• Open Access

Genome modifications in plant cells by custom-made restriction enzymes


(fax 972 8 647 2992; email tzfira@bgu.ac.il)


Genome editing, i.e. the ability to mutagenize, insert, delete and replace sequences, in living cells is a powerful and highly desirable method that could potentially revolutionize plant basic research and applied biotechnology. Indeed, various research groups from academia and industry are in a race to devise methods and develop tools that will enable not only site-specific mutagenesis but also controlled foreign DNA integration and replacement of native and transgene sequences by foreign DNA, in living plant cells. In recent years, much of the progress seen in gene targeting in plant cells has been attributed to the development of zinc finger nucleases and other novel restriction enzymes for use as molecular DNA scissors. The induction of double-strand breaks at specific genomic locations by zinc finger nucleases and other novel restriction enzymes results in a wide variety of genetic changes, which range from gene addition to the replacement, deletion and site-specific mutagenesis of endogenous and heterologous genes in living plant cells. In this review, we discuss the principles and tools for restriction enzyme-mediated gene targeting in plant cells, as well as their current and prospective use for gene targeting in model and crop plants.


Gene targeting was developed into a powerful tool for basic cell research in Saccharomyces cerevisiae (Scherer and Davis, 1979), then was latterly adapted for use in other model organisms including certain fungal species, mouse embryonic stem cells, Drosophila and human cell lines (Baribault and Kemler, 1989; Venken and Bellen, 2005; Porteus, 2007; Hall et al., 2009; Laible and Alonso-Gonzalez, 2009; Tenzen et al., 2009). Ever since these developments, scientists have been seeking to develop similar methods for plant species (Puchta, 2002; Hanin and Paszkowski, 2003; Reiss, 2003; Puchta and Hohn, 2005, 2010; Tzfira and White, 2005; Porteus, 2009; Weinthal et al., 2010; Vainstein et al., 2011). Yet, gene targeting by homologous recombination (HR), i.e. the replacement of a native genomic sequence with a foreign DNA molecule, which in principle can occur in every living eukaryotic cell, has been found to be difficult or nearly impossible to achieve in plant cells. Why is it that most foreign DNA molecules, which are typically transformed into plant cells by Agrobacterium-mediated gene transfer or direct plasmid transfer, are directed to integration via non-homologous end joining (NHEJ) and not HR, irrespective to their internal sequence? This question remains mostly unanswered. In line with the predominant repair of broken DNA fragments by NHEJ in plants (Ray and Langer, 2002; Britt and May, 2003; Lieberman-Lazarovich and Levy, 2011), integrations of a random nature predominate those involving homology direction for transforming DNA molecules of an Agrobacterium-transferred DNA (T-DNA) or plasmid DNA form (Draper et al., 1988; Tzfira et al., 2004; Li et al., 2005; Banta and Montenegro, 2008; Ziemienowicz et al., 2008; Dafny-Yelin et al., 2009; Shiboleth and Tzfira, 2011). Scientists have thus devised several experimental approaches to overcoming these biological limitations, including: (i) extensive screening for rare HR-mediated foreign DNA integration events (Miao and Lam, 1995; Kempin et al., 1997); (ii) developing novel selection methods for eliminating NHEJ-mediated integration events while recovering HR events (Terada et al., 2002, 2007; Yamauchi et al., 2009); and (iii) reprogramming the plant cell’s DNA repair machinery in favour of HR, by expressing bacterial and yeast genes (Hanin et al., 2000; Reiss et al., 2000; Shaked et al., 2005). While proven useful for obtaining cells and plants with modified transgenic and native genome sequences, these approaches have yet to be tested and applied to a wide range of genes and plant species.

Studies in animal and plant cells revealed that induction of genomic double-strand breaks (DSBs) can lead to enhanced HR (reviewed in Carroll, 2004; Lieberman-Lazarovich and Levy, 2011; Porteus, 2009; Puchta, 2005). It was thus suggested that if a site-specific DSB can be induced in the target gene, then foreign DNA molecules might be directed to integration into the break site via HR and not just NHEJ. Indeed, researchers reported on increased HR-mediated genomic repair upon expression of naturally occurring homing endonucleases in transgenic plants that had been engineered to carry an artificial site for these enzymes (Puchta et al., 1993; Siebert and Puchta, 2002; D’Halluin et al., 2008). Furthermore, several research groups demonstrated that T-DNA molecules, which shared no homology to the target site, could be captured, albeit by NHEJ, at genomic DSBs (Salomon and Puchta, 1998; Chilton and Que, 2003; Tzfira et al., 2003). Collectively, these observations suggested that genomic DSBs may act as traps for the integration of foreign DNA molecules, and that once directed to DSBs, these molecules could potentially be directed to integration via HR. Naturally, the implementation of DSB-dependent foreign DNA integration (by either NHEJ or HR) cannot solely depend on natural homing endonucleases and requires the development of enzymes with novel specificities. Indeed, in the past few years, several breakthroughs have been made in the development of novel restriction enzymes (NREs) and their use for genome editing in eukaryotes, including plants. NRE-mediated genome engineering in animals, animal cells and human cells led to a variety of outcomes, including, for example, site-specific mutagenesis (Foley et al., 2009; Ochiai et al., 2010; Takasu et al., 2010), gene replacement by HR (Urnov et al., 2005; Beumer et al., 2006), site-specific integration (Moehle et al., 2007) and chromosomal deletion (Lee et al., 2010).

Enabling routine and efficient NRE-mediated genome engineering in plant cells is likely to revolutionize the introduction of novel traits in crop plants, which is currently achieved by various plant genetic transformation approaches (Que et al., 2010; Vega- Sanchez and Ronald, 2010; Ali et al., 2011; Wally and Punja, 2011). Thus, for example, site-specific mutagenesis might be used as an alternative to RNA interference for silencing native genes in crop species, while gene replacement by HR could be harnessed to introduce genes with modified and improved agronomically important alleles. Gene replacement by HR might also be used to replace a native gene’s regulatory elements with engineered ones, as a means of enhancing the gene’s expression in the modified plants. Site-specific integration, by either HR or NHEJ, can assist in directing transgenes to specific genomic locations for more stable and predictable expression in transgenic crops. Finally, gene and chromosomal deletions could be used to eliminate unwanted transgenic sequences, native genes or gene clusters from crop plants.

Several excellent reviews have been published in the past several years that focus on NRE-mediated genome engineering and its uses for genome editing in animals, animal cells and human cells (Porteus and Carroll, 2005; Cathomen and Joung, 2008; Remy et al., 2010; Urnov et al., 2010). Here, we focus on NRE-mediated genome modification in plants. We introduce the reader to the different types of enzymes used for genome editing and the tools for their validation and expression in plant cells, while presenting examples from various reports that describe the editing of model and crop plant genomes. In our review, we only briefly introduce the reader to other pioneering and important studies in which successful gene targeting has been achieved by means of novel strategies and selection methods (Table 1), and we refer the reader to the extensive reviews that focus on them (Hohn and Puchta, 1999; Kumar and Fladung, 2001; Puchta, 2002, 2003; Britt and May, 2003; Iida and Terada, 2005).

Table 1.   Examples of HR-mediated gene replacement in plants
SpeciesTargetDonor DNASelectionReferences
  1. APH(3′)II, aminoglycoside 3′-phosphotransferase II; Sur, acetolactate synthase; hpt, hygromycin phosphotransferase; TGA3, a basic leucine zipper (bZIP)-like transcription factor-encoding gene; AGL5, agamous-like5 MADS-box gene; Gln1, glutamine synthetase; Pzf, gene encoding plant member of the RING-finger family of zinc-binding proteins; CHS, chalcone synthase; PPO, protoporphyrinogen oxidase; ADH, alcohol dehydrogenase; Wx, waxy gene encoding for starch granule-bound starch synthase; MET, maintenance DNA methyltransferase; CodA, cytosine deaminase; DT-A, Diphtheria toxin A fragment; EASE, egg-apparatus-specific enhancer.

