Indiscriminate ssDNA cleavage activity of CRISPR-Cas12a induces no detectable off-target effects in mouse embryos

Newly discovered characteristics like “collateral effect” or trans-cleavage in CRISPR-Cas13 and CRISPR-Cas12 systems have enabled their usage in nucleic acid detection (Gootenberg et al. 2017, 2018; Chen et al. 2018). The collateral RNA cleavage of Cas13a has been reported to be harmful for cell development (Wang et al. 2019; Buchman et al. 2020). As a representative gene editor of CRISPRCas12 system, CRISPR-Cas12a (Cpf1) holds great potential for therapeutic applications in the future (Zetsche et al. 2015; Koo et al. 2018; Campa et al. 2019). However, when used for genome editing in mammalian cells, target-activated Cas12a has the risk to cleave transiently exposed ssDNA during replication, transcription and homology-directed repair processes (Chen et al. 2018) (Fig. 1A), raising the concern of its therapeutic applications. Therefore, the potential off-target effects caused by the indiscriminate ssDNA cleavage activity of Cas12a need to be carefully investigated. Recently, we developed a new approach called GOTI (Genome-wide Off-target analysis by Two-cell Injection) to detect off-target effects without the interference of singlenucleotide polymorphisms (SNPs) in individuals (Zuo et al. 2019). In this study, we designed an optimized method called genome-wide off-target analysis by twin blastomeres (GOAT) for off-target edits detection. Briefly, mouse embryos were separated into two embryos at two-cell stage, and then gene editing tools such as BE3, ABEmax and Cas12a were injected into one of the twin embryos (Figs. 1B and S1A). To increase the pregnancy efficiency, twin embryos were co-transferred with two ICR embryos to the pseudopregnant mouse. When the twin embryos developed to embryonic day 12.5 (E12.5), twin embryos and ICR embryos were distinguished by their eye colors and a SNP site on Tyr gene (Fig. S1B and S1C). The edited embryo was distinguished from the unedited twin embryo by high editing efficiency to induce indels and nucleotide substitutions on the target sites. Whole-genome sequencing (WGS) was performed on the genomic DNA of twin embryos, separately. Then single-nucleotide variants (SNVs) and indels were called in the injected sample, with its twin uninjected one as the reference (Figs. 1B and S1A). GOAT could distinguish the injected and un-injected embryos directly, while GOTI relies on massive FACS to separate edited cells from unedited cells. In addition, GOAT could also avoid the leak of two-cell injection, false-negative FACS sorting and inferior developmental competition ability of the injected blastomere. To test the effectiveness of GOAT system, we included three groups in our study: GFP, BE3, and ABEmax groups (Fig. S1A). The developmental rate of twin embryos to blastocysts was more than 90%, and twin embryos developed to E12.5 was 23.0% ± 3.1% (n = 5; Table S1). WGS was conducted separately for the twin embryos at an average depth of 30 to confirm on-target editing efficiency and analyze the potential genome-wide off-target effects (Table S2). The activities of BE3 and ABEmax were confirmed by the high on-target efficiencies to introduce nucleotide substitutions (Figs. S1D, S2 and S3). For the off-target effects, we found 14 SNVs and 0 indel per embryo on average in the GFP-injected group (Figs. 1C, 1D, S4 and Tables S3, S4, S5). For the BE3-injected embryos, we found 210 SNVs per embryo on average, 15 times more than those of the GFP group (P = 0.0025; Figs. 1C, S4 and Tables S3, S6). By contrast, indels showed no differences between BE3 and GFP groups (Fig. 1D). We observed that about 86% of SNVs were mutated from C to T, or G to A (Figs. 1E and S5), consistent with the results of GOTI method (Zuo et al. 2019). We also analyzed the offtarget effects of ABEmax using GOAT. An average of 18 SNVs and 0 indel were detected in each embryo, similar to the number found in the GFP-injected group (P = 0.57; Figs. 1C, 1D, S4 and Tables S3, S4). Together, these results suggest that GOAT is a comparable approach to detect genome-wide off-target effects in mouse embryo comparing with GOTI. We further used GOAT to analyze the genome-wide offtarget effects of two commonly used Cas12a (LbCas12a and AsCas12a). Similarly, LbCas12a or AsCas12a mRNA and their crRNAs targeting Dmd or Tp53 gene were injected into one of the twin embryos (Fig. 1B). The activities of LbCas12a and AsCas12a were confirmed by the high efficiency


In vitro transcription (IVT) of gene editing tools
To generate IVT template of BE3, BE3 was amplified from plasmid with BE3-F and BE3-R primer pairs and T7 promoter at 5' end. After amplification, BE3 PCR products were gel-purified and used as IVT template. BE3 mRNA were in vitro transcribed using mMACHINE T7 ULTRA kit (Life Technologies) following the manufacture's instruction. BE4max mRNA, ABEmax mRNA, LbCas12a mRNA, AsCas12a mRNA and GFP mRNA were generated in the same way. Guide RNA used in the study were in vitro transcribed from PCR-generated template using MEGA shortscript T7 kit (Life Technologies). After in vitro transcription, both mRNA and sgRNA were purified with MEGA clear kit (Life Technologies) and eluted in RNase-free water. Guide RNA sequences and primers used in the study for generating IVT templates were listed as follows. Guide RNA sequence Guide RNA name Sequence (5'-3')

