machine. For standard VMR assays, the larvae had a 30 min settling period exposed to background light levels from the machine (30% power—Range 1). The larvae then underwent intervals of lights‐on/lights‐off/lights‐on/lights‐off/lights‐on with each interval having a duration of 20 min and a light level of 30% power—Range 1. For dark adapted VMR assays, the experiment was carried out in the same way but instead of a settling period of 30 min in background illumination the larvae were dark adapted in the VMR machine for 3 h. Data were analyzed using Microsoft Excel and GraphPad prism. Average activity (ms/s) was calculated as the sum of the ‘middur’ and ‘burdur’ values from the output Excel file. This represents the time in ms that a larva is above the lower movement threshold. For ON and OFF peak graphs, activity for both lights on or lights off transitions were averaged per larvae before being plotted on the graph. Average activity during the lights on period was calculated as the average per larvae of the activity during the second and third lights on period, ignoring the first. Average activity during the lights off period was calculated as the average per larvae of the activity during both lights off periods. Peak activity following light transitions was calculated by finding the peak activity in the 5 s following both light transitions and averaging them per larvae.

2.8. Humanized Transgenics Lines

A plasmid containing the CDS of human RLBP1 was obtained from the DNASU plasmid repository (Plasmid ID:HsCD00313676, DNASU.org). This sequence was sub‐cloned into the p‐3′ E entry vector using the Gateway BP clonase II enzyme mix (Invitrogen). The following primers were used to amplify RLBP1 from the template plasmid: attB2_hsRLBP1_F—5′‐GGGGACAGCTTTCTTGTACAAAGTGGCCATGTCAGAAGGGGTGGGCACGTTCC‐3′, attB3r_hsRLBP1_R—5′‐GGGGACAACTTTGTATAATAAAGTTGATCAGAAGGCTGTGTTCTCAGCTTGG‐3′. The final expression vector was created using the Gateway LR clonase II enzyme mix (Invitrogen). In this LR reaction, the following plasmids were used: p5′E‐rpe65a ([35]), pME‐GFP (no stop), p3′E‐RLBP1, and pDEST‐Tol2. This reaction yields an expression vector with N‐terminal GFP tagged human RLBP1 under the control of the zebrafish rpe65a promoter. The CDS of RLBP1 and flanking regions were sequenced to ensure the expression vector assembled as intended. Mutant RLBP1 expression vectors were created using the Q5 Site‐Directed Mutagenesis Kit (NEB E0554S). Site‐directed mutagenesis of the expression vector was unsuccessful, so it was instead carried out on the p3′E‐RLBP1 entry vector (using the same kit) and the LR reaction was repeated as before to yield mutant expression vectors. The following primers were used for SDM; F.R151Q‐5′‐CCTCTCTAGTCAGGACAAGTATGGCC‐3′, R.R151Q—5′‐ACACCAGGGTAGCCAGCT‐3′, F.A237V—5′‐CCGGTTCAAAGTCATCCACTTCA‐3′, and R.A237V—5′‐GCTGGGAAGGAATCCTGG‐3′. Expression vectors were sequenced to ensure the intended mutation was present. For Tol2 transgenesis, tol2 mRNA was created by in vitro transcription from the pCS2‐TP plasmid using the mMessage mMachine SP6 transcription kit (Invitrogen AM1340). mRNA was then purified using the RNeasy mini kit (Qiagen 74104). Expression vector plasmids were prepared for microinjection using the plasmid midi kit (Qiagen 12143). rlbp1b ^−/− zebrafish eggs were injected in the cell body at the one‐cell stage with ~2 nL of the Tol2 mRNA (12.5 ng/μL) and the desired expression vector (50 ng/μL). Larvae were selected on day 1 for a GFP positive pineal gland, and F0 fish were raised to sexual maturity. Founder fish were identified and outcrossed to WT stocks to create stable lines. PCR analysis and sequencing of these stable lines using the following primers confirmed that the correct transgene was present GFP_PCRseqTol2F—5′‐AGCTCAAGCTTCGAATTCTGC‐3′, GFP_PCRseqTol2R—5′‐GATGGTACCGTAAAACGACGG‐3′. Sequencing results were aligned to the template expression vector in Benchling (Benchling.com) and transgenic lines were deemed valid if the sequencing region covered the entirety of the rlbp1 CDS without errors (Figure S1).

2.9. Zebrafish Transgenic Lines

The zebrafish rlbp1b CDS was amplified from an eye cDNA library created from in‐house wildtype zebrafish using the following primers zf.rlbp1b.AttB2r—5′‐GGGGACAGCTTTCTTGTACAAAGTGGACATGGCGGTTGTTAGTGGAACATTTC‐3′ zf.rlbp1b.AttB3–5′‐GGGGACAACTTTGTATAATAAAGTTGTCTAGTCAAACAGATGTGCTGCTGTG‐3′. An expression vector with N‐terminal GFP tagged zebrafish rlbp1b under the control of the zebrafish rpe65a promoter was then created as described for the human gene. The rlbp1b CDS and surrounding regions were sequenced to ensure the expression vector was constructed as intended. Compared to the reference sequence, 6 mutations were present in our plasmid, but all were confirmed to be silent mutations with no impact on the protein sequence. This expression vector was mutated using site‐directed mutagenesis as described for the human gene using the following primers zfR151Q_F—5′‐CCTATCCAGCCAGGACAAATATGGC‐3′ and zfR151Q_R—5′‐ATTCCTGGGTATCCGGCCTC‐3′. Transgenic lines were created and validated as described for the human gene.

2.10. Retinoid Profiling

For retinoid analysis, larvae were euthanized in tricaine (MS‐222) prior to collection under dim red light (50 < 131 hpf larvae per replicate). Samples were snap frozen in liquid nitrogen and stored at −80°C. For normal phase high‐performance liquid chromatography (HPLC) analysis of retinoids, all retinoid extractions were carried out under dim red light in a dark room.

