Site-directed nucleases: a paradigmshift in predictable, knowledge-basedplant breedingNancy Podevin1*, Howard V. Davies2, Frank Hartung3,Fabien Nogue4, and Josep M. Casacuberta5

1 The European Food Safety Authority (EFSA), Via Carlo Magno 1A, Parma, Italy2 The James Hutton Institute, Invergowrie, Dundee DD2 5DA, UK3 Julius Kuhn-Institut (JKI), Federal Research Centre for Cultivated Plants, Institute for Biosafety in Plant Biotechnology,

Erwin-Baur-Str. 27, D-06484 Quedlinburg, Germany4 INRA AgroParisTech, IJPB, UMR 1318, INRA Centre de Versailles, route de Saint Cyr, 78026 Versailles CEDEX, France5 Center for Research in Agricultural Genomics (CSIC-IRTA-UAB-UB), Campus UAB, Cerdanyola del Valle s 08193 Barcelona, Spain

Review

Glossary

Double-strand break (DSB): a break in double-stranded DNA in which both

strands have been cleaved. It can originate from byproducts of cell metabolism

such as reactive oxygen species, from environmental exposure to irradiation,

chemical agents, or UV light. DSBs also occur as intermediates in various

biological events, such as V(D)J recombination, transposon movement or

meiotic recombination. See Box 2 for an explanation of the repair mechanisms

for DSBs.

Genetically modified plants (GMPs): see ‘Transgenic’

Homologous recombination (HR): error-free repair of a DSB in DNA, in which

the broken DNA molecule is repaired using a homologous sequence. See Box 4

for details on DNA repair mechanisms.

Indels: Indel is the combined artificial word for either an insertion or a deletion

which results from DNA damage by an error-prone DNA-repair mechanism or

an error during DNA replication.

Nonhomologous end joining (NHEJ): a major pathway of DSB repair, in which

Conventional plant breeding exploits existing geneticvariability and introduces new variability by mutagene-sis. This has proven highly successful in securing foodsupplies for an ever-growing human population. The useof genetically modified plants is a complementary ap-proach but all plant breeding techniques have limita-tions. Here, we discuss how the recent evolution oftargeted mutagenesis and DNA insertion techniquesbased on tailor-made site-directed nucleases (SDNs)provides opportunities to overcome such limitations.Plant breeding companies are exploiting SDNs to devel-op a new generation of crops with new and improvedtraits. Nevertheless, some technical limitations as wellas significant uncertainties on the regulatory status ofSDNs may challenge their use for commercial plantbreeding.

SDNs in plant breedingPlants and their derived food products are the cornerstone ofhuman nutrition, and since Neolithic times plant breedinghas increased yields and improved quality and nutritionalvalue. Plant breeding has therefore delivered to feed anever-growing population, although significant challengesremain [1]. Plant breeding is based on the introduction ofgenetic variability, historically using a ‘cross the best withthe best and hope for the best’ approach but development incontemporary genetics and genomics has driven resurgencein the search for novel sources of variation from germplasmcollections (wild species and landraces). Human interven-tion has also introduced new variation using agents such asirradiation and chemical mutagenic compounds which, inthe past 20 years, has been complemented by transgenesis,

0167-7799/$ – see front matter

� 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tibtech.2013.

03.004

Corresponding author: Casacuberta, J.M. (josep.casacuberta@cragenomica.es)* Current address: European Molecular Biology Laboratory, Meyerhofstrasse 1,

Heidelberg, Germany.Keywords: site-directed nuclease; zinc-finger nucleases (ZFN); transcription activator-like effector nucleases (TALEN); meganuclease; plant breeding; genetically modifiedorganism (GMO).

resulting in the production of genetically modified plants(GMPs, see Glossary).

Modern agriculture is challenged with feeding a globalpopulation (predicted to reach �9 billion by 2025) but withreduced availability of natural resources (productive land,water, and nutrients) and concurrent with negativeimpacts of changing climates (e.g., extreme weather andnew/emerging pests and diseases). Primary productionagriculture and plant breeding will require an expansivetoolkit to meet these challenges because there is no ‘onesize fits all’ solution. Conventional plant breeding can belimited in providing the genes and alleles required to meetthe challenges, whereas conventional mutation breedingrequires time consuming and expensive screens on largepopulations. Moreover, these methods are serendipitous,which limits their use to the production of plants with geneknockouts. With respect to transgenesis, one of the limita-tions is the complexity of the regulatory process for GMPs

the ends are directly ligated without the need for a homologous template.

NHEJ consists of two subpathways: canonical end-joining (C-EJ) and

alternative end-joining (A-EJ). See Box 4 for details on DNA repair mechan-

isms.

Site-directed nucleases (SDNs): nucleases that can be directed to specific DNA

sequences through a DNA binding activity (see Box 1 for different types of

SDNs).

Transgenic: in the context of this paper, transgenic refers to plants containing a

gene or genes that have been inserted using a genetic modification/

engineering technique, and that are referred to as GMPs.

Trends in Biotechnology, June 2013, Vol. 31, No. 6 375

Box 1. Different types of SDNs (MNs, ZFNs, and TALENs)

MNs

MNs are naturally occurring endodeoxyribonucleases encoded by

mobile introns of a wide range of organisms that are characterised

by a large recognition site (DNA sequences of typically 20–30 bp)

[53]. The best studied MNs and the most widely used in research

and genome engineering include I-SceI (discovered in the mito-

chondria of baker’s yeast Saccharomyces cerevisiae), I-CreI (from

the chloroplasts of the green alga Chlamydomonas reinhardtii) and

I-DmoI (from the archaebacterium Desulfurococcus mobilis) [54].

The DNA-binding site, which contains the catalytic domain, is

composed of two components located at either side of the DNA

cleavage site [53] (see Figure 1 in main text). The use of MNs has

been limited by the repertoire available. However, tailor-made MNs

have been developed either by modifying the specificity of existing

natural MNs through the introduction of a small number of

variations in the amino acid sequence of the natural recognition

site [55], or by fusing protein domains from different enzymes [56].

ZFNs and TALENs

In contrast to MNs, ZFNs and TALENs are chimeric enzymes that

combine a nuclease (frequently based on the FokI enzyme) and a

DNA-binding domain derived from ZF proteins or TAL effectors. A

single finger-like structure of a ZF DNA-binding domain recognises a

specific 3-bp sequence [25,57]. As each ZF binds a different

sequence, they can be linked together as independent modules,

resulting in a peptide that can be designed to bind a predetermined

DNA site [25,57]. They therefore usually target sequences of 9 bp.

