The enzyme Cas9 is used in genome editing to cut selected DNA sequences, but it also creates breaks at off-target sites. Protein engineering has now been used to make Cas9 enzymes that have minimal off-target effects. See Article p.490
For the past 30,000 years, humans have been genetically engineering the wolf through selective breeding, preserving some forms of genes and eliminating others to produce the dog. Now, two studies (one on page 490 of this issue1 and one in Science2) have used genetic engineering to tame a different type of wild creature — a nuclease enzyme called Cas9. In doing so, they have markedly reduced the enzyme's undesirable natural tendencies, but have preserved its ability to cut DNA in an RNA-guided manner. This feat of molecular domestication is great news for practitioners of genome editing, in which DNA sequences in cells or organisms are changed to scientists' specifications efficiently and accurately; such precise editing requires highly targetable nucleases3.
In its natural 'wild' state, Cas9 is part of the bacterial immune system. When a bacterium is infected by a parasite such as a virus, the organism's cellular machinery cuts up and retains pieces of the invader's DNA, storing the sequences in a region of the bacterium's own genome called a CRISPR locus4. Cas9 then polices the bacterium for repeat invaders by carrying with it an RNA copy of a sequence stored in the CRISPR locus (for simplicity, this RNA is referred to here as a guide RNA, or gRNA). The enzyme compares intracellular DNA to the sequence in the gRNA, and if there is a match, Cas9 cuts the invading DNA. Attackers evade detection by changing their DNA sequence, so Cas9 evolved to cut incoming DNA even if its sequence is a less-than-perfect match to the gRNA.
Studies of this and other bacterial defence mechanisms have had a major impact on genome editing. In this process, a nuclease cuts DNA inside the cell and, as this DNA break is being repaired, the desired edit (disruption, correction or insertion of a gene) takes place3,4,5. The first genome-editing experiments made use of another class of nuclease, zinc-finger nucleases (ZFNs)5, but the discovery that Cas9 is led by a gRNA dramatically expanded the scale and scope of genome-editing applications for research purposes6, because the gRNA enables easy and relatively efficient enzyme programming.
Cas9 evolved to defend a bacterium that has a genome 1,800 times smaller than the human genome, and cutting DNA sequences that are imperfectly matched to the gRNA is an adaptation to its natural battlefield. As a consequence, when Cas9 was brought in from the wild and placed in human cells, it introduced genetic changes to unintended stretches of DNA in addition to editing the gene of interest7. Imagine a short-sighted witness to a crime attempting to identify the perpetrator in a police line-up, relying not only on the facial features that make the criminal unique but also on those shared with other people, such as gender or height. This could lead to a case of mistaken identity, because a weak match to the witness's fuzzy mental image could be reinforced by a match on shared features. Similarly, wild-type Cas9 finds its target not only using the sequence-specific gRNA, but also by grasping onto the DNA backbone, which is the same in any gene.
In the present studies, Kleinstiver et al.1 and Slaymaker et al.2 set out to tame Cas9 using a thoughtful and well-executed approach that relied on an atom-by-atom understanding of how the enzyme binds to and cuts DNA8. (A similar study of how zinc fingers bind DNA9 ultimately provided the basis for the first genome-editing experiments.) The groups reasoned that, by engineering Cas9 such that its interactions with the DNA backbone were weakened, they could force the enzyme to rely to a greater extent on the gRNA–DNA pairing to recognize and cut its target (Fig. 1).
Kleinstiver and colleagues tested the editing specificity of their resulting enzyme, which they dubbed high-fidelity Cas9 (Cas9-HF), in a cancer-cell line. They programmed Cas9-HF with gRNAs for seven different stretches of human DNA: in six cases it edited only the intended target, and in the remaining stretch the enzyme was weakly distracted by only one other position in the DNA. By comparison, wild-type Cas9 cut at multiple unintended sequences when tested with gRNAs for these seven gene targets. Crucially, for 75% of targets tested, Cas9-HF was just as potent a genome editor as its wild ancestor. Slaymaker and colleagues followed the same overall principle to produce an enzyme that they called enhanced specificity Cas9 (eCas9), although the details of their engineering and analyses differed.
These 'domesticated' Cas9 enzymes are sure to be used by laboratories the world over. The immediate impact will be to shorten the time it takes to complete a genome-editing experiment, because the need to check for undesired edits will be reduced. Cas9 has been used to systematically scan many human genes at a time for those that underlie a trait of interest10, and such experiments will now be more efficient. In agriculture — especially in species that have extended life cycles, such as crops or cattle — the use of enhanced Cas9 might obviate the need to perform time-consuming crosses to obtain a pristinely edited organism.
Genome editing was first applied in the clinic in 2009, when an ex vivo ZFN-based approach was used to edit certain immune cells from people with HIV11. This approach has since been used to treat more than 80 patients, and has a good safety record. Last December, the first clinical trial of in vivo gene editing, a ZFN-based approach for treating haemophilia12, passed review by the US Food and Drug Administration. ZFNs have already been engineered to a level of specificity that is comparable to that of Cas9-HF (see go.nature.com/mkl6v1), and they have passed regulatory hurdles for use in clinical trials in both in vivo and ex vivo applications. The current studies inspire confidence that the scope of clinical genome editing will continue to expand. Advances in this field offer the promise of engineering genetic cures for many diseases — a prospect that is both encouraging and within our reach.Footnote 1
Kleinstiver, B. P. et al. Nature 529, 490–495 (2016).
Slaymaker, I. M. et al. Science 351, 84–88 (2016).
Carroll, D. Annu. Rev. Biochem. 83, 409–439 (2014).
Doudna, J. A. & Charpentier, E. Science 346, 1258096 (2014).
Urnov, F. D., Rebar, E. J., Holmes, M. C., Zhang, H. S. & Gregory, P. D. Nature Rev. Genet. 11, 636–646 (2010).
Bolukbasi, M. F., Gupta, A. & Wolfe, S. A. Nature Methods 13, 41–50 (2015).
Tsai, S. Q. et al. Nature Biotechnol. 33, 187–197 (2015).
Jiang, F. & Doudna, J. A. Curr. Opin. Struct. Biol. 30, 100–111 (2015).
Pavletich, N. P. & Pabo, C. O. Science 252, 809–817 (1991).
Shalem, O., Sanjana, N. E. & Zhang, F. Nature Rev. Genet. 16, 299–311 (2015).
Tebas, P. et al. N. Engl. J. Med. 370, 901–910 (2014).
Sharma, R. et al. Blood 126, 1777–1784 (2015).
The author declares competing financial interests: F.U. is a full-time employee of Sangamo BioSciences, Inc.
About this article
Clinical Journal of the American Society of Nephrology (2020)
European Heart Journal (2020)
Nature Methods (2017)
Best Practice & Research Clinical Gastroenterology (2017)