Skip to main content
NCBI home page
Science Advances logoLink to Science Advances
. 2025 Mar 26;11(13):eadu7334. doi: 10.1126/sciadv.adu7334

Systematic high-throughput evaluation reveals FrCas9’s superior specificity and efficiency for therapeutic genome editing

Rui Tian 1,†, Xun Tian 2,†, Meiying Yang 3,†, Yuping Song 4,†, Tingting Zhao 1, Chaoyue Zhong 1, Wei Zhu 1, Ping Zhou 5,*, Zhiqiang Han 6,*, Zheng Hu 7,*
PMCID: PMC11939069  PMID: 40138428

Abstract

CRISPR-Cas9 systems have revolutionized genome editing, but the off-target effects of Cas9 limit its use in clinical applications. Here, we systematically evaluate FrCas9, a variant from Faecalibaculum rodentium, for cell and gene therapy (CGT) applications and compare its performance to SpCas9 and OpenCRISPR-1. OpenCRISPR-1 is a CRISPR system synthesized de novo using large language models (LLMs) but has not yet undergone systematic characterization. Using AID-seq, Amplicon sequencing, and GUIDE-seq, we assessed the on-target activity and off-target profiles of these systems across multiple genomic loci. FrCas9 demonstrated higher on-target efficiency and substantially fewer off-target effects than SpCas9 and OpenCRISPR-1. Furthermore, TREX2 fusion with FrCas9 reduced large deletions and translocations, enhancing genomic stability. Through screening of 1903 sgRNAs targeting 21 CGT-relevant genes using sequential AID-seq, Amplicon sequencing, and GUIDE-seq analysis, we identified optimal sgRNAs for each gene. Our high-throughput screening platform highlights FrCas9, particularly in its TREX2-fused form, as a highly specific and efficient tool for precise therapeutic genome editing.

FrCas9 enables precise therapeutic genome editing with enhanced efficiency and minimal off-target effects.

INTRODUCTION

CRISPR-Cas9 systems have become indispensable tools for genome editing, offering unprecedented precision and efficiency in modifying genetic sequences (1, 2). The most widely used variant, SpCas9 from Streptococcus pyogenes, is favored for its robust on-target activity (3). However, its use in therapeutic settings is hindered by substantial off-target effects, which can lead to unintended genomic alterations and potential safety concerns (4–8). This has driven the development of additional CRISPR variants with improved specificity and reduced off-target activity to meet the stringent requirements of clinical applications.

Among these emerging alternatives, FrCas9 from Faecalibaculum rodentium has garnered attention due to its distinct Protospacer Adjacent Motif (PAM) recognition (NNTA) and potentially higher specificity compared to SpCas9 (9, 10). In addition, recently developed systems like OpenCRISPR-1, which was synthesized de novo based on predictions from large language models, have been designed to enhance precision and minimize off-target effects (11). Despite these advancements, a direct comparison of these CRISPR systems across diverse genomic contexts is still needed to determine their relative on-target efficiency and off-target profiles, which are critical for their deployment in cell and gene therapy (CGT).

To address this gap, we conducted a comprehensive evaluation of FrCas9, SpCas9, and OpenCRISPR-1, focusing on both on-target activity and off-target effects across various loci. Using high-throughput techniques including AID-seq (adaptor-mediated off-target identification by sequencing) (12), Amplicon sequencing (13), and GUIDE-seq (genome-wide, unbiased identification of Double-Strand Breaks (DSBs) enabled by sequencing) (7), we systematically assessed the cutting efficiency and specificity of each system. Our results reveal that FrCas9 consistently exhibits superior on-target activity and markedly lower off-target effects compared to both SpCas9 and OpenCRISPR-1, making it a strong candidate for clinical genome editing applications.

In addition to the intrinsic properties of each Cas9 variant, we also explored strategies to enhance FrCas9’s precision through systematic single guide RNA (sgRNA) design optimization (14, 15) and three prime repair exonuclease 2 (TREX2) fusions (16, 17). Modifying the spacer length and introducing hairpin structures in the sgRNA backbone yielded marked improvements in editing efficiency. Similarly, fusing TREX2 to FrCas9 at different termini resulted in substantial reductions in chromosomal translocation rates, further enhancing its safety profile. These modifications were tested across multiple human and mouse loci to validate their effects on editing outcomes.

Together, our study provides a comprehensive assessment of the on-target and off-target activities of FrCas9, SpCas9, and OpenCRISPR-1, offering valuable insights into the suitability of these systems for therapeutic genome editing. Our findings underscore the potential of FrCas9, particularly when coupled with optimized sgRNA designs and TREX2 modifications, as a promising tool for safe and efficient clinical applications.

RESULTS

FrCas9 demonstrates enhanced on-target efficiency and superior specificity relative to SpCas9 and OpenCRISPR-1

In this study, we compared the on-target efficiency and off-target specificity of three CRISPR-Cas systems—FrCas9, SpCas9, and OpenCRISPR-1—across eight genomic loci using GUIDE-seq (Fig. 1A and table S1) (7). The results showed clear differences in performance between these systems.

Fig. 1. Comparative on-target efficiency and off-target specificity of FrCas9, SpCas9, and OpenCRISPR-1.

Fig. 1.

Open in a new tab

(A) The on-target reads and off-target sites of FrCas9, SpCas9, and OpenCRISPR-1 for eight genomic loci, generated by GUIDE-seq in HEK293T cells. The sgRNA and PAM ranges of SpCas9 and OpenCRISPR-1 (20-nt sgRNA and 3-nt PAM) and FrCas9 (22-nt sgRNA and 4-nt PAM) were marked. GUIDE-seq read counts of each site were shown on the right side. (B) Summary of GUIDE-seq on-target reads of FrCas9, SpCas9, and OpenCRISPR-1 at the above eight sites. (C) Summary of GUIDE-seq off-target counts of FrCas9, SpCas9, and OpenCRISPR-1 at the above eight sites. (D) The on:off ratio of GUIDE-seq reads. N = 8 sites. The ratio is defined by log2 ratio of (on-target reads + 1) to (off-target reads + 1).

FrCas9 consistently achieved higher on-target activity compared to SpCas9 and OpenCRISPR-1. For instance, at the Ring Finger Protein 2 locus 1 (RNF2-1) site, FrCas9 had an on-target read count of 32,408, while SpCas9 and OpenCRISPR-1 only had 14,297 and 2147, respectively (Fig. 1, A and B). This pattern was also seen at other sites, such as human embryonic kidney (HEK) 293 site 2, where FrCas9 displayed a read count of 11,612, outperforming the other two systems.

In terms of off-target effects, FrCas9 showed fewer off-target sites overall (Fig. 1, A and C). For example, at the GRIN2B-3 and RUNX1-3 loci, FrCas9 did not produce any detectable off-target events, indicating its higher specificity. This was further supported by the log2 ratio of (on-target reads + 1) to (off-target reads + 1), where the addition of 1 accounted for cases with zero reads to avoid division by zero errors. Across all loci, FrCas9 consistently maintained a higher log2 ratio compared to SpCas9 and OpenCRISPR-1, with average values of 12.85 for FrCas9, 8.53 for SpCas9, and 5.89 for OpenCRISPR-1, further emphasizing FrCas9’s enhanced specificity (Fig. 1D).

The comparative analysis revealed distinct differences in the performance of SpCas9, OpenCRISPR-1, and FrCas9. OpenCRISPR-1, despite being engineered for increased specificity, consistently demonstrated lower on-target cleavage efficiency across all eight loci compared to SpCas9. In addition, OpenCRISPR-1 exhibited higher off-target activity, as indicated by a greater number of unintended cleavage sites. These results suggest that while OpenCRISPR-1 was designed to enhance specificity, it did not achieve superior performance in practice, showing both reduced efficiency and increased off-target effects. In contrast, FrCas9 displayed higher on-target efficiency and fewer off-target events, demonstrating a more balanced profile of specificity and activity. Furthermore, when compared with various high-fidelity SpCas9 variants (eSpCas9, SpCas9-HF1, SuperFiCas9, and xCas9), FrCas9 maintained substantially higher on-target activity while preserving its specificity advantage (fig. S1).