The enzyme FokI functions as a dimer, therefore, two ZFN units are

needed to cleave the DNA and the complete recognition site for

ZFNs consists of 18 bp.

The TAL effectors from Xanthomonas are trans-kingdom tran-

scription factors. They are injected by the bacteria into plant cells

where they activate the expression of specific plant genes that aid

bacterial colonisation and spread [58,59]. The DNA-binding domain

of TAL effectors contains 33–35 amino acid sequence repeats which

include a repeat-variable di-residue (RVD) at residues 12 and 13,

determining their specificity in DNA binding [59]. This simple

relation, whereby each repeat binds a specific nucleotide, has

facilitated the engineering of specific DNA-binding domains by

selecting a combination of repeat segments containing the appro-

priate RVD [60,61]. The number of repeats and the sequence of the

RVD determine the length and sequence of the target sequence that

will be recognised.

Future SDN developments

Recently, several groups have exploited the CRISPR (clustered

regularly interspaced short palindromic repeat) immune system of

prokaryotes [62] for successful targeting of the nuclease Cas9

(CRISPR associated 9) by synthesised guide RNAs to specific loci in

eukaryotic genomes [63–65]. This system works in principle like the

above described SDNs but the binding part consists of two small

RNAs that form a partially double-stranded RNA. This guide RNA

works also as an artificially fused chimera and can be designed and

synthesised to target a specific sequence of a given genome. The

cleaving part of this new SDN is the codon optimised Cas9 protein,

which forms a complex with the chimeric guide RNA and introduces

a DSB in the DNA at the specific target side.

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

together with detailed, time-consuming and expensivesafety analysis. Thus, the technology is primarily thedomain of big companies developing profitable traits inmajor crops. Regulatory oversight for GMPs has beenpartly justified by the imprecision of the transgenesistechnique (random insertion in the genome and potentialunintended effects). In recent years techniques have beendeveloped for genome modification based on the use ofSDNs for targeted mutagenesis of genes or targeted inser-tion of DNA sequences. The precise manipulation of gen-omes possible with SDN techniques may overcome some ofthe constraints associated with conventional or transgenic-based breeding techniques, although SDN techniques stillhave some limitations. Also, uncertainties remain over thefuture application of SDN approaches given impendingdecisions on their legal status and required regulatoryoversight [2].

In this review we briefly summarise the current knowl-edge of SDN techniques and their use in plants, togetherwith their potential use to introduce new traits in commer-cial crops. Moreover, we compare SDN techniques withrespect to the methods currently used in plant breedingand discuss the potential safety issues related to the use ofthese techniques and the appropriate safety evaluationand regulatory oversight of the products obtained by thesetechniques.

SDNs: modes of action and uses for genomemodificationSDN techniques are based on the introduction of double-strand breaks (DSBs) at specific locations within the ge-nome via tailor-made SDNs. SDNs include meganucleases(MNs), zinc-finger nucleases (ZFNs) and transcription-ac-tivator like (TAL) effector nucleases (TALENs) (Box 1).These SDNs have been developed and tested using bothplants and animal systems and they have been used for thetargeted modification of both plant and animal genomes.

The three different types of nucleases have advantagesand disadvantages [3]. The small size of MNs makes themeasier to deliver and their high specificity results in lessdamage to the cell (i.e., lower frequency of DSBs at off-target sites). However, the fact that their DNA binding andenzymatic activities cannot be easily separated, as well asthe complex array of protein/DNA-specific interactionsthey establish, makes them less flexible and more difficultto engineer [4]. Conversely, the modular nature of theDNA-binding motifs of ZFNs, and in particular TALENs,makes them highly flexible, facilitating the engineering ofDNA-binding modules for virtually any sequence of inter-est. However, a lower specificity has been observed withZFNs compared with MNs and TALENs, leading to signif-icant off-target DSBs and higher levels of cellular damage[5–8].

The outcome of the DSBs introduced by SDNs willdepend on the mechanism used for DNA repair (Box 2)and on the presence of homologous DNA to be used as arepair template (Figure 1).

A large number of different purposes and possible out-comes exist for SDNs. To simplify their descriptions andfacilitate a discussion on possible regulatory oversightof derived products, a classification into three different

376

groups has been proposed [9] (Figure 1). The SDN-1 tech-niques use SDNs to generate site-specific random muta-tions; in most cases via a single DSB that plants repairmainly by nonhomologous end joining (NHEJ) [10,11]. TheNHEJ mechanism can result in unfaithful repair, creatingsmall nucleotide deletions and/or insertions (Indels) (Box2). The SDN-1 techniques can also be designed to introducetwo DSBs, in which case a deletion of the region betweenthe two target sites can be obtained. Thus deletions ofregulatory regions, exons/introns, a whole gene or evenparts of chromosomes are possible by introducing two

Box 2. DSB repair mechanisms

Most of the pathways involved in DSB repair are well studied and

components involved have been identified in yeast and mammalian

systems. After induction of a DSB (e.g., due to bleomycin, irradia-

tion, or SDN), the break can be repaired by different pathways,

mostly depending on the extent of resection [progressive single-

stranded (ss)DNA degradation] of the 30 DNA end (Figure I). The

pathways are as follows. (i) C-EJ, which is dependent on KU70/80

and requires no homology. This process seals the broken ends and

assures genome stability. (ii) A-EJ, which is independent of KU70/80.

This process can operate after short or long DNA resection, resulting

in shorter or longer deletions. (iii) HR, which needs a long ssDNA

resection process and the loading of RAD51 (RADiation sensitive

51 protein) on ssDNA to facilitate an active search for homologous

DNA [66]. The DSB is occupied first by the MRN (Mre11–Rad50–

Nbs1) complex and KU70/80 and can be repaired by C-EJ. However,

the release of both complexes can lead directly to A-EJ, which needs

only short ssDNA resection but can also work later on extensively

resected ends. A-EJ is highly error prone, resulting in genomic

instability. Alternatively, to maintain genome stability, the HR

pathway can be activated via BRCA2. HR needs an extensive ssDNA

resection step in which EXO1 (Exonuclease 1 protein) and a RECQ-

helicase (Recombination Q helicase protein) [e.g., BLM (Bloom

syndrome protein) or SGS1 (Slow Growth Suppressor protein)] are

crucial [67].

