Credit: GETTY

The ability of engineered zinc finger nucleases (ZFNs) to target a specific DNA site for modification makes them an attractive option for genome manipulation in basic and applied research. Two studies of ZFN specificity now show that the in vivo off-target sites of ZFN cleavage cannot be fully predicted by in vitro or in silico analysis. This work argues for greater scrutiny of the off-target events of clinically applied ZFNs but also offers strategies for improving ZFN designs.

Pattanayak and colleagues developed an in vitro method to assay the capacity of two ZFNs to cleave each of 1011 potential target sites. In the case of the CCR5-224 ZFN, which is being used in an anti-HIV clinical trial, 37 of the cleaved sites were also present in the human genome; for VF2468, which targets the human VEGFA promoter, there were over 2,500. But is there evidence that the ZFN-induced cuts at the predicted sites also occur in vivo? This seems to be the case for one-third of the targets that were tested in cultured K562 human cells (10 out of 34 for CCR5-224 and 32 out of 90 for VF2468). The extent of cleavage activity at these off-target sites was much lower than at the intended target site and was dependent on ZFN concentration; a noteworthy finding was that an off-target site of CCR5-224 is the promoter of the cancer gene BTBD10.

The study by Gabriel and colleagues described a new way of tagging the ZFN target sites to examine the in vivo genome-wide specificity of four ZFNs. The authors used an integrase-defective lentiviral vector (IDLV) to trap DNA double strand breaks (DSBs); as the linear IDLV genome is ligated into the DSB when the break is repaired, it can mark transient and otherwise undetectable DSBs.

In this study, human K562 cells were transduced with IDLV and four sets of engineered ZFN dimers. Although most IDLV integration occurred at or around the expected ZFN target site, integration sites were also found at other recurring locations, usually within 47 bp of a site with over 66% homology to the intended ZFN target.

In both studies, sequence analysis of the off-target cleavage sites was useful for highlighting the functional elements of the ZFN that are responsible for higher tolerance to mutational changes in the target site. This information could be used to optimize ZFN design to reduce off-target activity. Strategies for increasing target fidelity include reducing the binding affinity of the ZFN, lowering ZFN expression and engineering ZFN affinity for binding sites that differ at more than 3 bp from the most closely related sequence in the genome.

Given the potential therapeutic applications of ZFNs, their DNA cleavage specificity has already received much scrutiny; however, these are the first studies to examine specificity in an unbiased, genome-wide fashion. This work highlights the importance of carrying out the assays with ZFN heterodimers (which have a different specificity to monomers or homodimers) and of using in vivo assays (Gabriel et al. showed that the ranking of in vivo cleavage sites could not be predicted by sequence homology to the intended target site).