origins of programmable nucleases for genome engineering · 2017-02-07 · origins of programmable...

Perspec�ve

Srinivasan Cha

0022-2836/© 2015 The(http://creativecommons.o

Origins of Programmable Nucleases forGenome Engineering

ndrasegaran1 and Dana C
arroll 2
1 - Department of Environmental Health Sciences, Johns Hopkins School of Public Health, 615 North Wolfe Street, Baltimore,MD 21205, USA2 - Department of Biochemistry, University of Utah School of Medicine, 15 North Medical Drive East, Salt Lake City, UT 84112, USA

Correspondence to Srinivasan Chandrasegaran and Dana Carroll: [email protected]; [email protected]://dx.doi.org/10.1016/j.jmb.2015.10.014Edited by T. K. Lu

Abstract

Genome engineering with programmable nucleases depends on cellular responses to a targeted double-strand break (DSB). The first truly targetable reagents were the zinc finger nucleases (ZFNs) showing thatarbitrary DNA sequences could be addressed for cleavage by protein engineering, ushering in the break-through in genomemanipulation. ZFNs resulted from basic research on zinc finger proteins and the FokI restrictionenzyme (which revealed a bipartite structure with a separable DNA-binding domain and a non-specific cleavagedomain). Studies on the mechanism of cleavage by 3-finger ZFNs established that the preferred substrates werepaired binding sites, which doubled the size of the target sequence recognition from 9 to 18 bp, long enough tospecify a unique genomic locus in plant andmammalian cells. Soon afterwards, a ZFN-inducedDSBwas shown tostimulate homologous recombination in cells. Transcription activator-like effector nucleases (TALENs) that arebased on bacterial TALEs fused to the FokI cleavage domain expanded this capability. The fact that ZFNs andTALENs have been used for genomemodification of more than 40 different organisms and cell types attests to thesuccess of protein engineering. The most recent technology platform for delivering a targeted DSB to cellulargenomes is that of the RNA-guided nucleases, which are based on the naturally occurring Type II prokaryoticCRISPR-Cas9 system. Unlike ZFNs and TALENs that use protein motifs for DNA sequence recognition,CRISPR-Cas9 depends on RNA–DNA recognition. The advantages of the CRISPR-Cas9 system—the ease ofRNA design for new targets and the dependence on a single, constant Cas9 protein—have led to its wide adoptionby research laboratories around the world. These technology platforms have equipped scientists with anunprecedented ability to modify cells and organisms almost at will, with wide-ranging implications across biologyand medicine. However, these nucleases have also been shown to cut at off-target sites with mutagenicconsequences. Therefore, issues such as efficacy, specificity and delivery are likely to drive selection of reagentsfor particular purposes. Human therapeutic applications of these technologies will ultimately depend on risk versusbenefit analysis and informed consent.

© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction

With the advent of high-throughput DNA sequenc-ing and powerful computer algorithms, the genomesequences of many organisms, including human,have been deciphered. However, understanding thefunctions of the genes requires sequence editingeither by deleting or by modifying them individuallyand then studying the resulting mutant phenotypes.

Authors. Published by Elsevier Ltd. Thisrg/licenses/by-nc-nd/4.0/).

The programmable nucleases take advantage ofnatural cellular pathways of DNA repair for the intro-duction targeted sequence changes.DNA double-strand breaks (DSBs) are potentially

lethal to cells, which have two broad classes ofmechanisms to repair them: non-homologous endjoining (NHEJ) and homology-directed repair (HDR).NHEJ often results in variable lengths of insertion anddeletion mutations (indels) and thus can be used to

is an open access article under the CC BY-NC-ND licenseJ Mol Biol (2016) 428, 963–989

http://creativecommons.org/licenses/by-nc-nd/4.0/

http://creativecommons.org/licenses/by-nc-nd/4.0/

http://dx.doi.org/

Fig. 1. Genome engineeringsing programmable nucleases.FNs, TALENs and CRISPR-Cas9re used to induce a targeted DSBt the desired chromosomal locus.ither NHEJ or HDR, one of the twoellular repair pathways, is thensed to repair the DSB. NHEJould be used to knockout genes,hile HDR could be used either forene correction or to introducerecise alterations into the genomeirected by an investigator-providedomologous DNA template. Figuredapted from Ramalingam et al.,

2013.

964 Review: Programmable Nucleases for Genome Engineering

knockout genes (Fig. 1). Nuclease-induced breaks areefficiently mutated likely because perfectly re-joinedsequences can andwill be re-cleaved until they acquirean indel, at which point they can no longer be cleaved.HDR normally relies on recombination with homolo-gous sequences in anundamagedchromatid, but it canbe diverted to use a homologous donor DNA templateprovided by the experimenter. This leads to theintroduction of precise alterations to the genome,which are specified by the template (Fig. 1). Genetargeting by HDR is not an efficient process in cells ofhigher eukaryotes—typically only one in approximatelya million treated cells undergoes the desired genomemodification [1,2].Experiments using rare-cutting meganucleases

(e.g., I-SceI) [3] showed that stimulation of bothlocal mutagenesis and incorporation of homologousdonor sequences can be achieved by inducing anintentional DSB. Restriction endonucleases, the mo-lecular scissors that revolutionized molecular biologyand the biotechnology industry in the seventies byenabling cloning, are not useful for delivering a targetedchromosomal DSB, since they recognize short DNAsites (usually 4- to 8-bp sites), which occur much toofrequently within most genomes. The recognitionspecificity of meganucleases proved too difficult totailor to desired target sites. Thus, generation of atargeted DSB remained the rate-limiting step in thedevelopment of HDR technology for genome engineer-ing of plant and mammalian cells, including humancells. Therefore, techniques had to be developed for ageneralmeansof delivering a targeted genomicDSBata unique chromosomal locus in order to stimulate HDRat that site with exogenously added donor DNA.This article gives our personal perspectives on the

origins of programmable nucleases [zinc fingernucleases (ZFNs), transcription activator-like effectornucleases (TALENs) and RNA-guided CRISPR-Cas9] for genome engineering, highlights key mile-stone publications of the field (Fig. 2) and discusses

uZaaEcucwgpdha

the biological and medical applications of thesetechnologies.

Reflections on the Origins of ProgrammableNucleases and Their Applications to HumanGenomeEngineering—SrinivasanChandrasegaran

In 1986, as a newly hired assistant professor withan independent laboratory in the Department ofEnvironmental Health Sciences at the Johns Hop-kins School of Public Health, I initiated the researchon chimeric restriction enzymes with the aim togenerate designer site-specific endonucleases thatcould target specific genes within cells. Type IIrestriction enzymes are not useful for this task, sincethey typically recognize 4- to 8-bp palindromic DNAsites and on average cut DNA once every 4096–65,536 bases [4]. For genome engineering, weneeded enzymes that recognize sequences of 16–18 bp in length to make a unique targeted DSBwithin the human genome of 3 × 109 bp. Therefore,a long-term goal of the restriction–modification(R–M) enzyme field had been to generate novelrestriction endonucleases with longer recognitionsites either by mutating or by engineering existingType II enzymes.

Functional domains in FokI restriction endonuclease

Restriction enzymes have a dual function, namelyDNA recognition and DNA cleavage. In the case ofType II enzymes, these functions overlap each other.Attempts to generate new specificities, particularlylonger recognition sites, by genetic manipulation oftheexistingType II enzymeswereunsuccessful. This isprobably due to the fact that multiple mutations areneeded before a change in specificity can be achieved.Alternatively, since the DNA recognition and catalyticfunctions overlap each other in Type II enzymes,attempts to changeamino acid residues responsible for

Fig. 2. Publication milestones for genome engineering using programmable nucleases.

965Review: Programmable Nucleases for Genome Engineering

sequence specificity may also affect the catalyticactivity. Changes in the DNA-binding domain mayalter the geometry of the catalytic sites, which is likelyaccompanied by a drop in cleavage activity by severalorders of magnitude making the mutant enzymesinactive. Therefore, we reasoned that the commonType II enzymes are not the best substrates forchanging sequence specificity.To circumvent this problem, we chose to study a

Type IIS restriction enzyme, FokI, that recognizesthe non-palindromic pentadeoxyribonucleotide5′-GGATG-3′:5′-CATCC-3′ in duplex DNA andcleaves 9/13 nucleotides downstream of the recog-nition site. We speculated that two separable proteindomains are likely to be present within FokI: one forsequence-specific recognition of DNA (FR) and theother one for the endonuclease activity (FN). Oncethe DNA-binding domain is anchored at the recog-nition site, a signal is transmitted to the endonucle-ase domain probably through allosteric interactions,for the cleavage to occur. If this were the case, wereasoned that the Type IIS enzymes are the idealcandidates for changing sequence specificitiesbecause one may be able to swap the recognitiondomains with other naturally existing DNA-bindingproteins that recognize longer sequences.Since none of the Type IIS R–M systems was

cloned at that time, we started by cloning the FokIR–M system from Flavobacterium okeanokoites (Wuand Chandrasegaran, 1989, unpublished results).Two other laboratories also independently clonedthe FokI R–M system [5,6]. We used PCR tore-design the FokI endonuclease gene to over-express the 66-kDa enzyme in Escherichia coli toobtain ~50 mg/L of the purified enzyme. The

prevailing opinion at that time was that FokI, whichrecognizes an asymmetric sequence and is abouttwice as large as the Type II enzymes (like EcoRIand BamHI), functions as a monomer while the TypeII enzymes that recognize a symmetric sequencefunction as homodimers.We performed proteolytic fragment analysis of

FokI endonuclease using trypsin, which revealed aseparable 41-kDa N-terminal DNA-binding domain(FR) and a 25-kDa C-terminal domain (FN) withnon-specific DNA cleavage activity [7]. We con-firmed the results by making C-terminal deletionmutants of FokI restriction endonuclease [8]. Wethen extended the predicted α-helix linker at thejunction of the DNA-binding domain and the cleav-age domain by one turn by inserting 4 amino acidsinto native FokI to move the cut site furtherdownstream by 1 bp, that is, from 9/13 bp (for thenative enzyme) to 10/14 bp (for the mutant enzyme)[9,10]. This further confirmed the boundaries of thefunctional domains within FokI. Waugh and Sauerlater showed that single-amino-acid substitutionscould uncouple the DNA binding and scissionactivities of FokI [11,12]. We analyzed the DNA-binding mode of FokI restriction enzyme in theabsence of magnesium ions by DNA footprinting,which showed a lack of protection at the cut site[8,13].Wah et al. later reported the crystal structures of

the native FokI and FokI bound to DNA [14,15]. Thestructures confirmed the modular nature of FokIendonuclease (Fig. 3a). They also revealed that thecleavage domain is sequestered by the recognitiondomain, thereby, restricting it from making any DNAcontacts. The structures were consistent with the

Fig. 3. Crystal structure of FokI and FokI bound to DNA.(a) Structure of FokI–DNA complex and FokI enzymealone. In both structures, the FokI cleavage domainpiggybacks on the recognition domain. (b) Native FokIcrystallizes as a dimer. The dimer interface is at the FokInuclease domains, which is formed by two salt bridgesbetween arginine (R) and aspartic acid (D) residues of theFokI monomers. Figure adapted from Wah et al., 1998.


DNA footprinting analysis, confirming that thecleavage domain did not make any sequence-specificDNA contacts at the cut site. Thus, the crystalstructure was amazingly in complete agreementwith the model derived from rigorous biochemicalstudies.

First chimeric restriction endonuclease

The bipartite nature of FokI endonucleasesuggested that it might be feasible to construct chimericnucleases with novel sequence specificities by linkingother naturally occurring DNA-binding proteins to thecleavage domain of FokI endonuclease. This indeedproved to be the case. We constructed the first“chimeric” restriction endonuclease by linking theDrosophila Ubx homeodomain to the cleavage domainof FokI [16]. We then demonstrated the creation of

other novel site-specific endonucleases by linking zincfinger proteins (ZFPs) (see below) and the N-terminal147 amino acids of the yeastGal4 protein, respectively,to the cleavage domain of FokI [17,18]. Thus, all threeeukaryotic DNA-binding motifs known at that time,namely the helix–turn–helix motif, the zinc finger (ZF)motif and the basic helix–loop–helix protein containinga leucine zipper motif, were converted into novel site-specific endonucleases [16–18]. Such engineeredchimeric nucleases were shown to make specificcuts in vitro very close to their cognate sites [16–19]. Yang-Gyun Kim, working as a post-doctoralfellow in Alex Rich's laboratory, reported theconstruction of Z–DNA conformation-specific endo-nuclease [20,21].

ZFNs (1996)

ZF motifs

Klug and co-workers reported the discovery ofrepetitive zinc-binding domains in transcription factorIIIA from Xenopus oocytes, which were later termedZF motifs [22]. Each ZF is composed of ~30 aminoacid residues containing two invariant pairs ofcysteines and histidines that bind a zinc atom. ZFsare highly prevalent in eukaryotes. Pavletich andPabo reported the crystal structure of a set of threefingers from the mouse transcription factor Zif268 incomplex with its target DNA, and this gave a glimpseof how the ZFs recognize their cognate sites (Fig. 4)[23]. The structure of the ZF was consistent with theproposed structure for the zinc-binding domains [24].The 30-amino-acid Cys2His2 ZF fold is a unique ββαstructure that is stabilized by a zinc ion [23]. Each ZFusually recognizes 3- to 4-bp sequence and bindsDNA by inserting the α-helix into the major groove ofthe double helix. The crystal structure suggestedthat amino acids within the α-helix (positions −1, +1,+2, +3, +5 and +6) of the ZF could be changed whileconserving the remaining amino acids as a consen-sus backbone to generate ZFs with new sequencespecificities (Fig. 4). Most ZFs make contact withtheir target 3-bp site; however, when there is anaspartic acid residue present at +2 position of theα-helix, it can enforce an adenine or a cytosine baseoutside the 3-bp site at the next base on thenon-contact strand of DNA via a cross-strandcontact, changing the ZFs recognition to a 4-bpsite (Fig. 4). This ZF contact outside the 3-bp sitefurther influences the specificity of neighboring ZFs,complicating the generation of ZFPs by simplemodular design, where each ZF recognizes a tripletsequence. Therefore, design and selection of eachZF has to be performed in a context-dependentfashion to obtain highly sequence specific ZFPs,which is laborious and time consuming. Normally,

Fig. 4. DNA recognition by ZFPs. (a) Structure of a single ZF, (b) DNA recognition by ZFs and (c) Structure of 3-fingerZif268 bound to its cognate site. Figure adapted from Pabo et al., 2001, and Miller and Pabo, 2001.


three to six such ZFs are linked together in tandem togenerate a ZFP that binds to a 9- to 18-bp target site(Fig. 4).With the publication of the crystal structure of

Zif268–DNA complex, scientists attempted to iden-tify unique ZFs that specifically recognize each ofthe 64 possible DNA triplets using phage display[25–35]. Inherent in this concept was the assumptionthat each ZF within a ZFP acts independent of itsneighbors to bind its cognate sequence. Scientistsreasoned that, by linking different ZFs selected insuch a way in tandem, one would be able to designZFPs that bind to any desired target DNA site withina complex genome. Thus, one could unravel theZFP–DNA recognition code, if one exists in nature.Early selection schemes kept the first and the thirdmodules of a 3-finger ZFP constant to generate alibrary of 3-finger ZFP mutants with new sequencespecificities by varying the amino acids within theα-helix (positions −1, +1, +2, +3, +5 and +6) of thesecond ZF while conserving the remaining aminoacids as a consensus backbone. Selection of thevariants that bound strongly to a desired 9-bp targetsite was performed using phage display to identifyZFs that supposedly bound specifically to 5′-GNN-3′,5′-ANN-3′, 5′-CNN-3′ and 5′-TNN-3′ triplets [31–35].Although early attempts at modular design of3-finger ZFPs using ZFs recognizing 5′-GNN-3′

triplets were somewhat successful, further attemptsto generate highly specific ZFPs by including otherZFs identified for the 5′-ANN-3′, 5′-CNN-3′ and5′-TNN-3′ triplets did not yield the requisite specificityfor their target sites. A high degree of failure ensuedwith these ZFPs, which became apparent whenthese ZFPs were fused to FokI cleavage domain toform ZFNs (see below). It soon became clear that therecognition of DNA by the ZFs was not truly modularas one had expected and that each ZF's recognitionwas greatly influenced by its neighbors [36–39].Generation of a library of mutants by changing of theamino acids within the α-helices (positions −1, +1,+2, +3, +5 and +6) simultaneously of all three ZFs toincorporate the contribution from its neighbors wassimply not feasible because of the huge number ofmutants in such a library (1.27 × 1025) and due tothe practical limit of E. coli transformation efficiency(~109 to 1011).Once the limitations of the modular assembly of

