Target-specific precision of CRISPR-mediated genome editing

Background

Cas9 is an RNA-guided DNA endonuclease that operates in the CRISPR (clustered regularly interspaced short palindromic repeats) bacterial adaptive immune mechanism. In these bacteria, short lengths of DNA from plasmids or bacteriophages are transcribed to CRISPR RNAs (crRNAs) which then provide the specificity for the Cas9 endonuclease to destroy the invading pathogen. Cas9 recognition of foreign DNA relies on crRNA, which contains a 20-nucleotide recognition region known as a protospacer, and a tracrRNA which hybridises with the crRNA. The role of the crRNA-tracrRNA complex can be performed by a single guide RNA (sgRNA) and this strategy is now widely used for genome editing in mammalian cells. The Cas9-sgRNA complex binds to homologous genomic DNA where aprotospacer adjacent motif (PAM) sequence (e.g. NGG for Cas9 from S. pyogenes) is present downstream of the target sequence. Cas9 induces a double-stranded break that is then repaired by endogenous DNA repair processes and can be further manipulated by inclusion of repair templates.

When no template for repair is provided, double-stranded breaks induced by Cas9 are repaired using error-prone repair pathways. These pathways introduce frameshift insertions or deletions (indwells) that disrupt the open reading frame and generate non-functional proteins, phenocopying gene knockout. A previous paper showed that the outcome of these repair events is not random and disruption of specific target sites can have a preferred repair outcome.

A current focus of research is finding ways to improve the design of sgRNAs to successfully target the locus of interest while minimizing off-target events elsewhere in the genome. In silico prediction tools now allow researchers apply our existing knowledge of the sequence patterns that correlate with high efficiency sgRNA activity but these remain imperfect.

Figure 1A: Experimental Design

Key findings

In their preprint, Chakrabarti and colleagues examine the pattern of indels generated during CRISPR-Cas9-mediated gene editing in the absence of a repair template. The authors assessed repair of 1492 target sites in 450 genes in HepG2 cells using a pooled lentiviral library of sgRNAs predicted to have high activity and confirm that the outcome of editing is non-random in biological replicates. Single nucleotide (nt) indels occurred most frequently but there was a long tail in the length of indels and the preferred indel length for some target sites was as long as 56 nt. Almost 90% of indels produced frameshift but some target sites showed in-frame indel preference, suggesting that they should be avoided for gene KO studies. This, along with the observation that multiple sgRNAs seemed to have lower activity than predicted suggests that prediction algorithms can be further refined through better understanding of the activity of Cas9 in human cells. The pattern of indels at different sites varied with some having one strong preference and others having little preference between dozens of possibilities; the finding that only one-fifth of target sites (‘precise targets’) have a greater than 50% probability of inducing one specific indel is significant as predicting the specific outcome of genome editing for the remaining target sites is not easy.

Further characterisation of precise targets revealed that editing at these was more efficient. Precise targets were more likely to be insertions and more likely to be a single nucleotide in length while imprecise targets favoured deletions. Microhomology around the indel appeared to be a feature of deletions, consistent with repair by the MMEJ pathway. Some single nucleotide insertions also showed a preference for a common base suggesting that the nucleotide choice is not random. The preferred base was homologous to nucleotide -4 from the PAM sequence, which is usually one nucleotide upstream of the cleavage site. Moreover, precise targets showed base preference at the -4 position: when the target has an “A” or a “T” in the -4 position, repair is likely to result in a highly recurrent insertion but when it is a “G”, deletions are more likely and repair is less predictable.

These data clarify the role of DNA sequence in repair precision after Cas9 cleavage but the failure of algorithms based solely on sequence to predict indel profiles accurately suggests that other factors might influence the indel profiles of target sites. In this regard, the preprint finds a role for chromatin structure. Addition of a HDAC inhibitor or an EZH2 inhibitor altered the indel profiles observed for the same target sites by increasing and decreasing indel formation, respectively. The extent of these changes was similar to when DNA repair pathways are pharmacologically manipulated supporting the importance of chromatin structure. Some changes in the relative frequency of specific indels at targets sites were observed but were broadly similar to the untreated conditions. That said, the authors demonstrate that for some target sites, altering the chromatin state can change the most frequent indel and favour some indels over others. Future work should address the hypothesis that there is more complex interplay between chromatin state and DNA repair pathway choice than is currently appreciated.

Conclusions

• DNA sequence features affect the indel profiles generated after CRISPR-Cas9 gene editing.
• Chromatin structure also affects indel formation and inducing histone acetylation improves the efficiency of editing.
• Establishing and targeting precise sites will maximize the likelihood of desirable outcomes in experimental and clinical applications of gene editing.

Further Reading

Taheri-Ghahfarokhi, A., Taylor, B. J. M., Nitsch, R., Lundin, A., Cavallo, A. L., Madeyski-Bengtson, K., Karlsson, F., Clausen, M., Hicks, R., Mayr, L. M. et al.(2018). Decoding non-random mutational signatures at Cas9 targeted sites. Nucleic Acids Res 46, 8417-8434.

Allen, F.R., Crepaldi, L.R., Alsinet-Armengol, C. ,Strong, A., Kleshchevnikov, V., Pietro De Angeli, P., Palenikova, P., Kosicki, M., Bassett, A.R., Harding, H. et al.(2018). Mutations generated by repair of Cas9-induced double strand breaks are predictable from surrounding sequence. bioRxiv.

Questions for Authors

Q1. The finding that the chromatin state of a particular target influences targeting efficiency is interesting as it is not obvious how you could incorporate this into existing prediction algorithms as the optimal sgRNA to knockout a gene in two different cell types is likely to be different. Do you think the indel patterns observed would be consistent in other cell lines, for example?

Q2. What are the implications for the progression of CRISPR-based technologies to the clinic if editing of the same target can vary between cell types or even between the same cell type in two patients?

Q3. Your preprint caused a stir on Twitter as it acknowledges Her Majesty Queen Elizabeth II for starting your sequencing run. Could you explain a bit more?

Tags: crispr-cas9, dna repair, gene editing

doi: https://doi.org/10.1242/prelights.4951

Read preprint (2 votes)