Gene therapy for RAB28: What can we learn from zebrafish?

The eye is particularly suited to gene therapy due to its accessibility, immunoprivileged state and compart-mentalised structure. Indeed, many clinical trials are underway for therapeutic gene strategies for inherited retinal degenerations (IRDs). However, as there are currently 281 genes associated with IRD, there is still a large unmet need for effective therapies for the majority of IRD-causing genes. In humans, RAB28 null and hypo-morphic alleles cause autosomal recessive cone-rod dystrophy (arCORD). Previous work demonstrated that restoring wild type zebrafish Rab28 via germline transgenesis, specifically in cone photoreceptors, is sufficient to rescue the defects in outer segment phagocytosis (OSP) observed in zebrafish rab28 -/- knockouts (KO). This rescue suggests that gene therapy for RAB28 -associated CORD may be successful by RAB28 gene restoration to cones. It also inspired us to critically consider the scenarios in which zebrafish can provide informative pre-clinical data for development of gene therapies. Thus, this review focuses on RAB28 biology and disease, and delves into both the opportunities and limitations of using zebrafish as a model for both gene therapy development and as a diagnostic tool for patient variants of unknown significance (VUS).


General introduction
Inherited retinal degenerations (IRDs) are a group of disorders which collectively account for the leading cause of vision loss in those of working age (Heath Jeffery, 2021).IRDs are characterised predominantly by the progressive dysfunction and death of light-sensing photoreceptor cells, and the supporting retinal pigment epithelium (RPE) cells (Duncan, 2018).IRDs are genetically heterogeneous, and there are currently 281 genes associated with IRD (Amato, 2021;RetNet).Causative genes can be inherited in autosomal dominant, autosomal recessive or X-linked patterns, resulting in varying degrees of symptom manifestation in family pedigrees (Nash et al., 2015).
Cone-rod dystrophies (CORD) constitute a subgroup of IRDs with a prevalence of 1-9/100,000, and characterised by the initial loss of cone cells, followed by progressive loss of rod photoreceptors (Gill et al., 2019).The clinical hallmarks of CORD include poor visual acuity, impaired colour vision and photophobia, with later-onset impairment of night and peripheral vision (Gill et al., 2019;Nassisi, et al., 2022).The most common cause of autosomal recessive CORD is mutations in the ABCA4 gene, accounting for 30-60% of cases (Manitto, 2015).The remaining CORD cases are caused by mutations in >30 genes, with approximately two thirds of these genes associated with autosomal recessive disease (Manitto, 2015).In 2013, the small ciliary G-protein RAB28 was identified as the first Rab protein to be associated with CORD (Roosing, 2013).In humans, RAB28 null and hypomorphic alleles cause autosomal recessive cone-rod dystrophy (arCORD) (Fig. 1).A potential explanation for the underlying pathology is that loss of Rab28 function is linked with defects in outer segment phagocytosis (OSP).This highly-regulated process is one means by which the retina controls photoreceptors to maintain vision.OSP involves engulfment and digestion of demarcated distal tips of the photoreceptor outer segments by the RPE.This enables removal of waste products but also recycling of molecular resources in the retina, preventing accumulation of toxic byproducts and maintaining vision.We are interested in the translation of RAB28 therapies from preclinical models into clinical trials.Therefore, in this review, we discuss relevant previous studies on Rab28 biology and delve into the bespoke, and arguably unanticipated, opportunities for using the zebrafish model as both a diagnostic tool and a preclinical model for RAB28 gene therapy development.
Treatment options for IRDs are scarce, with only one European Medicines Agency (EMA) and U.S. Food and Drug Administration (FDA) approved therapy developed to date (Nuzbrokh et al., 2021).Luxturna (also called voretigene neparvovec-rzyl), a gene replacement therapy for Leber congenital amaurosis (LCA) patients with mutations in the RPE65 gene, was authorised for use in December 2017, providing a treatment option for approximately 6% of patients with LCA (Maguire et al., 2021).The eye is particularly suited to gene therapy due to its accessibility for drug delivery, immunoprivileged status and compartmentalised structure.Approval of Luxturna provided a huge advancement for the IRD field, and there are currently up to 20 other gene therapy candidates undergoing clinical trials (Amato, 2021;Garafalo, 2020).However, notably in recent follow-up studies, clinical cases are reported where patients developed inflammation (Parrulli et al., 2022;Kessel et al., 2022) and chorioretinal atrophy following injection with voretigene neparvovec-rzyl (Kolesnikova, 2022;Gange, 2022), in addition to several patients experiencing a waning of the initial visual improvement observed following treatment (Kessel et al., 2022).These outcomes were not unanticipated based on earlier studies which reported unabated photoreceptor degeneration despite initial improved vision, similar declines in preserved RPE in treated and untreated fellow eyes and occurrence of a serious adverse event, a localized intraretinal immune response, following RPE65 gene therapy (Dimopoulos, 2018;Cideciyan, 2013).These reported poor outcomes, not to mention the significant economic costs of Luxturna (Darrow, 2019) suggest we still have much to resolve before IRD gene therapy can be implemented widely.