Individual replicates (50 < 131 hpf larvae) were stored at −80°C, thawed on ice, and homogenized in 500 μL of PBS using a glass tissue grinder (Kontes). Protein concentration was determined using 50 μL of sample and a Micro BCA protein assay kit (Pierce). To the remaining 450 μL of homogenate, 25 μL of 5% SDS (~0.2% SDS final concentration) and 50 μL of brine were added, and samples were briefly mixed. Hydroxylamine hydrochloride (500 μL of 1.0 m solution in pH 7.0 PBS) was added to generate retinal oximes, and samples were vortexed and incubated at room temperature for 15 min. The aqueous phase was quenched and diluted using 2 mL of cold methanol. The samples were twice extracted by the addition of 2 mL aliquots of hexane, brief vortexing, and centrifugation at 3000 g for 5 min to separate the phases. Pooled hexane extracts were added to 13 × 100 mm borosilicate test tubes and evaporated to dryness under a stream of nitrogen. Dried samples were dissolved in 100 μL of hexane and analyzed by normal‐phase HPLC using an Agilent 1100 series chromatograph equipped with a Supelcosil LC‐Si column (4.6 × 250 mm, 5 μm) using a 0.2%–10% dioxane gradient in hexane at a flow rate of 2 mL per min. The eluted mobile phase was analyzed using a photodiode‐array detector. Spectra (210–450 nm) were acquired for all eluted peaks. The identity of eluted peaks was established by comparison with spectra and elution times of known authentic retinoid standards. Retinoid amounts were quantitated by comparing their respective peak areas to calibration curves established with retinoid standards.

2.11. Proteomic Profiling

2.11.1. Sample Preparation

Four biological replicates per genotype were collected. Eyes were dissected and stored at −80°C. Tissue pellets were resuspended in 50 μL of ice‐cold 8 M urea/50 mM Tris–HCL with phosphatase and protease inhibitors (Roche). Samples were sonicated three times for 10 s at 10% power using a Branson Sonifier. Protein concentrations were assessed using a Nanodrop (Denovix), and 200 μg of protein was carried forward for digestion. Protein samples were reduced with 8 mM dithiothreitol (DTT) and incubated at 1200 rpm and 30°C for 30 min. Alkylation was performed by adding 20 mM iodoacetamide and incubating under the same conditions for 30 min in the dark. Samples were trypsin digested with a Thermo Fisher Scientific KingFisher Flex using a high‐throughput semi‐automated label‐free trypsin digestion protocol developed at the Novo Nordisk Foundation Center for Protein Research [36].

2.11.2. Mass Spectrometry

1 μg of pooled sample was analyzed using a Bruker timsTof Pro mass spectrometer connected to an Evosep One liquid chromatography system. Tryptic peptides were resuspended in 0.1% formic acid and loaded onto an Evosep tip. The Evosep was configured to pick up each tip, elute, and separate the peptides using a set chromatography method (30 samples a day) [37]. The mass spectrometer was operated in positive ion mode with a capillary voltage of 1300–1600 V, dry gas flow of 3 L/min, and a dry temperature of 180°C. All data was acquired with the instrument operating in a data dependent analysis parallel accumulation serial fragmentation mode (dda‐PASEF). Trapped ions were selected for ms/ms using parallel accumulation serial fragmentation (PASEF). A scan range of (100–1700 m/z) was performed at a rate of 4 PASEF MS/MS frames to 1 MS scan with a cycle time of 0.53 s [38]. The resultant file was used to create the dia‐PASEF method within Bruker timsControl software. The scan mode “dia‐PASEF” was selected, and the pooled sample dda‐PASEF file was opened in the window editor in the MS/MS tab. Once opened, the adjustable rhomboid was used to select the area of the heat map where the identifiable peptides (central region in the heat map containing peptides with charge states from +2 to +5) can be found. In summary, the dia‐PASEF settings used were mass width 26.0 Da, mass overlap 1.0, mass steps per cycle 35, mobility overlap 0.00, mass range 325.2–1201.2 m/z. All samples were acquired using data independent analysis parallel accumulation serial fragmentation (dia‐PASEF) [39].

2.12. Data Analysis

Data acquired using dia‐PASEF were analyzed using DIA‐NN 2.1.0 Academia (Data‐Independent Acquisition by Neural Networks) [40]. The Danio rerio subset from The Uniprot Swissprot/Trembl database was used to generate a spectral library within DIA‐NN (library free mode). Specific search settings included cysteine carbamidomethylation enabled as a fixed modification, maximum missed cleavages 1, min precursor +2, max precursor +4, neural network (cross validated) was used, cross‐run normalization was set to retention time dependent, and library generation was set to ID's, RT & IM profiling. Precursor FDR was set to 1%.

3. Results

3.1. Zebrafish rlbp1b Is Predominantly Expressed in the RPE and rlbp1a in Müller Glia

CRALBP functions in the visual cycle as an 11‐cis retinoid carrier protein in the RPE and Müller glia (Figure 1A). Our ambition was to understand the requirement of RPE‐expressed CRALBP in visual behavior. The single human RLBP1 gene is duplicated as rlbp1a and rlbp1b paralogues in zebrafish (Figure 2A). There are contradictory reports regarding the retinal cell type expression patterns of zebrafish rlbp1a and rlbp1b [14, 26, 27, 32, 41, 42]. To resolve this matter, recent publicly available zebrafish scRNAseq databases were interrogated. In a scRNAseq dataset of adult zebrafish eyes [33], rlbp1b shows predominant expression primarily in the RPE cell population whereas rlbp1a shows predominant expression primarily in the Müller glia cell population (Figure 1B–D). In agreement, from interrogation of the Daniocell scRNAseq database of larval zebrafish [43], the Müller glia cell population display robust rlbp1a expression, and the RPE cell population display robust rlbp1b expression (Figure 1E–G). Cell‐type expression is not fully mutually exclusive, as low levels of rlbp1a and rlbp1b are reported in the RPE and Müller glia populations, respectively (Figure 1H,I). From the Daniocell database, the onset of rlbp1a and rlbp1b expression is equivalent, starting at 48 hpf and persisting at similar levels until 120 hpf, the last time point analyzed.

FIGURE 1.

Zebrafish rlbp1a is expressed in the Müller glia and rlbp1b is expressed in the RPE. (A) Schematic diagram of the role of CRALBP in the visual cycle. Created using biorender.com. (B) tSNE clustering of scRNAseq data from [33]. (C) Expression of rlbp1a from the [33] dataset showing expression in Müller glia (red dots). (D) Expression of rlbp1b from the [33] dataset showing expression in the RPE (red dots). (E) UMAP projection of gene expression from the Daniocell database (daniocell.nichd.nih.gov). (F) UMAP projection of rlbp1a expression from Daniocell showing expression in Müller glia (blue to red colouration). (G) UMAP projection of rlbp1b expression showing expression in the RPE (blue to red colouration). (H) Time course of rlbp1a expression in the RPE and Müller glia during development from Daniocell. (I) Time course of rlbp1b expression in the RPE and Müller glia during development from Daniocell.

FIGURE 2.

