5.2.2. Category 02: Visual Transduction

Overview. Mutant alleles of genes encoding proteins responsible for light detection comprise a second category of models (Category 02: Visual Transduction; Figure 1c, Table S1). The multistep phototransduction process that detects light and amplifies this signal is similar in rod and cone cells, but the specific proteins that catalyze many of the steps are often unique to each cell type [141]. Phototransduction is initiated by the response of opsin-based light-sensitive G protein coupled receptors that are covalently linked to vitamin A retinal as a cofactor. The receptor rhodopsin (RHO) is expressed exclusively in rod cells and is optimized to detect dim green light. Cone pigments that detect short or medium wavelength visible light (OPN1SW and OPN1MW, respectively) are exclusively expressed in cone cells, in some retinal regions coordinately within the same cell. These receptors constitute >90% of OS protein and are localized to the disc membranes.

Light activation of RHO or cone pigments causes the bound retinal to isomerize from an 11-*cis* to an all-*trans* configuration, ultimately leading to its release from the receptor by hydrolysis. Isomerization results in a conformational change in the protein that alters its interaction with a bound heterotrimeric G protein, transducin, activating the exchange of GTP for GDP bound to the α subunit of this protein. In turn, activated α transducin-GTP binds the inhibitory γ subunits of phosphodiesterase 6, releasing it from the α and β subunits of this complex, which are thereby activated to catalyze the conversion of cGMP to GMP. The ensuing reduction in cGMP levels in the OS closes the cGMP-gated cation channel, slowing the influx of Na<sup>+</sup> and Ca2+ ions, which hyperpolarizes the plasma membrane of the OS and, ultimately, the entire cell. Hyperpolarization causes Ca2+ channels to close at the cell synapse, which leads to a decrease in the calcium-dependent release of glutamate-containing vesicles into the synapse and activates postsynaptic bipolar neurons.

The process is regulated to ensure the highest sensitivity to illumination. Following its activation, rhodopsin is quenched by the action of arrestin, which binds to bleached opsin molecules that are phosphorylated by rhodopsin kinase. Resetting of the cell following the light flash requires the formation of cGMP from GTP, catalyzed by a membrane-bound guanylate cyclase, the subsequent closing of the cGMP-gated cation channel, and the restoration of electrolyte distribution across the plasma membrane as achieved by ion pumps and transporters. Hydrolyzed retinal is passed from the OS to the RPE as part of the visual cycle (see below), where it is re-isomerized and returned to the PR cell to regenerate bleached opsin. An additional visual cycle involving Müller cells contributes to the regeneration of cone pigments.

Visual pigments. Profound e ffects on PR viability are observed due to mutations that a ffect rod cells, which represent 97% of the PR population. Mouse models bearing *Rho* alleles exhibit semidominant and recessive rod cell loss phenotypes that vary greatly in the onset and rate, consistent with the variety of possible disease mechanisms that have been proposed for RHO mutations over decades of study. For example, some missense alleles in Table S1, such as those that encode the Pro23His, Cys110Tyr, Tyr178Cys, and Cys185Arg variants [21,142–146] may support a hypothesis that excessive RHO misfolding in the endoplasmic reticulum induces cellular stress pathways that lead to PR cell loss [147]. Although the pathways linking misfolded RHO to cell death are not fully resolved, recent studies of the Pro23His variant in cultured cells and in rats [148] or mice [145,146] sugges<sup>t</sup> that stress pathways induced by the unfolded protein response are protective, and raise the possibility that increased intracellular calcium due to ER stress may cause cell death [146]. Misfolding may also explain the partial mislocalization of RHO Glu150Lys to the IS [149]. However, in this mutant, much of the protein appears to be correctly exported to the OS, where it leads to irregularly shaped and disorganized discs, possibly due to a defect in higher-order RHO organization [149]. Pro23His RHO also disrupts the orientation of discs during their morphogenesis, possibly through similar e ffects on higher-order structure [150].

By contrast, the e ffect of the Gln344Ter variant (Table S1), which is correctly folded but includes sequence extensions at the C-terminus that interfere with export to the OS [151], as well as the graded effect of heterozygous or homozygous knockout alleles *Rhotm1Jlem* and *Rhotm1Phm* [152,153] or the premature truncation mutant Arg107Ter (Table S1), provides evidence that a steady flow of RHO to the OS is essential for PR cell viability. These observations fit an emerging view that a proteostasis network, incorporating not only cellular stress pathways but also protein tra fficking and degradation, regulates the cellular protein balance to ensure viability [147,154]. According to this view, a failure to sort vesicles bearing RHO from the Golgi to the periciliary membrane, or a partial or complete loss of the protein, leads to protein imbalance in the IS. This imbalance may induce cellular stress responses and also a ffect the tra fficking of other molecules destined for the OS, such as other phototransduction proteins, lipids, and vitamin A, resulting in cellular toxicity. Finally, the RHO Asp190Asn variant (Table S1) appears to tra ffic properly to the OSs but may have structural defects that lead to constitutive signaling [155], which has been linked to PR degeneration [156]. The same mechanism may account for the e ffect of *Rho* mutants that result in rapid degeneration upon bright illumination [157] but were not included in Table S1 due to the dependence of the mutant phenotype on an environmental perturbation (see Discussion). Future studies of these and other models may resolve or converge the many proposed hypotheses to explain RHO-associated RD.