Non-genetic approaches for IRD treatment include the use of stem cells, drugs, and light-sensitive prostheses, the vast majority of which are not approved, but in preclinical or clinical trials (Duncan, 2018;Hinkle et al., 2021;Sundaramurthi, 2020).The basis of gene therapy is to ameliorate the impact of the underlying genetic defect by either restoring or enhancing the function of a protein (gene addition or gene editing), or by silencing the expression of a gene causative or linked to the pathobiology (gene silencing or gene editing).Gene therapy can be implemented in broadly 3 different manners; ex vivo, in vivo or in situ, with development often requiring individual approaches for each genetic variant (Papanikolaou & Bosio, 2021).However, with a more personalised approach, often fewer patients benefit.This is the case for RAB28-related CORD which raises barriers for gene therapy development as approved products are very costly and the patient population to conduct clinical trials is relatively small and globally distributed.
At first sight, the zebrafish appears to have relatively little to offer in preclinical development of a RAB28 gene therapy.For the canonical approach of viral vectors for gene augmentation, there are appropriate concerns that the viral serotypes do not transduce the same cells in zebrafish as in the human retina Furthermore, human promoters may not have conserved regulation of cell-type expression and the human protein may not have conserved function in the zebrafish retina.However, in the context that gene therapy is a continuum, in which RAB28 gene/variant pathogenicity must be confirmed prior to a patient enrolling in a RAB28 gene augmentation trial (or receiving a future approved gene therapy), and that the appropriate retinal cell type(s) in which to restore the gene must be determined, we consider the zebrafish model to provide cost-effective and scalable opportunities during preclinical development.

Rab28 and IRD
An understanding of the specific clinical IRD features associated with RAB28 as well as the expression and function of Rab28 is important to provide context for RAB28 gene therapy.Roosing et al. first reported homozygous nonsense variants in RAB28 as a cause of arCORD (CORD18) in a German family and a Moroccan Jewish family (Roosing, 2013); and since then a total of 9 pathogenic alleles (p.Ser23Phe, p. Arg137* p.Gln189*, p.Leu13*, p.Thr26Valfs4*, p.Asp68His, p. Cys217Trp and an additional splice site mutation), all associated with autosomal recessive pedigrees, have been published (Table 1) (Roosing, 2013;Riveiro-Álvarez, 2015;Lee, 2017;Jespersgaard, 2020;Iarossi, 2020;Villanueva-Mendoza, 2021).Patients harbouring pathogenic mutations in the RAB28 gene present common phenotypic features, such as highly reduced visual acuity, debilitated colour vision, macular atrophy and acute photophobia (Jespersgaard, 2020).More recent studies also report polydactyly in some RAB28 patients (p.Asp68His & p. Gly19Arg).However, it is yet to be confirmed whether polydactyly is truly associated with defects in RAB28 function, or whether it is a result of a second unknown mutation in this family (Jespersgaard, 2020).
The human RAB28 gene is located on chromosome 4 (4p15.33).There are also two Rab28 pseudogenes found on chromosome 9 and X (Jespersgaard, 2020).Human RAB28 has 3 splice isoforms, each displaying variations at the C-terminus.Isoform 1 and 2 both possess prenylation motifs that differ in their C-termini, both possess prenylation motifs and differ in their C-terminal-most 29 amino acids, resulting in CaaX motif disparity (Manitto, 2015).Isoform 3 is postulated to be unprenylated as it is missing the majority of the C-terminus and the CaaX motif.Isoforms 2 and 3 are highly expressed in the retina, albeit all 3 isoforms show expression in a multitude of tissues (Manitto, 2015).The first missense mutation reported, p.C217W, leads to a defunct CaaX motif, due to loss of the essential cysteine residue (Papanikolaou & Bosio, 2021).Loss of prenylation appears to impede membrane association of RAB28, suggesting that membrane binding is essential for RAB28 functionality (Papanikolaou & Bosio, 2021).
