CRISPR/Cas9 Genome Surgery For Retinal Diseases

Abstract

Retinal diseases that impair vision can impose heavy physical and emotional burdens on patients’ lives. Currently, clustered regularly interspaced short palindromic repeats (CRISPR) is a prevalent gene-editing tool that can be harnessed to generate disease model organisms for specific retinal diseases, which are useful for elucidating pathophysiology and revealing important links between genetic mutations and phenotypic defects. These retinal disease models are fundamental for testing various therapies and are indispensible for potential future clinical trials. CRISPR-mediated procedures involving CRISPR-associated protein 9 (Cas9) may also be used to edit genome sequences and correct mutations. Thus, if used for future therapies, CRISPR/Cas9 genome surgery could eliminate the need for patients with retinal diseases to undergo repetitive procedures such as drug injections. In this review, we will provide an overview of CRISPR/Cas9, discuss the different types of Cas9, and compare Cas9 to other endonucleases. Furthermore, we will explore the many ways in which researchers are currently utilizing this versatile tool, as CRISPR/Cas9 may have far–reaching effects in the treatment of retinal diseases.

1. Introduction:

Inherited retinal diseases affect approximately 1 in 3,000 people worldwide and include a number of progressive disorders involving photoreceptor degeneration [1,2]. Many retinal diseases ultimately lead to irreversible blindness. Currently, there are no proven cures for patients suffering from retinal diseases, but developments in medical technologies, pharmacology, and gene and regenerative therapies have yielded promising results [3,4].

The retina contains several layers of cells that convert light stimuli into electrical signals through phototransduction [5]. Photoreceptors capture photons of light and convert them into electrical signals [6], which are then amplified and transmitted to bipolar cells, inner neuron cells (i.e. amacrine, horizontal), and ganglion cells. The ganglion cells then transmit these electrical impulses to the brain through the optic nerve [7]. The flow of signal from the photoreceptors to the brain is disrupted as photoreceptors degenerate, often leading to irreversible blindness.

One method of treating recessive retinal diseases lies in gene augmentation therapy, which involves supplementing the patient with a copy of the wild type (WT) gene to allow for diseased cells to create functional protein products [8]. Adeno-associated virus (AAV) vectors are often used to deliver the appropriate gene due to their low toxicity and lack of pathogenicity [1]. Gene augmentation therapy in humans showed promising results in the treatment of Leber congenital amaurosis (LCA) caused by mutations in the retinal pigment epithelium-specific 65-kDa protein (RPE65) gene [9–11]. Researchers demonstrated the long-term survival of the AAV vectors and consistent expression of therapeutic genes in animal disease models [9–11]. On December 19, 2017, the U.S. Food and Drug Administration (FDA) approved the AAV therapy (named “Luxturna,” or voretigene neparvovec-rzyl) to treat LCA in humans after the phase 3 trial demonstrated safety and efficacy [10,12].

Despite its successes as a treatment for LCA, AAV gene augmentation therapy currently lacks the ability to treat dominant diseases, as the addition of an exogenous functional gene via gene therapy is insufficient to override the dysfunctional proteins produced by the dominant mutant allele. This is problematic because there are numerous dominant retinal diseases, including but not limited to autosomal dominant retinitis pigmentosa (adRP), pattern macular dystrophies, and Best vitelliform macular dystrophy [13–15]. For this reason, scientists have turned to alternative methods such as CRISPR/Cas9 genome surgery to overcome this hurdle.

CRISPR/Cas9 can serve as a fundamental tool for translational research on retinal diseases affecting widespread populations, such as retinitis pigmentosa (RP), LCA, and age-related macular degeneration (AMD) [16–18]. CRISPR/Cas9 technology has been used to create knock-in and knock-out animal disease models geared towards developing treatments for inherited retinal diseases. In this review, we will provide an overview of CRISPR/Cas9, discuss the different types of Cas9 being used in research today, and compare CRISPR/Cas9 to transcription activator-like nucleases (TALENs) and Zinc Finger Nucleases (ZFNs). Next, we will explore how CRISPR/Cas9 has been used to study disease mechanisms of retinal diseases as well as its potential for application in future clinical treatments.

