Get Help Sign In

Nominating off-target sites for CRISPR-Cas9 genome editing experiments

Chaudhari HG, Penterman J, Whitton HJ, et al. Evaluation of homology-independent CRISPR-Cas9 off-target assessment methods. The CRISPR Journal. 2020;3(6):440–453.

After a CRISPR genome editing experiment is performed, editing sites must be identified and analyzed. This study compares three commonly used off-target site nomination methods and provides tips and recommendations to achieve reproducible, reliable results.


CRISPR genome editing uses targeted guide RNA (gRNA) that directs the Cas9-RNA complex to a genomic DNA locus. After a CRISPR genome editing experiment has been performed, it is important to determine if the intended edits were made and if there are unwanted edits at off-target locations. This is particularly important when a CRISPR-Cas9 genome editing approach is considered for therapeutic applications.  Next generation sequencing (NGS) is the recommended method for full investigation of CRISPR edits, but sequencing the entire genome (whole genome sequencing, WGS) can be cost-prohibitive and time-consuming. Thus, targeted amplicon sequencing became an effective alternative to whole genome sequencing, since it enables sequencing of specific areas of the genome more rapidly and cost-effectively.

To perform targeted NGS, suspected off-target CRISPR editing sites must be nominated for sequencing. In this study, Chaudhari and colleagues compare the performance of three commonly used methods for nominating off-target edit (OTE) sites: GUIDE-seq, CIRCLE-seq, and SITE-seq.


A CRISPR genome editing experiment was performed in HEK293T/Cas9 cells. Following the CRISPR experiment, genomic DNA was extracted from the cells, and GUIDE-seq, CIRCLE-seq, and SITE-seq assays were performed in triplicate. Targeted amplicon libraries were generated using rhAmpSeq™ CRISPR Panels and the rhAmpSeq CRISPR Library Kit. Libraries were then sequenced on Illumina platforms. The results were compared to in silico nomination methods using the CCTop algorithm and validated with hybridization capture. The researchers compared more than 75,000 nominated off-target sites for eight gRNAs to genome-wide assays to evaluate false-negative nomination rate.

Hybridization capture was used for validation because it can accurately quantify large indels using long probes, and, like multiplexed amplicon sequencing, it has a high correlation between indel frequency differences.


Of the three methods, CIRCLE-seq nominated the most OTE sites, and GUIDE-seq nominated the fewest. CIRCLE-seq and SITE-seq had the most overlap, likely due to the use of genomic DNA as starting material, compared to GUIDE-seq, which relies on tag integration in live cells.

OTE sites nominated by more than one assay had lower edit distance, or the distance between the editing target and the OTE, than single-assay-nominated sites, suggesting a higher likelihood of being true gRNA-dependent sites. GUIDE-seq was the least reproducible, with high variation across the three replicates. SITE-seq had the highest level of reproducibility, but it was gRNA-dependent. SITE-seq also had the highest percentage of OTE sites with reads as high as on-target reads, suggesting a lack of correlation with cellular Cas9 activity. This suggests that SITE-seq may result in more false-negative OTEs.

In terms of precision, GUIDE-seq had the fewest false positives and CIRCLE-seq had the most. GUIDE-seq and CIRCLE-seq correlated read count and indel frequencies, while SITE-seq did not. Editing sites that were missed had very low indel frequencies.

The authors recommended GUIDE-seq in situations where the edited cell type was readily available and CIRCLE-seq if it was not. Due to the lack of read count and indel frequency, SITE-seq was not recommended.

The authors suggested the following to improve on-target editing:

  1. Use computational predictors when designing your gRNA. IDT offers the Alt-R® CRISPR-Cas9 guide RNA Design Tool.
  2. Use both computational and empirical methods to nominate potential editing sites. GUIDE-seq is recommended for ex vivo and CIRCLE-seq is recommended for in vivo experiments.
  3. Use multiple replicates, and sequence to a depth of at least 1000 reads.
  4. Evaluate nominated sites based on rational selection criteria rather than choosing an arbitrary number of sites to evaluate.
  5. Use appropriate statistical methods.
  6. Recognize areas of potential sequence artifacts.
  7. Analyze indel patterns to confirm they are CRISPR-mediated.
  8. Consider the context of the edits in relation to gene function.

Chaudhari et al. conclude their report by recommending performance assessment of target sites nominated by different methods (homology-dependent and homology-independent) in addition to adoption of the above practices for obtaining more accurate results. 


Published Jan 27, 2021