Artificial homologous recombination, which relies on the cellular recombination machinery, is laborious and time-consuming. Alternative methods, using exogenous nucleases that can be programmed to cut genomic DNA at unique sites, originated from genetic engineering studies in the mid-1990s. The aim was to create artificial site- specific endonucleases containing a modular DNA-binding domain joined to a DNA cleaving domain. The DNA-binding domain would be designed to act as a protein guide sequence, one that binds to a desired target DNA sequence.
Initially, the guide sequences were composed of DNA-binding elements from zinc finger transcription factors. They could position a coupled DNA-cleaving domain (from a bacterial endonuclease) at a desired target site in the genome. Thereafter, other types of DNA-binding protein elements were used as guide sequences. Subsequently, methods were developed that used RNA guide sequences instead of protein guide sequences. They have the huge advantage of being very much easier to carry out.
Exploiting repair of double-strand breaks in DNA
In genome editing with programmable endonucleases, the first, crucially important aim is to make a double-strand break in a pre-determined target DNA sequence in a population of desired cells. We describe below how that is carried out. But first we explain the ultimate aim of the genome editing procedures, which is to exploit how cells carry out repair of the unnatural double-strand break, and how to identify cells in which the DNA repair has resulted in a desirable sequence change at the target site. The two major path ways used to repair unnatural double-strand breaks are listed below.
• Nonhomologous end-joining (NHEJ). Unnatural double-strand breaks can be fatal for cells, and all cells have an emergency repair mechanism in which the priority is to quickly join the two broken ends before the DNA fragments drift apart. The process is error-prone, however, and joining of the two fragments frequently occurs with loss or gain of a small number of nucleotides.
• Homologous recombination (HR). This option is most readily available to dividing cells after S phase, when the DNA has duplicated to form closely associated sister chromatids. The cells can make flawless repairs of a double-strand break on a chromatid by using an invading DNA strand from the intact sister chromatid to act as a template DNA for new DNA synthesis (see Figure 1).
Fig1. Homologous recombination-mediated repair of double-strand DNA breaks. The double-strand break (DSB) in the chromatid at top is repaired using as a template the undamaged DNA strands in the sister chromatid (note, to make the mechanism easier to represent, the upper chromatid is shown in an unusual format; the 3′ → 5′ strand is placed above the 5′ → 3′ strand). The first step is to cut back the 5′ ends at the double-strand break to leave protruding single-strand regions with 3′ ends. Following strand invasion, each of the single-strand regions forms a duplex with an undamaged complementary DNA strand from the sister chromatid, which acts as a template for new DNA synthesis (newly synthesized DNA copied from the sister chromatid DNA is highlighted in yellow). Following DNA synthesis, the ends are sealed using DNA ligase. The repair is highly accurate because for both broken DNA strands, undamaged sister chromatid DNA strands act as a template to direct incorporation of the correct nucleotides during DNA synthesis.
In mammalian cells, NHEJ is the more commonly used method of repairing double strand breaks, and genome editing strategies often rely on natural errors made during NHEJ-based DNA repair (Figure 2A). In these cases, the double-strand break can be designed to occur in an early coding exon of a specific target gene in a population of cells. The cells are screened to identify those in which the DNA repair has introduced short deletions or insertions that cause a frameshift in the translational reading frame.
Fig2. Genome editing using programmable nucleases depends on natural errors, or artificial intervention, during repair of an artificial double strand break in DNA. Programmable nucleases are designed to be able to make a double-strand break at a unique target DNA sequence in the genome (top). (A) The break is often repaired by nonhomologous end-joining (NHEJ), which is prone to making errors, notably small deletions or insertions that may be desirable if the intention is to inactivate a gene. (B) The break may alternatively be repaired using homologous recombination (HR), and in this case an additional transgene can provide an altered sequence that is incorporated during DNA repair. The repair begins with a 5′ → 3′ exonuclease (exo) that trims back the broken ends (resection). That leaves room for invasion by a donor DNA strand (provided by a homologous DNA sequence) that acts as a template for new DNA synthesis. As a result of the repair, the original sequence is replaced by a copy of the donor DNA (see Figure 11.6 for the mechanism). By providing a transgene with a homologous sequence but carrying an altered sequence, the repaired sequence at the target site is a faithful replica of the donor sequence. The example here shows introduction of a single nucleotide change, but much more extensive changes can be made.
