Gene editing is a form of genetic engineering where an organism’s DNA is edited through insertions or deletions. Gene editing technologies have been utilized to create NHP IRD models such as the CRISPR-Cas RP model and achromatopsia model. In the following segments, the mechanisms of gene editing technologies will be explained and the two NHP models will be discussed.
Non-homologous end joining (NHEJ) and homologous recombination (HR) are two major pathways for the repair of double-stranded breaks (DSB) in DNA that were discovered in the late 1900s. These processes are the basis of gene editing and can be exploited
[8]. NHEJ modifies broken DNA ends and joins them up without factoring in any homology, resulting in unwanted sequence insertions or deletions. NHEJ is useful for gene knock-out. Meanwhile, HR relies on an undamaged DNA strand to guide the repair of the DSB, leading to the reformation of a sequence which closely resembles the original. By providing a synthetic DNA template strand, the HR mechanism can be exploited to correct any unwanted sequences, making it useful for editing specific DNA sequences.
While the initial generation of gene editing tools such as Zinc Finger Nucleases (ZFN) and Transcription Activator-Like Effector Nucleases (TALEN) were difficult to design, the emergence of simpler and more predictable CRISPR-based gene editing tools has revolutionized the field. Apart from therapeutic development, these tools have also accelerated the creation of animal models
[9].
Table 1 provides a comparison of gene editing technologies. In the following segment, the mechanisms of these tools will be explained and the IRD NHP models that have been created using these tools will be discussed.
2.1.1. Zinc Finger Nucleases (ZFN)/Transcription Activator-Like Effector Nucleases (TALEN)
ZFN and TALEN were conceptualized from studying the Fok1, Type II (S) restriction enzyme. As compared to other restriction enzymes, Fok1 is unique as it has separate DNA-cleavage and DNA-recognition domains. Most importantly, the DNA-cleavage domain has no specificity and can work independently as long as it is guided to a DNA strand
[10].
In the 1990s, Chandrasegaran et al. showed that the DNA-cleavage could be redirected by substituting the natural DNA-recognition domain with zinc finger (ZF) domains, one of the most common DNA-binding domains in mammals
[11][12]. By combining the Cys
2His
2 ZF, which can bind up to 30 amino acids
[13], to Fok1, many different genetic sequences can be cleaved to allow recombination to occur
[14]. In a ZFN, two ZFN proteins have to be created as Fok1 requires dimerization to function. Each ZF DNA binding domain can recognize 3 bps. By combining multiple ZFs together, a longer nucleotide sequence can be recognized. In general, ZFN are designed to demonstrate around 18–36 bp of specificity.
The next generation of gene editing tools to be developed were TALEN. In nature, TALE proteins activate plant genes to support the virulence of Xanthamonas, a plant pathogenic bacteria. The TALE protein is comprised of three domains: (1) amino-terminal with a transport signal, (2) DNA-binding domain made of 34 repeating sequences of amino acids arranged in tandem and (3) carboxyl-terminal with a nuclear localization signal and transcription activation domain
[15][16]. The DNA-binding domain contains two hypervariable amino acids, which are known as the repeat variable diresidue. These determine the nucleotide-binding specificity of each repeat
[17].
Similar to ZFNs, TALE proteins are fused to Fok1 restriction enzyme to form TALEN. Compared to ZFN, there are specific TALE proteins that recognize only 1 bp rather than 3 bp in the DNA-binding domain. The TALE proteins can be joined together to create a highly specific and modifiable tool to target almost any DNA sequence in the genome. With up to 30 to 40 BP of specificity, TALEN is, in theory, still the most specific gene-editing tool available
[18].
Between the late 1990s and early 2002s, both ZFN and TALEN were rapidly adopted for targeted genetic engineering as they provided an effective method for gene knockout and gene editing. Apart from therapeutic applications, these technologies were also used to generate animal models of various diseases. In the IRD realm, ZFN and TALEN have been used to create small animal models such as mice
[19] and zebrafish
[20][21][22]. However, to date ZFN and TALEN have not been used to create IRD NHP models.
While TALEN provided far superior sequence specificity to ZFN, the use of both technologies still required the use of significant protein engineering methods, which was resource- and time-consuming. Furthermore, TALEN, while providing superior specificity, has a large protein size. Hence, choosing an appropriate vehicle for TALEN delivery has been challenging. These challenges may explain a lack of ZFN- and TALEN-based IRD NHP models.
2.1.2. Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR-Associated (CRISPR/Cas) Methods
CRISPR/Cas systems can be found in almost 90% of all bacteria, and archaea and provide significant immunity against viruses
[23]. These systems are made of arrays of repeated sequences, interspersed by spacers, which are short, 20–50 bp long, non-repetitive DNA segments. The spacers are portions of the viral genome that are added to the CRISPR sequence during infection. These spacer arrays can be transcribed and eventually processed into small CRISPR RNA (crRNA) to recognize invading nucleic acids and mark them for eventual degradation
[24][25]. In summary, CRISPR/Cas system works in three phases: (1) integration of spacer sequences; (2) processing of CRISPR locus transcript and maturation of crRNA; and (3) DNA or RNA interference
[26].
