Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) sequences were firstly found in bacteria (Ishino et al., 1987) and definitely designation in 2002 (Jansen et al., 2002). However, scientists confirmed that CRISPR sequence functions to benefit bacteria to resist viral infections in 2007 (Barrangou and Horvath, 2017). The CRISPR family contains two main types of systems, which includes several Cas proteins. The first type of system is usually found in bacteria and archaea, such as I, III, and IV Cas proteins, which function by forming multi-subunit protein-crRNA (CRISPR RNA) effector complex. The second type of system contains II, V, and VI types, which could perform target editing relying on a single crRNA-guided protein, i.e., a single multidomain protein that exercises function (Makarova et al., 2011). Herein, the second CRISPR system is more convenient to carry out the gene editing. The detail reported Cas proteins in the second system is listed in Table 1. CRISPR Cas9 is the earliest, most widespread, and the most mature technology among second CRISPR system (Makarova et al., 2017). The CRISPR Cas9 technology includes a small guide RNA (sgRNA) used to target the desired DNA molecule and Cas9 protein, a non-specific CRISPR nuclease which could cleave double-stranded DNA molecules. The Cas9 protein has two cleavage-active domains: the HNH which cleaves the DNA strand complementary to the crRNA, and RuvC domains which cleaves the non-complementary strand (Lu et al., 2021). sgRNA is composed of a trans-activating crRNA (tracrRNA) sequence that can bind to the Cas9 protein and a crRNA containing a specific sequence with about 20 nt in length that is complementary to the target sequence, and the remaining sequence of crRNA complementary to tracrRNA. Therefore, this technology searches for the motif sequence complementary to the crRNA of gRNA in the target DNA molecule with a PAM sequence behind it. Subsequently, Cas9 protein cleaves this motif and results in the formation of double-strand breaks (DSB). Afterwards, knock-out and knock-in occur via non-homologous end joining and homologous recombination in the presence of the exogenous donor sequence in cell (Figure 1A), respectively (Kaulich et al., 2015).

CRISPR Cas9 and related CRISPR Cas systems have been continuously optimized since their emergence, with three main areas: (1) achieving more precise single base editing, (2) reducing off-target effects, and (3) improving editing efficiency. As for the optimization of its delivery approach, the main focus is to find smaller and more portable Cas proteins, which can be effectively packaged using related lentiviral or AAV vectors and more conducive to entering into cells. To reduce off-target efficiency, scientists found that converting one of the two key amino residues in RuvC I of RuvC to alanine (D10A or H840A) could produce Cas9 nickase (Cas9n). This nickase can only cleave the DNA chain that is complementary to the crRNA and cannot cleave the non-complementary DNA chain, thus reducing off-target effects and maintaining high-efficiency gene editing. Furthermore, if the RuvC catalytic domain is mutated by point mutation (D10A) and the HNH catalytic domain is mutated by point mutation (H840A), the Cas9 protein will completely lose its nuclease activity and form dead Cas9 (dCas9), which can only bind to the target gene under the guidance of sgRNA (Qi et al., 2013).

Based on the above findings, Liu et al. fused the cytidine deaminase with nCas9 or dCas9 protein, successfully converting cytidine C to uridine U, and achieved C-to-A and corresponding G-to-T conversion in DNA replication, which was termed as cytidine base editor (CBE). Later, CBE was continuously optimized to enable efficient and permanent C:G to T:A base pair conversion in bacteria, yeast, plants, zebrafish, mammalian cells, mice, and even human embryos (Komor et al., 2016, 2017; Kim et al., 2017). Likewise, by fusing the modified adenine deaminase with nCas9 protein, the system is capable of converting adenine (A) to inosine (I) on ssDNA, and inosine (I) is recognized and read as guanine (G) during DNA replication, ultimately achieving direct substitution of A: T to G:C base pairs. This system is termed as adenine base editor (ABE) (Gaudelli et al., 2017; Komor et al., 2018; Figure 1B). The application of combination of CBE and ABE can effectively perform transition of four types of bases, whereas it is still unable to achieve transversion of the other eight types of bases as well as bases insertion and deletion. Liu team then developed the prime editor (PE) which can achieve change of 12 types situation and precise insertion (up to 44 bp) and deletion (up to 80 bp) of bases without relying on DSB and donor sequence (Anzalone et al., 2019; Figure 1B). In this system, a pegRNA was added to the 3′ end of the sgRNA, which can complement the broken target DNA 3′ end to initiate the reverse transcription. It also contains target point mutations or insertion–deletion mutations to achieve precise base editing (Figure 1B). Later, many studies innovated and optimized single-base editing, and three teams successfully achieved simultaneous CBE and ABE in a single gene editing, which develop a new double-base gene editor (Grünewald et al., 2020; Sakata et al., 2020; Zhang et al., 2020). Yang et al. recently developed a new adenine base editor (AYBE), which can firstly achieve the transversion of adenine (Tong et al., 2023b).