FrCas9 exhibits stringent PAM specificity and reduced off-target activity compared to SpCas9 and OpenCRISPR-1

We used AID-seq (12) to further comprehensively evaluate the specificity and efficiency of FrCas9, SpCas9, and OpenCRISPR-1 across various genomic loci, including 1825 loci for FrCas9 (922 human and 903 mouse) and 1819 loci for SpCas9 and OpenCRISPR-1 (924 human and 895 mouse) (table S2). Genomic DNA was extracted from HEK293T cells for human loci and N2a cells for mouse loci.

The AID-seq data revealed that FrCas9 achieved notably higher on-target read counts, averaging 734.07 reads per site (Fig. 2A), compared to 327.75 for SpCas9 and 652.03 for OpenCRISPR-1 (table S2). This indicates that FrCas9 has a robust on-target cleavage capability, highlighting its effectiveness in precisely targeting specific genomic loci.

Fig. 2. Comprehensive AID-seq analysis of on-target and off-target activities of FrCas9, SpCas9, and OpenCRISPR-1.

Fig. 2.

Open in a new tab

(A) Average on-target read counts per locus for FrCas9, SpCas9, and OpenCRISPR-1 across 1825 loci from AID-seq data. FrCas9 was analyzed at 1825 loci (922 human and 903 mouse), and SpCas9 and OpenCRISPR-1 were analyzed at 1819 loci (924 human and 895 mouse). Genomic DNA was extracted from HEK293T cells for human loci and N2a cells for mouse loci. (B) Average number of off-target sites per locus for FrCas9, SpCas9, and OpenCRISPR-1 from AID-seq data. (C) Average top three off-target reads per site as measured by AID-seq for FrCas9, SpCas9, and OpenCRISPR-1. (D) Log10 ratio of (on-target reads + 1) to (off-target reads + 1) calculated from AID-seq data for FrCas9, SpCas9, and OpenCRISPR-1. (E to G) Violin plots illustrating the distribution of off-target read counts relative to the number of mismatches between the protospacer and PAM sequences for FrCas9 (E), SpCas9 (F), and OpenCRISPR-1 (G). (H to J) Distribution of off-target reads across various PAM sequences according to AID-seq data for FrCas9 (H), SpCas9 (I), and OpenCRISPR-1(J). (K to M) Heatmaps generated from AID-seq data depicting off-target PAM preferences for FrCas9 (K), SpCas9 (L), and OpenCRISPR-1 (M).

In terms of off-target activity, FrCas9 exhibited the fewest off-target sites, with an average of 9.7 sites per locus, substantially lower than the 117.62 for SpCas9 and 76.72 for OpenCRISPR-1 (Fig. 2B and table S2). In addition, the average top three off-target reads per site for each enzyme were calculated. FrCas9 had an average of 196.32 off-target reads per site, which was considerably lower than the 295.49 reads for SpCas9 and 267.74 reads for OpenCRISPR-1 (Fig. 2C and table S2). The log10 ratio of (on-target reads + 1) to (off-target reads + 1) was highest for FrCas9 (4.12), compared to SpCas9 (−3.95) and OpenCRISPR-1 (−2.06), indicating superior specificity (Fig. 2D). The distribution of off-target read counts relative to the number of mismatches between the protospacer and PAM sequences was further analyzed (Fig. 2, E to G). FrCas9 demonstrated consistent low off-target activity across increasing mismatch levels (Fig. 2E), indicating a stringent recognition specificity. In contrast, SpCas9 and OpenCRISPR-1 displayed greater tolerance for mismatches (Fig. 2, F and G), resulting in substantially higher off-target activities.

Next, we conducted PAM specificity analysis. The distribution of off-target reads across different PAM sequences (Fig. 2, H to J) revealed that FrCas9 has a strong preference for NNTA PAM sequences, constituting 93.93% of off-target activity, with a secondary presence at NNAA (3.58%) and NNTG (1.38%). In contrast, SpCas9 showed predominant activity at NGG (76.89%), with additional activity at NGA (12.23%), NAG (3.05%), NGT (2.81%), NCG (1.75%), and NTG (0.92%). This broader PAM preference is consistent with prior studies of SpCas9’s flexibility in recognizing various PAM sequences (18). OpenCRISPR-1 exhibited broader PAM compatibility, with significant off-target activity observed at NGG (69.33%), NGA (17.1%), NAG (4.63%), NGT (3.83%), and NTG (1.28%). The notable similarity in PAM preferences between OpenCRISPR-1 and SpCas9 further underscores the structural relationship between the two systems.

To validate the robustness of our PAM preference analysis and address potential technical biases in AID-seq detection, we compared our SpCas9 PAM profiling results with published High-Throughput PAM Determination Assay (HT-PAMDA) data, which uses an orthogonal PAM detection method independent of DSB capture. The concordance between our observed SpCas9 PAM distribution and those reported using HT-PAMDA suggests that while technique-specific biases may exist in AID-seq, they do not substantially affect the overall PAM preference profiles detected.

Last, we drawn detailed heatmaps (Fig. 2, K to M), which provide comprehensive insights into the off-target PAM preferences of each CRISPR system. FrCas9’s off-target activity was narrowly focused, primarily targeting NATA and NGTA PAM, illustrating a high level of specificity (Fig. 2K). Conversely, SpCas9 and OpenCRISPR-1 demonstrated broader off-target profiles, reflecting their ability to recognize a wider range of PAM sequences (Fig. 2, L and M), which may contribute to their higher off-target potential. The close resemblance in PAM profiles between SpCas9 and OpenCRISPR-1 indicates that the latter system, despite being engineered for enhanced specificity, retains many of the structural and functional characteristics of SpCas9.

Enhanced editing efficiency of FrCas9 through sgRNA design optimization

To improve FrCas9’s editing efficiency, we systematically modified the sgRNA backbone structure (14, 15) and optimized the spacer length. We first evaluated the editing efficiency of 11 sgRNAs targeting Human Papillomavirus (HPV) 16 loci in SiHa cells, comparing the wild-type (WT) sgRNA backbone to a modified sgRNA incorporating a short hairpin (ACUUCGGU) structure (Fig. 3A and table S1). Although individual sgRNAs showed variability (fig. S2A), the short hairpin backbone exhibited a slightly higher average indel rate of 20.5%, compared to 19.8% for the WT backbone (P = 0.93; fig. S2B). This indicates that the short hairpin structure modestly improves sgRNA performance in FrCas9-mediated editing.

Fig. 3. Optimization of sgRNA design for enhanced FrCas9 editing efficiency.

Fig. 3.

Open in a new tab

(A) Schematic representation of the modified sgRNA backbones, featuring long hairpin (GGACUUCGGUCC) and short hairpin (ACUUCGGU) structures at stem loop 2. (B) Editing efficiencies of FrCas9 using sgRNAs with long and short hairpins across 30 human genomic loci in HEK293T cells by Amplicon sequencing. (C) Comparative efficiency data indicating a slightly higher average indel rate for the short hairpin variant (60.9%) compared to the long hairpin (59.4%) (P = 0.42, two-sided paired Wilcoxon rank-sum test). (D) Comparison of editing efficiencies for sgRNAs with 21- and 22-bp spacer lengths, using a short hairpin backbone, across 25 human genomic loci in HEK293T cells by Amplicon sequencing (P = 0.0035, two-sided paired Wilcoxon rank-sum test). (E) Quantitative analysis showing the number of loci where each spacer length demonstrated higher editing efficiency. The 21-bp spacer outperformed the 22-bp spacer in 19 loci, while the 22-bp spacer was more efficient in 6 loci (P < 0.05). (F) Distribution of deletion sizes observed in FrCas9 editing products across 25 human genomic loci with spacer = 21, highlighting a predominance of 1-bp deletions. (G) Distribution of insertion sizes observed in FrCas9 editing products across 25 human genomic loci with spacer = 21, indicating that 1-bp insertions are the most frequent.

On the basis of the performance of the short hairpin in SiHa cells, we extended our analysis to HEK293T cells to compare the short hairpin with a long hairpin (GGACUUCGGUCC) structure across 30 genomic loci (Fig. 3A and table S1). The short hairpin outperformed the long hairpin, with a slight but consistent increase in editing efficiency (60.9% versus 59.4%, P = 0.42) (Fig. 3, B and C). These results further support the use of the short hairpin for optimized sgRNA design in FrCas9 applications.