TobaccoDefective APH(3′)IIPlasmidRestoration of kanamycin resistancePaszkowski et al. (1988)
TobaccoDefective APH(3′)IIT-DNARestoration of kanamycin resistanceOffringa et al. (1990)
TobaccoSurT-DNAResistance to chlorsulfuronLee et al. (1990)
ArabidopsisDefective hptPlasmidRestoration of hygromycin resistanceHalfter et al. (1992)
ArabidopsisTGA3T-DNAPositive resistance and GUS expressionMiao and Lam (1995)
ArabidopsisAGL5T-DNAPositive selection and PCR screeningKempin et al. (1997)
LotusGln1, PzfT-DNAPositive (nptII) and negative (codA) selectionThykjaer et al. (1997)
ArabidopsisCHST-DNAPositive (nptII) and negative (codA) selectionGallego et al. (1999)
ArabidopsisPPOT-DNAPositive and PCR screeningHanin et al. (2001)
RiceSurT-DNAResistance to butafenacil selectionEndo et al. (2007)
ArabidopsisADHT-DNAPositive (nptII) and negative (codA) selectionXiaohui Wang et al. (2001)
RiceWxT-DNAPositive (hpt) and negative (DT-A) selectionTerada et al. (2002)
RiceADH2T-DNAPositive (hpt) and negative (DT-A) selectionJohzuka-Hisatomi et al. (2008)
RiceMET1aT-DNAPositive (hpt) and negative (DT-A) selectionYamauchi et al. (2009)
35S:Rad54-ArabidopsisCRUCIFERINT-DNAGFP screeningShaked et al. (2005)
EASE:Rad54-ArabidopsisCRUCIFERINT-DNARFP screeningEven-Faitelson et al. (2011)

HR-mediated gene targeting in plants

Over two decades ago, Paszkowski et al. (1988) demonstrated that foreign DNA, which typically integrates by NHEJ at random locations across the target plant genome (Tzfira et al., 2004; Ziemienowicz et al., 2008; Dafny-Yelin et al., 2009), can integrate at specific locations via HR, albeit at very low frequencies. The rare HR-mediated foreign DNA events were detected by correction of a mutated selection gene in transgenic tobacco plants. More specifically, the authors first produced transgenic tobacco carrying a non-functional aminoglycoside 3′-phosphotransferase-encoding transgene (APH(3′)II), retransformed these plants with a linear plasmid that had been engineered to carry the correcting sequences, and selected for kanamycin-resistant plants. While proven to be feasible, the rate of HR-mediated gene targeting observed in this pioneering research was extremely low (on the order of 10−4). A similarly low rate of HR-mediated gene targeting (8.4 × 10−5) was also reported during the targeting of native sequences (the acetolactate synthase-encoding genes SurA and SurB) in tobacco plants, where T-DNAs were used as the transforming molecules (Lee et al., 1988). There too, a selection-based strategy was deployed for detection of rare HR-mediated T-DNA integration events, and cells with targeted acetolactate synthase-encoding genes were selected based on their ability to survive in the presence of the herbicide chlorsulfuron. A positive selection approach was also used in early studies of gene targeting in Arabidopsis plants, where Halfter et al. (1992) targeted and repaired a mutated hygromycin phosphotransferase gene (hpt) in transgenic Arabidopsis plants. Targeting of non-selective genes was reported by Miao and Lam (1995) and Kempin et al. (1997), who targeted the TGA3 [a basic leucine zipper (bZIP)-like transcription factor] locus and the AGL5 MADS-box gene, respectively, in Arabidopsis plants. Both groups used targeting vectors in which a kanamycin selection gene was cloned between regions of homology to the target gene on the foreign DNA molecule and used them to target and recover kanamycin-resistant cells. It should be noted that most of the kanamycin-resistant events derived from NHEJ and not HR-mediated foreign DNA integration. Thus, Miao and Lam (1995) also incorporated the GUS-encoding reporter gene outside TGA3′s regions of homology and used GUS staining, while Kempin et al. (1997) used PCR-based screening to differentiate between targeted and ectopic integration events.

The above-mentioned studies and others, which also included targeting of native agricultural crops and not just model plants (Table 1), demonstrated the feasibility of HR-dependent gene targeting in plants. Nonetheless, the extremely low natural frequency of HR-dependent gene targeting in plant cells prompted scientists to explore the possibility of modulating the plant DNA repair machinery as a means of enhancing the rate of HR in plant cells. Modulation of the plant DNA repair machinery can potentially be achieved by overexpressing heterologous or native DNA repair genes, or by down-regulating native DNA repair genes. Reiss et al. (2000), for example, demonstrated that overexpression of the bacterial protein RecA can stimulate sister chromatid exchange, although not HR-mediated gene targeting, in transgenic plants. In other reports, increased intrachromosomal HR was observed in MIM (a family member of structure maintenance of chromosomes proteins)-overexpressing plants (Hanin et al., 2000). Increased intrachromosomal HR was also observed in rad50 (a homologue of a yeast protein that is part of a complex that functions in both HR and NHEJ DNA repair pathways) (Gherbi et al., 2001) and Chromatin assembly factor 1 (CAF-1) (Endo et al., 2006) mutants. Increased intrachromosomal and extrachromosomal recombination was observed in RuvC (a bacterial protein that is involved in resolving Holliday junctions)-overexpressing plants (Shalev et al., 1999). More recently, Shaked et al. (2005) reported on a significant increase in the rate of HR-mediated gene targeting in transgenic Arabidopsis which overexpressed a chromatin-remodelling protein (RAD54) from yeast: the authors developed a GFP-based gene-targeting assay that allowed for quick visual screening of the targeting events in Arabidopsis seeds and reported on a HR-dependent gene-targeting frequency of up to 10−1 in RAD54-transgenic Arabidopsis (Shaked et al., 2005). More than 500 targeted plants were produced in that study, clearly indicating the potential of this approach for increasing the rate of HR-mediated gene targeting in plant species. More recently, Even-Faitelson et al. (2011) reported that expression of RAD54 under the control of the egg-apparatus-specific enhancer (EASE) also results in enhanced frequency of HR-dependent gene targeting in Arabidopsis. While the enhancement was slightly lower than that in constitutively expressing RAD54-transgenic plants (Shaked et al., 2005), one clear advantage of this approach lies in the reduced risk of transgenic silencing that is often associated with use of the 35S promoter in Arabidopsis (Mlotshwa et al., 2010). Whether high rates of gene targeting can be obtained in the above-described transgenic and mutant plants or by overexpression of other DNA repair proteins from yeast, plants and others has yet to be determined.

In their study, Shaked et al. (2005) and Even-Faitelson et al. (2011) also demonstrated the importance of devising strategies for efficient screening of targeting events. Indeed, novel transformation vectors have also been developed to assist with positive selection of both rare targeting events and frequent random integration events, while deploying negative selection against the latter. This positive–negative selection approach, which was originally developed as a tool for the selection of targeting events in non-selectable genes in mouse stem cells (Mansour et al., 1988), was first deployed by Thykjaer et al. (1997), who developed a set of gene-targeting vectors (using nptII and codA for positive and negative selection, respectively) for Gln1 and Pzf in Lotus japonica (Table 1). Nonetheless, while 185 negatively selected calluses were obtained, no conclusive evidence of successful gene-targeting events was provided. Similarly, Gallego et al. (1999) used a combination of htp and codA in an attempt to target the chalcone synthase-encoding gene in Arabidopsis cell suspension culture. Here too, negatively selected calluses were obtained, but with no detectable gene-targeting events. Successful positive–negative selection-based gene targeting was first reported by Xiaohui Wang et al. (2001), who targeted the alcohol dehydrogenase (ADH)-encoding gene in Arabidopsis root culture. Positive–negative selection vectors (using hpt and DT-A for positive and negative selection, respectively) were also proven useful for targeting Waxy and Adh2 in rice (Terada et al., 2002, 2007; Johzuka-Hisatomi et al., 2008) and for the production of mutant rice plants that were targeted at their methyltransferase-encoding gene (Yamauchi et al., 2009).

While relying on the plant’s natural or modulated DNA repair mechanisms and using negative–positive selection vectors have proven useful for the targeting of various native and transgenic genes (Table 1), such systems are still challenging and may be difficult to reproduce for the targeting of other genes and in other plant species (reviewed in Hohn and Puchta, 1999; Kumar and Fladung, 2001; Puchta, 2002, 2003; Britt and May, 2003; Iida and Terada, 2005; Puchta and Hohn, 2005; Porteus, 2009; Weinthal et al., 2010; Vainstein et al., 2011). In the last decade, several advances have been made in the development of NREs and in their deployment for induction of genomic DSBs in eukaryotic cells. Because genomic DSBs can be repaired by HR or NHEJ, NRE-mediated genome editing may provide the much needed technological breakthrough in gene targeting in plant cells.

Gene targeting by induction of genomic DSBs

Studies have shown that DSBs can stimulate the HR DNA repair machinery. DSBs can also be repaired by the error-prone NHEJ DNA repair machinery, which can lead to site-specific mutagenesis and the capture of foreign DNA (reviewed in Carroll, 2004; Puchta, 2005; Porteus, 2009). Thus, for example, induction of DSBs by X-ray (Tovar and Lichtenstein, 1992), excision of transposable elements (Xiao and Peterson, 2000) or expression of HO and I-SceI endonucleases (Chiurazzi et al., 1996; Siebert and Puchta, 2002; Orel et al., 2003) have all been shown capable of enhancing intrachromosomal HR in plant cells. In another example, DSB induction led to enhancement of extrachromosomal HR, as demonstrated by expression of I-SceI expression in plant cells (Puchta et al., 1996), and the capture of foreign DNA molecules by NHEJ into the break sites, as demonstrated by expression of I-SceI- and I-CeuI (Salomon and Puchta, 1998; Chilton and Que, 2003; Tzfira et al., 2003). These observations (Table 2), as well as reports in other eukaryotic cells, suggest that NREs can potentially be used for a wide range of genome modifications in plant cells, as illustrated in Figure 1.