DSB

Shortresec�on

C-EJNHEJ

Key:

A-EJ HR

Extensiveresec�on

BRCA2DNA-PKcs

RPAParp1

Ku80 Lig3

MRN CtlP

Ku70 BLM-Exo1

Lig4RAD51

TRENDS in Biotechnology

Figure I. Repair of a double-strand break (DSB). Ligation of nonhomologous end-joining (NHEJ) does not require sequence homology. For canonical end-joining (C-EJ),

the DNA ends are recognised by Ku70/Ku80 heterodimer and DNA-dependent protein kinase catalytic subunit (DNA-PKcs). Activation of the latter allows the recruitment

of ligase 4 for the ligation step. For alternative end-joining (A-EJ), unprotected DNA ends are degraded by the MRN complex revealing complementary

microhomologies that can anneal. Filling in or nicking of the single strand gap completes the end-joining. These last steps are dependent on PARP1 and ligase 3.

Homologous recombination (HR) uses an intact homologous sequence to repair the DSB and is initiated by resection of single-stranded DNA. The MRN complex, in

association with CtIP, initiates resection, which allows the recruitment of other nucleases, for the formation of single-stranded overhangs that are covered by the RPA

protein. BRCA2 will remove RPA and load the RAD51 protein for the invasion and reparation of the injured strand of the homologous sequence. Abbreviations: MRN

complex, Mre11–Rad50–Nbs1 complex; PARP1, Poly (ADP-ribose) polymerase 1; CtIP, CtBP-interacting protein; RPA, Replication protein A; BRCA2, breast cancer type 2

susceptibility protein; RAD51, RADiation sensitive 51 protein.

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

DSBs at different sites using either one or two SDNs [12–14]. The induction of two DSBs can also lead to duplica-tions and inversions of the sequences located between thetwo DSBs. SDN-1 also allows the induction of a transloca-tion event. In plants only the displacement into a targetedlocus of a transgene already present in the plant has beenachieved to date [15].

The SDN-2 techniques are based on the use of anexternally supplied DNA template (usually called donorDNA) for the repair that proceeds by homologous recombi-nation (HR) (Box 2). Donor DNA contains the targeteddesired mutations (a few nucleotide substitutions or shortinsertion/deletions) but also requires high sequence iden-tity to the endogenous gene to allow HR. A successful HRevent substitutes the endogenous sequence by the donorDNA introducing the desired mutations. Thus, modifica-tions in gene sequences can be targeted, for example, torepair undesirable spontaneous mutations (targeted genecorrection) or to introduce targeted mutations and newphenotypes (gene/allele replacement).

Finally, the SDN-3 techniques aim to introduce long DNAfragments (e.g., transgenes) at a predefined locus using adonor DNA (that includes the sequence to be introducedtogether with flanking DNAs showing homology to the targetlocus) in combination with an SDN. This facilitates thetargeted integration of DNA, which otherwise would inte-grate randomly in the genome at naturally occurring DSBs.

Uses of SDNs to develop next generation plants forcommercialisationSDNs have been used in plants both for site-specific muta-genesis (SDN-1 and SDN-2) and for targeted insertion offoreign DNA (SDN-3) [16]. Model plants have usually beenused to develop this technique, but SDNs are increasinglybeing used in crop plants and an important number ofpatents covering these techniques and their uses in plantsalready exist [9,17]. Here, we briefly review the recentadvances of the use of SDN techniques in plants, as wellas their possible use by breeding companies to develop newcommercial plants.

377

Random repair with gainor loss of base pairs

SDN1Gene off

Gene modifica�on atone or more posi�ons

SDN2Gene edit

DNA inser�onSDN3

Add gene

Ligate

DSB

Paste Paste

Donor DNA Donor DNA

HRNHEJ HR

No donorDNA

Chromosome

SDN

+ DonorDNA

+ DonorDNA

Bindingdomain

Key:

Nucleasedomain

TRENDS in Biotechnology

Figure 1. Different site-directed nuclease (SDN) techniques (SDN-1, 2, and 3). An SDN complex is shown at the top in association with the target sequence. The repair can

take place via nonhomologous end-joining (NHEJ) or homologous recombination (HR) using the donor DNA. SDN-1 can result in site-specific random mutations by NHEJ.

In SDN-2, a homologous donor DNA is used to induce specific nucleotide sequence changes by HR. In SDN-3 DNA is integrated in the plant genome via HR.

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

SDN-1 approaches using ZFNs have been used in Ara-bidopsis, tobacco, petunia, and soybean, whereas TALENshave been used in Arabidopsis, tobacco, and rice, and MNsin maize [16,18]. The first application of ZFN enzymes inplants used a reporter sequence newly incorporated intothe plant genome to screen for ZFN-derived mutants [16].Subsequently, several articles have reported site-specificmutations using ZFN constructs stably integrated into theplant genome [19–23]. Although in some cases short dele-tions accounted for 80% of the ZFN-induced mutations[24], in others nucleotide substitutions made up 70% ofall mutations induced [19]. The first ZFN-mediated geneknockouts in an endogenous target were of the tobaccoacetolactate synthase gene (ALS) named SuRA [25].

TALENs have been used to induce targeted mutations inprotoplasts from Arabidopsis and tobacco [26,27], and to-bacco leaves [28]. The use of an engineered MN in plants wasfirst reported in maize with targeted mutagenesis at thenative liguleless locus [29]. A recent study [30] has revealedTALEN-mediated mutation of the binding site for the natu-ral TAL effector AvrXa7 or PthXo3 of Xanthomonas oryzaepv. oryzae upstream of the rice bacterial blight susceptiblegene Os11N3. The obtained mutants did not induce theOs11N3 in response to the TAL effectors and were resistantto the pathogen. The plants were morphologically identicalto the wild-type, suggesting that the mutations did notinterfere with the normal expression of the gene.

SDN-2 techniques have also been used to correct non-functional reporter genes previously introduced into theplant [16] but also to modify native genes. For example,

378

ZFNs [31] and TALENs [32] introduced specific pointmutations in the acetolactate synthase genes (als SuRAand SuRB) to confer resistance to imidazolinone and sul-fonylurea herbicides.