Next, building on the optimized short hairpin backbone, we then evaluated the effect of spacer length on editing efficiency by comparing 21- and 22-bp spacers across 25 human genomic loci in HEK293T cells (table S1). The 21-bp spacer consistently outperformed the 22-bp spacer, achieving a mean editing efficiency of 59.7%, compared to 53.1% for the 22-bp spacer (P = 0.0035; Fig. 3, D and E). These results indicate that the 21-bp spacer is more effective at maximizing FrCas9’s on-target activity.

Last, we conducted an in-depth analysis of the indels generated by FrCas9 and found that its indel distribution was similar to that reported for SpCas9 (19–21). Small indels, particularly 1-bp deletions and insertions, dominated the outcomes, with 1-bp deletions accounting for 55% of total deletions and 1-bp insertions comprising 90% of total insertions (Fig. 3, F and G). These findings suggest that FrCas9 induces precise and minimal genetic changes, which is advantageous for applications requiring high editing accuracy.

TREX2 fusion enhances FrCas9 editing efficiency and reduces chromosomal translocations

Recent research has shown that TREX2 fusion to SpCas9 can significantly lower the translocation rates associated with SpCas9-mediated genome editing (17). To explore whether this strategy could similarly improve FrCas9’s performance, we generated two fusion variants: TREX2-FrCas9 (TREX2 fused to the N terminus of FrCas9) and FrCas9-TREX2 (TREX2 fused to the C terminus of FrCas9), linked by a GGGGS linker (Fig. 4A and table S1). These constructs were evaluated at three genomic loci (RNF2, HEK293 site 2, and GRIN2B-3) to assess their impact on editing efficiency, indel profile, and translocation frequency.

Fig. 4. Evaluation of FrCas9 and its TREX2 fusion variants for editing efficiency and specificity.

Fig. 4.

Open in a new tab

(A) Schematic representation of the fusion constructs: TREX2-FrCas9 (N-terminal fusion) and FrCas9-TREX2 (C-terminal fusion). (B) Editing efficiencies of FrCas9, TREX2-FrCas9, and FrCas9-TREX2 at three genomic loci (RNF2, HEK293 site 2, and GRIN2B-3) as determined by Amplicon sequencing. Data show the distribution of indel types (deletions, insertions) across the three loci. (C) Deletion size distribution for FrCas9 and its TREX2 variants across the tested loci. (D to F) Results from PEM-seq showing editing efficiency (D), proportion of deletions (E), and general translocation rates (F) for each construct. (G to I) Translocation patterns for FrCas9, TREX2-FrCas9, and FrCas9-TREX2 at the GRIN2B-3 locus. CMV, cytomegalovirus.

The indel rates and types generated by FrCas9, TREX2-FrCas9, and FrCas9-TREX2 were evaluated using Amplicon sequencing. The results show a clear distinction in the indel profiles among the three variants (Fig. 4B). TREX2-FrCas9 demonstrates a higher overall indel rate across all three loci (RNF2, HEK293 site 2, and GRIN2B-3) compared to FrCas9 and FrCas9-TREX2, indicating an enhanced editing efficiency for this variant. Both TREX2-FrCas9 and FrCas9-TREX2 exhibit an increased proportion of deletions (Del) and a decreased proportion of insertions (Ins) relative to FrCas9 (Fig. 4B). This shift suggests that the presence of TREX2 alters the balance of DNA repair pathways, favoring the generation of deletions over insertions (21, 22). Specifically, TREX2-FrCas9 displayed the highest deletion rate, indicating that the N-terminal TREX2 fusion has a more pronounced effect on the DNA repair process.

To further understand the nature of these deletions, we analyzed the size distribution of the deletions produced by each variant. The data showed that both TREX2 fusions resulted in a higher proportion of larger deletions (6 to 10 bp for FrCas9-TREX2 and 11 to 15 bp for TREX2-FrCas9) compared to FrCas9 (Fig. 4C). TREX2-FrCas9, in particular, exhibited a broader deletion size range, including a significant increase in deletions within the 11- to 20-bp range. This suggests that TREX2 fusion enhances exonuclease activity, generating larger deletions than FrCas9 alone.

Given the observed impact of TREX2 on deletion profiles, we further analyzed translocation frequencies using PEM-seq (primer extension-mediated sequencing) (23). Consistent with Amplicon sequencing, PEM-seq data confirmed that TREX2 fusion significantly influences editing efficiency and indel patterns. TREX2-FrCas9 exhibited the highest editing efficiency at 27.1%, followed by FrCas9 (22.5%) and FrCas9-TREX2 (16.8%) (Fig. 4D). Notably, the editing efficiencies detected by PEM-seq were generally lower than those measured by Amplicon sequencing, which can be attributed to the tendency of Amplicon sequencing to overestimate editing efficiency by preferentially amplifying shorter DNA fragments, thereby inflating the proportion of edited sequences (24). PEM-seq, with its ability to capture a more comprehensive range of repair outcomes, provides a more accurate reflection of true editing events across the genome (8, 23).

In addition to editing efficiency, PEM-seq data revealed similar trends in deletion proportions (Fig. 4E and table S3), with TREX2-FrCas9 again showing the highest proportion of deletions (25.6%), followed by FrCas9-TREX2 (15.5%) and FrCas9 (12.9%). Furthermore, TREX2 fusions led to a significant reduction in translocation rates compared to unmodified FrCas9. Specifically, TREX2-FrCas9 exhibited the lowest translocation rate at 0.43%, representing a 6.35-fold reduction compared to FrCas9’s rate of 2.73%. FrCas9-TREX2 followed closely with a translocation rate of 0.55%, a 4.96-fold reduction compared to FrCas9 (Fig. 4, F to I; fig. S3; and table S3). These findings indicate that TREX2 fusion, especially when attached to the N terminus, stabilizes the DNA repair process, reducing chromosomal rearrangements and promoting genomic integrity.

The consistent results from Amplicon sequencing and PEM-seq highlight the robustness of TREX2’s impact on FrCas9-mediated genome editing. TREX2 fusion, especially at the N terminus, boosts editing efficiency, promotes larger and more controlled deletions, and significantly reduces off-target translocations. These findings suggest that TREX2-FrCas9 is a promising tool for high-precision genome editing, where both enhanced efficiency and genomic stability are crucial.

To further validate TREX2-FrCas9’s clinical potential, we evaluated its performance at the BCL11A enhancer, a clinically validated target site for β-hemoglobinopathies treatment (table S1), using CD34+ hematopoietic stem and progenitor cells (HSPCs) from mobilized peripheral blood. PEM-seq analysis revealed that TREX2-FrCas9 achieved significantly higher editing efficiency (62.56% versus 43.28%) while maintaining remarkable genomic stability compared to SpCas9 Casgevy (fig. S3, G to I, and table S3). Most notably, TREX2-FrCas9 demonstrated a 6.2-fold reduction in the translocation rate (0.85% versus 5.24%) and a 22.8-fold decrease in the close inversion rate (0.08% versus 1.82%), addressing major safety concerns in therapeutic genome editing. The reduction in genomic instability was further evidenced by TREX2-FrCas9’s substantially decreased frequency of large deletions (>100 bp, 0.79% for TREX2-FrCas9 versus 3.30% for SpCas9) and characteristic editing pattern favoring deletions (60.84% versus 24.79%) while minimizing unwanted insertions (0.79% versus 11.43%). This pronounced enhancement of both efficiency and safety profiles at a clinically relevant target in therapeutic cell types provides compelling evidence for TREX2-FrCas9’s potential in clinical applications.

FrCas9 exhibits high specificity and robust editing efficiency in CGT-relevant genes to support its therapeutic potential

To assess the therapeutic potential of FrCas9 for CGT, we implemented a comprehensive high-throughput screening approach, integrating AID-seq, Amplicon sequencing, and GUIDE-seq to systematically evaluate on-target activity and off-target effects across 21 CGT-relevant genes (fig. S4). These genes included AAVS1, B2M, BATF, CCR5, CD52, CD7, CD96, CIITA, CTLA4, HAVCR2, hROSA26, LAG3, PDCD1, PTPN1, RASA2, Rogi1, Rogi2, SOCS1, TIGIT, TRAC, and TRBC (25–31). For each gene, we designed a customized sgRNA library to ensure comprehensive coverage, enabling a robust analysis of FrCas9’s genome editing efficiency and specificity at these therapeutically relevant loci.

