An improved zinc-finger nuclease architecture forhighly specific genome editingJeffrey C Miller1, Michael C Holmes1, Jianbin Wang1, Dmitry Y Guschin1, Ya-Li Lee1, Igor Rupniewski1,Christian M Beausejour1,2, Adam J Waite1, Nathaniel S Wang1, Kenneth A Kim1, Philip D Gregory1,Carl O Pabo1,2& Edward J Rebar1Genome editing driven by zinc-finger nucleases (ZFNs) yields high gene-modification efficiencies (>10%) by introducing arecombinogenic double-strand break into the targeted gene. The cleavage event is induced using two custom-designed ZFNs thatheterodimerize upon binding DNA to form a catalytically active nuclease complex. Using the current ZFN architecture, however,cleavage-competent homodimers may also form that can limit safety or efficacy via off-target cleavage. Here we develop animproved ZFN architecture that eliminates this problem. Using structure-based design, we engineer two variant ZFNs thatefficiently cleave DNA only when paired as a heterodimer. These ZFNs modify a native endogenous locus as efficiently as theparental architecture, but with a >40-fold reduction in homodimer function and much lower levels of genome-wide cleavage.This architecture provides a general means for improving the specificity of ZFNs as gene modification reagents.Zinc-finger nucleases (ZFNs) are rapidly emerging as versatile reagentsfor gene modification. These hybrid restriction enzymes, which linkthe cleavage domain of FokI to a designed zinc-finger protein (ZFP)1,2,may be used to introduce a variety of custom alterations into thegenomes of eukaryotic cells. Examples range from precise sequenceedits3–6to the targeted integration of entire genes7.ZFNsinitiatetheseevents by introducing a double-strand break at the site chosen formodification. If an exogenous repair template is also supplied thensequence alterations encoded in this donor may be incorporated intothe genome by homology-directed repair (HDR)8. Targeting of thecleavage event, which is central to ZFN specificity and versatility, ismediated by the ZFP domain. This DNA-binding domain has beencharacterized in great detail9–11and may be engineered to recognize awide variety of chosen DNA sequences12–16.ZFNs may offer a general method for engineering the genomes ofdiverse species as the requisite DNA repair pathways are highlyconserved17. ZFN-stimulated gene modification has been demonstratedin plants18,19,insects4,6,20, roundworms21and human cells3,5,withendogenous gene correction rates as high as 18%5.Moreover,donor-free delivery of ZFNs may be used for the targeted disruption ofendogenous genes4,6,20. The generality of ZFN-mediated gene modifi-cation raises the prospect of applications in diverse fields including cropengineering, therapeutic gene correction, cell line customization forbiologics production and the development of nonmurine models ofhuman disease. Realizing the full potential of these approaches, how-ever, will require the development of ZFN architectures and designstrategies that yield efficient cleavage while minimizing off-target effects.An important feature of the current ZFN architecture is that DNAcleavage requires nuclease dimerization. ZFNs, as well as wild-typeFokI, interact through their cleavage domains and are inactive asmonomers22–25. As a consequence, cleavage of a typical target requiresthe design of two different ZFNs for binding to adjacent half-sites(Fig. 1a,b). Moreover, the dimerization interaction is quite weak; FokIremains monomeric at concentrations of at least 15 mM and dimerizesonly when bound to its specific target23,26. From the standpoint oftargeting specificity, an advantage of this arrangement is that cleavagerequires simultaneous binding of both ZFNs to their respective half-sites. This has enabled the construction of ZFN dimers with cleavagespecificities of up to 24 bp5, a length that exceeds the target size ofother proposed gene modification agents such as the meganuclease I-Sce I27and that offers the prospect for unique targeting within thehuman genome. The development of ZFNs that specifically cleaveeven longer targets may be anticipated, as individual ZFPs have beensuccessfully designed for sequences as long as 18 bp28–31.Although the requirement for dimerization opens up the possibilityof restricting cleavage to very long and rare sequences, it alsointroduces a problem arising from the fact that protein-proteininteractions mediated by the wild-type FokI cleavage domain are notthemselves selective for the heterodimer species (Fig. 1c). As aconsequence, the expression of any ZFN heterodimer (e.g., Lwt/Rwtin Fig. 1c) also yields two side-product homodimers (Lwt/Lwtand Rwt/Rwt) that will be irrelevant for gene modification but maynonetheless limit safety or effectiveness by cleaving off-targetsequences. Moreover, if one ZFN is less specific than the other, theReceived 31 January; accepted 4 June; published online 1 July 2007; doi:10.1038/nbt13191Sangamo BioSciences, Inc., Pt. Richmond Tech Center, 501 Canal Blvd., Suite A100 Richmond, California 94804, USA.2Present addresses: De´partement depharmacologie, Centre de Recherche, CHU Ste-Justine 3175, Coˆte Ste-Catherine, Montre´al, Quebec H3T 1C5, Canada (C.M.B.) and Department of Systems Biology,Harvard Medical School, 200 Longwood Avenue, WAB 536, Boston, Massachusetts 02115, USA (C.O.P.). Correspondence and requests for materials should be addressedto E.J.R. ([email protected]).778 VOLUME 25 NUMBER 7 JULY 2007 NATURE BIOTECHNOLOGYARTICLES© 2007 Nature Publishing Group http://www.nature.com/naturebiotechnologyhomodimer of the less specific ZFN will tend to be the dose-limitingspecies in terms of off-target effects. Evidence of this phenomenon wasseen in a recent study, in which two out of three designed ZFNheterodimers exhibited substantial toxicity that limited effectivenessand that was, in each case, due to homodimer formation of a singleZFN6. Moreover, mutating the active site in one such ZFN relieved itstoxicity, indicating that homodimer cleavage properties, rather thanDNA binding, mediated the toxic effect6.The creation of FokI cleavage domain variants that preferentiallyheterodimerize would provide a general solution to the problem ofhomodimer formation while preserving the advantages of dimeriza-tion-dependent cleavage (Fig. 1d). Here, we describe the developmentof such cleavage domain variants using a strategy of iterative structure-based design followed by screening for HDR-driven gene correction inmammalian
View Full Document
Unlocking...