2. Overview of the CRISPR/Cas9 System:

2.1. Background of CRISPR/Cas9

The CRISPR/Cas system was first discovered in prokaryotes in the 1990s, and its role in adaptive immunity was further characterized in the 2000s [19–22]. By incorporating a copy of the pathogenic viral DNA into their own DNA sequences, bacteria retain a genetic memory of former viral invaders and utilize their CRISPR/Cas machinery to cleave and destroy DNA of future viral invaders. Cas9 binds to the DNA target site by recognizing a CRISPR RNA (crRNA) bound to a trans-activating crRNA (tracrRNA) strand. The crRNA binds to the complementary DNA sequence, or spacer sequence, which lies next to a G-rich protospacer adjacent motif (PAM) [23,24]. Then Cas9 makes a double-stranded break (DSB) to the target DNA using its HNH and RuvC nuclease domains [23,25,26]. In 2012, Jinek et al. demonstrated that a single guide RNA (sgRNA)—created by joining crRNA and tracrRNA with a hairpin-shaped linker—effectively mimics the tracrRNA:crRNA complex and enables a more efficient and versatile design [27].

2.2. Different Types of Cas9

A number of Cas9 endonucleases have been discovered thus far. Streptococcus pyogenes Cas9 (SpCas9) is a frequently utilized variant of Cas9 spanning 4.2 kb [20,27,28]. Due to its large size, however, researchers have explored alternative Cas9 strains that fit the carrying capacity of AAV (~4.7 kb) [24,28–30]. Ran et al. reported Staphylococcus aureus Cas9 (SaCas9) to have a gene editing efficiency comparable to that of SpCas9, with the added benefit of being over 1kb shorter in length [28]. In 2017, Kim et al. reported Campylobacter jejuni (CjCas9) to be an even shorter Cas9 orthologue (2.95 kb) that demonstrated highly specific in vivo gene editing [30]. In 2015, Zetsche et al. reported a unique single RNA-guided endonuclease, Cpf1, which differs in structure from Cas9 but is capable of facilitating gene-editing in a mammalian cell line [24]. Unlike Cas9, Cpf1 has: 1) a crRNA but no tracrRNA domain 2) a T-rich PAM instead of a G-rich PAM, and 3) a RuvC-like endonuclease that creates a staggered double stranded break and DNA overhang. After analyzing 16 Cpf1-family proteins, authors claimed that the endonucleases from Acidominococcus sp. Bv3L6 and Lachnospiracae bacterium ND2006 display the best efficacy for gene editing in human cells [24].

Catalytically dead Cas9 (dCas9) possesses inactive nuclease domains due to mutations in RuvC and HNH [31]. As such, it employs the DNA targeting and binding function of Cas9 while lacking the ability to create DSBs. Fusion to an effector domain, however, allows dCas9 to affect levels of gene expression [32]. When fused to Krüppel associated box (KRAB), a transcription repressor domain, the dCas9-KRAB complex blocks the transcription of target sequences in a modular and precise fashion [32]. This gene knockdown technique is called CRISPR interference (CRISPRi) because dCas9 blocks RNA Polymerase (RNAP) activity during the initiation of translation and elongation in place of cleaving a targeted site like WT Cas9 [31]. A highly efficacious endonuclease for gene knock down, dCas9 is capable of repressing gene expression in human HEK293 cells [31,33] and exhibits high specificity, reversibility, and an impressive on-target rate [31]. Although dCas9 expands the application of CRISPR/Cas9, it is challenging to use in vivo, requiring two AAVs to account for the size of both dCas9 and effector domain (e.g. KRAB) [34]. The effector is usually fused to an adaptor protein (e.g. Pumilio/FBF) and packaged into an AAV separately from dCas9 [35]. The dCas9 is bound to the effector/adaptor following delivery to the host cell [36] .

In efforts to minimize off-target effects without sacrificing on-target efficiency, researchers have created other Cas9 variations by mutating the protein’s catalytic domains. An aspartate to alanine (D10A) mutation in the RuvC catalytic domain converts Cas9 from a nuclease into a nickase (Cas9n), which produces single-stranded DNA cuts [37]. A histidine to alanine (H840A) mutation in the HNH domain likewise results in a Cas9 nickase, Cas9H840A [27,38–40]. The combined activity of two sgRNAs and two Cas9ns, each of which cleaves a single DNA strand, allows for “double nicking” or “paired nicking,” a phenomenon that results in a double-stranded DNA break [40]. In one study, gene editing with paired Cas9n lowered the off-target mutagenesis rate in human cells by 50 to 1,000 fold compared to WT Cas9 [40].

2.3. DNA Repair: Non-Homologous End Joining (NHEJ) vs. Homology Directed Repair (HDR)

Cells harbor two distinct repair mechanisms in response to DSBs in the DNA. During Homology Directed Repair (HDR), a process that occurs during the S phase of dividing cells, a homologous DNA sequence is used as a template for the reconstruction of broken DNA [41]. Researchers harness this mechanism to edit DNA sequences of interest by introducing homologous donor templates, resulting in the double crossover of engineered exogenous DNA sequences into the endogenous DNA. A single stranded oligodeoxynucleotide (ssODN) template strand can be used to guide the repair of CRISPR/Cas9-initiated DSBs and allow for the insertion or deletion of designated nucleotides. This can be particularly valuable for experiments that seek to correct a mutation or create a specific mutation in a disease model [42].