Table 2.   Examples of restriction enzyme-mediated genome editing in plants
SpeciesTargetZFN/donor delivery systemEnzymeSelection and outcomeReferences
  1. CHN50, endochitinase; IPK, inositol 1,3,4,5,6-pentakisphosphate 2-kinase; SuR, acetolactate synthase; Hax3-box, a promoter region of Hax3 (member of the AvrBs3 family of transcription activator-like effectors (TALEs)); ADH, alcohol dehydrogenase; TT4, transparent testa4; DLC, Dicer-like; RDR, RNA-dependent RNA polymerase; HEN, hua enhancer; LG1, Liguleless 1; ABI4, ABA-insensitive 4; pat, phosphinothricin N-acetyltransferase; NHEJ, non-homologous end joining; VAGE, virus-aided gene expression; ZFN, zinc finger nucleases; TALEN, TALE nucleases.

TobaccoTransgeneT-DNA, transient expressionI-SceIPositive/negative selection, site-specific, NHEJ-mediated integrationSalomon and Puchta (1998)
TobaccoTransgeneT-DNA, transient expressionI-CeuIReconstitution of defective kanamycin-resistant gene, NHEJ-mediated integrationChilton and Que (2003)
TobaccoTransgeneT-DNA, transient expressionI-SceIPCR, reconstitution of defective gus gene, NHEJ-mediated integrationTzfira et al. (2003)
BY2TransgeneT-DNA, transient expressionZFNReconstitution of defective pat gene, HR-mediated integrationCai et al. (2009)
BY2CHN50T-DNA, transient expressionZFNBialaphos-resistant BY2 isolates, HR-mediated integrationCai et al. (2009)
TobaccoCHN50T-DNA, transient expressionZFNBialaphos-resistant plants, HR-mediated integrationCai et al. (2009)
MaizeIPK1T-DNA, transient expressionZFNBialaphos-resistant plants, HR-mediated integrationShukla et al. (2009)
TobaccoTransgeneVAGEZFNSite-specific mutagenesis, reconstitution of GUS expressionMarton et al. (2010)
PetuniaTransgeneVAGEZFNSite-specific mutagenesis, reconstitution of GUS expressionMarton et al. (2010)
TobaccoSuRA, SuRBPlasmid, transient expressionZFNHR-mediated integration, resistance to herbicidesTownsend et al. (2009)
ArabidopsisTransgeneTransgenic, heat-shock inducibleZFNSite-specific mutagenesis, PCRLloyd et al. (2005)
TobaccoTransgeneT-DNA, transient expressionZFNSite-specific mutagenesis, reconstitution of GUS expressionTovkach et al. (2009)
ArabidopsisTransgeneTransgenic, heat-shock inducibleZFNSite-specific mutagenesis, reconstitution of GUS expressionTovkach et al. (2009)
TobaccoTransgeneT-DNA, transient expressionZFNReconstitution of kanamycin resistance and GUS expressionWright et al. (2005)
TobaccoHax3-boxT-DNA, transient expressionTALENSite-specific mutagenesis, reconstitution of GUS expressionMahfouz et al. (2011)
ArabidopsisTransgeneTransgenic, tissue specificZFNSite-specific mutagenesis, reconstitution of GUS expressionEven-Faitelson et al. (2011)
ArabidopsisADH1Plasmid, transient expressionTALENSite-specific mutagenesis, PCRCermak et al. (2011)
TobaccoTransgeneTransgenic, overexpressionZFNTransgene removal, elimination of GUS genePetolino et al. (2010)
ArabidopsisADH1, TT4Transgenic, oestrogen inducibleZFNSite-specific mutagenesis of two alleles, PCRZhang et al. (2010)
Soya beanTransgeneTransgenic tissues, oestrogen inducibleZFNSite-specific mutagenesis, PCRCurtin et al. (2011)
Soya beanDCL, RDR, HENTransgenic tissues, oestrogen inducibleZFNSite-specific mutagenesis, PCRCurtin et al. (2011)
Soya beanDCL4bTransgenic tissues, oestrogen inducibleZFNSite-specific mutagenesis, heritable mutation, PCRCurtin et al. (2011)
MaizeLG1 promoterTransgenic, overexpressionI-CreISite-specific mutagenesis, heritable mutation, PCRGao et al. (2010)
ArabidopsisABI4Transgenic, heat-shock inducibleZFNSite-specific mutagenesis, heritable mutation, ABA-insensitive phenotypeOsakabe et al. (2010)
ArabidopsisTransgeneTransgenic, tissue-specific, chemically induced and constitutive expressionZFNSite-specific mutagenesis, HR-mediated integrationde Pater et al. (2009)
Figure 1.

 Strategies for novel restriction enzymes (NRE)-mediated genome editing in plants. Inducing a single double-strand breaks (DSB) at the target sequence (shown here in green) by a single NRE (which is coded by one or two peptides) can lead to non-homologous end joining (NHEJ)-mediated site-specific mutagenesis. The latter is typically characterized by a single-nucleotide change (top panel, left) or deletion of one to several nucleotides (top panel, centre). Concurrent induction of two physically remote DSBs by a pair of NREs (i.e. up to four monomers) in a given chromosome may lead to chromosomal deletion (top panel, right). When combined with the delivery of a donor DNA (shown here in red), induction of a single DSB can lead to HR-mediated gene replacement (bottom panel, left) or gene insertion (bottom panel, right). Donor DNA can also integrate via NHEJ, leading to gene insertion (bottom panel, right) or gene replacement (bottom panel, centre). The latter, however, requires the concurrent induction of two DSBs by a pair of NREs.

Gene targeting via induction of genomic DSBs depends not only on the availability of the proper NREs but also on tools for their optimal delivery into the target cells as well as methods for the regeneration and selection of targeted plants. Naturally, selecting the proper NREs depends on the type of application for which they are destined. Thus, for example, site-specific mutagenesis, deletion or replacement of native genes will require the construction of novel enzymes capable of targeting those native sequences. Such enzymes will need to be designed, constructed and validated both in vivo and in vitro before they can be used for gene-targeting experiments in target plants. Targeting of transgenic sequences, on the other hand, can potentially be achieved using existing enzymes if these sequences are predesigned to carry unique recognition sequences. Here too, the NREs, and in particular those that have been designed and used for targeting experiments in non-plant species, will need to be tested and validated in living plant cells before their deployment for targeting experiments. Once designed, constructed and validated, users of this technology are presented with a variety of options for enzyme delivery into the target plants. Stable and transient expression of the NREs, with or without the co-delivery of foreign DNA, has all been used for targeting native and transgenic sequences in plants. It should be noted that screening and selection of targeted plants, and in particular the targeting of native sequences, may be technically challenging, and a variety of selection methods and screening schemes have been deployed for the detection of targeted plants. Below, we describe the key features of the different NREs used for targeting experiments in plants, as well as the main methods for their validation. We then discuss methods for NRE delivery and selection of targeted plants using examples from recent targeting experiments in plants.

Reagents and methods for induction of genomic DSBs

Novel restriction enzymes

There are currently three types of NREs suitable for inducing site-specific genomic DSBs in plants and other organisms. The first type includes homing endonucleases, which are also referred to as rare cutters or mega-nucleases. Homing endonucleases are restriction enzymes that have none or just a few natural recognition sites in the genomes of many eukaryotes. These nucleases are highly specific and show precise cleavage of their target. Rare cutters have been successfully used for targeting transgenic and native sequences in plants (Puchta, 1998; Salomon and Puchta, 1998; Chilton and Que, 2003; Tzfira et al., 2003; Yang et al., 2009; Gao et al., 2010) (Table 2) and other organisms. Nevertheless, while engineered rare cutters have been used to target native genes, re-engineering rare cutters for novel specificities has proven rather difficult (Smith et al., 2006; Fajardo-Sanchez et al., 2008; Arnould et al., 2011). More recently, Barzel et al. (2011) developed a unique algorithm and database that allow detecting a much wider range of naturally occurring homing endonucleases with specificities to various human genes. Whether similar enzymes with specificities to plant genomes can be detected and whether they can be used for plant genome editing have yet to be determined; however, a recent report by Gao et al. (2010), who used a designed I-CreI to target a native sequence adjacent to the LG1 gene promoter, demonstrated the potential of homing endonucleases for editing native plant sequences.