The first evidence that an SDN-induced break increasesthe frequency of targeting was obtained in tobacco using anMN [16]. More recently the use of SDNs (MNs and ZFNs) tofacilitate the targeted integration of transgenes has beenreported in tobacco [18,33], maize [34], and Arabidopsis[20]. In all three cases the transgenes were targeted to anartificial locus previously integrated in the genome.

Targeted gene insertion into a natural chromosomaltarget site has only been reported in maize [35] andtobacco [33]. In the latter, the transgene was targeted tothe stress-related CHN50 endochitinase gene, which ishighly expressed in the stationary phase of suspensionculture cells, and could therefore be particularly suitablefor recombinant protein production.

In addition to the examples reported above, the SDNtechnique is being used by plant breeding companies tointroduce new traits in crops. Companies are using SDNs(in particular MNs and ZFNs) in maize, oilseed rape, andtomato from the research phase through to phase 3 ofcommercial development [9]. Although the companies donot disclose the traits involved, the techniques are clearlyconsidered advantageous for accelerated breeding of newphenotypes [9].

There are many traits that could be improved using SDNs(Box 3). For targeted mutagenesis (SDN-1 and SDN-2) thesecould include: herbicide tolerance (via modifications to the

Box 3. Potential uses of SDNs

Potential uses are inumerable. DNA shuffling can generate

sequence variation producing proteins with desirable properties

(kinetics, substrate specificity, temperature or pH optima, and ligand

binding) including improved disease resistance (e.g., R gene

shuffling [68]). SDNs can target the most effective changes required.

Furthermore, different techniques such as TILLING and EcoTILLING

and next-generation sequencing (NGS)/whole genome sequence

analysis are identifying candidate genes and sequences polymorph-

isms that can be used in SDN applications for targeted phenotype

development. Examples [69] include:

(i) virus resistance (targeting translation initiation factors [36]);

(ii) herbicide tolerance (target – acetolactate synthase gene [70]);

(iii) lowering antinutritional compounds (erucic acid in Brassicas –

targeting fatty acid elongases) and allergens (target – con-

glutin genes in peanut) [71,72];

(iv) improved nutritional value via elevated carotenoids, modified

carotenoid balance (target – zeaxanthin epoxidase) [73];

(v) modified starches and fats for food and non-food uses

(targeting starch synthases, branching enzymes, and fatty acid

desaturases) [74,75];

(vi) longer shelf life/reduced wastage [targets include aminocyclo-

propane (ACC) oxidase and polygalacturoanse] [76];

(vii) improved quality by reducing enzymic (target polyphenol

oxidases) and nonenzymic browning (high quality and low

acrylamide potato – target invertase genes) [77];

(viii) yield benefits via modified RuBisCO genes, increasing catalytic

activity and/or decreasing oxygenation activity and improved

seed set (e.g., producing six row barley by targeting home-

odomain leucine zipper genes) [78,79];

(ix) improved biomass conversion for biofuels (lower lignin-target

caffeic acid O-methyltransferase gene) [80].

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

endogenous EPSPS, ALS, AHAS, or IPK genes), changedcomposition (e.g., reduced levels of allergens or phytic acids,to improve the nutritional value of animal feed, modifiedstarch for industrial use or to lower glycaemic index), orresistance to biotic or abiotic stresses. With respect to thelatter, targeted mutagenesis is envisaged for virus resis-tance via modification of the eiF4E gene in tomato [36],nematode resistance via modification of the Gro1-4 gene inpotato [37], and abiotic cold stress tolerance via modificationof the DREB1 genes in rice [38,39]. This should encourage anincreased focus on SDN approaches. The targeted insertionof transgenes (SDN-3) facilitates more predictable expres-sion (targeting to high expression regions or decreasing therisk of gene silencing) while minimising impacts on thegenome (avoiding insertion into genes or regulatoryregions). Moreover, SDN-3 may be used for stacking trans-genes at the same locus, allowing their mobilisation intoother germplasm by crossing as a single locus [18].

Potential safety issues associated with SDN-basedapproaches compared with other plant breedingplatformsSDNs can be used to introduce random or predefinedmutations at specific loci (SDN-1 or SDN-2) or to targettransgenes to a predefined locus (SDN-3). The SDN-basedtechniques are thus related to physical or chemical muta-genesis techniques used in conventional breeding (SDN-1and SDN-2) and to transgenesis (SDN-3). Therefore, whenanalysing the safety of SDN-derived products, a compari-son with risks associated with other techniques currentlyused to produce commercial crops (e.g., conventional mu-tagenesis and transgenesis) is highly relevant.

Mutations are key drivers of genome variability andevolution and therefore important in plant breeding. Spon-taneous mutations accumulate naturally due to, for exam-ple, replication errors, movement of transposons and DNArepair following exposure to UV light (Box 4). Plant bree-ders also generate new variability using mutagenesistechniques that are based on the fact that the repair ofmost of the DNA damage induced chemically (e.g., via ethylmethyl sulfonate; EMS) or physically (e.g., via irradiation)also relies on the cellular repair mechanisms, including theNHEJ pathway, that can be error prone [40] (Box 2). Thisoften results in mutations at the repair locus. Mutagenesishas been widely used by breeders since the second half ofthe last century and in 2012 the FAO/IAEA mutant varietydatabase listed 3200 officially released cultivars in over200 plant species (http://mvgs.iaea.org). After randomlymutagenising the plant genomes, forward genetic screenscan select mutants with improved traits for inclusion inbreeding programmes. As the induced mutagenesis israndom, many unwanted mutations occur. However (andwhen not limited by the reproductive biology of the spe-cies), these mutations can be segregated away during theselection and breeding processes.

The DNA lesions induced by SDN techniques will berepaired by one of the DNA repair pathways of the cell (Box2), as would be the case in those exploited by conventionalbreeding (spontaneous or induced mutations). Therefore,the resulting mutations will be indistinguishable. Al-though designed to cleave DNA only at predefined posi-tions, SDNs may also cleave DNA at unintended locationswith a certain probability. Off-target activity of the SDNdepends on the specificity of the SDN and the frequency ofsequences similar to the SDN recognition site in the ge-nome. The off-target cleavage is the main reason for cellu-lar toxicity observed with ZFNs [6–8]. Different strategiesare followed to decrease off-target mutations including theuse of obligate heterodimers [41,42], cleavage domains ofother restriction enzymes [43], or DNA nickases fused tothe specific DNA-binding domains of SDNs instead of thecurrently used restriction enzymes [44,45]. In any case, thefrequency of off-target mutations caused by SDNs would bewell below the frequency of unwanted mutations resultingfrom chemical or physical mutagenesis agents, for exam-ple, one mutation per 150 kbp can be effected by EMStreatment [46]. This reduces the possibility of unintendedeffects potentially having an impact on the safety of thefinal product. Furthermore, as in conventional breeding,unintended mutations can be segregated away during theselection and breeding process.

