5.2.8. Category 08: Signaling

Molecules such as growth factors/cytokines, hormones, neurotransmitters, and extracellular matrix proteins, or alternatively, mechanical stimuli, are examples of signals used to communicate environmental changes to the cell. Surface or intracellular (e.g., nuclear) receptors recognize the signals and e ffect changes within the cell, often setting in motion amplifying transduction cascades that mediate responses such as activation or inhibition of protein activity or migration to di fferent cellular localizations. Further, signals can also be transmitted from the cell to other cells, for example, through neurotransmitters. Since intra- and intercellular communication is crucial for the proper development or function of cells, it is not surprising that a large number of mutations in cellular signaling lead to defects in retinal development, which in turn a ffects PR survival. For example, vascular development is a ffected in mutants bearing disruptions in *Fzd5, Lrp5, Ndp,* and *Tspan12*—all components of the Wnt signaling pathway. Integral membrane frizzled receptors, of which they are 10, together with coreceptors, LRP5 and LRP6, mediate canonical Wnt signaling [355]. Thus, conditional *Fzd5* null mutants develop microphthalmia, coloboma and persistent fetal vasculature, and late-onset progressive RD [356] and *Lrp5* mutants exhibit similar vascular and retinal phenotypes [357]. Mice that are null for NDP, a ligand for FZD4, exhibit delayed retinal vasculature development, retrolental masses, disorganization of the ganglion cell layer, and occasionally focal areas of ONL absence at later stages of the disease [358]. TSPAN12, mediates NDP-FZD4-LRP5 signaling in the retinal vasculature, where it localizes, and a mutation leads to vascular defects that phenocopies disruptions in *Ndp*, *Fzd4,* and *Lrp5*, and at 3 months exhibits a 50% loss of PRs [359]. In all of these models, it is likely that the loss of PRs is caused by the aberrant retinal vasculature having secondary e ffects on PRs. A review by Hackam suggests that Wnt signaling may a ffect the apoptotic pathway and neurotrophin release, dysregulation of which may a ffect PR survival [360]. MFRP, which bears a CRD domain shared by all frizzled proteins, also leads to PR degeneration when disrupted [361,362], as does the human knock-in allele, p.S163R, of its bicistronic partner, CTRP5 [363]. The exact role or function of either protein has ye<sup>t</sup> to be fully elucidated.

Like the frizzled-associated proteins whose pathological e ffects on PRs are likely to be mediated through an aberrant retinal vasculature, other signaling molecules, *Ptpn11* and *Fyn*, appear to mediate their e ffects on PRs through another cell type as well, in this case, Müller glia cells, and PRKQ through the RPE. A *Six3-cre* mediated conditional knockout of *Ptnp11* [364] leads to altered ERK and MAPK signaling in Müller glia and alteration in their adhesive capabilities. FYN, a Src-kinase membrane associated tyrosine kinase, localizes to Müller glia cells, and FYN deficiency leads to altered adhesion properties of Müller cells and retinal dysmorphology [365]. PRKCQ, a serine threonine protein kinase, which localizes to the lateral surface of the RPE cells, causes a reduction in adhesion between the apical processes of the RPE and OSs when it is disrupted. The reduction in adhesion may be responsible for the retinal detachment and subsequent PR loss observed in this model [366].

The family of PI3Ks or phosphoinositide 3-kinases, made up of catalytic and regulatory subunits, function to phosphorylate the inositol ring of phosphatidylinositol and thereby regulate growth, proliferation, di fferentiation, motility, survival, and intracellular tra fficking. For example, it mediates insulin-stimulated increase in glucose uptake and glycogen synthesis and responds to signals such as FGFRs and PDGFRs. Conditional knockouts of *Pik3cb*, encoding a catalytic subunit [367] and *Pik3r1*, encoding a regulatory subunit [368], using the cone-specific CRE, Tg(OPN1LW-cre)4Yzl, lead to progressive cone PR loss. IRS2, necessary for the integration of signals from insulin and IGF1 receptors, causes an early-onset, moderately paced PR loss [369]. Additionally, a targeted conditional allele of PDGFRB developed diabetic retinopathy like features with angiogenesis, proliferative DR-like lesions, pericyte drop out, and eventual PR loss [370]. Disruption of MAP3K1, a serine/threonine kinase, which participates in the ERK, JNK, and NF-κB signaling pathways, leads to retinal laminar and vascular defects, aberrant RPE, and PR cell death [371]. SEMA4A, a transmembrane protein, also causes PR loss, most probably through its e ffects on endosomal sorting [372].

