Are Scientists Substantially Undercounting Gene-Editing Errors?

The standard gene-editing tool, CRISPR-Cas9, frequently produces a type of DNA mutation that ordinary genetic analysis misses, claims new research published in the journal Science Advances. In describing these findings the researchers called such oversights “ serious pitfalls ” of gene-editing. In all, the new results suggest that gene-editing is more error-prone than thought and, further, that identifying and discarding defective and unwanted outcomes is not as easy as generally supposed.

Derived originally from the bacterium streptococcus pyogenes, CRISPR-Cas9 is a DNA cutting and targeting system developed in 2012, that can cut and paste nearly any gene into any organism, a scientific breakthrough that opened up new opportunities for cell and organism manipulation.

CRISPR-Cas9 was adapted from a naturally occurring genome editing system in bacteria. The bacteria capture snippets of DNA from invading viruses and use them to create DNA segments known as CRISPR arrays. The CRISPR arrays allow the bacteria to “remember” the viruses (or closely related ones). If the viruses attack again, the bacteria produce RNA segments from the CRISPR arrays to target the viruses’ DNA. The bacteria then use Cas9 or a similar enzyme to cut the DNA apart, which disables the virus.

The CRISPR-Cas9 system works similarly in the lab. Researchers create a small piece of RNA with a short “guide” sequence that attaches (binds) to a specific target sequence of DNA in a genome. The RNA also binds to the Cas9 enzyme. As in bacteria, the modified RNA is used to recognize the DNA sequence, and the Cas9 enzyme cuts the DNA at the targeted location. Although Cas9 is the enzyme that is used most often, other enzymes (for example Cpf1) can also be used. Once the DNA is cut, researchers take advantage of the cell’s own DNA repair machinery to add or delete pieces of genetic material, or to make changes to the DNA by replacing an existing segment with a customized DNA sequence. The whole assembly is often just called CRISPR.

Other gene-editing methods exist (e.g. Zn Finger, TALENs ). However, because of the flexibility of its RNA targeting mechanism, CRISPR, in particular, has been the subject of enormous excitement in the biotech and agricultural research sectors.

CRISPR has mostly been used to create genetic mutations or to insert foreign DNA at desired locations in a genome. Nevertheless, other applications, like gene drives, have also been mooted. Despite the excitement, as Friends of the Earth has summarized, only one gene-edited product, a high oleic oil soybean made by a company called Calyxt, can be found on the market.

For many uses, however, gene-editing with CRISPR is insufficiently precise and a great deal of research is currently oriented towards fixing this defect.

Much of CRISPR’s lack of precision derives from the the fact that, though it is called “editing,” CRISPR and related techniques are really only cutting enzymes. They have no DNA repair function. This means that when repairs are made by the cell to the DNA at the cut site (and the cut must be repaired for the cell to survive) they are largely out of the control of the experimenter. Ten independent editing events will, therefore, give ten different mutations at the same location in the genome.

Thus, at a very basic level, each mutation created at the target site is likely to be unique, even to the extent that DNA from other species may end up being unexpectedly incorporated into the edited genome. To add to this uncertainty, different genome locations, different cell types, different species, and different versions of CRISPR can all influence the kinds of genetic alterations at the target site.

In some applications — primarily basic research — lack of precision of this kind is not necessarily a major problem. In crop breeding, for example, cells or organisms containing undesirable alterations or off-target mutations can, in theory, be detected and discarded.

But in many other applications, such as in medicine and commercial products, only more-or-less-complete precision is acceptable, for reasons of safety. Inaccurate editing of human cells in an early gene therapy trial once resulted in 2 of 11 treated children developing leukemia due to off-target effects and led to the trial being shut down.

The question of whether researchers and developers of edited organisms could or would adequately detect and discard undesirable mutations is a live concern. Recombinetics is a commercial company that, in 2016, created a hornless cow it claimed was the intended result of a precise gene edit. But FDA researchers who examined the company’s own DNA sequence data were subsequently able to show that both of the independently edited calves contained, at the site of the edit, entire antibiotic resistance genes.

By the time FDA was able to show this, however, offspring of the calves were already incorporated into a Brazilian breeding program. This breeding program has now been abandoned.

The new Science Advances research, published on Feb 12, directly addresses whether CRISPR researchers can, in fact, detect aberrant edits.

The German and Chinese researchers who authored the the paper edited mouse oocytes (i.e. embryos) with the added step (compared to simple cutting) of adding a stretch of DNA (the donor DNA) which they hoped would become integrated at the cut site. What they unexpectedly found, however, is that, at a high proportion of target sites, complex insertions of the desired DNA occurred. Rather than simply integrating single copies of the donor DNA into the cut site, DNA integrations were commonly head-to-tail arrangements of multiple copies. As the paper states:

“Overall, we conclude that the repetitive head-to-tail integration of the donor DNA template is a common by-product of the CRISPR-Cas9-mediated HDR-based genome editing process, regardless of the donor DNA template size, sequence composition, or strandedness of the template (dsDNA or ssDNA).” [ds=double stranded; ss=single stranded]

By “common” the researchers meant that, in one experiment, among 34 edited mice, six contained head-to-tail insertions. In other experiments 30 of 49 mice contained head-to-tail insertions.

In other words, complex and aberrant DNA insertions were common findings. Importantly, they occurred in multiple experiments, meaning this seems to be true regardless of what DNA was inserted or which stretch of the genome it was inserted into. This in itself is a very significant finding.

Even more notable, however, was that these complex genetic rearrangements were rarely detected by standard analytical methods. The authors called this finding “disturbing.” They wrote: “conventionally applied PCR analysis, in most cases, failed to identify these multiple integration events, which led to a high rate of falsely claimed precisely edited alleles.” Undetected, such aberrant events “would undermine the validity of studies,” they said.

In experimental settings, this is undoubtedly true. But for the general public, a more important implication exists. With companies and biohackers hoping to bring genome-edited products rapidly ( and without regulatory scrutiny ) to the market, this research represents a significant cautionary tale; especially since the authors speculate that their results probably apply equally to other editing methods, such as TALENs and Zn Finger nucleases.