Unexpectedly High Levels of Inverted Re-Insertions Using Paired sgRNAs for Genomic Deletions

Round 1

Reviewer 1 Report

The paper is concerned with a common scenario, where the use of a dual guide approach is probably the most used strategy for creating knockouts by Crispr Cas. 

The key finding here is that inversions can be common in cellular transfection.

There is the technique, used in mice and rat zygotes described as - CRISpr MEdiated REarrangement mecanisms (CRISMERE) where inversion is noted to be an outcome from the use of paired guides, evidence for on target effects that the authors cite in their excellent introduction also show that inversion is one of the outcomes to expect.  

The high incidence of inversion in the transfections summarised here undoubtedly represents a surprising finding.  As the authors detail, this is a useful finding, inversion will should lead to the required loss of expression. The paper details a simple means to screen for the event.  The authors aim here is to raise awareness of this possibility and in my opinion the paper will succeed in that aim.

The authors point out that the validation of any allele needs to account for all likely possibilities. Time and resources can be saved by looking for inversions within a transfected cell population, particularly where cell lines with more a complex genetic make up are used. These are valuable points to make.

To state that inversion is a common outcome though requires sufficient backing from evidence.  That would ideally cover multiple cell lines and multiple targets.  Transfection methods are less likely to influence the proportion of inversions seen, but variance there too would be valuable. 

The paper details the outcome of 7 cell lines and 10 targets. On transfection methods, plasmid and RNP based methods are used, all experiments involve transfection reagents, though 3 different reagents are used, alternatives of electroporation weren’t assessed.  While that’s a way from being comprehensive, the key aim, to raise awareness, isn’t undermined by not being more comprehensive, it is sufficient to suggest that inversion should be looked for whatever the target, cell line or transfection method.

The figures and tables are all well explained and necessary for the paper.  The methods are sufficiently described.

Author Response

Reviewer 1

The paper is concerned with a common scenario, where the use of a dual guide approach is probably the most used strategy for creating knockouts by Crispr Cas. The key finding here is that inversions can be common in cellular transfection.

There is the technique, used in mice and rat zygotes described as - CRISpr MEdiated REarrangement mecanisms (CRISMERE) where inversion is noted to be an outcome from the use of paired guides, evidence for on target effects that the authors cite in their excellent introduction also show that inversion is one of the outcomes to expect.  

We thank the reviewer for pointing out the above paper (which I missed out on) in which dual sgRNA based CRISMERE is successfully used for generation of in vivo model systems. We have incorporated this quotation into our introduction (line 95 in the revised manuscript, highlighted in yellow) and discuss it in the context of our findings (lines 285 to 288 in the revised manuscript, highlighted in yellow, with an additional reference included).

The high incidence of inversion in the transfections summarised here undoubtedly represents a surprising finding. As the authors detail, this is a useful finding, inversion will should lead to the required loss of expression. The paper details a simple means to screen for the event. The authors’ aim here is to raise awareness of this possibility and in my opinion the paper will succeed in that aim.

The authors point out that the validation of any allele needs to account for all likely possibilities. Time and resources can be saved by looking for inversions within a transfected cell population, particularly where cell lines with more a complex genetic make-up are used. These are valuable points to make.

We thank the reviewer for his shared view that such a technical note can be important for the scientific community aiming to generate CRISPR edited model systems. In our view, it is important/vital to consider all possible outcomes to be able to generate the best possible model system.

To state that inversion is a common outcome though requires sufficient backing from evidence. That would ideally cover multiple cell lines and multiple targets. Transfection methods are less likely to influence the proportion of inversions seen, but variance there too would be valuable. The paper details the outcome of 7 cell lines and 10 targets. On transfection methods, plasmid and RNP based methods are used, all experiments involve transfection reagents, though 3 different reagents are used, alternatives of electroporation weren’t assessed. While that’s a way from being comprehensive, the key aim, to raise awareness, isn’t undermined by not being more comprehensive, it is sufficient to suggest that inversion should be looked for whatever the target, cell line or transfection method.

We fully agree with Reviewer 1’s statement on comprehensiveness, both in terms of amount of cell lines and targets as well as on delivery options. As stated, we investigated this phenomenon in 7 cell lines at 10 different target loci, using plasmids and RNPs and either lipofection or nucleofection. The most important point in our view is that we consistently detect inverted re-insertions seemingly irrespective of above-mentioned parameters. We agree with the comment that most probably the delivery of the CRISPR/Cas9 machinery might have only a very limited influence on the outcome of the cell endogenous NHEJ pathway. We decided to insert a paragraph into the discussion (lines 303 to 306 in the revised manuscript, highlighted in yellow).

We are thankful to the comment of reviewer 1 that our data fully support the key aim, to raise awareness that such events are not uncommon and should be taken into account when validating genome engineered cells. And that such consideration will help correctly classifying cell lines as “WT”, “heterozygous” or “homozygous” modified cell lines. We decided to insert a further sentence into the discussion to make this a stronger point, also in line to comments we received from reviewer 2 (line 309 to line 318 in the revised manuscript, highlighted in green). We also included a sentence in the conclusion part to stress the fact that more cell lines, target regions and delivery protocols need to be systematically screened for inverted re-insertions to support a statement that such events are a “common outcome” (lines 303 to 306 in the revised manuscript, highlighted in yellow).