The predictability of the outcome of SDN techniques iseven higher in the case of SDN-2, because the mutations tobe introduced in the target locus are predefined by thesequence of the DNA used as a repair donor.

In contrast with the mutagenic techniques historicallyused in plant breeding, SDN-based mutagenesis usuallyrequires the introduction of the genes encoding the SDNsby transgenesis. However, the transgene can be segregatedaway during the selection and breeding processes becauseits presence is not required for the maintenance of theimproved trait. Moreover, SDNs can also be deliveredtransiently, avoiding the integration of any transgene

379

Box 4. DNA damage

Cells are constantly exposed to environmental agents and endogen-

ous processes that inflict damage to DNA (Figure I). Six main

pathways coexist to deal with DNA damage [81]. Three excision

repair pathways exist to cope with DNA single strand breaks: (i) base

excision repair (BER) will repair single-strand breaks, base modifica-

tions, and abasic sites; (ii) nucleotide excision repair (NER) will repair

bulky DNA adducts (pyrimidine dimer); and (iii) DNA mismatch repair

(MMR) will repair mismatch and base modifications. DNA DSBs are

handled by three DNA repair pathways: HR, C-EJ, and A-EJ, also

known as microhomology mediated end joining (MMEJ) [82]. C-EJ

and A-EJ can be grouped under the denomination of NHEJ.

Conventional breeding performed by mutagenesis based on EMS

or ionising irradiation will produce multiple base modifications or

DSBs that can lead to major chromosomal aberrations, chromosome

losses, and large deletions. By contrast, SDNs will produce mostly

targeted DSBs.

Mismatch and basemodifica�ons,

inser�ons, dele�ons,DNA breaks

Endogenous damage:

• Replica�onerrors

• Transposableelements

• Reac�ve oxygenspecies

• Enzymes(SDN)

• Chemicalagents

• Ionizing and UVradia�on

Exogenous damage:

DNA

dam

age

Dam

age

orig

in

Inser�ons,dele�ons,

DNA breaks

Base modifica�ons,abasic sites, DNA

breaks

DNA doublestrand breaks

Base modifica�ons(e.g., EMS), DNAcrosslinks (e.g.,

cispla�n), DNA breaks(e.g., bleomycin)

Pyrimidine dimer (e.g., UV),base modifica�ons, abasic

sites, DNA breaks,transloca�ons, large dele�ons(e.g.,fast neutron, γ- or X-rays)

TRENDS in Biotechnology

Figure I. Sources of DNA damage. DNA, our genetic fingerprint, is constantly subjected to mechanisms that can cause damage. These may be driven endogenously

(e.g., via errors in DNA replication, transposable element movement, or oxidative stress resulting from metabolic processes) or exogenously via factors such as

radiation or chemical agents that result in mutations or damage to nucleotide bases, DNA crosslinking, single-strand breaks (SSBs) and double-strand breaks (DSBs).

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

[16], or directly as proteins [47]. An interesting approach isthe use of virus vectors to transiently deliver the SDNs, asdemonstrated with petunia using the tobacco rattle virusas a vector to deliver a ZFN that allowed ZFN-mediatedmutations to be recovered in the progeny without requiringa transgenic step [48].

In addition to introducing mutations at specific loci,SDNs can be used to introduce long stretches of DNA (e.g.,transgenes) into the genome (the SDN-3 technique). Inthese cases it is pertinent to compare the process andoutcomes with transgenesis, because the aim and the finalproducts will be similar. However, although in transgen-esis the DNA is integrated into natural DNA breaks thatoccur at random positions in the genome (often close toactively transcribed genes [49]), SDN-3 techniques targetDNA insertion to a predefined site in the genome. Target-ing of the transgene in a predefined locus can optimise thegenomic environment for gene expression and minimisehazards associated with the disruption of genes and/orregulatory elements in the recipient genome. Moreover,because the junctions between the transgene and therecipient genome are predefined by the sequence of thedonor used, the SDN-3 approach should avoid the creationof new open reading frames with similarity to toxins orallergens. Therefore, the use of the SDN-3 techniquecould be seen as a way to decrease substantially theseperceived risks.

Regulatory considerationsAs already stated, SDN techniques can have multipleadvantages for plant breeding. SDN-1 and SDN-2approaches can create new gene variants with an unprec-edented specificity and predictability, and SDN-3

380

approaches can make the production of GMPs more pre-dictable and could be of great interest for the developmentof stacked events. Academic laboratories, consortia, andspecialised biotechnology service providers are improvingthe design of SDNs and the methods used to deliver theminto the cell to make them more efficient. Plant breedingcompanies have started to use these techniques to developa new generation of crops with new traits. However, thereare species-specific limitations. The design of the SDNtargets requires knowledge of the genome sequence, whichis still lacking for some important crops (and certainly for‘orphan’ crops). Also, the delivery of the SDNs oftenrequires efficient transformation and regeneration proto-cols that are not available for all plants [18,50]. Important-ly, the future of these techniques is in the balance asuncertainty remains over the legal status of plants gener-ated by SDN-based approaches and the regulatory over-sight required. For GMPs, it can currently take almost 6years and $35 million to generate the data for a regulatorydossier [9], limiting the use of this technology to high-valuecrops and traits developed by the major agrobiotechnologycompanies. Thus, finding the appropriate level of regula-tory oversight to ensure the safe use of these new techni-ques while keeping development costs at a minimum is ofgreat importance.