### *5.3. Category 09: Transcription Factors*

In mice, cone and rod PRs are born and develop between approximately E12 and P0, and approximately E13.5 and P7, respectively, from the same multipotent retinal progenitor cell (RPC) pool [373]. PR development, orchestrated by a network of transcription factors, is divided into five phases: proliferation of multipotent RPCs, restriction of RPC competence, cell fate specification, expression of genes important for PR function, and finally, PR structural maturation [374,375]. RB1 and E2F1 function by controlling the G1 to S phase transition in the cell cycle; RB1 plays an inhibitory role until activated by phosphorylation, balancing cell proliferation and cell fate specification [376]. OTX2 is critical for fate determination, while CRX is necessary for terminal PR di fferentiation and acts at di fferent steps in PR development. Transcriptional factors important for rod PR subtype specification include RORβ, NRL, and NR2E3, and for generation of the cone subtypes, TRβ2 and RXRγ [374,375]. Further, transcription factors regulate the expression of other transcription factors in the network (e.g., CRX interacts with *Nrl*, *Rorb,* and *Mef2d*, to name a few, to mediate rod di fferentiation, cone di fferentiation, and proteins necessary for the maturation of the PR, respectively).

The importance of transcription factors in retinal development has been explored in many studies resulting in a number of mouse models with di fferent disease phenotypes (MGI JAX). In many cases, disruption of transcription factors, especially those a ffecting earlier phases of PR development lead to a reduction in the total number of retinal cells generated. We have only included within this category those disrupted transcription factors that eventually lead to PR degeneration. Interestingly, the onset of degeneration of the PR transcription factor models is highly variable—14 days to 2 months of age—and appears to be dependent upon the method used to generate the model and possibly background strain, as variation of severity and onset di ffers among di fferent models of the same gene. Interestingly, the *Crx* models provide a series that recapitulate the clinical diagnoses of autosomal dominant cone-rod dystrophy, Leber congenital amaurosis, and late-onset dominant retinitis pigmentosa. *CrxRip* heterozygotes showed 34% degeneration at five weeks, compared to mice homozygous for the mutation, which reached 55% degeneration at the same age [377]. *Crxtm1.1Smgc* was also noted to have a heterozygous disease presentation more similar to a cone-rod dystrophy, while the homozygous mutant presented with a disease phenotype similar to Leber congenital amaurosis with 70% loss of PRs at one month of age [378].

Other transcriptional factors necessary for the proper maturation of the PRs, such as, MEF2D, shown to be important in regulating transcription of OS and synaptic proteins [379], or NRF1 [380], important in mitochondrial biogenesis also develop PR degeneration when disrupted. Finally, there are transcriptional factors that are important in the development or function of supporting cells such as

ONECUT1 for horizontal cells [381] and MITF for RPE and/or choroidal melanocytes [382], which affect PR survival when disrupted.

### *5.4. Category 10: DNA Repair, RNA Biogenesis, and Protein Modification*

Among the many disrupted genes that lead to PR degeneration, several instances have been documented in genes necessary for producing fully functional proteins, from transcription through post-translational modification. Defects in these genes are likely to impact the function of many other genes that they act upon, and hence, have a greater effect. Since they play a central and basic role, when disrupted they often lead to prenatal lethality in mice, and the adult phenotype is unknown unless a conditional knockout or hypomorphic allele is generated. For example, disruption of DNA repair genes such as *Ercc1*, RNA splicing genes such as *Prpf3, Prpf6, Prpf8, Prpf31*, and *Bnc2*, and miRNA processing genes, *Dicer1* and *Dgcr8* are prenatal lethal in a homozygous state [34,383]. In contrast, homozygous null alleles of *Bmi1* [384] and *Msi1* [385], both involved in repression of regulatory genes in embryonic development, are viable, suggesting potential compensatory mechanisms for the functional loss of these genes. Thus, germline, conditional or hypomorphic models were considered in this category.

