Deep learning models incorporating endogenous factors beyond DNA sequences improve the prediction accuracy of base editing outcomes.

Adenine base editors (ABEs) and cytosine base editors (CBEs) enable the single nucleotide editing of targeted DNA sites avoiding generation of double strand breaks, however, the genomic features that influence the outcomes of base editing in vivo still remain to be characterized. High-throughput datasets from lentiviral integrated libraries were used to investigate the sequence features affecting base editing outcomes, but the effects of endogenous factors beyond the DNA sequences are still largely unknown. Here the base editing outcomes of ABE and CBE were evaluated in mammalian cells for 5012 endogenous genomic sites and 11,868 genome-integrated target sequences, with 4654 genomic sites sharing the same target sequences. The comparative analyses revealed that the editing outcomes of ABE and CBE at endogenous sites were substantially different from those obtained using genome-integrated sequences. We found that the base editing efficiency at endogenous target sites of both ABE and CBE was influenced by endogenous factors, including epigenetic modifications and transcriptional activity. A deep-learning algorithm referred as BE_Endo, was developed based on the endogenous factors and sequence information from our genomic datasets, and it yielded unprecedented accuracy in predicting the base editing outcomes. These findings along with the developed computational algorithms may facilitate future application of BEs for scientific research and clinical gene therapy.

Undetectable off-target effects induced by FokI catalytic domain in mouse embryos.

The FokI catalytic domain can be fused to various DNA binding architectures to improve the precision of genome editing tools. However, evaluation of off-target effects is essential for developing these tools. We use Genome-wide Off-target analysis by Two-cell embryo Injection (GOTI) to detect low-frequency off-target editing events in mouse embryos injected with FokI-based architectures. Specifically, we test FokI-heterodimers fused with TALENs, FokI homodimers fused with RYdCas9, or FokI catalytic domains alone resulting in no significant off-target effects. These FokI genome editing systems exhibit undetectable off-target effects in mouse embryos, supporting the further development of these systems for clinical applications.

Cytosine base editors induce off-target mutations and adverse phenotypic effects in transgenic mice.

Base editors have been reported to induce off-target mutations in cultured cells, mouse embryos and rice, but their long-term effects in vivo remain unknown. Here, we develop a Systematic evaluation Approach For gene Editing tools by Transgenic mIce (SAFETI), and evaluate the off-target effects of BE3, high fidelity version of CBE (YE1-BE3-FNLS) and ABE (ABE7.10F148A) in ~400 transgenic mice over 15 months. Whole-genome sequence analysis reveals BE3 expression generated de novo mutations in the offspring of transgenic mice. RNA-seq analysis reveals both BE3 and YE1-BE3-FNLS induce transcriptome-wide SNVs, and the numbers of RNA SNVs are positively correlated with CBE expression levels across various tissues. By contrast, ABE7.10F148A shows no detectable off-target DNA or RNA SNVs. Notably, we observe abnormal phenotypes including obesity and developmental delay in mice with permanent genomic BE3 overexpression during long-time monitoring, elucidating a potentially overlooked aspect of side effects of BE3 in vivo.

Loss of function of CMPK2 causes mitochondria deficiency and brain calcification.

Brain calcification is a critical aging-associated pathology and can cause multifaceted neurological symptoms. Cerebral phosphate homeostasis dysregulation, blood-brain barrier defects, and immune dysregulation have been implicated as major pathological processes in familial brain calcification (FBC). Here, we analyzed two brain calcification families and identified calcification co-segregated biallelic variants in the CMPK2 gene that disrupt mitochondrial functions. Transcriptome analysis of peripheral blood mononuclear cells (PBMCs) isolated from these patients showed impaired mitochondria-associated metabolism pathways. In situ hybridization and single-cell RNA sequencing revealed robust Cmpk2 expression in neurons and vascular endothelial cells (vECs), two cell types with high energy expenditure in the brain. The neurons in Cmpk2-knockout (KO) mice have fewer mitochondrial DNA copies, down-regulated mitochondrial proteins, reduced ATP production, and elevated intracellular inorganic phosphate (Pi) level, recapitulating the mitochondrial dysfunction observed in the PBMCs isolated from the FBC patients. Morphologically, the cristae architecture of the Cmpk2-KO murine neurons was also impaired. Notably, calcification developed in a progressive manner in the homozygous Cmpk2-KO mice thalamus region as well as in the Cmpk2-knock-in mice bearing the patient mutation, thus phenocopying the calcification pathology observed in the patients. Together, our study identifies biallelic variants of CMPK2 as novel genetic factors for FBC; and demonstrates how CMPK2 deficiency alters mitochondrial structures and functions, thereby highlighting the mitochondria dysregulation as a critical pathogenic mechanism underlying brain calcification.

Optimization of the base editor BE4max in chicken somatic cells.

Advanced animal reproductive and breeding biotechnology has made it possible to alter traits or create new genetic resources by the direct knock-in or knock-out of target genes. Base editing technology can achieve single-base mutations without double-stranded DNA breaks, and is a promising tool for use in the genetic modification and breeding of livestock. However, the application of base editors (BEs) in chicken has not been optimized. We evaluated the efficacy of BE4max in chicken somatic cells (DF-1). The key element of BE4max, cytosine deaminase (APOBEC), was optimized for chicken. The base editing efficiency of the optimized chBE4max editor, compared with the original BE4max editor, was improved by 10.4% ± 4.6. By inhibiting the expression of the uracil DNA glycosylase-related gene methyl binding domain protein 4 (MBD4) by siRNA in chicken DF-1 cells, the editing efficiency was enhanced by 4.43% ± 1.4 compared to the control. These results suggest that this editor may have applications in poultry breeding studies.

Optimization of C-to-G base editors with sequence context preference predictable by machine learning methods.

Efficient and precise base editors (BEs) for C-to-G transversion are highly desirable. However, the sequence context affecting editing outcome largely remains unclear. Here we report engineered C-to-G BEs of high efficiency and fidelity, with the sequence context predictable via machine-learning methods. By changing the species origin and relative position of uracil-DNA glycosylase and deaminase, together with codon optimization, we obtain optimized C-to-G BEs (OPTI-CGBEs) for efficient C-to-G transversion. The motif preference of OPTI-CGBEs for editing 100 endogenous sites is determined in HEK293T cells. Using a sgRNA library comprising 41,388 sequences, we develop a deep-learning model that accurately predicts the OPTI-CGBE editing outcome for targeted sites with specific sequence context. These OPTI-CGBEs are further shown to be capable of efficient base editing in mouse embryos for generating Tyr-edited offspring. Thus, these engineered CGBEs are useful for efficient and precise base editing, with outcome predictable based on sequence context of targeted sites.