Corrections

PHYSICS, BIOPHYSICS AND COMPUTATIONAL BIOLOGYCorrection for “In silico investigation of intracranial blast miti-gation with relevance to military traumatic brain injury,” byMichelle K. Nyein, Amanda M. Jason, Li Yu, Claudio M. Pita,John D. Joannopoulos, David F. Moore, and Raul A. Rado-vitzky, which appeared in issue 48, November 30, 2010, of ProcNatl Acad Sci USA (107:20703–20708; first published November22, 2010; 10.1073/pnas.1014786107).The authors note that reference 19 appeared incorrectly. The

updated reference appears below. This error does not affect theconclusions of the article.

19. Moss WC, King MJ, Blackman EG (2009) Skull flexure from blast waves: A mechanismfor brain injury with implications for helmet design. Phys Rev Lett 103:108702.

www.pnas.org/cgi/doi/10.1073/pnas.1018365108

NEUROSCIENCECorrection for “A unifying model for timing of walking onsetin humans and other mammals,” by Martin Garwicz, MariaChristensson, and Elia Psouni, which appeared in issue 51, De-cember 22, 2009, of Proc Natl Acad Sci USA (106:21889–21893;first published December 14, 2009; 10.1073/pnas.0905777106).The authors note the following statement should be added

to the Acknowledgments: “We also acknowledge the support ofThe Knut and Alice Wallenberg Foundation (Project KAW2004.0119) and the Medical Faculty at Lund University.”

www.pnas.org/cgi/doi/10.1073/pnas.1018282108

PLANT BIOLOGYCorrection for “Site-directed mutagenesis in Arabidopsis usingcustom-designed zinc finger nucleases,” by Keishi Osakabe,Yuriko Osakabe, and Seiichi Toki, which appeared in issue 26,June 29, 2010 of Proc Natl Acad Sci USA (107:12034–12039; firstpublished May 27, 2010; 10.1073/pnas.1000234107).The authors wish to note that they inadvertently copied text in

their introductory paragraph and in the first three sentences oftheir second paragraph from reference 15 [Lloyd A, Plaisier CL,Carroll D, Drews GN (2005) Targeted mutagenesis using zinc-finger nucleases in Arabidopsis. Proc Natl Acad Sci USA 102:2232–2237] without proper attribution. The authors apologizefor the oversight.

www.pnas.org/cgi/doi/10.1073/pnas.1017337108

www.pnas.org PNAS | January 4, 2011 | vol. 108 | no. 1 | 433

CORR

ECTIONS

Site-directed mutagenesis in Arabidopsis usingcustom-designed zinc finger nucleasesKeishi Osakabea, Yuriko Osakabeb, and Seiichi Tokia,c,1

aPlant Genetic Engineering Research Unit, Division of Plant Sciences, National Institute of Agrobiological Sciences, Tsukuba 305-8602, Japan; bGraduate Schoolof Agricultural and Life Sciences, The University of Tokyo, Tokyo 113-8657, Japan; and cKihara Institute of Biological Research, Yokohama City University,Yokohama 244-0813, Japan

Edited by Joseph R. Ecker, Salk Institute, La Jolla, CA, and approved May 3, 2010 (received for review January 9, 2010)

Site-directed mutagenesis in higher plants remains a significanttechnical challenge for basic research andmolecular breeding. Here,we demonstrate targeted-gene inactivation for an endogenousgene in Arabidopsis using zinc finger nucleases (ZFNs). EngineeredZFNs for a stress-response regulator, the ABA-INSENSITIVE4 (ABI4)gene, cleaved their recognition sequences specifically in vitro, andZFN genes driven by a heat-shock promoter were introduced intothe Arabidopsis genome. After heat-shock induction, gene muta-tions with deletion and substitution in the ABI4 gene generatedvia ZFN-mediated cleavage were observed in somatic cells at fre-quencies as high as 3%. The homozygote mutant line zfn_abi4-1–1for ABI4 exhibited the expected mutant phenotypes, i.e., ABA andglucose insensitivity. In addition, ZFN-mediated mutagenesis wasapplied to the DNA repair-deficientmutant plant, atku80. We foundthat lack of AtKu80, which plays a role in end-protection of dsDNAbreaks, increased error-prone rejoining frequency by 2.6-fold, withincreased end-degradation. These data demonstrate that an ap-proach using ZFNs can be used for the efficient production of mu-tant plants for precision reverse genetics.

dsDNA breaks | Ku80 | mutation | nonhomologous end joining

Amajor focus of plant biotechnology is geneticmodificationandimprovement of crop plants.With this aim inmind, large-scale

genome analyses have been performed for the model plant Ara-bidopsis, and also formany important crop plants species, includingrice, maize, wheat, soybean, and tomato. The enormous amount ofgenome-sequence information now available has intensified theneed for methods that can use this data to generate targetedmodifications in plant genes. Site-directed mutagenesis methodscould be used experimentally to investigate plant gene function orfor genetic modification in plant cells. This is especially importantfor species that lack readily available mutant collections, includingmost crops.The most widely used site-directed mutagenesis strategy is

gene targeting (GT) via homologous recombination (HR). Effi-cient GT procedures have been available for more than 20 y inyeast (1) and mouse (2). Successful GT has also been achieved inArabidopsis and rice plants (3–6). Typically, GT events occur ina fairly small proportion of treated mammalian cells (approxi-mately 1% of the total random integration events in mouse EScells). However, GT efficiency is extremely low in higher plantcells [0.01–0.1% of the total number of random integration events(7)]. The lowGT frequencies reported in higher plants are thoughtto result from competition between HR and nonhomologous endjoining (NHEJ) for repair of dsDNA breaks (DSBs), whereasthe main pathway of DSB repair in higher plants seems to beNHEJ (8, 9). As a consequence, the ends of a donor molecule arelikely to be joined by NHEJ rather than participating in HR, thusreducing GT frequency. There is extensive data indicating thatDSBs repair byNHEJ in higher plants is error-prone.Often, DSBsare repaired by end-joining processes that generate insertionsand/or deletions (10, 11). Taken together, these observationssuggest that NHEJ-based strategies might be more effective thanHR-based strategies for targeted mutagenesis in higher plants.