Review of genes in this category suggested that a DNA damage response network to ensure transcription in the face of DNA lesions might be required for PR cell maintenance. DNA lesions, such as pyrimidine dimers, interstrand crosslinks, or double-strand breaks (DSBs), are induced by many mechanisms that include UV radiation or free radicals. Repair of such damage is essential for DNA replication and, of particular importance for long-lived post-mitotic neuronal cells, transcription [386–388]. Proteins encoded by *Bmi1, Dgcr8, Dicer1, Elp1, Ercc1, Ercc6, Msi1, Sirt6, Top2b, Ubb,* and *Uchl3* are known to participate in the DNA damage response [387,389–398], some in transcription-coupled DNA repair. For example, BMI1 represses transcription at sites of UV-induced DNA damage to allow repair [389]; ELP1 is a required component of the Elongator complex [399], which couples RNA polymerase II to an alkyladenine glycosylase that initiates base excision repair [392]; ERCC6 promotes DSB repair in actively transcribed regions by displacing RNA polymerase from the lesion site [387], and DGCR8 interacts with both RNA polymerase II and ERCC6 to mediate transcription-coupled nucleotide excision repair of UV-induced DNA lesions [390]. Intriguingly, topoisomerase TOP2B, which creates DSBs during transcriptional activation [396], has been identified as a key regulator of transcription during the last stages of postnatal PR development [400]. Thus, DSBs in PR cells may arise in part from transcriptional activation of genes that encode components destined for the OS. Additionally supporting the importance of DNA repair to PR maintenance, Category 01 gene *Atr* encodes a master regulator of the DNA damage response that has surprisingly been linked to retinal degenerative disease and localized to the cilium [401]. Further, Category 03 gene *Nmnat1* encodes an enzyme that synthesizes nicotinamide adenine dinucleotide in the nucleus, which may regulate the large-scale polyADP-ribosylation of protein targets at sites of DNA damage [402]. Mutations in the genes encoding these proteins all result in PR cell loss [230,384,385,400,401,403–411]. Mutations in five of these genes as included in Figure 6 (*Cwc27, Ercc1, Ercc6, Sirt6,* and *Ubb*) caused moderate to slow progression of PR cell loss (D50 ≥ 2 months), consistent with a steady accumulation of unresolved DNA damage with age. The rapid PR cell loss observed in *Atrtm1Ofc* mice (D50 = 13 days) may reflect its direct involvement in OS development [401] in addition to the DNA damage response.

Due to the high percentage of alternatively spliced genes in the human retina [412,413], it is not surprising that mutations in mRNA splicing genes: *PRPF3, PRPF4, PRPF6, PRPF8, PRPF31, PDAP1,* and *BNC2* have been shown to lead to PR degeneration in humans [3]. In fact, in human retinal disease, 14% of disease genes are categorized as playing a role in RNA metabolism [383]. Interestingly, heterozygous humanized alleles of *PRPF3* and *PRPF8* and the null allele of *Prpf31* in mice do not recapitulate PR degeneration observed in humans but rather exhibit late-onset RPE degeneration [414]. In contrast, a hypomorphic allele of the mRNA splicing gene, *Cwc27*, with reduced viability, does lead

to moderate onset PR degeneration [415]. The differences observed among species require additional studies to unravel the complexities that govern genetic interactions.

Post-translational modification, which occurs by adding modifying molecules to amino acids or removing or altering these modified amino acids, is important for proper folding, transport/trafficking, localization, function, regulation, and/or degradation of proteins. Examples of post-translational modifications include phosphorylation, glycosylation, acetylation, ubiquitination, sumoylation, methylation, and lipidation [416]. Kinases that affect activity by mediating phosphorylation states are described elsewhere, however, post-translational modification genes affecting glycosylation and lipidation/prenylation are prominent among those that lead to PR degeneration. For example, the encoded proteins of *Fkrp*, *Large1, Pomt1,* and *PomgnT1*, necessary for the glycosylation of alpha-dystroglycan, essential for formation of the dystroglycan complex and for proper retinal lamination, lead to moderate rates of PR degeneration when disrupted. Prenylation is critical for proper trafficking and localization of retinal proteins. Of the three genes important in the prenylation and postprenylation processes, conditional loss of *Rce1* leads to an absence of phosphodiesterase subunits PDE6A, PDE6B, and PDE6C from the rod OS, probably due to a failure to prenylate one or more of these proteins [417]. By contrast, ablation of *Icmt* does not appear to affect phosphodiesterase transport but rather results in lowered levels of prenylated proteins GNAT1, PDE6G, and GRK1 [418], which are essential PR proteins. The null mutation of farnesyl-diphosphate farnesyltransferase 1, which adds a farnesyl group to the cysteine of the CAAX amino acid motif is prenatal lethal, but as a conditional tissue specific knockout may result in the same PR effects.

Two additional types of post-translational modification involve glycylation and glutamylation of proteins essential for normal connecting cilia function. Disruption of *Ttll3*, encoding a protein-glycine ligase necessary for glycylation of tubulin, results in an absence of glycylation in PR cells, shortening of the connecting cilia, and slow PR cell loss [419]. Interestingly, PR tubulin glutamylation increased in *Ttll3* mutant mice. TTLL5, tubulin tyrosine ligase like 5, adds glutamate residues on proteins. Sun et al. [420] reported that *Ttll5* disruption leads to late onset, slowly progressive PR cell loss that phenocopied retinal disease observed in *Rpgr* mutants. Perhaps this is not surprising as these investigators determined that TTLL5 glutamylates RPGR, a modification that is necessary for normal RPGR function in the PR cilium. *Agtpbp1* encodes a metallocarboxypeptidase that deglutamylates target proteins. Its disruption in *pcd* mutants leads to abnormal tubulin glutamylation [419] and an accumulation of vesicles in the interphotoreceptor space [421], indicating the importance of proper post-translational modification for PR survival.