ZFPs became apparent [29,37], scientists began todevelop alternate strategies to select individual fingersin the context of their neighbors to generate highlyspecific ZFPs for the desired target sites. Two suc-cessful strategies were “sequential” selection and“bipartite” selection. In sequential selection, individ-ual ZFs are selected in the context of theirneighbors, thereby, circumventing the constraints


of the modular assembly [29,40,41]. However, thisrequires construction of multiple ZFP libraries andmultiple selections for each and every ZFP that isdesired. The bipartite approach is a variation ofthe sequential selection, which utilizes twopre-generated ZFP libraries wherein one-and-a-halffingers of a 3-finger ZFP are partially randomized atthe key residues (α-helix positions −1, +1, +2, +3, +5and +6) that make contact with DNA [40,41]. TheN-terminal half of the ZFP is randomized in one librarywhile the C-terminal half is randomized in the other.Selection is performed in parallel using the 5′ and 3′halves respectively of the target sequence. Selectionsfrom the individual libraries are then recombined andselected again using the full target site to obtain theZFP with the desired specificity. Although both theseapproaches yield high-affinity ZFPs, they are too laborintensive and cumbersome to perform routinely.Furthermore, the selection approaches become un-tenable with increasing number of ZFs within ZFPs,since this will result in exponential increase in thenumber of ZFP mutants, all of which cannot besampled in a single phage display selection experi-ment. Improved DNA-binding specificity of polyzincfinger peptides was also achieved by using strings oftwo finger units [40]. Greater target specificity wasachieved by the way in which ZF arrays areconstructed—linking three 2-finger domains ratherthat two 3-finger domains—resulting in far greatertarget specificity through increased discriminationagainst mutated or closely related sequences.The phage display strategies require several

rounds of ZFP selection and they do not occur inan in vivo setting. Alternate cell-based selectionstrategies such as the bacterial one- and two-hybridsystems, as well as the yeast two-hybrid system, linkthe ZFP–DNA interaction to the transcriptionalactivation of a reporter gene [42–45]. The advantageof these systems over the phage display system isthat the high-affinity ZFPs for the desired target sitescan be identified in a single round of selection and itis performed in an in vivo setting, even though thebacterial systems do not account for chromatin. Aswith all genetic reporter systems, the protein activity(i.e., the binding affinity of a ZFP) in a particular cloneis coupled to the transcription of a reporter,assuming that the protein concentration in the celldoes not vary greatly from clone to clone. The cell-based selection strategies are discussed in detailelsewhere [46].

Creation of ZFNs as “programmable” nucleases

ZFPs offered an attractive framework for designingchimeric restriction enzymes with tailor-madesequence specificities (Fig. 5a). We reasoned thatone could create designer nucleases that will cut DNAat any preferred site by making ZFP fusions to theFokI cleavage domain.We started by engineering two

novel chimeric nucleases (which were later renamedas ZFNs) by fusing two 3-finger proteins, Sp1-QNRand CP-QDR with different sequence specificities, tothe FokI cleavage domain to form Sp1-QNR-FN andCP-QDR-FN, respectively [17]. Their cleavage spec-ificities were tested by digesting the 48-kb λ genome.Both ZFNs cut at specific sites within the λ genome.CP-QDR-FN preferentially bound 5′-GAG GAGGCT-3′, which is one of the four predicted consensussites present in the λ genome. Sp1-QNR-FN did notbind to any of the four predicted consensus sites in theλ genome but preferentially bound to 5′-GAG GGATGT-3′ that occurs only once in the λ genome [17].Later, we showed that the fusion of the FokI cleavagedomain to a ZFP did not influence the sequencespecificity of the ZFP and did not alter its bindingaffinity significantly [47]. During the performance ofthese experiments, one observation that neededmore scrutiny stood out: cleavage by ZFNs requireda large excess of the purified enzyme over the DNAsubstrate carrying a single recognition site, hinting tothe possibility of ZFN dimer formation (see below).

Mechanism of DNA cleavage by ZFNs

As we constructed additional ZFNs using ZFPsassembled from published ZFs of known tripletsequence specificities and tested them, they alsodid not cut at the predicted consensus target sitesbut elsewhere in the λ genome. We observed thatthe cut sites, however, had some resemblance toinverted repeats that appeared to contain partial ZFNrecognition sites. This implied that the FokI cleavagedomain may be binding at degenerate sites andpossibly functioning as a dimer. Soon afterwards,the structure of native enzyme was published in1998 [15], which showed FokI crystallized as a dimerand the protein–protein interaction occurred at thedimer interface formed by the cleavage domains viatwo symmetrical salt bridges between arginine(487R) and aspartic acid (483D) residues (Fig. 3b).The kinetics of substrate cleavage by FokI restrictionenzyme provided additional support that FokI dimer-ization is required for DNA cleavage [48,49].After seeing our publication on the creation of

ZFNs in Proceedings of the National Academy ofSciences [17], Dana Carroll approached us for acollaboration to test stimulation of homologousrecombination (HR) in frog oocytes by a ZFN-induced DSB. Since our laboratory focus was alsoto target specific genes within cells [16], we agreedto such collaboration. Initially, we purified two ZFNswith known sequence specificities and shipped themby FEDEX to the Carroll laboratory for testingenhancement of intramolecular recombinationevents in an extrachromosomal plasmid substrateencoding ZFN target sites, by microinjection of ZFNsand substrate into frog oocytes. Since the purifiedZFNs were not very stable, we purified the ZFNs

(a) ZFNs

(b) TALENs

(c) CRISPR/Cas9

Fig. 5. A schematic diagram depicting programmable nucleases recognizing their target sites. (a) 4-finger ZFNs, (b)TALENs and (c) RNA-guided CRISPR-Cas9. Figure adapted from Ramalingam et al., 2013.


almost every 3 weeks or so to ship them to Utah.This was quite tedious and a time-consuming effort,especially for a small laboratory like ours. Later, weprovided Carroll laboratory with our ZFN constructs,which they used to develop other ZFN derivatives(by swapping the ZFPs) to use in their Drosophilaexperiments. Our collaboration with the Carrolllaboratory lasted about 5 years and resulted in twonoteworthy publications [50,51], setting the stage forgenome engineering using programmablenucleases.

We investigated in more detail the requirementsfor double-strand cleavage, both in vitro [50] and invivo [51] (see below), by Zif-QQR-FN and Zif-QNK-FN enzymes that recognize 5′-GGG GAA GAA-3′and 5′-GGG GCG GAA-3′ sites, respectively. Weconstructed and tested a collection of varying numberand orientation of their canonical binding sites encodedin a plasmid. At enzyme:substrate ratio close to 1, asingle copy of the recognition site did not supportcleavage. Only paired inverted sites in the tail-to-tailorientation showed efficient double-strand cleavage,


establishing that these were the preferred substratesfor ZFNs (Fig. 5) [50]. The cleavage efficiency of theinverted repeats also showed an exponential depen-dence on ZFN concentration. At substantially higherenzyme:substrate ratios, both ZFNs cut DNA thatcarried a single copy of their recognition site. At highenzyme concentrations, one molecule of ZFN isprobably bound specifically to the recognition site.Dimerization of the cleavage domains occurs throughinteraction either with a ZFN molecule that is boundnon-specifically to DNA in close proximity or with a ZFNmolecule in solution, promoting double-strand cleav-age. Substitution of the amino acid residues [arginine(R) or aspartic acid (D)] that form the symmetricalsalt bridges at the dimer interface with alanine (A)completely abolishedZFNcleavage activity, confirmingthat dimerization of the FokI cleavage domains isrequired for activity [50].We also examined the kineticsof DNA cleavage by Zif-QNK-FN in detail to gain insightinto how ZFNs cleave DNA and how two inverted sitespromote double-strand cleavage [52]. Substratecleavage was not first order with respect to theconcentration of Zif-QNK-FN, indicating that dou-ble-strand cleavage required dimerization of thecleavage domain. Thus, these findings that cleav-age domains of ZFNs must dimerize to effectefficient double-strand cleavage complementedthe observations from the natural FokI enzyme[48,49]. In contrast to ZFNs, which preferred pairedinverted repeats as substrates for efficient cleavage,the natural FokI enzyme cut single sites with veryhigh efficiency. The requirement for inverted pairedbinding sites in close proximity distinguishes ZFNsfrom FokI. In the latter case, it appears thatdimerization occurs between monomers bound toquite distant sites on the DNA [48,49].

ZFPs fusion to other functional moieties

While we were the first to fuse the FokI cleavagedomain to ZFPs with a focus to develop ZFNs astools for genome engineering, other laboratorieshave shown that ZFPs could be fused to otherfunctional moieties such as activator and repressordomains to form hybrid proteins that function eitheras zinc finger transcription activators (ZFAs) or aszinc finger transcription repressors (ZFRs) withincells [28,31,53–55]. Tim Bestor's group showedthat cytosine methylation could be targeted topre-determined sequences by fusing a CpG-specificDNA methyltransferase to Zif268 to form zinc fingermethylases (ZFMs) [56].The mechanism of action by ZFNs (Fig. 5) is quite

distinct from those of ZFAs, ZFRs and ZFMs. TwoZFNs are required to bind to two recognition sites inan inverted tail-to-tail orientation to promote anefficient double-strand cleavage [50,51]. ZFAs,ZFRs and ZFMs, on the other hand, bind to a singlerecognition site for their activity [54–56]. The

implications from the observation that ZFNs requiretwo copies of an inverted site to produce a DSB wastremendous; it meant that 3-finger ZFNs effectivelyhave an 18-bp recognition site that is large enough totarget a unique genomic address for cleavage withinplant and mammalian cells. It also signified that twoZFNs with different sequence specificities couldcollaborate to cleave in tandem to produce a DSBwhen their binding sites are appropriately positionedand oriented with respect to each other within agenome (Fig. 5) [50,51,57]. Later, we developed asimilar concept for split heterodimeric DNA methyl-ases as a platform for creating designer ZFMs fortargeted DNA methylation in cells, in collaborationwith the Ostermeier laboratory at The Johns HopkinsUniversity [58].

Stimulation of HR in frog oocytes by targetedcleavage using ZFNs

We tested the ZFNs for their ability to find and cleavetheir target sites in living cells [51]. The engineeredextrachromosomal plasmid DNA substrates and thenucleases when injected into Xenopus laevis oocytenuclei showed DNA cleavage and subsequent HR.Specific cleavage required two inverted copies of theZFN recognition site in close proximity, reflecting theneed for dimerization of the FokI cleavage domainsfor double-strand cleavage; these results wereconsistent with the in vitro studies [50]. CleavedDNA substrates were activated for HR in oocytes,and under optimum conditions, almost 100% of thesubstrate recombined, even though the substrateDNA was assembled into chromatin. We alsoshowed that two ZFNs, namely Zif-QQR-FN andZif-QNK-FN with different binding specificities, couldcollaborate in vivo to stimulate recombination bycleaving at the inverted hybrid recognition site [51].Because the recognition specificity of ZFPs can bealtered experimentally, this foundational work estab-lished the potential for inducing targeted recombinationin a variety of organisms and ushered in the era ofgenome engineering.After the successful collaborative effort with the

Carroll laboratory, our laboratory focus turned totherapeutic applications of ZFNs for targetedcorrection of disease-causing mutations in humanstem cells, particularly those of monogenic dis-eases [59–65]. The Carroll laboratory went on topursue applying the technology in Drosophila andother organisms (see Dana's reflections below)[66–70].

Genome engineering of mouse melanocytes andhuman cells using designed ZFNs

It was quite clear to us early on that facileproduction of ZFNs and rapid characterization oftheir in vitro sequence-specific cleavage properties


was a pre-requisite before ZFN-mediated genetargeting could become an efficient and effectivepractical tool for widespread use in biotechnology.Around 2000, we proceeded to use modular designto engineer ZFNs that target specific endogenoussequences within two mouse genes (mTYR andmCFTR) and two human genes (hCCR5 andhDMPK), respectively [59]. We were particularlyinterested in targeting the CCR5 locus of the humangenome for two reasons: (1) CCR5 is a co-receptorinvolved in HIV-1 infection of macrophages and Tcells. Homozygous inactivating mutations ofCCR5 are present in a subset of healthy humans,and CCR5 was thought to be dispensable fornormal cellular differentiation and function [71].We reasoned that CCR5 could potentially be anattractive therapeutic target for future gene ther-apy to treat and/or prevent HIV/AIDS by knockingout both alleles of the CCR5 gene in patient-derived hematopoietic pluripotent stem cells. (2)More importantly, we reasoned that ZFN-mediatedinactivating mutations of CCR5 should not affectthe biology of human pluripotent stem cells (hPSCs)and CCR5 could potentially serve as a safe harborlocus for targeted addition of a therapeutic gene forfunctional protein complementation (i.e., for geneactivation) in cells with the corresponding recessivemonogenic defects.The ZFN designs for the four targets were based

largely on ZFs that recognize 5′-GNN-3′ triplets. Weshowed that the engineered ZFN constructs recog-nize their respective cognate DNA sites encoded ina plasmid substrate in a sequence-specific mannerand induce a DSB [59]. However, when we testedthe mTYR ZFNs for gene correction at the endog-enous locus of albino mouse melanocytes by HDR,they were too toxic to cells. Although we observedpigmented cells by microscopy, as early as 4 dayspost-nucleofection of albino mouse melanocytesusing mTYR ZFPs fused to wild-type FokI cleavagedomains and the correcting donor, the number ofpigmented cells started to decline rapidly with time.Similarly, when we tested hCCR5 ZFNs for CCR5gene disruption in human cells by NHEJ, they alsoproved to be highly toxic to cells. Subsequent attemptsto reduce ZFNs toxicity by regulated expression ofZFNs using Tet-Off system in melanocytes andmodel the allosteric control of FokI nuclease domaininto ZFNs by incorporating more of the linker regionwere not successful [72].In 2003, Mathew Porteus in David Baltimore's

laboratory published a gene targeting reportersystem in cultured human cells based on HDR ofa mutant GFP gene that was interrupted by a targetfor 3-finger ZFNs of known sequence specificity[73]. Using the 3-finger ZFNs that we provided tohim, he showed that the expression of ZFNs in thepresence of a correcting donor led to a consider-able enhancement of GFP-positive cells via HDR.

He also observed a decline of GFP-positive cellswith time, indicating ZFN toxicity due to off-targetcleavage [73].In 2005, the first successful targeting of an

endogenous gene in human cells was accomplishedby scientists at Sangamo Biosciences [74]. Theychose to target the IL-2Rγ gene that is foundmutated in some X-linked severe combined immu-nodeficiency patients. They designed a pair of4-finger ZFNs that was very effective in inducingboth mutations by NHEJ and sequence replace-ments by HDR. They also showed that furtherrefinement of ZFNs for specificity improved thetargeting efficiency [74].ZFN toxicity, through cleavage at off-target sites,

was first shown in experiments with Drosophila [66].Several laboratories including ours independentlyobserved ZFN toxicity while performing gene target-ing in mammalian cells [72–74]. Since only twocopies are available for most genes in cells fortargeting, we reasoned that the ZFN toxicity wasmore likely from induction of multiple DSBs in thegenome due to ZFNs binding to inverted partial ordegenerate sites. The toxicity issue becomes moreacute especially with increasing concentration ofZFNs in cells due to their continued expression fromthe transfected plasmids encoding them. ZFNtoxicity could be lowered either by selecting highlyspecific ZFPs or by increasing the number of ZFswithin ZFNs to increase sequence specificity.Alternatively, one could re-design the dimer inter-face of the FokI cleavage domains to generateheterodimer variants that will actively cleave only atheterodimer binding sites and not at the homodimeror single sites. We had previously shown that theactivity of the ZFNs could be abolished by mutatingthe amino acid residues that form the salt bridges atthe FokI dimer interface [50]. Both of theseapproaches were successfully employed by scien-tists to increase the sequence specificity of ZFNsand lower cytotoxicity [75–77].After seeing the reports from Sangamo scientists

[75] and the Cathomen laboratory [76] in 2007, wegenerated mTYR and hCCR5 ZFNs with obligateheterodimer FokI nuclease domain variants (seebelow) to reduce toxicity. We tested the re-designedpairs of 3-finger hCCR5-ZFNs and mTYR-ZFNsusing the GFP gene targeting reporter system.Both readily yielded GFP-positive cells, indicatinggene correction by HDR; they also showed greatlyreduced cytotoxicity [62]. These studies clearlyestablished that the sequence specificity of thedesigned ZFNs was the major determinant of ZFNactivity for efficient gene targeting and reducedcytotoxicity. The potential of ZFN technology appli-cations for human therapeutics thus depended onthe ability to produce ZFNs that cleave the targetsequences with exquisite sequence specificity andlow cytotoxicity.


