A scientific article reporting unintended effects of the CRISPR gene-editing technique has riled executives of companies hoping to commercialize the technology. But they’re arguing on the wrong side of the science, reports Claire Robinson

UPDATE March 2018: The Schaefer et al (2017) article has been retracted.

Two gene-editing companies are attacking a scientific article[1] that caused their stocks to plummet, claiming that it is filled with errors and saying it should not have been published. In letters sent to Nature Methods, representatives of Intellia Therapeutics and Editas Medicine criticized the article, which reported that the gene-editing tool CRISPR had caused unexpected mutations in the genomes[2] of mice. In a syndrome that will be familiar to GMWatch readers, the CEO of one of the companies has even demanded that the editor of the journal retract the article.

But the findings reported in the article, along with other recent research papers that also report unintended effects of CRISPR gene editing, show that the companies are arguing on the wrong side of the science.

Scientific article hurt companies’ bottom line

The letter from Intellia stated that the explanation offered by the authors of the article for the unexpected effects in the mouse genomes “requires activities that Cas9 is not known to possess”. Intellia CEO Nessan Bermingham told Technology Review: “This publication has garnered a significant level of media and public attention resulting in significant damage. Given the issues around the design and interpretation I believe it is appropriate that the Nature Methods editorial board retract this paper.”

Retraction Watch commented, “This isn’t the first time we’ve seen corporations calling for the retraction of articles that aren’t good for their bottom line.”

Unintended mutations

GMWatch reported on the offending scientific article in June. The article presented research findings that using CRISPR to generate GM mice by injection into fertilized eggs led to mutations that were unexpected, not just in quantity but also in quality. Not only was the number of mutations far greater than expected, amounting to well over a thousand, but most of the mutations resulting from the CRISPR process were of a type that had never before been reported.

None of these DNA mutations were in locations predicted by the computer programs that are widely used by researchers to screen the genome of an organism to look for potential off-target effects of CRISPR. Among the top fifty sequences predicted for off-target effects – which would have been picked up by the computer programs – none were mutated.

The researchers sequenced the entire genome of the GM mice and identified all the mutations. They found that a number of mutations – 164 and 128 respectively in each of the two gene-edited mice – were of the expected type: insertions or deletions (indels) of DNA base units at the CRISPR target site. Most of the mutations identified, 1736 and 1696 respectively in each mouse, were single DNA base unit alterations. That means that one base unit type had been replaced with another, giving rise to what is known as a single nucleotide variation or SNV.

In addition, the vast majority of the CRISPR-associated mutations, 117 indels and 1397 SNVs, were the same in the two GM animals studied. This indicates a targeted and non-random process. The rate of CRISPR-associated mutations was also much higher than the known natural rate of genome mutation in mice. Furthermore, a comparison with the mouse genome variation database, collated from the study of 36 mouse strains, found that none of the SNVs in the GM mice were present in any of these strains. This clearly shows that the SNVs were the result of CRISPR off-target effects.

CRISPR unknowns

Commenting on the findings, the London-based molecular geneticist Dr Michael Antoniou said: “The occurrence of CRISPR-induced SNV mutations was totally unexpected, since currently there is no known mechanism by which this genome editing tool can bring about SNV changes. At face value, this suggests that CRISPR may possess molecular mechanisms of action that provoke hitherto unsuspected reactions in the genome.

“This tells us that we may not know all there is know about the mechanisms of action of the CRISPR-Cas9 system.”

Intellia responds

In its response to the scientific article, Intellia suggested that the SNVs were not caused by CRISPR but by a natural process called spontaneous deamination.

Deamination is the removal of an amino group from a molecule by enzymes known as deaminases. It can take place in both proteins and DNA. Depending on the DNA base unit that has been subject to deamination, this can result in either repair of the damage or the substitution of one DNA base unit for another. The body uses deamination to break down excess proteins to generate energy. At the level of DNA, deamination is thought to be a driver of gene variation in an evolutionary context.

Commenting on Intellia’s hypothesis, Dr Antoniou said, “It is highly unlikely that the deamination process would occur at exactly the same sites in two independently generated transgenic mice, as was seen in the study.

“Also, if the sites that resulted in SNVs were naturally prone to this kind of modification, then they would already be present in the mouse genome database. But they are not. This clearly indicates that they are novel and linked with the genetic modification procedure.”

Study could be expanded

For those who still doubt that the supposedly precise CRISPR technology could result in so many unexpected effects, Dr Antoniou says it would be relatively easy to repeat the study using more mice and sequencing the genomes of the parents. This would address the objections of critics that
1.    There were too few mice in the study (two edited mice plus one control)
2.    The genomes of the parents should have been sequenced and used as the controls, rather than an unedited mouse of unknown relationship with the edited mice, and
3.    The mutations were the result of natural processes and not CRISPR.

However, the possibility that further studies might cast more light on the findings is not a good reason to retract this or any other paper.

