Targeted gene addition into a specified location in the human genome using designed zinc finger nucleases Erica A. Moehle, Jeremy M. Rock, Ya-Li Lee, Yann Jouvenot, Russell C. DeKelver, Philip D. Gregory, Fyodor D. Urnov*, and Michael C. Holmes Sangamo BioSciences, Inc., Point Richmond Technology Center, 501 Canal Boulevard, Suite A100, Richmond, CA 94804 Communicated by Carl O. Pabo, Harvard Medical School, Boston, MA, December 27, 2006 (received for review November 14, 2006) Efficient incorporation of novel DNA sequences into a specific site in the genome of living human cells remains a challenge despite its potential utility to genetic medicine, biotechnology, and basic research. We find that a precisely placed double-strand break induced by engineered zinc finger nucleases (ZFNs) can stimulate integration of long DNA stretches into a predetermined genomic location, resulting in high-efficiency site-specific gene addition. Using an extrachromosomal DNA donor carrying a 12-bp tag, a 900-bp ORF, or a 1.5-kb promoter-transcription unit flanked by locus-specific homology arms, we find targeted integration fre- quencies of 15%, 6%, and 5%, respectively, within 72 h of treat- ment, and with no selection for the desired event. Importantly, we find that the integration event occurs in a homology-directed manner and leads to the accurate reconstruction of the donor- specified genotype at the endogenous chromosomal locus, and hence presumably results from synthesis-dependent strand an- nealing repair of the break using the donor DNA as a template. This site-specific gene addition occurs with no measurable increase in the rate of random integration. Remarkably, we also find that ZFNs can drive the addition of an 8-kb sequence carrying three distinct promoter-transcription units into an endogenous locus at a fre- quency of 6%, also in the absence of any selection. These data reveal the surprising versatility of the specialized polymerase machinery involved in double-strand break repair, illuminate a powerful approach to mammalian cell engineering, and open the possibility of ZFN-driven gene addition therapy for human genetic disease. gene therapy protein production somatic cell genetics Tmotif2H2eukarya he C zinc finger (1), the most abundant DNA recognition in (2, 3), is highly amenable to engineering for the recognition of virtually any DNA sequence (4���6). These properties have been successfully exploited to enable the mod- ulation of gene expression via their application as designed transcription factors (ZFP-TFs) (7), as well as direct modifica- tion of the DNA itself via engineered zinc finger nucleases (ZFNs) for human gene correction (8). The latter process, based on work from several laboratories including our own (9���16), overcomes the exceedingly low frequency of spontaneous ho- mologous recombination in mammalian cells, which until re- cently has made the targeted modification of human genome sequence in vivo impractical (17, 18). Although this limitation has been addressed in settings where drug-based selection schemes can be applied (19, 20), it is restricted to particular cell types, e.g., fibroblasts and mouse embryonic stem cells. Such traditional ������gene targeting������ requires the construction of elabo- rate vectors, a 6- to 8-week regimen of treatment with two distinct selective agents, and the isolation of individual cell clones by limiting dilution, only a subset of which carries the desired targeting event (18). ZFN-mediated gene correction (8), in contrast, occurs at high frequency without selection, is applicable to a broad range of primary and transformed cells, and does not require cell cloning because it invokes a natural process of genetic information transfer via a double-strand break (DSB). A DSB evoked by a stalled DNA replication fork or by an environmental insult is normally eliminated via end-joining (21) or homology-directed repair (HDR). The latter is a specialized form of homologous recombination that transfers genetic information to the broken chromosome from a DNA molecule of related sequence (22���25). Indeed, we have earlier shown that targeting a DSB to a specific site in the genome with engineered ZFNs (Fig. 1A) transfers single-base-pair changes from a donor plasmid into the chro- mosome with efficiencies that can exceed 20% (16). However rapid and efficient, gene correction is a localized event, and a single DSB, whether induced by a homing endo- nuclease (26) or by a ZFN (M.C.H., Y.