RAB28 is a member of the Rab family (66 proteins in humans), a subgroup of the larger Ras-related small GTPase family that act as molecular switches in many different biological pathways and processes (Li & Marlin, 2015).Like most G-proteins, Rabs consist of a multi-fold β-sheet structure flanked by α-helices, and cycle between active "GTPbound" and inactive "GDP-bound" states, each having distinct conformations (Pfeffer, 2005).Rabs also contain a C-terminal-specific motif not found in the wider Ras family that is essential for protein-membrane and protein-protein associations (Blacque et al., 2018).Rab proteins are heavily involved in intracellular and plasma membrane organisation and vesicle trafficking, regulating central aspects of membrane fusion, fission, budding, and tethering, as well as protein cargo incorporation (Blacque et al., 2018;Zhen & Stenmark, 2015;Brown & Pfeffer, 2010;Wandinger-Ness & Zerial, 2014).Rabs achieve these functions by interacting with a myriad of effector proteins such as membrane coat components, molecular motors and SNARE proteins (Hutagalung & Novick, 2011).Jensen et al. 2016 reported that Rab28 is conserved in ciliated eukaryotes, with the exception of Drosophila melanogaster (Carter, 2020).
RAB28 shows less sequence similarity to other Rab proteins in comparison to what is normally observed between members of the Rab family (~32% vs. >40%) (Lee, 2017).Rab28 is farnesylated rather than Table 1 Pathologic mutations in human RAB28 and the associated clinical symptoms.CRD: cone-rod dystrophy.Amino acid conservation details whether the wildtype amino acid at the location of the pathogenic variant in human RAB28, is conserved as the identical amino acid in zebrafish.* indicates a stop codon.geranylgeranylated, possessing a CaaX motif at its C-terminus, which is essential for membrane binding (Roosing, 2013).The importance of the switch regions in Rab28 protein function is highlighted by identified patient variants occurring within one of the GTP and Mg (Duncan, 2018) + interacting motifs, where the serine residue is predicted to hydrogen bond with the bound guanosine nucleotide (Lee, 2017).Interactome analysis in the RAB28 KO mouse model identified a number of Rab28 interactants, encompassing the voltage-gated potassium channel subfamily J member 13 (KCNJ13) and a lipidated intraflagellar trafficking chaperone protein-phosphodiesterase 6 δ-subunit (PDE6D), which facilitates transport of prenylated proteins.By generating cone-photoreceptor specific transgenic models expressing Rab28, Carter et al. (2020) observed that zebrafish GFPtagged Rab28 expression is almost solely in the outer segments under the gnat2 promotor.A protein interaction screen using the cone-specific transgenics revealed GFP-tagged zebrafish Rab28 also interacts with Pde6d in addition to several phototransduction cascade components, encompassing opsins, phosphodiesterase 6C and guanylate cyclase 2D (Carter, 2020).More recently, a predicted loss of function Rab28 mutation, discovered through genome sequencing (Thr26Asn) was postulated to decrease affinity for GTP-binding by in-silico experiments (Iarossi, 2020).However, this mutation is yet to be validated using functional assays.
RAB28 is reported to be involved in the ESCRT (endosomal sorting complex required for transport) pathway in trypanosomes, allowing for delivery of lysosomal protein cargo and endosomal proteins (Lumb et al., 2011).In rat endothelial cells, Rab28 is reported to regulate activation of NF-κB, by directly interacting with the NF-κB subunit p65.The Rab28-p65 interaction is proposed to then facilitate NF-κB nuclear translocation, as in the absence of Rab28, less nuclear translocated NF-kB is observed (Jiang, 2013).In mouse skeletal muscle and rat adipocytes, Rab28 was identified as a substrate for Tre-2/BUB2/CDC16 (TBC) domain family member 1 (TBC1D1) and 4 (TBC1D4), regulating the cellular translocation of GLUT4 (Zhou, 2017).Phosphorylation of TBC1D1 and TBC1D4 are postulated to inactivate the GAP domain, resulting in increased Rab28 GTP-binding (Zhou, 2017).Similar to the Rab28 KO mouse model where the absence of PDE6D results in failure of RAB28 localisation to the outer segments, highly modified cilia (Akella, 2020); in Caenorhabditis elegans (C.elegans) (Hutagalung & Novick, 2011) lack of the PDE6D orthologue PDL-1 results in the failure of RAB-28 to localise to sensory cilia.Previous work on C. elegans uncovered that RAB-28 works in a cilia-specific manner to negatively regulate extracellular vesicle (EV) biogenesis and shedding, in addition to the discovery that localisation of C. elegans RAB-28 is nucleotide-dependent (Akella, 2020;Jensen, 2016).Little is known about the regulators and effectors of Rab28, aside from myotubularin-related protein 5 and 13, guanine nucleotide exchange factors (GEF) that are members of the DENN domain proteins (Yoshimura et al., 2010).
