Gene therapy approaches to slow or reverse blindness from inherited retinal degeneration: growth factors and optogenetics

Abstract

To date, clinical gene therapy efforts for inherited retinal degeneration (IRD) have focused largely on gene replacement. The large number of genes and alleles causing IRD, however, makes this approach practical only for the most common causes. Additionally, gene replacement therapy cannot reverse existing retinal degeneration. Viral-mediated gene therapy can be used for two other approaches to slow or reverse IRD. First, by driving intraocular expression of growth factors or neuroprotective proteins, retinal degeneration can be slowed. Second, by expressing light-sensitive proteins (either microbial channelopsins or mammalian G-protein coupled opsins) in preserved inner retinal neurons, light sensitivity can be restored to the blind retina. Both approaches have advanced substantially in the past decade, and both are nearing clinical tests. This review surveys recent progress in these approaches.

Introduction

Macular degeneration is the most prevalent cause of blindness in the developed world, while hereditary retinal degenerations are the leading congenital cause of blindness. Both groups of diseases have genetic underpinnings. For macular degeneration, specific alleles of genes including complement factor H and genes in the chromosome 10q26 region including ARMS2 and HTRA1 confer increase risk of disease. However, these gene associations are not Mendelian; disease occurs in many patients who carry low risk alleles, while other patients with high risk alleles never develop disease. It is unclear at present whether gene replacement or editing would alter the course of disease. Hereditary retinal degenerations are generally Mendelian, and may follow autosomal dominant, autosomal recessive, or X-linked inheritance. The challenge in hereditary retinal degeneration, particularly for the most common retinitis pigmentosa family of diseases, is the large number of genes and specific mutations that can result in the final common pathway of outer retinal degeneration. It is currently estimated that there are at least 3100 alleles of 190 different genes capable of causing retinal degeneration. While targeted gene replacement has been successful for treatment of Leber congenital amaurosis due to RPE65 mutation,^1,2 the prospect of developing, validating, and gaining approval for specific gene therapy for each of the gene defects causing retinitis pigmentosa is daunting and likely not achievable. Further, gene replacement therapy may not be successful for autosomal dominant causes (particularly when the disease mechanism is due to a hypermorphic or antimorphic allele), where gene editing approaches would need to be considered.

Other forms of gene therapy may have more general applicability. It has become clear over the past two decades that hereditary outer retinal degeneration follows a final common pathway which can be non-cell autonomous. For example, in the most frequently encountered forms of retinitis pigmentosa, the causative mutation is expressed only in rod photoreceptors (for instance, rhodopsin mutations or phosphodiesterase 6b mutations), but vision loss becomes profound when there is secondary degeneration of the cone photoreceptors. Gene therapies that lead to constitutive expression of trophic factors may slow these retinal degenerations non-specifically. Alternatively, gene therapy can be used for functional recovery or repair. These approaches include expression of photoreceptive molecules in preserved cells in retinal degenerated retinas such as retinal ganglion cells (i.e. optogenetics). Finally, gene therapy can be used to attempt to drive retinal repair, by causing transdifferentiation of retinal precursor stem cells to replace damaged cells in the retina. In this review, each of these approaches will be considered in turn.

Trophic and neuroprotectant factor gene therapy for attenuating hereditary retinal degeneration

The clinical course of most hereditary retinal degenerations is prolonged. For many forms of retinitis pigmentosa, degeneration begins in the peripheral rods leading to nyctalopia and reduced peripheral visual field. Later in disease, macular cones become dysfunctional and die, leading to loss of central vision. Given 1.) the very slow mechanisms and timecourse of degeneration³ (in many cases decades) and 2.) the non-cell autonomous nature of degeneration (for instance, loss of cones in disease where the mutant gene is expressed exclusively in rods), many investigators have posited that trophic secreted factors may influence the timecourse of degeneration.^4,5 This has led over the years to a number of attempts to slow retinal degeneration by forcing intraocular expression of secreted neurotrophic or neuroprotective factors using gene therapy vectors. In each case, the strategy employed has been to clone the growth factor or neuroprotectant in question into a non-pathogenic, replication-incompetent viral vector (almost always a variant of adeno-associated virus [AAV], although replication incompetent lentiviral vectors may also be employed) and express the secreted factor non-specifically in ocular tissues.

Ciliary neurotrophic factor (CNTF).