The other mechanism of DSB repair, Non-Homologous End Joining (NHEJ), can occur in dividing cells as well as in post-mitotic non-dividing cells [41]. Through NHEJ, the two halves of DNA join together—often inaccurately—without the presence of homologous sequences nearby. This machinery frequently leads to random insertions or deletions (indels) of DNA bases, and genes may become nonfunctional or malfunctional (i.e. knocked out) [4].

In mammalian cells, NHEJ has been reported to have a higher efficiency than HDR [43]. In cycling cells, NHEJ repairs 75% of DSBs while HDR repairs the remaining 25% [43]. While this 3:1 ratio of NHEJ to HDR has been a widely accepted phenomenon, Miyaoka et al. in 2016 contrastingly found higher levels of HDR compared to NHEJ when single Cas9 nickases or dual Cas9 nickases were introduced to HEK293T cells [44].Their results suggested that parameters such as gene locus, nuclease type (single nickase Cas9 vs. dual Cas9 nickase vs. TALEN) and cell type (HeLa cells, HEK293T cells, and human iPSCs) have a significant effect on the NHEJ:HDR ratio.

2.4. Risk of Off-Target Effects

When Cas9 makes DSBs at sites other than those targeted by the sgRNA, off-target mutagenesis can occur. Even with mismatches between sgRNA and target DNA, SpCas9 may make DSBs and cause unwanted mutations in off-target locations because SpCas9 can tolerate mismatches depending on: 1) the number of sgRNA-target DNA mismatches, 2) the location of these mismatches within the guide sequence, and 3) the spacing between mismatches within the guide sequence [45]. In 2013, Fu et al. analyzed over 200 candidate sites for off-target mutagenesis in six locations amongst four genes in U2OS:EGFP cells [46]. Candidate sites were chosen if they differed from on-target sites by a range of one to six mismatches. By way of T7 Endonuclease I assay, 24 sites demonstrated a high off-target mutagenesis rate induced by Cas9-mediated NHEJ. Analysis of those candidate sites in other human cells—HEK293 and K562 cells—revealed a high off-target rate in the same 24 sites.

2.5. CRISPR/Cas9 Versus Other Endonucleases (ZFNs and TALENs)

There are currently three predominant tools used for gene-editing: Zinc finger nucleases (ZFNs), transcription activator-like nucleases (TALENs), and CRISPR/Cas9. ZFNs are chimeric restriction endonucleases that contain finger-like structures with one zinc atom per unit of zinc finger module and a FokI restriction enzyme with separate DNA binding and DNA cleavage domains [47–50]. The FokI recognition domains are programmable, so scientists can induce the enzyme to cleave a specific target sequence. Each zinc finger aids in the recognition of the target sequence. After the FokI cleavage domain dimerizes, it makes a double-stranded cut in the DNA [51,52]. Compared to CRISPR, which uses sgRNA to target specific DNA sequences, zinc finger modules are more difficult to design and have sub-optimal affinity and specificity [53]. Furthermore, finding a target site that is amenable to ZFNs can be challenging, as open-source ZFN programs can only target DNA sequences that are 200 base pairs apart [54].

TALENs are also nucleases that can recognize specific DNA sequences and make double-stranded cuts [55]. Similar to zinc fingers, TALEN proteins work in pairs, and each TALEN is comprised of a FokI nuclease and a DNA-binding domain that uses amino acid units (up to 30 repeats of 33–35 amino acids) to recognize DNA nucleotides [4,56]. TALEN pairs are large in size, thus reducing the number of viral vectors that can be used for their delivery into cells [4]. Furthermore, TALEN binding sites must start with a T, which presents an additional limitation to their design [57].

TALENs and CRISPR/Cas9 each present their own advantages and drawbacks. To target a DNA sequence of interest using TALENs, scientists must create two new TALEN genes, which can be difficult and laborious [37]. CRISPR/Cas9, on the other hand, is significantly easier to program, because scientists can simply design and order sgRNA oligos that are specific to the target site. TALENs cleave in a nonspecific fashion in the linker region between TALEN binding sites, whereas Cas9 always induces double-stranded cuts three bp upstream to the PAM [27]. There can be limitations to different PAM configurations, however, which scientists must consider when designing sgRNAs. Moreover, although CRISPR/Cas9 is capable of multiplex genome editing to target multiple loci at once through the co-delivery of multiple sgRNAs [38], this may require the use of dual AAVs due to the limited carrying capacity of the virus [28].

