High-fidelity CRISPR-Cas9 nucleases have no detectable off-target mutations

"Our creation of a Cas9 variant that brings off-target effects to levels where we can no longer detect them, even with the most sensitive methods, provides a substantial advance for therapeutic applications in which you want to accurately hit your target without causing damage anywhere else in the genome," says J. Keith Joung, MD, PhD, associate chief for Research and the Jim and Ann Orr MGH Research Scholar in the MGH Department of Pathology, senior author of the Nature paper. "But its impact will also be incredibly important for research applications because off-target effects can potentially confound the results of any experiment. As a result, we envision that our high-fidelity variant will supplant the use of standard Cas9 for many research and therapeutic applications."

Used to create targeted DNA breaks at which genetic changes can be introduced, CRISPR-Cas9 nucleases combine a bacterial DNA-cutting enzyme called Cas9 with a short guide RNA sequence that can bind to the target DNA sequence. While easier to use than previous gene-editing tools, CRISPR-Cas9 nucleases have a well-characterized and significant limitation. As described in 2013 studies led by Joung and others, CRISPR-Cas9 nucleases can induce off-target DNA breaks at sites that resemble the on-target sequence. Subsequent investigations by Joung's team and others have reduced but never completely and consistently eliminated these off-target effects.

Joung and his colleagues hypothesized that reducing interactions between Cas9 and the target DNA might more completely eliminate off-target effects while still retaining the desired on-target interaction. The MGH team focused on the fact that certain portions of the Cas9 enzyme itself can interact with the backbone of the target DNA molecule. Pursuing an observation originally made by co-lead author Vikram Pattanayak, MD, PhD, of MGH Pathology, the team altered four of these Cas9-mediated contacts by replacing the long amino acid side-chains that bind to the DNA backbone with shorter ones unable to make those connections. "Our previous work suggested that Cas9 might bind to its intended target DNA site with more energy than it needs, enabling unwanted cleavage of imperfectly matched off-target sites," says Pattanayak. "We reasoned that, by making substitutions at these four positions, we could remove some of that energy to eliminate off-target effects while still retaining full on-target activities."

Co-lead author Benjamin Kleinstiver, PhD, of the MGH Molecular Pathology Unit and Michelle Prew, a research technician in Joung's lab, then tested all 15 possible variants in which any combination of one, two, three or four of those amino acid side-chains were altered and found that one three-substitution and one four-substitution variant appeared to show the greatest promise in discriminating against mismatched target sites while retaining full on-target activities in human cells.

The researchers then more fully characterized the four-substitution variant, which they called SpCas9-HF1 (Sp for the Streptococcus pyogenes bacteria, which is the source of this widely used Cas9, and HF for high-fidelity). They found that this variant induced on-target effects comparable to those observed with the original unaltered SpCas9 when used with more than 85 percent of 37 different guide RNAs they tested. Using GUIDE-Seq, a highly sensitive system Joung's lab developed in 2014 to detect off-target CRISPR-Cas9 effects across the genome, the team found that, while nucleases combining unaltered SpCas9 with seven different guide RNAs induced as many as 25 off-target mutations, use of SpCas9-HF1 produced no detectable off-target effects with six of those guide RNAs and only one off-target site with the seventh. These results were further confirmed using targeted deep-sequencing experiments.

Joung's team also found that SpCas9-HF1 could reduce off-target effects when targeting atypical DNA sites characterized by repeat sequences of one or two nucleotides -- sites that are typically subject to many off-target mutations. They developed additional derivatives of SpCas9-HF1 -- called HF2, HF3 and HF4 -- which could eliminate the few residual off-target effects that persisted with the HF1 variant and a small number of guide RNAs. "If SpCas9-HF1 using a certain guide RNA still produces a handful of off-target effects that are particularly difficult to eliminate, it may be possible to engineer new variants that get rid of even those effects," says Joung, who is a professor of Pathology at Harvard Medical School.

The researchers also showed that SpCas9-HF1, like its naturally occurring counterpart, could be combined with other useful alterations that extend its utility. Previous work from the Joung lab published last summer in Nature had shown that introducing a series of amino acid substitutions could expand the targeting range of unaltered SpCas9. In the current study, the authors show that introducing these same alterations into SpCas9-HF1 also extended the targeting range of the high-fidelity variant. "These results show that these variants should be broadly useful to anyone currently using CRISPR-Cas9 technology," says Kleinstiver. "They can easily be used in place of wild-type SpCas9 and provide a highly effective method for reducing off-target mutations to undetectable levels."