The spin-off technologies of the CRISPR system are not only focused on the optimization of editing DNA bases, but in the regulation of gene expression without changing the genome sequence. DNA editing directly modifies genome sequence permanently and has potential off-target probability, which poses serious clinical application risks. While the regulation of gene expression is usually mild and reversible, and its application in disease treatment can also make up for the shortcomings of DNA editing. The existing CRISPR systems for regulating gene expression include RNA editing systems of Cas13-related proteins (Table 1) and DNA modification regulation systems in which dCas9 protein is coupled with various regulatory factors (Figure 1C). Binding dCas9 to the transcription start site (TSS) of a gene can block the start of transcription, thereby inhibiting gene expression (Figure 1C). Binding dCas9 coupled with transcriptional suppressors or activators can inhibit or activate downstream target gene transcription, respectively (Sen and Thummer, 2022). Fusing some methyltransferases and demethylases to dCas9 protein can selectively regulate the DNA methylation of gene promoter, regulating gene expression (Figure 1C). Cas 13 family is an RNA-dependent RNA endonuclease (Table 1). It can specifically cleave the target RNA to inhibit gene expression in the presence of a PFS sequence in target RNA. Cas13 proteins mainly includes four subtypes, namely CRISPR-Cas13a, b, c, d (Abudayyeh et al., 2017). Among them, CasRx (RfxCas13d) has received widespread attention for its higher efficiency, lower off-target rate. More importantly, Cas13d has a smaller size compared to other family members, which can be more easily packaged into viral vectors, thus having better delivery advantages and application prospects (Kushawah et al., 2020). Cas13d has been reported to achieve effective gene silencing in mouse liver (He et al., 2020). However, research has also found that the CRISPR-Cas13 editing system has significant collateral degradation effects while cleaving RNA sequences, that is, after Cas13 binds to the target RNA and is activated, it can degrade bystander RNAs to some extent. To address this problem, Yang et al. designed a dual fluorescence reporting system to detect collateral effects in mammalian cells and seek for Cas13 variants (Tong et al., 2023a). The team found that the mutated Cas13 variants Cas13d-N2V8 and Cas13X-M17YY had virtually eliminated collateral effects when editing RNA. In addition, researchers have also discovered two new proteins in the Cas13 protein family, termed as Cas13x and Cas13y, and developed a new RNA editing technology based on them. CRISPR Cas13x and CRISPR Cas13y have stronger knockdown activity than previous Cas13a and Cas13b system, and some derived truncated types can achieve RNA single-base editing after fusion with RNA single-base enzymes (Xu et al., 2021; Table 1). Yang et al. also developed a new CRISPR-Cas12f system (enOsCas12f1 and enRhCas12f1) in mammalian cells with high activity, broad targeting range, and high fidelity. After removing the cleavage activity, denOsCas12f1 is an epigenetic editor and gene expression activator (Kong et al., 2023), which shows strong regulatory activity in mammalian cells.