3. CRISPR/Cas9 in Retinal Diseases:

3.1. Using CRISPR/Cas9 for retinal disease modeling

Retinitis pigmentosa (RP) is a form of retinal dystrophy that impairs vision due to the loss of photoreceptor cells, which can lead to irreversible blindness [58]. RP affects 1 in 4,000 live births and is genetically heterogeneous, with 3,000 mutations and at least 60 genes implicated in causing the disease phenotype [59]. Because of the severe heterogeneity of RP, CRISPR/Cas9 is an efficacious and cost-effective tool that can aid in the production and design of animal models for studying retinal diseases (Table 1) [60]. Furthermore, because it is a highly efficient system, CRISPR/Cas9 is also advantageous for retinal disease research that aims to link genotype with phenotype.

Table 1.

CRISPR/Cas9 utilized in retinal disease modeling.

Gene Studied	Experiment	Outcome	Significance	References
Reep6	Knock-in mutation matching RP patient; HDR	Mice recapitulates photoreceptor degeneration seen in RP patient	REEP6 is important in retinal homeostasis; mutations in REEP6 may lead to RP	[61]
Pde6bY347X	Repair Y347X while keeping Xmv-28 intact (both are mutated in Rd1 mice); HDR	Restoration of retinal structure and neural function	Y347X is the causative variant of RP in rd1 mice	[62]
Rp9	Knockout of Rp9 and knock-in of mutation of an RP patient; HDR	Significantly less proliferation and migration of mutated cells than wild-type cells; Fscn2 expression is significantly affected	Mutated RP9 is associated with the expression of Fscn2	[63]
Kcnjl3	Use CRISPR/Cas9 to generate mosaic mouse models for LCA and study function of Kcnj13 ; HDR	Areas of KCNJ13 expression exhibit thinning outer nuclear layer (ONL) and significant rhodopsin mislocalization	Creating mosaic model of LCA using CRISPR/Cas9 can overcome barrier of Kcnj13 homozygous knockout being postnatal lethal and allow for the study of Kcnj13’s potential role in LCA	[65]

CRISPR/Cas9 has frequently been used to model human RP in murine disease models [61]. Using whole-exome sequencing, whole-genome sequencing, and Sanger sequencing, Arno et al. discovered biallelic mutations in Receptor Expression Enhancer Protein 6 (REEP6) in seven patients with RP [61]. To confirm the hypothesis that mutations in REEP6 could account for retinitis pigmentosa, they used CRISPR/Cas9 to create a knock-in mouse model that matched the REEP6 mutation p.Leu135Pro from one of their RP patients. In the Reep6^L135P/L135P homozygous knock-in mice, the significant loss of outer nuclear layers (ONLs) and dysfunction of rod photoreceptors identified by full-field scotopic electroretinography (ERG) matched the phenotype of the REEP6 mutation in patients. This suggests that REEP6 is important in rod-related retinal homeostasis and that mutations in REEP6 may lead to RP.

Wu et al. used CRISPR/Cas9 to show that the Y347X mutation is the causative variant of RP in rodless (rd1) mice, which are the most commonly studied model for RP [62]. There are two mutations identified within the Pde6b gene of Rd1 mice: Y347X nonsense point mutation and Xmv-28 insertion. Using CRISPR/Cas9 to repair the Y347X mutation while keeping the Xmv-28 insertion intact, they observed a functional rescue of the retina. Optical coherence tomography and hematoxylineosin-stained sections demonstrated that CRISPR-corrected mice had restored ONLs. Furthermore, ERG results from CRISPR-corrected mice revealed neurophysiology that closely resembled that of wild-type mice and significantly differed from that of unedited Y347X mutant mice. The identification of Y347X as the causative variant for retinal degeneration provides great translational utility for potential future therapeutic strategies.

In order to uncover the role of the pre-mRNA splicing factor RP9 protein in RP, Lv et al. utilized CRISPR/Cas9 to manipulate the Rp9 gene in a murine cone cell line (661 W cells) [63]. They generated two groups of cone cell lines with Rp9 gene-editing, one with a Rp9 knockout and the other with a Rp9 knock-in point mutation, c.A386T, which is analogous to a point mutation from a RP patient. In both gene-edited cell lines, the proliferation and migration of the cells were significantly less than that of wild-type cells, as respectively identified by MTT cell proliferation assay and in vitro scratch assay. Furthermore, differential gene expression analysis revealed that Fscn2, an RP associated gene, was significantly down-regulated in both mutant cell lines, suggesting that there may be a relationship between Rp9 and RP.