With the realization that modular assembly oftendoes not yield highly specific ZFNs, we developed aone-hybrid bacterial reporter system for interrogationof ZFP–DNA interactions, in collaboration with MarcOstermeier's laboratory at Johns Hopkins University[45]. While this approach yielded specific 3-fingerZFNs, it was tedious and time consuming like othersimilar efforts (see above). Later studies by otherlaboratories confirmed our findings that ZFNsgenerated by modular assembly does not alwaysyield highly specific ZFNs [78].

High-throughput generation of ZFNs

For high-throughput generation of ZFNs, furtherrefinements to cell-based screening strategies forZFPs were needed. They include selecting a desiredZFP from a restricted library of functional ZFPswhere each member has been previously shownexperimentally to bind to its target with high affinityand specificity. The functional ZFPs are chosen froma collection that has been accumulated in adatabase. Two such strategies are oligomerizedpool engineering (OPEN) reported by the Zinc FingerConsortium led by Keith Joung [79] and context-dependent assembly (CoDA) [80]. The OPENstrategy utilizes an archive of zinc finger pools,each consisting of a small number (~95 or fewer) ofdifferent fingers designed to bind to a particular3-bp subsite. For OPEN selection, a combinatoriallibrary of multi-finger arrays from these pools isgenerated for a target 9-bp sequence of interest.Members of this library, which bind efficiently to thetarget site, are then isolated using a bacterialtwo-hybrid selection system. This approach hasbeen shown to identify multi-finger ZFPs thatpossess high affinities and specificities for thetarget site. However, OPEN requires that a newlibrary of ZFP variants to be constructed andscreened for each and every new target site, whichis tedious and time consuming, limiting its utility forwidespread use. CoDA essentially is an extension ofthe OPEN method in that it uses the hundreds offunctional ZFPs that were generated and tested in theOPEN database to simplify the selection of a 3-fingerZFP for a new target site. It works as follows: for eachcentral finger, one identifies 15–20 context-dependentcombinations of N-terminal finger 1 and finger 2 and15–20 context-dependent combinations of C-terminalfinger 2 and finger 3, each of which has been shown tofunction well in a bacterial two-hybrid DNA-bindingassay. The rationale here is that joining these setsthrough a common central finger will be sufficient toinclude any context-dependent interactions betweenall three fingers of the ZFP that is eventually selectedfor the desired target site. The authors reported asuccess rate of N75% for ZFPs selected using CoDA[80]. Using a variety of approaches including thebipartite approach, Sangamo Biosciences has as-

sembled a proprietary database of functional ZFPsand ZF pairs from which they appear to be able toselect and refine highly specific ZFPs for a desiredtarget site.

Zinc finger nickases (ZFNickases)

Unlike engineered ZFNs, custom ZFNickasescleave only one pre-determined DNA strand of atargeted site. Conversion of ZFNs into ZFNickases isachieved by inactivating the catalytic activity of onemonomer in a ZFN dimer [81,82]. ZFNickasespossess robust strand-specific nicking activity invitro. Furthermore, ZFNickases also stimulate ho-mology directed repair at their nicking site in humancells, albeit at a frequency lower than that by DSBsusing ZFNs [83]. Importantly, ZFNickases appear toinduce greatly reduced levels of mutagenic NHEJ attheir target nicking site as compared to ZFNs; this islikely because nicks are not efficient substrates forNHEJ.

TALENs (2010)

Discovery of TALE motifs

TALENs [84–86] resulted from the fortuitousdiscovery of a novel TALE DNA-binding modulethat is found in plant virulence factors from Xantho-monas bacteria [87,88]. The TALE central repeatdomain consists of repeating units of 33–35 aminoacids (Fig. 5b). Each repeat is largely identicalexcept for two highly variable amino acids atpositions 12 and 13, referred to as the repeatvariable di-residues (RVDs). While each ZF recog-nizes 3–4 bases, each TALE motif recognizes asingle nucleotide, and the recognition specificity isdetermined by the RVD (amino acid residue NIrecognizes A, HD recognizes C, NG or HGrecognizes T and NN recognizes G or A) (Fig. 5b).Unlike the ZFs, the recognition of DNA by individualTALEmodules appears to be largely independent ofneighboring modules. The DNA recognition codethus provides a one-to-one correspondence be-tween the array of amino acid repeats and thenucleotide sequence of the DNA target (Fig. 5)[89,90]. This simple DNA recognition code and themodular nature of TALE motifs made them ideal forconstructing custom nucleases [91].

Transcription activator-like effector nucleases

The TALENs are designed in a modular stylesimilar to ZFNs, building largely upon the experiencegained from ZFN development (Fig. 5b). TALENsare also able to target unique loci in complexmammalian genomes. While ZFNs use the ZFs as


the DNA-binding modules, TALENs utilize TALEmotifs as the DNA recognition modules, but both usethe FokI catalytic domain as the DNA cleavagemodule. Many laboratories including ours haveshown that TALENs have the same efficiency ofcutting as ZFNs when targeted to the same genomiclocus but often appear to have slightly lowercytotoxicity [63,65].Although TALENs are easier to generate than ZFNs,

the genes encoding TALENs are about three timeslarger than ZFNs; this is because the TALE motifs aresimilar in size to ZFs but recognize a single base, whileZFs recognize 3- to 4-bp sequences. Furthermore, theinvariant and highly repetitive nature of TALE consen-sus sequences, except for the two amino acid residuesknownasRVDs,makes itmore difficult to assemble thegenes encoding for TALENs in E. coli. Virus-mediateddelivery of large highly repetitive genes encodingTALENs into mammalian cells is also problematic[92]. Initial commercial pricing for TALENs was alsohigh (~$5000 per target), putting it beyond the reach ofsmall laboratories. But, before TALENs could reallytake hold as a viable alternative to ZFNs, theCRISPR-Cas9 approach was already on thehorizon.

CRISPR-Cas9 (2012)

Discovery of adaptive immunity in bacteria andarchaea

Several excellent recent reviews describing thediscovery of adaptive immunity in bacteria and thedevelopment of CRISPR-Cas9 for genome engi-neering are available in literature [93–96]. Briefly,bacteria and archaea have evolved an adaptivedefense mechanism that uses the CRISPR-Cassystem to degrade complementary sequences pres-ent within invading viral and plasmid DNAs. Type IICRISPR-Cas systems rely on the integration offoreign DNA fragments into clustered regularly inter-spaced short palindromic repeat (CRISPR) loci upontranscription and processing result in shortCRISPR RNAs (crRNAs), which then anneal to atrans-activating crRNA (tracrRNA) enabling Casproteins to direct sequence-specific degradation ofthe foreign DNA [97].

Applications of CRISPR-Cas9 forgenome engineering

Gasiunas and co-workers reported that the Cas9–crRNA complex functions as an RNA-guided endo-nuclease with RNA-directed target sequence recog-nition and protein-mediated DNA cleavage, pavingthe way for engineering of universal programmableRNA-guided DNA endonucleases [98]. In elegant

work, Doudna, Charpentier and co-workers showedthat Cas9 endonuclease-mediated cleavage de-pends on tracrRNA and can also function efficientlyusing a fusion of crRNA and tracrRNA to form asingle guide RNA (sgRNA), which greatly simplifiedthe application of CRISPR-Cas9 system for genomeengineering (Fig. 5c) [99]. Several groups thenshowed that they could engineer the Type II bacterialCRISPR-Cas9 system to function with customsgRNA in human cells to direct sequence-specificcleavage [100–103]. The targeting efficiency of theendogenous loci in human cells was comparable toor better than that observed with TALENs or ZFNstargeting the same loci. These laboratories alsoshowed that, upon simultaneous introduction ofmultiple sgRNAs into human cells, they could achievemultiplex gene editing of targeted loci of the genome.They also generated Cas9 mutant nucleases callednickases, to generate only single-strand breaks at atargeted locus to promote HR while minimizingNHEJ-mediated mutagenesis [100,101]. Thus, theimprovements to the CRISPR-Cas9 system thatinclude nicking enzymes, catalytically inactive Cas9fusion to FokI cleavage domain [104,105] andsplit-Cas9 [106], have followed a very similar path asthat blazed by ZFN development and ZFP methyl-transferase fusions [58].Unlike ZFNs and TALENs, which require design

and protein engineering of two nucleases, CRISPR-Cas9 depends on RNA–DNA recognition for ge-nome engineering. RNA design is much simpler andeasier when compared to protein engineering of twonucleases each recognizing a half-site of a genomictarget sequence as in the case of ZFNs or TALENs.The advantages of the CRISPR-Cas9 systeminclude its ease of RNA design for new targets, thedependence on a single, constant Cas9 protein andthe ability to address many targets simultaneouslywith multiple guide RNAs. These have led to its wideadoption in research laboratories around the world.The CRISPR-Cas9 methodology is also very inex-pensive, making it very affordable by small labora-tories. Because of these features, many successfulapplications of the technology to genome engineer-ing have rapidly ensued.Not long after the demonstration that CRISPR-

Cas9 was effective in mammalian cells, severalgroups noted that target mutagenesis was oftenaccompanied by cleavage and mutagenesis atsecondary sites [107–110]. The level of off-targeteffects varied considerably among different targets,perhaps as a function of sgRNA design. Thespecificity of Cas9 cleavage has been enhancedby the use of paired nickases, which enforcerecognition by two sgRNAs [109,163]. This makesCas9 resemble ZFNs and TALENs. Another ap-proach has been to truncate the sgRNAs so that 17–19 bases, rather than 20, match the target [111]. Thethinking here is that one or a few mismatches in a


shorter duplex will have a larger effect on stability.Design of sgRNAs with a GG motif at the 3′ end oftarget specific sequences also appears to dramat-ically enhance genome editing by CRISPR-Cas9 inCaenorhabditis elegans [112].

Genome engineering of hPSCs usingprogrammable nucleases

hPSCs are especially well-suited for genomeengineering using programmable nucleases be-cause they can undergo extensive cell culturemanipulations while still maintaining their pluripo-tency and genome stability. The discovery thatsomatic cells can be reprogrammed to humaninduced pluripotent stem cells (hiPSCs) furtherenhanced the prospect of using genome engineeringto correct genetic diseases in patient-derived cells.For these reasons, autologous transplantation ofdisease-corrected hiPSC-derived precursors washailed as the future of regenerative medicine. Thispotential for personalized cell therapy led manylaboratories to concentrate their efforts to this end.Several laboratories have used the nuclease plat-forms for gene disruption and gene correction of avariety of precursors and disease-specific hiPSCs[113–124].After completing the work on genome engineering

of human cells [62], our laboratory focus turned tohiPSCs with the aim to address the following: (1)could we generate genetically well-defined hiPSCsby targeted addition of pluripotency genes to theCCR5 locus of human fibroblasts and cord bloodmononuclear cells, using programmable nucleases?This is in contrast to existing methods that userandom integration of stem cell factor genes into thehuman genome for reprogramming. (2) Is it possibleto generate hiPSCs with bi-allelic CCR5 knockoutusing this approach, to potentially treat HIV? (3) Afterreprogramming, could the remaining wild-type CCR5allele in single-allele CCR5-modified (heterozygous)hiPSCs be used for gene activation—that is expres-sion of a therapeutic gene? (4) Could the generationof hiPSCs be performed with simultaneous genecorrection using programmable nucleases to simplifytreatment?We used ZFN-mediated gene targeting to gener-

ate both single-allele and bi-allele CCR5-modifiedhiPSCs from human lung fibroblasts (IMR90 cells)and human primary cord blood mononuclear cells bysite-specific insertion of stem cell transcription factorgenes flanked by loxP sites at the endogenousCCR5 locus [63]. Subsequent Cre recombinasetreatment of the CCR5-modified hiPSCs resulted inthe removal of the pluripotency transgenes, leavinga loxP site in place. Further, we achieved geneticengineering of the single-allele CCR5-modifiedhiPSCs by site-specific addition of the large CFTRtranscription unit to the remaining CCR5 wild-type

allele, using CCR5-specific ZFNs. CFTR wasexpressed efficiently from the endogenous CCR5locus of the CCR5-modified hiPSCs [63].Later, we successfully generated both cystic

fibrosis (CF) and Gaucher's disease (GD) hiPSCs,respectively, from CF [homozygous for CFTRΔF508mutation] and Type II GD [homozygous for β-gluco-cerebrosidase 1448T N C mutation] patient-derivedfibroblasts, using CCR5-specific TALENs (Fig. 6)[65]. Site-specific addition of loxP-flanked Oct4/Sox2/Klf4/Lin28/Nanog gene cassette at the en-dogenous CCR5 locus of primary fibroblasts inducedreprogramming, giving rise to both mono-allele (het-erozygous) and bi-allele CCR5-modified hiPSCs [65].Like many other laboratories before us, we also

demonstrated site-specific correction of sickle celldisease (SCD) mutation at the endogenous HBBlocus of patient-specific hiPSCs [TNC1 line that ishomozygous for mutated β-globin alleles (βS/βS)],using HBB-specific TALENs (Fig. 6) [65]. SCD-cor-rected hiPSC lines showed gene conversion of themutated βS into the wild-type βA in one of the HBBalleles, while the other allele retained the mutantsequence. Excision of the loxP-flanked DNA cas-sette from the SCD-corrected hiPSC lines resulted insecondary heterozygous βS/βA hiPSCs, whichexpress the wild-type (βA) transcript to 30–40% levelas compared to uncorrected (βS/βS) SCD hiPSCswhen differentiated into erythroid cells [65]. ZFN/TALEN-mediated generation and genetic correctionof disease-specific hiPSCs did not induce anyoff-target mutations at closely related sites.The results from these studies suggest that it is

feasible to use ZFN/TALEN/CRISPR-Cas9-evokedstrategies to (1) generate precisely targeted genet-ically well-defined patient-specific hiPSCs, (2) re-shape the function single-allele CCR5-modifiedhiPSCs by targeted addition and expression oftherapeutic genes from the CCR5 chromosomallocus for autologous cell-based transgene correctiontherapy to possibly treat various recessive mono-genic human diseases and (3) correct disease-caus-ing mutations in hiPSCs. It also points to the futurepossibility of combining generation of hiPSCs withsimultaneous gene correction in a single treatment.The CRISPR-Cas9 system is probably best suited toaccomplish this. Several technical challenges thatneed to be addressed before successful translationof these strategies to human therapeutics still remain.For example, protocols for quantitative differentiation ofhiPSCs into hematopoietic stem cells and isolation ofpurified hematopoietic stem cells away from anyresidual contaminating hiPSCs for transplantation areyet to be worked out.In a review article published in Biological Chem-

istry in 1999 [57], we had anticipated the upcominggenome engineering revolution in biology andmedicine by the application of a ZFN-inducedtargeted DSB in cells, including gene correction in

(a) Scheme for generation of disease-specific hiPSCs

TALEN binding site

Cre-recombinase treatment

Secondary CCR5-modified GD hiPSCs

Donor

Cellular genome

WT CCR5 Cellular genome

WT CCR5

CCR5 modified GD hiPSCs

HR

Gauchers disease (GD) fibroblast

CCR5-specific TALENs

CCR5 CCR5LoxP

WT CCR5

WT CCR5

(b) Generation of disease-specific hiPSCs

CF fibroblast

CF hiPSCs

OCT4 DAPI

DAPI

DAPI

DAPI

SOX2

NANOG

TRA-1-60

OCT4/DAPI

SOX2/DAPI

NANOG/DAPI

TRA-1-60/DAPI

Immunostaining of GD hiPSCsGD hiPSCs

GD fibroblast

Fig. 6. TALEN-mediated generation of CF and GD hiPSCs from patient-derived fibroblasts. (a) Schematic diagramdepicts the two-step protocol that was used to generate GD hiPSCs. In a first step, the donor containing five stem cell factorgenes (OSKLN: Oct4, Sox2, Klf4, Lin8 and Nanog) and eGFP was inserted at the safe harbor CCR5 locus (usingCCR5-specific TALENs) to reprogram the patient-derived fibroblasts. In a second step, the loxP site-flanked donor wasexcised from the CCR5-modified CF and GD hiPSCs by treatment with Cre recombinase. CCR5-specific ZFNs have alsobeen used similarly to generate of hiPSCs from human fibroblasts and cord blood cells. (b) Bright field images of themorphology of CF and GD hiPSCs generated using CCR5-specific TALENs. Characterization of GD hiPSCs by Oct4/Sox2/Nanog/Tra-1-60 immunostaining and 4′,6-diamidino-2-phenylindole staining are also shown. (c) TALEN-mediatedcorrection of SCD hiPSCs. Mono-allele correction of homozygous HBB mutation of patient-specific SCD hiPSCs (TNC1line) was achieved using HBB-specific TALENs and wild-type HBB donor construct. Mono-allele gene correction wasconfirmed by sequencing the HBB locus. Parts of figure adapted from Ramalingam et al., 2014.