Some studies have constructed models by induced pluripotent stem cells (iPSCs) from somatic cells of PD patients with clear genetic mutations that cause the disease. The iPSCs were differentiated into neural stem cells which can simulate the relevant phenotypes of neuronal cells in the brain of PD patients. Mutations in some genes have been found in both familial and sporadic PD cases, such as the LRRK2, SNCA, and PARKIN genes (Satake et al., 2009; Simón-Sánchez et al., 2009; Nalls et al., 2011). Therefore, using cell models with these genetic mutations is more likely to reveal the pathogenesis of PD. Researchers found that neural stem cells differentiated from iPSC cells of patient with LRRK2 p.G2019S mutation showed increased susceptibility to proteasome stress, and had transgenerational defects in nuclear envelope organization, clonal expansion, and neuronal differentiation (Liu et al., 2012). Subsequently, using knock-in technology to change the point mutation to wild-type, the above phenotype can be reversed, which proposed that nuclear morphology alteration as a clinical diagnostic feature of PD. Further research found that LRRK2 forms a complex with a mitochondrial outer membrane protein Miro in iPSC cells, which then removes Miro. Once the removal of Miro is affected in LRRK2 p.G2019S mutation cells, it will delay mitochondrial autophagy, impairing cell respiration and metabolism (Hsieh et al., 2016). Mitochondrial dysfunction is a key aspect of Parkinson’s disease, and mutations in the PRKN/PARK2 gene have been reported to be associated with early-onset familial PD. This gene can regulate mitochondrial function and autophagy processes (Narendra et al., 2008). Researchers edited iPSCs cells from two PD patients with PARK2 gene mutations and found that neurons differentiated from iPSCs showed increased oxidative stress and enhanced Nrf2 pathway activity. In addition, neural cells showed abnormal mitochondrial morphology and impaired mitochondrial homeostasis (Imaizumi et al., 2012). In another study, scientists identified multiple pathogenic mutations in UQCRC1 in both familial and sporadic PD patients. After knock-in of these mutations in human dopaminergic SH-SY5Y cell lines, axonal degeneration and mitochondrial respiratory chain dysfunction were found (Lin et al., 2020).

PINK1 and Parkin form a central signaling axis that plays an important role in controlling the mitochondrial autophagy process in dopaminergic neurons. In pink1 null Drosophila, knocking down UCHL1 gene using RNAi can rescue the PD-related pathogenesis. Specifically, the loss of UCH deubiquitination promotes mitochondrial autophagy by activating the expression of AMPK and ULK1 (Ham et al., 2021). In Drosophila, researchers used CRISPR-Cas9 to knock out tango14, the homologous gene of NUS1, and found that the lifespan of Drosophila was shortened, and cholesterol accumulation appeared in dopaminergic neurons, which confirms that this gene is associated with the occurrence of PD due to lipid metabolism abnormalities (Xue et al., 2022).

In zebrafish, researchers constructed a dj-1 null strain targeting exon 1 using CRISPR Cas9. It was found that dj-1^−/− zebrafish developed normally in the early stages, but showed lower levels of tyrosine hydroxylase, skeletal muscle respiratory failure, and lower body weight as development progressed (Edson et al., 2019). Proteomic analysis of the brains from dj-1^−/− zebrafish revealed downregulation of proteins related to mitochondrial metabolism, autophagy, stress response, redox regulation, and inflammatory response. Then, researchers developed a new, unbiased computational method to classify the movement disorders of adult dj-1^−/− zebrafish (Hughes et al., 2020). Mitochondrial calcium uniporter (MCU) participates in mitochondrial dysfunction and cell death caused by excitotoxicity, inflammation, and oxidative stress by regulating mitochondrial calcium uptake, and plays an important role in PD (Liao et al., 2017). Soman et al. generated a zebrafish model with double knockout of mcu and pink1 using the CRISPR-Cas9 system. Compared with pink1^−/− zebrafish, (pink1; mcu)^−/− zebrafish showed a higher number of dopaminergic neurons and a higher mitochondrial membrane potential. In addition, mitochondrial sphericity was restored, and animals were protected from PD MPTP neurotoxin. Therefore, mcu may be an effective target for the treatment of PD (Soman et al., 2019).