In 2012, when Doudna and Charpentier et al. demonstrated the ability to program the Cas9 system from
Streptococcus pyogenes to function as a RNA-guided DNA endonuclease
[27], several groups, including Feng Zhang et al., continued modifying the system for application in mammalian cells
[28][29]. Today’s CRISPR/Cas systems are extremely simple, with only a single guide RNA (sgRNA) and a Cas protein. The sgRNA is typically 98–100 bp long, with the 5′ end having a protospacer that recognizes the sequence of interest and the 3′ end having a transactivating cRNA (tracrRNA). With this structure, the sgRNA guides the Cas protein to the protospacer adjacent motif (PAM). PAM is a short DNA sequence (2–6 bp long) that follows the DNA region targeted for cleavage by the CRISPR/Cas system. It is generally 3–4 nucleotides downstream from the cut site. Once the Cas protein induces DSB, further gene repair mechanisms will then take place. As compared to ZFN and TALEN, CRISPR/Cas systems do not require complex protein engineering methods. Hence, the design of such systems is simpler and less resource-intensive, making it widely available to many life science laboratories. Its simplicity has also revolutionized animal model development. To date, NHP models for achromatopsia and RP have been generated using CRISPR/Cas technology. These NHP models will be discussed in the subsequent paragraphs.
In a 2021 study by Li et al., the adeno-associated virus (AAV) serotype shH10 was used as a vector to deliver CRISPR/
Staphylococcus Aureus Cas9 (SaCas9) to knock out the
RHO gene in the rod photoreceptors of
Macaca mulatta in vivo, with the aim of generating a macaque model of retinitis pigmentosa. sgRNAs targeting the first exon of the
RHO gene were designed to achieve a high rate of complete gene knockout. SaCas9 was chosen over the conventional Streptococcus pyogenes Cas9 as the former is about 1 kb shorter and would therefore be able to fit within the packaging limitation of AAV (about 4.85 kb), allowing for the construction of both Cas9 and sgRNA into one AAV vector for high co-transduction rate. Each sgRNA under the control of the U6 promoter was individually cloned with SaCas9 under the control of the human synapsin I (hSyn) promoter to drive neuron-specific expression. In vitro, the cleavage efficiency of sgRNA was about 50%. In each test eye, three subretinal injections of AAV/ShH10-hSyn-SaCas9-U6-sgRNA1, 2, 3 plasmids were given, and about 10–20% of the retina was determined to be infected by AAV by immunohistochemistry. Significant indel-existing reads were found at the desired location, suggesting likely dysfunctional production of
RHO proteins. Furthermore, no mutations were detected at potential predicted off-target loci. Morphological studies of the virus-infected Cas9-
RHO retinae showed distinct photoreceptor degeneration, with reduced rhodopsin expression to ~47% that of control retinae, as well as reduced opsin (long-, mid- and short-wavelength) expression to ~27% that of control retinae, suggesting secondary loss of cone photoreceptors. There was complete loss of ONL in the macula after 8 months, indicating progressive photoreceptor degeneration. Furthermore, on FA, hyperfluorescent areas where virus was injected subretinally were seen, suggestive of leakage of retinal telangiectasia. On OCT, the ellipsoid zone was either disrupted or absent, while total retinal thickness and photoreceptor thickness of infected macula and periphery were observed to significantly decrease over time. On transmission electron microscopy, abnormal subcellular structures of infected photoreceptors were seen, with vacuolated mitochondria, shortened and disorganized rod discs, and strong cell apoptosis. In agreement with the morphological findings, ex vivo ERG testing showed significantly decreased photoresponse in infected areas compared to non-infected areas. Taken together, this research provided convincing evidence of the generation of an NHP RP model that closely mimicked class A RP disease in humans, with demonstrable loss of RHO protein, early rod photoreceptor degeneration, thinning retinae and reduced physiological functions
[30].
In 2020, Lin et al. reported the use of AAV9 as a vector to deliver CRISPR/
Streptococcus pyogenes Cas9 (SpCas9) to knock out the
CNGB3 gene in the cone photoreceptors of
Macaca fascicularis in vivo, with the aim of generating a macaque model of achromatopsia. sgRNAs targeting exon 6 of the
CNGB3 gene were designed, and the sgRNA with the highest targeting efficiency was cloned into a AAV9 vector under the control of the U6 promoter. SpCas9 was chosen instead of the smaller SaCas9 as there were more SpCas9 protospacer adjacent motifs in exon 6, and it was driven by the small but less efficient elongation factor promoter in order to fit the AAV packaging limit. Unlike the
RHO knockout research where a single Cas9-
RHO shH10 vector was used, this research used two separate AAV9 vectors in view of the larger SpCas9 size—one for SpCas9 and the other for the sgRNA. A premix of both vectors was subsequently injected subretinally into three separate sites in each test eye, and about 12–14% of the retina was determined to be infected by both immunohistochemistry and single-cell sequencing of isolated cones. This partial knockout of the
CNGB3 gene was demonstrated to result in consistent reduction of mfERG response at D90 post-injection but not overall retinal function as measured by ffERG, suggesting cone dysfunction in central macula is consistent with an achromatopsia phenotype
[31].