The second type of NREs are zinc finger nucleases (ZFNs)—artificial restriction enzymes that consist of a synthetic C2H2 zinc finger DNA-binding domain fused to the DNA cleavage domain of the restriction enzyme FokI (Durai et al., 2005; Porteus and Carroll, 2005; Cathomen and Joung, 2008). In ZFNs, the DNA-binding domain is typically composed of three to five ‘finger’ modules, each of which is capable of recognizing a 3-bp-long sequence. Typically, ZFNs function as dimers, and the two monomers are designed to bind within a distance of 5–7 nucleotides from each other, producing an enzyme capable of targeting an extremely long, unique DNA sequence (Durai et al., 2005; Porteus and Carroll, 2005; Cathomen and Joung, 2008). Several methods have been developed for the assembly of novel ZFNs, including ‘modular assembly’ (Wright et al., 2006; Ramirez et al., 2008), the oligomerized pool engineering (OPEN) strategy (Maeder et al., 2008, 2009) and the context-dependent assembly method (Sander et al., 2011). In addition, ZFNs constructed by Sangamo BioSciences’s proprietary technology are available through Sigma-Aldrich CompoZr service (St. Louis, MO, USA). It is important to note that while ZFN-mediated genome modification has proven useful in a variety of organisms (e.g. Urnov et al., 2005, 2010; Beumer et al., 2006; Carroll et al., 2008; Meng et al., 2008; Minczuk et al., 2008; Lei et al., 2011; Young et al., 2011), many ZFNs have been reported to be toxic (Bibikova et al., 2002; Porteus, 2006; Pruett-Miller et al., 2008), most likely due to the creation of non-target DSBs. Regulating ZFN protein levels in the infected cell and redesigning the FokI cleavage domain to create obligate heterodimers (Szczepek et al., 2007; Pruett-Miller et al., 2009; Guo et al., 2011; Ramalingam et al., 2011) are two available strategies that address the potential toxicity of ZFN expression in living cells. In contrast to the complexity involved in re-engineering natural homing endonucleases (Smith et al., 2006; Fajardo-Sanchez et al., 2008; Arnould et al., 2011), the assembly of ZFNs with novel specificities is much simpler and a relatively large number of plant species, transgenes and native genes have been targeted using these enzymes (Table 2). Future advances in plant genome editing are therefore likely to depend on the further development of ZFN design, assembly, validation and expression technologies, tools and protocols.

The third type of NREs are the TALE nucleases (TALENs)—artificial restriction enzymes that are similar in structure and mode of action to ZFNs, except that their DNA-binding domain is based on transcription activator-like effectors (TALEs) (Cermak et al., 2011; DeFrancesco, 2011; Hockemeyer et al., 2011; Li et al., 2011b; Mahfouz et al., 2011; Mussolino et al., 2011; Wood et al., 2011). In nature, TALEs are produced by Xanthomonas, delivered into the plant cell nucleus and transcriptionally activate gene expression in infected cells (Gu et al., 2005). TALEs typically contain several amino-acid repeats that bind DNA. Each repeat contains 34 highly conserved amino acids with variability in residues 12 and 13, known as repeat variable di-residues (RVDs). Interestingly, each nucleotide in the target site is recognized by these RVDs, suggesting a single repeat per nucleotide. Deciphering TALEs’ DNA-binding code might enable the development of a new type of nuclease with novel specificities, and several reports have demonstrated that these enzymes can be used for genome editing in plant species (Table 2).

Users of NRE-mediated gene-targeting technology can thus select from a relatively wide repertoire of enzyme types and methods for their design and assembly for targeting experiments. We refer the readers to several excellent reviews and research papers for in-depth discussions and technical details on these NREs (Wright et al., 2006; Maeder et al., 2008, 2009; Ramirez et al., 2008; Cermak et al., 2011; Li et al., 2011b; Mahfouz et al., 2011; Sander et al., 2011).

Validation systems for functional novel restriction enzymes

Once designed, newly assembled NREs need to be validated before they can be used in gene-targeting experiments. Owing to their central role in genome editing in various animals, as well as in both animal and human cell lines, several systems have been developed to validate the functionality of newly assembled ZFNs, both in vitro and in vivo. While we focus in this section on methods and tools for ZFN analysis and validation, many of these systems and assays can be adapted for analysis and validation of engineered homing endonucleases and TALENs.

A prerequisite for using any type of NRE for gene-targeting experiments is that they will be able to efficiently digest their target sequence. Several in vitro digestion assays have been developed to evaluate the digestion properties of newly assembled NREs. Thus, for example, an in vitro transcription–translation assay was developed by Mani et al. (2005) to investigate the cleavage properties of ZFNs. In their assay, the authors constructed dedicated bacterial expression vectors that facilitated the construction of ZFNs by fusing newly assembled zinc finger DNA-binding domains to the FokI domain, as well as target plasmids that had been engineered to carry ZFN target sites. Digestion of the target plasmids by ZFN was analysed by gel electrophoresis. In vitro digestion assays have also been used to characterize rare-cutting restriction enzyme specificities. Thus, for example, over 20 different in vitro target sequences were characterized for I-CreI and I-PpoI endonucleases (Argast et al., 1998), and over 60 possible in vitro target sites were characterized for I-TevI (Bryk et al., 1993). More recently, Tovkach et al. (2011) used an in vitro digestion assay to investigate ZFN target site specificities. The authors constructed a dedicated ZFN expression plasmid, used it to express the enzyme and tested the enzyme’s specificities using a collection of target plasmids that had been engineered to carry a set of possible target sites (Figure 2) predicted from public data on ZFN recognition helices (Dreier et al., 2001; Kolb et al., 2005). In another example, Cathomen and Sollu, (2010) described a simple and rapid method for in vitro analysis of several parameters of ZFN activity, including DNA-binding specificity, dimerization kinetics and catalytic activity. Users can thus choose from a selection of plasmids and assays for initial validation and characterization of their novel enzymes.

Figure 2.

 Example of an in vitro digestion assay of a newly assembled novel restriction enzymes (NRE). The predicted zinc finger nucleases (ZFN)-10 monomer target site-binding degeneracy was tested (panel a) using a collection of target pSAT6-TS substrate plasmids (panel b), which were engineered to carry modified semi-palindromic ZFN-10 target sites (ZFN-10-TS) (panel c). The digestion efficiency of different target sites (indicated as per cent relative to the original ATGGATGCA monomer-binding site, panel c), determined by gel electrophoresis (panel d), may be useful for predicting possible off-site in vivo gene targeting. Image reproduced from Tovkach et al. (2011) with permission.

Another approach to investigating and validating the targeting properties of newly assembled ZFNs was described by Wright et al. (2006), who produced a battery of reagents and protocols that are useful for assembly of ZFNs by modular assembly. More specifically, the authors described the use of a bacterial two-hybrid reporter assay, in which the binding activity of the ZFN DNA-binding domains was measured by activation of a lacZ reporter gene. Supported by a comprehensive set of vectors suitable for assembly of novel zinc finger DNA-binding proteins and for expression of ZFNs in plant and animal expression vectors, this system facilitates the introduction of validated ZFNs for targeting experiments in plant and animal systems. The bacterial two-hybrid reporter assay has also been used to validate various enzymes developed by the OPEN system (Wright et al., 2006; Maeder et al., 2009) and, more recently, by the context-dependent assembly (CoDA) method (Sander et al., 2011), yielding enzymes capable of targeting genes in tobacco and soya bean, and other, non-plant species.

It is important to note that the above-described and other in vitro and in vivo assays only provide first evidence of the activity of newly assembled NREs; further validation may be required before they can be deployed for genome-editing experiments. Indeed, other types of assays have been developed for more direct evaluation of NREs (mainly ZFN) activity in living cells. The in vivo gfp repair assay, for example, was developed as a means to assess the ability of newly assembled ZFNs to target a chromosomally amended transgene in mammalian cells (Porteus and Baltimore, 2003). In this assay, ZFN activity was measured by induction of HR-mediated repair of a mutated, non-functional gfp that was engineered to contain a ZFN target site. Similarly, HR-mediated repair of chromosomally embedded MEL1 (alpha-galactosidase-encoding gene) and lacZ, in which ZFN target sites had been engineered into their disrupted coding sequences, was used for the analyses of ZFN activity in yeast cells (Doyon et al., 2008; Townsend et al., 2009). It is worth noting that yeast and mammalian cell repair assays have been used to validate ZFNs targeting the inositol-1,3,4,5,6-pentakisphosphate 2-kinase (IPK)-encoding gene (IPK1) in maize plants (Shukla et al., 2009) and the ALS-encoding genes (SuRA and SuRB) in tobacco plants (Townsend et al., 2009). In addition, the yeast HR-mediated lacZ repair assay has been used for in vivo analysis of novel TALENs (Christian et al., 2010; Cermak et al., 2011; Li et al., 2011a).

Plant-specific validation assays and experimental strategies have also been developed. Wright et al. (2005), for example, demonstrated that ZFNs can be used for HR-mediated gene replacement in plant cells using a mutated gus::nptII fusion marker gene repair assay. The authors first incorporated gus::nptII, which was missing parts of the GUS- and NPTII-coding sequences and was engineered to carry a ZFN recognition site, into the chromosome of tobacco plants. ZFN expression, coupled with the delivery of a donor, repair DNA, resulted in regeneration of kanamycin-resistant and GUS-expressing calluses. This system was used not only the validation of ZFN activity in the target cell, but also for assessing the recombination of the foreign (donor) DNA into the break site. In another example, Cai et al. (2009) validated ZFN activity in a tobacco BY2 cell line by measuring intrachromosomal recombination of a fragmented GFP-encoding sequence: a vector carrying two partially overlapping fragments of the GFP-coding sequence, separated by a spacer DNA that was engineered to carry one type of ZFN recognition site (designated here as ZFN-1), was incorporated into the tobacco BY2 cell line genome. ZFN-1 activity was assayed by monitoring GFP expression in cells transformed by ZFN-1-expressing construct. Furthermore, the assay vector was also engineered to carry the 3′ end of the selectable marker phosphinothricin N-acetyltransferase (pat) and two other (identical) ZFN recognition sites (designated here as ZFN-2), which flanked the fragmented GFP-encoding sequence. This design allowed the activity of ZFN-2 to be measured by the regeneration of glufosinate-resistant cells upon delivery of the ZFN-2 expression construct and a pat repair donor DNA. The clear advantage of the above assays is that they allow not only for detection of ZFN activity in target cells but also for the regeneration of targeted tissues. The latter may facilitate the molecular characterization of targeting events in whole plants and/or tissues.