Mutations in GTP cyclohydrolase 1 (GCH1) may lead to the development of PD. In zebrafish, after knocking out gch1 using CRISPR Cas9 technology, zebrafish showed monoamine neurotransmitter defects, movement defects at 8 days post fertilization (dpf), and death at 12 dpf. Tyrosine hydroxylase (Th) protein was up-regulated, but there was no loss of dopaminergic (DA) neurons (Larbalestier et al., 2022). ATP12A2 is an autosomal recessive pathogenic gene for juvenile PD, also known as Kufor-Rakeb syndrome. Atp13a2^−/− zebrafish were also established by CRISPR Cas9 gene editing. The number of TH⁺ neurons in the posterior lobe and the locus coeruleus of atp13a2^−/−zebrafish was significantly reduced, indicating dopaminergic neuron degeneration, lysosomal dysfunction, and intracellular transport disorders (Nyuzuki et al., 2020). PARL, which encodes presenilin-associated rhomboid-like protein (PARL), has been found to contribute to mitochondrial morphology, function and is associated with familial PD. PARL is mitochondrial inner membrane protease that acts on many mitochondrial proteins involved in mitochondrial morphology, apoptosis, and mitophagy. After knocking out this gene in zebrafish using CRISPR Cas9, dopamine neurons were lost in the brain, and tyrosine hydroxylase transcription levels decreased, leading to impaired olfaction and decreased motor parameters in zebrafish (Merhi et al., 2021).

Mouse has been widely used to construct models related to PD. Researchers have generated a model called MCI-Park model in mice by CRISPR Cas9 to produce dopamine neurons that lack NDUSF2, which encodes mitochondrial complex I. Mice lacking NDUSF2 showed neurodegenerative changes (González-Rodríguez et al., 2021). Point mutations in the vacuolar protein sorting 35 gene (VPS35) are associated with an autosomal dominant late-onset PD (PARK17). Homozygous deletion of Vps35 by CRISPR Cas9 resulted in survival disadvantage and significantly reduced DA release in the caudate putamen of adult homozygous Vps35 mutant mice (Ishizu et al., 2016). Cyclin-dependent kinase 5 (CDK5) negatively regulates dopamine signaling in the striatum, playing a critical role in circadian rhythm disruption and sleep disorders. After knocking out Cdk5 using CRISPR Cas9, mice showed defects in motor activity and interference with activity/rest behavior, and the number of dendritic length and functional synapses in the mouse brain were down-regulated (Zhou et al., 2022). Peroxidized phospholipids caused by ferroptosis have been reported to be associated with the onset of some PD cases. PNPLA9, as a hydrolytic enzyme, can preferentially hydrolyze peroxidized phospholipids. In mice, knocking out Pnpla9 resulted in progressive PD motor disorders and accumulation of peroxidized phospholipids (Sun et al., 2021). Although gene editing of several genes can simulate some symptoms of PD, editing classic genes related to PD has not been successful in mice. Even if simultaneously knocking out three hot genes related to PD, including Prkn/Pink1/Dj-1, no neurodegenerative phenotype was observed in mice, even in older mice (Kitada et al., 2009). This suggests that these hot genes related to PD may have functional differences in the nervous systems between humans and mice.