### *5.5. Category 11: Immune Response*

As resident immune cells, microglia survey the retina constantly, presumably with the goal of removing unwanted debris and responding to damage arising from environmental and/or genetic stressors. They respond to damage by eliciting various responses that can range from regenerative to inflammatory depending on the type of injury. Thus, although microglia are unlikely to be instigators in RD, it may well be the case that microglia influence the severity of responses to ocular damage depending on mutations. Mutations in several genes central to the immune system lead to PR degeneration in mouse models. *Airetm1.1Doi* show early onset PR degeneration with 20% of ONL thickness loss at 10 weeks with rapid progression to 60% ONL thickness loss by 18 weeks [422]. *C3ar1tm1Cge* mutants show very slow PR degeneration with about 20% loss at 14 months [423]. *Cd46tm1Atk* show different rates of PR nuclei loss in male and female mice with 23% and 31% at 12 months of age, respectively [424]. Mutations in *Cx3cr1*, normally expressed in immune cells including microglia, were associated with PR cell loss. Homozygous *Cx3cr1tm1Litt* [425] and *Cx3cr1tm1Zm* [426] mice on the same C57BL/6J background showed similar rates of PR degeneration with 30% and 40% loss, respectively, at 16–18 months of age. However, *Cx3cr1tm1Zm* mice on the BALB/cJ background show complete nuclei loss at 4 months of age [426]. *Cxcr5tm1Lipp* causes late onset PR degeneration with 20% loss of ONL thickness at 17 months of age and RPE disorganization [427], whereas ablation of *Irf3* and *Igfbp3* showed mild PR degeneration at 2–4 months of age, about 10–14% [428,429]. *Ccl2* and *Ccr2* mutations also led to PR degeneration and fundus lesions, ONL loss in some areas and development of neovascular lesions, resembling phenotypes of AMD [430]. *Cfhtm1Mbo* was shown to have an impairment of rod and cone function by ERG and 29% decreased thickness of Bruch's membrane; however, rod opsin was distributed normally and no significant reduction in the number of PR cells was observed [431]. *Cfhtm1.1Song* demonstrated retinal whitening and cotton wool spots by fundus imaging [432]. Other genes involved in immune function that also showed PR degeneration as conditional knockouts encode transforming growth factor beta receptor II (*Tgfbr2*) [433] and aryl hydrocarbon receptor (*Ahr*) [434].

### *5.6. Omitted Models with PR Abnormalities that May be of Interest*

Based on the exclusion criteria described in the Methods section, a number of models with PR abnormalities caused by single gene mutations were not included in our final Table S1. Since we narrowly defined PR degeneration models as post-developmental loss of PR nuclei, some models, which were described with only OS alterations or ERG di fferences, were not included. For example, mice bearing a spontaneous point mutation in the *Ttc26hop* [435] that leads to the generation of a stop codon, Tyr430Ter, were reported to show OS shortening at one year of age with no PR loss. Likewise, ectopic expression of cone opsins in rod OSs led to scotopic ERG abnormalities but not PR degeneration in *Samd7tmlTFur* mice at 12 months of age [436]. The many allelic variants that cause ERG abnormalities without PR cell loss are listed in the MGI database and can be accessed through a phenotype query.

### *5.7. Factors Leading to Phenotypic Variability*

### 5.7.1. E ffects of Allelic Heterogeneity

Allelic heterogeneity is frequently a cause of phenotypic variability. For mouse models, this is often encountered when comparing a knockout model with spontaneous or induced mutations that still allow a protein to be produced. The latter would primarily be hypomorphic alleles due to amino acid substitutions, some splicing mutations that leave alternate splice forms intact and some C-terminal truncating mutations, which may retain some protein function. Often the knockout allele will be the more severe, presumably because in addition to the loss of protein function, the loss of the protein itself may cause secondary defects such as the failure to form a molecular complex that normally needs the native protein to form.

Mutations in the voltage gated calcium channel, *Cacna1f*, cause congenital stationary night blindness in humans due to abnormal neurotransmitter release in PR synapses. A null mutation in the *Cacna1f* gene ( ΔEx14–17) leads to an absent b-wave, abnormal PR synapses, lack of Ca2+ response in PR terminals and PR degeneration to 8 rows in the ONL at 8 months [314]. In contrast, an Ile756Thr amino acid substitution found in human patients and introduced into mouse, led to a di fferent phenotype with reduced b-wave, some intact ribbon synapses, a strong abnormal Ca2+ response, and a more severe degeneration (3–4 rows at 8 months of age [314]). Here the human allele represents a gain-of-function mutation that in addition to the loss of the original enzyme activity results in a new activity, or causes cell stress, which then induces additional phenotypes and makes the disease presentation more severe.