Generation and confirmation of zebrafish Cralbp CRISPR knockout lines. (A) Schematic of CRISPR approach used to generate rlbp1a and rlbp1b knockout lines. (B) Representative genotyping gels for PCR genotyping of rlbp1a and rlbp1b knockout lines where amplicons of 346 bp and 245 bp (rlbp1a) or 469 bp and 248 bp (rlbp1b) represent wildtype and deletion alleles, respectively. (C) Brightfield images of rlbp1a, or rlbp1b, and double knockout lines. (D) Cralbp immunostaining in 5 dpf larval zebrafish eyes showing expression in Müller glia and RPE. (E) Cralbp immunostaining in the rlbp1a knockout line showing loss of expression in Müller glia and robust expression in the RPE. (F) Cralbp immunostaining shows the opposite pattern in the rlbp1b knockout line with loss of expression in the RPE but persistent expression in the Müller glia. Magenta staining represents staining with the RLBP1 antibody (15356‐1‐AP), blue represents staining with Hoechst for nuclei.

3.2. CRISPR‐Cas9 Generation of rlbp1a ^−/−, rlbp1b ^−/−, Single and Double Knockout Zebrafish Lines

Establishment of a germline knockout of the RPE‐enriched rlbp1b has previously been reported [32] (Figure 2A). Here, an additional knockout of the Müller glia enriched zebrafish rlbp1a was generated using targeted CRISPR mutagenesis (Figure 2A). Two guides were designed to cut rlbp1a at exon 3, yielding a short deletion of 101 bp confirmed by PCR genotyping (Figure 2B). Disruption of rlbp1b transcription was confirmed by qPCR [32]. Double knockout rlbp1a ^−/−;rlbp1b ^−/− zebrafish lines were created by incrossing each single knockout line. Neither the single nor the double knockout lines show any gross morphological abnormality at ~125 hpf (Figure 2C). Immunostaining of retinal cryosections from < 131 hpf rlbp1a and rlbp1b knockout larvae confirms the expression patterns reported in the scRNAseq databases, with the Müller glia expressed rlbp1a knockout showing expression only in the RPE, and the rlbp1b knockout showing absent expression in the RPE, but staining in Müller glia and staining of RPE and Müller glia in wildtype controls (Figure 2D–F). These images confirm loss of Cralbpa or Cralbpb protein expression in the respective knockout lines and confirm that rlbp1a is expressed in the Müller glia while rlbp1b is expressed in the RPE.

3.3. Elimination of RPE Expressed rlbp1b/Cralbpb Results in Selective Deficits in Optokinetic Visual Behavior Under Dim Light Conditions

To assess the requirement of RPE‐expressed CRALBP for visual behavior, rlbp1b ^−/− zebrafish larvae were first analyzed under standard OKR conditions (Figure 3A). This consists of raising zebrafish larvae at 27°C at 452 lx under a 14:10 Light:Dark cycle, before < 131 hpf larvae are immobilized in 9% methylcellulose and presented with a rotating drum with a black and white striped pattern consisting of 0.02 CPD, as previously described [44]. Under these standard OKR conditions, the striped drum is illuminated to a level of 36 000 lx. The response of rlbp1b ^−/− and wildtype siblings was equivalent under these conditions, with an average of 21 and 19 saccades/min, respectively (Figure 3B). This was initially surprising based on attenuated electroretinograms previously reported in knockouts of RPE‐expressed Cralbp [14].

FIGURE 3.

RPE expressed rlbp1b knockout larvae have an OKR visual behavior defect specifically in dim light. (A) Schematic showing raising conditions, followed by the type of OKR assay. (B) Standard light OKR assay showing no difference between rlbp1b ^−/− larvae with wildtype siblings. Four biological replicates n ≥ 17 larvae per group. (C) Dim light OKR assay comparing rlbp1b ^−/− larvae with WT siblings. Four biological replicates n ≥ 16 larvae per group. p < 0.0001 unpaired two tailed t‐test. (D) No difference is seen in dim light OKR assays between rlbp1a ^−/− larvae with WT siblings. Three biological replicates n ≥ 10 larvae per group. (E) Dim light OKR assay showing a significant difference between rlbp1a ^−/−;rlbp1b ^−/− larvae and rlbp1b ^−/− larvae. Four biological replicates n ≥ 17 larvae per group. p = 0.0018 unpaired two tailed t‐test. (F) Schematic showing raising conditions including 24 h bright light treatment starting on 99 hpf, followed by type of OKR assay. (G) 24 h bright light treatment followed by standard light OKR assay showing no difference between rlbp1a ^−/−;rlbp1b ^−/− larvae and rlbp1b ^−/− larvae. Three biological replicates n ≥ 11 larvae per group. (H) 24 h bright light treatment followed by dim light OKR assay showing a significant difference between WT siblings with rlbp1b ^−/− larvae. Three biological replicates n ≥ 10 larvae per group. p < 0.0001 unpaired two tailed t‐test. All OKRs were carried out with a 0.02 CPD drum on < 131 hpf larvae. All graphs show mean ± standard deviation.

A review of the clinical presentations of inherited RLBP1‐related disease reveals a common poor vision phenotype in dim light levels or night blindness [6, 45]. Thus, a bespoke ‘dim light OKR’ assay was established using identical rearing light levels up to < 131 hpf but with lower levels of light illumination, recorded as 5.6 lx, during the OKR assay. Under these dim light OKR conditions rlbp1b ^−/− larvae show a significantly (p < 0.0001, two tailed t‐test) decreased (~50%) number of saccadic responses (9 saccades/min) compared to wildtype siblings (20 saccades/min) (Figure 3C). This visual deficit is specific to rlbp1b ^−/− larvae as rlbp1a ^−/− larvae display an equivalent dim light OKR as wildtype siblings (Figure 3D). In double knockouts, the dim light OKR of rlbp1a ^−/−;rlbp1b ^−/− larvae (6 saccades/min) is significantly (p < 0.0018, two tailed t‐test) lower (62%) than rlbp1b ^−/− knockouts (9 saccades/min), suggesting a synthetic genetic interaction wherein rlbp1a also supports dim light visual behavior but this function is not required when rlbp1b is functional (Figure 3E).

In contrast to dim light visual responses, we queried if exposure to an extended period of intense light (81 000 lx over 24 h at 99–123 hpf) revealed any visual behavior phenotypes (Figure 3F). However, there is no difference in response between the rlbp1b ^−/− and rlbp1a ^−/−;rlbp1b ^−/− knockout lines after extended bright light exposure followed by a standard OKR assay (Figure 3G). Furthermore, these rlbp1b ^−/− and rlbp1a ^−/−;rlbp1b ^−/− knockout lines respond at the same level as expected from wildtype larvae (~20 saccades/min) suggesting that bright light treatment does not cause a visual behavior disruption in the absence of Cralbp (Figure 3G). When this bright light treatment is repeated followed by the dim light OKR assay, rlbp1b ^−/− larvae show a significant (p < 0.0001, two tailed t‐test) reduction (67%) in visual response compared to wildtype siblings (Figure 3H). This mean OKR reduction following bright light treatment is slightly larger than the reduction in OKR response under standard conditions, but is not significantly different (p = 0.3375, two tailed t‐test).