In spite of the fact that gene editing in small rodents is convenient, it is difficult to reproduce significant neurodegeneration including loss of neurons as PD patients (Deng and Siddique, 2000; Deng et al., 2018). Therefore, it is not conductive to using mice to find therapeutic methods for PD. Recently, some large animal models of PD have been successfully constructed (Yang W. et al., 2021). These animal models are able to better simulate the phenotype of human PD, providing new insights into the pathogenesis of PD. Although the emotional and cognitive abilities of pigs are not as similar to those between monkeys and humans, their brain structure is roughly similar to that of humans. In addition, pigs reach sexual maturity (around 6 months) earlier than monkeys, have a shorter gestation period (about 4 months), and give birth to over 10 offspring in one litter. Endogenous genes in pigs can be edited through CRISPR Cas9 combined with somatic cell nuclear transfer (SCNT) to produce knock-out or knock-in models. All of these advantages make pigs a promising alternative large animal model for investigating human diseases (Lunney et al., 2021). Researchers used CRISPR Cas9 and SCNT technology to establish a Bama miniature pig models with Snca missense mutations (p.E46K, p.H50Q, p.G51D). However, PD-specific phenotype was not present such as immunopositivity for SNCA and loss of dopamine neurons in the substantia nigra (Zhu et al., 2018). TALEN combined with SNCA was utilized to construct Park7 knock-out pigs, these pigs were all dead after birth and DJ-1 protein was found significantly inhibited in various tissues (Yao et al., 2014). Another study used CRISPR Cas9 combined with SCNT to prepare Park2 and Pink1 double knock-out pigs, however, like the mouse model, the double knock-out pigs also showed no clinical signs of PD at 7 months old (Zhou et al., 2015). Wang et al. (2016) generated triple knock-out (Parkin, Dj-1, and Pink1) Bama miniature pigs, but the piglets did not exhibit the PD clinical phenotypes at 10 months of age. Considering the higher similarity of brain structure between monkey and human, gradual studies have been conducted using monkey as PD models in recent years. Li et al. performed the first knock-out of pink1 in monkey and found that knocking out pink1 gene resulted in part of monkey died and remaining monkeys displayed severe degeneration and death of neural cells in the brain (Yang et al., 2019a,b). Subsequently, researchers found Pink1 genes was expressed at a very low level in mice, whereas at a high level in primate brain tissues, suggesting it is unique to human or primates. This may explain the reason behind the inconsistent results of gene editing of Pink1 in mouse and monkey. Moreover, Pink1 was found not aggregating in mitochondria, but rather in the cytoplasm. Under non-stressful conditions, Pink1 functions to phosphorylate relevant neuronal proteins in the brain, protecting neurons from damage (Yang et al., 2022). It challenged the previous notion that Pink1 mutation causes PD due to mitochondrial dysfunction and autophagy blockade, suggesting the importance of developing drugs targeting protein phosphorylation in treating PD due to Pink1 mutation. Chen et al. developed a modified and optimized CRISPR Cas9n to knock out Pink1 gene in embryonic cells of Cynomolgus monkey. But the monkeys did not show a phenotype of neurodegenerative disease (Chen et al., 2021). Nevertheless, this study did not detect the expression of Pink1 in the brain, so it is not clear whether the lack of effect is due to editing heterozygosity. In another research, stereotactic injection of AAV to deliver CRISPR Cas9 system targeting Pink1 and Dj-1 to the monkey brain, severe loss of dopaminergic neurons in the substantia nigra and accumulation of a-synuclein pathology were observed (Li et al., 2021). Hence, once the expression of Pink1 in the brain, especially in the substantia nigra, is knocked out in monkeys, the development of loss of dopamine neurons in the substantia nigra should be a clear consequence.

Scientists have designed a one-step lentiviral vector targeting Snca intron 1, which fused the catalytic domain of DNA methyltransferase 3A (DNMT3A) with dCas9 to carry out gene regulation (Wüllner et al., 2016; Kantor et al., 2018). Application of this system to hiPSC-derived dopaminergic neurons from PD patients with SNCA triploidy resulted in a fine down-regulation of SNCA mRNA and protein. Furthermore, reduced SNCA protein level by CRISPR-dCas9-DNMT3A system rescued disease-related cellular phenotypic features of SNCA triplicated dopaminergic neurons, such as mitochondrial ROS production and low cell viability (Wüllner et al., 2016). This suggested that combining the regulatory region sequences of genes with CRISPR-dCas9 technology may become a new epigenetic-based treatment method for PD. Apart from iPSC cells, human embryonic stem cells (hESCs) have also been used for research. Chen et al. have used CRISPR-Cas9n editing to remove SNCA gene from hESCs. These stem cells were then able to be converted into dopamine-producing neuronal cells in vitro, and were sensitive to the pS129-αSyn positive SNCA protein aggregates without forming toxic clumps. This study may benefit young patients with PD and patients with aggressive PD (Chen et al., 2019).