Within an allelic series of amino acid substitutions there are also frequently gradations of phenotypic severity. If a protein has several functional domains, mutations in di fferent domains may lead to distinct phenotypes. In addition, some mutations can lead to an abnormal tertiary structure of the protein. Such structural changes can lead to a failure to interact with binding partners or substrates/ligands or change the nature of such interactions [437]. Structural changes can also a ffect export of the protein from the endoplasmic reticulum (ER) and result in ER stress and eventually apoptosis of the cell [438].

One of the larger allelic series available is for human PRPH2 with more than 150 disease causing mutations reported [439]. Although only the secondary structure of the protein is available, some clustering of disease phenotypes is apparent. For example, the area around amino acids 190–220

on the intradiscal loop 2 is enriched for mutations causing autosomal dominant retinitis pigmentosa. This area is thought to interact with ROM1. Mutations leading to macular degeneration are more frequently present between amino acids 142 and 172. However, some macular degeneration and autosomal dominant retinitis pigmentosa mutations are also found elsewhere in the protein [439,440]. Once a 3D structure is available, we may find that the disease specific mutations may well be in spatial proximity and a clearer picture of the genotype-phenotype relation may be revealed.

Allelic heterogeneity can also arise from the intron/exon structure of the gene itself. Many genes produce several distinct transcripts through alternative splicing of their exons [441]. These differing transcripts can each produce proteins, which possess unique functions. For example, the *Rpgrip1* gene produces two splice variants that code for proteins that differ at their C-terminus, a full-length transcript and a shorter transcript encompassing exons 1–13 plus three additional C-terminal amino acids. An insertion between exons 14 and 15 of the full-length transcript leads to PRs with vertically stacked OS discs [442], whereas, a chemically induced mutation in the splice acceptor site in intron 6 that leads to a loss of both splice variant forms results in a failure to develop OSs altogether [443].

Despite the promise of genotype–phenotype correlation analyses to aid in the functional annotation of retinal proteins as well as in the diagnosis and prognosis of retinal degenerative diseases, few allelic series are ye<sup>t</sup> available. In humans the analysis is complicated by the fact that environment and genetic background effects can confound the allelic effect. In animal models, large allelic series are not ye<sup>t</sup> available.

Until recently allelic heterogeneity posed a problem for the generation of mouse models for human retinal diseases because only transgenesis and the generation of knockout models by homologous recombination were available. The removal of the gene products using knockouts can only model recessive or haploinsufficiency diseases, and often the complete lack of the protein will lead to embryonic lethality.

Transgenic models are associated with their own set of problems. Depending on the transgene integration site, the expression of the transgene can be reduced or cellularly restricted. Integration into an unrelated gene can disrupt expression of that gene and cause a phenotype that is not related to the transgene. The use of directed transgene insertion into safe sites, such as the Rosa26 locus (*Gt(ROSA)26Sor*) provides a workaround for some of these problems, although the choice of a promoter that faithfully mimics the native expression is still a difficult process. For these reasons, transgenic mouse models were not included in this review.

With the advent of CRISPR/Cas9 technology to produce precise cuts in genomic DNA, and the ability to perform gene editing through homology directed repair, it is now feasible to recreate human mutations in the mouse and directly probe for the phenotypic effects of allelic heterogeneity [444]. Comitato et al. present an interesting phenotype comparison of transgenic and knock-in rhodopsin P23H models [445].

### 5.7.2. Effects of Genetic Interactions

Gene interaction, or epistasis, is frequently observed during genetic analysis when two or more alleles at different loci combine to alter the onset, type, or severity of disease phenotypes. Such phenotype altering interactions arise from the organization of proteins and RNAs into macromolecular complexes and/or biochemical and regulatory pathways and networks. For example, consider hypomorphic mutations in two proteins that are components of a linear enzymatic pathway. Individually the reduced activity may not greatly impact the flux through the pathway, but combined in the same cell, the pathway flux may be reduced and become severe enough to induce a disease phenotype due to a lack of sufficient pathway product. Alternatively, a mutation may impair Pathway A, so that a disease phenotype arises. A second mutation may arise in a Pathway B that allows it to compensate for the malfunction in Pathway A and thus reduce the severity of the original disease phenotype. Mutations of this latter type of interacting mutations are called suppressor mutations and are extremely useful because they directly identify potential drug targets whose manipulation may be used to treat disease.