Indeed, expression of I-Sce I, a rare cutting restriction enzyme, hasbeen shown to introduce mutations at I-Sce I cleavage sites inArabidopsis and tobacco (12). Nevertheless, the use of restrictionenzymes is limited to rarely occurring natural recognition sites orto artificial target sites. To overcome this problem, zinc fingernucleases (ZFNs) have been developed. ZFNs are chimeric pro-teins composed of a synthetic zinc finger–based DNA bindingdomain and a DNA cleavage domain. By modification of the zincfinger DNA binding domain, ZFNs can be specifically designed tocleave virtually any long stretch of dsDNA sequence (13, 14). AnNHEJ-based targeted mutagenesis strategy was developed re-cently in several organisms by using synthetic ZFNs to generateDSBs at specific genomic sites (15–19). Subsequent repair of theDSBs by NHEJ frequently produces deletions and/or insertions atthe joining site.To our knowledge, two groups have successfully applied ZFNs

to genetically modify genes in zebrafish embryos by using specificzinc finger motifs engineered to recognize distinct DNA se-quences (16, 17). The ZFN-encoding mRNA was injected intoone-cell embryos and a high percentage of animals carried thedesired mutations and phenotypes. These latter studies demon-strated that ZFNs can specifically and efficiently create heritablemutant alleles at loci of interest in the germ line, and that ZFN-induced alleles can be propagated in subsequent generations.Although precise genetic modification using ZFNs has been

successfully applied to higher plants (15, 20, 21), to our knowledge,only a study of Lloyd et al. (15) presented a detailed analysis of theNHEJ-based targeted-mutagenesis strategy with a model systemusing a synthetic target site for a previously reported three-finger–type ZFN_QQR (22) in the Arabidopsis genome. Thus, the nextstep in the establishment of this approach is to target endogenousgene loci in the genome of higher plants. In addition, further in-vestigation into the precise conditions and effectors required forapplication of ZFNs in plants would also be invaluable.In this report, we show that ZFNs can efficiently cleave and

stimulate mutations at an endogenous target gene in Arabidopsis.For this demonstration, we selected the ABA-INSENSITIVE4(ABI4) gene as a target gene. ABI4 encodes a member of theERF/AP2 transcription factor family and plays a role in regu-lating abscisic acid (ABA) (23), which controls a number of ag-ronomically important traits, including plant responses to abioticstress and seed development (24). We achieved targeted muta-genesis at a rate of approximately 0.26% to 2.86% in Arabidopsissomatic cells, and transmission of the induced mutation in the

Author contributions: K.O. and S.T. designed research; K.O. and Y.O. performed research;K.O. and Y.O. analyzed data; and K.O., Y.O., and S.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

See Commentary on page 11657.1To whom correspondence should be addressed. E-mail: stoki@affrc.go.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1000234107/-/DCSupplemental.

12034–12039 | PNAS | June 29, 2010 | vol. 107 | no. 26 www.pnas.org/cgi/doi/10.1073/pnas.1000234107

target gene to subsequent generations. The mutant line showedthe expected mutant phenotypes. In addition, we applied ZFN-mediated targeted mutagenesis to the NHEJ-deficient Arabi-dopsismutant atku80. We found that deficiency of Ku80 enhanceserror-prone rejoining of ZFN-cleavage ends, with increased end-degradation. These results indicate that ZFNs can form thebasis of a highly efficient method for site-directed mutagenesis ofhigher plant genes.

ResultsCustom Designed ZFNs Targeting the ABI4 Gene. To demonstrateZFN-mediated site-directed mutagenesis in Arabidopsis usingthe ABI4 gene as a target gene, we first identified full consensusZFN target sites [5′-NNCNNCNNC(Nx4∼7)GNNGNNGNN-3′(N = A, C, G, and T)] in ABI4 (Fig. 1). A combination of thethree ZF arrays for 5′-GGAGGAGGA-3′ (ZF_AAA) and 5′-G-TGGCGGCG-3′ (ZF_TCC) targeting ABI4 was designed usingthe zincfingermodules for 5′-GNN-3′ triplets reported byLiu et al.(25) and Segal et al. (26) (SI Materials and Methods, Fig. S1, andTable S1). ZF arrays assembled in the pGP-FB vector were testedin quantitative bacterial two-hybrid (B2H) assays. According toLacZ reporter assay results, all assembled ZF arrays showed highaffinity for their target, with transcription activated by more thanfourfold (Fig. S2). Thus, these modular assembled ZF proteinswere used in the construction ofZFNs.We next performed in vitroDNA digestion assay with ZFN proteins synthesized using an invitro translation systemwithwheat germextracts (Fig. S3A andB).We confirmed that the combination ofZFN_AAAandZFN_TCCdigested the target sequence of the ABI4 gene (Fig. S3 C and D).