CNTF is a 200 amino acid secreted protein initially identified in chicken embryonic ciliary nerve cultures. CNTF is a particularly interesting protein with respect to retinal degeneration. Intravitreal injection of recombinant CNTF protein into the vitreous of several mouse retinal degeneration models led to short-term rescue of photoreceptors.⁶ CNTF delivered by recombinant adenovirus delayed photoreceptor degeneration in rds/rds and rd/rd models,^7,8 and conferred life-long preservation of cone numbers in Rho^−/− mouse model of retinal degeneration.⁹ Expression of CNTF by recombinant AAV virus was able to rescue photoreceptor degeneration in rat models of retinal degeneration as well.¹⁰ However, Liang et al¹¹, in applying AAV-mediated gene therapy for CNTF to both mouse and rat retinal degeneration models, noted anatomic preservation but did not see functional improvement in ERGs of these animals; indeed, in the rat S334ter rhodopsin transgenic model, the electroretinograms (ERGs) of treated animals were actually lower amplitude than those of controls. CNTF application was ultimately studied in human clinical trial for retinal degeneration (including both retinitis pigmentosa and non-exudative macular degeneration) utilizing an encapsulated cell secretion technology.^12,13 Although early phase results were promising in terms of both retinal thickness preservation and visual function,¹⁴ 1 year results for human clinical trials showed preservation of thickness in retinitis pigmentosa patients but actually showed reversible worsening of function in high dose implants¹². This device remains in development for possible use in macular telangiectasis.¹⁵

Brain-derived neurotrophic factor (BDNF)

BDNF is another member of the neurotrophin family. This 116 amino acid secreted protein has also been demonstrated to be a strong survival factor for neurons. As with other members of this family, direct injection of BDNF into the vitreous has a transient survival benefit in light-induced retinal degeneration.^16,17 It is also expressed natively in several cell types in the retina. Gauthier et al.¹⁸ generated an adenoviral vector expressing BDNF and targeted expression to Muller glia. The authors found that in light-induced degeneration in albino rats, this vector had a substantial protective effect, increasing the number of surviving photoreceptors 10 days after light administration by ~50%. This was accompanied by a more modest improvement in ERG amplitude.

Basic fibroblast growth factor (FGF).

FGF is another family of trophic proteins that are expressed endogenously in the eye and upregulated with injury. Lau et al. expressed FGF-2 under adeno-associated virus in S334ter-4 transgenic rats.¹⁹ Similar to the results observed for CNTF, photoreceptor number appeared better preserved in animals treated with the growth factor; however, there was no demonstrable effect on electrophysiologic response. FGF-2 has also been used in a simian lentiviral-derived vector in combination with pigment epithelial-derived factor (PEDF) in both Royal College of Surgeons rats with outer retinal degenerative disease, as well as retinal degeneration slow (rds or peripherin^−/−) mice.²⁰ Each agent by itself was able to slow degeneration and delay loss of function alone, but together the agents appeared to have synergistic effect.

Glial-derived neurotrophic factor (GDNF).

GDNF was discovered by virtue of its ability to enhance survival of dopaminergic neurons in vitro. It is a member of the transforming growth factor-superfamily. Expression using an AAV vector preserved retinal thickness and delayed declines in electroretinography in the S334ter-4 transgenic rat model of retinal degeneration²¹. Further development of this approach entailed driving GDNF expression specifically from Muller glial cells. When Dalkara et al. utilized this approach in S334-ter transgenic rats, they noted substantial and sustained increased retinal thickness and ERG amplitude in these animals.²² Further, the suppressive effect on ERG of long-term gene therapy with CNTF did not appear with long-term GDNF therapy.²³

Pigment epithelial-derived growth factor (PEDF).

PEDF factor is a trophic factor initially purified from retinal pigment epithelium, and thus a prime candidate for a local trophic factor for photoreceptor survival. Miyazaki et al.²⁴ expressed PEDF via a simian lentiviral construct in Royal College of Surgeons rats with retinal degeneration. This group found significant preservation of outer retinal nuclei as well as improved electroretinograms compared with vector-only controls at 4 weeks and 8 weeks post injection. (However, it should be noted that ERG amplitudes were still ~10% of normal by 8 weeks even in treated animals). AAV-expressing PEDF has also been investigated for ocular use, but in tandem with anti-vascular endothelial growth factor (VEGF) in a laser-induced neovascularization model. In this case, the PEDF is used as an anti-angiogenic agent and appears to be synergistic with reduction in VEGF.²⁵ PEDF has also been engineered into adenoviral vectors, and used in phase 1 clinical studies as an anti-angiogenic treatment for exudative macular degeneration.²⁶ Although the treatment appeared well tolerated and phase 1 results looked promising, this treatment avenue does not appear to have advanced.

Rod-derived cone viability factor (RdCVF).