In the past few years several opinion articles have beenpublished discussing the possible legal status of SDN tech-niques [2,46,47]. It appears that there is growing agreementthat for SDN-1 techniques that introduce SDNs as proteinsor RNAs (no transgene is involved) the derived productsshould be considered for exclusion from genetically modifiedorganism (GMO) legislation. In the EU, this is also the viewof several competent authorities of individual member

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

states, such as the German ZKBS (http://www.bvl.bund.de/SharedDocs/Downloads/06_Gentechnik/ZKBS/02_Allgemeine_Stellungnahmen_englisch/01_general_subjects/zkbs_general_ZFN_1.pdf?__blob=publicationFile&v=3) and COGEM from The Netherlands (http://www.co-gem.net/index.cfm/en/publications/publicatie/zinc-finger-on-the-pulse-developments-and-implications-of-zinc-finger-technology). Similar opinions were voiced by mostparticipants at a European Joint Research Centre meet-ing in 2011, which included representation fromArgentina, Australia, Canada, Japan, the EU, and SouthAfrica [2]. In the USA, the Department of Agriculturehas informed crop companies that plants obtained usingSDN-1 techniques, together with those in which anytransgene coding for the SDN has been removed by segre-gation, will not be regulated (http://www.aphis.usda.gov/biotechnology/reg_loi.shtml) [51].

With respect to plants obtained by SDN-2 and SDN-3 theUS Department of Agriculture has intimated a case-by-caseapproach [51]; an approach preferred by some countries(e.g., Argentina) for SDN-2 [9]. In the EU, some competentauthorities (e.g., the German ZKBS), as well as a majority ofthe members of the New Techniques Working Group con-sidered that SDN-2-derived products should also be consid-ered for exclusion from the EU Directive that regulatesGMOs when the changes introduced are small and whereno stably introduced transgene is involved in the process(http://www.bvl.bund.de/SharedDocs/Downloads/06_Gen-technik/ZKBS/02_Allgemeine_Stellungnahmen_englisch/05_plants/zkbs_plants_new_plant_breeding_techniques.pdf?__blob=publicationFile&v=2). For SDN-3, most of thecountries seem to consider plants derived using this tech-nique to be GMOs. In Europe, the European Food SafetyAuthority (EFSA) has issued an opinion on SDN-3, indicat-ing that although plants obtained by SDN-3 techniques aresimilar to GMOs with respect to the introduced trait, thetargeted nature of SDN-3 may minimise the hazards linkedto the insertion of the transgene into the genome [52].

If products derived from SDN techniques are indeeddesignated as GMPs, a specific detection method is legallyrequired to identify the products unambiguously. For mostSDN-3 products this should be relatively easy, because thenewly introduced DNA is relatively long, but may beextremely challenging for some SDN-1 and SDN-2 pro-ducts that may differ in only a few nucleotides. Even whena method could be developed for detection of introducedmutations, it will be impossible to discriminate them fromthe same mutations obtained by conventional mutagenesisapproaches [2,9].

Concluding remarks and outlookSDN techniques are already proven approaches for site-specific mutagenesis and targeted DNA insertion inplants. The specificity they offer holds real promise forthe development, with predictable outcomes, of a newgeneration of improved crop varieties with a minimisedpotential for unintended effects that could have an impacton safety. However, some technical limitations as well asregulatory uncertainty may challenge SDN use for com-mercial plant breeding. The main technical limitations arerelated to both the genomic information needed for the

design of the SDNs, which is not available for all crops, andthe existence of an efficient delivery method, which is alsolimited to specific genotypes, even within the same spe-cies. However, information on plant genomes is growingvery rapidly and there is active research in new deliverymethods for SDNs. Indeed the field is developing at a fastpace and new methods developed in the animal field, suchas the CRISPR approach, might in the future also be usedto improve crop plants. With respect to the regulatoryclassification of the products obtained by SDN techniques,consensus among the different regulatory bodies and sta-keholders has not been reached yet, in particular, in theEU. For SDN-3 approaches most regulatory bodies seem toview them as improved GMOs, whereas products devel-oped using SDN-1 may, in the future, be considered as non-GMOs for legislative purposes and particularly when notransgene is involved. However, absolute clarity and con-sensus is proving elusive at present with SDN-2 a partic-ular case in point, because it may fall somewhere betweenthe two extremes of SDN-1 and SDN-3. There is an urgentneed for a discussion, based on scientific data as much aspossible, on the regulatory framework for these techni-ques that may clarify their future and allow the invest-ment the development of these promising techniquesneeds.

AcknowledgementsThe authors want to thank the members of the EFSA New TechniquesWorking Group for seminal discussions at the origin of this manuscript,and Elisabeth Waigmann for critical reading of the manuscript.

Disclaimer statementJosep Casacuberta, Howard Davies, and Fabien Nogue are currently adhoc members of the Molecular Characterisation Working Group of EFSAGMO Risk Assessment Panel. Nancy Podevin was, until recently amember of the unit supporting the EFSA GMO Panel. All of the authorswere members or supported a working group set up by EFSA to developthe EFSA GMO Panels final opinion on SDN–3 following a mandate fromthe European Commission. The opinion is published: [The EFSA GMOPanel. (2012) Scientific opinion addressing the safety assessment ofplants developed using zinc finger nuclease 3 and other site-directednucleases with similar function. EFSA J. 10, 2943 (http://dx.doi.org/10.2903/j.efsa.2012.2943)]The authors do not believe these involvements conflict with the paper tobe published in Trends in Biotechnology. The opinions in the manuscriptare those of the authors and not of the EFSA.

References1 Stamp, P. and Visser, R. (2012) The twenty-first century, the century of

plant breeding. Euphytica 186, 585–5912 Lusser, M. and Rodrıguez-Cerezo, E. (2012) Comparative regulatory

approaches for new plant breeding techniques. Workshop proceedings.European Commission. JRC Technical Report EUR 25237 EN

3 DeFrancesco, L. (2011) Move over ZFNs. Nat. Biotechnol. 29, 681–6844 Taylor, G.K. et al. (2012) LAHEDES: the LAGLIDADG homing

endonuclease database and engineering server. Nucleic Acids Res.40, W110–W116

5 Mussolino, C. et al. (2011) A novel TALE nuclease scaffold enables highgenome editing activity in combination with low toxicity. Nucleic AcidsRes. 39, 9283–9293

6 Cornu, T.I. et al. (2008) DNA-binding specificity is a major determinant ofthe activity and toxicity of zinc-finger nucleases. Mol. Ther. 16, 352–358

7 Pattanayak, V. et al. (2011) Revealing off-target cleavage specificities ofzinc-finger nucleases by in vitro selection. Nat. Methods 8, 765–770