3.4. RPE Expressed rlbp1b/Cralbpb Is Also Required for Canonical Larval Zebrafish Visual Motor Responses

To complement the OKR deficits identified in larvae lacking the RPE‐expressed rlbp1b, we assessed their visual motor response (VMR), a locomotor behavioral response to light‐change [34]. This allows us to quantify (i) overall locomotor activity, (ii) overall activity only during the light on, or only during the light off periods, and (iii) peak locomotor activity when the light comes on (MAX ON) or off (MAX OFF) (Figure 4A).

FIGURE 4.

rlbp1b knockout larvae have a deficit in VMR behavior when dark adapted. (A) Schematic of raising and experimental parameters for standard VMR assays using < 131 hpf rlbp1b ^−/− knockout zebrafish larvae. (B) Graph showing the overall average activity of rlbp1b ^+/+ and rlbp1b ^−/− larvae (n ≥ 30 larvae per group, three biological replicates). (C) Average activity from 25 s before and after the both lights on transitions. (D) Average activity from 25 s before and after both lights off transitions. (E) Average activity per larvae during light periods of the experiment. (F) Average activity per larvae during dark periods of the experiment. (G) Average peak activity in the 5 s following both lights on transitions per larvae. (H) Average peak activity in the 5 s following both lights off transitions per larvae. (I) Schematic of raising and experimental parameters for 3 h dark adapted VMR assays using < 131 hpf rlbp1b ^−/− knockout zebrafish larvae. (J) Graph showing the overall average activity of rlbp1b ^+/+ and rlbp1b ^−/− larvae (n ≥ 36 larvae per group, three biological replicates). (K) Average activity from 25 s before and after both lights on transitions. (L) Average activity from 25 s before and after both lights off transitions. (M) Average activity per larvae during light periods of the experiment. (N) Average activity per larvae during dark periods of the experiment. (O) Average peak activity in the 5 s following both lights on transitions per larvae. (P) Average peak activity in the 5 s following both lights off transitions per larvae. p = 0.0099, Mann–Whitney U test. All graphs show mean ± standard deviation except C, D, K, L which show the standard error of the mean.

Under standard VMR conditions, rlbp1b ^−/− larvae behave equivalently to wildtype siblings in terms of overall activity with mean ± standard deviation movements of 0.007 ± 0.011 ms/s and 0.010 ± 0.014 ms/s, respectively. Average movement over the course of a VMR experiment is shown in (Figure 4B). This graph shows that the larvae show less movement during the lights‐on period compared to the light‐off period, with characteristic spikes in activity immediately following light changes (Figure 4C,D). Close overlap in the trace lines representing rlbp1b ^−/− larvae and WT siblings indicates that there is no major difference in behavior between the two genotypes, with no significant difference in overall activity between both genotypes. Further analysis of larval movement consisted of plotting the average movement or max movement of individual larvae over a defined period of time. No significant change in average movement is seen between rlbp1b ^−/− larvae and WT siblings during the light ON periods, with mean ± standard deviation movements of 0.00069 ± 0.002 ms/s and 0.0014 ± 0.0035 ms/s from rlbp1b ^−/− larvae and WT siblings, respectively (Figure 4E). Average activity during the lights OFF periods was also not significantly different at 0.016 ± 0.016 ms/s and 0.023 ± 0.019 ms/s from rlbp1b ^−/− larvae and WT siblings, respectively (Figure 4F). The average peak activity in the 5 s following light changes also shows no significant difference between rlbp1b ^−/− larvae and WT siblings (Figure 4G,H). During the lights‐on peak, rlbp1b ^−/− larvae and WT siblings show average movements of 0.115 ± 0.096 ms/s and 0.084 ± 0.083 ms/s, respectively (Figure 4G). During the lights‐off peak, rlbp1b ^−/− larvae and WT siblings show average movements of 0.178 ± 0.089 ms/s and 0.167 ± 0.091 ms/s, respectively (Figure 4H).

We hypothesized that dark adapting the larvae for 3 h prior to the assay may reveal a VMR deficit in rlbp1b ^−/− larvae (Figure 4I). Under these conditions, the overall activity plots of rlbp1b ^−/− larvae and WT siblings were broadly similar (Figure 4J). Further analysis shows no significant difference in overall activity between both genotypes. The movement of larvae during the experiment is similar to that of a non‐dark adapted VMR, with more movement in the dark periods compared to the light periods and spikes in activity at light changes (Figure 4K,L). However, the average activity during the dark period is increased, relative to that of the non‐dark adapted VMR. In the more specific analyses showing average activity at light changes, there appears to be no difference between rlbp1b ^−/− larvae and WT siblings during the dark to light transition (Figure 4K). However, the response of rlbp1b ^−/− larvae appears to be attenuated compared to WT siblings during the light to dark transition (Figure 4L).

In the quantitative analyses, no change in average movement is seen between rlbp1b ^−/− larvae (0.0015 ± 0.00196 ms/s) and WT siblings (0.0010 ± 0.001 ms/s) during the light ON period (Figure 4M). There was also no significant difference in average activity during the lights OFF periods with 0.038 ± 0.022 ms/s and 0.048 ± 0.028 ms/s from rlbp1b ^−/− larvae and WT siblings respectively (Figure 4N). In relation to the peak activity in the 5 s following the dark to light changes, no significant difference between rlbp1b ^−/− larvae (0.1065 ± 0.09124 ms/s) and WT siblings (0.1147 ± 0.09971 ms/s) was observed (Figure 4O). However, the Max OFF activity during the 5 s following the light to dark transition is significantly different between rlbp1b ^−/− larvae and WT siblings (Mann–Whitney test, p = 0.0099), with average movements of 0.125 ± 0.098 ms/s and 0.171 ± 0.079 ms/s respectively (Figure 4P).

3.5. Retinoid Profiles Are Altered in < 131 hpf rlbp1b ^−/− Larvae

Following identification of altered visual behavior responses in rlbp1b ^−/− larvae, it was hypothesized that visual retinoids would be altered. Here, retinoid profiling was performed using light adapted 120–124 hpf whole larvae at midday (Figure 5A). This revealed a significant (two tailed t‐test, p = 0.0443) decrease (36%) in the level of total retinoids between WT and rlbp1b ^−/− larvae (Figure 5B). This total decrease predominantly stems from significantly (two tailed t‐test, p = 0.0015) lower (62%) levels of 11cRal between WT and rlbp1b ^−/− larvae (Figure 5C). Levels of atRAL were also significantly (two tailed t‐test, p = 0.0015) reduced (69%) between WT and rlbp1b ^−/− larvae (Figure 5D). Levels of trace retinoids 13cRal and 9cRal showed a non‐significant reduction in rlbp1b ^−/− larvae compared to WT (Figure 5E,F). There was also a non‐significant trend to increase in levels of retinal palmitate in rlbp1b ^−/− larvae compared to WT (Figure 5G,H). There was no significant difference between atROL levels between the WT and rlbp1b ^−/− larvae (Figure 5I).

