Correcting visual loss by genetics and prosthetics

Abstract

Retinal degenerations account for the majority of untreatable causes of blindness. Advances in gene delivery vectors, CRISPR/Cas9-based gene editing systems, and electronic engineering have led to a wide range of strategies for correcting visual loss. Here, we provide an overview of retinal gene therapy, gene editing, optogenetics and retinal prostheses using examples from recent clinical trials and pre-clinical studies.

Introduction

Visual loss from retinal degeneration accounts for the majority of untreatable causes of blindness. Retinal degenerations represent a large and heterogenous group of diseases, with genetic factors often playing crucial pathogenic roles. These range from monogenic disorders (such as the inherited retinal dystrophies, including retinitis pigmentosa) to those with proven genetic risk factors (such as age-related macular degeneration). The majority of vision loss in retinal degenerations, whatever the cause, ultimately result from the loss of photoreceptors or retinal pigment epithelial (RPE) cells in the outer retina. The emergence of efficient and safe viral vectors (in particular, the adeno-associated virus, AAV) capable of transducing these two cell types has led to major advances in retinal gene replacement therapy over the last decade. Once the outer retina has degenerated, it is interesting to observe that the inner retinal neural circuity from bipolar cells to retinal ganglion cells (RGCs) often remain intact. This gives rise to further opportunities to salvage vision in advanced disease through direct stimulation of the remaining inner retinal cells either via an optogenetic approach (artificially-induced expression of a photosensitive protein) or by implantation of an electronic retinal prosthesis. In addition to genetic and prosthetic approaches, direct replacement of lost retinal cells by transplantation of human embryonic or induced pluripotent stem cell-derived RPE or photoreceptors is under development, which has been reviewed elsewhere [Zarbin 2019].

Targeted approaches: from gene replacement to gene editing

The recombinant AAV viral vector is a replication-incompetent viral particle consisting of an icosahedral protein capsid encasing a single-stranded DNA genome, which can be a maximum 4.7 kb including the coding gene of interest, transcriptional regulatory elements and inverted terminal repeats (Fig. 1) [Samulski 1982, Laughlin 1983]. Upon transduction of a target mammalian cell, the single-stranded AAV DNA undergoes second strand synthesis and subsequently persists long-term within the host nucleus in the form of monomeric or concatemeric episomes [Afione 1996, Fisher 1997]. Compared with integration into the host genome, episomes are associated with minimal risk of causing oncogenic transformation. The non-pathogenic nature of AAV and the variety of viral serotypes available which can efficiently and stably transduce non-replicating cells (e.g. photoreceptors and RPE) make it a highly attractive vector for retinal gene therapy [Vandenberghe 2013]. Autosomal recessive or X-linked recessive inherited retinal dystrophies arise from loss-of-function mutations in single genes, thus delivering a working copy of the disease-causing gene to the outer retina using an AAV vector forms the basis of many current retinal gene therapy trials (Table 1).

Figure 1.

Therapeutic approaches to treating visual loss by (i) adeno-associated viral (AAV) vector-mediated gene therapy administered via either intravitreal or subretinal injections, or (ii) electronic retinal prosthesis which could be implanted at different anatomical locations (epiretinal, subretinal or suprachoroidal) within the eye. A schematic of an AAV particle is shown with its icosahedral protein capsid encasing a single-stranded DNA genome. ITR, inverted terminal repeat; RGC, retinal ganglion cell; BC, bipolar cell; PR, photoreceptor; RPE, retinal pigment epithelium; CH, choroid.

Table 1.

Current retinal gene therapy clinical trials using adeno-associated viral (AAV) vectors. Note: age-related macular degeneration (AMD) is a complex disease with a range of environmental and genetic risk factors, as well as a variety of animal models that recapitulate some but not all features of the human disease [Pennesi 2012]. RPE, retinal pigment epithelium; PR, photoreceptors; BC, bipolar cells; RGC, retinal ganglion cells; coRPGR, codon-optimised RPGR; CNV, choroidal neovascularisation; sFLT1, splice variant of vascular endothelial growth factor receptor; CFI, complement factor I.

