Rodent Models of Retinal Degeneration: From Purified Cells in Culture to Living Animals

Increased Intraocular Pressure

Increase of intraocular pressure (IOP) is a main risk factor in the development of glaucoma, and animal models of glaucoma can be created by increasing the IOP in the eye through different strategies. Thus, the disease was modeled by transiently increasing the intraocular pressure for a limited period (1 h) in the anterior chamber (Mohand-Said et al. 1997). In addition, increase of IOP can be induced by cauterization of episcleral veins (Froger et al. 2012). A similar result was obtained by laser photocoagulation of episcleral and limbal veins (Salinas-Navarro et al. 2010). Injection of polystyrene or magnetic beads in the anterior chamber was also used to block the mouse trabecular meshwork, thereby raising the IOP (Cone et al. 2010; Mao et al. 2013). In clinical practice, many different topical treatments that lower IOP have proved efficient in attenuating disease progression toward blindness. However, unfortunately, even when controlling the pressure, some patients still experience retinal ganglion cell loss and additional effective treatments for this highly prevalent disease are sorely needed.

Light Damage

Light damage models are widely used to investigate intracellular mechanisms that lead to photoreceptor cell death, a common feature in retinitis pigmentosa and age-related macular degeneration. Photochemical damage to the retina occurs after short or prolonged exposures of mild or intense light in mice and rats. The “white light damage” model targets photoreceptors, whereas the “blue light damage” model (excitation peak 440 nm) induces primarily RPE cell photosensitization. These models have been shown to have a major impact on photoreceptor and RPE function, inducing photochemical damage and leading to cell death (Wu et al. 2005). It is also possible to couple transgenic models sharing cellular features of AMD with light damage. Thus, transgenic Abca4^−/− or Abca4^−/−Rdh8^−/− mice that accumulate RPE lipofuscin at increased levels are more susceptible to retinal light damage than wild-type mice (Maeda et al. 2008, 2009; Wu et al. 2014; Fontaine et al. 2016).

Laser-Induced Subretinal Inflammation

Laser treatments were applied to generate choroidal neovascularization (CNV) as a model of the neovascular (“wet”) form of age-related macular degeneration (Combadière et al. 2007). This model allowed analysis of different treatment modalities including anti-VEGF treatments and studying the role of inflammation in CNV formation. It involves the induction of CNV using laser photocoagulation that penetrates the Bruchs membrane, which mimics the pathological processes seen in diseases such as age-related macular degeneration (AMD) (Grossniklaus et al. 2010). Subretinal inflammation is induced in pigmented mice with a laser mounted on a slit lamp. The injury induces a neutrophil and macrophage (macrophage/microglial cell) infiltration that can be quantified using Ly-6G and IBA-1 immunostainings on retinal and choroidal flat mounts or by FACS analysis at different time points after laser injury. Ly-6G⁺ neutrophils are significantly recruited to the injury site, reaching their maximum at 10 h and being cleared at day 3. IBA-1⁺ macrophages are recruited at day 3, reach a maximum at day 4, and diminish thereafter. Agonists and antagonists can be administered at different time points after laser injury by intravitreal injection (Camelo et al. 2012; Levy et al. 2015) or systemic treatment (Lavalette et al. 2011). Even sizeable molecules (such as antibodies) penetrate the post–laser injury retina and can affect subretinal inflammation. Choroidal neovascularization is evaluated by angiography or by OCT in vivo and by immunostained flat mounts postmortem. Although the trigger to induce neovascularization is very different from the human disease, mechanisms of downstream subretinal inflammation and neo-angiogenesis are likely similar. For instance, aflibercept, which blocks murine as well as human VEGF, inhibits CNV in this model similar to its effect in humans (Ma et al. 2016).

Oxygen-Induced Retinopathy

Ischemic retinopathies, including retinopathy of prematurity (ROP), vein occlusion, and diabetic retinopathy (DR) are all characterized by primary retinal ischemia. If left untreated, they can progress to neuronal dysfunction and proliferative forms of retinopathies causing rapid and irreversible vision loss. Unlike humans, adult rodents do not develop a proliferative response to experimental ischemic situations even following laser-induced vein occlusion or diabetes induction. To address the mechanisms of proliferative retinopathies, Smith et al. described a model of oxygen-induced retinopathy in newborn mice (Smith et al. 1994). The treatment relies on exposing neonatal mice to alternating periods of high and low oxygen levels. This results in retinal vascular abnormalities resembling those observed in human retinopathy of prematurity and diabetic retinopathy. As a consequence, it has since become a widely used model for studying the pathophysiology and potential treatments of ischemic retinopathies.