AID-seq served as the initial screening platform, assessing 1983 sgRNAs, with the number of sgRNAs per gene ranging from 6 (for TRBC) to 483 (for Rogi1) (table S4). Among the 1983 sgRNAs tested, 1884 (94.5%) showed detectable on-target activity, with a mean read count of 6440.9. The off-target analysis revealed that 725 sgRNAs (36.6%) had ≤10 off-target sites, with 347 (17.5%) showing ≤5 off-target sites, and notably, 19 sgRNAs (1.0%) demonstrating no detectable off-target activity. In addition, 1029 sgRNAs (51.9%) exhibited ≤15 off-target sites, indicating that a majority of the tested guides maintained relatively specific targeting (fig. S5, A and B). sgRNAs were ranked then by their log10 ratio of (on-target reads + 1) to (off-target reads + 1), facilitating the identification of high-specificity candidates (Fig. 5A). For instance, sgRNAs targeting CIITA (e.g., CIITA-397) and Rogi1 (e.g., Rogi1-5886) demonstrated high specificity, with log10 ratio of (on-target reads + 1) to (off-target reads + 1) exceeding 4, along with high on-target read counts, making them optimal for therapeutic applications. The top five sgRNAs identified via AID-seq for each gene were further validated using Amplicon sequencing, which quantified the editing efficiency at each target locus (Fig. 5B and table S5). Substantial variability in editing efficiency was observed across different sgRNAs and target loci. However, FrCas9 consistently demonstrated robust genome editing capabilities, achieving editing efficiencies exceeding 70% in genes such as B2M, hROSA26, Rogi2, AAVS1, CD7, PDCD1, and CCR5. Notably, sgRNAs targeting hROSA26 exhibited near 100% editing efficiency, further underscoring FrCas9’s potential as an efficient gene editing tool for CGT applications (Fig. 5B).

Fig. 5. Systematic screening of FrCas9 for on-target and off-target activities across 21 CGT-related genes.

Fig. 5.

Open in a new tab

(A) AID-seq screening: Log10-transformed ratio of on-target reads (+1) to off-target reads (+1) for sgRNAs targeting 21 CGT-related genes. sgRNAs per gene ranged from 6 (TRBC) to 483 (Rogi1). Red dots indicate the top five sgRNAs per gene selected for further validation. (B) Amplicon sequencing: On-target editing efficiency of the top five sgRNAs per gene, identified by AID-seq. Red dots highlight the top three sgRNAs selected for GUIDE-seq analysis. (C) GUIDE-seq on-target efficiency: GUIDE-seq measurement of on-target reads across 63 genomic loci in HEK293T cells for the top three sgRNAs per gene. (D) GUIDE-seq off-target detection: Number of off-target sites detected by GUIDE-seq across the same 63 loci for FrCas9. (E) GUIDE-seq on-to-off-target ratio: Log10-transformed ratio of on-target reads (+1) to off-target reads (+1), representing the specificity of each sgRNA.

To validate our experimental findings, we performed computational prediction of potential off-target sites using Cas-OFFinder. For five randomly selected targets (PDCD1-4, CTLA4-57, CD96-208, CD7-13, and B2M-80), Cas-OFFinder predicted thousands of potential off-target sites (ranging from 6822 to 21,191) with parameters set as PAM = NNTA, mismatch < 6, DNA bulge = 1, and RNA bulge = 1. However, AID-seq experimentally identified only 6 to 20 sites per target, with 22 to 50% overlap with computational predictions (fig. S5C). This substantial discrepancy between computational prediction and experimental validation further emphasizes the importance of experimental methods for accurate off-target assessment.

To comprehensively assess genome-wide specificity, we performed GUIDE-seq on the top three sgRNAs for each gene, selected on the basis of the highest editing efficiencies observed in the Amplicon sequencing results (Fig. 5B). GUIDE-seq analysis revealed that 34 of the 63 sgRNAs (54%) showed no detectable off-target sites, highlighting the high specificity of FrCas9 (Figs. 5C and 6 and table S5). Of the remaining sgRNAs, 16 had only 1 off-target site, with low read counts ranging from 2 to 24. Only 2 sgRNAs exhibited 2 off-target sites, and 5 sgRNAs presented 3 off-target sites, with the highest off-target read count being just 18, indicating minimal off-target activity (Figs. 5C and 6). Notably, only 1 sgRNA displayed more than 3 off-target sites, with a total of 17 off-target events, further demonstrating the limited off-target effects associated with FrCas9.

Fig. 6. Selected optimal sgRNAs for each gene.

Fig. 6.

Open in a new tab

For each gene, the most optimal sgRNA, showing the highest editing efficiency and lowest off-target activity, is displayed. The figure shows GUIDE-seq read counts and Amplicon sequencing indel rates for the selected sgRNAs, providing a comparison of on-target performance and validation of editing efficiency.

The consistency of the results across AID-seq, Amplicon sequencing, and GUIDE-seq underscores the robustness and reliability of FrCas9 as a genome editing tool. The majority of sgRNAs demonstrated either no off-target effects or low-frequency off-target events, further reinforcing FrCas9’s suitability for precise genome modifications. This high level of specificity, coupled with robust editing efficiency across multiple CGT-relevant genes, positions FrCas9 as a superior choice for clinical applications requiring precise and safe genome editing with minimal unintended genetic alterations.

DISCUSSION

This study provides robust evidence for the advantages of FrCas9 as a genome editing tool for CGT applications, demonstrating its superior on-target activity and specificity compared to SpCas9 and OpenCRISPR-1. FrCas9 consistently achieved higher on-target read counts and exhibited fewer off-target events across multiple genomic loci, likely due to its unique PAM recognition and stringent target site discrimination. These characteristics make FrCas9 particularly well-suited for therapeutic applications, where minimizing off-target effects is critical to ensure safety and efficacy.

The enhanced specificity of FrCas9 may be partially attributed to its NNTA PAM requirement, as supported by our Cas-OFFinder analysis showing 3.2- to 22.3-fold fewer potential off-target sites compared to SpCas9 (table S7). While SpCas9’s NGG PAM is frequently found in GC-rich regions that often correspond to highly expressed, more accessible genes, FrCas9’s NNTA PAM naturally occurs more frequently in Adenine-Thymine (AT)-rich regions. Previous studies have demonstrated that AT-rich regions tend to adopt more compact chromatin conformations with lower accessibility (32). This inherent difference in PAM distribution patterns and associated chromatin states likely contributes to FrCas9’s reduced off-target activity by limiting potential binding sites in highly accessible genomic regions. This PAM-related advantage, combined with FrCas9’s intrinsic structural features, collectively contributes to its superior specificity profile.

The modification of FrCas9 with TREX2 further enhances its potential for clinical use by reducing large deletions and translocations, which are major concerns in therapeutic genome editing (24, 33–35). Our results indicate that both N-terminal and C-terminal TREX2 fusions with FrCas9 increase editing efficiency and favor a shift toward more precise deletions, thereby improving genomic stability (17). This reduction in translocation rates, especially with TREX2-FrCas9, suggests that these modifications could provide a safer genome editing tool for clinical applications, addressing key challenges associated with DNA repair fidelity.

While TREX2-FrCas9 fusion demonstrates enhanced editing efficiency and reduced translocation rates, it is important to acknowledge its limitation regarding the increased frequency of short deletions. As shown in our data (Fig. 4C), TREX2-FrCas9 exhibits a higher proportion of deletions in the 11- to 20-bp range compared to unmodified FrCas9. These short deletions can lead to a higher likelihood of frameshift mutations, which may be undesirable for applications requiring precise gene correction or in-frame editing. However, this characteristic makes TREX2-FrCas9 particularly well-suited for applications where gene disruption is the desired outcome, such as knockout studies or therapeutic strategies requiring complete loss of protein function [e.g., disruption of Programmed Cell Death protein 1 (PD-1) or Cytotoxic T-Lymphocyte Associated protein 4 (CTLA4) in Chimeric Antigen Receptor T cell (CAR-T) cells (36, 37)]. For applications requiring precise gene correction or modification, we recommend using unmodified FrCas9 or exploring alternative strategies such as base editing (38) or prime editing (39). This understanding allows researchers to choose the most appropriate tool based on their specific experimental or therapeutic goals.