Overall, it is apparent that variants in RAB28 are linked to inherited human blindness in a rare form of cone-rod dystrophy.In addition to the RAB28 variants classified as pathogenic, ClinVar reports 50 variants of uncertain significance and 1 variant with potential conflicting interpretation (Landrum, 2018).Functional resolution of the pathogenicity of these variants is crucial for patients to participate in future RAB28 gene therapies.In addition, the retinal cell type(s) in which to restore RAB28 gene expression for optimal therapeutic benefit need to be elucidated.

Gene replacement as a therapeutic option for Rab28-CRD and rationale for incorporation of zebrafish models
An obvious approach to treat RAB28 related inherited CORD is to develop gene therapies.To date, there is no clinical trial investigating RAB28 gene therapy.The rarity of CORD18, the unresolved RAB28 patient variants, and the concerns over long-term benefits of Luxturna are some of the multifactorial challenges.A commercial barrier is that a CORD18 gene therapy may have high impact, but only in a small subset of IRD patients.However, conceivably, enhancing the process of OSP linked to Rab28 is of broader therapeutic relevance to other IRDs and macular degenerations.Current studies are testing the gene therapy potential of replacing MERTK, another gene associated with OSP and retinal degeneration.Preclinical studies in which an AAV vector containing a functional human MerTK gene (cDNA) was injected into a l MERTK deficient rat model showed positive results, providing rescue of OSP for all time points examined up to 6.5 months, in addition to an improvement in ERG-response amplitude (LaVail, 2016).Initial Phase I human clinical trials also showed that this intervention was well tolerated in patients with MERTK-related RP and subretinal injection of rAAV2-VMD2-hMERTK is not associated with major side effects; however, observed improvements in visual acuity waned 2 years after treatment (Ghazi, 2016).The results of the Phase 2 study have yet to be published.As there are currently no therapeutic strategies in clinical trials for RAB28-CORD, we focus here on what knowledge can be gained in zebrafish Rab28 models in order to progress human RAB28 gene therapies.
Model organisms showing similar ocular physiology to humans are necessary for researchers to fully comprehend the physiological processes ongoing in the retina, and in order to identify phenotypes associated with causative IRD genes.The precise molecular mechanisms initiating RAB28-CORD associated cell death is still unclear.To allow for the development of future therapeutic strategies, it is essential to understand the pathological mechanism at hand.Zebrafish (Danio rerio) are small freshwater teleosts, belonging to the family Cyprinidae, which has become a powerful vertebrate model organism (Ribas & Piferrer, 2014).Though more phylogenetically distant from humans in comparison to rodent models, for 76% of human disease associated genes there is a zebrafish ortholog, in addition to an ~70% overall sequence identity (Howe, 2013).Zebrafish development is relatively fast, with adult fish capable of breeding at 2-4 months of age (Lieschke & Currie, 2007).DNA sequence information and data mining tools allow for the elucidation of both conserved and novel genes in zebrafish (Howe, 2013).This along with advancements in CRISPR Cas9, and other gene editing technologies has made generating knockout lines routine (Krueger & Morris, 2022;Liu et al., 2019).Transgenic models to assess protein localisation and function can be obtained using methodologies such as Tol2 transposase-mediated transgenesis (Kawakami et al., 2016;Kwan, 2007); while transient knockdowns can be established using antisense oligonucleotide morpholinos (Timme-Laragy et al., 2012).Such methodologies have greatly enhanced the application of the zebrafish as a model for biomedical research and translational studies.Importantly, zebrafish eye development is highly similar to human and other vertebrate models (Glass and Dahm, 2004).In particular, the zebrafish retina recapitulates features of the human retina, with high levels of conservation in terms of organisation, morphology and function, despite the larval zebrafish retina lacking a typical macula (Hong & Luo, 2021;Yoshimatsu et al., 2020).Highly developed visual function in zebrafish is partially determined by their natural ecology, driving a necessity for refined visual acuity and rapid larval visual development (Yoshimatsu et al., 2020;Neuhauss, 2003) due to a diurnal lifestyle and shallow habitats with predators (Richardson et al., 2017;Colwill & Creton, 2011).Due to the prevalence of cone photoreceptors in zebrafish (Allison, 2010;Glaviano et al., 2016); they are preferable models for the interrogation of genes functioning in cone photoreceptors (Ward, 2020) as well as disorders which primarily affect these cells, such as cone-rod dystrophy.There have been a multitude of zebrafish IRD models generated over the past 3 decades, with further models added to the list each year, strengthening the utility of zebrafish in the field of IRD modelling and treatment development.The IRD models have been developed using techniques such as CRISPR Cas9 and transgenesis, or as was more common previously using mutagens such as n-ethyl-n-nitrosourea (ENU) and ethyl methane sulphonate (EMS) in forward genetic screens (Noel et al., 2022).