Alternatively, the genome editing strategies exploit homologous recombination mediated repair (sometime called homology-directed repair). Although artificial homologous recombination normally occurs at very low frequencies, a double-strand break in DNA massively increases the frequency of homologous recombination, by a factor of as much as 104 or more. Normally, the invading template strand whose sequence is copied would be from a natural homologous sequence (usually from a sister chromatid). However, the repair can be manipulated by providing an artificial donor sequence— a double-stranded plasmid transgene or a single-stranded oligonucleotide—that is designed to be highly homologous to the target sequence but containing some desired sequence change. By copying this artificial donor sequence during DNA repair, the original target sequence is replaced by the sequence of the donor strand, introducing the desired sequence change (Figure2B).
Genome editing using site-specific endonucleases containing a DNA-cleavage domain and a modular DNA-binding domain
A restriction endonuclease works by recognizing a specific short sequence in a DNA molecule, and then cleaving it. In most restriction enzymes, the DNA-binding and DNA cleavage activities are coincident, mapping to the same part of the protein. But some enzymes, such as FokI, a type IIS restriction endonuclease, have separable DNA-binding and DNA-cleavage domains. FokI works as a dimer, cleaving DNA at a short distance from a nonpalindromic and asymmetric recognition sequence.
The two methods described below take advantage of FokI’s properties: an isolated FokI DNA-cleavage domain cuts DNA randomly, but joining it to an artificial protein guide sequence that will recognize and bind a long, specific DNA sequence results in a site- specific endonuclease. A significant amount of genetic engineering is required. To make the guide sequence, a series of coding DNA sequences for modular DNA-binding domains with known sequence specificity must be joined together. The assembled guide sequence must then be ligated to a coding sequence for a FokI DNA-cleavage domain. The resulting DNA is incorporated into a transgene and expressed within a cell to make a site-specific endonuclease with a target-specific protein guide sequence covalently joined to a FokI DNA-cleavage domain. A pair of hybrid endonucleases is used to make a double-strand break at the target site (Figure 3A).
Fig3. Genome editing using programmable nucleases containing zinc finger and TALE DNA-binding modules. (A) The aim is to express transgenes to make a pair of site-specific endonucleases, each containing a monomeric FokI DNA-cleavage domain plus a protein guide sequence, an unrelated DNA-binding domain designed to recognize a specific nucleotide sequence. The FokI enzyme naturally works in bacterial cells as a dimer, using interacting DNA-cleavage domain monomers to make an asymmetric double-strand break. Here, the pair of site-specific endonucleases has mutated FokI DNA-cleavage domains (labeled + and −). They are designed to work as heterodimers, and co-operate to make an asymmetric double-strand break at a defined chosen target site, with the DNA-binding domains designed to bind to a sequence on the left (L) part or right (R) part of the target sequence. (B) The site-specific endonuclease has guide sequences constructed by combining zinc finger (ZNF) modules, each recognizing a specific trinucleotide sequence. In this example, the target sequence specificity is based on a 24-nucleotide sequence (12 nucleotides on the left and 12 nucleotides on the right, shown as filled black circles). (C) The site-specific endonuclease has guide sequences constructed by combining TALE (transcription activator-like effector) modules, each recognizing a single nucleotide.
Zinc finger nucleases
Here, the protein guide sequence is made by using genetic engineering to link together a series of C2H2 (Cys2 /His2) zinc fingers, the most common DNA-binding motifs in mammalian transcription factors. Each C2H2 zinc finger has about 23 amino acids (the name comes from its finger shape and the role of a central Zn atom that coordinates two cysteine and two histidine residues). Individual zinc fingers in C2H2 transcription factors are separated by 7–8 amino acids, and each zinc finger binds to a specific trinucleotide sequence. By using genetic engineering, it became possible to prepare site-specific endonucleases with multiple zinc finger mod ules to produce a combination that can specifically bind to 9- or 12- nucleotide target sequences (Figure 3B).
There are some major downsides. First, generating individual zinc finger nucleases by combining different modules is laborious. Additionally, not all possible trinucleotides have an available zinc finger module that can bind to them, and there can be sequence context issues when assembling a modular series of zinc fingers (the sequence specificity may not simply be a combination of the individual trinucleotide specificities). As a result, the choice of target sequences may be limited. Some highly effective zinc finger nucleases have been made, but the general difficulties described above have prompted interest in a more versatile alternative: TALENs.
TALENs (transcription activator-like effector nucleases)
Like zinc finger nucleases, TALENs are site-specific endonucleases that have a FokI DNA-cleavage domain and a protein guide sequence made up of modular DNA-binding sequences. In this case, the DNA-binding sequences come from transcription factors made by certain bacteria that are plant pathogens (notably Xanthomonas and related genera). These proteins—conservatively named TALEs (transcription activator-like effectors)—have a DNA-binding domain consisting of a series of tandem 34-amino acid repeats, with each repeat binding to a specific type of nucleotide in DNA. The repeats have highly conserved sequences, but key differences at amino acid residues 12 and 13 dictate the specificity of nucleotide binding.
