ABE/CBE Transcriptome Off-Target Editing Detection

ABE/CBE Transcriptome Off-Target Editing Detection

1. Background

Single-base editors (Base Editors) fuse deaminases with Cas9 nickases or catalytically inactive Cas proteins to achieve precise base conversion at targeted sites without inducing DNA double-strand breaks. The two most widely used classes of base editors are: Adenine Base Editors (ABE), which employ TadA deaminase to convert A to I (read as G), enabling precise A-to-G substitution; and Cytosine Base Editors (CBE), which utilize deaminases from the APOBEC/AID family to convert C to U (read as T), enabling precise C-to-T substitution. This technology has demonstrated significant value in gene therapy, disease model construction, and functional genomics research. However, the intrinsic sequence tolerance of deaminases or incomplete specificity of gRNAs may lead to "off-target editing"—unintended base modifications at non-target genomic or transcriptomic sites—posing potential safety risks and experimental interference. The transcriptomic off-target mechanism of base editors is specifically manifested as follows: the deaminases fused to CBE and ABE (e.g., APOBEC family or TadA) are biochemically capable of acting on single-stranded RNA. Their free diffusion within cells allows them to act independently of gRNA guidance, directly contacting mRNA and other single-stranded RNA substrates in the transcriptome, catalyzing unintended C→U or A→I base modifications, which can lead to amino acid substitutions, premature termination, and other functional abnormalities at the translational level. A schematic diagram is shown below:

Figure 1. Schematic diagram of the deaminase substrate action mechanism in CBE base editors

In recent years, regulatory guidelines related to gene therapy (such as China NMPA's 2022 "Technical Guidelines for Pharmaceutical Research and Evaluation of In Vivo Gene Therapy Products" and FDA's 2024 updated guidance) have emphasized the need for systematic off-target assessment strategies, encompassing bioinformatics prediction, DNA-level off-target detection, and RNA-level off-target validation, to comprehensively evaluate the specificity and risks of editing systems.

Transcriptome mRNA off-target editing detection, as an indispensable component of this framework, primarily targets the non-directed editing activities of ABE and CBE deaminases at the RNA level (A-to-I/G and C-to-U/T conversions, respectively). Through high-depth RNA sequencing and bioinformatics analysis of edited samples, one can directly assess whether predicted off-target sites exhibit aberrant editing in transcripts. The technical advantages include:

(1) Functional Directness: Detects off-target events in transcriptionally active regions, better reflecting the potential impact on gene expression and function;

(2) High Sensitivity: Capable of identifying low-frequency off-target signals (sensitivity typically reaching 0.1%), suitable for early safety screening;

(3) Mechanistic Insight: Can distinguish between sgRNA-dependent and sgRNA-independent off-target events, and differentiate the RNA off-target profiles of ABE (A-to-I/G) and CBE (C-to-U/T), providing a basis for understanding deaminase activity;

(4) Optimization Guidance: Validates predicted off-target sites and provides data support for optimizing ABE/CBE deaminase variants (e.g., high-fidelity ABEmax, evoAPOBEC), gRNA design, and delivery systems.

GeneRulor, leveraging its proprietary high-depth transcriptome sequencing and targeted analysis pipeline, has developed a systematic mRNA single-target off-target editing detection platform, designed to provide customers with regulatory-compliant, scientifically rigorous off-target risk assessment solutions that support the research, development, optimization, and safety translation of gene editing tools.

2. Detection Principle

Transcriptome mRNA single-target off-target editing detection technology is based on high-depth RNA sequencing. It targets bioinformatically predicted high-risk off-target sites to detect whether editing (e.g., base substitutions, insertions/deletions) occurs at the transcript level, thereby validating the authenticity and functional impact of predicted off-target sites.

Detection Workflow:

(1) Off-target Site Prediction: The Cas-OFFinder tool is used to predict potential off-target sites with sequences similar to the target site, prioritizing sites with ≤6 mismatches located in transcriptionally active regions.

(2) RNA Extraction and Library Construction: High-quality total RNA is extracted from edited samples, and target transcripts are enriched using oligo(dT) magnetic beads for mRNA enrichment or rRNA depletion, followed by high-depth sequencing library construction (recommended sequencing depth: ≥50M reads).

(3) High-Depth Sequencing: DNBSEQ-T7 is used for PE150 paired-end sequencing to ensure sufficient sequencing depth at predicted off-target sites (per-site coverage depth ≥1,000×).

(4) Off-target Editing Detection: Sequencing data are aligned to the reference genome; variant calling (SNV/Indel calling) is performed at predicted off-target sites; off-target editing frequencies are calculated and compared against control groups to validate the authenticity of off-target events.

Data Analysis Strategy:

(1) On-target Editing Efficiency Verification: The editing efficiency at the intended target site is first assessed to confirm successful editing. The editing frequency at the target site is calculated (e.g., C-to-T conversion rate for CBE, A-to-G conversion rate for ABE).