This factor is unique in having been discovered as a retinal-derived trophic factor specifically identified for its properties in enhancing cone survival; it was discovered in a screen for proteins enhancing cone-enriched cultures of chicken embryos.^4,27 This gene is a splice variant of the nucleoredoxin-like 1 (Nxnl1) gene, and protein acts via binding to Basigin-1 (BSG1), a transmembrane protein expressed by photoreceptors, which in turn binds to the GLUT1 glucose transporter, increasing glucose metabolism by cones and enhancing their survival.²⁸ AAV vectors expressing RdCVF, delivered systemically or intravitreally, were capable of slowing cone death and increasing cone electroretinogram function in both rd10 and transgenic P23H mouse models of retinitis pigmentosa.²⁹ Further, in the same study, a second vector expressing the RdCVFL splice variant showed improved rod survival in these models, and the combination appeared additive for rescue of rd10 mice. As the RdCVF family of proteins appears to be a native trophic pathway for maintenance of cone function in retinitis pigmentosa, this approach appears promising as a means for slowing cone degeneration.

Neuroprotectants and immune modulators

While the growth factors described in the previous section were all identified specifically as survival factors, several other proteins have been investigated that have neuroprotectant effects separate from other identified physiologic functions. Proinsulin has been investigated as a neuroprotectant for retinal degeneration. Fernandz-Sanchez et al.³⁰ expressed proinsulin in an AAV1 vector and tested for effect of injection at P20 on disease course in transgenic P23H rhodopsin rats, which have a strong, dominant degeneration phenotype. This group showed high expression levels in the eye and showed substantial rescue of both photoreceptor number as well as ERG amplitude. Further, at an ultrastructural level, this group demonstrated preservation of outer retinal synapses. Given the results with RdCVF discussed previously, it is of interest whether these therapies would be synergistic in improving glucose utilization in cones, which appears to be critical for survival following rod degeneration.

The Wnt pathway is involved in many developmental and repair functions in mammals. The pathway is complex, with multiple potential ligands binding to co-receptors LRP5/6 and Frizzled, activating gene expression of specific pathways through beta-catenin. Pan-retinal expression of ligand Wnt3a using adenoviral gene therapy partially rescued both morphology and function in the rd10 retinal degeneration model.³¹ In the same study, Wnt activation via constitutively active beta-catenin targeted to Muller glia also improved photoreceptor survival and reduced glial and neuronal remodeling.

One of the more creative approaches to non-specific slowing of disease is the viral-mediated expression of erythropoietin. This growth factor, which is critical to red blood cell development, is also expressed by Mueller cells in the retina, and appears to be cytoprotective in situ. Tao et al.^32,33 expressed erythropoietin under CMV promoter in an AAV-2 vector, demonstrating preserved retinal architecture and preserved functional signaling on multi-electrode array recording ex vivo. (It should be noted this is one of the first studies to employ MEA in the study of rescue of retinal degeneration; this technique may offer more sensitivity and information than classical electroretinography). Interestingly, erythropoietin may be administered intranasally, and appears to protect photoreceptors when delivered by that route in a pharmacologic model of retinal degeneration.³⁴

Finally, analysis of the potential neuroprotectant endothelin-2 suggests unexpected complexity in local and systemic effects of these potential agents.³⁵ Endothelin-2 is a vasoactive peptide initially purified from vascular endothelium. Endothelin-2 knockout mice showed increase survival of photoreceptors in two mouse models of retinal degeneration (rd1 and P347S transgenic). However, the authors of this same study saw rescue of degeneration in eyes treated with AAV5-expressed endothelin-2. This suggests that systemically produced endothelin-2 is deleterious for retinal degeneration, while locally produced endothelin-2 appears neuroprotective.

Immune mechanisms involving activation of microglia have been postulated to contribute to the degeneration seen in outer retinal degeneration.^36,37 Certain cytokines, particularly CXCL3, will counter this activation. Wang et al.³⁸ generated AAV8 constructs with soluble CXCL3, driven by the bestrophin enhancer to force expression in the retinal pigment epithelium, and tested these constructs in rd1, rd10, and Rho^−/− mice. In each case mice expressing this construct showed substantial rescue of cones in the central retina. A modest improvement in ERG sensitivity was also noted in these animals. Nearly identical results have also been observed using AAV vectors expressing isoforms of transforming growth factor beta (TGF-beta).³⁹ Taken together, these results suggest that antagonizing microglial activation using viral-mediated gene transfer may be a general technique for slowing degeneration in multiple models.

In addition to use of gene therapy as a mechanism to overexpress potentially disease-slowing trophic factors, the development of localized gene knock-down or knock-out by gene therapy (for instance through use of guide RNA or CRISPR-CAS9-based mechanisms^40,41) may have application to the slowing of retinal degeneration. For example, Guo et al. have demonstrated that knockout of chemokine receptor 2 (the receptor for monocyte chemoattractant protein 1 [MCP-1] can significantly alleviate photoreceptor cell death in the rd10 mouse model of retinal degeneration;⁴² knockdown of this protein may non-specifically slow degeneration. Similarly, Yu et al. ⁴³ have shown that AAV-mediated knockdown of the Nrl gene via CRISPR/Cas9 improves rod survival and preserves cone function in several mouse models of retinitis pigmentosa including Rho^−/−mice and human transgenic rhodopsin P347S mice. As there are thousands of potential knockdown targets for this approach, the use of a large-scale, inducible-progenitor stem cell (iPSC) retinal organoid screen will be essential to identifying and evaluating optimal candidates.^44,45

