Zinc-finger nucleases (ZFNs) are fusions of the non-specific cleavage domain from the FokI restriction endonuclease with custom-designed Cys2
zinc-finger proteins (ZFPs)1
. These chimeric nucleases induce sequence-specific DNA double-strand breaks (DSBs) that can be repaired by error-prone non-homologous end joining (NHEJ) to yield small alterations at targeted genomic loci. This strategy has enabled highly efficient gene disruption in numerous cell types2, 3
and model organisms4, 5
and has facilitated the progress of targeted gene therapy in humans6, 7
. Despite these advances and more recent methodological improvements8–10
, there remains a need for new methods that can improve the utility of these enzymes. The development of safe and effective ZFN delivery methods is of particular importance, as many viral and non-viral ZFN gene delivery systems may hinder the continued advancement of this technology. In particular, viral vectors11
are time-consuming to produce and can be associated with undesirable side-effects, such as insertional mutagenesis, while non-viral DNA and mRNA delivery systems are restricted to certain cell types and have been reported to show toxicity12, 13
and low efficiency14
. To address this problem, we set out to develop a simple alternative to conventional ZFN delivery systems by investigating the direct delivery of purified ZFN proteins to cells.
We began by introducing protein transduction domains into the established ZFN architecture. For this, the cell-penetrating peptide sequence from the HIV-1 TAT protein or a polyarginine motif was genetically fused to the N-termini of ZFNs designed to target the human CCR5
. These ZFNs, however, were consistently difficult to express or purify in quantities sufficient for analysis in cell culture (data not shown). Following these results, and based on the observation that ZFP-DNA binding domains carry a net positive charge (), we hypothesized that ZFNs might penetrate the cell in the absence of additional modification. ZFNs designed to target the CCR5
gene and lacking any transduction domain were expressed in Escherichia coli
and purified to homogeneity from either the soluble or the insoluble fractions (Supplementary Fig. 1
). In vitro
analysis confirmed that functional ZFN proteins with similar DNA cleavage profiles could be obtained by either method (Supplementary Fig. 2
and Supplementary Note
ZFN proteins are cell permeable and induce targeted mutagenesis in human cells
To determine the ability of ZFN proteins to penetrate cells and stimulate mutagenesis, we generated a fluorescence-based reporter system to measure ZFN-induced DSBs (). This system uses an integrated enhanced green fluorescent protein (EGFP) gene whose expression has been interrupted by a frameshift mutation introduced by a strategically placed ZFN cleavage site. ZFN proteins that penetrate reporter cells can induce DSBs at this target site and drive the introduction of small insertions and deletions in the EGFP locus by NHEJ. Because NHEJ is a stochastic process, approximately one-third of these mutational events (+ 2, 5, 8 …bps or − 1, 4, 7 …bps) will restore the frame and EGFP function.
Direct application of ZFN proteins to reporter cells resulted in a dose-dependent increase in EGFP fluorescence, with maximum activity (6% EGFP-positive cells) achieved after treatment with 2 μM ZFN proteins (). By comparison, transient transfection of ZFN expression plasmids under saturating conditions resulted in ~7% EGFP-positive cells (Supplementary Fig. 3
). We observed no difference in activity between ZFN proteins purified from the soluble fraction or inclusion bodies (Supplementary Fig. 4
). At all ZFN concentrations evaluated, the use of transient hypothermic culture conditions9
enhanced the efficiency of mutagenesis nearly twofold (). Extended periods of incubation (>60 min) did not increase the frequency of genome editing (Supplementary Fig. 5
). Consecutive protein treatments, however, did increase the percentage of EGFP-positive cells (). Notably, repeated treatment with ZFN proteins over three days using transient hypothermic conditions yielded ~12% EGFP-positive cells (). Sequence analysis of isolated EGFP-positive cells verified targeted mutagenesis, revealing the presence of the anticipated ZFN-induced insertions and deletions in the EGFP locus ().
To determine the contribution of each ZFN component to cellular penetration, we incubated cells with fluorescently labeled ZFN or FokI cleavage domain proteins (Supplementary Fig. 6
). Fluorescence was observed in cell lysate following treatment with ZFN – in the presence or absence of a nuclear localization sequence – but not with FokI cleavage domain, suggesting that zinc-finger domains facilitate cellular internalization.