Engineered ZFNs Stimulate Mutations at Target Sequences inArabidopsis Cells. To determine whether induction of ZFN ac-tivity could digest the Arabidopsis genome in vivo and inducemutations at the recognition sequence in Arabidopsis cells, weintroduced the ZFN expression vector pP1.2gfbPhsZFN_ABI4(Fig. 2A) into the Arabidopsis genome via Agrobacterium-mediated transformation. In this study, the Arabidopsis heatshock protein HSP18.2 gene promoter (28) was used to drive theexpression of ZFNs. This expression cassette allows inducibleexpression of ZFNs transiently upon heat shock, and thus hasthe benefit of avoiding cell toxicity (29). We transformed Ara-bidopsis plants (ecotype Col-0) with pP1.2gfbPhsZFN-ABI4 andselected for transgenic lines. We refer to lines with introducedpP1.2gfbPhsZFN-ABI4 as zfn_abi4-1, -2, -3, -5, -7, -9, -11, -12,and -13. To induce ZFN expression, 12-d-old plants from eachtransgenic line were subjected to a heat pulse at 40 °C for 90min; genomic DNA was isolated from true leaves of these plants48 h after heat shock. We also isolated genomic DNA from trueleaves of these lines before heat induction as a control. To de-

termine whether induction of ZFN activity could induce muta-tions at its recognition sequence, we developed a method ofidentifying mutations coupled with a mismatch-specific endonu-clease: Surveyor nuclease. PCR using primers flanking the ZFNrecognition sequence was performed on genomic DNA extractedfrom cells treated with ZFNs. PCR products were cloned intopCR-TOPO vector, and plasmid clones possessing ZFN-medi-ated mutations were identified using the Surveyor nuclease assaydescribed in Materials and Methods. The DNA sequences ofclones showing a positive signal in the Surveyor nuclease assaywas determined to verify that these clones contained mutationswithin the ZFN target sequence. The mutation frequency in so-matic cells was calculated as the number of clones containingmutations per total number of randomly picked clones. Assummarized in Table 1, in heat-treated seedlings, 0.26% to 2.86%of clones contained ZFN-induced mutations in the ZFN targetsequence. In contrast, control seedlings, in which ZFN expressionwas not induced by heat pulse, gave rise to no clones containinga mutation. Of 47 clones sequenced, 33 (70%) contained sub-stitution mutations, and 14 (30%) were simple deletions of 1 to 3bp in the target sequence of zfn_abi4 lines (Fig. 2B and Table S2).These data indicate that ZFNs can efficiently stimulate mutationsat their recognition sequences in Arabidopsis cells.

Functional Analysis of Mutant Plants for the ABI4 Locus Produced byZFN-Mediated Mutagenesis. We further determined transmission

Fig. 1. Consensus ZFN target sites in the Arabidopsis ABA-INSENSITIVE4(ABI4) gene. The schematic representation of the ABI4 gene is shown at thetop. Asterisk indicates the position of the mutation in the abi4 mutant. AP,AP2 domain; S/T, serine- and threonine-rich domain; Q, glutamine-rich do-main; Acid, activation domain. Target sites of ZFN monomers are highlightedwith gray bars. The putative cleavage sites are shown by arrows. FokI-DD,mutated Fok I nuclease domain (R487D); FokI-RR, mutated Fok I nucleasedomain (D483R).

C

B

A

Fig. 2. Alteration of amino acid sequences after ZFN digestion in the ABI4gene. (A) Structure of T-DNA for the ZFN expression binary vector. LB, leftborder sequence of T-DNA; RB, right border sequence of T-DNA; gfbsd2, GFP +blasticidin S-resistance gene fusion expression cassette; Phsp18.2, ArabidopsisHSP18.2 gene promoter; AAA, ZFN_AAA; 2A, self-cleaving 2A peptide derivedfrom Thosea asigna (27); TCC, ZFN_TCC; Tp5, Arabidopsis polyA-binding pro-tein PAB5 gene terminator. The gfbsd2 expression cassette is driven by 2×cauliflower mosaic virus 35S gene promoter and the nopaline synthase geneterminator. (B) Examples of repaired DNA sequences at ZFN target sites (bold)of the ABI4 genes in transgenic lines zfn_abi4-1 and zfn_abi4-9. The putativecleavage region is shown in lowercase letters. Mutations are indicated inmagenta-colored characters and deletions are shown as hyphens. (C) Alter-ation of amino acid sequences after ZFNs digestion of theABI4 gene. Two lines(zfn_abi4-1–1 and zfn_abi4-9-1) were recovered from initially detected mu-tated T1 plants for the ABI4 gene. After establishing the T3 generation, mu-tation sites were reconfirmed. Target sites of ZFN monomers are underlined.The putative site of the double-strand break is double-underlined. Predictedamino acid sequences of this region of the ABI4 gene in the two mutant linesare shown at bottom.