8 Gabriel, R. et al. (2011) Genome-wide determination of double-strandbreaks reveals high specificity of zinc finger nucleases. Hum. GeneTher. 21, 1371

381

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

9 Lusser, M. et al. (2012) Deployment of new biotechnologies in plantbreeding. Nat. Biotechnol. 30, 231–239

10 Roth, N. et al. (2012) The requirement for recombination factors differsconsiderably between different pathways of homologous double-strandbreak repair in somatic plant cells. Plant J. 72, 781–790

11 Waterworth, W.M. et al. (2011) Repairing breaks in the plant genome:the importance of keeping it together. New Phytol. 192, 805–822

12 Petolino, J.F. et al. (2010) Zinc finger nuclease-mediated transgenedeletion. Plant Mol. Biol. 73, 617–628

13 Sollu, C. et al. (2010) Autonomous zinc-finger nuclease pairs fortargeted chromosomal deletion. Nucleic Acids Res. 38, 8269–8276

14 Lee, H.J. et al. (2012) Targeted chromosomal duplications andinversions in the human genome using zinc finger nucleases.Genome Res. 22, 539–548

15 Fauser, F. et al. (2012) In planta gene targeting. Proc. Natl. Acad. Sci.U.S.A. 109, 7535–7540

16 Tzfira, T. et al. (2012) Genome modifications in plant cells by custom-made restriction enzymes. Plant Biotechnol. J. 10, 373–389

17 Parisi, C. et al. (2013) Analysing patent landscapes in plantbiotechnology and new plant breeding techniques. Transgenic Res. 22,15–29

18 Curtin, S.J. et al. (2012) Genome engineering of crops with designernucleases. Plant Genome 5, 42–50

19 Osakabe, K. et al. (2010) Site-directed mutagenesis in Arabidopsisusing custom-designed zinc finger nucleases. Proc. Natl. Acad. Sci.U.S.A. 107, 12034–12039

20 de Pater, S. et al. (2009) UN-induced mutagenesis and gene-targetingin Arabidopsis through Agrobacterium-mediated floral diptransformation. Plant Biotechnol. J. 7, 821–835

21 Even-Faitelson, L. et al. (2011) Localized egg-cell expression of effectorproteins for targeted modification of the Arabidopsis genome. Plant J.68, 929–937

22 Tovkach, A. et al. (2009) A toolbox and procedural notes forcharacterizing novel zinc finger nucleases for genome editing inplant cells. Plant J. 57, 747–757

23 Zhang, F. et al. (2010) High frequency targeted mutagenesis inArabidopsis thaliana using zinc finger nucleases. Proc. Natl. Acad.Sci. U.S.A. 107, 12028–12033

24 Lloyd, A. et al. (2005) Targeted mutagenesis using zinc-fingernucleases in Arabidopsis. Proc. Natl. Acad. Sci. U.S.A. 102, 2232–2237

25 Maeder, M.L. et al. (2008) Rapid ‘‘Open-Source’’ engineering ofcustomized zinc-finger nucleases for highly efficient gene modification.Mol. Cell 31, 294–301

26 Cermak, T. et al. (2011) Efficient design and assembly of customTALEN and other TAL effector-based constructs for DNA targeting.Nucleic Acids Res. 39, e82

27 Zhang, Y. et al. (2013) TALENs enable efficient plant genomeengineering. Plant Physiol. 161, 20–27

28 Mahfouz, M.M. et al. (2011) De novo-engineered transcriptionactivator-like effector (TALE) hybrid nuclease with novel DNAbinding specificity creates double-strand breaks. Proc. Natl. Acad.Sci. U.S.A. 108, 2623–2628

29 Gao, H.R. et al. (2010) Heritable targeted mutagenesis in maize using adesigned endonuclease. Plant J. 61, 176–187

30 Li, T. et al. (2012) High-efficiency TALEN-based gene editing producesdisease-resistant rice. Nat. Biotechnol. 30, 390–392

31 Townsend, J.A. et al. (2009) High-frequency modification of plant genesusing engineered zinc-finger nucleases. Nature 459, 442–445

32 Zhang, Y. et al. (2012) TALENs enable efficient plant genomeengineering. Plant Physiol. 161, 20–27

33 Cai, C.Q. et al. (2009) Targeted transgene integration in plant cellsusing designed zinc finger nucleases. Plant Mol. Biol. 69, 699–709

34 D’Halluin, K. et al. (2008) Homologous recombination: a basis fortargeted genome optimization in crop species such as maize. PlantBiotechnol. J. 6, 93–102

35 Shukla, V.K. et al. (2009) Precise genome modification in the cropspecies Zea mays using zinc-finger nucleases. Nature 459, 437–441

36 Piron, F. et al. (2010) An induced mutation in tomato eIF4E leads toimmunity to two potyviruses. PLoS ONE 5, e11313

37 Milczarek, D. et al. (2011) Suitability of molecular markers for selectionof potatoes resistant to Globodera spp. Am. J. Potato Res. 88, 245–255

38 Tondelli, A. et al. (2011) Inside the CBF locus in Poaceae. Plant Sci.180, 39–45

382

39 Mao, D. and Chen, C. (2012) Colinearity and similar expression patternof rice DREB1s reveal their functional conservation in the cold-responsive pathway. PLoS ONE 7, e47275

40 Tuteja, N. et al. (2009) Genotoxic stress in plants: shedding light on DNAdamage, repair and DNA repair helicases. Mutat. Res. 681, 134–149

41 Doyon, Y. et al. (2011) Enhancing zinc-finger-nuclease activity withimproved obligate heterodimeric architectures. Nat. Methods 8, 74–79

42 Miller, J.C. et al. (2007) An improved zinc-finger nuclease architecturefor highly specific genome editing. Nat. Biotechnol. 25, 778–785

43 Fonfara, I. et al. (2012) Creating highly specific nucleases by fusion ofactive restriction endonucleases and catalytically inactive homingendonucleases. Nucleic Acids Res. 40, 847–860

44 Ramirez, C.L. et al. (2012) Engineered zinc finger nickases inducehomology-directed repair with reduced mutagenic effects. NucleicAcids Res. 40, 5560–5568

45 Wang, J. et al. (2012) Targeted gene addition to a predetermined site inthe human genome using a ZFN-based nicking enzyme. Genome Res.22, 1316–1326