A more simplified reporter gene correction assay was developed by Tovkach et al. (2009), who monitored ZFN expression in plant cells by NHEJ-mediated repair of a mutated gus gene. The assay, which was part of a battery of ZFN functional assays (Figure 3), was composed of gus engineered to carry a ZFN target site and a stop codon at its 5′ end. By incorporating the mutated gus expression cassettes into a binary vector transformation system (Chung et al., 2005; Tzfira et al., 2005), the assay could be performed by transient and stable transfection and even in whole plants (Figure 3). As these assays are based on reconstruction of GUS, which is widely used in many plant species as a sensitive transformation marker (Jefferson et al., 1987; Hull and Devic, 1995), and on the use of Agrobacterium-mediated transformation systems, they can potentially be quickly adapted for the analysis of NREs in a wide variety of plant species and target cells. Indeed, the transgene repair assay proved instrumental during the development of a virus-based ZFN delivery system for gene targeting in tobacco and petunia plants, described further on (see also Figure 4). Furthermore, the mutated gus gene recovery assay has been recently adapted by Mahfouz et al. (2011) for in vivo analysis of TALENs in tobacco cells.

Figure 3.

 A set of in-planta ZFN validation assays and their vector systems. Three distinct assays were developed by Tovkach et al. (2011) for step-by-step validation of newly assembled zinc finger nucleases (ZFNs) (panel a). The assays are based on combining the mutated GUS expression cassette (designated as ZFN-TS*::GUS) with a ZFN expression cassette and monitoring for reconstruction of GUS expression in plant cells. The mutated GUS-coding sequence (exemplified in panel b) is designed to carry a stop codon within the spacer of a target sequence. Incorrect repair of the ZFN-induced double-strand breaks (DSB) can lead to reactivation of GUS expression by non-homologous end joining (NHEJ)-mediated elimination of the stop codon. In the T-DNA repair assay, both the mutated GUS expression cassette and a constitutive zinc finger nucleases (ZFN) expression cassette are cloned into a single plant transformation vector (i.e. pRCS2.[ZFN][ZFN-TS*::GUS]), and GUS expression is monitored in transiently infected cells, as shown for tobacco-infected leaves (panel c). In the transgene repair assay, the mutated GUS expression cassette is first integrated into the chromosome of the plant cell using a stable transformation vector (i.e. pRCS2.[KAN][ZFN-TS*::GUS]), and the cells are then retransformed with a constitutive ZFN expression vector (i.e. pRCS2.[ZFN]). In the whole-plant DNA repair assay, a stable plant genetic transformation vector (i.e. pRCS2.[KAN][hsp.ZFN][ZFN-TS*::GUS]) is designed to carry the mutated GUS expression cassette and an inducible ZFN expression cassette (e.g. using heat-shock-inducible promoter, hsp). The vector is then used for the production of transgenic plants in which ZFN activity can be monitored in various organs and at different developmental stages, as shown in panel d. Panels reproduced from Tovkach et al. (2011) with permission.

Figure 4.

 Viral vector-mediated site-specific mutagenesis in plant cells. Petunia plants were engineered to carry a mutated, non-functional GUS reporter gene, which was activated upon infection by virus-aided gene expression (VAGE) of zinc finger nucleases (ZFNs). The ability of the viral vector to travel into newly developed tissues (i.e. buds and primordia) allowed the generation of mutated plantlets in which the reconstructed GUS gene was stably inherited in the next generation (shown here by GUS-expressing seedlings). Panels reproduced from Marton et al. (2010) with permission.

Molecular screening of targeting events

While visual and selectable assays can be used for preliminary analysis and validation of NREs in vivo and in vitro, molecular analysis (i.e. cloning and sequencing) of the targeting events is essential for final proof of the stability of these events in the targeted genome. Molecular analysis can also be useful for estimating the efficiency of targeting, as well as for providing molecular information on the outcome of the targeting events. Lloyd et al. (2005), for example, demonstrated the usefulness of direct molecular analysis for determining the frequency of gene targeting in Arabidopsis plants. In their assay, an EcoRI recognition site was engineered into the ZFN recognition sequences of an artificial target site that was incorporated into the Arabidopsis genome. Because ZFN-mediated targeting often resulted in interruption of the EcoRI recognition site, digestion of total plant DNA with EcoRI before its amplification by PCR allowed for enrichment of targeted plant DNA molecules. This method not only allowed for estimating the efficiency of targeted mutagenesis in Arabidopsis but also assisted with manual selection of mutated Arabidopsis plants via sampling of putative mutated inflorescence stems (Lloyd et al., 2005). Molecular analysis can also be used to validate NRE activity on native genome sequences. Zhang et al. (2010), for example, validated the activity of ZFN designed to target ADH1 and TT4 by PCR amplification of NlaIII- (for ADH1) or NspI (for TT4)-digested total DNA from Arabidopsis protoplasts, which had been transiently transformed by ZFN expression constructs. As NlaIII and NspI were naturally present within the ZFN recognition target sites of ADH1 and TT4, respectively, the isolation of PCR fragments that were not cleaved by NlaIII or NspI allowed for the detection of ZFN-induced mutagenesis events in Arabidopsis protoplasts. These assays were later adapted for validation of several ZFNs designed to target soya bean genes in transgenic soya bean hairy root tissues (Curtin et al., 2011), for validation of TALENs designed to target ADH1 in Arabidopsis protoplasts and for isolation of ZFN-mediated mutagenesis events in transgene sequences in tobacco, petunia and Arabidopsis plants (de Pater et al., 2009; Tovkach et al., 2009; Marton et al., 2010).

Other molecular approaches have been taken to evaluate ZFN activity in various living cells, including those of plants. The CEL I endonuclease assay, for example, has been used for analysis of ZFN activity in animal and human cell lines (Lombardo et al., 2007; Miller et al., 2007; Maeder et al., 2008; Santiago et al., 2008) and can be adapted, in principle, for plant research. In this assay, ZFN activity is detected by co-amplification of mutated and wild-type target sequences, followed by denaturation and re-annealing of the mixed amplified sequences and their digestion with CEL 1. Detection of fragment polymorphism among the digested fragments indicates the presence of mutated sequences. In a similar approach, Osakabe et al. (2010) used Surveyor nuclease (a key enzyme in the transgenomic Surveyor® mutation detection kit from Transgenomics, Omaha, NE, USA) to detect non-homogeneity among PCR fragments derived from a mixture of mutated and wild-type target sequences that were amplified from transgenic Arabidopsis plants engineered to carry ZFNs for the targeting of ABI4. Finally, because site-specific mutagenesis often results in insertions and/or deletions at the break site, in vivo NRE expression may lead to size polymorphism of the targeted DNA fragment, which can be detected by DNA sequencing. Indeed, pyrosequencing was used to analyse targeting events in tobacco protoplasts (Townsend et al., 2009) and in cultured maize cells (Shukla et al., 2009), which had been transiently transformed by ZFNs for SuR and IPK, respectively.

Enzyme expression systems and genome targeting in plants

There are two distinct strategies for the targeting of plant genomes (Figure 1). In the first, targeting is solely dependent on the use of NREs, while in the second, targeting also requires the delivery of a foreign donor DNA molecule. With the exception of single-monomer NREs (e.g. when targeting semi-palindromic sequences by ZFN or TALEN monomers or when using certain homing endonucleases), targeting experiments call for simultaneous expression of two or more NRE monomers in a single target cell. Delivery of two monomers can be achieved by co-delivery of two independent expression vectors, by using a dual-gene transformation vector or bicistronic transformation vectors. All of these vectors can potentially be used in transient and stable transformation experiments, with or without delivery of a donor DNA molecule. In addition, viral vectors have been developed for transient NRE expression in plant cells, as we describe further on.