Exon 1GAG

Corrected allele(βA)

Mutant allele (βS)

GTG

SCD-Corrected hiPSCs

SCD-Corrected erythroid cells

(c) TALEN-mediated correction of SCD hiPSCs

Fig. 6 (continued).


hPSCs as a form of targeted gene therapy: “Theavailability of chimeric nucleases, a new type ofmolecular scissors that target a specific site withinthe human genome, will likely contribute and greatlyaid the feasibility of genome engineering and, inparticular, ex vivo gene therapy using stem cells”.Through tireless efforts of numerous scientistsaround the world, most of the predictions about theapplications using custom nucleases have cometrue.ZFNs sparked the genome engineering revolution

by first establishing that targeted modification ofcustom sites in mammalian cells including thehuman cells was possible. The development ofcompeting TALENs and CRISPR-Cas9 systems hasfueled the genome engineering revolution further bymaking nuclease-based genome engineering muchsimpler and easier and thereby extending it to smalllaboratories. It is now possible to target specificgenes for cleavage within human cells and togenerate mutant organisms on demand for diseasemodels (see Dana's reflections below). Perhaps, thegreatest impact of genome engineering usingprogrammable nucleases will likely be felt inagriculture and animal husbandry, provided thatGMO organisms become accepted (see Dana'sreflections below). Also, ex vivo genome engineeringof pluripotent stem cells using ZFNs is inching itsway to the clinic, largely through the efforts ofSangamo scientists and their collaborators. Be-cause of the potential for off-target effects, humantherapeutic applications will likely depend on riskversus benefit analysis and informed consent. But,we are yet to define what an “acceptable” or“minimum” risk is. We define the minimum risk

profile as a list of NHEJ mutations in regions of thehuman genome that, if they result from treatment,would not adversely affect the well-being of thepatient. When this profile is worked out, targetedgenome engineering using pluripotent stem cells asa form of cell-based gene therapy could be usedroutinely in clinical practice, signifying a paradigmshift in the treatment of human diseases, with thepotential to “cure” some of them.

Reflections on the Origins and Uses ofProgrammable Nucleases for GenomeEngineering—Dana Carroll

Zinc finger nucleases

For me, the story of the programmable nucleasesbegins with envy—envy of those people working withyeast and mice, who could make targeted genomicmodifications with relative ease and success. In bothsystems, the modifications depended on HR be-tween a genomic target and an exogenous donorDNA supplied by the experimenter. Also in bothsystems, the absolute frequency of the recombina-tion events was very low, but the desired productscould be recovered by strong selection. To translatethis type of gene targeting to other organisms, it wasclear that the frequency of recombination had to beimproved.It seemed clear that the barrier to enhancing

recombination was the fact that an intact genomictarget was essentially inert. Recombination is a modeof DNA repair, and if there is no damage, there is no


motivation to recombine. David Segal and I conceivedin the abstract of reagents that would have separateDNA recognition and damaging elements [125]. Incollaboration with Peter Glazer and colleagues at YaleUniversity, we worked on triplex-forming oligonucleo-tides linked toapsoralenmoiety that creates interstrandcross-links—potentially recombinagenic damage[126,127]. Some enhancement was seen, and elabo-rations of this approach are still being pursued [128],but the frequencies were still rather low.By the late 1980s and mid-1990s, several studies

had made it clear that a frank, intentional DNA DSBstimulates recombination with a homologous donorDNA. Also, in many systems, repair of the break byNHEJ often leads to production of local insertion anddeletion mutations. The pioneering experiments withthe meganucleases HO and I-SceI required theintroduction of the corresponding recognition sitesinto the genome before cleavage by the subse-quently provided enzyme [3,129–131]. To general-ize this approach to arbitrary targets, we neededcleavage reagents that could be programmed torecognize novel sequences while retaining highspecificity.When the paper from the Chandrasegaran labo-

ratory describing the ZF-FokI fusions appeared in1996 [17], we thought that they had promise. Weknew about ZFs and their modular style of DNArecognition, and linkage to a non-specific cleavagedomain created just the type of reagent we envi-sioned. I quickly phoned Chandra and established acollaboration to test the capabilities of these chimericrestriction enzymes (now ZFNs) in eukaryotic cells.As described in more detail by Chandra, we learnedabout the requirement for dimerization of thecleavage domain and the utility of paired bindingsites. We also examined the relationship betweenthe linker separating the domains of the protein andthe spacing between ZF binding sites and made amolecular model to account for the findings [50].Importantly for subsequent research, we showed

that the ZFNs very effectively cleaved their sub-strates in the nucleus of a living cell—the X. laevisoocyte—and initiated efficient recombination ofappropriately designed substrates [51]. It was notcertain ahead of time that this would work. While ZFsare components of many natural eukaryotic tran-scription factors, FokI is a bacterial enzyme; thus, itscleavage domain would never have been challengedwith a substrate assembled into chromatin. Remark-ably, both cleavage and recombination approached100% in this system.

ZFNs in Drosophila

Once the basic requirements for ZFN cleavagewere established, it was time to try them in a wholeorganism on an endogenous genomic target. Thispresented a number of potential stumbling blocks.

Success in the Xenopus oocyte experiments indi-cated that chromatin per se would not be a barrierbut that system was essentially synthetic. The targetwas in a plasmid DNA that had been injected intothe nucleus and, although assembled into chroma-tin, might not have been identical to chromosomalstructures. The ZFNs were ones of known specific-ity, and the targets were constructed to match them;genuine genomic targets would require new ZFcombinations that might or might not be effective. Inwhatever organism we decided to test the ZFNs, wewould have to deliver them and designed donorDNAs appropriately.We made the fortunate choice to talk with Kent

Golic about applying the ZFNs in the fruit fly,Drosophila melanogaster. He and Yikang Ronghad just devised a gene targeting method for fliesthat was based on producing a linear donor DNA insitu [132]. We proposed to enhance this approach bycutting the target with ZFNs, which would(if successful) yield targeted mutations via NHEJ,as well as increasing the frequency of homologousrepair.Kent recommended that we try the yellow (y) gene

as our initial target. The mutant phenotype is veryeasily scored, even in somatic mosaics, since itsproduct acts cell autonomously; he and Rong hadtargeted it successfully in their experiments. MarinaBibikova found a site in the second exon of y thatlooked like a plausible target for two 3-finger ZFNs. Itwas composed of two 9-bp sequences made up ofGNN triplets, for which ZFs had been identified inCarlos Barbas's laboratory and separated by theoptimal distance of 6 bp. Starting with plasmidssupplied by Carlos and Dave Segal, who was by thena post-doctoral in that laboratory, Marina producedcoding sequences for the required 3-finger sets andlinked each of them to the coding sequence of the FokIcleavage domain from the Chandrasegaran plasmids[17]. Following Kent's advice, these ZFN genes—yAand yB—were placed in P element vectors under thecontrol of a Drosophila heat shock promoter.In the Golic laboratory, we established yA and yB

transgenes in the genomes of separate stocks.Marina crossed yA flies with yB flies and subjectedthe larval offspring to a heat shock. Initially, all theheat-shocked larvae died, and this was true of larvaecarrying only yA but not yB alone. When wemoderated the heat shock temperature to 34 °C,rather than 37 °C, most of the larvae survived andemerged as healthy adult flies [66].The y gene is on the Drosophila X chromosome;

thus, we hoped that we might see evidence ofmutations in males after ZFN induction. In the firstbatch of heat shock survivors, Marina saw a fly with amottled cuticle that she thought might indicatesomatic mutagenesis [66]. I was not experiencedenough to make a definitive call; thus, we invitedKent to the microscope. After a quick look, he stood


up and said, without emotion, “If I were you, I'd bepretty excited”. On the strength of that one fly, Ibought celebratory wine for Kent's laboratory and myown. We found several more somatic mosaics inthat first experiment and showed that the mutationwas heritable in some cases, indicating that we hadsuccessfully targeted the germ line. DNA sequenc-ing of several of the induced mutations showed thatthey were exactly the indels expected for error-proneNHEJ [66].We then combined expression of the y ZFNs with

the Golic method for delivering a marked donor DNA[132]. Remarkably, we achieved quite good levels ofHR with only a single copy of the donor delivered insitu, either in circular or (even better) in linear form[67]. The enhancement over providing the lineardonor alone was up to 100-fold, establishing theprinciple that has guided nuclease-based genomeengineering ever since. ZFNs for the next twoDrosophila genomic targets we attacked (brown,bw and rosy, ry) were also successful [70]. Isometimes wonder what we would have done if theZFNs for y had failed. Would we have persistedor concluded that the approach was not going towork?In a later study, we showed that we could achieve

efficient ZFN targeting—both NHEJ and HDR—bydirect embryo injection of mRNAs for the proteinsand an appropriate donor DNA [133]. This obviatedthe need for extensive strain construction, and theinjections could be made into any strain background.The approach of embryo injection is also a method ofchoice for many other organisms.

ZFNs in other organisms

There was nothing about the ZFN-mediatedstimulation of targeted mutagenesis and genereplacement that was specific to flies, except themode of delivery. Therefore, we were eager to try theapproach in other organisms that lacked a usefulgene targeting method. A collaboration with GaryDrews' laboratory demonstrated that ZFNs workedwell in the model plant, Arabidopsis thaliana [68]. Inthis case, we targeted an integrated syntheticsubstrate, not an endogenous genomic target, withcharacterized ZFNs and only assessed NHEJmutagenesis. Nonetheless, this showed that ZFN swere active in plants and that the challenge ofdelivery could be solved [68]. At the same time, DanVoytas and colleagues demonstrated the utility ofZFNs in tobacco cells [134]. These experiments alsoemployed a synthetic target and ZFs of knownspecificity but demonstrated homologous repair andthe regeneration of whole plants from modified calli.The nematode, C. elegans, was and remains a

very popular experimental organism for studies ofdevelopment, neurobiology and other areas, butresearchers were reliant on random mutagenesis for

genetic studies. Jason Morton and I collaboratedwith Erik Jorgensen and Wayne Davis to demon-strate high levels of cleavage and mutagenesis byZFNs at both a synthetic extrachromosomal targetand an endogenous genomic target [69]. The latter,like our experiments with Drosophila genes, involvedthe construction of novel sets of ZFs. We werefrustrated, however, with our inability to produceheritable mutations. C. elegans was notorious forresisting the expression of transgenes in the germline, and we never overcame that limitation despiteconsiderable effort. Particularly with the advent ofthe more efficient TALEN and CRISPR-Cas9 plat-forms, many laboratories have now made germ-linemodifications—both by NHEJ and by homologousrepair—in the worms [135–137]. With the idea ofevading RNA interference in the C. elegans germline, I suggested to Jin-Soo Kim the approach ofinjecting pre-formed Cas9-sgRNA ribonucleoproteininstead of DNA vectors or mRNA. He and hiscollaborators showed this to work well [138], andribonucleoprotein delivery has since been used veryeffectively in mammalian cells [139,140].The first application of ZFNs tomammalian cells was

made by Matt Porteus in David Baltimore's laboratory[73]. He inactivated a GFP gene by inserting recogni-tion sites for characterized ZF sets and showed thatcleavageby the correspondingenzymesstimulated therestoration of fluorescence by recombination with adonor DNA fragment. The first attack on a humangenomic target was reported by Urnov et al. in 2005[74]. They achieved a homologous replacementfrequency of almost 20% at the IL-2Rγ locus and, inthe process, signaled the interest of Sangamo Biosci-ences in tackling human disease genes.Another ZFN success story was the application to

rats [141,142]. That organism was, and still is, incommon use for physiological studies, including drugtesting, but had no useful embryonic stem cells andthus no gene targeting method comparable to that inmice. Geurts et al. producedmutations in an integratedGFP gene and two endogenous loci by injecting ZFNplasmids or mRNAs into one-cell embryos and wereable to demonstrate germ-line transmission of themodification [141]. ZFNs were also adopted forgenome engineering by embryo injection in mice,where they have reduced the time and expense ofgenerating knockins and knockouts compared to theclassical embryonic stem cell technology [143–145].Effective ZFN mutagenesis was demonstrated in avariety of other species, including zebrafish [146,147]and other model organisms, crop plants and foodanimals (reviewed in Refs. [148–150]).

Successor Technologies

Although ZFNs were put to good use in manyorganisms, the technology was not broadly adopted


largely because of the challenge of designing newZF sets for new genomic targets. We were veryfortunate that our first three designs for Drosophilatargets all worked because we eventually hadsuccess with only 5 out of 12 attempts. ValidatedZFNs were made available for purchase, but only ata price that was prohibitive for many laboratories.Several different schemes were developed forselection of ZF combinations, but these requiredconsiderable expertise and resources, as Chandrahas described above. The anticipated code of ZF–DNA recognition [37] never materialized.Great excitement was generated by the discovery

of a simple recognition code in a novel class ofproteins, the transcription activator-like effectors(TALEs or TAL effectors) [87,88]. When somepathogenic bacteria, particularly of the genusXanthomonas, infect plants, they inject these TALEproteins into host cells, where they enter thenucleus, bind to promoter regions of host genesand alter their activity in ways that promote theinfection. The DNA-binding regions of these proteinsare composed of tandem 34-amino-acid repeats.When enough protein–target pairs had been char-acterized, it emerged that each repeat modulerecognizes a single base pair, and the most commonrepeats comprise a simple, robust recognition code.Well before any structural analysis was performed toconfirm the mode of recognition [89,90], the analogyto ZFs was obvious, and several groups linked setsof TALE modules to the FokI cleavage domain,producing TALENs that were very effective intargeted genome cleavage [84–86]. Like ZFNs, theTALENs must be delivered in pairs to satisfy therequirement for dimerization.When crystal structures of TALE clusters bound to

DNA were published [89,90], they clarified how 34amino acids (somewhat more than a single ZF) couldbe folded and polymerized to contact individual basepairs. Surprisingly, only one of the two amino acids thatconstitute the recognition code actually projects into themajor groove; the other side chain bends back andmakes contacts that stabilize the structure of themodule. Furthermore, only two of the modules (for Cand G) make hydrogen bond contacts with thecorresponding base pairs; the others establish speci-ficity by steric complementary and van der Waalsinteractions. One of the features of TALE recognition isthat consecutive modules have much more contactwith each other than is seen with ZFs. This may be thereason for the observation that single mismatchesbetween the protein andDNAhave a stronger effect onaffinity than might be expected [151]. If one or twomodules are out of register, this affects the interactionwith neighboring modules and disrupts the structure inthe vicinity. In this view, the energy of TALE bindingcomes from the scaffold that DNA provides for afavorable protein structure and relatively little fromdirect protein–DNA interactions.