Embryo collection, splitting and transplantation
Eight-week-old B6D2 F1 (C57BL/6 crossing with DBA/2 mice) female mice were superovulated by intraperitoneal injection of 0.75 IU PMSG on the first day and 0.75 IU hCG forty-eight hours after PMSG injection. Immediately after hCG administration, superovulated female mice were mated with 8-15 weeks old B6D2 F1 male mice. Zyogtes were collected from the mated mice oviducts 20 hours post hCG injection. After zygotes reached late 2-cell stage (46-49 hours post hCG injection), twin blastomeres in each embryo were split into two individual blastomeres in a droplet of M2 (Millipore) medium containing 5 μg/ml cytochalasin B (CB). Briefly, zona pellicuda of the 2-cell stage embryo was partially ablated by lazer to remove one of the blastomere out of the 2-cell stage embryo while leaving another one in the same zona pellicuda with 45 μm diameter-wide needle. Removed blastomere was then transferred to vacant recipient zona pellicuda. In this way, two artificial twin embryos were generated for separate injection of mRNA and sgRNA afterwards. For injection, BE3 mRNA (50ng/μl), BE4max mRNA(50ng/μl), ABEmax mRNA(50ng/μl), LbCas12a mRNA (50ng/μl) or AsCas12a mRNA (50ng/μl) and sgRNA (50ng/μl) were co-injected into one of the artificial twin embryos by FemtoJet microinjector (Eppendorf) with constant flow settings. All the embryos were cultured in KSOM+AA (Millipore) at 37℃ and 5% CO2 until blastocyst stage for transplantation. The twin blastocysts were inspected for high quality before being picked for transferring to the oviducts of 0.5dpc pseudopregnant surrogate mice. In addition to the twin embryos, two E1.5 ICR embryos were co-transferred to improve pregnant efficiency.

Twin embryos collection and genotyping
Cesarean surgery was performed for surrogate mice at pregnancy day 12.5. After surgery, embryos were dissected out of yolk sac with tweezers removing extra-embryonic tissues and washed with PBS for 5-10 times in the dish. Each E12.5 embryo was divided into 2 parts, one part was used for Sanger sequencing to evaluate on-target editing efficiency and the other part was used for whole genome sequencing. Embryos were lysed by lysis buffer with proteinase K at 37℃ for 1h, and then incubated at 95℃ for 30min to inactivate the proteinase K. Genotyping PCR was performed for 35cycles at 95℃ for 30s, 60℃ for 30s and 72℃ for 1min, respectively. PCR products were analyzed by Sanger sequencing to differentiate the edited from unedited twin embryos. Genotyping primers were listed in the following table.
Primers used for genotyping Sequence(5'-3') Fluorophore quencher (FQ)-labeled reporter assays 30 nM LbCas12a (NEB, M0653S) was pre-assembled with 50 nM of Dmd-targeting crRNA in 37°C for 10 min. Different concentration of dsDNA were added to the pre-assembled Cas12a and crRNA to incubate for 60 min in a 20 μL reaction system.

Whole genome sequencing (WGS)
Twin embryos were collected from uterus of sacrificed surrogate female male for genomic DNA extraction. Genomic DNA was extracted using DNeasy blood and tissue kit (catalog number 69504, Qiagen) and twin embryo samples were sequenced at an average depth of 30x using 150bp paired-end Illumina X-Ten platform. Fastp (V0.20.0) were used to filter the low-quality reads with parameters '-q 20 -u 40 -M 0 -n 5 -1 80 -w 32'. Bwa-mem (0.7.16a) was used to align the clean reads to mm10 reference genome. (Li 2013) Samtools (1.6) was used to sort the mapped BAM files and GATK (4.0.12.0) was used to mark the duplicated reads. (Li et al. 2009;McKenna et al. 2010) Four major algorithms Strelka (2.9.x), Lofreq (v2.13), Mutect2 (v4.0.12.0) and Scalpel (v0.5.4) (Saunders et al. 2012;Wilm et al. 2012;Cibulskis et al. 2013;Fang et al. 2016) were used to identify the SNVs and indels. To reduce the computational burden, an optimized pipeline was used to identify the variants. Firstly, Strelka was used to indentify the genome-wide SNVs and Indels. The regions 200bp upstream and downstream of the variants identified by Strelka were treated as candidate regions. Secondly, Lofreq and Mutect2 were used to calculate SNVs of the candidate regions identified by Strelka, respectively. Scalpel and Mutect2 were used to identify the indels of the candidate regions, respectively. We also applied the CasOFFinder for the identification of potential sgRNA-dependent off-target SNVs and indels. The adjacent 400bp region of these variants were also used as candidate regions for indel detection by Scalpel. Only variants identified by all the three algorithms and with more than 10% allele frequencies were used for the following analysis. To strictly control the quality of the variants, we removed variants overlapped with UCSC repeat regions or reported in dbSNP (v140) database. Bedtools (v2.29.2) were used to perform the overlapping of variants (Quinlan and Hall 2010). Bam-readcount (V0.8.0) were used to calculate base frequency of the SNVs.   The rare untargeted mutations in un-injected group were likely caused by sequence errors. Figure S4. Venn diagrams of SNVs in each embryo analyzed by Lofred, Mutect2 and Strelka. SNVs were called by Lofred, Mutect2 and Strelka, separately. Common SNVs called by all the three algorithms were defined as true SNVs. The overlapping SNVs with allele frequencies less than 10% were removed from the following analysis.    Each 2-cell stage embryo was separated into two blastomeres to get twin blastomeres. *When the twin embryos developed to blastocyst stage, the well developing twin blastocysts were transferred to pseudopregnant mother.