In addition to the vascular abnormalities observed in human retinopathy of prematurity, persistent rod photoreceptor dysfunction is observed in adulthood despite the clinical resolution of vascular abnormalities (Fulton et al. 2009). A similar observation is made in the mouse and rat oxygen-induced retinopathy models (Liu et al. 2005; Vessey et al. 2011). This long-term loss of rod photoreceptor function, while cone function remains intact, has been attributed to differential metabolic adaptation of rods and cones to ischemic stress (Fulton et al. 2009). Overall, while the oxygen-induced retinopathy model is widely used to study pathological retinal neovascularization, it also allows for the investigation of rod photoreceptor responses to acute ischemia as well as long-term modifications of rod photoreceptor activity.

Pharmacological Models

Different pharmacological treatments were shown to induce retinal degeneration. For instance, treatments limiting taurine transport by introducing β-alanine or guanidinoethane sulfonate in the diet were first reported to induce photoreceptor degeneration (Pasantes-Morales et al. 1983). More recently, similar treatments were shown to preferentially affect cone photoreceptors and retinal ganglion cells (Gaucher et al. 2012; García-Ayuso et al. 2019). This more precise description of the damaging effect of taurine deficiency resulted from the investigation of the retinal toxicity of vigabatrin, an antiepileptic drug still prescribed in infantile spams, which was shown to induce taurine deficiency in rodents and patients (Jammoul et al. 2009). It was first reported to generate a destruction of the outer nuclear layer in albino animals (Butler et al. 1987), and we showed that the earlier cone and retinal ganglion cell degeneration is related to taurine deficiency associated with phototoxicity (Duboc et al. 2004; Jammoul et al. 2009, 2010).

In glaucoma, the loss of retinal ganglion cells was often attributed to glutamate excitotoxicity (Lipton 2003). Supporting this hypothesis, memantine, an N-methyl-D-aspartate (NMDA) receptor antagonist, was found to be neuroprotective in different animal models of glaucoma (WoldeMussie et al. 2002). As a consequence, glaucomatous conditions are often modeled by intraocular injection of a glutamate NMDA receptor agonist (Torero Ibad et al. 2011). However, memantine was not found to be efficient in the treatment of glaucomatous patients (Danesh-Meyer and Levin 2009).

Pharmacological treatments with streptozotocin or alloxan were also applied to induce diabetes in rodents (Barber et al. 1998; Park et al. 2003; Martin et al. 2004; Gaucher et al. 2007). These treatments allowed to demonstrate neuronal apoptosis or a microglial reaction in the treated animals. The neurodegenerative process in these animal models was very variable from photoreceptors to retinal ganglion cells. These models also showed some alterations of the vascular system occurring with time (Song et al. 2004), but in general they do not develop diabetic retinopathy similar to that which can be observed in human diabetic patients.

Naturally Occurring Models

Many mice display spontaneous phenotypes of retinal degeneration in the photoreceptor layer, and they were named numerically in the order that they were identified: rd1, rd2, rd10, etc. (Chang et al. 2002). The rd1 mouse (Pde6b^rd1, also known as the rd mouse) was described in 1924 by Keeler as “leading to absence of the visual cells (rods), the external nuclear layer, and the external molecular layer” (Keeler 1924). Subsequently, it was found that the phenotype results from a spontaneous recessive mutation in the β-subunit of the rod-phosphodiesterase (PDE) gene Pde6b^rd1 (Bowes et al. 1990). The model is associated with massive death of rods in the first weeks of postnatal life, followed by the death of cones, thus mimicking an early-onset severe retinal degeneration as in patients affected by retinitis pigmentosa, who can in some cases have a mutation in the homolog human gene (Farber 1995). rd10 mice carry a different recessive mutation in the same rod-PDE gene, leading to a similar but somewhat milder degeneration profile with slower rod loss (between P20 and P25) and slower secondary cone degeneration (Chang et al. 2002). The rd1 mouse was used in the laboratory of Jean Bennett to demonstrate the therapeutic potential of subretinal injection of a recombinant replication-defective adenovirus that contained the murine eDNA for wild-type β PDE, Ad.CMV β PDE (Bennett et al. 1996).

