Discovered as an immune system against viral infection in domain bacteria and archaea, clustered regularly interspaced short palindromic repeats (CRISPR) system has quickly become a crucial tool in biological research. Long before it became the focus of debate because of its use to generate gene-edited babies (1), scientists recognized CRISPR system as an efficient, accurate and programmable nuclease system capable to induce double strand breaks (DSBs) in various organisms, therefore with a high potential as a versatile tool for scientific studies as well as a powerful tool for medicinal and agricultural improvement. Already, there are abundant data on the potential of CRISPR system for application in crop improvement (2) as an alternative to genetically modified organisms (GMO) (3). Moreover, clinical trials to cure cancer patients by using CRISPR edited T cells are ongoing in the United States (ClinicalTrials. gov registry number: NCT03399448) and China (registry number: NCT03545815).

From restriction enzyme technique (Fig. 1A), the ability to manipulate genomic DNA in living cells had a pivotal role in the history of biological research. In the 1980s, directed mutation technique via homologous recombination revolutionized the field by allowing directed mutation in mammalian cells (4). Homologous recombination is performed by introducing trans-acting DNA material, usually containing a selection marker, by flanking homologous sequence matching target genomic DNA (Fig. 1B). This technology led to pivotal discoveries at that time. As an example, directed gene deletion by homologous recombination of mice stem cells allowed subsequent generation of transgenic mouse bearing deletion in genes of interest (5). However, this technique efficiency was characteristically low (6). Thus, further researches focused on overcoming those limitations, for example by using negative selection markers such as thymidine kinase or diphtheria toxin fragment A. However, efficiency increase remained modest [reviewed in (7)]. In the 1990s, Rouet, et al. discovered that I-SceI, a Saccharomyces cerevisiae derived endonuclease, can introduce a DSB in mammalian cells (8). This leaded to the discovery of homing endonucleases that can be used to induce lateral transfer of an intervening sequence via DSB (Fig. 1C) (9). One of the major limitations with these homing endonucleases was that their recognition sites are relatively longer than most restriction enzymes (14-40 bp), limiting the number of suitable targets, prompting researches aiming to diversify homing enzymes recognition sites (10-12).

Figure 1. Representative cartoon of the evolution of the directed mutation technique from enzyme restriction (A), homologous recombination (B), homing endonuclease (C), ZFN/TALEN (D) and the CRISPR/Cas9 system (E).

In 1992, Fok I, another nuclease from Flavobacterium okeanokoites came to focus as its active domain and binding domain was identified (13). Interestingly, Fok I active domain requires homodimerization for activation, that is a pair of DNA-protein binding to flanking site to trigger its nuclease activity. Conversely, zinc finger motifs were known as short motifs in regulatory proteins that only need a three bases recognition site to bind the DNA (14). Thus, fusion of zinc finger DNA binding domain to Fok I nuclease domain (15, 16) allowed development of chimeric enzymes, or zinc finger nucleases (ZFNs), that can induce targeted DSB as a pair (17). Gene manipulation of the ZFNs domain allowed targeted gene editing in a broad range of organisms, opening novel experimental and therapeutic possibilities (18). Later in 2009, transcription activator-like effector (TALE) proteins from the plant pathogenic bacteria Xanthomonas was shown to have two hypervariable amino acid residues that can recognize a single base pair (19), leading to the development of a chimeric protein with TALE binding domain and Fok I nuclease domain, or TALEN (Fig. 1D) (20). While the development of the ZNFs and TALENs allowed targeted gene manipulation in living cells, because of theirs mandatory cloning and protein modification steps needed to program target loci, the use of these techniques were still restricted. In contrast, the newly emerging CRISPR/Cas9 technology uses a short guide RNA to direct its endonuclease activity, that is more convenient to manipulate (Fig. 1E).

CRISPR system is originally an RNA-mediated defense system in bacteria and archaea (21). While first discovered in Escherichia (E.) coli in the 1980s (22), the term CRISPR was later coined as its repeat motif was identified among other prokaryotes (23). CRISPR motif contains several nearly palindromic 30 bp repeats interspaced by ∼36 bp non-repetitive spacers, with Cas genes nearby (24). In 2005, three independent reports showed that CRISPR various spacers are present in various prokaryotes strains, including among others, Streptococcus, Sulfololus, Eschericia and Listesria genus, and that those spacers are from mobile genetic elements such as viruses and phages, implying that CRISPR motifs are part of a major prokaryotic defense mechanism toward foreign genetic elements (25-27).