In general, identification of genetic interactors can be useful for placing the primary mutated gene in a biological context and help to define its cellular and organismal function. Often, the known function of a gene and its biology can sugges<sup>t</sup> candidate interacting genes. Similar to the first hypothetical interaction case above, mutations in two proteins involved in iron homeostasis, ceruloplasmin (CP), a ferroxidase associated with transferrin transport across the plasma membrane, and hephaestin (HEPH), implicated in iron transport across cells, individually do not show obvious PR degeneration. Combined in a double mutant mouse model, however, they lead to iron overload in the retina and subsequent RPE abnormalities and PR degeneration [446]. Another example involves two proteins necessary for retinoid recycling, ABCA4 and RDH8. Mutations in each alone do not show any phenotype; combined they cause all-*trans*-retinoid accumulation and PR degeneration [296]. Since previous studies had suggested that activation of TLR3 may lead to inflammation and mediating apoptosis [447], the authors explored the role of *Tlr3* in their *Abca4*/*Rdh8* double mutant model. Importantly, adding a targeted mutation of *Tlr3* to make a triple mutant mouse resulted in rescue of PR cells [448]. Here then the *Tlr3* mutation acts as a suppressor of the degenerative phenotype of the *Abca4*/*Rdh8* double mutant.

Additional interacting gene pairs have been found that affect PR degeneration, among them *MertktmlGrl*; *Tyro3tmlGrl* [449], *Cep290rd16*; *Bbs4*tm1Vcs [450], *Cep290rd16*; *Mkks*tm1Vcs [451], *Rpgrtm1Tili*; *Cep290rd16* [452], *Cngb1tm1.1Biel*; *Cnga3tm1Biel*; *Hcn1tm2Kndl* [453], *Crb1tm1.1Wij*; *Crb2tm1.1Wij* [454], *Dio3tm1Stg*; *Dio2tm1Vag* [455], and *Ercc6tm1Gvh*; and *Xpatm1Hvs* [456].

In addition to testing candidate interacting genes, methods have been developed to identify such interactors in an unbiased fashion that is illustrated below.

Effects of genetic background. For the calcium channel gene *Cacna1f* mentioned above, there is a third allele available. Chang et al. [457] reported the phenotype of the *nob2* mutation, an out-of-frame insertion of a transposable element into the *Cacna1f* gene, which is predicted to cause a truncation after 32 amino acids. The authors demonstrated by western blot that this is a null mutation and no protein is detected. Compared to the ΔEx14–17 null mutation, however, the phenotype of *nob2* is much milder with no apparent PR degradation [457]. The most likely explanation for this discrepancy can be deduced from the fact that the *nob2* mutation arose on the AxB6 recombinant inbred strain, a strain whose DNA is composed of alternate segments derived from C57BL/6J and A/J. It is likely that the A/J strain carries one or more modifier loci that suppress the PR degeneration induced by a *Cacna1f* null mutation.

Upon outcrossing an inbred strain carrying a mutation that leads to a particular phenotype with a different inbred strain, it is frequently observed that the phenotype of the offspring differs from that of the parents. This was often encountered in the past when knockout alleles were created in embryonic stem cells derived from strain 129/Sv and the founder animals were then made congenic on the C57BL/6 background. An early example is a study of a homozygous *Rho* knockout that was shown to lose PR nuclei significantly faster on the 129Sv background than on the C57BL/6 background [458]. Corresponding differences were also found in the number of apoptotic nuclei and in ERG responses. It was concluded that the B6 strain carries protective alleles of modifier genes that lead to a slower rate of PR degeneration [458]. Alternatively, it is also possible that 129Sv carries modifier alleles that accelerate degeneration.

Other inbred strains have also been reported to modify retinal phenotypes. For example, a targeted mutation of *Rp1* (*Rp1tm1Eap*) only showed moderate PR degeneration as an incipient congenic (N6) on the A/J strain background, but not on C57BL/6J or DBA/1J backgrounds [459]. ONL dysplasia and excess blue cone formation caused by loss of *Nr2e3* in C57BL/6J are suppressed by the genetic backgrounds of CAST/EiJ, AKR/J, and NOD.NON-*H2nb1* strains [460].

In principle, all inbred strains will carry modifier alleles. However, which strain modifies a particular mutation will depend on the primary mutation. It should be emphasized that an inbred strain represents a single genotype. In order to model the phenotypic spectrum of a human disease-causing mutation, many inbred strain backgrounds would have to be examined. Recently, advanced genetically diverse mouse populations have become available, such as the collaborative cross (CC) or the diversity outcross (DO) populations, that allow for more e fficient modeling of human populations compared to the classical inbred strains [461,462].

