A major goal for gene therapy is to correct genetic defects in patient-derived stem cells, which could be put back into the patient to alleviate disease symptoms. However, the most widely used approach involves the genomic insertion of a transgene, which — owing to the non-specific nature of insertion — runs the risk of harmful side-effects. Advances using site-specific DNA-cutting enzymes are now showing promise for safer forms of gene therapy.

Zinc-finger nucleases (ZFNs) are engineered DNA editing enzymes that consist of a DNA binding zinc-finger domain and the nuclease domain of a restriction endonuclease. These enzymes produce double-stranded breaks, which can be targeted to a site of choice by altering the zinc-finger DNA binding specificity. ZFNs have been used previously to correct point mutations by inducing homologous recombination that brings about gene conversion from a donor DNA.

This approach avoids the insertion of new material into the genome. However, it requires the efficient delivery of three different constructs — two ZFNs (as the enzymes act as heterodimers) and the donor DNA — and current systems for delivery into clinically relevant target cells are highly inefficient. To overcome this problem, Lombardo and colleagues used an integrase-defective lentiviral vector (IDLV), which can infect most primary cells and deliver transgenes efficiently, but doesn't integrate into the host genome. For three human cell types, the authors showed that delivery of the necessary constructs using three IDLVs resulted in the desired gene editing outcome. For the different cell types, a variable but significant proportion of cells showed gene correction. In cases in which an endogenous gene was edited, there was normal expression of the gene product.

Could the uses of ZFNs be extended by using them for site-specific gene addition? The authors showed that this can be achieved by using a construct in which the gene to be added is flanked by sequences that are homologous to the target site. This was achieved for the IL2RG gene (which is mutated in severe combined immunodeficiency), allowing a promoterless region of this gene to be inserted within the endogenous locus. The result was expression of the transgene under the control of the IL2RG promoter, suggesting that physiological expression could be achieved for gene therapy.

The methods described above have a limitation in that new ZFNs would need to be designed for different mutations and different target genes. In the case of gene addition, this could be overcome by finding a genomic location into which any transgene could be inserted, using the same ZFNs in each case. The authors picked the CCR5 gene as a potential insertion site, as homozygous null mutations at this gene seem to have no ill effects in humans. The authors successfully achieved specific and efficient gene addition at this site in a range of human cell lines, including both haematopoietic progenitors and embryonic stem cells. In the latter case, expression of an inserted GFP transgene was maintained after induction to a neural progenitor fate, suggesting that this approach does not affect the differentiation and self-renewal properties of stem cells.

Thorough safety assessments will be needed before such approaches can enter the clinic, and questions about possible off-target effects must be addressed. However, this study highlights an exciting range of possible applications for ZFNs in gene therapy.