Gene therapy for stem-cell based approaches to retinal degeneration

The relationship between stem cell-based approaches and gene therapy approaches to vision restoration and preservation in retinal degeneration is substantial. While review of the former is outside the scope of the present work, there are uses of gene therapy within the stem cell domain that show promise for slowing of retinal degenerative disease or even restoration of lost cells. Jung et al.⁴⁶ introduced CNTF by lentiviral transfection into neuronal stem cells, and then grafted these cells into the vitreous of rd1 and rd10 mice. This approach significantly attenuated photoreceptor degeneration (although function was not measured). One potential advantage of this approach is the ability to optimize expression in an in vitro system prior to introduction. However, such an approach may also face challenges inherent to stem cell introduction into the eye, including degree of integration of graft and potential for tumor formation. Perhaps more intriguing is use of viral gene therapy to reprogram existing cells (particularly Muller glia) into retinal progenitor cells in situ. The proneural factor Achaete-scute homolog 1 (Ascl1 or Mash1) has the ability to do this when transfected into Muller glial cells.^47–49 Similarly, Otx2 expression induces differentiation of rat Muller cell-derived retinal stem cells into photoreceptor fates when transduced ex vivo with lentiviral vector.⁵⁰ There is great potential for combining gene therapy and endogenous stem cell function for in situ retinal repair.

Challenges facing non-specific gene therapy disease amelioration approaches to retinal degeneration

Despite the many successes in animal models in slowing hereditary retinal degeneration via non-specific gene therapy, significant challenges remain in translating this work to clinical use. Many of the gene therapy vectors used in these studies were AAV vectors that required subretinal injection for function. The only gene therapy for ocular use thus far approved by the US Food and Drug Administration (Luxterna ®, Spark Therapeutics) requires subretinal injection for administration. Performing this procedure in the atrophic retina of patients with more advanced retinitis pigmentosa may be challenging. More recent work has attempted to utilize in vivo evolution and capsid selection to create viruses with specific cellular tropism following intravitreal injection.^51,52 Optimization of vectors may also be challenging due to changes in viral tropism during degeneration. Such changes have been documented for lentivirus, for example, comparing light-damaged retina with Rho^−/− animals, where in the latter transduction becomes less efficient with increased degree of degeneration.⁵³ Viral gene therapy approaches are also subject to generation of non-specific and specific immune responses. Adeno-associated viruses are in general circulation, and pre-existing (as well as acquired) antibodies may limit effectiveness of gene therapy in specific individuals.⁵⁴ Perhaps most challenging is proving efficacy in human clinical trials. The hurdles for approval of human gene therapy are substantial; as noted previously, despite this technology having been established in animal models for more than 20 years, only a single ocular gene therapy product has been approved for use in the US to date. As retinitis pigmentosa has a very slow timecourse, demonstration of therapeutic protective effect may take many years. Further complicating the development of these therapies is choice of appropriate endpoint. As noted previously, several growth factor therapies in animal models preserved anatomy without improving physiologic function. A purely anatomical endpoint is unlikely to be recognized by regulatory agencies. As peripheral visual field endpoints have difficulties with reproducibility, and as electroretinography shows a very slow decline in many hereditary retinopathies, development of ultra-sensitive in vivo functional assays that are strongly predictive of clinical course may be critical to advancing this field.

Functional restoration of hereditary degeneration via gene therapy with photoreceptive proteins

The approach of expressing trophic factors via gene therapy is directed at slowing retinal degeneration, but offers little utility for individuals who have become blind due to loss of central photoreceptors. While the gene therapy approaches to endogenous stem cell transdifferentiation have potential for functional restoration, to date this has not been accomplished in an animal model of advanced disease. Three other approaches do offer the prospect of non-specific functional recovery: optoelectronics, small molecule photoswitch agents, and gene therapy using photoreceptive proteins. As the scope of the present review is limited to gene therapy approaches, only the latter will be discussed. The other approaches have been recently reviewed.^55,56

In advanced hereditary retinal degenerations, the rods and cones degenerate nearly completely, but bipolar, amacrine, and retinal ganglion cells remain viable. The fundamental concept of photoreceptor gene therapy is to drive expression of photoreceptive proteins – typically members of the opsin family – in preserved middle- or inner-retinal cells such as bipolar cells or retinal ganglion cells. Expression of these proteins confers light sensitivity upon these cells, restoring light responses to the blind retina. Three classes of photopigments have been utilized in these approaches to date: channel opsins, non-visual (rhabdomeric) opsins, and ciliary opsins. All are members of the 7-transmembrane opsin family, using retinaldehyde as chromophore, but each differs in its properties.