However, as with all model organisms, there are research limitations to zebrafish.Despite high levels of conservation, they display greater evolutionary distance from humans in comparison to rodents and higher-order primates.As zebrafish underwent whole genome duplication, there is a high level of gene redundancy (Howe, 2013).Over evolution, a gene may have diverted from its original function or both paralogues can be sub-functionalized.As such, in zebrafish studies, paralogues can compensate for loss of another gene, with single knockouts displaying wildtype or more subtle phenotypes than expected (Buglo, 2020;Peng, 2019).However, this trait can also be seen as advantageous as it can result in the ability to study cell-specific gene functions.Notably, only a single Rab28 ortholog is identified in zebrafish (Carter, 2020).
In order to use zebrafish to expand knowledge of potential gene therapies for RAB28, a robust phenotype related to Rab28 in zebrafish is desired.OSP is essentially a waste removal/recycling process in the retina, which aims to prevent the accumulation of photo-oxidative compounds and maintain photoreceptor health and viability.It involves engulfment and phagocytosis of demarcated tips of the photoreceptor OS.This integral retinal process is highly regulated and necessary for normal retinal function.For a detailed description of the OSP process see (Mazzoni et al., 2014;Lakkaraju, 2020;Kwon and Freeman, 2020;Moran et al., 2022).The zebrafish retina has 2 OSP peaks; a dawn peak at Zeitgeber Time (ZT) 4, with Zeitgeber time signifying hours after light onset, and a dusk peak at ZT 17, reported to consist of phagocytosed cone and rod photoreceptor OS at each time-point (Moran et al., 2022;Lewis et al., 2018;Moran, 2022).Moran et al. 2022 directly assessed the integrity of OSP in the zebrafish rab28 KO throughout the day (Moran, 2022); assessing samples obtained at various time points corresponding to expected zebrafish RPE phagosome peaks and troughs (Lewis et al., 2018).Analysis by electron microscopy revealed absence of zebrafish Rab28 alters both known peaks of OSP, leading to an approximate reduction of 50% of the number of RPE phagosomes compared to the wildtype control (Fig. 2 a,b).Samples collected at "non-peak" times affirmed that there is no phase-shifting of peaks in the zebrafish rab28 KO, nonetheless, a basal level of RPE phagosomes remained at all time-points analysed in both wildtype siblings and rab28 KO, revealing that Rab28 is required specifically for the burst of RPE phagosomes at dawn and dusk peaks, rather than for regulation of basal levels of OSP (Moran et al., 2022).Zebrafish rab28 KO also display visual cycle defects in the retina, with reduced levels of the visual chromophore 11-cis-retinaldehyde from 15 dpf and an increase in the toxic vitamin A derivative A2PE at the adult stage (Moran, 2022) (Fig. 3).Nevertheless, rab28 mutant zebrafish appear to have no apparent deficits to their visual behaviour up to 21 dpf, a normal electroretinogram (ERG) response at 2 mpf and also display no retinal degeneration up to 18 mpf (Carter, 2020;Moran, 2022).Therefore, the OSP and visual cycle phenotypes observed in the zebrafish rab28 knockout are best suited for evaluation of RAB28 gene therapy approaches.