The high-throughput screening strategy used in this study, integrating AID-seq (12), Amplicon sequencing (13), and GUIDE-seq (7), is broadly applicable to the selection of sgRNAs for any CRISPR-based system. By identifying sgRNAs with both high on-target activity and minimal off-target effects, this approach provides a reliable and scalable method for optimizing sgRNA selection across diverse genomic targets. Its applicability to all CRISPR platforms makes it a valuable tool for advancing precision genome editing in both research and clinical settings. A comprehensive comparison of these techniques and their strategic implementation has been provided in table S6.

A critical consideration in therapeutic genome editing is defining acceptable safety thresholds for off-target effects. Recent regulatory approvals, such as CASGEVY by the Food and Drug Administration, have helped establish concrete benchmarks, with off-target frequencies of ≥0.2% designated as high-frequency events requiring detailed characterization. This aligns with the EMA’s guidance (EMA/CAT/GTWP/671639/2008 Rev. 1) on gene therapy medicinal products, which emphasizes comprehensive off-target analysis. While the scientific community often uses more stringent detection thresholds of ≥0.01% (12, 40–42), these must be considered alongside the technical limitations of current sequencing platforms, which have an inherent error rate of ≥0.1% (43).

In this context, our findings demonstrate that FrCas9 achieves remarkably low off-target rates that align well with therapeutic requirements. The observation that 54% of analyzed sgRNAs showed no detectable off-target events, combined with the minimal translocation rate of 0.43% achieved with TREX2-FrCas9, suggests that this system could meet the stringent safety requirements for therapeutic applications. These characteristics, coupled with its high on-target efficiency, position FrCas9 as a promising tool for clinical genome editing, although specific safety assessments will need to be conducted on a case-by-case basis in alignment with regulatory guidelines.

MATERIALS AND METHODS

Cell culture and genomic DNA preparation

HEK293T [American Type Culture Collection (ATCC), CRL-11268] and N2a (ATCC, CCL-131) cell lines were maintained in Dulbecco’s Modified Eagle's Medium (DMEM) (Gibco, 11995065) supplemented with 10% fetal bovine serum (Gibco, 10099141) and 1% penicillin-streptomycin (10,000 U/ml) (Gibco, 15140122) at 37°C in a humidified atmosphere containing 5% CO2. Cells were seeded at a density of 2 × 105 cells per well in six-well plates. Genomic DNA was extracted 72 hours posttransfection using the DNeasy Blood & Tissue Kit (QIAGEN, 69504) following the manufacturer’s protocol. DNA concentration and purity were measured using a NanoDrop spectrophotometer.

Plasmid construction

sgRNAs were designed using the CRISPR Design Tool [Massachusetts Institute of Technology (MIT)] to minimize off-target effects. The sgRNA sequences were synthesized and cloned into the pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid (Addgene) for SpCas9, with equivalent plasmids used for FrCas9 and OpenCRISPR-1. The OpenCRISPR-1 gene, synthesized by GeneWiz (Suzhou), was directly substituted for SpCas9 in the pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid, while the sgRNA backbone remained unchanged.

For FrCas9, sgRNAs with both 21- and 22-bp spacers were cloned into the pX330-U6-Chimeric_BB-CBh-FrCas9 plasmid to assess spacer length optimization. In addition, plasmids incorporating short hairpin (ACUUCGGU) and long hairpin (GGACUUCGGUCC) structures were constructed on the pX330-U6-Chimeric_BB-CBh-FrCas9 backbone to enhance sgRNA stability. The TREX2 gene, also synthesized by GeneWiz (Suzhou), was cloned either to the N terminus or C terminus of FrCas9 using a flexible GGGGS linker to generate TREX2-FrCas9 and FrCas9-TREX2 constructs. All constructs were verified by Sanger sequencing to ensure proper cloning and sequence integrity.

Last, sgRNAs targeting 21 CGT-relevant genes (including AAVS1, B2M, BATF, CCR5, CD52, CD7, CD96, CIITA, CTLA4, HAVCR2, hROSA26, LAG3, PDCD1, PTPN1, RASA2, Rogi1, Rogi2, SOCS1, TIGIT, TRAC, and TRBC) were cloned into the pX330-U6-Chimeric_BB-CBh-FrCas9 plasmid using a 22-bp spacer and the WT sgRNA backbone for further validation and testing. The complete sgRNA sequences and plasmid maps are provided in tables S1 and S2, respectively.

sgRNA library oligo synthesis and in vitro transcription

A comprehensive sgRNA library was synthesized to cover multiple target sites within the 21 CGT-relevant genes. Oligonucleotides encoding the sgRNA sequences were synthesized by Twist Bioscience. The oligos were polymerase chain reaction (PCR)–amplified using high-fidelity DNA polymerase, and the amplified products were gel-purified. These PCR products served as templates for in vitro transcription, which was carried out using the HiScribe T7 High Yield RNA Synthesis Kit [New England Biolabs (NEB), E2040S] according to the manufacturer’s protocol. The transcribed sgRNAs were purified using the Monarch RNA Cleanup Kit (NEB, T2030L). The concentration and quality of the sgRNAs were assessed using an Agilent Bioanalyzer and a Qubit RNA HS Assay Kit (Thermo Fisher Scientific, Q32852).

Cell transfection and genome editing

Cells were transfected with sgRNA/Cas9 plasmids using Lipofectamine 3000 (Thermo Fisher Scientific, L3000015). For each well of a six-well plate, 2 μg of plasmid DNA and 3 μl of Lipofectamine 3000 reagent were diluted separately in 125 μl of Opti-MEM medium (Gibco, 31985070). After a 5-min incubation at room temperature, the DNA and Lipofectamine mixtures were combined and incubated for an additional 20 min to form DNA-lipid complexes. The complexes were then added dropwise to the cells. After 72 hours, cells were harvested for genomic DNA extraction or further downstream analysis. Transfection efficiency was monitored using a green fluorescent protein reporter plasmid cotransfected with Cas9/sgRNA constructs.

CD34+ cell culture and electroporation

CD34+ HSPCs were obtained from mobilized peripheral blood of healthy donors (Junx Bio, catalog no. PBSC001). CD34+ cells were thawed and cultured in StemSpan SFEM II medium supplemented with Stem Cell Factor (SCF) (50 ng/ml), Thrombopoietin (TPO) (50 ng/ml), Flt3-ligand (50 ng/ml), and interleukin-6 (50 ng/ml) for 24 to 48 hours before electroporation. For ribonucleoprotein (RNP) complex formation, purified Cas9 protein was preincubated with sgRNA targeting the BCL11A enhancer (sgRNA sequence in table S1) at a molar ratio of 1:1.5 in electroporation buffer for 15 min at room temperature. CD34+ HSPCs (2.5 × 105 cells per reaction) were electroporated with preassembled RNP complexes using the Lonza 4D-Nucleofector X Unit with program DZ-100 and a P3 primary cell nucleofection kit according to the manufacturer’s protocol. Postelectroporation, cells were immediately transferred to prewarmed culture medium and maintained at 37°C with 5% CO2 for 48 hours before genomic DNA extraction.

Protein purification of FrCas9, SpCas9, OpenCRISPR-1, and TREX2-FrCas9

The purification of FrCas9, SpCas9, OpenCRISPR-1, and TREX2-FrCas9 proteins was carried out using standard affinity chromatography. The coding sequences for each Cas9 variant were cloned into the pTBX1 vector, which includes a C-terminal intein tag and a chitin-binding domain to facilitate purification. The vectors were transformed into Escherichia coli BL21 codon plus (DE3) arginine, isoleucine, proline, and leucine codon-optimized (RIPL) cells (Agilent Technologies, catalog no. 230280) for protein expression, and the transformed cells were grown in Terrific Broth medium (Sangon, B540123) at 37°C until the optical density at 600 nm reached 0.6 to 0.8. Protein expression was induced by adding 0.1 mM isopropyl-β-d-thiogalactopyranoside (Sigma-Aldrich, catalog no. I6758), followed by incubation at 20°C for 16 to 18 hours to achieve optimal protein expression levels. After induction, the cells were harvested by centrifugation at 5000g for 15 min, and the pellets were resuspended in lysis buffer containing 20 mM Hepes-NaOH (pH 8.0) (Sigma-Aldrich, catalog no. H3375), 500 mM NaCl (Thermo Fisher Scientific, catalog no. S271-3), 1 mM EDTA (Invitrogen, catalog no. 15575020), 0.1% Triton X-100 (Sigma-Aldrich, catalog no. T8787), and 1 mM phenylmethylsulfonyl fluoride (Sigma-Aldrich, catalog no. P7626). Cell lysis was performed via sonication, and the lysates were clarified by centrifugation at 15,000g for 30 min at 4°C. To remove cellular DNA from the lysate, polyethylenimine (Sigma-Aldrich, catalog no. 408727) precipitation was applied. The clarified lysate was then loaded onto a chitin affinity column for protein purification.

