Previous work had shown the long-term presence of degenerating cones in RP patients and mouse models (17–19). These cones are often referred to as ‘dormant cones’ once their outer segment is significantly diminished. Such cones retaining the nucleus and inner segment and perhaps remnants of outer segments are also referred to as dysflective cones or more generally as remnant cones (34). These cones may be the cells most likely to respond to therapies designed to prevent cell death and restore visual function. Cone photoreceptor reactivation studies using optogenetics showed the feasibility of restoring vision in high light conditions using microbial opsin-based optogenetics (17–19). However, the high light requirements and the potential immunogenicity of using an opsin from prokaryotic origin are inherent disadvantages of such an approach. An interesting alternative is the use of mammalian opsins for vision restoration, however all work in this field has so far been focused on inner retinal neurons (23, 25, 35–37). After exploring the phototransduction cascade in two RCD mouse models, we propose, a new gene therapy strategy based on remaining endogenous cone opsins (Fig. 1). Indeed, we revealed that cone opsin and cone arrestin remain expressed in cone cell bodies at late disease stages both in mouse models of RP and in human RP patients. Based on this information, it is plausible to insert a channel acting via Gi/o proteins recruited by the activation of the remaining opsin and thereby creating an alternative ‘short phototransduction cascade’ within the cone photoreceptor (Fig. 1B). Such phototransduction cascade provides increased light responses as long as endogenous cone opsins are still present in viable cones. We propose that in early stages, while transducin is still present, the activation of the opsin by a light stimulus recruits the α subunit of transducin leaving the βγ subunits available for activation of GIRK2 channels, generating additional hyperpolarization (Fig. 1B). Alternatively, or additionally in later stages when transducin is no longer present, the opsin can recruit other G proteins present in degenerating cones targeting GIRK2 channels, subsequently allowing the efflux of potassium ions at the resting membrane potential of degenerating cones (17). K+ efflux via GIRK2 channel hyperpolarizes cone photoreceptors in response to light modulating glutamate release and light responses in two RCD mouse models by ERG and OKT. Since remaining opsin in cone cell bodies is still functional within its regular spectrum, the insertion of GIRK2 in all cones with PR1.7 promoter leads to light responses following the spectral properties of each of the opsins, therefore, allowing the preservation of color vision. We thus anticipate that our approach will provide, for the first time, color-vision restoration with both high acuity and sensitivity.
A clear advantage of microbial opsins is their robustness and millisecond scale kinetics (20, 38). For systems using other opsins, it should be considered that in order to respond to another light stimulus, the cascade has to be deactivated to recover light sensitivity. In absence of this, cones may stay hyperpolarized after GIRK2 channel activation limiting their ability to modulate synaptic transmission at a movie rate compatible with motion vision. In our case, signal termination in the cones was made possible thanks to the cone arrestin that is still maintained at advanced stages of the disease in both RCD models and patient retinas. This is readily visible in the 10Hz-flicker ERG traces showing responses of the retina during repetitive light stimuli and also by the improved optokinetic reflex of treated mice.
Our observations of ‘dormant cones’, i.e. degenerating cones with diminished outer segments and light sensitivity, in human RP retinas is consistent with previous reports (4, 17, 39–41). Li et al. (4) reported somata with very short or absent OS reactive for cone opsins, recoverin, and transducin-∝ cones in several RCD patients while Lin et al. (41) also found cones with abnormal, diminished OS that were positive for recoverin immunostaining in one RCD patient. Busskamp et al. (17) showed the presence of such cones using OCT in RCD patients. We show here the presence of this type of cones in the maculae of 3 of 4 RP patients studied. Importantly, we found co-expression of cone opsins and cone arrestin in the same cells, which supports our GIRK2 gene therapy approach as these two phototransduction cascade proteins are required for our cone reactivation strategy using GIRK2. Last, the fact that incorporation of GIRK2 enhances existing light responses in cones even prior to complete outer segment loss offers the possibility to implement this gene therapy in mid stages of the disease.
Based on our investigation of the proportion of potential eligible patients in our RCD cohort from the Quinze-Vingts hospital we found that roughly one quarter of RP patients with low to no light perception can be eligible for GIRK2 therapy to restore light sensitivity in their dormant cone population. Examination of the inner and outer segment structures in RCD patients using AO imaging revealed refractive changes in the inner and outer segments of the foveal cones suggesting it is possible to distinguish between cones with diminished outer segments, inner segments and light insensitive ‘dormant’ cone populations using this technique. The combination of the above-mentioned imaging techniques along with more recently described indicators of cone cell function (such as fundus autofluorescence imaging (42)) can be used to select patient populations for this type of gene therapy.
Despite functional improvements in GIRK2 expressing cones, our treatment does not stop the degeneration of cones. However, retinal degeneration in mice is much faster than in humans, thus a few months of therapeutic efficiency in mice may be equal to several years in humans. We recorded a decrease in the response of treated cones to light stimuli, which was consistent with decrease in cone numbers and the fact that we did not transduce all cones due to subretinal injection further limiting the beneficial effect. AAV vectors showing better lateral spread can be used to increase transduced cone numbers beyond the bleb (19). In order to increase the therapeutic window, neurotrophic factors can be implemented alongside our approach. Indeed, AAV-mediated secretion of neurotrophic factors such as the rod-derived cone viability factor (RdCVF) have been shown to delay cone cell death and may be combined with GIRK2-mediated sensitization (5).
Guanine nucleotide-binding proteins act as molecular switches inside a multitude of cells (43). They have a seven-transmembrane domain receptor, and their activity is regulated by factors that control their ability to bind and hydrolyze GTP to GDP. The most commonly identified G proteins in the retina are heterotrimeric and are composed of three subunits (α, β, γ). They are activated by light sensitive G protein-coupled receptors (GPCR) such as rhodopsin or cone opsin and transmit the message by activating other proteins in the cascade. In photoreceptors (PRs), the G protein that activates the cascade (transducin) belongs to the Gi/o family inhibiting the production of cyclic guanosine monophosphate (cGMP) from GTP and therefore inducing hyperpolarization of photoreceptor membrane and a subsequent decrease of glutamate release (44). Recently, the expression of a medium wavelength-sensitive cone opsin in the retinal ganglion cells has been shown to restore high light sensitive vision with adaptation (21). Even though the G protein coupled mechanism used by cone opsin in retinal ganglion cells remains unknown at this stage, such results further show the potential of vertebrate cone opsins in vision restoration. They also suggest that cone opsins are not specific to a single type of G protein with the ability to activate target channels in different neurons (45). Various ion channels can be activated via a G protein starting with the light-stimulated opsin. These findings altogether lead us to believe that our combined approach with GIRK2 or a similar dual approach with a target channel and opsin could be implemented in other subtypes of neurons broadening the field of applications towards other targets (27).