Base Editing Safety Assessment Solution-1

Base Editing Safety Assessment Solution

1. Industry Background and Regulatory Requirements

Base Editing technology represents a major breakthrough in the gene therapy field following CRISPR/Cas9. Unlike traditional nuclease-dependent gene editing, base editors achieve direct chemical conversion of single bases at precise genomic locations without creating double-strand breaks (DSB) by fusing deaminase with deactivated or nickase Cas9 (nCas9). Current mainstream base editing systems include Cytosine Base Editors (CBE) and Adenine Base Editors (ABE): CBE can convert C·G base pairs to T·A, while ABE can convert A·T base pairs to G·C, covering over 60% of known pathogenic point mutation sites clinically, demonstrating tremendous clinical translation potential in genetic disease treatment and tumor immunotherapy cell modification.

However, the off-target effects of base editing systems possess unique complexity, presenting challenges far exceeding traditional CRISPR tools for safety assessment. Base editing off-target effects can be divided into two major categories: First is sgRNA-dependent off-targeting, triggered by Cas9 mismatch binding at homologous genome sequences, similar to traditional CRISPR off-target mechanisms and predictable; Second is sgRNA-independent off-targeting, a risk unique to base editing—the deaminase domain in CBE and ABE can randomly bind single-stranded DNA (such as genome DNA exposed at replication forks) or RNA, causing genome-wide random base conversions and transcriptome-level RNA editing. This type of off-targeting cannot be predicted through sequence homology and represents a key safety concern for regulatory agencies.

The FDA in "Human Gene Therapy Products Incorporating Human Genome Editing" (2024) and the "Technical Guiding Principles for Non-clinical Research and Evaluation of Gene Therapy Products (Trial)" explicitly require comprehensive off-target risk assessment for gene editing products, emphasizing the use of multiple methods for genome-wide analysis to reduce bias in identifying potential off-target sites. We employ a multi-level, multi-dimensional comprehensive detection strategy, using donor samples from multiple sources and establishing scientific experimental groupings. This strategy can provide sufficient scientific evidence for preclinical safety assessment of gene editing products, ensuring IND application materials meet regulatory requirements and support smooth clinical trial initiation.

2. Industry Pain Points Analysis

Safety assessment of base editing products faces multi-level technical and operational challenges in actual execution. Compared with traditional CRISPR products, its unique dual off-target mechanism makes the construction of assessment systems more complex.

3. Comprehensive Safety Assessment Solution

Our base editing safety assessment solution is supported by three core technical platforms. First is the DNA-level off-target detection platform, which employs software prediction, WGS, Detect-seq (CBE editing), and Ino-seq (ABE editing) technologies for the different off-target mechanisms of CBE and ABE, achieving comprehensive detection coverage of sgRNA-dependent and sgRNA-independent off-targeting. Second is the RNA-level off-target detection platform, which captures random RNA editing events triggered by base editor deaminases at the transcriptome level through transcriptome sequencing (RNA-seq) specialized analysis workflows, filling the blind spots of DNA detection. Finally, the off-target site validation platform combines targeted enrichment sequencing (TES) with deep amplicon sequencing for stepwise validation, performing high-sensitivity quantitative validation of candidate off-target sites, forming a complete evidence chain meeting regulatory requirements.

3.1 Overall Technical Roadmap

[Image placeholder - insert base editing safety evaluation flowchart]

Figure 1. Base Editing Safety Assessment Comprehensive Technical Roadmap

3.2 Core Service Module Details

Module 1: Specialized Software Prediction

Software prediction is the first step in off-target assessment and an important prerequisite for experimental detection. Unlike general CRISPR prediction tools, we employ dedicated deep learning prediction models (CBEdeepoff / ABEdeepoff) for CBE and ABE respectively, comprehensively considering factors such as sequence mismatches, base editing preferences, and chromatin accessibility. We perform risk grading of sgRNA-dependent off-target sites across the genome, outputting candidate site lists.

Module 2: 100x WGS Off-target Detection

WGS can scan the entire genome without bias and is one of the fundamental methods required by regulatory agencies for off-target assessment. It should be noted that WGS itself has limited detection sensitivity and cannot serve alone as a sufficient means for base editing off-target assessment. It must be used in combination with high-sensitivity dedicated detection methods such as Detect-seq and Ino-seq to meet FDA and CDE regulatory requirements for multi-dimensional orthogonal validation.

We employ ≥100X depth for simultaneous library construction and sequencing of edited and control samples. After quality control filtering, we use GATK + Freebayes + VarScan three-tool joint calling with intersection to maximize false positive elimination. Subsequently, we overlay BE-specific filtering to extract CBE characteristic variations (C→T/G→A) or ABE characteristic variations (A→G/T→G) from all SNVs. Based on whether variant sites are located near sgRNA homologous sequences, results are classified as sgRNA-dependent or sgRNA-independent off-targeting, outputting high-confidence off-target site lists for integrated analysis with subsequent dedicated detection method results.

Module 3: Base Editing-Specific Off-target Detection

• Detect-seq Detection (CBE Editing)

The core principle of Detect-seq is specific capture of deoxyuridine (dU), the CBE editing intermediate. The workflow has three steps: ① Capture intermediate products—specifically label dU and enrich DNA fragments containing dU; ② Signal amplification—normal cytosines near dU are replaced with 5-formylcytosine (d5fC), which after chemical labeling converts to T. During sequencing, continuous C→T mutation signals appear within approximately 150bp range centered on dU around the editing site, forming clear differentiation from background noise; ③ Precise localization—based on continuous C→T characteristic signals combined with bioinformatics analysis, locate CBE off-target sites, covering all types of off-targeting including sgRNA-dependent, independent, editing outside protospacer regions, and target strand editing.