Other naturally occurring mouse models of retinal degeneration include the rds (retinal degeneration slow) mouse, because of a mutation in the Prph gene (rd2). The retinal phenotype of Prph2 Rd2/Rd2 mouse consists of complete failure to develop photoreceptor outer segments and apoptotic loss of photoreceptor cells. This model allowed for demonstration of successful reversal of the retinal degeneration by gene augmentation therapy using a recombinant adeno-associated virus (AAV) vector encoding a Prph2 transgene, showing stable generation of normal outer segments (Ali et al. 2000). The RD3 protein was reported to negatively modulate the activity of guanylate cyclase (Chen et al. 2022). The rd5 or tubby mouse (tub) displays not only retinal degeneration but also maturity-onset obesity, insulin resistance, and hearing deficits (Noben-Trauth et al. 1996). The rd6 phenotype is due to a mutation in the membrane-type frizzled-related protein (MFRP), which is normally expressed in the retinal pigment epithelium and ciliary body (Fogerty and Besharse 2011). The retinal degeneration 7 (rd7) is related to a mutation in the transcription factor Nr2e3, the mutation of which results in the production of photoreceptors with a hybrid rod and cone cell type responsible for enhanced S-cone syndrome phenotype (Corbo and Cepko 2005). The rd8 mutation in the CrB1 gene is often discovered in different lines of C57BL/6N or embryonic stem cells such that it can generate confusing retinal phenotypes in transgenic animals and should be actively screened for (Mattapallil et al. 2012). The rd9 mouse carries a base-pair duplication in the purine reach domain (ORF) of the retinitis pigmentosa GTPase regulator (RPGR) thereby providing an appropriate model of X-linked retinitis pigmentosa (Thompson et al. 2012). The rd11 model is caused by mutation in the lysophosphatidylcholine acyltransferase-1 (LPCAT1) gene coding for a phospholipid biosynthesis/remodeling enzyme (Friedman et al. 2010). The rd12 mouse, which represent a model for Leber congenital amaurosis (Redmond and Hamel 2000), was used to demonstrate efficacy of gene therapy when reintroducing the correct gene sequence of RPE65 (Pang et al. 2005). Similarly, the rd16 mouse, which represents a model for CEP290-elicited ciliogenesis defect, was used to assess a possible therapeutic intervention (Mookherjee et al. 2018).

Apart from these mouse models, a very common animal model in many research laboratories is the Royal College of Surgeon (RCS) rat, which shows an accumulation of debris below the retina. The retinal degeneration in this model is caused by a recessive mutation in the Mertk gene (D'Cruz et al. 2000; Nandrot et al. 2000), which is expressed in RPE cells, and plays an important role in phagocytosis of photoreceptor outer segments as part of their natural turnover (Vollrath et al. 2001). The failure of phagocytosis leads to accumulation of debris under the retina, leading to secondary degeneration of the photoreceptors. As such, it is often considered as a model of age-related macular degeneration, because photoreceptor loss follows upon primary dysfunction of the RPE. However, it is actually an inherited retinal disorder, and, in humans, mutations in the Mertk gene cause an early and severe form of retinitis pigmentosa or Leber congenital amaurosis. Gene augmentation therapy was shown to attenuate the course of disease in the RCS rat (Vollrath et al. 2001). Later, washout of the debris by subretinal surgery was also shown to prolong the survival of photoreceptors in this model (Lorach et al. 2018).

Finally, the DBA2/2J mouse serves as a common model for glaucoma because pigment dispersion in the anterior chamber and trabecular meshwork induces a major increase in intraocular pressure. Despite the high variability of this model, it was used to demonstrate efficacy of different molecules for the prevention of glaucoma such as memantine, timolol, or taurine (Schuettauf et al. 2002; Froger et al. 2012).

Induced Genetic Models

To increase the frequency of mutagenesis in mouse colonies, N-ethyl-N-nitrosourea-induced (ENU) mutagenesis was applied and then followed by screening for mouse models of diseases. This approach was applied with great success to generate animals with clear eye phenotypes including retinal degeneration (Thaung et al. 2002). However, recent advances in molecular genetic techniques now allow to manipulate the genome with high precision and provide the ability to generate specific models of retinal diseases, especially models of rare diseases. These include gene knockouts but also knockin models in which specific mutations can be introduced into specific genes. The number of these transgenic animal models has increased exponentially in recent years, and therefore they cannot be fully reviewed here. A major example is the P23H rat and mouse, which reproduce a highly prevalent human mutation in the rhodopsin gene (Machida et al. 2000; Chrysostomou et al. 2009; Kolomiets et al. 2010; Orhan et al. 2015). This animal model was used to assess different therapeutic strategies such as the effect of the rod-derived cone viability factor RdCVF on cone survival (Yang et al. 2009). Similarly, the Abca4-KO mouse model has been largely utilized to understand the etiology of Stargardt disease or age-related macular degeneration (Weng et al. 1999). For the large ABCA4 gene, which exceeds the loading capacity of AAV vectors, a lipid nanoparticle-based approach has demonstrated effectiveness for in vivo delivery in Abca4^−/− mice (Sun et al. 2020, 2022). Development of therapies for Usher syndrome type 1 (USH1)—a major cause of combined inherited deafness–blindness in humans—has been hampered because mouse models only develop deafness but no retinal degeneration, as is the case in both harmonin (Ush1c^−/−) or sans (Ush1g^−/−) knockout mice (Trouillet et al. 2018). In fact, cone degeneration was observed only when transferring the transgene in an albino background (Trouillet et al. 2018), thereby suggesting a role for light and oxidative stress in the degenerative process of Usher retinal disease. This result was consistent with a previous study showing vulnerability of photoreceptors in the shaker 1 mouse (Ush1b mutant) to moderate light exposure (Peng et al. 2011).