The potential challenge to building new TALEarrays—because of their repeat structure, they tendto be unstable and cannot be PCR amplified—wasrapidly overcome by several assembly schemes,and new TALEN designs can be produced routinely.New designs are very often effective. In our study inDrosophila, only 2 of 17 pairs tested failed to givevery usable levels of cleavage and mutagenesis[152]. In a larger study in cultured human cells,Reyon et al. had success at 84 of the 96 endogenousgenomic targets they attacked with TALENs [91].These and other papers demonstrated that the TALErecognition code is robust, TALE assembly isroutine, TALEN activity is reliably high and mostTALENs work as long as they can be deliveredeffectively.But pity the poor TALENs. Only 3 years after the

elucidation of the TALE recognition code, theCRISPR-Cas platform arrived on the scene [99].These tools emerged slowly from studies of oddarrays of short repeat sequences that are found inmany bacterial and archaeal genomes [93–96]. Asdescribed initially in the E. coli K12 genome,repeated sequences of a few dozen base pairswere found separated by unique spacer sequencesof similar length. Dubbed CRISPRs, these arrayswere accompanied by unique sets of protein-codinggenes, called cas for CRISPR associated. Nearly20 years after their discovery, the spacers werefound to correspond to sequences from bacterialviruses and plasmids, and this led to the hypothesis,since confirmed, that CRISPR-Cas was an adaptiveimmune system that defends against invading DNA[153]. How the spacers are captured into the arraysin the first place remains a subject of research.In the simplest of the CRISPR systems, called Type

II, the CRISPR repeats are transcribed into a longRNA that is processed into smaller pieces—crR-NAs—that carry part of the repeat sequence and partof a single spacer [93–96,154]. Binding of a transact-ing RNA (tracrRNA) produced from elsewhere withinthe CRISPR locus creates a complex that associateswith the Cas9 protein. This tripartite structure is anenzyme that locates its DNA target by the presence ofa specific short sequence, called a PAM, for proto-spacer-adjacent motif, and the presence just up-stream from the PAM sequence of a sequencecomplementary to the 5′ end of the spacer fragmentin the crRNA (Fig. 5c). Two nuclease active sites inCas9 then cut the two strands of the target DNA,inactivating the invader. For the widely used Cas9protein of Streptococcus pyogenes, the PAM that isrecognized in the target nGG, where n can be anynucleotide, and the standard length of complementarysequence in the crRNA is 20 nucleotides.In 2012, Jinek et al. showed that the components

of this system could be used in purified form to cutarbitrary DNA sequences by incorporating thecorresponding 20-nucleotide target sequence at


the 5′ end of the crRNA [99]. The same group alsofused the critical portions of the crRNA and tracrRNAinto a single guide RNA (sgRNA). In combinationwith the Cas9 protein, this comprises a simpletwo-component system for targeted cleavage. Hav-ing the experience with ZFNs and TALENs asprecedents, with almost no delay, a number ofgroups deployed the CRISPR tools in mammaliancells [100–103], in zebrafish [155], in mice [156], infruit flies [157,158] and in a number of othereukaryotic genomes. The remarkable simplicity ofCRISPR-Cas rapidly made it a favorite in laborato-ries around the world. No protein engineering wasrequired, a teenager could design new sgRNAs fornew targets and multiple targets could be attackedsimultaneously. The high success rate helpedestablish CRISPRs as the programmable nucleaseof choice for research. As many people have said,this platform democratized genome engineering.

Off-Target Effects

The first ZFNs for a genuine genomic targetrevealed an issue that has continued to concernthe field. As noted above, expressing the ZFNs forthe yellow gene too enthusiastically was lethal toflies [66]. This turned out to be a property of only oneof the proteins, yA; when it was expressed alone, itwas toxic. We later showed that this was due tocleavage, since a mutation in the nuclease activesite relieved the toxicity (and any efficacy, of course)[70]. This suggested that the ZFs of yA were able tobind and form dimers at sequences other than thosefor which they were designed, and cleavage atsecondary targets was the cause of the lethality.Off-target cleavage has been confirmed in manyother situations, with TALENs and Cas9 as well, byvisualizing multiple repair foci induced by thenucleases and by analyzing genomic sequencesfor induced mutations.In the case of CRISPR, it may be that tolerance of

one or a few mismatches is an important feature ofthe system [159]. Viral genomes evolve rapidly, andallowance for this fact provides immunity towardviruses related, but not identical, to the ones thatestablished the CRISPR array. Something similarmay be true of TALEs; in the ongoing war betweenpathogen and host, the bacterium needs scope torespond to sequence changes at the target in thehost promoter. Perhaps similarly, eukaryotic tran-scription factors that depend on recognition by ZFstypically show a consensus binding sequence that isnot rigidly matched at all their targets. Thus, a lack ofperfect discrimination against off-target sequencesmay be inherent in the recognition components of allthree platforms.Schemes have been designed to enhance the

specificity of all the programmable nucleases. For

ZFNs and TALENs, mutations have been introducedat the dimer interface of the FokI cleavage domainthat allow heterodimers to form but not homodimers[75–77,160]. This means that ZFNs such as yA areno longer toxic, and it has been quite effective inreducing off-target cleavage with many other de-signs. With CRISPR-Cas9, truncating the guidesequence has a positive effect on specificity[111,161] apparently because a small number ofmismatches in a shorter duplex has a largerdestabilizing effect at secondary sites. Kim andco-workers found that having two extra Gs at the 5′end of the sgRNA that do not match the targetimproved specificity without impairing on targetefficacy [162]. This maneuver may also somewhatdestabilize the RNA–DNA hybrid because the extranucleotides are not comfortably accommodated intothe Cas9 binding pocket. A mutation in one of thenuclease active sites of Cas9 converts it into anicking enzyme. Providing two sgRNAs directed totargets in close proximity on the DNA leads to pairednicking, which has proved quite effective at inducingboth NHEJ and HDR while demanding recognitionby both guides simultaneously [109,163]. I particu-larly like this scheme, as it essentially convertsCRISPR-Cas9 into an analog of ZFNs and TALENs,which are naturally “paired nickases”. I still find itremarkable, however, that nicks up to 100 bp apartare still processed similar to a DSB in cells. Finally,Cas9 inactivated at both active sites has been fusedto the FokI cleavage domain, producing an evencloser analog to ZFNs [104,105]. This construct alsorequires simultaneous recognition of two bindingsites, with a consequent increase in specificity.How much off-target cleavage matters depends on

the situation in which the programmable nucleasesare used. For many research applications, thedesigned alteration in the target is most important,and accommodation can be made for things goingon in the background. To attribute a phenotype to anengineered change, however, one must use inde-pendently derived mutated alleles, do extensiveout-crossing or rescue the phenotype by comple-mentation. For applications to food organisms,and particularly to human therapy, stringent criteriamust be applied, and out-crossing is not a sensibleoption.When cloned cells or whole organisms are used,

whole genome sequencing can reveal all inducedsequence changes. It must be remembered, how-ever, that such changes accrue naturally during DNAreplication in every cell cycle. The only way toconfidently identify nuclease-induced off-target mu-tations is to find indels (which are less commonduring replication) at sites sufficiently related to thedesigned target to be plausible recognition sites.Whole genome sequencing of mixed cell popula-tions would have to be very deep in order to revealinfrequent, but potentially harmful, modifications.


Identifying secondary targets and assessing theirsignificance is challenging. Initially, people scannedthe genome for related sequences and analyzedthem individually by amplification and sequencing.More recent methods capture cleavage sites morebroadly, and they are quite powerful and revealing,but each has some limitations. The methods havebeen reviewed very recently [164,165], and I haveonly a few additional comments.

1. Lentiviral capture [166,167] is cumulative—that is, the products are not re-cuttable; thus,they accumulate with time. If on-targetcleavage is near saturation, off-target cap-ture will continue to occur; thus, the lattermay be exaggerated. This method is notapplicable to all cells and organisms.

2. GUIDE-seq [161] captures breaks individu-ally, but because the introduced duplexoligonucleotide has protected ends, thejunctions show almost no insertions ordeletions in the target. How this affectscapture and the analogy to indel formationis not known. Like lentiviral capture, thismethod is cumulative.

3. The LAM-PCR HTGTS method of Frock etal. uses the designed target to capturesecondary targets based on induced trans-locations [168]. It has the advantage that itdoes not depend on introduction of anythingexogenous beyond the nucleases them-selves, and the junctions observed aretypical of NHEJ. On the other hand, it onlyreports on events in cells that have seen atleast two more or less simultaneous breaks,and this may over-estimate off-target cuts. Itis also cumulative.

4. Digenome-seq [162] requires a lot of verydeep sequencing.

5. BLESS takes a snapshot of breaks presentat the particular moment at which the cellsare harvested [169,170]. This will be influ-enced by the timing relative to nucleasedelivery.

6. Someone should compare these methodson the same set of samples.

NHEJ versus HDR

All of the nuclease platforms depend on cellularrepair activities for their ultimate outcomes. As notedabove, the major pathways of DSB repair in mostorganisms are NHEJ and HDR; and in mostexperimental situations, NHEJ dominates. For sim-ple targeted mutagenesis, this is fine, but it is often

the case that a particular sequence replacementtemplated by a donor DNA is the desired product. Infavorable circumstances, HDR may account for 20–30% of the repair products, and those products canbe isolated by molecular screening; but there areorganisms and cell types in which this frequency ismuch lower, sometimes less than 1%. The balancebetween HDR and NHEJ can be shifted by disablingcomponents of the major NHEJ pathway. Forexample, we found that knocking out DNA ligaseIV in Drosophila raised the proportion of HDR productsfrom around 20% to about 65%, sometimes evenhigher [133,171,172]. In the absence of ligase IV, indelmutations characteristic of NHEJ were still recov-ered; thus, there are clearly alternative routes to thistype of repair [173]. Other methods of reducingligase IV activity have also shown promise inmammalian cells and embryos [174–176]. Becausehomologous repair naturally occurs only in the S andG2 phases of the cell cycle, when a sister chromatidis present, cell synchronization has been used tofavor HDR with some success, but not in all cells[74,177]. A general solution to the challenge ofenhancing HDR may not exist due to differencesamong cell types, the overall complexity of repairmechanisms and the difficulty in identifying factorsthat limit each type of repair.

Societal Applications

One of the most exciting aspects of ZFNs,TALENs and CRISPR-Cas9 is how broadly theyhave been used effectively [148–150]. They readilymodify genomic targets in model organisms (somecited above), disease organisms, disease vectors,crop plants, livestock and primates, includinghumans (e.g., see Refs. [63,65,116,122] and [178–187]). I am optimistic about the prospects in all thesearenas.The programmable nucleases offer opportunities

for making beneficial changes in the genomes offood organisms with precision and safety. In someplant and animal species, loci that provide toleranceto heat and drought have been identified. To moderatethe already irreversible effects of climate change and tostabilize the food supply in threatened regions, we canintroduce the beneficial sequences into the genomesof other breeds with molecular approaches, withoutlaborious and expensive crossing and back-crossing[188]. While this is technically genetic modification, itwould not involve the introduction of whole genes inmost cases and certainly not genes from other species.Muchof the current opposition toGMOs is basedon theproprietary ownership of the organisms and thepolitically insensitive way that seed companies havetaken advantage of crop producers. These consider-ations have over-shadowed the benefits and essentialsafety of the crops. What would have happened if the

Table 1. A comparison of programmable nuclease platforms

Programmable nucleases(year of introduction)

ZFNs(1996)

TALENs(2010)

CRISPR-Cas9(2012)

(1) Sequence recognition Protein–DNA Protein–DNA RNA–DNA(2) Design and selection Tedious and time consuming

(protein engineering)Easier than ZFNs

(protein engineering)Simple

(RNA design)(3) Commercial pricing Very expensive Expensive Cheap(4) Targeting efficiency Variablea Moderate Highb

(5) Off-target effects Variablea Low Moderateb

(6) Multiple targets Difficult Difficult Easy(7) Viral delivery Easy Moderate Moderate

a Depending on ZFN constructs, targeting efficiency and off-target effects can vary from high to low.b These entries are somewhat subjective in nature. As with ZFNs, targeting efficiency and off-target effects can vary depending on

constructs and selected target site designs.


first GM crop had been golden rice and it had beenprovided essentially without conditions to farmers in thethird world? I doubt that much outcry would have beenheard.Livestock improvement is also being undertaken

with nucleases [184,185]. Beyond this, there areopportunities to use genome-edited pigs, in particular,tomodel humangenetic diseases.Mousemodels havebeen very informative, but in many cases, they do notmimic the human phenotype. In size and physiology,the pig is a closer match. In addition, pig organs havebeenused for xenotransplantation, but outcomes sufferfrom immunological rejection. Methods are beingdeveloped to grow essentially human organs in pigsto overcome this limitation.Finally, of course, there are applications to human

therapy under development. Using ZFNs preparedat Sangamo Biosciences, one Phase I clinical trial isalready complete [123]. Circulating T cells wereisolated from HIV-infected patients, and the cellswere treated ex vivo with ZFNs targeting CCR5,which codes for a co-receptor required for infectionby most HIV-1 strains. The cells were re-implanted inthe individuals from whom they came, and thesituation was monitored over a period of about8 months and continued up to 42 months. Thepatients were on anti-retroviral drug therapy duringmost of the trial; thus, the efficacy of the ZFNtreatment per se was not assessed. Gratifyingly,there were no adverse effects attributable to theZFN treatment; if off-target mutations were gener-ated, they were benign. This experience highlightsseveral aspects of applying programmable nucle-ases to the treatment of humans. Ex vivo treatmentof cells is much easier than delivering the nucleasesto tissues in vivo, it is easy to see how nucleasetreatment can be combined with stem cell therapiesonce they are developed to a clinical stage andknockouts via NHEJ are still easier than correctionby HDR. The next step for the CCR5 trial is to applythe ZFNs to earlier precursors or even stem cells[122,189].

Which platform will ultimately be best for thesesocietal applications? This is not a trivial questionto answer (see Table 1). For most such purposes—clinical or agricultural—an engineered nuclease willbe used repeatedly on many different individuals.Efficacy and specificity are required, but all theplatforms have this capability, and generating thenuclease will be a small part of the overall program;thus, other considerations come into play. Asillustrated by the CCR5 trial, some highly evolvedZFNs are very specific. TALENs seem to have thehighest inherent specificity [190–193], but there areadjustments for ZFNs and CRISPR-Cas9, as de-tailed above. The ZFNs are considerably smallerthan TALENs or Cas9; thus, delivery may be easierin some contexts. Only a single, constant protein isrequired in the CRISPR case, and multiple sgRNAscan be delivered simultaneously, if several targetsneed to be addressed. Smaller Cas9 proteins frombacteria other than S. pyogenes also seem effective[170,194], and this may ameliorate some deliveryproblems.

Genome Engineering of Human Embryos

In principle, whole-body correction of a diseasegene could be achieved by editing in humanfertilized eggs, but there are many ethical andscientific issues that must be addressed beforetaking this route. The experiments with monkeysdescribed above show that the fundamental tech-nology is at hand. In addition, a very recent paperreported the use of CRISPR-Cas9 in non-viablehuman embryos [195]. Clinical applications, howev-er, require extensive prior consideration and broaddiscussion.First, targeted changes made in zygotic genome

are permanent and heritable, unlike changes madein somatic cells. Do we know enough about theconsequences of such alterations to undertakethem? Second, off-target modifications will also be


permanent. Are we confident of the specificity of theprogrammable nucleases? We would argue, not yet.Third, this is a completely novel type of humanintervention into our own genetics. Is society ready toadopt this audacious approach? Fourth, there arealternative ways to treat genetic diseases, and invitro fertilized embryos can be screened as analternative to genome modification. What diseases,if any, are candidates for germ-line editing? Fifth,there will be people who want to use germ-lineediting for trivial and cosmetic traits, raisingthe specter of eugenics. How can this powerfulyet daunting technology be confined to legitimateuses?Both of us believe, with many other scientists, that

human germ-line modification—especially of DNA inthe nucleus, the material that makes us who weare—should remain out of bounds for research andexperimentation at present [196–198]. Mitochondrialtransplantation to treat mitochondrial diseases,which has been approved by the British government,is also heritable but involves only a small number ofnon-nuclear genes.

