High levels of AAV vector integration into CRISPR-induced DNA breaks

Using next-generation sequencing, we show that adeno-associated virus (AAV) DNA can cut-and-paste into CRISPR-induced breaks in the host genome.

In the seven years since CRISPR was brought to the forefront of molecular biology[1], it has emerged as one of the most widely used (and hotly anticipated) tools of genome editing and gene therapy, with human trials now beginning (clinicaltrials.gov: NCT03872479). CRISPR is a system for cutting DNA at specific sites. It utilizes a short guide RNA (gRNA) complementary to the target DNA region, and an enzyme (usually Cas9) that uses that gRNA to home in on the matching region of DNA and cut it. This allows for precise disruption of genes, including harmful mutations for therapeutic purposes.

In order to deliver this to target cells efficiently, many researchers use adeno-associated virus (AAV). This small virus has become the most widely used vector for delivering gene therapy, with three AAV-based therapies now on the market. It allows for the safe and stable expression of therapeutic genes in a wide array of cells types by delivering a small payload of DNA into cells’ nuclei. AAV used for gene therapy lacks dedicated machinery to integrate into the genome, but it has been shown that AAV’s DNA cargo can integrate into double-stranded breaks (DSBs) created by other means[2]. Therapeutic AAV integrates at a frequency of <0.1%. 

“While analyzing sequencing results from mice injected with AAV vectors to CRISPR disrupt an important dominant deafness gene, Tmc1, we observed unique viral sequences in the target gene. This was surprising as we really did not expect AAV genome to be there” said Bence György, co-senior author, regarding the discovery that spurred this project into being. “However, it took years for us to properly understand and characterize AAV integration. Particularly challenging was the description of genome wide AAV integration due to the difficulty of identifying true fusion reads between AAV and genomic sites. An analysis performed by Killian Hanlon revealed the true extent of AAV genome-wide integration.”

 A multidisciplinary effort, we collaborated with experts in the brain, inner ear and muscle to examine AAV integration in multiple organs. Surprisingly, we found that, in vitro and in vivo in mice, AAV can stably integrate into on-target CRISPR cut sites with up to 47% efficiency (that is, up to half of CRISPR-mediated DSBs result in AAV integration). We then ran extensive analyses to determine if this activity caused an increase in genome-wide integration of AAV. Importantly, this was not found to be the case – AAV integration rates were only heightened at the CRISPR cut site, and controls lacking a gRNA showed no integration increase over baseline. 


Of these integrations, most were small, consisting of only a handful of bases from AAV inverted terminal repeats (ITRs, repetitive regions of DNA that define the borders of the AAV vector). However, a significant minority of these were found to be much longer (>50 bases). We designed an AAV carrying a miniature genome (450bp, ~10% of AAV’s max capacity), which enabled us to determine exactly how much of the AAV vector was integrating into the genome. We saw that in some cases the entire AAV genome is capable of integrating, and that all observed cases of integration featured at least one ITR, which we deemed may be a necessary requirement. This suggested a hypothesis of integration (see Fig. 1): after a CRISPR-induced DSB, a broken ITR gets captured by the genome; then, the vector is chewed back in some fashion, possibly by exonucleases; finally, repair enzymes facilitate the complete integration of the vector. More work is required to understand this mechanism in detail!

Figure 1: Hypothesis of AAV vector DNA integration after CRISPR-mediated DNA cleavage.

The study of CRISPR – and how it interacts with AAV – is of increasing importance as the technology is moved towards the clinic. What we have demonstrated here is that, in all organs tested (brain, muscle and inner ear) AAV can stably integrate into CRISPR cut sites with high affinity, and that anywhere from single bases from entire genomes can be integrated. Critically, this increase is not reflected at off-target sites (as long as gRNAs are well-designed), and so does not indicate a significant safety risk.

Written by Killian Hanlon, Casey Maguire and Bence György

Link to the paper: https://doi.org/10.1038/s41467-019-12449-2

Please sign in or register for FREE

If you are a registered user on Nature Portfolio Bioengineering Community, please sign in