The etiology of age-related macular degeneration is multifactorial and associated with various genetic and environmental factors. The need to develop effective treatments for AMD has led to the generation of multiple rodent models by expressing genetic mutations commonly associated with the disease (Pennesi et al. 2012; Ratnapriya and Chew 2013; Fletcher et al. 2014; Veleri et al. 2015). Although there is no model that fully recapitulates all of the features of human AMD (particularly in view of the fact that rodents and indeed most species except primates and certain types of birds do not possess a macular structure), these models have helped in the discovery of pathophysiological mechanisms underlying AMD, such as inflammation and immune dysregulation, chronic oxidative damage, and alterations in lipid and carbohydrate metabolism, among others (Calippe et al. 2017; Beguier et al. 2020). Polymorphisms in the gene for complement factor H (CFH) B, C2, C3, and C9 have been associated with either conferring protection or susceptibility to the development and progression of AMD. The Cfh^−/− mouse was used to study some aspects of the complement factor pathway in the pathogenesis of AMD, but one should bear in mind that the genetic variants associated with AMD do not lead to down-regulation of CFH. These mice also display systemic consumption of C3, deposition of C3a in the glomerular basement membrane and glomerulonephritis, contrary to the increase of C3 and C3a observed in AMD. In the retina, they develop mild loss of photoreceptors, reduced rod function, and increased deposition of complement in the photoreceptor layer (Coffey et al. 2007). Decreased visual acuity and reduction in rod-driven ERG a- and b-wave responses were found at two years of age, together with thinning of the Bruch's membrane (Coffey et al. 2007), which is not typical of AMD. Transgenic CFH Y402H mice (Cfh^−/− mice expressing the human AMD-related polymorphism CFH 402H under control of the human ApoE promoter) show a large number of drusen-like deposits at the age of 1 yr, increased number of microglial cells and macrophages in the subretinal space, thickening of Bruch's membrane, and basement membrane deposition of C3d as compared to wild-type mice (Ufret-Vincenty et al. 2010). However, they did not show photoreceptor atrophy even under a high-fat diet but they showed clear differences from mice expressing the “wild-type” CFH 402Y protein (Landowski et al. 2019). Transgenic mice overexpressing C3 show several features of AMD such as proliferation and migration of endothelial cells within the retina, disruption of the RPE with migration of pigmented cells into the retina, complement deposition, and atrophy of the photoreceptor outer segments. However, the C3-overexpressing animals demonstrate increased incidence of retinal detachments that is not consistent with AMD (reviewed in Pennesi et al. 2012). These cfh mouse models of AMD developed by Bowes Rickman and collaborators have enabled the unveiling of novel mechanisms contributing to AMD pathophysiology, including correlation between CFH levels and retinal integrity (Ding et al. 2015) and a link between the complement system and lipid pathways (Ding et al. 2015; Toomey et al. 2015; Landowski et al. 2019; Landowski and Bowes Rickman 2022). Importantly, they serve as tools for validation of potential therapeutic targets in AMD (for review, see Kelly et al. 2020; Landowski and Bowes Rickman 2022).

