Quantitative characterization of on-target and off-target variation induced by CRISPR+Cas9 systems at the single-cell resolution.

Isai Salas-Gonzalez1, Victor Weigman1

1: BioSkryb Genomics, Durham, NC, USA

Introduction: Genome editing using CRISPR-Cas9 has revolutionized biomedical research by enabling the modification of DNA at targeted locations. The capacity of these systems to modify specific loci is dependent on the ability of Cas9 to induce DNA cleavage at genomic sites that exhibit precise base pairing with designed single guide RNAs (sgRNA). CRISPR/Cas systems are, conceptually, targeted editing tools, but the prevalence DNA edits that occur beyond the targeted gene (off-target effects) are increasingly observed. In silico tools have been developed to identify sgRNA guide-specific genome-wide potential off-target sites. Existing limitations of single cell genome enrichment have limited ability to confirm these in vitro, urging the need of unbiased methods of detection and validation. The goal of our study was to develop a bioinformatics workflow that could identify and prioritize on- and off-target gene editing events from single-cell whole genome sequencing data.

Methods: We leveraged previously published data in which CD34+ cord blood cells and U2OS sarcoma cells were transfected with by two previously described sgRNAs (EMX1 and VEGFA) and Cas9. Mock treated and Cas9-only transfected cells served as a control. Single cells were isolated and primary template-directed amplification was used to amplify genomes from single cells. Libraries were created and deep sequencing (~20X coverage) was performed. Secondary bioinformatics analysis was performed using the WGS pipeline available in BaseJumper (BioSkryb Genomics). We called CNV events using Ginkgo, translocations were called using an ensemble approach overlapping events called by Manta and GRIDSS. In silico prediction of off-target events was done with Cas-OFFinder allowing up to 7 mismatches and 1 indel.

Results: The EMX1 sgRNA induced more on-target mutations versus VEGFA sgRNA (5/5 vs 3/5 CD34+ cells and 5/5 vs 1/5 U2OS cells). Both guides exhibited heterogeneity in their on-target induced changes. Further, we characterized ~3200 reproducible off-target indels and 18 translocations from which only ~15% could be predicted in silico. Additionally, we identified ~150 mega bases across both cell types that exhibited significant fluctuation in their ploidy when exposed to the CRISPR+Guide treatments.

Conclusions: We have developed a single-cell whole genome sequencing workflow for detecting indels, CNV, and SV alterations gene-edited cells. Overall, the data presented showcase the hidden heterogeneity of on-target modifications induced by CRISPR-Cas9 systems and the emergence of prevalent genome-wide off-target events impossible to determine with in silico approaches.