AID-seq for high-throughput specificity screening

AID-seq was performed to assess the specificity and efficiency of FrCas9, SpCas9, and OpenCRISPR-1. Genomic DNA from HEK293T and N2a cells was first fragmented to 400 to 700 bp using a Bioruptor sonication system (Diagenode). Fragmented DNA was purified using 1× magnetic beads (Generulor, GR4003) and eluted in Tris-EDTA (TE) buffer.

End-repair and A-tailing were performed using the KAPA HTP library preparation kit (Roche, KK8234), followed by adapter ligation with custom ODN + i7 + U adapters. After ligation, samples were purified using magnetic beads. To minimize false positives, the ligated DNA was subjected to three sequential rounds of exonuclease digestion using a cocktail of three nucleases: lambda exonuclease (NEB, M0262S), exonuclease I (NEB, M0293S), and exonuclease III (NEB, M0206S). This three-step exonuclease digestion was specifically implemented to remove unligated or partially ligated adapters and ensure the highest possible specificity by eliminating false positives. After each round of digestion, DNA was purified using magnetic beads (Generulor, GR4003).

For in vitro cleavage assays, recombinant FrCas9, SpCas9, and OpenCRISPR-1 proteins were preincubated with their respective sgRNAs in NEB3.1 buffer (NEB, B7203S) at room temperature for 10 min to form RNP complexes. Purified genomic DNA (125 ng) was added to the reaction, and cleavage was allowed to proceed at 37°C for 2 hours. The reaction was stopped by adding proteinase K (NEB, P8107S) and incubating at 55°C for 30 min.

After cleavage, the DNA was purified using magnetic beads (Generulor, GR4003). The purified DNA was directly processed using the KAPA HTP library preparation kit (Roche, KK8234) for A-tailing and biotinylated adapter ligation. End repair was intentionally omitted to avoid potential read loss and experimental failure. Following adapter ligation, the biotinylated DNA fragments were captured using streptavidin magnetic beads (Generulor, GR4001) to enrich for the desired cleavage products. Postcapture, two rounds of nested PCR were carried out to selectively amplify the target regions. The libraries were sequenced on the MGI2000 platform, and sequencing reads were aligned to the human (hg38) and mouse (mm10) reference genomes using Bowtie2. Data were analyzed using open-source AID-seq software (https://github.com/yuyanwong/AID-seq).

GUIDE-seq for genome-wide off-target analysis

GUIDE-seq was used to systematically map genome-wide off-target effects of FrCas9, SpCas9, and OpenCRISPR-1. The comparative analysis of these three CRISPR systems was conducted on eight well-established endogenous loci frequently cited in the literature. In addition, a separate GUIDE-seq experiment was performed to assess off-target effects for FrCas9 across 21 CGT-relevant genes.

For the comparative analysis, plasmids encoding sgRNAs targeting the eight endogenous loci were equally pooled for electroporation, allowing a direct comparison between FrCas9, SpCas9, and OpenCRISPR-1. In contrast, for the 21 CGT-relevant genes, 3 sgRNAs were designed per gene, resulting in a total of 63 sgRNAs, which were divided into five subpools based on genomic loci. Each subpool was electroporated separately for FrCas9, as this portion of the study focused exclusively on FrCas9’s off-target effects across these 21 genes.

Electroporation was performed using a 100-μl reaction system on the LONZA 4D Nucleofector (SF Cell Line 4D-Nucleofector X Kit S, V4XC-2032), with 2 μl of double-stranded oligonucleotide (dsODN; 100 pmol/μl) to label DNA double-strand breaks and 2 μg of plasmid DNA for each transfection. The pooled GUIDE-seq approach was used for the 8-gene comparison, while FrCas9 was tested with each of the five subpools for the 21-gene experiment.

Seventy-two hours after electroporation, genomic DNA was extracted and fragmented to 300 to 500 bp using a Bioruptor sonicator (Diagenode). The library construction for GUIDE-seq was performed according to established protocols from the literature (44). Enrichment for dsODN-tagged DNA fragments was achieved via two rounds of PCR. The first round used primers specific to the dsODN and adjacent genomic regions, and the second round added sequencing adapters and sample-specific indices. The prepared libraries were sequenced on the MGI2000 platform. Data were analyzed using open-source guideseq software (https://github.com/aryeelab/guideseq).

To ensure fair comparison between the three CRISPR systems, sequencing data for all samples was normalized to 1 Gb before analysis. This normalization step eliminates potential bias from varying sequencing depths and allows direct comparison of on-target and off-target read counts between FrCas9, SpCas9, and OpenCRISPR-1.

PEM-seq for detection of large deletions and translocations in TREX2 fusion variants

PEM-seq (24, 45) was used to detect and quantify large deletions and translocations in cells transfected with FrCas9, FrCas9-TREX2 (C-terminal fusion), and TREX2-FrCas9 (N-terminal fusion), following a previously published protocol (36). The purpose of this analysis was to assess how TREX2 fusions influence genomic stability, specifically by measuring the frequency of large deletions and translocations.

Seventy-two hours posttransfection, genomic DNA was extracted from HEK293T cells using a standard genomic DNA extraction kit. The extracted DNA was fragmented to 300 to 700 bp using sonication. Linear primer extension was performed using primers specific to regions adjacent to predicted breakpoints, which allowed the capture of extended DNA fragments linked to translocation or deletion events. The biotinylated extended DNA was captured using streptavidin-coated magnetic beads (Generulor, GR4001), followed by ligation to sequencing adapters. Two rounds of PCR amplification were performed to enrich for extended DNA fragments, ensuring the detection of breakpoints. Libraries were prepared following a published PEM-seq protocol, ensuring optimal detection of structural variations, and sequenced on the MGI2000 platform. Sequencing data were processed using the open-source PEM-Q pipeline (available at https://github.com/liumz93/PEM-Q) to detect large deletions and translocations. The editing efficiency for each CRISPR variant was calculated as the proportion of edited reads relative to total reads. The detailed editing outcomes are provided in table S3.

Amplicon sequencing for on-target editing efficiency

Amplicon sequencing was used to quantify the on-target editing efficiency. Genomic DNA was extracted from transfected cells, and target regions flanking the sgRNA binding sites were amplified using two rounds of PCR, both using KAPA HiFi HotStart ReadyMix (Roche, KK2601) for high-fidelity amplification. The first-round PCR amplified the specific target regions, followed by magnetic bead-based purification to remove contaminants. In the second round of PCR, sample-specific indices were introduced to enable multiplexing. After the second PCR, the products were purified via gel extraction to ensure high specificity of the amplified fragments. The gel-purified products were then sequenced on the MGI2000 platform. Sequencing reads were analyzed with CRISPResso2 (46) to quantify indel frequencies at each target site, providing a precise measure of the editing efficiency for each CRISPR variant.

Statistical analysis

Data were analyzed using GraphPad Prism 8 and R software. For comparisons between two groups (such as sgRNA backbone structures and spacer lengths), two-sided paired Wilcoxon rank-sum tests were performed. For multiple comparisons, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used. Data are presented as mean ± SD, with a significance threshold set at P < 0.05. Statistical significance was indicated as follows: *P < 0.05, **P < 0.01, and ***P < 0.001.