Transient expression systems

In their pioneering study, Salomon and Puchta (1998) used Agrobacterium-mediated genetic transformation to transiently express I-SceI in transgenic tobacco plants and discovered that the I-SceI-encoding T-DNA molecules can integrate into the I-SceI-induced DSB. Whether it was the actual I-SceI-expressing T-DNA or another I-SceI-coding T-DNA molecule that co-entered the target cell and integrated into the break site was not determined, but in later reports (Chilton and Que, 2003; Tzfira et al., 2003), co-transformation of a donor DNA with homing endonuclease-expressing T-DNAs was used to target a second T-DNA (donor) molecule into the genomic DSBs (Table 2). The high frequency of non-selective NHEJ-mediated site-specific T-DNA integration (2.58%) (Tzfira et al., 2003) suggests that T-DNA molecules may be preferentially directed to genomic DSBs (Tzfira et al., 2004; Ziemienowicz et al., 2008; Dafny-Yelin et al., 2009). It is important to note that I-SceI- and I-CeuI-targeted plants (Chilton and Que, 2003; Tzfira et al., 2003) that were free of restriction enzyme-encoding T-DNA were recovered, indicating that using T-DNA molecules can be useful for transient NRE expression and delivery of donor DNA in plants.

Co-transformation of donor and ZFN-expressing DNA molecules was also used by Wright et al. (2006), who targeted and repaired a non-functional reporter gene by HR-mediated gene replacement in tobacco cells (Table 2). The authors first produced a collection of target transgenic tobacco plants, in which the non-functional gus::nptII, which was also engineered to carry a ZFN recognition site, was integrated at different genomic locations in different transgenic lines. Next, transgenic protoplasts were co-transformed with ZFN-expressing and donor plasmid DNA molecules and were selected for HR-mediated kanamycin resistance. The average frequency of kanamycin-resistant calluses was 4.5 × 10−3. The frequency of the kanamycin-resistant calluses was over threefold higher in co-transformation experiments of ZFN and donor DNA, compared with transformation of just donor DNA. Note that the authors recovered kanamycin-resistant callus in the absence of ZFN-expressing T-DNA as well, by promoter trapping of the nptII sequence. This experimental scheme, which allowed the authors to estimate the ratio between HR and NHEJ targeting events as 1:5.9, further validated the usefulness of co-transformation approaches for transient expression of NRE and delivery of donor DNA molecules into the target plants.

Co-transformation was also used in targeting experiments in BY2 tobacco cells, where a donor DNA was directed to a pre-integrated defective reporter gene as well as to the endogenous locus (Cai et al., 2009). For targeting of the transgenic locus, which was pre-engineered to carry disabled gfp and pat genes and was randomly integrated into the genome of BY2 tobacco cells, a single ZFN monomer-expressing T-DNA molecule was co-transformed with a donor T-DNA molecule into the transgenic BY2 lines. HR-mediated targeting events were selected and identified by replacement of the disabled gfp and reconstitution of a functional pat expression cassette. The latter derived from precise recombination between the donor and the disabled pat gene on the BY2 chromosome, as determined by isolation of bialaphos-resistant lines, Southern blot analysis and DNA sequencing of selected events. For targeting of a native locus in BY2 cells (i.e. the endochitinase gene CHN50), a pair of CHN50-specific ZFNs and a donor DNA were co-delivered into the target cells. To facilitate the transient co-expression of the two monomers of CHN50 ZFN in the target cells, the authors constructed a dual-gene expression cassette by flanking the 2A sequence with the CHN50 ZFN monomer-coding sequences and expressing the fusion under the control of the strong and constitutive CsVMV promoter. Here too, Agrobacterium-mediated genetic co-transformation was used for delivery of the CHN50-ZFN and donor DNA into BY2 cells, as well as into tobacco leaf disc cells. A dozen targeted bialaphos-resistant BY2 isolates and five transgenic tobacco plants were produced, and PCR and sequence analyses confirmed that they derived from HR-mediated gene replacement events. Although Southern blot analysis revealed that some of the targeted lines also carried randomly integrated donor DNA, these studies further supported the notion that co-transformation is a viable system for the delivery of NRE and donor DNA molecules into target plants. Indeed, the co-transformation approach was also used for co-delivery of dimer ZFNs, designed to target the ALS-encoding genes SuRA and SuRB in tobacco (Maeder et al., 2008, 2009). The enzymes were delivered as individual plant expression plasmids into tobacco protoplasts together with a third plasmid, carrying a constitutive kanamycin-resistant selection cassette, which was used to aid with the selection of transformed cells. Three out of 66 kanamycin-resistant transgenic plants carried a single-base deletion within the SuRA target sequence, but no SuRB mutants were detected (Maeder et al., 2008). This was later attributed not only to the design of the NRE but also to other possible factors such as chromatin structure and DNA methylation, as suggested in a study where both SuR genes were co-targeted using a different ZFN (Townsend et al., 2009).

Co-targeting of SuRA and SuRB in tobacco was not limited to detection of mutation events by pyrosequencing, but was also achieved by HR-mediated gene replacement (Townsend et al., 2009). Using the co-transformation approach, in which pairs of SuR-ZFN monomers and a third (donor) DNA molecule (which was partially homologous to the SuR loci, but carried specific mutations) were delivered into a target protoplast, HR-mediated targeted herbicide-resistant calluses and plants were recovered (Townsend et al., 2009). The estimated rate of ZFN-induced genome editing ranged between 2.4–5.3%, of which 0.2–4.0% was attributed to HR-dependent recombination between donor DNA molecules and SuR loci. Note that over 2% of the targeting events occurred more than 1300 bp away from the ZFN cleavage site (Townsend et al., 2009). Similarly, Cai et al. (2009) observed targeting events at a distance of 3000 bp from the ZFN-induced DSBs. Taken together, these observations suggest that a certain flexibility may exist when designing NRE for gene targeting. It is also worth noting that among the 47 different herbicide-resistant calluses obtained by co-transformation of donor and ZFN-expressing constructs, 19 were modified at multiple SuRA loci (by either HR or NHEJ) and 10 fully developed SuRA and SuRB double mutants were generated (Townsend et al., 2009). This study also demonstrated that co-targeting of similar genes can potentially be achieved by a single pair of ZFN monomers. Indeed, co-targeting of a duplicated gene using a single pair of ZFN monomers was later reported in soya bean (Curtin et al., 2011), as described further on.

HR-mediated gene targeting by co-transformation of a ZFN expression construct and donor DNA was also proven instrumental for targeting IPK in maize plants (Shukla et al., 2009). Maize plants carry two IPK paralogs (IPK1 and IPK2) that share 98% sequence identity in their coding sequences. Several ZFNs have been developed to target IPK1, selected for targeting based on its expression pattern. Deep-sequencing analysis of DNA extracted from maize cell culture transiently transformed by ZFN expression vectors was used for initial screening of the different ZFNs (Shukla et al., 2009). Two distinct IPK1-homologous donor DNA molecules were constructed: the autonomous donor molecule was designed to carry a complete pat expression system, while the non-autonomous donor carried a promoterless pat. Targeting by the non-autonomous vector was dependent on functional trapping of pat by the IPK1 promoter. Co-delivery of ZFN with donor DNA resulted in HR-mediated gene replacement and recovery of herbicide-resistant calluses in which one or both of the IPK1 alleles were targeted, as confirmed by molecular analysis (Shukla et al., 2009). Interestingly, the number of herbicide-resistant calluses was lower, while the frequency of HR-mediated targeting events (out of total targeted and random integration events) was higher for the non-autonomous donor than for the autonomous one. Furthermore, while calluses obtained from random integration events were characterized by multiple insertions, those obtained from HR-mediated integration events were characterized by a single, site-specific integration of the pat gene. As DSBs may act as ‘hot spots’ for NHEJ-mediated T-DNA integration (Tzfira et al., 2004; Ziemienowicz et al., 2008; Dafny-Yelin et al., 2009), we suggest that the ZFN expression construct may not have been delivered into the former cells, thus leaving the donor DNA molecules free to integrate at random, naturally occurring, but short-lived DSB sites. ZFN expression on the other hand may leave the DSBs available for integration for longer periods of time, and perhaps direct several T-DNAs to the break site, where only one T-DNA molecule will eventually integrate via HR. While we can only speculate on the possible mechanism by which donor DNA molecules were directed to HR and not NHEJ in targeted maize cells, these observations demonstrate the feasibility of obtaining HR-mediated targeted plants, which are free of additional, randomly integrated, foreign DNA molecules.

It should be noted that while IPK2 is nearly identical to IPK1, it remained intact in five independent IPK1-targeted plants, as determined by sequence analysis. In addition, sequence analysis of five more of the most probable off-target sites for the IPK1 ZFNs revealed that they were all true to type. These observations, as well as genotyping, segregation analysis and phenotypical characterization of mutant lines and their progeny, indicate that targeted plants with single and well-defined foreign DNA integration can be produced by transient expression of highly specific ZFNs and co-transformation of donor DNA.

Transgenic expression systems

A transgenic approach was used by Lloyd et al. (2005) who were the first to demonstrate ZFNs’ applicability for site-specific mutagenesis using a genomically inserted target site in Arabidopsis plants. Transgenic Arabidopsis plants were produced in which the QQR ZFN (Bibikova et al., 2001) was driven under the control of a heat-shock promoter. Heat treatment was used to induce QQR ZFN expression at specific developmental stages (i.e. in 10-day-old plants) and the presence of the EcoRI site within the ZFN recognition site facilitated the molecular analysis of targeting events, as already described. By comparing the number of DNA fragments with disrupted and true-to-type EcoRI sites (both amplified from undigested DNA extracted from heat-shock-induced plants), the authors estimated the rate of QQR-induced mutations to be as high as 0.2 per target. Furthermore, sequencing analysis suggested that most of the ZFN-induced mutations could potentially lead to functional gene knockout and that about 10% of the offspring from the heat-shock-induced plants carried mutations. T1-mutated seedlings were most likely obtained from mutations in the early-stage L2 cells of the shoot apical meristem of the heat-shock-induced plants.