After TALE proteins were identified that could bind to each of the four nucleotides, TALE guide sequences could be assembled by genetic engineering to specifically bind to any desired sequence. TALENs can therefore be designed to make a double-strand break at any target site of interest (Figure 3C). Note that although TALENs usually cut to give 5′ overhanging ends (like zinc finger nucleases), the first of the TALE DNA binding repeats binds to the most 5′ nucleotide of the recognition sequence (whereas the first zinc finger of zinc finger nucleases binds to the most 3′ nucleotide of the recognition sequence).
The much greater versatility of TALENs has meant that they have become the most popular way of carrying out genome editing with a hybrid endonuclease. They also show a high degree of sequence specificity. However, the genetic engineering required to produce TALENs to bind to target sites is still quite laborious.
Genome editing using RNA-guided endonucleases in the CRISPR-Cas system
Genome editing using site-specific endonucleases is greatly disadvantaged by the need to genetically engineer a new endonuclease for each target sequence. However, an alternative method—genome editing using RNA-guided endonucleases—has recently been developed and has rapidly become the method of choice. Its big advantage is that there is no need for complex genetic engineering: a single endonuclease can be used for all applications, and designing RNAs with target-specific guide sequences is comparatively quick and simple. Developed very recently, the method is quickly transforming genome engineering and has huge potential.
The natural function of CRISPR-Cas systems
The CRISPR-Cas system is a type of prokaryotic adaptive immune system used by the great majority of archaea and many bacteria as a defense system against invading viruses and plasmids. Cells have one or more CRISPR loci plus related genes forming a Cas (CRISPR-associated) operon. The central components are short, sequence-specific RNAs that detect foreign nucleic acids and an endonuclease that makes a double-strand break in the genome of the viral/plasmid invader, leading to its destruction. According to the components, there are at least six types of CRISPR-Cas system, but the simple type II system, notably that of Streptococcus pyogenes, is the one that has been most exploited in genome editing. The final stage of the defense system—involving recognition and cleavage of invading viruses and plasmids—is known as interference, and is preceded by two preparation stages, as listed below.
• Adaptation (spacer acquisition). The prokaryote captures short DNA segments (20–50 bp) known as “protospacers” from invading viruses and plasmids. It then inserts them into its genome as “spacer” sequences between copies of a similarly sized repeat sequence at certain loci. Containing clustered repeats interspersed with spacer sequences from viruses and plasmids, these loci are called CRISPR loci (CRISPR = clustered regularly interspaced short palindromic repeats). Individual spacer sequences stored in CRISPR loci act as a type of memory of a previous viral or plasmid invasion, and they allow the host cell to respond effectively to a subsequent invasion by the same type of invader.
• Expression and maturation. Transcription of a CRISPR locus produces a precursor RNA with sequences corresponding to each repeat unit and spacer. Thereafter, RNA cleavage and maturation produce short CRISPR RNAs (crRNAs) that have a single 5′ transcribed spacer sequence (guide sequence) plus a 3′ transcribed repeat sequence. Other components needed for interference are transcribed from genes in the Cas operon (Figure 4A).
• Interference. A single crRNA, with its transcribed spacer sequence, acts as a guide RNA, recruiting the endonuclease to cleave a viral or plasmid DNA with the same type of spacer sequence and separated by one nucleotide from a short protospacer-associated motif (PAM), such as NGG in the case of the S. pyogenes Cas9 system. In the type II CRISPR-Cas system, a crRNA recruits the Cas9 endonuclease with the help of a go-between, a trans-activating RNA (tracRNA) that can bind both crRNA and Cas9; the resulting complex is directed to its target by base pairing between the guide sequence of the crRNA and the target site (Figure 4B).
