Comparative analysis of CRISPR off-target activity discovery tools followingex vivoediting of CD34⁺hematopoietic stem and progenitor cells

Abstract

While CRISPR-based editing most often occurs at DNA sequences with perfect homology to the guide RNA (gRNA), unintended editing can occur at highly homologous regions (i.e., off-target (OT) sites). Due to the pace at which genome editing therapies are approaching clinical applications, there is an emerging need to define effective workflows for investigating OT editing effects. A number of homology-dependent, in silico-based prediction methods and wet lab-based empirical methods exist to investigate OT editing, but few have been subjected to analytical assessment or head-to-head comparison in human primary cells using an ex vivo editing process optimized for high-fidelity gene editing. Therefore, we sought to compare publicly available in silico tools (COSMID, CCTop, and Cas-OFFinder) as well as empirical methods (CHANGE-Seq, CIRCLE-Seq, DISCOVER-Seq, GUIDE-Seq, and SITE-Seq) in the context of ex vivo hematopoietic stem and progenitor cell (HSPC) editing. To do so, we edited CD34+ HSPCs using 11 different guide RNAs (gRNAs) complexed with HiFi Cas9, then performed targeted next-generation sequencing of ~200-site panels containing a range of nominated OT sites identified by in silico and empirical methods. We identified an average of 0.45 OT sites per gRNA at an indel detection limit of 0.5%. This study confirmed the marked improvement in specificity with HiFi Cas9 compared to wild-type Cas9 without compromising on-target activity when delivered as an RNP. Additionally, all HiFi Cas9 OT sites using a standard 20nt gRNA were identified by all OT detection methods with one exception (SITE-seq did not identify an OT generated by an AAVS1 gRNA). This resulted in high sensitivity for the majority of OT nomination tools, however due to the large number of false positives called by most methods, in silico-based COSMID and empirical methods DISCOVER-Seq and GUIDE-Seq attained the highest positive predictive value. We did not find the empirical methods identified off-target sites that were not also identified by bioinformatic methods when delivered as an RNP complex. Finally, this study supports that refined bioinformatic algorithms could be developed that maintain both high sensitivity as well as positive predictive value which would enable more efficient identification of potential off-target sites without compromising a thorough examination for any given gRNA.

CRISPR nuclease off-target activity and mitigation strategies

“…In spite of these myriad strategies for detection of off-target indels, recent work has shown that ex vivo editing in HSPCs elicits very few bona fide off-target sites (<1 true off-target site per gRNA) when using clinically relevant workflows ( Cromer et al, 2022b ). Of bona fide sites, all were highly homologous to the target sequence and predicted by the majority of in silico methods included in the study.…”

“…While the degree to which tissue-of-origin could be gleaned from this approach has yet to be fully explored, the investigation of on- and off-target CRISPR activity at expected cleavage sites in cfRNA could determine whether intended or unintended genome editing results in changes to the transcriptome. Since sites of CRISPR off-target activity typically reside in intergenic regions of the genome with no known functional significance ( Cromer et al, 2022b ), it is possible that CRISPR activity will be detected in cfDNA, but not in transcribed cfRNA. This could be an important measure to assay efficacy and safety of in vivo editing immediately following therapeutic delivery as well as over time.…”

“…While loss of fitness will likely result in drop out of the cell carrying the undesired event, gains of fitness are of greater concern due to the possibility of oncogenicity. Although site-specific off-target effects are infrequent ( Cromer et al, 2022b ), in the event that they do occur, the likelihood of directly disrupting another gene is small (only 1% of the human genome is coding DNA and only 7.2% of predicted off-target sites for exon-targeting gRNAs fall in exonic regions). And while modifications to non-coding DNA sequences may alter gene expression patterns or modify elements with as-yet-unknown important functions ( ENCODE Project Consortium, 2012 ) interpreting non-coding genomic disruptions is difficult.…”

Whole genome sequencing of CCR5 CRISPR-Cas9-edited Mauritian cynomolgus macaque blastomeres reveals large-scale deletions and off-target edits

“…To assess the feasibility of detecting off-target mutations by WGS, we utilized the in silico Cas-OFFinder tool to identify off-target sites based on sequence homology to the gRNA sequences and the total number of off-target sites when allowing for up to 9 mismatches between the gRNA and off-target sequence is provided in Supplementary Table S4 . While off-target editing can occur at sites with up to 4 mismatches ( Pattanayak et al, 2013 ; Cromer et al, 2022 ), we focused on assessing the WGS data at 93 predicted off-target regions that contained 2 or 3 mismatches which occurs more commonly. De novo mutations within individual blastomeres were identified by WGS in 16 predicted off-targeted sites of which 7 were located within genes and nine were located in intergenic regions ( Table 2 ).…”

“…These results confirmed that off-target editing could be assessed by WGS in individual blastomeres, although we used a biased in silico method that relied on assessing targets with sequence homology and did not evaluate potential targets with greater than three mismatches. Additional in silico nominated targets should be evaluated to fully assess the impact of off-target editing as CRISPR-Cas9 cleavage can occur at off-target sites with up to four mismatches ( Pattanayak et al, 2013 ; Cromer et al, 2022 ). Moreover, unbiased methods that survey the whole genome without prior knowledge or prediction of sequence homology would be more informative, yet there is not a current superior method or technique for this analysis ( Chaudhari et al, 2020 ; Atkins et al, 2021 ).…”