Notably, zebrafish and mouse Rab28 knockout models have overlapping and distinct phenotypes (Table 2).Knockout of murine Rab28 also disrupts OSP, but this is associated with a subsequent loss of visual function (Ying, 2018).In the Rab28 KO mouse, failure to prune old lamellae led to the accumulation of membranous material at cone tips and eventual degeneration and death of the cones, followed by rods [36].The murine model therefore appears better aligned to develop RAB28 gene therapies, and one has to question if there is a value-added role for the zebrafish model.However, there are >90 RAB28 variants of uncertain significance reported in ClinVar (Landrum, 2018).The costeffectiveness and the scalability of the zebrafish model are better positioned than the mouse model to resolve these variants.In addition, the paucity of cone photoreceptors in rod-dominated, nocturnal rodents is a disadvantage to the cone-rich zebrafish retina (Noel et al., 2022).Indeed, we consider that combining the zebrafish and mouse models is the better approach to optimise RAB28 gene therapy.
A key question in relation to gene therapy is in which cell types does the replacement gene need to be expressed to be most effective.In the mouse retina, Rab28 is expressed in the RPE, and rod and cone photoreceptors (Ying, 2018;Hoang, 2020).However, the precise spatial and temporal expression of rab28 in the zebrafish retina is unknown.To our knowledge, no zebrafish RNA sequencing data has been published which references the cell-type expression of rab28.Published microarray data compared the transcript expression profile of flow-sorted cone photoreceptors to the rest of the retina in adult zebrafish (Glaviano et al., 2016).While rab28 expression is detected in both sample sets, it was significantly elevated (20% higher) in the cone-only sample.This points to a potentially greater requirement of RAB28 in cones than other cells in the retina.Therefore, in zebrafish germline gene therapy experiments, we hypothesised that cone-specific expression of Rab28 would rescue the OSP defect.TEM sections collected from transgenic animals at the dawn OSP peak revealed that germline Rab28 gene replacement to selectively express wildtype (GFP-tagged) zebrafish Rab28 in cone photoreceptors was sufficient to restore the dawn OSP peak in rab28 KOs to wildtype levels, suggesting zebrafish Rab28 plays a more important role in cone photoreceptors compared to other retinal cells (Carter, 2020;Moran, 2022).This important finding indicates that Rab28 in cone photoreceptors plays a cell-autonomous role in the regulation of OSP peaks, with the data suggesting that Rab28 expressed in the cone photoreceptors is necessary to facilitate peaks of OSP (Fig. 2 c).From a therapeutic perspective, rescue of the peak OSP defect by cone-specific expression of Rab28, suggests that gene therapy for RAB28-associated CRD may be successful by RAB28 gene restoration to cones.

Emerging opportunities for human RAB28 translational studies in zebrafish
Opportunities exist in zebrafish for enhanced translational impact by generating humanised zebrafish lines.For example, if the human RAB28 gene is capable of rescuing zebrafish rab28 -/-OSP defects, it suggests that the mechanism by which Rab28 functions is conserved between humans and fish.This would support the use of fish as a good model for RAB28 related diseases.Notably, the human RAB28 protein (isoform two) shows 76% identity to the zebrafish Rab28 protein (Carter, 2020).However, there is a risk that the human RAB28 gene could fail to rescue the zebrafish Rab28 KO phenotype due to divergences in zebrafish and human Rab28 protein conformations or ability to interact with molecular partners.