Microbial opsin gene therapy.

Microbial or Type I opsins are native to prokaryotes and unicellcular eukaryotes such as the alga Chlamydomonas. These proteins have a seven transmembrane structure binding an all-trans retinaldehyde chromophore, but unlike opsins found in the animal kingdom, these proteins directly gate ion channels. Following absorption of light (for the widely used Channelrhodopsin 2 [ChR2] molecule, with peak absorption wavelength of 480 nm, in the visible blue spectrum), the all-trans retinaldehyde isomerizes to 13-cis and a conformational change in the protein opens a non-specific cation channel. In physiologic conditions, this results in a net influx of positive ions (principally sodium and hydrogen) with resultant depolarization of the cell. Other variants of channelopsins have been discovered including halorhodopsin, isolated from halobacteria of the Archea kingdom. Halorhodopsin also utilizes a seven-transmembrane structure with retinaldehyde chromophore, but gates an anion (chloride) channel resulting in hyperpolarization following activation, and silencing of the involved cell.

The first use of ChR2 to restore light sensitivity to the degenerated retina was reported by Bi et al. in 2006.⁵⁷ This group put a ChR2-green fluorescent protein fusion into adeno-associated virus under a strong CMV/beta-actin promoter, which resulted in widespread expression in the rd1 inner retina. Patch clamp recording of retinal ganglion cells showed substantial light-dependent currents after viral infection. However, the brightness of light necessary to trigger this response was quite high, with approximately 1 × 10¹⁷ photons/cm²/s of 460 nm light for half-maximal stimulation, which would correspond to extremely bright outdoor illumination levels (although some cells had detectable responses to light 2–3 orders of magnitude dimmer). Further, this group demonstrated projection of infected cells to the lateral geniculate nuclei, and demonstrated light-dependent visual evoked potential in these animals. In follow-up studies, Ivanova et al.⁵⁸ showed that expression of ChR2 was stable for up to 18 months in rd1 mice and that expression of the transgene by AAV2 vector did not appear to be toxic to retinal ganglion cells.

A similar approach has been taken by Lagali et al.⁵⁹ and subsequently Mace et al.⁶⁰, who both targeted ChR2 to on bipolar cells in the mouse specifically using a promoter from the mGluR6 gene, which is expressed exclusively in ON-bipolar cells. Such an approach has the theoretical advantage over ganglion cell-based expression of recapitulating some middle-retinal circuity, perhaps producing more natural ganglion cell firing patterns in response to natural scenes. Histologic staining confirmed the cell-specificity of this approach in both cases. Use of standard ChR2 resulted in sensitivities in the ~7 × 10¹⁶ photons/cm²/s range while Mace et al.’s use of the humanized ChR2 resulted in improved sensitivity, with half of cells showing detectable responses at ~5 × 10¹⁵ photons/cm²/s. Both groups further studied visual evoked potential (VEP) and demonstrated presence of restored VEP with both on- and off- responses to bright blue light. Blind mice treated with this vector showed improved light avoidance to bright blue light (although this experiment was done in mice still expressing melanopsin, which can help mediate this effect particularly in younger animals).⁶¹ A nearly identical approach was taken by Doroudchi et al.,⁶² who employed an AAV8 vector to express ChR2-GFP fusion protein in ON-bipolar cells via the mGluR6 promoter following subretinal injection. This group showed successful rescue of light responses in rd10 mouse retina, with intensities ~5 × 10¹⁶ photons/cm²/s for 50% firing activity. This group also tested this vector in rd1 and rd16 mice and found rescue of visual performance tasks such as water maze (using very bright LED light sources) in both models of retinal degeneration. Further modifications of the AAV8 by targeted mutagenesis (using chimeric AAV2/8) has produced vectors capable of driving bipolar cell expression using non-specific promoters such as chick beta-actin.⁶³

Tomita et al.^64,65 fused an N-terminal fragment of the ChR2 gene to the Venus fluorescent protein and inserted into an AAV2, and injected intravitreally into the eyes of 6-month-old dystrophic RCS (rdy/rdy) rats. Visual evoked potentials were elicited from RCS rats six weeks after injection. Optomotor responses also improved after the AAV2-ChR2V injection. Expression of ChR2V was observed mainly in the retinal ganglion cells. In subsequent studies of the same paradigm, Isago et al.⁶⁶ noted that rescued responses following AAV2-ChR2 rescue were significantly lower in animals treated at 10 months of age compared with those treated at 6 months of age. The authors of this study hypothesized that retinal injury from longer-standing degeneration might be responsible for the lower efficacy of animals treated at older age. This group also investigated the immune responses driven by AAV2-ChR2 expression in rats following subretinal injection.⁶⁷ While antibodies were detected against both the viral vector and the channelopsin after treatment, levels were felt to be too low to establish chronic inflammation, and overt inflammation was not observed after one month following treatment.

