NEW YORK (GenomeWeb) – Scientists have designed a Cas9 nuclease that exhibits the effectiveness of the gold-standard Streptococcus pyogenes CRISPR/Cas9 editing system but shows effectively no detectable off-target activity, according to a new study published today in Nature.
The scientists, led by Benjamin Kleinstiver, Vikram Pattanayak, and Keith Joung from Massachusetts General Hospital and Harvard Medical School, engineered an S. pyogenes Cas9 (SpCas9) variant with mutations at four amino acid residues that are thought to be in contact with the phosphate backbone of the target DNA.
In an enhanced GFP disruption assay, this Cas9 mutant showed greater than 90 percent efficiency. The scientists named this protein SpCas9 high-fidelity variant number 1 (SpCas9-HF1) and went on to show that it had comparable on-target efficiency compared to wild-type SpCas9 when used to target both EGFP and several endogenous human genes.
"We expected it would be improved in its specificity," Joung told GenomeWeb. "What was striking was just how good it was. In many cases all of the off-target effects caused by wild-type Cas9 drop to undetectable levels," when looked for using GUIDE-seq, an unbiased genome-wide approach for finding off-target effects of CRISPR/Cas9 genome editing developed in Joung's lab.
"This is really the first demonstration where we can consistently reduce off-targets to an undetectable level," he said.
Recently, CRISPR researchers have come up with several strategies that show improved specificity, but off-target effects have always persisted at low levels. Shortening the section of the guide RNA complementary to the target can reduce off-target effects but newer approaches involve engineering new Cas9 variants.
In December, Feng Zhang of the Massachusetts Institute of Technology and the Broad Institute described several improved Cas9 variants dubbed enhanced S. pyogenes Cas9s (eSpCas9s) that reduced binding to the non-target strand of DNA.
"[SpCas9-HF1] is a great demonstration of how a structural and mechanistic understanding of the Cas9 enzyme can continually improve our genome engineering tools," Winston Yan, a researcher in the Zhang lab and a co-author on the eSpCas9s paper told GenomeWeb in an email. "It seems like the amino acid substitutions reported here might be complementary to those that are described in our paper, so perhaps there is room for further optimization of specificity."
SpCas9-HF1 wasn't quite perfect; it did, in a single case, show an off-target cut, in a gene where there was a single base mismatch between the target and the off-target site. But the study shows an identical situation — a site with a single base mismatch relative to the target — in which SpCas9-HF1 didn't make the off-target cut. Moreover, a researcher could hedge against the possibility with careful guide RNA (gRNA) selection.
"For a lot of gRNAs, when you look at the potential off-target sites in the genome, there are very few sites that are just off by one," Joung said. "If you really want to try to minimize off-targets you could use [SpCas9-HF1] but also choose gRNAs that don't have sites in the genome that are only off by one. It's not hard to find those sites."
The GUIDE-seq assay has a low limit of detection — from approximately .1 percent to .01 percent of alleles — but it's still impossible to know whether the variant truly eliminates off-target activity or if GUIDE-seq just can't see it. It's an important question and one that's not likely to be answered soon, because the detection limit is largely due to the error rate of the sequencing technology used.
The study's authors also used targeted deep sequencing to verify their GUIDE-seq results, but Joung said genome resequencing is not a practical way to look for off-target effects, especially low-frequency ones. Both GUIDE-seq and the deep sequencing were performed using Illumina MiSeq technology.
Joung said the CRISPR/Cas9 field needs to come up with better, more sensitive off-target detection methods, especially as it moves towards clinical applications of genome editing. GUIDE-seq could detect mutation frequencies between one in every 1,000 to 10,000 alleles, but for applications that involve treating millions of cells or more, that won't cut it.
Until the field invents a new detection method, SpCas9-HF1 may be the best tool available to researchers. "Right now we have this issue that we can't tell if we are able to improve it," Joung said. "If there are off-target effects below the detection limit of GUIDE-seq, then we can try to improve it further. Until we have that capability [to find them], we're kind of stuck."
Joung has submitted a patent application for the variant, but he has also submitted plasmids expressing SpCas9-HF1 to the research repository AddGene, which will be available immediately.
"Anybody can incorporate these amino acid substitutions into their vectors," he said. "We envision that a lot of people will be using this going forward for both research and therapeutic applications as well."