Another gene-based translational opportunity in zebrafish is in IRD diagnostics, for example to model the impact of RAB28 coding variants found in patients.Genome sequencing is now more routinely available to IRD patients (Dockery et al., 2021); however, interpretation of the variants can be challenging for patients, clinicians and researchers if the pathogenicity of the variant is unknown.Some information related to pathological IRD implications can be predicted by mapping the location of a variant onto known protein domains and applying variant pathogenicity prediction software (Gunning, 2021).However, currently no platform exists to functionally investigate variants of unknown significance (VUS) in RAB28.Zebrafish potentially provide an appropriate vertebrate model for overcoming this challenge.Zebrafish are amenable to large-scale genetic modification due to their ex-utero development, short generation time, and efficient transposon based transgenics systems (Kawakami, 2007).Previous studies show that zebrafish can efficiently express human cDNAs and this method has been used to model patient variants of other genes (Collery et al., 2013;Lysko, 2022;Cuervas-Mons et al., 2019;Zink, 2022).Studies with humanised zebrafish models are summarised in Table 3.One study modelled IRD in fish by creating transgenic lines expressing autosomal dominant human RETGC-1 variants (Collery et al., 2013).A separate study used a mutant version of the human rhodopsin gene to model retinitis pigmentosa (Nakao et al., 2012).Aside from IRD, zebrafish have been used to model variants in the genes encoding human insulin and tau (Lopez, 2017;Cosacak, 2017;Eames et al., 2013).These studies focused on dominant variants, however, genetic complementation studies with human RAB28 transgenes is applicable for investigating recessive variants.One previous study used an F0 approach to screen human NRG1 variants for pathogenicity (Lysko, 2022).In this system, eggs from a cross of nrg +/-heterozygous zebrafish were injected with a Tol2 transgenic construct expressing WT or patient variant forms of NRG1 (Lysko, 2022).Rescue of the KO phenotype was then examined in KO larvae expressing the transgene.This approach is much faster than establishing a transmitting germline but expression of the transgene is mosaic in the F0 compared to uniform in a stable line.The F0 assay can be carried out in several days whereas establishing a stable line would take at least 4 months (Kawakami, 2007).Other diagnostic approaches to assess the pathogenicity of human gene variants in zebrafish used microinjections of mRNA to express WT genes and variants (Eijkenboom, 2019).This system is again swifter than establishing a stable line (several days vs 4 months), but it only allows for expression during early development (Sertori, 2022).Apart from transgenic methods, CRISPR knock-in technologies could introduce patient variant mutations to regions of a zebrafish gene conserved with humans (Ranawakage, 2020) placing the variant under the control of the endogenous promoter.The efficiency of CRISPR knock-ins in zebrafish is controversial, but one study reported between 17 and 29% of F0 fish transmitted a knock-in allele to their progeny (Ranawakage, 2020).However, this is not as high as the established Tol2 transgenic system of which 70% of F0 fish are reported to transmit the transgene to their offspring (Kawakami, 2007).While the reported efficiency of CRISPR knock-ins are high enough to generate stable lines they may not be at the level sufficient to carry out F0 knock-in screens of patient variants.Interestingly, we previously observed that transgenic expression of zebrafish Rab28 containing a T26N variant was capable of rescuing the OSP phenotype observed in the zebrafish rab28 KO background (Moran, 2022).There were lower levels of OSP phagosomes in the GDP-locked (T26N) transgenics compared to lines expressing the wildtype zebrafish rab28 transgene, however this difference was not significant statistically at the age analysed.Notably, T26N RAB28 is predicted by in-silico analysis to be a loss-of-function pathogenic variant (Iarossi, 2020) We have postulated that the T26N zebrafish transgene may be GDP-preferring in lieu of being GDP-locking and therefore may be hypomorphic (Moran, 2022); Alternatively, this variant may not be pathogenic.Thus, in order to focus on the patient RAB28 variants we consider it germane to focus on humanised zebrafish lines.
The process of developing a humanised zebrafish model to screen the pathogenicity of patient variants in RAB28 is technically feasible and applicable to other IRD genes.The most established system for generating transgenic zebrafish, the Tol2 transgenesis system works on the principle of injecting a one-cell stage zebrafish egg with mRNA encoding a transposase and a plasmid encoding a transgene (Kawakami, 2007).This system is closely allied with the multisite Gateway cloning system which gives access to a large array of promoters and protein tags (Kwan, 2007).To create a Tol2 expression vector, step one is to obtain cDNA of the gene of interest in a Gateway entry vector.Once the human gene is in the correct entry vector it can be combined with specific promoters and/ or tags and assembled into the final expression vector (Kwan, 2007).Our preliminary data suggest that zebrafish can express GFP-tagged human RAB28 protein in cones (Fig. 4).It is as yet unknown as to whether the human RAB28 functions equivocally to zebrafish Rab28.If complementation is observed, then patient variants can be generated from the expression vector via site directed mutagenesis (Carter, 2020).For recessive genes like RAB28 the human transgene is introduced into knockout zebrafish.Advancements in F0 knockout technology allows knockout zebrafish to be created directly from injected eggs instead of establishing a stable line (Kroll, 2021).F0 rescue experiments have also been described using the Tol2 system (Lysko, 2022).An intriguing prospect would be the ability to combine the F0 CRISPR KO strategy with Tol2 expression of a human transgene to allow rapid screening of patient variants from recessive genes.This could be particularly useful for genes that present an impaired survival phenotype when knocked out but may present a less severe phenotype when expressing a human missense variant.