FIGURE 5.

rlbp1b knockout larvae have altered retinoid metabolism. (A) Schematic of experimental conditions where 50 whole larvae were collected at midday day 5 following standard raising conditions. Following this, retinoids were profiled by HPLC. All graphs represent three biological replicates for each sample. (B) Total retinoid levels are significantly reduced in rlbp1b knockout larvae (p = 0.0443, two tailed t‐test). (C) 11‐cis‐retinal levels are significantly reduced in rlbp1b knockout larvae (p = 0.0015, two tailed t‐test). (D) all‐trans‐retinal levels are significantly reduced in rlbp1b knockout larvae (p = 0.0263, two tailed t‐test). (E, F) levels of less abundant 13‐cis‐retinal and 9‐cis‐retinal do not show any significant change in rlbp1b knockout larvae. (G, H) levels of retinal esters do not show any significant change in rlbp1b knockout larvae. (I) There is no significant change in levels of all‐trans‐retinol in rlbp1b knockout larvae. All graphs show mean ± standard deviation.

3.6. Proteomic Profiling of rlbp1b Knockout Zebrafish Eyes Reveals Differentially Expressed Proteins Related to Vitamin A Metabolism, Ferroptosis, and Lipid Metabolism and Storage

Following the identification of visual behavior and retinoid metabolism defects in the rlbp1b knockout line, we decided to carry out unbiased proteomic profiling to examine which proteins and pathways are dysregulated in Cralbpb knockout, whole adult eyes.

From 9107 proteins detected, 163 were determined to be differentially expressed based on statistical significance (p < 0.05) and a fold change cut off of 2 (Figure 6A). Individual biological replicates for each genotype clustered closely together as shown by heatmap analysis (Figure 6B). Further analysis revealed 69 upregulated proteins and 94 downregulated proteins in the rlbp1b knockout compared to wildtype. This includes Cralbpb (−2.3 fold reduction) confirming successful knockout of rlbp1b in this line (Figure 6C). Gene ontology enrichment analysis using the DAVID tool revealed significant enrichment of differentially expressed proteins for the biological processes ‘visual perception’, ‘phototransduction’, and ‘lipid storage’ and molecular function term ‘retinol binding’ (Figure 6D,E). These processes are particularly relevant to the loss of rlbp1b as they are linked to the role of Cralbp as a visual cycle protein and the known accumulation of lipid droplets in Rlbp1 knockout models and patients with mutations in RLBP1.

FIGURE 6.

Proteomic profile of rlbp1b knockout zebrafish eyes shows defects in proteins related to vitamin A metabolism, ferroptosis, and lipid metabolism/storage. (A) Summary statistics of differentially expressed proteins identified in this experiment. (B) Heatmap showing clustering of differentially expressed proteins in wildtype and rlbp1b knockout zebrafish eyes. (C) Volcano plot showing differentially expressed upregulated and downregulated proteins in rlbp1b knockout zebrafish eyes. (D) DAVID gene ontology enrichment analysis showing enriched ‘Biological Process’ terms in this dataset. (E) DAVID gene ontology enrichment analysis showing enriched ‘Molecular function’ terms in this dataset. (F) Selected vitamin A metabolism proteins that are differentially expressed (Red—upregulated, Blue—downregulated) in the rlbp1b knockout or not statistically different in the rlbp1b knockout. (G) KEGG pathway for Retinol metabolism with differentially expressed proteins highlighted. (H) Differentially expressed Ferroptosis proteins in this dataset. (I) KEGG pathway for Ferroptosis with differentially expressed genes highlighted. (J) Differentially expressed Lipid metabolism and storage genes in this dataset.

Differentially expressed proteins involved in vitamin A metabolism are shown in Figure 6F. From this key visual cycle proteins lecithin retinol acyltransferase (Lrat) (3.76 fold), retinol dehydrogenase 5 (Rdh5) (3.67 fold), and the vitamin A transporter receptor for retinol uptake STRA6 (Stra6) (2.93 fold) are significantly upregulated in the rlbp1b knockout. The zebrafish orthologs of Rgr, a key photoisomerase of the non‐canonical visual cycle, are also significantly upregulated (Rgra; 2.84 fold change, Rgrb, 2.2 fold change). Levels of other visual cycle proteins are upregulated in the rlbp1b knockout with fold changes of 1.9 and 1.6, respectively, seen for Rpe65a and Cralbpa. The positions of differential expressed visual cycle proteins in the KEGG Retinol Metabolism pathway is shown in Figure 6G. Notably, proteins regulating ferroptosis were also significantly differentially expressed in rlbp1b knockouts (Figure 6H). The positions of these genes in the KEGG ferroptosis pathway is shown in Figure 6I. Finally, the proteomic dataset also revealed the presence of differentially expressed proteins involved in lipid metabolism and storage (Figure 6J). This includes upregulation (3.45 fold change) of the zebrafish homolog of fatty acid CoA ligase Acsl3 (Acsl3) an enzyme that converts free long‐chain fatty acids into fatty acyl‐CoA esters. The zebrafish homolog of diacylglycerol O‐acyltransferase 1 (Dgat1), an enzyme responsible for the synthesis of triglycerides is also upregulated (2.03 fold change). Both Acsl3 and Dgat1 have been linked to the formation of lipid droplets and the suppression of ferroptosis [46, 47]. Also of interest is the robust upregulation of α‐, β‐, and γ‐crystallins in the rlbp1b knockout (Figure S3). These are structural proteins present in the lens but have protein chaperone functions in other tissues [48]. The heat shock protein family member HSP70 is also upregulated following loss of rlbp1b (Figure S3).

