A new study, published in Nature Biotechnology, investigates the use of CRISPR/Cas9 to make simple cuts in DNA in mouse cells and subsequent on-target mutagenesis due to the inaccuracy of the type of DNA repair process employed.
Dr Helen O’Neill, Programme Director, Reproductive Science and Women’s Health, University College London, said:
“It has long been understood that certain DNA repair processes, such as the one assessed here (non-homologous end joining), can have unpredictable outcomes, whether innate (occurring naturally in cells) or induced (through genome editing). The research presented here looks more thoroughly at the expected unpredictability of this repair process following intended CRISPR/Cas9 cutting and naturally suggests that this is acknowledged in others’ analyses, especially in a future clinical setting.
“Of course, when assessing a new technology like this, a multitude of variables need to be accounted for; cell type, the intended DNA cut, the guides used to locate this genomic cutting site and, quite importantly, the choice of molecular scissors, of which there are many. Cas9 was the chosen molecular scissors in this study and while different cell types were analysed, the genetic architecture of cells in vitro (in a dish), will never truly reflect their behaviour in a body (in vivo).
“When mutations are made in a genome, as were made in three different genes in this paper, we would anticipate that this would have a ripple effect in the genomic landscape. This paper gives an appreciation of the extent of this for the chosen mutations in their chosen cell types but the technology is ever-improving.
“It is well understood that sufficient sequencing would take place prior to the introduction of this technology clinically, but this paper should encourage researchers to thoroughly analyse their intended editing.”
Dr Christophe Galichet, Senior Laboratory Research Scientist at The Francis Crick Institute, said:
“The data produced in the paper are very sound and well supported. However, one can wonder how the type of mutations describe in the paper, some of which being large deletions, were missed in previous studies. Indeed, unwanted mutations when using Crispr/Cas have been a recurrent topic of discussion. Previous studies have analysed the extent of unwanted mutations when using Crispr/Cas, using whole genome sequencing or other methods, and did not find evidence of off-target mutation or large DNA deletions/ insertions.
“The paper exacerbates the need of better understanding how cells repair the damage caused by the cut in the genome. Furthermore, it is essential to understand whether the unforeseen mutations described in the paper is present in all cell types, for all genes (actively expressed, not expressed) and/or for all Crispr/Cas9 guides. Also differences between in vitro (Petri dish) and in vivo (animals) should be investigated. Indeed, one could think that cells in culture could be subject to stress (ie oxidative stress) which could lead to changes in cell behaviour.
“The paper does not remove the validity of the use of Crispr in clinical and non-clinical studies. A greater understanding of the Crispr/Cas9 system outcomes will lead to a safer use, as we will know what to expect. In non-clinical studies, where research animals are used, specific genetic mutations are selected for and thus deletion/insertion/chromosomal rearrangement which could be misread as wildtype allele, like described in the paper, will be discarded.”
Prof Robin Lovell-Badge FMedSci FRS, Group Leader, The Francis Crick Institute, said:
“It is hard to know whether the few results in this short paper are significant or not. The authors claim that they find unsuspected DNA damage associated with the use of “CRISPR” genome editing (Cas9 and gRNAs). This has been seen before, albeit rarely, and is obviously something that would need to be checked in any clinical application of the methods. The authors suggest that these alterations in DNA, which tend to be at or close to the target site, occur at much higher levels than anticipated. The paper is technically very sound – indeed the work is of high quality. However, it is not at all clear that the specific protocols used in the paper would relate much to any sensible use of genome editing clinically.
“The authors focus only on genome editing that relies on non-homology end joining (NHEJ) endogenous DNA repair mechanisms. These are known to be untrustworthy compared to methods that rely on homology directed repair to substitute one DNA sequence with another, and especially those that use “base editing” (and DNA mismatch repair), both of which can be used to repair mutant genes.
“Many experiments seem to have involved selection of cells exhibiting loss of target gene function. When gRNAs were used that recognised exonic (protein coding) sequences the efficiencies were understandably much higher than when they recognised intronic sequences (that largely contain “junk” DNA). However, most of the data presented in the paper were on the latter, where the typical small insertions or deletions (INDELs) produced by NHEJ repair (especially after a single double-strand break in DNA) will not cause loss of gene function. Therefore, it is perhaps not surprising that the authors most often found large deletions, rearrangements or insertions.
