Off-target effects in CRISPR-Cas genome editing for human therapeutics: Progress and challenges

Abstract

Targeted nucleases, primarily CRISPR-Cas-based systems, have revolutionized genome editing by enabling precise modification of target genes or transcripts. Many pre-clinical and clinical studies leverage this technology to develop treatments for human diseases; however, substantial off-target genotoxicity concerns delay its clinical translation. Despite the development of a wide array of tools, assays, and technologies aimed at identifying and quantifying off-target effects, the absence of standardized guidelines leads to inconsistent practices across studies. This review highlights the key challenges and potential solutions in ensuring the safety of gene editing studies for therapeutic applications, focusing on gRNA design, off-target sites prediction, and off-target activity measurement.

Keywords: CRISPR-Cas9; MT: Clinical Applications; gene therapy; genome editing; off-target; safety.

Graphical abstract

Figure 1

CRISPR-Cas DNA editors General structure of Cas9 and Cas12a nucleases, BEs and PEs, either with or without serine integrase domain fusion. DD, deaminase domain; nCas, nicking Cas; PAM, protospacer adjacent motif; RT, reverse transcriptase; SI, serine integrase; UGI, uracyl-glycosilase inhibitor; Y, A-or-C bases. Created with BioRender.com.

Figure 2

DSB repair pathways and small-molecule inhibitors The repair of DSBs follows three major pathways: HDR, NHEJ, MMEJ. (A) In the HDR pathway, MRN and CtIP initiate DNA end resection; RAD51 then replaces RPA on single-stranded DNA, enabling homology-based strand invasion and template-directed repair. HDR inhibitors include AICAR and B02, targeting RAD52 and RAD51, respectively. (B) In the NHEJ repair pathway, Ku70/Ku80 binds to DNA ends, DNA-protein kinase catalytic subunit promotes end processing if needed, and XRCC4-DNA ligase IV complex seals the DNA break. (C) In the MMEJ pathway, PARP-1 competes with Ku70/80 for DSB binding, MRN/CtIP mediates limited resection, and POLQ aligns microhomologous sequences, facilitating error-prone synthesis and repair by the LIG3/XRCC1 complex. Created with BioRender.com.

Figure 3

Mechanisms and consequences of OTA (A) Schematic illustration of gRNA/DNA binding at an on-target site (left) and OT sites, which may involve mismatches, DNA bulges, or RNA bulges. Illustration is based on SpCas9. (B) DNA repair outcomes at on-target or OT sites and their potential consequences. Shown are insertion (INS), deletion (DEL), translocation (TRANS), inversion (INV), and large deletion (L DEL). Resulting effects include exonic frameshift mutations, disruption of intronic regulatory regions (e.g., non-coding RNAs [ncRNAs]), and gene fusions. Red lightning bolts indicate potential disruption of gene expression; the yellow caution sign denotes oncogenic potential. (C) Illustration of various strategies to mitigate OT effects, including optimized gRNA design, high-fidelity (HF) Cas variants, and alternative nucleases (nCas9, Cas12, BE, and PE). Created with BioRender.com.

Figure 4

Recommended workflow for CRISPR-based genome editing and genomic safety characterization (A) Design phase: personalized gRNA design using in silico prediction tools (e.g., CRISPRme, Cas-OFFinder, COSMID, CRISPRon/off), selection of high-fidelity or engineered Cas variants, and optimization of delivery strategy (e.g., RNP, lipid nanoparticle, viral or non-viral vectors) for in vivo or ex vivo editing. (B) Preclinical OT nomination and validation: potential OTs are predicted (CRISPRme, Cas-OFFinder, CRISPRoff) and experimentally validated via orthogonal assays (e.g., CIRCLE-seq, Digenome-seq, GUIDE-seq, DISCOVER-seq). Quantification is performed by targeted deep sequencing (e.g., rhAmpSeq) and analyzed with tools like CRISPECTOR or CRISPResso2. (C) SVs and genomic rearrangement assessment: monitoring of large deletions, translocations, and other rearrangements using designated methods such as CAST-seq, PEM-seq, ddPCR, FISH, long-read sequencing. (D) Functional and safety validation: edited cells are evaluated via in vitro assays (cytotoxicity, tumorigenicity, differentiation, lineage tracking) and/or in vivo xenotransplantation into humanized mouse models to evaluate long-term safety, engraftment, and functional performance. (E) Regulatory and manufacturing considerations: all analyses are conducted under GMP conditions, using orthogonal methodologies aligned with FDA/EMA guidelines, with traceable data archiving for potential retrospective safety review. RNP, ribonucleoprotein; EMA, European Medicines Agency; FDA, US Food and Drug Administration. Created with BioRender.com.