3.7. Assessment of Humanized Zebrafish Lines to Resolve Patient RLBP1 Variant Pathogenicity

Following identification of a robust visual behavior phenotype in larval rlbp1b ^−/− zebrafish, we became interested in leveraging the zebrafish model and Tol2 transgenesis to assess human RLBP1 patient variants as conducted for other genes (reviewed in [49]). The first strategy to assess patient variant pathogenicity was to use a dim light OKR complementation assay with human CRALBP proteins tagged with eGFP, expressed in the RPE of rlbp1b ^−/− zebrafish (Figure 7A). Human CRALBP protein is 71% identical to zebrafish Cralbpb at the amino acid level, suggesting high functional conservation (Figure 7B). The first 36 residues of the CRAL/TRIO domain do not show as much conservation in percentage identity as the rest of the domain (Figure 7B). This is likely due to a high stringency applied to the alignment; indeed, the NCBI conserved domain search tool reports the CRAL/TRIO and Sec14 domains as conserved (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). Human CRALBP is known to have a structure comprising a N‐terminal region with 5 α‐helices, a core retinoid binding cavity made up of a β‐sheet with 1 antiparallel and 4 parallel strands, and a C‐terminal region containing a 3‐layer αβα sandwich [50]. The zebrafish Cralbp structures show close similarity with the human structure (Figure 7C), indicating that the structural features described above are conserved in the zebrafish. Of note, the Cralbpb protein is 10 residues shorter than the human protein, as seen in the C terminus of the protein structure (Figure 7C). Notably, structural modeling of known RLBP1 patient variants predicted no significant structural changes (Figure 7D), including in the p.R151Q mutation known to abolish retinoid binding [5]. This highlights a limitation of in silico pathogenicity prediction tools and shows why functional assays are needed.

FIGURE 7.

Human RLBP1 represents a good target for a humanized zebrafish approach. (A) Schematic of the humanized zebrafish lines approach used in this study. (B) Alignment of Cralbp protein sequences using the TCoffee alignment tool. Domains mapped to these sequences manually from the NCBI conserved domain search tool. Pathogenic mutations in CRALBP are taken from [6]. (C) Structural model of human CRALBP and zebrafish paralogs created using the SWISS‐MODEL web tool. (D) Structural model of human CRALBP and missense variants p.R151Q and p.A237V using the SWISS‐MODEL web tool.

3.8. Transgenic Expression of EGFP‐Tagged WT Human CRALBP Is Not Sufficient to Rescue Dim Light Vision in the rlbp1b Zebrafish Knockout

Humanized transgenic zebrafish lines were generated using the expression vectors shown in (Figure 8B) and corresponding larvae assayed using the dim light OKR assay (Figure 8A). In the presence of the transgene tg(rpe65a:EGFP‐Hsa.RLBP1), rlbp1b ^−/− larvae still show a significant (p < 0.001, Mann–Whitney test) 60% reduction in dim light OKR compared to WT siblings (Figure 8C). The lack of rescue produced by the WT human CRALBP EGFP tagged transgene precluded assessment of human RLBP1 variants in a complementation assay. However, we could still test if CRALBP disease variant lines modulate the dim light OKR. Notably, rlbp1b ^−/− larvae expressing RLBP1 p.R151Q or RLBP1 p.A237V show a significant reduction in OKR response (77% and 57%, respectively) compared to wildtype siblings (rlbp1b ^+/+) expressing the same human transgenes (Figure 8D,E). This reduction in OKR response is the same as in rlbp1b ^−/− larvae expressing the WT human transgene or non‐transgenic larvae. This indicates that expression of these mutant forms of CRALBP do not cause any additional impairment of larval OKR responses (Figure 8D,E).

FIGURE 8.

The EGFP‐tagged zebrafish rlbp1b gene but not the human RLBP1 gene can rescue the dim light OKR in rlbp1b knockout larvae. (A) Schematic of standard raising and dim light OKR approach used to assess rescue in humanized zebrafish lines. (B) Schematic of expression vectors used to create humanized EGFP‐tagged RLBP1 zebrafish lines used in this study. (C) The EGFP‐tagged human RLBP1 gene is not sufficient to rescue the dim light OKR phenotype in rlbp1b knockout larvae (p < 0.0001, Mann Whitney test). Three biological replicates n ≥ 16 larvae per group. (D, E) The expression of missense RLBP1 disease variants p.R151Q and p.A237V do not cause any additional impairment of the dim light OKR response in rlbp1b knockout larvae. p.R151Q 3 biological replicates n ≥ 12 larvae per group (p < 0.0001, two tailed t‐test). p.A237V 3 biological replicates n ≥ 11 larvae per group (p < 0.0047, Welch's t‐test). (F) Schematic of expression vectors used to create zebrafish lines expressing EGFP‐tagged rlbp1b used in this study. (G) F0 rescue of the dim light OKR phenotype using the tg(rpe65a:eGFP‐rlbp1b) transgene in the rlbp1b knockout larvae. Three biological replicates n = 24 rlbp1b ^−/− larvae per group (p = 0.0086, two tailed t‐test). (H) Rescue of the dim light OKR phenotype in F1 larvae stably expressing tg(rpe65a:eGFP‐rlbp1b). Three biological replicates n = 27 rlbp1b ^−/− larvae per group (p < 0.0001, two tailed t‐test). (I) Stable expression of the tg(rpe65a:eGFP‐rlbp1bp.R151Q) fails to rescue the dim light OKR phenotype in rlbp1b knockout larvae. No significant difference between non‐transgenic larvae and larvae with the tg(rpe65a:eGFP‐rlbp1bp.R151Q) transgene. Significant difference between larvae with the wildtype vs. p.R151Q mutant transgene (p < 0.0053, Kruskal–Wallis test). Three biological replicates n = 25 rlbp1b ^−/− larvae per group. All graphs show mean ± standard deviation.

3.9. Transgenic Expression of Wildtype, but Not R151Q Zebrafish Cralbpb Is Sufficient to Rescue Vision in Knockout Zebrafish; Providing a Platform for Human RLBP1 Variant Pathogenicity Prediction

Instead of expressing human CRALBP variants, the alternative strategy adopted was to express zebrafish Cralbpb with patient variants. Schematics of expression vectors used to generate zebrafish lines expressing the fish rlbp1b gene and disease‐relevant variants are shown in (Figure 8F). Importantly, the Tg(rpe65a:eGFP‐rlbp1b) transgene could rescue rlbp1b ^−/− even in the F0 mosaic generation. In these experiments, rlbp1b ^−/− larvae injected with and screened for the presence of the Tg(rpe65a:eGFP‐rlbp1b) transgene have a significantly (p = 0.0086, two tailed t‐test) higher (36%) dim light OKR response compared to buffer‐injected siblings (Figure 8G). This was subsequently confirmed in F1 larvae stably expressing the T g( rpe65a:eGFP‐rlbp1b) transgene, with transgenic larvae showing a significant (p < 0.001, Welch's t‐test) increase (43%) in dim light OKR response compared to siblings without the transgene (Figure 8H). To assess the ability of this platform to evaluate the pathogenicity of missense mutations in RLBP1, another zebrafish line was created expressing the zebrafish Cralbpb protein with the known loss‐of‐function p.R151Q mutation. F1 rlbp1b ^−/− larvae expressing the T g( rpe65a:eGFP‐rlbp1bp.R151Q) transgene show no rescue of the rlbp1b ^−/− dim light OKR response (Figure 8I). In rlbp1b ^−/− larvae with the T g( rpe65a:eGFP‐rlbp1bp.R151Q) transgene, there is no significant difference (p = 0.6279, Kruskal–Wallis test) between their dim light OKR response and that of non‐transgenic rlbp1b ^−/− larvae (Figure 8I). Additionally, there remains a significant (p = 0.0053, Kruskal–Wallis test) difference (37%) between rlbp1b ^−/− larvae with the T g( rpe65a:eGFP‐rlbp1b) transgene and those with the T g( rpe65a:eGFP‐rlbp1bp.R151Q) transgene (Figure 8I).