“Whether these large DNA alterations were indeed caused just by the genome editing components is also unclear. For most of the experiments reported, and all of those shown in the main paper, the genome editing components were introduced using a PiggyBac transposon system, which itself involves components that can modify DNA sequences, notably a very efficient transposase that frequently leads to insertions and transpositions. (The authors state that they had some similar results when they just introduced gRNA and Cas9 (as RNPs), but these data are shown in Supplementary Data that were not made available for comment.)
“The data in the paper mostly uses mouse embryonic stem cells, which probably have DNA repair mechanisms and properties different from most somatic cell types that would be the targets of somatic gene therapy using genome editing methods. Some experiments made use of an immortalised pigmented retinal epithelial cell line, however, the immortalisation process may have given these cells properties quite distinct from normal somatic cells. The only cells of slight relevance to somatic gene therapy treatments used in the paper are mouse bone marrow stem cells, but these had already been genetically altered to contain a constitutively active Cas9 gene. The latter will tend to drive genome editing repeatedly until the gRNAs are no longer present or the targeted gene has been so disrupted that the gRNAs no longer recognise it. In this case, 61 out of 96 clones did not have any large deletions of rearrangements, which is actually quite reassuring. (Again, this is part of the supplementary data, so I have not been able to see it.)
“Finally, the authors only test three genes, two of which are X-linked and one is an introduced GFP gene. None of these may be typical.
“In conclusion, this paper highlights the need when conducting experiments involving genome editing, and particularly for clinical applications, to verify that the alterations to the DNA sequence are those, and only those, that had been designed to occur. This has always been obvious. But the results give no reason to panic or to lose faith in the methods when they are carried out by those who know what they are doing.”
Dr Francesca Forzano, Consultant in Clinical Genetics and Genomics, Guy’s & St Thomas’ NHS Foundation Trust, said:
“The scientists at the Wellcome Sanger Institute have identified collateral genomic damage resultant from the use of CRISPR/Cas9 gene editing. Using long-read sequencing and long range PCR genotyping they could detect large deletions and genomic rearrangements on-target, and genomic rearrangements distal to the cut site in a large proportion of cells examined.
“This important work demonstrates that this technique is much less safe than previously thought, and suggests that the techniques previously used to monitor the efficiency and safety of CRISPR/Cas9 exploitation were not entirely adequate.
“This work represents a milestone in the gene editing field and signpost that more caution should be exerted in the application of this technique and more research is needed before considering any possible clinical application.
“This is a good quality research piece. Results fit to the existing evidence and brings new data and robust evidence regarding a collateral damage of this technique. The data will have a great impact in this field, as it suggests different techniques for monitoring on and off site effects should be used. They also demonstrate that this technique can be less safe than previously anticipated. Researchers should take this work into account when looking into clinical applications using this technique.”
Dr Lydia Teboul, Head of Molecular and Cellular Biology at the Mary Lyon Centre, MRC Harwell Institute, said:
“The article by Kosicki and colleagues describes an in-depth analysis of the range of genomic changes that occur at a locus targeted by CRISPR/Cas9 with the aim to create a deletion.
“They demonstrate a variable outcome, in particular with larger lesions than they had anticipated. The concept of unpredictability of the outcome of genome by CRISPR/Cas9 is currently arising in a series of articles both in cell culture and embryos±.
“This is the recognition that we still have to learn a lot about the system (which has only been used for a few years) and just as importantly about the way cells repair cuts in their DNA. The technology still holds great promises for our developing therapeutic strategies that involve making genetic changes in cells and tissues. With these important studies we are learning about the precautions that we will need to take as we use the powerful genome editing tools for therapy.”
* ‘Repair of CRISPR–Cas9-induced double-stranded breaks leads to large deletions and complex rearrangements’ by Michael Kosicki et al. will be published in Nature Biotechnology on Monday 16th July.
All our previous output on this subject can be seen at this weblink: http://www.sciencemediacentre.org/tag/genome-editing/
Dr Helen O’Neill: “No conflicts of interest.”
Dr Christophe Galichet: “The Wellcome Trust, that supported the paper, is also a founding partner of the Francis Crick Institute.”
Prof Robin Lovell-Badge: “No conflicts of interest.”
Dr Lydia Teboul: “I am managing the genome engineering team at the Mary Lyon Centre. My team has worked on questions similar to the topic raised in the paper and published articles showing the unpredictability of the outcome of genome editing in mouse embryos. Also, the Mary Lyon Centre has a long history of collaboration with the Wellcome Sanger Institute and in particular with Professor Bradley’s team, as part of large-scale genetics programmes.”
None others received.