Transgenic ZFN expression was also used by Tovkach et al. (2009, 2010), who developed a whole-plant DNA repair assay (Figure 3) and used it to demonstrate that in addition to small deletions and/or insertions, ZFN-mediated site-specific mutagenesis can also lead to single-nucleotide replacement at the break site (Tovkach et al., 2009). Single-nucleotide changes were also observed in tobacco and petunia plants, obtained by ZFN expression by viral vectors, as well as by transgenic expression of different ZFNs in Arabidopsis plants, as we describe further on (Marton et al., 2010; Osakabe et al., 2010; Zhang et al., 2010). Transgenic ZFN expression was also instrumental for induction of HR-mediated gene replacement in Arabidopsis when a donor DNA was delivered by flower-dip transformation into transgenic plants (de Pater et al., 2009). More specifically, a target locus, which was composed of functional pat and gfp genes as well as a unique recognition site for two ZFN monomers, was first incorporated into the genome of Arabidopsis plants (de Pater et al., 2009). The ZFN expression cassettes (driven by Rps5 tissue-specific, tamoxifen-inducible, or constitutive 35S promoters) were also stably integrated (either as single or dual ZFN expression constructs) into the target plants. Interestingly, only 2% of the cells of the transgenic plants that were engineered for ZFN overexpression (using the constitutive CaMV 35S promoter) were mutated (de Pater et al., 2009). This low rate of site-specific mutagenesis led the authors to examine whether retransformation of the ZFN-expressing transgenic plants with a donor DNA might lead to its integration via HR. Rps5-ZFN-transgenic plants were selected for HR-mediated targeting experiments, as the rate of site-specific mutagenesis in these plants (where ZFN expression was controlled under the Rps5 promoter, which is active in dividing cells and early embryos) was even lower than in 35S-ZFN-transgenic plants (de Pater et al., 2009). A donor DNA, which was engineered with partial homology to the transgenic target locus and also carried a hygromycin-resistant selection gene, was delivered into Rps5-ZFN by flower-dip transformation: 3000 hygromycin-resistant transgenic Arabidopsis plants were produced and screened by PCR for targeting events (de Pater et al., 2009). Two lines were targeted by HR, and a third line may have derived from a HR recombination event, followed by release and re-integration of the recombinant T-DNA into a different genomic location. The relatively high targeting frequency of 10−3 suggested that the strategy of delivering foreign donor DNA as T-DNA molecule into transgenic, ZFN-expressing plants might be useful for HR-mediated targeting. While the targeted lines also carried randomly integrated T-DNA molecules, these, as well as the ZFN-expressing cassettes, could potentially be segregated out by sexual crosses.

More recently, Zhang et al. (2010) and Osakabe et al. (2010) reported site-specific mutagenesis of native genes in Arabidopsis plants. Here too, the authors elected to use a transgenic approach, but used different types of promoters to express their ZFNs in the target plants. Zhang et al. (2010), who targeted ADH1 and TT4, produced hygromycin-resistant transgenic Arabidopsis plants in which the ADH1 and TT4 ZFNs were expressed under the control of an oestrogen-inducible promoter. For each target gene, a pair of ZFN monomers was expressed as an in-frame fusion with the T2A peptide. For induction of site-specific mutagenesis, T0 seeds were germinated in the presence of 17β-oestradiol; 10-day-old hygromycin-resistant seedlings were then tested for the presence of ZFN-induced ADH1 and TT4 mutations by PCR analysis and DNA sequencing. Deletions, insertions and a single-nucleotide substitution event were observed, with an estimated somatic mutation rate of 16% of the alleles for ADH1 and 7% of the alleles for TT4, as estimated by sequencing of cloned PCR fragments (Zhang et al., 2010). Mutated Arabidopsis plants were then produced by collecting seeds from 17β-oestradiol-induced T1 plants that were left to develop, grow and set seed: 11 of 16 (69%) of the T1 plants yielded adh1 mutants, of which 18% (2 of 11) were mutated at both alleles, as determined by segregation analysis of their progeny. In addition, 10 of 30 (33%) T1 plants yielded tt4 mutants, all of which were mutated in both alleles. The high rate of site-specific mutagenesis showed in this study could be attributed in part to the use of the strong 17β-oestradiol promoter, which can potentially reach higher expression levels than the 35S constitutive promoter (Zuo et al., 2000).

While a heat-shock-induced promoter yielded a lower frequency (10%) of mutated progeny (Zhang et al., 2010), it was still found useful for the production of abi4 mutant lines (Osakabe et al., 2010): Arabidopsis plants were genetically engineered to express the ABI4-ZFNs under a heat-shock-inducible promoter. The enzyme monomers were expressed as a single unit fused by the 2A peptide. Surveyor nuclease assay was used to detect mutations in somatic cells of nine ABI4-ZFN-transgenic lines, with an estimated rate of 0.26–2.86%. Interestingly, 70% of the cloned mutations were classified as substitution mutations, while the remaining 30% were classified as short (1–3 bp long) deletions. When allowed to grow, develop and set seed, two of the heat-shock-induced ABI4-ZFN-transgenic lines produced heterozygous mutant plants, as determined by Surveyor nuclease assay. More specifically, 7 and 3 of the 96 seeds surveyed for each line exhibited a single-base deletion or substitution mutation, respectively, from which T3 homozygous mutants were obtained.

More recently, Even-Faitelson et al. (2011) reported that ZFN expression under the control of EASE leads to site-specific mutagenesis, as determined by reconstruction of a mutated GUS gene. More specifically, the authors revealed that by confining ZFN expression to the egg cell (which is the target of the Agrobacterium T-DNA during flower-dip transformation, Ye et al., 1999), mutated Arabidopsis plants were recovered, which had been derived from targeting events occurring at early stages of embryo development. Clear advantages to this approach are that it does not necessitate exogenous stimuli for ZFN induction, and it restricts ZFN expression to the target tissue, thereby minimizing the risk of ZFN toxicity to the rest of the plant. Another important advantage pointed out by the authors is that as every mutant obtained by EASE-mediated ZFN derives from an independent germinal event, the use of this approach is likely to produce a wide range of new mutated alleles.

Stable NRE expression was also used for targeted mutagenesis in maize and soya bean, with a designed endonuclease and ZFN, respectively (Gao et al., 2010; Curtin et al., 2011). A single-chain homing endonuclease, composed of two designed I-CreI monomers fused into a single polypeptide, was expressed under the maize ubiquitin promoter in transgenic maize (Gao et al., 2010). A total of 781 transgenic plants were screened for putative mutations by PCR, of which 23 lines exhibited monoallelic, biallelic or a chimeric mix of mutations. Sequence analysis of selected T0 plants revealed a wide range of mutations, from deletions of just a couple of base pairs to large (up to 71 bp) deletions. Segregation and PCR analyses of progeny of selected T0 plants confirmed the authenticity and stability of the designed I-CreI-induced mutagenesis in maize. For targeting experiments in soya bean, ZFN-mediated mutagenesis was first assayed in transgenic Agrobacterium rhizogenes-induced hairy roots (Curtin et al., 2011). The roots were also engineered to express 2A-fused monomers of various ZFNs under the control of the oestrogen-inducible promoter. PCR analysis of digested DNA from 17β-oestradiol-induced roots and DNA sequencing of selected clones indicated that the transgenic root assay is efficient and reliable for analyses of ZFNs in soya bean tissues. Stable ZFN expression was also the method of choice for production of mutated soya bean plants: a binary plant transformation vector carrying ZFNs targeting two paralogs of DLC4 was used to produce transgenic soya bean plants in which ZFN expression was induced during in vitro cultivation. Two mutated plants were obtained, one carrying a mutation in DLC4a and the other in DLC4b. Both plants, which were either heterozygous or chimeric for the mutated alleles, were allowed to grow, develop and set seed; while the dlc4a mutant exhibited severe developmental abnormalities, the dlc4b mutant produced a large number of seeds and its progeny were successfully used for recovery of homozygous and heterozygous DLC4b mutant lines. Note that 1 out of 24 T1 heterozygous dlc4b lines was ZFN-free, as determined by PCR analysis, indicating that transgene removal can be achieved by segregation or sexual crosses.