In 2007, Barrangou, et al. reported that CRISPR spacers provide prokaryotes resistance against corresponding phage by mechanisms involving neighbor Cas genes (28). Following this key experiment, researchers quickly started to elucidate the mechanism of CRISPR/Cas system. In a CRISPR/Cas system inserted in E. coli, it was shown that a complex of five Cas proteins is required for maturation of a 61 bp CRISPR RNA (crRNA) that comprise a spacer flanked by two repeat sequences (29). This led to an early hypothesis that crRNA may form a secondary structure (30). One of the first speculations was that these small RNAs act by similar mechanism to the well-studied small RNA interfering system, that is by RNA-RNA interaction between spacer and target RNA leading to foreign RNA degradation (25). However, modifying the intron sequence within the target sequence abolished the CRISPR/Cas defense system, thus demonstrating that the target of crRNA is not mRNA but DNA (31). In the same year, it was shown that the repeats that are adjacent to spacers are critical for the S. thermophilus defense mechanism (32, 33) and that Cas9, a S. thermophilus Cas protein, cleaved the foreign DNA at exactly the same position relative to these repeats that were later named Protospacer Adjacent Motif (PAM) (34). Various eukaryotes have specific PAM sequences that are recognized by specific Cas genes. Further investigations were focused on S. pyogenes Cas 9 (spCas9) based CRISPR/Cas9 system, as it requires a relatively short 5’-NGG PAM sequence for its activity.

Parallel to these studies, another focus of this field was to elucidate the mechanism underlying crRNA maturation, as many organisms with a working CRISPR/Cas9 system lacked essential Cas proteins thought to be essential for crRNA maturation in previous studies. Trans-activating CRISPR RNA (tracrRNA) were first identified as the third most abundant class of transcripts in the S. pyogenes. As their name suggests, these small RNAs transcribed from locus adjacent to the CRISPR, and more importantly, were shown to be essential for crRNA maturation (35), and later on, for Cas9 nuclease activity. Thus, tracrRNAs were recognized as essential components of this system (36, 37).

In 2011, as all the crucial components of the CRISPR/Cas9 systems were identified, Siksnys’ lab successfully reconstructed a working CRISPR/Cas9 system in E. coli from S. thermophilus, demonstrating for the first time that the CRISPR is transferable between organisms (38). Subsequently in 2012, two independent groups demonstrated the potential of the CRISPR/Cas9 system as a biological tool. Siksnys and his colleagues showed that in vitro, purified his-tagged Cas9 and custom designed spacers can introduce a DSB at a locus that is 3 bases away from PAM site (39). Conversely, Charpentier and her colleagues showed that the cut site is programmable, using a S. pyogenes Cas9 expressed in E. coli with in vitro transcribed crRNA and tracrRNA, and also a fused single-guide RNA (sgRNA), which is now widely used in gene editing. sgRNA is a crRNA and tracrRNA hybrid RNA that comprises a stem loops structure, repeat versus anti-repeat duplex, a S. pyogenes specific PAM site (5’-NGG-3’) directly adjacent to 20 bp complementary to the target sequence (36). Both groups recognized the potential of their results, stating the potential use of this system as an RNA programmable DNA editing technique. Further studies showed that the CRISPR can be adapted in vivo for eukaryotes, notably human cells (40-43). Shortly after its potential as a genome editing tool was presented, the CRISPR/Cas9 system was widely adopted by the scientific community because of its ease in programming, yet high specificity to perform gene editing at target sites, with more than 15,000 papers published, and more than 4,500 in 2019 (Fig. 2).

Figure 2. Graph depicting the number of the CRISPR/Cas9 related publications yearly by MeSH (Medical Subject Headings) 1980 to March, 2020.

While the CRISPR/Cas9 system is quickly becoming the tool of choice to perform targeted mutation, there are still limitations to address. One of these was raised and assessed early from the groups that developed the CRISPR technology (44-47). Off-target, that is nuclease activity at sites other than the programmed sites, can occur in the CRISPR/Cas9 system [the widely used CRISPR system, Class 2 type II from the S. pyogenes (48)] if the sgRNA binding allows mismatches. Cleavage occurs after the third nucleotide from the PAM within the corresponding target sequence (49). It was shown that sgRNA mismatch can be tolerated up to five base pairs, depending on the sgRNA and Cas9 amount used (45). Moreover, it was demonstrated by chromatin immunoprecipitation-sequencing (ChIP-seq) that binding sites of the deactivated Cas9 (dCas9)/sgRNA complex are less stringent after the fifth base adjacent to the PAM site, resulting in multiple off-target bindings (50, 51), preferably in the open chromatin region (52). However, while the dCas9/sgRNA allows non-specific binding, it was suggested that the guide RNA and the target PAM distal site sequence interaction is necessary for the Cas9 cleavage activity (53), thus decreasing the occurrence of off-targets in vivo compared to the occurrence of binding identified by the ChIP-seq. Indeed, relative to the large number of off-target dCas9 binding loci, only few or no insertion-deletions mutation (InDels) were observed at these off-target sites (49, 51).