Based on the often profound e ffect of *Rho* variants on rod cell viability, it might be expected that cone pigment variants would similarly cause cone PR cell loss. However, cones remain viable for more than 1.5 years in homozygous *Opn1swtm1Pugh* mice, which show a 1000-fold decrease in transcript and produce no detectable OPN1SW by immunoblotting, histochemistry, or single-cell recording of light responses [158]. Likewise, cones are viable for at least 10 months in homozygous *Opn1mwtm1a(EUCOMM)Wtsi* knockout mice, despite an absence of OPN1MW in immunoblotting and immunohistochemical studies [159]. These studies sugges<sup>t</sup> fundamental di fferences in the cellular sensitivity of rod and cone cells to visual pigment deficiency. They also highlight the concern that reactivity to antibodies against cone opsins or other cone cell markers may be abolished even though the cells remain viable, and therefore may not be as reliable as counting cone nuclei [160] to assess cell loss.

Transducins. Rod transducin subunits α, β, and γ (encoded by *Gnat1*, *Gnb1*, and *Gngt1,* respectively) form the heterotrimeric G protein complex that is essential for propagating the signal from light-activated rhodopsin. *Gnat1* knockout mice have attenuated rod responses and model congenital stationary night blindness (CSNB) [161]. Although slow PR loss was reported for this model, our measurement of ONL thickness at four weeks of age based on reported images yielded a value of 90% of wild type, matching the author's value at 13 weeks [161] and suggesting an early developmental di fference rather than progressive cell loss. In support of this finding, others using the same strain reported ONL thickness was 85% of wild type at eight weeks of age with no evidence of significant cell loss up to 52 weeks of age [162]. By contrast, IRD2 mice, which are homozygous for a *Gnat1irdr* allele predicted to yield a prematurely truncated polypeptide, exhibit significant rod PR cell loss (Table S1) accompanied by late cone cell loss and reduced rod-specific ERG responses [163]. Homozygous *Gnat1irdr* mice may recapitulate recessive rod-cone dystrophy, which has recently been linked to human *GNAT1* variants predicted to encode prematurely truncated proteins [164–166]. The *Gnat1irdr* allele was discovered independently in *rd17* mice at JAX, suggesting a founder e ffect [167,168].

*Gnb1* knockout mice have not been studied due to embryonic and perinatal lethality. However knockout alleles of the gene encoding rod γ transducin, *Gngt1tm1Dgen* and *Gngt1tm1Ogk*, result in PR loss that is more rapid than in *Gnat1* mutants [169,170]. In these strains, GNGT1 deficiency is accompanied by a 6- to 50-fold post-translational reduction of GNAT1 and GNB1, indicating a key role of the transducin γ subunit in complex assembly. *Gngt1tm1Dgen*-associated degeneration is rescued by heterozygous *Gnb1Gt(prvSStrap)4B8Yiw* mice [171], which express retinal GNB1 at 50% of wild type levels. This result suggests that the toxicity of GNGT1-deficiency is due to an excess of improperly assembled GNB1, which is targeted for degradation but exceeds the capacity of the proteasome [171]. This observation supports the proteostasis network model of PR degeneration [154].

Among genes encoding cone transducin subunits α, β, and γ (*Gnat2, Gnb3,* and *Gngt2*), only *Gnat2* alleles have been reported to cause PR loss. A progressive reduction of cone cell ERG responses and a 27% decrease in PNA-positive cells at 12 months of age in homozygous *Gnat2tm1Erica* mice (Table S1) is consistent with cone PR loss [172]. However, cone nuclei were not counted directly, so it is possible that cone cell loss is less pronounced than reported. The predicted GNAT2 Asp173Gly substitution in this model may alter guanine nucleotide binding [172], although how this change might cause cell loss is unresolved. Interestingly, mislocalized cone opsin OPN1MW in this model suggests endoplasmic reticulum stress, which is often associated with PR degeneration. *Gnat2cpfl3* mice (Table S1) show no cone cell loss for at least 14 weeks but exhibit a slow loss of rod cells [173]. In contrast to these models, a recently developed *Gnat2* knockout strain abolishes GNAT2 function without PR loss or dysmorphology in the oldest mice examined at 9 months of age [174]. Although human *GNAT1-*variants are a rare cause of achromatopsia [175], a stationary congenital colorblindness, the clinical presentation is variable and some cases are associated with a reduction in visual acuity with age [176] that may sugges<sup>t</sup> progressive cone cell loss. The available mouse alleles may help to identify disease mechanisms that contribute to this phenotypic variability.

