CERSI Collaborator: University of California San Francisco: Matthew Porteus, MD, PhD; M. Kyle Cromer, PhD

FDA Collaborators: Center for Biologics Evaluation and Research: Zhaohui Ye, PhD

Project Start Date: 08/2018
Project End Date: 02/2023

Regulatory Science Challenge

With the advent of CRISPR-based genome editing, researchers can create a new class of medicines, ones that correct the genetic basis of disease at the DNA level. The advantage of CRISPR over prior technology is the ability to precisely and easily edit DNA at an intended genomic locus, or specific location. However, the introduction of CRISPR editing tools applied to living cells carries with it the possibility of editing similar but unintended sites in the genome, some of which could have functional significance. To address this, a variety of bioinformatic and biologic off-target activity prediction tools have been developed. In spite of tremendous potential of these tools to treat currently intractable disorders, there has not been a thorough comparison of these off-target detection tools in their ability to identify sites that harbor true off-target activity.

Project Description and Goals

To address the current scientific need, we have performed targeted deep sequencing of a large number of potential off-target sites which were identified for CRISPR guide RNAs targeting a total of 11 different genomic regions. In doing so, researchers will be able to assess the sensitivity and specificity of the most prominent bioinformatic and biologic off-target detection tools that are currently available, which will help scientists at all stages (from bench to bedside) determine the most appropriate methodology necessary to discover true sites of CRISPR off-target activity.

Research Outcomes/Results

Researchers performed deep sequencing of the most informative potential off-target sites for nine established CRISPR guide RNAs. The major findings of this study are:

• Delivering Cas9 in a clinically relevant fashion (transiently in human primary cells) is associated with little off-target activity (9 off-target sites out of over 2,000 sites assayed)
• High-fidelity Cas9 dramatically reduces off-target activity compared to wild-type Cas9
• For editing conditions using high-fidelity Cas9, all off-target sites were identified by all bioinformatic prediction tools, suggesting limited added value of laborious empirical methods

Research Impacts

These results highlight the inherent tradeoff between sensitivity and specificity for any given off-target prediction method. Overall, each research group should consider feasibility of sequencing a manageable list of mutations (likely on the order of tens rather than hundreds or thousands of candidate sites). Each group must also consider the feasibility of replicating a particular empirical method, especially when bioinformatic prediction tools identified all detectable off-target sites when using high-fidelity genome editing tools. Overall, we believe these data can be used to help adjust bioinformatic methods to improve sensitivity and specificity and offer practical recommendations for scientists at all stages of clinical development.