Leber Congenital Amaurosis (LCA) is another retinal disease with extensive heterogeneity and it causes severe vision loss in young children [64]. To study the potential role of KCNJ13 in LCA, Zhong et al. used CRISPR/Cas9 to target the putative start codon of Kcnj13 at exon 2 [65]. Kcnj13 null mice die at P1 and provide no chance for characterization because Kcnj13 may be lethal. To overcome this hurdle, CRISPR/Cas9 constructs were injected into zygotes to generate mosaic mice with different genotypes (N/N, N/W, or W/W where N = null Kcnj13 allele and W = wild-type allele) in varying cells. Tissue staining revealed that areas of Kcnj13 expression exhibited thinning ONLs and significant rhodopsin mislocalization. Thus, the KCNJ13 gene may have a causal role in LCA.

3.2. CRISPR/Cas9 Applications

Gene therapy often involves viral vectors for the delivery of therapeutic genes both in vivo and in vitro [1]. For in vivo delivery into the eye, vectors can be delivered via subretinal injection or intravitreal injection. For subretinal injections, a glass pipette containing the cell suspension is introduced through the scleral opening [66]. The needle must be positioned parallel to the eye wall so that it can reach the subretinal space without puncturing the retina. If the injection is done correctly, a temporary bleb detachment will form. The bleb allows the vector to transduce retinal cells, which subsequently express the delivered therapeutic gene [67]. Intravitreal injection, or injection into the vitreous, is less invasive, but it incurs more inflammation than subretinal injection. As mentioned previously, AAVs are frequently used as the vector of choice for genetic therapeutic editing, and CRISPR/Cas9 can be used to efficiently edit genes of interest for retinal diseases (Table 2).

Table 2.

CRISPR/Cas9 utilized in gene therapy.

Gene Studied	Experiment	Outcome	Significance	References
S334ter	Mutate the S334ter gene in transgenic rats; NHEJ	Mouse models for RP had improved visual function and did not demonstrate retinal degeneration	Potential for use in human clinical trials in the future	[69]
RHO	Edit RHO in vivo by electroporating CRISPR/Cas9 plasmids into transgenic mice carrying mutated human RHO; NHEJ	Transgenic mice demonstrated disrupted P23H-RHO	Demonstrates efficacy and feasibility of studying human genes in murine disease models of RP	[71]
Rho	Selective editing of mutant P23H allele in Rhodopsin; NHEJ	CRISPR-edited Rho+/P23H mice had improved retinal function and slower photoreceptor degeneration compared to controls	Proof of concept that CRISPR can target mutant allele while leaving healthy allele intact; great potential for therapy	[73]
RPGR	Create XLRP patient-specific iPSCs, then use CRISPR/Cas9 to correct the RPGR gene; HDR	iPSCs with corrected RPGR gene	Potential for future iPSC- based transplantation therapies to treat XLRP	[75]
Nrl	Knockout of Nrl in post-mitotic photoreceptors in retinas of 3 separate disease models for rod-specific gene mutations; NHEJ	Affected rods take on cone-associated characteristics; Preservation of Cone function	NRL is associated in rod cell fate even in post-mitotic photoreceptors	[59]
Mertk	Knock-in exon 2 of Mertk in RCS rat model for RP; HITI	Partial rescue of vision; Better preservation of ONL thickness and better ERG responses compared to controls	Earlier intervention time could overcome partial rescue	[76]
VEGF-A	Edit exon 1 of VEGF-A in human RPE cell line ARPE-19; NHEJ	Cell line had insertion- deletions and decreased levels of Vegf-A protein	Proof of concept shows that CRISPR/Cas9 can edit genes involved in retinal diseases more efficiently than other methods (antibodies, decoy proteins)	[79]
Vegfa	Subretinal injection of Vegfa-specific Cas9 into AMD mouse models	Experimental mice demonstrated smaller areas of CNV after laser injury compared to controls	CRISPR/Cas9 has the potential to locally treat degenerative diseases (i.e. AMD)	[77]
VEGFR2	Disrupt VEGFR2 in oxygen induced retinopathy (OIR) and laser-injury-induced choroid neovascularization (CNV) mice; NHEJ	CRISPR-edited mice had less CNV than controls	Disruption of VEGFR2 suppresses neovascularization	[80]
VEGFR2	Compare disruption of VEGFR2 in human retinal microvascular endothelial cells(HRECs) via CRISPR/Cas9 vs. aflibercept or ranibizumab; NHEJ	CRISP R-treated cells demonstrated less VEGF-stimulated activity than did aflibercept or ranibizumab-treated cells	CRISPR/Cas9 has great potential to treat patients with pathologic angiogenesis (i.e. AMD and PDR patients)	[78]
VEGFR2	CRISPR-mediated depletion of VEGFR2 in HRECs to study effects on angiogenesis; NHEJ	CRISPR-edited HRECs demonstrated characteristics of blocked angiogenesis	CRISPR-mediated depletion of VEGFR2 could lead to therapies for patients with AMD and PDR	[18]
CEP290	Creation of cellular model of LCA10 and editing of the gene to fix the splice mutation; HDR	Creation of healthy cell line	Potential to treat patients with cell line of wild-type CEP290	[81]