Disease	Disease gene	Target	Animal model	AAV vector	Clinical trial stage
Leber congenital amaurosis 2	RPE65	RPE	RPE65 mutant (Briard) dog	AAV2-RPE65	FDA approved, AAV5 and 4 vectors under investigation
Autosomal recessive retinitis pigmentosa	MERTK	RPE	Mertk -/- mouse	AAV2-MERTK	Phase I
Choroideremia	REP1	RPE	Rep1 -/- mouse (no disease), Rep1 -/- zebrafish	AAV2-REP1	Phase II-III
Achromatopsia	CNGA3	PR	Cnga3 -/- mouse	AAV8-CNGA3	Phase I/II
Achromatopsia	CNGB3	PR	Cngb3 -/- mouse	AAV8-CNGB3	Phase I/II
Autosomal recessive retinitis pigmentosa	PDE6B	PR	Pde6b mutant (rd1) mouse, PDE6B mutant dog	AAV5-PDE6B	Phase I/II
X-linked retinitis pigmentosa	RPGR	PR	Rpgr -/- mouse, XLPRA1 and XLPRA2 dogs	AAV8-coRPGR	Phase II
X-linked retinoschisis	RS1	BC	Rs1 -/- mouse	AAV8-RS	Phase I/II
Leber hereditary optic neuropathy	ND4	RGC	Mutant Nd4 mouse	AAV2-ND4	Phase II-III
Neovascular AMD		RPE	Laser-induced CNV mouse	AAV2-sFLT1	Phase I/II
Atrophic AMD		RPE	A range of mouse models	AAV2-CFI	Phase I/II

Luxturna (voretigene neparvovec) for RPE65-associated Leber congenital amaurosis (an autosomal recessive blinding disease) has recently become the first gene therapy approved by the FDA. RPE65 is a critical enzyme of the visual cycle normally expressed in RPE cells, and its deficiency leads to childhood-onset retinal degeneration and profound blindness. The gene therapy approach consists of subretinal injection of an AAV serotype 2 vector containing the complementary DNA (cDNA) for human RPE65 driven by a constitutive CBA (cytomegalovirus enhancer and chicken beta-actin) promoter. Phase I to III clinical trials for Luxturna have consistently demonstrated its ability to bring about sustained improvements in low luminance vision in otherwise blind patients [Maguire 2009; Bennett 2016; Russell 2017].

Similar gene replacement strategies have been used to treat a number of other inherited retinal dystrophies at various stages of clinical trial. Choroideremia is an X-linked retinal dystrophy caused by loss-of-function mutations in Rab escort protein 1 (REP1). REP1 plays an important role in intracellular vesicular trafficking and its deficiency leads to progressive degeneration of the RPE and photoreceptors, visual field constriction, and blindness in middle age (Fig. 2A). A phase I/II clinical trial of subretinal gene therapy using an AAV2 vector to deliver human REP1 to the outer retina has shown sustained preservation or improvement of visual acuity over 2 to 5 years in patients with advanced disease [MacLaren 2014, Xue 2018]. Unlike pre-clinical testing in mouse models, one of the challenges with gene therapy in patients is that the exact levels of retinal cell transduction and transgene expression in vivo are unknown, therefore it remains to be seen whether REP1 gene replacement is sufficient to halt anatomical degeneration over time and whether or not early intervention (at a young age) might lead to superior outcomes.

Figure 2.

Fundal images of the retina in health and disease. (A) Optomap ultra-widefield fundal image (Optos plc, Dunfermline, UK) of a normal retina. Note inferior eye lash artefact. (B) Optomap retinal image of a patient with choroideremia showing widespread RPE and photoreceptor loss leading to visible choroidal vasculature. (C) Optomap retinal image of a patient with RPGR-associated X-linked retinitis pigmentosa, showing peripheral bone spicules and attenuated blood vessels. (D) Optomap retinal image of an Alpha AMS subretinal electronic implant situated under the macula of a patient with end-stage retinitis pigmentosa.