Acknowledgements

S.C.: I thank Ham Smith for getting me interestedin restriction enzymes. I am grateful to the excellentgraduate students (Lin Li, Yang-Gyun Kim, J. Cha,Jeff Smith, Mala Mani and Joy Wu), talentedpost-doctoral fellows (Sundar Durai, KarthikeyanKandavelou, Narayana Annaluru, SivaprakashRamalingam and Heloise Muller) and many un-named undergraduate students, whose hard workmade the research on designer nucleases possible. Ithank Narayana Annaluru and Sivaprakash Rama-lingam for preparing the figures. I am deeplyindebted to several of my colleagues at JohnsHopkins Medical Institutions and elsewhere fortheir constant support and encouragement and forunderstanding my private and independent nature.Research on chimeric nucleases has been support-ed by grants from US National Institutes of Healthand Maryland Stem Cell Research Fund. D.C.: I amparticularly grateful to the people who have workedin my laboratory on the programmable nucleases. Ithas also been my pleasure to learn about the biologyof many different organisms through collaborationsand discussions with other investigators. Work in mylaboratory has been supported by research grantsfrom the US National Institutes of Health, mostrecently R01 GM078571. I need to disclose that Ireceive license royalties from Sangamo Biosci-ences, Inc., and I am a member of the ScientificAdvisory Board of Recombinetics, Inc., that is pursuingnuclease-based applications to livestock.

Received 8 August 2015;Received in revised form 12 October 2015;

Accepted 15 October 2015Available online 23 October 2015

Keywords:CRISPR-Cas9;gene therapy;

genome editing;transcription activator-like effector nucleases (TALENs);

zinc finger nucleases (ZFNs)

Abbreviations used:DSB, double-strand break; ZFN, zinc finger nuclease;

ZFP, zinc finger protein; HR, homologous recombination;TALEN, transcription activator-like effector nuclease;

NHEJ, non-homologous end joining; HDR,homology-directed repair; R–M, restriction–modification;ZF, zinc finger; ZFA, zinc finger transcription activator;

ZFR, zinc finger transcription repressor; ZFM, zinc fingermethylase; OPEN, oligomerized pool engineering; CoDA,

context-dependent assembly; RVD, repeat variabledi-residue; hPSC, human pluripotent stem cell; hiPSC,

human induced pluripotent stem cell; CF, cystic fibrosis;GD, Gaucher's disease; SCD, sickle cell disease.

References

[1] S.L. Mansour, K.R. Thomas, M.R. Capecchi, Disruption ofthe proto-oncogene int-2 in mouse embryo-derived stemcells: a general strategy for targeting mutations to non-selectable genes, Nature 336 (1988) 348–352.

[2] M.R. Capecchi, Altering the genome by homologousrecombination, Science 244 (1989) 1288–1292.

[3] P. Rouet, F. Smih, M. Jasin, Introduction of double-strandbreaks into the genome ofmouse cells by expression of a rare-cutting endonuclease, Mol. Cell. Biol. 14 (1994) 8096–8106.

[4] A. Pingoud, G.G. Wilson, W. Wende, Type II restrictionendonuclease—A historical perspective and more, NucleicAcids Res. 42 (2014) 7489–7727.

[5] K. Kita, H. Kotani, H. Sugisaki, M. Takanami, The FokIrestriction-modification systems. I. Organization and nucle-otide sequences of the restriction and modification genes,J. Biol. Chem. 264 (1989) 5751–5756.

[6] M.C. Looney, L.S. Moran, W.E. Jack, J.R. Feehery, J.S.Benner, et al., Nucletide sequence of the FokI restriction-modification system: separate strand specificity domains forthe methyltransferase, Gene 80 (1989) 193–208.

[7] L. Li, L.P. Wu, S. Chandrasegaran, Functional domains inFokI restriction endonuclease, Proc. Natl. Acad. Sci. U. S.A. 89 (1992) 4275–4279.

[8] L. Li, L.P. Wu, R. Clarke, S. Chandrasegaran, C-terminaldeletion mutants of FokI restriction endonuclease, Gene133 (1993) 79–84.

[9] L. Li, S. Chandrasegaran, Alteration of the cleavagedistance of FokI restriction endonuclease by insertionmutagenesis, Proc. Natl. Acad. Sci. U. S. A. 90 (1993)2764–2768.

http://refhub.elsevier.com/S0022-2836(15)00606-3/rf0005































[10] Y.-G. Kim, L. Li, S. Chandrasegaran, Insertion and deletionmutants of FokI restriction endonuclease, J. Biol. Chem.269 (1994) 31978–31982.

[11] D.S. Waugh, R.T. Sauer, Single amino acid substitutionsuncouple the DNA binding and strand scission activities ofFokI endonuclease, Proc. Natl. Acad. Sci. U. S. A. 90 (1993)9596–9600.

[12] D.S. Waugh, R.T. Sauer, A novel class of FokI restrictionendonuclease mutants that cleave hemi-methylated sub-strates, J. Biol. Chem. 269 (1994) 12298–12303.

[13] A. Yonezawa, Y. Sugiura, DNA binding mode of class-IISrestriction endonuclease FokI revealed by DNA footprintinganalysis, Biochim. Biophys. Acta 1219 (1994) 369–379.

[14] D.A. Wah, J.A. Hirsch, I. Schildkraut, A.K. Aggarwal,Structure of multimodular FokI endonucleases bound toDNA, Nature 388 (1997) 97–100.

[15] D.A. Wah, J. Bitinaite, I. Schildkraut, A.K. Aggarwal,Structure of FokI has implications for DNA cleavage, Proc.Natl. Acad. Sci. U. S. A. 95 (1998) 10564–10569.

[16] Y.-G. Kim, S. Chandrasegaran, Chimeric restriction endo-nuclease, Proc. Natl. Acad. Sci. U. S. A. 91 (1994) 883–887.

[17] Y.-G. Kim, J. Cha, S. Chandrasegaran, Hybrid restrictionenzymes: Zinc finger fusions to FokI cleavage domain,Proc. Natl. Acad. Sci. U. S. A. 93 (1996) 1156–1160.

[18] Y.-G. Kim, J.J. Smith, M. Durgerha, S. Chandrasegaran,Chimeric restriction enzyme: Gal4 fusion to FokI cleavagedomain, Biol. Chem. 379 (1998) 489–495.

[19] Y.-G. Kim, Y. Shi, J.M. Berg, S. Chandrasegaran, Site-specific cleavage of DNA-RNA hybrids by zinc finger/FokIcleavage domain fusions, Gene 203 (1997) 43–49.

[20] Y.G. Kim, P.S. Kim, A. Herbert, A. Rich, Construction of a Z-DNA-specific restriction endonuclease, Proc. Natl. Acad.Sci. U. S. A. 94 (1997) 12875–12879.

[21] Y.G. Kim, K. Lowenhaupt, T. Schwartz, A. Rich, Theinteraction between Z-DNA and Zab domain of double-stranded RNA adenosine deaminase characterized usingfusion nucleases, J. Biol. Chem. 274 (1999) 19081–19086.

[22] J. Miller, A.D. McLachlan, A. Klug, Repetitive zinc-bindingdomains in the protein transcription factor IIIA from Xenopusoocytes, EMBO J. 4 (1985) 1609–1614.

[23] N.P. Pavletich, C.O. Pabo, Zinc finger-DNA recognition:crystal structure of a Zif268-DNA complex at 2.1 Å, Science252 (1991) 809–817.

[24] J.M. Berg, Proposed structure for the zinc-binding domainsfrom transcription factor IIIA and related proteins, Proc. Natl.Acad. Sci. U. S. A. 85 (1988) 99–102.

[25] E.J. Rebar, C.O. Pabo, Zinc finger phage: affinity selection offingers with new specificities, Science 263 (1994) 671–673.

[26] Y. Choo, A. Klug, Selection of DNAbinding sites for zinc fingersusing rationally randomized reveals coded interactions, Proc.Natl. Acad. Sci. U. S. A. 91 (1994) 11168–11172.

[27] Y. Choo, A. Klug, Toward a code for the interactions of zincfinger with DNA: selection of randomized zinc fingersdisplayed on phage, Proc. Natl. Acad. Sci. U. S. A. 91(1994) 11163–11167.

[28] Y. Choo, I. Sanchez-Garcia, A. Klug, In vivo repression by asite-specific DNA-binding protein designed against anoncogenic sequence, Nature 372 (1994) 642–645.

[29] H.A. Griesman, C.O. Pabo, A general strategy for selectinghigh affinity zinc finger proteins for diverse DNA target sites,Science 275 (1997) 657–661.

[30] J.S. Kim, C.O. Pabo, Getting handhold on DNA: design of poly-zinc finger proteins with femtomolar dissociation constants,Proc. Natl. Acad. Sci. U. S. A. 95 (1998) 2812–2817.

[31] D.J. Segal, B. Drier, R.R. Beerli, C.F. Barbas, Towardscontrolling gene expression at will: selection and design ofzinc finger domains recognizing each of the 5′-GNN-3′ DNAtarget sequences, Proc. Natl. Acad. Sci. U. S. A. 96 (1999)2758–2763.

[32] B. Drier, D.J. Segal, C.F. Barbas, Insights into the molecularrecognition of the 5′-GNN-3′ family of DNA sequences byzinc finger domains, J. Mol. Biol. 303 (2000) 489–502.

[33] B. Drier, R.R. Beerli, D.J. Segal, J.D. Flippin, C.F. Barbas,Development of zinc finger domains for recognition of the 5′-ANN-3′ family of DNA sequences and their use in theconstruction of artificial transcription factors, J. Biol. Chem.276 (2001) 29466–29478.

[34] Q. Liu, Z. Xia, X. Zhong, C.C. Case, Validated zinc fingerprotein designs for all 16 GNN DNA triplet targets, J. Biol.Chem. 277 (2002) 3850–3856.

[35] B. Drier, R.P. Fuller, D.J. Segal, C.V. Lund, et al.,Development of zinc finger domains for recognition of the5′-CNN-3′ family of DNA sequences and their use in theconstruction of artificial transcription factors, J. Biol. Chem.280 (2005) 35588–35597.

[36] M. Islan, Y. Choo, A. Klug, Synergy between adjacent zincfingers in sequence-specific DNA recognition, Proc. Natl.Acad. Sci. U. S. A. 94 (1997) 5617–5621.

[37] S.A. Wolfe, H.A. Griesman, E.I. Ramm, C.O. Pabo, Analysis ofzinc fingers optimized via phage display: evaluating the utility ofa recognition code, J. Mol. Biol. 285 (1999) 1917–1934.

[38] C.O. Pabo, E. Peisach, R.A. Grant, Design and selection ofnovel Cys2His2 zinc finger proteins, Annu. Rev. Biochem.70 (2001) 313–340.

[39] J.C. Miller, C.O. Pabo, Rearrangement of side chains in aZif268 mutant highlights the complexities of zinc finger-DNArecognition, J. Mol. Biol. 313 (2001) 309–315.

[40] M. Moore, A. Klug, Y. Choo, Improved DNA bindingspecificity from polyzinc finger peptides by using strings oftwo finger units, Proc. Natl. Acad. Sci. U. S. A. 98 (2001)1437–1441.

[41] M. Islan, A. Klug, Y. Choo, A rapid, generally applicablemethod to engineer zinc fingers illustrated by targeting theHIV-1 promoter, Nat. Biotechnol. 19 (2001) 656–660.

[42] J.K. Joung, E.I. Ramm, C.O. Pabo, A bacterial two-hybridselection system for studying protein–DNA and protein–protein interactions, Proc. Natl. Acad. Sci. U. S. A. 97 (1996)7382–7387.

[43] J.K. Joung, E.I. Ramm, C.O. Pabo, A bacterial two-hybridselection system for studying protein–DNA and protein–protein interactions, Proc. Natl. Acad. Sci. U. S. A. 97 (2000)7382–7387.

[44] J.A. Hurt, S.A. Thibodeau, A.S. Hirsh, C.O. Pabo, J.K.Joung, Highly specific zinc finger proteins obtained by directdomain shuffling and cell-based selection, Proc. Natl. Acad.Sci. U. S. A. 100 (2003) 12271–12276.

[45] S. Durai, A. Bosley, A.B. Abulencia, S. Chandrasegaran, M.Ostermeier, A bacterial one-hybrid selection system forinterrogating zinc finger-DNA interactions, Comb. Chem.High Throughput Screen. 9 (2006) 301–311.

[46] S. Durai, M. Mani, K. Kandavelou, M. Porteus, S.Chandrasegaran, Zinc finger nucleases: Custom designedmolecular scissors for genome engineering of plant andmammalian cells, Nucleic Acids Res. 33 (2005)5978–5990.

[47] J.J. Smith, J. Berg, S. Chandrasegaran, A detailed study ofsubstrate specificity of a chimeric restriction enzyme,Nucleic Acids Res. 27 (1999) 674–681.


































































































































[48] J. Bitinaite, D.A. Wah, A.K. Aggarwal, I. Schildkraut, FokIdimerization is required for DNA cleavage, Proc. Natl. Acad.Sci. U. S. A. 95 (1998) 10570–10574.

[49] E.S. Vanamee, S. Santagata, A.K. Aggarwal, FokI requirestwo specific DNA sites for cleavage, J. Mol. Biol. 309 (2001)69–78.

[50] J.J. Smith, M. Bibikova, F. Whitby, A.R. Reddy, S.Chandrasegaran, D. Carroll, Requirements for double-strand cleavage by chimeric restriction enzymes with zincfinger DNA-recognition domain, Nucleic Acids Res. 28(2000) 3361–3369.

[51] M. Bibikova, D. Carroll, D.J. Segal, J.K. Trautman, J. Smith,Y.-G. Kim, S. Chandrasegaran, Stimulation of homologousrecombination through targeted cleavage by a chimericnuclease, Mol. Cell. Biol. 21 (2001) 289–297.

[52] M. Mani, J. Smith, K. Kandavelou, J.M. Berg, S.Chandrasegaran, Binding of two zinc finger nucleasemonomers to two specific sites is required for effectivedouble-strand DNA cleavage, Biochem. Biophys. Res.Commun. 334 (2005) 1191–1197.

[53] J.S. Kim, C.O. Pabo, Transcriptional repression by zincfinger pepetides: Exploring the potential for applicationsin gene therapy, J. Bio l . Chem. 272 (1997)29795–29800.

[54] J.S. Kim, J. Kim, K.L. Cepek, P.A. Sharp, C.O. Pabo,Design of TATA box-binding protein/zinc finger fusions fortargeted regulation of gene expression, Proc. Natl. Acad.Sci. U. S. A. 94 (1997) 3616–3620.

[55] R.R. Beerli, D.J. Segal, B. Drier, C.F. Barbas, Towardscontrolling gene expression at will: Specific regulation of theerbB-2/HER-2 promoter by suing polydactyl zinc fingerproteins constructed from modular building blocks, Proc.Natl. Acad. Sci. U. S. A. 95 (1998) 14628–14633.

[56] G.L. Xu, T. Bestor, Cytosine methylation targeted topredetermined sequences, Nat. Genet. 17 (1997) 376–378.

[57] S. Chandrasegaran, J. Smith, Chimeric restriction enzymes:what is next? Biol. Chem. 380 (1999) 841–848.

[58] G. Meister, S. Chandrasegaran, M. Ostermeier, Heterodi-meric DNA methyltransferases as a platform for creatingdesigner zinc finger methyltransferases for targeted DNAmethylation in cells, Nucleic Acids Res. 38 (2010)1749–1759.