3.2.1. CRISPR/Cas9 Applications for Retinitis Pigmentosa (RP)

CRISPR/Cas9 has demonstrated the potential to treat autosomal dominant RP (adRP), which is commonly caused by mutations in the rhodopsin gene (RHO) [68]. In a proof-of-principle experiment of in vivo CRISPR/Cas9 gene editing, Bakondi et. al used S334ter rats (Rho^S334/WT) as an adRP model [69]. The dominant S334ter mutation results in a serine substitution at the 334^th amino acid position, which generates an early termination sequence, or stop codon. Consequently, the photoreceptors in S334ter rats degenerate rapidly and progressively, starting 11 days after birth [69]. This murine phenotype mirrors the human adRP phenotype, which is caused by the mistrafficking of rhodopsin [69,70]. S334ter rats were subretinally injected with sgRNA and Cas9 plasmid, designed to target and disrupt the S334ter mutation in the mutant Rho allele (Rho^S334) while leaving the function of the native wild type Rho allele (Rho^WT) unabated. The researchers found that CRISPR/Cas9 indeed selectively ablated Rho^S334, rendering it silent and nonfunctional. Subsequently, this prevented retinal degeneration and improved visual function in the treated rats, as shown through morphology and visual acuity data.

Latella et. al successfully used CRISPR/Cas9 to edit Rhodopsin in vivo by electroporating CRISPR/Cas9 plasmids into transgenic mice carrying a mutated human RHO gene for Pro23His (P23H) [71,72]. The P23H mutation is the most common Rhodopsin mutation to cause adRP [73]. First, they designed sgRNAs that would target and disrupt exon 1 of P23H-RHO by way of indels and frameshift mutations. After confirming their proof of principle in vitro, they injected fourteen P23H RHO transgenic mice with their two sgRNAs constructs and used Sanger sequencing to confirm indels after CRISPR/Cas9 editing. This strategy showcased the efficacy and feasibility of studying human genes in murine disease models. Furthermore, it has great potential for application to future clinical trials, which may involve gene editing for patients with autosomal dominant diseases other than adRP. Most recently, Li et. al used CRISPR/Cas9 to do allele-specific editing in P23H RHO transgenic mice without the limitations of PAM placement, further expanding the feasibility for CRISPR/Cas9 to be used in potential therapies [74].

In a proof-of-concept experiment, Giannelli et al. also selectively targeted the mutant P23H allele in Rho^+/P23H mice while leaving the healthy allele intact [73]. After delivery of CRISPR/Cas9 machinery, Rho^+/P23H mice demonstrated cleavage in the P23H Rho allele but not in the WT Rho allele. This resulted in CRISPR-edited mice with improved retinal functions and slower photoreceptor degeneration compared to controls.

In addition to disrupting mutated genes, CRISPR/Cas9 may also be used to edit and repair mutations to ameliorate disease phenotypes. Bassuk et al. used CRISPR/Cas9 in induced pluripotent stem cells (iPSCs) to repair a genetic mutation that causes X-linked retinitis pigmentosa (XLRP) [75]. They created patient-derived iPSCs from a patient with an RPGR c.3070G>T point mutation, which was corrected using CRISPR/Cas9 with a 13% success rate. As a proof-of-concept experiment, this study is a step forward for CRISPR/Cas9 precision medicine, which may involve patient-derived iPSCs for gene therapy.

Given that RP leads to secondary cone loss from non-cell autonomous effects of rod dysfunction, the preservation of cone cells by ablation of rod fate-determining genes is of great interest. As such, the neural retina leucine zipper (Nrl) gene may be a potential target for treating patients with RP [59]. It has been reported that the ablation of Nrl leads to the loss of rod fate and acquisition of cone characteristics. Furthermore, Nrl ablation consequently improved rod cell survival in the presence of rod-specific gene mutations, presumably preventing secondary cone loss. To prove this hypothesis, Yu et al. utilized CRISPR/Cas9 to knock down Nrl in the post-mitotic photoreceptors of three different murine RP models with rod-specific gene mutations: Rho −/− mice (Rhodopsin null mice), Rd10 mice with a missense mutation in Pde6B, and RHO P347S transgenic mice with a dominant mutation in the human rhodopsin gene. This resulted in morphological and physiological enhancements in all three models, which suggested that rod cells gained cone-like characteristics, cone cells gained enhanced preservation, and vision improved after Nrl knockdown. Thus, Nrl is a feasible candidate for future clinical treatments of RP and may be an alternative to treating individual mutations in different genes.

