A CRISPR spin-off causes unintended typos in DNA

Even the best editor sometimes introduces typos. That’s true whether the editor is human or a version of the much-heralded gene-editing tool CRISPR.

One type of CRISPR gene editor that changes individual DNA bases, rather than cutting DNA, introduces more unwanted mutations than expected in mouse embryos and rice plants, researchers report. Those mistakes occurred in places where the tool wasn’t supposed to make changes. Another tested base editor, however, didn’t make the undesirable edits. The results were described in two studies published online February 28 in Science.

Researchers hope to use CRISPR base editors to make improvements to crops or correct genetic diseases in people one day. But the new findings suggest that some base editors still have challenges to overcome before being safe for use.

It is necessary to rigorously put the editors through their paces, says chemical biologist David Liu, a Howard Hughes Medical Institute investigator at Harvard University. His team originally created the base editors tested in the two studies, but he was not involved in either study. “The community needs these worst-case scenario pressure tests so we can make sure there’s a good margin of safety when these agents do enter clinical trials,” he says. 

CRISPR/Cas9 is a molecular scissors that cuts DNA at precise locations. Researchers have used the gene-editing tool, introduced in 2012, to squash mosquito populations in the lab, tame ground cherries into easier-to-grow crops and alter countless other animals and plants. And last year, a scientist in China reported that he had edited DNA in two babies using CRISPR (SN: 12/22/18, p. 20). But researchers are concerned that the tool is still not safe enough to use in people.

In recent years, scientists have devised versions of CRISPR/Cas9 that don’t cut DNA, but can chemically alter individual DNA bases (SN: 11/25/17, p. 7), like a pencil erasing and correcting a mistake. DNA bases — represented by the letters A, C, G and T —carry genetic information about building an organism. Changes, or mutations, in just one of these letters can sometimes have disastrous consequences, such as diseases or disorders.

CRISPR “base editors” are designed to correct these disease-causing mutations. Base editors have been thought to be safer than the DNA-cutting editors, because they don’t actually cut DNA. But no one knows how often these tools make “off-target” genetic changes.

In the mouse study, researchers in China and California devised an experiment to test the number of off-target edits made by three types of CRISPR gene editors. For each test, researchers injected one of two cells in early mouse embryos with one of the editors. A genetic trick allowed the researchers to make cells that got the editor — and all the cells that arose from those cells — glow red, while unedited cells remained colorless. The team then separated edited and unedited cells from mouse embryos and deciphered the DNA from each group.

Cells edited with a DNA-cutting version of CRISPR/Cas9 contained no more mutations than unedited cells did, indicating that the enzyme Cas9 was cutting only where directed. Similarly, a base editor that changes the DNA base A (for adenine) to G (guanine) didn’t cause additional mistakes, the researchers found.

But a base editor that changes the DNA base C (for cytosine) to T (thymine) caused mutations 20 times as often as mutations arose in the unedited cells. The unwanted editing was still relatively rare, changing an average of 283 extra bases in each embryo. That’s a typo introduced at one in every 20 million bases, Liu calculates. Many of those mistakes happened in genes that were turned on or in DNA that was being copied when the base editor was introduced to the cell. Researchers studying the editors in rice also found additional mistakes from the cytosine, but not adenosine, base editor.

The cytosine editor probably needs some additional tweaks before it can be used safely in people, says Lars Steinmetz, a geneticist at Stanford University and a coauthor of the mouse embryo study. “All this additional information and knowledge that we gain about how well [the editors] work brings us a step closer to using them to their full potential,” he says.

The fault for the mutations made by the cytosine editors probably doesn’t lie with Cas9, says Liu, also at the Broad Institute of MIT and Harvard. He and colleagues created both of the tested base editors by bolting an editing enzyme to a version of Cas9 that can no longer cut DNA (SN: 9/3/16, p. 22). Then Cas9 and a guide RNA direct the editing enzyme where to make changes.   

But like a toddler throwing forbidden treats into a shopping cart when parents aren’t looking, the cytosine-altering enzyme can grab single-stranded DNA that it gets close to and make edits on its own, Liu says. Single-stranded DNA temporarily exists where enzymes pull apart the rungs of the DNA ladder to turn on genes or to copy DNA. The tighter the base editor binds to DNA, the more likely the editor is to get a grip on random stretches of DNA and introduce mistakes.

Newer versions of the cytosine base editors don’t bind to DNA as tightly as the one tested in these studies, Liu says. Unpublished evidence from his lab suggests that those more recent versions will produce fewer typos, he says.