Cytosine base editor 4 but not adenine base editor generates off-target mutations in mouse embryos

Deaminase base editing, using CRISPR-Cas systems, offers a precise method for installing or correcting point mutations in genomes. This technique holds immense potential for treating genetic disorders characterized by single-nucleotide alterations, which represent the majority of known human phenotypic variants. While promising, the off-target effects of these base editors remain a critical concern, particularly in vivo. Previous studies have highlighted the high fidelity of adenine base editors (ABEs) but have revealed substantial off-target mutations induced by cytosine base editors (CBEs) like BE3. This study aims to evaluate the genome-wide fidelity of the improved CBE, BE4, and compare it to ABE using an unbiased approach in mouse embryos. Understanding the off-target profiles of these editors is crucial for their safe and effective application in therapeutic settings. The use of a family-based trio cohort in whole-genome sequencing allows for the precise identification of de novo mutations introduced by the base editors, providing a more rigorous assessment of their fidelity compared to previous studies that may have been confounded by population averaging of heterogeneous editing in vitro.

The development of base editors, such as ABE and CBE, represents a significant advancement in gene editing technology, offering a more precise and less disruptive method compared to traditional CRISPR-Cas9 systems that introduce double-stranded DNA breaks. Early studies demonstrated the remarkable potential of ABEs for high-fidelity editing, although unexpected C-to-T conversions at target sites were reported. Conversely, CBEs, initially exemplified by BE3, were shown to induce a considerable number of off-target base changes throughout the genome. The development of BE4, incorporating improvements such as a longer linker and a uracil DNA glycosylase inhibitor (UGI), aimed to enhance the specificity of CBEs. Previous studies using whole-genome sequencing (WGS) in rice and mouse embryos had suggested varying degrees of off-target effects for different base editors, highlighting the need for further investigation and optimization of these tools for in vivo applications. These studies also emphasized the importance of analyzing individual edited genomes rather than relying solely on population-averaged data from in vitro experiments.

This study employed a family-based trio cohort design, involving whole-genome sequencing of parents and their offspring. A total of 84 mice (one male and eight females, and their progeny) were sequenced, encompassing 22 ABE-edited founders, 16 ABE-edited offspring, and 21 non-injected controls. The same sgRNA targeting the Wap gene was used for both ABE and BE4 editing. Injected embryos were generated via in vitro fertilization (IVF) and microinjection of mRNA encoding either ABE or BE4 along with the sgRNA. Following implantation into pseudopregnant foster mothers, the resulting offspring were genotyped and subjected to WGS. WGS data (600x coverage) were analyzed using the GATK best practices pipeline for germline mutations, incorporating stringent filtering steps to minimize false positives in identifying single nucleotide variants (SNVs) and insertions/deletions (indels). These filtering steps included considerations for read depth, allele frequency, proximity to indels, and overlapping with repetitive elements. Complex indels were analyzed using Lumpy. Statistical analysis using R software was performed to compare the mutation frequencies between the ABE-edited, BE4-edited, and control groups. Off-target analyses were also performed using CRISPOR to predict potential off-target sites based on the sgRNA sequence.

The study's key finding is the significant difference in off-target mutation rates between ABE and BE4. BE4-edited mice exhibited a considerably higher number of de novo SNVs and indels compared to ABE-edited mice and non-injected controls. This increase in off-target mutations was statistically significant. Specifically, BE4 caused unintended proximal off-target mutations and deletions in four out of nine founders, while none of the 13 ABE-edited embryos with the same sgRNA showed such effects. While some rare SNVs coincided with potential in silico-predicted off-target sites, the majority of off-target events appear independent of the sgRNA used. The analysis clearly showed that BE4 had a substantially higher frequency of de novo mutations than ABE. Furthermore, the study highlighted the importance of analyzing individual genomes instead of pooled cell populations, as this approach can mask the presence of off-target mutations in a subset of cells.

The results demonstrate a stark contrast in fidelity between ABE and BE4 in mouse embryos. The significantly higher incidence of off-target SNVs and indels in BE4-edited mice raises serious concerns regarding its use, especially in therapeutic applications. These off-target effects could lead to unintended consequences, potentially masking or confounding the intended effects of the editing. The high fidelity of ABE, while encouraging, still warrants caution due to the observed rare but notable C-to-T conversions at target sites. The finding that most off-target events appear independent of the sgRNA suggests inherent differences in the editing mechanisms of ABE and BE4, and it underscores the importance of careful consideration of potential off-target effects during the design and application of base editing strategies. This study reinforces the importance of rigorous genome-wide analysis to assess the fidelity of base editors, especially for in vivo applications where the potential for unintended consequences is magnified.

This study demonstrates that ABE exhibits significantly higher fidelity than BE4 in mouse embryos, highlighting the need for further optimization of CBEs to reduce off-target effects. While ABE shows promise for therapeutic applications, the occurrence of rare aberrant C-to-T conversions requires ongoing attention. Future research should focus on further refining CBE technology to improve its specificity and minimize off-target effects while maintaining efficacy. The study emphasizes the crucial role of individual genome analysis in assessing the safety and efficacy of base editing.

The study's focus on a single sgRNA limits the generalizability of the findings to other target sequences. While stringent filtering steps were employed to minimize false positives, the possibility of residual undetected off-target mutations remains. The relatively small sample size of BE4-edited mice might limit the statistical power of certain analyses.