[59] M. Mani, K. Kandavelou, J.F. Dy, S. Durai, S.Chandrasegaran, Design, engineering and characteriza-tion of zinc finger nucleases, Biochem. Biophys. Res.Commun. 335 (2005) 447–457.

[60] K. Kandavelou, M. Mani, S. Durai, S. Chandrasegaran,“Magic” scissors for genome surgery, Nat. Biotechnol. 23(2005) 686–687.

[61] J. Wu, K. Kandavelou, S. Chandrasegaran, Custom-designed zinc finger nucleases: what is next? Cell. Mol.Life Sci. 64 (2007) 2933–2944.

[62] K. Kandavelou, S. Ramalingam, V. London, M. Mani, J. Wu, V.Alexeev, C.I. Civin, S. Chandrasegaran, Targetedmanipulationof mammalian cells using designed zinc finger nucleases,Biochem. Biophys. Res. Commun. 388 (2009) 56–61.

[63] S. Ramalingam, V. London, K. Kandavelou, L. Cebotaru, W.Guggino, C.I. Civin, S. Chandrasegaran, Generation andgenetic engineering of human induced pluripotent stemcells using designed zinc finger nucleases, Stem Cells Dev.22 (2013) 595–610.

[64] S. Ramalingam, N. Annaluru, S. Chandrasegaran, ACRISPR way to engineer the human genome, GenomeBiol. 14 (2013) 107–110.

[65] S. Ramalingam, N. Annaluru, K. Kandavelou, S.Chandrasegaran, TALEN-mediated generation and geneticcorrection of disease-specific hiPSCs, Curr. Gene Ther. 14(2014) 461–472.

[66] M. Bibikova, M. Golic, K.G. Golic, D. Carroll, Targetedchromosomal cleavage and mutagenesis in Drosophila usingzinc-finger nucleases, Genetics 161 (2002) 1169–1175.

[67] M. Bibikova, K. Beumer, J.K. Trautman, D. Carroll,Enhancing gene targeting using designed zinc fingernucleases, Science 300 (2003) 764.

[68] A. Lloyd, C.L. Plaisier, D. Carroll, G.N. Drew, Targetedmutagenesis using zinc-finger nucleases in Arabidopsis,Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 2232–2237.

[69] J. Morton, M.W. Davis, E.M. Jorgensen, D. Carroll, Inductionand repair of zinc-finger nuclease-targeted double-strandedbreaks in Caenorhabditis elegans somatic cells, Proc. Natl.Acad. Sci. U. S. A. 103 (2006) 16370–16375.

[70] K. Beumer, G. Battacharyya, M. Bibikova, J.K. Trautman, D.Carroll, Efficient gene targeting in Drosophila with zincfinger nucleases, Genetics 172 (2006) 2391–2403.

[71] Y. Huang, W.A. Paxton, S.M. Wolinsky, A.U. Neumann,et al., The role of mutant CCR5 allele in HIV-1 transmissionand disease progression, Nat. Med. 2 (1996) 1240–1243.

[72] Mala Mani, Engineering and applications of zinc fingernucleases (PhD thesis) Department of EnvironmentalHealth Sciences, Johns Hopkins School of Public Health,2005.

[73] M.H. Porteus, D. Baltimore, Chimeric nucleases stimulategene targeting in human cells, Science 300 (2003) 763.

[74] F.D. Urnov, J.C. Miller, Y.L. Lee, C.M. Beausejour, J.M.Rock, et al., Highly efficient endogenous human genecorrection using designed zinc-finger nucleases, Nature435 (2005) 646–651.

[75] J.C. Miller, M.C. Holmes, J. Wang, D.Y. Guschin, et al., Animproved zinc-finger nuclease architecture for highly specificgenome editing, Nat. Biotechnol. 25 (2007) 778–785.

[76] M. Szczepek, V. Brondani, J. Buchel, L. Serrano, et al.,Structure-based redesign of the dimerization interfacereduces the toxicity of zinc-finger nucleases, Nat. Biotech-nol. 25 (2007) 786–793.

[77] S. Ramalingam, K. Kandavelou, R. Rajenderan, S.Chandrasegaran, Creating designed zinc finger nucleaseswith minimal cytotoxicity, J. Mol. Biol. 405 (2011) 630–641.

[78] C.L. Ramirez, J.E. Foley, D.A.Wright, F. Muller-Lerch, et al.,Unexpected failure rates of modular assembly of engi-neered zinc fingers, Nat. Methods 5 (2008) 374–375.

[79] M.L. Maeder, S. Thibodeau-Beganny, A. Osiak, D.A.Wright, R.M. Anthony, et al., Rapid “open-source” engi-neering of customized zinc-finger nucleases for highlyefficient gene modification, Mol. Cell 31 (2008) 294–301.

[80] J.D. Sander, E.J. Dahlborg, M.J. Goodwin, L. Cade, F.Zhang, et al., Selection-free zinc-finger-nuclease engineer-ing by context-dependent assembly (CoDA), Nat. Methods8 (2011) 67–69.

[81] C.L. Ramirez, M.T. Certo, C. Mussolino, M.J. Goodwin, T.J.Cradick, et al., Engineered zinc finger nickases inducehomology-directed repair with reduced mutagenic effects,Nucleic Acids Res. 40 (2012) 5560–5568.

[82] E. Kim, S. Kim, D.H. Kim, B.S. Choi, I.Y. Choi, J.S. Kim,Precision genome engineering with programmable DNA-nicking enzymes, Genome Res. 22 (2012) 1327–1333.

[83] M.T. Certo, B.Y. Ryu, J.E. Annis, M. Garibov, J. Jarjour,et al., Tracking genome engineering outcome at individualDNA breakpoints, Nat. Methods 8 (2011) 671–676.


































































































































[84] M. Christian, T. Cermark, E.L. Doyle, C. Schmidt, F. Zhang,A. Hummel, A.J. Bogdanove, D.F. Voytas, Targeting DNAdouble-strand breaks with TAL effector nucleases, Genetics186 (2010) 757–761.

[85] J.C. Miller, S. Tan, G. Qiao, K.A. Barlow, et al., A TALEnuclease architecture for efficient genome editing, Nat.Biotechnol. 29 (2011) 143–148.

[86] C. Mussolino, R. Morbitzer, F. Lütge, N. Dannemann, T.Lahaye, T. Cathomen, A novel TALE nuclease scaffoldenables high genome editing activity in combinationwith low toxicity, Nucleic Acids Res. 39 (2011)9283–9293.

[87] M.J. Moscou, A.J. Bogdanove, A simple cipher governsDNA recognition by TAL effectors, Science 326 (2009)1501.

[88] J. Boch, H. Scholze, S. Schornack, A. Landgraf, S. Hahn, S.Kay, T. Lahaye, A. Nickstadt, U. Bonas, Breaking the codeof DNA binding specificity of TAL-type III effectors, Science326 (2009) 1509–1512.

[89] A.N. Mak, P. Bradley, R.A. Cernadas, A.J. Bogdanov, et al.,The crystal structure of TAL effector PthXo1 bound to itsDNA target, Science 335 (2012) 716–719.

[90] D. Deng, C. Yan, X. Pan, M. Mahfouz, et al., Structural basisfor sequence-specific recognition of DNA by TAL effectors,Science 335 (2012) 720–723.

[91] D. Reyon, S.Q. Tsai, C. Khayter, J.A. Foden, J.D. Sander,et al., FLASH assembly of TALENs for high-throughputgenome editing, Nat. Biotechnol. 30 (2012) 460–465.

[92] M. Holkers, I. Maggio, J. Liu, J.M. Janssen, et al.,Differential integrity of TALE nuclease genes followingadenoviral and lentiviral vector gene transfer into humancells, Nucleic Acids Res. 41 (2013) e63.

[93] J.A. Doudna, E. Charpentier, The new frontier of genomeengineering with CRISPR-Cas9, Science 346 (2014)1077–1086.

[94] P.D. Hsu, E.S. Lander, F. Zhang, Development andapplications of CRISPR-Cas9 for genome engineering,Cell 157 (2014) 1262–1278.

[95] R. Barrangou, L.A. Marraffini, CRISPR-Cas systems:Prokaryotes upgrade to adaptive immunity, Mol. Cell 54(2014) 234–244.

[96] J.D. Sander, J.K. Joung, CRISPR-Cas systems for editing,regulating and targeting genomes, Nat. Biotechnol. 32(2014) 347–355.

[97] H. Nishimasu, F.A. Ran, P.D. Hsu, S. Konermann, S.I.Shehata, N. Dohmae, R. Ishitani, F. Zhang, O. Nureki,Crystal structure of Cas9 in complex with guide RNA andtarget DNA, Cell 156 (2014) 935–949.

[98] G. Gasiunas, R. Barrangou, P. Horvath, V. Siksnys, Cas9-crRNA ribonucleoprotein complex mediates specific DNAcleavage for adaptive immunity in bacteria, Proc. Natl.Acad. Sci. U. S. A. 109 (2012) E2579–E2586.

[99] M. Jinek, K. Chylinski, I. Fonfara, M. Hauer, J.A. Doudna, E.Charpentier, A programmable dual-RNA-guided DNA en-donuclease in adaptive bacterial immunity, Science 337(2012) 816–821.

[100] P. Mali, L. Yang, K.M. Esvelt, J. Aach, M. Guell, J.E.DiCarlo, J.E. Norville, G.M. Church, RNA-guided humangenome engineering via Cas9, Science 339 (2013)823–826.

[101] L. Cong, F.A. Ran, D. Cox, S. Lin, R. Barretto, N. Habib,P.D. Hsu, X. Wu, W. Jiang, L.A. Marraffini, F. Zhang,Multiplex genome engineering using CRISPR/Cas systems,Science 339 (2013) 819–823.

[102] S.W. Cho, S. Kim, J.M. Kim, J.S. Kim, Targeted genomeengineering in human cells with Cas9 RNA-guided endo-nuclease, Nat. Biotechnol. 31 (2013) 230–232.

[103] M. Jinek, A. East, A. Cheng, S. Lin, E. Ma, J. Doudna, RNA-programmed genome editing in human cells, eLife (2013),http://dx.doi.org/10.7554/eLife.00471.001.

[104] S.Q. Tsai, N. Wyvekens, C. Khayter, J.A. Foden, V. Thapar,et al., Dimeric CRISPR RNA-guided FokI nucleases forhighly specific genome editing, Nat. Biotechnol. 32 (2014)569–576.

[105] J.P. Guilinger, D.B. Thompson, D.R. Liu, Fusion ofcatalytically inactive Cas9 to FokI nuclease improves thespecificity of genome modification, Nat. Biotechnol. 32(2014) 577–582.

[106] A.V. Wright, S.H. Sternberg, T.W. Taylor, B.T. Staahl, et al.,Rational design of a split-Cas9 enzyme complex, Proc. Natl.Acad. Sci. U. S. A. 112 (2015) 2984–2989.

[107] Y. Fu, J.A. Foden, C. Khayter, M.L. Maeder, D. Reyon,et al., High-frequency off-target mutagenesis induced byCRISPR-Cas nucleases in human cells, Nat. Biotechnol. 31(2013) 822–826.

[108] P.D. Hsu, D.A. Scott, J.A. Weinstein, F.A. Ran, S.Konermann, et al., DNA targeting specificity of RNA-guidedCas9 nucleases, Nat. Biotechnol. 31 (2013) 827–832.

[109] P. Mali, J. Aach, P. Benjamin Stranges, K.M. Esvelt, M.Moosburner, et al., Cas9 transcriptional activators for targetspecificity screening and paired nickases for co-operativegenome engineering, Nat. Biotechnol. 31 (2013) 833–838.

[110] V. Pattanayak, S. Lin, J.P. Guilinger, E. Ma, J.A. Doudna,D.R. Liu, High-throughput profiling of off-target cleavagereveals RNA-programmed Cas9 specificity, Nat. Biotech-nol. 31 (2013) 839–843.

[111] Y. Fu, J.D. Sander, D. Reyon, V.M. Cascio, J.K. Joung,Improving CRISPR-Cas9 specificity using truncated guideRNAs, Nat. Biotechnol. 32 (2014) 279–284.

[112] B. Farbound, B.J. Meyer, Dramatic enhancement ofgenome editing by CRIPR-Cas9 through improved guideRNA design, Genetics 199 (2015) 959–971.

[113] E.A. Moehle, J.M. Rock, Y.L. Lee, Y. Jouvenot, et al.,Targeted gene addition into a specified location in thehuman genome using designed zinc finger nucleases, Proc.Natl. Acad. Sci. U. S. A. 104 (2007) 3055–3060.

[114] E.E. Perez, J. Wang, J.C. Miller, Y. Jouvenot, K.A. Kim,et al., Establishment of HIV-1 resistance in CD4+ T cells bygenome editing using zinc finger nucleases, Nat. Biotech-nol. 26 (2008) 808–816.

[115] D. Hockemeyer, F. Soldner, C. Beard, Q. Gao, et al.,Efficient targeting of expressed and silent genes in humanESCs and iPSCs using zinc-finger nucleases, Nat. Bio-technol. 27 (2009) 851–857.

[116] D. Hockemeyer, H. Wang, S. Kiani, C.S. Lai, et al., Geneticengineering of human pluripotent cells using TALE nucle-ases, Nat. Biotechnol. 29 (2011) 731–734.

[117] K. Yusa, S.T. Rashid, H. Strick-Marchand, I. Varela, P.Q.Liu, D.E. Paschon, et al., Targeted gene correction ofalpha1-antitrypsin deficiency in induced pluripotent stemcells, Nature 478 (2011) 391–394.

[118] Y. Yao, B. Nashun, T. Zhou, L. Qin, et al., Generation ofCD34(+) cells from CCR5-disrupted human embryonic andinduced pluripotent stem cells, Hum. Gene Ther. 23 (2011)238–242.

[119] N. Ma, B. Liao, H. Zhang, L. Wang, Y. Shan, Y. Xue, et al.,Transcription activator-like effector nuclease (TALEN)-mediated gene correction in integration-free β-thalassemia




































































http://dx.doi.org/10.7554/eLife.00471.001





























































induced pluripotent stem cells, J. Biol. Chem. 288 (2013)34671–34679.

[120] P.K. Mandal, L.M.R. Ferreira, R. Collins, T.B. Meissner, C.L.Boutwell, et al., Efficient ablation of genes in humanhematopoietic stem and effector cells using CRISPR/Cas9, Cell Stem Cell 15 (2014) 643–652.

[121] M.D. Hoban, G.J. Gregory, M.C. Mendel, Z. Romero, M.L.Kaufman, et al., Correction of sickle-cell disease mutation inhaematopoietic stem/progenitor cells, Blood 125 (2014)2597–2604.

[122] P. Genovese, G. Schiroli, G. Escobar, T. Di Tomaso, C.Firrito, A. Calabria, et al., Targeted genome editing inhuman repopulating haematopoietic stem cells, Nature 510(2014) 235–240.

[123] P. Tebas, D. Stein, W.W. Tang, I. Frank, S.Q. Wang, et al.,Gene editing of CCR5 in autologous CD4 T cells of personsinfected with HIV, N. Engl. J. Med. 370 (2014) 901–910.

[124] M.D. Hoban, G.J. Cost, M.C. Mendel, Z. Romero, et al.,Correction of the sickle cell disease mutation in humanhematopoietic stem/progenitor cells, Blood 125 (2015)2597–2604.

[125] D. Carroll, D.J. Segal, Induced recombination in modelsystems, Curr. Res. Mol. Ther. 1 (1998) 593–598.

[126] A.F. Faruqi, M.M. Seidman, D.J. Segal, D. Carroll, P.M.Glazer, Recombination induced by triple helix-targeted DNAdamage in mammalian cells, Mol. Cell. Biol. 16 (1996)6820–6828.

[127] D.J. Segal, A.F. Faruqi, P.M. Glazer, D. Carroll, Processingof targeted psoralen crosslinks in Xenopus oocytes, Mol.Cell. Biol. 17 (1997) 6645–6652.

[128] J.Y. Chin, P.M. Glazer, Repair of DNA lesions associatedwith triplex-forming oligonucleotides, Mol. Carcinog. 48(2009) 389–399.