4. Discussion

This study resolves conflicting reports on the nomenclature and localization of the zebrafish rlbp1a and rlbp1b genes. Here, using two different scRNAseq databases and immunostaining of the knockout lines, our findings confirm that rlbp1a is expressed in the Müller glia, whereas rlbp1b is localized to the RPE. These discrepancies may have arisen due to misannotations in the ZFIN database (Zfin.org) where rlbp1a and rlbp1b have past names listed as rlbp1b and rlbp1a respectively. These findings provide greater clarity for the field and facilitate future research.

We have leveraged zebrafish genetic models to dissect the distinct contributions of RPE‐expressed rlbp1b and Müller glia‐expressed rlbp1a to visual behavior and retinoid metabolism. Our results demonstrate that loss of the RPE‐enriched rlbp1b expression results in a pronounced deficit in dim light–mediated OKR (~5 lx) but not in OKRs performed under standard light conditions (~36 k lx). This dim light OKR deficit is specific to rlbp1b, as Müller glia‐expressed rlbp1a do not have an impairment. These results align with recent studies [14, 30] which suggest that CRALBP plays a more prominent role in the visual cycle through the RPE than the Müller glia. The Müller glia‐expressed rlbp1a form does appear to contribute to the visual cycle, as dim light OKR responses from the double rlbp1a:rlbp1b knockout show a significant reduction compared to the single rlbp1b knockout. This is in line with previous reports of additive effects in the double rlbp1a:rlbp1b knockout [14].

The role of rlbp1a in the Müller glia‐dependent visual cycle is likely linked to the photoisomerase Rgr (reviewed in [51]). As Rgr is thought to support the visual cycle in periods of constant photopic illumination, we assayed if loss of Cralbp revealed related phenotypes by subjecting larvae to 24 h of intense light starting at 99 hpf and continuing until immediately before the OKR. Under these conditions, no OKR deficit is seen in the rlbp1b or double rlbp1a:rlbp1b knockout. Here, the rlbp1b knockout represents an isolated system where the contribution of Cralbp to the visual cycle is through the Müller glia expressed rlbp1a. The absence of an OKR phenotype in the double rlbp1a:rlbp1b knockout suggests that the light treatment did not sufficiently impair the visual cycle to induce a detectable behavioral deficit. As we have previously shown that the rlbp1b knockout has a dim light OKR when raised under normal light conditions we decided to subject these larvae to the same intense light treatment followed by dim light OKR. Under these conditions we see the same level of reduction in the rlbp1b knockout as when the assay is performed under standard conditions. This shows that exposure to a period of elevated levels of light does not further affect visual behavior.

After identifying a robust dim light OKR phenotype in the rlbp1b knockout, we examined whether these larvae exhibited deficits in their VMR. Similar to the OKR results, the rlbp1b knockout line does not exhibit a VMR phenotype under standard conditions. However, a deficit is observed in the light‐off peak movement when larvae are dark‐adapted for 3 h before the assay. The VMR is a robust visual behavior assay with more utility than the OKR assay for higher throughput and drug screening applications [34]. However, we opted to use only the OKR assay, as it provides a more direct measure of visual behavior in rlbp1 knockout lines. There has been one other report of OKRs under low light levels, which were used to examine the contribution of rods to larval zebrafish vision under very low light levels [52].

This study provides the first evidence of a visual behavior phenotype in a rlbp1 knockout zebrafish model. It was unexpected that the RPE‐expressed rlbp1b knockout exhibited no OKR phenotype under standard conditions, given prior reports of an ERG defect [14]. The choice to carry out dim light OKR assays was inspired by and mirrors closely the night blindness symptoms seen in patients with RLBP1 mutations [6, 45]. As rlbp1b acts in the visual cycle, we hypothesized that the reason for this OKR deficit was due to impaired retinoid metabolism. Retinoid profiling of < 131 hpf rlbp1b knockout larvae raised under standard conditions revealed significant decreases in total retinoids, 11cRAL, and atRAL. This is in line with previous reports in zebrafish and mouse models of Rlbp1 deficit [3, 14, 30]. However, it is interesting that despite having these decreased levels of retinoids, rlbp1b knockout larvae do not have a defect in standard OKR conditions. It is likely that the level of retinoids is sufficient for visual behavior under bright light conditions but falls below a certain threshold in dim light conditions and fails to elicit a visual behavior response. Notably, a greater than 50% reduction in the visual chromophore 11cRal does not induce an OKR phenotype under bright light conditions. This underscores the importance of light levels in visual behavior assays when investigating mutants with potential visual cycle disruption.

To gain mechanistic insights into factors in the eye that become dysregulated following Cralbpb loss, a proteomic profile was generated from rlbp1b knockout zebrafish eyes. This provides effective validation of the knockout line, showing that the Cralbpb protein is significantly downregulated, while the Cralbpa protein showed a 1.6‐fold upregulation. The rlbp1b knockout line shows disruption of proteins related to vitamin A metabolism, ferroptosis, and lipid storage/metabolism (Figure 9). These include upregulation of the zebrafish orthologues of the key visual cycle proteins Lrat and Rdh5. Lrat acts upstream of Cralbp binding to 11cRal (Figure 9) and therefore its upregulation could be explained by increased levels of its substrate or product [53]. Rdh5 upregulation following Cralbpb loss is likely also compensatory, to respond to reduced levels of its product 11cRal (Figure 9). Levels of the Rpe65 homolog Rpe65a are also increased, with a fold change of 1.9. The levels of the vitamin A transporter Stra6 are also significantly upregulated following rlbp1b loss [54]. This suggests a mechanism where the RPE responds to decreased levels of 11‐cis retinoids by upregulating proteins responsible for vitamin A intake and visual cycle metabolism in the RPE.

FIGURE 9.

Schematic of key upregulated genes in vitamin A metabolism and lipid metabolism pathways in the rlbp1b knockout.