The figures and tables are all well explained and necessary for the paper. The methods are sufficiently described.

We thank reviewer 1 for his comments.

Reviewer 2 Report

The manuscript, presented as a technical note, points to a less known "byproduct" of dual gRNA CRISPR/Cas9, the reinverted re-insertion, adding new knowledge to the field. However, the observed phenomenon has not been studied before in more detail, probably because screening for cell clones harboring the desired deletion is the main objective. Generation of KO models would make use of this if a heterozygous genotype is desired. However, if a full KO model is the goal, characterizing the deleted locus to investigate whether the deletion occur beyond the targeted loci is more appropriate. Furthermore, a deep-sequencing is required to validate the model.

The authors focus on wild-type clones for a second PCR screening, in the case of double gRNA approach genomic deletions. Such clones are usually ignored if a full KO model is the objective. Therefore, the significance of the findings and the impact of the research should be carefully reconsidered and presented.

Author Response

Reviewer 2:

The manuscript, presented as a technical note, points to a less known "byproduct" of dual gRNA CRISPR/Cas9, the reinverted re-insertion, adding new knowledge to the field. However, the observed phenomenon has not been studied before in more detail, probably because screening for cell clones harboring the desired deletion is the main objective. 

This indeed has been the motivation of putting this manuscript together. As mentioned, screening of CRISPR/Cas9 genome engineered cell lines requires an ever-growing range of analyses since more and more unintended/unplanned events are being discovered. Being aware of inverted re-insertions and their consequence allows applying a simple QC measure that allows detection of unwanted NHEJ events; and will ultimately help generating more accurate research data.

Generation of KO models would make use of this if a heterozygous genotype is desired. However, if a full KO model is the goal, characterizing the deleted locus to investigate whether the deletion occur beyond the targeted loci is more appropriate. Furthermore, a deep-sequencing is required to validate the model.

I am not entirely sure if we understood the full extent of this point and I am addressing this in a broader context. Inverted re-insertions do contribute not only to the genotypes of full KO cell lines but also have significant consequences if present in cells initially classified as WT by using a PCR with external primers. Such genome engineered “WT” cells are generally considered the best possible control lines. Those “WT” cells have been treated with sgRNA/Cas9 and as such should harbour any range of potential off-target hits when compared to the desired HET or HOM deleted line. If inverted re-insertions were not detected, those cells could actually represent HET allelic status and as such would significantly skew interpretation of data. In addition, screening cells that have been originally classified as HET could turn out to be a functional KO if an inverted re-insertion happened. In our opinion, this is an important issue with two consequences. First of all, it helps selecting the appropriate control lines (WT or HET) that underwent the same engineering procedure (as summarized in lines 303 - 307 in the original manuscript). Second, it significantly helps scaling down the number of clones to be screened when establishing a full KO in polyploid lines. In fact, we point to the fact that several full KOs generated in tetraploid lines were thanks to functional nulls generated by inverted re-insertions (lines 259 to 261 in the original manuscript, quoted “In line with this, we observe that a large proportion of full KO models in tetraploid HeLa and 4T1 cell lines have been generated as combinations of KO and re-inserted inversions, at 30% and 100%, respectively”). We decided to re-phrase the paragraph (line 309 to line 318 in the revised manuscript, highlighted in green) to make our point more apparent.

We fully agree with reviewer 2 that for final validation of the generated model system a deep sequencing approach is the ultimate quality control that will ensure correct modifications in all alleles and across the wider genomic context. In our manuscript we recommend the use of targeted locus amplification to uncover hidden modifications (lines 327 to 329 in the original manuscript, quoted here “To be fully aware of any other OFF-targeting events, a comprehensive genome wide analysis such as targeted locus amplification [30] is highly recommended”). However, to make this point clearer we decided to insert a statement into the discussion (lines 341 to 343 in the revised manuscript, highlighted in green, with an additional reference).

The authors focus on wild-type clones for a second PCR screening, in the case of double gRNA approach genomic deletions. Such clones are usually ignored if a full KO model is the objective. Therefore, the significance of the findings and the impact of the research should be carefully reconsidered and presented.

We apologize if we were misleading in our presentation. Our intention is to raise awareness that inverted re-insertions happen irrespective of cell lines and target loci. The point we wanted to raise is that such events have the potential to critically affect interpretation of scientific data. This will affect generation of clonal KO cell lines in three ways, and in most aspects has already been addressed in our answer above. “WT” cell lines treated with CRISPR/Cas9 are considered the “perfect” control for comparison of genetically modified cell lines, since any potential phenotype apparent due to off-targeting is taken into account. Inverted re-insertions can lead to a functional HET allelic setup, and, when mis-annotated as WT lines, can severely flaw interpretation of data. The same point applies for isolation and analysis of heterozygously deleted lines that could turn out to be full KO (if inverted re-insertion occurred).

Additionally, generation of gene knockouts in tri- or tetraploid lines is difficult to achieve. Our aim is to raise awareness of the possibility that cell lines genotyped as HET could contain undetected inverted re-insertions (and as such could be contributing to the pool of full KO alleles). Such an event can be detected by a simple and cost-effective screen and has the potential to significantly lower the amount of cell lines that need to be screened. We decided to re-phrase our conclusions’ part (i.e. the significance of our findings) to make sure our aims and considerations are described more accurately (lines 352 to 361 in the revised manuscript, highlighted in green).