-L.L., and F.D.U., unpublished data), can allow efficient correction of mutations only within an 200-bp window surrounding the break. The complex mutational spectrum underlying many human mono- genic diseases would therefore require tailoring ZFNs to each cluster of mutations. This requirement has prompted us to investigate the feasibility of using ZFNs to drive site-specific ������gene addition,������ specifically, the integration of long DNA segments into a predetermined locus. Both medical (gene ther- apy) and industrial (e.g., engineering cell lines for protein production) gene addition is currently achieved via random integration of the transgene into the genome, a process that presents safety concerns from a clinical perspective (27) and is costly and time-consuming in industrial applications because of chromatin-based effects on expression of a randomly integrated transgene (28). A considerable effort notwithstanding (29, 30), only limited progress has been made so far in controlling the location of gene insertion, and extensive screening or selection for the desired event is almost invariably a prerequisite. The present work shows that efficient, site-specific gene addition into a predetermined endogenous locus in human cells can occur in the absence of selection. We show that if a ZFN-cleaved locus is provided with an engineered template that consists of novel genetic information flanked by appropriate regions of target site homology, then break repair occurs via Author contributions: E.A.M. and J.M.R. contributed equally to this work E.A.M., J.M.R., P.D.G., F.D.U., and M.C.H. designed research E.A.M., J.M.R., Y.-L.L., Y.J., R.C.D., and F.D.U. performed research E.A.M., J.M.R., F.D.U., and M.C.H. analyzed data and P.D.G., F.D.U., and M.C.H. wrote the paper. Conflict of interest statement: C.O.P. is chair of the Scientific Advisory Board for Sangamo BioSciences, Inc. E.A.M., J.M.R., Y.-L.L., Y.J., R.C.D., P.D.G., F.D.U., and M.C.H. are full-time employees of Sangamo BioSciences, Inc. Freely available online through the PNAS open access option. Abbreviations: ZFN, zinc finger nuclease DSB, double-strand break HDR, homology- directed repair SDSA, synthesis-dependent strand annealing. *To whom correspondence should be addressed. E-mail: furnov@sangamo.com. This article contains supporting information online at www.pnas.org/cgi/content/full/ 0611478104/DC1. �� 2007 by The National Academy of Sciences of the USA www.pnas.org cgi doi 10.1073 pnas.0611478104 PNAS February 27, 2007 vol. 104 no. 9 3055���3060 APPLIED BIOLOGICAL SCIENCES

HDR in a manner most consistent with the synthesis-dependent strand annealing (SDSA) model of DSB repair (31, 32). Our data illuminate the surprising versatility of the currently unidentified DNA polymerase machinery involved in HDR: we observe accurate and highly efficient transfer of up to 8,000 bp of genetic information from an episomal donor to an endogenous locus in human cells in the absence of selection. Results Efficient ZFN-Induced Tag Integration into a Native Chromosomal Locus. In earlier work (16) we showed that a ZFN-induced DSB can transfer a single-base-pair change into an endogenous locus from a plasmid donor in human cells at a frequency of 20% in the absence of selection. Such ������gene correction������ invokes a process formally known as ������short-patch gene conversion������ (24). Studies in budding yeast (33) and Drosophila, however, have indicated that considerably larger segments of genetic informa- tion can be transferred in a homology-dependent way from one DNA molecule to another in the wake of a DSB. For example, the best-studied such system, mating type switching in Saccha- romyces cerevisiae, relocates 700 bp in this manner (34). As originally proposed by Gloor and colleagues (31) based on their analysis of P-element excision in Drosophila, providing a novel DNA stretch within a donor molecule, such that it becomes invisible to the HDR machinery during the homology search, could allow a break-driven integration event to occur via a process termed SDSA (see Discussion for a detailed explanation) (32, 33). Evidence from budding yeast (35) and Drosophila (31, 36) supports the notion that SDSA is one of the two pathways of homology-based DNA break repair in mitotic cells (an alterna- tive process, known as the ������double Holliday junction,������ appears to operate in some settings) (32). From a somatic cell genetics perspective, SDSA-based reso- lution of a nuclease-induced break could, in principle, allow the transfer of extended stretches of genetic information to endog- enous chromosomal locations from a plasmid donor. In our studies of ZFN-driven gene correction, however, we found that relatively minor sequence differences between the chromosomal locus being targeted and the homology arms contained on the donor molecule significantly lower genome editing frequency. This ������donor���target������ divergence penalty was imposed even when the mismatches were only a few base pairs from the position of the DSB (Y.