• Ino-seq Detection (ABE Editing)

Ino-seq achieves genome-wide off-target detection by tracking deoxyinosine (dI) produced by ABE. The core principle of this method is based on the specificity of endonuclease V (EndoV): EndoV can recognize inosine and precisely cleave DNA at the second phosphodiester bond at its 3' end.

The specific workflow is as follows: ① Extract genomic DNA from cells treated with ABE and perform mechanical fragmentation; ② Ligate adapters to fragment ends, remove DNA fragments without adapters through damage repair and exonuclease treatment; EndoV introduces specific nicks at inosine-containing sites; ③ After heat denaturation, single-strand DNA binding protein (SSB) stabilizes single-strand structure. Cleaved single strands without biotin are removed via streptavidin magnetic beads, thereby enriching inosine-containing DNA; ④ After ligating bridge adapters with unique molecular identifiers (UMI), perform PCR amplification to complete library construction; ⑤ During sequencing, inosine sites are fixed at the 2nd base downstream of the P7 sequencing primer.

Module 4: Transcriptome Off-target Detection

Detect off-target editing events triggered by base editors at the transcriptome level through whole transcriptome sequencing (RNA-seq) technology. The specific analysis workflow is: First perform high-throughput sequencing of RNA samples to obtain transcriptome data; then through rigorous bioinformatics analysis workflows including sequence quality control, alignment to reference genome, identification of base variant sites, filtering of background variations such as SNPs, ultimately identify true RNA editing sites. During analysis, distinguish editing events in Alu repeat regions from non-repeat regions, and comprehensively statistically annotate 12 types of base editing (such as A-to-G, C-to-U, etc.), thus capturing random RNA editing events caused by base editor deaminases at the whole transcriptome level, achieving unbiased detection and quantitative assessment of RNA off-target effects.

Module 5: Off-target Site Validation

Using candidate off-target sites obtained from preliminary detection as input, custom-design and synthesize probes for each site. The experimental workflow is: After fragmentation, genomic DNA undergoes dA-tailing followed by ligation of double-ended adapters containing UMI (Unique Molecular Identifiers) (UMI-L/UMI-R). After PCR amplification, perform solution-phase hybridization with probes targeting off-target sites, then capture with streptavidin magnetic beads, wash, PCR amplify, and complete targeted deep sequencing and data analysis. Introduction of UMI enables precise removal of PCR duplicates during bioinformatics analysis, effectively eliminating false positives caused by PCR amplification bias, significantly improving detection accuracy of low-frequency single-base off-target events.

For CBE, input sites come from candidate sites screened through three layers of detection: CBEdeepoff prediction, WGS, and Detect-seq (sgRNA-dependent off-targeting + sgRNA-independent off-target sites). Probe design covers corresponding C→T editing window regions, focusing on validation of cytosine editing events. For ABE, input sites come from candidate sites screened through ABEdeepoff prediction, WGS, and Ino-seq three-layer detection. Probe design covers corresponding A→G editing window regions, focusing on validation of adenine editing events. Both can simultaneously cover hundreds of candidate sites in a single experiment, achieving high-throughput preliminary screening.

4. Service Models

5. Customer Value

6. Our Advantages

• Full-Chain Technical Coverage: One of the few domestic enterprises simultaneously possessing CBE and ABE dedicated off-target detection (Detect-seq/Ino-seq), whole genome sequencing, transcriptome RNA off-target detection, and comprehensive validation method service capabilities, greatly reducing customer supplier management and communication costs. Adopt standardized ISO9001/CNAS certified processes to ensure detection result accuracy and reliability, meeting the latest regulatory requirements of FDA and CDE.

• Dedicated Technical Platform: For the unique off-target mechanisms of base editors, independently establish Detect-seq (CBE-specific) and Ino-seq (ABE-specific) detection platforms. Compared with general off-target detection methods, can more comprehensively cover all types of off-targeting including sgRNA-dependent, independent, protospacer-external editing, and target strand editing. Simultaneously equipped with RNA-level off-target detection, constructing a DNA+RNA dual-dimensional safety assessment system.

• Rich CGT Industry Experience: Have successfully served multiple leading CGT enterprises, deeply understanding regulatory requirements and technical challenges of base editing product IND filing, able to provide professional detection scheme design, data interpretation, and regulatory strategy recommendations. We not only provide detection data but also complete filing material packages meeting regulatory requirements, helping customers smoothly pass IND review.

• Flexible Service Models: Support IND filing full-process packages and individual services, can provide customized solutions based on specific customer needs and budgets. Equipped with dedicated project managers for full-process tracking, rapid response to customer needs, ensuring efficient project advancement and on-time delivery.

7. References

[1] "Human Gene Therapy Products Incorporating Human Genome Editing" 2024

[2] "Technical Guiding Principles for Pharmaceutical Research and Evaluation of In Vivo Gene Therapy Products (Trial)" 2022

[3] "Technical Guiding Principles for Non-clinical Research and Evaluation of Gene Therapy Products (Trial)" 2021

[4] Lei, Z., Meng, H., Lv, Z. et al. Detect-seq reveals out-of-protospacer editing and target-strand editing by cytosine base editors. Nat Methods 18, 643-651 (2021). https://doi.org/10.1038/s41592-021-01172-w