Another study also finds unintended effects of CRISPR

Other recent research findings confirm that those who claim that CRISPR is precise and predictable are in line for a rude awakening.

A study by a different research group[3] looked at the molecular consequences of 17 CRISPR gene-editing events in four different gene regions of the mouse genome.

The researchers found that CRISPR editing resulted in unexpected types of indels at all 17 sites in the mouse genome.

This was a surprise, as the DNA repair mechanism following the cutting of the DNA by CRISPR (known as non-homologous end-joining) is generally believed to result in approximately nine base unit deletions or insertions of a few base units. However, the authors found that depending on the site being targeted, the size of the deletion was unexpectedly large – up to 600 base units of DNA. This was particularly the case where the site targeted harboured DNA repeats – that is, stretches of DNA with repetition in the DNA sequence.

Furthermore, the authors demonstrated for the first time that the deletion resulting from the DNA repair was asymmetrically located compared with the actual CRISPR cut site. This means it was almost invariably located either upstream or downstream from the cut site, rather than symmetrically spanning the cut site, as previously assumed.

Such large deletions from a single CRISPR cut had not been clearly defined in previous studies. The authors mention that this was because the technology that is generally used to determine the extent of the deletion (known as polymerase chain reaction or PCR) had not been appropriately applied; that is, the PCR analysis was used in such a way that it would only detect small indels.

This analytical failure has two consequences:
1.    The intended alteration in a given genome region could have been far more extensive than intended and thus unexpectedly damage elements in the genome – for example, sections of targeted genes, neighbouring genes to the target site or their regulatory elements
2.    Large deletions would have been missed altogether.

The authors recommended that whole genome sequencing is necessary to look for such unintended effects, rather than the standard PCR screening method, which could miss them.

Third study finds unexpectedly large deletions

Yet another recently published study[4] has also unexpectedly found large deletions resulting from single CRISPR-induced cuts, in some cases in excess of 500 base units of DNA. In some cases, subregions of genes (“exons”) that carry information for the protein(s) for which they encode were deleted. This resulted in the formation of novel gene structures encoding truncated forms of proteins.

On the level of a whole living organism, such novel proteins could either be benign or harmful.

Implications for food safety

These three studies highlight hitherto unsuspected outcomes of genome editing. If these were to occur in the new generation of genome edited crops and foods, they could result in disturbances to many gene functions, in addition to any intended change. This could result in alterations to plant biochemistry, which in turn could result in unexpected toxicity or allergenicity.

Therefore gene-edited foods need to be analyzed with “omics” technologies (transcriptomics to look at gene expression, metabolomics to look at metabolites, and proteomics to look at the protein profile). The aim is to determine the full extent of the changes in the plant genome, both at a DNA and a gene function level. In addition, these foods need to be tested for toxicity and allergenicity in long-term studies in animals, followed by dose-escalation studies in humans.

Until such studies are done, no one can claim that CRISPR is safe, precise, or predictable.

Comparison with other gene-editing tools

The main reason that more papers are being published on the off-target effects of CRISPR than the other editing tools is because of its huge and rapid uptake due to the fact that it is cheap, quick, and easy to apply. Even a group that has relatively few financial resources can use it. It is being used extensively by medical researchers who hope to develop new gene therapies for diseases, as well as by plant biotechnologists.

But could other gene-editing tools, such as ZFN and TALEN, produce equally large deletions at the DNA cut site?

Dr Antoniou believes that this is likely. This is because once the gene-editing tool, whether it be CRISPR, ZFN or TALEN, has made the intended cut in the DNA, its job is done. At that point, a natural cellular process, such as non-homologous end joining, takes over and attempts to repair the DNA by joining up the cut ends. During this process, unexpectedly large insertions and especially deletions can occur. And clearly they will occur independently of whichever tool was used to make the initial cut in the DNA.

Dr Antoniou said: “In all likelihood, the ZFN and TALEN systems will also result in some instances of larger-than-expected deletions at the targeted cut site. We don’t know for sure, because researchers have not looked at the extent of the deletions at the target sites to the same degree as has recently been reported for CRISPR.”

While we wait for the data to come in, we cannot afford to allow companies like Intellia to remove scientific information from the record simply because it hurts their bottom line.

References

1. Schaefer KA et al (2017). Unexpected mutations after CRISPR–Cas9 editing in vivo. Nature Methods 14, 547–548. doi:10.1038/nmeth.4293 (retracted March 2018)
2. Genome = the total DNA base unit sequence.
3. Shin HY et al. (2017). CRISPR/Cas9 targeting events cause complex deletions and insertions at 17 sites in the mouse genome. Nature Communications 8, Article number: 15464. doi:10.1038/ncomms15464
4. Mou H et al. (2017). CRISPR/Cas9-mediated genome editing induces exon skipping by alternative splicing or exon deletion. Genome Biology 18:108. DOI: 10.1186/s13059-017-1237-8.