Acknowledgments

Funding: This work was supported by National Natural Science Foundation of China (32171465 and 32371541 to Z.Hu.; 82102392 to R.T.; 82172584 to X.T.; 82372672 to Z.Han), China Postdoctoral Science Foundation (2023 M734090 to R.T.; 2023M744121 to C.Z.; 2023M734091 to T.Z.), Key Technology R&D Program of Hubei (2024BCB057 to X.T. and Z.Hu.), Guangdong Foundation for Basic and Applied Basic Research Foundation Regional Joint Fund (2021B1515140063 to P.Z.), National key research and development program (2024YFC2707404) and Guangxi Natural Science Foundation (2024GXNSFBA010045 to M.Y.).

Author contributions: Conceptualization: R.T. Methodology: R.T. and X.T. Investigation: M.Y. and Y.S. Visualization: R.T., C.Z., T.Z., and W.Z. Supervision: Z. Hu, Z.Han., and P.Z. Writing—original draft: R.T. Writing—review and editing: Z. Hu, Z.Han., P.Z., M.Y., and Y.S.

Competing interests: R.T., C.Z., T.Z., and W.Z. are employees of Generulor Co. Ltd. The authors declare that they have no competing interests.

Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The sequencing data have been deposited in the Sequence Read Archive (SRA) under accession number PRJNA1215777.

Supplementary Materials

The PDF file includes:

Figs. S1 to S5

Legends for tables S1 to S7

sciadv.adu7334_sm.pdf (2.3MB, pdf)

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S7

sciadv.adu7334_tables_s1_to_s7.zip (1.9MB, zip)