In BV2-G cells, knocking out of GMF significantly reduced oxidative stress by reducing ROS production and calcium flux. Furthermore, GMF deficiency significantly reduced nuclear translocation of NRF2, which regulates HO-1 and ferritin activation, the expression of cyclooxygenase 2 (COX2) and nitric oxide synthase 2 (NOS2) in BV2 microglia. GMF may regulate iron metabolism by regulating NRF2-HO1 and ferritin expression, thereby mediating protein aggregation in microglial cell homeostasis associated with progression of PD (Selvakumar et al., 2019).

In a PD rat model with overexpression of SNCA (p. A53T), the significant rescue of α-synuclein overexpression, reactive astrogliosis, dopamine neuron degeneration, and Parkinsonian motor deficits was observed after knocking out SNCA gene using CRISPR Cas9 (Yoon et al., 2022). GWAS analysis of the chromosome 7p locus associated with SNCA identified an interaction between non-metastatic melanoma protein B (GPNMB) and SNCA. It furtherly showed that absence of GPNMB resulted in the loss of the ability to internalize α-synuclein fibrils and develop α-synuclein pathology in iPSC-derived neurons (Diaz-Ortiz et al., 2022). In addition, compared to 59 healthy individuals, GPNMB was elevated in the plasma of 731 Parkinson’s patients. And these Parkinson’s patients had more severe symptoms, indicating that GPNMB is a novel PD risk and biomarker, and may serve as a potential target for PD treatment.

P13 is a novel protein involved in mitochondrial oxidative phosphorylation and its overexpression can induce mitochondrial dysfunction and cell apoptosis. Knocking out P13 can alleviate toxin-induced mitochondrial dysfunction and apoptosis in dopaminergic SH-SY5Y cells. Furthermore, researchers constructed a mouse strain with heterozygous deletion of P13 and found that heterozygous knock-out of P13 prevented toxin-induced motor deficits and loss of dopamine neurons in the substantia nigra (Inoue et al., 2018). These results indicated that the regulation of P13 may be a new target for therapeutic interventions in PD. In Prkn^−/− mice, exhaustive exercise leaded to inhibition of mitochondrial autophagy and a significant increase in the levels of inflammatory factors in the serum. Notably, knocking out STRING gene via CRISPR Cas9, the phenotype can be rescued (Sliter et al., 2018), which suggested that STRING is a target gene that regulates mitochondrial stability in PD’s patients with PRKN mutation. After knocking out NDUSF2 which is responsible for formation of mitochondrial complex I, mitochondria were damaged but neurons remained intact for a relatively long time in mice. Further studies revealed that these neurons could release dopamine through the cytosol and dendrites in the nigrostriatal fraction to maintain specific motor functions, despite the absence of dopamine in the striatum (González-Rodríguez et al., 2021). Researchers subsequently designed gene therapies targeting the substantia nigra, which enables neurons in the substantia nigra to convert levodopa to dopamine, these facilitates improved treatment with levodopa in the advanced stages of PD.

Recently, some teams have proposed an ideal strategy for endogenous neuronal regeneration via in situ trans-differentiation of glial cells into neurons. Using CRISPR Cas9, cord blood-derived mesenchymal stem cells (UCB-MSC) could secret soluble RAGE. After transplantation of these UCB-MSC into striatum of PD mice, the mice showed a significant reduction in neuronal cell death and increased motility in striatum and substantia nigra (Lee et al., 2019). Qian et al. (2020) and Zhou et al. (2020) treated PD model mice by editing PTBP1 using shRNA and CRISPR CasRx system, respectively. After suppressing the expression of PTBP1 in astrocytes, these cells can efficiently trans-differentiate in situ into functional neurons within weeks to months, which improves the motor function of PD mouse models. While a recent repeat experiment of knocking down PTBP1 with shRNA packaged by AAV did not obtain the similar results. Neither signs of trans-differentiation of astrocytes to neurons nor increasing the number of neurons or decreasing the number of astrocytes was observed over half a month to 3 months. Furthermore, down-regulation of PTBP1 in astrocytes did not improve cognitive function, reduce synaptic damage and Aβ/tau pathology in AD model mice (Guo et al., 2022). This suggested the demand for a more systematic, rigorous and stable strategy for future research on trans-differentiation of central nervous system (CNS) cells.