Fig4. Components of the CRISPR Cas system: a prokaryotic adaptive immune system based on RNA-guided endonucleases. The example here is a generalized type II system using a Cas9 nuclease. (A) A CRISPR locus has multiple copies of a specific host-cell repeat sequence (R) with interspersed DNA sequences (“spacers”) captured from the genomes of previously invading viruses or plasmids. Transcription and processing of a CRISPR locus produces a series of short CRISPR RNAs (crRNAs), each containing one transcribed spacer sequence at its 5′ end (which will act as a guide sequence) plus an adjoining sequence transcribed from a neighboring repeat. (B) The interference mechanism. Maturation of each crRNA requires a trans-activating RNA (tracRNA) that will act as a common bridge to link a crRNA to the Cas9 endonuclease: the tracRNA 5′ sequence hybridizes to the transcribed CRISPR repeat sequence at the 3′ end of the crRNA, and the tracRNA 3′ sequence has binding sites for the Cas9 endonuclease. The resulting crRNA tracRNA–Cas9 complex is guided by the 5′ transcribed spacer sequence of the crRNA (guide sequence) to hybridize to a complementary protospacer DNA sequence (of an invading virus/plasmid of the same type as the one from which the spacer was captured). Binding occurs just upstream of a short protospacer-associated motif (PAM) in the virus/plasmid DNA (see text). Thereafter, the Cas9 nuclease cleaves both DNA strands (cleavage sites shown by yellow triangles), leading to degradation of the virus/plasmid DNA.
In most CRISPR-Cas systems, a Cas1–Cas2 protein dimer is important in acquiring spacers and inserting new spacers into a CRISPR locus (at the proximal end, with duplication of one repeat). The spacers appear to be acquired from degraded DNA intermediates in the cell that arise during repair of double-strand breaks. Accidental acquisition of spacers from “self” DNA (the cell’s genome) could be detrimental if it results in degradation of self DNA by the CRISPR-Cas system, but the possibility of autoimmunity is minimized by a mechanism that distinguishes self DNA from nonself DNA (viruses and plasmids), as described by Amitai & Sorek (2016) (PMID 26751509) under Further Reading.
Genome editing using CRISPR-Cas
The S. pyogenes type II CRISPR-Cas system was the first to be modified for genome editing purposes, and has been extensively used because of its simplicity. Transgenes express the Cas9 nuclease and an artificial hybrid guide RNA with features of both a crRNA and a tracRNA. At its 5′ end, the hybrid RNA has a guide sequence ~20 nucleotides long designed to hybridize to a sequence at the target site adjoining a suitable protospacer-associated motif (PAM). At its 3′ end it has a Cas9-binding sequence from the tracRNA sequence, allowing it to recruit the Cas9 nuclease and transport it to the target site. Once deposited, Cas9 cleaves both DNA strands of the target sequence, using different DNA-cleavage domains for the two strands (Figure 5A).
Fig5. CRISPR-Cas genome editing uses an RNA-guided endonuclease to create a site-specific double-strand break in a genome. (A) The basic method requires transgenes to express a Cas9 endonuclease (or equivalent) plus a single RNA designed to have a specific ~20-nucleotide guide sequence at its 5′ end and binding sites for the Cas9 endonuclease at its 3′ end (similar to the crRNA–tracRNA combination in natural systems—see Figure 8.17). The guide sequence is designed to hybridize to a genome sequence that has a closely flanking protospacer associated motif (PAM) specific for the endonuclease (NGG in the case of Cas9—see text). The Cas9 endonuclease has two cleavage domains, an N-terminal domain (RuvC) and a centrally located domain (HNH), that are responsible for making cuts on the individual strands, as indicated (yellow triangles indicate cleavage sites). (B) Greater specificity is possible using two guide sequences and a pair of mutated Cas9 endonucleases (“nickases”) that each cleave just one strand. When Cas9-coding DNA is mutated to produce an aspartate-to-alanine change at amino acid position 10 (D10A), the RuvC cleavage domain is inactivated, but the HNH cleavage domain is able to cut one DNA strand. The pair of guide sequences are selected to bind to opposing DNA strands at neighboring positions (left and right target sites, just like when using zinc finger nucleases and TALENs).
The CRISPR-Cas9 genome editing system is both simple and quick but the target specificity is not so high as that of TALENs. Correct base pairing at nucleotides near the 5′ end of the guide sequence does not seem to be important but the seed sequence, the sequence closest to the PAM, is very important and correct base pairing here is critically important. As a result, the effective region for base pairing may be substantially less than 20 nucleotides (simply extending the length of a single guide sequence does not increase the sequence specificity).
Various modifications have been made to reduce the chances of off-target effects (where double-strand breaks occur elsewhere in the genome in addition to the desired target site). One way is to mutate one of the two cleavage domains of the Cas9 nuclease so that the modifed enzyme becomes a nickase; that is, it cuts a single DNA strand instead of cutting both strands. For example, the RuvC cleavage domain located at the N-terminal region of the Cas9 nuclease can be made catalytically inactive by a D10A substitution, and an H840A substitution inactivates the central HNH cleavage domain. A pair of guide RNAs and a pair of nickases can be used to cleave the two DNA strands at neighboring regions of a target DNA sequence to increase the specificity (Figure 5B).