Researchers investigated various paths toward improving the Cas9 on-target efficiency. A straightforward method was to optimize Cas9 and gRNA amount and proportion (45), or directly deliver the Cas9 protein and sgRNA as ribonucleoprotein into cells (54). Additionally, while studies of the Cas9 variants from other organisms than the S. pyogenes have mostly contributed to broaden the scope of CRISPR technology application by diversifying available the PAM motifs (55), some of those variants may be also be used to optimize target specificity (56). While the consensus is that the probability to discover new subtypes more efficient than the known Cas9 is low, deeper understanding of Cas proteins variants and discovery of related proteins may contribute to technical enhancements (57). For example, Cas12a, formally known as Cpf1 (58), is annotated as a class 2 type V CRISPR system (59, 60). It requires only a single crRNA to introduce a staggered DSB, thus requires a simpler guide RNA and can be used to control insert orientation by its staggered DSB. CRISPR/Cpf1 has been used to generate targeted knock-out mice without any off-target effects observed, suggesting that it has potential as a DNA editing tool with efficiency that is at least comparable to the Cas9 system (61). Some of the Cas9 variants even shows to have increased specificity of targeting site, among other improvements such as PAM diversification and increase of efficiency (Table 1). However, it should be noted that off-target should still be carefully considered, as a recent study using reporter activation assay to investigate editing efficiency shows that both CRISPR/Cas9 and CRISPR/Cpf1 tolerate off-target mismatch mutation, especially in PAM-distal region of investigated target (63).

Table 1

Table 1 List of Cas9 variants developed to enhance specificity of sgRNA targeting

Name	Modification	Note
Cas9n or D10A Cas9 nickase	D10A	Developed by targeted mutation to induce rare DSB from nucleotide nickase (38), its paired D10A usage has reduced off-target and increased efficiency (62)
SpCas9-HF	N497A, R661A, Q695A, Q926A	Alteration of amino acid at gDNA interacting domain increase specificity (57) based on structural data from Cas9/gDNA complex crystallization (59)
espCas9(1.0)/(1.1)	K810A/K848A, K1003A, R1060A	Modification of amino acid at interacting with noncomplementary DNA strand based on structural data from Cas9/gDNA complex crystallization (61)
HypaCas9	N692A, M694A, Q695A, H698A	Alteration in REC3 domain, identified by single-molecule Förster resonance energy transfer experiments (68)
HifiCas9	R691A	Identified by unbiased bacterial screening method (69)
evoCas9	M495V, Y515N, K526E, R661Q	Yeast based screening of random mutation in the REC3 domain (70)
SniperCas9	F5395, M763I, K890N	Directed evolution in a E. coli system to screen for accurate and efficient nuclease (71)
xCas9-3.6	E108G, S217A, A262T, S409I, E480K, E543D, M694I, E1219V	Phage assisted evolution to screen for Cas9 variants to diversify PAM sites. Those variants are also more specific to target sites (72)
xCas9-3.7	A262T, R324L, S409I, E480K, E543D, M694I, E1219V

Recently, new techniques using dCas9 variants fused to proteins are emerging as promising tools in gene editing. For example, base editing is a technique that uses dCas9 fused to a nucleobase deaminase enzyme or a DNA glycosylase able to convert a single base pair in targeted site, enabling precise point mutation (64). Prime editing, or search-and-replace genome editing is a technique that uses a dCas9 fused to a reverse transcriptase domain and a modified gRNA to insert a designed sequence within target site, therefore enabling precise sequence insertion without donor DNA (65). While they do not induce DSBs, these editing techniques are also prone to the same off-target issue as CRISPR/Cas9 system. Indeed, it was shown that base editing occurs at off-target sites in a frequency ranging from 0.07% to 100% in 38-58% genes in human cell (66). Interestingly, it was shown that improvement in base editing on-target efficiency can be achieved by optimizing the base editing domain rather than the dCas9 domain (67).