We evaluated the efficacy of this approach for the disruption of endogenous genes by treating human embryonic kidney (HEK) 293 and human acute monocytic leukemia (THP1) cell lines, as well as primary adult human dermal fibroblast (HDF) and primary CD4+
T cells with ZFN proteins targeting the CCR5
gene. These ZFNs utilized the high-activity Sharkey
. Analysis of DNA isolated from each cell type with the Surveyor nuclease assay revealed efficient and dose-dependent endogenous CCR5
gene disruption (). HEK293 and HDF cells subjected to three consecutive treatments with 2 μM ZFN proteins exhibited gene disruption frequencies >24%, while CD4+
cells subjected to three consecutive treatments with 0.5 μM ZFN proteins exhibited gene disruption frequencies >8%. As observed in the reporter system, the frequency of gene disruption increased with repeated protein treatments (Supplementary Fig. 7
). Sequence analysis of cloned CCR5
alleles amplified from each treated cell type confirmed the presence of ZFN-induced insertions and deletions in the CCR5
gene (Supplementary Fig. 8
Modification of endogenous human genes by direct delivery of ZFN proteins
To investigate the cleavage specificity of ZFNs using this approach, we evaluated the activity of the CCR5 ZFN proteins against nine previously described6, 15
off-target cleavage sites in HEK293 cells (Supplementary Fig. 9
). In direct comparison to Lipofectamine-mediated transient transfection of ZFN expression plasmids, we found that cells subjected to consecutive protein treatments demonstrated a marked decrease in ZFN activity at every off-target site, including the CCR2
locus. Notably, no detectable ZFN activity was observed at three of these loci. Western blot analysis revealed complete degradation of delivered ZFN proteins less than 4 h after application, while cells transfected with ZFN expression plasmids produced high-levels of protein continuously from 16 to 72 h post-transfection (Supplementary Fig. 10
), indicating that the differences in cleavage specificity could be attributed to the short half-lives of transduced ZFN proteins and that limiting the duration of ZFN exposure inside cells is a viable method for minimizing toxicity16
. Consistent with these degradation kinetics, cells treated with ZFN proteins exhibited maximum activity at 8 h, whereas cells expressing ZFNs from plasmid DNA showed maximum activity at 48 h (Supplementary Fig. 10
In order to examine the breadth of this technique, we treated Chinese hamster ovary (CHO) cells with ZFN proteins designed to target the DHFR
. These ZFNs utilized various specialized cleavage domains, including Sharkey
and the evolutionarily optimized DS/RR obligate heterodimeric architecture10
. Reduced levels of functional DHFR protein, as determined by fluorescein-labeled methotrexate-based flow cytometry analysis, were observed in CHO cells following three consecutive treatments with DHFR ZFN proteins (). Notably, CHO cells incubated with ZFNs containing Sharkey
mutations exhibited a >12% reduction in functional DHFR. Sequence analysis of cloned DHFR
alleles amplified from cells treated with ZFN proteins validated these percentages and confirmed the presence of ZFN-induced insertions and deletions in the DHFR
gene (Supplementary Fig. 11
). Examination of DHFR protein levels in expanded clonal populations indicated biallelic DHFR
gene disruption frequencies >7% (Supplementary Fig. 12
), demonstrating that constitutive ZFN expression from plasmid DNA is not required for high-frequency biallelic modifications and instead can be achieved using directly applied ZFN proteins.
We observed no appreciable toxicity in HEK293 or HDF cells treated with ZFN proteins (). Toxicity was also not detected in CHO cells incubated with ZFN proteins containing either the wild-type cleavage domain or the DS/RR architecture (). However, we measured decreased proliferation in CHO and THP1 suspension cells incubated with >1 μM ZFN proteins containing Sharkey mutations (). Toxicity was also observed, qualitatively, in CD4+ cells subjected to consecutive treatments with >1 μM ZFN proteins, suggesting that sensitive cell types may require protein to be administered in consecutive low doses to minimize potential toxic effects.
We have demonstrated the intrinsic cell-penetrating capabilities of the standard ZFN architecture and have shown that direct delivery of ZFNs as proteins can be used to disrupt the expression of endogenous genes in a variety of mammalian cell types, including primary CD4+ T cells and primary adult human dermal fibroblasts, which are frequently used to generate induced pluripotent stem cells. In contrast to methods that require ZFN expression from DNA, ZFN protein delivery leads to comparatively fewer off-target cleavage events and does not carry the risk of insertional mutagenesis, making this method suitable for genome editing applications in which minimizing cellular toxicity or maintaining genetic integrity is of particular importance, such as the in vitro modeling of human diseases and the ex vivo modification of non-transformed human cell types. We show that this method can also be used to modify difficult-to-transfect cell types, including patient-derived leukemia cell lines and primary human lymphocytes, supporting the use of this technique in place of viral-mediated gene delivery for inducing gene knockouts in cultured cells for reverse genetics and drug discovery. As methods for engineering cell-permeability into proteins improve, we anticipate that protein delivery and the benefits afforded therein will be extended to other designer nucleases, including TALENs.