46 Sasaki, T. et al. (2012) Unusual case of apparent hypermutation inArabidopsis thaliana. Genetics 192, 1271–1280

47 Gaj, T. et al. (2012) Targeted gene knockout by direct delivery of zinc-finger nuclease proteins. Nat. Methods 9, 805–807

48 Vainstein, A. et al. (2011) Permanent genome modifications in plantcells by transient viral vectors. Trends Biotechnol. 29, 363–369

49 Fu, F.F. et al. (2009) Studies on rice seed quality through analysis of alarge-scale T-DNA insertion population. Cell Res. 19, 380–391

50 Barampuram, S. and Zhang, Z. (2011) Recent advances in planttransformation. In Plant Chromosome Engineering (Birchler, J.,ed.), pp. 1–35, Humana Press

51 Waltz, E. (2012) Tiptoeing around transgenics. Nat. Biotechnol. 30,215–217

52 European Food Safety Authority (2012) Scientific opinion addressingthe safety assessment of plants developed using zinc finger nuclease3 and other site-directed nucleases with similar function. EFSA J.10, 2943

53 Stoddard, B.L. (2011) Homing endonucleases: from microbial geneticinvaders to reagents for targeted DNA modification. Structure 19, 7–15

54 Hafez, M. and Hausner, G. (2012) Homing endonucleases: DNAscissors on a mission. Genome 55, 553–569

55 Arnould, S. et al. (2011) The I-CreI meganuclease and its engineeredderivatives: applications from cell modification to gene therapy.Protein Eng. Des. Sel. 24, 27–31

56 Grizot, S. et al. (2010) Generation of redesigned homing endonucleasescomprising DNA-binding domains derived from two different scaffolds.Nucleic Acids Res. 38, 2006–2018

57 Carroll, D. (2011) Genome engineering with zinc-finger nucleases.Genetics 188, 773–782

58 Bogdanove, A.J. et al. (2010) TAL effectors: finding plant genes fordisease and defense. Curr. Opin. Plant Biol. 13, 394–401

59 Boch, J. and Bonas, U. (2010) Xanthomonas AvrBs3 family-type IIIeffectors: discovery and function. Annu. Rev. Phytopathol. 48, 419–436

60 Boch, J. et al. (2009) Breaking the code of DNA binding specificity ofTAL-type III effectors. Science 326, 1509–1512

61 Moscou, M.J. and Bogdanove, A.J. (2009) A simple cipher governs DNArecognition by TAL effectors. Science 326, 1501

62 Jinek, M. et al. (2012) A programmable dual-RNA-guided DNAendonuclease in adaptive bacterial immunity. Science 337, 816–821

63 Cong, L. et al. (2013) Multiplex genome engineering using CRISPR/Cassystems. Science 339, 819–823

64 Jinek, M. et al. (2013) RNA-programmed genome editing in humancells. Elife 2, e00471

65 Mali, P. et al. (2013) RNA-guided human genome engineering via Cas9.Science 339, 823–826

66 Grabarz, A. et al. (2012) Initiation of DNA double strand break repair:signaling and single-stranded resection dictate the choice betweenhomologous recombination, non-homologous end-joining andalternative end-joining. Am. J. Cancer Res. 2, 249–268

67 Nimonkar, A.V. et al. (2011) BLM-DNA2-RPA-MRN and EXO1-BLM-RPA-MRN constitute two DNA end resection machineries for humanDNA break repair. Genes Dev. 25, 350–362

68 Wulff, B.B. et al. (2001) Domain swapping and gene shuffling identifysequences required for induction of an Avr-dependent hypersensitiveresponse by the tomato Cf-4 and Cf-9 proteins. Plant Cell 13, 255–272

Review Trends in Biotechnology June 2013, Vol. 31, No. 6

69 Kurowska, M. et al. (2011) TILLING: a shortcut in functional genomics.J. Appl. Genet. 52, 371–390

70 Shimizu, M. et al. (2011) Application of mutated acetolactate synthasegenes to herbicide resistance and plant improvement. In Herbicides,Theory and Applications (Larramendy, M., ed.), InTech

71 Tian, B.M. et al. (2011) Decreasing erucic acid level by RNAi-mediatedsilencing of fatty acid elongase 1 (BnFAE1.1) in rapeseeds (Brassicanapus L.). Afr. J. Biotechnol. 10, 13194–13201

72 Foley, R.C. et al. (2011) Identification and characterisation of seedstorage protein transcripts from Lupinus angustifolius. BMC PlantBiol. 11, 59

73 Yang, Q.J. et al. (2012) Functional characterization of the Gentianalutea zeaxanthin epoxidase (GlZEP) promoter in transgenic tomatoplants. Transgenic Res. 21, 1043–1056

74 Crofts, N. et al. (2012) Lack of starch synthase IIIa and high expressionof granule-bound starch synthase I synergistically increase theapparent amylose content in rice endosperm. Plant Sci. 193, 62–69

75 Chi, X.Y. et al. (2011) Genome-wide analysis of fatty acid desaturases insoybean (Glycine max). Plant Mol. Biol. Rep. 29, 769–783

76 Yun, S.S. et al. (2011) Isolation of fruit ripening genes from Caricapapaya var. Eksotika I cDNA libraries. J. Trop. Agric. Food Sci. 39,203–211

77 Toevs, E. et al. (2011) An industry perspective of all-native andtransgenic potatoes. AgBioForum 14, 14–19

78 Whitney, S.M. et al. (2011) Advancing our understanding and capacityto engineer nature’s CO2-sequestering enzyme, Rubisco. Plant Physiol.155, 27–35

79 Sakuma, S. et al. (2010) Duplication of a well-conservedhomeodomain-leucine zipper transcription factor gene in barleygenerates a copy with more specific functions. Funct. Integr.Genomics 10, 123–133

80 Fu, C.X. et al. (2011) Genetic manipulation of lignin reducesrecalcitrance and improves ethanol production from switchgrass.Proc. Natl. Acad. Sci. U.S.A. 108, 3803–3808

81 Ciccia, A. and Elledge, S.J. (2010) The DNA damage response: makingit safe to play with knives. Mol. Cell 40, 179–204

82 Symington, L.S. and Gautier, J. (2011) Double-strand break endresection and repair pathway choice. Annu. Rev. Genet. 45, 247–271

383