Osakabe et al. PNAS | June 29, 2010 | vol. 107 | no. 26 | 12035

PLANTBIOLO

GY

SEECO

MMEN

TARY

of mutations in zfn_abi4 lines. If mutations are induced in theL2 cells of the shoot apical meristem, these cells possessingmutations form a sector in the primary inflorescence that incor-porates germ line cells. To detect such mutation sectors, genomicDNA extracted from several flower clusters, including the ter-minal flower, of each zfn_abi4 line was analyzed for the presenceof mutations at the cleavage site in ABI4 with the Surveyor nu-clease assay.We detected mutations in two of nine lines, zfn_abi4-1 and zfn_abi4-9 (Fig. S4). We further collected progeny seeds(approximately 300 seeds) from the primary inflorescence of thetwo lines, and scored progeny seedlings for the presence of ZFN-induced mutations within ABI4. Approximately 100 seeds wereplated and germinated to collect true leaves from individualplants for the extraction of genomic DNA. By using the Surveyornuclease assay, progeny seedlings with mutations at the cleavagesite were identified: seven plants from 96 of randomly selectedplants in zfn_abi4-1 (Fig. S4). Sequence analysis revealed that allseven progeny seedlings showed a single base deletion at the sameposition, which produced a stop codon after Thr158 of ABI4 (Fig.2C). It was thought that they are heterozygous mutants for theABI4 locus, according to the Surveyor nuclease assay, becauseheteroduplex DNA of the ABI4 fragment was detected with thisassay by using genomic DNA from only a single mutant plant.These results suggest that a single cell harboring the mutationcreated by ZFN-cleavage produced a sector, and the sector pro-duced germ line cells, which exist as chimeras in reproductivetissue. In zfn_abi4-9, progeny seedlings containing mutations atthe cleavage site were also identified: three plants from 96 ofrandomly selected plants (Fig. S4). All three were substitutionmutations at the same position, which caused an amino acidchange at Val159 (V159 to Ile, Fig. 2C) and correspond tomutants heterozygous for theABI4 locus as seen in line zfn_abi4-1according to sequencing and Surveyor nuclease analyses. Forsubsequent phenotypic analysis, we selected a single plant eachfrom zfn_abi4-1 and zfn_abi4-9 progeny plants, named zfn_abi4-1–1 and zfn_abi4-9–1, respectively. T3 progenies of zfn_abi4-1–1and zfn_abi4-9–1 were produced to select plants homozygous forthese mutations. T3 seeds from T2 plants of these two lines seg-regated with a 1:2:1 ratio of homozygous mutation to heterozy-gous mutation to WT, following Mendelian rules (homo:hetero:WT, 9:19:12, n= 40, χ2 = 0.55, P= 0.760 in zfn_abi4-1–1; homo:hetero:WT, 9:23:8, n = 40, χ2 = 0.95, P = 0.622 in zfn_abi4-9-1).

ABI4 is an ERF/AP2 transcription factor and plays a role inABA and sugar signaling during seed development ofArabidopsis.Previous studies have shown that abi4 mutants are insensitive toABA and high concentrations of sugars during seedling de-velopment (30–32). To investigate the functionality of zfn_abi4mutations in Arabidopsis, we next asked whether zfn-abi4 mutantsshow a phenotype similar to that of the abi4mutant. zfn_abi4-1–1and zfn_abi4-9–1 mutant plants exhibited no morphologicalalterations under normal growth conditions; however, the germi-nation rates of the zfn_abi4-1–1 plants and abi4, but not those ofzfn-abi4-9–1 plants, were higher than those ofWTplants onABA-containing medium (Fig. 3 A and C). The zfn_abi4-1–1 mutantplants were also able to grow on a high concentration of glucose(Fig. 3B and Fig. S5A), suggesting that this mutant showed ABAand glucose insensitivity. Germination of zfn-abi4-1–1 was alsotolerant to NaCl and mannitol as seen in abi4 (Fig. S5 B and C).

Lack of Ku80 Reduces the Fidelity but not the Efficiency of NHEJ. Asdescribed earlier, after ZFN-ABI4 cleavage, the majority of ZFN-mediated mutations resulted in substitution-type mutations with-out large deletion or insertions (Table 1, Table S2, and Fig. 2B).The Arabidopsis protein AtKu80 plays a role in end-protection ofDSBs. We hypothesized that a lack of Ku protein activity wouldresult in drastic modifications such as large deletion and/or in-sertion at DSB ends. To test this hypothesis, we explored thetype and frequency of mutations in the NHEJ-deficient Arabi-dopsis mutant plant, atku80 (ecotype WS background, ref. 33).Heterozygous AtKu80/atku80 plants were transformed withpP1.2gfbPhsZFN-ABI4, and three T1 transgenic lines of hetero-zygous AtKu80/atku80 plants were recovered by PCR genotyping.Segregated WT and homozygous atku80/atku80 plants were re-covered from three lines (lines 80–1, 80–2, and 80–3). These linesshowed a 3:1 ratio of segregation for blasticidin-SR to blasticidin-SS when T2 seeds were plated on selection medium. Afterobtaining T3 seeds, we recovered plant lines segregating for WT(AtKu80/AtKu80) and homozygous for blasticidin-SR, and lineshomozygous for atku80/atku80 andblasticidin-SR from80–1, 80–2,and 80–3, respectively. These lines were used to test for ZFN-mediated mutagenesis. Fig. 4 shows a comparison of the mutationfrequency and spectrum in the segregatingWT and atku80 plants.As seen in Fig. 4A, themutation frequency of ecotypeWs is similarto that seen earlier with ecotype Col-0 (Table 1). Furthermore, wefound that the mutation frequency in atku80 was comparable tothat in WT (Fig. 4A). It is interesting to note that the fraction of

Table 1. Frequency of ZFN-induced mutations in the ABI4 gene

Treatment/line

Frequency of mutations,% (no. of clones

tested)

Deletion mutations/totalnumber of mutations

(percentage of deletionmutations*)

Heat inducedzfn-abi4-1 2.86 (384) 4/11 (36.3%)zfn-abi4-2 1.30 (384) 1/5 (20.0%)zfn-abi4-3 1.82 (384) 2/7 (28.6%)zfn-abi4-5 0.78 (384) 1/3 (33.3%)zfn-abi4-7 0.52 (384) 0/2 (0%)zfn-abi4-9 2.34 (384) 2/9 (22.2%)zfn-abi4-11 1.82 (384) 3/7 (42.9%)zfn-abi4-12 0.52 (384) 1/2 (50%)zfn-abi4-13 0.26 (384) 0/1 (0%)

Not inducedzfn-abi4-1 0 (768) –

zfn-abi4-9 0 (768) –

zfn-abi4-11 0 (768) –

*The percentage of deletion mutations was calculated as follows: percent-age of deletion mutations = no. of deletion mutations/total no. of muta-tions × 100.