REFERENCES AND NOTES

  • 1.Pacesa M., Pelea O., Jinek M., Past, present, and future of CRISPR genome editing technologies. Cell 187, 1076–1100 (2024). [DOI] [PubMed] [Google Scholar]
  • 2.Wang J. Y., Doudna J. A., CRISPR technology: A decade of genome editing is only the beginning. Science 379, eadd8643 (2023). [DOI] [PubMed] [Google Scholar]
  • 3.Hsu P. D., Scott D. A., Weinstein J. A., Ran F. A., Konermann S., Agarwala V., Li Y., Fine E. J., Wu X., Shalem O., Cradick T. J., Marraffini L. A., Bao G., Zhang F., DNA targeting specificity of RNA-guided Cas9 nucleases. Nat. Biotechnol. 31, 827–832 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Wienert B., Wyman S. K., Richardson C. D., Yeh C. D., Akcakaya P., Porritt M. J., Morlock M., Vu J. T., Kazane K. R., Watry H. L., Judge L. M., Conklin B. R., Maresca M., Corn J. E., Unbiased detection of CRISPR off-targets in vivo using DISCOVER-seq. Science 364, 286–289 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Wu J., Zou Z., Liu Y., Liu X., Zhangding Z., Xu M., Hu J., CRISPR/Cas9-induced structural variations expand in T lymphocytes in vivo. Nucleic Acids Res. 50, 11128–11137 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Manghwar H., Li B., Ding X., Hussain A., Lindsey K., Zhang X., Jin S., CRISPR/cas systems in genome editing: Methodologies and tools for sgRNA design, off-target evaluation, and strategies to mitigate off-target effects. Adv. Sci. 7, 1902312 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Tsai S. Q., Zheng Z., Nguyen N. T., Liebers M., Topkar V. V., Thapar V., Wyvekens N., Khayter C., Iafrate A. J., GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases. Nat. Biotechnol. 33, 187–197 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Hunt J. M. T., Samson C. A., Rand A. D., Sheppard H. M., Unintended CRISPR-Cas9 editing outcomes: A review of the detection and prevalence of structural variants generated by gene-editing in human cells. Hum. Genet. 142, 705–720 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.He Y., Han Y., Ma Y., Liu S., Fan T., Liang Y., Tang X., Zheng X., Wu Y., Zhang T., Qi Y., Zhang Y., Expanding plant genome editing scope and profiles with CRISPR-FrCas9 systems targeting palindromic TA sites. Plant Biotechnol. J. 22, 2488–2503 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Cui Z., Tian R., Huang Z., Jin Z., Li L., Liu J., Huang Z., Xie H., Liu D., Mo H., Zhou R., Lang B., Meng B., Weng H., Hu Z., FrCas9 is a CRISPR/Cas9 system with high editing efficiency and fidelity. Nat. Commun. 13, 1425 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.J. A. Ruffolo, S. Nayfach, J. Gallagher, A. Bhatnagar, J. Beazer, R. Hussain, J. Russ, J. Yip, E. Hill, M. Pacesa, A. J. Meeske, P. Cameron, A. Madani, Design of highly functional genome editors by modeling the universe of CRISPR-cas sequences. bioRxiv 590591 [Preprint] (2024). 10.1101/2024.04.22.590591. [DOI]
  • 12.Tian R., Cao C., He D., Dong D., Sun L., Liu J., Chen Y., Wang Y., Huang Z., Li L., Jin Z., Huang Z., Xie H., Zhao T., Zhong C., Hong Y., Hu Z., Massively parallel CRISPR off-target detection enables rapid off-target prediction model building. Med 4, 478–492.e6 (2023). [DOI] [PubMed] [Google Scholar]
  • 13.Bennett E. P., Petersen B. L., Johansen I. E., Niu Y., Yang Z., Chamberlain C. A., Met Ö., Wandall H. H., Frödin M., INDEL detection, the ‘achilles heel’ of precise genome editing: A survey of methods for accurate profiling of gene editing induced indels. Nucleic Acids Res. 48, 11958–11981 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kocak D. D., Josephs E. A., Bhandarkar V., Adkar S. S., Kwon J. B., Gersbach C. A., Increasing the specificity of CRISPR systems with engineered RNA secondary structures. Nat. Biotechnol. 37, 657–666 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Riesenberg S., Helmbrecht N., Kanis P., Maricic T., Pääbo S., Improved gRNA secondary structures allow editing of target sites resistant to CRISPR-Cas9 cleavage. Nat. Commun. 13, 489 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Güell M., Sharper scissors: How TREX2 fine-tunes CRISPR-Cas9 nickases for precise editing. Mol. Ther. Nucleic Acids 35, 102119 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Yin J., Lu R., Xin C., Wang Y., Ling X., Li D., Zhang W., Liu M., Xie W., Kong L., Si W., Wei P., Xiao B., Lee H.-Y., Liu T., Hu J., Cas9 exo-endonuclease eliminates chromosomal translocations during genome editing. Nat. Commun. 13, 1204 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Walton R. T., Hsu J. Y., Joung J. K., Kleinstiver B. P., Scalable characterization of the PAM requirements of CRISPR-Cas enzymes using HT-PAMDA. Nat. Protoc. 16, 1511–1547 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Cheong C., Varani G., Tinoco I. Jr., Solution structure of an unusually stable RNA hairpin, 5'GGAC(UUCG)GUCC. Nature 346, 680–682 (1990). [DOI] [PubMed] [Google Scholar]
  • 20.Xue C., Greene E. C., DNA repair pathway choices in CRISPR-Cas9-mediated genome editing. Trends Genet. 37, 639–656 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Shi X., Shou J., Mehryar M. M., Li J., Wang L., Zhang M., Huang H., Sun X., Wu Q., Cas9 has no exonuclease activity resulting in staggered cleavage with overhangs and predictable di- and tri-nucleotide CRISPR insertions without template donor. Cell Discov. 5, 53 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Allen F., Crepaldi L., Alsinet C., Strong A. J., Kleshchevnikov V., De Angeli P., Páleníková P., Khodak A., Kiselev V., Kosicki M., Bassett A. R., Harding H., Galanty Y., Muñoz-Martínez F., Metzakopian E., Jackson S. P., Parts L., Predicting the mutations generated by repair of Cas9-induced double-strand breaks. Nat. Biotechnol. 37, 64–72 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Wang Y., Feng Y.-L., Liu Q., Xiao J.-J., Liu S.-C., Huang Z.-C., Xie A.-Y., TREX2 enables efficient genome disruption mediated by paired CRISPR-Cas9 nickases that generate 3'-overhanging ends. Mol. Ther. Nucleic Acids 34, 102072 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Yin J., Liu M., Liu Y., Wu J., Gan T., Zhang W., Li Y., Zhou Y., Hu J., Optimizing genome editing strategy by primer-extension-mediated sequencing. Cell Discov. 5, 18 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Park S. H., Cao M., Pan Y., Davis T. H., Saxena L., Deshmukh H., Fu Y., Treangen T., Sheehan V. A., Bao G., Comprehensive analysis and accurate quantification of unintended large gene modifications induced by CRISPR-Cas9 gene editing. Sci. Adv. 8, eabo7676 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Aznauryan E., Yermanos A., Kinzina E., Devaux A., Kapetanovic E., Milanova D., Church G. M., Reddy S. T., Discovery and validation of human genomic safe harbor sites for gene and cell therapies. Cell Rep. Methods 2, 100154 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Carnevale J., Shifrut E., Kale N., Nyberg W. A., Blaeschke F., Chen Y. Y., Li Z., Bapat S. P., Diolaiti M. E., O’Leary P., Vedova S., Belk J., Daniel B., Roth T. L., Bachl S., Anido A. A., Prinzing B., Ibañez-Vega J., Lange S., Haydar D., Luetke-Eversloh M., Born-Bony M., Hegde B., Kogan S., Feuchtinger T., Okada H., Satpathy A. T., Shannon K., Gottschalk S., Eyquem J., Krenciute G., Ashworth A., Marson A., RASA2 ablation in T cells boosts antigen sensitivity and long-term function. Nature 609, 174–182 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Dimitri A., Herbst F., Fraietta J. A., Engineering the next-generation of CAR T-cells with CRISPR-Cas9 gene editing. Mol. Cancer 21, 78 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Foy S. P., Jacoby K., Bota D. A., Hunter T., Pan Z., Stawiski E., Ma Y., Lu W., Peng S., Wang C. L., Yuen B., Dalmas O., Heeringa K., Sennino B., Conroy A., Bethune M. T., Mende I., White W., Kukreja M., Gunturu S., Humphrey E., Hussaini A., An D., Litterman A. J., Quach B. B., Ng A. H. C., Lu Y., Smith C., Campbell K. M., Anaya D., Skrdlant L., Huang E. Y., Mendoza V., Mathur J., Dengler L., Purandare B., Moot R., Yi M. C., Funke R., Sibley A., Stallings-Schmitt T., Oh D. Y., Chmielowski B., Abedi M., Yuan Y., Sosman J. A., Lee S. M., Schoenfeld A. J., Baltimore D., Heath J. R., Franzusoff A., Ribas A., Rao A. V., Mandl S. J., Non-viral precision T cell receptor replacement for personalized cell therapy. Nature 615, 687–696 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Gao Q., Dong X., Xu Q., Zhu L., Wang F., Hou Y., Chao C., Therapeutic potential of CRISPR/Cas9 gene editing in engineered T-cell therapy. Cancer Med. 8, 4254–4264 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Mueller K. P., Piscopo N. J., Forsberg M. H., Saraspe L. A., Das A., Russell B., Smerchansky M., Cappabianca D., Shi L., Shankar K., Sarko L., Khajanchi N., Vonne Denne N., Ramamurthy A., Ali A., Lazzarotto C. R., Tsai S. Q., Capitini C. M., Saha K., Production and characterization of virus-free, CRISPR-CAR T cells capable of inducing solid tumor regression. J. Immunother. Cancer 10, e004446 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Weber E. W., Maus M. V., Mackall C. L., The emerging landscape of immune cell therapies. Cell 181, 46–62 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Dekker J., GC- and AT-rich chromatin domains differ in conformation and histone modification status and are differentially modulated by Rpd3p. Genome Biol. 8, R116 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Kosicki M., Tomberg K., Bradley A., Repair of double-strand breaks induced by CRISPR-Cas9 leads to large deletions and complex rearrangements. Nat. Biotechnol. 36, 765–771 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Leibowitz M. L., Papathanasiou S., Doerfler P. A., Blaine L. J., Sun L., Yao Y., Zhang C.-Z., Weiss M. J., Pellman D., Chromothripsis as an on-target consequence of CRISPR-Cas9 genome editing. Nat. Genet. 53, 895–905 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Tsuchida C. A., Brandes N., Bueno R., Trinidad M., Mazumder T., Yu B., Hwang B., Chang C., Liu J., Sun Y., Hopkins C. R., Parker K. R., Qi Y., Hofman L., Satpathy A. T., Stadtmauer E. A., Cate J. H. D., Eyquem J., Fraietta J. A., June C. H., Chang H. Y., Ye C. J., Doudna J. A., Mitigation of chromosome loss in clinical CRISPR-Cas9-engineered T cells. Cell 186, 4567–4582.e20 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Schelker R. C., Gattinoni L., Releasing the brake: CTLA-4 loss turbocharges CAR T cells. Immunity 56, 2180–2182 (2023). [DOI] [PubMed] [Google Scholar]
  • 38.Averina O. A., Kuznetsova S. A., Permyakov O. A., Sergiev P. V., Current knowledge of base editing and prime editing. Mol. Biol. 58, 571–587 (2024). [PubMed] [Google Scholar]
  • 39.Rupp L. J., Schumann K., Roybal K. T., Gate R. E., Ye C. J., Lim W. A., Marson A., CRISPR/Cas9-mediated PD-1 disruption enhances anti-tumor efficacy of human chimeric antigen receptor T cells. Sci. Rep. 7, 737 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Cameron P., Fuller C. K., Donohoue P. D., Jones B. N., Thompson M. S., Carter M. M., Gradia S., Vidal B., Garner E., Slorach E. M., Lau E., Banh L. M., Lied A. M., Edwards L. S., Settle A. H., Capurso D., Llaca V., Deschamps S., Cigan M., Young J. K., May A. P., Mapping the genomic landscape of CRISPR-Cas9 cleavage. Nat. Methods 14, 600–606 (2017). [DOI] [PubMed] [Google Scholar]
  • 41.Hlozánek I., The influence of ultraviolet-inactivated Sendai virus on Marek’s disease virus infection in tissue culture. J. Gen. Virol. 9, 45–50 (1970). [DOI] [PubMed] [Google Scholar]
  • 42.Lazzarotto C. R., Malinin N. L., Li Y., Zhang R., Yang Y., Lee G. H., Cowley E., He Y., Lan X., Jividen K., Katta V., Kolmakova N. G., Petersen C. T., Qi Q., Strelcov E., Maragh S., Krenciute G., Ma J., Cheng Y., Tsai S. Q., CHANGE-seq reveals genetic and epigenetic effects on CRISPR-Cas9 genome-wide activity. Nat. Biotechnol. 38, 1317–1327 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Foox J., Tighe S. W., Nicolet C. M., Zook J. M., Byrska-Bishop M., Clarke W. E., Khayat M. M., Mahmoud M., Laaguiby P. K., Herbert Z. T., Warner D., Grills G. S., Jen J., Levy S., Xiang J., Alonso A., Zhao X., Zhang W., Teng F., Zhao Y., Lu H., Schroth G. P., Narzisi G., Farmerie W., Sedlazeck F. J., Baldwin D. A., Mason C. E., Performance assessment of DNA sequencing platforms in the ABRF next-generation sequencing study. Nat. Biotechnol. 39, 1129–1140 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Malinin N. L., Lee G., Lazzarotto C. R., Li Y., Zheng Z., Nguyen N. T., Liebers M., Topkar V. V., Iafrate A. J., Le L. P., Aryee M. J., Joung J. K., Tsai S. Q., Defining genome-wide CRISPR-Cas genome-editing nuclease activity with GUIDE-seq. Nat. Protoc. 16, 5592–5615 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Liu Y., Yin J., Gan T., Liu M., Xin C., Zhang W., Hu J., PEM-seq comprehensively quantifies DNA repair outcomes during gene-editing and DSB repair. STAR Protoc. 3, 101088 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Clement K., Rees H., Canver M. C., Gehrke J. M., Farouni R., Hsu J. Y., Cole M. A., Liu D. R., Joung J. K., Bauer D. E., Pinello L., CRISPResso2 provides accurate and rapid genome editing sequence analysis. Nat. Biotechnol. 37, 224–226 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figs. S1 to S5

Legends for tables S1 to S7

sciadv.adu7334_sm.pdf (2.3MB, pdf)

Tables S1 to S7

sciadv.adu7334_tables_s1_to_s7.zip (1.9MB, zip)

Articles from Science Advances are provided here courtesy of American Association for the Advancement of Science

RESOURCES