Modifier screens. Modifier screens are a tool to identify genes that modify phenotypic traits caused by a particular mutation. The disease modifying properties of inbred strains have been used for many decades to identify the underlying modifier genes by using genetic crosses, marker assisted genetic mapping of modifying loci, and positional cloning or more recently high throughput whole exome or whole genome sequencing approaches. For example, when B6.Cg-*Nr2e3rd7* homozygotes are outcrossed to CAST/EiJ, AKR/J, or NOD.NON-*H2nb1* and then the F1 mice intercrossed, homozygous *Nr2e3rd7* mice of the F2 generation are found that unlike the parental B6.Cg-*Nr2e3rd7* homozygotes have fewer spots on fundus examination and no PR layer dysplasia in histological sections [460]. This phenotypic variability is caused by the genetic interaction between the *Nr2e3rd7* disease allele and variants of so-called modifier genes that are specific to the outcross partner strain. Several quantitative trait loci (QTL) on chromosomes 7, 8, 11, and 19 were mapped [460]. Generation of a congenic line carrying the Chr11 modifier, along with further fine mapping, reduced the critical genomic interval to 3.3 cM. Several candidate genes were sequenced and a single nucleotide polymorphism was found in a nuclear receptor gene, *Nr1d1,* that is predicted to lead to an Arg409Gln amino acid change. Causality was confirmed by phenotypic rescue of the *rd7*-associated phenotypes by in vivo electroporation of a wild-type *Nr1d1* expression construct [463].

Several other modifiers have been mapped and identified based on inbred strain di fferences. For example, mapping crosses have been carried out for *rd3* (BALB/cJ and C57BL/6J, [464]), *rd1* (C3H/HeOu and FVB/N, [465]), *Crb1* (C57BL/6N and C57BL/6JOlaHsd, Chr15, [466]), *Mfrp* (B6.C3Ga and CAST/EiJ, Chr 1, 6, and 11 [467]), and *Tub* and *Tulp1* (C57BL/6J and AKR/J, *Mtap1a*, [468]).

Although not ye<sup>t</sup> widely used as a means to explore retinal biology, a very e fficient way to identify modifier genes is the use of a sensitized mutagenesis screen in which a male mouse carrying a mutation of interest is given a chemical mutagen and its o ffspring are examined for any change in the original phenotype. O ffspring carrying a potential mutation is backcrossed to the unmutagenized parental inbred strain to test for heritability and to reduce the mutational load. Mutations are identified using whole exome sequencing of the pheno-deviant mouse. This approach avoids the limited genetic diversity of inbred strains since in principle all genes can be mutated. An example of the utility of mutagenesis to search for modifier genes is the identification of a suppressor mutation in *Frmd4b* that prevents the PR dysplasia and external limiting membrane fragmentation observed in *Nr2e3rd7* mutant mice [469].

### 5.7.3. E ffects of Environment on PR Degeneration

PR cell loss has been shown to be induced by a number of environmental factors such as light, diet, and smoking in combination with particular genotypes. Perhaps not surprisingly, light exposure in some models bearing mutations in genes that function directly or in an ancillary fashion in the visual transduction pathway trend toward hastening PR degeneration [470,471]. For example, transgenic mice bearing the rhodopsin VPP mutation, widely used in visual transduction studies, is susceptible to light-exacerbated PR degeneration [472]. Likewise, mice carrying a homozygous *Prom1* null mutation are particularly susceptible to light-induced degeneration. At eye opening, with exposure to light, degeneration initiates at P14, and all PRs are gone by P20, whereas dark rearing from P8 to P30 leads to significant preservation of PRs [471]. Dark-rearing has also been demonstrated to delay PR degeneration in *Slc6a6tm1Dhau* (10% loss vs. 90% loss in normal vivarium lighting at three weeks of age) [473] or have no e ffect in C57BL/6-*Mitfmi-vit*/J homozygotes [474]. In some situations, light may actually trigger the disease phenotype, as is the case in *Sag* knockout mice [198,199], with three Class B1 Rhodopsin missense mutations, *Tvrm1* and *Tvrm4* [157] or *Tvrm144* [18], and in null mutation models of *Rdh12* [475], *Asic2* [476], *Myo7a* [477], *Whrn* [478], or *Akt2* [479]. *Sag* mutants must be reared in the dark to observe any PR cells. Under normal vivarium lighting conditions, the other light-sensitive

mouse models do not show PR degeneration or only a slight shortening of OS at one year of age, as in the those carrying *Rho* alleles *Tvrm1, Tvrm4,* or *Tvrm144,* and in retinol dehydrogenase (*Rdh12*) mutant mice. However, exposure to bright light or rearing under cyclic moderate-lighting, even subjecting mice to fundus examination, leads to PR degeneration. A comprehensive list of animal models and the effects of dark-rearing or light exposure can be found in reference [470].

Like light exposure, smoking and high fat intake have been proposed to have a negative impact on retinal function by increasing oxidative stress and inflammation in PR and RPE cells [480]. Smoking has been implicated as a major risk factor in the development of age-related macular degeneration in humans [481,482], and the results have been replicated in mouse models as well. Smoking leads to increased oxidative stress and inflammation in B6 mice [483] and in the presence of *Nfe2l2* deficiency [484]. Likewise, combinations of smoking and high fat intake in the presence of an *ApoB* mutation that promotes production of the APOB100 isoform [485] leads to significant loss of PRs [484]. Further, high-fat diet intake for certain genotypes, such as mutations of *Ldlr* [486] or certain alleles of *Apoe* [487], has been shown to compromise PR integrity in mice.