Specific considerations are needed to modify the gene replacement strategy in some types of inherited retinal dystrophies. For instance, the most common X-linked retinitis pigmentosa is caused by mutations within the Retinitis Pigmentosa GTPase Regulator (RPGR) gene (Fig. 2B). RPGR is found at the photoreceptor connecting cilia where it plays a 'gating' function required for the maintenance of outer segments. The majority of pathogenic mutations in RPGR are concentrated within an alternatively spliced C-terminal exon, called open reading frame 15 (ORF15), which has a highly repetitive GA-rich sequence. In order to produce a stable AAV vector that expresses the photoreceptor-specific splice form of RPGR^ORF15, cryptic splice sites within the cDNA had to be removed and individual codons within ORF15 modified to reduce GA content and repetitiveness without altering the amino acid sequence – a process known as codon optimisation [Fisher 2017, Beltran 2017]. An AAV serotype 8 vector expressing codon-optimised human RPGR which can target photoreceptors is currently in phase II/III clinical trial. Preliminary results have been highly encouraging, indicating that at the optimal therapeutic dose, RPGR gene replacement could lead to reversal of visual field loss in patients with X-linked retinitis pigmentosa [Cehajic-Kapetanovic 2020].

Whilst most retinal gene therapies are currently delivered via subretinal injection, which requires high precision surgery and removal of the vitreous humor [Xue 2016], novel AAV vector serotypes are being engineered to improve their transduction efficacy, tissue tropism, and ability to penetrate the retina following simple intravitreal administration [Dalkara 2013]. Currently, intravitreal administration of AAV vectors works well in rodent models which have a small vitreous cavity and large lens, but are less effective at transducing the retina in primates due to the thicker inner limiting membrane restricting transit of viral particles. In addition, the large volume of vitreous in humans means a high dose of vector is generally required to achieve therapeutic concentrations at the retinal surface. This could significantly increase the risk of intraocular inflammation [Bouquet 2019]. Gene replacement therapies for X-linked retinoschisis (XLRS) and Leber hereditary optic atrophy (LHON) have been delivered via intravitreal injections in clinical trials to target bipolar cells and RGCs of the inner retina, respectively. XLRS arises from deficiency of retinoschisin (RS1), a secreted protein that helps to stabilise the rod-bipolar cell synaptic junction. AAV-mediated augmentation of human RS1 expression in the Rs1-knockout mouse led to improved retinal structure (reduced retinal cavities) and function (electroretinogram response), while early results from a phase I human trial has demonstrated potential closure of parafoveal schisis cavities in at least one patient [Bush 2016, Cukras 2018]. The results of gene replacement in LHON has been more difficult to interpret. LHON poses a unique challenge in retinal gene therapy as the causal gene, NADH-ubiquinone oxidoreductase 4 (ND4), is normally expressed in the mitochondria, where it encodes a subunit of Complex I of the respiratory chain. Since AAV gene therapy can only deliver the genetic payload into the host nucleus, there are currently two alternative strategies for expressing ND4 in the mitochondria: (i) by adding a mitochondrial localisation signal to the ND4 messenger RNA so that nuclear-transcribed mRNA could be translocated to the mitochondria for local translation, or (ii) by adding a mitochondrial localisation signal to the ND4 protein so that nuclear-translated ND4 could be imported into the mitochondria. Both approaches are currently being tested in clinical trials [Guy 2017, Bouquet 2019]. So far, eyes treated with gene therapy for LHON have exhibited little to no improvement in visual acuity. However, interpretation of the treatment effect has been made difficult by the natural variability in the age of onset, rate of progression, level of sight loss, and timing of second eye involvement between individuals with the same genetic mutation [Man 2016].

In contrast to targeted replacement of single gene deficiencies in inherited retinal dystrophies, different strategies will be required to treat more complex retinal degenerations. For instance, age-related macular degeneration (AMD) is the most common cause of sight loss in patients over the age of 65 in the industrialised world. A range of environmental and genetic risk factors have been identified which implicate aging, oxidative stress, lipid metabolism and neuroinflammation in the pathogenesis of AMD [Handa 2019]. One strategy to alleviate the worst type of AMD (wet AMD) in which neovascularisation leads to subretinal bleeding and acute deterioration of vision is to use AAV-mediated gene therapy to express an exogenous secreted anti-vascular endothelial growth factor (anti-VEGF) protein in the retina [Constable 2016]. So far it is unclear whether this surgical approach offers any significant clinical advantage over the standard care of repeated intravitreal anti-VEGF antibody injections. More recently, a novel disease-modifying approach to control complement-driven retinal inflammation in atrophic (dry) AMD by Gyroscope Therapeutics has entered phase I/II clinical trial. It remains to be seen whether AAV-mediated local expression of complement factor I in the RPE cells could protect them from complement-mediated cell death thus slow down dry AMD progression.