A unique approach for ZFN-mediated transgene deletion was taken by Petolino et al. (2010) who crossed ZFN-overexpressing plants with target transgenic plants, which were engineered to carry a GUS expression cassette that was flanked by recognition sites for the ZFN. Both types of transgenic plants were homozygous for the transgene. A large number of T1 × T1 GUS-negative hybrid plants were obtained, with a frequency of about 35% for one particular cross. PCR and sequencing analyses confirmed that the GUS cassette had indeed been removed, as well as the presence of insertions and deletions at the DSB site. Segregation analysis of F2 progeny demonstrated the heritability of the gus gene deletion trait. This approach can potentially be applied for deletion of native target sequences and chromosomes, as well as in other sexually propagated plants. It is important to note that while deletion of artificial genes (Figure 1) can be achieved using a single ZFN enzyme or even a single ZFN monomer, deletion of chromosomal fragments and NHEJ-mediated native gene replacement may require the co-expression of four ZFN or TALEN monomers, which can be achieved by co-transformation of multiple T-DNA molecules into plant cells (Dafny-Yelin and Tzfira, 2007; Naqvi et al., 2010) or by assembly of polyprotein transformation vectors (Halpin et al., 1999; El Amrani et al., 2004). However, the low efficiency of co-transformation and the formation of complex integration patterns which are often associated with multi-T-DNA and multi-plasmid transformation systems (Windels et al., 2008; Ziemienowicz et al., 2008), and the complexity of assembling large polyprotein transformation vectors, may pose a technical challenge for co-transformation of more than two monomers into plant cells. More recently, Tovkach et al. (2009) described the assembly of a vector system for ZFN validation and expression, which enables the expression of up to four independent ZFN monomer expression cassettes from a single T-DNA molecule. The system, which was built on the basis of a modular plasmid assembly system (Chung et al., 2005; Tzfira et al., 2005; Tovkach et al., 2010), can be easily adapted for assembly of new ZFNs into plant expression vectors, for expression of ZFN monomers under the control of different promoters and for stable transformation using different types of plant selection markers (Figure 5). Furthermore, because ZFN and TALEN monomers share their basic structure, it is likely that this system can be modified and adapted for expression of multi-TALEN monomers, as well as for expression of single and dual peptide-modified homing endonucleases.

Figure 5.

 General structure of a modular multi-ZFN expression system. The system supports the assembly of ZFN expression cassettes (plasmids 2–5) with or without a plant selection expression cassette (plasmid 1) onto the T-DNA region of the plant binary vector pRCS2. The promoter and terminator regions on each expression cassette can be exchanged by any other regulatory sequences, as shown by the exchange of 35S promoter (35SP) and terminator (35ST) with Rubisco promoter (rbcP) and terminator (rbcT) or heat-shock promoter (hspP) and nopaline-synthase terminator (nosT). A PCR-amplified DNA-binding domain (ZFP) can be fused with the FokI domain (illustrated in Plasmid 2). Three selection cassettes (hpt, nptII and bar) can be used for the production of transgenic plants.

Non-transgenic viral expression systems

Although NRE-mediated site-specific mutagenesis, deletion and chromosomal deletion do not require the transfer of a donor DNA into the target plant cells (Figure 1), using plasmid- or T-DNA-mediated gene transfer for transient NRE expression may render the plants transgenic owing to the nature of the direct gene delivery tool. An alternative, non-transgenic approach for NRE expression in plant cells was recently developed by Marton et al. (2010), who demonstrated that RNA-based viral vectors can be used for efficient delivery of ZFNs into a wide range of organs of tobacco and petunia plants. These authors first produced transgenic tobacco and petunia plants, which had been engineered to carry a mutated gus reporter gene (Figure 2B). The plants were then infected with a ZFN-expressing Tobacco rattle virus (TRV) vector, and the virus’s ability to travel from cell to cell, express the ZFNs and mutate the genome of these cells was monitored by GUS expression (Figure 4) and sequencing of targeting events. This indirect, virus-aided gene expression (VAGE) system (Vainstein et al., 2011) enabled the production of non-transgenic mutated tobacco and petunia plants. The stability of the mutation was confirmed by its transmission to the next generation, as confirmed by GUS expression (Figure 4) and DNA sequencing (Marton et al., 2010). As TRV can infect a large number of plant species (Marton et al., 2010; Vainstein et al., 2011) and can travel into a wide variety of cells, tissues and organs, it can potentially be adapted for non-transgenic genome modification in various model and crop plants (Vainstein et al., 2011). Furthermore, because TRV can also travel into existing growing buds, mutations at meristems of growing and developing plants may produce mutated seedlings, similar to the production of mutated seedlings from transgenic Arabidopsis plants (Lloyd et al., 2005; Osakabe et al., 2010; Zhang et al., 2010). The latter might simplify the production of mutated plants by recovering mutated plantlets from infected adult plants without the need for regeneration through tissue culture (Vainstein et al., 2011).

Conclusions and future prospects

Methods for genome modification will likely have a profound effect on gene discovery and the introduction of novel traits into crop plants. The advances in NRE technology (and in particular ZFN technology) for genome editing in animals, animal cells and human cells (reviewed by Carroll, 2004; Porteus and Carroll, 2005; Camenisch et al., 2008; Cathomen and Joung, 2008; Remy et al., 2010; Urnov et al., 2010) are being embraced by plant biologists and biotechnologists, who have demonstrated the use of this advanced technology for genome modification in plant species.

Designing NREs (i.e. ZFNs, TALENs and modified homing endonucleases) is a prerequisite for any gene-targeting experiment, and users of this technology enjoy a wide range of publicly available tools, reagents and protocols for the assembly of novel ZFNs and TALENs (Dreier et al., 2001; Liu et al., 2002; Mani et al., 2005; Carroll et al., 2006; Wright et al., 2006; Sander et al., 2007, 2011; Maeder et al., 2008; Cermak et al., 2011; Lei et al., 2011; Li et al., 2011a, 2011b; Ramalingam et al., 2011; Reyon et al., 2011). ZFNs may also be purchased from Sigma-Aldrich, although current licensing agreements with Sangamo BioSciences and Dow Agrosciences (http://www.dowagro.com) appear to restrict their use for animal research. Worth noting is that while re-engineering homing endonucleases for novel specificities may be technically challenging and time-consuming (Arnould et al., 2006; Smith et al., 2006; Fajardo-Sanchez et al., 2008), existing enzymes may well be usable for targeting of transgenic sequences. Indeed, both artificial and naturally occurring enzymes have been shown useful for genome editing in plant species. Nevertheless, the number of plant species that have been genetically manipulated using this technology (relative to the number of species, which have been genetically transformed) is still small (Table 2).

While enzyme design and assembly are central to the use of this technology for genome editing, methods and tools for enzyme validation and efficient expression in plant cells are no less important to achieving successful targeting events. Here too, users of this technology can choose from a wide range of validation methods and can select an expression system, which will best fit their target plant and its final application. Thus, for example, transgenic overexpression of NRE may be useful for the production of targeted Arabidopsis and other model plants, but may be less desirable for targeting in commercial crop plants. In another example, transient expression of two NRE monomers by co-delivery of two plasmid or T-DNA molecules with or without a third (donor) DNA molecule has proven useful for targeting in several species. Yet, co-transformation is a relatively inefficient process, which may not be applicable to different economically important species (Dafny-Yelin and Tzfira, 2007; Naqvi et al., 2010). Furthermore, this procedure may result in unintentional integration of the transformed molecules into the plant genome, rendering it transgenic. As co-transformation often results in complex integration patterns (Windels et al., 2008; Ziemienowicz et al., 2008), breeding them out by sexual crossing may be difficult and tedious. VAGE, which has been shown useful for delivering ZFNs and other proteins to growing and regenerating tissues of infected plants (Vainstein et al., 2011), may provide a viable alternative for direct gene transfer experiments, at least towards the production of mutated plants. As RNA viral vectors do not integrate into the plant genome, mutated plants derived from VAGE of NREs are likely to be classified as non-transgenic (Marton et al., 2010; Vainstein et al., 2011), which may be advantageous for genome engineering in commercial plant species.

Further progress in NRE-mediated gene-targeting technology is likely to be achieved by further development of novel enzymes, as well as expression and selection tools, and by its deployment for targeting transgenic and native gene target sequences in a wide range of plant species. Nonetheless, it is important to note that the success of this technology may also depend on a better understanding of plant genetic transformation processes and mechanisms of DNA repair in plant cells. Unveiling the mechanisms by which T-DNA molecules are directed to genomic DSBs, the possible effects of DNA structure on the accessibility of NRE to target sequences, the dynamics of DSB repair and the role of specific plant proteins in HR and NHEJ may all contribute to the much needed progress in efficient procedures for genome editing of plant cells.


Our laboratories are supported by US-Israel Binational Agricultural Research and Development (BARD) grants US-4150-08 and US-4322-10 (to Tzvi Tzfira and Alexander Vainstein, respectively), Israel Science Foundation grants 269/09 and 432/10 (to Alexander Vainstein), a BARD Postdoctoral Fellowship FI-409-08 (to Dan Weinthal) and Israeli Chief Scientist Office grants 42440, 44377 (to Ira Marton and Amir Zuker). Alexander Vainstein is an incumbent of the Wolfson Chair in Floriculture. Dan Weinthal has a Kreitman Foundation PostDoctoral Fellowship.