Suzuki et al. were able to partially rescue the vision of a rat model for RP using homology-independent targeted integration (HITI) [76]. As the name suggests, the process is independent of HDR, relying instead on the NHEJ pathway and possessing the ability to occur in post-mitotic cells. Unlike HDR, which requires long homology arms, the homologous template sequence for HITI is short, spanning only 5–25 bp. The RP rat models were RCS rats with a homozygous deletion in the Mertk gene from intron 1 to exon 2. RCS rat eyes subretinally injected with AAV-rMertk-HITI vector, containing constructs of the missing Mertk exon 2, were compared to RCS rat eyes subretinally injected with AAV-rMertk-HDR vector as controls. The eyes injected with HITI-AAVs demonstrated successful knock-in of exon 2 and significantly higher levels of Mertk mRNA than HDR-AAV injected eyes or non-injected controls, suggesting the efficacy of the HITI strategy. Furthermore, ONL thickness was better preserved in HITI-AAV rats compared to controls. ERG from 4 weeks post-injection exhibited improved ERG b-wave responses and cone responses in HITI-AAV rats, equating to a partial visual rescue. As subretinal injection was performed three weeks after birth, the authors concluded that full visual rescue could be made possible by earlier treatment intervention times.

3.2.2. CRISPR/Cas9 Applications for Age-related Macular Degeneration (AMD) and Proliferative Diabetic Retinopathy (PDR)

CRISPR/Cas9 may also be utilized to decrease levels of vascular endothelial growth factor (VEGF) in future clinical trials involving patients with age-related macular degeneration (AMD) or proliferative diabetic retinopathy (PDR). AMD is the leading cause of blindness in elderly populations of developed countries [77]. Excessive VEGF-A causes AMD by way of angiogenesis, or the formation of new blood vessels. Abnormal angiogenesis is also associated with proliferative diabetic retinopathy (PDR) [78]. Current clinical methods for targeting VEGF-A include antibodies and recombinant decoy receptor proteins [77]. However, these methods are costly and require frequent injections (at least seven times per year). CRISPR/Cas9 would be a beneficial alternative to these methods as it could target the VEGF-A gene and directly reduce the secretion of VEGF-A in human retinal pigment epithelieum (RPE) cells.

Yiu et al. used lentiviruses to deliver CRISPR/Cas9 and sgRNAs targeting exon 1 of VEGFA in the human RPE cell line ARPE-19 [79]. T7 Endonuclease I mismatch detection assays demonstrated successful indel formations for five different sgRNAs, and enzyme-linked immunosorbent assay (ELISA) confirmed decreased levels of VEGF-A protein for the sgRNAs with the highest indel frequencies. Furthermore, in vivo angiogenesis, monitored by human endothelial cell tube formation, was significantly decreased in CRISPR-edited cells compared to controls.

Kim et al. utilized a different approach to reduce the pathological levels of Vegfa [77]. First, they compared the efficiency of plasmid transfection and direct Cas9 delivery in two different cell lines. They found that Cas9 protein delivery yielded a higher efficiency rate than plasmid transfection, because it generated a higher indel mutagenesis rate. Furthermore, they discovered that the labeled Cas9 proteins, when injected subretinally into adult mice, were degraded completely by endogenous proteases 3 days post-injection. Because of their high efficiency and fast in vivo turnover, directly delivered Cas9 protein may be highly advantageous for in vivo applications because the process evades antibody and cell-mediated adaptive immune responses. Using the direct Cas9 delivery method, Kim et al. subretinally injected the Vegfa-specific Cas9 into mouse models of AMD and observed that the experimental mice demonstrated smaller areas of choroidal neovascularization (CNV) after laser injury compared to mouse controls, which were injected with Rosa26-specific Cas9.