Phosphodiesterase 6. Rod phosphodiesterase 6 consists of a catalytic αβ complex encoded by *Pde6a* and *Pde6b* and two inhibitory γ subunits encoded by *Pde6g*. The control of cGMP levels by this enzyme is expected to a ffect both PR function and viability, as cGMP has a central role in the phototransduction cascade and PR cell metabolism [177], and elevated cGMP levels have been linked to PR cell loss [178]. Indeed, *Pde6a* and *Pde6b* mutants show depressed ERG responses at an early age and rapid PR loss with D50 values of 11–30 days (Figure 6, Table S1). A study of *Pde6a* mutations on the same strain background made use of an allelic series that varied in disease severity [179]. The order of disease progression due to the alleles reported in this study, *nmf282* (Val685Met; fastest) > *tm1.1Bewi* (Arg562Trp) > *nmf363* (Asp670Gly; slowest), is the same as assessed by D50 (Figure 6). This allelic series led to a correlation of more rapid PR degeneration with an increased number of cGMP-positive PR cells [179]. The same trend in the progression of disease in *Pde6anmf282* and *Pde6anmf363* mice was found earlier [180], but an opposite cGMP result was obtained, possibly due to the assessment of total retinal cGMP rather than a count of cGMP-positive PR cells [179] (a 0.1-month di fference in the D50 of *Pde6anmf363* mice measured in the two studies may reflect strain di fferences that might also contribute to the di fference in findings). The later study also combined two alleles that matched human *PDE6A* variants to create a compound heterozygote [179], mirroring the more typical situation in human genetic disease. Further, the allelic series highlighted a non-apoptotic cell death mechanism involving calpain rather than the expected caspase-mediated apoptotic process [179]. Both elevated cGMP and calpain activation have been observed in other mouse RD models [181]. Thus, allelic series as used in these studies are informative for assessing disease mechanisms and identifying potential di fferences in treatment e fficacy that may reflect disease severity.

Of the *Pde6b* alleles described, *Pde6brd1* and *Pde6brd10* have been used most extensively as PR degeneration models. *Pde6brd10* disease develops later, providing a longer window of opportunity to test therapeutic e fficacy (Figure 6). The *Pde6batrd1* model has an even slower progression (D50 = 0.71) than *Pde6brd10* mice (D50 = 0.65), which may make it more attractive for assessing the variation in treatment with disease severity (Figure 6, Table S1). Finally, loss of the inhibitory subunit in homozygous *Pde6gtm1Go*ff mice did not lead to an expected increase in catalytic activity; instead PDE6G was found to be essential for activation and possibly stable assembly of the holoenzyme [182].

Cone phosphodiesterase 6 includes two catalytic α subunits encoded by *Pde6c* and two inhibitory γ subunits encoded by *Pde6h*. The *Pde6ccpfl1* mutation leads to severely reduced cone ERG response at three weeks and progressive cone PR loss with age [15] as determined by counting cone nuclei (Bo Chang, unpublished data, presented in Table S1). This model mimics achromatopsia in humans, which is sometimes accompanied by cone PR cell loss [183]. Surprisingly, *Pde6h* knockout mice show no detectable functional cone loss or degeneration, likely due to the expression of the *Pde6g* subunit in mouse cones, which may compensate for PDE6H loss [184]. Variants in human *PDE6H* cause achromatopsia [185,186] but cone cell loss has not been reported.

Cyclic nucleotide gated channels and cation exchanger. The decrease in cGMP levels resulting from PDE6 activation leads to the closing of cyclic nucleotide cation channels in the OS plasma membrane of both rods and cones. Channel closing diminishes the inward flux of Na<sup>+</sup> and Ca2+ ions that maintain the PR cell in a hyperpolarized state. The rod protein encoded by *Cnga1* and *Cngb1* is an <sup>α</sup>3β1 heterotetramer, in which the β subunit is a long isoform, CNGB1a [187,188]. *Cnga1* mutations have not ye<sup>t</sup> been described. Rod OSs of homozygous *Cngb1tm1.1Biel* mice yield no detectable CNGB1a or CNGA1, and rapid PR loss is observed [189]. Together with evidence that CNGA1, but not CNGB1a, is capable of self-oligomerizing in heterologous expression systems, this result suggests that CNGB1 plays a critical role in stabilizing CNGA1 for channel assembly during synthesis in the secretory pathway and/or subsequent transport to the OS. Although the mechanisms leading to PR cell loss are unknown, low intracellular Ca2+ may overactivate guanylyl cyclase and cause toxicity due to elevated cGMP [189].

