To the editor: Zinc-finger nucleases (ZFNs) consist of an engineered zinc-finger array fused to a nuclease domain. Dimers of ZFNs can create targeted double-strand DNA breaks, which can stimulate highly efficient gene targeting in many cell types1 (Supplementary Fig. 1 online). We found that the modular assembly method of engineering zinc-finger arrays has an unexpectedly higher failure rate than previously reported. Modular assembly advocates linking individual zinc fingers, each of which typically binds to a 3-bp ‘subsite’(Supplementary Fig. 2 online). Two large-scale surveys have suggested that modular assembly is highly effective (100%(ref. 2) and 60%(ref. 3) success rates) for making three-finger arrays designed to bind 9-bp target sites. The Zinc Finger Consortium recently assembled an archive of 141 previously published finger modules3–5 encoded on a standardized platform6. Our initial experiences using these reagents suggested that modular assembly was inefficient (Supplementary Discussion and Supplementary Table 1 online).

To perform a larger-scale test, we assembled 168 zinc-finger arrays designed for 104 diverse target DNA sites (Supplementary Table 2 and Supplementary Methods online). We tested these domains for DNA binding using a bacterial two-hybrid (B2H) assay6, which accurately identifies arrays that lack activity as ZFNs in human cells (Supplementary Discussion and Supplementary Figs. 3 and 4 online). For 79 of the 104 target sites, we did not obtain a single three-finger array that scored positively in the B2H assay (overall failure rate of~ 76%; Fig. 1a and Supplementary Table 2). Notably, modular assembly was far less effective for target sites composed of two, one, or no GXX subsites (where X is any base) compared with those composed of three GXX subsites (Fig. 1a). Additionally, because ZFNs function as dimers, we would expect failure rates for making a functional ZFN pair to be even higher (Supplementary Discussion). Notably, these values are all likely underestimates of actual failure rates because not all zinc-finger arrays that are positive in the B2H assay will be active as ZFNs in human cells (Supplementary Fig. 4).