C

BA

Fig. 3. The ZFN targeted mutation in ABI4 in Arabidopsis confers ABA-insensitive and stress tolerance phenotypes. (A and B) abi4, zfn-abi4, and WTplants were grown for 10 d on medium containing ABA (A) and glucose (B).(C) Seed germination rates of abi4, zfn_abi4, and WT plants at various con-centrations of ABA. Germination was assessed at d 2 after sowing. SD valueswere calculated from three individual experiments. n = 30 seeds per experi-ment. WT (■), abi4 (□), zfn_abi4-1–1 (○), zfn_abi4-9–1 (▲).

12036 | www.pnas.org/cgi/doi/10.1073/pnas.1000234107 Osakabe et al.

error-prone end-joining pathway(s) repair events increased byapproximately 2.6 times in atku80 cells (Fig. 4 B and C and TableS3). In particular, deletions more than 4 bp in length increased inatku80 drastically in as many as 70% of the mutants (Fig. 4B).Thus, we conclude that AtKu80 plays a role in end-protection ofenzymatically induced DSBs, although other end-joining pathway(s) repair DSB ends efficiently in Arabidopsis cells.

DiscussionGene knockout/inactivation is the most powerful tool for deter-mining gene function or to permanently modify phenotype for mo-lecular breeding in higher plants. Currently available methods forgene disruption in higher plants are limited by their efficiency, timeto completion, and the potential for confounding off-target effects.This study demonstrates the feasibility of aZFN-based approach

to site-directed mutagenesis in an endogenous gene ofArabidopsisusingABI4 (23) as a target gene. Themutation frequency achievedat theZFN target sites in theABI4 gene was sufficient to allow easyidentification of mutants following heat-induced ZFN expression(Table 1). We further confirmed transmission of mutations to thenext generation. Two transgenic lines (zfn_abi4-1 and zfn_abi4-9)froma total of nine lines gave rise to transmittedmutations (Fig. 2Cand Fig. S4), and mutations in these lines were further transmittedto their progeny following Mendelian rules. The physiologicalphenotypes of mutant lines produced via ZNF-mediated cleavagein terms of stress and ABA responses were confirmed as beingsimilar to those of mutants previously obtained by using the othermethod. zfn_abi4-1–1 plants clearly showed insensitivity to ABAand glucose (Fig. 3 and Fig. S5).

The mutation frequency achieved here clearly could enable anefficient site-directed mutagenesis procedure in higher plants, incontrast to conventional GT procedures without the help ofZFNs, which have much lower mutation frequencies; typicallyfewer than 10−7 GT events per cell or 10−6 to 10−4 GT events perintegration event (15). The dramatically higher mutation fre-quencies with the ZFN-mediated procedure can be comparedwith those obtained by HR-mediated procedures, supporting theview that the error-prone end-joining pathway for DSBs repair ishighly active in Arabidopsis somatic cells. The frequency of mu-tagenesis identified here (Table 1) is comparable to, albeit slightlylower than, that identified for similar experiments with ZFNsreported by others. Heat-inducible expression of ZFN-QQR (22)in Arabidopsis produced approximately 7% mutated progenyseedlings with an artificial target site (15). Maeder et al. (34), byusing induced ZFN SR2163, found a mutation frequency of 5%at the tobacco SuRA locus. These differences might be a result ofdifferences in protein stability and/or activity of ZFNs. In addi-tion, accessibility of ZFNs to chromatinized DNA might dependon ZFN design and each gene locus. These, and perhaps otheryet-uncharacterized, features could affect mutation frequency.We found many substitution mutations at the edge of the pu-

tative DNA cleavage sequence (Fig. 2B and Tables S2 and S3).One possible mechanism that might explain this observation isthe single-base insertion coupled with single-base deletion causedby the MRE11/RAD50 complex. This complex has exonucleaseactivity, but can delete a few bases, and functions to process DSBends to make blunt ends (8). It is interesting to note that themutation profile obtained in this study is slightly different fromthat found in previous studies with I-Sce I (12) and ZFNQQR(15) for target cleavage sites at the synthetic and exogenous in-tegrated gene loci of Arabidopsis. These studies showed that themajority of mutations were highly varied, with longer deletionsand/or insertions of 1 to 20 bp, and sometimes much longer (12,15). One possible explanation for these differences is that bindingactivity to chromatinized target DNA of ZFNs could differ ineach study, thus triggering different types of repair pathways orfactors. Thus, longer end-degradation at the cleavage sites mayhave occurred in other studies compared with ours. We thought itpossible that there might be higher activity of end-protection byDNA-PK/Ku at the cleavage sites under our conditions in a targetgene locus–dependent manner and depending on the cleavageactivity of the specific ZFNs used.We compared ZFN-mediated mutations in WT and Ku80-

deficient mutant plants. As shown in Fig. 4 B and C, our dataclearly indicate that deficiency of AtKu80 increases deletion sizeat the cleavage site ofABI4 as expected. We also found that a highlevel of NHEJ mutagenesis was maintained in the absence ofKu80, suggesting that a secondary, inaccurate repair pathway(s)function in Arabidopsis cells (Fig. 4A). Schulte-Uentrop et al. (35)reported similar results with Ku80-deficient mouse cells; efficientend-joining activity was observed in both WT and Ku80-deficientmouse cells after introducing DSBs enzymatically. However,deletions at the break site in Ku80-deficient mouse cells weresignificantly longer than in WT mouse cells.A recent study suggested that the DNA-PK–dependent path-

way involving Ku80 competes with the DNA repair pathway in-volving PARP-1/XRCC1/DNA ligase III (LigIII) in human cells(36). Counterpart genes for PARP1 and XRCC1 genes have beenidentified in the Arabidopsis genome, but a counterpart gene forthe LIGIII gene is lacking. Recently, Waterworth et al. (37)reported that Arabidopsis DNA ligase I plays an important rolein DSB repair as well as in the repair pathway of single-strandedDNA breaks. Thus, it is highly possible that ligase I participatesin the alternative end-joining repair pathway instead of LigIII,and that this pathway acts as a backup pathway for NHEJ inKu80-deficient Arabidopsis cells. In addition, the microhomology-mediated end joining pathway involving PARP-1/XRCC1/LigIII