Due to the limited cargo capacity of the AAV vector capsid (up to 4.7 kb of DNA), many large transgenes cannot be packaged within a single AAV particle. However, a number of important retinal disease-causing genes could potentially be split into two overlapping upstream and downstream fragments, and delivered using two AAV vectors (e.g. ABCA4 [8.9kb] in Stargardt maculopathy and MYO7A [6.5kb] in Usher syndrome type 1B). Dual vector gene therapy is intrinsically less efficient compared with single vector therapy as it requires successful co-transduction and recombination of two gene fragments within the target cell. In spite of this, the protein expression levels achievable may be sufficient to slow down or stop disease progression [Trapani 2015, McClements 2019].

A major limitation of simple gene replacement or augmentation is that it would not be able to treat autosomal dominant retinal dystrophies in which the mutation on a single allele causes a toxic gain of function. Approximately 50% of retinitis pigmentosa are autosomal dominant and mutations in rhodopsin (RHO), the light-sensitive G-protein-coupled receptor (GPCR) present in rods, account for 30-40% of these. A promising approach to gene therapy for RHO-associated autosomal dominant retinitis pigmentosa is to use a 'block-and-replace' strategy. Cideciyan et al. constructed an AAV vectors which delivers a short hairpin RNA (shRNA) that silences endogenous RHO gene expression while introducing a codon-modified version of wild-type RHO that is resistant to the RNA interference [Cideciyan 2018]. This gene therapy vector has been tested in a canine model of RHO-associated retinitis pigmentosa, which showed preservation of retinal structure and function up to 8-months follow-up.

An alternative approach to treating autosomal dominant retinal dystrophies is to use a clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 gene editing system. The CRISPR guide RNA (gRNA) can direct a Cas9 endonuclease to a highly specific DNA sequence within the genome adjacent to a protospacer adjacent motif (PAM) where it will create a double-strand break. Repair of the DNA break by non-homologous end joining leads to insertions or deletions that would disrupt the target gene. By packaging the CRISPR gRNA and Cas9 enzyme-encoding transgene within AAV and delivering the gene editing vector into the subretinal space, it is possible to specifically disrupt the disease-causing allele in the photoreceptors or RPE [Suzuki 2016; Yu 2017; McCullough 2019]. Furthermore, modification of the CRISPR/Cas9 components, for instance by fusing an enzymatically inactive Cas9 endonuclease to a transcriptional repressor, could allow specific knockdown of the disease-causing gene expression without creating any permanent change to the genome. Such a system has been used to transcriptionally inactivate the Nrl gene, a master regulator of rod fate determination. As a result, the rods became reprogrammed into cone-like cells, rendering them relatively resistant to retinitis pigmentosa-specific mutations [Moreno 2018]. With a rapidly expanding array of CRISPR genome editing molecular tools, many new therapeutic possibilities are emerging, including precise DNA base editing (to correct mutations or alter splicing), RNA editing (to correct errors within mRNA), and targeted DNA integration (to introduce large gene fragments) [Doudna 2020]. While gene editing technology holds enormous potential to revolutionise the field of gene therapy, further work is needed to investigate its potential off-target effects, long-term clinical safety, and ethical implications.

Generic approaches: from optogenetics to retinal prostheses

Over 200 retinal disease-causing genes have been mapped so far (https://sph.uth.edu/RETNET/), and developing tailored genetic therapies for each condition would represent a mammoth task. In contrast to disease-specific gene therapies, optogenetics offers the potential to restore visual function broadly across many types of retinal degenerations, even when there is advanced photoreceptor loss. The optogenetic approach aims to restore visual function by introducing a light sensitive protein into the surviving non-light sensitive cells of the inner retina (i.e. bipolar cells or RGCs). Several candidate proteins are being explored, ranging from microbial-derived opsins (e.g. channelrhodopsin and halorhodopsin) to mammalian opsins (e.g. melanopsin, rhodopsin or medium-wave cone opsin) and engineered opsins (e.g. light-gated ionotropic glutamate receptor, LiGluR) [Lin 2008; Caporale 2011; Cehajic-Kapetanovic 2015; De Silva 2017; Berry 2019]. These proteins can be broadly classified into light-sensitive ion channels or GPCRs, which have pros and cons in terms of their light sensitivity and response speed. Pre-clinical trials of optogenic therapies in murine models of retinitis pigmentosa have provided proof-of-principle in restoring basic visual function to blind animals, and human clinical trials are currently underway.