-L.L., F.D.U., and M.C.H., unpublished data), suggesting a mechanistic distinction of this phenomenon from the well established requirement for donor���target isogeny in conventional gene targeting (18). Based on the assumption that SDSA is the major break repair pathway for endogenous loci in mammalian cells, we hypothesized that confining this divergence to the position corresponding to the DSB could lower this penalty (Fig. 1 A and C). To investigate this issue, we built a donor DNA plasmid engineered to introduce a 4-aa tag in-frame with an endogenous locus (Fig. 1C). In this construct, 750-bp homology arms isogenic to the IL2R locus are interrupted with a 12-bp stretch engi- neered to introduce four new amino acids, RAKR, i.e., a furin cleavage site, in-frame with the endogenous IL2R ORF (for ease of subsequent detection, this tag was engineered to carry a StuI recognition site). Importantly (Fig. 1C), the tag was placed between Asn228 and Pro229, i.e., precisely at the position of the ZFN-induced break. We introduced this donor plasmid along with an expression plasmid encoding designed ZFNs engineered to introduce a DSB in exon 5 of IL2R into K562 cells. After 72 h of culturing the cells in normal medium and in the absence of any selection, we harvested genomic DNA and measured the fre- quency of tag integration by a highly quantitative PCR-based assay (16) that measures the percentage of total chromatids that have acquired an StuI site. In agreement with expectation, no measurable tag integration into the chromosome was observed in the absence of ZFNs (Fig. 1B, third lane). In contrast, 15% of the chromatids were StuI-sensitive in cells exposed both to the donor plasmid and the ZFNs (Fig. 1B, fourth lane). We isolated, cloned, and sequenced the cognate stretch of the X chromosome from this cell sample this analysis (Fig. 1C) demonstrated the precise ������copy-pasting������ of the four codons from the donor plasmid into the chromosome, i.e., the reconstruction of the donor-specified genotype at an endogenous locus in human cells. The speed (72 h) and frequency (15%) of this genome editing event was identical to that we had earlier observed in transferring single-base-pair changes to the chromosome (16). These data suggest that a properly placed exogenous DNA sequence can be transferred with high efficiency and in the absence of selection from a plasmid donor into a native locus in mammalian cells by inducing a DSB at the target using designed ZFNs. A ZFN-Invoked DSB Can Be Used to Target the Efficient Integration of an ORF and an Expression Cassette in the Absence of Selection. The high efficiency of ZFN-driven tag transfer from a plasmid donor Fig. 1. A ZFN-induced DSB leads to efficient, homology-based tag transfer into a native chromosomal locus. (A) Experimental outline and a schematic of the process whereby a ZFN-induced DSB is repaired by using an extrachromo- somal donor as a template. (B) PCR-based measurements of ZFN-driven tag integration frequency into the IL2R locus in K562 cells. Cells were left untransfected (first lane, ������neg.������ for negative control) or were transfected with an expression cassette for ZFNs that induce a DSB at exon 5 of IL2R (16) (second lane), and donor plasmids carrying a 12-bp tag flanked by 750-bp homology arms, in the absence (third lane) and presence (fourth lane) of the IL2R ZFNs. Genomic DNA was extracted 72 h later. The IL2R locus was amplified by 20 cycles of PCR in the presence of radiolabeled dNTPs by using primers that hybridize to the chromosome outside of the donor homology arms, and the PCR products were digested with StuI, resolved by 10% PAGE, and autoradiographed. The percentage of StuI-sensitive DNA is indicated below the fourth lane. (C) Sequence analysis of ZFN-edited chromatids. The primary DNA sequence, and the amino acid sequence it encodes, of exon 5 of the human IL2R gene, along with the target sites of the designed ZFNs, are indicated. The central portion of the donor sequence, along with the tag, is shown below. A representative chromatogram of the DNA sequence of one of the chromatids obtained from sample 4 (in B) is provided, showing the chromosomal sequence to be altered precisely in the manner specified by the donor, i.e., by copy-pasting of codons for four new amino acids in-frame with the endogenous ORF. Note that an additional silent SNP (Pro229 CCA3CCT), introduced for cloning purposes, is also transferred from the donor. 3056 www.pnas.org cgi doi 10.1073 pnas.0611478104 Moehle et al.