Ganjawala, Pan, and colleagues have addressed the issue of low sensitivity of native ChR2 by generating specific directed mutants with higher sensitivity.^68,69 While the original channelopsins used for vision restoration were derived from Chlamydomonas, ChRs from the alga Chloromonas oogama have higher light sensitivity. Patch clamp recordings in human embryonic kidney cells transfected with these channels suggested substantial responses as low as the 10¹⁵ photons/cm²/s range (about 2 orders of magnitude better than the original ChRs), although on- and off-kinetics were somewhat slower, averaging ~0.25 s for on and 0.5 s for off. However, despite these kinetics, retinal ganglion cells expressing these pigments were able to track flicker stimuli as fast as 10–15 Hz. When fused with GFP and injected in an AAV2-derived vector using a non-specific strong promoter, most retinal ganglion cells were successfully transduced. The action spectrum of these channels based on multi-electrode array recordings of these mice revealed peak at 460 nm. One synthetically mutated Chloromonas channelopsin, with three targeted site mutations (H94E, L112C, K264T) had markedly increased sensitivity and showed firing activity of most cells in the 10¹³ photons/cm²/s range. Mice lacking rod, cone, and melanopsin function rescued with these opsins showed substantial recovery of optokinetic reflex responses, with estimated acuities approximately 25–50% of wild-type animals.

Another approach to improving the channelrhodopsins has been the use of chimeras of Volvox and Chlamydomonas channelrhodopsins.70 The former has a much broader action spectrum relative to native ChR2, although its conductance is weaker. The chimera provides the improved spectrum with channel properties closer to ChR2, with responses seen using this single pigment between 460 nm and 640 nm. Further work with modified Volvox channelrhopsin demonstrated that co-injection of this construct with a construct for ChR2 resulted in broad spectrum (blue to red) rescue of light responses in rats with light-dependent outer retinal degeneration.⁷¹

In many forms of retinal degeneration, the rods degenerate first, and cones follow with loss of cone outer segments substantially preceding cone cell death. Busskamp et al.⁷² demonstrated that cones with outer segment loss could be rescued by AAV-mediated halorhodopsin expression. Unlike channelrhodopsins, which gate cation channels, halorhodopsins gate anion channels and cause hyperpolarization of cells following light activation. This recapitulates native photoreceptor function, where light activation of ciliary opsins activates transducing leading to closure of cyclic nucleotide-gated channels and hyperpolarization. This group was able to demonstrate restoration of downstream retinal circuitry including direction selectivity and cortical responses. Interestingly, sensitivity of responses at the level of individual photoreceptors was fairly low (on the order of 10¹⁶ photons/cm²/s) but improved at the level of ganglion cells presumably due to integrated responses. Kamar et al.⁷³ tested halorhodopsin in cultured post-mortem human retinas, which remained viable for about 7 days in culture. Cones in these retinas lost their outer segments, and were successfully optogenetically rescued using a lentivirus expressing halorhodopsin driven by the human arrestin promoter, with recapitulation of normal signaling including horizontal cell responses.

Mammalian opsin gene therapy.

One significant challenge of the channelrhodopsin approach is the lack of signal amplification afforded by these photopigments. Since each photon absorption can only open a single ion channel, dynamic range of restored responses are limited. Animal opsins have evolved to overcome this limitation by working through G-proteins, which allow a marked signaling cascade to amplify signals. It is the G-protein coupled receptor (GPCR) cascade through rod transducin, for example, that allows for macroscopic cell-level responses at the level of single photon absorption.⁷⁴ The inner retina of mammals expresses several opsins natively, including the rhabdomeric photoreceptive molecule melanopsin.⁷⁵ Lin et al.⁷⁶ expressed melanopsin broadly in retinal ganglion cells using an AAV vector, and demonstrated widespread expression. This treatment enhanced pupillary light responses in rd1 mice, and restored behavioral light responses. However, the kinetics of the pigment are very slow, particularly off-times, which are on the order of seconds, which likely limits the utility of melanopsin for functional vision restoration.

In another approach utilizing mammalian opsin gene therapy, Gaub et al.^77,78 expressed mammalian rhodopsin fused to yellow fluorescent protein in an AAV2 vector, using the mGluR6 promoter that drives expression to ON-bipolar cells. Despite the lack of transducin expression outside photoreceptors, this group showed robust ganglion cell firing in response to 510 nm light (peak absorption for rhodopsin) with normalized responses occurring at approximately 10–100 fold less light intensity than for ChR2. (This result highlights the promiscuity of G-protein coupled receptors, where rhodopsin appears to be utilizing another GPCR for signal transduction in bipolar cells). One of the challenges of this approach is that the latency to firing appeared somewhat slow, taking from ~0.1 to 0.4 s after exposure depending on the intensity of the initial light stimulus.