A more challenging opportunity is to apply zebrafish for pre-clinical studies of gene therapy targeted to somatic cells of the eye, so as to mimic the canonical gene therapies for IRD in humans.To the authors knowledge, however, no viral/non-viral vector gene therapy approach has been used in a zebrafish model of genetic disease.Several viral and non-viral vectors exist that can efficiently transduce zebrafish cells.Herpes simplex type 1 virus has been shown to efficiently transduce cells in the adult zebrafish brain following direct injection into brain tissue (Zou et al., 2014).Adenovirus type 5 transduces zebrafish brain cells when injected into larvae and this expression lasts for at least 30 days post injection (Gulías, 2019).These viral vectors were used in human gene therapy but not for gene delivery to the retina/retinal pigment epithelium where adeno-associated viruses (AAV) are preferred (Ong et al., 2019).Non-viral nanoparticle based vectors can deliver genetic material to zebrafish cells.Putrescine-sphingomyelin nanoparticles delivered microRNAs and plasmid DNA to zebrafish embryos (Gulías, 2019).In a separate study P-P-R 9 nanoparticles delivered siRNAs to the adult zebrafish heart following injection into the thoracic cavity (Wang, 2019).As these vectors function in zebrafish, it is likely that somatic gene replacement therapies will be trialled in zebrafish.This methodology has an advantage compared to transgenic methods for proof of concept gene replacement therapy as it shows that adult somatic cells can express the gene of interest and are still functional in the process of interest.

Conclusion
In summary, there is an unmet need to develop therapies for CORD18, associated with RAB28.Gene therapy is a logical approach for this inherited condition.On initial consideration, zebrafish appear to have limited use for RAB28 gene therapy development compared to the mouse model which displays retinal degeneration.However, the defects in OSP and visual cycle observed in zebrafish Rab28 knockouts provide an opportunity to functionally resolve, at scale, in a vertebrate conedominated eye, the pathogenicity of the many reported RAB28 variants of uncertain significance. Footnote.
We mention multiple Rab28 conventions throughout the manuscript.Where possible we strive to use the species specific term, however when referring to the protein in relation to multiple species we use the term Rab28.
Footnote Table 1

Fig. 1 .
Fig. 1.Comparison of RAB28 across species.(a) Multiple sequence alignment of human RAB28 isoform two compared with Rab28 from different model organisms.Alignment was generated using ClustalW.Colours represent percentage identity and were annotated using Jalview.(b) Phylogenetic tree of Rab28 across different model organisms.This phylogenetic tree was constructed from the alignment by Jalview using the neighbour joining method with the Blosum62 substitution matrix.

Fig. 2 .
Fig. 2. (a) Schematic of wildtype zebrafish showing regular OSP with function rab28.(b) Schematic of zebrafish rab28 KO with OSP defects which can be rescued to wildtype levels upon reintroduction of Rab28 solely to the cone photoreceptors (c) as shown in Moran et al. 2022.Created using biorender.com.

Fig. 3 .
Fig. 3. Working Model of the effect of rab28 KO on the zebrafish retina.From the published results, the hypothesis stands that A2PE is elevated within the rab28 KO photoreceptor OS due to the defective OSP phenotype and lower levels of Abca4b in rab28 KO.Accumulation and formation of A2PE may contribute to photoreceptor cell dysfunction.(Kim et al., 2006) Figure created using BioRender.com.

Fig. 4 .
Fig. 4. (a) Schematic representation of creation of humanised zebrafish lines.(b) 5 dpf larval zebrafish treated with PTU.Left, rab28 -/-non transgenic larvae.Right, rab28 -/-larvae transgenically expressing GFP tagged human RAB28.Arrows show the left eye of each larva.(c) Expression vectors used to create RAB28 humanised zebrafish lines.Multiple sequence alignments showing identical amino acids in all model organisms for site 23 and 217.S23F and C217W mutations mapped to RAB28 isoform2 protein domains.

Table 2
Comparison of various phenotypes observed in rab28 KO zebrafish and RAB28 KO mouse models.
A.L.Moran et al.

Table 3
Previous studies using humanised zebrafish lines to model pathogenicity of patient missense variants.
detailing species nomenclature.