The majority of pharmacological or dietary interventions that have been reported in the relationship to PR degeneration in mouse models are associated with the goal of increasing vitamin A derivative availability [488–490] or reducing oxidative stress [491,492] in the retina. Heritable mutations in enzymes, such as LRAT or RPE65, required for processing of vitamin A within the retina are known to cause early onset RD due a deficiency of the 11-*cis*-retinal chromophore. E fficacy of treatment with 9-*cis*-retinal derivatives of mice with null mutations in *Lrat* and *Rpe65* mice is thoroughly discussed in a review by Perusek and Maeda [488,489]. Administration of antioxidants has in some cases improved PR survival. *Rs1tm1Web* homozygous females or hemizygous males fed a diet high in DHA [493] or *Pde6brd10* mice fed lutein and zeaxanthin [494] showed a significant PR preservation. Further, injections of a mixture of antioxidants—alpha tocopheral, ascorbic acid, alpha-lipoic acid, and/or Mn(III)tetrakis porphyrin—were able to slow the loss of cone/rod PRs in *Pde6brd1* [495], and *Pde6brd10* mice and in mice with a rhodopsin Q344ter mutation [492]. Environmental enhancement of *Pde6brd10* mice was able to significantly reduce PR loss presumably by reducing retinal oxidative stress [496].

### *5.8. Relationship to Human Disease Genes*

Of the 273 retinal degenerative disease genes in RetNet [3] for which mouse homologs exist, mouse models are available for 110 or 40% of them, including both germline and conditional mutants (Figure 7). Through our survey, we found 120 additional genes, in which mutations lead to PR degeneration. These genes could serve as candidates for ye<sup>t</sup> to be identified human retinal diseases. The available mouse models, for the most part, recapitulate the human disease phenotype well and permit mechanistic and therapeutic studies. However, apparent failures of mouse models do occur. When mutations in *MFRP* were first identified in humans [497], mice were thought to be a poor model because unlike humans [498], mice were previously reported to develop PR degeneration [499], and the microphthalmia and hyperopia found in human patients had not been reported in homozygous *Mfrprd6* mice. In subsequent years, numerous human patients have been identified that do show a degenerative phenotype [500] and hyperopia was detected both in a mouse model carrying a human *MFRP* c.498\_499insC allele [501] and the original *Mfrprd6* mouse (our unpublished observations). An important family of deaf–blindness diseases, Usher syndrome, was also thought to be poorly recapitulated in mice, because early models like the shaker-1 mouse had only the characteristic hearing loss, but no retinal degeneration [502]. Later, however, it was found that moderate light exposure does result in photoreceptor degeneration in shaker-1 mice [477]. In addition, a knock-in of the Acadian *USH1C* c.216G>A mutation into the mouse *Ush1c* gene recapitulates both deafness and retinal degeneration phenotypes [503]. In many cases, discordance between the human and mouse phenotypes can be attributed to insu fficient information about variation in the human disease, or to allelic e ffects (knockout vs. hypomorph or gain of function, expression of alternatively splice isoforms), or strain background (modifier genes) in the mouse models. Such shortcomings in mouse models can often be addressed by testing multiple models, including human disease alleles, and by using multiple genetic backgrounds.

**Figure 7.** Comparison of the number of RD genes identified in human (RetNet) and mouse as listed in Table S1 and summarized in Figure 3. The total number of genes in the RetNet database that cause monogenic disease and have mouse homologs is indicated, as is the total for which conditional or germline mutations have been associated with PR cell loss in mice, as described in this review. Numbers within the overlapping areas of the diagram represent genes present in both RetNet and Table S1; the remaining numbers represent genes that are unique to the indicated category.

Although humans and mice share about 98% of their genes, species differences do exist and need to be considered when selecting a model. Examples of vision-related genes that mice lack are *EYS*, *ARMS2*, and *CETP*. Species differences are the result of different evolutionary histories; humans and mice have encountered different pathogens, resulting in adaptations of our respective immune systems. Mice have different nutritional requirements, resulting in differences in lipid metabolism. Additionally, mouse eyes are adapted to a nocturnal life, resulting in a rod dominated retina with no macula. Nevertheless, mice possess all of the same retinal cell types necessary for vision and the vast majority of the same genes, and even when missing genes are introduced into mice they result in relevant phenotypes. For example, in a transgenic mouse model for Stargardt-like macular degeneration 3 due to a mutation in *Elovl4*, PR cell loss occurs in the central retina in a pattern that resembles the human disease [504]. For the many retinal diseases still in need of models, including complex diseases such as AMD or diabetic retinopathy, it remains the case that valuable new insights into disease mechanism and basic eye biology can still be obtained from mouse studies.