Our research in the field of AMD showed that under physiological conditions, the subretinal space is immunosuppressive and devoid of mononuclear phagocytes, a family of cells that includes monocytes and infiltrating and resident macrophages (reviewed in Guillonneau et al. 2017). A common feature of early and advanced dry and neovascular forms of AMD is the activation of macrophages in the inner retina and the chronic accumulation of subretinal mononuclear phagocytes. We showed that the homeostatic elimination of mononuclear phagocytes that infiltrate the photoreceptor cell layer depends on thrombospondin 1 (TSP1)-mediated activation of their CD47 receptor. Accordingly, Thbs1^−/− and Cd47^−/− mice develop age-related subretinal mononuclear phagocyte accumulation similar to AMD patients. Importantly, we demonstrated that both major genetic AMD risk factors, the CFH H402 variant and 10q26 risk haplotype, inhibit TSP1-mediated CD47 activation and MP elimination, promoting chronic pathogenic inflammation (Calippe et al. 2017; Beguier et al. 2020). Our findings described a comprehensive mechanism of how these two prominent AMD-risk haplotypes, the variant of complement factor H and a minor haplotype of the chromosome 10q26, account for most of the genetic risk for AMD and promote AMD pathogenesis. This work provides a rationale for therapeutic restoration of immune privilege in the subretinal space to induce the resolution of subretinal chronic inflammation that promotes photoreceptor destruction and blindness in AMD. In agreement with these results, several mouse models targeting inflammatory mediators and apolipoproteins have been generated and reported to partially replicate AMD features, including subretinal drusen-like accumulations, thickening of Bruch's membrane, an increase in autofluorescence and lipofuscin granules, and photoreceptor degeneration (Combadière et al. 2007). However, the Rd8 mutation of the Crb1 gene present in many mouse strains appears sufficient to generate most of these ocular phenotypes (Mattapallil et al. 2012). Unless explicitly excluding the rd8 mutation, it is therefore impossible to conclude which genetic alteration truly provokes AMD-like features. Alternatively, different models of oxidative damage leading to inflammation have been developed based on the particular susceptibility of the retina to oxidative injury associated with its high metabolic demand, high concentration of oxidizable polyunsaturated fatty acids, and presence of photosensitive molecules (such as rhodopsin or lipofuscin). Among them, ceruloplasmin/hephaestin^−/− mice, Sod1^−/− mice, Sod2^−/− and Sod2 knockdown mice, and OXYS rat are commonly used models, allowing us to evaluate the role of oxidative stress in the pathogenesis of AMD (Fletcher et al. 2014). Exposure to cigarette smoke, hydroquinone, high fat diet, and blue light also produced many AMD-like features (see light damage model above).

Myopia can also have a genetic origin (Tedja et al. 2018) such that many mutated mouse models were found to develop or to have increased susceptibility to develop myopia under application of different optical devices (Zeitz et al. 2023). For instance, mutations causing complete congenital stationary night blindness (cCSNB) in patients and animal models were often found to induce myopia (Zeitz et al. 2023). Thus, a substantial increase of the myopic shift was reported in Gpr179^−/− mice compared to wild-type littermates following a 3-wk-long lens-induced myopia protocol (Wilmet et al. 2022). The adult Gpr179^−/− mice have a significant decrease in both retinal dopamine and 3,4-dihydroxyphenylacetic acid, in agreement with previous observations that an alteration of the dopaminergic system correlated with an increased susceptibility to lens-induced myopia (Schwahn et al. 2000; Feldkaemper and Schaeffel 2013; Landis et al. 2020).

Cre-lox Gene Targeting

The Cre-lox system is a genetic tool permitting efficient and accurate control over the location and timing of gene expression and generation of mouse strains in which a transgene is either inducible or expressed only in certain tissues (Sauer and Henderson 1990; Lakso et al. 1992). Pioneering work by Botond Roska's group described approximately 100 mouse lines with either defined strata or, more specifically, retinal cell types marked with GFP allowing stratification-based screening for characterizing neuronal circuitry (Siegert et al. 2009). This strategy not only allowed us to sort specific cell types for genetic analysis but also to optically target cells for recording and functional characterization. For the anatomical analysis of the circuit, the approach can be combined with the Brainbow technology (Livet et al. 2007). This transgenic system allows for stochastic expression of multiple fluorescent protein genes through Cre-lox recombination (Livet et al. 2007; Dumas et al. 2015, 2022).

This strategy to selectively ablate gene expression in a specific cell type was applied to demonstrate expression of the nucleoredoxin-like-1 (Nxnl1) gene in cone photoreceptors (Mei et al. 2016). Indeed, when using a transgenic line expressing Cre recombinase under the control of a cone opsin promoter, cones of these mice were dysfunctional and degenerated by 8 mo of age. Similarly, silencing the expression of RdCVFL in cone-enriched culture resulted in reduced cell viability. These experiments indicated that the Nxnl1 gene protects cones by two distinct pathways, by RdCVFL produced by cones themselves and by RdCVF released from rods (Léveillard et al. 2004; Aït-Ali et al. 2015). These results have provided the rationale for a future therapy combining both RdCVF and RdCVFL expression by gene therapy.