As an alternative to biological modification of inner retinal cells in advanced retinal degeneration, several types of retinal prostheses have been developed to provide direct electrical stimulation to the inner retina, thus bypassing the need for photoreceptors. The basic requirements for a retinal prosthesis are the abilities to receive and convert light input into electrical output, to be safe and durable for intraocular implantation, and to provide useful visual function. Numerous prosthetic devices and their iterations have been tested over the past decade of which three examples will be discussed to illustrate the main technical approaches. The Argus II epiretinal prosthesis (Second Sight Medical Products Inc, Sylmar, CA, USA) is the first device to receive CE mark and FDA approval. It consists of an external spectacle-mounted camera linked to a portable image processor and communication coil. This coil transmits radiofrequency signal to an internal receiver coil implanted over the surface of the sclera. The signal is decoded and sent via a transscleral cable to an intraocular retinal stimulator (an array of 60 electrodes spanning ~20 degrees of visual field) which is pinned to the retinal surface. A multicentre clinical trial of Argus II in patients with end-stage retinitis pigmentosa showed that the device enables basic visual tasks, such as grating visual acuity, high contrast object localisation and detection of direction of motion. The durability of the device generally exceeded 5 years and the best reported grating visual acuity is around 1.8 logMAR equivalent (6/379 Snellen) [da Cruz 2016; Schaffrath 2019].

The Alpha AMS (Retinal Implant AG, Reutlingen, Germany) is a subretinal implant that received CE mark in 2016. It consists of a 1600 photodiode amplifier-electrode array (spanning ~15 degrees of field), which is inserted into the subretinal space where it acts both as light sensor and electric stimulator. The array is connected to a polyimide foil that exits the eye via the sclera and powered by an electromagnetic induction coil recessed into the retroauricular skull bone, similar to a cochlear implant. Advantages of this design include ‘physiological’ location of the photodiode array within the subretinal space thus placing it in direct contact with the surviving bipolar cells, lack of any unsightly spectacle camera, and large number of electrodes providing potential for higher resolution. The chief disadvantage is the long surgical implantation time (approximately 8 hours). The best documented grating visual acuity with Alpha AMS has been equivalent to around 0.95 logMAR (6/60 Snellen), although significant variations exist over time and between patients [Stingl 2017; Edwards 2018].

Thirdly, a suprachoroidal retinal prosthesis has been devised by the Bionic Vision Australia team. In the first generation device, an array of 22 stimulating electrodes within a silicone substrate implanted in the suprachoroidal space is connected to a percutaneous connector behind the ear via a helical wire. The connector allows the implant to be plugged into an external spectacle-mounted camera and power supply. The suprachoroidal location of the implant allows continued function of any residual photoreceptors while enabling concomitant electrical stimulation, and takes advantage of choroidal atrophy in retinitis pigmentosa to maintain reasonable proximity between the prosthesis and inner retinal cells. The surgery is also simpler than a subretinal implant, taking between 3 to 4 hours. Clinical trial of the 22 electrode device achieved good safety profile and Landolt-C visual acuity equivalent of around 2.62 logMAR (20/8397 Snellen) [Ayton 2014]. A second generation device with 44 electrodes is currently under clinical trial.

Conclusions

In summary, a wide range targeted and generic approaches to treating visual loss from retinal degenerations are under development. These have been driven by advances in genetic diagnosis (through next generation sequencing), AAV vector technology, and electronic/computer engineering. While gene replacement therapies are proving efficacious for a number of recessive inherited retinal dystrophies, CRISPR/Cas9-based gene editing is likely to see increasing application in the treatment of autosomal dominant diseases. Characterisation of the large number of retinal disease-causing genes is improving our understanding of retinal physiology while offering the potential for personalised therapies for visual loss.