Following the success of this method, Berry et al.⁷⁹ expressed the human middle wavelength cone pigment (M-cone) under the human synapsin promoter (driving expression in retinal ganglion cells), and packaged this in an AAV2 variant (AAV2/2(4YF)) which could be injected intravitreally. This group found light responses of isolated rd1 retinas at irradiances of ~10⁻³ mW/cm² (corresponding to ~10¹² photons/cm²/s) with responses visible to flashes as short as 25 ms, both significant improvements over ChR-mediated responses. This system was able to restore visual responses in these animals, allowing them to discriminate vertical from horizontal green bars presented on a tablet computer screen in a learned shock avoidance paradigm. This group further showed restoration of visually guided exploratory behavior under light conditions comparable to indoor roomlight.

Photoswitch approaches using synthetic receptors.

One of the more interesting approaches utilizing gene therapy for vision restoration is the ‘two component’ photoswitch. Photoswitch compounds are photoisomerizable moieties covalently linked to pharmacologic agents. When coupled to voltage-gated potassium channels, these compounds can induce light-dependent action potential firing in retinal ganglion cells, thus achieving similar ends as opsin-based gene therapy. The development of these compounds is beyond the scope of this review, but have been recently reviewed elsewhere.⁵⁶ Caporale et al.⁸⁰ expressed the modified ionotropic glutamate receptor iGluR6 via AAV2 in rd1 mouse retinas under a human synapsin promoter (called LiGluR). This strongly targeted expression to retinal ganglion cells. The iGluR6 receptor was then triggered using the synthetic photoswitch maleimide-azobenzene-glutamate (MAG). Interestingly, this approach leads to visible light-dependent silencing of retinal ganglion cells (with activation occurring in the near-UV spectrum). These signals were able to restore pupillary light responses in triply mutant rod transducing/cone cyclic nucleotide gated channel/melanopsin mice, and to generate visual evoked potentials in these animals. Further, treated animals were able to navigate a water maze after viral infection and MAG injection.

Further development of this concept utilized a second-generation photoswitch (MAG0₄₆₀) that works on the same synthetic ionotropic glutamate receptor.⁸¹ Compared with the original MAG photoswitch, which had on responses at 380 nm and off responses at 500 nm, the second-generation photoswitch activates at 445 nm, and does not require a second wavelength for reversal, instead terminating spontaneously in dark. When targeted to either retinal ganglion cells on ON-bipolar cells, LiGulR activated with MAG0₄₆₀ resulted in robust blue-light-dependent firing. This combination appeared to restore both innate and learned light-dependent behavior including light-dark shuttle box and water maze in rd1 animals treated with AAV-LiGluR and activated with MAG0₄₆₀. Further, this vector was tested in the rcd1 canine retina in vitro and shown to restore light-dependent firing activity in this larger animal model of retinal degeneration.

Other receptors can be similarly targeted using synthetic photoswitch ligands. While ionotropic glutamate receptors, like microbial opsins, directly gate ion channels, the metabotropic family of glutamate receptors couple to G-proteins. Berry et al.⁸² engineered a metabotropic glutamate receptor based on mGluR2, called SNAG-mGluR2. Expressed under synapsin promoter in AAV2/2, this construct expressed in retinal ganglion cells. The synthetic photoswitch BGAG (similar instruction to MAG) conferred light dependent silencing to retinal ganglion cells under blue light stimulation. Kinetics were fast with responses seen to light pulses as short as 25 ms, leading to ability to track 8 Hz stimuli. Sensitivity was fair with responses seen on multi-electrode array recording at ~10¹⁵ photons/cm²/s. Behaviorally, rescue of rd1 animals with SNAP-mGluR2 led to visual behavioral rescue in a visually-cued shock avoidance paradigm. Perhaps most interestingly, SNAP-mGluR2 with BGAG (with its light-dependent silencing) could be combined with LiGluR with MAG0₄₆₀ with light-dependent activation. Individual retinal ganglion cells would express one or both constructs. rd1 mice infected with this combination interestingly showed both on and off responses reminiscent of native retinal ganglion cell responses.

Clinical development of opsin-based gene therapy.

At the time of writing, three clinical trials using channelrhodopsin for visual recovery are underway. RST-001 (Allergan) is an adenoviral-delivered modified ChR2 under development for treatment of advanced retinitis pigmentosa (clinicaltrials.gov identifier NCT02556736). This Phase I/II study opened in 2015 and Phase I closed to recruitment in 2020. Results have not been presented to date. ChrimsonR is an engineered red-shifted channelrhodopsin that has been shown to restore light responses in retinal organoid culture⁸³, rodents⁸⁴, as well as non-human primates⁸⁵. ChrimsonR has been packaged in an AAV7m8 vector and is also currently in early phase clinical trials along with an adjunctive stimulating device (GenSight, clinicaltrials.gov identifier NCT03326336). Finally, ChronosFP is another algal channelrhodopsin, derived from Stigeoclonium helveticum, with fast kinetics.⁸⁶ AAV-driven expression is being tested for late stage retinitis pigmentosa (Bionic Sight, clinicaltrials.gov identifier NCT04278131). While all three studies are active(and RST-001 has closed enrollment) no results have been published from any of these studies to date.