Additionally, this proteomic profile shows upregulation of proteins that negatively regulate ferroptosis. This includes the zebrafish homolog of the protein Acsl3, which acts as a suppressor of ferroptosis [55]. The rlbp1b knockout proteomic profile is also enriched for proteins involved in lipid metabolism and storage. Dgat1, a key enzyme for lipid droplet formation that acts directly downstream of Acsl3, was also significantly upregulated [47]. This is unsurprising given reports of lipid droplet accumulation in rlbp1 knockout models and the presence of white lesions in fundus images of patients with RLBP1 mutations [6].

The combined presence of lipid droplets related proteins and ferroptosis related proteins is intriguing and suggests a possible mechanism involved in RLBP1 pathogenesis. Ferroptosis has recently been reported as a key player in the progression of both age related macular degeneration (AMD) and RP. However, recent reports suggest that the lipid droplets confer a cell with resistance to ferroptosis [56, 57]. ACSL3, which is upregulated following rlbp1b loss, is an unfavorable prognostic marker in cancer due to its role in suppressing ferroptosis [46]. Additionally ACSL3 suppression has been reported to induce ferroptosis in hippocampal neurons [58]. Dgat1, an enzyme responsible for lipid droplet formation that acts downstream of Acsl3 has also been shown to confer resistance to ferroptosis in a lipid droplet dependent manner [47, 59, 60]. Taken together, this suggests that the presence of lipid droplets in patients with mutations in RLBP1 may be due to upregulation of Acsl3 and Dgat1. These lipid droplets may be partly conferring resistance to the ferroptosis dependent cell death commonly seen in AMD and RP. This suggests that treatment with inhibitors of lipid droplet formation may sensitize RPE cells to ferroptosis mediated cell death. Other clusters of proteins upregulated in the rlbp1b knockout include members of the crystallin family and the heat shock protein family member HSP70. These proteins are interesting as they have general protein chaperone roles in response to cell stress, suggesting that they are upregulated in response to cellular stress caused by the loss of rlbp1b [48, 61]. Additionally, the presence of heat shock protein family members and their related proteins is interesting as it has previously been demonstrated that the expression levels of RLBP1 is linked to the level of a different heat shock protein family member HSP90 [62].

Using the robust dim light OKR visual behavior assay identified here and the genetic malleability of zebrafish, we endeavored to explore the utility of humanized zebrafish models as pathogenicity prediction tools. For this system to be effective the human version of the RLBP1 gene needed to be sufficient to rescue the rlbp1b knockout phenotype. Human genes have been expressed in zebrafish for this purpose before, with previous successful and unsuccessful reports published (reviewed in [49]). The work presented here represents the first attempt at using this approach to model a recessive inherited retinal disease in zebrafish. Unfortunately, the expression of an EGFP‐tagged WT human RLBP1 gene was not sufficient to rescue the dim light OKR deficit in rlbp1b knockout larvae. This results in the inability to use this approach to model mutations for pathogenicity using the human gene. The failure of the human gene to rescue could have been for several reasons including incompatibility in protein–protein interactions with zebrafish proteins, interference from the GFP tag, or expression differences as while an RPE specific promoter was used it was not the endogenous rlbp1b promoter. As lines expressing missense mutant versions of the human protein were created, they were assessed to see if they caused any additional effect on the dim light OKR of rlbp1b knockout larvae. No additional negative effect was seen in these human missense mutant lines which is consistent with reports in the literature that report these mutations as recessive loss of function.

Rather than examining why the human gene failed to rescue the rlbp1b knockout phenotype, we assessed whether an EGFP‐tagged zebrafish rlbp1b gene was sufficient. Using the same expression system, we were able to see a significant rescue of the dim light OKR in larvae expressing the fish gene in both the F0 generation and the F1 generation. This indicates that the expression system is sufficient to rescue the knockout phenotype, but it is the human gene that is incompatible with zebrafish. To further show that this system is sufficient to model mutations in RLBP1, we created a transgenic line expressing the fish rlbp1b containing the p.R151Q, which is known to be pathogenic in humans. Larvae expressing this p.R151Q mutant transgene show no rescue of their OKR response, with no significance between them and non‐transgenic larvae. These results validate the zebrafish model as a tool for pathogenicity prediction and establish it as the first in vivo system for assessing RLBP1 mutations.

Author Contributions

John D. Fehilly: conceptualization, data curation, formal analysis, investigation, methodology, validation, and visualization. Tess McCann: data curation, formal analysis, investigation, methodology, validation, and visualization. Grace Ruddin: data curation, formal analysis, investigation, methodology, and validation. Joanna J. Kaylor: data curation, formal analysis, investigation, methodology, and validation. Hannah Grenville: data curation, formal analysis, investigation, methodology, and validation. Rebecca Ward: data curation, formal analysis, investigation, methodology, and validation. Kieran Wynne: data curation, formal analysis, investigation, methodology, and validation. Alicia Gómez Sánchez: data curation, formal analysis, investigation, methodology, and validation. Elin Strachan: data curation, formal analysis, investigation, methodology, validation. Gabriel H. Travis: conceptualization and methodology. Ross F. Collery: methodology. Breandán N. Kennedy: conceptualization, data curation, formal analysis, funding acquisition, methodology, project administration, resources, supervision, validation, and visualization. All authors were involved in drafting and revising the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1. PCR and sequencing of transgenic lines used in this study. (A) Schematic of location of PCR and sequencing primers relative to the transgenes used in this study. (B‐F) Agarose gels showing PCR product size and accompanying sequencing data with mutations highlighted for the transgenes; (B) Tg(rpe65a:EGFP‐Hsa.RLBP1), (C) Tg(rpe65a:EGFP‐Hsa.RLBP1_R151Q), (D) Tg(rpe65a:EGFP‐Hsa.RLBP1_A237V), (E) Tg(rpe65a:EGFP‐rlbp1b), and (F) Tg(rpe65a:EGFP‐rlbp1b_R151Q).

Figure S2. Top 45 downregulated proteins in the rlbp1b knockout compared to wildtype.

Figure S3. Top 45 upregulated proteins in the rlbp1b knockout compared to wildtype.

Fehilly J. D., McCann T., Ruddin G., et al., “Germline Disruption of Retinal Pigment Epithelium‐Expressed Zebrafish rlbp1b ^−/− Results in Selective Dim Light Visual Behavior Deficits and Provides a Screening Platform for Evaluating the Pathogenicity of Human RLBP1 Variants,” The FASEB Journal 39, no. 12 (2025): e70754, 10.1096/fj.202500600R.

Data Availability Statement

All data is contained in the figures and results section of this manuscript. The proteomic profiling data will be available in an online repository pending publication.