[129] N. Rudin, E. Sugarman, J.E. Haber, Genetic and physicalanalysis of double-strand break repair and recombination inSaccharomyces cerevisiae, Genetics 122 (1989) 519–534.

[130] A. Plessis, A. Perrin, J.E. Haber, B. Dujon, Site-specificrecombination determined by I-SceI, a mitochondrial group Iintron-encoded endonuclease expressed in the yeastnucleus, Genetics 130 (1992) 451–460.

[131] A. Choulika, A. Perrin, B. Dujon, J.-F. Nicolas, Induction ofhomologous recombination in mammalian chromosomes byusing the I-SceI system of Saccharomyces cerevisiae, Mol.Cell. Biol. 15 (1995) 1968–1973.

[132] Y.S. Rong, K.G. Golic, Gene targeting by homologousrecombination in Drosophila, Science 288 (2000)2013–2018.

[133] K.J. Beumer, J.K. Trautman, A. Bozas, J.L. Liu, J. Rutter,et al., Efficient gene targeting inDrosophila by direct embryoinjection with zinc-finger nucleases, Proc. Natl. Acad. Sci.U. S. A. 105 (2008) 19821–19826.

[134] D.A. Wright, J.A. Townsend, R.J. Winfrey Jr., P.A. Irwin, J.Rajagopal, et al., High-frequency homologous recombina-tion in plants mediated by zinc-finger nucleases, Plant J. 44(2005) 693–705.

[135] A.J. Wood, T.W. Lo, B. Zeitler, C.S. Pickle, E.J. Ralston,et al., Targeted genome editing across species using ZFNsand TALENs, Science 333 (2011) 307.

[136] A.E. Friedland, Y.B. Tzur, K.M. Esvelt, M.P. Colaiacovo,G.M. Church, J.A. Calarco, Heritable genome editing in C.elegans via a CRISPR-Cas9 system, Nat. Methods 10(2013) 741–743.

[137] T.W. Lo, C.S. Pickle, S. Lin, E.J. Ralston, M. Gurling, et al.,Heritable genome editing using TALENs and CRISPR/Cas9

to engineer precise insertions and deletions in evolutionarilydiverse nematode species, Genetics 195 (2013) 331–348.

[138] S.W. Cho, J. Lee, D. Carroll, J.S. Kim, J. Lee, Heritablegene knockout in C. elegans by direct injection of Cas9-sgRNA ribonucleoproteins, Genetics 195 (2013)1177–1180.

[139] S. Kim, D. Kim, S.W. Cho, J. Kim, J.S. Kim, Highly efficientRNA-guided genome editing in human cells via delivery ofpurified Cas9 ribonucleoproteins, Genome Res. 24 (2014)1012–1019.

[140] S. Lin, B.T. Staahl, R.K. Alla, J.A. Doudna, Enhancedhomology-directed human genome editing by controlledtiming of CRISPR/Cas9 delivery, eLife 3 (2014) e04766.

[141] A.M. Geurts, G.J. Cost, Y. Freyvert, B. Zeitler, J.C. Miller,et al., Knockout rats via embryo microinjection of zinc-fingernucleases, Science 325 (2009) 433.

[142] T. Mashimo, Gene targeting technologies in rats: zinc fingernucleases, transcription activator-like effector nucleasesand clustered regularly interspaced short palindromicrepeats, Develop. Growth Differ. 56 (2014) 46–52.

[143] I.D. Carbery, D. Ji, A. Harrington, V. Brown, E.J. Weinstein,et al., Targeted genome modification in mice using zinc-finger nucleases, Genetics 186 (2010) 451–459.

[144] M. Meyer, M. Hrabe de Angelis, W. Wurst, R. Kuhn, Genetargeting by homologous recombination in mouse zygotesmediated by zinc-finger nucleases, Proc. Natl. Acad. Sci.U. S. A. 107 (2010) 15022–15026.

[145] X. Cui, D. Ji, D.A. Fisher, Y. Wu, D.M. Briner, E.J. Weinstein,Targeted integration in rat and mouse embryos with zinc-fingernucleases, Nat. Biotechnol. 29 (2011) 64–67.

[146] X. Meng, M.B. Noyes, L.J. Zhu, L.D. Lawson, S.A. Wolfe,Targeted gene inactivation in zebrafish using engineeredzinc-finger nucleases, Nat. Biotechnol. 26 (2008) 695–701.

[147] M. Doyon, J.M. McCammon, J.C. Miller, F. Faraji, et al.,Heritable targeted gene disruption in zebrafish usingdesigned zinc-finger nucleases, Nat. Biotechnol. 26(2008) 702–708.

[148] D.J. Segal, J.F. Meckler, Genome engineering at the dawnof the golden age, Annu. Rev. Genomics Hum. Genet. 14(2013) 135–158.

[149] D. Carroll, Genome engineering with targetable nucleases,Annu. Rev. Biochem. 83 (2014) 409–439.

[150] Y. Peng, K.J. Clark, J.M. Campbell, M. Panetta, et al.,Making designer mutants in model organisms, Develop-ment 142 (2014) 4042–4054.

[151] A. Juillerat, G. Dubois, J. Valton, S. Thomas, S. Stella, et al.,Comprehensive analysis of the specificity of transcriptionactivator-like effector nucleases, Nucleic Acids Res. 42(2014) 5390–5402.

[152] K.J. Beumer, J.K. Trautman, M. Christian, T.J. Dahlem,C.M. Lake, et al., Comparing ZFNs and TALENs for genetargeting in Drosophila, G3 3 (2013) 1717–1725.

[153] K.S. Makarova, N.V. Grishin, S.A. Shabalina, Y.I. Wolf, E.V.Koonin, A putative RNA-interference-based immune sys-tem in prokaryotes: computational analysis of the predictedenzymatic machinery, functional analogies with eukaryoticRNAi, and hypothetical mechanisms of action, Biol. Direct 1(2006) 7.

[154] B. Wiedenheft, S.H. Sternberg, J.A. Doudna, RNA-guidedgenetic silencing systems in bacteria and archaea, Nature482 (2012) 331–338.

[155] W.Y. Hwang, Y. Fu, D. Reyon, M.L. Maeder, S.Q. Tsai,et al., Efficient genome editing in zebrafish using a CRISPR-Cas system, Nat. Biotechnol. 31 (2013) 227–229.


































































































































[156] H. Wang, H. Yang, C.S. Shivalila, M.M. Dawlaty, A.W.Cheng, et al., One-step generation of mice carryingmutations in multiple genes by CRISPR/Cas-mediatedgenome engineering, Cell 153 (2013) 910–918.

[157] A.R. Bassett, C. Tibbit, C.P. Ponting, J.-L. Liu, Highlyefficient targeted mutagenesis of Drosophila with theCRISPR/Cas9 system, Cell Rep. 4 (2013) 220–228.

[158] S.J. Gratz, A.M. Cummings, J.N. Nguyen, D.C. Hamm, L.K.Donohue, et al., Genome engineering ofDrosophilawith theCRISPR RNA-guided Cas9 nuclease, Genetics 194 (2013)1029–1035.

[159] D. Carroll, Staying on target with CRISPR-Cas, Nat.Biotechnol. 31 (2013) 807–809.

[160] Y. Doyon, T.D. Vo, M.C. Mendel, S.G. Greenberg, J. Wang,et al., Enhancing zinc-finger-nuclease activity with improvedobligate heterodimer architectures, Nat. Methods 8 (2011)74–79.

[161] S.Q. Tsai, Z. Zheng, N.T. Nguyen, M. Liebers, V.V. Topkar,et al., GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases, Nat. Biotech-nol. 33 (2015) 187–197.

[162] D. Kim, S. Bae, J. Park, E. Kim, S. Kim, et al., Digenome-seq: genome-wide profiling of CRISPR-Cas9 off-targeteffects in human cells, Nat. Methods 12 (2015) 237–243.

[163] F.A. Ran, P.D. Hsu, C.Y. Lin, J.S. Gootenberg, S.Konermann, et al., Double nicking by RNA-guided CRISPRCas9 for enhanced genome editing specificity, Cell 154(2013) 1380–1389.

[164] T. Koo, J. Lee, J.S. Kim, Measuring and reducing off-targetactivities of programmable nucleases including CRISPR-Cas9, Mol. Cell 38 (2015) 475–481.

[165] A. Hendel, E.J. Fine, G. Bao, M.H. Porteus, Quantifying on-and off-target genome editing, Trends Biotechnol. 33 (2015)132–140.

[166] R. Gabriel, A. Lombardo, A. Arens, J.C. Miller, et al., Anunbiased genome-wide analysis of zinc finger nucleasespecificity, Nat. Biotechnol. 29 (2011) 429–435.

[167] X. Wang, Y. Wang, X. Wu, J. Wang, Y. Wang, et al.,Unbiased detection of off-target cleavage by CRISPR-Cas9and TALENs using integrase-defective lentiviral vectors,Nat. Biotechnol. 33 (2015) 175–178.

[168] R.L. Frock, J. Hu, R.M. Meyers, Y.J. Ho, E. Kii, F.W. Alt,Genome-wide detection of DNA double-strand breaksinduced by engineered nucleases, Nat. Biotechnol. 33(2015) 179–186.

[169] N. Crosetto, A. Mitra, M.J. Silva, M. Bienko, N. Dojer, et al.,Nucleotide-resolution DNA double-strand break mapping bynext-generation sequencing, Nat. Methods 10 (2013)361–365.

[170] F.A. Ran, L. Cong, W.X. Yan, D.A. Scott, J.S. Gootenberg,et al., In vivo genome editing using Staphylococcus aureusCas9, Nature 520 (2015) 186–191.

[171] A. Bozas, K.J. Beumer, J.K. Trautman, D. Carroll, Geneticanalysis of zinc-finger nuclease-induced gene targeting inDrosophila, Genetics 182 (2009) 641–651.

[172] K.J. Beumer, J.K. Trautman, K. Mukherjee, D. Carroll,Donor DNA utilization during gene targeting with zinc-fingernucleases, G3 3 (2013) 657–664.

[173] L. Deriano, D.B. Roth, Modernizing the nonhomologousend-joining repertoire: alternative and classical NHEJ sharethe stage, Annu. Rev. Genet. 47 (2013) 433–455.

[174] P. Singh, J.C. Schimenti, E. Bolcun-Filas, A mousegeneticist's practical guide to CRISPR applications, Genet-ics 199 (2015) 1–15.

[175] T. Maruyama, S.K. Dougan, M.C. Truttmann, A.M. Bilate,J.R. Ingram, H.L. Ploegh, Increasing the efficiency ofprecise genome editing with CRISPR-Cas9 by inhibition ofnonhomologous end joining, Nat. Biotechnol. 33 (2015)538–542.

[176] V.T. Chu, T. Weber, B. Wefers, W. Wurst, S. Sander, et al.,Increasing the efficiency of homology-directed repair forCRISPR-Cas9-induced precise gene editing in mammaliancells, Nat. Biotechnol. 33 (2015) 543–548.

[177] S. Lin, B.T. Staahl, R.K. Alla, J.A. Doudna, Enhancedhomology-directed human genome engineering by con-trolled timing of CRISPR/Cas9 delivery, eLife 3 (2014)e04766.

[178] M. Ghorbal, M. Gorman, C.R. Macpherson, R.M. Martins, A.Scherf, J.J. Lopez-Rubio, Genome editing in the humanmalaria parasitePlasmodium falciparum using the CRISPR-Cas9 system, Nat. Biotechnol. 32 (2014) 819–821.

[179] S.M. Sidik, C.G. Hackett, F. Tran, N.J. Westwood, S.Lourido, Efficient genome engineering of Toxoplasmagondii using CRISPR/Cas9, PLoS One 9 (2014) e100450.

[180] M. DeGennaro, C.S. McBride, L. Seeholzer, T. Nakagawa,E.J. Dennis, et al., orco mutant mosquitoes lose strongpreference for humans and are not repelled by volatileDEET, Nature 498 (2013) 487–491.

[181] A. Aryan, M.A. Anderson, K.M. Myles, Z.N. Adelman,TALEN-based gene disruption in the dengue vectorAedes aegypti, PLoS One 8 (2013) e60082.

[182] V.K. Shukla, Y. Doyon, J.C. Miller, R.C. DeKelver, E.A.Moehle, et al., Precise genome modification in the cropspecies Zea mays using zinc-finger nucleases, Nature 459(2009) 437–441.

[183] W. Haun, A. Coffman, B.M. Clasen, Z.L. Demorest, A. Lowy,et al., Improved soybean oil quality by targeted mutagenesisof the fatty acid desaturase 2 gene family, Plant Biotechnol.J. 12 (2014) 934–940.

[184] D.F. Carlson, W. Tan, S.G. Lillico, D. Steverakov, C.Proudfoot, et al., Efficient TALEN-mediated gene knockoutin livestock, Proc. Natl. Acad. Sci. U. S. A. 109 (2012)17382–17387.

[185] W. Tan, D.F. Carlson, C.A. Lancto, J.R. Garbe, D.A.Webster, et al., Efficient nonmeiotic allele introgression inlivestock using custom endonucleases, Proc. Natl. Acad.Sci. U. S. A. 110 (2013) 16526–16531.

[186] Y. Niu, B. Shen, Y. Cui, J. Wang, et al., Generation of gene-modified Cynomolgus monkeys via Cas9/RNA-mediatedgene targeting in one-cell embryos, Cell 156 (2014)836–843.

[187] H. Liu, Y. Chen, Y. Niu, K. Zhang, Y. Kang, et al., TALEN-mediated gene mutagenesis in Rhesus and Cynomolgusmonkeys, Cell Stem Cell 14 (2014) 323–328.

[188] W.S. Tan, D.F. Carlson, M.W. Walton, S.C. Fahrenkrug,P.B. Hackett, Precision editing of large animal genomes,Adv. Genet. 80 (2012) 37–97.

[189] L. Li, L. Krymskaya, J. Wang, J. Henley, A. Rao, et al.,Genomic editing of the HIV-1 coreceptor CCR5 in adulthematopoietic stem and progenitor cells using zinc fingernucleases, Mol. Ther. 21 (2013) 1259–1269.

[190] Q. Ding, Y.K. Lee, E.A. Schaefer, D.T. Peters, A. Veres,et al., A TALEN genome-editing system for generatinghuman stem cell-based disease models, Cell Stem Cell 12(2013) 238–251.

[191] Y. Kim, J. Kweon, A. Kim, J.K. Chon, J.Y. Yoo, et al., Alibrary of TAL effector nucleases spanning the humangenome, Nat. Biotechnol. 31 (2013) 251–258.


































































































































[192] C. Smith, A. Gore, W. Yan, L. Abalde-Atristain, et al., Whole-genome sequencing analysis reveals high specificity forCRISPR/Cas9 and TALEN-based genome editing in humaniPSCs, Cell Stem Cell 15 (2014) 12–13.

[193] A. Veres, B.S. Gosis, Q. Ding, A. Ragavendran, H. Brand,et al., Low incidence of off-target mutations in individualCRISPR-Cas9 and TALEN targeted human stem cellclones detected by whole-genome sequencing, Cell StemCell 15 (2014) 27–30.

[194] B. Zetsche, J. Gootenberg, O.O. Abudeyyeh, I.M. Slaymaker,et al., Cpf1 is a single RNA-guided endonuclease of a class 2CRISPR-Cas system, Cell 163 (2015) 1–13.

[195] P. Liang, Y. Xu, X. Zhang, C. Ding, R. Huang, et al.,CRISPR\Cas9-mediated gene editing in human tripronuc-lear zygotes, Protein Cell 6 (2015) 363–372.

[196] E. Lanphier, F. Urnov, S.E. Haeker, M. Werner, J.Smolenski, Don't edit the human germline, Nature 519(2015) 410–411.

[197] D. Baltimore, P. Berg, M. Botchan, et al., A prudent pathforward for genomic engineering and germline genomemodification, Science 348 (2015) 36–38.

[198] E.S. Lander, Brave new genome, N. Engl. J. Med. 373(2015) 5–8.
























origins of programmable nucleases for genome engineering · 2017-02-07 · origins of programmable...

Documents