(2) Predicted Off-target Site Editing Detection: For each bioinformatically predicted off-target site, RNA-level editing is assessed. Off-target editing frequencies are calculated, and sites with frequencies significantly above background mutation rates are identified as true off-target sites. Based on prediction results, off-target sites are further classified as sgRNA-dependent or sgRNA-independent.

(3) Functional Impact Assessment of Off-target Editing: Genomic annotations of off-target editing sites (exonic, intronic, UTR, promoter, etc.) are analyzed; the impact of off-target editing on protein-coding sequences (synonymous, missense, and nonsense mutations) is evaluated to predict functional consequences.

(4) Tumor-related Gene Risk Assessment: Based on COSMIC, ONCOGENE, and other databases, the genes harboring off-target sites are evaluated for oncogene or tumor suppressor gene status to assess potential carcinogenic risks.

Compared to conventional Sanger sequencing or targeted amplicon sequencing, high-depth transcriptome sequencing enables simultaneous detection of dozens to hundreds of predicted off-target sites, significantly increasing detection throughput. The high sequencing depth (≥1,000× per site) enables detection of low-frequency off-target events (sensitivity as low as 0.1%), far exceeding traditional methods.

Figure 2. Workflow for base editor off-target detection based on high-depth mRNA sequencing

3. Technical Advantages

3.1 High-Throughput Coverage of Predicted Off-target Sites

(1)A single experiment can simultaneously detect dozens to hundreds of bioinformatically predicted off-target sites

(2)Covers all predicted off-target sites in transcriptionally active regions, with priority identification of functional off-target events

(3)Significantly improves detection efficiency compared to per-site Sanger sequencing or qPCR validation

3.2 High-Sensitivity Low-Frequency Off-target Detection

(1)Recommended sequencing depth ≥50M clean reads to ensure sufficient coverage at predicted off-target sites (≥1,000× per site)

(2)High-depth sequencing enables detection of low-frequency off-target editing events with sensitivity as low as 0.1%

(3)Professional variant calling algorithms (e.g., GATK, VarScan2) are employed to eliminate sequencing errors and background mutation noise

3.3 Functional Off-target Assessment

(1)Detects off-target editing in transcriptionally active regions, reflecting the actual impact of off-target effects on cellular function

(2)Assesses the impact of off-target editing on protein-coding sequences (synonymous/missense/nonsense mutations) to predict functional consequences

(3)Integrates tumor-related gene databases to evaluate the carcinogenic risk of off-target sites

4. Application Scenarios and Service Advantages

4.1 Application Scenarios

(1) IND Filing Support for Gene Editing Products: Provides transcriptome off-target editing detection reports compliant with regulatory requirements, verifying the authenticity of bioinformatically predicted off-target sites and meeting the data requirements of NMPA/FDA for single-target editing off-target safety evaluation.

(2) gRNA Design Optimization and Screening: Compares off-target editing profiles across different gRNA designs to select gRNA sequences with minimal off-target effects and optimize editing strategies.

(3) Safety Evaluation of Editing Tools: Assesses the RNA-level off-target editing profiles of different editing tools to verify their safety advantages.

(4) Preclinical Safety Research: Evaluates single-target gene editing off-target editing in cell lines, organoids, or animal models, providing safety evidence for clinical trial design.

(5) Cell Therapy Product Quality Control: Performs transcriptome off-target editing detection on engineered cell products such as CAR-T and TCR-T to ensure editing specificity and product safety.

4.2 Service Advantages

(1) Experienced Technology Platform: Having completed thousands of gene editing-related transcriptome projects, we have accumulated extensive experience in off-target editing detection, covering multiple editing tools including CRISPR/Cas9, Adenine Base Editors (ABE), Cytosine Base Editors (CBE), and Prime Editors, with particularly deep technical expertise in ABE/CBE RNA off-target detection.

(2) Dual-System Quality Certification: Simultaneously operating under ISO9001 quality management system and CNAS laboratory accreditation standards, ensuring traceable data quality.

(3) Professional Bioinformatics Team: Staffed with experienced bioinformatics analysts providing customized off-target site prediction and analysis solutions, including editing type-specific off-target detection strategies.

(4) Regulatory Submission Support: Provides standardized reports compliant with ICH, NMPA, and FDA requirements, directly supporting IND/BLA submissions.

(5) Integrated One-stop Service: Can integrate off-target site prediction, DNA-level off-target detection, and RNA-level off-target editing detection, providing multi-dimensional off-target assessment.