BA

C

Fig. 4. Efficiency andfidelity of NHEJ repaired DNA after ZFN cleavage in theABI4 gene in atku80 cells. (A) The frequency of end-joining after induction ofDSBs by ZFNs. The relative error-prone end-joining frequency was derivedfrom the number of DNA clones positive for the Surveyor nuclease assay pernumber of DNA clones tested (n = 480 for each experiment). Experimentswere performed with three independent lines, and data are presented asmean ± SD. (B) Distribution of length of deletions at individual junctions.Deletions are defined as the sum of base pairs lost at both sides of the DSB.(C) Examples of repaired DNA sequences obtained from genomic DNA of WTand atku80 plants after ZFN cleavage. ZFN recognition sites are depicted inbold. Putative cleavage regions are shown in lowercase characters. Mutationsare shown inmagenta-colored characters and deletion are shown as hyphens.

Osakabe et al. PNAS | June 29, 2010 | vol. 107 | no. 26 | 12037

PLANTBIOLO

GY

SEECO

MMEN

TARY

(38) might also function as an error-free repair pathway. ZFNsproduce 5′-cohesive ends with perfectly 4b-matched sequences,and these fragments could be employed for microhomology-mediated end joining without any modification following ZFNscleavage. Furthermore, in the case of Ku80-deficient cells, exo-nuclease 1 could function efficiently compared with WT cells (39).Thus, the 3′-single-strand tail produced by exonuclease 1 mightalso lead to the HR repair pathway for restriction enzyme-mediated cleavage, although this pathway is restricted to the S-to-G2 phase of the cell cycle (40).In conclusion, we have established a method for ZFN-mediated

site-directed mutagenesis in Arabidopsis. By using this system, wecompared profiles of mutations between WT and Ku80-deficientmutant cells. We found that a deficiency in Ku80 increased the fre-quency of mutations showing longer deletions of bases at the repairsites. Thedata indicate the existence of an alternativeNHEJpathwayinArabidopsis, and that this pathway is highly active under conditionsof deficiency of the Ku80-dependent NHEJ pathway. Our data fur-ther suggest that couplingofZFNsandKu80-deficient cell lines couldbe a powerful tool with which to create many type of mutations, in-cluding substitutions, and short and long deletions of bases, at thegene of interest. In addition, this method can be further applied tocrop plants such as rice, maize, wheat, soybean, and tomato, as se-quence information and transformation systemsare available in thesespecies. The ZFN technology can be easily applied to establish mu-tation lines of genes of interest for molecular breeding.

Materials and MethodsPlant Material and Growth. Arabidopsis thaliana ecotypes Columbia (Col-0)and Wassilewskija (Ws) were used in this study. Plants were grown on a soilmixture of equal parts of vermiculite and commercial soil (Super Mix; SakataSeed) in a growth chamber at 22 °C. Seeds were germinated and incubatedon plates containing half-strength Murashige and Skoog (MS) medium so-lidified with 0.25% Gelrite (MS Gelrite plate; Wako) in a growth chamber at22 °C. Seeds of the abi4 mutant were obtained from the Arabidopsis Bi-ological Resource Center. The T-DNA insertion mutant line for AtKU80 waskindly provided by C. M. Bray (University of Manchester, Manchester, UK;ref. 33). The PCR genotyping of atku80 plants was performed according toWest et al. (33).

Construction of ZFNs. ZFNswereconstructedusingthemodularassemblymethodpreviously reportedbyWright etal. (41)with slightmodifications (fordetails, seeSI Materials andMethods and Fig. S1). Amino acid sequences of helix motif in ZFare used as follows: QRAHLER forfinger 1, QSGHLQR forfinger 2, andQRAHLERfor finger 3 in ZF_AAA; and RSDALTR for finger 1, RSDDLQR for finger2,RSDDLQR for finger 3 in ZF_TCC. The coding sequences of ZF proteins werecloned between the Xba I and Bam HI sites of the pGB-FB vector (41).

Quantitative B2H Assay. B2H assays were performed according to the protocolof Wright et al. (41). Plasmid DNA and bacterial strains for B2H were obtainedfrom Addgene.

In Vitro DNA Digestion Assay. ZFN proteins were synthesized using in vitrotranslation system with wheat germ extracts. Details of protein synthesis andpurificationandinvitrodigestionassayareprovidedinSIMaterialsandMethods.

Expression of ZFNs in Arabidopsis Cells. The plasmid pP1.2gfPhsZFN_ABI4 wastransformed into Agrobacterium tumefaciens strain GV3101 by electroporationand introduced into Arabidopsis by the floral dipping method (42). Transgenicplants were selected on medium containing 1.5 μg/mL blasticidin S, and by theappearance of GFP fluorescence (43). Seedlings were grown on plates at 22 °Cfor 12 d. The plates thenwerewrapped in plastic wrap and immersed inwater at40 °C for90min. Seedlingsweregrown foranadditional 48hat 22 °CbeforeDNAextraction. T1 transformants following heat pulse were transplanted to soil andallowed to self-pollinate to obtain the next generation. T2 seeds from zfn_abi4lines were germinated on selective medium.