Finally, optimization can be achieved by sgRNA configuration. The five base pairs at proximity of the PAM region of the sgRNA are known as ‘seed regions’ more stringent in guiding the Cas9 complex to its target (51). While distal sequences from the PAM region are necessary to the Cas9 activity (53), those ‘seed regions’ context are crucial in determining the binding specificity. For example, U-rich seeds sequence increases specificity (51, 68) while high GCs content decreases Cas9 activity (68, 69). G is more favorable and C is less favorable as the base directly after the PAM. In contrast, at the fifth base from the PAM site, C is more preferred. From the 9th to 10th distal sequence from the PAM, A is favorable. At the 18th base from PAM, C is less favorable (51, 68, 70, 71). Those criteria can be used to design a target site for the gene of interest with minimum putative off-targets.

In order to use CRISPR/Cas9 system as a therapeutic tool, common agreement is that the risk of off-target should be assayed in a case specific manner. The impact of off-target mutation in patients will differ greatly based on the genes and tissues affected, as factors such as differential expression between tissues and pathological effect of genes differ greatly (72). Thus, to enable practical application, another focus was the development of tools that facilitate analysis of the off-targets using whole-genome sequencing (WGS) with improvement in cost and efficiency. To this end, several methods were developed including BLESS/BLISS which label and detect breaks in situ (73), Guide-Seq which incorporates oligo within DSBs as priming targets (74), Digenome-seq that perform WGS in nuclease digested sample to detect random modification (75), qDSB-Seq which compare DSBs ‘spike’ between two samples upon random DSBs (76) and recently DISCOVER-seq which track Cas9 binding site using Cas9-ChIP and WGS (77).

Another putative side effect of the targeted CRISPR/Cas9 system is on-target mis-regulation. The underlying mechanism of gene mutation by CRISPR/Cas9 systems is that Cas9 induces DSB in the genome that triggers repair pathways via the non-homologous joint end repair (NHEJ) or the less frequent homology directed repair (HDR) that occurs if a homologous template is nearby (78). In eukaryote, DSBs occur relatively frequently because of reactive oxygen species, radiation, replication error or mechanical stress. Thus, proteins that are involved NHEJ are intrinsically active (79). Repair by NHEJ usually results in imperfect repairs with insertion or deletion mutations (InDels), leading to a frameshift mutation that consequently results to a premature termination codon (PTC) within coding region. PTC triggers cell inherent nonsense-mediated mRNA decay (NMD) mechanism leading to complete knock out of the targeted gene by mRNA degradation within seconds (80, 81). On-target CRISPR/Cas9 mediated gene silencing is usually achieved by this mechanism.

However, a recent investigation showed that ∼50% of the cell lines from previous studies did not result in targeted gene knock out, but rather caused the production of truncated functional proteins. To reduce on-target mis-regulation, the authors recommended selecting target sites avoiding the internal ribosomal re-entry site, as InDels in those sites may result in the production of pseudo-mRNA. Also, exon splicing enhancers site should be avoided as target site as their deletion may result in exon skipping, thus generating truncated proteins rather than knock out (82). While other studies applied this exon skipping capability to introduce alteration in the targeted genes (83-86), the consensus is that in addition to the off-target mutations, these on-target mis-regulations should be carefully evaluated before application (84, 86).

For now, CRISPR/Cas9 system is known as the most convenient method to program target sites for mutation among developed techniques. Thus, CRISPR/Cas9 system is quickly becoming a prominent tool for basic research as well as for clinical and agricultural purposes. In this review, we discussed a few of the many studies that led to its development. Its basic principle is that it induce a targeted DBS in the genome that can go through two inherent mechanisms, NHEJ that ligate the break without a homologous template and HDR that use a homologous template, therefore that is less error-prone but has lower efficiency compared to NHEJ. Thus, NHEJ remains the most commonly used pathway despite its putative on/off-target side effect. Recent improvements have been initiated to increase the specificity of the Cas9 targeted DSB as well as to develop techniques to detect off-target at large scale, crucial to evaluating its safety for clinical and agricultural applications. While off-target mutation can be detected using large-scale analyses, on-target mis-regulation can only be assessed after mutation in a case specific manner. This shows that the CRISPR/Cas9 possible side-effects should be carefully assayed before application, and there is room for improving this highly effective targeted mutation technique.

This study was supported by grants awarded to the JWO (NRF-2017H1D3A1A01052995 and NRF-2016R1D1A1B0393 5382) and the JS (NRF-S201806S00067) by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology. We apologize to colleagues whose work could not be cited because of space limitations.

The authors have no conflicting interests.