Huang et al. tested the effects of CRISPR/Cas9-mediated disruption of VEGFR2 in vivo [80]. VEGFR2 plays a key role in angiogenesis due to its involvement in processes such as VEGF-mediated neovascularization (NV) and microvascular permeability. A CRISPR system was designed to deplete VEGFR2 in two mouse models: oxygen-induced retinopathy mice (OIR) and laser-injury-induced choroid neovascularization (CNV) mice. To create OIR mice, pups are placed in a hyperoxic environment of 75% oxygen for 5 days to significantly decrease vessel number and block vessel growth in the retina. When re-exposed to normal air, the retinas experience vessel regrowth and NV. CRISPR-mediated disruption of VEGFR2 inhibited NV in OIR mice; CRISPR-edited mice demonstrated significantly less retinal NV and more avascular areas than controls. VEGFR2 depletion was also tested in the laser-injury-induced choroid NV model, a model for exudative AMD. Dual AAVs carrying Cas9 and sgRNA were injected intravitreally after laser injury, and CRISPR-edited mice exhibited less CNV than controls. Taken together, these in vivo experiments suggest that disruption of VEGFR2 suppresses NV.

Huang et al. reported that CRISPR/Cas9-mediated disruption of VEGFR2 was more efficient at blocking VEGF-stimulated activity than aflibercept or ranibizumab, drugs that block VEGF from binding to VEGFR2 [78]. After using CRISPR/Cas9 to disrupt VEGFR2 in human retinal microvascular endothelial cells (HRECs), the CRISPR-edited HRECs demonstrated less VEGF-stimulated activation of Akt. Akt is a serine/threonine kinase fundamental to angiogenetic responses such as cell proliferation and migration. The CRISPR-edited cells also exhibited less VEGF-stimulated proliferation, migration, and tube formation than aflibercept or ranibizumab-treated cells.

Using a dual AAV system, Wu et al. used CRISPR/Cas9 to disrupt VEGFR2 and deplete VEGFR2 levels in HRECs to study its effects on angiogenesis [18]. Following infection with rAAV5 vectors carrying SpCas9 and VEGFR2 sgRNA, HRECs demonstrated characteristics of blocked angiogenesis such as inhibition of VEGF-induced activation of Akt. CRISPR-edited HRECs did not proliferate after stimulation by VEGF. In a scratch wound-healing assay, VEGF failed to induce migration in the CRISPR-edited HRECs. CRISPR-edited HRECs also exhibited blocked VEGF-induced tube formation in a collagen-based tube formation assay.

3.2.3. CRISPR/Cas9 Applications for Leber’s Congenital Amaurosis (LCA)

Ruan et al. used CRISPR/Cas9 to sidestep issues with AAV therapy in studying potential treatment methods for LCA10, the most prevalent subtype of Leber’s Congenital Amaurosis (LCA) [81]. LCA10 is frequently caused by an intronic splice mutation in the CEP290 gene. However, the gene is so large that it exceeds the carrying capacity for AAVs in AAV-mediated gene therapy. Therefore, Ruan et al. turned to CRISPR/Cas9 to aid them in their two-pronged study approach. First, they created a cellular disease model of LCA10 with the appropriate splice mutation and found that CRIPSR/Cas9 could successfully ablate the mutation and generate increased levels of wild-type CEP290, thus “rescuing” the mutation. Then, they used a murine disease model and discovered that using a dual system—AAV in tandem with CRISPR/Cas9—could delete pathogenic fragments of Cep290 successfully.

Conclusion:

In this review, we summarized the utility and efficiency of the CRISPR/Cas9 system. Compared to protein-mediated gene editing such as TALEN, which requires the creation of two new TALEN genes, CRISPR/Cas9 is relatively cheaper and simpler to design, as oligos (sgRNAs) are inexpensive and can be easily customized to target genes of interest [54,56,57]. In addition, CRISPR/Cas9 is highly specific due to the precise activity of Cas9, which dependably cleaves 3 bp away from the PAM [27]. New variations of Cas9 such as Cas9n and Cas9H840A have allowed for multiplex gene editing and increased efficiency and versatility, all while decreasing off-targeting effects [38].

CRISPR/Cas9 has rapidly gained prominence as a tool that can be used to model diseases, elucidate links between phenotype and genotype, and repair mutations in vivo and in vitro. Its utility and efficacy has been demonstrated in multiple studies that apply CRISPR/Cas9 to models of RP, LCA, PDR, and AMD. As research expands the applications of CRISPR technology, CRISPR/Cas9 may soon be a viable system that is employed not only as an investigational tool, but also as a clinical therapy to treat retinal diseases.

Highlights:

• CRISPR-Cas9 may be more efficacious and cost effective than TALENs or ZFNs

• CRISPR-Cas9 holds great promise for studying disease models of retinal disorders, helping researchers discover links between genotype and phenotype

• CRISPR-Cas9 can also be used for gene therapy to edit and repair genetic mutations