The cone channel encoded by *Cnga3* and *Cngb3* functions as an <sup>α</sup>2β2 tetramer. Due to the absence of downstream synaptic signaling associated with channel defects, mutations in both genes result in a loss of cone ERG responses modeling achromatopsia. In addition, the alleles included in Table S1, *Cnga3cpfl5, Cnga3tm1Biel*, *Cngb3cpfl10,* and *Cngb3tm1Dgen* result in cone PR degeneration as assessed by marker analysis, although confirmation of cell loss by a direct nuclear count was lacking in some studies. The mechanism of cell death is unknown in these models, but by analogy may involve elevated cGMP as hypothesized in rods.

A critical component of phototransduction is SLC24A1 (also called NCKX1), which exports sodium and calcium ions in exchange for potassium. This activity is responsible for the decrease in intracellular Ca2+ upon closing of the cGMP-gated channels. Homozygous *Slc24a1tm1Xen* mice exhibit slow degeneration, possible due to malformation of OS discs [190].

Guanylyl cyclase and activating proteins. Photoreceptor guanylyl cyclases function as homodimers encoded by two genes in mice, *Gucy2e*, and *Gucy2f*. In the homozygous *Gucy2etm1Gar* model, D50 was >12 months (Figure 6), indicating very slow rod PR cell loss, while cone cell numbers decreased rapidly to 33% of controls in 5 weeks [191]. Cone loss with rod preservation has been observed in Leber congenital amaurosis cases linked to variants of the human *Gucy2e* ortholog, *GUCY2D* [192]. However, *Gucy2etm1Gar* mice are not considered to model this disease because rod ERG function, though diminished, is still detectable [191]. Although *Gucy2f* knockout did not cause PR cell loss, double knockout of both guanylyl cyclase genes resulted in moderate degeneration [193]. Rod and cone ERG responses were abolished in this model, suggesting that the residual function in *Gucy2etm1Gar* mice was due to compensatory activity expressed from *Gucy2f*. The mechanism of PR cell loss in these models is unlikely to involve elevated cGMP as the enzymes needed for its production are ablated. The post-translational downregulation of other phototransduction proteins in double-knockout mice [193] may indicate a disruption of the proteostasis network that could explain PR cell loss.

Guanylyl cyclase activator proteins provide a feedback loop to restore cGMP levels. When intracellular Ca2+ is high, these proteins inhibit guanylyl cyclase; when Ca2+ levels are low, they switch to an activating Mg<sup>2</sup>+-bound conformation that promotes cGMP synthesis. This Ca2<sup>+</sup>-sensitive regulation permits PR cells to reestablish cGMP levels following light exposure due to lowered intracellular Ca2+, thereby resetting the cell for another stimulus. Double knockout of *Guca1a* and *Guca1b*, which encode the activator proteins in both rods and cones, had no detectable effect on retinal morphology up to eight months of age [194]. However, homozygous *Guca1atm1.1Hunt* mice, which have a Glu155Gly missense substitution identical to one found associated with a severe dominant cone dystrophy [195], result in rapid loss of cones and subsequently rods (Figure 6, Table S1). This mutation, like others associated with the human disease, may constitutively activate guanylyl cyclase due to a defect in calcium sensing [196], leading to cytotoxic accumulation of cGMP.

Recovery from light stimuli. Mechanisms to terminate the phototransduction cascade and recover the PR cell for additional stimuli include the phosphorylation of activated RHO by a *Grk1-*encoded kinase and the binding of *Sag-*encoded arrestin to the phosphorylated RHO. The binding of SAG limits transducin access to RHO and thereby prevents further activation of transducin and downstream processes. Significantly, defects in either gene induce photoreceptor cell loss, likely due to the accumulation of excess cGMP arising from unregulated active RHO. Early studies aimed at elaborating the role of the SAG or GRK1 proteins used mice raised in the dark [197,198], as typical vivarium cyclic light–dark rearing conditions were described as leading to rapid degeneration. Subsequent studies of homozygous *Sagtm1Jnc* [199] or homozygous *Grk1tvrm207* mice [200] reveal slow PR cell loss with D50 > 10 months under normal rearing conditions.
