• Home /
  • Blog /
  • Should I be worried about CRISPR off-target effects?

Should I be worried about CRISPR off-target effects?

Jun 17, 2016 11:06:42 AM No Comments

Increasingly the literature suggests that CRISPR's potential for off-target effects is not as “bad” as originally thought. Here are a few things to set your mind at ease if you’re working with the CRISPR-Cas9 system:

1. CRISPR/Cas9 is actually much more specific than might be predicted from just sequence homology

At the end of 2014 Cencic et al used CHIP-seq to characterise genome wide binding of catalytically inactive Cas9 directed by two different sgRNA targeting the Trp53 locus.

Critically, when they then analysed the 43 bound sites for evidence of wild type Cas9 cutting activity they observed editing only at the on-target site and 1 off target site.

The paper concludes that it whilst sequence homology (particularly at the PAM proximal region) will result in binding of the sgRNA/Cas9 complex, complementarity at the PAM distal region is required for cleavage activation.

In other words, the binding and cleavage activities of Cas9 can be considered two separate events, with cleavage actually more stringent than binding.

2. CRISPR/Cas9 off target cleavage is predictable in silico

Having just told you that sequence homology cannot necessarily predict off target cleavage, I’m now going to tell you that it can. Or at least – when off target cleavage occurs it does so in a relatively predictable manner.

Three papers published recently use slightly different approaches to determine genome wide off target cleavage in an unbiased manner:

  • Tsai et al – relies on capture of double-stranded oligodeoxynucleotides into double strand break (DSBs) (the GUIDE-seq method)
  • Wang et al – in a similar fashion uses an integrase defective lentivirus to identify sites of double strand break
  • Kim et al – uses sequencing of in vitro Cas9-digested genomic DNA to identify off target sites (Digenome-seq method)

Wang et al were able to use their approach to identify instances of off target cleavage that were not predicted in silico. However, closer inspection of the sites finds that base skipping results in a significant increase in homology, and this phenomenon can now be included in off-target prediction algorithms.

This ability to predict off target sites accurately remains key to selecting the best (least promiscuous) guides, as in each paper off target cleavage can occur with the same (and in some cases higher) frequency than on target cleavage. Tsai et alobserve off target cleavage as high as 60% and Kim et al as high as 83%.

3. There is no perfect system

While the potential for off target effects with CRISPR/Cas9 should inform our experimental design, it’s important to keep in mind that no approach to studying the phenotypic effects of loss and gain of function is perfect.

For example, cancer cell lines provide a robust in vitro model system with which to study biology, but are by their very nature genetically unstable, and the attribution of phenotypic observations to genetic observations should be done cautiously, and preferably using multiple cell lines.

Similarly RNAi, widely used for modelling loss of function, also has its limitations - the potential for off target effects with siRNA or shRNA are well known, with short stretches of sequence complementarity in the seed region driving widespread silencing of unintended transcripts (Jackson et al).

How to proceed…

So what does all this mean for you, the scientist, wondering how confident you should be in the cell line you’ve just generated and the phenotypes you’re observing? Well what we can take from the above papers is that while off target cleavage does occur, it’s occurring predictably and, provided you select a guide with low predicted off-target potential, with low frequency.

With this in mind the following guidance can be followed:

  • If at all possible use the genomic sequence of the specific cell line or organism you’re targeting to predict off target sites. The reference sequence is a statistical average and may not represent accurately your cell line of choice – in other words you may be missing things if you use it.
  • Select a guide with low predicted off target potential – this probably goes without saying! You may also wish to factor into your selection where predicted off target sites are located (in coding or non-coding region), and the degree of homology therein.
  • Isolate multiple, independent targeted clones – whilst off target cleavage can occur, the likelihood of it occurring in the same place in independent clones is very low.
  • The use of a second independent guide RNA to derive an independent clone or reconstitution of a knockout clone with a wild-type cDNA can also further add confidence and will allow you to derive clear genotype/phenotype relationships.
  • Although sequencing prices are dropping every day, performing whole genome sequencing on every cell line you’ve targeted may still be cost prohibitive. Further to this, analysing a data set from a WGS run and spotting small changes made by Cas9 is not an insignificant undertaking. A suitable compromise therefore is to take the off target predictions from sgRNA design tools such as gUIDEBook, and simply sequence the sites in your targeted clones and parental for comparison.

Finally, it is worth keeping in mind that many of the concerns about CRISPR off target effects stem from a desire to use the technology in a therapeutic setting (where even a very rare off-target mutation may have a deleterious consequence). If you’re working in a non-clinical cell biology lab, then these concerns arguably become less important, and so each scientist should know which “level of concern” is appropriate for their purpose.

At Horizon we apply stringent guide selection criteria for all of projects, selecting guides as far as possible with the lowest off-target scores. Further to this, for our cell line projects we generate multiple, independent clones which are highly unlikely to contain the same off target modifications, and can be used to corroborate any phenotypic observations.

Read our tips on guide RNA selection

#Gene editing, #CRISPR