Analysis of Genome Editing at ZFN Target Sites. To analyze mutations ZFN-induced mutations, the 685-bp region surrounding the ZFN_AAA/ZFN_TCCpair site forABI4was amplified by PCRwith the high fidelity DNA polymeraseKOD-Plus (Toyobo). PCR products were analyzed with the Surveyor MutationDetection Kit (Transgenomic). The detail for PCR and the nuclease assay con-ditions were described in SI Materials and Methods. To determine sequencesof ZFN-induced mutations, the initial PCR products were cloned into the pCR-TOPO vector (Invitrogen) and Escherichia coli colonies containing ABI4 genefragments were randomly selected and replicated. Twenty-four colonies werebatched and used for colony PCR. PCR products were analyzed with the Sur-veyor nuclease assay. Typical results are shown in Fig. S6. The batches con-taining the mutations according to the Surveyor nuclease assay werereanalyzedwith each single colony. Finally, the plasmid from the single colonyshowing a mutation signal according to the Surveyor nuclease assay was iso-lated and its sequence determined. To determine transmission of ZFN-inducedmutations, progeny seeds from lines zfn_abi4-1 and -9 were plated on MSmedium. Genomic DNA was extracted individually from true leaves of 96randomly picked plants. Twelve DNA samples were batched and used for theSurveyor nuclease assay. Batches containing the mutations according to theSurveyor nuclease assay were reanalyzed with each single DNA sample.

Seed Germination and Root Growth Assays. Germination assays were per-formed as previously described (44, 45). Seeds were surface-sterilized, sus-pended in sterile 0.1% agar, and placed on MS plates containing 1% sucrose,0–0.5 μMABA, 6% glucose, 150mMNaCl, or 300mMmannitol and 0.8% agarand grown in a 16/8 h light/dark cycle at 22 °C. Germination (i.e., radicalemergence) was scored at various times. Root growth assays were performedas follows. Seeds were surface-sterilized as described earlier and germinatedon MS plates containing 1% sucrose and 0.8% agar at 22 °C. After 5 d, seed-lings were transferred to plates containing various concentrations of NaCl,turned 90°, and incubated for additional 5 d. Root length was measured atdifferent times after seed plating.

ACKNOWLEDGMENTS.We thank the Arabidopsis Biological Resource Centerfor providing the abi4 mutant seeds. We thank Drs. M. Kimura (RIKEN WakoInstitute) for providing the gfbsd2 gene and C. M. Bray (University of Man-chester) for providing the T-DNA insertion line for AtKU80. We also thank K.Amagai, R. Aoto, C. Furusawa, E. Ozawa, A. Nagashii, and F. Suzuki for theirtechnical help. This work was supported by a Program for Promotion of BasicResearch Activities for Innovative Biosciences grant from the Bio-OrientedTechnology Research Advancement Institution of Japan (to S.T.).

1. Rothstein R (1991) Targeting, disruption, replacement, and allele rescue: IntegrativeDNA transformation in yeast. Methods Enzymol 194:281–301.

2. Capecchi MR (1989) Altering the genome by homologous recombination. Science 244:1288–1292.

3. Hanin M, et al. (2001) Gene targeting in Arabidopsis. Plant J 28:671–677.4. Terada R, Urawa H, Inagaki Y, Tsugane K, Iida S (2002) Efficient gene targeting by

homologous recombination in rice. Nat Biotechnol 20:1030–1034.5. Endo M, Osakabe K, Ichikawa H, Toki S (2006) Molecular characterization of true and

ectopic gene targeting events at the acetolactate synthase gene in Arabidopsis. PlantCell Physiol 47:372–379.

6. Endo M, et al. (2007) Molecular breeding of a novel herbicide-tolerant rice by genetargeting. Plant J 52:157–166.

7. Terada R, Johzuka-Hisatomi Y, Saitoh M, Asao H, Iida S (2007) Gene targeting byhomologous recombination as a biotechnological tool for rice functional genomics.Plant Physiol 144:846–856.

8. Ray A, Langer M (2002) Homologous recombination: Ends as the means. Trends Plant Sci 7:435–440.

9. Britt AB, May GD (2003) Re-engineering plant gene targeting. Trends Plant Sci 8:90–95.10. Gorbunova VV, V, Levy AA (1999) How plants make ends meet: DNA double-strand

break repair. Trends Plant Sci 4:263–269.

11. Britt AB (1999) Molecular genetics of DNA repair in higher plants. Trends Plant Sci 4:20–25.

12. Kirik A, Salomon S, Puchta H (2000) Species-specific double-strand break repair andgenome evolution in plants. EMBO J 19:5562–5566.

13. Kim YG, Cha J, Chandrasegaran S (1996) Hybrid restriction enzymes: Zinc fingerfusions to Fok I cleavage domain. Proc Natl Acad Sci USA 93:1156–1160.

14. Cathomen T, Joung JK (2008) Zinc-finger nucleases: The next generation emerges.Mol Ther 16:1200–1207.

15. Lloyd A, Plaisier CL, Carroll D, Drews GN (2005) Targeted mutagenesis using zinc-finger nucleases in Arabidopsis. Proc Natl Acad Sci USA 102:2232–2237.

16. Doyon Y, et al. (2008) Heritable targeted gene disruption in zebrafish using designedzinc-finger nucleases. Nat Biotechnol 26:702–708.

17. Meng X, Noyes MB, Zhu LJ, Lawson ND, Wolfe SA (2008) Targeted gene inactivationin zebrafish using engineered zinc-finger nucleases. Nat Biotechnol 26:695–701.