Challenges to vision restoration using gene therapy.

One of the challenges of gene therapy restorative approaches is that resulting vision will not be native. As noted previously, the light intensities necessary to stimulate optogenetic components are, at best, in the 10¹² photons/cm²/s. Human rod vision has a sensitivity approximately one million times lower. Thus, under mesopic and scotopic conditions, these optogenetic approaches will require the use of adjunctive devices to stimulate the expressed photoreceptor proteins (reviewed in⁸⁷). The dynamic range of human vision is almost 9 logs of intensity, while most optogenetics show 2–3 logs of sensitivity. Thus, dynamic range will also need to be compressed substantially. There are significant issues in human vision for retinal ganglion cell or bipolar cell activation, as the anatomy of the fovea results in displaced bipolar and retinal ganglion cells relative to the cones in their circuit. As such, optogenetic restoration will result in a central scotoma as there will be no photoreceptive cells in the anatomic center of vision. This can be managed with a video device that redirects the central image to the perifoveal inner retinal cells. Color remains a challenge as well; most optogenetic technologies in development have a relatively narrow action spectrum and would be blind to native stimuli outside of the pigment’s sensitivity. Again, an assistive video device will be required to ‘translate’ the spectral palette into an image at peak sensitivity for the opsin (similar to how infrared goggles are used by normal-sighted individuals). The representation of color will also be extremely challenging as, at present, there is no genetic means for distinguishing retinal ganglion cells with different color center-surround fields. It is unclear at present what color percepts will be generated by stimulation of all color channels simultaneously. Finally, the kinetics of optogenetic activation are not identical to native, which may lead to motion-blur or confusion. These issues may also be mitigated by assistive video devices. However development of these is challenging in animal models due to limited behavioral assays, and will be contingent on successfully expressing opsins in retinas of human subjects with retinitis pigmentosa.

A more general challenge in to all gene therapy approaches is the potential for toxicity and immugenicity in the viral vectors. Although AAV-based vectors have passed FDA safety requirements for RPE65 gene therapy for Leber Congenital Amaurosis (Luxterna©, Spark Therapeutics), studies of subretinal AAV in mice suggests substantial toxicity following subretinal injection of vectors with strong, non-specific promoters such as chick beta actin or CMV.⁸⁸ Further, while many studies have suggested low level immune responses to these vectors, some primate preclinical studies have had substantial intraocular inflammation, for both suprachoroidal and intravitreal injection routes.^89,90 Despite the eye being an ‘immune privileged’ site, intravitreal injection can also induce systemic neutralizing antibodies, which may limit prospects for sequential bilateral treatment for patients.⁹⁰ As AAV is known to activate toll-like-receptor 9 (TLR-9); directed engineering of capsids may offer a means for producing less inflammatory vectors.^91,92

Conclusions

The ability to drive expression of specific genes in the eye through virally-mediated gene therapy has opened broad possibilities for treatment of hereditary eye disease, beyond gene replacement therapy. This technology can be used as a personal pharmaceutical factory, allowing the patient to manufacture and deliver specific proteins within the eye. This approach makes pharmacologic treatment of degenerative disease using growth factors or neuroprotectants feasible, where biologically active proteins can be delivered continuously for long periods of time. The ‘proof-of-concept’ for this approach, using a variety of trophic factors in an array of animal models of hereditary retinal degeneration, has been well established as detailed in this review. However, significant challenges still loom. As noted previously, the regulatory hurdles for these treatments will be significant; not least among these is demonstrating a lack of toxicity for a gene product meant to be expressed for years in the eye. Selecting an optimal candidate among the many proteins that appear protective is also non-trivial. Head-to-head comparisons may be challenging; for just the 10 proteins discussed in this review, there are 45 different pairwise comparisons. The experience with CNTF in retinal degeneration is also instructive; anatomic rescue of degeneration may not correlate with significant functional improvement. The use of gene therapy as a way to augment cell function, by expressing light-sensitive proteins in cells that are not normally photoreceptive, is also an extremely promising approach for functional recovery. Again, proof-of-concept has been achieved for multiple photoreceptive proteins in multiple animal models of outer retinal blindness. Significant challenges remain in improving sensitivity and overcoming the non-native information coding conferred by these photoreceptors. Despite this, it appears feasible that the age-old goal of reversing blindness may be achieved by these means in the foreseeable future.