5. Example Report of Transcriptome Single-Target Off-Target Editing Detection

GeneRulor provides comprehensive transcriptome off-target editing detection reports that meet regulatory requirements. The following example, based on a single-target base editor (CBE) editing project, illustrates the core content of the report (reports for other types of single-target editing projects follow a similar structure):

5.1 Data Quality Assessment

Sequencing data statistics, base quality distribution, GC content analysis, and alignment rate assessment are provided to ensure data quality meets analytical requirements. This includes pre-QC and post-QC data tables. Verification of whether sequencing depth meets the requirements for predicted off-target site detection (≥50M reads; average coverage depth at predicted off-target sites ≥1,000×) is included.

Example raw sequencing data

Example QC and alignment data

Example sequencing depth data

5.2 On-target Editing Efficiency Verification

The editing efficiency at the intended target site is assessed to confirm successful editing. CBE editing example: The C-to-T conversion frequency at the target site is calculated, and an IGV visualization of target site editing efficiency is provided. The editing profile of the surrounding sequence is shown (e.g., editing window analysis when multiple C bases are present).

Example on-target site detection results

Figure 3. IGV visualization of sequencing coverage and C-to-U editing events at the CBE-edited on-target site

5.3 Off-target Site Editing Detection Results

For each off-target site, detailed editing detection results are provided: sequencing coverage depth, editing frequency (edited reads / total reads), and background mutation rate in the control group. CBE editing example: The C-to-U conversion frequency at each predicted off-target site is assessed, and true off-target sites significantly exceeding background levels are identified. Bar charts of off-target editing frequency and distribution maps of off-target sites are provided.

Example off-target site detection results

Figure 4. Number of 12 types of base editing events

Figure 5. Manhattan plot of editing efficiency distribution at editing sites

5.4 Functional Impact Analysis of Off-target Sites

The functional impact of detected off-target editing sites is analyzed: genomic annotation (exonic/intronic/UTR/promoter/intergenic regions), impact on protein-coding sequences (synonymous/missense/nonsense/frameshift mutations), and protein functional domain impact analysis. CBE editing example: For C-to-U edits in exonic regions, the predicted functional consequences are assessed.

Example functional impact results of off-target sites

Figure 6. Functional annotation of editing sites

5.5 Predicted Off-target Site List

A complete list of all bioinformatically predicted off-target sites is provided, including: genomic coordinates, mismatch count, mismatch positions, gene information, and genomic annotations (exonic/intronic/UTR/promoter, etc.). CBE editing example: Predicted off-target sites containing C bases are listed with priority.

Example predicted off-target site results

Based on combined analysis of predicted and detected off-target sites, off-target sites are further classified as sgRNA-dependent or sgRNA-independent.

5.6 Tumor-related Gene Risk Assessment

Based on COSMIC, ONCOGENE, and other databases, the genes harboring off-target sites are evaluated for oncogene or tumor suppressor gene status. A list of tumor-related genes and their off-target editing status is provided, with emphasis on high-risk off-target events (e.g., nonsense mutations in tumor suppressor genes, activating mutations in oncogenes).

5.7 Off-target Editing Summary and Risk Assessment

A summary table of off-target editing detection is provided: total number of predicted off-target sites evaluated, number of sites with detected off-target editing, number of high-frequency off-target sites (editing frequency ≥1%), number of functional off-target sites (missense/nonsense mutations in exonic regions), and number of high-risk off-target sites (in tumor-related genes). A comprehensive assessment of the off-target risk level (low/medium/high) for single-target editing is provided along with optimization recommendations (e.g., replacing the gRNA, optimizing editing conditions, etc.).

6. Transcriptome Off-target Editing Detection Service Content

Service Timeline: Standard workflow: 30–40 business days (from receipt of qualified samples to report delivery, including off-target site prediction)

7. Sample Requirements

Notes:

(1) All RNA samples must meet the quality standards above. RNA stabilization reagents or −80°C storage with dry ice shipping is recommended.

(2) Clients may also ship cell or tissue samples for RNA extraction.

(3) Complete editing information (target sequence, gRNA, editing tool type, etc.) must be provided to enable accurate off-target site prediction and editing detection.

(4) For special editing types or specific analytical requirements, please communicate with GeneRulor's technical team in advance.

Contact: Phone: 400-6309596 | Technical Support: service@generulor.com

8. References

[1] National Medical Products Administration, Center for Drug Evaluation. (2022). Technical Guidelines for Pharmaceutical Research and Evaluation of In Vivo Gene Therapy Products (Trial).

[2] FDA. (2024). Human Gene Therapy Products Incorporating Human Genome Editing - Guidance for Industry.

[3] Grunewald, J., et al. (2019). Transcriptome-wide off-target RNA editing induced by CRISPR-guided DNA base editors. Nature, 569(7756), 433-437.

[4] Kim, D., et al. (2019). Genome-wide target specificities of CRISPR RNA-guided programmable deaminases. Nature Biotechnology, 37(4), 430-435.

[5] Zuo, E., et al. (2019). Cytosine base editor generates substantial off-target single-nucleotide variants in mouse embryos. Science, 364(6437), 289-292.