18. Beumer KJ, et al. (2008) Efficient gene targeting in Drosophila by direct embryoinjection with zinc-finger nucleases. Proc Natl Acad Sci USA 105:19821–19826.

19. Santiago Y, et al. (2008) Targeted gene knockout in mammalian cells by usingengineered zinc-finger nucleases. Proc Natl Acad Sci USA 105:5809–5814.

20. Shukla VK, et al. (2009) Precise genome modification in the crop species Zea maysusing zinc-finger nucleases. Nature 459:437–441.

12038 | www.pnas.org/cgi/doi/10.1073/pnas.1000234107 Osakabe et al.

21. Townsend JA, et al. (2009) High-frequency modification of plant genes usingengineered zinc-finger nucleases. Nature 459:442–445.

22. Smith J, et al. (2000) Requirements for double-strand cleavage by chimeric restrictionenzymes with zinc finger DNA-recognition domains. Nucleic Acids Res 28:3361–3369.

23. Finkelstein RR, Wang ML, Lynch TJ, Rao S, Goodman HM (1998) The Arabidopsisabscisic acid response locus ABI4 encodes an APETALA 2 domain protein. Plant Cell 10:1043–1054.

24. Hirayama T, Shinozaki K (2007) Perception and transduction of abscisic acid signals:Keys to the function of the versatile plant hormone ABA. Trends Plant Sci 12:343–351.

25. Liu Q, Xia Z, Zhong X, Case CC (2002) Validated zinc finger protein designs for all 16GNN DNA triplet targets. J Biol Chem 277:3850–3856.

26. Segal DJ, Dreier B, Beerli RR, Barbas CF, 3rd (1999) Toward controlling geneexpression at will: Selection and design of zinc finger domains recognizing each ofthe 5′-GNN-3′ DNA target sequences. Proc Natl Acad Sci USA 96:2758–2763.

27. Donnelly ML, et al. (2001) The ‘cleavage’ activities of foot-and-mouth disease virus 2Asite-directed mutants and naturally occurring ‘2A-like’ sequences. J Gen Virol 82:1027–1041.

28. Takahashi T, Naito S, Komeda Y (1992) The Arabidopsis HSP18.2 promoter/GUS genefusion in transgenic Arabidopsis plants: A powerful tool for the isolation ofregulatory mutants of the heat-shock response. Plant J 2:751–761.

29. Porteus MH (2006) Mammalian gene targeting with designed zinc finger nucleases.Mol Ther 13:438–446.

30. Finkelstein RR, Gibson SI (2002) ABA and sugar interactions regulating development:Cross-talk or voices in a crowd? Curr Opin Plant Biol 5:26–32.

31. Gazzarrini S, McCourt P (2001) Genetic interactions between ABA, ethylene and sugarsignaling pathways. Curr Opin Plant Biol 4:387–391.

32. León P, Sheen J (2003) Sugar and hormone connections. Trends Plant Sci 8:110–116.33. West CE, et al. (2002) Disruption of the Arabidopsis AtKu80 gene demonstrates an

essential role for AtKu80 protein in efficient repair of DNA double-strand breaks invivo. Plant J 31:517–528.

34. Maeder ML, et al. (2008) Rapid “open-source” engineering of customized zinc-fingernucleases for highly efficient gene modification. Mol Cell 31:294–301.

35. Schulte-Uentrop L, El-Awady RA, Schliecker L, Willers H, Dahm-Daphi J (2008) Distinctroles of XRCC4 and Ku80 in non-homologous end-joining of endonuclease- andionizing radiation-induced DNA double-strand breaks. Nucleic Acids Res 36:2561–2569.

36. Wang M, et al. (2006) PARP-1 and Ku compete for repair of DNA double strand breaksby distinct NHEJ pathways. Nucleic Acids Res 34:6170–6182.

37. Waterworth WM, et al. (2009) DNA ligase 1 deficient plants display severe growthdefects and delayed repair of both DNA single and double strand breaks. BMC PlantBiol 9:79.

38. McVey M, Lee SE (2008) MMEJ repair of double-strand breaks (director’s cut): Deletedsequences and alternative endings. Trends Genet 24:529–538.

39. Tomita K, et al. (2003) Competition between the Rad50 complex and the Kuheterodimer reveals a role for Exo1 in processing double-strand breaks but nottelomeres. Mol Cell Biol 23:5186–5197.

40. Kikuchi K, et al. (2009) Bloom DNA helicase facilitates homologous recombinationbetween diverged homologous sequences. J Biol Chem 284:26360–26367.

41. Wright DA, et al. (2006) Standardized reagents and protocols for engineering zincfinger nucleases by modular assembly. Nat Protoc 1:1637–1652.

42. Clough SJ, Bent AF (1998) Floral dip: A simplified method for Agrobacterium-mediated transformation of Arabidopsis thaliana. Plant J 16:735–743.

43. Ochiai-Fukuda T, et al. (2006) A fluorescent antibiotic resistance marker for rapidproduction of transgenic rice plants. J Biotechnol 122:521–527.

44. Osakabe Y, et al. (2005) Leucine-rich repeat receptor-like kinase1 is a key membrane-bound regulator of abscisic acid early signaling in Arabidopsis. Plant Cell 17:1105–1119.

45. Arenas-Huertero F, Arroyo A, Zhou L, Sheen J, León P (2000) Analysis of Arabidopsisglucose insensitive mutants, gin5 and gin6, reveals a central role of the planthormone ABA in the regulation of plant vegetative development by sugar. Genes Dev14:2085–2096.

Osakabe et al. PNAS | June 29, 2010 | vol. 107 | no. 26 | 12039

PLANTBIOLO

GY

SEECO

MMEN

TARY